No title

WATER RESEARCH A Journal of the International Water Association

Editor-in-Chief Mark van Loosdrecht Delft University of Technology and KWR Watercycle Research The Netherlands E-mail: [email protected]

Editors Jean-Claude Block Université H. Poincaré, Nancy I France David Dixon University of Melbourne Australia Hiroaki Furumai University of Tokyo Japan Xiaodi Hao Beijing University of Civil Engineering and Architecture China Gregory Korshin University of Washington USA Anna Ledin Formas Sweden Eberhard Morgenroth ETH Zürich and Eawag, Swiss Federal Institute of Aquatic Science and Technology Switzerland Wolfgang Rauch University of Innsbruck Austria Maria Reis Universidade Nova de Lisboa/FCT Portugal Hang-Shik Shin Korea Advanced Institute of Science and Technology Korea Thomas Ternes Bundesanstalt für Gewässerkunde Germany Stefan Wuertz University of California, Davis USA

Associate Editors Andrew Baker University of New South Wales Australia Damien Batstone University of Queensland Australia

Guang-Hao Chen Hong Kong University of Science & Technology China Tom Curtis University of Newcastle upon Tyne UK Francis de los Reyes III North Carolina State University USA Ana Deletic Monash University USA Rob Eldridge Glen Waverley Australia Monica B. Emelko University of Waterloo Canada Rosina Girones University of Barcelona Spain Stephen Gray Victoria University Australia Kate Grudpan Chiang Mai University Thailand Edwin Herricks University of Illinois at Urbana-Champaign USA Peter Hillis AECOM Australia Hong-Ying Hu Tsinghua University China Bruce Jefferson Cranfield University UK Ulf Jeppsson Lund University Sweden Sergey Kalyuzhnyi RUSNANO Russian Federation Jae-Hong Kim Georgia Institute of Technology USA Günter Langergraber University of Natural Resources and Applied Life Sciences, Vienna (BOKU) Austria

Yoshihiko Matsui Hokkaido University Japan Armand Maul Université de Metz France Max Maurer Eawag Switzerland How Yong Ng National University of Singapore Singapore Satoshi Okabe Hokkaido University Japan Jong Moon Park Pohang University of Science & Technology Korea

Shang-Lien Lo National Taiwan University Taiwan Dionisis Mantzavinos Technical University of Crete Greece

Anastasios Zouboulis Aristotle University of Thessaloniki Greece

Susan Richardson US Environmental Protection Agency USA Miguel Salgot University of Barcelona Spain David Sedlak University of California, Berkeley USA Jean-Philippe Steyer LBE-INRA France Masahiro Takahashi Hokkaido University Japan Nathalie Tufenkji McGill University Canada Kai Udert Eawag Switzerland Vayalam Venugopalan BARC Facilities India Eduardo von Sperling Federal University of Minas Gerais Brazil Julie Zilles University of Illinois at Urbana-Champaign USA

Editorial Office E-mail: [email protected]

Publication information: Water Research (ISSN 0043-1354). For 2011, volume 45 is scheduled for publication. Subscription prices are available upon ­request from the publisher or from the Elsevier Customer Service Department nearest you or from this journal’s website (http://www.elsevier.com/locate/watres). Further information is available on this journal and other Elsevier products through Elsevier’s website (http://www.elsevier.com). Subscriptions are accepted on a prepaid basis only and are entered on a calendar year basis. Issues are sent by standard mail (surface within Europe, air delivery outside Europe). Priority rates are available upon request. Claims for missing issues should be made within six months of the date of dispatch. Author Enquiries: For enquiries relating to the submission of articles (including electronic submission) please visit this journal’s homepage at http://www.elsevier.com/ locate/watres. Contact details for questions arising after acceptance of an article, especially those relating to proofs, will be provided by the publisher. You can track accepted articles at http://www.elsevier.com/trackarticle. You can also check our Author FAQs at http://www.elsevier.com/authorFAQ and/or contact Customer Support via http://support.elsevier.com. Orders, claims, and journal enquiries: Please contact the Elsevier Customer Service Department nearest you: St. Louis: Elsevier Customer Service Department, 3251 Riverport Lane, Maryland Heights, MO 63043, USA; phone: (877) 8397126 [toll free within the USA]; (+1) (314) 4478878 [outside the USA]; fax: (+1) (314) 4478077; e-mail: [email protected] Oxford: Elsevier Customer Service Department, The Boulevard, Langford Lane, Kidlington OX5 1GB, UK; phone: (+44) (1865) 843434; fax: (+44) (1865) 843970; e-mail: [email protected] Tokyo: Elsevier Customer Service Department, 4F Higashi-Azabu, 1-Chome Bldg, 1-9-15 Higashi-Azabu, Minato-ku, Tokyo 106-0044, Japan; phone: (+81) (3) 5561 5037; fax: (+81) (3) 5561 5047; e-mail: [email protected] Singapore: Elsevier Customer Service Department, 3 Killiney Road, #08-01 Winsland House I, Singapore 239519; phone: (+65) 63490222; fax: (+65) 67331510; e-mail: [email protected] Application for membership of International Water Association should be made to: Executive Director, IWA, Alliance House, 12 Caxton Street, London SW1H 0QS, U.K. (Tel.: +44 207 654 5500; Fax: +44 207 654 5555; e-mail: [email protected]; website: http://www.IWAhq.org.uk). Registered Charity (England) No. 289269. Individual membership is available from £30 upwards. For details contact IWA.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 6 8 3 e4 6 9 9

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Integrated application of upflow anaerobic sludge blanket reactor for the treatment of wastewaters Muhammad Asif Latif, Rumana Ghufran, Zularisam Abdul Wahid, Anwar Ahmad* Faculty of Civil Engineering & Earth Resources, University Malaysia Pahang (UMP), Lebuhraya Tun Razak, 26300 Gambang, Kuantan, Pahang, Malaysia

article info

abstract

Article history:

The UASB process among other treatment methods has been recognized as a core method

Received 27 December 2010

of an advanced technology for environmental protection. This paper highlights the treat-

Received in revised form

ment of seven types of wastewaters i.e. palm oil mill effluent (POME), distillery wastewater,

24 May 2011

slaughterhouse wastewater, piggery wastewater, dairy wastewater, fishery wastewater

Accepted 31 May 2011

and municipal wastewater (black and gray) by UASB process. The purpose of this study is to

Available online 12 June 2011

explore the pollution load of these wastewaters and their treatment potential use in upflow anaerobic sludge blanket process. The general characterization of wastewater, treatment

Keywords:

in UASB reactor with operational parameters and reactor performance in terms of COD

UASB reactor

removal and biogas production are thoroughly discussed in the paper. The concrete data

Industrial wastewater

illustrates the reactor configuration, thus giving maximum awareness about upflow

Agro wastewater

anaerobic sludge blanket reactor for further research. The future aspects for research

Municipal wastewater

needs are also outlined. ª 2011 Elsevier Ltd. All rights reserved.

COD Biogas

Contents 1. 2. 3.

4. 5.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1. Characterization and environmental impacts of wastewaters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Treatment potential of UASB process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Operation and performance of UASB reactor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1. Organic loading rates and COD removal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2. Flow rate and hydraulic retention time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3. Upflow velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4. pH . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5. Operating temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.6. Mixing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Research needs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

* Corresponding author. Tel.: þ60 9 5493012; fax: þ60 9 5492998. E-mail address: [email protected] (A. Ahmad). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.049

4684 4684 4685 4688 4688 4692 4693 4693 4693 4694 4694 4695 4695 4695

4684

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 6 8 3 e4 6 9 9

Nomenclature AF AS BOD COD CODdiss CRT CUASB FOG HRT IUASB L OLR Q SBR SCOD

1.

anaerobic film activated sludge biochemical oxygen demand (g L1) chemical oxygen demand (g L1) dissolved chemical oxygen demand (g L1) cell residence time (days) upflow anaerobic sludge blanket at Canakkale fat, oil and grease (g L1) hydraulic retention time (days) upflow anaerobic sludge blanket at Istanbul length (cm) organic loading rate (kgCOD m3 d1) flow rate (L d1) sequencing batch reactor soluble chemical oxygen demand (g L1)

Introduction

At the beginning of 21st century, the world is facing environmental crisis in terms of water quality and global warming, caused by continuous population growth, industrialization, food production practices, increased living standards and poor water use strategies. The rapid industrialization, urbanization, and population growth resulted in increasing volumes of untreated domestic and industrial wastewater being discharged into the rivers and canals and consequently deteriorating surface and groundwater quality. The polluted water quality would prolong to affect the groundwater, threatening to drinking water safety and thus the health of urban and rural residents along with adverse effect on ecosystem, particularly aquatic life and biospheres. The lack of wastewater management has direct impact on biological diversity of the aquatic ecosystems, disrupting the fundamental integrity of our life support systems, on which a wide range of sectors from urban development to food production and industry depend. It is essential to consider the wastewater management as a part of integrated, ecosystem based management to operate across sectors and borders, freshwater and marine. The wastewater is a mixture of sewage, agricultural drainage, industrial waste effluents and hospitals discharge. Untreated wastewater may contain different range of pathogens including bacteria, parasites, and viruses, toxic chemicals such as heavy metals and organic chemicals from agriculture, industrial and domestic sources (Andrew et al., 1997; Drechsel and Evans, 2010). In order to minimize the environmental contaminants and health hazards, the treatment of these pollutants needs to be brought down to permissible limits for safe disposal of wastewater (Poots et al., 1978; Manju et al., 1998).

1.1. Characterization and environmental impacts of wastewaters The production of palm oil from the fruit Elaeis guineensis is the main industry in Southeast Asia (Ma, 2000). The effluent during

SS T-P T-N TKN TS TSS TUASB TVS UASB V VS Vs VSS Vup W WW

suspended solids (g L1) total phosphorous (g L1) total nitrogen (g L1) total Kjeldahl nitrogen (g L1) total solids (g L1) total suspended solids (g L1) upflow anaerobic sludge blanket at Tekirdag total volatile solids (g L1) upflow anaerobic sludge blanket reactor volume (L) volatile solids (g L1) superficial velocity (m h1) volatile suspended solids (g L1) upflow velocity (m h1) width (cm) wastewater

and after processing contains large amount of free and dissolved oil and fatty acids, crude oil solids, starches, proteins and plant tissues (Cheah et al., 1998). Similarly, distillery effluents are highly polluted and fall under medium to high-strength wastewaters as many kinds of raw materials are used for different types of alcohols (Ince et al., 2005). Effluent from a wine distillery consists first and foremost of organic acids with a lofty soluble biodegradable chemical oxygen demand (COD) fraction of 98% (Moosbrugger et al., 1993). The seasonal nature of distillery industries creates specific problems for the treatment processes in terms of wine distillery wastewater (Coetzee et al., 2004; Euse´bio et al., 2004). Moreover, slaughterhouse wastewater holds high amount of suspended and colloidal components in the form of fats, proteins and cellulose, which have adverse effects on the environment (Lettinga et al., 1997; Nu´n˜ez and Martı´nez, 1999). These organic matters can be treated by means of anaerobic digestion as they have high concentrations of biodegradable organic contents, sufficient alkalinity, and suitable concentrations of phosphorus, nitrogen, and micronutrients for the bacterial growth (Masse´ and Masse, 2001). Single-phase, upflow anaerobic sludge blanket (UASB) type anaerobic digesters are considered to be impractical, because the fat present may form thick foam inside the reactor, compromising the operation (Chen and Shyu, 1998; del Pozo et al., 2000; Torkian et al., 2003; Barreto, 2004). Moreover, accumulations of suspended solids lead to a reduction in methanogenic activity and biomass washout (Sayed and de Zeeuw, 1988; Hansen and West, 1992). Discharge of slaughterhouse wastewater without treatment is degrading the aquatic environment and polluting the irrigation water (Michael et al., 1988). Piggery wastes are also distinguished as rich in organic matters and pathogenic organisms. The disposal of piggery wastes without ample treatment can have a drastic effect on the environment and human health (Sa´nchez et al., 2005). This waste is a mixture of manure (feces and urine) and food waste from instance swill and sugar cane molasses (Sa´nchez et al., 2001). However, UASB reactor has rarely been used for the treatment of piggery waste because of rich nitrogenous

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 6 8 3 e4 6 9 9

compounds, the active biomass densification, granular process, and consequently the microorganism retention in the reactor difficult. The UASB process seems to be very efficient for the treatment of carbohydrate containing wastewaters. Limited work has been done on the application of this reactor for the treatment of piggery wastes (Sa´nchez et al., 2005). Dairy wastewater is generated either from milk or cheeseproducing industries. Dairy wastewater, from the milk industry has COD of 30 g L1 whereas, in case of cheeseproducing industries the generated wastewater generally contains surplus cheese whey with a COD value of 50 g L1 (Nemerow, 1987). Dairy wastewater is generally characterized by its relatively high temperature and variation in organic contents. Due to very small and medium size dairy industries which normally do not have an economic incentive to further use of cheese whey, it is necessary to choose the whey as waste stream (Mockaitis et al., 2005). Due to high organic contents, whey causes several environmental problems for surface water and soil while disposing (Patel and Madamwar, 1998; Kosseva et al., 2003). Dairy wastewaters may also cause serious problems in terms of organic load on the local municipal wastewater treatment plant (Papachristou and Lafazanis, 1997). Dairy wastewater has been widely treated using coagulation/flocculation and sedimentation process. The main shortcomings of these methods are high coagulant cost, large amount of sludge production, and the poor removal of dissolved COD. Therefore, biological treatment is usually recommended for such wastewaters (Vidal et al., 2000). Several studies have been done with the aim of adapting anaerobic high-rate digesters, especially UASB reactors for the treatment of dairy wastewater. Conventionally, UASB reactors are inoculated with granular sludge that has high methanogenic activity. It has been proved that it was not possible to maintain granular biomass with dairy wastewater in long term operation (Marques et al., 1990; Yang and Anderson, 1993). Fullscale UASB reactor has been successfully engaged for the treatment of dairy wastewater (Malina and Pohland, 1992). The use of a laboratory-scale hybrid UASB reactor for treatment of dairy wastewater at an operational temperature of 30  C was previously investigated by Ozturk et al. (1993). The treatment of multiple fat containing wastewaters should be done in UASB reactors inoculated with flocculent biomass that has a high hydrolytic and acidogenic capacity (Sayed, 1987). In fishery wastewaters the contaminants are undefined mixtures of mostly organic substances. It is difficult to characterize the extent of the problem created by this wastewater as it depends on the effluent potency, wastewater discharge rate and engrossing capacity of the receiving water bodies (Gonzalez, 1996). Moreover, wastewaters generated from fishmeal industry contain high organic suspended solids, proteins and salinity close to seawater (Vidal et al., 1997). Recently, anaerobic granular sludge treatment of fishmeal wastewaters is supposed to be an optimal process. However, few studies have been performed so far on the application of anaerobic granular sludge to the removal of organic pollutants in highly saline wastewaters (Jeison et al., 2008; Lei et al., 2008). During aerobic treatment of hyper saline wastewaters that is under high saline conditions, cell plasmolysis, deficiency of filamentous micro-organisms and lack of protozoa will occur simultaneously (Lei et al., 2008). Although, findings

4685

still have many disagreements. The sludge degranulation was really observed while treating saline tannery soaked liquor in UASB reactor (Huang et al., 2009). Black water contains half the load of organic material in domestic wastewater with a major fraction of the nutrients as nitrogen and phosphorus (Otterpohl et al., 1999; KujawaRoeleveld and Zeeman, 2006). The risk of dispersion of diseases, due to exposure to micro-organisms in the water, will be a critical phase if the water is to be reused for toilet flushing or irrigation. In contrast, gray water has a high potential of reuse because it contains the major fraction (approximately 70%) of domestic wastewater and remains relatively less polluted (Leal et al., 2007). There is a risk that micro-organisms in the water will be increased in the form of aerosols that are generated when the toilets are flushed (Feachem et al., 1983; Christova-Boal et al., 1996; Albrechtsen, 1998). Both inhaling and hand to mouth contact can be dangerous (Ince et al., 2005). The resulting effluent stream from the process is polluted with a COD of 20e30 mg L1 and pH of 3e4 (Wolmarans and de Villiers, 2002). In warm climate countries, the high-rate anaerobic process (like UASB) shows satisfactory treatment performance, even for diluted domestic wastewater, with many advantages, including reduction of green house gas emissions, reduced excess sludge productions, stabilized sludge and low space requirements as compared to conventional digestion systems (van Lier and Huibers, 2004). General characterization of different wastewaters used in UASB reactor is shown in Table 1. During the treatment of complex wastewaters containing significant amounts of fat (e.g. slaughterhouse, dairy), the continuous operation of UASB reactors has shown to cause problems of scum and sludge layer on top of the reactors with succeeding biomass washout (Hwu, 1997; Petruy, 1999; Nadais et al., 2005a,b). The high COD accumulation in sludge bed has also been reported to lead unstable performance of reactors on long run. Research on anaerobic degradation of complex fat containing wastewater showed that the initial removal mechanism was mainly adsorption (Riffat and Dague, 1995; Hwu, 1997; Nadais et al., 2003). The sharp adsorption of organic matter occurs in the sludge bed but is not followed by an immediate biological degradation of the adsorbed organic matter since the kinetics of biological degradation is much slower than the adsorption phenomena (Nadais, 2002). As a consequence heavy accumulation of organic matter in continuous treatment systems has been observed both at lab and full-scale UASB reactors treating complex fat containing substrates. According to Jeganathan et al. (2006) the factor that most influences high-rate reactor performance is the FOG (fats, oils and greases) accumulation rather than the FOG concentration in the reactor feed. The substrates that pose more problems in the anaerobic degradation of dairy wastewater are the fatty matters and the long chain fatty acids resulting from milk fat hydrolysis, especially oleic acid (Petruy, 1999).

2.

Treatment potential of UASB process

Anaerobic treatment of wastewaters is nowadays widely accepted as a proved technology and extensively used. One of the main factors leading to the success of anaerobic treatment

4686

Table 1 e General properties of wastewaters used for UASB process. Type of wastewater

COD

BOD

TS

TSS

VSS

T-P

T-N

pH

Oil & grease

Reference

21 BOD5

70

45

e

e

5

11

Siang (2006)

Distillery WW Wine distillery WW Vinasse Raw spent wash Molasses WW Alcohol Distillery

100e120 3.1e48 97.5 as SCOD 37.5 80.5 11e33

30 0.21e8.0 42.23 e e 6e16

51.5e100 11.4e32 3.9 2.82 109 e

e 2.4e5.0 e e e e

2.8 1.2e2.8 e e 2.5 e

e 0.24e65.7 e 0.24 e 0.3e0.7

e 0.1e64 e 2.02 1.8 0.12e0.25

3e4.1 3.53e5.4 4.4 4.2 5.2 4e7

e e e e e 3

Nataraj et al. (2006) Bustamante et al. (2005) Martin et al. (2002) Ramana et al. (2002) Jimenez and Borja (1997) Ince et al. (2005)

Dairy WW

3.38 5.4e77.3

1.94 e

1.56 3.9e58.9

0.83

0.022

0.051 as TKN 0.5e5.6

7.9 4.3e8.7

0.26 0.4e5.7

Tawfik et al. (2008) Kalyuzhnyi et al. (1996)

74.5

e

e

9.38 as SS

0.75 3.1e48.7 as VS 8.3

0.124

0.15 as TKN

3.92

e

Ergu¨der et al. (2001)

0.12 1.73 0.82 0.49e0.65

e 5.99 5 e

e e e e

e e e 0.012e0.02

7.3 6.2

5.5 2.3

2.7 e

0.82 5.3

0.01 0.04

6.6 6.6

e 1.0 as FOG 0.66 as FOG 0.05e0.1 as grease 0.31 0.6

Slaughterhouse WW

4.4

e

3.9

0.40 as SS

0.60 as VS

e

e 0.21 as TKN 0.13 as TKN 0.12e0.18 as TKN 0.08 0.21 as Organic N e

6.85 6.4 6.9 6.8e7.1

Poultry slaughter WW Slaughterhouse WW

25 e e 40e50 as SS 0.94 6.3

Huang et al. (2009) Prasertsan et al. (1994)

Slaughterhouse WW

0.17 3.3 1.5 1.5e2.2

6.8

e

Seif and Moursy (2001)

Piggery manure and WW

4.8e12.6

56e58

1.9e43.2

1.72e33.1

0.2e1.52

5e5.9

e

Sa´nchez et al. (1995)

6 7.8

e e

Sa´nchez et al. (2005) Angelidaki et al. (2002)

6.35 e e

0.26 as FOG e e

Halalsheh et al. (2008) Tandukar et al. (2005) Moawad et al. (2009)

Fishery WW Fish canning

Muncipal WW

30e70 95 50

11e30 22 25 BOD3

30e65 35 40.5

9e25 as SS 12 as SS 18 as SS

e e 0.02

0.5e0.9 e 0.75

3.5e4.5 4.35 4.7

5e13 as oil 10.6 e

Borja et al. (1996) Chaisri et al. (2007) Ma (1999)

10.19 65

e

7.21 59

1.64 53

1.17 38

0.42 3

0.8 as Organic N 0.34 6.8

2.6 0.53 0.45

1.1 0.24 0.22

e e e

0.85 0.26 as SS 0.19

e e e

0.02 e 0.0034

0.13 as TKN 0.046 as TKN 0.05 as TKN

All values are in g L1 except pH.

Sayed et al. (1984) Cha´vez et al. (2005) Caixeta et al. (2002)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 6 8 3 e4 6 9 9

80

e e 34 as TVS 37

Palm oil mill effluent

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 6 8 3 e4 6 9 9

was the introduction of high-rate reactors in which biomass retention and liquid retention are uncoupled (Lettinga et al., 1980; de Zeeuw, 1988). High-rate anaerobic reactors, that can retain biomass, have a high treatment capacity and hence low site area requirement (Droste, 1997). Several processes have thus been developed to operate anaerobic digestion reactors, each of them having several advantages. Musee et al. (2006) identified waste-generating mechanisms, analyzed the causes, and then derived options for feasible waste minimization alternatives. One of the most common is the UASB process that has successfully been used to treat a variety of wastewaters, but is often limited by poor biodegradability of complex organic substrates (Goodwin and Stuart, 1994; Seghezzo et al., 1998; Goodwin et al., 2001; Wolmarans and de Villiers, 2002; Coetzee et al., 2004). The UASB reactor exhibits positive features, such as high organic loading rates (OLRs), short hydraulic retention time (HRT) and a low energy demand (Borja and Banks, 1994; Metcalf and Eddy, 2003). Anaerobic sludge in UASB reactors spontaneously immobilizes into well settling granular sludge. It has been widely adopted for treatment of medium to highstrength industrial wastewaters (Lettinga and Hulshoff Pol, 1991; Fang et al., 1995). UASB reactor was developed by Lettinga et al. (1980) whereby this system has been successful in treating a wide range of industrial effluents including those with inhibitory compounds. The underlying principle of the UASB operation is to have an anaerobic sludge which exhibits good settling properties (Lettinga, 1995) and efficiently retains complex microbial consortium without the need for immobilization on a carrier material (for example, as a biofilm) by formation of biological granules with good settling characteristics. Performance depends on the mean cell residence time and reactor volume depends on the hydraulic residence time, therefore, UASB reactor can efficiently convert organic compounds of wastewater into methane in small ‘high-rate’ reactors. Approximately 60% of the thousands of anaerobic full-scale treatment facilities worldwide are now based on the UASB design concept, treating a various range of industrial wastewaters (Jantsch et al., 2002; Karim and Gupta, 2003). Moreover, previous research studies also indicate the feasibility of this process to treat domestic effluents (Behling et al., 1997; Singh and Viraraghavan, 2000). The key feature of this system is the microbial aggregation into a symbiotic multilayer structure called a granule and retention of highly active biomass with good settling abilities in the reactor (Schmidt and Ahring, 1996). Improved process knowledge and operational details on formation of stable granules have made the possibility of high organic loadings and resulting in a more sustainable operation. The long hydraulic retention times are known to be unfavorable for sludge granulation in UASB reactors (Alphenaar et al., 1993) whereas, very short hydraulic retention times give rise to possibility of biomass washout. Both scenarios are unfavorable to good performance of the UASB reactor, although granulation has been reported to be necessary for successful domestic wastewater treatment in UASB reactors (Aiyuk and Verstraete, 2004; van Haandel et al., 2006). The success of the UASB reactor also relies on the establishment of a dense sludge bed in the bottom of the reactor

4687

where all biological processes take place. This sludge bed is basically formed by accumulation of incoming suspended solids and bacterial growth. In upflow anaerobic systems, under certain conditions, it was also observed that bacteria can naturally aggregate in flocks and granules (Hulshoff Pol et al., 1983; Hulshoff Pol, 1989). These dense aggregates are not susceptible to washout from the system under practical reactor conditions. Retention of active sludge, either granular or flocculent, within the UASB reactor enables good treatment performance at high organic loading rates. The main reason for the success of the UASB reactor is its relatively high treatment capacity compared to other systems (Driessen and Yspeert, 1999). Natural turbulence caused by the influent flow rate and biogas production provides good wastewaterbiomass contact in UASB systems (Heertjes and van der Meer, 1978). Therefore, less reactor volume and space are required while, at the same time, high grade energy is produced as biogas. Several configurations can be imagined for a wastewater treatment plant including a UASB reactor. In any case, there must be a sand trap, screens for coarse material, and drying beds for the sludge. The UASB reactor may replace the primary settler, the anaerobic sludge digester, the aerobic step (activated sludge, trickling filter, etc.), and the secondary settler of a conventional aerobic treatment plant. However, the effluent from UASB reactors usually needs further treatment, in order to remove remnant organic matter, nutrients and pathogens. This posttreatment can be accomplished in conventional aerobic systems like stabilization ponds, activated sludge plants, and others. The economics of anaerobic treatment in UASB reactors were thoroughly discussed by Lettinga et al. (1983). The advantages and disadvantage of UASB reactors are shown in Table 2. In particular, the UASB reactor is a reliable and simple technology for wastewater treatment (van Haandel and Lettinga, 1994). Several full-scale plants have been in operation and many more are presently under construction, especially under tropical or subtropical conditions (van Haandel et al., 2006). The UASB system has become the most widely applied reactor technology for high-rate anaerobic treatment of industrial effluents. Its relative high treatment capacity compared to other systems permits the use of compact and economic wastewater treatment plants. Compared to aerobic system, it has slow growth rate, mainly associated with methanogenic bacteria. Therefore, it requires a long solids retention time, and also only a small portion of the degradable organic waste is being synthesized to new cells. So far, UASB process technique has been applied for the treatment of palm oil mill effluent, distillery wastewater, slaughterhouse wastewater, piggery wastewater, dairy wastewater, fishmeal process wastewater, municipal wastewater, potato waste leachate, coffee production wastewater, petrochemical wastewater, low strength wastewaters like real cotton processing wastewater and synthetic wastewater. The key design parameters of UASB reactor used by different researchers are shown in Table 3. The most commonly used operational parameters like pH, mixing; operational temperature, hydraulic retention time and organic loading rates are extensively discussed in Table 4 along with COD removal and biogas production.

4688

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 6 8 3 e4 6 9 9

Table 2 e Advantages and disadvantages of UASB reactor. Advantages 1. Good removal efficiency can be achieved in the system, even at high loading rates and low temperatures. 2. The construction and operation of these reactors is relatively simple and low demand for foreign exchange due to possible local production of construction material, plant components, spare parts and low maintenance. 3. Anaerobic treatment can easily be applied on either a very large or a very small scale. 4. When high loading rates are accommodated, the area needed for the reactor is small thus reducing the capital cost. 5. As far as no heating of the influent is needed to reach the working temperature and all plant operations can be done by gravity, the energy consumption of the reactor is less. Moreover, energy is produced during the process in the form of methane. 6. Reduction of CO2 emissions due to low demand for foreign (fossil) energy and surplus energy production. 7. Much less bio-solids waste generated compared with aerobic process because much of the energy in the wastewater is converted to a gaseous form and resulting in very little energy left for new cell growth. 8. The sludge production is low, when compared to aerobic methods, due to the slow growth rates of anaerobic bacteria. The sludge is well stabilized for final disposal and has good dewatering characteristics. It can be preserved for long periods of time without a significant reduction of activity, allowing its use as inoculum for the startup of new reactors. 9. Can handle organic shock loads effectively. 10. Low nutrients and chemical requirement especially in the case of sewage, an adequate and stable pH can be maintained without the addition of chemicals. 11. Macronutrients (nitrogen and phosphorus) and micronutrients are also available in sewage, while toxic compounds are absent.

Disadvantages 1. Pathogens are only partially removed, except helminthes eggs, which are effectively captured in the sludge bed. Nutrients removal is not complete and therefore a post-treatment is required. 2. Due to the low growth rate of methanogenic organisms, longer startup takes before steady state operation, if activated sludge is not sufficiently available. 3. Hydrogen sulphide is produced during the anaerobic process, especially when there are high concentrations of sulfate in the influent. A proper handling of the biogas is required to avoid bad smell and corrosion. 4. Post-treatment of the anaerobic effluent is generally required to reach the surface water discharge standards for organic matter, nutrients and pathogens. 5. Proper temperature control (15e35  C) required for colder climates.

3. Operation and performance of UASB reactor 3.1.

Organic loading rates and COD removal

Organic loading rate is an important parameter significantly affecting microbial ecology and performance of UASB systems. This parameter integrates reactor characteristics, operational characteristics, and bacterial mass and activity

into the volume of media (Torkian et al., 2003). Various studies have proven that higher OLRs will reduce COD removal efficiency in wastewater treatment systems (Patel and Madamwar, 2002; Torkian et al., 2003; Sa´nchez et al., 2005). However, gas production will increase with OLR until a stage when methanogens could not work quick enough to convert acetic acid to methane. Moreover, organic loading rate can also be related to substrate concentration and HRT, thus a good balance between these two parameters has to be obtained for good digester operation. Short HRT will reduce the time of contact between substrate and biomass. Palm oil mill effluent treatment has been successful with UASB reactors, achieving COD removal efficiency up to 98.4% with the highest operating OLR of 10.63 kgCOD m3 d1 (Borja and Banks, 1994). However, reactor operated under overload conditions with high volatile fatty acid content became unstable after 15 days. Due to high amount of POME discharge daily from milling process, it is necessary to operate treatment system at higher OLR. Borja et al. (1996) implemented a twostage UASB system for POME treatment with the objective of preventing inhibition of granule formation at higher OLRs without removing solids from POME prior to treatment. This method is desirable since suspended solids in POME have high potential for gas production while extra costs for sludge disposal can be avoided. Results from this study showed the feasibility of separating anaerobic digestion into two stages (acidogensis and methanogenesis) using a pair of UASB reactors. The methanogenic reactor was found to adapt quickly with the feed from the acidogenic reactor and also tolerate higher OLRs. It was suggested that OLR of 30 kgCOD m3 d1 could ensure an overall of 90% COD reduction and efficient methane conversion. UASB reactor is advantageous for its ability to treat wastewater with low suspended solid content (Kalyuzhnyi et al., 1998) and provide higher methane production (Kalyuzhnyi et al., 1996). Whereas, the packing material in anaerobic filter reactors clogged because of suspended solids and resulted in less biogas production (Stronach et al., 1987). Moreover, suspended and colloidal components of POME in the form of fat, protein, and cellulose have an adverse impact on UASB reactor performance and can cause deterioration of microbial activities and washout of the active biomass (Borja and Banks, 1994; Torkian et al., 2003). However, the reactor might face long startup periods if seed sludge is not granulated. A study by Goodwin et al. (1992) has proved that reactors seeded with granulated sludge can achieve high performance levels within a shorter startup period. It could also acclimatize quickly to gradual increase of OLR (Kalyuzhnyi et al., 1996). The use of two identical UASB reactors by Goodwin and Stuart (1994) operated in parallel as duplicates for 327 days for the treatment of malt whisky pot ale, achieved COD reductions of up to 90% for at influent concentrations of 3.5e5.2 g L1. When the OLRs of 15 kgCOD m3 d1 and above were used, the COD removal efficiency dropped to less than 20% in one of the duplicate reactors. A mesophilic two-stage system consisting of an anaerobic filter (AF) and an UASB reactor was found suitable for anaerobic digestion of distillery waste, enabling better conditions for the methanogenic phase (Blonskaja et al., 2003). An advanced version of UASB system was reported by Driessen and Yspeert (1999), wherein they

4689

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 6 8 3 e4 6 9 9

Table 3 e UASB reactor specifications and working parameters. Type of wastewater

Phase

Reactor Diameter Height Flow volume (L) (cm) (cm) rate (L d1)a

HRT (d)

Upflow Sampling velocity (m h1)a ports

Reference

Palm oil Two mill effluent Single Single

12 5 10 14

13 9 9e12 18.5

90 78 140 52

11.76 4.9 3 383.26b

1 1 3.33 0.53

e e e 0.59

6 6 4 3

Distillery WW

Single Two Three IUASB-Single TUASB-Single CUASB-Single

1.05 10.2 2.3 143,000 476,000 190,000

6.2 9 5 1102 2010 1270

34.8 160 83 1500 1500 1500

0.5 4 1.84 47,667 119,000 38,000

2 2.5 1.25 3 4 5

e e 2 e e e

e 15 e e e e

Goodwin et al. (2001) Laubscher et al. (2001) Keyser et al. (2003),c Ince et al. (2005)

Dairy WW

Phase 1 Phase 2 Single Intermittent Single

12.3 3.2 31.7 6 5

15 9 15.4 9.4 10

70 50 170 86.4 70

3.5 2.13

3.5 1.5 e 0.5 1

e e 0.11 e e

4 3 e 5

Luostarinen and Rintala (2005) Nadais et al. (2005a,b) Nadais et al. (2005a) Tawfik et al. (2008)

Single

7.85

10

100

e

e

5

Huang et al. (2009)

Slaughterhouse Single WW Single Three Single Two

31,840 3 7.2 1000 2

260 6.7 15 e 8

600 85 41 e 15

96,485 24 7.9e12.4 10,000 0.59

0.33 0.13 0.91e0.58 0.1 3.4

e e e 0.33e1.0 e

5 e 3 9 6

Sayed et al. (1984) Cha´vez et al. (2005) Caixeta et al. (2002) Torkian et al. (2003) Ruiz et al. (1997)

Piggery WW

5 3.78 800  103 1 2

15 6W, 6L 940 3.9 5.7

30 105 1160 84 78

1 6 e 3 4.37

5 0.63 e 0.33 0.46

e 2 as Vs e e e

e 6 e 4 4

Sa´nchez et al. (2005) Huang et al. (2005) Miranda et al. (2005) Hendriksen and Ahring (1996)

2.5

6

100

2

1.25

e

5

40 25 55 8

16 11 28 10

240 300 89 100

121 119 172 11.4

0.33 0.21 0.32 0.7

e e e e

e e 1 5

Threed Single

46 15.7

15W, 25L 10

125 200

e 80

0.17e0.13 0.196

0.31e0.43 0.426

4 7

Single

2.3

5

90

6

0.33

1

e

 dag  and Ag Sponza (2005) El-Gohary and Nasr (1999) Behling et al. (1997) Singh and Viraraghavan (1998) Moawad et al. (2009) Uemura and Harada (2000) Aiyuk and Verstraete (2004)

Fishery WW

Single Single Single Two

Municipal WW Two and waste Single Two Single Single

a b c d

12 5

Borja et al. (1996) Chaisri et al. (2007) Siang (2006)

Some values calculated by Eqs. (2) and (3). Total flow rate (influent and recycled). Design adopted from Trnovec and Britz (1998). Only UASB data.

used an internal circulation reactor characterized by biogas separation in two stages within a reactor with a high height/ diameter ratio and the gas driven internal effluent circulation. This system could handle high upflow liquid and gas velocities making possible treatment of low strength effluents at short hydraulic retention times as well as treating high-strength effluents from brewery at very high volumetric loading rates up to 35 kgCOD m3 d1. A three-phased UASB reactor used by Caixeta et al. (2002) for slaughterhouse wastewater treatment at an OLR of 2.7e10.8 kgCOD m3 d1 and average COD removal efficiencies of 85, 84 and 80% and BOD5 of approximately 95% at three different HRT of 22, 18 and 14 h, respectively. Syutsubo et al. (1997) reported a COD loading of 30 kgCOD m3 d1 with

a COD removal efficiency of 85% at sludge loading rates (SLRs) up to 3.7 gCOD g1 VSSd1 using thermophilic reactors (Syutsubo et al., 1998). Organic loading rates (OLRs) up to 104 kgCOD m3 d1 have been reported for anaerobic digestion of sugar substrate under thermophilic conditions (Wiegant and Lettinga, 1985). Torkian et al. (2003) concluded that results under steady state condition where OLRs were between 13 and 39 kgSCOD m3 d1 and HRT of 2e7 h. Removal efficiencies in the range of 75e90% were achieved at feed SCOD concentrations of 3e4.5 g L1. According to Soto et al. (1997), excellent stability and high treatment efficiency can be achieved with hydraulic residence times as low as 2 h at an OLR of 6 kgCOD m3 d1 with the percent COD removals being 92e95%. Sayed et al. (1987) treated effluent

Type of wastewater

Reactor type

Phase

OLR Influent (kgCOD COD (g L1) m3 d1)

HRT (days)

Temperature COD ( C) removal (%)

Biogas (L d1)

CH4 (L d1)

Average CH4%

Reference

POME

POME POME POME

UASB UASB UASB

Single Two Single

42.5 30.6 50

10.63 30 15.5

4 1.02 3.33

35 35 28

96 90 80.5

11.5 10 14

6.9 7 7

60 70 50

Borja and Banks (1994) Borja et al. (1996) Chaisri et al. (2007)

Distillery WW

Recalcitrant distillery WW Distillery WW

UASB

Single

10

19

0.53

55

<67

6.4

3.5

55

Harada et al. (1996)

AF-UASB

Two

UASB

Single

UASB UASB

Two Single

4 2.2 10.2 4.69 17.2 18

19e10 20 2.1 7 1.22 1.67

36  1.5 36  1.5 35 35 35  1.5 34e36

47 93 93 88 92 90  3

0.091e0.39 5.3 4.7 1.3 310e e

0.06e0.26 3.45 e e 238f e

65 65 e e 77f e

Blonskaja et al. (2003)

Malt whisky distillery pot ale Malt whisky WW Grape wine distillery WW Grain distillation WW Winery effluent Raki & Cognac distilleries

8.51e16.8a 13.6b 21.05c 32.86d 20.92 30

UASB UASB IUASB TUASB CUASB

Two Three Single

5.1g 6.4 33 32 23

18.4 5.1 11 8.5 4.5

0.28 1.25 3 4 5

35 35 36  1 36  1 36  1

90  3 86 85 60e80 70e80

e 2.3h 0.078i 0.078i 0.071i

e e 0.045 0.045 0.041

e e 74 74 74

UASB UASBSeptic UASB-AS UASB UASB UASB

Single Phase 1 Phase 2 Single Single Intermittent Single

37 0.63 0.36 2.01 79 13.5 1.8

6.2 0.179 0.24 3.4 7.5 22 13.5

6 3.5 1.5 1 0.66 2 0.13

35 20 10 35 35  1 35  1 30  2

98 73 64 69 74 97  1 >90

e e e e e e e

e e e e 16 54 e

e e e e e e e

UASB

12.48 12.48 55.1 17.8 2.72

12.48 12.48 11.1 8.9 8

1 1 4.95 2 0.33

35  1 35  1 e 35  1 e

90 90 95 87.3 80e90

e e e e e

e e 23.4j 0.27k e

e e e e e

Nadais et al. (2006)

UASB UASB UASB

Continuous Intermittent Two Two Single

UASB UASB UASB

Single Single Three

0.33 0.19 as CRT 0.92 0.75 0.58 0.1 3.4

33 37

70 95 89 90 86 90 93

10,000 e 11.9 10.9 10.6 e 1.03

6500 e e e e 280m 0.6

65e70 e e e e e 70.6

Sayed et al. (1984) Cha´vez et al. (2005),l Caixeta et al. (2002)

Five Two

3.5 28.7 4.6 8.7 10.8 30 2.23

20 24.7 35

UASB UASB-AFo

1.2 5.5 4.2 6.5 6.3 2.87 7.6

Torkian et al. (2003) Ruiz et al. (1997)

Single Two

8.12 2

1.62 3.17

5 0.63

30e35 30  1

75 91

4.1 1.51

2.37n 0.6

57.8 39.7

Sa´nchez et al. (2005) Huang et al. (2005)

Dairy WW Dairy palour WW Dairy & Domestic WW Dairy WW Dairy WW Digested cowdung slurry Dairy WW

Fishery WW

Cheese whey Dairy manure Mixed sardine and tuna canning

Slaughterhouse Slaughterhouse waste WW Poultry slaughter WW Slaughterhouse waste

Slaughterhouse WW Slaughterhouse WW Piggery WW

Piggery waste UASB Pre-settled piggery WW UASB-AS

Goodwin et al. (2001) Uzal et al. (2003) Wolmarans and de Villiers (2002) Laubscher et al., 2001 Keyser et al. (2003) Ince et al. (2005)

Gavala et al. (1999) Sari and Jukka (2005) Tawfik et al. (2008) Nadais et al. (2005a,b) Nadais et al. (2005a) Ramasamy et al. (2004)

Ergu¨der et al. (2001) Garcı´a et al. (2008) Palenzuela-Rollon et al. (2002)

(continued on next page)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 6 8 3 e4 6 9 9

Dairy WW

4690

Table 4 e Operation and performance of UASB reactor with various wastewaters.

Table 4 (continued ) Type of wastewater Municipal WW

a Acidogenic phase. b Methanogenic phase. c With 70% pot ale. d With 100% pot ale. e At influent of COD 16 g L1. f Stoichiometric calculations for CH4. g At controlled conditions. h At OLR of 6.3 kgCOD m3 d1. i As per gVSS. j LCH4 L1 of cheese whey. k LCH4 g1CODremoved. l All data in terms of BOD. m As per SCOD. n Per liter of influent. o Only UASB data. p 1st run data, total 3 runs. q Phase 3 data.

Phase

Influent OLR COD (kgCOD (g L1) m3 d1)

HRT (days)

Temperature COD ( C) removal (%)

Biogas (L d1)

CH4 (L d1)

Average CH4%

Reference

UASB-CSTR Twop

20

16

4.5

37  3

79

9.5

5.7

60.05

 dag  and Sponza (2005) Ag

UASB UASB-SBRo UASB-AS UASB UASB AF-UASBo

0.39 0.37 0.56 3.2 0.15e0.5 0.47e1.23

1.21 2.93 0.09 1.05 0.77e2.55 1.4e3.7

0.32 0.125 0.17 0.42 0.196 0.33

30 e e 20  1 25e13 12e23

85 57 85 86 68  4 <50

26 e e 1.97 e e

e e e 1.1 3.5 e

e e e 79 e e

Behling et al. (1997) Moawad et al. (2009) von Sperling et al. (2001) Singh and Viraraghavan (1998) Uemura and Harada (2000) Sawajneh et al. (2010)

Single Three Fiveq Single Single Two

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 6 8 3 e4 6 9 9

Municipal landfill leachate Domestic WW Domestic WW Municipal WW Municipal WW Sewage WW Sewage WW

Reactor type

4691

4692

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 6 8 3 e4 6 9 9

from meat processing plant in a granular UASB reactor and achieved COD removal efficiency of 55e85% with HRT of 0.5e0.6 days at volumetric loading rate of 11 kgCOD m3 d1. Dague and Pidaparti (1992) concluded that operation of reactor with hydraulic retention time of 8.8 days and OLR of 0.33 kgBOD5 m3 d1 yielded a BOD5 removal efficiency of 85e90% and biogas production of 0.51 m3CH4 kg1CODremoved. Two-phased UASB-septic tank used by Luostarinen and Rintala (2005) with high removal of organic matter for onsite treatment of synthetic black water (OLR 0.301 kgCOD m3 d1) and dairy parlor wastewater (OLR 0.191 kgCOD m3 d1) at low temperatures (10e20  C). Moreover, CODdiss removal was around 70% at 15  C and 10  C indicating good biological activity of the reactor sludge. Gavala et al. (1999) concluded that an OLR of 6.2 gCOD L1 d1 (diluted to 37 gCOD L1, with an HRT to 6 d) may be safely used for treating dairy wastewater and could be increased up to 7.5 gCOD L1 d1. Above that OLR, reduced performance is observed; while for nondiluted dairy wastewater, an HRT in excess of 30 d is required. According to Palenzuela-Rollon et al. (2002) the application of UASB system is a promising treatment option for fish processing wastewater. They determined the performance of USAB reactor for the treatment of mixed sardine and tuna canning effluent at varying lipid levels. They stated that at low lipid level (203e261 mg L1, 9% of total COD) approximately 78% COD removal and 61% COD conversion to methane can be achieved with an OLR of 2.3 gCOD L1 d1 and at HRT of 7.2  2.8 h. In the case of high-lipid wastewater a two-step UASB was recommended where the total COD removal and conversion to methane were 92% and 47%, respectively. Punal and Lema (1999) have used a 380 m3 UASB reactor for the treatment of fish-canning factory wastewater. The wastewater was a mixed effluent of tuna, sardine and mussel processing. The total alkalinity of more than 3 gCaCO3 L1 was maintained to operate the system properly and allow biomass to resist load shocks. An HRT of 2 days was maintained and the OLR was varied from 1 kgCOD m3 d1 to 8 kgCOD m3 d1. The efficiency of the system is dependent on the nature of the wastewater as shown in Table 4. The organic loading rate can be calculated by the following equation,

OLR ¼

CODin HRT

(1)

where OLR ¼ Organic loading rate (gCOD L1 d1), CODin ¼ Influent COD (g L1), HRT ¼ Hydraulic retention time (days).

3.2.

Flow rate and hydraulic retention time

Flow rate is also an important operating parameter which upholds the hydraulic retention time. A lot of data has been published regarding flow rate which also can justify by Eq. (2). Borja et al. (1996) worked for POME at a maximum flow rate of 11.76 L d1 and 4.9 L d1 at 24 h HRT for two-stage UASB reactor. The difference in flow rate at same HRT is due to change in volume of the reactor from 12 to 5 L. Experiment carried out by Siang (2006) for POME, where low HRT of 12.7 h was worked out at recycling mode of UASB and maintained a flow rate of 383.26 L d1. This low HRT might be due to less

height of the reactor (52 cm) and bigger diameter (18.5 cm). The purpose of describing height/diameter combination is that, during upflow anaerobic process if diameter will be too big then there is a chance of liquid channeling in the reactor. Moreover, because of channeling, the influent stream may not be in full contact with reactor biomass which will result in low conversion of organic matter into fatty acids and finally biogas production. So, bigger reactor diameter does not encourage more biogas production except sludge washout because of poor mixing within the reactor. On the other hand, comparatively more height may encourage substrate mixing which leads to proper contact of influent with micro-organisms which results in more organic matter conversion into biogas. Chaisri et al. (2007) and Laubscher et al. (2001) used almost 10 L UASB reactors with flow rates of 3 L d1 and 4 L d1 respectively. The difference in flow rate is due to the change in HRT only which might be designed according to the type of wastewater, where, in first case POME was used as a substrate and distillery wastewater for later one. Ince et al. (2005) worked out at three full-scale UASB reactors at 3e4 d HRT while change in flow rates is due to the different reactor volumes. Torkian et al. (2003) used a pilot scale UASB reactor for the treatment of slaughterhouse wastewater. They kept flow rate of 1000 L d1 at only 2.4 h HRT. Tawfik et al. (2008) studied on dairy wastewater and used a flow rate of 5 L d1 and at 1 d HRT while reactor height was also sufficient to run experiment at higher HRTs. In another study, Sa´nchez et al. (2005) used piggery wastewater as substrate and worked at very less flow rate of 1 L d1 while keeping HRT of 5 days. They used 30 cm high and 15 cm diameter UASB reactor. This higher HRT for lower height might has different reasons like, This long HRT for less height might has different reasons like, the use of raw wastewater rich in organic contents, or may be less dilution has been done for startup of UASB reactor, or, they used very less or no nutrients for running the experiment. Huang et al. (2005) used 105 cm high rectangular UASB reactor and worked only at 15 h HRT with a flow rate of 6 L d1. The success of UASB reactor also done by using municipal wastewater, Moawad et al. (2009) worked at two hydraulic retention times 4 h and 3 h but they did not mention the flow rate in their study. By using Eq. (2), the flow rate could be  dag  270e353 L d1 while they used rectangular UASB rector. Ag and Sponza (2005) studied degradation of municipal wastewater at two-stage UASB reactor of same volume (2.5 L) and designed flow rate was 2 L d1 at 1.25 d HRT. The low flow rate accompanied due to less reactor volume (2.8 L). Complete detail of UASB reactor configuration with flow rate and HRT designed by various researchers is shown in Table 3. Some data regarding flow rate is calculated by Eq. (2). By following Eq. (2), it is clear that flow rate is inversely proportional to HRT but it is understood that volume has direct relation with flow rate. For designing purpose, we can calculate the flow rate by following equation. Q¼

V HRT

(2)

where Q ¼ Flow rate of influent stream (L d1), V ¼ Volume of the reactor (L), HRT ¼ Hydraulic retention time (days).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 6 8 3 e4 6 9 9

3.3.

Upflow velocity

The upflow velocity (Vup) is also an important operational parameter in upflow digesters. It maintains the mixing and hydraulic retention time of the substrate and biomass. A few researches are done on upflow velocity as limited data had shown by few researchers. Upflow velocity is directly proportional to reactor height and inversely proportional to hydraulic retention time at fix HRT or reactor volume, as shown in Eq. (3). It determines the appropriate mixing of biomass with the height of the reactor with or without channeling. The permissible limit of upflow velocity is 0.5e1.5 m h1 as described by many researchers. Siang (2006) maintained a 0.59 m h1 Vup at HRT of 13 h. Keyser et al. (2003) reported a 2 m h1 Vup at HRT of 1.25 days which is quite higher than limits but they used distillery wastewater as substrate and little fluctuations might be possible with different types of wastewaters. While treating dairy wastewater, Nadais et al. (2005a,b) reported a 0.11 m h1 Vup without showing hydraulic retention time. Torkian et al. (2003) treated slaughterhouse wastewater and applied Vup from 0.33 to 1.0 m h1 while keeping a constant HRT of 2.4 h. They used a pilot scale UASB reactor of 1 m3 capacity and height was not shown in their study. While treating municipal wastewater in UASB reactor, Uemura and Harada (2000) reported 0.426 m h1 Vup at 4.7 h of HRT. Recently, Moawad et al. (2009) used upflow velocities of 0.31e0.43 m h1 at HRT of 4e3 h. They used a rectangular shaped UASB reactor for the treatment of municipal wastewater. Upflow velocity can be determined by Eq. (3) on the basis of HRT and height of the reactor. Vup ¼

h HRT

(3)

where Vup ¼ Upflow velocity of influent stream (m h1), h ¼ height of the reactor (m), HRT ¼ Hydraulic retention time (h). The upflow velocity can also be calculated by flow rate and cross-sectional area of the reactor, Vup ¼

Q A

(4)

where Vup ¼ Upflow velocity of influent stream (m h1), Q ¼ Flow rate of influent stream (m3 h1), A ¼ Reactor’s crosssectional area (m2).

3.4.

pH

The microbial community in the anaerobic digester is sensitive to the changes of pH and methanogens are affected to a greater extent (Grady et al., 1999). An investigation by Beccari et al. (1996) confirmed that methanogenesis is strongly affected by pH. As such, methanogenic activity will decrease when pH in the digester deviates from the optimum value. Optimum pH for most microbial growth is between 6.8 and 7.2 while the pH values less than 4 and more than 9.5 are not tolerable (Gerardi, 2006). Several cases of reactor failure have been reported in various studies of wastewater treatment due to accumulation of high concentration of volatile fatty acid, causing a drop in pH which inhibited methanogenesis (Patel

4693

and Madamwar, 2002; Parawira et al., 2006). Thus, volatile fatty acid concentration is an important parameter to monitor to guarantee reactor performance (Buyukkamaci and Filibeli, 2004). It was found that digester could tolerate acetic acid concentrations up to 4000 mg L1 without inhibition of gas production (Stafford, 1982). To control the level of volatile fatty acid in the system, alkalinity has to be maintained by recirculation of treated effluent (Borja et al., 1996; Najafpour et al., 2006) to the digester or addition of lime and bicarbonate salt (Gerardi, 2003). While treating slaughterhouse wastewater in UASB reactor, the pH should be essentially constant, varying between 7.5 and 8.5. The pH of the slaughterhouse influent is normally corrected to 7.0 but it has little variation due to acid forming bacteria. Caixeta et al. (2002) maintained the pH of influent stream by adding sodium bicarbonate in first run. However, in further two runs the influent was fed to the reactor without any pH adjustment. The slaughterhouse wastewater exhibited high buffering capacity without requiring pH correction any more. Sandberg and Ahring (1992) investigated the influence of high pH on anaerobic degradation of fish processing wastewater in a UASB reactor. According to Boone and Xun (1987) most methanogenic bacteria have optima for growth between pH 7 and 8, whereas VFA degrading bacteria have lower pH optima. The optimal pH for mesophilic biogas reactor is 6.7e7.4 (Clark and Speece, 1971). A study by Sandberg and Ahring (1992) demonstrated that fish condensate can be treated well in a UASB reactor from pH 7.3 to 8.2. When the pH was increased slowly to 8.0 or more 15e17% drop in COD removal occurred. Acetate was the only carbon source in the condensate that accumulated upon increasing the pH. More than 99% of VFA and TMA in process wastewater were degraded up to pH 7.9. It was concluded that gradual pH increment was essential in order to achieve the necessary acclimatization of the granules and to prevent disintegration of the granules and that the pH should not exceed 8.2. Aspe et al. (2001) modeled the ammonia-induced inhibition phenomenon of anaerobic digestion and concluded that methanogenesis was the most inhibited stage.

3.5.

Operating temperature

Temperature is an important operating parameter for anaerobic degradation process. The influence of temperature on microbial growth and biodegradation rate can be described by the Arrhenius equation (Batstone et al., 2002; Hao et al., 2002; Siegrist et al., 2002). Operation of anaerobic reactors under thermophilic conditions offers a number of advantages such as increased reaction rates and improved biodegradability of organic compounds (Rintala, 1997; Kim et al., 2002). However, startup and operation of a thermophilic reactor is cumbersome due to the high sensitivity of thermophilic microorganisms to variations in OLR, influent composition, reactor pH, and other factors. It is generally assumed that a transition from mesophilic to thermophilic conditions is accompanied by a significant (over 80%) and lengthy (over 4 days) decrease in methane production due to adaptation of methanogens to thermophilic temperatures (van Lier et al., 1992; Visser et al., 1993) Nevertheless, mesophilic methanogenic populations were shown to tolerate short-term temperature are increases

4694

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 6 8 3 e4 6 9 9

(Speece and Kem, 1970; Ahn and Forster, 2002) or sludge exchange between mesophilic and thermophilic reactors (Song et al., 2004). The cost benefit analysis for POME treatment system that utilizes biogas for electricity generation and digester effluent for land application also showed a faster payback (Yeoh, 2004). POME is discharged at temperatures around 80e90  C (Zinatizadeh et al., 2006) which actually makes treatment at both mesophilic and thermophilic temperatures feasible especially in tropical countries like Malaysia. Various studies have been conducted to investigate the feasibility of operating wastewater treatment systems in the thermophilic temperature range such as sugar, high-strength wastewater (Wiegant et al., 1985; Wiegant and Lettinga, 1985) and POME (Cail and Barford, 1985; Choorit and Wisarnwan, 2007). It is reported that operation at thermophilic temperature gives better results than mesophilic temperature after startup because methane producing bacteria produced at mesophilic temperature facilitate in high methane production at thermophilic temperature ranges (Cail and Barford, 1985). High production of methane was also observed from the treatment of sugar wastewater in this higher temperature range. Effect of temperature on the performance of anaerobic digestion was investigated. Yu et al. (2002) found that substrate degradation rate and biogas production rate at 55  C were higher than operation at 37  C. Studies have reported that thermophilic digesters are able to tolerate higher OLRs and operate at shorter HRT while producing more biogas (Ahn and Forster, 2002; Kim et al., 2006; Yilmaz et al., 2008). However, failure to control temperature increase can result in biomass washout (Lau and Fang, 1997) with accumulation of volatile fatty acid due to inhibition of methanogenesis. At high temperatures, production of volatile fatty acid is higher compared to mesophilic temperature range (Yu et al., 2002). Many operators prefer to have digesters operating in mesophilic temperature due to better process stability. Nevertheless, investigation on digester stability by Kim et al. (2002) proved that disadvantages of thermophilic digesters can be resolved by keeping microbial consortia in close proximity. Full-scale thermophilic (50e55  C) anaerobic digestion of wastewater from an alcohol distillery was reported by Vlissidis and Zouboulis (1993). More than 60% removal of COD was achieved with 76% of biogas comprising of methane thus making it a valuable fuel. Stevens and Schulte (1979) studied the effect of the temperature at solids retention times of 6e55 days at organic volumetric loading rates of 0.61e4.81 kgVS m3 d1, in a complete mixed anaerobic digester. They concluded that at organic rates in the range of 0.61e1.80 kg VS m3 d1 and temperatures lower than 25  C, the operation is proceeded satisfactorily. Low temperature digestion was found to require twice as long retention time as with satisfactory production and composition of gas. In another study, Sa´nchez et al. (2001) observed the effect of temperature and substrate concentration on the anaerobic batch digestion of piggery wastewater. The study compared the process at mesophilic temperature (35  C) with temperatures in the range of 16.8e29.5  C, and influent concentrations in the range of 3.3e26.3 gTCOD L1. The process at mesophilic temperature was more stable than at an ambient temperature, obtaining higher values of removal efficiency.

3.6.

Mixing

Mixing provides good contact between microbes and substrates, reduces resistance to mass transfer, minimizes buildup of inhibitory intermediates and stabilizes environmental conditions (Grady et al., 1999). When mixing is inefficient, overall rate of process will be impaired by pockets of material at different stages of digestion, whereby every stage has a different pH and temperature (Stafford, 1982). Mixing can be accomplished through mechanical mixing, biogas recirculation or through slurry recirculation (Karim et al., 2005a). Investigations have been done to observe the effects of mixing to the performance of anaerobic digesters. It was found that mixing improved the performance of digesters treating waste with higher concentration (Karim et al., 2005b) while slurry recirculation showed better results compared to impeller and biogas recirculation mixing mode (Karim et al., 2005c). Mixing also improved the gas production as compared to unmixed digesters (Karim et al., 2005b). Intermittent mixing is advantageous over vigorous mixing (Stafford, 1982; Kaparaju et al., 2008), where this has been adopted widely in large-scale municipal and farm waste digesters (Stafford, 1982). Sludge granules are formed due to fluidization (Guiot et al., 1992). Fluidization is achieved by mixing of the sludge by the flow and gas release. Rapid mixing is not encouraged as methanogens can be less efficient in this mode of operation (Gerardi, 2003). However, Karim et al. (2005b) mentioned that mixing during startup is not beneficial due to the fact that digester pH will be lowered, resulting in performance instability as well as leading to a prolonged startup period. Mixing in palm oil mills which depend on biogas produced (Ma and Ong, 1985) is less efficient compared to mechanical mixing as digesters are not perfectly mixed. The upflow reactors with big diameters can face the problem of channeling where upflow velocity, sometimes, cannot improve the mixing of more viscous substrate. So, mixing becomes the important functional parameter for such cases. Further investigation on the effects of mixing should be commenced to obtain a suitable mode of mixing for better digester performance.

4.

Research needs

The application of UASB reactors for the treatment of wastewaters is limited so far to regions with constant and relatively warm temperature conditions. The success of UASB reactors is mainly dependent on OLR, HRT and operating temperatures. Operating temperature can be a fixed parameter with minor fluctuations but the key factors (OLR and HRT) determine the ultimate amount of hydrolysis and methanogenesis in a UASB system especially at early stages. Organic loading rate is still a challenge for researchers to produce maximum biogas and high COD removal. The COD removal or biogas production is inhibited by the accumulation of fatty acid where, sometimes, system becomes unstable and results in sludge washout. Accumulation of undegraded SS may also induce a reduction in the methanogenic activity of the sludge,

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 6 8 3 e4 6 9 9

a deterioration of bacterial aggregates, and the formation of scum layers, leading to overloading of the reactor. When too long SRTs are necessary and, therefore, only low loading rates can be accommodated, a two-step system with additional sludge stabilization is to be considered. Despite considerable work devoted to the clarification of the mechanisms of suspended solids removal and hydrolysis, both the physical and the biological processes in the first step of a two-stage anaerobic treatment need further research. The removal of suspended solids will depend on factors like HRT, Vup, and sludge bed distinctiveness, and also on the characteristics of the suspended solids themselves. Mathematical modeling of the system, including physical and biological processes, can help to expand more approach into the process, and will certainly provide a balance for the adequate management of the sludge retention in UASB reactors. A model should also provide a basis for deciding for one or two-stage anaerobic systems according to wastewater characteristics and atmospheric conditions. Mathematical modeling can be extremely valuable in orienting future research on UASB technology, and can serve as a design tool for the expansion, relocate, and distribution of anaerobic technology for direct wastewater treatment. The treatment of different wastewater although can permit wastewaters to flow along with the normal surface water but biogas production should also be keep in account. The cost benefit ratio of the UASB reactor technology can be further decreased if more biogas will be produced.

5.

Conclusions

Upflow anaerobic sludge blanket reactor is an efficient wastewater treatment technology that connects anticipated anaerobic decomposition to lessen the waste volume and produce biogas. It has been broadly applied to the treatment of wastewaters from agricultural and industrial operations. Depending on the starting point, the waste stream may contain inhibitory or toxic substances such as ammonia, sulfide, heavy metals and organics. Accumulation of these substances may cause reactor suffering or failure, as pointed out by reduced biogas production or methane content. Because of the difference in anaerobic micro-organisms, wastewater composition, and experimental methods and conditions, results from previous investigations on inhibition of anaerobic processes vary considerably. The reactor failure is also mainly concerned with operating parameters such as OLR, HRT, pH, food to micro-organisms ratio, VFA to alkalinity ratio and the flow rate. On spot and time to time analysis of incoming wastewater stream and biomass within the reactor are important for both lab and large-scale UASB reactors.

Acknowledgment This work is carried out by the Faculty of Civil Engineering and Earth Resources of University Malaysia Pahang (Malaysia). It is the part of a Research Management Center (RMC) research project on integrated application and design of upflow

4695

anaerobic sludge blanket reactor funded by University Malaysia Pahang. Information provided by the library staff is gratefully acknowledged.

references

 dag  , O.N., Sponza, D.T., 2005. Anaerobic/aerobic treatment of Ag municipal landfill leachate in sequential two-stage up-flow anaerobic sludge blanket reactor (UASB)/completely stirred tank reactor (CSTR) systems. Proc. Biochem. 40, 895e902. Ahn, J.H., Forster, C.F., 2002. A comparison of mesophilic and thermophilic anaerobic upflow filters treating paperepulpeliquors. Proc. Biochem. 38, 257e262. Aiyuk, S., Verstraete, W., 2004. Sedimentological evolution in an UASB treating SYNTHES, a new representative synthetic sewage, at low loading rates. Bioresour. Technol. 93, 269e278. Albrechtsen, H.J., 1998. Water Consumption in Residences. Microbiological Investigations of Rain Water and Greywater Reuse Systems. Miljøstyrelsen (Miljø-og Energiministeriet) og Boligministeriet, ISBN 87-985613-9-1 (in Danish). Alphenaar, P.A., Visser, A., Lettinga, G., 1993. The effect of liquid upward velocity and hydraulic retention time on granulation in UASB reactors treating wastewater with a high sulphate content. Bioresour. Technol. 43 (3), 249e258. Andrew, B., Xiaodong, S., Edyveam, G.J., 1997. Removal of coloured organic matter by adsorption onto low cost-waste material. Water Res. 31, 2084e2092. Angelidaki, I., Ahring, B.K., Deng, H., Schmidt, J.E., 2002. Anaerobic digestion of olive oil mill effluents together with swine manure in UASB reactors. Water Sci. Technol. 45 (10), 213e218. Aspe, E., Marti, M.C., Jara, A., Roeckel, M., 2001. Ammonia inhibition in the anaerobic treatment of fishery effluents. Water Environ. Res. 73 (2), 154e164. Barreto, C.O., 2004. Tratamento de efluentes na indu´stria frigorı´fica e Parte 3. Revista da Carne 327, 138e141. Batstone, D.J., Keller, J., Angelidaki, I., Kalyuzhnyi, S.V., Pavlostathis, S.G., Rozzi, A., Sanders, W.T.M., Siegrist, H., Vavilin, V., 2002. Anaerobic Digestion Model No 1 (ADM1). IWA Publishing, London, UK. Beccari, M., Bonemazzi, F., Majone, M., Riccardi, C., 1996. Interaction between acidogenesis and methanogenesis in the anaerobic treatment of olive oil mill effluents. Water Res. 30, 183e189. Behling, E., Diaz, A., Colina, G., Herrera, M., Gutierrez, E., Chacin, E., Fernandez, N., Forster, C.F., 1997. Domestic wastewater treatment using a UASB reactor. Bioresour. Technol. 61 (3), 239e245. Blonskaja, V., Menert, A., Vilu, R., 2003. Use of two-stage anaerobic treatment for distillery waste. Adv. Environ. Res. 7, 671e678. Boone, D.R., Xun, L., 1987. Effects of pH, temperature and nutrients on propionate degradation by a methanogenic enrichment culture. Appl. Environ. Microbiol. 53, 1589e1592. Borja, R., Banks, C., 1994. Anaerobic digestion of palm oil mill effluent using an up-flow anaerobic sludge blanket (UASB) reactor. Biomass Bioener., 381e389. Borja, R., Banks, C.J., Sa´nchez, E., 1996. Anaerobic treatment of palm oil mill effluent in a two-stage up-flow anaerobic sludge blanket (UASB) reactor. J. Biotechnol. 45, 125e135. Bustamante, M.A., Paredes, C., Moral, R., Moreno-Casalles, J., Pe´rezEspinosa, A., Pe´rez-Murcia, M.D., 2005. Uses of winery and distillery effluents in agriculture: characterization of nutrient and hazardous components. Water Sci. Technol. 51 (1), 145e151. Buyukkamaci, N., Filibeli, A., 2004. Volatile fatty acid formation in anaerobic hybrid reactor. Proc. Biochem. 39, 1491e1494. Cail, R.G., Barford, J.P., 1985. Thermophilic semi-continuous anaerobic digestion of palm-oil mill effluent. Agricul. Wastes 13, 295e304.

4696

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 6 8 3 e4 6 9 9

Caixeta, C.E.T., Camarota, M.C., Xavier, A.M.F., 2002. Slaughterhouse wastewater treatment: evaluation of a new three-phase separation system in a UASB reactor. Bioresour. Technol. 81, 61e69. Chaisri, R., Boonsawang, P., Prasertsan, P., Chaiprapat, S., 2007. Effect of organic loading rate on methane and volatile fatty acids productions from anaerobic treatment of palm oil mill effluent in UASB and UFAF reactors. Songklanakarin J. Sci. Technol. 2, 311e323. Cha´vez, P.C., Castillo, L.R., Dendooven, L., Escamilla-Silva, E.M., 2005. Poultry slaughter wastewater treatment with an up-flow anaerobic sludge blanket (UASB) reactor. Bioresour. Technol. 96, 1730e1736. Cheah, S.C., Ma, A.N., Ooi, L.C.L., Ong, A.S.H., 1998. Biotechnological applications for the utilisation of wastes from palm oil mills. Fat Sci. Technol. 90, 536e540. Chen, T.H., Shyu, W.H., 1998. Chemical characterization of anaerobic digestion treatment of poultry mortalities. Bioresour. Technol. 63, 37e48. Choorit, W., Wisarnwan, P., 2007. Effect of temperature on the anaerobic digestion of palm oil mill effluent. Elect. J. Biotechnol. 10, 376e385. Christova-Boal, D., Eden, R.E., McFarlane, S., 1996. An investigation into grey water reuse for urban residential properties. Desalination 106, 391e397. Clark, R.H., Speece, R.E., 1971. The pH tolerance of anaerobic digestion. Adv. Water Pollut. Res. 1, 1e14. Coetzee, G., Malandra, L., Wolfaardt, G.M., Viljoen-Bloom, M., 2004. Dynamics of microbial biofilm in a rotating biological contactor for the treatment of winery effluent. Water SA 30 (3), 407e412. Dague, R., Pidaparti, S.R., 1992. Anaerobic sequencing batch treatment of swine wastes. In: Proceedingsof 46th Purdue Industrial Waste Conference, Chelsea, pp. 751e823. de Zeeuw, W., 1988. Granular sludge in UASB reactors. In: Lettinga, G., Zehnder, L.I.B., Grotenhuis, J.T.C., Hulshoff Pol, L. W. (Eds.), Granular Anaerobic Sludge; Microbiology and Technology. Pudoc Wageningen, The Netherlands, pp. 132e145. del Pozo, R., Diez, V., Beltra´n, S., 2000. Pretreatment of anaerobic of slaughterhouse wastewater that uses fixed-film reactors. Bioresour. Technol. 71, 143e149. Drechsel, P., Evans, A.E.V., 2010. Wastewater use in irrigated agriculture. Irrig Drainage Syst. 24, 1e3. Driessen, W., Yspeert, P., 1999. Anaerobic treatment of low, medium and high strength effluent in agro-industry. Water Sci. Technol. 40, 221e228. Droste, R.L., 1997. Theory and Practice of Water and Wastewater Treatment. John Wiley and Sons, New York. El-Gohary, F.A., Nasr, F.A., 1999. Cost effective pre-treatment of wastewater. Water Sci. Technol. 39 (5), 97e103. Ergu¨der, E.H., Tezel, U., Gu¨ven, E., Demirer, G.N., 2001. Anaerobic biotransformation and methane potential in batch and UASB reactors. Waste Manag. 21, 643e650. Euse´bio, A., Petruccioli, M., Lageiro, M., Federici, F., Duarte, J.C., 2004. Microbial characterization of activated sludge in jet-loop bioreactors treating winery wastewaters. J. Ind. Microbiol. Biotechnol. 31 (1), 29e34. Fang, H.H.P., Li, Y.Y., Chui, H.K., 1995. UASB treatment of wastewater with concentrated mixed VFA. ASCE J. Environ. Eng. 121 (2), 153e160. Feachem, R.G., Bradley, D.J., Garelick, H., Mara, D.D., 1983. Sanitation and Disease. Health Aspects of Excreta and Wastewater Management. The World Bank, pp. 16e21. Garcı´a, H., Rico, C., Garcı´a, P.A., Ric, J.L., 2008. Flocculants effect in biomass retention in a UASB reactor treating dairy manure. Bioresour. Technol. 99, 6028e6036. Gavala, H.N., Kopsinis, H., Skiadas, I.V., Stamatelatou, K., Lyberatos, G., 1999. Treatment of dairy wastewater using an

upflow anaerobic sludge blanket reactor. J. Agric. Eng. Res. 73, 59e63. Gerardi, M.H., 2003. The Microbiology of Anaerobic Digesters. Wiley-Interscience, New Jersey, pp. 51e57. Gerardi, M.H., 2006. Wastewater Bacteria. Wiley-Interscience, New Jersey, pp. 19e31. Gonzalez, J.F., 1996. Wastewater Treatment in the Fishery Industry. FAO Fisheries Technical Paper (FAO), No. 355/FAO. Fisheries Dept, Rome (Italy). Goodwin, J.A.S., Stuart, J.B., 1994. Anaerobic digestion of malt whisky distillery pot ale using upflow anaerobic sludge blanket reactors. Bioresour. Technol. 49, 75e81. Goodwin, J.A.S., Finlayson, J.M., Low, E.W., 2001. A further study of the anaerobic biotreatment of malt whisky distillery pot ale using an UASB system. Bioresour. Technol. 78, 155e160. Goodwin, J.A.S., Wase, D.A.J., Forster, C.F., 1992. Pre-granulated seeds for UASB reactors: how necessary are they? Bioresour. Technol. 41, 71e79. Grady Jr., C.P.L., Daigger, G.T., Lim, H.C., 1999. Biological Wastewater Treatment, second ed., revised and expanded. Marcel Dekker, Inc., New York. Guiot, S.R., Arcand, Y., Chavarie, C., 1992. Advantages of fluidization on granule size and activity development in upflow anaerobic sludge bed reactors. Water Sci. Technol. 26, 897e906. Halalsheh, M., Dalahmeh, S., Sayed, M., Suleiman, W., Shareef, M., Mansour, M., Safi, M., 2008. Grey water characteristics and treatment options for rural areas in Jordan. Bioresour. Technol. 99, 6635e6641. Hansen, C., West, G.T., 1992. Anaerobic digestion of rendering waste in an upflow anaerobic sludge blanket digester. Bioresour. Technol. 41, 181e185. Hao, X., Heijnen, J.J., van Loosdrecht, M.C.M., 2002. Sensitivity analysis of a biofilm model describing a one-stage completely autotrophic nitrogen removal (CANON) process. Biotechnol. Bioeng. 77 (3), 266e277. Harada, H., Uemura, S., Chen, A.C., Jayadevan, J., 1996. Anaerobic treatment of a recalcitrant distillery wastewater by a thermophilic UASB reactor. Bioresour. Technol. 55, 215e221. Heertjes, P.M., van der Meer, R.R., 1978. Dynamics of liquid flow in an up-flow reactor used for anaerobic treatment of wastewater. Biotechnol. Bioeng. 20, 1577e1594. Hendriksen, H.V., Ahring, B.K., 1996. Integrated removal of nitrate and carbon in an upflow anaerobic sludge blanket (UASB) reactor: operating performance. Water Res. 30 (6), 1451e1458. Huang, J.S., Wu, C.S., Chen, C.M., 2005. Microbial activity in a combined UASB-activated sludge reactor system. Chemosphere 61, 1032e1041. Huang, L., Zhang, B., Gao, B., Feng, L., 2009. Application of anaerobic granular sludge to treatment of fishmeal industry wastewaters under highly saline conditions. In: International Conference on Energy and Environ. Technol., 16e18 October, Guilin, China, pp. 433e436. doi:10.1109/ ICEET.2009.343. Hulshoff Pol, L.W., 1989. The phenomenon of granulation of anaerobic sludge. PhD thesis, Wageningen Agricultural University, Wageningen, The Netherlands. Hulshoff Pol, L.W., de Zeeuw, W.J., Velzeboer, C.T.M., Lettinga, G., 1983. Granulation in UASB-reactors. Water Sci. Technol. 15, 291e304. Hwu, C.S., 1997. Enhancing anaerobic treatment of wastewaters containing oleic acid. PhD thesis, Agricultural University of Wageningen, Wageningen, The Netherlands. Ince, O., Kolukirik, M., Oz, N.A., Ince, B.K., 2005. Comparative evaluation of full-scale UASB reactors treating alcohol distillery wastewaters in terms of performance and methanogenic activity. J. Chem. Technol. Biotechnol. 80, 138e144.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 6 8 3 e4 6 9 9

Jantsch, T.G., Angelidaki, I., Schmidt, J.E., de Hvidsten, B.E.B., Ahring, B.K., 2002. Anaerobic biodegradation of spent sulphite liquor in a UASB reactor. Bioresour. Technol. 84 (1), 15e20. Jeganathan, J., Nakhla, G., Bassi, A., 2006. Long-term performance of high-rate anaerobic reactors for the treatment of oily wastewater. Environ. Sci. Technol. 40, 6466e6472. Jeison, D., Del Rio, A., Van Lier, J.B., 2008. Impact of high saline wastewaters on anaerobic granular sludge functionalities. Water Sci. Technol. 57 (6), 815e819. Jimenez, A.M., Borja, R., 1997. Influence of aerobic pretreatment with Penicillium decumbens on the anaerobic digestion of beet molasses alcoholic fermentation wastewater in suspended and immobilized cell bioreactors. J. Chem. Technol. Biotechnol. 69, 193e202. Kalyuzhnyi, S., de los Santos, L.E., Martinez, J.R., 1998. Anaerobic treatment of raw and preclarified potatoemaize wastewater in a UASB reactor. Bioresour. Technol. 66, 195e199. Kalyuzhnyi, S.V., Skylar, V.I., Davlyatshina, M.A., Parshina, S.N., Simankova, M.V., Kostrikina, N.A., Nozhevnikova, A.N., 1996. Organic removal and microbiological features of UASB-reactor under various organic loading rates. Bioresour. Technol. 55 (1), 47e54. Kaparaju, P., Buendia, I., Ellegaard, L., Angelidaki, I., 2008. Effects of mixing on methane production during thermophilic anaerobic digestion of manure: lab-scale and pilot-scale studies. Bioresour. Technol. 99, 4919e4928. Karim, K., Gupta, S.K., 2003. Continuous biotransformation and removal of nitrophenols under denitrifying conditions. Water Res. 37, 2953e2959. Karim, K., Klasson, K.T., Hoffmann, R., Drescher, S.R., DePaoli, D. W., Al-Dahhan, M.H., 2005a. Anaerobic digestion of animal waste: effect of mixing. Bioresour. Technol. 96, 1607e1612. Karim, K., Hoffmann, R., Klasson, K.T., Al-Dahhan, M.H., 2005b. Anaerobic digestion of animal waste: effect of mode of mixing. Water Res. 39, 3597e3606. Karim, K., Hoffmann, R., Klasson, T., Al-Dahhan, M.H., 2005c. Anaerobic digestion of animal waste: waste strength versus impact of mixing. Bioresour. Technol. 96, 1771e1781. Keyser, M., Witthuhn, R.C., Ronquest, L.C., Britz, T.J., 2003. Treatment of winery effluent with upflow anaerobic sludge blanket (UASB) e granular sludges enriched with Enterobacter sakazakii. Biotechnol. Lett. 25 (22), 1893e1898. Kim, J.K., Oh, B.R., Chun, Y.N., Kim, S.W., 2006. Effects of temperature and hydraulic retention time on anaerobic digestion of food waste. J. Biosci. Bioeng. 102, 328e332. Kim, M., Ahn, Y.H., Speece, R.E., 2002. Comparative process stability and efficiency of anaerobic digestion; mesophilic vs. thermophilic. Water Res. 36, 4369e4385. Kosseva, M.R., Kent, C.A., Lloyd, D.R., 2003. Thermophilic bioremediation strategies for a dairy waste. J. Biochem. Eng. 15, 125e130. Kujawa-Roeleveld, K., Zeeman, G., 2006. Anaerobic treatment in decentralised and source-separation based sanitation concepts. Rev. Environ. Sci. Biotechnol. 5, 115e139. Lau, I.W.C., Fang, H.H.P., 1997. Effect of temperature shock to thermophilic granules. Water Res. 31, 2626e2632. Laubscher, A.C.J., Wentzel, M.C., Le Roux, J.M.W., Ekama, G.A., 2001. Treatment of grain distillation wastewater in an upflow anaerobic sludge bed (UASB) system. Water SA 27 (4), 433e444. Leal, L.H., Zeeman, G., Temmink, H., Buisman, C., 2007. Characterisation and biological treatment of greywater. Water Sci. Technol. 56 (5), 193e200. Lei, F., Xing, L.Z., Yang, H.C., Sheng, J., Huang, L., 2008. Research on the treatment for fishmeal wastewater by coagulation process. Indus. Saf. Environ. Prot. 34, 20e23. Lettinga, G., 1995. Anaerobic digestion and wastewater treatment systems. Antonie Leeuwenhoek 67, 3e28. Lettinga, G., Hulshoff Pol, L.W., 1991. UASB-process design for various types of wastewaters. Water Sci. Technol. 24 (8), 87e107.

4697

Lettinga, G., Field, J., van Lier, J., Zeeman, G., Hulshoff Pol, L.W., 1997. Advanced anaerobic wastewater treatment in the near future. Water Sci. Technol. 35 (10), 5e12. Lettinga, G., Hobma, S.W., Hulshoff Pol, L.W., de Zeeuw, W., de Jong, P., Grin, P., Roersma, R., 1983. Design operation and economy of anaerobic treatment. Water Sci. Technol. 15, 177e195. Lettinga, G., van Velseo, A.F.M., Hobma, S.W., de Zeeuw, W., 1980. Use of the upflow sludge blanket (USB) reactor concept for biological wastewater treatment, especially for anaerobic treatment. Biotechnol. Bioeng. 22, 699e734. Luostarinen, S.A., Rintala, J.A., 2005. Anaerobic on-site treatment of black water and dairy parlour wastewater in UASB-septic tanks at low temperatures. Water Res. 39 (2e3), 436e448. Ma, A.N., 1999. Innovations in management of palm oil mill effluent. The Planter 75 (881), 381e389. Kuala Lumpur. Ma, A.N., 2000. Environmental management for the palm oil industry. Palm Oil Develop., 1e10. Ma, A.N., Ong, A.S.H., 1985. Anaerobic digestion of palm oil mill. PORIM Bull. 4, 35e45. Malina, I.F., Pohland, F.G., 1992. Anonymous, Biogas technology in the Netherlands, anaerobic waste and wastewater treatment with energy production. In: Design of Anaerobic Processes for the Treatment of Industrial and Municipal Wastes. Technomic, Lancaster, PA, pp. 119e120. Manju, G.N., Raji, C., Anirudhan, T.S., 1998. Evaluation of coconut husk carbon for the removal of arsenic from water. Water Res. 32, 3062e3070. Marques, M.D., Cayless, S., Lester, J., 1990. Process aiders for startup of anaerobic fluidised bed systems. Environ. Technol. 11, 1093e1105. Martin, M.A., Raposo, F., Borja, R., Martin, A., 2002. Kinetic study of the anaerobic digestion of vinasse pretreated with ozone, ozone plus ultraviolet, and ozone plus ultraviolet light in the presence of titanium dioxide. Proc. Biochem. 37, 699e706. Masse´, D.I., Masse, L., 2001. The effect of temperature on slaughterhouse wastewater treatment in anaerobic sequencing batch reactors. Bioresour. Technol. 76, 91e98. Metcalf, Eddy, 2003. Wastewater Engineering, fourth ed. McGraw Hill Inc., New York, p. 10. Michael, N.N., Terry, W.S., Graig, L.B., 1988. Anaerobic contact pretreatment of slaughterhouse wastewater. Proc. Ind. Waste Conf., 1987, 42nd, p. 647. Miranda, L.A.S., Henriques1, J.A.P., Monteggia, L.O., 2005. A fullscale UASB reactor for Treatment of pig and cattle slaughterhouse wastewater with a High oil and grease content. Braz. J. Chem. Eng. 22 (4), 601e610. Moawad, A., Mahmoud, U.F., El-Khateeb, M.A., El-Molla, E., 2009. Coupling of sequencing batch reactor and UASB reactor for domestic wastewater treatment. Desalination 242, 325e335. Mockaitis, G., Ratusznei, S.M., Rodrigues, J.A.D., Zaiat, M., Foresti, E., 2005. Anaerobic whey treatment by a stirred sequencing batch reactor (ASBR): effects of organic loading rates and supplemented alkalinity. J. Environ. Manag., 1e9. Moosbrugger, R.E., Wentzel, M.C., Ekama, G.A., Marais, G.R., 1993. Treatment of wine distillery waste in UASB systems e feasibility, alkalinity requirements and pH control. Water Sci. Technol. 28 (2), 45e54. Musee, N., Lorenzen, L., Aldrich, C., 2006. Decision support for waste minimization in wine-making processes. Environ. Prog. 25 (1), 56e63. Nadais, H., 2002. Treatment of dairy wastewater in UASB reactors with intermittent operation (in Portuguese). Ph.D. thesis, University of Aveiro, Aveiro, Portugal. Nadais, H., Capela, I., Arroja, L., 2006. Intermittent vs continuous operation of upflow anaerobic sludge bed reactors for dairy wastewater and related microbial changes. Water Sci. Technol. 54 (2), 103e109.

4698

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 6 8 3 e4 6 9 9

Nadais, H., Capela, I., Arroja, L., Duarte, A., 2003. Biosorption of milk substrates onto anaerobic flocculent and granular sludge. Biotechnol. Prog. 19, 1053e1055. Nadais, H., Capela, I., Arroja, L., Duarte, A., 2005a. Treatment of dairy wastewater in UASB reactors inoculated with flocculent biomass. Water SA 31 (4), 603e608. Nadais, H., Capela, I., Arroja, L., Duarte, A., 2005b. Optimum cycle time for intermittent UASB reactors treating dairy wastewater. Water Res. 39, 1511e1518. Najafpour, G.D., Zinatizadeh, A.A.L., Mohamed, A.R., Hasnain Isa, M., Nasrollahzadeh, H., 2006. High-rate anaerobic digestion of palm oil mill effluent in an upflow anaerobic sludge-fixed film bioreactor. Proc. Biochem. 41, 370e379. Nataraj, S.K., Hosamani, K.M., Aminabhavi, T.M., 2006. Distillery wastewater treatment by the membrane based nanofiltration and reverse osmosis. Water Res. 40, 2349e2356. Nemerow, N.L., 1987. Dairy wastes. In: Industrial Water Pollution. R E Krieger, Malabar, FL, pp. 378e391. Nu´n˜ez, L.A., Martı´nez, B., 1999. Anaerobic treatment of slaughterhouse wastewater in an expanded granular sludge bed (EGSB) reactor. Water Sci. Technol. 40 (8), 99e106. Otterpohl, R., Albold, A., Oldenburg, M., 1999. Source control in urban sanitation and waste management: ten systems with reuse of resources. Water Sci. Technol. 39, 153e160. Ozturk, I., Eroglu, V., Ubay, G., Demir, I., 1993. Hybrid up-flow anaerobic sludge blanket reactor (HUASBR) treatment of dairy effluents. Water Sci. Technol. 28, 77e85. Palenzuela-Rollon, A., Zeeman, G., Lubberding, H.J., Lettinga, G., Alaerts, G.J., 2002. Treatment of fish processing wastewater in a one- or two-step upflow anaerobic sludge blanket (UASB) reactor. Water Sci. Technol. 45 (10), 207e212. Papachristou, E., Lafazanis, C.T., 1997. Application of membrane technology in the pre-treatment of cheese dairies wastes and co-treatment in a municipal conventional biological unit. Water Sci. Technol. 36, 361e367. Parawira, W., Murto, M., Zvauya, R., Mattiasson, B., 2006. Comparative performance of a UASB reactor and an anaerobic packed-bed reactor when treating potato waste leachate. Renew. Energy 31, 893e903. Patel, P., Madamwar, D., 1998. Surfactant in anaerobic digestion of salty cheese whey using upflow fixed film reactor for improved biomethanation. Proc. Biochem. 33 (2), 199e203. Patel, H., Madamwar, D., 2002. Effects of temperature and organic loading rates on biomethanation of acidic petrochemical wastewater using an anaerobic upflow fixed-film reactor. Bioresour. Technol. 82, 65e71. Petruy, R., 1999. Anaerobic treatment of protein, lipid and carbohydrate containing wastewaters using the EGSB technology. Ph.D. thesis, Agricultural University of Wageningen, Wageningen, The Netherlands. Poots, V.J.P., Mackay, G., Healy, J.J., 1978. Removal of basic dye from effluent using wood as an adsorbent. J. Water Pollut. Contr. Federation 50, 926e935. Prasertsan, P., Jung, S., Buckle, K.A., 1994. Anaerobic filter treatment of fishery wastewater. World J. Microbiol. Biotechnol. 10, 11e13. Punal, A., Lema, J.M., 1999. Anaerobic treatment of wastewater from a fish-canning factory in a full-scale upflow anaerobic sludge blanket (UASB) reactor. Water Sci. Technol. 40 (8), 57e62. Ramana, S., Biswas, A.K., Kundu, S., Saha, J.K., Yadava, R.B.R., 2002. Effect of distillery effluent on seed germination in some vegetable crops. Bioresour. Technol. 82, 273e275. Ramasamy, E.V., Gajalakshmi, S., Sanjeevi, R., Jithesh, M.N., Abbasi, S.A., 2004. Feasibility studies on the treatment of dairy wastewaters with up-flow anaerobic sludge blanket reactors. Bioresour. Technol. 93, 209e212.

Riffat, R., Dague, R., 1995. Laboratory studies on the anaerobic biosorption process. Water Environ. Res. 67 (7), 1104e1110. Rintala, J.A., 1997. Thermophilic anaerobic treatment of industrial process waters and wastewaters. Microbiology 66 (5), 699e704. Ruiz, I., Veiga, M.C., de Santiago, P., Blfizquez, R., 1997. Treatment of slaughterhouse wastewater in a UASB reactor and an anaerobic filter. Bioresour. Technol. 60, 251e258. Sa´nchez, E., Borja, R., Travieso, L., Martı´n, A., Colmenarejo, M.F., 2005. Effect of organic loading rate on the stability, operational parameters and performance of a secondary upflow anaerobic sludge bed reactor treating piggery waste. Bioresour. Technol. 96, 335e344. Sa´nchez, E., Borja, R., Weiland, P., Travieso, L., 2001. Effect of substrate concentration and temperature on the anaerobic digestion of piggery waste in tropical climates. Process Biochem. 37, 483e489. Sa´nchez, E., Monroy, O., Ca˜nizares, R.O., Travieso, L., 1995. Comparative study of piggery waste treatment by upflow sludge beds anaerobic reactors and packed bed reactors. J. Agric. Eng. Res. 62, 71e76. Sandberg, M., Ahring, B.K., 1992. Anaerobic treatment of fish meal process waste-water in a UASB reactor at high pH. Appl. Microbiol. Biotechnol. 36, 800e804. Sari, A.L., Jukka, A.R., 2005. Anaerobic on-site treatment of black water and dairy parlour wastewater in UASB-septic tanks at low temperatures. Water Res. 39, 436e448. Sawajneh, Z., Al-Omari, A., Halalsheh, M., 2010. Anaerobic treatment of strong sewage by a two stage system of AF and UASB reactors. Water Sci. Technol. 61 (9), 2399e2406. Sayed, S., 1987. Anaerobic treatment of slaughterhouse wastewater using the UASB process. Ph.D thesis, Agricultural University of Wageningen, Wageningen, The Netherlands. Sayed, S., de Zeeuw, W., 1988. The performance of a continuously operated flocculated sludge UASB reactor with slaughterhouse wastewater. Biol. Wastes 24, 213e226. Sayed, S., de Zeeuw, W., Lettinga, G., 1984. Anaerobic treatment of slaughterhouse waste using a flocculant sludge UASB reactor. Agricul. Wastes 11, 197e226. Sayed, S., van Campel, Lettinga, L., 1987. Anaerobic treatment of slaughterhouse waste water using a granular sludge UASB reactor. Biol. Wastes 21, 213e226. Schmidt, J.E., Ahring, B.K., 1996. Granular sludge formation in upflow anaerobic sludge blanket (UASB) reactors. Biotechnol. Bioeng. 49 (3), 229e246. Seghezzo, L., Zeeman, G., Van Lier, J.B., Hamelers, H.V.M., Lettinga, G., 1998. A review: the anaerobic treatment of sewage in UASB and SGSB reactors. Bioresour. Technol. 65, 175e190. Seif, H., Moursy, A., 2001. Treatment of slaughterhouse wastes. In: Sixth International Water Technology Conference, IWTC, Alexandria, Egypt. Siang, L.C., 2006. Biodegradation of oil and grease in upflow anaerobic sludge blanket reactor for palm oil mill effluent treatment. Masters degree thesis. Universiti Teknologi Malaysia, Malaysia. Siegrist, H., Vogt, D., Garcia-Heras, J.L., Gujer, W., 2002. Mathematical model for meso- and thermophilic anaerobic sewage sludge digestion. Environ. Sci. Technol. 36, 1113e1123. Singh, K.S., Viraraghavan, T., 1998. Start-up and operation of UASB reactors at 20 C for municipal wastewater treatment. J. Ferment. Bioeng. 85 (6), 609e614. Singh, K.S., Viraraghavan, T., 2000. Performance of UASB reactor at 6 to 32  C in municipal wastewater treatment. Water Quality Res. J. Can. 35 (1), 113e124. Song, Y.C., Kwon, S.J., Woo, J.H., 2004. Mesophilic and thermophilic temperature co-phase anaerobic digestion compared with single-stage mesophilic- and thermophilic digestion of sewage sludge. Water Res. 38, 1653e1662.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 6 8 3 e4 6 9 9

Soto, M., Ligero, P., Vega, A., Rui, I., Veiga, M.C., Blazquez, R., 1997. Sludge granulation in UASB digesters treating low strength wastewaters at mesophilic and psychrophilic temperatures. Environ. Technol. 18 (11), 1133e1141. Speece, R.E., Kem, J.A., 1970. The effect of short-term temperature variations on methane production. J. Water Pollut. Contr. Federation 42, 1990e1997. Stafford, D.A., 1982. The effects of mixing and volatile fatty acid concentrations on anaerobic digester performance. Biomass 2, 43e55. Stevens, M.A., Schulte, D.D., 1979. Low temperature digestion of swine manure. J. Environ. Eng. Div. ASCE 105 (EE1), 33e42. Stronach, S.M., Rudd, T., Lester, J.N., 1987. Start-up of anaerobic bioreactors on high strength industrial wastes. Biomass 13, 173e197. Syutsubo, K., Harada, H., Ohashi, A., 1998. Granulation and sludge retainment during start-up of a thermophilic-UASB reactor. Water Sci Technol. 38 (8e9), 349e357. Syutsubo, K., Harada, H., Ohashi, A., Suzuki, H., 1997. Effective start-up of thermophilic UASB reactor by seeding mesophilically-grown granular sludge. Water Sci. Technol. 36 (67), 391e398. Tandukar, M., Uemura, S., Machdar, I., Ohashi, A., Harada, H., 2005. A low-cost municipal sewage treatment system with a combination of UASB and the “fourth-generation” downflow hanging sponge reactors. Water Sci. Technol. 52 (1e2), 323e329. Tawfik, A., Sobheyb, M., Badawya, M., 2008. Treatment of a combined dairy and domestic wastewater in an up-flow anaerobic sludge blanket (UASB) reactor followed by activated sludge (AS system). Desalination 227, 167e177. Torkian, A., Eqbali, A., Hashemian, S.J., 2003. The effect of organic loading rate on the performance of UASB reactor treating slaughterhouse effluent. Res. Conserv. Recyc. 40, 1e11. Trnovec, W., Britz, T.J., 1998. Influence of organic loading rate and hydraulic retention time on the efficiency of a UASB reactor treating a canning factory effluent. Water SA 24 (2), 147e152. Uemura, S., Harada, H., 2000. Treatment of sewage by a UASB reactor under moderate to low temperature conditions. Bioresour. Technol. 72, 275e282. Uzal, N., Gokacay, C.F., Demirer, G.N., 2003. Sequential anaerobic/ aerobic biological treatment of malt whisky wastewater. Proc. Biochem. 39, 279e286. van Haandel, A.C., Lettinga, G., 1994. Anaerobic Sewage Treatment; A Practical Guide for Regions with a Hot Climate. John Wiley and Sons, Chichester, England. van Haandel, A., Kato, M.T., Cavalcanti, P.F.F., Florencio, L., 2006. Anaerobic reactor design concepts for the treatment of domestic wastewater. Rev. Environ. Sci. Biotechnol. 5 (1), 21e38. van Lier, J.B., Huibers, F.P., 2004. Agricultural use of treated wastewater: the need for a paradigm shift in sanitation and

4699

treatment. Risk Assessment of Re-use on Groundwater Quality. In: Proceedings of Symposium HS04 Held During IUGG2003 at Sapporo, Japan, July 2003. IAHS Publ. 285. van Lier, J.B., Grolle, K.C.F., Stams, A.J.M., de Macario, E.C., Lettinga, G., 1992. Startup of a thermophilic upflow anaerobic sludge bed (UASB) reactor with mesophilic granular sludge. Appl. Microbiol. Biotechnol. 37, 130e135. Vidal, G., Aspe´, E., Martı´, M.C., Roeckel, M., 1997. Treatment of recycled wastewaters from fishmeal factory by an anaerobic filter. Biotechnol. Lett. 19, 117e121. Vidal, G., Carvalho, A., Mendez, R., Lema, J.M., 2000. Influence of the content in fats and proteins on the anaerobic biodegradability of dairy wastewaters. Bioresour. Technol. 74, 231e239. Visser, A., Gao, Y., Lettinga, G., 1993. Effects of short-term temperature increases on the mesophilic anaerobic breakdown of sulfate containing synthetic wastewater. Water Res. 27, 541e550. Vlissidis, A., Zouboulis, A.I., 1993. Thermophilic anaerobic digestion of alcohol distillery wastewaters. Bioresour. Technol. 43, 131e140. von Sperling, M., Freire, V.H., de Lemos, C.A., Chernicharo, 2001. Performance evaluation of a UASB e activated sludge system treating municipal wastewater. Water Sci. Technol. 43 (11), 323e328. Wiegant, W.M., Lettinga, G., 1985. Thermophilic anaerobic digestion of sugars in upflow anaerobic sludge blanket reactors. Biotechnol. Bioeng. 27, 1603e1607. Wiegant, W.M., Classen, J.A., Lettinga, G., 1985. Thermophilic anaerobic digestion of high strength wastewaters. Biotechnol. Bioeng. 27, 1374e1381. Wolmarans, B., de Villiers, G.H., 2002. Start-up of a UASB effluent treatment plant on distillery wastewater. Water SA 28 (1), 63e68. Yang, J., Anderson, G., 1993. Effects of wastewater composition on stability of UASB. J. Environ. Eng. 119 (5), 958e977. Yeoh, B.G., 2004. A technical and economic analysis of heat and power generation from biomethanation of palm oil mill effluent. Electric Supply Industry in Transition: Issues and Prospect for Asia, 20e60. Yilmaz, T., Yuceer, A., Basibuyuk, M., 2008. A comparison of the performance of mesophilic and thermophilic anaerobic filters treating papermill wastewater. Bioresour. Technol. 99, 156e163. Yu, H.Q., Fang, H.H.P., Gu, G.W., 2002. Comparative performance of mesophilic and thermophilic acidogenic upflow reactors. Proc. Biochem. 38, 447e454. Zinatizadeh, A.A.L., Mohamed, A.R., Abdullah, A.Z., Mashitah, M. D., Hasnain Isa, M., Najafpour, G.D., 2006. Process modeling and analysis of palm oil mill effluent treatment in an up-flow anaerobic sludge fixed film bioreactor using response surface methodology (RSM). Water Res. 40, 3193e3208.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 0 0 e4 7 1 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Including greenhouse gas emissions during benchmarking of wastewater treatment plant control strategies Xavier Flores-Alsina, Lluı´s Corominas, Laura Snip, Peter A. Vanrolleghem* modelEAU, De´partement de ge´nie civil et ge´nie des eaux, Pavillon Adrien-Pouliot, Universite´ Laval, 1065, Avenue de la Me´decine, Que´bec, G1V 0A6 QC, Canada

article info

abstract

Article history:

The main objective of this paper is to demonstrate how greenhouse gas (GHG) emissions

Received 18 October 2010

can be quantified during the evaluation of control strategies in wastewater treatment

Received in revised form

plants (WWTP). A modified version of the IWA Benchmark Simulation Model No 2 (BSM2G)

15 April 2011

is hereby used as a simulation case study. Thus, the traditional effluent quality index (EQI),

Accepted 21 April 2011

operational cost index (OCI) and time in violation (TIV) used to evaluate control strategies

Available online 29 April 2011

in WWTP are complemented with a new dimension dealing with GHG emissions. The proposed approach is based on a set of comprehensive models that estimate all potential

Keywords:

on-site and off-site sources of GHG emissions. The case study investigates the overall

Activated sludge modeling

performance of several control strategies and demonstrates that substantial reductions in

Benchmarking

effluent pollution, operating costs and GHG emissions can be achieved when automatic

Global warming

control is implemented. Furthermore, the study is complemented with a scenario analysis

Model-based evaluation

that examines the role of i) the dissolved oxygen (DO) set-point, ii) the sludge retention

Multi-criteria decision making

time (SRT) and iii) the organic carbon/nitrogen ratio (COD/N) as promoters of GHG emis-

Process control

sions. The results of this study show the potential mechanisms that promote the formation

Sustainability

of CO2, CH4 and N2O when different operational strategies are implemented, the existing synergies and trade-offs amongst the EQI, the OCI and TIV criteria and finally the need to reach a compromise solution to achieve an optimal plant performance. ª 2011 Published by Elsevier Ltd.

1.

Introduction

The increasing demands on effluent quality at lower operational costs have promoted the development of new technologies and the implementation of control concepts to improve the overall performance of wastewater treatment plants (WWTP). Full-scale applications have shown the feasibility of automatic control in aeration systems, chemical dosage and recycle flows (Oennerth et al., 1996; Ingildsen et al., 2002; Devisscher et al., 2002; Olsson et al., 2005). Dynamic simulation studies have also been used to compare the performance of different control strategies (Zhao et al., 1995; Spanjers et al., 1998; Corominas et al., 2006; Stare et al.,

2007; Flores-Alsina et al., 2009; Machado et al., 2009) or to evaluate them before full-scale implementation (Ayesa et al., 2006). Plant-wide operation has also been introduced to take into account the interactions between the processes (Gujer and Erni, 1978; Lessard and Beck, 1993; Jeppsson et al., 2007). However, the increasing interest for greenhouse gas (GHG) emissions from wastewater treatment leads to re-think the traditional engineering approaches by adding this new dimension. Therefore, new tools are needed to estimate the GHG emissions and evaluate different operation schemes that prevent or minimize their generation in WWTP. During the last years, the scientific community has developed some mathematical tools to estimate/evaluate the

* Corresponding author. Tel.: þ1 418 656 5085; fax: þ1 418 656 2928. E-mail address: [email protected] (P.A. Vanrolleghem). 0043-1354/$ e see front matter ª 2011 Published by Elsevier Ltd. doi:10.1016/j.watres.2011.04.040

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 0 0 e4 7 1 0

Nomenclature A ADM AE AER AOB ANOX ASM BOD BSM2 CH4 CO2 CO2e COD DO EC EQI GHG HEnet KFNA KLa

alternative anaerobic digestion model aeration energy (kWh$day1) aerobic section ammonium oxidizing bacteria anoxic section activated sludge model biochemical oxygen demand (g m3) Benchmark Simulation Model No 2 methane (kg CH4$day1) carbon dioxide (kg CO2$day1) equivalent carbon dioxide (kg CO2e$day1) chemical oxygen demand (g m3) dissolved oxygen concentration (g m3) consumption of external carbon source (kgCOD$day1) effluent quality index (kg pollution$day1) greenhouse gas net heating energy (kWh$day1) inhibition constant for free nitrous acid (g N m3) volumetric oxygen transfer coefficient (day1)

generation of GHG in WWTP. However, these methods appear to be unsuitable to evaluate WWTP control strategies for several reasons. First of all, the current approaches are based on steady state calculations, i.e. empirical approaches (e.g. IPCC, 2006; LGO, 2008; NGER, 2008) or comprehensive models (Bridle et al., 2008; Bani Shahabadi et al., 2009; Pagilla et al., 2009), without taking into account the wastewater treatment dynamics. Thus, it is not possible to consider how changes in the influent load (daily, weekly, seasonal), temperature (winter/summer) and operating conditions (DO, SRT, COD/N ratios,) influence the production/emission of GHG. Second of all, some of these estimations are focused on particular wastewater sections/ compounds/technologies and do not consider the whole. For example, Cakir and Stenstrom (2005), Keller and Hartley (2003) and Monteith et al. (2005) study the contribution of aerobic degradation of carbonaceous biochemical oxygen demand (CBOD) to GHG emissions. Other investigations such as those by von Schulthess and Gujer (1996); Hiatt and Grady (2008a,b), and Foley et al. (2010) quantify the N2O emissions in aerobic-anoxic activated sludge plants. Batstone et al. (2002) and Greenfield and Batstone (2003) evaluated methane (CH4) and carbon dioxide (CO2) emissions under anaerobic conditions. Nevertheless, with the aforementioned approaches it is difficult to have the overall carbon footprint in terms of CH4, CO2 and N2O when the plant is running under a certain operational mode. Third of all, none of these approaches include multi-criteria evaluation combining GHG emissions information with rigorous quantification about the effluent quality and operational costs, enabling to quantify and compare their overall sustainability. In order to overcome these limitations and to include the GHG emissions during the evaluation of WWTP control strategies, the authors suggest combining pseudo-empirical equations and mechanistic models. This approach explores influent, effluent and operational variables at each simulation

ME MP N NHþ 4 NO N2O NOB NO 2 NO 3 PE PI Qe Qcarb Qintr Qr QW SP SRT TIV TKN TN TSS WWTP

4701

mixing energy (kWh$day1) methane production (kgCH4$day1) nitrogen ammonium (g N m3) nitric oxide (g N m3) nitrous oxide (g N m3) nitrite oxidizing bacteria nitrite (g N m3) nitrate (g N m3) pumping energy (kWh$day1) proportional integral controller effluent flow (m3$day1) external carbon source flow rate (m3$day1) internal recycle flow rate (m3$day1) external recirculation flow rate (m3$day1) waste flow rate (m3$day1) sludge production (kgTSS$day1) sludge retention time (days) time in violation (%) total Kjeldahl nitrogen (g m3) total nitrogen (g m3) total suspended solids (g m3) wastewater treatment plant

step. Then operational procedures or control strategies that may cause favorable conditions for GHG emissions can be quantified, providing an additional dimension to the traditional effluent quality, economical and legal criteria. The main objective of this paper is therefore to demonstrate how GHG can be quantified during the evaluation of control strategies. This paper details the rationale of how the different sources of CO2, CH4 and N2O can be taken into account dynamically within the evaluation procedure. The performance of this approach is then evaluated alongside a number of simulated scenarios where a modified version of the IWA Benchmark Simulation Model (BSM2G) is studied under open-loop and closed-loop regime.

2.

Methods

2.1. Wastewater treatment plant under study and evaluation criteria The WWTP under study has the same layout as the IWA BSM2 proposed by Nopens et al. (2010) (see a schematic representation in Fig. 1). The activated sludge unit is a modified Ludzack-Ettinger configuration consisting of five tanks in series. Tanks 1 (ANOX1) and 2 (ANOX2) are anoxic with a total volume of 3000 m3, while tanks 3 (AER1), 4 (AER2) and 5 (AER3) are aerobic with a total volume of 9000 m3. The circular secondary settler (SEC2) has a surface area of 1500 m2 with a total volume of 6000 m3. The BSM2 plant also contains a primary clarifier (PRIM), a sludge thickener (THK), an anaerobic digester (AD), a storage tank (ST) and a dewatering unit (DW). The primary clarifier is modeled in accordance with Otterpohl and Freund (1992) and Otterpohl et al. (1994). A modified version of the ASM1 (Henze et al., 2000), based on the

4702

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 0 0 e4 7 1 0

ASMN suggested by (Hiatt and Grady, 2008a) is chosen as the biological process model. This model incorporates two nitrifying populations e ammonia oxidizing bacteria (AOB) and nitrite oxidizing bacteria (NOB) e using free ammonia and free nitrous acid, respectively as their substrates. The model incorporates also four step denitrification, (sequential reduction of nitrate to nitrogen gas via nitrite, nitric oxide, and nitrous oxide), using individual reaction specific parameters. The parameter values suggested in Hiatt and Grady (2008a) were used, except for the KFNA (inhibition constant for free nitrous acid) that was reduced from 1 104 (used for high nitrogen loads) to 1 106 g m3 (used for low nitrogen loads) to promote NOB growth (Snip, 2010; Corominas et al., 2010). To account for seasonal variability, liquid-gas saturation constants, kinetic parameters, transfer coefficients and equilibrium reactions are temperature dependent. Stripping equations for the gases were implemented as in Foley et al. (2010). The double exponential settling velocity function of Taka´cs et al. (1991) is used to model the secondary settling process through a one-dimensional model consisting of ten layers. Regarding the thickener and dewatering units, these are modeled as ideal, continuous processes with no biological activity, and a constant percentage of TSS in the concentrated sludge flows leaving the thickening and dewatering units. The widely recognized Anaerobic Digestion Model No. 1 (ADM1) (Batstone et al., 2002) is the dynamic anaerobic digestion model implemented. The interfaces presented in Nopens et al. (2009) have been modified (see Corominas et al., 2010) to link the biological model and ADM1, by considering COD and N balances for all oxidized nitrogen compounds. Finally, the influent wastewater composition follows the principles outlined in Gernaey et al. (2006). Further information about the BSM2 layout and the description of the process models can be found in Jeppsson et al. (2007). A set of evaluation criteria is used to compare the simulation results in the BSM2. The overall pollution removal of the plant is obtained by calculating the effluent quality index (EQI), which is expressed in units of kg pollution$day1. Compared to Nopens

et al., (2010), the EQI was modified to include the different  oxidized nitrogen forms (NOx ¼ NO 3 þ NO2 þ NO þ N2O) on the receiving water (see Eq. 1): EQI ¼

1 ðt2  t1 Þ$1000

t2 ¼609 Z days

ðBSS $TSSe ðtÞ þ BCOD $CODe ðtÞ t1 ¼245 days

þ BTKN $TKNe ðtÞ þ BNOX $NOX;e ðtÞ þ BBOD5 $BODe ðtÞÞQe ðtÞ$dt

where TSSe, CODe, TKNe, SNOx,e and BODe represent, respectively, the total suspended solids, the chemical oxygen demand, the total Kjeldahl nitrogen, the oxidized nitrogen concentration and the biochemical oxygen demand in the effluent. Qe is the effluent flow rate and t time. The weights for the different pollutants are: BSS ¼ 2, BCOD ¼ 1, BTKN ¼ 30, BNOx ¼ 10 and BBOD5 ¼ 2. The operational cost index (OCI) is calculated as the weighted sum of the aeration energy (AE), the pumping energy (PE), the consumption of external carbon source (EC), the sludge production (SP), mixing energy (ME), heating energy (HE) and methane production (MP) as shown in Eq. 2 (Nopens et al., 2010).   OCI ¼ AE þ PE þ 3$EC þ 3SP þ ME  6MP þ max 0; HEnet

2.2.

Estimation of the greenhouse gas (ghg) emissions

The comprehensive approach suggested by Monteith et al. (2005) and extended in Bridle et al. (2008) is used to estimate all potential GHG emissions from the studied WWTP that cannot be obtained from the explicit results of the modified BSM2. A comprehensive description of the methodology can be found in Snip (2010) and Corominas et al. (2010).

INFLUENT

EFFLUENT WASTEWATER PRIM

SEC2

ANOX2

AER1

(2)

The final part of the evaluation procedure involves the calculation of the percentage of time (%) that the plant is in 3 ), TN violation for five different pollutants: NHþ 4 (4 g N m 3 3 (18 g N m ), COD (100 g COD m ), BOD (10 g COD m3) and TSS (30 g TSS m3). More details about the “time plant in violation” (TIV) criterion can be found in Copp (2002).

WASTEWATER

ANOX1

(1)

AER2

AER3

THK DW

AD SLUDGE FOR ST

Fig. 1 e Flow scheme of the treatment plant under study.

DISPOSAL

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 0 0 e4 7 1 0

The overall GHG evaluation comprises the estimation of the following GHG emissions: i) direct secondary treatment, ii) sludge processing, iii) net power, iv) embedded GHG emissions from chemical use and finally v) sludge disposal and reuse. In order to deal with the different nature of the generated GHG (CO2, CH4 and N2O), they are converted in units of CO2 equivalent (CO2e). It is important to highlight that the presented procedure only takes into account GHG emissions produced within the WWTP limits. - Direct secondary treatment emissions. The emission from the activated sludge section includes the CO2 generated from biomass respiration and BOD oxidation, the N2O generated from nitrogen removal and the CO2 credit from nitrification. The first two processes are estimated following the methodology proposed by Monteith et al. (2005). N2O emissions are given by the modified ASMN model. It is important to highlight that N2O production is only considered during heterotrophic denitrification. Other important pathways like N2O production during nitrification are not considered in this study (see discussion). Finally, the credit from nitrification is calculated with the factor 0.31 kg of CO2consumed (kg N nitrified)1 (Tchobanoglous et al., 2003). - Sludge processing. The emissions of GHG during sludge treatment are mainly generated in the anaerobic digester. Direct biogas CO2 and CH4 emissions are quantified using ADM1. In this case it is assumed that the biogas is fed directly into a gas-fired combustion turbine converting the CH4 into CO2 and generating electricity and heat (in turn used to heat the anaerobic digester). The CO2 generated during anaerobic digestion and the CO2 produced in the combustion are released to the atmosphere. - Net power GHG. The total energy consumption is quantified using the OCI defined in Eq 2. The credit refers to the electricity generated by the turbine and it is calculated by using a factor for the energy content of the methane gas (50,014 MJ (kg CH4)1) and assuming a 43% efficiency for electricity generation, which is a reasonable value according to Saravanamuttoo et al. (2009). The net power is the difference between the total energy consumption and the credit. - Chemicals. The embedded GHG emissions associated with chemicals used at the WWTP have been limited to the external carbon source. These emissions are estimated by using the emission factor of 1.54 g CO2e$g methanol1 (Dong and Steinberg, 1997). - Sludge disposal and reuse. CO2 emissions associated with trucking of bio-solids are quantified by multiplying the truck movements by the distance to the reuse (150 km to agriculture, 20 km to compost and 144 km to forestry). The CO2 emissions by mineralization are calculated based on the sludge mass times the carbon concentration times the factor of CO2 to carbon. It is assumed that 38% of sludge goes to agriculture, 45% to a compost site and 17% to forestry (Bridle et al., 2008).

2.3.

Implemented control strategies

A default open-loop control strategy (A0), referred to as BSM2G open-loop control, is the reference case and has the following characteristics: Qintr ¼ 61,944 m3$day1 (internal recycle flow

4703

rate), QW ¼ 400 m3$day1 (waste flow rate), Qr ¼ 20,648 m3$day1 (external recirculation flow rate), Qcarb ¼ 5 m3$day1 (external carbon source addition rate) and kLa1 ¼ kLa2 ¼ 2 day1 and kLa3 ¼ kLa4 ¼ kLa5 ¼ 140 day1 (aeration intensity, represented as the volumetric oxygen transfer coefficient) respectively. A low KLa is expected in the anoxic zone (and consequently some stripping) as a side-effect of mixing (von Schulthess and Gujer, 1996). Next, three different control strategies (A1, A2 and A3) are implemented and compared to the base case. The first strategy (A1) is based on a simple PI loop controlling the dissolved oxygen (DO) in the 2nd aerobic tank (AER2) through manipulation of the aeration flow in AER1,2 and 3 (KLa3, 4 and 5) (setpoint ¼ 2 g O2 m3). KLa5 is set to half the value of KLa3 and KLa4. In A2, a controller of the nitrate (NO 3 ) in the 2nd anoxic tank (ANOX2) manipulating the internal recycle flow rate (Qintr) is added to A1 (set-point ¼ 1 g N m3). Finally, A3 completes A2 with a cascade PI ammonium (NH4þ) controller that manipulates the DO set-point (set-point ¼ 2 g N m3). Additionally, in A1, A2 and A3 two waste flow rates in SEC2 are imposed depending of the time of the year in order to keep the biomass in the system during the winter period, i.e. starting/end date, (QW ¼ 300 m3 day1). For the rest of the simulation time the waste flow is set to 450 m3 day1. The DO sensor is assumed to be close to ideal with a response time of 1 min in order to prevent unrealistic control þ applications. The NO 3 and the NH4 sensors have a time delay of 10 min, with zero mean white noise (standard deviation of 0.5 gN m3) (Rieger et al., 2003). Finally, the aeration system (KLa), is defined with significant dynamics assuming a response time of 4 min. As discussed in Corominas et al. (2010) all dynamic simulations (609 days) are preceded by a steady state simulation (200 days) but only the data generated during the final 364 days (t) of dynamic simulation are used for plant performance evaluation.

3.

Case study control evaluation

The results obtained for the open (A0) and closed-loop (A1, A2 and A3) simulations have been evaluated with respect to effluent quality, operating costs, legal criteria and GHGs production (see Table 1 and Table 2). Compared to alternative A0, the DO controller implemented in alternatives A1, A2 and A3 substantially reduces the aeration energy (AE) and the OCI by up to 6%. In the same order of magnitude (up to 7%) EQI is also improved in all closed-loop strategies. This is mainly due to a more efficient use of the aeration system adapting the airflow rate to the oxygen demand for organic matter and nitrogen removal. The implementation of a NO 3 controller in alternatives A2 and A3 slightly improved denitrification efficiency and reduced both the effluent total nitrogen concentration (TN) and the percentage of violation in terms of TN (TIV_TN). Since the internal recycle (see the increase of PE in Table 1) between AER3 and ANOX1 is strongly influenced by the influent load and process performance, the NO 3 controller ensures an optimal use of the incoming organic matter for denitrification. Finally, the effect of the NHþ 4 controller implemented in A3 must be mentioned. The variation of the DO set-point in AER2 according to the nitrification needs substantially reduces the ammonium peaks (see TIV_NHþ 4 in

4704

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 0 0 e4 7 1 0

Table 1 e Effluent quality, operational cost and legal criteria for the different control strategies.

Total Kjeldahl Nitrogen (TKN) Total Nitrogen (TN) Chemical oxygen demand (COD) Biochemical oxygen demand (BOD5) Total suspended solids (TSS) Effluent quality index (EQI) Sludge production (SP) Aeration energy (AE) Pumping energy (PE) External Carbon addition (EC) Mixing energy (ME) Net Heating energy (HEnet) Methane production (MP) Operational cost index (OCI) Time Time Time Time Time

in violation in violation in violation in violation in violation

for for for for for

TN (TIV_TN) COD (TIV_COD) ammonium (TIV_NHþ 4) TSS (TIV_TSS) BOD5 (TIV_BOD5)

A0

A1

A2

A3

Units

3.83 15.04 49.19 3.14 17.51 6448

3.97 13.62 41.73 3.16 15.31 6239

4.05 13.12 49.71 3.15 15.31 6172

3.99 12.39 49.76 3.17 15.31 5995

g N m3 g N m3 g COD m3 g COD m3 g TSS m3 kg poll$dayL1

2703 5627 447 2000 768 4289 1141 14,107

2674 4843 442 2000 768 4247 1126 13,324

2673 4821 467 2000 768 4248 1125 13,323

2674 5048 496 2000 771 4247 1126 13,580

kg TSS$day1 kWh$day1 kWh$day1 kg COD$day1 kWh$day1 kWh$day1 kg CH4$day1 e

5.72 0.06 18.94 0.27 0.23

2.17 0.07 19.44 0.30 0.23

1.09 0.06 20.83 0.30 0.23

1.35 0.06 5.40 0.30 0.23

Table 1) at the expense of higher aeration energy (AE). No substantial differences are observed in the rest of effluent (COD, BOD5 and TSS) and economic (SP, EC, HEnet and MP) criteria. The plant under control can reduce the GHG emissions with up to 9.6% (from 1.142 to 1.032 kg CO2e m3 treated wastewater). The main differences are found in the direct secondary emissions and in the power consumption (see Table 2). A significant reduction of emitted N2O is observed in the control strategies with a DO controller (from 16.35 kg N2O$day1 for A0 to 10.09, 10.92 and 13.96 kg N2O$day1 for A1, A2 and A3) since the DO controller prevents the system from nitrite accumulation (see a detailed discussion in the following section). In addition, it is important to notice the differences between alternatives A2 (NO 3 controller) and A3 (NHþ 4 controller) in terms of N2O emissions. The performance of the NO 3 controller is heavily affected by temperature.

% % % % %

During winter time, when nitrification is lower, it is necessary to increase the internal recycle in order to maintain the desired set-point in the anoxic section because there is a lower production of nitrates in the aerobic zone. As a consequence, it also increases the quantity of oxygen transported from AER3 to ANOX1 worsening the overall denitrification efficiency. As N2O is an intermediate in the denitrification process, incomplete denitrification leads to N2O emissions. In the case of the cascade NHþ 4 controller (A3) the increase of N2O is due to a couple of reasons. Firstly, the sudden increase of aeration (KLa) in the aerated zones to smoothen the ammonium peaks, increases the quantity of DO returning into the anoxic zone. As in the previous case (NO 3 controller), this situation leads to problems in the anoxic zone (incomplete denitrification). Secondly, long periods with low aeration intensity (see the concomitant increase of the mixing and pumping energy in Table 1) lead to increased nitrite accumulation due to DO

Table 2 e GHG emissions for the different control strategies. A0

A1

A2

A3

Units

0.180 0.212 0.012 0.236 0.616

0.191 0.219 0.012 0.146 0.544

0.191 0.219 0.012 0.158 0.555

0.191 0.219 0.012 0.202 0.599

kg CO2e m3 kg CO2e m3 kg CO2e m3 kg CO2e m3 kg CO2e mL3

0.079 0.152 0.231

0.078 0.150 0.228

0.078 0.150 0.228

0.078 0.150 0.228

kg CO2e m3 kg CO2e m3 kg CO2e mL3

0.311 0.310 0.001

0.276 0.306 L0.030

0.276 0.306 L0.030

0.287 0.306 L0.019

kg CO2e m3 kg CO2e m3 kg CO2e mL3

Total Embedded GHG emissions from Chemicals use Total Sludge disposal and reuse GHG emissions

0.099 0.194

0.099 0.191

0.099 0.191

0.099 0.191

kg CO2e m3 kg CO2e m3

Total GHG emissions (GHG total)

1.142

1.032

1.044

1.100

kg CO2e mL3

Biomass respiration BOD oxidation Credit nitrification N2O emissions Total direct secondary treatment emissions CO2 emissions from digestion CH4 emissions from digestion Total Sludge processing GHG emssions Power Credit Power Total Net power GHG emissions

4705

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 0 0 e4 7 1 0

4.1. Scenario 1: effect of the DO concentration in the aerated section

limitation of NOB activity. Again, an increased nitrite concentration favors N2O production. A more efficient aeration system in A1, A2 and A3 reduces the off-site CO2 emissions due to lower power consumption (as already mentioned for the operational cost index). Note that the implementation of these three basic controllers does not suppose any change in the addition of external carbon source. Surprisingly, the different waste flow patterns in the closed-loop strategies do not have a big impact on the CO2 derived from sludge treatment and reuse. For this reason the GHG emissions due to sludge processing, sludge disposal and reuse and the embedded emissions from chemicals use remain almost at the same value. Finally, it has to be mentioned that the GHG estimations obtained in this study (0.87e1.00 kg CO2e$m3) are within the range of values presented in Bridle et al. (2008) (0.9, 1.6 and 2.2 kg CO2e$m3) and in Pagilla et al. (2009) (from 0.34 to 1.25 kg m3).

Scenario analysis

Bio-treatment

Net Power

Sludge processing

Chemical use

Sludge disposal

0,6 0,5 0,4 0,3 0,2 0,1 0,0

10

0

Q

ca rb

=

= ca rb

40 0/ 5 = w

= w Q

Total=1.142

Total=0.865

Q

50 20 0/ 3

=

=

Scenario 1

Total=0.992 50

Total=1.092

3

1

Total=1.097

DO

De fa ul

Total=1.022

DO

Total=1.044 A2 )

-0,1

t(

GHG emissions (kg CO2e·m-3 treated wastewater)

A scenario analysis, which focuses on analyzing GHG emissions as well as the previously defined environmental, economical and legal criteria, is also included to investigate which variables are worth looking at for control. Therefore, for exemplary purposes, we analyze how the results of alternative A2 are affected by changing some of its settings in three scenarios. Scenario 1 evaluates the plant performance at the DO set-points of 1 and 3 g O2$m3. Scenario 2 changes the sludge retention time by either increasing (QW ¼ 400/ 550 m3$day1) or decreasing (QW ¼ 200/350 m3$day1) the winter/summer waste flow patterns. Finally, in Scenario 3 the COD/N ratio is changed by modifying the dosage of external carbon source (Qcarb) from 5 m3$day1 to 0 and 10 m3$day1. First, the effect on the GHG emissions of the three scenarios is studied in subsection one, two and three. In the fourth subsection, the conflicting results for the other criteria are discussed.

Q

4.

The airflow rate in the aerobic section (AER) is essential in activated sludge treatment because it promotes the growth of the heterotrophic and autotrophic bacateria that will oxidize organic matter and nitrogen. Hence, Scenario 1 examines the effect of the overall GHG emissions when the dissolved oxygen set-point is above and below the default value of 2 g O2 m3. In order to do such analysis, the plant is simulated under different airflow rates and the emitted GHG quantified for each simulation. The results of the simulations show that low DO concentrations (Fig. 2, DO ¼ 1 g O2 m3) lead to a reduction of the CO2 production thanks to the lower energy consumption but a very slight increase of the direct secondary treatment emissions compared to the default A2. The responsible of the increase are the higher N2O emissions caused by the accumulation of NO 2 (Fig. 3b) due to incomplete nitrification (Fig. 3a). Lower DO concentrations in the nitrification reactor cause growth limitation of AOB and especially NOB. Thus, the resulting high NO 2 concentrations in the anoxic reactor transported by the internal recycle lead to lower denitrification rates and accumulation of NO and N2O. Fig. 3 presents one year of data starting July 1st (day 245) and with the summer holiday period (where the load from industries is significantly reduced) between days 270 and 300. The yearly dynamics also show an increase in the ammonia concentration in the winter period with a consequent increase of the nitrite accumulation. At a DO set-point of 3 g O2 m3 (Fig. 2) the production of CO2 is increased due to higher energy consumption (see net power values). In addition, more N2O is released due to incomplete denitrification caused by recirculation of DO from the aerobic to the anoxic reactor. As a consequence there is another increase of the overall secondary treatment emissions compared to the base case. No substantial changes are observed in GHG due to sludge treatment, sludge reuse or embedded use of chemicals. The large impact of the dissolved oxygen concentration in both N2O and CO2 emissions indicates that process control is required in the

Scenario 2

Scenario 3

Fig. 2 e Breakdown of the GHG emissions for the different evaluated scenarios.

4706

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 0 0 e4 7 1 0

L Fig. 3 e Dynamic evolution of the effluent NHD 4 (a) and NO2 (b) in AER1 when the DO set-point is changed.

nitrification tanks with a “moderate” set-point around 2 g O2 m3. In that way, complete nitrification is ensured, at the same time avoiding oxygen returns via internal recycle.

4.2.

to these two factors it can be concluded that for this specific system lower sludge ages (high waste flow patterns) allow reducing GHG emissions i.e. N2O is non crucial, less energy consumptions, more energy credit.

Scenario 2: effect of the sludge retention time (SRT) 4.3.

The second scenario (Scenario 2) analysis investigates how the emissions of GHG change when the winter/summer waste flow rate (QW) pattern is modified. At lower SRT (QW ¼ 400/ 550 m3 day1, SRT around 12 days) there is a slight increase of the GHG emissions due to sludge treatment and disposal because the amount of TSS going to the sludge line increases (Fig. 4a) compared to the default A2 (QW ¼ 300/450 m3 day1) It is important to highlight that the additional CH4 in the digester (Fig. 4b) comes with higher energy credit, i.e. less offsite CO2 generation, because more energy can be produced from the digester biogas. Note that it is possible to appreciate the changes in the quantity of TSS going to the sludge line and the methane produced in the digester during the summer and winter periods. In this specific case, SRT (together with dissolved oxygen concentration) is long enough for nitrite oxidation to proceed. Thus, a potential increase of N2O production is avoided as it has been reported in other studies (Hanaki et al., 1992). At higher SRT (QW ¼ 200/350 m3 day1, SRT around 18 days) there is an increase of the non-N2O GHG emission produced in the bio-reactor, off-site emission due to electrical use i.e. higher aeration intensities and less energy credit produced in the anaerobic digester (see Fig. 2). Thanks

Scenario 3: effect of the COD/N ratio

The last scenario (Scenario 3) examines plant performance at different COD/N ratios in the biological reactor by modifying the addition of the external carbon source flow rate (Qcarb). As in the previous cases, higher (Qcarb ¼ 10 m3$day1) and lower (Qcarb ¼ 0 m3$day1) values are compared to the default conditions (Qcarb ¼ 5 m3 day1). A higher COD/N ratio (Qcarb ¼ 10 m3$day1) substantially increases GHG emissions in the secondary treatment (see Fig. 2) i.e. from biomass respiration and BOD oxidation. Due to the increased COD load the quantity of TSS produced is higher and thus GHG emissions due to sludge processing and disposal are also increased. Finally, the off-site CO2 emission from chemical and energy use are very high, which make this scenario the worst in terms of GHG emissions. At low COD/N ratio (Qcarb ¼ 0 m3$day1) the total emissions are extremely low (17,861 kg CO2$day1) compared to the default A2 (21,558 kg CO2$day1). This is due to the fact that zero emissions are associated to chemicals and there is a significant decrease of the direct secondary emissions and the energy-related emissions. However, secondary treatment GHG emission related N2O are higher. From this study some important points can be put forth. It is confirmed

Fig. 4 e Dynamic evolution of the quantity of TSS going to the sludge line (a) and the CH4 produced in the anaerobic digester (b) when the waste flow pattern (QW) is changed.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 0 0 e4 7 1 0

4707

Fig. 5 e Dynamic evolution of the effluent NOL 3 (a) and N2O released from ANOX2 (b) when the Qcarb is changed. that limited availability of biodegradable organic carbon increases N2O emissions during incomplete denitrification (Schultness and Gujer, 1996; Hiatt and Grady, 2008b; Kampschreur et al., 2009) (see Fig. 5a and b). However, in overall terms the decrease in endogenous respiration and BOD oxidation is more important than the increase in N2O emitted. Thus, in terms of GHG emissions it seems that is better not to add carbon.

4.4. Synergies and trade-offs amongst ghg emissions and effluent, cost and legal criteria After analyzing the influence of some operational parameters and identifying low emission control strategies, it is necessary to study their synergies and trade-offs amongst the “traditional” effluent (EQI), economic (OCI) and legal criteria (TIV). Table 3 presents the results of the total GHG emissions with EQI, OCI and TIV for the previously defined scenarios. From these results it can be seen that the scenario with the lowest GHG emissions (Qcarb ¼ 0 m3$day1) is the worst in terms of effluent quality (see values of EQI). This is mainly due to the limited biodegradable organic carbon making denitrification almost impossible (check the high values in TIV_TN). Conversely, the scenario with the highest denitrification rates promotes the highest GHG emissions and it is extremely expensive to operate due to the high carbon addition (see OCI values). The scenario with the best effluent quality involves a high DO set-point (DO ¼ 3 g O2 m3), but again it implies a high operation cost and GHG emissions due to increased aeration energy (see Table 3 and Fig. 2). Finally, when it comes to sludge retention time, a low waste flow rate enhances nitrification but also the operation cost due to aeration (see TIV_NHþ 4 and OCI values). On the other hand, when a lower SRT is selected it is possible to achieve relatively low operational cost and GHG emissions but the effluent quality will suffer due to a reduced nitrifying activity. It is important to highlight that for this specific case, the SRT is long enough for nitrification to proceed (see that TIV_NHþ 4 is around 33%). Thus potential problems with NO 2 accumulation and consequently N2O emissions are avoided. From the above it is clear that a balancing act is necessary to find the best operating conditions to satisfy the four categories of criteria evaluated here, i.e. effluent quality, economic, legal and GHG criteria.

5.

Discussion

The results of this study provide several points of discussion. Firstly, from a process engineering point of view it is possible to see that to ensure sustainable wastewater treatment operation; one must guarantee a sufficiently long SRT and an adequate dissolved oxygen concentration in the biological reactor for AOB and NOB to grow. Too high oxygen set-points and SRT increase the operational cost and GHG emissions. On the other hand, too low oxygen levels and SRT provoke nitrification failures. In this case study, COD is limiting the denitrification process and addition of organic carbon is necessary to prevent eutrophication. However, too high external carbon dosage rates in the biological reactor increase the cost of operation as well as the overall GHG emissions. Secondly, the potential adverse effects of certain operational procedures are highlighted. These effects are normally not considered with current state-of-the-art evaluation methods because they do not consider dynamics during quantification of GHG emissions in the simulation procedure. For example, the ammonium controller implemented in A3 substantially improved the overall nitrification efficiency by adapting the airflow rate to the oxygen demand for organic matter and nitrogen removal. However, when the GHG emission were quantified, it was possible to observe an increased N2O emission due to the long periods the cascade controller set the DO to low values. A low DO concentration in the nitrification tank will lead to increased N2O production. This fact has important implications from a process engineering point of view. As previously stated by Kampschreur et al. (2009) the trend of WWTP to decrease their energy consumption could be adverse towards the greenhouse effect: even though it decreases CO2 emissions from aeration energy, this could be countered by increased N2O emissions due to its 300-fold stronger greenhouse effect. Another interesting point comes from Scenario 3, where it was possible to see e just in terms of GHG emissions e that a higher external carbon source reduced the overall N2O emissions. Nevertheless, when all the other potential GHG emissions were evaluated as well, the undesirable side-effect of such addition, i.e. an increase of the off-site CO2 emissions due to chemical and energy use, was easily detected. The results of such analysis depend largely on the model selected to perform the study. When modeling activated sludge plants, there is often a disagreement about the best

4708

1.142 6143 20,393 1.09 0.45 27.73 1.29 0.67 0.865 8405 6515 99.52 0.06 6.45 0.26 0.22 0.992 6602 13,102 2.62 0.05 33.27 0.26 0.23 1.097 5915 14,059 1.8 0.06 6.43 0.30 0.23 1.022 7653 12,743 11.13 0.06 51.39 0.31 0.23

1.092 6484 13,391 1.18 0.81 9.65 1.65 0.95

Qcarb ¼ 10 m3 day1 Qcarb ¼ 0 m3 day1 QW ¼ 400/550 m3 day1 QW ¼ 200/350 m3 day1 DO ¼ 3 g O2 g m3 DO ¼ 1 g O2 g m3

model to apply in each case. The representation of the biomass decay (Siegrist et al., 1999), oversimplification of settling models (Bu¨rger et al., 2011; De Clercq et al., 2009) or the possible aerobic/anoxic mechanisms leading to N2O emissions (Ahn et al., 2010; Yu et al., 2010; Lu and Chandran, 2010) are still issues under discussion. In view of the study at hand the latter is probably the most important modification to be expected from future research. No attempt was made at this stage to model AOB-related N2O production because no consensus exists yet on the way the observations should be modeled. In the same way, the list of possible GHG sources is not complete. For example the methane generated in the anaerobic digester that remains in the liquid phase and it is recycled to the inlet of the WWTP. Upstream and downstream sources are not considered either. For instance it is known that CH4 can be formed in the sewer system (Guisassola et al., 2009) and afterwards be stripped in the treatment plant (influent, pumping station, aeration tank). Other potential sources are N2O emissions due to nitrogen discharges with the effluent or from sludge disposal (Ahn et al., 2010). For this reason, the reader should be aware that the results of this study depend on the assumptions made by the authors and presented in the methods section. In that respect it should be emphasized that the objective of the approach presented in this paper is not to predict GHG emissions with absolute accuracy. The main objective of this paper is to provide a better picture of the overall WWTP performance with this new dimension dealing with GHG emissions. With the use of this platform it is now possible to see how effluent standards, economic considerations and the causes of GHG emissions are entangled. Given this complexity, the authors advocate use of multi-objective/ multi-criteria evaluation techniques (Flores-Alsina et al., 2008) in order to include all these different factors during the decision making process. Also, the simulation values need to be interpreted with care. Scenarios create combinations of conditions and, for instance, the limitation of denitrification capacity may be due to a combination of SRT and DO limitations. The absolute values obtained in the scenario analyses should not be taken as such, but the qualitative results obtained can, as presented above. Lastly, including this type of analysis in WWTP simulation studies is actively encouraged because it can give better guidance to decision makers, process engineers and wastewater professionals on the sustainability of different treatment options. Thus, it is possible to evaluate the capabilities of the control strategy or operational procedure to handle effluent, economical, legal and GHG related issues.

1.044 6172 13,323 1.09 0.06 20.83 0.30 0.23

6.

GHG total EQI OCI TIV_TN TIV_COD TIV_NHþ 4 TIV_TSS TIV_BOD5

Scenario 2 Scenario 1 A2

Table 3 e Total GHG emission, EQI, OCI and TIV values for the different evaluated scenarios.

Scenario 3

Units

kg CO2e m3 kg poll$day1 e % % % % %

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 0 0 e4 7 1 0

Conclusions

This paper has complemented the traditional effluent quality, operational cost and legal criteria used for evaluation of control strategies in WWTP with a new dimension dealing with GHG gases. The authors have applied this approach that evaluates and quantifies the different sources of GHG gases using dynamic modeling in different control strategies and

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 0 0 e4 7 1 0

scenarios. The key findings are summarized in the following points:  By implementing controllers it is possible to reduce the overall GHG emissions as well as to improve effluent quality and reduce operational costs.  A sufficiently high DO set-point is required to ensure complete nitrification, but it should not be too high as this would (i) increase oxygen recycle to the anoxic zone with N2O emissions due to incomplete denitrification and (2) energy consumption for aeration with concomitant increase in off-site CO2 emissions. Low concentration of oxygen could lead to high N2O emissions due to nitrite accumulation although off-site CO2 are reduced.  A rather low SRT should be favored because it reduces GHG emissions by increasing the CO2 credit due to energy production by the anaerobic digester CH4 even though that also increased the sludge treatment related emissions. A high SRT substantially increases GHG emissions from secondary treatment.  Increasing the carbon source addition increases the GHG emissions due to increased sludge production that results in increased endogenous respiration, sludge treatment and disposal and both chemicals and energy use. However the N2O emissions are reduced.  It is necessary to find a compromise solution between effluent quality, costs, legal and GHG criteria to reach sustainable modes of operation. In this case study, it was possible with a DO set-point of 2 g O2 m3 and moderate SRT and external carbon source addition rates.

Acknowledgements Lluı´s Corominas received the “Juan de la Cierva” scholarship from the Ministry of Spain. Peter Vanrolleghem holds the Canada Research Chair on Water Quality Modeling. The authors acknowledge the financial support obtained through the TECC project of the Que´bec Ministry of Economic Development, Innovation and Exports (MDEIE).

references

Ahn, J.H., Kim, S., Park, H., Rahm, B., Pagilla, K., Chandran, K., 2010. N2O emissions from activated sludge processes, 2008e2009: results of a national monitoring survey in the United States. Environ. Sci. Technol. 44, 4505e4511. Ayesa, E., De la Sota, A., Grau, P., Sagarna, J.M., Salterain, A., Suescun, J., 2006. Supervisory control strategies for the new WWTP of Galindo-Bilbao: the long run from the conceptual design to the full-scale experimental validation. Water Sci. Technol. 53 (4e5), 193e201. Bani Shahabadi, M., Yerushalmi, L., Haghighat, F., 2009. Impact of process design on greenhouse gas (GHG) generation by wastewater treatment plants. Water Res. 43, 2679e2687. Batstone, D.J., Keller, J., Angelidaki, I., Kayuznyi, S.V., Pavlostathis, S.G., Rozzi, A., Sanders, W.T.M., Siegrist, H., Vavilin, V.A., 2002. Anaerobic Digestion Model No 1. IWA Publishing, London, UK. IWA STR No 13.

4709

Bridle, T., Shaw A., Cooper S., Yap K.C., Third K., Domurad M. 2008. Estimation of greenhouse gas emissions from wastewater treatment plants. In: Proceedings IWA World Water Congress 2008, Vienna, Austria. September 7e12, 2008. Bu¨rger, R., Diehl, S., Nopens, I., 2011. A consistent modelling methodology for secondary settling tanks in wastewater treatment. Water Res. 45, 2247e2260. Cakir, F.Y., Stenstrom, M.K., 2005. Greenhouse gas production: a comparison between aerobic and anaerobic treatment technology. Water Res. 39 (17), 4197e4203. Copp, J.B., 2002. The Cost Simulation Benchmark: Description and Simulator Manual. Office for Official Publications of the European Community, Luxembourg. Corominas, L., Sin, G., Puig, S., Traore, A., Balaguer, M., Colprim, J., Vanrolleghem, P.A., 2006. Model-based evaluation of an online control strategy for SBRs based on OUR and ORP measurements. Water Sci. Technol. 53 (4e5), 161e169. Corominas, Ll., Flores-Alsina X., Snip L., Vanrolleghem P.A., 2010. Minimising overall greenhouse gas emissions from wastewater treatment plants by implementing automatic control. In: Proceedings 7th IWA Leading-Edge Conference on Water and Wastewater Technologies. Phoenix, AZ, USA, June 2e4, 2010. Devisscher, M., Bogaert, H., Bixio, D., Van de Velde, J., Thoeye, C., 2002. Feasibility of automatic chemicals dosage control: a fullscale evaluation. Water Sci. Technol. 45, 4e5. 445e442. De Clercq, J., Nopens, I., Defrancq, J., Vanrolleghem, P.A., 2009. Extending and calibrating a mechanistic hindered and compression settling model for activated sludge using indepth batch experiments. Water Res. 42 (3), 781e791. Dong, Y., Steinberg, M., 1997. Hynol e an economical process for methanol production from biomass and natural gas with reduced CO2 emission. Int. J. Hydrogen. Energy. 22 (10e11), 971e977. Flores-Alsina, X., Sin, G., Rodriguez-Roda, I., Gernaey, K.V., 2008. Multicriteria evaluation of wastewater treatment plant control strategies under uncertainty. Water Res. 42 (17), 4485e4497. Flores-Alsina, X., Comas, J., Rodriguez-Roda, I., Poch, M., Gernaey, K., Jeppsson, U., 2009. Evaluation of plant wide control strategies including the effects of filamentous bulking sludge. Water Sci. Technol. 60 (8), 2093e2103. Foley, J., de Haas, D., Yuan, Z., Lant, P., 2010. Nitrous oxide generation in full-scale biological nutrient removal wastewater treatment plants. Water Res. 44, 831e844. Gernaey, K.V., Rosen, C., Jeppsson, U., 2006. WWTP dynamic disturbance modelling e an essential module for long-term benchmarking development. Water Sci. Technol. 53 (4e5), 225e234. Greenfield, P.F., Batstone, D.J., 2003. Anaerobic digestion: impact of future previous term greenhouse next term gases mitigation policies on methane generation and usage. Water Sci. Technol. 52 (1e2), 39e47. Guisassola, A., Sharma, K.R., Keller, J., Yuan, Z., 2009. Development of a model for assessing methane formation in rising main sewers. Water Res. 43 (11), 2874e2884. Gujer, W., Erni, P., 1978. The effect of diurnal ammonium load variation on the performance of nitrifying activated sludge processes. Prog. Water Technol. 10 (5/6), 391e407. Hanaki, K., Hong, Z., Matsuo, T., 1992. Production of nitrous oxide gas during denitrification of wastewater. Water Sci. Technol. 26 (1e12), 1027e1036. Henze, M., Gujer, W., Mino, T., van Loosdrecht, M.C.M., 2000. Activated Sludge Models ASM1, ASM2, ASM2d and ASM3. IWA Scientific and Technical Report No 9. IWA Publishing, London, UK. Hiatt, W.C., Grady Jr., C.P.L., 2008a. An updated process model for carbon oxidation, nitrification, and denitrification. Water Environ. Res. 80, 2145e2156.

4710

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 0 0 e4 7 1 0

Hiatt, W.C., Grady Jr., C.P.L., 2008b. Application of the activated sludge model for nitrogen to elevated nitrogen conditions. Water Environ. Res. 80, 2134e2144. Ingildsen, P., Jeppsson, U., Olsson, G., 2002. Dissolved oxygen controller based on on-line measurements of ammonium combining feed-forward and feedback. Water Sci. Technol. 45 (4e5), 453e460. IPCC, 2006. 2006 IPCC Guidelines for National Greenhouse Gas Inventories. Available at. Intergovernmental Panel on Climate Change. http://www.ipccnggip.iges.or.jp/public/2006gl/index. html. Jeppsson, U., Pons, M.N., Nopens, I., Alex, J., Copp, J.B., Gernaey, K. V., Rosen, C., Steyer, J.P., Vanrolleghem, P.A., 2007. Benchmark simulation model no 2 e general protocol and exploratory case studies. Water Sci. Technol. 56 (8), 287e295. Kampschreur, M.J., Temmink, H., Kleerebezem, R., Jetten, M.S.M., van Loosdrecht, M.C.M., 2009. Nitrous oxide emission during wastewater treatment. Water Res. 43 (17), 4093e4103. Keller, J., Hartley, K., 2003. Greenhouse gas production in wastewater treatment: process selection is the major factor. Water Sci. Technol. 47 (12), 43e48. Lessard, P., Beck, M.B., 1993. Modelling of the activated sludge process: a case of Norwich plant. Water Res. 27 (6), 963e978. LGO (2008). Local government operations protocol for the quantification and reporting of greenhouse gas emissions inventories. Version 1.0, September 2008. Available at: www. theclimateregistry.org/resources/protocols/local-governmentoperations-protocol Lu, H., Chandran, K., 2010. Factors promoting emissions of nitrous oxide and nitric oxide from denitrifying sequencing batch reactors operated with methanol and ethanol as electron donors. Biotechnol. Bioeng. 106 (3), 390e398. Machado, V.C., Gabriel, D., Lafuente, J., Baeza, J.A., 2009. Cost and effluent quality controllers design based on the relative gain array for a nutrient removal WWTP. Water Res. 43 (20), 5129e5141. Monteith, H.D., Sahely, H.R., MacLean, H.L., Bagley, D.M., 2005. A rational procedure for estimation of greenhouse-gas emissions from municipal wastewater treatment plants. Water Environ. Res. 77, 390e403. NGER, 2008. National greenhouse and energy reporting (measurement) technical guidelines 2008 v1.1. July 2008 Available at. Australian Government Department of Climate Change. http://www.climatechange.gov.au/reporting/. Nopens, I., Batstone, D.J., Copp, J.B., Jeppsson, U., Volcke, E., Alex, J., Vanrolleghem, P.A., 2009. An ASM/ADM model interface for dynamic plant-wide simulation. Water Res. 43 (7), 1913e1923. Nopens, I., Benedetti, L., Jeppsson, U., Pons, M.-N., Alex, J., Copp, J.B., Gernaey, K.V., Rosen, C., Steyer, J.-P., Vanrolleghem, P.A., 2010. Benchmark simulation model no 2: finalisation of plant layout and default control strategy. Water Sci. Technol. 62 (9), 1967e1974. Oennerth, T.B., Nielsen, M.K., Stamer, C., 1996. Advanced computer control based on real and software sensors. Water Sci. Technol. 33 (1), 237e245.

Olsson, G., Nielsen, M.K., Yuan, Z., Lynggaard-Jensen, A., Steyer, J.P., 2005. Instrumentation, Control and Automation in Wastewater Systems. IWA Publishing, London, UK. Otterpohl, R., Freund, M., 1992. Dynamic models for clarifiers of activated sludge plants with dry and wet weather flows. Water Sci. Technol. 26 (5e6), 1391e1400. Otterpohl, R., Raak, M., Rolfs, T. 1994. A mathematical model for the efficiency of the primary clarification. In: Proceeding of 17th IAWQ Biennial International Conference, 24e29 July, Budapest, Hungary. Pagilla, K., Shaw, A., Kunetz, T., Schiltz, M. 2009. A systematic approach to establishing carbon footprints for wastewater treatment plants. In: Proceedings of WEFTEC 2009, Orlando, Florida, USA, October 10e14, 2009. Rieger, L., Alex, J., Winkler, S., Boehler, M., Thomann, M., Siegrist, H., 2003. Progress in sensor technology - progress in process control? Part I: sensor property investigation and classification. Water Sci. Technol. 47 (2), 103e112. Saravanamuttoo, H., Rogers, G.F.C., Cohen, G.F.C., Straznicky, P., 2009. Gas Turbine Theory. Pearson Education Limited, Halcrow, UK. von Schulthess, R., Gujer, W., 1996. Release of nitrous oxide (N2O) from denitrifying activated sludge: verification and application of a mathematical model. Water Res. 30 (3), 521e530. Siegrist, H., Brunner, I., Koch, G., Con Phan, L., Van Chieu, L., 1999. Reduction of biomass decay under anoxic and anaerobic conditions. Water Sci. Technol. 39 (1), 129e137. Snip, L. 2010. Quantifying the greenhouse gases emissions of wastewater treatment plants. Msc thesis. Department of Agrotechnology and Food Science. Wageningen University. De´partement de ge´nie civil. Universite´ Laval (Available at http://edepot.wur.nl/138115). Spanjers, H., Vanrolleghem, P.A., Nguyen, K., Vanhooren, H., Patry, G.G., 1998. Towards a simulation-benchmark for evaluating respirometry-based control strategies. Water Sci. Technol. 37 (12), 219e226. Stare, A., Vrecko, D., Hvala, N., Strmcnik, S., 2007. Comparison of control strategies for nitrogen removal in an activated sludge process in terms of operating costs: a simulation study. Water Res. 41 (9), 2004e2014. Taka´cs, I., Patry, G.G., Nolasco, D., 1991. A dynamic model of the clarification thickening process. Water Res. 25 (10), 1263e1271. Tchobanoglous, G., Burton, F.L., Stensel, H.D., 2003. Wastewater Engineering: Treatment, Disposal and Reuse. McGraw-Hill, New York. Yu, R., Kampschreur, M.J., van Loosdrecht, M.C.M., Chandran, K., 2010. Mechanisms and specific directionality of autotrophic nitrous oxide and nitric oxide generation during transient anoxia. Environ. Sci. Technol. 44 (4), 1313e1319. Zhao, H., Isaacs, S.H., Soeberg, H., Kummel, M., 1995. An analysis of nitrogen removal and control strategies in an alternating activated-sludge process. Water Res. 29 (2), 535e544.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 1 1 e4 7 2 1

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

The effect of hydraulic retention time on granular sludge biomass in treating textile wastewater Khalida Muda a,*, Azmi Aris a,**, Mohd Razman Salim a, Zaharah Ibrahim b, Mark C.M. van Loosdrecht c, Azlan Ahmad a, Mohd Zaini Nawahwi b a

Department of Environmental Engineering, Faculty of Civil Engineering, Universiti Teknologi Malaysia, 81310 Skudai, Johor, Malaysia Department of Biological Sciences, Faculty of Biosciences and Bioengineering, Universiti Teknologi Malaysia, 81310 Skudai, Johor, Malaysia c Department of Biotechnology, Delft University of Technology, Julianalaan 67,2628 BC Delft, The Netherlands b

article info

abstract

Article history:

The physical characteristics, microbial activities and kinetic properties of the granular

Received 12 February 2011

sludge biomass were investigated under the influence of different hydraulic retention

Received in revised form

times (HRT) along with the performance of the system in removal of color and COD of

7 May 2011

synthetic textile wastewater. The study was conducted in a column reactor operated

Accepted 12 May 2011

according to a sequential batch reactor with a sequence of anaerobic and aerobic reaction

Available online 20 May 2011

phases. Six stages of different HRTs and different anaerobic and aerobic reaction time were evaluated. It was observed that the increase in HRT resulted in the reduction of organic

Keywords:

loading rate (OLR). This has caused a decrease in biomass concentration (MLSS), reduction

Textile wastewater

in mean size of the granules, lowered the settling ability of the granules and reduction of

SBR

oxygen uptake rate (OUR), overall specific biomass growth rate (ı`overall), endogeneous decay

Anaerobic/aerobic reaction

rate (kd) and biomass yield (Yobs, Y). When the OLR was increased by adding carbon sources

Biokinetics

(glucose, sodium acetate and ethanol), there was a slight increase in the MLSS, the granules mean size, ı`overall, and biomass yield. Under high HRT, increasing the anaerobic to aerobic

Color

reaction time ratio caused an increase in the concentration of MLSS, mean size of granules and lowered the SVI value and biomass yield. The `ıoverall and biomass yield increased with the reduction in anaerobic/aerobic time ratio. The HRT of 24 h with anaerobic and aerobic reaction time of 17.8 and 5.8 h respectively appear to be the best cycle operation of SBR. Under these conditions, not only the physical properties of the biogranules have improved, the highest removal of color (i.e. 94.1
0.6%) and organics (i.e. 86.5
0.5%) of the synthetic textile dyeing wastewater have been achieved. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Granulation techniques for dye degradation processes have been reported by many researchers (van der Zee, 2003; Isik and Sponza, 2008). Most of the studies focused on the application of anaerobic process since major decolorization process occurs under this condition (Isik and Sponza, 2005; Somasiri et al., 2008).

Nevertheless, complete mineralization of dye containing wastewater requires both anaerobic and aerobic conditions. Several studies have therefore been conducted to investigate the removal efficiency under both operating conditions with series of anaerobic and aerobic reactor system (Isik and Sponza, 2004a, 2004b). Isik and Sponza (2008) reported that more than 91% and 84% removal efficiency of COD and color

* Corresponding author. Tel.: þ60 7 5531522; þ60 7 5531581; fax: þ60 7 5566157. ** Corresponding author. E-mail addresses: [email protected] (K. Muda), [email protected] (A. Aris). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.012

4712

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 1 1 e4 7 2 1

respectively, were observed when a mixed dye compounds in synthetic textile wastewater were treated using partially anaerobic granule/activated sludge in a sequential anaerobic/ aerobic reactor system. Previous study also shows that the overall percentage of color removal in textile wastewater treatment is very much affected by the hydraulic retention time (HRT) particularly during anaerobic phase as the cleavage of the eN¼Ne bond reported to occur more under anaerobic condition (Pasukphun and Vinitnantharat, 2003; Buitron et al., 2004; Ong et al., 2008). Color removal has been found to increase as the HRT of anaerobic phase increases (Buitron et al., 2004; Li and Xi, 2004; Van der Zee and Villaverde, 2005). However, different types of biomass (i.e. suspended cells, biofilm or granules) used in the treatment system may have different HRT requirements and will affect the design of the system. Successful cultivation of microbial granular sludge using synthetic and raw textile dyeing wastewater in a single reactor column operating under intermittent anaerobic and aerobic reaction phases has been reported earlier (Muda et al., 2010; Ibrahim et al., 2010). While the developed granules appear to be capable in treating the textile wastewater, their performance under different range of HRT is still unknown. The paper discusses how the granules properties, reactor performance, and biokinetics can be changed due to the variation in the HRT of the reactor.

2.

Materials and methods

2.1.

Granular biomass

2.3.

Reactor set-up

The schematic illustration of the reactor set-up is given in Fig.1. The design of the reactor system was based on Wang et al. (2004) and Zheng et al. (2005) with several modifications. The height of the reactor column was 100 cm with 8 cm internal diameter. The column was designed to be operated with 4-L working volume. The wastewater enters the reactor from the bottom of the column. A fine air bubble diffuser was placed at the bottom of the column for aerobic phase. Air bubbles was introduced into the reactor intermittently and controlled by a timer. The withdrawal of the wastewater took place via an outlet port located at 40 cm height from the bottom of the reactor which gave 50% of the volumetric exchange rate.

2.4.

Analytical methods

The effect of HRT was investigated with respect to the changes in the physical characteristics, microbial activity and removal performance of the granular biomass in the reactor system. Physical characteristics include particle size distribution, settling velocity (SV) and sludge volume index (SVI). The particle size distribution was measured throughout the experiment using wet sieves analysis where the volume of each different granular size was expressed as a fraction of the total volume of the granule. The SV was determined by recording the average time taken for an individual granule to settle at a certain height in a glass column filled with tap water. The SVI was determined based on Beun et al. (1999). The bed volume was obtained by measuring the bed height of the sludge biomass that settle within 5 min in the reactor

The development of the microbial granules has been explained in detail in Muda et al. (2010). The size of the microbial granules selected for this experiment was in the range of 0.3e2.5 mm.

2.2.

Wastewater composition

Synthetic textile dyeing wastewater which was applied in this study was prepared according to Tan et al. (2005) and Sirianuntapiboon and Srisornsak (2007) with several modifications. The wastewater contains a mixture of an equal ratio (based on the COD concentration) of glucose (0.5 g L1), ethanol (0.125 g L1) and sodium acetate (0.5 g L1) as the carbon sources. Other components including NH4Cl 0.16 g L1, KH2PO4 0.23 g L1, K2HPO4 0.58 g L1, CaCl2∙2H2O 0.07 g L1, MgSO4∙7H2O 0.09 g L1, EDTA 0.02 g L1 and trace solution 1 mL L1 were also added in the media solution. Several trace elements recommended by Smolders et al. (1995) were used as the essential mineral content. A total concentration of 50 mg L1 of Sumifix Black EXA, Sumifix Navy Blue EXF and Synozol Red K-4B were used as mixed dyes in this study. These azo dyes represent some of the dyestuffs that are commonly used in dyeing cotton, polyester and polyacrylic fabrics (Correia et al., 1994). The COD and BOD of the synthetic wastewater were 1270 mg/L and 312 mg/L, respectively with COD over BOD ratio of about 4.1 indicating the relatively non-biodegradable nature of the wastewater.

Fig. 1 e Schematic layout of the reactor system (Wang et al., 2004; Zheng et al., 2005).

4713

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 1 1 e4 7 2 1

Table 1 e Detail of experimental condition of the reactor system. Stage

I II III IV V VI

Days covered

Phase (h) 1st

OLR (kgCOD m3 day1)

6 12 24 24 24 24

2.5 1.3 0.6 0.8 0.8 0.8

2nd

Anaerobic

Aerobic

Anaerobic

Aerobic

1.42 2.92 5.92 5.92 8.92 2.92

1.42 2.92 5.92 5.92 2.92 8.92

1.42 2.92 5.92 5.92 8.92 2.92

1.42 2.92 5.92 5.92 2.92 8.92

49 43 51 43 46 46

HRT (h)

OLR ¼ X/VtotalVadd/t, where X ¼ COD concentration of the influent (mg L1); Vadd ¼ Volume of influent added in each cycle operation (mL); Vtotal ¼ Total working volume of the experiment (mL); t ¼ Hydraulic retention time (h).

after the aeration phase stopped and was multiplied with the surface area of the inner reactor column. Then, the obtained bed volume was divided with the dry weight of the biomass in the reactor. The microbial activity of the microbial granules was conducted by measuring the oxygen utilization rate (OUR) at different HRT. The OUR (mg DO L1 h1) measurements were performed by following the Standard Methods (APHA, 2005). Sludge retention time (SRT) was determined according to Beun et al. (1999). Other parameters such as mixed liquor suspended solid (MLSS), mixed liquor volatile suspended solid (MLVSS), COD and color were analyzed according to Standard Methods (APHA, 2005).

in reduction of the organic loading rate (OLR) supplemented into the reactor from 2.5 to 0.6 kg COD m3 day1. The HRT for Stage III to VI was kept constant, i.e. 24 h, but the durations of anaerobic and aerobic reaction phases were varied. From stage III onwards, the OLR was increased to 0.8 kg COD m3 day1 by increasing the concentration of the carbon sources in the synthetic textile dyeing wastewater. The temperature of the treatment system was kept constant at 30.0
2.0  C while the pH throughout the experiment was between 6.3 and 8.0.

3.

Results and discussions

2.5.

3.1.

Physical profile of the reactor System

Experimental procedures

The reactor system was operated according to a sequential batch reactor (SBR). A complete cycle of the operation comprised of fill (15 min), react (6e24 h), settle (5 min), decant (5 min) and idle (5 min) steps. The react step consisted of alternate anaerobic and aerobic reaction phase. During the anaerobic reaction phase, circulation of wastewater was carried out using peristaltic pump (ColeeParmer System Model, 6e600 rpm) at a flow rate of 18 L h1. Air was diffused into the reactor using an air pump at a superficial air velocity of 2.6 cm s1 throughout the aerobic react phase. The details of the experimental conditions are shown in Table 1. The HRT of the experiment was increased from 6 to 24 h in Stage I to Stage III. The increase in the HRT resulted

The changes in the biomass concentration at different HRT (Stage I to III) are shown in Table 2 and Fig. 2. It is apparent that the biomass concentration (MLSS) in the reactor decreases from about 35 g L1 to 25 g L1 as the HRT increases from 6 to 24 h. The reduction of the biomass concentration in the reactor is expected to be due to the reduction of OLR applied in the reactor system (i.e. from 2.5 to 0.6 kg COD m3 day1) as a result of the increase in HRT. Reduction in the OLR means less carbon was supplied to the microorganisms in the reactor and hence, less growth is taking place. When the OLR was increased to 0.8 kg COD m3 day1, the MLSS started to increase again to 30.5
3.4 g L1 and 31.6
3.7 g L1 in Stage IV and V. Similarly, the ratio of the

Table 2 e Biomass concentrations at different stages of the experiment. Reaction Phase

Stage I

Anaerobic (h) Aerobic (h) MLSS (g L1) MLVSS (g L1) MLVSS/MLSS Effluent (VSS g L1) SRT (day)

2.8 2.8 35.3 31.9 0.90 0.34 27.6


1.6
1.8
0.16
13.4

II 5.8 5.8 28.7 24.5 0.85 0.31 42.4


0.6
2.2
0.11
10.2

III

IV

V

VI

11.8 11.8 25.2
1.8 18.5
2.2 0.73 0.26
0.19 78.9
23.9

11.8 11.8 30.5
3.4 26.0
3.4 0.85 0.34
0.11 70.1
23.9

17.8 5.8 31.6
3.7 22.4
2.0 0.71 0.33
0.10 72.5
23.3

5.8 17.8 23.3
0.8 20.2
0.8 0.87 0.55
0.22 41.6
18.4

4714

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 1 1 e4 7 2 1

Since the system was operated with no sludge wastage at any of the operating time, Eq. (1) can be simplified as Eq. (2) (Tchobanoglous et al., 2004; Liu and Tay, 2007). q¼

Fig. 2 e Profile of biomass concentration at different stages of the experiment. (C) MLSS, (,) MLVSS. Stage I: anaerobic (2.8 h): aerobic (2.8 h); Stage II: anaerobic (5.8 h): aerobic (5.8 h); Stage III and Stage IV: anaerobic (11.8 h): aerobic (11.8 h); Stage V: anaerobic (17.8 h): aerobic (5.8 h); Stage V: anaerobic (5.8 h): aerobic (17.8 h).

volatile biomass (MLVSS) to total biomass (MLSS) reduced from Stage I to III mainly due to decrease in the OLR as the HRT increased from 6 to 24 h. It was also observed that when the duration of the anaerobic phase was increased, the MLVSS/MLSS ratio was decreased (i.e. 0.71). The decrease in MLVSS/MLSS ratio may indicate an increase of inert particles within the granules. The same observation was reported by Panswad et al. (2001) that increase of inert solids in the biomass was observed when the system was exposed to high anoxic/anaerobic condition in the SBR cycle. When the duration of aeration phase was increased up to 18 h, the biomass started to reduce again (Stage VI) and increase of VSS (Table 2) in the effluent was observed. This indicates that too long of aerobic reaction phase is not suitable for granular biomass system as the turbulent caused by the aeration may rupture the granules causing them to leave the reactor during decant phase due to their small size and long settling time. The sludge retention time (SRT) of the biomass in the SBR system can be calculated by using Eq. (1) (Tchobanoglous et al., 2004; Liu and Tay, 2007). q¼

Xvss Vr tc Xd Vd þ Xe Ve

(1)

Xvss Vr tc Xe Ve

(2)

Based on Eq. (2), the SRT of the reactor system increases from 27.6
13.4 to 78.9
30.8 d when the period of the HRT increases from 6 to 24 h (Stage I to III). This is mainly due to the decrease of MLSS and also MLVSS caused by lower OLR as previously explained. With HRT of 24 h, increase of anaerobic reaction phase up to 18 h (Stage IV to V) has slightly increased the SRT from 70.1
23.9 to 72.5
23.3 d. The SRT value changes in each stage of the experiment. According to Wijffels and Tramper (1995), the favorable sludge age for high removal efficiency for COD and nitrification process is more than 4 days. Based on the SRT obtained, this granular system is capable of simultaneous nitrification process and COD removal. Since the treatment goal is to remove recalcitrant dyeing compound, the SRT value of all stages evaluated in this experiment was in the acceptable range for the degradation of xenobiotic compounds (Grady et al., 1999).

3.2.

Physical properties of the granular biomass

The effects of HRT variation (6e24 h) on the physical properties of the granules are given in Table 3. In general, the mean size of the granules reduces with the increase in the HRT which resulted in higher SVI and lower SV values. The mean granular size in the reactor was the largest (i.e. 843
44 ı`m) during Stage I which has the shortest HRT and highest OLR. When the HRT increases from 6 to 24 h, the OLR was reduced from 2.5 to 0.6 kg COD m3 day1. This condition may contribute to the smaller granules formation due to the reduction in the food supply. Furthermore, as the HRT increases from 6 to 24 h, the duration of the aeration phase also increases from about 3 to 12 h. Hence, the reduction in the mean granular size may also be due to the long exposure of the granules to the shear force imposed by aeration process. When the OLR was increased from 0.6 to 0.8 kg COD m3 day1from Stage III to IV, the mean granular size slightly increased due to increase in food supply. Fig. 3 shows the particle size distribution of granular biomass in the reactor at each stage of the experiment. The figure shows that the particle size distribution was clearly affected by the HRT and aeration time which imposed shear force to the granules.

Table 3 e Physical properties of the granular biomass at different stages of experiment. Reaction Phase

Anaerobic (h) Aerobic (h) Mean size (mm) SVI (mL g1) SV (m h1)

Stage I

II

III

IV

V

VI

2.8 2.8 843
44 13.1
0.4 41.3
3.1

5.8 5.8 590
55 18.8
1.5 35.1
0.8

11.8 11.8 440
40 21.4
1.6 24.5
1.1

11.8 11.8 567
79 16.8
1.3 28.4
1.3

17.8 5.8 575
46 15.5
1.3 33.4
2.5

5.8 17.8 385
22 24.8
0.9 21.3
0.5

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 1 1 e4 7 2 1

4715

In Stage V, although the HRT was similar to Stage IV (i.e. 24 h), the mean granular size was observed to slightly increase. This is apparently due to the shorter aerobic phase (i.e. 6 h) as the anaerobic phase was prolonged to 18 h. This shows that bigger granules could still be maintained in the reactor at high HRT provided that the aerobic phase is reduced by increasing the anaerobic phase. This condition seems to be suitable for treating textile wastewater that requires both anaerobic and aerobic phases. Increase in the size of the granular biomass in Stage V may be achieved by the equilibrium between the growth factor and detachment of the granules due to the shear force effect. The intermittent anaerobic and aerobic reaction phase will influence the production and consumption rate of EPS that eventually causes the changes on the surface charges of the granular biomass (Foster, 1991). A reasonable amount of EPS production with suitable shear force imposed by the aeration rate may lead to the successful formation and growth of granular biomass (Li et al., 2006). The SVI and SV of the granular sludge were used to evaluate the granular settling ability. It is anticipated that bigger granules would have higher SV and hence, reduce the SVI value, indicating good settling ability. The SVI value changes with the same pattern as the granular biomass concentration as well as the mean particle size of the granules. As the particles of the granular biomass decreased in size, the SVI value increased. The SVI value improved when the anaerobic reaction phase was prolonged in Stage V indicating such reaction pattern would help to develop granules with better settling profile. The SVI value improved as the size of the granules become bigger. It could also be possible that the increased accumulation of inert particles within the granules have contributed to improvement of SVI properties of the granular biomass. Fig. 4 shows the profile of SVI throughout the experiments. The SVI value in Stage V was reduced from 16.8
1.3 mL g1 (in Stage IV) to 15.5
1.3 mL g1. This is expected to be due to the accumulation of more inert solids within the granules as shown with small ratio of MLVSS/MLSS in Stage V (0.71). Despite changes in HRT that caused decrease in the granular sizes, the SVI values of the whole experiments were good except for Stage VI. During Stage VI, the prolonged duration

of the aerobic phase (i.e. 17.8 h) which was operated at high superficial air velocity (2.5 cm s1), caused the granular biomass to rupture. At this stage, the size of the granular biomass became smaller causing the settleability of the particles to reduce and was demonstrated with the increase in the SVI value.

Fig. 3 e Distribution of size particles at different stages of the experiment. Stage I: anaerobic (2.8 h): aerobic (2.8 h); Stage II: anaerobic (5.8 h): aerobic (5.8 h); Stage III and Stage IV: anaerobic (11.8 h): aerobic (11.8 h); Stage V: anaerobic (17.8 h): aerobic (5.8 h); Stage V: anaerobic (5.8 h): aerobic (17.8 h).

Fig. 4 e Profile of sludge volume index throughout the experiment. Stage I: anaerobic (2.8 h): aerobic (2.8 h); Stage II: anaerobic (5.8 h): aerobic (5.8 h); Stage III and Stage IV: anaerobic (11.8 h): aerobic (11.8 h); Stage V: anaerobic (17.8 h): aerobic (5.8 h); Stage V: anaerobic (5.8 h): aerobic (17.8 h).

3.3.

Microbial activity and biokinetics

3.3.1.

Oxygen uptake rate

Microbial activity was measured based on the oxygen uptake rate during the aerobic phase in a one complete cycle operation. The OUR was measured several times before each stage ended. Figs. 5 and 6 show the profiles of the OUR during aerobic phase for Stage I to VI. The results of the OUR measurement show that most of the external substrate was consumed within the first 30 min of each aerobic reaction phase. This is shown by the sharp decrease of the oxygen consumption at the initial stage of the first aerobic phase for all Stage I to VI of the experiment. The OUR profiles of Stage I to Stage III (Fig. 5) show that the OUR at the beginning of each aeration phase reduces as the HRT increases. This is possibly due to the reduction in the OLR as the HRT increases. As the organic loading is reduced, less oxygen is therefore required. After a sharp increase of OUR at the beginning of each aerobic phase in all stages (Stage I to VI), the OUR was constantly low until the end of the aerobic phase. The low value of OUR indicates that most of the external substrates have been consumed. These figures (5 and 6) show that no further degradation took place although the HRT was extended. During the aerobic phase, after all of the external substrate was consumed, the microorganisms in the reactor undergo starvation phase where endogenous respiration took place. At the beginning of the second aerobic reaction phase, a relatively smaller increase in the OUR could be observed. The increase in the OUR measurement was believed probably to be due to the mineralization of amines, the byproducts of dye degradation during the anaerobic reaction phase.

4716

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 1 1 e4 7 2 1

Fig. 5 e OUR profile of (a) Stage I (Aerobic phase 2.84 h), (b) Stage II (Aerobic phase 5.84 h) and (c) Stage III (Aerobic phase 11.84 h).

Fig. 6 shows the Stage IV to VI which was operated with the same HRT and organic loading but different in the anaerobic and aerobic reaction phase HRT ratio. As the duration of the anaerobic reaction phase was increased, the short pulse of the OUR in the second aerobic phase has increased from 96 to 130 mg L1 h1 in Stage IV and V, respectively. However, the OUR in the second aerobic phase for Stage VI has reduced to less than 80 mg L1 h1. These changes were postulated to be related to the length of second stage of anaerobic reaction phase. Prolong of the anaerobic reaction phase has increased the production of amines which was further degraded under aerobic phase causing an increase in

the OUR. This condition is vice verse when the length of anaerobic stage was shorten (Stage VI).

3.3.2.

Biokinetic parameters

The total solid biomass concentration in a biological reactor system is governed by the rate of substrates utilization and biomass production by the microorganisms. The rates of such processes which are known as the biokinetic parameters would give prediction on the performance of the biological process in wastewater treatment. The understanding and information on the rate of biological reactions and basic principles governing the growth of microorganisms are

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 1 1 e4 7 2 1

4717

Fig. 6 e OUR profile of (a) Stage IV (Aerobic phase 11.84 h), (b) Stage V (Aerobic phase 5.84 h), (c) Stage VI (Aerobic phase 17.84 h).

very important in developing an effective design and operation of the biological reactor system (Tchobanoglous et al., 2004). In this study, the biokinetic parameters of the granular sludge were also investigated in relation to the effect of different HRTs. The biokinetic parameters that were investigated are the overall specific biomass growth rate (ı`overall), endogenous decay rate (kd), observed biomass yield (Yobs), and theoretical biomass yield (Y). The calculations for the biokinetic parameters are according to the equations listed in Table 4 and are based on Liu and Tay (2007) and Chen et al. (2008). The results of biokinetic parameters for all stages in this experiment are given in Table 5. When the experiment moved from Stage I to III, the SRT was increased from 27.6
13.4 to 78.9
30.8 d which have caused the ı`overall to reduce from 0.036 to 0.013/d. The results are in accordance with Li et al. (2006) that sludge biomass will lose their bioactivity when the SRT is increased. The reduction of the ı`overall as the HRT increased was also observed by Liu and

Tay (2007). As mentioned earlier, the OLR was reduced when the HRT was increased from 6 to 24 h (from Stage I to III). The reduction in the OLR may also contribute to the reduction of ı`overall from Stage I to III. The ı`overall of Stage IV and V was the same when the SRT of these two stages slightly increased from 70.1
23.9 to 72.5
23.3 d, respectively. The ı`overall of Stage VI increased as the SRT was reduced to 41.6
18.4 d although Stage VI was operated with the same HRT as Stage IV and V. The reduction of the SRT in Stage VI may be contributed by the increase in the sludge washout that was shown by the increase in the suspended solids concentration in the effluent discharge. The rate of biomass lost due to endogenous respiration is represented by endogenous decay rate kd, as given in Eq. (4). The OUR that was measured during the last 10 min or before the second aeration phase stop of one cycle operation was used to calculate the kd. As the HRT increased from 6 to 24 h (Stage I to III), the kd values reduced. However, since the

4718

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 1 1 e4 7 2 1

Table 4 e Coefficient of biokinetic parameters. Biokinetic coefficient Overall specific biomass growth rate

Units

Formula

Per day

moverall ¼

1 q

(3)

q ¼ sludge retention time Endogenous decay rate

Per day

kd ¼

dO2 =dt OX  M

(4)

dO2/dt ¼ oxygen uptake rate (mg/L h) Ox ¼ theoretical chemical oxygen demand which is assume as 1.42 mg O2/mg biomass M ¼ biomass concentration (mg VSS/L) Observed biomass yield

mg VSS/mg COD

Yobs ¼

Xe Ci  Ce

(5)

Xe ¼ Effluent volatile solid concentration (g VSS/L) Ci ¼ COD concentration in the influent (mg/L) Ce ¼ COD concentration in the effluent (mg/L) Theoretical biomass yield

mg VSS/mg COD

Y ¼ Yobs ð1 þ kd $qÞ

reduction was also very small, the kd can be considered as constant when the HRT was increased. Furthermore, the kd value during 24 h HRT of Stage III to V can also be considered constant (i.e. 0.0075 to 0.0076/d). It can thus be concluded that the kd is considered constant throughout the experiment. The kd values calculated from this study were very small as compared to the kd values of aerobic granules (Chen et al., 2008) and of the activated sludge (Tchobanoglous et al., 2004). The observed biomass yield Yobs is the ratio of the biomass production rate to the substrate removal rate and is calculated according to Eq. (5). The Yobs is one of the most important parameter used in biological kinetic models. Eq. (5) is derived from the equation below (Liu and Tay, 2007): Yobs ¼

  ðXvss2  Xvxss1 ÞVf þ Xe Ve tc ðCi  Ce ÞVe =tc

(7)

Then Eq. (5) is simplified from Eq. (7) when the reactor system reached steady state and the biomass was maintained at a constant value (Chen et al., 2008). The Yobs can be used to describe the sludge productivity which relates to the net sludge production. The results in Table 5 show that the sludge production is inversely related to the value of SRT as shown in Stage I to

(6)

III. As SRT increased, the Yobs, value decreased. Since the biomass activity is reduced when the SRT increased, this has caused the reduction of the biomass yield. The results obtained from this experiment are in accordance with the ones reported by van Loosdrecht and Hence (1999). It is well known that the net sludge production in an activated sludge system decreases with increasing sludge age. The biokinetic parameters could give a good indication for the system performance. It can be used as a basis for the design and product optimization of a system reactor. The Yobs value of Stage IV to V, decreased from 0.269 to 0.217 mg VSS/mg COD as the SRT of those stages was increased from 70.1
23.9 to 72.5
23.3 d, respectively. Although Stage IV and V were operated with the same HRT, the ratio of anaerobic/aerobic reaction phase was different. It shows that when the ratio of anaerobic/aerobic time was increased, the Yobs decreased. The theoretical Y value is calculated using Eq. (6). It is expected that the theoretical Y value will be higher as compared to Yobs. The difference between Yobs and theoretical Y value is contributed by several factors including the endogenous metabolism of the microorganisms, predation, death and as well as lysis process. The theoretical Y value obtained in this study shows the same pattern as given by the Yobs. The results for the Yobs and theoretical Y value obtained

Table 5 e Kinetic coefficients of granular sludge at different stages of the experiment. Kinetic coefficients of granules sludge Observed specific biomass growth rate (moverall) (per day) Endogenous decay rate kd (per day) Observed biomass yield (Yobs) (mg VSS/mg COD) Theoretical biomass yield Y (mg VSS/mg COD)

Stage I

Stage II

Stage III

Stage IV

Stage V

Stage VI

0.036 0.0096 0.316 0.399

0.024 0.0086 0.298 0.395

0.013 0.0075 0.242 0.385

0.014 0.0075 0.269 0.410

0.014 0.0076 0.217 0.338

0.024 0.0060 0.412 0.515

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 1 1 e4 7 2 1

in this experiment are within the typical reported range of conventional activated sludge system (Al-Malack, 2006; Tchobanoglous et al., 2004).

3.4. Chemical oxygen demand and color removal performance The profile for COD concentration in the influent, effluent and removal performance for all six stages of experiment is given in Fig. 7. The biogranular system shows consistent COD degradation performance with 84.2
0.9% removals after about 50 days of start-up period (acclimatization phase). The overall performance was almost consistent despite the decrease in biomass concentration and OLR as mentioned earlier. When the OLR was increased from 0.6 kg COD/m3∙day to 0.8 kg COD/m3∙day on the 194th day of the experiment (Stage III to IV), the COD removal efficiency increased from about 84.4
0.4% at the end of Stage III (day 193) to 90.7
0.2% at the end Stage IV (day 236). Mohan et al. (2007) reported that the performance efficiency of the SBR system was found to be affected by the operating OLR. The SBR system operating at higher OLR resulted with a high substrate uptake rate at the end of the cycle period. This was also observed by Ong et al. (2005). An increase in the percentage of COD removal efficiency was also observed when the period of anaerobic phase was increased from 12 h to 18 h. The removal increased from 90.7
0.2% in Stage IV to 94.1
0.6% in Stage V. Pasukphun and Vinitnantharat (2003) claimed that the increase in the nonaeration phase in the SBR system would cause an alteration in the population of anaerobic microorganisms in the system which is expected to produce good COD and color removal for textile wastewater. However, according to Kapdan and Oztekin (2006), when the duration of anaerobic phase is too long, the contribution of aerobic reaction phase might be

Fig. 7 e Profile of COD removal performance of the reactor system at different stages of the experiment. (B) Influent COD; (-) Effluent COD, (:) COD removal. Stage I: anaerobic (2.8 h): aerobic (2.8 h); Stage II: anaerobic (5.8 h): aerobic (5.8 h); Stage III and Stage IV: anaerobic (11.8 h): aerobic (11.8 h); Stage V: anaerobic (17.8 h): aerobic (5.8 h); Stage V: anaerobic (5.8 h): aerobic (17.8 h).

4719

decreased. This is possibly due to the toxic effect of aromatic amines produced during dye degradation. Owing to the condition in the SBR system where different reaction phases occur in the same column, too long of anaerobic reaction periods will cause high accumulation of aromatic amine in the same compartment. High concentrations of aromatic amines may inhibit the activity of aerobic microorganisms during the aerobic phase. In this study, even though the anaerobic reaction phase was extended up to 18 h, there was no reduction in COD removal. This shows that there was no inhibition on the activity of aerobic microorganisms by the long accumulation of the byproduct produced from anaerobic degradation of the dye compound. It might be that the concentration of dye used during this experiment was not that high to produce enough concentration of the aromatic amines that may cause toxic effect toward the microorganisms within the biogranules. Furthermore, the biogranules might not be affected by the dyestuff degradation byproducts due to the structural form of the biogranules. The biogranules structure which consisted of EPS acts as a shield for microorganisms within the granules against any shock loading or toxic compound. At the final stage (Stage VI) of the experiment, a surge drop of COD removal efficiency was observed. As the aeration time was increased from 6 to 18 h, the COD removal reduced from 94.1
0.6% to 82.6
0.8%. The drop in the COD removal efficiency was possibly due to the increase in biomass loss into the effluent. The MLSS in Stage VI was 23.3
0.8 g/L as compared to 31.6
3.7 g/L observed in the previous stages. Color removal was observed to increase from 66.7
1.6% to 76.5
0.8% as the HRT increased from Stage I to Stage III. Increase in the HRT allows longer contact time between the granules and the wastewater resulting in better color removal. Furthermore, when the OLR was increased from 0.6 kg COD/ m3∙day (Stage III) to 0.8 kg COD/m3∙day (Stage IV), improvement in color removal from 76.5
0.8% to 83.1
1.4% was observed. This may be caused by the increase in the microbial population. Ong et al. (2005) reported that the percentage of color removal efficiency increased by 16% in anaerobic and 50% in aerobic SBR reactor systems when the OLR rate was increased from 2.66 to 5.32 g COD/L∙day. An increase from 82% to 90% of color removal efficiency was observed by Talarposhiti et al. (2001) when the COD loading was increased in a two-phase anaerobic packed bed reactor from 0.25 to 1 kg COD/m3∙day. Since more color removal took place in anaerobic condition (van der Zee et al., 2001; Dos Santos et al., 2007), the percentage of color removal was once again increased from Stage IV (83.1
1.4%) to Stage V (86.5
0.5%) when the anaerobic reaction phase was extended from 12 to 18 h of the 24 h reaction cycle. Improved decolorization process that occurs during the anaerobic stage enhances the overall wastewater biodegradation since more readily biodegradable substances could be degraded in the following aerobic treatment (Stolz, 2001). Fig. 8 shows the profile of the color removal performance. With respect to the mechanisms that are involved in color degradation, the addition of electronedonating substrate could considerably improve the decolorization reductive rate as reported by Dos Santos et al. (2005). Their studies using

4720

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 1 1 e4 7 2 1

Fig. 8 e Profile of color removal performance of the reactor system at different stages of the experiment. (A) Influent color, (-) Effluent color, (B) Color removal. (100 ADMI z 1 PteCo). Stage I: anaerobic (2.8 h): aerobic (2.8 h); Stage II: anaerobic (5.8 h): aerobic (5.8 h); Stage III and Stage IV: anaerobic (11.8 h): aerobic (11.8 h); Stage V: anaerobic (17.8 h): aerobic (5.8 h); Stage V: anaerobic (5.8 h): aerobic (17.8 h).

anaerobic and aerobic sequential wastewater treatment system indicated that the anaerobic stage was the main step for color degradation while the aerobic phase acted as the polishing step and enhancement in COD removal. Higher initial COD concentration did not improve color removal but caused deterioration in COD removal in the anaerobicaerobic SBR system (Kapdan and Oztekin, 2006). Pasukphun and Vinitnantharat (2003) reported that the duration of the anaerobic phase should be long enough to obtain better COD and color removal. Increase in the HRT would provide enough time for the degradation of the organics and inter-metabolites of textile wastewater in anaerobic or/and anaerobic/aerobic systems (Isik and Sponza, 2008). From this study, it shows that having longer anaerobic (18 h) and shorter aerobic (6 h) reaction phase resulted with the highest removal for color and slight improvement in the efficiency of COD removal.

3.5.

Conclusions

Several conclusions can be derived from the study. They are as follows:  The granular biomass concentration in the reactor reduces as the HRT increases which is mainly due to the reduction in the OLR. However, with HRT of 24 h, the biomass concentration slightly improved when the anaerobic reaction phase was longer than aerobic reaction. The ratio of MLVSS/MLSS reduces when the anaerobic/aerobic reaction phase was set with 17.8/5.8 h which may be due to the increased accumulation of inert particles within the granules. Although with increase in the HRT, the concentration of granular biomass can be improved with the increase in the anaerobic reaction time and reduce in the aerobic reaction time.

 The size and the SVI of the granular sludge reduced as the HRT of the system increased due to increase in the aeration time that resulted with the disintegration of the granules. Prolonged starvation condition may cause instability of the granular structure that lead to disruption of the granules. However the size and the SVI value were improved with the increase in the OLR and anaerobic reaction time and reduce in the aerobic phases.  Increase in the HRT resulted with an increase in the SRT. Since the SRT is inversely related to the moverall, increase in the HRT will cause a reduction in the moverall. Increase in the HRT has caused a reduction in the bioactivity of the granular sludge shown by the reduction of the moverall, Yobs and Y values. A slight increase in the SRT was observed with increase in the anaerobic/aerobic time ratio. This has caused a reduction in the Yobs and Y values but the moverall is considered constant. The kd is also considered constant throughout the experiment.  The percentage of COD removal in this study was not likely affected by the increase in the HRT which was mainly due to the decrease in the granular biomass and OLR. However, the COD and color removal was improved with prolong on the anaerobic reaction phase.

Acknowledgments The authors wish to thank the Ministry of Science, Technology and Innovation (MOSTI), Ministry of High Education (MOHE) and Universiti Teknologi Malaysia for the financial supports of this research (Grants No.: 79137, 78211 and 75221).

references

Al-Malack, M.H., 2006. Determination of biokinetic coefficients of an immersed membrane bioreactor. Journal of Membrane Science 271, 47e58. American Public Health Association, 2005. Standard Methods for the Examination of Water and Wastewater, twenty first ed. American Public Health Association, Washington, DC, USA. Beun, J.J., Hendricks, A., van Loosdrecht, M.C.M., Morgenroth, E., Wilderer, P.A., Heijnen, J.J., 1999. Aerobic granulationin a sequencing batch reactor. Water Research 33, 2283e2290. Buitron, G., Quezada, M., Moreno, G., 2004. Aerobic degradation of the azo dye acid red 151 in a sequencing batch biofilter. Bioresource Technology 92, 143e149. Chen, Y., Jiang, W., Liang, D.T., 2008. Biodegradation and kinetics of aerobic granules under high organic loading rates in sequencing batch reactor. Apply Microbiology Biotechnology 79, 301e308. Correia, V.M., Stephenson, T., Judd, S.J., 1994. Characterisation of textile wastewaters e A review. Environmental Technology 15, 917e928. Dos Santos, A.B., Madrid, M.P., Stams, A.J.M., van Lier, J.B., Cervantes, F. J., 2005. Azo dye reduction by mesophilic and thermophilic anaerobic consortia. Biotechnology Progress 21, 1140e1145. Dos Santos, A.B., Cervantes, F.J., van Lier, J.B., 2007. Review paper on current technologies for decolourisation of textile wastewaters: perspectives for anaerobic biotechnology. Bioresource Technology 98, 2369e2385.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 1 1 e4 7 2 1

Foster, C.F., 1991. Anaerobic upflow sludge blanket reactors: aspects of their microbiology and their chemistry. Journal of Biotechnology 17, 221e231. Grady, C.P.L., Daigger, G.T., Lim, H.C., 1999. Biological Wastewater Treatment, second ed. Marcel Dekker, New York. Ibrahim, Z., Amin, M.F.M., Yahya, A., Aris, A., Muda, K., 2010. Characteristics of the developed granules containing selected decolourising bacteria for the degradation of textile wastewater. Water Science Technology 61 (5), 1279e1288. Isik, M., Sponza, D.T., 2004a. A batch kinetic study on decolorization and inhibition of reactive black 5 and direct brown 2 in an anaerobic mixed culture. Chemosphere 55, 119e128. Isik, M., Sponza, D.T., 2004b. Anaerobic/aerobic sequential treatment of a cotton textile mill wastewater. Journal of Chemical Technology and Biotechnology 79, 1268e1274. Isik, M., Sponza, D.T., 2005. Effects of alkalinity and co-substrate on the performance of an up-flow anaerobic sludge blanket (UASB) reactor through decolorization of Congo Red azo dye. Bioresource Technology 96 (5), 633e643. Isik, M., Sponza, D.T., 2008. Anaerobic/aerobic treatment of a simulated textile wastewater. Separation and Purification Technology 60, 64e72. Kapdan, I.K., Oztekin, R., 2006. The effect of hydraulic residence time and initial COD concentration on color and COD removal performance of the anaerobiceaerobic SBR system. Journal of Hazardous Materials 136, 896e901. Li, Y., Xi, D.L., 2004. Decolorization and biodegradation of dye wastewaters by a facultative-aerobic process. International Journal of Environmental Sciences 11 (6), 372e377. Li, Z.H., Kuba, T., Kusuda, T., 2006. The influence of starvation phase on the properties and the development of aerobic granules. Enzyme and Microbial Technology 38, 670e674. Liu, Y.Q., Tay, J.H., 2007. Cultivation of aerobic granules in a bubble column and an airlift reactor with divided draft tubes at low aeration rate. Biochemical Engineering Journal 34 (1), 1e7. Mohan, S.V., Rao, N.C., Sarma, P.N., 2007. Simulated acid azo dye (acid black 210) wastewater treatment by periodic discontinuous batch mode operation under anoxiceaerobiceanoxic microenvironment conditions. Ecological Engineering 3 (1), 242e250. Muda, K., Aris, A., Salim, M.R., Ibrahim, Z., Yahya, A., van Loosdrecht, M.C.M., Ahamd, A., Nawahwi, M.Z., 2010. Development of granular sludge for textile wastewater treatment. Water Research 44, 4341e4350. Ong, S.A., Toorisaka, E., Hirata, M., Hano, T., 2005. Treatment of azo dye orange II in aerobic and anaerobic-SBR systems. Process Biochemistry 40, 2907e2914. Ong, S.A., Toorisaka, E., Hirata, M., Hano, T., 2008. Combination of adsorption and biodegradation processes for textile effluent treatment using a granular activated carbon-biofilm configured packed column system. Journal of Environmental Sciences 20, 952e956. Panswad, T., Iamsamer, K., Anotai, J., 2001. Decolorisation of azoreactive dye by polyphospate and glycogen-accumulating organisms in an anaerobic-aerobic sequencing batch reactor. Bioresource Technology 76, 151e159. Pasukphun, N., Vinitnantharat, S., 2003. Degradation of organic substances and reactive dye in an immobilized-cell sequencing batch reactor operation on simulated textile wastewater. Journal of Environmental Science and Health, Part A: Toxic/Hazardous Substances and Environmental Engineering 38 (10), 2019e2028. 1532e4117.

4721

Sirianuntapiboon, S., Srisornsak, P., 2007. Removal of disperse dyes from textile wastewater using bio-sludge. Bioresource Technology 98, 1057e1066. Smolders, G.J.F., Klop, J., van Loosdrecht, M.C.M., Heijnen, J.J., 1995. A metabolic model of the biological phosphorus removal process. Effect of the sludge retention time. Biotechnology and Bioengineering 48, 222e233. Somasiri, W., Li, X.F., Ruan, W.Q., Jian, C., 2008. Evaluation of the efficacy of upflow anaerobic sludge blanket reactor in removal of colour and reduction of COD in real textile wastewater. Bioresource Technology 99, 3692e3699. Stolz, A., 2001. Basic and applied aspects in the microbial degradation of azo dyes. Applied Microbiology and Biotechnology 56, 69e80. Talarposhti, A.M., Donnelly, T., Anderson, G.K., 2001. Colour removal from a simulated dye wastewater using a twophase anaerobic packed bed reactor. Water Research 35, 425e432. Tan, N.C.G., van Leeuwen, A., van Voorthuizen, E.M., Peter Slenders, P., Prenafeta-Boldu, F.X., Temmink, H., Lettinga, G., Field, J.A., 2005. Fate and biodegradability of sulfonated aromatic amines. Biodegradation 16, 527e537. Tchobanoglous, G., Burton, F.L., Stensel, H.D., 2004. Wastewater Engineering: Treatment, Disposal and Reuse, fourth ed. McGraw-Hill Companies Inc, New York. van Loosdrecht, M.C.M., Hence, M., 1999. Maintenance, endogenous respiration, lysis, decay and predation. Water Science Technology 39 (1), 107e117. Van der Zee, F.P., Villaverde, S., 2005. Combined anaerobicaerobic treatment of azo dyes e A short review of bioreactor studies. Water Research 39, 1425e1440. van der Zee, F.P., Lettinga, G., Field, J.A., 2001. Azo dye decolourisation by anaerobic granular sludge. Chemosphere 44 (5), 1169e1176. van der Zee, F.P., Bisschops, I.A.E., Lettings, G., Field, J.A., 2003. Activated carbon as an electron acceptor and redox mediator during the anaerobic biotransformation of azo dyes. Environmental Science and Technology 37, 402e408. Wang, Q., Dua, G., Chena, J., 2004. Aerobic granular sludge cultivated under the selective pressure as a driving force. Process Biochemistry 39, 557e563. Wijffels, R.H., Tramper, J., 1995. Nitrification by immobilized cells. Enzyme and Microbial Technology 17, 482e492. Zheng, Y.M., Yu, H.Q., Sheng, G.P., 2005. Physical and chemical characteristics of granular activated sludge from a sequencing batch airlift reactor. Process Biochemistry 40, 645e650.

Abbreviations q: Solid retention time (d) Xvss: Volatile solid concentration in the reactor system (g VSS/L) Vr: Working volume of the SBR system (L) Xd: Biomass concentration of manually discharged mixture (g VSS/L) Vd: Manually discharge mixture volume (L) Xe: Effluent volatile solid concentration (g VSS/L) Ve: Effluent volume of the SBR operating cycle (L) tc: Cycle time of the SBR operation (d) XVSS1: Volatile solid concentration at the beginning of cycle operation in SBR reactor (g VSS/L) XVSS2: Volatile solid concentration at the end of cycle operation in SBR reactor (g VSS/L)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 2 2 e4 7 3 6

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Role of biodegradation in the removal of pharmaceutically active compounds with different bulk organic matter characteristics through managed aquifer recharge: Batch and column studies Sung Kyu Maeng a,*, Saroj K. Sharma b, Chol D.T. Abel b, Aleksandra Magic-Knezev c, Gary L. Amy d a

Water Research Center, Korea Institute of Science and Technology, P.O. Box. 131, Cheongryang, Seoul 130-650, South Korea UNESCOeIHE Institute for Water Education, P.O. Box 3015, 2601 DA Delft, The Netherlands c Het Waterlaboratorium, J.W. Lucasweg 2, 2031 BE Haarlem, The Netherlands d King Abdullah University of Science and Technology, Thuwal 23955-6000, Saudi Arabia b

article info

abstract

Article history:

Natural water treatment systems such as bank filtration have been recognized as providing

Received 4 March 2011

effective barriers in the multi-barrier approach for attenuation of organic micropollutants

Received in revised form

for safe drinking water supply. In this study, the role of biodegradation in the removal of

26 May 2011

selected pharmaceutically active compounds (PhACs) during soil passage was investigated.

Accepted 28 May 2011

Batch studies were conducted to investigate the removal of 13 selected PhACs from

Available online 7 June 2011

different water sources with respect to different sources of biodegradable organic matter. Neutral PhACs (phenacetine, paracetamol, and caffeine) and acidic PhACs (ibuprofen,

Keywords:

fenoprofen, bezafibrate, and naproxen) were removed with efficiencies greater than 88%

Bank filtration

from different organic matter water matrices during batch studies (hydraulic retention

Bulk organic matter

time (HRT): 60 days). Column experiments were then performed to differentiate between

Managed aquifer recharge

biodegradation and sorption with regard to the removal of selected PhACs. In column

Organic micropollutants

studies, removal efficiencies of acidic PhACs (e.g., analgesics) decreased under conditions

Pharmaceutically active compounds

of limited biodegradable carbon. The removal efficiencies of acidic PhACs were found to be less than 21% under abiotic conditions. These observations were attributed to sorption under abiotic conditions established by a biocide (20 mM sodium azide), which suppresses microbial activity/biodegradation. However, under biotic conditions, the removal efficiencies of these acidic PhACs were found to be greater than 59%. This is mainly attributed to biodegradation. Moreover, the average removal efficiencies of hydrophilic (polar) neutral PhACs (paracetamol, pentoxifylline, and caffeine) with low octanol/water partition coefficients (log Kow less than 1) were low (11%) under abiotic conditions. However, under biotic conditions, removal efficiencies of the neutral PhACs were greater than 98%. In contrast, carbamazepine persisted and was not easily removed under either biotic or abiotic conditions. This study indicates that biodegradation represents an important mechanism for the removal of PhACs during soil passage. ª 2011 Elsevier Ltd. All rights reserved.

* Corresponding author. Tel.: þ82 2 958 6769; fax: þ82 2 958 6854. E-mail address: [email protected] (S.K. Maeng). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.043

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 2 2 e4 7 3 6

1.

Introduction

During recent years, there has been growing concern over the increased detection of organic micropollutants (OMPs) such as pharmaceutically active compounds (PhACs), endocrine disrupting compounds (EDCs), and personal care products (PCPs) in drinking water and the aquatic environment (Focazio et al., 2008; Ku¨mmerer, 2009; Calisto and Esteves, 2009). The growing use of PhACs, EDCs, and PCPs for human and veterinary purposes has increased the frequency of detection of these compounds in water supplies and environment (Heberer, 2002; Jjemba, 2006). Currently, the total consumption of PhACs, EDCs, and PCPs in the world is not known because many of these compounds vary significantly among different nations with respect to applications and consumption (Ku¨mmerer, 2008). The development of new analytical procedures and instruments has enabled analysts to quantify the levels of environmental contamination even at very low concentrations. This has led to an increasing number of OMPs being detected in the aquatic environment and drinking water (Snyder et al., 2004). Generally, combinations of PhACs, EDCs, and PCPs are present in the environment. Various combinations of these compounds might have different and possibly synergistic impacts on public health and/or aquatic life as compared to the presence of only a single compound (Ku¨mmerer, 2009). The main route for the release of PhACs and metabolites (i.e., transformation products formed during ingestion) into the aquatic environment is through excretion in association with both urine and feces discharged into sewage treatment plants (Ying et al., 2009). Non-point sources such as overland flow (i.e., run-off) by heavy rainfall or land drainage in agriculture can also deliver PhACs such as veterinary medicines to surface water or groundwater (Boxall et al., 2004). However, little is known about the fate of PhACs and metabolites during drinking water treatment processes and in the aquatic environment (Mompelat et al., 2009). There is a possibility that potentially harmful pollutants such as EDCs are present in drinking water sources. This possibility has resulted in increased research on PhACs using advanced technologies in order to ensure reliable supplies of safe drinking water (Kim et al., 2007; Madden et al., 2009; Mechlinski and Heberer, 2005; Mompelat et al., 2009; Yangali-Quintanilla et al., 2010a). The removal of PhACs from water by advanced water treatment technologies is relatively costly and contributes to a high unit cost in the water treatment process. Managed aquifer recharge (MAR) treatment processes such as riverbank filtration (RBF), lake bank filtration (LBF), and artificial recharge (AR) are robust and cost-effective treatments, which provide a degree of OMP removal (Maeng et al., 2011a). Previous field studies have shown that RBF, LBF, and AR are effective for OMP removal (Gru¨nheid and Jekel, 2005; Heberer et al., 2004; Maeng et al., 2010; Massmann et al., 2008; Mechlinski and Heberer, 2005; Schmidt et al., 2007). The framework for assessment of OMPs removal using guidelines and quantitative structure activity relationship (QSAR) model was tried to provide more understanding of the behavior of OMPs during MAR (Maeng et al., 2011b) PhACs are removed via various mechanisms during soil passage. The most important mechanism is biodegradation.

4723

The biodegradability of PhACs is an important characteristic that can be used to assess their fate under environmental conditions and the risks related to their presence in the environment (Cunningham, 2008). Biodegradation of PhACs is the most desirable removal mechanism because it is a sustainable process and potentially forms end products consisting of inorganic compounds (i.e., mineralization) (Howard, 2000). Another important mechanism involved in the removal of PhACs during MAR is sorption. This mechanism has an impact on the bioavailability of PhACs. The octanol/water partition coefficient (Kow) is often used to assess the sorption potential and distribution behavior of OMPs in the aquatic environment. However, Kow may not accurately describe the distribution behavior between soil and water for some acidic PhACs because of electrostatic interactions. Many non-steroidal antiinflammatory drugs (NSAIDs) and lipid regulators are acidic PhACs and remain in ionized forms at environmentally relevant pH levels (Cunningham, 2008). Therefore, the acid dissociation constant (pKa) of acidic PhACs and pH of the environment are important to understand the fate (ionized and unionized forms) of acidic PhACs during soil passage. Previous studies have attempted to examine the removal of PhACs by sorption, photolysis, and biodegradation in wastewater, secondary effluent, and surface water (Heberer  et al., 2009; Ternes, 1998; Ternes et al., 2004; Radjenovic et al., 2004; Yangali-Quintanilla et al., 2010b; Ying et al., 2009). However, few studies have investigated the effects of the biodegradable fraction of natural organic matter (NOM) on PhAC removal by using a suite of innovative analytical tools (e.g., fluorescence excitationeemission matrices [FeEEM] and liquid chromatography with online organic carbon detection [LC-OCD]). In addition, even studies comparing sorption and biodegradation in the removal of PhACs during soil passage are few. Therefore, the objective of this study was to investigate the role of biodegradation in the removal of 13 selected PhACs of different classes from different water matrices, i.e., with different organic matter characteristics, during soil passage.

2.

Materials and methods

2.1.

Chemicals

A total of 13 selected PhACs (gemfibrozil, diclofenac, bezafibrate, ibuprofen, fenoprofen, naproxen, ketoprofen, clofibric acid, carbamazepine, phenacetine, paracetamol, pentoxifylline, and caffeine) were used to prepare stock solutions. Working solutions were prepared from the stock solutions and spiked into different experimental mixtures. Milli-Q water (Advantage A10; Millipore) was used to prepare 100 mg/L stock solutions. All PhACs under investigation were of analytical grade and purchased from SigmaeAldrich, Germany. The physicochemical properties of the selected compounds are shown in Table 1. For acidic pharmaceuticals, log DpH¼8 (distribution coefficient) was used as an indicator of their hydrophobicity. According to Cunningham (2008), a chemical with a log D value of less than 1 (a hydrophilic compound) may not sorb onto organic matter or bioconcentrate at all, in

4724

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 2 2 e4 7 3 6

Table 1 e List of PhACs studied and their properties. Name Gemfibrozil Diclofenac Bezafibrate Ibuprofen Fenoprofen Naproxen Ketoprofen Clofibric acid Carbamazepine Phenacetine Paracetamol Pentoxifylline Caffeine

MW (g/mol)

pKa

log Kowa

log Db (pH ¼ 8)

Classification @ pH ¼ 8c

250.3 296.2 361.8 206.3 242.3 230.3 254.3 214.6 236.3 179.2 151.2 278.3 194.2

4.7 4.2 3.6 4.9 4.5 4.2 4.5 3.2 n.a. n.a. n.a. n.a. n.a.

4.77 4.51 4.25 3.97 3.9 3.18 3.12 2.88 2.45 1.67 0.27 0.29 0.07

2.22 1.59 0.69 1.44 1.11 0.05 0.41 1.08 e e e e e

Ionic Ionic HydrophiliceIonic Ionic Ionic HydrophiliceIonic HydrophiliceIonic HydrophiliceIonic Neutral Neutral Hydrophilic-Neutral Hydrophilic-Neutral Hydrophilic-Neutral

a KOWWIN v.1.67 (US EPA, 2009). b ADME/Tox WEB software (http://www.acdlabs.com). c For acidic pharmaceuticals: hydrophobic: log D > 3, hydrophilic: log D < 1 at pH 8; for neutral pharmaceuticals: hydrophobic: log Kow > 3, hydrophilic: Kow < 1.

contrast to a chemical with a log D value of equal to or greater than 3 (a hydrophobic compound), which may significantly bioconcentrate or sorb onto/into soil organic matter. For neutral pharmaceuticals, the log Kow value was used in manner as the log D value for acidic pharmaceuticals. Easily biodegradable synthetic organic matter (SOM) was used as an external carbon source for a batch experiment. It was prepared from aldehydes (200 mg/L formaldehyde and 60 mg/L glyoxal) and carboxylic acids (800 mg/L sodium acetate and 600 mg/L sodium formate) and then used for column studies. The composition of the SOM was based on the formation of organic by-products from ozonation (Urfer and Huck, 2001). However, the concentrations of organic by-products used in this study were higher than the concentrations reported for a full-scale drinking water treatment plant using ozonation (Urfer and Huck, 2001). In this manner, the effects of organic byproducts from ozonation on the removal of PhACs can be estimated.

2.2.

Batch experimental setups

Batch experiments were performed in 1-L glass bottles containing silica sand (grain size, 0.8e1.25 mm) to assess the attenuation of selected PhACs in different sources of water. Simulations of long residence time included 30 days under oxic conditions followed by 30 days of anoxic conditions during the MAR process. Five batch reactors were used together in triplicate for 5 different samples: (1) water from River Meuse, The Netherlands (MR); (2) River Meuse water spiked with synthetic organic matter (SOM) as an additional carbon source (MR þ SOM); (3) secondary effluent (SE) from a wastewater treatment plant in Hoek van Holland, The Netherlands; (4) water from an experimental container used for cultivation of the common reed Phragmites australis (CR); and (5) nonchlorinated tap water (NCTW), Delft, The Netherlands (assimilable organic carbon, approximately 10 mg/L). For the CR sample, P. australis was obtained from the shore of a lake located in Delft, the Netherlands and transplanted into a custom-designed container. The water level of the container was maintained at the container surface. The plants

were incubated under constant temperature (16  C) for a 14-h light period. The CR sample was collected from the valve located at the bottom of the container. The purpose of investigating water in the presence of CR for batch reactors was to simulate water with high levels of humic substances. The CR sample exhibited high levels of aromaticity as indicated by SUVA analysis and contained organic matter, which originated from soil. Initially, batch reactors fed with the MR, MR þ SOM, SE, and CR samples were bioacclimated with their respective influents until the reactors stabilized (acclimated) with respect to reduction of dissolved organic carbon (DOC). Each reactor was then spiked with selected PhACs in concentrations ranging from 1.8 mg/L to 5.4 mg/L. All batch reactors were placed on a shaker table and rotated at 100 rpm under oxic conditions (30 days) followed by anoxic conditions (30 days) for hydraulic retention time (HRT) of 60 days. Generally, redox processes during MAR produce oxic conditions followed by anoxic conditions along the flow transect. Therefore, after the first 30 days under oxic conditions, nitrogen gas was used to remove dissolved oxygen (DO) from the batch reactors using diffusers, which were designed for a high-performance liquid chromatography (HPLC) system. The diffusers lowered the DO to less than 0.2 mg/L in the reactors. Nitrogen gas was very gently introduced to prevent disturbances in the batch reactors. After 60 days, nitrate concentrations were not completely removed from the reactors (data not shown). There were no electron acceptors used other than oxygen and nitrate. A blank experiment was also performed with approximately the same concentration of PhACs used for batch reactors in Milli-Q water without sand to investigate the loss of target PhACs during the entire experimental period. This would indicate the existence of other mechanisms for removal of measurable PhACs such as sorption to the surface of glassware. The average loss of PhACs was determined to be less than 7%.

2.3.

Column experimental setups

Double-walled soil columns (SC) (XK50/30; Amersham Pharmacia Biotech, Sweden) with an inner diameter of 50 mm and

4725

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 2 2 e4 7 3 6

a length of 300 mm were used. The inner part of the columns was packed with silica sand (0.8e1.25 mm), while the outer part was used to control the ambient temperature by flowing water from a chiller. All SCs were kept in a dark room to minimize the potential effects of photodegradation and run under oxic conditions supplying air into glass bottles containing influents (i.e., feed water). Four sand columns (SC1, SC2, SC3, and SC4) were deployed to assess the role of biodegradation in the removal of PhACs, and the results were compared with those of the sorption experiments. All 4 columns were connected to a chiller, and operating temperatures were maintained between 16  C and 17  C. The empty bed contact time (EBCT) for the SC setup was 17 h. This time was determined by a tracer study using NaCl with a hydraulic loading rate of 0.64 m/day. Concentrations of selected PhACs in influents prepared for columns were between 1.2 and 8.1 mg/L. Before introducing selected PhACs, SC1, and SC2 were bioacclimated with MR for 2 months. As in the case of the above-described batch experiment, the acclimation process was continued until both columns (SC1 and SC2) were stabilized with respect to reduction of DOC. After the acclimation period, NCTW and MR with PhACs served as feed to SC1 and the SC2, respectively. SC1 was used to assess the biodegradation of selected PhACs under low microbial conditions in NCTW, which contains low levels of biodegradable organic matter. If the few selected PhACs are removed by cometabolism, the limited biodegradable carbon in NCTW would diminish their removal efficiencies by SC1. In contrast to the acclimation period of 60 days for SC1 and SC2, SC3 only had an acclimation period of 10 days in order to have less active biomass associated with the sand. SC4 was packed with fresh sand. Demineralized water (DW) containing sodium azide (NaN3) at a concentration of 20 mM was used to maintain the column under abiotic conditions and to investigate the removal of PhACs by sorption. Sodium azide is a biocide, which is frequently used to suppress microbial activity in columns used for abiotic experiments (i.e., sorption experiments) (Chen et al., 2008; Liang et al., 2006). Adenosine triphosphate (ATP) and DOC concentrations were measured to verify that the SC4 column was maintained under abiotic conditions. Water samples were taken from influent and effluent of SCs to investigate the characteristics of dissolved organic matter and the removal of PhACs during soil passage. After collecting water samples, all columns were dismantled to collect sand samples investigating active microbial biomass associated with sand.

2.4.

Analytical methods

Dissolved organic matter (DOM) in all samples was characterized within 3 days after samples were collected and stored at 5  C after filtration through a 0.45-mm filter (Whatman, Dassel, Germany) to prevent biodegradation of DOM. The concentration of DOM was determined as dissolved organic carbon (DOC) by a total organic carbon analyzer (TOC-VCPN (TN), Shimadzu, Japan). The characteristics of DOM were elucidated by various analytical methods including FeEEM, LC-OCD (DOCeLabor Dr. Huber, Karlsruhe, Germany) and specific ultraviolet absorbance (SUVA). In FeEEM, all samples were adjusted to 1 mg/L of DOC by diluting the samples with

Milli-Q water. The samples were then measured using a FluoroMax-3 spectrofluorometer (HORIBA Jobin Yvon, Edison, NJ, USA). FeEEM spectra were obtained at excitation wavelengths of 240e450 nm at 10-nm intervals. The emission wavelengths were between 290 nm and 500 nm at 2-nm intervals. The water Raman peak at 348 nm was used to confirm the stability of the spectrofluorometer, and FeEEM spectra were corrected using blank subtraction. Table 2 shows the characteristics of organic matter expressed as regions determined by distinct wavelengths of excitation and emission (Coble, 1996; Henderson et al., 2009; Leenheer and Croue, 2003). The LC-OCD system uses a liquid chromatography method to determine the molecular weight (MW) distribution and classification of DOM as biopolymers, humic substances, building blocks, low-MW acids, and neutral compounds. More information about this system has been provided by Huber and Frimmel (1992). UV absorbance was measured at 254 nm by a UV/Vis spectrophotometer (UVe2501PC, Shimadzu, Japan), and SUVA was calculated by dividing UV254 absorbance by the DOC concentration to indicate the relative aromaticity of organic matter. More information on characterization tools for organic matter using DOC, FeEEM, LC-OCD and SUVA are provided in the recent study (Sharma et al., 2011). ATP was used to determine the active microbial biomass (AMB) associated with sand in this study (Oades and Jenkinson, 1979). Wet sand samples of 1e2 g, collected from batch and column studies, were suspended in autoclaved tap water (50 mL). High-energy sonication at 40-W power was applied to detach the biomass (Branson We250D Sonifier with a D¼ 5-mm microtip). The biomass concentration was determined as the concentration of ATP in the suspension obtained from the sonication process. A single 2-min sonication treatment was adequate to obtain more than 90% of the attached biomass represented as ATP (data not shown). A detailed description of methods and materials used for ATP extraction is provided by Magic-Knezev and van der Kooij (2004). For PhAC measurements, autotrace SPE workstations from Caliper Life Sciences GmbH (Ru¨sselsheim, Germany) were used for solid phase extraction. High-performance liquid chromatographyelectrospray ionization tandem mass spectrometry analyses were performed using an HPLC system 1100, Series II

Table 2 e Characteristics of organic matter using 3 key fluorescence peaks in FeEEM. Excitation Emission (nm) (nm) P1 250e260 237e260

380e480 400e500

P2 330e350 300e370

420e480 400e500

P3 270e280 275 280

320e350 340 350

Description Primary humic-like substances

References

(Leenheer and Croue, 2003) (Coble, 1996) Secondary humic-like (Leenheer and substances Croue, 2003) (Coble, 1996) Tryptophan-like peak (Leenheer and (protein-like substances) Croue, 2003) (Coble, 1996) (Henderson et al., 2009)

4726

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 2 2 e4 7 3 6

from Agilent Technologies (Waldbronn, Germany) equipped with an API 2000 triple quadrupole mass spectrometer (PE Sciex, Langen, Germany) with electrospray ionization (ESI) under atmospheric pressure. PhACs measurements of samples were conducted in TZW: DVGW-Technologiezentrum Wasser (Karlsruhe, Germany). Limits of quantification (LOQ) for PhACs were 10 ng/L in undiluted water samples, and samples were diluted 5-fold for analysis. In this study, LOQ and limits of detection (LOD) for selected PhACs were 50 ng/L and 20 ng/L, respectively. More information about the PhAC detection method is provided in the previous study (Sacher et al., 2008).

3.

Results and discussion

3.1. Removal of selected PhACs from different water sources: batch experiments 3.1.1.

Dissolved organic matter characteristics

Five batch reactors were used to investigate attenuation of PhACs using different water samples (MR, MR þ SE (1:1), SE, NCTW, and CR) spiked with PhACs. Samples were taken from each batch reactor at the beginning and end of the batch study (HRT: 60 days), and the pH, DOC, UV254 nm, and SUVA were determined. Table 3 shows the organic matter characteristics of the influents and effluents in the batch studies. BDOC60-day was defined as the change in DOC within 60 days of HRT. The SE sample exhibited the highest BDOC60-day value of 6.3  0.5 mg/L followed by the CR (4.6  1.9 mg/L), MR þ SOM (2.8  0.4 mg/L), and MR (1.5  0.2 mg/L) samples. BDOC60-day values for batch reactors were higher than the BDOC5-day values measured during the acclimation period because the slowly biodegradable organic matter requires longer residence times for degradation (data not shown). SUVA values were 3.6, 3.4, and 3.1 L mg1 m1 for the CR, SE, and MR samples, respectively. These values indicate the presence of significant levels of aromatic compounds in the organic matter of each sample. The low SUVA value observed for the MR þ SOM sample (2.4 L mg1 m1) is attributed to the aliphatic nature of SOM added into the MR. The 60-day HRT period in bioactive sand produced an increase of SUVA (CR: 3.9 L mg1 m1; SE: 5.0 L mg1 m1; MR: 4.3 L mg1 m1;

MR þ SOM: 4.7 L mg1 m1) due to biodegradation of the aliphatic structure of DOM in the samples. Previous studies have demonstrated similar results (Cha et al., 2004; Xue et al., 2009). Cha et al. (2004) showed that SUVA values increase in effluents from a column filled with natural river soil as a result of biodegradation of aliphatic organic matter. The dominant fraction of organic matter in CR appeared to be the high content of humic substances, which caused only a small increase (0.3 L mg1 m1) in the SUVA value during the 60-day HRT period. FeEEM spectra of influent and effluent samples were measured to probe the characteristics of DOM during the batch studies. As previously mentioned, a given FeEEM spectrum shows fluorescence intensity peaks at known wavelengths such that a fluorescence intensity peak at higher values of excitation and emission wavelengths indicates the presence of humic-like substances, whereas a fluorescence intensity peak at lower values of excitation and emission wavelengths indicates the presence of protein-like substances (Amy and Drewes, 2007). Three peaks were identified: a primary humic-like peak, P1 (lex/em ¼ 250e270/420e440 nm), a secondary humic-like peak, P2 (lex/em ¼ 330e340/ 420e440 nm), and a protein-like peak, P3 (lex/em ¼ 270e280/ 320e340 nm). Based on these peaks, the differences between fluorescence intensities at time ¼ 0 day and time ¼ 60 days were calculated and compared to characterize the transformation of DOM. Differential FeEEM spectra from batch reactors are shown in Fig. 1. After the 60-day HRT period, there were substantial decreases in P3 for the MR, MR þ SOM, and SE samples. According to a previous study by Hudson et al. (2008), the tryptophan-like peak (protein-like peak, lex/em ¼ 275/ 340 nm) was found to be an accurate surrogate for the presence and relative proportions of bioavailable organic matter. In this study, this tryptophan-like peak was found at lex/ em ¼ 270e280/320e340 nm. In contrast, the fluorescence intensity of P1 and P2 increased for the MR, MR þ SOM, and SE samples but did not increase for the CR sample. Table 4 exhibited increases in P1 and P2 by 27% and 34%, respectively, while 52% of P3 was decreased from the MR sample. In addition, 36% of the DOC was removed from the MR sample. For the MR þ SOM sample, fluorescence intensity increased by 63% and 64% for humic-like peaks P1 and P2, respectively. On the other hand, the removal of the protein-

Table 3 e Characteristics of influent and effluent for batch reactors.

NCTW MR MR þ SOM CR SE

Influent Effluent Influent Effluent Influent Effluent Influent Effluent Influent Effluent

a DOC60-days - DOC0-days.

pH

DOC (mg/L)

BDOC60-day (mg/L)a

% BDOC60-day relative to DOC

SUVA (L mg1 m1)

7.6  0.1 8.3  0.1 7.7  0.1 8.3  0.1 7.8  0.1 8.4  0.1 7.9  0.1 8.5  0.1 7.6  0.1 8.6  0.1

2.0  0.2 1.5  0.1 3.6  0.1 2.1  0.2 5.3  0.1 2.5  0.4 14.6  0.5 10.1  1.4 14.8  1.6 8.5  1.1

0.5  0.2

25

1.5  0.2

42

2.8  0.4

53

4.6  1.9

31

6.3  0.5

43

e e 3.1  0.1 4.3  0.7 2.4  0.1 4.7  0.9 3.6  0.2 3.9  0.6 3.4  0.4 5.0  0.8

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 2 2 e4 7 3 6

4727

Fig. 1 e Differential FeEEM spectra MR(a,b), MR D SOM(c,d), SE(e,f), CR(g,h). Left: effluent spectrum is subtracted from its feed water spectrum (degradation), Right: feed water spectrum is subtracted from its effluent spectrum (formation).

like peak P3 was equal to that of the MR sample (52%), and DOC removal for the MR þ SOM sample was 44%. For the SE sample, the fluorescence intensity for P1 and P2 increased by 32% and 25%, respectively (Table 4). Again, the protein-like peak P3 was reduced by 50%; 34% of DOC was removed. Relatively, small increases of P1 (3%) and P2 (6%) are observed for the CR sample. Moreover, 18% of DOC was removed, and P3 was lowered by 48%. These results indicate that the fluorescence intensity values for P1 and P2 are not correlated with the reduction in DOC concentrations over the 60-day HRT period. In fact, this result provides an indication of the biotransformation of DOM, and the degree of DOM biotransformation in the CR sample was relatively low compared to that of the MR, MR þ SOM, and SE samples. According to Saadi et al. (2006), new fluorescing materials associated with DOM (fluorescing DOM) are formed in effluent-inoculated samples during long-term biodegradation over 60 days. It was demonstrated that certain organic components in DOM were biotransformed into fluorescing DOM. Such increases in fluorescence were also observed in our study with respect to P1 and P2 (humic-like substances) for the MR, MR þ SOM, and SE samples during the 60-day HRT

period. According to our previous column studies, the fluorescence intensity of P1 and P2 did not increase during a short HRT (5 days; Maeng et al., 2008). Therefore, the formation of fluorescing DOM was minimal during the short HRT period, and this fraction was small relative to the degradation of fluorescing DOM. However, during the long 60-day HRT period, slowly biodegradable compounds were found to contribute significantly to the formation of fluorescing DOM at a greater extent than the degradation of fluorescing DOM. Moreover, it was found that formation of fluorescing DOM is limited by the HRT period that allows slowly biodegradable organic matter to be degraded but is also limited by the DOM that can be easily biotransformed into new fluorescing DOM. The FeEEM of the MR þ SOM sample showed relatively high increases for P1 and P2 (63% and 64% for humic-like peaks P1 and P2, respectively) as compared to that of the other water samples. This is attributed to SOM (organic by-products from ozonation) that were easily biotransformed into new fluorescing DOM. Ogawa et al. (2001) discovered that marine bacteria are able to generate refractory DOM (i.e., fluorescing aromatic organic matter) from labile compounds (glucose and glutamate). On the other hand, the formation of fluorescing

4728

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 2 2 e4 7 3 6

Fig. 1 e (continued)

DOM was not observed in CR samples because of the aromatic nature of the organic matter and the fact that aromatic compounds are not readily biodegradable. Fluorescence difference spectra between influent and effluent samples were calculated and are shown in Fig. 1. Two types of difference spectra were calculated between influent and effluent samples. First, the spectrum of each effluent from the MR, MR þ SOM, SE, and CR samples was subtracted from their influent spectrum to represent degradation of fluorescing NOM (Fig. 1a, c, e and g). The spectrum of each influent

from the MR, MR þ SOM, SE, and CR samples was then subtracted from the corresponding effluent spectrum to represent formation of fluorescing NOM (Fig. 1b, d, f and h). As shown in Fig. 1, the degradation of fluorescing DOM mainly occurred in P3 (protein-like substances). However, a significant increase of fluorescing DOM occurred in P2 during the long HRT period. LC-OCD analyses of the MR, MR þ SOM, SE, and CR samples were performed. Five distinguishable peaks appeared in the LC-OCD chromatograms, and it was determined that these peaks are characteristic of 5 different classes of organic

Table 4 e Intensity changes in fluorescence peaks (P1, P2, and P3) characterized organic matter characteristics (contact time: 60 day, D: increase, e: decrease). Samples

MR MR þ SOM SE CR

Intensity changes in fluorescence peaks (P1, P2, and P3), % P1 (lex/em ¼ 250e270/420e440 nm)

P2 (lex/em ¼ 330e340/420e440 nm)

P3 (lex/em ¼ 270e280/320e340 nm)

27 (þ) 63 (þ) 32 (þ) 3 (þ)

34 (þ) 64 (þ) 25 (þ) 6 (þ)

52 () 52 () 50 () 48 ()

4729

44 65 97 96 92 97 97 63 0 97 91 97 96 2.0 1.1 0.1 0.1 0.1 0.1 0.1 1.5 4.5 0.1 0.1 0.1 0.1 3.6 3.1 3.2 2.8 1.2 3.1 3.0 4.0 4.5 3.8 1.1 3.1 2.6 76 83 98 97 92 97 97 85 0 97 0 97 97 0.76 0.57 0.1 0.1 0.1 0.1 0.1 0.6 5.6 0.1 0.1 0.1 0.1 3.2 3.3 4.0 3.3 1.2 3.4 3.1 4.1 3.9 3.0 0.1 3.3 3.4 84 97 98 97 93 98 97 46 15 98 98 89 98 0.5 0.1 0.1 0.1 0.1 0.1 0.1 2.8 4.6 0.1 0.1 0.1 0.1 3.2 3.4 4.1 3.6 1.5 4.1 2.9 5.2 5.4 4.7 4.1 0.9 4.6 86 94 94 94 88 95 94 68 0 98 97 94 97 0.27 0.1 0.1 0.1 0.1 0.1 0.1 0.7 2.3 0.1 0.1 0.1 0.1 1.9 1.8 1.8 1.8 0.8 2.0 1.7 2.2 2.3 4.0 3.2 1.8 3.6 28 10 90 98 94 94 21 0 0 97 96 50 96 2.8 3.7 0.3 0.1 0.1 0.1 2.2 4.7 4.6 0.1 0.1 1.5 0.1 3.9 4.1 3.0 4.3 1.7 1.8 2.8 3.9 3.1 3.5 2.4 3.0 2.8 Gemfibrozil Diclofenac Bezafibrate Ibuprofen Fenoprofen Naproxen Ketoprofen Clofibric acid Carbamazepine Phenacetin Paracetamol Pentoxifylline Caffeine

Eff. (ug/L) Inf. (ug/L)

Eff. (ug/L)

(%)

Inf. (ug/L)

Eff. (ug/L)

(%)

Inf. (ug/L)

Eff. (ug/L)

(%)

Inf. (ug/L)

Eff. (ug/L)

(%)

Inf. (ug/L)

CR SE MR þ SOM MR NCTW PhACs (ug/L)

Fig. 2 e Organic carbon concentrations with respect to biopolymers, humic substances, building blocks, neutrals and acids using LC-OCD (batch study, contact time: 60 days).

Table 5 e Summary of selected PhACs observed in batch studies (influent [Inf.], effluent [Eff.] and removal efficiency[%]).

carbon fractions: (i) biopolymers, (ii) humic substances, (iii) building blocks, (iv) neutrals, and (v) low molecular weight (MW) acids. Each fraction, determined from the chromatogram, represents a concentration of organic carbon (Fig. 2). LCOCD indicated that biopolymers (MW > 20,000 Da) were almost completely removed from all samples because the biopolymer fraction mainly comprised biodegradable organic matter in the form of proteins and polysaccharides. LC-OCD showed a similar removal trend with respect to organic matter fractions in the MR and MR þ SOM. MR and MR þ SOM reactors used the same source of water (River Meuse) except that MR þ SOM contained an additional external carbon source (SOM). This additional carbon source was also identified in LC-OCD chromatograms, which led to higher concentrations in acids and neutral fractions (Fig. 2). About 1.5 mg/L and 2 mg/L of humic substances were detected in the MR and MR þ SOM samples after the 60-day HRT period, respectively. A higher fraction of humic substances was measured in the MR þ SOM compared to the MR, and the SUVA of the MR þ SOM sample (SUVA ¼ 4.7 L mg1 m1) was also higher than that of the MR (SUVA ¼ 4.3 L mg1 m1). This increase in the humic fraction and SUVA appears to correspond to the formation of fluorescing DOM for P2 in the FeEEM spectra. Jarusutthirak and Amy (2007) observed the transformation of organic compounds during biological processes using glucose as a sole carbon source for bacterial growth and demonstrated that glucose (MW, 180 Da) is converted to higher MW organic matter during biological processes. Therefore, the biodegradation of labile organic matter (lowMW glucose) in the MR þ SOM could be attributed to the formation of fluorescing humic substances (P2) and humic fraction with an MW of 350 Da. The SE sample was found to have a relatively high concentration of biopolymers (MW > 20,000 Da) with respect to the MR, MR þ SOM, and CR samples because soluble microbial products (SMPs) were found to be present in the SE sample (Fig. 2). SMPs are components of biopolymers containing polysaccharide- and protein-like substances and

(%)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 2 2 e4 7 3 6

4730

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 2 2 e4 7 3 6

represent products derived from substrate metabolism during the growth of biomass in a biological wastewater treatment process (Jarusutthirak and Amy, 2007). In contrast, the removal of DOC from the CR sample mainly occurred in the humic substance fraction according to the LC-OCD analysis. This process can be better explained by adsorption rather than by biodegradation because of the refractory properties of humic substances in the CR sample. Moreover, AMB, measured by ATP, of the CR sample was found to be lower than that of the MR and MR þ SOM samples. This suggests that biodegradation was relatively low and did not contribute significantly to the removal of DOC. The AMB values measured as ATP concentrations associated with sand for the NCTW, SE, MR þ SOM, MR, and CR samples were 2.1, 10.2, 7.9, 6.4, and 4.5 ng ATP/cm3, during the 60-day HRT period, respectively. ATP concentrations associated with sand were relatively low as compared to that of sand obtained from a slow sand filter (18e93 ng ATP/cm3, MagiceKnezev and van der Kooij, 2004) after the long HRT (60 days). This long period promoted carbon-limiting conditions in the batch reactors. However, the average ATP concentrations remained correlated to the BDOC concentrations introduced in the batch reactors except for that of the CR sample. In the CR sample, BDOC, measured as an average DOC removal within 60 days, was higher than BDOC in the MR and MR þ SOM samples. However, the ATP concentration associated with sand from the CR sample was relatively low. This result implies that DOC was not utilized for energy or microbial growth by microorganisms. Therefore, it appears that DOC was simply adsorbed onto the sand. In addition, the low ATP concentration observed in the CR sample corresponded to the LC-OCD results in that the humic fraction was mainly removed during the HRT period. This suggests that biodegradation was relatively low and did not play a role in the removal of DOC. Trulleyov and Rul (2004) showed that there is a certain overestimation in BDOC results because of adsorbed DOC, which is resistant to biodegradation. This could be the case for the CR sample in this study. A comparison between ATP concentrations and DOC reduction in the different water samples suggests that there was an overestimation of BDOC results for the CR sample. A slight increase in SUVA, the relatively low transformation of organic matter (FeEEM), and the low AMB value of the CR sample support the refractory characteristics of DOM. Therefore, BDOC, measured as an average value representing removal of DOC, may not be an applicable measure to represent the amount of organic matter

utilized by microorganisms when a sample contains high levels of humic substances.

3.1.2.

Pharmaceutically actively compounds

In batch reactors, selected neutral PhACs with low Kow (log Kow < 2, Table 1) (phenacetine, paracetamol, and caffeine) and acidic PhACs (ibuprofen, fenoprofen, bezafibrate, and naproxen) exhibited removal efficiencies greater than 88% from the different water samples (which have different organic matter characteristics) with respect to BDOC and humic content under the 60-day HRT period. These compounds are mostly analgesics with caffeine being the only exception. However, the removal efficiencies for gemfibrozil and clofibric acid, both of which are lipid regulators, were found to be low. Such low removal efficiencies might be caused by the biodiversity of microorganisms, biodegradable carbon-limiting conditions, or adsorption competition between humic substances and selected PhACs. Moreover, removal efficiencies of diclofenac, gemfibrozil, and clofibric acid from batch reactors containing NCTW significantly decreased to 10%, 28%, and 0%, respectively. The NCTW had a low BDOC concentration relative to that of the other water samples, and there are fewer humic substances in the NCTW, which lead to less potential competition between humic substances and PhACs for binding sites. Thus, the adsorption capacity of PhACs may be greater in NCTW than in the other batch reactors. This may lead to more favorable conditions for adsorption of selected PhACs to sand. However, the removal efficiencies of diclofenac, gemfibrozil, and clofibric acid observed in NCTW indicate a dependence upon the amount of BDOC available. Therefore, co-metabolism may play an important role in the removal of diclofenac, gemfibrozil, and clofibric acid. Lim et al. (2008) showed that biotransformation rates of some PhACs (diclofenac and gemfibrozil) increase as initial BDOC concentrations of wastewater increase. This result suggests that the amount of BDOC is related to the removal of PhACs. On the other hand, neutral PhACs with low Kow (log Kow < 0.3, Table 1), which include compounds such as paracetamol, phenacetine, and caffeine, were removed even in the NCTW (i.e., biodegradable carbon-limited conditions). The BDOC concentration in the NCTW was 0.5 mg/L, and similar results were obtained in column studies, which are described in the latter part of this study. Among the target PhACs analyzed, carbamazepine (an anticonvulsant) was the most persistent compound and was not influenced by the characteristics of different types of

Table 6 e Summary of characteristics for SC1, SC2, SC3 and SC4.

SC1 (NCTW) SC2 (MR) SC3 (MR) SC4, (DW þ NaN3)

Influent Effluent Influent Effluent Influent Effluent Influent Effluent

pH

DOC (mg/L)

7.6 8.1 7.9 8.3 8.0 8.3 6.8 7.2

2.0 1.5 5.9 2.7 5.2 2.4 1.3 1.3

Average ATP (ng ATP/cm3)

SUVA (L mg1 m1)

0.5

30

3.2

102

2.8

91

2.5 2.3 2.9 3.3 2.4 3.1 13.0 13.5

BDOC (mg/L)

< 0.1

0.6

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 2 2 e4 7 3 6

4731

Fig. 3 e FeEEM spectra: influent: SC1(a), SC2(c) and SC3(e); effluent: SC1(b), SC2(d) and SC3(f), EBCT: 17-h (biotic conditions). organic matter. According to previous studies, carbamazepine has persisted in the environment, and its removal efficiency was found to be less than 10% in most wastewater treatment

plants (Stamatelatou et al., 2003; Ternes, 1998; Ternes et al., 2004; Zhang et al., 2008). Clara et al. (2004) proposed that carbamazepine could be used as a marker for anthropogenic

4732

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 2 2 e4 7 3 6

influences in the aquatic environment. PhACs concentrations in influent and effluent samples of batch studies are summarized in Table 5.

3.2. Comparison of biodegradation and sorption for removal of selected PhACs 3.2.1.

Characteristics of dissolved organic matter

Table 6 summarizes pH, DOC, and SUVA values and ATP concentrations obtained from the column studies. The average value representing the removal of DOC from NCTW was found to be 0.5 mg/L. Like the batch reactor experiments described previously, the increase in SUVA for the MR columns (SC1 and SC2) also indicated preferential removal of aliphatic organic matter. DOC removal by SC4 (DW þ NaN3) was less than 0.02 mg/L. This value is mainly attributable to the level of experimental error. The detection limit of the TOC analyzer used in this study was 0.05 mg/L; thus the removal efficiencies observed from SC4 (DW þ NaN3) were negligible. ATP concentrations associated with sand were measured to ensure that all columns were operated under biotic (SC1, SC2, and SC3) or abiotic (SC4) conditions. ATP measurements for each SC require dismantling of the column to obtain sand samples. Average ATP concentrations associated with sand for column depths of 10 mm and 290 mm were 30, 102, and 0.6 ng ATP/cm3 for SC1, SC2, and SC4, respectively, at the end of the experiment. Figs. 3 and 4 indicate the FeEEM spectra of samples obtained from column studies. Fig. 3b shows that there was no reduction in FI for P1 and P2, while the FI for P3 was reduced by 38% in NCTW (SC1). The fluorescence intensity of P3 indicates that some of the protein-like substances, which were present in the NCTW have been removed. However, the fluorescence intensity of P3 in the NCTW was relatively low as compared to that of other samples. Moreover, fluorescence intensities for P1, P2, and P3 were relatively low in the NCTW because the level of organic matter was low. Fig. 3d indicates removal efficiencies of 9%, 34.4%, and 54.5% for P1, P2, and P3, respectively, for SC2 (MR, 60 days of acclimation). Fig. 3f

indicates removal efficiencies of 7.7%, 8.7%, and 24.5% for P1, P2, and P3, respectively, for the MR column acclimated for 10 days (SC3). SC2 and SC3 have high fluorescence intensity of P3 relative to the NCTW. Fig. 4 shows that the FeEEM spectra obtained for the DW þ NaN3 (SC4, abiotic conditions) have peaks with significantly low fluorescence intensity. The fluorescence intensity of P3 detected in samples from SC4 may be contributed by fluorescing PhACs. FeEEM spectra observed for samples from SC4 were substantially lower than those from SC2 and SC3. Thus, the fluorescence intensity reduction in SC4 was negligible as a result of very low DOC and in the presence of a biocide in SC4. Fig. 5 shows the LC-OCD results of SC1, SC2, SC3, and SC4. Biopolymers defined by LC-OCD for SC1 and SC2 were removed with 83% and 57% efficiencies, respectively. The fraction representing humic substances in the NCTW sample was relatively low, and these results correspond to the interpretations of the FeEEM spectra where P1 and P2 (humic-like substances) were found to be low. Even though different acclimation times were provided for SC2 (60 days) and SC3 (10 days), the extent of removal of the different organic matter fractions was found to be similar between SC2 and SC3 after 60 days of operation. The samples from SC3 were taken after 60 days of operation. AMB values for SC3 are close to the AMB values for SC2. ATP concentrations associated with sand from SC2 and SC3 were 102 and 91 ng ATP/cm3 of sand, respectively (Table 6). For SC4 (DW þ NaN3), the humic substances fraction was not detected by LC-OCD, and there was no change in the organic matter fractions at this point of the study.

3.2.2.

Pharmaceutically actively compounds

PhACs concentrations from influent and effluent samples collected during column studies are summarized in Table 7. Fig. 6 shows that removal efficiencies of phenacetine, pentoxifylline and caffeine (hydrophilic neutral, log Kow < 1) were above 91% for SC2. However, for SC3, the short acclimation time (10 days) decreased the removal of phenacetine and pentoxifylline. Moreover, the removal efficiencies of gemfibrozil, diclofenac, and fenoprofen (acidic PhACs) were 36%,

Fig. 4 e FeEEM spectra: influent: SC4(a), effluent: SC4(b), EBCT: 17-h (abiotic conditions).

4733

29 0 2 32 0 33 27 46 0 15 10 6 17 4.1 4.2 4.1 1.7 3.3 3.9 2.2 3.6 4.4 4.0 3.7 5.0 3.9 5.8 4.0 4.2 2.5 2.8 5.8 3.0 6.7 4.4 4.7 4.1 5.3 4.7 0 5 93 41 36 94 98 6 1 78 68 95 95 3.2 3.6 0.3 1.0 3.9 0.3 0.1 2.9 6.7 1.3 2.6 0.3 0.3 2.5 3.8 4.5 1.7 6.1 4.8 4.0 3.1 6.8 5.8 8.1 6.3 6.4 36 20 92 78 49 92 93 11 0 90 98 98 97 2.1 3.3 0.4 0.4 2.1 0.4 0.3 3.3 4.9 0.4 0.1 0.1 0.1 3.3 4.1 4.8 1.8 4.1 4.9 4.0 3.7 4.8 4.2 5.0 4.8 3.7 7 0 49 63 33 77 88 20 0 95 98 98 94 3.9 4.1 2.4 0.7 2.6 0.7 0.6 3.2 6.0 0.2 0.1 0.1 0.2 4.2 4.0 4.7 1.9 3.9 3.1 5.0 4.0 4.4 4.1 4.8 4.4 3.6 Gemfibrozil Diclofenac Ibuprofen Fenoprofen Bezafibrate Naproxen Ketoprofen Clofibric acid Carbamazepine Phenacetine Pentoxifylline Paracetamol Caffeine

Removal (%) Eff. (ug/L) Removal (%) Eff. (ug/L) Inf.(ug/L) Eff. (ug/L) Inf. (ug/L)

Eff. (ug/L)

Removal (%)

Inf. (ug/L)

SC2, MAR SC1, NCTW PhACs (ug/L)

20%, and 78% for SC2 (60-day acclimated column) and 0%, 5%, and 41% for SC3 (10-day acclimated column), respectively. This implies that the removal of these compounds is correlated with microbial activity (ATP) associated with the sand. However, at the same time, the organic carbon content associated with the sand may have been different due to the different acclimation period. Thus, the removal efficiencies observed for selected PhACs compounds were probably affected by the different organic carbon content associated with the sand. Therefore, an additional experiment was carried out using SC1 and SC2 columns, which had the same acclimation time (i.e., approximately the same organic carbon content) to investigate the impact of low microbial activity by introducing low BDOC water (NCTW). In addition, the role of BDOC in the removal PhACs was also determined from SC1 and SC2. After 2 months of acclimation with MR, the influent to SC1 was changed from MR to NCTW (low BDOC). BDOC concentrations for NCTW (SC1) and MR (SC2) were found to be 0.5 mg/L and 3.2 mg/L, respectively. Low BDOC in NCTW resulted in low microbial activity in SC1 (ATP ¼ 30 ng ATP/cm3 of sand) compared to that of MR (SC2, ATP ¼ 102 ng ATP/cm3 of sand). Removal efficiencies of acidic PhACs such as analgesics decreased as result of the low BDOC (Fig. 7). However, hydrophilic neutral compounds such as paracetamol, pentoxifylline, phenacetine, and caffeine were still removed with efficiency above 91% by SC1. Microorganisms appear to be capable of removing paracetamol, pentoxifylline, phenacetine, and caffeine in NCTW. In order to determine if these compounds were removed by biodegradation in NCTW, an abiotic experiment was necessary to examine the removal of PhACs by sorption. In SC4 (DW þ NaN3), an average removal efficiency of 21% was observed for acidic PhACs under abiotic conditions. However, under biotic conditions, an average removal efficiency of these compounds increased, and it was 59% (SC2) (Table 7). These results confirm the importance of biodegradation processes for removal of acidic PhACs during soil passage. Moreover, the removal efficiencies of hydrophilic

Table 7 e Summary of selected PhACs in columns studies (Influent (In.) and effluent (Eff.)).

Fig. 5 e The change of organic matter fractions (biopolymers, humic substances, building blocks, neutrals and acids) determined by LC-OCD for SC1, SC2, SC3 and SC4.

Removal (%)

SC3, Acclimation e 10 day

Inf.(ug/L)

SC4 (abiotic)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 2 2 e4 7 3 6

4734

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 2 2 e4 7 3 6

120 60-day 10-day 100

Removal (%)

80

60

40

20

G em

fib ro zil D ic lo fe n Ib up ac r Fe ofe n no pr o Be fen za f Ke ibra te to pr of en N ap ro C xe lo f n C ar ibric ba ac m id az e Ph pin en e ac Pe et in nt e ox i Pa fyll in ra ce e ta m ol C af fe in e

0

Fig. 6 e Impact of acclimation period on the removal of wastewater-derived compounds (SC2: 60 days, SC3: 10 days, EBCT: 17 h).

120 Abiotic

NCTW

MR

100

Removal (%)

80

60

40

20

- The humic-like fluorescing DOM was formed during batch study (HRT: 60 days). This was due to the transformation of DOM during biodegradation. - Biopolymers degrade more rapidly than the other fractions of DOM according to LC-OCD measurements and measurements of active microbial biomass associated with sand, which were determined from the ATP analyses. These results indicate the availability of BDOC. - Neutral PhACs (phenacetine, paracetamol, and caffeine) and acidic PhACs (ibuprofen, fenoprofen, bezafibrate, and naproxen) exhibit removal efficiencies greater than 88% from various water samples in the batch studies. - The removal efficiencies of phenacetine, pentoxifylline, gemfibrozil, bezafibrate, and fenoprofen were found to be increased with acclimation time. - The removal efficiencies of acidic PhACs (analgesics) decreased under conditions where biodegradable carbon is limited. Co-metabolism may play an important role in the removal of acidic PhACs. - The removal efficiencies of acidic PhACs (ibuprofen, fenoprofen, ketoprofen, and naproxen) and neutral PhACs (paracetamol, pentoxifylline, phenacetine, and caffeine) were less than 33% under abiotic conditions. However, removal efficiencies of these compounds were greater than 75% under biotic conditions. This is mainly attributed to biodegradation. - The finding that hydrophilic neutral PhACs (paracetamol, pentoxifylline, and caffeine) can be removed under conditions of limited biodegradable carbon suggests that these compounds may have been used as a carbon source for the growth of microorganisms. - Carbamazepine persists in batch and column studies. - Based on the results of batch and column studies, biodegradation was found to be an important mechanism for removing PhACs during soil passage.

G em

fib

ro zil D ic lo fe na Ib c up r Fe ofe n no pr o Be fen za f Ke ibra te to pr of en N ap ro C xe lo f n C ar ibric ba ac m id az e Ph pin en e a Pe cet in nt e ox Pa ifylli ne ra ce ta m ol C af fe in e

0

Fig. 7 e Impact of microbial activity on wastewater-derived compounds (abiotic: SC4, NCTW: SC1, MR: SC2, EBCT: 17 h).

neutral PhACs (pentoxifylline, paracetamol and caffeine) were less than 17%. These substantially low removal efficiencies observed for SC4 were attributed to inactivation of microorganisms by sodium azide. However, under biotic conditions, removal efficiencies of the hydrophilic neutral PhACs were greater than 98% (SC4). As a result, the hydrophilic neutral PhACs easily passed through the column. This implies that the dominant removal mechanism for these compounds is biodegradation.

4.

Conclusions

Based on the results obtained in this study, the following conclusions can be drawn.

Acknowledgment We would like to acknowledge the help of Theo van der Kaaij and Ineke van der Veer-Agterberg (HetWaterlaboratorium) for the support on LC-OCD/OND and ATP measurements. We would like to express our gratitude to Dr. Sacher from DVGWTechnologiezentrum Wasser, Germany for PhACs measurements. This work was financially supported by EU SWITCH Project No. 018530-2 under the Sixth Framework Programme.

references

Amy, G., Drewes, J., 2007. Soil aquifer treatment (SAT) as natural and sustainable wastewater reclamation/reuse technology: fate of wastewater effluent organic matter (EfOM) and trace organic compounds. Environ. Monit. Assess. 129 (1e3), 19e26. Boxall, A.B.A., Kay, P., Blackwell, P.A., Fogg, L.A., 2004. Fate of Veterinary Medicines Applied to Soils. Pharmaceuticals in the Environment. Springer, Berlin. Calisto, V., Esteves, V.I., 2009. Psychiatric pharmaceuticals in the environment. Chemosphere 77 (10), 1257e1274.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 2 2 e4 7 3 6

Cha, W., Choi, H., Kim, J., Kim, I.S., 2004. Evaluation of wastewater effluents for soil aquifer treatment in South Korea. Water Sci. Technol. 50 (2), 315e322. Chen, H., Chen, S., Quan, X., Zhao, H., Zhang, Y., 2008. Sorption of polar and nonpolar organic contaminants by oilcontaminated soil. Chemosphere 73 (11), 1832e1837. Clara, M., Strenn, B., Kreuzinger, N., 2004. Carbamazepine as a possible anthropogenic marker in the aquatic environment: investigations on the behaviour of carbamazepine in wastewater treatment and during groundwater infiltration. Water Res. 38 (4), 947e954. Coble, P.G., 1996. Characterization of marine and terrestrial DOM in seawater using excitationeemission matrix spectroscopy. Mar. Chem. 51 (4), 325e346. Cunningham, V.L., 2008. Special characteristics of pharmaceuticals related to environmental fate. In: Ku¨mmerer, K. (Ed.), Pharmaceuticals in the Environment. Springer, Berlin. Focazio, M.J., Kolpin, D.W., Barnes, K.K., Furlong, E.T., Meyer, M.T. , Zaugg, S.D., Barber, L.B., Thurman, M.E., 2008. A national reconnaissance for pharmaceuticals and other organic wastewater contaminants in the United States e II) Untreated drinking water sources. Sci. Total Environ. 402 (2e3), 201e216. Gru¨nheid, S., Jekel, M., 2005. Fate of Trace Organic Pollutants during Bank Filtration and Groundwater Recharge, In. Proceeding of 5th International Symposium on Management of Aquifer Recharge, 10e16 June 2005, Berlin, Germany. Heberer, T., 2002. Occurrence, fate, and removal of pharmaceutical residues in the aquatic environment: a review of recent research data. Toxicol. Lett. 131 (1e2), 5e17. Heberer, T., Mechlinski, A., Fanck, B., Knappe, A., Massmann, G., Pekdeger, A., Fritz, B., 2004. Field studies on the fate and transport of pharmaceutical residues in bank filtration. Ground Water Monit. R. 24 (2), 70e77. Henderson, R.K., Baker, A., Murphy, K.R., Hambly, A., Stuetz, R.M., Khan, S.J., 2009. Fluorescence as a potential monitoring tool for recycled water systems: a review. Water Res. 43 (4), 863e881. Howard, P.H., 2000. Biodegradation. Handbook of Property Estimation Methods for Chemicals. Environmental Health Sciences CRC Press LLC, Boca Raton. Huber, S., Frimmel, F.H., 1992. A liquid chromatographic system with multi-detection for the direct analysis of hydrophilic organic compounds in natural waters. Fresenius’ J. Anal. Chem. 342 (1e2), 198e200. Hudson, N., Baker, A., Ward, D., Reynolds, D.M., Brunsdon, C., Carliell-Marquet, D., Browning, S., 2008. Can fluorescence spectrometry be used as a surrogate for the Biochemical Oxygen demand (BOD) test in water quality assessment? An example from South West England. Sci. Total Environ. 391 (1), 149e158. Jarusutthirak, C., Amy, G., 2007. Understanding soluble microbial products (SMP) as a component of effluent organic matter (EfOM). Water Res. 41 (12), 2787e2793. Jjemba, P.K., 2006. Excretion and ecotoxicity of pharmaceutical and personal care products in the environment. Ecotox. Environ. Safe. 63 (1), 113e130. Kim, S.D., Cho, J., Kim, I.S., Vanderford, B.J., Snyder, S.A., 2007. Occurrence and removal of pharmaceuticals and endocrine disruptors in South Korean surface, drinking, and waste waters. Water Res. 41 (5), 1013e1021. Ku¨mmerer, K., 2008. Pharmaceuticals in the environmenteA brief summary. In: Ku¨mmerer, K. (Ed.), Pharmaceuticals in the Environment. Springer, Berlin. Ku¨mmerer, K., 2009. The presence of pharmaceuticals in the environment due to human use - present knowledge and future challenges. J. Environ. Manage. 90 (8), 2354e2366. Leenheer, J.A., Croue, J.-P., 2003. Characterizing aquatic dissolved organic matter. Environ. Sci. Technol. 37 (1), 18Ae26A.

4735

Liang, C., Dang, Z., Xiao, B., Huang, W., Liu, C., 2006. Equilibrium sorption of phenanthrene by soil humic acids. Chemosphere 63 (11), 1961e1968. Lim, M.-H., Snyder, S.A., Sedlak, D.L., 2008. Use of biodegradable dissolved organic carbon (BDOC) to assess the potential for transformation of wastewater-derived contaminants in surface waters. Water Res. 42 (12), 2943e2952. Madden, J.C., Enoch, S.J., Hewitt, M., Cronin, M.T.D., 2009. Pharmaceuticals in the environment: good practice in predicting acute ecotoxicological effects. Toxicol. Lett. 185 (2), 85e101. Maeng, S.K., Sharma, S.K., Amy, G., Magic-Knezev, A., 2008. Fate of effluent organic matter (EfOM) and natural organic matter (NOM) through riverbank filtration. Water Sci. Technol. 57 (12), 1999e2007. Maeng, S.K., Ameda, E., Sharma, S.K., Gru¨tzmacher, G., Amy, G., 2010. Organic micropollutant removal from wastewater effluent-impacted drinking water sources during bank filtration and artificial recharge. Water Res. 44 (14), 4003e4014. Maeng, S.K., Sharma, S.K., Lekkerkerker-Teunissen, K., Amy, G., 2011a. Occurrence and fate of bulk organic matter and pharmaceutically active compounds in managed aquifer recharge: a review. Water Res. 45 (10), 3015e3033. Maeng, S.K., Sharma, S.K., Amy, G., 2011b. Framework for assessment of organic micropollutant (OMP) removals during managed aquifer recharge and recovery (MAR): in “Riverbank filtration for Water Security in Desert Countries”. In: Ray, C., Shamrukh, M. (Eds.), NATO Science for Peace and Security Series. Springer, Dordrecht, The Netherlands. Magic-Knezev, A., van der Kooij, D., 2004. Optimisation and significance of ATP analysis for measuring active biomass in granular activated carbon filters used in water treatment. Water Res. 38 (18), 3971e3979. Massmann, G., Dunnbier, U., Heberer, T., Taute, T., 2008. Behaviour and redox sensitivity of pharmaceutical residues during bank filtration - Investigation of residues of phenazone-type analgesics. Chemosphere 71 (8), 1476e1485. Mechlinski, A., Heberer, T., 2005. Fate and Transport of Pharmaceutical Residues during Bank Filtration. Proceeding of ISMAR 2005, Berlin. Mompelat, S., LeBot, B., Thomas, O., 2009. Occurrence and fate of pharmaceutical products and by-products, from resource to drinking water. Environ. Int. 35 (5), 803e814. Oades, J.M., Jenkinson, D.S., 1979. Adenosine triphosphate content of the soil microbial biomass. Soil Biol. Biochem. 11 (2), 201e204. Ogawa, H., Amagai, Y., Koike, I., Kaiser, K., Benner, R., 2001. Production of refractory dissolved organic matter by bacteria. Science 292, 917e920. , J., Petrovic, M., Barcelo, D., 2009. Fate and Radjenovic distribution of pharmaceuticals in wastewater and sewage sludge of the conventional activated sludge (CAS) and advanced membrane bioreactor (MBR) treatment. Water Res. 43 (3), 831e841. Saadi, I., Borisover, M., Armon, R., Laor, Y., 2006. Monitoring of effluent DOM biodegradation using fluorescence, UV and DOC measurements. Chemosphere 63 (3), 530e539. Sacher, F., Ehmann, M., Gabriel, S., Graf, C., Brauch, H.-J., 2008. Pharmaceutical residues in the river Rhine-results of a onedecade monitoring programme. J. Environ. Monit. 10, 664e670. Schmidt, C.K., Lange, F.T., Brauch, H.J., 2007. Characteristics and evaluation of natural attenuation processes for organic micropollutant removal during riverbank filtration, In. Proceeding of Regional IWA Conference on Groundwater Management in the Danube river Basin and other Large River Basins, Belgrade. Sharma, S.K., Maeng, S.K., Nam, S.-N., Amy, G., 2011. Characterization tools for differentiating NOM from EfOM. In:

4736

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 2 2 e4 7 3 6

Peter, W. (Ed.), Treatise on Water Science. Aquatic Chemistry and Microbiology, vol. 3. Elsevier Publications, Oxford. Snyder, S.A., Leising, J., Westerhoff, P., Yoon, Y., Mash, H., Vanderford, B., 2004. Biological and physical attenuation of endocrine disruptors and pharmaceuticals: implications for water reuse. Ground Water Monit. R. 24 (2), 108e118. Stamatelatou, K., Frouda, C., Fountoulakis, M.S., Drillia, P., Kornaros, M., Lyberatos, G., 2003. Pharmaceuticals and health care products in wastewater effluents: the example of carbamazepine. Water Sci. Technol. 3 (4), 131e137. Ternes, T.A., 1998. Occurrence of drugs in German sewage treatment plants and rivers. Water Res. 32 (11), 3245e3260. Ternes, T.A., Herrmann, N., Bonerz, M., Knacker, T., Siegrist, H., Joss, A., 2004. A rapid method to measure the solid-water distribution coefficient (Kd) for pharmaceuticals and musk fragrances in sewage sludge. Water Res. 38 (19), 4075e4084. Trulleyov, S.K., Rul, M., 2004. Determination of biodegradable dissolved organic carbon in waters: comparison of batch methods. Sci. Total Environ. 332 (1e3), 253e260. US EPA, 2009. Estimation Programs Interface Suite for Microsoft Windows, V. 4.00. United States Environmental Protection Agency, Washington, DC, USA.

Urfer, D., Huck, P.M., 2001. Measurement of biomass activity in drinking water biofilters using a respirometric method. Water Res. 35 (6), 1469e1477. Xue, S., Zhao, Q.-L., Wei, L.-L., Ren, N.-Q., 2009. Behavior and characteristics of dissolved organic matter during column studies of soil aquifer treatment. Water Res. 43 (2), 499e507. Yangali-Quintanilla, V., Sadmani, A., McConville, M., Kennedy, M., Amy, G., 2010a. A QSAR model for predicting rejection of emerging contaminants (pharmaceuticals, endocrine disruptors) by nanofiltration membranes. Water Res. 44 (2), 373e384. Yangali-Quintanilla, V., Maeng, S.K., Fujioka, T., Kennedy, M., Amy, G., 2010b. Proposing nanofiltration as acceptable barrier for organic contaminants in water reuse. J.Membr. Sci. 362 (1e2), 334e345. Ying, G.-G., Kookana, R.S., Kolpin, D.W., 2009. Occurrence and removal of pharmaceutically active compounds in sewage treatment plants with different technologies. J. Environ. Monit. 11, 1498e1505. Zhang, Y., Geissen, S.-U., Gal, C., 2008. Carbamazepine and diclofenac: removal in wastewater treatment plants and occurrence in water bodies. Chemosphere 73 (8), 1151e1161.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 3 7 e4 7 5 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Application of system dynamics for developing financially self-sustaining management policies for water and wastewater systems R. Rehan a, M.A. Knight a,*, C.T. Haas a, A.J.A. Unger b a b

200 University Avenue, Department of Civil and Environmental Engineering, University of Waterloo, ON, Canada N2L 3G1 200 University Avenue, Department of Earth and Environmental Sciences, University of Waterloo, ON, Canada N2L 3G1

article info

abstract

Article history:

Recently enacted regulations in Canada and elsewhere require water utilities to be finan-

Received 8 December 2010

cially self-sustaining over the long-term. This implies full cost recovery for providing water

Received in revised form

and wastewater services to users. This study proposes a new approach to help water

16 April 2011

utilities plan to meet the requirements of the new regulations. A causal loop diagram is

Accepted 3 June 2011

developed for a financially self-sustaining water utility which frames water and waste-

Available online 5 July 2011

water network management as a complex system with multiple interconnections and feedback loops. The novel System Dynamics approach is used to develop a demonstration

Keywords:

model for water and wastewater network management. This is the first known application

Waste water collection

of System Dynamics to water and wastewater network management. The network simu-

Water distribution

lated is that of a typical Canadian water utility that has under invested in maintenance.

Infrastructure management

Model results show that with no proactive rehabilitation strategy the utility will need to

System dynamics

substantially increase its user fees to achieve financial sustainability. This increase is

Financial sustainability

further exacerbated when price elasticity of water demand is considered. When the utility pursues proactive rehabilitation, financial sustainability is achieved with lower user fees. Having demonstrated the significance of feedback loops for financial management of water and wastewater networks, the paper makes the case for a more complete utility model that considers the complexity of the system by incorporating all feedback loops. Crown Copyright ª 2011 Published by Elsevier Ltd. All rights reserved.

1.

Introduction

Municipal water and wastewater systems deliver clean water to residents, businesses, and industries and collect contaminated water (wastewater) for treatment and disposal. The health and prosperity of cities depend on well-functioning “out of sight” and often “out of mind” water and wastewater networks. In North America the assigned service life of buried distribution and collection pipes is often 50e75 years (Ministry of the Environment Ontario, 2007; CBO, 2002) even

though in some cases these pipes have been in service for more than 100 years. In North America, many cities are faced with the challenge of managing aging water and wastewater infrastructure with limited fiscal and personnel resources while ensuring that adequate levels of service are provided to consumers and customers. In Canada, recent federal and provincial government legislation requires public water agencies to be financially accountable by mandating new reporting requirements. New regulations include the Canadian Institute of Chartered

* Corresponding author. Tel.: þ1 519 581 8835. E-mail addresses: [email protected] (R. Rehan), [email protected] (M.A. Knight), [email protected] (C.T. Haas), aunger@ uwaterloo.ca (A.J.A. Unger). 0043-1354/$ e see front matter Crown Copyright ª 2011 Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.06.001

4738

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 3 7 e4 7 5 0

Accountants Public Sector Accounting Board (PSAB) statement PS3150 that requires all municipalities, starting in January 2009, to report all tangible capital assets along with their depreciation on financial statements (CICA, 2007) and Province of Ontario Regulation 453/07 (Ministry of the Environment Ontario, 2007), developed under the Safe Drinking Water Act, that requires all public utilities prepare and submit yearly reports on the current and estimated future condition of water and wastewater infrastructure. The later also requires the preparation and publication of long-term water and wastewater sustainability financial plans. This is related to the concept of “sustainable urban water” emerging in other parts of the world. A key principle for these plans is that revenues should be sufficient to pay all expenses of providing services (Ministry of the Environment Ontario, 2007). In the United States, the Governmental Accounting Standards Board (GASB) Statement 34, in France Accounting Standard M49, and in Australia, the Australian Accounting Research Foundation Standard 27 specifies similar accounting practices to PSAB (see FHWA, 2000; Howard, 2001; and Barraque and LeBris, 2007). Over the past several years many researchers have developed decision support tools to aid water utilities manage their water and wastewater networks. These tools include some or a combination of activities such as: registration of data related to infrastructure components; assessment and grading of the asset conditions; analysis of data for predicting remaining service life; comparison of costs of repair/rehabilitation alternatives over their life cycles; and, prioritization of rehabilitation activities that ensure maximum benefits at minimum costs (Grigg, 2003). The following provides an overview of management tools developed for water distribution networks. Shamir and Howard (1979) developed one of the first age based models to predict water main failure rates and Deb et al. (1998) developed the KANEW model using the concept of a survival function, which is a statistical predictor of useful life of a group of pipes belonging to the same class (e.g. age, material, and diameter). Kleiner et al. (1998) modeled the performance of a water distribution network by incorporating both the deterioration of structural integrity and hydraulic capacity. This approach is used to identify optimal rehabilitation strategies that minimize the total costs of rehabilitation and all maintenance over the planning horizon. Hadzilacos et al. (2000) present a prototype decision support system (DSS) called UtilNets for water pipes. This model facilitates rehabilitation of critical watermains based on reliability based life predictions. The DSS provides an aggregate structural, hydraulic, water quality, and service profile of a network along with an assessment of the required rehabilitation expenditures. Burn et al. (2003) employ a non homogeneous Poisson burst count model for predicting failure rates of pipes and developed PARMS-PLANNING which analyses expenditures and costs over a range of strategies. Moglia et al. (2006) developed PARMS-PRIORITY to add calculations for risk, failure predictions, cost assessment, scenario evaluation, and data exploration. In Saegrov (2005) KANEW is developed into CARE-W, a more comprehensive DSS that has modules for the assessment of performance indicators, prediction of pipe failures, and water supply reliability. Results generated from

these modules are utilized in two further modules that allow for planning long-term investment needs and annual rehabilitation project selection and ranking. Giustolisi et al. (2006) developed a polynomial regression method to predict the burst rates of watermains. The policy option explored is comparison of the reduction in burst rates after pipes’ replacement versus the cost of replacement. Dandy and Engelhardt (2006) applied a multi objective genetic algorithm approach to develop trade off curves between economic cost and reliability for replacement schedules of water pipes. Tabesh et al. (2009) present artificial neural network and neuro-fuzzy system models. This study found the artificial neural network model superior in terms of predicting pipe failure rate and for the assessment of mechanical reliability in water distribution networks. Kleiner et al. (2010) present a pipe failure prediction model and optimize renewal investments by taking into account costs that include adjacent infrastructure and economies of scale. The development of wastewater (sewer) network management tools is discussed in the following section. Wirahadikusumah and Abraham (2003) use probabilistic dynamic programming in conjunction with a Markov chain model to perform life cycle cost analysis of sewers. Savic et al. (2006) use evolutionary polynomial regression to develop models for predicting wastewater blockage events and collapse failures. Saegrov (2006) develops CARE-S, a corresponding framework to CARE-W for wastewater network rehabilitation decision making. CARE-S is a comprehensive DSS that combines several tools relevant to wastewater infrastructure management into a single platform. Younis and Knight (2010a) present a continuation ratio model that can be used for risk-based policy development for maintenance management of wastewater collection systems. Their proposed model can be used in devising appropriate intervention plans and optimum network maintenance management strategies based on pipelines age, material type, and internal condition grades. Younis and Knight (2010b) show that a cumulative logit model can be used to determine wastewater pipelines’ service life, predict future condition states, and estimate networks’ maintenance and rehabilitation expenditures. Halfawy et al. (2006) reviewed the following commercial municipal asset management systems: Synergen, CityWorks, MIMS, Hansen, RIVA, Infrastructure 2000, and Harfan. They found the majority of existing commercial asset management software to focus on operational management (e.g., work orders, service requests) with little or no functionality to support long-term renewal planning decisions (e.g., deterioration modeling, risk assessment, life cycle cost analysis, asset prioritization). From the reviewed systems, RIVA, Harfan, and Infrastructure 2000 implemented some level of support for long-term renewal planning of specific assets, mainly pavement. The other four systems include condition assessment and rating modules. Most of these commercial software tools now incorporate PSAB and other legislation annual reporting requirements and have improved strategic long range asset, risk and budget management by forecasting the full life cycle of infrastructure assets. They also generate a life cycle cost and risk profile for each asset, determine the events that should be scheduled each period, as well as, the

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 3 7 e4 7 5 0

impact on cost, condition, risk and capacity. None of these tools are water and wastewater asset specific management tools. Currently, no integrated water and wastewater decision support tool exists that considers the impact of feedback loops and complex interactions between integrated water, wastewater, financial and social sectors. Englehardt et al. (2003) state that when considering the financial sustainability of a water utility, it is vital to include the whole life cycle costs associated with network operation, maintenance, and rehabilitation. Linerand and deMonsabert (2010) indicate that the application of the triple bottom line (TBL) also requires utilities to analyze alternatives to address conflicting goals of economics (financial), environmental, and social issues. This study proposes a novel interconnected municipal water and waste water asset management framework using a System Dynamics model. This management framework will assist water utilities in whole life cycle cost analysis and to address triple bottom line principles. The paper first demonstrates complex interconnections and feedback loops between the physical infrastructure, financial and consumer sectors. The paper then describes the use and application of System Dynamics modeling for integrated water and wastewater network pipeline asset management. To the authors knowledge this is the first known application of System Dynamics to integrated water and wastewater asset management. This is then followed by the development of a basic aggregated water and wastewater System Dynamics demonstration model that is used to model the significance of complex interconnections and feedback loops on management decisions. A fully integrated water and wastewater model can be developed that includes water and wastewater pipe network, access chambers (manholes), laterals, valves, hydrants, treatment plants, etc, using the proposed System Dynamics approach. The development of a fully integrated model is deemed to be beyond the scope of this paper. Burnside (2005) noted that water distribution and wastewater collection networks together constitute approximately 75 percent of the costs of a municipal water system. Since water distribution and wastewater collection networks are the majority of the utility costs, the cost of water and wastewater treatment will not be considered in this paper analysis. The demonstration System Dynamics model is then used to show the impact of three specific management strategies on the utilities’ financial sustainability over the long-term. Three specific scenarios are discussed. First, the utility is assumed to under invest in the water distribution and wastewater collection networks by not paying for capital works needed to replace deteriorated buried pipes. Second, the utility is assumed to adopt a 1% annual replacement rate strategy. This strategy is motivated by the assumption that the average pipe lifespan is 100 years. Therefore, the entire network will be effectively replaced once every 100 years. Third, the utility is assumed to adopt a strategy by which no more than 5% of its network is in the poorest condition state. For each of the above three scenarios, three variations are considered which reflect: (A) a constant user fee and with no constraints on the utility’s fund balance, i.e. revenues do not need to equal expenses; (B) a variable user fee and with a zero

4739

funds balance, i.e. revenues equal expenses; and (C) a variable user fee, zero funds balance and price elasticity of water demand.

2. Modelling the complexity of water and wastewater network management The concept of interconnected components and complex system behavior for urban water systems is well recognized. For example, Grigg and Bryson (1975) presented a simulation model that is comprised of four interconnected sectors e financial accounting, water balance, water use, and population growth. Kotz and Hiessl (2005) demonstrated dynamic system interdependencies and used an agent-based modeling approach to simulate technical innovation processes in these systems. Guest et al. (2010) studied interactions among sustainability aspects related to decentralized wastewater treatment systems using a qualitative System Dynamics approach. Ahmad and Prashar (2010) also use a System Dynamics model to study interconnections among population growth, land use changes, water demand, and water availability. Adeniran and Bamiro (2010) modeled the interconnections among Finance, Production, Distribution, and Operation & Maintenance sectors of a municipal water supply system. This model does not include water and wastewater physical infrastructure. Management of municipal water and wastewater networks is a complex problem. Sterman (2000)states that the interaction of feedback loops is responsible for complex system behavior. When a component inside a feedback loop is changed, the perturbation traverses along the loop resulting in a change to the originating component (Hannon and Ruth, 1994). When a change in the originating component causes a change in other components that strengthens the original process, the feedback loop is termed a positive or a selfreinforcing loop. If the response of other components along the loop counteracts the original change, a negative or balancing loop is deemed to exist (Hannon and Ruth, 1994). In this section, feedback loops related to water and wastewater network management are identified using Fig. 1 causal loop diagram (CLD). In a CLD, relationships between variables are depicted using arrows with a positive (þ) or negative () sign placed besides the arrow head to indicate link polarity. A positive link polarity implies that “if a cause increases, the effect increases above what it would otherwise have been” and vice versa (Sterman, 2000). Similarly, a negative link polarity “means that if the cause increases, the effect decreases below what it would otherwise have been” and vice versa (Sterman, 2000). A simplified CLD for municipal water and wastewater network management is shown in Fig. 1. Names of feedback loops are in bold font and thick curved arrows around loop names indicate the direction of causation. The objective of presenting the CLD in Fig. 1 is to frame the scope of the System Dynamics model that is developed as part of this research. The System Dynamics model that is presented later in this work implements only a subset of the Fig. 1 causal loops. We describe the “bigger picture” of what major causal loop dependencies exist within a water and wastewater network

4740

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 3 7 e4 7 5 0

Fig. 1 e Feedback loops in water and wastewater network management.

from a management perspective. For this paper, we limit our discussion on causal loops that illustrate sustainable financial management strategies and demonstrate the complexity of the system.

2.1.

Feedback loop in infrastructure deterioration (R1)

Reinforcing loop R1 (Fig. 1) represents the typical deterioration process for physical infrastructure. It shows that the rate of deterioration of infrastructure is a function of its existing condition, which in turn, determines the condition of the infrastructure. If the condition of an infrastructure component increases (e.g., on a scale of 1e5, where 5 is a poor state and 1 is the best state), an increase in the deterioration rate occurs. A higher deterioration rate then leads to further deterioration of the infrastructure. Thus, a cycle is established in which infrastructure deterioration occurs at an accelerated rate. Wirahadikusumah and Abraham (2003) report a similar process of deterioration.

2.2.

Feedback loop in infrastructure rehabilitation (B1)

The exponential deterioration of infrastructure caused by loop R1 is mitigated by a balancing loop, B1. If infrastructure condition deteriorates (increases), the network’s service

performance will decline as a result. For example, deteriorated watermains cause more discoloured water events and watermain breaks. Similarly, reduced hydraulic capacity of deteriorated wastewater pipes will result in frequent backups. Increased complaints by consumers due to poor service performance of watermain and wastewater pipes will increase pressure on utility managers to improve the infrastructure condition by employing rehabilitation techniques. Increased rehabilitation works translate into improved infrastructure condition, closing the loop. Thus, deterioration in infrastructure condition, in a functional society, will ultimately drive improvement.

2.3.

Feedback loop in revenue generation (R2)

A water utility is financially self-sustaining when its revenues equal or exceed its expenses. When its fund balance (revenues minus expenditures) fall below a threshold value, the utility will often increase revenues by increasing user fees. Consumers can respond to an increase in user fees by reducing water consumption. The reduction in water use is often characterized by time delays (Fortin et al., 2001). For the more prevalent case where the utility charges its customers on the basis of consumed volume of water, a decrease in water consumption will reduce revenues. Lower revenues will result

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 3 7 e4 7 5 0

in a decreased fund balance. A self-reinforcing loop is established where an initial rise in user fees will ultimately cause user fees to increase more. It should be noted that this selfreinforcing feedback loop may not operate indefinitely as constraints on one or more parameters around the loop may be triggered that stop growth. For instance, once the minimum water demand (due to social or technological limits) is reached, further decreases may not occur regardless of user fees increases.

2.4.

Feedback loop in user fees adjustments (B2)

The operation of reinforcing loop R2 can be constrained by the existence of a balancing feedback loop B2. This feedback loop represents the limitations imposed by the socio-political environment on utility managers. In Canada, urban water and wastewater systems are publically owned. Therefore, user fees increases have to be approved by municipal councils which are sensitive to voters’ feedback. When user fees are increased, it causes a reduction in customers’ willingness to accept a further fee hike. Reduced willingness to accept a fee hike implies that future user fees will be lower than what would otherwise have been. Loop B2 is connected to loop B1 through the willingness to accept a fee hike. MacDonald et al. (2003) report that consumers are willing to pay positive amounts of money in return for a water supply service that would be more reliable and less prone to service interruptions. Since a deteriorated infrastructure system will cause increased service interruptions, it is reasonable to suggest that increased deterioration will increase consumers’ willingness to accept a fee hike. An increased willingness to accept a fee hike will result in increased user fees.

2.5.

Feedback loop in capital expenditures (B3)

Increased rehabilitation of infrastructure will increase the utility’s capital expenditures. This in turn reduces the availability of funds for further rehabilitation works. With a lower fund balance, infrastructure rehabilitation is decreased.

2.6.

Feedback loop in operational expenditures (R3)

This feedback loop is comprised of the following variables: Infrastructure Condition, Operational Expenditures, Funds Balance, and Infrastructure Rehabilitation. When the infrastructure condition deteriorates (increases), operational expenditures will increase due to the need for more frequent pipe flushing and emergency repairs. Pumping costs (due to reduced hydraulic capacity) will also increase. Deteriorated condition is also associated with water leakage in case of watermains and infiltration in case of sanitary sewers. Both these scenarios entail additional costs for the utility. An increase in operational expenditures will lower the funds balance and in turn the funds available for rehabilitation. With less rehabilitation, the condition of infrastructure will deteriorate further resulting in the cycle of deterioration to accelerate. The above discussion shows that water and wastewater infrastructure management involves multiple interacting

4741

feedback loops. To date, no model is available that captures the dynamic complexity arising due to these feedback loops. Therefore, a novel contribution of this study is to develop a System Dynamics model that can be used for strategic network management.

3.

System dynamics modeling

System Dynamics is a feedback based object oriented modeling paradigm developed by Forrester (1958) to model complex systems. The basic building blocks for System Dynamics models are: stocks, flows, converters, and connectors (Fig. 2). Stocks represent accumulations e both physical and non-physical. Examples of physical stocks are inventory of pipes, amount of water in a reservoir, etc. A non-physical stock is the consumer’s level of satisfaction with a water utility service. Stocks represent the ‘traces’ left by an activity. Material in a stock exists at a given point in time and persists even when activities end. Flows represent activities or actions in a stock that transport quantities into or out of a stock instantaneously or over time. Examples of flows are daily consumption of water, rate at which pipes move from one condition grade to another, monthly revenues or expenditures of a utility, etc. Mathematically the relationship between stocks and flows can be described using the following integral form (Sterman, 2000): Zt ½InflowðsÞ  OutflowðsÞds þ Stockðto Þ

StockðtÞ ¼

(1)

t0

where to is the initial time, t is the current time, Stock (to) is the initial value of the stock, Inflow (s) and Outflow (s) are flow rates into and out of a stock at any time s between the initial time to and current time t. Inflow (s) and Outflow (s) have the units of Stock (t) divided by time. Equation (2) determines the net rate of change of a stock with time (Sterman, 2000) dðStockÞ=dt ¼ InflowðtÞ  OutflowðtÞ

(2)

Fig. 3 shows a demonstration System Dynamics model for a hypothetical water utility that contains three sectors: physical infrastructure, consumer and finance. In Fig. 3 the connectors (arrows) establish relationships among various elements of the model and move information as inputs for

Fig. 2 e Building blocks of system dynamics models.

4742

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 3 7 e4 7 5 0

Fig. 3 e Demonstration system dynamics model for a water utility.

decisions or actions and converters (circles) house graphical and built-in functions. Examples of converters are pipe deterioration curves and demand curves for water usage.

4. Demonstration system dynamics model for water and wastewater network management To quantitatively highlight the significance of physical infrastructure, consumer and finance sector interconnections and feedback loops on strategic water utility management decisions the hypothetical demonstration System Dynamics model, shown in Fig. 3, is presented. It should be noted that this demonstration model is not a fully developed water utility model and is not deemed ready for utility management. The presented model is deemed sufficient to make the case for the development of a fully integrated System Dynamics model that can be used by utility managers for strategic decision making over the short and long-term. The following sections describe construction of the physical infrastructure, consumer and financial sectors of the demonstration model.

4.1.

Physical infrastructure sector

The physical infrastructure sector includes water and wastewater network pipes. Although the modeling framework allows for the development of separate water and wastewater pipe stocks, these stocks are aggregated in the demonstration model, for simplicity, into five Condition Group stocks (Condition Group 20, Condition Group 40, Condition Group 60, Condition Group 80, and Condition Group 100). Each condition group is assigned an average condition grade using an arbitrary scale that varies from 0 to 100.

Younis and Knight (2010a,b), Tabesh et al. (2009), and Savic et al. (2006) report that the deterioration of watermains and wastewater pipes depends upon several factors and that many different types of deterioration functions can be implemented to represent pipeline deterioration from one condition state to another. In the demonstration model each pipe is allowed to move from one condition state to the next (worse) condition state using flow functions such as Deterioration 20 to 40, Deterioration 40 to 60 etc as shown in Fig. 3. Although, any type of deterioration function can be implemented into Deterioration flows, a simple age based deterioration function is implemented in the demonstration model e each pipe is allowed to reside in a Condition Group stock for an average period of 20 years before moving into the next Condition Group stock. The reasons for implementing this simple age based deterioration process are: 1) age is commonly reported in the published literature to be strongly correlated to pipe condition and 2) the aggregation of the water and wastewater pipe segments into the same Condition Group stocks does not allow for the implementation of separate water and wastewater pipeline deterioration functions. Current Canadian government guidelines (e.g., Ministry of the Environment Ontario, 2007) indicate the service life for various civil infrastructure assets. For wastewater pipelines, the service life ranges from 40 to 75 years with limited or no asset deterioration knowledge (Ministry of the Environment Ontario, 2007). The flexible System Dynamics model architecture allows for the pipe average service life to be set to any value. To represent typical Canadian practice, the average service life is set to 100 years. Pipe renewal and replacement is represented using flow Renewal and user specified input Rehab Fraction. During each simulation time step, flow Renewal moves pipes from stock

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 3 7 e4 7 5 0

Condition Group 100 to stock Condition Group 20 using the lesser of Rehab Length and the total length of pipes in stock Condition Group 100. Most utilities will have set performance criteria for making rehabilitation investment decisions such as reducing recurring expenditures (OpEx) and ensuring levels of service to its customers (minimum service disruptions, watermain breaks, wastewater blockages, adequate water supply pressures, etc). Although performance criteria are not included in the proposed demonstration model, they can be implemented in a fully developed System Dynamics model. In the demonstration model poor service levels can be associated with the length of pipes in each Condition Group stock as will be explained in Section 4.3 below. The current demonstration model is formulated so that all rehabilitation activity only removes pipes from stock Condition Group 100. In practice, existing pipes may be repaired and/or renovated to extend their service life. In a fully developed System Dynamics model pipe repair and renovation activities can be formulated by providing additional flows similar to the flow Renewal. For example, if a rehabilitation technique extends the service life of a Condition 80 pipe by 20 years, then this can be modeled by adding a flow from Condition Group 80 to Condition Group 60. Converter Rehab Length determines the total pipe rehabilitation length and converter Average Condition determines the weighted average condition for all network pipes using Equation (3). P PipeLengthsi  i Average Condition ¼ Pi i PipeLengthsi  i

Fig. 4 e Change in water demand implemented over the adjustment period.

which is the product of population served by the utility and per capita Water Demand, represents the volume of water that is billable and hence earns revenue for the utility. If a large proportion of network is in poor condition, significant volumes of water may be lost due to leakage. In this case the total volume of treated water pumped into the network will be higher than the Total Water Consumption. Additional costs associated with leaked water are included in operational expenditures as explained in the following section.

(3)

where i is the condition state and is equal to 20, 40, 60, 80 and 100. PipeLengthsi is the length of pipes in condition group i.

4.2.

4743

Consumer sector

The consumer sector estimates the water demand and use during the simulation period for a constant population. The average daily volume of water consumed per person is determined using stock Water Demand and flow Demand Change. Demand Change is a function of price Elasticity Of Demand, Minimum Demand and User Fees. Lipsey and Chrystal (1999) define Price Elasticity of Demand as the percentage change in demanded quantity of a good divided by the corresponding percentage change in price. Thus, the function Demand Change decreases Water Demand if user fees increase. The rationale for the water demand decrease is that consumers will implement water conservation measures (i.e. retrofitting of plumbing fixtures and the installation of water conserving appliances) to reduce water costs as user fees increase. It is also assumed that once water conservation measures are implemented that they will be permanent. Thus, water demand is assumed to remain constant at its minimum attained level even when user fees decrease. Price induced changes in water consumption are not instantaneous and occur over time. As shown in Fig. 4, a time delay parameter Demand Adjustment Period is implemented using a low initial rate of water consumption change followed by an accelerated rate of change that is followed by a low rate of change. The converter Minimum Demand is used to set a minimum water demand limit. Total Water Consumption

4.3.

Finance sector

The finance sector has two separate but interconnected stockflow structures-Funds Balance and User Fees. The stock Funds Balance represents the net funds at the end of each simulation time step and is replenished through Revenue inflow. For this analysis a constant volumetric user fee regime is used. The utility’s Revenue is calculated by multiplying the water volume consumed during a simulation time step by the user fees. Capital expenditures, CapEx, represent rehabilitation costs to move pipes from stock Condition Group 100 (poorest condition state) to stock Condition Group 20 (best condition state). Flow CapEx is calculated by multiplying the length of pipes moving through flow Renewal (Physical Infrastructure Sector discussed in Section 4.1 above) and unit price of rehabilitation (UnitPriceCapEx, dollars per unit length). Operational expenditures, OpEx, represent the cost of unaccounted water loss, treatment of infiltrated groundwater, water and wastewater treatment costs, pumping costs, maintenance expenditures (such as those incurred on flushing of pipes and minor repairs), and emergency expenditures (repair breaks and blockages, etc). Since operational costs increase with worsening pipe condition state the UnitPriceOpEx (the operational cost per unit length of a completely new pipe) is multiplied by the Condition Multiplier OpEx. In the demonstration model the exponential Condition Multiplier OpEx, shown in Fig. 5, is implemented to increase operational expenditures with increases in the Average Condition determined using Equation (3). Stock User Fees tracks the price per unit volume of water charged to consumers. User Fees are maintained at a constant

4744

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 3 7 e4 7 5 0

Fig. 5 e Condition multiplier for operational expenditures.

level throughout a simulation or allowed to vary at each time step. A financially self-sustainable utility implies maintaining a “zero” Funds Balance. This means that revenues at each time step are equal to operational and capital expenditures. To set revenues equal to expenditures, stock User Fees is adjusted using inflow User Fee Hike and outflow User Fee Decline. Equations used to develop the demonstration model are presented in Appendix 1.

5.

Demonstration model simulations

5.1.

Initial conditions and assumptions

Using the System Dynamic model, described in Section 4, a number of simulations are preformed to explore the impact of the interconnections and feedback loops on a hypothetical water and wastewater utility. The hypothetical utility is assumed to maintain 700 km of pipes which serve 100,000 consumers. This assumption is consistent with data reported in Burnside (2005). For this analysis, the pipe network length and customer base are considered constant over the simulation period. This assumption is deemed valid for the case where expansion of the pipe network is funded through development charges. Inflation is not considered in these simplified demonstrations. It is also assumed that the utility manager is only responsible for the water distribution and wastewater collection network. This analysis represents the scenario where the linear networks are owned and managed by a lower tier of municipal government and the water and wastewater treatment plants are managed by an upper tier government. In this case the upper tier government sells water and charges the lower water utility for treatment of discharged wastewater back to the upper tier. This case is applicable to several Canadian municipalities.

Table 1 provides the initial distribution of pipes in each condition group stock. All pipes are assumed to have an average service life of 100 years. The initial and minimum water demand are set at 300 and 200 L per capita per day (lpcd) respectively, which are in accordance with data reported in Environment Canada (2006). Capital and operational expenditure unit prices are set at $1,000 and $50 per metre, respectively, which are in accordance with cost functions reported in Burnside (2005). These unit prices are assumed constant during the simulations. Thus, the rate of appreciation of costs (inflation rate) is equal to the project depreciation rate needed to discount all costs to present value. A user fee of $3.75 per m3 is used to set initial revenues equal to expenditures. Heare (2007) suggests that estimation of full long-term costs of water services requires a time horizon of a century or more. For this analysis we also used a 100 year simulation period. Table 2 provides a summary of the three scenarios with variations that are described in the introduction. The demonstration model is used to explore three scenarios with three annual rehabilitation strategies: (Scenario 1) no capital works expenditure to rehabilitate water and wastewater pipes within the network; (Scenario 2) a 1% annual rehabilitation strategy that will replace the entire network every 100 years, assuming the average age of the pipe is 100 years; and (Scenario 3) no more than 5% of the network with pipes in Condition Group 100, which implies an annual rehabilitation rate of 1.18% of the network. For each scenario (Case A) user fees are maintained at $3.75 per m3 or (Case B) allowed to change so that revenues equal expenditures at each time step or (Case C) allowed to change so that revenues equal expenditures but with price elasticity of demand for water. Boland et al. (1984) indicate that price elasticity for residential water demand varies between 0.2 and 0.5. For this study, price elasticity is set at 0.35. For the price elastic simulations, a 20-year water demand adjustment period is applied.

5.2.

Simulation results

The zero percent rehabilitation strategy (1A, 1B and 1C) is a “do nothing” reactive maintenance management strategy where pipes are fixed at the time of failure. Scenario 1 simulation results are provided in Fig. 6. Fig. 6a shows the pipe network average condition along with the percentage of pipe network in each of the pipe condition stocks over a 100-year simulation period. This figure shows the network to have an initial average condition of 49 and that the average condition increases rapidly to 88 by year 60 and finally to 97 in year 100. Fig. 6a shows that the

Table 1 e Initial distribution of pipes in various condition groups. Pipe Groups

Length (kilometers) Fraction of Network (%)

Condition 20

Condition 40

Condition 60

Condition 80

Condition 100

140 20

280 40

140 20

105 15

34 5

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 3 7 e4 7 5 0

Table 2 e Summary of simulation scenarios. Scenario

1A 1B 1C 2A 2B 2C 3A 3B 3C

Rehabilitation Strategy (% of network replaced)

Zero Funds Balance Enforced

Price Elasticity of Demand (%/%)

0.00 0.00 0.00 1.00 1.00 1.00 1.18 1.18 1.18

No Yes Yes No Yes Yes No Yes Yes

0.00 0.00 0.35 0.00 0.00 0.35 0.00 0.00 0.35

percentage of pipes in stock Condition Group 100 increases rapidly from 5 to 89 percent in 100 years. Fig. 6b shows capital and operational expenditures along with the net funds balance over the 100-year simulation period. This figure shows the following:  Capital work expenditures are nil over the entire simulation period. This is reasonable since no funds are invested to rehabilitate the pipes.  Annual operational expenditures increase from $42 to $67 million. This is deemed reasonable as operational expenditures will increase with average network condition and the trend of the operational expenditures follows the average network condition curve in Fig. 6a.

4745

 Curve A in Fig. 6b shows that the funds balance initially starts at zero and then decreases rapidly to 1.5 billion dollars at 100 years, while curves B and C show that the net fund balance remains constant at zero over the entire 100 year simulation period. This confirms that the implemented Zero Fund Balance routine works as designed. The zero fund balance is accomplished by adjusting the user fees so that revenue equals to expenses in each time step. Fig. 6c shows the per cubic metre user fees of water and water and wastewater services over the 100 year simulation period. Curve A shows that the unit price of water is constant at $3.75 per m3 in real dollar terms. Curves B and C show how the user fees changes to create a zero funds balance without and with price elasticity respectively. Both curves B and C show an increasing user fee with time. This increasing user fee is required to increase revenue in step with increasing operational expenditures that result from deteriorating infrastructure. Curves B and C follow the same trend up to approximately 15 years where Curve C shows a rapid increase in user fee compared to Curve B. By the year 100, a price elasticity of 0.35 requires a user fee of $7.6 per m3 to balance the funds while zero price elasticity requires a user fee of $6.2 per m3. Fig. 6d shows the water demand over the 100 year simulation period. This figure shows that the water demand is constant at 300 lcpd when price elasticity is not enforced (curves A and B). Curve C shows that enforcing price elasticity results in the water demand decreasing from 300 lcpd to 242 lcpd in year 100. It should be noted that the utility’s revenues

Fig. 6 e Simulation results for scenarios 1.

4746

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 3 7 e4 7 5 0

are a function of water usage and reduced water consumption reduces revenues. To maintain a zero fund balance (revenues ¼ expenses) a higher user fee is required with price elasticity. The higher user fee ($7.6 vs $6.2 per m3) with price elasticity enforced is deemed reasonable. Figs. 7 and 8 show simulation results when proactive annual pipe network rehabilitation strategies are used. Specifically, Fig. 7 presents results for Scenario 2 which involves a 1% annual rehabilitation rate, and represents 100 percent pipe replacement in 100 years. Fig. 8 presents results for Scenario 3 where the annual rehabilitation rate is increased to 1.18% so that no more than 5% of the network has pipes in Condition Group 100 for the entire 100 year simulation period. For the proactive pipe rehabilitation scenarios (Scenarios 2 and 3), pipes in stock Condition Group 100 are rehabilitated in accordance with the rehabilitation criteria (i.e. 1.0% or 1.18%). In all simulations the length of pipe rehabilitated is set to the maximum of the length set by the rehabilitation strategy or the length of pipes in stock Condition Group 100. Fig. 7a shows the average network condition increases from 49 to 59 in 100 years. Fig. 8a shows the average network condition starts at 49 and slowly increases to 53 by the end of the simulation. It should be noted that the no rehabilitation strategy resulted in an average network pipe condition of 97 at year 100. Fig. 7a shows 5 percent of the network in stock Condition Group 100 at year 0 decreases to 3% at year 15 then increases to 18% at year 100. Fig. 8a shows stock Condition Group 100 decreases to 0% at year 15, remains constant at 0% until year 37 then increases linearly to 5% at year 100. An increasing stock Condition Group 100 indicates that more pipe lengths are arriving into the stock than rehabilitated and

a decreasing stock trend indicates that more pipes are rehabilitated than arriving into the stock. For the no rehabilitation option the percentage of network in stock Condition Group 100 increases rapidly to 89% in 100 years. Figs. 7 and 8b show that the fund balance is zero over the 100 year simulation period when the Zero Fund Balance routine is implemented (curves B and C). When the user fee is maintained at $3.75 per m3, the funds balance decreases to $0.9 billion when the rehabilitation is set at 1.0% and 1.18%. These funds balance deficits are significantly less than the $1.5 billion for the no rehabilitation strategy. Fig. 7b shows that the annual operational expenditures increase linearly from $41 to $44 million in year 100 while Fig. 8b shows the annual operational expenditures increase from $41 to $42 million in 100 years. Final year operational expenditures of $44 and $42 million are significantly less than the $66 million required for the no rehabilitation strategy. Figs. 7 and 8b show annual capital expenditures for the 1.0% and 1.18% rehabilitation options are $7.0 and $8.3 million respectively. For the no rehabilitation option the capital expenditure costs are $0 annually. Figs. 7 and 8c shows changes in annual user fees over the simulation period for the 1.0% and 1.18% rehabilitation scenarios operated on a financially self-sustaining basis (zero funds balance). The impact of price elasticity is shown in curves B and C. For both rehabilitation strategies, the user fee generally increases linearly to $4.6 per m3 when no price elasticity is considered and increases with a decreasing slope to $5.5 per m3 when price elasticity is considered. Figs. 7 and 8d show that water demand is constant at 300 lpcd when no price elasticity is considered (curves A and B)

60

User Fees ($/m3 )

60

40

40

20

20

0 0

c

80

20

40

60

80

0 100

Time (years)

7

C

5.5

5

B

4.6

4

A

3.75

3

0.0

40 Funds Balance OpEx CapEx

20

-0.5

A

-1.0 0 -1.5 20

40

60

80

100

Time (years)

d

8

6

B,C

0

Water Demand (lpcd)

80

0.5 60

A,B

3 00

30 0

2 80

28 0

C

2 60

26 0

2 40 0

20

40

60

Time (years)

80

100

24 0 0

20

40

60

Time (years) Fig. 7 e Simulation results for scenarios 2.

80

100

Funds Balance (billion $)

100 Average Condition Condition 20 Condition 40 Condition 60 Condition 80 Condition 100

Network's Average Condition

Pipe Groups (% of Network)

100

OpEx and CapEx (million $)

b

a

4747

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 3 7 e4 7 5 0

60

User Fees ($/m3 )

60 40

40

20

20

0 100

0 0

c

80

20

40

60

80

Time (years)

0.5 60

B,C

Funds Balance OpEx CapEx

20

7 6

C

5.5

5

B

4.6

4

A

3.75

3

A

0 20

40

80

-1.5 100

A,B

30 0

60

Time (years) 3 00

2 80

28 0

C

2 60

26 0

2 40 0

20

40

60

80

-0.5

-1.0

d

8

0.0

40

0

Water Demand (lpcd)

80

b

Funds Balance (billion $)

Average Condition Condition 20 Condition 40 Condition 60 Condition 80 Condition 100

OpEx and CapEx (million $)

100

100

Network's Average Condition

Pipe Groups (% of Network)

a

100

0

Time (years)

20

40

60

80

24 0 100

Time (years)

Fig. 8 e Simulation results for scenarios 3.

and initially rapidly decreases then levels off when price elasticity is considered. Final water demand for the 1.0% and 1.18% scenarios is 257 and 252 lpcd respectively. These values are higher than the 242 lpcd water demand for the no rehabilitation strategy (curve C in Fig. 6d).

6.

Discussion

Table 3 provides a summary of all simulation results at year 100. Regulations in Canada are forcing utilities to be financially sustainable and similar pressures are likely to occur or have occurred in developed countries. Case 1A shows that a constant user fee of $3.75 per m3 with no annual

rehabilitation strategy will result in the utility having a deficit of $1.5 billion at year 100. To make the utility financially sustainable, user fees need to be increased to $6.13 per m3 by year 100 (65% increase). When price elasticity is considered, users fees need to be increased to $7.59 per m3 by year 100 (102% increase). When a proactive annual pipeline rehabilitation strategy of 1.0% is adopted (Case 2B), a self-sustainable user fee of $4.65 per m3 is required at year 100. This represents a 24% increase in user fees. If price elasticity is considered (Case 2C), a user fee of $5.46 per m3 is required at year 100. This represents an increase of 46%. When the annual rehabilitation strategy is 1.18%, the simulation results for user fees are similar to the 1.0% rehabilitation cases.

Table 3 e Summary of results at year 100. Scenario

1A 1B 1C 2A 2B 2C 3A 3B 3C

Final User Fee ($/m3)

Funds Balance (billion $)

Final Water Demand (lpcd)

Cumulative Operational Expenditures (billion $)

Cumulative Capital Expenditures (billion $)

Cumulative Total Expenditures (billion $)

Network Average Condition

3.75 6.13 7.59 3.75 4.65 5.46 3.75 4.59 5.48

1.48 0.00 0.00 0.88 0.00 0.00 0.88 0.00 0.00

300 300 242 300 300 256 300 300 251

5.57 5.57 5.57 4.29 4.29 4.29 4.17 4.17 4.17

0.00 0.00 0.00 0.70 0.70 0.70 0.82 0.82 0.82

5.57 5.57 5.57 4.99 4.99 4.99 4.99 4.99 4.99

97 97 97 59 59 59 53 53 53

4748

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 3 7 e4 7 5 0

For the no rehabilitation strategy the total expenditure over 100 years is $5.57 billion. When 1.0% or 1.18% annual rehabilitation is adopted, the total expenditure over 100 years is $4.99 billion. This represents a $0.58 billion (10%) saving with significantly lower user fees at year 100. It is worth noting that cumulative expenditures at the end of the simulation are the same for the 1.0% and 1.18% rehabilitation strategies even though annual capital and operational expenditures are different. This is due to maintaining the network in a better condition state which reduces operational costs. When price elasticity is included, an increase in user fees causes water consumption to decrease (curve C representing elastic water demand in Figs. 6e8d). Reduced volume of water billed to customers yields lower revenues than required to match expenditures. Hence, funds balance decreases and user fees need to be increased. Due to the influence of loop R2 (Section 2.3), curve C (variable user fee with elastic demand) in Figs. 6e8c moves away from curve B (variable user fee with inelastic demand). This widening of gap between curves B and C continues for the case of no rehabilitation (Fig. 6c). However, in cases with proactive rehabilitation, the departure of curve C from curve B decreases and finally stops (slope of curve C in Figs. 7 and 8c decreases to finally become zero). This slowing trend of departure is due to feedback loop R3 (Section 2.6). This loop is not operative for scenario 1 because for that scenario, one of variables along loop R3 i.e. Infrastructure Rehabilitated remains zero. For scenarios 2 and 3, however, Infrastructure Rehabilitated continuously increases. As a result, infrastructure condition decreases (improves) which in turn causes operational expenditures to decrease. With reduced operational expenditures, fund balance increases. Since fund balance is an element common to both loops R2 and R3, the influence of loop R2 is mitigated by loop R3. The above discussion highlights the influence of two feedback loops (R2 and R3) on water and wastewater network management. Thus, this study demonstrates the complexity of the system. To model a complete system, more feedback loops need to be added to the model. For example in the demonstration model, to achieve zero fund balance, user fees were adjusted without any constraints. However, it may not be politically possible to implement large user fee hikes instantaneously. Thus, feedback loop B2 needs to be included in a more complete model. Once it is recognized that user fees may not always be at desired levels, it then follows that the constraint of funds available for infrastructure rehabilitation must be included. Thus, inclusion of loop B2 would necessitate capturing the influence of feedback loop B3. Similarly, another simplifying assumption in the demonstration model is to aggregate water and wastewater pipes. In a complete model, pipes can be classified according to criteria such as material, age and diameter. With such additional details, it is possible to incorporate deterioration curves to model movement of pipes among various stocks (Section 4.1). Accordingly, feedback loop R1 needs to be included. Finally, there may be other important feedback loops in addition to the ones discussed in Section 2 that are required to capture the complex and dynamic behavior of water and wastewater network management. For example, Canadian municipalities are allowed to borrow for financing capital

projects. Such a financing mechanism involves additional feedback loops to be considered. Once a complete model is validated and calibrated, it can be used to develop strategic plans to ensure water utilities are financially self-sustainable over the long-term.

7.

Conclusions

The following conclusions are drawn from this study: 1. New regulations in Canada mandate that water utilities are managed such that they are financially self-sustainable over the long-term. 2. Existing infrastructure management systems and tools reported in the literature are not capable of helping Canadian municipalities meet the requirements of the new regulations. 3. A causal loop diagram is developed that demonstrates water and wastewater network management is a complex system with many interconnections and feedback loops. This is the first known causal loop diagram developed for a financially self-sustainable water utility. 4. The System Dynamics approach is deemed an acceptable modeling method for water and wastewater network management. 5. A demonstration System Dynamics model is developed that highlights the significance of interconnections and feedback loops. This is the first known application of System Dynamics to water and wastewater network management. 6. A complete System Dynamics model needs to be constructed, validated and calibrated for a water utility before it is used to determine financial sustainability.

Acknowledgments We gratefully acknowledge the financial support provided by the Natural Sciences and Engineering Research Council of Canada (NSERC), the Ontario Graduate Scholarship, the University of Waterloo, and the Centre for Advancement of Trenchless Technologies located at the University of Waterloo. We also acknowledge the City of Niagara Falls and the City of Waterloo for their kind provision of data and sharing of valuable insights on water utility management.

Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.06.001.

references

Adeniran, E.A., Bamiro, O.A., 2010. A System Dynamics Strategic Planning Model for a Municipal Water Supply Scheme, Proc.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 3 7 e4 7 5 0

28th International Conference of the System Dynamics Society, Seoul (Korea). 25e29 July 2010. http://www. systemdynamics.org/conferences/2010/proceed/papers/ P1017.pdf (accessed 20.10.10.). Ahmad, S., Prashar, D., 2010. Evaluating municipal water conservation policies using a dynamic simulation model. Water Resources Management 24 (13), 3371e3395. Barraque, B., LeBris, C., 2007. Water sector regulation in France. CESifo DICE Report 5 (2), 4e12. Boland, J.J., Dziegielewski, B., Baumann, D.D., Opitz, E.M., 1984. Influence of Price and Rate Structures on Municipal and Industrial Water Use. Report submitted to the U.S. Army Corps of Engineers, Institute for Water Resources, Fort Belvoir (Virginia-USA). http://www.iwr.usace.army. mil/inside/products/pub/iwrreports/84-C-2.pdf accessed 13.05.10. Burn, S., Tucker, S., Rahilly, M., Davis, P., Jarrett, R., Po, M., 2003. Asset planning for water reticulation systems e the PARMS model. Water Science and Technology-Water Supply 3 (1e2), 55e62. Burnside, 2005. Water and Wastewater Asset Cost Study. Ministry of Public Infrastructure Renewal (Ontario-Canada), prepared by R. J. Burnside & Associates Limited. http://www.mei.gov.on. ca/en/pdf/infrastructure/water/water-asset-cost-study-e.pdf accessed 30.07.09. CBO, 2002. Future Investment in Drinking Water and Wastewater Infrastructure, A Congressional Budget Office (CBO) study. The Congress of the United States. http://www.cbo.gov/ftpdocs/ 39xx/doc3983/11-18-WaterSystems.pdf accessed 18.01.09. CICA, 2007. Guide to Accounting for and Reporting Tangible Capital Assets e Guidance for Local Governments and Local Government Entities that Apply the Public Sector Handbook. Public Sector Accounting Group of the Canadian Institute of Chartered Accountants (CICA). http://www.psab-ccsp.ca/ other-non-authoritative-guidance/item14603.pdf accessed 18. 04.08. Dandy, G.C., Engelhardt, M.O., 2006. Multi-objective trade-offs between cost and reliability in the replacement of watermains. Journal of Water Resources Planning and Management 132 (2), 79e88. Deb, A.K., Hasit, Y.J., Grablutz, F.M., Herz, R., 1998. Quantifying Future Rehabilitation and Replacement Needs of Watermains. AWWA Research Foundation, Denver, CO, p. 156. Englehardt, M., Savic, D., Skipworth, P., Cashman, A., Saul, A.J., Walters, G.A., 2003. Whole life cycle costing: application to water distribution network. Water Science and Technology Water Supply 3 (1e2), 87e93. Environment Canada, 2006. Municipal Water and Wastewater Survey. http://www.ec.gc.ca/Water-apps/MWWS/en/ publications.cfm accessed 13.05.10. FHWA, 2000. Primer: GASB 34. Technical report. Federal Highway Administration (FHWA), Office of Asset Management, U.S. Department of Transportation,. http://isddc.dot.gov/OLPFiles/ FHWA/010019.pdf accessed 07.04.11. Forrester, J.W., 1958. Industrial dynamics: a major breakthrough for decision makers. Harvard Business Review 36 (4), 37e66. Fortin, M., Slack, E., Loudon, M., Kitchen, H., 2001. Financing Water Infrastructure. A report commissioned by the Walkerton Inquiry (Ontario-Canada). http://www.ontla.on.ca/ library/repository/mon/1000/10295208.pdf accessed 05.09.10. Giustolisi, O., Laucelli, D., Savic, D.A., 2006. Development of rehabilitation plans for watermains replacement considering risk and cost-benefit assessment. Civil Engineering and Environmental Systems 23 (3), 175e190. Grigg, N.S., Bryson, M.C., 1975. Interactive simulation for water dynamics. ASCE Journal of Urban Planning and Development 101 (1), 77e92.

4749

Grigg, N.S., 2003. Water, Wastewater, and Stormwater Infrastructure Management. Lewis Publishers, Boca Raton (Florida-USA), p. 242. Guest, J.S., Skerlos, S.J., Daigger, G.T., Corbett, J.R.E., Love, N.G., 2010. The use of qualitative system dynamicsto identify sustainability characteristics of decentralized wastewater management alternatives. Water Science and Technology 61 (6), 1637e1644. Hadzilacos, T., Kalles, D., Preston, N., Melbourne, P., Camarinopoulos, L., Eimermacher, M., Kallidromitis, V., Frondistou-Yannas, S., Saegrov, S., 2000. UtilNets: a watermains rehabilitation decision-support system. Computers, Environment and Urban Systems 24 (3), 215e232. Halfawy, M.M.R., Newton, L.A., Vanier, D.J., 2006. Review of commercial municipal infrastructure asset management systems. Electronic Journal of Information Technology in Construction 11, 211e224. Hannon, B., Ruth, M., 1994. Dynamic Modeling. Springer-Verlag, New York (USA), p. 248. Heare, S., 2007. EPA communique´ e achieving sustainable water infrastructure. Journal of American Water Works Association 99 (4), 24e26. 28. Howard, R.J., 2001. Infrastructure asset management under Australian accounting Standard 27 (AAS27). Proceedings of the Institution of Civil Engineers Municipal Engineer 145 (4), 305e310. Kleiner, Y., Adams, B.J., Rogers, J.S., 1998. Selection and scheduling of rehabilitation alternatives for water distribution systems. Water Resources Research 34 (8), 2053e2061. Kleiner, Y., Nafi, A., Rajani, B., 2010. Planning renewal of watermains while considering deterioration, economies of scale and adjacent infrastructure. Water Science and Technology Water Supply 10 (6), 897e906. Kotz, C., Hiessl, H., 2005. Analysis of system innovation in urban water infrastructure systems: an agent-based modeling approach. Water Science and Technology Water Supply 5 (2), 135e144. Linerand, B., deMonsabert, S., 2010. Balancing the triple bottom line in water supply planning for utilities. Journal of Water Resources Planning and Management. doi:10.1061/(ASCE)WR. 1943-5452.0000128. Lipsey, R.G., Chrystal, K.A., 1999. Principles of Economics, ninth ed. Oxford University Press, New York (USA), p. 656. MacDonald, D.H., Barnes, M., Bennett, J., Morrison, M., Young, M. D., 2003. What Consumers Value Regarding Water Supply Disruptions: A Discrete Choice Analysis?, Commonwealth Scientific Industrial Research Organisation (CSIRO), Glen Osmond (Australia), 18 p. http://www.cmis.csiro.au/Mary. Barnes/pdf/WTP_200305.pdf accessed 19.11.08. Ministry of the Environment Ontario, 2007. Toward Financially Sustainable Drinking-Water and Wastewater Systems. Financial Plans Guideline EBR Registry Number: 010e0490. Ministry of Environment, Ontario, Canada. http://hdl.handle. net/1873/9243 accessed 21.04.08. Moglia, M., Burn, S., Meddings, S., 2006. Decision support system for water pipeline renewal prioritisation. Electronic Journal of Information Technology in Construction 11, 237e256. Savic, D., Giustolisi, O., Berardi, L., Shepherd, W., Djordjevic, S., Saul, A., 2006. Modelling sewer failure by evolutionary computing. Water Management Journal 159 (2), 111e118. Saegrov, S. (Ed.), 2005. CARE-W: Computer Aided Rehabilitation for Water Networks. IWA Publishing, London, UK, p. 208. Saegrov, S. (Ed.), 2006. CARE-S: Computer Aided Rehabilitation of Sewer and Storm Water Networks. IWA Publishing, London, UK, p. 160. Shamir, U., Howard, C.D.D., 1979. An analytical approach to scheduling pipe replacement. Journal of American Water Works Association 71, 248e258.

4750

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 3 7 e4 7 5 0

Sterman, J., 2000. Business Dynamics: Systems Thinking and Modeling for a Complex World. Irwin/McGraw-Hill, Boston (Massachusetts-USA), p. 982. Tabesh, M., Soltani, J., Farmani, R., Savic, D., 2009. Assessing pipefailure rate and mechanical reliability of water distribution networks using data-driven modeling. Journal of Hydroinformatics 11 (1), 1e17. Wirahadikusumah, R., Abraham, D.M., 2003. Application of dynamic programming and simulation for sewer

management. Engineering Construction and Architectural Management 10 (3), 193e208. Younis, R., Knight, M.A., 2010a. Continuation ratio model for the performance behavior of wastewater collection networks. Tunnelling and Underground Space Technology 25, 660e669. Younis, R., Knight, M.A., 2010b. Probability model for investigating the trend of structural deterioration of wastewater pipelines. Tunnelling and Underground Space Technology 25, 670e680.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 5 1 e4 7 5 7

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Effect of chlorination on endotoxin activities in secondary sewage effluent and typical Gram-negative bacteria Huang Huang a, Qian-Yuan Wu a, Yang Yang a, Hong-Ying Hu a,b,* a

State Key Joint Laboratory of Environmental Simulation and Pollution Control, School of Environment, Tsinghua University, Beijing 100084, PR China b Graduate School at Shenzhen, Tsinghua University, Shenzhen 518055, PR China

article info

abstract

Article history:

Wastewater reuse is a viable and attractive method to address water shortage problems.

Received 20 November 2010

However, wastewater can have high endotoxin activity. Endotoxins are toxic inflammatory

Received in revised form

agents and are considered a risk factor for wastewater reuse. In this study, the effect of

29 May 2011

chlorination on endotoxin activity in secondary sewage effluent was evaluated by Limulus

Accepted 13 June 2011

(Tachypleus tridentatus) Amebocyte Lysate assay. It was found that chlorination could not

Available online 21 June 2011

decrease endotoxin activity of secondary effluent effectively under the conditions employed in this study. Chlorination of a pure cultured Gram-negative bacterium (Escher-

Keywords:

ichia coli), and a Gram-negative bacterium isolated from secondary sewage effluent,

Endotoxin

resulted in a significant increase in endotoxin activity, suggesting that the presence of

Chlorination

Gram-negative bacteria contributed substantially to endotoxin activity, masking any

Secondary sewage effluent

potential reduction that may be attributable to chlorination. Furthermore, the activities of

Gram-negative bacteria

both free and cell-bound endotoxins in pure culture increased significantly during chlori-

Wastewater reuse

nation due to cell wall damage induced by chlorination. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

The contradiction between water shortage and increasing water consumption is one of the major social and economic problems in many countries (Higgins et al., 2002). Thus wastewater reuse has gained particular importance in the sustainable management of water resources (Zhang and Farahbakhsh, 2007; Sun et al., 2009). Wastewater, however, contains pathogens and toxic chemicals, making the use of reclaimed water risky. Eliminating pathogens in reclaimed water has been of great concern (Zhang and Farahbakhsh, 2007; Costa´n-Longaresa et al., 2008), however, the harmful components of microorganisms such as endotoxins have not yet entered widespread consciousness. Endotoxins are complexes of lipopolysaccharide (LPS) that constitute the outer layer of the cell wall of most Gram-

negative bacteria and some cyanobacteria (Williams, 2007; Anderson et al., 2002; Rapala et al., 2002; Gehr et al., 2008; O’Toole et al., 2008). LPS molecules occupy an area of approximately three-quarters of the total outer surface area of a bacterial cell (Williams, 2007), and the dry weight of LPS is about 3.6% of a bacterial cell (Narita et al., 2005). LPS complexes are macromolecules composed of three main regions: Lipid A, core polysaccharide and O antigens. The biological activity of endotoxins is dependent on the structure of Lipid A (Brandenburg and Wiese, 2004; Gorbet and Sefton, 2005). As endotoxins can be released from cells by multiplication, death and lysis, endotoxins occur in both cell-bound and free forms in aquatic systems (Mattsby-baltzer et al., 1991; Anderson et al., 2002; Venter et al., 2006; Parikh and Chorover, 2007), and both forms have biological activity (Morrison et al., 1994; Anderson et al., 2002).

* Corresponding author. Graduate School at Shenzhen, Tsinghua University, Shenzhen 518055, PR China. Tel.: þ86 10 6279 4005; fax: þ86 10 6279 7265. E-mail address: [email protected] (H.-Y. Hu). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.06.013

4752

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 5 1 e4 7 5 7

Endotoxins are toxic inflammatory agents that can activate numerous cellular and humoral mediated systems (Morrison et al., 1994; Liao et al., 2010). Several clinical symptoms, including fever, diarrhea, vomiting, cough, breathing difficulties, shock, intravascular coagulation, and even death, can be induced by endotoxins (Morrison et al., 1994; Anderson et al., 2002; Liebers et al., 2008). Only 1e10 ng/kg body weight (intravenously), roughly equal to 10e100 endotoxin units (EU)/ kg body weight when assuming a conversion factor of 10 EU/ ng, can induce fever (Anderson et al., 2002). Recently, endotoxins have been found to augment the toxicity of some chemicals such as microcystins (Roth et al., 1997; Best et al., 2002). Accordingly, some standards on the threshold limits for endotoxins have been proposed. The China Pharmacopoeia, the British Pharmacopoeia and the United States Pharmacopoeia have established limits of 0.25 EU/mL for injection water (Chinese Pharmacopoeia Commission, 2005; Anderson et al., 2002). The Dutch Health Council has proposed a health-based airborne endotoxin exposure limit of 50 EU/m3 (Health Council of the Netherlands, 1998). Previous studies have indicated that high endotoxin activity can be detected in wastewater treatment plant effluents, typically ranging from 300 to 20,000 EU/mL (O’Toole et al., 2008; Guizani et al., 2009), therefore, the control of endotoxins should be considered for the safety of wastewater reuse. Chlorination has been widely used in wastewater reclamation as the last barrier, aiming to inactivate pathogenic microorganisms and reduce the transmission of waterborne infectious diseases. Accordingly, research on chlorination has focused on the inactivation of microorganisms, but information about the effect of chlorination on endotoxins is limited. Previous studies mainly focused on the removal efficiency of endotoxins in drinking water by chlorination. They found that chlorination only had a small effect on endotoxin inactivation under typical drinking water disinfection conditions (Rapala et al., 2002; Gehr et al., 2008; Anderson et al., 2003). The composition of wastewater is more complex than that of drinking water, and the chlorination conditions employed in wastewater treatment differ from those of drinking water. Therefore, the influence of chlorination on endotoxin activity in wastewater requires advanced study. The aim of this study was to assess the effect of chlorination on endotoxin activity in secondary sewage effluent. The effect of chlorination on endotoxin activity in pure cultured Gram-negative bacteria was also evaluated to understand the changes of the endotoxin activity in the effluent.

2.

Materials and methods

2.1.

Glassware and solutions preparation

All glassware used in this study was made endotoxin-free by heating at 250  C for 1 h. Other products such as pipette tips and microplates were purchased endotoxin-free (Xiamen Houshiji Ltd, China). Solutions related to the endotoxin activity assay, such as phosphate buffer solution (PBS) and chlorine stock solution, were prepared in Milli-Q UF water, which was tested and found to be endotoxin-free.

2.2.

Water samples

The non-disinfected wastewater samples were collected from the secondary effluent of a domestic wastewater treatment plant using an anaerobic-anoxic-oxic process. The samples were placed in a container filled with ice, transported to the laboratory, and stored at 4  C in the laboratory prior to the chlorination experiments performed within 12 h. The characteristics of the secondary effluent samples were measured and shown as follow: pH: 7.7e8.0, TOC: 6.9e7.1 mg/L, COD: 112e117 mg/L, Ammonia nitrogen: 15e17 mg/L, TSS: 4.8e7.7 mg/L.

2.3.

Strains and culture

Escherichia coli (E. coli) CGMCC 1.3373 obtained from the China General Microbiological Culture Collection Center was used in this study. An unidentified strain of Gram-negative bacteria isolated from the secondary effluent sample described above was also used. The isolated strain was subsequently identified as belonging to the Genus Enterobacter. The frozen stock suspension of the bacteria was inoculated in nutrient broth, and the subculture was grown at 37  C for 16 h at 150 rpm. The bacterial cells were then harvested by centrifugation at 10,000 rpm for 10 min at 4  C. After removing the broth, the cell pellet was rinsed and resuspended in PBS (pH 7.4). The bacterial suspension was then used for the chlorination experiments.

2.4.

Chlorination

Chlorination experiments were conducted to investigate the effects of contact time and chlorine dosage on endotoxin activity. To evaluate the effect of contact time, a chlorine dosage of 10 mg/L was selected, and contact time ranged from 10 to 300 min. To evaluate the effect of chlorine dosage, a contact time of 30 min was selected, and the chlorine dosage was changed from 0 to 50 mg/L. The chlorination experiments were performed in 100 mL Erlenmeyer flasks with glass stoppers. The flasks were mixed in a constant temperature incubator shaker (Harbin Donglian Electronic & Technology Development Co., Ltd, China.) at 25  C at 150 rpm. The samples of 50 mL secondary effluent or bacterial suspension were poured into the flask; except for the chlorine-free control, the appropriate amount of sodium hypochlorite was added. At the end of the pre-determined contact time, excess sodium thiosulfate was added to quench any remaining chlorine residual. The flasks were mixed for another 5 min to ensure that all chlorine was quenched. Afterward, 1.5 mL aliquots were removed for bacterial enumeration, and 2 mL aliquots were transferred into a pyrogen-free glass tube covered and laid in a 4  C refrigerator for the endotoxin activity assay (O’Toole et al., 2009).

2.5.

Endotoxin activity assay

Endotoxin activity was analyzed using chromogenic Limulus (Tachypleus tridentatus) Amebocyte Lysate (LAL) endpoint assay kits (Xiamen Houshiji Ltd, China). Approximately 2e3 serial dilutions (1:100e1:100,000) of each sample in pyrogen-free water were analyzed based on the expected endotoxin

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 5 1 e4 7 5 7

concentrations. The dilutions of the samples were mixed with LAL, incubated for 8 min at 37  C, mixed with the chromogenic substrate, and incubated for an additional 6 min; the reaction was stopped with 20% acetic acid. The absorbance of the developed yellow color was determined at 405 nm in a microplate reader (Molecular Devices, Inc, USA, MD SpectraMax M5). The endotoxin activity was obtained from the standard curve (0.1e1 EU/mL) using endotoxin standards from the E. coli O111:B4 strain. The differences in endotoxin activity between different samples were considered statistically significant when p < 0.05 according to the Student’s t-test. To investigate the changes in free and cell-bound endotoxin concentrations after chlorination, the free and bound endotoxin activities of the samples were assayed, respectively. The E. coli suspension samples (before and after chlorination) were first assayed for their (total) endotoxin activity. Afterward, 200 mL suspensions were centrifuged at 12,000 g for 10 min, and the supernatants were assayed for free endotoxin activity (Evans et al., 1978; Anderson et al., 2002). The supernatants were then removed, and the cell pellets were rinsed and resuspended in 200 mL pyrogen-free water (Milli-Q UF water). The new suspensions were assayed as bound endotoxin activity (Jorgensen et al., 1973; Venter et al., 2006).

2.6.

Bacteria count

The total number of bacteria in the secondary effluent samples was determined by the pour plate method. The samples were diluted and plated on nutrient agar. Then the plates were incubated at 37  C for 24 h (Guo et al., 2009). The cell counts of E. coli or the isolated Gram-negative bacterium in pure cultures were enumerated using the same method. The differences in bacteria counts between different samples were considered statistically significant when p < 0.05 according to the Student’s t-test.

2.7.

Transmission electron microscopy (TEM)

The E. coli suspensions, before and after chlorination, were prepared for TEM. The cells were centrifuged at 5000 rpm for 10 min. The pellet was then fixed, rinsed, dehydrated, infiltrated, and embedded. Thin 50e70 nm sections were cut with a glass knife on a Lecia EM UC6 ultramicrotome. After staining with lead citrate and uranyl acetate, the samples were imaged using a Hitachi H-7650 TEM.

3.

Results and discussion

3.1. Effect of chlorination on endotoxin activity in secondary effluent In order to investigate the effect of chlorination on endotoxin activity in secondary sewage effluent, different chlorination conditions were conducted. Unless specified otherwise, endotoxin activity is reported as the total activity. Fig. 1 shows the effect of contact time on the endotoxin activity and total bacteria count in secondary effluent. The chlorine dose was 10 mg/L, and the total chlorine residuals ranged from 8.6 mg/L after 10 min to 6.2 mg/L after 120 min. The total bacteria count

4753

Fig. 1 e Changes of endotoxin activity and total bacteria count versus contact time during chlorination of secondary effluent. Numbers above the symbols represent the total chlorine residuals (mg/L) after corresponding contact time. Asterisks indicate that the endotoxin activity or bacteria count of the chlorinated samples are significantly different from that of the original sample ( p < 0.05). Error bars represent the standard deviation based on triplicate analyses.

sharply decreased from 105 to 102 CFU/mL after chlorination. The initial endotoxin activity was about 1000 EU/mL. The endotoxin activity did not decrease over the 120 min contact time, indicating that increasing chlorination time is not an effective method to inactivate the endotoxins in secondary effluent. The effect of chlorine dosage on endotoxin activity in the secondary effluent was also investigated, as shown in Fig. 2. The total chlorine residuals after 30 min of contact time were 4.3, 7.9 and 39 mg/L at 5, 10 and 50 mg/L chlorine doses, respectively; And the total bacteria count was reduced to 101, 90 and 81 CFU/mL respectively. However, the endotoxins were not inactivated regardless of the chlorine dose. In fact, the endotoxin activity slightly increased at 5 mg/L chlorine dose ( p < 0.05). This suggests that increasing the chlorine dosage cannot inactivate the endotoxin in the secondary effluent. In a previous study by Anderson et al. (2003), a linear decrease of endotoxin activity with chlorination time was observed but only over long periods of exposure on the order of days to weeks (inactivation rate was 1.3e1.4 EU/(mL h)). In that study a pure endotoxin standard in endotoxin-free distilled water was examined in the absence of bacteria and other chemicals which could interfere with the actual measurement of the endotoxin degradation rate associated with free chlorination. In a separate study, when raw water collected following a full-scale drinking water treatment plant filter was chlorinated at doses used for drinking water disinfection, it was found that the endotoxin activity fluctuated with Ct (i.e., the product of the free chlorine residual and contact time) values, and the removals of endotoxins were only 11e25% but in some cases actually increased (Gehr et al.,

4754

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 5 1 e4 7 5 7

Fig. 2 e Changes of endotoxin activity and total bacteria count versus chlorine dosage during chlorination of secondary effluent. Numbers above the symbols represent the total chlorine residuals (mg/L) at corresponding chlorine dosage. Asterisks and error bars represent the same items as those in Fig. 1.

2008). The Gehr et al. study suggests that chlorine would have little effect on endotoxins under real drinking water treatment conditions. In this study, unfiltered secondary effluent was investigated. The components of secondary effluent are much more complex than in the waters used in the studies referred to above which utilized deionized water, drinking water, or raw surface water. The effluent contained not only various chemicals but also microorganisms. The microorganisms in the effluent may have kept the endotoxin activity from decreasing after chlorination. Therefore, the effect of chlorination on endotoxin activity from microorganisms, especially Gram-negative bacteria, needs to be further evaluated.

Fig. 3 e Changes of endotoxin activity and cell counts versus contact time during chlorination of pure cultured Gram-negative bacteria. Asterisks and error bars represent the same items as those in Fig. 1.

Fig. 4 e Changes of endotoxin activity and cell counts versus chlorine dosage during chlorination of pure cultured Gram-negative bacteria. The ranges of total chlorine residuals in the E. coli suspension: 1.8e2.8 mg/L at 5 mg/L chlorine dose, 5.8e7.5 mg/L at 10 mg/L chlorine dose, 15.2e17.3 mg/L at 20 mg/L chlorine dose, 43.2e46.7 mg/L at 50 mg/L chlorine dose; The ranges of total chlorine residuals in the isolated strain suspension: 2.1e2.3 mg/L at 5 mg/L chlorine dose, 6.7e7.2 mg/L at 10 mg/L chlorine dose, 17.1e18.6 mg/L at 20 mg/L chlorine dose. Asterisks and error bars represent the same items as those in Fig. 1.

3.2. Effect of chlorination on endotoxin activity of the pure cultured Gram-negative bacteria Two Gram-negative bacterial strains were selected: E. coli and a strain isolated from the secondary sewage effluent studied in this research. The strains were disinfected at different contact times and at different chlorine doses. Fig. 3 shows the cell counts and endotoxin activities of the two bacteria at different contact times. After chlorination for 30 min, the cell

Fig. 5 e Changes of endotoxin activities and cell counts of E. coli after chlorination at 50 mg/L chlorine dosage for 30 min. Asterisks and error bars represent the same items as those in Fig. 1.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 5 1 e4 7 5 7

counts of E. coli decreased to the detection limit, suggesting that chlorination can effectively inactivate E. coli. However, the endotoxin activity of E. coli increased from 275 to 1662 EU/ mL within the initial 30 min, and then remained constant over time. Since the cell counts and endotoxin activity of chlorinated E. coli did not change with contact time increasing from 30 to 300 min, the isolated strain samples were tested with shorter contact times (0e60 min). Similar results were observed for the chlorination of the isolated strain. With the increase of contact time, the cell counts of the isolated strain decreased from 108 CFU/mL to undetectable, whereas the endotoxin activity increased gradually from 970 to 9669 EU/mL. The effect of chlorine dosage on endotoxin activity and cell counts of the two bacteria is shown in Fig. 4. With chlorine addition ranging from 0 to 20 mg/L, the endotoxin

4755

activity of E. coli gradually increased from about 200 to 1000 EU/mL, increasing about 4 times; when the dosage increased to 50 mg/L, the endotoxin activity remained stable at 1000 EU/mL. The increase of endotoxin activity after chlorination of pure cultured bacteria was confirmed by the results of the isolated strain. With chlorine addition ranging from 0 to 20 mg/L, the cell counts of the isolated strain decreased from more than 107 CFU/mL to undetectable, and the endotoxin activity increased significantly from 463 to 8387 EU/mL. As the contact time or chlorine dosage increased, the endotoxin activity gradually increased with the inactivation of the pure cultured Gram-negative bacteria. It could be deduced that the effect of chlorination on the Gram-negative bacteria contributed to the increase of endotoxin activity in chlorinated secondary effluent. However, the endotoxin activity of

Fig. 6 e TEM images of E. coli cells (magnification, 40,0003 for a and c, 100,0003 for b and d). (a), (b) E. coli cells before chlorination. (c), (d) E. coli cells after 30 min contact time, at an initial chlorine dose of 50 mg/L (the range of total chlorine residual: 29.4e32.9 mg/L), arrows show perforations within the cell wall.

4756

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 5 1 e4 7 5 7

secondary effluent did not increase as regularly as that of pure cultures, probably because of the lower concentration of Gram-negative bacteria and more complex composition of the secondary effluent. The increase of endotoxin activity in pure culture after chlorination implied that chlorination might increase the risk of endotoxin in wastewater, especially in waters with high numbers of Gram-negative bacteria.

3.3. Changes of cell-bound and free endotoxin activities during chlorination In order to determine why the endotoxin activity of pure cultured Gram-negative bacteria increased during chlorination, the cell-bound and free endotoxin activities of E. coli were measured. A high dose of chlorine (50 mg/L, 30 min) was selected. The cell counts decreased, and the total endotoxin activity significantly ( p < 0.05) increased after chlorination. The activities of both cell-bound and free endotoxins also increased significantly ( p < 0.05), more than 2 times and 6 times, respectively (Fig. 5). Endotoxins are shed in large amounts during cell death (Gorbet and Sefton, 2005), and chlorination can cause the death of bacteria. Therefore, after chlorination, more free endotoxins were released in the supernatant, and the activity of free endotoxin increased. Anderson et al. (2008) also showed that some endotoxins were released after chlorination. It was found that chlorination could inactivate endotoxin standard at an inactivation rate of 1.4 EU/(mL h) when using 100 mg/L chlorine (Anderson et al., 2003), indicating that chlorination can break down some free endotoxins. However, the inactivation rate was very low, and considering the contact time, the reduction of free endotoxins would be negligible in this study. It is interesting to note that the activity of cell-bound endotoxin also increased after chlorination. Endotoxins are released from cells after chlorination, and thus the amount of cell-bound endotoxins should decrease, but the activity of cell-bound endotoxin increased. The increase of activity might be due to the effect of chlorination on the active group in endotoxin molecules. The biologically active Lipid A portion of the endotoxin molecule is embedded in the outer membrane of the cell, whereas the polysaccharide chain extends into the cell’s environment (Williams, 2007). At the dosages applied in this study, the chlorine may have damaged bacterial cell walls which in turn could have exposed Lipid A to the environment, activating the LAL, leading to an increase in the activity of cell-bound endotoxin. This was supported by pre- and post-chlorination TEM images of E. coli (Fig. 6). As can be seen, the cell walls were much thinner, and perforations within the cell wall can be seen following chlorination, implying the release of endotoxin molecules and exposure of the active Lipid A group.

4.

Conclusions

In this study, the influence of chlorination on endotoxin activities of secondary sewage effluent and pure cultured

Gram-negative bacteria was investigated. The following conclusions are made: (1) The endotoxin activity of the secondary sewage effluent did not decrease after chlorination. Neither increasing contact time nor increasing chlorine dosage could reduce the endotoxin activity of the effluent. (2) Endotoxin activity of pure cultured E. coli and the isolated Gram-negative bacterium increased significantly after chlorination, implying that chlorination might increase the concentration of endotoxin in wastewater, especially waters with high numbers of Gram-negative bacteria. (3) The activities of free and cell-bound endotoxins in pure cultured E. coli both increased after chlorination. The release of endotoxin molecules from cells led to the increase of free endotoxin activity, and the presence of previously un-exposed Lipid A from cell-bound endotoxin may have contributed to the increase of cell-bound endotoxin activity as detected by the LAL assay.

Acknowledgments This study was funded by the Chinese National Science Fund for Distinguished Young Scholars (No. 50825801) and the National High-tech R&D Program of China (863 Program) (No. 2008AA062502). The authors thank Ying Li from the Center of Biomedical Analysis of Tsinghua University for the TEM operation.

references

Anderson, W.B., Slawson, R.M., Mayfield, C.I., 2002. A review of drinking-water-associated endotoxin, including potential routes of human exposure. Can. J. Microbiol. 48 (7), 567e587. Anderson, W.B., Mayfield, C.I., Dixon, D.G., Huck, P.M., 2003. Endotoxin inactivation by selected drinking water treatment oxidants. Water Res. 37 (19), 4553e4560. Anderson, W.B., Mayfield, C.I., Huck, P.M., 2008. Endotoxin release from biologically active bench-scale drinking water anthracite/sand filters. J. Water Supply Res. Technol. Aqua 57 (8), 585e597. Best, J.H., Pflugmacher, S., Wiegand, C., Eddy, F.B., Metcalf, J.S., Codd, G.A., 2002. Effects of enteric bacterial and cyanobacterial lipopolysaccharides, and of microcystin-LR, on glutathione S-transferase activities in zebra fish (Danio rerio). Aquat. Toxicol. 60 (3e4), 223e231. Brandenburg, K., Wiese, A., 2004. Endotoxins: relationships between structure, function, and activity. Curr. Top. Med. Chem. 4 (11), 1127e1146. Chinese Pharmacopoeia Commission, 2005. in Chinese. Pharmacopoeia of the People’s Republic of China 2005, vol. II. Chemical Industry Press, Beijing. 76e82. Costa´n-Longaresa, A., Montemayora, M., Paya´na, A., Me´ndeza, J., Jofrea, J., Mujeriegob, R., Lucenaa, F., 2008. Microbial indicators and pathogens: removal, relationships and predictive capabilities in water reclamation facilities. Water Res. 42 (17), 4439e4448.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 5 1 e4 7 5 7

Evans, T.M., Schillinger, J.E., Stuart, D.G., 1978. Rapid determination of bacteriological water quality by using Limulus Lysate. Appl. Environ. Microbiol. 35 (2), 376e382. Gehr, R., Parent Uribe, S., Da Silva Baptista, I.F., Mazer, B., 2008. Concentrations of endotoxins in waters around the island of Montreal, and treatment options. Water Qual. Res. J. Can. 43 (4), 291e303. Gorbet, M.B., Sefton, M.V., 2005. Endotoxin: the uninvited guest. Biomaterials 26 (34), 6811e6817. Guizani, M., Dhahbi, M., Funamizu, N., 2009. Assessment of endotoxin activity in wastewater treatment plants. J. Environ. Monit. 11 (7), 1421e1427. Guo, M.T., Hu, H.Y., Liu, W.J., 2009. Preliminary investigation on safety of post-UV disinfection of wastewater: bio-stability in laboratory-scale simulated reuse water pipelines. Desalination 239 (1e3), 22e28. Health Council of the Netherlands, 1998. Dutch Expert Committee on Occupational Standards (DECOS). Endotoxins. Rijswijk: Health Council of the Netherlands. Publication No. 1998/03 WGD. Higgins, J., Warnken, J., Sherman, P.P., Teasdale, P.R., 2002. Survey of users and providers of recycled water: quality concerns and directions for applied research. Water Res. 36 (20), 5045e5056. Jorgensen, J.H., Carvajal, H.F., Chipps, B.E., Smith, R.F., 1973. Rapid detection of Gram-negative bacteriuria by use of the Limulus endotoxin assay. Appl. Microbiol. 26 (1), 38e42. Liao, V.H.C., Chou, W.C., Chio, C.P., Ju, Y.R., Liao, C.M., 2010. A probabilistic approach to quantitatively assess the inhalation risk for airborne endotoxin in cotton textile workers. J. Hazard. Mater. 177 (1e3), 103e108. Liebers, V., Raulf-Heimsoth, M., Bru¨ning, T., 2008. Health effects due to endotoxin inhalation (review). Arch. Toxicol. 82 (4), 203e210. Mattsby-Baltzer, I., Lindgren, K., Lindholm, B., Edebo, L., 1991. Endotoxin shedding by enterobacteria: free and cell-bound endotoxin differ in Limulus activity. Infect. Immun. 59 (2), 689e695. Morrison, D.C., Danner, R.L., Dinarello, C.A., Munford, R.S., Natanson, C., Pollack, M., Spitzer, J.J., Vogel, S.N.,

4757

McSweegan, E., 1994. Bacterial endotoxins and pathogenesis of gram-negative infections: current status and future direction. J. Endotoxin Res. 1 (2), 71e83. Narita, H., Isshiki, I., Funamizu, N., Takakuwa, T., Nakagawa, H., Nishimura, S.-I., 2005. Organic matter released from activated sludge bacteria cells during their decay process. Environ. Technol. 26 (4), 433e440. O’Toole, J., Sinclair, M., Jeavons, T., Leder, K., 2008. Alternative water sources and endotoxin. Water Sci. Technol. 58 (3), 603e607. O’Toole, J., Sinclair, M., Jeavons, T., Leder, K., 2009. Influence of sample preservation on endotoxin measurement in water. Water Sci. Technol. 60 (6), 1615e1619. Parikh, S.J., Chorover, J., 2007. Infrared spectroscopy studies of cation effects on lipopolysaccharides in aqueous solution. Colloids Surf. B 55 (2), 241e250. Rapala, J., Lahti, K., Ra¨sa¨nen, L.A., Esala, A.-L., Niemela¨, S.L., Sivonen, K., 2002. Endotoxins associated with cyanobacteria and their removal during drinking water treatment. Water Res. 36 (10), 2627e2635. Roth, R.A., Harkema, J.R., Pestka, J.P., Ganey, P.E., 1997. Is exposure to bacterial endotoxin a determinant of susceptibility to intoxication from xenobiotic agents? Toxicol. Appl. Pharmcol. 147 (2), 300e311. Sun, Y.X., Wu, Q.Y., Hu, H.Y., Tian, J., 2009. Effects of ammonia on the formation of THMs and HAAs in secondary effluent chlorination. Chemosphere 76 (5), 631e637. Venter, P., Abraham, M., Lues, J.F.R., Ivanov, I., 2006. The influence of sanitizers on the lipopolysaccharide composition of Escherichia coli O111. Int. J. Food Microbiol. 111 (3), 221e227. Williams, K.L., 2007. Endotoxins: Pyrogens, LAL Testing and Depyrogenation, third ed. Informa Healthcare USA Inc, N.Y, pp. 27, 29, 67. Zhang, K., Farahbakhsh, K., 2007. Removal of native coliphages and coliform bacteria from municipal wastewater by various wastewater treatment processes: Implications to water reuse. Water Res. 41 (12), 2816e2824.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 5 8 e4 7 6 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Net energy production associated with pathogen inactivation during mesophilic and thermophilic anaerobic digestion of sewage sludge Christopher Ziemba, Jordan Peccia* Department of Chemical and Environmental Engineering, Yale University, New Haven, CT 06520, USA

article info

abstract

Article history:

The potential for anaerobic digester energy production must be balanced with the

Received 14 March 2011

sustainability of reusing the resultant biosolids for land application. Mesophilic, thermo-

Accepted 15 June 2011

philic, temperature-phased, and high temperature (60 or 70

Available online 24 June 2011

digester configurations have been systematically evaluated for net energy production




C) batch pre-treatment

and pathogen inactivation potential. Energy input requirements and net energy production Keywords:

were modeled for each digester scheme. First-order inactivation rate coefficients for

Biosolids

Escherichia coli, Enterococcus faecalis and bacteriophage MS-2 were measured at each digester

Pathogens

temperature and full-scale pathogen inactivation performance was estimated for each

Energy

indicator organism and each digester configuration. Inactivation rates were found to increase dramatically at temperatures above 55
C.

Biogas Methane

Modeling full-scale performance using retention times based on U.S. EPA time and

Reactivation

temperature constraints predicts a 1e2 log inactivation in mesophilic treatment, and a 2e5 log inactivation in 50e55
C thermophilic and temperature-phased treatments. Incorporating a 60 or 70
C batch pre-treatment phase resulted in dramatically higher potency, achieving MS-2 inactivation of 14 and 16 logs respectively, and complete inactivation (over 100 log reduction) of E. coli and E. faecalis. For temperatures less than 70
C, viability staining of thermally-treated E. coli showed significantly reduced inactivation relative to standard culture enumeration. Due to shorter residence times in thermophilic reactors, the net energy production for all digesters was similar (less than 20% difference) with the 60 or 70
C batch treatment configurations producing the most net energy and the mesophilic treatment producing the least. Incorporating a 60 or 70




C pre-treatment phase can

dramatically increase pathogen inactivation performance without decreasing net energy capture from anaerobic digestion. Energy consumption is not a significant barrier against improving the pathogen quality of biosolids. ª 2011 Published by Elsevier Ltd.

1.

Introduction

More than 7 million dry tons of sewage sludge are produced annually in the U.S. (Beecher et al., 2007). This figure is expected to increase as more communities move to

centralized sewage collection systems, activated sludge-based nutrient removal processes become more prevalent, populations served by sewers grow, and anaerobic digestion is developed as a renewable energy source. The U.S. EPA encourages the treatment and beneficial reuse of stabilized

* Corresponding author. Tel.: þ962 203 432 4385; fax: þ962 203 432 4387. E-mail address: [email protected] (J. Peccia). 0043-1354/$ e see front matter ª 2011 Published by Elsevier Ltd. doi:10.1016/j.watres.2011.06.014

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 5 8 e4 7 6 8

sewage sludge (biosolids) as a solution to the high costs and environmental impacts of incineration and land filling. Approximately 55% of treated sewage sludge in the U.S. is reused, primarily as a soil conditioning product and fertilizer (Beecher et al., 2007). While the agricultural benefits of land applying biosolids are well documented (Evanylo et al., 2008; Khaleel et al., 1981), there is widespread concern over pathogen exposure to residents in communities that surround land applications sites (Lewis and Gattie, 2002; NRC, 2002). Pathogen inactivation during anaerobic digestion is an integral component for ensuring the safety and sustainability of biosolids reuse. With limited information from which to estimate risk, the U.S. EPA’s part 503 biosolids regulations were based on using best available treatment technology to address pathogen concerns (USEPA, 1999). These regulations established class A and class B treatment standards and prescribed stabilization methods necessary to meet each designation. Class A biosolids must undergo an EPA-approved treatment process yielding fecal coliform concentrations less than 1000 colony forming units (CFU) per dry gram, or less than 3 most probable number (MPN) Salmonella sp. in 4 dry grams. The resulting class A biosolids can be sold and utilized without restriction. Class B pathogen reduction goals are less stringent, requiring fecal coliform concentrations less than 2  106 CFU or MPN per dry gram. Class B biosolids still contain human pathogens following treatment, therefore site restrictions and reductions in vector attraction are also required to reduce pathogen exposure to the public. The most common treatment technology for meeting class B requirements is anaerobic digestion operating in the mesophilic range of 35e40
C. Achieving class A standards through digestion usually involves increasing digester temperature to between 50 and 55
C for thermophilic anaerobic digestion (TAD). Temperature-phased anaerobic digestion (TPAD) is a hybrid process which commonly features a shorter TAD phase, for hydrolysis, pathogen inactivation, and sometimes acetogenesis, followed by a mesophilic anaerobic digestion (MAD) phase. These current anaerobic digestion configurations have known limitations for inactivating pathogens. MAD inactivation efficiency is limited and only results in one or two log removal of fecal indicators in full-scale operation over a 15e40 day residence time (Gantzer et al., 2001; Guzma´n et al., 2007; Pedersen, 1981). Although log fecal coliform reduction is significantly greater during traditional thermophilic processes, (50 or 55
C) and potentially on the order of four logs, increasing evidence suggests that TAD may not be reliable at permanently inactivating some bacterial pathogen indicators. TAD processes may induce a viable but nonculturable (VBNC) condition, from which some bacteria may later recover. Such reactivation behavior has been demonstrated during high-speed centrifugal dewatering of thermophilically digested biosolids in both Escherichia coli and fecal enterococci (Higgins et al., 2007; Qi et al., 2007; Sahlstrom et al., 2004; Viau and Peccia, 2009a). Reactivation has not been reported in pasteurized biosolids. Mechanistic evidence of ribosomal unfolding at different temperatures (Lee and Kaletunc, 2002; Mackey et al., 1991) suggest that fecal coliform bacteria can be fully inactivated at temperatures at and above approximately 60
C.

4759

Redesigning digesters to employ an effective pathogen inactivation mechanism that does not significantly increase energy consumption is essential to ensuring the safe and economical reuse of biosolids and the development of biogas as a sustainable alternative fuel. We hypothesize that incorporating a 60
C or higher initial phase into an anaerobic digestion process will significantly improve pathogen inactivation performance over mesophilic or thermophilic treatment, and that the energy efficiencies produced by a shorter residence time and more effective hydrolysis will allow for this greater inactivation without impacting net energy production value. This hypothesis has been evaluated by estimating pathogen indicator inactivation rate coefficients in sludge as a function of temperature, identifying common (MAD, TAD, TPAD) and alternative digestion schemes (60 and 70
C pre-treatment), and modeling the pathogen reduction and net energy production of each scheme. Rather than a strict focus on meeting current regulations, this research seeks to understand the energy costs associated with decreasing the pathogen load in land applied biosolids.

2.

Materials and methods

2.1.

Selection of indicator organisms

Escherichia coli was chosen as a test organism because it is a member of the fecal coliform class, upon which the U.S. EPA 503 regulations for biosolids pathogen quality are predominantly based. Enterococcus faecalis is a representative member of the fecal enterococci group, which are gram-positive and have been shown to be more resistant to temperature inactivation than fecal coliforms (Viau and Peccia, 2009a). MS-2 is a commonly studied male-specific (Fþ) coliphage. A compelling case has been presented in the literature that malespecific coliphages have value as an indicator of fecal contamination and pathogenic virus inactivation due to their similarity to human enteric viruses in terms of structure and persistence through treatment (Funderburg and Sorber, 1985; Havelaar et al., 1993; Nappier et al., 2006).

2.2. Batch temperature inactivation experimental procedure Batch testing was conducted in 125 ml crimp-top serum bottles (Wheaton, Millville, NJ, USA) containing 74.5 ml of mesophilically digested sewage sludge, adjusted to 6% solids in phosphate buffered saline solution (PBS, 0.14 M NaCl, 0.01 M phosphate, and 0.003 M KCl, American Bioanalytical, Natick, MA). The digested sludge was obtained locally from a municipal wastewater treatment plant that utilized activated sludge treatment, MAD stabilization and belt filter press dewatering. Sludge characteristics are typical and presented in Supplementary Data Table S-1. Sludges were prepared by autoclaving for 30 min, homogenizing in a blender for 20 min and adjusting pH to 7.5. Bottles were capped, purged with nitrogen and acclimated to experimental temperature in either an incubator (T ¼ 37
C) or a water bath (T ¼ 50, 55, 60 or 70
C). Each reactor bottle was anaerobically spiked with 0.5 ml of one of three pathogen indicator organisms, E. coli, E.

4760

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 5 8 e4 7 6 8

faecalis or MS-2 bacteriophage. Well mixed conditions were established initially by vigorous shaking and maintained by orbital shaking at 40 RPM throughout the experiments. Reactors were sampled periodically by 18-gauge needle-tipped syringe, and cultures immediately enumerated. Due to rapid inactivation rates and slow characteristic mixing times for the 6% solids solution, inactivation experiments for E. coli and E. faecalis at 60 and 70
C were conducted in a series of 1 ml syringes. Sludge media was processed as in the serum bottle reactors described above, with the exception that the inoculum was added and the bottles well mixed while at room temperature. The inoculated sludge media was then distributed into 1 ml syringes, capped, and submerged in a water bath at 60 or 70
C. After waiting 60 s for the 60
C reactors and 80 s for the 70
C reactors to reach temperature (heating times were independently verified), syringes were individually removed at 5e10 s intervals and bacteria were enumerated. All inactivation experimental conditions were tested in duplicate. Log-transformed data were pooled and first-order inactivation rate constants and associated standard errors of fit were estimated by least squares regression. To determine the potential for VBNC behavior, E. coli batch experiments were recreated as above, at 50, 55, 60, and 70
C using PBS in place of sludge. Each bottle was inoculated with E. coli and incubated for times corresponding to w4 logs of culture-based inactivation. Bottles were cooled in a 25
C water bath and concentrations of viable cells were determined by staining with 5-cyano-2,3-ditolyltetrazolium chloride (CTC, SigmaeAldrich, St. Louis, MO). Viable staining procedure consisted of incubating 100 ml of sample in the dark for 90 min at 37
C with 1/100 strength TSB at a final concentration of 5 mM CTC. Viable cell counts were standardized to total cell count, by independently staining with SYTO-9 (Invitrogen, Carlsbad, CA). Total cell count procedure consisted of incubating 100 ml of sample in the dark for 20 min at room temperature with a final SYTO-9 concentration of 5 mM. CTC and SYTO-9 stained cells were each enumerated using standard microscopy procedures and appropriate fluorescent filters (Hobbie et al., 1977). Each temperature was tested in triplicate at a residence time corresponding to a 4 log loss of culturability. Each residence time was estimated and confirmed by plate count, employing the methods described below.

2.2.1.

Inoculum preparation

Overnight cultures of E. coli (ATCC#15597, American Type Culture Collection, Manassas, VA) were grown at 37
C in 5 ml of tryptic soy broth (TSB, Difco Inc., Detroit, MI)). A 50 ml aliquot of this overnight culture was reinoculated in 50 ml of TSB and incubated at 37
C under well mixed conditions for 5 h to reach mid-log phase. The bacteria were washed three times by repeated centrifugation at 5000 g for 10 min and resuspension in PBS, and finally suspended in 5 ml of PBS. Bacterial concentration was determined by direct fluorescence microscopy using 40 , 6-diamidino-2-phenylindole (DAPI) staining (SigmaeAldrich, St. Louis, MO) and standard counting procedures (Hobbie et al., 1977). Preparation of E. faecalis (ATCC#19433) was identical to that of E. coli, except brain heart infusion broth (Himedia Inc., Mumbai, India) was used. The male-specific bacteriophage MS-2 (ATCC 15597-B1) was prepared using a modified double agar layer (DAL) method

(USEPA, 2001). In a hot water bath at 48
C, 1 ml of MS-2 stock and 1 ml of log phase E. coli (ATCC#15597) were added to 4 ml of molten 0.7% tryptic soy agar (TSA, Difco Inc., Detroit, MI)) in a 30 ml test tube. The test tube was immediately removed from the water bath, gently rolled and poured onto a 10 cm diameter TSA plate. The DAL plates were allowed to cool to room temperature, inverted and incubated at 37
C for 24 h. To elute phages, 2 ml of PBS was gently added to each plate and again incubated at 37
C for 1 h. The PBS wash was then collected and centrifuged at 5000 g for 10 min. The supernatant was passed through a 0.45 mm syringe filter to remove bacteria (Whatman Inc. Florham Park, NJ, USA), and MS-2 phages were concentrated using a centrifugal membrane (Millipore Inc. Billerica, MA, USA). Final MS-2 concentration was determined by repeating the DAL method on dilutions from the new stock solution and counting the number of plaques in the E. coli lawn for dilutions yielding between 30 and 300 plaques. Inoculum concentrations were on the order of 109 CFU/ml for bacteria and 1012 PFU/ml (plaque forming units) for MS-2.

2.2.2.

Enumeration by culturing

Reactor samples were serially diluted in PBS to achieve between 30 and 300 colony or plaque forming units per culture plate. E. coli was plated on mFC agar (Difco Inc., Detroit, MI) and incubated at 44.5
C for 24 h (APHA et al. 2005). E. faecalis was plated on mEI agar (Difco Inc., Detroit, MI) and incubated at 41.5
C for 48 h (USEPA, 2002). Enumeration of MS-2 was conducted by the same DAL method described above, with serial dilutions of 0.45 mm filtered reactor samples added to the E. coli and molten TSA in place of the MS-2 Stock. All plating was performed in duplicate.

2.3.

Digester configurations

Table 1 presents seven digestion configurations selected to represent commonly used mesophilic and thermophilic digester temperatures, plus 60
C and 70
C proposed pretreatment schemes. The selected residence times are based on U.S. EPA part 503 class A and class B pathogen regulations. Class B MAD treatment at 37
C is modeled at the U.S. EPA minimum of 15 days residence time (USEPA, 1994). For digesters operating at or above 50
C, with solids content less than 7%, the EPA-mandated minimum residence times to achieve class A standards are governed by a time and temperature relationship (USEPA, 1994). Application of this relationship yields a minimum residence time of 5 days in a digester at 50
C. TAD residence times are extended to 15 days to allow more complete solids conversion and to better reflect the longer residence times often employed in practice (Viau and Peccia, 2009b). The residence times in our 50 and 55
C TPAD configurations are based on the 5 day minimum requirement at 50
C, followed by the standard 15 days in MAD. At 60 and 70
C, the EPA mandated minimum residence times are approximately 5 and 0.5 h respectively for pathogen inactivation. Table 1 presents a more conservative and internationally established residence time of 1 h for 70
C pasteurization, and the specified 5 h at 60
C. Both 60 and 70
C treatments must be paired with a mesophilic phase to achieve acceptable volatile solids (VS) conversion and gas production. The second phase is modeled as a 15 day MAD reactor.

4761

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 5 8 e4 7 6 8

Table 1 e Time and temperatures of the seven anaerobic digester configurations evaluated. MAD and TAD configurations are single stage processes. (cstr) denotes completely mixed stirred tank reactor configuration. Configuration

Phase 1


Phase 2

Temperature ( C)

Residence Time (days)

Temperature ( C)

37 50 55 50 55 60 70

15 (cstr) 15 (cstr) 15 (cstr) 5 (cstr) 5 (cstr) 0.208 (5 h batch) 0.042 (1 h batch)

37 37 37 37

MAD TAD 50 TAD 55 TPAD 50 TPAD 55 60 batch þ MAD 70 batch þ MAD

Each digestion scheme has been configured to incorporate heat exchangers to capture waste heat and reduce total energy costs, a common trend in modern digester design (Greer, 2007; Zupancic and Ros, 2003). The efficiency (the difference between the final and initial temperatures in the hot stream divided by the difference in initial temperatures of the cold and hot streams) of counter-flow heat exchangers are typically 50% (Kepp et al., 2000). For single-phase MAD and TAD configurations listed in Table 1, heat from the digester effluent sludge is transferred to the digester influent sludge by a counter-flow heat exchanger. In TPAD, 60
C batch, and 70
C batch configurations, incoming sludge receives heat from two sets of counter-flow heat exchangers. Here, influent sludge is first preheated by captured heat from the effluent of the second (MAD) phase and once again heated by the waste heat from sludge leaving the first (thermophilic CSTR, 60 or 70
C batch) phase.

2.4.

Energy balances

The amount of energy produced by an anaerobic digester per metric ton of wet sludge, is defined as the per ton energy content of the biogas produced minus the per ton energy input demands to operate the digester. This input demand includes energy required to heat sludge to digester operating temperature and the energy required to compensate for heat losses during operation. Secondary costs such as stirring and secondary products such as biological heat generation during digestion are relatively insignificant (Lu¨bken et al., 2007) and are omitted from the net energy calculation. All analysis assumes a 6% solids content.

2.4.1.

Heat-up energy demand

The amount of heat (kWh) required per wet metric ton (1000 kg) to heat-up sludge is the difference between the initial and desired temperatures multiplied by the specific heat capacity of 6% solids sludge, 1.117  103 kWh/kg
C (Metcalf and Eddy, 2003). The initial temperature of sludge flowing into a digester is the temperature at the previous source plus temperature gained through heat exchanger recovery. Incoming sludge to each digester configuration is initially assumed to be 15.6
C (Metcalf and Eddy, 2003).

2.4.2.




Heat loss from reactors

The rate of heat loss from a reactor, (q_ reactor, W) is described in equation (1) as the sum of heat loss rates through the floor, walls and roof.

q_ reacter ¼

X

Residence Time (days)

15 15 15 15

 Usurface Asurface ðTreacter  Tout Þ



(cstr) (cstr) (cstr) (cstr)

(1)

all surfaces

The heat loss through each surface is the product of the overall heat transfer coefficient through the surface (Usurface, W/m2/
C), the surface area (Asurface, m2) and the difference in inside (Treacter,
C) and outside (Tout,
C) surface temperatures. For the purpose of defining surface areas, plausible reactor geometries have been selected to accommodate a theoretical flow of 1.2  102 metric tons (0.45 million gallons) of wet sludge per day (Metcalf and Eddy, 2003), which corresponds to a wastewater flow of approximately 3  107 L (9 million gallons) per day. Digesters are conventionally cylindrical in shape, drawing to a point at the bottom. For 15 day residence time reactors, dimensions are set at 18 m in diameter, 6 m deep at the sides and 9 m deep in the middle. The dimensions of 5 day residence time reactors preserve the same ratio between the diameter, side depth and mid depth, scaled down to 1/3 volume. Wall construction for 15 and 5 day residence time reactors is set at 0.3 m thick concrete in contact with air,
resulting in an overall heat transfer coefficient of 4.9 W/m2 C (USEPA, 1979). The floor is also 0.3 m thick concrete, in contact with dry earth, resulting in an overall heat transfer coefficient
of 0.34 W/m2 C, (USEPA, 1979). Selecting a 0.225 m thick fixed concrete cover achieves an overall heat transfer coefficient of
3.3 W/m2 C (USEPA, 1979). Representative external temperature values are 11.5
C for soil and 8.6
C for air. In batch reactor configurations at 60 or 70
C, the sludge is treated using three batch reactors. The dimensions are 2.9 m diameter, 1.9 m side depth and 2.6 m mid depth for each 60
C reactor and 1.5 m diameter, 1.6 m side depth and 2.1 m mid depth for each 70
C reactor. Both sets of batch reactors are constructed using 10 mm thick steel walls, floor and roof (Le et al., 2002), each in contact with air, resulting in an overall
heat transfer coefficient, Usurface, of 5.167 W/m2
C (USEPA, 1979). It is assumed that batch reactors operate constantly and operate at capacity. We assume 15 min fill and empty times for 1 h residence time reactors and 30 min fill and empty times for 5 h residence time reactors. Heat losses from batch reactors during filling and emptying phases are conservatively modeled to be identical to heat losses during full operation. The overall amount of heat loss per wet ton of sludge for a specific reactor configuration (qreacter, Wh/ton treated sludge) is depicted in Eq. (2), and determined by multiplying the heat loss rate (q_ reacter , W, Eq. (1)) by the residence time of interest (qreactor ) and then dividing by the reactor volume (Vreactor, L) and the density of sludge (rsludge, kg/L). The density of 6% sludge is 1.01 kg/L (Metcalf and Eddy, 2003).

4762

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 5 8 e4 7 6 8

( qreacter ¼

2.4.3.

P

)   Usurface Asurface ðTreactor  Tout Þ qreactor

all surfaces

Vreactor rsludge

(2)

Biogas production

Biogas production was estimated using a generalized organic waste fermentation equation that models digested sludge as C10H19O3N and utilizes CO2 as the terminal electron acceptor, Eq. (3) (Rittmann and McCarty, 2001).     C10 H19 O3 N þ 18  22:5fs  12:5fe H2 O/ 6:25fe CH4       þ 9  10fs  6:25fe CO2 þ 2:5fs C5 H7 O2 N þ 1  2:5fs NHþ 4   þ 1 þ 2:5fs HCO (3) 3 The net fraction of sludge organic matter converted to cell mass is represented as fs and the net fraction of this organic matter that is utilized for cellular energy is represented by fe Values of f s ¼ 0.07 (and f e ¼ 0.93) were estimated by previously published relationships based on a mass balance for volatile solids in a CSTR (Rittmann and McCarty, 2001) and using the initial fraction converted into cells fso ¼ 0:11, the biodegradable fraction of cellular biomass f d ¼ 0.8, an endogenous decay rate b ¼ 0.05 day1, and the steady state CSTR residence time Ө of 15 days. A common VS conversion value of 56% was assumed for mesophilic digestion at 15 days (Metcalf and Eddy, 2003). Operating single phase digesters at thermophilic temperatures or the presence of an initial thermophilic phase typically improves hydrolysis and increases the bioavailability of digestible material. Increases over MAD in VS destruction due to either thermophilic digestion or the presence of a thermophilic phase have been reported to range from 7% to 11% (Ge et al., 2010; Salsabil et al., 2010; Shimp et al., 2003). The TAD, TPAD, and 60 and 70
C pretreatment cases presented here are modeled conservatively as achieving a 7% increase in VS conversion (63% overall) over MAD treatment alone. The raw energy value of methane (kWh) produced per ton of digested sludge under each digester configuration can be calculated by multiplying together the solids content (kg/kg), the fraction VS of TS (kg/kg), the VS conversion efficiency of the reactor (kg/kg), the number of moles of methane produced per mole of converted sludge (mole/mole), the lower heating value of methane (LHVCH4 ; kWh=mole), and divided by the molecular weight of C10H19O3N (Msludge, g/mol), Eq. (4). Incoming sewage sludge is modeled as 6% solids (Shimp et al., 2003), with a 70% VS content (Metcalf and Eddy, 2003). The lower heating value of methane is 0.223 kWh/mole (Perry and Green, 1984).

 Raw Energy Produced ¼ Metric Ton Digested Sludge

% Solids Content

efficiencies associated with electricity generation, thus the final value of the biogas includes only a 2% loss in total biogas energy value for removal of water content by refrigeration (Krich et al., 2005) and a 12% reduction from the raw energy value due to combustion efficiency (Bekkering et al., 2010).

3.

Results

3.1. Pathogen indicator inactivation kinetics determined in anaerobic batch reactors Inactivation rate coefficients for E. coli, E. faecalis and MS-2, were measured in anaerobic batch reactors at 37, 50, 55, 60, and 70
C (Fig. 1a,b,c). Inactivation profiles for each organism and temperature are presented in Supplementary Data figures S-1, S-2 and S-3. Increasing reactor temperature was found to increase inactivation rate coefficients in each test organism and the magnitude of this increase was much greater at temperatures above 50
C. Inactivation rate coefficients for E. coli and E. faecalis in the 50e55
C range are not statistically different ( p < 0.05) within each temperature, achieving 1.4 h1 for E. coli vs. 1.0 h1 for E. faecalis at 50
C and 6.8 h1 for E. coli vs. 6.7 h1 for E. faecalis at 55
C. These rate coefficients were 1e3 times greater and statistically different ( p < 0.05) than that of MS-2 at 50 and 55
C. Increasing temperature from 55 to 60
C yields a dramatic increase in bacterial inactivation rate coefficient (significant at p < 0.0001), and a smaller increase for MS-2. The inactivation rate coefficient for E. coli is 94 times greater at 60
C than at 55
C (650 h1) and 25 times greater in E. faecalis (177 h1). The rate coefficient for MS-2, in contrast, only doubled to 6 h1. For bacterial inactivation at 70
C, it was determined that the 80 s required to heat up the sludge to 70
C was sufficiently lethal to achieve complete inactivation at our limit of detection. Therefore, no precise data are reported for E. coli or E. faecalis inactivation at 70
C. We can conservatively estimate inactivation rate coefficients based on the log reduction from the seeded concentration (typically 109 CFU/ml) to the method limits of detection (2  102 CFU/ml) over the 80 s interval to be greater than 106 h1 for both bacteria. MS-2 inactivation at 70
C increased more modestly to 36 h1. At each digestion temperature MS-2 inactivation rate constants are significantly lower ( p < 0.05) than the corresponding values for E. coli and E. faecalis. The inactivation kinetics obtained from batch reactor testing have been inserted into first-order mass balance models for CSTR or batch reactors to predict inactivation performance for the seven residence time and reactor

    VS mole CH4 VS Conversion ðLHVCH4 Þ Efficiency % mole VS TS   6   10 g Msludge metric ton

Finally, extracting energy value from biogas requires usespecified purification of the gas. This model compares energy values in terms of heat and therefore does not include

(4)

temperature configurations listed in Table 1. MAD (37
C) inactivation performance is limited to 1.1 log in E. coli, 1.6 log in E. faecalis and 0.7 log in MS-2 (Fig. 2). Reductions in

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 5 8 e4 7 6 8

4763

Fig. 2 e Estimated digester performance for E. coli, E. faecalis and MS-2 as log reductions in culturable concentrations under various treatment schemes. Arrows on E. coli and E. faecalis bars for 60 and 70
C batch plus MAD indicate dramatically higher log reduction values, in excess of 100 log.

Fig. 1 e First-order inactivation rate coefficients for E. coli, E. faecalis and MS-2 as a function of temperature. Experiments were carried out in batch reactors under anaerobic conditions in autoclaved MAD biosolids at 6% solids content and pH 7.5. Rate coefficients are extrapolated by regression of the culturable concentration measurements at each experimental temperature as a function of time. Error bars represent standard error. Insets in Figs. 1a, b and c display subsets of the full plot in order to observe the impact of temperature on inactivation rate coefficients below 60 or 70
C.

thermophilic single stage and phased digesters ranged from 2 to 4 logs in TAD at 50 or 55
C and TPAD at 50
C configurations. TPAD treatment at 55
C displayed slightly greater inactivation with 4.0 log in E. coli, 4.5 log in E. faecalis and 3.2 log in MS2. Projected inactivation at 60
C is significantly greater than in MAD, TAD or TPAD configurations, resulting in inactivation

predictions of 13.7 log for MS-2 and greater than 100 log for E. coli and E. faecalis. At 70
C, E. coli and E. faecalis are again projected to have greater than 100 log reduction and MS-2 is predicted to be reduced by 16.4 log. Inactivation projections for batch reactors do not include additional inactivation occurring during filling and emptying phases. Inactivation of E. coli at temperatures of 50, 55, 60 and 70
C was also accessed by CTC viability staining and is presented in Fig. 3. The initial viability in the E. coli inoculum was 68% of total cells. After incubation at each temperature at times standardized to a 4 log loss of culturability, losses in viability were observed to be less than the losses in culturability in E. coli. A reduction of 0.76 log viability was achieved at 50
C, 1.31 log at 55
C and 2.14 log at 60
C. At 70
C the average loss of viability was 2.93 log. The losses in viability at 50 and 55
C are significantly reduced from losses at 60
C ( p < 0.001) and 70
C ( p < 0.01). The 70
C loss of viability value of 2.93 log is not significantly less ( p ¼ 0.063) than the 3.7 log loss exhibited in autoclaved E. coli. The 70
C and autoclaved E. coli reductions are at the upper value of log reductions that can be observed by microscopy.

3.2.

Energy associated with inactivation

The total heating demand of each reactor has been divided by the expected pathogen inactivation performance to show the relative heating demands per log inactivation, Table 2. Total heat demand is the sum of energy to heat the sludge up to the digester temperature and the energy required to overcome heat losses from the digester to maintain the desired temperature. Due to the high inactivation rates at thermophilic temperatures, and the shorter residence times, increasing the temperature decreases the amount of heating energy required per log removal. Rankings of heat demand per log removal are the following: MAD > TAD > TPAD>>60 and 70
C pretreatment plus MAD. Net energy output (kWhr/metric ton) is defined as the amount of usable energy produced in biogas minus the total heat demand. Each digestion configuration produces more

Log Reduction in Viable Concentration

4764

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 5 8 e4 7 6 8

3.5

1.5

information presented here demonstrates the feasibility and the sustainability advantages of operating anaerobic digesters at increased temperatures. This work is only an initial step. Implementation at the full-scale must involve comprehensive pilot testing and operational experience to ensure all challenges of digester performance (e.g. VS destruction, limited foaming, inactivation of multiple types of human pathogens, etc.) are met.

1.0

4.1.

3.0 2.5 2.0

0.5 0.0

50

55

60

70

Temperature (°C)

Fig. 3 e Log reduction of viable E. coli as determined by CTC staining after batch temperature treatment at residence times corresponding to w4 log loss of cultivability. Error bars represent standard error of three independent experiments.

energy than it consumes, and generally the amounts of energy produced are similar across all configurations, with energy produced in the most productive configuration only 19% greater than the least (Fig. 4). Heating up sludge and heat losses account for approximately 25% (range 19%e32%) of the total energy produced in these digesters. The batch pretreatment configurations produce the most net energy with both 60 and 70
C pretreatment variations producing about 120 kWh per wet metric ton of treated sludge.

4.

Discussion

The results of this study reveal two important concepts for operating anaerobic digesters to produce a pathogen-free, sustainable biosolids product. First, batch temperature inactivation studies revealed the dramatic increase in inactivation potential of operating a digester at or above 60
C. Second, this significantly greater pathogen reduction can be achieved without a decrease in digester net energy production. The

Inactivation kinetics

The major concern surrounding land applying class B biosolids is exposure of workers and nearby residents to infectious pathogens (NRC, 2002). Currently 39 of 50 U.S. states have local or statewide restrictions on land application (Beecher et al., 2007). Restrictions adopted typically include the use of buffer zones between biosolids-applied land and residential areas. The value of using set-back distances for risk reduction however, is limited as aerosol transport studies suggest that for most set-back distances (usually less than 50 m), there is less than 1 log reduction in infectious pathogen exposure (Low et al., 2007). Multiple log reductions in pathogen content and potential risk, therefore, must be accomplished during sludge stabilization. Temperatures above 55
C have traditionally not been considered in anaerobic digestion. Exceeding 55
C in single stage digesters may cause an imbalance between acetogenesis and methanogenesis and lead to instability or digester failure. Incorporating 60 or 70
C temperatures in the first phase of multi-phase treatment, or as a batchpretreatment step, expands the range of operating temperatures without negatively impacting process stability or solids conversion. This study demonstrates a sharp increase in bacterial first-order inactivation rate coefficients in pure culture batch reactors at temperatures of 60
C and above. Such behavior is consistent with the onset of permanent ribosome damage in E. coli occurring at temperatures in the vicinity of 60
C as determined by differential scanning calorimetry (DSC) (Lee and Kaletunc, 2002; Mackey et al., 1993). Unlike temperatures above 60
C, DSC plots at 37, 50 and 55
C do not reveal permanent conformational changes in the structure of the cell, which would indicate effective and permanent inactivation (Lee and Kaletunc, 2002; Mackey

Table 2 e Inactivation performance efficiency expressed as the demand for heating energy (kWh) per wet ton of treated sludge for a given reactor configuration divided by the log removal calculated for each pathogen indicator in that reactor configuration. Values for E. coli and E. faecalis batch plus MAD configurations at 60 and 70
C are based on conservative estimates for inactivation rate coefficients and therefore presented here as upper bounds. Configuration

MAD TAD 50 TAD 55 TPAD 50 TPAD 55 60 batch þ MAD 70 batch þ MAD

Heating Energy Demand (kWh/wet metric ton treated sludge)

27 42 47 37 39 31 29

Heating Energy Demand to Achieve 1 log Expected Inactivation (kWh/wet metric ton treated) E. coli

E. faecalis

24 15 14 11 10 < 0.1 < 0.0001

17 16 14 10. 8.7 < 0.1 < 0.0001

MS-2 39 19 16 16 11 2.2 1.7

4765

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 5 8 e4 7 6 8

Energy Costs and Benefits (kWh per metric ton treated sludge)

300 250 200

Heat-up Cost Heat Loss Cost Biogas Value Net Energy

150 100 50 0 -50

MAD

TAD 50

TAD 55

TPAD 50

TPAD 55

60+MAD

70+MAD

Treatment Scheme

Fig. 4 e Energy required and production per wet metric ton of treated sludge for various treatment schemes. Heat-up requirement refers to the initial heating to bring the sludge up to temperature in the first or only phase of the digester. Heat loss refers to heat energy lost from the digester or digester phases which must be replaced. Biogas production is expressed as heat energy produced and reflects losses for combustion efficiency and moisture removal.

et al., 1991, 1993). Lower inactivation rates and documented full-scale complications with VBNC behavior at 50 or 55
C are likely caused by the absence of a true bacterial inactivation mechanism. The differences between the culturability loss and viability loss (Fig. 3) at temperatures below 60
C suggests that E. coli treated in this temperature range may be more prone to VBNC and reactivation behavior previously observed in full-scale, thermophilic digesters. Loss of viability was clearly below the maximum observable loss (3.7 log) for 50 (0.76 log) and 55
C (1.31 log) temperatures. At 60
C, the exposure time which causes a 4 log reduction in E. coli culturability corresponds to a 2.14 log loss in viability, which is also significantly less than the 3.7 log reduction in the autoclaved control. The threshold for permanent ribosomal damage in E. coli is approximately 60
C. It is possible that the variation in this value, or variation in experiment placed the bacteria just below this lethal threshold. Experiments at 70
C are more certainly lethal in terms of permanent ribosomal damage, but due to the very rapid inactivation (4 log loss of culturability in 16 s) the higher variation between replicates at this temperature can likely be attributed to experimental imprecision. The useful range of CTC staining is limited by minimum detection limits of microscopy and non-specific interactions between the stain and high concentrations of inactivated cells. The maximum CTC-derived log reduction in viability observed was 3.7 log in an autoclaved negative control. While further investigation using different types of organisms is necessary to confirm these effects, this work does suggest that digestion or treatment above 60
C provides some advantage for reducing potential coliform reactivation. Particular attention should be paid to exactly where the threshold of permanent ribosomal damage falls for bacteria of concern. Finally, the sharp inactivation rate increase in bacteria at 60
C was not observed in MS-2 phage, which does not contain ribosomes. The greater resistance to thermal inactivation also supports the use of MS-2 or coliphages as a more conservative pathogen inactivation indicator than fecal coliforms. Additionally, the inactivation rate coefficients observed for MS-2 are either similar or more resistant

to heat inactivation than the rate coefficients reported for Ascaris suum (Aitken et al., 2005; Popat et al., 2010). The relationship between inactivation rate constant and digester temperature has also been investigated with Arrhenius plots for each indicator organism, and is presented in the Supplementary Data Figure S-4. The Arrhenius plot for MS-2 displays linearity through the entire temperature range (25e70
C). This indicates that the activation energy barrier for MS-2 inactivation is constant through this temperature range, and dominated by a single mechanism, likely protein denaturation. Arrhenius plots for E. coli and E. faecalis however, show a linear region at lower temperatures and then a marked increase at temperatures beyond 55
C. This behavior supports the conclusion that a threshold temperature exists in bacteria whereby vital proteins or ribosomes become completely and irreversibly denatured. For the important indicators considered here, this temperature is above 55
C. The shape of these Arrhenius plots is also consistent with the markedly increased loss of bacteria viability observed above 60
C (Fig. 3). The log inactivation projections presented here for common thermophilic and mesophilic temperatures are consistent with previous bench and full-scale inactivation observations. First-order inactivation rate coefficients for E. coli, E. faecalis and MS-2 are similar to previously reported values in several studies, conducted in water or manure, at temperatures ranging from 35 to 60
C (Aitken et al., 2007; Nappier et al., 2006; Olsen and Larsen, 1987; Spinks et al., 2006). Additionally, observations in full-scale sludge digester systems for fecal coliforms have similarly demonstrated 1 to 2 log reductions in MAD, 3-4 log in TAD or TPAD and compete inactivation by pasteurization at 70
C (Bagge et al., 2005; Sidhu and Toze, 2009; Viau and Peccia, 2009b).

4.2. Comparing energy efficiencies between digester configurations An energy balance for each digester configuration is dominated by the value of biogas produced. This assessment

4766

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 5 8 e4 7 6 8

agrees with previous modeling studies that have suggested net energy gains can be achieved by adding 70
C pretreatment steps at 1e5 day residence times (Bolzonella et al., 2007; Lu et al., 2008). Energy requirements for heating consume between 19 and 32% of this biogas value depending on the reactor configuration. While no optimized energy balance comparing thermophilic and mesophilic digestion conditions has been published, our estimate of MAD heating requirements consuming 21% of biogas energy production (30% if losses associated with heat recovery are neglected) compares well with the previously published 35% for a MAD system which neglects the potential for process heat recovery (Lu¨bken et al., 2007). These similarities and the fact that the heating requirement and biogas models are well accepted for estimating digester performance suggest that our model provided heating and biogas production in the ranges that are observed in full-scale practice and is useful for making comparisons between digester configurations. All 7 digestion configurations analyzed produce comparable amounts of net energy per unit of treated sludge, which indicates that 60 or 70
C pretreatment can be implemented without forfeiting energy production. While this result is not intuitive, there are two reasons for why high temperature digestion can be achieved without sacrificing energy output. First, the heating costs in a reactor are a combination of the heat input to bring the sludge up to the correct temperature and the additional heat needed to compensate for heat losses to the environment. Increasing the reactor temperature increases the required initial heat input as well as the rate of heat loss. However this effect is mitigated by the dramatic reduction in residence times and reactors sizes used at higher temperatures due to more rapid inactivation. Secondly, our model incorporates increases in biogas energy yield (based on 7% increase in total VS destruction over MAD alone) for configurations incorporating 50, 55, 60 or 70
C treatment. Operating single phase digesters at thermophilic temperatures or the presence of an initial thermophilic phase typically improves hydrolysis and increases the bioavailability of digestible material (Bolzonella et al., 2007; Ferrer et al., 2008; Lu et al., 2008). A representative increase of 7% in total VS conversion for thermophilic TPAD treatment has been noted in fullscale digesters relative to MAD treatment alone (Shimp et al., 2003). Slightly higher 9 and 10% increases have been observed in bench-scale testing for 90 min of 60
C pretreatment (Salsabil et al., 2010) and short residence time (2 day) TPAD systems (Ge et al., 2010) respectively, each relative to MAD treatment alone. Though not knowing the specific increase in biogas production due to each pretreatment regime introduces some model uncertainty, the impact on this uncertainty is minimal. If our model did not include an increase in biogas production for higher temperature treatments, the decrease in net energy associated with adding pretreatment to MAD would only be a loss of 1% at 70
C or 3% at 60
C, which does not impact our broader conclusions about relative net energy production between thermophilic and mesophilic processes. Net energy calculations for all reactor configurations are also affected by environmental assumptions, such as air, soil, and initial sludge temperatures. However as the net energy calculations

shift across different environmental temperature ranges, the relative difference in expected net energy between our reactor configurations of interest only differs by 1 or 2 percent from the relationship expressed in Fig. 4.

5.

Conclusion

Traditionally, anaerobic digestion has been designed and optimized with the goals of increased gas production and solids destruction. Optimizing for the inactivation of pathogens has been given little attention. The results presented here describe the potential for significantly greater inactivation of bacterial and viral pathogens in biosolids and demonstrate that it is possible to do so while working within currently available configurations and without increases in energy costs.  The temperature-based inactivation rates observed in batch reactors demonstrate a dramatic increase in pathogen destruction potential when a 60 or 70
C phase is included. At 60
C, inactivation rates doubled in MS-2 phage and increased 2 orders of magnitude for enteric bacteria surrogates over rates for the more common 50 to 55
C thermophilic conditions. More dramatic increases in inactivation rates were achieved at 70
C. Thermal treatment of E. coli to produce w4 logs of culturable inactivation demonstrated that at temperatures less than 70
C, the loss of cell viability was not commensurate to the loss of culturability.  Although significantly larger reductions in pathogen indicators are observed in the high temperature regimes, the net energy production in the 60 or 70
C pretreatment systems was not significantly different than the net energy produced in more conventional systems that operate in mesophilic (w37
C) and thermophilic (50e55
C) ranges.

Appendix. Supplementary material Supplementary data related to this article can be found online at doi:10.1016/j.jorganchem.2011.03.010.

references

Aitken, M.D., Sobsey, M.D., Blauth, K.E., Shehee, M., Crunk, P.L., Walters, G.W., 2005. Inactivation of Ascaris suum and poliovirus in biosolids under thermophilic anaerobic digestion conditions. Environmental Science & Technology 39 (15), 5804e5809. Aitken, M.D., Sobsey, M.D., Van Abel, N.A., Blauth, K.E., Singleton, D.R., Crunk, P.L., Nichols, C., Walters, G.W., Schneider, M., 2007. Inactivation of Escherichia coli O157:H7 during thermophilic anaerobic digestion of manure from dairy cattle. Water Research 41 (8), 1659e1666. APHA, AWWA, WEF, 2005. Standard Methods for the Examination of Water and Wastewater. D.C, Washington. Bagge, E., Sahlstro¨m, L., Albihn, A., 2005. The effect of hygienic treatment on the microbial flora of biowaste at biogas plants. Water Research 39 (20), 4879e4886. Beecher, N., Crawford, K., Goldstein, N., Kester, G., LonoBatura, M., Dziezyk, E., 2007. A National Biosolids Regulation,

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 5 8 e4 7 6 8

Quality, End Use & Disposal Survey. North East Biosolids and Residuals Association. Bekkering, J., Broekhuis, A.A., van Gemert, W.J.T., 2010. Optimisation of a green gas supply chain - a review. Bioresource Technology 101 (2), 450e456. Bolzonella, D., Pavan, P., Zanette, M., Cecchi, F., 2007. Two-phase anaerobic digestion of waste activated sludge: effect of an extreme thermophilic prefermentation. Industrial and Engineering Chemistry Research 46 (21), 6650e6655. Evanylo, G., Sherony, C., Spargo, J., Starner, D., Brosius, M., Haering, K., 2008. Soil and water environmental effects of fertilizer-, manure-, and compost-based fertility practices in an organic vegetable cropping system. Agriculture, Ecosystems & Environment 127 (1e2), 50e58. Ferrer, I., Ponsa´, S., Va´zquez, F., Font, X., 2008. Increasing biogas production by thermal (70
C) sludge pre-treatment prior to thermophilic anaerobic digestion. Biochemical Engineering Journal 42 (2), 186e192. Funderburg, S.W., Sorber, C.A., 1985. Coliphages as indicators of enteric viruses in activated sludge. Water Research 19 (5), 547e555. Gantzer, C., Gaspard, P., Galvez, L., Huyard, A., Dumouthier, N., Schwartzbrod, J., 2001. Monitoring of bacterial and parasitological contamination during various treatment of sludge. Water Research 35 (16), 3763e3770. Ge, H., Jensen, P.D., Batstone, D.J., 2010. Pre-treatment mechanisms during thermophilic-mesophilic temperature phased anaerobic digestion of primary sludge. Water Research 44 (1), 123e130. Greer, D., 2007. Financing an anaerobic digester. BioCycle 48 (12), 44e48. Guzma´n, C., Jofre, J., Montemayor, M., Lucena, F., 2007. Occurrence and levels of indicators and selected pathogens in different sludges and biosolids. Journal of Applied Microbiology 103 (6), 2420e2429. Havelaar, A.H., Van Olphen, M., Drost, Y.C., 1993. F-specific RNA bacteriophages are adequate model organisms for enteric viruses in fresh water. Applied and Environmental Microbiology 59 (9), 2956e2962. Higgins, M.J., Chen, Y.C., Murthy, S.N., Hendrickson, D., Farrel, J., Schafer, P., 2007. Reactivation and growth of non-culturable indicator bacteria in anaerobically digested biosolids after centrifuge dewatering. Water Research 41 (3), 665e673. Hobbie, J.E., Daley, R.J., Jasper, S., 1977. Use of nuclepore filters for counting bacteria by fluorescence microscopy. Applied and Environmental Microbiology 33 (5), 1225e1228. Kepp, U., Machenbach, I., Weisz, N., Solheim, O.E., 2000. Enhanced stabilisation of sewage sludge through thermal hydrolysis three years of experience with full scale plant. Water Science & Technology, 89e96. Khaleel, R., Reddy, K.R., Overcash, M.R., 1981. Changes in soil physical properties due to organic waste applications: a review. Journal of Environmental Quality 10 (2), 133e141. Krich, K., Augenstein, D., Batmale, J., Benemann, J., Rutledge, B., Salour, D., 2005. Biomethane from Dairy Waste: a Sourcebook for the Production and Use of Renewable Natural Gas in California (Prepared for Western United Dairymen). Le, M.S., Mayhew, M.E., Back, P.A., 2002. Effectiveness of secondary digesters as a pathogen controller in winter. Water and Environment Journal 16 (4), 292e295. Lee, J., Kaletunc, G., 2002. Evaluation of the heat inactivation of Escherichia coli and Lactobacillus plantarum by differential scanning calorimetry. Applied and Environmental Microbiology 68 (11), 5379e5386. Lewis, D., Gattie, D.K., 2002. Pathogen risks from applying sewage sludge to land. Environmental Science and Technology 36 (13), 286Ae293A. Low, S.-Y., Baertsch, C., Paez-Rubio, T., Kucharski, M., Peccia, J., 2007. Off-site exposure to respirable aerosols produced during

4767

the disk incorporation of class B biosolids. Journal of Environmental Engineering 133, 897e994. Lu, J., Gavala, H.N., Skiadas, I.V., Mladenovska, Z., Ahring, B.K., 2008. Improving anaerobic sewage sludge digestion by implementation of a hyper-thermophilic prehydrolysis step. Journal of Environmental Management 88 (4), 881e889. Lu¨bken, M., Wichern, M., Schlattmann, M., Gronauer, A., Horn, H., 2007. Modelling the energy balance of an anaerobic digester fed with cattle manure and renewable energy crops. Water Research 41 (18), 4085e4096. Mackey, B.M., Miles, C.A., Parsons, S.E., Seymour, D.A., 1991. Thermal-denaturation of whole cells and cell components of Escherichia coli examined by differential scanning calorimetry. Journal of General Microbiology 137, 2361e2374. Mackey, B.M., Miles, C.A., Seymour, D.A., Parsons, S.E., 1993. Thermal denaturation and loss of viability in Escherichia coli and Bacillus stearothermophilus. Letters in Applied Microbiology 16 (2), 56e58. Metcalf and Eddy, 2003. Wastewater Engineering: Treatment and Reuse. Mc Graw-Hill, New York. Nappier, S.P., Aitken, M.D., Sobsey, M.D., 2006. Male-specific coliphages as indicators of thermal inactivation of pathogens in biosolids. Applied and Environmental Microbiology 72 (4), 2471e2475. NRC, 2002. Biosolids Applied to Land: Advancing Standards and Practices. National Research Council of the National Academies, Washington D.C. Olsen, J.E., Larsen, H.E., 1987. Bacterial decimation times in anaerobic digestions of animal slurries. Biological Wastes 21 (3), 153e168. Pedersen, D.C., 1981. Density Levels of Pathogenic Organisms in Municipal Wastewater Sludge: a Literature Review. U.S. EPA Municipal Environmental Research Laboratory, Cincinnati, OH. EPA 600-2-81-170. Perry, R.H., Green, D.W., 1984. Perry’s Chemical Engineers’ Handbook. McGraw-Hill Book Co., New York. Popat, S.C., Yates, M.V., Deshusses, M.A., 2010. Kinetics of inactivation of indicator pathogens during thermophilic anaerobic digestion. Water Research 44 (20), 5965e5972. Qi, Y., Dentel, S.K., Herson, D.S., 2007. Increases in fecal coliform bacteria resulting from centrifugal dewatering of digested biosolids. Water Research 41 (3), 571e580. Rittmann, B.E., McCarty, P.L., 2001. Environmental Biotechnology: Principles and Applications. McGraw-Hill, New York. Sahlstrom, L., Aspan, A., Bagge, E., Danielsson-Tham, M.-L., Albihn, A., 2004. Bacterial pathogen incidences in sludge from Swedish sewage treatment plants. Water Research 38, 1989e1994. Salsabil, M.R., Laurent, J., Casellas, M., Dagot, C., 2010. Technoeconomic evaluation of thermal treatment, ozonation and sonication for the reduction of wastewater biomass volume before aerobic or anaerobic digestion. Journal of Hazardous Materials 174 (1e3), 323e333. Shimp, G.F., Rowan, J.M., Long, D.W., Santha, H., 2003. Anaerobic digestion. retooling an old process to meet a ‘Class A’ objective. Water Environment and Technology 15 (5), 45e49. Sidhu, J.P.S., Toze, S.G., 2009. Human pathogens and their indicators in biosolids: a literature review. Environment International 35 (1), 187e201. Spinks, A.T., Dunstan, R.H., Harrison, T., Coombes, P., Kuczera, G., 2006. Thermal inactivation of water-borne pathogenic and indicator bacteria at sub-boiling temperatures. Water Research 40 (6), 1326e1332. USEPA, 1979. Process Design Manual Sludge Treatment and Disposal. U.S. EPA Office of Research and Development, Washington, D.C. EPA 625-1-79-011. USEPA, 1994. Land Aplication of Sewage Sludge. A Guide for Land Appliers on the Requirements of the Federal Standard for the

4768

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 5 8 e4 7 6 8

Use or Disposal of Sweage Sludge 40 CFR Part 503, U.S. EPA Office of Enforcement and Compliance Assurance, Washington, D.C EPA 831-B-93002b. USEPA, 1999. Environmental Regulations and Technology: Control of Pathogens and Vector Attraction in Sewage Sludge. U.S. EPA Office of Research and Development, Washington, D.C. EPA 625-R-92013. USEPA, 2001. Method 1601: Male-specific (Fþ) and Somatic Coliphage in Water by Two-Step Enrichment Proceedure. U.S. EPA Office of Water, Washington, D.C. EPA 821-R-01e030. USEPA, 2002. Method 1600: Enterococci in Water by Membrane Filtration Using Membrane-Enterococcus Indoxyl-b-D-

Glucoside Agar (MEI). U.S. EPA Office of Water, Washington, D.C. EPA 821-R-02e022. Viau, E., Peccia, J., 2009a. Evaluation of the enterococci indicator in biosolids using culture-based and quantitative PCR assays. Water Research 43 (19), 4878. Viau, E., Peccia, J., 2009b. A survey of wastewater indicators and human pathogen genomes in biosolids produced by class A and class B stabilization treatments. Applied and Enviornmental Microbiology 75, 164e174. Zupancic, G.D., Ros, M., 2003. Heat and energy requirements in thermophilic anaerobic sludge digestion. Renewable Energy 28 (14), 2255e2267.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 6 9 e4 7 8 1

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Magnetic binary oxide particles (MBOP): A promising adsorbent for removal of As (III) in water Rajesh M. Dhoble b, Sneha Lunge a, A.G. Bhole c, Sadhana Rayalu a,* a

Environmental Materials Division, National Environmental Engineering Research Institute, Nagpur, M.S., India Civil Engineering Department, Priyadarshini Indira Gandhi College of Engineering, Nagpur, M.S., India c Civil Engineering Department, Visvesvaraya National Institute of Technology, Nagpur, M.S., India b

article info

abstract

Article history:

Magnetic binary oxide particles (MBOP) synthesized using chitosan template has been

Received 7 March 2011

investigated for uptake capacity of arsenic (III). Batch experiments were performed to

Received in revised form

determine the rate of adsorption and equilibrium isotherm and also effect of various rate

10 June 2011

limiting factors including adsorbent dose, pH, optimum contact time, initial adsorbate

Accepted 16 June 2011

concentration and influence of presence cations and anions. It was observed that uptake of

Available online 28 June 2011

arsenic (III) was independent of pH of the solution. Maximum adsorption of arsenic (III) was w99% at pH 7.0 with dose of adsorbent 1 g/L and initial As (III) concentration of 1.0 mg/L at

Keywords:

optimal contact time of 14 h. The adsorption equilibrium data fitted well to Langmuir and

Arsenic (III)

Freundlich isotherm. The maximum adsorption capacity of adsorbent was 16.94 mg/g.

Batch adsorption

With increase in concentration of Ca2þ, Mg2þ from 50 mg/L to 600 mg/L, adsorption of As

Isotherm

(III) was significantly reduced while for Fe3þ the adsorption of arsenic (III) was increased

Magnetic adsorbent

with increase in concentration. Temperature study was carried out at 293 K, 303 K and 313 K reveals that the adsorption process is exothermic nature. A distinct advantage of this adsorbent is that adsorbent can readily be isolated from sample solutions by application of an external magnetic field. Saturation magnetization is a key factor for successful magnetic separation was observed to be 18.78 emu/g which is sufficient for separation by conventional magnate. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Arsenic is one of the several common occurring toxic metals found in environment and designated by United State Environmental Protection Agency (USEPA) under clean water acts (Martinson and Reddy, 2009). The sources of arsenic and many other trace elements in ground water are from geochemical reactions, weathering of rocks, industrial wastewater discharge, agriculture uses of arsenical pesticides, discharges from coal fired thermal power plants, herbicides and fertilizers etc. Arsenic concentrations are very low in major rock-forming silicates, 0.05e2.3 mg kg1, and in carbonates, 1e8 mg/kg. The

highest arsenic concentrations (20e200 mg/kg) are typically observed in organic and sulphide-rich shales, sedimentary ironstones, phosphatic rocks, and some coals (Smedley and Kinniburgh, 2002). In rural areas of India and Bangladesh, ground water is the main source for drinking water through dug well and tubewells. It has been reported that in India many districts in West Bengal are suffering from arsenic problem. Recent studies carried out in northeastern England revealed arsenic enrichment within the urban and industrially affected rivers. Arsenic concentration in rural area averaged between 0.6 and 0.9 mg/L and in between 3.2 and 5.6 mg/L for the rivers influenced by industrial discharges (Escudero et al., 2009).

* Corresponding author. E-mail addresses: [email protected] (R.M. Dhoble), [email protected] (S. Lunge), [email protected] (S. Rayalu). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.06.016

4770

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 6 9 e4 7 8 1

Arsenic occurs in different forms and inorganic trivalent form of arsenic (III) is the most toxic among the various forms of arsenic present in natural water. The common valencies of geogenic arsenic in ground water sources are As (III) (arsenite) and As(V) (arsenate). The inorganic hydrolysed species are 2 3 and H3AsO4, present as H3AsO3, H2 AsO 3 , HAsO3 , AsO3  2 3 H2 AsO4 , HAsO4 and AsO4 (Gupta et al., 1997). As per United State Environmental Protection Agency (USEPA) and International Agency for Research on Cancer, arsenic is classified as a human carcinogen. Arsenic in drinking water indicates that arsenic could cause liver, lung, kidney and skin cancer (Smith et al., 1992). An acute high dose of arsenic by oral intake cause gastrointestinal irritation resulting in difficultly in swallowing, thirst, abnormal high blood pressure and convulsions (Pontius et al., 1994). It has been reported that arsenic may cause neurological damage to those who drink water contaminated with slightly greater than 0.1 mg/L of As (III). The lowest arsenic concentration in water sample producing dermatosis was found to be 0.2 mg/L (Chakraborti and Saha, 1987). The total quantity of arsenic consumed per day and the duration of exposure are very important factors. WHO provisional guideline value for arsenic in drinking water is 10 mg/l (Who, 2004). Various physio-chemical treatment methods have been adopted to remove arsenic from drinking water both in laboratory and field condition below the MCL (maximum contamination level). Among a variety of technologies (including precipitation, coagulation, membrane separation, ion exchange, lime softening and adsorption), adsorption and coagulation are believed to be the cost effective method. Although coagulation with iron and aluminium salts is more effective, the requirement of skilled operator and the introduction of contaminants into the water limit its application in small community and household levels. Since solid adsorbents are easy to handle and are appropriate for use in country side where high arsenic ground water mostly occurs, adsorption has received much attention on As removal. Iron containing substances have been widely investigated to remove arsenic from aqueous solution due to their high specific surface area, including Mn-substituted Fe oxyhydroxide (Lakshmipathiraj et al., 2006), granular ferric hydroxide (Banerjee et al., 2008), ferrihydrite (Jessen et al., 2005), goethite (Sun and Doner, 1998), zero valent iron (Nikolaidis et al., 2003), Ce(IV)- doped Fe oxide (Zhang et al., 2003), copper oxide incorporated alumina (Pillewan et al., 2011), natural hematite and natural siderite (Guo et al., 2007). This study investigates the feasibility of the magnetic adsorbent for trivalent arsenic As(III) removal from aqueous solution. The main objectives are (i) to understand the As(III) adsorption kinetics, (ii) to evaluate the influence of temperature, pH and coexisting anions on the As(III) removal capacities; and (iii) to describe and explain some important thermodynamic parameters.

2.

Materials and methods

2.1.

Reagents

All chemicals were analytical grade. All stock and fresh solutions were prepared in deionized water for entire study.

Standards for calibration were prepared from As (III) standard reference sodium (Meta) arsenite. Stock solution (1000 mg/L) was prepared from sodium arsenite and frozen to prevent oxidation. Solutions of As (III) of 100 mg/L were prepared in every fortnight and working solutions of As (III) were prepared according to experiment requirements. pH was adjusted by standard acid and base solutions of 0.1 N HCl and 0.1 N NaOH respectively. For the study of effect of adsorption due to presence of background ions in ground water, Ca(NO3)4$H2O, MgSO4.7H2O, FeSO4.7H2O, NaCl, Na2SO4, NaNO3and Na2HPO4 (Merck India) salts were used. Effect of temperature on arsenite adsorption was studied by varying temperature.

2.2.

Synthesis

The Magnetic binary oxide particles (MBOP) were synthesized by template method. 27 g of chitosan was dissolved in 900 ml of 5% acetic acid with constant stirring on mechanical stirrer for 1 h. In one beaker 84.78 g of aluminium nitrate was dissolved in 100 ml of distilled water and in another beaker 97.89 g of ferrous nitrate was dissolved in 100 ml of distilled water. Aluminium nitrate and ferrous nitrate solutions added to the chitosan gel with stirring for 180 min. The resulting Al-Fe-chitosan slurry was added drop wise into NH4OH solution (50% v/v) under vigorous stirring, using a syringe pump. The gel macro spheres formed were allowed to stabilize in NH4OH solution for 60 min. The beads were separated from the NH4OH solution and washed with deionised water and dried at 70  C for 24 h in oven. The dried beads were calcined at 450  C for 6 h in muffle furnace. Finally the calcined product was subjected to multiple washing with deionised water and dried at 80  C.

2.3.

Methods

MBOP was characterized by using different techniques like Xray diffraction, Scanning electron microscopy (SEM), Wave length energy dispersive analysis of X-ray (WDAX), and BET surface area analysis. Magnetic properties of adsorbent were examined by Vibrating sampler magnetometer (VSM). The Xray pattern of adsorbent was recorded on Rigaku X-ray diffractometer. SEM was performed by using JEOL-6380A for analysing the surface morphology of the material. Composition of material was determined by wave length energy dispersive analysis of X-ray (WDAX). Surface area of adsorbent was measured by Brauner, Emmett and Teller (BET) method using micromeritics ASAP 2010 surface area analyzer.

2.4.

Batch study

Batch adsorptions were carried out by shaking 50 ml of arsenic (III) samples in a controlled rotary shaking machine (Model no. CIS-24, Remi Instruments, Mumbai, India) at a speed of 150 rpm. The solution was taken in glass stopper bottles of 125 ml capacity. The dose of adsorbent and arsenite concentration was varied within feasible parameter range. The solution was then filtered through Whatman filter paper no. 42 and the filtrate was analyzed for residual arsenic after adsorption in Atomic absorption spectrophotometer hydride vapour generator (AASHVG-1) 6300 Shimadzu Japan 2007. All adsorption experiments were conducted at a room

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 6 9 e4 7 8 1

temperature of 30  2  C to investigate the effect of various parameters like adsorbent dose, initial arsenic concentration, presence of interfering ions and pH etc. The specific amount of arsenic adsorbed was calculated from the following equation qe ¼ ðC0  Ce Þ 

V W

(1)

and %Removal ¼

ðC0  Ce Þ  100 Ce

4771

NaOH. According to the WHO guidelines, the normal range of pH in drinking water lies in the range 6.0e8.5, and is mostly dependent upon the geological characteristics of the soil and weather conditions. Also, pH of the solution is one of the major factors which significantly affect the As (III) adsorption. Hence, it is necessary to study the effect of pH on removal of As (III) from water. The effect of background ions i.e. cations and anions commonly present in ground water was also studied with different proportions.

(2)

where qe is the adsorption amount (mg/g) in the solid at equilibrium; Co, Ce are initial and equilibrium concentrations of arsenic (mg/L), respectively; V is volume (ml) of the aqueous solution and W is the mass (g) of adsorbent used in the experiments. The effect of pH on As (III) removal was studied by adjusting the pH of the solution using 0.1 N HCl and 0.1 N

2.5.

Kinetic study

In order to estimate equilibrium adsorption rate for the uptake of As (III) by MBOP, time dependent sorption studies were conducted. Adsorption kinetics was monitored by adding known weight of MBOP into 50 ml of 1 mg/L arsenic solution at 293 K, 303 K and 313 K stirred at 150 rpm. A portion of solution

Fig. 1 e (a) XRD of MBOP, (b) SEM of MBOP, (c) SEM of MBOP after arsenic (III) adsorption, (d) FTIR of MBOP, (e) VSM magnetization curves for MBOP, (f) Photographs of MBOP attracted by magnetic bar a) before adsorption b) after adsorption.

4772

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 6 9 e4 7 8 1

Fig. 1 e (continued).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 6 9 e4 7 8 1

was withdrawn from the vessel at predetermined time intervals and filtered. The filtrate as analyzed for residual concentration of As (III) using Atomic absorption spectrophotometer.

that the MBOP is attracted by the magnetic bar. This further confirms that MBOP is magnetic nature.

3.2. 2.6.

Thermodynamics

Thermodynamic parameters of adsorption including standard free energy change (ΔG ), standard Enthalpy change (ΔH ), and standard entropy change (ΔS ) were calculated at 293 K, 303 K and 313 K temperature.

3.

Results and discussion

3.1.

Characterization of MBOP

The prepared MBOP is reddish brown coloured granular adsorbent. The X-ray diffraction spectrum pattern of the MBOP did not show any sharp peak (Fig. 1a), thereby indicating the amorphous nature of the product. The surface morphology of MBOP before and after arsenic adsorption was studied from SEM. It can be seen from Fig. 1b the adsorbent has expected large number of porous structure which indicates the adsorbent may have a high surface area and high adsorption capacity. These large pores are formed by the elimination of chitosan template during calcinations step of synthesis. After arsenic adsorption the surface morphology of MBOP remains unchanged suggesting physical adsorption of arsenic (Fig. 1c). The BET surface area, average pore size and total pore volume of MBOP was observed to be 123.28 m2/g, 61.59 Ǻ, and 0.1732 cm3/g, respectively. The radius of arsenite (0.58 Ǻ) is much smaller than the pore size of the MBOP. This may facilitate increased dispersion of arsenite in the inner layer of the granular MBOP. The chemical analysis of the product gave iron, aluminium and oxygen contents as 42.6%, 16.69%, and 34.21%, respectively as analyzed by WDAX. The FTIR spectrum of MBOP (Fig. 1d) indicates the presence of predominant peaks at 3512.38, 3321.24 cm1 (eOH and eNH stretching vibrations), 2900.67and 2342.24 cm_1 (eCH stretching vibration, 1648.20 cm1 (eNH bending vibration in eNH2)), 1378.91 cm1 (eNH deformation vibration in eNH2). The low intensity band at 1062.40 cm1 is attributed to FeeOH structural vibration. The band between 400 and 450 cm1 could be due to the superposition of the characteristic stretching bands of aluminium oxide. The bands observed between 1100 and 500 cm1 could be characteristic vibrations of aluminium oxide.

3.1.1.

4773

Effect of dose of MBOP

The influence of adsorbent dose on As (III) removal at a fixed initial arsenic concentration of 1 mg/L and neutral pH is shown in Fig. 2. It was noticed that percentage removal of As (III) increased from 27% to 99% with an increase in adsorbent dose from 0.01 g/L to 2.0 g/L respectively which is due to the higher active site/As (III) ratio. However, it was noticed that after a dosage of 0.5 g/L, there was no significant change in the percentage removal of As (III). Usually a point of intersection in this graph is considered as the optimum dose as this point represents balance between % As (III) removal and adsorption capacity. The intersection point is at 0.03 g/L however dose of 1 g/L has been selected as it was required to bring down the arsenic level below 0.01 mg/L (10 ppb) as per WHO guidelines. Adsorbent dose of 1 g/L was used for further study.

3.3.

Effect of pH on As (III) uptake

pH plays significant role in adsorption-based water treatment processes, because proton concentration can strongly modify the redox potential of sorbates and sorbents, enhance dissolution of the adsorbent material and modify chemical speciation of the adsorbates as well as surface charge of adsorbent (Escudero et al., 2009). The effect of pH on arsenic removal by MBOP was studied over a broad pH range of 3e11 with adsorbent dose of 1 g/L; initial concentration 1 mg/L, shaking speed of 150 rpm and contact time of 24 h. The effect of pH on arsenic (III) adsorption is shown in Fig. 3. pH of the arsenic contaminated ground water is normally reported between 7 and 9 and there is drastic reduction in the uptake capacity of most of the adsorbents in the pH above 7. It is evident from Fig. 3 that there was no significant effect of pH on As (III) adsorption over a wide range of pH 3 to 9 which is highly advantageous for practical operation. This may be due to the specific chemical reaction interaction between adsorbate and adsorbent surface. Arsenic and (III) exists in non-ionic (H3AsO3) and anionic (H2 AsO1 3 ) form in the pH range 2e7 and 7.5e9 respectively. In the HAsO2 3

Magnetic property of MBOP

A distinct advantage of MBOP is that adsorbent can readily be isolated from solution by application of an external magnetic field. Fig. 1e shows the VSM magnetization curves for MBOP at room temperature. MBOP exhibited typical superparamagnetic behaviour, characterized by strong magnetic susceptibility. Saturation magnetization is a key factor for successful magnetic separation. Ma et al. observed that saturation value of 16.3 emu/g was sufficient for magnetic separation with a conventional magnet (Ma et al., 2005). Thus, the saturation magnetization value achieved with MBOP was high (18.78 emu/g) enough for magnetic separation. Fig. 1f shows

Fig. 2 e Effect of adsorbent dose on As(III) removal.

4774

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 6 9 e4 7 8 1

Fig. 3 e Effect of pH on As (III) adsorption by MOA.

acidic range, when non-ionic As (III) species comes in contact with the adsorbent, the adsorbent surface is charged positively which helps in conversion of non-ionic arsenic to its anionic form, which in turns assists in the adsorption process (Kundu and Gupta, 2006). However, slight decrease in arsenic (III) adsorption capacity was observed above pH 9 which may be due to i) competition of excessive OH ˉ ions for adsorption, ii) negative surface charge of adsorbent at alkaline conditions and iii) negatively charged H2 AsO 3 species starts dominating. When neutral species of H3AsO3 exists then the maximum removal of arsenic (III) occurs. In alkaline medium, the negatively charged H2 AsO1 3 species start dominating and surface also tends to acquire negative charges. This tendency of adsorbate species and adsorbent surface will continue to increase with increase of pH causing a gradual increase in repulsive forces between the surface and adsorbate species resulting in a decrease of adsorption (Rajan et al., 2009). The results obtaned are nearly similar to those described by Kundu and Gupta. (2006). Keeping in view practical operating conditions and drinking water standard for pH, pH 7.0 appears to be optimal and has been used in the entire study.

3.4.

Effect of initial concentration

Fig. 4 shows the effect of initial arsenic concentration on adsorption of As (III) by MBOP. It is observed that the percent removal of As (III) decreases while the equilibrium As (III) adsorption capacity increases with the increase in initial As (III) concentration. This decrease of percent arsenic (III) removal may be attributed to the fact that at higher As (III) concentration; the number of active sites on adsorbent surface is not enough to accommodate arsenic (III) ions. However, at low As (III) concentration, the ratio of surface active sites to total As (III) is high and therefore As (III) ions can interact with the active sites on adsorbent surface sufficiently.

3.5.

Effect of background ions

In ground water several ionic species are present and these ions may interfere in the uptake of arsenic by the adsorbent

Fig. 4 e Effect of initial As (III) concentration on As(III) removal by MBOP.

through the competitive binding or complexation (Bhaskar et al., 2006). Batch equilibrium experiments were conducted to find the individual effect of cations (Ca2þ, Mg2þ and Fe3þ) 3  and anions (SO2 4 , PO4 , Cl and NO3 ) on adsorption of As (III). The salts used in this study includes MgSO4$7H2O, Ca(NO3)2.4H2O, FeSO4$7H2O, Na2SO4, NaCl and Na2HPO4. The percentage removal of As (III) was compared with samples having no background ions. As it is evident from Fig. 5a, with increase in concentration of Ca2þ, Mg2þ from 50 mg/L to 600 mg/L, adsorption of As (III) was reduced significantly from 85.37 to 57.72% and 84.81 to 68.49% respectively and for similar concentration of Fe3þ the adsorption of arsenic (III) increased from 37% to 95%. However iron concentration in ground water is generally very low; hence effect on arsenic adsorption at low concentration was also studied. Increase in Fe3þ concentration from 2 mg/L to 25 mg/L resulted in increase of As (III) removal from 13.64% to 34.7% (Fig. 5b). Fig. 5c shows the effect of presence of anions. It was 2 ions has no observed that the presence of NO 3 and SO4 significant effect on As (III) adsorption in concentration range of 100e800 mg/L. In presence of Cl- ions, decrease in As (III) adsorption was observed compared to As (III) adsorption in blank (without cations and anions). It was observed that at Clconcentration of 100 mg/L, there is drastic decrease in As (III) adsorption than As (III) adsorption at Cl- concentration >100 mg/L. In presence of PO-4 ions, decrease in As (III) adsorption was observed. At lower concentration of 100 mg/L, it was observed that PO-4 does not interfere much in As (III) adsorption. However at PO-4 concentration > 100 mg/L, nega tive effect of PO 4 ions was observed up to 600 mg/L PO4 .  Further increase in PO4 concentration upto 800 mg/L has no significant effect on As (III) adsorption.

3.6.

Adsorption isotherms

In order to study the dominant adsorption mechanism and to compute various adsorption parameters three isotherm

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 6 9 e4 7 8 1

4775

Fig. 5 e a: Effect of presence of Ca, Mg and Fe on arsenic removal. b: Effect arsenic removal in presence of Fe at low concentration. c: Effect of background anions on arsenic removal by MBOP.

models namely Langmuir, Freundlich and D-R isotherms were used. The Langmuir adsorption model can be represented in linear form as follows: 1 1 1 1  þ ¼ qe qmax b Ce qmax

(3)

Where qmax is the maximum amount of the arsenic ion per unit weight of adsorbent to form a complete monolayer on the surface bound at high Ce, while b is a constant related to the affinity of the binding sites. qe represent a particle limiting adsorption capacity when the surface is fully covered with solute. Langmuir parameters, qmax and b were calculated from the slope and intercept of the linear plots of 1/qe vs 1/Ce (Fig. 6a). The Freundlich model indicates the heterogeneity of the adsorbent surface and considers multilayer adsorption. The linear form of Freundlich adsorption model is as follows (Pillewan et al., 2011):


log qe ¼ log KF þ 1=nlog ðCe Þ

(4)

Where KF and 1/n are Freundlich constants, related to adsorption capacity and adsorption intensity (heterogeneity factor) respectively. The values of KF and 1/n were obtained from the slope and intercept of the linear Freundlich plot of log qe vs log Ce (Fig. 6b). In order to predict the adsorption efficiency of the process, the dimensionless quantity (r) was calculated by using the following equation. r¼

1 1 þ bC0

(5)

Where C0 and b are the initial concentration of arsenic and Langmuir isotherm constant. If the value of r < 1, it represents favourable adsorption while greater than 1.0 represents unfavourable adsorption.

4776

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 6 9 e4 7 8 1

towards iron and also it is reported that metal oxide incorporation or coating increase the zeta potential to more positive values resulting in enhanced anion sorption (Maliyekkala et al., 2009). The mean free energy (E ) was of adsorption was calculated from the kads value using the following equation: 0:5

E ¼ ð2kÞ

(8)

As shown in Table 1, the E value is 5.0 kJ/mol for As (III) on the MBOP medium. The numerical value of mean free energy is in the range of 1e8 and 9e16 kJ/mol for physical and chemical adsorption respectively (Saeed, 2003). In the present study the E value is less than 8 kJ/mol which is within the energy range of physical adsorption which implies that the type of adsorption is physical.

3.7.

Adsorption kinetics

Kinetic models are used to examine the rate of the adsorption process and potential rate-controlling step. The capability of the pseudo-first-order and pseudo-second-order kinetic model was examined in this study. The pseudo-first-order equation of Lagergren is generally expressed as follows (Pillewan et al., 2011):
Kad t log qe  q ¼ log qe  2:303

Fig. 6 e Adsorption isotherm a) Langmuir fit, b) Freundlich fit for arsenic adsorption by MBOP.

Where qe and q (both in mg/g) are the amount of arsenic adsorbed per unit mass of adsorbent at equilibrium and time “t” respectively. The adsorption rate constant (Kad) for arsenic sorption was calculated from the slope of the linear plot log (qeq) vs time (t) as shown in Fig. 7a. The pseudo-second-order model is also commonly used to predict the kinetic parameters linear for of which can be written as (Pillewan et al., 2011)

The DeR isotherm model considers more common features of adsorption and is not based on homogenous monolayer adsorption like Langmuir. The linear form of DeR isotherm model is as follows (Pillewan et al., 2011):

and

ln qe ¼ ln Qm  kads 32

h ¼ kq2e

(6)

(9)

t 1 t ¼ þ qt h qe

(10)

(11)

and   1 3 ¼ RTln 1 þ Ce

(7)

Where Qm is the theoretical adsorption capacity (mg/g), kads is a constant related to adsorption energy, Ɛ is polyani potential, R is gas constant (kJ/mol. K), T is temperature (K ). Results of experimental data were fitted to these three isotherms models to determine which model most accurately described adsorption by the adsorbent. The results of various adsorption parameters obtained from these isotherm model are also presented in Table 1. On comparison of the fitness of the three isotherms it is evident that for As (III) the experimental data were well fitted to Langmuir model followed by Freundlich and D-R models signifying the monolayer adsorption of arsenic on uniform surface. The values of adsorption capacity for MBOP obtained from the Langmuir model was 16.94 mg/g. The significantly high adsorption capacity is probably due to increased affinity of arsenic

Table 1 e Adsorption isotherm parameters for As(III) adsorption by MBOP. Isotherm parameters

293 K

303 K

313 K

Langmuir isotherm qmax (mg/g) b (L/mg) R2

23.256 7.167 0.997

16.949 5.9 0.996

14.925 3.941 0.991

Freundlich Isotherm Kf (mg/g) 1/n R2

5.57 0.787 0.993

5.007 0.835 0.988

4.216 0.846 0.983

DeR Isotherm Kads Qm (mg/g) R2 E (KJ/mol)

-2E-8 24.16 0.948 5.0

-2E-8 19.7 0.931 5.0

-2E-8 15.64 0.928 5.0

4777

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 6 9 e4 7 8 1

of pseudo-first-order (Eq. (5)) and pseudo-second-order (Eq. (7)) models are presented in Fig. 7b. The values of kad, k and h and correlation coefficients obtained from the linear plots are also presented in Table 2. It is apparent from the values of correlation coefficients that the pseudo-first-order kinetic model fitted well as compared to pseudo-second-order model. Sorption of a liquid adsorbate on porous solid adsorbent can be modelled by diffusion models, which can be particle diffusion and intra-particle pore diffusion model. The particle diffusion model can be written as (Pillewan et al., 2011): 

Ct Ce

ln

 ¼ kp t

(12)

Where kp is the particle diffusion coefficient (min1). The value of kp can be obtained by slope of the plot between ln (Ct/Ce) and t (Fig. 8a). The intra-particle pore diffusion model given by Weber and Morris is also commonly used to characterize the sorption data. In order to test the contribution of intra-particle pore diffusion on the adsorption process, the rate constant for intra-particle pore diffusion was obtained by using following equation. According to this model, if the rate limiting step is diffusion of adsorbate within the pores of adsorbent particle (intra-particle diffusion) a graph between amount of adsorbate adsorbed and square root of time should give a straight line passing through the origin. The equation can be written as (Pillewan et al., 2011): qt ¼ ki t1=2

Fig. 7 e Kinetic a) Pseudo-first order model, b) Pseudosecond order model for arsenic removal by MBOP.

Where qt is the amount of arsenic adsorbed at time t (mg/g), qe is the amount of arsenic adsorbed at equilibrium (mg/g), h is the initial sorption rate (mg/g min). The values of qe (1/slope), k (slope2/intercept) and h (1/intercept) can be calculated from the plots of t/qt versus t and given in Table 2. The linear plots

(13)

Where ki (mg/g min1/2) is the intra-particle pore diffusion rate constant, qt amount of arsenic adsorbed per unit mass of adsorbent at any time t, was plotted as a function of square root of time t1/2 (Fig. 8b). The plots of linear forms of particle diffusion and intra-particle pore diffusion models are given in Fig. 8 and b respectively for As(III) and the values of different parameters are given in Table 2. The values of R2 for intraparticle pore diffusion model are closer to unity indicating that intra-particle pore diffusion of adsorbate is contributing more towards rate determining step. However, in case of intra-particle diffusion model the lines are not passing through the origin, which reveals that the adsorption of arsenic on MBOP is a complex process involving surface

Table 2 e Various kinetic and diffusion parameters for As(III) adsorption by MBOP. Lagergren parameters T (K) 293 303 313

Pseudo-second-order parameters 1

1

2

Kad (min )

R

0.002303 0.002303 0.002303

0.99 0.99 0.99

k (g mg

293 303 313

min )

h (mg g1 min1)

R2

0.00591 0.005817 0.005599

0.99 0.98 0.97

0.004916 0.004243 0.003583

Particle diffusion model T (K)

1

Intra-particle pore diffusion model

Kp (min1)

R2

Ki (mg g1 min1/2)

R2

0.009 0.03 0.122

0.86 0.89 0.88

0.033 0.035 0.037

0.967 0.982 0.972

4778

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 6 9 e4 7 8 1

3.9.

Thermodynamic parameters

Thermodynamic parameters of adsorption namely standard free energy change (ΔG ), standard enthalpy change (ΔH ), and standard entropy change (ΔS ) were calculated using standard methods. Standard free energy change (ΔG ) is given by the equation (Sekar et al., 2004): DG ¼ RT ln ðK0 Þ

(15)



Where ΔG standard free energy change of sorption (KJ/mol), T the temperature in Kelvin and R is universal gas constant (8.314 J/mol K) and K0 is the thermodynamic equilibrium constant equal to qmax  b of Langmuir isotherm (Sekar et al., 2004). The standard enthalpy change (ΔH ), and standard entropy change (ΔS ) was calculated using following equation (Sekar et al., 2004): ln ðK0 Þ ¼

DS DH  R RT

(16)

Where (ΔH ) is standard enthalpy change (KJ/mol) and (ΔS ) is standard entropy change (KJ/mol K). The values of ΔH and ΔS were obtained from the slope and intercept of the Vant Hoff’s plot of ln(K0) against 1/T (Fig. 10). These values are observed to be 39.674 kJ/mol and 0.0928 kJ/mol K respectively. The negative values of ΔH indicated the exothermic nature of the sorption process. Negative value of entropy change (ΔS ), indicate a greater order of reaction during the adsorption of As (III). It may be due to the fixation of As (III) to the exchanger sites resulting in a decrease in the degree of freedom of the system. The negative values of ΔG 12.46.798, 11.601, 10.603 kJ/mol at all temperatures studied indicated feasibility and spontaneity of the sorption reaction.

3.10. Fig. 8 e a) Particle diffusion model, b) WebereMorris plot for arsenic adsorption by MBOP.

Adsorption of As(V) and field trial

Most of the ground water or other water supplies contain both As species III and V. Hence it is necessary to study the removal

adsorption, inter-particle diffusion and intra-particle diffusion all contributing towards the rate of sorption.

3.8.

Mass transfer coefficient

Mass transfer analysis for the removal of arsenic (III) was carried out using the following equation:  ln

     Ct 1 MK 1 þ MK  ¼ ln  b$Ss$t C0 1 þ MK 1 þ MK MK

(14)

Where K is the constant obtained by multiplying Qmax and b (L/g), M is the mass of the adsorbent per unit volume of particle free adsorbate solution (g/L). Ss is the outer surface of adsorbent per unit volume of particle free slurry (L/cm) and b is the mass transfer coefficient (cm/min). ln((Ct/C0)1/ (1 þ MK)) versus t for the temperature of 293, 303 and 313 K gives the straight line of slope ((1 þ MK )/MK )bSs and the value of mass transfer coefficient b was calculated from the slope of the plots and was found to be as 1.871  105, 2.807  105and 3.743  105 cm/s, respectively (Fig. 9).

Fig. 9 e Estimation of mass transfer coefficient of MBOP for arsenic (III) removal.

4779

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 6 9 e4 7 8 1

Table 3 e Physio-chemical parameters of field water before and after treatment with MBOP. Parameter

Fig. 10 e Vant Hoff’s plot for removal of arsenic by MBOP.

of As(V) by MBOP. Adsorption capacities were calculated using Langmuir isotherm model (Fig. 11). MBOP has adsorption capacity of 11.31 mg/g for As(V). Adsorption capacity of MBOP was calculated for arsenic III and V when these species exist together in equimolar ratio. For this, deionized water was spiked with arsenic III and V (w1 mg/L) and equilibrium concentration of arsenic III and V was monitored at different dose of MBOP. Adsorption capacity of 11.4 and 11.14 mg/g was obtained for arsenic III and V respectively which indicates that MBOP has equal affinity for arsenic III and V. Considering the practical application, MBOP was also tested for arsenic removal in ground water contaminated with arsenic III and V. The physico-chemical parameters of ground water before and after adsorption are given in Table 3. At an adsorbent dose of 1 g/L, MBOP has reduced the arsenic III and V concentration in ground water from 983.71 to 998.91 mg/L to 7.44 and 9.89 mg/L (992.6 mg/L to 9.81 mg/L), which is below WHO permissible limit for arsenic. Also there was no major change in the other water quality parameter of water after removal of arsenic by MBOP. The residual concentration of iron and aluminium ion in treated water was 0.206 mg/L and 0.0018 mg/L respectively, which is lower than the standard set by WHO for drinking water indicating that the MBOP can be

Fig. 11 e Isothem plot for adsorption of As(V), As (III) in (III D V) system, As (V) in (III D V).

pH Colour Odour Turbidity Alkalinity Total hardness Conductivity Chloride Sulphate TDS Ca2þ Mg2þ Fe2þ Aluminium As (V) Arsenic (III)

Unit

Before After treatment treatment

6.58 Hazans 0.4 Odourless NTU 3.9 mg/L 120 CaCO3 170 ms/cm 110 mg/L 33.74 mg/L 146 mg/L 350.58 mg/L 29 mg/L 19.2 mg/L 0.215 mg/L BDL ppb 998.91 ppb 983.71

6.67 0.2 Odourless 4.1 110 160 98 28.93 138 340.56 31 9.6 0.206 0.018 9.89 7.44

Permissible limit 6.5e8.5 5 Unobjectionable 5 200 300 250 200 500 75 30 0.3 0.03 10 10

used for treatment of arsenic III and V contaminated drinking water.

4.

Mechanism of arsenic adsorption

From XRD it is clear that the product is amorphous. It has been reported that the adsorption of arsenic on amorphous metal oxides is through formation of inner sphere surface complexes which are mainly attached as bidentate linkages with some monodentat linkages (Maliyekkala et al., 2009). It has also been reported that adsorption following formation of inner sphere complex are not much influenced by pH and ionic strength (Goldberg and Johnston, 2001). Removal of arsenic by MBOP remains almost constant at different pH (Fig. 3) which also indicates that the adsorption of arsenic on MBOP is through formation of inner sphere complex through formation of AseO linkages.

Fig. 12 e Desorption of MBOP using NaOH.

4780

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 6 9 e4 7 8 1

Table 4 e Comparison between various adsorbent used for arsenic removal on the basis of adsorption capacity. Adsorbent Iron-impregnated chitosan granular Modified Native Cellulose Fibre Molybdate-impregnated chitosan beads (MICB) Iron oxide coated sponge Copper Oxide Incorporated Mesoporous Alumina Activated alumina Iron oxide coated sand Activated alumina Iron oxide impregnated activated alumina CuO Magnetic binary oxide particle (MBOP)

5.

pH

Initial concentration of As (III) (mg/L)

Ref.

8.0 8.0 5.0

1.007 10 10

6.48 8.96 1.98

Gang et al., 2010 Tian et al., 2011 Chen et al., 2008

7.3 7.0

1.0 1.0

3.85 2.61

Nguyen et al., 2010 Pillewan et al., 2011

7.0 7.5 7.6 12.0

e 0.4 1.0 1.4

3.5 0.029 0.18 0.734

Lin and Wu, 2001 Gupta et al., 2005 Singh and Pant, 2004 Shugi et al., 2004

8 7.0

0.1e100 1.0

Desorption study

Considering the practical applicability in field it is desirable that an adsorbent should be fully regenerated and reused so that it can be put into cyclic use in a cost effective manner. Regeneration of arsenic saturated adsorbents has been achieved using either alkali or strong acids including desorption of arsenic from MBOP. To study the regeneration, MBOP was first saturated with arsenic by shaking the adsorbent with initial arsenic concentration of 1 mg/L and adsorbent dose of 1 g/L for 24 h. This adsorption cycle was repeated till the adsorbent get saturated. Regeneration studies were conducted by shaking the required quantity of arsenic saturated MBOP with different concentrations of NaOH for 1 h. The results of regeneration studies are presented in Fig. 12. As evident from the results the amount of As (III) leached decreases at NaOH concentration of 5% and 10%. With NaOH concentration of 1% and 2% almost all the As (III) desorbed (more than 95% regeneration) in 1 h resulting in complete regeneration of MBOP and MBOP retained the original adsorption capacity after one complete adsorption desorption cycle, confirming the reusability of MBOP for arsenic removal. The results of regeneration studies suggest that the MBOP can be used in a continuous flow for removal of arsenic.

6.

Adsorption capacity qmax (mg/g)

Comparison with other adsorbents

A comparison has been made between MBOP and previously reported adsorbents for arsenic removal (Table 4). For comparison, Langmuir adsorption capacity was considered. An analytical comparison shows that MBOP is better than many other adsorbents except CuO in terms of adsorption capacity with additional feature of magnetic separation. Cost analysis has been done for the production of MBOP. The cost of MBOP works out to be Rs. 320 per Kg which includes cost of raw materials and process cost. The water treatment cost was calculated on the basis of dose of MBOP required to treat 100 L

26.9 16.94

Martinson and Reddy, 2009 Present work

of water with arsenic concentration 0.001e2.5 mg/L which works out to be Rs. 32. The cost MBOP and water treatment cost appears to be very economic.

7.

Conclusion

Magnetic adsorption process provides a cost effective and environmentally benign water treatment process. Consequently, a novel binary oxide with magnetic property was prepared, characterized and applied for in the removal arsenic as a model contaminant in water. The material was effective in arsenic removal in water. The arsenic uptake was rapid initially up to 3 h and then gradually slows down as it reaches to equilibrium. Arsenic uptake depends on the initial concentration, adsorbent dose and temperature and independent upon the pH. The adsorption kinetics follows pseudo-second order kinetic model. The adsorption process was exothermic in nature. The equilibrium data fitted well to the Langmuir and Freundlich model. Adsorption of As (III) was not affected remarkably by presence of other anions like chloride, nitrate, sulphate etc. Up to 95% of the adsorbed arsenic on MBOP was desorbed using 1% and 2% NaOH solution. The adsorbent retained the original adsorption capacity after one complete adsorption desorption cycle, confirming the reusability of MBOP for arsenic removal. Further tests are still required to determine the robustness of the material and fix bed column studies for practical application of material.

Acknowledgement We thank Director, NEERI for providing research facilities. We thankfully acknowledge to the Council of Scientific and Industrial Research (CSIR) as one of the author Mrs. Sneha Lunge is CSIR Senior Research Fellow (SRF). We thankfully acknowledge to Sophisticated Analytical Instrument facilities, IIT Chennai for VSM test.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 6 9 e4 7 8 1

references

Banerjee, K., Amy, G.L., Prevost, M., Nour, S., Jekel, M., Gallagher, P. M., Blumenschein, C.D., 2008. Kinetic and thermodynamic aspects of adsorption of arsenic onto granular ferric hydroxide (GFH). Water Research 42, 3371e3378. Bhaskar, P.B., Gupta, A.K., Ayoob, S., Kandu, S., 2006. Investigation of arsenic (V) removal by modified calcined bauxite. Colloids and Surfaces, 237e245. Chakraborti, A.K., Saha, K.C., 1987. Arsenical dermatosis from tube well water in West Bengal, India. Journal of Medical Research 85 (3), 326e334. Chen, C.-Y., Chang, T.-h., Kuo, K.-T., Chen, Y.-F., Chung, Y.-C., 2008. Characteristics of molybdate-impregnated chitosan beads (MICB) in terms of arsenic removal from water and the application of a MICB-packed column to remove arsenic from wastewater. Bioresource Technology 99, 7487e7494. Escudero, C., Fiol, N., Villaescusa, I., Bollinger, J.C., 2009. Arsenic removal by waste metal (hydr)oxide entrapped into calcium alginate beads. Journal of Hazardous Materials 164, 533e541. Gang, D., Deng, B., Lin, L., 2010. As (III) removal using an ironimpregnated chitosan sorbent. Journal of Hazardous Materials 182, 156e161. Goldberg, S., Johnston, C.T., 2001. Mechanisms of arsenic adsorption on amorphous oxides evaluated using macroscopic measurements, vibrational spectroscopy, and surface complexation modeling. Journal of Colloid and Interface Science 234 (1), 204e216. Guo, H.M., Stuben, D., Berner, Z., 2007. Adsorption of arsenic (III) and arsenic(V) from groundwater using natural siderite as the adsorbent. Journal of Colloid Interface Science 315, 47e53. Gupta, V.K., Saini, V.K., Jain, N., 2005. Adsorption of As (III) from aqueous solutions by iron oxide-coated sand. Journal of Colloid and Interface Science 288, 55e60. Gupta, A. Datta, Bandopadhyay, A., Pratip, 1997. Technologies and Options for Arsenic Removal IWWA, Annual Convention, Calcutta 1-9. Jessen, S., Larsen, F., Koch, C.B., Avin, E., 2005. Sorption and desorption of arsenic to ferrihydrite in a sand filter. Environmental Science and Technology 39, 8045e8051. Kundu, S., Gupta, A.K., 2006. Adsorptive removal of As (III) from aqueous solution using iron oxide coated cement (IOCC): evaluation of kinetics, equilibrium and thermodynamic models. Separation and Purification Technology 51 (2), 165e172. Lakshmipathiraj, P., Narasimhan, B.R.V., Prabhakar, S., Bhaskar, R.G., 2006. Adsorption studies of arsenic on Mnsubstituted iron oxyhydroxide. Journal of Colloid Interface Science 304, 317e322. Lin, T.F., Wu, J.K., 2001. Adsorption of arsenite and arsenate within activated alumina grains: equilibrium and kinetics. Water Research 35 (8), 2049e2057. Ma, Z.Y., Guan, Y.P., Liu, H.Z., 2005. Synthesis and characterization of micron-sizedmonodisperse supermagnetic polymer particles with amino groups. Journal of Polymer Science and Polymer Chemistry 43, 3433e3439.

4781

Maliyekkala, S.M., Philip, L., Pradeep, T., 2009. As(III) removal from drinking water using manganese oxide-coated-alumina: performance evaluation and mechanistic details of surface binding. Chemical Engineering Journal 153, 101e107. Martinson, C.A., Reddy, K., 2009. Adsorption of arsenic (III) and arsenic (V) by cupric oxide nanoparticles. Journal of Colloid and Interface Science 336, 406e411. Nguyen, T.V., Vigneswaran, S., Ngo, H., Kandasamy, J., 2010. Arsenic removal by iron oxide coated sponge: experimental performance and mathematical models. Journal of Hazardous Materials 182, 723e729. Nikolaidis, N.P., Dobbs, G.M., Lackovic, J.A., 2003. Arsenic removal by zerovalent iron: field, laboratory and modeling studies. Water Research 37, 1417e1425. Pillewan, P., Mukherjee, S., Roychowdhury, T., Das, S., Bansiwal, A., Rayalu, S., 2011. Removal of As(III) and As(V) from water by copper oxide incorporated mesoporous alumina. Journal of Hazardous Material 186, 367e375. Pontius, F.W., Brown, K.G., Chen, C.J., 1994. Health implication of arsenic in drinking water. Journal of American Water Work Association 86, 52e63. Rajan, D., Talat, M., Hasan, S.H., 2009. Biosorption of arsenic from aqueous solution using agriculture residue rice polish. Journal of Hazardous Materials 166, 1050e1059. Saeed, M.M., 2003. Adsorption profile nd thermodynamic paramerters of the preconcnentration of Eu (III) on 2thenoyltrifluoroacetone loaded plyurethane (PUR) foam. Journal of Radioanalysis and Nuclear Chemistry 256, 73. Sekar, M., Sakthi, V., Rengaraj, S., 2004. Kinetics and equilibrium adsorption study of lead(II) onto activated carbon prepared from coconut shell. Journal of Colloid Interface Science 279, 307e313. Shugi, K., Tony, S., Pant, K.K., 2004. Adsorption of As(III) from aqueous solution onto iron oxide impregnated activated alumina. Water Quality Research Journal of Canada 39 (3), 258e266. Singh, T.S., Pant, K.K., 2004. Equilibrium kinetics and thermodynamics studies for adsorption of As (III) on activated alumina. Separation and Purification Technology 36, 139e147. Smedley, P.L., Kinniburgh, D.G., 2002. A review of the source, behaviour and distribution of arsenic in natural waters. Applied Geochemistry 17, 517e568. Smith, A.H., Hopenhayn-Rich, C., Bates, M.N., Goeden, H.M., Piccoitto, I.H., Duggan, H.M., 1992. Cancer risk from arsenic in drinking water. Environmental Health Perspective 97, 259e267. Sun, X., Doner, H.E., 1998. Adsorption and oxidation of arsenite on goethite. Soil Science 163, 278e287. Tian, Ye., Wu, M., Liu, R., Wang, D., Lin, X., Liu, W., Ma, L., Li, Y., Huang, Y., 2011. Modified native cellulose fibersdA novel efficient adsorbent for both fluoride and arsenic. Journal of Hazardous Materials 185, 93e100. Who, 2004. Guideline for drinking water quality. In: Recommendation, third ed.. World Health Organization, Geneva. Zhang, Y.M., Yang, M., Huang, X., 2003. Arsenic (V) removal with a Ce (IV)-doped iron oxide adsorbent. Chemosphere 51, 945e952.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 8 2 e4 7 9 2

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Abiotic properties of landfill leachate controlling arsenic release from drinking water adsorbents Mengling Y. Stuckman, John J. Lenhart*, Harold W. Walker Department of Civil and Environmental Engineering and Geodetic Science, The Ohio State University, Columbus, OH 43210, United States

article info

abstract

Article history:

In this study, As leaching from five arsenic bearing solid residuals (ABSRs) comprised of the

Received 10 February 2011

iron hydroxide adsorbent Bayoxide E33 used in long-term operations was evaluated in

Received in revised form

leaching trials using California Waste Extraction Test (CalWET) and Toxicity Characteristic

22 April 2011

Leaching Protocol (TCLP) leachate solutions, a landfill leachate (LL), and synthetic leachate

Accepted 20 June 2011

(SL). The initial As loading of the media, which reflects the influence of source water

Available online 30 June 2011

chemistry and varying treatment conditions at the point of removal, strongly influenced the magnitude of As release. The chemical composition of the leachate also influenced As

Keywords:

release and demonstrated the relative importance of different release mechanisms,

Arsenic

namely media dissolution, pH-dependent sorption/desorption, and ion exchange. The

Drinking water treatment

CalWET solution, which partially dissolved the iron-based media, resulted in 100 times

Adsorbents

more As release than did the TCLP solution, which did not dissolve the media. The LL had

Landfill leachate

a higher pH than the TCLP solution, and even though its organic carbon content was lower

Release mechanisms

it tended to release more As. Tests with the SL were conducted to determine the influence of variations in leachate pH, phosphate, bicarbonate, sulfate, silicate, and natural organic matter (NOM). Release increased at high pH, in the presence of high concentrations of phosphate and bicarbonate, and in the presence of high NOM concentrations. For pH, this reflects the pH-dependence of sorption reactions, whereas for the anions and NOM, direct competition appeared important. Similar to the CalWET solution, excess NOM dissolved portions of the media thereby facilitating As release. In general, our results suggest that estimating As release into landfills will remain a challenge as it depends upon As loading, which reflects site-specific properties, and the composition of the leachate, which varies from landfill to landfill. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Due to its known toxic and carcinogenic effects, the presence of arsenic in drinking water is currently regulated in the United States at a level of 10 mg/L (USEPA, 2005). To meet this limit, single-use adsorbents are likely to be utilized to remove arsenic from impacted small water systems, which in the U.S. comprise 92% of the affected systems (USEPA, 2005). As

a result, an estimated 10,000 tons of arsenic bearing solid residuals (ABSRs) will be generated every year (Impellitteri and Scheckel, 2006). Evidence suggests most ABSRs will pass regulatory tests, such as the U.S. Environmental Protection Agency’s (USEPA’s) Toxicity Characteristic Leaching Protocol (TCLP), and thus will end up in municipal solid waste landfills. However, several studies suggest that regulatory leaching tests like the TCLP are inappropriate for estimating oxoanion

* Corresponding author. Tel.: þ1 614 688 8157; fax: þ1 614 292 3780. E-mail address: [email protected] (J.J. Lenhart). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.06.024

4783

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 8 2 e4 7 9 2

release and thus underestimate arsenic release (Hooper et al., 1998; Ghosh et al., 2004). The resulting As-laden leachate presents several problems in terms of treatment and management. For example, treatment of the leachate might have to be modified to ensure As removal such that discharge options remain unimpacted. In addition, land application of the As-laden sludge generated from leachate treatment may no longer be permitted resulting in an increase in the overall disposal costs (Williams et al., 2006). Lastly, under certain conditions the leachate itself could potentially result in the pollution of groundwater (Christensen et al., 2001). Compared to the extensive research detailing mechanisms responsible for arsenic treatment via adsorption to iron oxides and iron-based adsorptive media (e.g., Dixit and Hering (2003) and Ng et al. (2004)), significantly less research exists regarding the processes and mechanisms controlling arsenic release from these same materials (Ghosh et al., 2006). The factors likely to influence arsenic release under landfill conditions include pH and redox potential variations, anion exchange, interactions with natural organic matter (NOM) and biotransformation processes. Increasing the pH promotes As release by enhancing repulsion between the adsorbent surface and the increasingly negatively charged arsenic species (Ghosh et al., 2006). Reducing the redox potential promotes the formation of neutrally-charged As(III) species, which are more weakly bound to the adsorbent surface at neutral pH values (Pierce and Moore, 1982). Anions are known to compete with As during adsorption processes (Meng et al., 2000; Holm, 2002); however, little is known of the ability of these same anions to influence the stability of adsorbed As. Ghosh et al. (2006) evaluated the influence of phosphate, bicarbonate, sulfate, silicate, and NOM on the stability of adsorbed As(V) and determined that phosphate, due to its

similar structure, most effectively exchanges with sorbed arsenate. NOM also induces arsenic release because it directly competes with arsenic anions for surface sites (Grafe et al., 2001) and upon adsorption to the media facilitates its dissolution (Fendorf et al., 2010). NOM also complexes As, either directly (Ko et al., 2004) or through metal-bridging mechanisms (Lin et al., 2004). Finally, under certain conditions NOM acts as an electron shuttle and reduces As(V) to As(III) (Scott et al., 1999). Although not the focus of this study, microbiological processes are also very important determinants of As fate (deLemos et al., 2006; Tufano et al., 2008). Leachate composition varies among landfills reflecting differences in waste composition, landfilling technology and landfill age (Christensen et al., 2001). As Pohland and Kim (1999) describe, a complex sequence of biologically, chemically and physically mediated events occur during waste stabilization within a landfill characterized by five distinct phases. Arsenic release from ABSRs likely only occurs during two of these phases, the acid formation phase (Phase III) and the methane fermentation phase (Phase IV). During the acid formation phase, the pH in the landfill decreases from 7 to nearly 5 (See Table 1) and high concentrations of dissolved organic carbon (DOC) are produced (Harmsen, 1983). Under these conditions, metal species such as calcium, magnesium and iron are released into the leachate. Arsenic release during Phase III is suppressed by the decrease in pH and oxic conditions and simultaneously promoted by the large DOC concentrations and subsequent iron dissolution (deLemos et al., 2006). In the methane fermentation phase, the system becomes anoxic, alkalinity increases, and DOC decreases as fatty acids and other readily assimilated carbon sources are anaerobically transformed into methane (Christiansen et al., 1998). Arsenic release during Phase IV is likely promoted by

Table 1 e Typical ranges in landfill leachate composition (left columns) and the measured composition of landfill leachate, TCLP, CalWET and background synthetic leachate used in this work (Italics columns, maximum relative error was less than 5%). All units are in mg/L except pH. Overall Rangea

Key Phase Rangeb Phase III

pH Bicarbonate Phosphate Sulfate Silicate Manganese TOC Iron Ammonia-N Magnesium Calcium Sodium Potassium

4.5e9 34e15050 0.11e234 105e4900 5.1e51 0.03e1400 76e40000 3e5500 50e2200

70e7700b 50e3700b

Landfill Leachatef TCLP Leachate CalWET Leachate Background SL

Phase IV

4.5e7.5

7.5e9

70e1750

10e420

0.3e65 6000e60000e 20e2100

0.03e45 500e4500e 3e280

50e1150 10e2500

40e350 20e600

7.3 1650c 1.55 73 67.9 0.01 486 0.85 200 82.4 134 814 184

4.9 1410c NDd ND 0.02 0.02 2860 ND

5.0 3160c ND ND 0.13 ND 12400 ND

0.57 0.12 328 ND

0.52 0.11 849 ND

7.3 1500c 1.50 75 50 ND 400 1 200 100 100 800 200

a Ghosh et al. (2006). b Ehrig (1989). c Alkalinity as mg/L CaCO3. d Not detected with ICP-OES, Fe <0.001 mg/L, Mn <0.001 mg/L, S <0.1 mg/L, K <0.1 mg/L, P <0.001 mg/L. e Represents COD value. f The initial As concentration in landfill leachate was 0.013  0.002 mg/L. The initial As concentration in the other leachate solutions was below the detection limit.

4784

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 8 2 e4 7 9 2

the elevated pH and reducing conditions (Herbel and Fendorf, 2006) while also being confounded by the production of more reactive reduced-phase secondary iron minerals (Tufano and Fendorf, 2008). With few exceptions (Impellitteri and Scheckel, 2006; Nagar et al., 2010), the work evaluating arsenic release from ABSRs does not utilize media actually used to treat Ascontaining natural waters. This exposure to heterogeneous natural waters likely alters the physicochemical characteristics of the media as well as its potential to release or retain arsenic under landfill conditions to an extent that differs from media saturated with As under laboratory conditions. In this paper, we evaluated arsenic release from ABSRs under abiotic conditions using five samples collected from long-term treatment operations. These media were all comprised of the same material (Bayoxide E33), allowing us to determine how variations in source water chemistry and arsenic treatment processes impact arsenic loading and subsequent arsenic release under landfill conditions. The samples were subjected to different leaching trials to examine the effects of contact time, leachate test conditions and leachate type on arsenic release. Leachate solutions included those used in regulatory tests, a landfill leachate and a synthetic leachate. The synthetic leachate composition was tailored to investigate the influence of pH, competitive anions and NOM. Our results demonstrated that leachate composition and initial As loading in the media were the two main abiotic factors impacting As release from ABSRs. Leachate with high pH, elevated NOM and high concentrations of phosphate and bicarbonate enhanced As release the most.

2.

Materials and methods

Unless stated otherwise, all solutions were prepared from ACS-grade or trace-metal-grade reagents purchased from Fisher Scientific. Water for all samples and solutions was supplied from a deionized (DI) water system (Milli-Q Plus,

Millipore). All plastic or glassware was bathed in 10% nitric acid over night followed by a thorough rinse with DI water. All analyses were subject to quality assurance and quality control according to Standard Methods (Eaton et al., 1995). Details of these procedures are included in the Supplementary Material.

2.1.

ABSRs

An unused sample of Bayoxide E33 and five ABSRs that originated from four different states, New Mexico (NM), New Hampshire (NH), Arizona (AZ) and Texas (TX1 and TX2), were provided courtesy of Battelle Memorial Institute (Columbus, OH). These samples were collected from long-term arsenic removal operations conducted by the USEPA’s Arsenic Removal Technology Demonstration Program (USEPA, 2008a). Differences in shipping and sample preservation (USEPA, 2008a, b, c, 2009) resulted in the NM sample being received dry and the remaining samples received wet. TX1 and TX2 were from the downstream tank and upstream tank, respectively, of a two-tank system at the TX test site (USEPA, 2008a). As summarized in Table 2, the source water composition and treatment conditions differed at each site. Bayoxide E33 is an iron-based adsorbent containing 90.1% goethite. Its reported removal efficiency is 99% at pH 5.5e8.5 for source water with arsenic less than 100 ppb (USEPA, 2008b; AdEdge, 2009). Prechlorination was used at all sites, as E33 is less effective at removing As(III) (USEPA, 2008a, b, c, 2009).

2.2.

Leaching tests

All ABSRs were subjected to the following leaching tests: (1) kinetic trials with regulatory leachates and landfill leachate, and (2) batch tests with synthetic leachate (SL). All leaching tests were conducted in duplicate at 23  C. Before each leaching test, vessels containing wet ABSRs were tumbled to ensure sample homogeneity. Samples were extracted from the vessels and the remaining free liquid (0e8% v/v) was removed by vacuum filtration. Thus, all tested solids were

Table 2 e Comparison of sorbent characteristics and source water chemistry. Maximum relative error for elemental composition was 11.3%. Parameters

E33

NM

NH

AZ

TX1

TX2

Water content (%) BET surface area (m2/g) Bed volume of breakthrough at 10 ppba Contact time (months) Elemental composition of ABSR (mg/g)b

0 124

0 142 40600 12 2.15 2.52 0.32 7.6 0.022 0.009 158

67.1 139 17000 12 2.14 21.7 0.28 8.4 0.046 0.088 38

63.9 132 39180 30 7.53 1.69 0.15 7.2 0.050 N/D 13

43.4 NAc NA 6 2.25 3.34 0.10 7.8 0.029 0.058 2

67.7 NA

Source water quality (mg/L)a

As Mn S pH As Mn SO4

NDd 1.69 0.08 NA NA NA NA

3.07 9.91 0.17

a Source water quality was reported from USEPA Final Performance Evaluation Report at NM (USEPA, 2008a), NH (USEPA, 2009) and AZ (USEPA, 2008c) and Six-Month Evaluation Report at TX (USEPA, 2008b). b Elemental composition was obtained by ICP-OES. c Not available because of lack of measurement or lack of data in the reports. d Not detected, <0.1 mg/L.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 8 2 e4 7 9 2

weighed without free liquid. All post-leaching solutions were filtered through 0.45 mm glass fiber membranes and stored with minimal headspace at 4  C before analysis. Background concentrations of As and Fe present in the initial leachate were subtracted from the post-leaching As and Fe concentrations.

2.2.1. Kinetic tests with regulatory leachates and landfill leachate A series of four tests, lasting 72 h, was conducted to evaluate As release by the two regulatory leachate solutions and a landfill leachate under different physical conditions (e.g. liquid-to-solid (L:S) ratio, headspace, and agitation method). Each test used nine identical samples prepared and agitated in the same way, but individually sacrificed at designated time intervals. The first set of tests, designated TCLP, used extraction fluid #1 from the Toxicity Characteristic Leaching Procedure as described in USEPA Method 1311 (2007). This solution, with a pH of 4.93  0.05, was prepared in 1000 mL batches using 5.7 mL of glacial acetic acid and 64.3 mL of 1 mol/L NaOH. Due to limited sample size, the prescribed L:S ratio of 20:1 was met by using 1.0  0.01 g solids and 20.0 mL leachate. The samples were agitated on an end-over-end rotator in ambient air. The second set of tests, designated CalWET, used a 0.2 mol/L sodium citrate solution at pH 5.0  0.1, prepared following the California Waste Extraction Test procedures (California EPA, 2006). The desired L:S ratio of 10:1 was met by equilibrating 2.0  0.01 g solids with 20.0 mL of leachate. The sample containers were purged with nitrogen for 30 min, tightly sealed to minimize introduction of air and placed on a shaker table operating at 300 rpm for 48 h. The final two series of tests, designated as LLTCLP and LLCalWET, used landfill leachate (LL) collected from the Solid Waste Authority of Central Ohio’s Franklin County landfill. This leachate was stored in individual 1 L autoclaved acid-washed polypropylene bottles with zero headspace at a temperature of 4  C. When removing leachate from each bottle, the remainder was either discarded or the bottle was purged with nitrogen for 30 min prior to sealing in order to maintain anaerobic conditions. The physical leaching conditions for the LLTCLP tests (20:1 L:S ratio, ambient air, and rotator) were the same as those used in the TCLP trial. Those for LLCalWET were the same as the CalWET tests (10:1 L:S ratio, nitrogen purged, and shaker).

2.2.2.

Batch tests using synthetic leachate (SL)

Batch tests investigating chemical factors for As release (pH, TOC concentration and competitive anion concentration) were conducted with 2.0  0.01 g of the ABSRs and 20 mL of a synthetic leachate (SL). The background SL, with the composition noted in Table 1, was similar in composition to LL and provided a constant matrix with which to analyze the effects of changing pH, TOC concentration and competitive anion concentration. To investigate the effects of pH, the leaching solution consisted of the background SL of pH 5, 7 or 9. The influence of TOC and competitive anions was investigated using leaching solutions comprised of the background SL that excluded the analyte of interest. The specific analyte was then added at concentrations of 30e15000 mg/L for bicarbonate, 0.15e300 mg/L for phosphate, 75e300 mg/L for sulfate,

4785

2.5e100 mg/L for silicate, and 400e40000 mg/L for TOC, which spanned the range expected in landfill leachate (See Table 1). To adjust the SL pH HNO3 or NaOH were used. The mixtures were purged with nitrogen for 30 min and equilibrated for 48 h on a shaker table at 300 rpm before sampling. Humic acid, purchased from Aldrich, was used as the representative carbon source. Prior to use, the humic acid was purified following a modification of the procedure described by Nash and Choppin (1980). Briefly, 20 g of the humic acid was dissolved in 1 L 0.4 mol/L NaOH and stirred over night. The insoluble solid residue was removed by filtration (Whatman GF, pore size: 1.2 mm). The humic acid in the filtrate was precipitated by reducing the pH to 2 by adding 6 mol/L HCl. This precipitate was collected by centrifugation at 3780 g for 100 min and subsequently rinsed with DI water to remove excess acid. This dissolution and precipitation cycle was repeated 5 times until the final product was completely dissolved in deionized water at pH 7. This process reduced the concentration of metals in the humic acid to below their respective detection limits by inductively coupled plasma optical emission spectrometry (ICP-OES). The stock purified humic acid solution was concentrated by oven-evaporation at a temperature of 50  C for 48 h and measured to contain 60,000 mg/L TOC using a TOC analyzer (Shimadzu TOC-5000A).

2.3.

Analysis

2.3.1.

ABSR characterization

The water content of the ABSRs was measured by weighing samples before and after being air-dried and oven-dried at 110  C over night. No difference was found between the two drying methods. The BET surface area was measured using a Micromeritics high-speed surface area analyzer. The elemental composition of the ABSRs, after acid digestion following USEPA Method 3051 (1994), was determined using ICP-OES (Varian Vista AX). The elements analyzed included all TCLP regulated elements plus those that comprised the ABSRs (USEPA, 2008b). These characterizations were conducted in triplicate. X-ray powder diffraction (XRD, Philips PW1710 diffractometer) was used to identify crystalline phases in the air-dried solid samples.

2.3.2.

Leaching solution characterization

Bulk leachate properties, such as pH and alkalinity of the TCLP leachate, CalWET leachate and landfill leachate were measured in triplicate following methods in Eaton et al. (1995). All pre- and post-leaching solutions were subjected to elemental analysis by ICP-OES. Total As concentration measurements were measured in duplicate with a graphite furnace atomic absorption spectrometer (Varian Spectr AA 880Z). Total organic carbon (TOC) was measured for selected pre- and post-leaching solutions using the TOC analyzer.

3.

Results and discussion

3.1.

ABSR characterization

As received, NH, AZ and TX2 contained in excess of 60% water with visible free liquid. TX1 was moist with a water content of

4786

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 8 2 e4 7 9 2

43.4% but free liquid was not visible. NM was received dry. The BET surface area of the ABSRs (Table 2) was similar to the reported value of 142 m2/g (USEPA, 2008a) and compared well to reported values of 150 m2/g for high-surface area goethite (Cornell and Schwertmann, 1996) as well as other adsorbents such as activated alumina (Lin and Wu, 2001) and granular ferric hydroxide (Badruzzaman et al., 2004). XRD analyses confirmed the major mineral component of the media was goethite and that the crystalline structure was not altered through drinking water treatment (Fig. S1). The elemental composition of the media coincided with source water quality (Table 2). Except for manganese, the elemental compositions of TX1 and TX2 were similar, because they shared the same water source location. Manganese in the source water was highest at the NH site and second highest for the TX site, hence NH and TX2 contained higher Mn than the other samples. The high Mn loading likely reflects Mn (II) removal as MnOx coatings on the media in the presence of free chlorine (Knocke et al., 2010). Evidence in support of manganese precipitation was noted by the fact that the NH sample was darker in appearance than the other samples (Fig. S2). Sulfate adsorbs to iron oxides, particularly at low pH (Wilkie and Hering, 1996), and consequently the high sulfate source water (NM) had the highest media S content. The initial arsenic loading in the ABSRs on a dry-weight basis varied from 2.14 mg/g to 7.53 mg/g, reflecting differences in the source water composition and treatment conditions. Of all of the sites tested, the AZ site had the highest source water As concentration (0.05 mg/L). This, in combination with the high bed volume to breakthrough at 10 ppb (Table 2), resulted in the AZ sample having the highest arsenic loading. The impact of the source water As content on the initial As loading was apparent when evaluating the NM sample, which had a lower arsenic loading than AZ despite the similar bed volume to breakthrough. The pH of the source water appeared to influence the initial As loading (Table 2). Optimal arsenate removal occurs in the pH range 6e8 and thus the USEPA recommends pH adjustment (USEPA, 2009). This adjustment did not occur for the NH sample, because the carbon dioxide injection module failed (USEPA, 2009). The adsorption of As(V) to iron oxides decreases with increasing pH as the surface charge shifts from positive to negative thereby repelling the predominantly anionic arsenate species (Dixit and Hering, 2003). Thus, even though As in the source water at NH site was similar to that at the AZ site (Table 2), the As loading was significantly less. In addition to the lack of pH adjustment, the NH sample had a much lower bed volume to breakthrough (Table 2) and was subject to frequent backwashing (USEPA, 2009). The backwashing was used to address column pressure problems at the NH site and although the vendor recommends backwashing once per month, backwashing was conducted 28 times in 33 weeks (USEPA, 2009). An additional consequence of the frequent backwashing and maintenance at the NH site was a 46% loss of media during the run (USEPA, 2009). A similar media loss of 45% was reported at the NM site (USEPA, 2008b). According to the USEPA (USEPA, 2008b), the loss at the NM site reflected the media used had a small size and lacked integrity. These lost small-sized media grains could have carried a disproportionate load of adsorbed arsenic perhaps explaining the lower As loading for the NM

sample. Contact times were also quite variable and likely also influenced the As loading. This was most apparent in the low loading for the TX sample because the system was operated for just 6 months and did not achieve breakthrough (USEPA, 2008a). The presence of the MnOx coating on the NH sample also likely influenced loading, but due to the other operational issues at this site its impact was impossible to discern. Overall, multiple factors resulted in the ABSRs from the NH, AN and TX sites having lower As loadings than the AZ site.

3.2.

Leachate characterization

Details of the composition of the TCLP leachate, CalWET leachate, landfill leachate (LL) and synthetic leachate (SL) are summarized in Table 1. The high TOC and mildly acidic pH of the TCLP leachate and CalWET leachate are typical for landfill leachate created during the acid formation phase (Ghosh et al., 2004). The TCLP and CalWET leachate lack large concentrations of potentially competitive anions (e.g., sulfate) and di- or tri-valent cations (e.g. Mn2þ, Fe3þ, Mg2þ and Ca2þ) commonly present in acid formation phase leachate. Compared to literature reported values, the LL TOC was low and pH was high, suggesting the landfill sampled was in the methane fermentation phase (Christensen et al., 2001). The LL anion and cation concentrations were consistent with those for this phase and were used as target values for preparing the synthetic leachate. The initial As concentrations of all leachate solutions was below detection limits except for LL, which contained 0.013  0.002 mg/L As.

3.3. Kinetic tests with regulatory leachates and landfill leachate Arsenic release occurred quickly and appears to reach equilibrium by approximately 24 h (Figs. 1 and 2), similar to the time reported by Ghosh et al. (2006). Comparing the data for 48 h (Table 3), we note that AZ released more As than the other samples, regardless of the leaching conditions. Release for the other samples varied, but was always significantly less than that from the AZ sample (see also example data set in Fig. 1 and complete data in Fig. S3). This appears to closely correspond to the initial As loading (Table 2) since AZ had the highest As loading and it subsequently released approximately tenfold more As than the other samples. The negative values presented for Fe and As release in Table 3 indicate a decrease in the concentration from that initially present in the leachate solutions (see Table 1), perhaps suggesting resorption of the analyte of interest. Arsenic release under CalWET conditions was much higher than under the other conditions (Table 3). A similar result was observed for iron release as well. The main component of the CalWET solution was citrate, and at acidic pH values (<6.5) citrate not only sorbs strongly onto goethite and other iron oxides thereby potentially exchanging for sorbed arsenic, it also induces mineral dissolution (Shi et al., 2009a; Filius et al., 1997). At higher pH values (6.5e8.5), citrate sorption decreases and competition with arsenic and citrate-promoted dissolution cease to occur (Shi et al., 2009a, b). Therefore, the roughly two-order of magnitude higher release of As during the CalWET trials likely reflects the combined impacts of citrate

4787

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 8 2 e4 7 9 2

18

As released (mg/L)

15 12 NM NH AZ TX1 TX2

9 6 3 0

0

6

12

18

24

30

36

42

48

54

60

66

72

Time (h) Fig. 1 e Arsenic release as a function of time under CalWET conditions for all samples.

0.07 0.06

sorption and citrate-promoted iron dissolution. The initial As loading appears to limit release, however, as only one sample (AZ) released As above the current regulatory threshold in the United States of 5 mg/L. This was under CalWET conditions as release in the TCLP trials was low (See Table 3). This difference in As release between the two regulatory solutions was consistent with the literature (Hooper et al., 1998; Ghosh et al., 2004) and reflects the primary component of the TCLP solution (acetate) has little affinity for oxide surfaces at a pH of 5 and thus it neither dissolves iron oxides (Shi et al., 2009b) nor promotes arsenic release (Halim et al., 2004). The concentration of arsenic released was divided by the initial arsenic loading in order to determine the percent As release (Fig. 2) allowing the comparison of the results for LLTCLP and LLCalWET. From this, we noted that differences in release as a function of the physical conditions of the tests were minor as differences in agitation method or headspace atmosphere in the LLTCLP and LLCalWET trials did not result in appreciable differences in As release. This was particularly

0.07

a

0.06 0.05 % As released

% As released

0.05 0.04 0.03 0.02 0.01

0.04 0.03 0.02 0.01

0.00

0.00

-0.01

-0.01 0 6 12 18 24 30 36 42 48 54 60 66 72 Time (h)

0 6 12 18 24 30 36 42 48 54 60 66 72 Time (h) 0.07 0.06

0.07

c

0.06

0.04 0.03 0.02 0.01

0.04 0.03 0.02 0.01

0.00

0.00

-0.01

-0.01 0 6 12 18 24 30 36 42 48 54 60 66 72 Time (h)

0.06

d

0.05 % As released

% As released

0.05

0.07

b

0 6 12 18 24 30 36 42 48 54 60 66 72 Time (h)

e

% As released

0.05 0.04 0.03 0.02 0.01 0.00 -0.01

0 6 12 18 24 30 36 42 48 54 60 66 72 Time (h)

Fig. 2 e Percent As released as a function of time under LLTCLP (-), LLCalWET (C) and TCLP (B) conditions for (a) NM, (b) NH, (c), AZ, (d) TX1 and (e) TX2.

4788

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 8 2 e4 7 9 2

Samples

Elements released

NM

As released Fe released As released Fe released As released Fe released As released Fe released As released Fe released

NH AZ TX1 TX2

LLTCLP LLCalWET TCLP CalWET 0.001 0.32 0.048 0.31 0.104 0.56 0.010 0.37 0.000 0.84

0.003 0.59 0.054 0.50 0.315 0.43 0.031 0.41 0.036 0.03

0.021 0.13 0.001 0.13 0.085 0.21 0.049 0.23 0.008 0.10

0.816 570 2.58 720 13.9 250 1.88 300 2.86 560

notable when compared to the large differences observed with the different leachate solutions (Table 3). With the impact of physical differences in the trials being minimal, differences in the release of As in the LL trials from those with the regulatory leaching solutions (Table 3 and Fig. 2) mainly reflect differences in the pH and TOC in different leachate solutions (See Table 1). Likely, this reflects not only the significant difference in the TOC concentrations, but also differences in the properties of the molecules that comprised the TOC. Arsenic release, as a percent of that initially bound to the ABSR, in the presence of the landfill leachate under CalWET or TCLP conditions was similar in magnitude to that under TCLP condition (Fig. 2). There was no general trend with respect to the order of release. For some samples, AZ and NH, the landfill leachate released more As than the TCLP leachate, whereas for others, NM, TX1 and TX2, the reverse was true. This suggests our landfill leachate behaves similar to the acetate solution used in the TCLP even though its composition and pH was significantly different (see Table 1). It also suggests that unlike the citrate solution used in the CalWET, the organic acids associated with the TOC in this specific landfill leachate dissolves very little of the media. These results differ from those reported by Ghosh et al. (2004) and Hooper et al. (1998) who both reported a significant increase in As release when landfill leachate was applied in the leaching tests instead of the TCLP or CalWET solutions. Presumably, this reflects differences in both the composition of ABSRs and the aggressiveness of landfill leachate collected from different facilities. Hooper et al. (1998) report results with a landfill leachate with a TOC of 1220 mg/L that generated As release greater than the TCLP leachate under the same leaching conditions. Ghosh et al. (2004) used a landfill leachate with a low TOC (160 mg/L), but higher total dissolved solids (TDS) of 3600 mg/L, and also observed higher As release than the TCLP leachate under the same leaching conditions. In this work, the landfill leachate had a low TOC (119 mg/L) and a TDS of 3095 mg/L, which was similar to those reported by Ghosh et al. (2004). Unfortunately, the specific anion compositions of the leachate samples from Hooper et al. (1998) and Ghosh et al. (2004) were not reported so we cannot directly compare the relative importance of anions impacting arsenic release such as phosphate or carbonate. The observed differences could also simply reflect differences in the samples. Ghosh et al. (2004), for example, used activated alumina and granular ferric hydroxide, pre-

equilibrated in the laboratory with elevated levels of arsenic for relatively short periods of time, in their study. Hooper et al. (1998) used actual waste material, but this waste was taken from a variety of industrial sources (e.g., mine tailings and burn ash) as well as one sample identified as water precipitate. Specific details of the waste samples other than their composition were not provided making more detailed comparisons impossible (Hooper et al., 1998; Ghosh et al., 2004).

3.4.

Synthetic leachate batch tests

To better understand how specific leachate properties affect arsenic release, additional experiments were carried out with synthetic leachate of varying composition. Arsenic release upon exposure to SL exhibited pH-dependence, with increasing release observed as the pH increased from 5 to 9 (Fig. 3). Similar trends in the pH-dependence of As adsorption and desorption to mineral surfaces have been reported (Dixit and Hering, 2003; Grafe et al., 2001) and reflect decreasing affinity of arsenic for mineral surfaces at elevated pH. The change was particularly evident in increasing the pH from 7 to 9, where As release from the NM, NH and AZ samples increased by 330%, 596% and 1100%, respectively (Fig. 3). Arsenic release also exhibited a marked dependence on the TOC concentration of the synthetic leachate. Increasing the TOC from 400 mg/L to 40,000 mg/L promotes As release by approximately one order of magnitude (Fig. 4). At a TOC of 40,000 mg/L, As release from the AZ sample neared the regulatory limit of 5 mg/L; however, this amount still only corresponded to 0.6% of the initial loading. Other samples with lower initial loadings (e.g., NH and TX2) released higher percentages of total As, but lesser amounts in magnitude. This result coincided with stronger Fe and Mn release from NH and TX2 (Fig. 4b and c), suggesting TOC-induced mineral dissolution as well as anion exchange promoted As release. The decrease in iron concentration at higher TOC (Fig. 4b) reflected the precipitation of humic-iron solids during the pretreatment of samples with 5% nitric acid prior to the elemental analyses.

2.0

pH5

1.5 As released (mg/L)

Table 3 e Arsenic and iron released in 48 h leaching tests. All units are mg/L. Maximum relative error was 7.3%.

pH7 pH9 1.0

0.5

0.0 NM

NH

AZ

Samples Fig. 3 e Effect of pH on arsenic release from different samples with synthetic leachate over 48 h. Agitation was conducted with a shaker table and a nitrogen headspace was applied.

4789

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 8 2 e4 7 9 2

1.6

a

% As released

1.4 1.2 1.0 0.8 0.6 0.4

Table 4 e - TOC comparison in selected samples before and after leaching tests with TOC amended synthetic leachate. All units are mg/L. Maximum relative error was 1.82%. Designed TOC

400

2000

4000

20000

40000

SL before leaching AZ sample NH sample

331 112 262

1820 950 927

3010 1960 1800

13000 11900 10200

41800 39900 38900

0.2 0.0 100

1000 10000 TOC (mg/L)

100000

Fe released (mg/L)

250

b

200 150 100 50 0 100

1000

10000 TOC (mg/L)

300

c

250 Mn released (mg/L)

50000

200 150 100 50 0 100

1000 10000 TOC (mg/L)

100000

Fig. 4 e Effect of TOC on (a) arsenic, (b) iron, and (c) manganese release with synthetic leachate for samples NM (,), NH (C), AZ (:), TX1 (;) and TX2 (B) at pH 7 for 48 h shaking with nitrogen headspace. The Y indicates the concentration in the background synthetic leachate.

Humic acids are highly surface active and their affinity for a surface depends upon the type, density, and acidity of functional groups on the organic macromolecules, the properties of the surface, and the suspension pH (Vermeer and Koopal, 1999). We observed significant humic acid sorption that increased in magnitude as the TOC increased from 400 mg/L to 4000 mg/L (Table 4). Above this TOC level, however, further sorption was not observed as the available sites for the sorption of humic acid on the media were likely saturated. This plateau coincided with an arsenic release plateau (Fig. 4a). Arsenic release induced by the sorbed humic acid likely occurred via direct competition for sorption sites and/or through humic acid-promoted dissolution of the media. Support for competition as a release mechanism was drawn

from the literature evaluating As sorption, where it has been shown that As(V) competes with anionic humic acid functional groups for sorption sites, particularly at neutral pH values (Grafe et al., 2001, 2002; Wang and Mulligan, 2006). Extensive sorption of humic acid also alters the properties of the underlying surface and these changes could induce release. For example, adsorbed humic acid alters the charging behavior of the underlying mineral potentially making it less attractive to As (Ko et al., 2004; Wang and Mulligan, 2006). Humic substances are also known to induce mineral dissolution in a variety of settings (Ghosh et al., 2006; Fendorf et al., 2010, deLemos et al., 2006), evidence of which exists in our data (Fig. 4b and c). Like observed in the CalWET trials, such dissolution enhances arsenic release. The released cations (e.g., Fe3þ and Mn2þ) likely bound by humic acid, could enhance As solubility through the formation of metalbridging complexes (Lin et al., 2004). Bicarbonate at concentrations that ranged from 30 to 15000 mg/L caused a 10e25 fold increase in As release (Fig. 5a). This range of concentrations was much larger than those of the other anions, but was comparable to that of the TOC (Table 1). Although the most likely route of As release was through the direct exchange of bicarbonate with As on the surface of the media (Appelo et al., 2002; Anawar et al., 2004; Meng et al., 2000), bicarbonate also forms complexes with arsenic (Kim et al., 2000). These two processes are important at elevated bicarbonate concentrations and are thought to be somewhat responsible for arsenic mobilization in the bicarbonatedominated reducing aquifers of Bangladesh and other parts of the world (Fendorf et al., 2010; Anawar et al., 2004). Thus, their impact on As release in bicarbonate-rich landfill leachate could be important. Of the anions evaluated, phosphate altered As release behavior most significantly as evident in the nearly 1.8% release of arsenic from the AZ sample (10.2 mg/L As) in the presence of 300 mg/L phosphate (Fig. 5b). Except for the NM sample, where the increase was by a factor of 4, As release increased by 20e35 fold over just a 200 fold increase in phosphate concentration. This compares to a maximum increase in release of 25 times for bicarbonate over a much larger range (500 fold) in anion concentration. The amount of As released from the ABSRs increased in a roughly linear fashion with phosphate sorption (Fig. 6), suggesting direct competition for sorption sites on the media. Exchange was not on a molar equivalent, however, as it ranged from 2.6  104 to 0.056 mol As released per mol phosphate sorbed (Fig. 6). This suggests that only a fraction of the As was accessible to exchange with phosphate or that phosphate can sorb without desorbing As, particularly when the initial As loading was low. Zhao and Stanforth (2001) and Ghosh et al. (2006) also observed that

4790

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 8 2 e4 7 9 2

0.5

% As released

0.4

% As released

2.2 2.0 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 -0.2

a

0.3 0.2 0.1 0.0 -0.1 10

100

1000

10000

50000

b

0.1

c % As released

% As released

0.12 0.11 0.10 0.09 0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0.00 50

100

1000

1

10

100

Phosphate (mg/L)

Bicarbonate (mg/L) 0.12 0.11 0.10 0.09 0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0.00

5000

d

1

Sulfate (mg/L)

10

100

200

Silicate (mg/L)

Fig. 5 e Effect of competitive anions (a) bicarbonate, (b) phosphate, (c) sulfate, and (d) silicate on arsenic release for samples NM (,), NH (C), AZ (:), TX1 (;) and TX2 (B). Leaching tests were conducted at pH 7 with 10:1 liquid: solid ratio and nitrogen headspace on a shaker table for 48 h. The Y indicates the concentration in the background synthetic leachate.

the molar ratio of exchanged As during phosphate sorption was much less than one at low levels of As loading. The AZ sample released proportionately more As per mol phosphate adsorbed than the other samples, which was consistent with its much higher surface loading. Sulfate released significantly less arsenic than did phosphate or bicarbonate (Fig. 5c). At pH 7, we observed limited sorption of sulfate to the media (Fig. S5a), which was consistent with literature results indicating sulfate lacks affinity for the surface of iron oxides (Wilkie and Hering, 1996). The negligible influence of sulfate on arsenic adsorption/

220 200

2

NM y=0.44+0.26x, R =0.72 2 NH y=0.21+7.40x, R =0.88 2 AZ y=-2.4+56.8x, R =0.84 2 TX1 y=0.04+2.71x, R =0.85 2 TX2 y=-0.38+5.34x, R =0.82

As released (µ mol/L)

180 160 140 120 100 80

desorption was also consistent with results presented by Ghosh et al. (2006) and Meng et al. (2000). Unlike sulfate, silicate was sorbed to the surface; however, under the conditions studied it had little impact on As release (Fig. 5d). The sorption of silicate onto the media only occurred when silicate concentrations in SL were greater than 25 mg/L (Fig. S5b). Below this concentration the release of mediaassociated silica still exceeded that added. In competitive adsorption experiments, silicic acid sorption modifies the surface potential of the underlying oxide surface while also blocking reactive sites for binding (Waltham and Eick, 2002). The result of this competition is a reduction in the rate and extent of As adsorption (Waltham and Eick, 2002; Swedlund and Webster, 1999). Ghosh et al. (2006) indicate on a per mol basis that silicate-induced release of arsenic exceeds that for bicarbonate. Our results suggest that there was little difference in As release when evaluating the two anions on a per molar basis. However, since the typical landfill silicate concentration is much less than the bicarbonate concentration (see Table 1), the influence of silica likely pales in comparison to that of bicarbonate.

60 40

4.

20 0 -20 0

1

2

3

Phosphate sorbed (mmol/L) Fig. 6 e Relationship between As release versus phosphate adsorbed. The trend lines represent 5-point linear fit to the individual data sets.

Conclusions

From all the factors that influence As release from ABSRs used in long-term treatment operations, the initial As loading in the ABSRs was more significant than variation in the leachate composition. The initial As loading was impacted by the source water chemistry and arsenic treatment operation, which differed for the sites evaluated. The sample releasing the most As (AZ) had the highest initial As loading due to its

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 8 2 e4 7 9 2

being recovered from a site where treatment was conducted over a longer period of time with few interruptions as well as having a source water with high As concentration. Leachate composition also influenced As release. Solutions with elevated pH, presence of excess TOC available to dissolve the media or compete with sorbed As, large concentrations of phosphate or bicarbonate contribute most to As release. The effects of agitation methods, L:S ratio and headspace atmosphere were negligible under the conditions studied. In general, arsenic release potential dramatically increased for ABSRs heavily laden with arsenic. This could present a problem in the future if the regulatory limit decreases from 5 mg/L to 1 mg/L to correspond with the recently reduced MCL. It also presents a problem in terms of treatment design in that utilizing media to its capacity and thereby increasing its arsenic load could not only result in the media not passing regulatory tests it could lead to excessive release of the retained arsenic to the landfill leachate. The failure of the TCLP to simulate the methane fermentation phase of a landfill represents a concern as well because the leachate under this condition will have a high pH and high concentration of competitive anions that we (and others) have shown enhances release. This work does not address the potential biotransformation of iron or arsenic under landfill conditions and thus additional investigations, particularly emphasizing microbemediated iron dissolution and arsenic reduction, are merited.

Acknowledgments This research was supported by the Ohio Water Development Authority as well as the Environmental Science Graduate Program and Department of Civil and Environmental Engineering and Geodetic Science at The Ohio State University. We thank Dr. Abraham Chen, Ms. Lily Wang and Dr. Ryan Fimmen from Battelle Memorial Institute for providing spent media samples and Ms. Maria Segovia at Solid Waste Authority of Central Ohio for assistance in collecting the landfill leachate samples.

Appendix. Supplementary material Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.06.024.

references

AdEdge, 2009. Small Water System Solutions APU & Modular System Applications Adsorption Media-Arsenic Reduction. Adedge Technologies, Inc., Buford, GA. Anawar, H.M., Akai, J., Sakugawa, H., 2004. Mobilization of arsenic from subsurface sediments by effect of bicarbonate ions in groundwater. Chemosphere 54 (6), 753e762. Appelo, C.A.J., Van Der Weiden, M.J.J., Tournassat, C., Charlet, L., 2002. Surface complexation of ferrous iron and carbonate on ferrihydrite and the mobilization of arsenic. Environmental Science & Technology 36 (14), 3096e3103.

4791

Badruzzaman, M., Westerhoff, P., Knappe, D.R.U., 2004. Intraparticle diffusion and adsorption of arsenate onto granular ferric hydroxide (GFH). Water Research 38 (18), 4002e4012. California, E.P.A., 2006. Waste Extraction Test (WET) Procedures. Section 25141, Health and Safety Code. Office of Environmental Analysis and Regulations, California EPA, California. Christiansen, J.S., Engesgaard, P., Bjerg, P.L., 1998. A Physically and Chemically Heterogeneous Aquifer: Field Study and Reactive Transport Modelling. International Association of Hydrological Sciences, Wallingford, Germany. Christensen, T.H., Kjeldsen, P., Bjerg, P.L., Jensen, D.L., Christensen, J.B., Baun, A., Albrechtsen, H.-J., Heron, G., 2001. Biogeochemistry of landfill leachate plumes. Applied Geochemistry 16 (7e8), 659e718. Cornell, R.M., Schwertmann, U., 1996. The Iron Oxides: Structure, Properties, Reactions, Occurrences, and Uses. VCH Verlagsgesellshaft, Weinheim, Germany. deLemos, J.L., Bostick, B.C., Renshaw, C.E., Sturup, S., Feng, X., 2006. Landfill-stimulated iron reduction and arsenic release at the coakley superfund site (NH). Environmental Science & Technology 40 (1), 67e73. Dixit, S., Hering, J.G., 2003. Comparison of arsenic(V) and arsenic(III) sorption onto iron oxide minerals: implications for arsenic mobility. Environmental Science & Technology 37 (18), 4182e4189. Eaton, A.D., Clesceri, L.S., Greenberg, A.E., 1995. Standard Methods for the Examination of Water and Wastewater. American Public Health Association, American Water Works Association, Water Environment Federation, Hanover, Maryland. Ehrig, H.J., 1989. Water and element balances of landfills. In: Baccini, P. (Ed.), The Landfill. Springer, Berlin/Heidelberg, pp. 83e115. Fendorf, S., Michael, H.A., van Geen, A., 2010. Spatial and temporal variations of groundwater arsenic in South and Southeast Asia. Science 328 (5982), 1123e1127. Filius, J.D., Hiemstra, T., Van Riemsdijk, W.H., 1997. Adsorption of small weak organic acids on goethite: modeling of mechanisms. Journal of Colloid and Interface Science 195 (2), 368e380. Ghosh, A., Mukiibi, M., Ela, W., 2004. TCLP underestimates l arsenic from solid residuals under landfill conditions. Environmental Science & Technology 38 (17), 4677e4682. Ghosh, A., Sa´ez, A.E., Ela, W., 2006. Effect of pH, competitive anions and NOM on the leaching of arsenic from solid residuals. Science of the Total Environment 363 (1e3), 46e59. Grafe, M., Eick, M.J., Grossl, P.R., 2001. Adsorption of arsenate (V) and arsenite (III) on goethite in the presence and absence of dissolved organic carbon. Soil Science Society of America Journal 65 (6), 1680e1687. Grafe, M., Eick, M.J., Grossl, P.R., Saunders, A.M., 2002. Adsorption of arsenate and arsenite on ferrihydrite in the presence and absence of dissolved organic carbon. Journal of Environmental Quality 31 (4), 1115e1123. Halim, C.E., Scott, J.A., Natawardaya, H., Amal, R., Beydoun, D., Low, G., 2004. Comparison between acetic acid and landfill leachates for the leaching of Pb(II), Cd(II), As(V), and Cr(VI) from cementitious wastes. Environmental Science & Technology 38 (14), 3977e3983. Harmsen, J., 1983. Identification of organic-compounds in leachate from a waste tip. Water Research 17 (6), 699e705. Herbel, M., Fendorf, S., 2006. Biogeochemical processes controlling the speciation and transport of arsenic within iron coated sands. Chemical Geology 228 (1e3), 16e32.

4792

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 8 2 e4 7 9 2

3 Holm, T.R., 2002. Effects of CO2 3 /bicarbonate, Si, and PO4 on arsenic sorption to HFO. Journal of AWWA 94 (4), 174e181. Hooper, K., Isker, M., Sivia, G., Hussein, F., Hsu, J., DeGuzman, M., Odion, Z., Ilejay, Z., Sy, F., Petreas, M., Simmons, B., 1998. Toxicity characteristic leaching procedure fails to extract oxoanion-forming elements that are extracted by municipal solid waste leachates. Environmental Science & Technology 32 (23), 3825e3830. Impellitteri, C.A., Scheckel, K.G., 2006. The distribution, solidphase speciation, and desorption/dissolution of As in waste iron-based drinking water treatment residuals. Chemosphere 64 (6), 875e880. Kim, M.-J., Nriagu, J., Haack, S., 2000. Carbonate ions and arsenic dissolution by groundwater. Environmental Science and Technology 34 (15), 3094e3100. Ko, I., Kim, J.-Y., Kim, K.-W., 2004. Arsenic speciation and sorption kinetics in the As-hematite-humic acid system. Colloids and Surfaces A: Physicochemical and Engineering Aspects 234 (1e3), 43e50. Knocke, W.R., Zuravnsky, L., Little, J.C., Tobiason, J.E., 2010. Adsorptive contactors for removal of soluble manganese during drinking water treatment. Journal American Water Works Association 102 (8), 64e68. Lin, T.F., Wu, J.K., 2001. Adsorption of arsenite and arsenate within activated alumina grains: equilibrium and kinetics. Water Research 35 (8), 2049e2057. Lin, H.-T., Wang, M.C., Li, G.-C., 2004. Complexation of arsenate with humic substance in water extract of compost. Chemosphere 56 (11), 1105e1112. Meng, X., Bang, S., Korfiatis, G.P., 2000. Effects of silicate, sulfate, and carbonate on arsenic removal by ferric chloride. Water Research 34 (4), 1255e1261. Nagar, R., Sarkar, D., Makris, K.C., Datta, R., 2010. Effect of solution chemistry on arsenic sorption by Fe- and Al-based drinking-water treatment residuals. Chemosphere 78 (8), 1028e1035. Nash, K.L., Choppin, G.R., 1980. Interaction of humic and fulvic acids with Th(IV). Journal of Inorganic and Nuclear Chemistry 42 (7), 1045e1050. Ng, K.-S., Ujang, Z., Le-Clech, P., 2004. Arsenic removal technologies for drinking water treatment. Reviews in Environmental Science and Biotechnology 3 (1), 43e53. Pierce, M.L., Moore, C.B., 1982. Adsorption of arsenite and arsenate on amorphous iron hydroxide. Water Research 16 (7), 1247e1253. Pohland, F.G., Kim, J.C., 1999. In situ anaerobic treatment of leachate in landfill bioreactors. Water Science and Technology 40 (8), 203e210. Scott, D.T., McKnight, D.M., Blunt-Harris, E.L., Kolesar, S.E., Lovely, D. R., 1999. Quinone moieties act as electron acceptors in the reduction of humic substances by humics-reducing microorganisms. Environmental Science & Technology 32 (19), 2984e2989. Shi, R., Jia, Y., Wang, C., 2009a. Competitive and cooperative adsorption of arsenate and citrate on goethite. Journal of Environmental Sciences 21 (1), 106e112.

Shi, R., Jia, Y., Wang, C., Yao, S., 2009b. Mechanism of arsenate mobilization from goethite by aliphatic carboxylic acid. Journal of Hazardous Materials 163 (2e3), 1129e1133. Swedlund, P.J., Webster, J.G., 1999. Adsorption and polymerisation of silicic acid on ferrihydrite, and its effect on arsenic adsorption. Water Research 33 (16), 3413e3422. Tufano, K.J., Reyes, C., Saltikov, C.W., Fendorf, S., 2008. Reductive processes controlling arsenic retention: revealing the relative importance of iron and arsenic reduction. Environmental Science & Technology 42 (22), 8283e8289. Tufano, K.J., Fendorf, S., 2008. Confounding impacts of iron reduction on arsenic retention. Environmental Science & Technology 42 (13), 4777e4783. USEPA, 1994. Method 3051: Microwave assisted acid digestion of Sediment, Sludge, Soils and Oils. USEPA, 2005. Treatment Technologies for Arsenic Removal. National Risk Management Research Laboratory, Cincinnati, OH. USEPA., 2007. Method 1311: Toxicity characteristic leaching procedure. USEPA, 2008a. Arsenic removal from drinking water by adsorptive media U.S. EPA Demonstration Project at Oak Manor municipal Utility District at Alvin, TX Six-mMonth Evaluation Report. USEPA Contract 68-C-00e185 Task Order 0029. USEPA, 2008b. Arsenic Removal from Drinking Water by Adsorptive Media U.S. EPA Demonstration Project at Desert Sands MDWCA, NM Final Performance Evaluation Report. USEPA Contract 68-C-00e185 Task Order 0019. USEPA, 2008c. Arsenic Removal from Drinking Water by Adsorptive Media U.S. EPA Demonstration Project at Rimrock, AZ Final Performance Evaluation Report. USEPA Contract 68C-00e185 Task Order 0019. USEPA, 2009. Arsenic Removal from Drinking Water by Adsorptive Media U.S. EPA Demonstration Project at Rollinsford, NH Final Performance Evaluation Report. USEPA Contract 68-C-00e185 Task Order 0037. Vermeer, A.W.P., Koopal, L.K., 1999. Charge adjustments upon adsorption of a weak polyelectrolyte to a mineral oxide: The hematite-humic acid system. Journal of Colloid and Interface Science 212 (1), 176e185. Waltham, C.A., Eick, M.J., 2002. Kinetics of arsenic adsorption on goethite in the presence of sorbed silicic acid. Soil Science Society of America Journal 66 (3), 818e825. Wang, S.L., Mulligan, C.N., 2006. Effect of natural organic matter on arsenic release from soils and sediments into groundwater. Environmental Geochemistry and Health 28 (3), 197e214. Wilkie, J.A., Hering, J.G., 1996. Adsorption of arsenic onto hydrous ferric oxide: effects of adsorbate/adsorbent ratios and cooccurring solutes. Colloids and Surfaces A: Physicochemical and Engineering Aspects 107, 97e110. Williams, A.G.B., Scheckel, K.G., Tolaymat, T., Impellitteri, C.A., 2006. Mineralogy and characterization of arsenic, iron, and lead in a mine waste-derived fertilizer. Environmental Science & Technology 40 (16), 4874e4879. Zhao, H.S., Stanforth, R., 2001. Competitive adsorption of phosphate and arsenate on goethite. Environmental Science & Technology 35 (24), 4753e4757.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 9 3 e4 8 0 2

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

The nature of the carbon source rules the competition between PAO and denitrifiers in systems for simultaneous biological nitrogen and phosphorus removal Javier Guerrero 1, Albert Guisasola*, Juan A. Baeza 2 Departament d’Enginyeria Quı´mica, Escola d’Enginyeria, Universitat Auto`noma de Barcelona, 08193 Bellaterra (Barcelona), Spain

article info

abstract

Article history:

The presence of nitrate in the theoretical anaerobic reactor of a municipal WWTP aiming

Received 8 February 2011

at simultaneous C, N and P removal usually leads to Enhanced Biological Phosphorus

Received in revised form

Removal (EBPR) failure due to the competition between PAO and denitrifiers for organic

15 June 2011

substrate. This problem was studied in a continuous anaerobiceanoxiceaerobic (A2/O)

Accepted 17 June 2011

pilot plant (146 L) operating with good removal performance and a PAO-enriched sludge

Available online 30 June 2011

(72%). Nitrate presence in the initially anaerobic reactor was studied by switching the operation of the plant to an anoxiceaerobic configuration. When the influent COD

Keywords:

composition was a mixture of different carbon sources (acetic acid, propionic acid and

Carbon source

sucrose) the system was surprisingly able to maintain EBPR, even with internal recycle

EBPR

ratios up to ten times the influent flow rate and COD limiting conditions. However, the

Nitrate

utilisation of sucrose as sole carbon source resulted in a fast EBPR failure. Batch tests with

OHO

different nitrate concentrations (0e40 mg L
1) were performed in order to gain insight into

PAO

the competition for the carbon source in terms of P-release or denitrification rates and P-

VFA

release/C-uptake ratio. Surprisingly, no inhibitory or detrimental effect on EBPR performance due to nitrate was observed. A model based on ASM2d but considering two step nitrification and denitrification was developed and experimentally validated. Simulation studies showed that anaerobic VFA availability is critical to maintain EBPR activity. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Biological nutrient removal (BNR) is considered the most economical and sustainable technology to meet the increasingly stricter discharge requirements in wastewater treatment. Although many wastewater treatment plants (WWTP) have already adapted their operation to meet stringent nutrient discharge limits, many others do not satisfy these requirements due to failures in the BNR processes. WWTP configurations integrating Enhanced Biological Phosphorus

Removal (EBPR) and nitrification/denitrification for simultaneous C, N and P removal require an anaerobic reactor after the inlet and the presence of an aerobic reactor before the settling process to promote Polyphosphate Accumulating Organisms (PAO) growth. Nitrification in the aerobic reactor may result in nitrate recirculation to the anaerobic reactor through the external recycle. This availability of nitrate in the supposedly anaerobic reactor is one of the most reported causes of EBPR failure in full-scale WWTP. Some studies (Kuba et al., 1994; Patel and Nakhla, 2006) indicated that the presence

* Corresponding author. Tel.: þ34 93 581 1879; fax: þ34 93 581 2013. E-mail addresses: [email protected] (J. Guerrero), [email protected] (A. Guisasola), [email protected] (J.A. Baeza). 1 Tel.: þ34 93 581 4798; fax: þ34 93 581 2013. 2 Tel.: þ34 93 581 1587; fax: þ34 93 581 2013. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.06.019

4794

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 9 3 e4 8 0 2

of nitrate prevented anaerobic P-release and thus EBPR activity, which only occurred after nitrate depletion (<1 mg L
1). Two different hypotheses have been reported in the literature so far. On one hand, some studies (Van Niel et al., 1998; Saito et al., 2004) linked the detrimental effect of nitrate on EBPR to the presence of some denitrification intermediates (nitrite or nitric oxide) which would have an inhibitory effect on PAO. On the other hand, the EBPR failure could be due to an insufficient amount of organic carbon, COD, as substrate for nutrient removal. The presence of nitrate would trigger the activity of ordinary heterotrophic organisms (OHO) which would reduce nitrate using COD as electron donor and result in less COD available for PAO growth. Several external carbon sources have been studied to balance the abovementioned COD deficiency in wastewaters (Gerber et al., 1986; Jones et al., 1987; Winter, 1989; Appeldoorm et al., 1992; Isaacs et al., 1994; Hallin et al., 1996). Among those, acetic acid was suggested as the most effective carbon source for improving BNR. However, Cho and Molof (2004) reported that acetic acid was preferentially degraded by denitrifying bacteria over PAO, which were outcompeted for the carbon source. For this reason, the delicate balance between organic carbon, N and P levels has a major impact to enhance the P removal in BNR systems. In this sense, EBPR failures from urban wastewater with low or medium organic content have been reported (Tasli et al., 1999). Another factor to consider in the effectiveness of EBPR is the nature of the carbon source that plays the electron donor role. Randall et al. (1997) proved that the presence of volatile fatty acids (VFA) was imperative to obtain a high P-removal capacity. In addition, Pijuan et al. (2004) and Oehmen et al. (2006) also proved that propionic acid favoured PAO enrichment. Regarding the electron acceptor, a fraction of PAO called denitrifying PAO (DPAO), can uptake effectively phosphorus linked to denitrification under anoxic conditions using polyhydroxyalkanoates (PHA) previously accumulated under anaerobic conditions. Hence, DPAO can be useful for achieving simultaneous phosphorus and nitrogen removal from wastewaters with carbon shortage (Kerrn-Jespersen and Henze, 1993; Kuba et al., 1996). Some authors recently found that PAO can be divided into two types with different denitrifying capabilities (Carvalho et al., 2007; He et al., 2007; Flowers et al., 2009). One clade (named IA) was able to couple nitrate reduction with phosphorus uptake, but another (named IIA) could only use nitrite in addition to oxygen.

-

The overall objective of this work was to study the role played by the nature of the carbon source in the intricate competition between PAO and OHO for the organic substrate under anoxic conditions. An anoxic/aerobic modified Ludzack-Ettinger (MLE) continuous pilot plant (146 L) for simultaneous biological organic matter, N and P removal was operated with different internal recycle ratios to study the detrimental effect of nitrate presence in the anaerobic reactor. Different organic matter concentrations and compositions were also used at different steps to induce EBPR failure. Finally, a mathematical model to describe the behaviour of the system was developed and validated. Different scenarios were simulated to obtain a better understanding on the role of the carbon source on EBPR feasibility under anoxic and aerobic conditions.

2.

Material and methods

2.1.

Pilot-plant description

The pilot plant (146 L) consisted of three continuous stirred tank reactors and one settler (Fig. 1). The plant was initially operated with the classical anaerobiceanoxiceaerobic (A2/O) configuration for simultaneous C, N and P removal. The first reactor (R1, 28L) was anaerobic so that PAO were selected against other OHO. Nitrate entering to the second reactor (R2, 28L) with QRINT was denitrified by either OHO or DPAO. The third reactor (R3, 90 L) worked under aerated conditions and complete organic matter and P removal took place together with nitrification. The settler (50 L) produced an effluent stream and a biomass enriched stream which was returned to the system through the external recycle (QREXT). Mixed liquor was withdrawn daily from the three reactors in order to keep a desired sludge retention time (SRT), around 15  2 days. The influent (QIN) flow-rate was 140 L d
1 while QREXT was maintained around 125 L d
1 during all the experiments. This configuration was maintained during 4 months under steady-state conditions. However, during most of this work, the plant was working with MLE configuration (anoxiceaerobic) when QRINT was moved to R1 (Fig. 1), which did not behave anaerobic anymore. This configuration, typical of systems designed for only biological C and N removal, was chosen for gaining insight into the effect of nitrate entering to the anaerobic phase on the P removal. In fact,

-

Fig. 1 e A2/O and MLE pilot plant configurations.

4795

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 9 3 e4 8 0 2

in most of the cases, nitrate was completely depleted in R1 as COD resulting in an anoxiceanaerobiceaerobic configuration. The three reactors were monitored on-line with DO (Desin DO2-WW), pH (Desin EPH-M11) and temperature (Pt-100) probes connected to signal converters (Desin TM-3659). R3 was also equipped with ammonium and nitrate probes (Hach Lange scNH4D and scNO3D) and an on-line phosphate analyser (Hach Lange PHOSPHAX) with a sample filtration system (Hach Lange FILTRAX). On-line data were measured with a data acquisition card (Advantech PCI-1711) connected to a PC, with LabWindows CVI 2009 software for process monitoring and control. The data acquisition card had several analog and digital outputs for actuation over the pumps, stirrers and valves. The pH was controlled at 7.25  0.05 using an on-off controller with sodium carbonate (1 M) dosage. Dissolved oxygen (DO) in R3 was controlled at 1.75  0.25 mg DO L
1 with an on/off controller. Synthetic influent was prepared from a concentrated feed (Table 1) that was diluted (20:1) with tap water. The concentration of organic matter in the influent was different throughout the study (Table 2). The micronutrients composition was adapted from Smolders et al. (1994). Sludge from municipal WWTP from Manresa (Barcelona) was used to inoculate the pilot plant. PAO content was analysed by FISH quantification resulting in less than 2% of the total biomass.

2.2.

Batch experiments

Off-line batch experiments were performed in a magnetically stirred vessel (2 L). This system could be operated either under anaerobic/anoxic or aerated conditions by sparging nitrogen or oxygen gas, respectively. This gas was supplied through a microdiffuser which ensured good transfer from gas to liquid phase. The gas flow was controlled with a mass flowmeter (Bronckhorst HiTec 825) to ensure a constant flow.

Table 1 e Concentrated feed composition. Composition Macronutrients Sodium acetate (C2H3O2Na)a Sodium propionate (C3H5NaO2)a Sucrose (C12H22O11)a Ammonium chloride (NH4Cl)a Dipotassium phosphate (K2HPO4)a Potassium phosphate (KH2PO4)a Micronutrients Magnesium sulphate (MgSO4$7H2O) Calcium chloride (CaCl2$2H2O) Potassium chloride (KCl) Ferric chloride (FeCl3$6H2O) Potassium iodide (KI) Boric acid (H3BO3) Cobalt chloride (CoCl2$6H2O) Manganese chloride (MnCl2$4H2O) Zinc sulphate (ZnSO4$7H2O) Sodium molybdate (Na2MoO4$2H2O) Copper sulphate (CuSO4$5H2O) EDTA (C10H16N2O8)

g L
1 2.20/4.39 1.38/2.77 0.94/1.87 3.06 0.74 0.29 0.88 1.40 0.38 1.50b 0.18b 0.15b 0.15b 0.12b 0.12b 0.06b 0.03b 10.00b

a Main components: 4/8 g COD L
1 (37.5% acetate, 37.5% propionate and 25% sucrose), 0.8 g N L
1 y 0.2 g P L
1. b Trace solution: 1 mL introduced per L of influent.

Table 2 e Pilot plant conditions for each experimental step. Experiment

Step Step Step Step Step Step

0 I II III IV Va

Influent composition, mg L
1 (COD:N:P)

QRINT flowrate, L d
1

400:40:10 400:40:10 400:40:10 400:40:10 200:40:10 400:40:10

420 420 840 1400 420 420

QRINT/ Plant QIN configuration

3 3 6 10 3 3

A2/O MLE MLE MLE MLE MLE

a Sucrose was used as sole carbon source.

The pH (WTW Sentix 81) and DO (WTW CellOx 325) probes were connected to a multiparametric terminal (WTW INOLAB 3) which was in turn connected via RS232 to a PC allowing for data monitoring and storage. This software also manipulated a high precision microdispenser (Crison Multiburette 2S) for pH control with acid/base addition. More detailed information about this equipment can be found in Guisasola et al. (2007). The batch experiments aimed at studying the competition for influent COD between OHO and PAO under anoxic conditions and the inhibitory effect of nitrate on EBPR. The procedure followed was : i) the vessel was filled with biomass (around 2000 mg TSS L
1) from the pilot plant when it was operating under A2/O conditions, which was left under aerobic conditions for 12 h to ensure PHA depletion and thus, to obtain the maximum P-release after carbon source addition; ii) a pulse of acetic acid (350 mg COD L
1) at different nitrate concentration (0
1 or 40 mg NeNO
3 L ) was added under nitrogen-sparging conditions and the major components (COD, P, NeNO
2, NeNO
3 ) were monitored. After COD depletion, a second pulse of nitrate was added (20 mg L
1) in the experiments with nitrate to monitor the anoxic P-removal. Finally, the system was switched to aerobic conditions after 10 h to monitor aerobic P removal.

2.3.

Analytical methods

Ammonium was analysed by means of a continuous flow analyzer based on a potentiometric determination of ammonia (Baeza et al., 1999). Nitrate and nitrite were analysed with ionic chromatography (DIONEX ICS-2000). Phosphate was measured by a phosphate analyser (PHOSPHAX sc) based on the Vanadomolybdate yellow method, where a two-beam photometer with LEDS measured the phosphate specific yellow colour. Organic matter, mixed liquor total suspended solids (TSS) and mixed liquor volatile suspended solids (VSS) were analysed according to APHA (1995). Fluorescence in situ hybridisation (FISH) technique (Amman, 1995) coupled with confocal microscopy was used to quantify the biomass distribution as in Jubany et al. (2009). Hybridizations were performed using at the same time a Cy3labelled specific probe and Cy5-labelled EUBMIX for most bacteria (Dains et al., 1999). Ammonium oxidising bacteria (AOB) detection was performed with Nso190 specific probe while NIT3 probe was used for nitrite oxidising bacteria (NOB) hybridization (Jubany et al., 2009). PAOMIX probe was used to quantify most of PAO bacteria (Crocetti et al., 2000) and the

4796

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 9 3 e4 8 0 2

methodology proposed by Flowers et al. (2009) was used to hybridize PAO clade IA and clade IIA. GAOMIX, DF1MIX and DF2MIX probes were used to quantify glycogen accumulating organisms (GAO) according to the methodology described in Guisasola et al. (2009).

2.4.

Model description

The model used (see Supplementary information) is an extension of the well-known Activated Sludge Model 2d (ASM2d) proposed by IWA that describes the different processes occurring in a system for simultaneous biological organic matter and nutrient removal (Henze et al., 2000). The major extension was the inclusion of nitrite as a state variable. Nitrite is a key intermediate to understand the behaviour of the different PAO fractions since some of them can use either nitrate and/or nitrite as electron acceptor. Moreover, nitrite may have an inhibitory effect on EBPR as some authors reported (Van Niel et al., 1998; Saito et al., 2004). The anoxic P-uptake with nitrite as electron acceptor could be also considered according to the different denitrifying capabilities of DPAO biomass (He et al., 2007; Flowers et al., 2009). Hence, nitrification was modelled as a two-step process, including AOB and NOB. Denitrification was also described in two steps (nitrate to nitrite and nitrite to nitrogen gas) to understand the COD fate under anoxic conditions and the possible substrate competition with PAO. The extended model included 21 compounds, which were divided into soluble or particulate, and 28 processes. The process kinetics, stoichiometry and parameter values matrix can be found in the Supplementary information. The model was integrated with Matlab using the ode15s function, a variable order method recommended for stiff systems. The parameter estimation of the new processes considered was carried out by using NeldereMead Simplex search method ( fminsearch Matlab function). The settler was modelled using the model of Taka´cs et al. (1991). The starting point for each simulation was the steady state of the system under A2/O during 100 days in order to obtain the initial conditions of the plant (Step 0).

3.

Results

3.1.

Pilot-plant operation

Tables 2 and 3 describe, respectively, the different plant configurations and its corresponding steady-state effluent composition. Step 0 corresponds to the starting point of this study, the A2/O configuration (see Fig. 1). In the following steps (Step IeIII), the plant configuration was moved to a MLE

configuration for a better assessment of the detrimental effects of NOX (nitrate or denitrification intermediates) in the EBPR performance. The value of QRINT was gradually increased among these periods. Different FISH analyses were performed at the end of each step in order to quantify PAO sludge content. Fig. 2 shows the experimental profiles of the main compounds of the influent and the effluent during the steps 0 to III. High N and P removal (around 80% and 98% respectively, as Table 3 shows) was achieved with the conventional A2/O configuration despite a little amount of NOX entering
1 where nitrate was with the external recycle (8 mg NeNO
X L the predominant compound). Under anaerobic conditions (R1), 13.6% of the total COD inlet was consumed for denitrification of the recycled nitrate, while 53% was taken up by PAO resulting in P-release. These percentages were calculated assuming default growth yields: the recycle of 1 mg NeNO
3 to the anaerobic phase would consume 7.6 mg COD, whereas released would consume 2.5 mg COD as VFA 1 mg PePO
3 4 (Henze et al., 2000). FISH quantification performed during A2/O step clearly indicated the development of an enriched PAO sludge (72% of PAO) comparing with PAO content in the start-up of the plant (see Section 2.1.). Therefore, the existing PAO were able to coexist with denitrifying OHO in the A2/O configuration despite the nitrate inlet. This observation contrasted to the common textbook knowledge that a strict anaerobic phase is mandatory to achieve EBPR and that NOX presence in the anaerobic reactor can be detrimental to EBPR success (Simpkins and Mclaren, 1978; Van Niel et al., 1998; Henze et al., 2008). When the plant configuration was changed from A2/O to MLE (Step I, Table 2), N and P removal efficiencies slightly decreased to 74% and 97%, respectively (Table 3). The increase of QRINT during step II resulted in a decrease of the effluent NOX (more than 40%) as more NOX was brought to R1 to be denitrified. However, the subsequent increase of the internal recycle (Step III) did not result in a important decrease in the NOX (less than 15%) effluent content because the COD concentration became limiting under anoxic conditions. The measured effluent COD (Table 3) could be related to inert organic components. Surprisingly, the net-P removal efficiency was never affected during the abovementioned MLE operation (i.e. P removal was never lower than 85% despite the increase in the phosphorus concentration at the end of the step III) suggesting that COD was preferentially consumed for EBPR. This fact was corroborated with the FISH quantification obtained for steps IeIII (Table 3), where PAO population did not show an important shift during the experiments. Then, a COD-limited influent (200 mg COD L
1) was proposed in step IV to gain more insight into the substrate competition between PAO and

Table 3 e Steady-state effluent composition and PAO percentage in the sludge at the end of each experimental step. Experiment Step Step Step Step

0 I II III

COD (mg COD L
1) 22.4  23.2  22.8  18.4 

0.1 0.8 1.4 0.7


1 NeNHþ 4 (mg N L )

0.21  0.07 0.08  0.01 <0.05 0.32  0.15


1 NeNO
X (mg N L )

7.89 9.48 5.46 4.68

   

0.25 0.85 0.46 0.80


1 PePO
3 4 (mg P L )

0.21 0.26 0.23 1.09

   

0.01 0.04 0.05 0.67

PAO biomass (%) 72  77  68  71 

9 5 5 5

4797

Step 0

70

Step II

Step I MLE QRint:Inf 3:1

A2O

60

Step III MLE QRint:Inf 10:1

MLE QRint:Inf 6:1

50

600 500 400

40 300 30 200

20

COD (mg · L-1)

+ -3 -1 P-PO4 , N-NH4 and N- NOX (mg · L )

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 9 3 e4 8 0 2

100

10 0

0 0

5

10

15

20

25

30 35 Time (days)

40

45

50

55

60

65

Fig. 2 e Influent and effluent concentrations during the experimental steps 0eIII. (;) COD inlet, (7) COD outlet, (-) ammonium inlet, (,) ammonium outlet, (C) phosphorus inlet, (B) phosphorus outlet and (>) NOX outlet.

OHO (Fig. 3). Again, EBPR was not significantly affected by the COD decrease (P effluent concentration was always lower than 1.5 mg P L
1). On the contrary, the denitrification process was limited by the carbon source and NOX effluent concentration increased from 7 mg NeNOX L
1 to 15 mg NeNOX L
1 (nitrate was around 97% of the total NOX). This observation was again surprising and in clear disagreement with textbook knowledge that OHO should outcompete PAO in an anoxic COD-limited scenario. Finally, the PAO capacity to outcompete OHO related to the nature of the carbon source was studied in step V (Fig. 3). For this purpose, sucrose was used as sole carbon source. The Premoval capacity was progressively lost after only 4 days resulting in a high P effluent concentration. Most of the sucrose (70%) was oxidised by denitrifying OHO since NOX effluent concentration showed a slightly decreasing trend.

3.2.

nitrate in the anaerobic phase (Table 4) showed that nitrate presence did not significantly reduce the P-release rate. These results seem to be in disagreement with some literature (Chuang et al., 1996; Artan et al., 1998) which reported that nitrate presence reduces the P-release rate. An extended version of ASM2d, detailed in section 2.4 and in the Supplementary information, was used for a better understanding of the causes of the experimentally observed EBPR non-deterioration. The proposed model with default ASM2d parameters predicted EBPR failure when the QRINT increased, in contrast to the experimental data. Fig. 4 was used for model calibration purposes and the biomass diversity was fixed according to FISH quantification results: 72% of PAO, 4.0% of AOB, 6.7% of NOB and 0.5% of GAO (not included in the model). The rest (18%) were considered OHO. The parameters obtained after the model calibration process are presented in Table 5.

Batch experiments and model calibration 3.3.

20

Pilot plant simulations

The utilisation of the model with the calibrated parameters allowed a proper description of the experimental results when increasing the QRINT during Steps IeIII (Fig. 5). No EBPR failure was predicted in coincidence with the experimental results,

20

Step IV

Step V

16

16

12

12

8

8

4

4

P-PO

4

-3

+ -1 , N-NH and N- NO (mg · L ) 4 X

A batch experiment where a pulse of acetic acid and nitrate was simultaneously added to our sludge (see Section 2.2) is shown in Fig. 4. P-release occurrence was not avoided with high nitrate concentrations when acetic acid was the carbon source. Moreover, batch tests conducted with and without

0

0 0

1

2

3 4 Time (days)

5

6

7

0

1

2

3

4

5

6

7

Time (days)

Fig. 3 e Effluent composition and model predictions with a low COD inlet (Step IV) and results with sucrose as sole carbon source (Step V). (,) ammonium, (>) NOX and (C) phosphorus. Dotted line belongs to the model prediction for phosphorus, dashed line to ammonium and solid line to NOX.

4798

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 9 3 e4 8 0 2

120

60

200

40

P-PO 4

100

Table 5 e Parameters obtained in the model calibration of the batch experiment with acetate and nitrate.

-1 HAc (mg COD·L )

Anoxic conditions

80

300

-3

-1 and N- NO (mg · L ) X

100

Aerobic conditions

400 Nitrate pulse: 20 ppm

20 0

0 0

2

4

6 8 Time (h)

10

12

14

Fig. 4 e Experimental batch test (2L) for model calibration purposes (;) COD, (>) NOX and (C) phosphorus. Dotted line belongs to the phosphorus behaviour described by the model, solids line to NOX and dashed line to COD.

even when a low COD content influent was used (step IV, Fig. 3). Moreover, a simulation based study was performed in order to investigate the competition between PAO and OHO for the carbon source. Two different issues were analysed: i) the effect of COD influent concentration (with the same sucrose/VFA ratio shown in Table 1) on the competition of PAO and OHO ii) how the EBPR could be affected by the nature of the carbon source (i.e. different VFA and sucrose ratios were simulated) under COD limiting conditions (200 mg L
1). Each scenario was simulated during 7 days to mimic the experimental conditions of steps IV and V. As the model predicted (Fig. 6), the EBPR failure was lower with the calibrated model as the capacity of PAO to outcompete OHO for the carbon source was strengthened with the estimated parameters.

ASM2d value (20  C)

Calibrated value

Units

qPHA

3.00

5.00

qPP

1.50

0.60

qPAO hNO3 ; PAO hNO2 ; PAO a hNO3 ; OHO hNO2 ; OHO a

1.00 0.60 e 0.80 e

0.56 0.07 0.90 0.90 0.90

mg XPHA mg
1 X
1 PAO d mg XPP mg
1 X
1 PAO d d
1 e e e e

Parameters

a These parameters do not appear in ASM2d model (Henze et al., 2000).

rate, which was not detected when comparing experiments with and without nitrate. Another hypothesis to explain the lower P/C ratios obtained with nitrate presence and acetic acid as sole carbon source is the simultaneous oxidation of COD by OHO with nitrate as electron acceptor. Denitrification activity, i.e. nitrate reduction, was also observed without COD presence (Fig. 4) which could be attributed to the presence of DPAO. The denitrification capabilities of the different existing PAO is discussed in Oehmen et al. (2010) according to the classification proposed by Flowers et al. (2009), which specified that every PAO population (clade IA and clade IIA) could denitrify from nitrite, but only some of them (clade IA) could do this process from nitrate. These observations are in agreement with our results. We detected that part of our PAO could denitrify from nitrate since FISH quantification resulted in 22.3% of clade IA and 77.6% of clade IIA from the total PAO population quantified with PAOMIX probe.

4.2.

4. 4.1.

Feasibility of P removal in MLE system

Discussion Batch experiments

The simultaneous presence of nitrate and organic matter did not prevent net-P removal, i.e. P-release and subsequent Puptake. However, lower P/C ratios (Table 4) were obtained when nitrate was present. It could be argued that simultaneous P-release and anoxic-P-uptake was occurring, but this hypothesis should lead to a certain decrease in the P-release

Table 4 e Major transformations obtained in the batch studies with acetic acid. Initial NeNO
3 concentration 0 mg L
1 40 mg L
1
1
1 P-Release Rate (g PePO
3 d ) 4 g TSS
3 Nitrate uptake rate (g NeNO g TSS
1 d
1) COD uptake rate (g COD g TSS
1 d
1) P-release/C-uptake (P mmol/C mmol)

1.27 e 2.35 0.43

1.22 0.20 3.01 0.35

EBPR deterioration was not observed after switching its configuration to MLE despite the increased NOX inlet and thus, high P removal was obtained. These results seem to challenge the widely accepted idea that denitrifying OHO outcompete PAO when competing for the electron donor. The carbon source used in this work is a combination of propionic acid, acetic acid and sucrose. Different results can be found in the literature depending on the carbon source used. Some authors (Kuba et al., 1994; Patel and Nakhla, 2006) indicated that P-release should only occur when denitrifica
1 tion is completed (NeNO
3 < 1 mg L ) with most of the carbon sources except with acetic acid. With the latter, some reports have found that simultaneous nitrate reduction and P-release is observed (Gerber et al., 1986; Chuang et al., 1996; Artan et al., 1998). In any case, OHO is usually considered to have a preference for acetic acid in the denitrification process (Cho and Molof, 2004; Elefsiniotis et al., 2004). With respect to propionic acid, it is considered as an optimum carbon source for PAO growth. Some authors observed that propionate may be relatively easily sequestered and metabolized by PAOs compared to other microorganisms (Pijuan et al., 2004; Oehmen et al., 2006). Finally, sucrose is a complex carbon

4799

+ -3 P-PO4 , N-NH4 and N- NOX -1 (mg · L )

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 9 3 e4 8 0 2

60 50 40

+ -3 P-PO4 , N-NH4 and N- NOX -1 (mg · L )

Step III MLE QRint:Inf 10:1

30 20

Step 0 A2O

10 0 0 60

5 Step 0

50

A2O

40

10

15

20

25

30

Step I MLE QRint:Inf 3:1

35 40 Reactor 2 Step II MLE QRint:Inf 6:1

45

35 40 Reactor 3

45

50

55

60

65

70

Step III MLE QRint:Inf 10:1

30 20 10 0 0

+ -3 P-PO4 , N-NH4 and N- NOX -1 (mg · L )

Reactor 1 Step II MLE QRint:Inf 6:1

Step I MLE QRint:Inf 3:1

60

5 Step 0

50

A2O

40

10

15

20

25

30

Step I MLE QRint:Inf 3:1

50

55

60

65

70

Step III MLE QRint:Inf 10:1

Step II MLE QRint:Inf 6:1

30 20 10 0 0

5

10

15

20

25

30

35 40 Time (days)

45

50

55

60

65

70

Fig. 5 e Model validation. Pilot plant behaviour and model predictions when increasing QRINT (Steps 0eIII). (,) ammonium, (>) NOX and (C) phosphorus. Dotted line belongs to the phosphorus model prediction, dashed line to ammonium and solid line to NOX.

source of the synthetic wastewater. Under anaerobic conditions, fermentative bacteria (FB) should be consuming sucrose rather than PAO or OHO and produce VFA, while in presence of nitrate it could be an alternative carbon source for denitrifying bacteria. Our experimental results indicate that, under nitrate limited conditions (i.e. with less nitrate entering the anaerobic zone than the amount required to oxidise all the influent COD)

35

A

30

N-NOX-and P-PO4-3 (mg·L-1)

N-NOX-and P-PO4-3 (mg·L-1)

35

25 20 15 10 5 0 100

most of the COD is consumed linked to the EBPR process (Step IeIII). PAO population remained more or less constant during the different steps as FISH quantification shown (Table 3). This was observed under continuous nitrate entrance for more than 60 days. Batch experiments showed that P-release rate did not differ when nitrate was added (Table 4). Under CODlimited conditions (i.e. in step IV the amount of nitrate entering the anaerobic reactor was enough to oxidise all the

B

30 25 20 15 10 5 0

150

200

250

300 COD (mg·L-1)

350

400

0%

25%

50%

75%

100%

VFA content of the total organic carbon source

Fig. 6 e Simulation and experimental results to study the effect of influent COD content (A) and the nature of the carbon source (B) in the EBPR process. (>) belong to NOX and (B) to phosphorus. White symbols represent the default values of ASM2d, black symbols the calibrated model and grey symbols the experimental values.

4800

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 9 3 e4 8 0 2

influent COD) PAO were able to uptake carbon source rapidly and a high P removal was observed (Fig. 3). NOX content increased since OHO could not denitrify completely the influent NOX due to substrate limitations (i.e. the complete denitrification of NOX present in the anaerobic phase would have required a COD inlet of 390 mg L
1, 195% higher than the experimental COD influent concentration). These results do not agree with the preferential utilization of acetic acid by OHO reported by Cho and Molof (2004). This capacity of PAO to outcompete OHO for the carbon source is intrinsically linked to the nature of the organic matter and to the population distribution in the sludge (i.e. the high amount of PAO present in the system). Step V (Fig. 3) showed that the utilization of a complex carbon source (i.e. sucrose) could favour the OHO denitrification process against EBPR. Phosphorus effluent concentration increased from 0.57 mg L
1 to 6.63 mg L
1 while NOX outlet presented a decreasing trend. Under those conditions, PAO biomass needed the VFA produced by FB in sucrose fermentation. When, sucrose and NOX coexisted most of the sucrose was oxidised by denitrifying OHO (around 70% of the COD influent content). This fact could explain the EBPR failure when sucrose was the sole carbon source in contrast to the situation when VFA were added.

4.3.

Model validation and simulation based study

The experimental results obtained could not be reliably described using the extended AMS2d model with default parameters. For that reason, a calibration process was performed (Table 5). The higher maximum rate of PHA storage (qPHA) obtained after the calibration process could suggest that the PAO population could be more effectively consuming VFA than with the standard ASM2d values (Henze et al., 2000). This increase is necessary to describe that PAO would be more favoured than OHO in terms of VFA competition. Consequently, P-release capacity was almost not affected by NOX presence. The reduction factor for denitrification from nitrate for PAO (sNO3, PAO) indicates a low capacity to denitrify from nitrate to nitrite. However, it was enough to obtain the denitrification rates registered in the experimental data. This fact was in agreement with FISH quantification results when PAO clade IA and clade IIA were quantified (22.3% and 77.6% respectively of the total PAO bacteria). Finally, the results of the model calibration also indicated a lower P-uptake rate and lower PAO growth rate than the standard values of ASM2d, which did not affect the EBPR process in our case. Fig. 6 shows the results obtained in the model based study, where several scenarios with the default and calibrated parameters (Table 5) were simulated. The non-calibrated model predicted the total EBPR failure (Fig. 6A) when COD content was below 200 mg L
1 (i.e. P effluent concentration was the same as in the influent). In contrast, only a partial EBPR failure was observed with the calibrated model, even under strong COD limiting conditions (100 mg L
1). It should be noted that the denitrification process was more limited by the influent COD reduction with the calibrated model resulting in an effluent with high nitrate content. The PAO capacity to outcompete OHO for the carbon source would explain this fact. Steady-state experimental points were also included in Fig. 6. As can be observed, model

predictions properly described the experimental phosphorus values. However, less NOX effluent content was experimentally obtained in contrast to model predictions, suggesting that denitrification process was more efficient at practice. These divergent results could be explained if one takes into account that the model was calibrated when the pilot plant was operated under A2/O conditions. FISH quantification results showed an increase on PAO clade IA population after step III (from 22.3% to 35.6%) and thus, an increase in the denitrification capacity would explain the experimental results. Regarding the results where the nature of the carbon source was analysed (Fig. 6B), the predictions of both models were quite different. The P removal was almost negligible and denitrification was not deteriorated with the different influent tested with the default parameters, because it is assumed that the carbon source is preferentially used for denitrification rather than for EBPR. With the calibrated model the simulations results showed that P and N removal presented an inverse behaviour. EBPR capacity was highly affected by the VFA influent content and thus, P removal decreased as the VFA influent content also decreased. On the contrary, denitrification was favoured when the influent was enriched in a complex carbon source (sucrose). These results may be very helpful in view the designing new systems for simultaneous biological C, N and P removal when the influent wastewater composition is known. Also, it could be used to explain the reasons why some EBPR failures are observed with nitrate presence and how to solve them.

5.

Conclusions

This work shows how the nature of the carbon source rules the competition between PAO and denitrifying OHO in systems for simultaneous biological nitrogen and phosphorus removal. After switching the operation of an A2/O pilot plant to MLE operation, no inhibitory effect on EBPR due to NOX presence in the anaerobic phase was observed. When the carbon source presented a high VFA content, PAO could outcompete OHO even under anoxiceaerobic configuration. Heterotrophic denitrification activity was more affected than EBPR when working with low influent COD. EBPR failed when a more complex compound (sucrose) was used as a sole carbon source. In that case, NOX presence had an inhibitory effect in EBPR, not to inhibit the P-release process itself but to prevent the fermentation process for the VFA production. A model was developed and experimentally validated which explained the EBPR feasibility with nitrate presence under anoxiceaerobic conditions. The model calibration allowed a better understanding of the experimental results in terms of kinetic parameters of PAO and OHO. The simulation of different scenarios evidenced again that the PAO population could be more effective than OHO consuming VFA even under anoxic conditions. Although no EBPR failure was observed even under COD limiting conditions (100 mg L
1), VFA presence was demonstrated as the key point to trigger EBPR activity. The calibrated model was validated as a helpful tool to set the limits to avoid EBPR failures linked to nitrate presence.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 9 3 e4 8 0 2

Acknowledgments Javier Guerrero is grateful for the grant received from the Spanish government. This work was supported by the Spanish Ministerio de Ciencia y Tecnologı´a (CTM2010-20384). The authors are members of the GENOCOV research group (Grup de Recerca Consolidat de la Generalitat de Catalunya, 2009 SGR 815).

Appendix. Supplementary information Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.06.019.

references

Amman, R.I., 1995. In situ identification of microorganisms by whole cell hybridization with RNA-targeted nucleic acid probes. In: Ackkermans, A.D.L., van Elsas, J.D., de Bruijn, F.J. (Eds.), Molecular Microbial Ecology Manual. Kluwer Academic Publications, Dordrecht, Holland. APHA, 1995. Standard Methods for the Examination of Water and Wastewater, nineteenth ed.. American Publishers Health Association, Washington, DC, USA. Appeldoorm, K.J., Kortstee, J.J., Zehnder, A.J.B., 1992. Biological phosphate removal by activated sludge under defined conditions. Water Res. 26, 453e460. ¨ rgu¨r, N., Orhon, D., 1998. The fate of Artan, N., Tasli, R., O phosphate under anoxic conditions in biological nutrient removal activated sludge systems. Biotechnol. Lett. 20 (11), 1085e1090. Baeza, J., Gabriel, D., Lafuente, J., 1999. An expert supervisory system for pilot WWTP. Environ. Modell. Softw 14, 383e390. Carvalho, G., Lemos, P.C., Oehmen, A., Reis, A.M., 2007. Denitrifying phosphorus removal: linking the process performance with the microbial community structure. Water Res. 41, 4383e4396. Cho, E., Molof, A.H., 2004. Effect of sequentially combining methanol and acetic acid on the performance of biological nitrogen and phosphorus removal. J. Environ. Manage. 73, 183e187. Chuang, S.H., Ouyang, C.F., Wang, Y.B., 1996. Kinetic competition between phosphorus release and denitrification on sludge under anoxic conditions. Water Res. 30 (12), 2961e2968. Crocetti, G.R., Hugenholtz, P., Bond, P.L., Schuler, A., Keller, J., Jenkins, D., Blackall, L.L., 2000. Identification of polyphosphate-accumulating organisms and design of 16S rRNA-directed probes for their detection and quantitation. Appl. Environ. Microbiol. 66 (3), 1175e1182. Dains, H., Bru¨hl, A., Amann, R., Schleider, K.H., Wagner, M., 1999. The domain-specific probe EUB338 is insufficient for the detection of all bacteria: development and evaluation of a more comprehensive probe set. Syst. Appl. Microbiol. 22 (2), 434e444. Elefsiniotis, P., Wareham, D.G., Smith, M.O., 2004. Use of volatile fatty acids from an acid-phase digester for denitrification. J. Biotecnol. 114 (3), 289e297. Flowers, J., He, S., Yilmaz, S., Noguera, D., McMahon, K.D., 2009. Denitrification capabilities of two biological phosphorus removal sludges dominated by different ‘Candidatus Accumulibacter’ clades. Environ. Microbiol. Rep. 1 (6), 583e588.

4801

Gerber, A., Mostert, E.S., Winter, C.T., Villiers, R.H., 1986. The effect of acetate and other short-chain carbon compounds on the kinetics of biological nutrient removal. Water S.A. 12 (1), 7e12. Guisasola, A., Vargas, M., Marcelino, M., Lafuente, J., Casas, C., Baeza, J.A., 2007. On-line monitoring of the enhanced biological phosphorus removal processes using respirometry and titrimetry. Biochem. Eng. J. 35, 371e379. Guisasola, A., Qurie, M., Vargas, M.M., Casas, C., Baeza, J.A., 2009. Failure of an enriched nitrite-DPAO population to use nitrate as an electron acceptor. Process Biochem. 44, 689e695. Hallin, S., Rothman, M., Pell, M., 1996. Adaptation of denitrifying bacteria to acetate and methanol in activated sludge. Water Res. 30, 1445e1450. He, S., Gall, D.L., McMahon, K.D., 2007. “Candidatus Accumulibacter” population structure in enhanced biological phosphorus removal sludge as revealed by polyphosphate kinase genes. Appl. Environ. Microbiol. 73 (18), 5865e5874. Henze, M., Gujer, W., Mino, T., van Loosdrecht, M., 2000. Activated Sludge Models ASM1, ASM2, ASM2d, ASM3. IWA Publishing, London. Henze, M., van Loosdrecht, M., Ekama, G., Brdjanovic, D., 2008. Biological Wastewater Treatment. IWA Publishing, London, ISBN 1843391880, pp. 162e169. Isaacs, S.H., Henze, M., Søeberg, H., Kummel, M., 1994. External carbon source addition as a means to control an activated sludge nutrient removal process. Water Res. 28, 511e520. Jones, P.H., Tadwalkar, A.D., Hsu, C.L., 1987. Enhanced uptake of phosphorus by activated sludge: effect of substrate addition. Water Res. 21, 301e308. Jubany, I., Lafuente, J., Carrera, J., Baeza, J.A., 2009. Automated thresholding method (ATM) for biomass fraction determination using FISH and confocal microscopy. J. Chem. Technol. Biotechnol. 84 (8), 1140e1145. Kerrn-Jespersen, J.P., Henze, M., 1993. Biological phosphorus uptake under anoxic and aerobic conditions. Water Res. 27, 617e624. Kuba, T., Wachtmeister, A., van Loosdrecht, M.C.M., Heijnen, J.J., 1994. Effect of nitrate on phosphorus release in biological phosphorus removal systems. Water Sci. Technol. 30 (6), 263e269. Kuba, T., van Loosdrecht, M.C.M., Heijnen, J.J., 1996. Phosphorus and nitrogen removal with minimal COD requirement by integration of denitrifying dephosphatation and nitrification in a two-sludge system. Water Res. 30 (7), 1702e1710. Oehmen, A., Saunder, A.M., Vives, M.T., Yuan, Z., Keller, J., 2006. Competition between polyphosphate and glycogen accumulating organisms in enhanced biological phosphorus removal systems with acetate and propionate as carbon source. J. Biotechnol. 123, 22e32. Oehmen, A., Lopez-Vazquez, C.M., Carvalho, G., Reis, M.A.M., van Loosdrecht, M.C.M., 2010. Modelling the population dynamic and metabolic diversity of organisms relevant in anaerobic/ anoxic/aerobic enhanced biological phosphorus removal processes. Water Res. 44, 4473e4486. Patel, J., Nakhla, G., 2006. Interaction of denitrification and P removal in anoxic P removal systems. Desalination 201, 82e99. Pijuan, M., Saunders, A.M., Guisasola, A., Baeza, J.A., Casas, C., Blackall, L.L., 2004. Enhanced biological phosphorus removal in a sequencing batch reactor using propionate as the sole carbon source. Biotechnol. Bioeng. 85 (1), 56e67. Randall, A.A., Benefield, L.D., Hill, W.E., Nicol, J.P., Boman, G.K., Jing, S.R., 1997. The effect of volatile fatty acids on enhanced biological phosphorus removal and population structure in anaerobic/aerobic sequencing batch reactors. Water Sci. Technol. 35 (1), 153e160. Saito, T., Brdjanovic, D., van Loosdrecht, M.C.M., 2004. Effect of nitrite on phosphate uptake by phosphate accumulating organisms. Water Res. 38, 3760e3768.

4802

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 9 3 e4 8 0 2

Simpkins, M.J., Mclaren, A.R., 1978. Consistent biological phosphate and nitrate removal in an activated sludge plant. Progr. Water Tech. 10 (5e6), 433e442. Smolders, G.J.F., van der Meij, J., van Loosdrecht, M.C.M., Heijnen, J.J., 1994. Model of the anaerobic metabolism of the biological phosphorus removal process: stoichiometry and pH influence. Biotechnol. Bioeng. 42 (1994), 461e470. Taka´cs, I., Patry, G.G., Nolasco, D., 1991. A dynamic model of the clarification thickening process. Water Res. 25 (10), 1263e1271.

Tasli, R., Orhon, D., Artan, N., 1999. The effect of substrate composition on the nutrient removal potential of sequencing batch reactors. Water S.A. 25 (3), 337e344. Van Niel, E.W.J., Appeldoorn, K.J., Zehnder, A.J.B., Kortstee, G.J.J., 1998. Inhibition of anaerobic phosphate release by nitric oxide in activated sludge. Appl. Environ. Microbiol. 64, 2925e2930. Winter, C.T., 1989. The role of acetate in denitrification and biological phosphate removal in modified Bardenpho systems. Water Sci. Technol. 21, 375e385.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 0 3 e4 8 1 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Comparison of low-cost and engineered materials for phosphorus removal from organic-rich surface water Treavor H. Boyer*, Amar Persaud, Poulomi Banerjee, Pedro Palomino Department of Environmental Engineering Sciences, University of Florida, P.O. Box 116450, Gainesville, FL 32611-6450, USA

article info

abstract

Article history:

Excess phosphorus (P) in lakes and rivers remains a major water quality problem on

Received 26 February 2011

a global scale. As a result, new materials and innovative approaches to P remediation are

Received in revised form

required. Natural materials and waste byproduct materials from industrial processes have

11 June 2011

the potential to be effective materials for P removal from surface water. Advantages of

Accepted 19 June 2011

natural and waste byproduct materials include their low-cost, abundant supply, and

Available online 28 June 2011

minimal preparation, especially compared with engineered materials, such as ion exchange resins and polymeric adsorbents. As a result, natural and waste byproduct

Keywords:

materials are commonly referred to as low-cost materials. Despite the potential advan-

Alum sludge

tages of low-cost materials, there are critical gaps in knowledge that are preventing their

Eutrophication

effective use. In particular, there are limited data on the performance of low-cost materials

Ion exchange

in surface waters that have high concentrations of natural organic matter (NOM), and there

Natural organic matter

are no systematic studies that track the changes in water chemistry following treatment

Phosphate

with low-cost materials or compare their performance with engineered materials. Accordingly, the goal of this work was to evaluate and compare the effectiveness of lowcost and engineered materials for P removal from NOM-rich surface water. Seven lowcost materials and three engineered materials were evaluated using jar tests and minicolumn experiments. The test water was a surface water that had a total P concentration of 132e250 mg P/L and a total organic carbon concentration of 15e32 mg C/L. Alum sludge, a byproduct of drinking water treatment, and a hybrid anion exchange resin loaded with nanosize iron oxide were the best performing materials in terms of selective P removal in the presence of NOM and minimum undesirable secondary changes to the water chemistry. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Natural materials (e.g., clay and limestone) and waste byproduct materials from industrial processes (e.g., water treatment residuals, fly ash, and iron and steel slags) have been previously studied for remediation of a wide range of contaminants in water and sediment, including phosphate (PO4), arsenic, mercury, and other heavy metals (Bailey et al.,

1999; Hovsepyan and Bonzongo, 2009; Sibrell et al., 2009; Nagar et al., 2010). An advantage of natural and waste byproduct materials is that the materials are inexpensive and locally available. The reuse of waste byproduct materials for a beneficial purpose, such as water treatment, is also attractive as an alternative to disposal. In this work, natural and waste byproduct materials are defined as “low-cost” materials because the only cost associated with the materials is

* Corresponding author. Tel.: þ1 352 846 3351; fax: þ1 352 392 3076. E-mail address: [email protected] (T.H. Boyer). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.06.020

4804

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 0 3 e4 8 1 4

shipping or transportation, and there is an abundant supply of the materials. The focus of this work is the removal of phosphorus (P) from natural surface water by low-cost materials, and comparison of the treatment efficiency between low-cost and engineered materials. It is well-known that PO4 is the form of P that is most bioavailable and readily removed by physical-chemical treatment processes. Elevated PO4 concentrations in surface waters are a concern because of eutrophication, which remains a major water quality problem on a global scale (Dubrovsky et al., 2010; Harrison et al., 2010). PO4 inputs to surface waters include wastewater treatment plant effluent, agricultural drainage, and stormwater runoff. Furthermore, low PO4 concentrations in point and non-point sources can result in high mass loads of PO4 to receiving waters over time. Finally, the biogeochemistry of PO4 in surface waters is complicated by interactions with aquatic natural organic matter (NOM) and competing inorganic anions, such as chloride and sulfate. Common physical-chemical treatment unit processes used to remove PO4 include adsorption, coagulation, ion exchange, and precipitation. A concise review of PO4 removal by engineered materials (i.e., ion exchange resins and polymeric sorbents) and low-cost materials is provided to place this work in the context of previous work. A variety of novel materials have been synthesized and tested for PO4 removal including Zr(IV)-immobilized polymer chelating resin (Zhu and Jyo, 2005), Cu(II)- and Fe(III)-immobilized polymer anion exchange resin (Blaney et al., 2007; Pan et al., 2009; Sengupta and Pandit, 2011), and Cu(II)- and Fe(III)-immobilized porous silica (Chouyyok et al., 2010). All of the engineered materials function by a similar mechanism in which PO4 forms an inner-sphere complex with the immobilized metal. The engineered materials can achieve substantial PO4 removal with PO4 in treated water as low as 10 mg P/L depending on testing conditions (e.g., Chouyyok et al., 2010). Other common anions in surface water, such as chloride and sulfate, form outeresphere complexes with the functional groups of polymer anion exchange resins, which is a less selective mode of removal (Pan et al., 2009). Low-cost materials that have been previously studied for PO4 removal include acid mine drainage sludge (Sibrell et al., 2009), iron and steel slags (Baker et al., 1998; Zeng et al., 2004; Mortula et al., 2007; Xiong et al., 2008; Pratt et al., 2009), aluminum and iron drinking water treatment residuals (Makris et al., 2005; Yang et al., 2006; Mortula and Gagnon, 2007), fly ash (Ugurlu and Salman, 1998), limestone (Baker et al., 1998; Mortula et al., 2007), biologically produced iron oxides (Rentz et al., 2009), and palygorskite clay (Gan et al., 2009). The work by Sibrell et al. (2009) systematically evaluated acid mine drainage sludge as an adsorption media for P removal from agricultural wastewater. Acid mine drainage sludge was able to reduce the effluent P concentration to w50 mg P/L over 40,000 bed volumes of treatment. The lowcost materials previously studied in the literature remove PO4 by two main mechanisms: adsorption or precipitation. Adsorption typically takes place as an inner-sphere complex between PO4 and a metal oxide (e.g., Yang et al., 2006), whereas precipitation of sparingly soluble PO4-containing minerals is triggered by increasing pH and increasing

concentration of divalent cations, such as calcium and magnesium (e.g., Baker et al., 1998). It is important to recognize that high concentrations of NOM can limit PO4 removal by both engineered and low-cost materials by occupying adsorption sites, complexing metal ions in solution, and inhibiting precipitation (Guan et al., 2006; Weng et al., 2008; Qualls et al., 2009). Although low-cost materials have attractive characteristics, there are several gaps in knowledge pertaining to the use of low-cost materials for water treatment. Areas that lack data include testing low-cost materials in NOM-containing natural waters as opposed to NOM-free synthetic model waters; evaluating P concentrations that are representative of natural surface waters (i.e., mM); quantifying the secondary changes in water chemistry as a result of treatment using low-cost materials; and comparing the performance of low-cost materials with engineered materials such as ion exchange resins and polymeric adsorbents. The overall goal of this work is to evaluate the effectiveness of low-cost materials for the removal of P from NOM-rich surface water. The specific objectives of this work are: (1) to evaluate the properties of the low-cost materials; (2) to evaluate the changes in water chemistry following treatment with low-cost materials; (3) to compare the P removal efficiency of low-cost and engineered materials; (4) to discuss real-world applications of low-cost materials for P removal.

2.

Materials and methods

2.1.

Surface waters

Experiments were conducted using water from Lake Jesup and Sanford Avenue Canal. Lake Jesup is located in Seminole County, FL, USA, and is part of the middle basin of the St. Johns River. Lake Jesup has a surface area of w43 km2 and drains a watershed of w354 km2 (Gao, 2006). The watershed of Lake Jesup is highly urbanized, and as a result, the lake is impaired by high concentrations of total P (TP) and total nitrogen (Gao, 2006). Sanford Avenue Canal is w3.2 km in length, and drains both swamp and urban areas before entering Lake Jesup (Seminole County Water Atlas, 2011). Water samples from Lake Jesup and Sanford Avenue Canal were collected by the St. Johns River Water Management District (SJRWMD) in January, March, April, and June 2009. Samples were delivered to the Department of Environmental Engineering Sciences at the University of Florida and stored at 4  C until testing.

2.2.

Materials

2.2.1.

Low-cost materials

Seven low-cost materials were tested: drinking water treatment alum sludge (DWT-AS; Peace River Manasota Regional Water Supply Authority, Arcadia, FL, USA), drinking water treatment ferric sludge (DWT-FS; David L. Tippin Water Treatment Facility, Tampa, FL, USA), Class F fly ash (CFFA; Boral Materials Technologies, Tampa, FL, USA), granulated blast furnace slag (GBFS; Civil & Marine Inc., Cape Canaveral, FL, USA), basic oxygen furnace steel slag (BOFSS; Levy Enterprises, Valparaiso, IN, USA), recycled concrete (RC; Florida

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 0 3 e4 8 1 4

Concrete Recycling Inc., Gainesville, FL, USA), and limestone (LS; Florida Rock Industries Inc., Gainesville, FL, USA). Both DWT-AS and DWT-FS contained an unknown amount of powdered activated carbon, which was used prior to the coagulation process. No cleaning steps were performed on the low-cost materials. The only preparation of the materials was drying under ambient laboratory conditions and crushing. It is acknowledged that the materials could have different amounts of inherent moisture following the drying process. Nevertheless, drying under ambient conditions is how the materials are expected to be used in real-world applications. The low-cost materials, excluding CFFA, were crushed in a mortar and pestle and sieved through U.S. Standard Sieves 30 and 40 to give particle size range of 420e595 mm. CFFA, which was a powder, was used as received. All low-cost materials were weighed and dosed as dry material.

2.2.2.

Engineered materials

Three strong-base, macroporous, polymeric ion exchange resins were evaluated: phosphate selective resin (PSR; formerly PhosXnp by SolmeteX; now LayneRT by Layne Christensen), MIEX (Orica Watercare), and Dowex22 (Dow Chemical). PSR is a polymer resin impregnated with iron oxide nanoparticles that was developed specifically for PO4 removal (particle diameter ¼ 300e1200 mm; manufacturer data). PSR has been tested for PO4 and arsenic removal (Blaney et al., 2007; Sylvester et al., 2007; Pan et al., 2009; Sengupta and Pandit, 2011). MIEX is a polyacrylic resin that is used in water treatment for NOM removal (particle diameter ¼ 20e500 mm; manufacturer data). Dowex22 is a polystyrene resin used for industrial applications (particle diameter ¼ 300e1200 mm; manufacturer data). The ion exchange resins were used as received and stored in deionized (DI) water. The resins were measured using a graduated cylinder and dosed as volume of wet settled resin.

2.3.

Jar tests

Jar tests were conducted using a Phipps and Bird PB-700 jar tester at laboratory temperature. Raw water collected in January and March 2009 was used in the jar tests. Two liters of raw water was added to each jar. Pre-determined doses of low-cost materials (0.5, 1, 2, 4, and 8 g/L) and engineered materials (0.5, 1, 2, and 4 mL/L) were measured and added to each jar (all materials were not tested at all doses). The following constant mixing speeds were used: 100 rpm for DWT-AS, CFFA, and all engineered materials, and 200 rpm for DWT-FS, GBFS, BOFSS, RC, and LS because these materials were denser. All jar tests were conducted for 60 min, with samples collected after 5 min mixing (no settling), 30 min mixing (no settling), and 60 min mixing (with 30 min undisturbed settling). All material doses were tested in duplicate. All results are average values of duplicate samples with error bars showing one standard deviation, unless noted otherwise. Raw and treated water samples were measured for pH, turbidity, PO4, TP, UV absorbance at 254 nm (UV254), total organic carbon (TOC), chloride, and sulfate. TOC and UV254 were used as surrogates for NOM concentration and chemistry (Abbt-Braun et al., 2004). Samples for PO4, UV254, chloride, and sulfate were vacuum-filtered through 0.45 mm nylon

4805

membrane filters (Millipore), which were pre-rinsed with 500 mL of DI water and 20 mL of sample.

2.4.

Mini-column experiments

Mini-column experiments were conducted at laboratory temperature using a glass column with 10 mm inner diameter and 50 mm length (Omnifit). The column was filled with 1 mL of wet settled material: DWT-AS, DWT-FS, MIEX, or PSR. No additional changes to the particle size of the materials were made. Raw water collected in April and June 2009 was used in the mini-column experiments. Water was pumped up-flow through the column at 2 mL/min using a peristaltic pump (MasterFlex L/S ColeeParmer). The column set-up in nondimensional parameters was 1 bed volume (BV) ¼ 1 mL wet settled material and flow rate ¼ 2 BV/min. The flow rate was chosen to achieve the desired contact time and was not intended to approach equilibrium conditions. The column setup was cleaned with 120 BV of DI water before each experiment. Raw water was filtered through a 1.6 mm binder-free glass fiber filter (Whatman GF/A) to minimize clogging of the column. Samples were collected at pre-determined times over the duration of a column experiment. A sample volume of 120 mL was collected, so each sample represented a composite of the previous 120 BV of treatment. A column experiment was stopped when the effluent PO4 concentration divided by the influent PO4 concentration was w0.5. All samples were analyzed for the same parameters as the jar tests.

2.5.

Analytical methods

2.5.1.

Water samples

ACS reagent-grade purity chemicals and DI water were used to prepare all standard solutions. Analytical methods for pH, chloride, sulfate, UV254, and TOC are described elsewhere (Apell and Boyer, 2010). The total specific UV254 absorbance (TSUVA254) was defined as UV254/TOC. Turbidity was measured on a Lamotte 2020 turbidimeter that was calibrated daily with a 1 NTU standard. Samples for PO4 and TP were sent to the UF/ IFAS Analytical Services Laboratory in 20 mL scintillation vials, and were analyzed following Method 365.1 (U.S. EPA, 1993). Samples for TP analysis were acidified in the laboratory to pH <2 with sulfuric acid for preservation, after which autoclave digestion using ammonium persulfate and sulfuric acid was performed at the UF/IFAS Analytical Services Laboratory.

2.5.2.

Low-cost materials

Low-cost materials were digested following Method 3050B using heat plus nitric acid followed by hydrogen peroxide oxidation (U.S. EPA, 1996) to determine total recoverable metals. Depending on the low-cost material, 100e1000 mg of material was digested and the digestion procedure was performed with three samples of each material. The digested samples were measured on an ICP-AES (Thermo Jarrell Ash) according to Method 6010C (U.S. EPA, 2007). All data reported >0 mg/kg had a concentration in the digested sample above the minimum detection limit of the ICP-AES. Standard Reference Material 2709 (NIST, 2002) was digested and analyzed following the procedure described above to quantify the accuracy of the procedure. Standard Reference Material 2709

4806

12 55 5.1 32

a Water filtered through Whatman GF/A 1.6 mm filter prior to analysis. Not measured (nm).

0.72 nm 7.6

100

141

1.63

33 186 4.1 15 0.86 nm 7.8

215

250

0.594

35 176 4.5 18 0.69 3 7.5

138

201

0.818

24 36 23 108 173 117 3.6 3.0 4.9 16 17 22 0.15 0.05 0.65 4 19 2 7.5 7.5 7.3

Lake Jesup (January) Lake Jesup (March) Sanford Avenue Canal (January) Sanford Avenue Canal (March) Sanford Avenue Canal (Aprila) Sanford Avenue Canal (Junea)

17 10 88

111 208 132

0.579 0.522 1.07

TSUVA254 (L/mg C m) TOC (mg C/L) UV254 (1/cm) PO4/TP TP (mg P/L) PO4 (mg P/L) Turbidity (NTU) pH Location (month 2009)

Table 1 e Water quality for Lake Jesup and Sanford Avenue Canal water used in jar tests and mini-column experiments.

Cl (mg/L)

SO2 4 (mg/L)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 0 3 e4 8 1 4

is reported to have the following fraction of total recoverable metals (median % weight/weight (w/w)): 2.6% Al, 1.5% Ca, 3.0% Fe, 0.32% K, 1.4% Mg, and 0.068% Na (NIST, 2002). Laboratory digestion and analysis of Standard Reference Material 2709 gave the following total recoverable metals (% w/w): 2.4% Al, 1.7% Ca, 2.8% Fe, 0.42% K, 1.2% Mg, and 0.078% Na, which are in good agreement with the reported NIST (2002) data. Ag, As, Ba, Cd, Cr, and Pb are a subset of the elements that were measured in this work, and are also used to define the toxicity characteristic of solid waste (Electronic Code of Federal Regulations, 2011). A solid waste material containing greater than 100 mg Ag/kg solid, 100 mg As/kg solid, 2000 mg Ba/kg solid, 20 mg Cd/kg solid, 100 mg Cr/kg solid, or 100 mg Pb/kg solid has the potential to exceed the regulatory level of the toxicity characteristic, and this will depend on the results of the Toxicity Characteristic Leaching Procedure (TCLP; U.S. EPA, 1992). A solid waste material containing less than the amounts specified above will not exceed the toxicity characteristic because there is not sufficient metal in the solid waste to leach into water. X-ray diffraction (XRD) was performed as described in Harris and White (2008). Briefly, powder samples were mounted on plexiglass cavity sample holders. Each sample was scanned from 2 to 60 2q at 2 per min using CuKa radiation. The instrument was a computer-controlled X-ray powder diffractometer equipped with stepping motor and graphite crystal monochromator. Results were reported in graphical form and the dominant mineral phases were estimated. Point of zero charge (pHPZC) was determined by adding 0.5 g of material to 5 mL of N2-sparged DI water, shaken for 24 h in a sealed vial, and then measured for pH (Tennant and Mazyck, 2007). pHPZC values are the average of duplicate samples. BET surface area, pore diameter, and pore volume were measured on a Quantachrome NOVA 1200 Gas Sorption Analyzer (Byrne and Mazyck, 2009).

3.

Results

3.1.

Surface waters

The water quality for Lake Jesup and Sanford Avenue Canal are shown in Table 1. Both Lake Jesup and Sanford Avenue Canal have near neutral pH and similar concentrations of TP, TOC, chloride, and sulfate. However, the chemistry of P and NOM differ for the two surface waters. Lake Jesup has a low PO4/TP ratio (0.05e0.15) and an intermediate TSUVA254 (3.0e3.6 L/mg C m). The low fraction of PO4 suggests that P is incorporated into biomass. The biological activity in Lake Jesup is also reflected in its higher turbidity (maximum 19 NTU) relative to Sanford Avenue Canal (maximum 3 NTU). In contrast to Lake Jesup, Sanford Avenue Canal has a high PO4/ TP ratio (0.65e0.86) and a high TSUVA254 (4.1e5.1 L/mg C m). The high fraction of PO4 suggests that PO4 runoff from urban areas is entering the canal. In addition, the high TSUVA254 is indicative of terrestrial plants with a high aromatic carbon content being the precursor material for NOM (Weishaar et al., 2003). Sanford Avenue Canal water was used for all jar tests and mini-column experiments because it had a high fraction

4807

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 0 3 e4 8 1 4

Table 2 e Total recoverable metals composition of low-cost materials (mg/kg)a. Element

DWT-AS e 1.4E5 1.8E2 3.9E1 2.7E1 e 3.8E3 e e 1.2E2 4.2E2 4.5E3 7.4E2 9.9E2 1.6E2 8.7E1 9.7E3 1.3E1 1.3E1 e 3.6E1 4.8E2 4.4E1 1.6E2

Ag Al As B Ba Be Ca Cd Co Cr Cu Fe K Mg Mn Mo Na Ni Pb Sb Sn Sr V Zn

DWT-FS

GBFS e 5.7E4 1.4E0 7.6E1 3.2E2 6.5E0 1.8E5 e 3.3E0 2.2E1 4.2E0 2.2E3 3.4E3 2.1E4 1.8E3 e 1.7E3 1.4E0 8.3E-1c e e 2.8E2 2.5E1 3.9E1

4.3E-1 2.5E3 1.8E1 5.1E0 8.7E1 e 2.6E4 6.5E0 1.9E1 9.0E0 8.4E0 1.5E5 2.9E2 8.6E2 2.4E2 1.1E2 e 1.3E1 1.2E1 7.6E0 4.5E0 9.2E1 4.8E2 6.3E2

BOFSS b

1.3E0 1.4E4 4.2E-1b 3.1E1 5.9E1 e 1.0E5 4.3E0 3.1E0 6.4E2 2.8E1 1.2E5 5.8E2 2.7E4 6.7E3 1.0E1 e 6.4E0 7.6E0 2.5E0 9.0E-1 1.3E2 7.6E2 1.7E2

RC

LS

e 4.3E3 5.5E0 2.4E1 3.2E1 3.8E-1 1.4E5 2.0E-1 1.9E0 1.4E1 1.5E1 3.9E3 4.8E2 1.7E4 6.5E1 2.8E0 2.0E3 7.2E0 6.1E0 1.2E0b 2.9E0 2.1E2 1.5E1 7.0E1

e 7.2E2 2.7E0 1.4E1 1.6E0 e 1.5E5 e e 1.0E1 5.3E0 8.8E2 3.1E2 8.0E4 4.6E1 1.7E0c 2.6E3 3.7E0 1.9E0 e 1.3E0 1.6E2 9.7E0 6.6E2

a Mean value of triplicate samples unless noted otherwise. b Value of single sample. c Mean value of duplicate samples.

of PO4, which is the fraction of P that is readily removed by adsorption and precipitation as reviewed in Section 1.

3.2.

Low-cost materials

3.2.1.

Material properties

The elemental composition of the low-cost materials based on total recoverable metals is shown in Table 2. XRD crystal structure, pHpzc, and physical structure of the low-cost materials are shown in Table 3. DWT-AS. Al (14% w/w) and Na (1.0% w/w) were the most abundant elements, with all other elements < 0.5% w/w. DWT-AS contained 180 mg As/kg solid and 120 mg Cr/kg solid, which could theoretically exceed the regulatory level for As and Cr for the toxicity characteristic depending on the results of the TCLP. DWT-AS cannot exceed the regulatory level for Ag, Ba, Cd, or Pb for the toxicity characteristic based on its elemental composition. Quartz was the only crystal structure

identified by XRD. DWT-AS was characterized by an acidic surface (pHpzc < 7) and a porous structure (BET surface area > 200 m2/g). DWT-FS. Fe (15% w/w) and Ca (2.6% w/w) were the most abundant elements, with all other elements < 0.3% w/w. DWTFS cannot exceed the regulatory level for Ag, As, Ba, Cd, Cr, or Pb for the toxicity characteristic based on its elemental composition. DWT-FS was non-crystalline based on XRD analysis. DWT-FS was characterized by an acidic surface (pHpzc < 7) and a nonporous structure (BET surface area < 10 m2/g). GBFS. Ca (18% w/w), Al (5.7% w/w), and Mg (2.1% w/w) were the most abundant elements, with all other elements < 0.4% w/w. GBFS cannot exceed the regulatory level for Ag, As, Ba, Cd, Cr, or Pb for the toxicity characteristic based on its elemental composition. Calcite was the only crystal structure identified by XRD. GBFS was characterized by a basic surface (pHpzc > 9) and a nonporous structure (BET surface area < 10 m2/g).

Table 3 e Additional properties of low-cost materials. Material

XRD crystal structure

pHPZC

BET surface area (m2/g)

˚) Pore diameter (A

Pore volume (cm3/g)

DWT-AS DWT-FS GBFS BOFSS RC LS CFFA

Q NS C Ag, C, L C, D, Q C, D Q

6.2 6.9 9.3 12.5 12.0 9.5 11.6

227 6.7 2 nm 3.5 nm 2.7

76.6 69.1 279 nm 161 nm 69.3

0.44 0.01 0.014 nm 0.014 nm 0.004

Ag ¼ aragonite, C ¼ calcite, D ¼ dolomite, L ¼ larnite, NS ¼ nonecrystalline Q ¼ quartz, nm ¼ not measured.

4808

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 0 3 e4 8 1 4

GBFS BOFSS DWT-AS

LS DWT-FS RC

PO4 , C/C0

0.8 0.6 0.4 0.2 0

b

0

10

20

0

10

20

0

10

20

30 40 time (min)

50

60

10.5 10 9.5 9 8.5 8 7.5 7 6.5

c

Jar tests

The change in PO4, pH, and UV254 as a function of mixing time is shown in Fig. 1. All data are for jar tests using Sanford Avenue Canal water and the low-cost materials at 4 g/L. Fig. 1a shows the data as normalized PO4 concentration to allow for comparison of removal efficiency at different initial PO4 concentrations. CFFA added PO4 to the treated water with a final PO4 concentration of 218 mg P/L (results not shown). GBFS and LS showed <20% PO4 removal at 5 min and no additional removal at 60 min. DWT-FS and BOFSS showed w50% PO4 removal at 60 min. The slope of the PO4 curves for DWT-FS and BOFSS suggest that greater PO4 removal is possible at longer contact times. DWT-AS and RC showed the highest level of treatment achieving 68% and 81% PO4 removal, respectively. The corresponding final PO4 concentrations were 28 mg P/L (DWT-AS) and 27 mg P/L (RC). DWT-AS reached its maximum PO4 removal in 5 min, while RC showed increasing PO4 removal with increasing time similar to BOFSS. Fig. 1b shows the change in pH with time. Replicate data for pH are shown to avoid the confusion of taking the arithmetic mean of a log-transformed parameter. pH decreased by approximately 0.5 pH units for DWT-FS and DWT-AS, while pH increased by less than 0.5 pH units for GBFS and LS. pH increased by approximately 2 pH units following treatment with CFFA (results not shown), BOFSS, and RC. A majority of the pH increase by CFFA and RC took place within 5 min, while BOFSS showed a trend of increasing pH with time. Fig. 1c shows the change in UV254 absorbance with time. DWT-FS was the only material to increase the UV254

1.2 1

30 time (min)

40

50

60

30

40

50

60

1.6 1.4 1.2

UV254 (1/cm)

3.2.2.

a

pH

BOFSS. Fe (12% w/w), Ca (10% w/w), Mg (2.7% w/w), Al (1.4% w/w), and Mn (0.70% w/w) were the most abundant elements, with all other elements < 0.10% w/w. BOFSS contained 640 mg Cr/kg solid, which could theoretically exceed the regulatory level for Cr for the toxicity characteristic depending on the results of the TCLP. BOFSS cannot exceed the regulatory level for Ag, As, Ba, Cd, or Pb for the toxicity characteristic based on its elemental composition. Aragonite, calcite, and larnite were the mineral phases identified by XRD. BOFSS was characterized by a basic surface (pHpzc > 12). The pore structure of BOFSS was not measured. RC. Ca (14% w/w) and Mg (1.6% w/w) were the most abundant elements, with all other elements < 0.5% w/w. RC cannot exceed the regulatory level for Ag, As, Ba, Cd, Cr, or Pb for the toxicity characteristic based on its elemental composition. Calcite, dolomite, and quartz were the mineral phases identified by XRD. RC was characterized by a basic surface (pHpzc w12) and a nonporous structure (BET surface area < 10 m2/g). LS. Ca (15% w/w) and Mg (8.0% w/w) were the most abundant elements, with all other elements < 0.3% w/w. LS cannot exceed the regulatory level for Ag, As, Ba, Cd, Cr, or Pb for the toxicity characteristic based on its elemental composition. Calcite and dolomite were the mineral phases identified by XRD. LS was characterized by a basic surface (pHpzc > 9). The pore structure of LS was not measured. CFFA. The elemental composition of CFFA was not determined. Quartz was the only mineral phase identified by XRD. CFFA was characterized by a basic surface (pHpzc > 11) and a nonporous structure (BET surface area < 10 m2/g).

1 0.8 0.6 0.4 0.2 0 time (min)

Fig. 1 e Time-varying change in (a) PO4, (b) pH, and (c) UV254 for GBFS (triangle), LS (plus), BOFSS (diamond), DWT-FS (square), DWT-AS (circle), and RC (ex) during jar tests using 4 g/L of low-cost material and Sanford Avenue Canal water collected January 2009 (GBFS, LS, DWT-FS, DWT-AS) and March 2009 (BOFSS, RC). PO4 and UV254 are mean values of duplicate samples with error bars showing one standard deviation. pH is data for duplicates.

absorbance of the treated water. The increase in UV254 absorbance could be due to release of NOM or iron or both (Weishaar et al., 2003). GBFS and LS did not change the UV254 absorbance of the treated water. CFFA (results not shown), BOFSS, and RC showed low removal of UV254-absorbing substances (<15% at 60 min), with a majority of the removal achieved within the first 5 min of treatment. DWT-AS was

4809

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 0 3 e4 8 1 4

2.4 2.2 2 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0

PO4 TP TP (theo.) UV254 TOC

from the jar tests was analyzed for TP, so any P-containing low-cost materials or precipitates would be digested and measured during TP analysis. Attrition or precipitation is also supported by the increase in turbidity of the treated waters for all low-cost materials (final turbidity (NTU) at 60 min: GBFS ¼ 10, LS ¼ 29, DWT-FS ¼ 28, BOFSS ¼ 17, DWT-AS ¼ 13, RC ¼ 36) relative to Sanford Avenue Canal water (2e3 NTU). GBFS, LS, BOFSS, DWT-AS, and RC showed preferential removal of PO4 relative to TOC (see Fig. 2). RC showed the greatest difference in removal with 81% PO4 removal and <5% TOC removal. The selective removal of PO4 by the low-cost materials is an important result considering the TOC concentration was 18e22 mg C/L (w1e2 mM), whereas the TP concentration was 132e201 mg P/L (w4e6 mM). DWT-FS increased the TOC concentration of the treated water similar to the UV254 results. There was no change in the chloride concentration of the treated water for all low-cost materials (results not shown). However, DWT-AS and DWT-FS increased the sulfate concentration of the treated water by a factor of >2 (results not shown). This is because both DWT-AS and DWT-FS were sulfate-based metal salts used for coagulation.

3.2.3.

Mini-column experiments

DWT-AS and DWT-FS were tested in mini-column experiments using Sanford Avenue Canal water collected in April and June 2009. The results of a mini-column experiment using DWT-AS and Sanford Avenue Canal water collected in April 2009 is the focus of this section (see Fig. 3). The normalized effluent PO4 concentration (i.e., C/C0) exceeded 0.5 between 4320 and 5760 BV treated. The effluent TP concentration closely tracked the effluent PO4 concentration, and there was no observed problem with TP release into the treated water as seen in Fig. 2. UV254 and TOC reached breakthrough (i.e., C/ C0 > 1) between 1080 and 1560 BV treated. The chloride concentration was essentially unchanged by DWT-AS during the column experiment, while the sulfate concentration was approximately doubled over the first 120 BV treated and then returned to its initial concentration. The pH of the treated 2

PO4 UV254 Cl-

1.8 1.6

TP TOC SO4--

1.4 GBFS

LS

DWT-FS BOFSS DWT-AS low-cost materials

1.2

RC

Fig. 2 e Final normalized concentrations after 60 min jar tests using 4 g/L of low-cost material and Sanford Avenue Canal water collected January 2009 (GBFS, LS, DWT-FS, DWT-AS) and March 2009 (BOFSS, RC). Legend: PO4 (solid blue), TP (narrow upward diagonal lines), TP (theo.) (wide upward diagonal lines), UV254 (solid black), TOC (dots). All data are measured values except TP (theo.), which is equal to (initial TP L initial PO4 D final PO4)/(initial TP). All measured data are mean values of duplicate samples with error bars showing one standard deviation. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

C/C0

C/C0

the only material to remove a substantial fraction of UV254absorbing substances (>50%), with a majority of the removal coming in the first 5 min of treatment. DWT-AS was also the only material in which removal of PO4 and UV254 followed similar trends, i.e., substantial removal in the first 5 min of treatment. Finally, all low-cost materials showed greater removal of PO4 than UV254-absorbing substances after 60 min of treatment, except CFFA which released PO4. The final normalized concentrations of PO4, TP, TP (theo.), UV254, and TOC in Sanford Avenue Canal water following treatment by the low-cost materials at 4 g/L and 60 min are shown in Fig. 2. All concentrations are experimentally measured values expect TP (theo.), which is the theoretical TP concentration in the treated water due to PO4 removal only. TP (theo.) is equal to (initial TP  initial PO4 þ final PO4)/ (initial TP). For brevity, the final normalized concentrations are referred to as either measured or theoretical. No data are presented for CFFA because it added PO4 to the treated water. All low-cost materials removed PO4 to some extent (i.e., C/ C0 ¼ 0.19e0.93). However, the measured TP concentrations were >1 (i.e., C/C0 ¼ 1.01e2.14) for all low-cost materials except RC, which indicates that TP was higher in the treated water than raw water. The increase in TP was unexpected because all materials removed PO4 and would be expected to show some TP removal. This is illustrated by the theoretical TP concentrations, which were <1 for all low-cost materials. In addition, the measured TP concentration for RC was <1, but greater than the theoretical TP concentration. The discrepancy between the measured and theoretical TP concentrations suggest that attrition of the low-cost materials or precipitation of P-containing particles, or both, are adding TP to the treated water. The unfiltered supernatant

1 0.8 0.6 0.4 0.2 0 0

1000

2000

3000

4000

5000

6000

7000

treated

Fig. 3 e Results of mini-column experiment using DWT-AS and Sanford Avenue Canal water collected April 2009. Legend: PO4 (solid triangle), TP (open triangle), UV254 (solid square), TOC (open square), chloride (plus), and sulfate (ex). Bed volume [ 1 mL and empty bed contact time [ 0.5 min. Data from single column run.

4810

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 0 3 e4 8 1 4

Engineered materials

3.3.1.

Jar tests

The change in PO4, pH, and UV254 as a function of mixing time using 1 mL/L of engineered material and Sanford Avenue Canal water is shown in Fig. 4. Fig. 4a shows the normalized PO4 concentration. PO4 removal at 5 min was MIEX (10%) > Dowex22 (7%) > PSR (1%), whereas PO4 removal at 60 min was PSR (61%) > Dowex22 (29%) > MIEX (26%). A similar trend for PO4 removal was observed at a resin dose of 4 mL/L with MIEX (44%) > PSR (27%) at 5 min and PSR (80%) > MIEX (57%) at 60 min (results not shown). The results illustrate that MIEX has a faster rate of ion exchange for PO4, while PSR has a greater affinity for PO4. Fig. 4b shows the change in pH with time. The pH of the treated water changed by less than  0.5 pH units during the jar tests. The small change in pH is an expected result for strong-base anion exchange resins. Fig. 4c shows the change in UV254 absorbance with time. All engineered materials removed UV254-absorbing substances with UV254 removal at 5 min greater than UV254 removal at 60 min, e.g., PSR showed 70% UV254 removal at 5 min versus 22% UV254 removal at 60 min. In addition, all engineered materials showed greater removal of UV254 than PO4 at 5 min. At 60 min, however, PO4 removal was greater than UV254 removal for PSR and Dowex22, while UV254 removal was greater than PO4 removal for MIEX. Similar results were observed at a resin dose of 4 mL/L with UV254 removal > PO4 removal for all resins at 5 min, and PO4 removal > UV254 removal for PSR and Dowex22 at 60 min and UV254 removal > PO4 removal for MIEX at 60 min. The rate and extent of removal of PO4 and UV254 suggests that P, NOM, and other anions (e.g., sulfate) are competing for ion exchange sites on the resin. The final normalized concentrations of PO4, TP, TP (theo.), UV254, and TOC in Sanford Avenue Canal water following treatment by the ion exchange resins at 1 mL/L and 60 min are shown in Fig. 5. The measured TP concentration was <1 for all engineered materials, which was expected based on PO4 removal. The measured TP concentration was approximately equal to the theoretical TP concentration for Dowex22 and MIEX, which indicates that PO4 is the sole fraction of TP that is removed by Dowex22 and MIEX. In contrast, the measured TP

PSR Dowex22 MIEX

1.2

PO4 , C/C0

1 0.8 0.6 0.4 0.2 0 0

b

pH

3.3.

a

10

20

30

40

50

60

40

50

60

30 40 time (min)

50

60

time (min) 8

7.5

7 0

c

10

20

30 time (min)

0.9 0.8 0.7

UV254 (1/cm)

water decreased by 0.3e1.3 pH units over the duration of the column run (results not shown). The mini-column experiment using DWT-AS was repeated for Sanford Avenue Canal water collected in June 2009 (results not shown). The trends for P, NOM, and inorganic ions were very similar between the two column experiments. For example, the normalized effluent PO4 concentration exceeded 0.5 between 4320 and 5760 BV treated, and TOC and UV254 reached breakthrough at 600 and 2880 BV treated, respectively. A mini-column experiment using DWT-FS and Sanford Avenue Canal water collected in June 2009 was also conducted (results not shown). The normalized effluent PO4 concentration exceeded 0.5 between 0 and 120 BV treated. The effluent concentrations UV254 and TOC reached breakthrough between 0 and 120 BV treated. Thus, the jar tests and mini-column experiments showed effective P removal by DWT-AS.

0.6 0.5 0.4 0.3 0.2 0.1 0 0

10

20

Fig. 4 e Time-varying change in (a) PO4, (b) pH, and (c) UV254 for PSR (circle), Dowex22 (ex), and MIEX (triangle) during jar tests using 1 mL/L of engineered material and Sanford Avenue Canal water collected March 2009. PO4 and UV254 are mean values of duplicate samples with error bars showing one standard deviation. pH is data for duplicates.

concentration was less than the theoretical TP concentration for PSR, which suggests that other forms of P in addition to PO4 are being removed, such as condensed polyphosphates and dissolved organic P. Finally, the preference of the resins for P relative to NOM varied. PSR and Dowex22 followed a similar removal efficiency with PO4 > TP > UV254 z TOC. In addition, PSR showed greater removal of P and NOM than Dowex22. MIEX removed less P and more NOM than PSR and Dowex22, and had a distinctly different removal efficiency with UV254 z TOC > PO4 z TP.

4811

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 0 3 e4 8 1 4

1

PO4

TP

TP (theo.)

UV254

1.2

TOC

0.9

1

0.8 0.7

0.8 C/C0

C/C0

0.6 0.5 0.4

0.6 0.4

0.3

0.1

TP TOC SO4--

0 0

0 PSR

Dowex22 engineered materials

1000

2000

Dowex22 and MIEX increased the chloride concentration and decreased the sulfate concentration of the treated water (results not shown). These results are expected for ion exchange using chloride as the mobile counter ion. PSR removed both chloride and sulfate (results not shown). PSR is regenerated in a mixed NaCleNaOH solution (Blaney et al., 2007), which allows chloride and hydroxide to act as mobile counter ions and explains why both chloride and sulfate were removed.

Mini-column experiments

PSR and MIEX were tested in mini-column experiments using Sanford Avenue Canal water collected in April and June 2009. The results of mini-column experiments using PSR and Sanford Avenue Canal water collected in April 2009 are the focus of this section (see Fig. 6). The data in Fig. 6 are the average values of duplicate mini-column experiments with error bars showing one standard deviation. There was good agreement between the duplicate mini-column experiments as illustrated by the error bars. The normalized effluent PO4 concentration exceeded 0.5 between 4320 and 5760 BV treated. The effluent TP concentration closely tracked the effluent PO4 concentration, with the difference between the normalized concentrations of TP and PO4  0.1. The close agreement between PO4 and TP is a result of PO4 being the major fraction of TP in the raw water (i.e., PO4/TP ¼ 0.86). UV254 and TOC approached near-breakthrough at 2880 BV treated with final normalized concentrations of 0.90e0.96. Chloride and sulfate were removed over the first 120 BV, and then were unchanged after the first 120 BV treated with the

3000

4000

5000

6000

7000

treated

MIEX

Fig. 5 e Final normalized concentrations after 60 min jar tests using 1 mL/L of engineered material and Sanford Avenue Canal water collected March 2009. Legend: PO4 (solid blue), TP (narrow upward diagonal lines), TP (theo.) (wide upward diagonal lines), UV254 (solid black), TOC (dots). All data are measured values except TP (theo.), which is equal to (initial TP L initial PO4 D final PO4)/ (initial TP). All measured data are mean values of duplicate samples with error bars showing one standard deviation. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

3.3.2.

PO4 UV254 Cl-

0.2

0.2

Fig. 6 e Result of mini-column experiment using PSR and Sanford Avenue Canal water collected April 2009. Legend: PO4 (solid triangle), TP (open triangle), UV254 (solid square), TOC (open square), chloride (plus), and sulfate (ex). Bed volume [ 1 mL and empty bed contact time [ 0.5 min. Data are mean values of duplicate samples with error bars showing one standard deviation.

effluent concentration approximately equal to the influent concentration. The pH of the treated water increased by up to 0.6 pH units over the duration of the column run (results not shown). The mini-column experiment using PSR was repeated for Sanford Avenue Canal water collected in June 2009 (results not shown). The trends for P, NOM, and inorganic ions were similar between the PSR mini-column experiments using Sanford Avenue Canal water collected in April and June 2009, despite the changes in water quality. For example, the normalized effluent PO4 concentration fluctuated between 0.44 and 0.55 over 5760e8640 BV treated, and UV254 and TOC fluctuated around w0.9 after 600e1080 BV treated. Duplicate minicolumn experiments using MIEX and Sanford Avenue Canal water collected in June 2009 were also conducted (results not shown). The normalized effluent PO4 concentration exceeded 0.5 between 120 and 600 BV treated, while UV254 was <0.8 when the experiment ended at 1560 BV treated. Thus, batch (i.e., jar test) and continuous-flow (i.e., mini-column) experiments showed that PSR had a greater affinity for P relative to NOM. Furthermore, PSR achieved substantial removal of PO4 and TP over entire column runs in contrast to negligible removal of TOC, UV254, chloride, and sulfate after 120 BV. The jar test and mini-column results for PSR indicate that P is removed through an inner-sphere complex with iron oxide whereas the other anions are removed through an outer-sphere complex with the anion exchange functional groups. These observations are also consistent with the literature (e.g., Pan et al., 2009).

4.

Discussion

4.1.

Summary of low-cost materials

The experimental results allow for discussion of the low-cost materials on the basis of material properties, P removal, and

4812

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 0 3 e4 8 1 4

secondary changes in water chemistry. CFFA added PO4 to the treated water, and GBFS and LS showed negligible PO4 removal. These materials will not be discussed further because of their poor performance. BOFSS, DWT-FS, DWT-AS, and RC achieved PO4 removal in the range of 47e81% in jar tests. However, the TP concentration of the treated water was increased for BOFSS, DWT-FS, and DWT-AS in jar tests. RC showed TP removal, but the removal was less than the theoretical removal predicted based on PO4 removal. The explanation for the TP results obtained from the jar tests is attrition of the low-cost materials and/or precipitation of P-containing particles. This is supported by the mini-column experiments using DWT-AS in which the TP closely tracked PO4. Thus, the TP data for the low-cost materials using jar tests should be interpreted with caution. BOFSS, DWT-FS, DWT-AS, and RC all showed secondary changes in water chemistry. BOFSS and RC increased the pH of the treated water from pH 7.6 to pH 10.2 and pH 9.5, respectively. The increase in pH is a reflection of the material properties, such as pHPZC z 12 and aragonite (CaCO3), calcite (CaCO3), dolomite (CaMg(CO3)2), and larnite (Ca2SiO4) mineral phases. DWT-FS added UV254-absorbing substances and TOC to the treated water. XRD analysis characterized DWT-FS as non-crystalline, which is in agreement with amorphous ferric floc and weakly bound NOM. DWT-FS and DWT-AS added sulfate to the treated water because the precursor materials were metal sulfate salts. The pH decreased slightly for DWT-FS and DWT-AS due to the acidic nature of the metal salts. The mini-column experiments using DWT-AS showed that sulfate was released over the first 1000 BV and then returned to the background sulfate concentration for the remaining 5000 BV of treatment. This suggests that DWTAS could be cleaned prior to service to minimize leaching of sulfate. The results summarized in this subsection provide insights into the likely modes of PO4 uptake by the low-cost materials. RC and BOFSS have similar material properties dominated by calcium and magnesium minerals (i.e., carbonates and oxides). Dissolution of these minerals results in an increase in pH (>9) and increase in dissolved calcium and magnesium. The dissolution process creates favorable conditions for precipitation of sparingly soluble calcium phosphates (e.g., Baker et al., 1998), and the adsorption of PO4 to calcite and other mineral surfaces (e.g., Karageorgiou et al., 2007). Negligible TOC removal by RC and BOFSS is also consistent with the precipitation and adsorption mechanisms. DWT-AS and DWT-FS consisted of amorphous aluminum (hydr)oxides and amorphous ferric (hydr)oxides, respectively, and ligands such as sulfate and NOM. DWT-AS removed both PO4 and NOM (measured as UV254 and TOC) and released Hþ and sulfate, which suggests a ligand exchange process similar to that described by Yang et al. (2006). Furthermore, mini-column data showed that substantial uptake of PO4 by DWT-AS continued after sulfate, UV254, and TOC reached breakthrough, which supports ligand exchange between PO4 and aluminum (hydr)oxides. Minicolumn data for DWT-FS showed a similar trend as DWTAS but with less PO4 uptake, which is consistent with the literature (Makris et al., 2005).

4.2. Comparison of low-cost materials with engineered materials DWT-AS showed the greatest removal of PO4 and minor secondary changes in water chemistry among the low-cost materials, so it was selected for comparison with the engineered materials. DWT-AS showed the following order of removal: PO4 z TP > UV254 z TOC with initial release of sulfate, negligible change in chloride, and slight decrease in pH. DWT-AS also showed near-maximum removal of PO4 in w5 min at 100 rpm mixing. Among the engineered materials there were differences in the rate and affinity of removal. PSR and Dowex22 showed a similar order of removal as DWT-AS: PO4 > TP > UV254 z TOC. However, PSR showed greater removal of PO4 than Dowex22 because of selective inner-sphere complex between PO4 and ferric oxide in PSR, whereas PO4 uptake by Dowex22 was less selective ion exchange. In contrast to PSR and Dowex22, MIEX showed removal efficiency of UV254 z TOC > PO4 z TP. The results for MIEX illustrate its selective ion exchange uptake of NOM relative to many inorganic anions (Boyer et al., 2008). Overall, DWT-AS and PSR functioned by a similar removal mechanism (i.e., ligand exchange/inner-sphere complex), showed similar removal efficiency for P, and resulted in similar changes in water chemistry.

4.3.

Real-world applications of low-cost materials

The work described in this paper is part of a larger effort to develop innovative approaches to P remediation of surface water. The motivation for the larger project is to treat tributaries that contribute large PO4 loads to lakes. Thus, a floating treatment system is envisioned for tributary treatment. The advantages of a floating treatment system include no dependency on land access to the tributary and it is easily maneuverable in a lake or tributary. Accordingly, the goal of the larger project was to design, build, and test a pilot-scale floating treatment system that would combine both biological and physical-chemical unit processes. The biological unit process was designed as a horizontal-flow constructed wetland. The physical-chemical unit process was designed as an up-flow fluidized bed that contained either low-cost or engineered material. Tributary or lake water would flow in series through the horizontal-flow constructed wetland which would target particulate P, followed by the up-flow fluidized bed which would target dissolved PO4. Both DWT-AS and PSR were selected based on the results in this paper for testing in the floating treatment system. The results of P removal from lake water by the floating treatment system are the focus of a subsequent publication. Finally, the results in this paper are expected to advance the use of low-cost materials for the remediation of arsenic-contaminated water due to the similar chemistry of arsenate and phosphate (e.g., Hongshao and Stanforth, 2001; Nagar et al., 2010). Furthermore, low-cost materials have been previously evaluated for uptake of mercury and other heavy metals (e.g., Bailey et al., 1999; Hovsepyan and Bonzongo, 2009), however, the competing effects of NOM were not considered. The results described in this paper will allow previous data to be reevaluated for water matrices that contain organic matter.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 0 3 e4 8 1 4

5.

Conclusions

The goal of this work was to evaluate and compare the effectiveness of low-cost and engineered materials for P removal from surface water. The major conclusions of this work are as follows:  Low-cost materials are not homogeneous and show a wide range of P removal efficiencies. The best performing lowcost materials for PO4 removal were BOFSS, DWT-AS, DWT-FS, and RC. However, many of the low-cost materials caused secondary changes in water chemistry that would not be desirable from a treatment perspective. DWT-FS added organic material to the treated water, whereas BOFSS and RC increased the pH of the treated water by >2 pH units.  The mode of testing the low-cost materials was important in evaluating their performance. Results from jar tests showed an increase in turbidity and essentially no TP removal, although substantial PO4 removal was achieved for some low-cost materials. The most likely cause of the turbidity increase and unexpected TP results was attrition of the low-cost materials. This is supported by the minicolumn experiments which did not show a turbidity increase and TP removal closely tracked PO4 removal.  The engineered materials showed different rates of removal, and different removal affinities for P relative to NOM. PSR and Dowex22 showed greater removal of P than NOM, whereas MIEX showed greater removal of NOM than P.  The best performing materials for P removal were DWT-AS (low-cost material) and PSR (engineered material). Both of these materials also showed minimal undesirable secondary changes in water chemistry.

Acknowledgments This work was funded in part by SJRWMD contract number 25104 Lake Jesup Total Phosphorus Removal Treatment Technologies Floating Island Pilot Project. The authors thank project manager Dr. Sherry Brandt-Williams for her assistance throughout the project. This manuscript was improved by the helpful comments of two anonymous reviewers.

references

Abbt-Braun, G., Lankes, U., Frimmel, F.H., 2004. Structural characterization of aquatic humic substances e the need for a multiple method approach. Aquatic Sciences 66, 151e170. Apell, J.N., Boyer, T.H., 2010. Combined ion exchange treatment for removal of dissolved organic matter and hardness. Water Research 44, 2419e2430. Bailey, S.E., Olin, T.J., Bricka, R.M., Adrian, D.D., 1999. A review of potentially low-cost sorbents for heavy metals. Water Research 33, 2469e2479. Baker, M.J., Blowes, D.W., Ptacek, C.J., 1998. Laboratory development of permeable reactive mixtures for the removal

4813

of phosphorus from onsite wastewater disposal systems. Environmental Science & Technology 32, 2308e2316. Blaney, L.M., Cinar, S., SenGupta, A.K., 2007. Hybrid anion exchanger for trace phosphate removal from water and wastewater. Water Research 41, 1603e1613. Boyer, T.H., Singer, P.C., Aiken, G.R., 2008. Removal of dissolved organic matter by anion exchange: Effect of dissolved organic matter properties. Environmental Science & Technology 42, 7431e7437. Byrne, H.E., Mazyck, D.W., 2009. Removal of trace level aqueous mercury by adsorption and photocatalysis on silica-titania composites. Journal of Hazardous Materials 170, 915e919. Chouyyok, W., Wiacek, R.J., Pattamakomsan, K., Sangvanich, T., Grudzien, R.M., Fryxell, G.E., Yantasee, W., 2010. Phosphate removal by anion binding on functionalized nanoporous sorbents. Environmental Science & Technology 44, 3073e3078. Dubrovsky, N.M., Burow, K.R., Clark, G.M., Gronberg, J.M., Hamilton, P.A., Hitt, K.J., Mueller, D.K., Munn, M.D., Nolan, B.T., Puckett, L.J., Rupert, M.G., Short, T.M., Spahr, N. E., Sprague, L.A., Wilber, W.G., 2010. The Quality of Our Nation’s Waters-Nutrients in the Nation’s Streams and Groundwater, 1992e2004: U.S. Geological Survey Circular, 1350, p. 174. Electronic Code of Federal Regulations, 2011. Title 40: Protection of Environment, Part 261 e Identification and Listing of Hazardous Waste, Subpart C e Characteristics of Hazardous Waste, x 261.24 Toxicity Characteristic. Gan, F.Q., Zhou, J.M., Wang, H.Y., Du, C.W., Chen, X.Q., 2009. Removal of phosphate from aqueous solution by thermally treated natural palygorskite. Water Research 43, 2907e2915. Gao, X., 2006. TMDL Report: Middle St. Johns Basin, Lake Jesup, WBIDs 2981/2981A, Nutrients/Unionized Ammonia. Florida Department of Environmental Protection, Bureau of Watershed Management, Watershed Planning and Coordination Section, Tallahassee, FL. Guan, X.H., Shang, C., Chen, G.H., 2006. Competitive adsorption of organic matter with phosphate on aluminum hydroxide. Journal of Colloid and Interface Science 296, 51e58. Harris, W.G., White, G.N., 2008. X-ray diffraction techniques for soil mineral identification. In: Ulery, A., Drees, R. (Eds.), Methods of Soil Analysis: Part 5 e Mineralogical Methods, Soil Sci. Soc. Am. Madison, WI. pp. 81e115. Harrison, J.A., Bouwman, A.F., Mayorga, E., Seitzinger, S., 2010. Magnitudes and sources of dissolved inorganic phosphorus inputs to surface fresh waters and the coastal zone: a new global model. Global Biogeochemical Cycles 24 (1), 1e16. Hongshao, Z., Stanforth, R., 2001. Competitive adsorption of phosphate and arsenate on goethite. Environmental Science & Technology 35, 4753e4757. Hovsepyan, A., Bonzongo, J.C.J., 2009. Aluminum drinking water treatment residuals (Al-WTRs) as sorbent for mercury: Implications for soil remediation. Journal of Hazardous Materials 164, 73e80. Karageorgiou, K., Paschalis, M., Anastassakis, G.N., 2007. Removal of phosphate species from solution by adsorption onto calcite used as natural adsorbent. Journal of Hazardous Materials 139, 447e452. Makris, K.C., Harris, W.G., O’Connor, G.A., Obreza, T.A., Elliott, H. A., 2005. Physicochemical properties related to long-term phosphorus retention by drinking-water treatment residuals. Environmental Science & Technology 39, 4280e4289. Mortula, M., Gibbons, M., Gagnon, G.A., 2007. Phosphorus adsorption by naturally-occurring materials and industrial byproducts. Journal of Environmental Engineering and Science 6, 157e164. Mortula, M.M., Gagnon, G.A., 2007. Phosphorus adsorption-and oven dried alum residual solids in fixed bed column,

4814

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 0 3 e4 8 1 4

experiments. Journal of Environmental Engineering and Science 6, 623e628. Nagar, R., Sarkar, D., Makris, K.C., Datta, R., 2010. Effect of solution chemistry on arsenic sorption by Fe- and Al-based drinking-water treatment residuals. Chemosphere 78, 1028e1035. NIST, 2002. Standard Reference Material 2709, 2710 and 2711 Addendum. Standard Reference Materials Program. NIST, Gaithersburg, MD. Pan, B.J., Wu, J., Pan, B.C., Lv, L., Zhang, W.M., Xiao, L.L., Wang, X. S., Tao, X.C., Zheng, S.R., 2009. Development of polymer-based nanosized hydrated ferric oxides (HFOs) for enhanced phosphate removal from waste effluents. Water Research 43, 4421e4429. Pratt, C., Shilton, A., Haverkamp, R.G., Pratt, S., 2009. Assessment of physical techniques to regenerate active slag filters removing phosphorus from wastewater. Water Research 43, 277e282. Qualls, R.G., Sherwood, L.J., Richardson, C.J., 2009. Effect of natural dissolved organic carbon on phosphate removal by ferric chloride and aluminum sulfate treatment of wetland waters. Water Resources Research 45. Rentz, J.A., Turner, I.P., Ullman, J.L., 2009. Removal of phosphorus from solution using biogenic iron oxides. Water Research 43, 2029e2035. Seminole County Water Atlas, 2011. Sanford Ave Canal. Retrieved February 6, 2011 from. http://www.seminole.wateratlas.usf. edu/river/default.asp?wbodyid¼1045&wbodyatlas¼river. Sengupta, S., Pandit, A., 2011. Selective removal of phosphorus from wastewater combined with its recovery as a solid-phase fertilizer. Water Research 45, 3318e3330. Sibrell, P.L., Montgomery, G.A., Ritenour, K.L., Tucker, T.W., 2009. Removal of phosphorus from agricultural wastewaters using adsorption media prepared from acid mine drainage sludge. Water Research 43, 2240e2250. Sylvester, P., Westerhoff, P., Mooller, T., Badruzzaman, M., Boyd, O., 2007. A hybrid sorbent utilizing nanoparticles of hydrous iron oxide for arsenic removal from drinking water. Environmental Engineering Science 24, 104e112.

Tennant, M.F., Mazyck, D.W., 2007. The role of surface acidity and pore size distribution in the adsorption of 2-methylisoborneol via powdered activated carbon. Carbon 45, 858e864. Ugurlu, A., Salman, B., 1998. Phosphorus removal by fly ash. Environment International 24, 911e918. U.S. EPA, 1992. Method 1311: Toxicity Characteristic Leaching Procedure (Revision 0, July 1992). In SW-846. U.S. EPA, Washington, DC. U.S. EPA, 1993. Method 365.1: Determination of Phosphorus by Semi-Automated Colorimetry (Revision 2, August 1993). U.S. EPA, Cincinnati, OH. U.S. EPA, 1996. Method 3050B: Acid Digestion of Sediments, Sludges, and Soils (Revision 2, December 1996). In SW-846. U.S. EPA, Washington, DC. U.S. EPA, 2007. Method 6010C: Inductively Coupled PlasmaAtomic Emission Spectrometry (Revision 3, February 2007). In SW-846. U.S. EPA, Washington, DC. Weishaar, J.L., Aiken, G.R., Bergamaschi, B.A., Fram, M.S., Fujii, R., Mopper, K., 2003. Evaluation of specific ultraviolet absorbance as an indicator of the chemical composition and reactivity of dissolved organic carbon. Environmental Science & Technology 37, 4702e4708. Weng, L.P., Van Riemsdijk, W.H., Hiemstra, T., 2008. Humic nanoparticles at the oxide-water interface: interactions with phosphate ion adsorption. Environmental Science & Technology 42, 8747e8752. Xiong, J.B., He, Z.L., Mahmood, Q., Liu, D., Yang, X., Islam, E., 2008. Phosphate removal from solution using steel slag through magnetic separation. Journal of Hazardous Materials 152, 211e215. Yang, Y., Zhao, Y.Q., Babatunde, A.Q., Wang, L., Ren, Y.X., Han, Y., 2006. Characteristics and mechanisms of phosphate adsorption on dewatered alum sludge. Separation and Purification Technology 51, 193e200. Zeng, L., Li, X.M., Liu, J.D., 2004. Adsorptive removal of phosphate from aqueous solutions using iron oxide tailings. Water Research 38, 1318e1326. Zhu, X.P., Jyo, A., 2005. Column-mode phosphate removal by a novel highly selective adsorbent. Water Research 39, 2301e2308.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 1 5 e4 8 2 6

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Photolytic and photocatalytic transformation of methadone in aqueous solutions under solar irradiation: Kinetics, characterization of major intermediate products and toxicity evaluation Cristina Postigo a,*, Carla Sirtori b, Isabel Oller b, Sixto Malato b, Manuel Ignacio Maldonado b, Miren Lo´pez de Alda a, Damia` Barcelo´ a,c,d a

Institute of Environmental Assessment and Water Research, (IDAEA-CSIC), Department of Environmental Chemistry, C/Jordi Girona, 18-26, 08034 Barcelona, Spain b Plataforma Solar de Almerı´a (PSA-CIEMAT), Carretera Sene´s, km 4, 04200 Tabernas, Almerı´a, Spain c Catalan Institute for Water Research (ICRA), Parc Cientı´fic i Tecnolo`gic de la Universitat de Girona, Edifici H2O, 17003 Girona, Spain d King Saud University, Box 2454, Riyadh 11451, Saudi Arabia

article info

abstract

Article history:

The present manuscript describes the transformation and mineralization of methadone

Received 4 April 2011

(MET) in aqueous solutions (demineralized water (DW) and synthetic municipal wastewater

Received in revised form

effluent (SWeff)) by natural solar irradiation and two solar photocatalytic processes:

6 June 2011

heterogeneous photocatalysis with titanium dioxide (TiO2) and homogeneous photo-

Accepted 20 June 2011

catalysis by photo-Fenton. Direct solar irradiation resulted in almost complete trans-

Available online 26 June 2011

formation of MET in the investigated matrices after 20 h of normalized irradiation time. MET photocatalytic transformation required shorter illumination times in DW compared to

Keywords:

SWeff. Only 16 and 36 min of solar illumination were required during photo-Fenton and

Methadone

photocatalysis with TiO2, respectively, to transform MET completely in SWeff. Mineraliza-

Solar photo-Fenton

tion of the dissolved organic carbon took place only during photocatalytic treatments.

Solar TiO2 photocatalysis

Kinetics parameters were calculated for processes comparison. Additionally, photo-

Solar photolysis

transformation intermediates generated during each treatment were investigated and

Phototransformation products

characterized by means of ultra-performance liquid chromatography coupled to

Liquid chromatography-tandem mass

quadrupole-time of flight tandem mass spectrometry (UPLC-QqTOF-MS/MS). The main MET

spectrometry

phototransformation pathways were observed to be hydroxylation, and fragmentation and cyclization. According to the Vibrio fischeri bioassay, the acute toxicity of the generated phototransformation products was not relevant, since the observed inhibition percentages of bacterial bioluminescence were always below 30% after 30 min of sample contact. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Methadone (MET) is a synthetic opioid mainly used to treat opioids addiction. Like many other illicit and licit drugs, this

substance has been usually found in environmental waters, due to its low elimination efficiency in wastewater treatment plants that base the secondary treatment on activated sludge processes. MET efficiency removals have been reported to be

* Corresponding author. Tel.: þ34 93 400 61 00; fax: þ34 93 204 59 04. E-mail address: [email protected] (C. Postigo). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.06.027

4816

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 1 5 e4 8 2 6

usually below 40% (Boleda et al., 2009; Terzic et al., 2010). Indeed, higher levels of MET have been occasionally found in effluent wastewaters compared to influent wastewaters (Boleda et al., 2007). The presence of this compound in treated wastewaters and in river surface waters reaches levels up to 735 ng/L and 18 ng/L, respectively (Boleda et al., 2009). MET residues have been also detected in surface waters from lakes at maximum levels of 2.5 ng/L (Berset et al., 2010; Castiglioni and Zuccato, 2010) and in groundwaters at levels below 0.5 ng/L. Moreover, MET residues have been quantified below 3 ng/L in treated drinking water (Boleda et al., 2009). These concentrations, although very low, may pose an environmental risk to aquatic organisms, not yet investigated. Reducing aquatic environmental levels of MET requires the use of more efficient water treatment technologies. In this respect, advanced oxidation processes (AOPs) are being considered as an alternative to conventional water treatments. AOPs, which are characterized by the production of oxidative species, mainly hydroxyl radicals (HO), have been observed to successfully remove and even mineralize organic microcontaminants, such as pharmaceuticals and pesticides, present in aqueous solutions (Klavarioti et al., 2009). However, they are not fully implemented since they would increase water treatment costs, which can be reduced by using catalytic AOPs (as TiO2 and photo-Fenton) and sunlight as source of irradiation (Malato et al., 2009). Compound oxidation by means of HO, which also occurs in the environment, may involve several reactions that take place in a non selective way: eletrophilic addition at unsaturated CeC bonds or in aromatic rings (HO þ R / HOR), electron transfer (HO þ R / Rþ þ HO), and hydrogen abstraction at CeH, NeH and OeH bonds (HO þ RH / R þ H2O). Additionally, organic radicals generated may react with atmospheric oxygen to form peroxy radicals that generate further oxidative transformations (R þ O2 / RO2 / CO2 þ products) that contribute to the complete mineralization of the organic matter (Legrini et al., 1993; Malato et al., 2009). In the present work, the performance of two photocatalytic processes in the transformation and mineralization of MET aqueous solutions was investigated. These treatments, namely heterogeneous photocatalysis with TiO2 and homogeneous photocatalysis by photo-Fenton, were assisted by solar irradiation, and their application has been described in detail elsewhere (Malato et al., 2009). Photocatalytic experiments were run at pilot plant scale by means of compound parabolic collectors (CPCs). The chemical solar transformation of MET was compared to its natural solar transformation. All experiments were performed on two aqueous matrices (demineralized water (DW) and simulated effluent wastewater (SWeff)) to evaluate the effect of the water matrix on the transformation process. In this context, the main objectives of this study were to evaluate the transformation kinetics of MET under the investigated treatments in DW and SWeff, to identify and characterize the main phototransformation products originated in each process, and to evaluate the acute toxicity of the photoproducts generated. The study was performed at higher than environmental concentrations in order to facilitate the study of the kinetics reactions and the identification of the photoproducts generated. This aspect may slightly affect the phototransformation

kinetics but not the transformation pathways observed, since they are governed by the active oxidant species involved, and not by the contaminant concentration. To the authors’ knowledge this is the first study in investigating the photocatalytic transformation of MET in water.

2.

Experimental

2.1.

Chemicals and reagents

Methadone hydrochloride salt (purity, >99%) was provided as a concession for research purposes (2009C00124) by the Division of Narcotic Drugs and Psychotropic Substances of the Spanish Agency of Pharmaceuticals and Medical Products. 2-Ethylidene1,5-dimethyl-3,3-diphenylpyrrolidine (EDDP) (perchlorate salt, purity >99%) was purchased from Cerilliant (Round Rock, TX, U.S.A). DW used in the experiments was obtained from the Plataforma Solar de Almerı´a (PSA) demineralization plant (conductivity < 10 mS/cm, Cl‾ ¼ 0.7e0.8 mg/L, NO 3 < 0.2 mg/L, organic carbon < 0.5 mg/L). Demineralized water was also used to generate SWeff. The chemical composition of the SWeff is detailed in Table S1 (supplementary material), and was derived from the guidelines established by the U. S. Environmental Protection Agency (USEPA, 2002) and the OECD for moderately-hard synthetic freshwater and synthetic sewage (OECD, 2001). Heterogeneous photocatalysis was performed using TiO2 Degussa P-25 (Frankfurt, Germany). Reagents used in the photo-Fenton experiments were iron sulfate heptahydrate (FeSO4·7H2O) and hydrogen peroxide (H2O2) (30% w/v), and those used for pH adjustment were sulfuric acid (H2SO4) and sodium hydroxide (NaOH). All of them were purchased from Panreac (Barcelona, Spain). High performance liquid chromatography (HPLC)-grade acetonitrile (Merck, Darmstadt, Germany) and water produced by a Mili-Q ultra-pure water system from Millipore (Milford, MA, USA) were used for HPLC analyses. Ultra-performance liquid chromatography (UPLC)-grade acetonitrile and water (Merck) were used in the analyses with UPLC coupled to quadrupole-time of flight tandem mass spectrometry (UPLCQqTOF-MS/MS). Formic acid (purity, 98%) added to the chromatographic mobile phase was acquired from Fluka (Buchs, Switzerland).

2.2. Hydrolysis, photolysis and solar photocatalysis experiments All hydrolysis and phototransformation experiments were carried out during summer at the PSA (latitude 37 N, longitude 2.4 W). Concerning hydrolysis and photolysis experiments, individual solutions of MET were prepared by dissolving the compound in DW and SWeff at an initial concentration of 10 mg/L in 5 L Pyrex beakers. This value is much higher than reported environmental concentrations (Boleda et al., 2009; Fatta-Kassinos et al., 2011); however, it was chosen as MET initial concentration (MET0) for better evaluation of transformation kinetics and photointermediates generated. The

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 1 5 e4 8 2 6

beakers containing the MET solutions were kept in the dark at room temperature during hydrolysis experiments and they were exposed to direct sunlight for 6 days during the photolysis experiments. Samples were taken periodically after water solutions stirring. A CPC reactor was used for the photochemical assays. The photoreactor is composed of two modules of eight Pyrex glass tubes mounted on a fixed platform tilted 37 (local latitude), providing a total irradiated area of 3 m2. The total volume in each experiment was 35 L, but only 22 L were irradiated. At the beginning of all photochemical experiments, homogenization of the MET0 concentration and the reagents added to the process was done with the photoreactor covered to avoid any photoreaction during preparation. In the TiO2 heterogeneous photocatalytic experiments, after addition of the drug to the photoreactor, the system was well homogenized for 15 min. Subsequent addition of the catalyst (TiO2, 200 mg/L) required also homogenization of the system for 15 min more. A homogenized sample was collected before uncovering the photoreactor and starting the photocatalytic experiment to check MET0 concentration. In the photo-Fenton experiments, after homogenization of MET0 concentration in the photoreactor, the pH of the water was adjusted with sulfuric acid (H2SO4, 2N) in order to carry out the photo-Fenton reaction at a fixed pH, between 2.6 and 2.8. After 15 min of homogenization a sample was taken to confirm the pH, and afterward, the iron salt (2 mg/L) was also added and well homogenized for 15 min more. A homogenized sample was collected also before adding the first dose of H2O2, which was kept always in excess (5e20 mg/L) during the experiment, and prior to uncover the photoreactor. All the photochemical experiments were performed at different days between 9 am and 4 pm. Solar ultraviolet (UV) radiation was measured by a global UV radiometer (KIPP & ZONEN, model CUV3) mounted on a platform tilted 37 (the same as the CPC reactor). Comparison of the data obtained with diverse photochemical experiments carried out on different days is possible using Equation (1) as described elsewhere (Malato et al., 2003); where tn is the experimental time for each sample, Vi is the illuminated volume, VT is the total volume, UV is the average solar UV radiation measured during Dtn, and t30w is the normalized illumination time that refers to a constant solar UV power of 30 W/m2 (the typical solar UV power on a perfectly sunny day around noon). t30w;n ¼ t30w;n1 þ Dtn

2.3.

UV Vi ; Dtn ¼ tn  tn1 30 VT

(1)

Analytical determinations

All analytical determinations were performed on PTFE filtered samples (0.22 mm). MET levels were monitored by reversephase liquid chromatography coupled to UV detection using a HPLC-UV system (Agilent Technologies, series 1100). The mobile phase used consisted of a linear gradient of a mixture of acetonitrile/water with formic acid (25 mM) (10/90, initial conditions) and the stationary phase was a Gemini C18 column (150  3 mm, 5 mm) from Phenomenex (CA, USA). UV detection of MET was done at l ¼ 200.4 nm. Mineralization was evaluated by measuring the dissolved organic carbon (DOC) of

4817

filtered water samples with a Shimadzu-5050A TOC analyzer, which was calibrated with standard solutions of potassium hydrogen phthalate. Ammonium concentration was determined with a Dionex DX-120 ion chromatograph (IC) equipped with a Dionex Ionpac CS12A 4  250 mm column. Anion concentrations (NO 3 and carboxylates) were measured with a Dionex DX-600 ion chromatograph using a Dionex Ionpac AS11-HC 4  250 mm column. In order to ensure that the photo-Fenton reactions take place, Fe2þ/Fe3þ and H2O2 must be continuously present in the system. In this respect, and following ISO 6332, total iron concentration was monitored in water samples by colorimetric determination with 1,10-phenanthroline using a Unicam-2 spectrophotometer. The concentration of H2O2 was analyzed by means of a fast and simple spectrophotometric method, which is based on the measure of the red orange peroxovanadium cation formed when H2O2 reacts with ammonium metavanadate (Nogueira et al., 2005). After evaluating MET transformation kinetics with the photo-Fenton treatment, a controlled photo-Fenton reaction was performed by adding periodically small amounts of H2O2 (0.01 mM) to the system. Samples used to identify phototransformation products and to perform the toxicity studies were collected once the added amount of H2O2 was consumed. Samples used to identify phototransformation products were 20-fold concentrated by means of solid phase extraction (SPE) with a Baker vacuum system (J.T. Baker, The Netherlands) onto previously conditioned (5 mL of MeOH and 5 mL of demineralized water) Oasis HLB cartridges (6cc/ 200 mg, 30 mm) (Waters, Milford, MA). Analyte elution was performed with 4 mL þ 4 mL of MeOH. The eluted volume was dried under N2 and then reconstituted to 1 mL with water/ MeOH (90/10, v/v). Identification of phototransformation products generated during the different treatments was performed by means of UPLC-QqTOF-MS/MS using a Waters Acquity UPLC system coupled to a Waters/Micromass QqToF-MicroTM (Waters/ Micromass, Manchester, UK). Chromatographic separation was performed on a Waters Acquity BEH C18 column (2.1  100 mm, 1.7 mm) that was kept in a column oven at 30  C. The mobile phase consisted of a linear gradient of A: acetonitrile and B: 25 mM aqueous formic acid for analyses performed in positive electrospray ionization (PI) mode. A linear gradient of A: acetonitrile and B: water was applied for analyses carried out in the negative ionization (NI) mode. Full-scan analyses carried out on selected samples in NI mode did not show significant peaks compared to blank samples, thus further MS and MS2 analyses were performed in the PI mode. Acquisition in full scan mode was performed with a capillary voltage of 3000 V in the range m/z 50e700 at different cone voltages (15 V, 25V and 35 V) to look for potential generation of dimers. Dimerization of the organic compound or its phototransformation products has been previously observed during phototransformation experiments (Konstantinou et al., 2010). MS2 analyses at different collision energies (10e40 eV) were carried out on identified protonated molecules [M þ H]þ in order to get structural information. The collision gas applied in the fragmentation cell was argon. Data were collected in the centroid mode, with a scan time of 0.3 s and an interscan delay time of 0.1 s,

4818

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 1 5 e4 8 2 6

and with a full width at half maximum (FWHM) resolution of 5000. Other MS parameters were set as follows: 600 L/h for the desolvation gas at a temperature of 350  C, 50 L/h for the cone gas, 120  C as source temperature. A valine-tyrosine-valine (Val-Tyr-Val) solution (m/z of [M þ H]þ ¼ 380.2185) was used to tune the instrument and also as lock mass to achieve mass accuracy. Elemental compositions and accurate masses of the protonated molecules and their fragments were determined by means of MassLynx V4.1 software.

2.4.

Acute toxicity evaluation

Acute toxicity of MET and their solar phototransformation products was evaluated on selected not preconcentrated samples with Biofix Lumi-10, a commercial bioassay based on inhibition of the bioluminescence emitted by the marine bacteria Vibrio fischeri. The inhibition of light emission was measured after sample contact periods of 5, 15 and 30 min, as detailed in ISO 11348-3:2007. To avoid further transformation, the Fenton process was quenched by controlling the amount of H2O2 added to the system, as aforementioned. Quenching in samples collected during TiO2 photocatalysis and photolysis was achieved by filtration and storage in the dark at 4  C until toxicity evaluation (performed in less than 24 h).

3.

Results and discussion

3.1.

Methadone hydrolysis and photolysis

1.0

1.0

0.8

0.8

0.6

0.6

0.4

0.4

DOC/DOC0

MET/MET0

No decrease on MET0 was observed after 22 h of keeping DW and SWeff solutions in the dark at room temperature. On the contrary, MET0 was completely transformed during photolysis experiments, as it is shown in Fig. 1. Note that t30w was calculated using a variation of Equation (1), where Vi/VT was

MET/MET0 in DW DOC/DOC0 in DW

0.2

MET/MET0 in SWeff

0.2

DOC/DOC0 inSWeff

0.0

0.0 0

10

20

30

40

50

60

t 30w(hours)* Fig. 1 e MET transformation (MET/MET0) and mineralization (DOC/DOC0) during photolysis experiments performed in DW and SWeff. (*the normalized illumination time (t30w) was calculated with a variation of Equation (1), where Vi/VT [ 1).

equal to 1 as these experiments were performed in transparent beakers where the complete volume was illuminated. Direct sun-light exposure of MET solutions resulted in a decrease of more than 90% of MET0 after 10 and 17 h of sunlight irradiation time in DW and SWeff matrices, respectively. Photolytic transformation rates of MET were considered negligible compared with photocatalytic transformation rates, because the photolytic transformation of MET0 requires much longer times (hours vs. min, see Section 3.2). MET transformation during photolysis is mainly derived by photonic reactions, since hydroxyl radicals are not produced and other so oxidant species, such as superoxide radicals, are unlikely to be generated because water solutions were not aerated. This statement is confirmed by the fact that DOC levels were stable throughout the entire photolysis experiments, which were run for up to 6 days (z62 h of normalized irradiation time) in both investigated matrices, and phototransformation products differ to some extent from those generated during photocatalytic treatments (see section 3.4).

3.2.

Solar photocatalysis

3.2.1.

Heterogeneous photocatalysis with TiO2

Phototransformation and mineralization of MET in DW and SWeff with TiO2 heterogeneous photocatalysis is shown in Fig. 2a. After system homogenization and before starting the TiO2 photocatalytic treatment, a decrease in MET0 of about 7% and 10% was observed in DW and SWeff, respectively, due to adsorption of the compound onto the TiO2 particles. MET was completely transformed in DW after 23 min of solar photocatalytic treatment whereas 13 min more were necessary for its complete transformation in SWeff. Overall, DOC mineralization occurred at a slower rate than compound transformation. In this respect it must be clarified that the DOC measured in SWeff (DOC0 z 30 mg/L) is mainly generated by the added peptone, meat extract and urea (see Table S1), substances that hide the mineralization behavior of MET in SWeff. On the other hand, more than 90% of the DOC (DOC0 z 8 mg/L) present in DW, which is exclusively generated by the presence of MET, was mineralized.

3.2.2.

Homogeneous photocatalysis by photo-Fenton

Transformation of MET in the investigated water matrices with photo-Fenton treatment is shown in Fig. 2b. Phototransformation of MET with solar photo-Fenton was in general very fast. Only 4 min of photo-Fenton treatment were sufficient to transform completely MET0 in DW. Complete transformation of MET0 in SWeff required four times more of normalized irradiation time (16 min). About 80% of DOC mineralization was achieved in DW containing MET after 75 min. The mineralization rate during photo-Fenton treatment was initially higher than with TiO2 heterogeneous photocatalysis. However, similar mineralization amounts were reached at the end of both photocatalytic treatments. In this respect, the mineralization curve in the photo-Fenton treatment of MET DW solutions reached a plateau at 80%, which is slightly lower than that observed during heterogeneous photocatalysis with TiO2 (90%). Usually

4819

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 1 5 e4 8 2 6

1.0

0.8

0.8

0.6

0.6

0.4

0.4

0.2 0.0 0

50

100

150

b

1.0

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

0.2

0.0

0.0

MET/MET0

1.0

0

200

25

50

75

DOC/DOC0

1.0

DOC/DOC0

MET/MET0

a

0.0 100 125 150 175 200

t30w(min)

t30w(min)

Fig. 2 e MET transformation (MET/MET0) and mineralization (DOC/DOC0) during a) heterogeneous photocatalysis with TiO2 and b) homogeneous photocatalysis by photo-Fenton in DW and SWeff (t30w: normalized illumination time, see Equation (1)).

photo-Fenton is faster than TiO2 photocatalysis in DOC mineralization, but not necessarily higher mineralization levels are obtained, as it has been previously reported (Malato et al., 2002; Sirtori et al., 2009). This finding could be attributed to the formation of more recalcitrant carboxylic acids during the photo-Fenton treatments due to the formation of stable complexes with Fe (Pignatello et al., 2006). The same level of DOC mineralization was achieved in SWeff, compared with DW, but after more than 2 h of photo-Fenton treatment (190 min). Mineralization of MET SWeff solutions with photoFenton was comparatively higher than with TiO2 photocatalysis, since mineralization efficiency may be inhibited in the latter by phosphate ions, which are known to adsorb onto TiO2 particles, and carbonates, which behave as HO scavengers (effect not present during photo-Fenton treatment due to acidic pH). Other inorganic constituents of the SWeff matrix, such as sulfate and chloride anions, may behave as HO scavengers in both investigated photocatalytic treatments (Chong et al., 2010). 



3.3.

Kinetics of phototransformation reactions

The photocatalytic transformation of MET with TiO2 followed apparent first order kinetics, as is usual in heterogeneous photocatalysis when initial concentration is low enough and no catalyst saturation occurs. Since phototransformation intermediates generated in the process could also be competitive on the surface of the TiO2 and their concentration, like MET concentration, changes throughout the reaction up to their mineralization, the transformation process follows a LangmuireHinshelwood type mechanism and reaction kinetics could be described by Equation (2) (Herrmann, 1999):

r¼

1 þ KC þ

kr KC Pn i¼1 Ki Ci ði ¼ 1; nÞ

(2)

where kr is the reaction rate constant, K is the reactant (MET) adsorption constant, C is MET concentration at any time, and Ki and Ci are the adsorption constant and the concentration at any time, respectively, of the phototransformation products i. When C0 (10 mg/L of MET) is low enough, Equation (2) can be simplified (1 þ KCþ S. ¼ 1) to a first order reaction rate equation (see Equation (3)), which was also confirmed by the linear behavior of ln(C0/C) as a function of t30w: r ¼ kap C

(3)

In the photo-Fenton treatment and working at such concentration of MET, the main reaction that governs the transformation of this compound is that happening between HO and MET, being hydroxyl radicals at constant concentration. The concentration of HO , like in heterogeneous photocatalytic treatments with TiO2, depends on the Fe concentration, which was maintained constant in all experiments, and the photons entering in the photoreactor. Using Equation (1) the radiation entering the photoreactor was normalized. Based on the aforementioned, the rate equation of the photo-Fenton process can be written as: 



r ¼ kHO ½HO C ¼ kap C

(4)

where C is MET concentration, kHO is the photo-Fenton reaction 0 rate constant and kap is a pseudo first order constant that takes into consideration that under the commented experimental conditions, HO concentration could be considered constant. In the light of the results, and as it is shown in Table 1, it can be concluded that transformation of MET is faster with the 

4820

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 1 5 e4 8 2 6

Table 1 e Kinetic parameters obtained for phototransformation of MET in DW and SWeff with solar photocatalytic treatments (kap: pseudo first order reaction rate constant and r2: coefficient of determination, t30w, 75%DOC: normalized irradiation time required to mineralize 75% of the initial DOC). DW 1

TiO2 Photocatalysis Photo-Fenton

2

SWeff 1

kap (min )

r

t30w,75%DOC (min)

kap (min )

r2

t30w,75%DOC (min)

0.227 1.034

0.9761 0.9739

55 50

0.110 0.264

0.9653 0.972

>200 170

photo-Fenton treatment than with TiO2 heterogeneous photocatalysis, being about four times and two times faster in DW and in SWeff water, respectively. This assessment is supported by the half-life time of MET observed in TiO2 heterogeneous photocatalysis experiments (3 min in DW and 8 min in SWeff) and in photo-Fenton experiments (0.5 min in DW and 4 min in SWeff). The observed results may be explained by

the higher solar light harvesting that the photo-Fenton process presents as compared to TiO2 photocatalysis, which finally produces larger quantities of HO in less time. In solar photo-Fenton treatments, the effective wavelength can reach up to 600 nm depending on the presence of different iron complexes, whereas in TiO2 photocatalysis it is below 390 nm. Additionally, contrary to photo-Fenton treatments, in TiO2

Table 2 e Accurate mass measurement of protonated molecules and fragment product ions of MET and its phototransformation intermediates obtained with UPLC-ESI-QqTOF-MS/MS analyses. Comp.a

tR(min)

MET

5.98

b P263 (EMDP)

5.68

P277 (EDDP)

5.45

P293 (OH-EDDP)

4.26 4.44 4.46 4.59

c

P309

Precursor ion Product ion

Molecular formula

þ

[M þ H] [M þ HeC2H7N]þ [M þ HeC5H13N]þ [M þ HeC14H23N]þ [M þ HeC14H21NO]þ [M þ H]þ [M þ HeC2H5] [M þ HeCH3eC2H5]þ [M þ H]þ [M þ HeC2H5] [M þ HeCH3eC2H5]þ [M þ HeCH3eC6H5]þ [M þ H]þ [M þ HeC2H5] [M þ HeC H3eC2H5]þ [M þ HeCH3eCH3-C2H5] [M þ HeCH3eC6H5]þ [M þ HeCH3eC6H5O]þ [M þ H]þ [M þ HeC2H7N]þ M þ HeC2H7NeH2O]þ [M þ HeC5H13N]þ [M þ HeC13H19NO]þ [M þ HeC14H21NO]þ [M þ H]þ [M þ HeC2H7N]þ [M þ HeC5H13N]þ [M þ HeC6H13NO]þ [M þ HeC14H23N]þ [M þ HeC14H23NO]þ [M þ H]þ [M þ HeC2H7N]þ [M þ HeC5H13N]þ [M þ HeC6H13NO]þ [M þ HeC14H23NO]þ 



5.36 5.49 5.54 5.67

P325 (OH-MET)

4.83 4.96 5.01 5.07

P341

3.70 3.94 4.15





C21H28NO C19H21O C16H15O C 7H5O C 7H7 C19H22N C17H17N C16H14N C20H24N C18H19N C17H16N C13H16N C20H24NO C18H19NO C17H16NO C16H13NO C13H16NO C13H16N C21H28NO C19H21O C19H19 C16H15O C 8H9 C 7H7 C21H28NO2 C19H21O2 C16H15O2 C15H15O C7H5O2 C 7H5O C21H28NO3 C19H21O3 C16H15O3 C15H15O2 C7H5O2

Mass (m/z)

Error

DBE

Experimental

Calculated

(mDa)

(ppm)

310.2168 265.1604 223.1131 105.0344 91.0551 264.1759 235.1353 220.1120 278.1915 249.1523 234.1292 186.1286 294.1863 265.1478 250.1244 235.0992 202.1226 186.1294 310.2181 265.1607 247.1500 223.1132 105.0699 91.0549 326.2114 281.1550 239.1060 211.1115 121.0294 105.0340 342.2055 297.1490 255.1030 227.1069 121.0290

310.2171 265.1592 223.1123 105.0340 91.0548 264.1752 235.1361 220.1126 278.1909 249.1517 234.1283 186.1283 294.1585 265.1467 250.1232 235.0997 202.1232 186.1283 310.2171 265.1592 247.1487 223.1123 105.0704 91.0548 326.2120 281.1542 239.1072 211.1123 121.0290 105.0340 342.2069 297.1491 255.1021 227.1072 121.0290

0.3 1.2 0.8 0.4 0.3 0.7 0.8 0.6 0.6 0.6 0.9 0.3 0.5 1.1 1.2 0.5 0.6 1.1 1.0 1.5 1.3 0.9 0.4 0.1 0.6 0.8 1.2 0.8 0.4 0.0 1.4 0.1 0.9 0.3 0.0

1.0 4.5 3.6 3.8 3.3 2.6 3.4 2.7 2.2 2.4 3.8 1.6 1.7 4.1 4.8 2.1 3.0 5.9 3.2 5.7 5.3 4.0 3.8 1.1 1.8 2.8 5.0 3.8 3.3 0.0 4.1 0.3 3.5 1.3 0.0

8.5 9.5 9.5 5.5 4.5 9.5 10.0 10.5 9.5 10.0 10.5 6.5 9.5 10.0 10.5 11.0 6.5 6.5 8.5 9.5 10.5 9.5 4.5 4.5 8.5 9.5 9.5 8.5 5.5 5.5 8.5 9.5 9.5 8.5 5.5

a If no symbol is written, the phototransformation product was identified in samples collected during photolysis and the two investigated photocatalytic treatments. b Only generated during the two investigated photocatalytic treatments. c Only generated during photolysis of MET.

4821

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 1 5 e4 8 2 6

photocatalysis, reactions do not take place in dark zones (Malato et al., 2009). As the mineralization does not follow simple models like first or zero order kinetics, overall rate constants cannot be calculated. Therefore, the normalized irradiation time necessary to mineralize 75% of the initial DOC was provided to compare experiments (see Table 1), and in this case, photo-Fenton was also more effective than TiO2 photocatalysis in compound mineralization. Lower transformation and mineralization rates of the compound in SWeff compared to DW systems indicate the non-selective attack of HO, which also react with other inorganic and organic species that are present in the SWeff, being the obtained kinetics more realistic than in DW experiments.  The formation of nitrogen inorganic species (NHþ 4 and NO3 ) and carboxylic acids produced from MET transformation was monitored by IC during the photocatalytic treatments carried out with DW. These measurements were not performed in SWeff experiments, as the chemical composition of this matrix will interfere in the results obtained. In the light of the results, the heteroatoms present in the molecule of MET were  released as NHþ 4 and NO3 . During photocatalytic treatments of MET aqueous solutions initial NO 3 levels (<0.6 mg/L) increased up to 1.17 mg/L by the end of the experiments, whereas initial NHþ 4 levels increased from levels below 0.1 mg/L to 0.4 mg/L. Taking into account that MET contains one atom of N and that 10 mg/L of MET (0.45 mg/L of N) were decomposed, it can be concluded that 80% of organic N was mineralized. Main carboxylic acids accumulated during these processes were oxalic acid and maleic acid, which reached levels up to 2 mg/L.

full scan of the former evidenced the existence of a greater number of peaks, and hence phototransformation products, than the latter and also because the lower intensity of some of the phototransformation intermediates observed in nonpreconcentrated samples affected negatively the mass accuracy of the results. Total ion current (TIC) chromatograms obtained after full-scan MS analyses of representative preconcentrated DW samples are shown as supplementary information in Fig. S1. Experimental and theoretical masses (m/z), the error between them in mDa and ppm, the double bond equivalent (DBE), and the proposed elemental composition of the protonated phototransformation intermediates and their main fragment ions formed during phototransformation of MET are shown in Table 2. The calculated accurate masses of parent ions, which were present in all cases in their protonated form [M þ H]þ, were obtained by constraining molecule elements as follows: C: 0e45, H: 0e80, N: 0e2, O: 0e8. A tolerance of 5 ppm in the error between the measured and the calculated accurate mass was considered in all but a few cases (where it was slightly higher) in order to guarantee the correct assignment of the molecular formula of major ions (Ferrer and Thurman, 2003). The evolution of the most abundant phototransformation intermediates is shown in Fig. 3. During solar photolysis, photointermediates levels continually increase as the MET0 decreases, whereas during solar photocatalysis, the highest levels of photointermediates were observed when more than half of the MET0 was transformed and they were readily phototransformed afterward. Overall, all identified phototransformation intermediates of MET were present in both investigated aqueous matrices. However, some of them were exclusively generated by one of the investigated treatments, as it has been indicated in Table 2. In this respect, up to four isobaric compounds (P309) were identified in water samples collected only during solar photolysis. The protonated molecule of these compounds presented the same exact mass, and in consequence, the same molecular formula, than the protonated molecule of

Identification of phototransformation intermediates of MET was performed on preconcentrated DW and SWeff samples rather than on non-preconcentrated ones because analysis in

P309 (tR=5.36)

b

100

40 P325 (tR=4.83)

P309 (tR=5.49)

P325 (tR=4.96)

60

P325 (tR=5.07)

20

MET 40 10

30

P325 (tR=5.07)

(Photo-product/MET0)*100

P325 (tR=4.83) P325 (tR=5.01)

P325 (tR=5.01)

80 (MET/MET0)*100

(Photo-product/MET0)*100

P309 (tR=5.67)

P277 P341 (tR=3.70) P341 (tR=3.94)

60

MET 40

10

0 -20

0

20

40

t30w (hours)*

60

P325 (tR=4.83) P325 (tR=5.01) P325 (tR=5.07)

30

0 0

10

20

30

40

t30w (min)

100 200

80

P341 (tR=3.70) P341 (tR=3.94) P341 (tR=4.15)

60

MET

20

40 10 20

20

0

100

P325 (tR=4.96)

80

P341 (tR=4.15)

20

20

0

40

P325 (tR=4.96)

P309 (tR=5.54)

30

c

100

(Photo-product/MET0)*100

40

(MET/MET0)*100

a

(MET/MET0)*100

3.4. Major phototransformation intermediates and phototransformation routes

0

0 0

5

10

15

20

t30w (min)

Fig. 3 e Time evolution of the major phototransformation intermediates of MET generated in DW during: a) hydrolysis (shadowed part) and solar photolysis, b) solar heterogeneous photocatalysis with TiO2 and c) solar photo-Fenton (t30w: normalized illumination time, in the case of photolysis it was calculated with a variation of Equation (1) where Vi/VT [ 1 and in the case of photocatalysis with TiO2 and solar photo-Fenton it was calculated with Equation (1)).

4822

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 1 5 e4 8 2 6

MET ([M þ H]þ ¼ 310.2168, C21H28NO). However, as it is shown in Fig. 4, the fragmentation patterns of those compounds labeled as P309 (tR ¼ 5.36, 5.49, 5.54 and 5.67) are different from that observed for MET (tR ¼ 5.98) at a given collision energy, e.g., 20 eV, whose identity was confirmed with the analysis of a MET standard solution. The fragmentation of MET was characterized by the presence of two main fragment ions with calculated exact masses of m/z 265.1592 and 105.0340 as main fragments ions, which gave as best fit formula C19H21O and C7H5O, respectively. The ion at m/z 265 corresponds to the loss of dimethylamine (HN(CH3)2, 45 uma), whereas the ion at m/z 105 is believed to correspond to a benzoyl cation formed after structural reordering (Joyce et al., 2004). Fragmentation of the isobaric compounds at tR ¼ 5.36, 5.49, 5.54 and 5.67 resulted in the ion at m/z 223.1123 (C16H15O) as main fragment ion. This ion, which may result from the sequential loss of

dimethylamine and the alkene CH2 ¼ CHCH3 was also present in the fragmentation spectrum of MET, but at a lower intensity (Joyce et al., 2004). The fragmentation of some of the compounds labeled as P309 favors also the formation of ions present in the MET spectrum, such as m/z 265 and m/z 247.1487 (C9H19), the latter generated by further loss of a molecule of H2O. Thus, similarities in the fragmentation spectra of the compounds labeled as P309 and that obtained for MET suggest that compounds P309 are positional isomers of MET. The protonated molecule of P277, compound generated during all investigated treatments presented an accurate mass of m/z 278.1915 (calculated exact mass of m/z 278.1909), which gave the best fit molecular formula C20H24N. This compound was identified as EDDP, one of the main urinary metabolites of MET (Ferrari et al., 2004; Goldstein and Brown,

Fig. 4 e Chromatogram and spectra obtained with MS/MS experiments of m/z 310 at a collision energy of 20 eV in a DW sample collected during the solar photolysis experiment.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 1 5 e4 8 2 6

2003) and its identity was confirmed with the MS and MS/MS analysis of a standard solution of EDDP, as it is shown in Fig. S2 as supplementary information. The phototransformation product P263 was observed only in water samples collected during photocatalytic processes. The calculated exact mass of its protonated molecule, obtained with an error below 5 ppm, m/z ¼ 264.1752 (C19H22N), fits with another urinary metabolite of MET, 2-ethyl-5-methyl3,3-diphenylpyrroline (EMDP), which is a N-demethylated form of EDDP. The main ions observed during its fragmentation (m/z 235 and m/z 220) correspond to the sequential loss of an ethyl and a methyl group, thus following the same fragmentation pattern as EDDP (see Table 2), and confirm previously published findings (Kelly et al., 2005). During the investigated photolytic and photocatalytic treatments, up to four isobaric compounds, labeled as P293, whose protonated molecule presents a calculated exact mass of m/z of 294.1585 that corresponds to the molecular formula C20H24NO, were also formed (see Table 2). Note that these phototransformation products are generated with less intensity during photolysis than during photocatalysis. Their fragmentation spectra, the molecular formula of their protonated form and the fragment ions generated during MS2 analysis (shown as supplementary material in Figs. S3 and S4) suggest that they are monohydroxilated derivatives of EDDP. The main fragment ions observed during MS/MS analyses of the protonated molecule of the P293 compounds, m/z 265.1467 e C18H19NO and m/z 250.1232 e C17H16NO, as well as the parent ion, are similar to those observed for EDDP plus 16 uma, i.e., an oxygen atom (see Figs. S2 and S4). Despite the fact that the fragmentation spectra obtained do not allow identifying the exact position of the OH group in the molecules, hydroxylation seems to occur at the phenyl groups of EDDP. This assessment is supported by the presence of the fragment ion with m/z 186.1283 e C13H16N, which could be generated by the sequential loss of a methyl group and a hydroxylated phenyl group. The investigated treatments also generated monohydroxylated (P325) and dihydroxylated (P341) derivatives of MET. Concerning monohydroxylated species, several isobaric compounds with m/z 326.2120 (C21H28NO2) were detected, and the fragmentation spectra and fragmentation pathway of the most intense m/z 326 ions have been provided as supplementary material in Fig. S5. These compounds are also the major phototransformation products generated during the investigated treatments (see Fig. 3 and Fig. S1). MS/MS analyses of P325 revealed ions at m/z 281.1542 e C19H21O2 and 239.1072 e C16H15O2 as major fragments. These fragment ions correspond to the main MET fragment ions with m/z 265 and m/z 223 plus an additional atom of oxygen (16 uma) (see Table 2). Additionally, the presence of the fragment ions at m/z 105 (major fragment ion of MET) and m/z 121 (105 þ 16 um), which could correspond to a benzoyl cation (Castiglioni et al., 2008) and a hydroxybenzoyl cation, respectively, may indicate that the HO attack (in the case of photocatalytic treatments) took place at one of the phenyl groups of the molecule. Despite the fact that dihydroxylated derivatives of MET, labeled as P341, were also found in water samples collected during MET photolysis, they were generated with higher intensity during photocatalytic treatments (see Fig. 3 and

4823

Fig. S1). Up to three isobaric compounds presented a calculated exact mass of m/z 342.2069 ([M þ H]þ e C21H28NO3) in all cases with an error below 5 ppm. MS/MS analyses showed the ion at m/z 121 as major fragment ion. As it was discussed for compounds P325, this fragment ion is indicative of a hydroxylated phenyl group. Therefore, with the structural information obtained (fragmentation spectra shown as supplementary material in Fig. S6), it is possible to predict that one of the HO attacks occurred at one of the phenyl groups of the molecule. However, MS/MS analyses do not provide enough information to assess that the other HO attack may have taken place at the other phenyl group or in another part of the molecule because other major fragments were obtained (m/z 297.1491 e C19H21O3 and m/z 255.1030 e C16H15O3), which corresponded to main MET fragment ions plus two atoms of oxygen (m/z 265 þ 32 and m/z 223 þ 32). Fig. 5 shows the main proposed phototransformation pathways of MET during solar photocatalytic treatments. Considering the identified photointermediates and their abundance, the primary phototransformation route would lead to the multistep hydroxylation of MET (P325 y P341) via HO attack. Multi-step hydroxylations were also identified as one of the main photocatalytic transformation route of COC in aqueous solutions (Postigo et al., 2011). Further transformation would proceed through fragmentation and cyclization of MET (EDDP and EMDP). Reisch and Schildgen reported that MET in its solid form or in aqueous solution reacts in presence of UV light mainly via fragmentation which produces propionaldehyde (C2H5CHO) and N,N-dimethyl-4,4,-diphenylbut-3-en-2amine ([M þ H]þ ¼ 251.1674 uma e C18H21N), and cyclization, which generates 2-ethylideno-5-methyl-3,3-diphenyl-tetrahydrofuran ([M þ H]þ ¼ 264.1514 uma e C19H20O) (Reisch and Schildgen, 1972a, b). However, these compounds were not found in the investigated samples by means of UPLC-ESIQqToF-MS/MS, perhaps because they were not retained in

Fig. 5 e Proposed phototransformation pathway of MET in aqueous solution during solar photocatalytic treatments (MET: methadone, EDDP: 2-ethylidene-1,5-dimethyl-3,3diphenylpyrrolidine, EMDP: 2-ethyl-5-methyl-3,3diphenylpyrroline, OH-MET: hydroxymethadone). * Identity confirmed with the analysis of a commercial analytical standard solution.

4824

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 1 5 e4 8 2 6

b 50

50 1.0

45

Inhibition of bioluminescence (%)

c 50

40

1.0

MET=0 MET0

40 MET=0 MET0

35

40

0.8

30

1.0

MET=0 MET0

0.8

30

0.8

30

0.6

0.6

0.6

25 20

20

20

0.4

0.4

0.4

DOC/DOC0 (mg/L)

a

15 10

0.2

10

0.2

10

0.2

5 0

0.0

S 0 S1 S 2 S 3 S 4 S 5 S 6 S7 S 8 S 9

-

t *

30w

+

0

0.0

S0 S1 S2 S3 S4 S5 S6 S7

-

t 30w

+

0

0.0

S 0 S1 S 2 S 3 S 4 S 5

-

t

30w

+

Fig. 6 e Inhibition of V. fischeri bioluminescence (%) after 30 min of contact with selected DW samples collected during a) solar photolysis, b) solar heterogeneous photocatalysis, and c) photo-Fenton experiments. (White dots indicate the DOC content of each sample).

SPE or they could be formed but also rapidly degraded by HO. Identified phototransformation routes of MET in DW and in SWeff were equal. Further transformation of most of these compounds into aliphatic products, e.g., carboxylic acids before complete mineralization would imply cleavage of their benzene rings. Dimmers of MET or its identified phototransformation products were not observed to be formed in any case.

3.5.

Toxicity evaluation

The acute toxicity of MET and its phototransformation intermediates was monitored with the V. fischeri toxicity test. Bioluminescence inhibition percentages obtained in selected DW samples during the photolysis, heterogeneous photocatalysis with TiO2 and photo-Fenton are shown in Fig. 6. Bioluminescence inhibition values were usually below 30%, which indicates that MET and its intermediates have low acute toxicity effects on the tested bacteria. Overall, higher inhibition percentages were obtained in DW samples compared to SWeff samples, where stimulation percentages were mainly observed. This fact may be attributed to the higher content of organic matter in SWeff than in DW, which may be available for the cells as a food resource. In the light of the results, bioluminescence emitted by the bacteria seemed to be more inhibited during photolysis than in solar photocatalytic treatments. As it is shown in Fig. 6, complete disappearance of MET did not have a strong influence on inhibition percentage, which points out the generated photointermediates as the main responsible for bioluminescence inhibition. This is also confirmed by the initial samples (S0), which only contained MET and did not inhibit the bacterial bioluminescence and by comparing the evolution of the

generated photo-intermediates (see Figs. 3 and 6). On one hand, many photo-intermediates generated during photolysis were not further transformed and thus, they, and their associated toxicity, remained in solution at the end of the treatment. On the other hand, photo-intermediates generated during photocatalytic treatments are completely transformed and substantially mineralized at the end of the treatment, which is in agreement with the disappearance of the acute toxicity. Photo-Fenton treatment seems to produce less acute toxic effects. However, since the same photointermediates were identified during both photocatalytic treatments, this fact may be attributed to the rapid transformation of the most harmful chemical species during photo-Fenton treatment, or to the generation of more toxic, still unidentified, species during TiO2 photocatalysis.

4.

Conclusions

MET in aqueous solutions did not experience hydrolysis, but it was substantially transformed in the presence of solar irradiation. Compared with direct solar photolysis, transformation rate of MET in aqueous solutions was highly increased with the solar photocatalytic processes investigated i.e. photo-Fenton and heterogeneous photocatalysis with TiO2 (min vs hours). Among the studied photocatalytic treatments, photo-Fenton was observed to be slightly more efficient than TiO2 photocatalysis in MET transformation and mineralization. This work confirms that the studied AOPs are good alternatives to decontaminate waters containing MET. The use of natural solar irradiation increases the costeffectiveness of photocatalytic processes, and thus, facilitates their potential application.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 1 5 e4 8 2 6

MET transformation rates observed in SWeff are more realistic than those observed in DW, since other chemical species are present in solution and also react with the generated HO . However, lower transformation rates may be expected in real waters with high organic carbon loads. The presence of natural organic matter should not affect to the transformation pathway here reported, though other transformation routes may also take place due to the formation of reactive oxygen species (ROS) other than HO. However, in this respect, it is important to remark that the formation of ROS during AOPs is minor compared to the production of HO radicals. Additionally, the treatment of lower MET initial concentrations would benefit the transformation rates and would result in less toxic treated samples, since the amount of phototransformation products generated will be also reduced. 

Acknowledgments The research leading to these results has received funding from the European Community’s Seventh Framework Programme ([FP7/2007-2013]) under grant agreement n. 265264, the Spanish Ministry of Science and Innovation through the projects SCARCE (Consolider-Ingenio 2010 CSD2009-00065) and CEMAGUA (CGL2007-64551/HID), and the “Programa de Acceso y Mejora de Grandes Instalaciones Cientı´ficas Espan˜olas” (Plataforma Solar de Almerı´a, GIC-05-17). It reflects only the author’s views. The Community is not liable for any use that may be made of the information contained therein. Cristina Postigo acknowledges the European Social Fund and AGAUR (Generalitat de Catalunya, Spain) for their economical support through the FI pre-doctoral grant. Carla Sirtori wishes to thank the CAPES foundation-Ministry of Education for the Ph.D. Research grant (BEX Processo: 37630-05-6). Special thanks to Agustin Carrio´n and Elisa Ramos for assistance during experiments execution.

Appendix. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.06.027.

references

Berset, J.D., Brenneisen, R., Mathieu, C., 2010. Analysis of licit and illicit drugs in waste, surface and lake water samples using large volume direct injection high performance liquid chromatography - electrospray tandem mass spectrometry (HPLC-MS/MS). Chemosphere 81 (7), 859e866. Boleda, M.R., Galcera´n, M.T., Ventura, F., 2007. Trace determination of cannabinoids and opiates in wastewater and surface waters by ultra-performance liquid chromatographytandem mass spectrometry. Journal of Chromatography A 1175 (1), 38e48. Boleda, M.R., Galcera´n, M.T., Ventura, F., 2009. Monitoring of opiates, cannabinoids and their metabolites in wastewater, surface water and finished water in Catalonia, Spain. Water Research 43 (4), 1126e1136.

4825

Castiglioni, S., Zuccato, E., Chiabrando, C., Fanelli, R., Bagnati, R., 2008. Mass spectrometric analysis of illicit drugs in wastewater and surface water. Mass Spectrometry Reviews 27 (4), 378e394. Castiglioni, S., Zuccato, E., 2010. Illicit drugs in the environment: emerging contaminants and indicators of drug abuse. Integrated Environmental Assessment and Management 6 (1), 186e187. Chong, M.N., Jin, B., Chow, C.W.K., Saint, C., 2010. Recent developments in photocatalytic water treatment technology: a review. Water Research 44, 2997e3027. Fatta-Kassinos, D., Meric, S., Nikolau, A., 2011. Pharmaceutical residues in environmental waters and wastewaters: current state of knowledge and future research. Analytical and Bioanalytical Chemistry 399 (1), 251e275. Ferrari, A., Coccia, C.P.R., Bertolini, A., Sternieri, E., 2004. Methadone - metabolism, pharmacokinetics and interactions. Pharmacological Research 50 (6), 551e559. Ferrer, I., Thurman, E.M., 2003. Liquid chromatography/time-offlight/mass spectrometry (LC/TOF/MS) for the analysis of emerging contaminants. Trends in Analytical Chemistry 22 (10), 750e756. Goldstein, A., Brown, B.W., 2003. Urine testing in methadone maintenance treatment: applications and limitations. Journal of Substance Abuse Treatment 25 (2), 61e63. Herrmann, J.M., 1999. Heterogeneous photocatalysis: fundamentals and applications to the removal of various types of aqueous pollutants. Catalysis Today 53 (1), 115e129. Joyce, C., Smyth, W.F., Ramachandran, V.N., O’Kane, E., Coulter, D.J., 2004. The characterisation of selected drugs with amine-containing side chains using electrospray ionisation and ion trap mass spectrometry and their determination by HPLC-ESI-MS. Journal of Pharmaceutical and Biomedical Analysis 36 (3), 465e476. Kelly, T., Doble, P., Dawson, M., 2005. Chiral analysis of methadone and its major metabolites (EDDP and EMDP) by liquid chromatography-mass spectrometry. Journal of Chromatography B 814 (2), 315e323. Klavarioti, M., Mantzavinos, D., Kassinos, D., 2009. Removal of residual pharmaceuticals from aqueous systems by advanced oxidation processes. Environment International 35 (2), 402e417. Konstantinou, I.K., Lambropoulou, D.A., Albanis, T.A., 2010. In: Fatta-Kassinos, D., Bester, K., Ku¨mmerer, K. (Eds.), Xenobiotics in the Urban Water Cycle: Mass Flows, Environmental Processes, Mitigation and Treatment Strategies Environmental Pollution. Springer, Netherlands, pp. 179e194. Legrini, O., Oliveros, E., Braun, A.M., 1993. Photochemical processes for water treatment. Chemical Reviews 93 (2), 671e698. Malato, S., Blanco, J., Ca´ceres, J., Ferna´ndez-Alba, A.R., Agu¨era, A., Rodrı´guez, A., 2002. Photocatalytic treatment of water-soluble pesticides by photo-Fenton and TiO2 using solar energy. Catalysis Today 76 (2e4), 209e220. Malato, S., Blanco, J., Vidal, A., Alarco´n, D., Maldonado, M.I., Ca´ceres, J., Gernjak, W., 2003. Applied studies in solar photocatalytic detoxification: an overview. Solar Energy 75 (4), 329e336. Malato, S., Ferna´ndez-Iba´n˜ez, P., Maldonado, M.I., Blanco, J., Gernjak, W., 2009. Decontamination and disinfection of water by solar photocatalysis: recent overview and trends. Catalysis Today 147 (1), 1e59. Nogueira, R.F.P., Oliveira, M.C., Paterlini, W.C., 2005. Simple and fast spectrophotometric determination of H2O2 in photo-Fenton reactions using metavanadate. Talanta 66 (1), 86e91. OECD, 2001. Test No. 303: Simulation Test - Aerobic Sewage Treatment - A: Activated Sludge Units ; B: Biofilms. OECD Guidelines for the Testing of Chemicals, Section 3:

4826

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 1 5 e4 8 2 6

Degradation and Accumulation, OECD Publishing, Paris. Available at. http://www.oecd-ilibrary.org/environment/testno-303-simulation-test-aerobic-sewage-treatment-aactivated-sludge-units-b-biofilms_9789264070424-en accessed in May 2011. Pignatello, J.J., Oliveros, E., Mackay, A., 2006. Advanced oxidation processes for organic contaminant destruction based on the Fenton reaction and related chemistry. Critical Reviews in Environmental Science Technology 36 (1), 1e84. Postigo, C., Sirtori, C., Oller, I., Malato, S., Maldonado, M.I., Lo´pez De Alda, M.J., Barcelo´, D., 2011. Solar transformation and photocatalytic treatment of cocaine in water: kinetics, characterization of major intermediate products and toxicity evaluation. Applied Catalysis B: Environmental 104 (1e2), 37e48. Reisch, J., Schildgen, R., 1972a. Light-induced fragmentation of D(-)-methadone hydrochloride in aqueous solution. Archiv der Pharmazie 305, 40e48.

Reisch, J., Schildgen, R., 1972b. Photolysis of crystalline D(-)-methadone hydrochloride; radiolysis of aqueous solutions of D-(-)-methadone hydrochloride. Archiv der Pharmazie 305, 49e53. Sirtori, C., Zapata, A., Malato, S., Gernjak, W., Ferna´ndez-Alba, A. R., Agu¨era, A., 2009. Solar photocatalytic treatment of quinolones: intermediates and toxicity evaluation. Photochemical and Photobiological Sciences 8 (5), 644e651. Terzic, S., Senta, I., Ahel, M., 2010. Illicit drugs in wastewater of the city of Zagreb (Croatia) - Estimation of drug abuse in a transition country. Environmental Pollution 158 (8), 2686e2693. USEPA, 2002. Methods for Measuring the Acute Toxicity of Effluents and Receiving Waters to Freshwater and Marine Organisms. EPA-821-R-02e012, USEPA, Washington D.C. Available at. http://water.epa.gov/scitech/methods/cwa/wet/ upload/2007_07_10_methods_wet_disk2_atx.pdf accessed in May 2011.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 2 7 e4 8 3 6

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Fate of antibiotics in activated sludge followed by ultrafiltration (CAS-UF) and in a membrane bioreactor (MBR) Eyal Sahar a,*, Rami Messalem b, Haim Cikurel c, Avi Aharoni c, Asher Brenner a, Manuel Godehardt d, Martin Jekel d, Mathias Ernst e a

Unit of Environmental Engineering, Faculty of Engineering Sciences, Ben-Gurion University of the Negev, P.O. Box 653, 84105 Beer-Sheva, Israel b Zuckerberg Institute for Water Research, Ben-Gurion University of the Negev, Sede Boker Campus 84990, Israel c Mekorot Israel Water Works, 9 Lincoln Street, 61201 Tel-Aviv, Israel d TU Berlin, Chair of water quality control, Sec. KF4, Strasse des 17. Juni 135, D-10623 Berlin, Germany e TU Berlin, Centre for Water in Urban Areas, KF4, Strasse des 17. Juni 135, D-10623 Berlin, Germany

article info

abstract

Article history:

The fates of several macrolide, sulphonamide, and trimethoprim antibiotics contained in

Received 4 October 2010

the raw sewage of the Tel-Aviv wastewater treatment plant (WWTP) were investigated

Received in revised form

after the sewage was treated using either a full-scale conventional activated sludge (CAS)

20 June 2011

system coupled with a subsequent ultrafiltration (UF) step or a pilot membrane bioreactor

Accepted 20 June 2011

(MBR) system. Antibiotics removal in the MBR system, once it achieved stable operation,

Available online 30 June 2011

was 15e42% higher than that of the CAS system. This advantage was reduced to a maximum of 20% when a UF was added to the CAS. It was hypothesized that the

Keywords:

contribution of membrane separation (in both systems) to antibiotics removal was due

Antibiotics

either to sorption to biomass (rather than improvement in biodegradation) or to

Organic micropollutants

enmeshment in the membrane biofilm (since UF membrane pores are significantly larger

Membrane bioreactor

than the contaminant molecules). Batch experiments with MBR biomass showed a mark-

Municipal wastewater

edly high potential for sorption of the tested antibiotics onto the biomass. Moreover,

Sorption

methanol extraction of MBR biomass released significant amounts of sorbed antibiotics.

Ultrafiltration

This finding implies that more attention must be devoted to the management of excess sludge. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Organic micropollutants (OMPs) have become a major health issue in terms of sewage treatment quality due to their potentially harmful impacts on the internal systems of a variety of organisms (Damstra et al., 2002; Khan et al., 2004). The current strong interest in OMPs has developed in reaction to: 1) natural population growth and an increased standard of

living, both of which have contributed to overall increases in quantities of sewage, the most concentrated source of OMPs (Heberer et al., 2002; Kreuzinger et al., 2004; Paxeus, 2004); 2) the increased frequency of indirect potable wastewater reuse, which enhances the transport of the relatively stable OMP molecules from the municipal effluents to the ground water, from where they can become a health threat to biological systems, including those of humans (Daughton and Ternes,

* Corresponding author. E-mail addresses: [email protected] (R. Messalem), [email protected] (A. Aharoni), [email protected] (A. Brenner), mathias.ernst@ tu-berlin.de (M. Ernst). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.06.023

4828

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 2 7 e4 8 3 6

1999; Kolpin et al., 2001; Nghiem and Scha¨fer, 2005); 3) technological improvements to achieve nanoscale measurement capabilities in the analytical devices used to measure OMPs (Ternes and Joss, 2006). The main processes effective in OMP removal include advanced oxidation processes, adsorption to activated carbon, biological treatments comprising conventional activated sludge (CAS) and membrane bioreactor (MBR), the latter of which incorporates microfiltration or ultrafiltration (UF), and high pressure separation technologies, such as nanofiltration and reverse osmosis (RO) (Ternes and Joss, 2006). Previous studies comparing the removal efficiencies of several micropollutants by MBR and CAS produced diverse data that showed no definite advantage of one method over the other. For example, Clara et al. (2004) reported almost identical removal efficiencies for ethinylestradiol (60e70%), ibuprofen (90e95%), and galaxolide (60e80%) and showed that the CAS performed 20% better than the MBR for diclofenac (30% and 50% removal rates, respectively). Ternes (1998) found that the CAS removed 69% of diclofenac while Zwiner and Frimmel (2000) observed only 1e5% removal levels for the same drug. Gobel et al. (2007), who tested macrolide and sulfonamide antibiotics removal using MBR and CAS, also obtained diverse results. They measured similar removal rates in the range of 95e100% for erythromycin (ERY), clarithromycin (CLA), and trimethoprim (TMP) in both systems. Removal rates for roxithromycin (ROX), however, were slightly better in the CAS (80% as opposed to 65% by the MBR), while the MBR removed sulfamethoxazole (SMX) significantly better than the CAS (95% vs. 60%, respectively). Gobel et al. (2007) concluded that for the MBR, only a small portion of the removal was caused by sorption (5e10%) while biodegradation, which was strongly influenced by the sludge retention time (SRT), played a major role. Le-Minh et al. (2010) stated that the similarities of SRT and hydraulic retention time (HRT) between CAS and MBR leads to comparable antibiotic removal levels. The papers mentioned above lead to the conclusion that the variance of micro-pollutant removal efficiencies is based on two factors: the first is related to the environmental conditions, including temperature, mixed liquor suspended solids (MLSS) concentration, SRT, raw sewage content, and the microbial community. The second factor is related to the characteristics of each compound, such as contaminant concentration and hydrophobic/hydrophilic nature of the molecule, among others. These operational variables must be

adequately defined to obtain results that will lend themselves to comparison (Le-Minh et al., 2010). This paper compares the OMP removal rates obtained for a full-scale CAS process coupled with a UF pilot with those of a pilot MBR system. The aim of the research was to estimate the antibiotics removal efficiencies of the two pilots and to determine the fates and removal mechanisms of selected antibiotics by running batch adsorption tests using the MBR biomass. Both the MBR and CAS-UF systems are located in the Shafdan wastewater treatment plant (WWTP) in Tel-Aviv, Israel, treating the same raw sewage.

2.

Materials and methods

2.1.

Processes tested

Designed for biological nutrient removal (BNR), the CAS tested in this study was a full-scale, single-sludge system that treats 330,000 m3/day. The UF pilot set up to treat a small portion of the Shafdan’s effluents was based on immersed hollow fibers (ZeeWeed-1000 technology) with 24 modules, and it operated under a stable flux of 45e47 L per meter2 per hour (LMH). During the study period, it was maintained at a relatively low transmembrane pressure of 3e10 kPa (see Table 1). The MBR pilot plant was also constructed as a BNR system designed to remove nitrogen and phosphorus. Its membrane separation section was based on immersed hollow fibers (ZeeWeed-500 technology), which included two membranes (total area of 2 m2), and it was operated with a flux of 10e20 LMH during the study. The influent passed through an 800-mm filter before entering the bioreactor, which contained a 40-L anaerobic zone, a 80-L anoxic zone, and a 120-L aerobic zone. Bioreactor effluent entered the 110-L UF membrane chamber next. MBR operation was divided into two periods: biomass buildup (3.8e10.4 g/L MLSS) for 10 months and stable operation at 10
1 g/L MLSS for 6 months. The MBR was inoculated from the CAS biomass at a starting concentration of 2.5 g/L. During biomass buildup, the flow was gradually increased from 15 to 40 L/h and the transmembrane pressure was maintained between 10 kPa during the summertime and 24 kPa during the winter (without a temperature correction). Nutrient removal was enhanced by circulating 0.5e1 Q (Q represents the influent flow) semi-treated wastewater from

Table 1 e Operational parameters in CAS/UF and MBR. Parameter

HRT [h] MLSS [g/L] SRT [d] Capacity [m3/h] Flux [LMH] Filtration/BW [min] Filtration/BW [m3/h]

CAS/UF

MBR

CAS

UF

Biomass build-up

Stable operation

14e16 2e3 2e4 N.R. N.R. N.R. N.R.

N.R. N.R. N.R. 45-47 45e47 50/1 45/45

9e12 3.8e10.4 >70 0.02e0.04 10e20 5/0.5 (0.04e0.02)/(0.022e0.044)

9 9e11 40 0.04 20 5/0.5 0.04/0.044

N.R. e Not Relevant; BW e Back-wash.

4829

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 2 7 e4 8 3 6

Table 2 e Removals of organic compound and nutrients in the CAS/UF and MBR systems (n [ 98, 62, 36 for CAS/UF, biomass buildup and stable operation, respectively). Parameter

CAS/UF

MBR Biomass build-up

Raw sewage (mg/L) COD NH4eN NO3eN NO2eN PO4eP

925
57 42.3
4.8 e e 10.1
1.8

Stable operation

Effluent (mg/L)

Removal (%)

800 mm filtration (mg/L)

Effluent (mg/L)

Removal (%)

Effluent (mg/L)

Removal (%)


2.5
0.7
0.9
0.4
0.4

96.8
1.5 93.5
3.3 e e 85.1
7.9

435
27 43
5.2 <0.5 <0.5 9.4
1.2

24.6 3.6 15.3 0.83 5.25








94.3
1.5 91.6
7.4 e e 44.1
32.1

19.3
0.43
1.1
0.4
2.8


94.2
2 99.0
0.8 e e 70.2
20.0

27.7 2.9 1.2 0.4 1.1

5.2 2.7 9.2 0.3 2.3

2.3 0.3 0.5 0.2 1.5


Represents the range of concentrations and removal rates.

the outlet of the anoxic zone to the inlet of the anaerobic zone and 4Q from the membrane tank to the inlet of the anoxic zone.

2.2.

Analytical procedures

A comparison of the operational parameters of the CAS/UF with those of the MBR shows that MLSS concentration and SRT were markedly higher in the MBR (Table 1). We expected this finding to enhance significantly OMP biodegradation due to higher variety of bacteria species attributed to increased SRT. In addition, the CAS capacity was more than 1000 times higher than that of the MBR, which, however, was run at a much higher SRT. The relatively high SRT (>70 d) needed to buildup sufficient biomass in the MBR was reduced to 40 d once stable operation conditions were achieved. In addition to the standard test parameters (Table 2) that were evaluated according to standard methods (Standard Methods 2540D and 2540E), two kinds of antibiotics were concentrated by solid-phase extraction (SPE) and analyzed by LC/MS/MS in both Germany (TUB) and Israel (Gush Katif Lab) according to the protocol described by Asmin et al. (2006). The types of antibiotics collected included: macrolides (ERY, ROX, and CLA), sulfonamides [SMX, sulfamethazine (SMZ)], and TMP. The higher macrolide concentrations in the raw sewage of Israel relative to most sulfonamides (except SMX) can be attributed to the higher consumption of the former, which is estimated by the European Surveillance of Antimicrobial Consumption (ESAC) to be three times higher than that for sulfonamides (3 vs. 1 defined daily dose per 1000 inhabitants per day e ESAC, 2005). The correlation between a compound’s concentration and its removal potential depends on the background solution content as described by Yangali et al. (2009). A sludge background, as opposed to a distilled water background, contains a variety of compounds at different concentrations and with different properties. This situation can lead to a matrix effect, which, in turn, has an influence on OMP removal (Ternes and Joss, 2006). The increase in biomass concentration changes the background characteristics constantly, and therefore, it is also expected to change the removal ratio. Each sampling campaign for antibiotics analysis lasted 24 h, during which a composite sample was established by an

automatic sampler set to pump 150 ml once an hour. The purpose of the composite sampling is to avoid the daily fluctuations in the antibiotics concentration. In addition, the samplings were taken only during mid-week (Monday to Wednesday) due to the significant changes in antibiotics concentrations on weekends, as reported by Ternes and Joss (2006). Excess sludge from the MBR was disposed of once a day in the morning. No excess sludge was discarded during the 24-h sampling campaign. The laboratory analysis results showed that the limit of detection (LOD) for sulfonamides and macrolides were 5 and 10 ng/L, respectively. The limit of quantification (LOQ) for sulfonamides and macrolides were 25 and 50 ng/L, respectively. Analytical reproducibility of duplicates was up to 18.7%.

3.

Results and discussion

3.1.

Research approach

This paper is divided into two parts: the first, done to evaluate the antibiotics removal efficiencies of the two pilots, covers MBR and CAS/UF testing as ‘black boxes’ without mass balance by testing the influent and effluent for antibiotics (Section 3.5). It should be noted that the accuracies of mass balance assessments of solutions with low antibiotics concentrations and in which the biomass contains the target compounds is limited. The second part of the paper includes tests with MBR biomass to distinguish between biodegradation and sorption removal mechanisms by applying certain conditions (Sections 3.6e3.7). In this case, mass balance was preserved by increasing the antibiotics concentration significantly (10 mg/L instead of the tens or few hundreds of ng/L typically found in the raw sewage) and by providing conditions that favored sorption to the stabilized MBR biomass over biodegradation. As can be seen in Table 5, desorption of antibiotics during the batch experiment would have had a negligible effect on the results.

3.2.

Removal of organic compounds and nutrients

Both wastewater treatment systems exhibited high rates of organic compound removal and complete nitrification (Table 2). Only partial denitrification and a relatively low rate of

4830

Table 3 e Characteristics of the tested macrolides, sulfonamides, and trimethoprim (Ternes and Joss, 2006). Macrolides

Roxithromycin (ROX)

Clarythromycin (CLA)

Erythromycin (ERY)

C38H69NO13

C41H76N2O15

C37H67NO13

MW, g Log Kow pKa

747.953 2.75 8.8

837.047 3.16 8.9

733.93 3.06 8.9

Sulfonamides and trimethoprim

Sulfamethazine (SMZ)

Sulfamethoxazole (SMX)

Trimethoprim (TMP)

C10H11N3O3S

C14H18N4O3

253.279 0.89 5.6

278.328 0.91 7.2

Structure

Structure

MW, g Log Kow pKa

290.32 0.9 7.4

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 2 7 e4 8 3 6

C12H14N4O2S

4831

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 2 7 e4 8 3 6

Table 4 e MBRs’ permeate concentrations and improvement in antibiotics removal after stabilizing the operation conditions at MLSS of 10 gr/L (n [ 6). Units ERY ROX CLA SMX SMZ TMP

Biomass buildup (ng/L)

Stable operation (ng/L)

Average improvement (%)

Biomass buildup (C/C0)

Stable operation (C/C0)

243.8
90 675.6
420 136.6
43.5 204.4
68.6 >LOQ 17.3
17.3

147.1
45.3 145.6
78.5 95.7
27.3 91.2
38.2 3
3 2.1
2.1

39.6 78.4 29.9 55.4 e 87.8

0.245
0.1 0.445
0.055 0.317
0.028 0.439
0.11 e 0.641
0.078

0.148
0.046 0.139
0.075 0.078
0.022 0.303
0.127 0.138
0.138 0.059
0.059


Represents the range of concentrations.

phosphorus removal were observed in the MBR due to difficulty to preserve anaerobic and anoxic conditions during biomass buildup. Complete denitrification and high phosphorus removal were achieved in the MBR once stable operation conditions were established. The improvement in phosphorus removal is explained by the increase in excess sludge amount (0.5e6 L and 6e8 L during the biomass buildup stage and the stable operation, respectively). The results shown in Table 2 are slightly different from those presented in Sahar et al. (2010), since more up-to-date data is included in the current table.

3.3. Expected removal mechanism based on antibiotics characteristics The main removal mechanisms in biological treatment are biodegradation and sorption, both of which depend on bacteria and on the characteristics of the OMPs (Ternes and Joss, 2006; Le-Minh et al., 2010). Although an increase in biomass concentration may enhance OMP biodegradation, it does not necessarily enhance sorption, due to the dynamic equilibrium in sorption/desorption conditions and the limited number of newly formed binding sites. Another factor that can significantly affect OMP removal is the SRT (which is directly related to biomass concentration), since high SRTs favor the development of diverse bacterial populations, some of which are capable of bioaccumulating and biodegrading more complex organic molecules (Clara et al., 2004; Le-Minh et al., 2010). An increase in SRT causes a reduction in the

Table 5 e Average antibiotics concentrations in the raw sewage and after desorption from MBR MLSS by methanol (n [ 3). Units ERY ROX CLA SMX SMZ TMP

Elusion from MLSS Average concentration with Methanol (ng/g MLSS) in raw sewage (ng/L) 86.5 83.7 95.2 49.3 2.4 0.5









14.2 8.9 11.3 13.7 0.9 0.3

880 1102 1293 367 21 39









558 605 312 153 15 12


Represents the range of the antibiotics concentration load on biomass and concentration, respectively.

food to microorganism (F/M) ratio, thereby reducing the bacteria population growth rate (and excess amounts of sludge). As a result, the concentration of renewable binding sites is decreasing, and therefore, theoretically, the MBR should be less effective at removing OMPs by sorption to sludge. Practically speaking, the process is not that straightforward, and environmental factors such as background fluid content and the contaminant concentration may stimulate significant sorption, as will be explained in Section 3.7. Although all the antibiotics used in the study are originally targeted to destroy bacteria, sulfonamides and macrolides possess different structural characteristics (Table 3) and mechanisms of action. While the macrolides each have one ring with side chains or sugars, every sulfonamide has two relatively small rings connected by a sulfa atom and nitrogen bonds. Their structural differences are expressed in the sulfonamides’ more polar and hydrophilic nature, relative to macrolide characteristics (log Kow < 1 and log Kow > 3, respectively). Ternes and Joss (2006) stated that compounds with log Kow  3 are considered potentially bioaccumulative, and therefore, the main macrolide removal mechanism was hypothesized to be sorption to sludge (bioaccumulation should be more dominant for ERY and CLA, with log Kow of 3.06 and 3.16, respectively, and less for ROX since its log Kow is 2.75). In contrast, we expected sulfonamides to be readily biodegraded due to their hydrophilic nature and relatively low log Kow. Our conjectures about antibiotic removal mechanisms are supported by the different resistance mechanisms bacteria possess against antibiotics; To prevent macrolides from bonding and to thwart their lethal effects, bacteria must be able to methylate their ribosomes and to operate an active protein pump to eject the unwanted molecules from the bacteria cell (Tenson et al., 2003). Sufonamides, on the other hand, can be degraded by enzymatic activity (Bajpai et al., 2000).

3.4. Effect of MLSS concentration on antibiotics removal in the MBR Biomass buildup is an inherent stage in starting or restarting of an MBR. Therefore, it is important to identify MBR capability in removing OMPs during that stage, which can last a few days up to several weeks. The acclimation period required by the MBR to raise its biomass concentration enabled us to characterize the effect of MLSS concentration on removal efficiency. During acclimation, MLSS concentration was increased from

4832

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 2 7 e4 8 3 6

180

SMX

100

160 80

140

TMP

TMP-% removal

60 100 80 40

% removal

SMX-% removal

concentration (ng/L)

120

60 40

20

20 0 3000

4000

5000

6000

7000

8000

9000

0 11000

10000

MLSS concentration (mg/L)

Fig. 1 e Sulfamethoxazole and trimethoprim concentrations and removal rates in MBR treatment during MLSS buildup.

3.8 to 10.4 g/L. At the biomass buildup stage, each sampling campaign was set after a week of stable operation at the desired MLSS concentration. On the one hand, extending the duration at each sampling point would have probably affected the antibiotics removal, due to the changes in the biomass characteristics, but on the other hand, a biomass buildup stage is short, as mentioned before, and hence, the results can indicate the actual removal of the antibiotics in MBR during that stage. Due to the relatively short biomass buildup time, the data can also be referred to as an average (as presented in Fig. 3). Throughout the entire acclimation period, TMP was removed completely. As apposed to that, the concentration of SMX in the MBR effluent gradually decreased from 157 to 71 ng/L (54% improvement), with the increase of MLSS concentration (Fig. 1). The feed concentrations of SMX during

the biomass buildup varied between 184 and 251 ng/L. SMZ was not included in Fig. 1, since its concentrations during the biomass buildup stage were below the LOQ. In contrast to the SMX, the macrolides showed no correlation between the biomass concentration and the antibiotics removal rate (Fig. 2). One can assume that an increase in biomass concentration should increase the OMPs removal, either by biodegradation or by sorption, since more bacteria mass is involved. Nevertheless, it is deemed during the unstable conditions of biomass buildup in the MBR that adsorption and desorption rates change significantly according to the compound’s characteristics, its concentration, and the background fluid content and conditions (matrix), especially in cases of low compound concentrations relative to background compound concentrations. Therefore, the fluctuations in Fig. 2 support the assumption that macrolide

2.0

100

ROX

CLA-% removal ERY-% removal

80 1.5

60 1.0 40

0.5 20

ROX-% removal

0.0 3000

4000

5000

6000

7000

8000

9000

10000

0 11000

MLSS concentration (mg/L)

Fig. 2 e Macrolide concentrations and removal rates in MBR treatment.

% removal

ERY

Normelized concentration (C/C 0)

CLA

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 2 7 e4 8 3 6

MBR (MLSS 2.8-10.6 g/L) RR>99%

TMP

CAS/UF (MLSS 2.3-2.5 g/L)

RR=45% RR=70%

SMX

RR=52% RR=68%

ROX

RR=80% RR=61%

ERY

RR=71% RR=86%

CLA

RR=94%

0

200

400

600

800

1000

Antibiotics concentration (ng/L)

Fig. 3 e Effluent concentrations and removal rates of macrolides and sulfonamides in the CAS/UF and MBR treatments during the MLSS buildup period in the MBR (n [ 11). The error bars represent the range of the antibiotics concentrations.

antibiotics are strongly affected by changes in the background conditions. Due to the variation in the feed concentrations of macrolides (ERY: 327e674 ng/L, ROX: 682e2764 ng/L, CLA: 442e2235 ng/L), the results in Fig. 2 are presented as normalized concentration (C/C0). Despite those variations, the permeate concentration range was relatively low (up to 250 ng/L for ERY and CLA and up to 350 ng/L for ROX). Therefore, it can be hypothesized that in relatively high biomass concentrations (MLSS > 4000 mg/L), OMP removal is quite effective and does not change much with further increases of biomass concentration. An additional step was taken to distinguish between sorption and biodegradation as described in Section 3.6.

3.5.

Comparison between CAS/UF and MBR

The discussion about the MBR is divided according to its main periods of operation: acclimation (biomass growth) and stable operation, the latter occurring when the MLSS concentration remains at approximately 10 g/L. During acclimation, the average MBR removal rate of SMX was 18% better relative to the same rate in the CAS/UF (Fig. 3). For TMP, that advantage was even higher (54%). This can be explained by the relatively high SRT in the MBR, which allows for the enrichment of different bacteria types capable of more efficient antibiotics removal, as claimed by several studies (Clara et al., 2004; Kreuzinger et al., 2004; Ternes and Joss, 2006). SMZ results are not included since in most cases, its concentration was below LOQ. The macrolides, on the other hand, were removed slightly better in the CAS/UF (10e12%), indicating that both SRT and MLSS concentrations did not markedly affect their removal mechanism. It is important to indicate that the production rate of binding sites in MBR and CAS is different by definition (due to different SRT, MLSS concentration, and F/M ratio). In this study, the differences might have been even higher since the COD concentration at the influent to the MBR was half that at the feed point to the CAS (due to the 800

4833

micron filter in the MBR e see Table 2). A relatively low binding site production rate could explain the limited macrolide removal in the MBR during biomass buildup. After the MBR had stabilized and operated steadily for a month at a high MLSS level, the improvement in antibiotics removal was remarkable, as can be seen in Table 4 (from 30% for CLA to around 80% for ROX and almost 90% for TMP). This improvement may be related to the reduction in SRT, which increased the production of new binding sites and enabled the rate of sorption to increase. In addition, the well developed and stable bacteria colonies could have increased biodegradation of the antibiotics. The improvement during the stable operation is also noticeable when the change in feed concentration is eliminated by presenting the normalized concentration (Table 4, columns 5, 6). The stabilized MBR demonstrated a better removal rate of 15e42% relative to the CAS for all of the antibiotics tested (Fig. 4). Each sampling repetition was done after an additional month of stable operation. As mentioned before, MBR nutrient removal capacity also increased dramatically, indicating that the treatment process had stabilized. These results may lead to the conclusion that the higher MLSS (or SRT), which is theoretically associated with the development of a more diverse population of bacteria, caused the improvement in MBR performance compared to CAS. The addition of UF to the CAS system, however, significantly increased (by up to 28%) the removal of all tested antibiotics (Fig. 5). UF filtration of synthetic mixtures containing various OMPs has shown that alone, it is not a significant barrier to these compounds since the UF membrane pore sizes are 100 times larger than the micro-pollutant molecules (Radjenovic et al., 2007). The addition of UF for the secondary effluents had a negligible effect on the MLSS concentration, SRT, and the bacterial composition in the CAS system. Therefore, it is hypothesized that the OMP removal mechanisms enhanced by UF addition were due to the action of the biofilm that formed on the UF membrane surface, which incidentally made the “bio-membrane” a tighter physical (enmeshment) and chemical (sorption) barrier (Yangali et al., 2009; Sahar et al., 2010). UF operation comprised 50 minuets filtration followed by 1 min back-wash (as opposed to 5 min filtration followed by 30 s back-wash in the MBR). This means that biofilm formation (bio-fouling) is a result of the long filtration periods inherent in the daily operation of the UF system. The addition of UF to the CAS system, therefore, reduced the relative advantage the MBR had over the CAS (þUF) to a maximum of 20%. In fact, the two systems exhibited the same removal rate for CLA (94%). Thus, we theorize that the ability of an MBR to remove macrolides and sulfonamides is only slightly better than that of a CAS/UF system. Such a conclusion also indicates that biodegradation is probably not the main removal mechanism in these two systems.

3.6.

Batch adsorption experiments with MBR biomass

A series of batch adsorption experiments were carried out to determine which of the two competing mechanisms, biodegradation or adsorption, was dominant. Preliminary tests using heat (105  C) or toxin (sodium azide) to inhibit the

4834

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 2 7 e4 8 3 6

100

ERY-H2O

80

CLA SMZ SMX TMP

MBR removal (%)

ROX

60

40

20

0 0

20

40

60

80

100

CAS removal (%)

Fig. 4 e Antibiotics removal in MBR (at stable MLSS concentration of 10 g/L) vs. in CAS (n [ 6). X axis and Y axis error bars represent the ranges of antibiotics removal in the CAS and MBR treatments, respectively.

biological activity (as was done by Andersen et al., 2005) conferred upon the biomass characteristics of a passive sorbent (comparable to activated carbon). Therefore, it was decided to use the MBR biomass without further intervention and to favor adsorption over biodegradation by limiting contact time. Once stable operation was achieved, biomass (at a concentration of 10 g/L) was taken from the MBR system. The biomass was divided between four 1-L beakers and diluted with MBR permeate to the following concentrations: 10, 8, 6, and 4 g/L MLSS. The reason for using MBR permeate instead of synthetic effluent, as described by Nyholm et al. (1996) and Andersen et al. (2005), was to minimize the effect of ion concentrations on the bacteria cells. Each beaker was spiked with a synthetic solution of six different antibiotics, each at a concentration of 10 mg/L. The experiments were run for a relatively short time (1 h), during which the beakers were mixed at 100 rpm.

The batch experiment results indicate that sorption, and not biodegradation, was probably the main removal mechanism for both the sulfonamides and for the macrolide antibiotics (Fig. 6), in contrast to studies mentioned by Le-Minh et al. (2010). With the exception of SMX in an MLSS concentration of 4 g/L (64% removal), macrolides and the sulfonamides were removed at efficiencies of >82% at all MLSS levels. No correlation was found between the MLSS concentration and the macrolide removal rate, as was stated in Section 3.4.

3.7.

Desorption experiments with methanol

To verify the sorption hypothesis and identify desorption potential, methanol was used as a solvent to extract sorbed antibiotics from biomass samples taken out from the MBR during steady state operation. Mixed liquor (1 L, 10 g/L MLSS)

100

80

ERY-H2O

CLA SMZ SMX TMP

MBR removal (%)

ROX

60

40

20

0 0

20

40

60

80

100

CAS/UF removal (%)

Fig. 5 e Antibiotics removal in MBR (at stable MLSS concentration of 10 g/L) vs. in CAS/UF (n [ 6). X axis and Y axis error bars represent the ranges of antibiotics removal in the CAS/UF and MBR treatments, respectively.

4835

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 2 7 e4 8 3 6

100 4 gr/L MLSS

8 gr/L MLSS 10 gr/L MLSS

% Removal

6 gr/L MLSS

80 60 40 20 0 ERY

ROX

CLA

SMZ

SMX

TMP

Fig. 6 e Antibiotics removal by MBR sludge in batch adsorption tests (initial concentration of 10 mg/L for each antibiotic, n [ 3). The error bars represent the range of the antibiotics removal.

was centrifuged and separated into its solid and liquid phases. The solid phase was washed three times with tap water (0.1 L each time) to insure that the only source of desorbed antibiotics would be the solid phase, as described by Andersen et al. (2005). The solid phase was added to a 1-L beaker and was filled with methanol which was selected for this task due to its dual hydrophobicehydrophilic nature and its efficient performance in SPE procedures to desorb organic contaminants, such as the selected antibiotics, from cartridge polymers (Ternes and Joss, 2006). Solvent/biomass suspensions were mixed at 100 rpm for 1 h followed by a second solideliquid phase separation by centrifuge. The liquid was analyzed for antibiotics as described in Section 2.2 above. The results in Table 5 demonstrate release of large amounts of desorbed macrolides (83.7e95.2 ng/g MLSS), a medium amount of desorbed SMX (49.3 ng/g MLSS), and low amounts of desorbed SMZ and TMP (2.4e3.9 ng/g MLSS). There is a noticeable correlation between the desorbed compounds and their original concentrations in the raw sewage (except for TMP, which had the lowest desorption rate but a higher concentration than the SMZ in the raw water). These results indicate that sludge is an efficient sorbent for antibiotics, and strengthen the proposed hypothesis regarding the significance of this mechanism in Activated Sludge systems.

Table 6 e Linear correlation between molecular characteristics of sulfonamides and macrolides and their removal rates. Molecular characteristics

pKa M.W. log Kow RR ¼ Removal rate.

Linear correlation value () CAS/UF

MBR

0.64 0.55 0.56

0.48 0.1 0.1

3.8. Effect of compound characteristics on antibiotics removal It is common to estimate a removal rate according to several molecular characteristics such as M.W., pKa, and log Kow. Such studies were done by Yangali et al. (2009) and Comerton et al. (2007). In these papers, the experiments were performed for each compound separately using a distilled water solution under markedly divergent conditions to those in the current study in which pilot plants treated raw sewage. The similarity between macrolides and sulfonamides in terms of M.W. and log Kow (see Table 3) encouraged us to look for a linear correlation. From Table 6, it seems that as the SRT value and MLSS concentration increased, the linear correlation between the molecular characteristics and antibiotics removal decreased. Although the CAS/UF exhibited moderate correlations (0.55e0.56) for M.W. and log Kow, the MBR correlations were significantly lower (0.1).

4.

Conclusions

The efficiency and mechanisms of antibiotics removal from wastewater were tested in CAS/UF and MBR pilot plants treating the same raw sewage of the Tel-Aviv region in Israel. It was found that under stable operation the MBR demonstrated higher removal efficiency over the CAS for all the tested antibiotics. The incorporation of UF after CAS improved significantly the antibiotics removal, thereby reducing the difference between the removal efficiencies of the two systems. It is therefore hypothesized that the biofilm formed on the membrane (in both MBR and UF system) makes it’s separation characteristic tighter related to organic compounds and thus contributes to an enhanced removal of the antibiotics. This may explain the findings indicating better removals of various OMPs in MBR systems compared to CAS systems. However, it means also that the MBR advantage is probably due to the bio-membrane removal mechanism rather then to superior biodegradation.

4836

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 2 7 e4 8 3 6

Batch experiments with MBR biomass showed that the biomass possessed an extremely high sorption potential for all the antibiotics (>82% for sulfonamides and >92% for macrolides at an MLSS concentration of 4, 6, 8, or 10 g/L). This was further demonstrated by methanol extraction, of solidphase biomass taken out of the MBR, which released significant amounts of antibiotics into the liquid phase. It is therefore theorized that sorption mechanism (to both suspended and membrane-attached biomass) is also a significant removal mechanism. These results indicate that greater attention should be devoted to excess sludge management due to its antibiotics accumulating potential.

Acknowledgments This work was supported by research funds from the BMBFMOST Cooperation in Water Technologies Grant WT704 and a scholarship to E. Sahar from the Israel Water Authority. The advice given and correction work done on the manuscript by Christian Kazner, RWTH Aachen, and AVT is greatly appreciated.

references

Andersen, H.R., Hansen, M., Kjolholt, J., Lauridsen, F.S., Ternes, T., Sorensen, B.H., 2005. Assessment of the importance of sorption for steroid estrogen removal during activated sludge treatment. Chemosphere 61, 139e146. Asmin, J., Benner, J., Ernst, M., Fink, J., Hollender, J., Hein, A., Krauss, M., McArdell, C., Reemtsma, T., Ternes, T., 2006. Analytical procedure for antibiotics and their metabolites, organic tracers and selected organic trace pollutants, Deliverable D3.1 of the EU-Project RECLAIM WATER, Water reclamation technologies for safe artificial groundwater recharge, pp. 6e17. Bajpai, A.K., Rajpoot, M., Mishra, D.D., 2000. Studies on the correlation between structure and adsorption of sulfonamides compounds. Colloids and Surfaces A: Physicochemical and Engineering Aspects 168 (3), 193e205. Clara, M., Strenn, B., Ausserleitner, M., Kreuzinger, N., 2004. Comparison of the behavior of selected micro pollutants in a membrane bioreactor and a conventional wastewater treatment plant. Water Science and Technology 50 (5), 29e36. Comerton, A.M., Andrews, R.C., Bagley, D.M., Yang, P., 2007. Membrane adsorption of endocrine disrupting compounds and pharmaceutically active compounds. Journal of Membrane Science 303, 267e277. Damstra, T., Barlow, S., Bergman, A., Kavlock, R., Van-der Kraak, G., 2002. REPIDISCA-Global Assessment of the State-ofthe-science of Endocrine Disruptors. International Programme on Chemical Safety. World Health Organization. Daughton, C.G., Ternes, T.A., 1999. Pharmaceuticals and personal care products in the environment: agents of subtle change? Environmental Health Perspectives 107 (6), 907e938. ESAC, 2005. European Surveillance Of Antimicrobial Consumption (ESAC): outpatient antibiotic use in Europe, 1998e2005. Euro Surveillance 12 (12), 306e309. Gobel, A., McArdell, C.S., Joss, A., Siegrist, H., Giger, W., 2007. Behavior of sulfonamides, macrolides and trimethoprim in

different wastewater treatment technologies. Science of the Total Environment 372, 361e371. Heberer, Th., Reddersen, K., Mechlinski, A., 2002. From municipal sewage to drinking water: fate and removal of pharmaceutical residues in the aquatic environment in urban areas. Water Science Technology 46 (3), 81e88. Khan, S.J., Wintgens, T., Sherman, P., Zaricky, J., Scha¨fer, A.L., 2004. Removal of hormones and pharmaceuticals in the advanced water recycling demonstration plant in Queensland, Australia. Water Science and Technology 50 (5), 15e22. Kolpin, D.W., Furlong, E.T., Meyer, M.T., Thurman, E.M., Zaugg, S. D., Barber, L.B., Buxton, H.B., 2001. Pharmaceuticals, hormones, and other organic wastewater substances in U.S. streams, 1999e2000: a national reconnaissance. Environmental Science and Technology 36 (6), 1202e1211. Kreuzinger, N., Clara, M., Strenn, B., Kroiss, H., 2004. Relevance of the sludge retention time (SRT) as a design criteria for wastewater treatment plants for the removal of endocrine disruptors and pharmaceuticals from wastewater. Water Science and Technology 50 (5), 149e156. Le-Minh, N., Khan, S.J., Drewes, J.E., Stuetz, R.M., 2010. Fate of antibiotics during municipal water recycling treatment processes. Water Research 44 (15), 295e323. Nghiem, D.L., Scha¨fer, A., 2005. Critical risk points of nanofiltration and reverse osmosis processes in water recycling applications. Desalination 187, 303e312. Nyholm, N., Berg, U., Ingerslev, F., 1996. Environmental Project No. 337 e Activated Sludge Biodegradability Simulation Test Environmental Project 337, Milijostyrelsen, Copenhagen, Denmark. Paxeus, O., 2004. Removal of selected non-steroidal antiinflammatory drugs (NSAIDs), gemfibrozil, carbamazepine, b-blockers, trimethoprim, and triclosan in conventional wastewater treatment plants in EU countries and their discharge to the aquatic environment. Water Science and Technology 50 (5), 253e260. Radjenovic, J., Petrovic, M., Barcelo, D., 2007. Analysis of pharmaceuticals in wastewater and removal using membrane bioreactor. Analytical and Bioanalytical Chemistry 387, 1365e1377. Sahar, E., Ernst, M., Godehardt, M., Hein, A., Herr, J., Kazner, C., Melin, T., Cikurel, H., Aharoni, A., Messalem, R., Brenner, A., Jekel, M., 2010. Comparison of two treatments for the removal of selected organic micropollutants and bulk organic matter: conventional activated sludge (CAS) followed by ultrafiltration (UF) vs. membrane bioreactor (MBR). Water Science and Technology 63 (4), 733e740. Standard Methods 2540D current ed., Total Suspended Solids Dried at 103e105 C. Standard Methods 2540E current ed., Fixed and Volatile Solids Ignited at 550  C. Tenson, T., Lovmar, M., Ehrenberg, M., 2003. The mechanism of action of macrolides, lincosamides and streptogramin B reveals the nascent peptide exit path in the ribosome. Journal of Molecular Biology 330, 1005e1014. Ternes, T.A., 1998. Occurrence of drugs in German sewage treatment plants and rivers. Water Research 32 (11), 3245e3260. Ternes, T.A., Joss, A., 2006. Human Pharmaceuticals, Hormones and Fragrances: the Challenge of Micropollutants in Urban Water Management. IWA, London. Yangali, Q.V., Fujioka, T., Kennedy, M., Amy, G., 2009. Is nanofiltration a robust barrier for organic micro pollutants?. In: Proceedings of the IWA Membrane Technology Conference & Exhibition 2009, 1e3 September 2009, Beijing, China. Zwiner, C., Frimmel, F., 2000. Oxidative treatment of pharmaceuticals in water. Water Research 34 (6), 1881e1885.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 3 7 e4 8 4 3

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Detection of microsporidia in drinking water, wastewater and recreational rivers Fernando Izquierdo a, Jose´ Antonio Castro Hermida b, Soledad Fenoy a, Mercedes Mezo b, Marta Gonza´lez-Warleta b, Carmen del Aguila a,* a

Universidad San Pablo CEU, Laboratorio de Parasitologı´a, Facultad de Farmacia, Urbanizacio´n Monteprı´ncipe, CP 28668 Boadilla del Monte, Madrid, Spain b Centro de Investigaciones Agrarias de Mabegondo, Laboratorio de Parasitologı´a, Instituto Galego de Calidade Alimentaria-Xunta de Galicia, Carretera AC-542 de Betanzos a Meso´n do Vento, Km 7.5, CP 15318 Abegondo (A Corun˜a), Spain

article info

abstract

Article history:

Diarrhea is the main health problem caused by human-related microsporidia, and water-

Received 26 January 2011

borne transmission is one of the main risk factors for intestinal diseases. Recent studies

Received in revised form

suggest the involvement of water in the epidemiology of human microsporidiosis.

22 June 2011

However, studies related to the presence of microsporidia in different types of waters from

Accepted 22 June 2011

countries where human microsporidiosis has been described are still scarce. Thirty-eight

Available online 2 July 2011

water samples from 8 drinking water treatment plants (DWTPs), 8 wastewater treatment plants (WWTPs) and 6 recreational river areas (RRAs) from Galicia (NW Spain) have been

Keywords:

analyzed. One hundred liters of water from DWTPs and 50 L of water from WWTPs and

Drinking water treatment

RRAs were filtered to recover parasites, using the IDEXX Filta-Max
system.

plant (DWTP)

Microsporidian spores were identified by Weber’s stain and positive samples were

Wastewater treatment

analyzed by PCR, using specific primers for Enterocytozoon bieneusi, Encephalitozoon intesti-

plant (WWTP)

nalis, Encephalitozoon cuniculi, and Encephalitozoon hellem. Microsporidia spores were identi-

Recreational river area (RRA)

fied by staining protocols in eight samples (21.0%): 2 from DWTPs, 5 from WWTPs, and 1

IDEXX Filta-Max

from an RRA. In the RRA sample, the microsporidia were identified as E. intestinalis.

Microsporidia Encephalitozoon intestinalis

To the best of our knowledge, this is the first report of human-pathogenic microsporidia in water samples from DWTPs, WWTPs and RRAs in Spain. These observations add further evidence to support that new and appropriate control and regulations for drinking, wastewater, and recreational waters should be established to avoid health risks from this pathogen. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Waterborne transmission is one of the main risk factors for intestinal diseases causing an important morbidity and mortality worldwide. However, it is surprising that even though known agents that produce intestinal disease, such as

Giardia and Cryptosporidium, are frequently transmitted by water (Graczyk et al., 2007b), over 50% of waterborne infections are produced by unknown agents (Dowd et al., 1998). This finding is of special interest if we bear in mind that, for economic and environmental reasons, spreading sewage sludge on agricultural lands has increased during recent years

* Corresponding author. Tel.: þ34 91 372 47 96/84; fax: þ34 91 351 04 96. E-mail addresses: [email protected] (F. Izquierdo), [email protected] (J.A. Castro Hermida), [email protected] (S. Fenoy), [email protected] (M. Mezo), [email protected] (M. Gonza´lez-Warleta), [email protected] (C.del Aguila). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.06.033

4838

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 3 7 e4 8 4 3

(Rimhanen-Finne et al., 2004). This might affect not only the circulation of recognized pathogens such as Cryptosporidium and Giardia, but also emerging pathogens, such as microsporidia. Moreover, the results obtained in different studies carried out to establish the quality of depuration end products associated with the presence of parasites seem to be contradictory (Straub et al., 1993; Wiandt et al., 2000; Caccio et al., 2003; Graczyk et al., 2007a). The general impression is that treatment to obtain sewage sludge end products has demonstrated a high efficacy of pathogen removal. However, as viable pathogens have been detected in these end products, they could be considered a serious health threat (Graczyk et al., 2007a). On the other hand, it is important to understand that the presence of human pathogens in surface water may suggest the presence of living environmental reservoirs, such as domestic and wild animals. Among the latter, aquatic birds may play an important role in the transmission of different pathogens (Slodkowicz-Kowalska et al., 2006). Microsporidia are obligate intracellular eukaryote pathogens that may cause infection in both vertebrate and invertebrate hosts. Diarrhea is the most frequent health problem caused, mainly in immunocompromised people. The transmission routes indicated are via airborne, person-to-person, zoonotic, and waterborne means (Didier et al., 2004; Graczyk et al., 2007c). Waterborne transmission of microsporidian spores has not yet been appropriately addressed in epidemiological studies, due to the small size of spores (1e4 m) (Mathis et al., 2005). Their presence, associated with waterborne outbreaks and also with recreational and river water, has rarely been documented (Sparfel et al., 1997; Dowd et al., 1998, 2003; Cotte et al., 1999; Fournier et al., 2000; Thurston-Enriquez et al., 2002; Coupe et al., 2006; Graczyk et al., 2007b, 2007c; Lucy et al., 2008). On the other hand, the demonstration of waterborne microsporidian spores of species known to infect humans, proceeding from common waterfowl which have unlimited access to surface waters, has only recently been documented (Slodkowicz-Kowalska et al., 2006). In spite of this, microsporidia are recognized category B biodefense agents on the National Institutes of Health list, and the transmission of microsporidian spores is seriously considered by American agencies concerned with the quality of drinking water (Nwachcuku and Gerba, 2004). These pathogens have been included in the Contaminant Candidate List of the U.S. Environmental Protection Agency ((EPA), 1998) because spore identification, removal, and inactivation in drinking water are technologically challenging, and human microsporidial infections are difficult to treat (SlodkowiczKowalska et al., 2006; Graczyk et al., 2007b). In Europe, the regulation related to the quality of sanitary water for human consumption is adapted from Directive 98/ 83/EEC (Communities, 1998), which specifies the need to detect fecal bacterial indicators and also establishes a water turbidity limit to determine the presence of Cryptosporidium or other microorganisms and parasites, when considered appropriate by authorities. However, microsporidia are not specifically monitored. The quality of bathing water and the use of sewage sludge in agriculture are governed by Directives 76/160/EEC

(Community, 1976) and 86/278/EEC, respectively (Community, 1986). However, parasites are not covered by these directives, so microsporidia are not routinely monitored. Finally, the use of regenerated water has recently been regulated in our country, (R.D 1620/2007). However, although Giardia, Cryptosporidium, and helminth eggs are included, there is no mention of the search for microsporidia. Considering that this type of water is planned for use in urban, agricultural, industrial, recreational, and environmental practices, this may represent a sanitary risk for users. The present work studies, for the first time, the presence of microsporidia in different types of water in Spain.

2.

Materials and methods

2.1.

Water sampling

Thirty-eight water samples from 8 drinking water treatment plants (DWTPs), 8 wastewater treatment plants (WWTPs) and 6 water samples from recreational river areas (RRAs) from Galicia (NW Spain) were analyzed (Fig. 1). The water treatment carried out in all the DWTPs included coagulation, flocculation, and clarification through sedimentation, filtration, and disinfection by chlorination. Neither UV treatment nor ozonation was carried out in any of DWTPs included in the study. The main processes in the selected WWTPs consisted of a primary treatment (screening, storage and preconditioning) and a secondary treatment (coagulation and flocculation, sedimentation and decantation). A tertiary treatment (UV or ozone) was not carried out. All water sampling areas were located in areas with high livestock (cattle and sheep) activity, predominantly cattle farming. One hundred liters of water from DWTPs and 50 L from WWTPs and RRAs were collected. In all cases, water samples were concentrated by the IDEXX Filta-Max
system for the capture and recovery of Cryptosporidium sp and Giardia sp., following the 1623 method used by the United States Environmental Protection Agency (USEPA) (U.S.E.P.A., 2005). Water was collected using a portable water pump connected to a foam filter module, following the manufacturer’s instructions and USEPA 1623 (U.S.E.P.A., 2005). Organisms were recovered by elution in a final volume of 5 ml.

2.2.

Staining methods

Thin smears from concentrated water samples were prepared and stained, using Weber’s chromotrope-based stain (Weber et al., 1992) to detect microsporidia. However, this method cannot determine the species of microsporidia.

2.3.

DNA extraction and purification

To determine the microsporidian species, DNA from unpreserved concentrated water samples was extracted following the methods described earlier (del Aguila et al., 1997a). DNA was extracted by bead disruption of spores using the FastDNA-Spin kit, according to the manufacturer’s instructions (Bio 101, Carlsbad, Calif.). PCR inhibitors were removed using the QIAquick PCR kit (QIAGEN, Chatsworth, CA).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 3 7 e4 8 4 3

4839

Fig. 1 e Geographical location of the sampling points in relation to the 11 municipalities in Galicia (NW Spain), where the following water samples were obtained: recreational river areas (RRAs; no. [ 6); influent and final effluent from drinking water treatments plants (DWTPs; no. [ 8) and wastewater treatment plants (WWTPs; no. [ 8). * Locations where microsporidia were detected.

2.4.

PCR amplification

Microsporidial-small subunit rRNA (SSU-rRNA) coding regions were amplified, using the following species-specific primers: EBIEF1/EBIER1 for Enterocytozoon.bieneusi (Da Silva et al., 1996), SINTF/SINTR for Encephalitozoon intestinalis (Da Silva et al., 1997), EHELF/EHELR for Encephalitozoon hellem (Visvesvara et al., 1994), and ECUNF/ECUNR for Encephalitozoon cuniculi (De Groote et al., 1995). The PCR amplification was carried out with the GenAmp kit (PerkineElmer Cetus, Norwalk, CT), according to manufacturer’s procedures and the conditions for the reaction described previously (Da Silva et al., 1997). Purified samples were tested for the presence of PCR inhibitors, as described previously (Da Silva et al., 1997). Amplification products were analyzed by 2% agarose gel electrophoresis and visualized by ethidium bromide staining (Da Silva et al., 1997).

ovoid and ranged from 0.9 to 1.6 mm (Fig. 2; Table 1). In only one of the positive treatment plants, microsporidian spores were detected solely in the final effluent. It was in the WWTP of Municipality No.6 (Fig. 1, Table 2). On the other hand, in Municipality No. 8, microsporidian spores were detected in the influent water of both DWTPs and WWTPs studied (Fig. 1). From all RRAs, only one case of microsporidial contamination was detected by Trichrome stain (Municipality No. 7). DNA amplification of positive samples in the staining technique, with specific primers for the four most common microsporidia infecting humans, allowed us to confirm the presence of microsporidian species in the water sample from an RRA (Municipality No. 7). The microsporidia were identified as E. intestinalis, showing the diagnostic band of 528 bp in the agarose gels. No positive samples for E. bieneusi, E. cuniculi, or E. hellem were detected. No PCR inhibitors were detected (Fig. 2).

4. 3.

Discussion

Results

Eight samples (21.0%) out of 38 water samples (2 from DWTPs, 5 from WWTPs and 1 from an RRA) showed a variable number of spores that stained pinkish red when the Weber’s stain was used. The characteristic morphology of microsporidian spores with a clear vacuole-like polar end was observed; they were

Microsporidia, are emerging pathogens related to diarrhea in both immunocompetent and immunosuppressed patients, which in the last few years, have been recognized as water contaminants, but their small size makes their detection difficult. Using Weber’s stain, microsporidia spores were detected in 8 water samples (21.0%): 2 samples from DWTPs, 5

4840

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 3 7 e4 8 4 3

Fig. 2 e A: Microsporidia spores stained with modified trichrome stain proceeding from RRA (Municipality No 7). B: PCR amplification from RRA (Municipality No 7) using specific primers for E. intestinalis, M: 100 pb DNA ladder. Lane 1: DNA extracted. Lane 2: DNA 1/10 dilution. Lane 3: DNA 1/100 dilution. Lane 4: positive control. Lane 5: negative control. RRA: Recreational River Area.

from WWTPs and 1 from an RRA. The species E. intestinalis could be identified by PCR methods only in the sample proceeding from the RRA. It is necessary to point out that, although the PCR technique is the most sensitive method for species identification, the main problem involved is the appearance of false-negative results, due to a low parasite DNA concentration, and the presence of PCR inhibitors (Da Silva et al., 1997). In the case of microsporidia, the presence of extruded spores (non viable spores with no DNA) in samples may be one additional reason for a low parasitic DNA concentration, possibly influenced by treatments of DWTPs and WWTPs. Finally, it is important to consider that the water samples that tested positive with the staining methods may not necessarily amplify with the specific primers used, due to the presence of microsporidia other than the species studied. To date, no agreement in the methods used to concentrate microsporidia from water samples has been reached (Sparfel et al., 1997; Fournier et al., 2000; Thurston-Enriquez et al., 2002; Li et al., 2003; Hoffman et al., 2007; Kwakye-Nuako et al., 2007). To our knowledge, in only one previous study, IDEXX Filta-Max
was used for the concentration step, although the system was considered unsuitable for detecting microsporidia, based on the scarce recovery percentage (Stine et al., 2005). However, the level of detection obtained in our study (20.1% of samples) would be sufficient to include it among techniques useful in detecting microsporidia. Studies in

Table 1 e Results obtained by Trichrome stain and PCR from drinking water treatment plants (DWTPs), wastewater treatment plants (WWTPs) and recreational river areas (RRAs) in the municipalities included in the study. No: number of samples analyzed. Type of Water (No.)

Positive Samples (%) Trichrome stain

DWTP (16) WWTP (16) RRA (6) Total (38)

2 5 1 8

(12.5%) (31.2%) (16.6%) (21.0%)

PCR e e E. intestinalis E. intestinalis

different types of water have shown the presence of microsporidia such as E. bieneusi, E. intestinalis, E. hellem, Vittaforma corneae, and Pleistophora, affecting humans (Avery and Undeen, 1987; Dowd et al., 1998, 2003; Fournier et al., 2000; Thurston-Enriquez et al., 2002; Graczyk et al., 2007a, 2007b; Lucy et al., 2008). To the best of our knowledge, our results are the first report of human-pathogenic microsporidia in water samples from Spain. It is important to bear in mind that in our country microsporidia have been related to human diarrhea in HIV positive (1.2e13.9%) and negative patients (5.1e17.02%) (Subirats et al., 1996 del Aguila et al., 1997b; Gainzarain et al., 1998; Lo´pez-Ve´lez et al., 1999; Lores et al., 1999, 2002a, 2002b; Abreu-Acosta et al., 2005); and 5.4% of blood-donors showed seropositivity for Encephalitozoon sp. (del Aguila et al., 2001). Additionally, human-related microsporidia have been identified in a high percentage (20.9%) in pigeons from urban parks (Haro et al., 2005), reinforcing the convenience of studies to discern the implication of waterborne transmission in the epidemiology of these parasites. A low contamination by microsporidia in DWTPs (only two cases) was detected, compared with that shown in WWTPs (5 cases). The contamination detected in the DWTPs was only found in the influent water but not in the final effluent. Although the number of DWTPs positive for microsporidia was low, the absence of this parasite in the final effluent in all cases would suggest that the treatment used effected its removal. To date, there are no similar studies on DWTPs, although E. intestinalis have been demonstrated in drinking waters (Dowd et al., 2003). In one of the positive samples from WWTPs, microsporidian spores were detected in the final effluent. Previous studies have shown the presence of human microsporidia in a tertiary effluent (Dowd et al., 1998) or in sewage sludge end products or wetland outfalls (Graczyk et al., 2007a), which may be explained because they are potentially resistant to disinfection (Dowd et al., 1998) or because these parasites would be propagated by dogs, livestock and visiting wildlife (Graczyk et al., 2009a). Only one of the microsporidia-positive samples from RRAs could be confirmed as E. intestinalis by PCR analysis. Previous PCR studies have shown the presence of human-pathogenic

4841

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 3 7 e4 8 4 3

Table 2 e Results obtained by Trichrome stain and PCR of microsporidia in the influent and effluent samples from drinking water treatment plants (DWTPs) and wastewater treatment plants (WWTPs) in the municipalities included in the study. No.: number of samples analyzed. Plant (No.)

DWTP (8) WWTP (8)

Influent

Effluent

Positive samples (%)

Trichrome stain

PCR

Positive samples (%)

Trichrome stain

PCR

2 (25.0%) 4 (50.0%)

2 4

0 0

0 1 (12.5%)

0 1

0 0

microsporidia, such as E. bieneusi and E. intestinalis, in surface and recreational waters (Sparfel et al., 1997; Dowd et al., 1998, 2003; Fournier et al., 2000; Coupe et al., 2006; Graczyk et al., 2007c; Lucy et al., 2008). The pathways of microsporidia infections, modes, or routes of transmission, and the knowledge of the epidemiology are still uncertain, although recent studies point to a zoonotic origin (Didier et al., 2004; Haro et al., 2005). The fact that wildlife that inhabits or visits rivers or wetland systems can significantly contribute to humanpathogenic microsporidia was previously suggested (Graczyk et al., 2007c, 2009b). The presence of E. intestinalis in an RRA reinforces this idea, since wild animals, including waterfowl, have unlimited access to surface waters of the area under study (Slodkowicz-Kowalska et al., 2006; Graczyk et al., 2009b). It is notable that Municipalities No. 6, 7, and 8 have a very high livestock density, mainly of bovine origin, with a density of cattle population double that of the human population in those areas. Although the livestock have no access close to river water manure is frequently washed away from these areas along well-defined drainage paths during rainfall events, and the cows typically have free access to nearby streams. In this scenario, both livestock manure and grazing cattle may contribute to contamination of the rivers. Although there are no data on microsporidial infection of bovines in Spain percentages between 13% and 15% have been described in United States and Korea (Santin et al., 2004; Lee, 2007). In our study, we could not establish the viability of microsporidian spores detected. However, in previous studies, the viability of microsporidia spores after water treatments has been demonstrated. In a study on water proceeding from sewage sludge end products, Graczyk et al. (2007a) observed that most spores identified were potentially viable using fluorescent in situ hybridization method (FISH). This method indicated that the viability of microsporidia should be considered even though the parasitic load is low, taking into account their long-term environmental survival, and their serious implications in human health (Weber et al., 1994). Therefore, studies to discern the possible source-tracking of contamination and the viability of these parasites after treatments are necessary, as the water obtained in WWTPs could be discharged into a river or be used for urban, agricultural, industrial, recreational, or environmental practices and might contribute to the contamination of the environment with the consequent risk to human health. In addition, we must bear in mind that the ID50 for microsporidia in humans is still unknown. However, previous reports indicated that in animals the minimal infectious dose is very low (Graczyk et al., 2010). Our results are the first report on human-related microsporidia in different kind of water from Spain, and warn of the

possibility that exposure to recreational waters could play a role in the epidemiology of human microsporidiosis. Nevertheless, more complete epidemiology studies are needed to understand the origin and the contribution of microsporidia water contamination to human diarrhea.

5.

Conclusions

This study shows the presence of human-related microsporidia in water samples, highlighting the potential role of water in microsporidiosis epidemiology. The difficulties observed for the microsporidia species determination in these kinds of samples have made us aware of the need for the development and standardization of good laboratory methods for an easier and more accurate detection of microsporidia in water samples. This is a necessary first step that would contribute to the development of a monitoring programme to carry out source-tracking, risk assessment and linked epidemiology studies to better understand these pathogens.

Acknowledgments We are grateful to L. Hamalainen for help in the preparation of the manuscript. This work was supported by the Ministerio de Ciencia e Innovacio´n, within the Programa Nacional de Recursos y Tecnologı´as Agroalimentarias (RTA2010-00003-0000) and by grants from the Fundacio´n San Pablo-CEU 03/08.

references

(EPA), 1998. U.E.P.A Announcement of the drinking water contaminant candidate list: notice. Fed. Regist. 63, 10272e10287. Abreu-Acosta, N., Lorenzo-Morales, J., Leal-Guio, Y., CoronadoAlvarez, N., Foronda, P., Alcoba-Florez, J., Izquierdo-Dı´az, N., del Aguila, C., Valladares, B., 2005. Enterocytozoon bieneusi (microsporidia) in clinical samples from immunocompetent individuals in Tenerife, Canary Islands, Spain. Trans. R Soc. Trop. Med. Hyg. 99, 848e855. Avery, S.W., Undeen, A.H., 1987. The isolation of microsporidia and other pathogens from concentrated ditch water. J. Am. Mosq Control Assoc. 3, 54e58. Caccio, S.M., De Giacomo, M., Aulicino, F.A., Pozio, E., 2003. Giardia cysts in wastewater treatment plants in Italy. Appl. Environ. Microbiol. 69, 3393e3398. Communities, E., 1998. Council Directive 98/83/EC on the quality of water intended for human consumption. Off. J. Eur. Communities 330, 32e54.

4842

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 3 7 e4 8 4 3

Community, E., 1976. Council Directive concerning the quality of bathing water. Off. J. Eur. Communities, 1e9. Community, E., 1986. On the protection of the environment, and in particular of the soil when sewage sludge is used in agriculture. Off. J. Eur. Communities, 1e13. Cotte, L., Rabodonirina, M., Chapuis, F., Bailly, F., Bissuel, F., Raynal, C., Gelas, P., Persat, F., Piens, M.A., Trepo, C., 1999. Waterborne outbreak of intestinal microsporidiosis in persons with and without human immunodeficiency virus infection. J. Infect. Dis. 180, 2003e2008. Coupe, S., Delabre, K., Pouillot, R., Houdart, S., SantillanaHayat, M., Derouin, F., 2006. Detection of Cryptosporidium, Giardia and Enterocytozoon bieneusi in surface water, including recreational areas: a one-year prospective study. FEMS Immunol. Med. Microbiol. 47, 351e359. Da Silva, A.J., Schwartz, D.A., Visvesvara, G.S., de Moura, H., Slemenda, S.B., Pieniazek, N.J., 1996. Sensitive PCR diagnosis of infections by Enterocytozoon bieneusi (microsporidia) using primers based on the region coding for small-subunit rRNA. J. Clin. Microbiol. 34, 986e987. Da Silva, A.J., Slemenda, S.B., Visvesvara, G.S., Schwartz, D.A., Wilcox, C.M., Wallace, S., Pieniazek, N.J., 1997. Detection of Septata intestinalis (Microsporidia) Cali et al. 1993 using polymerase chain reaction primers targeting the small submit subunit ribosomal RNA coding region. Mol. Diagn. 2, 47e52. De Groote, M.A., Visvesvara, G., Wilson, M.L., Pieniazek, N.J., Slemenda, S.B., daSilva, A.J., Leitch, G.J., Bryan, R.T., Reves, R., 1995. Polymerase chain reaction and culture confirmation of disseminated Encephalitozoon cuniculi in a patient with AIDS: successful therapy with albendazole. J. Infect. Dis. 171, 1375e1378. del Aguila, C., Lopez-Velez, R., Fenoy, S., Turrientes, C., Cobo, J., Navajas, R., Visvesvara, G.S., Croppo, G.P., Da Silva, A.J., Pieniazek, N.J., 1997a. Identification of Enterocytozoon bieneusi spores in respiratory samples from an AIDS patient with a 2year history of intestinal microsporidiosis. J. Clin. Microbiol. 35, 1862e1866. del Aguila, C., Navajas, R., Guribindo, D., Ramos, J.T., Mellado, M.J., Fenoy, S., Mun˜oz-Fernandez, M.A., Subirats, M., Pieniazek, N.J., 1997b. Microsporidiosis in HIV-positive children in Madrid (Spain). J. Eukaryot. Microbiol. 44, 84Se85S. del Aguila, C., Rueda, C., de la Ca´mara, C., Fenoy, S., 2001. Seroprevalence of anti-Encephalitozoon antibodies in spanish immunocompetent subjects. J Eukaryot. Microbiol., 75Se78S. Didier, E.S., Stovall, M.E., Green, L.C., Brindley, P.J., Sestak, K., Didier, P.J., 2004. Epidemiology of microsporidiosis: sources and modes of transmission. Vet. Parasitol. 126, 145e166. Dowd, S.E., Gerba, C.P., Pepper, I.L., 1998. Confirmation of the human-pathogenic microsporidia Enterocytozoon bieneusi, Encephalitozoon intestinalis, and Vittaforma corneae in water. Appl. Environ. Microbiol. 64, 3332e3335. Dowd, S.E., John, D., Eliopolus, J., Gerba, C.P., Naranjo, J., Klein, R., Lopez, B., de Mejia, M., Mendoza, C.E., Pepper, I.L., 2003. Confirmed detection of Cyclospora cayetanesis, Encephalitozoon intestinalis and Cryptosporidium parvum in water used for drinking. J. Water Health 1, 117e123. Fournier, S., Liguory, O., Santillana-Hayat, M., Guillot, E., Sarfati, C., Dumoutier, N., Molina, J., Derouin, F., 2000. Detection of microsporidia in surface water: a one-year follow-up study. FEMS Immunol. Med. Microbiol. 29, 95e100. Gainzarain, J.C., Canut, A., Lozano, M., Labora, A., Carreras, F., Fenoy, S., Navajas, R., Pieniazek, N.J., da Silva, A.J., del Aguila, C., 1998. Detection of Enterocytozoon bieneusi in two human immunodeficiency virus-negative patients with chronic diarrhea by polymerase chain reaction in duodenal biopsy speciment. Clin. Infect. Dis. 21, 392e398. Graczyk, T.K., Lucy, F.E., Tamang, L., Miraflor, A., 2007a. Human enteropathogen load in activated sewage sludge and

corresponding sewage sludge end products. Appl. Environ. Microbiol. 73, 2013e2015. Graczyk, T.K., Sunderland, D., Tamang, L., Lucy, F.E., Breysse, P.N., 2007b. Bather density and levels of Cryptosporidium, Giardia, and pathogenic microsporidian spores in recreational bathing water. Parasitol. Res. 101, 1729e1731. Graczyk, T.K., Sunderland, D., Tamang, L., Shields, T.M., Lucy, F.E., Breysse, P.N., 2007c. Quantitative evaluation of the impact of bather density on levels of human-virulent microsporidian spores in recreational water. Appl. Environ. Microbiol. 73, 4095e4099. Graczyk, T.K., Lucy, F.E., Mashinsky, Y., Andrew Thompson, R.C., Koru, O., Dasilva, A.J., 2009a. Human zoonotic enteropathogens in a constructed free-surface flow wetland. Parasitol. Res. 105, 423e428. Graczyk, T.K., Lucy, F.E., Tamang, L., Mashinski, Y., Broaders, M.A., Connolly, M., Cheng, H.W., 2009b. Propagation of human enteropathogens in constructed horizontal wetlands used for tertiary wastewater treatment. Appl. Environ. Microbiol. 75, 4531e4538. Graczyk, T.K., Sunderland, D., Awantang, G.N., Mashinski, Y., Lucy, F.E., Graczyk, Z., Chomicz, L., Breysse, P.N., 2010. Relationships among bather density, levels of human waterborne pathogens, and fecal coliform counts in marine recreational beach water. Parasitol. Res. 106, 1103e1108. Haro, M., Izquierdo, F., Henriques-Gil, N., Andres, I., Alonso, F., Fenoy, S., del Aguila, C., 2005. First detection and genotyping of human-associated microsporidia in pigeons from urban parks. Appl. Environ. Microbiol. 71, 3153e3157. Hoffman, R.M., Wolk, D.M., Spencer, S.K., Borchardt, M.A., 2007. Development of a method for the detection of waterborne microsporidia. J. Microbiol. Methods 70, 312e318. Kwakye-Nuako, G., Borketey, P., Mensah-Attipoe, I., Asmah, R., Ayeh-Kumi, P., 2007. Sachet drinking water in accra: the potential threats of transmission of enteric pathogenic protozoan organisms. Ghana Med. J. 41, 62e67. Lo´pez-Ve´lez, R., Turrientes, M.C., Garro´n, C., Montilla, P., Navajas, R., Fenoy, S., del Aguila, C., 1999. Microsporidiosis in travelers with diarrhea from the tropic. J. Travel Med. 6, 223e227. Lee, J.H., 2007. Prevalence and molecular characteristics of Enterocytozoon bieneusi in cattle in Korea. Parasitol. Res. 101, 391e396. Li, X., Palmer, R., Trout, J.M., Fayer, R., 2003. Infectivity of microsporidia spores stored in water at environmental temperatures. J. Parasitol. 89, 185e188. Lores, B., Arias, C., Fenoy, S., Iglesias, I., Ocampo, A., Miralles, C., Garcı´a Estevez, J.M., del Aguila, C., 1999. Enterozytozoon bieneusi: a common opportunistic parasite lungs of HIVpositive patients? J. Eukaryot. Microbiol. 46, 6Se7S. Lores, B., del Aguila, C., Lo´pez-Miragaya, I., Torres, J., Arias, C., 2002a. Detection of Enterocytozoon bieneusi by PCR in immunocompetent Spanish patients with diarrhea and pneumonia. Rev. Iber Parasitol. 62, 43e49. Lores, B., Lo´pez-Miragaya, I., Arias, C., Fenoy, S., Torres, J., del Aguila, C., 2002b. Intestinal microsporidiosis due to Enterocytozoon bieneusi in elderly human immunodeficiency virus-negative patients Vigo, Spain. Clin. Infect. Dis. 34, 918e921. Lucy, F.E., Graczyk, T.K., Minchin, D., Tamang, L., Miraflor, A., 2008. Biomonitoring of surface and coastal water for Cryptosporidium, Giardia and human virulent microsporidia using molluscan shellfish. Parasitol. Res. 103, 1369e1375. Mathis, A., Weber, R., Deplazes, P., 2005. Zoonotic potential of the microsporidia. Clin. Microbiol. Rev. 18, 423e445. Nwachcuku, N., Gerba, C.P., 2004. Emerging waterborne pathogens: can we kill them all? Curr. Opin. Biotechnol. 15, 175e180.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 3 7 e4 8 4 3

R.D, 2007. Reutilizacio´n de aguas depuradas. Boletı´n Oficial del Estado 294, 50639e50658. Rimhanen-Finne, R., Vuorinen, A., Marmo, S., Malmberg, S., Hanninen, M.L., 2004. Comparative analysis of Cryptosporidium, Giardia and indicator bacteria during sewage sludge hygienization in various composting processes. Lett. Appl. Microbiol. 38, 301e305. Santin, M., Trout, J.M., Fayer, R., 2004. Prevalence of Enterocytozoon bieneusi in post-weaned dairy calves in the eastern United States. Parasitol. Res. 93, 287e289. Slodkowicz-Kowalska, A., Graczyk, T.K., Tamang, L., Jedrzejewski, S., Nowosad, A., Zduniak, P., Solarczyk, P., Girouard, A.S., Majewska, A.C., 2006. Microsporidian species known to infect humans are present in aquatic birds: implications for transmission via water? Appl. Environ. Microbiol. 72, 4540e4544. Sparfel, J.M., Sarfati, C., Liguory, O., Caroff, B., Dumoutier, N., Gueglio, B., Billaud, E., Raffi, F., Molina, J.M., Miegeville, M., Derouin, F., 1997. Detection of microsporidia and identification of Enterocytozoon bieneusi in surface water by filtration followed by specific PCR. J. Eukaryot. Microbiol. 44, 78S. Stine, S.W., Vladich, F.D., Pepper, I.L., Gerba, C.P., 2005. Development of a method for the concentration and recovery of microsporidia from tap water. J. Environ. Sci. Health A Tox. Hazard. Subst. Environ. Eng. 40, 913e925. Straub, T.M., Pepper, I.L., Gerba, C.P., 1993. Hazards from pathogenic microorganisms in land-disposed sewage sludge. Rev. Environ. Contam. Toxicol. 132, 55e91.

4843

Subirats, M., Gonza´lez-Castelao, G., Aguilera, O., Moody, A., Visvesvara, G., Verdejo, J., Baquero, M., del Aguila, C., 1996. Diagnosis of 4 cases of intestinal microsporidiosis in AID patients. Enferm Infecc Microbiol. Clin. 14, 533e537. Thurston-Enriquez, J.A., Watt, P., Dowd, S.E., Enriquez, R., Pepper, I.L., Gerba, C.P., 2002. Detection of protozoan parasites and microsporidia in irrigation waters used for crop production. J. Food Prot. 65, 378e382. U.S.E.P.A, 2005. Method 1623: Cryptosporidium and Giardia in Water by Filtration/IMA/FA. Visvesvara, G.S., Leitch, G.J., da Silva, A.J., Croppo, G.P., Moura, H., Wallace, S., Slemenda, S.B., Schwartz, D.A., Moss, D., Bryan, R. T., et al., 1994. Polyclonal and monoclonal antibody and PCRamplified small-subunit rRNA identification of a microsporidian, Encephalitozoon hellem, isolated from an AIDS patient with disseminated infection. J. Clin. Microbiol. 32, 2760e2768. Weber, R., Bryan, R.T., Owen, R.L., Wilcox, C.M., Gorelkin, L., Visvesvara, G.S., 1992. Improved light-microscopical detection of microsporidia spores in stool and duodenal aspirates. The enteric opportunistic infections working group. N. Engl. J. Med. 326, 161e166. Weber, R., Bryan, R.T., Schwartz, D.A., Owen, R.L., 1994. Human microsporidial infections. Clin. Microbiol. Rev. 7, 426e461. Wiandt, S., Grimason, A.M., Baleux, B., Bontoux, J., 2000. Efficiency of wastewater treatment plants at removing Giardia sp. cysts in southern France. Schriftenr Ver Wasser Boden Lufthyg 105, 35e42.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 4 4 e4 8 5 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Recovery of toxic metal ions from washing effluent containing excess aminopolycarboxylate chelant in solution Hiroshi Hasegawa a,*, Ismail M.M. Rahman a,b,*, Masayoshi Nakano a, Zinnat A. Begum a, Yuji Egawa a, Teruya Maki a, Yoshiaki Furusho c, Satoshi Mizutani d a

Graduate School of Natural Science and Technology, Kanazawa University, Kakuma, Kanazawa 920-1192, Japan Department of Chemistry, University of Chittagong, Chittagong 4331, Bangladesh c GL Sciences, Inc., Nishishinjuku 6-22-1, Shinjuku, Tokyo 163-1130, Japan d Graduate School of Engineering, Osaka City University, Sugimoto 3-3-138, Sumiyoshi-Ku, Osaka 558-8585, Japan b

article info

abstract

Article history:

Aminopolycarboxylate chelants (APCs) are extremely useful for a variety of industrial

Received 21 February 2011

applications, including the treatment of toxic metal-contaminated solid waste materials.

Received in revised form

Because non-toxic matrix elements compete with toxic metals for the binding sites of

13 May 2011

APCs, an excess of chelant is commonly added to ensure the adequate sequestration of

Accepted 23 June 2011

toxic metal contaminants during waste treatment operations. The major environmental

Available online 2 July 2011

impacts of APCs are related to their ability to solubilize toxic heavy metals. If APCs are not sufficiently eliminated from the effluent, the aqueous transport of metals can occur

Keywords:

through the introduction of APCs into the natural environment, increasing the magnitude

Metal recovery

of associated toxicity. Although several techniques that focus primarily on the degradation

Aminopolycarboxylate chelants

of APCs at the pre-release step have been proposed, methods that recycle not only the

Non-destructive separation

processed water, but also provide the option to recover and reuse the metals, might be

Solid phase extraction

economically feasible, considering the high costs involved due to the chelants used in

Molecular recognition technology

metal ion sequestration. In this paper, we propose a separation process for the recovery of

Washing effluents

metals from effluents that contain an excess of APCs. Additionally, the option of recycling

Wastewater treatment

the processed water using a solid phase extraction (SPE) system with an ion-selective immobilized macrocyclic material, commonly known as a molecular recognition technology (MRT) gel, is presented. Simulated effluents containing As(V), Cd(II), Cr(III), Pb(II) or Se(IV) in the presence of APCs at molar ratios of 1:50 in H2O were studied with a flow rate of 0.2 mL min
1. The ‘captured’ ions in the SPE system were quantitatively eluted with HNO3. The effects of solution pH, metal-chelant stability constants and matrix elements were assessed. Better separation performance for the metals was achieved with the MRT-SPE compared to other SPE materials. Our proposed technique offers the advantage of a nondestructive separation of both metal ions and chelants compared to conventional treatment options for such effluents. ª 2011 Elsevier Ltd. All rights reserved.

* Corresponding authors. Tel./fax: þ81 76 234 4792. E-mail addresses: [email protected] (H. Hasegawa), [email protected] (I.M.M. Rahman). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.06.036

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 4 4 e4 8 5 4

1.

Introduction

Aminopolycarboxylate chelants (APCs) are used in a variety of industrial processes, for example, metal plating or finishing, textile and paper manufacturing, industrial cleaning, and water softening (Conway et al., 1999; Nowack and VanBriesen, 2005). They have also been applied to the remediation of toxic metal-contaminated solid waste materials (Raghavan et al., 1991; Grasso, 1993; Abumaizar and Khan, 1996; Peters, 1999; Roundhill, 2001; Chang et al., 2007). Because ethylenediaminetetraacetic acid (EDTA) forms strong watersoluble chelant complexes with most toxic metals (Egli, 2001; Nowack and VanBriesen, 2005; Lestan et al., 2008), it has been utilized most often among the APCs. Although APCs have received widespread acclaim for their excellent metal-binding capacities, the pre- and posttoxicities of APCs and related environmental consequences evoke many concerns (Rahman et al., 2011a). When APCs are released into aquatic environments, they may induce the remobilization of metal ions from soils and sediments into the water phase (Means et al., 1980; Nowack and VanBriesen, 2005), therefore extending the residence time of the metals. When APCs enter the environment, the exposure effects from APCs are likely to persist for a long time because of their poor photoe, chemoe and biodegradability (Means et al., 1980; Kari and Giger, 1995; Egli, 2001; Nowack, 2002; No¨rtemann, 2005). Additionally, in most cases, the toxicity threshold values of APCs increase with metal complexation (Sillanpa¨a¨ and Oikari, 1996; Sorvari and Sillanpa¨a¨, 1996; Sillanpa¨a¨, 2005). APCs can also contribute to eutrophication by increasing the total nitrogen content and phosphate solubility in interstitial waters (Horstmann and Gelpke, 1991; Erel and Morgan, 1992; Hering and Morel, 2002). Legislative agencies have become more concerned about eco-environmental consequences due to the increasing use of APCs, and increasingly stringent environmental regulations have been imposed (Grundler et al., 2005; van Ginkel and Geerts, 2005). Therefore, the treatment of industrial effluents and metal-contaminated wastewaters from other sources containing APCs is a prerequisite before they can be safely discharged. The characteristics and concentrations of both the added chelant and metals in the source solutions are important factors to consider when determining methods of treatment (Juang et al., 1999). A degradation treatment of APCs in solution is considered when the concentration falls below 1 mM (Juang and Wang, 2000a), and several methods have been proposed to obliterate and reduce the concentration of chelant in discharge waters (Krapfenbauer and Getoff, 1999; Mun˜oz and von Sonntag, 2000; Bucheli-Witschel and Egli, 2001; Ra¨mo¨ and Sillanpa¨a¨, 2001; Sillanpa¨a¨ and Pirkanniemi, 2001; Pirkanniemi et al., 2007). However, the recovery and reuse of APCs and metals become the main concern for concentrations above 5 mM in solution (Juang and Wang, 2000b) because the cost of chelants is a critical issue surrounding their use in metal ion sequestration (Kim and Ong, 1999; Lim et al., 2005; Lestan et al., 2008). An electrochemical reduction treatment followed by membrane separation (Juang and Wang, 2000b; Are´valo et al., 2002), a precipitation treatment with zero-valent metals (Lee and Marshall, 2002) or the addition of suitable reagents (e.g.,

4845

NaOH, Ca(OH)2, Na2S, FeSO4, FeCl3, NaH2PO4, Na2HPO4, or diethyldithiocarbamate) (Tu¨nay and Kabdasli, 1994; Chang, 1995; Steele and Pichtel, 1998; Hong et al., 1999; Kim and Ong, 1999; Xie and Marshall, 2001; Di Palma et al., 2003; Lim et al., 2005) are potential techniques proposed for the recovery of metal ions from metal-chelant solutions. Operational problems, such as membrane fouling, membrane degradation, considerable costs or the inherent stability of metal-chelant complexes in solution, are some drawbacks of the proposed separation techniques (Kim and Ong, 1999; Di Palma et al., 2003; Lim et al., 2005). Most of the proposed separation techniques are also based on equimolar solutions of metals and APCs (Chang, 1995; Kim and Ong, 1999; Juang and Wang, 2000b), while washing effluents from metalcontaminated solid-waste treatment processes are often characterized by a large excess of free APCs in solution or APCs that are combined with other competitive ions in the waste (Di Palma et al., 2003; Lestan et al., 2008). A technique that ensures the effortless selective separation of metal ions and recycling of processed water, including APCs, may therefore be economically beneficial (Lim et al., 2005; Lestan et al., 2008). The separation of metal ions from complex aqueous matrices using solid sorbent materials, a process known as solid phase extraction (SPE), has increased in popularity in recent years. SPE possesses the capability to interact with a variety of metal ions, and it has also been shown to interact with fairly specific selectivity to one particular ion of interest (Nickson et al., 1995; Hosten and Welz, 1999; Ghaedi et al., 2006, 2007; 2008; Rahman et al., 2011b,c). SPE systems have not been used extensively for the separation of metal ions from wastewaters containing APCs because APCs compete with SPE materials for complexation of metal ions, which causes a remarkable decrease in the extraction efficiency (Hasegawa et al., 2010, 2011). In this work, we propose a technique for the separation of toxic metal ions from synthetic effluents containing a large excess of APCs in solution. An ion-selective SPE system with immobilized macrocyclic material, commonly known as molecular recognition technology (MRT) gel (Bradshaw et al., 1988; Izatt et al., 1994, 1995), was used to achieve a quantitative recovery of metal ions. Unique features of the proposed separation process include the non-destructive recovery of toxic metal ions from the excess APC-containing aqueous matrix and the one-step clean-up of the effluent with an option for recycling the processed water.

2.

Material and methods

2.1.

Instrumentation

An inductively coupled plasma optical emission spectrometer (ICP-OES) (iCAP 6300, Thermo Fisher Scientific Inc., MA, USA), composed of an EMT duo quartz torch, glass spray chamber and concentric glass nebulizer, was used for the chemical analysis of metals. The operating conditions for the ICP-OES were as follows: the RF power at the torch was 1.15 kW, the plasma gas flow was 12 L min
1, the auxiliary gas flow was

4846

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 4 4 e4 8 5 4

1 L min
1, the nebulizer gas flow was 0.5 L min
1, and the integration time was 30 s. A fully automated high-performance liquid chromatography (HPLC) system (TOSOH 8020, Tosoh, Tokyo, Japan) was used for the analysis of NTA, EDTA and DTPA. The HPLC system was composed of the following components: a DP8020 pump, an AS-8021 auto sample injector, a CO-8020 column oven, a PD-8020 UVeVIS detector, PD-8020 data processing software, and TSK-gel ODS-80TM octadecylsilica columns (4.6 mm i.d.  250 mm and 4.6 mm i.d.  150 mm). The mobile phase solution consisted of 5 mM ammonium dihydrogenphosphate (pH 2.4) and was pumped at a flow rate of 0.5 mL min
1 at 25  C. The injection volume was 20 mL, and detection was performed at 254 nm. SPE was performed on a GL-SPE vacuum manifold kit (for 12 samples) (GL Sciences, Tokyo, Japan) combined with an air pump (CAS-1; AS ONE, Osaka, Japan). A Navi F-52 pH meter (Horiba Instruments, Kyoto, Japan) and a combination electrode were used for pH measurements. A Barnstead 4-housing E-Pure water purification system (Barnstead/Thermolyne, Dubuque, IA, USA) was used to prepare deionized water, which is referred to as EPW hereafter.

2.2.

Materials

Analytical grade commercial products were used without further purification. Stock solutions (10 mM) of As(V), Cd(II), Cr(III), Pb(II) and Se(IV) were prepared from sodium arsenate heptahydrate (Na2HAsO4$7H2O; Kanto Chemical, Tokyo, Japan), cadmium (II) nitrate tetrahydrate (Cd(NO3)2$4H2O; Wako Pure Chemical, Osaka, Japan), chromium (III) nitrate nonahydrate (Cr(NO3)3$9H2O; Wako Pure Chemical, Osaka, Japan), lead (II) nitrate (Pb(NO3)2; Wako Pure Chemical, Osaka, Japan) and sodium selenite (NaSeO3; Wako Pure Chemical, Osaka, Japan). Chelant stock solutions (10 mM) were prepared from nitrilotriacetic acid ((HOCOCH2)3N, NTA; Kanto Chemical, Tokyo, Japan), disodium dihydrogen ethylenediamine tetraacetate dihydrate (C10H14N2Na2O8$2H2O, EDTA; Kanto Chemical, Tokyo, Japan) and diethylenetriamineN,N,N0 ,N00 ,N000 -pentaacetic acid (C14H23N3O10, DTPA; Dojindo Laboratories, Kumamoto, Japan). A multi-element solution (PlasmaCAL, SCP Science, Que´bec, Canada) containing 21 metals (Al, Ba, Be, Bi, Ca, Cd, Co, Cu, Fe, Ga, In, Mg, Mn, Ni, Pb, Sc, Sr, Ti, V, Y, and Zn) in 5% HNO3 was used to check the effects of diverse ions. Solutions of working standards ranging from mM to mM were prepared by dilution with EPW on a weight basis. The experimental pH ranged from 4 to 9 and was adjusted using either 1 M HCl or 1 M NaOH. MES (2-(N-morpholino) ethanesulfonic acid monohydrate, C6H13NO4S$H2O; SigmaeAldrich, St. Louis, MO, USA), HEPES (N-2-HydroxyethylpiperazineN0 -2-ethanesulfonic acid, C8H18N2O4S; Nacalai Tesque, Kyoto, Japan), and TAPS (N-Tris(hydroxymethyl)methyl-3aminopropanesulfonic acid, C7H17NO6S; MP Biomedicals, Solon, OH, USA) were used as buffer reagents for pH 4e6, 7e8 and 9, respectively. Aqueous solutions of 10 mM chelating ligands in the appropriate buffer were spiked with 200 mM of As(V), Cd(II), Cr(III), Pb(II) or Se(IV) to prepare the samples. Different types of SPE materials were used, including an MRT gel, three chelating resins, and two ion exchange resins.

The MRT gel type was AnaLig TE-01 (silica gel base containing crown ether functional groups; GL Sciences, Tokyo, Japan). The chelating resins were Chelex-100 (styrene divinylbenzene base containing iminodiacetic acid functional groups; Bio-Rad Laboratories, Hercules, CA, USA), NOBIAS Chelate PA-1 (hydrophilic methacrylate base containing polyamino-polycarboxylic acid functional groups; Hitachi High-Technologies, Tokyo, Japan), and NOBIAS Chelate PB-1 (divinylbenzene/methacrylate polymer base containing polyamino-polycarboxylic acid functional groups; Hitachi High-Technologies, Tokyo, Japan). The ion exchange resins were NOBIAS Ion SA-1 (hydrophilic methacrylate base containing quaternized amine functional groups; Hitachi High-Technologies, Tokyo, Japan) and NOBIAS Ion SC-1 (hydrophilic methacrylate base containing sulfonic acid functional groups; Hitachi High-Technologies, Tokyo, Japan). Low-density polyethylene bottles (Nalge Nunc, Rochester, NY, USA), perfluoroalkoxy (PFA) tubes and micropipette tips (Nichiryo, Tokyo, Japan) were used throughout the experiments. Before use, laboratory wares were first soaked in an alkaline detergent (Scat 20X-PF, Nacalai Tesque, Kyoto, Japan) overnight, and then in 4 M HCl overnight, followed by rinsing with EPW after each step. Certified reference material (CRM) BCR-713 (effluent wastewater) from the European Commission Joint Research Centre, Institute of Reference Materials and Measurements (EC-JRC-IRMM), along with spiked soil washing solution (i.e., natural arsenic-contaminated soil from Hokkaido, Japan that was treated with 10 mM EDTA and spiked with a known amount of metal ions, followed by 6 h of shaking at room temperature) and spiked ‘real’ water samples (i.e., tap water from Kakuma, Kanazawa University, Kanazawa, Japan and water from Asano River, Kanazawa, Japan) were used for process validation. Cellulose membrane filters of 0.45 mm pore size (Advantec, Tokyo, Japan) were used for the pre-separation step filtration treatment of the soil washing solution and the ‘real’ water samples.

2.3.

Experimental setup

SPE materials were packed into 5 mL columns, and the columns were cleaned with 1 M HNO3 (8 mL) and EPW (6 mL). MES, HEPES or TAPS buffer solution (32e40 mL, 2 mL each loading) was allowed to flow through the column to ensure desired pH conditions from 4 to 9 in the SPE columns. A total of 4 mL sample solution with pH already preadjusted with 0.1 M buffer solution (MES, HEPES or TAPS, whichever was appropriate) was passed through the SPE column at a controlled flow rate of 0.2 mL min
1, and the column effluent was collected. The next step involved washing the column with EPW to remove the analyte fraction that was not retained. The final step was the elution of the ‘captured’ analyte with HNO3 (1 and 6 M) from the SPE system. The metal concentrations in the sample, column effluent, wash effluent and elution effluent were measured using the ICP-OES. The terms extraction and recovery were used to explain the separation performance of the SPE systems and were calculated from the analyte concentrations in the column effluent, wash effluent and elution effluent. The extraction ratio (%) of each column for individual metal species was calculated by comparing the numbers of moles of analyte in

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 4 4 e4 8 5 4

the elution effluent with the cumulative number of moles of analyte present in all the effluents. The cumulative number of moles of analyte recovered from all fractions (i.e., column effluent, wash effluent and elution effluent) was compared with the numbers of moles of analyte in the solution that was loaded onto the column to calculate the recovery ratio (%). Three replicates for each of the experiments were performed, and the average values were reported. The workflow sequence for the separation of metal ions using SPE columns followed by ICP-OES determination is shown schematically in Fig. 1.

3.

Results and discussion

3.1. Comparative evaluation of MRT-SPE and other commercial SPE materials 3.1.1.

Extraction and recovery behavior

Aqueous solutions containing toxic metal ions and APCs (NTA, EDTA and DTPA) in 1:50 molar ratios were treated with the MRT-SPE (AnaLig TE-01) and other commercial SPE materials (Chelex-100, NOBIAS Chelate PA-1, NOBIAS Chelate PB-1, NOBIAS Ion SA-1, and NOBIAS Ion SC-1) to compare the separation efficiencies at optimized conditions. As shown in Fig. 2, when we evaluated the metal separation performance of the SPE columns with or without APCs, we concluded that excess chelant in solution resulted in considerable performance variations of the SPEs. It was also apparent that the MRT-SPE ensured quantitative extraction of the toxic metal ions from the aqueous solution with or without APCs. However, an exception should be noted for the aqueous systems containing NTA and Pb(II), which exhibited an extraction rate below 77% for all of the SPE systems. The MRTSPE demonstrated superior extraction efficiency for EDTArich metal-fortified aqueous solutions when compared with other SPE systems, where the extraction rates were 60%. Separation of Cr(III) or Pb(II) from DTPA-rich aqueous solutions was quantitative for all the SPE systems, and aqueous systems with Pb(II) displayed similar behavior, even when no chelant was present in solution. The complete recovery of the metal ions that were ‘captured’ in the SPE columns was achieved with the MRT-SPE, while exceptions for As(V)- and Cr(II)spiked solutions without chelant were observed for some of the commercial SPE materials other than MRT-SPE.

4847

As(V) and Se(IV) have no known affinity for the APCs used here. However, those ions were simultaneously extracted with the APCs in solution, which subsequently reduced the extraction efficiencies of the SPEs (Fig. 2). These limitations were minimized with the use of MRT-SPE because the quantitative maximum extraction followed by recovery was achieved, compared with the other SPE systems.

3.1.2.

Effect of the metal-chelant stability constant

APCs (i.e., NTA, EDTA or DTPA) form water-soluble metal complexes of high thermodynamic stability (Lim et al., 2005) of varying metal-chelant stability constants (KML) with Cd(II), Cr(III) or Pb(II) (Table 1), which may influence the separation performance of the SPE materials. The effect of the metalchelant complexes’ conditional stability constants (K’ML, at pH 7) on the performance of MRT-SPE and other commercial SPE materials was studied for the extraction of Cd(II), Cr(III) or Pb(II) ions from chelant-rich, metal-spiked aqueous system (Fig. 3). AnaLig TE-01 demonstrated better effectiveness than the other SPE materials (i.e., Chelex-100, NOBIAS Chelate PA-1, NOBIAS Chelate PB-1, NOBIAS Ion SA-1, NOBIAS Ion SC-1) for Cd(II), Cr(III) or Pb(II) separation from EDTA-rich aqueous solutions. Comparable separation performances for Cd(II) or Pb(II) were observed for excess DTPA-containing solutions. The Pb(II) extraction rate with MRT-SPE from NTA-rich mixtures was only 57.5  1.9%, but none of the SPE columns were capable of ensuring its quantitative extraction. It is likely that the separation between metals and chelants (i.e., the extraction of metal) will be easier when the stability constant of the metal-chelant complex is low. The K’ML (at pH 7) of the Pb(II)-NTA complex (8.82) in the aqueous matrix was lower than that for EDTA and DTPA, and the quantitative maximum Pb(II) extraction rate was expected from NTA-containing solutions as it was obtained for EDTA and DTPA. However, Pb(II) oxide has a propensity to precipitate at neutral pH. Such precipitation is facilitated as a result of the lower affinity between NTA and Pb, which has a significant effect on the extraction capacity of the SPE system. Although the K’ML of Cd(II)-NTA complex (7.10) was also comparatively low, Cd(II) ions remain soluble in the aqueous matrix at pH 7 and have no such effect on the extraction performance. In general, we note that MRT-SPE can effectively be used to separate metal ions from the chelant-rich aqueous solutions for metal-chelant stability constants up to 18.8, which is the KML value for Pb(II)-DTPA, with exception of the behavior of

Fig. 1 e Schematic workflow diagram of the separation process.

4848

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 4 4 e4 8 5 4

Fig. 2 e Comparative performance of different SPE columns. The sample solutions were composed of 200 mM As(V), Cd(II), Cr(III), Pb(II), or Se(IV). The chelant was 10 mM NTA, EDTA, DTPA or EDDS and the matrix was H2O. The solution pH was 7, the sample volume was 4 mL, the flow rate was 0.2 mL minL1, and the elution solution consisted of 1 M HNO3 (2 mL) D 6 M HNO3 (1 mL) D EPW (1 mL) (n [ 3).

4849

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 4 4 e4 8 5 4

Table 1 e Acid dissociation constants (pKa), stability constants (KML) and conditional stability constants (K’ML) of metaleligand complexes at 25  C (m [ 0.1)a. APCs

KML

pKa

NTA EDTA DTPA

pKa1

pKa2

pKa3

pKa4

1.81 2.00 2.0

2.52 2.69 2.70

9.66 6.13 4.28

10.19 8.60

2þ

3þ

K’ML (at pH 7) 2þ

2þ

pKa5

Cd

Cr

Pb

Cd

Cr3þ

Pb2þ

10.50

9.76 16.5 19.0

NAb 23.4 NAb

11.48 18.0 18.8

7.10 13.3 15.5

e 20.2 e

8.82 14.8 15.3

a Martell et al. (2004). b NA ¼ Not available. Data not available in the critically selected NIST database.

Pb(II) with NTA. The MRT-SPE appeared as the solitary potential option for the separation of toxic metal ions from aqueous solutions containing an excess of EDTA, which is the most widely used APC for metal-contaminated waste treatment.

3.2.

Effect of variables on the performance of MRT-SPE

3.2.1.

pH

The separation performance of the AnaLig TE-01 SPE column was studied as a function of pH and was described in terms of extraction and recovery rate (Fig. 4). The experimental conditions utilized EDTA, considering its frequent use among the APCs. Therefore, the study was restricted to the pH range from 4 to 9 because of the insufficient solubility of EDTA at very low pH in aqueous media (Ueno et al., 1992). The increasing solubility of silica gel with increasing pH (Vogelsberger et al., 1992), which may dissolve the silica gel base support of AnaLig TE-01 column, was also a concern. Nearly similar extraction patterns for As(V) or Se(IV) were

observed with or without EDTA in solution, which established that the excess chelant in the aqueous system had no significant influence on the solubility or separation aptitude of those metals. However, a significant drop in the extraction rate of As(V) or Se(IV) above a pH of 8 was observed, which may have been due to increased concentrations of the competitive ions (OH
or HL3
) in the system. An extraction rate of 98% for Cd(II), Cr(III) and Pb(II) from pH 5 to 7 was attained from metal-fortified solutions containing an excess of chelant, while the changes in the recovery rates were insignificant in terms of pH. The decrease in the extraction rate at pH <5 or >7 can be attributed either to an excess of Hþ ions in the acidic region or OH
/HL3
ions in the basic region, respectively. Subsequent experiments with the MRT-SPE column were conducted at pH 7 to minimize any possible effects from the competitive ions.

3.2.2.

Sample loading flow rate

The loading flow rates of metal-fortified sample solutions have a significant influence on metal retention rates in SPE

Fig. 3 e Effect of metal-chelant stability constants on the performance of SPE materials. The sample solutions were composed of 200 mM Cd(II), Cr(III) or Pb(II), and the chelant was 10 mM NTA, EDTA or DTPA. The matrix was H2O, the solution pH was 7, the sample volume was 4 mL, the flow rate was 0.2 mL minL1, and the elution solution consisted of 1 M HNO3 (2 mL) D 6 M HNO3 (1 mL) D EPW (1 mL) (n [ 3).

4850

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 4 4 e4 8 5 4

Fig. 4 e Extraction and recovery performance of the MRT-SPE column as a function of pH, with or without chelant. The sample solutions were composed of 200 mM As(V), Cd(II), Cr(III), Pb(II), or Se(IV). The chelant was 10 mM EDTA, and the matrix was H2O. The pH ranged from 4 to 9, the sample volume was 4 mL, the flow rate was 0.2 mL minL1, and the elution solution consisted of 1 M HNO3 (2 mL) D 6 M HNO3 (1 mL) D EPW (1 mL) (n [ 3).

columns (Bag et al., 1998). The effects of sample loading flow rates were studied in the range of 0.2e5 mL min
1. A gradual decrease in retention capacities of the MRT-SPE column was observed with increasing flow rates above 0.25 mL min
1 (Fig. 5). A constant retaining capability of the MRT-SPE column at the initial loading period is indicated by such behavior; therefore, a flow rate of 0.2 mL min
1 was applied for subsequent experiments.

3.2.3.

Eluent

Eluent selected for a particular separation process should be capable of extracting the analyte, thereby facilitating its quantitative determination (Chen et al., 2009). Analytes retained in the MRT-SPE column were eluted with HNO3 (4 mL) of varying concentrations (0.1e6 M), which all displayed constant recovery rates for eluent concentrations above 0.5 M (Fig. 6). However, IBC Advanced Technologies (2007)

Fig. 5 e Effect of sample loading flow rates on the separation performance of the MRT-SPE column. The sample solutions were composed of 200 mM As(V), Cd(II), Cr(III), Pb(II), or Se(IV). The chelant was 10 mM EDTA, and the matrix was H2O. The pH was 7, the sample volume was 4 mL, the flow rate ranged from 0.2 to 5 mL minL1, and the elution solution consisted of 1 M HNO3 (2 mL) D 6 M HNO3 (1 mL) D EPW (1 mL) (n [ 3).

4851

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 4 4 e4 8 5 4

Fig. 6 e Effect of eluent concentration on the separation performance of the MRT-SPE column. The sample solutions were composed of 200 mM As(V), Cd(II), Cr(III), Pb(II), or Se(IV). The chelant was 10 mM EDTA, and the matrix was H2O. The solution pH was 7, the sample volume was 4 mL, the flow rate was 0.2 mL minL1, and the elution solution consisted of 0.1e6 M HNO3 (3 mL) D EPW (1 mL) (n [ 3).

recommended the use of 5 M acids for the elution of bound ions in the TE-01 SPE column. Hence, a combination of 1 M HNO3 (2 mL) and 6 M HNO3 (1 mL) was selected as the eluent for subsequent experiments to ensure the complete elution of the analyte when treated with TE-01.

3.3.

Effect of diverse metal ions

The interference caused by complexing species results in significant problems towards the quantitative extraction of analytes (Prabhakaran and Subramanian, 2003). To examine the separation efficiency of MRT-SPE in the presence of various interfering metal ions, studies were performed using PlasmaCAL multi-element metal ion solutions spiked with the target metal ions and APCs. EDTA was used as the representative APC because EDTA has most often been utilized among the APCs, owing to its capacity to form water-soluble chelant complexes with most toxic metals (Egli, 2001; Nowack and VanBriesen, 2005; Lestan et al., 2008). The metalto-chelant ratio was maintained at 1:50, and the final solutions were allowed to equilibrate for 24 h before analysis. The extraction and recovery rates demonstrated the superior ion-

Table 2 e Separation performance of the MRT-SPE column in the presence of various interfering metal species in the matrixa. Species As(V) Cd(II) Cr(III) Pb(II) Se(IV)

Extraction (%) 99.6  101  98.7  100  97.7 

3.4 4.7 3.9 2.5 3.6

Recovery (%) 100  100  99.4  97.8  102 

4.2 1.6 1.1 3.4 2.1

a Sample solutions were composed of 200 mM As(V), Cd(II), Cr(III), Pb(II), or Se(IV). The chelant was 10 mM EDTA, and the matrix was H2O. The matrix ions included Al, Ba, Be, Bi, Ca, Cd, Co, Cu, Fe, Ga, In, Mg, Mn, Ni, Pb, Sc, Sr, Ti, V, Y, and Zn. The solution pH was 7, the sample volume was 4 mL, the flow rate was 0.2 mL min
1, and the elution solution consisted of 1 M HNO3 (2 mL) þ 6 M HNO3 (1 mL) þ EPW (1 mL) (n ¼ 3).

selective separation performance of the MRT-SPE in the presence of large concentrations of matrix components (Table 2).

3.4.

Retention capacity of the MRT-SPE

The stability of the SPE system during the separation process can be determined from its retention capacity, which is calculated from the breakthrough volume (i.e., the volume of sample that causes the target analyte to be eluted from the SPE material) and the analyte concentration (Yu et al., 2003). Metal-spiked sample solutions were passed through the MRTSPE column, eluted and subjected to ICP-OES analysis to estimate the retention capacity expressed in terms of mmol of analyte captured in 1 g of SPE material. The retention capacities of the MRT-SPE (mmol g
1) at pH 7 were as follows: 0.44  0.04 for As(V), 0.41  0.06 for Cd(II), 0.05  0.02 for Cr(III), 0.48  0.06 for Pb(II), and 0.34  0.05 for Se(IV). The matrix was H2O, the flow rate was 0.2 mL min
1, and the elution solution consisted of 2 mL of 1 M HNO3, 1 mL of 6 M HNO3, and 1 mL of EPW.

3.5.

Regeneration ability of the MRT-SPE

The regeneration ability of the MRT-SPE was investigated with sample solutions spiked with 200 mM of As(V) or Pb(II) ions and 10 mM of EDTA in aqueous matrix. Again, the flow rate was

Table 3 e Separation of metals from certified reference material BCR-713 (effluent wastewater). Certified dataa Species As Cd Cr Pb Se

This work
1

Value (mg L )

Species

Value (mg L
1)

    

As(V) Cd(II) Cr(III) Pb(II) Se(IV)

8.7  0.8 5.2  0.7 NDb NDb NDb

9.7 5.1 21.9 47 5.6

1.1 0.6 2.4 4 1.0

a Certified by EC-JRC-IRMM (European Commission Joint Research Centre, Institute of Reference Materials and Measurements). b ND ¼ Not detected.

4852

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 4 4 e4 8 5 4

Table 4 e Separation of metals from spiked samples of ‘real’ waters and soil washing effluent. Sample

Tap Water River Water Soil washing effluent

Added (mg L
1)


1

Found (mg L ) Recovery (%) Found (mg L
1) Recovery (%) Found (mg L
1) Recovery (%)

As(V)

Cd(II)

Cr(III)

Pb(II)

Se(IV)

15.0

22.5

10.4

41.4

15.8

14.9  0.3 99.4 15.1  0.5 100.6 15.0  0.3 100.3

18.9  0.9 83.9 22.8  0.7 101.5 22.5  0.9 100.3

10.5  1.5 101.1 10.6  0.2 101.9 10.6  0.5 101.5

40.8  3.1 98.3 41.6  0.8 100.3 41.5  0.9 100.2

15.9  1.2 100.4 15.8  0.3 100.0 13.8  1.4 87.5

0.2 mL min
1, and the elution solution contained 2 mL of 1 M HNO3, 1 mL of 6 M HNO3, and 1 mL of EPW. The extraction rates of the fresh column (As(V): 99.0  0.1; Pb(II): 100  0.1) and after 100 cycles (As(V): 97.2  4.1; Pb(II): 98.4  0.3) were evaluated to conclude that more than 100 loading and elution cycles could be performed using MRT-SPE without any loss of analytical performance.

3.6.

Accuracy and applications

3.6.1.

Recovery of metals from certified reference material

EC-JRC-IRMM CRM, namely BCR-713 (effluent wastewater), spiked with 10 mM of EDTA (pH maintained at 7 with HEPES buffer), was used to evaluate the accuracy of the proposed separation process (Table 3). The recovery rates for As(V) and Cd(II) were 89.7 and 101.4%, respectively, while Cr(III), Pb(II) or Se(IV) were not detected.

3.6.2. Recovery of metals from ‘real’ water samples and soil washing effluent The proposed separation process was applied to the analysis of local natural water samples (i.e., both tap water and river water) and soil washing effluent. The samples were spiked with known amounts of As(V), Cd(II), Cr(III), Pb(II) or Se(IV) and 10 mM of EDTA, followed by MRT-SPE separation and ICP-OES analysis (Table 4). Recoveries at varying rates (99e101% for As(V), 84e102% for Cd (II), 101e102% for Cr(III), 98e100% for Pb(II), and 88e100% for Se(IV)) from metal-spiked excess chelant-containing solutions were observed.

4.

Conclusion

The recoveries of As(V), Cd(II), Cr(III), Pb(II) and Se(IV) from simulated washing effluents containing an excess of APCs (i.e., NTA, EDTA or DTPA) was accomplished with an ionselective immobilized macrocyclic material, commonly known as MRT gel. The MRT-SPE system showed optimum separation performance in the pH range of 5e7. Quantitative extraction occurred using a sample loading flow rate of 0.2 mL min
1, and the ‘captured’ metal ions were eluted with a mixture of 1 and 6 M HNO3. The MRT-SPE was stable during operation and enabled more than 100 loading and elution cycles to be performed without any loss of analytical performance. The non-destructive treatment of chelant-enriched metal-contaminated effluent with the subsequent option to recycle the processed water and metal ions are the major focal points of the proposed separation process.

Acknowledgement This research was partially supported by Grants-in-Aid for Scientific Research (K22042) from the Ministry of the Environment, Japan.

references

Abumaizar, R., Khan, L.I., 1996. Laboratory investigation of heavy metal removal by soil washing. J. Air Waste Manag. Assoc. 46, 765e768. Are´valo, E.F., Stichnothe, H., Tho¨ming, J., Calmano, W., 2002. Evaluation of a leaching process coupled with regeneration/ recycling of the extractant for treatment of heavy metal contaminated solids. Environ. Technol. 23, 571e581. Bag, H., Lale, M., Tu¨rker, A.R., 1998. Determination of iron and nickel by flame atomic absorption spectrophotometry after preconcentration on Saccharomyces cerevisiae immobilized sepiolite. Talanta 47, 689e696. Bradshaw, J.S., Bruening, R.L., Krakowiak, K.E., Tarbet, B.J., Bruening, M.L., Izatt, R.M., Christensen, J.J., 1988. Preparation of silica gel-bound macrocycles and their cation-binding properties. J. Chem. Soc. Chem. Comm., 812e814. Bucheli-Witschel, M., Egli, T., 2001. Environmental fate and microbial degradation of aminopolycarboxylic acids. FEMS Microbiol. Rev. 25, 69e106. Chang, F.C., Lo, S.L., Ko, C.H., 2007. Recovery of copper and chelating agents from sludge extracting solutions. Sep. Purif. Technol. 53, 49e56. Chang, L.Y., 1995. A waste minimization study of a chelated copper complex in wastewater e treatability and process analysis. Waste Manag. 15, 209e220. Chen, D., Huang, C., He, M., Hu, B., 2009. Separation and preconcentration of inorganic arsenic species in natural water samples with 3-(2-aminoethylamino) propyltrimethoxysilane modified ordered mesoporous silica micro-column and their determination by inductively coupled plasma optical emission spectrometry. J. Hazard. Mater. 164, 1146e1151. Conway, M., Holoman, S., Jones, L., Leenhouts, R., Williamson, G., 1999. Selecting and using chelating agents. Chem. Eng. 106, 86e90. Di Palma, L., Ferrantelli, P., Merli, C., Biancifiori, F., 2003. Recovery of EDTA and metal precipitation from soil flushing solutions. J. Hazard. Mater. 103, 153e168. Egli, T., 2001. Biodegradation of metal-complexing aminopolycarboxylic acids. J. Biosci. Bioeng. 92, 89e97. Erel, Y., Morgan, J.J., 1992. The relationships between rockderived lead and iron in natural waters. Geochim. Cosmochim. Ac 56, 4157e4167. Ghaedi, M., Ahmadi, F., Shokrollahi, A., 2007. Simultaneous preconcentration and determination of copper, nickel, cobalt and lead ions content by flame atomic absorption spectrometry. J. Hazard. Mater. 142, 272e278.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 4 4 e4 8 5 4

Ghaedi, M., Asadpour, E., Vafaie, A., 2006. Sensitized spectrophotometric determination of Cr(III) ion for speciation of chromium ion in surfactant media using a-benzoin oxime. Spectrochim. Acta A 63, 182e188. Ghaedi, M., Shokrollahi, A., Kianfar, A.H., Mirsadeghi, A.S., Pourfarokhi, A., Soylak, M., 2008. The determination of some heavy metals in food samples by flame atomic absorption spectrometry after their separationpreconcentration on bis salicyl aldehyde, 1,3 propan diimine (BSPDI) loaded on activated carbon. J. Hazard. Mater. 154, 128e134. Grasso, D., 1993. Hazardous Waste Site Remediation: Source Control. Lewis Publishers, Boca Raton, FL. Grundler, O.J., van der Steen, A.T.M., Wilmot, J., 2005. Overview of the European risk assessment on EDTA. In: Nowack, B., VanBriesen, J.M. (Eds.), Biogeochemistry of Chelating Agents. American Chemical Society, Washington, DC, pp. 336e347. Hasegawa, H., Rahman, I.M.M., Kinoshita, S., Maki, T., Furusho, Y., 2010. Non-destructive separation of metal ions from wastewater containing excess aminopolycarboxylate chelant in solution with an ion-selective immobilized macrocyclic material. Chemosphere 79, 193e198. Hasegawa, H., Rahman, I.M.M., Kinoshita, S., Maki, T., Furusho, Y., 2011. Separation of dissolved iron from the aqueous system with excess ligand. Chemosphere 82, 1161e1167. Hering, J.G., Morel, F.M.M., 2002. Kinetics of trace metal complexation: role of alkaline-earth metals. Environ. Sci. Technol. 22, 1469e1478. Hong, P.K.A., Li, C., Banerji, S.K., Regmi, T., 1999. Extraction, recovery, and biostability of EDTA for remediation of heavy metal-contaminated soil. J. Soil Contam. 8, 81e103. Horstmann, U., Gelpke, N., 1991. Algal growth stimulation by chelatisation risks associated with complexants in P-free washing agents. Rev. Intl. Oceanogr. Med. 260, 101e104. Hosten, E., Welz, B., 1999. Evaluation of an immobilised macrocyclic material for on-line column preconcentration and separation of cadmium, copper and lead for electrothermal atomic absorption spectrometry. Anal. Chim. Acta 392, 55e65. IBC Advanced Technologies, 2007. AnaLig Data Sheet: TE-01 and TE-02. IBC Advanced Technologies, Inc., American Fork, UT. Izatt, R.M., Bradshaw, J.S., Bruening, R.L., Bruening, M.L., 1994. Solid phase extraction of ions of analytical interest using molecular recognition technology. Am. Lab. 26 28C-28M. Izatt, R.M., Bradshaw, J.S., Bruening, R.L., Tarbet, B.J., Bruening, M. L., 1995. Solid phase extraction of ions using molecular recognition technology. Pure Appl. Chem. 67, 1069e1074. Juang, R.S., Wang, S.W., 2000a. Electrolytic recovery of binary metals and EDTA from strong complexed solutions. Water Res. 34, 3179e3185. Juang, R.S., Wang, S.W., 2000b. Metal recovery and EDTA recycling from simulated washing effluents of metal-contaminated soils. Water Res. 34, 3795e3803. Juang, R.S., Wang, S.W., Lin, L.C., 1999. Simultaneous recovery of EDTA and lead(II) from their chelated solutions using a cation exchange membrane. J. Membr. Sci. 160, 225e233. Kari, F.G., Giger, W., 1995. Modeling the photochemical degradation of ethylenediaminetetraacetate in the river Glatt. Environ. Sci. Technol. 29, 2814e2827. Kim, C., Ong, S.-K., 1999. Recycling of lead-contaminated EDTA wastewater. J. Hazard. Mater. 69, 273e286. Krapfenbauer, K., Getoff, N., 1999. Comparative studies of photoand radiation-induced degradation of aqueous EDTA. Synergistic effects of oxygen, ozone and TiO2 (acronym: CoPhoRaDe/EDTA). Radiat. Phys. Chem. 55, 385e393. Lee, C.C., Marshall, W.D., 2002. Recycling of complexometric extractants to remediate a soil contaminated with heavy metals. J. Environ. Monit. 4, 325e329.

4853

Le stan, D., Luo, C.L., Li, X.D., 2008. The use of chelating agents in the remediation of metal-contaminated soils: a review. Environ. Pollut. 153, 3e13. Lim, T.T., Chui, P.C., Goh, K.H., 2005. Process evaluation for optimization of EDTA use and recovery for heavy metal removal from a contaminated soil. Chemosphere 58, 1031e1040. Martell, A.E., Smith, R.M., Motekaitis, R.J., 2004. NIST Standard Reference Database 46: NIST Critically Selected Stability Constants of Metal Complexes Database (Version 8.0 for Windows). Texas A&M University, College Station, TX. Means, J.L., Kucak, T., Crerar, D.A., 1980. Relative degradation rates of NTA, EDTA and DTPA and environmental implications. Environ. Pollut. B. 1, 45e60. Mun˜oz, F., von Sonntag, C., 2000. The reactions of ozone with tertiary amines including the complexing agents nitrilotriacetic acid (NTA) and ethylenediaminetetraacetic acid (EDTA) in aqueous solution. J. Chem. Soc. Perk. T. 2, 2029e2033. Nickson, R.A., Hill, S.J., Worsfold, P.J., 1995. Analytical perspective. Solid phase techniques for the preconcentration of trace metals from natural waters. Anal. Proc. 32, 387e395. No¨rtemann, B., 2005. Biodegradation of chelating agents: EDTA, DTPA, PDTA, NTA, and EDDS. In: Nowack, B., VanBriesen, J.M. (Eds.), Biogeochemistry of Chelating Agents. American Chemical Society, Washington, DC, pp. 150e170. Nowack, B., 2002. Environmental chemistry of aminopolycarboxylate chelating agents. Environ. Sci. Technol. 36, 4009e4016. Nowack, B., VanBriesen, J.M., 2005. Chelating agents in the environment. In: Nowack, B., VanBriesen, J.M. (Eds.), Biogeochemistry of Chelating Agents. American Chemical Society, Washington, DC, pp. 1e18. Peters, R.W., 1999. Chelant extraction of heavy metals from contaminated soils. J. Hazard. Mater. 66, 151e210. Pirkanniemi, K., Metsa¨rinne, S., Sillanpa¨a¨, M., 2007. Degradation of EDTA and novel complexing agents in pulp and paper mill process and waste waters by Fenton’s reagent. J. Hazard. Mater. 147, 556e561. Prabhakaran, D., Subramanian, M.S., 2003. Selective extraction and sequential separation of actinide and transition ions using AXAD-16-BTBED polymeric sorbent. React. Funct. Polym. 57, 147e155. Raghavan, R., Coles, E., Dietz, D., 1991. Cleaning excavated soil using extraction agents: a state-of-the-art review. J. Hazard. Mater. 26, 81e87. Rahman, I.M.M., Hossain, M.M., Begum, Z.A., Rahman, M.A., Hasegawa, H., 2011a. Eco-environmental consequences associated with chelant-assisted phytoremediation of metalcontaminated soil. In: Golubev, I.A. (Ed.), Handbook of Phytoremediation. Nova Science Publishers, Inc., New York, pp. 709e722. Rahman, I.M.M., Begum, Z.A., Nakano, M., Furusho, Y., Maki, T., Hasegawa, H., 2011b. Selective separation of arsenic species from aqueous solutions with immobilized macrocyclic material containing solid phase extraction columns. Chemosphere 82, 549e556. Rahman, I.M.M., Furusho, Y., Begum, Z.A., Izatt, N., Bruening, R., Sabarudin, A., Hasegawa, H., 2011c. Separation of lead from high matrix electroless nickel plating waste solution using an ion-selective immobilized macrocycle system. Microchem. J. 98, 103e108. Ra¨mo¨, J., Sillanpa¨a¨, M., 2001. Degradation of EDTA by hydrogen peroxide in alkaline conditions. J. Clean. Prod. 9, 191e195. Roundhill, D.M., 2001. Extraction of Metals from Soils and Waters. Kluwer Academic/Plenum Publishers, New York.

4854

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 4 4 e4 8 5 4

Sillanpa¨a¨, M., Oikari, A., 1996. Assessing the impact of complexation by EDTA and DTPA on heavy metal toxicity using microtox bioassay. Chemosphere 32, 1485e1497. Sillanpa¨a¨, M., Pirkanniemi, K., 2001. Recent developments in chelate degradation. Environ. Technol. 22, 791e801. Sillanpa¨a¨, M.E.T., 2005. Distribution and fate of chelating agents in the environment. In: Nowack, B., VanBriesen, J.M. (Eds.), Biogeochemistry of Chelating Agents. American Chemical Society, Washington, DC, pp. 226e233. Sorvari, J., Sillanpa¨a¨, M., 1996. Influence of metal complex formation on heavy metal and free EDTA and DTPA acute toxicity determined by Daphnia magna. Chemosphere 33, 1119e1127. Steele, M.C., Pichtel, J., 1998. Ex-situ remediation of a metalcontaminated superfund soil using selective extractants. J. Environ. Eng.-ASCE 124, 639e645. Tu¨nay, O., Kabdasli, N.I., 1994. Hydroxide precipitation of complexed metals. Water Res. 28, 2117e2124.

Ueno, K., Imamura, T., Cheng, K.L., 1992. Handbook of Organic Analytical Reagents. CRC Press, Boca Raton, FL. van Ginkel, C.G., Geerts, R., 2005. Full-Scale biological treatment of industrial effluents containing EDTA. In: Nowack, B., VanBriesen, J.M. (Eds.), Biogeochemistry of Chelating Agents. American Chemical Society, Washington, DC, pp. 195e203. Vogelsberger, W., Seidel, A., Rudakoff, G., 1992. Solubility of silica gel in water. J. Chem. Soc. Faraday T. 88, 473e476. Xie, T., Marshall, W.D., 2001. Approaches to soil remediation by complexometric extraction of metal contaminants with regeneration of reagents. J. Environ. Monit. 3, 411e416. Yu, C.H., Cai, Q.T., Guo, Z.X., Yang, Z.G., Khoo, S.B., 2003. Inductively coupled plasma mass spectrometry study of the retention behavior of arsenic species on various solid phase extraction cartridges and its application in arsenic speciation. Spectrochim. Acta B. 58, 1335e1349.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 5 5 e4 8 6 5

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Treatment of nanofiltration and reverse osmosis concentrates: Comparison of precipitative softening, coagulation, and anion exchange Sarah E.H. Comstock, Treavor H. Boyer*, Katherine C. Graf Department of Environmental Engineering Sciences, University of Florida, P.O. Box 116450, Gainesville, FL 32611-6450, USA

article info

abstract

Article history:

Disposal and treatment of concentrate from nanofiltration (NF) and reverse osmosis (RO) are

Received 31 December 2010

major challenges to implementing membrane treatment processes. Intermediate treatment

Received in revised form

of membrane concentrate, between primary and secondary membrane stages, has the

16 May 2011

potential to increase membrane recovery rates and decrease the volume of concentrate

Accepted 23 June 2011

produced. To achieve this, however, there is a need to better understand treatment of

Available online 2 July 2011

membrane concentrate. As a result, this work systematically evaluated lime softening, ferric sulfate coagulation, and magnetic ion exchange (MIEX) as individual, intermediate treatment

Keywords:

processes for membrane concentrate. Six membrane concentrates, from NF and RO, with

Calcium

varying concentrations of calcium, dissolved organic matter (DOM), and sulfate were chosen

Dissolved organic carbon

for this study. Maximum removal of calcium was achieved by lime softening, whereas

Membrane concentrate

maximum removals of DOM and sulfate were achieved by MIEX. The results of this work

Membrane fouling

show that intermediate treatment of NF/RO concentrate is capable of producing treated

MIEX

concentrate with water quality approximately equal to the initial source water. ª 2011 Elsevier Ltd. All rights reserved.

Sulfate

1.

Introduction

Nanofiltration (NF) and reverse osmosis (RO) are increasingly being used to treat water sources, such high hardness and organic-rich surface and ground waters, which are difficult to treat by conventional methods (Mohammadesmaeili et al. (2010b); Gabelich et al. (2011)). However, one of the main challenges for NF/RO is the treatment and disposal of membrane concentrate (Van der Bruggen et al. (2003); Greenlee et al. (2009)). In addition, treatment of NF/RO concentrate to remove sparingly soluble minerals and increase recovery rates is essential for NF/ RO treatment plants to achieve zero liquid discharge (Heijman et al. (2009); Mohammadesmaeili et al. (2010b)). One approach to increase recovery rates is intermediate treatment of primary NF/RO concentrate to remove scale-forming minerals, such as

divalent cations and sulfate, and foulants, such as dissolved organic matter (DOM), before sending the concentrate through a second stage of NF/RO units (Mohammadesmaeili et al. (2010a); Gabelich et al. (2011); Greenlee et al. (2011)). Improving the recovery rates of NF/RO processes results in decreased operating costs because the cost of disposing of NF/RO concentrate can be substantial in some cases and may even preclude membrane treatment as an option (Ning and Troyer (2009)). NF/ RO concentrate treatment may also be necessary before final disposal to a wastewater treatment plant, surface water, or deep-well injection so as to not have an adverse effect on receiving waters (Roberts et al. (2010)). Treatment of NF/RO concentrate can be divided into two categories: (1) removal of excess water (e.g., evaporation ponds) and (2) removal of specific chemicals (Van der Bruggen

* Corresponding author. Tel.: þ1 352 846 3351. E-mail address: [email protected] (T.H. Boyer). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.06.035

4856

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 5 5 e4 8 6 5

et al. (2003); Greenlee et al. (2009)). The focus of this review is the removal of specific chemicals, such as divalent cations, sulfate, and DOM, from NF/RO concentrate. The motivation for removing divalent cations and sulfate is to prevent mineral precipitation, e.g., magnesium hydroxide, calcium carbonate, calcium sulfate, barium sulfate, and strontium sulfate (Shih et al. (2005); Lin et al. (2006); Jarusutthirak et al. (2007); Rahardianto et al. (2007); Greenlee et al. (2010a); Lin et al. (2011)), on NF/RO membrane surfaces. Furthermore, various fractions of DOM and DOM-calcium complexes are known to foul NF/RO membrane surfaces (Yoon et al. (1998); Li and Elimelech (2004); Jarusutthirak et al. (2007); Tran et al. (2007); Jin et al. (2009)). Precipitative softening is the most well-documented method for treatment of NF/RO concentrate generated during drinking water treatment (Gabelich et al. (2007); Rahardianto et al. (2007); Qu et al. (2009), Greenlee et al. (2010a, 2010b), Mohammadesmaeili et al. (2010a, 2010b); Gabelich et al. (2011)). For example, Gabelich et al. (2007) evaluated intermediate treatment of RO concentrate by precipitation and showed that > 95% recovery was achieved by adding NaOH to primary RO concentrate, allowing the precipitate to settle, and following with secondary RO treatment of the treated concentrate. The precipitation step removed scaling precursors including calcium, barium, strontium, magnesium, and silica. Subsequent work by Gabelich et al. (2011) demonstrated the feasibility of intermediate precipitative softening of RO concentrate at the pilot scale. There is a lack of literature examining other physicalchemical processes for treatment of NF/RO concentrate produced during drinking water treatment. For example, coagulation and anion exchange have been studied as pretreatment processes for NF/RO to minimize fouling (Shon et al. (2008); Verliefde et al. (2009); Cornelissen et al. (2010)), and both processes have been investigated for treatment of RO concentrate produced during wastewater treatment (Dialynas et al. (2008)). However, coagulation and anion exchange have not been systematically studied as intermediate treatment processes for primary NF/RO concentrate produced during drinking water treatment. Furthermore, there is no previous research comparing the performance of precipitative softening, coagulation, and anion exchange for intermediate treatment of membrane concentrate. Filling these gaps in knowledge is important because it will allow for selection of treatment process based on the desired water chemistry of the treated NF/RO concentrate. Accordingly, the goal of this work was to compare precipitative softening, coagulation, and anion exchange as intermediate treatment processes for primary NF/RO concentrate to reduce mineral scaling and organic fouling potentials in subsequent use. The specific objectives of this work were to: (i) evaluate the chemistry of NF/RO concentrates generated during drinking water treatment of high hardness and high DOM groundwater; (ii) evaluate precipitative softening using lime for removal of hardness cations, DOM, and, sulfate; (iii) evaluate coagulation using ferric sulfate for removal of DOM; (iv) evaluate anion exchange using MIEX resin for removal of DOM and sulfate; and (v) discuss the implications of this work on the treatability of NF/RO concentrate. The results of this

research should be applicable to physical-chemical treatment of other high ionic strength waste streams, especially those high in DOM.

2.

Experimental

2.1.

Membrane concentrate

All membrane concentrate samples examined in this work were from municipal drinking water treatment plants in Florida, USA. All membrane concentrate samples were collected using sampling ports on the membrane units, transported to the laboratory, stored at 4  C, and tested within 3 months of collection. Membrane concentrate was collected from five different NF plants (NF-1, NF-2, NF-3, NF-4, and NF-5) and one RO plant (RO-1). Membrane concentrate samples were collected twice from each facility during May and July 2010, except for NF-5 which was collected once during May 2010. Source (i.e., feed) water corresponding to each membrane concentrate was collected once during May 2010.

2.2.

Treatment chemicals

Calcium oxide (CaO; Acros Organics; 97þ% purity) was used for lime softening, ferric sulfate (Fe2(SO4)3$9H2O; Fisher Scientific; technical grade) was used for coagulation, and MIEX resin in the chloride form (MIEX-Cl; Orica Watercare) was used for anion exchange. Sodium carbonate (Na2CO3; Fisher Scientific; certified ACS) was used for alkalinity addition in select lime-soda softening experiments. Chemicals used in analytical methods are described below. Deionized (DI) water was used to prepare all chemical reagents and standards, and to dilute samples when necessary. Glassware was cleaned with laboratory detergent, if necessary soaked overnight in a 6% nitric acid solution, and rinsed three times with DI water.

2.3. Precipitative softening, coagulation, and anion exchange jar tests A Phipps & Bird PB-700 jar tester was used to conduct all experiments. One liter of NF/RO concentrate was added to each jar during treatment. The lime softening procedure was followed as described in Hsu and Singer (2009). The lime doses were 375, 625, and 1250 mg/L as CaO. These doses were achieved by adding 15 mL of concentrated lime slurry of concentration 0.5, 0.74, and 1.49 M, respectively, to the jars. Two limesoda softening procedures were also tested: pre-lime-softening Na2CO3 addition and post-lime-softening Na2CO3 addition. The Na2CO3 doses were chosen based on the calcium concentration remaining in NF-1 and NF-4 concentrates treated by 625 mg/L CaO. Doses of w1.3 g/L Na2CO3 (NF-1) and w2.6 g/L Na2CO3 (NF4) were used. Na2CO3 was added to untreated membrane concentrate (pre-lime-softening) or softened membrane concentrate (post-lime-softening), and mixed at 100 rpm until dissolved. The lime softening procedure was followed using a dose of 625 mg/L CaO. The coagulation procedure was followed as described in Comstock et al. (2010). Ferric sulfate was selected as the coagulant because it showed greater removal of dissolved

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 5 5 e4 8 6 5

organic carbon (DOC) than ferric chloride or aluminum sulfate in previous work by the authors (Comstock et al. (2010)). One limitation of ferric sulfate is the addition of sulfate to the treated concentrate, which is discussed in Sections 3.3 and 4.4. The doses investigated were 1.79, 4.48, and 8.95 mmol/L as Fe(III), which corresponds to 100, 250, and 500 mg/L as Fe(III), respectively. Anion exchange resin was measured as the volume of wet settled resin in a graduated cylinder. The doses were 5, 10, and 20 mL/L MIEX-Cl resin. Virgin resin was regenerated prior to use following Apell and Boyer (2010). The resin was mixed in a NaCl solution for 10 min at 100 rpm and then rinsed in DI water for 10 min at 100 rpm using the jar tester described above. The anion exchange procedure was as follows: mix at 100 rpm for 30 min and settle for 5 min. Samples were collected at 5, 10, and 20 min (no settling) and 30 min (after settling). Samples at 5, 10, and 20 min were removed directly from the jar using a syringe fitted with a PVDF 0.45 mm filter (Millex-HV, Millipore) and analyzed for UV absorbance at 254 nm (UV254) only. All jar tests were conducted by dosing treatment chemicals (i.e., lime, ferric sulfate, and MIEX-Cl resin) in duplicate. All data are the mean values of duplicate samples with error bars showing one standard deviation.

2.4.

Analytical methods

All source water, untreated membrane concentrate, and treated membrane concentrate (at final contact time for MIEX) were analyzed for pH, DOC, UV254, total hardness, and calcium. The following additional measurements were made: source water and untreated membrane concentrate (alkalinity, Sr, Ba, sulfate, and total dissolved solids (TDS)), untreated and treated membrane concentrate (electrical conductivity), softened membrane concentrate (alkalinity and sulfate), and MIEXtreated (final contact time) membrane concentrate (sulfate). Unfiltered sample was used for pH. All other measurements were performed on samples filtered through 0.45 mm nylon membrane filters (Millipore). pH, DOC, and UV254 were analyzed as described in Comstock et al. (2010), and sulfate was measured by ion chromatography as described in Apell and Boyer (2010). The calcium concentration was measured as calcium hardness following Standard Method 3500-Ca D (Clesceri et al. (1989)). However, instead of using a 0.2% Eriochrome Blue Black R indicator, a 0.5% Eriochrome Blue Black R indicator was used to make the color change more noticeable. Dilutions were needed for most samples to prevent precipitation when increasing the pH. Total hardness was determined following Standard Method 2340 (Clesceri et al. (1989)). Due to the high hardness of membrane concentrate, dilutions were necessary for most samples. Calmagite was used as the indicator. The magnesium concentration was calculated as the difference in total hardness and calcium hardness. Alkalinity was determined by following Standard Method 2320 (Clesceri et al. (1989)). Standard 0.02 N sulfuric acid was used as the titrant and bromcresol green was used as the indicator. For all titrations, once the end point was reached, this value was recorded and a few more drops of titrant were added to ensure that the color change was complete. Sr and Ba were analyzed by acidifying samples to pH < 2 with concentrated nitric acid (Trace Metal

4857

Grade, Fisher Scientific) and measured on an ICP-AES (Thermo Jarrell Ash) as described in Method 6010C (U.S. EPA, 2007). An Orion model 115A þ TDS meter equipped with an Orion conductivity cell 011510 was used to measure TDS. Electrical conductivity was measured using an ECTestr11 þ portable electrode (Eutech Instruments).

3.

Results

3.1.

Source waters and NF/RO concentrates

Water treatment conditions relevant to the NF/RO concentrates are given in Table 1. The source waters were from three different aquifer systems in FL, USA. The membrane treatment plants used similar pre-treatment steps and had recovery rates of 80e90%. The maximum theoretical concentration factor (CFmax) was calculated based on the percent recovery of the membrane process with the assumption of 100% rejection of solute by the membrane; solute rejection < 100% would decrease CFmax. The membrane plants used primary NF/RO with disposal of the membrane concentrate. Water chemistry for the source waters and untreated membrane concentrates is given in Table 2. All source waters were fresh groundwater characterized by high total hardness (>200 mg/L as CaCO3), high TDS (>250 mg/L), and high DOC (>7.5 mg C/L; expect source water for RO-1). The membrane concentrates had a composition similar to brackish water, e.g., total hardness > 850 mg/L as CaCO3 and TDS > 900 mg/L. All of the NF concentrates had a very high DOC (30e78 mg C/L). CFs were calculated by dividing the concentration in untreated membrane concentrate by the concentration in source water (results not shown). A limitation of estimating the CFs this way is that chemicals deposited on the membrane surface are not accounted for. Nevertheless, it is useful to compare the concentration-based CFs with CFmax. In general, CFs for divalent cations, sulfate, and DOM (i.e., DOC and UV254) approached CFmax for all membrane concentrates. It is not clear why NF-4 and NF-5 concentrates had CFs for sulfate much greater than CFmax. Both membrane plants used the same antiscalant, which contains the sulfonate functional group. However, the doses of antiscalant are too low to account for the sulfate concentrations in the membrane concentrate. Alkalinity and TDS were generally concentrated to a lesser extent than CFmax. The CFs were consistent with NF/RO having a higher rejection of divalent ions and organic macromolecules than monovalent ions.

3.2.

Precipitative softening

3.2.1.

Lime softening

Results for lime softening jar tests are shown in Figs. 1 and 2. Normalized concentrations of calcium, DOC, and sulfate as a function of lime dose are shown in Fig. 1. pH before and after lime softening is shown in Fig. 2. NF-2, NF-3, NF-5 and RO-1 showed optimum removal of calcium at a dose of 625 mg/L CaO. The corresponding calcium removal was 69%e74% and final pH varied from 6.9 to 10.6. NF-1 and NF-4 showed increasing calcium concentration as the CaO dose was increased from 375 to 1250 mg/L and final pH > 10.5 for all CaO doses. The increasing calcium concentration and high pH

4858

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 5 5 e4 8 6 5

Table 1 e Treatment conditions relevant to NF and RO drinking water plants. Concentrate

Aquifer

Membrane type (Manufacturer)

NF-1

Biscayne

ESNA1-LF (Hydranautics)

NF-2

Biscayne

ESNA1-LF (Hydranautics)

NF-3

Biscayne

NF-4

Tamiami

ESNA1-LF2 (Hydranautics) OSMOMUNIeNFe365 (GE)

NF-5

Tamiami

TFC 8923 S-400 (Koch)

RO-1

Floridan

TFC 9921 S-400 (Koch)

Pre-treatment

Acid addition (sulfuric acid; target pH 5.8); filtration (sand separators and 5 mm cartridge filters) Acid addition (sulfuric acid; target pH 6.6); antiscalant addition (Nalco PC-1850T; applied dose 1.5 mg/L); filtration (5 mm cartridge filters) Filtration (pressure filters and 5 mm cartridge filters) Acid addition (sulfuric acid; target pH 5.2); antiscalant addition (American Water Chemicals A102plus; applied dose 1.7 mg/L); filtration (5 mm cartridge filters) Acid addition (sulfuric acid), antiscalant addition (American Water Chemicals A102plus; applied dose 2.5 mg/L); filtration (5 mm cartridge filters) Acid addition (sulfuric acid; target pH 5.0); antiscalant addition (GE Betz Hypersperse MDC700; applied dose 2.0 mg/L); filtration (manganese greensand pressure filters and 5 mm cartridge filters)

suggest that the added CaO did not precipitate as CaCO3 due to insufficient alkalinity. The alkalinity of NF-1 and NF-4 was 270 and 62 mg/L as CaCO3, respectively, whereas the other concentrates had alkalinity values ranging from 925 to 1170 mg/L as CaCO3. The results for NF-1 and NF-4 are specific to the CaO doses tested and may not be applicable at lower CaO doses. Improved calcium removal for NF-1 and NF-4 was achieved by adding Na2CO3 to the concentrates either pre- or post-lime-softening, as discussed in Section 3.2.2. Magnesium removal was inconsistent at 375 and 625 mg/L CaO because the pH was generally less than 10. Magnesium removal varied from 67% to 98% at 1250 mg/L CaO because the pH was greater than 11 for all concentrates (results not shown). All concentrates showed decreasing DOC concentration with increasing CaO dose. DOC removal was 15%e24% at the optimum dose for calcium removal (i.e., 625 mg/L CaO for NF2, NF-3, NF-5, and RO-1). Corresponding removal data for UV254 was 22%e35%. Furthermore, UV254 removal was greater than DOC removal for all concentrates and all CaO doses. Preferential removal of UV-absorbing DOM during lime softening of membrane concentrate is consistent with favorable interactions between aromatic DOM and CaCO3 (Lin et al. (2005); Russell et al. (2009)). Maximum removal of DOC was achieved at 1250 mg/L CaO at which all membrane

Recovery rate

CFmax

Disposal

80%

5

Deep-well injection

90%

10

Wastewater treatment plant

80e85%

5e6.7

85%

6.7

Wastewater treatment plant Deep-well injection

85%

6.7

Deep-well injection

80%

5

Reclaimed water irrigation system

concentrates had final pH > 11 and showed substantial removal of magnesium. Furthermore, DOC removal was greater than calcium removal at 1250 mg/L CaO for NF-2, NF-3, NF-5, and RO-1. DOM removal during lime softening can occur by multiple mechanisms including adsorption onto CaCO3 surfaces, DOM-calcium complexation and co-precipitation, and DOM adsorption/co-precipitation with magnesium (Russell et al. (2009)). The results suggest that DOC adsorption to CaCO3 and DOM-calcium co-precipitation are the active modes of removal at 375 and 625 mg/L CaO, whereas magnesium is largely responsible for DOC removal at 1250 mg/L CaO. Further research is needed to determine the operative mechanisms for DOM removal during lime softening of NF/RO concentrate due to the complex water chemistry. Sulfate removal was 0%e14% at the optimum dose for calcium removal, and decreasing or increasing lime dose did not affect sulfate removal. Precipitation of CaSO4, and to a lesser extent SrSO4 and BaSO4, were the likely modes of sulfate removal. A limitation of the lime softening experiments was that the mineral composition of the precipitates was not characterized. Mineral composition data could be used to confirm the aqueous removal data presented in this section.

4859

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 5 5 e4 8 6 5

Table 2 e Water quality of source waters and membrane concentrates. Parameter pH

EC

Ie

TDS

Alkalinity

Sulfate

Total hardness

Mg

Ca

Sr

Ba

DOC

UV254

Units a

e eb ec mS/cma mS/cmb mS/cmc Ma Mb Mc mg/La mg/Lb mg/Lc mg/L as mg/L as mg/L as mg/La mg/Lb mg/Lc mg/L as mg/L as mg/L as mg/La mg/Lb mg/Lc mg/La mg/Lb mg/Lc mg/La mg/Lb mg/Lc mg/La mg/Lb mg/Lc mg/La mg/Lb mg/Lc 1/cma 1/cmb 1/cmc

CaCO3a CaCO3b CaCO3c

CaCO3a CaCO3b CaCO3c

NF-1

NF-2

NF-3

NF-4

NF-5

RO-1

6.3 6.6 6.6 nm 1944 1664 e 0.03 0.03 259 979 nm 68 270 250 153 693 585 214 1071 870 0 15 9.4 86 404 333 0.59 2.6 nm 14 56 nm 9.5 45 30 0.35 1.6 1.2

6.8 7.5 7.5 nm 2260 3180 e 0.04 0.05 271 1380 nm 216 1170 1360 28 198 358 242 1596 1822 3.9 54 26 91 550 687 0.71 4.1 nm 30 111 nm 7.7 52 51 0.30 1.8 1.9

7.0 7.8 7.8 nm 1842 1902 e 0.03 0.03 284 940 nm 226 950 940 15 75 71 250 1050 1032 1.7 18 2.7 97 391 409 0.43 1.8 nm 36 102 nm 12 66 56 0.46 2.3 2.1

7.2 5.6 5.5 nm 3700 3720 e 0.06 0.06 407 1880 nm 322 62 66 35 2650 2640 343 2505 2328 16 144 100 111 766 767 0.59 4.5 nm 14 116 nm 8.5 61 58 0.26 1.8 1.8

7.1 7.6 ed nm 2590 ed e 0.04 ed 306 1180 ed 282 1050 ed 8 160 ed 287 1454 ed 11 61 ed 97 481 ed 0.23 1.0 ed 20 85 ed 12 78 ed 0.45 2.7 ed

6.8 7.8 7.7 nm 4170 3850 e 0.07 0.06 540 2190 nm 194 925 950 35 179 151 307 1717 1478 7.9 76 50 110 562 510 0.29 1.6 nm 30 127 nm 2.6 13 12 0.055 0.28 0.27

Not measured (nm). a Source water collected May 2010. b Membrane concentrate collected May 2010; used for lime softening (375, 625, 1250 mg/L CaO), coagulation (100, 250, 500 mg Fe(III)/L), and MIEX (5, 10 mL/L; expect NF-2 at 10 mL/L). c Membrane concentrate collected July 2010; used for lime-soda softening and MIEX (NF-2 at 10 mL/L, 20 mL/L). d Second batch of NF-5 not collected. e Ionic strength (I) ¼ 1.6  105(EC) (Edzwalk and Tobiason (2011)).

3.2.2.

Lime-soda softening

Lime-soda softening experiments were performed using NF-1 and NF-4 as shown in Fig. 3. All lime softening experiments were conducted at 625 mg/L CaO, which was the optimum dose for calcium removal determined in Fig. 1. Both pre- and post-lime-softening Na2CO3 addition resulted in substantial removal of calcium (93%e99%) compared to lime softening without Na2CO3 addition, which increased the calcium concentration. Calcium removal during lime-soda softening was greater than calcium removal from NF-2, NF-3, NF-5 and RO-1 at the optimum CaO dose. The final pH and DOC removal were not affected by Na2CO3 addition. For example, the corresponding pH values for NF-1 were pH 11.5 (no Na2CO3 addition), pH 12 (pre-lime-softening Na2CO3 addition), and pH

11.7 (post-lime-softening Na2CO3 addition). DOC removal in the lime softening with Na2CO3 addition was within w5% of DOC removal for lime softening with no Na2CO3 addition. For example, the corresponding DOC removals for NF-1 were 50% (no Na2CO3 addition), 45% (pre-lime-softening Na2CO3 addition), and 56% (post-lime-softening Na2CO3 addition).

3.3.

Coagulation

The results of the coagulation jar tests using ferric sulfate are shown in Fig. 4. NF-2, NF-3, NF-5, and RO-1 all showed decreasing DOC concentration as the ferric sulfate dose increased from 100 to 500 mg Fe(III)/L, with maximum DOC removal at 500 mg Fe(III)/L. The corresponding maximum DOC

4860

a

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 5 5 e4 8 6 5

14

2 NF-1

NF-2

NF-3

13

NF-4

NF-5

RO-1

12

NF-2 NF-5

NF-3 RO-1

11 pH

Calcium, C/C0

1.5

NF-1 NF-4

1

10 9 8 7

0.5

6 5

0 0

b

250

500

750

1000

0

1250

CaO (mg/L)

250

500

750

1000

1250

CaO (mg/L)

Fig. 2 e pH of untreated and lime softened NF/RO concentrate. Data are the mean of duplicate samples.

1

DOC, C/C0

0.8 0.6 0.4 0.2

NF-1

NF-2

NF-3

NF-4

NF-5

RO-1

0 0

250

500

750

1000

1250

CaO (mg/L)

c

1.2

Sulfate, C/C0

1 0.8

(results not shown). However, dissolved Fe can increase UV254 absorbance (Weishaar et al. (2003)), which can make the results difficult to interpret. The difference in coagulation performance among the concentrates is explained by examining the pH after coagulation as shown in Fig. 4b. Coagulation with Fe(III) is most effective in the pH range of 4.5e6 (Matilainen et al. (2010)). NF1 and NF-4 had much lower alkalinity than the other concentrates (see Table 2), and as a result, the pH decreased substantially in NF-1 and NF-4 with the addition of 100 mg Fe(III)/L. Thus, NF-1 achieved the highest DOC removal at 100 mg Fe(III)/L because the pH was in the effective range for ferric coagulation; DOC removal in NF-1 and NF-4 decreased as the ferric sulfate dose increased because the high doses of ferric sulfate decreased the pH to < 3; and DOC removal in NF2, NF-3, NF-5, and RO-1 increased as the ferric sulfate dose

0.6 0.4

1.4

0.2

NF-1

NF-2

NF-3

NF-4

NF-5

RO-1

Calcium

DOC

1.2 1

0

250

500 750 CaO (mg/L)

1000

1250

Fig. 1 e Final normalized concentrations of (a) calcium, (b) DOC, and (c) sulfate following lime softening of NF/RO concentrate. Data are the mean of duplicate samples; error bars show one standard deviation (typically inside symbol).

C/C0

0

0.8 0.6 0.4 0.2 0 NF-1 a

removals were 80%e84% for NF-2, NF-3, and NF-5 and 58% for RO-1. In contrast, NF-1 and NF-4 had a minimum DOC concentration at the lowest ferric sulfate dose, and increasing DOC concentration as the ferric sulfate dose increased. As a result, maximum DOC removal was achieved at 100 mg Fe(III)/L. The corresponding maximum DOC removals were 81% (NF-1) and 57% (NF-4). Removal of UV254-absorbing organic matter was generally greater than DOC removal

NF-1 b

NF-1 c

NF-4 a

NF-4 b

NF-4 c

Lime softening procedure

Fig. 3 e Effect of Na2CO3 addition on final normalized concentrations of calcium and DOC following lime softening of NF-1 and NF-4 concentrates. Lime softening procedure was 625 mg/L CaO: (a) no Na2CO3 addition, (b) pre-lime softening Na2CO3 addition, and (c) post-lime softening Na2CO3 addition. Data are the mean of duplicate samples; error bars show one standard deviation.

4861

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 5 5 e4 8 6 5

1

NF-1 NF-3 NF-5

DOC, C/C0

0.8

NF-2 NF-4 RO-1

0.6 0.4 0.2 0 0

b

100

200

300

400

500

Ferric sulfate (mg Fe(III)/L) 9

NF-1 NF-3 NF-5

8

NF-2 NF-4 RO-1

7

pH

6 5 4 3 2 0

100

200

300

400

500

Ferric sulfate (mg Fe(III)/L)

Fig. 4 e Ferric sulfate coagulation of NF/RO concentrate: (a) final normalized concentration of DOC and (b) pH. Data are the mean of duplicate samples; error bars for DOC show one standard deviation (typically inside symbol).

increased because the pH approached the effective pH range for ferric coagulation. Removal of calcium by coagulation was <10% for all concentrates and ferric sulfate doses (results not shown). Sulfate was not measured for samples treated by coagulation because using ferric sulfate increased the sulfate concentration. The results described in this section are specific to ferric sulfate; additional research is needed to determine the applicability of these results to ferric chloride and aluminum sulfate.

3.4.

Anion exchange

Kinetic data for uptake of UV254-absorbing organic matter by MIEX-Cl resin is shown in Fig. 5. Results are shown for 10 mL/L MIEX-Cl resin; results for the other resin doses are not shown because they followed a similar trend. Greater than 50% (i.e., 52%e82%) of UV254-absorbing organic matter was removed during the first 5 min of treatment. UV254 removal increased by an additional 3e31% from 5 min to 30 min. Total UV254 removal at 30 min was 67%e93%. NF-3 had the greatest UV254 removal at 30 min, and also had the lowest sulfate concentration and lowest TDS. NF-4 and RO-1 had the lowest removals of UV254, and had the highest sulfate concentration (NF-4) and the highest TDS (RO-1).

Normalized concentrations of DOC and sulfate as a function of MIEX-Cl resin dose (at 30 min contact time) are shown in Fig. 6. DOC removal was 43%e83% at 10 mL/L MIEX-Cl resin (see Fig. 6a). MIEX resin showed preferential removal of UV254absorbing organic matter relative to DOC, which is well documented in the literature (Boyer et al. (2011)). Decreasing the MIEX-Cl resin dose to 5 mL/L decreased DOC removal by 4e12%, and doubling the MIEX resin dose to 20 mL/L increased DOC removal by 4e16%. Similar to Fig. 5, NF-3 (lowest sulfate concentration, lowest TDS) had the greatest DOC removal, while NF-4 (highest sulfate) and RO-1 (highest TDS) had the lowest DOC removals at all resin doses. Sulfate removal varied widely as a function of membrane concentrate chemistry and MIEX-Cl resin dose (see Fig. 6b). For example, sulfate removal was 8%e59% at 10 mL/L MIEX-Cl resin. An important consideration for ion exchange treatment of high ionic strength waters, such as membrane concentrate, is the ion exchange capacity of a resin. MIEX-Cl resin doses of 5, 10, and 20 mL/L have the ion exchange capacity to remove approximately 125, 250, and 500 mg/L of sulfate, respectively, based on MIEX-Cl resin capacity of 0.52 meq/mL (Boyer and Singer (2008)). The initial sulfate concentration in all concentrates (except NF-3) was greater than the resin capacity at 5 mL/L, and as a result, sulfate removal was <10% except for NF-3 (31%). The initial sulfate concentration in NF-1 and NF-4 was still greater than the resin capacity at 20 mL/L, which limited sulfate removal. For example, 19% sulfate removal from NF-4 at 20 mL/L was approximately equal to the maximum capacity of the resin. DOC removal was greater than sulfate removal for all concentrates and all resin doses. MIEX resin showed <10% calcium removal for all concentrates and resin doses (expect NF-2). Calcium removal for NF-2 was 9% (5 mL/L) to 16% (20 mL/L). The mode of calcium removal was likely anion exchange via complexation with DOM, however, further research is needed to verify this mechanism.

1 0.8

UV254, C/C0

a

NF-1 (1.6)

NF-2 (1.9)

NF-3 (2.3)

NF-4 (1.8)

NF-5 (2.7)

RO-1 (0.28)

0.6 0.4 0.2 0 0

10

20

30

Time (min) Fig. 5 e Normalized UV254 absorbance following MIEX-Cl treatment (10 mL/L) of NF/RO concentrate. UV254 absorbance (1/cm) of untreated concentrate given in parentheses. Data are the mean of duplicate samples; error bars show one standard deviation (typically inside symbol).

4862

a

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 5 5 e4 8 6 5

1

NF-1 NF-3 NF-5

DOC, C/C0

0.8

NF-2 NF-4 RO-1

0.6 0.4 0.2 0 0

5

10

15

20

MIEX-Cl (mL/L)

b

4.2.

Effect of water chemistry and antiscalants

1.2 1

Sulfate, C/C0

on maximum removal, was anion exchange (19%e81% at 20 mL/L MIEX-Cl for all concentrates) > softening (no removal to 13% at 1250 mg/L CaO for all concentrates). Using anion exchange to remove sulfate is an important contribution of this work because sulfate is typically considered a competitive species with DOC (e.g., Fu and Symons (1990); Boyer and Singer (2006)). However, the anion exchange results show >50% DOC removal and high sulfate removal. Coagulation increased the sulfate concentration because ferric sulfate was chosen as the coagulant based on maximum DOC removal. Lime softening was the only process to remove calcium, DOC, UV254-absorbing organic matter, and sulfate. However, anion exchange showed greater removals of DOC, UV254, and sulfate than lime softening.

0.8 0.6 0.4 NF-1 NF-3 NF-5

0.2

NF-2 NF-4 RO-1

0 0

5

10

15

20

MIEX-Cl (mL/L) Fig. 6 e Final normalized concentrations of (a) DOC and (b) sulfate following MIEX-Cl treatment (at 30 min contact time) of NF/RO concentrate. Data are the mean of duplicate samples; error bars show one standard deviation (typically inside symbol). Standard deviation for RO-1 (sulfate, 20 mL/L) [ 0.34; all other standard deviations equal to 0.04 or less.

4.

Discussion

4.1.

Summary of removal performance

Calcium, DOC, alkalinity, sulfate, and TDS are all important parameters to consider for treatment of membrane concentrate as illustrated in this study. Another factor that has been reported to affect the treatment efficiency of membrane concentrate, especially precipitation processes, is antiscalants (e.g., Greenlee et al. (2010a, 2010b, 2011)). Many water utilities add antiscalants to the feed water prior to membrane treatment to prevent mineral precipitation on membrane surfaces. Greenlee et al. (2010b) showed that antiscalants could interfere with precipitation during softening of synthetic membrane concentrate. However, this study showed that water chemistry had a greater impact on treatment efficiency than antiscalants. For example, NF-4 and NF-5 both used AWC A102plus as an antiscalant, yet the concentrates responded very differently to lime softening and coagulation treatment due to the source water chemistry (i.e., NF-4 had low alkalinity). Additionally, NF-1 and NF-3 did not use antiscalants, yet the concentrates also responded very differently to lime softening and coagulation treatment because NF-1 had low alkalinity. Furthermore, MIEX-Cl treatment achieved greater DOC removal in NF-3 compared to NF-1, which was due to NF3 having a lower sulfate concentration and lower TDS than NF-1. Although antiscalants are expected to have a secondary effect of treatment efficiency, the results of this work show that water chemistry should be the first consideration when evaluating a process for NF/RO concentrate treatment.

4.3. The following summary is based on the experimental range tested and may not be applicable to conditions outside of this range. Lime softening achieved optimum calcium removal at 625 mg/L CaO (69%e74% in NF-2, NF-3, NF-5, and RO-1 concentrate), and lime-soda softening achieved the highest removal of calcium (93%e99% at 625 mg/L CaO in NF-1 and NF4 concentrate). Coagulation and anion exchange showed minor removal of calcium (<10%). The order of DOC removal, based on maximum removal, was anion exchange (51%e87% at 20 mL/L MIEX-Cl for all concentrates) z coagulation (NF1 ¼ 81% and NF-4 ¼ 57% at 100 mg Fe(III)/L; 58%e84% at 500 mg Fe(III)/L for remaining concentrates) > softening (41%e66% at 1250 mg/L CaO for all concentrates). Furthermore, all treatment processes showed greater removal of UV254-absorbing organic matter than DOC. The order of sulfate removal, based

Treating NF/RO concentrate to source water quality

Figs. 7 and 8 synthesize the results from Section 3 by comparing the water composition of treated NF/RO concentrate with the corresponding source water. The motivation for these figures is that treating NF/RO concentrate to source water quality would allow for secondary NF/RO to operate under the same conditions as primary NF/RO, which are assumed to be desirable. Treated NF/RO concentrate with water chemistry similar to source water should also be easier to dispose of. Calcium sulfate (CaSO4$2H2O) was used as a surrogate to evaluate the inorganic water quality of the NF/ RO concentrates and source waters (see Fig. 7). Precipitation of calcium sulfate on membrane surfaces is well-studied as described in the Introduction. The caption for Fig. 7 provides details for the calculation of the saturation index (SI). The

NF/RO concentrate Saturation Index

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 5 5 e4 8 6 5

10

processes reduced the DOC concentration in the treated concentrate. Both coagulation (ferric sulfate) and anion exchange (MIEX) were able to reduce the DOC concentration in a majority of the concentrates to the approximate DOC of the source water. Thus, precipitation and anion exchange were able to treat NF/RO concentrate to water quality approaching that of source water. This is an important contribution of the work because a majority of the previous literature is focused solely on precipitation as the method for intermediate treatment of membrane concentrate.

Untreated Precipitation (*625 mg/L CaO) Anion exchange (20 mL/L MIEX) y=x Saturation line

1

0.1

0.01

0.001 0.001

4.4. 0.01

Fig. 7 e Comparison of calcium sulfate (CaSO4$2H2O) saturation index in NF/RO concentrate and source water. Saturation index [ ion activity product 3 Ks0; ion activity product [ (gD)2[Ca2D][SO24 ], where gD is activity coefficient for divalent ion estimated using Gu ¨ ntelberg equation; ionic strength (Table 2) used to estimate gD; gD assumed equal to 1 in source water; log Ks0 [ L4.58 (Stumm and Morgan (1996)). *Lime softening at 625 mg/L CaO for NF-2, NF-3, NF5, and RO-1; lime-soda softening (pre-softening Na2CO3 addition) at 625 mg/L CaO for NF-1 and NF-4.

untreated NF/RO concentrates have a greater SI than the source water because both calcium and sulfate were concentrated by the membrane process. Both precipitation (lime softening) and anion exchange (MIEX) were able to reduce the SI in the treated concentrate, and for several concentrates, the SI approached that in the source water. DOC was used to evaluate the organic water quality of the NF/RO concentrates and source waters (see Fig. 8). DOC in the untreated NF/RO concentrates was concentrated 5 times or greater relative to the source waters. All three treatment

NF/RO concentrate DOC (mg/L)

Untreated Precipitation (1250 mg/L CaO) Coagulation (*500 mg Fe(III)/L) Anion exchange (20 mL/L MIEX) y=x y = 5x

90 80 70 60

Conclusions

This work provides a systematic study of common drinking water treatment processes for the treatment of NF and RO membrane concentrates. The treatment processes investigated were lime softening, ferric sulfate coagulation, and anion exchange using MIEX-Cl resin. The major findings are as follows:

40 30 20 10 0 5 10 Source water DOC (mg/L)

The advantages of the treatment processes tested in this study are described in Sections 3 and 4. Furthermore, Figs. 7 and 8 summarize the major impacts of the treatment processes on inorganic and organic water chemistry. Despite these advantages, there are limitations to the processes tested. Both lime softening and ferric sulfate were conducted at high chemical doses, which would generate substantial quantities of sludge. Lime softening and coagulation require the appropriate alkalinity and pH to perform correctly, and as a result, may require additional chemicals. Beneficial reuse of the sludge, e.g., mineral recovery or land application, would lessen the burden of disposing of the sludge. Anion exchange generates a small volume but highly concentrated waste brine. A major motivation for intermediate treatment of NF/RO concentrate is to avoid having to dispose of waste membrane concentrate. Anion exchange would reduce the total volume of waste to dispose of; nevertheless, there would still be a liquid waste stream. Both the sludge and anion exchange brine could further accumulate chemicals that were concentrated in the membrane concentrate, such as arsenic. This is problematic because it can make an undesirable waste product into a hazardous waste. Thus, this brief discussion illustrates that the conservation of mass principle must be considered when evaluating processes for intermediate treatment of NF/RO concentrate.

5.

50

0

Limitations of treatment processes tested

0.1

Source water Saturation Index

100

4863

15

Fig. 8 e Comparison of DOC in NF/RO concentrate and source water. Line y [ x represents equal DOC concentrations in NF/RO concentrate and source water. Line y [ 5x represents concentration factor of 5 in NF/RO concentrate. *Coagulation at 500 mg Fe(III)/L for NF-2, NF-3, NF-5, and RO-1, and 100 mg Fe(III)/L for NF-1 and NF-4.

 Optimum lime softening conditions for calcium removal showed the following order of removal: calcium > UV254 > DOC > sulfate. Excess lime softening showed a decrease in calcium removal, but an increase in removal of magnesium and DOC. Concentrate containing insufficient alkalinity to form CaCO3(s) required the addition of Na2CO3 to achieve calcium removal. DOC removal was the same with and without Na2CO3 addition. Lime softening was able to reduce the calcium sulfate saturation index in treated concentrate to approximately the saturation index of source water.

4864

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 5 5 e4 8 6 5

 For coagulation, the order of removal was UV254 > DOC. This trend was not applicable to concentrates with low alkalinity at high ferric sulfate doses due to dissolution of Fe(III). Coagulation is not recommended for treatment of membrane concentrate because the method is sensitive to alkalinity and pH, and does not remove inorganic chemicals.  For anion exchange treatment, the order of removal was UV254 > DOC > sulfate. Anion exchange was a more robust process compared to lime softening and coagulation in terms of variations in concentrate chemistry, such as alkalinity and pH. One caution was that the ion exchange capacity of a resin can be a limiting factor for sulfate removal from high ionic strength waters. Anion exchange removed both organic and inorganic chemicals, and as a result, decreased the calcium sulfate saturation index and reduced the DOC concentration to the level of source water.  Calcium was removed by lime softening only. The order of DOC removal was anion exchange z coagulation > lime softening. The order of sulfate removal was anion exchange > lime softening.  Alkalinity, sulfate, and TDS adversely affected treatment efficiency, with lime softening and coagulation affected by alkalinity and anion exchange affected by sulfate and TDS. The initial composition of the membrane concentrate had a greater effect on treatment efficiency than the presence of antiscalants.

Acknowledgments The authors thank the staff at the water treatment facilities and Mike Witwer from CH2M Hill for their assistance with sample collection and process information. The authors also thank Orica Watercare for providing MIEX resin.

references

Apell, J.N., Boyer, T.H., 2010. Combined ion exchange treatment for removal of dissolved organic matter and hardness. Water Research 44 (8), 2419e2430. Boyer, T.H., Singer, P.C., 2006. A pilot-scale evaluation of magnetic ion exchange treatment for removal of natural organic material and inorganic anions. Water Research 40, 2865e2876. Boyer, T.H., Singer, P.C., 2008. Stoichiometry of removal of natural organic matter by ion exchange. Environmental Science & Technology 42, 608e613. Boyer, T.H., Graf, K.C., Comstock, S.E.H., Townsend, T.H., 2011. Magnetic ion exchange treatment of stabilized landfill leachate. Chemosphere 83 (9), 1220e1227. Clesceri, L.S., Greenberg, A.E., Trussell, R.R. (Eds.), 1989. Standard methods for the examination of water and wastewater. APHA. Comstock, S.E.H., Boyer, T.H., Graf, K.C., Townsend, T.G., 2010. Effect of landfill characteristics on leachate organic matter properties and coagulation treatability. Chemosphere 81 (7), 976e983. Cornelissen, E.R., Chasseriaud, D., Siegers, W.G., Beerendonk, E.F., van der Kooij, D., 2010. Effect of anionic fluidized ion exchange (FIX) pre-treatment on nanofiltration (NF) membrane fouling. Water Research 44 (10), 3283e3293.

Dialynas, E., Mantzavinos, D., Diamadopoulos, E., 2008. Advanced treatment of the reverse osmosis concentrate produced during reclamation of municipal wastewater. Water Research 42 (18), 4603e4608. Edzwalk, J.K., Tobiason, J.E., 2011. Chemical principles, source water composition, and watershed protection. In: Edzwald, J. K. (Ed.), Water Quality and Treatment: a Handbook on Drinking Water. McGraw-Hill Inc, New York, NY. Fu, P.L.K., Symons, J.M., 1990. Removing aquatic organic substances by anion exchange resins. Journal American Water Works Association 82, 70e77. Gabelich, C.J., Williams, M.D., Rahardianto, A., Franklin, J.C., Cohen, Y., 2007. High-recovery reverse osmosis desalination using intermediate chemical demineralization. Journal of Membrane Science 301 (1e2), 131e141. Gabelich, C.J., Rahardianto, A., Northrup, C.R., Yun, T.I., Cohen, Y., 2011. Process evaluation of intermediate chemical demineralization for water recovery enhancement in productionscale brackish water desalting. Desalination 272, 36e45. Greenlee, L.F., Lawler, D.F., Freeman, B.D., Marrot, B., Moulin, P., 2009. Reverse osmosis desalination: water sources, technology, and today’s challenges. Water Research 43, 2317e2348. Greenlee, L.F., Testa, F., Lawler, D.F., Freeman, B.D., Moulin, P., 2010a. The effect of antiscalant addition on calcium carbonate precipitation for a simplified synthetic brackish water reverse osmosis concentrate. Water Research 44, 2957e2969. Greenlee, L.F., Testa, F., Lawler, D.F., Freeman, B.D., Moulin, P., 2010b. Effect of antiscalants on precipitation of an RO concentrate: metals precipitated and particle characteristics for several water compositions. Water Research 44 (8), 2672e2684. Greenlee, L.F., Testa, F., Lawler, D.F., Freeman, B.D., Moulin, P., 2011. Effect of antiscalant degradation on salt precipitation and solid/liquid separation of RO concentrate. Journal of Membrane Science 366, 48e61. Heijman, S.G.J., Guo, H., Li, S., van Dijk, J.C., Wessels, L.P., 2009. Zero liquid discharge: Heading for 99% recovery in nanofiltration and reverse osmosis. Desalination 236 (1e3), 357e362. Hsu, S., Singer, P.C., 2009. Application of anion exchange to control NOM interference on lime softening. Journal American Water Works Association 101 (6), 85e94. 14. Jarusutthirak, C., Mattaraj, S., Jiraratananon, R., 2007. Influence of inorganic scalants and natural organic matter on nanofiltration membrane fouling. Journal of Membrane Science 287 (1), 138e145. Jin, X., Huang, X.F., Hoek, E.M.V., 2009. Role of specific ion interactions in seawater RO membrane fouling by alginic acid. Environmental Science & Technology 43, 3580e3587. Li, Q.L., Elimelech, M., 2004. Organic fouling and chemical cleaning of nanofiltration membranes: measurements and mechanisms. Environmental Science & Technology 38, 4683e4693. Lin, Y.P., Singer, P.C., Aiken, G.R., 2005. Inhibition of calcite precipitation by natural organic material: Kinetics, mechanism, and thermodynamics. Environmental Science & Technology 39, 6420e6428. Lin, C.J., Shirazi, S., Rao, P., Agarwal, S., 2006. Effects of operational parameters on cake formation of CaSO4 in nanofiltration. Water Research 40, 806e816. Lin, N.H., Shih, W.Y., Lyster, E., Cohen, Y., 2011. Crystallization of calcium sulfate on polymeric surfaces. Journal of Colloid and Interface Science 356, 790e797. Matilainen, A., Vepsa¨la¨inen, M., Sillanpa¨a¨, M., 2010. Natural organic matter removal by coagulation during drinking water treatment: a review. Advances in Colloid and Interface Science 159, 189e197. Mohammadesmaeili, F., Badr, M.K., Abbaszadegan, M., Fox, P., 2010a. Byproduct recovery from reclaimed water reverse

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 5 5 e4 8 6 5

osmosis concentrate using lime and soda-ash treatment. Water Environment Research 82, 342e350. Mohammadesmaeili, F., Badr, M.K., Abbaszadegan, M., Fox, P., 2010b. Mineral recovery from inland reverse osmosis concentrate using isothermal evaporation. Water Research 44, 6021e6030. Ning, R.Y., Troyer, T.L., 2009. Tandom reverse osmosis process for zero-liquid discharge. Desalination 237 (1e3), 238e242. Qu, D., Wang, J., Wang, L.L., Hou, D.Y., Luan, Z.K., Wang, B.Q., 2009. Integration of accelerated precipitation softening with membrane distillation for high-recovery desalination of primary reverse osmosis concentrate. Separation and Purification Technology 67 (1), 21e25. Rahardianto, A., Gao, J.B., Gabelich, C.J., Williams, M.D., Cohen, Y., 2007. High recovery membrane desalting of low-salinity brackish water: Integration of accelerated precipitation softening with membrane RO. Journal of Membrane Science 289, 123e137. Roberts, D.A., Johnston, E.L., Knott, N.A., 2010. Impacts of desalination plant discharges on the marine environment: a critical review of published studies. Water Research 44, 5117e5128. Russell, C.G., Lawler, D.F., Speitel, G.E., 2009. NOM coprecipitation with solids formed during softening. Journal American Water Works Association 101 (4), 112e124. Shih, W.Y., Rahardianto, A., Lee, R.W., Cohen, Y., 2005. Morphometric characterization of calcium sulfate dihydrate (gypsum) scale on reverse osmosis membranes. Journal of Membrane Science 252, 253e263. Shon, H.K., Vigneswaran, S., Cho, J., 2008. Comparison of physicochemical pretreatment methods to seawater reverse osmosis:

4865

Detailed analyses of molecular weight distribution of organic matter in initial stage. Journal of Membrane Science 320, 151e158. Stumm, W., Morgan, J.J., 1996. Aquatic Chemistry. John Wiley & Sons, Inc., New York. Tran, T., Bolto, B., Gray, S., Hoang, M., Ostarcevic, E., 2007. An autopsy study of a fouled reverse osmosis membrane element used in a brackish water treatment plant. Water Research 41, 3915e3923. U.S. EPA, 2007. Method 6010C: Inductively Coupled PlasmaAtomic Emission Spectrometry (Revision 3, February 2007). In SW-846. U.S. EPA, Washington, DC. Van der Bruggen, B., Lejon, L., Vandecasteele, C., 2003. Reuse, treatment, and discharge of the concentrate of pressuredriven membrane processes. Environmental Science & Technology 37 (17), 3733e3738. Verliefde, A.R.D., Cornelissen, E.R., Heijman, S.G.J., Petrinic, I., Luxbacher, T., Amy, G.L., Van der Bruggen, B., van Dijk, J.C., 2009. Influence of membrane fouling by (pretreated) surface water on rejection of pharmaceutically active compounds (PhACs) by nanofiltration membranes. Journal of Membrane Science 330, 90e103. Weishaar, J.L., Aiken, G.R., Bergamaschi, B.A., Fram, M.S., Fujii, R., Mopper, K., 2003. Evaluation of specific ultraviolet absorbance as an indicator of the chemical composition and reactivity of dissolved organic carbon. Environmental Science & Technology 37 (20), 4702e4708. Yoon, S.H., Lee, C.H., Kim, K.J., Fane, A.G., 1998. Effect of calcium ion on the fouling of nanofilter by humic acid in drinking water production. Water Research 32, 2180e2186.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 6 6 e4 8 7 6

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Evaluation and application of anion exchange resins to measure groundwater uranium flux at a former uranium mill site Valerie Stucker a, James Ranville a,*, Mark Newman b, Aaron Peacock d, Jaehyun Cho c, Kirk Hatfield b a

Department of Chemistry and Geochemistry, Colorado School of Mines, Golden, CO, USA Department of Civil and Coastal Engineering, University of Florida, Gainesville, FL, USA c Department of Environmental Engineering Sciences, University of Florida, Gainesville, FL, USA d Microbial Insights Inc., Rockford, TN, USA b

article info

abstract

Article history:

Laboratory tests and a field validation experiment were performed to evaluate anion

Received 16 December 2010

exchange resins for uranium sorption and desorption in order to develop a uranium

Received in revised form

passive flux meter (PFM). The mass of uranium sorbed to the resin and corresponding

22 June 2011

masses of alcohol tracers eluted over the duration of groundwater installation are then

Accepted 23 June 2011

used to determine the groundwater and uranium contaminant fluxes. Laboratory based

Available online 13 July 2011

batch experiments were performed using Purolite A500, Dowex 21K and 21K XLT, Lewatit S6328 A resins and silver impregnated activated carbon to examine uranium sorption and

Keywords:

extraction for each material. The Dowex resins had the highest uranium sorption, followed

Ion exchange

by Lewatit, Purolite and the activated carbon. Recoveries from all ion exchange resins were

Contaminant

in the range of 94e99% for aqueous uranium in the environmentally relevant concentra-

Passive flux meter

tion range studied (0.01e200 ppb). Due to the lower price and well-characterized tracer

Sorption

capacity, Lewatit S6328 A was used for field-testing of PFMs at the DOE UMTRA site in Rifle, CO. The effect on the flux measurements of extractant (nitric acid)/resin ratio, and uranium loading were investigated. Higher cumulative uranium fluxes (as seen with concentrations > 1 ug U/gram resin) yielded more homogeneous resin samples versus lower cumulative fluxes (<1 ug U/gram resin), which caused the PFM to have areas of localized concentration of uranium. Resin homogenization and larger volume extractions yield reproducible results for all levels of uranium fluxes. Although PFM design can be improved to measure flux and groundwater flow direction, the current methodology can be applied to uranium transport studies. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Uranium is a contaminant of concern at many DOE former mining and milling sites. With the passing of the Uranium

Mill Tailings Radiation Control Act (UMTRCA) in 1978, tailings were removed to repositories and the surface soil was remediated from 24 inactive uranium mill sites. In Rifle, Colorado, the site of a former uranium and vanadium mill,

* Corresponding author. E-mail address: [email protected] (J. Ranville). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.06.030

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 6 6 e4 8 7 6

the contamination remains in the soils and shallow groundwater due to uranium leached from the tailings prior to removal. Monitored natural attenuation (MNA) was the chosen method for remediating the uranium below ground surface (USEPA, 1999). Surface remediation was completed in 1996, but groundwater treatment is ongoing. Uranium, vanadium, and selenium are the main contaminants of concern at this site (DOE, 2008). The Rifle, CO site is situated directly adjacent to the Colorado River and has a semi-arid climate, receiving on average 300 mm of precipitation. The first 20e30 feet below ground surface is Colorado River alluvium, which covers the sands, silts and clays of the Wasatch Formation. A more complete description of the Rifle mill site geology and hydrology can be found elsewhere (Yabusaki et al., 2007). By natural flushing, soluble uranium discharges under a natural gradient to the Colorado River where it is further diluted to background levels. Initial groundwater modeling predicted this process to take 100 years to reach background levels (DOE, 2008). Decrease in uranium concentration is, however, occurring slower than anticipated and other remediation strategies are being investigated. Since 2002, biostimulation experiments have been performed at the site to stimulate microbial reduction of uranium (VI), which is the soluble, mobile form, to immobile uranium (IV) using acetate injections (Anderson et al., 2003; Lovley et al., 1991; Ortiz-Bernad et al., 2004; Yabusaki et al., 2007). To determine the effectiveness of the biostimulation, it is necessary to measure the uranium concentrations upgradient and downgradient of the injection wells. It is also valuable to know the rate of groundwater flow. A rapid flow combined with a high soluble contaminant concentration is of greater concern than a slow moving plume of similar contaminant concentration. To be considered successful, bioremediation usually needs to achieve low soluble contaminant concentrations, however a high concentration may be allowed to remain if little or no flow occurs. Examining the concentration alone may lead to an incomplete understanding of the system. Measuring contaminant flux (Jc), which is the product of the volumetric water flux, or specific discharge (q0) and the fluxaveraged contaminant concentration (c) provides better insight into the outcome of remediation efforts (Hatfield et al., 2004). There are limitations with the current methods for estimating flux. Concentrations are typically measured independently of the specific discharge, possibly at separate sampling times or location, leading to significant variations and errors in flux calculations. Flux estimates generally rely on modeled flow rates or laborious and intrusive hydrologic tests (tracer injections or pump tests). To acquire more reliable information, the passive flux meter (PFM) was developed to provide simultaneous direct measures of the specific discharge (qo) and contaminant mass flux (Jc) (Annable et al., 2005; Hatfield et al., 2004). Contaminant measurements made using PFM represent cumulative local fluxes, which can then be used to estimate flux-averaged aqueous concentrations (c), according to Equation (1), for comparison to measured aqueous concentrations. Jc ¼ qo c

(1)

4867

PFM schematics and the details of the calculations are provided by Hatfield et al. (2004). The PFMs were initially used to measure the flux of organic contaminants and used silver impregnated (antimicrobial) granulated activated carbon (GAC Ag) as the sorbent material (Annable et al., 2005). Previously, anion exchange resins such as Dowex 21K and 1-X8, Purolite A500, A600 and A520E, and Lewatit K 6367 have been proven to remove uranium from aqueous solutions (Kolomiets et al., 2005; Phillips et al., 2008) under a large range of pH conditions, from acidic to alkaline, with the most effective sorption observed near neutral pH to alkaline conditions. Chelating resins (Merdivan et al., 2001; Pesavento et al., 2003), such as Chelex 100 and Amberlite XAD-16, and cation exchange resins, such as Dowex-50 (Schumann et al., 1997), have been shown to be most effective at sorbing aqueous uranium in acidic solutions. Due to the near neutral pH at the Rifle site and the expected dominance of anionic uranyl carbonato species, this work focused on anion exchange resins previously shown to be effective. The goal of the overall project was to develop a PFM to quantify aqueous uranium fluxes at the Rifle site that can be used to evaluate the effectiveness of biostimulation by comparing upgradient and downgradient uranium fluxes. To have an effective PFM, the material must quantitatively sorb all of the uranium passing through it, be capable of releasing the sorbed uranium upon extraction, and predictably retain and release the chosen tracers, preferably without microbial growth occurring on the resin. Because the PFM will be used to evaluate the effectiveness of biostimulation an additional constraint for this application was that elution of resident tracers must not affect results of the other experiments being performed simultaneously at the site. To keep total remediation and monitoring costs at a minimum, resin cost is also a factor for making the final choice. Therefore resins will be evaluated for uranium sorbed and desorbed, tracer capacity and finally, cost. In this specific study uranium speciation and water chemistry conditions (pH, alkalinity, major ions, etc.) were investigated to find the best sorbent material for use in the PFMs at Rifle, CO.

2.

Methods

A sample of background Rifle groundwater was analyzed for speciation. Modeling tests were performed to determine the aqueous uranium (VI) speciation using Visual MINTEQ, Version 3.0, over a pH range of 3e10 using the Multi-problem/ Sweep function. Only aqueous species were included and sorption and precipitation were ignored. Data on metal and ligand concentrations were obtained from inductively coupled plasma atomic emission spectroscopy (ICP-AES) using a Perkin Elmer 3000 for cation metals, ion chromatography (IC) using a Dionex ICS-90 system for major anions, a Shimadzu TOC analyzer for dissolved organic carbon (DOC) and carbonate was determined by alkalinity titration. The concentrations of these major components in Rifle groundwater can be found in Table 1. These concentration values were used as input for the model and updated constants for the calcium and magnesium species provided by Dong and Brooks (2006) were used to update the database.

4868

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 6 6 e4 8 7 6

Table 1 e Concentrations of components used to model uranium speciation. Only major components affecting uranium species are included. Component

Concentration

U(VI) Ca2þ Naþ Mg2þ SO2 4 CO2 3 Cl NO 3 DOC

2.1.

226 ug/L 262 mg/L 202 mg/L 128 mg/L 795 mg/L 168 mg/L 192 mg/L 12 mg/L 4.5 mg/L

Materials

Numerous sorbents have been developed and used for the removal of uranium from water at near neutral pHs. Organic resins with quaternary amine sites, strong anion-exchange resins, bind uranyl complexes strongly and are available from several manufacturers, for example the Dowex 21K and Purolite A series. (Barton et al., 2004; Chanda and Rempel, 1992a,b; Jelinek and Sorg, 1988; Kolomiets et al., 2004; Vaaramaa et al., 2000) Purolite A500, Dowex 21K and 21K XLT, Lewatit S6328 A and silver impregnated granular activated carbon (GAC Ag) were investigated as possible sorbent materials in the PFMs. The Lewatit resin has been previously used for anion capture in PFMs (Cho et al., 2007). These resins were expected to have good sorption for the uranyl carbonate anion complexes predicted in the Rifle groundwater. Each resin was examined for uranium sorption and desorption and tracer elution properties relative to price. Table 2 shows a comparison of the resin properties and prices at the time of the experiment (fall 2008). Resins were used in the chloride form and rinsed thoroughly five times with deionized water and allowed to air dry. No cleaning procedure was necessary for the GAC Ag. Adsorption studies were initially done in new, sterile 15 mL polypropylene centrifuge tubes (BD, Franklin Lakes, NJ, PN 352196) using water with added sodium bicarbonate (1 mM) to simulate Rifle alkalinity and pH. Some wall-sorption of uranium was observed using these tubes. After 24 h, polypropylene had sorbed 8% of the uranium present, and 25%

after 72 h. Polystyrene tubes (BD, PN 352095) were then used for comparison. Polystyrene tubes showed less uranium sorption to the walls than did the polypropylene tubes with 17% sorbed after 72 h. Uranium sorption to the tubes increased over time for both materials; however, based on the near complete uranium recovery and mass balance from the resin after material transfer, we concluded that uranium preferentially sorbed to the resin regardless of time or tube material and the tube sorption seen earlier in the blanks was deemed negligible in actual samples. Remaining experiments were done using polystyrene as a precaution to ensure most uranium was sorbing to the resin. Cleaning with nitric acid (3% v/v) was effective in removing all uranium from tubes. All uranium solutions were made by diluting a 1000  3 ppm U (in 2% nitric acid) ICP standard (High-Purity Standards, Charleston, SC). Dilutions were done using NanoPure deionized water for neutral pH, and trace metal grade nitric or hydrochloric acids (Fisher Scientific) for the lower pH standards. Sodium bicarbonate, used to create the groundwater simulant, and the silver nitrate used to impregnate the GAC were ACS grade.

2.2.

Adsorption batch studies

Sorbent materials were weighed (100  5 mg) into centrifuge tubes. Uranium standard solutions at concentrations of 5, 10, 50, 100 and 300 ppb were prepared in 1 mM sodium bicarbonate (pH 7.3  0.1) to simulate Rifle groundwater. Uranium concentrations in Rifle groundwater average around 200 ppb. The effects of pH, which will influence the aqueous uranium speciation, were tested by lowering the pH with nitric and hydrochloric acid to pH 3.8. 10 mL of uranium solution were added to each tube containing resin. Each concentration was examined in triplicate at both pHs. The samples were placed on a shaker for 24 h to allow mixing and ion exchange. After the 24-h equilibration period, the resins were allowed to settle and the overlying solutions were decanted into new centrifuge tubes, leaving the resins for the next extraction step. In some cases this procedure was repeated with a 72 h equilibration time to verify that complete ion exchange had been achieved over the shorter time period. Since the PFMs will be in place in the aquifer for weeks at a time, testing the sorption for shorter times was deemed unnecessary based on estimated residence times (of groundwater passing through PFM) that were greater

Table 2 e Comparison of sorbent materials tested using data provided by manufacturer fact sheets. *Prices listed are values given for educational groups and may not reflect current industry prices. Resin Purolite A500 Dowex 21K 1630 Dowex 21K XLT Lewatit S6328 A GAC Ag

Type Type 1 strong base anion Type 1 strong base anion Type 1 strong base anion Type 1 strong base anion activated carbon

Functional Group Exchange Capacity Ionic Form Water Content Specific Price* $/ft3 (min.eq/L) (as shipped) (%) Gravity quaternary amine

1.15

chloride

53e58

1.08

163.00

quaternary amine

1.2

chloride

50e58

1.06

449.50

quaternary amine

1.4

chloride

50e60

1.08

384.50

quaternary amine

1.0

chloride

58e63

1.06

150.00

carbon

unknown

none

5

266.00

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 6 6 e4 8 7 6

than 24 h. The solutions were analyzed for uranium by inductively coupled plasma-mass spectrometry (ICP-MS).

2.3.

Extraction

A 1% nitric acid solution (10 mL) was added to all tubes with resins and shaken overnight. By lowering the pH to less than 1, cation, becomes the dominant form and the uranyl, UO2þ 2 should be extracted from the anion exchange resin. Furthermore, the high level of nitrate should also cause desorption through competition for anion exchange sites. After 24 h, the acidic solution was decanted, verified to still be acidic (pH <2) by indicator paper, and analyzed by ICP-MS. The same extraction procedure was used for field PFMs, using larger volumes as described later. A uranium mass balance was computed using the values of desorbed uranium and the uranium remaining in solution following the sorption experiments.

2.4.

ICP-MS analysis

A Perkin Elmer Elan 6100 ICP-MS was used for all analyses. All samples were acidified with nitric acid prior to analysis and introduced simultaneously into the argon plasma with a 40 ppb In internal standard using a Gilson peristaltic pump. A check standard was used every ten samples to verify instrument performance to be within 10% of the true value. Instrument calibration was done using uranium standards made in 1% nitric acid (Fisher Optima grade). Running conditions complied with manufacturer recommendations and standard quality controls were practiced.

2.5.

Resin selection

The ideal resin should have a high uranium sorption capacity that is linear over the concentration range of interest. Linearity allows for predictability in performance and ease in calculations. Linear partition sorption isotherms were developed from the results of each resin batch experiment. The linear partition equation is given as: Ca ¼ Kd Cs

(2)

where Ca is the concentration of uranium adsorbed to the resin (ug/kg). Kd is the sorption equilibrium coefficient (kg/kg) and Cs is the concentration of uranium remaining in solution (ug/kg). Sorption (determined from Kd and average percentage sorbed over linear range) and desorption properties (determined by percentage of sorbed uranium extracted with acid), were evaluated. The data was also analyzed using Langmuir and Freundlich isotherms to investigate non-linearity. Fitting the data to the Freundlich isotherm: Ca ¼ KCns

(3)

provided a better fit than the Langmuir isotherm. If the exponential term, n, is equal to one, this equation reduces to the linear isotherm equation. This linear form was found to be sufficient to describe resin sorption with the exception of the Purolite resin.

4869

Along with the use of ion exchange resin for uranium sorption, a second media consisting of granular activated carbon (GAC) pretreated with a suite of short chain alcohol tracers is used to measure volumetric water flux as presented in Hatfield et al. (2004) and Annable et al., 2005. Additional details of device construction and sampling is provided in Section 3.4 Field Application of this paper. Flow through experiments were performed in a benchscale three-dimensional aquifer model using similar methods as presented in Hatfield et al. (2004) and Cho et al., 2007. These tests were used to confirm performance of the resin and estimate the requisite flow convergence terms as presented in Hatfield et al. (2004).

2.6.

Field studies

Resins were tested for microbial growth to ensure minimal interference with uranium sorption and desorption. As a precaution for field samples, resins were pretreated with silver nitrate solution to add ionic silver to the resin for antibacterial properties. The solutions were then analyzed by ICP-AES for residual silver (the difference assumed to be sorbed to the resin). Silver was added to the resin to be comparable to the silver concentration (w0.03% by weight Ag:resin) on the GAC-Ag used in the initial PFM studies reported by Hatfield et al. (2004). PFMs were installed in Rifle over a three week period in the summer and the late fall of 2009. The meters were removed from the wells, and a number of vertical segments were separated and homogenized by complete mixing of each sample before being split for uranium, tracer, and microbial analyses. For the data presented in this paper PFM were deployed in 4-inch wells, and each PFM was constructed with six alternating segments of granular activated carbon (GAC) and Lewatit resin (three segments of GAC and three segments of Lewatit). The bottom segment of each PFM was composed of GAC. The PFM were retrieved and sampled with a deployment length of 23 days. Tracer analysis from the GAC segments were used for groundwater flux estimates, and the Lewatit segments were used for uranium flux estimates. Uranium was extracted from the resins using additions of 10 mL of 1% nitric acid, which were replicated until no additional uranium removal was observed as determined by ICPMS. Initially, small amounts of resin were used for the extracts to minimize waste produced and mirror the lab experiments. Samples of 100e150 mg and 1e1.5 g of resin were extracted for uranium and concentration values were compared between the different resin sample masses. For the 100 mg resin samples, this took one acid extraction with a second to verify complete uranium removal. Poor reproducibility was observed at lower concentrations (<1 ug U/g resin). Three acid extractions were necessary for the 1 g resin samples, but this larger resin amount had better reproducibility, regardless of uranium concentration or flux. In extractions performed later in the project, 4 g resin samples were extracted in 40 mL of 1% nitric acid, which provided sufficient acid extractant for multiple analyses for uranium and other groundwater anions, while also giving a more representative sample size for improved reproducibility at low

4870

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 6 6 e4 8 7 6

fluxes. The uranium concentration data presented here will be used to calculate the uranium flux through the PFM.

3.

Results

3.1.

Speciation

Visual MINTEQ computations suggest that over the environmentally relevant pH range (7e8), most of the uranium is present as uranyl carbonato anions or uranyl calcium triscarbonato neutral species (d in Fig. 1). In fact, across the pH range, all of the dominant species are neutral (b,c, and d) as uranyl sulfate, UO2SO4, at lower pHs, uranyl carbonate, UO2CO3, at the mid range, and uranyl calcium triscarbonato above pH 5.5. The uranyl cation, UO2þ 2 , becomes more important at the lower pHs (see a in Fig. 1), and this is the basis for removal of uranium from the anion exchange resins using nitric acid. Despite the speciation being predominantly neutral aqueous species, the resins still show strong uranium sorption, demonstrating that anion speciation is not necessarily needed for sorption to an anion exchange resin. More complete geochemical modeling of this system, including sensitivity and error analyses on these samples, can be found in Leavitt et at 2010 (in review).

3.2.

Adsorption

All of the sorbent materials tested showed nearly complete uranium adsorption at the near neutral pH in the artificial groundwater sodium bicarbonate solutions. The GAC Ag, which had the lowest sorption at the higher concentrations as seen in Fig. 2, still sorbed an average of 89% of the uranium in solution. Both of the Dowex resins showed very good linear sorption across the range of concentrations investigated. These resins sorbed an average of 99% of the uranium in solution. The only

difference between the resins is the mesh size, from 16 to 30 for the 21K, and around 30 (but with more uniform size) for the 21K XLT. At the examined range of uranium concentrations, there is no advantage to the smaller resin beads. Lewatit S6328 A also had a linear sorption isotherm, but with a slightly lower sorption percentage, averaging 95% with a Kd of 2000 (kg/kg) for aqueous uranium. These linear sorption isotherms are desirable over the concentration range, so the amount sorbed can easily be used to calculate mass flux. Purolite A500 shows increased uranium sorption at higher concentrations, a pattern that is different from the other resins. Repeating the procedure for Purolite over 72 h rather than 24 h showed the same trend, with a slight increase in sorption at higher concentrations. While equilibrium had been reached for the other resins after 24 h, the Purolite resin was still sorbing uranium; an average of 94% of the uranium sorbed in 24 h, and 97% had sorbed over 72 h. This same trend was seen in sulfuric acid solutions used by Kolomiets et al. (2005). The other resins tested by this group had reached equilibrium within 25 h, while the Purolite A500 had not (Kolomiets et al., 2005). While the linear isotherm has a good R2 value (0.91), it was not a linear trend, evidenced by a trend in the residuals, which was not seen with the other resins. Using the Freundlich isotherm equation, this was confirmed with an exponent statistically greater than 1 (1.6  0.1). Currently, we do not have a clear explanation for this phenomenon. One possibility is a surface precipitation process that could be tested using infrared or Raman spectroscopy, but this was not done in this experiment since there were other resins with simpler adsorption trends to use in the PFM. An overview of the sorption calculations is shown in Table 3. While there could be some improvement in the R squared values and the errors on the slope for our sorption data, the errors were deemed unimportant relative to instrumental errors. With the exception of Purolite A500, each of the other sorbent materials produced Freundlich exponents near one, with only minor statistical differences. The residuals from the

Fig. 1 e Rifle speciation modeled using Visual MINTEQ.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 6 6 e4 8 7 6

4871

Fig. 2 e Sorption isotherms for all materials tested comparing mass uranium sorbed to the resin (mg U/g resin) versus to mass uranium remaining in solution (mg/kg). Each point represents an average of the three replicates with error bars omitted for clarity (values given in text). Lines indicate the linear sorption isotherm from which Kd values were calculated.

linear partition isotherms were also investigated to verify linearity and showed no trend that would indicate a deviation from linearity. Thus, the linear isotherms were presented. When resins were tested using groundwater collected from the Rifle, CO field site, all of the anion exchange resins performed in a similar manner as in the lab experiments with quantitative sorption at 99% sorbed. This matched the sorption performance seen with Dowex resins in laboratory studies, suggesting improved sorption with all resins in natural waters. Each of the resins maintained the original groundwater pH of 7.2, with only a minor drop to 7.1 for the Dowex resins. The activated carbon did not follow the same pattern of uranium sorption in Rifle groundwater. The pH increased slightly to 7.4. It sorbed around 33% of the calcium in the water, but it removed only 5% of the uranium. This result ruled out the GAC as a sorbent choice for the field PFMs, but it could still be considered in the use of tracers for groundwater flux measurements. This work focused on U(VI) sorption since the solubility of U(IV) is so low and was not expected to influence the results. It has been suggested that

Table 3 e Results of resin sorption studies. Kd was determined from the linear slopes, and errors on those slopes are shown along with the R squared value of the linear fit. Resin Purolite A500 Dowex 21K 1630 Dowex 21K XLT Lewatit S6328 A GAC Ag

Kd

Error

R squared

% Sorbed

1400 8800 12000 2000 800

160 1500 1700 70 130

0.91 0.82 0.86 0.97 0.84

94 99 99 95 89

U(IV) complexation to organic ligands can increase solubility, but since the groundwater samples showed complete sorption and we are only interested in total uranium quantification, U(IV) vs U(VI) speciation was not investigated further. At a lower pH (3.8), uranium cations dominate speciation, and so little uranium adsorption would be expected to sorb to the resins. This was not the case for the Purolite and Lewatit resins when pH was adjusted using nitric acid. With each of the other resins, much lower masses of uranium were sorbed at pH 3.8 than at neutral pH with the same bicarbonate concentrations. Purolite A500 and Lewatit S6328 A both showed the same 94e95% sorption at the lower pH. Gu et al. (2004) had examined Purolite resins and suggested that this phenomenon may be due to the concentration of nitrate on the resin surface allowing for aided uranium adsorption since the resin, has a very high affinity for nitrate anions. All of the resins have quaternary amine groups, but they suggest that the triethylamine groups have a higher affinity for nitrate than the trimethylamine groups present in the Dowex resins. In another experiment pH was lowered using HCl to test if concentrated nitrate aided uranium sorption, and very different results were obtained. As was expected with the acidic pH, there was very little uranium sorption to the resins. As an additional test to confirm this nitrate-aided sorption, we exchanged the chloride ions already present on the resins with nitrate ions by rinsing the resins in excess sodium nitrate before testing the resins for uranium sorption at the lower pH. Uranium sorption was at 12% for Lewatit and 20% for Purolite; these lower sorption results do not suggest a nitrate aided sorption. One possibility is that there might be a uranylenitrate complex in solution that is sorbing to the resin rather than the uranyl species sorbing to concentrated nitrates on the resin. More work is necessary to fully understand these results.

4872

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 6 6 e4 8 7 6

A 1% nitric acid solution (pH w1) was effective at extracting uranium from all resins. Essentially all of the sorbed uranium was desorbed into the acidic solution. This was not the case for the GAC Ag. Repeat acid extractions as well as bicarbonate and carbonate extraction solutions were used, but only a combined, inconsistent 40e70% of the total uranium could be removed from the GAC using the different extraction solutions. This result again confirms the elimination of the GAC Ag as the sorbent of choice for the uranium PFMs since it would be nearly impossible to quantify uranium passing through the meter without a more efficient and reliable extraction method.

a

3.3.

Based on sorption capacity, tracer studies and resin price, Lewatit S6328 A was the chosen resin for the field PFMs. GAC had been eliminated due to poor extraction recovery and lower sorption than the other resins, and Purolite was eliminated due to the unexplained sorption phenomenon at higher concentrations. Table 3 presents a summary of the results from the resin sorption tests. While the Dowex resins exhibited better uranium sorption, the price difference for the 3e4% improvement in sorption, given the lower uranium concentrations at Rifle, was deemed unnecessary for this

Uranium Extracted from PFM 4 100 mg Total 100 mg extraction 1 100 mg extraction 2 100 mg extraction 3 1 g Total 1 g extraction 1 1 g extraction 2 1 g extraction 3 1 g extraction 4

4

ug U / g resin

Resin choice

2

0 Section 2

Section 4

Section 6

Resin sample ID

b

Uranium Extracted from PFM 6

ug U / g resin

0.8

100 mg Total 100 mg extraction 1 100 mg extraction 2 100 mg extraction 3 1 g Total 1 g extraction 1 1 g extraction 2 1 g extraction 3 1 g extraction 4

0.6

0.4

0.2

0.0 Section 2

Section 4

ZSection 6

Resin sample ID Fig. 3 e a) Uranium extracted from PFM#4 and b) from PFM #6. Since Lewatit was alternated with GAC to obtain tracer information, resin samples were 2, 4 and 6 (top) in the depth sequence. The left-most bar (black) in each series represents the total uranium concentration removed from the resin after all extractions using the two different masses.

4873

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 6 6 e4 8 7 6

3.4.

Field application

For the initial field test six PFM were deployed in six 4-inch wells. Each PFM was constructed using a new flux pod design having six alternating segments of granular activated carbon (GAC) and Lewatit resin (three pods or segments of GAC and three pods of Lewatit). The bottom segment of each PFM was GAC. The flux pods are self-contained segments of sorbent that can be stacked on a common center tube. The pods can then be retrieved individually for sampling on site or packaged for shipment to an analytical laboratory for remote sampling and analysis. The objective for using individual segments of sorbent was two-fold: to test the individual capability of the Lewatit resin to capture uranium under field conditions while using the GAC to estimate groundwater flux and to test the new modular flux pod design. This was a necessary step to test the efficiency of the Lewatit as a sorbent under field conditions. However, the alternating segment design does not allow for measurement of water flux and uranium flux at the exact same vertical location. One design modification implemented following this experiment was to develop a three-layer prototype which will allow both groundwater and uranium flux measurements at the same vertical location and mixed media sorbents are now also being tested. The PFM were retrieved and sampled after a deployment of 23 days (approximately three weeks). Tracer analysis from the GAC segments were used for groundwater flux estimates, and the Lewatit segments were used for uranium flux estimates. For comparison and validation of PFM performance, water samples were collected in each well four days prior to PFM deployment. Silver additions to the Lewatit resin matched the concentrations on the alternating GAC layers. Solutions decanted after silver addition indicated that the Lewatit resin had in fact been coated to 0.03% silver by weight. This prevented microbial growth on the resin while installed at Rifle. After coating with silver (98% of silver sorbed to the resin), lab experiments were conducted to ensure that the silver was not being released from the resin and uranium sorption remained the same as without added silver. Treating Rifle groundwater with the silver coated resin showed release of silver to be 0.3% of

Uranium Extracted from Replicate Samples

Total Extraction 1 Extraction 2 Extraction 3

4

ug U / g resin

experiment, especially given the improved performance in the actual groundwater samples. If it was necessary to get more accurate flux measurements, without having to make a minor correction for incomplete sorption, either Dowex resin would be a better choice. Using this method, the greatest uncertainty a flux measurement lies in the up to 10% error of the ICP-MS, not the slight loss that may be seen by using the Lewatit resin over one of the Dowex resins. Since the resins did not reach a sorption capacity, it would be difficult to use these results to choose the best resin for use in high uranium flux contaminated sites. One consideration for testing samples is that as the U fills the strongest resin binding sites, the apparent partitioning K will change (decrease) and the actual U flux would be underestimated. However, based on the levels of U bound to the resin (Fig. 3), we are in fact well within the experimental range (Fig. 2), so this is not a problem for this work.

2

0 High 1

High 2

Medium 1 Medium 2

Low 1

Low 2

Resin sample Fig. 4 e The results of using larger amounts of resin (4 g) and acid (40 mL) for extractions. Samples were chosen for this test based on the initial results so that a range of concentrations would be presented. High, medium and low represent the concentration/flux ranges expected from the first set of tests.

Fig. 5 e Measured fluxes for PFM 4 showing vertical distribution of uranium mass flux and volumetric water flux (specific discharge or Darcy Velocity). The PFM fluxaveraged uranium concentration for this well was 167.25 mg/L which compares well with the aqueous uranium concentration (180.95 mg/L) measured in the well.

4874

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 6 6 e4 8 7 6

Table 4 e Summary of uranium mass flux, volumetric water flux (specific discharge or Darcy Velocity), flux-averaged concentration for all PFM segments in all wells, and measured aqueous concentrations for all wells. From Passive Flux Meter (PFM)

Minimum Maximum Standard Deviation Mean

Uranium Mass Flux (mg/cm2 day)

Specific discharge (Darcy Velocity) (cm/day)

Flux averaged uranium concentration (mg/L)

Measured aqueous uranium concentration (mg/L)

0.13 3.57 0.84 0.13

2.27 6.83 1.38 5.00

57.62 615.31 140.31 188.43

171.00 192.30 8.84 180.63

Sample size (n) for each data set: For PFM: n ¼ 3 PFM segments (samples) per 6 wells ¼ 18, For aqueous samples ¼ n ¼ 1 sample per 6 wells ¼ 6.

the sorbed silver and 99% of aqueous uranium sorbed to the resin, which is unchanged from the sorption to resin with no silver addition, making it suitable for use at the field site. We are unsure of the mechanism by which Agþ sorbs to an anion exchange resin, perhaps it is related to the matrix material. A complete understanding of this was not necessary, however, since we were only interested in just adding it for an antimicrobial purpose that would not interfere with uranium sorption. Uranium was extracted from the resins taken from the Rifle PFMs by the same 1% nitric acid solution used for lab studies. As mentioned before, small quantities of resin were compared to a larger sample to determine homogeneity in sample and method during the first sampling trips. Fig. 3 shows the results of the multiple extractions. If the samples are homogeneous, the final total concentration of uranium extracted from both quantities of resin should be the same. There were some minor discrepancies that can be attributed to either ICP-MS error or sample heterogeneity. PFMs that were exposed to higher water and uranium fluxes were more likely to have a homogeneous composition of uranium on the resin (totals match in Fig. 3a). Lower flows yielded locally concentrated uranium in the PFM on the upgradient side and were more difficult to completely homogenize leading to inconsistent concentration values in samples less than 1 g (see Fig. 3b). Using a large quantity of resin, or the whole sample for the extraction, will improve the accuracy of the uranium masses used for flux calculations. Samples were reanalyzed using the 4 g/40 mL method listed, and replicate samples showed excellent reproducibility with a maximum difference between replicates of 5%, even at the lower concentrations of uranium (Fig. 4). All samples from subsequent trips were analyzed in this manner. Future PFM design may examine water flow direction (which changes with the river water stage) by cross-sectional quartering of samples. This may also be useful for samples with localized concentrated uranium.

3.5.

Summary of field results

Observed variations and trends in measured fluxes were consistent amongst wells and similar to the results provided for PFM 4 in Fig. 5, which shows the vertical distribution of uranium mass flux and volumetric water flux (specific discharge). It can be observed that vertical trends in uranium

flux tend to agree with water flux, and the resulting fluxaveraged concentration for this well (167.25 mg/L) compares favorably with the aqueous concentration measured in the well (180.95 mg/L). Table 4 provides a summary of uranium mass flux, volumetric water flux (specific discharge), flux-averaged concentration for all PFM segments, and measured aqueous concentrations. The summary provides a comparison of fluxaveraged uranium concentrations estimated from all PFM segments across all wells to aqueous uranium concentrations measured in each well. One key point to observe is the similarity of mean concentrations while noting the larger standard deviation for PFM-based flux-averaged estimates. This is because the aqueous concentrations represent the volumetric average within the entire borehole, while the flux-averaged concentrations show much higher resolution with respect to variation with depth (as shown in Fig. 5). The similar mean values for flux-averaged and measured aqueous uranium concentration provides a positive validation for PFM measurement of uranium flux while also providing additional detail with regard to the vertical variation of flux within the well and surrounding formation. As seen in Fig. 3a, when larger resin samples were extracted for uranium, the second extraction removed the most uranium, whereas the smaller resin amounts demonstrated expected extraction patterns, with the most uranium removed in the first extraction and less in each subsequent extraction. This pattern was seen only in the first sampling trip in samples with high uranium fluxes. This is possibly due to high concentrations of organics binding to the resin and preventing release of uranium in the first extraction. To determine other anions sorbed to the resin and possible desorption interferences, a larger resin sample and volume of acid were used to have sufficient volume for both ICP-MS and ICP-AES analyses. This larger volume also saw better consistency in low flux samples and was used for all future extractions. It was also observed that significant sulfate was removed, with less in each subsequent extraction. Based on the acid strength and nitrate concentration, there were still enough exchangeable nitrate anions to remove all uranium and sulfate together, even with the 2:1 nitrate:sulfate molar charge ratio, supporting the theory that organics are binding and blocking the sites. However, if the binding strength of sulfate to Lewatit resin is much higher than that of nitrate, an excess of 5 times as many exchangeable ions of nitrate to

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 6 6 e4 8 7 6

sulfate may not have been sufficient, also contributing to the pattern of uranium removal.

4.

Conclusions

Anion exchange resins provide an effective material for use in a passive flux meter under the water chemistry conditions at the old mill site in Rifle, CO. The Dowex 21K resins had the highest uranium sorption capacity, but are more costly than some of the other resins (i.e. Lewatit) which have sufficient uranium sorption for this purpose and concentration range. Despite geochemical computations indicating mostly neutral uranium species, anion exchange resins were very useful as uranium sorbents. These resins may be good for anionic and neutral species, but more work is needed to understand the sorption mechanisms and speciation covering a larger pH range. Coating the resins with silver nitrate does not affect uranium sorption, and it prevents microbial growth on the resins which could negatively impact the flux measurements and concentrations obtained. PFMs that were placed in high flux areas at the Rifle field site allowed for simple, reliable flux measurements due to homogeneity in the resin samples removed from the PFM; uranium passed through the whole PFM. At sampling locations where uranium fluxes were low the mass of extracted uranium was much more variable from small resin samples and therefore contributed to the error in the flux calculations. When homogenization was incomplete, some of these small resin samples were taken from the upgradient side with more uranium and others were from the downgradient side with little, if any, uranium. The best extraction method found for good reproducibility with sufficient sample size and minimal waste uses 40 mL of a 1% nitric acid solution and 4 g of resin. This finding will be useful in future PFM applications to quantify uranium mass contaminant fluxes in groundwater. Results from an initial field experiment demonstrated that calculated flux-averaged uranium concentrations compared well with measured aqueous uranium concentrations under field conditions, which provided positive validation for use of ion exchange resins for quantifying uranium flux. Ultimately, these flux measurements will provide insight into the mobility of uranium and effectiveness of current remediation strategies employed at the contaminated former mill site in Rifle, Colorado.

Acknowledgments This work was funded by the Department of Energy’s Environmental Remediation Sciences Program (ERSP), U.S. Department of Energy (Grant Number DE-FG02-08ER64585) and through a fellowship provided by the Department of Education’s Graduate Assistance in Areas of National Need (GAANN) program. Special thanks go to Melanie Newman for help with preparing extractions, Steve Cabaniss for providing a background on ion exchange resins, and Janet Leavitt for advice on speciation calculations.

4875

references

Annable, M.D., Hatfield, K., Cho, J., Klammler, H., Parker, B.L., Cherry, J.A., Rao, P.S.C., 2005. Field-scale evaluation of the passive flux meter for simultaneous measurement of groundwater and contaminant fluxes. Environmental Science and Technology 39, 7194e7201. Anderson, R.T., Vrionis, H.A., Ortiz-Bernad, I., Resch, C.T., Long, P. E., Dayvault, R., Karp, K., Marutzky, S., Metzler, D.R., Peacock, A., White, D.C., Lowe, M., Lovley, D.R., 2003. Stimulating the in situ activity of Geobacter species to remove uranium from the groundwater of a uranium-contaminated aquifer. Applied Environonmental Microbiology 69, 5884e5891. Barton, C.S., Stewart, D.I., Morris, K.S., Bryant, D.E., 2004. Performance of three resin-based materials for treating uranium-contaminated groundwater within a PRB. Journal of Hazardous Materials 116, 191e204. Chanda, M., Rempel, G.L., 1992a. Removal of uranium from acidic sulfate-solution by ion-exchange on poly(4-vinylpyridine) and polybenzimidazole in protonated sulfate form. Reactive Polymers 17, 159e174. Chanda, M., Rempel, G.L., 1992b. Uranium sorption behavior of a macroporous, quaternized poly(4-vinylpyridine) resin in sulfuric-acid medium. Reactive Polymers 18, 141e154. Cho, J.Y., Annable, M.D., Jawitz, J.W., Hatfield, K., 2007. Passive flux meter measurement of water and nutrient flux in saturated porous media: bench-scale laboratory tests. Journal of Environmental Quality 36, 1266e1272. DOE, 2008. Rifle, Colorado, Processing Sites and Disposal Site. http://www.lm.doe.gov/Rifle/Disposal/Documents.aspx (accessed 28.07.08). Dong, W.M., Brooks, S.C., 2006. Determination of the formation constants of ternary complexes of uranyl and carbonate with alkaline earth metals (Mg2þ, Ca2þ, Sr2þ, and Ba2þ) using anion exchange method. Environmental Science and Technology 40, 4689e4695. Gu, B.H., Ku, Y.K., Jardine, P.M., 2004. Sorption and binary exchange of nitrate, sulfate, and uranium on an anionexchange resin. Environmental Science and Technology 38, 3184e3188. Hatfield, K., Annable, M., Cho, J., Rao, P.S.C., Klammler, H., 2004. A direct passive method for measuring water and contaminant fluxes in porous media. Journal of Contaminant Hydrology 75, 155e181. Jelinek, R.T., Sorg, T.J., 1988. Operating a small full-scale ionexchange system for uranium removal. Journal American Water Works Association 80, 79e83. Kolomiets, D.N., Konopleva, L.V., Sheremet’ev, M.V., Golubeva, T. E., Shatalov, V.V., 2004. Improvement of the sorption technology for extracting uranium from solutions and pulps. Atomic Energy 97, 463e467. Kolomiets, D.N., Troshkina, I.D., Sheremet’ev, M.F., Konopleva, L.V., 2005. Sorption of uranium from sulfuric acid leaching solutions by strongly basic anion exchangers. Russian Journal of Applied Chemistry 78, 722e726. Lovley, D.R., Phillips, E.J.P., Gorby, Y.A., Landa, E.R., 1991. Microbial reduction of uranium. Nature 350, 413e416. Merdivan, M., Duz, M.Z., Hamamci, C., 2001. Sorption behaviour of uranium(VI) with N, N-dibutyl-N’-benzoylthiourea impregnated in Amberlite XAD-16. Talanta 55, 639e645. Ortiz-Bernad, I., Anderson, R.T., Vrionis, H.A., Lovley, D.R., 2004. Vanadium respiration by Geobacter metalireducens: novel strategy for in situ removal of vanadium from groundwater. Applied Environmental Microbiology 70, 3091e3095.

4876

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 6 6 e4 8 7 6

Pesavento, M., Biesuz, R., Alberti, G., Sturini, M., 2003. Characterization of the sorption of uranium(VI) on different complexing resins. Analytical and Bioanalytical Chemistry 376, 1023e1029. Phillips, D.H., Gu, B., Watson, D.B., Parmele, C.S., 2008. Uranium removal from contaminated groundwater by synthetic resins. Water Research 42, 260e268. Schumann, D., Andrassy, M., Nitsche, H., Novgorodov, A.F., Bruchertseifer, H., 1997. Sorption behaviour of uranium on cation and anion exchange resins from HCl/HF-containing aqueous solutions: model experiments for the determination of chemical properties of element 106 (Seaborgium). Radiochimica Acta 79, 217e220.

USEPA, 1999. In: Use of Monitored Natural Attenuation at Superfund, RCRA Corrective Action, and Underground Storage Tank Sites. OSWER. Vaaramaa, K., Lehto, J., Jaakkola, T., 2000. Removal of U-234, U238, Ra-226, Po-210 and Pb-210 from drinking water by ion exchange. Radiochimica Acta 88, 361e367. Yabusaki, S.B., Fang, Y., Long, P.E., Resch, C.T., Peacock, A.D., Komlos, J., Jaffe, P.R., Morrison, S.J., Dayvault, R.D., White, D.C., Anderson, R.T., 2007. Uranium removal from groundwater via in situ biostimulation: field-scale modeling of transport and biological processes. Journal of Contaminant Hydrology 93, 216e235.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 7 7 e4 8 8 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Limited filamentous bulking in order to enhance integrated nutrient removal and effluent quality Wen-De Tian a,b, Wei-Guang Li a,b,*, Hui Zhang a,b, Xiao-Rong Kang a,b, Mark C.M. van Loosdrecht c,d a

School of Municipal and Environmental Engineering, Harbin Institute of Technology, Harbin 150090, China State Key Laboratory of Urban Water Resource Environment, Harbin Institute of Technology, Harbin 150090, China c Department of Biotechnology, Delft University of Technology, Julianalaan 67, 2628 BC Delft, The Netherlands d KWR Watercycle Research Institute, Groningenhaven 7, 3422 PE Nieuwegein, The Netherlands b

article info

abstract

Article history:

Limited filamentous bulking has been proposed as a means to enhance floc size and make

Received 23 December 2010

conditions more favorable for simultaneous nitrification/Denitrification (SND). Moreover

Received in revised form

a slightly heightened SVI is supposed to increase the removal of small particulates in the

23 June 2011

clarifier. Integrated nitrogen, phosphorus and COD removal performance under limited

Accepted 25 June 2011

filamentous bulking was investigated using a bench-scale plug-flow enhanced biological

Available online 6 July 2011

phosphorus removal (EBPR) reactor fed with raw domestic wastewater. Limited filamentous bulking in this study was mainly induced by low DO levels, while other influencing

Keywords:

factors associated with filamentous bulking (F/M, nutrients, and wastewater characteris-

Limited filamentous bulking

tics) were not selective for filamentous bacteria. The optimum scenario for integrated

Dissolved oxygen

nitrogen, phosphorus and COD removal was achieved under limited filamentous bulking

Settleability (SVI)

with an SVI level of 170e200 (associated with a DO of 1.0e1.5 mg/L). The removal effi-

Simultaneous nitrification

ciencies of COD, TP and NHþ 4 eN were 90%, 97% and 92%, respectively. Under these

and denitrification

conditions, the solideliquid separation was practically not affected and sludge loss was

EBPR

never observed. A well-clarified effluent with marginal suspended solids was obtained. The results of this study indicated the feasibility of limited filamentous bulking under low DO as a stimulation of simultaneous nitrification/denitrification for enhancing nutrient removal and effluent quality in an EBPR process. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Enhanced biological phosphorus removal (EBPR) in activated sludge systems characterized by high removal efficiency, economy, environmentally-friendly operation, and potential phosphorus recovery (Barat and van Loosdrecht, 2006; Martı´ et al., 2010), has become a popular and widespread technology in wastewater treatment plants (WWTPs). However,

filamentous sludge bulking has been reported for the EBPR process (Vaiopoulou et al., 2007) leading to sludge loss and poor solideliquid separation, and therefore result in subsequent upsets and deterioration in removal performance. Abundant previous researches were mainly focused on the study of the prevention, control and modeling of filamentous bulking in various activated sludge systems (Cenens et al., 2000; Martins et al., 2004b; Gulez and de los Reyes, 2009).

* Corresponding author. School of Municipal and Environmental Engineering, Harbin Institute of Technology, Harbin 150090, China. Tel./fax: þ86 451 8628 3003. E-mail addresses: [email protected] (W.-D. Tian), [email protected] (W.-G. Li), [email protected] (M.C.M. van Loosdrecht). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.06.034

4878

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 7 7 e4 8 8 4

Nomenclature WWTP A/O SBR UCT EBPR A2O BCFS EPS F/M A/V SND

wastewater treatment plant anoxic-oxic sequencing batch reactor university of cape town enhanced biological phosphorus removal anaerobic-anoxic-oxic biologisch-chemische-fosfaat-stikstof verwijdering extracellular polymeric substances food/micro-organisms ratio surface-to-volume simultaneous nitrification and denitrification

Specific (Martins et al., 2003a, b; Wanner et al., 2010) and nonspecific (Liao et al., 2004; Martins et al., 2004a) methods as well as various biological selectors (Van Loosdrecht et al., 1998; Vaiopoulou and Aivasidis, 2008) were developed to prevent and control filamentous bulking. Still the costs and the need for chemicals and operational control are the intractable issues, and the general cost effective and easy control solution has not been adopted by the plant operators. Recently, Peng et al. (2008) advocated an energy-saving method of limited filamentous bulking under low DO condition for the first time using anoxic-oxic (A/O) process and domestic wastewater. Thereafter Guo et al. (2010) developed the energy-saving theory and method of limited filamentous bulking, which minimizes energy consumption by taking advantage of higher oxygen transfer rate obtained under low DO level. However, hitherto the utilization of limited filamentous bulking induced by controlling low DO levels for the EBPR process and its influence on overall process performance were marginally documented. The lower DO concentration not only maintains a favorable anoxic environment but also can be favorable for simultaneous nitrification and denitrification (SND). Many researchers have been attracted by the simultaneous nitrification and denitrification (SND) technique because of its simplified process design and smaller anoxic zone, as well as no requirements of external carbon source and alkalinity while minimizing the need for sludge recycles (Ajay et al., 2006; Fu et al., 2009). The biological and physical explanations for SND are the coexistence of denitrifiers and autotrophic nitrifiers and oxygen gradients within activated sludge flocs caused by the limitation of oxygen diffusion (Guo et al., 2005; Chiu et al., 2007). The oxygen gradients in biological floc lead to an interior anoxic microenvironment, which facilitates SND. The larger floc diameter has advantages to limit relative penetration depth of oxygen in the floc and therefore generate oxygen gradients and microbial process stratification, which has been described in the literatures (Chu et al., 2004; Li and Bishop, 2004; Pe´rez et al., 2005). Andreadakis (1993) showed that microenvironments within the floc for SND was better formed with the floc size of 50e100 mm than 10e70 mm. Pochana and Keller (1999) found that the nitrogen removal efficiency via SND was increased by 31% when the average floc size increased from 40 mm to 80 mm. Some researches assumed that a bit larger floc diameter was likely to promote the SND due to diffusional limitation of oxygen in

ORP BOD COD TN SVI MLSS SRT MLVSS TP SS HRT DO FISH

oxidation-reduction potential biochemical oxygen demand, mg L1 chemical oxygen demand, mg L1 total nitrogen, mg L1 sludge volume index, ml g1 mixed liquid suspended solids, g L1 solid retention time, d mixed liquor volatile suspended solids, g L1 total phosphate, mg L1 suspended solids, mg L1 hydraulic retention time, dissolved oxygen, mg L1 fluorescence in situ hybridization

the floc (Zhu et al., 2007; Guo et al., 2009). Dissolved oxygen is important for development of filamentous microorganism (Martins et al., 2003b, 2004a). Having a limited growth of filamentous bacteria will also achieve a better effluent quality due to the irregular filamentous morphology which gives a better filtering out of suspended particles (Wile´n and Balme´r, 1999; Guo et al., 2009, 2010). The main objective of the present study is therefore to demonstrate limited filamentous bulking under low DO as a stimulation of SND to enhance biological nutrient removal and effluent quality. Experiments were carried out in a benchscale EBPR process, designed according to a BCFS process (Van Loosdrecht et al., 1998). The removal performance of nitrogen, phosphorus and chemical oxygen demand (COD) were investigated at different SVI values in aerobic compartment.

2.

Materials and methods

2.1.

Reactor configuration and experimental setup

Long-term experiments were performed in a bench-scale EBPR process, designed according to a BCFS process, as shown in Fig. 1. The reactor was made of plexiglas with a total working volume of 27 L, which was separated in four functional compartments by removable plastic sheets. The working volume of anaerobic, anaerobic selector, anoxic, anoxic/oxic and aerobic compartment is 5.4 L, 1.2 L, 5.4 L, 9 L and 6 L, respectively. A mechanical mixer was used in nonaerated zones to provide well mixed conditions. Aeration was supplied at the bottom of the aerobic compartments by an air compressor. A clarifier with a working volume of 9 L was used for solideliquid separation. The surface loading of the clarifier was 0.36 m3/m2 h. The flow rates of influent, return sludge, and the two internal recycle flows were controlled by four peristaltic pumps (Lange Z1515-100M, China).

2.2.

Wastewater composition

The bench-scale EBPR reactor was fed with raw domestic sewage. There were no extra chemicals added. The compositions are described as follows. CODcr: 183.5e367 mg/L; NHþ 4 eN:

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 7 7 e4 8 8 4

4879

Fig. 1 e Schematic diagram of bench-scale EBPR reactor. (1) Influent tank; (2) feed pump; (3) mechanical mixer; (4) check valve; (5) diffuser; (6) airflow meter; (7) air compressor; (8) return nitrified liquor pump; (9) return denitrified liquor pump; (10) secondary clarifier; (11) effluent; (12) waste sludge; and (13) return sludge pump.

46.3e62.5 mg/L; TN: 48.6e71.5 mg/L; TP 5.3e10.9 mg/L; SS: 151e200 mg/L; pH: 7.0e7.7. The standard deviations of CODcr, NHþ 4 eN, TN and TP of 90 samples are 18.7, 4.2, 4.5 and 0.8, respectively.

2.3.

Experimental operational conditions and procedures

The reactor was inoculated with activated sludge collected from Wenchang municipal wastewater treatment plant (A/O process) of Harbin, PR China, the initial concentration of sludge in the reactor was set at 4 g/L, and the reactor was fed with diluted raw domestic sewage for seven days to obtain constant colonization and accumulation of microorganism. Afterwards, the reactor was operated in a continuous plugflow mode fed with raw domestic sewage. The reactor gradually stabilized after the acclimatization of 35days at the room temperature 22  3  C, DO of 1.5 mg/L, hydraulic retention time (HRT) of anaerobic compartment (1.8 h), anaerobic selector (0.4 h), anoxic compartment (1.8 h), anoxic/oxic compartment (3 h), aerobic compartment (2 h) and solid retention time (SRT) of 15days. During the steady-state periods, the recycling rate of sludge, nitrified liquor and denitrified liquor was set at 1.0, 2.0 and 1.5 time of total influent flow rate, respectively. Then various investigations were conducted with different DO levels in the aerobic compartment.

2.4.

Analytical methods

 Ammonia nitrogen (NHþ 4 eN), nitrate nitrogen(NO3 eN), nitrite eN), total phosphorus (TP) chemical oxygen nitrogen (NO 2 demand (COD), mixed liquor suspended solids (MLSS), mixed liquor volatile suspended solids (MLVSS), alkalinity and sludge volume index (SVI) were measured according to the standard methods for the examination of water and wastewater (APHA, 1998). The total nitrogen (TN) concentration was determined with LiquiTOCII (Elementar, Germany). The DO was measured by SG6-ELK SevenGo Pro (Mettler Toledo, Switzerland). The ORP, pH and temperature were measured with HI-8424 pH meter (HANNA, Italy). Periodically microscopic observations of sludge samples from the aerobic compartment were performed with an Olympus IX51 inverted microscope (Tokyo, Japan) and the microscopic analysis was according to the reference manuals (Eikelboom, 2000; Jenkins et al., 2003).

3.

Results

3.1. Occurrence and control of limited filamentous bulking Extensive experiments have been performing for the optimization of process parameters such as volume ratios of interactive functional compartments, various recycling rate, SRT and DO in the past thirteen months. The occurrence of limited filamentous bulking was observed when the volume ratio of anaerobic, anoxic, anoxic/oxic and aerobic compartment was 1:1:1.7:1.1. Herein, DO concentration of the anoxic/oxic zone was constant in the range of 0.3e1.0 mg/L. Limited filamentous bulking is defined as sludge with an SVI of 140e250. The proliferation of filamentous micro-organisms was found by periodic microscopic observations. However, the removal performance of TN and TP adversely enhanced, and COD removal was stable during the period of limited filamentous bulking, in parallel with a case of good sludge settleability. Particularly the poor solideliquid separation and the loss of sludge were never observed. The phenomenon was consistent with the findings of a previous study (Guo et al., 2010). For the further research, limited filamentous bulking induced by DO level was studied. The relationship between settleability (SVI) and DO level under the prerequisite of suitable nutrients (N, P), food/micro-organisms ratio (F/M), wastewater characteristics, pH and temperature, is presented in Fig. 2. There is an expected relationship between DO and SVI. Experimental results showed that it is easy to control the limited filamentous bulking by adjusting aerobic DO levels in this bench-scale plug-flow EBPR reactor. Subsequently, the effect of limited filamentous bulking on the pollutants removal performance was focused at different aerobic DO levels (1.0e1.5, 1.5e2.0 and 2.0e3.0), which will be discussed further below.

3.2. Overall performance of nutrient removal under limited filamentous bulking 3.2.1.

COD removal

The organic loading rate of the reactor was in the range of 0.20e0.39 kgCOD kgMLSS1 d1, which in general is supposed not to lead to filamentous bulking (Chudoba et al., 1974; Wanner et al., 2010). Fig. 2(a) presented the COD removal efficiency under limited filamentous bulking in the bench-

4880

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 7 7 e4 8 8 4

D Fig. 2 e The removal efficiencies of COD, NHD 4 eN, TP at different SVI periods (a: COD removal; b: NH4 eN removal; c: TP removal). The nutrient removal efficiencies were compared under the condition of normal settleability and limited filamentous bulking.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 7 7 e4 8 8 4

scale EBPR reactor. The observed trends of COD removal efficiencies and effluent COD are very similar even though the SVI increased from 100 to 200, which means the COD removal performance is not affected by limited filamentous bulking. Previous researches have showed that COD removal efficiency was slightly interfered under filamentous bulking caused by overpopulation of Haliscomenobacter hydrossis (Kotay et al., 2010). The influent COD concentration had an average value of 260 mg/L ranging from 184 to 367 mg/L, whereas effluent concentrations had an average value of 30 mg/L fluctuating between 17 and 40 mg/L. Total COD removal efficiency with had an average value of 90% ranging from 82% to 95% during the steady-state period.

3.2.2.

Nitrogen removal

The NHþ 4 eN removal efficiency under limited filamentous bulking was evaluated, as illustrated in Fig. 2(b) and Fig. 3. For clear comparison, two vertical dotted lines were used to divide the figure into three sections based on the different SVI values. The influent NHþ 4 eN concentrations with an average value of 54 mg/L fluctuated between 46 and 63 mg/L, while effluent concentrations with an average value of 2.8, 3.7 and 4.7 at respective SVI values. From the Fig. 2(b), it could be readily observed that the trends of NHþ 4 eN removal efficiency slightly declined with the increasing SVI values, the average removal efficiency was 95%, 94% and 92% at respective SVI values. Denitrifying nitrogen removal efficiency remained stable under different DO levels, however, the TN removal efficiency enhanced since the ongoing occurrence of simultaneous nitrification and denitrification (SND) with SVI scale of 170e200 (associated with a DO of 1.0e1.5 mg/L) in aerobic compartment, which could obviously be observed by comparing the section of SND in Fig. 3. A similar result was reported by (Third et al., 2003) who also observed that SND increased during aerobic famine period in an SBR at low DO (<2 mg/L). From Fig. 3 it could be apparently found that SND never occurred at DO of 1.5e2.0 and 2.0e3.0 mg/L, this result is consistent with the record which reported that SND was not

Fig. 3 e Nitrogen mass balance at different SVI periods. The corresponding distribution of total nitrogen under the condition of normal settleability and limited filamentous bulking was displayed respectively.

4881

able to be achieved when DO was above 1.5 mg/L (Guo et al., 2010). Moreover, the higher concentration of nitrate in aerobic compartment at DO of 1.5e2.0 and 2.0e3.0 mg/L led to the upset of foregoing anoxic and anaerobic environment, and hence result in the lower phosphorus removal efficiency. Therefore, the preferable DO concentration of aerobic compartment for SND nitrogen removal was 1.0e1.5 mg/L in this bench-scale EBPR reactor.

3.2.3.

TP removal

The TP removal efficiency under limited filamentous bulking was investigated, as shown in Fig. 2(c), which was divided by vertical dotted lines into three sections according to the respective SVI periods. It could be distinctly observed that the TP removal efficiency steadily enhanced with increasing SVI, the lowest, average and highest removal efficiency was recorded as 74%, 90% and 97% with an SVI of 100e140,140e170 and 170e200, respectively. The influent TP had a mean value of 7.7 mg/L ranged between 5.5 and 11 mg/L, whereas effluent concentrations had a mean value of 2.23, 0.84 and 0.25 mg/L, correspondingly. Constant lower removal efficiency occurred at an SVI level of 100e140 (associated with a DO of 2e3 mg/L), which was mainly attributed to the chain of reactions in the reactor caused by DO level. Since the recycling of nitrified liquor and denitrified liquor were set for denitrification and denitrifying dephosphatation and for the reutilization of BOD, respectively. Therefore, strict anaerobic environment could be disturbed if the concentration of nitrate exceed 0.1 mg N/L in the anaerobic compartment (Van Loosdrecht et al., 1998), and the trespass of a higher DO in anoxic compartment would interfere the denitrifying dephosphatation. Reversely, the higher TP removal efficiency was obtained at higher SVI values of 140e170 and 170e200 (i.e. limited filamentous bulking). Moreover, low DO level can be favorable for maintaining denitrifying dephosphatation and indirectly upholding a comfortable anaerobic environment for efficient phosphorus release and associated substrate uptake. Notwithstanding, to guarantee the solideliquid separation in the secondary clarifier and a satisfactory effluent, DO level is not as lower as better and therefore an optimum settleability (SVI) should be dominated by moderate DO level based on the critical point(1.0e1.5 mg/L in this study). In particularly, phosphorus removal should be integrated with nitrification and denitrification, while DO level is a critical operating parameter of these procedures. Furthermore, the optimal SVI under limited filamentous bulking induced by DO level isn’t an absolute value, which still depends on the configuration of the process and other uncertain factors in practice (Eikelboom, 2000).

4.

Discussion

4.1.

Microscopic observation aspect

The filament index (FI) is a measure of the number of filamentous micro-organisms in activated sludge, which is established by comparing the microscopic image of the sludge with a series of reference photograph of the five FI classes at a low magnification. The predominance of filaments was mainly distinguished according to intrinsic morphological

4882

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 7 7 e4 8 8 4

characteristics of the filamentous micro-organisms viz. mobility, branching, filament shape, filament length, attached growth, septa or transverse walls, cell diameter and sheath etc. (Eikelboom, 2000). The observations in this study showed that FI was maintained between 1 and 2 under limited filamentous bulking which means the effect of the filaments on the settling velocity of the sludge is limited (Eikelboom, 2000). Sludge sample under limited filamentous bulking was staining with DAPI fluorochrome and then observed using the fluorescence microscope, the typical epifluorescence micrographs are showed in Fig. 4. The dominant filamentous bacteria was identified as H. hydrossis due to its needle-like appearance in a pin cushion with straight filaments protruding from the flocs, as presented in Fig. 4(a), and which was within the floc structure. In addition, minor S. natans characterized by

straight or smoothly curved filaments with no/tree-like false branching, round-ended and rod shaped cells and clearly visible cell septa with indentations was also recognized as the secondary filamentous bacteria, as shown in Fig. 4(b). S. natans can radiate outward from the floc surface into the bulk solution and results in a high SVI by inter-floc bridging. In this study, the presence of H. hydrossis was mainly caused by low DO, while S. natans was likely to be caused by low DO and long retention time of sludge in secondary clarifier. Nevertheless, H. hydrossis population is usually limited present in domestic treatment plants and it can develop en masse in industrial plants where many easily biodegradable compounds are present in the influent (Eikelboom, 2000). Indeed, low DO tends to cause filamentous bulking by S. natans, type 1701 and H. hydrossis (Jenkins et al., 2003). Moreover, a few amount of Eikelboom Type 0041 with much attached growth was also observed which was beneficial as the backbone structure for the flocs in Fig. 4(c). However, M. parvicella and Type 021N were apparently not observed. The results of this study are in line with the research of Gaval and Pernelle, 2003, in which the dominant filamentous bacteria identifying by morphological criteria and FISH were H. hydrossis and S. natans under respective oxygen deficiency condition, and small Type 021N was also observed. But Guo et al., 2010 demonstrated that Eikelboom Type 0041 was the dominant filamentous bacterium and few Type 021N and M. parvicella were also detected, however, H. hydrossis and S. natans were never observed.

4.2.

Fig. 4 e Epifluorescence micrographs of DAPI stained filament of (a) H. hydrossis, (b) S. natans and (c) Eikelboom Type 0041. The length of the bars corresponds to: a and b 10 mm; c 20 mm. The filamentous bacterium was observed under limited filamentous bulking.

Limited filamentous bulking

It has been hypothesized that filamentous micro-organisms serve as a backbone for the flocs to provide more binding sites for the attachment of free cells or smaller aggregates by extracellular polymeric substances (EPS) (Cenens et al., 2000; Liao et al., 2011). The principle of limited filamentous bulking is to keep a moderate imbalance between floc-forming and filamentous bacteria, which has a slight advantage for filamentous bacteria. This allows a better enmeshing of tiny particles or free flocs, and the larger floc diameter under limited filamentous bulking implies diffusion resistance of oxygen inside the flocs is larger which facilitate SND (Martins et al., 2004a), although filamentous bacteria extend from the flocs make flocs a bit incompact and porous, but denser. Moreover, filamentous micro-organisms have the competitive advantages to access organic substrate based on A/V hypothesis (Jenkins et al., 2003) and diffusion-based selection (Martins et al., 2004a, 2010; Lou and de los Reyes, 2008) since the intrinsic morphological property (preferential growth of one or two directions) facilitates a large contact area and an easy penetration rate. In addition, the kinetic selection hypothesis (Chudoba et al., 1974) can also well explain it because of the lower affinity constant of filamentous bacteria than floc-forming bacteria that low DO concentrations favor the growth of filamentous bacteria. If we can balance the advantages and disadvantages of filamentous bacteria under low DO condition, such as limited filamentous bulking for a good effluent, energy-saving undoubtedly been achieved. The experiments in this study are according to above theory. Although general believe is that the activated sludge with more filamentous bacteria is negative for wastewater

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 7 7 e4 8 8 4

operations (Jin et al., 2003), it was found that the solideliquid separation was practically not affected and no sludge loss was observed even with the highest SVI (200) condition under limited filamentous bulking induced by DO level. In the case of bio-P sludge, the density of the flocs might however compensate for the effect of the larger number of filaments (Eikelboom et al., 1998), and less production of sludge attributing to the characteristic of denitrifying dephosphatation in the EBPR reactor (Beun et al., 2000). In addition, a stable wellclarified effluent with marginal suspended solids (<8 mg/L) was obtained, which can be attributed to morphological characteristics of filamentous bacteria (i.e. enmeshment mechanism). The results of this study also show that limited filamentous bulking is repeatable and controllable. Therefore, the technique of limited filamentous bulking could be an alternative solution for enhancing nutrient removal and effluent quality though the control strategy and engineering capital in practice are still in question.

5.

Conclusions

This study investigated the integrated nitrogen, phosphorus and COD removal performance in a bench-scale plug-flow EBPR reactor under limited filamentous bulking. Experimental work lasted for about 200 days to study the removal performance of the EBPR process at different SVI periods. The results show that limited filamentous bulking induced by low DO level could enhance biological nutrient removal and achieve a well-clarified effluent. Moreover, low DO concentration associated to limited filamentous bulking is of importance for energy-saving. The optimum scenario for integrated nitrogen, phosphorus and COD removal was achieved under limited filamentous bulking at an SVI level of 170e200(associated with a DO of 1.0e1.5 mg/L), and the corresponding respective removal efficiencies of COD, TP, NHþ 4 eN were 90%, 97% and 92%. In addition, the solideliquid separation was practically not affected and no sludge loss was observed even at the highest SVI (200) condition, which was likely to attribute to the heavy flocs of bio-P sludge and less production of sludge of denitrifying dephosphatation.

Acknowledgments The authors gratefully acknowledge the financial support provided by National Water Pollution Control and Management Technology Major Projects of China (No. 2009ZX07317-008).

references

APHA, 1998. Standard Methods for Examination of Water and Wastewater, 20th ed. American Public Health Association, Washington, DC. Ajay, P., Jesse, Z., George, N., 2006. Simultaneous carbon, nitrogen and phosphorous removal from municipal wastewater in

4883

a circulating fluidized bed bioreactor. Chemosphere 65 (7), 1103e1112. Andreadakis, A.D., 1993. Physical and chemical properties lf activated sludge flocs. Water Research 27 (11), 1707e1714. Beun, J.J., Paletta, F., Van Loosdrecht, M.C.M., Heijnen, J.J., 2000. Stoichiometry and kinetics of poly B hydroxybutyrate metabolism under denitrifying conditions in activated sludge cultures. Biotechnology and Bioengineering 67, 379e389. Barat, R., van Loosdrecht, M.C.M., 2006. Potential phosphorus recovery in a WWTP with the BCFS process: interactions with the biological process. Water Research 40, 3507e3516. Chiu, Y.C., Lee, L.L., Chang, C.N., Chao, A.C., 2007. Control of carbon and ammonium ratio for simultaneous nitrification and denitrification in a sequencing batch bioreactor. International Biodeterioration and Biodegradation 59, 1e7. Chu, K.H., van Veldhuizen, H.M., van Loosdrecht, M.C.M., 2004. Respirometric measurement of kinetic parameters: effect of activated sludge floc size. Water Science and Technology 48 (8), 61e68. Chudoba, J., Blaha, J., Madera, V., 1974. Control of activated sludge filamentous bulking eIII. Effect of sludge loading. Water Research 8 (4), 231e237. Cenens, C., Smets, I., Van Impe, J., 2000. Modeling the competition between floc-forming and filamentous bacteria in activated sludge waste water treatment systems. Part II. A prototype mathematical model based on kinetic selection and filamentous backbone theory. Water Research 34, 2535e2541. Eikelboom, D.H., Andreadakis, A., Andreasen, K., 1998. Survey of the filamentous population in nutrient removal plants in four European countries. Water Science and Technology 37 (4/5), 281e290. Eikelboom, D.H., 2000. Process Control of Activated Sludge Plants by Microscopic Investigation. IWA Publishing, London, UK. Fu, Z., Yang, F., An, Y., Xue, Y., 2009. Simultaneous nitrification and denitrification coupled with phosphorus removal in an modified anoxic/oxic-membrane bioreactor (A/O-MBR). Biochemical Engineering Journal 43, 191e196. Gaval, G., Pernelle, J.J., 2003. Impact of the repetition of oxygen deficiencies on the filamentous bacteria proliferation in activated sludge. Water Research 37, 1991e2000. Gulez, G., de los Reyes, F.L., 2009. Multiple approaches to assess filamentous growth in activated sludge under different carbon source conditions. Journal of Applied Microbiology 106, 682e691. Guo, H.Y., Zhou, J.T., Su, J., et al., 2005. Integration of nitrification and denitrification in airlift bioreactor. Biochemical Engineering Journal 23, 57e62. Guo, J.H., Peng, Y.Z., Wang, S.Y., Zheng, Y.N., Huang, H.J., Wang, Z.W., 2009. Long term effect of dissolved oxygen on partial nitrification performance and microbial community structure. Bioresource Technology 100, 2796e2802. Guo, J.H., Peng, Y.Z., Peng, C.Y., Wang, S.Y., Chen, Y., Huang, H.J., Sun, Z.R., 2010. Energy saving achieved by limited filamentous bulking sludge under low dissolved oxygen. Bioresource Technology 101, 1120e1126. Jenkins, D., Richard, M.G., Daigger, G.T., 2003. Manual on the Cause and Control of Activated Sludge Bulking, Foaming, and Other Solids Separation Problems, third ed. Lewis Publishers, NY, USA. Jin, B., Wile´n, B.M., Lant, P., 2003. A comprehensive insight into floc characteristics and their impact on compressibility and settleability of activated sludge. Chemical Engineering Journal 95 (1e3), 221e234. Kotay, S.M., Datta, T., Choi, J., Goel, R., 2010. Biocontrol of biomass bulking caused by Haliscomenobacter hydrossis using a newly isolated lytic bacteriophage. Water Research. doi:10.1016/j. watres.2010.08.038.

4884

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 7 7 e4 8 8 4

Li, B., Bishop, L., 2004. Micro-profiles of activated sludge floc determined using microelectrodes. Water Research 38 (5), 1248e1258. Liao, J., Lou, I., delosReyes, F.L., 2004. Relationship of speciesspecific filament levels to filamentous bulking in activated sludge. Applied and Environmental Microbiology 70 (4), 2420e2428. Liao, B.Q., Lin, H.J., Langevin, S.P., Gao, W.J., Leppard, G.G., 2010. Effects of temperature and dissolved oxygen on sludge properties and their role in bioflocculation and settling. Water Research 45 (2), 509e520. Lou, I., de los Reyes, F.L., 2008. Clarifying the roles of kinetics and diffusion in activated sludge filamentous bulking. Biotechnology and Bioengineering 101, 327e336. Martins, A.M.P., Heijnen, J.J., van Loosdrecht, M.C.M., 2003a. Effect of feeding pattern and storage on the sludge settleability under aerobic conditions. Water Research 37 (11), 2555e2570. Martins, A.M.P., Heijnen, J.J., Van Loosdrecht, M.C.M., 2003b. Effect of dissolved oxygen concentration on sludge settleability. Applied Microbiology Biotechnology 62 (5e6), 586e593. Martins, A.M.P., Pagilla, K., Van Loosdrecht, M.C.M., 2004a. Filamentous bulking sludge e a critical review. Water Research 38 (4), 793e817. Martins, A.M.P., Picioreanu, C., Heijnen, J.J., van Loosdrecht, M.C. M., 2004b. Three-dimensional dual-morphotype species modeling of activated sludge flocs. Environmental Science and Technology 38 (21), 5632e5641. Martins, A.M.P., Karahan, O., van Loosdrecht, M.C.M., 2010. Effect of polymeric substrate on sludge settleability. Water Research. doi:10.1016/j.watres.2010.07.055. Martı´, N., Pastor, L., Bouzas, A., Ferrer, J., Seco, A., 2010. Phosphorus recovery by struvite crystallization in WWTPs: influence of the sludge treatment line operation. Water Research 44, 2371e2379.

Peng, Y.Z., Guo, J.H., Wang, S.Y., Chen, Y., 2008. Energy saving achieved by limited filamentous bulking under low dissolved oxygen: derivation, originality and theoretical basis. Journal of Environmental Sciences 29 (12), 3342e3347. Pe´rez, J., Picioreanu, C., van Loosdrecht, M.C.M., 2005. Modeling biofilm and floc diffusion processes based on analytical solution of reaction-diffusion equations. Water Research 39, 1311e1323. Pochana, K., Keller, J., 1999. Study of factors affecting simultaneous nitrification and denitrification (SND). Water Science and Technology 39 (6), 61e68. Third, K.A., Burnett, N., Cord-Ruwisch, R., 2003. Simultaneous nitrification and denitrification using stored substrate (PHB) as the electron donor in an SBR. Biotechnology and Bioengineering 83, 706e720. Van Loosdrecht, M.C.M., Brandse, F.A., de Vries, A.C., 1998. Upgrading of waste water treatment processes for integrated nutrient removal e the BCFS process. Water Science and Technology 37 (9), 209e217. Vaiopoulou, E., Melidis, P., Aivasidis, A., 2007. Growth of filamentous bacteria in an enhanced biological phosphorus removal system. Desalination 213, 288e296. Vaiopoulou, E., Aivasidis, A., 2008. A modified UCT method for biological nutrient removal: configuration and performance. Chemosphere 72, 1062e1068. Wanner, J., Kragelund, C., Nielsen, P., 2010. In: Seviour, R., Nielsen, P.H. (Eds.), Microbial Ecology of Activated Sludge. IWA Publishing, London. Wile´n, B.M., Balme´r, P., 1999. The effect of dissolved oxygen concentration on the structure, size and size distribution of activated sludge flocs. Water Research 33 (2), 391e400. Zhu, G.B., Peng, Y.Z., Wu, S.Y., Wang, S.Y., Xu, S.W., 2007. Simultaneous nitrification and denitrification in step feeding biological nitrogen removal process. Journal of Environmental Science 19, 1043e1048.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 8 5 e4 8 9 5

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Guided adaptive optimal decision making approach for uncertainty based watershed scale load reduction Yong Liu a, Rui Zou b,*, John Riverson b, Pingjian Yang a, Huaicheng Guo a a

College of Environmental Science and Engineering, Peking University, The Key Laboratory of Water and Sediment Sciences, Ministry of Education, Beijing 100871, China b Tetra Tech, Inc. 10306 Eaton Place, Ste. 340, Fairfax, VA 22030, USA

article info

abstract

Article history:

Previous optimization-based watershed decision making approaches suffer two major

Received 12 January 2011

limitations. First of all, these approaches generally do not provide a systematic way to

Received in revised form

prioritize the implementation schemes with consideration of uncertainties in the water-

29 April 2011

shed systems and the optimization models. Furthermore, with adaptive management, both

Accepted 26 June 2011

the decision environment and the uncertainty space evolve (1) during the implementation

Available online 3 July 2011

processes and (2) as new data become available. No efficient method exists to guide optimal adaptive decision making, particularly at a watershed scale. This paper presents

Keywords:

a guided adaptive optimal (GAO) decision making approach to overcome the limitations of

Adaptive management

the previous methods for more efficient and reliable decision making at the watershed

Guided adaptive optimal approach

scale. The GAO approach is built upon a modeling framework that explicitly addresses

Risk explicit interval

system optimality and uncertainty in a time variable manner, hence mimicking the real-

linear programming

world decision environment where information availability and uncertainty evolve with

Load reduction

time. The GAO approach consists of multiple components, including the risk explicit

Lake Qionghai Watershed

interval linear programming (REILP) modeling framework, the systematic method for prioritizing implementation schemes, and an iterative process for adapting the core optimization model for updated optimal solutions. The proposed approach was illustrated through a case study dealing with the uncertainty based optimal adaptive environmental management of the Lake Qionghai Watershed in China. The results demonstrated that the proposed GAO approach is able to (1) efficiently incorporate uncertainty into the formulation and solution of the optimization model, and (2) prioritize implementation schemes based on the risk and return tradeoff. As a result the GAO produces more reliable and efficient management outcomes than traditional non-adaptive optimization approaches. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

The need for load reduction to restore aquatic ecosystem health has been well recognized over the past decades (National Research Council, 2001; USEPA, 2001; Diaz and Rosenberg, 2008), and it has been considered necessary to

identify optimal nutrient management strategies both environmentally sound and economically effective (DePinto et al., 2004; Liu et al., 2008). Uncertainty due to data limitation and the stochastic nature of environmental systems presents a difficult challenge for developing and implementing optimal load

* Corresponding author. E-mail address: [email protected] (R. Zou). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.06.038

4886

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 8 5 e4 8 9 5

reduction management schemes (National Research Council, 2001). Various methods have been developed to account for the uncertainty in environmental optimization models (Chang et al., 1997; Freedman and Nemura, 2004; Fiedler et al., 2006; Bazargan, 2007). Although many of these methods appeared to be able to effectively handle uncertainties in the solution process, they all suffer the limitation that the model formulation and optimal solutions are assumed to be essentially static; however, implementation of management schemes is a dynamic process that may influence both the model formulation and the optimal state of a solution. Even though some optimization models such as multi-stage linear programming and dynamic programming appear to be dynamic in their formulation, they are essentially static in their predictive ability because the model formulations (and therefore, the resulting space of possible optimal solutions) remain the same with each successive solution iteration toward an optimal solution. The contradiction between the static nature of decision support models and the dynamic nature of the implementation processes can be illustrated using a hypothetic watershed management problem. When solving a comprehensive optimization model in a watershed that consists of 500 decision variables, the solution leads to optimal/near-optimal solutions for all the 500 decision variables. In other words, there are up to 500 management measures that need be implemented in order to achieve the environmental target. Implementation of these management measures, however, is a dynamic process where some of the management measures would be implemented before others. Furthermore, implementing each management measure would need a certain amount of time and financial resources, and many of these measures cannot be implemented simultaneously. In practice, it is highly possible that after implementing some of the management measures, new information about cost and environmental responses will become available, and might show that certain key coefficients in the original optimization model were inaccurate, making some of the original optimal solutions invalid. In such a case, if decision makers continue to implement management schemes based on the original optimal solutions and assumptions, the culminating management strategy would become ineffective and non-optimal. A preferred alternative to the traditional decision making approach is the adaptive management approach, which provides an effective way to cope with the uncertainties in environmental decision making (Walters, 2007; Harrison, 2007b). Adaptive management was proposed in 1970s as a “learning by doing” process to reduce uncertainty in decision making and correct decision error early in the process (Holling, 1978; Walters, 1986). Adaptive management is a continuous learning process in which decision-making is supported by models while management alternatives are modified when new information becomes available (Marttunen and Vehanen, 2004; Gregory et al., 2006b). Both the traditional environmental decision making and the adaptive management approaches recognize uncertainty. However, traditional methods try to use existing knowledge about the system uncertainty to formulate the best future course of action; while adaptive management explicitly evolves the decision making in the implementation processes by refining

management alternatives over time (McLain and Lee, 1996; Gregory et al., 2006a; Walters, 2007). More importantly, adaptive management is an incremental approach in which each proposed decision is viewed as an experiment in which the corresponding outcomes will be monitored, evaluated, responded and adapted to better reflect the studied systems (Harrison, 2007a,b; Pahl-Wostl, 2007). Adaptive management has been widely used in climate change studies (Brooks et al., 2005; Maciver and Wheaton, 2005; IPCC, 2007), fisheries management (Marttunen and Vehanen, 2004), water resource management (Pahl-Wostl, 2007), marine reserve evaluation (Grafton et al., 2005), and others. However, no study has reported explicit application of adaptive management in load reduction; USEPA (1991) proposed a ‘phased approach’ for Total Maximum Daily Load (TMDL), which is similar, but not completely the same, to the idea of adaptive management (Freedman and Nemura, 2004). Although there is wide acceptance of the idea of adaptive management in watershed pollution control and load reduction, there lacks a systematic approach for achieving optimal adaptive management in the decision making process. Most of the previous adaptive management research has focused on the “adaptive” process, but have overlooked the “optimization” perspective. Likewise, most of the “optimization” research has overlooked the “adaptive” process of problem formulation. A systematic approach for optimal adaptive management of watershed system should have three major traits. First, it should be able to explicitly handle uncertainty in the optimization model. Second, it should be able to handle relatively large scale problems that are expected for watershed scale analysis. Third, it should be able to provide guidance for prioritizing management measures to achieve reliable and efficient decisions. There is very little literature that addresses optimal adaptive management in environmental systems. For example, Harrison (2007a,b) proposed a two-stage adaptive approach with Bayesian Programming (BP) for optimal adaptive management of a river basin, in which new data were collected after implementation of the first-stage decision to update the priors for the second decision stage using Bayesian analysis. This approach was shown to have merits over earlier approaches in terms of computational efficiency and capability of handling relatively larger problems; however, it still can be computational prohibitive for watershed-scale problems because the BP is essentially a type of stochastic method where a large number of numerical realizations are needed to achieve model solutions. In addition, the BP and other previous researches provide no systematic way to prioritize the implementation schemes given uncertainties in the watershed systems and the optimization models. More importantly, the BP approach is not a full optimization-based adaptive management approach because it does not address cost effectiveness in an uncertainty-based decision making framework. The absence of an effective optimization approach and the lack of a systematic method to prioritize implementation reveal the need for a new methodology for uncertainty-based watershed-scale optimal adaptive management. This paper presents an uncertainty-based adaptive management framework that has been developed for optimal load reduction

4887

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 8 5 e4 8 9 5

decision making. A recently developed risk explicit interval linear programming (REILP) framework was used as the platform for formulating the uncertainty based optimal load reduction model (Zou et al., 2010; Liu et al., 2010). An adaptive management decision making approach was then proposed as a way to assist with prioritizing the implementation order of optimal solutions obtained at each stage of decision making, while the optimal solution at each stage is adaptively obtained through the process of implementation and data assimilation. This approach was applied to a real-world nutrient load reduction case; the system performance was compared and analyzed with and without applying the proposed approach.

2.

Materials and methodology

s:t: A
X  B


(2)

X0

(3)





where, f is objective function; C
¼ ½c
1 ; c2 ; .; ci ; .; cn  and g ði ¼ 1; 2; .; n; j ¼ 1; 2; .; mÞ are the interval coeffiA
¼ fa
ji cients defined as the lower bound and upper bound estimated from data; X is the unknown decision variables;

T represents the right-hand side B
¼ ½b
1 ; b2 ; .; bm  constraints. Eqs. (1)e(3) can be decomposed into two sub-models corresponding to the lower and upper bounds of the objective function and solved using standard LP algorithms (Tong, 1994). Upon completing the ILP solution for the lower and upper bound of the optimal objective function, the corresponding REILP model can be formulated as (Zou et al., 2010):

2

2.1. Uncertainty based optimization modeling framework

min xi ¼ 4i 4

3     þ  þ  5 lij aij  aij xj þ hi bi  bi

(4)

j¼1

Linear Programming (LP) has been one of the most widely applied mathematical programming techniques for assisting optimal environmental decision making during the past decades (Dantzig, 1955; Huang et al., 1992; Chang et al., 1997; Bazargan, 2007). When considering uncertainty, various types of LP approaches were developed to deal with uncertaintybased decision making problems within a LP frameworkestochastic linear programming (SLP), fuzzy linear programming (FLP) and interval linear programming (ILP) are the most widely researched and applied approaches. Among the three types, the ILP models have been identified by many previous studies, to be the most suitable for handling uncertainty in practical environmental systems analysis because it has the lowest data requirement that is compatible with the data availability in most real cases (Chang et al., 1997; Chinneck and Ramadan, 2000; Fiedler et al., 2006; Ozdemir and Saaty, 2006). Nevertheless, traditional ILP approaches that present optimal solutions in interval numbers have been found to be ineffective in supporting practical decision due to the problems of infeasibility, non-optimality, and the inability to relate decisions to risks (Zou et al., 2010; Liu et al., 2010). As a result, these approaches cannot provide a platform for systematically guiding an optimal adaptive management decision making process. To overcome the limitations of the traditional ILP approaches, Zou et al. (2010) developed a REILP approach to fully explore the uncertainty space defined by an ILP model for optimal solutions that explicitly relates system performance to decision risk. This study shows that the REILP would serve as an effective mathematical framework for optimal adaptive management decision making at a watershed scale. The following text details the formulation of the uncertainty based optimization modeling framework, the solution algorithm, and the methods for prioritizing implementation and guiding adaptive management decision making. A typical ILP model can be formulated as below (Tong, 1994):

Min f
¼ C
X

n X

(1)

s:t:

n X

  þ þ  cþ j xj  m  fopt  l0 fopt  fopt

(5)

j¼1

bþ i 

n X

aij xj  xi ;

ci

(6)

j¼1

l0 ¼ lpre

(7)

0  lij  1

(8)

xj  0;

cj

(9) l0 ðcþ j

c j Þxj ;

 fopt

þ fopt

where, m ¼  and are the lower and upper bound of the optimal solutions for the objective function of the original ILP model; l0 is the system aspiration level with a value between 0 and 1; lij and hi are real numbers between 0 and 1; xi is the risk function of constraint i, which is defined P þ   as xi ¼ nj¼1 lij ðaþ ij  aij Þxj þ hi ðbi  bi Þ; x is the risk function of the entire system; and lpre is the prescribed system aspiration level; 4 is a general arithmetic operator which can be a simple addition, a weighted addition, simple arithmetic mean, weighted arithmetic mean, or a max operator. The risk function forms a platform for risk and system return tradeoff such that when xi takes values greater than 0, some or all constraints would be relaxed for higher system return (Zou et al., 2010). The REILP model (4) to (9) can be solved using a nonlinear programming algorithm at each pre-specified aspiration level to form a decision front for risk-based decision making.

2.2. REILP-based implementation prioritization approach (IPA) The IPA plays a critical role in the optimal adaptive management decision making. As noted earlier, adaptive decision making is preferred over the traditional static approach because of uncertainties that exist in the decision making process. Adaptive management would allow system analysts and decision makers to update model results and

4888

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 8 5 e4 8 9 5

management schemes based on newly obtained data and refined characterization of uncertainty. Adaptive management is an experimental “learn by doing” approach to reduce uncertainty and improve the cost-effectiveness of management alternatives. In the context of uncertainty and experimental methods, decision makers need to make choices on what decision variables need to be implemented with higher priority either due to consideration of limited resources or interactions between different schemes (Gregory et al., 2006b). During the implementation process, those management schemes that are relatively insensitive to uncertainties should be implemented with higher priority than sensitive ones because they are less likely to be invalidated by newly obtained information. In other words, through implementing at an earlier stage those decision variables with lower risks, it is possible to acquire additional information to reduce the uncertainty and risk associated with those decision variables with higher risks, hence reduce the overall risks of the entire load reduction decision and implementation program. With this consideration, the IPA is formulated on the basis of a sensitivity analysis of the REILP solutions. In the REILP modeling framework, the impact of uncertainty on the optimal solutions is reflected in the aspiration levels (l0), which represents different levels of trade-offs between the cost of management schemes and the risk of potentially violating water quality standards due to uncertainties in the system. A higher aspiration level indicates that the decision makers’ would adopt management schemes that incur lower cost while accepting higher risk of non-compliance in water quality standards, and vice versa. At each aspiration level, a distinct set of optimal solutions can be obtained during decision making. Although it is possible to obtain optimal solutions for any aspiration levels between zero and one, an overly accurate stipulation of aspiration levels is not meaningful to decision makers who usually think in a quasi-qualitative way. In real-world decision processes, evaluations of risk or performance of a system are usually conducted on either a 10-level or 5-level scale to better emulate human brain function. In the present study, we propose to categorize the risk-cost tradeoff on a five-level scale, which are Level I: extremely conservative, Level II: moderately conservative, Level III: intermediate, Level IV: moderately aggressive and Level V: extremely aggressive. These five levels are respectively corresponding to the aspiration levels of [0.0, 0.2], [0.2, 0.4], [0.4, 0.6], [0.6, 0.8] and [0.8, 1.0]. An IPA is then devised based on the fivelevel scale. The proposed IPA consists of two steps, including an overall sensitivity analysis on system return with regard to system risk for achieving the desired tradeoff level (DTL), and an individual sensitivity analysis for each decision variable to decide the implementation priority under DTL.

2.2.1.

Step 1: DTL determination

The key information needed by decision makers to determine the DTL is the sensitivity of system return (the value of the objective function in the original ILP model) to the decision risk. For example, let’s assume that a REILP model was solved at the tradeoff level III and IV, and that the system return at

level IV is significantly better than that of level III, while the decision risk (in the form of the value of the risk function in the REILP model) is only slightly higher. In such a case, decision makers would likely choose level IV as the DTL for decision making. On the other hand, if the decision risk of level IV is significantly higher than that of level III, then decision makers might be more cautious and tend to choose level III as the DTL. Let r represents the decision risk level, which is equivalent to the Normalized Risk Level (NRL) in the REILP framework. The overall sensitivity of the system return ( f ) with regard to decision risk r can be expressed as:  df ¼ dr

d

Pn



i¼1 ci xi

dr

¼

n X dðci xi Þ i¼1

dr

(10)

Eq. (10) measures how system return interacts with the decision risk given the uncertainty present in the model formulation. A higher value of the sensitivity means that an increase in risk would be accompanied by large system gains, which reflects a situation where decision makers might consider taking the slightly increased risk for the chance of larger gains. On the other hand, a smaller value means that the same increase in decision risk would be accompanied by only small system gains, reflecting a situation where decision makers might be reluctant to take the risk for only minor gains. In practice, Eq. (10) can be evaluated across the neighboring tradeoff levels to obtain the distribution of system return-risk tradeoff efficiency. A cross-category analysis of the sensitivity can be produced to help decide the DTL reflecting the optimal tradeoff between risk and system performance. Let lt denotes an aspiration level corresponding to the tradeoff level t. The corresponding risk level is denoted as r ¼ rt, and the corresponding coefficient ci ¼ cti ; and the optimal solutions obtained by the REILP at lt are xi ¼ xti , where i ¼ 1, 2, ., n. Eq. (10) can thus be numerically evaluated at tradeoff level t using a finite difference approach. A backward finite difference expression of Eq. (10) is: df t Df t f t  f t1 ¼ ¼ drt Drt rt  rt1

(11)

where t ¼ 2, 3, ., 5 The sign of the value obtained from Eq. (11) depends on whether the original ILP model was minimization or maximization. To measure the absolute sensitivity of system return to risk level for the sake of simplicity, an absolute value of Eq. (11) is taken to define the Combined Sensitivity Coefficient at level t (CSCt) as:       df t  Df t  f t  f t1        CSCt ¼  t  ¼  t  ¼  t  dr  Dr   r  rt1 

(12)

Note that each tradeoff level represents a range of aspiration levels; therefore, the objective function value and risk level of each category t should be calculated before evaluating the value for the CSCt. Here we propose to use the average values for the objective function values and risk levels of each

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 8 5 e4 8 9 5

category to represent the corresponding system return and risk level values, leading to: Pm

j¼1 fj

ft ¼

m

¼

fLt þ fUt 2

(13)

ZUt

4889

each decision variable to the system return with regard to the decision risk level, i.e., the absolute individual sensitivity (AIS) can be expressed as:   dðci xi Þ  AISi ¼  dr 

(15)

rdl r ¼ t

Lt

Ut  L t

(14)

where, m is the total number of aspiration levels within tradeoff level t where REILP solutions are available; Lt and Ut represent the lower and upper bound aspiration levels for the tradeoff level; fLt and fUt are the lower and upper bounds of system return within tradeoff level t. Eq. (13) holds because the variation of system return values between aspiration levels follows a linear relationship by definition in REILP formulation, making it straightforward to evaluate the value of ft. However, because decision risk is a nonlinear function of aspiration level, an integration process is needed to evaluate the value for rt. In practice, Eq. (14) can be evaluated using a numerical integration method such as Simpson’s rule (Atkinson, 1989). With Eqs. (13) and (14), the CSC of each tradeoff level can be evaluated using Eq. (12), and the calculated CSC can be provided to decision makers to help decide a DTL upon which the optimal implementation plan will be based. Take a decision process starting from level I as example. Let’s assume that the CSC for level II is very large. It is anticipated that decision makers would likely step up from level I to II for significantly higher system return with relatively smaller risk. The decision makers can then evaluate the CSC for level III, which is assumed to also have a large value. Under such a circumstance, the decision makers might be encouraged to further step up to level III while checking the CSC value for level IV. Let’s further assume that the CSC value for level IV is a considerably smaller number, which means to further enhance system return would mean being subjected to significantly higher risk. Under these circumstances, decision makers would likely decide to stay on level III unless they have special reasons to step up to level IV. Since each tradeoff level is defined by a range of aspiration levels, the representative aspiration level for the DTL needs to be determined for the subsequent analysis. To avoid further complexity, we propose to use the mid-value of the range of aspiration levels for each tradeoff level as the representative aspiration level to reflect a preference for balanced decision within a specific risk tolerance category. As such, the aspiration level corresponding to the five tradeoff levels are 0.1, 0.3, 0.5, 0.7, and 0.9, respectively. In practice, system analysts and decision makers might choose to use different interval range for each category, hence different representative aspiration level for each category. In the latter text, the representative aspiration level corresponding to the DTL is referred to as the desired aspiration level (DAL).

2.2.2.

Step 2: implementation order determination

After determining the DAL, the corresponding optimal solutions can be used as the basis for formulating management schemes. At the DAL, the sensitivity of the contribution of

The value of AISi at a specific DAL can be evaluated within the corresponding DTL using a centered finite difference scheme as:   clþ1=2h xlþ1=2h  cl1=2h xl1=2h    i i i AISi ¼  i    rlþ1=2h  rl1=2h

(16)

where h is the aspiration level range of the DTL. A large AISi value indicates that the contribution of the corresponding decision variable to the total cost is sensitive to the uncertainty in the system representation. It is apparent that a robust implementation scheme should be associated with the decision variables that are less sensitive to the uncertainty in the model, i.e., those having small values of AIS. Therefore, the implementation process should follow the order that the more robust solutions are to be implemented prior to the less robust ones. To allow direct comparison of the relative sensitivity of each individual decision variable with all others, a relative AIS (RAIS) is defined as: AISi RAISi ¼ Pn i¼1 AISi

(17)

Ordering RAISi from the smallest to the largest, one can easily obtain decision variables which are least sensitive to the uncertainty, which would also tend to be more robust. These solutions should be implemented prior to others.

2.3.

Adaptive management scheme

The basic idea of adaptive management is to recognize the presence of uncertainty and adjust the modeling and practical implementation with time as new information becomes available. We propose a four-step approach for a risk-based optimal adaptive watershed management.

2.3.1.

Step 1: REILP formulation and DTL determination

This step involves formulating and solving the ILP and REILP models for a specific watershed-scale load reduction problem, and then conducting the analysis shown in Eqs. (10)e(14) to determine the DTL.

2.3.2.

Step 2: implementation order determination

This step involves calculating the AISi and RAISi for each decision variable using Eqs (16) and (17) to determine the implementation order. The decision variables with low RAISi values are considered robust and will be placed a higher priority in implementation. At this stage, it is possible that decision makers would incorporate additional stipulations regarding budget and implementation periods for the prioritized decision variables. In practice, multiple implementation

4890

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 8 5 e4 8 9 5

periods might be needed, considering the availability of financial resources and construction efforts.

2.3.3. Step 3: staged implementation and optimization analysis Throughout each implementation period, new datasets, monitoring data, and other information may become available to better quantify the coefficients in the ILP and REILP models. As modeling-based decisions are implemented, newly collected and monitored data would be available to help increase the accuracy of system representation and reduce uncertainty in the model formulation. Therefore, the optimization model should be updated with the new information to obtain enhanced decision support capability. In this step, the ILP and REILP models developed in Step 1 and used in Step 2 are updated and solved to produce new optimal solutions, after which the analyses in Step 2 are to be repeated to obtain the DTL, implementation order, and new management schemes for the subsequent period.

2.3.4. Step 4: repeating until accomplishment of load reduction goal This step involves repeating Steps 1 to 3 whenever new data become available in the implementation process until the completion of the anticipated management program.

3.

Case study

3.1.

Study area: Lake Qionghai Watershed

Lake Qionghai, the second largest freshwater lake in Sichuan Province, China, is on top of the agenda for water quality protection in the local area (Liu et al., 2008). The watershed is divided into 20 sub-watersheds (Supporting Materials). Lake Qionghai is experiencing rapid eutrophication caused by excessive watershed nutrient loading. A long-term watershed nutrient loading reduction program was initiated to restore the water quality and to protect the aquatic ecosystem. Previously, integrated watershed optimization analysis was conducted in a static decision manner, which is considered insufficient to support the long-term management goal (Liu et al., 2008, 2010). In this study, the proposed optimal adaptive management approach is applied to this system to illustrate the procedure and advantage.

3.2.

3.3.

Adaptive management: initial stage

The ILP and REILP model equations represent the uncertainty based optimization model built upon the currently available information (Supporting Materials). The model can be solved to provide solutions guiding the decision making for the initial stage of the adaptive management. Note that in traditional optimal decision making framework, the solutions obtained in this stage were previously used to formulate long-term implementation plan, which will later be shown to be ineffective and unreliable. Fig. 1 presents the optimal solutions showing the tradeoff between system performance (total cost, f ) and the risk levels (Liu et al., 2010). As shown, for lpre ¼ 0, the risk level is 0 by definition, which corresponds to the upper bound solutions of the ILP model with a total cost of f ¼ $1.546 billion. The upper bound solution reflects the situation that when the decision makers are willing to spend $1.546 billion to treat the pollution sources in the watershed, the risk of violating the water quality target in the lake would be at the minimum. On the other hand, the risk level for lpre ¼ 1.0 reaches the maximum value, suggesting that a most aggressive decision corresponds to the lower-bound solutions of the ILP model with a total cost of f ¼ $0.773 billion. This extreme solution means that if the decision makers are willing to spend the lower amount of $0.773 billion for the load reduction based on an extremely optimistic interpretation of uncertainty in the optimization model, there would be a pretty good chance that the water quality target would still be violated. In Fig. 1, the tradeoff curve shows a sharp decrease of total cost ( f ) when the risk level starts to increase from zero. This decent of cost, however, gradually levels off until reaching the point after which any further increase in risk only result in relatively small reductions in total cost. The optimal adaptive decision making starts with determining the order of implementation for each decision variable. For this purpose, Eqs. (12)e(14) were used to calculate CSCt. The calculated CSCt value for level I to II is 2199.5, for II to III 2022.9, for III to IV 1384.8, and for IV to V 605.0. Since the CSCt values for level I to II and II to III are similar to each other,

ILP and REILP model formulation

The goal of optimal nutrient load reduction analysis is to minimize the implementation cost while meeting the water quality targets. Since significant uncertainty exists in the system characterization and no data were available to specify the probabilistic or possibilistic distribution of model coefficients, interval numbers were used to represent the uncertainty in a LP framework, resulting in an ILP model for the watershed. In our previous studies, a multi-stage ILP-based and REILP-based optimization model were developed to deal with the nutrient load reduction issues in Lake Qionghai Watershed in three 5-year periods (Liu et al., 2008, 2010; Supporting Materials).

Fig. 1 e Trade-off curve of total cost versus normalized risk level (NRL) of decision.

4891

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 8 5 e4 8 9 5

either of them can be chosen as the DTL in practice depending on the decision makers’ preference. In this study, the highest CSCt value of 2199.5 was chosen to determine the DTL for illustrative purposes to suggest that it is most effective to formulate a management scheme based on conservative interpretation (tradeoff level II) of the uncertainty in the ILP model (Supporting Materials). After determining the DTL, the AIS and RAIS were calculated and presented in Table 1. As shown, solutions for all the decision variables except those corresponding to j ¼ 6, 8 and 10 have very small RAIS. Therefore, all decision variables with small RAIS are considered robust with regard to uncertainty in the ILP model. Based on this analysis, the solutions for these robust decision variables would be set at a higher priority for implementation than the solutions for the decision variables corresponding to j ¼ 6, 8, and 10. It is noted that to fully implement all these robust solutions would require approximately $1.3 billion, which is far beyond the funds available at the current stage. In contrast, the total funds approved for the load reduction projects in the first three years are approximately $200 million, which also cannot meet the investment requirements for implementing all of the robust solutions; therefore, decision makers need to further prioritize the implementation order based on the available budget, RAIS of each decision variable, and the associated construction times and maintenance costs. This analysis led to the selection of the four decision variables corresponding to j ¼ 1, 2, 4, and 7 to be implemented in the first three years.

3.4.

Adaptive decision making

In real-world practice, decision making is always a dynamic process as the system representation continually changes and new information becomes available. In some cases the new data and information might provide justification for the validity of the parameter values used in the initial optimization model, indicating that the original optimal solutions can still be valid for the next stage of implementation. However, in cases where the new data and information indicates that the parameter values in the original optimization model are inaccurate, the original optimal solutions for decision variables not yet being implemented might become invalid or suboptimal for guiding the next stage of implementation. In such cases, it is desirable to obtain updated optimal solutions that better characterize the changing system. For watershed management, examples of new data and information include: (1) water quality monitoring data characterizing the watershed and lake system, (2) an updated water quality model linking watershed loadings to in-lake water quality response, or (3) a more accurate cost function

acquired through the construction and maintenance of the treatment facility built during the initial stage of implementation. In the case of Lake Qionghai, the original optimization model was configured using a water quality response matrix derived from a simple box model of the lake. If a sophisticated eutrophication model is developed after the initial stage, which provides a more accurate response matrix, the optimization model would need to be updated using the newly available information. For illustrative purposes, this paper presents a hypothetical condition where new information becomes available that convinced the decision makers to find updated optimal decisionsdthe new information will help to reduce the previous uncertainties and therefore shrink the parameter intervals. The hypothetical decision conditions are as follows: (1) after the first implementation period, more information becomes available to update the cost functions

ðIIC
j and ASCj Þ for decision variables; (2) TEC was refined using new water quality monitoring and load data; (3) after evaluating the performances of the implemented load reduction schemes in the initial stage, the decision makers decide to refine their anticipated goals, leading to the updates

of APR
j , RFLi and RLR , etc. (please refer to the Supporting Materials). To support the second stage of adaptive decision making, the original ILP/REILP system was updated using the aforementioned new information, and the updated model was then solved for new solutions (Supporting Materials). Note that for those decision variables already implemented during the first stage, no updated solution were generated because they are now part of the physical reality in the system. Therefore, they are labeled as “NaN”. The corresponding AIS, RAIS and cost for the second stage REILP was calculated and shown in Table 2. The results revealed that j ¼ 3, 9 and 8 should be on top of the list for being implemented in the following years since they have relatively low RAIS values. The decision making will follow a similar adaptive process as in the first implementation period, which we will not further elaborate here. In subsequent implementation periods, the ILP/REILP model will be updated whenever new data and information become available to guide the optimal decision making process over time.

4.

Results and discussions

Fig. 2 compares the total cost and load reduction that resulted from the original non-adaptive REILP optimization model (NAO) and the guided adaptive REILP optimization model. It suggests that by applying the guided adaptive management approach, the total cost for load reduction can be reduced

Table 1 e AIS, RAIS and cost for each decision variable.

6

AIS (10 ) RAIS Cost ($ million)

j¼1

j¼2

j¼3

j¼4

j¼5

j¼6

j¼7

j¼8

j¼9

j ¼ 10

97.64 0.05 132.02

40.03 0.02 55.38

24.26 0.01 37.79

0.31 0 0.79

164.58 0.08 79.89

1015.83 0.48 321.05

0 0 0.02

315.09 0.15 422.66

17.39 0.01 34.6

435.17 0.21 230.09

4892

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 8 5 e4 8 9 5

Table 2 e AIS, RAIS and cost for each decision variable based on updated REILP model.

6

AIS (10 ) RAIS Cost ($ million)

j¼3

j¼5

j¼6

j¼8

j¼9

j ¼ 10

21.67 0.01 37.79

365.81 0.14 79.89

1526.10 0.57 321.05

268.49 0.10 422.66

17.19 0.01 34.60

479.93 0.17 230.09

from $1.31 billion to $1.18 billion (a savings of about 10%) due to the incorporation of new information throughout the implementation period. Correspondingly, the total load reduction amount between the two scenarios changes from the 222.4 tonnes/year to 221.12 tonnes/year. Compared with the original NAO model results, the costs and load reduction for j ¼ 5, 6, 8 and 10 and i ¼ 3, 9, 10, 11 and 12 varied noticeably, suggesting that the optimal solutions of the NAO model that had a higher range of uncertainty, were no longer valid given the updated, more accurate characterization of the system. The reason why the guided adaptive optimal approach would produce a lower long-term cost than the static approach is that with the new information incorporated, the original optimal solutions might become obsolete and no longer optimal in the updated model framework. To further demonstrate the advantage of the adaptive approach, the original REILP solutions at various aspiration levels were entered into the updated REILP framework to determine the maximum achievable aspiration levels (MAAL). The MAAL is defined as the maximum aspiration level, which when exceeded, would cause the original REILP solutions to become infeasible in the updated model. The introduction of the MAAL into the analytic framework provides a consistent way of evaluating the performance of different REILP solutions (Liu et al., 2010). After determining the MAAL, the corresponding risk and system performance ( f ) at lpre ¼ MAAL was then

calculated by incorporating the original REILP solutions into the modified REILP modeling framework, allowing for direct comparison of the solutions of the adaptive REILP model at various risk levels (Fig. 3). As shown, at each specific decision risk (any given point along the X-axis), the decision based on the original static REILP model solutions incurs a higher total cost than the corresponding one based on the adaptive approach. Similarly, at each specific total cost (any given point along the Y-axis), the decision based on the original REILP suffers a higher risk of violating water quality standards than the one based on the adaptive approach. The results of this analysis demonstrate one of the main points of this paper: the traditional optimization-based watershed management approaches which use static optimal solutions to guide longterm management plan can be ineffective at achieving cost reduction and risk control. In contrast, the proposed adaptive optimal approach was shown to be able to provide improved decision support for cost effectiveness and risk control simultaneously. Another major intention of this paper is to fill in the knowledge gap of traditional adaptive management by developing a systematical way to conduct adaptive management that is optimal with regard to evolving uncertainties in the system characterization. To distinguish our approach to the conventional adaptive management concept, our approach is hence referred to as the “guided adaptive optimal” (GAO) approach, which features the capability of conducting guided prioritization of implementation order. Here we will illustrate the advantage of the GAO approach by contrasting it against the unguided adaptive optimization approach (UGAO) that is adaptive, but without guidance for prioritizing the implementation order at each adaptive stage. While the GAO was able to perform an informed decision by selecting j ¼ 1, 2, 4, and 7 to be implemented in the first three years based on the RAIS calculation, the UGAO could only proceed no better than randomly picking values, for example j ¼ 2, 3, 4 and 5, to be

Fig. 2 e Total cost and load reduction summarized by strategy and subwatershed for the non-adaptive REILP versus guided adaptive optimal approach.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 8 5 e4 8 9 5

Fig. 3 e The total cost versus normalized risk level of decision for non-adaptive optimal REILP versus the guided adaptive optimal solutions.

implemented in the first three years due to lack of systematic guidance. The resulting solutions for the GAO and UGAO comparison are shown in Fig. 4. With the systematic guidance for adaptive management, the GAO results in optimal solutions for the subsequent implementation stage that is different from those of UGAO. With the GAO approach, the total cost for load reduction would be $1.18 billion, which is lower than the $1.22 billion of the UGAO approach (a savings of only about 3%). However, the risk level of the decisions obtained by the GAO approach is only 0.099 (versus 0.106 for the UGAO approach), which represents lower risk level. By using a systematic process to prioritize the implementation order, the GAO approach produced a more cost-effective and safer (more reliable) management decision than the more

4893

traditional unguided adaptive optimal management approach. Finally, when evaluating solutions associated with the NAO, UGAO and GAO, it becomes evident how the selected strategies for different sub-watersheds determine the predicted costs and load reductions. Fig. 5 shows that although the average relative distribution over the study area for selected strategies is similar, there are some big differences between selected solutions among some of the sub-watersheds. Subwatersheds i ¼ 1, 3, 5, 8, 12, and 16 had some of the most notable differences between scenarios. For example, the GAO solution completely forwent lake riparian vegetation covers ( j ¼ 10) in subwatershed 12 in favor of increased extreme erosion restoration ( j ¼ 3) elsewhere in subwatershed 3 and wastewater treatment plant enhancements ( j ¼ 5) in subwatershed 8. Within subwatershed 8, wastewater treatment plant enhancement was also prioritized over NPS controlling measures in the GAO solution. On the other hand, the UGAO solution invested significantly more money on NPS controlling measures than wastewater treatment in subwatershed 8. As part of a wider planning effort for the entire study area, the GAO-selected solutions resulted in a net cost savings for comparable load reduction at a lower risk level. By minimizing the risk of selecting ineffective solutions, the proposed guided adaptive optimization approach has a greater potential to deliver optimal solutions in the long run than both non-adaptive and unguided optimization approaches. For larger systems with more management options or even subsequent implementation rounds of the same system, the advantages of the guided approach would likely become even more pronounced in terms of cost savings, load reduction, and risk management. The major limitation of the presented approach is that it is not based on a full simulation-optimization framework, therefore, it is best fit for systems with insufficient data

Fig. 4 e Total cost and load reduction summarized by strategy and subwatershed for unguided adaptive versus guided adaptive optimal approaches.

4894

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 8 5 e4 8 9 5

Fig. 5 e Cost distribution by strategy (J) for non-adaptive, unguided, and guided adaptive optimal approaches for entire study area and selected sub-watersheds (I).

available to accurately quantity the nonlinear cause-andeffect responses between management measures and water quality. In cases where data are sufficient to quantify the nonlinear cause-and-effect responses, and where the nonlinearity is strong and non-negligible, it would be more desired to apply a direct simulation-optimization framework to support the adaptive management decision-making process. Such methodological framework will be the focus of follow-up researches.

5.

Conclusions

This study developed a GAO approach for risk-based decision making for nutrient reduction at the watershed scale. The methodological procedure of the GAO approach was demonstrated through a case study application of the Lake Qionghai Watershed in China. The advantages of the GAO approach were further demonstrated by comparing the resulting solutions using: (1) the non-adaptive model versus the guided adaptive model, and (2) the unguided adaptive model versus the guided adaptive model. The conclusions obtained in this study are as follows: 1) Traditional non-adaptive optimization-based decision support approaches are not effective in a dynamic management environment; therefore, an adaptive optimal approach offers a better way to dynamically take into account new information for updating the system characterization and quantifying uncertainty. In an adaptive management process, the order of implementation can have a significant impact on the system performance. Traditional unguided adaptive management approaches

lack a systematic way that simultaneously addresses optimization while prioritizing the implementation order. The proposed GAO approach provides a solid analytical framework and a systematic way to guide optimal decision making in a changing decision environment by addressing the evolution of uncertainty with newly available information during the implementation process, and providing timely guidance to subsequent implementation phases for a more reliable outcome. 2) The case study shows that with the GAO approach, the resulted decision making is (a) more cost-effective and (b) has lower risk liability than the case when a static optimization-based decision making was used. The case study also demonstrates that the adaptive process alone is not guaranteed to produce optimal management results; in contrast, the systematic prioritization component of the GAO provides a sound basis for informed decision making with regard to the implementation order, resulting in higher cost-effectiveness and lower risk-liability than a more conventional, unguided adaptive management approach. The model results shows that the total GAO cost for load reduction would be $1.18 billion, which is lower than the $1.22 billion of the UGAO approach. 3) While the case study example demonstrated an improving cost-effectiveness and risk reduction trend between the non-adaptive, unguided, and guided adaptive optimization approaches in the Lake Qionghai Watershed, only two rounds of adaptive management were performed. It is expected that the differences between these approaches would become even more pronounced (1) after subsequent management rounds, or (2) in a larger watershed with more management strategy options or more compliance constraints.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 8 5 e4 8 9 5

4) The ILP and REILP mathematical framework appears to be an effective analytic basis for handling the adaptive decision support problem given their direct consideration of uncertainty in both the formulation and the solution, coupled with the capability of generating a risk-based decision front upon which to conduct in-depth tradeoff analysis for guiding the adaptive management process.

Acknowledgments This paper was supported by the “China National Water Pollution Control Program” (2008ZX07102-001), Los Angeles County Department of Public Works, and Research Fund for the Doctoral Program of Higher Education of China (20100001120020).

Appendix. Supplementary material The supplementary data associated with this article can be found in the on-line version at doi:10.1016/j.watres.2011.06.038.

references

Atkinson, Kendall A, 1989. An Introduction to Numerical Analysis, second ed. John Wiley & Sons. Bazargan, M., 2007. A linear programming approach for aircraft boarding strategy. European Journal of Operational Research 183 (1), 394e411. Brooks, N., Adger, W.N., Kelly, P.M., 2005. The determinants of vulnerability and adaptive capacity at the national level and the implications for adaptation. Global Environmental Change 15, 151e163. Chang, N.B., Chen, H.W., Shaw, D.G., Yang, C.H., 1997. Water pollution control in a river basin by interactive fuzzy interval multi-objective programming. Journal of Environmental Engineering ASCE 12 (123), 1208e1216. Chinneck, J.W., Ramadan, K., 2000. Linear programming with interval coefficients. Journal of the Operational Research Society 51, 209e220. Dantzig, G.B., 1955. Linear programming under uncertainty. Management Science 1, 197e206. DePinto, J.V., Kaur, J., Larson, W.M., Atkinson,J.F., 2004. LOTOX2 Model Documentation e In Support of Development of Load Reduction Strategies and a TMDL for PCBs in Lake Ontario, Submitted to USEPA e Region 2 by Limno-Tech, Inc., Ann Arbor, MI. 122 pp. Diaz, R.J., Rosenberg, R., 2008. Spreading dead zones and consequences for marine ecosystems. Science 321, 926e929. Fiedler, M., Nedoma, J., Ramik, J., Rohn, J., Zimmermann, K., 2006. Linear Optimization Problems with Inexact Data. Springer, New York. Freedman, P.L., Nemura, A.D., 2004. Viewing total maximum daily loads as a process, not a singular value: adaptive watershed management. Journal of Environmental Engineering-ASCE 130 (6), 695e702. Grafton, R., Kompas, T., Schneider, V., 2005. The bioeconomics of marine reserves: a selected review with policy implications. Journal of Bioeconomics, 161e178.

4895

Gregory, R., Failing, L., Higgins, P., 2006a. Adaptive management and environmental decision making: a case study application to water use planning. Ecological Economics 58 (2), 434e447. Gregory, G., Ohlson, D., Arvai, J., 2006b. Deconstructing adaptive management: criteria for applications to environmental management. Ecological Applications 16, 2411e2425. Harrison, K.W., 2007a. Test application of Bayesian programming: adaptive water quality management under uncertainty. Advances in Water Resources 30 (3), 606e622. Harrison, K.W., 2007b. Two-stage decision-making under uncertainty and stochasticity: Bayesian Programming. Advances in Water Resources 30 (3), 641e664. Holling, C.S. (Ed.), 1978. Adaptive Environmental Assessment and Management. John Wiley and Sons, New York. Huang, G.H., Baetz, B.W., Patry, G.G., 1992. A grey linear programming approach for municipal solid waste management planning under uncertainty. Civil Engineering Systems 9, 319e335. IPCC, 2007. Climate change 2007: impacts, adaptation and vulnerability. Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. In: Parry, M.L., Canziani, O.F., Palutikof, J.P., Linden, P. J.v.d., Hanson, C.E. (Eds.). Cambridge University Press, Cambridge, UK, p. 976. Liu, Y., Guo, H.C., Zhou, F., Qin, X.S., Huang, K., Yu, Y.J., 2008. Inexact chance-constrained linear programming model for optimal water pollution management at the watershed scale. Journal of Water Resources Planning and Management-ASCE 134 (4), 347e356. Liu, Y., Zou, R., Guo, H.C., 2010. A risk explicit interval linear programming model for uncertainty-based nutrient-reduction optimization for the Lake Qionghai watershed. Journal of Water Resources Planning and Management-ASCE 137 (1), 83e91. Maciver, D.C., Wheaton, E., 2005. Tomorrow’s forests: adapting to a changing climate. Climatic Change 70, 273e282. Marttunen, M., Vehanen, T., 2004. Toward adaptive management: the impacts of different management strategies on fish stocks and fisheries in a large regulated lake. Environmental Management 33 (6), 840e854. McLain, R.J., Lee, R.G., 1996. Adaptive management: promises and pitfalls. Environmental Management 20, 437e448. National Research Council, 2001. Assessing the TMDL Approach to Water Quality Management. National Academy Press, Washington, D.C. Ozdemir, M.S., Saaty, T.L., 2006. The unknown in decision making what to do about it. European Journal of Operational Research 174 (1), 349e359. Pahl-Wostl, C., 2007. The implications of complexity for integrated resources management. Environmental Modeling and Software 22, 561e569. Tong, S.C., 1994. Interval number, fuzzy number linear programming. Fuzzy Sets Systems 66, 301e306. USEPA, 1991. Guidance for Water Quality-Based Decisions: The TMDL Process. Development and Implementation of the TMDL, EPA Office of Water, Washington, D.C. http://water.epa. gov/lawsregs/lawsguidance/cwa/tmdl/dec3.cfm (accessed 20. 08.10). USEPA, 2001. Protocol for developing pathogen TMDLs. EPA 841-R00e002, first ed. EPA Office of Water, Washington, D.C. Walters, C., 1986. Adaptive Management of Renewable Resources. McMillan, New York. Walters, C.J., 2007. Is adaptive management helping to solve fisheries problems? AMBIO 36 (4), 304e307. Zou, R., Liu, Y., Liu, L., Guo, H.C., 2010. REILP approach for uncertainty based decision making in civil engineering. Journal of Computing in Civil EngineeringeASCE 24 (4), 357e364.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 9 6 e4 9 0 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Evaluation of suitable chlorine bulk-decay models for water distribution systems Ian Fisher a,*, George Kastl b, Arumugam Sathasivan c a

Watervale Systems Pty Ltd, PO Box 318, Potts Point, NSW 1335, Australia MWH Australia, Level 3, Boundary Street, South Brisbane, QLD 4101, Australia c Department of Civil Engineering and Construction, Curtin University of Technology, GPO Box U1987, Perth, WA 6845, Australia b

article info

abstract

Article history:

Maintaining the chlorine residual is a major disinfection goal for many water distribution

Received 17 December 2010

systems. A suitable general chlorine bulk-decay model is required for simulation of chlo-

Received in revised form

rine profiles in networks to assist disinfection planning/management efficiently. The first-

21 March 2011

order model is unsuitable due to inaccuracy and inability to represent rechlorination.

Accepted 26 June 2011

Three potentially suitable, simple, reactant models were compared. The single-reactant

Available online 2 July 2011

model was found to be unsuitable, as it was inaccurate when restricted to using a single set of invariant parameters. The two-reactant model was more suitable than the variable-

Keywords:

rate-coefficient model, although both models were accurate under the same restriction.

Chlorine decay

The two-reactant model was then calibrated against datasets consisting of multiple decay

Reactant model

tests for five distinctly different waters. It accurately predicted data reserved for validation

Bulk water

over the chlorine concentration range of 0e6 mg/L, using a single set of invariant param-

Distribution system

eters, and is therefore the simplest, generally suitable model for simulating chlorine

Network

profiles in distribution system networks. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Drinking water is generally disinfected to protect public health. The goal of secondary disinfection is to maintain a disinfectant residual throughout the distribution system, so that a nominated residual is achieved even at the system extremities (e.g., WHO, 2004; USEPA, 2002). This provides some degree of protection against contamination, as well as limiting bacterial regrowth (LeChevallier, 1999). As chlorine is the most widely used secondary disinfectant, its efficient use is of considerable importance to water utilities worldwide. Most of the chlorine dosed is consumed in reactions with other substances remaining in the water after treatment, particularly dissolved organic matter (DOM). In some waters, there is also substantial chlorine consumption by inorganic

substances, such as ammonia, iron and manganese. Only free chlorine is considered in this paper, as that form of chlorine is a substantially stronger oxidant than any other form likely to be present. Such losses of chlorine, also known as chlorine “decay”, have two undesirable consequences. Firstly, a substantially greater concentration of chlorine must be established at the entry to the distribution system, to achieve the management goal of a nominated residual at system extremities. However, too high an initial dose will generate taste/odour complaints from upstream consumers, requiring booster doses (rechlorination) at intermediate locations to enable the initial dose to be lowered. Secondly, reactions with some of the components of DOM create by-products that are known to be harmful to human health. Consequently, regulation of by-products, in the USA and internationally (USEPA,

* Corresponding author. Tel.: þ61293572310. E-mail address: [email protected] (I. Fisher). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.06.032

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 9 6 e4 9 0 8

2006; WHO, 2004), further constrains planners and managers to minimize the amount of chlorine dosed, while still achieving microbial control. Software for simulating flows in complex pipe/tank networks has been in common use to plan and manage distribution systems for more than 15 years (e.g., Rossman, 1994). More recently, these packages have included models of chlorine decay to plan and manage the balancing act required to achieve secondary disinfection goals. A conceptual partitioning of chlorine decay is commonly made and embodied in network modelling packages (e.g., Powell et al., 2000; Rossman, 1994). The “bulk reaction” is the decay of chlorine due to its reaction with substances including DOM, which remain in the bulk water after treatment. The decay due to reaction of chlorine with the pipe wall, the biofilm that grows on it and corrosion products and other particles that adhere to it, is termed the “wall reaction”. It is seldom sufficiently emphasized that bulk-decay needs to be accurately modelled first, as it is a characteristic only of the transported water, independent of the system. In contrast, the wall decay (chlorine loss) can only be quantified as differences between in-system measurements of chlorine and values calculated from a bulk-decay (only) model at corresponding points and times in the network model. If there were no error in the in-system measurements, any error in bulkdecay prediction becomes an equal error of opposite sign in the wall decay estimate (Fisher et al., 2011). Consequently, the wall decay should only be evaluated after a sufficiently accurate model of bulk decay is derived. Then a model of the instantaneous wall decay rate can be devised to fit the wall decay (loss) estimates and both models of instantaneous decay rate can be integrated simultaneously to yield predictions of the chlorine concentration over time in the system. An additional advantage of this separate estimation of bulk and wall decay for planning/management is the separation of water character from system character, so that changes to water routes, pipe networks and reservoirs do not require changes to the bulk reaction model parameters in different system scenarios. Similarly, changes to source water characteristics do not necessarily require recalibration of the entire system model, although it is likely that the wall decay will eventually change in response. An accurate bulk-decay model is needed to detect such a response and to quantify it. Even if the wall decay does change, separation of water character from system character still minimizes calibration/scenario simulation effort. The model most widely used in network software is still the first-order (FO) decay of Equation (1) (Johnson, 1978) in which bulk-decay rate is proportional to chlorine concentration remaining. dcCl =dt ¼ KcCl

(1)

where cCl is chlorine concentration [mg Cl/L] in a water sample t [h] after initial dosing and K is the FO reaction rate coefficient or “chlorine decay” coefficient [h1]. This model has long been recognized (Boccelli et al., 2003; Kastl et al., 1999; Clark, 1998; Haas and Karra, 1984) as a poor approximation to real decay behaviour (the shape of the decay curve) over the time water typically spends traversing a distribution system, because decay is really a complex set of

4897

reactions between chlorine and numerous organic (and inorganic) compounds in the bulk water. Before more suitable bulk-decay models for planning/management of disinfection in distribution networks can be identified, the issues to be addressed by system modelling need to be defined, so that the features required in a bulk-decay model can be explicitly specified. Disinfection planning/management issues form a multilevel hierarchy, in the sense defined by Haimes et al. (1975). A major feature of such hierarchies is that a problem at one level can only be satisfactorily solved if a solution procedure has already been established for the problem at the next lower level. Determination of the initial chlorine concentration (ICC) required at system entry to achieve a target concentration at system extremities can be regarded as the basic (level 0) issue in the hierarchy of disinfection issues. This “basic ICC” is determined from system simulations that generate residual concentrations at the extremities, for trial values of ICC selected from the normal operational range (say 0e5 mg/L). An accurate bulk-decay model is an essential element in such simulations. The trial ICC value that results in the target residual being slightly exceeded at all system extremities is then identified as the “basic ICC”. Other major issues (levels) in the hierarchy are: 1. basic ICCs for different temperatures (seasons), which requires a model that includes the effect of temperature; 2. optimal locations and doses for rechlorination; 3. optimal chlorination regime for multiple sources (blending), which requires a model in which decay rates in each source are specified; 4. direct representation of disinfection by-product (DBP) formation, which requires an augmented decay model that generates DBPs, and 5. tradeoffs between additional reactant removal (treatment) and chlorine doses downstream, for which a linked model of the effect of removal is a prerequisite. For maximum efficiency of system modelling, the same decay model “structure” would be used to address issues at all levels of the hierarchy. This generality would minimize the amount of software development and ownership required of utilities and would enable practitioners to develop a high level of expertise in model calibration and scenario simulation. On the other hand, a general model must represent accurately the decay behaviour exhibited by a specific water, through selection of the most appropriate set of values of the coefficients (parameters) that are built into the model structure. This selection process should focus exclusively on minimizing the mismatch between model predictions and data from simple decay tests on the relevant water because there are no sufficiently strong, consistent relationships between model parameters and other water quality variables such as DOM, even for a single water (e.g., Powell et al., 2000). Nor should such relationships be expected, due to the wide range of reacting substances that remain in different waters, even after treatment. The suitability criteria relevant to finding basic ICCs (briefly summarised in Table 1) are first justified in terms of planning/ management issues. Then, the simplest models that are

4898

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 9 6 e4 9 0 8

Table 1 e Basic criteriaa for suitability of bulk chlorine decay models to assist disinfection planning and management in a water distribution network. Criterion 1 2 3

4

5

Description Accurate prediction of chlorine residual over typical maximum system travel times Simplicity (least number of state variables and parameters), after meeting Criterion 1 Actual initial chlorine concentration (ICC) is settable as ICC in model; i.e., the ICC equivalent to the chlorine dose is applied at t ¼ 0 in the model, rather than some lower concentration being applied after an initial loss period to improve the fit to data Parameter values invariant over the maximum time taken for water to travel through the distribution system Parameter values invariant for any ICC in the operational range

a See Section 1.1 for detailed definition and justification of these criteria.

potentially more suitable than the FO model are identified from the literature for further evaluation.

1.1. Criteria for suitability of bulk chlorine decay models for planning/management The first requirement of a decay model is that it accurately predicts the chlorine concentration as it decreases towards the chosen (low) target level at system extremities. Criterion 1 is that the decay model must be sufficiently accurate over a time scale of the maximum travel time, which may be several hundred hours in some systems. Secondly, on grounds of scientific principle and efficiency of modelling itself, the simplest model with sufficient accuracy should be identified (Criterion 2) and used in practice. The number of state variables and parameters included in a dynamic model are taken to be reasonable measures of complexity. Most longer-term models (e.g., FO) commence after some initial period of loss that is too fast to be described by the model (e.g., Boccelli et al., 2003). The model ICC (the concentration at the end of the initial loss period), and the time at which it occurs after dosing, can only be estimated arbitrarily. For system modelling, this implies repeated re-estimation of the length of the initial loss period, the initial loss rate and the associated reduction in model ICC, to appropriately compensate for inaccurate representation of chlorine decay and the impact of time-varying system demand. This is highly inefficient when a large number of scenarios are to be examined, as in planning studies. If, instead, the initial loss is ignored (i.e., the longer-term model parameters are applied from time zero), then large errors are generated during the initial loss period, which can be propagated downstream, especially if there are reservoirs or pipe junctions near the initial dosing point. Consequently, for efficient model calibration/scenario runs and implementation of results operationally, Criterion 3 is that the actual concentration resulting from the initial chlorine dose, at system entry, can be set as the ICC in a network simulation.

Parameter invariance over time (Criterion 4) is essential in system models because the travel time to a particular point in a network varies with demand. If a parameter took on a different value at different times after dosing, then a check would need to be made on each parcel of water at each timestep to determine whether a new value should be set. This facility is not generally available in system modelling software. Furthermore, at any junction having more than one inflow, water parcels from each inflow are mixed. If each inflow could have a different value of the parameter at a given time, then the parameter value for the mixed outflow(s) would be indeterminate. Whether the volume-weighted average, for example, could be used would need to be determined for each water, over a range of different volume ratios. This is most unlikely to be adopted generally, so that parameter invariance over time is required instead. It should be noted that the use of a constant decay rate for some initial loss period before a longer-term model (e.g., FO) commences, effectively constitutes a multi-stage model, which does not meet Criterion 4. Decay models embedded in network simulation software also need to use an invariant set of parameter values to characterize bulk decay over the operating range of ICCs (Criterion 5), presuming that water character immediately prior to dosing is not changing with time. This invariance also establishes confidence in the model robustness. It is also needed for efficient assessment of major changes to system configuration, when bulk parameters would otherwise need to be re-estimated for each alternative configuration being considered. Several studies (e.g., Powell et al., 2000; Hua et al., 1999; Kie´ne´ et al., 1998) have shown that the FO decay coefficient is not invariant, even over the range of ICCs encountered operationally.

1.2. Review of bulk chlorine decay models against criteria The FO model is a member of a group of models which define the decay rate as a function of chlorine only. These models are generally unable to meet one or more of the suitability criteria in Table 1 (Fisher et al., 2011). Such models are also unsuitable because they are inherently incapable of representing the slower decay (at a given residual concentration) that occurs after rechlorination, unless they use a value for the decay coefficient that is smaller than the original coefficient value. This would be highly inefficient in modelling practice, because the post-rechlorination decay coefficient would need to be reestimated for every different booster location, booster dose and system flow rate that is to be considered in a scenario simulation. As rechlorination is an issue higher in the planning/ management hierarchy, only those models capable of adequately representing it are considered further in this paper. To enable efficient modelling of rechlorination issues, a model needs to include explicitly at least one substance that reacts with chlorine (hereafter termed a “reactant” or reacting “species”). Jadas-He´cart et al. (1992) showed that the rate at which a single notional constituent reacted with chlorine in a Parisian treated water could be represented over 150 h by a second-order model that is bilinear in each constituent

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 9 6 e4 9 0 8

(Equation (2)), but only after an initial loss period of 4 h. This single-reactant (SR) model is identical to the classical kinetic model of chemical reaction between two pure substances, according to the “law” of mass action: dcCl =dt ¼ k  cCl  R

(2)

where R is the concentration of the substance reacting with chlorine [mg R/L] and k is the second-order reaction rate coefficient [L/mg R/h]. Clark (1998) commenced with the same rate equation and Rossman (2000) provided a means to implement this decay model from time zero within the EPANET software. The model was fitted to 11 single decay tests on five different waters (of which all but one were treated). In all cases, a good fit was achieved. However, in seven cases, the tests were conducted for only 50 h or less, which is far less than the maximum travel time in many distribution systems. In several cases, decay was measured over a limited concentration range, rather than until concentration was nearly zero. The time at which the first chlorine measurement was taken was not precisely related to the time of dosing. Each of these limitations may greatly reduce the overall curvature, which is the feature of a full chlorine decay curve that is difficult to represent accurately. Boccelli et al. (2003) showed that the Clark (1998) model was unable to represent the full decay curve adequately over 125 h in four different waters, even for ICCs as low as 1.2 mg/L. They then used the idea of an “initial loss” devised by Jadas-He´cart et al. (1992), but found that the period before the second-order model could be applied was different in different waters. That is, the SR model could not meet Criterion 2 (setting the model ICC to the actual ICC) while meeting the requirement for longterm accuracy (Criterion 1). For one water tested by Boccelli et al. (2003), parameters were derived from decay tests with three different ICCs, but the model was fitted to each test independently. As even a single decay test could not be adequately represented over the full period of the test, the SR model is even less likely to represent all three decay curves adequately with a single set of parameter values. For a different water, Kastl et al. (1999) showed that when the SR model was fitted to the low-ICC test, the same amount of reactant was consumed too early in the high-ICC test. That is, the SR model did not meet Criterion 5 (parameter invariance over the ICC range). Multiple-reactant (or species) models contain rate equations that describe the decrease in the concentration of each reactant over time, usually as a function of the concentration of one or more of the reactants. These models generally assume that the law of mass action applies to all reactions, in its simplest form (i.e., the bilinear form in Equation (2)). Although each reaction may have the same form as the Clark (1998) model, the inclusion of more than one reaction can account for a much wider spectrum of behaviour, while maintaining invariance of parameters. The general second-order (bilinear) model for m reactants is then X ki  cCl  Ri (3) dcCl =dt ¼  where Ri is the concentration of the ith reacting substance, i ¼ 1,2 . m [mg Ri/L] and ki is the ith reaction rate coefficient [L/mg Ri/h].

4899

Jonkergouw et al. (2009) proposed a variable rate coefficient (VRC) model, in which chlorine decay rate was proportional to both the chlorine concentration (cCl) and the sum of the concentration of all reacting substances (SRi in Equation (3)). It therefore appears to be a second-order SR model. However, the rate of change of the time-varying decay coefficient (kt) was approximated by an empirical function with parameters kmin and a. For each of six waters, the model was fitted to either two or three decay tests concurrently. Each test had a different ICC. The decay tests mostly extended until the chlorine residual was nearly exhausted or the decay rate was almost zero. However, substantially lower ICCs than those measured were used to obtain the best fit (accuracy) across the full period of decay, often by about 0.5 mg/L and up to 1.5 mg/ L. As presented, the VRC model did not meet Criterion 3 and some of the parameters are not readily interpretable. Kastl et al. (1999) simplified the general second-order model (Equation (3)) by using only the first two terms; i.e., by including notional fast- and slow-reacting reducing agents involved in second-order reactions with chlorine over periods typical of travel times in long distribution systems (Equations (4)e(6)). Hereafter, this is termed the two-reactant (2R) model. The second-order reaction rates and resulting chlorine decay rate are given by: dcF ¼ kF  cCl  cF dt

(4)

dcS ¼ kS  cCl  cS dt

(5)

dcCl dcF dcS ¼ þ dt dt dt

(6)

where cF and cS are the concentrations of fast and slow reducing agents respectively, and kF and kS are the decay rate coefficients for the fast and slow reactions. In contrast to the coefficients of the empirical function of Jonkergouw et al. (2009), the parameters of the 2R model are readily interpretable. These are the initial concentrations of the two notional reactants (c0F and c0S) and their respective decay rate coefficients (kF and kS), which can be estimated straightforwardly using software that is built specifically for such a purpose (e.g., Reichert, 1998, 1994), provided the appropriate decay test data have already been generated. The reactants are represented only in terms of the amount of chlorine they react with (i.e., as mg Cl-equivalent/L). This allows the simplest stoichiometry to be assumed (one molecule of chlorine reacts with one molecule of “chlorine-equivalent” reactant). As each notional reactant represents a complex mixture of (unknown) reducing agents, the alternative of estimating some average stoichiometry for each reactant is not feasible. Any substantial deviation in the actual stoichiometry of some of the compounds involved will be absorbed into the model parameter values during their estimation. In addition to meeting the accuracy criterion better than four other models of comparable or lower complexity, Kastl et al. (1999) showed that the estimated parameter values were invariant over time and with respect to ICC over the range of 0e4 mg/L. The 2R model therefore met all the criteria for suitability specified in Table 1 for the single water modelled.

4900

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 9 6 e4 9 0 8

The three reviewed models (SR, VRC and 2R) are the simplest models capable of addressing the rechlorination (level 3) issue. Given the uncertainties from different studies regarding their capability to find basic ICCs (level 0 issue), their performance was first compared, using the same dataset for calibration of each model. Then a detailed validation of the most suitable model was carried out for a wider range of waters. The degree to which this model meets the suitability criteria was also evaluated from the validation results.

2.

Methodology

2.1.

Chlorine decay test data

The model comparison used data from decay tests conducted at the Harbin Institute of Technology on water from Harbin, China. The experimental method was detailed in Jonkergouw et al. (2009) and is similar to that described below for Greenvale water. Four other waters, from Australia and the USA, were also used for model validation. These ranged from raw water from large reservoirs and an artesian aquifer to surface waters treated with conventional and advanced processes. It should be noted that chlorinated raw water is still a water source for many distribution systems e both Greenvale and Warragamba were examples at the time their decay tests were carried out. At least three decay tests were conducted on each water at a constant temperature. The basic features of the waters, the experimental data and their sources are summarised in Table 2. The raw water sample from Greenvale Reservoir (Melbourne, Australia) was transported on ice by air freight to a Sydney laboratory, where five sub-samples in new 1.25 L PET bottles were placed in a water bath maintained at 20  C (0.5  C) for the duration of the decay tests. A chlorine stock solution was prepared from MilliQ pure water and 10% sodium hypochlorite. The volume required to establish the appropriate ICC was added to each sub-sample and a MilliQ blank for each ICC. Immediate measurement of chlorine in blank samples provides an accurate estimate of the ICCs, as very little chlorine decay occurs in them. Free chlorine measurements were made on all sub-samples with Hach chlorine

pocket colorimeters, which use the DPD method, at times between which free chlorine concentrations dropped mostly by less than 0.5 mg/L. Typical precision (95% confidence interval) of low-range free chlorine measurement is 0.05 mg/ L (Hach Company, 2006). The sample from Warragamba Reservoir (Sydney, Australia) was tested earlier at the same laboratory, so the procedure was somewhat different. In particular, free chlorine measurement used the amperometric method, which has precision similar to that of DPD colorimeters (Harp, 2002). A full description of the method is given in Kastl et al. (1999). Water from Rocky Creek Reservoir (New South Wales, Australia) undergoes ultra-filtration prior to treatment with ozone and BAC followed by buffering to pH 8.2 with lime/CO2. The sample was taken prior to final disinfection and tested by the local laboratory according to the method used for Greenvale water. The water from the Wanneroo artesian aquifer (Perth, Western Australia) was untreated and tested using the DPD colorimetric method with an experimental error of 1e2% RSD (Warton et al., 2006). The test data for Wanneroo and Harbin water were extracted from the figures in the papers cited, using digitizing software.

2.2.

Model parameter estimation

The user of a model needs some convenient way to estimate the parameter values that allow the model to best represent chlorine decay in the water of interest. These “optimal” values are derived by minimizing the sum of squared differences between chlorine concentrations measured during decay tests and the model estimates at corresponding times during simulation of these tests. Fisher and Kastl (1996) and all subsequent applications of their multiple-reactant models have used the AQUASIM software package (Reichert, 1998, 1994) to derive optimal parameter values. It provides a user-friendly, modular system for dynamic simulation of complex multiple reactions in a set of reactors, which may or may not be linked by hydraulic flows. It also provides special facilities for deriving the set of parameter values that result in the best fit of model predictions to a given dataset. In particular, it enables the

Table 2 e Summary of decay test data used for calibration of models. Water

Treatment ICC range No. tests Temp. ( C) (mg/L)

Greenvale Reservoir (GV) None Warragamba Reservoir (WG) None Harbin (HB) Not given

1e4 1e4 1e3

5 3 3

Rocky Creek (RC) Wanneroo artesian aquifer (WR)

1e5 4e10

4 4

Advanced None

a Used only in model comparison (Table 3b). b Range around time of sampling. c DOC.

20 25 3.5a 15.5 28a 20 Not given

TOC (mg/L)

pH

Alkalinity (mg/L)

Test procedure

2.6 6.5e7 3.7 6.9 Not given Not given

Not given 32 Not given

See Method section Kastl et al. (1999) Jonkergouw et al. (2009)

2e3b 1.8c

35e45b 95

See Method section Warton et al. (2006)

7.5e8b 8.0

4901

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 9 6 e4 9 0 8

simultaneous optimization of a single set of model parameter values from a set of batch reactors, in which the same set of reactions is occurring, but under different conditions in each. This is exactly the requirement for fitting a multiple-reactant model to decay tests conducted at a constant temperature and commencing from different ICCs. The AQUASIM package was used to derive optimal parameter values for the 2R models fitted to the six datasets in this paper. The AQUASIM software calculates the sum of squared differences between each experimental data point and the corresponding model prediction, assuming an initial set of parameter values. Using the simplex technique (Nelder and Mead, 1965) or the secant algorithm (Ralston and Jennrich, 1978) combined with the active set technique (Gill et al., 1981), it then systematically varies the parameter values to search for the set that produces the best fit (i.e., minimizes the sum of squared differences). There were no constraints placed on the parameter values, except for non-negativity. As one of the suitability criteria was the ability to set the actual ICC used in the decay tests, these values (listed in Table 3) were set during the parameter optimization procedure.

2.3.

Model comparison and validation

The comparison between the SR, VRC and 2R models used the dataset for water from Harbin, China, which comprised decay tests with three different ICCs at each of three different temperatures. The three models were initially calibrated against data from a single temperature (15.5  C), to provide the most straightforward comparison of their suitability against the criteria of Table 1. The models were then recalibrated against the full dataset, while constraining the initial concentration of reactant(s) to be identical in all decay tests because a sub-sample of the same water was used in each test. The subsequent validation used datasets derived from the five waters described above. Each set comprised decay tests for at least three different ICCs at a single temperature.

Initially, the model was calibrated using the whole of a dataset, to show whether an invariant set of parameter values provided adequate estimation of all decay test data. Then the model was validated by recalibrating to only the decay tests having the highest and lowest ICC in that dataset. Predictions of the decay test data not used in the calibration were then compared with the corresponding measured data. These predictions were also compared with the corresponding estimates from the models fitted to the whole dataset.

3.

Results and discussion

3.1.

Model comparison

Optimal values for the parameters in each of the SR, VRC and 2R models were estimated for Harbin water by minimizing c2, the sum (over all relevant decay tests) of the squared differences between measured and modelled chlorine concentrations at each measurement time, divided by the variance(s) of measurements. (Variance of all chlorine measurements was set to unity, so that it had no influence on the minimization.) This minimization was readily achieved using the two optimization techniques available within the AQUASIM parameter estimation procedure. The simplex technique appeared to find an initial solution more readily than the secant method, but the latter seemed to find a more optimal solution than the former; i.e., the secant method seemed more capable of moving past local optima. Each minimization was restarted each time an optimum was found, as recommended by Reichert (1998), until no further improvement in c2 was obtained. For about a tenth of these minimizations, at least one other starting point was used; i.e., the procedure was repeated from a markedly different point in parameter space. Occasionally a third starting point was used. In all cases, the same optimal parameter set was obtained, provided the secant method was used as the final optimization method.

Table 3 e Optimal parametera values and associated c2 for SR, VRC and 2R models calibrated using Harbin data. Calibration data (a) 15.5  C only SR VRC 2R

ICCs (mg/L)

R0/R0/c0S (mg/L)b

k/kmin/kS (L/mg Ri/h)

e/a/c0F (mg/L)

e/k0/kF (L/mg Ri/h)

c2

2.25 7.83 2.22

0.345 0.00617 0.0363

e 23.8 1.09

e 0.907 1.91

2.47 0.175 0.209

e e e

e e e

6.26

2.20 1.84 12.7

0.565

2.53 2.65 6.34

0.844

1.3, 2.4, 2.8

(b) Same initial reactant concentrations at all temperatures 1.1, 1.3, 2.5 2.13 SR 3.5  C 1.3, 2.4, 2.8 SR 15.5  C 1.8, 2.4, 2.9 SR 28  C

0.306 0.421 0.638

VRC 3.5  C VRC 15.5  C VRC 28  C

1.1, 1.3, 2.5 1.3, 2.4, 2.8 1.8, 2.4, 2.9

3.73

0.0107 0.0239 0.0861

2R 3.5  C 2R 15.5  C 2R 28  C

1.1, 1.3, 2.5 1.3, 2.4, 2.8 1.8, 2.4, 2.9

2.07

0.0242 0.0506 0.126

12.5 10.1 16.6 0.953

a Parameters defined in Equations (4)e(6). b Parameter names are listed in same order in the heading as they are in the models; i.e., SR/VRC/2R.

4902

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 9 6 e4 9 0 8

The SR model has two state variables (cCl and R) and two parameters (R0 and k). Boccelli et al. (2003) added a third parameter during model fitting, by also optimising the initial concentration of chlorine, c0Cl. The VRC model has three state variables (cCl, R, and k) and four parameters (R0, k0, kmin and a). Jonkergouw et al. (2009) also optimised c0Cl, effectively adding a fifth parameter. Allowing c0Cl to vary during calibration required the arbitrary selection of either R0 or k0, which Jonkergouw et al. (2009) resolved by an ill-defined relationship between R0 and DOM. The 2R model also has three state variables (cCl, cF and cS) and four parameters (c0F, c0S, kF and kS). Kastl et al. (1999) set c0Cl to measured values in their model comparison of chlorine decay in a single water. For the closest possible comparison between models, the SR and VRC models were refitted to the 15.5  C Harbin data with c0Cl set to measured values. R0 could then be simultaneously optimised in the VRC model, rather than arbitrarily relating it to DOM. Optimised parameter and c2 values for all calibrated models are shown in Table 3a and model estimates of the data are shown in Fig. 1. The c2 value obtained from the VRC model was 0.18, compared with 0.21 for the 2R model; i.e., both models estimate the data very well. Using accuracy as a criterion, there is little to choose between them. In contrast, the c2 value obtained for the SR model (2.47) was an order of magnitude higher, indicating a relatively poor fit, as confirmed by Fig. 1A. The incomplete reaction in the high-ICC test due to exhaustion of the reactant, also noticeable in the results of Kastl et al. (1999), is even more obvious here.

The Harbin dataset was selected because three comparisons could be done on the same water (at different temperatures). The initial amount of any reactant must then be identical in all cases. The SR model has one initial reactant concentration (R0) and only one other parameter (k) to be optimised at each temperature (i.e., a total of four parameters). The optimised parameter and c2 values are given in Table 3b. The poor fit indicated by the high c2 value (6.26) is confirmed in Fig. 2A. Exhaustion of the reactant occurs at all temperatures. Constraining R0 to a realistic value e greater than the maximum amount of chlorine consumed e increased c2 to 9.17, indicating an even poorer fit to the data. The VRC model also has one initial reactant concentration (R0) and three other parameters, so that a total of ten parameters required optimization over all temperatures. The 2R model has two initial reactant concentrations (c0F and c0S) and two other parameters; hence a total of eight parameter values to be optimised. Consequently, the VRC model should estimate the data substantially better than the 2R model does. When the parameters were optimised (Table 3b), c2 for the VRC model was 0.57 compared with 0.84 for the 2R model, which confirmed slightly better estimation by the VRC. Although both models overestimated data from two of the decay tests at 28  C (Fig. 2F and I), estimates from either model are acceptable. In contrast, the SR model again does not provide estimates of adequate accuracy. The suitability criteria in Table 1 were specified for only the basic issue of the disinfection issues hierarchy. A model

Fig. 1 e Decay tests and model simulations for Harbin water at 15.5  C: (A) SR model; (B) VCR model, and (C) 2R model, where points are measured data, curves are results from calibrated models (in Table 3a). (D) 2R model calibrated to minimal dataset (see Fig. 3 for legend).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 9 6 e4 9 0 8

4903

Fig. 2 e Decay tests and model simulations for Harbin water at 3.5  C, 15.5  C and 28  C: (A)e(C) SR model; (D)e(F) VRC model, and (G)e(I) 2R model. Points are measured data, curves are results from calibrated models (in Table 3b).

suitable for addressing other issues will also need parameter values that are invariant with respect to temperature. The 2R model can be immediately extended to meet this criterion by incorporating the classical Arrhenius variation of chemical reaction rate with temperature. This implies a monotonic increase in reaction rate with temperature, which the optimised parameter values (kF and kc) of the 2R Harbin model exhibit (see Table 3b). Jonkergouw et al. (2009) noted that the VRC parameters (except R0) are temperature dependent, but there is no consistent relationship with temperature evident in the VRC case, except for kmin. Consequently, the 2R model is suitable to address issues higher in the hierarchy, whereas the temperature dependency of the parameters in the VRC model remains to be established. Using the 2R model, there is also an obvious way to represent the mixture of two (or more) source waters in the same network. Each source is represented by a separate pair of equations of the same form as Equations (4) and (5). At any mixing point, the flow-weighted average of the concentrations of each reactant provides a proper estimate of the mixed concentrations and there is no adjustment needed to the invariant rate parameters estimated from decay tests on the individual sources. Jonkergouw et al. (2009) claimed that the same could be achieved with the VRC model, but the

assumption that a flow-weighted average of each of their parameters at any mixing point is a proper estimate of the corresponding parameters in the mixture is questionable, and would need to be demonstrated. For both these reasons, the 2R model is the more suitable model to address disinfection planning/management issues and is therefore the only one considered here for validation in a wider range of waters.

3.2.

Initial calibration of 2R model for other waters

Optimal values for the four parameters in the 2R model (Equations (4)e(6)) were estimated for each of the other four waters, using the same procedure as the model comparison. The optimal parameter and c2 values derived are given in Table 4a and the decay curves obtained from simulations using these values are shown in relation to the data in Figs. 3Ae6A. For waters tested with ICCs up to 4 mg/L (Harbin, Greenvale and Warragamba), excellent estimations of all decay curves were obtained from the model (Figs. 1C, 3A and 4A respectively, and low corresponding c2 values in Table 4a). For Rocky Creek water (maximum ICC 5 mg/L), very good estimates were

4904

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 9 6 e4 9 0 8

Table 4 e Optimal parametera values, and associated c2, RMSEb and R2c values for calibrated two-reactant models. Water

ICCs mg/L

(a) Calibration to entire datasets 1.3, 2.4, 2.8 Harbin 15.5  C Greenvale 1.0,1.5, 2, 3, 4 Warragamba 1.0, 2.1, 3.7 Rocky Creek 1.2, 2.5, 3.8, 4.8 Wanneroo 4.0, 6.0, 8.0, 10.0 Wanneroo restricted 4.0, 6.0, 8.0

c0S mg/L kS L/mg Ri/h c0F mg/L kF L/mg Ri/h c2 calib. No. data points RMSE all data 2.22 2.66 1.79 1.63 2.47 1.84

0.0363 0.00623 0.0286 0.000164 0.00363 0.00821

1.09 0.837 0.606 0.678 2.34 2.24

(b) Calibration only using decay tests with minimum and maximum ICCs 1.3, 2.8 1.96 0.0458 1.13 Harbin 15.5  C Greenvale 1.0, 4.0 2.87 0.00366 1.13 Warragamba 1.0, 3.7 1.96 0.0130 0.819 Rocky Creek 1.2, 4.8 1.79 0.000173 0.612 Wanneroo 4.0, 10.0 2.57 0.00531 2.31 Wanneroo restricted 4.0, 8.0 1.80 0.0105 2.24 a b c d

1.91 0.237 1.57 0.00440 1.72 1.87

0.209 0.170 0.365 1.12 3.33 1.91

2.09 0.141 0.758 0.0131 1.69 1.77

c2 calib.d 0.0591 0.0334 0.274 0.396 1.34 1.50

42 38 30 60 84 63 c2 all data 0.287 0.268 0.415 1.57 5.61 2.06

R2

0.07 0.07 0.11 0.14 0.20 0.17

0.980 0.993 0.961 0.850 0.965 0.969

0.08 0.08 0.12 0.16 0.26 0.18

0.973 0.989 0.956 0.789 0.941 0.966

Parameters defined in Equations (4)e(6). Root mean square error. Coefficient of determination. c2 calib. values in Table 4b are calculated using only the data involved in calibration.

obtained (Fig. 5A) except for the minimum-ICC curve after 100 h. Wanneroo water was tested with much higher ICCs than those of the other waters (Table 4a). Consequently, the initial (fast) decay is generally much greater and the time taken to approach zero residual is much longer. This more extreme behaviour is outside the normal operating range of ICCs, but constitutes a more severe test for any model to describe. Good model estimates were obtained for the decay curves with ICCs of 10 and 6 mg/L (Fig. 6A). The data for ICC 8 mg/L was under-estimated by about 0.2 mg/L, while data for ICC 4 mg/L was overestimated by about 0.5 mg/L. These estimation errors are comparable with the experimental error for high-range chlorine measurements by the DPD colorimetric method (Harp, 2002). They could also be partly due to the implicit higher weighting given to the residuals in the period 1 < t < 5 h by the much higher rate of measurement in this period. It should also be noted that Warton et al. (2006) used four parameters to characterize each decay test, in contrast to the four parameters used by the 2R model to characterize the entire dataset.

3.3.

Validation of the 2R model

For each water, the 2R model was then calibrated using only the minimum amount of data necessary e decay tests with two different ICCs. The decay tests with highest and lowest ICC were chosen on the assumption that interpolation to other conditions would be more reliable than extrapolation. The optimal parameter values and associated c2 obtained for each model are given in Table 4b. The calibrated models were then validated by predicting the decay curves for other tests with ICCs different from those used in model calibration. The decay curves estimated for the calibration data (solid black lines) and the predicted curves for the validation data (broken grey lines) are shown in Figs. 1D and 3Be6B. For Greenvale water, the residuals estimated by the model for the two tests used in the recalibration, and the predicted decay curves for the three intermediate ICCs (Fig. 3B), are difficult to distinguish from the curves generated from the model calibrated against all data (Fig. 3A). For Harbin and Warragamba water, a similar outcome was obtained, except that the model over-predicted the intermediate-ICC curve by

Fig. 3 e Decay tests and 2R model simulations for Greenvale water: (A) calibration using all data; (B) calibration using only minimum and maximum ICC data. Filled black points are calibration data; black curves are model estimates; open grey points are validation data, and broken grey curves are model predictions.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 9 6 e4 9 0 8

4905

Fig. 4 e Decay tests and 2R model simulations for Warragamba water: (A) calibration using all data; (B) calibration using only minimum and maximum ICC data. Filled black points are calibration data; black curves are model estimates; open grey points are validation data, and broken grey curves are model predictions.

about 0.1e0.2 mg/L for part of the time (Figs. 1D and 4B). As expected when fewer data are used, recalibration of the model to the Rocky Creek data for only the maximum and minimumICC tests produced better estimates of the calibration data (Fig. 5B) than those in Fig. 5A, obtained by using all the Rocky Creek data for calibration. This implies poorer prediction of the data not used in calibration. However, in Fig. 5B, both estimated and predicted curves still closely matched the measurements, except at three measurement times in the period 20 < t < 50 h, which appear to be inconsistent with the remaining data. Root mean squared errors (RMSE) are given in Table 4. This model error is of similar magnitude to the measurement error for pocket colorimeters. Hach Company (2006) quoted typical precision (95% CI) of 0.05 at 1 mg/L free chlorine and 0.2 at 5 mg/L. Consequently, for ICCs less than 5 mg/L, it was found that a 2R model using only the data from tests with minimum and maximum ICCs was sufficiently precise to interpolate to intermediate-ICC values between the conditions used for calibration. This finding is based on the close agreement between the grey curves in Figs. 1D, 3B, 4B and 5B (which are model predictions of data not used in the calibration) with the black curves in Figs. 1C, 3A, 4A and 5A respectively (which are model estimates of the same data when it is included in the calibration).

Wanneroo water was tested with ICCs in the much higher 4e10 mg/L range. Better estimates of high-ICC and low-ICC decay data (Fig. 6B) were obtained than those derived previously (Fig. 6A). However, the model under-predicted the intermediate-ICC data by about 0.5e0.7 mg/L, compared with 0.3e0.5 mg/L previously. The results in Fig. 6B suggested that it was the high-ICC decay that was faster than expected, especially in the longer term. The model was therefore recalibrated against only the second-highest and lowest- ICC data. As well as providing good estimates of the calibration data, the data for the intermediate ICC (6 mg/L) was then predicted very well (Fig. 6C). The highest-ICC (10 mg/L) data was then increasingly over-predicted for t > 50 h, by up to 1 mg/L. This is not considered to be important as it is well outside the ICC range of interest. A possible explanation for the behaviour at high ICC is that additional compound(s) in the water become reactants, as the concentration of chlorine increases. Regardless of the cause, it would be prudent to conduct at least one decay test more than those with highest and lowest ICC that are used for model calibration, when the range of ICCs to be modelled extends above 5 mg/L. Data from the additional test(s) could then be used to confirm whether intermediate-ICC data was well predicted by the initially calibrated model. If this prediction is inadequate, then recalibration over a more restricted ICC

Fig. 5 e Decay tests and 2R model simulations for Rocky Creek water: (A) calibration using all data; (B) calibration using only minimum and maximum ICC data. Filled black points are calibration data; black curves are model estimates; open grey points are validation data, and broken grey curves are model predictions.

4906

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 9 6 e4 9 0 8

Fig. 6 e Decay tests and 2R model simulations for Wanneroo water: (A) calibration using all data; (B) calibration using only minimum [4 mg/L] and maximum ICC [10 mg/L] data, and (C) calibration using only minimum [4 mg/L] and high [8 mg/L] ICC data. Filled black points are calibration data; black curves are model estimates; open grey points are validation data, and broken grey curves are model predictions.

range can be attempted. Such recalibration was successful in the Wanneroo case over the 4e8 mg/L range of ICC, as confirmed by the good prediction of the intermediate-ICC test data within that range. A quantitative comparison can be made between the c2 values in Table 4(a) with those calculated for all data in Table 4(b) to obtain the increase in modelling error that resulted from using a minimal data set. The increases are small, except for the full Wanneroo dataset. The increase in that case also becomes small, when the 10mg/L decay curve is removed. Coefficients of determination (R2) were also calculated for all calibrated models (Table 4). Values were greater than 0.94, except for Rocky Creek. That is, less than 6% of the variation in any of the data sets was unexplained by the models. The lower values (around 0.8) for Rocky Creek are due to the much smaller total variation to be explained, as well as the greater variability of measured points, especially near the starting time. The parameter values in Table 4 can be regarded as characteristics of each water. These values showed both similarities and marked differences between waters. The initial concentration of the fast reactant (c0F) was more than double that of the corresponding slow reactant (c0S), except for Wanneroo water, for which they were about equal. The greater amount of fast reactant resulted in the much higher decay that occurred in that water in the short term. For all waters, the fast decay rate coefficient (kF) was more than an order of magnitude greater than the corresponding slow coefficient (kS). The effect of the high degree of treatment on Rocky Creek water was reflected in both coefficients being more than an order of magnitude smaller than those of any other water.

Shang et al. (2008) more recently produced the multi-species extension (MSX) to EPANET, in which numerous simultaneous reactions between multiple species can be included at rates flexibly defined by the user. Consequently, users can now set up models with two (or more) reactants within EPANET or the Innovyze derivative product H2OMapMSX, which provides an additional Windows menu interface.

3.5.

The 2R model has been shown here to be capable of addressing the basic issue in a hierarchy of increasingly complex system planning/management issues, after it is embedded in a network model. It also appears to be suitable to address the higher level issues of temperature influence and rechlorination. With minor augmentation, the 2R model also becomes capable of direct representation of DBP concentrations throughout a distribution system (Fisher et al., 2004). Such capability could be used by water utilities in the derivation of least-cost measures to meet the more stringent limitation of DBPs which is required within a few years by the Stage 2 Disinfection By-Product Rule (USEPA, 2010). The effect of treatment on bulk decay has been successfully incorporated into the optimization process for derivation of parameter values in the 2R model. The specific case of enhanced coagulation was considered by combined application of a model for DOM removal (Kastl et al., 2004) and the 2R model for chlorine decay (Fisher et al., 2004).

4. 3.4.

Other capabilities of the two-reactant model

Conclusions

Implementation in distribution system models

When the 2R model was developed, network modelling software, such as EPANET and its commercial derivatives, only incorporated bulk chlorine decay rates that were functions of chlorine concentration. As there were no network simulation software packages capable of handling multiple reactants (species), this and a more complex four-reactant model were incorporated directly into EPANET to produce DSMtool (Fisher et al., 2004), but this package was not widely distributed outside Australia.

From model comparisons in the literature, it is clear that the FO model of chlorine decay in bulk water is not fit for the purpose of modelling to assist disinfection planning/ management in distribution systems. It does not meet the criteria for suitability to address even the basic issue in the hierarchy of disinfection issues and is inherently unable to represent rechlorination. From model comparisons made here, the SR model did not meet all these criteria and the VRC and 2R models are therefore the simplest models that do meet them. The 2R model

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 9 6 e4 9 0 8

was selected for validation because it showed potential to represent temperature effects (in addition to rechlorination), whereas the VRC model did not. For each water, optimal parameter values were readily obtained for the 2R model, which were invariant over the entire ICC range in each dataset. Furthermore, the ICCs set in the model were those actually measured, which avoids the need (later) to continually re-estimate that parameter in different system simulation scenarios. For three waters, chlorine residuals estimated from the calibrated model closely matched all decay test data (ICC range 1e4 mg/L). For the two waters tested over a higher ICC range (up to 10 mg/L), a similar close match was obtained for all but the lowest-ICC test in each case. When recalibrated against a “minimal” dataset consisting of only the test data with lowest and highest ICCs, the chlorine residuals estimated by the 2R model more closely matched the (fewer) calibration data used. For the first three waters, model predictions of the test data not used in the calibration were almost as good as those estimated from models fitted to the entire datasets. For the fourth water, model estimates of the low-ICC test data were considerably improved, while model predictions of the test data for the two intermediate ICCs were almost as good as those from the previous calibration. The 2R model was therefore considered validated for the ICC range of 0e5 mg/L, even when a “minimal” set of data is used for calibration. As this range covers the conditions encountered in most distribution systems, the 2R model is entirely suitable to address the basic issue in the disinfection planning/management hierarchy. While software such as AQUASIM provides the capability to find optimal parameter values for multiple-reactant decay models, the MSX extension to network modelling software provides the capability to implement such bulk-decay models in distribution network models. This provides an unprecedented opportunity for the network software user to more realistically plan and manage distribution systems to achieve disinfection (and other water quality) goals.

Acknowledgements Greenvale and Warragamba datasets were supplied by Sydney Water Corporation (SWC). They were generated earlier by the authors. The Greenvale dataset was an SWC contribution to the research program of the Australian Cooperative Research Centre for Water Quality and Treatment. The Rocky Creek dataset was supplied by Rous Water, NSW, Australia. The Harbin dataset was obtained from the American Chemical Society (License 2406160334024) and the Wanneroo dataset was obtained from Elsevier (License 2472920761136).

references

Boccelli, D., Tryby, M., Uber, J., Summers, S., 2003. A reactive species model for chlorine decay and THM formation under rechlorination conditions. Water Research 37 (11), 2654e2666.

4907

Clark, R., 1998. Chlorine demand and TTHM formation kinetics: a second order model. Journal of Environmental Engineering, ASCE 124 (1), 16e23. Fisher, I., Kastl, G., 1996. Numerical modelling of water quality in distribution systems. In: Proceedings, Australian Water and Wastewater Association WaterTECH Conference, May 1996, Sydney, Australia, p. 461ff. Fisher, I., Kastl, G., Sathasivan, A., Chen, P., van Leeuwen, J., Daly, R., Holmes, M., 2004. Tuning the enhanced coagulation process to obtain best chlorine and THM profiles in the distribution system. Water Science and Technology: Water Supply 4 (4), 235e243. Fisher, I., Kastl, G., Sathasivan, A., Jegatheesan, V., 2011. Suitability of chlorine bulk decay models for planning and management of water distribution systems. Critical Reviews in Environmental Science and Technology 41 (20). doi:10.1080/ 10643389.2010.495639. Gill, P., Murray, W., Wright, M., 1981. Practical Optimization. Academic Press, London. Haas, C., Karra, S., 1984. Kinetics of wastewater chlorine demand exertion. Journal of the Water Pollution Control Federation 56 (2), 170e173. Hach Company, 2006. Pocket Colorimeter II Analysis Systems e Instruction Manual, fourth ed., Loveland, CO, pp. 59570e59588. Haimes, Y., Hall, W., Freedman, H., 1975. Multiobjective Optimization of Water Resources Amsterdam & New York. Harp, D., 2002. Current Technology of Chlorine Analysis for Water and Wastewater. In: Technical Information Series d Booklet, vol. 17. Hach Company, Loveland, CO. Hua, F., West, J., Barker, R., Forster, C., 1999. Modelling of chlorine decay in municipal water supplies. Water Research 33 (12), 2735e2746. Jadas-He´cart, A., El Moher, A., Stitou, M., Bouillot, P., Legube, B., 1992. The chlorine demand of a treated water (in French). Water Research 26 (8), 1073e1084. Johnson, J., 1978. Measurement and persistence of chlorine residuals in natural waters. In: Jolley, R.L., et al. (Eds.), Water chlorination: environmental impact and health effects, vol. 1. Ann Arbor Science, Ann Arbor, MI, p. 37ff. Jonkergouw, P., Khu, S.-T., Savic, D., Zhong, D., Hou, X., Zhao, H.B., 2009. A variable rate coefficient chlorine decay model. Environmental Science and Technology 43 (2), 408e414. Kastl, G., Fisher, I., Jegatheesan, V., 1999. Evaluation of chlorine kinetics expressions for drinking water distribution modelling. Journal of Water Supply: Research and Technology e Aqua 48 (6), 219e226. Kastl, G., Sathasivan, A., Fisher, I., van Leeuwen, J., 2004. Modeling DOC removal by enhanced coagulation. Journal of American Water Works Association 96 (2), 79e89. Kie´ne´, L., Lu, W., Le´vi, Y., 1998. Relative importance of phenomena responsible for chlorine decay in drinking water systems. Water Science and Technology 38 (6), 219e227. LeChevallier, M., 1999. The case for maintaining a disinfectant residual. Journal of American Water Works Association 91 (1), 86e94. Nelder, J., Mead, R., 1965. A simplex method for function minimization. Computer Journal 7 (4), 308e313. Powell, J., Hallam, N., West, J., Forster, C., Simms, J., 2000. Factors which control bulk chlorine decay rates. Water Research 34 (1), 117e126. Ralston, M., Jennrich, R., 1978. DUD e a derivative-free algorithm for nonlinear least squares. Technometrics 20 (1), 7e14. Reichert, P., 1994. AQUASIM e a tool for simulation and data analysis of aquatic systems. Water Science and Technology 30 (2), 21e30. Reichert, P., 1998. AQUASIM 2.0 e User Manual. Swiss Federal Institute for Environmental Science and Technology, Du¨bendorf, Switzerland.

4908

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 8 9 6 e4 9 0 8

Rossman, L., 1994. EPANET Users Manual Version 1.1. US Environmental Protection Agency, Cincinnati, Ohio. Rossman, L., 2000. EPANET 2 Users Manual. US Environmental Protection Agency, Cincinnati, Ohio. Shang, F., Uber, J., Rossman, L., 2008. Modeling reaction and transport of multiple species in water distribution systems. Environmental Science and Technology 42 (3), 808e814. USEPA, 2002. National Primary Drinking Water Regulations, 40 CFR Part 141, p. 436.

USEPA, 2006. Stage 2 Disinfectants and Disinfection Byproducts Rule (Stage 2 DBPR) 71 FR 388 January 4, 2006, vol. 71, no. 2. USEPA, 2010. Stage 2 Disinfectants and Disinfection Byproducts Rule Consecutive Systems Guidance Manual Office of Water (4607M) EPA 815-R-09e017. Warton, B., Heitz, A., Joll, C., Kagi, R., 2006. A new method for calculation of the chlorine demand of natural and treated waters. Water Research 40 (15), 2877e2884. WHO, 2004. Guidelines for Drinking Water Quality, third ed. World Health Organisation, Geneva, Switzerland.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 0 9 e4 9 2 2

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Denitrification and dissimilatory nitrate reduction to ammonium (DNRA) in a temperate re-connected floodplain F. Sgouridis a, C.M. Heppell a,*, G. Wharton a, K. Lansdown a,b, M. Trimmer b a b

School of Geography, Queen Mary University of London, Mile End Road, London E1 4NS, United Kingdom School of Biological and Chemical Sciences, Queen Mary University of London, London E1 4NS, United Kingdom

article info

abstract

Article history:

The relative magnitudes of, and factors controlling, denitrification and dissimilatory

Received 26 January 2011

nitrate reduction to ammonium (DNRA) were measured in the soil of a re-connected

Received in revised form

temperate floodplain divided into four different land management zones (grazing grass-

26 April 2011

land, hay meadow, fritillary meadow and a buffer zone). Soil samples were collected from

Accepted 26 June 2011

each zone to measure their respective potentials for nitrate attenuation using

Available online 8 July 2011

the surface and at depth in the soil column and additional samples were collected to

15

N both at

measure the lability of the organic carbon. Denitrification capacity ranged between 0.4 and Keywords:

4.2 (mmol N g
1 dry soil d
1) across the floodplain topsoil and DNRA capacity was an order

Denitrification

of magnitude lower (0.01e0.71 mmol N g
1 d
1). Land management practice had a signifi-

DNRA

cant effect on denitrification but no significant effects were apparent for DNRA. In this

Floodplain

nitrogen-rich landscape, spatial heterogeneity in denitrification was explained by differ-

Lability

ences in lability and the magnitude of organic carbon associated with different manage-

Land use management

ment practices (mowing and grazing). The lability of organic carbon was significantly

River restoration

higher in grazing grassland in comparison to other ungrazed areas of the floodplain, and consequently denitrification capacity was also highest in this area. Our results indicate that bacteria capable of DNRA do survive in frequently flooded riparian zones, and to a limited extent, compete with denitrification for nitrate, acting to retain and recycle nitrogen in the floodplain. Exponential declines in both denitrification and DNRA capacity with depth in the floodplain soils of a hay meadow and buffer zone were controlled primarily by the organic carbon content of the soils. Furthermore, grazing could be employed in re-connected, temperate floodplains to enhance the potential for nitrate removal from floodwaters via denitrification. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

To date, the main aim of many floodplain restoration schemes has been to increase the flood storage capacity of the floodplain, thereby providing sustainable, inexpensive flood defence through downstream flow attenuation and, as a side effect, ecological benefits for the floodplain in terms of

increased biodiversity (Blackwell and Maltby, 2005; Acreman et al., 2007). Re-instating river-floodplain connectivity has also been proposed as a means of improving the water quality of rivers, through, for example, nitrate removal from overbank floodwater, which is considered an important floodplain service (Tockner and Stanford, 2002). Denitrification has been shown to be the dominant pathway for the permanent

* Corresponding author. Tel.: þ44(0) 207 882 2768; fax: þ44(0) 208 981 6276. E-mail address: [email protected] (C.M. Heppell). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.06.037

4910

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 0 9 e4 9 2 2

removal of nitrate in temperate floodplain environments (Baker and Vervier, 2004; Forshay and Stanley, 2005), and is pervasive because the main conditions for the occurrence of the process (i.e. organic carbon supply for microbial respiration, an absence of oxygen and the supply of nitrate) are met during flooding with nitrate-rich water. Despite the wealth of evidence for nitrogen removal from the hydric surface and subsurface soils of floodplains and riparian buffer zones (Peterjohn and Correll, 1984; Ambus and Lowrance, 1991; Groffman et al., 1992; Pinay et al., 1993, 2007; Hefting et al., 2003), only a handful of studies have provided evidence for nitrogen removal in restored floodplains (Hoffmann et al., 1998; Sheibley et al., 2006; Orr et al., 2007), and investigated the factors controlling nitrogen removal processes in these re-connected environments (Tockner et al., 1999; Olde Venterink et al., 2003; Van Der Lee et al., 2004). And to ensure the success of future river-floodplain restoration projects aimed specifically at nitrogen removal (Brettar et al., 2002; Hoffmann and Baattrup-Pedersen, 2007) the land management regimes implemented following restoration need careful consideration. Traditional land use management of temperate, lowland agricultural grasslands includes mowing for hay, and grazing by ungulates (Robson et al., 2007). Higher denitrification potential in areas affected by grazing cattle (Frank and Groffman, 1998; Meneer et al., 2005; Patra et al., 2005; Philippot et al., 2009) has been attributed to changes in hydraulic soil properties through soil compaction and in nutrient availability through the deposition of urine and faeces and limited plant N uptake after defoliation (Luo et al., 1999; Meneer et al., 2005). Mowing has been shown to stimulate microbial activity (Patra et al., 2006; Ilmarinen et al., 2009) by increasing the availability of C and N, while the release of easily extractable carbon via rhizo-deposition (Gavrichkova et al., 2008) has been positively related to enhanced denitrification (Robson et al., 2007). In contrast, arable farming has been shown to negatively affect the availability of labile organic carbon in soils (Bowman et al., 1990; Groffman et al., 1993; Boyer and Groffman, 1996; Ullah and Faulkner, 2006) and therefore significantly decrease denitrification potential (Groffman et al., 1993; Cavigelli and Robertson, 2000; Ullah and Faulkner, 2006). Thus, there is a need to better understand the factors that control nitrate removal in the topsoil of restored floodplains and how these factors may be affected by land use management practices, following restoration. Furthermore, to our knowledge none of the studies in restored floodplains have, as yet, quantified denitrification in the vertical dimension, and the impact this might have on nitrogen removal from floodwaters, despite the importance of quantifying the cycling of terrestrial N at depth in the soil for the purpose of constructing comprehensive soil N budgets (Krug and Winstanley, 2002). Dissimilatory nitrate reduction to ammonium (DNRA) is another potentially significant component of the nitrogen cycle in temperate re-connected floodplain environments (Burt et al., 2010). Whereas the conditions promoting DNRA and heterotrophic denitrification are similar (absence of oxygen, and available nitrate and organic substrates), denitrification represents a permanent nitrogen removal pathway, while DNRA is a nitrogen-conserving mechanism that

transforms nitrate to another more bio-available inorganic-N form, ammonium (NHþ 4 ). Recently, the role of DNRA has been established in nitrogen limited tropical forest soils (Silver et al., 2001, 2005; Pett-Ridge and Firestone, 2005; Pett-Ridge et al., 2006; Huygens et al., 2007; Ru¨tting et al., 2008; Templer et al., 2008) where DNRA was shown to be at least three times higher than denitrification, or in some cases accounted for nearly 99% of nitrate reduction (Huygens et al., 2007). In contrast, DNRA has been shown to account for only 5e15% of nitrate reduction in temperate freshwater environments, such as riparian fens and wetlands (Ambus et al., 1992; Matheson et al., 2003; Scott et al., 2008), and a similar range in paddy soils (Yin et al., 2002), with denitrification responsible for the majority of nitrate removal through the production of N2 gas. Riparian zones have been confirmed as active sites of DNRA but the factors controlling its magnitude and extent have not yet been verified (Davis et al., 2008; Woodward et al., 2009). In the context of river-floodplain restoration projects, it is desirable to understand whether DNRA might be a significant process in temperate floodplains, and whether DNRA needs to be considered in schemes designed to optimise nitrogen removal. Consequently, the main objectives for the present study were to investigate the relative magnitudes of, and the factors controlling, both denitrification and DNRA capacity in a reconnected floodplain within a temperate agricultural landscape. By studying spatial and vertical variations in these microbially-mediated processes in floodplain zones subject to different management regimes (grazing and mowing) we sought to understand the influence of management practices on the magnitude of denitrification and DNRA.

2.

Methods

2.1.

Study site description

The study site comprises a 2 km stretch of the River Cole floodplain (NGR: SU 208970, 50 ha, Fig. 1). The valley of the River Cole is floored with alluvium. Thames gravels and Oxford Clay underlie the alluvium, and the gravels locally outcrop in the river banks (Sear and White, 1994). The main soil associations in the area are Pelo-Stagnogley/Stagnogley, Brown Redzina, Gleyic brown calcareous earths, and grey Redzinas. The predominant topsoil textures in the floodplain are silty clay and silty clay loam with the silt and clay fractions ranging between 68 and 88% of the particle size distribution. The site was part of a EU-Life Demonstration Project for damaged rivers and floodplains, completed in 1995 (Holmes and Nielsen, 1998); the main objectives of the Life project being the restoration of the river and floodplain in terms of physical features, flood storage, habitat diversity, and visual appearance (Janes et al., 1999). This was achieved by both new channel creation and reshaping of existing channels. During this process, channel width was reduced by c. 5 m, and channel length was increased by about 8%. Bankfull capacity was reduced by raising bed levels by about 1 m. This aimed to restore more frequent seasonal flooding to adjacent fields which would be farmed less intensively (Janes et al., 1999). Flooding was allowed through the connection of the new

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 0 9 e4 9 2 2

4911

Fig. 1 e The floodplain of the River Cole (Coleshill, Oxfordshire, UK) illustrating elements of the EU-Life demonstration restoration project, and current land management zones: FM [ Fritillary Meadow; HM [ Hay Meadow; BZ [ Buffer Zone and GG [ Grazing Grassland (reproduced with permission from the River Restoration Centre).

channel with a former drainage ditch surrounding the hay meadows (labelled HM in Fig. 1) in the downstream section. Fig. 1 illustrates the various elements of the restoration project along the restored reach that was the focus of this study. Following the implementation of the restoration plans, flood frequency and river valley inundation have increased considerably. During the period of the study (2006e2008), commencing eleven years after the restoration, the downstream floodplains were inundated 13% and 28% of the year, exceeding the predicted duration of overbank flooding (2e3 weeks every winter) by the restoration project. Due to the lowangle topography and the clay character of the floodplain alluvium, saturated conditions persisted for long periods of time, typically ranging between 5 and 26 continuous days.

The four land management zones of the restored River Cole reach are shown in Fig. 1. The Grazing Grassland (GG) is the area between the new meandering river and the retained mill leat in which the dominant land use is grazing by cattle. The Buffer Zone (BZ) area is part of a buffer strip maintained between the arable land used for winter cereals (oil seed rape and flax) on the west and the new meandering section of the river downstream from Coleshill Bridge. The arable land to the east of the downstream section of the river was reclaimed during the EU-Life project and turned into hay meadow (HM), where grazing is not allowed and hay is cut once a year. Overbank flooding is encouraged by the connection of the old drainage ditch, surrounding the HM, to the meandering river. Here, overflowing of the re-connected drainage ditch has

4912

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 0 9 e4 9 2 2

created a linear depression on the floodplain spreading across the HM with direction from east to west. The Fritillary Meadow (FM), north of the HM, has been conserved and flooding is also encouraged from both the river and the drainage ditch.

2.2.

Sampling strategy

To investigate the potential for nitrogen removal across the different land management zones of the re-connected floodplain, a stratified systematic soil sampling approach was followed. Each land management zone was divided into three sampling transects and five topsoil (0e20 cm) samples were collected per transect; a total of 15 samples per floodplain zone. Soil samples were collected in the FM in July 2006, in the GG and the BZ in June 2007 and in the HM in July 2007. The GG, BZ and HM samples were all collected within seven days following a 4-day period of inundation by floodwater. The FM samples were collected from topsoil which had been flooded approximately three months prior to sampling, and had been subjected to wetting and drying cycles due to rainfall events between flooding and sampling. Samples were collected with the aid of small metal cores (38 mm I.D., 23 cm long, ELE International, Bedfordshire, UK). The samples were stored at 4  C overnight. The next day visible stones and roots were removed manually and the top 0e10 cm of the soil was well homogenised by mixing before analysis. In order to investigate the effect of vertical changes in soil stratigraphy and electron donor availability on potential nitrate reduction pathways, eight soil samples were collected from six depths (0e10; 10e20; 20e30; 50e70; 80e100 and 100e120 cm) by hand augering. Bulk density rings were also collected for each depth increment, while the storage, transport and processing of the soil samples followed that of the surface soil samples.

2.3.

Soil properties

The soil samples were analysed for soil bulk density, porosity, moisture content, organic matter content (by Loss on Ignition) and water filled pore space (WFPS) following the methods described in Rowell (1994), Heiri et al. (2001) and Linn and Doran (1984) respectively. The total organic carbon (TOC) and total nitrogen (TN) content of the soil samples were determined with elemental analysis (Elemental Analyser; Flash EA 1112, Thermo-Finnigan, Bremen, Germany). For determining the TOC content, all the samples were acidified prior to analysis in order to remove any carbonates (Hedges and Stern, 1984). Sample TOC and TN contents were calculated from an L-aspartic acid (C4H7NO4) standard (C: 36.09%; H: 5.30%; N: 10.52%; O: 48.08%), using the Eager 300 software and linear regression. The precision as a coefficient of variation was better than 1%. All soil samples (except for the FM
þ floodplain area) were extracted for NO
3 , NO2 and NH4 anal
ysis with 2 M KCl (Rowell, 1994). The analysis for NO
3 , NO2 þ and NH4 was performed on a segmented continuous flow auto analyser (SANþþ, SKALAR, Delft, The Netherlands) according to standard colorimetric techniques (Kirkwood, 1996). Peak integration and calibration with linear regression was performed with the FlowAccess software.

2.4.

Denitrification capacity

To measure denitrification we incubated anoxic samples of soil with 15NO
3 without any addition of organic carbon and refer to this simply as denitrification capacity (Myrold and Tiedje, 1985; Yeomans et al., 1992; Well et al., 2005). This method represents the active denitrifier biomass of the soil at the time of sampling and is closely related to field denitrification rates (Smith and Tiedje, 1979; Myrold and Tiedje, 1985). Such an approach was well suited for our comparisons between sites (Groffman et al., 2006) and, further, the use of 15 N-labelled nitrate avoids the complications of acetylene block (reviewed in Groffman et al., 2006). Each homogenised soil sample was divided into four subsamples by weighing approximately 1.1 g of field moist soil into gas-tight vials fitted with butyl rubber septa (3 mL, Exetainer vial; Labco Ltd., High Wycombe, United Kingdom). The vials were then transferred to an anoxic glove box (Belle Technology, UK) and homogeneous slurries prepared at a 1:1 ratio of soil to water by addition of 1 mL of degassed (oxygen free nitrogen, OFN) synthetic river water (Smart and Barko, 1985). The vials were then sealed, removed from the glove box and pre-incubated on rollers (Spiromix, Thermo) in the dark at room temperature (22e25  C) for 24 h. This preincubation aims to remove any traces of oxygen and ambient 14NO
3 from the porewater and we depleted the background nitrate pool by 96.3 (SE 0.7) % on average. Following pre-incubation, the first sub-sample was left nonenriched as a reference for natural abundance 15N and ambient nitrate and microbial activity was inhibited by injecting ZnCl2 (100 mL 50% w/v) through the septum. The remaining three subsamples were then enriched by injecting a concentrated and degassed (OFN) stock of labelled 15NO
3 15 (75 mL of 2.4 mM Na15NO
N atom %] SigmaeAldrich, 3 [99.3 Poole, UK) through the septa using a gas-tight syringe (Hamilton, 1750 RN, 500 mL, VWR International) and returned to the rollers. The final concentration of nitrate in the slurry samples was 90 mM on average, which was lower than the average ambient porewater concentration of nitrate (500 mM  SE 0.05) but still vastly in excess of apparent KM values for denitrifying bacteria (typically <2 mM NO
3 ). For the topsoil samples the vials were sacrificed every 2 h for 6 h by injecting ZnCl2 through the septa as above and in the subsurface samples, where activity was much lower, every 6 h for 24 h; this regime ensured a good linear production of 15N2 gas throughout the incubation. At the end of the experiment, all of the slurry samples were gently centrifuged for 5 min at 3000 rpm to remove any soil particles from the underside of the septum which could potentially block the needle and interfere with the analysis (see below). Samples of the headspace (50 mL) in each vial were injected using an autosampler (Multipurpose Sampler MSP2, Gerstel, GmbH, Germany) into an elemental analyser (Flash EA 1112, Thermo-Finnigan, Bremen, Germany) interfaced (ConFlo III Interface, Thermo-Finnigan, Bremen, Germany) with a continuous flow isotope ratio mass spectrometer (CF/IRMS, Finnigan MAT DeltaPlus, Thermo-Finnigan, Bremen, Germany). The mass spectrometer was calibrated with N2 in helium over airequilibrated water at 22  C and the mass charge ratios for m/z 28, m/z 29 and m/z 30 (28N2, 29N2, and 30N2 respectively) were

4913

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 0 9 e4 9 2 2

measured. Precision as a coefficient of variation was better than 1%. After pre-incubation of the samples and enrichment with 29 N2 in the 15N2 gas 99.2% atom 15NO
3 , the proportion of should have been less than 1.5%, assuming random pairing of 14 N and 15N through denitrification and the absence of anammox (Hauck et al., 1958; Thamdrup and Dalsgaard, 2002). Indeed, the 24 h pre-incubation of the samples was so effective that only trace amounts of 29N2 were detected with the CF/ IRMS and we could be confident that denitrification was the sole source of 15N2 in our samples. This enabled simplification of the calculations used in the isotope pairing technique (Nielsen, 1992) whereby denitrification could be estimated solely from the production of 30N2 according to and adapted from Thamdrup and Dalsgaard (2000): p mmol N g
1 d


1

¼ 2: 
1

X

# "   30  30 N N2 P 2
P N2 sample N2 reference N2


1

sample

p DNRAmmol Ng


1

d


1

# "   29  29 N2 N2 P ¼
P N2 sample N2 reference #! "    30 30 N2 N2 (2)
P þ 2: P N2 sample N2 reference X Ve  N2 sample : a
1 : Vs : : m
1 : t
1 Vs

where p DNRA mmol N g
1 d
1 is the production of 15N in ammonium; the fractions 29N2/SN2 and 30N2/SN2 represent the signal ratio for 29N2 and 30N2 to total signal SN2 (total signal for m/z ratios 28, 29 and 30 for either sample or reference, respectively); a is the calibration factor as above; Vs is the volume of the sub-sample analysed (L); Ve is the total volume of the KCl extract (L); m the dry mass of the soil sample (g) and t (days) the duration of linear p29N2 and p30N2 production.

(1)

: a
1 : V : m
1 : t
1 15

where p mmol N g d was the production of N measured as N2; the fraction 30N2/SN2 represents the signal ratio for 30N2 to total signal SN2 (total signal for m/z ratios 28, 29 and 30 for either sample or reference, respectively); a is the calibration factor (signal/mmol N2 L
1) obtained by measuring vials containing atmospheric air; V the volume of the sample (L); m the dry mass of the soil sample (g) and t (days) the duration of linear p30N2 production.

2.6. Lability of organic carbon and potential methane production

30

2.5.

DNRA capacity

We used the same slurry samples as for the denitrification assay to measure the DNRA capacity of the soil using an adaptation of a combined microdiffusion-hypobromite oxidation method (Risgaard-Petersen et al., 1995). Having quantified the 15N2 gas production in the headspace the vials were opened and any NHþ 4 in the slurries extracted with 2 M KCl (4:1 ratio). The KCl extract was then filtered (0.45 mm PTFE syringe filter, Eurolab, VWR International) and 1 mL transferred into a new 3 mL gas-tight vial and the remaining sample frozen (
18  C) for later colorimetric analysis of ammonium. Reaction-needles were prepared by flattening the tip of a syringe needle and adding 50 mL of hypobromite iodine solution (prepared according to Rysgaard and RisgaardPetersen, 1997) into the luer of each needle. The ‘reaction needles’ were then transferred tip down into the vials containing the KCl extracts, which were then capped. To decrease the background of atmospheric N2 in the samples and also to remove any 15N-labelled N2 carried over from denitrification, the vials were flushed with He (CP grade) for 10 min. Ammonium ions were then converted to volatile NH3 by increasing the pH of the KCl extract to >12 by addition of 50 mL 12 M NaOH through the septum of each vial. The vials were then agitated gently and left for at least 24 h at 22  C in order to allow NH3 to react with the hypobromite and be oxidised to N2. The recovery and time to complete oxidation was calibrated against standards of known amounts of 15NHþ 4 . The isotopic composition of N2 in the headspace of the vials was then analysed as above and DNRA capacity in soil samples calculated according to:

Several studies have measured the anaerobic mineralisation rate of organic carbon (Bijay et al., 1988; Drury et al., 1998; Simek et al., 2000; Hill and Cardaci, 2004; Ullah and Faulkner, 2006; Dodla et al., 2008) as an expression of the fraction of labile organic carbon available for nitrate reduction (Gale et al., 1992). In the present study, we measured the anaerobically mineralisable organic carbon through the evolution of CO2 without addition of nitrate. Therefore, the method selected for this study represents a combined measurement of the relative availability of labile organic carbon and ambient availability of electron acceptors for anaerobic respiration and fermentation. Field moist soil (5.5 g) was weighed into gas-tight vials (12.5 mL, as above), which were then transferred to the anoxic glove box as above and capped. The soil samples were then incubated at room temperature (constant 22  C) and the headspace sampled every 48 h for 21 days. A 100 mL of headspace was withdrawn from the headspace using a gas-tight syringe in an auto-sampler (Multipurpose Sampler MSP2, Gerstel, GmbH, Germany) and injected into a GC-FID (7890A GC Agilent Technologies Ltd., Cheshire, UK) (Column 40  C, detector 300  C). Headspace concentrations of CO2 were calculated from peak areas using an electronic integrator (ChemStation, software), and were calibrated against known standards (Scientific and Technical Gases Ltd, Staffs, UK). Precision, as a coefficient of variation, was better than 1% for CO2. Production of CO2 was calculated from the first 17 days of the incubation over which time the accumulation CO2 was linear and expressed as mmol CO2 g
1of dry soil day
1. The lability of organic carbon was calculated as the rate of carbon mineralised per day as a proportion of the total organic carbon content of the sample (Dauwe et al., 2001).

2.7.

Statistical analysis

All data were tested for normality and homogeneity of variance with the KolmogoroveSmirnov test and the Levene

4914

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 0 9 e4 9 2 2

statistic respectively. If data did not meet the requirements they were log-transformed before statistical analysis. Comparison of independent samples’ variance was performed primarily with One-way ANOVA combined with the Least Significant Difference (LSD) post hoc test for the assessment of inter-sample group differences. The variance of those samples that were not log-normally distributed was tested with the non-parametric KruskaleWallis test (for more than two groups of samples) or the ManneWhitney U-test (between two groups of samples). Multivariate linear regression was used to explore the factors controlling denitrification and DNRA in the floodplain topsoils (0e10 cm depth, all land management zones) and in the subsurface soils (10e120 cm depth, HM and BZ only) using forward, stepwise regression to determine the predictor variables. In each case sample datasets were tested for multicollinearity using Pearson correlation coefficients and variance inflation factor (VIF), and residual autocorrelation using Durbin-Watson’s test. All statistical analyses were performed using SPSS 11.5 for Windows (SPSS 2002, Chicago, Illinois, USA).

3.

Table 2 e Comparison of physicochemical properties’ variance between floodplain areas. df; degrees of freedom between groups of samples, F; F statistic, c2; c2 statistic P; probability level. Where the variable is annotated with * the non-parametric KruskaleWallis test has been used instead of One-way ANOVA. Topsoil (0e10 cm) One-way ANOVA and KruskaleWallis

Bulk density Organic carbon* Total nitrogen* C/N ratio Nitrate Ammonium

df

F or c2

P

3 3 3 3 2 2

3.9 4.9 27.6 109.0 6.7 3.7

<0.05 >0.05 <0.01 <0.01 <0.01 <0.01

3.2. Variation in rates of nitrate reduction and carbon lability in the topsoil Denitrification capacity varied between 0.4 and 4.2 mmol N g
1 dry soil d
1 in the different land management zones of the floodplain. On average, denitrification capacity was highest (ANOVA; F ¼ 5.96, df ¼ 3, P < 0.01) in the GG (1.6  0.3 mmol N g
1 d
1), lowest in the FM (0.7  0.07 mmol N g
1 d
1), and intermediate in the BZ and HM (1.1  0.1 and 1  0.1 mmol N g
1 d
1 respectively) (Fig. 2a). DNRA activity ranged between 0.01 and 0.71 mmol N g
1 d
1 and was in an order of magnitude lower than denitrification on average, whilst no significant differences (ANOVA; F ¼ 2.08, df ¼ 3, P > 0.05) were found between the different floodplain areas (Fig. 2b). The relative importance of denitrification versus DNRA, expressed by the ratio of mean denitrification/DNRA capacity, decreased in the order FM > HM > BZ < GG (Fig. 2c); although the difference between floodplain areas was not statistically significant (ANOVA; F ¼ 1.31, df ¼ 3, P > 0.05). Rates of anaerobic CO2 production were highest in the GG at up to 12.4 mmol CO2 g
1 dry soil d
1 (ANOVA; df ¼ 3, F ¼ 11.6, P < 0.001), with no significant differences between the remaining three zones (FM, HM and BZ; data not shown), where the highest rate was only 2.4 mmol CO2 g
1 dry soil d
1 in the BZ. The lability of organic carbon (Fig. 3a and b) calculated as the ratio of CO2 production to soil organic carbon content, was markedly higher in the GG in comparison to the other sites (FM, HM and BZ; ANOVA; F ¼ 23.7, df ¼ 3, P < 0.01). There was a significant positive correlation between both denitrification and DNRA activities and the lability of organic carbon across the different floodplain areas (r ¼ 0.75 and r ¼ 0.49, respectively, P < 0.001).

Results

3.1. Variation in topsoil physical and chemical characteristics in the reconnected floodplain The physico-chemical properties of the topsoil (0e10 cm) in the four floodplain areas are summarised in Table 1 and the results of the One-way ANOVA or KruskaleWallis test are given in Table 2. Dry bulk density was significantly higher in the buffer zone (BZ) and the grazing grassland (GG) compared to the fritillary meadow (FM) and hay meadows (HM). The soil organic carbon content was not different between the floodplain areas and the values (1.5%e12%) are characteristic of mineral agricultural soils. Total nitrogen (TN) content varied between 0.33 and 0.91%, being lowest in the FM then increasing in the order HM < BZ ¼ GG. Consequently, the organic C/N ratios were significantly lower in the GG and BZ areas (3e7:1), compared to the HM and FM (8e16:1). Soil samples were analysed for nitrate and ammonium in all areas apart from the FM, and both the soil nitrate and ammonium content followed the same pattern as the TN content between the floodplain areas (HM < BZ ¼ GG).

Table 1 e Soil physico-chemical properties in the four land management zones of the River Cole. Data are means ± standard error (SE) in parenthesis (n [ 15). Similar lower case letters indicate no significant statistical differences between land management zones. Missing data are replaced by a (L) sign. Dry Bulk Density Clay content Organic carbon Total Nitrogen Ratio C/N (g cm
3) Fritillary Meadow (n ¼ 15) Hay Meadow (n ¼ 15) Buffer Zone (n ¼ 15) Grazing Grassland (n ¼ 15)

0.8 (0.05)b 0.7 (0.02)b 1 (0.07)a 0.9 (0.05)a

(%) 43 31 36 29

(1.5)a (3.3)b (0.6)c (1.6)b

(%) 7 5 5 7

(0.7)a (0.4)a (0.3)a (0.6)a

(0.07)a (0.03)b (0.02)c (0.09)c

NHþ 4 -N

(mmol g
1) (mmol g
1)

(%) 0.6 0.4 0.9 0.9

NO
3 -N

10 10 4 5

(0.1)a (0.5)a (0.1)b (0.3)b

e 0.7 (0.12)a 1.4 (0.36)b 1.5 (0.32)b

e 0.2 (0.05)a 0.6 (0.21)b 1.4 (0.81)ab

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 0 9 e4 9 2 2

4915

3.3. Variation in subsurface soil physical and chemical characteristics Changes in soil physico-chemical properties with depth in the BZ and HM areas are summarised in Fig. 4. Dry bulk density generally increased with depth, and was consistently higher in the BZ compared to the HM (Fig. 4a). NO
3 eN content also decreased with depth through the profile, and the decrease was more marked in the BZ in comparison to the HM (Fig. 4b). The organic carbon content declined steadily with depth through the A horizon. In the BZ there was no further decline in organic carbon content below 30 cm depth, whereas in the HM the organic carbon content continued to decrease throughout the B and C soil horizons (Fig. 4c).

3.4. Variation in rates of nitrate reduction and carbon lability in the subsurface Fig. 5 illustrates the vertical distribution of denitrification capacity and DNRA in the subsurface of the BZ and HM. The highest denitrification rate was observed in the top 10 cm of the soil and subsequently decayed exponentially with depth in both areas (Fig. 5a). The mean denitrification rate in both the BZ and HM decreased from 3.71 to 1.96 mmol N g
1 dry soil d
1 within 10 cm of the topsoil and then to 0.13 mmol N g
1 dry soil d
1 by 120 cm. There was no significant difference in denitrification between the HM and BZ subsurface, apart from at 20 cm and 120 cm. DNRA activity was measured throughout the subsurface of both areas (Fig. 5b). The mean rate of DNRA between BZ and HM decreased from 0.25 to 0.12 mmol N g
1 dry soil d
1 within 10 cm of the topsoil, and to 0.01 mmol N g
1 dry soil d
1 by 120 cm. The ratio of denitrification to DNRA was relatively constant between 10 cm and 70 cm, while it increased deeper into (100 cme120 cm) the BZ area (Fig. 5c). In the HM, the ratio of denitrification to DNRA increased between 20 and 30 cm and 100 and 120 cm. The production of CO2 followed a similar exponential decay from between 0.8 and 1.4 mmol CO2 g
1 dry soil d
1 for the top of the BZ and HM, respectively, to <0.2 mmol CO2 g
1 dry soil d
1 at 120 cm (Fig. 6a). The lability of organic carbon remained comparatively constant with depth throughout the A horizon, and then increased in the B and C horizons of the HM, where it was greater overall throughout the soil column (Fig. 6b).

3.5. Factors controlling denitrification and DNRA activity in the floodplain

Fig. 2 e Mean nitrogen transformation rates in the different land management zones (0e10 cm depth): (a) Denitrification capacity, (b) DNRA capacity, (c) Ratio of Denitrification versus DNRA: FM [ Fritillary Meadow; HM [ Hay Meadow; BZ [ Buffer Zone and GG [ Grazing Grassland. Significant differences indicated with different lower case letters. Error bars indicate standard error (SE), n [ 15.

Multiple linear regression was performed to investigate the factors controlling denitrification and DNRA activity in the topsoil (0e10 cm) of all of the floodplain areas. Accordingly, lability and the quantity of organic carbon were the most important factors controlling denitrification in the topsoil (Table 3), explaining 56% of the variability in denitrification capacity across all land management zones of the floodplain. Nitrogen species, lability, C:Nitrate and texture (% clay) together accounted for 74% of variability in DNRA (Table 3). A combination of four predictor variables: organic carbon content, lability, bulk density and C:Nitrate could explain 84% of the variability in denitrification capacity within the soil subsurface, whilst a combination of organic carbon and bulk

4916

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 0 9 e4 9 2 2

Fig. 3 e Mean carbon transformation rates in the different land management zones (0e10 cm depth): (a) Production of carbon dioxide (b) Lability of organic carbon in the four floodplain management zones of the River Cole: FM [ Fritillary Meadow; HM [ Hay Meadow; BZ [ Buffer Zone and GG [ Grazing Grassland. Significant differences indicated with different lower case letters. Error bars indicate standard error (SE), n [ 15.

density explained 36% of the variability in DNRA between 10 and 120 cm depth in the soil. The inclusion of the organic C to nitrate ratio (Tiedje, 1988) as a possible factor controlling DNRA did not alter the outcome of the regression model.

4.

Discussion

The total nitrogen and nitrate contents of the floodplain are typical of those found in lowland soils subject to high agricultural intensity, despite the fact that N applications to the floodplain ceased following the restoration project, eleven years prior to this study. Mean total nitrogen content exceeded values typical of improved grassland and permanent fertilised pasture in the UK (Cardenas et al., 2010), whilst being comparable to

Fig. 4 e Soil physical and chemical properties in the subsurface of the BZ (full circles) and HM (open circles) floodplain management zones of the River Cole: (a) bulk density, (b) nitrate content, (c) % organic carbon. Data are means ± 1 SE, n [ 8.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 0 9 e4 9 2 2

Fig. 5 e Mean nitrogen transformation rates in the subsurface of the BZ (full circles) and HM (open circles) floodplain management zones: (a) Denitrification capacity, (b) DNRA capacity, (c) Ratio of denitrification versus DNRA capacity. Data are means ± 1 SE, n [ 8.

4917

chronically-loaded riparian zones across Europe (Hefting et al., 2006). Nitrogen enrichment of the floodplain is most likely to be caused by regular flooding of soils by nitrate-rich stream water (with an annual mean of 429  28 mmol NO3-N L
1) characteristic of lowland catchments in the UK (Jarvie et al., 2010), with the concurrent deposition of N-rich sediments (Pinay et al., 2007). Fertiliser inputs from surrounding land, and the additional organic-N inputs from grazing ungulates are likely to contribute to the significantly higher TN and nitrate concentrations found in the BZ and GG, respectively, in comparison to the other land management areas (FM & HM). Denitrification capacity across the floodplain was in the upper range of reported values for Europe (Well et al., 2005), and comparable to rates reported for floodplains with a similar soil texture in lowland England (Burt et al., 1999). Soil texture, in terms of silt þ clay content across the floodplain exceeded the 60% threshold at which Pinay et al. (2007) measured the highest denitrification in a pan-European study. The River Cole has a temperate, maritime flow regime driven by seasonal variations in rainfall and evaporation resulting in a short periodicity of floodedrainage cycles within the reconnected floodplain (Heppell unpub.). Hefting et al. (2004) have shown that such a magnitude of flood frequency can enhance nitrogen removal efficiency, whilst high nitrogen loading rates from floodwater maintains nitrogen availability. Comparisons of nitrate concentrations in the River Cole (ranging from 190 to 525 mmol NO3-N L
1) and in the soil porewater of the A and B horizons (ranging from 35 to 150 mmol NO3-N L
1) during the study also suggest that the reconnected floodplain may benefit river and soil water quality due to the associated high rates of denitrification. The multiple linear regression analysis suggested that variations in the quantity and lability of organic carbon across the floodplain surface are the main factors controlling denitrification capacity, and the inclusion of soil nitrate concentration did not improve the predictive power of the model (Table 3). In effect nitrate was in excess throughout the floodplain and the availability and quality of organic carbon as electron donors regulated denitrification, in keeping with other studies (Ambus and Lowrance, 1991; Hedin et al., 1998; Hill et al., 2000; Ostrom et al., 2002; Brettar and Ho¨fle, 2002; Vidon and Hill, 2004; Hefting et al., 2006). Our findings highlight the important role that carbon lability has to play in controlling denitrification in a nitrogen-rich floodplain, with fairly homogeneous organic carbon content. Accordingly, denitrification capacity was highest in the extensive, unfertilised GG, where the organic carbon was most labile. In the other zones, all of which are mowed on an annual basis, organic carbon in the topsoil was less than half as labile as in the GG. Two topsoil samples in the GG were significant outliers (3e5 times >SD from mean) in terms of their nitrate content and carbon lability, and the chemistry of these samples may be explained by hotspots caused by patches of urine or manure. However, the overall results of the regression are not significantly influenced by these outliers, which may indicate a wider spatial effect of grazing on lability and denitrification in the topsoil. Whilst differences in carbon availability in soils through management regime (grazing and/or mowing) can give rise to variations in genetic structure of nitrate reducing bacterial

4918

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 0 9 e4 9 2 2

Fig. 6 e Mean carbon transformation rates in the subsurface of the BZ (full circles) and HM (open circles) floodplain management zones: (a) Production of carbon dioxide (b) Lability of organic carbon. Data are means ± 1 SE, n [ 8 apart from (a) (HM depths of 100 and 120 cm), where outliers >10 times the standard deviation have been removed.

populations (Williams et al., 2000), the linkages between diversity and activity of the microbial communities are not fully understood. Elevated ammonium concentrations due to grazing activities are also a significant factor known to affect microbial community diversity in grassland soils, although these changes in diversity have not yet been associated with concurrent changes in activity (Patra et al., 2006). However,

our regression analysis suggested that elevated ammonium and nitrate concentrations in the GG were not a significant factor responsible for the enhanced denitrification activity in this zone. Alternatively, it is possible that inter-zone differences in plant species (resulting from grazing activity) are responsible for the observed variation in labile organic carbon. Patra et al. (2006) demonstrated that plant species have a significant influence on denitrification activity in grazed grassland soils, which could be explained by quantitative and qualitative differences in root exudates and plant litter inputs beneath different plant species, amongst other factors (Steltzer and Bowman, 1998; Burgmann et al., 2005). Drying and wetting cycles, arising from both rainfall events and flooding, can alter the response of microbial processes (such as denitrification) in a soil not only due to changes in soil nutrient availability but also due to changes in microbial communities potentially arising from physiological stress (Fierer and Schimel, 2002, 2003; Fromin et al., 2010). The topsoil of the re-connected River Cole floodplain is subjected to frequent drying and wetting cycles caused by both rainfall and inundation by floodwater. The GG, BZ and HM samples were all collected during a particularly prolonged period (over 8 months) of regular flooding. From April to July the periodicity of flooding was approximately monthly, and each flood event was of 4/5 days duration. Samples from all three land management zones were exposed to a similar hydrological regime, so inter-site differences in periodicity of wetting and drying is unlikely to be responsible for the observed differences in denitrification capacity in this instance. The samples from the FM were collected in July 2006 under drier soil conditions compared to the samples obtained from the other three sites in June/July 2007. It is possible that differences in moisture content could account for the reduced denitrification capacity in the FM compared with the other sites. However, Fromin et al. (2010) have found that denitrifier communities subjected to frequent dryewet fluctuations have a higher resilience to drought than those in more buffered situations. Furthermore, a separate experiment designed to compare denitrification capacity under ‘wet’ and ‘dry’ conditions in this floodplain (data not shown) showed no significant difference in denitrification capacity in thirty topsoil samples taken in near-saturated conditions following floodwater recession compared to thirty samples collected twelve days later under drier conditions (c. 10% difference in soil moisture content between wet and dry conditions).

Table 3 e Multiple Linear Regression Models to investigate the factors controlling denitrification capacity and DNRA in the floodplain. Independent variable Log DC topsoil Log DC depth Log DNRA topsoil Log DNRA depth

Equation
0.105 þ (0.278  log lab) þ (0.027  log % OC)
0.085 þ (0.825  log %OC) þ (0.183  log lab)
(0.619  log BD)
(0.073  log C:NO3)
0.234 þ (0.188  log NO3)
(0.264  log NO2) þ (0.095 log lab) þ (0.007  C:NO3) þ (0.002  % clay) 0.035 þ (0.087  log %OC)
(0.107  log BD)

r2

D- F ratio VIF n W

0.563 1.367 25.803 1.042 59 0.837 1.835 109.068 1.986 91 0.742 1.591 21.257 1.834 43 0.359 1.944 24.378 1.392 90

DC ¼ Denitrification capacity; D-W ¼ Durbin-Watson statistic; VIF ¼ Variance Inflation Factor; lab ¼ lability; OC ¼ organic carbon; BD ¼ bulk density.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 0 9 e4 9 2 2

In contrast to the pronounced variation in lability in the topsoil across the floodplain areas, lability remained fairly constant with depth in the BZ, but increased in the C horizon of the HM which comprises Oxford clays characterised by organic deposits of a marine origin (Fig. 6b). The amount of organic carbon and its lability did not co-vary with depth in these soils so the effects of these potential controls on anaerobic mineralisation and denitrification (both processes mediated by anaerobic bacterial communities) are decoupled. Our results indicate that the exponential decay in denitrification capacity and DNRA with increasing depth is controlled primarily by organic carbon content. A similar exponential decay of denitrification with depth has been reported in a number of other floodplain studies (Burt et al., 1999; Cle´ment et al., 2002; Hill et al., 2004), whereas we are aware of only one other study in a temperate riparian setting which has considered variation in DNRA with depth (Davis et al., 2008). We propose that the exponential decline in organic carbon content with depth results in a decrease in bacterial biomass in the soil. Dauwe et al. (2001) suggested that where sediment CO2 production was below a critical limit of 0.1 mmol CO2 g wet soil h
1, then, regardless of the lability of the organic matter, anaerobic populations and their efficiencies declined. Such rates were characteristic of our soil column below 30e50 cm and, nitrate reduction would have been restricted. In contrast to denitrification, DNRA was not significantly different in the topsoil of the various land management zones. These values fall in the lower range of rates reported in the surface of tropical N-limited forest soils (Silver et al., 2001). Denitrification and DNRA were significantly positively correlated (r ¼ 0.57, n ¼ 59, P < 0.01), suggesting that both processes were being regulated by similar environmental conditions and reactants (Tiedje, 1988). In the topsoil, DNRA was primarily controlled by organic carbon content, with soil texture, although it should be noted that the relatively low explanatory power of the regression models for DNRA suggest that other unexplained factors also come into play. DNRA has been shown to be mediated by obligate anaerobes (e.g. Clostridium spp.), facultative anaerobes (e.g. Escherichia coli) and some aerobes (e.g. Bacillus spp.) (Tiedje, 1988; Cole, 1996). In addition, a good deal of evidence suggests that it is favoured by reduced redox and both a plentiful supply organic carbon and the relative lability or quality of that organic carbon; though the true microbiology and regulatory dynamics of DNRA in soils remains uncertain (Tiedje, 1988; Yin et al., 2002). A greater clay content of soil will most likely generate low oxygen conditions within the immobile regions of fine textured soils, through which diffusion of oxygen is very slow and such conditions may suit a more obligate anaerobic physiology of any DNRA community in floodplain topsoils (Ambus et al., 1992; Scott et al., 2008; Sotta et al., 2008). In tropical soils, DNRA is reputed to be adapted to fluctuating redox regimes and also to be able to withstand unfavourable periods of higher oxygen concentration (PettRidge and Firestone, 2005; Silver et al., 2005; Pett-Ridge et al., 2006). Our results indicate that bacteria capable of DNRA also survive, and to a limited extent compete for resources with denitrifying bacteria, in temperate grassland topsoils which are subjected to frequent wetting and drying cycles, with concurrent fluctuating oxygen conditions.

5.

4919

Conclusions

Our results show that the re-connection of the floodplain of the River Cole could contribute to water quality improvements with respect to nitrate, due to the potential for significant heterotrophic denitrification across the entire floodplain, irrespective of land management practice. In this instance, and as demonstrated by other studies, enhancing overbank flooding through restoration has created the necessary conditions for denitrification to occur. The key factor controlling denitrification in an N-rich landscape such as this is organic carbon. Furthermore our findings indicate that an important secondary control on denitrification in such landscapes is the lability of the organic carbon. Spatial heterogeneity in denitrification across the floodplain topsoil was explained by differences in the availability of organic carbon arising from alternative management practices. Accordingly, in agricultural catchments post-restoration management practices, such as extensive grazing on an unimproved floodplain, that enhance both the quantity and the availability of the organic carbon, could be employed to enhance the potential for nitrate removal from floodwater across the floodplain, and should be the subject of further research. Grazing, however, should be accompanied by fencing to prevent the poaching of the river banks by cattle and the concurrent introduction of nitrogen to the river. Finally, this research demonstrates that although the potential for DNRA exists in temperate N-rich riparian soils, it operates at significantly lower rates than denitrification, and is associated with fine textured soils subject to frequent saturation which would give rise to fluctuating oxygen conditions. Therefore, the majority of nitrate in floodwaters infiltrating the soil is reduced to nitrogen gas and is removed rather than conserved in the system as NHþ 4 (which could easily re-enter the soil N cycle via plant uptake and subsequent die-back).

Acknowledgements The authors are grateful to the UK River Restoration Centre and the National Trust at Coleshill for their guidance and advice and to the tenant farmer for permitting the installation of research equipment at the River Cole floodplain. This research was funded by a joint Ph.D. studentship to Fotis Sgouridis by Queen Mary University of London and the Alexander S. Onassis Public Benefit Foundation (Greece).

references

Acreman, M.C., Fisher, J., Stratford, C.J., Mould, D.J., Mountford, J.O., 2007. Hydrological science and wetland restoration: some case studies from Europe. Hydrol. Earth Syst. Sci. 11 (1), 158e169. Ambus, P., Lowrance, R., 1991. Comparison of denitrification in two riparian soils. Soil Sci. Soc. Am. J. 55, 994e997. Ambus, P., Mosier, A.R., Christensen, S., 1992. Nitrogen turnover rates in a riparian fen determined by 15N dilution. Biol. Fertil. Soils 14, 230e236.

4920

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 0 9 e4 9 2 2

Baker, M.A., Vervier, P., 2004. Hydrological variability, organic matter supply and denitrification in the Garonne River ecosystem. Freshw. Biol. 49, 181e190. Bijay, S., Ryden, J.C., Whitehead, D.C., 1988. Some relationships between denitrification potential and fractions of organic carbon in air-dried and field-moist soils. Soil Biol. Biochem. 20 (5), 737e741. Blackwell, M.S.A., Maltby, E., 2005. Ecoflood Guidelines: How to Use Floodplains for Flood Risk Reduction. EVK1-CT-2002e80017. Bowman, R.A., Reeder, J.D., Lober, R.W., 1990. Changes in soil properties in a central plains rangeland soil after 3, 20, and 60 years of cultivation. Soil Sci. 150, 851e857. Boyer, J.N., Groffman, P.M., 1996. Bioavailability of water extractable organic carbon fractions in forest and agricultural soil profiles. Soil Biol. Biochem. 28 (6), 783e790. Brettar, I., Ho¨fle, M.G., 2002. Close correlation between the nitrate elimination rate by denitrification and the organic matter content in hardwood forest soils of the Upper Rhine floodplain (France). Wetlands 22 (2), 214e224. Brettar, I., Sanchez-Perez, J.M., Tre´molie`res, M., 2002. Nitrate elimination by denitrification in hardwood forest soils of the Upper Rhine floodplain e correlation with redox potential and organic matter. Hydrobiologia 469 (1), 11e21. Burgmann, H., Meier, S., Bunge, M., Widmer, F., Zeyer, J., 2005. Effects of model root exudates on structure and activity of a soil diazotroph community. Environ. Microbiol. 7, 1711e1724. Burt, T.P., Pinay, G., Sabater, S., 2010. What we still need to know about the ecohydrology of riparian zones? Ecohydrology 3, 373e377. Burt, T.P., Matchett, L.S., Goulding, K.W.T., Webster, C.P., Haycock, N.E., 1999. Denitrification in riparian buffer zones: the role of floodplain hydrology. Hydrol. Process. 13, 1451e1463. Cardenas, L.M., Thorman, R., Ashlee, N., Butler, M., Chadwick, D., Chambers, B., Cuttle, S., Donovan, N., Kingston, H., Lane, S., Dhanoa, M.S., Scholefield, D., 2010. Quantifying annual N2O emission fluxes from grazed grassland under a range of inorganic fertiliser nitrogen inputs. Agric. Ecosyst. Environ. 136, 218e226. Cavigelli, M.A., Robertson, G.P., 2000. The functional significance of denitrifier community composition in a terrestrial ecosystem. Ecology 81 (5), 1402e1414. Cle´ment, J.C., Pinay, G., Marmonier, P., 2002. Seasonal dynamics of denitrification along topohydrosequences in three different riparian wetlands. J. Environ. Qual. 31, 1025e1037. Cole, J.A., 1996. Nitrate reduction to ammonia by enteric bacteria: redundancy, or a strategy for survival during oxygen starvation? FEMS Microbiol. Lett. 136, 1e11. Dauwe, B., Middelburg, J.J., Herman, P.M.J., 2001. Effect of oxygen on the degradability of organic matter in subtidal and intertidal sediments of the North Sea area. Mar. Ecol. Prog. Ser. 215, 13e22. Davis, J.H., Griffith, S.M., Horwath, W.R., Steiner, J.J., Myrold, D.D., 2008. Denitrification and nitrate consumption in an herbaceous riparian area and perennial ryegrass seed cropping system. Soil Sci. Soc. Am. J. 72 (5), 1299e1310. Dodla, S.K., Wang, J.J., DeLaune, R.D., Cook, R.L., 2008. Denitrification potential and its relation to organic carbon quality in three coastal wetland soils. Sci. Total Environ. 407, 471e480. Drury, C.F., Oloya, T.O., McKenney, D.J., Gregorich, E.G., Tan, C.S., van Luyk, C.L., 1998. Long-term effects of fertilization and rotation on denitrification and soil carbon. Soil Sci. Soc. Am. J. 62, 1572e1579. Fierer, N., Schimel, J.P., 2002. Effects of drying-wetting frequency on soil carbon and nitrogen transformations. Soil Biol. Biochem. 34, 777e787. Fierer, N., Schimel, J.P., Holden, P.A., 2003. Influence of dryingrewetting frequency on soil bacterial community structure. Microb. Ecol. 45, 63e71.

Forshay, K., Stanley, E., 2005. Rapid nitrate loss and denitrification in a temperate river floodplain. Biogeochemistry 75 (1), 43e64. Frank, D.A., Groffman, P.M., 1998. Denitrification in a semi-arid grazing ecosystem. Oecologia 117, 564e569. Fromin, N., Pinay, G., Montuelle, B., Landais, J.M., Ourcival, J.M., Joffre, R., Lensi, R., 2010. Impact of seasonal sediment desiccation and rewetting on microbial processes involved in greenhouse gas emissions. Ecohydrology 3, 339e348. Gale, P.M., Reddy, K.R., Graetz, D.A., 1992. Mineralization of sediment organic-matter under anoxic conditions. J. Environ. Qual. 21 (3), 394e400. Gavrichkova, O., Moscatelli, M.C., Grego, S., Valentini, R., 2008. Soil carbon mineralization in a Mediterranean pasture: effect of grazing and mowing management practices. Agrochimica 52 (5), 285e296. Groffman, P.M., Gold, A.J., Simmons, R.C., 1992. Nitrate dynamics in riparian forests: microbial studies. J. Environ. Qual. 21, 666e671. Groffman, P.M., Rice, C.W., Tiedje, J.M., 1993. Denitrification in a tallgrass Prairie landscape. Ecology 74 (3), 855e862. Groffman, P.M., Altabet, M.A., Bohlke, J.K., Butterbach-Bahl, K., David, M.B., Firestone, M.K., Giblin, A.E., Kana, T.M., Nielsen, L. P., Voytek, M.A., 2006. Methods for measuring denitrification: diverse approaches to a difficult problem. Ecol. Appl. 16 (6), 2091e2122. Hauck, R.D., Melsted, S.W., Yankwich, P.E., 1958. Use of N- isotope distribution in nitrogen gas in the study of denitrification. Soil Sci. 86, 287e291. Hedges, J.I., Stern, J.H., 1984. Carbon and nitrogen determinations of carbonate-containing solids. Limnol. Oceanogr. 29 (3), 657e663. Hedin, L.O., von Fisher, J.C., Ostrom, N.E., Kennedy, B.P., Brown, M.G., Robertson, G.P., 1998. Thermodynamic constraints on nitrogen transformations and other biogeochemical processes at soil-stream interfaces. Ecology 79 (2), 684e703. Hefting, M.M., Bobbink, R., De Caluwe, H., 2003. Nitrous oxide emission and denitrification in chronically nitrate-loaded riparian buffer zones. J. Environ. Qual. 32, 1194e1203. Hefting, M.M., Cle´ment, J.C., Dowrick, D.J., Cosandey, A.-C., Bernal, S., Cimpian, C., Tatur, A., Burt, T.P., Pinay, G., 2004. Water table elevation controls on soil nitrogen cycling in riparian wetlands along a European climatic gradient. Biogeochemistry 67, 113e134. Hefting, M.M., Bobbink, R., Janssens, M.P., 2006. Spatial variation in denitrification and N2O emission in relation to nitrate removal efficiency in a N-stressed riparian buffer zone. Ecosystems 9, 550e563. Heiri, O., Lotter, A.F., Lemcke, G., 2001. Loss on ignition as a method for estimating organic and carbonate content in sediments: reproducibility and comparability of results. J. Paleolimn. 25, 101e110. Hill, A.R., Devito, K.J., Campagnolo, S., Sanmugadas, K., 2000. Subsurface denitrification in a forest riparian zone: interactions between hydrology and supplies of nitrate and organic carbon. Biogeochemistry 51 (2), 193e223. Hill, A.R., Cardaci, M., 2004. Denitrification and organic carbon availability in Riparian wetland soils and subsurface sediments. Soil Sci. Soc. Am. J. 68 (1), 320e325. Hill, A.R., Vidon, P.G.F., Langat, J., 2004. Denitrification potential in relation to lithology in five headwater Riparian zones. J. Environ. Qual. 33 (3), 911e919. Hoffmann, C.C., Pedersen, M.L., Kronvang, B., Ovig, L., 1998. Restoration of the rivers Brede, Cole and Skerne: a joint Danish and British EU-LIFE demonstration project, IV d Implications for nitrate and iron transformation. Aquat. Conserv. 8, 223e240. Hoffmann, C.C., Baattrup-Pedersen, A., 2007. Re-establishing freshwater wetlands in Denmark. Ecol. Eng. 30, 157e166.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 0 9 e4 9 2 2

Holmes, N.T.H., Nielsen, M.B., 1998. Restoration of the rivers Brede, Cole and Skerne: a joint Danish and British EU-LIFE demonstration project, I e Setting up and delivery of the project. Aquat. Conserv. 8, 185e196. Huygens, D., Ru¨tting, T., Boeckx, P., Cleemput, O.V., Godoy, R., Muller, C., 2007. Soil nitrogen conservation mechanisms in a pristine south Chilean Nothofagus forest ecosystem. Soil Biol. Biochem. 39, 2448e2458. Ilmarinen, K., Mikola, J., Nissinen, K., Vestberg, M., 2009. Role of soil organisms in the maintenance of species-rich seminatural grasslands through mowing. Restor. Ecol. 17 (1), 78e88. Janes, M., Holmes, N., Haycock, N., 1999. Manual of River Restoration Techniques. The River Restoration Centre, Silsoe, UK. Jarvie, H., Wither, P.J.A., Bowes, M.J., Palmer-Felgate, E.J., Harper, D.M., Wasiak, K., Wasiak, P., Hodgkinson, R.A., Bates, A., Stoate, C., Neal, M., Wickham, H.D., Harman, S.A., Armstrong, L.K., 2010. Streamwater phosphorus and nitrogen across a gradient in rural-agricultural land use intensity. Agric. Ecosyst. Environ. 135 (4), 238e252. Kirkwood, D.S., 1996. Nutrients: Practical Notes on their Determination in Seawater. ICES, Copenhagen, Denmark. Krug, E.C., Winstanley, D., 2002. The need for comprehensive and consistent treatment of the nitrogen cycle in nitrogen cycling and mass balance studies: 1. Terrestrial nitrogen cycle. Sci. Total Environ. 293, 1e29. Linn, D.M., Doran, J.W., 1984. Effect of water-filled pore space on carbon dioxide and nitrous oxide production in tilled and nontilled soils. Soil Sci. Soc. Am. J. 48, 1267e1272. Luo, J., Tillman, R.W., Ball, P.R., 1999. Grazing effect on denitrification in a soil under pasture during two contrasting seasons. Soil Biol. Biochem. 31, 903e912. Matheson, F.E., Nguyen, M.L., Cooper, A.B., Burt, T.P., 2003. Shortterm nitrogen transformation rates in riparian wetland soil determined with nitrogen-15. Biol. Fertil. Soils 38 (3), 129e136. Meneer, J.C., Ledgard, S., McLay, C., Silvester, W., 2005. Animal treading stimulates denitrification in soil under pasture. Soil Biol. Biochem. 37, 1625e1629. Myrold, D.D., Tiedje, J.M., 1985. Establishment of denitrification capacity in soil: effects of carbon, nitrate and moisture. Soil Soil Biol. Biochem. 17 (6), 819e822. Nielsen, L.P., 1992. Denitrification in sediment determined from nitrogen isotope pairing. FEMS Microbiol. Ecol. 86, 357e362. Olde Venterink, H., Hummelink, E., Van Den Hoorn, M.W., 2003. Denitrification potential of a river floodplain during flooding with nitrate-rich water: grasslands versus reedbeds. Biogeochemistry 65, 233e244. Orr, C.H., Stanley, E.H., Wilson, K.A., Finlay, J.C., 2007. Effects of restoration and reflooding on soil denitrification in a leveed midwestern floodplain. Ecol. Appl. 17 (8), 2365e2376. Ostrom, N.E., Hedin, L.O., von Fisher, J.C., Robertson, G.P., 2002. Nitrogen transformations and NO
3 removal at a soil-stream interface: a stable isotope approach. Ecol. Appl. 12 (4), 1027e1043. Patra, A.K., Clays-Josserand, A., Degrange, V., Grayston, S.J., Guillaumaud, N., Loiseau, P., 2005. Effects of grazing on microbial functional groups involved in soil N dynamics. Ecol. Monogr. 75, 65e80. Patra, A.K., Abbadie, L., Clays-Josserand, A., Degrange, V., Grayston, S.J., Guillaumaud, N., Loiseau, P., Louault, P., Mahmood, S., Nazaret, S., Philippot, L., Poly, F., Prosser, J.I., Le Roux, X., 2006. Effects of management regime and plant species on the enzyme activity and genetic structure of Nfixing, denitrifying and nitrifying bacterial communities in grassland soils. Environ. Microbiol. 8 (6), 1005e1016. Peterjohn, W.T., Correll, D.L., 1984. Nutrient dynamics in an agricultural watershed: observations on the role of a riparian forest. Ecology 65, 1466e1475.

4921

Pett-Ridge, J., Firestone, M.K., 2005. Redox fluctuation structures microbial communities in a wet tropical soil. Appl. Environ. Microbiol. 71 (11), 6998e7007. Pett-Ridge, J., Silver, W.L., Firestone, M.K., 2006. Redox fluctuations frame microbial community impacts on N-cycling rates in a humid tropical forest soil. Biogeochemistry 81, 95e110. Philippot, L., Cuhel, J., Saby, N.P.A., Cheneby, D., Chronakova, A. , Bru, D., Arrouays, D., Martin-Laurent, F., Simek, M., 2009. Mapping field-scale spatial patterns of size and activity of the denitrifier community. Environ. Microbiol. 11 (6), 1518e1526. Pinay, G., Roques, L., Fabre, A., 1993. Spatial and temporal patterns of denitrification in a Riparian Forest. J. Appl. Ecol. 30 (4), 581e591. Pinay, G., Gumiero, B., Tabacchi, E., Gimenez, O., Tabacchi-Planty, A. M., Hefting, M.M., Burt, T.P., Black, V.A., Nilsson, C., Iordache, V., Bureau, F., Vought, L., Petts, G.E., De´camps, H., 2007. Patterns of denitrification rates in European alluvial soils under various hydrological regimes. Freshw. Biol. 52 (2), 252e266. Risgaard-Petersen, N., Rysgaard, S., Revsbech, N.P., 1995. Combined microdiffusion-hypobromite oxidation method for determining Nitrogen-15 isotope in ammonium. Soil Sci. Soc. Am. J. 59, 1077e1080. Robson, T.M., Lavorel, S., Clement, J.C., Le Roux, X., 2007. Neglect of mowing and manuring leads to slower nitrogen cycling in subalpine grasslands. Soil Biol. Biochem. 39, 930e941. Rowell, D.L., 1994. Soil Science: Methods and Applications. Longman Scientific & Technical. Ru¨tting, T., Huygens, D., Muller, C., Van Cleemput, O., Godoy, R., Boeckx, P., 2008. Functional role of DNRA and nitrite reduction in a pristine south Chilean Nothofagus forest. Biogeochemistry 90, 243e258. Rysgaard, S., Risgaard-Petersen, N., 1997. A sensitive method for determining nitrogen-15 isotope in urea. Mar. Biol. 128 (2), 191e195. Scott, J.T., McCarthy, M.J., Gardner, W.S., Doyle, R.D., 2008. Denitrification, dissimilatory nitrate reduction to ammonium, and nitrogen fixation along a nitrate concentration gradient in a created freshwater wetland. Biogeochemistry 87, 99e111. Sear, D.A., White, R.A., 1994. The River Restoration Project: A geomorphological survey of the River Cole, Wiltshire. Department of Geography, University of Southampton. Sheibley, R.W., Ahearn, D.S., Dahlgren, R.A., 2006. Nitrate loss from a restored floodplain in the lower Cosumnes River, California. Hydrobiologia 571, 261e272. Silver, W.L., Herman, D.J., Firestone, M.K., 2001. Dissimilatory nitrate reduction to ammonium in upland tropical forest soils. Ecology 82 (9), 2410e2416. Silver, W.L., Thompson, A.W., Reich, A., Ewel, J.J., Firestone, M.K., 2005. Nitrogen cycling in tropical plantation forests: potential controls on nitrogen retention. Ecol. Appl. 15 (5), 1604e1614. Simek, M., Cooper, J.E., Picek, T., Santruckova, H., 2000. Denitrification in arable soils in relation to their physicochemical properties and fertilisation practice. Soil Biol. Biochem. 32, 101e110. Smart, R.M., Barko, J.W., 1985. Laboratory culture of submersed freshwater macrophytes on natural sediments. Aquat. Bot. 21, 251e263. Smith, M.S., Tiedje, J.M., 1979. Phases of denitrification following oxygen depletion in soil. Soil Biol. Biochem. 11, 261e267. Sotta, E.D., Corre, M.D., Veldkamp, E., 2008. Differing N status and N retention processes of soils under old-growth lowland forest in eastern Amazonia, Caxiuana, Brazil. Soil Biol. Biochem. 40, 740e750. Steltzer, H., Bowman, W.D., 1998. Differential influence of plant species on soil nitrogen transformations within moist meadow Alpine tundra. Ecosystems 1, 464e474.

4922

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 0 9 e4 9 2 2

Templer, P.H., Silver, W.L., Pett-Ridge, J., DeAngelis, K.M., Firestone, M.K., 2008. Plant and microbial controls on nitrogen retention and loss in a humid tropical forest. Ecology 89 (11), 3030e3040. Thamdrup, B., Dalsgaard, T., 2000. The fate of ammonium in anoxic manganese oxide-rich marine sediment. Geochim. Cosmochim. Acta 64 (24), 4157e4164. Tiedje, J.M., 1988. Ecology of denitrification and dissimilatory nitrate reduction to ammonium. In: Zehnder, A.J.B. (Ed.), Biology of Anaerobic Microorganisms. Wiley, New York, pp. 179e244. Tockner, K., Pennetzdorfer, D., Reiner, N., Schiemer, F., Ward, J.V., 1999. Hydrological connectivity, and the exchange of organic matter and nutrients in a dynamic river-floodplain system (Danube, Austria). Freshw. Biol. 41 (3), 521e535. Tockner, K., Stanford, J.A., 2002. Riverine floodplains: present state and future trends. Environ. Conserv. 29 (3), 308e330. Ullah, S., Faulkner, S.P., 2006. Denitrification potential of different land-use types in an agricultural watershed, lower Mississippi valley. Ecol. Eng. 28, 131e140. Van Der Lee, G.E.M., Venterink, H.O., Asselman, N.E.M., 2004. Nutrient retention in floodplains of the Rhine distributaries in the Netherlands. River Res. Appl. 20, 315e325.

Vidon, P., Hill, A.R., 2004. Denitrification and patterns of electron donors and acceptors in eight riparian zones with contrasting hydrogeology. Biogeochemistry 71 (2), 259e283. Well, R., Hoper, H., Mehranfar, O., Meyer, K., 2005. Denitrification in the saturated zone of hydromorphic soils e laboratory measurement, regulating factors and stochastic modeling. Soil Biol. Biochem. 37, 1822e1836. Williams, B.L., Grayston, S.J., Raid, E.J., 2000. Influence of synthetic sheep urine on the microbial biomass, activity and community structure in two pastures in the Scottish uplands. Plant Soil 225, 175e185. Woodward, K.B., Fellows, C.S., Conway, C.L., Hunter, H.M., 2009. Nitrate removal, denitrification and nitrous oxide production in the riparian zone of an ephemeral stream. Soil Biol. Biochem. 41, 671e680. Yeomans, J.C., Bremner, J.M., McCarty, G.W., 1992. Denitrification capacity and denitrification potential of subsurface soils. Commun. Soil Sci. Plant Anal. 23 (9&10), 919e927. Yin, S.X., Chen, D., Chen, L.M., Edis, R., 2002. Dissimilatory nitrate reduction to ammonium and responsible microorganisms in two Chinese and Australian paddy soils. Soil Biol. Biochem. 34, 1131e1137.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 2 3 e4 9 2 9

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Operation of an aquatic worm reactor suitable for sludge reduction at large scale Tim L.G. Hendrickx a,b,*, Hellen H.J. Elissen a, Hardy Temmink a,b, Cees J.N. Buisman a,b a b

Wetsus, Centre of Excellence for Sustainable Water Technology, P.O. Box 1113, 8900 CC Leeuwarden, The Netherlands Sub-department of Environmental Technology, Wageningen University, P.O. Box 17, 6700 AA Wageningen, The Netherlands

article info

abstract

Article history:

Treatment of domestic waste water results in the production of waste sludge, which

Received 1 December 2010

requires costly further processing. A biological method to reduce the amount of waste

Received in revised form

sludge and its volume is treatment in an aquatic worm reactor. The potential of such

21 April 2011

a worm reactor with the oligochaete Lumbriculus variegatus has been shown at small scale.

Accepted 26 June 2011

For scaling up purposes, a new configuration of the reactor was designed, in which the

Available online 2 July 2011

worms were positioned horizontally in the carrier material. This was tested in a continuous experiment of 8 weeks where it treated all the waste sludge from a lab-scale activated

Keywords:

sludge process. The results showed a higher worm growth rate compared to previous

Aquatic worm reactor

experiments with the old configuration, whilst nutrient release was similar. The new

Lumbriculus variegatus

configuration has a low footprint and allows for easy aeration and faeces collection,

Scaling up

thereby making it suitable for full scale application.

Continuous operation

ª 2011 Elsevier Ltd. All rights reserved.

Waste sludge reduction

1.

Introduction

Excess sludge is an inevitable product of biological treatment of municipal waste waters. Historically, waste sludge was applied in agriculture, making best use of its organic content and nutrients. However, sludge also contains contaminants, such as heavy metals and organic micropollutants, restricting the beneficial use of sludge. As a result, incineration is becoming the main final treatment option for waste sludge (IWA, 2007). The complete sludge processing chain involves transport of large volumes, which could be minimised by decreasing the amount of waste sludge or improving sludge thickening processes. The latter may, for example, be achieved by conditioning additives which enhance water removal (Dentel, 2001) or freeze/thawing to release bound water (Tuan and Sillanpa¨a¨, 2010). However, these require a substantial

input of chemicals and/or energy. Sludge reduction can be achieved by available disrupting techniques such as ozonation, high pressure homogenisation, thermal treatment and ultrasonic treatment (Foladori et al., 2010), where the lysis products are returned to the treatment plant (or anaerobically digested when available on-site). These techniques require a substantial input of energy and are therefore not sustainable. A biological approach is the use of aquatic worms (aquatic Oligochaeta), which has achieved considerable attention in recent years. Several researchers have focussed on separate worm reactors in which sludge reduction is achieved with free-swimming worms (Song and Chen, 2009; Guo et al., 2007), sessile worms (Tian and Lu, 2010; Huang et al., 2007; Elissen et al., 2006), or a combination of free and sessile worms (Wei and Liu, 2005). It was shown to be feasible to operate a worm reactor continuously with the sessile

* Corresponding author. Wetsus, Centre of Excellence for Sustainable Water Technology, P.O. Box 1113, 8900 CC Leeuwarden, The Netherlands. Tel.: þ31 582 843 000; fax: þ31 582 843 001. E-mail address: [email protected] (T.L.G. Hendrickx). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.06.031

4924

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 2 3 e4 9 2 9

aquatic worm Lumbriculus variegatus immobilised in a carrier material as shown in Fig. 1 (Hendrickx et al., 2009a). Recently, we have shown that next to sludge reduction, the main benefit of such an aquatic worm reactor is the improved dewaterability of the separately collected worm faeces (Hendrickx et al., 2010b). Both contribute to a decrease in the transportation costs in the sludge processing chain. This makes it particularly interesting for plants with relatively high sludge transportation costs, i.e. smaller waste water treatment plants (WWTPs) without sludge processing facilities. The main challenge is now to design a worm reactor that is suitable for application at full scale. Since the aquatic worms are immobilised on a surface area (the carrier material), the size of a reactor depends on the amount of area that can be installed per reactor volume. To optimize this, a new configuration was designed in which the worms are positioned horizontally in the carrier material. This is also closer to their orientation in the natural environment (where heads are buried into sediments and tails protrude upwards) when compared to their inverted positioning used in the original configuration (Elissen et al., 2006). This paper presents the experiments that were performed with the new and more compact configuration of the aquatic worm reactor. In a continuous experiment of 8 weeks all the waste sludge from a lab-scale activated sludge process was treated with this aquatic worm reactor. Sludge reduction, worm growth and nutrients release are evaluated and the implications of such a worm reactor at a WWTP are discussed.

Fig. 2 e Set-up for the experiment with the large continuous worm reactor. material (SEFAR Nitex) was 350 mm. Fresh sludge was added every 1e3 days and was circulated over the mesh cylinder. The effluent in the water compartment was aerated using a fine bubble diffuser (to maintain DO > 8 mg/L) and was replaced once a week. Worms were removed from the mesh once a week to weigh them. Worms fallen from the carrier into the water compartment were not accounted for in this experiment.

2.1.2.

2.

Materials and methods

2.1.

Experimental set-up

2.1.1.

Feasibility experiment

A feasibility experiment with the new reactor configuration was performed with activated sludge and effluent from the Leeuwarden municipal WWTP (The Netherlands). Sludge was first sieved (1 mm mesh) and effluent was filtered over black ribbon filters (12e25 mm, Schleicher and Schuell) before being used in the experiments. A small version of the reactor shown in Fig. 2 was used, with only one mesh cylinder with a diameter of 4 cm and a height of 30 cm. The mesh size of the carrier

Lab-scale activated sludge system

Activated sludge and effluent from a lab-scale activated sludge system were used in the experiments. This system treated pre-settled domestic sewage in a completely mixed aeration tank (50 L) followed by a settler. Waste water was frequently analysed for total Chemical Oxygen Demand (COD), total nitrogen and total phosphorus. Every other day, sludge was analysed for total and volatile suspended solids and effluent was analysed for total COD, ammonia, nitrate and phosphate. The system was operated at a sludge retention time (SRT) of 18 days by wasting a fixed volume of sludge directly from the aeration tank. The waste water flow rate was 78 L/d with an average total COD of 408 mg/L, which resulted in an average organic loading rate of 0.16 g COD/(g TSS$d). The average TSS and VSS were 4.0 and 2.9 g VSS/L, respectively.

Fig. 1 e Concept for aquatic worm reactor with worms immobilised in a carrier material and separation between sludge and faeces.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 2 3 e4 9 2 9

2.1.3.

Sequencing batch experiments

Sequencing batch experiments with a horizontal carrier material were performed as described in detail in Hendrickx et al. (2009a), using sludge and effluent from the lab-scale activated sludge system. Effluent was first filtered over black ribbon filters (12e25 mm, Schleicher and Schuell). Worms were counted and their wet weight (ww) was determined. A 350 mm polyamide mesh (SEFAR Nitex) with a surface area of 7.5 cm2 was used as carrier material. These experiments were required to accurately determine sludge consumption rate and the reduction percentage, which could not be done in the large continuous reactor.

2.1.4.

Large continuous worm reactor

Characteristics of the larger worm reactor are given in Table 1. Worms (29.8 g wet weight) were introduced in the worm reactor via the open top of the mesh cylinders. Waste sludge (2 L/d) from the activated sludge system was directly pumped to the inlet of the sludge compartment, i.e. the bottom of the mesh cylinders (Fig. 2). Effluent from the activated sludge system was collected in an overflowing bucket, from where it was pumped to the inlet of the water compartment. The effluent flow rate through the water compartment of the worm reactor was determined by the ammonia concentration and was decreased stepwise from 43 L/d to 2.8 L/d. Worm faeces were pumped from the bottom of the water compartment at a rate of 1.2 L/d. The water compartment was aerated using a diffuser (with an air flow rate of about 690 mL/min) inside a pipe. This visibly created mixing of the effluent in the water compartment, which distributed dissolved oxygen throughout the worm reactor, while the worm faeces still could settle. The outflow from the worm reactor was collected and analysed for ammonia, nitrate and phosphate. Sludge that was not consumed by the worms was not found in the worm outlet, but created a sludge bed inside the mesh cylinders. Collected worm faeces were analysed for TSS & VSS and its supernatant for ammonia, nitrate and phosphate. Waste sludge and worm faeces were occasionally analysed for total N and total P. At the end of the experimental run, all the worms in the mesh cylinders were collected and their wet weight was determined.

2.2.

Analyses

Wet weight (ww) of the worms was determined by placing the worms on a 150 mm polyamide mesh material (SEFAR Nitex). By pushing paper towelling against the back of the mesh, adhering water was removed from the worms. The ratio dry

Table 1 e Dimensions of the large worm reactor. Mesh size carrier material Mesh cylinders Surface area per cylinder Height mesh cylinder Diameter mesh cylinder Spacing between cylinders Volume sludge compartment Volume water compartment

mm # cm2 cm cm cm L L

350 3 1257 100 4 4 3.8 31

4925

weight/wet weight was previously determined at 0.13 (Elissen et al., 2006). The nitrogen and phosphorus content of the worms were 0.14 g N/g dry weight and 0.012 g P/g dry weight, respectively (Hendrickx et al., 2010b). Oxygen consumption by the worms was 4.9 mg O2/(g wet weight$d). TSS and VSS were determined according to Standard Methods (APHA, 1998) using black ribbon filters (12e25 mm, Schleicher and Schuell). COD, total N, total P and total ammonia (NHþ 4 þ NH3) were determined according to Standard Methods (APHA, 1998) using Dr Lange test kits. Nitrate 3 (NO 3 ) and phosphate (PO4 ) were determined according to Standard Methods (APHA, 1998) using ion chromatography (Metrohm 761 Compact IC). Temperature and dissolved oxygen (DO) concentration in the water compartment of the worm reactor were measured using an optical dissolved oxygen measurement probe (Oxymax W COS61, Endress and Hauser).

3.

Results and discussion

3.1.

Feasibility experiment

The feasibility experiment with the new worm reactor showed that worm growth took place in this configuration. In a period of 40 days, the worm biomass in the reactor increased from 9.8 to 18 g ww. The net worm growth rate (0.015 d1) was higher than found previously in a horizontal carrier (0.009e0.013 d1) and using the same sludge, though still below rates found for non-immobilised worms (0.026 d1) (Hendrickx et al., 2010a). As discussed previously, immobilising the worm in a mesh material could explain the lower growth rate compared to non-immobilised worms (Hendrickx et al., 2010a). The advantage of the current configuration was that the worms were positioned horizontally, which is closer their natural position (tail projecting upwards), which might be an explanation for the higher growth rate.

3.2.

Sequencing batch experiments

Sequencing batch experiments with a horizontally placed carrier material and with sludge (and effluent) from the labscale activated sludge system, showed a sludge consumption rate of 138 mg TSS/(g ww$d) when a worm density of 1.2 kg ww/m2 was used. TSS reduction in the batch experiments was 11% (16% based on VSS). The low reduction may be explained to the nutritional value of the sludge, caused by the low organic loading of the lab-scale activated sludge system. At the same time sludge consumption rate by the worms was high, which could have been a compensation for the low nutritional value. These results were used to estimate the size of a larger worm reactor that could treat the waste sludge from the lab-scale activated sludge system. This would require 58 g ww of worm biomass and a carrier surface area of 485 cm2. The large continuous worm reactor we used offered ample surface area (3770 cm2), thereby avoiding worm density to become a limiting factor for growth, as was shown previously (Hendrickx et al., 2010a). The larger reactor was started with a lower amount of worm biomass (29.8 g ww) to demonstrate that worm growth would be possible.

4926

3.3.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 2 3 e4 9 2 9

Larger continuous reactor

The larger continuous worm reactor was operated without any problems during the entire experimental period of nearly 8 weeks. The cumulative amounts of waste sludge fed to the worm reactor and collected worm faeces are shown in Fig. 3. In total 431 g TSS of waste sludge was fed to the worm reactor and 167 g TSS was collected as worm faeces. Sludge accumulation was observed as a sludge bed in the mesh cylinders, which was expected since the reactor was started with an insufficient amount of worms. The amount of sludge consumed by the worms was therefore estimated from the amount of collected worm faeces and a TSS reduction of 11%, as found in the batch experiments. This resulted in an estimated total sludge consumption of 187 g TSS and an estimated total sludge digestion by the worms of 20 g TSS. The estimated sludge consumption rate of 110 mg TSS/(g ww$d) during the last days of operation (based on faeces production rate and TSS reduction), was lower than the 138 mg TSS/ (g ww$d) in the sequencing batch experiment. A plausible explanation was the DO concentration of 6.7 mg/L in the water compartment, which was below the optimum concentration (>8.1 mg/L) required for maximum sludge consumption rate by the worms (Hendrickx et al., 2009a).

3.4.

Worm biomass

The large continuous worm reactor was started with 29.8 g ww of worms, divided over the three mesh cylinders. At the end of the eight weeks of operation, 49.5 g ww of worms was found in the mesh cylinders. During operation of the worm reactor a total of 6.7 g ww of worms had fallen from the mesh and was collected with the worm faeces. Thus, a total worm growth of 26.4 g ww was observed, which corresponded with a yield of 0.18 g dw/g TSS digested by the worms. This is higher than the yield of 0.13 g dw/g TSS digested found in the continuous worm reactor with a horizontal carrier material (Hendrickx et al., 2009b). The average worm net biomass growth rate was 0.014 d1, similar to the growth rate found in the feasibility experiment.

Visual inspection of the mesh cylinders showed that worms were situated along the entire sludge bed inside each mesh cylinder. By the end of the experiment, the total sludge bed height in each cylinder had increased to 20e45 cm. However, most of the worms (w80%) were situated in the top w 10 cm of the sludge bed. This corresponded to a worm density of 1.1 kg ww/m2 carrier material, which matched the stable worm density found in sequencing batch experiments with the same carrier material (Hendrickx et al., 2010a).

3.5.

Nutrients

Nutrient content of the sludge and collected worm faeces were determined during the last four weeks of operation. On average, the sludge from the activated sludge system contained 48 mg total N/g TSS and 15 mg total P/g TSS. The collected worm faeces contained 40 mg total N/g TSS and 16 mg total P/g TSS. Similar to experiments with sludge from a municipal WWTP (Hendrickx et al., 2010b), the phosphorus content of the faeces was higher than in the sludge, whereas the nitrogen content was lower. The amounts of released nutrients per g of digested TSS (Table 2) were similar to those found previously for sludge from a municipal WWTP. Contrary to those experiments, the ammonia release was low (9 mg NH4eN/g TSS digested) and the larger part of the nitrogen release was nitrate (46 mg NO3eN/g TSS digested). This showed that nitrification of the released ammonia had taken place in the worm reactor, which was observed throughout the entire experimental run. Whether this occurred in or near the sludge bed in the mesh cylinders or by nitrifiers in the worm faeces, was not further investigated. Nitrification can also be expected to take place in a full scale worm reactor. This will result in an increase of the oxygen demand of the worm reactor, but also in a decrease of the internal ammonia load on the WWTP. An advantage of nitrification in the worm reactor is that less effluent is required to keep the ammonia concentration low enough (<0.1 mg N/L of unionised ammonia) to prevent inhibition of the sludge consumption rate by the worms (Hendrickx et al.,

Table 2 e Calculated and measured nutrient balance over the worm reactor during the last four weeks of operation. The measured nutrient release is the balance between influent (sludge supernatant, effluent) and effluent (supernatant worm faeces and outflow worm reactor).

Fig. 3 e Cumulative amounts of added waste sludge and collected worm faeces from a continuous worm reactor, which daily received all the waste sludge from a lab-scale activated sludge system.

Consumed sludge Collected worm faeces Digested sludge Worm growth Calculated nutrient release Measured nutrient release

g TSS

gN

110 98

5.03 4.02

1.66 1.46

1.01 0.14 0.87

73

0.20 0.01 0.19

16

0.65

55

0.20

17

12 1.9

mg N/g TSS digested

gP

mg P/g TSS digested

4927

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 2 3 e4 9 2 9

2009a). This then also decreases the internal hydraulic load on the WWTP. Analysis of the supernatant of the worm faeces showed that the total ammonia concentration was high, up to 9.2 mg N/L. At the same time, the total ammonia concentration in the outlet of the worm reactor was very low (<0.5 mg N/ L) and appeared to be independent of worm activity (which increased in time, due to worm growth) and independent of the effluent replacement flow rate, which was decreased stepwise from 42 L/d to 2.8 L/d. This showed that the ammonia load released by the worms (and that was not nitrified), was removed from the reactor mainly with the supernatant of the worm faeces (Fig. 4), thus keeping the ammonia concentration in the water compartment low. Possibly ammonia adsorbed to worm faeces, as has also been reported to occur for sludge (Nielsen, 1996), although in our case this was about a factor 10 higher (2.9 mg NH4eN/g TSS of faeces). For phosphate the average concentration in the supernatant of the worm faeces was 7.2 mg PO4eP/L, only somewhat higher than in the outlet of the worm reactor (6.0 mg PO4eP/L). The dissolved phosphate load removed with the faeces remained constant during the entire experimental period (Fig. 4). The stepwise decrease in effluent replacement rate (with a constant phosphate concentration) caused the decrease in the phosphate load in the outflow from the worm reactor (Fig. 4).

3.6.

Air input

Inside the worm reactor, the effluent in the water compartment was aerated with a single bubble diffuser. Oxygen was also introduced to the reactor with the effluent, which had an average dissolved oxygen (DO) concentration of 5.8 mg/L. In the first weeks, the average DO concentration in the reactor was 7.7 mg/L, whilst during the last couple of weeks, this had dropped to 6.7 mg/L. This may have been caused by the increased oxygen consumption by the worms (as the amount of worm biomass increased over time), by the lower input of

Table 3 e Calculated changes in O2 consumption and O2 supply of the worm reactor between the start and at the end of the 8 week experimental period.

O2 consumption by worms O2 consumption by nitrification O2 supply by influent

Start

End

Increase

mg O2/d mg O2/d

146 0

243 333

þ97 þ333

mg O2/d

249

16

233

DO with the effluent that was pumped into the worm reactor (this flow rate was decreased stepwise over time), by oxygen consumption due to nitrification of the ammonia that was released by the worms and by the respiration of the sludge bed in the mesh cylinders. The temperature was constant at 20.6  0.9  C and could therefore not have caused the decrease in DO concentration. From Table 3 it is clear that the increased O2 consumption by nitrification and the lower supply of dissolved oxygen with the influent contributed much more towards the lower DO in the worm reactor, than the increased O2 consumption by the worms. When designing the oxygen supply to a full scale worm reactor, the large oxygen demand for nitrification will largely determine the required capacity. At the same time, the ammonia load from the worm reactor (and its oxygen requirement) towards the WWTP will be lower.

4.

General discussion

The new reactor configuration with a vertically placed carrier material demonstrated a stable operation over a period of eight weeks. Although only a low TSS reduction of 11% was achieved, worm faeces with a high settleability were continuously produced. The latter was shown to be the main benefit of a worm reactor (Hendrickx et al., 2010b).

Fig. 4 e Ammonia and phosphate load removed with the outflow from the worm reactor and with the worm faeces.

4928

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 2 3 e4 9 2 9

The new configuration has several advantages over the initial configuration with a horizontal carrier material. Firstly, a higher net growth rate of 0.014 d1 over eight weeks was achieved, compared to 0.013 d1 over three weeks for the horizontal carrier reactor described in Hendrickx et al. (2009b). Despite their horizontal orientation, still some worms fell from the carrier material. However, this was less than 0.5% per day of the amount of worms in the mesh cylinders, which was much lower than the 4% found in sequencing batch experiments with the same, but horizontally orientated, mesh material (Hendrickx et al., 2010a). A second advantage is the efficient use of reactor volume. With the new configuration 22.5 m2 of carrier material per m3 of reactor volume can be achieved by spacing the mesh cylinders at a distance of 4 cm. In contrast, to achieve 22.5 m2/m3 in a reactor with a flat horizontal carrier (as used in the original configuration), sections consisting of a water- and sludge compartment would have to be stacked. Each section could then only be 4 cm high; 2 cm for the sludge compartment and 2 cm for the water compartment. Such a small height would make sludge distribution and faeces collection very difficult. Tian and Lu (2010) reported a much lower surface to volume ratio of 6.6 m2/m3 for their worm reactor with Tubificidae in perforated panels. Their lower ratio can be explained by the fact that the panels need to turn during operation of the reactor. However, for a fair comparison between the two systems, the achieved sludge reduction in g TSS per reactor volume should be compared for the same sludge. Practical advantages of the new configuration are related to faeces collection (one collection system for a large number of mesh cylinders) and aeration (air bubbles cannot get trapped under the carrier material). Additionally, the combined collection of non-consumed sludge and effluent with worm metabolites excludes the need to control two levels in the reactor (one in the water compartment and one in the sludge compartment). A separate collection is not required as both effluent and non-consumed sludge need to be returned to the WWTP. The outflow of the worm reactor will contain the nutrients that are released by the worms and will, therefore, require treatment in the WWTP. Previously, we estimated the additional internal hydraulic and nitrogen loads to be 15e20 and 5%, respectively (Hendrickx et al., 2009a). From the current experiments it became clear that part of the internal nitrogen (ammonia) load will be nitrified in the worm reactor, thereby placing the additional oxygen requirement at the worm reactor. The resulting lower ammonia concentration in the worm reactor, will decrease the WWTP effluent requirement of the worm reactor (which is primarily determined by the ammonia concentration) and thus also decrease the internal hydraulic load on the WWTP. A worm reactor such as the one described in this paper is expected to be most interesting for smaller WWTPs with relatively high sludge handling costs. As an example, the results from the current experiments (110 mg TSS/(g ww$d) and 1.1 kg ww/m2) were used to estimate the size of a worm reactor for a 35,000 p.e. WWTP. For a waste sludge production of 1600 kg TSS/d, the footprint of the worm reactor would be 195 m2 (assuming 22.5 m2/m3 and a height of 3 m). This is roughly one tenth of the surface area of a settler of such a WWTP. This additional surface space is expected to be

available at small WWTPs, which are generally not located in densely populated areas. The new reactor configuration is suitable for further scaling up towards a pilot plant of several m3. Such a pilot plant can provide further design and operational details, such as optimal oxygen supply, faeces collection, worm harvesting and how to deal with influences of seasonal changes (in e.g. sludge composition and temperature).

5.

Conclusions

The new reactor configuration for sludge reduction with the immobilised aquatic worm L. variegatus allows for easy aeration of the water compartment, easy removal of worm faeces and a low footprint of the reactor and is therefore suitable for scaling up of the process. A continuous worm reactor directly treated the daily waste sludge from a lab-scale activated sludge reactor and was successfully operated during a period of nearly eight weeks. Net growth of worm biomass clearly took place in the worm reactor at a rate of 0.014 d1. Release of nutrients by the worms was 17 mg PO4eP/g TSS digested and 55 mg (NH4eN þ NO3eN)/g TSS digested. The ammonia released by the worms as a product of their metabolism, was partially converted to nitrate in the worm reactor (46 mg NO3eN/g TSS digested). The remaining ammonia load was removed from the worm reactor mainly with the small flow of worm faeces.

Acknowledgements The authors thank Bert Willemsen (Wageningen University) for his input in designing the new reactor configuration. This work was performed in the TTIW-cooperation framework of Wetsus, Centre of Excellence for Sustainable Water Technology (www.wetsus.nl). Wetsus is funded by the Dutch Ministry of Economic Affairs, the European Union Regional Development Fund, the Province of Fryslaˆn, the City of Leeuwarden and the EZ/Kompas program of the “Samenwerkingsverband Noord-Nederland”. The authors like to thank the participants of the research theme “Membrane Bioreactors” for their financial support.

references

APHA, 1998. Standard Methods for the Examination of Water and Wastewater, twentieth ed. American Public Health Association, Washington, DC. Dentel, S.K., 2001. Conditioning, thickening and dewatering: research update/research needs. Water Science and Technology 44 (10), 9e18. Elissen, H.J.H., Hendrickx, T.L.G., Temmink, H., Buisman, C.J.N., 2006. A new reactor concept for sludge reduction using aquatic worms. Water Research 40, 3713e3718. Foladori, P., Tamburini, S., Bruni, L., 2010. Bacteria permeabilisation and disruption caused by sludge reduction technologies evaluated by flow cytometry. Water Research 44, 4888e4899.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 2 3 e4 9 2 9

Guo, X., Liu, J., Wei, Y., Lin, L., 2007. Sludge reduction with Tubificidae and the impact on the performance of the wastewater treatment process. Journal of Environmental Sciences 19, 257e263. Hendrickx, T.L.G., Temmink, H., Elissen, H.J.H., Buisman, C.J.N., 2009a. The effect of operating conditions on aquatic worms eating waste sludge. Water Research 43, 943e950. Hendrickx, T.L.G., Temmink, H., Elissen, H.J.H., Buisman, C.J.N., 2009b. Aquatic worms eating waste sludge in a continuous system. Bioresource Technology 100, 4642e4648. Hendrickx, T.L.G., Temmink, H., Elissen, H.J.H., Buisman, C.J.N., 2010a. Design parameters for sludge reduction in an aquatic worm reactor. Water Research 44, 1017e1023. Hendrickx, T.L.G., Temmink, H., Elissen, H.J.H., Buisman, C.J.N., 2010b. Aquatic worms eat sludge: mass balances and processing of worm faeces. Journal of Hazardous Materials 177, 633e638. Huang, X., Liang, P., Quan, Y., 2007. Excess sludge reduction induced by Tubifex tubifex in a recylced sludge reactor. Journal of Biotechnology 127, 443e451.

4929

IWA, International Water Association, 2007. In: Spinosa, L. (Ed.), Wastewater Sludge: A Global Overview of the Current Status and Future Prospects. IWA Publishing, London 48. Nielsen, P.H., 1996. Adsorption of ammonium to activated sludge. Water Research 30, 762e764. Song, B., Chen, X., 2009. Effect of Aeolosoma hemprichi on excess activated sludge minimization. Journal of Hazardous Materials 162, 300e304. Tian, Y., Lu, Y., 2010. Simultaneous nitrification and denitrification process in a new Tubificidae-reactor for minimizing nutrient release during sludge reduction. Water Research. doi:10.1016/j.watres.2010.07.069. Tuan, P.-A., Sillanpa¨a¨, M., 2010. Effect of freeze/thaw conditions, polyelectrolyte addition, and sludge loading on sludge electro-dewatering process. Chemical Engineering Journal 164, 85e91. Wei, Y., Liu, J., 2005. The discharged excess sludge treated by Oligochaeta. Water Science and Technology 52 (10e11), 265e272.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 3 0 e4 9 3 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Occurrence of nine nitrosamines and secondary amines in source water and drinking water: Potential of secondary amines as nitrosamine precursors Wanfeng Wang a, Shuoyi Ren a, Haifeng Zhang a, Jianwei Yu a, Wei An a, Jianying Hu b, Min Yang a,* a

State Key Laboratory of Environmental Aquatic Chemistry, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing 100085, China b College of Environmental Science, Peking University, Beijing 100871, China

article info

abstract

Article history:

Due to their high carcinogenicity, the control of nitrosamines, a group of disinfection by-

Received 30 March 2011

products (DBPs), is an important issue for drinking water supplies. In this study,

Received in revised form

a method using ultra-performance liquid chromatography-electrospray ionization tandem

3 June 2011

mass spectrometry was improved for simultaneously analyzing nine nitrosamines in

Accepted 29 June 2011

source water and finished water samples of twelve drinking water treatment plants

Available online 22 July 2011

(DWTPs) in China. The method detection limits of the nine target analytes were

Keywords:

Of the nine nitrosamines, six (N-nitrosodimethylamine (NDMA), nitrosodiethylamine

Nitrosamines

(NDEA),

Secondary amines

N-nitrosomorpholine (NMor), N-nitrosodi-n-butylamine (NDBA), N-nitrosomethylethyl-

Disinfection by-products

amine (NMEA), and N-nitrosodiphenylamine (NDPhA)) were detected. The total nitrosa-

Chlorination

mine concentrations in source water and finished water samples were no detection-

Chloramination

42.4 ng/L and no detection-26.3 ng/L, respectively, and NDMA (no detection-13.9 ng/L and

UPLC-MS/MS

no detection-20.5 ng/L, respectively) and NDEA (no detection-16.3 ng/L and no detection-

0.2e0.9 ng/L for the source water samples and 0.1e0.7 ng/L for the finished water samples.

14.0 ng/L, respectively) were the most abundant. Meanwhile, the occurrence of nine secondary amines corresponding to the nine nitrosamines was also investigated. All of them except for di-n-propylamine were detected in some source water and finished water samples, and dimethylamine (no detection-3.9 mg/L and no detection-4.0 mg/L, respectively) and diethylamine (no detection-2.4 mg/L and no detection-1.8 mg/L, respectively) were the most abundant ones. Controlled experiments involving chloramination of four secondary amines confirmed that dimethylamine, diethylamine, morpholine and di-n-butylamine in water can form the corresponding nitrosamines, with diethylamine and morpholine showing significantly higher yields than dimethylamine which has already been identified as a precursor of NDMA. This study proved that diethylamine, morpholine and di-nbutylamine detected in raw water would be one of the important the precursors of NDEA, NMOR and NDBA, respectively, in drinking water. ª 2011 Elsevier Ltd. All rights reserved.

* Corresponding author. E-mail address: [email protected] (M. Yang). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.06.041

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 3 0 e4 9 3 8

1.

Introduction

Nitrosamines, as a group of emerging disinfection byproducts (DBPs) in drinking water particularly when chloramine is adopted as the disinfectant, have recently raised great concerns because of their high carcinogenic potency in comparison to conventional DBPs such as trihalomethanes (THMs) and haloacetic acids (HAAs). Until now, nine nitrosamines have been detected in drinking water, while N-nitrosodimethylamine (NDMA), N-nitrosodiethylamine (NDEA), and N-nitrosopyrrolidine (NPyr) are the most frequently detected compounds (Asami et al., 2009; CDPH, 2007; Charrois et al., 2007; Jurado-Sanchez et al., 2010; Zhao et al., 2006). Among these nine nitrosamines, six are listed in the third Unregulated Contaminant Monitoring Regulation by the US EPA (2007), and five are in the third version of the Contaminant Candidate List (CCL3) in 2008 (EPA, 2008). The US, Canada and some other countries have established temporary regulations for NDMA, NDEA, N-nitrosodi-n-propylamine (NDPA), and N-nitrosomorpholine (NMor) at regional or even national levels (Inspectorate, 2009; OMET, 2003; Planas et al., 2008; SDWA, 2002). To control conventional DBPs and maintain residual chlorine levels in tap water, chloramine (instead of chlorine) has been used for disinfection in some countries. In addition to direct doses of chloramine, some source waters contain relatively high concentrations of ammonia due to contamination by fertilizer and municipal wastewater (Turrentine, 1929; Wood, 2001); this leads to the formation of chloramine even when using chlorine as the disinfectant. In order to effectively control the formation of nitrosamines during water treatment, it is necessary to explore the potential precursors of these compounds. NDMA formation mechanisms during chlorination, chloramination, and ozonation processes have been extensively studied by many researchers (Chen and Young, 2008; Lorenzo et al., 2007; Oya et al., 2008; Zhao et al., 2006). It has generally been assumed that almost all precursors of NDMA contain one or more dimethylamine groups, and the NDMA formation potential is largely dependent on the structures of these precursors. Identified NDMA precursors in environmental water include dyes (such as N, N-dimethyl-pphenylenediamine), pesticides (such as diuron), pharmaceuticals and personal care products (such as ranitidine) and rubber vulcanizing agents (such as dimethyldithiocarbamate) (Chen and Young, 2008; Fiddler et al., 1972; Vocht et al., 2007; Oya et al., 2008). Recent research has shown that some secondary amines including dimethylamine (DMA) and diphenylamine (DPhA) are also important precursors for corresponding nitrosamines (19-21). In addition to DMA and DPhA, there are other secondary amines, such as diethylamine (DEA), methylethylamine (MEA), di-n-propylamine (DPA), di-nbutylamine (DBA), morpholine (Mor), pyrrolidine (Pyr) and piperidine (Pip), that are important intermediates with significant usage in the chemical and pharmaceutical industries (Kim et al., 2009; Sacher et al., 1997). These amines can also be formed via biodegradation of proteins, amino acids, and other nitrogen-containing compounds (Brink et al., 1990; Lorenzo et al., 2007; Pietsch et al., 2001; Sacher et al., 1997). The concentrations of DMA, DEA and Mor have been

4931

found to be no detection (ND) to 0.55, ND to 63.0, and ND to 2.5 mg/L, respectively, in surface waters in Germany (Akyuz and Ata, 2006). Pip (ND to 0.1 mg/L), Pyr (82e152 mg/L), DPhA (ND to 0.1 mg/L) and DBA (ND to 0.24 mg/L) have been detected in rivers in Turkey (Kamarei et al., 2010). These secondary amines may also serve as the precursors for corresponding nitrosamines in drinking water. However, few studies have focused on the relationship between the suspected secondary amine and corresponding nitrosamines except for DMA and DPhA (Schreiber and Mitch, 2006), although this information is necessary to better control the formation of nitrosamines in drinking water. Several analytical methods using liquid chromatographytandem mass spectrometry (LC-MS/MS) have been used to simultaneously determine nitrosamines in drinking water samples. However, LC-MS/MS shows relatively low sensitivity for some nitrosamines (Plumlee et al., 2008; Zhao et al., 2006). Although liquid chromatography coupled with a linear ion trap quadrupole (LTQ) Orbitrap high resolution MS (LC-MS/HRMS) has exhibited much higher sensitivities to many nitrosamines (Krauss and Hollender, 2008), an LC-MS/MS method with relatively high sensitivity is desirable for extensive environmental investigations. In this study, we improved an ultra-performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS) method to simultaneously analyze the nine nitrosamines with a high sample throughput and sensitivity. Using this method, we investigated the occurrence of the nine nitrosamines, including NDMA, N-nitrosomethylethylamine (NMEA), NPyr, N-nitrosopiperidine (NPip), NMor, NDEA, NDPA, N-nitrosodi-n-butylamine (NDBA) and NDPhA in source water and finished water from twelve drinking water treatment plants (DWTPs) in China. To better characterize the potential precursors, we (1) determined nitrosamine formation potential (FP) for all of the source water samples; and (2) studied the occurrence of nine secondary amines (DMA, DPhA, DEA, MEA, DPA, DBA, Mor, Pyr and Pip) in the source water and finished water samples of the twelve DWTPs and the formation of nitrosamines through chloramination of four extensively detected amines including DMA, DEA, Mor and DBA, respectively. To the best of our knowledge, this is the first systematic study to improve the knowledge of the occurrence of secondary amine precursors and their relationships with the occurrence of corresponding nitrosamines in source water and drinking water.

2.

Experimental section

2.1.

Reagents

A standard solution containing 1000 mg/mL each of NDMA, NMEA, NPyr, NPip, NMor, NDEA, NDPA, NDBA and NDPhA was purchased from Supelco (USA). [2H6] N-nitrosodimethylamine (NDMA-d6), [2H14] N-nitrosodi-n-propylamine (NDPA-d14) and [2H6] dimethylamine hydrochloride were obtained from Cambridge Isotope Laboratories (Andover, MA, USA). HPLCgrade methanol, acetonitrile and hexane were purchased from Fisher Chemical Co. (USA), and dichloromethane was purchased from CNW Technologies GmbH (Germany). Resprep EPA Method 521 cartridges (6 mL/2 g) were purchased

4932

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 3 0 e4 9 3 8

from Restek (Milford, MA, USA), and glass fiber pads (GF/F, 0.7 mm) were obtained from Whatman International Ltd. (Maidstone, UK). PhA, MEA, DPA, DBA, Mor, Pyr and Pip were purchased from Alfa Aesar. DMA and DEA used as hydrochlorides and benzenesulfonyl chloride (>99%) were obtained from SigmaeAldrich. Ammonium bicarbonate was purchased from Fluka (USA), and all other chemicals such as hydrochloric acid (37%) and sodium bicarbonate were obtained from Beijing Chemical Co. (China). Stock solutions for all standard substances were stored at
20  C.

2.2.

Sample collection

Source water and finished water samples were collected from twelve DWTPs during November and December, 2010. Water samples were sampled in inlets and outlets of each DWTP according to the hydraulic retention time (HRT) (about 2 h). Samples were collected in amber bottles and were maintained at a temperature of 10  C during transportation to the lab. Sodium thiosulfate (20 mg/L) was added to the finished water samples to quench any chlorine residue. Information of the twelve DWTPs, include the treatment processes, water quality characteristics, and disinfection types, are shown in Figure S4 and Table S3. Except for DWTP 3 which adopts the advanced water treatment process (ozone-activated carbon), all of the other DWTPs adopt conventional water treatment process. Chloramine disinfection was adopted by two plants (DWTPs 1 and 12), and chlorine disinfection was adopted by the other ten plants.

2.3.

Sample preparation for analyzing nitrosamines

The samples were vacuum filtered through a 0.7 mm glass fiber. After filtration, water samples (500 mL) were spiked with 20 ng/L surrogate standard (NDMA-d6) and were basified to pH 8.0 using sodium bicarbonate. The samples were extracted using Resprep EPA 521 cartridges (Planas et al., 2008). The cartridges were preconditioned with 10 mL of hexane, followed in sequence by 20 mL of dichloromethane, 20 mL of methanol and 20 mL of ultrapure water. The samples were passed through the cartridges at a flow rate of 3e5 mL/min under vacuum conditions. The cartridges were then dried with nitrogen gas. Analytes were eluted with 15 mL of dichloromethane at a flow rate of 2e3 mL/min, 400 mL of a water/ methanol solution (95:5, V/V) was added to the extracts, and the dichloromethane was completely removed using a rotary vaporator (Krauss and Hollender, 2008). Then 50 mL of 20 mg/L surrogate standard (NDPA-d14) was added and the sample volume was gravimetrically adjusted to 0.5 mL using ultrapure water. To remove possible solid particles, all samples were filtered through syringe filters (GHP Acrodisc 13 mm, 0.2 mm, PALL) prior to injection into the UPLC-MS/MS system.

2.4.

UPLC-ESI-MS/MS analysis for nitrosamines

In this experiment, a Waters ACQUITY UPLC system (Waters, USA) consisting of an ACQUITY UPLC binary solvent manager and an ACQUITY UPLC sample manager was used. Chromatographic separation of the compounds was performed at 30  C using an ACQUITY UPLC BEH C18 column (150 mm  2.1 mm,

1.7 mm particle size) (Waters, USA). Mobile phase A was methanol, and mobile phase B was 10 mmol/L ammonium bicarbonate in ultrapure water. The following gradient was used: 0e3 min, 5% A to 45% A; 3e5 min, 45% A to 95% A; 5e7.5 min, 95% A; 7.5e8 min, 95% A to 5% A; 8.0e13.0 min, re-equilibrate with 5.0% A. The flow rate of the mobile phase was 0.2 mL/min, and the injection volume was 30 mL. Analyses were performed using a Waters Micromass Quattro Premier XE detector equipped with an electrospray ionization source. Data acquisition was performed in the positive ion mode, and the optimized parameters were as follows: source temperature, 110  C; desolvation temperature, 400  C; capillary voltage, 4.0 kV; cone voltage, 28 V; desolvation gas flow, 850 L/h; cone gas flow, 50 L/h; and multiplier voltage, 650 V. Argon (99.999%) was used as the collision gas, and the argon pressure in the collision cell was maintained at 3.5e
3 mbar. Quantitative analysis was performed in the multiple reaction monitoring (MRM) mode. The optimal conditions for MS/MS analysis are listed in the Supporting Information Table S1. All of the data were acquired and processed using MassLynx 4.1 software.

2.5. Nitrosamine formation potential test for source waters To investigate the potential risks in drinking water, we studied the nitrosamine-FP of the twelve source water samples using a procedure employed for simulated distribution system (SDS) chlorination (Koch et al., 1991). The Cl2-demand (Cl2-D) of each sample was measured according to the APHA Standard Method prior to the experiments (APHA, 1995). For the determination of NH2Cl-demand (NH2Cl-D), a monochloramine solution was prepared according to a previous paper (Yang et al., 2007). The NH2Cl-D of each sample was then measured using a method similar with that of the Cl2-D measurement. The samples were chlorinated and chloraminated in 500 mL amberized vials by keeping a free chlorine or chloramine residual of 1.0 mg/L after storage at (25  1)  C for 24 h. The concentration of monochloramine was measured using the N, N-diethyl-p-phenylenediamine ferrous ammonium sulfate (DPD-FAS) method (APHA, 1992). Analyses of nitrosamines were then conducted after the residual chlorine and chloramine were quenched.

2.6.

Analysis of the secondary amines in water samples

The nine secondary amines were analyzed using GC/MS after derivation with BSC according to a previously developed method (Sacher et al., 1997). Details are given in Supporting Information.

2.7.

Chloramination of the secondary amines

Reaction solutions were maintained in batches of 1 L sealed amber bottles shielded from light. Unless otherwise specified, all reaction solutions were buffered with a mixture of phosphate buffer (pH 7.2). DMA, DEA, Mor or DBA were respectively added to and fully dissolved in 1 L of water solution to give a final concentration of 0.1 mM. The amine solutions were then reacted with 1.0 mM chloramines for 24 h. For a proper time

4933

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 3 0 e4 9 3 8

interval, one amber bottle was taken for the determination of the formed nitrosamines and residual secondary amines. An aliquot of 1 L of ultrapure water without amine precursors was also processed using the same procedure to serve as a control.

Results and discussion

3.1. Characterization of nitrosamines using UPLC-MS/ MS analysis The optimal UPLC-MS/MS conditions are important for the unequivocal identification of nitrosamines at very low levels in environmental samples. Since the ESI is largely dependent on the solvent conditions, the additives in the mobile phase were investigated. In this study, methanol/water containing 10 mM of ammonium bicarbonate were employed as the mobile phase since this mobile phase composition produced a three- to four-fold increase in the signal intensity, as compared to the methanol/water containing 2 mM or 10 mM ammonium acetate for some of the nitrosamines, particularly for NMor, as shown in Fig. 1(a) and 1(b) (Plumlee et al., 2008; Zhao et al., 2006). To obtain the precursor-product ion pairs for MRM detection, we first characterized the fragmentation behavior of the nine nitrosamines under tandem MS conditions. Table S1 shows the transitions of the nine nitrosamines produced under the optimized conditions. The precursor ion and two

S/N:PtP=38.72

NDMA 75>42.8

0

100

NPyr 100.9>54.8

100

0

%

%

NDEA 102.9>74.8

100

100

NPip 114.9>68.8

0 NDPA 130.9>88.8

S/N:PtP=106.66 100

0 100

S/N:PtP=458.03

NDPhA 199>168.8

S/N:PtP=1233.51

100

0

0 NDPhA 199>168.8

S/N:PtP=991.61

3.2 3.6 4.0 4.4 4.8 5.2 5.6 6.0 6.4 6.8 7.2 Time

100

NDPA 130.9>88.8

0 NDBA 159>56.8

%

S/N:PtP=233.02

%

NDBA 159>56.8

NPip 114.9>68.8

100

%

0

100

%

NDPA S/N:PtP=66.58 130.9>88.8

NDEA 102.9>74.8

0 S/N:PtP=108.85

0

0

5.26

%

%

100

%

NPip 114.9>68.8

NPyr 100.9 > 54.8

100 0

%

%

100

%

NPyr 100.9>54.8

0 S/N:PtP=39.76

NMEA 88.9>60.8

100 0

S/N:PtP=78.08

S/N:PtP=72.53

NDEA 102.9>74.8

%

% %

S/N:PtP=27.96

0

%

NMEA 88.9>60.8

0

0

NMor 116.9>85.7

3.85

0 S/N:PtP=54.83

0 S/N:PtP=33.05

%

100

%

%

100

NMEA 88.9>60.8

%

S/N:PtP=19.15

0

NMor 116.9>85.7

0

0

100

0 S/N:PtP=44.49

%

%

100

NDMA 75>42.8

3.34

%

NMor 116.9>85.7

100

100

100

0 S/N:PtP=13.66

100

NDMA 75>42.8

100

%

%

100

100

c

b S/N:PtP=15.77

100

Of the nine nitrosamines, all except for NPyr, NDPA and NPip were detected in source water and finished water samples as

%

a

3.2. Occurrence of nitrosamines in source water and finished water

NDBA 159>56.8

7.09

0 100

0 3.2 3.6 4.0 4.4 4.8 5.2 5.6 6.0 6.4 6.8 7.2 Time

NDPhA 199>168.8

7.19

%

3.

product ions were used as diagnostic ions (with the exception of NDEA, for which only one product ion was available). Confirmation of the target analytes was accomplished by comparing the retention time (<2%) and the ratio (within 20%) of the selected MRM ion transition with those of standards. The results of the spiking experiments in various water matrices are listed in Table S2, and the overall method recoveries for the target analytes except for NDPhA were between 52.2 and 113%, with a relative standard deviation (RSD) less than 17.6%. The method detection limits (MDLs) of the nine target analytes were between 0.2 and 0.9 ng/L for the source water samples and between 0.1 and 0.7 ng/L for the finished water samples (Table S2), which are considerably low in comparison with previous studies using HPLC-MS/MS for the analysis of nitrosamines. Plunlee et al. (2008) reported on the MDLs of seven nitrosamines (NDMA, NDEA, NDBA, NMEA, NDPA, NPyr and NPip) being 2 and 24 ng/L for natural surface water; Zhao et al. (2006) reported on the MDLs of nine nitrosamines being 0.1 and 10.6 ng/L for water samples. The MDLs in this study were in general comparable to those given by Krauss and Hollender (2008) for an LC-MS/HRMS based method (0.1e3.3 ng/L for finished water samples).

0

3.2 3.6 4.0 4.4 4.8 5.2 5.6 6.0 6.4 6.8 7.2 Time

Fig. 1 e UPLC-MS/MS chromatograms of a standard mix solution (a, b) and detected nitrosamines in water samples (c). (a): methanol/water containing 10 mM ammonium acetate as mobile phases; and (b): methanol/water containing 10 mM ammonium bicarbonate as mobile phase.

4934

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 3 0 e4 9 3 8

exemplified in the UPLC-MS/MS chromatograms of the detected nitrosamines in a water sample (Fig. 1(c)). NDMA, NDEA, NDBA and NMor were the most frequently detected compounds (Table S4). The total concentrations of the six detected nitrosamines in source water samples ranged from ND to 42.4 ng/L. NDMA was detected in four source water samples, with concentrations ranging from 6.4 to 13.9 ng/L, and NDEA (1.9e16.3 ng/L) and NDBA (1.0e19.9 ng/L) were detected in seven and six source water samples, respectively. NMor (1.1e2.8 ng/L), NDPhA (0.6e2.9 ng/L) and NMEA (1.0e1.2 ng/L) were detected in six, five and two samples, respectively. The concentration levels of NDMA are in the same range with those reported in previous studies (Asami et al., 2009; Charrois and Hrudey, 2007). As for the other nitrosamines, NMor was detected in raw water in Spain (2.8 ng/L), and in river water in Germany and France (92e114 ng/L) (Krauss and Hollender, 2008; Planas et al., 2008). The occurrence of these chemicals in source water indicates the presence of some discharging sources. Previous studies have shown that NDMA may be present in secondary effluents from sewage treatment systems and in effluents from industries including rubber, leather, pesticides, food processing, foundries and dyes (Chen and Young, 2008; Feng et al., 2009; Fiddler et al., 1985). NDEA, NDBA, NMor and NDPhA have been detected in emulsion products, such as condoms, baby bottle nipples, rubber, soothers and gloves, and in tobacco, meat and wines (Spatafore and Mcdemoit, 1991; Thilen and Shishoo, 2000; Verna et al., 1996). The relatively high concentrations of NDMA, NDEA and NDBA (11.9, 13.9 and 12.5 ng/L for NDMA; 16.3 and 7.9 ng/L for NDEA; 6.4 ng/L for NDBA) in source water samples from DWTPs 1, 10 and 12 were perhaps related with the contamination of source water as indicated by the relatively high ammonia concentrations (1.7, 0.38 and 0.17 mg/L, respectively). As shown in Table S4, six nitrosamines were also detected in the finished water samples, The concentrations ranged from 4.6 to 20.5 ng/L for NDMA (found in 7/12 samples), 1.9e16.3 ng/L for NDEA (9/12), 0.4e3.4 ng/L for NDBA (6/12), 1.1e1.7 ng/L for NMor (3/12), 1.1 ng/L for NMEA (1/12) and 3.3 ng/L for NDPhA (1/12), respectively. In comparison with the previous studies, the concentrations of NDMA detected in this study are low (max. 65 ng/L in finished water in Canada and 30 ng/L in the US) (Plumlee et al., 2008), but are close to those reported in Japan (ND-2.2 ng/L in summer and ND-10 ng/L in winter) (Asami et al., 2009). As for the other nitrosamines, the concentration levels are in accordance with previously reported studies (ND-12.9 ng/L for NDEA, 1.0e7.9 ng/L for NMor, and 1.86  0.13 ng/L for NDPhA) (Jurado-Sanchez et al., 2010; Planas et al., 2008; Zhao et al., 2006). Comparing to those in source water samples, the concentrations of NDMA in finished water samples from DWTPs 1 increased by 8.6 ng/L and those from DWTPs 5, 8, 9 and 11 increased from ND to 6.7, 4.6, 8.9 and 6.2 ng/L, respectively. The concentrations of NDEA also significantly increased in the samples from DWTP 11 (from 5.2 to 14.0 ng/L) and other DWTPs (from ND to <4 ng/L). Increase of concentrations of NDBA (DWTPs 2, 3, 5 and 8) and NDPhA (DWTP 6) was also observed. In general, these results suggest that disinfection by chlorine or chloramine during drinking water treatment could produce nitrosamines due to the presence of precursors,

which are consistent with previous reports (Chen and Young, 2008; Fiddler et al., 1972; Kemper et al., 2010; Oya et al., 2008; Zhou et al., 2009). The experiment of nitrosamine-FPs of source water also supports the above observation (Table S4). The FPs of NDMA, NDEA, NMor and NDBA for both chloramination and chlorination increased prominently in many source water samples, with the highest values being 108.5, 29.9, 20.2, and 19.5 ng/L, respectively. The FPs of the four nitrosamines during chloramination were in general higher than those found during chlorination, except for DWTPs 10 and 11. In DWTP 3, the concentrations of some nitrosamines (NDMA and NDEA) in finished water samples were much lower than the FPs of source water by chloramination; this could be attributed to the removal of precursors by ozoneactivated carbon treatment (ozone dose, 0.4 mg/L). The relatively high concentration of NDMA (20.5 ng/L) in finished water from DWTP 1 was perhaps attributed to chloramine disinfection. The FPs of NDMA by chlorination in DWTPs 1, 9, 10, 11 and 12 were relatively high (30.8, 31.7, 38.5, 21.0 and 33.3 ng/L, respectively), which could be related to the formation of chloramine due to the existence of ammonia in water (1.7, 0.31, 0.38, 0.23 and 0.17 mg/L, respectively). DWTP 11 contains 0.11 mg/L nitrite, which could also contribute to the formation of NDMA (Choi and Valentine, 2003). On the other hand, the decreases of some nitrosamines from source water to finished water samples were found in some DWTPs. The removal value of NDMA were 1.8, 4.1 and 13.9 ng/L in DWTPs 7, 12 and 10, respectively. For other nitrosamines, the removal efficiencies were 54.4e100% for NDEA (DWTPs 6, 7, 8 and 12), 39.3e100% for NMor (DWTPs 1, 6, 7 and 11), 95.5e100% for NDBA (DWTPs 1, 6, 7, 10 and 12), and 100% for NDPhA (DWTPs 1, 2, 4 and 7), respectively. Similar phenomena have also been reported in previous studies (Asami et al., 2009). Biodegradation in sand filters, or even decomposition by pre-chlorination and, chlorination during disinfection, might be responsible for the removal of nitrosamines (Eizember et al., 1979; Krauss et al., 2009). Further studies are necessary to clarify the removal mechanisms of nitrosamines.

3.3. Occurrence of the potential secondary amine precursors in source water and finished water The occurrence of the nine secondary amines was investigated in the twelve source water and finished water samples, and the results are shown in Table S7. DMA and DEA were detected in most of the samples, with concentrations in source water samples ranging from 0.2 to 3.9 mg/L (10/12) and 0.3e2.4 mg/L (8/12), respectively. The other amines existed at lower concentrations and detection frequency: the concentrations of MEA, Mor, Pyr, Pip, DBA and DPhA were 0.2e1.0 (4/12), 0.1 (5/12), 0.2 (1/12), 0.3 (1/12), 0.1e0.3 (5/12) and 0.1e0.2 mg/L (4/12), respectively, and DPA was not detected. The levels of these secondary amines in rivers are generally within the range reported in previous papers (Akyuz and Ata, 2006; Kamarei et al., 2010; Plumlee et al., 2008). The occurrence of secondary amines in source water could be due to discharges from chemical and pharmaceutical industries and municipal wastewater (Brink et al., 1990; Lorenzo et al., 2007;

Fig. 2 e Correlations between nitrosamines and the corresponding secondary amines in water samples.

4936

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 3 0 e4 9 3 8

Sacher et al., 1997). Aliphatic amines like DMA, DEA, etc., are important intermediates in chemical and pharmaceutical industries, and some of them are produced in quantities of more than 100 000 tons per year in Western Europe. The concentration of DMA in an industrial waste stream, for example, was reported to be up to 100 mg/L, and that of DEA was up to 30 mg/L (Sacher et al., 1997). In this study, some source water samples (DWTPs 1, 3, 4, 5, 7, 12) were taken from inland rivers, along which a lot of industrial plants are distributed. Waste streams discharged from some of these plants may contain secondary amines. DMA and DEA were also detected in most of the finished water samples, with concentrations ranging from 0.4 to 4.0 mg/L (11/12) and 0.1e1.8 mg/L (11/12), respectively. It is interesting that the concentrations of DMA, DEA and Mor in some finished water samples were higher than those found in source water samples. The increase of DMA concentrations was observed in samples from DWTPs 4, 8, 11 and 12 (from ND, 0.6 mg/L, 0.8 mg/L and ND to 0.4, 0.9, 3.9 and 0.5 mg/L), respectively. The increase of DEA concentrations was found in samples from DWTPs 12, 3, 4 and 1, respectively (from ND, ND, ND and 0.3 mg/L to 1.1, 1.6, 1.8 and 1.5 mg/L). These results suggest that some source water might contain CeN bond-bearing compounds that could be transformed into secondary amines during water treatment. It has been reported that some tertiary amines (e.g., trimethylamine, 3-(dimethylaminomethyl)indole, and 4-di methylaminoantipyrine) produce DMA during ClO2 treatment (Lee et al., 2007). Fig. 2 compares the concentrations of the secondary amines with the corresponding nitrosamines in both source water and finished water samples. In general, the presence of NDMA, NDEA, NMor and NDBA in source water, finished

Ym ¼

3.0

0.12 DEA

2.0

0.08

2.0

0.06

1.5

0.06

1.5

0.04

1.0

0.04

1.0

0.02

0.5

0.02

0.5

0.0

0.00

0.00 0

2

4

6

8

DEA (uM)

0.08

NDMA (nM)

0.10

0.0 0

10 12 14 16 18 20 22 24 Reaction time (h)

0.12

NDEA

3.0

2

4

6

8

10 12 14 16 18 20 22 24 Reaction time (h)

0.12

3.0

Mor 0.10

DBA

0.08

2.0

0.08

2.0

0.06

1.5

0.06

1.5

0.04

1.0

0.04

1.0

0.02

0.5

0.02

0.5

0.0

0.00

0.00 0

2

4

6

8

10 12 14 16 18 20 22 24 Reaction time (h)

DBA (uM)

0.10

NMor (nM)

2.5

Nmor

NDEA (nM)

2.5

2.5

0.10

Mor (uM)

(1)

NDBA

2.5

NDBA (nM)

DMA NDMA

CNt  100% ðCA0
CAt Þ

where CA0 is the initial amine concentration (mM), and CNt (nM) and CAt (mM) are the concentrations of nitrosamines and amines in solution at a specified reaction time (t), respectively. The 24-h molar yield of NDEA from DEA after chloramination was the highest (5.9%), followed by Mor (4.19%), DMA (0.74%) and DBA (0.17%), and the maximum concentrations of NDMA,

3.0

0.12

DMA (uM)

water, chlorinated and chloraminated water samples was observed when the corresponding secondary amines were present, suggesting that DMA, DEA, Mor and DBA, as potential precursors, might have contributed to the formation of the nitrosamines, respectively. In addition, the detectable rates of the nine nitrosamines in finished water samples were correlated positively with the detectable rates of the corresponding secondary amines in source water samples (r2 ¼ 0.87). In fact, DMA has been identified as one of the precursors of NDMA when chloramine is used as a disinfectant (Choi and Valentine, 2002; Mitch and Sedlak, 2002). However, until now, no evidence has shown that DEA, Mor and DBA are respectively the precursors of NDEA, NMor and NDBA during disinfection. Therefore, the formation of nitrosamines from the secondary amines during chloramination was investigated in controlled experiments (Fig. 3). The results show that, in all four solutions, the concentrations of the nitrosamines increased with decreasing concentrations of the secondary amines over time, indicating that the four amines are the potential precursors of their corresponding nitrosamines. The molar yields (Ym) of nitrosamine formation from the four secondary amines were calculated based on Equation (1) (Zhou et al., 2009).

0.0 0

2

4

6

8

10 12 14 16 18 20 22 24 Reaction time (h)

Fig. 3 e Nitrosamine formation from the secondary amines during chloramination; 0.1 mM amine (DMA, DEA, Mor and DBA) reacted with 1.0 mM chloramine, respectively.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 3 0 e4 9 3 8

NDEA, NMor and NDBA were 0.24, 1.18, 2.22 and 0.12 nM, respectively. The pseudo-first order reaction constants for DMA, DEA, Mor and DBA were calculated to be 0.025, 0.06, 0.31 and 0.004 h
1, respectively, showing that the reactivity of different secondary amines with chloramines were quite different. Zhou et al. (2009) attributed the differences in the reactivity of secondary amines (DMA and DPhA) in forming corresponding nitrosamines (NDMA and NDPhA) to the effect of pH and structural-dependent reactivity. The pKb values of DMA, DEA, Mor and DBA are 2.77, 3.30, 5.51 and 2.91, respectively, which might be able to explain the differences of the pseudo-first order reaction constants. The molar yield of NDMA from DMA was very low, which was in accordance with previous studies (Chen and Valentine, 2006; Mitch et al., 2003; Gerecke and Sedlak, 2003). In spite of the similar PKb values of DBA, DMA and DEA (2.77, 3.3 and 2.91, respectively), the molar yields of their corresponding nitrosamines were quite different. There might be some other mechanisms affecting the yields of nitrosamines, which requires further study. The high yields of NDEA from DEA and NMor from Mor may explain the significant increases of corresponding nitrosamines in some finished water or chloraminated water samples (NDEA in samples from DWTPs 3, 4, 6, 7, 9 and 11 and NMor in samples from DWTPs 1 and 12). In samples from DWTPs 3 and 4, NDEA was detected in finished water samples or chlorinated/chloraminated water samples even though the corresponding secondary amine DEA was not detected in source water samples as shown in Table S7 and Fig. 2. From the fact that DEA was detected in finished water samples, it is speculated that DEA produced from some complicated CeN bond-bearing compounds was further transformed into NDEA during disinfection. Thus, we found that nitrosamines, particularly NDMA, NDEA, NMor and NDBA are prevalent in some source water and finished water in China, and their corresponding secondary amines could be one of their precursors based on both survey and simulation results using model chemicals. The findings of this study are useful for the establishment of nitrosamine control strategies in water supply industry.

Acknowledgments This work was supported by the National Natural Science Foundation of China (grant no. 21077118) and the National Special Funding Project for Water Pollution Control and Management of China (2009ZX07419-001, 2008ZX07421-004). The authors would like to express their gratitude towards Dr. Urs von Gunten from Swiss Federal Institute of Aquatic Science and Technology for his help with the valuable advice.

Appendix. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011. 06.041.

4937

references

Akyuz, Mehmet, Ata, Sevket, 2006. Simultaneous determination of aliphatic and aromatic amines in water and sediment samples by ion-pair extraction and gas chromatography emass spectrometry. Journal of Chromatography A 1129, 88e94. American Public Health Association, 1992. Standard Methods for the Examination of Water and Wastewater. American Public Health Association, 1995. Standard Methods for the Examination of Water and Wastewater. United Book Press. Asami, M., Oya, M., Kosaka, K., 2009. A nationwide survey of NDMA in raw and drinking water in Japan. Science of the Total Environment 407 (11), 3540e3545. Brink, B.T., Damink, C., Joosten, H.M.L.J., Huis in’t Veld, J.H.J., 1990. Occurrence and formation of biologically active amines in foods. International Journal of Food Microbiology 11, 73e84. California Department of Public Health, 2007. California Drinking Water: NDMA-Related Activities. http://www.cdph.ca.gov/ certlic/drinkingwater/Pages/NDMA./aspx. Charrois, J.W.A., Hrudey, S.E., 2007. Breakpoint chlorination and free-chlorine contact time: implications for drinking water Nnitrosodimethylamine concentrations. Water Research 41 (3), 674e682. Charrois, J.W.A., Boyd, J.M., Froese, K.L., Hrudey, S.E., 2007. Occurrence of N-nitrosamines in Alberta public drinkingwater distribution systems. Journal of Environmental Engineering 6 (1), 103e114. Chen, Z., Valentine, R.L., 2006. Modeling the formation of N-nitrosodimethylamine (NDMA) from the reaction of natural organic matter (NOM) with monochloramine. Environmental Science & Technology 40 (23), 7290e7299. 7297. Chen, W.H., Young, T.M., 2008. NDMA formation during chlorination and chloramination of aqueous diuron solutions. Environmental Science & Technology 42 (4), 1072e1077. Choi, J.H., Valentine, R.L., 2002. Formation of Nnitrosodimethylamine (NDMA) from reaction of monochloramine: a new disinfection by-product. Environmental Science & Technology 37, 4871e4876. Choi, J.H., Valentine, R.L., 2003. N-nitrosodimethylamine formation by free-chlorine-enchanced nitrosation of dimethylamine. Water Research 36 (4), 817e824. Eizember, R.F., Vogler, K.R., Souter, R.W., Cannon, W.N., Wege II, P.M., 1979. Destruction of nitrosamines. treatment of nitrosamines with various acids and halogens. The Journal of Organic Chemistry 44 (5), 784e786. Feng, D., Wang, H., Cheng, X., Wang, J., Ning, L., Zhou, Q., Zhou, Y., Yang, Q., 2009. Detection and toxicity assessment of nitrosamines migration from latex gloves in the Chinese market. International Journal of Hygiene and Environmental Health 212 (5), 533e540. Fiddler, W., Pensabene, J.W., Doerr, R.C., Wasserman, A.E., 1972. Formation of N-nitrosodimethylamine from naturally occurring quaternary ammonium compounds and tertiary amines. Nature, 236e307. Fiddler, W., Pensabene, J.W., Kimoto, W.I., 1985. Investigation of volatile nitrosamines in disposable protective gloves. American Industrial Hygiene Association Journal 46 (8), 463e465. Gerecke, A.C., Sedlak, D.L., 2003. Precursors of Nnitrosodimethylamine in natural waters. Environmental Science & Technology 37 (7), 1331e1336. Government of Ontario. Safe Drinking Water Act, 2002. Ontario Regulation 169/03, Schedule 2. http://www.ene.gov.on.ca/ envision/water/sdwa/legislation.htm. Inspectorate, D.W., 2009. Guidance on the Water Supply (Water Quality) Regulations 2000 Specific to N-Nitrosodimethylamine (NDMA) Concentrations in Drinking Water.

4938

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 3 0 e4 9 3 8

Jurado-Sanchez, B., Ballesteros, E., Gallego, M., 2010. Screening of N-nitrosamines in tap and swimming pool waters using fast gas chromatography. Journal of Separation Science 33 (4e5), 610e616. Kamarei, F., Ebrahimzadeh, H., Yamini, Y., 2010. Optimization of solvent bar microextraction combined with gas chromatography for the analysis of aliphatic amines in water samples. Journal of Hazard Materials 178, 747e752. Kemper, J.M., Walse, S.S., William, M.A., 2010. Quaternary amines as nitrosamine precursors: a role for consumer products? Environmental Science & Technology 44 (4), 1224e1231. Kim, M.K., Mah, J.H., Hwang, H.J., 2009. Biogenic amine formation and bacterial contribution in fish, squid and shellfish. Food Chemisty 116 (1), 87e95. Koch, B., Krasner, S.W., Sclimenti, M.J., Schimpff, W.K., 1991. Predicting the formation of DBPs by the simulated distribution system. Journal of the American Water Works Association, 62e70. Krauss, M., Hollender, J., 2008. Analysis of nitrosamines in wastewater: exploring the trace level quantification capabilities of a hybrid linear ion trap/orbitrap mass spectrometer. Analytical Chemistry 80 (3), 834e842. Krauss, M., Longree, P., Dorusch, F., Ort, C., Hollender, J., 2009. Occurrence and removal of N-nitrosamines in wastewater treatment plants. Water Research 43, 4381e4391. Lee, C., Schmidt, C., Yoon, J., Von Gunten, U., 2007. Oxidation of N-nitrosodimethylamine (NDMA) precursors with ozone and chlorine dioxide: kinetics and effect on NDMA formation potential. Environmental Science & Technology 41 (6), 2056e2063. Lorenzo, J.M., Martinez, S., Franco, I., Carballo, J., 2007. Biogenic amine content during the manufacture of dry-cured lacon, a Spanish traditional meat product: effect of some additives. Meat Science 77, 287e293. Mitch, W.A., Sedlak, D.L., 2002. Formation of Nnitrosodimethylamine (NDMA) from dimethylamine during chlorination. Environmental Science & Technology 36 (4), 588e595. Mitch, W.A., Gerecke, A.C., Sedlak, D.L., 2003. A Nnitrosodimethylamine (NDMA) precursor analysis for chlorination of water and wastewater. Water Research 37 (15), 3733e3741. Ontario Ministry of the Environment, 2003. Technical Support Document for Ontario Drinking Water Standards, Objectives and Guidelines. http://www.ene.gov.on.ca/envision/gp/ 4449e01.pdf (Toronto, Ontario). Oya, M., Kosaka, K., Asami, M., Kunikane, S., 2008. Formation of N-nitrosodimethylamine (NDMA) by ozonation of dyes and related compounds. Chemosphere 73 (11), 1724e1730. Pietsch, J., Sacher, F., Schmidt, W., Brauch, H.J., 2001. Polar nitrogen compounds and their behavior in the drinking water treatment process. Water Research 35 (15), 3537e3544. Planas, C., Palacios, O., Ventura, F., Rivera, J., Caixach, J., 2008. Analysis of nitrosamines in water by automated SPE and

isotope dilution GC/HRMS - occurrence in the different steps of a drinking water treatment plant, and in chlorinated samples from a reservoir and a sewage treatment plant effluent. Talanta 76 (4), 906e913. Plumlee, M.H., Lopez-Mesas, M., Heidlberger, A., Ishida, K.P., Reinhard, M., 2008. N-nitrosodimethylamine (NDMA) removal by reverse osmosis and UV treatment and analysis via LC-MS/ MS. Water Research 42 (1e2), 347e355. Sacher, F., Lenz, S., Brauch, H.J., 1997. Analysis of primary and secondary aliphatic amines in waste water and surface water by gas chromatography mass spectrometry after derivatization with 2,4-dinitrofluorobenzene or benzenesulfonyl chloride. Journal of Chromatography A 764 (1), 85e93. Schreiber, I.M., Mitch, W.A., 2006. Occurrence and fate of nitrosamines and nitrosamine precursors in wastewaterimpacted surface waters using boron as a conservative tracer. Environmental Science & Technology 40 (10), 3203e3210. Spatafore, R., Mcdemoit, L., 1991. Near-IR reflectance analysis quantifies polyolefin additives. Platsics Compound 14 (6), 68e71. Thilen, M., Shishoo, R., 2000. Optimization of experimental parameters for the quantification of polymer additives using SFE/HPLC. Journal of Applied Polymer Science 76 (6), 938e946. Turrentine, J.W., 1929. Synthetic ammonia in the fertilizer industry. Journal of Chemical Education, 894e898. Washington, D.C. US EPA, 2007. http://www.epa.gov/IRIS/subst/0042.htm. US EPA, 2008. http://www.epa.gov/OGWDW/ccl/ccl3.html. Verna, L., Whysner, J., Williams, G.M., 1996. Nnitrosodiethylamine mechanistic data and risk assessment: bioactivation, DNA-adduct formation, mutagenicity, and tumor Initiation. Pharmacology & Therapeutics 71 (1e2), 57e81. Vocht, F. de., Burstyn, I., Straif, K., Vermeulen, R., Jakobsson, K., Nichols, L., Peplonska, B., Taeger, D., Kromhout, H., 2007. Occupational exposure to NDMA and NMor in the European rubber industry. Journal of Environmental Monitoring 9 (3), 253e259. Wood, C.M., 2001. Influence of feeding, exercise, and temperature on nitrogen metabolism and excretion. Fish Physiology 20, 201e238. Yang, W., Chen, J., Li, X., Liang, H., 2007. Study on polluted raw water by pre-oxidation of chloramine. Journal of Harbin University 23 (2), 148e152. Zhao, Y.Y., Boyd, J., Hrudey, S.E., Li, X.F., 2006. Characterization of new nitrosamines in drinking water using liquid chromatography tandem mass spectrometry. Environmental Science & Technology 40 (24), 7636e7641. Zhou, W.J., Boyd, J.M., Qin, F., Hrudey, S.E., Li, X.F., 2009. Formation of N-nitrosodiphenylamine and two new ncontaining disinfection byproducts from chloramination of water containing diphenylamine. Environmental Science & Technology 41 (21), 8443e8448.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 3 9 e4 9 5 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Removal kinetics of organic compounds and sum parameters under field conditions for managed aquifer recharge Bernd Wiese a,b,*, Gudrun Massmann f,g, Martin Jekel c, Thomas Heberer e,h, Uwe Du¨nnbier d, Dagmar Orlikowski a, Gesche Gru¨tzmacher a a

Berlin Centre of Competence for Water, gGmbH, Cicerostraße 24, 10709 Berlin, Germany Helmholtz Centre Potsdam, German Research Centre for Geosciences, Telegrafenberg, 14473 Potsdam, Germany c Department of Water Quality Control, Institute for Environmental Engineering, Technical University Berlin, Sekr. KF 4, Strasse des 17. Juni 135, 10623 Berlin, Germany d Berlin Water Company, Motardstraße 35, 13629 Berlin, Germany e Federal Office of Consumer Protection and Food Safety, Department 3: Veterinary Drugs, Mauerstrasse 39-42, 10117 Berlin, Germany f Carl von Ossietzky Universita¨t Oldenburg, Working Group Hydrogeology and Landscape Hydrology, 26111 Oldenburg, Germany g Freie Universita¨t Berlin, Working Group Hydrogeology, Malteser Str. 74-100, 12249 Berlin, Germany h Technical University of Berlin, Institute of Food Chemistry, Sekr. TIB 4/3-1, Gustav-Meyer-Allee 25, 13355 Berlin, Germany b

article info

abstract

Article history:

Managed aquifer recharge (MAR) provides efficient removal for many organic compounds

Received 28 February 2011

and sum parameters. However, observed in situ removal efficiencies tend to scatter and

Received in revised form

cannot be predicted easily. In this paper, a method is introduced which allows to identify

24 June 2011

and eliminate biased samples and to quantify simultaneously the impact of (i) redox

Accepted 28 June 2011

conditions (ii) kinetics (iii) residual threshold values below which no removal occurs and (iv)

Available online 22 July 2011

field site specifics. It enables to rule out spurious correlations between these factors and therefore improves the predictive power. The method is applied to an extensive database

Keywords:

from three MAR field sites which was compiled in the NASRI project (2002e2005, Berlin,

Bank filtration

Germany). Removal characteristics for 38 organic parameters are obtained, of which 9 are

Removal

analysed independently in 2 different laboratories. Out of these parameters, mainly phar-

Kinetics

maceutically active compounds (PhAC) but also sum parameters and industrial chemicals,

Trace organics

four compounds are shown to be readily removable whereas six are persistent. All partly

Multi-tracer

removable compounds show a redox dependency and most of them reveal either kinetic

Statistical analysis

dependencies or residual threshold values, which are determined. Differing removal effi-

Threshold concentration

ciencies at different field sites can usually be explained by characteristics (i) to (iii).

Residual concentration

ª 2011 Elsevier Ltd. All rights reserved.

Pharmaceutical residues

1.

Introduction

Managed Aquifer Recharge (MAR) describes intentional infiltration, treatment and storage of water in aquifers. It provides efficient removal for many organic water parameters

(compounds and sum parameters) but it is a difficult task to quantify removal under field conditions: Observed concentrations often tend to scatter and may be biased by subsurface mixing of different waters, e.g. ambient groundwater or bank filtrate of different age. The removal efficiency is affected by

* Corresponding author. Berlin Centre of Competence for Water, gGmbH, Cicerostraße 24, 10709 Berlin, Germany. E-mail address: [email protected] (B. Wiese). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.06.040

4940

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 3 9 e4 9 5 0

different environmental parameters, such as redox potential and residence times (e.g. Stuyfzand et al., 2007; Heberer et al., 2008; Baumgarten et al., 2011). In addition, factors such as threshold values (residual concentrations below which no removal occurs) and field site specifics may have an impact but have so far achieved little attention in the literature. In this context, the surface water concentration at the time of infiltration is crucial. Furthermore, observed removal efficiencies depend on the analytical limit of quantification (LOQ). If the interaction of these factors is not considered properly, the data evaluation might be biased by spurious correlations and lead to data misinterpretation. For example, removal efficiencies appear to depend on redox conditions or appear to be field site specific, while in fact, they are simply the consequence of longer flow times e which are site specific and will also affect redox conditions. Likewise, different sites may have different surface water concentrations, residence times and aquifer characteristics. On the other hand, field site specific removal may occur without correlation to one of the analysed factors. Frequently, removal efficiency is determined with strict a priori assumptions. Several methods comprise the calculation of a mean value per observation well (Denecke, 1997; Stuyfzand et al., 2007) or with box plots (Gru¨nheid et al., 2005; Heberer et al., 2008; Massmann et al., 2008a). Stuyfzand et al. (2007) explicitely considers redox conditions and the flow time to calculate removal with a half-life approach. However, for all these approaches data are aggregated at an early stage, transient effects are not considered explicitly and removal kinetics are inferred rather than deduced by the data. These effects are accounted for in complex reactive biogeochemical models (e.g. Greskowiak et al., 2006), that have a limited applicability because they are very time and cost consuming and therefore often not realizable. The presented method overcomes these difficulties, quantifies the removal efficiencies and kinetics and is applicable to large data bases. This allows to reduce the impact of outliers and statistical dispersion. The most advanced method of evaluation is described in Stuyfzand et al. (2007). However, for the currently available database (KWB, 2009) a method is developed that does not infer removal characteristics a priori, but determines the type of kinetic directly from the data. We present a method to quantify removal of organic water parameters during MAR. The method consists of the following elements: (i) The determination of flow time for each sample, based on observed tracer data and surface water concentrations, (ii) the identification and subsequent exclusion of mixed samples, (iii) the determination of organic parameter concentrations at the time of infiltration by flow time and retardation and (iii) the quantification of the removal as the difference between concentration at the time of infiltration and the actual groundwater concentration, taking account for the residence time. The evaluation is carried out individually for different redox conditions by a combination of a statistical analysis (providing precise condition dependent removal efficiency) and a graphical analysis to determine removal kinetics. The combination of both approaches allows to quantify significant removal processes and to rule out spurious correlations. The data for this study (KWB, 2009) were collected within the NASRI project between 2002 and 2005 at two bank

filtration (BF) sites (Tegel BF, Wannsee BF) and one basin aquifer recharge site (Tegel AR) in Berlin. In total, 38 organic parameters were analysed, of which 9 were measured in parallel by two working groups (Table 1). The working groups are TU-ORG (Jekel, 2006), TU Drug (Heberer and Jekel, 2006) and BWB (Du¨nnbier, Dlubek). Furthermore, it contains additional geochemical parameters, of which chloride, boron, 18O and temperature are used as tracers. Results of various aspects of the NASRI project at these field sites have previously been published. For the Tegel BF site results were reported on transient hydraulics and transport (Wiese and Nu¨tzmann, in press; Wiese, 2006), environmental tracers (Massmann et al., 2008c) and organic compounds (Gru¨nheid et al., 2005). Details of the Wannsee BF site were presented focussing on redox conditions (Massmann et al., 2008a), re-aeration due to water-level fluctuations (Massmann et al., 2008a,b,c; Kohfahl et al., 2009), phenazone type PhACs (Massmann et al., 2008a) and antimicrobial residues (Heberer et al., 2008). Transient hydraulics and geochemistry for the infiltration zone at the Tegel AR are given in Greskowiak et al. (2005), while redox conditions, Pharmaceutically Active Compounds (PhACs) as well as bulk organic carbon behaviour along a transect were described by Massmann et al. (2006) and Gru¨nheid et al. (2005). Finally, Greskowiak et al. (2006) modelled the reactive, redox dependant degradation of the analgesic phenazone at the Tegel AR site.

2.

Methods

2.1.

Analytical methods

The analysed compounds and their affiliation are listed in Table 1. The method of analysis and Limit of Quantification (LOQ) are listed in the third and fourth column. Several compounds are determined in parallel by 2 laboratories. Methods 1 and 2 were applied by the working group TU DRUGS. Method 1 (derivatizing with pentafluorobenzyl bromide) was primary used for analyzing the acidic residues like clofibric acid, diclofenac etc. For neutral compounds method 2 (silylation with N-(tert.-butyldimethylsilyl)-Nmethyl-trifluoroacetamide) was applied. The methods are described by Reddersen and Heberer (2003). Method 3 was applied by the working group TU Drugs. (Heberer et al., 2008) Method 4 was applied by the working group TU-ORG and is a combination of solid phase extraction (SPE) and MS/MSquantification. (Putschew et al., 2001; Hartig, 2000). Method 5 is carried out by the working group TU-ORG. It is describe by Storm et al. (1999). The analyses of method 6 were applied in the laboratory of the BWB. The method follows Zu¨hlke et al. (2004). The analyses of method 7 were applied in the laboratory of the BWB. The method follows DIN 38413-P10. Analyses of method 8 were carried out in the laboratory of the BWB. The method is developed within the BWB follows the internal protocol BWB 03-02. Chloride was analysed according to the DIN EN ISO 10304-1 (D20) method, Boron with the DIN EN ISO 11885 (E22) method at the BWB laboratory. Stable isotope measurements (d18O, dD)

4941

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 3 9 e4 9 5 0

Table 1 e List of analysed organic compounds, analytical methods and limits of quantification (LOQ). Compound

Affiliation

AAA (Acetylaminoantipyrine) AMDOPH (1-acetyl-1-methyl-2-dimethyl-oxamoyl-2-phenylhydrazide) AMPH (1-acetyl-1-methyl-2-phenylhydrazide) Bentazone Bezafibrate Carbamazepine (CBZ)

metabolite analgesic metabolite analgesic metabolite analgesic herbicide blood lipid regulator antiepileptic

Clarithromycin Clindamycin Clofibric acid Diclofenac DP (1,2-dihydro-1,5-dimethylpyrazol-3-one) EDTA (ethylenediaminetetraacetic acid) FAA (formylaminoantipyrine) Indometacine Iopromide Mecoprop MTBE (methyl-tertiary-butyl-ether) N-(phenylsulfonyl)-sarcosine (NPS) NDSA-Isomers (Naphthalenedisulfonic acid) o,p’-DDA (2-(2-chlorophenyl)-2-(4-chlorophenyl) acetic acid) PDP (1,2-dihydro-4-isopropyl-1,5-dimethyl-pyrazol-3-one) Phenazone p,p’-DDA (2,2-bis(4-chlorophenyl)acetic acid) Primidone Propyphenazone

antibiotic antibiotic metabolite of blood lipid regulator analgesic/anticonvulsants metabolite analgesic complexing agent metabolite analgesic analgesic/anticonvulsants x-ray contrast agent herbicide fuel additive metabolite of a corrosion inhibitor industrial chemicals metabolite of an insecticide metabolite analgesic analgesic metabolite of an insecticide analgesic/anticonvulsants analgesic/anticonvulsants

Roxithromycin Sulfadimidine (SDMD) Sulfamethoxazole (SMX)

antibiotic antibiotic bacteriostatic antibiotic

Trimethoprim

antibiotic

were carried out at the Alfred Wegener Institute, research Unit Potsdam. Method details are described in Meyer et al. (2000).

2.2.

Flow time

The flow time for each sample is determined by a 1-D transient multi-tracer approach (d18O, temperature and chloride). The tracer input signal of the surface water is shifted by a date dependent flow time to fit the breakthrough curve in each observation well. Most flow times are determined with d18O because it is an ideal tracer and shows pronounced seasonal surface water variations. However, monthly sample intervals set a lower limit of resolution of 1e2 months. Shorter flow times are determined with temperature for which a retardation factor of 2 is taken into account. Temperature is more suitable for short flow times since retardation doubles the time shift compared to conservative tracers and logging data often provides continuous time series with a much higher resolution than obtained by sampling. The shortest flow times which can be determined with this method are about two to three days. In cases in which sampling of surface water and groundwater started simultaneously the corresponding first concentrations of environmental tracer in the surface water are not known and flow time cannot be determined. Consequently, the samples are disregarded in further analysis.

Method of analysis

LOQ (ng/L)

6 1 6 6 1 1 2 6 3 3 1 1 6 7 6 1 4 1 8 1 5 1 6 6 1 2 1 6 3 3 3 4 3

50 5 50 50 5 50 5 50 0.2 0.1 10 5 50 2000 50 30 20 5 30 30 30 10 50 50 10 5 10 50 0.2 3 1 20 2

The underlying assumptions are that dispersion can be neglected and no preferential flow occurs. This is probably true, since dispersivity typically is one order of magnitude lower than the problem scale (Gelhar et al., 1992). The current analysis refers to observation wells, that basically show parallel flow and lower dispersion than abstraction wells with convergent flow. For maximum flow times around 100 days dispersion effects are in the order of 10 days, which is generally lower than the time scale of the regarded processes. Furthermore, preferential flow would be detected as mixing (see below).

2.3.

Mixing

Groundwater samples at bank filtration sites may be biased due to mixing of different waters, either ambient groundwater (Hiscock and Grischek, 2002) or bank filtrate of different age (Wiese, 2006). Mixing is identified by using the tracers chloride, boron and propyphenazone. Propyphenazone is an indicator for old bank filtrate at the field site Tegel BF (Reddersen et al., 2002). It is very sensitive because groundwater concentrations are up to about 10 times higher than presently found in surface water. Taking into account the variability in surface and groundwater as well as the measurement accuracy the following criteria for mixing were applied:

4942

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 3 9 e4 9 5 0


Boron: groundwater concentrations differ at least 10 mg/l or 10% from corresponding concentration at the time of infiltration
Chloride: groundwater concentrations differ at least 4 mg/l or 10% from the corresponding concentration at the time of infiltration
Propyphenazone: groundwater concentration of the sample is twice the limit of quantification (LOQ) and more than 25% higher than the corresponding concentration at the time of infiltration. If two of the criteria are fulfilled, the sample is classified as subsurface mixing and is not included in the further evaluation. Most mixed samples are identified at Tegel BF, due to vertical stratification of different ages of bank filtrate (Wiese and Nu¨tzmann, in press).

2.4.

Retardation study

Many organics are adsorbed to the aquifer material, hence their residence time is longer than the flow time of water and conservative tracers. The degree of sorption depends on the strength of the interaction between the compound and the adsorbing matrix and is described with a retardation factor. Retardation factors were determined with a literature study. Experimental retardation values are available for 13 parameters. Within the NASRI project 7 retardation factors were determined during column experiments (Jekel, 2006; Licht et al., 2005). Referenced values for comparable conditions (sandy aquifer, low sedimentary organic carbon) are used for further 6 parameters (Braids, 2001; Broholm et al., 2001; Buss et al., 2006; Grischek et al., 1997; Heberer and Jekel, 2006; Jørgensen et al., 1998; Mersmann et al., 2003; Scheytt et al., 2006; Stuyfzand et al., 2007; Tuxen et al., 2000). Eight retardation values are applied based on similarity considerations regarding the substances’ structural characteristics (Table 2). For 17 parameters no physically based values are available, hence a retardation factor equal to one is applied e thereby considering a worst case scenario. The residence times of the compounds and sum parameters are determined by multiplying the flow time with the retardation factor. The retardation has the underlying assumption of linear equilibrium sorption. This may not generally be true, but appears appropriate considering the data basis. In case evidence for more complex sorption mechanisms dominate, the analysis may be biased.

2.5.

Removal calculation

Removal can only be calculated if the respective surface water concentration is known. For each parameter and groundwater sample the time of infiltration is determined by starting at the sampling date and subtracting the residence time. The corresponding substance concentration at time of infiltration is acquired by linear interpolation between surface water concentrations. In cases the time of infiltration is less than 2 weeks prior to the first measured surface water concentration, this concentration is applied as source concentration. A time of 2 weeks is chosen because this is half of the monthly sampling interval. For earlier times of infiltration the

concentration becomes undefined, and therefore also the respective groundwater concentration is disregarded in further analysis. Rabs ¼ csw  cgw

(1)

where Rabs is the absolute removal [mg/l], csw the surface water concentration at the time of infiltration [mg/l] and cgw [mg/l] is the groundwater concentration. Removal is calculated as (i) the absolute and (ii) the relative difference between the concentration at the time of infiltration and the observed groundwater concentration.

2.6.

Redox classification

The predominant electron acceptors are used to classify the þþ ’ ‘Feþþ and redox state for each sample: ’O2‘, ‘NO 3 ’, ‘Mn ‘mixed’ with threshold values adapted from McMahon and Chapelle (2008). Samples classified as ‘mixed’ are not in equilibrium and exhibit electron acceptors of different zones simultaneously. The current classification only refers to the redox zone of the groundwater at sampling time, nevertheless some sampled water may have previously passed through other redox conditions, which is discussed in chapter 3.4.

2.7.

Statistical evaluation

Previous investigations have stated that in general, the removal of trace organics depends on redox conditions (Stuyfzand, 2007) and frequently also on field site characteristics (Schmidt, 2005). Furthermore, the calculated removal may also depend on how the values below the LOQ are treated. To account for these effects, the mean removal is calculated for different classifications, according to Eq. (2). Ri is the mean removal of class i, j is the counter of the class members (see below). P Ri ¼ P

Rabs;j

j

csw;j  100

(2)

j

Removal for the following classifications are calculated: þþ ’, ‘Feþþ’,
five classes of redox conditions ‘O2’, ‘NO 3 ’, ‘Mn mixed
four classes field site specifications: all data, Tegel BF, Tegel AR and Wannsee BF
two methods for regarding the limit of quantification (LOQ): for concentrations cgw < LOQ then either cgw ¼ 0 or cgw ¼ LOQ=2

This approach results in 20 classes with two values for each method regarding values below LOQ. In cases when the treatment of LOQ leads to a difference of less than 10%, the mean of both values is provided, when it is larger than 10% a range of values is provided, if the difference is higher than 50% the trend is described in words. The results for the first four redox classes for data of all field sites are presented in Table 2.

4943

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 3 9 e4 9 5 0

Table 2 e Observed removal efficiency of 38 organic parameters, given as percentage of surface water concentrations. Text entries are used if most data points are close to LOQ or few data points are available. Retardation values without asterisk denote literature derived values, one asterisk denotes an estimation based on structural similarity and physical properties. Two asterisks denote the initial assumption with a value of one. The right column indicates other factors which may affect removal: T: time, Th: threshold value, S: site specific removal, M: Metabolite. For details of the latter refer to Table 3. Removal by redox zone [%] O2 Sum parameters

Individual substances

AOBr (BWB) AOBr (TU-Org) AOI (BWB) AOI (TU-Org) AOX (BWB) DOC (BWB) DOC (TU-Org) SUVA (BWB) SUVA (TU-Org) UV_254 (BWB) UV_254 (TU-Org) 1,5 NDSA (TU-Org) 1,7 NDSA (TU-Org) 2,7-NDSA (TU-Org) AAA (BWB) AMDOPH (BWB) AMDOPH (TU-Drug) AMPH (BWB) Bentazone (TU-Drug) Bezafibrate (TU-Drug) Carbamazepine (BWB) Carbamazepine (TU-Drug) Clarithromycin (TU-Drug) Clindamycin (TU-Drug) Clofibric acid (TU-Drug) D-Erythromycin (TU-Drug) Diclofenac (TU-Drug) DP (BWB) EDTA (BWB) FAA (BWB) Indometacine (TU-Drug) lopromide (Tu-Org) Mecoprop (TU-Drug) MTBE (BWB) NPS (TU-Drug) p,p’-DDA (TU-Drug) PDP (BWB) Phenazone (BWB) p,p’-DDA (TU-Drug) Primidone (TU-Drug) Propyphenaz. (BWB) Propyphenaz. (TU-Drug) Roxithromycin (TU-Drug) Sulfadimidine (TU-Drug) Sulfamethox. (TU-Drug) Sulfamethox. (TU-Org) Trimethoprim (TU-Drug)

NO3

Mn

Fe

Retadation factor

Non-redox impact

16 23 37 78 17 32 59 68 4 13 38 58 8 31 56 63 17 25 36 58e84 29 28 29 26 32 35 34 31 3 3 8 12 1 9 9 9 30 25 20 16 32 32 27 24 6 4 3 6 47 57 29 20 54 47 28 22 95 98 90 56 25 4 4 6 26 1 10 1 increase (Mn, Fe) some removal (O2, NO3) between 22 and 88%, redox discontinous (O2, NO3, Mn, Fe) some removal (O2, NO3, Mn), less for Fe conditions 16 8 23 51

1** 1** 1** 1** 1** 1** 1** 1** 1** 1** 1** 1.05* 1.05 1,05* 1.8* 1.5* 1.5* 1.8* 1.1 1.5* 1.4

14

8

20

50

1.4

>96

>96

>96

>96

1**

95

97

98

27

1**

T,S

70

66

90

51

1.3

T

>97

>97

>97

>97

1**

T

1.4 1** 1.2 1.8* 1**

T M S T,M T S

91 85 some removal (O2, NO3) 11 6 94 93 some removal (O2, NO3, Mn, Fe)

61 61 increase (Mn, Fe) 14 15 less than O2/NO3 (Mn, Fe)

94 95 99 53e69 19 81e93 26 22 20 slight removal (O2, NO3, Mn, Fe) apparently persistent (O2, NO3, Mn, Fe) all groundwater values < LOD (O2, NO3, Mn) 91 73 46 slight removal (O2, NO3, Mn, Fe) 9 2 4 >58 56e95 42e79 84 67 48

9 0 to 54 9

1.1 1.1 1.05 1** 1** 1** 1 1** 1.2 2 2

>99

>99

1.3*

>99

>99

70 39 24

occurrence (Fe) 6

Th,S Th,S Th T,Th T,Th

T,Th T,Th Th Th T,M T,Th,M T,Th,M M T T,Th,S T,Th,S

S

M

S S

sometimes increase (O2, NO3)

often increase (Mn, Fe)

1**

41

47

74

89

1.05

T,S

31 >95

40 >95

77 >95

86 >95

1.05 1**

S

4944

2.8.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 3 9 e4 9 5 0

Graphical evaluation

The graphical evaluation is carried out for the same classes as the statistical evaluation. Values below LOQ are plotted as provided, either as LOQ/2 or as 0. Data are visualised by two scatter plots of concentration at time of infiltration csw and groundwater concentration cgw, both with the respective flow time as abscissa (e.g. Fig. 1). Different colours allow differentiating the three field sites. The LOQ is indicated as a horizontal line. The time dependent removal can be quantified with a halflife approach: cgw ¼ cthr þ ðc0  cthr Þ0:5t=t1=2 cgw ¼ c0

for c0 > cthr for c0  cthr

(3)

where cgw [mg/l] is the groundwater concentration, c0 [mg/l] is the concentration at the time of infiltration, cthr [mg/l] is a residual threshold concentration below which no removal occurs. t [d] is the flow time and t1=2 [d] is the half-life. Some organic water parameters show a removal during infiltration, which is faster than the shortest determined flow time. The removal of these can be better quantified with a linear regression of the form cgw ¼ c0 ða  btÞ

A linear removal in the subsurface is chosen because it empirically reflects the slope of the observed concentrations, but extrapolation might not be possible, for longer times an exponential approach may be more appropriate. The respective units differ for the parameters DOC (Dissolved Organic Carbon), UV254 (UltraViolet adsorption at 254 nm) and SUVA (Specific Ultraviolet Absorption at 254 nm). The approach of removal calculation is chosen to mitigate the impact of scattering and inaccuracies. It would be possible to calculate a removal percentage for each pair of surface and groundwater concentration, but the mean of the percentage value would be biased for low surface water concentrations and not reflect the net removal.

(4)

where a [-] is the fraction of removal that occurs rapidly during infiltration and b [1/d] defines the time dependent component.

3.

Results and discussion

The number of samples that are available for interpretation is reduced for two reasons: First, several samples are subject to mixing processes and second, for organic water parameters with high residence times the corresponding concentration at the time of infiltration is prior to the first sampling campaign and therefore no removal can be calculated. For these reasons 29% of the original data is not used. A number of 9992 concentration values remains for interpretation of which 42% þþ ’ belong to ‘O2’ conditions, 6% to ‘NO 3 ’ conditions, 7% to ’Mn þþ conditions and 14% to ’Fe ’ conditions. The latter condition

Fig. 1 e Removal characteristic for Carbamazepine (BWB) for oxic groundwater conditions. The plots show a) surface water concentration, b) observed groundwater concentration and c) simulated groundwater concentration (according to Eq. (3)). Each marker represents a sample. The envelope corresponds to a half life of 35 days with a threshold concentration of 0.3 mg/l.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 3 9 e4 9 5 0

4945

Table 3 e Compounds for which removal is impacted by factors other than the redox zone (Table 2). For parameters of the temporal kinetic refer to Eq. 3 and 4. Residual threshold concentrations are values below which no removal occurs. Site specific behaviour is related to significant differences in removal for at least one of the field sites and redox zone.

hardly occurs at Tegel AR (1 sample). The mixed redox state comprises 31% of the concentration values and is only used for consistency check. The results are of the combined statistical and graphical evaluation are summarized in Table 2. Results show that four parameters are readily removable (clarithromycin, D-erythromycin, roxithromycin, trimethoprim, (removal efficiency >95%) and 7 parameters are persistent (1,5-NDSA, bezafibrate, indometacine, NPS, p,p’DDA, sulfadimidine, removal efficiency < 30%). For all parameters which are in-between and only partly removed, the magnitude of removal depends on the redox conditions. The removal is usually monotonic with regard to redox, i.e. it either increases or decreases continuously. Hence, the number of samples in each redox class is generally large enough to smooth statistical noise

sufficiently. Minor discontinuities exist for some parameters (DOC, MTBE) but they are not regarded as significant.

3.1.

Redox conditions

A significant redox discontinuity exists for clofibric acid and mecoprop that is interpreted as a cross correlation effect to flow time. The similar chemical structure allows a comprehensive interpretation of both compounds’ behaviour (see below). Using the graphical interpretation it is determined whether other factors superpose the redox-dependent removal and the removal efficiency as presented in Table 2 may not be a predictive value. Graphical interpretation yields that for 20 compounds removal is affected by at least (i) residence time, or (ii) a residual threshold concentration, or (iii) site

4946

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 3 9 e4 9 5 0

characteristics (Table 3). These specifics are mainly determined for oxic conditions, for other redox conditions often insufficient data are available. Therefore, a missing entry in the respective column indicates that the particular behaviour could not be observed within the present dataset, but not necessarily that it can be ruled out.

3.2.

Determination of kinetics

Removal kinetics and potential cross correlation effects are also determined by means of the graphical evaluation. Three examples (Figs. 1e3) illustrate how removal kinetics are determined. Example 1: Carbamazepine (measured by BWB, Table 1) is removed by 16% under oxic conditions (Fig. 1, Table 2). Groundwater concentrations decrease with time, which can be described by an envelope with a T1/2 of 35 days and a cthr of 0.3 mg/l (Eq. (2)), Table 3). These parameters are also applied to simulate the groundwater concentrations starting from the surface water concentrations. The simulation results (Fig. 1c) are very similar to observed groundwater concentrations (Fig. 1b). Example 2: The removal of carbamazepine under ‘Feþþ’ reducing conditions is about 50%. The high removal is a combination of two effects: The mean flow time is 54 days (compared to 23 days for oxic conditions) and therefore removal is more efficient. Furthermore, the lower redox conditions enhance removal. While the mean flow time is similar for ‘Mnþþ’ and for ‘Feþþ’ reducing conditions, the removal increases from 20% to 50%. For ‘Feþþ’ reducing conditions the residual threshold is lower, i.e. groundwater concentrations down to values of 0.1 mg/l are observed (Fig. 2). The half life appears to be identical to oxic conditions. Example 3: The removal efficiency of sulfamethoxazole (measured by TU Drug) for oxic conditions is 41% (Fig. 3). Concentrations for long flow times are significantly lower than for shorter flow times. The removal characteristics can be described by an envelope that follows an 1st order decay with a half life of 30 days. Furthermore, calculating the decay of each observed surface water concentration applying Eq. (3) results in a point cloud with similar properties to the observed concentrations (compare Fig. 3b and c). The removal efficiency of sulfamethoxazole (measured by TU-Org) for oxic conditions is 31% and 10% points lower compared to the values measured by TU Drug. The difference can be explained by the kinetics which are determined. The removal efficiency is lower because the samples have a significantly lower flow time with a mean of 9 days compared to 26 days for the samples measured by TU Drug (i.e. different samples were measured). Results of both groups show minimum concentrations of about 0.05 mg/l, which could be an indication for a residual threshold concentration. Nevertheless this hypothesis cannot be validated for the present data, because the data can be described sufficiently by an exponential decay. The removal of sulfamexthoxazole for ‘Mnþþ’ and ‘Feþþ’ reducing conditions is 74% and 89%, respectively and much higher compared to ‘O2’ and ‘NO 3 ’ reducing conditions. Partly this can be attributed to flow times that are double as long, but also the removal processes are more effective.

Removal of sulfamethoxazole shows complex kinetics, Baumgarten et al. (2011) found residual threshold concentrations of 0.12 mg/l for source concentrations between 0.24 and 0.36 mg/l. For higher source concentrations the residual threshold concentration decreases, which is interpreted as result of microbial removal. Complementary, we observe considerable removal of 0.2 and 0.5 mg/l at Tegel BF and a no removal (i.e. residual threshold) of source concentrations around 0.12 mg/l at Wannsee BF. On the other hand the column studies show that removal is less efficient for lower redox conditions (Baumgarten et al., 2011), while we find it more efficient at the field sites (Table 2). This might be result of adaptation time, which are already more than two years for oxic conditions and might be even longer for lower redox. For the other compounds and sum parameters the removal characteristics are determined analogously (Table 2 and Table 3). For 10 compounds most groundwater concentrations are below LOQ or only few values are available, thus the removal cannot be calculated reliably. In these cases the trend is given verbally.

3.3.

Interpretation of residual threshold concentration

While the removal efficiency can be determined quite straightforward, the underlying mechanisms are much more difficult to determine and frequently remain unknown. Remaining initial concentrations from abiotic reactions tend to reach equilibrium and therefore tend to be proportional to the initial concentrations. A residual threshold concentration can be observed when microbial energetic use requires a minimum initial concentration (Schlegel, 1992). Therefore the existence of a residual threshold concentration is a hint for microbial activity which may imply energetic use of the respective compounds (Baumgarten et al., 2011). In case the compound is a cosubstrate in a microbial process, residual concentrations of the co-substrate can remain when the primary substrate is depleted. In this case a residual threshold concentration would be the result of enzymatic activity, that requires a minimum concentration of reagents. The environment has very low concentrations of biodegradable organic carbon. After the first metre, the DOC consumption is low and around 5e10 mg/d. Considering that DOC consists of a high number of different compounds which are removed, it appears possible that specialized bacteria exist which are able to use low concentrations of trace organics as energy source. The aquifer has been exposed to the trace organics during many years, which may promote microbiological specialisation. For SMX this behaviour has been observed by Baumgarten et al. (2011). The observed residual threshold conentrations support the hypothesis that microbial activity promotes removal. However, further research is necessary for to support the hypothesis and determine type and pathway. Several compounds with slow time dependent removal are quite polar (AAA, AMDOPH, clofibric acid, FAA, SMX), therefore their retardation may not be caused by classical adsorption but by more complex processes such as surface complexation. In this case sorption would potentially be highly variable since it would be affected by pH, redox and the availability of free

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 3 9 e4 9 5 0

4947

Fig. 2 e Removal characteristic for Carbamazepine (BWB) for FeDD groundwater conditions. The plots show a) surface water concentration, b) observed groundwater concentration and c) simulated groundwater concentration (according to Eq. (3)). Each marker represents a sample. The envelope corresponds to a half life of 35 days with a threshold concentration of 0.1 mg/l.

hydroxide or oxide surfaces. But without biological removal, complex sorption may induce a time shift only and no net decrease on a long term. Nevertheless, a net decrease is always observed. The long observation duration of 26 month with a maximum flow time of three month (for the relevant samples) strongly suggest, that this is representative and not just an intermediate effect. It may be possible, that surface complexation is followed by biodegradation, due to better adaptation and higher local concentrations of trace organics. It may also be induced by competitive adsorption, which leads to desorption effects. There are more basic studies needed to determine the extent of these combined mechanisms of adsorption and subsequent biodegradation.

3.4.

Cross correlation effects

Frequently removal efficiencies are expressed as percentage as mean value for a number of samples. Redox conditions have the most important impact, but additional factors such as flow time, residual threshold values or field site specifics may contribute to removal or even exceed the relevance of redox conditions. In the following examples are discussed, where spurious correlations bias an interpretation which is only based on redox conditions, and removal kinetics have to be considered. This analysis implies that the main residence time is in the sampled redox zone itself, which is generally true for the field

sites. Water infiltrating at deeper parts of the lakes exhibits lower redox conditions than water infiltrating at shallow shore-line areas. This was shown by Wiese and Nu¨tzmann (2009) for Tegel BF and by Massmann et al. (2008b) for Wannsee BF. This pattern continues in the groundwater, where redox zones are stratified and the general flow is horizontal (Massmann et al., 2008b). At Tegel BF a large unsaturated zone exists which partly contains atmospheric oxygen (Wiese and Nu¨tzmann, 2009) so that the flow path may include a passage of considerable length under oxic conditions, while the observed redox state is lower in the sample. This is relevant for clindamycine and iopromide. Both parameters are readily removable (>94%) with a half life of less than 3 days for ’O2’ to ’Mnþþ’ redox conditions at all field sites and groundwater concentrations rapidly fall below LOQ. For iron reducing conditions, however, they show apparent removal efficiencies of 97% and 96% at Tegel BF, but only 27% and 70% at Wannsee BF, respectively. At Tegel BF, a short residence time under more oxidizing conditions, distorts the observed removal for ’Feþþ’ reducing samples. Consequently, for both compounds, the removal efficiency for iron reducing conditions in Table 2 is not the mean of all sites but only refers to Wannsee BF. Less pronounced hints exist for other parameters such as 1,7 naphthalenedisulfonic acid (NDSA), diclofenac and sulfamethoxazole (SMX). The opposite effect may occur when oxic water previously has passed an area with lower redox

4948

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 3 9 e4 9 5 0

Fig. 3 e Removal characteristic for sulfamethoxazole (TU Drug) for oxic groundwater conditions. The plots show a) surface water concentration, b) observed groundwater concentration and c) simulated groundwater concentration (according to Eq. (3)). Each marker corresponds represents a sample. The envelope corresponds to a half life of 30 days.

conditions (McMahon and Chapelle, 2008). This is not considered to be relevant, as only 4 oxic samples show nitrate concentration <0.5 mg/l. Removal characteristics (Table 3) can only be discussed exemplarily due to restricted article length. Carbamazepine shows highest removal efficiency at Tegel BF (14% ‘O2’ increasing to 58% for ‘Feþþ’), while the maximum removal at Tegel AR and Wannsee BF is 20% and sometimes removal even shows negative values. Negative removal occurs only for classes with less than 5 members. Removal at Tegel BF is highest because carbamazepine shows time dependent removal (t1/2 ¼ 35 days) and the mean flow times for the redox classes vary between 27 day for ‘O2‘and 105 days for ‘Mnþþ’ conditions. Flow times for individual redox zones vary between 6 and 29 days At Tegel AR and between 12 and 24 days at Wannsee BF. In addition, carbamazepine shows residual threshold concentrations between 0.1 and 0.3 mg/l. Therefore, in addition to the flow time effect, the removal efficiency also increases with higher surface water concentration, which is around 1 mg/l at Tegel BF and 0.3 mg/l at Wannsee BF. This is one example where removal efficiencies are field site specific, but this can be attributed to removal kinetics which are generally valid. For clofibric acid (2-(4-Chlorophenoxy)-2-methylpropanoic acid) the highest removal occurs in the ‘Mnþþ’ zone which may be explained by a cross correlation effect with flow time. For oxic conditions time dependent removal is identified, but there are not enough values to determine kinetics for ‘NO 3’

and ‘Mnþþ’ reducing conditions. Nevertheless, such a kinetic appears reasonable. The ‘Mnþþ’ zone shows a mean flow time of 62 days, while the ‘O2’ and ‘NO 3 ’ zone only show 26 and 18 days, respectively and therefore a kinetic removal of clofibric acid under ‘Mnþþ’ is a reasonable explanation for the high removal. The mean flow time for ‘Feþþ’ reducing conditions is 55 days, hence the removal cannot be explained by flow time and therefore is least effective. Mecoprop ((RS)-2-(4-chloro-2-methylphenoxy)propanoic acid) shows similar behaviour as clofibric acid, with highest removal efficiencies for the zones ‘O2’ and ‘Mnþþ’, and longest flow times for the ‘Mnþþ’ zone. Hints exist for a time dependent removal kinetic, but it cannot be determined since the data quality is not as good as for clofibric acid, (the number of data points is 134 and 182, the ratio of mean surface water concentration to LOQ is 3 and 4, respectively). Both compounds have a very similar chemical structure. Compared to clofibric acid mecoprop has a methyl group at the aromatic ring, the side chain is identical. For both compounds the reactivity of the aromatic ring is low, so it is expected that the removal of both compounds occurs due to the same side chain reaction. Consequently and analogue to mecoprop, the highest removal for the ‘Mnþþ’ zone is interpreted mainly as cross correlation effect of highest flow time and temporal kinetic. However, the removal of mecoprop appears to be slightly lower than that of clofibric acid. This may be based in steric interference of the methyl group. Some

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 3 9 e4 9 5 0

indications exist that a faster removal under ‘Mnþþ’ conditions contributes to high removal rates. Samples with a flow time less than 25 days show >95 removal for clofibric acid (6 samples) and >79% removal for mecoprop (4 samples). However, the sample numbers are too low for a conclusive interpretation and from the geochemical point of view it appears improbable that removal is most effective for ‘Mnþþ redox conditions. Further research is necessary at this point. Propyphenazone concentrations are elevated under iron reducing conditions. This behaviour only appeared at the field site Tegel BF, where some samples with slight mixing must have been overlooked (Table 2, Table 3). For most parameters site specific removal can be explained by the abovementioned mechanisms. Only for EDTA and MTBE no conclusive interpretation has been found. For the six metabolites it has to be taken into account that their concentration may increase in the groundwater. The presented method allows to identify removal characteristics and to reduce uncertainty by eliminating ambiguous data. An uncertainty analysis is facilitated by statistical numbers and graphical representation, which are to be interpreted with expert knowledge. The determination of nonredox removal characteristics includes a subjective component. The assumption of redox-dependent removal is reasonable for all partly removable compounds but this assumption alone frequently does not allow a conclusive interpretation because the majority of compounds and parameters are also affected by factors other than redox. The removal process can be represented more accurately when time dependent kinetics and residual threshold concentrations are considered and the results have a relevantly higher predictive value.

4.

Conclusions

A method is developed which allows analysing the removal kinetics during bank filtration of organic water parameters. It is applicable to large data bases. The residence time and initial surface water concentration for each groundwater concentration is determined and the removal is calculated. This allows different types of evaluation and compared to previous approaches the data may be aggregated at a later stage. For a general overview removal efficiencies are evaluated for each substance under different redox conditions. The core innovation of the method is that different removal kinetics are distinguished and determined from the data itself so there is no need to assume kinetics a priori. It can be shown that these kinetics affect the removal for the majority of the water parameters. Thereby it is possible to interpret why removal efficiencies vary at different field sites.
The analysed organic parameters show removal efficiencies between 0% and >99%.
Redox conditions are a principal factor of impact for the removal since all partly removable parameters show a pronounced dependence on the redox conditions.
However, in many cases redox conditions alone are not sufficient to predict removal (Table 3). Removal kinetics frequently account for different removal efficiency at different field sites.

B B

B

4949

For 15 parameters the removal increases with flow time. Nine parameters show a residual threshold value below which removal becomes insignificant. To the authors knowledge it is the first study where residual threshold values are determined from field data. Three parameters show site specific behaviour that cannot be explained by removal kinetics.

The existence of residual threshold concentrations for several compounds is a hint for microbial removal activity. Nevertheless, further research is needed to determine type and pathway. Removal in this study only means that the analysed compounds can no longer be determined. Metabolites are not considered and should be taken into account when conducting an overall risk assessment.

Acknowledgements We thank the Berliner Wasserbetriebe (BWB) and Veolia Water for funding the NASRI (2002e2005) and IC-NASRI (2007e2009) project. Many thanks to Uwe Hu¨bner (TU Berlin) for proof reading.

references

Baumgarten, B., Ja¨hrig, J., Reemtsma, T., Jekel, M., 2011. Long term laboratory column experiments to simulate bank filtration: factors controlling removal of sulfamethoxazole. Water Research 45 (1), 211e220. Braids, O.C., 2001. MTBE - Panacea or problem. Environmental Forensics 2 (3), 189e196. Broholm, M.M., Rugge, K., Tuxen, N., Hojberg, A.L., Mosbaek, H., Bjerg, P.L., 2001. Fate of herbicides in a shallow aerobic aquifer: A continuous field injection experiment (Vejen, Denmark). Water Resources Research 37 (12), 3163e3176. Buss, S.R., Thrasher, J., Morgan, P., Smith, J.W.N., 2006. A review of mecoprop attenuation in the subsurface. Quarterly Journal of Engineering Geology and Hydrogeology 39, 283e292. Denecke, E., 1997. Evaluation of long-term measurements concerning the aerobic degradation performance of the subsoil passage of a water catchment at the lower Rhine. Acta Hydrochimica et Hydrobiologica 25 (6), 311e318. Gelhar, L.W., Welty, C., Rehfeldt, K.R., 1992. A critical review of data on field-scale dispersion in Aquifers. Water Resources Research 28 (7), 1955e1974. Greskowiak, J., Prommer, H., Massmann, G., Johnston, C.D., Nu¨tzmann, G., Pekdeger, A., 2005. The impact of variably saturated conditions on hydrogeochemical changes during artificial recharge of groundwater. Applied Geochemistry 20 (7), 1409e1426. Greskowiak, J., Prommer, H., Massmann, G., Nu¨tzmann, G., 2006. Modeling seasonal redox dynamics and the corresponding fate of the pharmaceutical residue phenazone during artificial recharge of groundwater. Environmental Science and Technology 40 (21), 6615e6621. Grischek, T., Neitzel, P., Andrusch, T., Lagois, U., Nestler, W., 1997. Verhalten von EDTA bei der Untergrundpassage und Ausweisung von Infiltrationsprozessen an der Elbe. Vom Wasser 89, 261e282.

4950

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 3 9 e4 9 5 0

Gru¨nheid, S., Amy, G., Jekel, M., 2005. Removal of bulk dissolved organic carbon (DOC) and trace organic compounds by bank filtration and artificial recharge. Water Research 39 (14), 3219e3228. Hartig, C., 2000. Analytik, Vorkommen und Verhalten aromatischer Sulfonamide in der aquatischen Umwelt. (Analysis and behaviour of aromatic sulfonamides in the aquatic environment). PhD Thesis, Technical University of Berlin, School of Process Sciences and Engineering. http:// opus.kobv.de/tuberlin/volltexte/2000/131/. Heberer, T., Jekel, M., 2006. Occurrence and fate of drug residues and related polar contaminants during bank filtration and artificial recharge. Technical Report. Heberer, T., Massmann, G., Fanck, B., Taute, T., Du¨nnbier, U., 2008. Behaviour and redox sensitivity of antimicrobial residues during bank filtration. Chemosphere 73 (4), 451e460. Hiscock, K.M., Grischek, T., 2002. Attenuation of groundwater pollution by bank filtration. Journal of Hydrology 266 (3e4), 139e144. Jekel, M., 2006. Organic substances in bank filtration and groundwater recharge - Process studies. Technical Report. Jørgensen, P.R., McKay, L.D., Spliid, N.H., 1998. Evaluation of chloride and pesticide transport in a fractured clayey till using large undisturbed columns and numerical modeling. Water Resources Research 34 (4), 539e553. Kohfahl, K., Massmann, G., Pekdeger, A., 2009. Sources of oxygen flux in groundwater during induced bank filtration at a site in Berlin, Germany. Hydrogeology Journal 17, 571e578. KWB, 2009. Natural and artificial systems of recharge and infiltration, revised project database. Technical Report. Licht, E., Wiese, B., Heberer, T., Gru¨tzmacher, G., 2005. Estimating of the solute transport parameters retardation factor and decay coefficient of pharmaceutical residues using the program Visual CXTFIT. 1, ISMAR - 5th international symposium on management of aquifer recharge. Massmann, G., Greskowiak, J., Du¨nnbier, U., Zuehlke, S., Knappe, A., Pekdeger, A., 2006. The impact of variable temperatures on the redox conditions and the behaviour of pharmaceutical residues during artificial recharge. Journal of Hydrology 328 (1e2), 141e156. Massmann, G., Du¨nnbier, U., Heberer, T., Taute, T., 2008a. Behaviour and redox sensitivity of pharmaceutical residues during bank filtration - Investigation of residues of phenazone-type analgesics. Chemosphere 71 (8), 1476e1485. Massmann, G., Nogeitzig, A., Taute, T., Pekdeger, A., 2008b. Seasonal and spatial distribution of redox zones during lake bank filtration in Berlin, Germany. Environmental Geology 54 (1), 53e65. Massmann, G., Su¨ltenfuß, J., Du¨nnbier, U., Knappe, A., Taute, T., Pekdeger, A., 2008c. Investigation of groundwater residence times during bank filtration in Berlin: a multi-tracer approach. Hydrological Processes 22 (6), 788e801. McMahon, P.B., Chapelle, F.H., 2008. Redox processes and water quality of selected principal aquifer systems. Ground Water 46 (2), 259e271. Mersmann, P., Scheytt, T., Heberer, T., 2003. Column experiments on the transport behavior of pharmaceutically active compounds in the saturated zone (Sa¨ulenversuche zum Transportverhalten von Arzneimittelwirkstoffen in der

wassergesa¨ttigten Zone). Acta Hydrochimica et Hydrobiologica 30 (5e6), 275e284. Meyer, H., Scho¨nicke, L., Wand, U., Hubberten, H., Friedrichsen, H., 2000. Isotope studies of hydrogen and oxygen in ground ice - experiences with the equilibration technique. Isotopes and Health Studies 36 (2), 133e149. Putschew, A., Schittko, S., Jekel, M., 2001. Quantification of triiodinated benzene derivatives and X-ray contrast media in water samples by liquid chromatography-electrospray tandem mass spectrometry. Journal of Chromatography A 930, 127e134. Reddersen, K., Heberer, T., 2003. Multi-compound methods for the detection of pharmaceutical residues in various waters applying solid phase extraction (SPE) and gas chromatography with mass spectrometric (GC-MS) detection. Journal of Separation Science 26 (15e16), 1443e1450. Reddersen, K., Heberer, T., Dunnbier, U., 2002. Identification and significance of phenazone drugs and their metabolites in ground- and drinking water. Chemosphere 49 (6), 539e544. Scheytt, T.J., Mersmann, P., Heberer, T., 2006. Mobility of pharmaceuticals carbamazepine, diclofenac, ibuprofen, and propyphenazone in miscible-displacement experiments. Journal of Contaminant Hydrology 83 (1e2), 53e69. Schlegel, H.G., 1992. Allgemeine Mikrobiologie. Thieme, Stuttgart, New York. Schmidt, C., 2005. Datenbank zum Verhalten organischer Spurenstoffe bei der Uferfiltration. DVGWTechnologiezentrum Wasser (TZW). Technical Report. Storm, T., Reemtsma, T., Jekel, M., 1999. Use of volatile amines as ion-pairing agents for the high-performance liquid chromatographic-tandem mass spectrometric determination of aromatic sulfonates in industrial waste water. Journal of Chromatography, A 854, 175e185. Stuyfzand, P., Segers, W., Rooijen, N., 2007. Behavior of Pharmaceuticals and Other Emerging Pollutants in Various Artificial Recharge Systems in the Netherlands. Phoenix, Arizona, USA. ISMAR. Tuxen, N., Tuchsen, P.L., Rugge, K., Albrechtsen, H.J., Bjerg, P.L., 2000. Fate of seven pesticides in an aerobic aquifer studied in column experiments. Chemosphere 41 (9), 1485e1494. Wiese, B., 2006. Spatially and temporally scaled inverse hydraulic modelling, multi tracer transport modelling and interaction with geochemical processes at a highly transient bank filtration site. PhD Thesis, Humboldt-University Berlin. Wiese, B., Nu¨tzmann, G., 2009. Transient Leakance and infiltration characteristics during Lake bank filtration. Ground Water 47 (1), 57e68. Wiese, B., Nu¨tzmann, G., in press. Calibration of spatial aquitard distribution using hydraulic head changes and regularisation. Journal of Hydrology. doi:10.1016/j.jhydrol.2011.07.015. Zu¨hlke, S., Du¨nnbier, U., Heberer, T., 2004. Detection and identification of phenazone-type drugs and their microbial metabolites in ground and drinking water applying solidphase extraction and gas chromatography with mass spectrometric detection. Journal of Chromatography A 1050 (2), 201e209.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 6 0 e4 9 7 2

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Fate of the pathogen indicators phage FX174 and Ascaris suum eggs during the production of struvite fertilizer from source-separated urine Loı¨c Decrey a,b, Kai M. Udert c, Elizabeth Tilley b, Brian M. Pecson a, Tamar Kohn a,* a

Ecole Polytechnique Fe´de´rale de Lausanne (EPFL), Environmental Chemistry Laboratory, School of Architecture, Civil and Environmental Engineering (ENAC), 1015 Lausanne, Switzerland b Department of Water and Sanitation in Developing Countries, Swiss Federal Institute of Aquatic Science and Technology (Eawag), 8600 Du¨bendorf, Switzerland c Department of Process Engineering, Swiss Federal Institute of Aquatic Science and Technology (Eawag), 8600 Du¨bendorf, Switzerland

article info

abstract

Article history:

Human urine has the potential to be a sustainable, locally and continuously available

Received 10 February 2011

source of nutrients for agriculture. Phosphate can be efficiently recovered from human

Received in revised form

urine in the form of the mineral struvite (MgNH4PO4$6H2O). However, struvite formation

30 June 2011

may be coupled with the precipitation of other constituents present in urine including

Accepted 30 June 2011

pathogens, pharmaceuticals, and heavy metals. To determine if struvite fertilizer presents

Available online 18 July 2011

a microbiological health risk to producers and end users, we characterized the fate of a human virus surrogate (phage FX174) and the eggs of the helminth Ascaris suum during

Keywords:

a low-cost struvite recovery process. While the concentration of phages was similar in both

Struvite

the struvite and the urine, Ascaris eggs accumulated within the solid during the precipi-

Urine separation

tation and filtration process. Subsequent air-drying of the struvite filter cake partially

Ascaris

inactivated both microorganisms; however, viable Ascaris eggs and infective phages were

Phage

still detected after several days of drying. The infectivity of both viruses and eggs was

Filter cake

affected by the specific struvite drying conditions: higher inactivation generally occurred

Moisture content

with increased air temperature and decreased relative humidity. On a logelog scale, phage inactivation increased linearly with decreasing moisture content of the struvite, while Ascaris inactivation occurred only after achieving a minimum moisture threshold. Sunlight exposure did not directly affect the infectivity of phages or Ascaris eggs in struvite cakes, though the resultant rise in temperature accelerated the drying of the struvite cake, which contributed to inactivation. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Recovering nutrients from human excreta can help to meet fertilizer demands while simultaneously reducing the pollution of water by untreated human waste (Langergraber and Muellegger, 2005; Larsen et al., 2007). Human excreta are an

especially important source of agricultural nutrients in developing countries, where the lack of affordable fertilizer is a major contributor to food shortages (Sanchez, 2002). Collecting human waste, processing it and selling the valueadded products can furthermore provide economic opportunities (Kone, 2010). Since human excreta and untreated

* Corresponding author. Tel.: þ41 21 693 0891; fax: þ41 21 693 8070. E-mail address: [email protected] (T. Kohn). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.06.042

4961

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 6 0 e4 9 7 2

wastewater contain pathogens, pharmaceuticals and heavy metals, adequate pre-treatment is a necessity to prevent the spread of diseases and contamination (WHO, 2006; Winker et al., 2009). One efficient and simple way to harvest nutrients from human excreta is to separate the collection of urine and faeces. Most of the excreted nutrients are contained within the urine: 85e90% of nitrogen, 50e80% of phosphorus, 80e90% of potassium and close to 100% of sulphur (Larsen and Gujer, 1996). Urine of healthy people is typically sterile, but a few pathogens (e.g. Schistosoma haematobium, Salmonella typhi, Salmonella paratyphi, Leptospira interrogans) can be found in the urine of infected people (Feachem et al., 1983). In addition, feacal material, which can contain a wide range of pathogens such as bacteria, protozoa, viruses and parasitic worms (WHO, 2006), can cross-contaminate source-separated urine. This occurs particularly in the case of the misuse or the poor maintenance of urine-diverting toilets (UDTs). A Swedish study showed that 8 out of 36 (22%) urine samples and 11 out of 30 (37%) bottom-sedimented sludge samples taken from source-separated urine storage tanks were contaminated with feacal material (Schonning et al., 2002). In Kenya, up to 720 Ascaris eggs per litre were found in the urine of poorly maintained UDTs in a school (Kraft, 2010). The harsh chemical conditions of stored urine, in particular the high pH and ammonia concentration caused by urea hydrolysis (Udert et al., 2003), promote the inactivation of microorganisms. These conditions lead to high levels of free ammonia (NH3), a constituent known to be biocidal for most organisms (Cramer et al., 1983; Jenkins et al., 1998; Pecson and Nelson, 2005). Temperature also affects the inactivation of microorganisms in stored urine. Various studies have characterized the survival of different pathogens over a range of temperatures (Table 1). Because of the long survival times of some pathogens at ambient temperatures, particularly viruses and helminths (Table 1), stored urine would benefit from additional treatment before being used as a fertilizer. For most applications, urine treatment is also necessary to concentrate and stabilize the nutrients or to remove micropollutants (Lienert et al.,

2007; Udert et al., 2006). Maurer et al. (2006) compiled an overview of possible technologies for urine treatment, some of which are potentially suitable for implementation in developing countries (Pronk and Kone, 2009). The majority of these technologies, however, have only been tested at lab scale. The most studied and applied process for nutrient extraction from urine is struvite precipitation (Etter et al., 2011). Struvite (MgNH4PO46H2O) is an effective phosphorus fertilizer (Romer, 2006), which can easily be produced from hydrolyzed urine. Struvite production mainly aims at recovering phosphorus from urine, since only a small percentage of the ammonia in the initial urine solution is recovered (Etter et al., 2011). Some struvite spontaneously precipitates during urine storage due to urea hydrolysis (Udert et al., 2003), though more significant quantities can be generated by adding magnesium-containing minerals. Struvite crystallization is a fast process, taking only a few minutes (Etter et al., 2011). After precipitation, struvite is usually recovered by filtration and subsequently air-dried at ambient temperatures. Struvite should not be heated above 40e55  C due to the risk of ammonia loss, which leads to a reduction of the N content of the final fertilizer product (Bhuiyan et al., 2008; Frost et al., 2004). An additional step may be included to form the powder into granules that can be more easily applied as fertilizer (Etter et al., 2011). Ronteltap et al. (2007) investigated the fate of pharmaceuticals and heavy metals during struvite production. In their experiments, only a small fraction of the spiked hormones and pharmaceuticals were incorporated into the struvite; more than 98% remained in the urine. Furthermore, struvite produced from stored urine contained significantly lower concentrations of heavy metals than commercially available fertilizers. To date, the fate of pathogens during struvite production has not been investigated. With this study, we aim to fill this gap. In particular, we determined the survival of two commonly used and easy to manipulate surrogates of human pathogens, namely phage FX174 as a proxy for human viruses and Ascaris suum as surrogate of the human helminth Ascaris

Table 1 e Inactivation rate of different pathogens types in (un-)diluted stored urine at pH around 9. Species

T ( C)

Dilution (urine:water)

Average inactivation rate (T90)a

Bacteria

Enterococcus faecalis, Salmonella Typhimurium

1:0

Viruses

Phage FX174

34 24 4 34 24 4 20 5 20 5 34 24 4

<1 day 1e2 days 2e6 days <6 days 12 days 120 days 35 days NOTDb <7 days 29 days 3 daysc 48 daysc >480 daysc

Type

Rotavirus Protozoa Helminths

Cryptosporidium Parvum Ascaris suum eggs

a Time for 90% or 1 log10 inactivation. b NOTD ¼ No observable trend in decay. c T99 ¼ time for 99% or 2 log10 inactivation.

1:0

1:2 1:2 1:0

Reference (Vinneras et al., 2008)

(Vinneras et al., 2008)

(Hoglund et al., 2002) (Hoglund and Stenstrom, 1999) (Nordin et al., 2009)

4962

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 6 0 e4 9 7 2

lumbricoides, during struvite production and drying. By using two of the most resistant pathogen classes, our goal was to provide a conservative estimate of pathogen inactivation during struvite formation. The concentrations and infectivity of these indicator organisms were measured in the precipitated struvite and during the processing of the struvite filter cake. The effect of air temperature, relative humidity (RH), sunlight exposure, and filter cake thickness was determined. The struvite production procedure was in accordance with the methods developed by Etter et al. (2011) for low-cost struvite production in Nepal.

2.

Materials and methods

2.1.

Stored urine

Undiluted, stored urine was obtained from the urine storage tank at Eawag’s main building in Du¨bendorf (Switzerland), which collects urine from waterless urinals and men’s NoMix toilets. The residence time of urine in the storage tank was between 37 and 47 days (Goosse et al., 2009). Urine samples were withdrawn from the tank immediately before the start of the experiments. Throughout the duration of this study (April to October 2010), six urine samples were taken from the storage tank to assess its composition. No significant differences were observed between the samples; therefore the urine composition is presented as the average of all samples in Table 2. Phosphate, sulphate and chlorine concentrations were measured by ion chromatography (Column Metrosep A Supp 4, Metrohm, Herisau, Switzerland); magnesium, calcium, potassium and sodium by inductively coupled plasma optical emission spectrometry (ICP-OES, Ciros, Spectro Analytical Instruments, Kleve, Germany); total ammonia (NH3 þ NHþ 4 ) by flow injection analysis (Application Note 5520, FOSS, Hillerød, Denmark); chemical oxygen demand (COD) with cuvette tests (Hach-Lange, Berlin, Germany) and total inorganic carbon (TIC) by means of a TOC-TN Analyser (IL 550, Hach-Lange, Berlin, Germany).

Table 2 e Average composition (at 25  C) of six urine samples taken from the Eawag storage tank. Average  S.D. PH [-] PO4 [mM] NH4 tot [mM] Mg [mM] Caa [mM] Cl [mM] K [mM] Na [mM] SO4 [mM] TICa,b [mM] CODa,c [mgO2/L] a Average of five samples. b Total inorganic carbon. c Chemical oxygen demand.

8.8  0.1 6.4  0.7 186  15 <0.5 0.4  0.1 91  10 41  5 83  10 7.4  1.2 104  11 4684  553

2.2.

Struvite production

Struvite was produced in a batch wise operation according to the field process developed by Etter et al. (2011). Urine volumes of 1e4.5 L were stirred in a glass beaker at room temperature with a magnetic stirrer at approximately 400 rpm and were spiked with microorganisms according to the procedure described below. Magnesium chloride hexahydrate (MgCl26H2O) was added at a Mg:P molar ratio of 1:1.1 and mixed for 10 min. Given an average P concentration of 6.4 mM in the stored urine (Table 2), approximately 1.5 g MgCl26H2O was added per L of urine. After the magnesium addition, the urine was filtered through a nylon fabric with irregular pore diameters of 18e240 mm. This filter material was chosen to be consistent with the material used in a field study in Nepal (Etter et al., 2011). This allowed us to approximate the conditions of struvite production used in the field. The discharge rate during filtration was kept as constant as possible at values ranging between 5 and 50 mL/min. At the beginning of filtration, the urine was filtered by gravity alone; later, a vacuum pump was used to maintain the same discharge rate. At the end of the filtration process, a strong vacuum was applied to maximize the drying of the struvite. Struvite produced from 1 L of urine and dried under laboratory conditions (23  1  C and 43  5% RH) yielded 0.3 to 1.3 g of solid. The filter cakes had a diameter of 38 mm, though its thickness varied with the volume of urine used: the average thickness of a 1 L cake was 1.5  0.7 mm, 2.1  0.9 mm for a 3 L cake, and 3.9  0.4 mm for a 4.5 L cake. The thickness of the cake increased from the centre to the rim. Ascaris eggs and phages were added to the urine as indicator organisms for human pathogens prior to the struvite production (for details see Sections 2.4 and 2.5).

2.3.

Struvite cake drying experiments

After struvite production, phage and Ascaris egg survival was monitored as a function of struvite drying under different combinations of temperature and relative humidity (Table 3). The drying conditions were set in two different ways. For the 20  C experiments, samples were stored in temperature and humidity-controlled rooms. For the experiments at 5  C (cold room) and 35  C (incubator), the samples were placed in small desiccators filled at the bottom with saturated salt solutions, as described by Winston and Bates (1960). Using NaBr and LiCl salts, in conjunction with water-soaked paper towels as a moisture source, a relative humidity of 85% and 35%, respectively, could be maintained. Temperature and relative humidity were measured regularly with small combined thermometer/hygrometers (Piccolo, Irox, Switzerland) that were placed in the desiccators with the samples. The mean temperature and relative humidity were determined by a weighted mean of the measured data (weighted by the elapsed time between measurements). The struvite cakes were protected from light with aluminium foil during the drying phase, unless noted otherwise. Periodic triplicate samples were taken and enumerated for infective phages and viable Ascaris eggs during three days following the production of the struvite cake (see below). This corresponds to the time frame encountered in field

4963

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 6 0 e4 9 7 2

Table 3 e List of the experimental parameters tested for phage FX174 (F) and Ascaris suum eggs (A). No. A B C D.1 D.2 E.1 E.2 E.3 E.4 E.5 F G H

Shade/Sun Shade Shade Shade Shade Shade Shade Shade Shade Shade Shade Shade Shade Sun

Temperature 

5 C (0) 5  C (0) 20  C (0) 20  C (0) 20  C (0) 20  C (0) 20  C (0) 20  C (0) 20  C (0) 20  C (0) 36  C (1) 35  C (1) 31  C (3)

Relative humidity 35% (9) 85% (3) 42% (2) 77% (4) 77% (4) 90% (0) 93% (1) 93% (1) 93% (1) 93% (1) 36% (8) 85% (3) <35%

Urine filtered 1 1 1 1 3 1 1 3 1

L L L L L L L L L 4.5 L 1L 1L 1L

Day of sampling 0, 1, 3, 10 0, 1, 2, 3, 10a 0, 1, 2, 3 0, 1, 2, 3 0, 1, 2, 3 0, 1, 2, 3 0, 1, 2, 3 0, 1, 2, 3 0, 1, 2, 3, 9 0, 1, 2, 3, 9 0, 6 h, 1, 3 0, 1, 2, 3 0, 1h, 2h, 5h

Microorganisms spiked F F, F, F, F F, F F F F F, F, F,

A A A A

A A A

a Sampling after 10 days only for phages.

experiments, where the struvite is dried for two to three days in between filtration and further processing (Etter et al., 2011). In addition, we conducted experiments to assess the inactivation of phages during longer drying periods (9e10 days).

2.4.

Phage enumeration

FX174 (ATCC 13706-B1) and their host Escherichia coli (ATCC 13706) were purchased from the German Collection of Microorganisms and Cell Cultures (DSMZ, Braunschweig, Germany). Phages were propagated in liquid culture according to DSMZ instructions, and were enumerated by the double layer agar method. Infective phage numbers are reported in plaque forming units (pfu). A phage stock solution of approximately 109 pfu/mL was produced, passed through 0.45 mm pore size filters to remove bacterial debris, inoculated with streptomycin (final concentration 2 mg/L) to prevent further bacterial growth and stored at 4  C. 10 mL of this concentrated stock were spiked into 1 L of stored urine to achieve an initial concentration of 106e107 pfu/mL of urine. Phage concentrations were determined in the struvite filter influent, effluent and in the struvite filter cake. Prior to enumeration, samples were diluted in phosphate buffered saline (PBS; 5 mM NaH2PO4H2O, 10 mM NaCl) at pH 7.5. For the enumeration of phages in the influent and effluent, samples were directly diluted by PBS and plated. For the enumeration of phages in struvite, 50e100 mg of struvite were carefully removed from the filter cake in small pieces ranging from the centre to the rim, dissolved in 1 mL citrate buffer (41.5 mM C6H5Na3O72H2O, 9.5 mM C6H8O7H2O, 10 mM NaCl) at pH 5.7 by vortexing during 30 s, further diluted in PBS, and plated. Control experiments confirmed that this process recovered all the phages from the struvite, that the phages did not replicate or aggregate in stored urine, and that they were not inactivated in the citrate buffer. Phage inactivation is expressed in terms of concentration of infective phage over initial concentration of infective phage (C/C0) throughout the study.

a homogeneous spiking solution containing approximately 104 eggs/mL by stirring 2 mL into 8 mL of milliQ water. The resulting exact concentration was determined from three samples of 150 mL enumerated under a microscope using worm egg counting slide (McMaster, JA Whitlock & Co, Australia). One mL of the spiking solution was then transferred into 0.5 L of urine, which was subsequently processed to produce struvite. Control experiments determined the ratio of eggs adhering to the walls of the beaker and the struvite filter setup (described in the Supplementary information), to derive the actual number of eggs suspended in the urine in relation to the number of eggs initially added with the spiking solution. Eggs in the struvite were enumerated by dissolving 50e100 mg of struvite in 1 mL of 0.1 M H2SO4 and counting the eggs in 3  150 mL of this solution as described above. The egg numbers were extrapolated to 1 mL to determine the egg concentration in the struvite cake. The viability of the eggs in struvite was assessed by mixing 50e100 mg of the struvite filter cake produced from 1 L stored urine with 1 mL of 0.1 M H2SO4. This solution was centrifuged at 4000 rpm for 3 min to sediment the eggs and the supernatant was removed and replaced by new sulphuric acid. This process was carried out five times to dissolve the struvite, and to lower the pH to around 2 to avoid microbial growth to prevent the formation of free ammonia (100 mg struvite/ml corresponds to approximately 400 mM total ammonia) during the subsequent incubation. The sedimented eggs were incubated in the dark in uncapped tubes for 4 weeks at 28  C (Pecson and Nelson, 2005). After incubation, a solution aliquot was examined under a microscope (magnification 10x and 40x) to assess the viability of 200 eggs. Only eggs that had developed to the full larval stage were counted as viable. Ascaris egg inactivation is expressed in terms of measured concentration of viable eggs over initial concentration of viable eggs (C/C0) throughout the study.

2.6. 2.5.

Drying curve and moisture content of struvite

Ascaris suum enumeration

A stock of 106 Ascaris eggs (Excelsior Sentinel Inc., Ithaca, NY) stored at 4 C in 0.5% formalin, was used to obtain

To assess the cake’s moisture content during drying, unspiked struvite cakes were produced and dried at the same time and under the same conditions as the ones spiked with

4964

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 6 0 e4 9 7 2

microorganisms. The loss of mass from these cakes over time was monitored with a digital balance to obtain a drying curve. The microorganism concentrations in and the moisture content of a wet solid (e.g. sludge) are generally expressed in units per mass of dry material. However, struvite undergoes thermal decomposition starting at 40e55  C due to the volatilization of water and ammonia bound in the struvite crystals (Bhuiyan et al., 2008; Frost et al., 2004); therefore the exact dry mass could not be determined by heating the samples. Instead, struvite cakes were allowed to dry until they reached mass equilibrium (me) under the targeted temperature and relative humidity conditions. Subsequently, the cakes were dried under reference laboratory conditions (23  1  C and 43  5% RH) until no further mass change was observed (mref). The gravimetric moisture content (qg; [g/g]) of the cake at any time during the drying process was related to mref by the following relationship, which can be interpreted as the mass of moisture available for evaporation over mass of the cake at equilibrium under reference conditions: qg ðtÞ ¼

mt  mref mref

(1)

where mt is the mass of the cake at time t. 10e15% of mref or 0.11e0.18 g moisture/g dry mass, was determined to be remaining moisture (see last paragraph in this section for methodology). In order to compare the different drying curves, the moisture ratio qg/qg0 was plotted against time, where qg0 is the moisture content at the beginning of the drying process (t ¼ 0). Similarly, the concentration of surviving microorganisms (C; [pfu/g] or [viable eggs/g]) was normalized to the laboratorydried struvite mass mref as follows:

used to calculate the mass of struvite in the filter cake, from which the remaining moisture content was then determined.

2.7.

Filter cakes were exposed to natural sunlight for 5 h (12:00 to 17:00) under a clear sky on 20 July, 2010, on the roof of the Eawag laboratory building (47 230 N, 8 370 E). Struvite filter cakes were placed in a glass beaker, the bottom of which was covered with cloth to prevent the samples from getting wet from any condensation that may have formed on the surface of the glass. The beaker was placed in a plastic bucket filled with ice in order to maintain the temperature in and around the filter cakes close to 30  C (Table 3, experiment H). The parts of the setup that did not need to be exposed to the sunlight were covered with aluminium foil to slow the melting of the ice. The air temperature was measured with a PT100 sensor (TST310, Endress þ Hauser, Reinach, Switzerland) set next to the filter cake and connected to a data logger with a resolution of one measurement per minute. We were not able to measure the relative humidity just above the glass beaker during the experiment; however, by comparing the drying curve of the sunlight-exposed cake to those obtained in the experiments at 35  C in the shade inside the laboratory, we can conclude that the relative humidity was lower than 35%. The cumulative radiant exposure (fluence over the range of 280e900 nm) throughout the experiment was calculated from periodic irradiance measurements obtained by means of a spectroradiometer (model ILT-900-R; International Light). After 1, 2 and 5 h, fluences of 2119, 3974 and 7953 kJ/m2 were determined, respectively.

2.8. mt Ct ¼ Ct measured mref

(2)

where Ct is the normalized concentration of the microorganisms at time t; and Ct_measured is the actual concentration measured at time t for a mass mt. In order to express the drying of the cakes as a function of time, we used the Lewis thin layer drying model to fit our data (Lewis, 1921; Panchariya et al., 2002): mt  me ¼ exp ðko tÞ m0  me

(3)

where ko [1/time] is an empirical parameter determined by an exponential fit. Drying curves from experiments A, B, F, G, H (Table 3) were fitted with the Lewis equation yielding an R2 value between 0.96 and 0.99. The balance precision was 1 mg, therefore moisture contents lower than 0.001 g/g were extrapolated from Equations (1) and (3) using the respective fitted drying curve derived for each experiment. Data for which the moisture content was estimated by fitting are indicated by an asterisk in Figs. 3 and 5. The mass of liquid in the wet filter fabric was considered negligible in our calculations. The typical moisture content in struvite cakes after drying under laboratory conditions was determined in 2 cakes made from 1 L of urine. Around 0.5 g of the cake was dissolved in 100 mL 1 M HCl, and the phosphorus and magnesium concentrations were determined. These concentrations were

Sunlight experiments

NH3 concentration

NH3 concentration, pH and ionic strength in the cake moisture during drying were estimated for the conditions exhibiting the highest inactivation rates (36  C/36% RH, see Figs. 2a and 5a). Besides the evaporation of water and volatilization of NH3, the loss of CO2 was also considered because of its effect on pH and ionic strength of the solution. Assuming that the loss of water, NH3 and CO2 were limited by the diffusion out of small airfilled pores, which act as bottleneck boundaries, we used the film model to approximate the gas exchange (Schwarzenbach et al., 2003) (see Supplementary information). The simulation of the NH3 concentration, pH and ionic strength was performed by first calculating the composition of the moisture based on Table 2 (especially NH3 and CO2 concentration) with the computer code EQ3/6 Version 8.0 (Wolery and Jarek, 2003). This code allows calculating the chemical speciation in high-strength solutions, using the Pitzer approach for ionic strength corrections. The computer code was enhanced with equilibrium constants and Pitzer parameters for the dominant phosphate complexes (Supplementary information). The solubility constant for struvite and its temperature dependency was taken from Ronteltap et al. (2007). More details about the extension of the computer code are given in the Supplementary information. Simulations were performed stepwise and the concentrations were corrected according to an estimated gas exchange (for details, see Supplementary information).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 6 0 e4 9 7 2

For the sake of consistency, EQ3/6 was also used to recalculate the NH3 concentrations in the experiments of other authors cited herein.

3.

Results and discussions

3.1.

Phage FX174 distribution

Fig. 1a shows the distribution of phages between urine influent and struvite for different initial phage concentrations and different volumes of urine. The absolute number of phages in struvite was 3 log10 units (1000-fold) lower than the initial number in the urine (Fig. 1a). This was the case for every volume of urine and initial phage number tested. When comparing phage concentrations on a struvite mass basis, assuming a urine density of 1.02 g/cm3 (Price et al., 1940), the values were approximately equal for urine and struvite (Fig. 1a). As the moisture content of the cake immediately after filtration was rather high, (1.81  0.30 g moisture/g dry mass, 1.29  0.11 g/g and 1.25 g/g for the cake made from 1 L, 3 L and 4.5 L respectively) the concentration of phages in the struvite cake can be mainly attributed to the presence of phages in the residual urine contained by the cake. The even distribution of the phage concentration between urine and struvite suggests that neither straining by the filter fabric and

4965

filter cake, nor phage adsorption onto struvite, occurred during the filtration process. The apparent lack of straining can be explained by the very small size of FX174 (27 nm diameter (Dowd et al., 1998)). The absence of phage adsorption can be explained by considering the electrostatic interactions between struvite and phages. It has been shown that struvite particles are negatively charged in struvite-saturated solutions (10 mM NaCl) at pH 9.0e10.0 (Bouropoulos and Koutsoukos, 2000). In the same study, it was shown that an increase in the Mg2þ bulk concentrations reduced the negative surface charge, eventually resulting in a positive charge, if Mg2þ concentrations were above 18 mM. In our experiments, less than 1 mM of Mg2þ remained in solution after struvite precipitation (data not shown). Other, mostly monovalent cations (Naþ, Kþ; Table 2), may have additionally shielded the negative charge of the struvite. Nevertheless, we assume that struvite was negatively charged under the conditions of our experiment. FX174 also carries a negative surface charge under alkaline conditions, since its isoelectric point (pI) is 6.6 (Michen and Graule, 2010). This implies that there was electrostatic repulsion between struvite and phages. Although adsorption was not predominant under the conditions of our experiments, it may have a more pronounced effect on viruses with a pI higher than FX174. However, only few viruses have been reported to have a pI greater than 7. Examples among those with a high pI are

Fig. 1 e Distribution of FX174 and Ascaris eggs between urine and struvite. Absolute number (left) and concentration of (a) FX174 and (b) Ascaris eggs in urine influent (black) and struvite filter cake (grey). Error bars depict standard deviation for experiments with 3 or more replicates. Number of experiments with FX174: 2 (1 L, low conc.), 9 (1 L), 2 (3 L) and 1 (4.5 L). 3 experiments with Ascaris eggs. The concentration in struvite is calculated based on the mass mref. For urine a density of 1.02 g/cm3 was used (Price et al., 1940).

4966

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 6 0 e4 9 7 2

poliovirus and rotavirus (Michen and Graule, 2010). These two viruses could thus be expected to be retained by struvite more strongly than the viruses investigated herein. In addition, solutions with higher amounts of Mg2þ resulting from a higher Mg2þ dosage could lead to more adsorption of viruses with low pI onto struvite. Dosages of up to 1.8 mol Mg/mol P were used to achieve more than 95% phosphorus recovery (Abegglen, 2008). For stored urine with a phosphorus content similar to our experiments (Table 2), the remaining Mg2þ concentration upon addition of this higher dosage would be as low as 5.5 mM after precipitation. This is considerably lower than the 18 mM that were reported for a neutral surface charge of struvite (Bouropoulos and Koutsoukos, 2000). Finally, it should be noted that the measured phage concentrations in the filter effluent were of the same order of magnitude as the concentrations measured in the influent. This indicates that no considerable inactivation took place during the mixing and the filtration step.

3.2.

Ascaris egg distribution

Of the eggs initially added to the influent urine, 10e15% were found to adhere to the beaker and the filter setup, 50e55% were retained by the struvite cake, and the remainder broke through into in the effluent. This resulted in a 100-fold higher concentration of Ascaris eggs in struvite compared to the influent urine (Fig. 1b). Thus, the retention of Ascaris eggs on the filter and in the filter cake was high but incomplete. Ascaris eggs, which have a size of 35e50  45e70 mm (Jensen et al., 2009), are similar in size to the filter pore size. This led to partial straining of the eggs (around 60% of the suspended fraction) by the filter and the filter cake as soon as the struvite cake started to form. There were no significant changes in the Ascaris egg viability in the control or in the struvite immediately after filtration (data not shown); the mixing and filtration steps therefore had no effect on the survival of the eggs.

3.3.

Phage FX174 inactivation during drying phase

3.3.1.

Effect of temperature and relative humidity

Phages in filter cakes made from 1 L of stored urine were dried at different combinations of temperature and relative humidity. The concentration of surviving phages was regularly enumerated over the course of several days. A decrease in infective phage concentrations was observed in all experiments (Fig. 2), though the degree of inactivation depended on the environmental conditions. To assess the reproducibility of the experiments, experiments at 20  C/90% RH were performed in triplicate. The maximum variation in phage concentration between replicates was 0.5e0.8 log10 units at each sampling point (Fig. 2c). Inactivation during struvite drying was fastest at high temperatures and low relative humidity and slowest at low temperature and high relative humidity. After 3 days of drying, inactivation spanned a range of 0.5 log10 units at 5  C/ 85% RH to 3.5 log10 units at 36  C/35% RH (Fig. 2). Inactivation rates gradually decreased over the course of 3 days (Fig. 2). Samples taken after 9 and 10 days confirmed this tendency (data not shown): a reduction of 0.03, 0.07 and 0.09 log10/day

was observed for drying times longer than three days at 5  C/ 85%, 20  C/93% and 5  C/35% respectively. Temperature variations showed a large effect on the inactivation curves at both low (Fig. 2a) and high relative humidity (Fig. 2b). The effect of relative humidity on the inactivation was less pronounced, yet also apparent (Fig. 2c). The observed phage inactivation was thus a combination of the effects of temperature and relative humidity. As temperature and relative humidity both influence the moisture content of the cake (qg; see Section 2.6, Equation (1)), we examined whether this parameter correlated with the inactivation of the organisms tested. Within a 3 log10 decrease of the moisture content of the cake, a linear relationship between the logarithms of moisture content and inactivation was found from filtration experiment using 1 L of urine (Fig. 3a). From this relationship, a dependency of virus inactivation on moisture of 0.61  0.042 log10 viruses inactivated (log C/C0) per log10 of moisture content (log qg) was determined (robust regression with bisquare weighting function, MATLAB; error indicates the 95% confidence interval). In other words, when the moisture content was reduced by 90%, 75% of the phages were inactivated. A correlation between virus inactivation and decreasing moisture content was also observed for poliovirus in soil (Yeager and Obrien, 1979), in sludge (Brashear and Ward, 1983; Ward and Ashley, 1977b) and poliovirus and coliphages (MS2, PRD1) in soil amended with sludge (Straub et al., 1992). Although none of the authors compared inactivation and moisture content on a logelog scale, their data suggest a logarithmic dependency. Notably, our experiments showed the same correlation of inactivation and moisture content independent of the temperature or relative humidity (Fig. 3a). This correlation can thus be used to predict inactivation over a wide range of environmental conditions. Similarly, Straub et al. (1992) observed a correlation between soil moisture and phages (MS2 and PRD1) inactivation which was independent of the drying temperature over a range of 15e40  C. In addition to the direct effect of water evaporation, possibly leading to inactivation by desiccation, the loss of moisture from the filter cake could result in inactivation for three other reasons: the formation of a solid-air-water interface, the change in pH and NH3 concentration, and the increase in ionic strength (i.e. osmotic pressure). Dewatering leads to the introduction of air and the creation of a solidwater-air interface within the struvite cake. Thompson and Yates (1999) showed that if viruses in a solution containing Teflon beads were mixed with air, virus inactivation occurred. This effect, however, was only observed for hydrophobic surfaces in combination with phage MS2, which is known to be more hydrophobic than FX174 (Shields and Farrah, 2002). FX174, in contrast, was not susceptible to this type of inactivation. Therefore, we do not expect that the solid-air-water interface had an effect on inactivation in our experiments. Secondly, the chemical composition of urine may cause inactivation due to the high pH and NH3 content, both of which have been shown to have virucidal properties (Cramer et al., 1983; Ward and Ashley, 1977a). The NH3 fraction of total ammonia (NH3 þ NHþ 4 ) increases with rising pH and rising temperature. For FX174, Vinneras et al. (2008) observed a 1 log10 inactivation after 5.7, 12 and 120 days at 34, 24 and

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 6 0 e4 9 7 2

4967

Fig. 2 e FX174 inactivation (left) and moisture ratio (right) in struvite cake as a function of time. Struvite was produced from 1 L of stored urine. Drying conditions: (a) Low RH (35e42%) and variable T (5  C (-), 20  C (-) and 36  C (-). (b) High RH (85e93%) and variable T (5  C (:), 20  C (:) and 35  C (:)). (c) Constant T (20  C) and variable RH (42% (-), 77% (C) and 90e93% (: [E.1], A [E.2], ; [E.4]; see Table 3). The dashed line (—) indicates the detection limit. Error bars depict standard deviation of triplicate enumeration of the same sample.

4  C, respectively, in urine containing 426 mM total ammonia (164, 102 and 34 mM NH3). In our experiments, we measured up to 95% moisture content reduction in the cake during drying. On this basis, an approximation of the NH3 concentration in the cake moisture was simulated for the conditions with the highest inactivation rate (36  C/36% RH). Fig. 4 shows that the NH3 concentration, which was initially around 80 mmol/kgwater (w80 mM) decreased with decreasing moisture content and stabilized below 1 mmol/kgwater after a moisture content reduction of 20%, corresponding to 30 min of drying. Despite this low NH3 concentration we measured a much higher inactivation rate than Vinneras et al. (2008).

Therefore, we do not expect that the NH3 caused inactivation in our experiments. The pH exhibited a similar dependence on moisture content. From its initial value of around 9, the pH rapidly decreased to approximately 6.7 within the first 20% of moisture loss. Additional drying caused only a small further decreased in pH (Fig. 4). pH values in the neutral range are not virucidal (Feng et al., 2003). We therefore conclude that the change in moisture pH did not contribute to inactivation. Finally, it has been shown that high ionic strength up to 2 M does not affect the inactivation of FX174 (Thompson and Yates, 1999). In our experiments, the estimated ionic strength upon drying of the struvite cake reached around 2 M

4968

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 6 0 e4 9 7 2

for the experiment at 36  C/36% RH (Fig. 4). Therefore we can infer that inactivation remains unaffected by ionic strength. In summary, neither the solid-air-water interface, the NH3 concentration, the pH, nor the increase in ionic strength was likely to account for the observed phage inactivation. Thus, desiccation remains the best parameter for rationalizing the observed relationship between the moisture content and phage inactivation. The moisture content therefore remains the best parameter to predict the level of inactivation at the filter cake size of our experiments (1.5  0.7 mm thickness). Thus, within the temperature range (5e36  C), relative humidity range (35e93%) and time scale (0e10 days) of our experiments, temperature and relative humidity should be seen as parameters that influence the phage inactivation rate by determining the moisture evaporation rate. Since the moisture content nearly reached its equilibrium value after three days (Fig. 2), and since higher temperatures and lower relative humidity led to lower moisture content and lower concentrations of infective viruses (Fig. 3), increasing the temperature or reducing relative humidity was more effective for pathogen inactivation than increasing the drying time.

3.3.2.

Effect of sunlight

Phages are known to be susceptible to direct inactivation by sunlight in the UVB range, as well as to indirect inactivation by higher wavelengths (Davies-Colley et al., 2000). FX174 was shown to be one of the most sensitive viruses to sunlight irradiation among bacteriophages and human viruses (Hijnen et al., 2006; Love et al., 2010). If this phage is not inactivated, more resistant viruses will not be affected either. To test if the presence of sunlight enhances phage inactivation during struvite drying, experiments were conducted under solar irradiation (experiment H, Table 3) and the resulting inactivation was compared to control experiments in the dark. After 5 h of exposure to natural sunlight (7953 kJ/m2, 31  C, RH < 35%), a 1.5 log10 inactivation was observed (data not shown), while after 6 h in the shade under similar experimental conditions (36  C, 35% RH) a 2 log10 inactivation was measured (Fig. 2a). A logelog comparison of phage inactivation and moisture content showed that the experiments conducted under sunlight adhered to the correlation developed for experiments conducted in the shade (Fig. 3b). This

Fig. 3 e FX174 inactivation as a function of the moisture content of the cake (qg): (a) comparison between filter cake

made with 1 L of stored urine and dried in the shade at different temperatures and relative humidity: 35e42% RH and 5  C (-), 20  C (-) and 36  C (-); 77% RH and 20  C (C); 85%e93% RH and 5  C (:), 20  C (: [E.1], A [E.2], ; [E.4]; see Table 3) and 35  C (:). Data correspond to those depicted in Fig. 2. (b) Filter cake dried in the sunlight (>) after 0, 1, 2 and 5 h sunlight exposure; (c) comparison between filter cake dried in the shade at 20  C and 77e93% RH and made from 3 L (empty symbols) and 4.5 L (solid symbols) of stored urine. Data for which qg was determined by extrapolation are indicated by (*) (see Section 2.6). The continuous line shown in all graphs represents the robust linear regression curve of the whole dataset “1 L/in the shade”. The dashed line (—) indicates the detection limit. Error bars depict standard deviation of triplicate enumeration of the same sample.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 6 0 e4 9 7 2

4969

Fig. 4 e Simulation of NH3 concentration (C), pH (,) and ionic strength (:) as a function of moisture reduction in the cake during drying at 36  C/36% RH.

indicates that, despite the thinness of the cake (1.5  0.7 mm), the sunlight did not penetrate into the cake to a sufficient extent to significantly contribute to inactivation. The main contribution of sunlight to the inactivation was an indirect one that resulted from increasing the temperature of the filter cake. Drying the cake in the sunlight can thus enhance inactivation by reducing the moisture of the cake more rapidly, but it must be considered that high temperatures will also cause the release of ammonia from the struvite crystals (Bhuiyan et al., 2008; Frost et al., 2004).

3.3.3.

Effect of filter cake thickness

To investigate the effect of filter cake’s thickness on drying and phage inactivation, cakes made from 3 to 4.5 L were dried at 20  C and 70e90% RH. Inactivation in these thicker cakes could also be linked to the moisture content with approximately the same correlation as the one obtained for the 1 L cakes (Fig. 3c). However, a longer drying time was required to reach the same moisture content as the cake made from 1 L of urine (data not shown). We furthermore caution that the infective virus and moisture content relationship was only tested for urine volumes up to 4.5 L. For the larger volumes commonly used in the field, this relationship remains to be confirmed.

3.4.

Ascaris suum egg inactivation during drying phase

3.4.1.

Effect of temperature and relative humidity

Ascaris suum eggs showed no significant inactivation at 5 and 20  C after 3 days of drying (0.003e0.01 log10 inactivation) (Fig. 5a). Only the experiments carried out at 35e36  C inactivated eggs to a considerable extent: 1.2 to more than 2 log10 reduction occurred after 3 days of drying. No significant difference was observed when varying relative humidity at 20  C, but at 35e36  C inactivation was faster at low relative humidity. At 36  C and 36% RH, less than one day was needed to achieve a 2 log10 inactivation, while at 35  C and 85% RH, the inactivation was only 1.2 log10 after 3 days. Other researchers have shown that high temperatures are required to inactivate Ascaris suum eggs in buffered ammonia-

Fig. 5 e Ascaris suum egg inactivation as a function of (a) time and (b) struvite moisture content (qg). Struvite cakes were made from 1 L of stored urine and dried under different conditions: 5  C and 85% RH (:); 20  C and 90% RH (:), 77% RH (C), 42% RH (-); 35  C and 85% RH (:), 36  C and 36% RH (-); sunlight 31  C, <35% RH (>). Data for which qg was determined by extrapolation are indicated by (*) (see Section 2.6). The dashed line (—) indicates the detection limit. Error bars depict standard deviation of triplicate enumeration of the same sample.

free solutions. Nordin et al. (2009) did not observe any significant inactivation over the course of one month at 24  C and 34  C in buffered saline solution. Pecson and Nelson (2005) observed a 2 log10 inactivation after 3 days at temperatures above 48  C, but found that inactivation was only minimally influenced by temperatures below 40  C over the same time period. Wharton (1979) demonstrated that the relative humidity has a strong effect on inactivation when Ascaris eggs are exposed to air: at high temperature (30  C), eggs were inactivated within 3

4970

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 6 0 e4 9 7 2

days at 0e34% relative humidity, and within 7 days at 75.5% relative humidity. At low temperature (16  C), the time required for inactivation was five to ten times slower. Yet, in more than 95% relative humidity, all eggs survived for more than 51 days regardless of temperature (Wharton, 1979). In the struvite cake, Ascaris eggs faced both wet and dry conditions. Immediately after filtration, all eggs were surrounded by urine. As the moisture front within the pores retreated, the eggs became increasingly exposed to air. This is in contrast to the conditions encountered by the phages, whose very small size may allow them to retreat with the moisture front and hence remain immersed in liquid for greater periods of time. Ascaris inactivation did not exhibit a linear correlation with the moisture content of the cake (Fig. 5b), but our observations suggest that inactivation started when the moisture content was reduced to a sufficient extent such that most eggs were exposed to air. In the case of a filter cake made from 1 L urine, this moisture content threshold was between 0.001 and 0.01 g/ g if normalized to mref, (recall that mref additionally has a remaining moisture content of 0.11e0.18 g moisture/g dry mass). Additional experiments are needed to determine the inactivation threshold more accurately. It is probable that eggs follow the inactivation kinetics observed for desiccation (Wharton, 1979) once the moisture threshold is reached. In this case, inactivation will also take place at temperatures below 36  C, if struvite drying is sufficiently long to reach a moisture content below the critical threshold. As for phages, it has been shown that exposure to NH3 causes Ascaris suum egg inactivation (Pecson and Nelson, 2005). At 36  C and pH 9.0, in a buffered solution containing approximately 570 mM total ammonia (248 mM NH3), Pecson and Nelson (2005) observed a 2 log10 inactivation after 3 days. A similar inactivation was confirmed by Nordin et al. (2009) who did experiments with urine at 34  C, pH 8.8e9.0 and a total ammonia concentration of 439 mM (169 mM NH3). Under the same pH and temperature conditions, but in diluted urine with 203 and 131 mM total ammonia (59 and 33 mM NH3, respectively), they observed 2 log10 inactivation after 6.3 and 8.5 days, respectively. As discussed above, the NH3 concentration in the cake moisture during drying at 36  C/36% RH evened out below 1 mmol/kgwater and the pH stabilized around neutral within 20% moisture content reduction (Fig. 4). Despite the neutral pH and low NH3 concentration, we measured a higher inactivation rate than Pecson and Nelson (2005) and Nordin et al. (2009). Thus, pH and NH3 concentration could not explain the observed inactivation. Osmotic stress was shown to have no effect on Ascaris eggs during embryonation (Matthews, 1985). Thus, inactivation in our experiment could not be attributed to the increase of ionic strength during struvite drying. In summary, Ascaris egg inactivation was affected only by desiccation resulting from the loss of moisture in the struvite cake at higher temperatures.

3.4.2.

exposure of a struvite cake to sunlight, no significant inactivation was observed (<0.02 log10 inactivation; data not shown). As for the phage experiments, we assume that most of the Ascaris eggs were well shielded within the struvite cake. In our experiments, the indirect effect of temperature increase and higher evaporation in the sunlight was more important for struvite cakes than the inactivation by solar radiation.

4.

Conclusions

In this study we investigated the fate of pathogens during two key steps of struvite recovery from urine: 1) struvite precipitation and filtration and 2) struvite drying. Phage FX174 and Ascaris suum eggs were used as surrogates for human viruses and helminths. Both microorganisms were retained in struvite, although only the large Ascaris suum eggs accumulated (100-fold more concentrated in struvite than in urine on a per mass basis), probably as a result of straining. The concentration of the smaller phage FX174 could be explained by the amount of phage-containing urine remaining in the struvite cake after filtration. The effluent produced as a result of struvite production contained high amounts of both microorganisms. If the effluent should be used as fertilizer, additional treatment is advisable to reduce the pathogen content. For filter cakes made from 1 L urine, the moisture content was shown to be the predominant parameter affecting microorganism inactivation. For phages, we found a linear relationship between the logarithms of microorganism inactivation and the moisture content, even for thicker cakes made from 3 to 4.5 L of urine. For Ascaris eggs, inactivation occurred only below a certain moisture content threshold (in our experimental conditions 0.001e0.01 g/g referring to mref) due to desiccation. Osmotic pressure, as well as biocidal effects of pH or NH3 in the cake moisture could be excluded as factors causing inactivation. Exposure of the struvite cakes to sunlight had no substantial direct effect on phage FX174 or Ascaris suum eggs. However, heating the struvite cakes in the sun accelerated moisture loss and thereby, inactivation. Tough our findings remains to be confirmed in the field and for larger scale struvite production, we recommend the following measures for struvite treatment to minimize pathogen concentrations, based on our conclusion that low moisture content is the main factor governing pathogen inactivation:  Dry filter cakes at elevated temperatures (e.g., in the sun) and/or low relative humidities; ensure, however, that temperatures in the cake do not exceed 40e55  C to prevent substantial ammonia loss.  Create thin struvite filter cakes to minimize the drying phase. Based on our observations for indicator organisms for human viruses and helminths, we propose that the following should be taken into account during struvite production:

Effect of the sunlight

Spindler (1940) showed that Ascaris eggs in shallow water and on a dry surface were completely inactivated by sunlight within a maximum of 9 h at 30e35  C. However, after 5 h of

 For viruses: high temperatures or low relative humidity during a short drying phase are more advisable than drying the struvite for a long time. Extending the drying phase

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 6 0 e4 9 7 2

beyond the time when a steady-state moisture content has been reached will not increase the inactivation significantly;  For Ascaris eggs: only high temperatures, preferably combined with low relative humidity, cause inactivation in the short term. For temperatures below 35  C, inactivation may occur if the cake is stored for several weeks. However, this assumption remains to be tested. Inactivation may also occur during urine storage, struvite storage and after applying the struvite to the soil. In order to have a complete evaluation of human health risk of struvite fertilizer produced from urine, further studies on inactivation during production and application of struvite are needed and should be linked in a quantitative microbial risk assessment. Nevertheless, with our study we could demonstrate that e given the appropriate temperature and relative humidity conditions - complete drying of struvite in ambient air, coupled with an appropriate urine storage time, would represent an efficient barrier for pathogens.

Acknowledgements This work was supported by Angel Fund of the Gemeinnu¨tzige Stiftung SYMPHASIS, Zu¨rich. The authors thank Thomas Egli, Thomas Fleischmann, Frederik Hammes, Stefan Koetzsch, Rudolf Schneebeli and Hans-Ulrich Weilenmann for providing the laboratory facilities and for valuable advice; Bastian Etter for the introduction to the struvite production process developed in Nepal; Michael Wa¨chter for the urine composition measurements and the modification of the EQ3/6 database; Brian Sinnet for the filter analysis; Claudia Baenninger, Karin Rottermann and Jean-David Teuscher for chemical analyses; Pietro Lura and Walter Trindler for access to the temperature and relative humidity regulated storage rooms, Pooja Manandar for testing the laboratory methodologies and Stefan Diener for the worm egg counting slides. We furthermore thank three anonymous reviewers for their helpful insights and comments.

Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.06.042.

references

Abegglen, C.K., 2008. Membrane bioreactor technology for decentralized wastewater treatment and reuse. PhD thesis, Nr. 17998, Swiss Federal Institute of Technology Zurich, Switzerland, doi: 10.3929/ethz-a-005745648. Bhuiyan, M.I.H., Mavinic, D.S., Koch, F.A., 2008. Thermal decomposition of struvite and its phase transition. Chemosphere 70, 1347e1356. Bouropoulos, N.C., Koutsoukos, P.G., 2000. Spontaneous precipitation of struvite from aqueous solutions. Journal of Crystal Growth 213, 381e388.

4971

Brashear, D.A., Ward, R.L., 1983. Inactivation of indigenous viruses in raw sludge by air drying. Applied and Environmental Microbiology 45, 1943e1945. Cramer, W.N., Burge, W.D., Kawata, K., 1983. Kinetics of virus inactivation by ammonia. Applied and Environmental Microbiology 45, 760e765. Davies-Colley, R.J., Donnison, A.M., Speed, D.J., 2000. Towards a mechanistic understanding of pond disinfection. Water Science and Technology 42, 149e158. Dowd, S.E., Pillai, S.D., Wang, S.Y., Corapcioglu, M.Y., 1998. Delineating the specific influence of virus isoelectric point and size on virus adsorption and transport through sandy soils. Applied and Environmental Microbiology 64, 405e410. Etter, B., Tilley, E., Khadka, R., Udert, K.M., 2011. Low-cost struvite production using source-separated urine in Nepal. Water Research 45, 852e862. Feachem, R.G., Bradley, D.J., Garelick, H., Mara, D.D., 1983. Sanitation and Disease: Health Aspects of Excreta and Wastewater Management. Wiley, New York. Feng, Y.Y., Ong, S.L., Hu, J.Y., Tan, X.L., Ng, W.J., 2003. Effects of pH and temperature on the survival of coliphages MS2 and Q beta. Journal of Industrial Microbiology & Biotechnology 30, 549e552. Frost, R.L., Weier, M.L., Erickson, K.L., 2004. Thermal decomposition of struvite - implications for the decomposition of kidney stones. Journal of Thermal Analysis and Calorimetry 76, 1025e1033. Goosse, P., Steiner, M., Neuenschwander, W., Udert, K.M., 2009. NoMix-Toilettensystem. Erste Monitoringergebnisse im Forum Chriesbach. Gas Wasser Abwasser 7, 567e574 (NoMix Toilet System, First Monitoring Results in Forum Chriesbach, in German). Hijnen, W.A.M., Beerendonk, E.F., Medema, G.J., 2006. Inactivation credit of UV radiation for viruses, bacteria and protozoan (oo) cysts in water: a review. Water Research 40, 3e22. Hoglund, C., Ashbolt, N., Stenstrom, T.A., Svensson, L., 2002. Viral persistence in source-separated human urine. Advances in Environmental Research 6, 265e275. Hoglund, C.E., Stenstrom, T.A.B., 1999. Survival of Cryptosporidium parvum oocysts in source separated human urine. Canadian Journal of Microbiology 45, 740e746. Jenkins, M.B., Bowman, D.D., Ghiorse, W.C., 1998. Inactivation of Cryptosporidium parvum oocysts by ammonia. Applied and Environmental Microbiology 64, 784e788. Jensen, P.K.M., Phuc, P.D., Konradsen, F., Klank, L.T., Dalsgaard, A., 2009. Survival of Ascaris eggs and hygienic quality of human excreta in Vietnamese composting latrines. Environmental Health 8, 57. Kone, D., 2010. Making urban excreta and wastewater management contribute to cities’ economic development: a paradigm shift. Water Policy 12, 602e610. Kraft, 2010. Final Sampling Report for Products from DoubleChamber UDDTs. EU-SIDA GTZ EcoSan Promotion Project. http:// www2.gtz.de/Dokumente/oe44/ecosan/en-eu-sida-gtz-ecosanpromotion-project-final-report-2010.pdf accessed June 2010. Langergraber, G., Muellegger, E., 2005. Ecological sanitation a way to solve global sanitation problems. Environment International 31, 433e444. Larsen, T.A., Gujer, W., 1996. Separate management of anthropogenic nutrient solutions (human urine). Water Science and Technology 34, 87e94. Larsen, T.A., Maurer, M., Udert, K.M., Lienert, J., 2007. Nutrient cycles and resource management: implications for the choice of wastewater treatment technology. Water Science and Technology 56, 229e237. Lewis, W.K., 1921. The rate of drying of solid materials. Journal of Industrial and Engineering Chemistry 13, 427e432. Lienert, J., Burki, T., Escher, B.I., 2007. Reducing micropollutants with source control: substance flow analysis of 212

4972

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 6 0 e4 9 7 2

pharmaceuticals in faeces and urine. Water Science and Technology 56, 87e96. Love, D.C., Silverman, A., Nelson, K.L., 2010. Human virus and bacteriophage inactivation in clear water by simulated sunlight compared to bacteriophage inactivation at a Southern California Beach. Environmental Science & Technology 44, 6965e6970. Matthews, B.E., 1985. The influence of temperature and osmotic stress on the development and eclosion of hookworm eggs. Journal of Helminthology 59, 217e224. Maurer, M., Pronk, W., Larsen, T.A., 2006. Treatment processes for source-separated urine. Water Research 40, 3151e3166. Michen, B., Graule, T., 2010. Isoelectric points of viruses. Journal of Applied Microbiology 109, 388e397. Nordin, A., Nyberg, K., Vinneras, B., 2009. Inactivation of ascaris eggs in source-separated urine and Feces by ammonia at ambient temperatures. Applied and Environmental Microbiology 75, 662e667. Panchariya, P.C., Popovic, D., Sharma, A.L., 2002. Thin-layer modelling of black tea drying process. Journal of Food Engineering 52, 349e357. Pecson, B.M., Nelson, K.L., 2005. Inactivation of Ascaris suum eggs by ammonia. Environmental Science & Technology 39, 7909e7914. Price, J.W., Miller, M., Hayman, J.M., 1940. The relation of specific gravity to composition and total solids in normal human urine. Journal of Clinical Investigation 19, 537e554. Pronk, W., Kone, D., 2009. Options for urine treatment in developing countries. Desalination 248, 360e368. Romer, W., 2006. Plant availability of P from recycling products and phosphate fertilizers in a growth-chamber trial with rye seedlings. Journal of Plant Nutrition and Soil Science. (Zeitschrift Fu¨r Pflanzenerna¨hrung Und Bodenkunde) 169, 826e832. Ronteltap, M., Maurer, M., Gujer, W., 2007. The behaviour of pharmaceuticals and heavy metals during struvite precipitation in urine. Water Research 41, 1859e1868. Sanchez, P.A., 2002. Ecology - soil fertility and hunger in Africa. Science 295, 2019e2020. Schonning, C., Leeming, R., Stenstrom, T.A., 2002. Faecal contamination of source-separated human urine based on the content of faecal sterols. Water Research 36, 1965e1972. Schwarzenbach, R.P., Gschwend, P.M., Imboden, D.M., 2003. Environmental Organic Chemistry, second ed. Wiley, Hoboken, N.J.

Shields, P.A., Farrah, S.R., 2002. Characterization of virus adsorption by using DEAE-sepharose and octyl-sepharose. Applied and Environmental Microbiology 68, 3965e3968. Spindler, L.A., 1940. Effect of tropical sunlight on eggs of ascaris suis (Nematoda), the large intestinal Roundworm of Swine. The Journal of Parasitology 26, 323e331. Straub, T.M., Pepper, I.L., Gerba, C.P., 1992. Persistence of viruses in desert soils amended with anaerobically digested sewagesludge. Applied and Environmental Microbiology 58, 636e641. Thompson, S.S., Yates, M.V., 1999. Bacteriophage inactivation at the air-water-solid interface in dynamic batch systems. Applied and Environmental Microbiology 65, 1186e1190. Udert, K.M., Larsen, T.A., Biebow, M., Gujer, W., 2003. Urea hydrolysis and precipitation dynamics in a urine-collecting system. Water Research 37, 2571e2582. Udert, K.M., Larsen, T.A., Gujer, W., 2006. Fate of major compounds in source-separated urine. Water Science and Technology 54, 413e420. Vinneras, B., Nordin, A., Niwagaba, C., Nyberg, K., 2008. Inactivation of bacteria and viruses in human urine depending on temperature and dilution rate. Water Research 42, 4067e4074. Ward, R.L., Ashley, C.S., 1977a. Identification of virucidal agent in wastewater-sludge. Applied and Environmental Microbiology 33, 860e864. Ward, R.L., Ashley, C.S., 1977b. Inactivation of enteric viruses in wastewater-sludge through dewatering by evaporation. Applied and Environmental Microbiology 34, 564e570. Wharton, D.A., 1979. Ascaris-Sp - water-Loss during desiccation of embryonating eggs. Experimental Parasitology 48, 398e406. WHO, 2006. Guidelines for the Safe Use of Wastewater, Excreta and Greywater - v. 4. Excreta and Greywater Use in Agriculture. World Health Organization, Geneva, ISBN 92 4 154686 7. Winker, M., Vinneras, B., Muskolus, A., Arnold, U., Clemens, J., 2009. Fertiliser products from new sanitation systems: their potential values and risks. Bioresource Technology 100, 4090e4096. Winston, P.W., Bates, D.H., 1960. Saturated solutions for the control of humidity in biological-research. Ecology 41, 232e237. Wolery, T.W., Jarek, R.L., 2003. Software User’s Manual EQ3/6 (version 8.0). Sandia National Laboratories, Albuquerque, New Mexico. Yeager, J.G., Obrien, R.T., 1979. Enterovirus inactivation in soil. Applied and Environmental Microbiology 38, 694e701.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 7 3 e4 9 8 2

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Spatio-temporal distribution of cell-bound and dissolved geosmin in Wahnbach Reservoir: Causes and potential odour nuisances in raw water Sabine Ja¨hnichen a,c,*, Kathrin Ja¨schke a, Falk Wieland a, Gabriele Packroff b, Ju¨rgen Benndorf a a

Institute for Hydrobiology, Dresden University of Technology, 01062 Dresden, Germany Wahnbach Reservoir Association, Siegelsknippen, 53721 Siegburg, Germany c Institute for Occupational and Social Medicine, Dresden University of Technology, 01062 Dresden, Germany b

article info

abstract

Article history:

In many lakes and reservoirs, problems caused by off-flavours are known to be particularly

Received 19 November 2010

associated with the occurrence of planktonic and benthic cyanobacteria. Frequently

Received in revised form

observed objectionable taste and odorous products of cyanobacteria are geosmin and

25 June 2011

2-methylisoborneol.

Accepted 30 June 2011 Available online 13 July 2011

Investigations focused on the littoral zone of Wahnbach Reservoir (Germany) revealed that benthic cyanobacteria were present in this oligotrophic drinking water reservoir. Benthic cyanobacteria were found in the depth horizon between 1.75 m and 11 m,

Keywords:

particularly on south-exposed slopes. This spatial distribution indicates a possible key role

Odour compounds

of the underwater light climate. Moreover, cell-bound and dissolved geosmin were

Off-flavours

detected in corresponding littoral samples. Both fractions were subjected to spatial and

Drinking water

primarily temporal variations with maximum concentrations at the end of summer.

Benthic cyanobacteria

However, a substantial lowering of the water level caused a diminution of cyanobacterial

Oligotrophic reservoir

growth. Due to the drawdown of the water level concentrations of cell-bound geosmin and pigments (as a proxy of cyanobacterial biomass) were remarkably reduced, and dissolved geosmin was never detected during this phase. Except for the influence of water level fluctuation no other abiotic variables had a significant influence on pigment and geosmin concentrations. From geosmin concentrations detected in the littoral zone, the probability of serious episodes of odour events in the raw water of the Wahnbach Reservoir was estimated. Hence, the probability that the raw water was affected by geosmin was minor, which was supported by routine flavour profiles. Nevertheless, the study shows that odorous episodes caused by benthic cyanobacteria are likely to develop even in an oligotrophic lake or reservoir when these cyanobacteria, and consequently odorous production, proliferate. In principle, such a proliferation cannot be excluded as nutrients are available from the sediment pore water, and underwater light at the sediment surface in the sublittoral is sufficiently high due to very low phytoplankton-induced turbidity under oligotrophic conditions. Thus, management-induced fluctuations of the water level seem to be the main control variable to generate light conditions at the sediment surface fluctuating in a given depth horizon faster than cyanobacteria can develop there. ª 2011 Elsevier Ltd. All rights reserved.

* Corresponding author. Dresden University of Technology, Institute for Hydrobiology, 01062 Dresden, Germany. E-mail address: [email protected] (S. Ja¨hnichen). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.06.043

4974

1.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 7 3 e4 9 8 2

Introduction

Episodes of taste and odour nuisance of water bodies have been reported throughout the world (for summary see Ju¨ttner and Watson, 2007; Krishnani et al., 2008). Among the multitude of compounds causing water quality problems the most prominent ones are the earthyemuddyemusty smelling terpenoids geosmin and 2-methylisoborneol (MIB). Odour threshold concentrations of geosmin and MIB are very low, thus concentrations at parts per trillion levels can already be detected by human olfactory sense. Geosmin and MIB are produced by planktonic and benthic aquatic cyanobacteria, and these metabolites may also originate from aquatic filamentous actinomycete bacteria (Watson, 2003; Smith et al., 2008). For a long time the odorous potential of benthic cyanobacteria has been underestimated (Izaguirre and Taylor, 2004; Watson, 2004), mainly due to the methodology of routine sampling in lakes and reservoirs (Wood et al., 2001). However, if the potential of several benthic and pelagic cyanobacterial taxa to produce geosmin or MIB was compared, it became obvious that the majority of the identified producers were benthic species (Watson, 2003; Ju¨ttner and Watson, 2007; Smith et al., 2008). Thus, it can be concluded that odorous problems in lakes and reservoirs may be mainly caused by these organisms. However, besides the taxonomical affiliation the environmental conditions also exert a strong (indirect) influence on the production of geosmin or MIB by controlling the cyanobacterial growth. The proliferation of benthic cyanobacteria in lakes and reservoirs depends primarily on the underwater light climate which is governed by incident solar radiation and turbidity in the water column. Besides sediment resuspension, turbidity is mainly caused by phytoplankton abundance which, in turn, is controlled by dissolved nutrients. Consequently, the littoral of oligotrophic lakes with ample light conditions was assigned to be the preferred habitat of benthic cyanobacteria (Lowe, 1996), and benthic contribution to total primary production was kova´ et al., found to be highest in such clear-water lakes (Poulic 2008). More eutrophic and turbid conditions impede phytobenthic primary production mainly due to the shading by the phytoplankton (Lowe, 1996; Liboriussen and Jeppesen, 2003). Thus, whereas pelagic geosmin production resulted in higher concentrations in a eutrophic lake compared to meso- and oligotrophic ones (Peter et al., 2009), benthic cyanobacteria were considered to be the primary source of water odour problems in a mesotrophic system (Watson and Ridal, 2004). In Wahnbach Reservoir, an oligotrophic water body used for drinking water supply, sporadic episodes of unpleasant odours in drinking water were documented during the last years. However, the origin of the muddy, musty smelling compounds remained unknown. Therefore, the objective of this study was to determine the sources and to analyse the spatio-temporal distribution of off-flavours in Wahnbach Reservoir. Secondly, influences of abiotic variables on the dynamics of the odour compounds will be investigated in order to gain insights to the underlying processes that trigger and modify benthic growth and odour production. Thirdly, the probability of off-flavour impacts on raw water will be estimated from odour concentrations in littoral zone.

2.

Methods

2.1.

Study site, sampling

The investigation was performed in Wahnbach Reservoir (Germany, 50
480 N, 7
170 E), an oligotrophic, dimictic or monomictic water body with a surface area of 2.0 km2, a volume of 40.9∙106 m3 and a maximum depth of 46 m. Samples were collected from the littoral (for locations see Fig. 1) from 2006 through 2009 between May and October. Sampling was performed between 8 am and 1 pm. Every sampling position was located with a GPS navigator (Garmin, Olathe, KS, USA). Phytobenthos were surveyed by diving several transects to a depth of about 10 m. The areal extent of algal mats was estimated visually at each sampling site. Benthic samples were taken from all areas where algal mats grew, otherwise a random sample was selected at a depth of 3 m. At each site, water was collected close to the sediment surface for chemical analyses, including total dissolved phosphorus (TDP), soluble reactive  phosphorus (SRP), nitrite ðNO 2 Þ, nitrate ðNO3 Þ, ammonium þ ðNH4 Þ, dissolved iron, and dissolved geosmin and MIB. Furthermore, water temperature, pH, and dissolved oxygen (O2) were recorded close to the sediment surface by WTW probes (WTW Instruments, Weilheim, Germany). Measurements of light attenuation were performed with a spherical quantum sensor (LiCor, Lincoln, NE, USA). At two pelagic sampling points (PA, PE see Fig. 1) depth profiles of chlorophyll

Fig. 1 e Sampling sites in Wahnbach Reservoir. All littoral sampling sites (L) which were investigated from 2006 through 2009 are indicated by a grey circle except for those during low water level phase in 2008 (white circles). Additional sampling sites in the pelagic zone (PA, PE) are shown.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 7 3 e4 9 8 2

a were measured using a submersible multichannel fluorescence probe (BBE Moldaenke, Kiel, Germany). In addition, water samples from different layers (epi-, meta- and hypolimnion) were taken to analyse dissolved geosmin and MIB.

2.2.

Sample preparation, analysis

2.2.1.

Nutrients

Water samples were filtered through a 0.45 mm filter to analyse for dissolved substances. Analyses were performed according to standard methods (Table 1).

2.2.2.

Identification of phytobenthos

Randomly selected samples of phytobenthos were microscopically characterised (Axiophot, Zeiss, Germany) by their morphological features according to Koma´rek and Anagnostidis (2005).

2.2.3. Quantification of phytobenthos: analysis of chlorophyll a, phycocyanin, phycoerythrin Phytobenthos samples were subdivided into samples for pigment analysis. For chlorophyll a analysis, a subsample was extracted in methanol, concentrations were determined based on excitation at 434 nm and emission at 667 nm using a luminescence spectrometer (LS 50 B, Perkin Elmer). For phycocyanin and phycoerythrin analysis, cells were disrupted by freezing/thawing and ultrasonication (Sonoplus W 70, Bandelin) with subsequent extraction of phycocyanin and phycoerythrin in phosphate buffer (pH 7, 12 h, 4
C) according to Otsuki et al. (1994). Concentrations of phycocyanin were measured at an emission of 645 nm with excitation wavelength of 625 nm. The purity of phycocyanin was evaluated by using the A620/A280 ratio (Minkova et al., 2003). Concentrations of phycoerythrin were measured at an emission of 570 nm with an excitation of 520 nm, phycocyanin and phycoerythrin concentrations were analysed using a luminescence spectrometer (LS 50 B, Perkin Elmer).

2.2.4.

Analysis of geosmin and MIB

For the quantification of cell-bound odorous compounds, the phytobenthos samples were first diluted in methanol in order to extract both cytosolic and protein-bound geosmin (Wu and Ju¨ttner, 1988; Ju¨ttner and Watson, 2007). Cell-bound and dissolved odour compounds were determined by solid phase microextraction (SPME) from a liquid solution on a DVB/ Carboxen/PDMS fibre (Supelco Inc.) followed by gas chromatography analysis (GC). SPME was performed using an automated device (Consept, PAS-Technologies, Germany) as follows: samples were shaken at 65
C for 20 min, then the

Table 1 e Chemical variables determined and techniques used. Variable total dissolved phosphorus soluble reactive phosphorus nitrite nitrate ammonium dissolved iron

Symbol

Reference

TDP SRP NO 2 NO 3 NHþ 4 Fe

EN ISO 15681-2 D46 EN ISO 15681-2 D46 DIN EN ISO 13395 D28 DIN EN ISO 13395 D28 DIN EN ISO 11732 E23 DIN EN ISO 11885 E22

4975

fibre was exposed in the head-space of the vial for 60 min in order to absorb the odour compounds. The fibre was transferred to the injection port of the gas chromatograph and desorbed in the splitless mode at 250
C for 2 min. Analysis was performed on an Agilent 6890N gas chromatograph equipped with flame ionization detector (GC/FID) using a DB-624 column (30  0.53  3, J & W) and helium as carrier gas (flow 4.4 ml min1). The operating conditions were as follows: Starting from 50
C (constant for 3 min) to 250
C at 10
C/min, finally the temperature was held at 250
C for 10 min. For the quantitative analysis, the system was calibrated within the concentration range from 50 ng L1 to 10 mg L1 for each analyte. Calibration standards for geosmin (Supelco Inc.) and MIB (Supeclo Inc.) were used. 2-isopropyl3-methoxypyrazine (Supelco Inc.) was added to each sample as internal standard.

2.3.

Statistical analysis

Principle component analysis was performed using the R 2.8.1 system for statistical analysis (R Development Core Team, 2008).

3.

Results

3.1. Physical and chemical variables in the littoral zone close to the sediment Physical and chemical variables were influenced by a lowering of the water level in 2008 due to restoration work at the dam (Table 2). Consequently, the sampling period can be subdivided in two phases: a rather long period with normal water level (from 2006 through the end of June 2008 and during 2009) and a short period characterised by a reduced water level in 2008. Phytobenthos were found at depths between 1.75 m and 11 m (mean 4.1 m). Water temperatures at the littoral sites were affected by sampling depths and seasonal influences. The concentrations of dissolved oxygen were normally above 6 mg L1 (only 3% of all measured concentrations were lower). However, O2-concentrations sporadically dropped down during the low water level phase. Phosphorus concentrations (SRP, TDP) were generally very low. Enhanced phosphorus concentrations were observed during the low water level phase. Conversely, nitrate was reduced during this period. The pH values ranged from 6.5 to 8.7 without remarkable changes during the low water phase, and concentrations of dissolved iron ranged from below 10 to greater than 1000 mg L1. Highest concentrations of dissolved iron were detected during the low water level phase. Light reached the near-bottom layer at all sampling sites. The lowering of the water level resulted in a decrease of light intensity.

3.2. Quantification of microphytobenthos using chlorophyll a, phycocyanin and phycoerythrin concentrations Microphytobenthos mainly consisted of cyanobacteria of the genera Oscillatoria, Phormidium and Pseudanabaena. A quantification was performed by analysing the concentrations of chlorophyll a, phycocyanin and phycoerythrin. In general, pigment

4976

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 7 3 e4 9 8 2

Table 2 e Water level, abiotic variables in the near-bottom layer of the littoral and pigment concentrations of microphytobenthos in Wahnbach Reservoir during the corresponding sampling periods from 2006 to 2009. All values (except of water level) are given as mean based on values from eight sampling sites (L1eL8 see Fig. 1) during the respective sampling period with minimum and maximum values (in parentheses). Sampling period

11.07e10.10.2006

12.06e09.10.2007

Water level (m above sea) Water temperature (
C) Dissolved oxygen (mg L1) SRP (mg L1)a TDP (mg L1)a Nitrate-N (mg L1) Ammonium-N (mg L1) Dissolved iron (mg L1)b pH Irradiance (mE m2 s1) Chlorophyll a (mg (mg wet weight sediment)1) Phycocyanin (mg (mg wet weight sediment)1) Phycoerythrin (mg (mg wet weight sediment)1)

118 (105e124)

119 (110e123)

14.8 (5.2e25.2) 9.5 (7.7e11)

01.07e04.11.2008 low water level phase

26.05e18.08.2009

121 (111e123)

103 (100e110)

119 (118e121)

16.6 (11e21.6) 10.0 (0.3e13.3)

12.6 (6.5e19.8) 11.1 (7.8e14.8)

16.9 (10.5e22.2) 7.9 (0.4e11.8)

14.9 (6.2e20.6) 11.8 (4.1e16.2)

<8 <8 (12) 2.0 (0.5e2.5) 141 (14e1190) 36 (<1036) 7.2 (6.5e8.7) no data

<8 <8 2.5 (2.3e2.9) 39 (17e87) 10 (<1026) 7.4 (6.6e8.6) no data

<8 <8 (7) 2.4 (1.2e2.8) 55 (18e126) 11 (<1031) 7.4 (6.5e8.3) 83 (0.3e180)

16 (<853) 42 (<899) 1.2 (0.02e2.2) 685 (22e3600) 214 (17e1754) 7.4 (6.9e8.4) 6 (0.1e27)

<8 <8 2.3 (1.9e2.6) 42 (19e125) <10 (11) 7.3 (6.6e7.9) 64 (4e190)

no data

242 (0.1e6275) max in September 39 (0.2e620) max in September 43 (0.2e1219) max in September

147 (2e405) max in May 0.5 (0.8e2) max in May 0.02 (0.02e0.03) max in May

105 (0.2e1619) max in September 0.5 (0.2e1.3) max in November not detected

210 (0.2e1438) max in August 11 (0.3e136) max in May 10 (0.1e108) max in May

no data no data

20.05e02.06.2008

a detection limit 8 mg L1. b detection limit 10 mg L1.

concentrations varied within and between the years (Table 2). Chlorophyll a was maximum in late summer. In 2007, concentrations of phycocyanin and phycoerythrin also reached their maximum during that time (September). However, a decrease of phycocyanin and phycoerythrin concentrations was apparent in 2008. The reduction of phycocyanin- and phycoerythrincontaining microphytobenthos was pronounced during the low water level phase. Phycoerythrin was actually below the analytical detection limit during this phase. In 2009, phycocyanin and phycoerythrin concentrations were lower than in 2007.

3.3. Cell-bound and dissolved geosmin in the littoral of Wahnbach Reservoir Both cell-bound and dissolved geosmin concentrations were detected in littoral samples. By contrast, MIB was never found. Extracellularily dissolved geosmin was detected in 25% of the samples, only in 2006 and 2007 (Fig. 2). Maximum concentrations occurred in August 2006 and 2007, respectively. Dissolved geosmin varied with time and littoral site, thereby the local variance (33% of total variance) matched almost the temporal variance (27% of total variance). Dissolved geosmin of the inflowing water was measured, however, geosmin was never detected. The highest values of cell-bound geosmin concentration were observed at the end of summer (Fig. 3). In 2008, cell-bound geosmin concentrations were remarkably reduced compared to 2007. This trend continued, as cell-bound geosmin was again very low during 2009. The variance of cell-bound geosmin was primarily influenced by the temporal variations (70% of total variance). Only 14% of the total variance was determined by the variance between the littoral sites. Nevertheless, cell-bound

concentrations at sites near the inflow were slightly enhanced compared to the others.

3.4. Multivariate characterisation of samples with/ without cyanobacteria and geosmin by principal component analysis (PCA) The first two dimensions of PCA explain 44% of the total variance (Fig. 4). Phases with high and low water level were separated by the first dimension, a reduction of the water level was associated with an increase in concentration of phosphorus species (SRP, TDP) and dissolved iron ( p < 0.001). Nitrate and dissolved oxygen were reduced during this phase ( p < 0.001). The second dimension specified samples by their chlorophyll a, phycocyanin and geosmin concentrations. Cell-bound geosmin correlates positively with dissolved geosmin ( p ¼ 0.0001). Both geosmin concentrations (cell-bound and dissolved) were enhanced when phycocyanin was high ( p ¼ 0.0002, p ¼ 0.025, respectively). No significant relation was found between geosmin concentrations and chlorophyll a ( p > 0.7). The impact of water level fluctuation on cyanobacteria, phycocyanin and geosmin was analysed in more detail. During the time of reduced water level phycocyanin was diminished ( p ¼ 0.09, Fig. 5). Periods with positive findings of benthic cyanobacteria were less frequent. Moreover, during this time concentrations of cell-bound geosmin were reduced ( p ¼ 0.03, Fig. 5). Dissolved geosmin was not detected (see 3.3). The lowering of the water level was further related to a reduction of the Secchi disk depth and to a lowering of light intensity which reached the sediment. A significant positive relation of phycocyanin to light intensity ( p ¼ 0.006) points to an influence of light on cyanobacteria growth. On the other hand, no significant relation of phycocyanin to Secchi disk

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 7 3 e4 9 8 2

4977

Fig. 2 e Concentration of dissolved geosmin in littoral samples from Wahnbach Reservoir during 2006 and 2007 (mean and standard error of all dates (right) and all sampling sites (left), respectively). In 2008, 2009 dissolved geosmin was not detected.

depth was found, and neither cell-bound nor dissolved geosmin was significantly influenced by light intensity. Rather geosmin concentrations were still low after the water level rose (see Fig. 3).

were obviously caused by a concentration effect due to the extremely reduced water volume.

4. 3.5. Worste-case estimation of odour events in Wahnbach Reservoir Geosmin was detected in the littoral, but never in the pelagic zone. By these findings, however, the potential occurrence of odour nuisances in raw water, which is abstracted from the hypolimnic pelagic zone, cannot be excluded. A worst-case situation is conceivable that under specific circumstances the total geosmin accumulated in the littoral could be released and homogeneously distributed throughout the pelagic zone. This worst-case estimation was specified from measured benthic geosmin concentrations. The calculation was performed by multiplying the proportion of the area covered with microphytobenthos (estimated by diving) relative to the volume of the reservoir by the area-related total (cell-bound plus dissolved) geosmin concentrations. Cumulative frequency distributions of these calculated concentrations can be compared with the odour threshold concentration ¨ mu¨r-O ¨ zbek and Dietrich (2005) (Fig. 6). Accordaccording to O ingly, the probability that the odour threshold of geosmin could be exceeded was low (2007) or lacking at all (2009). Only in 2008, the year with the low water level, a substantial probability (about 30%) did exist that geosmin concentrations in the raw water could exceed the odour threshold, although total geosmin concentrations in the littoral were relatively low (Figs. 2 and 3). Higher estimated pelagic geosmin concentrations in 2008 compared to the other years as shown in Fig. 6

Discussion

4.1. Spatio-temperal pattern of cell-bound and dissolved benthic geosmin The spatio-temporal distribution of geosmin in the littoral zone of Wahnbach Reservoir was investigated by a four-years survey revealing that geosmin episodes occurred most likely in late summer. Consistent with the cell-bound geosmin concentration, the dissolved fraction was also maximum in August in Wahnbach Reservoir in 2006 and 2007 (Figs. 2 and 3). Comparable investigations regarding the pattern of benthic geosmin in reservoirs are not known, other investigations were focused on geosmin in the pelagic zone or on benthic geosmin in rivers (Sabater et al., 2003). The pelagic studies in other lakes also revealed late-summer peaks of geosmin (e.g. Sugiura et al., 1998) or MIB (Westerhoff et al., 2005). Other studies, however, showed additional geosmin peaks during winter (Suigiura et al., 2004; Dzialowski et al., 2009) and recommended involving winter data in future predictions models (Dzialowski et al., 2009). Thus, only a study which comprehends the entire season can unambiguously clarify whether geosmin occurs in Wahnbach Reservoir during winter. However, since routine flavour profiles corresponded well to the analysis of dissolved geosmin during the sampling period of the present investigation and did not indicate perceivable odorous concentrations during winter, the occurrence of geosmin peaks in Wahnbach Reservoir in winter is rather unlikely.

4978

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 7 3 e4 9 8 2

Fig. 3 e Concentrations of cell-bound geosmin in littoral samples from Wahnbach Reservoir during 2007e2009 (mean and standard error of all dates (right) and all sampling sites (left), respectively). Note the different scaling of the ordinate.

Cell-bound geosmin increased during summer as phycocyanin did (compare Fig. 3 and Table 2). As Fig. 4 shows that these two variables are closely related it can be concluded that geosmin in Wahnbach Reservoir was primarily produced by phycocyanin-containing cyanobacteria. On the basis of a field study, it is difficult to decide whether cell-bound geosmin rose due to the increase of benthic cyanobacterial biomass, to environmental influences or to the change within the benthic community. Watson and Ridal (2004) ascribed seasonal dynamics to changes in cell production and/or to the succession of different taxa. Among different taxa a tremendous range in instrinsic capacity for production of metabolites such as geosmin and MIB was documented (Watson, 2003). Cell-bound geosmin in our study was calculated as specific concentration (per unit of chlorophyll a). Consequently, the increase in cell-bound geosmin was more likely due to enhanced cellular production or to an increase in the relative abundance of geosmin-producing species rather than to an increase in cyanobacterial biomass. In contrast to the temporal pattern, spatial variations of cell-bound geosmin were minor. Accordingly, it is concluded that conditions for growth and production at all sampling

sites were comparable. On the other hand, spatial inhomogeneities may occur when algal mats were detached and translocated due to the formation of gas bubbles by photosynthesis (Ju¨ttner and Watson, 2007). A detachment of cyanobacterial mats and the emanating floating might result in enhanced odour concentrations as it was observed by Vilalta and Sabater (2005). However, floating mats were never observed in Wahnbach Reservoir.

4.2. Direct and indirect influences on the production of benthic geosmin No significant influences of abiotic environmental variables on cell-bound and dissolved geosmin were found (Fig. 4 left). However, the lowering of the water level caused changes of different abiotic variables probably having an indirect influence on benthic geosmin. For instance, a reduction of cell-bound geosmin and phycocyanin was indicated by splitting geosmin and phycocyanin into two groups separated by water level (Fig. 5). In addition, during the low water level phase an increase of phosphorus and iron concentrations (see Table 2, Fig. 4) but also a decline of light intensity within the

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 7 3 e4 9 8 2

4979

Fig. 4 e Loadings (left) and scores (right) of littoral samples from Wahnbach Reservoir during 2006e2009 on the bidimensional plane defined by the first two components of the PCA. Space reduction from 14 to 2 dimensions (44% of the total variance). Samples were classified as samples with and without cyanobacteria, and with and without geosmin, respectively. Variables are cell-bound geosmin (geosminint), dissolved geosmin (geosminext), phycocyanin (pc), chlorophyll a (chla), water temperature (temp), dissolved iron (Fe), TDP, SRP, sampling depth, year, dissolved oxygen (O2), nitrate (NOL 3 ), water level, presence/absence of cyanobacteria (cyano).

near-bottom layer (Table 2) was observed. Moreover, turbidity due to resuspension of bottom sediment was enhanced (data not shown). Also in other lakes, under turbid (and more eutrophic) conditions the growth of benthic algae was observed to be limited by phytoplanktonic shading kova´ et al., 2008). A shift from benthic to pelagic (Poulic primary production with increasing eutrophication was revealed by Vadeboncoeur et al. (2001). As phytoplankton chlorophyll a in Wahnbach Reservoir was enhanced immediately after the water level was lowered (1.8 mg L1 from July through November 2008 compared to 0.7 mg L1 as an average in 2007 before lowering the water level) an adverse influence of phytoplankton by shading on benthic growth is most likely. Indeed, from net growth rates of benthic cyanobacteria in Wahnbach Reservoir (calculated from phycocyanin concentrations) it can be deduced that the decrease of cyanobacteria during the low water level phase was due to growth limitation rather than due a lower inocolum. Thus, benthic geosmin during the low water level phase was indirectly reduced because of enhanced phosphorus concentrations during this

phase. The indirect effect consists of phosphorus-favoured planktonic growth which results in a limitation of phytobenthos growth due to shading by phytoplankton. Concerning influences on geosmin production, it is important to differentiate between environmental impacts on growth of geosmin-producing cyanobacteria and direct influences on the cellular production of geosmin. Concerning the water quality problems, both processes are of interest. Regarding the understanding why and when during cyanobacteria growth geosmin (and other odorous metabolites) are produced, a differentiated analysis of direct and indirect effects is needed. In order to estimate a probable nuisance emanating from the metabolites, it is important to know the specific capacity of cells for metabolite production, possible triggers of metabolite production and growth characteristics of the producer (Watson et al., 2008). However, answers to these topics can only be obtained by performing growth experiments with geosmin-producing species, which was beyond the scope of this investigation.

4.3.

Fig. 5 e Phycocyanin and cell-bound geosmin concentration in dependence on water level. Boxplots with median and 25th, 75th percentiles, whiskers are the 10th, 90th percentiles, circles represent the 5th, 95th percentiles.

From benthic geosmin to geosmin in raw water

The potential of benthic cyanobacteria to produce odour compounds such as geosmin or MIB is well known for a rather long time (Lechevalier and Eren, 1970; Izaguirre et al., 1983; Berglind et al., 1983). Recent investigations corroborated and specified these findings (e.g. Sugiura et al., 1998; Watson and Ridal, 2004; Baker et al., 2006). However, the question may be raised whether or not this type of water quality problem can occur in an oligotrophic water body having very low phosphorus input due to a long-run phosphorus elimination plant (Bernhardt et al., 1985). In full accordance with this question, water quality problems in Wahnbach Reservoir due to offflavours were extremely rare. However, as already

4980

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 7 3 e4 9 8 2

Fig. 6 e Potential cumulative frequency of geosmin in Wahnbach Reservoir under the assumption that all cell-bound geosmin measured in the littoral would be released and together with the measured dissolved geosmin completely mixed into the entire reservoir water (worst-case scenario).

mentioned above a worst-case situation is conceivable that under specific circumstances the total geosmin produced in the littoral by benthic cyanobacteria could be released and homogeneously distributed throughout the pelagic zone. In order to validate the probability of such potential odorous events the pelagic geosmin concentrations (see Fig. 6) were estimated from measured dissolved and cell-bound geosmin concentrations in the littoral zone (Figs. 2 and 3). These concentrations are high (in the range of mg per litre or mg (mg chla)1, respectively). However, they are comparable to concentrations found in other studies (Jones and Korth, 1995; Ju¨ttner and Watson, 2007). Under the assumption that all benthic geosmin is released (worst-case situation), the probability that the odour threshold would be exceeded is subjected to a considerable interannual variation in Wahnbach Reservoir (see Fig. 6). Thus, the probability was low in 2007, a year which is characterised by a normal water level, and by relatively high geosmin and pigment concentrations (see Figs. 2 and 3, Table 2). However, the ratio between the littoral area covered by cyanobacteria and the reservoir water volume was low enough resulting in such a low probability. If the ratio increases as observed during the low water level phase in 2008, the probability rose although geosmin concentrations were relative low. Unexpectedly and surprisingly, the probability was zero in 2009, a year with normal water level. In contrast to the previous years the biomass of cyanobacteria was lower (measured by the phycocyanin and phycoerythrin concentrations, see Table 2) indicating that growth of cyanobacteria was limited after the reservoir was refilled. We assume a probable reason for growth limitation was the restricted sediment phosphorus availability following the intensive contact of the littoral sediment with air and subsequent iron oxidation during the low water level phase. Measured phosphorus concentrations in Wahnbach Reservoir were generally very low (Table 2). Within this concentration range the assumed decrease cannot be detected. However, the probably restricted phosphorus availability at the sediment surface is indicated by the lowest concentrations of dissolved iron during the investigation period (Table 2). Likewise, after refilling another reservoir a lack of dissolved phosphorus and

iron compounds within the upper centimetres of the sediment was detected by sediment peeper measurements (Bautzen Reservoir, Germany: M. Wetzel, Institute for Hydrobiology, TU Dresden, pers. comm). The estimation is a worst-case calculation because only a portion of geosmin is really dissolved (e.g. Watson et al., 2008), and losses due to decomposition or biodegradation (Westerhoff et al., 2005; Korth et al., 1992) were not taken into account. Nevertheless, the probability of odour events during the investigation period was obviously controlled by variables influencing the growth of cyanobacteria (light, phosphorus) and by the water volume of the reservoir which has, on the one hand, a direct dilution effect, and on the other hand, an indirect effect via the variables (light, phosphorus) controlling the growth of benthic cyanobacteria.

5.

Conclusions

The littoral slopes of Wahnbach Reservoir were shown to be colonised by benthic cyanobacteria, and some species were likely the source of the odour compound geosmin detected in littoral samples. Throughout this study, MIB was never detected in any sample. Although cell-bound specific geosmin concentrations (mean 22.6 mg (mg chla)1, maximum 1152 mg (mg chla)1) were comparable to concentrations found in other studies the occurrence of odour episodes in Wahnbach Reservoir was rare. From the pattern of geosmin production and release found in Wahnbach Reservoir odour events are, if any, most probable in late summer. A worst-case estimation of the probability of odour nuisances substantiate that the ratio of biomass of geosmin-producing benthic cyanobacteria to the reservoir water volume (or filling status) was decisive for the probability of an off-flavour impact on raw water. The probability is rather high when the water volume is low, and it is very low when either the water volume is high or the biomass of benthic cyanobacteria is reduced. If a high probability is predicted, a short-term change of the reservoir water level appears to be the optimal control measure that leads to a suppression of growth of benthic cyanobacteria either by an

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 7 3 e4 9 8 2

increase of water turbidity (drawdown) or sedimentphosphorus limitation (refilling after drawdown).

Acknowledgements We thank the laboratory of the Wahnbach Reservoir Association for analytical work, and Gerhard Linden (Wahnbach Reservoir Association) and the diving team from the German Lifeguard Association (Rhine-Sieg district) for supporting the sampling campaigns. Special thanks to Janice Hegewald for linguistic improvements, and to the two anonymous reviewers for helpful comments on an earlier version of the manuscript. The study has been supported by the Wahnbach Reservoir Association.

references

Baker, L.A., Westerhoff, P., Sommerfeld, M., 2006. Adaptive management using multiple barriers to control tastes and odors. J. Am. Water Works Assoc. 98 (6), 113e126. Berglind, L., Johnsen, I., Ormerod, K., Skulberg, O., 1983. Oscillatoria brevis (Ku¨tz.) Gom. and some other especially odouriferous benthic cyanophytes in Norwegian inland waters. Wat. Sci. Tech. 15 (6/7), 241e246. Bernhardt, H., Clasen, J., Hoyer, O., Wilhelms, A., 1985. Oligotrophierung stehender Gewa¨sser durch chemische Na¨hrstoff-Eliminierung aus den Zuflu¨ssen am Beispiel der Wahnbachtalsperre. Arch. Hydrobiol. 70, 481e533 (Monographische Beitra¨ge). Din En Iso 13395, 1996. Water quality e Determination of Nitrite Nitrogen and Nitrate Nitrogen and the Sum of Both by Flow Analysis (CFA und FIA) and Spectrometric Detection, German standard methods for the examination of water, waste water and sludge, p. D28. Din En Iso 11732, 2005. Water Quality e Determination of Ammonium Nitrogen e Method by Flow Analysis (CFA and FIA) and Spectrometric Detection, German standard methods for the examination of water, waste water and sludge, p. E23. Din En Iso 11885, 2009. Water Quality e Determination of Selected Elements by Inductively Coupled Plasma Optical Emission Spectrometry (ICP-OES), German standard methods for the examination of water, waste water and sludge, p. E 22. (former version: DIN EN ISO 11885 1998-04). Dzialowski, A.R., Smith, V.H., Huggins, D.G., deNoyelles, F., Lim, N.C., Baker, D.S., Beury, J.H., 2009. Development of predictive models for geosmin-related taste and odor in Kansas, USA, drinking water reservoirs. Water Res. 43, 2829e2840. En Iso 15681-2, 2004. Water quality e Determination of Orthophosphate and Total Phosphorus Contents by Flow Analysis (FIA and CFA) e Part 2: Method by Continuous Flow Analysis (CFA), German standard methods for the examination of water, waste water and sludge, p. D 46. Izaguirre, G., Hwang, C.J., Krasner, S.W., McGuire, M.J., 1983. Production of 2-methylisoborneol by benthic cyanophyta. Wat. Sci. Tech. 15 (6e7), 211e220. Izaguirre, G., Taylor, W.D., 2004. A guide to geosmin- and MIB-producing cyanobacteria in the United States. Water Sci. Tech. 49, 19e24. Jones, G.J., Korth, W., 1995. In-situ-production of volatile odor compounds by river and reservoir phytoplankton populations in Australia. Wat. Sci. Tech. 31 (11), 145e151.

4981

Ju¨ttner, F., Watson, S.B., 2007. Biochemical and ecological control of geosmin and 2-methylisoborneol in source waters. Appl. Environ. Microbiol. 73, 4395e4406. Krishnani, K.K., Ravichandran, P., Ayyappan, S., 2008. Microbially derived off-flavor from geosmin and 2-methylisoborneol: sources and remediation. Rev. Environ. Contam. Toxicol. 194, 1e27. Koma´rek, J., Anagnostidis, K., 2005. Cyanoprokaryota 2. Teil/ 2nd part: oscillatoriales. In: Bu¨del, B., Krienitz, L., Ga¨rtner, G., Schagerl, M. (Eds.), Su¨sswasserflora von Mitteleuropa 19/2. Elsevier/Spektrum, Heidelberg, p. 759. Korth, W., Ellis, J., Bowmer, K.H., 1992. The stability of geosmin and MIB and their deuterated analogues in surface waters and organic solvents. Wat. Sci. Tech. 25 (2), 115e122. Lechevalier, H., Eren, J., 1970. In: Shuval, H.I. (Ed.), Developments in Water Quality Research: Proceedings of the Jerusalem International Conference on Water Quality and Pollution Research. Ann Arbor-Humphrey Science Publ.. Liboriussen, L., Jeppesen, E., 2003. Temporal dynamics in epipelic, pelagic and epiphytic algal production in a clear and turbid shallow lake. Freshw. Biol. 48, 418e431. Lowe, R.L., 1996. Periphyton patterns in lakes. In: Stevenson, R.J., Bothwell, M.L., Lowe, R.L. (Eds.), Algal Ecology. Freshwater benthic ecosystems. Academic press, San Diego. Minkova, K.M., Tchernov, A.A., Tchorbadjieva, M.I., Fournadjieva, S.T., Antova, R.E., Busheva Ch, M., 2003. Purification of C-phycocyanin from Spirulina (Arthrospira) fusiformis. J. Biotechnol. 102, 55e59. Otsuki, A., Omi, T., Hashimoto, S., Aizaki, M., Takamura, N., 1994. HPLC fluorometric-determination of natural phytoplankton phycocyanin and its usefulness as cyanobacterial biomass in highly eutrophic shallow lake. Water Air and Soil Poll. 76, 383e396. ¨ mu¨r-O ¨ zbek, P., Dietrich, A.M., 2005. Determination of O temperature-dependent Henrys law constants of odorous contaminants and their application to human perception. Environ. Sci. Technol. 39, 3957e3963. Peter, A., Ko¨ster, O., Schildknecht, A., von Gunten, U., 2009. Occurrence of dissolved and particle-bound taste and odor compounds in Swiss lake waters. Water Sci. Technol. 43 (8), 2191e2200. kova´, A., Ha Poulic sler, P., Lysa´kova´, M., Spears, B., 2008. The ecology of freshwater epipelic algae: an update. Phycologia 47 (5), 437e450. R Development Core Team, 2008. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. URL. http://www.r-project.org/. Sabater, S., Vilalta, E., Gaudes, A., Guasch, H., Munoz, I., Romani, A., 2003. Ecological implications of mass growth of benthic cyanobacteria in rivers. Aquat. Microb. Ecol. 32 (2), 175e184. Smith, J.L., Boyer, G.L., Zimba, P.V., 2008. A review of cyanobacterial odorous and bioactive metabolites: impacts and management alternatives in aquaculture. Aquaculture 280, 5e20. Sugiura, N., Iwami, N., Inamori, Y., Nishimura, O., Sudo, R., 1998. Significance of attached cyanobacteria relevant to the occurrence of musty odor in Lake Kasumigaura. Wat. Res. 32, 3549e3554. Suigiura, N., Utsumi, M., Wei, B., Iwami, N., Okano, K., Kawauchi, Y., Maekawa, T., 2004. Assessment for the complicated occurrence of nuisance odours from phytoplankton and environmental factors in a eutrophic lake. Lake Res. Res. Managn. 9, 195e201. Vadeboncoeur, Y., Lodge, D.M., Carpenter, S.R., 2001. Whole-lake fertilization effects on distribution of primary production between benthic and pelagic habitats. Ecology 82 (4), 1065e1077. Vilalta, E., Sabater, S., 2005. Structural heterogeneity in cyanobacterial mats is associated with geosmin production in rivers. Phycologia 44, 678e684. Watson, S.B., 2003. Cyanobacterial and eukaryotic algal odour compounds: signals or byproducts? A review of their biological activity. Phycologia 42 (4), 332e350.

4982

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 7 3 e4 9 8 2

Watson, S.B., 2004. Aquatic taste and odor: a primary signal of drinking water integrity. J. Toxicol. Environ. Health A 67, 1779e1795. Watson, S.B., Ridal, J., 2004. Periphyton: a primary source of widespread and severe taste and odour. Water Sci. Technol. 49 (9), 33e39. Watson, S.B., Ridal, J., Boyer, G.L., 2008. Taste and odour and cyanobacterial toxins: impairment, prediction, and management in the great lakes. Can. J. Fish. Aquat. Sci. 65, 1779e1796.

Westerhoff, P., Rodriguez-Hernandez, M., Baker, L., Sommerfeld, M., 2005. Seasonal occurrence and degradation of 2-methylisoborneol in water supply reservoirs. Water Res. 39, 4899e4912. Wood, S., Williams, S.T., White, W.R., 2001. Microbes as a source of earthy flavours in potable water - a review. Int. Biodeterior. Biodegrad. 48 (1e4), 26e40. Wu, J.T., Ju¨ttner, F., 1988. Differential partitioning of geomin and 2-methylisoborneol between cellular-constituents in Oscillatoria tenuis. Arch. Microbiol. 150, 580e588.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 8 3 e4 9 9 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Network condition simulator for benchmarking sewer deterioration models A. Scheidegger, T. Hug, J. Rieckermann, M. Maurer* Eawag, Swiss Federal Institute of Aquatic Science and Technology, U¨berlandstrasse 133, 8600 Du¨bendorf, Switzerland

article info

abstract

Article history:

An accurate description of aging and deterioration of urban drainage systems is necessary for

Received 7 October 2010

optimal investment and rehabilitation planning. Due to a general lack of suitable datasets,

Received in revised form

network condition models are rarely validated, and if so with varying levels of success. We

30 June 2011

therefore propose a novel network condition simulator (NetCoS) that produces a synthetic

Accepted 1 July 2011

population of sewer sections with a given condition-class distribution. NetCoS can be used to

Available online 13 July 2011

benchmark deterioration models and guide utilities in the selection of appropriate models and data management strategies. The underlying probabilistic model considers three main

Keywords:

processes: a) deterioration, b) replacement policy, and c) expansions of the sewer network.

Sewerage

The deterioration model features a semi-Markov chain that uses transition probabilities

Deterioration model

based on user-defined survival functions. The replacement policy is approximated with

Semi-Markov chain

a condition-class dependent probability of replacing a sewer pipe. The model then simulates

Asset management

the course of the sewer sections from the installation of the first line to the present, adding

Pipe condition inspection

new pipes based on the defined replacement and expansion program. We demonstrate the usefulness of NetCoS in two examples where we quantify the influence of incomplete data and inspection frequency on the parameter estimation of a cohort survival model and a Markov deterioration model. Our results show that typical available sewer inventory data with discarded historical data overestimate the average life expectancy by up to 200 years. Although NetCoS cannot prove the validity of a particular deterioration model, it is useful to reveal its possible limitations and shortcomings and quantifies the effects of missing or uncertain data. Future developments should include additional processes, for example to investigate the long-term effect of pipe rehabilitation measures, such as inliners. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Urban drainage systems have an extended life span and are typically designed for future load conditions to be reached in about 25e50 years, although they may very well be in use much longer. Consequently, structures that were built in the 1960s and 1970s are reaching their end-of-life and the maintenance and proper rehabilitation of infrastructures is now becoming increasing important (Davies et al., 2001a; Maurer, 2009; Wirahadikusumah et al., 2001).

Rehabilitation measures are necessary because the deterioration of sewers can cause hydraulic and static deficiencies that give rise to urban flooding and safety problems (Clegg et al., 1989; Delleur, 1994; Djordjevic et al., 2005; Saegrov et al., 1999; Schmitt et al., 2004) as well as health hazards like groundwater pollution due to leaky sewers (Rutsch et al., 2008). In addition, however, inefficient maintenance strategies can lead to economic bottlenecks, often due to insufficient inspection planning (Baur and Herz, 2002; Herz and Lipkow, 2002). Therefore, the proper management of sewer assets

* Corresponding author. Tel.: þ41 44 823 5386; fax: þ41 44 823 5389. E-mail address: [email protected] (M. Maurer). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.008

4984

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 8 3 e4 9 9 4

Nomenclature a Unit used for year CS Condition state (1 ¼ best, 5 ¼ worst) t Point in time t.built Year of construction q ¼t  t.built Age, time since construction ðtÞ Condition state in year t of section i, built in t.built CSt:built i j ¼ 1,.,5 Index of condition states (1 ¼ best, 5 ¼ worst) Survival function, probability that a pipe section of Sj(q) age q has CS j or better Parameter vector of Sj(q) of Cohort model (Baur bj and Herz, 2002) Functional life-span of pipe section i, built in Tit:built t.built Time a pipe section remains in CS j TCS j TCS1/j Total time a pipe section has spent in CS j or better

challenges us to accurately assess the rehabilitation needs of urban drainage systems in the future. A wide variety of deterioration models has been proposed in the literature (see ‘Literature review’ section), with the general aim of predicting the deterioration process of populations of pipes or individual lines. Although the proposed models differ in i) the mathematical description of the deterioration process, ii) the data requirements, and iii) the mode of calibration, they all require some form of adjustment to real-world data to produce meaningful predictions of future condition states (CS). However, a recent review by Ana and Bauwens (2010) suggests that most of the proposed models fail to show that they can adequately forecast future condition states. We agree with the authors that a major reason for this failure is the lack of complete and reliable datasets. In addition to insufficient data management, considerable errors in the derivation of CS from CCTV inspections (Dirksen et al., 2007) can create substantial uncertainty in the available data. In this paper, we therefore propose a novel simulator which generates a synthetic dataset of sewer inventories that can be used to test deterioration models. We demonstrate by two examples how the synthetic data can help to identify weaknesses and errors due to missing data and a data bias. The main scientific innovation of our work is the simulator, which includes not only the deterioration but also relevant processes such as system growth and replacement of pipes, and can be used to generate datasets on different scales and with arbitrary resolution. This represents a significant improvement, because it i) allows scientists and utility managers to identify limitations of current deterioration models, and ii) guides data management in terms of appropriate time-intervals as well as attributes that have to be collected for reliable model applications. It is important to note that NetCoS cannot verify a model, but can reveal its limitations and weaknesses. The manuscript is structured as follows. First, we will briefly review current deterioration models and use this as a basis to describe our method in detail. We will then demonstrate the usefulness of this approach with two examples. Our results show that current datasets lead to a severe over-estimation of life-spans if information on past growth and replacement is

ðtÞ Probability vector of pipe section i, built in t.built, pt:built i probabilities for each CS in year t P(q) Transition probability matrix for a pipe section of age q ðtÞ ¼ Iðt < Tit:built Þ Indicator of pipe section i, built in dt:built i t.built is in use in year t rep rep rep rep rep prep ¼ ðpCS1 ; pCS2 ; pCS3 ; pCS4 ; pCS5 Þ Vector with replacement probabilities for each condition state. See Eq. (8) Lexp (t) Length of expansion in year t Length of all pipe sections replaced in year t Lrep (t) Number of pipe sections built in t.built Nt.built i ¼ 1,.,Nt.built Index of all pipe sections built in year t.built Length of pipe section i, built in t.built Lt:built i L Average length of a pipe section D Numerical time step for Markov chain (D< < 1 year)

not available. Ultimately, we discuss important benefits and limitations of our approach and draw our final conclusions.

2.

Literature review

As described above, information on future infrastructure conditions is essential for the efficient management of infrastructure assets. The Water Research Center systematically investigated the deterioration of rigid sewers in the UK and concluded that their deterioration is best described as a probabilistic process and determined by random events (WRc, 1986). Not surprisingly, a large number of deterioration models are described in the literature. In the following, we would like to give a short overview of the models available, the explanatory variables and the influence of data availability on the application of these models. The aim of this overview is to show that there is no shortage of modeling approaches, but a widespread lack of methods and tools to evaluate and validate these models. More detailed reviews can be found in Ana and Bauwens (2010), Kleiner and Rajani (2001), Korving (2004), Madanat et al. (1995) and Tran (2007).

2.1.

Sewer deterioration models

According to Morcous et al. (2002b), we can categorize the models into three mutually non-exclusive groups (see also overview in Ana and Bauwens, 2010):

2.1.1.

Polynomial-type models

These use continuous functions to describe the effect of factors affecting the condition, also called explanatory variables, of the assets. They can be based on physico-chemical deterioration processes, but often just use a correlation approach to gain additional insights. Examples are linear models (Madanat et al., 1995), exponential models (Wirahadikusumah et al., 2001), and polynomial regression models (Chughtai and Zayed, 2008; Savic et al., 2009). The single biggest disadvantage of these approaches is that continuous functions are inappropriate for representing discrete ordinal measures. Condition ratings

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 8 3 e4 9 9 4

generally only indicate a relative ordering with no or little meaning assigned to the distance between the condition ratings, which contradicts the fundamental assumption behind these continuous functions. An approach designed to overcome this disadvantage can be found in Madanat et al. (1995), who developed an ordered probit model to map a linear deterioration model to discrete condition values.

2.1.2.

Artificial intelligence models

These require no assumptions about the model structure and are purely data or information-driven. Model outputs are classified from a set of input variables by learning from the available data. Examples for such supervised learning algorithms are case-based reasoning (Hahn et al., 2002; Morcous et al., 2002a, 2002b), fuzzy-based approaches (Kleiner et al., 2006; Makropoulos et al., 2003; Yan and Vairavamoorthy, 2003) and neural networks (Najafi and Kulandaivel, 2005). These types of models can handle ordinal condition ratings and non-linear deterioration behavior. However, their obvious disadvantage is the large amount of data needed for training and calibration, and in the case of neural networks, sophisticated methods for the optimization procedure to find global optima. Tran (2007) found in a comparison of various deterioration models that a neural network was the most suitable model for predicting the condition changes of individual pipes for both structural and hydraulic deterioration.

2.1.3.

Stochastic or probabilistic models

These assume a probabilistic relationship between input and output data. The types of models commonly found in the deterioration literature are ordinal regression methods, stochastic duration models and Markov chain models. Ordinal (logistic) regression models are widely used to identify the explanatory variables for structural pipe deterioration (see below). Duration or time-based models are typically used in survival statistics and apply a probability distribution of time spent to undergo a state change. A strong benefit of these types of models is the sound handling of repeated observations and, depending on the model, censored data. Applications are described by Mishalani and Madanat (2002) using a Weibull distribution, and by Baur and Herz (2002) using an infrastructure-specific survival function (Herz, 1998). Markov chain models describe the probability of a transition from one condition state to another over a unit of time. An early paper adopting Markov chain based deterioration modeling from bridge management for sewers was written by Micevski et al. (2002). Mishalani and Madanat (2002) classify Markov chains as state-based models, whereas stochastic duration models are time-based models. They are mathematically similar, as it is possible to convert condition-state transition probabilities into probability density functions of state durations and visa versa (Mauch and Madanat, 2001).

2.2.

Data availability

The main challenge for Markov models is to estimate the transition probability matrix from scarce sample data. The

4985

calibration techniques depend strongly on the quality of the data and their frequency of collection. Proper calibration and testing of Markov models require condition data from three consecutive periods (Madanat and Ibrahim, 1995; Wirahadikusumah et al., 2001). Unfortunately, these are not typically available for urban drainage systems, which are hazardous environments and also difficult to access. In addition, condition data are only present as a snapshot or a sample of pipes inspected only once, and contain considerable errors (Ana and Bauwens, 2010; Baik et al., 2006; Kathula, 2001; Kleiner and Rajani, 2001; Tran, 2007; Wirahadikusumah et al., 2001). Furthermore, many utilities base their inspection frequency on a risk assessment or the relevance of the sewer line, and therefore reduce data availability even further (Woodhouse et al., 2008). Hafskjold and VanrenterghemRaven (2007) report from CARE-W, an earlier EU research project on sewer rehabilitation: “[.] even large utilities have very little overview of their physical assets, and in many cases the installation year, the pipe material, and even the location of pipes is unknown [.]. In other cases, data is present, but in a format which makes it very difficult to use for analyses [.]. Data may also be inaccurate or have lacunae.”

2.3. Explanatory variables for structural conditions of sewers The use of probabilistic models also raises the question of the relevant explanatory variables. As could be expected from the conclusion of the WRc report (WRc, 1986), there is not only a large number of processes that can lead to the degradation of sewer pipes, but also a correspondingly large number of explanatory variables (Harrell, 2001). This is a direct consequence of the fact that in general pipe networks have developed and evolved over many years of construction. This means that networks have been built by many different contractors using many different materials, construction methods and specifications. In addition, the site conditions, i.e. soil conditions, sewage composition, water table location, etc. can all impact the rate of deterioration. A long list of explanatory variables can be compiled from the literature, for example from the works of Ana and Bauwens (2010), Chughtai and Zayed (2008), Davies et al. (1999), and Mu¨ller (2002), and almost any combination of explanatory variables can be found. Besides the probabilistic nature of the deterioration process, this certainly also has something to do with the lack of availability of datasets with extensive parameter sets, and that correlation does not necessarily imply causality. Furthermore, statistical variable selection procedures may result in different models if the explanatory variables are strongly correlated. Nevertheless, there seems to be a core set of parameters that frequently show high relevance across various studies. Mu¨ller (2002) identified these as age (very high relevance), size, depth and material (all with high relevance), and similar variables are identified by many other authors. Dirksen and Clemens (2008) conclude from a not very systematic investigation that age is the most relevant explanatory variable for sewer decay. Baik et al. (2006) also found in their analysis of 90 km (545 data points) of sanitary sewer lines from San Diego that the ordered probit model predicts higher transition probabilities for older pipes. Furthermore, Wirahadikusumah et al. (2001) identified higher transition

4986

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 8 3 e4 9 9 4

probabilities for older sewers in the City of Indianapolis (USA), indicating that age is indeed a relevant explanatory variable. In contrast, Davies et al. (2001b) did not find a significant correlation between age and deterioration, although they used the property age as a proxy for sewer line age, and it is difficult to assess what influence this had on the result. Madanat et al. (1997) mention two issues in bridge condition data that are equally relevant for sewer condition observations: heterogeneity and non-homogenous or nonstationary Markov properties. The former indicates variances in the explanatory parameters, whereas the latter suggests that the transition probabilities are time-dependent.

2.4.

Conclusions from the literature review

The literature does not indicate a clear set of explanatory variables. This is not particularly surprising, due to the large variations between different networks and the stochastic character of the processes involved. Age or time-dependency seems to be of very high relevance, but not necessarily in all cases. Nevertheless, it seems that there is quite a compelling case for non-homogenous Markov models (e.g. semi-Markov chain, Kleiner and Rajani, 2001), where the option of timedependent transition probabilities exists. These can account for the frequent observation that over time the number of functioning lines decreases and the rate of this decrease is not uniform. Non-homogenous or semi-Markov chain models, whose transition probabilities are time dependent, are a widely used approach to describe the probabilistic behavior of sewer deterioration. They are conceptually appropriate for representing the generally used condition grading, which is a discrete ordinal measure. A major weakness of all these models is the availability of appropriate data for parameter estimation. In practice, this data can have substantial limitations and deviate substantially from a perfect dataset needed for calibration. For this reason, transition probabilities are frequently derived from expert opinions, with no means of verification by data. It is therefore extremely difficult to test the performance of deterioration models with real-world data. This conclusion is shared in the review by Ana and Bauwens (2010): “Among the nine cases reviewed, only four have provided validation analyses. In these few cases, the results showed varying levels of success. There is therefore much research needed in order to evaluate and to validate these models.” For this purpose, we developed the Network Condition Simulator which is capable of producing a ‘perfect’ synthetic dataset based on given survival functions. In the following, we demonstrate and discuss the usefulness of such datasets by means of two numerical experiments.

3. Network condition simulator (NetCoS) formulation 3.1.

Overview

The observed size and condition of a sewer network at a time t is the result of four main processes: deterioration of the sewer

pipes, decisions about replacing pipe sections, expansions of the sewer network, and the rehabilitation of sewers by temporary measures, such as inliners, which extend a section’s life span. In this section, we use the first three processes (illustrated in Fig. 1) to propose a network condition simulator (NetCoS) that emulates a hypothetical but complete set of sewer condition data. NetCoS generates a complete history of the network by simulating the time from the construction of the first sewer line to the current state for each pipe section within the sewer system. This includes the year of construction, the year of replacement, and the condition state in each year. It is important to note that the replaced (removed) pipes are not eliminated from the dataset to avoid biased estimations of the life-span. a) Deterioration (Step 1, Section 3.2): We define the deterioration by a set of survival functions (Fig. 2). A survival function defines the age-dependent probability that a pipe section is in a certain condition state (CS) or better. With no intervention, a section can only remain in the same CS or deteriorate to a poorer CS within a time step. The probabilities of changing the CS in the next time step are computed with a discrete semi-Markov chain. Based on these probabilities, a new CS is assigned randomly to each pipe section each year. In contrast to the homogenous Markov chain, where the time the process remains in the same state is geometrically distributed, this time can be arbitrarily distributed in semi-Markov chains (Barbu and Limnios, 2008; Ross, 1996). This property allows the deterioration for any kind of survival functions to be simulated. Although the resolution of the output is typically one year, the deterioration is calculated internally with a much smaller numerical time step D to obtain accurate results. The condition of a pipe section is represented by an ordinal inspection rating. Here, a rating with five levels is used (according to WRc, 1986): CS 1 stands for the best state and CS 5 for the worst one. b) Replacement (Step 2, Section 3.3): A decision about the replacement of a pipe section is made after each year. The replacement is modeled as a probabilistic process that depends on the condition class. This simulates, first, the fact that in reality it takes some time to detect a defective sewer line, and second, that non-defective pipes also have a certain probability of being replaced (e.g. due to capacity problems or in the course of a coordinated construction project). c) Extension (Step 3, Section 3.4): New pipes are added to the network due to expansion of the catchment area. By default, all newly built pipe sections are in the best condition (CS 1). It is important to realize that the combination of deterioration (a) and replacement (b) as a management decision leads to a distinction between physical life-span and functional lifespan. The latter is the time from the construction of a pipe section until it is replaced: it corresponds to the observed lifespan of a sewer and depends on the specific management strategy applied. As stated above, premature replacement could be due to hydraulic limitations or coordinated construction activities. The physical life-span is the time

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 8 3 e4 9 9 4

4987

Fig. 2 e Four survival functions define the deterioration from condition states (CS) 1 to 5. Each survival function Sj(q) gives the probability that a pipe section of age q has CS j or better. For example, the probability that a pipe section of age 70 has CS 4 can be read out from the vertical distance between S4(70) and S3(70).

a pipe section could be used until it reaches CS 5. This differentiation is not always relevant and can be ignored by setting the replacement probabilities in Step 2 to zero for all but the worst condition state.

3.2.

Deterioration model

The time TCS j a pipe section remains in CS j is treated as a random variable to model the different deterioration behavior of the pipe sections. Consequently, TCS1/j as the total time a section has spent in CS j or better, is also a random variable: TCS1/j ¼

j X

TCS j ; j ¼ 1; .; 4

(1)

k¼1

The probability that a pipe section of age q has CS j or better is expressed by the survival function Sj(q), which is related to the cumulative distribution function Fj (q) of TCS1/j:   P TCS1/j > q ¼ 1  Fj ðqÞ ¼ Sj ðqÞ

(2)

To describe the transition probability among the five CS used in this paper, four survival functions Sj(q), j ¼ 1,.,4 are required (Fig. 2). These survival functions can be chosen from any distribution as long they satisfy the condition: Sj ðqÞ > Sj1 ðqÞ; j ¼ 2; .; 4

Fig. 1 e Graphical representation of the calculation steps of NetCoS. For the sake of simplicity, only three condition states are shown in this graph (from CS 1 (good, dark color) to CS 3 (poor, light color)). a) The condition state distribution at the end of year t is shown where the lengths of all pipe sections with the same condition and year of construction are added up. b) The effect of deterioration and c) the effect of a replacement decision leading to d), the condition state distribution in the year t D 1. Additionally, the new pipe sections due to expansion are added.

(3)

for all q > 0. Common distributions for the life-span used for asset deterioration are the exponential, the log-normal, the Herz (see Eq. (9)) and the Weibull distribution. These survival functions define the proportions of the CS for pipe sections of the same age if no replacement of pipe sections occurs. Therefore, the four survival functions represent the distribution of the physical life-span. The difference Sj(q)  Sj1(q) is equal to the probability that a pipe of age q is in CS j. Fig. 2 shows an example where all new pipes are in the best CS. However, this is not a requirement for NetCoS. To predict the future condition of a single pipe section with a known CS at age q, a semi-Markov chain approach is used. Unlike homogeneous Markov chains, the transition probabilities of semi-Markov chains depend on the time the process has already spent in a specific state.

4988

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 8 3 e4 9 9 4

Given a probability vector pt:built ðtÞ ¼ ðpCS1 ðtÞ; .; pCS5 ðtÞÞT i for a section i built in t.built  t, where pCSj (t) stands for the probability that the section is in CS j at time t. Then the probability vector in t þ D can be calculated by:

At the end of year t, new pipe sections for all replaced pipe sections are built. The replacement probability for sections in poor condition is generally higher than for sections in good condition.

pt:built ðt þ DÞT ¼ pt:built ðtÞT Pðt  t:builtÞ ¼ pt:built ðtÞT PðqÞ i i i

3.4.

(4)

where P(t  t.built) ¼ P(q) is the age-dependent transition probability matrix. Assuming that a pipe section can deteriorate only one state per numerical time step D and that no improvement of pipe sections occurs, the transition probability matrix can be written in this simplified form: 1 0 0 0 1  p1;2 ðqÞ p1;2 ðqÞ B 0 1  p2;3 ðqÞ p2;3 ðqÞ 0 0 C C B 0 0 1  p3;4 ðqÞ p3;4 ðqÞ 0 C PðqÞ ¼ B C B @ 0 0 0 1  p4;5 ðqÞ p4;5 ðqÞ A 0 0 0 0 1 0

(5)

Pj,k (q) is the probability that a pipe section in CS j will change to CS k within the time step from q to q þ D. These probabilities are derived from the survival functions Sj(q) following the description in Kleiner (2001): 8 > > <

f1 ðqÞ$D S1 ðqÞ pj;jþ1 ðqÞ ¼ fj ðqÞ$D > > : Sj ðqÞ  Sj1 ðqÞ

;j ¼ 1 (6) ; j˛f2; 3; 4g

where fj(t) is the probability density function (pdf) of Fj(t) ¼ 1  Sj(t). The numerical time step D has to be small (e.g. D < 1/100 year), otherwise the approximation is not valid. For the procedure it is assumed that all newly built pipe sections are in the best condition state (CS 1). The condition ðt þ 1Þ is assigned state of a pipe section in the next year CSt:built i probabilistically with respect to the current condition CSit:built ðtÞ and the age of a pipe section. The probability vector ðt þ 1Þ for the CS next year is calculated by repeating Eq. pt:built i (4). Because the numerical time step D is smaller than one, D1 repetitions are required to calculate the prediction for the following year: ðt þ 1ÞT ¼ pt:built ðtÞT pt:built i i

1=D Y

Pðt  t:built þ mDÞ; D << 1

(7)

m¼1

The condition of the pipe section in the previous year t is ðtÞ is defined. known, so pt:built i

3.3.

Replacement

Replacements of pipe sections are based on the information available to the operator and the chosen management strategy. The operator might not know that the line is damaged until it has been inspected, or decides not to spend money even if the line is in bad condition. Additionally, there is a chance that pipes in all condition classes are replaced due to hydraulic capacity problems or to coordinated construction efforts with other utilities. The model emulates the effect of such behavior by assigning a replacement probability to every CS. The probabilities that a sewer line of a particular CS is replaced within the next year are expressed by the vector:  rep rep rep rep rep  prep ¼ pCS1 ; pCS2 ; pCS3 ; pCS4 ; pCS5

(8)

Network expansion

Typically, a sewer system is not built at once but is likely to have grown over several decades. In NetCoS, this is taken into account by an annual expansion of the sewer system by a length Lexp (t). At the end of year t, new pipe sections with a total length Lexp (t) are added to the inventory.

4.

Application of NetCoS

4.1.

Use of NetCoS

All deterioration models make explicit and implicit assumptions (e.g. the survival functions are Weibull distributed, no rehabilitation occurs, the data are free from errors, etc.). It would be crucial to be able to test the models with reliable datasets. However, as pointed out in Section 2, it is very hard or practically impossible to obtain a good dataset from mature sewer systems that have grown over several decades. Often, the data only represents a subset of the total dataset. For example, historic data about replaced sewer lines are very often not available. Also, if one accepts the fact that condition data are needed to make informed decisions, it is difficult to determine optimal inspection intervals. Are 10 years sufficient? Would it be better to first inspect each pipe every five years and subsequently increase the intervals if the condition does not change? NetCoS offers the opportunity to answer such questions by investigating the performance of deterioration models with respect to systematic errors, the efficiency of the parameter estimation or the influence of inadequate data. Examples are wrong interpretations of CCTV data, estimation of the parameter uncertainties based on the input data (see Section 4.3, Fig. 9.) or the loss of historical data for replaced sewer lines (see Section 4.2.3, Fig. 6.). These examples highlight some aspects that can be investigated using NetCoS. Fig. 3 illustrates a general procedure for the use of NetCoS: first, the survival functions and all other parameters of the simulator are defined. Then NetCoS is used to generate a complete dataset, from which only a subset is used to fit a deterioration model, typically implemented in a commercial software tool used by utilities. Finally, the survival functions estimated by the deterioration model can be compared with the predefined survival functions. Due to the stochastic nature of NetCoS, the model output can be seen as one realization of a random variable: two datasets generated with the same parameters are drawn from the same distribution but will be different. As a consequence, meaningful systematic errors and calculate confidence intervals are best computed from several repeated simulations. In the following sections, we illustrate this examination process with two examples of different types of deterioration models. The influence of the replacement probabilities and the choice of data are first examined for the cohort model presented by Baur and Herz (2002). As a second example,

4989

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 8 3 e4 9 9 4

Fig. 3 e Examination of a deterioration model using NetCoS.

a homogeneous Markov model with a maximum likelihood estimation as applied by Dirksen and Clemens (2008) is used.

It
4.2.

It


Example I: cohort survival model

i

4.2.1.

Model description

To apply the cohort survival model proposed by Baur and Herz (2002), the data of pipe sections with a similar expected deterioration behavior are grouped together in ‘cohorts’. This grouping is generally based on the construction methods applied and pipe materials used, and can be quantified by using one-and two-way analysis of variance (ANOVA) procedures as suggested by Kleiner and Rajani (1999), for instance. For each of these groups, a separate set of survival functions is fitted directly to the data. A special survival function for the network infrastructure has been suggested by Herz (1995):    Sj qbj ¼

bj;1 þ 1 bj;1 þ ebj;2 ðqbj;3 Þ

; j ¼ 1; .; 4

(9)

Every element of the parameter vector bj ¼ (bj,1,bj,2,bj,3) has to be positive as well as bj,2 < 1 and bj,3  q. The aging factor bj,1 affects the speed of the aging at the beginning. bj,2 determines the deterioration rate and bj,3 is the resistance time before any deterioration occurs. The parameters are estimated by fitting the survival functions Sj(qjbj), j ¼ 1,.,4 for the transition from CS j to CS j  1 to the percentage of pipe sections which have at least a CS j:

min bj

t X t:built¼t0

j ¼ 1;.;4

  abs Sj t  t:builtbj  

P

t:built It
PNt:built i¼1

Lt:built i

!! ; (10)

where t is the current year, t0 the construction year of the oldest pipe section and bj the parameter vector of the survival the length of all function Sj. Nt.built is the number and Lt.built i pipe sections built in year t.built. The length of all pipe sections built in t.built which are still in use at a time t and P t:built It
4.2.2.

1 ; if pipe section i;built in t:built is in use at t 0 ; else

(11)

Influence of the replacement policy

To examine the influence of the replacement policy on the estimation of the life-span, two scenarios were defined, which leads to parameter sets with different replacement probabilities (Table 1). To identify solely the effect of the replacement, we eliminated a possible bias due to different model formulations. We consequently applied the survival functions of the cohort model in Eq. (9) in Eq. (2) of NetCoS to generate virtual datasets with the predefined parameter sets given in Table 1. Two scenarios were defined: in Scenario A, only pipe sections in CS 5 are replaced with a probability of 0.3 a1, whereas in Scenario B there is a certain probability that a section is replaced before it has reached CS 5. The replacement probabilities are defined as vector prep and shown Table 1. together with the parameters (bi) of the survival functions.

Table 1 e Predefined parameter sets used by NetCoS for data generation. Scenario A replaces only pipes in CS 5 with a probability of 0.3 aL1. Scenario B allows the replacement of lines in better condition with decreasing probabilities. Scenario A (replacement of pipes in CS 5 only)

Scenario B (replacement of pipes in all CS possible)

b1 ¼ (50,0.05,0) b2 ¼ (18,0.05,0) b3 ¼ (5,0.05,0) b4 ¼ (0.7,0.05,0) prep ¼ (0,0,0,0,0.3) L ¼ 25 m Length built: 163 km Length active: 120 km Length replaced in CS 5: 43 km

b1 ¼ (50,0.05,0) b2 ¼ (18,0.05,0) b3 ¼ (5,0.05,0) b4 ¼ (0.7,0.05,0) prep ¼ (0,0.001,0.01,0.05,0.3) L ¼ 25 m Length built: 184 km Length active: 120 km Length replaced in CS 5: 23 km

4990

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 8 3 e4 9 9 4

4.2.3.

Fig. 4 e Results of the cohort model (Baur and Herz, 2002) for datasets generated with Scenario A (only pipe sections in CS 5 are replaced). The shaded areas indicate the predefined survival functions used in NetCoS. The solid lines are the means of the estimations from 50 generated datasets, the dashed lines the 10% and 90% quantiles. No visible bias is present.

Survival selection bias

In practice, pipe section inventories often comprise only the lines in use; historic data about the replaced pipe sections are missing. This means that the total length once built in P t:built t:built , is unknown. In the following, a particular year, N i¼1 Li we show the effect of only using the condition rating of the active sewer inventory. By considering the existing pipe sections only, the denominator of the fraction in Eq. (10) is replaced by this known length (see also Eq. (11)). The minimization problem then changes to: 0 min@ bj

11 P t:built It
0

i

j ¼ 1; .; 4

(12)

For both scenarios, datasets of a 120 year old sewer system with a constant annual expansion of 1000 m were simulated. The length of a pipe section L was set to 25 m. Grouping was not required, since the scenario assumes the same deterioration behavior for all pipe sections for the data generation. The survival functions were fitted to these data according Eq. (10). Because NetCoS is stochastic, this step was repeated 50 times. Our results for Scenario A show that the estimated survival functions (solid lines) match the predefined survival functions (the boundaries of the shaded areas) exactly, without any detectable systematic deviation (Fig. 4). For Scenario B, however, the life-span is systematically underestimated (Fig. 5). This effect is particularly pronounced for the transition from CS 4 to CS 5, i.e. S4 ðqjb4 Þ and to a lesser degree for the transition from CS 3 to CS 4. In Scenario B, a fraction of the pipe sections were replaced before reaching their physical life-span, therefore effectively shortening this life-span. This shows that the cohort model estimates functional life-spans and that their interpretation as physical life-spans would be incorrect.

The survival functions were fitted with Eq. (12) to the same 50 datasets generated for Scenario A (see Table 1). Our results show that all survival functions are overestimated (Fig. 6). As expected, this is particularly pronounced for S4, i.e. the transition from CS 4 to CS 5. What we had not expected, however, was the magnitude of the over estimation, which was more than 200 years for the median life span of the pipes in our didactical example. The fact that discarding historic data does not simply increase the noise for this particular model, but creates a systematic bias, is of particular importance. This bias comes from the fact that the early deteriorated pipes were replaced (according the probabilities in Table 1) and therefore eliminated from the dataset, so that the non-representative data of mainly robust long-living pipes are used for the estimation. This “selective survival bias”, which leads to a potentially massive over-estimation of the life-spans, has previously been reported by Ho¨rold and Baur (1999), but had not been previously quantified.

Fig. 5 e Results of the cohort model (Baur and Herz, 2002) for datasets generated with Scenario B (pipe section of all condition states can be replaced). The shaded areas indicate the predefined survival functions used in NetCoS. The solid lines are the means of the estimations from 50 generated datasets, the dashed lines the 10% and 90% quantiles. The physical life span is systematically underestimated.

Fig. 6 e Results of the cohort model (Baur and Herz, 2002) for datasets generated with Scenario A (only using data of pipe sections still in use). The shaded areas indicate the predefined survival functions used in NetCoS. The solid lines are the means of the estimations from 50 generated datasets, the dashed lines the 10% and 90% quantiles.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 8 3 e4 9 9 4

4.3.

Example II: Markov model

4.3.1.

Model description

4991

Dirksen and Clemens (2008) model the sewer conditions of a Dutch town using a discrete, homogeneous Markov chain and three condition classes. As a consequence, the probability of changing a condition class depends only on the current state, but not on the time already spent in this class. By adapting their method to five condition classes, the following transition probability matrix P is obtained: 0 B B B P¼B B @

1

P5

k¼2 p1;k

0 0 0 0

1

p P1;2 5

k¼3 p2;k

0 0 0

1

p1;3 p P2;3 5

k¼4 p3;k

0 0

p1;4 p2;4 p3;4 1  p4;5 0

1

p1;5 C p2;5 C C p3;5 C C p4;5 A 1

(13)

The probabilities pj,k that a pipe section changes from CS j to CS k, j < k, are estimated by maximizing the likelihood function (Eq. (14)), which requires data from two consecutive inspections of the same pipe sections. LCS j/CS k(n) is the total length of all pipe sections which have changed from CS j at the first inspection to CS k at the second inspection after n years. The logarithm of the likelihood function is: LL ¼

5 X 5 XX n

 n  log Pjk LCS

j/CS k ðnÞ

(14)

j¼1 k¼1

No survival functions are directly estimated for this Markov model. However, they can be derived as the predicted state probabilities of a pipe with age q which was in CS 1 at age 0.

4.3.2. Influence of inspection frequency and pipe section replacement In example II, NetCoS is used to examine the influence of the time span between two inspections. Again, two scenarios were investigated. Scenario C, where no replacements occur, and Scenario D, where pipes in CS 5 have a 0.3 a1 probability of being replaced: prep ¼ (0,0,0,0,0.3). The Markov model of Dirksen and Clemens (2008) requires exponential distributed residence times (time a pipe spent in a certain CS). To implement this in Eq. (2), we used an Erlang survival function, because the sum of n exponential distributed residence times with rate l has an Erlang survival function with a rate parameter l and a shape parameter n (Evans et al., 2000):

Sj ðqjlÞ ¼

j1 X

1 elq ðlqÞk ; l; q  0; j ¼ 1; .; 4 k! k¼0

Fig. 7 e Maximum likelihood fit of the Markov model on data generated without replacement. The shaded areas indicate the predefined survival functions used in NetCoS. The solid lines are the means of the estimations of 150 simulated datasets, the dashed lines are the 10% and 90% quantiles. No systematic bias is detected.

Our results for Scenario C show that, where no sewer pipes are replaced, the survival functions estimated by the Markov model match the predefined survival functions well. There is no bias (Fig. 7) and the inspection interval consequently has no influence on the parameter estimation. However, the lifespan is systematically overestimated for Scenario D (some replacement when in worst condition) (Fig. 8). This is because a fraction of the pipe sections with CS 5 were replaced after the first inspection and could not therefore be considered in the parameter estimation. These results demonstrate that the shorter the time between inspections the smaller the bias of the survival function, since fewer pipes in CS 5 were discarded from the dataset (Fig. 9). NetCoS can be used to quantify this systematic deviation. Using the dataset from Scenario D, the time between inspections was varied between 1 and 10 years. The average bias of S4(q) (distance (a) at 50% probability in Fig. 8) and its variance (distance (b) at 50% probability in Fig. 8) for different inspection intervals is plotted in Fig. 9. It can be seen that the bias and the variance of the estimation increase with longer inspection intervals. If, in our didactical example, an average bias of less

(15)

where q is the age of a pipe section and l the rate parameter of the exponential distributions. For Scenarios C and D, we also generated datasets for a 120 year old sewer system with an expansion of 1000 m a1 and a pipe section length of 25 m. The rate for all Erlang survival functions was set to l ¼ 4.591‧102 a1, which corresponds to a median physical life-span of 80 a (transition CS4 to CS 5). Because NetCoS is stochastic, 150 datasets were simulated. Again, historic data were discarded so that only data for ‘active’ pipe sections were used in the fitting of the Markov model.

Fig. 8 e Maximum likelihood fit of the Markov model on data generated with replacement and 10 years between inspections. The shaded areas indicate the predefined survival functions used in NetCoS. The solid lines are the means of the estimations of 150 simulated datasets and the dashed lines the 10% and 90% quantiles. The arrow (a) shows the bias, the arrow (b) the interquantile range. The model systematically overestimates the average (physical) life-span.

4992

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 8 3 e4 9 9 4

Fig. 9 e Average bias of the median of the survival function S4(q) (transition between CS 4 and CS 5) against the time between inspections. The average bias was calculated from 150 generated datasets (with replacement) for each point.

then 15 years is desired, we see that each pipe should be inspected every three years. Common inspection cycles of approximately 10 years will most probably lead to a bias of approximately 40 years. However, these results depend strongly on the number of class changes between inspections and therefore on the number of sewer sections (i.e., the length of the sewer system) and the average age of the network.

5.

Discussion

Our results show that our novel network condition simulator, NetCoS, is useful for testing existing sewer deterioration models and quantifying phenomena related to different data qualities and inspection intervals. In the following, we will further discuss the use of NetCoS to evaluate deterioration models and its strengths and limitations.

5.1.

Use of NetCoS to evaluate deterioration models

In our examples, we used NetCoS to produce well-defined datasets. These then served as input data for estimating the parameters of two different deterioration models so that the weaknesses of these models with respect to imperfect data could be illustrated. In the first example, we demonstrated that the cohort survival model is sensitive to premature replacement of sewer lines and to discarded data of replaced pipes. A particularly important result is that for this specific model discarding historic data does not simply increase noise, but creates a systematic bias that cannot in most cases be compensated by statistical means. This “selective survival bias” had previously been reported by Ho¨rold and Baur (1999). In the second example, we analyzed the effect of prematurely replaced pipes and inspection frequency on the parameter estimation for a homogeneous Markov chain model. The results showed that the Markov chain model is also sensitive to prematurely replaced pipes (Fig. 8) and that the systematic deviation depends on the inspection frequency (Fig. 9). In both examples, NetCoS was used to create artificial datasets in order to systematically explore the characteristics

of specific deterioration models and identify their weaknesses. We focused on two issues and ignored the influence of uncertainty, errors and biases in the data, e.g. the conversion of visual CCTV material into condition classes. However, as NetCoS can be used to test the sensitivity of the parameter estimation process for a specific model, it can help to identify quality requirements for the reproducibility of CCTV-derived condition data. In reality, it is rare to have accurate and complete information about the history of all pipe sections. Our proposed approach using NetCoS allows the effect of incomplete datasets to be quantitatively evaluated in combination with a specific deterioration model of interest, or a model to be potentially found that can generate unbiased estimates of life-spans on the basis of the available data. This is especially important if commercially available asset management software is used whose underlying model cannot be modified by the user and, in addition, is often not transparently documented. These examples allow conclusions to be drawn about the requirements for proper data management. A particularly important fact is that datasets without information about replaced pipes (historic data) not only leads to greater uncertainty in the estimate, but gives it a systematic bias. Although the value of historic data is very often not recognized in practice, our results demonstrate that the most effective and reasonably cost-effective remedy in all these examples is to keep historic data in the utilities’ database.

5.2.

Strengths and limitations of NetCoS

The current version of NetCoS considers three processes (deterioration, replacement and expansion), whereas most existing models only consider the deterioration for predicting future sewer conditions. However, the version of NetCoS presented here still represents a very simple model, which can be readily extended, for example to model rehabilitation measures, such as inliners, or to investigate the influence of other explanatory variables such as materials, pipe diameters, installation depths or construction techniques. Two important strengths of NetCoS are: i) its flexibility, and ii) its modular model structure. First of all, its great flexibility enables the user to generate more complex data, e.g. by

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 8 3 e4 9 9 4

introducing CCTV interpretation noise, in order to address additional sources of uncertainty. Another reason for its flexibility it that transition probabilities are derived from survival functions: NetCoS can be used to describe any probabilistic deterioration process; the method is not limited to certain types of distributions, as any continuous survival function can be applied. For example, if there is evidence that a certain group of sewer pipes has a longer life span, a specific, skewed survival function could be used to model this specific group. An additional advantage of NetCoS is its modular structure: NetCoS can be readily extended to include other processes or replace current modules with more sophisticated implementations. For example, the replacement probabilities can be made time-dependent to mimic a change in past management policy, or more complex models can be used for replacement decisions. Different pipe materials with specific deterioration behaviors can be simulated by defining separate sets of survival functions for each material and choosing the material of new pipe sections in the growth process. Similarly, the impact of heavy traffic or zones of unstable ground can be considered. In this regard, NetCoS can also be adapted to include further explanatory variables. However, it suffers from the limitation that representative survival functions may not be available for systems involving largely unknown or highly variable deterioration processes. Unfortunately, given the current state of knowledge, this is mostly the case for all sewer networks. As stated above, we also fully acknowledge the fact that it cannot be verified whether the synthetic dataset produced by NetCoS is indeed a realistic representation of real-life data or not. As a consequence, NetCoS cannot provide information about the performance of a specific deterioration model in an absolute sense. In this sense, the synthetic datasets enable the user to construct relevant test cases for the quantitative analyses of candidate models as illustrated in our examples. We think that NetCoS is a useful tool for analyzing a large variety of relevant questions concerning the quality of modelbased estimations of sewer deterioration and life-span (e.g. as proposed by Ana and Bauwens, 2010). We hope that NetCoS will help to increase awareness of data and modeling issues in sewer deterioration modeling and ultimately to improve the effective and efficient management of these expensive infrastructure assets.

6.

Conclusions

 Quantitative maintenance and rehabilitation planning of our network infrastructure is becoming increasingly important. For sewer sections, this requires the application of deterioration models that can predict their future condition. Although several deterioration models have been proposed in the literature, current investigations show that validation studies are hardly ever performed. As this seems to be due to a lack of reliable asset data, there is a great need for research tools to assess deterioration models and guide data management.  In this paper, we therefore propose a novel Network Condition Simulator (NetCoS) which can be used to generate synthetic datasets of sewer inventories on different scales

4993

and with arbitrary resolution. Unlike most current models, it not only simulates the deterioration but also considers the expansion of the network and the replacement of pipes of any condition.  On the basis of the results from computer experiments, we conclude that the quality of the model-based estimation of the survival functions depends strongly on the representativeness or completeness of the data used. We demonstrate that discarding historic data in the cohort survival model does not simply increase noise, but creates a systematic bias. This leads to an over-estimation of the physical lifespan of more than 200 years for a typical scenario. For a homogenous Markov model, the bias depended on the frequency of CCTV investigation, increasing from approximately 10 to 40 years with inspection intervals from 1 to 10 years.  The approach presented here using a data simulator allows scientists and utility managers to identify the weaknesses of current deterioration models and develop appropriate data management strategies for reliable model applications.  NetCoS can easily be adapted to more complex deterioration models or specific cases. The link to the freely available R source code can be found in the Supplementary Material.

Appendix. Supplementary material Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.07.008.

references

Ana, E.V., Bauwens, W., 2010. Modeling the structural deterioration of urban drainage pipes: the state-of-the-art in statistical methods. Urban Water Journal 7 (1), 47e59. Baik, H.S., Jeong, H.S., Abraham, D.M., 2006. Estimating transition probabilities in Markov chain-based deterioration models for management of wastewater systems. Journal of Water Resources Planning and Management 132 (1), 15e24. Barbu, V.S., Limnios, N., 2008. Semi-Markov Chains and Hidden Semi-Markov Models Toward Applications. Springer, New York. Baur, R., Herz, R., 2002. Selective inspection planning with ageing forecast for sewer types. Water Science and Technology 46 (6e7), 389e396. Chughtai, F., Zayed, T., 2008. Infrastructure condition prediction models for sustainable sewer pipelines. Journal of Performance of Constructed Facilities 22 (5), 333e341. Clegg, D., Eadon, A.R., Fiddes, D., 1989. UK state-of the-art e sewerage rehabilitation. Water Science and Technology 21 (10e11), 1101e1112. Davies, J.P., Clarke, B.A., Whiter, J.T., Cunningham, R.J., 2001a. Factors influencing the structural deterioration and collapse of rigid sewer pipes. Urban Water 3 (1e2), 73e89. Davies, J.P., Clarke, B.A., Whiter, J.T., Cunningham, R.J., Leidi, A., 2001b. The structural condition of rigid sewer pipes: a statistical investigation. Urban Water 3 (4), 277e286. Davies, J.P., Whiter, J.T., Clarke, B.A., Ockleston, G.O. and Cunningham, R.J., 1999. Application of interaction matrices to the problem of sewer collapse. In: Proceedings 11th European sewage and refuse symposium. Liquid wastes section, May 1999, Munich, Germany.

4994

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 8 3 e4 9 9 4

Delleur, J.W., 1994. Sewerage failure, diagnosis and rehabilitation. In: URBAN Drainage Rehabilitation Programs and Techniques. ASCE, New York, pp. 11e28. Dirksen, J., Clemens, F.H.L.R., 2008. Probabilistic modeling of sewer deterioration using inspection data. Water Science and Technology 57 (10), 1635e1641. Dirksen, J., Goldina, A., ten Veldhuiz, J.A.E. and Clemens, F.H.L.R., 2007. The role of uncertainty in urban drainage decisions: uncertainty in inspection data and their impact on rehabilitation decisions. In: Proceedings IWA 2nd Leading Edge Conference on Strategic Asset Management, 17e19 October 2007, Lisbon, Portugal. Djordjevic, S., Prodanovic, D., Maksimovic, C., Ivetic, M., Savic, D., 2005. SIPSON - Simulation of interaction between pipe flow and surface overland flow in networks. Water Science and Technology 52 (5), 275e283. Evans, M., Hastings, N., Peacock, B., 2000. Statistical Distributions, third ed. Wiley-Interscience, New York. Hafskjold, L.S., Vanrenterghem-Raven, A., 2007. Experiences from the application of care-w, computer aided rehabilitation of water networks. In: Proceedings Combined International Conference of Computing and Control for the Water Industry, CCWI2007 and Sustainable Urban Water Management, SUWM2007, 3e5 September 2007, De Montfort University, Leicester, UK. Hahn, M.A., Palmer, R.N., Merrill, M.S., Lukas, A.B., 2002. Expert system for prioritizing the inspection of sewers: knowledge base formulation and evaluation. Journal of Water Resources Planning and Management 128 (2), 121e129. Harrell, F.E., 2001. Regression Modeling Strategies: With Applications to Linear Models, Logistic Regression, and Survival Analysis, first ed. Springer, Berlin. Herz, R., 1995. Alterung und Erneuerung von Infrastrukturbesta¨nden - Ein Kohortenu¨berlebensmodell (Aging and renewal of infrastructure e a cohort survival model). Jahrbuch fu¨r Regionalwissenschaft 14e15, 5e29. Herz, R.K., 1998. Exploring rehabilitation needs and strategies for water distribution networks. Journal of Water Supply Research and Technology AQUA 47 (6), 275e283. Herz, R.K., Lipkow, A., 2002. Life cycle assessment of water mains and sewers. Water Science and Technology: Water Supply 2 (4), 51e58. Ho¨rold, S., Baur, R., 1999. Modelling sewer deterioration for selective inspection planning e case study Dresden. In: Proceedings of the 13th European Junior Scientist Workshop, September 1999, Rathen, Germany. Kathula, V.S., 2001. Structural distress condition modeling for sanitary sewers. Ph.D. thesis. Louisiana Tech University, Ruston, LA, USA. Kleiner, Y., 2001. Scheduling inspection and renewal of large infrastructure assets. Journal of Infrastructure Systems 7 (4), 136e143. Kleiner, Y., Rajani, B., 1999. Using limited data to assess future need. Journal American Water Works Association 91 (7), 47e61. Kleiner, Y., Rajani, B., 2001. Comprehensive review of structural deterioration of water mains: statistical models. Urban Water 3 (3), 131e150. Kleiner, Y., Sadiq, R., Rajani, B., 2006. Modelling the deterioration of buried infrastructure as a fuzzy Markov process. Journal of Water Supply Research and Technology AQUA 55 (2), 67e80. Korving, 2004. Probabilistic assessment of the performance of combined sewer systems. Ph.D. thesis. TU Delft, The Netherlands. Madanat, S., Ibrahim, W.H.W., 1995. Poisson regression models of infrastructure transition probabilities. Journal of Transportation Engineering 121 (3), 267e272. Madanat, S., Mishalani, R., Ibrahim, W.H.W., 1995. Estimation of infrastructure transition probabilities from condition rating data. Journal of Infrastructure Systems 1 (2), 120e125.

Madanat, S.M., Karlaftis, M.G., McCarthy, P.S., 1997. Probabilistic infrastructure deterioration models with panel data. Journal of Infrastructure Systems 3 (1), 4e9. Makropoulos, C.K., Butler, D., Maksimovic, C., 2003. Fuzzy logic spatial decision support system for urban water management. Journal of Water Resources Planning and Management 129 (1), 69e77. Mauch, M., Madanat, S., 2001. Semiparametric hazard rate models of reinforced concrete bridge deck deterioration. Journal of Infrastructure Systems 7 (2), 49e57. Maurer, M., 2009. Specific net present value: an improved method for assessing modularisation costs in water services with growing demand. Water Research 43 (8), 2121e2130. Micevski, T., Kuczera, G., Coombes, P., 2002. Markov model for storm water pipe deterioration. Journal of Infrastructure Systems 8 (2), 49e56. Mishalani, R.G., Madanat, S.M., 2002. Computation of infrastructure transition probabilities using stochastic duration models. Journal of Infrastructure Systems 8 (4), 139e148. Morcous, G., Rivard, H., Hanna, A.M., 2002a. Case-based reasoning system for modeling infrastructure deterioration. Journal of Computing in Civil Engineering 16 (2), 104e114. Morcous, G., Rivard, H., Hanna, A.M., 2002b. Modeling bridge deterioration using case-based reasoning. Journal of Infrastructure Systems 8 (3), 86e95. Mu¨ller, K., 2002. Entwicklung eines allgemein anwendbaren Verfahrens zur selektiven Erstinspektion von Abwasserkana¨len und Anschlussleitungen. Teil A: Wissenschaftliche Untersuchungen (Development of a generally applicable methodology for selective inspection of sewer networks. Part A: scientific studies). Final report of the Institute of Urban Water Management ISA, RWTH Aachen University, Aachen, Germany. Najafi, M., Kulandaivel, G., 2005. Pipeline condition prediction using neural network models. In: Proceedings ASCE Pipeline Division Specialty Conference e PIPELINES 2005, 21e24 August 2005, Houston, TX, USA. Ross, S.M., 1996. Stochastic Processes, second ed. Wiley, New York. Rutsch, M., Rieckermann, J., Cullmann, J., Ellis, J.B., Vollertsen, J., Krebs, P., 2008. Towards a better understanding of sewer exfiltration. Water Research 42 (10e11), 2385e2394. Saegrov, S., Melo Baptista, J.F., Conroy, P., Herz, R.K., Legauffre, P., Moss, G., Oddevald, J.E., Rajani, B., Schiatti, M., 1999. Rehabilitation of water networks: survey of research needs and on-going efforts. Urban Water 1 (1), 15e22. Savic, D.A., Giustolisi, O., Laucelli, D., 2009. Asset deterioration analysis using multi-utility data and multi-objective data mining. Journal of Hydroinformatics 11 (3e4), 211e224. Schmitt, T.G., Thomas, M., Ettrich, N., 2004. Analysis and modeling of flooding in urban drainage systems. Journal of Hydrology 299 (3e4), 300e311. Tran, H.D., 2007. Investigation of deterioration models for stormwater pipe systems. Ph.D. thesis. Victoria University, Melburne, VIC, Australia. Wirahadikusumah, R., Abraham, D., Iseley, T., 2001. Challenging issues in modeling deterioration of combined sewers. Journal of Infrastructure Systems 7 (2), 77e84. Woodhouse, J., Williams, A., Dixon, M., Male, S., Davies, R., Mustafa, M., Bryan, N., Jay, P., Thompson, A., 2008. PAS 55-1: 2008 Asset Management. Part 1: Specification for the Optimized Management of Physical Assets. The Woodhouse Partnership Ltd, Kingsclere, UK. WRc, 1986. Sewer Rehabilitation Manual, second ed. Water Research Centre Plc, UK. Yan, J.M., Vairavamoorthy, K., 2003. Fuzzy approach for pipe condition assessment. In: Proceedings ASCE International Conference on Pipeline Engineering and Construction: New Pipeline Technologies, Security, and Safety, 13e16 June 2003, Baltimore, MD, USA.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 9 5 e5 0 0 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Solid respirometry to characterize nitrification kinetics: A better insight for modelling nitrogen conversion in vertical flow constructed wetlands Ania Morvannou a, Jean-Marc Choubert a,*, Marnik Vanclooster b, Pascal Molle a a b

Cemagref, UR MALY, 3 bis quai Chauveau - CP 220, F 69336 Lyon, France Earth and Life Institute Environmental Sciences, Universite´ Catholique de Louvain, Croix du Sud 2 Box 2, B 1348 Louvain-la-Neuve, Belgium

article info

abstract

Article history:

We developed an original method to measure nitrification rates at different depths of

Received 8 February 2011

a vertical flow constructed wetland (VFCW) with variable contents of organic matter

Received in revised form

(sludge, colonized gravel). The method was adapted for organic matter sampled in con-

23 June 2011

structed wetland (sludge, colonized gravel) operated under partially saturated conditions

Accepted 1 July 2011

and is based on respirometric principles. Measurements were performed on a reactor,

Available online 13 July 2011

containing a mixture of organic matter (sludge, colonized gravel) mixed with a bulking agent (wood), on which an ammonium-containing liquid was applied. The oxygen demand

Keywords:

was determined from analysing oxygen concentration of the gas passing through the

Nitrification

reactor with an on-line analyzer equipped with a paramagnetic detector. Within this paper

Solid respirometry

we present the overall methodology, the factors influencing the measurement (sample

Vertical flow constructed wetlands

volume, nature and concentration of the applied liquid, number of successive applications), and the robustness of the method. The combination of this new method with a mass balance approach also allowed determining the concentration and maximum growth rate of the autotrophic biomass in different layers of a VFCW. These latter parameters are essential inputs for the VFCW plant modelling. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Wastewater treatment based on the technology of constructed wetlands (CW) is an attractive technology for small communities (<2000 PE). The simplicity of low-cost operations and the reliability as well as efficiency of their treatment are often compatible with the limited resources that can be spent on water treatment in these communities (Kadlec, 2000). An innovative type of vertical flow constructed wetland (VFCW) was developed by Cemagref, France (Molle et al., 2005). The VFCWs contain gravel rather than sand and is combined with a preliminary settling tank. For a long-term operation period,

mean removal efficiencies of the VFCW system yield 90% for chemical oxygen demand (COD) and total suspended solids (TSS), and 60e80% for total kjeldahl nitrogen (TKN). Despite the recent progress of optimization, the nitrification efficiency on the first stage of this VFCW still remains incomplete. To remediate to this, the VFCW is upgraded with a second vertical stage, but this increases the total surface area needed. Alternatively, optimizing the first stage should still be possible, as suggested by a recent study carried out under constraining conditions (Molle et al., 2008), but requires a deeper analysis and a better understanding of the biological turn-over rates.

* Corresponding author. Tel.: þ33 4 72 20 87 87. E-mail address: [email protected] (J.-M. Choubert). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.004

4996

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 9 5 e5 0 0 4

Partially saturated filter beds are often considered as “black boxes” because of the complexity and the limited understanding of the processes that take place under partially saturated conditions. Conceptual models, based on physical and biochemical process knowledge, improve the insight on the way partially saturated filter beds function and improve their design and operation (Langergraber et al., 2008). As a result, some process based models of wetland systems were recently developed, allowing to refine the description of the water flow and solute transport processes in the filter medium. For instance CWM1 (Langergraber et al., 2009),   or CW2D in the Hydrus software (Sim unek et al., 1999;   Langergraber and Sim unek, 2005), combine equations describing the biological processes of growth and decay of biomass with a multi-component reactive transport model. In these cases, the biological process descriptions are inferred from the activated sludge model ASM (Henze et al., 1987). However, due to model complexity and hence the small parsimonious status of these models (i.e. high number of parameters to fit), the calibration of these conceptual VFCW models remains a complicated task, in particular for the biological component of the model. Indirect parameter determination methods through inverse modelling of the inflow/ outflow pollutants fluxes of a VFCW was suggested to support the calibration of such a model (Langergraber, 2007). Though, for a calibration that also fits with the state-of-the-art biological process knowledge, the combination of inverse modelling with direct parameter determination techniques is suggested. Parameters inferred from direct determination experiments can therefore be used to generate reliable prior estimates of biological parameters, which are subsequently further refined during model inversion. By doing so, estimated parameters comply with prior estimated parameter intervals and overall parameter uncertainty is reduced. Dynamic measurement techniques (i.e. under continuous air supply) are widely used when modelling a WWTP process. The turnover rates of the different fractions of organic matter (e.g. rapidly/slowly degradable organic matter), and the parameters of the heterotrophic biomass contributing to the degradation, are determined using the monitored time evolution of COD (Lasaridi and Stentiford, 1998; Stricker et al., 2003; Scaglia and Adani, 2008) or the oxygen uptake rate under liquid conditions (Spanjers and Vanrolleghem, 1995; Lagarde et al., 2005). Other authors adapted the concepts of activated sludge models to characterize the turn-over of solid wastes and the partially saturated filtration medium (Adani et al., 2004;

Tremier et al., 2005; Andreottola et al., 2007; Ortigara et al., 2011). Yet, to study the conversion of ammonium into nitrates by autotrophic biomass (nitrification), it is still necessary to design a new technique that allows conducting measurement in partially saturated media. Indeed, the available techniques consist either in the estimation of the nitrate production rates (WERF, 2003; Dold et al., 2005; Choubert et al., 2009), either in the measurement of the oxygen uptake rate under inhibiting conditions (Surmacz-Gorska et al., 1995) or either in the measurements of the pH-evolution (Ficara and Rozzi, 2001). All these techniques lead to significant disturbance of the biofilm nitrification activity in partially saturated conditions. The purpose of this paper is to present a novel solid respirometric methodology, designed to measure the maximum nitrification rate in a VFCW. The novelty is based on the fact that the solid respirometric method allows handling matrices in partially saturated conditions, such as those found in the porous media of a VFCW. The effects of the addition of substrates, the hydraulic loading rates, and the injected nitrogen load on the measured nitrification rate are assessed. Special attention is given to the interpretation of the nitrogen release and storage phenomenon. Subsequently, the robustness of the developed method is determined. Finally, the method is applied to assess turn-over parameters of the autotrophic biomass at different depths of a full-scale VFCW, that would be useful for models.

2.

Materials and methods

The solid respirometric equipment was set-up for partially saturated samples collected from a VFCW. The equipment is first described in part 2.1. Next in part 2.2, the method for measuring the maximum nitrification rate as well as the data for evaluating its robustness are presented. Finally, the application of the method for a full-scale VFCW plant is illustrated in part 2.3.

2.1.

The solid respirometer

2.1.1.

Equipment and operating conditions

The respirometric system (Fig. 1) was adapted from a concept developed for characterizing household waste (Tremier et al., 2005). The system consisted of six stainless steel, temperature controlled (double envelope), cylinders of an overall height of 0.40 m and an internal diameter of 0.22 m. The temperature of

Fig. 1 e Solid respirometer device.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 9 5 e5 0 0 4

the reactor was maintained at 20  C with a cooler (Lauda WKL903). The VFCW sample was mechanically mixed with a woody bulking agent (wood chips of average size 3  2  0.5 cm) that was previously immersed in tap water for 24 h, using an end-over-end tumbler (Firlabo, operated at 60 rpm for 2 min). For colonized gravel from the VFCW (e.g. d10 ¼ 2.46 mm; UC ¼ 1.39; initial porosity of 40%), the volumetric ratio bulking agent/sample yielded 50/50, allowing to maximize both porosity and oxygen uptake. For sludge, a volumetric ratio bulking agent/sample of 75/25% was used. The application of a mechanical mixing was found to be essential to obtain a good repeatability. A total volume of 3 L of the mixture was then introduced in each cell of the solid respirometer, and the reactors were hermetically closed. Each reactor was subsequently subjected to a continuous supply of ambient airflow (0.5e1 L/min) mixed with recycled air (7e10 L/min). These conditions led to a ratio between recycled and ambient air between 10 and 20, allowing to obtain perfect mixed conditions and no limiting oxygen rates (Berthe et al., 2007). Gas flow rates applied to every reactor were quantified by volumetric sensors (Gallus, 2000; Actaris). After a previous cooling stage (Peltier, PKE 511), allowing to eliminate the humidity of the gases, the outgoing gases of every reactor were analyzed with an on-line analyzer (Servomex 4900) equipped with a paramagnetic detector for oxygen (0e100%) and with optical sensors for the analysis of carbon dioxide (CO2, 0e3000 ppm), methane (CH4, 0e500 ppm) and nitrous oxide (N2O, 0e500 ppm). The outgoing gas of each reactor was analyzed every hour for 8 min, and the incoming gas was analyzed every hour for 12 min.

2.1.2.

Sensitivity of the method

The bulking agent improved the porosity and thus the aeration efficiency of the sample (Tremier et al., 2005), without extra oxygen uptake. It also maintained the humidity for the growth of microorganisms (Berthe et al., 2007). The calibration of the gas analyzer was carried out twice a week and before starting a new experimental campaign. This guaranteed a gas measurement precision of 0.01%. To improve the reliability of the method, connecting pipes were changed regularly to limit supplementary oxygen uptake caused by microorganisms that develop in the connection pipes of the recirculation line (oxygen uptake called “intrusive” uptake). Moreover, the total “intrusive” oxygen uptake (blank test) of each reactor and connecting pipe were systematically measured before filling the reactors. This intrusive uptake was subtracted from the uptake measured with the mixture introduced in the reactor. With this protocol, a relative standard deviation between two replicates smaller than 20% was obtained.

2.1.3.

Respirometric index

In the following equations, the oxygen uptake rates were expressed in terms of dry matter (DM). For the measurement of DM, the water of collected samples (sludge and colonized gravel) was evaporated until dried in a drying oven (105  C) for at least 24 h. DM was calculated by the difference of weights before and after drying according to the standardized method published by APHA (2005). Duplicate determination coincided with the 5% of the measured weight.

4997

The difference in the oxygen concentration between the out reactor entrance (Oin 2 ðtÞ, [%]) and the reactor exit (O2 ðtÞ, [%]) provided the oxygen uptake rate by mass unit (noted DRI, [mg O2/kg DM/h]) as shown in Eq. (1). This variable was related to the biomass activity within a reactor. The total consumed oxygen mass during the period tF (CO(tF), [mg O2/kg DM]) was defined by integrating DRI (t) between t ¼ 0 to tF, as shown in Eq. (2) h i out in Oin 2 ðtÞ  O2 ðtÞ  Qair  60  1000  32 dO2 ðtÞ ¼ DRIðtÞ ¼ Vmolar  Msample dt

(1)

ZtF COðtF Þ ¼

DRIðtÞ:dt

(2)

0 in with Qair [L/min], the gas flow rate applied to a reactor; Vmolar(T ) [L/mol] the molar volume of the outgoing air considering its temperature (T [ C]); Msample [kg DM] the mass of dry sample to characterize; and tF the length of the experiment. The conversion factors are as follows: 60 to convert from minute to hour; 1000 to convert from g to mg O2; 32 to convert from mol O2 to g O2. Fig. 2 illustrates examples of the measured time course of DRI(t) and CO(tF) for three wastes like sand, colonized gravel and sludge. For sand containing few biodegradable organic matter (lab-scale columns fed with synthetic wastewater, Rolland et al., 2009), CO(tF ¼ 200 h) ranges between 480 and 1000 mg O2/kg DM. For such low CO(tF), the difference between replicates was about 40% which prevented using the proposed method for samples with a too low amount of organic matter. For colonized gravel sampled at the first stage of a VFCW (situated in Evieu (France) described in Molle et al. (2008), we found CO(tF ¼ 200 h) that ranged between 3840 and 6400 mg O2/kg DM. As for sludge from a drying reed bed situated in Andancette (France), described in Troesch et al. (2009), we found CO(tF ¼ 200 h) that ranged between 48 000 and 320 000 mg O2/kg DM. For both organic matrices, the difference between replicates was lower than 20%.

2.2.

Determining the nitrification rate

2.2.1. index

Liquid addition, percolate release, and nitrification

To study nitrification, each reactor was equipped with a spraying system (2 L/min) allowing a uniform supply of liquid solution to the sample. To limit the dissolution of the oxygen when spraying, and to limit the sur-saturation phenomenon for oxygen, the liquid was first aerated during 12 h and, subsequently, at rest for 1 h before being injected in a reactor. A typical time course of DRI(t) recorded with an addition of liquid is shown in Fig. 3. The DRImax was determined by the most important difference between the oxygen concentrations entering and leaving the reactor. The CO(tF) (Eq. (2)) was equal to the surface area located between the DRI(t) curve and the horizontal line.

2.2.2.

Calculating nitrification rates from DRImax and CO(tF)

The actual oxygen uptake rate (dO2 =dt, [mg O2/kg DM/h]) in local conditions can be calculated using conceptual models, as

4998

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 9 5 e5 0 0 4

Fig. 2 e Examples of DRI(t) and CO(t) results for carbon degradation in three organic wastes (colonized sand and gravel, sludge).

shown in Eq. (3). These conceptual models are based on activated sludge models (Henze et al., 1987) and extended to   VFCW (Sim unek et al., 1999). They require using five of the following parameters: the maximum autotrophic growth rate (mA max, [1/h]), the cellular yield coefficient (YA, [g CODbiomass/ g NH4eNnitrified]), the saturation coefficient for oxygen (KANs;O2 , [mg O2/L]), nitrogen (KANs,NH4, [mg NH4eN/L]), and inorganic phosphorus (KANs,IP, [mg P/L]). Also the four following variables are used: the oxygen concentration (CO2 , [mg O2/L]), the ammonium (CNH4 , [mg N/L]), the inorganic phosphorus concentration (CIP, [mg P/L]) and the biomass concentration (CXA, [mg CODbiomass/kg DM]): dO2 4:57  YA CO2 CNH4 ¼ mA;max $ $ dt YA KANs;O2 þ CO2 KANs;NH4 þ CNH4 CIP $ $CXA KANs;IP þ CIP

(3)

When no limiting conditions occur in terms of oxygen, nitrogen and phosphorus, the previous equation can be simplified into:   dO2 4:57  YA ¼ $mA;max $CXA ¼ DRImax ¼ Rs;max ¼ r:Rv;max dt max YA (4) From the value of DRImax Eq. (4), the maximum nitrification rates Rs,max [mg O2/kg DM/h] were determined. This value was linked to the maximum volumetric nitrification rate RV,max [mg O2/Lsample/h] using the concentrations of solids in the layer, noted r [g DM/Lsample]. CO(tF) and DRImax gave the nitrified mass using a conversion factor of 4.3 g O2 consumed for 1 g NH4eN nitrified (¼4.57-YA). An alternative method in determining the nitrified mass was to carry out a mass balance on chemical analysis data. After passing through the mixture made of the bulking agent and sample, each percolated solution was collected and immediately filtered. The composition of the percolate was determined by using chemical analysis COD, NH4eN, NO3eN and NO2eN, according to standardized methods (APHA, 2005).

2.2.3.

the concentration and the volume of the liquid solution on nitrification was studied in detail for the colonized gravel from a VFCW. Two types of liquids were studied to measure the nitrification rate: an ammonium-free tap-water to study the need to clean the organic matter from CW before measuring nitrification; and tap water containing ammonium. The chemicals that were considered in the liquids were ammonium chloride (NH4Cl) to feed nitrifiers, and sodium bicarbonate (NaHCO3) to prevent from mineral carbon limitation. Different concentrations and volumes of injected water were evaluated in order to study the role of the mass of nitrogen applied on nitrification rate. The storage and depletion (release) phenomenon of nitrogen were studied following the chemical analysis of batches. Samples of colonized gravel collected within the first stage of the VFCW from a plant located in Evieu (France) (Molle et al., 2008), were mixed with the bulking agent (see 2.1.1), and were characterized in the reactors of the solid respirometer device at 20  C. The DRI (t) curves were determined under the following supply conditions (Table 1): e injections of three different volumes (250, 500 and 1000 mL) with two concentrations of ammonium (30 and 60 mg NH4eN/L), implying four thresholds nitrogen concentrations (7.5, 14.3, 28.5, 60.0 mg N). Among them,

Experimental conditions tested to adjust liquid injection

The strategy carried out to adjust, evaluate and apply the solid respirometric method is summarized in Table 1. The effects of

Fig. 3 e Respirogram obtained after addition of ammonia for the nitrification study.

4999

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 9 5 e5 0 0 4

Table 1 e Strategy to adjust, evaluate and apply the solid respirometric method. Evaluation of the method eTesting for the robustness and comparison with “classical” protocols or literature e

Design of the solid respirometric method for nitrification e Testing of the experimental conditions on respiration rates e e Impact of the injected volume of liquid; e Impact of injections of ammonium-free water, and subsequent injections of water with ammonium; e Impact of injections of water with ammonium only (no previous addition of ammonium-free water); e Impact of injected mass of nitrogen.

Application to a VFCW plant

e Comparison with NO3eN production rate in liquid conditions (batch test) with biofilm (taken off from gravels by a mechanical step), immersed in liquid and excess of oxygen; e Comparison with NH4eN uptake and NO3eN release with analysis in percolates of respirometer; e Comparison with literature values from other methods.

two threshold concentrations (14.3 and 28.5 mg N) were applied with two different volumes (Fig. 4); e injections of ten successive volumes of 1 L of liquid; six of them were carried out with ammonium-free water (tap water). Then, four batches were carried out with water containing ammonium (60 mgNH4eN/L) and bicarbonate in stœchiometric quantity (2 batches/day); e injections of six successive volumes of 1 L of water containing ammonium (90 mg NH4eN/L) and bicarbonate in stœchiometric quantity (2 batches/day); and, e injections with different masses of nitrogen (8, 15, 28, 55, 90 mg) applied to five samples of colonized gravel, and determined after the application of a succession of a 1 L-batch with water containing ammonium. For each experimental condition detailed above, the maximum specific nitrification rates (Rs,max, [mg O2/kg DM/h]) were determined as explained in 2.2.2. Moreover, the mass of released ammonium and nitrates was calculated from the chemical composition of the liquid that percolated.

2.2.4. Other conventional protocols for comparing solid respirometry results The maximum volumetric nitrification rates (RS,max, [mg O2/ kg DM/h]) obtained from solid respirometry experiments were

e Measurement of the maximum nitrification rate (Rv,max) [with DRI measured after four successive injections of 1 L of liquid solution (90 mgNH4eN/L); e Mass balance calculated with input and output nitrogen fluxes calculated [inflow rate (19 m3/d) and nitrified mass (420 g Nnitrified/d)]. This aims at characterizing the autotrophic degradation process in a full-scale VFCW; e Determination of mA,max

compared to the values obtained with a common protocol generally applied to evaluate the maximum nitrification rate of an activated sludge. The batch test consisted in measuring the NO3eN production rate in liquid conditions (noted RS,liq) under non limiting conditions (mixing, aeration, substrate availability) as detailed in the protocol published by Water Environment Research Foundation (WERF, 2003). The batch test was applied after a mechanical removal of the biofilm composed of organic matter and biomass, attached to the gravel. The biofilm was removed by a rigorous mixing 1 L of colonized gravel placed in 3 L of water. This step was repeated three times successively, until clean gravel was obtained. The nitrification rate was converted as detailed in 2.2.2. The experiment was carried out for 5 samples of colonized gravel, collected every two weeks.

2.3. Determining autotrophic biomass parameters developed in a VFCW plant In the VFCW, three horizontal layers were differentiated: only sludge deposit (1st layer, h1 ¼ 20 cm of thickness), colonized gravel with biofilm (2nd layer, h2 ¼ 30 cm of thickness), and almost clean gravel (3rd layer, h3 ¼ 30 cm of thickness) with a negligible activity. The section below describes sampling, measurement, and calculation strategies that were 35

50 N = 30 mg/L

N = 30 mg/L

N = 60 mg/L

N = 60 mg/L

60.0

10 5

14.3

28.5

1000 mL

28.5

15 1000 mL 500 mL

1000 mL

14.3

0

20

500 mL 250 mL

1000 mL 500 mL

10

500 mL 250 mL

20

25

250 mL

30

250 mL

CO (mgO2 / kgDM )

40

DRImax (mgO2 / kgDM / h)

30

0 7.5

Injected nitrogen mass (mg)

7.5

60.0

Injected nitrogen mass (mg)

Fig. 4 e CO(tF) (left) and DRImax (right) measured for different batches loaded with different nitrogen load.

5000

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 9 5 e5 0 0 4

implemented in order to determine the turn-over parameters of the autotrophic biomass,.

2.3.1.

Sampling and preparation

Samples of sludge and colonized gravel were shovelled up at two locations of the first stage of the VFCW. Hence, one sample was collected for each 15 m2 of filter. Both sampling points were located at 2 m of the feeding point, providing raw waste water input to the VFCW. The duplicate samples collected for each layer are merged, then subsequently prepared with the bulking agent (as described in 2.1), and introduced in the reactors of the solid respirometer device.

2.3.2. Determining the maximum nitrification rate (RS,max) with solid respirometry The DRI(t) and DRImax were determined at 20  C for the two samples (sludge and colonized gravel) after four successive batches of 1 L, containing a solution of 90 mg NH4eN/L, was applied. The maximum specific nitrification rates, RS,max(h1) and RS,max(h2) [mg N/g DM/h], were then determined. The concentrations of solids in each layer, r1 and r2 [g DM/Lsample], were used to convert the individual values into volumetric nitrification rates noted RV,max (h1) and RV,max (h2). By weighting with the respective thickness (h1 and h2) of each layer, the global value (for the two first layers), RV,max [mg N/Lsample/h], was calculated.

2.3.4. Calculating the maximum autotrophic growth rate (mA,max) Using the previous calculations explained in 2.3.3. (CXA) and in 2.3.2. (global RV,max), the maximum autotrophic growth rate (mA,max [h1]) was determined with Eq. (6).

mA;max ¼ YA $

h1 $Rv;max ðh1 Þ þ h2 $Rv;max ðh2 Þ 1 $ MXA h1 þ h2 V1 þ V2

(6)

with YA the cellular yield coefficient (0.24 g CODbiomass/ g NH4eNnitrified, Henze et al., 1987).

3.

Results and discussion

3.1. Impact of injecting liquid and consequences on measuring nitrification

2.3.3. Calculating the autotrophic concentration (CXA) by means of a nitrogen mass balance The daily net growth of the autotrophic biomass that accumulated in the system was modelled with an approach adapted from Nowak et al. (1994) presented in Eq. (5). It used the daily nitrified nitrogen fluxes in a wastewater treatment plant as determined from the flow rates and the concentrations of different nitrogen forms of raw and treated water. The amount of the autotrophic biomass produced each day by nitrification, MXA(t), [g COD], was estimated by the sum of daily nitrified nitrogen (4Nnitrified, [g Nnitrified/d]), assuming that the autotrophic biomass input from the influent was negligible. The fraction that disappears each day was estimated from a decay process (bA, [1/d]) and from the evaluation of the mass removed by treated effluent (4XAeffluent, [g COD/d]), as there was no regular withdrawal of the excess sludge, but resting period with no feed. Moreover uptake of nitrogen by reeds and denitrification were considered as negligible as shown by Molle et al. (2008) and Tanner (1996). The typical decay rate value (bA) was chosen similar to thoses determined for secondary sludge from an activated sludge process [0.0071 h1 (¼0.17 d1) at 20  C (Dold et al., 2005)], as it is not easily measurable. The daily net growth of the autotrophic biomass yields: dMXA ðtÞ ¼ YA :4Nnitrified  bA :MXA ðtÞ  4XAeffluent dt

concentration of 100 g CODbiomass, the successive mass balances converged towards a reliable value, independent of the initial value (Nowak et al., 1999). The global concentration of autotrophic biomass accumulated in the VFCW (CXA) was calculated by dividing the mass of autotrophic biomass (MXA) by the volume of the active biological reactor (V1þV2), located from the surface layer to the bottom of the second layer of gravel (h1þh2 ¼ 50 cm height), assuming that the third layer (30 cm at the bottom of VFCW) had no biological activity (almost clean gravel).

(5)

This equation was applied to the VFCW from Evieu that treated 19 m3/d of water and yielded a daily nitrogen nitrification flux (4Nnitrified) of 420 g Nnitrified/d (14 g Nnitrified/m2.d). This latter value was computed from concentrations of different nitrogen forms in daily average, flow proportional, composite samples. Assuming an initial autotrophic

The conditions of injections were studied as described in part 2.2.3.

3.1.1. Impact of the injected volume and the concentration of nitrogen Fig. 4 presents CO(tF) and DRImax for batches of 250 mL, 500 mL and 1 L of tap water, containing either a concentration of 30 or 60 mg NH4eN/L with bicarbonate, for a nitrogen mass applied of 7.5e60 mg N. Relative standard deviation indicates the discrepancy around the mean values. For a given injected nitrogen mass of 14.3 and 28.5 mg N, the CO(tF) and the DRImax remain rather constant and independent of the injected volumes or the nitrogen concentrations that was applied. No differences were observed for different injected volumes. Also, no influence of the hydraulic retention time was observed. Only the injected nitrogen mass had an impact on CO(tF) and consequently on the nitrification rate. No influence of the hydraulic retention time was measured. Further investigations are presented in part 3.1.4. When the injected volume increased from 250 to 500 mL, and from 500 mL to 1 L, for a given nitrogen concentration (30 or 60 mg NH4eN/L), the CO(tF) and the DRImax increased by 20e50%, while the injected mass of nitrogen doubled. Similar observations were carried out when the nitrogen concentration increased from 30 to 60 mg NH4eN/L for a given injected volume of water (250, 500 or 1000 mL). Thus we can conclude that the increase was not proportional to the injected volume or to the NH4eN concentration, probably due to a limitation in the mixture coming from the oxygen transfer or due to the hydraulic retention time.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 9 5 e5 0 0 4

Nitrogen mass (mgN)

30 25

10 mg N. Consequently, depletion and then storage of nitrogen are two processes that disturb the measurement of the nitrification rate in the biofilm. The application of successive batches of water containing nitrogen only is thus studied in the following paragraph.

NH4-N injected NH4-N percolate NO3-N percolate N nitrified (calculated from CO)

20 15

3.1.3. Impact of adding only water containing ammonium (no previous ammonium-free water)

10 5 0

5001

1

2

3

4

5

Addition of ammonium-free water

6

7

8

9

10

Addition of water Number of with ammonium batches

Fig. 5 e NH4eN, NO3eN and equivalent CO(tF) measured (mg N) for the successive addition of water with or without ammonium.

3.1.2. Impact of adding “ammonium-free” water, and subsequent water containing ammonium Fig. 5 presents the mass of nitrate nitrogen (NO3eN) and ammonium (NH4eN) released in the percolates, when successive 1 L batches were applied to a respirometric reactor. Six batches of ammonium-free water (tap water) were first applied, and then four batches of water containing ammonium and bicarbonate were injected. The nitrified mass calculated from CO(tF) (conversion with 2.2.2.) is also shown. As the number of applied batches without ammonia increased, we measured a release of nitrates. The individual concentration of nitrates decreased from 10 mg NO3eN (batches 1 and 2) to 2 mg NO3eN (batches 5 and 6). At the same time, a small release was measured for ammonium (0.3 mg N for batches 1 and 2, and almost no release for the others). The nitrified mass, calculated from CO, decreased from 2 mg N, for batches 1 and 2, to 0.1 mg N for batches 5 and 6. When batches of water containing ammonium were applied, subsequently after batches with ammonium-free water, a storage process of nitrogen occurs through the reactor for an amount of 2e5 mg N per batch, varying in terms of the injected mass. Indeed, a removal of 10 mg NH4eN occured for batches 7 and 8, and 15 mg NH4eN for batches 9 and 10. Though, the released mass of nitrates was about 2 and 5 mg NO3eN for batches 7 to 10. The nitrified mass (calculated from CO) was between 5 and

Fig. 6 presents the results obtained with six successive 1 L batches of water containing ammonium and bicarbonate, with no previous additions of ammonium-free water. As the number of injections increased, we observed a decrease in the mass of nitrogen released through the percolates from 95 to 87 mg NH4eN. At the same time, the released mass of NO3eN increased from 1 to 6 mg NO3eN, and the nitrified mass (calculated from CO) increased from 1 to 5 mg N. In this protocol where water containing ammonium was used, we did not observe a storage of nitrogen during the injections. We still observed a difference of 1 mg N between the nitrates produced and the oxygen consumed by nitrification. This is in accordance with the accuracy of the respirometric method and chemical analysis. The CO(tF) increased from 7 to 17 mg O2/kg DM for the batches 1 to 3, and remained almost constant, approximately 17  2 mg O2/kg DM for batches 3 to 6 (Fig. 6, right). This observation suggests that a saturation was reached of the nitrogen stock in the biofilm, after application of at least four batches. However, the application of water containing ammonium, after previously cleaning the organic matter from CW with “ammonium-free” water, would require more than four batches to restore the depleted nitrogen stock.

3.1.4.

Influence of the injected nitrogen mass

Fig. 7 presents the values of mean and relative deviation of the specific oxygen uptake rate (RS,max, [mg O2/kg DM/h]), determined for five injected nitrogen masses in the range 8e90 mg N (noted LoadN, [mgN]). These values were obtained after applying four 1 L batches of water containing ammonium. The maximum oxygen uptake rate (RS,max) increased as the injected nitrogen mass increases. We observed that the uptake rate reached an approximate constant value of 34 mg O2/kg DM/h or 41 mg O2/L/h. Thus, the maximal nitrification rate of the sample of colonized gravel was reached in

Fig. 6 e Mass of NH4eN (injected and released), NO3eN (released), nitrified nitrogen (mg N) and CO(tF) measured for the successive addition of water with ammonium.

5002

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 9 5 e5 0 0 4

Rs,max and RS,liq (mgO2 /kgDM /h)

50

Solid respirometry Batch measurements

40 RS,max 30 (-0.052*loadN)

RS,max(loadN) = 34*[1-e

]

20

10

0 0

50

100

150

Injected nitrogen mass (Load N, mg)

Fig. 7 e Rvmax and Rvliq measured at different injected nitrogen mass.

such conditions (no oxygen limitation, 20  C). A logarithmic regression was performed using Eq. (7).

improved results either, due to the depletion and storage phenomenon.

  RS;max ðloadNÞ ¼ RS;max $ 1  eðkRv max $loadNÞ

3.2.2.

(7)

with RS,max the maximal oxygen uptake rate [mg O2/kg DM h], kRvmax the rate coefficient [1/mg N], and loadN [mgN] the injected nitrogen load. The following parameter values were estimated from this regression: RS,max ¼ 34 mg O2/kg DM/h and kRvmax ¼ 0.052/mg N. For further experimentations, we can chose to apply an injected nitrogen mass of 100 mg N to work at the conditions of the maximum level of nitrification rate, which corresponded to 40.8 g N/g DM injected.

3.2.

Evaluation of the proposed method

3.2.1.

Comparison with a conventional method

Nitrification rates obtained with the proposed solid respirometry (RS,max) were compared to batch measurements carried out by immersing the detached biofilm in liquid conditions (RS,liq) (see 2.2.4.). As shown in Fig. 7, the results of liquid conditions were located between 3 and 11 mg O2/kg DM/ h. A difference with the solid respirometry method of 5e30 mg O2/kg DM/h was obtained, suggesting that nitrifying biomass reacts differently. When partially saturated conditions were regularly applied, we thus suggest that the solid respirometry was more appropriate for assessing the nitrification rate in such systems. However, the solid respirometry was much more complex to implement. In addition, the use of nitrates produced and released through the percolates did not

Comparing existing literature values

In order to discuss our results, the values of maximum specific nitrification rates (RS,max, [mg O2/DM/h]) were converted into volumetric rates (RV,max, [mg O2/Lsample/h]), that were compared to literature values in Table 2. We observe that the maximum nitrification rate evaluated from the solid respirometric technique differed considerably with the results published by Andreottola et al. (2007) of 1.8 mg O2/Lsample/h. Notwithstanding the injected nitrogen amount was almost similar (90 mg), the respirometric method, carried out by Andreottola et al. (2007), involved passing liquid continuously through a colonized filtering column. On the contrary, our results are consistent with the maximum nitri  fication rate proposed by Langergraber and Sim unek (2005) of 30.5 mg O2/Lsample/h. They obtained maximum nitrification rates from inverse modelling with the HYDRUS-CW2D model, which used inflow/outflow pollutant fluxes from a VFCW.

3.3. Determining the concentration and the maximum growth rate of autotrophic biomass Table 3 presents the maximal volumetric nitrification rate (Rv,max) obtained with the solid respirometric technique applied to the samples collected and prepared as described in part 2.3.1. For the sludge (1st layer, r1 ¼ 187 g DM/Lsample), a maximum volumetric nitrification rate of 16.3 g N/Lsample/h was obtained. For the colonized gravel (2nd layer, r2 ¼ 114 g

Table 2 e Comparison of RV,max to literature values [mg O2/Lsample.h]. Method of RV,max Solid respirometry (partially measurement saturated conditions) Source (this study) Results

32e50 (mean ¼ 41) (SD ¼ 9; 2 values)

Nitrates production rate (batch liquid conditions) (this study)

Respirometrya (liquid conditions) (Andreottola et al., 2007)

Simulationsb (Langergraber and   Sim unek, 2005)

13e25 (mean ¼ 19) (SD ¼ 6; 5 values)

1.8

30.5

a Respirometric method applied to the liquid passing continuously through a colonized filtering column (Andreottola et al., 2007). b Calculated with a conversion of the predicted autotrophic biomass concentration (150 mgCOD/L once HYDRUS-CW2D was calibrated by   inverse modeling with inflow/outflow pollutants fluxes of a VFCW) using the values of mA,max and YA (Langergraber and Sim unek, 2005).

5003

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 9 5 e5 0 0 4

Table 3 e Maximum nitrification rate (Rv,max), concentration (CXA), fraction released by treated effluent and maximum growth rate (mA,max) for the autotrophic biomass. Layers

Overall VFCW 1st layer (h1 ¼ 20 cm sludge) 2nd layer (h2 ¼ 30 cm colonized gravels) 3rd layer (h3 ¼ 30 cm clean gravels)

ra [g DM/ Lsample]

Rv,max [g N/ Lsample/h]

CXA in reactor [mg CODbiomass/Lsample]

CXA in treated effluent [mg CODbiomass/L] (Fraction of the mass contained in reactor)

mA,max [1/h] (1/d )

143 187

19.5 16.3

123b 103c

0.04 (0.2 %) e

0.037 (0.88) e

114

21.7

137c

e

e

negligible

negligible

negligible

e

e

Dashed line when no value was determined. a Wet density of pure submerged biofilm is 1 kg/L (i.e. density of water) (Vigne et al., 2010). r includes the biofilm concentration in each layer, that is to say the concentration of dry matter once it was get rid off of dead leaves, roots, gravels. b Calculated according to 2.3.3 with mass balance. c Calculated with the value of mA,max and the individual values (each layer) of Rv,max (Eq. (4)).

DM/Lsample), a maximum volumetric nitrification rate of 21.7 g N/Lsample/h was obtained. Assuming a negligible nitrification rate for the deep layer of the VFCW, we calculated a global volumetric value RV,max of 19.5 g N/Lsample/h, from the two individual values weighted by the thickness of the two layers (part 2.3.2.). A negligible autotrophic biomass concentration of 0.04 mg CODbiomass/L released with total suspended solids (TSS) in the treated effluent was determined. This represented only 0.2% of the amount of autotrophic biomass stored in the VFCW. This level was very low compared to the usual values in treated water released by the activated sludge process (Feray et al., 1999; Vigne et al., 2010). A value of 123 mg CODbiomass/ Lsample was obtained for the global autotrophic biomass concentration (CXA) in the filter according to the method presented in 2.3.3. The maximum growth rate coefficient (mA,max) was calculated, using the approach presented in part 2.3.4. A value of 0.037 h1 (i.e. 0.88 d1) was obtained. This value is in accordance with usual values used for modelling of activated sludge process as reviewed by Choubert et al. (2009). Consequently, the concentration in autotrophic biomass in the two first layers of the VFCW was estimated to be 103 and 137 mg COD/Lsample, respectively. Therefore a comparison with the predicted values of the HYDRUS-CW2D model could be carried out. These values are very close to the values of 110 g   unek CODbiomass/Lsample predicted by Langergraber and Sim (2005) with the CW2D model.

4.

Conclusions

A novel respirometric tool was developed and applied to measure the nitrification rate in partially saturated, porous samples, collected from a VFCW, composed of sludge and colonized gravel. The best operating conditions for injecting liquids (volume, composition) were studied. The use of ammonium-free water before injecting water containing ammonium led to a depletion of the nitrogen storage in the

initial system and impeded correct nitrification rate measurements. The use of water containing ammonium only led to equilibrating the depletion/storage phenomenon. The comparison with liquid batch experiments indicates that the solid respirometry was more appropriate for the partially saturated conditions and was therefore closer to the conditions of their original medium. The method was compared to some conventional protocols and literature data. As compared to protocols using liquid conditions, the data were found to be more appropriate to fit within the range of simulated maximum nitrification rates reported in the literature. Indeed the autotrophic concentrations were closer to the values given by simulation once calibrated by inverse modelling with inflow/outflow pollutant fluxes of a VFCW. Then, the protocol was successfully applied to obtain the parameters of the autotrophic biomass of a VFCW. A value of maximum growth rate similar to the value of activated sludge process was determined. Finally, we conclude that the data provided with the solid respirometer are appropriate for a finer calibration of models of constructed wetlands.

Acknowledgement The authors thank Cle´ment Cre´tollier, Ame´lie de Sale´on, Nicolas Philippe, Vincent Nordmann, Domique Gorini and Loı¨c Richard for their valuable technical assistance in elaborating and running the respirometer, or for carrying out the chemical analyses. The authors also thank Ce´line Druilhe for the discussion in elaborating the research tool, and Ashley Tilghman-sibille for improvement of the English.

references

Adani, F., Confalonieri, R., Tambone, F., 2004. Dynamic respiration index as a descriptor of the biological stability of organic wastes. Journal of Environment Quality 33 (5), 1866e1876.

5004

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 9 9 5 e5 0 0 4

Andreottola, G., Oliveira, E., Foladori, P., Peterlini, R., Ziglio, G., 2007. Respirometric techniques for assessment of biological kinetics in constructed wetland. Water Science and Technology 56 (3), 255e261. APHA, AWWA, WPCF, 2005. Standard Methods for the Examination of Water and Wastewater, 21st ed. Amer Public Health Assoc., Amer. Water Works Assoc, Water Poll. Control Fed, Washington, DC. Berthe, L., Druilhe, C., Massiani, C., Tremier, A., de Guardia, A., 2007. Coupling a respirometer and a pycnometer, to study the biodegradability of solid organic wastes during composting. Biosystems Engineering 97 (1), 75e88. Choubert, J.M., Stricker, A.E., Marquot, A., Racault, Y., Gillot, S., He´duit, A., 2009. Updated activated sludge model no1 parameter values for improved prediction of nitrogen removal in activated sludge processes: validation at 13 full-scale plants. Water Environment Research 81 (9), 858e865. Dold, P.L., Jones, R.M., Bye, C.M., 2005. Importance and measurement of decay rate when assessing nitrification kinetics. Water Science and Technology 52 (10e11), 469e477. Feray, C., Volat, B., Degrange, V., Clays-Josserand, A., Montuelle, B., 1999. Assessment of three method for detection and quantification of nitrite-oxidizing bacteria and nitrobacter in freshwater sediments (MPN-PCR, MPN-Griess, Immunofluorescence). Microbial Ecology 37, 208e217. Ficara, E., Rozzi, A., 2001. pH-stat titration to assess nitrification inhibition. Journal of Environmental Engineering 127 (8), 698e704. Henze, M., Grady, C.P.L., Gujer, W., Marais, G.R., Matsuo, T., 1987. Activated Sludge Model N 1, IAWQ Scientific and Technical report No 1, London, ISSN: 1010e707X, 33 p. Kadlec, R.H., 2000. The inadequacy of first-order treatment wetland models. Ecological Engineering 15 (1), 105e119. Lagarde, F., Tusseau-Vuillermin, M.H., Lessard, P., He´duit, A., Dutrop, F., Mouchel, J.M., 2005. Variability estimation of urban wastewater biodegradable fractions by respirometry. Water Research 39 (19), 4768e4778. Langergraber, G., 2007. Simulation of the treatment performance of outdoor subsurface flow constructed wetlands in temperate climates. Science of the Total Environment 380 (1e3), 210e219.   Langergraber, G., Sim unek, J., 2005. Modelling variably saturated water flow and multicomponent reactive transport in constructed wetlands. Vadose Zone Journal 4 (4), 924e938. Langergraber, G., Giraldi, D., Mena, J., Meyer, D., Pena, M., Toscano, A., Brovelli, A., Korkusuz, E.A., 2008. Recent developments in numerical modelling of subsurface flow constructed wetlands. Science of the Total Environment 407 (13), 3931e3943. Langergraber, G., Rousseau, D.P.L., Garcia, J., Mena, J., 2009. CWM1: a general model to describe biokinetic processes in subsurface flow constructed wetlands. Water Science and Technology 59 (9), 1687e1697. Lasaridi, K.E., Stentiford, E.I., 1998. A simple respirometric technique for assessing compost stability. Water Research 32 (12), 3717e3723. Molle, P., Lie´nard, A., Boutin, C., Merlin, G., Iwema, A., 2005. How to treat raw sewage with constructed wetlands: an overview of the French systems. Water Science and Technology 51 (9), 11e21.

Molle, P., Prost-Boucle, S., Lie´nard, A., 2008. Potential for total nitrogen removal by combining vertical flow and horizontal flow constructed wetlands: a full-scale experiment study. Ecological Engineering 34 (1), 23e29. Nowak, O., Schweighofer, P., Svardal, K., 1994. Nitrification inhibition - a method for the estimation of actual maximum autotrophic growth rates in activated sludge systems. Water Science and Technology 30 (6), 9e19. Nowak, O., Franz, A., Svardal, K., Muller, V., Kuhn, V., 1999. Parameter estimation for activated sludge models with the help of mass balances. Water Science and Technology 39 (4), 3e120. Ortigara, A.R.C., Foladori, P., Andreottola, G., 2011. Kinetics of heterotrophic biomass and storage mechanism in wetland cores measured by respirometry. Water Science and Technology 64 (2), 409e415. Rolland, L., Molle, P., Lie´nard, A., Bouteldja, F., Grasmick, A., 2009. Influence of the physical and mechanical characteristics of sands on the hydraulic and biological behaviors of sand filters. Desalination 248 (1e3), 998e1007. Scaglia, B., Adani, F., 2008. An index for quantifying the aerobic reactivity of municipal solid wastes and derived waste products. Science of the Total Environment 394 (1), 183e191.   Sim unek, J., Senja, M., van Genuchten, M.Th., 1999. The HYDRUS2D Software Package for Simulating the Two-Dimensional Movement of Water, Heat, and Multiple Solutes in VariablySaturated Media. Version 2.0 Manual. U.S. Salinity Laboratory, USDA, Riverside, California, USA. Spanjers, H., Vanrolleghem, P., 1995. Respirometry as a tool for rapid characterization of wastewater and activated sludge. Water Science and Technology 31 (2), 105e114. Stricker, A.E., Lessard, P., Heduit, A., Chatellier, P., 2003. Observed and simulated effect of rain events on the behaviour of an activated sludge plant removing nitrogen. Journal of Environmental Engineering and Science 2 (6), 429e440. Surmacz-Gorska, J., Demuynck, C., Vanrolleghem, P., Verstraete, W., 1995. Nitrification process control in activated sludge using oxygen uptake rate measurements. Environmental Technology 16, 569e577. Tanner, C., 1996. Plants for constructed wetland treatment systems - a comparison of the growth and nutrient uptake of eight emergent species. Ecological Engineering 7 (1), 59e83. Tremier, A., de Guardia, A., Massiani, C., Paul, E., Martel, J.L., 2005. A respirometric method for characterising the organic composition and biodegradation kinetics and the temperature influence on the biodegradation kinetics, for a mixture of sludge and bulking agent to be co-composted. Bioresource Technology 96 (2), 169e180. Troesch, S., Lie´nard, A., Molle, P., Merlin, G., Esser, D., 2009. Treatment of septage in sludge drying reed beds: a case study on pilot-scale beds. Water Science and Technology 60 (3), 643e653. Vigne, E., Choubert, J.-M., Canler, J.-P., He´duit, A., Sorensen, K., Lessard, P., 2010. A biofiltration model for tertiary nitrification of municipal wastewaters. Water Research 44 (15), 4399e4410. Water Environment Research Foundation [WERF], 2003. Methods for Wastewater. Characterization in Activated Sludge Modeling. Report 99-WWF-3 , Alexandria, Virginia.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 0 5 e5 0 1 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Biodegradation of three selected benzotriazoles under aerobic and anaerobic conditions You-Sheng Liu a,b, Guang-Guo Ying a,b,*, Ali Shareef b, Rai S. Kookana b a

State Key Laboratory of Organic Geochemistry, Guangzhou Institute of Geochemistry, Chinese Academy of Sciences, Guangzhou 510640, PR China b CSIRO Land and Water, Water for a Healthy Country Flagship, PMB No.2, Glen Osmond, SA 5064, Australia

article info

abstract

Article history:

We examined the biodegradability of three benzotriazoles (benzotriazole: BT, 5-

Received 17 May 2011

methylbenzotriazole: 5-TTri and 5-chlorobenzotriazole: CBT) under aerobic and anaer-

Received in revised form

obic (nitrate, sulfate, and Fe (III) reducing) conditions. All three benzotriazoles were

29 June 2011

degraded by microorganisms under aerobic and anaerobic conditions. Both the biodegra-

Accepted 1 July 2011

dation efficiency and biodegradation products were dependent on the predominant

Available online 13 July 2011

terminal electron-accepting condition. Among the redox conditions studied, the shortest biodegradation half lives for BT and 5-TTri were 114 days and 14 days, respectively, under

Keywords:

aerobic condition. The shortest half-life for CBT was 26 days under Fe (III) reducing

Benzotriazole

condition. The longest biodegradation half lives for BT and CBT were 315 days and 96 days,

Biodegradation

respectively, under sulfate reducing condition, while that of 5-TTri was 128 days under

Aerobic

nitrate reducing condition. These results suggest that aerobic biodegradation is the

Anaerobic

dominant natural attenuation mechanism for BT and 5-TTri, while the most favorable

Redox conditions

process for CBT was anaerobic biodegradation. This study demonstrated that different predominant terminal electron-acceptors present in natural environment play a key role on the biodegradation of BT, 5-TTri and CBT, leading to specific biodegradability. This could have significant implications on in-situ biodegradation of the selected benzotriazoles in aerobic and anaerobic waters, soils and sediments. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Benzotriazoles (BTs), which include benzotriazole (BT) and its two derivatives 5-methylbenzotriazole (5-TTri) and 5chlorobenzotriazole (CBT), are commonly used as a corrosion inhibitor in dishwasher detergents and de-icing/antiicing fluids, an ultraviolet light stabilizer in plastics, and an antifogging agent in photography and airport. It is estimated that approximately 9000 tons per annum of BTs are produced and used in the United States alone, with a much greater

global production (Hart et al., 2004). Consequently, these compounds are widely detected in wastewaters and the receiving environments due to direct discharge or incomplete removal in wastewater treatment plants (WWTPs). Removal efficiencies in conventional WWTPs have been reported to be in the ranges of 13e62% for BT and 11e74% for 5-TTri (Hollingsworth et al., 2005; Giger et al., 2006; Voutsa et al., 2006; Weiss et al., 2006), with no published data available for CBT. Benzotriazole has been detected in surface water at concentrations up to 3.4 mg/L (Breedveld et al., 2003; Weiss and

* Corresponding author. State Key Laboratory of Organic Geochemistry, Guangzhou Institute of Geochemistry, Chinese Academy of Sciences, Guangzhou 510640, PR China. Tel./fax: þ86 020 85290200. E-mail addresses: [email protected], [email protected] (G.-G. Ying). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.001

5006

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 0 5 e5 0 1 4

Reemtsma, 2005; Giger et al., 2006; Reemtsma et al., 2006; Voutsa et al., 2006; Weiss et al., 2006; Kiss and Fries, 2009), in sediment and sludge at concentrations up to 198 ng/g (Zhang et al., 2011). Much higher concentrations of up to 126 mg/L for BT and 17 mg/L for 5-TTri have been found in groundwater from airport locations due to its heavy use as an antifogging agent (Cancilla et al., 1998, 2003). BTs are considered emerging contaminants due to potential adverse effects on aquatic species, microbial community and mammals (Davis et al., 1977; Sills et al., 1999; Pillard et al., 2001; Jia et al., 2006). From relative few available acute toxicity data, they suggest that BTs are relatively nontoxic, reported no-observed effect concentrations (NOEC) in freshwater and marine environments are usually in the mg/L range (Pillard et al., 2001; Kadar et al., 2010). However, many authors have commented that the chronic toxicity of these compounds still needs to be seriously concerned and a rigorous investigation of their chronic toxicity is necessary (Kadar et al., 2010; Harris et al., 2007; Farre et al., 2008). BTs in the environment are quite resistant to microbial degradation (Gruden et al., 2001; Weiss et al., 2006). With the exception of a few reports on removal of BTs in conventional WWTPs, very little is known about their biodegradation potential and mechanisms under environmental conditions (Breedveld et al., 2003; Weiss et al., 2006). Microbial respiration in aquatic systems and aquifers has been widely reported to take place via a variety of electron acceptors (Lovley and Phillips, 1986; Myers and Nealson, 1988; Blackburn and Blackburn, 1992; Canfield et al., 1993). Microbial degradation of organic compounds, especially substituted aromatic compounds, through microbial respiration processes can be influenced by the type of electron acceptors present in the aquatic environment (Gibson and Suflita, 1986; Kuhn et al., 1990; Haggblom et al., 1993; Kazumi et al., 1995; Krumholz and Sulflita, 1997; Milligan and Haggblom, 1999; Cortinas

et al., 2006). So far there has been little information available in the literature on the biodegradation of BTs under various redox conditions. This study investigated the biodegradation of BT, 5-TTri and CBT in laboratory batch studies under aerobic and anaerobic conditions. Fresh activated sludge and anaerobic digested sludge were used as inoculums for aerobic and anaerobic microcosms, respectively, in order to understand the fate of the three selected BTs in the environment and in WWTPs. Various reducing conditions were created to study the effect of electron acceptors on biodegradation of these three compounds by using the media amended with nitrate, sulfate or Fe (III) under anaerobic conditions. To the best of our knowledge, this is the first report on the biodegradation of the three BTs under various redox conditions.

2.

Materials and methods

2.1.

Chemicals

High purity standards benzotriazole (BT) (99%), 5methylbenzotriazole (5-TTri) (98%), and 5-chlorobenzotriazole (CBT) (98%) were obtained from SigmaeAldrich (Seelze, Germany), with their physicochemical properties shown in Table 1. Benzylcinnamate (99%, internal standard) was purchased from Dr. Ehrenstorfer GmbH (Augsburg, Germany). Fe (III) citrate, resazurin, Na-L lactate and NaN3 of analytical grade were obtained from SigmaeAldrich (St. Louis, MO, USA). NaNO3, Na2SO4 and Na2S of analytical grade were obtained from BDH (Kilsyth, Victoria, Australia). HPLC-grade methanol and dichloromethane were purchased from SigmaeAldrich (Seelze, Germany). Stock solutions (100 mg/L) of BT and 5-TTri were prepared in ultra pure water (Milli-Q), and CBT (100 mg/L) was prepared in methanol. All glassware was hand-washed with tap

Table 1 e Physicochemical properties and structures of three target benzotriazoles. Compound

Propertiesa

CAS number Molecular formula

MWb

Kocc (L/kg)

pKa

pKow

Henry’s law constant (atm-m3/mol)

benzotriazole (BT)

95-14-7

C6H5N3

119.13

62.3

8.37

1.44

1.47E-07

5-methylbenzotriazole (5-TTri)

136-85-6

C7H7N3

133.15

87.9

8.66

1.71

1.62E-07

5-chlorobenzotriazole (CBT)

94-97-3

C6H4ClN3

153.57

99.8

7.5/7.7

2.17

1.09E-07

a Source: http://www.syrres.com/what-we-do//databaseforms.aspx?id¼386. b MW, molecular weight. c Estimated by using EPIWEB 4.0 (KOCWIN), US EPA.

Structure

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 0 5 e5 0 1 4

water, rinsed with HPLC-grade water and methanol, and baked at 450  C for at least 4 h before use.

2.2.

Microcosms

Aerobic and anaerobic microcosms were prepared by using freshly obtained activated sludge and digested sludge as inoculums, respectively, from a WWTP in Adelaide, South Australia, with a daily treatment volume of 150 ML. The plant is equipped with activated sludge treatment (hydraulic retention time of 18 h) with a two-step feed plug flow design with anaerobic and anoxic zones (sludge retention time of 7 days), followed by six stabilization lagoons (with a cumulative hydraulic retention time of 27 days) and further polished using in-filter dissolved air flotation and chlorination. The mixed liquor suspended solids of fresh activated sludge and digested sludge were 28,500 mg/L and 29,000 mg/L, respectively. Incubation solutions with 10% of each inoculum (v/v) were prepared in minimal salts media consisting of KCl (1.3 g/L), KH2PO4 (0.2 g/L), NaCl (1.17 g/L), NH4Cl (0.5 g/L), CaCl$2H2O (0.10 g/L), MnCl2$6H2O and NaHCO3 (2.8 g/L) (Kazumi et al., 1995) and amended with trace salts and vitamins (Healy and Young, 1979). For anaerobic treatments, the medium was deoxygenated by boiling with nitrogen gas for 15 min and cooled in an anaerobic chamber.

2.3.

Biodegradation experiments

The biodegradation of BT, 5-TTri and CBT was evaluated individually under different redox conditions. Two 1 L glass Schott bottles with 500 mL media were set up for each testing treatment of aerobic and anaerobic- nitrate reducing, sulfate reducing and iron (III) reducing conditions. Sterile controls for each of the treatment were included. The test compounds were spiked into the incubation media at 1 mg/L in each treatment by pipetting 0.5 mL of each stock solution of BT, 5TTri and CBT. For the sterile controls of each treatment, the bottles containing the spiked media were autoclaved (120  C, 20 min) for three consecutive days followed by the addition of the metabolic inhibitor sodium azide (NaN3, 0.5% final concentration) to maintain sterility. Two replicate samples from each treatment were collected at pre-determined sampling time intervals (0, 7, 14, 21, 28, 35, 42, 49, 56, 70, 84 and 91 days following treatment).

2.3.1.

Aerobic treatment

In the batch test, 500 mL incubation solutions (10% of inoculums v/v) prepared in 1 L Shott bottles as described above, were incubated at 25  C in an orbital mixer incubator (Ratek OM11) with continuous shaking at 300 rmp. Aerobic condition was maintained by opening the caps three times a day in a laminar flow chamber.

2.3.2.

Anaerobic treatments

All the sample preparations for anaerobic treatments were conducted under an atmosphere of N2, inside an anaerobic chamber. 1 L bottles containing 500 mL incubation solutions with 10% of the anaerobic inoculum (v/v) were prepared as described above, and sealed with thick rubber stoppers and

5007

aluminum crimps. The incubation solution was either unamended with any of electron acceptors as the anaerobic 2 control (NO 3 : 0.03 mM; SO4 : 0.31 mM), or amended with either NaNO3 (20 mM), Na2SO4 (20 mM), or Fe (III) citrate (20 mM), as an electron acceptor to stimulate nitrate-reducing, sulfate-reducing or Fe(III) reducing conditions. Na2S (1 mM) and Na-lactate (10 mM) were added to all three reducing treatments serving as reducing agent and electron donor, respectively (Boopathy, 2002). Relatively high concentrations of the electron acceptors (nitrate, sulfate and iron (III)) have been used to achieve effective reducing conditions. To maintain anaerobic condition, the anaerobic chamber was flushed with a mixture of N2/CO2 gas (80:20, v/v). All the treatments contained the redox indicator resazurin at 0.0002% (w/w). The medium color turned from pink to colorless when the anaerobic condition in each vessel reached. Nitrate amended samples had to be supplemented with further nitrates (20 mM) to maintain the nitrate reducing condition after complete depletion during the period of incubation. At each sampling point the cultures were rigorously shaken and sampled with sterile syringes. All bottles were incubated in the dark at 25  C inside the anaerobic chamber.

2.4.

Extraction and analysis

2.4.1.

Analysis of BT, 5-TTri and CBT

For analysis of residual concentrations of each of the target analytes, a 15 mL aliquot of slurry was withdrawn from each treatment using glass syringe into glass culture tubes. The tubes were then screw capped using PTFE lined caps and centrifuged at 800 g for 20 min. The resulting supernatant was extracted/cleaned by using solid phase extraction (SPE) cartridges (Oasis HLB 6 mL 500 mg; Waters, Milford, MA, USA). The remaining pellets in the tubes were then extracted three times by ultrasonication with 2 mL methanol for 10 min, and after centrifugation the supernatant was diluted with Milli-Q water and put through the same SPE cartridge with the previous supernatant. All the extracts were analyzed by gas chromatography tandem mass spectrometry (GCeMS/MS, Agilent 7000A/ 7890A, USA). Target compounds in the samples were separated on an HP-5MS column (30 m  0.22 mm, 0.25 mm thickness) with helium as carrier gas at a linear flow rate of 1.656 mL/min. The GC oven temperature was programmed from 80  C (hold 2 min) to 280  C (hold 6 min) at a rate of 15  C/min. The injection port temperature was 280  C and transfer line temperature was 280  C. The MS/MS was operated in multiple reaction monitoring (MRM) mode. Ionization was carried out in high sensitivity electron impact (EI) mode at 70 eV, with the ion source temperature at 230  C. The MRM details of the target compounds are shown in Table S1. The target compounds were identified by comparing the retention times (within 2%) and the ratios (within 20%) of the two selected precursor-product ion transitions with those of the standards. Quantification was performed using the internal standard method (benzylcinnamate used as internal standard). Laboratory blanks were also analyzed along with the samples to assess potential sample contamination. Data acquisition was performed under Agilent Mass Hunter (Ver. B.03.01) application. Detailed extraction, analysis and

5008

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 0 5 e5 0 1 4

recovery test for the three compounds could be found in Supporting Information (SI).

2.4.2.

Analysis of biodegradation products

The biodegradation products of each target compound under different conditions (aerobic, nitrate reducing, sulfate reducing, iron (III) reducing and anaerobic control for BT; aerobic and anaerobic control for 5-TTri and CBT) after 91 days incubation were extracted (50 mL of slurry sample each) by using SPE as described earlier and analyzed using an Agilent 6890 gas chromatography coupled to a 5973 mass spectrometric detector (GCeMS) and a Thermo Finnigan TSQ Quantum liquid chromatography-tandem mass spectrometry (LC-MS/MS). Detailed instrumental conditions are given in SI. The mass spectrum of each peak in the total ion chromatograms (TIC) of GCeMS in scan mode was deconvoluted, and peaks were assigned identities using automated mass spectral deconvolution and identification system (AMDIS) (National Institute of Standards and Technology, Gaithersburg, MD, USA), which is able to identify chemical structures, estimate molecular weight, and generate chemical formulas for compounds corresponding to the respective peaks (Ausloos et al., 1999; Pongsuwan et al., 2007).

2.4.3.

Analysis of nitrate, sulfate and bacterial counting

In order to monitor the electron acceptors in the medium, nitrate and sulfate in the cultures were measured using a Dionex ICS-2500 ion chromatograph (Dionex. Sunnyvale, CA, USA), equipped with a 2 mm AS16 anion separation column and hydroxide eluent generated on line followed by conductivity detection after chemical suppression. Total numbers of culturable bacteria in the media of each BTs treatment were monitored on each sampling occasion using the most probable number technique (SI) (Ying et al., 2008).

3.

Results

3.1. Biodegradation of BT, 5-TTri and CBT under different redox conditions All three selected compounds (BT, 5-TTri and CBT) were found to be stable under sterile conditions. There was no hydrolysis occurring, and any removal due to volatilization during the testing period could also be accounted for based on the data for the sterile controls in each treatment. The biodegradation kinetics of the three target compounds followed the first-order model and the corresponding kinetic parameters including kinetic rate constant (k) and half-life (t1/2) are summarized in Table 2. Nitrate and sulfate as well as microbial activity were monitored during the incubation period. Anoxic NaNO3 or Na2SO4 solution was added to the NO 3 amended treatments,  or SO24 amended treatments to yield initial dissolved NO 3 concentrations of 16.12 mM (16.12 mM for BT, 17.01 mM for 5 TTri, 16.68 mM for CBT), or SO24 concentrations of 18.75 mM (18.75 mM for BT, 17.19 mM for 5-TTri, 16.98 mM for CBT), respectively. Nitrate reduction followed a time course with an initial rapid nitrate depletion (first 7 days) followed by a slow decrease in nitrate concentration after re-amendment of

Table 2 e Kinetic parameters for the biodegradation of benzotriazole (BT), 5-methylbenzotriazole (5-TTri) and 5chlorobenzotriazole (CBT) under aerobic, anaerobic control, nitrate reducing, sulfate reducing and Fe (III) reducing conditions. Compounda BT

5-TTri

CBT

Condition

k (mg/L/d)b

r2c

t1/2 (d)d

Aerobic Anaerobic control Nitrate reducing Sulfate reducing Fe (III) reducing Aerobic Anaerobic control Nitrate reducing Sulfate reducing Fe (III) reducing Aerobic Anaerobic control Nitrate reducing Sulfate reducing Fe (III) reducing

0.0061 0.0048 0.0029 0.0022 0.0042 0.0493 0.0121 0.0054 0.0079 0.0167 0.0081 0.0156 0.0089 0.0072 0.0268

0.861 0.908 0.927 0.913 0.899 0.893 0.906 0.948 0.955 0.960 0.835 0.948 0.988 0.970 0.841

114 144 239 315 165 14 57 128 88 41 86 44 78 96 26

a The initial concentration of each compound was 1 mg/L in all experiments. b Kinetic rate constant, which was predicted by using the firstorder reaction kinetic model (mean values from duplicate experiments were used in the calculation). c Correlation coefficient, which represents the fitness of the modeling data. d Half-life, calculated as (ln 2)/k.

nitrate during the remaining incubation period in the nitrate reducing treatments (Fig. S1). The depletion rate of sulfate for BT treatment under sulfate reducing conditions was similar to that for 5-TTri or CBT, showing very slow decreases in sulfate concentration during the incubation period (Fig. S2). Microbial activity in the treatments for BT increased after an initial stagnant phase of 14 days for most redox conditions (Fig. S3). The stagnant microbial phase seemed to have no significant impact on the biodegradation of BT as showed in Fig. 1. After the initial stagnant phase, the microbes maintained a stable growth. The total numbers of culturable aerobic bacteria (3.5  108 to 1.6  109) were found to be much higher than those of anaerobic bacteria (3.8  104 to 1.5  108).

3.1.1.

Benzotriazole (BT)

BT showed slow biological degradation in both aerobic and anaerobic microcosms within 91 days of incubation (Fig. 1). Differences in biodegradation rate for BT within the incubation period (91 days) were observed under different redox conditions: aerobic (46%) > anaerobic control (36%) > Fe (III) reducing (31%) > nitrate reducing (24%) > sulfate reducing (18%). All three reducing conditions (nitrate, sulfate and Fe (III) reducing) inhibited the anaerobic biodegradation of BT when compared to that of the anaerobic control condition, thus the redox conditions had obvious influences on the biodegradation of BT. The long half-lives of BT (114e315 d) in aerobic and anaerobic conditions (Table 2) suggested that BT was slowly biodegraded and would be inefficiently removed from the sewage system within limited hydraulic retention times. The biodegradation rate constant (k) values for BT under aerobic

5009

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 0 5 e5 0 1 4

a

a

120

120

100

100

80

C/C0 [%]

C/C0 [%]

80

60

60

40

40 20

Sterile Aerobic

20

0

7

14

21

Sterile Aerobic

0

28

35

42

49

56

63

70

77

84

91

0

98

7

14

21

35

42

49

56

63

70

77

84

91

98

77

84

91

98

Incubation time (days)

Incubation time (days)

b

28

b

120

120

100

100

C/C0 [%]

C/C0 [%]

80

80

60

40

Sterile Nitrate reducing Sulfate reducing Fe (III) reducing Anaerobic control

60

Sterile Nitrate reducing Sulfate reducing Fe (III) reducing Anaerobic control

20

0

40 0

7

14

21

28

35

42

49

56

63

70

77

84

91

0

98

7

14

21

28

35

42

49

56

63

70

Incubation time (days)

Incubation time (days) Fig. 1 e Aerobic (a) and anaerobic (b) biodegradation of Benzotriazole (BT) (initial concentration of 1 mg/L) by using 10% of activated sludge and digested sludge as inoculums, respectively. Error bars indicate standard deviations of the residual concentrations (n [ 2).

Fig. 2 e Aerobic (a) and anaerobic (b) biodegradation of 5-methylbenzotriazole (5-TTri) (initial concentration of 1 mg/L) by using 10% of activated sludge and digested sludge as inoculums, respectively. Error bars indicate standard deviations of the residual concentrations (n [ 2).

and anaerobic conditions were 0.0061 and 0.0048 mg/L/d, respectively, indicating that aerobic biodegradation of BT was slightly faster than anaerobic degradation.

The half-lives of 5-TTri under various redox conditions ranged from 14 days to 128 days (Table 2), which were much shorter than those of BT. This suggests that 5-TTri could be more susceptible to biodegradation and more effectively removed than BT in sewage treatment process.

3.1.2.

5-Methylbenzotriazole (5-TTri)

5-TTri was completely removed through biodegradation under aerobic conditions, and its removal under different redox conditions had the following order: aerobic (100%) > iron (III) reducing (76%) > anaerobic control (61%) > sulfate reducing (47%) > nitrate reducing (35%) within 91 days of incubation (Fig. 2). Compared to the anaerobic control, all the reducing conditions excepting Fe (III) reducing condition inhibited the biodegradation of 5-TTri. 5-TTri showed similar losses within the first 35 days, with 16% loss under nitrate reducing conditions and 19% loss under sulfate reducing conditions. The difference between the two treatments became more prominent with time after this period. This is possibly due to second supplementation of a high concentration of nitrate (20 mM) under the nitrate reducing condition.

3.1.3.

5-Chlorobenzotriazole (CBT)

The removal of CBT under different redox conditions within 91 days of incubation followed the following order: Fe (III) reducing (86%) > anaerobic control (71%) > nitrate reducing (53%) > aerobic (52%) > sulfate reducing (45%) (Fig. 3), indicating that the anaerobic conditions were in general more favorable for CBT to be degraded than aerobic conditions except for sulfate reducing conditions. This pattern of degradation under different redox conditions was similar to that in the case of 5-TTri. The half-lives of CBT under various redox conditions ranged between 26 days and 96 days, suggesting that CBT was much more biodegradable than BT under anaerobic

5010

a

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 0 5 e5 0 1 4

120

100

C/C0 [%]

80

60

40

20

Sterile Aerobic

0 0

7

14

21

28

35

42

49

56

63

70

77

84

91

98

77

84

91

98

Incubation time (days)

b

120

100

C/C0 [%]

80

60

40

20

Sterile Nitrate reducing Sulfate reducing Fe (III) reducing Anaerobic control

0

-20 0

7

14

21

28

35

42

49

56

63

70

Incubation time (days) Fig. 3 e Aerobic (a) and anaerobic (b) biodegradation of 5chlorobenzotriazole (CBT) (initial concentration of 1 mg/L) by using 10% of activated sludge and digested sludge as inoculums, respectively. Error bars indicate standard deviations of the residual concentrations (n [ 2).

conditions and under aerobic conditions as well (86 d for CBT, 114 d for BT) (Table 2). The results may also indicate that chlorine substitution in benzene ring of its chemical structure makes CBT less persistent than BT under anaerobic conditions.

3.2.

Biodegradation products of BT, 5-TTri and CBT

Preliminary assessment of various degradation products of BT in each treatment culture and of 5-TTri and CBT in aerobic and anaerobic control cultures was performed by GCeMS analysis of the dichloromethane/methanol extracts. Five degradation products including phenol (A), phthalic acid (B), 1-methyl benzotriazole (C), 1H-benzotriazole 4-methoxy (D) and 1Hbenzotriazole 5-methoxy(E) were identified for BT under aerobic conditions; and four degradation products i.e. phenol (A), 1-methyl benzotriazole (C), dimethyl benzylamine (F), carbazole (G) were identified under anaerobic conditions (Fig. S4 and Table S2). Two products 2,5-dimethyl benzoxazole

(H), and benzotriazole (I) and one product 5-chloro-2-methyl benzoxazole (L) were identified as the aerobic biodegradation products of 5-TTri and CBT, respectively (Fig. S5, Table S3 and Table S4). Three products benzotriazole (I), 2-methyl phenol (J), 2-methyl benzenamine (K) and two products benzotriazole (I), 2-methyl phenol (J) were identified as the main intermediates of 5-TTri and CBT under anaerobic conditions, respectively. Biodegradation products were also verified by LC-MS/MS analysis of the samples (Figs. S6, S7 and S8). For the products of BT under aerobic conditions, one main peak at m/z 150 was detected and identified as 1H-benzotriazole 4-methoxy (D) or 1H-benzotriazole 5-methoxy (E) (Fig. S6). For the products of BT under anaerobic conditions, only main peak at m/z 214 was detected by LC-MS/MS, but its structure has not been elucidated yet. For 5-TTri under aerobic conditions, only one peak at m/z 120 was detected by LC-MS/MS and identified as benzotriazole (I) in addition to its parent compound at m/z 134 (Fig. S7). Under anaerobic conditions, a peak with m/z 146 was detected as a transformation product of 5-TTri. For CBT under aerobic conditions, three peaks with m/z 146, 226 and 279 were detected in addition to CBT peak with m/z 154 (Fig. S8), while under anaerobic conditions three peaks with m/z 120, 146 and 148 were detected. It should be noted that the biodegradation products of BT, 5TTri and CBT after 91 days incubation were tentatively identified only based on GCeMS and LC-MS/MS results by using AMDIS program and NIST05 database. Further identification of degradation products by using synthetic standards, time-offlight mass spectrometry (TOF-MS) and NMR are needed to have better understanding of the transformation of these BTs under various redox conditions. Future study should also look into the toxicity of these degradation products.

3.3. Proposed biotransformation pathways for BT, 5TTri and CBT Based on the identified products, biotransformation pathways were tentatively proposed for the three target compounds (Figs. 4 and 5). Five biotransformation products (A: phenol, B: phthalic acid, C: 1-methyl benzotriazole, D: 1H-benzotriazole 4-methoxy and E: 1H-benzotriazole 5-methoxy) have been detected in aerobic biotransformation of BT (Fig. 4 and Table S2). We proposed that 1-methyl benzotriazole (C) was formed by methylation on nitrogen atom of the triazole chain of BT, and this kind of attack was also detected under nitrate reducing and anaerobic control conditions. The two isomers (D: 1H-benzotriazole 4-methoxy and E: 1H-benzotriazole 5methoxy) were formed via addition of OeCH3 group to the para and meta position carbon atom of benzene ring of BT. Phthalic acid (B) was formed via scission of triazole chain and then acidification of benzene ring. Phenol (A) was identified as the lowest molecular weight byproduct of BT. As phenol (A) and phthalic acid (B) are not likely to be direct products of BT, we propose that these compounds could be the secondary byproducts of BT. For anaerobic treatments for BT, 1-methyl benzotriazole (C) (Table S2) was also detected under the anaerobic conditions (anaerobic control and nitrate reducing), which suggests

5011

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 0 5 e5 0 1 4

OH

A O

CH3 N

OH N

C

-OH

OH

3

Aerobic

O

HC

O D

3

Scission -O-CH3

Scission

-CH3

N

N

O E

-C00H

N

N

HC

N B

H N

H N

H N N N

Anaerobic

BT

-CH3

-N2

-CH3

Scission

Polymerazation

Scission

CH3

CH3

N

N N

N C Scission

N N

HC 3

N

F

HC 3

C

CH3 -OH

H N

Scission -OH

F

N

G

CH3 -OH

Scission -OH

OH

OH

OH

OH

A

A

A

A

Anaerobic control

Nitrate-reducing

Sulfate-reducing

Fe(III)-reducing

Fig. 4 e Proposed schemes for the biotransformation of Benzotriazole (BT) under aerobic, anaerobic control, nitrate reducing, sulfate reducing and Fe (III) reducing conditions. The biotransformation products of BT were tentatively identified by GCeMS, and further confirmed by using the AMDIS and NIST05 database searching program. A: phenol; B: phthalic acid; C: 1-methyl benzotriazole; D: 1H-benzotriazole 4-methoxy; E: 1H-benzotriazole 5-methoxy-; F: dimethyl benzylamine; G: Carbazole.

initial methylation of BT occurring under both aerobic and anaerobic conditions. However, methylation and NeN scission of BT were inhibited under the nitrate, sulfate and iron (III) reducing conditions. Dimethyl benzylamine (F) was detected under sulfate reducing and anaerobic control conditions, and it was formed by NeN bond scission and methylation. Carbazole (G) was found under Fe (III) reducing conditions (Table S2) and formed by NeN bond scission followed by polymerization. Demethylation of 5-TTri to BT occurred under aerobic and anaerobic conditions (Fig. 5 and Table S3) and then BT

was transformed to 2-methyl benzenamine (K) via NeN bond scission and methylation, and further transformed to 2-methyl phenol (J) by release of nitrogen and addition of hydroxyl radicals in 5-TTri anaerobic biotransformation. Another pathway for 5-TTri under aerobic conditions was attack by oxygen and methyl radicals on the traizole chain leading to the formation of 2,5-dimethyl benzoxazole (H). Dechlorination of CBT to BT was the main pathway under the anaerobic conditions (Fig. 5 and Table S4). Then BT was further transformed to 2-methyl phenol (J). CBT had

5012

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 0 5 e5 0 1 4

O CH3 N

H3C -N2

H -O

H N

-CH3

N

-CH3

H N

I

N H3C

Aerobic

N

N 5-TTri

Anaerobic

-CH3

NH2 -N2

H N

OH -N

CH3

CH3

N -CH3

N -H

I

-OH

-CH3

H N

-O

O

N

CH3

N

Cl

J

K

Cl

-N2

CBT

Aerobic

N L

-Cl

OH

-N2

H N

CH3 Anaerobic

N N I

-CH3

-OH

J

Fig. 5 e Proposed schemes for the biotransformation of 5-methylbenzotriazole (5-TTri) and 5-chlorobenzotriazole (CBT) under aerobic and anaerobic conditions. The biotransformation products of 5-TTri and CBT were tentatively identified by GCeMS, and further confirmed by using the AMDIS and NIST05 database searching program. H: 2,5-dimethyl benzoxazole; I: benzotriazole; J: phenol, 2-methyl; K: benzenamine, 2-methyl; L: 5-chloro-2-methyl benzoxazole.

a similar biotransformation pathway with 5-TTri under the aerobic conditions. CBT was attacked by oxygen and methyl radicals on the triazole chain to form 5-chloro-2-methyl benzoxazole (L). Since the degradation pathways of these BTs were tentatively proposed based on the final degradation products, some intermediates formed during the incubation might not be identified. Further research should focus on the identification of degradation products as a function of time in order to have better understanding of their degradation pathways.

4.

Biodegradation mechanism

The results from this study demonstrated that the three selected BTs are biodegradable under aerobic and anaerobic conditions with some characteristically different half-lives based on the specific redox conditions involved. Both BT and 5-TTri showed the highest removal under aerobic conditions, while CBT had the highest removal under Fe (III) reducing conditions. However, BT was found very persistent with halflives more than 100 days under both aerobic and anaerobic

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 0 5 e5 0 1 4

conditions, suggesting its recalcitrance to biodegradation in the environment. BT was found more persistent than 5-TTri under aerobic conditions. The difference in degradation between the two compounds is consistent with their removal efficiencies in WWTPs (Voutsa et al., 2006; Weiss et al., 2006). However, CBT as the chlorinated derivative of BT had a similar degradation rates to BT under aerobic conditions. This suggests that initial demethylation of 5-TTri at the meta position might occur more easily than dechlorination of CBT under aerobic conditions. Under anaerobic conditions, CBT showed higher degradation rates than the other two target compounds (BT and 5-TTri). This is probably due to reductive dechlorination process for CBT. However, compared to the anaerobic controls, both nitrate and sulfate reducing conditions showed significantly inhibitory effects on biodegradation of all three compounds. Inhibition of anaerobic transformation of organic compounds by nitrate and sulfate has been reported previously (Milligan and Haggblom, 1999). Inhibition of dehalogenation of monoaromatics in the presence of sulfate was also well documented in the literature (Gibson and Suflita, 1986; Madsen and Aamand, 1991; Milligan and Haggblom, 1999). Significant faster biodegradation of 5-TTri and CBT under Fe (III) reducing conditions in the present study suggests that Fe (III) reducing condition as a common anaerobic terminal electron accepting process in aquifer systems (Finneran and Lovley, 2001; Bradley et al., 2002) are favorable for the degradation of 5-TTri and CBT, but not for BT in aquifer environment. This can explain higher persistence of BT than 5-TTri and CBT reported in some aquifer systems (Cancilla et al., 2003). The finding from this study suggest that redox conditions can significantly affect biodegradation rates of these three compounds could have significant environmental implications for their treatment in sewage treatment plants and groundwater systems where they are often found in high concentrations. Aerobic conditions are most favorable for BT and 5-TTri to be degraded, while Fe (III) reducing conditions are most favorable for CBT to be degraded. For anaerobic aquifers and other environments, addition of Fe (III) might improve the biodegradation of all three benzotriazoles.

5.

Conclusions

The results showed that the three selected BTs are biodegradable under aerobic and anaerobic conditions although the rate of degradation of the three compounds varied significantly. However, the longest biodegradation half lives for BT and CBT were 315 days and 96 days, respectively, under sulfate reducing condition; while that of 5-TTri was 128 days under nitrate reducing condition, suggesting their persistence in the environment. The predominant electron accepting process can affect the biodegradation rates and extents of three selected BTs under aerobic and anaerobic conditions. The biodegradation products of BT, 5-TTri and CBT under different electron acceptor conditions were tentatively identified for the first time. Specific redox condition can be created by amendment of certain electron acceptor to improve biodegradation of these benzotriazoles.

5013

Acknowledgments The authors would like to acknowledge the financial support from National Natural Science Foundation of China (NSFC 40821003, 20977092 and 40688001), Guangdong Provincial Natural Science Foundation (8251064004000001) and the Earmarked Fund from the State Key Laboratory of Organic Chemistry (sklog2009A02). We also thank China Scholarship Council and CSIRO Australia for the scholarships to YS Liu for his PhD project conducted at CSIRO laboratories. This is a Contribution No. 1365 from GIG CAS.

Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.07.001.

references

Ausloos, P., Clifton, C., Lias, S., Mikaya, A., Stein, S., Tchekhovskoi, D., Sparkman, O., Zaikin, V., Zhu, D., 1999. The critical evaluation of a comprehensive mass spectral library. Journal of the American Society for Mass Spectrometry 10 (4), 287e299. Blackburn, T.H., Blackburn, N.D., 1992. Model of nitrification and denitrification in marine sediments. FEMS Microbiology Letters 100 (1e3), 517e521. Boopathy, R., 2002. Anaerobic biotransformation of carbon tetrachloride under various electron acceptor conditions. Bioresource Technology 84 (1), 69e73. Bradley, P.M., Landmeyer, J.E., Chapelle, F.H., 2002. TBA biodegradation in surface-water sediments under aerobic and anaerobic conditions. Environmental Science and Technology 36 (19), 4087e4090. Breedveld, G.D., Roseth, R., Sparrevik, M., Hartnik, T., Hem, L., 2003. Persistence of the de-icing additive benzotriazole at an abandoned airport. Water, Air, and Soil Pollution: Focus 3 (3), 91e101. Cancilla, D.A., Baird, J.C., Rosa, R., 2003. Detection of aircraft deicing additives in groundwater and soil samples from Fairchild Air Force Base, a small to moderate user of deicing fluids. Bulletin of Environmental Contamination and Toxicology 70 (5), 868e875. Cancilla, D.A., Martinez, J., Aggelen, G.C.V., 1998. Detection of aircraft deicing/antiicing fluid additives in aperched water monitoring well at an international airport. Environmental Science and Technology 32 (23), 3834e3835. Canfield, D.E., Thamdrup, B., Hansen, J.W., 1993. The anaerobic degradation of organic matter in Danish coastal sediments; iron reduction, manganese reduction, and sulfate reduction. Geochimica et Cosmochimica Acta 57 (16), 3867e3883. Cortinas, I., Field, J.A., Kopplin, M., Garbarino, J.R., Gandolfi, A.J., Sierra-Alvarez, R., 2006. Anaerobic biotransformation of roxarsone and related N-substituted phenylarsonic acids. Environmental Science and Technology 40 (9), 2951e2957. Davis, L.N., Santodonato, J., Howard, P.H., Saxena, J., 1977. Investigations of selected potential environmental contaminants: benzotriazoles, Office of Toxic Substances, USEPA, EPA Document 560. 2-77-001. Farre, M., Perez, S., Kantiani, L., Barcelo, D., 2008. Fate and toxicity of emerging pollutants, their metabolites and transformation

5014

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 0 5 e5 0 1 4

products in the aquatic environment. TrAC Trends in Analytical Chemistry 27 (11), 991e1007. Finneran, K.T., Lovley, D.R., 2001. Anaerobic degradation of methyl ter-butyl ether (MTBE) and tert-butyl alcohol (TBA). Environmental Science and Technology 35 (9), 1785e1790. Gibson, S.A., Suflita, J.M., 1986. Extrapolation of biodegradation results to groundwater aquifers: reductive dehalogenation of aromatic compounds. Applied and Environmental Microbiology 52 (4), 681e688. Giger, W., Schaffner, C., Hans-Peter, E.K., 2006. Benzotriazole and tolyltriazole as aquatic contaminants. 1. Input and occurrence in rivers and lakes. Environmental Science and Technology 40 (23), 7186e7192. Gruden, C.L., Dow, S.M., Hernandez, M.T., 2001. Fate and toxicity of aircraft deicing fluid additives through anaerobic digestion. Water Environmental Research 73 (1), 72e79. Haggblom, M.M., Rivera, M.D., Young, L.Y., 1993. Influence of alternative electron acceptors on the anaerobic biodegradability of chlorinated phenols and benzoic acids. Applied Microbiology and Biotechnology 59 (4), 1162e1167. Harris, C.A., Routledge, E.J., Schaffner, C., Brian, J.V., Giger, W., Sumpter, J.P., 2007. Benzotriazole is antiestrogenic in vitro but not in vivo. Environmental Toxicology and Chemistry 26 (11), 2367e2372. Hart, D.S., Davis, L.C., Erickson, L.E., Callender, T.M., 2004. Sorption and partitioning parameters of benzotriazole compounds. Microchemical Journal 77 (1), 9e17. Healy, J.B., Young, L.Y., 1979. Anaerobic biodegradation of eleven aromatic compounds to methane. Applied Microbiology and Biotechnology 38 (1), 84e89. Hollingsworth, J., Sierra-Alvarez, R., Zhou, M., Ogden, K.L., Field, J. A., 2005. Anaerobic biodegradability and methanogenic toxicity of key constituents in copper chemical mechanical planarization effluents of the semiconductor industry. Chemosphere 59 (9), 1219e1228. Jia, Y., Bakken, L.R., Breedveld, G.D., Aagaard, P., Frostegard, A., 2006. Organic compounds that reach subsoil may threaten groundwater quality; effect of benzotriazole on degradation kinetics and microbial community composition. Soil Biology and Biochemistry 38 (9), 2543e2556. Kadar, E., Dashfield, S., Hutchinson, T.H., 2010. Developmental toxicity of benzotriazole in the protochordate Ciona intestinalis (Chordata, Ascidiae). Analytical and Bioanalytical Chemistry 396 (2), 641e647. Kazumi, J., Haggblom, M.M., Young, L.Y., 1995. Diversity of anaerobic microbial processes in chlorobenzoate degradation: nitrate, iron, sulfate and carbonate as electron acceptors. Applied Microbiology and Biotechnology 43 (5), 929e936. Kiss, A., Fries, E.F., 2009. Occurrence of benzotriazoles in the rivers main, Hengstbach, and Hegbach (Germany). Environmental Science and Pollution Research 16 (6), 702e710. Krumholz, L.R., Sulflita, J.M., 1997. Anaerobic aquifer transformation of 2,4-dinitrophenol under different terminal electron accepting conditions. Environmental Microbiology 3 (6), 399e403. Kuhn, E.P., Townsend, G.T., Suflita, J.M., 1990. Effect of sulfate and organic carbon supplements on reductive dehalogenation of

chloroanilines in anaerobic aquifer slurries. Applied and Environmental Microbiology 56 (9), 2630e2637. Lovley, D.R., Phillips, E.J.P., 1986. Organic matter mineralization with reduction of ferric iron in anaerobic sediments. Applied and Environmental Microbiology 51 (4), 683e689. Madsen, T., Aamand, J., 1991. Effects of sulfuroxy anions on degradation of pentachlorophenol by a methanogenic enrichment culture. Applied and Environmental Microbiology 57 (9), 2453e2458. Milligan, P.W., Haggblom, M.M., 1999. Biodegradation and biotransformation of dicamba under different reducing conditions. Environmental Science and Technology 33 (8), 1224e1229. Myers, C.R., Nealson, K.H., 1988. Bacterial manganese reduction and growth with manganese oxide as the sole electron acceptor. Science 240 (4857), 1319e1321. Pillard, D.A., Cornell, J.S., Dufresne, D.L., Hernandez, M.T., 2001. Toxicity of benzotriazole and benzotriazole derivatives to three aquatic species. Water Research 35 (2), 557e560. Pongsuwan, W., Fukusaki, E., Bamba, T., Yonetani, T., Yamahara, T., Kobayashi, A., 2007. Prediction of Japanese green tea ranking by gas chromatography /mass spectrometry-based hydrophilic metabolite fingerprinting. Journal of Agricultural and Food Chemistry 55 (2), 231e236. Reemtsma, T., Weiss, S., Mueller, J., Petrovic, M., Gonzalez, S., Barcelo, D., Ventura, F., Knepper, T.P., 2006. Polar pollutant entry into the water cycle by municipal wastewater: a European perspective. Environmental Science and Technology 40 (17), 5451e5458. Sills, R.C., Hailey, J.R., Neal, J., Boorman, G.A., Haseman, J.K., Melnick, R.L., 1999. Examination of low-incidence brain tumor responses in F344 rats following chemical exposures in national toxicology program carcinogenicity studies. Toxicologic Pathology 27 (5), 589e599. Voutsa, D., Hartmann, P., Schaffner, C., Giger, W., 2006. Benzotriazoles, alkylphenols and bisphenol A in municipal wastewaters and in the Glatt River, Switzerland. Environmental Science and Pollution Research 13 (5), 333e341. Weiss, S., Jakobs, J., Reemtsma, T., 2006. Discharge of three benzotriazole corrosion inhibitors with municipal wastewater and improvements by membrane bioreactor treatment and ozonation. Environmental Science and Technology 40 (23), 7193e7199. Weiss, S., Reemtsma, T., 2005. Determination of benzotriazole corrosion inhibitors from aqueous environmental samples by liquid chromatography electrospray ionization tandem mass spectrometry. Analytical Chemistry 77 (22), 7415e7420. Ying, G.G., Toze, S., Hanna, J., Yu, X.Y., Dillon, P.J., Kookana, R. S., 2008. Decay of endocrine disrupting chemicals in aerobic and anoxic groundwater. Water Research 42 (4e5), 1133e1141. Zhang, Z.F., Ren, N.Q., Li, Y.F., Kunisue, T., Gao, D., Kannan, K., 2011. Determination of benzotriazole and benzophenone UV filters in sediment and sewage sludge. Environmental Science and Technology 45 (9), 3909e3916.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 1 5 e5 0 2 6

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Carbon-sensitized and nitrogen-doped TiO2 for photocatalytic degradation of sulfanilamide under visible-light irradiation Penghua Wang a,b, Tao Zhou a,c, Rong Wang a,b, Teik-Thye Lim a,b,* a

School of Civil and Environmental Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798, Singapore Singapore Membrane Technology Centre, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798, Singapore c DHI-NTU Water and Environment Research Centre and Education Hub, Singapore 639798, Singapore b

article info

abstract

Article history:

A novel carbon-sensitized and nitrogen-doped TiO2 (C/NeTiO2) was synthesized by a facile

Received 10 April 2011

sol-gel method using titanium butoxide as both titanium precursor and carbon source, and

Received in revised form

nitric acid as nitrogen source. The calcination temperature had a great effect on the crystal

1 July 2011

phase structure, nitrogen incorporation into the TiO2 lattice and content of carbonaceous

Accepted 2 July 2011

species. The incorporated carbonaceous species could serve as photosensitizer, while the

Available online 18 July 2011

nitrogen doping could lead to the remarkable red shift of absorption edge of C/NeTiO2. The C/NeTiO2 calcinated at 300
C (T300) exhibited the highest photocatalytic activity for

Keywords:

sulfanilamide (SNM) degradation under irradiation of visible-light-emitting diode (vis-LED).

Titanium dioxide

The SNM photocatalytic degradation and mineralization were more efficient in acidic

Carbon-sensitizing

conditions due to the carbon photosensitizing effect. Insignificant inhibitory effects were

Nitrogen-doping

observed in the presence of chloride, nitrate and sulfate, while bicarbonate, phosphate and

Light-emitting diode

silica could inhibit the SNM mineralization to different degrees. Acetate, ammonium and

Sulfanilamide

sulfate were released during SNM mineralization. T300 exhibited good photochemical

Toxicity

stability and could be reused for 5 times with less than 10% decrease in the SNM removal efficiency. The acute toxicity of SNM solution could be reduced over prolonged photocatalysis according to the Microtox assay. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Sulfonamides are one class of antibiotics frequently used for human and veterinary medicine as antibacterial drugs and growth promoters (Ku¨mmerer, 2009a). However, sulfonamides and their metabolites could find their way into the aquatic environment, including runoff from municipal sewage or hospital wastewater system and animal feeding operations, infiltration from aquaculture activities, leaching from landfills and compost made from animal manure containing antibiotics. Their ubiquitous appearances in the sewage water, surface water, groundwater, drinking water

and sludge have been reported (Ku¨mmerer, 2009a, b; Le-Minh et al., 2010). As persistent pollutants, sulfonamides residues in the aquatic environment have become one of the most concerned issues with regard to public health. They exhibit potential toxicity to human beings and aquatic organisms, and are responsible for the emergence of antibiotic resistant bacteria and genes (Daughton and Ternes, 1999; Ku¨mmerer, 2009b; Munir et al., 2011). Recently, heterogeneous photocatalysis for sulfonamides degradation using TiO2 has become a potentially costeffective and environmentally sustainable treatment alternative (Abella´n et al., 2009; Baran et al., 2009; Beltra´n et al.,

* Corresponding author. School of Civil and Environmental Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798, Singapore. Tel.: þ65 6790 6933; fax: þ65 6791 0676. E-mail address: [email protected] (T.-T. Lim). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.002

5016

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 1 5 e5 0 2 6

2009; Hu et al., 2007; Le-Minh et al., 2010). TiO2 is the most suitable photocatalyst for applications in water treatment and reclamation due to its high photocatalytic activity, good chemical and biological stability, high energy efficiency, relatively low cost and non-toxicity (Fox and Dulay, 1993). Hydroxyl radicals (OH) formed on the TiO2 surface can completely mineralize a variety of pollutants even without chemical additives. However, due to its large band gap, TiO2 can be photoexcited only under UV irradiation (l < 387 nm) which makes up only 3e5% of the solar spectrum. Great efforts have been made to develop visible-light responsive TiO2 to harness solar energy for photocatalysis. Doping TiO2 with non-metals (C, N, S, F, B) (Asahi et al., 2001; Go´rska et al., 2008; Li and Shang, 2010; Rengifo-Herrera and Pulgarin, 2010; Subagio et al., 2010; Zaleska et al., 2008) is one of the most efficient methods to shift the spectral response of TiO2 from UV to visible-light region by narrowing its band gap. Visible-light photocatalytic activity could be also enhanced by carbon-sensitized TiO2, for which the carbon could be from titanium alkoxide and mainly exist as the CeC bonds (carbonaceous species), and CeO and O¼CeO bonds (carbonate species). Lettmann et al. (2001) prepared coke-containing TiO2 by a modified solegel process using different alkoxide precursors in the absence of any dopant. They stated that the highly condensed carbonaceous species embedded in the TiO2 matrix could be responsible for the photosensitization. Chou et al. (2006) and Go´rska et al. (2008) prepared visible-lightresponsive carbon-containing TiO2 using a sol-gel method, respectively. Both of them concluded that the incorporation of carbonaceous species in highly condensed and coke-like structure could play the role of sensitizer with an extended absorption band tail in the visible-light region. To our knowledge, there is no report about the synthesis and application of TiO2 with synergistic effects of both carbonsensitizing and nitrogen-doping. The objective of this study was to synthesize novel carbonsensitized and nitrogen-doped TiO2 (C/NeTiO2) using a facile solegel method under controlled calcination temperature. Sulfanilamide (SNM) was used as the target pollutant, and its chemical structure is shown as follows.

The effect of calcination temperature for C/NeTiO2 synthesis on SNM degradation and mineralization were investigated. The visible-light-emitting diodes (vis-LEDs) that emit white, blue, green and yellow lights were used as visible-light sources to investigate the photo-response of C/NeTiO2 in the visible-light region. Since inorganic anions are prevalent in aqueous phase, it is important to investigate their effect on the photocatalytic removal of organic pollutants. In this study, from a practical application point of view, the effects of pH and inorganic anions on the photocatalytic degradation and mineralization of SNM were investigated. The organic intermediates and end-products formed were monitored. Deactivation test was conducted to investigate the photochemical

stability of the photocatalyst. Microtox assay was carried out to evaluate the evolution of acute toxicity of SNM solution during photocatalysis.

2.

Experimental

2.1.

Synthesis of C/NeTiO2

All chemicals were used as received without further purification. Millipore Co. MilliQ (MQ) water with resistivity of 18.2 MU cm was used throughout the study unless otherwise stated. Titanium butoxide (Ti(OC4H9)4, 97%, SigmaeAldrich) was used as both titanium precursor and carbon source. Nitric acid (70%, Labscan) was used to catalyze the hydrolysis and condensation reactions and as nitrogen source. In a typical synthesis, 10 mL titanium butoxide was added dropwise into 50 mL absolute ethanol (99.9%, Merck) under magnetic stirring for 30 min at ambient temperature (26  1
C). Then a mixture of 3 mL MQ water, 3 mL absolute ethanol and 1 mL nitric acid was added dropwise into the above solution, followed by stirring for 6 h and aging overnight to obtain the gel. The gel was dried at 80
C for about 10 h in an oven, ground into fine powder, and calcinated at 200, 250, 300, 350 and 400
C for 2 h in a muffle furnace under air atmosphere to obtain the final products, which are denoted as T200, T250, T300, T350 and T400, respectively. Under these controlled calcination temperatures, different contents of carbonaceous species would be incorporated in the as-synthesized C/NeTiO2.

2.2.

Characterization of C/NeTiO2

X-ray diffraction (XRD) patterns were obtained from a Bruker D8 ADVANCE X-ray Diffractometer with Cu Ka radiation ˚ ) in a 2q range of 5e80
. The crystallite size and (l ¼ 1.5418 A phase content were obtained from TOPAS 2.0 software. The specific surface area was calculated using the BrunauerEmmett-Teller (BET) equation with a Quantachrome Autosorb1 instrument at 77 K. The particle morphology was observed with a JEOL-2100F transmission electron microscopy (TEM) at an accelerating voltage of 200 kV. The surface element content was collected using JEOL JSM6360 scanning electron microscope (SEM) equipped with energy dispersive X-ray spectroscopy (EDX). The surface chemical compositions and bonding states of the photocatalysts were probed by X-ray photoelectron spectroscopy (XPS) analysis on a KratosAxis Ultra spectrometer using a monochromatized Al Ka (1486.71 eV) X-ray source. All the spectra were calibrated with the adventitious carbon at 284.8 eV. The UVevis diffuse reflectance spectra (DRS) were obtained using a Perkin Elmer Lambda 35 UVevis spectrophotometer equipped with an integrating sphere assembly with BaSO4 as the reflectance standard.

2.3. Vis-LED photoreactor and evaluation of photocatalytic activity The vis-LED photoreactor used in this study has been reported in our previous paper (Wang and Lim, 2010). In this study, 500 mL borosilicate glass reactor with diameter of 80 mm and height of 110 mm was wrapped with a 2 m long LED flexible strip (SMD

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 1 5 e5 0 2 6

5050, 15W) comprising 60 LED units (beam angle 120
). Four different LED strips were used, emitting white light (LED-W, main emission wavelength l ¼ 450 nm), blue light (LED-B, l ¼ 465 nm), green light (LED-G, l ¼ 523 nm) and yellow light (LED-Y, l ¼ 589 nm), respectively. The visible-light and UV intensities of the LEDs were ca. 75 W m2 and <0.7 W m2, respectively, as measured by a digital power meter SP 1065 (Janco Impex) and an Accumax XRP-3000 radiometer. The changes of reactor temperature are insignificant throughout the experiment. A mirror cover was used to minimize light penetration from surrounding and water loss due to evaporation. The photocatalytic degradation of SNM was systematically evaluated with the C/NeTiO2 compared with the commercial TiO2 Degussa P25 (BET area ¼ 50  15 m2 g1 and average primary particle size ¼ 21 nm) at ambient temperature. Before turning on the LED-W, 400 mL suspension with photocatalyst dosage of 1.0 g L1 and initial SNM concentration of 5.0 mg L1 in MQ water was stirred in dark to reach adsorption equilibrium within 1 h. During photocatalysis, sample aliquots were collected at appropriate time intervals and filtered using 0.45 mm cellulose acetate syringe membrane filters for high performance liquid chromatography (HPLC) analysis. Changes in total organic carbon (TOC) were measured with a Shimadzu ASI-V TOC Analyzer. The effect of solution pH was investigated through pH adjustment using 1 M HCl and 1 M NaOH solutions. To investigate the influence of inorganic anions, NaCl, NaNO3, Na2SO4, NaHCO3, NaH2PO4 2H2O and Na2SiO3 9H2O were introduced into the SNM solution to obtain a concentration of 1 mM, respectively. The adsorption of SNM on T300 in the presence of these inorganic anions was also carried out.

2.4.

Analytical method

The HPLC analysis was performed with a Spherisorb ODS-2 column (150 mm  4.6 mm I.D., 5 mm) and an SPD-M20A diode array detector (DAD) using a Shimadzu LC system (LC20AD). Methanol (HPLC-grade, Merck) and MQ water was used as mobile phase A and B (vA/vB ¼ 40/60), respectively, at a flow rate of 0.5 mL min1. The oven temperature was maintained at 30
C. The detector wavelength was set at 260 nm and injection volume was 20 mL. Carboxylates (acetate, oxalate, formate and propionate) and inorganic anions (sulfate, nitrate and nitrite) were monitored by a Dionex ICS-1000 ion chromatography (IC) system equipped with a conductivity detector. An IonPac AS15 anionexchange column (4  250 mm) was employed with an elution containing 36 mM KOH at a flow rate of 1.2 mL min1. Ammonium was determined using the colorimetric nesslerization method (HACH Method 8038) with a HACH DR/2400 spectrophotometer. Zeta potential data were obtained using a Malvern NanoZS Zetasizer with T300 concentration of 0.3 g L1 in suspension. The pH value of the suspension was adjusted using 1 M HCl and 1 M NaOH solutions.

2.5.

Photocatalytic degradation kinetics

LangmuireHinshelwood model as defined by Eq. (1) has been widely applied for analysis of heterogeneous photocatalytic

5017

degradation kinetics of pollutants in the aqueous phase (Herrmann, 2005): dC kKC r0 ¼  ¼ dt 1 þ KC

(1)

where r0 is the initial reaction rate (mg L1 min1), C is the concentration of pollutants (mg L1), t is the reaction time (min), k is the LangmuireHinshelwood reaction rate constant (mg L1 min1), and K is the Langmuir adsorption equilibrium constant (L mg1). At a dilute concentration of pollutants (i.e., KC << 1), pseudo-first-order kinetics model can be assumed as shown in Eqs. (2) and (3): dC r0 ¼  ¼ kKC dt

(2)

  C ¼ kKt ¼ kapp t ln C0

(3)

where kapp is the apparent rate constant (min1), and C0 is the initial concentration of pollutants (mg L1).

2.6.

Initial quantum yield

Quantum yield is defined as the ratio of the reaction rate in molecules transformed to the efficient photonic flux in photons received (Herrmann, 1995). However, since the reaction rate is a function of the substrate concentration which may vary with light irradiation time, quantum yield appears to be a kinetic parameter which is time-dependent. The best estimation of quantum yield is given by the initial value as expressed in Eq. (4) (Chatzitakis et al., 2008): 40 ð%Þ ¼

r0  100 1000MwIA

(4)

where 40 is initial quantum yield, r0 can be calculated from kapp and C0 (r0 ¼ kappC0, mg L1 min1), Mw is the molecular weight of SNM (172.20 g mol1), IA is the irradiation intensity absorbed by the photocatalyst which is equal to the incident photon flux intensity (mol photons L1 min1). According to the quantum theory, IA can be calculated by Eqs. (5)e(7): IA ¼

MQ Vt

(5)

M¼

l hc  6:02  1023

(6)

Q ¼ 60  ISt

(7)

where M is the mole photons needed to derive 1 J of energy (mol photons J1), Q is the energy produced by light during reaction (J), V is the volume of the treated water in the reactor (L), t is the reaction time (min), l is the wavelength of the incident light (m), h is Plank’s constant (6.626  1034 J s1), c is the light velocity (3.0  108 m s1), 6.02  1023 is Avagadro’s number, I is the light intensity (W m2), S is the irradiation area (m2).

2.7.

Acute toxicity test

The acute toxicity of SNM solution during photocatalysis was examined with Vibrio fischeri (NRRL number B-11177) using

5018

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 1 5 e5 0 2 6

a Microtox Model 500 Analyzer (Azur Environment, Workingham, England) within 24 h after irradiation. V. fischeri was obtained as a freeze-dried reagent and stored at 20
C. Prior to the test, the freeze-dried V. fischeri was activated by hydration with a sterilized solution of sodium chloride. The sample was filtered using 0.45 mm cellulose acetate syringe membrane filter. The sample pH was adjusted to between 6 and 8 using NaOH or HCl and the sample osmolality was adjusted to 2% by adding high purity sodium chloride. V. fischeri emits light as a by-product of their cellular respiration and metabolic processes. Decrease in the respiration rate, corresponding to the inhibition of bioluminescence, will be caused by the presence of toxic compounds. The bioluminescence of V. fischeri was recorded after 5, 15 and 30 minincubation time at 15
C. The EC50 values (%v/v, the percentage of sample dilution that causes a 50% reduction in bioluminescence of V. fischeri) were calculated with MicrotoxOmni software.

3.

Results and discussion

3.1.

Characteristics of C/NeTiO2

3.1.1.

Structure and textural properties

For all the C/NeTiO2, obvious XRD diffraction peaks associated with the anatase phase appeared as shown in Fig. 1. Although there were no obvious XRD diffraction peaks for both brookite and rutile phase, the brookite phase was present in the C/NeTiO2 according to the phase content analysis with TOPAS 2.0 software, while the rutile phase did not exit. According to Carp et al. (2004), the calcination temperature in this study was not high enough for the transformation of TiO2 from the anatase to rutile phase. The diffraction peaks became sharper with the calcination temperature increased from 200 to 400
C, indicating the TiO2 crystal growth which resulted in the crystallite size increased from ca. 8.3e13.5 nm and the specific surface area decreased from 188 to 109 m2 g1 (Table 1). Also the anatase phase content increased with the calcination temperature, suggesting the crystallinity enhancement.

Fig. 1 e XRD patterns of the C/N TiO2 and P25.

Fig. 2a shows that the T300 particle sizes were ca. 10 nm which is in agreement with the XRD results. The indices of the spotted rings shown in the selected area electron diffraction (SAED) pattern of T300 validated the anatase tetragonal phase of TiO2. Fig. 2b presents the HRTEM image of T300 to further confirm the anatase-crystal nature. The lattice spacing (d) was about 0.35 nm between adjacent lattice planes, matching well with the distance between the (1 0 1) crystal planes of anatase TiO2.

3.1.2.

EDX analysis

Table 1 shows the result of EDX elemental microanalysis as obtained from the EDX spectra and elemental mappings. According to the EDX spectra, the C/NeTiO2 was mainly composed of C, Ti and O while the characteristic peak for nitrogen was not evidenced. However, nitrogen was detected to be present rather uniformly in the EDX elemental mappings of the C/NeTiO2 (See Supplementary Information for the case of T300). It is noted that the carbon content on the surface of the C/NeTiO2 decreased from 10 to 3% when the calcination temperature increased from 200 to 400
C, due to the increasing decomposition of carbon from the organic precursor. Correspondingly, the color of the C/NeTiO2 appeared to change from yellow to brown when the calcination temperature increased from 200 to 300
C, and changed to grayish and white when calcinated at 350
C and 400
C, respectively.

3.1.3.

XPS analysis

XPS spectra reveal the presence of C, N, O and Ti on the T300 surface (Fig. 3). Fig. 3a shows that the C 1s peaks were observed at 284.8, 286.1 and 288.9 eV with atomic percentages (at%, estimated from the relative area intensities of peaks) of 20.26, 5.68 and 2.05, respectively. The C1s peak at 284.8 eV is usually ascribed to the adventitious elemental carbon or carbon residues from the organic precursor (mainly CeC bonds) (Gu et al., 2008; Yang et al., 2008a). The C 1s peaks at 286.1 and 288.9 eV are assigned to the carbonate species, such as CeO, C]O, O]CeO and CeN bonds (Yang et al., 2008a). No peak was found around 282 eV, thus the possibility of carbon substitutional doping in the form of TieC bonds cannot be confirmed (Gu et al., 2008; Yang et al., 2008a). These carbonaceous species could act as photosensitizer to induce the visible-light absorption and response (Chou et al., 2006; Go´rska et al., 2008; Lettmann et al., 2001), which is responsible for the enhancement of visible-light photocatalytic activity. In addition, they could narrow the band gap by creating interface states (Treschev et al., 2008). Three peaks were observed from the N1s XPS spectra (Fig. 3b). The peak at 396.5 (0.05 at%) is ascribed to the TieN bonds resulted from the nitrogen substitutional incorporation into the TiO2 lattice (Popa et al., 2010; Yang et al., 2008a). The peak at 399.5 (0.53 at%) is assigned to the molecularly chemisorbed g-N2 (NeH, NeN, NeO, NeC bonds), such as TieNeO and TieOeN sites (Wang and Lim, 2010; Wei et al., 2008; Yang et al., 2008a). The peak at 402.1 eV (0.03 at%) is attributed to the molecularly adsorbed nitrogen species (i.e. NOx or NHx) (Yang et al., 2008a). These three peaks indicated that nitrogen was interstitially and substitutionally doped into the TiO2 lattices. Nitrogen doping is related to the band gap narrowing with red

5019

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 1 5 e5 0 2 6

Table 1 e Physicochemical properties of the C/N TiO2. TiO2

Crystallite sizea (nm)

Anatase contenta (%)

SSAb (m2 g1)

Surface element contentc (wt%) N

T200 T250 T300 T350 T400 a b c d

8.3 8.4 10.1 11.0 13.5

75 77 79 90 95

188 183 160 138 109

BLQ BLQ BLQ BLQ BLQ

d

C

Ti

O

9.76 7.33 5.42 4.72 3.19

40.36 41.84 51.95 46.03 54.99

49.88 50.83 42.63 49.25 41.82

Derived from XRD patterns. Specific surface area, calculated using the BET equation. Determined via EDX analysis. BLQ, below the limit of quantification.

shift of absorption edge to the visible-light region (Asahi et al., 2001; Wang and Lim, 2010), leading to the enhanced visiblelight photocatalytic activity. Fig. 3c shows one peak at 529.2 eV (41.34 at%) for TieO bonds and another at 530.9 eV (10.00 at%) for the oxygen atoms in the surface hydroxyl groups (HeO bonds) and/or in the carboxyl groups (CeO bonds) (Go´rska et al., 2008; Xu et al., 2010). Ti4þ oxidation state was confirmed by peaks at 458.2 (13.67 at%) and 463.9 eV (6.39 at%) for Ti 2p3/2 and Ti 2p1/2, respectively (Yang et al., 2008a) (Fig. 3d). No peak for Ti3þ was observed which is in agreement with the study by Go´rska et al. (2008) who found that TiO2 mainly contained Ti4þ at low calcination temperature. Ti 2p peak at around 455 eV was not observed to further indicate the absence of the TieC bonds (Gu et al., 2008).

3.1.4.

Optical properties

Fig. 4 indicates that the C/NeTiO2 exhibited remarkable red shift of absorption edge with long band tailings to the visiblelight region. The red shift phenomenon could be attributed to the interface states created by the carbonaceous species (Treschev et al., 2008), and/or the nitrogen incorporated into the TiO2 lattice through the mixing of N 2p states with O 2p states (Asahi et al., 2001; Wang and Lim, 2010). The extended band tailing in the visible-light region could originate from the carbonaceous species working as photosensitizer (Chou et al., 2006; Go´rska et al., 2008; Lettmann et al., 2001).

3.2.

Photocatalytic degradation

3.2.1. Effect of calcination temperature for C/NeTiO2 synthesis In dark without photocatalyst added, SNM concentration remained unchanged (5.0 mg L1) over 5 h, indicating that hydrolysis of SNM could be negligible. The kapp and 40 values under various experimental conditions are summarized in Table 2. Fig. 5a shows that <1% of SNM was photodegraded via photolysis (without photocatalyst) after 5 h of LED-W irradiation. The adsorption of SNM on the C/NeTiO2 could be negligible. The photocatalysts exhibited SNM photocatalytic degradation efficiencies in the order of T300 (95%) > T350 (65%) > T250 (46%) > T200 (29%) > T400 (14%) > P25 (5%) with the 40 decreased from 2.70 to 0.050%. P25 exhibited lower photocatalytic activity which can be validated by the result of Fig. 4 that the absorbance intensity of P25 in the visible-light region was lower than that of T300. In addition, compared to P25 (the pure TiO2), T300 was a carbon-sensitized and nitrogen-doped TiO2 as shown in Fig. 3. It is postulated that T300 will possess more visible-light photocatalytic activity than that of P25. The TOC removal (maximum at 70% for T300) was less than the photocatalytic degradation efficiency under the same experimental condition, indicating the incomplete mineralization of SNM along with the formation of organic intermediates.

Fig. 2 e (a) TEM and (b) HRTEM imagies of T300. Inset shows SAED pattern of T300.

5020

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 1 5 e5 0 2 6

Fig. 3 e XPS spectra of (a) C 1s, (b) N 1s, (c) O 1s and (d) Ti 2p region for T300.

The calcination temperature has a great effect on the crystal phase structure, incorporation of nitrogen into the TiO2 lattice and content of carbonaceous species, which is related to the visible-light photocatalytic activity. In this study, the optimal calcination temperature for photocatalytic degradation of SNM was 300
C (T300). T300 mainly contained anatase phase which is photocatalytically more active than rutile and brookite phase. In addition, a fraction of amorphous

Fig. 4 e DRS of the C/N TiO2 and P25.

phases were present in T300. In the study by Lettmann et al. (2001), they found that the amorphous phase may also carry photocatalytic activity. The nitrogen incorporated into the TiO2 lattice could result in the band gap narrowing and red shift of absorption edge to the visible-light region (Asahi et al., 2001; Wang and Lim, 2010). There are three possible mechanisms for the enhancement of photocatalytic activity by the carbonaceous species incorporated in T300. Mechanism 1. They could serve as photosensitizer to extend the absorption band tailing to the visible-light region (Chou et al., 2006; Go´rska et al., 2008; Lettmann et al., 2001). Moreover, the excited photosensitizer could inject an electron into the conduction band of T300, then the electron is transferred to the molecular oxygen adsorbed on the T300 surface, producing  O 2 and OH. Both O2 and OH are capable of degrading SNM (Dong et al., 2008). In order to investigate the role of carbonaceous species, benzoquinone (BQ) was used as an effective quencher of O 2 (Raja et al., 2005). As shown in Fig. 5b, the SNM removal efficiencies decreased with the increase of BQ concentrations in solution, reflecting that BQ prevented the degradation of SNM through consuming O 2 . This result confirmed the carbon photosensitizing effect that the carbonaceous species could promote the production of O 2. Mechanism 2. They could act as the lattice defect of TiO2 to form interface states between the anatase and brookite phases which could narrow the band gap effectively (Chou et al., 2006; Treschev et al., 2008).

5021

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 1 5 e5 0 2 6

Table 2 e kapp, r0 and 40 for SNM photocatalytic degradation under various experimental conditions. Anion

kapp (min1),  103

r0 (mg L1 min1),  103

40 (%)

Neutral

e

LED-W

Neutral

e

LED-W

Neutral

e

Neutral

e

1.0

LED-W LED-B LED-G LED-Y LED-W

e

1.0

LED-W

3.0  0.1 5.0  0.1 7.0  0.1 9.0  0.1 11.0  0.1 6.9  0.1

0.012 (0.821a) 0.202 (0.984) 2.27 (0.989) 1.34 (0.970) 10.9 (0.997) 4.15 (0.963) 0.614 (0.957) 10.9 (0.997) 6.70 (0.997) 4.10 (0.999) 2.86 (0.997) 2.06 (0.992) 1.02 (0.998) 2.41 (0.998) 5.68 (0.999) 10.9 (0.997) 13.3 (0.999) 14.7 (0.998) 10.9 (0.997) 6.80 (0.984) 2.18 (0.951) 0.677 (0.825) 14.2 (0.995) 11.7 (0.995) 7.36 (0.982) 6.73 (0.980) 6.19 (0.978) 9.94 (0.994) 9.31 (0.989) 10.3 (0.989) 5.78 (0.970) 7.13 (0.981) 4.40 (0.962)

0.06 1.01 11.4 6.68 54.3 20.7 3.07 54.3 67.0 81.9 85.8 103 5.08 12.0 28.4 54.3 66.4 73.4 54.3 34.0 10.9 3.38 71.2 58.6 36.8 33.6 31.0 49.7 46.5 51.3 28.9 35.7 22.0

0.003 0.050 0.564 0.332 2.70 1.03 0.152 2.70 3.33 4.07 4.27 5.13 0.252 0.598 1.41 2.70 3.30 3.64 2.70 1.63 0.466 0.128 3.54 2.91 1.83 1.67 1.54 2.47 2.31 2.55 1.44 1.77 1.09

C0 (mg L1)

TiO2 dosage (g L1)

Visible light

5.0

1.0

LED-W

1.0

T300

5.0 10 20 30 50 5.0

T300

5.0

0.10 0.25 0.50 1.0 1.5 2.0 1.0

T300

5.0

T300

5.0

TiO2 Photolysis P25 T200 T250 T300 T350 T400 T300

pH

Cl (1 mM) NO 3 (1 mM) SO2 4 (1 mM) HCO 3 (1 mM) H2PO 4 (1 mM) Silica (1 mM)

a R2, correlation of determination.

Mechanism 3. They could also play a role of electron scavenger to inhibit the electron-hole recombination (Dong et al., 2008; Lettmann et al., 2001). However, although T200 and T250 contained more carbon than T300, the photocatalytic activities of T200 and T250 were lower than that of T300. This indicated that there is an optimal content of carbonaceous species for highest visible-light photocatalytic activity. The excessive carbonaceous species on the T200 and T250 surface could absorb the visible light, thus reduced the visible-light photons available to generate electron-hole pairs, and consequently lowered their photocatalytic activity. T300 possessed the optimal content of carbonaceous species in this study. Therefore, further evaluations of photocatalytic activity as discussed in the following sections were focused on T300.

Another possible reason is that the organic intermediates and end products formed during photocatalysis might compete with the SNM molecules for the limited adsorption and photocatalytic sites on the T300 surface, causing inhibition of the SNM photocatalytic degradation to a certain degree. Additionally, LangmuireHinshelwood model was used for the degradation kinetics analysis. The linear plot of LangmuireHinshelwood model is shown in the inset of Fig. 6(a1). The LangmuireHinshelwood reaction rate constant and Langmuir adsorption equilibrium constant are 0.103 mg L1 min1 and 0.215 L mg1 (R2 ¼ 0.981), respectively. Fig. 6(a2) shows that the time-dependent TOC removal agreed with that of the SNM degradation. TOC removal decreased from 70 to 30% after 5 h of LED-W irradiation when the initial SNM concentration increased from 5.0 to 50 mg L1.

3.2.2.

3.2.3.

Effect of initial SNM concentration

Fig. 6(a1) illustrates the effect of initial SNM concentration on the SNM photocatalytic degradation kinetic. The kapp values decreased from 10.9 to 2.06  103 min1 and 40 values increased from 2.70 to 5.13% as the initial SNM concentration increased from 5.0 to 50 mg L1 (Table 2). It is presumed that a fraction of light might be absorbed by the SNM molecules rather than T300 for solution with high SNM concentration.

Effect of T300 dosage

It can be seen from Fig. 6(b1) that the photocatalytic degradation efficiencies increased with the T300 dosage up to 1.0 g L1 with kapp values of 10.9  103 min1 (Table 2). Increase of the T300 dosage from 0.10 to 1.0 g L1 led to a tenfold increase of 40 values from 0.252 to 2.70%, while a higher dosage resulted in slight increase of 40 values. With the increase of T300 dosage, the photocatalyst surface area

5022

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 1 5 e5 0 2 6

photosensitizing effect for SNM degradation and mineralization was more significant under LED-W and -B irradiations. However, a fraction of SNM was still removed under LED-Y irradiation, showing that T300 could be photoexcited by visible-light even with wavelength at around 600 nm. This confirmed the remarkable red shift of absorption edge of T300 as shown in Fig. 4.

3.2.5.

Fig. 5 e Photocatalytic degradation and mineralization of SNM using the C/N TiO2 and P25 (a) without and (b) with BQ in solution (C0: 5.0 mg LL1; photocatalyst dosage: 1.0 g LL1; irradiation light: LED-W; pH: neutral).

available for adsorption and photocatalytic degradation was increased, while the solution opacity was also increased leading to a decrease in the penetration of the photon flux through the reactor. Moreover, the loss in the surface area by the photocatalyst agglomeration at higher T300 dosage was also observed. As a result, the optimum T300 dosage in this study was 1.0 g L1 Fig. 6(b2) also shows that the TOC removal only increased marginally as the T300 dosage increased from 1.0 to 2.0 g L1.

3.2.4.

Effect of type of vis-LED

It was found that <1% of SNM was photodegraded after 5 h of LED-W, B, G and Y irradiations via photolysis. As shown in Fig. 6(c1), the degradation efficiency of SNM was enhanced by T300, namely, 95%, 84%, 42% and 13% of SNM was degraded after 5 h of LED-W, B, G and Y irradiations with kapp values of 10.9, 6.80, 2.18 and 0.677  103 min1, respectively. The corresponding TOC removal agreed with that of the SNM degradation and was 70%, 61%, 34% and 9% (Fig. 6(c2)), respectively. Apparently, the SNM degradation and mineralization efficiencies were higher under LED-W and B irradiations than those under LED-G and Y irradiations. Meanwhile, the carbon

Effect of solution pH

Fig. 6(d1) demonstrates that the photocatalytic degradation of SNM was more favorable in acidic conditions. After 5 h of LEDW irradiation at pH 3 and 5, >99% and 97% of SNM was photocatalytic degraded, respectively, while the corresponding TOC removal was 76% and 74% (Fig. 6(d2)). With the pH increasing from 3 to 11, the kapp values decreased from 14.2 to 6.19  103 min1 (Table 2) and the 40 values decreased from 3.54 to 1.54%. As an amphoteric material, the TiO2 surface is positively charged at a pH lower than point of zero charge (pzc) and negatively charged at a pH higher than pzc. The pzc of T300 was determined to be at pH 5.5. Therefore, T300 was positively charged at pH  5, while negatively charged at pH  7. Additionally, SNM molecule possesses pKa values of 2.40 and 10.43 (Baran et al., 2009). Thus, SNM is mainly neutral at pH 3, 5, 7 and 9, while negatively charged at pH 11 (Garcı´a-Gala´n et al., 2009). The adsorption experiment confirmed that SNM adsorption decreased with increasing pH, resulting from the net electrostatic repulsion between T300 and SNM. However, it can be postulated that the adsorption effect only played a minor role in the enhancement of photocatalytic activity, as the adsorption of SNM on T300 within the pH range investigated was insignificant. The incorporation of carbonaceous species in T300 contributed significantly to the visible-light photocatalytic activity. Generally, OH can be produced according to Eq. (8) in both acidic and alkaline conditions. In the case of low pH, the mechanism for OH formation in the presence of molecular oxygen is shown in the Eqs. (9)e(12), while in the case of high pH, direct generation of OH radicals (Eq. (13)) is possible (Malato et al., 2009). þ

H2 O þ h /  OH þ Hþ

(8)

O2 þ e /  O 2

(9)

þ O 2 þ H /HO2 

(10)

2HO2  /H2 O2 þ O2

(11)

 H2 O2 þ O 2 /  OH þ OH þ O2

(12)

þ

OH þ h /  OH

(13)

As photosensitizer, the carbonaceous species could promote the formation of O 2 (Eq. (9)) in acidic conditions through the injection of an electron into the conduction band followed by the reaction between the electron and molecular oxygen adsorbed on the T300 surface. Subsequently, both the formed O 2 and OH could be responsible for the photocatalytic degradation of SNM. The photocatalytic degradation

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 1 5 e5 0 2 6

5023

Fig. 6 e Effect of (a) initial SNM concentration (Inset shows linear plot fitting of Langmiur-Hinshelwood model), (b) T300 dosage, (c) Vis-LEDs, (d) solution pH and (e) inorganic anions on (1) photocatalytic degradation of SNM and (2) TOC removal.

5024

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 1 5 e5 0 2 6

of SNM still occurred at pH > pKa2 due to the OH originated from the reaction between OH/H2O and photogenerated holes as shown in the Eq. (13). Apparently, due to the carbon photosensitizing effect during which both OH and O 2 were formed as shown in Eqs. (9)e(12), the photocatalytic degradation efficiency of SNM was higher in acidic conditions than that in alkaline conditions.

3.2.6.

Effect of inorganic anions

The influence of the inorganic anions on the adsorption of SNM on T300 was negligible. As shown in Table 2 and Fig. 6(e1), the inorganic anions generally did not significantly inhibit the SNM photocatalytic degradation. The order of inhibitory effects induced by the anions follows silica > bicarbonate > phosphate > chloride, nitrate and sulfate. The TOC removal trend agreed with that of the SNM degradation (Fig. 6(e2)). It has been reported that anions, such as chloride, nitrate, sulfate, bicarbonate and phosphate could act as hþ and OH scavenger to form ionic radicals (Eqs. (14) and (15)) (Rinco´n and Pulgarin, 2004), which are less reactive than hþ and OH. Therefore, the presence of anions would inevitably inhibit the SNM mineralization to a certain extent.   þ  2   h þ Cl ; NO 3 ; SO4 ; HCO3 ; H2 PO4 / Cl; NO3 ; SO 4 ; HCO3 ; H2 PO4 

3.4.


   2   OHþ Cl ; NO 3 ; SO4 ; HCO3 ; H2 PO4
/ Cl  NO3 ; SO-4 ; HCO3 ; H2 PO4  þ OH

(14)

(15)

Silica showed a slightly more pronounced inhibitory effect for the SNM photocatalytic degradation. It is reported that silica can undergo polymerization at relatively low concentration, and polymeric or colloidal silica may be present through the condensation of monomers (Kohn et al., 2003; Yang et al., 2008b). As a result, the active sites of photocatalysts might be occupied by polymeric and/or colloidal silica surrounding the photocatalysts. In addition, the light could be intercepted by polymeric and/or colloidal silica, leading to the light screening effect. Therefore, the SNM photocatalytic degradation could be inhibited to some degree.

3.3.

Fig. 7 e Evolution of acetate, ammonium, sulfate and organic intermediate during photocatalysis using T300 (C0: 5.0 mg LL1; T300 dosage: 1.0 g LL1; irradiation light: LEDW; pH: neutral).

Deactivation test

To investigate the possibility of the leaching of carbonaceous species and nitrogen dopant from T300 and the poisoning of T300 by the organic intermediates and end-products, especially sulfate, deactivation test was carried out as follows. After 5 h of LED-W irradiation, the T300 solution was centrifuged, washed with MQ water for 3 times and collected for reuse. The result shows that 88% of SNM degradation and 57% of TOC removal were still achieved after five recycles (Fig. 8), indicating that the activity of T300 only decreased by less than 10%. This suggested that both the leaching of carbonaceous species and nitrogen dopant and the poisoning of T300 could be negligible during its 5 times of reuse for the SNM photocatalytic degradation. Apparently, T300 exhibited good photochemical stability during the photocatalytic degradation of SNM under vis-LED irradiation.

Evolution of carboxylates and inorganic anions

High degree of mineralization of a variety of organic pollutants can be achieved with photocatalysis due to the non-selectivity of OH. However, the complete mineralization for most antibiotics is difficult due to their structural stability (Westerhoff et al., 2005). In this study, a new chromatographic peak appeared during the HPLC analysis of the SNM solution after irradiation. The peak area for the organic intermediate increased first and then decreased during the 5 h of LED-W irradiation (Fig. 7). On the other hand, a slow increase of acetate concentration with irradiation time was observed. Nitrate was detected but its concentration was below the limit of quantification, while nitrite was not detected in the solution. About 86% of nitrogen was released as ammonium and about 52% of sulfur was released as sulfate after 5 h of LED-W irradiation. In addition, the concentration of both ammonium and sulfate remained almost constant after a rapid release within 1 h.

Fig. 8 e Deactivation of T300 during the photocatalytic degradation of SNM (C0: 5.0 mg LL1; T300 dosage: 1.0 g LL1; irradiation light: LED-W; pH: neutral). Inset shows the corresponding TOC removal after 5 h of irradiation.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 1 5 e5 0 2 6

5025

photochemical stability and its absorption onset could extend to about 600 nm. Due to the synergistic effects of carbonsensitizing and nitrogen-doping, T300 showed the highest photocatalytic activity for SNM degradation under LED-W irradiation. The acidic conditions were favorable for the SNM photocatalytic degradation. The presence of chloride, nitrate, and sulfate appeared to inhibit the SNM degradation insignificantly, while different degrees of inhibitory effects were observed in the presence of bicarbonate, phosphate and silica. This work shows that C-sensitized and N-doped TiO2 has the potential application for the mineralization and detoxification of pollutants in water even when irradiated by indoor vis-LED light.

Acknowledgments Fig. 9 e Acute toxicity of SNM solution during photocatalysis using T300 (C0: 5.0 mg LL1; T300 dosage: 1.0 g LL1; irradiation light: LED-W; pH: neutral).

3.5.

Acute toxicity evaluation

The acute toxicity evaluation of antibiotics is essential due to the possible formation of more toxic organic intermediates and end-products during treatment. Fig. 9 demonstrates the acute toxicity (expressed as EC50, lower EC50 value indicates higher acute toxicity to V. fischeri) of SNM solution as a function of irradiation time during visible-light photocatalysis using T300. The acute toxicity trends of SNM solution were similar for the cases of 5, 15 and 30 min-incubation time. The original SNM solution before visible-light irradiation showed acute toxicity to V. fischeri with EC50 value of w23%v/v. The acute toxicity reached the maximum after 1 h of irradiation, and then it began to decrease slowly due to the further mineralization. Since the concentrations of ammonium and sulfate remained almost constant during the 5 h of irradiation, and the concentration of organic intermediates continued to decrease with time, it is concluded that ammonium and sulfate might be not toxic to V. fischeri while the acute toxicity might be induced by the organic intermediates. For the acetate, it could be used as the nutrient for V. fischeri. In addition, the synergistic effects among the species in the solution on the acute toxicity should also be considered. However, with further irradiation for more than 5 h, the acute toxicity of SNM solution continued to decrease. After 12 h of irradiation, the SNM solution exhibited much lower acute toxicity than that of the original SNM solution before irradiation. It indicates that photocatalysis using T300 is favorable for the detoxification of SNM solution.

4.

National Research Foundation (NRF), Singapore, is acknowledged for the financial support through the project EWI RFP 0802-11 which aims to develop novel photocatalysts for water purification and wastewater treatment using visible and solar light.

Conclusions

Under controlled calcination temperature, different contents of carbonaceous species from titanium butoxide could be incorporated in the C/NeTiO2. T300 exhibited good

Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.07.002.

references

Abella´n, M.N., Gime´nez, J., Esplugas, S., 2009. Photocatalytic degradation of antibiotics: the case of sulfamethoxazole and trimethoprim. Catalysis Today 144 (1e2), 131e136. Asahi, R., Morikawa, T., Ohwaki, T., et al., 2001. Visible-light photocatalysis in nitrogen-doped titanium oxides. Science 293 (5528), 269e271. Baran, W., Adamek, E., Sobczak, A., et al., 2009. Photocatalytic degradation of sulfa drugs with TiO2, Fe salts and TiO2/FeCl3 in aquatic environment - kinetics and degradation pathway. Applied Catalysis B: Environmental 90 (3e4), 516e525. Beltra´n, F.J., Aguinaco, A., Garcı´a-Araya, J.F., 2009. Mechanism and kinetics of sulfamethoxazole photocatalytic ozonation in water. Water Research 43 (5), 1359e1369. Carp, O., Huisman, C.L., Reller, A., 2004. Photoinduced reactivity of titanium dioxide. Progress in Solid State Chemistry 32 (1e2), 33e177. Chatzitakis, A., Berberidou, C., Paspaltsis, I., et al., 2008. Photocatalytic degradation and drug activity reduction of chloramphenicol. Water Research 42 (1e2), 386e394. Chou, P.W., Treschev, S., Chung, P.H., et al., 2006. Observation of carbon-containing nanostructured mixed titania phases for visible-light photocatalysts. Applied Physics Letters 89 (13), 131919e131921. Daughton, C.G., Ternes, T.A., 1999. Pharmaceuticals and personal care products in the environment: agents of subtle change? Environmental Health Perspectives 107 (Suppl. 6), 907e938. Dong, F., Zhao, W.R., Wu, Z.B., 2008. Characterization and photocatalytic activities of C, N and S co-doped TiO2 with 1D nanostructure prepared by the nano-confinement effect. Nanotechnology 19 (36), 365607e365617.

5026

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 1 5 e5 0 2 6

Fox, M.A., Dulay, M.T., 1993. Heterogeneous photocatalysis. Chemical Reviews 93 (1), 341e357. Washington, D. C. Go´rska, P., Zaleska, A., Kowalska, E., et al., 2008. TiO2 photoactivity in vis and UV light: the influence of calcination temperature and surface properties. Applied Catalysis B: Environmental 84 (3e4), 440e447. Garcı´a-Gala´n, M.J., Silvia Dı´az-Cruz, M., Barcelo´, D., 2009. Combining chemical analysis and ecotoxicity to determine environmental exposure and to assess risk from sulfonamides. TrAC Trends in Analytical Chemistry 28 (6), 804e819. Gu, D.-E., Lu, Y., Yang, B.-C., et al., 2008. Facile preparation of micro-mesoporous carbon-doped TiO2 photocatalysts with anatase crystalline walls under template-free condition. Chemical Communications 21, 2453e2455. Herrmann, J.M., 1995. Heterogeneous photocatalysis: an emerging discipline involving multiphase systems. Catalysis Today 24 (1e2), 157e164. Herrmann, J.M., 2005. Heterogeneous photocatalysis: state of the art and present applications. Topics in Catalysis 34 (1e4), 49e65. Hu, L., Flanders, P.M., Miller, P.L., et al., 2007. Oxidation of sulfamethoxazole and related antimicrobial agents by TiO2 photocatalysis. Water Research 41 (12), 2612e2626. Ku¨mmerer, K., 2009a. Antibiotics in the aquatic environment a review - Part I. Chemosphere 75 (4), 417e434. Ku¨mmerer, K., 2009b. Antibiotics in the aquatic environment a review - Part II. Chemosphere 75 (4), 435e441. Kohn, T., Kane, S.R., Fairbrother, D.H., et al., 2003. Investigation of the inhibitory effect of silica on the degradation of 1,1,1Trichloroethane by granular iron. Environmental Science and Technology 37 (24), 5806e5812. Le-Minh, N., Khan, S.J., Drewes, J.E., et al., 2010. Fate of antibiotics during municipal water recycling treatment processes. Water Research 44 (15), 4295e4323. Lettmann, C., Hildenbrand, K., Kisch, H., et al., 2001. Visible light photodegradation of 4-chlorophenol with a coke-containing titanium dioxide photocatalyst. Applied Catalysis B: Environmental 32 (4), 215e227. Li, Q., Shang, J.K., 2010. Composite photocatalyst of nitrogen and fluorine codoped titanium oxide nanotube arrays with dispersed palladium oxide nanoparticles for enhanced visible light photocatalytic performance. Environmental Science and Technology 44 (9), 3493e3499. Malato, S., Ferna´ndez-Iba´n˜ez, P., Maldonado, M.I., et al., 2009. Decontamination and disinfection of water by solar photocatalysis: recent overview and trends. Catalysis Today 147 (1), 1e59. Munir, M., Wong, K., Xagoraraki, I., 2011. Release of antibiotic resistant bacteria and genes in the effluent and biosolids of five wastewater utilities in Michigan. Water Research 45 (2), 681e693.

Popa, M., Macovei, D., Indrea, E., et al., 2010. Synthesis and structural characteristics of nitrogen doped TiO2 aerogels. Microporous and Mesoporous Materials 132 (1e2), 80e86. Raja, P., Bozzi, A., Mansilla, H., et al., 2005. Evidence for superoxide-radical anion, singlet oxygen and OH-radical intervention during the degradation of the lignin model compound (3-methoxy-4-hydroxyphenylmethylcarbinol). Journal of Photochemistry and Photobiology A: Chemistry 169 (3), 271e278. Rengifo-Herrera, J.A., Pulgarin, C., 2010. Photocatalytic activity of N, S co-doped and N-doped commercial anatase TiO2 powders towards phenol oxidation and E. coli inactivation under simulated solar light irradiation. Solar Energy 84 (1), 37e43. Rinco´n, A.-G., Pulgarin, C., 2004. Effect of pH, inorganic ions, organic matter and H2O2 on E. coli K12 photocatalytic inactivation by TiO2: implications in solar water disinfection. Applied Catalysis B: Environmental 51 (4), 283e302. Subagio, D.P., Srinivasan, M., Lim, M., et al., 2010. Photocatalytic degradation of bisphenol-A by nitrogen-doped TiO2 hollow sphere in a vis-LED photoreactor. Applied Catalysis B: Environmental 95 (3e4), 414e422. Treschev, S.Y., Chou, P.-W., Tseng, Y.-H., et al., 2008. Photoactivities of the visible-light-activated mixed-phase carbon-containing titanium dioxide: the effect of carbon incorporation. Applied Catalysis B: Environmental 79 (1), 8e16. Wang, X., Lim, T.-T., 2010. Solvothermal synthesis of CeN codoped TiO2 and photocatalytic evaluation for bisphenol A degradation using a visible-light irradiated LED photoreactor. Applied Catalysis B: Environmental 100 (1e2), 355e364. Wei, F., Ni, L., Cui, P., 2008. Preparation and characterization of NeS-codoped TiO2 photocatalyst and its photocatalytic activity. Journal of Hazardous Materials 156 (1e3), 135e140. Westerhoff, P., Yoon, Y., Snyder, S., et al., 2005. Fate of endocrinedisruptor, pharmaceutical, and personal care product chemicals during simulated drinking water treatment processes. Environmental Science and Technology 39 (17), 6649e6663. Xu, Q.-C., Wellia, D.V., Sk, M.A., et al., 2010. Transparent visible light activated CeNeF-codoped TiO2 films for self-cleaning applications. Journal of Photochemistry and Photobiology A: Chemistry 210 (2e3), 181e187. Yang, J., Bai, H., Jiang, Q., et al., 2008a. Visible-light photocatalysis in nitrogenecarbon-doped TiO2 films obtained by heating TiO2 gel-film in an ionized N2 gas. Thin Solid Films 516 (8), 1736e1742. Yang, X., Roonasi, P., Holmgren, A., 2008b. A study of sodium silicate in aqueous solution and sorbed by synthetic magnetite using in situ ATR-FTIR spectroscopy. Journal of Colloid and Interface Science 328 (1), 41e47. Zaleska, A., Sobczak, J.W., Grabowska, E., et al., 2008. Preparation and photocatalytic activity of boron-modified TiO2 under UV and visible light. Applied Catalysis B: Environmental 78 (1e2), 92e100.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 2 7 e5 0 3 7

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

A simple optode based method for imaging O2 distribution and dynamics in tap water biofilms M. Staal a,*, E.I. Prest a,b, J.S. Vrouwenvelder b,c, L.F. Rickelt a, M. Ku¨hl a,d a

Marine Biological Section, Department of Biology, University of Copenhagen, Strandpromenaden 5, DK-3000 Helsingør, Denmark Wetsus, Centre of Excellence for Sustainable Water Technology, P.O. Box 1113, 8900 CC Leeuwarden, The Netherlands c Delft University of Technology, Department of Biotechnology, Julianalaan 67, 2628 BC Delft, The Netherlands d Plant Functional Biology and Climate Change Cluster (C3), Department of Environmental Science, University of Technology Sydney, Broadway NSW 2007, Australia b

article info

abstract

Article history:

A ratiometric luminescence intensity imaging approach is presented, which enables spatial

Received 2 February 2011

O2 measurements in biofilm reactors with transparent planar O2 optodes. Optodes consist of

Received in revised form

an O2 sensitive luminescent dye immobilized in a 1e10 mm thick polymeric layer on

25 May 2011

a transparent carrier, e.g. a glass window. The method is based on sequential imaging of the

Accepted 3 July 2011

O2 dependent luminescence intensity, which are subsequently normalized with lumines-

Available online 13 July 2011

cent intensity images recorded under anoxic conditions. We present 2-dimensional O2 distribution images at the base of a tap water biofilm measured with the new ratiometric

Keywords:

method and compare the results with O2 distribution images obtained in the same biofilm

Oxygen sensing

reactor with luminescence lifetime imaging. Using conventional digital cameras, such

Planar optodes

simple normalized luminescence intensity imaging can yield images of 2-dimensional O2

Lifetime

distributions with a high signal-to-noise ratio and spatial resolution comparable or even

Imaging

surpassing those obtained with expensive and complex luminescence lifetime imaging

Membrane fouling simulator

systems. The method can be applied to biofilm growth incubators allowing intermittent

Biofilm

experimental shifts to anoxic conditions or in systems, in which the O2 concentration is depleted during incubation. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Biofilms are surface-associated microbial communities exhibiting spatio-temporal heterogeneity in their structure, composition, physiology and chemical microenvironment (Costerton et al., 1995; Stewart and Franklin, 2008). Such communities represent the preferred lifestyle of many microbes in natural ecosystems, and biofilms also play important roles in more applied contexts such as waste water

treatment and industrial processes (Nicolella et al., 2000), chronic infections (Costerton et al., 1999; Hall-Stoodley et al., 2004), and e.g. corrosion and biofouling of materials (Ridgway and Flemming, 1996; Vrouwenvelder et al., 2008). The growth dynamics and complex structural heterogeneity of biofilms has been studied in great detail, especially through application of various microscopic techniques (Neu et al., 2010). However, a similar detailed mapping of the chemical landscape and dynamics in biofilms is lacking in

Abbreviations: ROI, region of interest; MFS, membrane fouling simulator; mil, 1 mil equals 25.4 mm; Ru-dpp, Ruthenium(II)-tris-4,7diphenyl-1,10-phenanthroline; PS, polystyrene. * Corresponding author. Tel.: þ45 3532098; fax: þ45 35321951. E-mail address: [email protected] (M. Staal). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.007

5028

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 2 7 e5 0 3 7

most studies. Electrochemical and fiber-optic microsensors (Revsbech, 2005; Ku¨hl, 2005) can provide very detailed information on local chemical dynamics, zonations of microbial processes and mass transfer processes in biofilms (e.g. de Beer et al., 1994; Stoodley et al., 1994; Ku¨hl and Jørgensen, 1992; Ku¨hl et al., 1996), but such 1-dimensional characterization is in most cases inadequate to assess the true spatial distribution and dynamics of the chemical microenvironment in heterogeneous biofilms at similar resolution as biofilm structure can be resolved (Jørgensen and Des Marais, 1990). Modeling of biofilm systems has shown that significant differences can be found between 1- and 2-dimensional descriptions of biofilms, since 1-dimensional models fail to describe spatial heterogeneity other than over depth (z-plane). It has also been argued from biofilm modeling approaches that it is sufficient to do 2-dimensional measurements for a proper description of spatial heterogeneity in microenvironments, since increasing model complexity from 2- to 3dimensions did not yield much extra insights (Eberl et al., 2000). However, there is still a lack of real datasets linking spatial biofilm heterogeneity to chemical heterogeneities and microenvironments. With the development of imaging techniques, several new non-invasive methods became available, e.g. Confocal Laser Scanning Microscopy (Neu et al., 2010), Raman Microspectroscopy (Ivleva et al., 2010), lifetime imaging (Glud et al., 1996; Hidalgo et al., 2009), Magnetic Resonance Microscopy (Wagner et al., 2010). These methods allow measurements of 2and 3-dimensional chemical and biological landscapes in biofilms. Molecular oxygen (O2) is a key parameter in biogeochemical and biological studies (Glud, 2008; Fenchel and Finlay, 2008) and the introduction of planar optodes for mapping 2dimensional O2 concentration in natural systems (Glud et al., 1996) was a big step forward for the study of the heterogeneity of O2 distribution and dynamics in sediments and biofilms (e.g. Glud et al., 1998, 1999; Ku¨hl et al., 2007). Such planar optode measurements are based on luminescent indicator dyes, immobilized in a polymeric matrix and cast onto a transparent carrier. The measuring principle relies on the dynamic quenching of indicator luminescence by O2. Both the luminescence intensity (I) and luminescence lifetime (s) vary reversibly with O2 concentration, and the process does not consume O2. The ideal response of such optical O2 sensors is described by the SterneVolmer relation (Bacon and Demas, 1987): I s 1 I0 s0 5 ¼ ¼ 1 þ KSV ½O2  ¼ ¼ s I0 s0 1 þ KSV ½O2  I

(1)

where s0 and s denote the luminescence lifetime in the absence and in the presence of O2 respectively; I0 and I denote the luminescence intensity in the absence or presence of O2; KSV is the bimolecular quenching coefficient of the dye in its specific polymeric matrix, and [O2] is the O2 concentration. In practice, most planar optodes exhibit a non-ideal SterneVolmer like response, which can be modeled with a two-component model (Carraway et al., 1991), where only a certain fraction of the O2 indicator dye remains quenchable upon immobilization. This relationship can be described by the equation: I s 1a þa ¼ ¼ I0 s0 1 þ Ksv C

(2)

where a is the non-quenchable fraction. Initial applications of O2 planar optodes involved simple luminescence intensity measurements in combination with planar optodes with a black O2 permeable overcoat to avoid optical artifacts from background light and sample backscatter (Glud et al., 1996). Application of transparent O2 optodes on glass allows direct alignment of O2 distribution to the structure of the sample (e.g. Holst and Grunwald, 2001; Ku¨hl et al., 2007, 2008, Staal et al., 2011) and nowadays oxygen imaging with planar optods makes almost entirely use of luminescent lifetime imaging systems (Holst et al., 1998; Oguri et al., 2006). Such systems are mostly custom built and relatively expensive (>30.000 V), which has been a bottleneck for the more widespread distribution of planar O2 optode methodology. Recently, however, a new ratiometric method has been published using a normal digital single-lens reflex camera as detector (Wang et al., 2010; Larsen et al., in press). In this method, correction for an uneven light field was accomplished by inclusion of a luminescent O2 insensitive reference dye in the planar optode matrix. Both dyes are excited by the same excitation source, but the reference dye emits light in the green region, and the O2 sensitive dye in the red region, coinciding with the green and red channels of common color camera CCD or CMOS detectors. A ratio of the red and green channel indicates the O2 concentration. Here we present a simple luminescence intensity imaging approach enabling 2-dimensional mapping of biomass and O2 concentration in a biofilm growth incubator using simple digital color cameras or monochromatic cameras combined with suitable emission filters. The method uses the ratio of the luminescence from a transparent planar optode under anoxia to the luminescence under experimental O2 conditions to correct for heterogeneity in excitation light. No reference dye is required. As a proof of principle, a heterotrophic biofilm was grown on top of a transparent planar optode, which created heterogeneity in O2 concentrations. We show that such a simple imaging approach yields information on the 2dimensional distribution of O2 in biofilms with a good signalto-noise ratio and with a spatial resolution of 36 mm/pixel comparable or better than in more elaborate lifetime based O2 imaging.

2.

Methods

The planar O2 optode used in this study was based on the luminescent O2 sensitive dye Ruthenium(II)-tris-4,7-diphenyl1,10-phenanthroline (Ru-dpp) immobilized in a polystyrene (PS) matrix (20 mg Ru-dpp/g PS). Such Ru-dpp based planar optodes are suitable for O2 imaging over a wide dynamic range (up to full O2 saturation) (Ku¨hl and Polerecky, 2008). The PS/ Ru-dpp mixture was dissolved in chloroform (3.3 w/w%) and cast onto a silanized glass plate (Ku¨hl et al., 2007). After slow evaporation of the solvent in a semi-closed container, this resulted in a w9 mm thick homogeneous O2 sensor layer on the complete glass window of the incubator (6 cm  28 cm). The coated glass plate was mounted as a window in a watertight flow-trough biofilm growth incubator, i.e., a membrane fouling simulator (MFS), where a polypropylene spacer mesh was placed on top of the optode surface. The

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 2 7 e5 0 3 7

biofilm growth chamber had external dimensions of 0.07 m  0.30 m  0.04m (Vrouwenvelder et al., 2006, 2007). Coupons of feed spacer, membrane and product spacer were placed in the MFS resulting in the same spatial dimensions and orientation as in spiral wound membrane elements applied in water treatment. For MFS experiments, membranes and 787 mm (31 mil) thick feed spacer sheets were taken from a new, unused spiral wound reverse osmosis membrane element. The development of fouling, i.e. biofilm formation, was monitored by a rhodamine tracer dye imaging technique (see below) and by measuring the pressure drop increase over the feed spacer channel of the MFS. During operation, the MFS window was covered with a light-tight lid to prevent growth of phototrophic organisms. The reactor system setup consisted of a pressure reducing valve, manometer, dosage point (for biodegradable compounds), MFS and flow controller (Fig. 1, Vrouwenvelder et al., 2007). An extra T-connection was placed in the tubing before the inlet of the biofilm monitor to allow injection of saturated sodium dithionite solution to create anoxic conditions, as well as injection of water colored by the dye rhodamine WT (Chrompton & Knowles OT/US 04029NS). The dye Rhodamine WT is an inert, non-adsorbing and stable tracer for flow visualization and is simple to quantify by light absorbance (Huettel et al., 1996). The MFS was sterilized for a period of 5 min with 70% ethanol before the start of the experiment. Bacteria in the biofilm originated from bacteria occurring in the drinking water system. The MFS was operated at 16  C. A pressure of 120 kPa was applied to avoid degassing. The feed water flow was 16 L h1 equal to a linear flow velocity of 0.16 m s1, representative for the linear flow velocity in the lead membrane element in full-scale installations containing spiral wound nanofiltration or reverse osmosis membrane elements (Vrouwenvelder et al., 2009a). The MFSs were operated single pass without (partial) recirculation. Pressure drop measurements were performed with a pressure difference transducer (Deltabar S: PMD70-AAA7FKYAAA, Endress & Hauser, The Netherlands). The calibrated measuring range was 0e50 kPa (Vrouwenvelder et al., 2009b). Concentrated substrate was dosed into the feed water (tap water), prior to the MFS by a peristaltic pump (Masterflex L/S

5029

brushless digital drive, HV 7523.70, 1.6e100 rpm, with an Easy Load II pump rotor, Applicon Analytical, The Netherlands) at a flow of 0.03 L h1 from a 5L stock solution reservoir. The dosage of substrate was checked periodically by measuring the water volume pumped over a defined time interval. The chemicals NaCH3COO, NaNO3 and NaH2PO4 were dosed in a mass ratio C:N:P of 100:20:10 with a concentration of 1.0 mg acetate-C L1 in the feed water entering the MFS. The substrates were dissolved in MilliQ water. To restrict bacterial growth in the substrate dosage bottle, the pH was set at 11 by NaOH dosage. Stock solution bottles were replaced every 5 days. The substrate flow rate (0.03 L h1) was low compared to the feed water flow rate (16 L h1). Thus, the effect of the substrate flow on the pH of the feed water was insignificant.

2.1.

Imaging systems and image calculations

Two different camera systems were used for O2 measurements: (1) A monochrome 12 bit fast gate-able cooled 1280X1024 CCD chip camera (2/3" chip), denoted as the PCO camerain this manuscript (SENSICAM-SENSIMOD, PCO AG, Kehlheim, Germany) and controlled by custom made acquisition software (Look@MOLLI, Holst and Grunwald, 2001) that can perform intensity based as well as lifetime based O2 imaging (Holst et al., 1998); (2) A 8 bit color 1280X1024 CMOS chip camera (1/2" chip), denoted as the mEye camerain this manuscript (USB mEye SE,UI-1540-C, IDS Imaging Development Systems GMBH, Obersulm, Germany) and controlled by the manufacturers acquisition software. The mEye camera can only perform intensity based O2 imaging. Both image acquisition programs allowed manual programming of the exposure time of the camera chips. The gain per color channel could be set manually in the software for the mEye camera allowing separate signal optimization of the three color channels. For measurements with the mEye camera, the camera was coupled via a C-mount to a 2X magnification objective with an additional focusing lens (Microbench basic set, Qioptiq, Germany); the same optical setup was also mounted on the PCO camera for direct comparisons. The focal distance was

Fig. 1 e Schematic description of the membrane fouling simulator (MFS) including a water mixing system, pressure-drop sensor, flow meter and the imaging setup for O2 detection. The camera was a cooled PCO monochrome cooled CCD camera or a mEye color CMOS camera. The trigger-delay box and the control of the LED driver (both indicated with dashed lines) are only present in the lifetime setup with the fast gate-able digital camera (PCO camera). The black arrows indicate the flow direction. There is no flow via the pressure-drop sensor.

5030

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 2 7 e5 0 3 7

w1.5 cm. A long pass filter (>590 nm) was placed in between the camera chip and the lens to exclude background and reflected blue excitation light. With a binning of 2, the resolution of this setup is w36 mm per pixel. Additional measurements with the PCO camera were done with a Xenoplan XNP 1.4/17 objective (Schneider-Kreutznach, Germany) equipped with a long pass filter (>590 nm, Schneider-Kreutznach, Germany) and distance macro rings (2 mm) on the PCO camera to shorten the minimum focal distance from 15 to 5e10 cm. The resolution with this setup is w72 mm per pixel. Background light was further diminished by a black card box mounted around the camera setup. For the O2 measurements with the PCO camera, two modulated power LEDs were used for excitation of the optode (1W Luxeon Star, 470 nm, Lumileds, San Jose; USA), as controlled by a custom built trigger-delay box. The two LEDs were mounted next to the lens. For lifetime imaging, the PCO camera was used in a modulated measuring mode, in which luminescence intensity images were acquired sequentially by integrating series of 3 ms imaging periods starting respectively at 0.1 (IW1) ms and 4.1 (IW2) ms after the 4 ms long excitation pulses, and followed by dark image acquisition (see also fig. X1 and table X-1, additional information). The total integration time per measurement period was 500 ms (see Holst et al., 1998 for a detailed description of the measuring method and program). Lifetime (s) images were subsequently calculated from the luminescence intensity images IW1 and IW2 images according to s¼

Dt  ln IW1=I



(3)

W2

where Dt is the time difference between IW1 and IW2. For the ratiometric imaging method, we calculated the O2 concentration from the ratio of luminescence images (I0/I ) taken under complete anoxic conditions (I0) and oxygenated conditions (I ), respectively (see also fig. X-1 and table X-1, additional information). In measurements with the PCO camera system, we also used the IW1 images to calculate O2 images based on the ratio I0W1/IW1, thus allowing direct comparison of image quality by both lifetime and intensity based imaging. Additionally, intensity based measurements were also done using the intensity image setting in Look@ MOlli, where the O2 dependent luminescence intensity is measured during the blue LED excitation period, in contrast to the Iw1 images taken 0.1 ms after the excitation flash. The integration time for the intensity images was shorter (100 ms) than for the lifetime measurements (to measure comparable intensities as in the lifetime images), while the excitation pulse length was equal (4 ms). The measuring time per pulse was 3 ms. Dark images were automatically taken and subtracted in the image acquisition program. For the mEye color camera measurements, two power LEDs (1W Luxeon Star, 470 nm, Lumileds, San Jose; USA) were used for excitation of the optode. The LED light intensities were controlled by a stable and adjustable DC voltage source (GPC3030DQ, GW Instek, Tucheng City, Taiwan) allowing precise control of current and voltage. The exposure time for the mEye camera was 200 ms (see also table X-1, additional information) and the gain for the blue, green and red channel were set to 80,

85 and 12, respectively, to optimize the gray value distribution over the three color channels. Calibration of the optode in the MFS was done by circulating water with different O2 concentrations at a stabilized temperature. The O2 concentration in the water was changed by flushing the water with a series of set gas mixtures, mixed by a PC controlled automated gas mixing system based on electronic mass flow controllers (SensorSense, Nijmegen, Netherlands).

2.2.

Rhodamine assay

Imaging of biofilm thickness in the MFS was performed to compare O2 distribution with biomass distribution. Imaging was performed with the mEye camera without the long pass filter and samples were illuminated with 2 warm white LED strips (Hide-a-lite, Electro Elco AB, Sweden). A volume of 20 ml of a dilute rhodamine WT solution was injected into the MFS to measure the biofilm thickness. Rhodamine WT absorbs green-orange light, and absorption was measured by the green channel of the camera. If the measurement was carried out immediately after injection of the solution, rhodamine was only present in the flow channels, i.e. places were no biofilm was present. Calibration of the relationship between absorption and thickness was done with a triangle shaped glass cuvette with a thickness range of 0e700 mm, filled with the rhodamine solution (appendix, Fig X-2 and X-3,). A membrane (equal to the one in the MFS) was placed behind the cuvette to have the same light reflection characteristics as in the MFS. Images of the calibration cuvette were made at the same camera distance and illumination geometry as used for the biofilm thickness measurements in the MFS. All image calculations (O2 sensing and rhodamine assay) were done in the freeware ImageJ (version 1.45a; http://rsb. info.nih.gov/ij). After import to ImageJ, the color RGB images were split. All images were converted to a 32 bit floating point format before initial thresholding. Thresholding was performed to exclude low intensity pixels, i.e. pixels within areas where no fluorescence was measured (for example areas with marker ink), from further calculations (i.e ratio calculations, lifetime calculations etc). Ratio images were calculated using the “process/ image calculator” option in ImageJ. The color camera images were split into their Red, Green and Blue channels prior to ratio calculations, by using the “image/color/split channels” option in ImageJ. False coloring of the images was applied to visualize/emphasize differences in O2 concentration. All data handling in ImageJ was done manually.

3.

Results

3.1.

Calibration

For calibration, lifetime images and luminescent intensity images of a clean optode (without biofilm) were recorded at 5 different O2 concentrations. From these images, we calculated ratio images for the ratios s0 =s, I0W1/IW1 of the PCO camera and the I0/I of the red channel of the mEye camera at the different O2 concentrations (Fig. 2), and we averaged the values of

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 2 7 e5 0 3 7

a region of interest (1.6  105 pixels), comprising half of the imaged area. Selection of the regions of interest was done to exclude areas in the image without oxygen sensitive dye as well as areas with pixel values below a threshold grey value (<30). All three imaging methods showed a linear relationship with O2 concentration (r2 > 0.99), indicating that the nonquenchable fraction (a) was small. While the calibration curves of the different methods showed some variations, the use of different linear correlations to convert I0/I into oxygen concentrations (see caption in Fig. 2) allowed us to compare the calculated O2 images based on measurements with the different methods.

3.2.

Comparison of different methods

To evaluate the performance of the ratiometric and the lifetime approach, we compared O2 images based on luminescence images taken with the PCO camera (2X lens mounted) using different calculation and measuring methods (Fig. 3). All pictures were taken from exactly the same area of the planar optode in the biofilm growth incubator (MFS). The biofilm in the MFS had been growing for a period of 10 days. The luminescence images under full anoxic conditions (I0 and I0w1) were taken after injection of saturated sodium dithionite solution into the MFS. Images (I and Iw1) and lifetime images (Iw1 and Iw2) in the presence of O2 were taken, while oxygenated medium was flowing through the MFS. The O2 distribution on the optode surface was highly heterogeneous due to variations in biofilm thickness. The biofilm partially blocked the water flow causing formation of

Fig. 2 e Calibration curves of the O2 dependent ratios of s0 =s ( y [ 1049.8x L 1048.5, circles, Ksv [ 9.53 3 10L4), the intensity I0W1/IW1 ( y [ 1164.5x L 1160.2, squares, Ksv [ 8.59 3 10L4) both measured with the PCO camera, and the I0/I ratio of the luminescence intensity in the red channel measured with the mEye camera ( y [ 974.8x L 973.6, triangles, Ksv [ 1.03 3 10L3). The linear correlation (r2) for all curves was >0.999. The points represent average values of 1.6 3 105 pixels in a square Regions Of Interest (ROI) in the center of the image. Error bars denote the standard deviations of the ratio values of the pixels.

5031

flow channels in the exopolymer matrix. All three O2 images showed the same spatial O2 distribution patterns, and the O2 concentrations were in the same range. However, the O2 images based on the ratio method I0W1/IW1 or I0/I (Fig. 3) showed much more details, as compared to the lifetime images. The biofilm in the MFS formed up to 0.8 mm thick structures due to the support of the spacer mesh (Fig. 4). Formation of such thick biofilms caused an increased pressure gradient within the MFS (data not shown) forcing medium through narrow channels. The flow channels were visualized by injection of rhodamine WT solution into the MFS (Fig. 4C). The mEye and many other color cameras allow manual gain setting of the separate color channels to optimize the pixel saturation per channel. For O2 sensing, we used the intensity histograms of each color channel in the camera program at anoxia (highest luminescence) to optimize the gain settings of every channel. In this way, no pixels were oversaturated in any of the channels, while ensuring optimal signal-to-noise ratio. With such optimized gain settings, it was possible to see differences in fluorescence intensity, even without splitting the three channels (Fig. 5A). However, proper calculation of O2 images from the color camera .tiff format images still required splitting of the three channels. We used the red channel for O2 calculation since the emission spectrum of the Ru-dpp indicator (max. emission 610 nm) coincided best with the spectral range of the red channel. There was also a visible change in the green channel upon changing O2 concentration, while the blue channel did not show much variation. There was no biofilm growth in the MFS during the calibration, and the O2 concentration was considered homogeneous within the field of view. However, there was a clear intensity difference visible (Fig. 5B), due to spatial variation in excitation light and lens effects (Fig. 5C). After calculation of the ratio image, such heterogeneity was gone (Fig. 5D, E) and the image ratio showed a homogeneous response of the planar optode area in the camera field of view to the different O2 concentrations. To compare the ratio images from the mEye setup with the lifetime images of the PCO setup, we analyzed a position in the MFS that was imaged by both camera systems, consecutively (Fig. 6). Both lifetime and ratio images measured with the mEye camera yielded relatively noisy images, as compared to the ratio I0W1/IW1 image and the I0/I image made with the PCO camera (Fig. 3B, D). However, a structured O2 distribution due to flow channel formation within the biofilm was still clearly visible. Closer inspection of the images showed that the O2 distribution did not match exactly in the two image types. This difference is most likely caused by repetitive changes in flow rate. In our setup, we had to stop the flow to inject the saturated dithionite solution for the anoxic measurement before the flow was started again. This can result in detachment of part of the biofilm structure. In our example, most of the structure stayed intact and the images are still comparable (Fig. 6).

3.3.

Heterogeneity of O2 distribution and consumption

The highest O2 concentrations were generally found in areas with no or little biofilm formation (Fig. 4D), while low O2 zones developed away from the flow channels in areas with thick

5032

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 2 7 e5 0 3 7

Fig. 3 e Images of O2 distribution in the biofilm growth incubator (MFS) as measured with a 12 bit PCO camera system with a 2X magnification microscope lens mounted. The O2 concentrations in the images are calculated from lifetime imaging (A), the ratio of the IW1 images (I0W1/IW1) under anoxic conditions (I0W1) and under oxygenated conditions (IW1) (B), and the ratio of the luminescence intensity (I0/I ) of images recorded under anoxic and oxygenated conditions, respectively (C). The color bar indicates the O2 concentration in mmol lL1. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

biofilm and little water movement. However, overlaying the O2 image with the biofilm distribution image also showed some zones with relative dense biofilm formation and relative high O2 concentrations, as well as some regions exhibiting low O2 concentrations without thick biofilm formation. The low O2 zones did not reach complete anoxia at the optode surface (Fig. 4A, B). No differences were found between the two imaging approaches. The average O2 concentration was 25  13 mmol l1 in the largest O2 depleted zone (Fig. 7A left side, Region Of Interest (ROI) 1), while the O2 concentration in other depleted zones was 52  13 mmol l1. The O2 concentration reached zero within these regions w60e90 s after the flow in the biofilm monitor was stopped. In comparison, the O2 concentration in the biofilm flow channels was 152  23 mmol l1 and became anoxic 100e150 s after the flow was stopped. We estimated the O2 depletion rate by measuring a series of O2 images after stopping the flow with a frequency of 1 image per 10 s, and by subsequent subtraction of images in such time series. Lifetime images were generally too noisy to acquire accurate O2 depletion rate images based on two sequential images, and averaging of 4e5 sequential images was necessary. However, the ratio (I0w1/Iw1) based O2 images were less noisy, and with these images it was possible to calculate O2 depletion rates images without averaging. The initial O2 depletion rate (Fig. 7) in the channels measured by the ratio method was w3.57  1.97 mmol l1 s1 (average value of 6.3  104 pixels), which was 2e4 times the depletion rate (1.30  1.39 mmol l1 s1 average value of 2.9  104 pixels) found in the low O2 zones (ROI1 excluded). However, this was partly caused by the difference in initial O2 concentration at the moment the flow was stopped (Fig. 7B). Oxygen depletion rates in the different channel regions were about 30% higher (1.85  1.98 mmol l1 s1) at w50 mmol l1 (average 50  24 mmol l1). The large zone with low O2 on the left site (ROI1) had a lower O2 depletion rate 0.65  1.31 mmol l1 s1 at 25  14 mmol l1 (3X104 pixels, while the other low O2 zones had a depletion rate of 1.00  1.28 mmol l1 s1 at an average O2 concentration of 29  12 mmol l1. The O2 depletion rates in the channels were 1.15  1.38 mmol l1 s1 at an average O2

concentration of 31  19 mmol l1. The O2 depletion rates seemed equal for all 10 ROI’s at O2 concentrations below w20 mmol l1 (Fig. 7B). The respiration rates found in this study are within the ranges found using micro-electrodes (0.3e5 mmol l1 s1) (Nielsen et al., 1990; Satoh et al., 2005).

4.

Discussion

4.1.

Comparison of the different imaging techniques

All three imaging methods showed a linear relationship between O2 concentration and so =s, I0w1/Iw1 and I0/I, indicating a low non-quenchable fraction of the indicator in the planar optodes used in this study. In principle, this linear relationship for the range 0e300 mmol l1 O2 allows a simple two point calibration. The calibration images made with the color camera (Fig. 5) showed that the ratio approach corrects for variation in the luminescence intensity due to heterogeneity in the light field and lens effects. This makes the method applicable for imaging in systems even when the light distribution of the excitation lights is not perfectly homogeneous. A prerequisite is, however, that the measured area can be made anoxic without physical change or movement of the setup. In our case, we created anoxic conditions of the monitored area by addition of dithionite via an injection port in the tubing or by stopping the flow of the medium. But even though we took care not to touch the camera or MFS, the image position sometimes changed in between capturing images, due to minute movement of the MFS relative to the camera. The shift was only 2 or 3 pixels, but is enough to create erroneous results. The risk of movement is much bigger in our system than in the lifetime setting, since nothing is physically touched in between the recording of the lifetime images. The time difference between the recording is much smaller than for the I0/I images. However, within ImageJ it can be easily checked whether the images are still aligned and, if necessary, the images can be repositioned based on chosen reference points within both images.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 2 7 e5 0 3 7

5033

Fig. 4 e Images of O2 concentration in an MFS harboring mature biofilms (10 days old). The O2 was measured using luminescence lifetime imaging (A), and by taking the ratio between luminescence intensity images measured under anoxia and under experimental conditions (B). The heterogeneity in O2 concentration is mostly caused by channel formation within the incubator. The color bar indicates the O2 concentration in mmol lL1. (C) Visualization of the water channels in the same position. The color bar indicates the water volume of moving water based on light absorption in the green channel after injection of rhodamine WT solution into the MFS. The color bar of C indicates the biofilm thickness in mm. After tresholding the image in panel C a mask is created with two different ranges of biofilm thicknesses. The white areas of the mask (D) indicate the open channel, the gray shaded zones indicate areas where the biofilm was 0.1e0.3 mm thick, while black zones indicate >0.3 mm thick biofilms. Panel E shows the O2 image (B) combined with the a mask (D) made of the biofilm distribution image. Panel F shows the oxygen image with the support mesh overlayed. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

The ratio I0/I images from the color camera did not have the same spatial accuracy as the ratio based oxygen images from the PCO camera. This is not surprising, given that the PCO camera had a more sensitive cooled CCD chip with a bit depth of 12, while the color camera had a simple CMOS chip with a depth of 8 bit. A bit depth of 12 results in a 16 times larger

dynamic range as compared to 8 bit cameras, and thus a much higher accuracy in the calculation of the ratio. However, the resolution, even with our relatively simple 8 bit color camera was almost as good as the resolution of the lifetime images. The accuracy may easily be improved by using better cameras e.g. with a bit depth of 12 per color channel. Cameras with a bit

5034

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 2 7 e5 0 3 7

Fig. 5 e Luminescence intensity images at different O2 levels (expressed as % O2 in gas phase) (A) from an O2 sensitive planar optode mounted in an MFS measured with the mEye camera (8 bit color). (B) the red channel of the color images, (C) the average gray values of the red channel, (D) images of the ratioI0/I. (D) the average value of the ratio (E). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

depth of 12 and cooled down to 30  C below ambient temperature can be found nowadays for prices starting under 1000 V (e.g. cameras from Tucsen inc., China or Optic Star, United Kingdom), which is substantially below costs and efforts involved in establishing a luminescent lifetime system. The O2 images calculated from lifetime measurements had a lower resolution than the images based on the ratio I0/I. The higher resolution of the I0/I ratio approach thus allowed visualization of O2 gradients, which were not clearly visible in the lifetime images. The reason for the difference in spatial

resolution may be caused by the exponential character of lifetime images. Taking the exponent of the ratio Iw1/Iw2, as is done in the lifetime images, increases the noise level of the oxygen image. There was no obvious difference in resolution between the I0w1/Iw1 and I0/I. Another imaging approach based on intensity images, rather than on lifetime was used in the very first planar optode applications (Glud et al., 1996, 1999), where Ksv and a in Equation (2) were determined by measuring luminescence intensity images at 0%, 20%, and 100% O2. Subsequently, the

Fig. 6 e Comparison of lifetime image obtained with the PCO camera (A), and a I0/I image (B) as calculated from the red channel images of the 8 bit mEye color camera from approximately the same area in the MFS. The I0/I was reduced in size to correct for differences in pixel size. Positioning and resizing was based on the black ink structure (upper left) drawn on the glass window. The color bar indicates the O2 concentration in mmol lL1. The black areas are caused by marks made with a black permanent marker on the window. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 2 7 e5 0 3 7

5035

during calibration, yielding a relatively flexible and robust, yet simple O2 imaging setup. In our case, the image ratio had a linear relationship to O2 concentration over 0e300 mmol l1, but also non-linear relationships can be used for the conversion within ImageJ. It has been found that other types of fluorescent dyes have nonlinear relationships with O2 concentration e.g. due to a substantial non-quenchable fraction of the immobilized indicator (Carraway et al., 1991). For planar optodes exhibiting a linear relationship between image ratios and O2 concentration, it is possible to calculate dynamic changes (e.g. during experimental lightedark shift or stop-flow) without making an image under anoxic conditions simply by calculating the ratio of two consecutive luminescence intensity images multiplied by the slope found in the O2 calibration curve.

4.1.1.

Fig. 7 e Relationship between O2 concentration and O2 depletion rate for 10 regions of interest (ROIs) in the MFS. Panel A shows an O2 distribution image with a selection of different ROIs from which the depletion rates were measured. Panel Bshows the relationship in the areas where water is flowing freely (ROI 9 and 10), as well as in the low O2 zones, characterized by the absence of free moving water (ROI 1-8). The points in the graphs indicate the average O2 depletion rates at average O2 concentration values of the ROI’s indicated in the legend.

Ksv and a images where used to convert experimental luminescence intensity images to O2 concentration images. While yielding high quality O2 images, this approach required the use of non-transparent planar optodes as pure intensity based O2 imaging is prone to light field variations and e.g. scattering artifacts from the sample structure (Holst et al., 1998). This method also requires the camera not be moved relative to the optode. In the setup presented here, the O2-dependent image ratio is calculated directly from the ratio of an image obtained under anoxic conditions and an image under a given O2 concentration. The conversion from ratio to O2 image is carried out by a linear correlation. This correlation can be determined prior to or afterward actual experiments and is not done at pixel level. Therefore it is not required that the optode has exactly the same position relative to the camera

O2 heterogeneity in the biofilm monitor

The O2 distribution was heterogeneous in the MFS after biofilm development. There was a good agreement with the distribution of free moving water and the higher O2 concentrations. It seems logic that O2 concentrations are higher in channels since little biomass is present in the channels, while water is flowing relatively fast. Advective transport is important in these regions, and fast flowing water will result in thin boundary layers, while little biomass is present. The O2 concentration at the optode surface depends on the thickness of the diffusive boundary layer on top of the biofilm, the thickness of the biofilm on top of the optode and the O2 consumption rate within the biofilm. However, some regions seemed to deviate from the relationship between water volume and higher O2 concentration. This may partly be explained by the fact that the rhodamine method to estimate the free moving water did not discriminate for the distance at which the water flows from the optode. When half of the space is filled with biofilm, the free moving water may flow directly over the optode, but it may also be that the biofilm is between the optode and the water. Both situations will give the same absorbance, but will result in different O2 concentrations at the optode surface. This illustrates the current limitations in our ability to experimentally resolve spatially complex chemical landscapes and mass transfer phenomena in biofilms. It was surprising that O2 was not fully depleted in regions without flowing medium. These regions did only become anoxic after stopping the flow in the MFS, while the lowest O2 concentration under flow conditions in the MFS was w25 mmol l1. Advective flow through the biofilm as reported in several studies (Costerton et al., 1999; Stewart and Franklin, 2008) could explain the observation. Since these channels are extremely small (a few mm) advective flow will be relatively slow due to a high resistance. This flow may have been overlooked by the rhodamine method used in this study. We only injected 20 ml rhodamine solution, prior to stopping the flow. This was enough to penetrate the main channels, but may not enough to show a flow in areas almost completely filled with biofilm. Another explanation may be that the diffusion of the organic substrate into the biofilm is slow, and not in balance with the diffusion of O2. This would result in the consumption of all organic material in the zones next to the channels, while

5036

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 2 7 e5 0 3 7

not all O2 is consumed. The stop-flow experiments showed that the O2 depletion rate was indeed relatively low in the biofilm zones (Fig. 4, left side). The O2 depletion rate was almost 2 times lower compared to the depletion rate in the channel regions, while the biomass was assumed to be much higher in the biofilm zones. This could indicate that the biofilm in the low O2 zones is either inactive or limited by organic substrates rather than by O2. However, there was still some consumption in these regions, as shown by the onset of anoxic conditions when flow was stopped. Actually, biofilm regions reached anoxia before the channel regions turned anoxic.

4.1.2.

O2 depletion and the stop-flow technique

The O2 concentrations in the medium at the inlet and outlet of the MFS were measured daily with an O2 fiber optode showing that the O2 concentration in the medium dropped by 25 mmol l1 during its residence in the MFS, while O2 imaging experiments were conducted. At the moment of the planar optode measurement, the flow was reduced from 16 to 2.7 L h1 due to resistance caused by the biofilm formation. This would result in a total O2 consumption rate of 18.7 nmol s1 in the MFS. The average O2 depletion rate measured using the stop-flow technique was 2.2 mmol l1 s1 for the monitored region. The free volume of the MFS was 6.5 ml, resulting in a total O2 consumption rate of the whole monitor of 14.3 nmol s1. This value is rather close to the value estimated by measuring the O2 decrease between the inlet and the outlet, although it is a bit lower. One explanation for the lower value may be that the O2 sensor used to measure O2 concentrations in the inlet and outlet measures in the tubing a bit out of the monitor. Since biofilms will also form in the tubing, this will add an extra consumption component to the system, resulting in a higher consumption rate. The stop-flow technique combined with the planar optode does not involve the tubing. Another explanation is that the surface area that was monitored does not reflect the behavior of the total incubator perfectly. In general, these two methods yield the same overall results, but the imaging method gives much more information on the heterogeneity in process rates. Many microelectrode studies are published, showing oxygen gradients and oxygen fluxes in biofilms, but volumetric O2 respiration rates of biofilms were determined in only few publications. Several volumetric respiration rates found with microelectrode studies were within the same range as the O2 depletion rates found in this study (0.3e5 mmol l1 s1, e.g. see Nielsen et al., 1990; Satoh et al., 2005). Therefore we feel that the O2 depletion rates found with the ratiometric imaging approach are representative for the respiration rates of the biofilm. More extreme values, ranging from 0.03 (Polerecky et al., 2005) to 50 mmol l1 s1 (De Beer and Costerton, 2006) have also been reported, but the differences found may rather result from environmental factors like temperature, availability of organic substrate, etc. than from the method used. With the planar optode it can be assumed that the oxygen depletion rate reflects the oxygen consumption rate of the thin layer just on top of the optode, especially when the oxygen decrease is measured during the first 10 s after the flow is stopped. Stopping the flow will disrupt the steady state O2 gradient and therefore reflect the

respiration rate (Staal et al., 2011). After 5e10 s the O2 depletion rate will become increasingly affected by a change in flux from overlaying layers.

5.

Conclusions

It can be concluded that the I0/I ratio approach for O2 imaging with transparent planar O2 optodes can be considered a good and more simple alternative to more elaborate luminescence lifetime imaging approaches. However, our approach can only be used in systems were the O2 concentration can be reduced to zero without physical movement of the MFS relative to the camera. Such mechanical stability can easily be achieved in most biofilm reactor setups. A very simple imaging setup for imaging O2 in biofilm reactors can be established from inexpensive commercial CCD or CMOS cameras and high intensity LEDs. The image analysis can be performed by powerful freeware such as ImageJ. In line with other recent ratiometric O2 imaging approaches (Wang et al., 2010; Larsen et al., in press), this simplifies O2 imaging and makes it more accessible for the research community. Here we presented O2 distribution images measured at the base of a biofilm, but with some modifications in the experimental setup it is also possible to measure the O2 distribution inside granules or biofilms.

Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.07.007.

references

Bacon, J., Demas, J., 1987. Determination of oxygen concentrations by luminescence quenching of a polymerimmobilized transition-metal complex. Analytical Chemistry 59, 2780e2785. Carraway, E.R., Demas, J.N., DeGraff, B.A., Bacon, J.R., 1991. Photophysics and photochemistry of oxygen sensors based on luminescent transition-metal complexes. Analytical Chemistry 63, 337e342. Costerton, J.W., Lewandowski, Z., Caldwell, D.E., Korber, D.R., Lappin-Scott, H.M., 1995. Microbial biofilms. Annual Review of Microbiology 49, 711e745. Costerton, J.W., Stewart, P.S., Greenberg, E.P., 1999. Bacterial biofilms: a common cause of persistent infections. Science 284, 1318e1322. De Beer, D., Costerton, J.W., 2006. Microbial biofilms. In: Dworkin, M., Falkow, S., Rosenberg, E., Schleifer, K.H., Stackebrandt, E. (Eds.), Prokaryoteseds. Springer, New York, pp. 904e937. De Beer, D., Stoodley, P., Roe, F., Lewandowski, Z., 1994. Effects of biofilm structures on oxygen distribution and mass transport. Biotechnology and Bioenginering 43, 1131e1138. Eberl, H.J., Picioreanu, C., Heijnen, J.J., van Loosdrecht, M.C.M., 2000. A three-dimensional numerical study on the correlation of spatialstructure, hydrodynamic conditions, and mass transfer and conversionin biofilms. Chemical Engineering Science 55, 6209e6222.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 2 7 e5 0 3 7

Fenchel, T., Finlay, B., 2008. Oxygen and the spatial structure of microbial communities. Biological Reviews 83, 553e569. Glud, R.N., Ramsing, N.B., Gundersen, J.K., Klimant, I., 1996. Planar optrodes: a new tool for fine scale measurements of two-dimensional O2 distribution in benthic communities. Marine Ecology Progress Series 140, 217e226. Glud, R.N., Santegoeds, C.M., de Beer, D., Kohls, O., Ramsing, N.B., 1998. Oxygen dynamics at the base of a biofilm studied with planar optrodes. Aquatic Microbial Ecology 14, 223e233. Glud, R.N., Ku¨hl, M., Kohls, O., Ramsing, N.B., 1999. Heterogenity of oxygen production and consumption in a photosynthetic microbial mat as studied by planar optodes. Journal of Phycology 35, 270e279. Glud, R.N., 2008. Oxygen dynamics of marine sediments. Marine Biology Research 4, 243e289. Hall-Stoodley, L., Costerton, J.W., Stoodley, P., 2004. Bacterial biofilms: from the natural environment to infectious diseases. Nature Reviews Microbiology 2, 95e108. Hidalgo, G., Burns, A., Herz, E., Hay, A.G., Houston, P.L., Wiesner, U., Lion, L.W., 2009. Functional tomographic fluorescence imaging of pH microenvironments in microbial biofilms by use of silica nanoparticle sensors. Applied and Environmental Microbiology 75 (23), 7426e7435. Holst, G., Grunwald, B., 2001. Luminescence lifetime imaging with transparent oxygen optodes. Sensors & Actuators, B: Chemical 74, 78e90. Holst, G., Kohls, O., Klimant, I., Konig, B., Ku¨hl, M., Richter, T., 1998. A modular luminescence lifetime imaging system for mapping oxygen distribution in biological samples. Sensors and Actuators B: Chemical 51, 163e170. Huettel, M., Ziebis, W., Forster, S., 1996. Flow-induced uptake of particulate matter in permeable sediments. Limnology and Oceanography 41, 309e322. Ivleva, N.P., Wagner, M., Horn, H., Niesner, R., Haisch, C., 2010. Raman microscopy and surface-enhanced Raman scattering (SERS) for in situ analsysis of biofilms. Journal of Biophotonics 3 (8e9), 548e556. Jørgensen, B.B., Des Marais, D.J., 1990. The diffusive boundary layer of sediments: oxygen microgradients over a microbial mat. Limnology & Oceanography 35, 1343e1355. Ku¨hl, M., Rickelt, L.F., Thar, R., 2007. Combined imaging of bacteria and oxygen in biofilms. Applied and Environmental Microbiology 73 (19), 6289e6295. Ku¨hl, M., Glud, R.N., Ploug, H., Ramsing, N.B., 1996. Microenvironmental control of photosynthesis and photosynthesis-coupled respiration in an epilithic cyanobacterial biofilm. Journal of Phycology 32, 799e812. Ku¨hl, M., Jørgensen, B.B., 1992. Microsensor measurements of sulfate reduction and sulfide oxidation in compact microbial communities of aerobic biofilms. Applied and Environmental Microbiology 58, 1164e1174. Ku¨hl, M., 2005. Optical microsensors for analysis of microbial communities. Methods in Enzymology 397, 166e199. Ku¨hl, M., Polerecky, L., 2008. Functional and structural imaging of phototrophic microbial commmunities and symbioses. Aqautic Microbial Ecology 53, 99e118. Larsen, M., Borisov S.M., Grunwald B., Klimant I. and Glud R.N. A simple and inexpensive high resolution color ratiometric planar optode imaging approach: application to oxygen and pH sensing. Limnology & Oceanography Methods, in press. Neu, T.R., Manz, B., Volke, F., Dynes, J.J., Hitchcock, A.P., Lawrence, J.R., 2010. Advanced imaging techniques for assessment of structure, composition and function in biofilm systems. FEMS Microbiology Ecology 72, 1e21. Nicolella, C., van Loosdrecht, M.C.M., Heijnen, S.J., 2000. Particlebased biofilm reactor technology. Trends in Biotechnology 18, 312e320.

5037

Nielsen, L.P., Christensen, P.B., Revsbech, N.P., Sørensen, N.P., 1990. Denitrification and oxygen respiration in biofilms studied with a microsensor for nitrous oxide and oxygen. Microbial Ecology 19, 63e72. Oguri, K., Kitazato, H., Glud, R.N., 2006. Platinum octaetylporphyrin based planar optodes combined with an UV-LED excitation light source: an ideal tool for highresolution O2 imaging in O2 depleted environments. MarineChemistry 100, 95e107. Polerecky, L., Franke, U., Werner, U., Grunwald, B., de Beer, D., 2005. High spatial resolution measurement of oxygen consumption rates in permeable sediments. Limnology and Oceanography: Methods 3, 75e85. Revsbech, N.P., 2005. Analysis of microbial communities with electrochemical microsensors and microscale biosensors. Methods in Enzymology 397, 147e166. Ridgway, H.F., Flemming, H.F., 1996. Membrane biofouling. In: Mallevialle, J., Odendaal, P.E., Wiesner, M.R. (Eds.), Water Treatment Membrane Processes. McGraw-Hill, New York, pp. 6.1e6.62. Satoh, H., Sasaki, Y., Nakamura, Y., Okabe, S., Suzuki, T., 2005. Use of microelectrodes to investigate the effects of 2chlorophenol on microbial activities in biofilms. Biotechnology and Bioengineering 91, 133e138. Staal, M., Borisov, S.M., Rickelt, L.F., Klimant, I., Ku¨hl, M., 2011. Ultrabright planar optodes for luminescence life-time based microscopic imaging of O2 dynamics in biofilms. Journal of Microbial Methods 85, 67e74. Stewart, P.S., Franklin, M.J., 2008. Physiological heterogeneity in biofilms. Nature Reviews Microbiology 6, 199e210. Stoodley, P., deBeer, D., Lewandowski, Z., 1994. Liquid flow in biofilm systems. Applied and Environmental Microbiology 60, 2711e2716. Vrouwenvelder, J.S., van Paassen, J.A.M., Wessels, L.P., van Dam, A.F., Bakker, S.M., 2006. The Membrane fouling simulator: a practical tool for fouling prediction and control. Journal of Membrane Science 281 (1e2), 316e324. Vrouwenvelder, J.S., Bakker, S.M., Cauchard, M., Le Grand, R., Apacandie, M., Idrissi, M., Lagrave, S., Wessels, L.P., van Paassen, J.A.M., Kruithof, J.C., van Loosdrecht, M.C.M., 2007. The membrane fouling simulator: a suitable tool for prediction and characterisation of membrane fouling. Water Science & Technology 55 (8e9), 197e205. Vrouwenvelder, J.S., Manolarakis, S.A., van der Hoek, J.P., van Paassen, J.A.M., van der Meer, W.G.J., van Agtmaal, J.M.C., Prummel, H.D.M., Kruithof, J.C., van Loosdrecht, M.C.M., 2008. Quantitative biofouling diagnosis in full scale nanofiltration and reverse osmosis installations. Water Research 42, 4856e4868. Vrouwenvelder, J.S., Hinrichs, C., Van der Meer, W.G.J., Van Loosdrecht, M.C.M., Kruithof, J.C., 2009a. Pressure drop increase by biofilm accumulation in spiral wound RO and NF membrane systems: role of substrate concentration, flow velocity, substrate load and flow direction. Biofouling 25, 543e555. Vrouwenvelder, J.S., Van Paassen, J.A.M., Kruithof, J.C., Van Loosdrecht, M.C.M., 2009b. Sensitive pressure drop measurements of individual lead membrane elements for accurate early biofouling detection. Journal of Membrane Science 338, 92e99. Wagner, M., Manz, B., Volke, F., Neu, T.R., Horn, H., 2010. Online3 assessment of biofilm development, sloughing and forced detachment in tube reactor by means of magnetic resonance microscopy. Biotechnology and Bioengineering 107 (1), 172e181. Wang, X., Meier, R.J., Link, M., Wolfbeis, O.S., 2010. Photographing oxygen distribution. Angewandte Chemie 122, 5027e5029.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 3 8 e5 0 4 6

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

The role of the microbial stringent response in excess intracellular accumulation of phosphorous in mixed consortia fed synthetic wastewater Muamar M. Al-Najjar a, Erik R. Coats b, Frank J. Loge a,* a b

Department of Civil and Environmental Engineering, University of California Davis, One Shields Avenue, Davis, CA 95616, USA Department of Civil Engineering, University of Idaho, Moscow, ID 83844-1022, USA

article info

abstract

Article history:

Four bench-scale sequencing batch reactors (SBRs) seeded with activated sludge were

Received 30 November 2010

operated under either fully oxic or anoxic/oxic conditions and fed synthetic wastewater

Received in revised form

containing either peptone or acetate. The function of each reactor was assessed through

24 May 2011

the measure of (i) soluble chemical oxygen demand, orthophosphate, ammonia, and

Accepted 3 July 2011

nitrate; and (ii) biomass concentrations of phosphorus, polyhydroxyalkanoate, guanosine

Available online 23 July 2011

tetraphosphate, adenosine monophosphate, adenosine diphosphate, and adenosine triphosphate. In all four reactors, the biomass concentration of phosphorous was corre-

Keywords:

lated statistically with the biomass concentration of ppGpp. The microbial consortia in all

Biological phosphorus removal

four reactors removed an appreciable quantity of phosphorous from solution (67e99%),

Wastewater treatment

and the net quantity of phosphorous removed from solution corresponded to the net

Phosphate accumulating organisms

increase in the biomass concentration of phosphorous. Hence, the microbial stringent

Polyhydroxyalkanoate

response (MSR) was associated with excess intracellular accumulation of phosphorous in

Guanosine tetraphosphate

mixed microbial consortia fed synthetic wastewater. With recognition of the potential role of the MSR in the removal of soluble phosphorous from wastewater, additional research may lead to further optimization of treatment technologies and the development of new treatment systems for the biological removal of phosphorus from wastewater. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

In pure microbial cultures, the storage of phosphorus has been shown to involve the transfer of the terminal phosphate of adenosine triphosphate (ATP) to polyphosphate (polyP) using polyP kinase 1 (PPK1) or polyP kinase 2 (PPK2) (Rao et al., 2009). Conversely, another set of enzymes (exopolyphosphatases (PPX)) hydrolyze polyP to orthophosphate (Ault-Riche´ et al., 1998). PPX has been found to exist under a common promoter with PPK (Ault-Riche´ et al., 1998), seemingly as

a means to regulate the polyP pool or otherwise prevent excess polyP accumulation. Inactivation of PPX could thus induce excess accumulation of polyP. Control of PPX has been associated with the microbial stringent response (MSR) (Kuroda et al., 1997). The MSR, a global regulatory system that functions as an important cellular survival mechanism under conditions of stress (Chatterji and Ojha, 2001), is characterized, in part, by accumulation of the alarmones guanosine tetra-(ppGpp) and penta-phosphate (pppGpp) (Kuroda et al., 1997). Synthesis of

* Corresponding author. Tel.: þ1 530 754 2297; fax: þ1 530 752 7872. E-mail address: [email protected] (F.J. Loge). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.006

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 3 8 e5 0 4 6

(p)ppGpp is regulated through expression of the RelA and/or SpoT (Potrykus and Cashel, 2008). The MSR is stimulated, in part, by unbalanced growth conditions, and can be elicited by subjecting cells to nutritional or osmotic stress (Ault-Riche´ et al., 1998), amino acid starvation (Kuroda et al., 1997), a downshift in aqueous pH (Wells and Gaynor, 2006), or changes in terminal electron acceptor (Glass et al., 1979; Mouery et al., 2006). From an environmental perspective, phosphorus is of great concern in protecting the quality of surface waters. Anthropogenic activities can result in the release of excess nutrients into aquatic environments that lead to advanced surface water body eutrophication (Pretty et al., 2003), and in many cases, phosphorus is often the limiting macronutrient for hyper-algal growth (Heathwaite and Sharpley, 1999). Although non-point source discharges arguably contribute large nutrient loads to surface water bodies (Powers, 2007), point source discharges from wastewater treatment facilities nonetheless receive the most attention due to their relative obtrusiveness and ease in which they can be and are regulated. Phosphorus can readily be removed from wastewater using mixed microbial consortia by applying an engineered process known as enhanced biological phosphorus removal (EBPR); however, molecular-level mechanisms controlling excess intracellular accumulation of phosphorous in EBPR have not yet been fully elucidated. EBPR theory stipulates that the imposed environmental conditions: (i) enrich for polyP accumulating organisms (PAOs) (Mino et al., 1987; Seviour et al., 2007), and (ii) promote the biological cycling of aqueous phosphorus (Comeau et al., 1986; Mino et al., 1987; Seviour et al., 2007). Anaerobically, PAOs hydrolyze intracellular polyP and release phosphate into wastewater, and assimilate and store organic carbon from wastewater as polyhydroxyalkanoates (PHA). Aerobically, PAOs assimilate aqueous phosphate and store it as intracellular polyP, and utilize PHA for energy, growth, polyP storage, and maintenance. The quantity of aqueous phosphate assimilated oxically exceeds that released by PAOs anaerobically, thereby reducing the bulk solution phosphorus concentration. A portion of the activated sludge is wasted from the EBPR process at the end of the oxic phase to achieve net phosphorus removal. EBPR is the most environmentally friendly method for removing phosphorus from wastewater (Coats et al., 2011). There is significant interest in better understanding how microorganisms accumulate phosphorus in EBPR in an effort to improve process performance. The goal of the study reported herein was to evaluate the potential role of the MSR in excess intracellular accumulation of phosphorous in mixed microbial consortia fed synthetic wastewater. Bench-scale sequencing batch reactors (SBRs) were operated under conditions specifically designed to elicit the MSR. However, the SBRs were not exclusively operated to simulate EBPR conditions (i.e., cyclical anaerobic/oxic environments). The function of each reactor was assessed through the measure of (i) soluble chemical oxygen demand (sCOD), orthophosphate, ammonia, and nitrate and (ii) biomass concentrations of phosphorus, PHA, ppGpp, adenosine monophosphate (AMP), adenosine diphosphate (ADP), and adenosine triphosphate (ATP).

2.

Materials and methods

2.1.

Operation of sequence batch reactors

5039

Four bench-scale SBRs were fabricated from an acrylic glass column with working volumes of 4 L. The SBRs were fed synthetic wastewater and inoculated with activated sludge derived from the University of California Davis (UCD) wastewater treatment plant (WWTP), which is operated as an extended aeration activated sludge process with a mean cell residence time (MCRT) of 20 days. Each laboratory reactor was operated with three feed-cycles per day, a hydraulic residence time (HRT) of 1 day and MCRT of 16 days, at a controlled room temperature of 211  C. Each feed-cycle consisted of 1.5 h settling, 0.25 h decanting, 0.25 h feed addition, and 6 h reaction. Two SBRs were operated with a fully oxic reaction period, while the other two SBRs were operated with a reaction period consisting of 2.5 h of anoxic conditions and 3.5 h of oxic conditions. Air was supplied (1.5 L min1) to provide oxygen and mixing, regulated by an air flow meter (ColeeParmer Instrument Co, Vernon Hills, IL, USA). Nitrogen gas was provided by a nitrogen tank with a regulator, and was used to facilitate a dissolved oxygen concentration of approximately zero during the anoxic period of the anoxic-oxic SBRs. MasterFlex pumps (L/S 7553-70, ColeeParmer Instrument Co, Vernon Hills, IL, USA) were used for feeding and decanting. The MasterFlex pumps were connected to digital timers (Model DT 17, Intermatic Incorporated, Spring Grove, IL, USA) to control the feed, reaction, settling, and decant periods.

2.2.

Synthetic wastewater

The synthetic wastewater contained per liter either 2 g peptone (Bacto peptone, BD, Franklin Lakes, NJ, USA) or 2 g sodium acetate (CH3COONa); 0.8 g NaHCO3; 0.25 g KH2PO4; 0.30 g NH4Cl; 0.42 g KCl; 0.42 g CaCl2$2H2O; 0.40 g MgSO4$7H2O; 5.5 mg C6H8O7$H2O; 3.03 mg FeCl3$6H2O, 0.5 mg H3BO3; 0.18 mg KI; 0.03 mg CuSO4$5H2O; 0.12 mg MnCl2$4H2O; 0.06 mg Na2MoO4$2H2O; 0.12 mg ZnSO4$7H2O; 0.15 mg CoCl2$6H2O; and 10 mg EDTA (Goel and Noguera, 2006). All chemicals were greater than 99.2% (w/w) grades and obtained from either Fisher Scientific (Fair Lawn, NJ, USA) or EMD Chemicals (Gibbstown, NJ, USA).

2.3.

Analytical methods

Reactor performance over the acclimation period (first six months of operation) was monitored every two weeks by sampling for soluble sCOD, phosphorus ðPO3 4  PÞ, ammonia (NH3eN), and nitrate ðNO 3  NÞ; and mixed liquor suspended solids (MLSS) at the beginning and at end of the entire feedcycle. Following the acclimation period, sampling was performed over the course of a single feed-cycle once every two months for a total time period of one year. During a typical sampling event for the anoxic/oxic SBRs, seven samples were collected: one each at the beginning and end; two samples during the anoxic period; one sample at the end of the anoxic period; and two samples during the oxic period. A 20 mL sample was recovered from each reactor for MLSS and mixed

5040

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 3 8 e5 0 4 6

liquor volatile suspended solids (MLVSS) analyses in accordance with Standard Method 2540 E (American Public Health Association, 2005). Another 12 mL sample was recovered from each reactor, filtered through sterilized 0.22-mm filters (Millipore Corporation, Billerica, MA, USA), and analyzed for soluble sCOD, phosphorus, ammonia, and nitrate. The sCOD test was performed in accordance with Standard Methods 5220-D (American Public Health Association, 2005) using Hach low-range ampules (Hach Company, Loveland, CO, USA). Ammonia and nitrate concentrations were measured using a continuous-flow inorganic nitrogen analyzer (Timberline Model 383, Timberline Instruments Inc., Boulder, USA). Concentrations of soluble phosphorus were measured using Hach method 8048 (Hach Company, Loveland, CO, USA). The Hach method is equivalent to method 4500-PE of Standard Methods (American Public Health Association, 2005). Cell dry weight (CDW) was estimated from a 10 mL sample, which was centrifuged (J6-HC, Beckman Coulter Inc., Fullerton, CA, USA) for 15 min at 4000 rpm and then rinsed once with Milli-Q water before drying at 105  C overnight. Dry biomass PHA content was determined by gas chromatography/mass spectrometry (GC/MS) as previously described by Braunegg et al. (1978). GC-MS was performed with split injection under an initial oven temperature of 40  C (2 min) ramped up to 200  C at 5  C min1 using a 30 m ZB-624 column (0.25 mm i.d., 1.4 mm film; Phenomenex, Torrance, CA, USA). The compounds (methyl ester derivatives) were scanned by comparing the MS spectra in Wiley 275 library to confirm PHA polymer composition. Additionally, the PHA polymers were identified by retention time and mass spectral matching 3-hydroxybutyric acid (3HB) and 3-hydroxyvaleric acid (3HV) and quantified based on external and internal standards. Total biomass PHA content was expressed on a dry weight basis (e.g., mass PHA (as COD)/mass of dry biomass, [w/w]). The mass of PHA was converted to a COD basis in accordance with Dionisi et al. (2004). Dry biomass phosphorus content was determined through sulfuric acid digestion of the sludge in accordance with Standard Methods 4500-PB (American Public Health Association, 2005) and subsequent quantification of soluble reactive phosphorus in accordance with Hach method 8048, using Hach lowrange ampules (Hach Company, Loveland, CO, USA). Total biomass phosphorous was expressed on a dry weight basis (e.g., mass phosphorous/mass of dry biomass, [w/w]). Dry biomass concentrations of ppGpp, AMP, ADP, and ATP were determined using high performance liquid chromatography (HPLC), as described by Neubauer et al. (1995) with the addition of 0.3 g of glass beads (diameter of 0.40e0.60 mm, Glasperlen, B. Braun Biotech International, Germany) during sonication to improve cell lysis. HPLC analyses of the extract samples were performed using a HP1100 series HPLC system (Agilent Technologies Inc., Santa Clara, USA) and an Agilent ChemStation system consisting of two pumps, an autosampler, and a diode array detector. Ion-pair reverse-phase chromatography was performed with a Supelcosil LC-18-T column (15 cm  3 mm, 3-mm particle size) connected to a Supelguard LC-18-T guard column (2 cm  3 mm) (Supelco, Bellefonte, USA). The temperature was set at 25  C, the flow rate of the eluent (HPLC buffer) at 0.50 mL min1, and absorbance detection at 252 nm. The injection volume was set at

10 mL, with a needle wash after injection. A sample containing only Milli-Q water and methanol was analyzed after every 10 samples as a quality assurance/quality control (QA/QC) measure. Total biomass ppGpp was expressed on a dry weight basis (e.g., mass ppGpp/mass of dry biomass, [w/w]). The adenylate energy charge (EC), which represents an estimate of the amount of metabolically available energy stored in the adenylate pool, was calculated as (Chapman et al., 1971):  ATP þ ð0:50  ADPÞ ðAMP þ ADP þ ATPÞ

 EC ¼

2.4.

(1)

Statistical analyses

Statistical correlations between the concentrations of soluble phosphorous, biomass ppGpp, biomass phosphorous, and biomass PHA were evaluated using the Pearson’s Product Moment Correlation Coefficient (r). For each statistical comparison in a given reactor, the value of r, and the associated p-value, was generated using Number Cruncher Statistical Software (Kaysville, Utah, USA) with 7 matched pairs of data obtained from 7 sampling events performed at distinct times over a given feed-cycle.

3.

Results

3.1.

Stabilization of SBRs

Stable operating conditions were reached within six months of operation for Reactors 1 (fully oxic fed peptone) and 2 (fully oxic fed acetate) and within three months for Reactors 3 (anoxic/ oxic fed peptone) and 4 (anoxic/oxic fed acetate). Once stable operating conditions were achieved, profiles of (i) soluble sCOD, orthophosphate, ammonia, and nitrate and (ii) biomass concentrations of phosphorus, PHA, ppGpp, AMP, ADP, and ATP over the course of a feed-cycle remained consistent throughout the 1-year duration of reactor operation. MLVSS in the SBRs stabilized at 3000e3200 mg L1 for Reactor 1; 3400e3800 mg L1 for Reactor 2; 3300e3700 mg L1 for Reactor 3; and 4000e4300 mg L1 for Reactor 4. Removal of soluble phosphorous in the SBRs stabilized at approximately 67, 99, 99, and 97% in Reactors 1, 2, 3, and 4, respectively.

3.2.

Performance of anoxic/oxic SBRs

During the anoxic period in the peptone-fed anoxic/oxic reactor (Reactor 3; Fig. 1): (i) sCOD was converted into PHA (Fig. 1a); (ii) biomass phosphorus was released into solution (Fig. 1c); (iii) the concentration of NH3eN increased (likely from ammonification associated with hydrolysis of peptone, which is an organic nitrogen rich carbon source); and (iv) the concentration of NO 3  N was below the detection limit for all but the first 10 min (i.e., the anoxic period was thus largely anaerobic) (Fig. 1b). During the subsequent oxic period: (i) biomass PHA was oxidized for microbial growth, phosphorus accumulation, and cell maintenance (Fig. 1a); (ii) aqueous phosphorus was incorporated into biomass (Fig. 1c); and (iii) the concentration of NH3eN decreased with a concurrent increase in NO 3  N (indicative of nitrification) (Fig. 1b).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 3 8 e5 0 4 6

5041

Fig. 1 e Concentrations of (a) sCOD and PHA, (b) ammonia and nitrate, (c) aqueous and biomass phosphorus, (d) ppGpp and EC over the course of one feed-cycle in the peptone-fed anoxic/oxic SBR (Reactor 3; results shown for samples collected after 1-year of SBR operation).

Overall, in regard to the anoxic/oxic cycling of phosphorus, PHA, and sCOD, the microbial consortium performed according to EBPR theory, and an appreciable quantity of phosphorous was removed from solution (ca. 99%). The MSR, reflected in a biomass ppGpp concentration, cycled inversely to the soluble phosphorus concentration (Fig. 1c and d). Applying a mass balance, the net increase in biomass phosphorous corresponded with the net quantity of phosphorous removed from bulk solution. The EC level, a measure of cellular energy, decreased from 0.98 to 0.81 over the anoxic period, a reflection of the limited ability of the consortium to generate ATP. When energetic conditions dramatically improved with the addition of oxygen, the EC subsequently increased to a value of 0.93 by the end of the oxic period (Fig. 1d). During the anoxic period of the acetate-fed anoxic/oxic reactor (Reactor 4; Fig. 2): (i) sCOD was converted into biomass PHA (Fig. 2a); (ii) biomass phosphorous was released into solution (Fig. 2c); (iii) the concentration of NH3eN decreased; and (iv) the concentration of NO 3  N was below the detection limit (e.g., the anoxic period was anaerobic) (Fig. 2b). In the subsequent oxic period: (i) biomass PHA was rapidly consumed (Fig. 2a); (ii) aqueous phosphorus was incorporated

into biomass (Fig. 2c); and (iii) the concentration of NH3eN decreased below the detection limit resulting in a nitrogen limitation (i.e., non-detectable concentrations of both NH3eN and NO 3  N) (Fig. 2b). Contrasted with the peptone-fed reactor (Reactor 3), the consortium fed acetate (Reactor 4) synthesized significantly more PHA, and also maintained significantly higher PHA reserves (Fig. 2a vs. Fig. 1a). Overall, in regard to the cycling of phosphorous, PHA, and sCOD, the consortium in Reactor 4 performed largely according to EBPR theory, and an appreciable quantity of phosphorous was removed from solution (ca. 97%). The net quantity of phosphorous removed from bulk solution corresponded to the net increase in biomass phosphorous, and the MSR, reflected in the biomass concentration of ppGpp, again cycled inversely to the soluble phosphorus concentration (Fig. 2c and d). Unlike the microbial consortium in Reactor 3, the microbial consortium in Reactor 4 experienced a nearly immediate limitation in inorganic nitrogen when the oxic period commenced. However, this limitation did not induce as much stress, reflected in the biomass concentrations of ppGpp, in Reactor 4 as Reactor 3 (Fig. 2d vs. Fig. 1d). Finally, the EC level again reflected the different energetic potentials within respective

5042

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 3 8 e5 0 4 6

Fig. 2 e Concentrations of (a) sCOD and PHA, (b) ammonia and nitrate, (c) aqueous and biomass phosphorus, (d) ppGpp and EC over the course of one feed-cycle in the acetate-fed anoxic/oxic SBR (Reactor 4; results shown for samples collected after 1-year of SBR operation).

anoxic and oxic environments: the EC level decreased from 0.84 to 0.69 over the anoxic period, and subsequently increasing to a value of 0.82 by the end of the oxic period (Fig. 2d).

3.3.

Performance of fully oxic SBRs

As a contrast to the cyclically operated anoxic/oxic Reactors 3 and 4, Reactors 1 and 2 were operated under fully oxic conditions. In Reactor 1 (peptone fed), the microbial consortium exhibited a feast response (i.e., rapid sCOD depletion concurrent with PHA synthesis) immediately after receiving fresh substrate; this response is consistent with feast-famine PHA synthesis (Dionisi et al., 2004). The feast period, which occurred within the first 10 min of the cycle, was characterized by: (i) a rapid decline in sCOD (Fig. 3a); (ii) an increase in biomass PHA (Fig. 3a); and (iii) an increase in NH3eN (likely from ammonification of the peptone) and a statistically insignificant decrease in NO 3  N (Fig. 3b). During the famine period (herein defined as occurring immediately following peak PHA synthesis): (i) biomass PHA was slowly consumed (Fig. 3a); (ii) aqueous phosphorus was incorporated into

biomass (Fig. 3c); and (iii) the concentration of NH3eN decreased with a concurrent increase in NO 3  N (indicative of nitrification) (Fig. 3b). PHA degradation and residual sCOD utilization occurred concurrently in the famine period, suggesting that either some of the sCOD could not be converted into PHA or non-PHA synthesizing microorganisms were also present in the consortium; sCOD was depleted within the first 3 h of the cycle, while some residual PHA remained at the end of the cycle. Overall, the microbial consortium in Reactor 1 did not perform according to EBPR theory, yet an appreciable quantity of phosphorous was removed from solution (ca. 67%). The MSR, reflected in the biomass concentration of ppGpp, cycled inversely to the soluble phosphorous concentration (Fig. 3c and d). Based on a mass balance analysis, the net quantity of phosphorous removed from bulk solution corresponded to the net increase in biomass phosphorous. The microbial EC increased from a value of 0.51e0.62 (Fig. 3d), a reflection of the energetic potential of the oxic environment. The microbial consortium in Reactor 2, which was operated identical to Reactor 1 but was fed acetate instead of peptone, generally cycled nutrients similarly to Reactor 1. The feast period, which lasted approximately the first 5 min of the

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 3 8 e5 0 4 6

5043

Fig. 3 e Concentrations of (a) sCOD and PHA, (b) ammonia and nitrate, (c) aqueous and biomass phosphorus, and (d) ppGpp and EC over the course of one feed-cycle in the peptone fed fully oxic SBR (Reactor 1; results shown for samples collected after 1-year of SBR operation).

cycle, was characterized by: (i) a rapid decline in sCOD (Fig. 4a); (ii) an increase in biomass PHA (Fig. 4a); (iii) a release of biomass phosphorus into solution (Fig. 4c); (iv) a rapid decrease in NH3eN (Fig. 4b); and (v) a non-detectable concentration of NO 3 -N (Fig. 4b). In contrast to Reactor 1, measurable phosphorus release was observed in Reactor 2 at the beginning of the feed-cycle (Fig. 4c). During the subsequent famine period: (i) biomass PHA was consumed (Fig. 4a); (ii) aqueous phosphorus was incorporated into biomass (Fig. 4c); and (iii) the concentration of NO 3  N decreased below the detection limit, ultimately resulting in a nitrogen limitation (non-detectable concentrations of both NH3eN and NO 3  N) (Fig. 4b). Unlike Reactor 1, the microbial consortium in Reactor 2 generally performed in accordance with EBPR theory (in terms of the cycling of phosphorous, PHA, and sCOD), despite the lack of anaerobic/oxic cycling, and an appreciable quantity of phosphorous was removed from solution (ca. 99%). The MSR, reflected in the biomass concentration of ppGpp, cycled inversely to the soluble phosphorous concentration (Fig. 4c and d). Based on a mass balance analysis, the net quantity of phosphorous removed from bulk solution corresponded to the net increase in

biomass phosphorous. The microbial EC increased from a value of 0.73e0.78 (Fig. 4d), likely a reflection of the energetic potential of an oxic environment.

3.4.

Statistical correlation of selected reactor parameters

The MSR, reflected in the biomass ppGpp concentration, was correlated statistically to the biomass concentration of phosphorous in all four of the reactors: for each individual reactor, the Pearson’s Product Moment Correlation Coefficient had a p-value less than 0.05 (Table 1). Additionally, the biomass ppGpp concentration was inversely correlated statistically ( p-values less than 0.05) to the soluble concentration of phosphorous in each of the four reactors (Table 1). The biomass PHA concentration was inversely correlated statistically ( p-values less than 0.05) to the biomass ppGpp concentration and the biomass phosphorous concentration in Reactors 2 and 4 (Table 1). The biomass PHA concentration was not inversely correlated statistically ( p-values greater than 0.05) to either the biomass ppGpp concentration or the biomass phosphorous concentration in Reactors 1 and 3 (Table 1).

5044

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 3 8 e5 0 4 6

Fig. 4 e Concentrations of (a) sCOD and PHA, (b) ammonia and nitrate, (c) aqueous and biomass phosphorus, and (d) ppGpp and EC over the course of one feed-cycle in the acetate fed fully oxic SBR (Reactor 2; results shown for samples collected after 1-year of SBR operation).

4.

Discussion

4.1. Excess intracellular accumulation of phosphorous and the MSR The MSR, reflected in the biomass ppGpp concentration, was correlated statistically with biomass concentration of phosphorous. Additionally, because the statistical analysis was performed with matched pairs of data collected at distinct time points over the course of a feed-cycle, the biomass ppGpp concentration was correlated to the intracellular cycling of phosphorous in reactors that cycled phosphorous (Reactors 2, 3, and 4). The microbial consortia in all four reactors removed an appreciable quantity of phosphorous from solution (ca. 67, 99, 99, and 97% in Reactors 1, 2, 3, and 4, respectively). The reactors that cycled phosphorous (Reactors 2, 3, and 4) removed the largest quantity from solution. The net quantity of phosphorous removed from solution corresponded to the net increase in biomass phosphorous. Excess intracellular accumulation of phosphorous is commonly defined as an

increase in intracellular polyP (Comeau et al., 1986; Mino et al., 1987). Given that intracellular polyP was not specifically assayed in this study, excess intracellular accumulation of phosphorous was defined as a net increase in biomass phosphorous. Hence, given that the MSR was correlated with biomass phosphorous, and there was a net increase in biomass phosphorous over a given feed-cycle, the MSR was associated with excess intracellular accumulation of phosphorous in mixed microbial consortia fed synthetic wastewater. For the reactors subjected to anoxic/oxic cycling (Reactors 3 and 4; Figs. 1 and 2), the MSR was likely induced individually or by a combination of the following factors: the ultimate lack of a terminal electron acceptor during the anoxic period; nutritional stress generated by the feast-famine cycling of organic carbon over the course of a given cycle; ultimate limitation of ammonia-nitrogen; and/or a cyclical limitation in inorganic phosphorus. For the fully oxic reactors fed peptone (Reactor 1; Fig. 3) and acetate (Reactor 2, Fig. 4), the MSR appeared to be triggered by either ammonia depletion or the feast-famine cycling of carbon over the course of a given feed-cycle.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 3 8 e5 0 4 6

Table 1 e Statistical correlation of selected parameters within each reactor. Comparison

ra

Soluble phosphorous vs. biomass ppGpp Reactor 1 0.93 Reactor 2 0.94 Reactor 3 0.88 Reactor 4 0.78 Biomass phosphorous vs. biomass ppGpp Reactor 1 0.90 Reactor 2 0.91 Reactor 3 3 0.85 Reactor 4 4 0.81 Biomass PHA vs. biomass ppGpp Reactor 1 0.61 Reactor 2 0.90 Reactor 3 0.38 Reactor 4 0.85 Biomass PHA vs. biomass phosphorous Reactor 1 0.48 Reactor 2 0.99 Reactor 3 0.72 Reactor 4 0.89

p-value 0.0022 0.0045 0.0087 0.038 0.0059 0.011 0.032 0.028 0.14 0.015 0.40 0.015 0.27 <0.001 0.10 0.0075

a Pearson’s Product Moment Correlation Coefficient. Seven matched pairs of data obtained from seven sampling events performed at distinct times over a given feed-cycle were used in generating each value of r.

4.2.

PHA and the MSR

The microbial consortia in all four reactors cycled PHA (defined as a temporal increase and then decrease in biomass PHA) over the feed-cycle. In the anoxic/oxic reactors (Reactors 3 and 4), the peak concentration of biomass PHA occurred shortly after the transition from the anoxic to oxic environment. In the fully oxic reactors (Reactors 1 and 2), the peak concentration of biomass PHA occurred within the first 10 min after introduction of fresh feed. The net yield, defined as the net increase in biomass PHA concentration from the start of a feed-cycle divided by the initial carbon substrate concentration, was 10, 5, 11, and 41% for Reactors 1, 2, 3, and 4, respectively. The highest yield was obtained in the anoxic/oxic reactor fed acetate. The intracellular cycling of PHA in mixed consortia fed peptone is not nearly as well documented in the literature as the intracellular cycling of PHA in mixed consortia fed acetate. Orhon et al. (2009) demonstrated that a mixed microbial consortium fed peptone is capable of cycling intracellular PHA in a feast-famine environment. Selected amino acids in peptone likely enter the PHA biosynthetic pathways through metabolic conversion into propionyl-CoA (Steinbuchel and Lutke-Eversloh, 2003). The propionyl-CoA is then converted to R-3-hydroxyvaleryl-CoA with b-ketothiolases and acetoacytylCoA reductases (Steinbuchel and Fuchtenbusch, 1998). The R-3hydroxyvaleryl-CoA serves as the substrate for PHA synthase. Acetate, in contrast, enters the PHA biosynthetic pathways through metabolic conversion into acetyl-CoA (Martin et al., 2006). The acetyl-CoA is then converted to R-3hydroxybutyric-CoA with b-ketothiolases and acetoacytylCoA reductases (Tsuge, 2002). The R-3-hydroxybutyric-CoA serves as the substrate for PHA synthase. In this study, the

5045

statistical correlation between biomass ppGpp and biomass PHA in the reactors fed acetate (Reactors 2 and 4) suggests that the MSR may have a regulatory role in the metabolisms associated with the conversion of acetate to PHA. The lack of a statistical correlation between biomass ppGpp and biomass PHA in the reactors fed peptone may suggest that the MSR does not have a regulatory role in the metabolisms associated with the conversion of amino acids to PHA. Additional research would be needed to establish a causal linkage between the MSR and intracellular cycling of PHA in PAOs.

5.

Conclusions

The MSR was associated with excess intracellular accumulation of phosphorous in mixed microbial consortia fed synthetic wastewater. With recognition of the potential role of the MSR in the removal of soluble phosphorous from wastewater, additional research may lead to further optimization of treatment technologies and the development of new treatment systems for the biological removal of phosphorus from wastewater.

Acknowledgments This material is based upon work supported by the National Science Foundation under Grant Number 0607329. Any opinions, findings, and conclusions expressed in this material are those of the authors and do not necessarily reflect the views of the funding agency.

references

American Public Health Association A.W.W, 2005. Standard Methods for Examination of Water and Wastewater, 21 ed. A Water Environment Federation, Washington, DC, USA. Ault-Riche´, D., Cresson, D., Tzeng, C., Kornberg, A., 1998. Novel assay reveals multiple pathways regulating stress-induced accumulations of inorganic polyphosphate in Escherichia coli. Journal of Bacteriology 180, 1841e1847. Braunegg, G., Sonnleitner, B., Lafferty, R.M., 1978. Rapid gas chromatographic method for the determination of poly-betahydroxybutyric acid in microbial biomass. European Journal of Applied Microbiology 6, 29e37. Chapman, A.G., Fall, L., Atkinson, D.E., 1971. Adenylate energy charge in Escherichia coli during growth and starvation. Journal of Bacteriology 108, 1072e1086. Chatterji, D., Ojha, A., 2001. Revisiting the stringent response, ppGpp and starvation signaling. Current Opinions in Microbiology 4, 160e165. Coats, E.R., Watkins, D.L., Kranenburg, D., 2011. A comprative environmental life cycle analysis for removing phosphorus from wastewater: biological versus physical/chemical processes. Water Environment Research 83, 750e760. Comeau, Y., Hall, K.J., Hancock, R.E.W., Oldham, W.K., 1986. Biochemical-model for enhanced biological phosphorus removal. Water Research 20, 1511e1521. Dionisi, D., Renzi, V., Majone, M., Beccari, M., Ramadori, R., 2004. Storage of substrate mixtures by activated sludges under dynamic conditions in anoxic or aerobic environments. Water Research 38, 2196e2206.

5046

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 3 8 e5 0 4 6

Glass, T.L., Holmes, W.M., Hylemon, P.B., Stellwag, E.J., 1979. Synthesis of guanosine tetra- and pentaphosphates by the obligately anaerobic bacterium Bacteroides Thetaiotaomicron in response to molecular oxygen. Journal of Bacteriology 137, 956e962. Goel, R.K., Noguera, D.R., 2006. Evaluation of sludge yield and phosphorus removal in a cannibal solids reduction process. Journal of Environmental Engineering 132, 1331e1337. Heathwaite, L., Sharpley, A., 1999. Evaluating measures to control the impact of agricultural phosphorus on water quality. Water Science Technology 39, 149e155. Kuroda, A., Murphy, H., Cashel, M., Kornberg, A., 1997. Guanosine tetra- and pentaphosphate promote accumulation of inorganic polyphosphate in Escherichia coli. Journal of Biological Chemistry 272, 21240e21243. Martin, H.G., Ivanova, N., Kunin, N., Warnecke, F., Barry, K.W., McHardy, A.C., Yeates, C., He, S., Salamov, A.A., Szeto, E., Dalin, E., Putnam, N.H., Shapiro, H.J., Pangilinan, J.L., Rigoutsos, I., Kyrpides, N.C., Blackall, L.L., McMahon, K.D., Hugenholtz, P., 2006. Metagenomic analysis of two enhanced biological phosphorus removal (EBPR) sludge communities. Nature Biotechnology 24, 1263e1269. Mino, T., Van Loosdrecht, M.C.M., Heijnen, J.J., 1987. Effect of phosphorus accumulation on acetate metabolism in the biological phosphorus removal process. In: Ramadori, R. (Ed.), Biological Phosphate Removal from Wastewaters. Pergamon Press, Oxford, pp. 27e38. Mouery, K.B.A., Rader, B.A., Gaynor, E.C., Guillemin, K., 2006. The stringent response is required for Helicobacter pylori survival of stationary phase exposure to acid and aerobic shock. Journal of Bacteriology 188, 5494e5500. Neubauer, P., Ahman, M., Tornkvist, M., Larsson, G., Enfors, S.O., 1995. Response of guanosine tetraphosphate to glucose fluctuations in fed-batch cultivations of Escherichia coli. Journal of Biotechnology 43, 195e204.

Orhon, D., Cokgor, E.U., Insel, G., Karahan, O., Katipoglu, T., 2009. Validity of monod kinetics at different sludge ages - peptone biodegradation under aerobic conditions. Bioresource Technology 100, 5678e5686. Potrykus, K., Cashel, M., 2008. (p)ppGpp still magical? Annual Reviews of Microbiology 62, 35e51. Powers, S.E., 2007. Nutrient loads to surface water from row crop production. International Journal of Lifecycle Analysis 12, 399e407. Pretty, J.N., Mason, C.F., Nedwell, D.B., Hine, R.E., Leak, S., Dils, R., 2003. Environmental costs of freshwater eutrophication in England and Wales. Environmental Science and Technology 37, 201e208. Rao, N.N., Gomez-Garcia, M.R., Kornberg, A., 2009. Inorganic polyphosphate: essential for growth and survival. Annual Reviews of Biochemistry 78, 605e647. Seviour, R.J., Mino, T., Onuki, M., 2007. The microbiology of biological phosphorus removal in activated sludge systems. FEMS Microbiology Reviews 27, 99e127. Steinbuchel, A., Fuchtenbusch, B., 1998. Bacterial and other Biological Systems for Polyester Production. Tibtech. October 16, 419e427. Steinbuchel, A., Lutke-Eversloh, T., 2003. Metabolic engineering and pathway construction for biotechnological production of relevant polyhydroxyalkanoates in microorganisms. Biochemical Engineering Journal 16, 81e96. Tsuge, T., 2002. Metabolic improvements and use of inexpensive carbon sources in microbial production of polyhydroxyalkanoates. Journal of Bioscience and Bioengineering 94, 579e584. Wells, D.H., Gaynor, E.C., 2006. Helicobacter pylori initiates the stringent response upon nutrient and pH downshift. Journal of Bacteriology 188, 3726e3729.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 4 7 e5 0 5 7

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Iodinated contrast media oxidation by nonthermal plasma: The role of iodine as a tracer Shirra Gur-Reznik, Sara P. Azerrad, Yana Levinson, Lilly Heller-Grossman, Carlos G. Dosoretz* Faculty of Civil & Environmental Engineering and Grand Water Research Institute, Technion-Israel Institute of Technology, Haifa 32000, Israel

article info

abstract

Article history:

The oxidation of trace pharmaceutical compounds in wastewater desalination streams by

Received 24 March 2011

nonthermal plasma (NTP) was evaluated. Brines from a two stage-RO pilot plant process as

Received in revised form

well as two sources of tertiary effluents, ultrafiltrated secondary effluents and membrane

3 July 2011

biological reactor effluents, were comparatively tested with ultra-pure water. The non-

Accepted 4 July 2011

ionic and ionic iodinated contrast media (ICM) compounds, iopromide (IOPr) and dia-

Available online 13 July 2011

trizoate (DTZ), respectively, were used as model compounds. The neurostabilizer drug carbamazepine (CBZ) was used for reference purposes. Based on deiodination profiles, two

Keywords:

distinct patterns of initial oxidation could be established for the ICM. The time profile of

Nonthermal plasma oxidation

deiodination and transformation paralleled for DTZ, indicating that transformation of the

Iodinated contrast media

aromatic ring is the main initial pattern of transformation. For IOPr, a considerable lag

Pharmaceutical active compounds

phase of deiodination was observed, suggesting that oxidation of the alkyl chains rather

Desalination brines

than ring oxidation is the main pattern of initial transformation. Although transformation

Tertiary effluents

rate of IOPr was higher compared to DTZ, the rate and degree of deiodination was higher

Trace contaminants

for DTZ than IOPr. Both ICM displayed a markedly lower susceptibility to NTP oxidation compared to CBZ. However, the kinetics of IOPr transformation seems to be less affected by the water matrixes, compared to DTZ and CBZ. Whereas NTP mediated oxidation of ICM followed first-order kinetics, a better fit to Harris model was found for CBZ. As a result of the NTP oxidation, treated brines and effluents displayed a substantial increase in biodegradability (measured as BOD). To conclude, NTP displayed a high potential for treating reluctant pharmaceuticals active compounds such as ICM, even at the background of relatively high DOC concentrations, as can be found in treated effluents and desalination brines, and with no need for chemical additives. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Continuous population growth, rising standards of living, industrialization and urbanization limit the freshwater available in arid and semi-arid regions. Wastewater reuse is

being increasingly emphasized as a strategy for conservation of limited resources of freshwater and as a mean of safeguarding the aquatic environment due to contaminants present in wastewater, including synthetic organic contaminants (SOCs) at nano and micro levels. In the case of Israel, in

* Corresponding author. Tel.: þ972 4 8294962; fax: þ972 4 8228898. E-mail address: [email protected] (C.G. Dosoretz). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.003

5048

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 4 7 e5 0 5 7

which approx. 80% of the treated effluents are reused, constituting roughly 50% of the total demand for agricultural irrigation (Inbar Committee, 2004), these issues are of special concern. Moreover, the increasing demand for freshwater will probably lead to greater incidences of indirect and direct water-reuse situations as the quality of the recycled water will rise and pollution of drinking water sources with SOCs will be of great concern in such situations (Oliver et al., 2005). Combining simplicity, versatility and continuous processing of aqueous streams, pressure driven-membrane separation systems, applying dense membranes such as reverse osmosis (RO) or nanofiltration (NF), are a promising generic treatment techniques for the reclamation of municipal wastewater effluents in general and specifically for treating SOCs (Comerton et al., 2008; Drewes et al., 2005; Gur-Reznik et al., 2011; Kimura et al., 2009). Over the recent decades the cost of membranes has decreased and fluxes have increased dramatically making membranes an attractive technological solution for water treatment. While purified water is obtained in a fast and continuous way with membrane filtration, micro and nanopollutants as well as other organic contaminants are concentrated in a relatively small volume in the brine stream. However, the cost of disposal or treatment of the resulting RO brine, which contains high concentration of salts, organics and biological components, constitutes a significant disadvantage of this technique (Van der Bruggen et al., 2003). Especially the elevated concentrations of organic micro and nanopollutants might possess a serious hazard with regard to potential (eco)-toxicological effect, if the brine is discharged directly into the aquatic environment. Thus, implementation of advanced oxidation technologies to reduce SOCs concentrations has been the subject of numerous studies (see reviews by Dalrymple et al. (2007); Malato (2008); Comninellis et al. (2008); Klavarioti et al. (2009)). These include several advanced oxidation processes (AOPs), aimed to generate hydroxyl radicals, ozonation and their combination. A relatively recent innovative advanced oxidation technology for water treatment includes among others nonthermal plasma (NTP) (Sunka et al., 1999). NTPs are usually created by electrical discharges between sets of electrodes (Grabowski, 2006). The NTP reactor that was used for this study utilizes high voltage and high current of short duration electrical pulses across sharpened edges of carbon fiber electrodes to form a corona discharge in the gas above the liquid surface (Even-Ezra et al., 2009), which is up to 10 times more efficient compared to corona generated within fluid (Grabowski, 2006). The corona discharge generates an efficient production of hydroxyl radicals, hydrogen peroxide, low ozone concentration and UV light with a spectral distribution similar to that of sunlight (from 200 to 1000 nm) (Gerrity et al., 2010; Locke et al., 2006; Pekarek, 2003). Moreover, these active species can be generated without the constant addition of costly chemicals or UV lamps, which require cleaning and are hindered by high turbidity and matrix absorbance (Gerrity et al., 2010). NTP is an attractive alternative for AOPs processes, since during the NTP process little energy is lost in heating the surrounding fluid and no additives are required. However, implementation of NTP technology for wastewater treatment is still an evolving technology and its application on brines has not been reported yet (Westerhoff et al., 2009; Zhou et al., 2011).

Even-Ezra et al. (2009) tested an experimental NTP system (Aquapure), similar to that employed in the present research, with tertiary effluent and contaminated groundwater in relation to N-nitrosodimethylamine (NDMA), trichloroethylene (TCE), methyl tert-butylether (MTBE), and 1,4-dioxane. They reported high removal efficiencies of over 90% at both laboratory and field scales and obtained energy yields falling within the range of common commercial AOPs systems. Gerrity et al. (2010) tested the degradation of seven PhACs and potential EDCs monitored in tertiary effluent (with dual media filtration as the tertiary step) and spiked surface water in batch and single-pass modes. Degradation rates were reported as a function of generator energy consumption. They also showed that the NTP by AquaPure has comparable energy requirements to more common AOPs. The present research evaluated the performance of NTP in combination with membrane desalination of wastewater effluents. Membrane filtration is a preferential platform to be combined with UV based AOPs since it enhances activity of oxidizing species by removing turbidity. In addition, filtration enhances the effectiveness of AOPs by lowering matrix consumption. Moreover, these water quality parameters are expected to be less significant in the context of water treatment plasma based technologies. The study focused on iopromide (IOPr) and diatrizoate (DTZ), two iodinated contrast media (ICM) compounds. Degradation and resistance of ICM to AOPs and ozonation has been a polemic issue, and therefore they appear to be excellent right-hand markers for process evaluation. The widely studied antiepileptic and mood stabilizer drug carbamazepine (CBZ) was used for reference purposes. Brines from a two-stage RO pilot plant process were comparatively tested with ultra-pure water (UPW), as well as two sources of tertiary effluents, ultrafiltrated secondary effluents and membrane biological reactor (MBR) effluents.

2.

Materials and methods

2.1.

Water matrixes

Ultra-pure water (UPW) was generated by distillation (Hamilton, GB) of RO-desalinated tap water followed by ion exchange/activated carbon adsorption and UV irradiation in a water purification system (Elga, England). MBR effluents were generated in ZW-10 reactor (Zenon) of 0.2 m3 capacity, equipped with a submerged UF membranes of approx. 0.04 mm pore size and 0.9 m2 filtration area, operating at the Technion Institute campus (Katz and Dosoretz, 2008). UF filtered secondary effluents and RO brines were generated at the Technion effluent desalination-pilot plant located at the wastewater treatment plant (WWTP) at Nir Ezyon (15 km south of Haifa, IL). The pilot plant comprises a four module-UF system (DOW) with MWCO of z80 kDa and 51 m2 filtration area followed by a two stages-RO unit (8” and 4” membranes with 7 m2 filtration area, Toray) operated at an average recovery rate of approx. 85%. Thus, two stages of RO brines were used for this research, the first stage is concentrated up to two folds, and the second up to five folds. The feedwater to the pilot system consisted of secondary effluent produced by

5049

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 4 7 e5 0 5 7

the Nir Ezyon WWTP. The quality parameters of the wastewater effluent and brines are presented in Table 1.

2.2.

Pharmaceuticals

Sodium diatrizoate, iopromide and carbamazepine were used as pharmaceutical model compounds. They were chosen since they are relatively omnipresent PhAC compounds representing two entirely different ends of stability to oxidation treatment. Moreover, they can be used as treatment indicators, especially the ionic ICM diatrizoate, due to their resistance to AOPs treatment. The compounds were purchased from SigmaeAldrich (Israel), and reported to be of high purity (at least 99%).

2.3.

Nonthermal plasma system

An NTP pilot-scale reactor developed by AquaPure Technologies, Ltd. (Upper Galilee, Israel) was employed (Fridman, 2010). Experiments were performed in the time frame of up to 6 h. The unit is essentially similar to that described by Even-Ezra et al. (2009) and Gerrity et al. (2010). In brief, it comprises the oxidation reactor, a high voltage generator, two electrodes: a metal one (ground electrode or anode) which constitutes the chamber floor and the second are carbon fibers (cathode) electrodes above, as well as an ozone injection unit and a recirculation tank (30 L). The reactor was operated in a batch mode at 200 L/h recycling flow. The water was recirculated within the thin channel formed beneath the electrodes. From the tank, the water is pumped through a Venturi-type injector at an air flow of 0.7 m3/h. The injector draws O3-enriched air (w0.1% by weight) from the reactor headspace to a static mixer, which then pumped back to the water tank. The plasma was generated above the water surface by high voltage/high frequency (up to 40 kV/500 Hz) of short electrical pulses

Table 1 e Initial water quality parameters for the different matrixesa. Matrix/Parameters pH EC [mS/cm] DOC [mg/L] A254 SUVA [L/mg$m] A280 BOD COD

Tertiary effluent MBR 7.5  0.6 UF 8.0  0.08 MBR 1.3  0.07 UF 1.4  0.05 MBR 7.3  0.2 UF 10.7  0.9 MBR 0.2  0.02 UF 0.3  0.03 MBR2.0  0.2 UF 2.2  0.4 MBR 0.1  0.01 UF 0.2  0.02 MBR 0.3  0.3 UF1.0  0.2 MBR 44  0.6 UF n.a

RO1 brines RO2 brines 7.6  0.3

7.7  0.4

3.1  0.1

5.6  0.3

20.0  3.7

49.1  2.5

0.6  0.1

1.5  0.2

2.9  0.2

3.0  0.2

0.5  0.1

1.3  0.2

1.7  0.9

1.9  0.3

79.0  9.0

137.0  38.0

a Values represent average  standard deviation of 3e7 replicates.

(nanoseconds) to generate a corona discharge. The ozone injector was activated and set to a water pressure of 2 bar. A spiking level of approx. 65  11 mg/L of each of the three model compounds was applied for all the matrixes. At the indicated times, samples of 8 ml (or 3 L for BOD and COD measurements, at the begging (before starting the corona) and at the end of an experiment) were taken for analyses. The resultant transformation of the mother molecules followed by LCMS/MS indicated disappearance/transformation of the model compounds.

2.4.

Analytical techniques

Chemical oxygen demand (COD) and biological oxygen demand (BOD) measurements were conducted according to Standard Methods (Eaton et al., 1995) 5210, 5220, respectively. pH and electrical conductivity (EC) were measured with Cyberscan Electrodes (PC 300 Series) (EUTECH Instruments, waterproof series). Dissolved organic carbon (DOC), defined as the total organic carbon (TOC) fraction which passes 0.45 mm filter, and ultraviolet absorbance (UV) at a wavelength of 254 nm (A254) and 280 nm (A280) were analyzed using a TOC analyzer (multi N/C 2000, Analytik Jena) and UVevisible spectrophotometer (Agilent 8453 series) with a 1 cm quartz cell, respectively. Specific UVA (SUVA), ratio of UVA to DOC, was calculated as well. PhACs were analyzed using liquid chromatography with electrospray tandem mass spectrometry (LC-MS3) according to Gur-Reznik et al. (2011). LCMS analyses ware performed on a system consisting in an Agilent 1200 HPLC (Hewellet Packard) system coupled with ion spray interface to an API 3200 (Applied Biosystems) triple quadrupole mass spectrometer. Electrospray ionization was used in a positive ion mode. The compounds were detected in the multiple-reaction monitoring (MRM) mode. A LiChroCART Purospher STAR RP-18 (Merck) endcapped column (4.6 mm  15 cm) with 5 mm pore size was used with a binary gradient performed according to Vanderford et al. (2003) with some custom modifications. (A) 0.1% formic acid (v/v) in water and (B) is 100% methanol. The gradient was as follows: 5% B was held for 2 min, than increased linearly to 100% during another 7 min and held for 3 min, afterward the gradient decreased linearly again to 5% and finally held for 8 min. The total run time was 15 min. The flow rate was 400 mL/min. An injection volume of 15 mL was used for CBZ, 20 mL for IOPr, and

Table 2 e Precursor and product ions used in LCMS analysis. Precursor ion [m/z] [M þ H]þ

Production [m/z]

Loss of

8.5

614.7

IOPr

7.94 þ 8.04

791.8

CBZ

10.5

237.1

361.2 233.1 300.0 572.8 194.1 193.1 192.0

2I HI, 2I 3HI,C3H9NO3 HI,C3H9NO2 CNHO CNH2O CNH3O

PhACs

Retention time [min]

DTZ

5050

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 4 7 e5 0 5 7

25 mL for DTZ. The quantification limits under these conditions were 0.1 mg/L (S/N > 6). Standard curves yielded goodness of fit (R2) greater than 0.99 within the experiments concentration range. In order to track efficiency of recoveries mainly due to matrix effects standard addition was performed randomly before LCMS analysis. Precursor and product ions used for the identification of CBZ and DTZ are given in Table 2. Since the recoveries were approximately 100%, matrix affect was negligible. Organic iodine was analyzed with a differential

adsorbable organic halogen (AOX) procedure followed by specific quantification of iodide, based on a previous report by Oleksy-Frenzel et al. (2000) with some custom modifications. Iodide was detected with ion chromatographic conductivity detector (881 compact IC pro, Metrohm), using Metrohm A supp 5-150 (with 4 mm pore size) column. 3.2 mM sodium carbonate and 1.0 mM sodium bicarbonate with 5% acetonitrile served as eluents. Organic iodide detection limit was 1.2 mg/L.

Fig. 1 e DTZ transformation (, and 3) and deiodination (C) profiles upon NTP oxidation for the different water matrixes. Dilution effect of RO1 brines on DTZ oxidation (1:2, ,; 1:5, 3) and deiodination (1:5, C) are also presented. Spiking concentration was 65 ± 11 mg/L. Values represent average ± standard deviation of 3 independent replicates.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 4 7 e5 0 5 7

5051

Fig. 2 e IOPr transformation (, and 3) and deiodination (C) profiles upon NTP oxidation for the different water matrixes. Spiking concentration was 69 ± 11 mg/L.

3.

Results and discussion

The performance of NTP was studied with four different water matrixes representing four different DOC and ionic strength levels (see Table 1). The time profiles of DTZ transformation and deiodination are presented in Fig. 1. It can be seen that

DTZ transformation is largely matrix dependent. Whereas a very high degree of DTZ disappearance from UPW was attained after 6 h (>99%), its transformation for tertiary effluents decreased down to approximately 70% as the DOC and EC levels increased (UF filtered secondary effluents or MBR effluents behaved similarly for all the PhACs tested), to w40% for RO1 brines and to w25% for RO2 brines. Interestingly,

Fig. 3 e CBZ transformation profiles upon NTP oxidation for the different water matrixes. Spiking concentration was 62 ± 11 mg/L. Values represent average ± standard deviation of 3 independent replicates.

5052

Table 3 e Comparative summary of DTZ and IOPr (ICM) and DTZ trasnformation treated with AOP/ozonation. AOP

Water matrix Initial concentration

Maximal transformation (%) DTZ

SE

2.1e5.7 mg/L

O3

SE Tap þ SE þ TE

1.8e5.4 mg/L 0.5e5.0 mg/L

Tap þ TE

100 mg/L and 10 mg/L

1 mg/L

SE Surface water UV irradiation þ Potassium UPW peroxydisulfate (KPS) UPW UV/TiO2 (P25 and Hombikat UV100) NTP TE

176e1600 ng/L

Several

Various

Various

NTP

UPW TE RO1 brines RO2 brines

65 mg/L

SE-secondary effluents TE-tertiary effluents.

60e400 mg/L 0.5e10 mg/L

w15e25% >80%

Reference

CBZ

5e15 mg/L O3 at pH ¼ 7.2, and a contact time of 18 min 50% 50% 95% 2e22 mg/L O3 at pH of effluent 0.5e5 mg/L O3 at pH ¼ 6.7, 7 and 7.3, Negligible w60% < 0.8 1/M$S 3$105 1/M$S and a contact time of 4.2 and 10 min e Undetectable e 10 mg/L after 30 min increased to 30 mg/L O3 (unknown pH), and a contact time of 60 min 50% 90% e 1e3 mg/L O3 at pH of SE, 7, 9, 12 and 20% 35%e55% a contact time of 2e10 min e w90% mineralization e 0e20 mM KPS at various pHs and a contact time of 80 min e w70% w90% 100 and 500 mg/L at pH ¼ 3.4e6.5 for approximately 3 min e e w90% frequency of 500 Hz, a voltage of 8.0 kV, and a recirculation rate of 8.0 L/min. Most ICM resulted refractory Relatively readily Various degradable >99% Undetectable Undetectable 5 min of 6 h (at pH of effluent) w70% >98% 40 min of 6 h (at pH of effluent) w40% w90% 2 h of 6 h (at pH of effluent) w25% e 2.5 h of 6 h (at pH of effluent) Undetectable

Ternes et al. (2003) Bahr et al. (2007) Huber et al. (2003, 2005) Putschew et al. (2007)

Seitz et al. (2008; 2006a; 2006b) Chan et al. (2010) Doll and Frimmel (2004) Gerrity et al. (2010) Ikehata et al. (2006) This work

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 4 7 e5 0 5 7

O3/UV/H2O2

IOPr

AOP features

5053

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 4 7 e5 0 5 7

Table 4 e First-order reaction rates (k) and half-life constants (t1/2).a PhACs/Matrix

DTZ 1

UPW Tertiary effluent RO1 brines RO2 brines

IOPr 2

1

CBZ 2

1

k [sec ]

t1/2 [sec]

R [%]

k [sec ]

t1/2 [sec]

R [%]

k [sec ]

t1/2 [sec]

R2 [%]

1.0  0.2 0.2  0.004 0.08  0.005 0.05

0.8  0.1 3.6  0.08 8.4  0.5 13.7

>99.2 >99.3 >97.9 97.5

1.2 0.6  0.1 0.4 0.3

0.6 1.2  0.3 1.8 2.0

99.4 <99.5 99.9 99.7

37  17.3 7.2  2.9 3.1  0.3 2.3

0.02  0.01 0.1  0.04 0.2  0.02 0.3

<73.8 <93.5 <86.7 94.7

a Values repesent average  standard deviation of 3 replicates.

the time profile of DTZ dehalogenation almost paralleled the pattern of DTZ transformation in all cases, suggesting that transformation of the aromatic ring is the main pattern of DTZ transformation. Dilution of the RO1 brines resulted in an almost proportional increase of the transformation rate and correspondent dehalogenation. The results for IOPr transformation and dehalogenation are presented in Fig. 2. Profiles and extent of transformation were considerable less affected by the matrix, achieving 98% for tertiary effluents and 90% for RO1 brines within 6 h compared to 100% transformation in less than 4 h for UPW. Strikingly, a considerable lag phase of deiodination was observed for IOPr, suggesting that dehalogenation is not involved in the initial transformation during NTP treatment. These findings can be correlated with reported ones regarding higher susceptibilities to oxidative transformation during ozonation of IOPr vs. DTZ (Bahr et al., 2007; Seitz et al., 2008; Ternes et al., 2003), most probably due to high lability of the aliphatic side chains of IOPr. Our results are supported by Putschew et al. (2007), who found that ozonation results in a rapid transformation of iopromide whereas the decrease of organic bound iodine was much lower. Seitz et al. (2008) found two major oxidation by-products of iomeprol during the initial stage of surface water ozonation. Aldehydes and carbonyl moieties are major functional groups in oxidation by-products of iomeprol which has a chemical structure very similar to IOPr. Accordingly, during the oxidative treatment at least one of these hydroxyl groups can be rapidly oxidized to form an aldehyde or ketone-like oxidation by-product. In addition, no significant mineralization of 10 ppm iomeprol could be achieved at pH 7 and 9, whereas at pH 12 approximately 40% of the initial iomeprol concentration was mineralized after 20 min of contact time with ozone (at 3 mg/L O3 dose). In contrast, Ning and Graham (2008) reported for ozone reactions with four ICM compounds (diatrizoate, iomeprol, iopromide, and iopamidol), the concomitant release of inorganic iodine and transformation, suggesting cleavage of the iodineecarbon bond on the aromatic ring. The proportion of iodine release was similar among the non-ionic ICM compounds but much greater for diatrizoate. Both of these reports demonstrated that ICMs are mainly prone to hydroxyl radicals during ozonation. Chan et al. (2010) examined the behavior of IOPr in UPW at the presence of photo-activated potassium peroxydisulfate. They reported complete transformation of 100 ppm IOPr after 30 min, and near-complete mineralization within 80 min. However, little mineralization was achieved during the first 10 min of the reaction, when the corresponding degradation of IOPr was substantial. A recent study by Jeong

et al. (2010) depicted deiodination as a reduction pathway of DTZ by g-irradiation in UPW, while the side chains of DTZ remained intact. Deiodination seems to be also the main route in the photo degradation pattern of IOPr. Though compared to the ionic ICM, DTZ, the non-ionic ICM, such as IOPr, involve also transformation of the side chains to form ketones. Fig. 3 presents the NTP oxidation of CBZ in the background of UPW, MBR effluents and RO1 brines, in which 100% transformation was achieved in approx. 5, 40 and 60 min, respectively. These results clearly depict that CBZ is much more susceptible to oxidation than ICM. Interestingly, as in the case of DTZ, a very strong influence of the water matrix on the time profile of transformation was evident. This behavior might be related to the initial transformation of the rings in both cases. The relative extent and rate of transformation of CBZ vs. ICM found here for NTP oxidation is in line with previous reports for other AOPs. A comparative summary of previous results in relation to ours is presented in Table 3, with special reference to ozonation, which was the most widely oxidation technology studied for ICM. Although scarce data was found for DTZ, it appears to be reluctant to transformation by almost all means tested. In addition, whereas all techniques were found effective for CBZ regardless of the background tested, a divergent effect was gathered for IOPr. Evaluation of the kinetic parameters of the transformation of DTZ, IOPr and CBZ for all the four water matrixes is presented in Table 4. NTP degradation of both ICM as well as CBZ followed first-order kinetics for all water matrixes, although ICM depicted a higher degree of goodness of fit (R2) than CBZ. Not surprisingly, the kinetics of transformation decreased when background DOC levels increased; with a difference of one order of magnitude between the different matrixes. As mentioned above, it seems that IOPr transformation kinetics is less affected by the background of the water matrix compared to DTZ and CBZ.

Table 5 e Harris model constants for carbamazepine. Matrix

UPW Tertiary effluents RO1 brines RO2 brines

CBZo [ppb] 69 a 0.014 b 2894 c 3.7 0.2 DOCo [ppm] 99.1 R2 [%]

47 0.021 9 3.5 6.4 99.9

73 0.014 0.60 2.6 7.1 99.5

73 0.014 0.08 2.5 23 99.7

344 0.003 0.01 1.9 51 99.4

CBZo and DOCo are the initial carbamazepine and total dissolved organic carbon concentrations, respectively.

5054

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 4 7 e5 0 5 7

The kinetics of CBZ degradation displayed a better fit (R2>99%) to for the Harris model (Eq. (1)), indicating that a pseudo-second order transformation takes place. Ce ¼

1 a þ btc

(1)

Where Ce is the concentration with time, t, after treatment; and a, b, c are constants. Table 5 details the different constants found for CBZ and the four water matrixes tested in this study. The different columns for the same matrix are independent repetitions for two different concentrations. From Eq. (1) and Table 5 it is clear that constants a and c are related to the both initial CBZ and DOC concentrations while the constant b seems to be principally influenced by the initial DOC concentration. Second-order kinetic rate constants for the reactions of CBZ and ICM compounds with ozonation were reported in other cases (Bahr et al., 2007; Huber et al., 2003; Ning and Graham, 2008). In contrast, the kinetics of NTP degradation of a series of SOCs for the same NTP systems as applied in this research was reported to be of first-order rate. Indeed, EvenEzra et al. (2009) tested the NTP system with tertiary effluent and contaminated groundwater in relation to NDMA, TCE, MTBE, and 1,4-dioxane. Gerrity et al. (2010) examined the degradation of meprobamate, dilantin, primidone, carbamazepine, atenolol, trimethoprim and atrazine from tertiary effluents (with dual media filtration as the tertiary step). They reported pseudo first-order degradation rate constants for these seven compounds in m3/kWh, with carbamazepine and

trimethoprim as the most degradable PhACs; while meprobamate, primidone, and atrazine were the most recalcitrant compounds. They saw consistent relative degradation rates between three different configurations tested. Based on this relative oxidation rate, the use of indicator compounds for evaluation of SOCs treatment has been suggested. In this sense, our results strongly support the use of ICM as such indicators, especially the ionic ICM such as DTZ. The rationale for this selection is: (i) They represent a family of haloaromatic compounds containing iodine which can be used as a specific endogenous labeling for tracking the compounds (by both specific iodine detection with AOI and its specific mass in LCMS analysis). (ii) They also comprise a broad range of chemical interactions, ranging from negative (hydrophilic) to fairly positive (hydrophobic) compounds. (iii) Based on their clinical application they are stable to biological metabolism and therefore released into the water system as intact pharmaceutical compounds (>80% of the ingested dose within the first 24 h according to Haiß and Ku¨mmerer (2006)). (iv) Thus, as opposed to other drugs which are secreted as transformation adducts, iodinated X-ray contrast media represent authentic model compounds. (v) Their varied relative resistance to the various AOPs (non-ionic vs. ionic). The fate of the background DOC during NTP treatment, i.e., mineralization, is presented in Fig. 4. The rate of DOC mineralization was in the order of 25% for the tertiary effluents and approx. 15% and 10% for the RO1 and RO2 brines, respectively. Compared to mineralization, a high SUVA

Fig. 4 e Fate of dissolved organic matter background upon NTP treatment. Upper panel: dissolved organic carbon (DOC) mineralization profiles; middle panel: SUVA reduction profiles; lower panel: change in absorbance at 280 nm (A280). Values represent average ± standard deviation of 3e5 independent replicates.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 4 7 e5 0 5 7

5055

Fig. 5 e Changes in the background organic matter upon NTP oxidation. Upper panels: BOD values; lower panels: COD values. The left column shows initial values before treatment and the right column the results following NTP treatment. Bars represent average ± standard deviation of up to five independent replicates. reduction was observed: 80  0.2%, 61  6% and 48  9% for tertiary effluents, RO1 and RO2 brines, respectively, depicting a considerable reduction of the total dissolved organic carbon aromaticity (Jarusutthirak et al., 2002); Gerrity et al. (2010) and Westerhoff et al. (2009) reported that the reduction in UVA254 exceeded the DOC removal. Since SUVA is the ratio of UVA at a wavelength of 254 nm to DOC, this observation is supported by the results presented here as well. UVA254 indicates favorable oxidation of unsaturated carbon bonds (Westerhoff et al., 2009). Moreover, a very high reduction in 280 nm wavelength absorption; 90  10%, 73  5% and 59  3% for tertiary effluents, RO1 and RO2 brines, respectively, was also observed. This most probably implies that transformation of the proteinic and phenolic DOC groups took place. Complete oxidation of all organics to carbon dioxide would be the highest level of treatment (Westerhoff et al., 2009), since the result is conversion to innocuous materials with no secondary disposal needs (Sunka et al., 1999). However, it would require a very high cost of treatment as far as energies and chemicals are concerned. Instead, an overall increase in digestibility of the remaining organic matter will enable biological treatment prior to disposal or the recycling of the end of the pipe streams back to the WWTP. We measured BOD of the remaining fractions following NTP oxidation to estimate the change in digestibility. Fig. 5 shows the results of BOD and COD measurements. As can be seen, a considerable increase in BOD, between 4 and 6 folds, was observed at the end of the experiments, following the NTP treatment. COD values decreased by 10e15 after the NTP treatment, as expected after oxidation. The relatively high variability of the results is probably mainly due to the low values measured and in part to the daily variability of the effluents among the different samples.

4.

Conclusions

Two patterns were found for NTP oxidation of DTZ and IOPr. The time profile of DTZ deiodination and transformation

paralleled, indicating that transformation of the aromatic ring is the main pattern of transformation. For IOPr, a considerable lag phase of deiodination was observed, suggesting that transformation of alkyl chains rather than ring oxidation is the main pattern of initial transformation. The rate and degree of deiodination was higher for DTZ than IOPr, which is inversely correlated to de-alkylation of the side chains. These two patterns may represent the two possible ICM transformation pathways achievable upon different oxidation treatments, although, the time sequence of these pathways might change for the different AOPs. The first one involves direct oxidation of the aromatic ring and deiodination, and the second is oxidation of the side chains preceding ring transformation. Both can lead to transformation by-products, either individually or in integrated ways. AOI was used here as a simple way to establish and differentiate between these two mechanisms. Both ICM displayed a markedly lower susceptibility to NTP oxidation compared to CBZ. However, the kinetics of IOPr transformation seems to be less affected by the water matrixes compared to DTZ and CBZ. NTP mediated oxidation of ICM followed first-order kinetics, and for CBZ a better fit for Harris model was found, suggesting a pseudo-second order kinetics. The rate and extent of transformation of the model compounds decreased as the background DOC levels increased, with a difference of one order of magnitude between the different water matrixes. In addition, a very high SUVA reduction and even a higher reduction in absorbance at 280 nm was observed, implying that transformation of the aromatic moieties and unsaturated carbon bonds, among others, takes place. As a result of the NTP oxidation, treated brines and effluents displayed a substantial increase in biodegradability (measured as BOD). Based on all the stated above, their relative resistance to oxidations processes and the iodine tracing, ICM appear as excellent right-hand markers for process evaluation and environmental regulation. In addition, they represent authentic markers since they are mostly (>80%) released into the water system as intact pharmaceutical compounds and

5056

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 4 7 e5 0 5 7

display relative high resistance to biological transformation, even at WWTP conditions. To conclude, the NTP treatment presented in this work displayed a high potential for treating reluctant PhACs, even at the background of relatively high DOC concentrations, as can be found in treated effluents and desalination brines, without the need for chemical additives as required in other AOPs.

Acknowledgments This work was funded by the Rieger foundation, the Grand Water Research Institute - Zakin award and the Joint German-Israeli Research Program BMBF-MOST (Contract No. WT 0903/2194).

references

Bahr, C., Schumacher, J., Ernst, M.F., Luck, F., Heinzmann, B., Jekel, M., 2007. SUVA as control parameter for the effective ozonation of organic pollutants in secondary effluent. Water Sci. Technol. 55, 267e274. Chan, T.W., Graham, N.J.D., Chu, W., 2010. Degradation of iopromide by combined UV irradiation and peroxydisulfate. J. Hazard. Mater. 181, 508e513. Comerton, M.A., Andrews, C.R., Bagley, M.D., Hao, C., 2008. The rejection of endocrine disrupting and pharmaceutically active compounds by NF and RO membranes as a function of compound and water matrix properties. J. Mem. Sci. 313, 323e335. Comninellis, C., Kapalka, A., Malato, S., Parsons, S.A., Poulios, I., Mantzavinos, D., 2008. Advanced oxidation processes for water treatment: advances and trends for R&D. J. Chem. Technol. Biotechnol. 83, 769e776. Dalrymple, O.K., Yeh, D.H., Trotz, M.A., 2007. Removing pharmaceuticals and endocrine-disrupting compounds from wastewater by photocatalysis. J. Chem. Technol. Biotechnol. 82, 121e134. Doll, T.E., Frimmel, F.H., 2004. Fate of pharmaceuticals-photodegrad ation by simulated solar UV light. Chemosphere 52, 1757e1769. Drewes, J.E., Bellona, C., Oedekoven, M., Xu, P., Kim, T.U., Amy, G., 2005. Rejection of wastewater-derived micropollutants in high-pressure membrane applications leading to indirect potable reuse. Environ. Prog. 24, 400e409. Eaton, A.D., Clesceri, L.S., Greenberg, A.E., 1995. Standard Methods for the Examination of Water and Wastewater. American Public Health Association, Washington, DC. Even-Ezra, I., Mizrahi, A., Gerrity, D., Snyder, S., Salveson, A., Lahav, O., 2009. Application of a novel plasma-based advanced oxidation process for efficient and cost-effective destruction of refractory organics in tertiary effluents and contaminated groundwater. Desal. Water Treat. 11, 236e244. Fridman, N., 2010 Minimization of bromate formation as part of the operation of a non-thermal-plasma method for oxidation of refractory organic compounds. MSc. thesis, Technion_Israel Institue of Technology, Israel. Gerrity, D., Stanford, B.D., Trenholm, R.A., Snyder, S.A., 2010. An evaluation of a pilot-scale nonthermal plasma advanced oxidation process for trace organic compound degradation. Water Res. 44, 493e504. Grabowski, L.R. (2006) Pulsed Corona in air for water treatment. PhD thesis, Eindhoven University of Technology, Netherlands. Gur-Reznik, S., Koren Menashe, I., Heller-Grossman, L., Rufel, O., Dosoretz, C.G., 2011. Influence of seasonal and operating conditions on the rejection of pharmaceutical active

compounds by RO and NF membranes. Desalination 277, 250e256. Haiß, A., Ku¨mmerer, K., 2006. Biodegradability of the X-ray contrast compound diatrizoic acid, identification of aerobic degradation products and effects against sewage sludge micro-organisms. Chemosphere 62, 294e302. Huber, M.M., Canonica, S., Park, G.-Y., von Gunten, U., 2003. Oxidation of pharmaceuticals during ozonation and advanced oxidation processes. Environ. Sci. Technol. 37, 1016e1024. Huber, M.M., Go¨bel, A., Joss, A., Herrmann, N., Lo¨ffler, D., McArdell, C.S., Ried, A., Siegrist, H., Ternes, T.A., van Gunten, U., 2005. Oxidation of pharmaceuticals during ozonation of municipal wastewater effluents: a pilot study. Env. Sci. Technol. 39, 4290e4299. Ikehata, K., Naghashkar, N.J., El-Din, M.G., 2006. Degradation of aqueous pharmaceuticals by ozonation and advanced oxidation processes: a review. Ozone: Sci. Eng. 28, 353e414. Inbar Committee, 2004. Upgraded Effluent Standards. Ministry of Environmental Protection, Israel. Jarusutthirak, C., Amy, G., Croue´, J.-P., 2002. Fouling characteristics of wastewater effluent organic matter (EfOM) isolates on NF and UF membranes. Desalination 145, 247e255. Jeong, J., Jung, J., Cooper, W.J., Song, W., 2010. Degradation mechanisms and kinetic studies for the treatment of X-ray contrast media compounds by advanced oxidation/reduction processes. Water Res. 44, 4391e4398. Katz, I., Dosoretz, C.G., 2008. Desalination of domestic wastewater effluents: phosphate removal as pretreatment. Desalination 222, 230e242. Kimura, K., Iwase, T., Kita, S., Watanabe, Y., 2009. Influence of residual organic macromolecules produced in biological wastewater treatment processes on removal of pharmaceuticals by NF/RO membranes. Water Res. 43, 3751e3758. Klavarioti, M., Mantzavinos, D., Kassinos, D., 2009. Removal of residual pharmaceuticals from aqueous systems by advanced oxidation processes. Environ. Intl. 35, 402e417. Locke, B.R., Sato, M., Sunka, P., Hoffmann, M.R., Chang, J.S., 2006. Electrohydraulic discharge and nonthermal plasma for water treatment. Ind. Eng. Chem. Res. 45, 882e905. Malato, S., 2008. Removal of emerging contaminants in wastewater treatment: removal by photo-catalytic processes. Hdb. Env. Chem. 5 (Part S/2), 177e197. Ning, B., Graham, N.J.D., 2008. Ozone degradation of iodinated pharmaceutical compounds. J. Environ. Eng. ASCE. 134, 944e953. Oleksy-Frenzel, J., Wischnack, S., Jekel, M., 2000. Application of ion-chromatography for the determination of the organicgroup parameters AOCl, AOBr and AOI in water. Fresenius J. Anal. Chem. 366, 89e94. Oliver, A., Lester, J.N., Voulvoulis, N., 2005. Pharmaceuticals: a threat to drinking water? Trends Biotech. 23, 163e167. Pekarek, S., 2003. Non-thermal plasma ozone generation. Acta Polytech. 43, 47e51. Putschew, A., Miehe, U., Tellez, A.S., Jekel, M., 2007. Ozonation and reductive deiodination of iopromide to reduce the environmental burden of iodinated X-ray contrast media. Water Sci. Technol. 56, 159e165. Seitz, W., Jiang, J.-Q., Schulz, W., Weber, W.H., Maier, D., Maier, M., 2008. Formation of oxidation by-products of the iodinated X-ray contrast medium iomeprol during ozonation. Chemosphere 70, 1238e1246. Seitz, W., Jiang, J.-Q., Weber, W.H., Lloyd, B.J., Maier, M., Maier, D., 2006a. Removal of iodinated X-ray contrast media during drinking water treatment. Environ. Chem. 3, 35e39. Seitz, W., Weber, W.H., Jiang, J.-Q., Lloyd, B.J., Maier, M., Maier, D., Schulz, W., 2006b. Monitoring of iodinated X-ray contrast media in surface water. Chemosphere 64, 1318e1324. Sunka, P., Babicky, V., Clupek, M., Lukes, P., Simek, M., Schmidt, J., Cernak, M., 1999. Generation of chemically active species by

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 4 7 e5 0 5 7

electrical discharges in water. Plasma Sources Sci. Technol. 8, 258e265. Ternes, T.A., Stu¨ber, J., Herrmann, N., McDowell, D., Ried, A., Kampmann, M., Teiser, B., 2003. Ozonation: a tool for removal of pharmaceuticals, contrast media and musk fragrances from wastewater? Water Res. 37, 1976e1982. Van der Bruggen, B., Lejon, L., Vandecasteele, C., 2003. Reuse, treatment, and discharge of the concentrate of pressure-driven membrane processes. Environ. Sci. Technol. 37, 3733e3738. Vanderford, B.J., Pearson, R.A., Rexing, D.J., Snyder, S.A., 2003. Analysis of endocrine disruptors, pharmaceuticals, and

5057

personal care products in water using liquid chromatography/tandem mass spectrometry. Anal. Chem. 75, 6265e6274. Westerhoff, P., Moon, H., Minakata, D., Crittenden, J., 2009. Oxidation of organics in retentates from reverse osmosis wastewater reuse facilities. Water Res. 43, 3992e3998. Zhou, T., Lim, T.-T., Chin, S.-S., Fane, A.G., 2011. Treatment of organics in reverse osmosis concentrate from a municipal wastewater reclamation plant: feasibility test of advanced oxidation processes with/without pretreatment. Chem. Eng. J. 166, 932e939.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 5 8 e5 0 6 2

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Influence of electric current on bacterial viability in wastewater treatment V. Wei a,*, M. Elektorowicz b, J.A. Oleszkiewicz a a

Department of Civil Engineering, University of Manitoba, 15 Gillson St., Winnipeg, Canada R3T 5V6 Department of Building, Civil and Environmental Engineering, Concordia University, 1455 de Maisonneuve Blvd. West, Montre´al, Canada H3G 1M8 b

article info

abstract

Article history:

Minimizing the influence of electric current on bacterial viability in the electro-

Received 26 April 2011

technologies such as electrophoresis and electrocoagulation is crucial in designing and

Received in revised form

operating the electric hybrid wastewater treatment system. In this study the biomass from

24 June 2011

a membrane bioreactor (MBR) was subjected to constant direct current and the bacterial

Accepted 5 July 2011

viability was monitored against electrical intensity, duration as well as the spatial vicinity

Available online 19 July 2011

related to the electrodes. It was found that the bacterial viability was not significantly affected (less than 10% of death percentage) when the applied electric current density (CD)

Keywords:

was less than 6.2 A/m2 after 4 h. The percentage of live cell dropped by 15% and 29% at CD

Electric current

of 12.3 A/m2 and 24.7 A/m2, respectively. The pH of electrolytic biomass fluid has shifted to

Wastewater treatment

alkaline (from nearly neutral to around pH 10) at CD above 12.3 A/m2, which could have

Electrochemistry

been the contributing factor for the bacterial inactivation. The temperature change in the

Microorganisms

electrolytic media at all current densities during 4 h of experiment was less than 2
C, thus

Bacterial viability

temperature effects were negligible. Bacteria experienced different micro-environments in the electrochemical reactor. Bacterial cells on the cathode surface exhibited highest death rate, whereas bacteria outside the space between electrodes were the least affected. It was concluded that in an electro-technology integrated wastewater treatment process, sufficient mixing should be used to avoid localized inactivation of bacterial cells. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

With pressing demand for more environmentally sustainable solutions to wastewater treatment, as promising alternatives the electro-technologies such as electrophoresis and electrocoagulation are becoming more attractive and increasingly integrated into the wastewater treatment processes due to their high efficiency, minimization of external chemical addition and ease of automation control (Yang et al., 2002; Molla and Bhattacharjee, 2005; Kim et al., 2007; Bani Melhem and Elektorowicz, 2010). Biological processes in the modern

wastewater treatment facilities rely on the mixed microorganism communities to remove organic matter and nutrients (N and P) from raw wastewater, therefore, influence of electric current on bacterial activity or viability has been one of the major concerns in applying these electro-technologies. Cells function best in their natural and optimum environment. When a stressor such as a strong electric current is added, the cell’s metabolism, physiology, shape and movement will be impacted (Jackman et al., 1999; Satoshi et al., 1997; She et al., 2006; Thrash and Coates et al., 2008). Sakakibara and Kuroda, 1993 reported that there was a linear

* Corresponding author. E-mail addresses: [email protected] (V. Wei), [email protected] (M. Elektorowicz), [email protected] (J.A. Oleszkiewicz). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.011

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 5 8 e5 0 6 2

relationship between the decrease of denitrification rate and the increase of the electric current applied. Alshawabkeh et al. (2004) investigated the effect of electro-stimulation on an aerobic culture and found that a small “window” of DC fields (between 0.57 and 1.14 V/cm) resulted in improvement in the biological removal of chemical oxygen demand (COD). Luo et al. studied cell surface properties of phenol-degrading bacteria in the presence of low, moderate and high currents. They concluded that at low current (<20 mA), there were no significant changes in cell surface properties such as surface hydrophobicity, electrostatic charge and cell shape. Exposure to a DC of more than 20 mA (up to 40 mA) did produce significant changes and caused an increase in surface hydrophobicity and flattening of the cells. Liu et al. (1997) explored the bactericidal mechanisms of low amperage (10e100 mA) electric current (DC) on Staphylococcus epidermidis and Staphylococcus aureus and demonstrated that an electric current as low as 10 mA introduced the antibacterial substances of H2O2 and chlorine, at the cathode and anode, respectively. Many changes can be observed in the cell’s physiology and structure in response to these compounds. Ultimately, if concentrations of these substances are too high, the cell may lose its viability altogether, which has been extensively used as a bactericidal approach. Loghavi et al. (2007) applied moderate electric field across microbial growth media of Lactobacillus acidophilus and observed that stress caused by the electric field induced an increase in the bacteriocin (proteinaceous toxinsproduced by bacteria) production and an increase in transmembrane conductivity and diffusive permeability of nutrients, surfactants, bacteriocin and autoinducers. Li et al. (2001) demonstrated that the metabolism of nitrifying bacteria was inhibited when electric current was above 2.5 A/m2 (in an activated sludge process) or 5 A/m2 (in biofilms) and the nitrification rate in a biofilm was reduced by 20%. Thiobacillus ferrooxidans and Acidiphilium SJH was found by Jackman et al. (1999) to be inactivated with current intensity of 20 mA/cm2 but the sulphur-oxidizing bacteria activity at high cell densities could be recovered after the electricity was turned off. Tokuda and Nakanishi, 1995 reported that direct electric current might sterilize the suspensions of some bacterial cells such as Escherichia coli IFO 3301 and Pseudomonas aeruginosa IFO 2689 and that their death rate was proportional to the current intensity. The research discussed above was conducted using pure bacterial cultures. No work has been done to systematically investigate the viability of heterotrophic bacteria in biological wastewater treatment. In this study, the mixed biomass from a membrane bioreactor was challenged by (DC) current supplied from a pair of immersed aluminium electrodes. Viability of the bacteria was measured at different current density and duration of exposure, as well as in the various zones in the vicinity of electrodes.

2.

5059

beaker, a pair of aluminium electrodes was inserted into the biomass. The effective electrode area was 9 cm  9 cm, the distance between the electrodes was 5 cm and the direct current electricity was supplied by a Kepco BOP 100-2D unit. 30 mL of biomass was taken out for various analyses. pH was measured using Accumet XL50 Dual Channel meter (Fisher Scientific), dissolved COD (sCOD) and dissolved TOC (STOC) were determined by HACH Spectrophotometer DR/2500 and Tekmar Dohrmann Phoenix 8000, respectively. Samples for dissolved parameters were filtered through 0.45 mm glass fibre filter. Viability of bacteria was measured using the LIVE/ DEAD BacLight Bacterial Viability kits (P/N L13152) supplied by Molecular Probes, Inc. (Eugene, Oregon, USA), and Bio-Tek PowerWave XS was used for the microplate reading of viability tests. The specific oxygen uptake rate (SOUR) was measured following the procedure in the APHA Standard Methods for the Examination of Water & Wastewater (21st edition) and dissolved oxygen (DO) was monitored by Orion Star and Star Plus Meter (Thermo Scientific). All tests were performed in triplicates, and no standard errors are more than 10%. The LIVE/DEAD BacLight bacterial viability kit from Molecular Probes was reported to be representative in determining the fraction of active cells (Boulos et al., 1999; Bunthof et al., 2001) and was therefore employed to test the bacterial viability in the presence of electric current in this research. The BacLight stain package contains two nucleic acid-binding stains with different spectral characteristics and cell penetration capacity: one is SYTO 9 green-fluorescent nucleic acid, and the other one is red-fluorescent nucleic acid, propidium iodine. SYTO 9 penetrates all bacterial membranes (both live and dead) freely and stains the cells green, while highly charged propidium iodide only penetrates damaged cell membranes and stains the cells red, consequently the SYTO 9 stain fluorescence intensity decreases. Simultaneous application of both stains thus enables measurement of the relative ratio of viable cells with an intact membrane and dead cells with a compromised membrane. Calibration was performed following the supplier’s protocol except E. coli suspensions which were replaced by the MBR biomass. A standard curve was prepared for each microplate and a typical one is shown in Fig. 1.

Materials and methods

Heterotrophic bacterial mass was collected from a membrane bioreactor (MBR) which contained about 6 g/L of total suspended solids (TSS). 900 mL of that biomass was placed in a 1 L

Fig. 1 e A typical standard calibration curve for bacterial viability test.

5060

When the AC or DC current is applied to a pair of electrodes (an anode and a cathode) placed in a microbial suspension, an electric field is generated and acts on the microbes between electrodes. If the electric field is beyond a certain threshold, it may significantly change the cell’s shape, surface hydrophobicity and net surface charge. The electric field may also affect the orientation of membrane components such as lipids. All of these effects can potentially result in irreversible permeabilization of the membrane and subsequent leakage of essential cytoplasmic constituents and decreasing respiratory rate (Chen et al., 2002). Along with the action of electric field on microbial cells, electrochemical redox reactions occur simultaneously on the electrode surfaces (Thrash and Coates, 2008): At anode, þ



2H2 O  4e /O2 ðgÞ þ 4H

(1)

0

E ¼ 1.23 V (vs. NHE) if nonreactive material such as platinum is used 

2Cl  2e /Cl2 ðgÞ

(2)

0

E ¼ 1.358 V (vs. NHE) if there is considerable amount of chloride ions AlðsÞ  3e /Al ðaqÞ 3þ

0

(3) 0

E ¼ 1.66 V (vs. NHE) if active material (e.g. Al) is used. E stands for the standard electrode potential. At cathode, 2H2 O þ 2e /H2 ðgÞ þ 2OH

105 100 95 90

0 A/cm2

85

3.7 A/cm2

80

6.2 A/cm2

75

12.3 A/cm2

70

24.7 A/cm2

65 60 1

2

3

4

Electric application time (h)

Fig. 2 e Effect of current intensity and duration on the relative live cell percentage.

biological degradation; (2). direct electrochemical oxidation of organic substances into carbon dioxide; (3). adsorption or encapsulation into the mesh-like precipitate of aluminium hydroxyl and (4). oxidation of the organics by electrochemically generated oxidants such as perchloride or hydrogen peroxide. In this research active metal aluminium is used as the anode material, so its dissolution reaction [Al (s)  3e / Al3þ (aq)] is electrochemically preferred to [2H2O  4e / O2(g) þ 4Hþ] due to higher standard oxidation potential. In an electrochemical reactor the applied electric voltage U can be calculated based on the following equation (Drees et al., 2003): U ¼ Eeq þ d*j/k þ anode overpotential þ cathode overpotential

(4)

0

E ¼ 0.828 V (vs. NHE) O2 ðgÞ þ 2Hþ þ 2e /H2 O2

Live cell percentage(%)

110

Results and discussions

(5)

E0 ¼ 0.695 V (vs. NHE)if the fluid is acidic. The above electrochemical reactions suggest potentially large pH change in the localized vicinity of the electrodes in a nonbuffered stationary system and the production of toxic hydrogen peroxide and chlorine or subsequent hypochlorous acid, both of which may penetrate into the interior of the cells and accelerate the inactivation process. In this study the cell viability in presence of aeration (3.4 L/L min) and electric currents of various intensities were investigated for duration of 4 h and the results are presented in Fig. 2. Aeration generates considerable mixing and exchange of liquid in the vicinity of the electrodes. When the applied electric density was less than 6.2 A/m2, there was no significant (<10%) live cell decrease compared to the initial biomass or the control sample. However, the fraction of live cell dropped 15% and 29% at the CD of 12.3 A/m2 and 24.7 A/m2, respectively. This suggests that under the studied conditions 6.2 A/m2 is the CD the bacteria were not affected by, and that higher CD could be detrimental to the cell. Fig. 3 shows that after 8 h of direct current application at CD of 6.2 A/m2 sCOD and STOC were removed by 78.1% and 75.8%, respectively. The mechanisms of COD and TOC removal may include (Moreno-Casillas et al., 2007): (1).

where Eeq ¼ equilibrium potential difference between the anode and the cathode (V) k ¼ fluid conductivity (mho/m) d ¼ distance between electrodes (m), j ¼ current density (A/m2). When U is large enough (8 V in this study), it may overcome the equilibrium potential difference, electrode overpotentials and ohmic potential drop and induce the electrochemical reactions in Equations (1)e(5). In addition, as shown in Fig. 4, the pH of the biomass increased significantly to a level hostile for the bacteria when current densities reached or exceeded

Concentration (mg/L)

3.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 5 8 e5 0 6 2

240 220 200 180 160 140 120 100 80 60 40 20 0

Dissolved TOC Dissolved COD

0

2

4

6

8 10 12 14 16 18 20 22 24

Electric application time (hr.) Fig. 3 e Effect of current duration on STOC and sCOD (at CD of 6.2 A/m2).

5061

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 5 8 e5 0 6 2

10 9.5 9 pH

8.5 3.7 A/m^2

8

6.2 A/m^2

7.5 7

12.3 A/m^2

6.5

24.7 A/m^2

6 0

1

2

3

4

Electric application time (h) Fig. 4 e Effect of current intensity and duration on pH.

6.0 5.5

SOUR (mg/g.h)

24

Temperature (ºC)

23.5

5.0 4.5 4.0 3.5

23 3.7 A/m2

22.5

6.2 A/m2 22 12.3 A/m2 21.5

24.7 A/m2

21 0

0.5

1

1.5 2 2.5 Current duration (h)

3

3.5

4

Fig. 6 e Effect of current intensity and duration on the biomass temperature.

stimulation of cells induces changes in DNA and protein synthesis, membrane permeability and cell growth and revealed that at low level current, bacterial activity and metabolism which were measured in terms of alcohol production were enhanced; the mechanism of these changes is still not well understood. In the research reported here the electro-stimulation effect was not observed.

110 Fraction of live cells (%)

12.3 A/m2. Fig. 5 demonstrated that the biomass’s SOUR dropped by 42% after 4 h of electric inactivation at current density of 24.7 A/m2. The temperature changes observed during application of electric current at room temperature were displayed in Fig. 6. The maximum change at all current densities during 4 h was less than 2
C. Therefore, the temperature changes monitored should not have caused any bacterial inactivation effect. Bacteria experience different micro-environment in an electrochemical reactor, especially when the reactor is not stirred or there is little mixing. As shown in Fig. 7, bacterial cells on the cathode surface were directly subjected to significantly elevated pH and action of electric field, consequently exhibiting highest death rate, whereas bacteria outside the space between electrodes had the highest viability because they were beyond influence of the electric field and are least affected by the toxicity of electrochemical byproducts. Therefore, for a wastewater treatment process in which an electro-technology is incorporated, a strong mixing is desirable to enhance dispersion and diffusion of microorganisms and prevent localized cell inactivation. Direct currents may also be used to stimulate bacterial activity and metabolism in a process called electrostimulation. Several studies have been conducted on the stimulatory effects of low level direct currents on microbial growth. Nakanishi et al. (1998) reported that electro-

100 90 80

Outside the space between electrodes

70

Between electrodes

60 50

On the cathode surface

40

3.0

30

2.5 2.0 0

1

2

3

4

Electric application time (hr.)

Fig. 5 e SOUR of biomass vs. current duration (current density 24.7 A/m2).

1

2

3

4

Time under electric field (h)

Fig. 7 e Bacterial viability in the different zones relative to electrodes when no mixing was applied in the reactor (current density [ 12.3 A/m2).

5062

4.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 5 8 e5 0 6 2

Conclusion

Mechanisms attributed to inactivation of bacteria in the presence of a DC current include oxidative stress, the production of harmful oxidants and irreversible membrane permeabilization. Using the LIVE/DEAD BacLight bacterial viability test kit this work demonstrated that viability of the microorganisms in an electrically enhanced MBR is dependent on duration of the current application and current density. Under the studied experimental conditions, the bacterial viability was not significantly (less than 10% of inactivation) affected when the applied current density was less than 6.2 A/ m2, the live cell percentage dropped by 15% and 29% at currents of 12.3 A/m2 and 24.7 A/m2, respectively. The pH of the electrolytic biomass mixed liquor has shifted significantly to alkaline above electric densities of 12.3 A/m2, which may be partially responsible for the bacterial inactivation. The maximum temperature change at all current densities during 4 h was less than 2
C, therefore temperature was virtually excluded as a factor in cell inactivation. Bacteria experiencing different micro-environments in the electrochemical reactor have distinctly differing viability. Cells on the cathode surface exhibited highest death rate, whereas bacteria outside the space between electrodes had the highest viability as they were least affected by the toxicity of electrochemical byproducts and electric field. It is important that beside optimized current density and duration, sufficient mixing should be provided for an electrokinetically enhanced hybrid wastewater treatment reactor to prevent localized cell inactivation.

Acknowledgements The work has been sponsored by the Canadian Natural Sciences and Engineering Research Council (NSERC) through the Strategic Projects Grants program. Support from Mr Mario Gagne´ from Lusine d’e´ puration d’Auteuil (City of Laval); Mr Pierre Purenne from La station d’e´ puration des eaux use’ es City of Montreal; Dr Yoomin Lee from AECOM Canada is gratefully acknowledged.

references

Alshawabkeh, A.N., Shen, Y., Maillacheruvu, K.Y., 2004. Effect of DC electric fields on COD in aerobic mixed sludge processes. Environmental Engineering Science 21 (3), 321e329. Bani Melhem, K., Elektorowicz, M., 2010. Development of a novel Submerged Membrane Electro-Bioreactor (SMEBR): performance for fouling reduction. Environmental Science & Technology 44 (9), 3298e3304. Boulos, L., Prevost, M., Barbeau, B., Coallier, J., Desjardins, R., 1999. LIVE/DEAD BacLight: application of a new rapid

staining method for direct enumeration of viable and total bacteria in drinking water. Journal of Microbiological Methods 37, 77e86. Bunthof, C.J., Schalkwijk, V.S., Meijer, W., Abee, T., Hugenholtz, J., 2001. Fluorescent method for monitoring cheese starter permeabilization and lysis. Applied Environmental Microbiology 6 (4), 4264e4271. Chen, X., Chen, G., Yue, P.L., 2002. Investigation on the electrolysis voltage of electrocoagulation. Chemical Engineering Science 57 (13), 2449e2455. Drees, K.P., Abbaszadegan, M., Maier, R.M., 2003. Comparative electrochemical inactivation of bacteria and bacteriophage. Water Research 37 (10), 2291e2300. Jackman, S.A., Maini, G., Sharman, A.K., Knowles, C.J., 1999. The effects of direct electric current on the viability and metabolism of acidophilic bacteria. Enzyme and Microbial Technology 24 (5e6), 316e324. Kim, J.O., Jung, J.T., Yeom, I.T., Aoh, G.H., 2007. Electric fields treatment for the reduction of membrane fouling, the inactivation of bacteria and the enhancement of particle coagulation. Desalination 202 (1e3), 31e37. Li, X.G., Cao, H.B., Wu, J.C., Yu, K.T., 2001. Inhibition of the metabolism of nitrifying bacteria by direct electric current. Biotechnology Letters 23 (9), 705e709. Liu, W.K., Brown, M.R.W., Elliott, T.S.J., 1997. Mechanisms of the bactericidal activity of low amperage electric current (DC). Journal of Antimicrobial Chemotherapy 39 (6), 687e695. Loghavi, L., Sastry., S.K., Yousef, A.E., 2007. Effect of moderate electric field frequency on growth kinetics and metabolic activity of Lactobacillus acidophilus. Biotechnology Progress 24 (1), 148e153. Molla, S.H., Bhattacharjee, S., 2005. Prevention of colloidal membrane fouling employing dielectrophoretic forces on a parallel electrode array. Journal of Membrane Science 255 (1e2), 187e199. Moreno-Casillas, H.A., Cocke, D.L., Gomes, J.A.G., Morkovsky, P., Parga, J.R., Peterson, E., 2007. Electrocoagulation mechanism for COD removal. Separation and Purification Technology 56, 204e211. Nakanishi, K., Tokuda, H., Soga, T., Yoshinaga, T., Takeda, M., 1998. Effect of electric current on growth and alcohol production by yeast cells. Journal of Fermentation and Bioengineering 85 (2), 250e253. Sakakibara, Y., Kuroda, M., 1993. Electric prompting and control of denitrification. Biotechnology and Bioengineering 42, 535e537. Satoshi, N., Norio, M., Hiroshi, S., 1997. Electrochemical cultivation of Thiobacillus ferrooxidans by potential control. Bioelectrochemistry and Bioenergetics 43, 61e66. She, P., Song, B., Xing, X.-H., Loosdrecht, M.V., Liu, Z., 2006. Electrolytic stimulation of bacteria Enterobacter dissolvens by a direct current. Biochemical Engineering Journal 28, 23e29. Thrash, J.C., Coates, J.D., 2008. Review: direct and indirect electrical stimulation of microbial metabolism. Environmental Science & Technology 42 (11), 3921e3931. Tokuda, H., Nakanishi, K., 1995. Application of direct current to protect bioreactor against contamination. Biosci. Biotech. Biochem 59, 753e755. Yang, G.C.C., Yang, T.Y., Tsai, S.H., 2002. A preliminary study on electrically enhanced crossflow microfiltration of CMP (chemical mechanical polishing) wastewater. Water Science & Technology 46 (11e12), 171e176.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3 e5 0 7 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Remediation of groundwater contaminated with MTBE and benzene: The potential of vertical-flow soil filter systems Manfred van Afferden a,*, Khaja Z. Rahman a, Peter Mosig a, Cecilia De Biase b, Martin Thullner b, Sascha E. Oswald c, Roland A. Mu¨ller a a

Centre for Environmental Biotechnology (UBZ), UFZeHelmholtz Centre for Environmental Research, Permoserstrasse 15, 04318 Leipzig, Germany b Department of Environmental Microbiology, UFZeHelmholtz Centre for Environmental Research, Permoserstrasse 15, 04318 Leipzig, Germany c Institute for Earth and Environmental Sciences, University of Potsdam, Potsdam, Germany

article info

abstract

Article history:

Field investigations on the treatment of MTBE and benzene from contaminated ground-

Received 23 May 2011

water in pilot or full-scale constructed wetlands are lacking hugely. The aim of this study

Received in revised form

was to develop a biological treatment technology that can be operated in an economic,

4 July 2011

reliable and robust mode over a long period of time. Two pilot-scale vertical-flow soil filter

Accepted 5 July 2011

eco-technologies, a roughing filter (RF) and a polishing filter (PF) with plants (willows), were

Available online 14 July 2011

operated independently in a single-stage configuration and coupled together in a multistage (RF þ PF) configuration to investigate the MTBE and benzene removal perfor-

Keywords:

mances. Both filters were loaded with groundwater from a refinery site contaminated with

Benzene

MTBE and benzene as the main contaminants, with a mean concentration of 2970  816

Groundwater remediation

and 13,966  1998 mg L1, respectively. Four different hydraulic loading rates (HLRs) with

Hydraulic loading rate

a stepwise increment of 60, 120, 240 and 480 L m2 d1 were applied over a period of 388

MTBE

days in the single-stage operation. At the highest HLR of 480 L m2 d1, the mean

Pilot-scale constructed wetland

concentrations of MTBE and benzene were found to be 550  133 and 65  123 mg L1 in the

Vertical-flow soil filter

effluent of the RF. In the effluent of the PF system, respective mean MTBE and benzene

Willow tree

concentrations of 49  77 and 0.5  0.2 mg L1 were obtained, which were well below the relevant MTBE and benzene limit values of 200 and 1 mg L1 for drinking water quality. But a dynamic fluctuation in the effluent MTBE concentration showed a lack of stability in regards to the increase in the measured values by nearly 10%, which were higher than the limit value. Therefore, both (RF þ PF) filters were combined in a multi-stage configuration and the combined system proved to be more stable and effective with a highly efficient reduction of the MTBE and benzene concentrations in the effluent. Nearly 70% of MTBE and 98% of benzene were eliminated from the influent groundwater by the first vertical filter (RF) and the remaining amount was almost completely diminished (w100% reduction) after passing through the second filter (PF), with a mean MTBE and benzene concentration of 5  10 and 0.6  0.2 mg L1 in the final effluent. The emission rate of volatile organic compounds mass into the air from the systems was less than 1% of the inflow mass loading rate. The results obtained in this study not only demonstrate the feasibility of vertical-flow soil filter systems for treating groundwater contaminated with MTBE and benzene, but can

* Corresponding author. Tel.: þ49 341 235 1848; fax: þ49 341 235 1830. E-mail address: [email protected] (M. van Afferden). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.010

5064

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3 e5 0 7 4

also be considered a major step forward towards their application under full-scale conditions for commercial purposes in the oil and gas industries. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Due to the widespread use of fuels, BTEX compounds (benzene, toluene, ethylbenzene, m-, o-, and p-xylene) and MTBE (methyl tertiary-butyl ether) are frequently detected groundwater contaminants, with releases occurring during their production, transportation and storage (Baehr et al., 1999; Deeb et al., 2000; Squillace et al., 1996). MTBE has received considerable attention in recent times as it migrates much more quickly through the soil than most of the petroleum distillates due to its high water solubility (up to 51 g L1, USEPA, 2004). Its presence in the environment is considered as a health and drinking water problem and classifies MTBE as a possible human carcinogen (Johnson et al., 2000). MTBE is relatively resistant to biological degradation under anaerobic conditions (Moreels et al., 2006), but several studies have shown a biodegradability under aerobic conditions (Deeb et al., 2000; Ferreira et al., 2006; Schmidt et al., 2004). Benzene is carcinogenic and the most water soluble BTEX compound. It can also be degraded by many microorganisms under aerobic conditions (Yerushalmi et al., 2002). The present limit concentrations established by the United States Environmental Protection Agency and the German guideline value are 200 mg L1 for MTBE and 1 mg L1 for benzene in drinking water (USEPA, 2005; DVGW, 2001). The physico-chemical properties especially the high water solubility and the low carbon adsorption coefficient of MTBE make it difficult to treat these organic contaminants by using conventional groundwater treatment technologies and represent some unique remediation challenges. The active exsitu remedial methods include air stripping and removal with granular activated carbon, vapour extraction, advanced chemical oxidation and multiphase high-vacuum extraction (Davis and Powers, 2000; Deeb et al., 2003; Sutherland et al., 2004; Wilhelm et al., 2002). However, the cost associated with the construction, maintenance and operation of these treatments diminishes their feasibility. Constructed wetland (CW) systems represent an effective and inexpensive option for treating municipal wastewater and becoming available due to their wide range of applications (Cooper, 1999; Kadlec and Wallace, 2009). They are also accepted as an alternative method to the commonly used engineering-based treatment technologies for the removal of organic contaminants from surface water or groundwater (Rubin and Ramaswami, 2001; Kassenga et al., 2004; Lorah and Voytek, 2004). In general, the vertical-flow constructed wetlands or soil filters are gaining popularity due to their greater oxygen transfer capacity and smaller size as compared to the horizontal-flow wetland systems (Cooper, 1999; Kadlec and Wallace, 2009). The findings of Eke and Scholz (2008) suggested that intermittently flooded vertical-flow constructed wetlands are able to effectively treat benzene from hydrocarbon-contaminated wastewater streams in the

presence of sufficient oxygen and fertilizer. But very little is known about the technical use of vertical-flow constructed wetlands for the removal of both MTBE and benzene from heavily contaminated groundwater. The SAFIRA-project (remediation research in regionally contaminated aquifers) is an interdisciplinary research project focussing on innovative remediation technologies to treat complex groundwater contamination. Within the framework of this research project, a pilot plant was constructed at a refinery in Leuna, Germany, aiming at the investigation and development of eco-technologies for the removal of volatile organic compounds. Since the groundwater treatment technology currently used in Leuna (pumpand-treat system associated with an air stripping and adsorption unit) is very expensive and requires high maintenance efforts, the aim of this work was to develop an alternative biological treatment technology that can be operated in an economic, reliable and robust mode over a long period of time. Therefore, a specially designed pilot-scale subsurface vertical-flow constructed wetland system was installed and operated at the Leuna site for field investigations on the removal of MTBE and benzene as the main groundwater contaminants. In order to identify the potential factors influencing the treatment efficiencies, the dynamics of MTBE and benzene were investigated using pilot-scale single-stage and multi-stage single-pass vertical-flow soil filter ecotechnologies with different hydraulic loading rates (HLRs) in this study. As far as we are aware, no such biological treatment system has been explored to date in pilot-scale facilities for treating MTBE and benzene compounds from contaminated groundwater using the planted and unplanted verticalflow soil filter systems, nor has the effect of the different hydraulic loading conditions been directly compared. The main objectives of this study were: (i) to explore the treatment performances of pilot-scale single-stage and multistage single-pass vertical-flow soil filter systems for removing MTBE and benzene from contaminated groundwater; (ii) to evaluate the potential effects of the different hydraulic loading rates (HLR) on the treatment efficiencies in both systems; and finally (iii) to assess the feasibility of applying a vertical-flow soil filter eco-technology to treat MTBE and benzene contaminated groundwater under full-scale conditions for commercial purposes.

2.

Materials and methods

2.1.

Site location and groundwater composition

The pilot-scale treatment facility was built near the Leuna refinery in the North-East of Germany in 2007. Due to accidental spills, improper handling (leaking underground storage tanks, pipelines, etc.), and damages due to heavy bombing

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3 e5 0 7 4

5065

during World War II, the groundwater in this area is heavily contaminated with high concentrations of different gasoline components (Martienssen et al., 2006). The fuel additive MTBE and benzene are the predominant groundwater contaminants at the site, with mean concentrations of 2970  816 and 13,966  1998 mg L1, respectively. The mean concentrations and standard deviations of the main organic and inorganic compounds present in the water and the geochemical characteristics of the influent groundwater observed during the investigation period are given in Table 1.

2.2.

Filter design

The two vertical-flow soil filters used in this study, the Roughing Filter (RF) and Polishing Filter (PF), consisted of two identical stainless steel containers (length: 2.30 m, width: 1.75 m, depth: 1.75 m), with a surface area of 4.025 m2 and a total volume of 7.04 m3. Both filters were filled with a granular material of different grain sizes and arranged in layers of varying configurations (Fig. 1). The filters were part of a larger pilot plant with central maintenance facilities and operated outdoors at the site, with their surface exposed to the local climatic conditions. The Roughing Filter (RF) consisted of three successive layers of filter packing materials: a cover layer on the top (25 cm), a main filter layer (120 cm) in the middle and a bottom layer (10 cm), which served as the drainage layer. The bottom drainage layer was separated from a 20 cm deep sump by a perforated steel plate. The cover layer was composed of coarse expanded clay material (8e16 mm), facilitating water distribution over the entire filter surface area and protecting

Table 1 e Influent groundwater characteristics based on samples collected during the whole experimental operation period of 20 months (from September 2008 to May 2010, except where noted). Parameter

MTBE Benzene Cl NHþ 4 SO2 4 PO3 4 Fe2þ Ca2þ Fetot Ptot Kþ Naþ Mg2þ Mn2þ O2 Eh s pHa Ta

Unit

mg L1 mg L1 mg L1 mg L1 mg L1 mg L1 mg L1 mg L1 mg L1 mg L1 mg L1 mg L1 mg L1 mg L1 mg L1 mV mS cm1 e  C

Inflow groundwater composition Mean

Standard deviation

Number of samples

2970.18 13,965.62 116.85 51.04 11.09 1.20 6.73 205.73 6.69 0.84 12.36 132.38 58.02 1.63 0.10 432.25 2.32 7.45 12.20

816.25 1997.88 9.96 9.34 8.95 0.75 2.36 14 1.57 0.18 0.87 8.03 3.20 0.23 0.07 161.7 0.40 0.35 3.11

484 469 44 44 44 44 44 44 43 44 44 44 44 44 57,075 57,935 57,946 54,046 54,046

a Online measurement from September 2008 to April 2010.

Fig. 1 e Schematic diagram of the roughing filter (RF; on the top) and the polishing filter (PF; on the bottom): (1) Inflow feeding pipe; (2) Distribution pipe; (3) Layered filter material; (4) Sump; (5) Plant biomass; (6) Drainage outlet pipe.

the surface of the main layer from erosion. The 25-cm thick cover layer was designed to reduce the emission of volatile organic compounds. The underlying main layer consisted of expanded clay material with a grain size in the range of fine gravel (3e6 mm). One reason for using such a gravel material was to prevent clogging due to a potential precipitation of iron and carbonate within the filter bed. The advantage of the larger pore spaces within these gravel particles reduced the chances of filter clogging and increased the possibility of applying higher hydraulic loads, which eventually facilitated this filter system to serve as a potential first treatment step. Finally, the drainage layer at the bottom consisted of crushed gravel (8e16 mm), which prevented the washing out of fine particles into the sump. The Polishing Filter (PF) comprised four successive layers. The 15-cm cover layer on the top consisted of a coarse expanded clay material (8e16 mm). The underlying main filter layer of 120 cm was filled with zeolite material (zeosoil; grain size 0e5 mm). The reason for using a finer material was that the proportion of the finer particles caused a greater surface area. Moreover, a longer hydraulic retention time is associated with a higher degradation of organic pollutants and a homogeneous distribution of the contaminated groundwater within this main filter layer. Zeolites have a larger surface area, a special texture and inner structure, as compared to conventional sand, and were therefore used within this filter system. However, their smaller pore spaces are associated with the risk of filter clogging and hence this filter system was designed to serve as a potential second treatment step. To facilitate better water discharge, the PF was constructed of two drainage layers underlying the main layer. The upper 20cm drainage layer consisted of crushed gravel (8e16 mm)

5066

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3 e5 0 7 4

and IV). The duration of each experimental phase was preset to guarantee that a representative number of samples were taken from each system. Detailed information on the operational strategies and loading schedules of both systems is listed in Table 2. During operational period 2 (days 388e611), the filters (RF and PF) were connected to each other and operated in series as a multi-stage single-pass vertical-flow filter system (RF þ PF). The RF was receiving the contaminated groundwater from the inflow storage tank and served as a first treatment step with a hydraulic loading rate (HLR) of 960 L m2 d1. The pre-treated groundwater from the effluent of the RF was then pumped into the second filter (PF) and passed through the second system at a hydraulic loading rate of 480 L m2 d1. The remaining 50% of the RF-effluent was sent to the nearby technical groundwater remediation plant for further treatment (stripping coupled with activated carbon adsorption) and then re-injected into the aquifer. With the highest HLR of 960 L m2 d1 in the RF system, we were interested to see if any hydraulic or technical problems occur, such as clogging, overloading, etc. This experimental phase V was run over a period of 223 days (days 388e611). Similarly to operational period 1 (single-stage configuration), both filters (RF and PF) were intermittently loaded with repeated pulses of groundwater (Table 2). The experiment started with period 1 in September 2008 and continued until the end of period 2 in May 2010. Willow trees on the PF system showed an active growth of their biomass, densely covering the whole filter surface area with green and healthy shoots before the start of the experiment.

followed by another 20-cm layer packed with even coarser crushed gravel (16e32 mm) and placed at the bottom of the filter. The PF was planted with white willows (Salix alba) on the top, with a density of around 5 plants m2. Trees of almost equal biomass (average height of 50 cm) and strength were obtained from a local supplier and uniformly planted at the end of August 2007. Willow trees were used due to their high biomass productivity, their relatively high resistance to organic contaminants, their ability to adapt to a broad range of climatic and site specific conditions, their broad reaching root systems, and their common use for phytoremediation (Mleczek et al., 2010; Rentz et al., 2005). The RF was unplanted in this investigation. The contaminated groundwater was injected from the top of each filter through a uniform distribution system of perforated PVC pipes, which was laid horizontally under the cover layer. Water drained through the filter media to the bottom of each basin, from where it was collected and discharged at the outflow by a PVC drainage pipe.

2.3.

Experimental conditions for filter operation

Contaminated groundwater was pumped by a timercontrolled pump into an anaerobic storage plastic container (Volume: w3 m3). Another timer-controlled pump distributed the water as intermittent loads through distribution pipes onto the surface of the two filter systems. This intermittent dosing of water was chosen to provide good oxygen transfer to the water phase (Kadlec, 2001). The pulse frequencies for the two filters under different experimental conditions are presented in Table 2. The experimental strategy was divided into two distinctly different operation periods. During operational period 1 (days 0e388), both filters (RF and PF) were operated independently as single-stage single-pass vertical-flow filter systems and received the influent groundwater separately from the same storage tank in parallel. Four different hydraulic loading rates (HLRs) were applied to the systems and increased stepwise (60, 120, 240 and 480 L m2 d1) over the period of 388 days comprising four different experimental phases (phase I, II, III

2.4.

Sampling and analysis

Concentrations of dissolved MTBE and benzene at the influent and effluent of each system were analysed online using a completely automated gas chromatograph (GC) equipped with a photoionisation detector (PID) (META Water sampling and analysis system WSS3; type: meta 3 HE II/PID, META, Messtechnische Systeme GmbH, Dresden, Germany). An Ultimetal column with a length of 25 m was used and the carrier gas was synthetic air, set at 5 bar. The oven and

Table 2 e Operation strategies and different experimental conditions (hydraulic loading schedules) of the vertical-flow soil filter systems during the whole investigation period. Period 1

2

Stage Single

Multiple

Phase

Duration (day)

I

0e86

II

86e235

III

235e297

IV

297e388

V

388e611

a Multi-stage combined system (RF þ PF).

Vertical filter

Volume of water per load (L)

Loading pulses per day ()

Injection interval (min)

HLR (L m2 d1)

RF PF RF PF RF PF RF PF

10 12 20 24 40 48 80 80

24 20 24 20 24 20 24 24

60 72 60 72 60 72 60 60

60 60 120 120 240 240 480 480

RFa PFa

80 60

48 32

30 45

960 480

5067

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3 e5 0 7 4

injection/detection port temperatures were 60 and 80  C, respectively. The detection limit for MTBE and benzene was 0.37 and 0.18 mg L1, respectively. The determination of detection and quantification limits of the calibration procedure was carried out in accordance with DIN 32645 (1994). Process control and data storage were carried out using the installed software (metaControl) that stored all the measurements and optional external signals on the hard disk of the attached PC. Known intermediate degradation products of MTBE, such as tert-butyl alcohol (TBA), tert-butyl formate (TBF) and aromatic hydrocarbons such as toluene, ethylbenzene, m-pXylene, o-Xylene, 1,3,5-Trimethylbenzene, 1,2,4Trimethylbenzene and Naphthalene were analysed in both influent groundwater and effluents from the filters by headspace gas chromatography and mass spectroscopic detection (HS-GCeMS). For headspace analysis, aqueous samples (10 ml) were stirred for 60 min at 70  C in headspace vials (20 ml) containing 2.5 g NaCl. Gas from the headspace (1 ml) was injected into a GC/MS (GC: Agilent 6890, MS: Agilent 5973) equipped with a 60 m HP1 column (Split injection 1:25, injection time 2 min). The time program was: 35  C for 6 min, to 120  C with 4  C/min and to 280  C with 20  C/min, held at 280  C for 5 min. The measuring time is 65 min per sample. The detection limit for TBA was 1.56 mg L1 and for other substances specified above was <1 mg L1.

2.5.

Emission measurement

In principle, the contaminated groundwater comes in contact with the atmosphere in both filter systems, and hence emissions of volatile organic substances are expected in the air during the treatment operation period. The volatile organic compounds (VOCs) were measured in terms of total organic carbon in a range of 0e100 mg TOC m3 using a mobile flame ionisation detector, FID 3-100 (JUM Engineering GmbH, Karlsfeld, Germany). The continuous flame ionisation chamber was heated up to 190  C. The measurements were performed at different heights (10, 20, 50, and 100 cm) in the air just above the centre (middle point) of each filter surface and also at same height immediately above the line of the inflow distribution pipe (inlet point) installed below the top layer of the filters. Moreover, the measurements were taken approx. 1e2 m downstream of each filter segment in the direction of the out-flowing wind (at 40 cm height; downwind) and approx. 5 m away from the filters against the wind direction (at 40 cm height; upwind) as a background value. The emission of VOCs in terms of TOC in mg m3 air was measured at an HLR of 480 L m2 d1 in both filter systems (experimental phase IV, single-stage operation). Measurements were taken in different measuring cycles over the RF and the PF system. Duration of each cycle was 60 min, which included an inflow feeding pulse with duration of 4e8 min and a continuous measurement of emission in the air at different specified heights. The emission of VOCs in terms of TOC in mg m3 from each measuring heights and also the background values were recorded over one feeding pulse interval in one cycle. The net emission at each particular height (measuring points) was calculated by subtracting the background value from the measured emission value attained at

that particular height. Four cycles were carried out for the emission estimation over the RF and only two cycles for the PF in this experimental phase with a same HLR in both the filters. Since wind can have a strong influence on the measurements, mobile walls were built around the filters to limit the movement of the air above the filter beds to a wind speed range of 0.1e0.5 m s1. The emitted mass of VOCs in each feeding pulse was also calculated with the assumption that the certain volume of water feeding on the filter segment per pulse was displacing the same volume of air which was coming out over the filter surface. Based on this assumption as a preliminary emission estimation study, the rate of emitted mass from each filter surface in terms of mg TOC m2 d1 and percentage of emission (%) from the inflow loading mass that goes in the atmosphere (air) were calculated.

3.

Results

3.1. Dynamics of MTBE and benzene: single-stage systems The influent and effluent dynamics of MTBE and benzene in the RF and PF system within the different experimental phases are shown in Figs. 2 and 3. During experimental phase I (days 0e86) with an HLR of 60 L m2 d1, the mean MTBE concentration in the effluent of the RF was detected to be 139  69 mg L1, which was below the limit value of 200 mg L1 for MTBE. In contrast, a relatively higher and wide range of MTBE concentration with a mean value of 332  680 mg L1 was

A B

Fig. 2 e Influent and effluent concentrations of MTBE in the A) RF and B) PF system during different experimental phases (IeIV) of operational period 1.

5068

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3 e5 0 7 4

A B

Fig. 3 e Influent and effluent concentrations of benzene in the A) RF and B) PF system during different experimental phases (IeIV) of operational period 1.

measured in the effluent of the PF system. In this phase, a mean reduction in the MTBE concentration of 97 and 93% was obtained, when the RF and PF single-stage system was used, respectively. Consequently, a mean concentration of benzene of 64  76 and 0.3  0.2 mg L1 in the effluent of the RF and the PF, respectively, was measured. An extremely high reduction in the benzene concentration (w100%) was observed in the PF system. During experimental phase II (days 86e235) with an HLR of 120 L m2 d1, the mean effluent MTBE concentration in the RF increased at the beginning and then steadily slowed down to a mean value of 399  318 mg L1. A relatively sharp decreasing tendency within the effluent MTBE concentration of the PF was observed in the middle part of phase II and continued with a very low concentration until the end of this phase (see Fig. 2A and B). However, a mean value of 91 and 93% reduction in the MTBE concentration was achieved in the effluent of the RF and the PF, respectively. Similarly to the effluent dynamics of MTBE, the effluent benzene concentration in the RF was increased gradually and then lowered down to a mean effluent concentration of 413  736 mg L1 from a mean influent concentration of 15,126  2382 mg L1. In the PF, no particular trend was seen in the dynamics of the effluent benzene concentration and a relatively higher mean value of 11  53 mg L1 with a great deviation was detected, as compared to the previous phase I (see Fig. 3A and B). In experimental phase III (corresponding to days 235e297) with an HLR of 240 L m2 d1, both the systems RF and PF started to develop differently as it was observed in the MTBE and benzene effluent dynamics. A mean effluent value of 402  222 mg L1 resulted in a mean MTBE-concentration

reduction of 84% in the RF, whereas in the PF, the effluent MTBE concentration sharply decreased almost immediately after changing the experimental phase and maintained a low concentration until the end of the phase. A mean value of 43  90 mg L1 resulted in a remarkable reduction (w99%) of the mean MTBE concentration in the PF system (Fig. 2A and B). In the case of benzene, the effluent concentration varied drastically in the RF even though there was a relatively constant influent and a very high mean effluent value of 401  803 mg L1 at the end of this experimental phase. In contrast, a highly efficient reduction (w100%) in the benzene concentration was monitored in the effluent of the PF, with a mean value of 0.3  0.2 mg L1 (Fig. 3A and B). At a higher HLR of 480 L m2 d1 in the next experimental phase IV (corresponding to days 297e388), a relatively constant effluent MTBE concentration was observed in the RF, with a mean value of 550  133 mg L1, which contributed to a mean MTBE-concentration reduction of 75% from the influent. No particular trend in the reduction of the MTBEconcentration values was detected within the effluent dynamics of the RF and a continuous fluctuation in the MTBE concentration with a wide range of values was observed in the effluent of the PF. Nevertheless, the mean effluent MTBE concentration of 49  77 mg L1 in the PF was nearly 11-fold lower than the mean MTBE concentration of 550  133 mg L1 in the RF (Fig. 2A and B). Similarly to the previous experimental phase III, the dynamics of benzene in the effluent of the RF and the PF showed a completely opposite trend. In the RF system, a rapid fluctuation in the benzene concentration values showing no particular reduction trend resulted in a mean effluent benzene concentration of 65  123 mg L1, whereas a relatively constant trend in concentration reduction was observed in the effluent of the PF. The mean value of 0.5  0.2 mg L1 in the effluent contributed to a highly efficient (w100%) reduction in the benzene concentration of the PF system, as compared to the RF (see phase IV; Fig. 3A and B).

3.2. Dynamics of MTBE and benzene: multi-stage system The dynamics of MTBE and benzene in both the influent and effluent of the combined multi-stage vertical-flow soil filter system (RF þ PF) during operational period 2, in the experimental phase V (corresponding to days 388e611), are shown in Fig. 4. The RF system as the first filter received contaminated groundwater at an HLR of 960 L m2 d1. The mean influent MTBE-concentration value of 2760  594 mg L1 was reduced to a mean effluent value of 831  318 mg L1, which resulted in a mean MTBE-concentration reduction of 69% in this treatment step. This effluent of the RF system was pumped intermittently onto the surface of the second filter (PF) at an HLR of 480 L m2 d1. The results demonstrated a remarkable (w99%) reduction in the MTBE concentration of the effluent of the PF with a mean value of 5  10 mg L1. Although the dynamics of the MTBE concentration in the effluent of the PF showed a rapid fluctuation in the values during this experimental phase, all the effluent concentration values were well below the limit value of 200 mg L1 for MTBE.

5069

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3 e5 0 7 4

0.6  0.2 mg L1 in the effluent of the second filter contributed to an almost complete (w100%) removal of benzene in this combined multi-stage system. The overall treatment performances obtained in both filters (RF and PF) during the whole operational period of this study are summarized in Table 3. The mean concentrations of intermediate degradation products of MTBE (TBA, TBF) and other aromatic hydrocarbons such as toluene, ethylbenzene, m-p-Xylene, o-Xylene, 1,3,5-Trimethylbenzene, 1,2,4-Trimethylbenzene and Naphthalene in the influent groundwater and effluents of both the RF and the PF are given in Table 4. Both TBA and TBF were detected with a low mean concentration value in the effluent of the RF and the PF system during the single-stage operational phase, but their concentrations were almost diminished or went below the detection limit in the final effluent after passing the multi-stage system. Similarly, in the final effluent of the multi-stage operational phase, the other aromatic hydrocarbons could not be detected due to a very low concentration value at the end (see Table 4).

A

B

3.3. Fig. 4 e Influent and effluent concentration along with the limit value of A) MTBE and B) benzene in the multiphase combined (RF D PF) system during experimental phase (V) of operational period 2 (days 388e611).

Emission estimation

The emissions of the volatile organic compounds (VOCs) in the air phase at several specified heights over the vertical-flow soil filter systems at a given day in several measurement cycles were registered and plotted on curves in this study. The measurements were recorded during the single-stage operation period at the same HLR of 480 L m2 d1 in both filters (experimental phase IV). An example of emission calculation in one measurement cycle over the RF and PF is given in Fig. 5A and B. In the RF, the inflow feeding pulse with duration of nearly 4 min contributed to an immediate displacement of inside trapped air to the filter surface and over a period of approximately 17 min, the displaced air disappeared and the emission level came back to the concentration at the background value until the end of the 60 min cycle (Fig. 5A). Emissions of VOCs were measured in this 17 min time duration and the obtained results showed that the highest emission with a concentration of 12.27 mg TOC m3 was recorded

In the case of benzene, the first filter (RF) received the influent groundwater with a mean benzene concentration of 13,527  1638 mg L1 and a drastic fluctuation was observed in the effluent benzene concentration values of this filter system. The values were spread out over a large range but a mean effluent value of 291  573 mg L1 resulted in a mean reduction in the benzene concentration of 98% from the influent, which did not meet the allowable limit value of 1 mg L1. However, after passing through the second filter (PF), a remarkably low and stable benzene concentration was detected in the effluent of the PF system. The mean value of

Table 3 e Summary of the treatment performances in the RF and the PF during the whole operational period of 611 days. Stage

Single

Phase

I II III IV

Multiple

V

Filter

MTBE

Benzene

Influent (mg L1)

Effluent (mg L1)

Removal (%)

n

Influent (mg L1)

Effluent (mg L1)

Removal (%)

n

RF PF RF PF RF PF RF PF

3953  298 4337  338 3850  680 4207  456 2635  490 3104  587 2214  266 2204  301

139  69 332  680 399  318 289  370 402  222 43  90 550  133 49  77

97 93 91 93 84 99 75 98

40 50 92 99 47 51 83 84

15,574  2800 18,695  1578 15,126  2382 17,030  2664 13,046  1463 14,856  1115 13,649  1142 13,052  2462

64  76 0.3  0.2 413  736 11  53 401  803 0.3  0.2 65  123 0.5  0.2

99 100 98 100 97 100 99 100

36 26 77 61 45 46 83 46

RFa PFa

2760  594 831  318

831  318 5  10

69 99

154 140

13,527  1638 291  573

291  573 0.6  0.2

98 100

154 100

a Multi-stage combined system (RF þ PF), n: number of samples.

5070

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3 e5 0 7 4

Table 4 e Concentration of intermediate degradation products of MTBE and other aromatic compounds analysed in the influent groundwater and the effluents of the RF and the PF system during the whole operational period of 611 days. Substance

Single-stage (phase IeIV) Influent (mg L1)

TBA TBF Toluene Ethylbenzene m-p-Xylene o-Xylene 1,3,5-Trimethylbenzene 1,2,4-Trimethylbenzene Naphthalene

53 4.1 8 50 76 6.5 4.7 393 68

        

12 1 1.4 37 53 2.9 3.6 141 22

Effluent, RF (mg L1) 13  2 1.1  1.4  1.3  1.2  1.6  2.6  1.2 

8 1.5 0.1 0.1 1.0 0.6 1.0 2 0.6

Multi-stage (phase V)

Effluent, PF (mg L1)

n

Influent (mg L1)

43 2  1.5 n.d. 1.1  0.2 1.3  0.5 n.d. n.d. 1.2  0.6 n.d.

24 24 24 22 22 23 21 24 24

41 2.7 6.7 31 57 6.7 2.4 252 36

6  1.3  0.7  14 7  0.9  1.5  63 9

Effluent, RF (mg L1)

Effluent, PF (mg L1)

n

13  3 1.3  0.5 n.d. 1.1  0.3 n.d. n.d. n.d. 4.1  3.1 n.d.

21 n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d.

12 11 11 11 11 11 9 12 12

n.d.: below the limit of detection (<1 mg L1), n: number of samples.

at the height of 10 cm over the inlet pipe (buried underneath the top layer of the RF). The concentration of VOCs decreased down to background level after this 17 min time period. Therefore, the net emission at the highest concentration measurement point was estimated as 10.55 mg TOC m3, which was achieved by subtracting the mean background value of 1.72  0.06 mg TOC m3 from the highest measured emission value (H10 Inlet data set, Fig. 5A). For better estimation in the PF system, the inflow feeding pulse duration was adjusted to nearly 8 min and measurement values were recorded over the next 8e10 min after the dosing. The maximum concentration of 3.18 mg TOC m3 was registered at the height of 10 cm over the central middle point

A

B

of the filter surface. The net emission at that highest concentration measurement point was calculated as 0.72 mg TOC m3, by subtracting the measured mean background value of 1.4  0.02 mg TOC m3 from the highest emission value measured in this measuring cycle over the PF (H10 Middle data set, Fig. 5B). The regional background levels were consistent during the whole emission measurement experiment with values in the range of 1e2 mg TOC m3 measured in the air. After each feeding pulse, it was assumed that a total volume of 80 L water (Table 2) was flushing on the filter surface and the same volume of 80 L entrapped air was coming out from the filter over the surface within a short time (8e20 min) and then disappeared. Based on this assumption for a preliminary emission calculation, it was observed that a mass of 42.77 mg TOC m2 d1 was emitted over the segment of the RF system. This was calculated quantitatively by the triangular area under the curve (actual emission measuring zone by connecting the H10 Inlet data set, Fig. 5A) multiplied by the HLR of 480 L m2 d1. After this particular emission zone within the curve, the dynamics of measured air emission came back to the concentration at the background level until the cycle ends and hence they were not taken into account in this calculation. Comparing to the inflow TOC mass loading rate of 7566.94 mg TOC m2 d1 to the filter bed, it was observed that only 0.45% of the inflow TOC mass was emitted over the surface and went into the surrounding atmosphere. Similar calculation approach by using the H10 Middle data set (Fig. 5B) estimated that the emitted VOCs mass percentage value was even lower (0.04%) in the PF system, as compared to the RF system under the same HLR. The summary of emission measurement calculations in each cycle over the RF and the PF system is given in Table 5.

4.

Discussion

4.1. MTBE and benzene removal performances: singlestage systems Fig. 5 e Concentration of VOCs emission from the surface of the A) RF and B) PF measured overall sampling heights in cycle 1 with the same HLR in the single-stage operation.

During the stepwise increase of the HLR (60e480 L m2 d1) in the first operational period of the single-stage systems, the RF

5071

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3 e5 0 7 4

Table 5 e Summary of the emission measurement cycles during a particular day over the surface of the RF and the PF system with the same HLR obtained in single-stage operation. Stage

Filter

Single (HLR: 480 L m2 d1)

RF

PF

Measuring cycle

Inflow mass loading (mg TOC m2 d1)

Emitted mass over segment (mg TOC m2 d1)

Emission (%)

1 2 3 4 1 2

7566.94

42.77 50.76 30.23 21.41 2.98 1.37

0.45 0.67 0.40 0.28 0.04 0.02

7277.39

clearly showed a decreasing tendency in the mean MTBE removal (%), which means the lower the HLR, the better the MTBE removal from this system (Table 3). Moreover, the overall reductions in MTBE concentrations were not sufficient to reach the limit value of 200 mg L1 (USEPA, 2005; DVGW, 2001). In contrast, the PF showed a better performance of increasing the MTBE removal (%) associated with increases in the HLR. At the highest HLR of 480 L m2 d1 in the PF, the mean effluent MTBE concentration was recorded to be 49  77 mg L1, which was nearly 7 times less than the mean effluent MTBE concentration (332  680 mg L1) at the lowest HLR of 60 L m2 d1 and well below the limit value of 200 mg L1 for MTBE. However, although the mean MTBE concentration of 49  77 mg L1 was below the limit value, a rapid fluctuation in the dynamics of the effluent MTBE concentration was observed and nearly 10% of the measured values were higher than the limit value of 200 mg L1 (see Fig. 2B). This indicates that the MTBE removal performance obtained in the PF system was sufficient, but not stable. In the case of benzene, a mean removal within a range of 97e99% was achieved in the RF, but the overall effluent concentrations were never below the limit value of 1 mg L1. In contrast, the PF system exhibited a highly efficient benzene concentration reduction and the effluent concentrations were predominantly found to be below 1 mg L1 (see Fig. 3A and B). The differences in the MTBE and benzene treatment performance observed in the RF and PF might be due to the differences in filter designs mainly defined by the different filtering media used as the main filter materials (expanded clay in the RF and zeosoil in the PF; Fig. 1). The obtained results show that a filter loaded with a fine zeosoil (0e5 mm) filter material and plants is more efficient in MTBE and benzene removal than a filter loaded with a coarse expanded clay material (3e6 mm) without plants. This is probably due to a higher reactive surface area, a better oxygen transfer and a higher hydraulic retention time (data not shown) within the filter loaded with fine materials. But finer materials have the disadvantage of a possible filter clogging and also the problem associated with water saturation at high hydraulic loads. Additionally the differences in effective depth and different compactions of the filter bed, different gas exchange rates and planteroot activity in the case of the PF that provide oxygen to the rhizosphere (Scholz, 2006) might explain the observed differences in treatment performance. However, more investigations are needed before making any concluding remarks on these particular assumptions.

An effective benzene biodegradation could be expected in the two RF and PF filter systems, since this pollutant has been degraded in environmental systems even under hypoxic conditions and treatment efficiencies for aerobic bioreactors up to 100% have been described (Yerushalmi et al., 2002). In contrast, MTBE biodegradation is reported to be by far not as effective as benzene biodegradation. The possible reasons might be that MTBE is more resistant to enzymatic attacks due to its tertiary carbon atom and the ether bond (Davidson and Creek, 2000). Moreover, it is reported that the biodegradation of MTBE might be inhibited due to the presence of cocontaminants, such as benzene, ethylbenzene, toluene and xylene (BTEX), and the accumulation of by-products from the biodegradation of BTEX compounds (Raynal and Pruden, 2008). An inhibition of MTBE biodegradation in the presence of BTEX due to a potential production of by-products has also been suggested by others (Deeb et al., 2001; Sedran et al., 2002). These studies have focused mainly on substrate inhibition (Park, 1999), by-product inhibition (Wilson et al., 2002) or competitive inhibition (Sedran et al., 2002). Therefore, the presence of BTEX compounds in the groundwater of the refinery site was expected to inhibit MTBE biodegradation in both the RF and PF systems. But these inhibitory effects on MTBE biodegradation could not be observed in both the filters during this study. Table 4 shows the presence of BTEX and other aromatic hydrocarbons in the influent groundwater with a mean concentration value and still a highly efficient MTBE-concentration reduction (93e98%) can be seen especially in the effluent of the PF during singlestage operation (Table 3). However, very little is known about the microbial community structure during the aerobic MTBE degradation in the presence of BTEX. The disappearance of MTBE metabolites, such as tert-butyl alcohol (TBA) and tert-butyl formate (TBF) indicated the potential of complete biodegradation within the filter systems. As can be seen in Table 4, the mean concentrations of TBA and TBF were remarkably decreased in the effluent of the RF and the PF system during the single-stage operation period and almost completely diminished or biodegraded after passing through the second filter (PF) during the multistage operation period.

4.2. MTBE and benzene removal performances: multistage system The second operational period with a multi-staged combined (RF þ PF) vertical-flow soil filter system runs very well with

5072

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3 e5 0 7 4

a highly efficient MTBE and benzene removal. Both MTBE and benzene effluent concentrations were always stable and well below the allowable limit value after the second treatment step. Moreover, the high HLR in the first filter (RF) was not accompanied by hydraulic or technical problems, such as clogging, overloading, etc., and it was thus very encouraging to a further increase of HLR, which could reduce the cost of the required land for the filter construction. Therefore, it can be concluded that the multi-stage combined vertical-flow soil filter system is more stable, more effective and a better option for the removal of MTBE and benzene from contaminated groundwater, as compared to the single-stage system. In principle, the overall decrease in the MTBE and benzene concentrations obtained from the vertical-flow soil filter systems can be caused by microbial degradation, sorption onto solid filter packing materials and volatilization. Moreover, in the planted filter system, it might also be caused by plant uptake followed by transport, transformation and phyto-volatilization. A long-term operation of the verticalflow soil filter systems is leading to an established adsorption/desorption balance and therefore, a removal by sorption onto the filtering media can be assumed to be negligibly small in this investigation. Biodegradation of MTBE and benzene is expected to be the most dominant process for the removal of these contaminants from the groundwater. However, the extent of degradation cannot be estimated accurately without a long-term and complete set of data in terms of volatilization and plant uptake from pilot-scale vertical-flow constructed wetlands. Eke and Scholz (2008) also concluded that the impacts of volatilization, biodegradation and adsorption on the benzene removal are often difficult to separate quantitatively from each other. For a long-term operation of the vertical-flow soil filters, the designed hydraulic loads need to be achieved by optimising the volume of the water in each loading pulse and the associated frequency, in order to increase the dewatering efficiency of the filters in the period of time between the intermittent pulses, and thus promoting oxygenation and achieving treatments with the highest possible level of efficiency.

4.3.

Emissions

After estimating the emissions of volatile organic compounds from the vertical-flow soil filter systems, the overall results indicated that the emissions from the planted PF systems in the air were much lower than those from the RF system and were only slightly above the background value (Fig. 5). By comparing to the inflow TOC mass loading rate, only a negligible amount (<1%) was emitted from the surface of the both RF and the PF systems. Therefore, with a highly efficient mean MTBE and benzene concentration reduction in the effluent of the RF and the PF system and almost a negligible emission rate of VOCs mass, it can be concluded that the biodegradation is the predominant removal pathway of both MTBE and benzene within the vertical-flow soil filter system treating contaminated groundwater. Volatilization of toxic organic hydrocarbons may be increased by technological problems, such as clogging and subsequent flooding, and may lead to serious air pollution

(Braeckevelt et al., 2008). But both of our vertical-flow soil filter systems were almost free of technical problems such as filter clogging, overloading, surface flooding, etc. Experimental investigations have shown that phyto-volatilization is a potential emission path for MTBE and benzene along with the direct volatilization via the soil surface of a constructed wetland (Reiche et al., 2010). However, more improved technical equipment is necessary for measuring both the VOCs concentration in the air and the volume of air emitted from the surface of the filter beds. Future investigations should be carried out with the purpose of a final evaluation of the volatilization rate of MTBE and benzene per unit area (m2) of the filter surface and with the aim of achieving a complete mass balance of organic compounds and discovering the role of the cover layer for protecting volatilization.

5.

Conclusions

The following conclusions can be drawn from the current study: 1. The Polishing Filter (PF) with a finer material and plants is more efficient in removing MTBE and benzene from contaminated groundwater, as compared to the Roughing Filter (RF) with a coarse material and without plants. 2. Factors, such as filter packing material, particle size, filter depth and loading rate, are playing an important role in achieving a robust filter operation for the removal of organic contaminants by vertical-flow soil filter systems. 3. The MTBE removal performance decreases with an increasing HLR in the RF, whereas the PF system is characterized by a remarkable MTBE and benzene removal performance at an increasing HLR. 4. At a higher HLR, the MTBE removal performance of a single-stage vertical-flow soil filter system is often stable, but not sufficient. 5. In general, a continuous reduction in both the MTBE and benzene concentration of the effluent indicates that the maximum treatment capacity is yet to be reached in both the RF and the PF systems. 6. The multi-stage combined vertical-flow soil filter system (RF þ PF) produces the most stable and sufficient effluent concentrations to reach the limit concentrations of MTBE and benzene for drinking water. 7. Since a negligible amount of volatile organic compounds is going in the air from our filter systems, therefore they are not considered to be a potential source of air pollution affecting the surrounding environment. 8. Since the vertical-flow constructed wetlands are accumulative systems (biomass, organic matter, calcareous material, etc.), it is of great importance to assess the optimal operation/design load of the filters treating MTBE and benzene and to predict the cases in which hydraulic overloads might be problematic for the filter longevity. 9. Our systems are designed to minimize clogging and after treating groundwater by using our technology, the water will not need to be amended further and can be released into an aquifer or discharged into any conventional drainage system.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3 e5 0 7 4

10. This novel groundwater remediation technology promises to be a cost effective remediation approach for treating groundwater contaminated with MTBE and benzene on a full-scale. 11. To prove the long-term stability and process optimization as well as to reach a sound economical and ecological evaluation for this new approach, a pilot-scale or fullscale operation over an extended period of time is needed. Further studies are intended: i) to focus on identifying the major microbial processes that lead to an aerobic biodegradation of organic contaminants, ii) to quantify volatilization, adsorption, absorption, mineralization and other removal mechanisms in large-scale vertical-flow soil filter systems treating MTBE and benzene, iii) to characterize and quantify plant uptake, phyto-sorption and phyto-volatilization of MTBE and benzene, iv) to explore the effects of seasonal temperature changes on the removal of MTBE and benzene, v) to investigate the effects of iron and calcium precipitations on the filter performance in treating MTBE and benzene contaminated groundwater, vi) to define the design criteria for the remediation of contaminated groundwater using verticalflow soil filter eco-technologies.

Acknowledgements This work was supported by the Helmholtz Centre for Environmental Research e UFZ in the scope of the SAFIRA II Research Programme (Revitalization of Contaminated Land and Groundwater at Megasites, sub-project ‘‘Compartment Transfer e CoTra’’) and funded by a grant from the German Federal Ministry of Education and Research (BMBF). The authors would like to thank Francesca Lo¨per, Grit Weichert, Sibylle Mothes, Heidrun Paschke, and Christina Petzold for their assistance in the field and laboratory work.

references

Baehr, A.L., Stackelberg, P.E., Baker, R.J., 1999. Evaluation of the atmosphere as a source of volatile organic compounds in shallow groundwater. Water Resources Research 35, 127e136. Braeckevelt, M., Mirschel, G., Wiessner, A., Rueckert, M., Reiche, N., Vogt, C., Schultz, A., Paschke, H., Kuschk, P., Kaestner, M., 2008. Treatment of chlorobenzenecontaminated groundwater in a pilot-scale constructed wetland. Ecological Engineering 33, 45e53. Cooper, P.F., 1999. A review of the design and performance of vertical-flow and hybrid reed bed treatment systems. Water Science and Technology 40, 1e9. Davidson, J.M., Creek, D.N., 2000. Using the gasoline additive MTBE in forensic environmental investigations. Environmental Forensics 1 (1), 31e36. Davis, S., Powers, S., 2000. Alternative sorbents for removing MTBE from gasoline-contaminated ground water. Journal of Environmental Engineering 126 (4), 354e360. Deeb, R.A., Chu, K.H., Shih, T., Linder, S., Suffet, I., Kavanaugh, M. C., Alvarez-Cohen, L., 2003. MTBE and other oxygenates:

5073

environmental sources, analysis, occurrence and treatment. Environmental Engineering Science 20 (5), 433e444. Deeb, R.A., Hu, H.-Y., Hanson, J.R., Scow, K.M., Alvarez-Cohen, L., 2001. Substrate interaction in BTEX and MTBE mixtures by an MTBE-degrading isolate. Environmental Science and Technology 35 (2), 312e317. Deeb, R.A., Scow, K.M., Alvarez-Cohen, L., 2000. Aerobic MTBE biodegradation: an examination of past studies, current challenges and future research directions. Biodegradation 11, 171e186. DIN 32645, 1994. “Chemische Analytik”, Nachweis-, Erfassungsund Bestimmungsgrenze e Ermittlung unter Wiederholungsbedingungen e Begriffe, Verfahren, Auswertung. Hrsg.: Deutsches Institut fu¨r Normung, BeuthVerlag, Berlin, pp. 1e20 (in German). DVGW, D.V.d.G.-u.W.e.V., 2001. Verordnung zur Novellierung der Trinkwasserverordnung vom 21. Mai 2001, Bonn, Germany (in German). Eke, P.E., Scholz, M., 2008. Benzene removal with vertical-flow constructed treatment wetlands. Journal of Chemical Technology Biotechnology 83, 55e63. Ferreira, N.L., Malandain, C., Fayolle-Guichard, F., 2006. Enzymes and genes involved in the aerobic biodegradation of methyl tert-butyl ether (MTBE). Applied Microbiology and Biotechnology 72 (2), 252e262. Johnson, R., Pankow, J., Bender, D., Price, C., Zogorsky, J., 2000. MTBE to what extent will past releases contaminate community water supply wells? Environmental Science and Technology 34, 210Ae217A. Kadlec, R.H., 2001. Thermal environments of subsurface treatment wetlands. Water Science and Technology 44 (11-12), 251e258. Kadlec, R.H., Wallace, S.D., 2009. Treatment Wetlands, second ed. Taylor and Francis Group, Boca Raton, USA, ISBN 978-1-56670526-4. Kassenga, G.R., Pardue, J.H., Moe, W.M., Bowman, K.S., 2004. Hydrogen thresholds as indicators of dehalorespiration in constructed treatment wetlands. Environmental Science and Technology 38 (4), 1024e1030. Lorah, M.M., Voytek, M.A., 2004. Degradation of 1,1,2,2tetrachloroethane and accumulation of vinyl chloride in wetland sediment microcosms and in situ porewater: biogeochemical controls and associations with microbial communities. Journal of Contaminant Hydrology 70, 117e145. Martienssen, M., Fabritius, H., Kukla, S., Balcke, G.U., Hasselwander, E., Schirmer, M., 2006. Determination of naturally occurring MTBE biodegradation by analysing metabolites and biodegradation by-products. Journal of Contaminant Hydrology 87, 37e53. Mleczek, M., Rutkowski, P., Rissmann, I., Kaczmarek, Z., Golinski, P., Szentner, K., Strazynska, K., Stachowiak, A., 2010. Biomass productivity and phytoremediation potential of Salix alba and Salix viminalis. Biomass and Bioenergy 34, 1410e1418. Moreels, D., Bastiaens, L., Ollevier, F., Merckx, R., Diels, L., Springael, D., 2006. Evaluation of the intrinsic methyl tertbutyl ether (MTBE) biodegradation potential of hydrocarbon contaminated subsurface soils in batch microcosm systems. FEMS Microbiology Ecology 49, 121e128. Park, K., 1999. Biodegradation of the Fuel Oxygenate, Methyl Tertbutyl Ether (MTBE), and Treatment of MTBE Contaminated Ground Water in Laboratory Scale Reactors. Ph.D. dissertation, State University of New Jersey, New Jersey, USA. Raynal, M., Pruden, A., 2008. Aerobic MTBE biodegradation in the presence of BTEX by two consortia under batch and semibatch conditions. Biodegradation 19, 269e282. Reiche, N., Lorenz, W., Borsdorf, H., 2010. Development and application of dynamic air chambers for measurement of volatilization fluxes of benzene and MTBE from constructed

5074

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3 e5 0 7 4

wetlands planted with common reed. Chemosphere 79, 162e168. Rentz, J.A., Alvarez, P.J.J., Schnoor, J.L., 2005. Benzo[a]pyrene cometabolism in the presence of plant root extracts and exudates: implications for phytoremediation. Environmental Pollution 136, 477e484. Rubin, E., Ramaswami, A., 2001. The potential for phytoremediation of MTBE. Water Research 35, 1348e1353. Schmidt, T.C., Schirmer, M., Weiss, H., Haderlein, S.B., 2004. Microbial degradation of methyl tert-butyl ether and tert-butyl alcohol in the subsurface. Journal of Contaminant Hydrology 70 (3e4), 173e203. Scholz, M., 2006. Wetland Systems to Control Urban Runoff. Elsevier, Amsterdam. Sedran, M.A., Pruden, A., Wilson, G.J., Suidan, M.T., Venosa, A.D., 2002. Effect of BTEX on degradation of MTBE and TBA by mixed bacterial consortium. Journal of Environmental Engineering 128 (9), 830e835. Squillace, P.J., Zogorski, J.S., Wilber, W.G., Price, C.V., 1996. Preliminary assessment of the occurrence and possible sources of MTBE in groundwater in the United States,

1993e1994. Environmental Science and Technology 30 (5), 1721e1730. Sutherland, J., Adams, C., Kekobad, J., 2004. Treatment of MTBE by air stripping, carbon adsorption, and advanced oxidation: technical and economic comparison for five groundwaters. Water Research 38, 193e205. USEPA, 2004. Technologies for Treating MTBE and Other Fuel Oxygenates. Office of Superfund Remediation and Technology Innovation, Washington, DC, USA. EPA-542/R-04-009. USEPA, 2005. List of Drinking Water Contaminants and MCLs From: http://water.epa.gov/drink/index.cfm. Wilhelm, M., Adams, V., Curtis, J., 2002. Carbon adsorption and air-stripping removal of MTBE from river water. Journal of Environmental Engineering 128 (9), 813e823. Wilson, R.D., Mackay, D.M., Scow, K.M., 2002. In situ MTBE biodegradation supported by diffusive oxygen release. Environmental Science and Technology 36, 190e199. Yerushalmi, L., Lascourreges, J.F., Guiot, S.R., 2002. Kinetics of benzene biotransformation under microaerophilic and oxygen-limited conditions. Biotechnology and Bioengineering 79 (3), 347e355.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 7 5 e5 0 8 3

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Reducing the startup time of aerobic granular sludge reactors through seeding floccular sludge with crushed aerobic granules Maite Pijuan a,b, Ursula Werner a, Zhiguo Yuan a,* a b

The University of Queensland, Advanced Water Management Centre (AWMC), Brisbane QLD 4072, Australia Catalan Institute for Water Research (ICRA), Scientific and Technological Park of the University of Girona, Girona 17003, Spain

article info

abstract

Article history:

One of the main challenging issues for the aerobic granular sludge technology is the long

Received 30 December 2010

startup time when dealing with real wastewaters. This study presents a novel strategy to

Received in revised form

reduce the time required for granulation while ensuring a high level of nutrient removal.

28 June 2011

This new approach consists of seeding the reactor with a mixture of crushed aerobic

Accepted 5 July 2011

granules and floccular sludge. The effectiveness of the strategy was demonstrated using

Available online 14 July 2011

abattoir wastewater, containing nitrogen and phosphorus at approximately 250 mgN/L and 30 mgP/L, respectively. Seven different mixtures of crushed granules and floccular sludge

Keywords:

at granular sludge fractions (w/w in dry mass) of 0%, 5%, 10%, 15%, 25%, 30% and 50% were

Aerobic granules

used to start eight granulation processes. The granulation time (defined as the time when

Startup

the 10th percentile bacterial aggregate size is larger than 200 mm) displayed a strong

Seeding

dependency on the fraction of granular sludge. The shortest granulation time of 18 days

Crushed granules

was obtained with 50% crushed granules, in comparison with 133 days with 5% crushed

Nutrient removal

granules. Full granulation was not achieved in the two trials without seeding with crushed

Nutrient rich wastewater

granules. In contrast to the 100% floccular sludge cases, where a substantial loss of biomass occurred during granulation, the biomass concentration in all other trails did not decrease during granulation. This allowed that good nitrogen removal was maintained in all the reactors during the granulation process. However, enhanced biological phosphorus removal was achieved in only one of the eight trials. This was likely due to the temporary accumulation of nitrite, a strong inhibitor of polyphosphate accumulating organisms. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Aerobic granulation has been extensively studied during the last decade using sequencing batch reactors (SBRs). Aerobic granules are dense microbial aggregates that present much faster settling velocities than conventional floccular sludge, thus allowing operating reactors at a higher biomass concentration and reducing the volumetric requirement for

phase separation (de Bruin et al., 2004; Liu and Tay, 2004). Many studies have been done exploring the granulation process using simple synthetic wastewaters where only carbon removal occurred (Tay et al., 2001a,b). In these cases, full granulation was achieved typically after several days of operation. Aerobic granules have also been used to treat various types of real wastewater including nutrient rich abattoir wastewater (Yilmaz et al., 2008; Cassidy and Belia,

* Corresponding author. E-mail address: [email protected] (Z. Yuan). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.009

5076

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 7 5 e5 0 8 3

2005), dairy effluent (Arrojo et al., 2004) and domestic wastewater (Ni et al., 2009; Liu et al., 2010). However, the formation of aerobic granules with nutrient removal capabilities can take several months. In Yilmaz et al. (2008), 170 days were needed to adapt aerobic granules developed with synthetic wastewater to abattoir wastewater. In Ni et al. (2009), 300 days were needed to achieve 85% granulation in a reactor treating domestic wastewater. Schwarzenbeck et al. (2004) spent 147 days to reach complete granulation in a reactor achieving carbon removal from malting wastewater. Figueroa et al. (2008) required 75 days to get mature granules with effluent from an anaerobic digester treating wastewater from a fishcanning factory and hardly any nitrogen removal was observed during the first 60 days. Such long startup times are a major challenge to be addressed for this technology to gain wide application in wastewater treatment. Some research has been conducted to enhance granule formation and reduce the startup time. Linlin et al. (2005) investigated the cultivation of aerobic sludge granules by seeding the reactor with anaerobic granules, taking advantage of the availability of anaerobic granules. However, these granules broke up under aerobic conditions and most of them were washed out before a new granulation occurred. Similarly, Muda et al. (2010) seeded anaerobic granules in floccular sludge to start the granulation process in a reactor treating synthetic textile wastewater. They also observed disintegration of the anaerobic granules during the first few days of operation; almost half of the sludge was washed-out from the reactor causing rapid decrease in biomass concentration. A novel startup strategy consisting of seeding floccular sludge with crushed aerobic granules is investigated in this study. Several combinations of crushed aerobic granules with floccular sludge were used to seed different SBRs treating abattoir wastewater. Size distribution, biomass concentration and nutrient removal performance were monitored for several months, which allowed assessing the effectiveness of the strategy in reducing the granulation time and in maintaining the nutrient removal capability during granulation. The effect of the size of crushed granules on granulation was also assessed.

2.

Materials and methods

2.1.

Sludge sources

Aerobic granules used as seeding sludge were sampled from a lab-scale sequencing batch reactor (SBR) treating abattoir wastewater. At the time of sampling, the reactor was achieving stable removal of soluble chemical oxygen demand (COD), soluble nitrogen (N) and soluble phosphorus (P) with efficiencies of 85%, 93%, and 89%, respectively, from abattoir wastewater (composition to be described below). The granules had a mean size of 769 mm. More details about the reactor operation, size distribution of granules and the nutrient removal performance can be found in Yilmaz et al. (2008). Floccular sludge used was obtained from a full-scale wastewater treatment plant (WWTP) performing biological COD, N and P removal from domestic wastewater in Queensland, Australia.

Aerobic granules used as a seeding sludge were manually crushed in order to reduce their size and obtain more particles from fewer granules. These granules were pressed through certified sieves. The pore size of the sieves varied between experiments, as will be further described blow.

2.2. Reactor operation to determine the effect of granular sludge fraction Seven SBRs, seeded with different mixtures of granular and floccular sludges, were operated to determine the effect of granular sludge fraction on the granulation time and reactor performance during startup. They each had a working volume of 2 L, with a diameter of 7 cm and a height of 76 cm. These reactors were operated in a temperature-controlled room (20e23
C). Their mixing was carried out via a combination of a magnetic stirring (200 rpm) and intermittent sparging of nitrogen gas (10 s on, 15 s off) during anaerobic/anoxic periods, or air in aerobic period with DO controlled between 1.5 and 2.0 mg/L. The gas flow rate was 1 L/min. Six reactors were seeded with a mixture of crushed granules and floccular sludge, with the fraction (weight/weight of dry biomass) of granular sludge being 5%, 10%, 15%, 25%, 30% and 50%, respectively. The 7th SBR was seeded with only floccular sludge. This reactor was inoculated twice in a period of 6 months. The wastewater loading per cycle was gradually increased from 0.25 to 0.5 L at the beginning of the reactor operation to 1 L, which increased the volumetric exchange ratio (VER) from 12.5 to 25%e50% and the organic loading rate (OLR) to 0.46, 0.91 and 1.83 g COD/Ld respectively. The nitrogen loading rate (NLR) also increased from 0.097 to 0.195 and 0.39 g TN/Ld with the increase of VER. The height of the effluent withdrawal was 6.5 cm, 13 cm and 26 cm when 12.5, 25 and 50% VER was applied respectively. At the same time, the settling time was progressively reduced. The initial settling time applied in all the reactors was 20 min. During the first 10 days the settling time was progressively reduced to 10 min and during the following 10 days, the settling time was reduced to 5 min. A further reduction to 2 min settling time was only applied when the 50th percentile of the particles in the reactor was higher than 200 mm. After the reactor achieved complete granulation, the settling time was reduced to 1.5 min and kept constant until the end of operation. The SBRs had an 8 h cycle consisting of the 1st anaerobic phase with a wastewater pulse feed (75%), 1st aerobic phase, 1st anoxic, 2nd anaerobic with wastewater pulse feed (25%), 2nd aerobic, 2nd anoxic, and a settling, decant and idle phase. The lengths of these periods were adjusted in each reactor depending on the treatment capabilities of the reactor, the wastewater loading and the sludge settling properties. The anaerobicanoxic to aerobic phases ratio was around 50%. pH in all reactors were recorded but not controlled, and fluctuated between 6.8 and 8.6. A sieve with a pore size of 500 mm was used to obtain crushed granules. The 10th, 50th and 90th percentiles of the crushed granules were 168 mm, 512 mm and 917 mm, respectively. The performance of the reactors was monitored with weekly cycle studies. During each study, liquid phase samples were collected at an interval of 30e60 min and immediately

5077

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 7 5 e5 0 8 3

filtered through disposable Millipore filter units (0.22 mm pore size) for the analyses of ammonia, nitrite and phosphate (methods given below). The mixed liquor suspended solids (MLSS) and mixed liquor volatile suspended solids (MLVSS) concentrations as well as size distribution of the particles were measured in triplicates three times per week, with methods to be further described below.

2.3. Reactor operation to determine the effect of the size of crushed granules A set of experiments was conducted to establish the effect of the size of the crushed granules on the granulation process. Intact granules were pressed through three certified sieves with pore size of 1000 mm, 500 mm and 180 mm, respectively to obtain granules with different size distributions, as shown in Table 1. The 500 mm case is the same as the one described in the previous section. Three SBRs were seeded with the above obtained crushed granules along with floccular sludge at a granular fraction of 30% (weigh/weigh of dry biomass), to determine the effect of the size of the crushed granules on the granulation process. The operation and monitoring of the reactors were as described above.

stored at 4
C. The characteristics of the anaerobic pond effluent are detailed in Table 2.

2.6.

Analyses

  Ammonium ðNHþ 4  NÞ, nitrate ðNO3  NÞ, nitrite ðNO2  NÞ 3 and orthophosphate ðPO4  PÞ were analysed using a Lachat QuikChem8000 Flow Injection Analyser (Lachat Instrument, Milwaukee). Total and soluble chemical oxygen demand (CODt and CODs, respectively), total Kjeldahl nitrogen (TKN), total phosphorus, mixed liquor suspended solid (MLSS) and volatile MLSS (MLVSS) were analysed according to the standard methods (APHA, 1995). VFAs were measured by PerkineElmer gas chromatography. To determine the size distribution of the particles in each SBR, 30 mL of mixed liquor were pumped through a Malvern laser light scattering instrument, Mastersizer 2000 series (Malvern Instruments, Worcestershire, UK). The technique of laser diffraction is based on the principle that particles passing through a laser beam will scatter light at an angle that is directly related to their sizes. This method represents a rapid and robust measurement of particle sizes present in a bulk. The measurable size range is 0.02e2000 mm. The granule morphology was qualitatively observed using a stereo microscope (Olympus SZH10).

2.4. Reactor operation to determine the effect of adding intact granules A set of experiments was conducted to establish the effect of adding intact instead of crushed granules on the granulation process. Two SBRs were seeded with intact granules and floccular sludge at a granular fraction of 20% (weigh/weigh of dry biomass). The difference between them was the size of the intact granules used in each SBR, as shown in Table 1. The operation and monitoring of both reactors were as described above.

2.5.

Abattoir wastewater

The wastewater used in the reactors operation was from a local abattoir in Queensland, Australia. At this site, the raw effluent passes through anaerobic ponds before being treated in an SBR for biological COD and N removal. Anaerobic pond effluent from the abattoir was collected on a weekly basis and

Table 1 e Size distribution of crushed granules obtained with three different sieves and 2 different types of intact granules used.

1000 mm sieve 500 mm sieve 180 mm sieve “Big” intact granules “Small” intact granules

10th percentile (mm)

50th percentile (mm)

90th percentile (mm)

359 168 101 923

704 512 235 1268

1215 917 499 1646

441

727

1184

3.

Results and discussion

3.1. Development of aerobic granules from 100% floccular sludge with abattoir wastewater Two different startup trials were carried out using floccular sludge as the sole inoculum. The settling time was progressively reduced in order to select for fast settling bacterial aggregates. This approach is commonly used and has been proved to enhance granulation (Qin et al., 2004; Cassidy and Belia, 2005). However, during the application of this strategy, a fast decrease in biomass concentration occurred in the reactor in both runs (Fig. 1). The appearance of some granules was observed when biomass concentration decreased from the initial 3.0 g MLSS/L to levels lower than 1.0 g MLSS/L. After that, granules increased in size but the biomass concentration failed to recover in 2.5 months. Also, full granulation was not achieved in neither run, as indicated by the fact that d(0.1) was below 200 mm at all times.

Table 2 e Characteristics of the wastewater used in reactor operation. Parameters CODtotal (mg/L) CODsoluble (mg/L) VFA (mg/L) TN (mg/L) TP (mg/L) NeNH4 (mg/L) NeNOx (mg/L) P-PO4 (mg/L)

Pond average 1081e1356 862e1137 650e800 245e275 36e47 200e254 0 31e40

5078

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 7 5 e5 0 8 3

granule size (um)

1400 1200

A

B

C

D

1000 800 600 400 200 0 4000 3500

mg/L

3000 2500 2000 1500 1000 500 0 0

10

20

30

40

50

60

70

80 0

10

20

30

40

50

60

70

80

Time (days)

Time (days)

Fig. 1 e Size distribution profiles (top) and MLSS & MLVSS (down) of the SBRs seeded with 100% floccular sludge: A, C - 1st run; B, D - 2nd run. Percentiles: ;d(0.9), B d(0.5), C d(0.1); , MLVSS, - MLSS.

Soluble COD removal was achieved in both runs with an efficiency of 99% even though the biomass concentration was low. During the first run (Fig. 2A), the volumetric exchange ratio (VER) was set to 33% and was kept constant. However, nitrification deteriorated due to a decrease of biomass in the SBR, which caused accumulation of NHþ 4 in the reactor. Nitrification did not recover until the reactor was stopped after 80 days of operation. The high-level of ammonium concentration could have inhibited the development of nitrifiers (Vadivelu et al., 2006). In run 2 (Fig. 2B), the initial VER applied was 17% (lower than in run 1) in order to avoid NHþ 4 accumulation. During the first 25 days, 90% N removal was achieved. The wastewater loading was then slightly increased, leading to a VER of 25%. The settling time was maintained at 8 min. However, the system could not cope with this increase in loading. The biomass concentration was decreasing, leading to decreased N removal. Although VER

mg N/L

250

3.2. Granulation and nutrient removal performance with mixed seed of floccular sludge and crushed aerobic granules Six different combinations of crushed granules and floccular sludge were used to start six aerobic granular sludge reactors. Fig. 3 shows the size distribution profiles of the SBRs. The 90th percentile was always substantially higher than the 50th and 10th percentiles due to the presence of the crushed granules. The 90th percentiles increased since the beginning of operation. After a period of time, which varied depending on the reactor, the 50th percentile started to increase to values 100

B

A

80

200 60 150 40 100 20

50 0

exchange ratio (%)

300

was reduced again, the performance did not substantially improve and biomass concentration reached very low levels (<500 mg/L). This run was stopped after 70 days of operation.

0 0

10

20

30

40

50

Time (days)

60

70

80 0

10

20

30

40

50

60

70

80

Time (days)

Fig. 2 e Nitrogen removal performance of the SBR seeded with 100% floccular sludge: A - 1st run; B - 2nd run. C NLNHD 4 influent, B NLNHD 4 effluent, ; NeNOx effluent, e exchange ratio.

5079

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 7 5 e5 0 8 3

1600

A

Particle size (um)

1400

B

C

1200 1000 800 600 400 200 0 0

20

40

60

80

100 120 140 160

0

20

40

60

1600

D

Particle size (um)

1400

80

100 120 140 160

0

20

40

time (days)

time (days)

60

80

100 120 140 160

time (days)

E

F

1200 1000 800 600 400 200 0 0

10

20

30

40

50

60

70

80

90

0

10

20

30

time (days)

40

50

t (days)

60

70

80

90

0

10

20

30

40

50

60

70

80

90

time (days)

Fig. 3 e Size distribution profiles of SBRs seeded with different percentages of crushed granules: A - 5%; B - 10%; C - 15%; D 25%; E - 30%; F - 50%. Percentiles: ;d(0.9), B d(0.5), C d(0.1). Note the different time scales for the lower and higher percentages of crushed granules.

higher than 200 mm, indicating an increase in the number of granules in the reactor. The reactor was considered fully granular when the 10th percentile achieved values higher than 200 mm, which is the minimum size for a particle to be considered an aerobic granule (de Kreuk et al., 2007). At this stage, the appearance of the sludge was completely granular with flocs barely observable. As an example, Fig. 4 shows the morphology of the sludge at the beginning of operation and after complete granulation in one of the reactors. The relationship between the time of granulation and the percentage of crushed granules in the seeding sludge is presented in Fig. 5. The dependency can be reasonably described by an exponential function.

3.3.

Biomass concentration and nutrient removal

In all reactors, the biomass wastage was not controlled but occurred through sludge washout during decanting. With this strategy, just the biomass with good settling properties remained in the reactor. Although some biomass washout occurred due to the progressive decrease in the settling time and the increase in the VER, the net biomass concentration did not decrease in any of the SBRs with mixed seeding (Fig. 6). When the majority of the biomass was granular, a faster increase in the biomass concentration was observed. All the SBRs started with a VER between 12.5 and 25%. Having a higher VER has been suggested to promote

Fig. 4 e Stereo microscope images of the morphology of the sludge at the beginning and in the last week of operation from the SBR seeded with 10% crushed granules. Scale bar [ 1 mm.

5080

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 7 5 e5 0 8 3

ammonium accumulated until nitrogen removal recovered. All the reactors achieved relatively stable nitrogen removal during the granulation period. A 50% VER (corresponding to 16 h HRT) was achieved in all cases with a nitrogen removal efficiency being 90%. In contrast, significant biological phosphorus removal was only achieved in the SBR seeded with 15% crushed granules (Fig. 8). Biological phosphorus removal from abattoir wastewater is challenging due to the high levels of ammonia in the influent producing high levels of nitrite and nitrate during nitrification, which are detrimental for establishing stable EBPR conditions (Lemaire et al., 2009; Pijuan and Yuan, 2010).

Time of granulation (days)

140 y=168.4*e-0.071x r2 = 0.90

120 100 80 60 40 20 0 0

20

40

60

80

100

% of granular seeding (w/w)

3.4. Effect of different sizes of crushed and intact granules

granulation because more supernatant can be discharged in one cycle, and more of the non-fast settling biomass is washed out from the reactor (Wang et al., 2006). However, a correlation between the increase in VER and the granulation time was not observed in this study. Different granulation times were observed in the reactors seeded with 5e15e25e30e50% crushed granules while the VER was increased in a similar fashion in all these reactors, reaching 50% VER by day 20 of operation (Fig. 7). When dealing with nutrient reach wastewater, the nutrient removal capability of the reactor has to be taken into account before increasing the VER. Fig. 7 shows the nitrogen removal performance of the six SBRs during the granulation period. Increase in wastewater loading was implemented progressively in all reactors ensuring that nitrogen removal was not compromised. In one case, the loading rate was decreased temporarily when

In order to investigate the possible effect of the size of the crushed granules on the overall granulation time, three SBRs were inoculated with 30% crushed granular sludge combined with 70% floccular sludge. The only difference between them was the size distribution of the crushed granules (Table 1). Fig. 9 shows the size distribution profiles of the three SBRs along their operation. The SBRs seeded with granules crushed through 500 mm and 1000 mm sieves achieved full granulation in 35e40 days. They displayed a similar increase in biomass concentration during granulation (Fig. 9). High-level of nitrogen removal (at around 90%) was maintained in both reactors throughout the experimental period (data not shown). In comparison, the SBR seeded with the smallest crushed granules (crushed through a 180 mm sieve) required almost 80 days to reach full granulation. In this reactor, biomass concentration had a slower increase compared to the other two reactors due to higher biomass lost during decanting periods. This also caused

g/L

Fig. 5 e Dependence of the granulation time on the percentage of crushed granules in the seeding sludge.

20,0 17,5 15,0 12,5 10,0 7,5 5,0 2,5 0,0

A

0

20,0

B

20 40 60 80 100 120 140 160 Time (days)

D

17,5

0

C

20 40 60 80 100 120 140 160

0

20 40 60 80 100 120 140 160

Time (days)

Time (days)

E

F

g/L

15,0 12,5 10,0 7,5 5,0 2,5 0

10 20 30 40 50 60 70 80 Time (days)

0

10 20 30 40 50 60 70 80 Time (days)

0

10 20 30 40 50 60 70 80 Time (days)

Fig. 6 e MLSS and MLVSS of SBRs seeded with different percentages of crushed granules: A - 5%; B - 10%; C - 15%; D - 25%; E 30%, F - 50%. B MLVSS; C MLSS. Note the different time scales for the lower and higher percentages of crushed granules.

5081

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 7 5 e5 0 8 3

A

300

B

100

C

80

200

60

150

40

100 20

50 50%

0 0

20

40

60

80 100 120 140 160

0

20

40

D

300

80 100 120 140 160

0

20

40

Time (days)

Time (days) 350

60

60

0 80 100 120 140 160

Time (days)

E

100

F

80

mg N/L

250 200

60

150

40

100 20

50

exchange ratio (%)

mg N/L

250

exchange ratio (%)

350

0

0 0

10

20

30

40

50

60

70

80

0

Time (days)

10

20

30

40

50

Time (days)

60

70

80 0

10

20

30

40

50

60

70

80

Time (days)

Fig. 7 e Nitrogen removal performance of SBRs seeded with different percentages of crushed granules: A- 5%; B- 10%; C- 15%; D B NH LN and NOxeN effluent; e exchange ratio. D- 25%; E- 30%, F- 50% C NHD 4 LN influent; 4

deterioration on nitrogen removal achieving only 40% removal during the first month (data not shown). To investigate the effect of adding intact granules instead of crushed granules in the seeding sludge, two SBRs were inoculated with 20% intact granules and 80% floccular sludge. The only difference between them was the size of the intact granules used as described in Table 1. The SBR inoculated with the smallest intact granules achieved fully granulation after 62 days of operation while the SBR inoculated with the biggest granules needed 102 days. These times are higher than the ones obtained when crushed granules are used (Fig. 5). This suggests that using crushed granules instead of intact granules has a positive effect on reducing the start-up time. However, more research is needed to fully understand the physical and microbial interactions between floccular sludge and crushed granules.

3.5.

Practical implications and future research

A major problem was encountered when an aerobic granular reactor was started with floccular sludge as the seed for the treatment of nutrient rich wastewater. Granules were not observed until the biomass concentration reached very low values (<1 g/L). In this process, nitrifiers, which are known to be slow growers, were washed out. The granular sludge thus obtained was able to remove COD only. Many other researchers have observed a similar situation when trying to develop granules using different types of wastewater. Arrojo et al. (2004) investigated the use of aerobic granules to treat industrial wastewater produced in a laboratory for analysis of dairy products. During the first 7 days of operation, they reported an almost washout of the suspended biomass from

the 2 SBRs they operated. Also, they only started to observe nitrification after 1 month of operation and N removal was only obtained after day 150. Figueroa et al. (2008) reported a similar situation when trying to develop granules with fish canning industry wastewater. A large part of biomass was lost during the first days of operation. Some granules were observed by day 45 but only COD removal could be achieved. The use of a mixture of crushed granules and floccular sludge substantially reduced the granulation time. Equally importantly, this strategy also enabled nitrogen removal through the granulation period. This was attributed to the stable biomass concentration during the startup of granulation. This study therefore delivered a strategy that enables the conversion of a floccular system into a granular one while maintaining its nitrogen removal capability. Obviously, a significant source of granular sludge is needed to enable the use of this strategy. As expected, the higher the crushed granules fraction is, the faster the granulation process. However, there are few granular wastewater treatment plants worldwide at present and therefore, the usage of a sludge combination with a lower percentage of crushed granules is more realistic. One option is to initially develop granular sludge at pilot scales. A gradual scaling-up may then be achieved. While the experimental data clearly demonstrated that adding crushed aerobic granules to floccular sludge substantially enhanced the granulation process, the mechanisms involved are not immediately clear. In comparison to the two trials with 100% floccular sludge, there was much less loss or even no loss of MLSS and MVSS during granulation in the cases where crushed granules were added. This indicated that the addition of crushed granules enhanced the settleability of

5082

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 7 5 e5 0 8 3

mg P/L

40

80

30

60

20

40

10

20

0 0

20 40 60 80 100 120 140 160 Time (days)

50

0

20 40 60 80 100 120 140 160 0 Time (days)

0 20 40 60 80 100 120 140 160 Time (days)

100

F

E

D

40

m g P/ L

100

C

B

exchange ratio (%)

A

80

30

60

20

40

10

20

exchange ratio (%)

50

0

0 0

10 20 30 40 50 60 70 80

0

10 20 30 40 50 60 70 80

0 10 20 30 40 50 60 70 80 90

Time (days)

Time (days)

Time (days)

Fig. 8 e Phosphorus removal performance of SBRs seeded with different percentages of crushed granules: A- 5%; B- 10%; C3L B PO 15%; D- 25%; E- 30%; F- 50%. C PO3L 4 LP influent; 4 LP; e exchange ratio.

flocs, and as a result a quick loss of floccular sludge, as occurred in the two 100%-floccular sludge trials, was avoided. It is possible that some form of granule-floc matrices developed when the granular and floccular sludges were mixed. Due to the incorporation of high-density granules, such aggregates could have had a settling velocity that was higher than that of typical flocs, and were thus retained in the system despite of the short settling time. Consequently, the nitrogen and COD removal capability of the sludge was maintained.

granule size (um)

1600 1400

The roles of the granule-floc matrices, and indeed crushed granules and flocs in general, in the subsequent granulation process remain to be clarified. Further research is required to elucidate the details of the interactions between seeding granules and flocs, and their respective roles in the granule development process. Crushed rather than intact granules were used in this study because they seemed to provide shorter start-up times. The size of these crushed granules was also found to impact

A

B

C

D

E

F

1200 1000 800 600 400 200 20,0 17,5 15,0

g/ L

12,5 10,0 7,5 5,0 2,5 0,0 0

10

20

30

40

50

Time (days)

60

70

80 0

10 20 30 40 50 60 70

Time (days)

80 0

10 20 30 40 50 60 70 80 90

Time (days)

Fig. 9 e Size distribution profiles (top) and MLSS and MLVSS (bottom) of the SBRs inoculated with 30% granules crushed through sieves with three different pore sizes: 1000 mm (A, D), 500 mm (B, E) and 180 mm (C, F).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 7 5 e5 0 8 3

on the speed of granulation. However, this study did not reveal the optimal size of the seeding sludge. Fig. 9 showed that, when the seeding granules were too small, the strategy became less effective. It could be speculated that a too large size could also reduce the effectiveness of the strategy due to reduced number of granules. Further experimental studies are required to fully reveal the dependency of granulation on the size of seeding granules.

4.

Conclusions

A novel startup strategy for aerobic granulation was proposed and demonstrated to significantly speed up the granulation process while maintaining the nitrogen removal capacity of the system. The strategy, which consists of seeding a granulation reactor with a mixture of crushed aerobic granules and floccular sludge, effectively avoids biomass loss that typically occurs during granulation. It thus maintains the pollutant removal capability of the sludge during granulation. The granulation time is strongly dependent on the fraction of the crushed granules in the seeding sludge, with a higher fraction leading to faster granulation. The interactions between crushed granules and floccular sludge during granulation and their contribution to the faster granulation remain to be elucidated.

Acknowledgements This work was funded by the Environmental Biotechnology Cooperative Research Centre (EBCRC) Pty Ltd, Australia and the Department of Innovation Industry Science and Research from the Australian Government via the International Science Linkages program. M. Pijuan acknowledges the Ramon y Cajal research fellowship (RYC-2009-04959) provided by the Spanish Government. The authors wish to thank Dr Victor Arias for his help with reactor operation.

references

Arrojo, B., Mosquera-Corral, A., Garrido, J.M., Mendez, R., 2004. Aerobic granulation with industrial wastewater in sequencing batch reactors. Water Research 38 (14e15), 3389e3399. Cassidy, D.P., Belia, E., 2005. Nitrogen and phosphorus removal from abattoir wastewater in a SBR with aerobic granular sludge. Water Research 39 (19), 4817e4823. de Bruin, L.M.M., de Kreuk, de Kreuk, M.K., van der Roest, H.F.R., Uijterlinde, C., van Loosdrecht, M.C.M., 2004. Aerobic granular sludge technology: an alternative to activated sludge? Water Science and Technology 49 (11e12), 1e7.

5083

de Kreuk, M.K., Kishida, N., Van Loosdrecht, M.C.M., 2007. Aerobic granular sludge e state of the art. Water Science and Technology 55 (8e9), 75e81. Figueroa, M., Mosquera-Corral, A., Campos, J.L., Mendez, R., 2008. Treatment of saline wastewater in SBR aerobic granular reactors. Water Science and Technology 58 (2), 479e485. Lemaire, R., Yuan, Z., Bernet, N., Marcelino, M., Yilmaz, G., Keller, J., 2009. A sequencing batch reactor system for highlevel biological nitrogen and phosphorus removal from abattoir wastewater. Biodegradation 20 (3), 339e350. Linlin, H., Jianlong, W., Xianghua, W., Yi, Q., 2005. The formation and characteristics of aerobic granules in sequencing batch reactor (SBR) by seeding anaerobic granules. Process Biochemistry 40, 5e11. Liu, Y., Tay, J.H., 2004. State of the art of biogranulation technology for wastewater treatment. Biotechnology Advances 22, 533e563. Liu, Y.Q., Moy, B., Kong, Y.H., Tay, J.H., 2010. Formation, physical characteristics and microbial community structure of aerobic granules in a pilot-scale sequencing batch reactor for real wastewater treatment. Enzyme and Microbial Technology 46, 520e525. Muda, K., Aris, A., Salim, M.R., Ibrahim, Z., Yahya, A., van Loosdrecht, M., Ahmad, A., Nawahwi, M.Z., 2010. Development of granular sludge for textile wastewater treatment. Water Research 44, 4341e4350. Ni, B.J., Xie, W.M., Liu, S.G., Yu, H.Q., Wang, Y.Z., Wang, G., Dai, X. L., 2009. Granulation of activated sludge in a pilot-scale sequencing batch reactor for the treatment of low strength municipal wastewater. Water Research. 43, 751e761. Pijuan, M., Yuan, Z., 2010. Development and optimization of a sequencing batch reactor for nitrogen and phosphorus removal from abattoir wastewater to meet irrigation standards. Water Science and Technology 61 (8), 2105e2112. Qin, L., Tay, J.H., Liu, Y., 2004. Selection pressure is a driving force of aerobic granulation in sequencing batch reactors. Process Biochemistry 39, 579e584. Schwarzenbeck, N., Erley, R., Mc Swain, B., Wilderer, P., Irvine, R., 2004. Treatment of malting wastewater in a granular sludge sequencing batch reactor (SBR). Acta Hydrochimica et Hydrobiologica 32 (1), 16e24. Tay, J.H., Liu, Q.S., Liu, Y., 2001a. Microscopic observation of aerobic granulation in sequential aerobic sludge blanket reactor. Journal of Applied Microbiology 91, 168e175. Tay, J.H., Liu, Q.S., Liu, Y., 2001b. The role of cellular polysaccharides in the formation and stability of aerobic granules. Letters in Applied Microbiology 33, 222e226. Vadivelu, V., Keller, J., Yuan, Z., 2006. Effect of free ammonia and free nitrous acid concentration on the anabolic and catabolic processes of an enriched nitrosomonas culture. Biotechnology and Bioengineering 95 (5), 830e839. Wang, H.L., Yu, G.L., Liu, G.S., Pan, F., 2006. A new way to cultivate aerobic granules in the process of papermaking wastewater treatment. Biochemical Engineering Journal 28, 99e103. Yilmaz, G., Lemaire, R., Keller, J., Yuan, Z., 2008. Simultaneous nitrification, denitrification, and phosphorus removal from nutrient-rich industrial wastewater using granular sludge. Biotechnology and Bioengineering 100, 529e541.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 8 4 e5 0 9 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Trihalomethane exposures in indoor swimming pools: A level III fugacity model Roberta Dyck a,*, Rehan Sadiq a, Manuel J. Rodriguez b, Sabrina Simard b, Robert Tardif c a

University of British Columbia Okanagan, School of Engineering, Kelowna, BC, Canada E´cole supe´rieure d’ame´nagement du territoire, Universite´ Laval, Que´bec, QC, Canada c Universite´ de Montreal, De´partement de sante´ environnementale et sante´ au travail, Montreal, QC, Canada b

article info

abstract

Article history:

The potential for generation of disinfection byproducts (DBPs) in swimming pools is high

Received 25 November 2010

due to the concentrations of chlorine required to maintain adequate disinfection, and the

Received in revised form

presence of organics introduced by the swimmers. Health Canada set guidelines for

6 May 2011

trihalomethanes (THMs) in drinking water; however, no such guideline exists for swim-

Accepted 5 July 2011

ming pool waters. Exposure occurs through ingestion, inhalation and dermal contact in

Available online 13 July 2011

swimming pools. In this research, a multimedia model is developed to evaluate exposure concentrations of THMs in the air and water of an indoor swimming pool. THM water

Keywords:

concentration data were obtained from 15 indoor swimming pool facilities in Quebec

Swimming pools

(Canada). A level III fugacity model is used to estimate inhalation, dermal contact and

Exposure assessment

ingestion exposure doses. The results of the proposed model will be useful to perform

Level III fugacity model

a human health risk assessment and develop risk management strategies including

Disinfection byproduct

developing health-based guidelines for disinfection practices and the design of ventilation

Monte Carlo simulations

system for indoor swimming pools. ª 2011 Elsevier Ltd. All rights reserved.

Trihalomethanes (THMs)

1.

Introduction

In Canada, swimming is a popular activity for leisure and exercise and is ranked fourth among leisure activities, after walking, gardening and home exercise. Many of the swimmers using indoor public pools are children, pregnant women and seniors who may be at greater risk for health effects from chemical exposures in swimming pool water; therefore, it is important to quantify the associated exposure and risk. The objective of this paper is to develop an integrated model to evaluate exposure concentrations of trihalomethanes (THMs) in the air and water of an indoor swimming pool facility (natatorium). The fugacity approach is used to develop a multimedia environmental exposure model to assess the inhalation and dermal contact exposures, and minor

ingestion exposure. The results of the proposed model will be useful to perform a human health risk assessment and develop risk management strategies including the development of health-based guidelines for disinfection practices and the design of ventilation system for swimming pools.

1.1.

Disinfection byproducts (DBPs)

Chlorine has been used to disinfect drinking water for over a century and has helped eliminate most waterborne diseases in developed countries, such as typhoid, and cholera. Approximately 90% of the water supply systems in Canada use chlorine for disinfection purposes (Health Canada, 2009). During the disinfection process, reactions between natural organic matter in the source water and chlorine added to the

* Corresponding author. E-mail address: [email protected] (R. Dyck). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.005

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 8 4 e5 0 9 8

water generate disinfection byproducts (DBPs), such as THMs and haloacetic acids (HAAs). Since the discovery of DBPs in 1974 (Rook, 1974), many studies have been done on their formation, prevalence and associated risks. Of the more than 600 DBPs identified to date, very few have been the subject of exposure and toxicological studies (Richardson et al., 2007). The DBPs that are currently most studied are THMs and HAAs in part due to the availability of exposure and toxicology studies, and in part due to the fact that THMs are present in the highest concentrations, followed by HAAs (Who, 2000). THMs consist of four distinct but related compounds: chloroform (CHCl3), bromodichloromethane (CHCl2Br), dibromochloromethane (CHClBr2), and bromoform (CHBr3). For the purposes of testing and regulation, the concentrations of these four compounds are often added to form a parameter commonly referred to as total trihalomethanes (TTHMs). The presence of DBPs in water has been linked to an increased risk of bladder cancer (Villanueva et al., 2007), reproductive effects (Nieuwenhuijsen et al., 2000), and immediate respiratory effects such as asthma (Levesque et al., 2006; Goodman and Hays, 2008; Jacobs et al., 2007; Thickett et al., 2002). Health Canada (2008) has set guidelines for some groups of DBPs in drinking water: namely, TTHMs (0.10 mg/L), HAAs (0.08 mg/L), bromate (0.01 mg/L), bromodichloromethane (0.016 mg/L) and chlorite (1 mg/L), however, no such guideline is in use for swimming pool waters.

et al., 2007). The potential for DBP formation in swimming pools is high due to the concentrations of chlorine required for disinfection and the presence of organic matter from swimmers in the form of hair products, lotion, mucus, skin excretions and urine (Kim et al., 2002).

1.3.

Exposure to DBPs

In a swimming pool, although children may ingest a small amount of pool water, exposure to DBPs is mainly through inhalation and dermal contact (Whitaker et al., 2003). The amount and impact of the exposure may be influenced by many factors including: the type and dose of disinfectant, the resultant concentration of DBPs in the water, the number of swimmers, temperature of water and air (Chu and Nieuwenhuijsen, 2002), ventilation rates in the building, duration of swimming, and the turbulence in the water generated by moving-water features such as fountains, water slides, wave pools, and hot tub jets (Hery et al., 1995). For chlorination THMs, inhalation may constitute a significant impact due to the volatility of the compounds. In this study, exposures to THMs through dermal contact, ingestion and inhalation are modeled using concentrations of THMs detected in water samples collected from indoor pools in Quebec.

2. 1.2.

5085

Data collection

DBPs formation in swimming pools

Disinfection in swimming pools is important to reduce the risk of exposure to pathogens present in the water that originate either from the source water or from the swimmers in the water. Viruses, bacteria, parasites and fungus present in the water can cause health effects and even death (WHO, 2006). Chlorine is the most common disinfectant used for indoor pools (WHO, 2006). Chlorine is added to the pool waters as chlorine gas, calcium/sodium hypochlorite, or through electrolytic generation of sodium hypochlorite. Regardless of the method of application, once the chlorine is in the water it forms hypochlorous acid which dissociates into hydrogen atoms and hypochlorite ions. The “free chlorine” residual is the sum of the concentrations of hypochlorous acid, (HOCl), hypochlorite ion (OCl
), and aqueous chlorine (Cl2(aq)) (Weaver et al., 2009). The residual chlorine is measured frequently to ensure adequate disinfection power is retained in the pool throughout operation. Regulations in Quebec require free chorine in swimming pools to be between 0.8 and 2.0 mg/L (Ministe`re du De´veloppement Durable, de l’Environnement et des Parcs (MDDEP), 2006). In British Columbia (BC), the required amount of free chlorine varies with pH: 0.5 mg/L for pH of 7.4e7.8, and 1.0 mg/L for pH of greater than 7.8 (BC, 2010). Other parameters that are measured and regulated in swimming pools include fecal coliforms, Escherichia coli, Pseudomonas aeruginosa, Staphylococcus aureus, alkalinity, hardness, pH and turbidity. The formation of other DBPs is dependent on the presence of precursors which are usually naturally occurring organic material. A recent investigation in swimming pools in Spain identified more than 100 different DBPs in swimming pools, including brominated DBPs which are generally more mutagenic and carcinogenic than chlorinated DBPs (Richardson

The concentration data used to model environmental exposures were obtained from a study by Simard (2008). Pools were selected in various areas of Quebec City served by several different potable water distribution systems. The 15 indoor pools chosen were those most attended in each area with similar disinfection practices (chlorination). Indoor pools are considered due to the comparative ease of modeling the air exchange in the environment to account for volatilization of DBPs and also due to the presumed increased risk from the enclosed indoor pool environment. Samples were collected from each pool once per month over a year (Jan. 2007eFeb. 2008). Water samples were stored in accordance with the protocol suggested by MDDEP (2009). Analysis for chloroform, BDCM, DBCM and bromoform was performed by gas chromatograph with a mass spectrophotometry column in compliance with US EPA methods 524.2 and 552.2 (US EPA, 1995a,b). The risk analysis software @Risk (Palisade, 2010) and the Excel add-in EasyFitXL Version 5.3 (MathWave Technologies, 2010) were used to fit probability density functions to the results of the analysis of the water samples. Lognormal distribution was found to be the best fitted distribution in all cases. A summary of the results of the laboratory analyses is presented in Table 1 along with the means and standard deviations of the lognormal distributions. These distributions were used as input into the Monte Carlo simulations in the model to represent the concentrations in the water in the pools in Quebec as well as the variability in that data.

3.

Fugacity model

The concept of fugacity is used for describing the conditions of equilibrium among multimedia environments. Fugacity is

5086

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 8 4 e5 0 9 8

Table 1 e Results of laboratory analyses. Analyte Chloroform BDCM DBCM Bromoform Temperature

Detection limit mg/L

Range mg/L

# of samples (out of 176) below detection limit

Mean (m)

Standard deviation (s)

0.3 0.4 0.4 0.5 e

12.93e215 0e23.94 0e27.13 0e19.23 26.5e31.1

0 11 142 173 NA

55.2 1.23 0.26 0.26 28.2

31.6 2.55 1.94 1.41 0.88

NA e not applicable.

defined as the escaping tendency of a chemical to leave one medium in preference for another. At the low concentrations expected for environmental sampling, there is a linear relationship between concentration, fugacity ( f ) and fugacity capacity (Z ): C ¼ fZ

(1)

The mass balance equations used in fugacity modeling can include terms that correspond to chemical and biochemical reactions, inter-media transport, diffusion between media, and advection in or out of an “evaluative environment”. The use of fugacity makes it possible to consider complex interrelationships between environmental media such as air, water, and soil, as well as sediment underlying water, suspended solids within the water, and aerosols or particulates within the air. Biota in the form of fish, plants or humans can also be included as media (Mackay, 2001). Fugacity offers advantages over the use of dual-phase partition coefficients when there are several media and many processes occurring. Traditional partition coefficients can only describe equilibrium conditions between two media at a time, while fugacity models can generate equations to consider all the media at the same time. Four levels of fugacity models have been proposed (Mackay, 2001). Level III deals with steady state, but includes flow and non-equilibrium conditions.

3.1.

Evaluative environment

The first step in fugacity modeling is the definition of the evaluative environment. In the case of an indoor swimming pool, the environment consists of the water in the pool, the air above the pool within the natatorium, and the people in the pool (biota). During the regular operation of most community swimming pools, water is filtered and chlorinated during recirculation of the water from the pool (BC, 2010). Some water is lost during back-washing of the filters and by evaporation, therefore fresh water is added at a rate of approximately 1% of the total pool volume per day or 30 L per swimmer (MDDEP, 2006). Some DBPs may be adsorbed to suspended solids in the pool and removed with the filtration; however, for this study that process is neglected. The concentration of DBPs is assumed to remain constant and the processes of DBP formation due to swimmers and chlorination, as well as losses to filtration are assumed to be represented by that constant concentration of DBPs. The other component of the environment, the air, is also re-circulated and filtered. Many ventilation systems in natatoriums are operated with the goal of maintaining constant and comfortable humidity. The American Society of Heating,

Refrigerating and Air-Conditioning Engineers (ASHRAE, 2007) recommends maintaining the humidity between 40 and 60% and the air temperature 2e4  C higher than the temperature of the water. For energy efficiency and protection of structural elements, the humidity is removed from the air and a large portion of the air is returned to the natatorium. It is unclear what effect dehumidifying the air has on the concentration of DBPs in the air. In Quebec, the recommended rate of ventilation using outside air is 9 m3/h m2, based on the area of the pool surface and surrounding deck. The total ventilation including recirculated and outside air should equal 4e6 air changes per hour (ACH). For the “typical” pool facility considered here, 10e15% of the air entering the natatorium is outside air and the rest is re-circulated. For the purposes of modeling, the outside air is assumed to have negligible concentrations of DBPs, which would be the case, provided the exhaust and air intake have sufficient separation between them. A previous study has shown that inhalation of DBPs in aerosols contributes significantly to the average daily dose; however, that study focussed on HAAs which are less volatile than THMs (Xu and Weisel, 2003). In the present study we assumed that THMs are sufficiently volatile to not exist in any appreciable concentrations within aerosols of respirable size (<10 mm). Another assumption made in this study is that the air in the natatorium is completely and instantaneously mixed. A degree of variation of concentration and temperature with height in natatoriums has been shown (Hsu et al., 2009); however, considering such variation increases the complexity of the model. A schematic of the evaluative environment is presented in Fig. 1. The compartments are

Fig. 1 e System schematic of with flows and exposure routes. Evaluative environment used in the fugacity model showing model compartments, human exposure routes, inter-media transport and advective transport processes.

5087

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 8 4 e5 0 9 8

numbered as used in the model. The inter-media transport, human exposure routes, and advection processes are also highlighted. Depending on the amount absorbed, inhaled and ingested by the humans, the removal of contaminants when the people leave the pool can also be considered an advection process. For the typical pool in Quebec considered in this study, the size of the pool and natatorium is assumed from the dimensions of the pools that were sampled (where they were known) and other similar pools. The size and recirculation parameters of the typical pool considered in the model are presented in Table 2.

3.2.

Physico-chemical properties

Once the evaluative environment has been described, the chemicals of concern must be characterized in order to model their behavior in the environment. Chemical properties of THMs were determined from chemical handbooks and literature (Mackay, 2006). Where several values were presented for a chemical property, the multiple data points were fitted to a probability density function using the decision support software @Risk. The inclusion of multiple data points for these parameters reflects a degree of uncertainty in some of the chemical properties that were used as input into the model. As such, it was desirable to propagate the uncertainty in each of these parameters through the model using Monte Carlo simulations so that the final model results reflected the uncertainties in all of the input parameters. Some of the chemical properties were the same in all sources, and in that case deterministic (or “crisp”) values were used as model input. The chemical properties are presented in Table 3, with the properties for which distributions were used shown in shading. The parameter Z, the fugacity capacity constant, has units mol/m3 Pa. Each compartment of the natatorium environment has a fugacity capacity that is determined by the chemical properties of the contaminant. The equations for Z are shown in Table 3.

In their paper on unified dermal uptake model, McKone and Howd (1992) present a unitless partition coefficient between water and skin, KM. The values of KM are related to the octanolewater partition coefficient, KOW, by the empirically-derived equation in the notes below Table 3. The calculated fugacity capacities and parameters are also presented in Table 3.

3.3.

Removal from the environment

In a typical environmental fugacity model, removal of contaminants from the environment can occur in various ways including reactions, diffusion, and advection. In the case of a swimming pool, it is assumed that the predominant reaction occurring is the formation of the DBPs, which is already incorporated in the constant water concentration. The advective processes include the flow of the re-circulating air and water, and the movement of people in and out of the pool. The maximum allowable capacity calculated for our assumed pool size is 417 bathers (MDDEP, 2006); however, the maximum number of swimmers observed in any of the pools during the sampling was 100. The number of swims per month and minute per month spent swimming vary by age. Data from the US EPA Exposure Factors Handbook (US EPA, 2009) was fitted to distributions for calculating the average minutes per swim in order to vary the duration of exposure by age. For simplicity in the fugacity model, 1 h duration was used as an average time of swimming as recommended by US EPA (2009). This is done so that the flow of people in and out of the pool is simplified for all age groups. Therefore we can calculate that every hour 100 people leave the pool with THMs absorbed into their skin. The ages of the 100 people were distributed according to their minutes swimming per month. The volume of the people in the pool was calculated using age specific exposure factors (US EPA, 2009) and a relationship between height, surface area and weight (Sendrov and Collison, 1966). The volume and surface area were adjusted for the assumption that people do not have their head submerged (Xu and Weisel, 2005).

Table 2 e Evaluative environment parameters for Quebec pools and swimming pool facility characteristics in Modena, Italy (Fantuzzi et al., 2001). Swimming pool

Pool surface area (m2)

Pool volume (m3)

Natatorium area (m2)

Air volume of natatorium (m3)

Ventilation rate (ACH)

Maximum # swimmers during sampling

4e6e

417f

Quebec typical poola

500b

1200c

770

9390d

Modena 1 Modena 2 Modena 3 Modena 4 Modena 5

312 312 þ 42g 300 312 250

420 500 þ 30g 420 670 565

700 600 450 440 525

3500 4200 3375 3000 3150

a b c d e f g

6 5 5 6 5

16 27 3 20 13

"half olympic sized pool" assumed to represent Quebec pools. 25 m  20 m. based on 2.4 m depth. based in 12.2 m (40 ft) ceiling. uniform distribution used. maximum bather load based on 1.2 m2 per bather (Ministe`re du De´veloppement Durable, de l’Environnement et des Parcs, 2006). additional dimensions given for wading pool in the same natatorium.

5088

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 8 4 e5 0 9 8

Table 3 e Estimates for fugacity capacities (after Mackay et al., 1985). Medium (compartment number)

Air (1) Z1 ¼ 1/RT (mol/Pa m3)

Chloroform BDCM DBCM Bromoform

4.04E-04 4.04E-04 4.04E-04 4.04E-04

Water (2) 3

Skin (3) S

S

Z2 ¼ 1/H or C /P (mol/Pa m3)

H (Pa m /mol) 390.87 242.8 119.60 59.56

2.56E-03 4.12E-03 8.36E-03 1.68E-02

KM (unitless)

Z3 ¼ KM/H (mol/Pa m3)

10.053 12.606 16.126 20.682

0.026 0.052 0.135 0.347

Notes: R ¼ 8.314 Pa m3/mol K; T ¼ absolute temperature (298 K); H ¼ Henry’s Law Constant (Pa m3/mol); CS ¼ aqueous solubility (mol/m3); PS ¼ vapor pressure (Pa); KM ¼ skin water partition coefficient ¼ 0.64 þ 0.25(KOW)0.8.

Based on the recommendations for operation of pool facilities made by ASHRAE (2007) the air in the natatorium is re-circulated at a rate of 56,340 m3/h. The percentage of recirculated air that is replaced with fresh air was calculated using 9 m3 of fresh air per m2 of surface area of the room. The flow of re-circulated water is disregarded because the concentration of THMs in the water is held constant. The movement of the people in and out of the pool is assumed to be 4.13 m3/h. These unconventional units are useful for calculating inter-media and advective flows in the model.

3.4.

Inter-media transport

Movement of THMs between media is governed by diffusive and mass transport processes. The net transfer rate is described by the following equation: N ¼ Dij  fi
Dji  fj


 mol h

(2)

where Dij is the transfer from medium i to medium j and Dji is the transfer from j to i. The transport coefficient is generally described by:
 mol D¼kAZ $h : Pa

(3)

where k is the mass transfer coefficient and A is the interphase area.

3.4.1.

Air and water

The mass transport between the air and the water was calculated using following relationship: D12 ¼

1 1 1 þ kG A12 Z1 kL A12 Z2

(4)

where kG is the air side mass transfer coefficient, kL is the water side mass transfer coefficient, and A12 is the area over

which the air and water are in contact. This relationship is based on the two-resistance theory presented by Mackay and Yeun (1983). The mass transfer coefficients, k, are physico-chemical properties that were generated based on formulas provided by Guo and Roache (2003). The air side mass transfer coefficient kG was calculated using seven formulas from the following sources: 1. 2. 3. 4. 5. 6. 7.

Bennett and Myers (1982), Sparks et al. (1996), Higbie (1935), Reinke and Brosseau (1997), Mackay and Matsugu (1973), Reinke and Brosseau (1997), Geankoplis (1993), Reinke and Brosseau (1997), Jayjock (1994), van Veen et al. (1999), Sparks et al. (1996) and Reinke and Brosseau (1997), Geankoplis (1993).

Using the distribution fitting function of the decision support software @Risk, the results were fitted to a lognormal distribution to be later used in performing Monte Carlo simulations. The water side mass transfer coefficient, kL, was determined using an equation from Southworth (1979) provided by Guo (2002) and Guo and Roache (2003). The transfer coefficients kG, kL, and kp are presented in Table 4. The probability distributions and their characteristic parameters are also provided where applicable.

3.4.2.

Water and skin

The transport coefficient for water and skin, D23, is described by: D23 ¼ Kp  A23  Z3

(5)

where Kp is the diffusion coefficient for skin and A23 is the surface area of the skin in contact with the water (McKone and Howd, 1992). The diffusion of the chemical into skin is also a two-resistance model, with the permeability coefficient, Kp.

Table 4 e Mass transfer coefficients. Transfer coefficient (m/h)

Distribution parameters

Chloroform

BDCM

DBCM

Bromoform

kG kL kp

Lognormal mean (std dev) Deterministic Deterministic

46.25 (20.70) 0.399 0.0296

44.87 (20.71) 0.341 0.0311

43.93 (20.78) 0.302 0.0349

43.21 (20.98) 0.274 0.0402

5089

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 8 4 e5 0 9 8

3.5.

Mass balance

3.6.

Mass balance equations were generated for each compartment. The equations were of the general form: Ei þ Gi CBi ¼ Gi Ci þ

X

Dij fi


X

Dji fj

(6)

where Ei ¼ direct emission to the compartment, GiCBi ¼ advective flow into the compartment, in this case the air re-circulated back into the natatorium with some of it removed and replaced with clean air, GiCi ¼ advective flow out of the compartment, SDijfi and SDjifj ¼ inter-media transport from compartment i to compartment j, and from compartment j to compartment i, respectively. The concentration of chemicals in the water, C2, is known and the concentrations C1 in the air, and C3 in the skin can be expressed as Ci ¼ Zi∙fi. The remaining mass balance equations are: For air: D12  C2 Z2 f1 ¼ 0:12  G1  Z1 þ D12

(7)

For Water: E2 ¼ D23 



 C2 C2
D23  f 3 þ D21 
D21  f 1 Z2 Z2

(8)

For Skin:

f3 ¼

D32  C2 Z2 G3  Z3 þ D32

(9)

Model validation

The proposed model was validated using a data set provided by Fantuzzi et al. (2001). In that study, water and air concentrations of the four THMs were measured for five swimming pools located in Modena, Italy. The air and water concentrations of THMs are presented in Table 5 and the pool facility characteristics are presented in Table 2. These concentrations and pool characteristics were used as input into the fugacity model. The equations given in the previous section were used to determine the concentrations in air and skin. The modeled air concentrations were compared to the air concentrations measured by Fantuzzi et al. (2001) as shown in Fig. 2. The normalized mean bias (NMB) and mean fractional bias (MFB), shown in Table 5, were calculated for each THM using the following equations: NMBð%Þ ¼

 PN  i¼1 Ypredicted
Ymeasured  100 PN i¼1 Ymeasured

(10)

MFBð%Þ ¼

  N 2  Ypredicted
Ymeasured 1X   N i¼1 Ypredicted þ Ymeasured

(11)

The modeled concentrations were closest to the measured concentrations for chloroform and total trihalomethanes with NMB and MFB from 1.7% to 17%, followed by BDCM which were 28% and 43%, respectively. Bromoform was only detected in one swimming pool in the air. The poorest fit was for the DBCM with NMB of 74% and MFB of 109%. The air recirculation characteristics used in the model for the Italian pools are presented in Table 2. The Italian guidelines for swimming pool facilities (Conferenza Permanente per i Rapporti tra lo Stato, 2003; Fantuzzi, 2011 personal communication) suggest air recirculation which results in air

Table 5 e THM concentrations in pools in Modena, Italy (Fantuzzi et al., 2001) and comparison of measured and modeled air concentrations for model validation.

Chloroform

BDCM

DBCM

Bromoform

TTHM

Water (mg/L) Air (mg/m3) Water (mg/L) Air (mg/m3) Water (mg/L) Air (mg/m3) Water (mg/L) Air (mg/m3) Water (mg/L) Air (mg/m3)

Measured Modeled Measured Modeled Measured Modeled Measured Modeled Measured Modeled

Modena 1

Modena 2

Modena 3

Modena 4

Modena 5

26 58.6 87.67 (37.81) 5.3 13 11.17 (4.79) 0.6 3.5 0.64 (0.3) 0.1 nd nd 32 75.1 99.54 (38.15)

47 42.00 40.15 (15.83) 4.3 4.80 2.29 (0.88) 1.3 1.20 0.35 (0.14) 0.1 nd nd 52.7 48.00 42.80 (15.86)

18.7 67.7 50.78 (18.11) 4.5 14.7 7.62 (2.62) 2 4.3 1.71 (0.63) 0.3 nd nd 25.5 86.7 60.24 (18.32)

68.4 43 73.25 (29.07) 2 2.9 1.35 (0.53) 0.4 0.3 0.14 (0.06) -nd nd nd 70.8 46.2 74.74 (29.07)

6.1 19 16.86 (6.33) 5 8 8.67 (3.19) 1.3 6 1.14 (0.46) 1.3 0.8 0.58 (0.25) 17.8 33 27.25 (7.08)

Values in parentheses are standard deviations. nd e not detected. NMB e normalized mean bias. MFB e mean fractional bias.

NMB

MFB

16.68

9.36


28.34


42.86


73.99


108.58

5.39

1.73

5090

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 8 4 e5 0 9 8

Fig. 2 e Comparison of measured air concentration to modeled air concentration for model validation.

velocities of no more than 0.10 m/s with an outdoor air component of at least 20 m3/h for each m2 of swimming pool surface area. The number of swimmers considered in the model for each facility were the maximum number of bathers reported by Fantuzzi (personal communication, 2011) during the sampling event. For pools 2 and 4, the model fit was poor when using the default assumed values for air recirculation. The model was re-run using 100% outside air for these pools, resulting in a better fit. In general, the fit for these two swimming pools was not as favorable. In the case of pool 2, there were two swimming pool basins in the natatorium, which would make it more difficult for the model to predict the air concentrations. The poor fit for pool 4 could not be completely explained; therefore, further analysis would be required to investigate how to best fit the model to various swimming pool and natatorium configurations. The ratio of air concentrations to water concentrations for these pools was lower in those two pools than the other three by a factor of 2e5, suggesting that the conditions specific to those two pools make the air concentrations much lower than the other three pools.

3.7. Results and discussion on fugacity modeling for Quebec pools To evaluate the exposure doses for the swimming pools in Quebec, Equations (6) through (9) were used to calculate the fugacities and to estimate the concentrations, amount, and % in each compartment, presented in Table 6. Because the concentration in the water was known, there was an unknown term E2, which represented the formation of the THMs. This number should be equal to the total amount of THMs leaving with the people, and vented from the exhaust fans. The fugacities, concentrations, mass, and advective and interphase transports for each compound are presented in Fig. 3. The partitioning of the THMs between the water and the air is consistent with other studies in which concentrations of chloroform in water and air were measured. In Table 7, the results of the model are compared to empirical data from the literature. Chloroform is presented because it is the most commonly measured THM in the literature. In Fig. 4, a plot of measured chloroform concentrations in water and air are

Table 6 e Fugacities, concentrations and emissions. Compound

Chloroform BDCM DBCM Bromoform

Fugacity in Pa f1

f2

f3

4.83E-03 4.98E-05 4.12E-06 1.78E-06

1.81E-01 1.82E-03 1.46E-04 6.05E-05

1.64E-03 1.75E-05 1.56E-06 7.45E-07

C1 mg/m3

C2 mg/L

C3 mg/m3

232.6 3.29 0.35 0.18

55.2 1.23 0.26 0.26

5266 148.0 44.40 66.17

E2 g/h 1.64 2.32E-02 2.62E-03 1.54E-03

5091

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 8 4 e5 0 9 8

Fig. 3 e Fugacity model results for the four THM species. Visual representation of the model results showing concentration, fugacity, amount (in grams) present and percent of total amount for each compartment of the model, and for each of the four THMs considered.

Table 7 e Comparison of literature with model results (concentrations in water and air). Data source Aggazzotti et al. (1990) Aggazzotti et al. (1993) Aggazzotti et al. (1995)

Aggazzotti et al. (1998)

Cammann and Hubner (1995)

Pool water mg/L

Pool air mg/m3

32.75 36.59 89.69 44.59 57.44 99.79 36.5 78.44 19.5 114.5 47.5 98.78 33.7 2.3

(10.76) (18.24) (31.24) (18.03) (11.03) (31.85) (5.10) (6.95) (2.65) (12.45) (1.02) (5.92) (9.6) (0.60)

209.52 (129.6) 138.68 (78.45) 261.74 (110.8) 310.33 (134.0) 120.52 (38.03) 376.54 (107.19) 96 (17.6) 253.78 (56.65) 48 (1.63) 459.5 (160.6) 302.5 (88.8) 387.83 (124.7) 91.7 (15.4) 10.5 (3.1)

0.8 11.38 10.12 23.6

(0.2) (1.17) (4.09) (2.42)

5.2 (1.5) 12.285 (2.61) 42.8 (11.95) 93.85 (8.75)

Data source Erdinger et al. (2004) Fantuzzi et al. (2001)

Jo (1994) Levesque et al. (1994)

Levesque et al. (2000) Model results using Simard (2008) data Model results using simulated water concentrations

Values given are the mean, values in parentheses are standard deviation, model results in bold.

Pool water mg/L

Pool air mg/m3

17.5 (3.81) 39.8 (21.7) 33.2 (24.6) 4.2 (1.3) 1.9 (2) 23.9 (6.6) 19.5 (7.5) 158.6 (7.5) 200 (10.7) 307.1 (16) 567.5 (5.0) 538.3 (21.4) 37.79 (21.02) 55.20

195 (25.1) 58 (22.1) 46.1 (18.6) 8.7 (5.1) 3.1 (2.3) 50.9 (2.2) 33.6 (12.8) 507 (45.8) 1490 (438.7) 1120 (157.1) 1296 (36.5) 1630 (196.4) 208.59 (91.12) 233

65 80 100 150

275 338 423 634

5092

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 8 4 e5 0 9 8

Fig. 4 e Chloroform air and water concentrations from model and previous studies. Chart showing the relationship between measured concentrations of chloroform in air and water in previous studies, and the model results of air concentrations for given water concentrations.

presented with a line representing the modeledair concentrations. The known water concentrations of THMs and modelled air concentrations can now be used in exposure assessment to determine the exposure doses for swimmers through dermal contact, inhalation and ingestion.

4.

Exposure assessment

In the equations for each exposure route below, some common terms are: EF ¼ exposure frequency (min/month)  12 months/year ED ¼ exposure duration (years) e to calculate chronic daily intake (CDI) BW ¼ body weight (kg) AT ¼ averaging time (yr) e for chronic only (70 yrs)

4.1.

In a swimming pool, the exposure routes for THMs are dermal contact, inhalation and ingestion. The exposures vary for different age groups based on average time swimming, body surface area, inhalation rate and rate of inadvertent ingestion of the pool water. The exposure factors used below are listed in Table 8.

Inhalation

The dose of inhaled THMs was calculated using the following equation:
 mg Ca  IRA  EF  ED
 Dose ¼ (12) days kg$d BW  AT  365 year where

Table 8 e Exposure factors (US EPA, 2009). Exposure factor

Body weights (kg) Time spent swimming (min/month) Respiration rates (m3/minute) Surface area of whole body (m2) Surface area of head (m2) Surface area of body, no head (m2) Water ingestion (mg/L)

Age

Mean Standard Mean Standard Mean Standard Mean Standard Mean Standard Mean Standard

deviation deviation deviation deviation deviation deviation

1e4

5e11

12e17

18e64

>65

14.53 2.41 81.53 133.1 2.11E-02 4.55E-03 0.63 0.07 0.085 0.0017 0.545 0.069 49

32.03 11.2 68.76 54.59 2.13E-02 4.82E-03 1.09 0.22 0.109 0.0042 0.98 0.21 49

56.56 16.58 87.41 83.4 2.54E-02 6.46E-03 1.59 0.27 0.111 0.0039 1.475 0.27 49

80 20.15 51.59 47.54 2.72E-02 7.25E-03 1.95 0.28 0.116 0.0035 1.836 0.28 21

74.49 16.16 53.67 59.45 2.56E-02 5.00E-03 1.88 0.24 0.115 0.0031 1.768 0.24 21

5093

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 8 4 e5 0 9 8

Ca ¼ concentration of contaminant in air (mg/m3) IRA ¼ receptor air intake rate (m3/min)

CW ¼ concentration of contaminant in water (mg/L) IR ¼ ingestion rate (ml/h) h ¼ 1:67  10
5 CF ¼ conversion factor 1012 Lml  601min

Lognormal distributions of the respiration rates, body weight, and minutes per month spent swimming for bathers of different ages were determined from the US EPA Exposure Factors Handbook (US EPA, 2009).

4.2.

Ingestion of pool water is a larger issue for children; however, the EFH (Dufour et al., 2006) provides ingestion rates for ages under sixteen and over 18. They were allocated to the ages in this study using one value for those under 17, and one value for those older than 18.

Dermal contact

The dose due to dermal contact with THMs in water was calculated using the fugacity based model. The concentrations of THMs in the skin (biota compartment) in the model was multiplied by age specific volume per swimmer:
 mg CB  V  EF  ED
  CF Dose ¼ (13) days kg$d BW  AT  365 year

4.4.

The exposure dose for each age group, each THM and each exposure route were determined using Monte Carlo simulations using the software @Risk (Palisade, 2010). The results presented in Table 9 are the daily dose in mg/kg day. The results are also illustrated in Fig. 5. The chronic daily intake (CDI) was calculated by weighting the age appropriate dose by the number of years out of the lifetime (70 years) in the age group.

where CB ¼ concentration of contaminant in swimmers skin (mg/L) V ¼ body volume of the swimmer (m3) 3 L h  601min ¼ 16:7 CF ¼ conversion factor 10 m3




4 7 6 þ dose5
11  þ dose12
17  CDI ¼ dose1
4  70 70 70 

47 5 þ dose>65  þ dose18
64  70 10

The surface areas and body volume for each age group were obtained from the EFH.

4.3.

Results and discussion of exposure assessment

(15)

The calculated CDIs for each THM are:

Ingestion

   

The dose of inadvertent ingestion of pool water was calculated using the equation:
 mg CW  IR  EF  ED
  CF Dose ¼ (14) days kg$d BW  AT  365 year where

chloroform 3.19E-01 mg/kg day BDCM 5.92E-03 mg/kg day DBCM 9.65E-04 mg/kg day bromoform 7.32E-04 mg/kg day

For the exposure doses presented in Table 9 and Fig. 5, the ingestion exposure was predominantly related to children.

Table 9 e Doses of THMs by age and exposure route (mg/kg day). THMs

Chloroform

BDCM

DBCM

Bromoform

Exposure route

(1e4)

(5e11)

(12e17)

>65

(18e64)

m

s

m

s

m

s

m

s

m

s

Inhalation Dermal contact Ingestion

9.62E-01 2.11E-01 8.74E-03

2.09Eþ00 4.86E-01 1.64E-02

4.33E-01 1.69E-01 3.91E-03

5.52E-01 2.34E-01 4.22E-03

3.23E-01 2.24E-01 2.45E-03

4.84E-01 3.49E-01 3.13E-03

1.43E-01 1.39E-01 4.35E-04

2.16E-01 2.11E-01 5.60E-04

1.47E-01 1.42E-01 4.78E-04

2.04E-01 2.00E-01 5.95E-04

Total

1.18Eþ00

2.51Eþ00

5.76E-01

6.90E-01

4.52E-01

6.33E-01

2.08E-01

2.92E-01

2.16E-01

2.82E-01

Inhalation Dermal contact Ingestion

1.32E-02 5.89E-03 1.91E-04

4.97E-02 1.99E-02 6.18E-04

6.15E-03 4.82E-03 8.82E-05

2.26E-02 1.30E-02 2.39E-04

4.38E-03 6.19E-03 5.32E-05

1.19E-02 1.63E-02 1.43E-04

1.92E-03 3.76E-03 9.33E-06

5.84E-03 9.66E-03 2.49E-05

2.08E-03 4.08E-03 1.07E-05

5.86E-03 1.16E-02 2.89E-05

Total

1.92E-02

6.89E-02

1.22E-02

3.10E-02

7.92E-03

1.98E-02

3.95E-03

9.67E-03

3.81E-03

1.04E-02

Inhalation Dermal contact Ingestion

1.39E-03 1.88E-03 4.28E-05

1.46E-02 2.46E-02 5.89E-04

6.45E-04 1.35E-03 1.79E-05

5.68E-03 7.33E-03 1.14E-04

6.62E-04 2.21E-03 1.40E-05

2.17E-02 4.00E-02 3.28E-04

1.96E-04 1.12E-03 1.88E-06

1.31E-03 6.29E-03 1.01E-05

2.16E-04 1.16E-03 2.21E-06

1.41E-03 6.39E-03 1.30E-05

Total

3.31E-03

3.93E-02

1.86E-03

9.93E-03

2.07E-03

4.48E-02

5.49E-04

3.07E-03

6.08E-04

3.57E-03

Inhalation Dermal contact Ingestion

6.81E-04 2.51E-03 3.94E-05

4.34E-03 1.52E-02 2.50E-04

3.13E-04 2.03E-03 1.80E-05

1.69E-03 1.19E-02 1.08E-04

2.18E-04 2.67E-03 1.07E-05

1.01E-03 1.61E-02 5.89E-05

1.07E-04 1.74E-03 2.13E-06

8.25E-04 1.37E-02 2.21E-05

1.16E-04 1.86E-03 2.38E-06

6.80E-04 1.10E-02 1.47E-05

Total

1.77E-03

1.07E-02

1.62E-03

6.57E-03

9.53E-04

5.16E-03

5.17E-04

4.72E-03

5.58E-04

3.33E-03

5094

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 8 4 e5 0 9 8

Fig. 5 e THM daily dose exposure by route and age. Graphical representation of exposure doses for each of the four THMs indicating total dose, dose for each exposure route, and exposures for each age group.

The contribution of inhalation, ingestion and dermal contact to the total dose changed for each compound as shown in Fig. 5. Note that the inhalation contribution decreases with decreasing water compartment fugacity ( fw). The proportion of total dose attributable to each exposure route is inconsistent among previous studies. Lindstrom et al. (1997) estimated 80% of total exposure resulting from dermal contact. In studies using scuba tanks to eliminate inhalation exposure, Erdinger et al. (2004) and Levesque et al. (1994) estimated the contribution from dermal contact to be less than inhalation, 1/3 and 24% respectively. In studies on exposure during showering, Jo et al. (1990) and Weisel and Jo (1996) estimated that the exposure is roughly equal for dermal and inhalation routes. The results of this study indicate that the contribution from each exposure route changes dramatically for each age group and for the four THMs considered. Research on THMs is often generalized from studies considering chloroform only; therefore, the exposure to the other THMs is incompletely understood and requires further study. Sensitivity analysis was conducted to find the most influential factors in the exposure doses for each age group for each exposure route, as shown in Table 10. THM concentration in water was the most important factor for all routes for all THMs except for chloroform, where minutes swimming were consistently more important than water concentration. The reason for this could be the much higher concentrations of chloroform present in water which would result in a smaller impact for variability.

4.5.

Comparison of models

The United States Environmental Protection Agency (US EPA) Office of Pesticide Programs developed a screening level

model, SWIMODEL, to assess swimmers’ exposures to chemicals by inhalation, ingestion, and dermal contact routes, as well as buccal/sublingual, nasal/orbital, and aural routes (US EPA, 2003). The model was based on the worst case exposure of swimmers to trihalomethanes (Beech, 1980). The user is recommended to input the air concentration of the chemical of concern, however, the model can also calculate the air concentration from water concentration and chemical properties using either Henry’s Law or Raoult’s Law. The model allows selection of swimmer age, gender and activity level (competitive or non-competitive swimming). The default exposure pathways considered by the model are inhalation, ingestion, and dermal contact with optional consideration of sublingual/buccal, nasal/orbital and aural, or any combination of these exposure routes.

Table 10 e Most sensitive factors for each exposure route. Ingestion

Inhalation

Water concentration Water concentration Minutes swimming Minutes swimming Body weight kG e inter-media transport coefficient

Dermal contact

Water concentration Minutes swimming KOW e octanol-water partition coefficient (chloroform only) Inhalation rate Body surface area Body weight Thickness of fully hydrated skin Henry’s law constant Body weight (for chloroform and bromoform) Air Changes per Volume, surface Hour (ACH) area and weight of other age groups Height

Air chloroform concentration (µg/m3)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 8 4 e5 0 9 8

Measured air concentration Dyck et al. modelled air concentration SWIMODEL Henry's law SWIMODEL Raoult's Law

Water Chloroform Concentration ( µg/L)

Fig. 6 e Comparison of models based on measured chloroform air concentrations.

The results of this study were compared to the results using SWIMODEL to illustrate the comparative fit to empirical data for each model. Using a range of concentrations of chloroform in water from empirical data shown in Fig. 3,

5095

SWIMODEL was used to calculate concentrations of chloroform in air. Fig. 6 shows the empirical data and modeled concentrations from this study; along with air concentrations determined using the two methods in SWIMODEL, Henry’s Law and Raoult’s Law. The air concentrations generated using the fugacity model appear to fit the empirical data better than the SWIMODEL results for either Henry’s Law or Raoult’s law. Also, the results from the two methods used by SWIMODEL differed by several orders of magnitude. Exposure doses were calculated using SWIMODEL with default values for a male, adult non-competitive swimmer. Fig. 7 shows the percentage of the total exposure to all four THMs that is attributable to each exposure route. This figure also provides a comparison between the results of the fugacity model and SWIMODEL. One major difference between the models is the use of additional exposure routes in SWIMODEL. Obviously, the distribution between exposure routes will be different when using additional routes, however, it appears that for some compounds the alternative routes (sublingual/ buccal, nasal/orbital and aural) have a far greater impact on exposure than the more commonly considered routes of

Fig. 7 e Comparison of proposed fugacity model with SWIMODEL.

5096

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 8 4 e5 0 9 8

ingestion, inhalation and dermal contact. Also for all THMs SWIMODEL predicts a much larger contribution to total exposure from ingestion than the fugacity model. The influence of these alternative exposure routes may be exaggerated in the SWIMODEL approach. Little literature was found to support the use of these exposure routes (Beech, 1980) and they are not among the exposures commonly recommended by the US EPA (1992) or Health Canada (2010).

of the model. Factors that should be explored include the impact on the results of the following parameters:

5.

An understanding of the partitioning and mass transport in the system of the pool environment and the risks generated could be used to calculate concentrations of chlorinated disinfection agents that can be safely added to the pool while balancing microbial risks with chemical risks.

Discussion

A number of assumptions were made in this model that may affect the outcome. For the dermal contact exposures, the thickness of hydrated skin was used in the model as a probabilistic parameter. According to McKone and Howd (1992), the absorption of organic compounds into the skin is influenced by the skin hydration rate, with the absorption rate changing over time and reaching a steady state condition for absorption. In this study, the absorption rate was assumed to be constant. It was also assumed that there is no threshold concentration under which there would be no absorption, and also that there is no maximum dose absorbed after which the body is saturated and no more can be absorbed. There is also a potential that a swimmer’s percentage of body fat could affect THM absorption and distribution in the body. In this study, the inhalation dose was calculated using the air concentration of THMs and the breathing rate for each age group. The breathing rates used in the model were for moderate activity. Intake of THMs can be measured using breath samples (Jo et al., 1990; Lindstrom et al., 1997; Aggazzotti et al., 1998) and the concentration of THM in the breath is correlated to the dose within the body. The inhalation pathway was not considered a mass transfer process for the fugacity model due to the difficulty in calculating the amount of THM exhaled, assuming that it would change with the dose that had been taken up by the body. In that case, the exhaled amount with change over the time that the swimmer is in the pool area and it would no longer be possible to maintain a steady state condition in the model. The swimming pool environment was assumed to be instantaneously and completely mixed both in the water and in the air. In the large volume of air in the natatorium, there is the potential for air currents caused by temperature gradients from the water surface to the ceiling, as well as near features such as hot tubs and saunas. Depending on the placement of air ventilation outlets and inlets, short circuits can develop in the air flow so that the mixing is not complete within the natatorium causing localized concentration of THMs close to the water surface. The inclusion of spatial variability would add to the complexity of the model with uncertain benefit. The air circulation system in a natatorium is typically designed to remove some of the moisture in the air. It is unknown what effect de-humidification has on the concentrations of THMs present in the air. In a further study, the exposure doses generated in this study with the use of a fugacity model can be used to determine cancer and non-cancer health risks, including respiratory and reproductive effects. Risk management recommendations will be made based on sensitivity analysis

 pool size, temperature and humidity,  number of bathers and activities they are involved in (high splashing, water slides etc.),  pool water recirculation rates  ventilation rates and water circulation rates, and  disinfection agent chosen and the amount used.

6.

Summary

Chlorination is important to preserve the health of swimmers in indoor pools. Microbial risks are effectively managed by the addition of chlorinated disinfectants to pool water. The generation of DBPs has been identified and characterized in many studies since their discovery. DBPs, specifically THMs, that are generated in the pool water are absorbed into the skin of swimmers, ingested by swimmers, and volatilized to the air where they can be inhaled by swimmers. The cancer and noncancer risks associated with such exposure have been established by previous research. A level III fugacity model has been developed to quantify the amount of THMs that are volatilized and absorbed by swimmers based on water concentration data collected from indoor pools in Quebec City. The modeled concentrations in air compared favorably with air concentrations measured in other studies. The resulting concentrations were used to determine exposure doses by dermal absorption, ingestion and inhalation exposure routes for 5 age groups: ages 1e4, ages 5e11, ages 12e17, ages 18e64 and ages >65. Future study is required to apply the doses to health risk formulas to determine the risks associated with swimming pool exposure to THMs. Risk management strategies should be developed that minimize THM exposure, without compromising disinfection efficiency.

Acknowledgments The authors gratefully acknowledge the input and assistance of Guglielmina Fantuzzi along with Elena Righi and Gabriella Aggazzotti for sharing their data as well as providing pool specifications and insight into the guidelines for operation of pool facilities in Italy. This paper presents the results of an ongoing research funded under Canada NSERC Discovery Grant program and UBC Okanagan internal Grant.

references

Aggazzotti, G., Fantuzzi, G., Tartoni, P.L., Predieri, G., 1990. Plasma chloroform concentrations in swimmers using indoor

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 8 4 e5 0 9 8

swimming pools. Archives of Environmental Health 45 (3), 175e179. Aggazzotti, G., Fantuzzi, G., Righi, E., Tartoni, P.L., Cassinadri, T., Predieri, G., 1993. Chloroform in alveolar air of individuals attending indoor swimming pools. Archives of Environmental Health 48 (4), 250e254. Aggazzotti, G., Fantuzzi, G., Righi, E., Predieri, G., 1995. Environmental and biological monitoring of chloroform in indoor swimming pools. Journal of Chromatography 710, 181e190. Aggazzotti, G., Fantuzzi, G., Righi, E., Predieri, G., 1998. Blood and breath analyses as biological indicators of exposure to trihalomethanes in indoor swimming pools. The Science of the Total Environment 217, 155e163. American Society of Heating, Refrigeration and air-Conditioning Engineers, 2007. ASHRAE HandbookdHVAC Applications. Available at. http://www.knovel.com/web/portal/browse/ display? EXT KNOVEL DISPLAY bookid¼2397 (accessed 09.04.09.). Beech, J.A., 1980. Estimated worst case trihalomethane body burden of a child using a swimming pool. Medical Hypotheses 6, 303e307. Bennett, C.O., Myers, J.E., 1982. Momentum, Heat, and Mass Transfer, third ed. McGraw-Hill, New York, NY (Cited by Guo 2002). British Columbia, 2010. Public Health Act Swimming Pool, Spray Pool and Wading Pool Regulations. Reg. 289/72 O.C. 4190/72. Queens Printer, Victoria BC. Cammann, K., Hubner, K., 1995. Trihalomethane concentrations in swimmers and bath attendants blood and urine after swimming or working in indoor swimming pools. Archives of Environmental Health 50 (1), 61e65. Chu, H., Nieuwenhuijsen, M.J., 2002. Distribution and determinants of trihalomethane concentrations in indoor swimming pools. Occupational and Environmental Medicine 59 (4), 243e247. Conferenza Permanente per i Rapporti tra lo Stato, le Province Autonome di Trento e di Bolzano. Accordo 16 gennaio 2003. GURI n. 51 del 3 marzo 2003. Dufour, A.P., Evans, O., Behymer, T.D., 2006. Water ingestion during swimming activities in a pool: a pilot study. Journal of Water and Health 04 (2), 425e430. Erdinger, L., Ku¨hn, K.P., Kirsch, F., Feldhues, R., Fro¨bel, T., Nohynek, B., Gabrio, T., 2004. Pathways of trihalomethane uptake in swimming pools. International Journal of Hygiene and Environmental Health 207, 571e575. Fantuzzi, G., Righi, E., Predieri, G., Ceppelli, G., Gobba, F., Aggazzotti, G., 2001. Occupational exposure to trihalomethanes in indoor swimming pools. The Science of the Total Environment 264, 257e265. Geankoplis, C.J., 1993. Transport Processes and Unit Operations, third ed. PTR Prentice Hall, Englewood Cliffs, NJ (Cited by Guo, 2002). Guo, Z., 2002. Review of indoor emission source models. Part 2. Parameter estimation. Environmental Pollution 120, 551e564. Guo, Z., Roache, N.F., 2003. Overall mass transfer coefficient for pollutant emissions from small water pools under simulated indoor environmental conditions. Annals of Occupational Hygiene 47 (4), 279e286. Guidelines for Canadian Drinking Water Quality: Summary Table. Available at, 2008. www.hc-sc.gc.ca/ewh-semt/pubs/watereau/sum_guide-res_recom/index-eng.php (accessed 02.01.10.). Health Canada, 2009. Guidelines for Canadian Drinking Water Quality: Chlorine Guideline Technical Document. Available at. www.hc-sc.gc.ca/ewh-semt/pubs/water-eau/chlorine-chlore/ index-eng.php (accessed 02.01.10). Health Canada, 2010. Federal Contaminated Site Risk Assessment in Canada, Part V: Guidance on Complex Human Health

5097

Detailed Quantitative Risk Assessment for Chemicals (DQRACHEM). Version 1.0, Draft. Health Canada, Contaminated Sites Division, Safe Environments Programme, Ottawa ON. Hery, M., Hecht, G., Gerber, J.M., Gendre, J.C., Hubert, G., Rebuffaud, J., 1995. Exposure to chloramines in the atmosphere of indoor swimming pools. Annals of Occupational Hygiene 39 (4), 427e439. Higbie, 1935. The rate of absorption of a pure gas into a still liquid during short periods of exposure. Transactions of American Institute of Chemical Engineers 31, 365e389. Cited by Reinke and Brosseau. 1997. Hsu, H.T., Chen, M.J., Lin, C.H., Chou, W.S., Chen, J.H., 2009. Chloroform in indoor swimming-pool air: monitoring and modeling coupled with the effects of environmental conditions and occupant activities. Water Research 43, 3693e3704. Jacobs, J.H., van Rooy, G.B.G.J., Meliefste, C., Zaat, V.C., Rooyackers, J.M., Heederik, D., Spaan, S., 2007. Exposure to trichloramine and respiratory symptoms in indoor swimming pool workers. The European Respiratory Journal: Official Journal of the European Society for Clinical Respiratory Physiology 29 (4), 690e698. Jayjock, M.A., 1994. Back pressure modeling of indoor air concentrations from volatilizing sources. American Industrial Hygiene Association Journal 55 (3), 230e235. Jo, W.K., 1994. Chloroform in the water and air of Korean indoor swimming pools using both sodium-hypochlorite and ozone for water disinfection. Journal of Exposure Analysis and Environmental Epidemiology 4 (4), 491e502. Jo, W.K., Weisel, C.P., Lioy, P.J., 1990. Routes of chloroform exposure and body burden from showering with chlorinated tap water. Risk Analysis 10 (4), 575e580. Kim, H., Shim, J., Lee, S., 2002. Formation of disinfection byproducts in chlorinated swimming pool water. Chemosphere 46 (1), 123e130. Levesque, B., Ayotte, P., LeBlanc, A., Dewailly, E., Prud’Homme, D., Lavoie, R., Allaire, S., Levallois, P., 1994. Evaluation of dermal and respiratory chloroform exposure in humans. Environmental Health Perspectives 102 (12), 1082e1087. Levesque, B., Ayotte, P., Tardif, R., Charest-Tardif, G., Dewailly, E., Prud’Homme, D., Gingras, G., Allaire, S., Lavoie, R., 2000. Evaluation of the health risk associated with exposure to chloroform in indoor swimming pools. Journal of Toxicology and Environmental Health-Part A 61 (4), 225e243. Levesque, B., Duchesene, J.F., Gingras, S., Lavoie, R., Prud’homme, D., Bernard, E., Boulet, L.P., Ernst, P., 2006. The determinants of prevalence of health complaints among young competitive swimmers. International Archives of Occupational and Environmental Health 80 (1), 32e39. Lindstrom, A.B., Pleil, J.D., Berkoff, D.C., 1997. Alveolar breath sampling and analysis to assess trihalomethane exposures during competitive swimming training. Environmental Health Perspectives 105 (6), 636e642. MDDEP, Ministe`re du De´veloppement Durable, de l’Environnement et des Parcs, 2006. Re`glement sur la qualite´ de l’eau des piscines et autres bassins artificiels (Loi sur la qualite´ de l’environnement). L.R.Q., c. Q-2, s. 46, 87, 109.1 and 124.1, Gouvernement du Que´bec. MDDEP Guide des me´thodes de pre´le`vement, de conservation et d’analyse des e´chantillons relatifs a l’e´valuation de la qualite´ de l’eau des piscines et autres bassins artificiels. DR-09-05, 2009, Gouvernement du Quebec. Mackay, D., 2001. Multimedia Environmental Models: The Fugacity Approach, second ed. Lewis Publishers, Boca Raton. Mackay, D., 2006. Handbook of Physical-Chemical Properties and Environmental Fate for Organic Chemicals. CRC/Taylor & Francis, Boca Raton FL.

5098

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 8 4 e5 0 9 8

Mackay, D., Matsugu, R.S., 1973. Evaporation rates of liquid hydrocarbon spills on land and water. Canadian Journal of Chemical Engineering 51 (4), 434e439. Mackay, D., Yeun, A.T.K., 1983. Mass transfer coefficient correlations for volatilization of organic Solutes from water. Environmental Science and Technology 17 (4), 211e217. Cited by Guo, 2002. Mackay, D., Paterson, S., Cheung, B., Neely, W.B., 1985. Evaluating the environmental behaviour of chemicals with a level III fugacity model. Chemosphere 14 (3/4), 335e374. McKone, T.E., Howd, R.A., 1992. Estimating dermal uptake of nonionic organic chemicals from water and soil: I. Unified fugacity-based models for risk assessments. Risk Analysis 12 (4), 543e557. MathWave Technologies, 2010. EasyFitXL. Version 5.3. From. http://www.mathwave.com/easyfitxl-distribution-fittingexcel.html. Nieuwenhuijsen, M.J., Toledano, M.B., Eaton, N.E., Fawell, J., Elliott, P., 2000. Chlorination disinfection byproducts in water and their association with adverse reproductive outcomes: a review. Occupational and Environmental Medicine 57 (2), 73e85. Palisade Corporation, 2010. @RISK for Excel, Risk Analysis Add-in for Microsoft Excel Version 5.5.1: Industrial Edition. Reinke, P.H., Brosseau, L.M., 1997. Development of a model to predict air contaminant concentrations following indoor spills of volatile liquids. Annals of Occupational Hygeine 41 (4), 415e435. Richardson, S.D., Plewa, M.J., Wagner, E.D., Schoeny, R., DeMarini, D.M., 2007. Occurrence, genotoxicity, and carcinogenicity of regulated and emerging disinfection byproducts in drinking water: a review and roadmap for research. Mutation Research 636, 178e242. Rook, J.J., 1974. Formation of haloforms during chlorination of natural waters. Water Treatment and Examination 23, 234e243. Sendrov, J., Collison, H.A., 1966. Determination of human body volume from height and weight. Journal of Applied Physiology 21, 167e172. Simard S., 2008. Occurrence des sous produits de la de´sinfection dans l’eau des piscines publiques de la ville de Que´bec. Masters thesis, Universite´ Laval. Southworth, G.R., 1979. The role of volatilization in removing polycyclic aromatic hydrocarbons from aquatic environments. Bulletin of Environmental Contamination and Toxicology 21, 507e514. Sparks, L.E., Tichenor, B.A., Chang, J., Guo, Z., 1996. Gas-phase mass transfer model for predicting volatile organic compound (VOC) emission rates from indoor pollutant sources. Indoor Air 6, 31e40. Thickett, K.M., McCoach, J.S., Gerber, J.M., Sadhra, S., Burge, P.S., 2002. Occupational asthma caused by chloramines in indoor swimming-pool air. European Respiratory Journal 19 (5), 827e832. US EPA, 1995a. Method 524.2 Measurement of Purgeable Organic Compounds in Water by Capillary Column Gas

Chromatography/Mass Spectrometry. Revision 4.1. National Exposure Research Laboratory, Office of Research and Development, Cincinnati. OH. US EPA, 1995b. Method 552.2 Determination of Haloacetic Acids in Drinking Water by LiquideLiquid Extraction and Gas Chromatography with Electroncapture Detection. National Exposure Research Laboratory, Office of Research and Development, Cincinnati. OH. US EPA, 2003. User’s Manual: Swimmer Exposure Assessment Model (SWIMODEL). Version 3.0. Office of Pesticide Programs, Washington, DC. US EPA, 2009. Exposure Factors Handbook: 2009 Update. External Review Draft EPA/600/R-09/052A. Office of Research and Development, National Center for Environmental Assessment, Washington, DC. United States Environmental Protection Agency (US EPA), 1992. Guidelines for Exposure Assessment. U.S. Environmental Protection Agency, Risk Assessment Forum, Washington, DC. 600Z-92/001. van Veen, M.P., Fortezza, F., Bloemen, H.J.T., Kliest, J.J., 1999. Indoor air exposure to volatile compounds emitted by paints: experiment and model. Journal of Exposure Analysis and Environmental Epidemiology 9 (6), 569e574. Villanueva, C.M., Cantor, K.P., Grimalt, J.O., Malats, N., Silverman, D., Tardon, A., Garcia-Closas, R., Serra, C., Carrato, A., Catano-Vinyals, G., Marcos, R., Rothman, N., Real, F.X., Dosemeci, M., Kogevina, M., 2007. Bladder cancer and exposure to water disinfection by-products through ingestion, bathing, showering, and swimming in pools. American Journal of Epidemiology 156 (2), 148e156. Weaver, W.A., Li, J., Wen, Y., Johnston, J., Blatchley, M.R., Blatchley III, E.R., 2009. Volatile disinfection by-product analysis from chlorinated indoor swimming pools. Water Research 43, 3308e3318. Weisel, C.P., Jo, W.K., 1996. Ingestion, inhalation, and dermal exposures to chloroform and trichloroethene from tap water. Environmental Health Perspectives 104 (1), 48e51. Whitaker, H.J., Nieuwenhuijsen, M.J., Best, N.G., 2003. The relationship between water concentrations and individual uptake of chloroform: a simulation study. Environmental Health Perspectives 111 (5), 688e694. WHO (World Health Organization), 2000. Disinfectants and Disinfectant By-products. World Health Organization, Geneva, Switzerland (Environmental Health Criteria 216). World Health Organization, 2006. Guidelines for Safe Recreational Water Environments Volume 2: Swimming Pools and Similar Environments Geneva, Switzerland. Xu, X., Weisel, C.P., 2003. Inhalation exposure to haloacetic acids and haloketones during showering. Environmental Science and Technology 37, 569e576. Xu, X., Weisel, C.P., 2005. Dermal uptake of chloroform and haloketones during bathing. Journal of Exposure Analysis and Environmental Epidemiology 15, 289e296.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 9 9 e5 1 0 9

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Succession of phytoplankton functional groups regulated by monsoonal hydrology in a large canyon-shaped reservoir Li-Juan Xiao a, Tian Wang a, Ren Hu a, Bo-Ping Han a,*, Sheng Wang b, Xin Qian b, Judit Padisa´k c a

Institute of Hydrobiology, Jinan University, Guangzhou 510632, China State Key Laboratory of Pollution Control and Resources Reuse, School of Environment, Nanjing University, Nanjing 210093, China c Department of Limnology, University of Veszpre´m, H-8200 Veszpre´m, P.0. Box 168, Hungary b

article info

abstract

Article history:

Liuxihe reservoir is a deep, monomictic, oligo-mesotrophic canyon-reservoir in the subtropical

Received 21 October 2010

monsoon climate region of southern China. Phytoplankton functional groups in the reservoir

Received in revised form

were investigated and a comparison made between the succession observed in 2008, an

1 July 2011

exceptionally wet year, and 2009, an average year. The reservoir shows strong annual fluctu-

Accepted 8 July 2011

ations in water level caused by monsoon rains and artificial drawdown. Altogether 28 func-

Available online 23 July 2011

tional groups of phytoplankton were identified, including 79 genera. Twelve of the groups were analyzed in detail using redundancy analysis. Because of the oligo-mesotrophic and P-limited

Keywords:

condition of the reservoir, the dominant functional groups were those tolerant of nutrient

Monsoonal rains

(phosphorus) deficiency. The predominant functional groups in the succession process were

Reservoir drawdown

Groups A (Cyclotella with greatest axial linear dimension < 10 mm), B (Cyclotella with greatest

Driving factors

axial linear dimension >10 mm), LO (Peridinium), LM (Ceratium and Microcystis), E (Dinobryon and

Functional groups

Mallomonas), F (Botryococcus), X1 (Ankistrodesmus, Ankyra, Chlorella and Monoraphidium) and X2

Algal morphology

(Chlamydomonas and Chroomonas). The development of groups A, B and LO was remarkably

Diversity

seasonal. Group A was dominant during stratification, when characteristic small size and high surface/volume ratio morphology conferred an advantage. Group LO was dominant during dry stratification, when motility was advantageous. Group B plankton exhibited a high relative biomass during periods of reduced euphotic depth and isothermy. Groups LM, E, F, X1 and X2 occasionally exhibited high relative biomasses attributable to specific environmental events (e.g. drawdown, changes in zooplankton community). A greater diversity of phytoplankton functional groups was apparent during isothermy. This study underscores the usefulness of functional algal groups in studying succession in subtropical impoundments, in which phytoplankton succession can be significantly affected by external factors such as monsoonal hydrology and artificial drawdown, which alter variables such as retention time, mixing regime and thermal structure and influence light and nutrient availability. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Phytoplankton functional groups, which “bracket together species with similar morphological and physiological traits and with

similar ecologies” (Reynolds et al., 2002) offer a supplementary system to Linnaean classification. So far 39 phytoplankton functional groups have been identified (Reynolds et al., 2002; Padisa´k et al., 2009). Two deductions can be made from the

* Corresponding author. E-mail addresses: [email protected], [email protected] (B.-P. Han). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.012

5100

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 9 9 e5 1 0 9

definition of functional groups: (1) a species belonging to a functionally well-adapted group of phytoplankton is likely to tolerate the constraining conditions of factor deficiency more successfully than individuals of a less adapted group, and (2) a habitat shown typically to be constrained by light, C or N or some other factor, is more likely to be populated by species with the appropriate functional properties (Reynolds et al., 2002). The habitat preferences, tolerances and dislikes of functional groups make them useful in recognizing particular environmental conditions or vice versa (Huszar et al., 2000; Kruk et al., 2002; Salmaso and Padisa´k, 2007; Zhang et al., 2007). Although any resource essential for algal growth can be potentially limiting, the main drivers of change in phytoplankton communities are usually identified as nutrient availability (in particular C, N, P, and Si for diatoms) and light (Tilman et al., 1982). Zooplankton grazing and parasite infections also appear to be important factors (Kagami et al., 2007). To adapt to their environment, phytoplankton species have developed morphological and physiological specializations (Reynolds, 2007). Size and shape expressed by GALD (greatest axial linear dimension) and S/V (surface/volume ratio) reflect their ability to obtain nutrients and light, resist sinking, and minimize grazing pressure (Lewis, 1976; Naselli-Flores and Padisa´k, 2007). Hence phytoplankton morphological traits are related to species’ life-strategies (Irwin et al., 2006), and they are the critical references used to describe and identify functional groups (Reynolds et al., 2002; Padisa´k et al., 2009). Functional group classification developed using temperate water communities but has been applied worldwide in many types of lake and reservoir (Reynolds et al., 2002; Padisa´k et al., 2009). In a tropical lake, Lake Catemaco (Mexico), strong grazing pressure by filter-feeding herbivorous fish, permanent mixing and low transparency, and nitrogen limitation contributed to the dominance of the SN group (represented by Cylindrospermopsis catemaco and C. philippinensis, Koma´rkova´ and Tavera, 2003). A study of a deep Mediterranean reservoir showed that light, mixing regime and nutrient availability were the driving factors of phytoplankton functional groups (Becker et al., 2010). In deep subtropical reservoirs located in non-monsoon regions where the phytoplankton functional group concept has been applied, a high degree of environmental stability resulted in the development of equilibrium phytoplankton communities (Huszar et al., 2003) almost independent of human use. In another two subtropical Brazilian reservoirs, short retention time was identified as the principal factor limiting phytoplankton development (Borges and Train, 2008). Overall, phytoplankton growth in the world’s lakes is determined mainly by availability of light and nutrients which thus also provide the primary influence over functional group assemblage, although grazing, climate and weather factors and human activity in watershed can also play important roles for phytoplankton succession. Liuxihe reservoir is a deep reservoir subject to a subtropical monsoon climate. It supplies drinking water to Guangzhou, a city of more than 14 million inhabitants. This intensive water consumption and the climatic regime have important effects on the ecosystem. In the present study, we apply the concept of phytoplankton functional groups to identify phytoplankton assemblage structure and its succession. We hypothesize that by regulating mixing regime and water

retention time, thereby indirectly influencing nutrient availability and light regime, monsoonal hydrology and artificial drawdown will play an important role in regulating phytoplankton functional group succession.

2.

Materials and methods

2.1.

Study site, local climate

Liuxihe reservoir (23
45 N，113
46 E ) is a canyon-shaped water body situated close to the Tropic of Cancer in South East Asia (SE Asia, Fig. 1). Its latitude means it is subject to both tropical and subtropical climate influences and experiences a heavy monsoon with seasonally contrasting patterns of precipitation. The dry season, when precipitation is rare, runs from October to March, while 80% of annual precipitation occurs in the wet season, from April to September. Annual evaporation is about 1300 mm, with a monthly maximum of 170 mm occurring in September. Annual average air temperature is 22
C, with the lowest air temperature (about 5
C) in January and highest in July (about 35
C). The reservoir’s surface area is about 15 km2, total volume 3.2  108 m3 , average depth 21.3 m, and maximum depth 73 m. The sampling station is located within the main lacustrine zone close to the dam, from which water is discharged as outflow. Maximum depth at the sampling site is 46 m. The routine outlet is located 16.5 m above the bottom of the sample station and 300 m upstream of the dam, and is therefore defined as a mid-outlet.

2.2.

Sampling and analysis

Sampling was carried out monthly between January 2008 and December 2009 at four depths: 0.5, 5, 10, and 20 m. Secchi depth (SD) was multiplied by 2.7 to provide an estimate of euphotic depth (Zeu) (Cole, 1994). Water temperature was measured at 1-m increments using a Yellow Spring Instrument during February to November, and historical records of water column temperature was used in December to January because of no observation in our sampling period (Lin et al., 2003; Lin et al., 2009). Water retention time (RT) was the ratio of volume to daily outflow. Chlorophyll-a (Chl-a) was extracted with acetone and measured by spectrophotometry after filtering 500-ml samples through a 0.45-mm acetate fiber mesh (Lorenzen, 1967; Lin et al., 2005). Nutrient analysis was carried out according to Chinese national standards for water quality (APHA, 1989) and assessed levels of soluble reactive phosphorus (SRP), nitrate (N-NOˉ3), ammonium (N-NHˉ4), total nitrogen (TN) and phosphorus (TP). All measurements were completed within 24 h of sampling. One-liter samples were preserved with 5% formalin and 1% Lugol’s and allowed to settle in a column bottle for more than two weeks. After the supernatant was siphoned off with a 2 mm diameter hose, the residue (25 ml) was collected and used for counting algae. At least 400 algal units (>2 mm) placed in a Sedgewick Rafter counting chamber were counted under an Olympus microscope with non-inverted optics at 400 magnification (APHA, 1989). The size of each counted algal cell was measured and three subsamples were counted as one

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 9 9 e5 1 0 9

5101

Fig. 1 e Location of Liuxihe reservoir and the sample site (6) near the dam.

sample. Phytoplankton were classified to the level of genus. The biovolume and cell surface area were calculated according to Hillebrand et al. (1999) and a specific gravity of 1 mg mm3 was assumed for calculating biomass. The phytoplankton biomass data from the four sampling depths was averaged over the 0e20 m water column. Chl-a concentration was used as an alternative approximation of phytoplankton biomass, while relative biomass and the ratio of population biomass to total phytoplankton biomass were used to evaluate the importance of different functional groups. Phytoplankton functional groups were identified according to Reynolds et al. (2002) and Padisa´k et al. (2009). Because of difficulty in species identification of centric diatoms, the organisms were allocated to functional groups according to size. In the current study, 10 mm was found to be the critical diameter affecting the response of diatoms to environmental conditions. Centric diatoms with a diameter <10 mm (<103 mm3 in volume) were assigned to group A, and those >10 mm (>103 mm3 in volume) in diameter fell into group B. The difference in nutrient contents in different water layers was tested by t-test and ANOVA with SPSS 15.0. Relative water column stability (RWCS) was calculated following Padisa´k et al. (2003c): RWCS ¼ (Dh  Ds)/(D4  D5), where Dh is water density at the bottom, Ds is the water density at surface, and D4 and D5 are water density at 4
C and 5
C respectively. In this study, relative whole water column stability (RWWCS) and relative upper water column stability (RUWCS) was calcualted: RWWCS ¼ (Dh  Ds)/(D4  D5), and RUWCS ¼ (D20  Ds)/(D4  D5), where, D20 is water density at 20 m depth. Water mixing depth (Zm) was the depth of upper water column with the same temperature. The formula for the Shannon-Weaver index (H) of functional group diversity is:

P H ¼ (ni/N)  Log2(ni/N), where i ¼ 1 to S, with S being the number of groups. The formula for functional group evenness (E) is: E ¼ H  Log2S, where ni is biomass of ieth group, and N is total biomass. The relationships between functional groups and environmental factors were analyzed using detrended correspondence analysis (DCA) and redundancy analysis (RDA), and tested with Monte Carlo simulation using CANOCO  4.5 (Leps and Smilauer, 2003).

3.

Results

3.1.

Physical conditions

Isothermal conditions in Liuxihe reservoir occurred during December and January, and the lowest water temperature occurred in January. Stratification began at the end of February and a maximum surface water temperature of 31.4
C was recorded from July to September. Thermocline depth increased as surface water temperature increased. There was a positive correlation between thermocline depth and water level (R2 ¼ 0.74, P < 0.01). There was a significant difference between the thermocline depths recorded in 2008 and in 2009 (t-test, P < 0.01), which correlated with differences in water level. During the stratification period the hypolimnion was never disrupted, even after strong drawdown events in August 2008 (Fig.2). Total precipitation in the experimental years of 2008 and 2009 was 2660 and 1635 mm respectively, and exceptionally high rainfall was observed between April and July 2008 (Fig. 3). Average retention times were 168 days in 2008 and 375 days in 2009. The shortest recorded RT (32 days) occurred in June 2008

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 9 9 e5 1 0 9

1200

1600 1400

1000

1200 800

1000

600

800 600

400

400 200

Retention time (day)

Precipitation (mm)

200

0

0 J J A S O N 20 D 09 -J F M A M J J A S O N D

coinciding with the highest precipitation and a lower RT in stratification period because of artificial operation (Fig. 3). Where the whole water column was assessed, a high relative water column stability (RWWCS) occurred from July to September of that year, during which time the surface water temperature was high. From May to June 2008, RWWCS was lower than for the same period in 2009, and coincided with higher precipitation in those monsoonal months. For the upper 20 m of the water column, relative water column stability (RUWCS) in the wet season (AprileOctober) was correlated (R2 ¼ 0.62, P < 0.05) with water level. In June 2008 lower surface water temperatures and a deeper thermocline resulted in a much lower RUWCS than in 2009 (Fig. 4). Water levels in Liuxihe reservoir depend on inflow from precipitation across the watershed during wet season and on drawdown. Drawdown took place in April of both years (1.5 m in 2008 and 4.2 m in 2009), and a further pronounced drawdown of about 7 m occurred in AugusteSeptember 2008 following a rapid rise in water level of about 15 m in MayeJuly. Average monthly evaporation in the region amounts to less than 170 mm in summer and 100 mm in spring, and the observed water drawdown was due mostly to human operations. The data show that inflow of precipitation lifted the metalimnion and cooled the epilimnion, while artificial drawdown removed cool water from the mid-outlet and resulted in warm water further down. Thus precipitation and artificial drawdown contributed to a strong sinking of the thermocline during the stratified period of 2008. There was a significant positive correlation between water level and Zeu (R2 ¼ 0.64, P < 0.01), which showed that water level affected the underwater light regime. The ratio of Zeu/Zm indicates light availability in the mixing layer. Because Zeu/Zm was <1 in SeptembereNovember of 2008

20 08 -J F M A M

5102

Fig. 3 e Precipitation in Liuxihe reservoir (dark bars), water retention time (A).

and in OctobereNovember of 2009, light became a limiting factor for phytoplankton in those periods. The limitation was greater in SeptembereNovember 2008 than in the same period in 2009, resulting in a shallower Zeu (Fig. 5). Even if the data were not available, it could be deduced that Zeu/Zm was <1 in DecembereJanuary because of the increased mixing depth. Four periods describing water mixing regime were recognized according to changes in mixing depth (Zm) and precipitation: (1) dry stratificationI in MarcheApril with Zeu/Zm  1; (2) wet stratification in MayeAugust with Zeu/Zm > 1; (3) dry stratificationII in SeptembereNovember with Zeu/Zm  1; and (4) dry isothermy in DecembereFebruary with Zeu/Zm < 1.

3.2.

Nutrient conditions

TP was shown to vary with depth in 2008, with concentrations at 10 and 20 m significantly higher (ANOVA, P < 0.05) than

Fig. 2 e Isolines of temperature profile and water level fluctuation in the lacustrine zone. of Liuxihe reservoir. The bottom of the reservoir is shown in grey.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 9 9 e5 1 0 9

3.3.

600

5103

Phytoplankton

500

RWCS

400 300 200 100

D

N

S O

J

J A

A M

N

20 D 09 -J F M

S O

J

J A

A M

20 08 -J F M

0

Fig. 4 e Relative water column stability (RWWCS: -; RUWCS: B).

those at 0.5 and 5 m, especially in wet season (ANOVA, P < 0.01). No difference in TP was observed between the four sampling depths in 2009. TP concentration in surface the waters (0.5 and 5 m) did not differ between 2008 and 2009, but the difference between years was significant at 10 m and at 20 m (ANOVA, P < 0.01 and P < 0.05 respectively). TN concentration did not differ between depths or between years (ANOVA, P > 0.05). The results indicate that greater sedimentation of inorganic particles occurred during 2008, especially in the wet season. The mass ratio of TN/TP was never lower than 10 in the upper layers (0.5 m and 5 m), indicating Liuxihe reservoir was phosphorus-limited. But the ratio was lower at 10 m and 20 m because of increased TP (Fig. 6). Soluble reactive phosphorus (SRP) did not differ significantly between the four depths. It remained below 10 mg L1 throughout the two year study period, except for a peak (>10 mg L1) at 10 m and 20 m depth during AugusteSeptember 2008 coinciding with the abrupt drawdown events taking place at that time. Dissolved nitrogen (DIN) was above 184 throughout the two years. DIN comprised more than 90% nitrate (N-NO 3 ) with the balance contributed by ammonium (N-NH 4 ). DIN increased and was stratified during the wet stratification periods, and the DIN maximum was at the middle of the water column depended on temperature (Fig. 7). The results show that DIN levels were driven collectively by precipitation, mixing regime, and thermal dynamics. The mass ratio of DIN/SRP was always higher than 20, with a mean ratio of 121, indicating that SRP was the limiting factor.

Twenty-eight phytoplankton functional groups including 79 different genera were identified during the experimental period 2008e2009 (Table 1). Annual average Chl-a concentrations were 2.6 mg L1 2.3 mg 1 L in 2008 and 2009 respectively, with a maximum of 5.1 mg L1 in 2008 and a minimum of 1.3mg L1 in 2009. Thus, Liuxihe reservoir can be categorized as an oligo-mesotrophic water body. An approximately bi-modal pattern of annual Chla concentration was observed, in which two annual periods with elevated Chl-a concentrations occurred prior to and at the end of the wet season. Chl-a fluctuated markedly during the wet season. There was a positive correlation between Chla and Zeu (R2 ¼ 64, P < 0.05) in 2009, but not in 2008 (P > 0.05). Functional groups contributing more than 5% of total phytoplankton biomass at least once during sampling period were classified as prevailing groups. Twelve such groups were recognized (namely groups A, B, F, NA, S1, X1, W2, LO, LM, E, X2, Y) and collectively they contributed more than 90% of total phytoplankton biomass in every sample. The composition of the phytoplankton community and process of succession were analyzed using these 12 prevailing functional groups (Fig.8). The dominant and co-dominant groups in the two study years were A, B, F, E, Lo, LM, X1, X2. Groups A, B and Lo were the common dominant groups and each showed seasonal dynamics. Group A was typically dominant during periods of wet stratification and dry mixing, especially in JuneeAugust when it contributed more than 40% of phytoplankton biomass. Group Lo dominated samples during the dry stratification period, especially in MarcheApril when it contributed to more than 40% of phytoplankton biomass. Group B codominance occurred during periods of isothermy and high precipitation. Groups F, E, LM, X1, X2 occurred occasionally as dominant or co-dominant groups (Table 2). The diversity of phytoplankton functional groups was not correlated with group numbers (P > 0.05). In the two years, the diversity index varied seasonally, with lowest diversity occurring during dry stratification and highest diversity recorded during isothermy. In 2008, there were three obvious low diversity indices in March and April. In 2009, there were three low diversity indices in April and October (Fig. 9).

3.4. 8- J 2 00 F M A M J

J

9- J O N D 2 00 F M A M J

A S

J

A S

O N D

3

4

0

Depth (m)

5 10 15 20

4

1

2

3

4

1

2

25

Fig. 5 e Euphotic depth (white bar), mixing depth (dark bar), and ratio Zeu/Zm (:) in Liuxihe reservoir and the four different period (1: dry stratificationI; 2: wet stratification; 3: dry stratificationII; 4: dry isothermy).

Redundancy analysis

A detrended correspondence analysis (DCA) of the 12 prevailing phytoplankton functional groups revealed that the maximum gradient length of the four axes was 2.6. A redundancy analysis (RDA) was subsequently selected to test the relationship between environmental factors and phytoplankton. The RDA ordination results for phytoplankton functional groups and environment variables on axes 1 and 2 are shown in Fig. 10. The first two axes account for 98% of species-environmental variables relation (axis 1: 70%; axis 2: 28%) and 32% of the groups variables（axis 1: 23%; axis 2: 9%). The relative eigenvalues of axis 1, 2, 3, 4 were: 0.23, 0.09, 0.005, 0.001, respectively. The sum of all canonical axes accounted for 33% (Zm: 10%; SRP: 7%; N-NO 3 : 5%; T: 4%; others: 7%) of total variance in phytoplankton groups. Zm and SRP were the

5104

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 9 9 e5 1 0 9

Fig. 6 e Profile of TP and TN concentration in Liuxihe reservoir.

most important environmental variables for explaining phytoplankton functional groups. However, Monte Carlo testing (499 permutations) revealed no correlation between the environmental variables and phytoplankton (F ¼ 0.27; P > 0.05) that also coincided with the periodical nature of environmental variation or interference from artificial operation. Ordination of the 12 phytoplankton functional groups distributed mostly in the SRP and Zm direction. Group A had a positive relationship with Zeu, B had a positive correlation with nitrate, groups Lo, LM, S1 were negatively correlated with Zm, and groups NA, X1, X2, E, Y had a positive correlation with SRP (Fig. 10).

4.

Discussion

Liuxihe reservoir undergoes a long period of stratification (9 months) each year, creating conditions that inhibit the internal circulation of nutrients. As surface water

temperature and water column stability increase, nutrient availability in the euphotic zone decreases. This scenario was illustrated in Sau reservoir, a Mediterranean water-supply reservoir with a canyon-shaped basin where stratification occurred during the dry season (Becker et al., 2010). But in Liuxihe reservoir, stratification occurs in the wet season. Monsoon precipitation enhances the external load of nutrients and improves nutrient availability in the surface waters. The abrupt drawdown observed in JulyeAugust 2008 lead to a shallower hypolimnion and increased SRP concentration at 10 and 20 m. In the dry season, water drainage reduced the water level dramatically and a greater concentration of TP occurred in FebruaryeMarch. Light availability is one of the main factors influencing phytoplankton growth (Reynolds, 1998). In Liuxihe reservoir, large fluctuations in water level prevent hydrophyte development in the littoral zone, and the bare shoreline therefore becomes a source of suspended particles. The canyon-shape of the reservoir allows these suspended particulates from

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 9 9 e5 1 0 9

5105

Fig. 7 e Profile of SRP and DIN concentration in Liuxihe reservoir.

the shore to diffuse easily to the pelagic zone and to reduce Zeu, subsequently reducing light availability (Weyhenmeyer et al., 1998). Water level in reservoirs is determined by precipitation in the drainage basin and water use. Therefore, monsoon and human operations influence the light regime by altering water level. The standing biomass of phytoplankton depends on a balance between growth and loss rates (Reynolds, 2006). Nutrients, light availability and water temperature are the main factors influencing phytoplankton growth, and flushing, grazing, sinking are the main factors influencing loss (Reynolds, 2006). During JuneeAugust in 2008, Chla concentration in Liuxihe reservoir underwent an obvious decrease. But, this period also saw an increase in SRP concentration and Zeu improved nutrient and light availability. The unchanged stability of the upper 20 m of the water column (RUWCS) indicated that the decline in Chl-a concentration was

a result of shorter retention time (<70 d) rather than cell sinking. Although grazing is an important factor in the loss of specific algae, it does not tend to impact significantly on total phytoplankton biomass (Padisa´k et al., 2003a). In Blelham Tarn, a small eutrophic lake in the English Lake District, it was observed that Chl-a concentration decreased with retention time when retention time was shorter than 70 d (Jones and Elliott, 2007). The higher flushing rate indicated by short retention times may account for the pronounced decline in Chl-a concentration during JuneeAugust in 2008. The PEG Model describes the seasonal succession of plankton in temperate lakes (Sommer et al., 1986). In this model, two peaks in phytoplankton biomass occur: one in spring, involving small edible algae, and the other in summer, involving large or colonial algae resistant to zooplankton grazing. Zooplankton grazing is an important factor influencing phytoplankton dominance in the model. In Liuxihe

5106

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 9 9 e5 1 0 9

Table 1 e Phytoplankton functional groups in Liuxihe reservoir. Group code

Genus included in the group

Group code

A B C D E G H2 K LM Y SN TC J

Attheya, Cyclotella, Rhizosolenia Cyclotella, Aulacoseira Asterionella Nitzschia, Synedra Dinobryon, Mallomonas Eudorina Anabaena Aphanocapsa, Aphanothece Ceratium, Microcystis Cryptomonas Cylindrospermopsis Gloeocapsa, Gloeothece Golenkinia, Actinastrum, Scenedesmus, Crucigenia, Pediastrum, Coelastrum, Tetraedron

N NA P TD S1 T TB W2 W1 X3 X2 X1 MP

LO

Peridinium, Peridiniopsis, Glenodinium, Gymnodinium, Merismopedia, Chroococcus

F

reservoir, two peaks of phytoplankton biomass occurred prior to and at the end of wet season (stratification period). In both peaks, edible Peridinium (GALD < 50 mm) and Cyclotella (GALD < 10 mm) were often dominant. This suggests that in Liuxihe reservoir abiotic factors are more critical to phytoplankton growth and increase of standing biomass than biotic ones. Phytoplankton community succession exhibits spatial and temporal variation dependant on environmental conditions (Reynolds and Irish, 1997; Padisa´k et al., 2003b). The efficiency of this environmental regulation depends not only on environmental variables but also on the adaptive strategies of different types of algae (Padisa´k et al., 2003a). Liuxihe reservoir is an oligo-mesotrophic water body, therefore its phytoplankton functional groups were dominated by E, F, A, B, Lo, LM, X2, all of which tolerate nutrient deficiency (Reynolds et al., 2002; Padisa´k et al., 2009). The high mass ratio of TN/TP and DIN/SRP confirmed that phytoplankton growth was most likely constrained by phosphorus (Reynolds, 2006). Group Lo, comprising Peridinium, and groups

Genus included in the group Cosmarium, Staurastrum, Staurodesmus Euastrum, Spondylosium Fragilaria, Closterium Oedogonium, Planktonema Limnothrix, Pseudanabaena, Planktothrix Geminella, Mougeotia Achnanthes, Navicula Trachelomonas Euglena Schroederia Chlamydomonas, Chroomonas Ankistrodesmus, Ankyra, Chlorella, Monoraphidium Chlorococcum, Cocconeis, Cymbella, Gomphonema, Surirella, Lyngbya, Oscillatoria, Ulothrix Treubaria, Botryococcus, Dictyosphaerium, Elakatothrix, Gloeocystis, Kirchneriella, Nephrocytium, Oocystis. Palmella, Elakatothrix, Selenastrum, Sphaerocystis, Westella, Dactylococcopsis

A and B, comprising centric diatoms, were the most common phytoplankton functional groups, thanks to their tolerance of low phosphorus conditions (Wynne and Berman, 1980; Tilman and Kilham, 1976; Grigorszky et al., 2006). During the two annual dry stratification periods, Lo was the dominant group in both study years. The two periods shared some similar environmental conditions: a moderate thermocline and mixing depth, 20 m > Zm > 5 m and Zm approximating Zeu, water temperature ranging from 20 to 27
C, a moderate relative water column stability, and SRP below 3 mg L1. In Luixihe, Lo was represented by the motile Peridinium which can obtain more nutrients from the hypolimnion when the thermocline is shallow, but this advantage is diminished when the thermocline is deep (Borics et al., 2005). Water temperature in the 20e27
C range is favorable for reproduction in this genus (Lindstro¨m, 1984; Grigorszky et al., 2006). The less storm and higher water retention time ensured cell division and motility of Peridinium circyum in more stable condition (Pollingher and Zemel, 1981) in dry season. Therefore motility, tolerance of low SRP, and a specific growth

Fig. 8 e Phytoplankton functional groups structure and seasonla dynamics in Liuxihe reservoir.

5107

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 9 9 e5 1 0 9

Table 2 e Dominant and co-dominant functional groups in Liuxihe reservoir and morphological traits of representative species or genera. Dominant and co-dominant functional groups Code A

Time of occurring

LM X1

Wet stratification and NoveDes in the two years MayeJun in 2008, JaneFeb, May and Des in 2009 Jan in 2008 Jan and Sep in 2008 FebeApr and OcteNov in the two years and Sep in 2009 Jan in 2009 Oct in 2008

X2

Des in 2008

B F E Lo

Representative species or genera and its traits of morphology Name Cyclotella spp.

Unicell in column

Immobile

<10

<103

1e2

Cyclotella spp

Unicell in column

Immobile

10e30

103e104

<1

Botryococcus sp. Dinobryon sp. Peridinium spp.

Colony Colony Unicell in ellipse

Immobile Immobile Motile

>30 30e55 15e50

>104 >104 103e104

<0.6 <0.5

Ceratium hirundinell Ankistrodesmus sp Chlorella sp. Chlamydomonas spp

With arms Filiform Unicell in ellipse Unicell in ellipse

Motile Immobile Immobile Motile

>100 10e100 <10 7e25

104e105 <103 <103 <104

<0.2 1e4 1e2 <1

20

2.5 2

15

1.5

10

1 5

0.5

J J A S O N D

0 J J A S O N 20 D 09 -J F M A M

-J F M A M

0 08

E and F. Group B included larger Cyclotella with GALD >10 mm and Melosira, a genus adapted to low light. Group B also showed high relative biomass at the beginning of the wet season when Zeu was low. Group LM included Ceratium and Microcystis, motile or buoyant taxa are tolerant of low light and sensitive to water flush. Groups E and F were made up of nonmotile Chlorophyta with mucilage, which have an elevated light threshold and function well in clear and deep mixing water (Becker et al., 2009). Group X1 includes small and slim single-celled Chlorophyta that can adapt to water stratification and low light

25

3

Group number

Shanonn-Weaver divesity

3.5

20

S/V (um1)

Motility

preference favored group LO during the dry stratification periods. Periods of wet stratification were characterized by increased water column stability, which separated nutrients recycling between surface and bottom water. Although the euphotic zone was deep, interference caused by precipitation and artificial drawdown reduced the stability of the upper water column, allowing phytoplankton to be exposed to strong fluctuations in light. During such periods, group A, represented by Cyclotella (GALD < 10 mm), became dominant. This diatom has a lower P half-saturation constant and light saturation level, hence it can compete well under P deficiency and tolerates frequent variations in light intensity (Tilman and Kilham, 1976; Tilman et al., 1982; Reynolds et al., 1994). Its small size, high S/V and cylindrical shape may significantly decrease its sinking rate (Reynolds, 2006). Sarmento et al. (2006) found the highly abundant group A centric diatoms of Lake Kivu (central Africa) were also tolerant of nutrient depletion and high light intensity during stratification. Therefore a low P half-saturation constant, ability to survive under fluctuating light conditions, and high resistance to sinking may all contribute to the high relative biomass of group A during wet stratification. Water mixed better during periods of isothermy than at other times, and low Zeu/Zm recorded during isothermy indicate that light becomes a limiting factor. The dominant groups in Liuxihe reservoir at such times included A, LO, B, LM,

GALD (mm)

V (mm3)

Morphology

Fig. 9 e Shannon-Weaver diversity (A) of phytoplankton functional group structure and group number (B) in the open water of Liuxihe reservoir.

Fig. 10 e RDA biplot of prevailing phytoplankton functional groups (dashed lines with arrowhead) and environmental variables (solid lines with unshaded arrowhead). SRP [ soluble reactive phosphorus; T [ water temperature at 0.5 m; Zm, mixing depth; Zeu, euphotic depth; WL [ water level; Pre [ precipitation; RT [ retention time; RUWCS [ relative water column stability in upper 20 m depth.

5108

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 9 9 e5 1 0 9

levels, but are sensitive to nutrient deficiency (Reynolds et al., 2002). The group increased in relative biomass only in AugusteOctober 2008, coinciding with an SRP peak observed in all layers because of a drawdown event. Group X2 included Chlamydomonas and Chroomonas, which can tolerate stratification but are sensitive to mixing and filter-feeding grazers. This group presented a high relative biomass in December 2008, but never became dominant in the assemblage. During the stratification period, the otherwise abundant species Daphnia galeata and Ceriodaphnia quadrangula became uncommon (Wang et al., 2011). We therefore suspect that reduced grazing was one of the reasons for the high relative biomass of group X2. The dominant groups occupied habitats in Liuxihe reservoir in line with the present classification of phytoplankton functional groups. The functional groups were described mainly by physical and chemical factors, in particular the availability of phosphorus and of light, making mixing depth and SRP the most important driving variables. Biodiversity is an important aspect of community structure. In Liuxihe reservoir it is clearly affected by seasonality but its variation can be masked by group evenness. The mechanisms regulating community diversity are complex, with disturbance and a number of limiting resources being the primary factors (Reynolds et al., 1993; Interladoi and Kilha, 2001). Intermediate disturbances, mostly caused by changes in the external environment (e.g. rain, wind), can affect species diversity, and each may have its own seasonality (Sommer, 1993). In Liuxihe, precipitation and artificial drawdown were important disturbance events. The reduced RUWCS and RT in the wet season of 2008 compared to 2009 indicated a greater influence of precipitation, which explained the higher biodiversity and evenness of phytoplankton functional groups in 2008. Interlandi and Kilha (2001) discovered a strong positive correlation between diversity and the number of resources at physiologically limiting levels, suggesting that an increased number of limiting resources (nutrient and light) leads to higher diversity. For SRP concentration, 10 mg L1 was considered as a limiting P concentration for some phytoplankton, and 3 mg L1 was considered limiting for most phytoplankton (Reynolds et al., 2000). Phosphorus in the Liuxihe reservoir was limiting because of low SRP concentration. During periods of dry isothermy, the increased depth of the mixing layer allowed light to become limiting. The increased number of limiting factors may therefore explain the higher diversity observed during dry isothermy.

5.

Conclusions

(1) The study demonstrated that the concept of phytoplankton functional groups can be applied in the understanding of phytoplankton succession in reservoirs located in the subtropics of SE Asia. (2) Light and phosphorus availability are the primary internal factors selecting predominant groups, and the phytoplankton succession process in the Liuxihe reservoir is mostly driven by the mixing regime. (3) Monsoon hydrology and artificial drawdown affect water retention time, thermal structure, mixing regime, light

regime and nutrient availability, which in turn influence phytoplankton standing biomass, community structure and stage of succession.

Acknowledgements Support by grants from Chinese NSF (U0733007) and Ph.D. Programs Foundation of Ministry of Education of China (200944011220009) is appreciated. We are grateful to Profs Henri Dumont of Ghent University and Dr. Chen Ken from Australia for reading the manuscript and providing comments. We also thank all colleagues and students in the field station for their help with sampling.

references

American Public Health Association, 1989. Standard methods for the examination of water and wastewater. American water works association and water pollution control federation, Washington, DC, USA. Becker, V., Huszar, V.L.M., Crossetti, L.O., 2009. Responses of phytoplankton functional groups to the mixing regime in a deep subtropical reservoir. Hydrobiologia 628, 137e151. Becker, V., Caputo, L., Ordo´n˜ez, J., Marce´, R., Armengol, J., Crossetti, L.O., Huszar, V.L.M., 2010. Driving factors of the phytoplankton functional groups in a deep Mediterranean reservoir. Water Research 44, 3345e3354. Borics, G., Grigorszky, I., Padisa´k, J., Barbosa, F.A.R., Doma, Z.Z., 2005. Dinoflagellates from tropical Brazilian lakes with description of Peridinium brasiliense sp. nova. Algological Studies 118, 47e61. Borges, P.A.F., Train, S., 2008. Spatial and temporal variation of phytoplankton in two subtropical Brazilian reservoirs. Hydrobiologia 607, 63e74. Cole, G.A., 1994. Textbook of Limnology. Waveland PressInc, Illinois. Grigorszky, I., Kiss, K.T., Be´res, V., Ba´csi, I., M-Hamvas, M., Ma´the´, C., Vasas, C., Padisa´k, J., Borics, G., Gligora, M., Borbe´ly, G., 2006. The effects of temperature, nitrogen, and phosphorus on the encystment of Peridinium cinctum, Stein (Dinophyta). Hydrobiologia 563, 527e535. Hillebrand, H., Du¨rselen, C.D., Kirschtel, D., Pollingher, U., Zohary, T., 1999. Biovolume calculation for pelagic and benthic microalgae. Journal of Phycology 35, 403e424. Huszar, V.L.M., Silva, L.H.S., Marinho, M., Domingos, P., SantAnna, C.L., 2000. Cyanoprokaryote assemblages in eight productive tropical Brazilian waters. Hydrobiologia 424, 67e77. Huszar, V., Kruk, C., Caraco, N., 2003. Steady state of phytoplankton assemblage of phytoplankton in four temperate lakes (NE USA). Hydrobiologia 502, 97e109. Interlandi, J.I., Kilha, S.S.M., 2001. Limiting resources and the regulation of diversity in phytoplantkon communities. Ecology 82, 1270e1282. Irwin, A.J., Finkel, Z.V., Schofield, O.M.E., Falkowski, P.G., 2006. Scaling-up from nutrient physiology to the size-structure of phytoplankton communities. Journal of Plankton Research 28, 459e471. Jones, I.D., Elliott, J.A., 2007. Modelling the effects of changing retention time on abundance and composition of phytoplankton species in a small lake. Freshwater Biology 52, 988e997.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 9 9 e5 1 0 9

Kruk, C., Mazzeo, N., Lagerot, G., Reynolds, C.S., 2002. Classification schemes for phytoplankton: a local validation of a functional approach to the analysis of species temporal replacement. Journal of Plankton Research 24, 901e912. Kagami, M., Bruin, A., Ibelings, B.W., Van Donk, E., 2007. Parasitic chytrids: their effects on phytoplankton communities and food-web dynamics. Hydrobiologia 578, 113e129. Koma´rkova´, J., Tavera, R., 2003. Steady state of phytoplankton assemblage in the tropical Lake Catemaco (Mexico). Hydrobiologia 502, 187e196.  Lep s, J., Smilauer, P., 2003. Multivariate Analysis of Ecological Data using CANOCO. Cambridge University Press. Lin, Q.-Q., Hu, R., Han, B.-P., 2003. Effect of hydrodynamics on nutrient and phytoplankton distribution in Liuxihe Reservoir. Acta Ecologica Sinica 23, 2278e2285. Lin, Sh.J., He, L.J., Huang, P.Sh., Han, B.P., 2005. Comparison and improvement on the extraction method for chlorophyll a in phytoplankton. Chinese Journal of Ecologic Science 24, 9e11. Lin, G.-E., Wang, T., Lin, Q.-Q., Han, B.-P., 2009. Spatial pattern and temporal dynamics of limnological variables in Liuxihe Reservoir, Guangdong. Chinese Journal of Lake Sciences 21, 387e394. Lewis, W.M., 1976. Surface/volume ratio: implication for phytoplankton morphology. Science 192, 885e887. Lindstro¨m, K., 1984. Effects of temperature, light, and pH on growth, photosynthesis and respiration of the dinoflagellate Peridinium cinctum fa. Westii in laboratory cultures. Journal of Phycology 20, 212e220. Lorenzen, C.J., 1967. Determination of chlorophyll and pheopigments: spectrophotometric equations. Limnology Oceanography 12, 343e346. Naselli-Flores, L., Padisa´k, J., 2007. Shape and size in phytoplankton ecology: do they matter? Hydrobiologia 578, 157e161. Padisa´k, J., Scheffler, W., Kasprzak, P.S., Koschel, R., Krienitz, K., 2003a. Spatial and temporal pattern of development and decline of the spring diatom populations in Lake Stechlin in 1999. Advances in Limnology 58, 135e155. Padisa´k, J., Barbosa, F., Koschel, R., Krienitz, L., 2003b. Deep layer cyanoprokaryota maxima in temperate and tropical lakes. Limnology 58, 175e199. Padisa´k, J., Scheffler, W., Kasprzak, P., Koschel, R., Krienitz, L., 2003c. Interannual variability in the phytoplankton composition of Lake Stechlin (1994-2000). Advances in Limnology 58, 101e133. Padisa´k, J., Crossetti, L.O., Naselli-Flores, L., 2009. Use and misuse in the application of the phytoplankton functional classification: a critical review with updates. Hydrobiologia 621, 1e19. Pollingher, U., Zemel, E., 1981. In situ and experimental evidence of the influence of turbulence on cell division processes of Peridinium cinctum forma westii (Lemm.) Lefe`vre. European Journal of Phycology 16, 28e287. Reynolds, C.S., Padisa´k, J., Sommer, U., 1993. Intermediate disturbance in the ecology of phytoplankton and the maintenance of species diversity: a synthesis. Hydrobiologia 249, 183e188.

5109

Reynolds, C.S., Descy, J.P., Padisa´k, J., 1994. Are phytoplankton dynamics in rivers so different from those in shallow lakes? Hydrobiologia 289, 1e7. Reynolds, C.S., Irish, A.E., 1997. Modelling phytoplankton dynamics in lakes and reservoirs: the problem of in-situ growth rates. Hydrobiologia 349, 5e17. Reynolds, C.S., 1998. What factors influence the species composition of phytoplankton in lakes of different trophic status? Hydrobiologia 369/370, 11e26. Reynolds, C.S., Dokulil, M., Padisa´k, J., 2000. Understanding the assembly of phytoplankton in relation to the trophic spectrum: where are we now? Hydrobiologia 424, 147e152. Reynolds, C.S., Huszar, V., Kruk, C., Naselli-Flores, L., Melo, S., 2002. Towards a functional classification of the freshwater phytoplankton. Journal of Plankton Research 24, 417e428. Reynolds, C.S., 2006. The ecology of phytoplankton. Cambridge University Press, London. Reynolds, C.S., 2007. Variability in the provision and function of mucilage in phytoplankton: facultative responses to the environment. Hydrobiologia 578, 37e45. Salmaso, N., Padisa´k, J., 2007. Morpho-functional groups and phytoplankton development in two deep lakes (Lake Garda, Italy and Lake Stechlin, Germany). Hydrobiologia 578, 97e112. Sarmento, H., Isumbisho, M., Descy, J.P., 2006. Phytoplankton ecology of Lake Kivu (eastern Africa). Journal of Plankton Research 28, 815e829. Sommer, U., Gliwicz, Z.M., Lampert, W., Duncan, A., 1986. The PEG-model of seasonal succession of planktonic events in fresh-waters. Archiv fur Hydrobiologie 106, 433e471. Sommer, U., 1993. Disturbance-diversity relationship in two lakes of similar nutrient chemistry but contrasting disturbance regimes. Hydrobiologia 249, 59e65. Tilman, D., Kilham, S.S., 1976. Phosphate and silicate growth and uptake kinetics of the diatom Asterionella formosa and Cyclotella meneghiniana in batch and semicontinuous culture. Journal of Phycology 12, 375e383. Tilman, D., Kilham, S.S., Kilham, P., 1982. Phytoplankton community ecology: the role of limiting nutrients. Annual Review of Ecology and Systematics 13, 349e372. Wang, T., Xiao, L.-J., Lin, Q.-Q., Han, B.-P., Dumont, H.J., 2011. Pelagic flatworm predation on daphniids in a subtropical reservoir: different effects on Daphnia galeata and on Ceriodaphnia quadrangula. Hydrobiologia 658, 139e146. Weyhenmeyer, G.A., Pettersson, K., Padisa´k, J., 1998. Quantitative relationships between planktonic biomass and organic/ inorganic resuspended particulate matter. Limnology 51, 201e212. Wynne, D., Berman, T., 1980. Hot water extractable phosphorus an indicator of nutritional status of Peridinium cinctum (Diophyceae) from Lake Kinneret (Israel)? Phycology 16, 40e46. Zhang, X., Xie, P., Chen, F.Z., Li, S.X., Qin, J.H., 2007. Driving forces shaping phytoplankton assemblages in two subtropical plateau lakes with contrasting trophic status. Freshwater Biology 52, 1463e1475.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 1 0 e5 1 2 2

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Characterizing chromophoric dissolved organic matter in Lake Tianmuhu and its catchment basin using excitation-emission matrix fluorescence and parallel factor analysis Yunlin Zhang a,*, Yan Yin a,b, Longqing Feng a, Guangwei Zhu a, Zhiqiang Shi c, Xiaohan Liu a,b, Yuanzhi Zhang d a

Taihu Lake Laboratory Ecosystem Research Station, State Key Laboratory of Lake Science and Environment, Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences, Nanjing 210008, China b Graduate University of Chinese Academy of Sciences, Beijing 100049, China c College of Environmental Science and Engineer, Hohai University, Nanjing 210098, China d Institute of Space and Earth Information Science, Yuen Yuan Research Center for Satellite Remote Sensing, The Chinese University of Hong Kong, Shatin, N.T., Hong Kong

article info

abstract

Article history:

Chromophoric dissolved organic matter (CDOM) is an important optically active substance

Received 22 February 2011

that transports nutrients, heavy metals, and other pollutants from terrestrial to aquatic

Received in revised form

systems and is used as a measure of water quality. To investigate how the source and

6 July 2011

composition of CDOM changes in both space and time, we used chemical, spectroscopic,

Accepted 11 July 2011

and fluorescence analyses to characterize CDOM in Lake Tianmuhu (a drinking water

Available online 28 July 2011

source) and its catchment in China. Parallel factor analysis (PARAFAC) identified three individual fluorophore moieties that were attributed to humic-like and protein-like

Keywords:

materials in 224 water samples collected between December 2008 and September 2009.

Chromophoric

The upstream rivers contained significantly higher concentrations of CDOM than did the

dissolved organic matter

lake water (a(350) of 4.27
2.51 and 2.32
0.59 m1, respectively), indicating that the rivers

Chemical oxygen demand,

carried a substantial load of organic matter to the lake. Of the three main rivers that flow

Fluorescence

into Lake Tianmuhu, the Pingqiao River brought in the most CDOM from the catchment to

Lake Tianmuhu

the lake. CDOM absorption and the microbial and terrestrial humic-like components, but

Parallel factor analysis

not the protein-like component, were significantly higher in the wet season than in other seasons, indicating that the frequency of rainfall and runoff could significantly impact the quantity and quality of CDOM collected from the catchment. The different relationships between the maximum fluorescence intensities of the three PARAFAC components, CDOM absorption, and chemical oxygen demand (COD) concentration in riverine and lake water indicated the difference in the composition of CDOM between Lake Tianmuhu and the rivers that feed it. This study demonstrates the utility of combining excitation-emission matrix fluorescence and PARAFAC to study CDOM dynamics in inland waters. ª 2011 Elsevier Ltd. All rights reserved.

* Corresponding author. Present address: Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences, 73 East Beijing Road, Nanjing, Jiangsu 210008, PR China. Tel.: þ86 25 86882198; fax: þ86 25 57714759. E-mail address: [email protected] (Y. Zhang). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.014

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 1 0 e5 1 2 2

1.

Introduction

Chromophoric dissolved organic matter (CDOM) is an optically active substance that plays many important roles in freshwater ecosystems including inhibiting the attenuation of ultraviolet radiation and moving nutrients, heavy metals and other pollutants from terrestrial to aquatic systems; it can also be used as an estimate of water quality (Hansell and Carlson, 2002; Coble, 2007; Yamashita and Jaffe´, 2008; Zhang et al., 2011). CDOM is an important pool of dissolved organic carbon (DOC) and participates in the global carbon cycle and in global warming (Battin et al., 2009) via direct photochemical mineralization of DOC to dissolved inorganic carbon (DIC, CO2, and CO) (Bertilsson and Tranvik, 2000; Xie et al., 2004; Johannessen et al., 2007; Shank et al., 2010). Organic and inorganic nutrients released by microbial metabolism and photodegradation of CDOM in upstream rivers can affect eutrophication of downstream lakes (Bushaw et al., 1996). Furthermore, CDOM interferes with most of the processes of drinking water treatment. It is responsible for unpleasant odor and taste of water, formation of carcinogenic disinfection by-products, fouling of filtration membranes, increased disinfectant demands, and microbial regrowth in water distribution networks (Baghoth et al., 2011; Bieroza et al., 2010). As a result many studies have been dedicated to understanding the source, cycling and fate of CDOM in aquatic environments. Because many of these processes are controlled by the structure, composition, and the relative abundance of CDOM, characterization of CDOM is an important factor in understanding and effectively managing our aquatic resources. Due to the complexity and heterogeneity of CDOM, delineating the behavior of different components of the overall CDOM dynamics is complicated. Spectroscopic techniques, especially fluorescence, provide information about the source and composition of CDOM at natural concentrations, without requiring isolation or concentration prior to analysis (Coble, 1996). Certain components of CDOM exhibit fluorescence, meaning that optical excitation by a certain wavelength triggers light emission at a longer wavelength. Recent reviews have considered the use of fluorescence spectroscopy to investigate the dynamics of CDOM in a variety of aquatic systems: marine ecosystems (Coble, 2007), freshwaters (Hudson et al., 2007; Fellman et al., 2010), and water treatment systems (Henderson et al., 2009). Early studies determined the quantity and quality of humic fluorescence using the emission spectra at a given excitation wavelength (350 or 355 nm) or the fluorescence index, the ratio of fluorescence at 470 nm to that at 520 nm resulting from excitation at 370 nm (Hoge et al., 1993; Cory and McKnight, 2005; Zhang et al., 2007). More recently, three-dimensional excitation-emission matrix (EEMs) fluorescence has been considered to be the simplest and most effective method for studying the composition and source of CDOM because of its simplicity, sensitivity and low cost (Fellman et al., 2010). However, the EEMs of CDOM from natural waters are composed of various types of overlapping fluorophores, making it very difficult to assess the dynamics of CDOM based solely on the traditional ‘peak picking’ technique (Coble, 1996).

5111

Recently, the combined techniques of EEMs fluorescence with parallel factor analysis (PARAFAC) and principal component analysis (PCA) have successfully evaluated the environmental dynamics (composition, source, and fate) of CDOM in diverse aquatic ecosystems (Stedmon et al., 2003; Stedmon and Markager, 2005a,b; Yamashita et al., 2008, 2010; Zhang et al., 2009a; Miller and McKnight, 2010). This approach provides a considerable advantage over traditional methods for interpreting the multi-dimensional nature of EEMs data sets because the EEMs are broken down into individual fluorescent components. The use of PARAFAC to characterize DOM fluorescence properties has been further accelerated by the development of a MATLAB-based tutorial and toolbox specifically for PARAFAC analysis of DOM fluorescence (Stedmon and Bro, 2008). In China, eutrophication is a serious environmental problem in many freshwater ecosystems, and characterization of the CDOM is an important component of effective management of water quality and control of eutrophication. Lake Tianmuhu, in eastern China, is a major freshwater resource that supplies drinking water for 700,000 people in the city of Liyang. The safety of this drinking water, however, is threatened by accelerating eutrophication due to rapid economic development and population increase over the last 20 years. Therefore, the physical, chemical, and biological parameters of the lake (such as turbidity, nutrients, and chlorophyll a concentration) are rigorously monitored, and some measures to control eutrophication, including inhibiting the development of aquaculture and increasing the area of upstream wetland, have been implemented since 2000. However, the concentration, composition, and source of CDOM for Lake Tianmuhu have not been characterized, despite the fact that CDOM is an important biogeochemical factor for drinking water resources. In the present study, we used spectral absorption and EEMs fluorescence measurements to characterize the spatial and temporal distribution of CDOM concentration, composition, and source in Lake Tianmuhu and its catchment basin. By examining the relationships between different fluorescing components and CDOM absorption coefficient and chemical oxygen demand (COD) concentration, we sought to understand the coupling process of CDOM and COD.

2.

Material and methods

2.1.

Study area

Meso-eutrophic Lake Tianmuhu in Jiangsu Province, in eastern China, (117 120 E, 31 080 N) has a water area of 12 km2, a mean depth of 10 m, a maximal water volume of 1.1  108 m3, and a basin area of 148.5 km2. The catchment area is dominated by agriculture (74%) and also contains wetlands (9%), meadows (9%), and urban development (5%).1 Approximately 80e90% of the volume of water entering Lake Tianmuhu is channeled through the Pingqiao River in the east (PQR) and the Zhongtian River (ZTR) in the south. l

Unpublished data, Li HP.

5112

2.2.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 1 0 e5 1 2 2

Water sampling

Eight water sampling cruises were conducted between December 2008 and September 2009; sampling was not done in January and August. The eight sample collections were divided into three seasonal groups according to rainfall and inflowing runoff (Fig. 1): dry season, December and September; dry-to-wet transition season, March, April, and May; and wet season, February, June, and July. Samples were collected at 28 sites, including 9 in Lake Tianmuhu (LTM1eLTM9) and 19 in the catchment (Fig. 2). The catchment sites were distributed among the three main inflowing rivers, the Pingqiao River (PQR1ePQR9), the Xiasong River (XSR1), and the Zhongtian River, (ZTR1eZTR5), and the western streams (LWS1eLWS4). Water samples were collected from 0 to 1.0 m depth in 2 L acid-cleaned plastic bottles and were held on ice in the field. After samples collection, the samples were immediately transported to the laboratory within 2 h. Most water samples were immediately filtered then stored in the dark at 4  C after they were transported to the laboratory. The remaining water samples were stored at 20  C and then filtered less than three days. To verify that the short-term freezing process did not result in any loss of chromophores, absorption spectra were collected prior to and after freezing to assure no change in spectral properties. All measurements were finished within 2 weeks.

2.3.

Absorption measurement

All samples were filtered twice at low pressure, first through a pre-combusted Whatman GF/F filter (0.7 mm) and then through a pre-rinsed 25 mm Millipore membrane cellulose filter (0.22 mm), into glass bottles that had been pre-combusted at 550  C for 6 h. Comparative experiment showed that the leaching of CDOM from the filter is not significant with only 6.7% increase for the lowest a(350) of 1.34 in our study. Absorption spectra were collected between 240 and 800 nm, at 1-nm intervals, using a Shimadzu UV-2450PC UVeVis recording spectrophotometer with matching 5-cm quartz cells at room temperature (20
2  C). Milli-Q water was used in the reference cell. Absorbance measurements at each wavelength (l) were baseline corrected by subtracting the absorbance at 700 nm. Absorption coefficients were obtained from the following equation:

Fig. 2 e Distribution of sampling sites in Lake Tianmuhu and its catchment basin. Lake Tianmuhu: LTM1eLTM9, Pingqiao River: PQR1ePQR9, Xiasong River: XSR1, Zhongtian River: ZTR1eZTR5, the western streams: LWS1eLWS4.

aðlÞ ¼ 2:303DðlÞ=r

where a(l) is the CDOM absorption coefficient at wavelength l, D(l) is the corrected optical density at wavelength l, and r is the cuvette path length in m. The detection limit of 0.0001 arbitrary units (AU) of the spectrophotometer corresponds to a CDOM absorption coefficient detection limit of 0.0046 m1 using the 5 cm cell. The range of optical densities in our analysis was 0.0290e0.3225 AU. The concentration of CDOM is expressed as a(350). The spectral slope ratio (SR) was defined as the ratio of the spectral slopes of the shorter (275e295 nm) to the longer (350e400 nm) wavelength ranges (Helms et al., 2008). The spectral slope of the CDOM absorption curve (S ) was calculated by non-linear regression over the 275e295 nm and 350e400 nm wavelength ranges according to the following equation (Stedmon et al., 2000): aðlÞ ¼ aðl0 Þexp½Sðl0  lÞ þ K

Fig. 1 e Monthly variations of rainfall and inflowing runoff for Lake Tianmuhu from October 2008 to September 2009.

(1)

(2)

where a(l) is the absorption coefficient at wavelength l, a(l0) is the absorption coefficient at a reference wavelength l0 of 440 nm and S is the spectral slope (a measure of the decrease in absorption with increasing wavelength). K is a background

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 1 0 e5 1 2 2

parameter, which accounts for baseline shifts or attenuation due to factors other than CDOM.

2.4.

Three-dimensional fluorescence measurement

The EEM fluorescence of CDOM was measured using a Hitachi F-7000 fluorescence spectrometer (Hitachi High Technologies, Tokyo, Japan) with a 700-voltage xenon lamp at room temperature (20
2  C). The scanning ranges were 200e450 nm for excitation and 250e600 nm for emission. Readings were collected in ratio mode (S/R) (the default mode of F-7000 fluorescence spectrometer) at 5 nm intervals for excitation, with 1-nm emission wavelengths, using a scanning speed of 2400 nm/min. The bandpass widths were 5 nm for both excitation and emission. Water Raman scatter peaks were eliminated by subtracting a Milli-Q water blank of the EEMs. The spectra were corrected for instrumental response according to the procedure recommended by Hitachi (Hitachi F-7000 Instruction Manual). First, excitation was calibrated with Rhodamine B as standard (quantum counter) and a single-side frosted red filter in excitation scan mode. The emission was then calibrated with a diffuser in synchronous scan mode. The excitation and emission spectra obtained over the range 200e600 nm were applied internally by the instrument (through FL Solutions 2.1 software) to correct the subsequent spectra. To eliminate the inner-filter effect, the EEMs were corrected for absorbance by multiplication of each value in the EEMs with a correction factor, based on the premise that the average path length of the absorption of the excitation and emission light was 1/2 of the cuvette length (McKnight et al., 2001). This correction is expressed mathematically as: FCorr ¼ FObs  10AEx þAEm =2

(3)

where FCorr and FObs are the corrected and uncorrected fluorescence intensities and AEx and AEm are the absorbance values at the current excitation and emission wavelengths. Daily variations in fluorescence intensity were calibrated and normalized in quinine sulfate units (QSU), where 1 QSU is the maximum fluorescence intensity of 0.01 mg/L of quinine (qs) in 1 N H2SO4 at the excitation wavelength (Ex; nm)/emission wavelength (Em; nm) ¼ 350/450 (Hoge et al., 1993; Zhang et al., 2007). Rayleigh scatter effects were removed from the dataset by excluding any emission measurements made at wavelengths  excitation wavelength þ 5 nm and at wavelengths  excitation wavelength þ 300 nm. Zero was added to the EEMs in the two triangle regions (emission wavelength  excitation wavelength þ 5 nm and excitation wavelength þ 300 nm) of missing data. The contour figures of EEMs were drawn using Origin 6.0.

2.5.

PARAFAC modeling

The PARAFAC statistically decomposes the complex mixture of DOM fluorophores into non-co-varying components, without any assumptions about their spectral shape or number (Stedmon et al., 2003). Since the pioneering work of

5113

Stedmon et al. (2003), the combination of EEMs and PARAFAC has been widely applied to characterize DOM from freshwater and marine aquatic environments (Cory and McKnight, 2005; Murphy et al., 2008; Yamashita et al., 2008; Kowalczuk et al., 2009) and in laboratory and mesocosm experiments (Stedmon and Markager, 2005a). The number of fluorescent components found using PARAFAC ranges from 3 to 13 for diverse freshwater and marine aquatic environments (Cory and McKnight, 2005; Stedmon and Markager, 2005a,b; Murphy et al., 2008; Yamashita et al., 2008; Borisover et al., 2009; Kowalczuk et al., 2009). The PARAFAC analysis in our study was performed in MATLAB using the DOMFluor toolbox for MATLAB, according to Stedmon and Bro (2008). We deleted excitation wavelengths from 200 to 225 nm and emission wavelengths from 250 to 300 nm and 550 to 600 nm. A total of 224 EEMs of water samples from Lake Tianmuhu and its catchment basin were used for PARAFAC analysis. An initial exploratory analysis was performed in which outliers were identified and removed from the dataset. Two samples were considered outliers and removed either because they contained some instrument error or artifact or because they were properly measured but were very different from the others (determined by calculating the leverage using DOMfluor).

2.6.

COD concentration

The COD was measured by titration with acidic potassium permanganate based on the procedures for “Monitoring and Analytical Method of Water and Waste Water” (State Environment Protection Administration of China, 2002).

2.7.

Statistical analyses

Statistical analyses (mean value, linear-fitting, non-linear fitting, and multiple regression) were performed with Statistical Program for Social Sciences (SPSS) 17.0 software. Differences in parameters were assessed with an independent samples t-test using a p-value of 0.05 to determine significance. To examine the relationships between the three variables of CDOM absorption, COD concentration, and the maximum fluorescence intensities, we used regression and correlation analyses, using a p-value of 0.05 to determine significance.

3.

Results and discussion

3.1.

Spatial distribution of CDOM and COD

The minimum, maximum, mean, and standard deviation of CDOM absorption at 350 nm and COD concentration in the four river regions and in Lake Tianmuhu in the 3 seasons (dry, dry-to-wet transition, and wet) are shown in Table 1. The spatial distributions of CDOM absorption coefficient a(350) and COD concentration in Lake Tianmuhu are presented in Fig. 3. The mean CDOM absorption coefficient a(350) for all seasons was highest at 4.92 m1 in the Pingqiao River, decreasing to 4.50 m1 in the western streams, 4.33 m1 in the

5114

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 1 0 e5 1 2 2

Table 1 e Spatial and seasonal variations of CDOM absorption a(350) (mL1) and COD concentration (mg/L) in the different study regions and seasons in Lake Tianmuhu and its catchment. PQR: Pingqiao River, XSR: Xiasong River, ZTR: Zhongtian River, LWS: lake western streams, LTM: Lake Tianmuhu.

All season

a(350) COD

Dry

a(350) COD

Dry-to-wet transition

a(350) COD

Wet

a(350) COD

PQR

XSR

ZTR

LWS

All rivers

LTM

Minemax Mean
SD Minemax Mean
SD

1.87e14.85 4.92
2.68 1.97e9.34 4.41
1.89

2.21e7.39 4.33
1.67 1.89e5.68 3.45
1.29

1.34e10.27 2.92
1.97 1.04e7.48 2.70
1.57

2.28e12.00 4.50
2.30 1.92e8.56 4.38
1.67

1.34e14.85 4.27
2.51 1.04e9.34 3.90
1.88

1.40e3.92 2.32
0.59 2.67e5.19 3.54
0.71

Minemax Mean
SD Minemax Mean
SD

2.56e7.48 4.16
1.40 2.39e5.20 3.62
0.81

3.75e4.84 4.30
0.77 2.47e5.68 4.08
2.27

1.47e3.92 2.35
0.70 1.04e5.07 2.16
1.26

2.28e5.14 3.80
0.81 1.92e7.04 3.87
1.66

1.47e7.48 3.62
1.33 1.04e7.04 3.31
1.36

2.19e3.59 2.82
0.40 3.24e4.99 3.73
0.50

Minemax Mean
SD Minemax Mean
SD

1.87e8.18 3.83
1.68 1.97e8.29 4.10
2.09

2.21e4.31 3.00
1.14 1.89e3.38 2.42
0.83

1.34e3.43 2.19
0.74 1.31e3.83 2.18
0.86

2.35e7.65 3.67
1.65 1.92e6.83 3.69
1.33

1.34e8.18 3.32
1.59 1.31e8.29 3.40
1.80

1.40e3.50 2.13
0.55 2.67e4.51 3.00
0.44

Minemax Mean
SD Minemax Mean
SD

1.89e14.85 6.51
3.35 2.41e9.34 5.23
1.96

4.01e7.39 5.67
1.69 3.78e4.21 4.05
0.24

1.73e10.27 4.04
2.79 1.82e7.48 3.58
1.94

2.69e12.00 5.79
2.98 3.63e8.56 5.41
1.57

1.73e14.85 5.66
3.17 1.82e9.34 4.77
1.95

1.61e3.92 2.16
0.56 2.69e5.19 3.95
0.73

Xiasong River, and 2.92 m1 in the Zhongtian River; the lowest value of 2.32 m1 was recorded in Lake Tianmuhu (Table 1). A similar spatial pattern was observed for the COD concentration of the four rivers. The variation in a(350) and COD concentration in the four different rivers was linked to the combination of social and economic development and land use in the sub-catchments. First, the Pingqiao River flows through Pingqiao town, the only town in the catchment of

Lake Tianmuhu. Although we had no sewage and wastewater data, it could be concluded that point source discharges and urban runoff of this town would have a marked effect on CDOM and COD in the Pingqiao River from other studies (Hook and Yeakley, 2005; Sickman et al., 2007). Second, urban development due to human activities occupied 5.9%, 7.4%, 5.9%, and 4.4% of the land area in the sub-catchments of the Pingqiao, the western streams, the Xiasong, and the

Fig. 3 e Spatial distribution of CDOM absorption a(350) (mL1) and COD concentration (mg/L) in Lake Tianmuhu.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 1 0 e5 1 2 2

Zhongtian, respectively.2 Previous studies have shown that the change of land use from forest and wetland to urban development can influence river CDOM concentration and characteristics (Wilson and Xenopoulos, 2008, 2009; Williams et al., 2010). The terrestrial CDOM input from the catchment will increase when more land is converted for agricultural and urban use. Additionally, more autochthonous, labile DOM will be produced or transformed from terrestrial sources due to enhanced bacterial production. Because both natural and anthropogenic features of the sub-catchment including the average slope, wetland area, residential area, and tillage methods would certainly affect CDOM and COD exports (Vidon et al., 2008), it was not possible to determine which of these variables has the most impact on CDOM and COD concentration in the rivers studied. The hydrological and biogeochemical processes that control the CDOM and COD spatial variation in the different rivers are complex and merit further investigation. The spatial distributions of CDOM and COD in each individual river were not completely consistent throughout the year. For example, in most months, the a(350) of the Zhongtian river increased in the downstream direction, with the lowest value at ZTR5, the most upstream site, and the highest value at ZTR1, the most downstream site. In June, however, the reverse trend was observed, with the highest a(350) value at ZTR5. The mean values of a(350) from all four rivers were significantly higher than those from the lake (t-test, p < 0.001) (Table 1), suggesting that river input was a potential important source of CDOM in Lake Tianmuhu. Although the COD concentration was generally slightly higher in the rivers than in the lake (Table 1), there was no significant difference between river and lake samples. This suggests that, while river input cannot be considered to be the dominant factor controlling COD concentration in the lake, the higher COD concentration in the rivers may contribute to the COD concentration in the lake. In Lake Tianmuhu, the mean values of a(350) and COD concentrations at sites 8 and 9 near the mouths of inflowing rivers were significantly higher than those at the other 7 sites (t-test, p < 0.01) (Fig. 3). In contrast, there were no significant spatial differences in either a(350) or COD at other sites in the lake. In general, the a(350) values and COD concentrations were highest in the most southern parts of the lake and decreased gradually from south to north (Fig. 3).

3.2.

Temporal variations of CDOM and COD

For all river samples, both a(350) and COD concentration were significantly higher in the wet season than in the dry and dryto-wet transition seasons (t-test, p < 0.01) (Table 1), suggesting a large amount of CDOM and COD input from the catchment and surrounding cities, along with high precipitation and runoff. This is consistent with observations from other catchments (Vidon et al., 2008; Fellman et al., 2009; Miller and McKnight, 2010). We also found that, when the data from all rivers are grouped together, the 7-day antecedent precipitation of each sample collection cruise was significantly and positively correlated to the mean a(350) (r2 ¼ 0.51, n ¼ 8, 2

Unpublished data, Li HP.

5115

p < 0.05). A strong, but not significant, positive correlation was found between the 7-day antecedent precipitation and the mean COD concentration (r2 ¼ 0.38, n ¼ 8, p ¼ 0.10). This suggests a quick transfer of CDOM and COD to the river as soon as runoff increases due to precipitation. Previously many studies conclude that increased precipitation and runoff would result in increased DOM export (Vidon et al., 2008; Fellman et al., 2009; Saraceno et al., 2009; Miller and McKnight, 2010). Thus, our results support the idea that hydrological processes control the quantity and quality of CDOM exported from the catchment. The COD concentration of the lake samples did not vary significantly with season (Table 1). The a(350) of the lake samples was generally lower in the wet season than in other seasons (Table 1), but the difference was not significant. The inconsistent seasonal variations of a(350) and COD concentration between the river and lake samples suggested that other CDOM removal processes and mechanisms are involved in the lake CDOM cycle, in addition to the river input. Many studies have shown that photochemical degradation and photobleaching are important CDOM removal mechanisms (Moran et al., 2000; Twardowski and Donaghay, 2002; Johannessen et al., 2007; Tzortziou et al., 2007; Helms et al., 2008; Zhang et al., 2009b). In a 12-day experiment in the nearby Lake Taihu, exposure of CDOM to natural solar radiation decreased the CDOM absorption values a(355) and a(280) by 29.8% and 20.8%, which is a significant difference from the initial values (Zhang et al., 2009b). Lake Tianmuhu experienced high intensity UV-B radiation during the summer wet season (June and July) that would significantly increase photobleaching in the surface water column and thereby decrease CDOM absorption in those months. The higher spectral slope ratio in the summer wet season (1.80
0.30) than in dryewet (1.69
0.33) and dry (1.47
0.16) seasons further confirmed that stronger CDOM photobleaching appeared in the summer wet season because many studies had observed the marked increase of the spectral slope ratio during the photobleaching process (Helms et al., 2008; Spencer et al., 2009; Zhang et al., 2009b). In addition, the lake surface received approximately 1.3  107 m3 of precipitation, accounting for 14.6% of the total water input, which partly diluted CDOM absorption.

3.3.

PARAFAC components

Three fluorescent components were identified by PARAFAC using the split-half validation procedure, a number similar to that found for Lake Kinneret and its catchment basin (Borisover et al., 2009). The EEMs spectra (aec) and the excitation and emission loadings (def) of the three components using PARAFAC modeling are shown in Fig. 4. Modeling of the largely overlapping excitation and emission loadings of the three components on the halves of the dataset and on the whole dataset showed that the three components could completely reproduce the CDOM fluorescence composition (Fig. 4). However, our results did not suggest that only three types of fluorophores were present in all samples or that all three components were present in each sample. The spectral characteristics of the three components identified in our samples were similar to those of CDOM previously reported in other aquatic environments using the

5116

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 1 0 e5 1 2 2

Fig. 4 e The PARAFAC model output showing fluorescence signatures of the three fluorescent components. The contour plots above present spectral shapes of the excitation and emission of the three components (aec). The line plots below present split-half validation results of the three components; excitation (left) and emission (right) spectra were estimated from two independent halves of the dataset (red and green lines), and the complete dataset (black lines) (def). A perfect validation is obtained if loadings from the two halves are identical. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

PARAFAC model (Stedmon et al., 2003; Cory and McKnight, 2005; Stedmon and Markager, 2005a,b; Yamashita et al., 2008; Borisover et al., 2009; Zhang et al., 2009a, 2010). The three components we identified from the fluorescence spectra were two microbial or terrestrial humic-like components (C1 and C2) and a protein-like component (C3). Component 1 exhibited a primary (and secondary) fluorescence peak at an excitation/emission wavelength of <230 (290) nm/410 nm (Fig. 4a, d). These fluorescence characteristics could be categorized as the previously defined marine humic-like peak M (Coble, 1996). This component was also similar to microbial oxidized components (Cory and McKnight, 2005) and a phytoplankton degradation release humic component due to microbial activity (Zhang et al., 2009a). The appearance of component 1 in our river samples suggests a microbial humic component originating from the microbial transformation products of terrestrially-derived organic matter. Previous studies have demonstrated that this component may be either terrestrially-derived or produced autochthonously in aqueous environments from terrestrial organic substrates (Murphy et al., 2008; Borisover et al., 2009; Zhang et al., 2010). Component 2 also exhibited primary and secondary fluorescence peaks, although they were red-shifted compared to component 1, occurring at 240 (350) nm/470 nm (Fig. 4b, e). This second component was categorized as a mixture of the traditional terrestrial humic-like peaks A and C (Coble, 1996). The spectral features were also similar to other reported terrestrial components (Stedmon et al., 2003; Cory and McKnight, 2005).

Component 3 exhibited a single excitation/emission wavelength pair of 270 nm/305 nm (Fig. 4c, f), similar to the tyrosine-like component previously reported (Yamashita et al., 2008; Zhang et al., 2010). No tryptophan-like component was identified in our study, which was surprising because it is often reported in other studies (Stedmon and Markager, 2005a,b; Yamashita et al., 2008; Zhang et al., 2010).

3.4. Spatial and temporal distribution of PARAFAC components The spatial and seasonal variations of the mean values of the maximum fluorescence intensities (Fmax) of each of the three fluorescent components are shown in Fig. 5. For all the water samples, the Fmax of the microbial humic-like component C1 was generally higher than those of the terrestrial humic-like component C2 and the protein-like component C3 (Fig. 5a). The spatial trends of the microbial humic-like component C1 and terrestrial humic-like component C2 were similar to that of a(350), with the highest values in the Piangqiao River and the western streams and the lowest values in Lake Tianmuhu (Fig. 5a, Table 1). This result suggests that the spatial variations of the humic-like components were controlled by the same factors as a(350), namely land use and human activities in the catchment. The mean values of Fmax(C1) and Fmax(C2) were significantly higher in the rivers than in the lake (t-test, p < 0.001) (Fig. 5a), suggesting that terrestrial river input was an important source of these two components in Lake Tianmuhu. The mean values of the protein-like component Fmax(C3) were very close in the different rivers (Fig. 5a),

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 1 0 e5 1 2 2

5117

Fig. 5 e Spatial and seasonal variations of Fmax values of the three components (C1, C2, and C3) in the different study regions of Lake Tianmuhu and its catchment. The error bar represents standard deviation. PQR: Pingqiao River, XSR: Xiasong River, ZTR: Zhongtian River, LWS: the western streams, LTM: Lake Tianmuhu.

suggesting that the factors that control the spatial variations of the protein-like component were different from those of the humic-like components. The protein-like component may represent freshly and autochthonously produced semi-labile CDOM from the plants and phytoplankton in the rivers. In addition, the abundances of protein-like components were recently reported to be related to bioavailability and microbial activity of CDOM (Fellman et al., 2009; Williams et al., 2010). The mean value of Fmax(C3) was only slightly higher in river samples than in lake samples, which was different from

Fmax(C1) and Fmax(C2) suggesting that the protein-like component C3 could be less derived from allochthonous input from the rivers. In Lake Tianmuhu, the mean values of Fmax(C1) and Fmax(C2) of sites 8 and 9 were significantly higher than those of other sites (t-test, p < 0.05) (Fig. 6). There were no significant spatial differences for Fmax(C1) and Fmax(C2) at other sites in the lake, only a gradual decreasing trend from the southern part to the northern part of the lake. Although there were no significant spatial differences in Fmax(C3) between any sites in

Fig. 6 e Spatial distribution of Fmax values (QSU) of the three components (C1, C2, and C3) in Lake Tianmuhu.

5118

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 1 0 e5 1 2 2

Table 2 e Linear correlation matrix of CDOM absorption coefficient a(350), COD concentration and the maximum fluorescence intensities of the three components (C1, C2 and C3). River

COD a(350) C1 C2 C3

COD

a(350)

1.00 0.63* 0.61* 0.59* 0.01

1.00 0.85* 0.93* 0.00

C1

1.00 0.88* 0.04***

Lake C2

1.00 0.03

C3

COD

a(350)

1.00

1.00 0.12** 0.05 0.00 0.14**

1.00 0.48* 0.37* 0.01

C1

1.00 0.69* 0.02

All C2

1.00 0.15**

C3

COD

a(350)

C1

C2

C3

1.00

1.00 0.55* 0.57* 0.51* 0.00

1.00 0.85* 0.93* 0.01

1.00 0.88* 0.05***

1.00 0.04***

1.00

*: p < 0.001; **: p < 0.01; ***: p < 0.05.

the lake, the Fmax(C3) values at sites 8 and 9 were generally lower than those of other sites (Fig. 6), in contrast to Fmax(C1) and Fmax(C2). The opposite spatial distributions of the humiclike and protein-like components in the lake further confirmed that the sources and controlling factors of the humic-like and protein-like components were completely different. Terrestrial river input was the most important source of the two humic-like components in Lake Tianmuhu. Thus, Fmax(C1) and Fmax(C2) gradually decreased from the rivers to the upstream part to the downstream part of the lake due to dilution. Photobleaching in the surface water column would also gradually decrease the microbially derived humiclike component because the water transparency increased from the upstream part to the downstream part of the lake due to sedimentation (data not shown). The higher spectral

slope ratio found for the lake samples (1.68
0.31) than for the river samples (1.12
0.23) and the gradual increase of the spectral slope from the upstream part to the downstream part of the lake (Fig. 7) indicated the increase of CDOM photobleaching from the rivers to the lake, and from the upstream part to the downstream part of the lake. Conversely, increased light exposure has been observed to promote the formation of oxidized quinone-like CDOM moieties in aquatic systems (Miller et al., 2009) that were less susceptible to photodegradation than the humic-like components (Cory et al., 2007). In addition, the higher Fmax(C3) in the northern part of the lake would correlated to the human activities in this region. The northern part of the lake is the center of tourism activities of National AAAA Class Senic Spots (Tianmu Lake Natural Scenery). First, the living sewage (kitchens and toilets) of the surrounding hotels would be partly discharged into the lake. Second, water entertainment activities would also produce some living sewage and oil. All the living sewage and oil were rich in the protein-like component (Hudson et al., 2007; Henderson et al., 2009).

3.5.

Fig. 7 e Spatial distributions of the spectral slope ratio (SR) in Lake Tianmuhu.

Temporal variations of PARAFAC components

The Fmax(C1) and Fmax(C2) values of the river samples were significantly higher in the wet season than in the dry and dryto-wet transition seasons (t-test, p < 0.01) (Fig. 5b, c, d). This result verifies that the humic-like components C1 and C2 were derived from terrestrial input from the catchment. Furthermore, there was a significant positive correlation between the 7-day antecedent precipitation of every sample collection cruise and the Fmax(C2) (r2 ¼ 0.61, n ¼ 8, p < 0.05) and a positive, but not significant, correlation to Fmax(C1) (r2 ¼ 0.43, n ¼ 8, p ¼ 0.07). This confirms that the terrestrial humic-like component C2 is more sensitive to the hydrological processes than is the microbial humic-like component C1. In contrast, there were no significant seasonal differences in Fmax(C3) (Fig. 5b, c, d) or positive correlation between 7-day antecedent precipitation and Fmax(C3), suggesting that the hydrological processes did not affect the distribution of the protein-like component. Thus, our results support the idea that hydrological processes controlled not only the quantity but also the quality of CDOM exported from the catchment. The humic-like components Fmax(C1) and Fmax(C2) of the lake samples did not vary significantly with season (Fig. 5b, c, d). In contrast, the protein-like component Fmax(C3) was significantly higher in the dry-to-wet transition season than in

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 1 0 e5 1 2 2

the dry and wet seasons (t-test, p < 0.01) (Fig. 5b, c, d), which caused the ratios of Fmax(C3)/Fmax(C1) and Fmax(C3)/Fmax(C2) to be significantly higher in the dry-to-wet transition than in the dry and wet seasons (t-test, p < 0.01).

3.6. Correlation among CDOM absorption, COD concentration and PARAFAC components Because of the difference in their CDOM composition, the river samples and lake samples were considered separately when determining the correlation among CDOM absorption, COD concentration and Fmax of PARAFAC components. The linear correlation matrix for river samples, lake samples, and all samples combined is presented in Table 2. There were highly significant positive linear relationships ( p < 0.001) between CDOM absorption, COD concentration, and Fmax for the two humic-like components (C1 and C2) for the river samples and for all samples (Table 2). However, no significant linear correlation was found between COD concentration and Fmax of C1 and C2 for the lake samples (Table 2). Similar significant positive correlations between CDOM absorption and COD concentration have been found for other rivers, streams, and lakes (Holbrook et al., 2006; Erlandsson et al., 2008; Yin et al., 2011). In general, closer correlations between CDOM absorption and COD concentration have been found when the water was seriously polluted and had a high CDOM absorption coefficient. Therefore, the CDOM absorption coefficient at 254 nm has often been used as a surrogate for COD and biological oxygen demand (BOD) testing of waste water and treated effluent (Baker and Inverarity, 2004; Henderson et al., 2009).

5119

In our study, the COD concentration was significantly and positively correlated to the humic-like components but not to the protein-like component for river samples. Our observations from natural rivers not seriously polluted by sewage wastewater (COD concentration less than 10 mg/L) were different from the results observed for waste water and treated effluent (COD concentration larger than 10 mg/L in most cases), which showed that the protein-like component was highly significantly correlated to COD and BOD concentrations (Reynolds, 2002; Lee and Ahn, 2004; Fu et al., 2007; Hur et al., 2010). Weak protein-like component and COD, BOD concentration correlations were often found in river water where COD and BOD values were small (Baker and Inverarity, 2004; Holbrook et al., 2006). Such discrepancies can be partly explained because COD includes both refractory and labile CDOM of fluorescent and non-fluorescent characters; therefore, correlations could be weakened by changes in the ratios of both humic/fulvic-like to tryptophan-like material and fluorescent to non-fluorescent CDOM. The CDOM absorption was also highly linearly correlated to Fmax of C1 and C2 for lake samples, but displayed a markedly low determination coefficient, compared with that of river samples and that of all samples combined (Table 2, Fig. 8). Furthermore, the slopes of the linear relationships between CDOM absorption and Fmax were lower for lake samples than for river samples (Fig. 8aed), suggesting a decrease of fluorescence percentage of CDOM for lake samples. This phenomenon can partly be attributed to CDOM photobleaching in Lake Tianmuhu. Because the lake is more transparent than the rivers, photobleaching in the surface water column would significantly increase, which was confirmed from the higher spectral slope ratio found for the

Fig. 8 e Linear relationships among the CDOM absorption coefficient a(350), and Fmax values of the two humic-like components (C1 and C2) from river (a, c, e) and lake samples (b, d, f) from Lake Tainmuhu and its catchment.

5120

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 1 0 e5 1 2 2

lake samples (1.68
0.31) than for the river samples (1.12
0.23). Previous studies have shown that fluorescent dissolved organic matter (FDOM) was more easily degraded by UV-B radiation (Zhang et al., 2009b), which would decrease the fluorescence percentage of CDOM for lake samples. Components 1 and 2 were highly linearly correlated with each other (r2 ¼ 0.88 for river samples, r2 ¼ 0.69 for lake samples, and r2 ¼ 0.88 for all samples) (Table 2, Fig. 8e, f), indicating that a common factor controls the concentrations of these components. However, the linear relationship between the maximum fluorescence intensities of these two components was substantially weaker for lake samples, and a different slope was obtained (slope ¼ 0.502 for river samples and slope ¼ 0.435 for lake samples, Fig. 8e, f). The different relationships documented in Fig. 8 indicated differences in the composition of humic-like matter between Lake Tianmuhu and its surrounding rivers. A similar result was also observed in Lake Kinneret and its catchment basin (Borisover et al., 2009). In the present study, non-significant or only weakly significant linear relationships were found between the maximum fluorescence intensities of C1, C2, and C3 (Table 2), indicating that the source of the humic-like and protein-like components was different, as has been observed in other aquatic environments (Borisover et al., 2009; Zhang et al., 2010).

4.

Conclusion

Our finding that CDOM absorption and microbial and terrestrial humic-like components were significantly higher for river samples than for lake samples indicated that river input was one of the important sources of CDOM in Lake Tianmuhu. The result that CDOM absorption and the levels of microbial and terrestrial humic-like components, but not the level of the protein-like component, were significantly higher in the wet season than in other seasons indicated that hydrological processes including rainfall and runoff could significantly impact the quantity and quality of CDOM from the catchment. The results of this study support previous observations of the concentration and composition of CDOM in the lake and its catchment basin. Three fluorescent components, comprising two humic-like components and a tyrosine-like component, were identified using EEMs and the PARAFAC model. There were highly significant linear relationships between CDOM absorption and the maximum fluorescence intensities of two humic-like components; the determination coefficient was lower for lake samples than for river samples. The different slopes and determination coefficients of the relationships between CDOM absorption and the maximum fluorescence intensities of the two humic-like components revealed a difference in the composition of CDOM between Lake Tianmuhu and its surrounding rivers. Significant positive correlations were found between CDOM absorption and COD concentration, with markedly higher determination coefficients for river samples. COD concentration was significantly and positively correlated to the humic-like components but not the proteinlike component for river samples suggesting that COD was easily coupled with the humic-like components in Lake

Tianmuhu catchment. Overall, our results demonstrate that absorption and EEMs may be a successful toolset for water quality monitoring and CDOM characterization of drinking water sources.

Acknowledgments This study was jointly funded by the Knowledge Innovation Project of the Chinese Academy of Sciences (KZCX2-YWQN312), the Major Projects on Control and Rectification of Water Body Pollution (2009ZX07101-013), the National Natural Science Foundation of China (grants 40971252, 40825004), the Provincial Nature Science Foundation of Jiangsu of China (BK2009336) the General Research Fund (CUHK454909), and the Hong Kong Innovation Technology Fund (ITS/058/09FP). We would like to thank Chen Weimin, Xie Chungang, Li Hengpeng for their help for sample collection in the field. We thank two anonymous reviewers, whose comments helped improve this manuscript.

references

Baghoth, S.A., Sharma, S.K., Amy, G.L., 2011. Tracking natural organic matter (NOM) in a drinking water treatment plant using fluorescence excitation emission matrices and PARAFAC. Water Research 45 (2), 797e809. Baker, A., Inverarity, R., 2004. Protein-like fluorescence intensity as a possible tool for determining river water quality. Hydrological Process 18 (15), 2927e2945. Battin, Y.J., Luyssaert, S., Kaplan, L., Aufdenkampe, K., Richter, A., Tranvik, L.J., 2009. The boundless carbon cycle. Nature Geoscience 2 (9), 598e600. Bertilsson, S., Tranvik, L.J., 2000. Photochemical transformation of dissolved organic matter in lakes. Limnology and Oceanography 45 (4), 753e762. Bieroza, M.Z., Bridgeman, J., Baker, A., 2010. Fluorescence spectroscopy as a tool for determination of organic matter removal efficiency at water treatment works. Drinking Water Engineering and Science 3 (1), 63e70. Borisover, M., Laor, Y., Parparov, A., Bukhanovsky, N., Lado, M., 2009. Spatial and seasonal patterns of fluorescent organic matter in Lake Kinneret (Sea of Galilee) and its catchment basin. Water Research 43 (12), 3104e3116. Bushaw, K.L., Zepp, R.G., Tarr, M.A., Schulz-Jander, D., Bourbonniere, R.A., Hodson, R.E., Miller, W.L., Bronk, D.A., Moran, M.A., 1996. Photochemical release of biologically available nitrogen from aquatic dissolved organic matter. Nature 381 (6581), 404e407. Coble, P.G., 1996. Characterization of marine and terrestrial DOM in seawater using excitation emission matrix spectroscopy. Marine Chemistry 51 (4), 325e346. Coble, P.G., 2007. Marine optical biogeochemistry: the chemistry of ocean color. Chemical Reviews 107 (2), 402e418. Cory, R.M., McKnight, D.M., 2005. Fluorescence spectroscopy reveals ubiquitous presence of oxidized and reduced quinones in dissolved organic matter. Environmental Sciences and Technology 39 (21), 8142e8149. Cory, R.M., McKnight, D.M., Chin, Y.P., Miller, P., Jaros, C.L., 2007. Chemical characteristics of fulvic acids from Arctic surface waters: microbial contributions and photochemical transformations. Journal of Geophysical Research 112 G04S51, doi:10.1029/2006JG000343.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 1 0 e5 1 2 2

Erlandsson, M., Buffam, I., Folster, J., Laudon, H., Temnerud, J., Weyhenmeyer, G.A., Bishop, K., 2008. Thirty-five years of synchrony in the organic matter concentrations of Swedish rivers explained by variation in flow and sulphate. Global Change Biology 14 (5), 1191e1198. Fellman, J.B., Hood, E., Edwards, R.T., D’Amore, D.V., 2009. Changes in the concentration, biodegradability, and fluorescent properties of dissolved organic matter during stormflows in coastal temperate watersheds. Journal of Geophysical Reserch 114 G01021, doi:10.1029/2008JG000790. Fellman, J.B., Hood, E., Spencer, R.G.M., 2010. Fluorescence spectroscopy opens new windows into dissolved organic matter dynamics in freshwater ecosystems: a review. Limnology and Oceanography 55 (6), 2452e2462. Fu, P., Wu, F., Liu, C., Wang, F., Li, W., Yue, L., Guo, Q., 2007. Fluorescence characterization of dissolved organic matter in an urban river and its complexation with Hg(II) ion. Applied Geochemistry 22 (8), 1668e1679. Hansell, D.A., Carlson, C.A. (Eds.), 2002. Biogeochemistry of Marine dissolved Organic Matter. Academic Press. Helms, J.R., Stubbins, A., Ritchie, J.D., Minor, E.C., Kieber, D.J., Mopper, K., 2008. Absorption spectral slopes and slope ratios as indicators of molecular weight, source, and photobleaching of chromophoric dissolved organic matter. Limnology and Oceanography 53 (3), 955e969. Henderson, R.K., Baker, A., Murphy, K., Hambly, A., Stuetz, R.M., Khan, S.J., 2009. Fluorescence as a potential monitoring tool for recycled water systems: a review. Water Research 43 (4), 863e881. Hoge, F.E., Vodacek, A., Blough, N.V., 1993. Inherent optical properties of the ocean: retrieval of the absorption coefficient of chromophoric dissolved organic matter from fluorescence measurements. Limnology and Oceanography 38 (7), 1394e1402. Holbrook, R.D., Yen, J.H., Grizzard, T.J., 2006. Characterizing natural organic material from the Occoquan Watershed (Northern Virginia, US) using fluorescence spectroscopy and PARAFAC. Science of the Total Environment 361 (1e3), 249e266. Hook, A.M., Yeakley, J.A., 2005. Stormflow dynamics of dissolved organic carbon and total dissolved nitrogen in a small urban watershed. Biogeochemistry 75 (3), 409e431. Hudson, N., Baker, A., Reynolds, D., 2007. Fluorescence analysis of dissolved organic matter in natural, waste and polluted waters e a review. River Research and Applications 23 (6), 631e649. Hur, J., Lee, B.M., Lee, T.H., Park, D.H., 2010. Estimation of biological oxygen demand and chemical oxygen demand for combined sewer systems using synchronous fluorescence spectra. Sensors 10 (4), 2460e2471. Johannessen, S.C., Pen˜a, M.A., Quenneville, M.L., 2007. Photochemical production of carbon dioxide during a coastal phytoplankton bloom. Estuarine, Coastal and Shelf Science 73 (1e2), 236e242. Kowalczuk, P., Durako, M.J., Young, H., Kahn, A.E., Cooper, W.J., Gonsior, M., 2009. Characterization of dissolved organic matter fluorescence in the South Atlantic Bight with use of PARAFAC model: interannual variability. Marine Chemistry 113 (3e4), 182e196. Lee, S., Ahn, K.H., 2004. Monitoring of COD as an organic indicator in waste water and treated effluent by fluorescence excitationeemission (FEEM) matrix characterization. Water Science and Technology 50 (8), 57e63. McKnight, D.M., Boyer, E.W., Westerhoff, P.K., Doran, P.T., Kulbe, T., Andersen, D.T., 2001. Spectrofluorometric characterization of dissolved organic matter for indication of precursor organic material and aromaticity. Limnology and Oceanography 46 (1), 38e48. Miller, M.P., McKnight, D.M., 2010. Comparison of seasonal changes in fluorescent dissolved organic matter among

5121

aquatic lake and stream sites in the Green Lakes Valley. Journal of Geophysical Research 115 G00F12, doi:10.1029/ 2009JG000985. Miller, M.P., McKnight, D.M., Chapra, S.C., 2009. Production of microbially-derived fulvic acid from photolysis of quinonecontaining extracellular products of phytoplankton. Aquatic Science 71 (2), 170e178. Moran, M.A., Sheldon, W.M., Zepp, R.G., 2000. Carbon loss and optical property changes during long-term photochemical and biological degradation of estuarine dissolved organic matter. Limnology and Oceanography 45 (6), 1254e1264. Murphy, K.R., Stedmon, C.A., Waite, T.D., Ruiz, G.M., 2008. Distinguishing between terrestrial and autochthonous organic matter sources in marine environments using fluorescence spectroscopy. Marine Chemistry 108 (1e2), 40e58. Reynolds, D.M., 2002. The differentiation of biodegradable and non-biodegradable dissolved organic matter in wastewaters using fluorescence spectroscopy. Journal of Chemical Technology and Biotechnology 77 (8), 965e972. Saraceno, J.F., Pellerin, B.A., Downing, B.D., Boss, E., Bachand, P.A. M., Bergamaschi, B.A., 2009. High-frequency in situ optical measurements during a storm event: assessing relationships between dissolved organic matter, sediment concentrations, and hydrologic processes. Journal of Geophysical Research 114 G00F09, doi:10.1029/2009JG000989. Shank, G.C., Zepp, R.G., Va¨ha¨talo, A., Lee, R., Bartels, E., 2010. Photobleaching kinetics of chromophoric dissolved organic matter derived from mangrove leaf litter and floating Sargassum colonies. Marine Chemistry 119 (1e4), 162e171. Sickman, J.O., Zanoli, M.J., Mann, H.L., 2007. Effects of urbanization on organic carbon loads in the Sacramento River, California. Water Resources Research 43 W11422, doi:10.1029/ 2007WR005954. Spencer, R.G.M., Stubbins, A., Hernes, P.J., Baker, A., Mopper, K., Aufdenkampe, A.K., Dyda, R.Y., Mwamba, V.L., Mangangu, A. M., Wabakanghanzi, J.N., Six, J., 2009. Photochemical degradation of dissolved organic matter and dissolved lignin phenols from the Congo River. Journal of Geophysical Research 114 G03010, 10.1029/2009JG000968. State Environment Protection Administration of China, 2002. Analysis of water quality monitoring standards of practice handbook, fourth ed. China Environment Science Press, Beijing, pp. 210e234 (In Chinese with English abstract). Stedmon, C.A., Bro, R., 2008. Characterizing dissolved organic matter fluorescence with parallel factor analysis: a tutorial. Limnology and Oceanography: Methods 6 (1), 1e6. Stedmon, C.A., Markager, S., 2005a. Tracing the production and degradation of autochthonous fractions of dissolved organic matter by fluorescence analysis. Limnology and Oceanography 50 (5), 1415e1426. Stedmon, C.A., Markager, S., 2005b. Resolving the variability in dissolved organic matter fluorescence in a temperate estuary and its catchment using PARAFAC analysis. Limnology and Oceanography 50 (2), 686e697. Stedmon, C.A., Markager, S., Bro, R., 2003. Tracing dissolved organic matter in aquatic environments using a new approach to fluorescence spectroscopy. Marine Chemistry 82 (3e4), 239e254. Stedmon, C.A., Markager, S., Kaas, H., 2000. Optical properties and signatures of chromophoric dissolved organic matter (CDOM) in Danish coastal waters. Estuarine, Coastal and Shelf Science 51 (2), 267e278. Twardowski, M.S., Donaghay, P.L., 2002. Photobleaching of aquatic dissolved materials: absorption removal, spectral alteration, and their interrelationship. Journal of Geophysical Research 107 (C8), 3091. doi:10.1029/1999JC000281. Tzortziou, M., Osburn, C.L., Neale, P.J., 2007. Photobleaching of dissolved organic material from a tidal marsheestuarine

5122

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 1 0 e5 1 2 2

system of the Chesapeake Bay. Photochemistry and Photobiology 83 (4), 782e792. Vidon, P., Wagner, L.E., Soyeux, E., 2008. Changes in the character of DOC in streams during storms in two midwestern watersheds with contrasting land uses. Biogeochemistry 88 (3), 257e270. Williams, C.J., Yamashita, Y., Wilson, H.F., Jaffe´, R., Xenopoulos, M.A., 2010. Unraveling the role of land use and microbial activity in shaping dissolved organic matter characteristics in stream ecosystems. Limnology and Oceanography 55 (3), 1159e1171. Wilson, H.F., Xenopoulos, M.A., 2008. Ecosystem and seasonal control of stream dissolved organic carbon along a gradient of land use. Ecosystems 11 (4), 555e568. Wilson, H.F., Xenopoulos, M.A., 2009. Effects of agricultural land use on the composition of fluvial dissolved organic matter. Nature Geoscience 2 (1), 37e41. Xie, H., Zafiriou, O.C., Cai, W., Zepp, R.G., Wang, Y., 2004. Photooxidation and its effects on the carboxyl content of dissolved organic matter in two coastal rivers in the Southeastern United States. Environmental Sciences and Technology 38 (15), 4113e4119. Yamashita, Y., Jaffe´, R., 2008. Characterizing the interactions between trace metals and dissolved organic matter using excitation emission matrix and parallel factor analysis. Environmental Sciences and Technology 42 (19), 7374e7379. Yamashita, Y., Jaffe´, R., Maie, N., Tanoue, E., 2008. Assessing the dynamics of dissolved organic matter (DOM) in coastal environments by excitation and emission matrix fluorescence and parallel factor analysis (EEM-PARAFAC). Limnology and Oceanography 53 (5), 1900e1908.

Yamashita, Y., Maie, N., Briceno˜, H., Jaffe´, R., 2010. Optical characterization of dissolved organic matter in tropical rivers of the Guayana Shield, Venezuela. Journal of Geophysical Research 115 G00F10, doi:10.1029/2009JG000987. Yin, Y., Zhang, Y.L., Liu, X.H., Zhu, G.W., Qin, B.Q., Shi, Z.Q., Feng, L.Q., 2011. Temporal and spatial variations of chemical oxygen demand in Lake Taihu, China, from 2005 to 2009. Hydrobiologia 665 (1), 129e141. Zhang, Y.L., Liu, M.L., Qin, B.Q., Feng, S., 2009a. Photochemical degradation of chromophoric dissolved organic matter exposed to simulated UV-B and natural solar radiation. Hydrobiologia 627 (1), 159e168. Zhang, Y.L., Qin, B.Q., Zhu, G.W., Zhang, L., Yang, L.Y., 2007. Chromophoric dissolved organic matter (CDOM) absorption characteristics in relation to fluorescence in Lake Taihu, China, a large shallow subtropical lake. Hydrobiologia 581 (1), 43e52. Zhang, Y.L., Van Dijk, M.A., Liu, M.L., Zhu, G.W., Qin, B.Q., 2009b. The contribution of phytoplankton degradation to chromophoric dissolved organic matter (CDOM) in eutrophic shallow lakes: field and experimental evidences. Water Research 43 (18), 4685e4697. Zhang, Y.L., Yin, Y., Zhang, E.L., Zhu, G.W., Liu, M.L., Feng, L.Q., Qin, B.Q., Liu, X.H., 2011. Spectral attenuation of ultraviolet and visible radiation in lakes in the Yunnan Plateau, and the middle and lower reaches of the Yangtze River, China. Photochemical and Photobiological Sciences 10 (4), 469e482. Zhang, Y.L., Zhang, E.L., Yin, Y., van Dijk, M.A., Feng, L.Q., Shi, Z. Q., Liu, M.L., Qin, B.Q., 2010. Characteristics and sources of chromophoric dissolved organic matter in lakes of the Yungui Plateau, China, differing in trophic state and altitude. Limnology and Oceanography 55 (6), 2645e2659.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 2 3 e5 1 2 9

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Phase transformation and its role in stabilizing simulated lead-laden sludge in aluminum-rich ceramics Xingwen Lu, Kaimin Shih* Department of Civil Engineering, The University of Hong Kong, Pokfulam Road, Hong Kong, Hong Kong SAR, China

article info

abstract

Article history:

This study investigated the mechanisms of stabilizing lead-laden sludge by blending it into

Received 4 April 2011

the production process of aluminum-rich ceramics, and quantitatively evaluated the

Received in revised form

prolonged leachability of the product phases. Sintering experiments were performed using

28 June 2011

powder mixtures of lead oxide and g-alumina with different Pb/Al molar ratios within the

Accepted 12 July 2011

temperature range of 600e1000
C. By mixing lead oxide with g-alumina at a Pb/Al molar

Available online 23 July 2011

ratio of 0.5, the formation of PbAl2O4 is initiated at 700
C, but an effective formation was observed when the temperature was above 750
C for a 3-h sintering time. The formation

Keywords:

and decomposition of the intermediate phase, Pb9Al8O21, was detected in this system

Sludge

within the temperature range of 800e900
C. When the lead oxide and g-alumina mixture

Lead

was sintered with a Pb/Al molar ratio of 1:12, the PbAl12O19 phase was found at 950
C and

Alumina

effectively formed at 1000
C. In this system, an intermediate phase Pb3(CO3)2(OH)2 was

Stabilization

observed at the temperature range of 700e950
C. Over longer leaching periods, both

Leaching behavior

PbAl2O4 and PbAl12O19 were superior to lead oxide in immobilizing lead. Comparing the leaching results of PbAl2O4 and PbAl12O19 demonstrated the higher intrinsic resistance of PbAl12O19 against acid attack. To reduce metal mobility, this study demonstrated a preferred mechanism of stabilizing lead in the aluminate structures by adding metalbearing waste sludge to the ceramic processing of aluminum-rich products. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

The contamination of ground and surface water with hazardous metals is a global problem and has become a growing threat to human health. The hazardous metals of particular concern include cadmium (Cd), chromium (Cr), copper (Cu), lead (Pb), manganese (Mn), mercury (Hg), nickel (Ni), vanadium (V) and zinc (Zn) (Gupta et al., 2009), which can enter the food chain through drinking water and crop irrigation (Al-Degs et al., 2001). In particular, lead is a widely found and nonbiodegradable heavy metal that tends to accumulate in the cells of living organisms, which causes severe damage to the kidneys, liver, and the nervous and reproductive systems of

humans (Gupta et al., 2011). Common sources of lead are the wastewater that comes from industries engaged in the manufacturing of lead batteries and oil-based paints, mining, plating, electronics and wood production (Jalali et al., 2002; Gupta et al., 2001; Conrad and Hansen, 2007). Current common techniques used to treat lead-containing wastewater include chemical precipitation, electrochemical reduction, ion exchange, coagulation, adsorption, and membrane processes (Husein et al., 1998; Lin and Navarro, 1999; Petruzzelli et al., 1999; Saeed et al., 2005; Doyurum and Celik, 2006; Ali and Gupta, 2007). However, most of these methods produce enormous amounts of residual sludge that causes secondary pollution and additional operational cost (Gupta and Suhas, 2009).

* Corresponding author. Tel.: þ86 852 2859 1973; fax: þ86 852 2559 5337. E-mail address: [email protected] (K. Shih). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.015

5124

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 2 3 e5 1 2 9

Sludge containing hazardous metal residues should be disposed of in controlled landfills. However, the high cost of this strategy, combined with the limited number of landfills capable of accepting toxic metal waste, makes the development of effective and economical treatment technologies essential. Many researchers have attempted to immobilize toxic metals using sorbents or cements and then correlating the performance directly with metal leachability (Kapoor and Viraraghavan, 1996; Lin et al., 1998; Bailey et al., 1999). However, solidification/stabilization technologies via sorption or cementation mechanisms may not be successful in the prevention of metal leaching in acidic environments (Bonen and Sarkar, 1995; Yousuf et al., 1995). Several processes of stabilizing radioactive waste in glass and ceramic materials through thermal reaction have been successfully demonstrated (Lewis et al., 1993; Wronkiewicz et al., 1997). However, the end products of these processes are radioactive and must be stored in geologic repositories. For non-radioactive hazardous metals, studies aiming to design a waste-to-resource strategy have shown the potential and mechanism of stabilizing them in ceramic products. For example, the reaction mechanism and incorporating efficiency for nickel- and copper-containing sludge with aluminum- or iron-rich precursors during the ceramic sintering processes were reported by Shih et al. (2006a, 2006b), Tang et al. (2010), and Hu et al. (2010). The intrinsic metal leachability of theses sintered product phases was also investigated and the results clearly indicated the importance of initiating certain key phase transformations to significantly reduce leaching of the metal from the products. Studies have shown that the calcining of lead oxide (PbO) on g-Al2O3 produces the various phases of magnetoplumbite-like structure (PbAl12O19) and lead aluminate (PbAl2O4), which can be used as catalysts for oxidative coupling with methane. However, the incorporation mechanism and phase transformation pathway have not been discussed in detail (Wendt et al., 1988; Park and Chang, 1993). Both PbAl12O19 and PbAl2O4 have been reported in the equilibrium phase diagram of the PbOeAl2O3 system (Kuxmann and Fischer, 1974), and the published PbOeAl2O3eSiO2 equilibrium phase diagram also confirmed the presence of PbAl12O19 and PbAl2O4 (Chen et al., 2001). These equilibrium studies have provided the opportunity to observe the potential interaction between PbO and aluminum-rich precursors under industrial sintering processes. Therefore, incorporating lead-laden sludge into ceramic sintering may be able to initiate a beneficial phase-transformation process to further stabilize lead-containing waste. The toxicity characteristic leaching procedure (TCLP), which was designed by the U.S. Environmental Protection Agency (EPA), is commonly used to assess the hazardous nature of metal-bearing waste (U.S. EPA, 1998; 1992). Using acetic acid (pH 2.88) as leaching fluid to simulate the presence of organic materials in municipal landfill leachate (Halim et al., 2004; U.S. EPA, 1998; MacKenzie et al., 2000), the TCLP simulates the worst possible scenario for co-disposing of wastes in landfill. However, due to the dependence of the TCLP on a single-point and short-term leaching result, its use in predicting leaching may result in misclassification of wastes, which consequently leads to the underestimation of long-term leachability or to unnecessary treatment cost

(Kosson et al., 2002). As a result, extending this standardized leaching procedure will be a useful way to observe the leaching behavior of metal-bearing waste under prolonged leaching environment. The objective of this work is to contribute a better understanding of the phase transformation to PbAl12O19 and PbAl2O4 during ceramic sintering, which may be the potential incorporation mechanisms of lead in aluminum-rich ceramics. The approach was to blend PbO with an alumina precursor for a short sintering process (3 h) with temperatures ranging from 600 to 1000
C. Furthermore, a prolonged leaching procedure modified from the TCLP was carried out to evaluate the long-term stabilization effects of lead in the two product phases.

2.

Materials and methods

PbO was purchased from SigmaeAldrich. The phase composition of the PbO powder was identified by X-ray diffraction (XRD) method as a mixture of a-PbO (litharge) and b-PbO (massicot) phases. PbO powder gave a measured surface area of 0.51 m2/g after 300
C heating and He-gas purging for 3 h degassing. The surface area was measured by a Beckman Coulter SA3100 Surface Area and Pore Size Analyzer using the BET method. The g-Al2O3 was prepared from PURAL SB powder fabricated by Sasol with an average particle size w45 mm. The phase of PURAL SB powder was identified by XRD as boehmite (AlOOH; ICDD PDF # 74-1875), and it was successfully converted to the g-Al2O3 phase after heat treatment at 650
C for 3 h (Zhou and Snyder, 1991; Wang et al., 2005). Lead incorporation experiments were conducted using PbO to simulate the high temperature phase of lead in sludge under sintering condition. The g-alumina precursor and PbO were mixed by ball milling in water slurry at Pb/Al molar ratios of 1:2 and 1:12 for 18 h. The slurry samples were then dried and homogenized by mortar grinding. The derived powder was pressed into 20-mm pellets at 650 MPa to ensure consistent compaction of the powder sample for the sintering process. The pellets were sintered at targeted temperatures from 600 to 1000
C for 3 h (Sun et al., 2001) and then quenched in air to room temperature. The total mass loss after sintering was less than 1 wt.%. After sintering, the samples were ground in an agate mortar and pestle to a particle size of no more than 10 mm for XRD analysis and leaching test. Phase transformation during sintering was determined by using the X-ray powder diffraction technique. The X-ray powder diffraction data of the samples were collected on a Bruker D8 Advance X-ray powder diffractometer equipped with a Cu Ka radiation and a LynxEye detector. The diffractometer was operated at 40 kV and 40 mA, and the 2q scan range was from 10
to 80
, with a step size of 0.02
and a scan speed of 0.3 s/step. Qualitative phase identification was done using Eva XRD Pattern Processing software (Bruker Co. Ltd.) by matching powder XRD patterns with those retrieved from the standard powder diffraction database of the International Centre for Diffraction Data (ICDD PDF-2 Release, 2008). After BET surface area measurement of the powder samples, the leachabilities of the single-phase samples were

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 2 3 e5 1 2 9

5125

tested using a leaching procedure modified from the U.S. EPA SW-846 Method 1311eTCLP with a pH 2.9 acetic acid solution (extraction fluid # 2) as the leaching fluid. Each leaching vial was filled with 10 ml of TCLP extraction fluid and 0.5 g of powder. The leaching vials were rotated end-over-end at 60 rpm for 0.75e23 days. At the end of each agitation period, the leachates were filtered with 0.2 mm syringe filters, the pH values were measured, and the concentrations of lead was analytically determined using a flame-type Perkin Elmer model 3300 atomic absorption spectrometer (Perkin Elmer Co. Ltd.).

3.

Results and discussion

3.1.

Formation of PbAl2O4

The two polymorphs of lead(II) oxide (the tetragonal form aPbO and the orthorhombic form b-PbO) could transform into one another at certain temperature or under certain pressure. The transition of low-temperature-phase a-PbO to b-PbO may occur when the temperature reaches 540
C (Wriedt, 1988). Moreover, the polymorphic transformation of b-PbO to a-PbO after ball-milling has been observed by Senna and Kuno (1971). As the sludge used for ceramic sintering preparation may experience both mechanical and thermal processes, the as-received and mixed-phase PbO was directly used as the raw material for sintering with g-Al2O3. To explore the phases that may appear in the high-lead concentration sintering, Fig. 1 demonstrated the XRD patterns of the products from the Pb/Al ¼ 1:2 mixture of PbO þ g-Al2O3 powder sintered at 600e1000
C for 3 h. The result showed that when sintered at 700
C, the formation of crystalline PbAl2O4 phase described by Eq. (1) was first observed. PbO þ g  Al2 O3 /PbAl2 O4

(1)

Geller and Bunting (1943) reported the formation of PbAl2O4 at temperature 600
C, but their experiment involved 1e2 months of dwelling time. Therefore, this difference may suggest that the formation of PbAl2O4 at temperatures below 700
C is largely limited by the prevailing slow diffusion, although it is thermodynamically feasible at temperatures above 600
C. The solidestate reaction is affected by both thermodynamic constraint and kinetic process. Below 700
C, the PbAl2O4 phase formed by the short sintering scheme might only be limited at the grain boundary of reactants, and this small quantity in the sample was not reflected in the XRD results. As the intensity of the PbAl2O4 phase increased when the sintering temperature increases, a significantly higher intensity of PbAl2O4 signal was achieved at 750
C, and no Bragg reflection of a-PbO or b-PbO phase was observed when the temperature was above 800
C. An intermediate product of Pb9Al8O21 (ICDD PDF # 73-1875) was found when the sintering temperature was around 750
C, but its Bragg diffraction peaks disappeared in the sample sintered at 900
C. Above 950
C, PbAl2O4 became the only product phase in the samples of this raw material system. The powder diffraction database shows that the strongest diffraction peak (2q ¼ 29.12
) of the PbAl2O4 phase overlapped

Fig. 1 e The XRD patterns of the PbO D g-Al2O3 system with a Pb:Al molar ratio of 1:2. The result shows the formation of PbAl2O4 (ICDD PDF # 85-1289) and Pb9Al8O21 (ICDD PDF # 73-1875) when sintering samples at temperatures between 600 and 1000
C for 3 h.

with the strongest of b-PbO (2q ¼ 29.08
) in the XRD patterns. As the second strongest peak of PbAl2O4 phase was located at 2q ¼ 19.96
and a major peak reflected from the (4 1 0) plane of Pb9Al8O21 was located at 2q ¼ 27.72
, the 2q range of 18.8e22.2
and 27.4e28.0
demonstrated the details in the phase transformation processes to PbAl2O4 and Pb9Al8O21 (Fig. 2). Within the 2q range of 18.8e22.2
, a few diffraction peaks of the PbAl2O4 phase could be observed to represent the formation of PbAl2O4 phase, and they are reflected by the (0 2 0), (0 1 1), (2 0 0), and (1 2 0) crystal planes of PbAl2O4 structure, which correspond to 2q of 19.21
, 19.96
, 20.99
, and 21.91
, respectively. A higher sintering temperature resulted in an increase in peak intensity, which indicates a continuing growth of the PbAl2O4 phase from 700 to 1000
C sintering. The Fig. 2(b) illustrates that the peak intensity reflected by the (4 1 0) planes of Pb9Al8O21 increased with an increase in sintering temperature before reaching its maximum at 800
C; this was followed by a decrease in intensity when the sintering temperature further increased. The formation of Pb9Al8O21 during the sintering process was probably initiated by the reaction between PbO and gAl2O3 in a non-equilibrium system and due to the insufficient sintering time at the lower temperature range. The efficiency of a homogeneous reaction generally depends on the

5126

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 2 3 e5 1 2 9

Fig. 2 e The XRD patterns (a) (2q [ 18.8e22.2
) and (b) (2q [ 27.4e28.0
) of the PbO D g-Al2O3 system with a Pb:Al molar ratio of 1:2. The results show the changes of peak intensity of the PbAl2O4 (ICDD PDF # 85-1289) and Pb9Al8O21 (ICDD PDF # 731875) phases in the products sintered at 600e1000
C for 3 h.

encountering rate between reactant molecules (Kukukova et al., 2009). The short sintering time and low temperature were not able to provide sufficient contact, and hence the formation of PbAl2O4 is incomplete and an intermediate compound appeared in the product. At higher temperatures, more intensive interaction between reactants was achieved and Pb9Al8O21 was not observed in the sintered samples. Therefore, to more effectively achieve the formation of PbAl2O4, the temperature of a short sintering scheme should be higher than 950
C.

3.2.

Formation of PbAl12O19

When the PbO molar content in the PbOeAl2O3 system was lower than 50%, lead dodecaaluminate (PbAl12O19; ICDD PDF # 80-1174) was the only product phase reported in equilibrium experiments (Kuxmann and Fischer, 1974; Chen et al., 2001). Therefore, a potential thermal reaction of incorporating lead by g-Al2O3 precursor at lower lead level is described by Eq. (2): PbO þ 6g  Al2 O3 /PbAl12 O19

(2)

Fig. 3(a) presents the XRD patterns of the 600e1000
C sintered PbO þ g-Al2O3 mixtures with a Pb/Al molar ratio of 1:12, and shows that the peaks of the PbAl12O19 phase appeared when the sintering temperature exceeded 950
C. The lowest temperature at which PbAl12O19 was observed after a short sintering scheme was about 200
C higher than that derived from the equilibrium experiment carried out by Kuxmann and Fischer (1974). Similarly, this discrepancy can be explained by the shorter sintering time and the potential diffusion barrier created by the newly formed PbAl12O19 layer between the PbO and g-Al2O3 grains. At 1000
C, substantial growth of PbAl12O19 phase was found and PbAl12O19 was the only lead-containing phase in the product. However, an intermediate phase, hydrocerussite (Pb3(CO3)2(OH)2; ICDD PDF # 73-4362), formed

in the products sintered at temperatures between 700
C and 950
C. The formation of poor crystalline Pb3(CO3)2(OH)2 phase was probably due to the instability of product phase(s) sintered at 700e950
C, which might be vulnerable to the attack of CO2 and moisture in the air during sample quenching. Nevertheless, when sintered at 1000
C, the significant conversion to the PbAl12O19 phase completely eliminated the formation of hydrocerussite in the product, and this result may also indicate superior stability of the PbAl12O19 phase in the sintered product. Further growth details of the PbAl12O19 phase can be carried out by observing its two major peaks located at 2q ¼ 18.81
and 2q ¼ 36.10
, which correspond to the (1 0 1) and (1 1 4) planes of the PbAl12O19 structure (Fig. 3(b)). Starting from 950
C, the initiation of poor crystalline PbAl12O19 phase from sintering the PbO þ g-Al2O3 mixture could be observed. Significant growth of the PbAl12O19 phase clearly occurred within the temperature range of 950e1000
C. This result confirms the potential of forming PbAl12O19 to incorporate lead into the aluminum-rich ceramics for systems with lower Pb/Al ratios. Therefore, because PbAl2O4 and PbAl12O19 were identified as the potential hosting phases for lead in the ceramic products, both of them were later analyzed by the prolonged toxicity characteristic leaching procedure to observe their leachabilities and leaching behavior.

3.3.

Leachabilities of product phases

To compare the effects of lead stabilization in different hosting forms, the use of single-phase samples in leaching experiments can further facilitate the interpretation of leachate data. In this study, both PbAl2O4 and PbAl12O19 single-phase samples were prepared from sintering the raw materials with Pb/Al molar ratios of 1:2 and 1:12 at 950 and 1000
C, respectively. The sintered products were then ground

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 2 3 e5 1 2 9

5127

Fig. 4 e The pH values of the leachates from PbO, PbAl2O4 (950
C/3 h sintered PbO D g-Al2O3 sample with Pb:Al molar ratio [ 1:2), and PbAl12O19 (1000
C/3 h sintered PbO D g-Al2O3 sample with Pb:Al molar ratio [ 1:12) powders. The leaching solution was TCLP extraction fluid # 2 (acetic acid solution) at pH 2.9. Each leaching vial was filled with 10 ml extraction fluid and 0.5 g powder sample, and then rotated end-over-end for between 0.75 and 23 days.

Fig. 3 e The XRD results of (a) products sintered from the PbO D g-Al2O3 samples with a Pb:Al molar ratio of 1:12 at temperatures between 600 and 1000
C for 3 h, and (b) the significant increase of PbAl12O19 after sintering at 1000
C.

into powder and confirmed by XRD to be the single-phase samples without observable reactant residues. In addition, the property of PbO powder used in sintering the raw material was also examined by the leaching experiment to compare it with those of aluminate phases. As the leaching process of solids is likely to be a surface reaction between solid and leachate, the observed lead concentrations in leachate are potentially proportional to the surface areas of samples. The BET surface areas of the powder samples before leaching were measured and yielded the values of 0.51 m2/g for PbO, 0.63 m2/ g for PbAl2O4, and 3.76 m2/g for PbAl12O19. Since solution pH is usually responsible for the lead leachability (Kim et al., 2011; Pereira et al., 2001), the leachate pH after the prolonged leaching tests was measured and shown in Fig. 4. Although both PbO and PbAl2O4 showed increases in leachate pH in the first couple of days, the PbAl2O4 leachates

stabilized at pH w6.5 whereas the pH values of the PbO leachates increased to pH w9.3. In contrast, the pH values of the PbAl12O19 leachates were kept close to the initial pH value of its leaching fluid throughout the entire leaching period. The increase of leachate pH is likely to be due to the dissolution of crystal cations through ion exchange with protons in the solution. This is usually accompanied by the destruction of the crystal structure by the acidic leaching fluid. Hence, the most significant increase in PbO leachate pH indicates that PbO is very vulnerable to proton-mediated dissolution. Similarly, the very limited change of PbAl12O19 leachate pH may also indicate its strong resistance to acid attack. The leaching process of metals from a solid sample is probably dominated by surface reactions and also influenced by the available amount of metals in the sample. In this study, we demonstrated the leachability of lead from the samples by normalizing the concentrations of lead in the leachates with respect to its weight percentage in the solid sample and the surface areas of the powder samples (Fig. 5). The normalized leached lead per surface area of sample (NLPbSA; m2) was calculated as follows, NLPbSA ¼ 106 

n CPb ,AWPb  k SW,SA,MWPhase

(3)

where n is the number of Pb atoms in each molecule; k is the ratio of sample weight (g) to extraction fluid volume (mL); CPb is lead concentration in leachate (mg/L); AWPb represents the atomic weight of lead; SW is sample weight (g); SA is the solid sample surface area (m2/g) and MWPhase represents the molecular weight of tested phase. The lead concentration observed in PbO leachate was w21 g/L at the end of 23-day leaching process. After

5128

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 2 3 e5 1 2 9

However, the sintering experiment revealed that a higher lead content phase, Pb9Al8O21, may occur as an intermediate between 800 and 900
C and when the Pb/Al molar ratio in the raw materials is around 1:2. In addition, when the raw materials have a lower Pb/Al molar ratio (1:12), another intermediate phase, Pb3(CO3)2(OH)2, was observed at the temperature range of 700e950
C. The approach we outline here provides an effective way of incorporating lead-laden sludge into aluminum-rich ceramic precursors by the formation of PbAl2O4 and PbAl12O19 in an optimal temperature range of 950e1000
C.

Fig. 5 e Leached lead normalized by the surface area and lead weight percentage of PbO, PbAl2O4, and PbAl12O19 samples. The surface areas of PbO, PbAl2O4, and PbAl12O19 were 0.51 m2/g, 0.63 m2/g, and 3.76 m2/g, respectively. Each leaching vial was filled with 10 ml extraction fluid # 2 (pH 2.9) and 0.5 g powder sample. The curve in the inset (right) further provides the details of the normalized lead concentrations in the PbAl12O19 leachates.

normalization, it was nearly three times higher than that from the PbAl2O4 samples (w10 g/L before normalization) and over three orders of magnitude higher than that from the PbAl12O19 samples (w80 mg/L before normalization) at the end of the leaching period. At leachate pH less than 10, the leachability of lead was reported to increase at lower pH values (Dubey and Townsend, 2004; Pereira et al., 2001). Comparing with the pH value of PbO leachate at w9.3, PbAl2O4 and PbAl12O19 leachate remained at a lower pH value of around 6.5 and 3.5 throughout the whole leaching process, but the concentrations of the leached lead in PbAl2O4 and PbAl12O19 leachate were still much lower than that of the PbO leachate at the end of leaching experiment. As shown in Fig. 5, the significant difference points out that both PbAl2O4 and PbAl12O19 have much higher intrinsic resistances to acid attack compared to the PbO phase. This result was also consistent with the observation in the changes of leachate pH. The inset of Fig. 5 further provides the details of normalized lead concentrations in PbAl12O19 leachate, which showed the high stability of PbAl12O19 in the prolonged leaching period. Therefore, the leaching results suggest that the phase transformation to lead aluminates, particularly to PbAl12O19 phase, is a highly effective lead-stabilization strategy.

4.

Conclusions

The results of this study indicate that the formation of lead aluminates, PbAl2O4 and PbAl12O19 phases, may be initiated by sintering the simulated lead-bearing sludge and aluminum-rich ceramic precursors at temperatures above 750 and 950
C. The findings of the prolonged TCLP leachability test show the superiority of both PbAl2O4 and PbAl12O19 over PbO in resisting proton-mediated dissolution. This result provides a promising strategy to further stabilize lead by incorporating it into its aluminate structures.

Acknowledgments We gratefully acknowledge the funding for this research provided by the General Research Fund Scheme of the Research Grants Council of Hong Kong (gs1) (HKU 716310E). The authors are also grateful to Ms. Vicky Fung for assisting us with the atomic absorption analysis.

references

Al-Degs, Y., Khraisheh, M.A.M., Tutunji, M.F., 2001. Sorption of lead ions on diatomite and manganese oxides modified diatomite. Water Research 35 (15), 3724e3728. Ali, I., Gupta, V.K., 2007. Advances in water treatment by adsorption technology. Nature Protocols 1 (6), 2661e2667. Bailey, S.E., Olin, T.J., Bricka, R.M., Adrian, D.D., 1999. A review of potentially low-cost sorbents for heavy metals. Water Research 33 (11), 2469e2479. Bonen, D., Sarkar, S.L., 1995. The effects of simulated environmental attack on immobilization of heavy metals doped in cement-based materials. Journal of Hazardous Materials 40 (3), 321e335. Chen, S., Zhao, B., Hayes, P.C., Jak, E., 2001. Experimental study of phase equilibria in the PbOeAl2O3eSiO2 system. Metallurgical and Materials Transactions B: Process Metallurgy and Materials Processing Science 32 (6), 997e1005. Conrad, K., Bruun Hansen, H.C., 2007. Sorption of zinc and lead on coir. Bioresource Technology 98 (1), 89e97. Doyurum, S., Celik, A., 2006. Pb(II) and Cd(II) removal from aqueous solutions by olive cake. Journal of Hazardous Materials 138 (1), 22e28. Dubey, B., Townsend, T., 2004. Arsenic and lead leaching from the waste derived fertilizer ironite. Environmental Science and Technology 38 (20), 5400e5404. Geller, R.F., Bunting, E.N., 1943. Report on the systems lead oxidealumina and lead oxide-alumina-silica. Journal of Research of the National Bureau of Standards 31 (5), 255e270. Gupta, V.K., Agarwal, S., Saleh, T.A., 2011. Synthesis and characterization of alumina-coated carbon nanotubes and their application for lead removal. Journal of Hazardous Materials 185 (1), 17e23. Gupta, V.K., Carrott, P.J.M., Ribeiro Carrott, M.M.L., Suhas, 2009. Low-Cost adsorbents: growing approach to wastewater treatment a review. Critical Reviews in Environmental Science and Technology 39 (10), 783e842. Gupta, V.K., Gupta, M., Sharma, S., 2001. Process development for the removal of lead and chromium from aqueous solutions using red mud - An aluminium industry waste. Water Research 35 (5), 1125e1134.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 2 3 e5 1 2 9

Gupta, V.K., Suhas, 2009. Application of low-cost adsorbents for dye removal e A review. Journal of Environmental Management 90 (8), 2313e2342. Halim, C.E., Scott, J.A., Natawardaya, H., Amal, R., Beydoun, D., Low, G., 2004. Comparison between acetic acid and landfill leachates for the leaching of Ph(II), Cd(II), As(V), and Cr(VI) from clementitious wastes. Environmental Science and Technology 38 (14), 3977e3983. Hu, C.Y., Shih, K., Leckie, J.O., 2010. Formation of copper aluminate spinel and cuprous aluminate delafossite to thermally stabilize simulated copper-laden sludge. Journal of Hazardous Materials 181 (1e3), 399e404. Husein, M.M., Vera, J.H., Weber, M.E., 1998. Removal of lead from aqueous solutions with sodium caprate. Separation Science and Technology 33 (12), 1889e1904. Jalali, R., Ghafourian, H., Asef, Y., Davarpanah, S.J., Sepehr, S., 2002. Removal and recovery of lead using nonliving biomass of marine algae. Journal of Hazardous Materials 92 (3), 253e262. Kapoor, A., Viraraghavan, T., 1996. Discussion: treatment of metal industrial wastewater by fly ash and cement fixation. Journal of Environmental Engineering 122 (3), 243. Kim, E.J., Herrera, J.E., Huggins, D., Braam, J., Koshowski, S., 2011. Effect of pH on the concentrations of lead and trace contaminants in drinking water: a combined batch, pipe loop and sentinel home study. Water Research 45 (9), 2763e2774. Kosson, D.S., Van Der Sloot, H.A., Sanchez, F., Garrabrants, A.C., 2002. An integrated framework for evaluating leaching in waste management and utilization of secondary materials. Environmental Engineering Science 19 (3), 159e204. Kukukova, A., Aubin, J., Kresta, S.M., 2009. A new definition of mixing and segregation: three dimensions of a key process variable. Chemical Engineering Research and Design 87 (4), 633e647. Kuxmann, U., Fischer, P., 1974. Lead monoxide-aluminum oxide, lead monoxide-calcium oxide, and lead monoxide-silicon dioxide phase diagrams. Erzmetall 27 (11), 533e537. Lewis, M.A., Fischer, D.F., Smith, L.J., 1993. Salt-occluded zeolites as an immobilization matrix for chloride waste salt. Journal of the American Ceramic Society 76 (11), 2826e2832. Lin, C.F., Lo, S.S., Lin, H.Y., Lee, Y., 1998. Stabilization of cadmium contaminated soils using synthesized zeolite. Journal of Hazardous Materials 60 (3), 217e226. Lin, S.W., Navarro, R.M.F., 1999. An innovative method for removing Hg2þ and Pb2þ in ppm concentrations from aqueous media. Chemosphere 39 (11), 1809e1817. MacKenzie, K.J.D., Temuujin, J., Smith, M.E., Angerer, P., Kameshima, Y., 2000. Effect of mechanochemical activation on the thermal reactions of boehmite (g-AlOOH) and g-Al2O3. Thermochimica Acta 359 (1), 87e94. Park, S.E., Chang, J.S., 1993. Oxidative coupling of methane over PbO/PbAl2O4 catalysts. Studies in Surface Science and Catalysis 75, 2233e2236. Pereira, C.F., Rodriguez-Piero, M., Vale, J., 2001. Solidification/ stabilization of electric arc furnace dust using coal fly ash:

5129

analysis of the stabilization process. Journal of Hazardous Materials 82 (2), 183e195. Petruzzelli, D., Pagano, M., Tiravanti, G., Passino, R., 1999. Lead removal and recovery from battery wastewaters by natural zeolite clinoptilolite. Solvent Extraction and Ion Exchange 17 (3), 677e694. Saeed, A., Iqbal, M., Akhtar, M.W., 2005. Removal and recovery of lead(II) from single and multimetal (Cd, Cu, Ni, Zn) solutions by crop milling waste (black gram husk). Journal of Hazardous Materials 117 (1), 65e73. Senna, M., Kuno, H., 1971. Polymorphic transformation of PbO by isothermal wet ball-milling. Journal of the American Ceramic Society 54 (5), 259e262. Shih, K., White, T., Leckie, J.O., 2006a. Spinel formation for stabilizing simulated nickel-laden sludge with aluminum-rich ceramic precursors. Environmental Science and Technology 40 (16), 5077e5083. Shih, K., White, T., Leckie, J.O., 2006b. Nickel stabilization efficiency of aluminate and ferrite spinels and their leaching behavior. Environmental Science and Technology 40 (17), 5520e5526. Sun, D.D., Tay, J.H., Cheong, H.K., Leung, D.L.K., Qian, G., 2001. Recovery of heavy metals and stabilization of spent hydrotreating catalyst using a glass-ceramic matrix. Journal of Hazardous Materials 87 (1e3), 213e223. Tang, Y., Shih, K., Chan, K., 2010. Copper aluminate spinel in the stabilization and detoxification of simulated copper-laden sludge. Chemosphere 80 (4), 375e380. USEPA., 1998. Applicability of the Toxicity Characteristic Leaching Procedure to Mineral Processing Wastes and its Salts. available from: http://www.epa.gov/osw/nonhaz/industrial/ special/mining/minedock/tclp/tcremand.pdf. USEPA., 1992. U.S. EPA Method 1311- Toxicity Characteristic Leaching Procedure and its Salts. http://www.epa.gov/osw/ hazard/testmethods/sw846/pdfs/1311.pdf available from: Wang, Y., Suryanarayana, C., An, L., 2005. Phase transformation in nanometer-sized g-alumina by mechanical milling. Journal of the American Ceramic Society 88 (3), 780e783. Wendt, G., Meinecke, C.D., Schmitz, W., 1988. Oxidative dimerization of methane on lead oxide-alumina catalysts. Applied Catalysis 45 (2), 209e220. Wriedt, H.A., 1988. The O-Pb (oxygen-lead) system. Bulletin of Alloy Phase Diagrams 9 (2), 106e127. Wronkiewicz, D.J., Bates, J.K., Buck, E.C., Hoh, J.C., Emery, J.W., Wang, L.M., 1997. Radiation Effects in Moist-air Systems and the Influence of Radiolytic Product Formation on Nuclear Waste Glass Corrosion. Argonne National Laboratory Report No. ANL-97/15. Argonne. Yousuf, M., Mollah, A., Vempati, R.K., Lin, T.C., Cocke, D.L., 1995. The interfacial chemistry of solidification/stabilization of metals in cement and pozzolanic material systems. Waste Management 15 (2), 137e148. Zhou, R.S., Snyder, R.L., 1991. Structures and transformation mechanisms of the eta, gamma and theta transition aluminas. Acta Crystallographica Section B 47 (5), 617e630.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 3 0 e5 1 4 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Evaluating sludge minimization caused by predation and viral infection based on the extended activated sludge model No. 2d Xiaodi Hao a,*, Qilin Wang a,b, Yali Cao a, Mark C.M. van Loosdrecht c a

Key Laboratory of Urban Stormwater System and Water Environment/R & D Centre for Sustainable Environmental Biotechnology, Beijing University of Civil Engineering and Architecture, Ministry of Education, 1 Zhanlanguan Road, Xicheng District, Beijing 100044, PR China b Advanced Water Management Centre (AWMC), The University of Queensland, St Lucia, Brisbane QLD 4072, Australia c Department of Biotechnology, Delft University of Technology, Julianalaan 67, 2628 BC Delft, The Netherlands

article info

abstract

Article history:

The Activated Sludge Model No. 2d (ASM2d) was extended to incorporate the processes of

Received 27 May 2011

both predation and viral infection. The extended model was used to evaluate the contri-

Received in revised form

butions of predation and viral infection to sludge minimization in a sequencing batch

11 July 2011

reactor (SBR) system enriching polyphosphate-accumulating organisms (PAOs). Three

Accepted 12 July 2011

individual decay processes formulated according to the general model rules were used in

Available online 22 July 2011

the extended model. The model was firstly calibrated and validated by different experimental results. It was used to evaluate the potential extent of predation and viral infection

Keywords:

on sludge minimization. Simulations indicate that predation contributes roughly two times

Predation

more to sludge minimization than viral infection in the SBR system enriching PAOs. The

Viral infection

sensitivity analyses of the selected key parameters reveal that there are thresholds on both

Sludge minimization

predation and viral infection rates, if they are too large a minimal sludge retention time is

Polyphosphate-accumulating

obtained and the effluent quality is deteriorating. Due to the thresholds, the contributions of predation and viral infection to sludge minimization are limited to a maximal extent of

organisms (PAOs) Sequencing

batch

reactor

(SBR)

about 21% and 9%, respectively. However, it should be noted that the parameters concerning predation and viral infection were not calibrated separately by independent

system Activated

sludge

model

No.

2d

experiment in our study due to the lack of an effective method, especially for the parameters regarding viral infection. Therefore, it is essential to better evaluate these

(ASM2d)

parameters in the future. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Sludge minimization has become a hot topic in biological wastewater treatment systems (BWTSs). Sludge production is directly related to endogenous processes (van Loosdrecht and Henze, 1999; Lopez et al., 2006; Schuler and Jassby, 2007; Hao et al., 2010a). Endogenous processes could be classified into

two levels: cell level and community level (Hao et al., 2010a). The former refers to endogenous respiration and the latter consists of predation by higher organisms, viral infection and other factors. Endogenous respiration has been mostly understood and applied in sludge minimization, such as extended aeration processes (oxidation ditches) and aerobic sludge digestion. Predation by higher organisms also plays an

* Corresponding author. Tel.: þ86 10 8060 4512; fax: þ86 10 6832 2128. E-mail address: [email protected] (X. Hao). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.013

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 3 0 e5 1 4 0

important role in sludge minimization and has been regularly studied (Curds, 1971; Cech et al., 1994; Lee and Welander, 1996; Ratsak et al., 1996; Ghyoot and Verstraete, 2000; Moussa et al., 2005; Akpor and Momba, 2010; Ni et al., 2010). However, a quantitative process description of the contribution of predation is still minimal. Viral infection is even less studied although the use of viruses in biomass minimization has been reported in literatures (Suttle, 1994; Suttle and Chan, 1994; Withey et al., 2005; Otawa et al., 2007; Kunin et al., 2008). To what extent do predation and viral infection contribute to sludge minimization and how do they affect the effluent quality, especially for biological nutrient removal (BNR) systems? Less residual sludge is expected in BWTSs, but less biomass in BWTSs might cause a deteriorated effluent quality in BNR systems. Therefore, any potential technical measures of predation and viral infection for sludge minimization have to be evaluated before practiced in engineering. The interactions in activated sludge processes are rather complex, therefore a process model could help in improving the interpretation of experimental results and increase our understanding of these endogenous processes. In the current activated sludge models decay coefficients remain unclear, and only one lumped process is described (Henze et al., 1987, 1999; Gujer et al., 1995, 1999; Rieger et al., 2001; Meijer, 2004). This makes it impossible to distinguish the roles of predation and viral infection from the process of decay by the existing models. For this reason, Moussa et al. (2005) and Ni et al. (2010) have developed their models of predation, but the model of viral infection has not been seen from the literature yet. It is really necessary to develop a model involving both predation and viral infection to evaluate the above concern of predation and viral infection in sludge minimization. Within the existing Activated Sludge Model No. 2d (Henze et al., 1999), three individual decay processes were introduced. The extended model was applied to simulate a sequencing batch reactor (SBR) system enriching polyphosphate-accumulating organisms (PAOs) for calibration and validation. Finally, the calibrated and validated model was applied to predict the contributions of predation and viral infection to sludge minimization.

2.

Materials and methods

2.1.

SBR system enriching PAOs

A 5-L fermenter with a 4-L working volume was used as a SBR system to selectively grow an enriched PAOs culture. The SBR continuously operated under alternating anaerobic/aerobic conditions, on the basis of a 6-h operational cycle (feeding: 5 min; anaerobic reaction: 125 min; aerobic reaction: 158 min; sludge extraction: 2 min; settling: 60 min; effluent extraction: 10 min). The SBR was constantly mixed with a stirrer at 150 rpm in feeding, anaerobic/aerobic reactions and sludge extraction phases. Aeration was provided with airflow of 120 L/min, which resulted in the dissolved oxygen (DO) levels generally at 0.1 mg/L in the anaerobic phase and at 3 mg/L in the aerobic phase. During both the anaerobic and aerobic phases, pH was maintained at 7.5  0.05 using 1-M NaOH and 0.5-M HCl

5131

(automatically fed by peristaltic pumps controlled by a computer), and the temperature was controlled at 22  0.5  C with an outer thermostat casing around the fermenter. In each cycle, 2-L synthetic wastewater was fed to the fermenter, resulting in 12-h HRT. SRT was maintained at 12 d, by extracting 83-ml mixed liquid from the fermenter during each cycle. Seeding sludge for the fermenter was taken from a pilotscale BNR system. After about 90 d, the fermenter reached to a steady state with an MLSS concentration of 2,845  35 mg/L. The carbon source in the synthetic influent was alternated between acetic and propionic acids with the influent COD level of 400 mg/L. The alteration of VFAs was to provide a selective advantage to PAOs over GAOs (Lu et al., 2006). During a switching cycle of 36 d (3 SRTs), the fermenter respectively operated for 24 d with the acetic influent and for 12 d with the propionic influent at the same SRT of 12 d (Lu et al., 2007). The COD/P ratio in all the influents was maintained at 20 (g COD/g P). The detailed composition of 2-L synthetic influent can be referred to Hao et al. (2010b). After a steady state was reached and some necessary data was measured (for model calibration), the SBR was switched to operate at SRT of 15 d and the influent COD level of 440 mg/L consisting of 30 d-acetic influent and 15 d-propionic influent, whereas other operation conditions were not changed (for model validation).

2.2.

Extended model

ASM2d was used as a model framework for this work; however, the glycogen metabolism of PAOs is not included in ASM2d. Since glycogen plays a vital role in the metabolic process of PAOs (Smolders et al., 1995; Lu et al., 2007), it was deemed necessary to be included into ASM2d, as done by previous researchers (Yagci et al., 2004; Schuler, 2005; Whang et al., 2007; Lopez-Vazquez et al., 2009). In order to include predation and viral infection, the decay process in ASM2d was split into three individual processes, that is, predationinduced decay, viral infection-induced decay and other factors-induced decay (e.g. toxic substances, natural cell death, etc.). Correspondingly, the stoichiometric matrix and kinetic rate expressions relating to predation and viral infection had to be added to the extended ASM2d. There are 13 state variables in the extended model: biomass for PAOs (X_PAO), polyphosphate for PAOs (X_PP), PHA for PAOs (X_PHA), glycogen for PAOs (X_Gly), ordinary heterotrophic organisms (OHOs: X_H), higher organisms (predators: X_P), inert particulate biomass (dead biomass: X_I), inert soluble substrate (S_I), slowly biodegradable substrate (X_S), fermentable/readily biodegradable substrate (S_F), fermentation product (S_A), phosphate (S_PO4), and dissolved oxygen (S_O2). As there was 60 mg/L of Allyl-N thiourea (ATU) (a nitrification inhibitor) in the influent (40 mg NHþ 4 -N/L) to the SBR system, nitrification could not happen and the influent NHþ 4 -N was only used for the bacterial anabolism, which was further confirmed by the low nitrite and nitrate concentrations in the SBR system at all times (below 0.1 mg N/L). For this reason, the state viable S_NH4 was deleted from the model. Finally, the extended model was established and shown in Tables 1 and 2, with the references of Henze et al. (1999), Gujer et al. (1999), Schuler and Jassby (2007) and Whang et al. (2007). The

5132

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 3 0 e5 1 4 0

Table 1 e Stoichiometric matrix for the extended model (I). Process Aerobic growth of X_PAO Aerobic storage of X_PP Aerobic storage of X_Gly Anaerobic storage of X_PHA Aerobic hydrolysis of X_S Anaerobic hydrolysis of X_S Fermentation of S_F Aerobic growth of X_H on S_A Aerobic growth of X_H on S_F Other factors-induced decay of X_H Other factors-induced decay of X_PAO Other factors-induced decay of X_PP Other factors-induced decay of X_PHA Other factors-induced decay of X_Gly Viral infection on X_P Viral infection on X_H Viral infection on XPAOe Predation on X_H Predation on XPAO

S_A

S_O2

S_F

S_I

S_PO4

11/Y_PAO

i_PBM

Y_PHApp

1

X_PAO 1

1Y_PHAgly 1

1 1/Y_H

11/Y_H 11/Y_H

Y_PO4 1f_SI

f_SI

i_PXS-i_PSI*f_SI-i_PSF*(1-f_SI)

1f_SI

f_SI

i_PXS-i_PSI*f_SI-i_PSF*(1-f_SI)

1

i_PSF i_PBM

1/Y_H

i_PSF/Y_H-i_PBM i_PBM-i_PXI*Fxi-(1-Fxi)*i_PXS 1

i_PBM-i_PXI*Fxi-(1-Fxi)*i_PXS 1 1 1 i_PBM-i_PXI* Fxi-(1Fxi)*i_PXS i_PBM-i_PXI*Fxi-(1Fxi)*i_PXS I_PBM*AaþDdi_PXI*Fxi*A-i_PXS*((1Fxi)*A þBbþCc) i_PBM-i_PXI*Fxi-Y_P*(1Fxi)*i_PBM i_PBM*A þ D-i_PXI* Fxi*A-i_PXS*((1Fxi)*A þ B þ C)

1 þ Y_P*(1Fxi) þ Fxi Y_P*((1Fxi)*A þ B þ C) A  BC  D þ Fxi*A

Decay of X_P

A A

i_PBM-I_PXI*Fxi-(1Fxi)*i_PXS

Stoichiometric matrix for the extended model (II) Process Aerobic growth of X_PAO Aerobic storage of X_PP Aerobic storage of X_Gly Anaerobic storage of X_PHA Aerobic hydrolysis of X_S Anaerobic hydrolysis of X_S Fermentation of S_F Aerobic growth of X_H on S_A Aerobic growth of X_H on S_F Other factors-induced decay of Other factors-induced decay of Other factors-induced decay of Other factors-induced decay of Other factors-induced decay of Viral infection on X_P Viral infection on X_H Viral infection on XPAO Predation on X_H Predation on XPAO Decay of X_P a b c d e

X_H

X_P

X_S

X_PP 1 Y_PO4

X_PHA

X_Gly

1/Y_PAO Y_PHApp Y_PHAgly Y_PHA

1 Y_Gly

X_I

1 1

X_H X_PAO X_PP X_PHA X_Gly

1 1 1

Fxi Fxi

1Fxi 1Fxi 1 1 1 1

1 1

A ¼ X_PAO/(X_PAO þ X_PHA þ X_Gly þ X_PP). B ¼ X_PHA/(X_PAO þ X_PHA þ X_Gly þ X_PP). C ¼ X_Gly/(X_PAO þ X_PHA þ X_Gly þ X_PP). D ¼ X_PP/(X_PAO þ X_PHA þ X_Gly þ X_PP). XPAO ¼ X_PAO þ X_PHA þ X_Gly þ X_PP.

Y_P*(1Fxi) Y_P*((1Fxi)*A þ B þ C) 1

1Fxi 1Fxi (1Fxi)*A þ B þ C

1Fxi

D

B

C

D

B

C

Fxi Fxi Fxi*A Fxi Fxi*A Fxi

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 3 0 e5 1 4 0

5133

Table 2 e Kinetic rate expressions for the extended model. Process Aerobic growth of X_PAO Aerobic storage of X_PP Aerobic storage of X_Gly Anaerobic storage of X_PHA Aerobic hydrolysis of X_S Anaerobic hydrolysis of X_S Fermentation of S_F Aerobic growth of X_H on S_A Aerobic growth of X_H on S_F Other factors-induced decay of X_H Other factors-induced decay of X_PAO Other factors-induced decay of X_PP Other factors-induced decay of X_PHA Other factors-induced decay of X_Gly Viral infection on X_P Viral infection on X_H Viral infection on X_PAO Predation on X_H Predation on X_PAO

Decay of X_P

Reaction rate equation mPAO*S_O2/(K_O2þS_O2)*S_PO4/(K_P þ S_PO4)*(X_PHA/X_PAO)/(K_PHA þ X_PHA/X_PAO)*X_PAO q_PP*(X_PHA/X_PAO)/(K_PHA þ X_PHA/X_PAO)*(K_max-X_PP/X_PAO)/(K_IPP þ K_max-X_PP/X_PAO)*S_O2/(K_O2 þS_O2) *S_PO4/(K_PS þ S_PO4)*X_PAO q_Gly*(X_PHA/X_PAO)/(K_PHA þ X_PHA/X_PAO)*(K_maxGly-X_Gly/X_PAO)/(K_IGly þ K_maxGly-X_Gly/X_PAO) *S_O2/(K_O2þS_O2)*X_PAO q_PHA*S_A/(K_A þ S_A)*K_O2/(K_O2þS_O2)*(X_PP/X_PAO)/(K_PP þ X_PP/X_PAO)*(X_Gly/X_PAO)/(K_Gly þX_Gly/X_PAO)*X_PAO K_h*S_O2/(K_O2þS_O2)*(X_S/X_H)/(K_X þ X_S/X_H)*X_H K_h*h

fe*K_O2/(K_O2þS_O2)*(X_S/X_H)/(K_X

q_fe*S_F/(K_fe þ S_F)*K_O2/(K_O2þS_O2)*X_H mH *S_O2/(K_O2þS_O2)*S_A/(K_A þ S_A)*S_A/(S_A þ S_F)*S_PO4/(K_P þ S_PO4)*X_H mH *S_O2/(K_O2þS_O2)*S_F/(K_F þ S_F)*S_F/(S_A þ S_F)*S_PO4/(K_P þ S_PO4)*X_H b_H*X_H b_PAO*X_PAO b_PP*X_PP b_PHA*X_PHA b_Gly*X_Gly Ev*X_P/(X_H þ X_PAO þ X_PHA þ X_Gly þ X_PP þ X_P)*X_P Ev*X_H/(X_H þ X_PAO þ X_PHA þ X_Gly þ X_PP þ X_P)*X_H Ev*(X_PAO þ X_Gly þ X_PHA þ X_PP)/(X_H þ X_PAO þ X_PHA þ X_Gly þ X_PP þ X_P)*(X_PAO þ X_PHA þ X_Gly þ X_PP) EP*X_H/(X_H þ X_PAO þ X_PHA þ X_Gly þ X_PP)*S_O2/(kO2P þ S_O2)*X_H EP*(X_PAO þ X_PHA þ X_Gly þ X_PP)/(X_H þ X_PAO þ X_PHA þ X_Gly þ X_PP)*S_O2/(kO2P þ S_O2)*(X_PAO þ X_PHA þX_Gly þ X_PP) b_P*X_P

definition and values of kinetic and stoichiometric coefficients are listed in Table 3.

2.3.

Simulation of the SBR system

The SBR system operated in a completely mixed compartment with a variable volume (2.0e4.0 L). To simulate the effluent, a second completely mixed compartment with a constant volume was introduced (Beun et al., 2001) into AQUASIM (Reichert et al., 1994). The aeration process was simulated by means of programming and Eqs. (1) and (2) (Zhang et al., 2000). if (t_period >¼ 2.1667) and (t_period <¼ 4.8333) then dS0 ðtÞ ¼ 82  ð7:95  SO ðtÞÞ dt

(1)

else dS0 ðtÞ ¼0 dt

þ X_S/X_H)*X_H

(2)

Where: t_period is a formula variable in AQUASIM, which can be expressed as t_period ¼ t mod 6; 2.1667 and 4.8333 indicate

the time when aeration begins and ends; SO (t) is the concentration of oxygen in liquid phase at time t; 82 is the oxygen transfer coefficient applied in our study; 7.95 is the maximum oxygen solubility in liquid phase at 22  C; The schematic diagram of a SBR configuration in AQUASIM is illustrated in Fig. S1.

2.4.

Calibration and validation of the extended model

The extended model was calibrated by the first set of experimental results (Fig. 1, SRT ¼ 12 d and COD ¼ 200 mg/L). To simplify the calibration procedure, the model calibration method is to change as few parameters as possible (Xu and Hultman, 1996). Six kinetic parameters (b_H, b_PAO, b_PHA, b_Gly, b_PP, and q_PP) were calibrated to fit mixed liquid volatile suspended solid (MLVSS), PAOs (XPAO), dead biomass (X_I), volatile fatty acid (VFA: S_A), phosphate (S_PO4) and glycogen contained in PAOs, according to the causality of the parameters to the model outputs and the relative uncertainty in the original parameter values (van veldhuizen et al., 1999; Ni et al., 2008). The selection of b_H, b_PAO, b_PHA, b_Gly and b_PP for calibration was due to the fact that the lumped decay process in the current ASM2d was divided into three individual

5134

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 3 0 e5 1 4 0

Table 3 e Definition and values of kinetic and stoichiometric coefficients involved. Coefficient

Description

Default value

Unit 1

b_H b_PAO b_P Ev

Other factors-induced decay coefficient of X_H Other factors-induced decay coefficient of X_PAO Decay coefficient of X_P Viral infection rate

0.018 0.0075 0.15 0.006

h h1 h1 h1

mH mPAO EP Fxi i_PBM i_PSF i_PSI i_PXI i_PXS f_SI b_PP b_PHA b_Gly K_A K_F K_fe K_h hfe K_IPP K_PP K_PS K_X K_P K_O2 K_max K_Gly

Maximum growth rate of X_H Maximum growth rate of X_PAO Maximum predation rate of X_P Fraction of inert COD generated in biomass lysis P content of biomass P content of S_F P content of S_I P content of X_I P content of X_S Production of S_I in hydrolysis Other factors-induced decay coefficient of X_PP Other factors-induced decay coefficient of X_PHA Other factors-induced decay coefficient of X_Gly Saturation coefficient for growth on S_A Saturation coefficient for growth on S_F Saturation coefficient for fermentation of S_F Hydrolysis rate constant Anaerobic hydrolysis reduction factor Inhibition coefficient for storage of X_PP Saturation coefficient for X_PP Saturation coefficient for phosphorus in storage of X_PP Saturation coefficient for X_S Saturation coefficient for S_PO4 Saturation/inhibition coefficient for oxygen Maximum ratio of X_PP/X_PAO Saturation coefficient for X_Gly

0.275 0.045 0.01 0.2 0.02 0.01 0 0.01 0.01 0 0.0075 0.0075 0.0075 4 4 4 0.1333 0.4 0.02 0.01 0.2 0.1 0.01 0.2 0.34 0.01

h1 h1 h1 g COD/g COD g P/g COD g P/g COD g P/g COD g P/g COD g P/g COD g COD/g COD h1 h1 h1 mg COD/L mg COD/L mg COD/L h1

K_PHA K_IGly

Saturation coefficient for X_PHA Inhibition coefficient for storage of X_Gly

0.01 0.02

g X_PHA/g X_PAO g X_Gly/g X_PAO

K_maxGly kO2P q_fe q_PP q_PHA q_Gly Y_H Y_PAO Y_PHA Y_PO4 Y_Gly Y_P Y_PHApp Y_PHAgly

Maximum ratio of X_Gly/X_PAO Affinity constant for oxygen of predators Maximum rate for fermentation Rate constant for storage of X_PP Rate constant for storage of X_PHA Rate constant for storage of X_Gly Yield coefficient of X_H Yield coefficient of X_PAO PHA produced for S_A uptake PP required for S_A uptake X_Gly consumption for S_A uptake Yield coefficient of X_P PHA required for storage of polyp PHA required for storage of X_Gly

0.4 0.2 0.1375 0.0667 0.4625 0.125 0.625 0.625 1.5 0.3 0.5 0.5 0.2 1.26

g X_Gly/g X_PAO mg O2/L g COD/g COD,h g X_PP/g X_PAO,h g X_PHA/g X_PAO,h g X_Gly/g X_PAO,h g COD/g COD g COD/g COD g COD/g COD g P/g COD g COD/g COD g COD/g COD g COD/g P g COD/g COD

processes in this study. Consequently, the lumped decay coefficients used in the current ASM2d (i.e. b_H, b_PAO, b_PHA, b_Gly and b_PP) needed to be calibrated. The calibration of q_PP was attributed to the fact that q_PP played an important role in the performance of a SBR system enriched PAOs, as also done by Lopez-Vazquez et al. (2009). By calibrating the above 6 kinetic parameters, the model was able to simulate the performance of SBR satisfactorily (Fig. 1). The criterion used for the determination of parameters and the detailed methods used for calculation of 95% confidence intervals are shown in the supplementary material (Smith et al., 1998).

g X_PP/g X_PAO g X_PP/g X_PAO mg P/L g X_S/g X_H mg P/L mg O2/L g X_PP/g X_PAO g X_Gly/g X_PAO

References Henze et al. (1999) Hao et al. (2010a) Moussa et al. (2005) Suttle (1994); Suttle and Chan (1994) Henze et al. (1999) Henze et al. (1999) Moussa et al. (2005) Gujer et al. (1999) Henze et al. (1999) Henze et al. (1999) Henze et al. (1999) Henze et al. (1999) Henze et al. (1999) Henze et al. (1999) Hao et al. (2010a) Hao et al. (2010a) Hao et al. (2010a) Henze et al. (1999) Henze et al. (1999) Henze et al. (1999) Henze et al. (1999) Henze et al. (1999) Henze et al. (1999) Henze et al. (1999) Henze et al. (1999) Henze et al. (1999) Henze et al. (1999) Henze et al. (1999) Henze et al. (1999) Schuler (2005); Whang et al. (2007) Henze et al. (1999) Yagci et al. (2004); Whang et al. (2007) Schuler (2005) Moussa et al. (2005) Henze et al. (1999) Henze et al. (1999) Meijer (2004) Schuler, (2005) Henze et al. (1999) Henze et al. (1999) Mino et al. (1987) Meijer (2004) Meijer (2004) Moussa et al. (2005) Henze et al. (1999) Smolders et al. (1995)

However, it should be noted that the parameters concerning predation and viral infection were not calibrated separately by independent experiment in our study. This is due to the fact that there is not an effective method to calibrate them for the time being, especially for the parameters regarding viral infection. As a result, it is essential to better evaluate these parameters in the future. The validation step was carried out with the calibrated input model parameters by a second set of the experimental results (Fig. 2, SRT ¼ 15 d and COD ¼ 220 mg/L), which were measured at Day 90 after the first set of experimental results were obtained.

5135

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 3 0 e5 1 4 0

A simulated

measured

420

150

350

VFA (mgCOD/L), PO -P (mg/L)

MLVSS, XPAO, X_I (mgCOD/L)

3000

175

125

2500

1500 1000 500

280

VFA(simulated) VFA(measured) PO4-P(simulated) PO4-P(measured) Gly(simulated) Gly(measured)

100

2000

75

210 140

50

70

25 0

0 MLVSS

XPAO

0 0

X_I

Gly(mgCOD/L)

B 3500

1

2

3 Time (h)

4

5

Fig. 1 e Simulated and measured concentrations of (A) mixed liquid volatile suspended solid (MLVSS), polyphosphateaccumulating organisms (XPAO), dead biomass (X_I) and (B) volatile fatty acid (VFA), phosphate, glycogen in the enriched PAOs system (SRT [ 12 d and COD [ 200 mg/L) at the steady state (the error bars indicate the standard deviations).

LIVE/DEAD staining

(APHA, 1995). Glycogen was determined on the basis of Anthrone Colorimetry (Jenkins et al., 1993). VFA was measured using 5-point pH titration method (Moosbrugger et al., 1993). The concentration of PAOs was determined according to Eq. (3) (Hao et al., 2009, 2010b).

The LIVE/DEAD BacLight bacterial viability kit (Type: L-7012, for microscopy and quantitative assays) was used to discriminate between viable cells and dead cells (Invitrogen Molecular Probes, 2004; Ziglio et al., 2002). The detailed procedure was described by Hao et al. (2009).

2.6.

XPAO ¼ MLVSS  LIVE  FISH

Where: XPAO is the concentration of viable PAOs; MLVSS is mixed liquid volatile suspended solid; LIVE is the ratio of viable bacteria to total bacteria (viable þ dead bacteria); FISH is the ratio of viable PAOs to total viable bacteria (Van der Vliet et al., 1994). The concentration of dead biomass was calculated according to Eq. (4).

Fluorescence in-situ hybridization (FISH)

Fluorescence in-situ hybridization (FISH) was used to determine the ratio of PAOs to the total viable bacteria, according to Amann (1995) and Hao et al. (2010b).

2.7.

Analytical methods

X I ¼ MLVSS  DEAD

A

(4)

Where: DEAD is the ratio of dead bacteria to total bacteria (viable þ dead bacteria), i.e., 1LIVE.

MLSS and MLVSS were analyzed according to the standard methods (APHA, 1995). The measurement of soluble orthophosphate (PO3 4 ) was performed spectrophotrometrically

B

4000

210

3500

simulated

measured

3000 2500 2000 1500 1000 500 0 ML VSS

XP AO

X_I

VFA (mgCOD/L), PO -P (mg/L)

MLVSS, XPAO , X_I (mgCOD/L)

(3)

500

180

400

150 VFA(simulated) VFA(measured) PO4-P(simulated) PO4-P(measured) Gly(simulated) Gly(measured)

120 90 60

300 200

Gly (mgCOD/L)

2.5.

100

30 0

0 0

1

2

3 Time (h)

4

5

Fig. 2 e Validation of the calibrated model on (A) mixed liquid volatile suspended solid (MLVSS), polyphosphateaccumulating organisms (XPAO), dead biomass and (B) volatile fatty acid (VFA), phosphate, glycogen (the error bars indicate the standard deviations).

5136

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 3 0 e5 1 4 0

3.

Results and discussion

3.1.

Model calibration

It was therefore used as a default in our model. The sensitivity of the viral infection coefficient on the effluent quality could be evaluated (Fig. 4B) after calibration and validation. With initial selected kinetic and stoichiometric parameters (Table 3), the first simulation was conducted and there was a deviation between the simulation (results not shown) and experimental results. For this reason, 6 kinetic parameters were calibrated (see section 2.4) to match the measured results. The optimal kinetic parameter values, their standard deviations and 95% confidence intervals are summarized in Table 4. With the calibrated kinetic parameters, the simulation results approached to the experimental ones (Fig. 1), which means that the calibration was successful.

The SBR system functioned like a biological phosphate removal system with VFAs uptake associated with phosphate release and glycogen consumption under the anaerobic condition, as well as phosphate uptake and glycogen formation under the aerobic condition (Fig. 1B). Reaching to a steady state was judged by the stable MLSS and MLVSS concentrations, and also by the constant dynamic patterns of VFAs, phosphate and glycogen in each operational cycle. In addition, a well settling sludge was obtained with a high fraction of PAOs in the total amount of bacteria detected by the oligonucleotide probe (approximately 91%). In the extended model, the decay coefficient (0.15 h1) of predators and the predation rate (0.01 h1) were derived directly from Moussa et al. (2005), which were determined by model calibration based on a lab-scale SBR system. The experimental system set up in this study was similar to that of Moussa et al. (2005) (synthetic medium, similar SBR and temperature; albeit a different community: PAOs versus nitrifiers). Also Ni et al. (2010) developed a model structure to simulate predation processes for a lab-scale SBR system. They obtained a different value for calibrated decay coefficient and predation rate (0.071 h1and 0.019 h1, respectively). Their predation rate was however defined on the basis of predator biomass rather than bacterial biomass as used by both Moussa et al. (2005) and our study. The different values on the predation rate inevitably resulted in the different calibrated decay coefficients in the two models. After comparing both sets of decay coefficients and predation rates, however, it can be easily identified that the two sets of parameters are actually in the same order of magnitude if the same biomass basis would be taken. Regarding the viral infection coefficient, there are a few references available. In the only references (Suttle, 1994; Suttle and Chan, 1994; Otawa et al., 2007), the value of about 0.006 h1 was referred on the basis of bacterial biomass was proposed.

3.2.

With the calibrated model and another operational condition (SRT ¼ 15 d and COD ¼ 220 mg/L) of the experiment, a model validation was conducted, as shown in Fig. 2. Compared with Fig. 1, higher MLVSS and glycogen contents could be observed in Fig. 2. Clearly, the calibrated model describes the experimental conditions well under the changed conditions. As a result, the calibrated and validated model could predict the performance of the enriched PAOs system, and thus the contributions of predation and viral infection to sludge production could be also evaluated.

3.3. Sensitivity of the predation rate on sludge minimization The calibrated and validated model was firstly applied to evaluate the contribution of predation to sludge minimization in the enriched PAOs system. With and without predation (the other parameters remained unchanged), the simulation results on the performance of the enriched PAOs system are shown in Table 5. As shown in Table 5, predation contributes to a reduced sludge production. Due to predation, the sludge concentration in the reactor is reduced by 600 mg COD/L. Correspondingly,

B

A 1.4 excess sludge

1.2

PO4-P

1100

discharge standard of PO4-P

1050

0.8 0.6

950

0.4

900

0.2

850

0.004

0.008 -1 Ep (h )

0.012

3000

1

1000

800 0.000

3500

0 0.016

PO4-P (mgP/L)

1150

MLVSS, XPAO, X_H, X_I (mgCOD/L)

1200 Excess sludge (mgCOD/d)

Model validation

2500

MLVSS XPAO X_H X_I

2000 1500 1000 500 0 0.000

0.004

0.008 -1 Ep (h )

0.012

0.016

Fig. 3 e Sensitivity of the predation rate (Ep) on (A) sludge production, effluent-P concentration and (B) mixed liquid volatile suspended solid (MLVSS), polyphosphate-accumulating organisms (XPAO), ordinary heterotrophic organisms (X_H), dead biomass (X_I).

5137

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 3 0 e5 1 4 0

B

A

3200

1.2 excess sludge

1

PO4-P discharge standard of PO4-P

1000

0.8

975

0.6

950

0.4

925

0.2

900 0.000

0.002

0.004 0.006 Ev (h-1)

0.008

0 0.010

PO4-P (mgP/L)

Excee sludge (mgCOD/d)

1025

MLVSS, XPAO, X_H, X_I (mgCOD/L)

1050

2800 MLVSS XPAO X_H X_I

2400 2000 1600 1200 800 400 0 0.000

0.002

0.004 0.006 -1 Ev ( h )

0.008

0.010

Fig. 4 e Sensitivity of the viral infection rate (EV) on (A) sludge production, effluent-P concentration and (B) mixed liquid volatile suspended solid (MLVSS), polyphosphate-accumulating organisms (XPAO), ordinary heterotrophic organisms (X_H), dead biomass (X_I).

the amount of residual sludge is reduced from 1,140 mg COD/ d to 937 mg COD/d due to predation. In other words, predation could contribute to 18% of sludge minimization. With predation, XPAO and X_H decrease considerably, which reflects predation of higher organisms on PAOs and OHOs. However, the fraction of active biomass in the reactor is 12% higher in the case without predation. Clearly, predation could lead to reducing sludge production, but the fraction of active biomass is negatively influenced. It is of interest to evaluate how far extra predation could be stimulated before the process fails. Predation removes bacteria and can be seen as a factor reducing the solid retention time or forcing the predated population to grow faster in order to maintain them in the process. For this purpose, further simulations were conducted to observe the sensitivity of the predation rate on both the sludge production and the effluent-P concentration (Fig. 3). As shown in Fig. 3A, the amount of residual sludge gradually decreases along with increased predation rate (Ep) (the other parameters remained unchanged), but the effluent-P concentration tends to suddenly increase over Ep ¼ 0.012 h1. This can be explained by the fact that the

amount of PAOs decreases due to increased predation and this results in a sudden increase in the phosphate concentration in the effluent when the maximal polyphosphate content in the PAOs is reached. At Ep ¼ 0.014 h1, the effluent phosphate concentration has reached to 0.5 mg P/L, which is actually a threshold keeping the effluent-P quality in the discharge standard (DSPMWTP, 2002). For this reason, the maximal predation rate in the system under the study should be limited below 0.014 h1 and the minimal amount of residual sludge production would be about 900 mg COD/d. In other words, sludge production could be reduced by a maximum at 21% with optimized predation. The simulated effects of the predation rate on the concentrations of MLVSS, XPAO, X_H and X_I are shown in Fig. 3B. There are indeed decreases in the amounts of MLVSS, XPAO and X_H and the amount of X_I increases gradually with increased Ep. This trend is in accordance with the results reported by Priya et al. (2007). There are two limits in an effective biological P-removal system. One limit is the maximal SRT, above which the sludge production is limiting the P-removal capacity as discussed above. This limit is effectively determined by the P/biodegradable-COD ratio in the influent. The other limit is maximal and optimal predation rate, above which PAOs will be washed out from the system. In the studied case, this would occur at a predation rate of

Table 4 e Parameter values of this study with the confidence intervals. Parameters b_H (h1) b_PAO (h1) b_PHA (h1) b_PP (h1) b_Gly (h1) q_PP (g X_PP/g X_PAO,h)

Values Standard Confidence calibrated deviations intervals (%)a 0.008 0.0015 0.0015 0.0015 0.0015 0.1

0.001529 0.000122 0.000339 0.000286 0.000346 0.005925

38 16 45 38 46 12

a 95% confidence intervals are presented in this column as absolute percentage of the parameter estimates, that is, (confidence interval/parameter)  100.

Table 5 e Simulation results with and without predation. Parameters MLVSS (mg COD/L) XPAO (mg COD/L) X_H (mg COD/L) X_I (mg COD/L) X_P þ X_S (mg COD/L) Fraction of active biomass (%) Excess sludge (mg COD/d)

With predation

Without predation

2 812 1351 434 992 35 65 937

3421 2105 524 775 17 77 1140

5138

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 3 0 e5 1 4 0

0.014 h1. Under these conditions, the sludge production would be reduced by 21%.

3.4. Sensitivity of the viral infection rate on sludge minimization As done above for predation, the contribution of viral infection to sludge production could be simulated with and without viral infection involved in the calibrated model (the other parameters remained unchanged), and the simulation results are shown in Table 6. Similar to predation, viral infection is also responsible for sludge production. The amount of residual sludge is reduced by 8% with viral infection involved. This is actually the first time (to our best knowledge) to demonstrate the contribution of viral infection to sludge minimization. Interestingly the amount of OHOs increases due to viral activity. This is opposite to the case of higher organisms’ activity where the relative contribution of OHOs decreases. This phenomenon seems reasonable and explainable. In an enriched PAOs system, the influent substrate (COD/VFAs) is almost all taken by PAOs to form PHA in the cells under the anaerobic condition. For this reason, OHOs can grow only by lysate of PAOs and/or carbon compounds excreted by PAOs (i.e., cryptic growth) in the aerobic phase (van Loosdrecht and Henze, 1999; Lu et al., 2006). When viral infection occurs, death of PAOs increases, which provides more substrate for OHOs. Of course, viral infection will also cause death of OHOs, but the balance between grow and death is obviously positive. Where higher organisms use the bacterial cells as substrate to produce their own biomass, viruses use the bacterial cell as machinery to produce new viruses. After release of the produced viruses in the medium, the dead cell remains and becomes substrate for OHOs. As a result, OHOs tend to increase with viral infection involved. The sensitivity of the viral infection rate on the amount of excess sludge and effluent-P are shown in Fig. 4A. In general, the amount of total sludge production decreases, and the concentration increases with increased viral effluent PO3 4 infection rate (Ev) (the other parameters remained unchanged), which is similar to the pattern for predation (Fig. 3A). At Ev  0.009 h1, the effluent PO34 concentration also reaches to a threshold in the studied system keeping the effluent-P quality (0.5 mg P/L, DSPMWTP, 2002). At this state, the amount of residual sludge is at about 920 mg COD/d, which is equivalent to a decrease of 9% in sludge production. The

Table 6 e Simulation results with and without viral infection involved. Parameters MLVSS (mg COD/L) XPAO (mg COD/L) X_H (mg COD/L) X_I (mg COD/L) X_P þ X_S (mg COD/L) Fraction of active biomass (%) Excess sludge (mg COD/d)

With viral infection

Without viral infection

2812 1351 434 992 35 65 937

3043 1897 258 850 38 72 1014

reduction in biomass production due to viral infection is less because it effectively leads to more growth of heterotrophs in the system. From a sludge reduction point of view, higher organisms’ activity would therefore be more beneficial than viral activity. The simulated correlations between the viral infection rate and MLVSS, XPAO, X_H as well as X_I are shown in Fig. 4B. As explained above, the amount of PAOs decreases and the amount of OHOs increases with increasing viral infection rate. Correspondingly, dead biomass and/or inert particulate increases along with increasing viral infection rate.

4.

Conclusions

ASM2d was extended incorporating predation and viral infection. The model was formulated with three individual processes for decay; and it was effectively calibrated with a set of experimental results and 6 kinetic parameters needing adjustment. It was validated against another set of experimental results. After calibration and validation of the extended model, the sensitivities of predation and viral infection to sludge minimization were respectively evaluated, and some points can be concluded. 1. Simulations reveal that predation is more effective in controlling the sludge production than viral infection. This is mostly related to the fact that dead cells after viral infection will be used for growth of ordinary heterotrophic organisms. 2. The sensitivity analyses indicate that there are the thresholds on both predation and viral infection rates, which are the deviating points on the P-effluent quality. Due to the thresholds, the contributions of predation and viral infection to sludge minimization are limited to the maximal extents of about 2 1% and 9%, respectively. 3. A better understanding of the processes of higher organisms and viral activity is recommended as a tool to minimize sludge production.

Acknowledgments The experimental study was financially supported by the Funding Project for Academic Human Resources Development in Institutions of Higher Learning under the Jurisdiction of Beijing Municipality (PHR20100508), by National Natural Science Foundation of China (50678017) and by KNAW (03CDP008), the Netherlands.

Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.07.013.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 3 0 e5 1 4 0

references Akpor, O.B., Momba, M.N.B., 2010. Relationship of protozoan biomass to phosphate and nitrate removal from activated sludge mixed liquor. Biotechnol. J. 5 (3), 304e313. Amann, R.I., 1995. In situ Identification of Micro-organisms by Whole Cell Hybridization with rRNA-targeted Nucleic Acid Probes. Molecular Microbial Ecology Manual. Kluwer Academic Publishers, Netherland. 1e15. APHA., 1995. Standard Methods for the Examination of Water and Wastewater. American Public Health Association, Washington DC, America. Beun, J.J., Heijnen, J.J., van Loosdrecht, M.C.M., 2001. N-removal in a granular sludge sequencing batch airlift reactor. Biotechnol. Bioeng. 75, 82e92. Cech, J.S., Hartman, P., Macek, M., 1994. Bacteria and protozoa population dynamics in biological phosphate removal systems (1994). Water Sci. Technol. 29 (7), 109e117. Curds, C.R., 1971. A computer-simulation study of predator-prey relationships in a single-stage continuous-culture system. Water Res. 5 (10), 793e812. Discharge Standard of Pollutants for Municipal Wastewater Treatment Plant (DSPMWTP), 2002. Ministry of Environmental Protection of the People’s Republic of China. Ghyoot, W., Verstraete, W., 2000. Reduced sludge production in a two-stage membrane-assisted bioreactor. Water Res. 34 (1), 205e215. Gujer, W., Henze, M., Mino, T., Matsuo, M.C., Wentzel, M.C., Marais, G.V.R., 1995. The activated sludge model No.2: biological phosphorus removal. Water Sci. Technol. 31 (2), 1e11. Gujer, W., Henze, M., Mino, T., van Loosdrecht, M.C.M., 1999. Activated sludge model No. 3. Water Sci. Technol. 39 (1), 183e193. Hao, X.D., Wang, Q.L., Zhang, X.P., Cao, Y.L., van Loosdrecht, M.C. M., 2009. Experimental evaluation of decrease in bacterial activity due to cell death and activity decay in activated sludge. Water Res. 43 (14), 3604e3612. Hao, X.D., Wang, Q.L., Zhu, J.Y., van Loosdrecht, M.C.M., 2010a. Microbiological endogenous processes in biological wastewater treatment systems. Crit. Rev. Environ. Sci. Technol. 40 (3), 239e265. Hao, X.D., Wang, Q.L., Cao, Y.L., van Loosdrecht, M.C.M., 2010b. Experimental evaluation of decrease in the activities of polyphosphate/glycogen-accumulating organisms due to cell death and activity decay in activated sludge. Biotechnol. Bioeng. 106 (3), 399e407. Henze, M., Grady, C.P.L., Gujer, W., Marais, G.V.R., Matsuo, T., 1987. Activated Sludge Model No. 1. IAWPRC, London. Henze, M., Gujer, W., Mino, T., Matsuo, T., Wentzel, M., Marais, G. v.R., van Loosdrecht, M.C.M., 1999. Activated sludge model No. 2d, ASM 2D. Water Sci. Technol. 39 (1), 165e182. Invitrogen Molecular Probes, 2004. LIVE/DEAD BacLight Bacterial Viability Kits. Manuals and Product Inserts. http://probes. invitrogen.com/media/pis/mp07007.pdf. Jenkins, D., Richard, M.G., Daigge, G.T., 1993. Manual on the Causes and Control of Activated Sludge Bulking and Foaming. Lewis Publishers, Chelsea. Kunin, V., He, S., Warnecke, F., Peterson, S.B., Garcia Martin, H., Haynes, M., Ivanova, N., Blackall, L.L., Breitbart, M., Rohwer, F., McMahon, K.D., Hugenholtz, P., 2008. A bacterial metapopulation adapts locally to phage predation despite global dispersal. Genome Res. 18 (2), 293e297. Lee, N.M., Welander, T., 1996. Use of protozoa and Metazoa for decreasing sludge production in aerobic wastewater treatment. Biotechnol. Lett. 18 (4), 429e434. Lopez, C., Pons, M.N., Morgenroth, E., 2006. Endogenous processes during long-term starvation in activated sludge performing

5139

enhanced biological phosphorous removal. Water Res. 40 (8), 1519e1530. Lopez-Vazquez, C.M., Oehmen, A., Hooijmans, C.M., Brdjanovic, D., Gijen, H.J., Yuan, Z., van Loosdrecht, M.C.M., 2009. Modeling the PAO-GAO competition: effects of carbon source, pH and temperature. Water Res. 43 (2), 450e462. Lu, H., Oehmen, A., Virdis, B., Keller, J., Yuan, Z., 2006. Obtaining highly enriched cultures of Candidatus accumulibacter phosphatis through alternating carbon sources. Water Res. 40 (20), 3838e3848. Lu, H., Keller, J., Yuan, Z., 2007. Endogenous metabolism of Candidatus accumulibacter phosphatis under various starvation conditions. Water Res. 41 (20), 4646e4656. Meijer, S.C.F., 2004. Theoretical and Practical Aspects of Modeling Activated Sludge Processes. Delft University of Technology, Delft, the Netherlands. Mino, T., Arun, V., Yoshiaki, T., Matsuo, T., 1987. Effect of phosphorus accumulation on acetate metabolism in the biological phosphorus removal process. In: Ramadori, R. (Ed.), Advances in Water Pollution Control-Biological Phosphate Removal from Wastewaters. Pergamon Press, Oxford, United Kingdom, pp. 27e38. Moosbrugger, R.E., Wentzel, M.C., Ekame, G.A., Marais, G.R., 1993. A 5 pH point titration method for determining the carbonate and SCFA weak acid/bases in anaerobic systems. Water Sci. Technol. 28 (2), 237e245. Moussa, M.S., Hooijmans, C.M., Lubberding, H.J., Gijzen, H.J., van Loosdrecht, M.C.M., 2005. Modeling nitrification, heterotrophic growth and predation in activated sludge. Water Res. 39 (20), 5080e5098. Ni, B.J., Yu, H.Q., Sun, Y.J., 2008. Modeling simultaneous autotrophic and heterotrophic growth in aerobic granules. Water Res. 42, 1583e1594. Ni, B.J., Rittmann, B.E., Yu, H.Q., 2010. Modeling predation processes in activated sludge. Biotechnol. Bioeng. 105 (6), 1021e1030. Otawa, K., Lee, S.H., Yamazoe, A., Onuki, M., Satoh, H., Mino, T., 2007. Abundance, diversity, and dynamics of viruses on microorganisms in activated sludge processes. Microb. Ecol. 53 (1), 143e152. Priya, M., Haridas, A., Manilal, V.B., 2007. Involvement of protozoa in anaerobic wastewater treatment process. Water Res. 41, 4639e4645. Ratsak, C.H., Maarsen, K.A., Kooijman, A.L.M., 1996. Effects of protozoa on carbon mineralization in activated sludge. Water Res. 30 (1), 1e12. Reichert, P., Ruchti, J., Simon, W., 1994. Aquasim 2.0. Swiss Federal. Institute for Environmental Science and Technology (EAWAG), Duebendorf, Switzerland. Rieger, L., Koch, G., Kuhni, M., Gujer, W., Siegrist, H., 2001. The EAWAG Bio-P module for activated sludge model NO. 3. Water Res. 35 (16), 3887e3903. Schuler, A.J., 2005. Diversity matters: dynamic simulation of distributed bacterial states in suspended growth biological wastewater treatment systems. Biotechnol. Bioeng. 91, 62e74. Schuler, A.J., Jassby, D., 2007. Distributed state simulation of endogenous processes in biological wastewater treatment. Biotechnol. Bioeng. 97 (5), 1087e1097. Smith, L.H., Mccarty, P.L., Kitanidis, P.K., 1998. Spreadsheet method for evaluation of biochemical reaction rate coefficients and their uncertainties by weighted nonlinear least-squares analysis of the integrated monod equation. Appl. Environ. Microbiol. 64 (6), 2044e2050. Smolders, G.J.F., van der Meij, J., van Loosdrecht, M.C.M., Heijnen, J.J., 1995. A structured metabolic model for anaerobic and aerobic stoichiometry and kinetics of the biological phosphorus removal process. Biotechnol. Bioeng. 47, 277e287.

5140

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 3 0 e5 1 4 0

Suttle, C.A., 1994. The significance of viruses to mortality in aquatic microbial communities. Microb. Ecol. 28, 237e243. Suttle, C.A., Chan, A.M., 1994. Dynamics and distribution of cyanophages and their effect on marine Synechococcus spp. Appl. Environ. Microbiol. 60, 3167e3174. Van der Vliet, G.M.E., Schepers, P., Schukkink, R.A.F., Van gemen, B., Klatser, P.R., 1994. Assessment of mycobacterial viability by RNA amplification. Antimicrob. Agents Chemother. 38 (9), 1959e1965. van Loosdrecht, M.C.M., Henze, M., 1999. Maintenance, endogenous respiration, lysis, decay and predation. Water Sci. Technol. 39 (1), 107e117. van veldhuizen., H.M., van Loosdrecht, M.C.M., Heijnen, J.J., 1999. Modelling biological phosphorus and nitrogen removal in a full scale activated sludge process. Water Res. 33 (16), 3459e3468. Whang, L.M., Filipe, C.D.M., Park, J.K., 2007. Model-based evaluation of competition between polyphosphate- and

glycogen-accumulating organisms. Water Res. 41 (6), 1312e1324. Withey, S., Cartmell, E., Avery, L.M., Stephenson, T., 2005. Bacteriophages e potential for application in wastewater treatment processes. Sci. Total Environ. 339 (1e3), 1e18. Xu, S.L., Hultman, B., 1996. Experiences in wastewater characterization and model calibration for the activated sludge process. Water Sci. Technol. 33, 89e98. Yagci, N., Insel, G., Artan, N., Orhon, D., 2004. Modelling and calibration of phosphate and glycogen accumulating organism competition for acetate uptake in a sequencing batch reactor. Water Sci. Technol. 50 (6), 41e50. Zhang, Z.J., Lin, R.C., Jin, R.L., 2000. Sewage Engineering, fourth ed. China Architecture and Building Press, Beijing, China. Ziglio, G., Andreottola, G., Barbesti, S., Boschetti, G., Bruni, L., Foladoria, P., Villa, R., 2002. Assessment of activated sludge viability with flow cytometry. Water Res. 36 (2), 460e468.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 4 1 e5 1 5 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Identification of cocaine and its metabolites in urban wastewater and comparison with the human excretion profile in urine Sara Castiglioni a,*, Renzo Bagnati a, Manuela Melis a, Deepika Panawennage b, Paul Chiarelli b, Roberto Fanelli a, Ettore Zuccato a a b

Department of Environmental Health Sciences, Mario Negri Institute for Pharmacological Research, via La Masa 19, 20156 Milan, Italy Department of Chemistry, Loyola University, 1032 W. Sheridan Rd., Chicago, IL 60660, USA

article info

abstract

Article history:

The most relevant human urinary metabolites of cocaine (nine metabolites) were

Received 10 February 2011

measured in urban wastewater in Italy and USA. A novel analytical method based on liquid

Received in revised form

chromatography tandem mass spectrometry allowed the identification of ecgonine,

5 July 2011

ecgonine methyl ester and the pyrolytic derivatives of cocaine in untreated wastewater.

Accepted 13 July 2011

The aim of this study was to verify whether the pattern of cocaine metabolites in waste-

Available online 23 July 2011

water reflected the human excretion profile in urine. The performance of the method was good, with recoveries higher than 60% and limits of quantifications in the low ng/L range.

Keywords:

The stability in untreated wastewater was assessed for all metabolites and the best storage

Urban wastewater

condition resulted freezing samples immediately after collection and keep them frozen

Cocaine

until analysis. All the selected compounds were measured in wastewater at concentrations

Mass spectrometry

up to 1.5 mg/L and their weekly loads were calculated during a five weeks monitoring

Human excretion profile

campaign in Milan (Italy). The profiles of cocaine metabolites in wastewater matched with

Sewage epidemiology

those in human urine reported in the literature, suggesting that measures in wastewater reflect the real human excretion and that wastewater analysis is suitable for assessing drug consumption. Benzoylecgonine was confirmed as the best target for estimating cocaine use by wastewater analysis, while cocaine itself should not be considered because its amount in wastewater is affected by other environmental sources such as transport, handling and consumption. Results suggested that the measurement of other metabolites in combination with benzoylecgonine might reflect 60% of an administered dose of cocaine providing also information on different patterns of use. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

In the last few years illicit drugs have been “discovered” as a novel class of environmental contaminants, and have attracted interest in different scientific disciplines such as analytical and environmental chemistry (Richardson, 2009; Zuccato and Castiglioni, 2009) and social sciences (EMCDDA,

2008; Frost et al., 2008). The main source of environmental contamination by illicit drugs is human consumption. In fact, after the ingestion of a “drug dose”, active parent compounds (unchanged parent drugs) or metabolites are excreted in consumers’ urine, entering urban wastewater and reaching sewage treatment plants (STPs). These substances can persist even in treated wastewater, and make their way into surface

* Corresponding author. Tel.: þ39 02 39014776; fax: þ39 02 39014735. E-mail address: [email protected] (S. Castiglioni). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.017

5142

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 4 1 e5 1 5 0

water (rivers, lakes, sea) or drinking water (Chiaia et al., 2008; Huerta-Fontela et al., 2008; Jones-Lepp et al., 2004; Postigo et al., 2008; Zuccato and Castiglioni, 2009). The first multiresidue analytical method to detect illicit drugs in urban untreated wastewater was set up by our group in 2006 (Castiglioni et al., 2006), and was employed to measure the most widely used classes of illicit drugs, i.e. cocaine and its main metabolites, amphetamines, opioids, and cannabinoids, in Italy and Switzerland. The method was based on solid phase extraction (SPE) and liquid chromatography tandem mass spectrometry (LC-MS-MS), which became the most common techniques to detect drugs of abuse in environmental samples (Castiglioni et al., 2008; Postigo et al., 2008). An interesting application of the measurement of illicit drugs in wastewater is “sewage epidemiology”, a novel approach for estimating drug consumption in a community by wastewater analysis (Daughton, 2011; Zuccato et al., 2005, 2008). This method is based on the principle that if drug residues are excreted by humans in substantial amounts and are sufficiently stable in wastewater, they can be easily detected in urban untreated wastewater, reflecting the amounts consumed by the population. Wastewater analysis has been used to estimate drug consumption in several case studies (van Nuijs et al., 2010), and drug consumption profiles obtained from wastewater analysis resulted similar to the national consumption profiles based on annual prevalence data. The correspondence between the pattern of drug metabolites detected in wastewater and their human excretion profile is one of the main assumptions on which the present method is based (Zuccato et al., 2008), but this correspondence has only been assessed for a few compounds so far. Moreover, a proper storage of samples is mandatory in order to avoid the degradation of the analytes that should be used for drug consumption estimation. The aim of this study was to verify whether the pattern of cocaine metabolites detected in wastewater reflects the human excretion profile in urine. We chose cocaine because this substance has a well known excretion profile in urine (Baselt, 2004) with specific metabolites for the different routes of administration. The complete set of cocaine metabolites normally detected in consumer’s urine was measured in untreated wastewater extending and revising our previous publications (Castiglioni et al., 2008). In particular, ecgonine is hard to extract from biological fluids such as plasma (Klingmann et al., 2001), urine or whole blood (Paul et al., 2005), because of its high polarity and water-solubility, and required a specific procedure for extraction from wastewater too. The novel method we have developed within this study allows the simultaneous analysis of ecgonine and other structurally-related metabolites of cocaine, i.e. ecgonine methyl ester, and the main pyrolytic products of cocaine (anhydroecgonine and anhydroecgonine methyl ester), in untreated wastewater. Hydrophilic interaction chromatography (HILIC) has been employed for these substances since it is a powerful technique for the retention of polar analytes offering a difference in selectivity compared to traditional reversed-phase chromatography (Hsieh, 2008). A similar method based on SPE and direct injection liquid chromatography was developed contemporaneously to measure these compounds in

wastewater (Bisceglia et al., 2010), but a lower method performance for ecgonine has been obtained. The selected compounds were measured in several STPs in Italy and USA, allowing the comparison of their levels in wastewater with those measured in urine (Paul et al., 2005) and with the human excretion profiles of cocaine reported in the literature. Measuring cocaine metabolites in urban untreated wastewater was also useful to identify the different routes of administration of cocaine and to distinguish the amounts of cocaine from human consumption and from other sources such as production, transport, and distribution. Finally, the best storage conditions for wastewater samples has been provided within this study through a stability test which assessed the degradation of each compound in untreated wastewater under different storage conditions.

2.

Materials and methods

2.1.

Selection of compounds

Cocaine is one of the most widely used illicit drugs worldwide, with between 16 and 21 million of consumers, i.e. 8% of the total drug users in the population aged 15 to 64 (UNODC, 2009) and has a wide range of human metabolites. Within this study the compounds selected were: cocaine (COC), benzoylecgonine (BZE), norbenzoylecgonine (NBE), norcocaine (NCO), cocaethylene (CET), ecgonine methyl ester (EME), ecgonine (ECG), anhydroecgonine (AEC), anhydroecgonine methyl ester (AME). BZE and EME are the primary human metabolites of COC and urinary excretion amounts to respectively 45% and 40% (mean percentages) of the dose consumed (Baselt, 2004; Ambre et al., 1984; Ambre et al., 1988). ECG is excreted in smaller amounts in urine, 1e8% of the dose consumed (Fish and Wilson, 1969; Smith et al., 2010), but can also be produced by the spontaneous hydrolysis of EME in urine (Baselt, 2004). COC is only partially excreted in the urine as unchanged drug, i.e. 1e9% of the administered dose, depending on the pH of the urine (Baselt, 2004). Other minor metabolites that form during the oxidative metabolism of COC and BZE are NCO, a pharmacologically active metabolite, and NBE (Maurer et al., 2006). In the presence of ethanol, COC can be trans-esterified, forming CET, which accounts for 0.7% of a dose in 24-h urine (Baselt, 2004). When COC is smoked as crack, i.e. the free base form of COC, the main pyrolytic product is AME, which is inhaled and form three metabolites in human body: AEC, anhydroecgonine ethyl ester, and noranhydroecgonine (Maurer et al., 2006). Our investigation included the main pyrolytic derivative of COC, AME, and one of its metabolites, AEC, that were detected in human urine (Paul et al., 2005).

2.2.

Chemicals and materials

The analytical standards of COC, BZE, NBE, NCO, CET, EME, ECG, AEC, AME and the deuterated analogues BZE-d3, COC-d3, CET-d8, ECG-d3, and EME-d3 were acquired from Cerilliant Corporation (Round Rock, Texas, USA). The standards, available as solutions in methanol or acetonitrile (1 or 0.1 mg/mL),

5143

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 4 1 e5 1 5 0

were diluted to 10 ng/mL with methanol or acetonitrile, and stored at 20  C in the dark. Working solutions (1 and 0.1 ng/ mL) containing all the substances to be analyzed were prepared before each analytical run and stored at 20  C in the dark. The deuterated compounds were used as internal standards (IS), and separate working solutions (1 ng/mL) were prepared before analysis. Methanol for pesticide analysis, acetonitrile for HPLC-MS and hydrochloric acid (37%) were from Carlo Erba (Italy); ammonium hydroxide solution (25%), acetic acid for LC-MS (>99%), ammonium formate (>99% dry matter) and formic acid (98%) were acquired from Fluka (Buchs, Switzerland). HPLC grade MilliQ water was obtained with a MILLI-RO PLUS 90 apparatus (Millipore, Molsheim, France). The cartridges used for solid phase extraction were 3-mL disposable Oasis MCX (60 mg), and 6-mL disposable Oasis MCX (150 mg) from Waters Corp., Milford, MA. The analytical HPLC columns used were an XTerra MS C18, 100  2.1 mm, 3.5 mm and an X-Bridge HILIC 100  2.1 mm, 3.5 mm from Waters Corp., Milford, MA.

2.3.

Wastewater samples

Composite 24-h water samples, obtained by pooling water collected by automatic sampling devices and taking 500 mL from each mixed sample, were collected after the primary entrance grids from the influents in different STPs in Italy (Milan, Como and four towns in Sardinia) and USA (Table 1). Sampling locations were chosen to investigate areas with different socio-economic patterns. The plants in Italy were located in the biggest business city in the north of Italy (Milan), with a high incoming flow of people for work and amusement, in a smaller but industrialized city in the same area (Como) and in an island (Sardinia). In Sardinia two cities were chosen on the cost (Cagliari and Olbia) being characterized by a high touristic influx, and two were chosen in the inland (Sassari and Nuoro). The STP in USA is the Stickney Water Reclamation Plant which processes wastewater from the south and west parts of Chicago. For each plant, composite samples were

collected every day for seven consecutive days. The sampling was repeated for five weeks in Milan, four weeks in Como, two weeks in Sardinia and one week in Chicago. All the samples were collected in dry weather (no rainfall) during MarchSeptember 2009 (Italy) and November-December 2009 (USA). Water samples were frozen and stored in the dark at 20  C. Before extraction, samples were filtered first on a glass microfiber filter GF/A 1.6 mm (Whatman, Kent, U.K.) then on a mixed cellulose membrane filter 0.45 mm (Whatman, Kent, U.K.).

2.4.

Solid phase extraction (SPE)

The procedures for the extraction and analysis of the selected compounds required their separation into two main analytical groups (Table 2). The first analytical group was extracted adapting the procedure described in a previous publication (Castiglioni et al., 2006) to an automated SPE system, GX-274 ASPEC (Gilson, Middleton, WI, USA). This method was not applicable to the second analytical group, especially to ECG, due to the low molecular weight and high water-solubility of this substance (Castiglioni et al., 2008). ECG extraction was previously performed from biological fluids and wastewater using cation exchange cartridges (Bisceglia et al., 2010; Klingmann et al., 2001; Paul et al., 2005) or Hysphere MM anion sorbent (Jagerdeo et al., 2008), but the recoveries were lower than 40%. The extraction procedure from untreated wastewater required an optimization by testing different cartridges and pH conditions: Oasis MCX (60 mg) at pH 2 and 6.0, Oasis WCX (60 mg) at pH 4.0 and 8.0, Oasis MAX (60 mg) at pH 8.0, Oasis HLB at pH 7.0, Evolute ABN at pH 7.0. The recoveries of ECG were poor in all the cases and the spiked amounts were found in the wastewater passed through the cartridge, indicating that ECG was not retained on the solid phases because of its high water-solubility. The extraction of the other compounds belonging to the second analytical group was best using an Oasis MCX cartridge (60 mg) at pH 2.0, so cartridges packed

Table 1 e Main characteristics of the wastewater treatment plants (STPs) investigated. STPs investigated

Sampling mode

Milan (Nosedo)

Volume proportional

Como

Time proportional every 20 min Time proportional every 15 min

Mean daily flow rate (m3/d)

Population served by the plant

Type of sewage

370,000

1,250,000

63,000

100,000

domestic þ industrial

Cagliari

96,000

300,000

domestic

Olbia

20,000

50,000

Sassari

36,000

130,000

7000

21,000

1,925,000

2,400,000

Sardinia

Nuoro Chicago (Stickney)

Volume Proportional

domestic

domestic þ industrial domestic þ industrial domestic domestic þ industrial

Wastewater treatment activated sludge þ disinfection (peracetic acid) activated sludge

activated sludge (hypochlorite) activated sludge (hypochlorite) activated sludge (hypochlorite) activated sludge (hypochlorite) activated sludge

þ disinfection þ disinfection þ disinfection þ disinfection þ disinfection

5144

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 4 1 e5 1 5 0

Table 2 e Recoveries of COC and its main metabolites (mean ± standard deviation-SD), instrumental quantification limits (IQL), and limits of quantification (LOQ) of the analytical method in untreated wastewater. Chemicals

Recovery  SD (%)

First analytical group BZE 107 NBE 85 COC 96 NCO 112 CET 109 Second analytical group ECG 63 EME 101 AEC 98 AME 62

IQL (pg/injected)

LOQ (ng/L)

    

9 5 5 7 4

22 7.8 23.7 19.8 16.6

2.1 0.6 2.1 2.5 1.3

   

1 8 2 4

60.1 15.5 11.2 26.5

7.2 2.6 2.2 7.5

with more sorbent (150 mg) were tested to increase the adsorption surface. In this way the extraction efficiency of ECG was highly improved. The volume of extraction was optimized for the retention of ECG on Oasis MCX (150 mg) by testing different volumes of untreated wastewater, i.e. 10, 20, 30, 40, and 50 mL. The best result was obtained with an Oasis MCX cartridge (150 mg, 6 cc) at pH 2.0 and extracting 20 mL of untreated wastewater. Briefly, untreated wastewaters (20 mL) were spiked with 20 ng of each IS and the pH was adjusted to 2.0 with 37% HCl. The cartridges were conditioned by washing with 12 mL methanol and 6 mL MilliQ water, and samples were passed through by gravity. Cartridges were then vacuum-dried for 10 min and eluted with 2 mL methanol and 2 mL 2% ammonia solution in methanol. The eluates were pooled and dried to 10 mL under a nitrogen stream.

2.5. Liquid chromatography (LC) - mass spectrometric analysis (MS-MS) The LC system consisted of two Series 200 pumps and a Series 200 auto sampler (PerkineElmer, Norwalk, CT). The first analytical group (Fig. S-1) was analyzed according to the method described previously (Castiglioni et al., 2006, 2008). The second group that includes small, extremely hydrophilic molecules hardly retained by reversed-phase chromatography using acidic buffers, was previously analyzed using two different chromatographic methods (Castiglioni et al., 2008): an XTerra C18 column with an ammonium formate buffer for EME and AME, and a PFP Propyl column with an acetic acid buffer for ECG and AEC. Despite good results for analytical standards, the chromatographic separation and resolution in wastewater samples were very poor and the entire procedure was time-consuming. Therefore, within this study we set out to find better and faster conditions for the simultaneous analysis of these compounds. A Luna CN column was initially tested, but gave the same results as before, and low reproducibility on successive analytical runs. An X-Bridge HILIC column was then tested, and gave the best results, allowing simultaneous analysis of all these compounds and giving good chromatographic resolution both for standards and wastewater samples. The Hydrophilic Interaction Liquid

Chromatography (HILIC) technique works similarly to a normal phase chromatography, i.e. the analytes are retained when gradients are applied with initial high percentage of organic phase and elution is done with a slightly acidic water solvent (no more than 50% of water solvent must be applied to the column). Under these conditions small polar compounds, like those considered here, can be retained by the stationary phase, allowing good chromatographic separation. The eluates concentrated to 10 mL were diluted with 90 mL acetonitrile, then centrifuged and transferred into glass vials. A gradient elution was used for chromatographic separation, using ammonium formate 5 mM in water acidified to pH 4 with formic acid as solvent A, and acetonitrile as solvent B at a flow rate of 300 mL/min. The elution started with 5% of eluent A and 95% of eluent B, followed by a 12-min linear gradient to 50% of eluent A, a 3-min isocratic elution and a 1-min linear gradient to 5% of eluent A, which was maintained for 7 min to equilibrate the column. The injection volume was 10 mL. Typical chromatograms obtained from the analysis of untreated wastewater are presented in Fig. 1. The mass spectrometric analysis was performed using an API 3000 triple quadrupole, equipped with a turbo ion spray source (AB - Sciex, Thornhill, Ontario, Canada). The analyses were done in positive ion mode for all the compounds and the instrumental conditions optimized for each compound are summarized in Table S-1. A detailed description of mass spectrometric analysis, including the collision-induced dissociation (CID) spectra of all the drugs, quantification procedures, and matrix effects control are reported elsewhere (Castiglioni et al., 2006, 2008).

2.6.

Performance of the method

The method performance was assessed in untreated wastewater for all the compounds. Recoveries and repeatability of the method were assessed by analyzing untreated wastewater samples in triplicate. Since relevant amounts of illicit drugs were already present in wastewater, the samples were spiked with 5 mg/L of BE, 2 mg/L of COC, ECG, EME, AEC, and AME, and 0.5 mg/L of the other drugs before extraction. Blank samples (mineral water) were analyzed within each analytical run to test for and correct bias. Instrumental quantification limits (IQL) were determined by direct injection of standard solutions with increasing amounts of each substance. The limits of quantification (LOQ) for the whole method were calculated directly from extracted samples of STP influents. LOQs were calculated as the concentrations giving peaks with a signal-tonoise ratio of 10 for the first group of analytes and 5 for the second group (Table 2). The linearity of the calibration curves was tested in the ranges normally measured in wastewater and a calibration curve was injected during each analytical run to check the linearity (correlation factors) and the instrumental repeatability. Intra- and interday instrumental repeatability and precision were also assessed by replicated injections of standard mixtures and wastewater samples.

2.7.

Stability in wastewater

The stability of the selected compounds in untreated wastewater was investigated to establish the best storage

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 4 1 e5 1 5 0

5145

Fig. 1 e HILIC-LC-MRM chromatograms of ecgonine, ecgonine methyl ester, anhydroecgonine and anhydroecgonine methyl ester in untreated wastewater. The chromatogram of anhydroecgonine methyl ester refers to a spiked sample (2 mg/L), the other chromatograms to real samples (80, 150 and 10 ng/L respectively for ecgonine, ecgonine methyl ester and anhydroecgonine).

conditions for wastewater samples before and during the analyses. Stability tests were run in triplicate on samples stored in glass bottles in the dark for one and three days, and on other samples processed with a freeze-and-thaw (F/T) cycle (performed overnight). No pH adjustment was performed during the stability tests to reproduce the real storage conditions. Untreated wastewater samples were spiked with 5 mg/L of BZE, 2 mg/L of COC, and 1 mg/L of the other drugs. Three aliquots of wastewater were stored for three days at 4  C, and analyzed after one and three days. Three identical aliquots were prepared separately and were subjected immediately to an F/T cycle. An additional sample without the drug spiking was used as control for each set of samples and was processed the same way. The analyses were carried out immediately after spiking (T0), after one and three days of incubation at 4  C (T1 and T3), and after the F/T cycle (F/T). The residual amounts of the substances were calculated by subtracting the amounts already present in the control sample.

3.

Results and discussion

3.1.

Performance of the method

The recoveries obtained in untreated wastewater were higher than 60% for all the compounds including ECG, with less than 9% variability (Table 2). To our knowledge, this is the best result obtained for extracting ECG from biological samples

(Klingmann et al., 2001; Paul et al., 2005) and wastewater (Bisceglia et al., 2010). The instrumental sensitivity was good and the IQLs ranged from 8 to 60 pg/injected. LOQs in untreated wastewater ranged between 0.6 and 2.8 ng/L, except for ECG (7.2 ng/L) and AME (7.5 ng/L) (Table 2). The analytical response was linear for all the compounds in the range of concentrations measured in wastewater and the interday correlation factors (r2) were higher than 0.9996 with standard deviations (SD) < 0.0003 (Table S-2). Instrumental repeatability was assessed using replicate injections of standard mixtures and wastewater and was generally lower than 10% (Table S-2).

3.2.

Stability in wastewater

The stability of COC and its metabolites in untreated wastewater was investigated at different storage conditions and the results are shown in Fig. 2. The first stability test involved the analysis of triplicate samples after an F/T cycle. The residual amounts were compared with those measured immediately after the drug spiking (T0). All the compounds were stable during this treatment, except for EME and COC which showed a slight decrease (about 22% less than the initial concentration). The second stability test involved the analysis of triplicate aliquots of the spiked samples after one (T1) and three days (T3) of storage at 4  C in the dark (Fig. 2). Again, the residual amounts were compared with those measured in the samples immediately after spiking (T0). COC, NCO and EME declined

5146

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 4 1 e5 1 5 0

Fig. 2 e Stability of the compounds in wastewater immediately after drug spiking (T0), after a freeze-and-thaw cycle (F/T), and after one day (T1) and three days (T3) of storage at 4  C.

substantially after one (39, 8 and 44% of the initial concentrations) and three days (81, 65 and 76%), and correspondent increases of BZE, NBE and ECG concentrations were observed after three days (þ21, þ5 and þ26% of the initial concentrations). The reactions probably occurring in untreated wastewater were therefore the demethylation of COC and EME to BE and ECG and the N-demethylation of COC and BE to NorCOC and NorBE. Further experiments are required to ascertain the actual transformation processes occurring during stability tests. After three days’ storage at 4  C the CET concentration dropped 38% from the initial concentration, while AEC and AME were quite stable. These results are in line with the stability tests conducted by another group for COC and BZE in wastewater at different storage conditions (Gonzalez-Marino et al., 2010). Our group also observed decreases in COC, NCO and CET concentrations in a previous investigation (Castiglioni et al., 2006), but although the experimental conditions were the same, they were lower than in this study. Since the only difference was the wastewater used, we assume that the nature and composition of the water deeply influences the biological degradation of these substances, and their stability must therefore be carefully checked before samples are stored. The stability of all the compounds we tested was good after an F/T cycle, but was low after one or three days at 4  C (Fig. 2). We decided therefore to freeze wastewater samples immediately after collection and keep them frozen (20  C) until analysis. Samples were processed immediately after thawing, avoiding storage at 4  C.

3.3. Occurrence of cocaine and metabolites in untreated wastewater COC and its main metabolites were measured in untreated wastewater in five medium-sized STPs in Como (Italy), Cagliari, Olbia, Sassari and Nuoro (Sardinia, Italy), and in two

large plants in Milan (Italy) and Chicago (USA) (Table 3). ECG and the pyrolytic derivatives of COC (AEC and AME) were measured in wastewater for the first time within this study. All the compounds were detected in wastewater, except ECG and AEC in Sardinia and AME. The total concentrations of COC and its metabolites (sum of the single concentrations) in untreated wastewater were 1.3, 0.64, 0.55, 0.61,0.26, 0.28 and 3.1 mg/L respectively in Milan, Como, Cagliari, Olbia, Sassari, Nuoro and Chicago. The most abundant compound was BZE, which was measured at concentrations up to 1.5 mg/L in Chicago. COC, EME and ECG reached concentrations in the hundreds of ng/L range, and the other metabolites were at lower concentrations. Similar levels of BZE and COC were reported in untreated wastewater from other European countries, such as Spain, Belgium, UK, Croatia, The Netherlands, and Germany (Hogenboom et al., 2009; Postigo et al., 2008; Terzic et al., 2010; Zuccato and Castiglioni, 2009). COC, BZE, NCO and NBE were also measured in seven STPs in USA (Chiaia et al., 2008), and their concentrations were in the same range as those in Chicago in this study. The ratio of COC to BZE was about 0.3 in Milan, Como and Sardinia and 0.55 in Chicago, and was higher than in a previous analytical campaign in Italy (0.15) (Zuccato et al., 2005), but similar to those found in several STPs in Belgium (0.4) (van Nuijs et al., 2009). This is higher than the ratio expected from the percentages of excretion of these substances in urine reported in the literature (1e9% of COC and 35e54% of BZE), which averages 0.1. This suggests a role for other sources of COC probably related to transport, handling and consumption of the drug. The main pyrolytic metabolite of COC, AME, which is mainly metabolized in the human body to AEC and other metabolites (Maurer et al., 2006), was never found in wastewater, but AEC was detected at measurable concentrations in Milan (35 samples), Como (28 samples) and Chicago (7 samples) (Table 3). The ratio of AEC (a metabolite of COC when

5147

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 4 1 e5 1 5 0

Table 3 e Mean levels (ng/L) of COC and its main metabolites and Relative Standard Deviation (RSD%) in untreated wastewater in Italy and USA detected during several successive campaigns carried out in 2009. Chemicals

Served people

Milan (five weeks monitoring)

Como (four weeks monitoring)

Sardinia (two weeks monitoring)

Chicago (one week monitoring)

Cagliari

Olbia

Sassari

Nuoro

1,250,000

101,000

300,000

50,000

130,000

21,000

2,400,000

BZE NBE COC NCO CET

712 (8) 20 (24) 255 (15) 4.0 (38) 6.4 (16)

380 (10) 12 (25) 98 (38) 3.2(12) 2.7(29)

316 (2.5) 11 (8) 138 (37) 3.5 (33) 3.0 (9)

337 (14) 11 (37) 131 (5) 3.3 (29) 5.1 (33)

149 (11) 5.5 (11) 48 (13) 0.9 (7) 0.4 (10)

153(8) 5.3 (6) 55 (16) 2.0 (24) 3.2 (14)

1553 (22) 51 (20) 868 (21) 28 (23) 23 (40)

ECG EME AEC AME

97 (8) 176 (11) 4.3 (29)
58 (35) 84 (29) 1.0 (23)




199 (23) 346 (23) 35 (27)
Total amount measured

1276 (18)

638 (20)

548 (21)

607 (20)

257 (20)

285 (21)

3103 (17)

smoked as crack) to BZE (the major metabolite of COC) was three times higher in Chicago than in Milan. This can suggests that in the time period under analysis, the percentage of COC used as crack was higher in Chicago than in Milan, in line with epidemiological investigations indicating a greater prevalence

of crack use in USA than in Italy (EMCDDA, 2009). Even if further research is needed to confirm these results, the AEC/ BE ratio seem to be useful as a rough estimate of the amount of COC smoked as crack in a community, and can therefore help to estimate how much COC is used as crack.

3.4. Weekly loads of cocaine and metabolites in untreated wastewater in Milan

Fig. 3 e A. Daily loads (g/day) of COC and metabolites in untreated wastewater in Milan. B. Comparison between the daily loads of COC, and BZE and the sum of all metabolites measured in wastewater. The loads are mean values (with standard deviation) of a five week monitoring campaign.

The daily loads (g/day) of COC and metabolites have been calculated in a five weeks monitoring campaign conducted in Milan, multiplying the concentrations of each compound in untreated wastewater by the mean daily flow rates (Fig. 3A). Generally, the loads increased during the weekend reaching a peak on Saturday, when BZE, EME, ECG, and NBE loads were 30% higher than in a weekday. The load of BZE was the highest, ranging from 250 g/day during the week to 350 g/day on Saturday. Lower loads were measured for EME and ECG, and were respectively 60 and 35 g/day during the week and 90 and 45 g/day on Saturday. Among the minor metabolites of COC, NBE was the most abundant (7e10 g/day) and CET showed a marked increase during the weekend (from 2.5 to 3.7 g/day). The daily loads of NCO and AEC were about 2 g/day without any relevant change during the weekend. Fig. 3B shows the weekly loads of COC compared to those of BZE and of the sum of all the measured metabolites (including BE). The daily load of COC ranged from 90 g/day in a weekday to 110 g/day on Saturday, thus following the same weekly pattern of its main metabolites. Despite this suggests a relationship with human urinary excretion, the total amount of COC found in wastewater is higher than that expected from human metabolism, as discussed in the previous paragraph, and it is therefore unsuitable to indicate human consumption. The loads of all metabolites were 30% higher than those of BZE alone (Fig. 3B), indicating that the measurement of additional COC metabolites in combination with BE could increase the percentage of an administered dose detected in wastewater. If BE alone represents about 45% of a dose of COC, all

5148

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 4 1 e5 1 5 0

metabolites can reflect 59e60%. Moreover, the analysis of different COC metabolites in wastewater might also help investigating the patterns of consumption. For instance, measuring CET might help to assess how frequently COC is consumed together with alcohol as observed for its weekly pattern of consumption in Milan (Fig. 3A).

3.5. Comparison of metabolic profiles in wastewater and in human urine The profile of COC metabolites in wastewater was compared with the profiles in human urine reported in the literature. The urinary profiles considered came either from pharmacokinetic studies conducted under controlled doseadministration conditions, and from spot urine analysis (Table 4). Table 4 (first column) reports the amount of each metabolite, as a percentage of the administered dose of COC, recovered in the 24e72 h urine in pharmacokinetic studies. The main limitation of these data is the relatively low number of subjects involved (2e6 subjects each study). Thus we decided to include in our comparison supplementary data from spot urine analysis (Table 4, second column). Each metabolite is reported as a percentage of the sum of the median levels of each metabolites plus COC measured in spot urine specimens. These percentages were calculated from data published recently by Paul et al., who measured 15 COCrelated compounds in 30 urine samples from living individuals and post-mortem specimens. The excretion profiles of COC metabolites and COC were calculated using median values because of the high variability of the concentrations

measured in spot urine samples (Paul et al., 2005). Similarly, the third column reports each metabolite as a percentage of the sum of the median concentrations of metabolites plus COC measured in wastewater. These percentages were calculated using the concentrations measured in Milan and Como (Italy), and Chicago (USA) because a similar pattern of COC consumption has been found in these cities. Due to the variable excretion ranges observed for some compounds in pharmacokinetics studies (i.e. cocaine, range 1e14%), only a general comparison among the excretion profiles was possible according to the aims of our study. Data from wastewater could be reasonably compared with profiles in urine (Table 4), since they indicate the profiles of excretion of the different metabolites of COC in the urine of single individuals (pharmacokinetic studies), and in wastewater, which collects the urine excreted by an entire community. BZE was confirmed as the most abundant metabolic product of COC, with excretion approximating 50% in all cases. Percentages were comparable also for other minor metabolites such as NBE, NCO and CET. On the contrary, some differences were observed for EME, which is indicated as the second main urinary metabolite of COC with 32e49% excretion in pharmacokinetic studies, and 17% in spot urine analysis, but was detected in wastewater at a lower percentage (12.7%), probably because of its limited stability in wastewater, as also assessed in our study (Fig. 2). The percentage of ECG found in wastewater (8%) was comparable with the figures from pharmacokinetic studies (1e8%), but lower than in spot urine analysis (21%). Since ECG is indicated as the main transformation product of EME during the

Table 4 e Comparison of the metabolic profiles of COC in human urine (from pharmacokinetic studies and spot urine analysis) and wastewater. Substances

BZE

NBE COC

NCO CET ECG EME

AEC AME 1

Percentages of COC and its metabolites from pharmacokinetic studies (2e6 subjects)

Percentages of COC and its metabolites from spot urine analysis (30 subjects)

Percentages of COC and its metabolites from wastewater analysis (entire community)

35e541-3 462 314 605,6,a 16e407 3.15 1e91-3 2e142 2e34 7.55,6,a 0.5e17 e 0.73 2-81 3.35,6,a 32e493 412 265,6,a 7e157 e 0.027

49

54.8

2.5 3.7

1.7 21.2

0.6 1.5 21

0.5 0.6 8.3

17

12.7

4.4 0.3

0.5 e

Fish and Wilson, 1969; 2Ambre et al., 1984; 3Baselt, 2004; 4Hamilton et al., 1977; 5Huestis et al., 2007; 6Smith et al., 2010; 7Cone et al., 1998. a Mean percentages calculated from the reported Cmax following controlled oral, intravenous, intranasal and smoked administration.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 4 1 e5 1 5 0

storage of urine samples (Klingmann et al., 2001), a delay between urine generation and sampling might explain this relatively high percentage of ECG and the low percentage of EME in spot urine. The percentage of COC measured in wastewater (21% of the total) was much higher than expected from pharmacokinetic studies (1e14% of the administered dose) or spot urine analysis (3.7% of the total). These results suggest COC might enter wastewater from other environmental sources besides human urinary excretion, as indicated by the presence of traces of COC on money (Armenta and de la Guardia, 2008) and airborne particulate (Cecinato et al., 2009; Postigo et al., 2009). The metabolic pattern of COC when smoked as crack depends on the pyrolytic conditions and composition of the drug dose (Maurer et al., 2006) and is therefore subjected to wide variability. The average percentages of these metabolites in spot urine samples of suspected consumers of COC by an unknown route were 4.4% and 0.3% of the total metabolites respectively for AEC and AME (Paul et al., 2005). AEC was lower in wastewater (0.5% of the total metabolites), as expected, since only a small proportion of consumers may use COC as crack, the majority being more likely to “snort” the drug.

4.

Conclusions

Profiles of COC metabolites in wastewater qualitatively and quantitatively matched with the profiles in human urine suggesting that concentrations of metabolites in wastewater reflect real human excretion and that wastewater analysis can be used to estimate collective urinary excretion of COC metabolites. Studying the stability of cocaine metabolites in wastewater is essential to figure out eventual losses that can occur in the sewer before wastewater enter STPs, during sampling and/or the storage of samples. BZE occurrence and stability in wastewater make it the best target for estimating community drug use by wastewater analysis. COC itself should not be considered for estimating COC consumption, as its amount in wastewater is probably affected by other environmental sources besides human metabolism, and this might cause overestimation of the real amounts consumed. Indications about COC consumption and pattern of use can also be highlighted by measuring metabolites other than BZE in wastewater. However, considering the documented interconversion of some metabolites, an extended panel of substances should be used, including all the metabolites measured in this study. In this case, the bulk of metabolites found in wastewater would reflect 59e60% of the administered dose and could also indicate if cocaine is smoked, snorted, and consumed together with alcohol. The novel approach based on wastewater analysis, that we call “sewage epidemiology”, is therefore suitable for assessing drug consumption in a population, providing real-time updates of the patterns of drug consumption, and preserving individual anonymity. It is also useful to identify different patterns of consumption in cities and countries and has several fields of application to complement epidemiological studies.

5149

Acknowledgement The authors are grateful to Milano Depur Spa, Como Depur Spa, Abbanoa Spa and the Metropolitan Water Reclamation District of Chicago (MWRDC) for assistance and collaboration for sampling respectively in Milan, Como, Sardinia and Chicago. The authors also thank the Como Local Health Agency and the Cagliari Local Health Agency (ASL 8) for supporting the investigations in Como and Sardinia.

Appendix. Supplementary data Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.07.017.

references

Ambre, J., Fischman, M., Ruo, T.I., 1984. Urinary excretion of ecgonine methyl ester, a major metabolite of cocaine in humans. J. Anal. Toxicol. 8, 23e25. Ambre, J., Ruo, T.I., Nelson, J., Belknap, S., 1988. Urinary excretion of cocaine, benzoylecgonine, and ecgonine methyl ester in humans. J. Anal. Toxicol. 12, 301e306. Armenta, S., de la Guardia, M., 2008. Analytical methods to determine cocaine contamination of banknotes from around the world. TrAC - Trends Anal. Chem. 27, 344e351. Baselt, R.C., 2004. Disposition of Toxic Drugs and Chemicals in Man. Biomedical Publications, Foster City, California, USA. Bisceglia, K.J., Roberts, A.L., Schantz, M.M., Lippa, K.A., 2010. Quantification of drugs of abuse in municipal wastewater via SPE and direct injection liquid chromatography mass spectrometry. Anal. Bioanal. Chem. 398, 2701e2712. Castiglioni, S., Zuccato, E., Chiabrando, C., Fanelli, R., Bagnati, R., 2008. Mass spectrometric analysis of illicit drugs in wastewater and surface water. Mass Spectrom. Rev. 27, 378e394. Castiglioni, S., Zuccato, E., Crisci, E., Chiabrando, C., Fanelli, R., Bagnati, R., 2006. Identification and measurement of illicit drugs and their metabolites in urban wastewater by liquid chromatography-tandem mass spectrometry. Anal. Chem. 78, 8421e8429. Cecinato, A., Balducci, C., Nervegna, G., 2009. Occurrence of cocaine in the air of the World’s cities. An emerging problem? A new tool to investigate the social incidence of drugs? Sci. Total Environ. 407, 1683e1690. Chiaia, A.C., Banta-Green, C., Field, J., 2008. Eliminating solid phase extraction with large-volume injection LC/MS/MS: analysis of illicit and legal drugs and human urine indicators in U.S. wastewaters. Environ. Sci. Technol. 42, 8841e8848. Cone, E.J., Tsadik, A., Oyler, J., Darwin, W.D., 1998. Cocaine metabolism and urinary excretion after different routes of administration. Ther. Drug Monit. 20, 556e560. Daughton, C.G., 2011. Illicit drugs: contaminants in the environment and utility in forensic epidemiology. Rev. Environ. Contam. Toxicol. 210, 59e110. EMCDDA (European Monitoring Centre for Drugs and Drug Addiction), 2008. Insights 9. http://www.emcdda.europa.eu/ publications/insights/wastewater. EMCDDA (European Monitoring Centre for Drugs and Drug Addiction), 2009. Cocaine and crack cocaine. In: Drug Situation in Europe. http://www.emcdda.europa.eu/situation/cocaine/3. Fish, F., Wilson, W.D., 1969. Excretion of cocaine and its metabolites in man. J. Pharm. Pharmacol. 21, 135e138S.

5150

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 4 1 e5 1 5 0

Frost, N., Griffiths, P., Fanelli, R., 2008. Peering into dirty waters: the potential and implications of a new approach to monitoring drug consumption. Addiction 103, 1239e1241. Gonzalez-Marino, I., Quintana, J.B., Rodriguez, I., Cela, R., 2010. Determination of drugs of abuse in water by solid phase extraction, derivatization and gas chromatography-ion traptandem mass spectrometry. J. Chrom. A 1217, 1748e1760. Hamilton, H.E., Wallace, J.E., Shimek, E.L., Land, P., Harris, S.C., Christenson, J.G., 1977. Cocaine and benzoylecgonine excretion in humans. J. Forensic Sci. 22, 697e707. Hogenboom, A.C., van Leerdam, J.A., de Voogt, P., 2009. Accurate mass screening and identification of emerging contaminants in environmental samples by liquid chromatography-hybrid linear ion trap orbitrap mass spectrometry. J. Chromatogr. A 1216, 510e519. Hsieh, Y., 2008. Potential of HILIC-MS in quantitative bioanalysis of drugs and drug metabolites. J. Sep. Sci. 31, 1481e1491. Huerta-Fontela, M., Galceran, M.T., Ventura, F., 2008. Stimulatory drugs of abuse in surface waters and their removal in a conventional drinking water treatment plant. Environ. Sci. Technol. 42, 6809e6816. Huestis, M.A., Darwin, W.D., Shimomura, E., Lalani, S.A., Trinidad, D.V., Jenkins, A.J., Cone, E.J., Jacobs, A.J., Smith, M.L., Paul, B.D., 2007. Cocaine and metabolites urinary excretion after controlled smoked administration. J. Anal. Toxicol. 3, 462e468. Jagerdeo, E., Montgomery, M.A., Lebeau, M.A., Sibum, M., 2008. An automated SPE/LC/MS/MS method for the analysis of cocaine and metabolites in whole blood. J. Chromatogr. B. Anal. Technol. Biomed. Life Sci. 874, 15e20. Jones-Lepp, T.L., Alvarez, D.A., Petty, J.D., Huckins, J.N., 2004. Polar organic chemical integrative sampling and liquid chromatography-electrospray/ion-trap mass spectrometry for assessing selected prescription and illicit drugs in treated sewage effluents. Arch. Environ. Contam. Toxicol. 47, 427e439. Klingmann, A., Skopp, G., Aderjan, R., 2001. Analysis of cocaine, benzoylecgonine, ecgonine methyl ester, and ecgonine by high-pressure liquid chromatography-API mass spectrometry and application to a short-term degradation study of cocaine in plasma. J. Anal. Toxicol. 25, 425e430. Maurer, H.H., Sauer, C., Theobald, D.S., 2006. Toxicokinetics of drugs of abuse: current knowledge of the isoenzymes involved in the human metabolism of tetrahydrocannabinol, cocaine, heroin, morphine, and codeine. Ther. Drug Monit. 28, 447e453. Paul, B.D., Lalani, S., Bosy, T., Jacobs, A.J., Huestis, M.A., 2005. Concentration profiles of cocaine, pyrolytic methyl ecgonidine

and thirteen metabolites in human blood and urine: determination by gas chromatography-mass spectrometry. Biomed. Chromatogr. 19, 677e688. Postigo, C., Lopez de Alda, M.J., Barcelo´, D., 2008. Analysis of drugs of abuse and their human metabolites in water by LC-MS2: a non-intrusive tool for drug abuse estimation at the community level. TrAC - Trends Anal. Chem. 27, 1053e1069. Postigo, C., Lopez De Alda, M.J., Viana, M., Querol, X., Alastuey, A., Artin˜ano, B., Barcelo´, D., 2009. Determination of drugs of abuse in airborne particles by pressurized liquid extraction and liquid chromatography-electrospray-tandem mass spectrometry. Anal. Chem. 81, 4382e4388. Richardson, S.D., 2009. Water analysis: emerging contaminants and current Issues. Anal. Chem. 81, 4645e4677. Smith, M.L., Shimomura, E., Paul, B.D., Cone, E.J., Darwin, W.D., Huestis, M.A., 2010. Urinary excretion of ecgonine and five other cocaine metabolites following controlled oral, intravenous, intranasal, and smoked administration of cocaine. J. Anal. Toxicol. 34, 57e63. Terzic, S., Senta, I., Ahel, M., 2010. Illicit drugs in wastewater of the city of Zagreb (Croatia)eestimation of drug abuse in a transition country. Environ. Pollut. 158, 2686e2693. UNODC (United Nations Office of Drugs and Crime), 2009. World Drug Report. http://www.unodc.org/documents/wdr/WDR_ 2009/WDR2009_eng_web.pdf. van Nuijs, A.L., Castiglioni, S., Tarcomnicu, I., Postigo, C., de Alda, M.L., Neels, H., Zuccato, E., Barcelo, D., Covaci, A., 2010. Illicit drug consumption estimations derived from wastewater analysis: a critical review. Sci. Total Environ.. doi:10.1016/j. scitotenv.2010.05.030 20598736. van Nuijs, A.L., Pecceu, B., Theunis, L., Dubois, N., Charlier, C., Jorens, P.G., Bervoets, L., Blust, R., Neels, H., Covaci, A., 2009. Spatial and temporal variations in the occurrence of cocaine and benzoylecgonine in waste- and surface water from Belgium and removal during wastewater treatment. Water Res. 43, 1341e1349. Zuccato, E., Castiglioni, S., 2009. Illicit drugs in the environment. Philos. Trans. A Math. Phys. Eng. Sci. 367, 3965e3978. Zuccato, E., Chiabrando, C., Castiglioni, S., Bagnati, R., Fanelli, R., 2008. Estimating community drug abuse by wastewater analysis. Environ. Health Perspect. 116, 1027e1032. Zuccato, E., Chiabrando, C., Castiglioni, S., Calamari, D., Bagnati, R., Schiarea, S., Fanelli, R., 2005. Cocaine in surface waters: a new evidence-based tool to monitor community drug abuse. Environ. Health 4, 14.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 5 1 e5 1 6 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Mercury and trace metal partitioning and fluxes in suburban Southwest Ohio watersheds Avani P. Naik, Chad R. Hammerschmidt* Department of Earth & Environmental Sciences, Wright State University, 3640 Colonel Glenn Highway, Dayton, OH 45435, USA

article info

abstract

Article history:

Many natural watersheds are increasingly affected by changes in land use associated with

Received 30 December 2010

suburban sprawl and such alterations may influence concentrations, partitioning, and

Received in revised form

fluxes of toxic trace metals in fluvial ecosystems. We investigated the cycling of mercury

18 May 2011

(Hg), monomethylmercury, cadmium, copper, lead, nickel, and zinc in three watersheds at

Accepted 14 July 2011

the urban fringe of Dayton, Ohio, over a 13-month period. Metal concentrations were

Available online 23 July 2011

related positively to discharge in each stream, with each metal having a high affinity for suspended particles and Hg also having a noticeable association with dissolved organic

Keywords:

carbon. Although not observed for the other metals, levels of Hg in river water varied

Organic carbon

seasonally and among streams. Yields of Hg from two of the catchments were comparable

Suspended solids

to that predicted for runoff of atmospherically deposited Hg (w25% of wet atmospheric

Waste water treatment facility

flux), whereas the third watershed had a significantly greater annual flux associated with

River

greater particle-specific and filtered water Hg concentrations, presumably from a point

Land use

source. Fluxes of metals other than Hg were similar among each watershed and suggestive of a ubiquitous source, which could be either atmospheric deposition or weathering. Results of this study indicate that, with the exception of Hg being increased in one watershed, processes affecting metal partitioning and loadings are similar among southwest Ohio streams and comparable to other North American rivers that are equally or less impacted by urban development. Relative differences in land use, catchment area, and presence or absence of waste water treatment facilities had little or no detectable effect on most trace metal concentrations and fluxes. This suggests that suburban encroachment on agricultural and undeveloped lands has either similarly or not substantially impacted trace metal cycling in streams at the urban fringe of Dayton and, by extension, other comparable metropolitan areas. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Trace-metal pollution of surface water is a concern in the United States (U.S. Environmental Protection Agency, 2009) and globally (World Health Organization, 2008) because of the potential toxicity of metals to both aquatic biota (Hare, 1992) and humans (Peraza et al., 1998; Mergler et al., 2007). Although

naturally occurring, a major fraction of trace metals in aquatic systems can be linked to anthropogenic sources. Cadmium (Cd), copper (Cu), lead (Pb), nickel (Ni), and zinc (Zn), for example, either are or have been used commonly in a variety of commercial products and industrial processes (Patterson, 1965; Burton and Pitt, 2002). Mercury (Hg) is derived largely from combustion of fossil fuels, namely coal (Pacyna et al.,

* Corresponding author. Tel.: þ1 937 775 3457. E-mail address: [email protected] (C.R. Hammerschmidt). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.023

5152

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 5 1 e5 1 6 0

2006), and owing to its long-range atmospheric distribution (Lamborg et al., 2002) is ubiquitous in the environment. Each of these metals are eventually mobilized to surface waters from either point or non-point sources (Shafer et al., 1997; Lawson et al., 2000; Tiefenthaler et al., 2008; Balogh et al., 2005; Eckley and Branfireun, 2008; Li et al., 2009). The bioavailability, bioaccumulation, and toxicity of trace metals in surface waters are related to their loadings, partitioning, and chemical speciation with organic and inorganic ligands. For many metals, the free ion is believed to be more bioavailable than complexed forms (Allen et al., 1980), although biotransformation to monomethylmercury (MMHg) increases availability and toxicity of Hg (Wiener et al., 2003). Watershed characteristics, differences in land use, and variations of discharge can affect levels of dissolved organic carbon (DOC) and suspended solids in streams that, in turn, influence transport and partitioning of metals (e.g., Hurley et al., 1995; Shafer et al., 1997, 1999; Balogh et al., 2005; Eckley and Branfireun, 2008; Tiefenthaler et al., 2008; Brigham et al., 2009; Li et al., 2009; Cloran et al., 2010). Many natural watersheds are increasingly affected by changes in land use associated with suburban sprawl. Annually in the United States, more than 2  106 acres of land on the urban fringe are developed for residential and commercial purposes (U.S. Department of Housing and Urban Development, 2000). Development can change the hydrology of watersheds, introduce anthropogenic sources of trace metals and other pollutants, alter contaminant transport and fate, and result in greater discharge of treated and untreated waste water to rivers, each of which can affect the cycling of trace metals in fluvial ecosystems. Despite the growing area and population of the urban fringe, little is known about tracemetal concentrations, partitioning, and fluxes in watersheds that have mixed land uses and an increasing suburban/residential component. The objective of this study was to investigate the cycling and fluxes of Hg, MMHg, Cd, Cu, Pb, Ni, and Zn in three watersheds with mixed and increasingly urban land uses in southwest Ohio near the Dayton metropolitan area.

2.

Materials and methods

2.1.

Water sampling

We examined Hg, MMHg, Cd, Cu, Pb, Ni, and Zn in three watersheds with mixed land uses near Dayton, Ohio (Fig. 1). This region has experienced a more than 30% increase in urban land use during the past three decades (Exurban Change Program, 2010), with the Dayton suburbs expanding into each of the three study watersheds: Wolf Creek, Holes Creek, and the upper Little Miami River. These streams have not been investigated previously for trace metals, and there is little information, in general, on metal cycling in the greater Ohio River watershed. Wolf and Holes Creeks drain dominantly urban/residential catchments (Ohio Environmental Protection Agency, 1997) and both have gauging stations (U.S. Geological Survey, 2010) near which river water was sampled. The watershed areas of Wolf and Holes Creeks upstream of the sampling locations are 92 and 25 km2, respectively. The upper Little Miami River drains a mixed crop

Fig. 1 e Watersheds of Wolf Creek, Holes Creek, and Little Miami River near the Dayton metropolitan region in Montgomery, Greene, and Warren Counties of southwest Ohio, USA. Watersheds are delineated by solid lines, sampling locations are marked with filled circles, and urbanized area of Dayton is shaded. Adapted from Ohio Department of Natural Resources (1999).

land/residential catchment and also receives effluent from multiple wastewater treatment facilities (WWTFs) that serve suburbs and small municipalities east of Dayton. Little Miami water was sampled at the U.S. Geological Survey gauging station in Spring Valley, OH, where the upstream watershed is 492 km2 (Schiefer, 2002). To investigate the potential influence of waste water discharge on trace metal loadings and biogeochemistry, Little Miami water also was sampled about 2 km downstream of the WWTF in Waynesville, OH, about 10 km downstream of the primary sampling location. Samples were collected at the downstream and upstream locations within 30 min of each other on each of the sampling days. Gauging stations in Wolf and Holes Creeks measure instantaneous discharge whereas the one in the Little Miami River measures instantaneous stage height. Each of the three watersheds are in the Till Plains region of the Central Lowland Province of Ohio, with the geology comprised mainly of glacial moraine deposits and Ordovician-Silurian age calcareous shales and limestones (Schiefer, 2002). River water was sampled on a near weekly basis from each of the watersheds over a 13-month period (April 2009eMay 2010). Conductivity, pH, and temperature were measured in situ with electrochemical probes and a thermocouple. Water was sampled with trace-metal clean techniques (Gill and Fitzgerald, 1985) into 2-L FEP Teflon bottles and transported promptly to Wright State University for processing and chemical analysis. Particulate and filter-passing (<0.2 mm)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 5 1 e5 1 6 0

5153

phases of metals in water were separated with polycarbonate membranes. Filtrate for total metals analysis was preserved by acidification to 0.2% with HNO3 (J.T. Baker Instra-Analyzed) and for MMHg determination by acidification to 0.2% with H2SO4 (J.T. Baker Instra-Analyzed). Separate aliquots were preserved for analysis of nitrate and chloride by ion chromatography (APHA et al., 1995) and DOC by an infrared combustion method (Sharp et al., 1995). Total suspended solids (TSS) were measured with standard methods (APHA et al., 1995). All equipment and bottles were cleaned rigorously with acid and rinsed with reagent-grade water (>18 MU-cm) prior to use.

known additions from aqueous samples averaged 87  16% (n ¼ 7). Limits of quantification were less than sample concentrations. Distribution coefficients (KD, L kg1) of trace metals were estimated as the ratio of metal concentrations associated with particles (ng kg1 dry weight) to that in 0.2-mm filtered water (ng L1). Particle-specific concentrations of each trace metal (ng kg1 dry weight) were calculated as the quotient of particle-associated metal (ng L1) and concentration of TSS (kg L1) in each sample.

2.2.

3.

Results and discussion

3.1.

Stream physicochemistry

Metals analysis

Total Hg in filtered water and particles was determined by cold-vapor atomic fluorescence spectrometry (CVAFS). Particles were digested with 4 N HNO3 in a 60  C water bath prior to analysis (Hammerschmidt and Fitzgerald, 2006). Filtered water and particle digestates for analysis of total Hg were oxidized with BrCl (Bloom and Crecelius, 1983) for > 12 h and pre-reduced with NH2OH prior to reduction with SnCl2 and Hg quantification by dual-Au amalgamation CVAFS (Fitzgerald and Gill, 1979; Bloom and Fitzgerald, 1988). Total Hg determinations were calibrated versus Hg0 taken from the headspace over pure liquid and verified by comparison to analyses of aqueous Hg2þ standards traceable to the U.S. National Institute of Standards and Technology (NIST). Recovery (1 SD) of aqueous Hg2þ averaged 106  12% (n ¼ 40) for filtered water and 102  12% (n ¼ 24) for particle digestates. Precision of duplicate measurements averaged 8.7% relative difference (range ¼ 0.1e30%; n ¼ 25) for filtered water and 10% relative difference (range ¼ 0.1e18%, n ¼ 48) for particles. Detection limits for total Hg were 0.01 ng L1 for both filtered water and particle digestates. MMHg was measured in filtered water only by flowinjection, gas-chromatographic CVAFS (Bloom, 1989; Tseng et al., 2004) after treatment with dilute sulfuric acid (Bowman and Hammerschmidt, 2011). Determinations of sample MMHg were calibrated with an aqueous solution of CH3HgCl that was standardized versus Hg0 and a NIST-traceable Hg2þ standard. Precision of procedurally duplicated MMHg determinations averaged 20% relative difference (range ¼ 1.5e41, n ¼ 27). Recovery of known MMHg additions from filtered river water averaged 93  8% (n ¼ 12). The detection limit for MMHg was about 0.005 ng L1. Particle and filtered Cd, Cu, Pb, Ni, and Zn were measured by inductively coupled plasma mass spectrometry (U.S. Environmental Protection Agency, 2007) with a PerkinElmer Elan 9000. Metals in filtered, acidified water were determined directly, and particle samples were digested with 4 N HNO3 prior to analysis (Hammerschmidt and Fitzgerald, 2006). All metals analyses were calibrated with standards traceable to the U.S. NIST. Quality assurance analyses included procedural and filtration blanks, replicate samples, and sample aliquots with known additions of metals. Precision of measurements varied among metals and increased with greater ambient concentrations. Precision of procedurally duplicated samples averaged 20% relative difference for filtered water (n ¼ 32) and 6% relative difference (n ¼ 35) for particles. Recoveries of

Water physicochemistry varied among the three streams over the 13-month sampling period (Supplementary Table S1). ANOVA tests showed that temperature ( p ¼ 0.35), pH ( p ¼ 0.07), and TSS ( p ¼ 0.96) were similar among all three streams, with each having an annual temperature range of about 1e24  C, a mean annual pH of about 8.3, and an average suspended sediment load of about 30 mg L1. Similarities of temperature and pH were expected because the watersheds have the same base geology, replete with calcareous shales and limestones that buffer the water (Schiefer, 2002), and are subject to the same weather variations and climate. In contrast, Holes Creek had greater conductivity than either of the other two streams (Tukey, p-values < 0.001). Nitrate was significantly different among each watershed (Tukey, p-values < 0.05), with the Little 1 Miami River (mean  SD annual NO 3 , 11.2  4.0 mg L ), the dominantly agricultural/residential watershed, having greater levels than Wolf (8.0  7.2 mg L1) and Holes (4.6  2.7 mg L1) Creeks. Differences in water physicochemistry were evident seasonally and corresponded with hydrography. Each stream had four distinct and coincident hydraulic periods; greatest discharge occurred during wet seasons indentified, for the purpose of this study, as spring (March 1 to June 30; Julian days 60e181) and fall (September 1 to November 30; Julian days 244e334) interspersed by drier periods in the summer and winter (Fig. 2). Significant differences in either stage height (Little Miami River; Tukey, p-values < 0.05) or volumetric discharge (Holes and Wolf Creeks; Tukey, p-values < 0.001) were observed among all seasons except between spring and fall, which had similarly high discharge. In each stream, pH was greater in winter and spring compared to summer and fall (Tukey, p-values < 0.05), presumably as a result of temperature-related differences in biological respiration. Conductivity was greatest during winter (Tukey, pvalues < 0.001), whereas DOC was greater during spring and fall, seasons of higher discharge, than in winter and summer in each stream (Tukey, p-values < 0.05). Increased conductivity in winter, especially in Holes Creek, is attributed to loadings of road salt and is evidenced by increased levels of chloride. DOC and TSS were related positively to either volumetric discharge or stage height in each river (Supplementary Figs. S1 and S2), which is typical of other streams (e.g., Brigham et al., 2009). Nitrate did not vary seasonally (ANOVA, p ¼ 0.051).

5154

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 5 1 e5 1 6 0

A

B

C

Fig. 2 e Temporal variation of unfiltered total Hg (solid line) and either discharge at Wolf and Holes Creeks or stage height above datum (dashed lines) at Little Miami River during the 13-month sampling period.

3.2.

Total Hg

Levels of total Hg differed among the three streams (Table 1). Total Hg in filtered water (<0.2 mm) was greater in Wolf Creek than in either Holes Creek or the Little Miami River (Tukey, p-values < 0.05). Increased total Hg in the Wolf Creek

watershed also was evident when filtered and particle concentrations were added to estimate an “unfiltered” concentration. Unfiltered total Hg was greater in Wolf Creek than in either of the other two watersheds (Tukey, pvalues < 0.05), which were not different from each other. Levels of total Hg in southwest Ohio streams are comparable to those at other locations in North America (Supplementary Table S2). On average, about 70% of Hg in the three streams was associated with particles (Table 1). Atmospheric deposition is a principal source of Hg to most locations (Fitzgerald et al., 1998), and it is presumed that atmospheric fluxes of Hg are comparable among each of these nearby watersheds that contain no major Hg emission sources. Accordingly, increased levels of total Hg in Wolf Creek are likely related to either point sources in the catchment or watershed characteristics that enhance delivery of Hg to the stream (Li et al., 2009). Total Hg varied seasonally in each stream (Fig. 2). Total Hg generally was lowest during periods of reduced stream discharge in summer and winter and increased during the spring and fall. Indeed, levels of filtered Hg were significantly greater in fall and spring than in winter (Tukey, p-values < 0.05). In each stream, unfiltered total Hg was related positively to instantaneous discharge (Fig. 3), similar to observations in other fluvial systems (Balogh et al., 2005; Brigham et al., 2009). The correlation between unfiltered Hg and water discharge can be dissected into two components that influence the transport of HgdDOC and TSS, both of which also were correlated positively with instantaneous discharge (Supplementary Figs. S1 and S2). Filtered total Hg was related to DOC in southwest Ohio streams (Fig. 4). Similar relationships have been observed in other rivers (Lyons et al., 2006; Brigham et al., 2009). DOC can serve as a transport vector for dissolved metals (Shafer et al., 1997) and such a relationship between filtered Hg and DOC might be expected given the high affinity of Hg2þ for dissolved organic ligands (Lamborg et al., 2003). However, the ratio of Hg:DOC is much less than expected if all of the Hg-binding organic ligands (L) were saturated with Hg. Measured L:DOC ratios in river water are about 10e80  106 (Lamborg et al., 2004). Accordingly, for river water that contains from 2 to 20 mg L1 DOC (Fig. 4), one would predict a dissolved L concentration of 2e130 nM, which is substantially greater than measured Hg in the filtered fraction (0.003e0.01 nM). This suggests that Hg-binding organic ligands associated with DOC are not fully saturated with Hg and may compete with ligands on suspended solids for Hg. Undersaturation of Hg-complexing ligands may result from competition by other metals for the same binding sites. Particulate total Hg also was related strongly to TSS (Fig. 5), which is consistent with observations elsewhere (Mason and

Table 1 e Annual mean (±1 SD) concentrations and distribution coefficients (KD) of Hg species in the three study streams (n [ 61 for each). MMHg was sampled from AprileDecember 2009 only (n [ 28e29). Stream Wolf Creek Holes Creek Little Miami R.

Filtered Hg (ng L1)

Particulate Hg (ng L1)

Particulate Hg (ng g1)

log KD Hg (L kg1)

Filtered MMHg (ng L1)

1.50  2.12 0.87  0.78 0.75  0.52

3.43  6.04 1.44  1.66 2.48  2.86

302  304 133  83 107  57

5.35  0.60 5.18  0.50 5.17  0.38

0.036  0.026 0.040  0.032 0.043  0.027

5155

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 5 1 e5 1 6 0

Wolf Holes Little Miami

-1

Particulate Hg (ng L )

100

10

1 2

r = 0.66 p < 0.001 Hg = 0.59[TSS] - 0.46 0.1 0.1

1

10

100

1000

TSS (mg L-1)

Fig. 3 e Unfiltered total Hg versus either instantaneous discharge for Wolf and Holes Creeks or instantaneous stage height for Little Miami River (LMR) (Wolf Creek: y [ 1.3x D 1.7, r2 [ 0.48; Holes Creek: y [ 0.8x D 1.5, r2 [ 0.77; LMR: y [ 6.4x L 3.8, r2 [ 0.44; p-values < 0.001).

Sullivan, 1998; Lawson and Mason, 2001; Balcom et al., 2008; Eckley and Branfireun, 2008; Brigham et al., 2009). Hg species are associated largely with the organic phase of particulate material (Hammerschmidt et al., 2004). It was surprising to find such similarity in the particle Hg versus TSS relationship among the three streams, given that the organic content of suspended particles can vary independently of TSS and that particulate organic matter is thought to be a better proxy than TSS for Hg (HydroQual, 2006; Mason and Sullivan, 1998; Mason et al., 1999). This suggests that the geochemical composition of particles is relatively consistent among the three streams, and this is supported by similarities of distribution coefficients for Hg (Table 1), which are within the range of those determined in other fluvial systems (Lawson et al., 2000; Lyons et al., 2006; Balcom et al., 2008; Brigham et al., 2009). Distribution coefficients of total Hg were related inversely to DOC (Fig. 6; r2 ¼ 0.23, p < 0.001). This trend indicates that proportionately more of the Hg is in the filter-passing phase of

Fig. 5 e Relationship between particulate total Hg and total suspended solids (TSS) in Wolf and Holes Creeks and Little Miami River.

water as DOC increases. This can be explained by ligands associated with DOC competing with particle-phase ligands for Hg, which would be expected given the likely undersaturation of Hg-binding ligands in filtered water. An alternative hypothesis is that the affinity of Hg for particles decreases under environmental conditions that correspond with greater DOC levels. We observed that both DOC and TSS are correlated positively with instantaneous discharge (Supplemental Figs. S1 and S2). One might expect that greater stream discharge would suspend large particles (e.g., sand) that have a lower affinity for Hg compared to organic material and thereby lower the KD (Hammerschmidt et al., 2004). Indeed, the distribution coefficient of total Hg also was related inversely to TSS (Fig. 7), which would support a “particledilution” hypothesis. However, particulate Hg concentration is related strongly to TSS concentration and, by extension, instantaneous discharge (Fig. 5). Such a relationship would not be expected if particle dilution were operative: Particle dilution would not result in particulate Hg levels increasing linearly with TSS. Together, these results imply that differences in the particle-water partitioning of Hg are influenced

6.5

Hg = 0.06[DOC] - 0.46 2 r = 0.27 p < 0.001

Wolf Holes Little Miami

6.0 -1

Hg log KD (L kg )

Wolf Holes Little Miami

-1

Filtered Hg (ng L )

100

10

1

5.5 5.0 4.5 4.0 3.5

2

r = 0.23 p < 0.001

3.0 0

0.1 0

5

10

15

20

5

10

15

20

DOC (mg L-1)

-1

DOC (mg L ) Fig. 4 e Filtered total Hg versus dissolved organic carbon (DOC) in Wolf and Holes Creeks and the Little Miami River.

Fig. 6 e Variation of the distribution coefficient (KD) of total Hg with dissolved organic carbon (DOC) in the three study streams: log KD [ L0.09[DOC] D 5.71.

5156

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 5 1 e5 1 6 0

Hg log KD (L kg-1)

7 Wolf Holes Little Miami

6

5

4 2

3 0.1

r = 0.57 p < 0.001 log KD = -0.64[TSS] + 5.90 1

10

TSS (mg L-1)

100

1000

Fig. 7 e Correlation of the distribution coefficient (KD) of total Hg with total suspended solids (TSS) in the three study systems.

by variations of either dissolved organic ligands or Hg-binding colloidal material in the <0.2 mm fraction, as suggested by Mason and Sullivan (1998). Particle-specific concentrations of Hg were highly variable within and among the three study streams (Table 1). Suspended solids in Holes Creek and the Little Miami River contained between about 10 and 400 ng g1 Hg. These levels are within the range of those measured in surface soils of Montgomery and Greene Counties (range, 1e370 ng Hg g1 dry weight; Tabatchnick, 2010). In contrast, suspended material in Wolf Creek contained from about 20 to 1700 ng Hg g1. Concentrations at the upper end of this range are much greater than can be explained from either common sources of anthropogenic contamination or atmospheric deposition alone and suggest that Wolf Creek watershed has either current or historic point sources of Hg, which is consistent with differences of Hg in filtered water among streams. The watershed of Wolf Creek, unlike the upper Little Miami River and Holes Creek, has hosted a variety of major industrial and manufacturing activities.

3.3.

MMHg

Filtered MMHg was measured during spring, summer, and fall only. Mean levels of MMHg were comparable among each of the

three streams and averaged about 0.04 ng L1 (Table 1). For the AprileDecember period, filtered MMHg was unrelated to filtered total Hg in each of the three streams ( p-values ¼ 0.1e0.8) and, among all three, averaged 5.6% (range, 0.11e35%) of filtered total Hg, which is comparable to that in rivers in Connecticut (2e5%, Balcom et al., 2004), Wisconsin (0.2e11%, Hurley et al., 1995), and Maryland (0.8e3%, Lawson and Mason, 2001). Absence of a correlation between filtered MMHg and total Hg suggests differences in either the partitioning or sources of MMHg and Hg(II). Mean MMHg levels in southwest Ohio streams are considerably less than those in agricultural watersheds of Minnesota (0.20e0.35 ng L1; Balogh et al., 2005) but within the range of those in Wisconsin (0.03e0.09 ng L1; Hurley et al., 1995) and Connecticut (0.04e0.40 ng L1; Balcom et al., 2004). Filtered MMHg was significantly less in summer than during spring or fall (Tukey, p-values  0.004), which did not differ. Similar seasonal differences of MMHg have been observed in other North American streams (Hurley et al., 1995; Lawson et al., 2000; Lawson and Mason, 2001; Balogh et al., 2005; Brigham et al., 2009) and attributed to production in watershed soils with subsequent mobilization during high discharge events in spring (Balogh et al., 2005). However, and as observed elsewhere (Lawson et al., 2000), no significant correlation was found between filtered MMHg and discharge in this study ( p ¼ 0.76).

3.4.

Other trace metals

Unlike Hg, levels of Cd, Cu, Pb, Ni, and Zn were not different among streams (Table 2). The only exceptions to this generalization were filtered Ni being less in Wolf Creek than the other two streams, filtered Cu being greater in Holes Creek, and particulate Cd being greater in Wolf Creek (Tukey, pvalues  0.05). Additionally, and in contrast to total Hg and MMHg, there was no detectable seasonal variation of trace metal concentrations, with the exception of filtered Pb being less in summer (Tukey, p-values < 0.05). Sources of these metals include atmospheric deposition, weathering, and local anthropogenic inputs. The relative similarity of trace metal concentrations among streams suggests that the dominant source is ubiquitous, which is consistent with inputs from either weathering or atmospheric deposition. Pb levels in rivers have been found to correspond well with that in wet atmospheric deposition (Lawson and Mason, 2001). Similar to Hg, each of the five other trace metals exhibited a high affinity for suspended solids. Particulate metal

Table 2 e Annual mean (±1 SD) concentrations of trace metals in filtered (<0.2 mm) and particulate phases of water from Wolf Creek, Holes Creek, and Little Miami River. Filtered water (mg L1)

Stream Cd

Cu

Ni

Pb

Zn

Wolf Creek Holes Creek Little Miami R.

0.03  0.04 0.03  0.02 0.02  0.02

1.22  0.60 2.25  1.48 1.25  0.53

0.03  0.05 0.04  0.04 0.05  0.06

4.06  8.53 3.22  3.76 3.98  2.91

Creek Holes Creek Little Miami R.

0.03  0.04 0.01  0.02 0.01  0.01

0.97  1.75 0.83  1.40 1.02  1.07

2.65  0.66 3.27  0.71 3.32  0.87 Particles (mg L1) 0.64  1.31 0.65  1.20 0.77  0.86

1.08  2.20 0.53  0.88 0.84  0.96

4.86  7.40 3.78  5.73 4.17  3.63

5157

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 5 1 e5 1 6 0

concentrations were related positively to TSS (Cd, r2 ¼ 0.35; Cu, r2 ¼ 0.36; Pb, r2 ¼ 0.61; Ni, r2 ¼ 0.59; Zn, r2 ¼ 0.55; p-values < 0.001) with results from all streams combined. Additionally, and as expected, unfiltered concentrations of each metal were correlated positively with instantaneous discharge (r2 ¼ 0.29e0.83; p-values < 0.05). With the exception of Pb, the affinity of each metal for suspended particles was less than that of Hg (Tables 1 and 3). In contrast to Hg, filtered Cd, Cu, and Zn were unrelated to DOC ( p ¼ 0.12e0.96) and filtered Pb and Ni were correlated inversely with DOC ( p-values < 0.05), albeit weakly (r ¼ 0.2 and 0.3, respectively). Absence of a positive correlation between these metals and DOC may be attributed to variability in the speciation/complexation of metals in the filtered fraction. Filtered metal concentrations and distribution coefficients in these southwest Ohio streams are within the range of those reported for rivers in non-urbanized watersheds of Wisconsin (Shafer et al., 1997) and mixed land use catchments in Maryland (Lawson et al., 2000).

3.5.

Impact of waste water treatment facilities

The Little Miami River, unlike Wolf and Holes Creeks, receives discharge from several WWTFs upstream of our primary sampling location. To investigate the potential impact of these facilities on water quality and metals loadings, we also sampled river water at a location 10 km downstream of our primary sampling location, about 2 km below the outfall of the WWTF at Waynesville, OH. None of the measured physicochemical variables or trace metal concentrations differed significantly between the upstream and downstream sampling locations (paired t-test, p-values > 0.3). These results suggest that the Waynesville WWTF was not a significant source of metals to the river and, by extension, we infer that other upstream facilities also do not substantially influence levels of trace metals in the Little Miami River.

3.6.

Hg fluxes

Watershed yields of Hg were estimated from measured concentrations and water discharges for each of the study catchments. Unfiltered total Hg concentrations were a function of discharge in each stream (Fig. 3). Accordingly, a dischargeweighted mean Hg concentration was estimated for each stream at the mean instantaneous volumetric discharge during the study period (Holes Creek ¼ 0.94 m3 s1; Wolf Creek ¼ 2.48 m3 s1). Watershed fluxes were estimated as the product of discharge-weighted mean concentrations (Holes Creek ¼ 2.31  1.12 ng L1; Wolf Creek ¼ 4.97  0.68 ng L1) and measured annual water discharge (Holes Creek ¼ 0.022 km3 y1;

Wolf Creek ¼ 0.070 km3 y1). Uncertainty of the dischargeweighted mean concentration is the 95% confidence interval of the linear regression of unfiltered total Hg versus discharge. Discharge-weighted mean concentrations were similar to the arithmetic mean in each stream (Holes ¼ 2.33 ng L1; Wolf ¼ 5.05 ng L1). Mass fluxes of Hg were normalized for upstream watershed area to estimate yields of 2.0  0.2 mg Hg m2 y1 for Holes Creek and 3.8  1.0 mg m2 y1 for Wolf Creek. The same approach was used to estimate Hg yields from Wolf and Holes Creeks watersheds during the period of high discharge in spring, when mean instantaneous volumetric discharges (Holes Creek ¼ 1.3 m3 s1; Wolf Creek ¼ 3.5 m3 s1) and discharge-weighted Hg concentrations (Holes Creek ¼ 2.61  0.27 ng L1; Wolf Creek ¼ 6.36  1.30 ng L1) were increased. Estimated watershed yields of Hg during this 122-d period account for 51% of the annual flux from Holes Creek watershed and 74% of the annual flux from Wolf Creek. Given that 45% (Holes) and 58% (Wolf) of the volumetric discharge occurs during this period, it appears that a disproportionate amount of Hg is mobilized during the spring from Wolf Creek. Instantaneous discharge was not measured at the Spring Valley sampling location on the Little Miami River; however, the mean volumetric discharge was measured to be 0.38  0.12 km3 y1 during a 15-y period from 1968 to 1983 (U.S. Geological Survey, 2010). Additionally, the annual water discharge of the Little Miami River at our sampling location can be estimated from the upstream watershed area and by assuming precipitation runoff in the Little Miami watershed is similar to that in Wolf (0.76 m y1) and Holes (0.86 m y1) Creeks. These values were estimated by dividing measured volumetric discharge by watershed area. The product of watershed area (492 km2) and predicted runoff velocity (0.8 m y1) for the Little Miami River equates to an estimated volumetric flux for the study period of 0.39 km3 y1, which is in good agreement with the 15-y mean noted above. The yield of Hg from the Little Miami River watershed was estimated as the product of arithmetic mean (SD) concentration of unfiltered Hg (3.17  3.02 ng L1) and annual volumetric flux normalized for watershed area. A dischargeweighted mean could not be determined for the Little Miami River, but we found discharge-weighted and arithmetic means to be similar in Wolf and Holes Creeks. The estimated yield of Hg from the Little Miami watershed is 2.4  2.3 mg m2 y1. Hg fluxes from the three study catchments are similar to those from the Connecticut River (2.4 mg m2 y1; Balcom et al., 2004) and watersheds in Oregon, Wisconsin and Florida (0.9e4.4 mg m2y1; Brigham et al., 2009) but at the lower end of estimates for other watersheds in Maryland, Wisconsin, and

Table 3 e Mean (±1 SD) distribution coefficients (KD) of trace metals in Wolf Creek, Holes Creek, and Little Miami River (n [ 37e61). log KD (L kg1)

Stream

Wolf Creek Holes Creek Little Miami R.

Cd

Cu

Ni

Pb

Zn

4.85  0.61 4.46  0.62 4.55  0.52

4.62  0.48 4.26  0.36 4.39  0.48

4.06  0.24 3.94  0.43 3.75  0.26

6.39  0.46 6.03  0.49 5.90  0.39

5.16  0.47 4.99  0.51 4.64  0.38

5158

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 5 1 e5 1 6 0

Table 4 e Annual trace metal yields (±95% CI) from Wolf Creek, Holes Creek and Little Miami River watersheds. Percent of annual flux during spring period is given in parentheses. Flux (mg m2 y1)

Stream

Wolf Creek Holes Creek Little Miami R.

Cd

Cu

Ni

Pb

Zn

0.040  0.010 (69) 0.033  0.005 (48) 0.027  0.010

1.6  0.2 (73) 2.7  0.4 (48) 1.8  1.0

2.5  0.1 (63) 3.4  0.1 (47) 3.0  0.5

0.8  0.2 (89) 0.5  0.1 (55) 0.7  0.5

6.0  1.3 (70) 5.7  1.1 (50) 6.4  2.2

Minnesota (Hurley et al., 1995; Lawson and Mason, 2001; Lawson et al., 2000; Balogh et al., 2005). Wet atmospheric deposition of Hg in southwest Ohio is estimated to be about 10 mg m2 y1 (Mercury Deposition Network, 2010). Prior studies have found that about 25% of Hg deposited atmospherically to terrestrial catchments is exported to receiving waters (Swain et al., 1992; Lawson and Mason, 2001; Engstrom et al., 1994). Accordingly, runoff of atmospherically deposited Hg is expected to be about 2.5 mg m2 y1 in the study watersheds. This estimated yield is consistent with measured fluxes in Holes Creek (2.0  0.2 mg m2 y1) and Little Miami River (2.4  2.3 mg m2 y1) watersheds and suggests that atmospheric deposition is the principal source of Hg to these catchments. Good agreement between predicted and measured fluxes in Holes Creek and the Little Miami River also supports our assessment that WWTFs along the Little Miami have little impact on Hg loadings. The yield of Hg from the Wolf Creek watershed (3.8  1.0 mg m2 y1) is greater than the other catchments and that predicted from wet atmospheric deposition alone. Greater export of Hg from Wolf Creek suggests that either atmospherically deposited Hg is more mobile in this watershed or there are additional sources of Hg. While neither hypothesis can be tested unequivocally in this study, increased levels of filtered Hg and the unusually high particle-specific Hg concentrations in Wolf Creek (Table 1) imply that there are point sources of Hg in the watershed. Water from Wolf and Holes Creeks and the Little Miami River eventually joins the Mississippi River and enters the Gulf of Mexico. If the processes and forces affecting Hg yields from southwest Ohio catchments (mean ¼ 2.7  0.9 mg Hg m2 y1) were generalized broadly to be representative of those

throughout the greater Mississippi watershed (3.0  106 km2), then we would predict an annual discharge of 8  3 tons Hg y1 to the Gulf of Mexico. A mass balance estimate suggests that the discharge of Hg from the combined Mississippi/ Atchafalaya River system is 7.3  3 tons y1 (Rice et al., 2009). This simple scaling argument suggests that the primary factors affecting Hg mobilization from catchments in southwestern Ohio are either representative or comparable to those throughout much of the midcontinent of North America.

3.7.

Watershed fluxes were calculated similarly for the other trace metals, both on an annual and seasonal basis (Table 4). Fluxes of Cd, Cu, Pb, Ni, and Zn were orders of magnitude greater than those of Hg in each of the three streams. Yields of each of the metals were similar among catchments and compare well with those for the Potomac and Susquehanna River watersheds (Lawson et al., 2000). Agreement among the study catchments further supports our assessment that WWTFs along the Little Miami have a minor impact on metal loadings. As observed for Hg, almost 50% or more of the trace metal flux occurred during spring, which is consistent with results from other watersheds (Tiefenthaler et al., 2008). Similarities of trace metal fluxes among the three watersheds suggest common sources, including, for example, atmospheric deposition and weathering. However, the Wolf Creek watershed appears to have a disproportionately high efflux of Cu, Pb, Zn, and Hg relative to water discharge (58%) during the spring period. This indicates that metals may be more mobile in Wolf Creek watershed compared to Holes Creek, where the annual proportion of metals flux during the spring (39e55%) is comparable to the fraction of volumetric discharge that occurs during this period (45%). The relative flux of metals from the Wolf Creek watershed during spring is correlated positively with distribution coefficients of individual metals (Fig. 8), which suggests that increased mobilization in spring is mediated principally by flushing of particles. Greater mobility in the Wolf Creek watershed could be attributed to a greater extent of paved surfaces that promote transport of particulate metals (Lyons et al., 2006; Eckley and Branfireun, 2008; Brigham et al., 2009).

4.

Fig. 8 e Relationship between partitioning coefficient (KD) and the fraction of annual trace-metal flux that occurs during the spring period in Wolf Creek.

Fluxes of other metals

Conclusion

Results of this study indicate that, with the exception of Hg that was increased modestly in Wolf Creek, processes affecting metal partitioning and loadings are similar among southwest Ohio streams. Relative differences in land use,

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 5 1 e5 1 6 0

catchment area, and presence or absence of WWTFs had little or no detectable effect on most trace metal concentrations and fluxes. Additionally, metal concentrations and fluxes in southwest Ohio streams are less than or within the range of those in other North American rivers that are equally or less impacted by urban development. This suggests that suburban encroachment on agricultural and undeveloped lands has either similarly or not substantially impacted trace metal cycling in streams at the urban fringe of Dayton and, by extension, other comparable metropolitan areas.

Acknowledgments We thank Geraldine Nogaro, Melissa Tabatchnick, Katlin Bowman, Lisa Romas, Matt Konkler, Robbie Weller, Will Ehresman, Jaclyn Klaus and Deepthi Nalluri for help with field sampling, sample preparation, and analyses. Geraldine Nogaro assisted with statistical analyses. Financial support was provided by Wright State University (gs1), which had no involvement in the collection, analysis, and interpretation of data.

Appendix. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.07.023.

references

Allen, H.E., Hall, R.H., Brisbin, T.D., 1980. Metal speciation: effects on aquatic toxicity. Environmental Science & Technology 14, 441e443. APHA, et al.(American Public Health Association, American Water Works Association, Water Environment Federation), 1995. Standard Methods for the Examination of Water and Wastewater, nineteenth ed. American Public Health Association, Washington, D.C. Balcom, P.H., Fitzgerald, W.F., Vandal, G.M., Lamborg, C.H., Rolfhus, K.R., Langer, C.S., Hammerschmidt, C.R., 2004. Mercury sources and cycling in the Connecticut River and long Island Sound. Marine Chemistry 90, 53e74. Balcom, P.H., Hammerschmidt, C.R., Fitzgerald, W.F., Lamborg, C. H., O’Connor, J.S., 2008. Seasonal distribution and cycling of mercury and methylmercury in the waters of New York/ New Jersey Harbor estuary. Marine Chemistry 109, 1e17. Balogh, S.J., Nollet, Y.B., Offerman, H.J., 2005. A comparison of total mercury and methylmercury export from various Minnesota watersheds. Science of the Total Environment 340, 261e270. Bloom, N., 1989. Determination of picogram levels of methylmercury by aqueous phase ethylation, followed by cryogenic gas chromatography with cold vapor atomic fluorescence detection. Canadian Journal of Fisheries and Aquatic Sciences 46, 1131e1140. Bloom, N.S., Crecelius, E.A., 1983. Determination of mercury in seawater at sub-nanogram per liter levels. Marine Chemistry 14, 49e59. Bloom, N., Fitzgerald, W.F., 1988. Determination of volatile mercury species at the picogram level by low-temperature gas

5159

chromatography with cold vapor atomic fluorescence detection. Analytica Chimica Acta 208, 151e161. Bowman, K.L., Hammerschmidt, C.R., 2011. Extraction of monomethylmercury from seawater for low-femtomolar determination. Limnology & Oceanography: Methods 9, 121e128. Brigham, M.E., Wentz, D.A., Aiken, G.R., Krabbenhoft, D.P., 2009. Mercury cycling in stream ecosystems: water column chemistry and transport. Environmental Science & Technology 43, 2720e2739. Burton Jr., G.A., Pitt, R.E., 2002. Stormwater Effects Handbook. Lewis Publishers, Boca Raton, FL. Cloran, C.E., Burton Jr., G.A., Hammerschmidt, C.R., Taulbee, K.W., Custer, K.W., Bowman, K.L., 2010. Effects of suspended solids and dissolved organic carbon on nickel toxicity. Environmental Toxicology & Chemistry 29, 1781e1787. Eckley, C.S., Branfireun, B., 2008. Mercury mobilization in urban stormwater runoff. Science of the Total Environment 403, 164e177. Engstrom, D.R., Swain, E.B., Henning, T.A., Brigham, M.E., Brezonik, P.L., 1994. Atmospheric mercury deposition to lakes and watersheds: a quantitative reconstruction from multiple sediment cores. Environmental Chemistry of Lakes and Reservoirs 237, 33e66. Exurban Change Program, 2010. http://exurban.osu.edu/index. htm (accessed 12.09.10.). Fitzgerald, W.F., Gill, G.A., 1979. Subnanogram determination of mercury by two-stage gold amalgamation and gas phase detection applied to atmospheric analysis. Analytical Chemistry 51, 1714e1720. Fitzgerald, W.F., Engstrom, D.R., Mason, R.P., Nater, E.A., 1998. The case for atmospheric mercury contamination in remote areas. Environmental Science & Technology 32, 1e7. Gill, G.A., Fitzgerald, W.F., 1985. Mercury sampling of open ocean waters at the picomolar level. Deep-Sea Research 32, 287e297. Hammerschmidt, C.R., Fitzgerald, W.F., Lamborg, C.H., Balcom, P. H., Visscher, P.T., 2004. Biogeochemistry of methylmercury in sediments of long Island Sound. Marine Chemistry 90, 31e52. Hammerschmidt, C.R., Fitzgerald, W.F., 2006. Methylmercury cycling in sediments on the continental shelf of southern New England. Geochimica et Cosmochimica Acta 70, 918e930. Hare, L., 1992. Aquatic insects and trace metals: bioavailability, bioaccumulation, and toxicity. Critical Reviews in Toxicology 22, 327e369. HydroQual, 2006. A model for the management and evaluation of concern in water, sediment, and biota in NY/NJ Harbor Estuary. Contaminant fate, and transport, and bioaccumulation sub-models (draft report). Contaminant assessment and reduction project, HRF0010. Hurley, J.P., Benoit, J.M., Babiarz, C.L., Shafer, M.M., Andren, A.W., Sullivan, J.R., Hammond, R., Webb, D.A., 1995. Influences of watershed characteristics on mercury levels in Wisconsin rivers. Environmental Science & Technology 29, 1867e1875. Lamborg, C.H., Fitzgerald, W.F., Damman, A.W.H., Benoit, J.M., Balcom, P.H., Engstrom, D.R., 2002. Modern and historic atmospheric mercury fluxes in both hemispheres: global and regional mercury cycling implications. Global Biogeochemical Cycles 16, 1104. Lamborg, C.H., Tseng, C.-M., Fitzgerald, W.F., Balcom, P.H., Hammerschmidt, C.R., 2003. Determination of mercury complexation characteristics of dissolved organic matter in natural waters through "reducible Hg" titrations. Environmental Science & Technology 37, 3315e3322. Lamborg, C.H., Fitzgerald, W.F., Skoog, A., Visscher, P.T., 2004. The abundance and source of mercury-binding organic ligands in Long Island Sound. Marine Chemistry 90, 151e163. Lawson, N.M., Mason, R.P., 2001. Concentration of mercury, methylmercury, cadmium, lead, arsenic, and selenium in the

5160

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 5 1 e5 1 6 0

rain and stream water of two contrasting watersheds in western Maryland. Water Research 35, 4039e4052. Lawson, N.M., Mason, R.P., Laporte, J.-M., 2000. The fate and transport of mercury, methylmercury, and other trace metals in Chesapeake Bay tributaries. Water Research 35, 501e515. Li, L.Y., Hall, K., Yuan, Y., Mattu, G., McCallum, D., Chen, M., 2009. Mobility and bioavailability of trace metals in the watersediment system of the highly urbanized Brunette watershed. Water Air and Soil Pollution 197, 249e266. Lyons, W.B., Fitsgibbon, T.O., Welch, K.A., Carey, A.E., 2006. Mercury geochemistry of the Scioto River, Ohio: impact of agriculture and urbanization. Applied Biogeochemistry 21, 1880e1888. Mason, R.P., Sullivan, K.A., 1998. Mercury and methylmercury transport through an urban watershed. Water Research 32, 321e330. Mason, R.P., Lawson, N.M., Lawrence, A.L., Leaner, J.J., Lee, J.G., Sheu, G.-R., 1999. Mercury in Chesapeake Bay. Marine Chemistry 65, 77e96. Mercury Deposition Network. 2010. http://nadp.sws.uiuc.edu/ mdn (accessed 09.14.10.). Mergler, D., Anderson, H.A., Chan, L.H.M., Mahaffey, K.R., Murray, M., Sakamoto, M., Stern, A.H., 2007. Methylmercury exposure and health effects in humans: a worldwide concern. Ambio 36, 3e11. Ohio Department of Natural Resources, 1999. Principal streams and their drainage areas. Division of Water, Columbus, OH. Ohio Environmental Protection Agency, 1997. Biological and water quality study of the middle and lower Great Miami River and selected tributaries, 1995 Ohio EPA Technical Report MAS/ 1996-12-8, Columbus, Ohio. Pacyna, E.G., Pacyna, J.M., Steenhuisen, F., Wilson, S., 2006. Global anthropogenic mercury emission inventory for 2000. Atmospheric Environment 40, 4048e4063. Patterson, C.C., 1965. Contaminated and natural lead environments of man. Archives of Environmental Health 11, 344e360. Peraza, M.A., Ayala-Fierro, F., Barber, D.S., Casarez, E., Rael, L.T., 1998. Effects of micronutrients on metal toxicity. Environmental Health Perspectives 106 (Suppl. 1), 203e216. Rice, G.E., Senn, D.B., Shine, J.P., 2009. Relative importance of atmospheric and riverine mercury sources to the northern Gulf of Mexico. Environmental Science & Technology 43, 415e422.

Schiefer, M.C., 2002. Basin descriptions and flow characteristics of Ohio streams. Bulletin 47. Ohio Department of Natural Resources, Columbus, OH. Shafer, M.M., Overdier, J.T., Hurley, J.P., Armstrong, D.E., Webb, D., 1997. The influence of dissolved organic carbon, suspended particulates, and hydrology on the concentration, partitioning and variability of trace metals in two contrasting Wisconsin watersheds. Chemical Geology 136, 71e97. Shafer, M.M., Overdier, J.T., Phillips, H., Webb, D., Sullivan, J.R., Armstrong, D.E., 1999. Trace metal levels and partitioning in Wisconsin rivers. Water Air and Soil Pollution 110, 273e311. Sharp, J.H., Benner, R., Bennett, L., Carlson, C.A., Fitzwater, S.E., Peltzer, E.T., Tupas, L.M., 1995. Analyses of dissolved organic carbon in seawater; the JGOFS EqPac methods comparison. Marine Chemistry 48, 91e108. Swain, E.B., Engstrom, D.R., Brigham, M.E., Henning, T.A., Brezonik, P.L., 1992. Increasing rates of atmospheric mercury deposition in midcontinental North America. Science 257, 784e787. Tabatchnick, M.D., 2010. Mercury speciation in temperate tree foliage. M.S. thesis, Wright State University, Dayton, OH. Tiefenthaler, L.L., Stein, E.D., Schiff, K.C., 2008. Watershed and land use-based sources of trace metals in urban storm water. Environmental Toxicology & Chemistry 27, 277e287. Tseng, C.-M., Hammerschmidt, C.R., Fitzgerald, W.F., 2004. Determination of methylmercury in environmental matrixes by on-line flow injection and atomic fluorescence spectrometry. Analytical Chemistry 76, 7131e7136. U.S. Department of Housing and Urban Development, 2000. The state of the cities 2000. U.S. Department of Housing and Urban Development, Washington, DC. U.S. Environmental Protection Agency, 2007. Method 6020A: Inductively coupled plasma-mass spectrometry. U.S. Environmental Protection Agency, Washington, DC. U.S. Environmental Protection Agency, 2009. Final national priorities list sites, 2009. U.S. Environmental Protection Agency, Washington, DC. U.S. Geological Survey, 2010. Real-time water data for the nation. http://waterdata.usgs.gov/nwis/rt (accessed 08.21.10.). Wiener, J.G., Krabbenhoft, D.P., Heinz, G.H., Scheuhammer, A.M., 2003. Ecotoxicology of mercury. In: Hoffman, D.J., Rattner, B.A., Burton Jr., G.A., Cairns Jr, J. (Eds.), Handbook of Ecotoxicology. CRC Press, Boca Raton, FL. World Health Organization, 2008. The world health report 2008. http://www.who.int/whr/2008/en/index.html.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 6 1 e5 1 7 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Reversible and irreversible low-pressure membrane foulants in drinking water treatment: Identification by principal component analysis of fluorescence EEM and mitigation by biofiltration pretreatment Sigrid Peldszus a,*, Cynthia Halle´ c, Ramila H. Peiris b, Mohamed Hamouda a,1, Xiaohui Jin a,1, Raymond L. Legge b, Hector Budman b, Christine Moresoli b, Peter M. Huck a,1 a

NSERC Chair in Water Treatment, Department of Civil and Environmental Engineering, University of Waterloo, 200 University Avenue West, Waterloo, Ontario N2L 3G1, Canada b Department of Chemical Engineering, University of Waterloo, Canada c Department of Hydraulic and Environmental Engineering, Norwegian University of Science and Technology (NTNU), S.P. Andersens 5, Trondheim NO-7491, Norway

article info

abstract

Article history:

With the increased use of membranes in drinking water treatment, fouling e particularly

Received 18 December 2010

the hydraulically irreversible type e remains the main operating issue that hinders

Received in revised form

performance and increases operational costs. The main challenge in assessing fouling

27 May 2011

potential of feed water is to accurately detect and quantify feed water constituents

Accepted 15 July 2011

responsible for membrane fouling. Utilizing fluorescence excitation-emission matrices

Available online 23 July 2011

(EEM), protein-like substances, humic and fulvic acids, and particulate/colloidal matter can be detected with high sensitivity in surface waters. The application of principal component

Keywords:

analysis to fluorescence EEMs allowed estimation of the impact of surface water constit-

Biofiltration

uents on reversible and irreversible membrane fouling. This technique was applied to

Ultrafiltration Fluorescence

experimental data from a two year bench-scale study that included thirteen experiments excitation

emission

investigating the fouling potential of Grand River water (Ontario, Canada) and the effect of

matrices

biofiltration pre-treatment on the level of foulants during ultrafiltration (UF). Results

Principal component analysis

showed that, although the content of protein-like substances in this membrane feed water

Protein-like substances

(¼ biofiltered natural water) was much lower than commonly found in wastewater appli-

Fouling control

cations, the content of protein-like substances was still highly correlated with irreversible fouling of the UF membrane. In addition, there is evidence that protein-like substances and particulate/colloidal matter formed a combined fouling layer, which contributed to both reversible and irreversible fouling. It is suggested that fouling transitions from a reversible to an irreversible regime depending on feed composition and operating time. Direct biofiltration without prior coagulant addition reduced the protein-like content of the

* Corresponding author. Tel.: þ1 519 888 4567; fax: þ1 519 746 7499. E-mail addresses: [email protected] (S. Peldszus), [email protected] (C. Halle´), [email protected] (R.H. Peiris), [email protected] (M. Hamouda), [email protected] (X. Jin), [email protected] (R.L. Legge), [email protected] (H. Budman), [email protected] (C. Moresoli), [email protected] (P.M. Huck). 1 Tel.: þ1 519 888 4567x33511; fax: þ1 519 746 7499. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.022

5162

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 6 1 e5 1 7 0

membrane feed water which in turn reduced the irreversible fouling potential for UF membranes. Biofilters also decreased reversible fouling, and for both types of fouling higher biofilter contact times were beneficial. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Membrane filtration in drinking water treatment has gained considerable momentum in the past decade. However, hydraulically irreversible fouling remains an important issue as it increases chemical cleaning frequency and operational costs while reducing membrane life time. Various fractions of organic matter such as biopolymers, polysaccharides, proteins, humic substances and also colloidal/particulate matter have been associated with UF membrane fouling (e.g. Amy, 2008; Jucker and Clark, 1994; Aoustin et al., 2001; Howe and Clark, 2002; Kimura et al., 2004; Halle´ et al., 2009; Peiris et al., 2010a), but a clear understanding of the role of these foulants during surface water filtration is still lacking. An indication for the complexity of membrane fouling is the recently reported synergistic effects of interactions between model foulants, i.e. inorganic colloids and polysaccharides or humic acids (Jermann et al., 2008). Transferring model solution results to actual surface water filtration remains a challenge, since model solutions can not truly capture the complex nature of natural organic matter and natural colloids (Buffle et al., 1998). Results from a multitude of wastewater fouling studies identifying wastewater constituents responsible for reversible and irreversible fouling vary widely and are in part contradictory (e.g. Drews, 2010). Besides, results from wastewater studies are only partially applicable to surface water filtration as these water matrices differ not only in the characteristics of their constituents but also in concentrations. Fouling studies involving natural water are challenging in part due to a lack of adequate analysis techniques for measuring specific organic matter fractions at the concentrations encountered in surface water. For example, spectrophotometric methods employed for polysaccharides or proteins in wastewater are not sensitive enough (Frølund et al., 1996; Drews, 2010). Liquid chromatography/organic carbon detection (LC/OCD) and fluorescence excitation and emission matrices (EEM) are the most promising techniques currently available for determining low levels of organic substances in water. LC/OCD quantifies biopolymers, which are comprised of polysaccharides and proteins (Huber et al., 2011), humic substances and some lower molecular weight fractions, whereas fluorescence EEM differentiates between humic and fulvic acid-like substances, protein-like substances, and colloidal/particulate matter with high sensitivity (e.g. Henderson et al., 2009; Her et al., 2003; Kimura et al., 2004; Liu et al., 2007). To capture the variability and seasonal changes in water quality it is important to analyze the full fluorescence EEMs as opposed to individual peaks (Sierra et al., 2005; Peiris et al., 2008). Principal component analysis (PCA) has been successfully employed for this purpose (Peiris et al., 2010a,b). Rapid biofiltration without prior coagulation has been demonstrated to be an effective, chemical-free, and robust

pre-treatment method that can reduce hydraulically reversible and irreversible membrane fouling by reducing the biopolymer concentration in the membrane feed (Halle´ et al., 2009). Reversible fouling increased proportionally with increased biopolymer concentration in the feed but for irreversible fouling such a correlation could not be established. It was hypothesized that biopolymer composition rather than absolute biopolymer concentration was critical for hydraulically irreversible fouling (Halle´ et al., 2009). Fluorescence EEM is able to capture a component of the biopolymers i.e. protein-like substances and should therefore be suitable for elucidating irreversible membrane fouling (Peiris et al., 2010a, b). The objective of this study was to identify surface water constituents responsible for hydraulically reversible and irreversible fouling of a polymeric UF membrane using fluorescence EEM. Moreover, reduction in irreversible fouling was to be related to changes in feed water quality after biofiltration pre-treatment. Fluorescence EEMs of membrane feed (i.e. biofilter effluent) and permeate were evaluated by using the entire fluorescence spectra for PCA.

2.

Material and methods

2.1.

Source water

Grand River water (GRW), southwestern Ontario, Canada, was used as source water the quality of which varied seasonally (e.g. TOC ranged from 5.3 to 7.7 mg C/L). Source water quality was also impacted by upstream discharge from a municipal sewage treatment plant and hence, biopolymers were detected throughout all seasons. Detailed data were provided in Halle´ et al. (2009).

2.2. Pretreatment: roughing filtration followed by biofiltration GRW was first filtered through an upflow roughing filter to reduce peak concentrations of suspended material and turbidity. The roughing filter effluent served as feed for two parallel biofilters operated in downflow mode at 5 m/h. The dual media biofilters (anthracite over sand) had empty bed contact times (EBCT) of 5 min for biofilter 1 (B1) and 14 min for biofilter 2 (B2). Further details can be found in Halle´ et al. (2009).

2.3.

Ultrafiltration membrane

This study set out to investigate membrane fouling and mitigation of fouling by biofiltration pre-treatment under conditions relevant to the drinking water industry. Therefore a bench-scale membrane system was chosen which simulated GE (formerly Zenon) UF membranes which have found

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 6 1 e5 1 7 0

widespread use in the drinking water industry throughout North America and eles where. By opting for this applied approach, membrane properties and system configuration were predefined as follows. A bench-scale UF membrane unit containing commercial hollow fiber PVDF membranes (molecular weight cut-off 400 kDa; membrane surface area 470 cm2; GE formerly Zenon, Oakville, Ontario, Canada) was operated in dead-end mode at a constant, temperature adjusted permeate flux corresponding to 57.5 L/m2h at 20  C. Fouling was monitored by measuring the increase in transmembrane pressure (TMP). The filtration cycle was fully automated and included a backpulse (i.e. backwash) with air sparging, partial draining and refilling of the tank resulting in an overall recovery of 94%. Experiments with B1 and B2 were performed sequentially since only one UF unit was available and hence, only one biofilter at a time could be used as pretreatment for UF. A typical experiment ran for 5 days. The manufacturer’s recommended maximum operating TMP (9 psi ¼ 62 kPa) was reached under severe fouling conditions which is referred to later as a high fouling event. The membrane module was chemically cleaned after each experiment and cleaning efficiency was confirmed by clean water permeability tests. A new UF membrane module was used for each season and in each season several 5 day experiments were performed. Further details are given in Halle´ et al. (2009).

2.4. Quantification of hydraulically reversible and irreversible fouling Irreversible fouling was calculated as the TMP difference between the start and the end of each experiment (i.e. after 5 days) by subtracting the initial TMP of the last cycle of operation from the initial TMP of the first cycle of operation and expressed as DTMPirr. It was not possible to determine DTMPirr reliably after short run times since the incremental increase in TMP due to irreversible fouling after for example 1 h filtration was too small to be measured reliably. Hydraulically reversible fouling was calculated as the average of 3 cycles, i.e. one before, during and after sampling, by subtracting the TMP at the beginning of a cycle from the TMP at the end of a cycle just prior to backwash.

2.5.

Fluorescence analysis

Fluorescence EEMs of the UF feed (i.e. biofilter effluents sampled before entering the membrane tank) and UF permeate, were acquired using a Varian Cary Eclipse Fluorescence Spectrophotometer (Palo Alto, CA) collecting 301 individual emission intensity values (within the 300e600 nm emission range) at sequential 10 nm increments of excitation wavelengths between 250 nm and 380 nm. Fluorescence spectra for ultrapure water obtained with the same instrumental parameters were subtracted from all spectra. Sample temperature during measurement was maintained at w22  C and sample pH was not adjusted, the pH ranged from 7.3 to 8.4. Fluorescence EEMs were obtained from 13 biofiltration/UF experiments at different filtration time intervals (i.e. 1, 24, 48 and 96 h; 128 samples in total). Samples were stored at 4  C until they were analyzed on the same day they were taken. A detailed description of the procedure can be found in Peiris

5163

et al. (2010a, b). Inner filtering and fluorescence quenching effects, which can be caused by high concentrations of organic material, metal ions or large differences in temperature or pH (Henderson et al., 2009), were considered to be negligible. Hence no corrections to fluorescence EEM intensity readings were made to account for these effects. The total organic carbon concentrations for samples in this study ranged from 5.3 to 7.7 mg C/L which was well below the limit of 25 mg C/L put forward by Hudson et al. (2008) as the lower limit at which these effects were observed. Metal ion concentrations were also low and pH and temperature were in a very narrow range throughout the entire study.

2.6. Fluorescence data pre-treatment and principal component analysis Data pre-treatment and PCA are described in detail in Peiris et al. (2010b). The initial data matrix (X) contained 128 rows (the water samples), and 4214 columns (intensity values of the corresponding fluorescence EEMs). PCA has been shown to be a valuable technique capable of analyzing multiple fluorescence EEMs over their entire excitation and emission range (Peiris et al., 2010a). Hence, these data were then analyzed by PCA thus extracting and visualizing the information residing in this initial multivariate data matrix. PCA resulted in three new variables, principal components (PCs), which were mutually independent and captured the major pattern of the original data matrix X (e.g. Eriksson et al., 2001). Resulting score plots and loading plots are provided in Peiris et al. (2010b). The three major PCs were found to be related to the major foulant fractions present in the feed water (loading plots in Peiris et al., 2010b). The PC1 loading plot showed peaks at similar locations to that of characteristic fluorescence EEM peaks of humic substances. Both humic (Excitation wavelength (Ex)/Emission wavelength (Em) 270nm/460 nm) and fulvic acids (Exc/Em 320nm/415 nm) were present in similar proportions as in the original Grand River water EEM, albeit shifted, whereas PC2 was dominated by the arrays of peaks located at similar locations (ranged from Ex/Em w 300 nm/300 nm to Ex/Em w 380 nm/380 nm and Ex/Em w 260nm/520 nm to Ex/Em w 300nm/600 nm) to that of Raleigh scattering regions, which were found to provide information related to colloidal/particulate matter present in water (Peiris et al., 2010b). The loading plot of PC3 displayed a negative peak at the location that corresponded to the fluorescence EEM peak (at Ex/Em w 280nm/330 nm) related to protein-like substances. This negative PC3 loading peak indicated that the PC3 score values are inversely proportional to the protein content in water, i.e. higher PC3 scores corresponds to lower protein content. In this paper, the term protein will be used instead of protein-like substances for PC3 with the understanding that its actual nature is comprehensively described by its loading plot (Peiris et al., 2010b). Score values for PC1, PC2 and PC3 were calculated for each water sample. LC/OCD analysis (data not shown) corroborates that for samples where the score values for PC1 and PC2 increased the corresponding LC/OCD fraction also increased. However, PC3 showed an inverse relationship meaning that PC3 score values decreased as the protein content increased. Hence, the membrane fouling behavior

5164

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 6 1 e5 1 7 0

could be evaluated using the PC scores as semi-quantitative indicators of the different foulant fractions. The calculation of the accumulated/rejected foulants at 1 h and at 24 h is based on the assumption that the difference in PC scores between feed (i.e. biofilter effluent ¼ influent membrane feed tank) and permeate (i.e. PCaccumulated/ rejected ¼ (PCfeed e PCpermeate)) is proportional to the amount of a specific foulant component being rejected at that point in time. Foulant components that are being rejected and contained in the membrane tank until it is drained, are in principal available as foulant mass for accumulation on the surface and/or within the pores of the membrane. It follows that the higher the rejection, the higher the foulant content available in the tank, and therefore the higher are the chances for actual accumulation of foulants on the surface and/or within the pores. Although the term foulant accumulated/ rejected is used throughout this manuscript, it could be interpreted as foulant component available for accumulation. This paper focuses on characterizing biofilter effluents which served as feed for the UF membranes. When examining the PC scores of these effluents, conclusions could be drawn with regards to the role that the different foulants played in hydraulically reversible and irreversible membrane fouling which are from here on termed reversible and irreversible fouling. Simultaneously the impact of biofiltration pre-treatment on specific foulant fractions could be assessed at two EBCTs (5 min (B1), 14 min (B2)) and related back to both types of fouling. Thirteen biofiltration/UF experiments were performed over 2 years giving a comprehensive data set covering all seasons. Experiments were run for 5 days with the exception of runs with high fouling events (n ¼ 6) where experiments had to be stopped after as little as 2 h when the maximum operating pressure of the membrane unit (9 psi) was reached.

3.

Results and discussion

3.1.

Impact of protein content on irreversible fouling

Proteins played an important role in the irreversible fouling of the UF membranes used in this study. From Fig. 1a it is evident that under normal operating conditions, excluding the high fouling events, there is a strong positive correlation between irreversible fouling and protein content. This strong correlation is remarkable since experiments were performed over a 2 year period with fluctuating raw water qualities similar to ones experienced at a full-scale plant. To maintain a constant flux, the operating TMP had to be increased proportionally as protein content in the UF feed increased (note that PC3 score and protein content are inversely related). This has not been reported previously for longer term UF membrane operations with surface water where organic carbon concentrations are much lower than typically encountered in wastewater. These findings are in agreement with results from short term, crossflow, flat sheet UF membrane experiments with the same surface water (Peiris et al., 2010b). Although there were substantial differences between the current study and these flat sheet experiments in terms of UF membrane characteristics (molecular weight cut-off, membrane material and

Fig. 1 e Impact of protein on irreversible fouling of UF membranes. Irreversible fouling quantified by TMP difference (DTMP [ TMPend of exp. e TMPstart of exp.) vs. (a) protein content of membrane feed, and (b) protein accumulation/rejection; 1 h: samples taken after 1 h of filtration; 24 h>: samples taken after 24 h or more of filtration; B1 [ biofilter 1 (EBCT 5 min); B2 [ biofilter 2 (EBCT 14 min); PC3B-effluent [ Biofilter effluent [ UFFeed; PC3Permeate [ UFEffluent.

hydrophobicity), flow regime and duration, in both cases protein was identified as an important factor contributing to irreversible fouling. This is also consistent with previous LC/ OCD results by Halle´ et al. (2009) who concluded that biopolymer composition rather than overall biopolymer concentration may be of higher relevance for irreversible fouling. From Fig. 1b it can be concluded that irreversible fouling is related to the accumulation/rejection of proteins since again strong correlations were observed. However, the exact fouling mechanism and the role of proteins during irreversible fouling in this surface water needs to be elucidated further. Proteins may be able to adsorb onto the membrane, a hypothesis which is supported by findings in filtration studies of proteins where strong interaction between proteins and

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 6 1 e5 1 7 0

membranes are known to exist during protein separation with membrane filtration systems. Similarly Her et al. (2007) reported accumulation of proteins on membranes. Overall, based on Fig. 1 it is apparent that proteins play an important role in irreversible fouling of the UF membranes utilized. Protein content in the membrane feed (Fig. 1a) and protein accumulation/rejection (Fig. 1b) both show a strong correlation with the extent of irreversible fouling with samples taken after 1 h and also taken after longer periods of time (24 h). The similarity of PC3 scores after 1 h and 24 h (Fig. 1a) shows that the protein content in the feed did not change substantially between sampling after 1 h and 24 h. The similarity in the extent of the correlations and the slopes in Fig. 1b indicate that protein accumulation/rejection leading to irreversible UF fouling did also not change substantially between 1 h and 24 h of filtration under normal operating conditions excluding high fouling events. This similarity implies the possibility of assessing and predicting the fouling potential of surface water based on protein content as assessed by fluorescence EEM during early stages of filtration, e.g. after 1 h. This finding is complementary to previous results (Peiris et al., 2010b) which were based solely on assessing PC score plots at different treatment stages and did not involve detailed analysis of reversible and irreversible membrane fouling. Experiments were defined as high fouling events when the maximum operating pressure of 9 psi was reached at which point further operation of the membrane unit was not possible. This occurred in 6 out of 13 experiments using either B1 or B2 effluents. For these events there was no significant correlation between irreversible fouling and the protein content in the UF feed (Fig. 1a) after 1 h of filtration as the upper boundary in terms of operating pressure had been reached. However, for these high fouling events there is a slight downward trend in the ΔTMP data after 24 h (Fig. 1b). Further discussion delineating high fouling events from experiments under sustainable operating conditions can be found in section 3.6.

3.2.

Impact of protein content on reversible fouling

Protein content in the membrane feed (Fig. 2a) was also linked albeit weakly to reversible fouling for sampling at 1 h and 24 h. This was surprising since proteins have in general been associated with irreversible fouling in wastewater investigations (e.g. Drews, 2010). However, the trends for relating protein content (Fig. 2a) to reversible fouling are by far weaker than those observed for irreversible fouling (Fig. 1a). Similarly, accumulation/rejection of protein (Fig. 2b) showed an increasing trend with an increase in reversible fouling after 1 h and 24 h of filtration. These trends were again much weaker than those observed for irreversible fouling (Fig. 1b). Those weak trends imply that proteins played a different role during irreversible fouling compared reversible fouling. From the weak correlations it may be inferred that other foulants in addition to protein were probably participating in the formation of a reversible, i.e. backwashable, fouling layer. Results for the other foulant components investigated in this study support the concept of the formation of a combined fouling layer probably involving colloidal/particulate matter rather than humic substances (see also section 3.6 combined fouling discussion).

5165

Fig. 2 e Impact of protein on the reversible fouling of UF membranes. Reversible fouling quantified by the TMP difference within a filtration cycle (average of 3 cycles i.e. before, during and after sampling) vs. (a) protein content of membrane feed, and (b) protein accumulation/rejection; 1 h: samples taken after 1 h of filtration; 24 h>: samples taken after 24 h or more of filtration; B1 [ biofilter 1 (EBCT 5 min); B2 [ biofilter 2 (EBCT 14 min); PC3B-effluent [ Biofilter effluent [ UFFeed; PC3Permeate [ UFEffluent.

The difference in reversible fouling for the same protein content (Fig. 2a) and the same amount of protein accumulated/rejected (Fig. 2b) after 1 and 24 h filtration time implies that the characteristics of the reversible, i.e. backwashable, fouling layer was changing over time, and that protein remained an important component. Fig. 2a indicates that for the same protein content in the feed, the resulting foulant layer was easier to backwash after 1 h of filtration than after >24 h. Similarly Fig. 2b shows that a higher DTMP was required to backwash the same amount of protein accumulated/rejected after 24 h compared to 1 h filtration time. When following the evolution of the protein content in the UF feed over time in individual experiments, it was observed that protein content in the feed fluctuated but that there was no discernable trend either up or down (detailed data not shown). However, protein content in the permeate decreased from 1 to 24 h filtration time in all experiments except for one

5166

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 6 1 e5 1 7 0

(detailed data not shown) thus indicating that protein rejection increased over time.

a

1.800

filtrationtime 1h 24h >

1.600

The colloidal/particulate content in the UF feed did not show any correlation with irreversible fouling (results not shown). This was expected since it has been well documented (e.g. Jermann et al., 2008). Similarly, no correlation was observed between colloidal/particulate accumulation/rejection and irreversible fouling. From Fig. 3 it is apparent though that the colloidal and particulate accumulation/rejection for high fouling events spanned a slightly higher range than is the case for B1 and B2 effluent experiments under sustainable operating conditions.

1.400

Reversible fouling increasing

3.3. Impact of colloidal and particulate matter on irreversible fouling

y = 0.0433x + 1.2871 R² = 0.34

1.000 0.800 0.600

y = 0.0457x + 0.7507 R² = 0.47

0.400 0.200 0.000 -10

-8

-6

-4

-2

0

2

4

6

8

(PC2 score) Colloidal /particulate content increasing

b

3.4. Impact of colloidal/particulate content on reversible fouling

1.8 1.6

ΔTMP (psi) Reversible fouling increasing

In terms of the effect of colloid/particulate matter on reversible fouling, results were similar to those observed for protein. Colloidal/particulate content and also their accumulation/ rejection showed increasing trends with respect to reversible fouling after 1 h and >24 h filtration times (Fig. 4a and b). Once more, trends were not very distinct with R2 values similar to the ones observed for protein. The fact though that colloid/ particulate trends were observed for reversible fouling (Fig. 4) but not for irreversible fouling (Fig. 3) is consistent with the literature where generally colloid/particulate matter is associated with reversible, i.e. backwashable, fouling (e.g. Howe and Clark, 2002; Peiris et al., 2010a). However, since only

1.200

1.4 1.2

y = 0.0174x + 1.0803 R² = 0.11

1 0.8 0.6

y = 0.04x + 0.4454 R² = 0.56

0.4

filtration time 1h 24h >

0.2 0 -5

0

5

10

15

20

25

30

(PC2B-effluent - PC2Permeate) Colloidal/particulate accumulation/rejection increasing

Fig. 4 e Impact of the colloid/particulate matter on the reversible fouling of UF membranes. Reversible fouling quantified by the TMP difference within a filtration cycle (average of 3 cycles i.e. before, during and after sampling) vs. (a) colloidal/particulate content of membrane feed, and (b) colloidal/particulate accumulation/rejection in the membrane. 1 h: samples taken after 1 h of filtration; 24 h>: samples taken after 24 h or more of filtration; B1 [ biofilter 1 (EBCT 5 min); B2 [ biofilter 2 (EBCT 14 min); PC2Beffluent [ Biofilter effluent [ UFFeed; PC2Permeate [ UFEffluent.

Fig. 3 e Impact of colloidal/particulate accumulation/ rejection on irreversible fouling of UF membranes. Irreversible fouling quantified by TMP difference (DTMP [ TMPend of exp. e TMPstart of exp.) vs. protein accumulation/rejection; 1 h: samples taken after 1 h of filtration; 24 h>: samples taken after 24 h or more of filtration; B1 [ biofilter 1 (EBCT 5 min); B2 [ biofilter 2 (EBCT 14 min); PC2B-effluent [ Biofilter effluent [ UFFeed; PC2Permeate [ UFEffluent.

weak trends were observed colloid/particulate matter alone could not be directly responsible for reversible fouling. Other foulants may have participated in the formation of a combined reversible fouling layer which, based on the above results, may also have contained proteins. Similar to the results for protein in section 3.2, correlations at 1 and 24 h filtration times were offset indicating that the fouling layer for the same colloid/particulate content was easier to backwash after 1 h compared to 24 h (Fig. 4a). Also, the same amount of colloidal/particulate accumulated/rejected (Fig. 4b) required higher DTMPs after 24 h compared to 1 h filtration time. Overall this again points to a change in the reversible, i.e. backwashable, fouling layer over time to which colloidal/particulate foulants seem to contribute.

5167

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 6 1 e5 1 7 0

30 filtration time 1h

25

(PC2B-effluent - PC2permeate)

Colloidal/particulate accumulation/rejection

Humic substances have been identified as a major contributor to irreversible fouling of UF membranes (Jucker and Clark, 1994; Jones and O’Melia, 2001; Peiris et al., 2010a), albeit for tighter membranes than the ones used in this study (e.g. Peiris et al., 2010a used UF membranes with an MWCO of 20 and 60 kDa). Hence, significant correlations were not observed between humic substance content in the feed and reversible or irreversible fouling (results not shown) for the UF membrane used in this study (MWCO 400 kDa). Accumulated/ rejected humic substances did also not show any correlations with reversible or irreversible fouling (data not shown). The similarity of the PC1 score values for humic substances, and LC/OCD results (data not shown) for feed and permeate both indicate that most of the humic substances passed through the 400 kDa UF membrane used in this study. In addition, the non existing correlations to reversible and irreversible fouling seem to point to little or no interactions between humic substances on the one hand and colloidal/particulate matter or protein on the other hand. Such an interaction would have assisted in the retention of humic substances on the membranes and hence, would have contributed to reversible and/or irreversible fouling.

a

24h >

normal operation

1h

24h >

high fouling events

20 15 10 5 0 0

5

10

15

20

25

- (PC3B-effluent - PC3 permeate) Protein accumulation/rejection

b

2.5 filtration time 1h

2.0

Feed turbidity (NTU)

3.5. Contribution of humic substance-like matter to UF membrane fouling

24h >

normal operation

1h

24h>

high fouling events

1.5

1.0

0.5

3.6. Irreversible fouling by combined protein and colloidal/particulate content Fig. 5a is an indirect analysis of the accumulated/rejected foulants as they pertain to normal operation and to high fouling events as depicted in Figs. 1 and 3. The composition of the accumulated/rejected foulants at a particular point in time was determined by subtracting the PC scores of the permeate from the PC scores for the biofilter effluent which served as influent to the membrane feed tank. This difference corresponded to the amount of foulant rejected, part of which was likely to accumulate either on the membrane surface or within membrane pores. Based on Fig. 5a, protein clearly is a key factor leading to high fouling events which are characterized by accumulation/rejection of protein above a threshold value of 4. All experiments which experienced high fouling (Fig. 5a; open symbols) showed a high amount of accumulated/rejected protein (>4) and also higher, yet strongly varying amounts of accumulated/rejected colloidal/ particulate matter. Composition of the accumulated/rejected foulants for experiments under normal operating conditions (Fig. 5a; solid symbols)differed distinctly in that the protein accumulation/rejection was lower (<4) but the accumulated/ rejected colloidal/particulate amount was again highly variable. These results emphasize the key role protein is playing in the fouling process and this is consistent with the correlation observed between irreversible fouling and protein content in the feed (Fig. 1a). The influence of membrane feed characteristics on high fouling events is depicted in Fig. 5b. Turbidity was chosen over colloidal/particulate score in part since it is a very accessible water quality parameter and in part because turbidity data were more tightly grouped compared to colloidal/particulate score values when plotting against protein scores. It is evident

0.0 -10

-5

0 5 Protein score (PC3) increasing

10

15

Protein content increasing

Fig. 5 e The role of protein and colloidal/particulate matter in high fouling events and under normal membrane operating conditions. (a) Composition of accumulated/ rejected foulants, (b) composition of membrane feed water i.e. biofilter effluents. 1 h: samples taken after 1 h of filtration; 24 h>: samples taken after 24 h or more of filtration.

that a high protein content (i.e. PC3 score < 0) in the feed led to high fouling events regardless of turbidity. However, when the protein content was lower (PC3 score >0) turbidity was the deciding factor for the occurrence of high fouling events which occurred once a threshold of 0.45 NTU was exceeded. These results point to potential interactions between protein and colloidal/particulate matter in that both may have contributed to the formation of a combined irreversible fouling layer especially under high fouling conditions. This is supported by results from recent fouling investigations involving the combination of model solutions of either carbohydrates or humic acids and inorganic colloids (Jermann et al.; 2008) which reported the formation of combined fouling layers and synergistic effects. A separate study involving UF of GRW using flat sheet cross-flow set-up also strongly indicated the potential interaction between colloidal/particulate and protein matter and its contribution towards membrane fouling (Peiris et al., 2011). Interactions between protein and inorganic colloid model solutions have also been investigated

5168

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 6 1 e5 1 7 0

in a high pressure membrane fouling study where interactions and synergistic effects on fouling were observed (Contreras et al.; 2009). The formation of a combined protein/ colloid fouling layer (Fig. 5a) is also consistent with results by Howe and Clark (2002). Based on the sequential fractionation of 5 different surface waters they identified very small colloids as key foulants in natural water which were comprised of organic and inorganic material. Mechanistically it may be postulated that protein and colloidal/particulate matter may aggregate as a consequence of concentration polarisation thus facilitating the formation of a combined fouling layer. Alternatively, naturally occurring colloids composed of inorganic colloids and rigid or flexible biopolymers may already be present in this surface water (Buffle et al., 1998) and contribute to a combined fouling layer. Detailed mechanistic studies, which were beyond the scope of this natural water filtration study, are necessary to further elucidate detailed fouling mechanisms. A complex, combined fouling phenomenon also explains why turbidity alone was not directly correlated to irreversible fouling (data not shown) as was expected. Although turbidity did play a role in the formation of a combined fouling layer and contributed to high fouling events, irreversible fouling under normal operating conditions was largely controlled by protein content (Fig. 1) if a certain feed water turbidity was not exceeded (Fig. 5b). Previous work (Halle´ et al., 2009) found that biopolymer concentrations were not directly correlated to irreversible fouling, but it was postulated that biopolymer composition must play an important role in irreversible fouling. The work presented here confirms this hypothesis in that proteins, which are detected together with polysaccharides in the biopolymer peak in the LC/OCD, were responsible for irreversible fouling of the membrane under normal operating conditions and proteins also played a key role under high fouling operating conditions.

3.7. Reversible fouling by combined protein and colloidal/particulate content None of the water constituents measured showed a strong direct correlation with reversible fouling which would have supported the assumption that a single foulant component was governing reversible fouling. Instead, both protein content and colloidal/particulate content seem to be involved in reversible fouling based on the trends shown in Figs. 2 and 4. The TMP required to backwash the foulant layer for the same protein or colloidal/particulate content in the feed (Figs. 2a and 4a) increased from 1 to >24 h filtration time. The same held true for accumulated/rejected protein and accumulated/rejected colloidal/particulate matter (Figs. 2b and 4b). This points to a combined, but this time reversible, i.e. backwashable, fouling layer where proteins interact with particles. This conclusion is also supported by reports that natural colloids can be of rather complex nature and that both organic and inorganic components contribute to their formation (Buffle et al., 1998). Also, based on the feed water quality identified from Fig. 5b, it may be concluded that for normal membrane operations either protein or a combination of protein and colloidal/particulate matter were not available in sufficient quantity to cause substantial irreversible fouling which otherwise would have

lead to high fouling events. However, as mentioned above, over time the character of the reversible fouling layer changed as it became more resistant to removal by backwashing. Similarly, as protein content in the feed increased, the fouling layer required higher pressures to be backwashed. In both cases it may be hypothesized that a reversible combined fouling layer was transitioning to an irreversible combined fouling layer. The exact mechanisms of the proposed interactions between colloids/particulates and proteins and the membrane remains to be elucidated. A possible explanation though may be that proteins first attached loosely to particles forming backwashable aggregates. These aggregates may have then transitioned either with time or with increased protein content in the feed into less backwashable aggregates where the protein portion of the aggregate may have attached irreversibly to the membrane surface.

3.8. Mitigation of irreversible and reversible fouling by chemically unassisted biofiltration Direct biofiltration as a pre-treatment for UF had a positive impact on membrane operations in terms of reversible and irreversible fouling. The biofilter with the longer contact time (B2) produced effluents with lower protein content than B1, presumably by either adsorbing or degrading the protein (Figs. 1a and 2a). Protein accumulation on the membrane was also lower with B2 effluent as feed compared to B1 (Figs. 1b and 2b). Fig. 1a shows that biofiltration is able to reduce irreversible fouling by reducing protein content in the feed since B2 effluents which had the longer contact time, caused less irreversible fouling than B1 effluents. Similarly, reversible fouling was reduced by biofiltration as indicated by less reversible fouling when using B2 effluent as feed compared to B1 (Fig. 2a and b). The weaker correlations for reversible fouling can be explained by the complex evolution of a combined fouling layer, as described above, where protein content seems to have a central role. Although not indicated in Fig. 4a and b, direct biofiltration also reduced colloidal/particulate content since the colloid/ particulate score was lower for B2 than B1 for the majority of the data points. This is consistent with Halle´ et al. (2009) who reported a reduction in turbidity by direct biofiltration. The lower colloidal/particulate content for B2 was associated with lower reversible fouling for the majority of the B2 effluents compared to B1 (data not shown). This may be explained by the participation of colloidal/particulate matter in the formation of a combined backwashable fouling layer. Reduction of reversible fouling by direct biofiltration may therefore be attributed to reduction in protein content and colloidal/ particulate content. Reduction in reversible fouling by direct biofiltration pre-treatment has been reported previously (Halle´ et al., 2009) and was linked to biopolymer concentration which are partially composed of protein (Huber et al., 2011).

3.9. Practical implications e prediction of irreversible fouling and irreversible fouling mitigation This and previous work (Peiris et al., 2010b) have shown that PCA analysis of fluorescence EEM data is a promising methodology for investigating and potentially predicting

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 6 1 e5 1 7 0

irreversible membrane fouling. For the biofiltration/UF treatment of Grand River water investigated here, a range of water qualities for sustainable UF membrane operations could be defined based on turbidity values and protein scores of the UF membrane feed (Fig. 5b). For this sustainable operating range, it should be possible to develop models for predicting the degree of irreversible fouling based on the strong correlation between protein content in the feed and irreversible fouling (Fig. 1a). Provided that similar relationships as depicted in Figs. 1a and 5b can be established for other surface waters this methodology has the potential to be extended into practice. Once a dataset for a particular water membrane combination has been acquired, for example through piloting, the feasibility of utilizing a certain membrane for a particular water source can be established. Optimization of operational conditions to account for changes in irreversible fouling due to fluctuating protein content throughout the seasons may also be possible. However, further research is necessary to accomplish these goals. The more immediate step would be confirmation at pilot scale. Given the above results, pre-treatment for low pressure membranes should target a reduction in protein content in the membrane feed in order to reduce, in particular, irreversible membrane fouling. This study and previous work (Halle´ et al., 2009) have shown that direct biofiltration pretreatment without prior coagulation holds the promise of being a feasible alternative to other pre-treatment technologies currently applied. In particular, direct biofiltration was able to reduce protein content which played a key role in irreversible fouling, thus making water sources impacted by upstream wastewater treatment plant discharges amenable to membrane filtration, and thereby decreasing energy consumption as lower TMPs may be used. Again confirmation at pilot scale and assessing feasibility with other waters are desirable next steps.

4.

Conclusions

Surface water impacted by upstream sewage discharge was pretreated using biofiltration without prior coagulation. Biofilter effluents were fed to a bench-scale, hollow fiber UF unit. Fluorescence EEMs obtained for UF feed (i.e. biofilter effluents) and permeate were evaluated with principal component analyses. The main conclusions relating membrane fouling to feed water quality are:  Protein content in the surface water feed was highly correlated with irreversible fouling of the polymeric UF membrane under normal operating conditions.  Both colloidal/particulate matter and protein content in the surface water feed contributed to reversible and irreversible fouling of the UF membrane, likely by forming a combined fouling layer. It is suggested that this fouling layer transitions from a reversible to an irreversible fouling layer depending on feed composition and operating time. Protein content in the feed seems to have a key role in the formation of the combined fouling layer.  Direct biofiltration without prior coagulant addition decreased the fouling potential for UF membranes by

5169

reducing protein content in the UF feed. Higher biofilter contact times resulted in lower protein content and thus lower irreversible fouling. Reversible fouling was also decreased by biofiltration pre-treatment.  Boundaries for high fouling events could be defined based on turbidity and protein PC scores of the feed. Upon confirmation for other waters, this may allow the prediction of the fouling potential of a particular surface water based on its turbidity and protein PC score.  PCA of fluorescence EEM has been proven to be a useful methodology making evaluation of membrane foulants in natural water possible. It is the only method currently available to provide a reliable estimate of protein content in surface water.  Humic substances did not contribute to reversible or irreversible fouling and were not found to be important in the formation of the combined fouling layer, although this may be different for other, e.g. tighter, UF membranes than the one used in this study.

Acknowledgements We acknowledge a number of contributors to this work including GE for the donation of UF modules, and the financial support of the Canadian Water Network, the Natural Sciences and Engineering Research Council of Canada (NSERC) including an NSERC Postgraduate scholarship to R.H. Peiris and the partners of the NSERC Industrial Research Chair in Water Treatment (P.M. Huck) for funding. The current Chair partners may be found at http://www.civil.uwaterloo.ca/watertreatment/. C. Halle´ was a Ph.D. student at the NSERC Chair in Water Treatment during the time that data for this paper was collected.

references

Amy, G., 2008. Fundamental understanding of organic matter fouling of membranes. Desalination 231 (1e3), 44e51. Aoustin, E., Scha¨fer, A.I., Fane, A.G., Waite, T.D., 2001. Ultrafiltration of natural organic matter. Separat. Purif. Technol. 22-23, 63e78. Buffle, J., Wilkinson, K.J., Stoll, S., Filella, M., Zhang, J., 1998. A generalized description of aquatic colloidal interactions: the three-colloidal component approach. Environ. Sci. Technol. 32, 2887e2899. Contreras, A.E., Kim, A., Li, Q., 2009. Combined fouling of nanofiltration membranes: mechanisms and effect of organic matter. J. Membr. Sci. 327, 87e95. Drews, A., 2010. Membrane fouling in bioreactors Characterizations, contradictions, causes and cures. J. Membr. Sci. 363, 1e28. Eriksson, L., Johansson, E., Kettaneh-Wold, N., Wold, S., 2001. Multi- and megavariate data analysis, principles and applications. Umetrics Academy, Umea, Sweden, ISBN 91973730-1-X, p. 533. Frølund, B., Palmgren, R., Keiding, K., Nielsen, P.H., 1996. Extraction of extracellular polymers from activated sludge using a cation exchange resin. Water Res. 30, 1749e1758.

5170

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 6 1 e5 1 7 0

Halle´, C., Huck, P.M., Peldszus, S., Haberkamp, J., Jekel, M., 2009. Assessing the performance of biological filtration as pretreatment to low pressure membranes for drinking water. Environ. Sci. Technol. 43 (10), 3878e3884. Henderson, R.K., Baker, A., Murphy, K.R., Hambly, A., Stuetz, R.M., Khan, S.J., 2009. Fluorescence as a potential monitoring tool for recycled water systems: a review. Water Res. 43 (4), 863e881. Her, N., Amy, G., McKnight, D., Sohn, J., Yoon, Y., 2003. Characterization of DOM as a function of MW by fluorescence EEM and HPLC-SEC using UVA, DOC, and fluorescence detection. Water Res. 37 (17), 4295e4303. Her, N., Amy, G., Plottu-Pecheux, A., Yoon, Y., 2007. Identification of nanofiltration membrane foulants. Water Res. 41 (17), 3936e3947. Howe, K.J., Clark, M.M., 2002. Fouling of microfiltration and ultrafiltration membranes by natural waters. Environ. Sci. Technol. 36 (16), 3571e3576. Huber, S.A., Balz, A., Abert, M., Pronk, W., 2011. Characterisation of aquatic humic and non-humic matter with size-exclusion chromatography - organic carbon detection - organic nitrogen detection (LC-OCD-OND). Water Res. 45 (2), 879e885. Hudson, N., Baker, A., Ward, D., Reynolds, D.M., Brunsdon, C., Carliell-Marquet, C., Browning, S., 2008. Can fluorescence spectrometry be used as a surrogate for the biochemical oxygen demand (BOD) test in water quality assessment? An example for South West England. Sci. Total Environ. 391 (1), 149e158. Jermann, D., Pronk, W., Ka¨gi, R., Halbeisen, M., Boller, M., 2008. Influence of interactions between NOM and particles on UF fouling mechanisms. Water Res. 42 (14), 3870e3878. Jones, K.L., O’Melia, C.R., 2001. Ultrafiltration of protein and humic substances: effect of solution chemistry on fouling and flux decline. J. Membr. Sci. 193 (2), 163e173.

Jucker, C., Clark, M.M., 1994. Adsorption of aquatic humic substances on hydrophobic ultrafiltration membranes. J. Membr. Sci. 97, 37e52. Kimura, K., Hane, Y., Watanabe, Y., Amy, G., Ohkuma, N., 2004. Irreversible membrane fouling during ultrafiltration of surface water. Water Res. 38 (14e15), 3431e3441. Liu, R., Lead, J.R., Baker, A., 2007. Fluorescence characterization of cross flow ultrafiltration derived freshwater colloidal and dissolved organic matter. Chemosphere 68 (7), 1304e1311. Peiris, B.R.H., Halle´, C., Haberkamp, J., Legge, R.L., Peldszus, S., Moresoli, C., Budman, H., Amy, G., Jekel, M., Huck, P.M., 2008. Assessing nanofiltration fouling in drinking water treatment using fluorescence fingerprinting and LC-OCD analyses. Water Sci. Technol. Water Supply 8 (4), 459e466. Peiris, R.H., Budman, H., Moresoli, C., Legge, R.L., 2010a. Understanding fouling behaviour of ultrafiltration membrane processes and natural water using principal component analysis of fluorescence excitation-emission matrices. J. Membr. Sci. 257 (1e2), 62e72. Peiris, R.H., Halle´, C., Budman, H., Moresoli, C., Peldszus, S., Huck, P.M., Legge, R.L., 2010b. Identifying fouling events in a membrane-based drinking water treatment process using principal component analysis of fluorescence excitationemission matrices. Water Res. 44 (1), 185e194. Peiris, R.H., Budman, H., Legge, R.L., Moresoli, C., 2011. Assessing irreversible fouling behavior of membrane foulants in the ultrafiltration of natural water using principal component analysis of fluorescence excitation-emission matrices. Water Sci. Technol. Water Supply 11(2), 179e185. Sierra, M.M.D., Giovanela, M., Parlanti, E., Soriano-Sierra, E.J., 2005. Fluorescence fingerprint of fulvic and humic acids from varied origins as viewed by single-scan and excitation/ emission matrix techniques. Chemosphere 58 (6), 715e733.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 7 1 e5 1 8 3

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Comparison of potentially pathogenic free-living amoeba hosts by Legionella spp. in substrate-associated biofilms and floating biofilms from spring environments Bing-Mu Hsu a,*, Chin-Chun Huang a, Jung-Sheng Chen a,b, Nai-Hsiung Chen a, Jen-Te Huang a a

Department of Earth and Environmental Sciences, National Chung Cheng University, 168 University Road, Minhsiung Township, Chiayi County 62102, Taiwan, ROC b Research and Diagnostic Center, Centers for Disease Control, Taipei, Taiwan, ROC

article info

abstract

Article history:

This study compares five genera of free-living amoebae (FLA) hosts by Legionella spp. in the

Received 18 April 2011

fixed and floating biofilm samples from spring environments. Detection rate of Legionella

Received in revised form

spp. was 26.9% for the floating biofilms and 3.1% for the fixed biofilms. Acanthamoeba spp.,

24 June 2011

Hartmanella vermiformis, and Naegleria spp. were more frequently detected in floating bio-

Accepted 15 July 2011

film than in fixed biofilm samples. The percentage of pathogenic Acanthamoeba spp. among

Available online 23 July 2011

all the genus Acanthamoeba detected positive samples was 19.6%. The potential pathogenic

Keywords:

italica) was 54.2% to all the Naegleria detected positive samples. In the study, 12 serotypes of

Acanthamoeba

possible pneumonia causing Legionella spp. were detected, and their percentage in all the

Naegleria spp. (for example, Naegleria australiensis, Naegleria philippinensis, and Naegleria

Hartmanella vermiformis

Legionella containing samples was 42.4%. The FLA parasitized by Legionella included

Naegleria

unnamed Acanthamoeba genotype, Acanthamoeba griffini, Acanthamoeba jacobsi, H. vermi-

Legionella

formis, and N. australiensis. Significant differences were also observed between the pres-

Biofilm

ence/absence of H. vermiformis and Legionella parasitism in FLA. Comparisons between the culture-confirmed method and the PCR-based detection method for detecting FLA and Legionella in biofilms showed great variation. Therefore, using these analysis methods together to detect FLA and Legionella is recommended. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

The Legionellae are Gram-negative and non-spore-forming bacilli, which are ubiquitous in natural and man-made aqueous environments, with at least forty-eight species and seventy serogroups identified (Percival et al., 2004). Approximately half of genuses Legionella are etiological agents of Legionnaires’ disease and Pontiac fever (Declerck et al., 2007; Diederen, 2008). Legionellae are able to reproduce at

temperatures between 25
C and 43
C and survive in temperatures of up to 55
Ce60
C; therefore, they are frequently reported from thermo water systems (Sanden et al., 1989; Leoni et al., 2001; Hsu et al., 2006). Legionellae can exist as free-living planktonic forms in the environment, but they are more commonly found as intracellular parasites of protozoans, especially for free-living amoebae (FLA)d including Acanthamoeba spp., Naegleria spp., Hartmanella spp., and Vahlkampfia spp. Studies have reported at least thirteen

* Corresponding author. Tel.: þ886 5 2720411x66218; fax: þ886 5 2720807. E-mail address: [email protected] (B.-M. Hsu). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.019

5172

4.2 11 2.8  102 8.9  102 7.1 1.6  102 104 104 105 104 105 105       5.5 4.9 3.8 1.1 3.5 1.3 41.4 38.7 38.5 39.3 40.8 37.3 40 5.7 11 41 7.8 13 3.62 8.24 7.62 7.42 8.11 8.22 3 14 12 2 0 14 7 15 8 0 1 2 5 6 5 5 4 20 10 12 25 3 4 16

Vahlkampfia Naegleria Hartmanella

33 34 37 10 11 35

25
080 24
510 24
490 23
060 22
410 22
050

N, N, N, N, N, N,

121
300 121
330 121
460 120
410 120
590 120
440

E E E E E E

7 2 12 3 2 14

Total coliforms HPC Temperature Turbidity pH

Water quality parameter (mean)

Positive number of Legionella

Peitou (A) Wulai (B) Giausi (C) Pauli (D) Zerben (E) Sichunsi (F)

The longitude and latitude of six spring recreation area (AeF) and the number of sampling sites in each area are displayed in Table 1. In each sampling site, the floating biofilm and fixed biofilm were taken. The sampling period was from June 2008 to May 2009. The floating biofilm samples from the spring water at the airewater interface shows organization of cells with the matrix at the outside. Therefore, 1 L of airewater interface (1 to 5-cm-deep) spring waters or containing an extensive 2- to 3mm-thick algal mat was chosen for each floating biofilm sample. In order to concentrate free-living amoebae (FLA) in floating biofilm, 1 L aliquots of the water sample were filtered through 45-mm diameter cellulose nitrate membranes (Pall, USA) with a pore size of 0.22 mm. The material was then scraped off the filters and diluted with PBS. The solution was transferred into a 50 ml conical centrifuge tube, and centrifuged (2000 g, 20 min). After aspirating the top solution, the remaining 5 ml pellet was preserved.

Acanthamoeba

Study location and sample preparation

Positive number of FLA

2.1.

Longitude/ Latitude

Materials and methods

No. of Sites

2.

Sampling Area (Code)

FLA species to support the intracellular replication of Legionella, which eventually lysing the host cells and returns to the environment (Stout and Yu, 1997; Greub and Raoult, 2004; Declerck et al., 2007). When the Legionella containing aerosol is inhaled by humans, it can replicate within alveolar macrophages and may cause infection (Rowbotham, 1980; Abu Kwaik et al., 1998). FLA is ubiquitous in soil and aquatic environments. Their profusion and diversity in the environment are strongly dependent on temperature, moisture, pH, and nutrient availability (Rodriguez-Zaragoza, 1994; Bass and Bischoff, 2001). The trophozoite forms of FLA feed phagocytotically on bacteria, fungi, and algae, and shelter them from antibiotic and disinfection treatments (Hoffmann and Michel, 2001; Greub and Raoult, 2004). In natural and man-made environments, bacteria are commonly found in biofilms (Costerton et al., 1987). Biofilms can be developed on solid substratums in contact with water (fixed biofilms) or at the airewater interface (floating biofilms). Not only does the FLA often live on biofilms, but Legionella species also survive as free-living planktonic cells within biofilm communities, where the biofilm matrices provide a habitat and nutrients (Harb et al., 2000; Fields et al., 2002). However, the comparison of living situations for Legionella and FLA in two types of biofilms is still limited. The aim of this study was to determine the prevalence of Legionella and FLA in fixed biofilm and floating biofilm communities of spring water. Therefore, a total of 160 parallel floating and fixed biofilm samples were obtained from six locations in Taiwan. The study also evaluated Legionella existing as free-living forms or as intracellular parasites of FLA in these two types of biofilms. Legionella spp. and five genera of FLA were analyzed with a culture-confirmed method, as well as the PCR-based detection method, combined with the molecular taxonomic identification method.

Table 1 e The mean level of water quality parameters and detected positive number of FLA and Legionella in six hot spring recreation areas of Taiwan.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 7 1 e5 1 8 3

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 7 1 e5 1 8 3

The fixed biofilm samples from the spring wateresolid interface were collected by scraping surface material from the inner wall of the water basins with a transwab packet (Creative Microbiologicals, LTD., Taiwan), which contains a swab stick with a color-coded cap and pre-labeled transport tube, used for safe and efficient transport of bacteria. To analyze FLA and Legionella in fixed biofilms, the biofilm that adhered to the swab was cut into small pieces before the fibers of the swab were vortexed in 5 ml of PBS.

2.2.

transferred onto buffered charcoaleyeast extract (BCYE) agar and then incubated for 5 days at 37
C. On BCYE, Legionellae are identified by their colonial morphology and the requirement of L-cysteine. After 5 days’ growth on BCYE, Legionella candidates have a characteristic “cut glass” appearance: grayewhite, glistening, convex and 3e4 mm in diameter. The presumptive colonies were picked by sterile loop, suspended in 1 ml of sterile PBS, heated at 95
C for 10 min, centrifuged at 9700 g, 10 min, and then the suspension was analyzed by PCR reaction for the presence of Legionella-specific genes.

Free-living amoebae detection procedure

Two different detection methods, FLA culture-confirmed and PCR-based detection, were followed in this study. For the FLA culture-confirmed method, 200 ml of resuspended samples of floating biofilm and fixed biofilm were subsequently spread on 1.5% non-nutrient agar (NNA) plates seeded with a heat-killed suspension of Escherichia coli, and incubated at 32
C for 14 days. During incubation time, the FLA candidates were transferred onto new NNA plates containing a lawn of heatkilled E. coli 1e3 times to avoid fungal contamination. The transfer times were determined by considering the fungal growth situation. The FLA candidates on NNA-E. coli plate were transferred to 5 ml glass tubes with PYG medium consisting of 2% proteose peptone, 0.2% yeast extract, 0.1 M glucose, and 1% Gibco Antibioticeantimycotic (Cat. No. 1524006, USA) and incubated at 32
C for 3e4 days. The samples that FLA stuck on the glass tubes were carefully poured out the medium and gently washed with PBS to discard the debris and the bacteria outside FLA. The procedures of PYG cultivation and PBS elution were repeated once. Finally, the FLA were harvested by vortex and resuspended in sterile PBS. Each sample was centrifuged to 300 ml at 9700 g, 10 min, and then DNA extraction was conducted by Nucleospin Tissue (Macherey-Nagel Inc., Germany), which is recommended for purification of total DNA from clinical samples with the highest sensitivity, and its suitability for use with FLA has been proven in our other studies (Hsu et al., 2009a, 2009b; Huang and Hsu, 2010a, 2010b). For the PCR-based detection method, 1 ml of resuspended samples of floating biofilm and fixed biofilm were directly extracted using Nucleospin tissue according to the kit provided manufacturer’s instructions manual.

2.3.

5173

Legionella detection procedure

To determine if Legionella was present in the spring biofilms and the living situation of Legionella, three detection methods, PCR-based detection, Legionella culture-confirmed, and Amoeba-intracellular Legionella culture-confirmed were compared in this study. The detection procedures of Amoebaintracellular Legionella culture-confirmed method and PCRbased detection method for Legionella were the same with detection procedures for FLA but used different PCR conditions. As for the procedures of Legionella culture-confirmed, 200 ml of resuspended solution of floating biofilms and fixed biofilms were applied heat and acid treatment to reduce the presence of contaminating unwanted bacterial species. Following decontamination the collected material is

2.4. PCR conditions, gel electrophoresis and sequence analysis The extracted total genomic DNA from the procedures of PCRbased detection method as well as Legionella and FLA cultivation procedures were then subjected to PCR reaction. The PCR reaction solution was prepared with 5 ml of the DNA templates together with PCR mixture to create a total volume of 50 ml. The PCR mixture had 5 ml 10 PCR Buffer (20 mM MgCl2), 1 ml dNTP Mix (10 mM of each dNTP), 200 pmol each of the oligonucleotide primers and 0.3 ml VioTaqTM DNA Polymerase (Viogene, 5 U/ul), as well as DNase-free deionized water. The PCR assay primers, PCR cycling conditions, and their fragments of target genes used for detecting Acanthamoeba spp., Hartmanella vermiformis, Naegleria spp., Legionella spp., Vahlkampfia spp., and Willaertia spp. in this study were displayed in Table 2. The PCR assay primers used in this study were designed by Miyamoto et al. (1997), Pe´landakis et al. (2000), Schroeder et al. (2001), Pe´landakis and Pernin (2002), Garstecki et al. (2005) and Kuiper et al. (2006). PCR products were identified with gel electrophoresis on a 2% agarose gel (Biobasic Inc. Canada) performed with 5 ml of the reaction solution. DNA fragments were confirmed using ethidium bromide staining (0.5 mg/ml, 10 min). A 100-bp DNA ladder was used as a DNA size marker. We used a Bio-Dye Teminator Cycle Sequencing Kit (Applied Biosystem, USA) for sequence analysis. The gene sequences were aligned by use of the DNAMAN software program (Lynnon Biosoft, Canada). Negative DNA controls (template DNA replaced with distilled water), and positive controls (Acanthamoeba, isolated from a clinical sample from National Cheng Kung University Hospital; Hartmanella ATCC 50237; Naegleria ATCC 22758, and Legionella pneumophila ATCC 33823) and sample DNA were analyzed in triplicate during each PCR run. In this study, controls for inhibition were run. Each sample was assayed twice with each primer set, once after it was seeded with a positive control that consisted of DNA template and once when it was unseeded. The seeded reactions were performed to determine if the sample would inhibit the PCR.

2.5. Physical parameters and microbiological parameters analysis Water temperature and its pH were measured in situ using a portable pH meter (D-24E, Horiba Co., Japan). Turbidity was measured using a ratio turbidimeter (HACH Co., Loveland, CO). Water samples taken for indicator microorganisms were collected in a 300 ml sampling bag (Nasco Whirl-Pak, USA).

5174

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 7 1 e5 1 8 3

Table 2 e Description of the primers used in PCR of Acanthamoeba, Hartmanella, Naegleria, Vahlkampfia, Willaertia, and Legionella. Pathogen

Target gene

Primer sequences

Acanthamoeba Acanthamoeba-specific JDP1 50 - GGC CCA GAT amplimer (ASA.S1) CGT TTA CCG TGA A-30 JDP2 50 -TCT CAC AAG CTG CTA GGG GAG TCA-30 Hartmanella 18S rDNA Hv1227F 50 -TTA CGA GGT CAG GAC ACT GT-30 Hv1728R 50 -GAC CAT CCG GAG TTC TCG-30 Naegleria Complete ITS Region ITS1 50 -GAA CCT GCG TAG GGA TCA TTT-30 ITS2 50 -TTT CTT TTC CTC CCC TTA TTA-30 Willaertia Complete ITS Region ITS1 50 -GAA CCT GCG TAG GGA TCA TTT-30 ITS2 50 -TTT CTT TTC CTC CCC TTA TTA-30 Vahlkampfia Complete ITS Region JITSF 50 -GTC TTC GTA GGT GAA CCT GC-30 JITSR 50 -CCG CTT ACT GAT ATG CTT AA-30 Legionella 16S rDNA LEG 225 50 -AAGATTA GCCTGCGTCCGAT-30 LEG 858 50 -GTCAAC TTATCGCGTTTGCT-30

Denaturation, annealing and extension temperature

Source

95
C

61
C

72
C

40

500

Schroeder et al., 2001

95
C

56
C

72
C

40

502

Kuiper et al., 2006

94
C

55
C

72
C

35

400e453

Pe´landakis et al., 2000

94
C

55
C

72
C

35

500

Pe´landakis and Pernin, 2002

95
C

52
C

72
C

35

700

Garstecki et al., 2005

95
C

62
C

72
C

35

654

Miyamoto et al., 1997

The samples were kept in coolers during transportation to the lab. Heterotrophic plate counts (HPC) were measured by the spread method. Total coliforms were measured by membrane filtration procedures with a differential medium described in the Standard Method for the examination of water and wastewater (Methods 9222 B and D) (APHA, 1995). The calculations of difference between the presence/absence of pathogenic FLA and Legionella and in the presence/absence of these pathogens based on various physical parameters and microbiological parameters were conducted using STATISTICA software (StatSoft, Inc., USA).

3.

Cycling Amplicon No. size (bp)

Results

3.1. Free-living amoeba prevalence in fixed biofilms and floating biofilms In this study, we collected the two types of parallel biofilms from each site. Table 3 gives the detecting results of Acanthamoeba spp., H. vermiformis, Naegleria spp., and Vahlkampfia spp. in fixed biofilms and floating biofilms by the PCR-based detection method and the culture-confirmed method combined with the molecular technique, except for Willaertia spp., not found in the study. Of the four genera of FLA in 160 sampling sties, Acanthamoeba spp., H. vermiformis, Naegleria spp., and Vahlkampfia spp. were found in the percentage of 25% (40/160), 43.8% (70/160), 26.3% (42/160), and 24.4% (39/160), respectively. While comparing the frequencies of four genera of FLA between the fixed biofilms and floating biofilms, the most FLA frequency in fixed biofilms was Vahlkampfia spp., followed by H. vermiformis, Acanthamoeba spp., and Naegleria

spp. The frequency of four genera of FLA in floating biofilms was H. vermiformis, Naegleria spp., Acanthamoeba spp., and Vahlkampfia spp. in series. The sampling sites containing the same genus of FLA in fixed biofilms and floating biofilms were rarely detected (ranging from 1.3% to 7.5%). Except for Vahlkampfia spp., the other three genera of FLA were more frequently detected in floating biofilms than in fixed biofilms. Table 1 shows the detected positive number of four genera of FLA in six spring recreation areas of this study. The four genera of FLA were found in all the spring recreation areas, except for Vahlkampfia spp., which was not detected in Area D. The highest percentage of FLA in each spring recreation area was H. vermiformis in Area A and C, Naegleria spp. in Area D and F, Vahlkampfia spp. in Area B, and Naegleria spp. as well as Vahlkampfia spp. in Area E. The results indicate that the hot spring environment was capable of supporting the essential elements for FLA growth. This study also includes a statistical test comparing the independent means of the values of water quality parameters to reveal any significant differences between the samples with and without FLA. Significant differences (ManneWhitney U test, P < 0.05) were observed between the presence/absence of Acanthamoeba and heterotrophic plate counts (HPC), H. vermiformis and HPC, as well as Naegleria and total coliforms in floating biofilms. The significant difference between Acanthamoeba and HPC, Vahlkampfia and total coliforms, Vahlkampfia and water temperature, as well as between Vahlkampfia and turbidity were also observed in solideliquid interface biofilms. The homologous relationships of all sequences of Acanthamoeba PCR products in this study and the Acanthamoeba reference strains from the NCBI GenBank were inferred by neighbor-joining analysis from pairwise comparisons of the

Method 1: PCR-based detection method; Method 2: FLA culture-confirmed method; no.: number; Method 1X2: FLA was detected positive by Method 1 and Method 2, simultaneously.; Method 1W2: FLA was detected positive by Method 1 or Method 2.; Biofilm 1X2: FLA was detected positive in Biofilm 1 and Biofilm 2 in each site, simultaneously.; Biofilm 1W2: FLA was detected positive in Biofilm 1 or Biofilm 2 in each site.

(40/160) (70/160) (42/160) (39/160) 25 43.8 26.3 24.4 3.1 (5/160) 7.5 (12/160) 1.9 (3/160) 1.3 (2/160) (25/160) (61/160) (42/160) (15/160) 15.6 38.1 26.3 9.4 1.9 (3/160) 12.5 (20/160) 1.9 (3/160) 0 (0/160) (19/160) (52/160) (24/160) (9/160) 11.9 32.5 15 5.6 (9/160) (29/160) (21/160) (6/160) 5.6 18.1 13.1 3.8 (20/160) (21/160) (3/160) (26/160) 12.5 13.1 1.9 16.3 1.9 (3/160) 1.9 (3/160) 0 (0/160) 0 (0/160) 5.6 9.4 0.6 16.3 Acanthamoeba Hartmanella Naegleria Vahlkampfia

8.8 (14/160) 5.6 (9/160) 1.3 (2/160) 0 (0/160)

(9/160) (15/160) (1/160) (26/160)

Biofilm 1W2 Biofilm 1X2 Method 1W2 Method 1X2 Method 2 Method 1 Method 1W2 Method 1X2 Method 2 Method 1

Sampling site percentage positivity (Positive no./Total no.) Floating biofilm (Biofilm 2) percentage positivity (Positive no./Total no.) Solideliquid interface biofilm (Biofilm 1) percentage positivity (Positive no./Total no.) Free-living Amoeba

Table 3 e The prevalence of four genera of free-living amoeba in biofilm and sampling sites, expressed as the percentages of positive samples detected with two analysis methods and the combined results.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 7 1 e5 1 8 3

5175

partial subunit 18S rRNA gene, the Acanthamoeba-specific amplimer (Fig. 1). The tree showed three main clusters. The first cluster included most of the Acanthamoeba sequences in the study with the reference strains of A. palestinensis (T2), A. lenticulata (T5), A. tubiashi (T8), A. castellanii (T4), A. healyi (T12), A. jacobsi (T15), A. griffini (T3), and A. polyphaga (T4). The reference strains of A. astronyxis (T7) and A. comandoni (T9) formed the second main cluster. The third cluster was composed of four sequences of Acanthamoeba products without any reference strains. The most frequently identified Acanthamoeba genotype was T15 (n ¼ 19), followed by T2 (n ¼ 13), and then A. polyphaga (T4, n ¼ 7). The remaining genotypes were relatively rare, such as T5 (n ¼ 2), T3 (n ¼ 2) and A. castellanii (T4, n ¼ 1). Except for four unnamed Acanthamoeba sequences cataloged into the third cluster, the remaining three sequences of Acanthamoeba PCR products, which could not be identified from the homologous tree, were cataloged as subgenotypes with T15 (two sequences) in the first cluster outgroup (one sequence). Fig. 2 displays the neighbor-joining analysis inferred from the relationships between the sequences of Naegleria PCR products and Naegleria reference strains. The homologous tree showed three main clusters. The first cluster included fortyone environmental sequences with most of the reference strains. The second cluster contained one reference strain, Naegleria gruberi. The third cluster contained seven PCR products without reference strains. The most frequently identified sequences of the Naegleria genotype were N. australiensis (n ¼ 21), followed by N. lovaniensis (n ¼ 5), and then uncultured Naegleria (n ¼ 4), N. philippinensis (n ¼ 3), N. clarki (n ¼ 2), N. italica (n ¼ 2) as well as N. andersoni (n ¼ 1). In the study, most PCR products of Naegleria were from the floating biofilms (n ¼ 45) and merely three PCR products were from fixed biofilms, identified as N. italica, and uncultured Naegleria. Although Naegleria fowleri was not identified from the spring biofilms in the study, the percentage of low virulence of Naegleria (for example N. australiensis, N. philippinensis and N. italica) was 54.2% (26/48) to all Naegleria containing samples. The Vahlkampfia PCR products and Vahlkampfia reference strains from NCBI GenBank were used to construct the phylogenetic tree shown in Fig. 3. The forty-one Vahlkampfia sequences were dispersed over two main genotypical clusters. Thirty-two sequences (78.0%) were identified as the same genogroup with Vahlkampfia avara and seven Vahlkampfia sequences (17.1%) were in the cluster with uncultured Vahlkampfia. The other two ITS sequences of genus Vahlkampfia, A19 floating biofilm and B29 floating biofilm, recovered in this study did not match those of other species deposited in the nucleotide database, and phylogenetic analysis inferred that these isolates might be novel genotypes of the Vahlkampfia.

3.2. Legionella prevalence in fixed biofilms and floating biofilms Table 1 shows the detected positive number of Legionella in six spring recreation areas of this study. Legionella was found in five of all six spring recreation areas. The spring recreation areas containing the highest percentage of Legionella were Area F (40%), followed by Area B (41.2%), Area C (32.4%), Area D (20%), and Area A (9.1%). Table 4 shows the results of Legionella

5176

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 7 1 e5 1 8 3

Fig. 1 e Phylogenetic relationships of Acanthamoeba PCR products and reference strains from the NCBI GenBank, inferred by neighbor-joining analysis from pairwise comparisons of Acanthamoeba-specific amplimer (ASA.S1) nucleotide sequences.

monitoring by three analysis methods in fixed biofilms and floating biofilms. Legionella was detected positive in forty-five of the 160 sites (28.1%), where forty-three sites were detected positive for Legionella in floating biofilms but merely five sites were detected positive for Legionella in fixed biofilms. In the 160 floating biofilm samples, thirty-two samples analyzed by

PCR-based detection method were positive for Legionella, eight samples analyzed by Legionella culture-confirmed method were positive for Legionella, and nine samples were positive for Legionella when analyzed by the Amoeba-intracellular Legionella culture-confirmed method. While detecting Legionella in the fixed biofilms, two samples analyzed by PCR-based

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 7 1 e5 1 8 3

5177

Fig. 2 e Phylogenetic relationships of Naegleria PCR products and reference strains from NCBI GenBank, inferred by neighbor-joining analysis from pairwise comparisons of complete ITS region nucleotide sequences.

detection method and three samples analyzed by Amoebaintracellular Legionella culture-confirmed method were positive, respectively. However, none of the samples were detected by the Legionella culture-confirmed method. When the Legionella-positive and -negative samples were compared with various water quality parameters, significant differences (ManneWhitney U test, P < 0.05) were observed between the presence/absence of Legionella and pH, as well as between the presence/absence of Legionella and turbidity.

Fig. 4 shows the homologous relationships of all sequences of Legionella PCR products from three analysis methods and the Legionella reference strains from NCBI GenBank were inferred by neighbor-joining analysis from pairwise comparisons of the subunit 16S rRNA gene. The most frequently identified Legionella species were the sequences in the same group with L. pneumophila as well as L. longbeachae (n ¼ 12), and uncultured Legionella spp. (n ¼ 12), followed by Legionella-like amoebal pathogens (n ¼ 4); the sequences in the same group

5178

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 7 1 e5 1 8 3

89 88 60 78

D9 floating biofilm (Method 1) B21 fixed-biofilm (Method 2)

B27 fixed-biofilm (Method 2) C31 fixed-biofilm (Method 2)

64

B28 fixed-biofilm (Method 2)

60

B23 fixed-biofilm (Method 2)

56

C30 fixed-biofilm (Method 2)

0

B22 fixed-biofilm (Method 2) D4 floating biofilm (Method 1)

3

A30 floating biofilm (Method 2) B20 fixed-biofilm (Method 2) B26 fixed-biofilm (Method 2) D8 floating biofilm (Method 1) B33 fixed-biofilm (Method 2) C32 fixed-biofilm (Method 2) C29 fixed-biofilm (Method 2) F29 floating biofilm (Method 1) A24 fixed-biofilm (Method 2) A33 fixed-biofilm (Method 2)

1 B31 fixed-biofilm (Method 2)

B18 fixed-biofilm (Method 2) B29 fixed-biofilm (Method 2) C20 fixed-biofilm (Method 2) B32 fixed-biofilm (Method 2) B17 fixed-biofilm (Method 2) 6 A23 fixed-biofilm (Method 2)

C33 fixed-biofilm (Method 2) V. avara (NCBI AJ698837) F35 floating biofilm (Method 1) 28

D1 floating biofilm (Method 1) B24 fixed-biofilm (Method 2) B30 fixed-biofilm (Method 2) B32 floating biofilm (Method 2) V. ciguana (NCBI AJ973126) V. inornata (NCBI AJ698838)

59 95

V. orchilla (NCBI AJ973127)

A19 floating biofilm (Method 2) 65

B29 floating biofilm (Method 2) F31 floating biofilm (Method 2) A21 fixed-biofilm (Method 2)

99

C2 floating biofilm (Method 2)

94

u. Vahlkampfiidae (NCBI EU812477)

51 26 32 35

F32 floating biofilm (Method 2) A31floating biofilm (Method 2) C7 floating biofilm (Method 2)

34 E2 fixed-biofilm (Method 2)

0.1

Fig. 3 e Phylogenetic relationships of Vahlkampfia PCR products and reference strains from NCBI GenBank, inferred by neighbor-joining analysis from pairwise comparisons of complete ITS region nucleotide sequences.

5179

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 7 1 e5 1 8 3

Table 4 e The prevalence of Legionella in biofilm and sampling sites, expressed as the percentages of positive samples detected with three analysis methods and the combined results. Analytical method Method 1 Method 2 Method 3 Method 1X2X3 Method 1W2W3

Solideliquid interface biofilm percentage positivity (Positive no./Total no.)

Floating biofilm percentage positivity (Positive no./Total no.)

Sampling site percentage positivity (Positive no./Total no.)

1.3 (2/160) 0 (0/160) 1.9 (3/160) 0 (0/160) 3.1 (5/160)

20 (32/160) 5 (8/160) 5.6 (9/160) 0 (0/160) 26.9 (43/160)

21.3 (34/160) 5 (8/160) 7.5 (12/160) 0 (0/160) 28.1 (45/160)

Method 1: PCR-based detection method; Method 2: Legionella culture-confirmed method; Method 3: Amoeba-intracellular Legionella cultureconfirmed method; Method 1X2X3: Legionella was detected positive by Method 1, Method 2, and Method 3, simultaneously.; Method 1W2W3: Legionella was detected positive by Method 1, Method 2, or Method 3.

with L. tucsonensis, L. gratiana, and L sainthelensi (n ¼ 3), and then L. worsleiensis (n ¼ 2), L. cherrii (n ¼ 2), L. quinlivanii serogroup 2 (n ¼ 2); the sequences in the same group with L. nautarum and L. londiniensis (n ¼ 2), and L. oakridgensis (n ¼ 2). Legionella lytica, L. parisiensis, L. feeleii, and the sequence in the same group with L. micdadei and L. jordanis were all detected once. Of the fourteen unnamed Legionella genotype samples, four clustered with L. pneumophila, L. longbeachae, L. worsleiensis as well as L. lytica, and two clustered with L. nautarum, L. londiniensis as well as L. oakridgensis, and one clustered with L. hackeliae and L. lansingensis.

3.3.

Legionella parasitism in free-living amoeba

The study used two steps of the FLA culture, the NNA-E. coli plate and the PYG medium, to exclude extracellular Legionella and obtained FLA-intracellular Legionella. Finally, Legionella and four genera of FLA were identified by molecular technique and Table 5 displays the results. Of the twelve FLA culture samples in which Legionella was detected positive, eight samples were detected positive for H. vermiformis, and three samples were detected positive for Acanthamoeba spp. and Naegleria australiensis, respectively. A statistical analysis was performed to detect any significant differences between the four pathogenic FLA-positive and -negative samples regarding the presence/absence of Legionella. Significant differences (Chi-square test, P < 0.05) were observed between FLAintracellular Legionella and non FLA-intracellular Legionella in terms of the existence of H. vermiformis indicate that H. vermiformis plays a role in the survival of FLA-intracellular Legionella. The species of FLA-intracellular Legionella in H. vermiformis included Leonella cherrii (n ¼ 2), uncultured Legionella spp. (n ¼ 2), Legionella-like amoebal pathogen (n ¼ 2), L. worsleiensis (n ¼ 1), and L. quinlivanii serotype 2 (n ¼ 1). The FLAintracellular Legionella in N. australiensis included uncultured Legionella spp., L. worsleiensis and Legionella-like amoebal pathogen. In the three Legionella-parasitic Acanthamoeba species, A. griffini, A. jacobsi and unnamed Acanthamoeba genotype, the intracellular Legionella, uncultured Legionella spp., L. pneumophila serotype 6 and Legionella-like amoebal pathogen, were detected, respectively. This paper was the first to find the characteristics of L. cherrii to parasitize amoeba.

4.

Discussion

Researches have commonly found the genus Acanthamoeba, Hartmanella, Naegleria, and Vahlkampfia in various environment water sources throughout the world (Schuster and Visvesvara, 2004). Among the FLA genus, some species of Naegleria and Acanthamoeba are pathogenic agents for humans and animals (Karanis et al., 2007). From a public health viewpoint, pathogenic FLA in the airewater interface of the spring environment presents a particular potential hazard to bathers for aerosol contamination than those in fixed biofilms. In the survey of Declerck et al. (2007) for Naegleria spp. and Acanthamoeba spp. in floating biofilms, the prevalence was 50e92% and 67e72%, higher than our detecting result. Declerck et al. (2007) also concluded that Willaertia spp. and Vahlkampfia spp. were hardly detected in floating biofilms. The study did not detect Willaertia spp. as positive in either floating biofilm nor in fixed biofilm, and Vahlkampfia spp. was detected positive in floating biofilm samples at a very low percentage compared to the percentage of Vahlkampfia spp. in fixed biofilms and the percentage of three other genera of FLA in floating biofilms. The detecting results of four genera of FLA by PCR-based detection method and culture-confirmed method were different. The H. vermiformis and Vahlkampfia spp. displayed higher detected percentage in the culture-confirmed method analysis of the two kinds of biofilms than by the PCR-based detection method. While detecting genus Acanthamoeba in fixed biofilms, the percentage obtained from the cultureconfirmed method was higher than the PCR-based detection method. However, while detecting Acanthamoeba spp. in floating biofilms, we obtained controversial results. While detecting genus Naegleria in the two kinds of biofilms, we obtained a similar percentage between the culture-confirmed method and the PCR-based detection method. The cultureconfirmed method used in this study was designed based on Acanthamoeba cultivation. The results indicate that the Acanthamoeba culture method significantly improves detection sensitivity of the genus Vahlkampfia, and H. vermiformis in biofilms. Although most FLA containing samples were detected positive by the culture-confirmed method, rarely have samples been detected positive by both the PCR-based detection method and the culture-confirmed method,

5180

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 7 1 e5 1 8 3

Fig. 4 e Phylogenetic relationships of Legionella PCR products and reference strains from NCBI GenBank, inferred by neighbor-joining analysis from pairwise comparisons of 16S rDNA nucleotide sequences.

simultaneously. This study recommends using the two analysis methods to survey the prevalence of FLA in biofilms. The six genotypes of Acanthamoeba identified in the current study are commonly found in aquatic and terrestrial environments reported in other researchers’ studies (Bottone, 1993;

Hewitt et al., 2003). Among them, the Acanthamoeba castellanii (T4), A. polyphaga (T4), and T3, have been recognized to cause Amoebic Keratitis, which is the only water-related syndrome caused by genus Acanthamoeba (Bottone, 1993; MarcianoCabral and Cabral, 2003; Yu et al., 2004). The other three

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 7 1 e5 1 8 3

Table 5 e Presence of amoeba-resistant Legionella and their hosts in biofilm. Sample Biofilm type Amoeba-resistant code Legionella A13 A18 B12

B21 B22 C23 D4 F28

F31 C2 C3 C18

Floating biofilm L. cherrii Floating biofilm L. cherrii Floating biofilm Legionella-like amoebal pathogen Floating biofilm Uncultured Legionella spp. Floating biofilm L. worsleiensis Floating biofilm L. quinlivanii SG2 Floating biofilm Uncultured Legionella spp. Floating biofilm Legionella-like amoebal pathogen Floating biofilm Uncultured Legionella spp. Fixed biofilm L. pneumophila Fixed biofilm L. lytica Fixed biofilm Legionella-like amoebal pathogen

Host H. vermiformis H. vermiformis H. vermiformis

H. vermiformis N. australiensis H. vermiformis N. tihangensis H. vermiformis H. vermiformis H. vermiformis N. tihangensis Acanthamoeba spp. A. griffini A. jacobsi e e

Acanthamoeba genotypes, T2, T5, and T15, are not considered as pathogenic Acanthamoeba. However, some papers have reported their infection and disease (Flint et al., 2003; Spanakos et al., 2006; Barete et al., 2007). In Taiwan, Acanthamoeba was the third most commonly isolated microbe among patients diagnosed with microbial infected keratitis (Chen et al., 2004). Therefore, if Acanthamoebae exists in the biofilms of spring water, it should be considered a potential health threat associated with human activities in spring recreation areas. About thirty species of Naegleria have been recognized based on sequencing data and N. fowleri has proven to be the only Naegleria species that is pathogenic to humans (Karanis et al., 2007). This study did not find N. fowleri in the biofilm samples. Although some researchers have concluded that N. fowleri is commonly detected in the aquatic environment worldwide, it is rarely reported in Asia (Marciano-Cabral et al., 2003; Visvesvara et al., 2007; Edagawa et al., 2009). N. australiensis, N. italica and N. philippinensis have displayed pathogenicity in a few laboratory animals because their virulence is lower than that of N. fowleri (De Jonckheere, 2002; Schuster, 2002). Seven unnamed Naegleria in the third cluster, and the other three sequences of unnamed Naegleria genotype were corroborated in the same subcluster with N. fowleri and N. lovaniensis. The results infer that the three sequences of unnamed Naegleria genotypes show physiological properties related to pathogenicity, and some researchers have proven this hypothesis in other studies (Stevens et al., 1980; De Jonckheere, 1994; Walochnik et al., 2005). N. gruberi, which was drawn in the second cluster, displayed great genetic heterogeneity with other Naegleria species and has been discussed in the study of Robinson et al. (1992). Recent studies have found novel genotypes of Vahlkampfia in the environmental samples of Japan and Thailand

5181

(Edagawa et al., 2009). Edagawa et al. (2009) inferred that unidentified Vahlkampfia like species are common in the aquatic environment of Asia. Sequence analysis of the rRNA ITS region has proven to differentiate the genus Naegleria and Vahlkampfia more effectively than the previously used Naegleria flagellation test (Ettinger et al., 2003). The standard method for detecting Legionella species involves isolation in a selective medium. However, the standard method obtains less sensitivity compared to the other two methods. This is because the Legionella standard method fails to detect viable cells that cannot be cultured, the loss of viability of Legionella after collection, and Legionella requires intracellular protozoan hosts to develop a natural life cycle  o et al., 2002). The three analysis (Catalan et al., 1997; Gardun methods in this study did not detect Legionella as positive in each sample. Therefore, the three analysis methods used together to detect Legionella in biofilms are recommended. Besides L. pneumophila serogroup 1e6, there are twenty-two Legionella spp. reported to be pneumonia agents (USEPA, 1999). The positive pathogenic Legionella detected in this study included L. cherrii, L. feeleii, L. gratiana, L. jordanis, L. longbeachae, L. lytica, L. micdadei, L. oakridgensis, L. parisiensis, L. quinlivanii, L. sainthelensi, and L. tucsonensis. The percentage of pathogenic Legionella in all the Legionella containing samples was 42.4%. This investigation obtained neither the uncultured Legionella spp. nor the Legionella-like amoebal pathogen from the Legionella culture-confirmed method. The twelve uncultured Legionella spp. were obtained from the PCR-based detection method (n ¼ 9) as well as the Amoeba-intracellular Legionella culture-confirmed method (n ¼ 3), and four Legionella-like amoebal pathogens were obtained from the PCR-based detection method (n ¼ 1) as well as the Amoeba-intracellular Legionella culture-confirmed method (n ¼ 3). In the survey of Declerck et al. (2007) on floating biofilms in anthropogenic and natural aquatic systems, the prevalence of Legionella spp. and L. pneumophila ranged between 70% and 100%, higher than this study. The collected samples in this study were mainly from man-made spring facilities. Legionella spp. has developed mechanisms to survive or reproduce by residing in biofilms or in FLA. A regular clean and running water source for spring water systems influences biofilm accumulation and FLA that may cause the descent of Legionella prevalence. Except for L. cherrii, studies have documented the ability of other FLA-intracellular Legionella spp. to parasitize amoeba (USEPA, 1999; Wullings and van der Kooij, 2006). Although twenty-six fixed biofilm and fifteen floating biofilm samples were detected positive for genus Vahlkampfia by the FLA culture method, none of the samples containing Vahlkampfia were detected with Legionella. This infers that it is difficult for Legionella to parasitize Vahlkampfia. In the research of Legionella parasitism in Vahlkampfia, Steele (1993) implicated the genus Vahlkampfia as the host of Legionella. However, the results of Taylor et al. (2009) and this study have yet to demonstrate that Legionella parasitizes the genus Vahlkampfia. In the study, two fixed biofilm samples were detected positive for FLAintracellular Legionella but not for the four genera of FLA. This infers that not all the Legionella parasitized hosts can be detected in our studies and the detection method remains questionable. Researchers worldwide have reported that the presence of Legionella in the aquatic environment depends on

5182

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 7 1 e5 1 8 3

the spectrum of parasitized-FLA present (Greub and Raoult, 2004). This study detected Legionella from forty-five out of 160 sampling sites. Among the forty-five Legionella detected positive sites, merely twelve sites (26.7%) were detected positive for FLA-intracellular Legionella. The low percentage of FLAintracellular Legionella infers that Legionella employs multiple survival strategies to persist in biofilms (Murga et al., 2001; Guerrieri et al., 2005). The results also revealed that the FLA-intracellular Legionella in floating biofilms were more frequently detected than in fixed biofilms. This suggests that nutrient conditions and FLA presence influence Legionella survival. Compared with floating biofilms, fixed biofilms can be treated as a nutrientrich system suitable for replicating most heterotrophic bacteria. However, Legionella’s inability to utilize L-cysteine may act as a limiting factor in extracellular growth (Ewann and Hoffman, 2006). Low nutrient conditions and high FLA prevalence within the floating biofilms may prompt Legionella to parasitize FLA (Taylor et al., 2009). FLA-intracellular Legionella is significant for human health because it parasitizes FLA for multiplication, frequently in floating biofilms, which more easily forms an aerosol, thereby increasing the risk of microorganism dissemination by human inhalation. FLA-intracellular Legionella may cause synergism due to a highly resistant pathogenicity to a disinfectant and the ability to resist destruction by human macrophages (Fritsche et al., 1998).

5.

Conclusions

(1) Vahlkampfia spp. was more frequently habituated in fixed biofilm than in floating biofilm samples; however, Acanthamoeba spp., H. vermiformis, Naegleria spp., and Legionella spp. yielded a controversial result. (2) When monitoring Legionella, we suggest using the Legionella culture-confirmed method, the PCR-based detection method, and the amoeba-intracellular Legionella cultureconfirmed method in concert. (3) The presence of A. castellanii, A. polyphaga, and A. griffin, lower pathogenic virulence Naegleria, N. australiensis, N. philippinensis, N. italic, and pathogenic Legionella in the biofilm of spring water should be considered a potential health threat associated with bathing. (4) H. vermiformis performs a function in the survival and growth of FLA-intracellular Legionella. (5) The Legionella parasitism in FLA included uncultured Legionella spp., L. pneumophila, L. cherrii, Legionella-like amoebal pathogen, L. worsleiensis, and L. quinlivanii SG 2. The hosted FLA included Acanthamoeba spp., A. griffini, A. jacobsi, H. vermiformis, and N. australiensis.

Acknowledgments This work was supported by a research grant from National Science Council (gs1) of Taiwan, R.O.C. (NSC97-2628-M-194001-MY3).

references

APHA, 1995. Standard Method for the Examination of Water and Wastewater, fifteenth ed. APHA, WEF and AWWA, Washington, DC. Abu Kwaik, Y., Gao, L.Y., Stone, B.J., Venkataraman, C., Harb, O.S., 1998. Invasion of protozoa by Legionella pneumophila and its role in bacterial ecology and pathogenesis. Appl. Environ. Microbiol. 64, 3127e3133. Barete, S., Combes, A., de Jonckheere, J.F., Datry, A., Varnous, S., Martinez, V., Ptacek, S.G., Caumes, E., Capron, F., France`s, C., Gibert, C., Chosidow, O., 2007. Fatal disseminated Acanthamoeba lenticulata infection in a heart transplant patient. Emerg. Infect. Dis. 13 (5), 736e738. Bass, P., Bischoff, P.J., 2001. Seasonal variability in abundance and diversity of soil gymnamoebae a short transect in southeastern USA. J. Eukaryot. Microbiol. 48, 475e479. Bottone, E.J., 1993. Free-living amebas of the genera Acanthamoeba and Naegleria: an overview and basic microbiological correlates. Mount. Sinai. J. Med. 60, 260e270. Catalan, V., Garcia, F., Moreno, C., Vila, M.J., Apraiz, D., 1997. Detection of Legionella pneumophila in wastewater by nested polymerase chain reaction. Res. Microbiol. 148, 71e78. Chen, W.L., Wu, C.Y., Hu, F.R., Wang, I.J., 2004. Therapeutic penetrating keratoplasty for microbial keratitis in Taiwan from 1987 to 2001. Am. J. Ophthalmol. 137 (4), 736e743. Costerton, J.W., Cheng, K.J., Geesey, G.G., Ladd, T.I., Nickel, J.C., Dasgupta, M., Marrie, T.J., 1987. Bacterial biofilms in nature and disease. Annu. Rev. Microbiol. 41, 435e464. Declerck, P., Behets, J., van Hoef, V., Ollevier, F., 2007. Detection of Legionella spp. and some of their amoeba hosts in floating biofilms from anthropogenic and natural aquatic environments. Wat. Res. 41, 3159e3167. De Jonckheere, J.F., 1994. Comparisons of partial SSU-rDNA sequences suggests revisions of species names in the genus. Naegleria. Eur. J. Protistol. 30, 333e341. De Jonckheere, J.F., 2002. A century of research on the amoeboflagellate genus. Naegleria. Acta Protozool. 41, 309e342. Diederen, B.M.W., 2008. Legionella spp. and Legionnaires’ disease. J. Infect. 56, 1e12. Edagawa, A., Kimura, A., Kawabuchi-Kurata, T., Kusuhara, Y., Karanis, P., 2009. Isolation and genotyping of potentially pathogenic Acanthamoeba and Naegleria species from tap-water sources in Osaka, Japan. Parasitol. Res. 105 (4), 1109e1117. Ettinger, M.R., Webb, S.R., Harris, S.A., McIninch, S.P., Garman, G. C., Brown, B.L., 2003. Distribution of free-living amoebae in James River, Virginia, USA. Parasitol. Res. 89, 6e15. Ewann, F., Hoffman, P.S., 2006. Cysteine metabolism in Legionella pneumophila: characterization of an L-cysteine-utilizing mutant. Appl. Environ. Microbiol. 72, 3993e4000. Fields, B.S., Benson, R.F., Besser, R.E., 2002. Legionella and Legionnaires’ disease: 25 years of investigation. Clin. Microbiol. Rev. 15, 506e526. Flint, J.A., Dobson, P.J., Robinson, B.S., 2003. Genetic analysis of forty isolates of Acanthamoeba group III by multilocus isoenzyme electrophoresis. Acta Protozool. 42, 317e324. Fritsche, T.R., Sobek, D., Gautom, R.K., 1998. Enhancement of in vitro cytopathogenicity by Acanthamoeba spp. following acquisition of bacterial endosymbionts. FEMS Microbiol. Lett. 166, 231e236.  o, R.A., Gardun  o, E., Hiltz, M., Hoffman, P.S., 2002. Gardun Intracellular growth of Legionella pneumophila gives increase to a differentiation form dissimilar to stationary-phase forms. Infect. Immun. 70, 6273e6283. Garstecki, T., Brown, S., De Jonckheere, J.F., 2005. Description of Vahlkampfia signyensis n. sp. (Heterolobosea), based on

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 7 1 e5 1 8 3

morphological, ultrastructural and molecular characteristics. Eur. J. Protistol. 41, 119e127. Greub, G., Raoult, D., 2004. Microorganisms resistant to free-living amoebae. Clin. Microbiol. Rev. 17 (2), 413e433. Guerrieri, E., Bondi, M., Ciancio, C., Borella, P., Messi, P., 2005. Micro- and macromethod assays for the ecological study of Legionella pneumophila. FEMS Microbiol. Lett. 252, 113e119. Harb, O.S., Gao, L.Y., Kwaik, Y.A., 2000. From protozoa to mammalian cells: a new paradigm in the life cycle of intracellular bacterial pathogens. Environ. Microbiol. 2, 251e265. Hewitt, M.K., Robinson, B.S., Monis, P.T., Saint, C.P., 2003. Identification of a new Acanthamoeba 18S rRNA gene sequence type, corresponding to the species Acanthamoeba jacobsi Sawyer, Nerad, and Visvesvara, 1992 (Lobosea: Acanthamoebidae). Acta Protozool. 42, 325e329. Hoffmann, R., Michel, R., 2001. Distribution of free-living amoebae (FLA) during preparation and supply of drinking water. Int. J. Hyg. Environ. Health 203, 215e219. Hsu, B.M., Chen, C.H., Wan, M.T., Cheng, H.W., 2006. Legionella prevalence in hot spring recreation areas of Taiwan. Wat. Res. 40, 3267e3273. Hsu, B.M., Lin, C.L., Shih, F.C., 2009a. Survey of pathogenic freeliving amoebae and Legionella spp. in mud spring recreation area. Wat. Res. 43, 2817e2828. Hsu, B.M., Ma, P.H., Liou, T.S., Chen, J.S., Shih, F.C., 2009b. Identification of 18S ribosomal DNA genotype of Acanthamoeba from hot spring recreation areas in the central range, Taiwan. J. Hydrol. 367, 249e254. Huang, S.W., Hsu, B.M., 2010a. Isolation and identification of Acanthamoeba from Taiwan spring recreation areas using culture enrichment combined with PCR. Acta Trop. 115, 282e287. Huang, S.W., Hsu, B.M., 2010b. Survey of Naegleria and its resisting bacteria-Legionella in hot spring water of Taiwan using molecular method. Parasitol. Res. 106, 1395e1402. Karanis, P., Kourenti, C., Smith, H., 2007. Waterborne transmission of protozoan parasites: a worldwide review of outbreaks and lessons learnt. J. Water Health 5, 1e38. Kuiper, M.W., Valster, R.M., Wullings, B.A., Boonstra, H., Smidt, H. , van der Kooij, D., 2006. Quantitative detection of the freeliving amoeba Hartmannella vermiformis in surface water by using real-time PCR. Appl. Environ. Microbiol. 72, 5750e5756. Leoni, E., Legnani, P.P., Bucci Sabattini, M.A., Righi, F., 2001. Prevalence of Legionella spp. in swimming pool environment. Wat. Res. 35, 3749e3753. Marciano-Cabral, F., Cabral, G., 2003. Acanthamoeba sp. as agents of disease in humans. Clin. Microbiol. Rev. 16, 273e307. Marciano-Cabral, F., MacLean, R., Mensah, A., LaPat-Polasko, L., 2003. Identification of Naegleria fowleri in domestic water sources by nested PCR. Appl. Environ. Microbiol. 69, 5864e5869. Miyamoto, H., Yamamoto, H., Arima, K., Fujii, J., Maruta, K., Izu, K., Shiomori, T., Yoshida, S., 1997. Development of a new seminested PCR method for detection of Legionella species and its application to surveillance of Legionellae in hospital cooling tower water. Appl. Environ. Microbiol. 63, 2489e2494. Murga, R., Forster, T.S., Brown, E., Pruckler, J.M., Fields, B.S., Donlan, R.M., 2001. Role of biofilms in the survival of Legionella pneumophila in a model potable-water system. Microbiology 147, 3121e3126. Pe´landakis, M., Pernin, P., 2002. Use of multiplex PCR and PCR restriction enzyme analysis for detection and exploration of the variability in the free-living amoeba Naegleria in the environment. Appl. Environ. Microbiol. 68, 2061e2065.

5183

Pe´landakis, M., Serre, S., Pernin, P., 2000. Analysis of the 5.8S rRNA gene and the internal transcribed spacers in Naegleria spp. and in N. fowleri. J. Eukaryot. Microbiol. 47, 116e121. Percival, S.L., Chalmers, R.M., Embrey, M., Hunter, P.R., Sellwood, J., Wyn-Jones, P., 2004. Legionella. In: Microbiology of Waterborne Diseases, first ed. Elsevier Academic Press, California, USA, pp. 145e153. Robinson, B.S., Christy, P., Hayes, S.J., Dobson, P.J., 1992. Discontinuous genetic variation among mesophilic Naegleria isolates: further evidence that N. gruberi is not a single species. J. Protozool. 39, 702e712. Rodriguez-Zaragoza, S., 1994. Ecology of free-living amoebae. Crit. Rev. Microbiol. 20, 225e241. Rowbotham, T.J., 1980. Preliminary report on the pathogenicity of Legionella pneumophila for freshwater and soil amoebae. J. Clin. Pathol. 33, 1179e1183. Sanden, G., Fields, B.S., Barbaree, J.M., Feeley, J.C., 1989. Viability of Legionella pneumophila in chlorine-free waters at elevated temperatures. Curr. Microbiol. 18, 61e65. Schroeder, J.M., Booton, G.C., Hay, J., Niszl, I.A., Seal, D.V., Markus, M.B., Fuerst, P.A., Byers, T.J., 2001. Use of subgenic 18S ribosomal DNA PCR and sequencing for genus and genotype identification of Acanthamoeba from humans with keratitis and from sewage sludge. J. Clin. Microbiol. 39, 1903e1911. Schuster, F.L., 2002. Cultivation of pathogenic and opportunistic free-living amebas. Clin. Microbiol. Rev. 15, 342e354. Schuster, F.L., Visvesvara, G.S., 2004. Free-living amoebae as opportunistic and non-opportunistic pathogens of humans and animals. Int. J. Parasitol. 34, 1001e1027. Spanakos, G., Tzanetou, K., Miltsakakis, D., Patsoula, E., Malamou-Lada, E., Vakalis, N.C., 2006. Genotyping of pathogenic Acanthamoebae isolated from clinical samples in Greece-report of a clinical isolate presenting T5 genotype. Parasitol. Int. 55, 147e149. Steele, T.W., 1993. Interactions between soil amoebae and soil Legionellae. In: Barbaree, A.M., Breiman, R.F., Dufour, A.P. (Eds. ), Legionella: Current Status and Emerging Perspectives. American Society for Microbiology, Washington, DC. Stevens, A.R., De Jonckheere, J.F., Willaert, E., 1980. Naegleria lovaniensis new species: isolation and identification of six thermophilic strains of a new species found in association with Naegleria fowleri. Int. J. Parasit. 10, 51e64. Stout, J.E., Yu, V.L., 1997. Legionellosis. N. Engl. J. Med. 337, 682e687. Taylor, M., Ross, K., Bentham, R., 2009. Legionella, protozoa, and biofilms: interactions within complex microbial systems. Microb. Ecol. 58, 538e547. USEPA, 1999. Legionella: Human Health Criteria Document. USWPA Office of Science and Technology/Office of Water, Washington, DC. Visvesvara, G.S., Moura, H., Schuster, F.L., 2007. Pathogenic and opportunistic free-living amoebae: Acanthamoeba spp., Balamuthia mandrillaris, Naegleria fowleri, and Sappinia diploidea. FEMS Immunol. Med. Microbiol. 50, 1e26. Walochnik, J., Mu¨ller, K.D., Aspo¨ck, H., Michel, R., 2005. An endocytobiont harbouring Naegleria strain identified as N. clarki De Jonckheere, 1994. Acta Protozool. 44, 301e310. Wullings, B.A., van der Kooij, D., 2006. Occurrence and genetic diversity of uncultured Legionella spp. in drinking water treated at temperatures below 15 degrees C. Appl. Environ. Microbiol. 72, 157e166. Yu, H.S., Kong, H.H., Kim, S.Y., Hahn, Y.H., Hahn, T.W., Chung, D. I., 2004. Laboratory investigation of Acanthamoeba lugdunensis from patients with keratitis. Invest. Ophthalmol. Vis. Sci. 45, 1418e1426.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 8 4 e5 1 9 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Impact of exopolysaccharides on the stability of silver nanoparticles in water S. Sudheer Khan, Amitava Mukherjee, N. Chandrasekaran* Centre for Nano-Biotechnology, School of Bio-Sciences and Technology, VIT University, Vellore 632014, India

article info

abstract

Article history:

The stability of commercial silver nanoparticles (SNPs) in aquatic environment plays

Received 22 January 2011

a significant role in its toxicity to the environment and to human health. Here, we have

Received in revised form

studied the impact of bacterial exopolysaccharides (EPS) to the stability of engineered

15 July 2011

SNPs. When nanoparticles are present in neutral water, the nanoparticles exhibited low

Accepted 17 July 2011

zeta potential and are least stable. However, in the presence of EPS (10e250 mg/L), the

Available online 23 July 2011

negative surface charge of nanoparticles increased and therefore the propensity of nanoparticles to aggregate is reduced. In UVevisible spectroscopic analysis a decrease in

Keywords:

absorbance at plasmon peak of SNPs (425 nm) was observed till the addition of 50 mg/L of

Silver nanoparticles

EPS, beyond that a blue shift towards 417 nm was observed. The adsorption of EPS was

Exopolysaccharides

confirmed by Fourier-transform infrared spectroscopy. The EPS adsorbed SNPs were more

Zeta potential

stable and exhibited the zeta potential of higher than
30 mV.

Adsorption

ª 2011 Elsevier Ltd. All rights reserved.

Stabilization

1.

Introduction

Usage of silver particles (SNPs) are increasing significantly in consumer products such as food packaging, textiles, paints, household appliances and medical devices including wound dressings and therapeutic devices. The increased production and use of engineered nanoparticles in recent years have drawn the attention of the scientific community to its toxicity and health impacts, and to the flow of nanoparticles in the environment. The indiscriminate use of SNPs in consumer products and industrial applications the nanoparticles, possibly a large amount of these will be discharged into environment (Nowack and Bucheli, 2007). The studies by Impellitteri et al. (2009) revealed that the SNPs can easily leak into waste water from SNPs impregnated clothes and the washing systems during washing. The release of nanoparticles in the sewage treatment plants was estimated to be 270 tonnes per year (Blaser

et al., 2008). Kaegi et al. (2010) reported the release of SNPs from wall painted with nanopaints. In aquatic habitats, the nanoparticles could accumulate in the fish body and directly enter into food chain (Sun et al., 2007). The nanoparticles have adverse effects on human health due to its smaller size and large surface area (AshaRani et al., 2009). Due to the ill effects of nanoparticles they are being included as a category of emerging potential toxic contaminants and the stability of these nanoparticles in environment needs to be investigated. In aquatic environments, the stability of nanoparticles may be affected by many factors, including pH and ionic strength. Previous studies have focused on the stability of metal oxide nanomaterials in the aquatic environment (Zhang et al., 2008, 2009), but the stability of SNPs in water is not well understood. The natural organic matter is known to have stabilizing effects on nanoparticles in water (Zhang et al., 2009). In the present study we have investigated the interaction of SNPs with exopolysaccharides (EPS) and its stability in

* Corresponding author. Tel.: þ91 416 2202624. E-mail addresses: [email protected], [email protected] (N. Chandrasekaran). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.024

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 8 4 e5 1 9 0

water. The EPS are produced by many bacterial and algal species biofilms, against toxic chemicals (Davey and O’Toole, 2000; Hunga et al., 2005). The results of our study would provide basic information on the stability of SNPs in natural waters.

2.

Materials and methods

2.1.

Materials

SNPs were obtained from Sigma Aldrich, USA. Nanoparticles were dispersed using an ultrasonic processor with a frequency of 132 kHz (Crest, USA). The EPS used in this study was extracted from the bacterium Bacillus pumilus (Accession no. GQ401238) isolated from sewage environment, Vellore, India. The organism was previously reported as SNPs resistant (Khan et al., 2011a). All the experiments were carried out in minimum of three replicates.

2.2.

Characterization of SNPs

The preliminary characterization of SNPs was done by UVevisible spectroscopy (Schimadzu UV - 1700, Japan). The size and morphology of the SNPs were analyzed by high resolution Transmission Electron Microscopy (TEM, Tecnai G20) and Scanning Electron Microscopy (FEI Sirion, Eindhoven, Netherlands). The samples were prepared by placing a drop of SNPs on a copper grid coated with a lacey carbon film and allowing it to dry in air. Mean particle size was analyzed from the digitized images with Image Tool software. The size distribution of particles in dispersion was analyzed by 90Plus Particle Size Analyzer (Brookhaven Instruments Corp., Holtsville, New York).

interaction was followed by centrifuging at 15,000 g for 10 min. The supernatant was collected and concentration of EPS left in the supernatant was quantified (Dubois et al., 1956). The effect of pH on adsorption was investigated in a pH range of 4e9 with an initial concentration of 250 mg/L EPS. The effect of salt concentration on adsorption was evaluated at different NaCl concentrations (0e1.5 M) (pH 7) with an initial concentration of 250 mg/L EPS. The amount of EPS adsorbed at equilibrium qe (mg/L) on SNPs was calculated from the following equation: qe ¼

2.5.

2.4.

Adsorption isotherms

For adsorption studies, the different concentrations of EPS (10e250 mg/L) were interacted with a fixed concentration of SNPs (20 mg/L) in ultrapure water for 4 h in a rotary shaker at 200 rpm and the pH was adjusted by 0.01 M HCl or NaOH solution. The experiments were carried out at pH 7. The

(1)

Adsorption kinetics

For determination of the adsorption kinetics, the EPS with an initial concentration of 250 mg/L was interacted with SNPs of 20 mg/L at pH 7. An aliquot of the interaction mixture was taken out at various time intervals, centrifuged and the concentration of EPS left in the supernatant was determined. The amount of adsorption at time t, qt (mg/mg), was calculated by:

Extraction and purification of EPS

EPS was extracted and purified according to the protocol described by Kumar et al. (2003). A loop full of bacterial culture was inoculated into 50 ml of LB broth, and it was allowed to grow for 24 h at room temperature, under shaking at 150 rpm. The culture was centrifuged at 10,000 g for 10 min and the supernatant was collected. To the supernatant, 10% (vol/vol) of saturated KCl solution was added as an electrolyte, precipitated with an equal volume of 95% ethanol, and kept at
20  C overnight. The precipitate was then removed by centrifugation at 10,000 g for 30 min at 4  C, washed twice with 95% ethanol, and air-dried. The pellet was dissolved in double-distilled water, dialyzed against distilled water for 72 h, precipitated with two volumes of acetone, and air-dried. The EPS was estimated by phenol-sulphuric method (Dubois et al., 1956) using glucose as standard (Torino et al., 2001).

ðC0
Ce ÞV W

where C0 and Ce (mg/L) are the concentrations of EPS at initial and equilibrium, respectively, V is the volume of the solution (L) and W is the mass of SNPs used (mg). All the tests were carried out in triplicate, and mean values of the results were reported. The experimental error limit was strictly kept within 5%. After the interaction of EPS with SNPs, the nanoparticles were subjected to lyophilization. Thereafter the lyophilized particles were subjected to fourier-transform infrared spectroscopy (FTIR).

qt ¼

2.3.

5185

ðC0
Ct ÞV W

(2)

where C0 and Ct (mg/L) are the concentrations of EPS at initial and the time t, respectively, V is the volume of the solution (L) and W is the mass of SNPs used (mg). The zeta potential and the particle size for the nanoparticles were noticed over the interaction period (72 h).

2.6.

Zeta potential measurement

The zeta potentials of SNPs, EPS and EPS adsorbed SNPs were measured using a Brookhaven Zeta 90Plus analyzer.

3.

Results and discussion

3.1.

Characterization of SNPs

UVevisible spectroscopy is the techniques used for the preliminary characterization of SNPs. UVevisible absorption spectra for the manufactured SNPs showed maximum absorbance at 425 nm. Size and morphology of SNPs were characterized by TEM and SEM. The microscopic images showed that SNPs were spherical in shape and polydispersed. The particle size distribution analysis of SNPs showed a mean diameter of 65  2.1 nm.

5186

3.2.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 8 4 e5 1 9 0

UVevisible spectral study

The uninteracted SNPs showed maximum absorbance at 425 nm. A decrease in absorbance of 425 nm observed at increasing concentrations of EPS. A Plasmon shift w1 nm (424 nm) was noted up to 50 mg/L of EPS, beyond that a blue shift (towards lower wavelength) in absorbance value by 5e10 nm was observed. The lowest absorbance value was noted at 417 nm for 250 mg/L EPS. The blue shift may be due to the enhanced electrostatic repulsion occurred between the particles due to electron density on the particle surfaces. The negatively charged EPS adsorbed onto the particle will restrict the free electrons of the SNPs within a smaller volume leading to an increased free electron density, thus a higher plasmon frequency (lower wavelength). Similar results were obtained by Ravindran et al. (2010).

3.3.

Adsorption isotherms

Many adsorption isotherm models are usually used to fit the adsorption data in order to obtain a linear regression data to predict the best-fit isotherm and the method of least squares has been used for finding the parameters of the isotherm. Two most commonly used isotherms Langmuir and Freundlich were employed in the present study. The Langmuir model is used to fit the experimental results. The Langmuir equation is expressed as qe ¼

qmax $Ka $Ce 1 þ Ka $Ce

(3)

where ‘Ce’ is the mass concentration of EPS in the supernatant (mg/L), ‘qe’ is the amount of EPS (mg) adsorbed per mg of SNPs, ‘qmax’ is the maximum amount of EPS at SNPs surface for a monolayer and ‘Ka’ is the adsorption constant (L/mg), reflects the affinity of EPS for SNPs surface. The experimental data were fitted to the Langmuir model, suggesting the monolayer adsorption of EPS onto SNPs. The high R2 values (linear regression coefficients) indicate that the Langmuir model predicts well the adsorption behavior of EPS on SNPs than Freundlich model. The EPS molecules had a significant number of negatively charged functional groups, which were attached on the surface of nanoparticles, formed the attractive coulombic interaction between the EPS and SNPs. The adsorption of EPS on SNPs was confirmed by FTIR analysis. The FTIR spectra of EPS capped SNPs is shown in Supplementary material Fig. S1.

3.4. EPS

Zeta potential and stability of SNPs in presence of

The stability of nanoparticles is mainly depending on the surface charge (Zhang et al., 2009). The particles with similarly charged surface were repelled each other due to electrostatic force of repulsion. Fig. 1a shows the zeta potential of EPS coated SNPs. In the absence of EPS, SNPs had positive zeta potential, þ10.23 mV at nearly neutral water. In the presence of EPS, SNPs exhibited negative zeta potential regardless of whether its original potential was positive. In 10 mg/L EPS, SNPs showed a zeta potential of
29.32 mV, beyond that SNPs exhibited

Fig. 1 e (a) The zeta potential of SNPs after 4 h interaction with different concentrations of EPS.(b) The average particle size of SNPs after interaction with different concentrations of EPS. ‘c’ represents the initial size of SNPs before interaction with EPS, ‘0’ represents the size of SNPs after 4 h interaction in absence of EPS and ‘10e250’ represents the size of SNPs after 4 h interaction with different concentrations of EPS (10e250 mg/L).

negative zeta potential greater than
30 mV. The result suggests that EPS imparted negative charge to SNPs surface. The effect of EPS on the stability of SNPs was evaluated by determining the sizes of nanoparticles with variable EPS concentrations (Fig. 1b). In the absence of EPS, The SNPs in water aggregated within 4 h, their average sizes increased to more than 1 mm (Fig. 1b). However, EPS could stabilize the nanoparticles effectively by preventing the aggregation/flocculation of SNPs. Release of SNPs to the environment potentially affects the bacterial species in the waste water treatment plants (Benn and Westerhoff, 2008) as well as bacterial communities in soil or natural waters (Neal, 2008; Bradford et al., 2009). Wigginton et al. (2010) reported that the adsorption of bacterial protein on the surface of SNPs increased bioavailability. The bacterial extracellular proteins have the ability to bind on SNPs that leads to its stability and transport in the environment (Khan et al., 2011b). Here we have studied the possible adsorption of bacterial EPS on SNPs and its stability in the environment. Dimkpa et al. (2011) studied the interaction of SNPs with an environmentally beneficial bacterium,

5187

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 8 4 e5 1 9 0

implications for understanding the stability, environmental transport, and bioavailability of SNPs in aquatic systems. The present study may also have significant implications for predicting the stability, transport and toxicity of a wide range of other related nanomaterials.

Adsorbed EPS (mg/mg)

160 140 120 100 80

3.5.

Kinetics of adsorption and stability of nanoparticles

60 40 20 0

0

20

40

60

80

Time (h) Fig. 2 e The adsorption kinetics of EPS on SNPs at an initial concentration of 250 mg/L.

To investigate the mechanism of adsorption and the potential rate controlling steps, pseudo-first-order (Lagergren, 1898) and pseudo-second-order (Ho and McKay, 1999) kinetics models were used. The adsorption kinetics of EPS on SNPs was carried out at an initial concentration of 250 mg/L EPS and the results are shown in Fig. 2. Linear form of pseudo-first-order kinetic equation is expressed as log qe
qt

Pseudomonas chlororaphis and its potential toxic effects. Similar results could also be observed from organically and inorganically coated nanoparticles once it is released into the environment. SNPs are ideal for incorporation into the membranes in order to reduce the biofouling of the polymeric membranes (Ng et al., 2010). This polymer coated SNPs are found to be stable in the environment and exhibited higher toxicity to bacterial species (El-Badawy et al., 2011). Dasari and Hwang (2010) reported on the stabilization of SNPs in the environment by humic acid and these stable nanoparticles were able to disrupt the natural aquatic bacterial assemblage (Dasari and Hwang, 2010) and bacterial biofilm (Fabrega et al., 2009). The toxicity of nanoparticle is related to surface properties and the nanoparticles can be functionalized with a monolayer or multilayer assembly of the desired hydrophobic or hydrophilic functional groups (Luo et al., 2010). The surface modified SNPs found to have high inhibitory effect on gram positive and gram negative bacterial species (Travan et al., 2011). Hence it can be suspected that, the EPS stabilized nanoparticles may also exhibit toxicity to environmentally beneficial bacterial species and aquatic organisms. Fabrega et al. (2011) evaluated the behavior, the biological effects and the route of uptake of SNPs to organisms and reported that very low concentration (ng/L) of SNPs can significantly affect prokaryotes, invertebrates and fish. The observations from this study have




¼ logqe


k1 t 2:303

(4)

where qe and qt are the amounts of adsorbed EPS on SNPs at equilibrium and at time t (mg/L), respectively, and k1 is the equilibrium rate constant of pseudo-first-order adsorption. The slope and intercept of the plot, log (qeeqt) versus t were used to obtain the pseudo-first-order rate constant k1 and equilibrium adsorption density qe. The pseudo-first-order rate constant k1 and qe determined from the model indicated that this model had failed to estimate qe since the experimental values of qe differs from estimated ones (Fig. 3a). Linear form of pseudo-second-order kinetic equation is expressed as t 1 t ¼
þ qt k2 qe 2 qe

(5)

The second order rate constant K2 and qe values were determined from the slopes and intercepts of the plots. The correlation coefficients R2 value is the indicative of the strength of the linear relationship. The theoretical qe values agree well with the experimental qe values, suggesting the adsorption data tend to follow pseudo-second-order kinetics (Fig. 3b), which relies on the assumption that chemisorption may be the rate limiting step. Here, we have studied the zeta potential of EPS capped SNPs over the interaction period (up to 72 h). Fig. 4a and

Fig. 3 e (a) The pseudo-first-order kinetics plotted as a function of log (qeLqt) vs time (h) and (b) The pseudo-second-order kinetics plotted as a function of t/qt vs time (h).

5188

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 8 4 e5 1 9 0

a

200

Equilibrium adsorption quantity, mg/mg solid

180 160 140 120 100 80 60 40 20 0

4

b

5

6

7

8

9

pH 40

Zeta Potential, mV

30 20 10

SNPs

0 -10

0

1

2

3

4

5

6

7

8

9

10 11 12 13 EPS

-20 -30 -40 -50

pH

Fig. 5 e a) Effect of pH on the adsorption of EPS on SNPs at an initial concentration of 250 mg/L EPS. (b) Zeta potential of EPS and SNPs at different pH values.

Fig. 4 e (a) The zeta potential and (b) average particle size of SNPs after interaction with 250 mg/L EPS for 72 h.

b shows the zeta potential and average particle size of EPS capped SNPs and uninteracted SNPs over the interaction period. The study suggests that EPS capped SNPs are highly stable and it exhibited the negative zeta potential of greater than
30 mV over the interaction period. But the particle size of uninteracted SNPs increased nearly to 2 mm. EPS are produced by many bacterial species from environmental habitats and are believed to protect bacterial cells by providing abundant binding ligands for nanoparticles (Miao et al., 2009). The EPS can stabilize nanoparticle dispersions and thus may exacerbate the toxicity of nanoparticles (Wilkinson and Reinhardt, 2005).

3.6.

Influence of pH on adsorption

The effect of pH on the adsorption of EPS onto SNPs was evaluated in a pH range of 4e9 and the results are shown in Fig. 5a. It shows that pH had a significant effect on adsorption. With the increase of pH from 4 to 9, the amount of adsorbed EPS on SNPs decreased significantly. A very low adsorption was observed at pH 9. A possible explanation for pH effect on adsorption may be related to the surface charge of nanoparticles and EPS. Fig. 5b shows the schematic illustration of electrostatic interaction between EPS and SNPs at different pH

values. The result shows that the EPS exhibited negative zeta potential in all the pH values. The nanoparticles exhibited positive zeta potential up to pH 8 and beyond that it turned to negative. The zeta potential value agrees well with the adsorption mechanism. With increase in pH, the zeta potential of SNPs decreased; thus the adsorption of EPS on SNPs was also decreased. The possible explanation is that, at pH above 8, the electrostatic force may large due to the negatively charge surface of both SNPs and EPS, and therefore, the electrostatic repulsion does not favor the adsorption of EPS on SNPs. In pH values up to 8, it shows that electrostatic interaction is one of the driving forces for the adsorption of EPS on SNPs. In this region, EPS has a negative charge while SNPs have a positive charge, the electrostatic repulsion between SNPs and EPS is weak, which can promote the adsorption of EPS on SNPs. The pH has a direct influence on zeta potential, and it was reported that the zeta potential of particles decreases with increase in pH (Peng et al., 2004; Jiang et al., 2009). The zeta potential of silica nanoparticles ranges from
20 to
50 mV dependent upon solution pH (Singh and Song, 2007). A pH dependant zeta potential was exhibited by oligochitosan stabilized SNPs, which ranges from 28 mV at pH 2.4 to
56 mV at pH 11 (Long et al., 2007).

3.7.

Influence of salt (NaCl) concentration on adsorption

In addition to pH, salt content is also an important factor affecting the stability of nanoparticles in natural water. Normally the salt concentration in the natural fresh waters found

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 8 4 e5 1 9 0

Equilibrium adsorption quantity, mg/mg solid

160

Appendix. Supplementary data

140

Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.07.024.

120 100

references

80 60 40 20 0 0

0.05

0.1

0.5

1

1.5

NaCl concentrations (M) Fig. 6 e Effect of salt (NaCl) concentrations on adsorption of EPS on SNPs at an initial concentration of 250 mg/L EPS.

to be negligible. The area where tannery effluents are releasing eg:-chromium tanneries, the salt concentration was found to be 0.1e0.2 M. The NaCl concentrations in the costal areas are 1.5 M or above. So we have chosen salt concentration range from 0 to 1.5 M. The effect of salt concentration on the adsorption of EPS on SNPs was studied with an initial concentration of 250 mg/L EPS and the results are shown in Fig. 6. Results show that, when salt concentration was increased from 0 to 0.1 M, a significant change on the adsorption was not observed. However, when NaCl concentration beyond 0.1 M, the adsorption was drastically decreased (Fig. 6). This might be due to the aggregation of SNPs that would result in less surface area to which EPS could adsorb. The average size of SNPs was determined in presence of NaCl. The size of SNPs was increased at increase in concentration of NaCl (Supplementary material Fig. S2). In the presence of EPS the size of the SNPs was almost similar up to 0.1 M concentration of NaCl, beyond that the particle size was increased nearly to 2 mm (Supplementary material Fig. S2).

4.

5189

Conclusion

In conclusions, EPS is an important factor that affects the stability of SNPs in water. The adsorption of EPS can greatly impart negative charge to SNPs surface and increase their surface potential. The environmental pH favors the adsorption of EPS on SNPs. The adsorption process was decreased with increase in pH and salt concentrations. Our study provides a context for understanding the stability of SNPs in natural water bodies and waste water treatment systems.

Acknowledgments Authors thank VIT University Chancellor, for providing us with funding to carry out our research.

AshaRani, P.V., Mun, G.L.K., Hande, M.P., Valiyaveettil, S., 2009. Cytotoxicity and genotoxicity of silver nanoparticles in human cells. ACS Nano 3, 279e290. Benn, T.M., Westerhoff, P., 2008. Nanoparticle silver released into water from commercially available sock fabrics. Environ. Sci. Technol. 42, 7025e7026. Blaser, S., Scheringer, M., MacLeod, M., Hungerbuhler, K., 2008. Exposure of Modeling of Nanosilver in the Environment nanoECO Conference, Monte Verita. Bradford, A., Handy, R.D., Readman, J.W., Atfield, A., Muhling, M., 2009. Impact of silver nanoparticle contamination on the genetic diversity of natural bacterial assemblages in estuarine sediments. Environ. Sci. Technol. 43, 4530e4536. Dasari, T.P., Hwang, H.M., 2010. The effect of humic acids on the cytotoxicity of silver nanoparticles to a natural aquatic bacterial assemblage. Sci. Total Environ. 408, 5817e5823. Davey, M.E., O’Toole, G.A., 2000. Molecular biofilms: from ecology to molecular genetics. Microbiol. Mol. Biol. Rev. 64, 847e867. Dimkpa, C.O., Calder, A., Gajjar, P., Merugu, S., Huang, W., Britt, D. W., McLean, J.E., Johnson, W.P., Anderson, A.J., 2011. Interaction of silver nanoparticles with an environmentally beneficial bacterium, Pseudomonas chlororaphis. J. Hazard. Mater. 188, 428e435. Dubois, M., Gilles, K.A., Hamilton, J.K., Peters, P.A., Smith, F., 1956. Colorimetric method for determination of sugars and related substances. Anal. Chem. 28, 350e356. El-Badawy, A.M., Silva, R.G., Morris, B., Scheckel, K.G., Suidan, M. T., Tolaymat, T.M., 2011. Surface charge-dependent toxicity of silver nanoparticles. Environ. Sci. Technol. 45, 283e287. Fabrega, J., Luoma, S.N., Tyler, C.R., Galloway, T.S., Lead, J.R., 2011. Silver nanoparticles: behaviour and effects in the aquatic environment. Environ. Int. 37, 517e531. Fabrega, J., Renshaw, J.C., Lead, J.R., 2009. Interactions of silver nanoparticles with Pseudomonas putida biofilms. Environ. Sci. Technol. 43, 9004e9009. Ho, Y.S., McKay, G., 1999. Pseudo-second order model for sorption processes. Proc. Biochem. 34, 451e456. Hunga, C.C., Santschia, P.H., Gillow, J.B., 2005. Isolation and characterization of extracellular polysaccharides produced by Pseudomonas fluorescens. Biovar II. Carbohyd. Polym 61, 141e147. Impellitteri, C.A., Tolaymat, T.M., Scheckel, K.G., 2009. The speciation of silver nanoparticles in antimicrobial fabric before and after exposure to a hypochlorite/detergent solution. J. Environ. Qual. 38, 1528e1530. Jiang, W., Mashayekhi, H., Xing, B., 2009. Bacterial toxicity comparison between nano- and micro-scaled oxide particles. Environ. Pollut. 157, 1619e1625. Kaegi, R., Sinnet, B., Zuleeg, S., Hagendorfer, H., Mueller, E., Vonbank, R., Boller, M., Burkhardt, M., 2010. Release of silver nanoparticles from outdoor facades. Environ. Pollut. 158, 2900e2905. Khan, S., Mukherjee, A., Chandrasekaran, N., 2011a. Silver nanoparticles tolerant bacteria from sewage environment. J. Environ. Sci. 23, 346e352. Khan, S.S., Srivatsan, P., Vaishnavi, N., Mukherjee, A., Chandrasekaran, N., 2011b. Interaction of silver nanoparticles (SNPs) with bacterial extracellular proteins (ECP) and its

5190

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 8 4 e5 1 9 0

adsorption isotherms and kinetics. J. Hazard. Mater.. doi:10. 1016/j.jhazmat.2011.05.024. Kumar, A., Sunish Kumar, R., Sakthivel, N., 2003. Compositional difference of the exopolysaccharides produced by the virulent and virulence-deficient strains of Xanthomonas oryzae pv. oryzae. Curr. Microbiol. 46, 251e255. Lagergren, S., 1898. About the theory of so-called adsorption of soluble substances. Kungliga Svenska Vetenskapsakademiens Handlinger 24, 451e465. Long, D., Wu, G., Chen, S., 2007. Preparation of oligochitosan stabilized silver nanoparticles by gamma irradiation. Radiat. Phys. Chem. 76, 1126e1131. Luo, J., Chan, W.B., Wang, L., Zhong, C.J., 2010. Probing interfacial interactions of bacteria on metal nanoparticles and substrates with different surface properties. Int. J. Antimicrob. Ag 36, 549e556. Miao, A., Schwehr, K.A., Xu, C., Zhang, S., Luo, Z., Quigg, A., Santschi, P.H., 2009. The algal toxicity of silver engineered nanoparticles and detoxification by exopolymeric substances. Environ. Pollut. 157, 3034e3041. Neal, A.L., 2008. What can be inferred from bacteriumnanoparticle interactions about the potential consequences of environmental exposure to nanoparticles. Ecotoxicology 17, 362e371. Ng, L.Y., Mohammad, A.W., Leo, C.P., Hilal, N., 2010. Polymeric membranes incorporated with metal/metal oxide nanoparticles: a comprehensive review. Desalination. doi:10. 1016/j.desal.2010.11.033. Nowack, B., Bucheli, T.D., 2007. Occurrence, behavior and effects of nanoparticles in the environment. Environ. Pollut. 150, 5e22. Peng, Z.G., Hidajat, K., Uddin, M.S., 2004. Adsorption of bovine serum albumin on nanosized magnetic particles. J. Colloid Interf. Sci. 271, 277e283. Ravindran, A., Singh, A., Raichur, A.M., Chandrasekaran, N., Mukherjee, A., 2010. Studies on interaction of colloidal Ag

nanoparticles with bovine serum albumin (BSA). Colloid Surf. B. 76, 32e37. Singh, G., Song, L., 2007. Experimental correlations of pH and ionic strength effects on the colloidal fouling potential of silica nanoparticles in crossflow ultrafiltration. J. Membr. Sci. 303, 112e118. Sun, H., Zhang, X., Niu, Q., Chen, Y., Crittenden, J.C., 2007. Enhanced accumulation of arsenate in carp in the presence of titanium dioxide nanoparticles. Water Air Soil Pollut. 178, 245e254. Torino, M.I., Taranto, M.P., Sesma, F., de Valdez, G.F., 2001. Heterofermentative pattern and exopolysaccharide production by Lactobacillus helveticus ATCC 15807 in response to environmental pH. J. Appl. Microbiol. 91, 846e852. Travan, A., Marsich, E., Donati, I., Benincasa, M., Giazzon, M., Felisari, L., Paoletti, S., 2011. Silver-polysaccharide nanocomposite antimicrobial coatings for methacrylic thermosets. Acta Biomater. 7, 337e346. Wigginton, N.S., DeTitta, A., Piccapietra, F., Dobias, J., Nesatyy, V. J., Suter, M.J.F., Latmani, R.B., 2010. Binding of silver nanoparticles to bacterial proteins depends on surface modifications and inhibits enzymatic activity. Environ. Sci. Technol. 44, 2163e2168. Wilkinson, K.J., Reinhardt, A., 2005. Contrasting roles of natural organic matter on colloidal stabilization and flocculation in freshwaters. In: Droppo, I.G., Leppard, G.G., Liss, S.N., Milligan, T.G. (Eds.), Flocculation in Natural and Engineered Environmental Systems. CRC Press, Boca Raton, FL, pp. 143e170. Zhang, Y., Chen, Y., Westerhoff, P., Crittenden, J.C., 2009. Impact of natural organic matter and divalent cations on the stability of aqueous nanoparticles. Water Res. 43, 4249e4257. Zhang, Y., Chen, Y., Westerhoff, P., Hristovski, K., Crittenden, J.C., 2008. Stability of commercial metal oxide nanoparticles in water. Water Res. 42, 2204e2212.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 9 1 e5 1 9 9

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Evaluation of enhanced coagulation pretreatment to improve ozone oxidation efficiency in wastewater Eric C. Wert a, Sarah Gonzales b, Mei Mei Dong b, Fernando L. Rosario-Ortiz b,* a b

Southern Nevada Water Authority (SNWA), P.O. Box 99955, Las Vegas, NV 89193-9955, USA Department of Civil, Environmental and Architectural Engineering, University of Colorado, Boulder, CO 80309, USA

article info

abstract

Article history:

Enhanced coagulation (EC) using ferric chloride was evaluated as a pretreatment process to

Received 16 February 2011

improve the efficiency of ozone (O3) for the oxidation of trace organic contaminants in

Received in revised form

wastewater. At the applied dosages (10e30 mg/L as Fe), EC pretreatment removed between

14 July 2011

10 and 47% of the dissolved organic carbon (DOC) from the three wastewaters studied. Size

Accepted 17 July 2011

exclusion chromatography (SEC) showed that EC preferentially removed higher apparent

Available online 23 July 2011

molecular weight (AMW) compounds. Subsequent O3 testing was performed using an

Keywords:

doses were reduced by 10e47% by the EC pretreatment process. Hydroxyl radical (HO
)

Ozone

exposure, measured by parachlorobenzoic acid (pCBA), showed 10% reduction when using

O3:DOC ratio of 1. Results showed that O3 exposures were similar even though the required

Wastewater

a FeCl3 dose of 30 mg/L, likely due to the lower O3 dose and decreased production of HO


Effluent organic matter (EfOM)

during the initial phase of O3 decomposition (t < 30 s). The oxidation of 13 trace organic

Enhanced coagulation

contaminants (including atenolol, carbamazepine, DEET, diclofenac, dilantin, gemfibrozil,

Ferric chloride

ibuprofen,

Pharmaceuticals

trimethoprim) was evaluated after EC and O3 treatment. EC was ineffective at removing

Water reuse

any of the contaminants, while O3 oxidation reduced the concentration of compounds

meprobamate,

naproxen,

primidone,

sulfamethoxazole,

triclosan,

and

according to their reaction rate constants with O3 and HO
. ª 2011 Published by Elsevier Ltd.

1.

Introduction

The application of ozone (O3) during wastewater treatment has been stimulated by the desire to remove trace organic contaminants and to comply with more stringent disinfection requirements. Several studies have been performed demonstrating the ability of O3 to oxidize a variety of pharmaceuticals, personal care products, and other trace organic contaminants during wastewater treatment (Hollender et al., 2009; Huber et al., 2003, 2005; Wert et al., 2009; Zimmermann et al., 2011). Other studies have documented coliform disinfection by O3 in secondary and tertiary treated effluents (Paraskeva and Graham, 2002). However, the efficiency of O3

processes in wastewater applications is impacted by high concentrations of dissolved organic carbon (DOC), which can result in increased O3 demand, faster O3 decay rates, and scavenging of hydroxyl radicals (HO
) (Rosario-Ortiz et al., 2008; Wert et al., 2009). The DOC in treated wastewater is described as effluent organic matter (EfOM), which is composed of recalcitrant dissolved organic matter (DOM) in addition to soluble microbial products (SMP) from biological treatment (Barker and Stuckey, 1999; Shon et al., 2006). Recent work has shown that the application of O3 for wastewater treatment is impacted not only by the concentration of EfOM, but also by its physicochemical properties. During a study of three

* Corresponding author. Tel.: þ1 303 492 7607. E-mail address: [email protected] (F.L. Rosario-Ortiz). 0043-1354/$ e see front matter ª 2011 Published by Elsevier Ltd. doi:10.1016/j.watres.2011.07.021

5192

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 9 1 e5 1 9 9

cost of additional coagulant usage and solids production versus the reduced energy consumption associated with a lower O3 dose.

Table 1 e Water quality characteristics for the three wastewaters. Water quality Parameter DOC pH UV254 Alkalinity Nitrite Nitrate Ammonia

Secondary wastewater Units

(mgC/L) (1/cm) (mg/L as CaCO3) (mgN/L) (mgN/L) (mgN/L)

Site A

Site B

Site C

8.71 7.25 0.163 87 0.059 4.47 0.062

13.44 7.02 0.181 193 < 0.05 0.061 24.2

7.02 7.21 0.146 62 < 0.05 13.8 0.038

wastewaters, differences in O3 decay rate, HO
exposure and overall oxidation of organic contaminants were attributed to variable properties of the EfOM, since the operational conditions were normalized to the DOC concentration (Wert et al., 2009). Other studies have shown that the HO
scavenging rate due to EfOM varies between wastewaters (Dong et al., 2010; Rosario-Ortiz et al., 2008), which can impact the oxidation efficiency of organic contaminants (Wert et al., 2009). Enhanced coagulation (EC) provides a treatment alternative to remove DOC prior to an ozonation process. EC with ferric chloride (FeCl3) or aluminum sulfate is already a common wastewater treatment process for the removal of particles, nutrients and organics. Studies have shown that EC preferentially removes higher molecular weight dissolved organic matter (DOM) during drinking water treatment (Allpike et al., 2005; Chow et al., 2008; Fabris et al., 2008). However, the effect of EC on the removal of DOC and subsequent effect on O3 and HO
exposure have not been studied during wastewater treatment. The objective of this study was to examine the effect of EC pretreatment to improve the efficiency of an ozonation process targeting the removal of organic contaminants. Three secondary treated wastewaters were evaluated to identify the effect of EC using FeCl3 on the removal of bulk DOC and specific molecular weight fractions. These wastewaters were then used to evaluate subsequent O3 process efficiency. The efficiency was evaluated based upon O3 demand and decay rates, HO
exposure, and oxidation of trace contaminants. The results could lead to site-specific evaluations comparing the

2.

Materials and methods

2.1.

Water quality

Secondary treated effluent samples were collected in 10 L capacity plastic containers from three wastewater treatment facilities (Sites A, B and C), and shipped overnight to the testing facility. Upon arrival, samples were collected for DOC, UV254, alkalinity, nitrite, nitrate, and ammonia measurements. Samples were processed within a week of collection and analytical tests were performed within two weeks. These samples were filtered through a 0.70 mm Whatman Glass Microfibre Filter (Type GF/F, No. 1825-047, Fisher Scientific Inc) prior to analysis. The remaining unfiltered sample was refrigerated at 4  C until coagulation and ozonation tests were performed the following day. Water quality information from the three plants is provided in Table 1.

2.2.

Enhanced coagulation

Enhanced coagulation (EC) experiments were performed using a six-paddle programmable jar testing apparatus (PB-900 Phipps & Bird, Richmond, VA, USA). Wastewater was transferred into a 2 L capacity acrylic jar to conduct the test (BKER2, Phipps & Bird, Richmond, VA, USA). A FeCl3 standard (80 mg Fe/ mL) was prepared from a sample of 40% solution used at a fullscale treatment facility (Kemira Water Solutions, Lawrence, KS, USA). Experiments were performed using dosages of 0, 10, 20, and 30 mg/L as Fe in the three wastewaters. FeCl3 was added and the samples were rapid mixed at 100 rpm for 2 min to promote coagulation, and then mixed at 30 rpm for 20 min to promote flocculation (Westerhoff et al., 2005). Once mixing was complete, the paddle was removed from the beaker and the samples were allowed to settle for up to 30 min. The settled samples were filtered through a 1.5 mm Whatman Glass Microfibre Filter (Type 934-AH, No. 1827-047, Fisher Scientific Inc.) to remove larger floc particles, and filtered again through a 0.70 mm

Table 2 e Water quality changes from FeCl3 coagulation. Waste water Site A

Site B

Site C

FeCl3 Dose (mgFe/L)

DOC (mgC/L)

% DOC Removal

UV254 (1/cm)

% UV254 Removal

SUVA (L/mgC cm)

pH

Alkalinity (mg/L as CaCO3)

0 10 20 30 0 10 20 30 0 10 20 30

8.71 7.82 7.47 6.45 13.44 11.65 10.62 9.69 7.02 6.00 4.77 3.75

0 10 14 26 0 13 21 28 0 15 32 47

0.163 0.143 0.128 0.106 0.181 0.172 0.150 0.134 0.146 0.123 0.102 0.079

0 12 21 35 0 5 17 26 0 16 30 46

1.87 1.83 1.71 1.64 1.35 1.48 1.41 1.38 2.08 2.05 2.14 2.10

7.25 7.13 7.03 6.72 7.02 7.00 6.85 6.68 7.21 7.01 6.71 6.32

87 84 73 64 193 181 172 159 62 48 37 29

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 9 1 e5 1 9 9

Whatman Glass Microfibre Filter (Type GF/F, No. 1825-047, Fisher Scientific Inc.). Samples were collected from the filtrate for DOC, UV254, nitrite, nitrate, ammonia, pH, alkalinity, and organic contaminants. Additional FeCl3 experiments were performed in order to produce a 2 L volume of coagulated, flocculated, and filtered water for subsequent O3 testing.

2.3.

5193

Ozonation at bench-scale

Bench-scale tests were performed using a batch reactor to generate a highly concentrated solution of dissolved O3. MilliQ water (Millipore, Billerica, MA, USA) was placed inside a water-jacketed beaker and chilled to 2  C using

Fig. 1 e SEC results after enhanced coagulation and filtration: (top left) Site A, (top right) Site B, and (bottom) Site C. (Note: Site A 30 mg/L not available).

5194

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 9 1 e5 1 9 9

a recirculating chiller. Once cooled, gaseous O3 was diffused into the water using an oxygen-fed generator (CFS-1A, Ozonia North America, Inc., Elmwood Park, NJ). The O3 stock solution concentrations and dissolved O3 residuals were measured using potassium indigotrisulfonate (APHA et al. 1998). The concentration of the stock solution was consistently around 85 mg/L of dissolved O3. During the testing, an aliquot of O3 stock solution was transferred into a wastewater sample inside a 1 L amber glass container with a repeating pipet dispenser. The O3 decay curves were generated by dispensing water from inside the reactor into 125 mL Erlenmyer flasks containing indigo solution at different time intervals. During the initial phase of ozonation (t < 30 s), the instantaneous O3 demand (IOD) was calculated as the difference between the applied dose and the dissolved residual at 30 s. During the second phase of ozonation (t > 30 s), the O3 exposure (CT) was calculated by integrating the dissolved residual concentration over time (!O3dt) (Rakness et al., 2005). Once the dissolved O3 residual had decayed to less than 0.05 mg/L, samples were collected for EfOM characterization, pH, UV254, and quantification of organic contaminants. Duplicate experiments were performed with the addition of 115 mg/L (0.74 mM) parachlorobenzoic acid (pCBA). The probe pCBA was used as a HO
exposure indicator since it reacts slowly with O3 and rapidly with HO
(Elovitz and von Gunten, 1999). During these experiments, O3 residual and pCBA samples were collected until the O3 residual had decayed to less than 0.05 mg/L. Samples for pCBA analysis were quenched with an aliquot of sodium thiosulfate. The Rct, defined as the ratio of HO
exposure to O3 exposure (!HO
dt/!O3dt), was measured for each testing condition to determine whether there were any changes in the yield of HO
(Elovitz and von Gunten, 1999). Precision tests for the quantification of Rct and ozone decay rate (kO3) were performed, yielding a relative standard deviation (RSD) of 4% and 7% respectively.

2.4.

254 nm. The mobile phase consisted of a phosphate buffer (0.0024 M NaH2PO4, 0.0016 M Na2HPO4 and 0.025 M Na2SO4) adjusted to pH 6.8  0.1. The flow rate was operated at 1.0 mL/min. A modified commercially available Sievers-800 DOC analyzer (General Electric, CO, USA) with 1.5 mL/min acid and oxidizer flow rate was used to monitor the DOC elution from the SEC column. An Agilent interface (model 35900e Palo Alto, CA, USA) was used to record voltage output from the DOC. The voltage output was linearly correlated to DOC analyzer signal. Polyethylene glycols (Fluka, Milwaukee, WI, USA) were used for AMW calibration.

3.

Results and discussion

3.1.

Enhanced coagulation

FeCl3 dosages of 10, 20, and 30 mg/L as iron removed DOC in wastewater samples from Site A (10e26%), Site B (13e28%), and Site C (15e47%) as shown in Table 2. Removal of DOC

Analytical methods

Water samples were collected, preserved, and refrigerated until analyzed. Water quality analysis was performed following standard methods (APHA et al., 1998). The concentrations of O3 in the stock solution and dissolved residuals were measured according to the indigo method described in Standard Methods 4500-O3 (APHA et al., 1998). Potassium indigotrisulfonate (SigmaeAldrich, St. Louis, MO USA) was used to generate indigo solutions with molar absorptivity of 20,000 M1cm1. The analysis of the organic contaminants was performed using liquid chromatography with tandem mass spectrometry (Trenholm et al., 2009), with a RSD of below 5.7%. The quantification of pCBA was performed using a HPLC with absorbance quantification, with minimum reporting limit of 5 mg/L, and an RSD of 1.7%. The characterization of the apparent molecular weight (AMW) distribution of the EfOM was performed using size exclusion chromatography (SEC) with UV254 and DOC quantification. An Agilent 1200 LC system (Palo Alto, CA, USA) with a Toyopearl HW-50 S 250  20 mm column (Chromatography, Rottenburg, Germany) was used with an injection volume of 2.0 mL. The detector consisted of a diode array from Agilent (Model 1200 Palo Alto, CA, USA) monitoring at a wavelength of

Fig. 2 e Dissolved O3 residual (DO3) decay after EC pretreatment with FeCl3: (a) Site A, (b) Site B, and (c) Site C.

5195

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 9 1 e5 1 9 9

Table 3 e Summary of O3 demand, exposure, and decay and Rct for the various testing conditions. Waste water Site A

Site B

Site C

FeCl3 Dose (mgFe/L)

O3 Dose (mgO3/L)

IOD (mgO3/L)

IOD (mgO3/mgC)

O3 CT (mgO3 min/L)

kO3 (1/sec)

kO3 (L/mgC sec)

Rct (108)

0 10 20 30 0 10 20 30 0 10 20 30

8.71 7.82 7.47 6.45 13.44 11.65 10.62 9.69 7.02 6.00 4.77 3.75

5.35 5.02 4.15 3.38 e 5.83 5.85 4.95 4.02 3.11 2.55 2.05

0.61 0.64 0.55 0.52 e 0.50 0.55 0.51 0.57 0.52 0.53 0.54

3.19 2.16 4.63 4.53 5.32 5.69 4.72 5.01 3.29 3.54 2.74 2.61

0.013 0.016 0.009 0.009 0.012 0.012 0.012 0.011 0.012 0.011 0.011 0.009

0.0015 0.0020 0.0013 0.0014 0.0009 0.0011 0.0011 0.0012 0.0018 0.0019 0.0023 0.0024

3.30 3.02 1.34 1.82 1.80 1.39 1.89 1.65 3.96 3.63 3.50 3.75

corresponded well with UV254 removal for all dosages and waters. Interestingly, SUVA values had very little response to the coagulant dose indicating the aromaticity of the wastewater was unchanged by coagulation. This contrasts with published information in drinking water treatment, which shows a decrease in aromaticity during enhanced coagulation (Chow et al., 2009). Other water quality changes associated with FeCl3 coagulation were reductions in pH and alkalinity. The differences in the percent removal of DOC among the three wastewaters were further investigated by characterizing the EfOM using SEC, with UV254 and DOC quantification. Fig. 1 presents the SEC chromatograms for the three wastewaters as a function of FeCl3 dose. The chromatograms were characterized by three regions, which have been described as organic colloids and polysaccarides (AMW > 10 kDa), humic-like susbtances (10 > AMW > 1 kDa) and low molecular weight acids (AMW < 1 kDa) in decreasing order of AMW (Nam et al., 2008). Results from Site C (Fig. 1) indicated that EC removed primarily the high AMW components of the EfOM. This removal was observed with both UV254 and DOC quantification, although it was more dramatic for the former. The DOC response decreased greatly in the high AMW range (approx. 40 kDa)

compared to the UV254 signal, indicating preferential removal of these larger fragments, which are also characterized by low aromaticity. The SEC chromatogram for Site A showed a substantial decrease in the UV254 signal, that was more homogeneous throughout the AMW range studied (Fig. 1). The decrease in the DOC response for this sample was less drastic, indicating that EC preferentially removed aromatic structures. In the case of Site B, there was a decrease in the high AMW as shown in the UV254 chromatogram, although the DOC response showed less removal. The SEC chromatograms for the samples from the three sites showed that EC preferentially removed the polysacchride and humic substance regions of the EfOM. The fraction of the EfOM with AMW less than 1 kDa was more recalcitrant, although the Site A sample showed removal in these regions. These findings are supported by other studies examining EC for drinking water (Allpike et al., 2005; Chow et al., 2008; Fabris et al., 2008). Chow et al., (2008) found larger AMW organic matter were easily removed with conventional EC processes and lower AMW compounds were nonremovable for the drinking water samples studied. In another study, Fabris et al. (2008) investigated efficacy of EC

Fig. 3 e Degradation of pCBA during O3 experiments. Error bars represent ± RSD (1.7%).

5196

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 9 1 e5 1 9 9

Fig. 4 e SEC plots for Site C. Secondary effluent before and after ozone (left) and secondary effluent, post coagulation (30 mg/L) and post ozone (right).

for NOM removal in two raw water sources, with one being more humic in nature than the other. The SEC UV signatures for recalcitrant NOM after EC for the two treated waters were similar, indicating the humic-like NOM was removed relatively easily. These findings regarding NOM coagulation were similar to EfOM behavior during this study, and explain the differences in the % removal of DOC among the three wastwaters.

3.2.

Ozone and hydroxyl radical exposure

Since EC treatment reduced the pH of the wastewater, the coagulated wastewater samples were pH adjusted to 7.16  0.02 using 0.5 M NaOH before bench-scale testing with O3. An O3:DOC ratio of 1.0 was used for all bench-scale tests. The O3 demand exerted by nitrite was neglected since the reported concentrations were near or below the detection limit (<0.05 mg/L). Therefore the O3 dose was equivalent to the DOC concentration after EC as shown in Table 2. Dissolved O3 residual decayed to less than 0.1 mg/L within 7 min in all three wastewaters (Fig. 2). The IOD, O3 CT, and kO3 are shown in Table 3. Normalized values for the IOD and kO3 are also shown to account for the differences in DOC concentration after EC. The IOD varied significantly depending on the wastewater site and EC testing conditions (the value of the IOD for bulk Site B was not obtained). When normalized on a per carbon basis, the IOD was consistently between 0.50 and 0.64 mgO3/mgC and did not change by more than 10%. Therefore, EC pretreatment provided a benefit in reducing the

overall IOD exerted by EfOM in the three wastewaters. CT values varied by 53%, 17%, and 26% at Sites A, B, and C, respectively, primarily due to differences in ozone exposure within the initial 2 min of reaction time (Fig. 2). The values for kO3 were similar as a function of the FeCl3 dose. The normalized values also remained constant with the exception of Site C, which saw a modest increase. Accounting for the different DOC concentrations, the O3 decay rate increased in the samples from Site B and Site C wastewaters as the FeCl3 dose increased. The HO
exposure was indirectly measured via the probe compound pCBA. Fig. 3 shows the decrease in pCBA concentration after 30 s, 2 min, and 6 min. These results show that most of the HO
oxidation occured during the initial phase of ozonation (t < 30 s). This is consistent with other research showing that the initial phase of ozonation is an advanced oxidation process (Buffle et al., 2006; Buffle and von Gunten, 2006). Furthermore, the amount of pCBA degradation decreased by approximately 10% as the FeCl3 dose increased to 30 mg/L. The reduction in HO
was attributed to a reduced O3 dose in addition to changes in the chemistry of the initial phase of ozonation due to modification of the EfOM composition. The Rct values were calculated from the O3 and pCBA decay rates to assess oxidation efficiency during the secondary phase of ozonation (t > 30 s) (Elovitz and von Gunten, 1999). The values for Rct remained constant for the conditions tested, as shown in Table 3, with the exception of the sample from Site A, where a reduction of approximately 50% was observed

5197

620 600 640 640 620 630 670 670 540 560 550 550 96 86 83 88 170 160 150 170 68 90 84 91 1300 1300 1300 1300 960 990 980 1000 810 810 870 890 140 130 130 130 120 110 110 110 73 68 80 80 120 120 110 110 380 380 350 330 <25 <25 <25 <25 270 280 270 280 230 230 240 230 120 130 130 130 <25 <25 <25 <25 87 91 99 90 <25 <25 <25 <25 510 490 500 520 2800 2700 2700 2600 170 170 170 170 220 210 200 220 150 150 160 150 71 73 81 75 150 150 140 140 200 200 210 190 89 78 81 82 100 100 110 99 480 490 470 500 47 47 44 43 320 330 310 320 250 260 290 260 200 200 190 200 1900 1600 1600 1600 2200 2200 2100 2300 450 370 420 420 Site C

Site B

Results from trace contaminant analysis showed poor removal of the detected compounds by EC treatment (Table 4). Musk ketone and atrazine were also analyzed but concentrations were below their respective reporting limits of 100 ng/L and 25 ng/L in the three wastewaters. These results coincided with other research that showed coagulation using aluminum

0 10 20 30 0 10 20 30 0 10 20 30

Trace organic contaminant removal

Site A

3.3.

Waste FeCl3 Atenolol Carbamazepine DEET Diclofenac Dilantin Gemfibrozil Ibuprofen Meprobamate Naproxen Primidone Sulfamethoxazole Triclosan Trimethoprim water Dose (mg/L as Fe)

after the application of 20e30 mg/L of FeCl3. Therefore, the wastewater from Site A experienced a lower HO
exposure per unit of O3 exposure as the FeCl3 dose increased. However there was minimal change in the Rct values calculated in the wastewaters from Sites B and C. These results indicate that the removal of EfOM by EC did not significantly change the dynamics of O3 decomposition or production of HO
in the second phase for these two wastewaters. This indicates that EC pretreatment had greater importance during the initial phase of ozonation (t < 30 s) rather than the secondary phase of ozonation (t > 30 s). Fig. 4 presents the SEC chromatograms for Site C before and after ozone for the non-coagulated water and the coagulated water (30 mg/L of FeCl3). The coagulation step was very effective at removing the higher AMW compounds. The application of ozone resulted in a shift toward lower AMW compounds as shown in Fig. 4b, although no significant mineralization of the EfOM was observed. These results indicate that the remaining higher AMW compounds were broken down into lower AMW compounds. The overall aromaticity of the EfOM also decreased according to the UV response, which was expected since O3 is known to preferentially react with aromatic groups (Westerhoff et al., 1999). The EfOM changes by EC can be related to the O3 and HO
exposure results presented previously. While maintaining an O3:DOC ratio of 1, the O3 dose was reduced by 10e47% while maintaining similar O3 and HO
exposures (Rct) during the secondary phase of ozonation (t > 30 s). These results indicate that the removal of high AMW components of the EfOM by EC had the greatest effect during the initial phase of ozonation (t < 30 s). During this initial phase, the majority of HO
exposure occurred, and the pCBA results showed a 10% decrease in HO
exposure as the FeCl3 dose was increased to 30 mg/L. Therefore, the removal of high AMW components from the EfOM by EC resulted in less formation of HO
during the initial phase of ozonation (t < 30 s). These results also coincide with previous reports showing that a wastewater containing higher AMW components exerted a greater O3 demand than other wastewaters containing a lower proportion of high AMW components when evaluated on an equivalent O3:DOC basis (Wert et al., 2009). The EfOM results also showed the transformation of high AMW compounds (40 kDa) into low AMW compounds by O3 (1e3 kDa). Based upon the previous discussion, most of this transformation was expected to take place rapidly during the initial phase of ozonation (t < 30 s). Therefore, the low AMW compounds (<3 kDa) remain during the second phase of ozonation (t > 30 s), which is dominated by radical chain reactions with either hydroxide or other specific NOM moieties (Buffle and von Gunten, 2006).

Table 4 e Trace contaminant concentration (in ng/L) after EC with FeCl3. The concentrations of ibuprofen were below detection limit for Sites A and C; for naproxen, the concentrations were below the detection limit for Site C.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 9 1 e5 1 9 9

5198

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 9 1 e5 1 9 9

Fig. 5 e Concentrations of meprobamate (MRL [ 10 ng/L) and DEET (MRL [ 25 ng/L) remaining after O3 oxidation. (Note: DEET concentrations below the MRL are shown with a hollow bar). sulfate removed < 20% of most trace contaminants from wastewater (Reungoat et al., 2010). During the O3 experiments, nine of the thirteen trace contaminants were removed to below their reporting limits after ozonation in all three wastewaters including diclofenac (<25 ng/L), gemfibrozil (<10 ng/L), ibuprofen (<25 ng/L), naproxen (<25 ng/L), triclosan (<25 ng/L), atenolol (<25 ng/L), carbamazepine (<10 ng/L), sulfamethoxazole (<25 ng/L), and trimethoprim (<10 ng/L). With the exception of ibuprofen (k00O3 ¼ 9.6 M1s1), these compounds are fast reacting with O3 with reaction rate constants greater than 103 M1s1 (Benner et al., 2008; Dodd et al., 2006; Huber et al., 2003, 2005; Suarez et al., 2007). DEET, meprobamate, dilantin, and primidone are all approximated to have a k00O3 < 10 M1s1 based on previous work by the authors (Wert et al., 2009). Therefore, the removal of these compounds becomes dependent upon HO
exposure. Dilantin and primidone had a few detectable concentrations after ozonation, which were near their reporting limit of 10 ng/L. Post O3 concentrations of meprobamate and DEET are shown in Fig. 5. The observed removal of these compounds was a function of HO
exposure, with a decrease in removal as the FeCl3 dose increased. These results are consistent with pCBA data previously reported in Fig. 3, and can be attributed to the lower O3 dose requirement and less HO
production during the initial phase of ozonation.

4.

Conclusions


EC using FeCl3 reduced the DOC concentration by 10e47% in the three wastewaters evaluated. The removal of DOC was linear with the applied FeCl3 dose and UV254. Although coagulant aids were not evaluated in this research, they may have improved the DOC removal efficiency of the coagulation/flocculation process.
SEC results indicated EC preferentially removed high AMW components of the EfOM in all three wastewaters.
The reduction in DOC concentration also reduced the IOD in all three wastewaters depending on the FeCl3 dose. When

normalized on a per carbon basis, the IOD was consistently between 0.50 and 0.64 mg O3/mg C. Therefore, EC pretreatment provided a benefit in reducing the IOD exerted by EfOM during the initial phase of ozonation (t < 30 s).
The O3 decay rates and CTs were similar for all samples before and after EC. The HO
exposures were reduced by up to 10%, even though the O3 dose was reduced by as much as 47%. The Rct values, which represent the formation of HO
per unit of O3 exposure during the second phase of O3 decomposition, did not change for two of the samples after EC. For the third sample, the value of RCT decreased by 50%.
Results from trace contaminant analysis showed poor removal of the detected compounds after EC treatment.
All of the organic contaminants that react rapidly with O3 (k00O3 > 103 M1s1) were removed below the detection limit during all conditions studied. For the compounds that react primarily with HO
, removal was a function of HO
exposure and coincided with pCBA results.

Acknowledgments The authors acknowledge the National Science Foundation (project number 0926396) for support for Mei Mei Dong and the Department of Civil, Environmental and Architectural Engineering at the University of Colorado, Boulder for partial support for Sarah Gonzales. The authors also thank the following staff members within the SNWA Water Quality Research and Development team: Samantha Stoughtenger for her assistance during the O3 testing, and Rebecca Trenholm for performing the analysis of pharmaceuticals. Finally, the authors acknowledge the contributions of participating utilities.

references

Allpike, B.P., Heitz, A., Joll, C.A., Kagi, R.I., Abbt-Braun, G., Frimmel, F.H., Brinkmann, T., Her, N., Amy, G., 2005. Size exclusion chromatography to characterize DOC removal in drinking water treatment. Environmental Science and Technology 39 (7), 2334e2342. APHA, AWWA and WEF, 1998. Standard Methods for the Examination of Water and Wastewater. American Public Health Association; American Water Works Association; and Water Environment Federation, Washington D.C. Barker, D.J., Stuckey, D.C., 1999. A review of soluble microbial products (SMP) in wastewater treatment systems. Water Research 33 (14), 3063e3082. Benner, J., Salhi, E., Ternes, T., von Gunten, U., 2008. Ozonation of reverse osmosis concentrate: kinetics and efficiency of beta blocker oxidation. Water Research 42 (12), 3003e3012. Buffle, M.-O., Schumacher, J., Salhi, E., Jekel, M., von Gunten, U., 2006. Measurement of the initial phase of ozone decomposition in water and wastewater by means of a continuous quenchflow system: application to disinfection and pharmaceutical oxidation. Water Research 40 (9), 1884e1894. Buffle, M.O., von Gunten, U., 2006. Phenols and amine induced HO generation during the initial phase of natural water ozonation. Environmental Science and Technology 40 (9), 3057e3063. Chow, C.W.K., Fabris, R., van Leeuwen, J., Wang, D.S., Drikas, M., 2008. Assessing natural organic matter treatability using high

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 9 1 e5 1 9 9

performance size exclusion chromatography. Environmental Science & Technology 42 (17), 6683e6689. Chow, C.W.K., van Leeuwen, J.A., Fabris, R., Drikas, M., 2009. Optimised coagulation using aluminium sulfate for the removal of dissolved organic carbon. Desalination 245 (1e3), 120e134. Dodd, M.C., Buffle, M.-O., von Gunten, U., 2006. Oxidation of antibacterial molecules by aqueous ozone: Moiety-specific reaction kinetics and application to ozone-based wastewater treatment. Environmental Science and Technology 40 (6), 1969e1977. Dong, M.M., Mezyk, S.P., Rosario-Ortiz, F.L., 2010. Reactivity of effluent organic matter (EfOM) with hydroxyl radical as a function of molecular weight. Environmental Science and Technology 44 (15), 5714e5720. Elovitz, M.S., von Gunten, U., 1999. Hydroxyl radical ozone ratios during ozonation processes. I-The R-ct concept. Ozone Science and Engineering 21 (3), 239e260. Fabris, R., Chowa, C.W.K., Drikas, M., Eikebrokk, B., 2008. Comparison of NOM character in selected Australian and Norwegian drinking waters. Water Research 42 (15), 4188e4196. Hollender, J., Zimmermann, S.G., Koepke, S., Krauss, M., McArdell, C.S., Ort, C., Singer, H., von Gunten, U., Siegrist, H., 2009. Elimination of organic micropollutants in a municipal wastewater treatment plan upgraded with a full-scale postozonation followed by sand filtration. Environmental Science and Technology 43 (20), 7862e7869. Huber, M.M., Canonica, S., Park, G.-Y., von Gunten, U., 2003. Oxidation of pharmaceuticals during ozonation and advanced oxidation processes. Environmental Science and Technology 37 (5), 1016e1024. Huber, M.M., Gobel, A., Joss, A., Hermann, N., Loffler, D., Mcardell, C.S., Ried, A., Siegrist, H., Ternes, T.A., von Gunten, U., 2005. Oxidation of pharmaceuticals during ozonation of municipal wastewater effluents: a pilot study. Environmental Science and Technology 39 (11), 4290e4299. Nam, S.-N., Krasner, S.W., Amy, G.L., 2008. Differentiating effluent organic matter (EfOM) from natural organic matter (NOM): impact of EfOM on drinking water sources. Advanced Environmental Monitoring, 259e270. Paraskeva, P., Graham, N.J.D., 2002. Ozonation of municipal wastewater effluents. Water Environment Research 74 (6), 569e581.

5199

Rakness, K.L., Najm, I., Elovitz, M., Rexing, D., Via, S., 2005. Cryptosporidium log-inactivation with ozone using effluent CT10, geometric mean CT10, extended integrated CT10 and extended CSTR calculations. Ozone Science and Engineering 27 (5), 335e350. Reungoat, J., Macova, M., Escher, B.I., Carswell, S., Mueller, J.F., Keller, J., 2010. Removal of micropollutants and reduction of biological activity in a full scale reclamation plant using ozonation and activated carbon filtration. Water Research 44 (2), 625e637. Rosario-Ortiz, F.L., Mezyk, S.P., Doud, D.F.R., Snyder, S.A., 2008. Quantitative correlation of absolute hydroxyl radical rate constants with non-isolated effluent organic matter bulk properties in water. Environmental Science and Technology 42 (16), 5924e5930. Shon, H.K., Vigneswaran, S., Snyder, S.A., 2006. Effluent organic matter (EfOM) in wastewater: Constituents, effects, and treatment. Critical Reviews in Environmental Science and Technology 36 (4), 327e374. Suarez, S., Dodd, M.C., Omil, F., von Gunten, U., 2007. Kinetics of triclosan oxidation by aqueous ozone and consequent loss of antibacterial activity: Relevance to municipal wastewater ozonation. Water Research 41, 2481e2490. Trenholm, R.A., Vanderford, B.J., Snyder, S.A., 2009. On-line solid phase extraction LC-MS/MS analysis of pharmaceuticals indicators in water: a green alternative to conventional methods. Talanta 79 (5), 1425e1432. Wert, E.C., Rosario-Ortiz, F.L., Snyder, S.A., 2009. Effect of ozone exposure on the oxidation of trace organic contaminants in wastewater. Water Research 43, 1005e1014. Westerhoff, P., Aiken, G., Amy, G., Debroux, J., 1999. Relationships between the structure of natural organic matter and its reactivity towards molecular ozone and hydroxyl radicals. Water Research 33 (10), 2265e2276. Westerhoff, P., Yoon, Y., Snyder, S., Wert, E., 2005. Fate of endocrine-disruptor, pharmaceutical, and personal care product chemicals during simulated drinking water treatment processes. Environmental Science & Technology 39 (17), 6649e6663. Zimmermann, S.G., Wittenwiler, M., Hollender, J., Krauss, M., Ort, C., Siegrist, H., von Gunten, U., 2011. Kinetic assessment and modeling of an ozonation step for full-scale municipal wastewater treatment: Micropollutant oxidation, by-product formation and disinfection. Water Research 45 (2), 605e617.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 0 0 e5 2 1 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Fast and highly efficient removal of dyes under alkaline conditions using magnetic chitosan-Fe(III) hydrogel Chensi Shen a, Yu Shen b, Yuezhong Wen a,c,*, Hongyu Wang b, Weiping Liu a,c a

Institute of Environmental Science, College of Environmental and Resource Sciences, Zhejiang University, Hangzhou 310058, China College of Civil Engineering and Architecture, Zhejiang University of Technology, Hangzhou 310032, China c Zhejiang Provincial Key Laboratory of Organic Pollution Process and Control, Zhejiang University, Hangzhou 310058, China b

article info

abstract

Article history:

The dyeing effluent of high alkalinity, which could not be treated efficiently by traditional

Received 23 February 2011

wastewater technologies, highlighted the need to explore a technically feasible, highly

Received in revised form

efficient and cost effective method. Thus, a fast and highly efficient method for the

15 May 2011

removal of dyes under alkaline conditions using magnetic chitosan-Fe(III) hydrogel was

Accepted 17 July 2011

proposed. Firstly, chitosan-Fe(III) hydrogel was prepared by a chelation procedure with

Available online 26 July 2011

cheap and environmentally friendly chitosan and iron salts. We characterized the sorption and desorption of C. I. Acid Red 73, a common type of anionic dye, on magnetic chitosan-

Keywords:

Fe(III) hydrogel, to understand its availability for alkaline dyeing wastewater. Sorption of

Magnetic

dye to chitosan-Fe(III) hydrogel was fast (adsorption could reach equilibrium in less than

Chitosan-Fe(III) hydrogel

10 min) in a wide pH range, and agreed well to the LangmuireFreundlich adsorption model

Removal

with a high maximum adsorption capacity of 294.5 mg/g under pH ¼ 12. Meanwhile, 1 mol/L

Dyes

NaOH was used to desorb the dye efficiently (desorption efficiency 94.4%) and 0.1 mol/L HCl

Alkaline

was applied to regenerate the chitosan-Fe(III) hydrogel. The results showed that the chitosan-Fe(III) hydrogel could retain its high efficiency after the desorption and regeneration. The common coexisting ions almost had no negative effect on the dye adsorption of chitosan-Fe(III) and the removals of a variety of anionic dyes suggest that the magnetic chitosan-Fe(III) hydrogel could efficiently adsorb both the acid and reactive dyes under alkaline condition. Overall, the results reported herein indicated that magnetic chtisoanFe(III) with high adsorption efficiency and strong magnetic property is very attractive and implies a potential of practical application for alkaline dyeing effluent treatment. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Many industries, such as textile, dyeing, paper and pulp, tannery and paint industries are big consumers of dyes, and hence the effluents of these industries tend to contain dyes in excessive quantities (Gupta and Suhas, 2009). The serious environmental problems caused by large amounts of dyeing

wastewater have caused significant concern because of its strong color, low ratio of BOD5/COD (around 20%), and the bio-recalcitrant components such as dyes and dyeing additives (Crini, 2006). Importantly, the dyeing effluents are often of high alkalinity depending on the processes used, such as the mercerization of cottons and caustic reduction of polyester fabrics (Choe et al., 2005). For instance, the pH of the

* Corresponding author. Institute of Environmental Science, College of Environmental and Resource Sciences, Zhejiang University, Hangzhou 310058, China. Tel./fax: þ86 571 8898 2344. E-mail address: [email protected] (Y. Wen). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.018

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 0 0 e5 2 1 0

dyeing discharge from textile industrial generally ranges from 9 to 11 (Pelegrini et al., 1999; Vlyssides et al., 2000; Ko¨rbahti and Tanyolac¸, 2008), and sometimes the pH of alkali reduction wastewater from artificial silk industry can even rise to more than 12 (Fan et al., 2007). Although a wide range of methods has been developed for the removal of synthetic dyes from wastewater to decrease their impact on the environment, such as adsorption (Chatterjee et al., 2005; Shen et al., 2011), chemical oxidation (Shen et al., 2010), microbiological or enzymatic decomposition (Wu et al., 2007), electrochemical treatment (Gupta et al., 2007), etc., some limitations still exist. Because of the high alkalinity, most of dyeing wastewater cannot be treated directly by these current treatment processes. And the pretreatments before the influents enter into traditional treatments system were inconvenient and even produce large amounts of solid wastes due to the use of chemicals for pH adjustment and coagulation (Selcuk, 2005). In order to solve this problem, ultrafiltration, nanofiltration and hyperfiltration were investigated to recover the dyes, chemicals, water, alkali, and energy from textile wastewater (Avlonitis et al., 2008). Although these methods are efficient for the treatment of dyeing waters, they are very costly and commercially unattractive. Additionally, alkali bacterial consortium was obtained by enrichment cultivation and was used to treat printing and dyeing wastewater (pH ¼ 11e12), but it is constrained by sensitivity toward toxicity of some chemicals (Yang et al., 2011). Meanwhile, adsorption has been found to be superior to other techniques for water reuse in terms of initial cost, flexibility and simplicity of design, ease of operation and insensitivity to toxic pollutants (Crini and Badot, 2008). Xu et al. proposed the conditioning of the chitosan beads with ammonium sulfate and applied it to removal the dyes under alkaline conditions (Xu et al., 2008). Its maximum adsorption capacity is about 80 mg/g, which is quite lower compared with the same adsorbent under acidic conditions. Thus, there is an urgent need for technically feasible, highly efficient and cost effective wastewater treatment technologies in this industry. Recently, some metal-binding biopolymers have received a great deal of attention, according to their chemical stability and high capacity in adsorption processes (Jang et al., 2007; Yoshitake et al., 2003; Wang et al., 2007; Cheng et al., 2010). It has been reported that iron-chitosan composites and Fe-crosslinked chitosan complex were applied to remove Cr(VI), As(III) and As(V) in wastewater (Nieto et al., 1992; Klepka et al., 2008). Cu(II) complex of dithiocarbamate modified starch was proposed to efficiently adsorb ionic dyes through strong chelating interactions. Nevertheless, their high adsorption efficiencies were also confined to acidic or neutral conditions. If the chelating interactions between dye molecules and metal complex center can be utilized in alkaline conditions, the naturally abundant biopolymer will imply attractive potential on dye removal process. In this study, a fast and highly efficient method for the removal of dyes under alkaline conditions using magnetic chitosan-Fe(III) hydrogel was proposed and by chelation with cheap and environmentally friendly chitosan and iron salts, a novel low-cost magnetic sorbent material was prepared. The applicability of magnetic chitosan-Fe(III) in alkaline

5201

dyeing effluents treatment was evaluated in view of the sorption kinetic and capacity, effects of the common coexisting ions, as well as the reuse of adsorbents. And the possible mechanism proposed herein may be also used to account for the high efficiency of dyes adsorption under alkaline conditions.

2.

Experimental

2.1.

Chemicals

FeCl3$6H2O, glutaraldehyde, and ethanol were obtained from National Medicines Corporation Ltd. of China. Chitosan was purchased from Zhejiang Golden-shell Biochemical Co., Ltd, Zhejiang, China (deacetylation degree ¼ 91.04%). Fe3O4 nano particles were obtained from Beijing Nachen S&T Ltd., China (OD ¼ 20 nm, purity>99.9%). C. I. Acid Red 73 (AR 73) and other dyes were commercial products (chemical structures are shown in Figure S1 in Supporting Information). Doubly distilled water was used throughout this study. Other chemicals were of laboratory reagent grade and used without further purification.

2.2.

Preparation of chitosan-Fe(III) hydrogel

The magnetic chitosan-Fe(III) hydrogel was prepared by modifying the literature procedure (Klepka et al., 2008). Chitosan powder (1.0 g) was dissolved in 0.1 M FeCl3 aqueous solution (50 mL), and the mixture was stirred at room temperature for 4 h. Subsequently, the Fe3O4 nano particles were added into the mixture and dispersed sufficiently. Next, the magnetic chitosan-Fe(III) complex precipitation was obtained after the addition of ethanol. Besides, the solid was washed with ethanol to remove an excess of FeCl3 and dried at 80  C. Then, it was placed in ethanol solution in contact with 5% of glutaraldehyde for 2 h, according to the chemical crosslinking of chitosan with glutaraldehyde occurs by Schiff’s reaction. Once again, the magnetic chitosan-Fe(III) hydrogel after chemical crosslinking was collected by magnetic separation and further washed with distilled water in order to remove the unreacted glutaraldehyde. At last, it was dried at 80  C and the chitosan-Fe(III) hydrogel was finally prepared. Due to that the Fe3O4 also contain a large amount of Fe(III), the non-magnetic hydrogel was applied to investigate this efficient adsorbent in order to avoid the effects brought by Fe3O4 nano particles. It was prepared according to the literature procedures (Klepka et al., 2008). The amount of Fe(III) in the chitosan-Fe(III) without Fe3O4 was 10.95% as determined by atomic absorption spectrometry on a Thermo Scientific iCE 3300 AA spectrometer.

2.3.

Adsorption and desorption procedures of dyes

All the dye adsorption experiments were performed in 50 mL flasks, which were sealed and agitated at 100 rpm in a thermostatic shaker maintained at 25  C. The typical reaction mixture was initiated with 10 mL of dye at 50 mg/L and 0.02 g of chitosan-Fe(III) hydrogel at the condition of pH ¼ 12. The alkaline condition was simulated with 0.01 mol/L NaOH

5202

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 0 0 e5 2 1 0

solution. In some cases, the initial pH value of the dye solution were adjusted to 6 to 12 by addition of 0.01 mol/L HCl and 0.01 mol/L NaOH, respectively. After the adsorption procedure, the dye-loaded non-magnetic chitosan-Fe(III) hydrogel was separated by centrifugation. In contrast, the magnetic chitosan-Fe(III) hydrogel with adsorbed dye could be easily separated by a magnetic field. For chitosan desorption experiments, the used adsorbents were resuspended in 25 mL of different eluents at 25  C for 30 min and separated by a magnetic field or centrifugation. Dye concentrations were analyzed using a Shimadzu UV-2401PC UVevis spectrometer (Tokyo, Japan) at its absorbance maximum.

2.5.

Surface morphology was studied with an electron microscope. The scanning electron micrographs (SEMs) of composite chitosan-Fe(III) hydrogel was obtained with a field emission scanning electron microscope (FEI, SIRON) at a voltage of 25.0 kV to test. The sample surfaces were gold-coated before analysis. The TGA studies were performed with Pyris TGA instruments (PerkineElmer, US) at a heating rate of 10  C/min under a N2 atmosphere.

3. 2.4.

Results and discussion

Adsorption isotherms measurements

The adsorption isotherms of dye on chitosan-Fe(III) in water were carried out using the batch slurry method. The slurry, containing 0.02 g of chitosan and 10 mL of AR 73 solution at various concentrations, was agitated at 100 rpm in a thermostatic shaker until equilibrium was reached at temperatures of 25  C, respectively. The amount of adsorbed AR 73, qe, was calculated by Eq. (1):

qe ¼

Characterization of the chitosan-Fe(III) hydrogel

ðC0  Ce ÞV M

(1)

where qe is the dye capacity in the sorbent at equilibrium (mg/g), C0 is the initial dye concentration in the liquid-phase (mg/L), Ce is the liquid-phase dye concentration at equilibrium (mg/L), V is the volume of solution (L), and M is the mass of sorbent used (g) (Wong et al., 2003).

3.1. Characterization of magnetic chitosan-Fe(III) hydrogel Scanning electron micrographs (SEMs) of chitosan-Fe(III) hydrogel with or without magnetism are shown in Fig. 1. It illustrates that the shape of the particles is irregular and relatively nonporous. The XPS spectrum of chitosan-Fe(III) hydrogel were provided in Figure S2 reveals carbon, oxygen, nitrogen, and iron are the predominant elements observed on the surface. Figure S2d shows the Fe 2p peaks at binding energies of 710.5 and 723.1 eV, with shakeup satellites at 715.9 and 731.5 eV, which could be attributed to Fe(III). Meanwhile, the N 1s and O1s spectrum of chitosan-Fe(III) complex are also recorded (Figure S2b and S2c). The broad peak of O1s can be fitted by two peaks at binding energies of 531.4 and 532.7 eV, respectively. The peak at low binding energy 531.1 eV is characteristics of eOH oxygen in chitosan which chelated with Fe

Fig. 1 e Scanning electron micrographs (SEMs) of chitosan-Fe(III) hydrogel and magnetic chitosan-Fe(III) hydrogel at low and high magnification: A1eA2, chitosan-Fe(III) hydrogel; B1-B2, magnetic chitosan-Fe(III) hydrogel.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 0 0 e5 2 1 0

(III), and the other peak at around 532.7 eV suggests the presence of eOH oxygen without chelation. For the spectrum of N 1s, the fitted peak with the lowest binding energy 398.9 eV is associated to the eNH2 nitrogen which interacted with Fe (III), and the other two peaks at 399.9 and 402.2 eV are the characteristics of nitrogen in eNH2 and eNHþ 3 group in chitosan. These results are consistent with the reported literature (Nieto et al., 1992; Ai et al., 2008; Klepka et al., 2008). The characteristic of magnetism provides an easy and efficient way to separate the chitosan-Fe(III) hydrogel from aqueous solutions. Thus, the Fe3O4 nano particles were added into the chitosan-Fe(III) complex, the magnetic chitosan-Fe(III) hydrogel was obtained. It can be seen that the magnetic chitosan-Fe(III) complex with adsorbed dye could be easily separated by a magnetic field, resulting in clean water (Fig. 2).

3.2.

Kinetics study of dye removal

Rapid interaction of the pollutants to be separated with the adsorbent is desirable and beneficial for practical adsorption applications. The kinetic behavior of the adsorption process was studied under alkaline condition (pH ¼ 12) at different initial dye concentrations using chitosan-Fe(III) and magnetic chitosan-Fe(III) (Fig. 3a and b). It can be observed from the figure that AR 73 uptake on chitosan-Fe(III) hydrogel was a very fast process. The amount of adsorption increased rapidly in the first 5 min, contributing to 90% of the ultimate adsorption amount, and then augmented slowly and approached the adsorption equilibrium in about 10 min. Besides, the total amount of AR 73 adsorbed increased with the increasing initial dye concentrations. This is due to that the higher initial adsorbent concentration provides higher driving force for the ions from the solution to the chitosanFe(III) hydrogel, resulting in more collisions between dye ions and active sites on the chitosan-Fe(III) complex. Although there is no significant improvement of dye removal efficiency using magnetic chitosan-Fe(III) hydrogel, compared with chitosan-Fe(III) hydrogel, the characteristic of magnetism

Fig. 2 e The separation of magnetic chitosan-Fe(III) hydrogel from dye solution.

5203

provides an easy and efficient way to separate the chitosanFe(III) hydrogel from aqueous solutions. To determine the rate-controlling and mass transfer mechanism, kinetic data were correlated to linear forms of the pseudo-first-order rate model and the second-order rate model (the equation and parameters of these two models are listed in Text S2 of Supporting Information). The values of the rate constants are determined from the intercepts of the curves and are given in Table 1. It can be observed from Table 1 that the change of the pseudo-first-order rate constant is irregular with an increase of initial dye concentration, but the second-order rate constant decreases with an increase of initial dye concentration. By correlation of the kinetic data with the above two rate models, the high degree of nonlinearity suggest the inability of the pseudo-first-order model to describe the kinetic profile of the adsorption process, whereas it was found that the plot of t/qt against time using different initial dye concentrations gives straight lines (Fig. 3c and d) with high correlation coefficients (Table 1). This indicates that the present sorption system follows predominantly the second-order rate model and the overall process appears to be controlled by chemisorptions (Chatterjee et al., 2005).

3.3.

pH sensitivity of magnetic chitosan-Fe(III) hydrogel

The pH of the dye solution plays an important role in the whole adsorption process and particularly on the effectiveness of treatment, influencing not only the surface charge of the adsorbent, the degree of ionization of the material present in the solution and the dissociation of functional groups on the active sites of the adsorbent, but also the solution dye chemistry (Crini, 2006). Thus, it is important to examine whether the adsorption was pH independent for a new developed adsorbent material. In order to investigate the pH sensitivity of magnetic chitosan-Fe(III) hydrogel, the effect of pH on dye adsorption is illustrated in Fig. 4. It appears that the chitosan-Fe(III) hydrogel, and magnetic chitosan-Fe(III) hydrogel possessed a higher adsorption efficiency compared to normal chitosan under the same pH condition. In the case of normal chitosan, 90% of AR 73 was adsorbed under pH ¼ 6, but the sharpest decline of removal efficiency occurred when pH increased from 6 to 8, and no apparent sorption was observed when pH was above 10. On the other hand, the adsorption efficiency of AR 73 on the chitosan-Fe(III) complex and magnetic chitosan-Fe(III) complex slightly decreased from 98.7% and 98.2% to 93.2% and 91.3% when the pH condition increased from 6 to 12. It is important to indicate that while the adsorption on chitosanFe(III) hydrogel was almost independent on the pH, the adsorption of dyes on normal chitosan was controlled by the acidity of the solution. During the dyes adsorption process, different kinds of interactions between chitosan and dyes such as chemical bonding, ion-exchange, hydrogen bonds, hydrophobic attractions, van der Waals force, physical adsorption, aggregation mechanisms, dyeedye interactions, etc., can act simultaneously (Crini and Badot, 2008). For normal chitosan, electrostatic attraction is now recognized as the main factor in anionic dyes adsorption. Decreasing the pH of the solution makes more protons available to protonate the amine group of

5204

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 0 0 e5 2 1 0

b

80 200mg/L 100mg/L 50mg/L 25mg/L

60

100

Adsorption Quantity (mg/g)

100

Adsorption Quantity (mg/g)

a

40

20

80 200mg/L 100mg/L 50mg/L 25mg/L

60

40

20

0

0 0

5

10

15

20

25

0

30

5

10

15

d

3.0 200mg/L 100mg/L 50mg/L 25mg/L

t/Qe (min•g/mg)

2.5 2.0

25

30

3.0 200mg/L 100mg/L 50mg/L 25mg/L

2.5

t/Qe (min•g/mg)

c

20

Time (min)

Time (min)

1.5 1.0 0.5

2.0 1.5 1.0 0.5 0.0

0.0 0

5

10

15

20

25

30

0

35

5

10

15

20

25

30

35

Time (min)

Time (min)

Fig. 3 e Removal kinetics of dyes (initial concentrations ranged from 25 mg/L to 200 mg/L): (a) time profile with chitosanFe(III) hydrogel (2 g/L); (b) time profile with magnetic chitosan-Fe(III) hydrogel (2 g/L); (c) fitting curve of pseudo-second-order for chitosan-Fe(III) hydrogel; (d) fitting curve of pseudo-second-order for magnetic chitosan-Fe(III) hydrogel.

chitosan, with the formation of a larger number of cationic amines, resulting in high pH sensitivity of chitosan. Nevertheless, under the alkaline condition, the free amino groups of chitosan could not be protonated, which did not facilitate electrostatic interaction between chitosan and the negatively charged anionic dyes. Therefore, in the case of chitosan-Fe(III) hydrogel, the chelating interaction between dye molecules and Fe(III) center may play a leading role instead of electrostatic interactions in AR 73 adsorption. This strong chelating interaction provided the chitosan-Fe(III)

hydrogel a larger adsorption capacity, and the high removal efficiency could remain in a wide range of pH from 6 to 12 in chitosan-Fe(III) hydrogel adsorption system.

3.4.

Adsorption isotherms under alkaline condition

The adsorption isotherm of AR 73 on chitosan-Fe(III) hydrogel was measured under various initial dye concentrations in the presence of 2.0 g/L of the adsorbent at pH ¼ 12. Langmuir, Freundlich, and LangmuireFreundlich isotherms were used to

Table 1 e Parameters of kinetics study for dyes adsorption onto chitosan-Fe(III) hydrogel. Adsorbent

Chitosan-Fe(III) complex

Magnetic chitosan-Fe(III) complex

Initial concentration (mg/L)

25 50 100 200 25 50 100 200

Pseudo-first-order equation 2

Pseudo-second-order equation

qe (mg/g)

K1 (1/min)

R

qe (mg/g)

K2 (g/mg$min)

R2

11.06 24.18 42.87 82.04 11.58 23.00 44.17 81.31

1.06 1.23 0.91 0.66 0.75 0.63 0.64 0.66

0.9993 0.9993 0.9977 0.9898 0.9986 0.9973 0.9893 0.9863

11.20 24.46 44.03 86.28 11.83 23.71 46.47 86.21

0.65 0.37 0.08 0.02 0.31 0.11 0.04 0.02

1.0000 1.0000 0.9998 0.9992 0.9998 0.9992 0.9996 0.9998

5205

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 0 0 e5 2 1 0

a

250

100

qe (mg/g)

Removal rate (%)

200

80

60

40

pH=6 pH=8 pH=10 pH=12

20

150

100

Langmuir model Freundlich model Langmuir-Freundlich model

50

0 0

0 0

5

10

15

20

25

Time (min)

Removal rate (%)

b

500

1000

1500

2000

2500

3000

3500

Ce (mg/L)

30

Fig. 5 e Adsorption isotherms fitted to Langmuir model (the dash line), Freundlich equation (the dash-dotted line) and LangmuireFreundlich model (the real line).

100

80

60

40

pH=6 pH=8 pH=10 pH=12

20

The LangmuireFreundlich isotherm fits best with the experimental data (correlation coefficient R2 ¼ 0.9929), whereas the low correlation coefficients (R2 ¼ 0.9615) show less agreement of Freundlich isotherm with the experimental data. The sorption capacity of the chitosan-Fe(III) hydrogel in alkaline condition was found to be 294.5 mg/g according to the fitted parameter of LangmuireFreundlich equation.

3.5.

Regeneration and reuse of chitosan-Fe(III) hydrogel

0 0

5

10

15

20

25

30

Time (min)

c

Removal rate (%)

100

80

60

40

pH=6 pH=8 pH=10 pH=12

20

0 0

5

10

15

20

25

30

Time (min) Fig. 4 e pH sensitivity of (a) normal chitosan, (b) chitosanFe(III) hydrogel, and (c) magnetic chitosan-Fe(III) hydrogel in dye removal. (Initial concentration 100 mg/L, 10 mL, T [ 25  C, adsorbent 0.02 g). fit the experimental data (the equation and parameters of these three isotherm models are listed in Text S3 of Supporting Information). The results of the experimental data fitted to these three equations and the parameters are shown in Fig. 5 and Table S1.

The regeneration of the adsorbent is important for lowering the cost of the adsorption process and for possibly recovering the pollutant extracted from wastewater. Thus, regeneration and reuse experiments were investigated in our study. The basic solutions as NaAc, NaOH, NaHCO3, Na2CO3, Na2HPO3, and NH4OH, the polar solvents as methanol, ethanol, and acetone, and the chelator EDTA with different concentrations were selected to find the optimal eluent for desorption. The use of methanol, ethanol and acetone solutions for the dye desorption is totally ineffective. The desorption efficiencies of 1 mol/L NaAc, NaHCO3, Na2CO3, Na2HPO3, NH4OH and EDTA were found to be 10.78%, 8.42%, 7.24%, 3.33%, 0.85%, and 6.02% respectively. Only the strong basic solution (1 mol/L NaOH) did help to desorb the dye efficiently (desorption efficiency 94.4%). However, according to the experiments in our study, in which 0.01, 0.05, 0.1, and 0.5 mol/L NaOH solutions with different concentration were applied, the chitosan-Fe(III) hydrogel could keep its high adsorption performance when the concentration of NaOH below 0.05 mol/L. Besides, the desorbed dyes were condensed during the desorption procedure, due to that only 25 mL 1 M NaOH could desorbed the dyes from chitosan-Fe(III) hydrogel efficiently, which the solution volume was reduced in half. Then, the dyes could come into reuse after adjusting the pH value. After the desorption process, the ultrapure water containing 0.1 mol/L HCl was applied to rinse the adsorbent, in order to regenerate the chitosan-Fe(III) complex. Then, the regenerated chitosan-Fe(III) hydrogel could be put into reuse. Meanwhile, during both the adsorption and desorption

5206

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 0 0 e5 2 1 0

Table 2 e Adsorption and Desorption Behaviors of C. I. acid red 73 on Magnetic chitosan-Fe(III) hydrogela. Cycle

Adsorption (%)

Desorption (%)

95.3 94.0 93.5 94.0 90.0

94.4 94.6 94.0 93.8 90.1

I II III IV V

a The initial concentration of dye is 50 mg/L, volume is 10 mL, pH ¼ 12, T ¼ 25  C, and the dosage of adsorbent is 0.02 g.

procedure, the magnetic chitosan-Fe(III) hydrogel could be separated by a magnetic field. The efficiency of chitosan-Fe(III) hydrogel undergoing five cycles are illustrated in Table 2. Reuse experiments showed that the adsorption efficiency of the chitosan-Fe(III) hydrogel with magnetism remained almost constant for five cycles of adsorption and desorption, which indicated that there were no irreversible sites on the surface of the adsorbent.

3.6. Effect of the coexisting ions and adsorption of other anionic dyes Due to the presence of the common ions coexisting with dye which may imply competition for available chelation sites, it is necessary to investigate the competitive influence of commonly coexisting anions and cations with dyes. The common cations in alkaline condition as Naþ, Ca2þ, and Kþ,    and the common anions as NO 3 , Cl , HPO4 , and SO4 were applied in the coexisting ions investigation. It is shown in Figure S3 in Supporting information, that the influences of Naþ, Ca2þ, and Kþ on the adsorption of dye were rather insignificant since they did not compete for the Fe(III) site in chitosan-Fe(III) complex. Similarly, the competitive influence   2 of NO 3 , Cl , HPO4 , and SO4 on dye adsorption can be ignored. Compared with dye molecules which contain sulfonate group,   2 suggest a weaker adsorption NO 3 , Cl , HPO4 , and SO4 mechanism via complexation. Thus, the common coexisting ions almost had no negative effect on the dye adsorption of chitosan-Fe(III) hydrogel under alkaline condition, which also

Table 3 e Adsorption efficiency of common anionic dyesa. Dyes

Removal rate (%)

Pictures of experiment

Chitosan-Fe(III)

Magnetic chitosan-Fe(III)

AB25 AB40 AB62 AB113

99.80 99.37 98.66 100.00

96.5 95.7 93.1 97.9

Control

AB193 AR73 RR24 RY2

97.85 96.28 99.27 97.96

96.3 97.5 98.5 95.2

Chitosan-Fe(III) hydrogel

RB74 RB194 RR11 RY18

98.48 99.04 98.74 93.62

97.2 98.5 97.8 94.1

Normal Chitosan

a The initial concentration of dye is 50 mg/L, volume is 10 mL, pH ¼ 12, T ¼ 25  C, and the dosage of adsorbent is 0.02 g.

5207

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 0 0 e5 2 1 0

Proposed mechanism

The experimental results in kinetics study clearly show that AR 73 uptake on chitosan-Fe(III) hydrogel follows predominantly the second-order rate model, suggesting the overall process was controlled by chemisorptions. Besides, it was a very fast process, which the adsorption could reach equilibrium in less than 10 min. According to the literature (Li et al., 2005), this type of adsorption behavior is typical of the specific adsorption process in which adsorption rate is usually dependent upon the number of available adsorption sites on the surface of the adsorbent and eventually controlled by the attachment of the AR 73 molecules on the surfaces. The high initial uptake rate and the short adsorption equilibrium time can be an indication that the surfaces of chitosan-Fe(III) hydrogel had a high density of active sites for dyes adsorption. Additionally, as can be observed in Fig. 4, chitosan-Fe(III) hydrogel could efficiently remove AR 73 under the condition of pH ¼ 12, while no apparent dyes sorption was observed in normal chitosan system. Under the alkaline condition, the free amino groups of chitosan could not be protonated since its point of zero potential lies within 6.5e6.7 (Guibal, 2005), which could not facilitate electrostatic interaction between chitosan and the negatively charged anionic dyes. Thus, in the chitosan-Fe(III) hydrogel system, the electrostatic interaction should not play the leading role in dyes removal. According to literatures (Bhatia and Ravi, 2000; Hernandez et al., 2008), chitosan-Fe(III) complex is a hexacoordinated Fe(III), and there is the entrance of 2 mol of monomeric sugar units of the ligand in the coordination sphere of Fe3þ. The complexation is mainly through the amino group (eNH2) and the hydroxyl group (eOH) of chitosan. Water molecules or other ions in solution would complete the coordination sphere of chitosan-Fe(III) complex with different stability (Figure S4). Hence, we speculated that the dye molecules, which contain sulfonate groups with negative charges, could change places with the water molecules in the coordinated sphere of chitosan-Fe(III) complex to chelate the Fe(III) center due to the higher stability, resulting in their stronger adsorption. The detailed information about near neighbor atoms around Fe and their distances to central Fe atom would be determined by

a

100 90 80

Weight (%)

3.7.

the analysis of extended X-ray absorption fine structure (EXAFS) data in our future study. To provide evidence of the dyes adsorption mechanism, thermogravimetric analysis was conducted for chitosanFe(III) hydrogel before and after dyes adsorption at pH 12. As shown in Fig. 6a, four weight loss stages can be distinguished from the TGA curves. For chitosan-Fe(III) complex, the mass loss was approximately 10% at 80  C, related to loss of water (Wong et al., 2003). The second loss, between 215 and 300  C, approximately 32%, related to the oxidative degradation and crosslinking reactions (Nieto et al., 1992). These processes are followed by glucose ring scissions (stage III) and carbonization of the material (stage IV). The weight loss is continuous in the last three stages. After the dyes loading to the chitosan-Fe(III), the thermal stability of chitosan-Fe(III) complex was enhanced. It shows a lower decomposition rate than that of chitosan-Fe(III) complex, indicating formation of a more stable complex. However, the dye-loaded chitosan-Fe(III) hydrogel did not show the weight loss of C. I. Acid Red 73 dyes between 25 and 800  C obviously, attributing to the small concentration of dyes in the sample (c.a. 25 mg/g). Especially, only 2.5% of mass loss appeared after 80  C, which obviously suggests that the bound water in chitosan-Fe(III) complex disappeared after the dyes adsorption. Additionally, as the differential thermogravimetry analysis (DTG) curves (Fig. 6b)

70 60 50 40 chitosan-Fe(III) AR 73 loaded chitosan-Fe(III)

30 20 0

100

200

300

400

500

600

700

800

Tempreture (°C)

b Derivatives Weight (%/m)

decided the possible of practical application for magnetic chitosan-Fe(III) hydrogel. It is well known that the complicated structures of dye molecules, which vary with respect to the organic chains and the numbers and positions of functional groups, are directly related to their adsorption behaviors. Thus, the same adsorption conditions used for C. I. Acid Red 73 were next applied to the removal of a variety of anionic dyes. The acid dyes, such as C. I. Acid Blue 25 (AB 25), C. I. Acid Blue 40 (AB 40), C. I. Acid Blue 62(AB 62), C. I. Acid Blue 113 (AB 113) and C. I. Acid Blue 193(AB 193), the reactive dyes, such as C. I. Reactive Yellow 2 (RY 2), C. I. Reactive Yellow 18 (RY18), 1 C. I. Reactive Red 11 (RR11), C. I. Reactive Red 24 (RR 24), C. I. Reactive Blue 194 (RB 194) and C. I. Reactive Blue 74 (RB 74), were chosen as prototypical dye pollutants. The results are shown in Table 3. The data suggest that the chitosan-Fe(III) hydrogel could efficiently adsorb both the acid and reactive dyes under alkaline condition.

1 0 -1 -2 -3 -4 -5

chitosan-Fe(III) AR 73 loaded chitosan-Fe(III)

-6 0

100

200

300

400

500

600

700

Tempreture (°C) Fig. 6 e TGA (a) and DTG (b) curves of chitosan-Fe(III) hydrogel with or without dyes loadling.

800

5208

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 0 0 e5 2 1 0

shown, the first peak at 80  C corresponding to the bound water further confirms that the water molecules chelated in dye-loaded chitosan complex was lost. And the second peak of chitosan-Fe(III) complex appears at 250  C for the oxidative degradation and crosslinking reactions, whereas the peak exhibits at higher temperature (280  C) for chitosan-Fe(III)dye. The lower weight loss of dye-loaded chitosan-Fe(III) hydrogels could be attributed to that a more thermal stable complex was formed through the chelating adsorption between dyes and chitosan-Fe(III) (Cheng et al., 2010). It demonstrates that the anionic dye strongly binds to the Fe(III) complex of chitosan and suggests that the chelating interactions play an important role in dyes removal under alkaline conditions. As the results shown in desorption studies, the use of polar solvents as methanol, ethanol and acetone solutions for the dye desorption is totally ineffective, but the strong basic solution (1.0 mol/L NaOH) did help to desorb the dyes efficiently (desorption efficiency 94.4%). When the concentration of free hydroxyl ions in solution was high enough, hydroxyl ions would break the chelating interaction between dyes and chitosan-Fe(III) and replaced the dyes to bind the Fe(III) center. Nevertheless, it should be noted that the hydroxyl ions below 0.05 mol/L would not produce negative impact on the chelation between dyes and chitosan complex according to the experiments indicated above. In the regeneration process, the ultrapure water containing 0.1 mol/L HCl was used. It could effectively regenerate the chitosan-Fe(III) hydrogel, the chitosan-Fe(III) hydrogel could remain the high removal performance for five cycles of adsorption and desorption. Overall, the main possible mechanism for the fast and efficient dyes adsorption under alkaline condition is proposed in Fig. 7. When the chitosan-Fe(III) hydrogel was added to the aqueous dye solution, the free dye molecules changed places

with the water molecules to chelate the Fe(III) center in hydrogel, resulting in their stronger adsorption (Fig. 7a). During the desorption processes (Fig. 7b), the hydroxyl ions of high concentration would break the chelating interaction between dyes and chitosan-Fe(III) and replaced the dyes to bind the Fe(III). Then, under the condition which contains large amount of Hþ, the water molecular replaced the hydroxyls to complete the coordination spheres and the chitosan-Fe(III) complex was recovered. Thus, the magnetic chitosan-Fe(III) hydrogel could quickly and efficiently remove the dyes under alkaline conditions via the chelation between sulfonic group of dye and Fe(III) center in chitosan-Fe(III) complex.

3.8.

Practical implication

This study revealed that the magnetic chitosan-Fe(III) hydrogel could efficiently remove the dyes under the alkaline conditions. In particular, the high speed of dyes adsorption process, which the adsorption could reach equilibrium in less than 10 min, is desirable and beneficial for practical adsorption applications. Meanwhile, the hydrogel with adsorbed dyes can be simply recovered from water with magnetic separations, which can hopefully simplify the water treatment process. In our experiments, 12 kinds of anionic dye containing sulfonic groups were chosen to examine the applied scope of chitosan-Fe(III) hydrogel, and all the dyes could be removed efficiently. On the other hand, this efficient adsorbent might be easily extended to the removal of other pollutants which contain anionic groups, according to the strong chelating interaction between Fe(III) center in chitosan complex and ions in solution. Additionally, the preparation of magnetic chitosan-Fe(III) hydrogel are very cheap and environmentally friendly as their main component, Fe3þ and

Fig. 7 e Proposed interactions mode between dyes and chitosan-Fe(III) hydrogel.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 0 0 e5 2 1 0

chitosan, are abundant and have no adverse effect on the environment. Besides, the effective regeneration of the adsorbent can further lower the cost of the adsorption process and possibly recover the pollutant extracted from wastewater. Therefore, this technically feasible, highly efficient and cost effective adsorbent is expected to have wide applicability in the removal of anionic dyes or even other anionic organic pollutants under the alkaline conditions.

4.

Conclusion

In this study, a novel low-cost magnetic sorbent material, chitosan-Fe(III) hydrogel, was prepared by chelation procedure with cheap and environmentally friendly chitosan and iron salts. It could quickly and efficiently remove the dyes under alkaline conditions (pH ¼ 12) via the chelation between sulfonic group of dye and Fe(III) center in chitosanFe(III) complex. The adsorption of dyes to chitosan-Fe(III) hydrogel was fast, which could reach equilibrium in less than 10 min, and agreed well to the LangmuireFreundlich adsorption model with a high maximum adsorption capacity of 294.5 mg/g. Additionally, the common coexisting ions almost had no negative effect on the dye adsorption of chitosan-Fe(III) hydrogel under alkaline condition. And this adsorbent could be regenerated and reused effectively for the adsorption of mostly common dyes. Further, the possible mechanism for dye removal was proposed, which involved the adsorption, desorption and regeneration processes. Thus, this technically feasible, highly efficient and cost effective adsorbent is very attractive and implies a potential of practical application for alkaline dyeing effluent treatment.

Acknowledgments This study was supported by the Qianjiang Talent Scheme, Zhejiang Province, China (No. 2011R10045), the National Basic Research Program of China (No. 2009CB421603) and the Key Laboratory of Industrial Ecology and Environmental Engineering, China Ministry of Education. We also thank for the help of SSRF (Shanghai Synchrotron Radiation Facility) in the characterization of the chitosan-Fe(III) complex.

Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.07.018.

references

Ai, Z., Cheng, Y., Zhang, L., Qiu, J., 2008. Efficient removal of Cr(VI) from aqueous solution with Fe@Fe2O3 coreshell nanowires. Environ. Sci. Technol. 42, 6955e6960.

5209

Avlonitis, S., Poulios, I., Sotiriou, D., Pappas, M., Moutesidis, K., 2008. Simulated cotton dye effluents treatment and reuse by nanofiltration. Desalination 221, 259e267. Bhatia, S.C., Ravi, N., 2000. A magnetic study of an Fechitosan complex and its relevance to other biomolecules. Biomacromolecules 1, 413e417. Chatterjee, S., Chatterjee, S., Chatterjee, B.P., Das, A.R., Guha, A.K., 2005. Adsorption of a model anionic dye, eosin Y, from aqueous solution by chitosan hydrobeads. J. Colloid Interface Sci. 288, 30e35. Cheng, R., Ou, S., Xiang, B., Li, Y., Liao, Q., 2010. Equilibrium and molecular mechanism of anionic dyes adsorption onto copper(II) complex of dithiocarbamate-modified starch. Langmuir 26, 752e758. Choe, E., Son, E., Lee, B., Jeong, S., Shin, H., Choi, J., 2005. NF process for the recovery of caustic soda and concentration of disodium terephthalate from alkaline wastewater from polyester fabrics. Desalination 186, 29e37. Crini, G., 2006. Non-conventional low-cost adsorbents for dye removal: a review. Bioresour. Technol. 97, 1061e1085. Crini, G., Badot, P.M., 2008. Application of chitosan, a natural aminopolysaccharide, for dye removal from aqueous solutions by adsorption processes using batch studies: a review of recent literature. Prog. Polym. Sci. 33, 399e447. Fan, L., Zhang, L., Shen, J., Wang, S., Chen, H., 2007. Study on recovery and refining of TA from alkali reduction wastewater. Desalination 206, 353e357. Guibal, E., 2005. Heterogeneous catalysis on chitosan-based materials: a review. Prog. Polym. Sci. 30, 71e109. Gupta, V.K., Jain, R., Varshney, S., 2007. Electrochemical removal of the hazardous dye Reactofix Red 3 BFN from industrial effluents. J. Colloid Interface Sci. 312, 292e296. Gupta, V.K., Suhas, 2009. Application of low-cost adsorbents for dye removal - a review. J. Environ. Manage. 90, 2313e2342. Hernandez, R.B., Franc, A.P., Yola, O.R., Lopez-Delgado, A., Felcman, J., Recio, M.A.L., Merce, A.L.R., 2008. Coordination study of chitosan and Fe3þ. J. Mol. Struct. 877, 89e99. Jang, M., Min, S.H., Park, J.K., Tlachac, E.J., 2007. Hydrous ferric oxide incorporated diatomite for remediation of arsenic contaminated groundwater. Environ. Sci. Technol. 41, 3322. Ko¨rbahti, B.K., Tanyolac¸, A., 2008. Electrochemical treatment of simulated textile wastewater with industrial components and Levafix Blue CA reactive dye: optimization through response surface methodology. J. Hazard. Mater. 151, 422e431. Klepka, M.T., Nedelko, N., Greneche, J.M., LawniczakJablonska, K., Demchenko, I.N., Slawska-Waniewska, A., Rodrigues, C.A., Debrassi, A., Bordini, C., 2008. Local atomic structure and magnetic ordering of iron in Fe-chitosan complexes. Biomacromolecules 9, 1586e1594. Li, N., Bai, R.B., Liu, C.K., 2005. Enhanced and selective adsorption of mercury ions on chitosan beads grafted with polyacrylamide via surface-initiated atom transfer radical polymerization. Langmuir 21, 11780. Nieto, J.M., Penichecovas, C., Delbosque, J., 1992. Preparation and characterization of a chitosan-Fe(III) complex. Carbohydr. Polym. 18, 221e224. Pelegrini, R., Peralta-Zamora, P., de Andrade, A.R., Reyes, J., Duran, N., 1999. Electrochemically assisted photocatalytic degradation of reactive dyes. Appl. Catal. B 22, 83e90. Selcuk, H., 2005. Decolorization and detoxification of textile wastewater by ozonation and coagulation processes. Dyes Pigm. 64, 217e222. Shen, C.S., Song, S.F., Zang, L.L., Kang, X.D., Wen, Y.Z., Liu, W.P., Fu, L.S., 2010. Efficient removal of dyes in water using chitosan microsphere supported cobalt (II) tetrasulfophthalocyanine with H2O2. J. Hazard. Mater. 177, 560e566. Shen, C.S., Wen, Y.Z., Kang, X.D., Liu, W., 2011. H2O2-induced surface modification: a facile, effective and environmentally

5210

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 0 0 e5 2 1 0

friendly pretreatment of chitosan for dyes removal. Chem. Eng. J. 166, 474e482. Vlyssides, A.G., Papaioannou, D., Loizidoy, M., Karlis, P.K., Zorpas, A.A., 2000. Testing an electrochemical method for treatment of textile dye wastewater. Waste Manage. 20, 569. Wang, X.L., Yang, K., Tao, S., Xing, B.S., 2007. Sorption of aromatic organic contaminants by biopolymers effects of pH, copper (II) complexation, and cellulose coating. Environ. Sci. Technol. 41, 185. Wong, Y.C., Szeto, Y.S., Cheung, W.H., McKay, G., 2003. Equilibrium studies for acid dye adsorption onto chitosan. Langmuir 19, 7888e7894. Wu, H.F., Wang, S.H., Kong, H.L., Liu, T.T., Xia, M.F., 2007. Performance of combined process of anoxic baffled reactor-

biological contact oxidation treating printing and dyeing wastewater. Bioresour. Technol. 98, 1501e1504. Xu, D., Hein, S., Loo, S.L., Wang, K., 2008. The fixed-bed study of dye removal on chitosan beads at high pH. Ind. Eng. Chem. Res. 47, 8796e8800. Yang, Q., Zhang, W., Zhang, H., Li, Y., Li, C., 2011. Wastewater treatment by alkali bacteria and dynamics of microbial communities in two bioreactors. Bioresour. Technol. 102, 3790e3798. Yoshitake, H., Yokoi, T., Tatsum, T., 2003. Adsorption behavior of arsenate at transition metal cations captured by amino-functionalized mesoporous silicas. Chem. Mater. 15, 1713.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 1 1 e5 2 1 7

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Development of a rapid DNA extraction method and one-step nested PCR for the detection of Naegleria fowleri from the environment Arine Fadzlun Ahmad a,b, James Lonnen a, Peter W. Andrew a, Simon Kilvington a,* a

Department of Infection, Immunity & Inflammation, University of Leicester, Medical Sciences Building, PO Box 138, University Road, Leicester LE1 9HN, UK b Department of Parasitology, Faculty of Medicine Building, University of Malaya, 50603 Kuala Lumpur, Malaysia

article info

abstract

Article history:

Naegleria fowleri is a small free-living amoebo-flagellate found in natural and manmade

Received 5 February 2011

thermal aquatic habitats worldwide. The organism is pathogenic to man causing fatal

Received in revised form

primary amoebic meningoencephalitis (PAM). Infection typically results from bathing in

18 July 2011

contaminated water and is usually fatal. It is, therefore, important to identify sites con-

Accepted 19 July 2011

taining N. fowleri in the interests of preventive public health microbiology. Culture of

Available online 27 July 2011

environmental material is the conventional method for the isolation of N. fowleri but requires several days incubation and subsequent biochemical or molecular tests to confirm

Keywords:

identification. Here, a nested one-step PCR test, in conjunction with a direct DNA extrac-

Naegleria fowleri

tion from water or sediment material, was developed for the rapid and reliable detection of

Nested PCR

N. fowleri from the environment. Here, the assay detected N, fowleri in 18/109 river water

PAM

samples associated with a nuclear power plant in South West France and 0/10 from

Ecology

a similar site in the UK. Although culture of samples yielded numerous thermophilic free-

Detection

living amoebae, none were N. fowleri or other thermophilic Naegleria spp. The availability of

DNA extraction

a rapid, reliable and sensitive one-step nested PCR method for the direct detection of N. fowleri from the environment may aid ecological studies and enable intervention to prevent PAM cases. Crown Copyright ª 2011 Published by Elsevier Ltd. All rights reserved.

1.

Introduction

Naegleria fowleri is a thermophilic, free-living amoebo-flagellate characterised by a life-cycle of trophozoite, flagellate and cyst stage (Ma et al., 1990; Carter, 1972; Marciano-Cabral, 1988; Schuster and Visvesvara, 2004; De Jonckheere, 2002; John, 1982). The organism is pathogenic to humans, causing primary amoebic meningoencephalitis (PAM) (Carter, 1972; MarcianoCabral, 1988; De Jonckheere, 2002; John, 1982; Martinez, 1993). Infection results from the instillation of the organism into the

anterior nares usually whilst bathing. From here the amoebae penetrate the nasal epithelium and olfactory nerves and migrate through the cribriform plate to invade the brain and meninges (Carter, 1978; Cain et al., 1981). PAM is almost invariably fatal with death occurring in 3e7 days following exposure (John, 1982; Cain et al., 1981). It is, therefore, important to identify sites containing N. fowleri in the interests of preventive public health microbiology. This requires methods that are both accurate and reliable for the differentiation of N. fowleri from other closely

* Corresponding author. Tel.: þ44 116 252 2950; fax: þ44 116 252 5030. E-mail address: [email protected] (S. Kilvington). 0043-1354/$ e see front matter Crown Copyright ª 2011 Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.025

5212

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 1 1 e5 2 1 7

related thermophilic Naegleria spp. Although not pathogenic, Naegleria lovaniensis resembles N. fowleri in tolerating growth up to 45  C, cytopathogenicity for tissue culture cells and antigenicity (Marciano-Cabral, 1988; Stevens et al., 1980). Although mouse pathogenicity was originally used to differentiate the species the recognition of Naegleria australiensis that was shown to be pathogenic for mice, albeit less so than N. fowleri, rendered the test nonspecific (John, 1982; De Jonckheere, 1981). The isolation of N. fowleri and N. lovaniensis involves the culture at 44  C of sediment and water concentrates on nonnutrient agar plates (NNA) seeded with a living suspension of Escherichia coli (NNA-E. coli) (Anon, 1990; Page, 1988; John and Howard, 1996). Presumptive Naegleria of are then picked from the plates and incubated at 37  C in deionised water or ¼ strength Ringer’s solution and observed for trophozoite transformation into the temporary flagellate stage (Anon, 1990; Page, 1988). Flagellate positive isolates must then be further characterised to differentiate the species. In natural thermal environments, N. lovaniensis tends to predominate and often necessitates the screening of large numbers of isolates when attempting to identify N. fowleri. Furthermore, it has been reported that the enflagellation test may not always be positive with Naegleria (Kilvington et al., 1991; De Jonckheere et al., 2001; Behets et al., 2003). Biochemical and molecular techniques have been developed for the differentiation of Naegleria spp. and associated identification of N. fowleri (De Jonckheere, 2002). These include monoclonal antibodies (Visvesvara et al., 1987; Flores et al., 1990; Reveiller et al., 2000), isoenzyme electrophoretic profiles (Kilvington, 1995; Pernin and Grelaud, 1989; De Jonckheere, 1982; Adams et al., 1989) and analysis of restriction fragment length polymorphisms from either whole-cell or PCR amplified DNA (De Jonckheere, 1987a; Kilvington and Beeching, 1995a; Pelandakis et al., 1998, van Belkum et al., 1992). More recently the rapid identification of N. fowleri by PCR methods have been developed (Kilvington and Beeching, 1995b; Pelandakis and Pernin, 2002; Qvarnstrom et al., 2006; Reveiller et al., 2002; Sparagano, 1993; Maclean et al., 2004). Whilst these methods enable the reliable and specific identification of N. fowleri and Naegleria spp., they typically require the primary culture and subculture of isolates from environmental samples. This is a time consuming process and can result in lack of detection sensitivity due to overgrowth from the more rapidly growing but non-pathogenic species such as N. lovaniensis and other FLA (De Jonckheere, 2002; Kilvington et al., 1991; Kilvington and Beeching, 1995a). Developments in the PCR, combined with improved DNA extraction methods, enable the direct detection of microorganisms in environmental samples without the need for primary culture isolation (Yeates et al., 1997; Fitzpatrick et al., 2010; Fierer et al., 2005; Tzeneva et al., 2009). This has been applied to the rapid detection of N. fowleri from the environment through application of conventional, nested and realtime PCR methods (Pelandakis and Pernin, 2002; Maclean et al., 2004; Puzon et al., 2009; Sheehan et al., 2003; Jamerson et al., 2009). Although PAM is a rare infection (111 cases confirmed in the USA since 1962 (Yoder et al., 2010) reports have suggested that the incidence is increasing (Yoder et al., 2010; Heggie, 2010). Accordingly, the rapid and reliable

identification of environmental habitats containing N. fowleri may enable the implementation of preventative public health measures to reduce the risk of PAM cases. Here we describe a DNA extraction and one-step nested PCR for the reliable detection of N. fowleri from thermally enriched environmental samples.

2.

Materials and methods

2.1.

N. fowleri nested PCR development

The nested PCR was developed from a cloned fragment of Naegleria folweri (MCM) DNA (pUC PB2.3), shown from hybridization and PCR analysis to be specific for the organism (Kilvington and Beeching, 1995a, b). The outer primers were: OP4F 50 -gcctttcttcggctcgcatg-30 and OP4R 50 -cttgagtgcacgccacttgat-30 while the internal primers were: IP4F 50 -caggaatgtcatcacac-30 and IP4R 50 -gaatgagtactcgttgc-30 . Testing of these primer sets alone and in combination against purified DNA from several strains of N. fowleri, Naegleria spp. and other organisms showed them to be specific only for the species. The Tm values for the outer and inner primer sets were calculated as 65  C and 54  C respectively, enabling the nested PCR to be conducted as a one-step procedure in a single tube to minimise carry over and cross-contamination. The PCR produce sizes were 767 bp and 506 bp for the outer and inner primer sets, respectively. PCR was performed in 40 ml volumes consisting 2X ReddyMix PCR Master Mix (ABgene, Surrey, UK), 0.1 mM of each outer primer, 0.5 mM of each internal primer and 4 ml of environmental DNA sample. The PCR conditions comprised: 94  C for 4 min, followed by 20 cycles of 94  C for 1 min, 60  C for 1 min, 72  C for 1 min, and then 35 cycles of 94  C for 1 min, 48  C for 1 min, 72  C for 1 min. A final elongation step of 72  C for 10 min was used. Amplicons were detected by electrophoresis, through 1.5% (w/v) agarose gels containing 0.5 mg/ml ethidium bromide in Tris-Acetate-EDTA buffer (Sambrook et al., 1989). N. fowleri positive PCR products were excised from the agarose gel, purified using NucleoSpin Extract II (MachereyeNagel, Germany) according to the suppliers instructions and sequenced by MWG-Biotech (Ebersberg, Germany) using the primer IP4R. DNA sequences were compared to the cloned N. fowleri DNA sequence using BLAST software (Altschul et al., 1990).

2.2.

Samples collection and DNA extraction

A total of 119 water samples (1 L) were collected from the River Tarn, South West France (109 samples) and the River Trent, United Kingdom (10 samples) between June 2008 and August 2009. Samples were collected into sterile polypropylene containers and processed within 5 days of collection. Depending on the turbidity of the water samples, processing was done either by filtration (clear samples) or direct centrifugation. Briefly, 750 ml of water was filtered slowly (approximately 5 min) through a 0.45 mm pore size cellulose nitrate membrane (Sartorius, Surrey, UK). Filtration was stopped when approximately 20 ml of water remained above the membrane. The residual water above the membrane and the membrane were added to a 50 ml sterile centrifuge tube so

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 1 1 e5 2 1 7

5213

that the sample surface faced inwards from the walls of the container. The tube was vortexed for 10 s and the surface debris removed from the membrane with sterile cotton-tipped swab and expressed into the tube. The membrane was then removed and the tube centrifuged at 1000 g for 10 min. The supernatant was discarded, leaving approximately 1 ml and the pellet resuspended by vortexed for 5 s. Turbid samples were centrifuged at 1000 g for 10 min and the pellets pooled and resuspended as described above. The resuspended deposits were inoculated over a series of NNA-E. coli as 5 small drops per plate. After allowing the drops to dry, the plates were sealed in polythene bags and incubated at 44  C. Plates were examined microscopically for up to 7 days for FLA trophozoites migrating away from the inocula. A part of each isolate growth was scraped from the plate with a disposable 1 ml bacteriological loop and inoculated into the well of a flat bottomed 96 well microtitre plate containing 100 ml of ¼ strength Ringer’s solution. The plate was sealed and incubated at 37  C and observed after 1e3 h for trophozoite morphological characteristics of the genus Naegleria and transformation into the flagellate stage (Page, 1988). Culture positive, enflagellating isolates are then identified as either N. fowleri or N. lovaniensis using a duplex species specific PCR (S. Kilvington, unpublished methodology). In addition, FLA that exhibited limax type trophozoite movement were also tested using this PCR to exclude the possibility that they were N. folweri or N. lovaniensis but failed to flagellate.

“Universal” 16S prokaryote primers 338F-16S (50 -ACTCCTACGGGNGGCNGCA-30 ), (Fierer et al., 2005) and 797Re16S (50 GGACTACCAGGGTATCTAATCCTGTT-30 ) (Nadkarni et al., 2002). These primers exhibit broad specificity against known prokaryote bacteria for this gene region (personal communication, Dr R. Free, University of Leicester, UK) and have been used previously to assess the quality of environmental DNA extracts for PCR amplification (Ahmad et al., 2011). The sensitivity of the N. fowleri nested PCR was determined using 100, 10 and 1 pg of purified genomic DNA in the assay. Samples that were negative by this PCR were reamplified using the same primer sets and conditions along with 1 ml from the initial reaction tube. The total volume was 20 ml consisting of 2X ReddyMix PCR Master Mix, 0.5 mM of each primer and 4 ml of environmental DNA sample. The PCR conditions comprised: 95  C for 5 min, followed by 30 cycles of 94  C for 30 s, 60  C for 30 s, 72  C for 40 s and a final elongation step at 72  C for 10 min. The expected PCR product is 500 bp. Based on the standard operating procedure for testing the samples stated that samples with negative results during the primary PCR should be reamplified (1 ml) using the same primers. However, if the results were still negative, the samples should be excluded from analysis using the N. fowleri nested PCR.

2.3.

Of the 119 samples extracted for DNA, 108 (91%) gave a positive reaction with the 16S bacterial PCR primers, resulting a product of expected 500 bp size (results not shown). Of the 11 negative samples, all were positive on reamplification of the original PCR tubes using the 16S primers and included in the study. All 119 water samples were culture negative for N. fowleri although other thermophilic FLA were isolated at 44  C incubation (Vahlkampfia sp., Hartmannella sp., Cashia sp., Vannella sp., Platyamoeba sp., Acanthamoeba sp. and an unidentified amoeba) from both the French (54/109) and UK samples (6/10). However, 18/119 (15%) of samples were positive for N. fowleri by the one-step nested PCR, giving the expected nested product of 506 bp (Fig. 1). All N. fowleri nested PCR positive samples were from samples collected from the River Tarn, South West France (18/109). For 4 of the positive reactions, sequencing of the DNA product showed a 99% homology to the DNA region used to design the PCR primers and further confirms that the nested PCR is specific in the amplification of N. fowleri DNA only. In the sensitivity studies, the nested PCR was able to detect 10 pg of purified genomic DNA of N. fowleri (results not shown). No reaction was obtained with 1 pg of DNA and, therefore, the sensitivity of the assay lies between 10 and >1 pg (equivalent to approximately 50 - >5 cells).

DNA extraction

The remainder of the pellet was mixed with 10 ml of UNSET lysis solution (urea 8 M, sarkosyl 2%, NaCl 0.15 M, EDTA 0.001 M, Tris pH 7.5 0.1 M) and vortexed for 1 min with 4 g of 0.25 mm glass beads (Jencons, Sussex, United Kingdom) (Hugo et al., 1992). The suspension was left at room temperature for 30 min, centrifuged at 3900 g for 10 min. The supernatant was then incubated at RT for 2 h or overnight at 4  C in a halfvolume of 30% (v/v) polyethylene glycol in 1.6 M NaCl (Yeates et al., 1997). Following centrifugation at 3900 g for 40 min, the resulting pellet was resuspended in 5 ml TE buffer (10 mM TriseHCl, 1 mM sodium EDTA, pH 8.0) and potassium acetate (7.5 M) added to 0.5 M and placed on ice for 5 min prior to centrifugation at 3900 g for 30 min at 4  C (Yeates et al., 1997). The supernatant was extracted twice with phenol:chloroform (1:1) and once with chloroform by centrifugation at 3900 g for 3 min at RT. The nucleic acids were precipitated with 0.8 volume of ice-cold isopropanol, pelleted by centrifugation at 3900 g for 30 min and washed twice with 70% (v/v) ethanol. The final pellet was dried and dissolved in 50e100 ml 10 mM Tris-0.1 mM EDTA (pH 8.0). The nucleic acid was further purified using a commercial kit (ZR soil microbe DNA kit, Zymo Research, CA, USA) following the manufacturer’s instructions, with the exception that the bead extraction stage was omitted. The presence of DNA was confirmed by electrophoresis through 1.5% (w/v) agarose gels containing 0.5 mg/ml ethidium bromide in Trisacetate-EDTA buffer (Sambrook et al., 1989). The purified DNA was stored at 20  C until required (1e2 weeks). To assess the suitability of the environmental DNA extracts for PCR, samples were tested using the modified

3.

4.

Results

Discussion

N. fowleri is found in thermal aquatic environments and can tolerate temperatures up to 46  C (John, 1982). Although N. fowleri is most likely to be isolated from sites where the

5214

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 1 1 e5 2 1 7

Fig. 1 e Agarose gels showing PCR products from water samples from the River Tarn, France, after extraction by the rapid DNA extraction method followed by one-step nested N. fowleri PCR. Lanes M: 100 bp size marker ladder. Lanes 1e2: positive N. fowleri nested PCR samples from French river water samples (arrow indicates expected 506 pb nested PCR product). Lane 3: positive control of N. fowleri DNA plasmid clone PB2.3 (upper arrow indicates outer primer set PCR product of 767 bp and lower arrow the 506 bp nested PCR product). Lane N: negative control (nanopure water).

temperature is above 30  C, the cysts can survive at 4  C for at least 12 months with retention of virulence by the excysted trophozoites (Warhurst et al., 1980). N. fowleri occurs worldwide and has been isolated from both natural and artificial thermally enriched habitats such as natural hot springs, fresh water lakes, domestic water supplies, chlorinated swimming pools, water cooling towers and effluent from industrial processes (Martinez and Visvesvara, 1997). Since PAM was first recognised in 1965, several hundred cases of PAM have been reported globally. Although PAM is a rare infection, 111 cases have been confirmed in the USA since 1962 (Yoder et al., 2010) and reports have suggested that the incidence is increasing (Yoder et al., 2010; Heggie, 2010). Clustering of cases can occur when a single site is the source of infection. In Usti, Czechoslovakia, 16 cases were associated with a public swimming pool (Cerva and Novak, 1968). Cases of PAM have been reported from Belgium and Czechoslovakia in persons swimming in warm effluent water from industrial processes (De Jonckheere, 1987b). One confirmed case of PAM occurred in Bath Spa, England in 1978 in a child who swam in a public bathing pool fed with water from the historic thermal springs that rise naturally in the City (Cain et al., 1981). Subsequent analysis confirmed the thermal springs to be the source of the infection (Kilvington et al., 1991). In South Western Australia, several cases were associated with the reticulated mains supply water (Dorsch et al., 1983). Cases linked to the domestic

water supply in Arizona, USA have also been reported (Marciano-Cabral et al., 2003). More recently, PAM has been associated with swimming in lakes or ponds in the USA and Italy (Yoder et al., 2010; Cogo et al., 2004). PAM is almost invariably fatal and emphasises the importance of identifying sources harbouring N. fowleri so that remedial action such as disinfection or prevention of bathing can be implement (Martinez, 1993; Vargas-Zepeda et al., 2005). Typically, detection relies on the culture of environmental material prior to subsequent genus identification and speciation of N. fowleri using biochemical or molecular techniques (Page, 1988; Behets et al., 2003; Kilvington, 1995; Kilvington and Beeching, 1995b; Lares-Villa and Hernandez-Pena, 2010). This can take several days or weeks to accomplish and the presence of N. fowleri on the primary isolation culture plates can be obscured by other, faster growing, FLA. PCR enables the rapid and specific identification of N. fowleri and has been applied successfully to the identification of the organism from the environment following culture isolation (Behets et al., 2003; Kilvington and Beeching, 1995b; Pelandakis and Pernin, 2002; Jamerson et al., 2009; Lares-Villa and Hernandez-Pena, 2010). In a nested PCR, the outer primers amplify DNA in the first round of reactions which act as templates for the internal primers in the second round of the PCR. This can significantly enhance the detection sensitivity of the assay and mitigate the effect of compounds inhibitory to the Taq polymerase present in the extracted DNA. Typically, a nested PCR is accomplished in two separate tube reactions, with the products of the first round of reaction being added to a separate second tube. This can increase the occurrence of false positive reactions from cross-contamination. The advantage of the nested PCR developed here is that the annealing temperatures (Tm) of the outer and inner primer sets are such that they react independently in the first and second round of the assay. This enables the reaction to be conducted in a single tube and greatly simplifies the procedure and reduces the risk of cross-contamination. Key to any PCR is the quality of the DNA used in the reaction. Environmental samples can contain inhibitory compounds, such as humic material, which can contaminate extracted DNA and inhibit the activity and efficiency of Taq polymerase (Yeates et al., 1997; Tsai and Olson, 1992). In this study, we report a reliable method for extracting DNA from water samples which is suitable for PCR amplification. Modification of a previously published method for environmental DNA extraction through the additional use of a commercial kit to further purify the DNA showed that 90% (108/119) were suitable for PCR using universal bacterial 16S primers or 100% when reamplification of the reaction was used. We have also used the method to successfully extract DNA from soil and mud samples for the detection of the pathogenic FLA Balamuthia mandrillaris (Visvesvara et al., 2007) by a nested PCR (Ahmad et al., 2011). Accordingly, the method may be suitable for the DNA extraction and PCR analysis for a wide range of both prokaryote and eukaryote organisms from the environment. In previous studies N. fowleri PCR assays have reported a detection sensitivity of 5 pg of purified DNA (which approximates to 25 amoeba) or 5 whole amoebae when spiked into environmental samples. For the nested PCR described in this study a positive reaction was obtained with 10 pg of

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 1 1 e5 2 1 7

purified DNA but not with the next dilution tested of 1 pg. Accordingly, the sensitivity of the assay lies between these values for purified DNA and is likely to be comparable to that reported previously. Although spiking of samples with whole amoebae was not done in this study, it is acknowledged that the sensitivity of the PCR is likely to be reduced due the possible presence of substances inhibitory to the reaction that may co-purify during the extraction procedure. The possibility that the positive results for the nested N. fowleri PCR are derived from the presence of dead organism in the environmental samples cannot be excluded by this technique. However, knowledge that an environment may harbour N. fowleri even in the nonviable state would still be of importance in ecological and preventative public health studies. Previous studies have shown the sensitivity of nested PCR in the detection of N. fowleri from the environment either with or without primary culture enrichment of the samples (Pelandakis and Pernin, 2002; Maclean et al., 2004; MarcianoCabral et al., 2003). The sensitivity of the environmental DNA extraction method in conjunction with the N. fowleri nested PCR developed here in the detection of the organism from two river water sites in France and the UK is clearly shown. None of the samples from either site grew N. fowleri on culture analysis but 18/119 (15%) were positive by the direct nested PCR, all from the French waters. Both sites have been shown previously to contain N. fowleri by culture isolation and the absence in the present study is unclear. However, regular samples in the previous year from the French site were also negative on culture for N. fowleri and the UK site had not been examined since 1997, suggesting that presence in the waters have declined or else fluctuate in detectable numbers (Kilvington and Beeching, 1995a, 1997; Pelandakis and Pernin, 2002). Another explanation may lie in the amount of material than can be analysed by both methods. Culture on NNA-E. coli plates permits only a relatively small amount of material to be processed (<1 g) compared with up to 10 g using the direct DNA extraction presented here. In addition, the presence of faster growing FLA can obscure the presence of N. fowleri and reliance on the flagellation test to screen for presumptive Naegleria spp. may be unreliable, resulting in false negative findings (De Jonckheere et al., 2001; Behets et al., 2003). A reliable and sensitive one-step nested PCR method for the direct detection of N. fowleri from environmental samples has been developed and evaluated. Future developments in the technology may include evaluation of faster DNA extraction from environmental samples and the use of real-time PCR methods that could also quantify the presence of N. fowleri in the processed samples and differentiate viable from dead organisms (Qvarnstrom et al., 2006; Robinson et al., 2006). Although it was shown here that the nested PCR was more sensitive than culture for the detection of the organism, culture isolation should always be attempted in conjunction, as the ability to perform molecular typing assays on isolates is valuable in studying the genetic diversity of N. fowleri and in epidemiological investigations to identify sources of infection (Kilvington and Beeching, 1995a, van Belkum et al., 1992; Pelandakis and Pernin, 2002). The ability to detect the causative agent of PAM in the environment may enable intervention to prevent cases of human infection and aid ecological studies.

5.

5215

Conclusion

A reliable and sensitive one-step nested PCR method for the direct detection of N. fowleri from environmental samples has been developed and evaluated. This approach represents a significant advance in the ability to rapidly and reliably detect this human pathogenic free-living amoeba in the environment. This will enable a greater understanding of the ecology of the organism that may, also, help prevent cases of human infection.

Acknowledgements This work was supported by a research funding to Arine Fadzlun Ahmad from the Ministry of Higher Education Malaysia and the University of Malaya, Malaysia. We are grateful to Dr Wayne Heaselgrave of the University of Leicester for assistance in collecting and processing of samples examined in this study.

references

Adams, M., Andrews, R.H., Robinson, B., Christy, P.B.P.R., Dobson, P.J., Blackler, S.J., 1989. A genetic approach to species criteria in the genus Naegleria using alloenzyme electrophoresis. International Journal for Parasitology 19, 823e834. Ahmad, A.F., Andrew, P.W., Kilvington, S., 2011. Development of a nested PCR for environmental detection of the pathogenic free-living amoeba Balamuthia mandrillaris. Journal of Eukaryotic Microbiology 58 (3), 269e271. Altschul, S.F., Gish, W., Miller, W., Myers, E.W., Lipman, D.J., 1990. Basic local alignment search tool. Journal of Molecular Biology 215 (3), 403e410. Anon, 1990. Isolation and Identification of Giardia Cysts, Cryptosporidium Oocysts and Free Living Pathogenic Amoebae in Water, etc. 1989-Methods for the Examination of Waters & Associated Materials. HMSO, London. Behets, J., Seghi, F., Declerck, P., Verelst, L., Duvivier, L., Van Damme, A., Ollevier, F., 2003. Detection of Naegleria spp. and Naegleria fowleri: a comparison of flagellation tests, ELISA and PCR. Water Science and Technology 47 (3), 117e122. Cain, A.R., Wiley, P.F., Brownell, B., Warhurst, D.C., 1981. Primary amoebic meningoencephalitis. Archives of Disease in Childhood 56 (2), 140e143. Carter, R.F., 1972. Primary amoebic meningo-encephalitis. Transactions of the Royal Society of Tropical Medicine and Hygiene 66, 193e213. Carter, S.A., 1978. Primary amoebic meningoencephalitis (A “new” disease associated with water pollution). International Journal of Environmental Studies 12, 199e205. Cerva, L., Novak, K., 1968. Amoebic meningoencephalitis: sixteen fatalities. Science 160, 92. Cogo, P.E., Scagli, M., Gatti, S., Rossetti, F., Alaggio, R., Laverda, A. M., Zhou, L., Xiao, L., Visvesvara, G.S., 2004. Fatal Naegleria fowleri meningoencephalitis, Italy. Emerging Infectious Diseases 10 (10), 1835e1837. De Jonckheere, J.F., Brown, S., Dobson, P.J., Robinson, B.S., Pernin, P., 2001. The amoeba-to-flagellate transformation test is not reliable for the diagnosis of the genus Naegleria. Description of three new Naegleria spp. Protist 152 (2), 115e121.

5216

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 1 1 e5 2 1 7

Dorsch, M.M., Cameron, A.S., Robinson, B.S., 1983. The epidemiology and control of primary amoebic meningoencephalitis with particular reference to South Australia. Transactions of the Royal Society of Tropical Medicine and Hygiene 77 (3), 372e377. Fierer, N., Jackson, J.A., Vilgalys, R., Jackson, R.B., 2005. Assessment of soil microbial community structure by use of taxon-specific quantitative PCR assays. Applied and Environmental Microbiology 71 (7), 4117e4120. Fitzpatrick, K.A., Kersh, G.J., Massung, R.F., 2010. Practical method for extraction of PCR-quality DNA from environmental soil samples. Applied and Environmental Microbiology 76 (13), 4571e4573. Flores, B.M., Garcia, C.A., Stamm, W.E., Torian, B.E., 1990. Differentiation of Naegleria fowleri from Acanthamoeba species by using monoclonal antibodies and flow cytometry. Journal of Clinical Microbiology 28, 1999e2005. Heggie, T.W., 2010. Swimming with death: Naegleria fowleri infections in recreational waters. Travel Medicine and Infectious Disease 8 (4), 201e206. Hugo, E.R., Stewart, V.J., Gast, R.J., Byers, T.J., 1992. Protocols in Protozoology. Society of Protozoology, Kansas. Jamerson, M., Remmers, K., Cabral, G., Marciano-Cabral, F., 2009. Survey for the presence of Naegleria fowleri amebae in lake water used to cool reactors at a nuclear power generating plant. Parasitology Research 104 (5), 969e978. John, D.T., 1982. Primary amoebic meningoencephalitis and the biology of Naegleria fowleri. Annual Review of Microbiology 36, 101e103. John, D.T., Howard, M.J., 1996. Techniques for isolating thermotolerant and pathogenic freeliving amebae. Folia Parasitologica 43, 267e271. De Jonckheere, J.F., 1981. Naegleria australiensis sp. nov., another pathogenic Naegleria from water. Protistologica XVII, 423e429. De Jonckheere, J.F., 1982. Isoenzyme pattrerns of pathogenic and non-pathogenic Naegleria spp. using agarose isoelectric focusing. Annales de l’Institut Pasteur Microbiology 133, 319e342. De Jonckheere, J.F., 1987a. Characterisation of Naegleria species by restriction endonuclease digestion of whole-cell DNA. Molecular and Biochemical Parasitology 24, 55e66. De Jonckheere, J.F., 1987b. Epidemiology. In: Rondanelli, E.G. (Ed.), Amphizoic Amoebae, Human Pathology. Piccin Nuova Libraria, pp. 127e147. De Jonckheere, J.F., 2002. A Century of research on the amoeboflagellate genus Naegleria. Acta Protozoologica 41, 309e342. Kilvington, S., 1995. Identification of Naegleria fowleri and other Naegleria spp. (free-living amoebae) using cellulose acetate membrane electrophoresis of glucose phosphate isomerase. FEMS Microbiology Letters 133 (3), 219e223. Kilvington, S., Beeching, J., 1995a. Identification and epidemiological typing of Naegleria fowleri with DNA probes. Applied and Environmental Microbiology 61 (6), 2071e2078. Kilvington, S., Beeching, J., 1995b. Development of a PCR for identification of Naegleria fowleri from the environment. Applied and Environmental Microbiology 61 (10), 3764e3767. Kilvington, S., Beeching, J., 1997. Detection of a novel restriction fragment length polymorphism type in Naegleria fowleri (freeliving amoeba) isolates from electricity power stations in England and France. European Journal of Protistology 33, 186e191. Kilvington, S., Mann, P.G., Warhurst, D.C., 1991. Pathogenic Naegleria amoebae in the waters of Bath: a fatality and its consequence. In: Kellaway, G.A. (Ed.), Hot Springs of Bath, pp. 89e96. Lares-Villa, F., Hernandez-Pena, C., 2010. Concentration of Naegleria fowleri in natural waters used for recreational

purposes in Sonora, Mexico (November 2007eOctober 2008). Experimental Parasitology 126 (1), 33e36. Ma, P., Visvesvara, G.S., Martinez, A.J., Theodore, F.H., Daggett, P. M., Sawyer, T.K., 1990. Naegleria and Acanthamoeba infections: review. Reviews of Infectious Diseases 12 (3), 490e513. Maclean, R.C., Richardson, D.J., LePardo, R., Marciano-Cabral, F., 2004. The identification of Naegleria fowleri from water and soil samples by nested PCR. Parasitology Research 93 (3), 211e217. Marciano-Cabral, F., 1988. Biology of Naegleria spp. Microbiological Reviews 52 (1), 114e133. Marciano-Cabral, F., MacLean, R., Mensah, A., LaPat-Polasko, L., 2003. Identification of Naegleria fowleri in domestic water sources by nested PCR. Applied and Environmental Microbiology 69 (10), 5864e5869. Martinez, A.J., 1993. Free-living amebas: infection of the central nervous system. Mount Sinai Journal of Medicine 60 (4), 271e278. Martinez, A.J., Visvesvara, G.S., 1997. Free-living, amphizoic and opportunistic amebas. Brain Pathology 7 (1), 583e598. Nadkarni, M.A., Martin, F.E., Jacques, N.A., Hunter, N., 2002. Determination of bacterial load by real-time PCR using a broad-range (universal) probe and primers set. Microbiology 148 (Pt 1), 257e266. Page, F.C. (Ed.), 1988. A New Key to Freshwater and soil Gymnamoebae. Freshwater Biological Association, Ambleside, Cumbria. Pelandakis, M., Pernin, P., 2002. Use of multiplex PCR and PCR restriction enzyme analysis for detection and exploration of the variability in the free-living amoeba Naegleria in the environment. Applied and Environmental Microbiology 68 (4), 2061e2065. Pelandakis, M., De Jonckheere, J.F., Pernin, P., 1998. Genetic variation in the free-living amoeba Naegleria fowleri. Applied and Environmental Microbiology 64 (8), 2977e2981. Pernin, P., Grelaud, G., 1989. Application of isoenzymatic typing to the identification of nonaxenic strains of Naegleria (Protozoa, Rhizopoda). Parasitology Research 75 (8), 595e598. Puzon, G.J., Lancaster, J.A., Wylie, J.T., Plumb, I.J., 2009. Rapid detection of Naegleria fowleri in water distribution pipeline biofilms and drinking water samples. Environmental Science and Technology 43 (17), 6691e6696. Qvarnstrom, Y., Visvesvara, G.S., Sriram, R., da Silva, A.J., 2006. Multiplex real-time PCR assay for simultaneous detection of Acanthamoeba spp., Balamuthia mandrillaris, and Naegleria fowleri. Journal of Clinical Microbiology 44 (10), 3589e3595. Reveiller, F.L., Marciano-Cabral, F., Pernin, P., Cabanes, P.A., Legastelois, S., 2000. Species specificity of a monoclonal antibody produced to Naegleria fowleri and partial characterization of its antigenic determinant. Parasitology Research 86 (8), 634e641. Reveiller, F.L., Cabanes, P.A., Marciano-Cabral, F., 2002. Development of a nested PCR assay to detect the pathogenic free-living amoeba Naegleria fowleri. Parasitology Research 88 (5), 443e450. Robinson, B.S., Monis, P.T., Dobson, P.J., 2006. Rapid, sensitive, and discriminating identification of Naegleria spp. by real-time PCR and melting-curve analysis. Applied and Environmental Microbiology 72 (9), 5857e5863. Sambrook, J., Fritsch, E.F., Maniatis, T., 1989. Molecular Cloning. A Laboratory Manual. Cold Spring Harbor, New York. Schuster, F.L., Visvesvara, G.S., 2004. Free-living amoebae as opportunistic and non-opportunistic pathogens of humans and animals. International Journal for Parasitology 34 (9), 1001e1027. Sheehan, K.B., Fagg, J.A., Ferris, M.J., Henson, J.M., 2003. PCR detection and analysis of the free-living amoeba Naegleria in hot springs in yellowstone and grand teton national Parks. Applied and Environmental Microbiology 69 (10), 5914e5918.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 1 1 e5 2 1 7

Sparagano, O., 1993. Differentiation of Naegleria fowleri and other Naegleriae by polymerase chain reaction and hybridization methods. FEMS Microbiology Letters 110 (3), 325e330. Stevens, A.R., De Jonckheere, J.F., Willaert, E., 1980. Naegleria lovaniensis new species: isolation and identification of six thermophilic strains of a new species found in association with Naegleria fowleri. International Journal of Parasitology 10, 51e64. Tsai, Y.L., Olson, B.H., 1992. Rapid method for separation of bacterial DNA from humic substances in sediments for polymerase chain reaction. Applied and Environmental Microbiology 58, 2292e2295. Tzeneva, V.A., Salles, J.F., Naumova, N., de Vos, W.M., Kuikman, P.J., Dolfing, J., Smidt, H., 2009. Effect of soil sample preservation, compared to the effect of other environmental variables, on bacterial and eukaryotic diversity. Research in Microbiology 160 (2), 89e98. van Belkum, A., De Jonckheere, J., Quint, W.G.V., 1992. Genotyping Naegleria spp. and Naegleria fowleri isolates by interrepeat polymerase chain reaction. Journal of Clinical Microbiology 30, 2595e2598. Vargas-Zepeda, J., Gomez-Alcala, A.V., Vasquez-Morales, J.A., Licea-Amaya, L., De Jonckheere, J.F., Lares-Villa, F., 2005.

5217

Successful treatment of Naegleria fowleri meningoencephalitis by using intravenous amphotericin B, fluconazole and rifampicin. Archives of Medical Research 36 (1), 83e86. Visvesvara, G.S., Peralta, M.J., Brandt, F.H., Wilson, M., Aloisio, C., Franko, E., 1987. Production of monoclonal antibodies to Naegleria fowleri, agent of primary amebic meningoencephalitis. Journal of Clinical Microbiology 25, 1629e1634. Visvesvara, G.S., Moura, H., Schuster, F.L., 2007. Pathogenic and opportunistic free-living amoebae: Acanthamoeba spp., Balamuthia mandrillaris, Naegleria fowleri, and Sappinia diploidea. FEMS Immunology and Medical Microbiology 50 (1), 1e26. Warhurst, D.C., Carman, J.A., Mann, P.G., 1980. Survival of Naegleria fowleri at 4  C for eight months with retention of virulence. Transactions of the Royal Society of Tropical Medicine and Hygiene 74, 832. Yeates, C., Gillings, M.R., Davison, A.D., Altavilla, N., Veal, D.A., 1997. PCR amplification of crude microbial DNA extracted from soil. Letters in Applied Microbiology 25 (4), 303e307. Yoder, J.S., Eddy, B.A., Visvesvara, G.S., Capewell, L., Beach, M.J., 2010. The epidemiology of primary amoebic meningoencephalitis in the USA, 1962e2008. Epidemiology and Infection 138 (7), 968e975.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 1 8 e5 2 2 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Occurrence and removal of pharmaceuticals and personal care products (PPCPs) in an advanced wastewater reclamation plant Xin Yang a,*, Riley C. Flowers b, Howard S. Weinberg b, Philip C. Singer b a

School of Environmental Science and Engineering, Sun Yat-sen University, Guangzhou, China Department of Environmental Sciences and Engineering, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599-7431, United States b

article info

abstract

Article history:

The occurrence of nineteen pharmaceutically active compounds and personal care prod-

Received 7 May 2011

ucts was followed monthly for 12 months after various stages of treatment in an advanced

Received in revised form

wastewater reclamation plant in Gwinnett County, GA, U.S.A. Twenty-four hour composite

13 July 2011

samples were collected after primary clarification, activated sludge biological treatment,

Accepted 21 July 2011

membrane filtration, granular media filtration, granular activated carbon (GAC) adsorption,

Available online 29 July 2011

and ozonation in the wastewater reclamation plant. Compounds were identified and quantified using high performance liquid chromatography/tandem mass spectrometry (LC-

Keywords:

MS/MS) and gas chromatography/mass spectrometry (GCeMS) after solid-phase extraction.

Pharmaceuticals and personal care

Standard addition methods were employed to compensate for matrix effects. Sixteen of the

products (PPCPs)

targeted compounds were detected in the primary effluent; sulfadimethoxine, doxycycline,

Wastewater

and iopromide were not found. Caffeine and acetaminophen were found at the highest

Ozonation

concentrations (w105 ng/L), followed by ibuprofen (w104 ng/L), sulfamethoxazole and DEET

Activated sludge biological

(w103 ng/L). Most of the other compounds were found at concentrations on the order of

treatment

hundreds of ng/L. After activated sludge treatment and membrane filtration, the concen-

GAC adsorption

trations of caffeine, acetaminophen, ibuprofen, DEET, tetracycline, and 17a-ethynylestradiol (EE2) had decreased by more than 90%. Erythromycin and carbamazepine, which were resistant to biological treatment, were eliminated by 74 and 88%, on average, by GAC. Primidone, DEET, and caffeine were not amenable to adsorption by GAC. Ozonation oxidized most of the remaining compounds by >60%, except for primidone and DEET. Of the initial 16 compounds identified in the primary effluent, only sulfamethoxazole, primidone, caffeine and DEET were frequently detected in the final effluent, but at concentrations on the order of 10e100 ng/L. Removal of the different agents by the various treatment processes was related to the physicalechemical properties of the compounds. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Pharmaceuticals and personal care products (PPCPs) are likely to be found in any body of water influenced by raw or treated

wastewater, including rivers, streams, lakes and impoundments, and ground waters, many of which are used as drinking water sources. Some PPCPs may cause ecological harm, such as endocrine disruption and development of

* Corresponding author. Tel.: þ86 2039332690. E-mail address: [email protected] (X. Yang). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.026

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 1 8 e5 2 2 8

antimicrobial resistance. The health effects of small amounts of these agents on humans over a lifetime of exposure are unknown at this time. Municipal wastewater is a significant source of PPCPs in the environment because many of them are not removed completely in conventional wastewater treatment plants. The literature shows a wide range of many classes of PPCPs (e.g., antibiotics, betablockers, antiepileptics, liquid regulators) in wastewater effluents and receiving waters (Kasprzyk-Hordern et al., 2008; Kolpin et al., 2002; Lindberg et al., 2005; Ternes et al., 2003). Conventional wastewater treatment processes, such as activated sludge treatment and subsequent clarification, are not effective at completely eliminating all PPCPs from wastewater (Westerhoff et al., 2005). Activated carbon adsorption, ozonation or advanced oxidation, and membrane separation are promising advanced treatment processes that are capable of removing many of the PPCPs commonly found in wastewater (Ikehata et al., 2008; Snyder et al., 2007; Westerhoff et al., 2005). For example, addition of 5 mg/L of powder activated carbon (PAC) with a 4-h contact time removed 50%e>98% of the volatile PPCPs analyzed by GC/MS/MS and 10%e>95% of the polar PPCPs analyzed by LC/MS/MS (Westerhoff et al., 2005). Ozone is extremely reactive with some pharmaceuticals, such as carbamazepine, diclofenac, estradiol and estrogen (Ikehata et al., 2008; Westerhoff et al., 2005). These pharmaceuticals have functional groups and structures such as phenolic groups, amines, thioether sulfurs, and activated aromatic rings that are readily attacked by ozone. These advanced processes have mostly been evaluated using laboratory batch tests. Few studies have examined PPCP removal in full-scale treatment plants containing the advanced wastewater treatment processes listed above (Dickenson et al., 2009; Hollender et al., 2009; Reungoat et al., 2010). Furthermore, most of the full-scale studies that have been done have utilized intermittent grab samples. No study has been conducted to evaluate PPCP removal using composite samples over an extended period of time, e.g., one year. The F. Wayne Hill Water Resources Center, a wastewater reclamation plant in Gwinnett County, GA, employs both conventional biological treatment and advanced treatment processes, including membrane filtration, granular media filtration, granular activated carbon (GAC) adsorption, and ozonation, to treat its wastewater. The treated wastewater is currently discharged into the Chattahoochee River which ultimately serves as a drinking water supply for several downstream communities, including the City of Atlanta. Prior to this study, no information was available on the occurrence levels or effect of treatment on PPCPs at this plant. The primary aim of this work was to investigate the removal of 19 PPCPs by various wastewater treatment processes, including activated sludge treatment, membrane filtration, granular media filtration, GAC adsorption, and ozonation, in an advanced wastewater treatment plant and to relate the extent of removal to physicalechemical properties of the targeted PPCP compounds. The originality of this work derives from the long time span (over one year) of monthly sampling, utilization of composite sample collection instead of collection of grab samples, the investigation of different treatment processes in a full-scale advanced wastewater treatment plant, and a comparison of observations with wellknown physicalechemical properties of the analytes.

2.

Materials and methods

2.1.

Selected PPCPs

5219

A diverse group of PPCPs widely reported to occur in aquatic systems was chosen for study (see Table 1). Compounds were chosen to represent different groups of PPCPs, such as antibiotics, antiepileptics, analgesic drugs, X-ray contrast agents, and personal care products such as insect repellants. One or two compounds representing different classes of antibiotics were selected, such as quinolones, tetracyclines, sulfonamides, diaminopyrimidines, and lincosamides. In each class, compounds that are frequently reported in wastewater were considered. Differences in structures and physicalechemical properties were also considered. As noted in Table 1, sulfamethoxazole, erythromycin, trimethoprim, lincomycin, caffeine, DEET, acetaminophen, and triclosan were among the 30 most frequently detected organic wastewater contaminants as reported by the US Geological Survey (Kolpin et al., 2002). Six of the compounds selected e ciprofloxacin, erythromycin, sulfamethoxazole, carbamazepine, ibuprofen, and diclofenac e were among the top 10 high priority pharmaceuticals identified in a European assessment of PPCPs (Voogt et al., 2008). Physical, chemical, and biological properties of the targeted PPCPs are listed in Table 1 and the structures of the PPCPs are given in Table S1.

2.2.

Description of treatment plant

The F. Wayne Hill Water Resources Center treats 227 thousands of cubic meters per day of wastewater from Gwinnett County, GA. The facility consists of primary clarification for removal of settleable solids; activated sludge treatment (sludge age of 12 days and mixed liquor suspended solids concentration of 3200 mg/L) to achieve removal of biochemical oxygendemanding organic compounds (BOD), nitrogen, and phosphorus; and secondary clarification. Flow then splits, with about 5% going to a second clarification tank after ferric chloride addition, followed by granular media filtration. The remaining 95% of the flow goes to submerged membrane microfiltration units. The flow is then combined and passes on to GAC adsorption beds (Calgon F-400) with an empty bed contact time of 15 min. The beds were in place for about three years at the time this study began. It was assumed that they were essentially exhausted with respect to adsorption capacity other than capacity produced by bio-regeneration. The final effluent passes through ozone contact chambers; ozone doses range from 0.75 to 2 mg/L, with an average of 1 mg/L. A schematic diagram of the facility is shown in Fig. 1. General wastewater quality parameters, e.g., COD (chemical oxygen demand), TSS (total suspended solids), ammonia-nitrogen, etc., are listed in Table S2 in the Supporting Information.

2.3.

Sample collection

Composite samples at the wastewater reclamation treatment plant were collected by plant personnel once each month over a 24-h period using an Isco (Teledyne Isco, Lincoln NE)

5220

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 1 8 e5 2 2 8

Table 1 e Targeted PPCPs in this study and their properties. Class Sulfonamide antibiotic Macrolide antibiotic Quinolone antibiotic Tetracycline antibiotic Other antibiotic Antiepileptic X-ray contrast agent Anti-inflammatory

Antiseptic Hormone Others

Log K[2] ow

Compound

pKa[1]

Sulfadimethoxine Sulfamethoxazolea,b Erythromycina,b Lincomycina Ciprofloxacinb Levofloxacin Doxycycline Tetracycline Trimethoprima Carbamazepineb Primidone Iopromide Ibuprofenb Acetaminophena Diclofenacb Triclosana 17a-Ethynylestradiol (EE2) Caffeinea DEETa

N/A 1.69, 5.57 8.88 7.60 6.43, 8.49 6.05, 8.22 N/A 3.30, 7.68, 9.69 7.12 13.90 N/A N/A 4.91 9.38 4.15 7.9 10.4 10.40 0.67

kOH (109 M1s1)

Log Kd

Kbiol

kO3 (M1s1)

N/A 0.89 3.06 0.56 0.28 0.39 N/A 1.3 0.91 2.45 0.91 2.05 3.97 0.46 4.51 4.76 3.67

N/A 2.4[3] N/A N/A 4.3[7] N/A N/A N/A 2.3[3] 0.1[9] N/A 1.0[9] 0.9[9] 2.6[12] N/A 1.2[9] N/A 2.5[9]

N/A <0.1[4] 0.5e1[4] N/A N/A N/A N/A N/A N/A <0.01[10] N/A 1e2.5[10] 21e35[10] N/A <0.1[10] N/A 7e9[4]

N/A 2.5  106 [5] N/A 3.3  105 [6] 1.9  104 [8] N/A N/A 1.9  106 [8] 2.7  105 [8] 3.0  105 [5] 1.04[11] <0.8[5] 9.1  1[5] 1.41  103 [13] 1.0  106 [5] 3.8  107 [14] 3.0  106 [5]

N/A 5.5  0.7[5] N/A N/A 4.1[8] N/A N/A 7.7[8] 6.9[8] 8.8  1.2[5] 6.7[11] 3.3  0.6[5] 7.4  1.2[5] 2.2 7.5  1.5[5] N/A 9.8  1.2[5]

0.07 2.18

N/A N/A

N/A N/A

0.82[15] N/A

6.9[16] 4.95[17]

pKa, negative log of acidity constant(s); Kow, octanol-water partition coefficient; Kd, sorption constant on activated sludge; Kbio, pseudo firstorder degradation rate constant (1 g SS1 day1); KO3, second-order rate constant with O3; KOH, second-order rate constant with OH radicals. References : [1] Howard and Meylan, 1997 [2] http://logkow.cisti.nrc.ca/logkow/search.html. [3] Gobel et al., 2005 [4] Suarez et al., 2008 [5] Huber et al., 2003 [6] Qiang et al., 2004 [7] Golet et al., 2003 [8] Dodd et al., 2006 [9] Ternes et al., 2004 [10] Joss et al., 2006 [11] Benitez et al., 2008 [12] Carballa et al., 2007 [13] Andreozzi et al., 2005 [14] Suarez et al., 2007 [15] Rosal et al., 2009 [16] Kesavan and Powers, 1985 [17] Song et al., 2009. a Among the 30 most frequently detected organic wastewater chemicals reported by US Geological Survey (Kolpin et al., 2002) b Among the top 10 high priority pharmaceuticals identified in a European assessment of PPCPs (Voogt et al., 2008).

automated refrigerated sampler on a flow-weighted basis from each of the sampling locations shown in Fig. 1. Samples of primary clarifier, membrane filtration, GAC adsorption, and final effluent were collected monthly from January to December 2008. Secondary clarifier effluent following activated sludge treatment, and granular media filtration effluent were collected from September to December 2008. Raw wastewater influent was not collected and analyzed due to complexity introduced by the high suspended solids content of these samples and difficulties associated with analysis of such samples. After collection, samples were transferred to 2.5 L amber glass bottles, packed with “blue ice” in an insulated cooler chest, and transported to the University of North Carolina at Chapel Hill (UNC) for overnight delivery. The bottles had

been previously treated with 5% dimethyldichlorosilane in toluene to minimize adsorption of analytes to the walls of the bottles. The treated bottles were rinsed with toluene, methanol and laboratory-grade water (LGW), and dried prior to use (Ye et al., 2007). Upon receipt at UNC, samples were filtered immediately through 0.45 mm nylon filters (Millipore, Billerica, MA) and the pH was adjusted to 2.5 with concentrated sulfuric acid. Samples were then stored at 4  C in the dark. The extracts were analyzed within one week of receipt.

2.4.

Standards were obtained from the following suppliers: iopromide (U.S. Pharmacopeia, Rockville, MD), primidone (MP

Membrane filtration

Chemical Addition

Grit chamber

Standards and reagents

95%

screen

Primary clarifier

Aeration tanks Secondary clarifier

5%

Chemical Addition

GAC

Ozone contactor

Chemical clarifier Composite sampling location – Monthly sampling for 12 months

Granular media filter

Composite sampling location – Monthly sampling for 4 months

Fig. 1 e Schematic diagram of the wastewater reclamation plant, showing sampling locations.

5221

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 1 8 e5 2 2 8

Biomedicals, LLC, Solon, OH), ciprofloxacin (ICN Biochemicals, Irvine, CA), levofloxacin, caffeine, and triclosan (Fluka, Buchs, Switzerland). Surrogate standards 13C6-sulfamethoxazole and 13 C2-erythromycin were obtained from Cambridge Isotope Laboratories (Andover, MA) and d3-caffeine was obtained from C/D/N Isotopes (Quebec, Canada). Internal standards simatone and fenoprop were purchased from Accustandard (New Haven, CT) and SigmaeAldrich (St. Louis, MO), respectively. All other standards were obtained from SigmaeAldrich (St. Louis, MO). LGW was prepared in a water purification system (Pure Water Solutions, Hillsborough, NC) which prefilters fines (1 mm), removes chlorine, reduces total organic carbon to less than 0.2 ppm with activated carbon, and removes ions to 18 MU with mixed-bed ion-exchange resins. Stock solutions of the standards were prepared by dissolving each compound in methanol or LGW.

2.5.

Analytical methods

Solid-phase extraction (SPE) methods were employed to concentrate the analytes from the aqueous samples. Triclosan and EE2 were extracted, derivatized and analyzed by gas chromatography/mass spectrometry (GC/MS) based on a previously published method by Stanford and Weinberg (Stanford and Weinberg, 2007). Other compounds were analyzed by liquid chromatography/tandem mass spectrometry (LC/MS/MS) using the method of Ye et al. (Ye et al., 2007) with the following modifications; 13C6-sulfamethoxazole, 13 C2-erythromycin and d3-caffeine were used as surrogate standards and simatone and fenoprop were used as internal standards. The detailed SPE methods for all PPCPs are described in the Supporting Information (Text S1). The method of standard additions was used to compensate for matrix effects. Each sample was split into seven aliquots, five of which were spiked with standard solutions of the target

PPCPs; the remaining two were not spiked. The spike levels were selected so that the calculated concentrations of the compounds fell into the ranges of the spiked additions. Spike levels of 10, 20, 50, 100, and 200 ng/L were employed for the final effluent from the wastewater plant. The spike levels for the other samples varied. The LC/MS/MS system consisted of a ProStar 210 solvent delivery module, a ProStar 430 autosampler, and a Varian 1200L triple quadrupole mass spectrometer (Varian Inc., Walnut Creek, CA). Chromatographic separation was achieved using a Pursuit C-18 column (15 cm  2 mm, 3 mm) and a C-18 guard column (3 cm  2 mm, 3 mm) supplied by Varian Inc. The optimal collision voltage for each of the precursor to product ion transitions of each compound is listed in Table S3 in the Supporting Information. The derivatized triclosan and EE2 were analyzed on an HP 5890 bench-top GC/MS (HewlettePackard) containing an HP 5972 mass selective detector (MSD) equipped with electron ionization (70 eV). The column was a DB-5MS fused silica capillary column (30 m  0.25 mm I.D. with 0.25 mm film thickness, J&W Scientific). Detailed information about the chromatographic and mass spectrometric methods, detection limits, quality assurance and control protocols are described in the Supporting Information (Text S2). Minimum reporting limits (MRLs) and recovery data are listed in Tables S4 and S5 in the Supporting Information, respectively.

3.

Results and discussion

3.1. PPCP concentration profiles across the wastewater reclamation plant Table 2 summarizes the monthly average, minimum and maximum concentrations of the targeted PPCPs in the primary, membrane filter, GAC and final effluent for the 12-

Table 2 e Summary of monthly average, minimum, and maximum concentrations of targeted PPCPs measured in 24-h composite samples of primary, membrane filter, GAC and final effluent (January to December, 2008). Compounds

Sulfamethoxazole Erythromycin Trimethoprim Lincomycin Ciprofloxacin Levofloxacin Tetracycline Carbamazepine DEET Primidone Diclofenac Triclosan EE2 Caffeine Acetaminophen Ibuprofen

Primary effluent

Membrane effluent

Average ng/L

Min ng/L

Max ng/L

Average ng/L

2600 340 610 21 620 460 160 230 1500 100 220 470 140 80000 80000 11000

1200 (140) 390 11 430 250 68 130 220 (60) 140 170 (24) 54000 37000 (3900)

3400 480 770 36 1100 900 310 440 4000 180 280 820 242 120000 130000 15000

420 270 280 14 130 140 <50 250 29 120 99 15 <20 65 <50 64

() Values are below MRLs, but meet quality control criteria. a Detected in a single sample.

GAC effluent

Final effluent

Min ng/L

Max ng/L

Average ng/L

Min ng/L

Max ng/L

Average ng/L

Min ng/L

Max ng/L

130 140 (69) 10 70 63

1600 410 530 20 240 250

210 13 11

1200 50 32

140 15

19

30

550 100 190 130 19

25 13 33

140 48 96

14

57

80 2 <10 <10 1 1 <10 1 18 46 <10 <10 <10 17 <50 <10

35 <10

100 13 90 (27) 13

670 28 21 <10 23 <10 <10 67 24 72 <10 <10 <10 36 <50 <10

38

98

(49)

78

<10 <10

16a 10a

<10 <10 25

12a 30 120

<10

50

5222

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 1 8 e5 2 2 8

month period from January to December, 2008. Figure S1(a)e(d) in the Supporting Information show PPCP concentrations in the primary, membrane, GAC and final effluent composite samples collected each month, respectively; the error bars shown represent the difference in concentration of the two duplicate unspiked samples. Sixteen of the nineteen target compounds were detected in the primary effluent. Doxycycline, sulfadimethoxine, and iopromide were below their MRL values in each sample. The compounds with the highest concentrations were caffeine, acetaminophen and ibuprofen which were found at levels of tens of mg/L. (See Table 2 and right side of Figure S1(a) and note change in scale.) Acetaminophen and ibuprofen are pain killers and are among the most widely used PhACs in the US. Caffeine is a stimulant and is also commonly used in beverages and foods. Similar levels of these compounds have been reported in the raw wastewater of other municipal wastewater treatment plants (Benotti and Brownawell, 2007; Buerge et al., 2003; Vieno et al., 2005). Sulfamethoxazole concentrations varied from about 1200 to 3400 ng/L, with an average concentration of 2570 ng/L. Previous reported levels of sulfamethoxazole in raw wastewater range from 520 to 9000 ng/L (Gobel et al., 2005; Hartig et al., 1999; Lindberg et al., 2005). The high concentrations of sulfamethoxazole could be due to its high rate of consumption for medical treatment purposes. Trimethoprim is always prescribed with sulfamethoxazole at a dose weight ratio of 0.2, and a trimethoprim/ sulfamethoxazole ratio of 0.23 by weight was found for most of the samples in this study. Concentrations of DEET, a widely used insect repellant, ranged from about 220 to 4000 ng/L, with an average level of 1500 ng/L. The measured concentrations of DEET in the primary effluent show a noticeable seasonal pattern, with concentrations being lowest in the winter months and peaking in the summer months, consistent with expected usage patterns for this insect repellant. (Values are not shown for July and August because the observed concentrations were above the highest spike levels (2000 ng/L) used to construct the calibration curve.) The other eleven PPCPs were found at concentrations below 1000 ng/L (1 mg/L). After activated sludge treatment and membrane filtration, most compounds were measured at concentrations below 1000 ng/L (see Figure S1(b) and Table 2). Tetracycline, EE2 and acetaminophen were below their respective MRLs of 50, 20 and 50 ng/L. Caffeine, ibuprofen and acetaminophen, which were found at the highest concentrations in the primary effluent, were all reduced to below 100 ng/L. A comparison of PPCP concentrations in the effluent from the secondary clarifiers, membrane filters, and granular media filters (GMF) for the 4 months in which composite samples were collected at all three of these locations (see Fig. 1) is shown in Figure S2 in the Supporting Information. In general, similar concentrations were found for all of the PPCPs in the clarifier and membrane filter effluents, implying that the PPCP removal shown in Figure S1(b) and Table 2 was due to biological treatment and subsequent solideliquid separation of the biological floc, not due to removal by membrane treatment. Microfiltration membranes are essentially a particle removal technology and, accordingly, do not reject soluble PPCPs. Since the analytical methods used in this study only

capture PPCPs in the aqueous phase, this is hardly surprising. Snyder et al. also found that microfiltration and ultrafiltration did not reject PPCPs to any appreciable degree (Snyder et al., 2007). Concentrations of PPCPs in GMF effluent were similar to those in secondary clarifier and membrane effluent, except for ciprofloxacin and levofloxacin, which were lower in the GMF effluent. For the GAC effluent, the compound with the highest levels was sulfamethoxazole, with concentrations ranging from 210 to 1200 ng/L. All other PPCPs were found at concentrations below 100 ng/L (except for one case of carbamazepine), as shown in Figure S1(c) and Table 2. Table 2 and Figure S1(d) show the PPCP concentrations in the final effluent, after ozonation. Sulfamethoxazole and primidone were detected in all of the monthly samples. Caffeine and DEET were detected frequently. The concentrations of sulfamethoxazole, primidone, DEET, and caffeine were in the range of 35e140, 25e120, <10e30, and <10e50 ng/L, respectively, and the average concentrations were respectively 78, 46, 21, and 20 ng/L. Erythromycin, ciprofloxacin, ibuprofen, and carbamazepine were found only occasionally in the final effluent. All of the other targeted PPCP compounds were below their MRLs.

3.2. Removal of targeted PPCPs by various unit processes Fig. 2 shows the concentrations of selected PPCPs across the treatment plant for illustrative purposes: (a) DEET; (b) sulfamethoxazole; and (c) primidone. Figure S3 in the Supporting Information shows concentrations of the other PPCPs across the treatment plant. Activated sludge treatment and microfiltration reduced the concentrations of DEET by several orders of magnitude (Fig. 2(a)). DEET appears to be relatively resistant to removal by GAC adsorption and ozonation. Significant removal of sulfamethoxazole (Fig. 2(b)) was achieved by biological treatment, but GAC treatment for sulfamethoxazole removal appeared to be limited. In fact, for a number of months, the concentration of sulfamethoxazole in the GAC effluent appeared to be higher than in the GAC influent. This is the only PPCP for which this pattern was observed. The cause of this apparent anomaly is not clear, but may be related to the anionic nature of sulfamethoxazole or to the presence and transformation of sulfamethoxazole metabolites. Table 1 indicates that, because of the low acidity constants for sulfamethoxazole, it is most likely present in anionic form. Anionic species tend to be poorly adsorbed by GAC, and it is possible that any sorbed sulfamethoxazole might have been displaced by other adsorbates. This pattern of adsorption and displacement may vary over time, depending on variations in influent water quality to the GAC bed and the extent of biodegradation reactions within the bed. Additionally, various human metabolites of sulfamethoxazole, such as N4-acetylsulfamethoxazole, are known to be present in municipal wastewater. N4-acetylsulfamethoxazole was measured using the same method as for sulfamethoxazole and analyzed by LC/MS/MS for two months during the latter phases of the study; the results are shown in Figure S4. Concentrations of N4-acetylsulfamethoxazole were greatly

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 1 8 e5 2 2 8

5000

a

DEET

Concentration (ng/L)

4000

90th percentile

2000

75th percentile

3.3.1.

Median

The percentage removal of each PPCP by activated sludge treatment and microfiltration is shown in Fig. 3. The removal is presented using box and whiskers plots which give the 10th, 25th, 50th (median), 75th, and 90th percentile values of the monthly composite analyses, as well as any outliers. Removal here refers to a reduction in concentration due to biodegradation and sorption (see below) since only measurements in the liquid phase were made. Also, it is recognized that the compound may be transformed to various products that may persist in the treated water so that the compound may not be completely removed; no analyses were performed for possible degradation products. Because, as noted above, membrane microfiltration is essentially a particle removal process, the values shown in Fig. 3 are attributed to removal by the microorganisms in the activated sludge process. PPCP removal in the activated sludge process is attributed to two mechanisms: sorption onto biological floc and biodegradation. Kd, the sorption constant which describes the partitioning of a compound between solid and aqueous phases, Kow, the widely used octanol-water partition coefficient which describes partitioning between lipophilic and hydrophilic phases, and Kbiol, the pseudo first-order biodegradation rate constant obtained from batch biodegradation tests, are listed in Table 1 for the various PPCPs examined, along with reference citations from which the values were obtained. The Kd values vary for sludges with different sludge ages and properties, and biodegradation rates for each of the compounds may differ due to different operating conditions at the treatment plant. The different PPCPs investigated in this study exhibited very different removal percentages. Acetaminophen, caffeine and ibuprofen, which were present at the highest concentrations in the primary effluent, were removed by more than 99%. Consistently high removal efficiencies for these compounds have also been reported by other researchers (Buerge et al., 2003; Miao et al., 2005; Vieno et al., 2005; Yu et al., 2006), attributable primarily to microbial degradation (Joss et al., 2006). No detectable tetracycline was found after activated sludge treatment. Though tetracycline has a high aqueous solubility and a low Kow, it has been shown to sorb readily onto various solids, such as soil particles (Lindsey et al., 2001; Sithole and Guy, 1987), and exhibited a high adsorption affinity for activated sludge biosolids (Kim et al., 2005). Significant removal of fluoroquinolone antibiotics was observed, an average of 80% for ciprofloxacin and 71% for levofloxacin. Sorption onto sludge floc has been found to be significant for ciprofloxacin, consistent with its high log Kd value of 4.3 reported by Golet et al. (2003). Although a Kd value for levofloxacin was not reported, the high Kd values for other fluoroquinolone antibiotics, such as norfloxacin, trovafloxacin, and gemifloxacin, suggest that levofloxacin is also strongly adsorbed. Due to the higher polarity of levofloxacin, its removal by adsorption was expected to be less than that of ciprofloxacin.

25th percentile 10th percentile 1000

b

Sulfamethxoazole

Concentration (ng/L)

3000

2000

1000

c Primidone 200

Concentration (ng/L)

PPCPs were apparent, other than the observation for DEET noted above.

3.3. Integrated assessment of removal of PPCPs by various processes

outlier 3000

5223

150

100

50

0

Primary effluent

Membrane GAC effluent effluent

Final effluent

Fig. 2 e Concentrations of (a) DEET; (b) sulfamethoxazole; and (c) primidone in primary, membrane filtration, GAC, and final effluent.

reduced during activated sludge treatment, and GAC treatment provided additional removal. It is possible that some of the N4-acetylsulfamethoxazole was transformed to sulfamethoxazole during GAC treatment. Transformations of other sulfamethoxazole metabolites confound the interpretation of sulfamethoxazole behavior during treatment. The results for primidone (Fig. 2(c)) show that it is essentially resistant to activated sludge treatment, GAC adsorption, and ozonation. Primidone was frequently detected in the final effluent of the treatment facility. An attempt was made to examine seasonal variations in the occurrence and removal of the PPCPs and also to see if there were any relationships with wastewater quality or operational parameters at the treatment plant. However, no seasonal variations in the occurrence and removal of the

Activated sludge treatment and microfiltration

5224

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 1 8 e5 2 2 8

100%

Removal efficiency

80% outlier

60%

90th percentile 75th percentile

40%

Median 25th percentile

20%

10th percentile

0% -20%

4

7

4

9

4

9

8

9

4

11

8

6

9

Tetracycline

EE2

DEET

triclosan

Sulfamethoxazole

Ciprofloxacin

Levofloxacin

diclofenac

Trimethoprim

Lincomycin

Erythromycin

Primidone

10 Carbamazepine

9

ibuprofen

Acetaminophen

N 9

Caffeine

N: number of samples

Fig. 3 e Removal efficiency of targeted PPCPs by activated sludge biological treatment and membrane filtration. (The number above the X-axis refers to the number of samples that were analyzed and met the quality assurance criteria.)

Removal of trimethoprim varied from 21% to 91%, with a median value of 58%, and removal of erythromycin varied from 41% to 72% with a median value of 4%. Values from the literature also show a large variation, from 40% to 70% removal for trimethoprim (Batt et al., 2006; Gobel et al., 2007; Lindberg et al., 2005; Kasprzyk-Hordern et al., 2009) and 14 to 49% removal for erythromycin (Gobel et al., 2007; KasprzykHordern et al., 2009). Higher removals have been reported for trimethoprim in nitrifying sludge, with longer sludge ages, than in conventional activated sludge (Batt et al., 2006). For sulfamethoxazole, the mean removal in this study was 92%, which is inconsistent with the findings by Joss et al. (2006) who reported a Kbiol value of less than 0.1 for sulfamethoxazole, suggesting it is minimally biodegradable. Different operating conditions in full-scale, continuous-flow treatment plants compared to the batch biodegradation tests conducted by Joss et al. (2006), such as differing dissolved oxygen levels, sludge ages, and wastewater characteristics, may be responsible for the observed differences in the two studies. N4-acetylsulfamethoxazole, a metabolite of sulfamethoxazole, was also reduced in concentration by 80e90%, presumably via biodegradation. Diclofenac was removed by 51e80% by activated sludge treatment. Removal percentages reported in the literature vary significantly. Some studies showed resistance of diclofenac to activated sludge treatment (Kasprzyk-Hordern et al., 2009), while other researchers reported a range of 9e80% removal (Kimura et al., 2007; Lindqvist et al., 2005; Yu et al., 2006). The sorption coefficient of 16 L/kg SS (log Kd ¼ 1.2) is

too low to expect significant attachment to sludge floc (Ternes et al., 2004). Poor biodegradation was also found in the batch tests by Joss et al. (2006). Zwiener and Frimmel (2003) reported that the anoxic-oxic ratios in microbial reactors may influence the efficiency of diclofenac removal, with better diclofenac degradation under anoxic conditions. The activated sludge process in the Gwinnett plant employs anoxic cells along with aeration cells to facilitate nitrogen and phosphorus removal. This may explain the high removal observed for diclofenac. Triclosan was eliminated by 96% due to biological treatment. Triclosan is relatively hydrophobic (log Kow ¼ 4.76), and the high sorption constant (log Kd ¼ 4.3) suggests that triclosan will be strongly sorbed onto sludge floc (Singer et al., 2002). Triclosan was also reported to undergo complete mineralization to CO2 as a result of biodegradation, with some incorporation into biomass (Federle et al., 2002). Therefore, it appears that both biodegradation and sorption contribute to triclosan removal. Good removal of DEET and EE2 were found in this study as a result of activated sludge treatment. DEET and EE2 have both been reported to be readily biodegradable (Joss et al., 2006; Rivera-Cancel et al., 2007). No significant removal of carbamazepine was found by activated sludge treatment, which is consistent with literature reports (Clara et al., 2004; Joss et al., 2005; Miao et al., 2005). The low Kd and Kbio values suggest that carbamazepine does not adsorb onto sludge biosolids and is not readily biodegradable. Although corresponding properties for primidone were not found in the literature, the results indicate that primidone is recalcitrant to activated sludge treatment.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 1 8 e5 2 2 8

In summary, sorption, biodegradation, or both could be the dominant pathway for the removal of PPCPs during activated sludge treatment, dependent on the physical, chemical, and biological properties of the specific PPCP.

applied to predict the adsorbability of an acidic or basic PPCP using the following equation, which takes into account the speciation of the compound at the pH of the wastewater: 0

KOW ¼

3.3.2.

GAC adsorption

Fig. 4 summarizes the PPCP removal results by GAC treatment. GAC effectively removed many of the PPCPs, with concentrations of some compounds decreasing by over 60% after GAC treatment. DEET, caffeine and primidone were relatively resistant to GAC treatment. As noted earlier, the GAC beds were in place for about three years at the time this study began. It was assumed that they were essentially exhausted with respect to adsorption capacity other than capacity produced by bio-regeneration. Biodegradation is known to occur in GAC beds (American Water Works Association, 1981), and the fact that the measured dissolved oxygen concentration in the effluent from the GAC beds was close to zero supports the presence of biological activity within the beds. Hence the GAC beds provide opportunities both for adsorption of the target compounds and biodegradation. Lincomycin, levofloxacin, trimethoprim, and ciprofloxacin, all of which exhibited high degrees of removal by GAC, were also readily removed by activated sludge treatment (see Fig. 3 and Table 1). Interestingly, some compounds, e.g. carbamazepine and erythromycin, which were not eliminated to any appreciable degree during activated sludge treatment, were eliminated quite well by the GAC, suggesting that there was still some sorption capability available in the beds, perhaps as a result of bio-regeneration. Adsorption of specific PPCPs onto GAC is dependent on the physical-chemical properties of the individual PPCPs. GAC tends to adsorb hydrophobic organic compounds; hydrophobicity is often characterized by the log of the octanol-water partition coefficient, Kow. A modified log Kow value (Kow0 ) was

Removal efficiency

100% 80%

-0.46 -1.39 1.66 1.88 0.66 -0.39

5225

KOW
1 þ 10^ ðpH  pKaÞ

Fig. 4 shows the percent removal of the targeted PPCPs that were detectable in the GAC influent and their corresponding log Kow0 values. Carbamazepine and erythromycin, with log Kow0 s of 3.05 and 2.45 respectively, exhibited high removal efficiencies while primidone and DEET, with respective log Kow0 s of 6.09 and 4.15, exhibited low removal efficiencies. Overall, however, no consistent pattern between log Kow0 and percent removal is apparent. Hence, it would appear that both adsorption and biodegradation contribute to the observed removal of PPCPs in the GAC beds. To assist in understanding the GAC removal results, supplementary batch adsorption tests were conducted using membrane effluent from the wastewater reclamation plant. 25 mg/L of pulverized F-400 GAC (the same carbon used in the full-scale plant) was added to the membrane effluent and mixed for 24 h, after which the pulverized carbon was separated from the suspension by filtration through 0.45 mm filters. The results are shown in Figure S5 in the Supporting Information by overlaying the removals during a single batch test with those obtained from the 12-month plant survey. The results confirm that primidone and DEET are the least readily removed by adsorption on activated carbon.

3.3.3.

Ozonation

Fig. 5 summarizes the removal efficiency of the PPCPs remaining in the GAC effluent by ozonation in the form of a series of box and whiskers plots. Also shown are the molecular structures for the PPCPs and their second-order rate constants with molecular ozone. As illustrated, ozonation effectively removed most of the remaining PPCPs in the GAC

-0.07

Log Kow '

60%

3.05

40%

2.45

20% 0%

9

10

3

7

8

9

11

10

9

Levofloxacin

diclofenac

ibuprofen

Trimethoprim

Ciprofloxacin

Erythromycin

Carbamazepine

Caffeine

Primidone

Lincomycin

N 8

-4.15 10 DEET

-6.09

-20%

Fig. 4 e Removal efficiency of targeted PPCPs by GAC. The number above the X-axis (N) refers to the number of samples that were analyzed and met the quality assurance criteria for reporting data. The values associated with each box or bar are the log of the modified octanol-water partition coefficients (log Kow0 ).

Fig. 5 e Removal of PPCPs by ozonation with their corresponding structures and second-order rate constants with molecular ozone.

5226

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 1 8 e5 2 2 8

effluent, although it is expected that most were transformed to other compounds. Sulfamethoxazole, primidone, caffeine and DEET were the only compounds routinely detected after ozonation; ciprofloxacin was detected on only two occasions. The removal of trimethoprim and carbamazepine was over 80%, and removal of sulfamethoxazole ranged from 67 to 94%. Rapid conversion of carbamazepine, trimethoprim, and sulfamethoxazole during ozonation has been reported by several groups of researchers (Nakada et al., 2007; Ternes et al., 2002). Nakada et al. (2007) demonstrated that 96% trimethoprim and 87% sulfamethoxazole were converted at an applied ozone dose of 3 mg/L and 27 min of contact in a secondary wastewater effluent after sand filtration (DOC ¼ 3.3 mg/L) at pH 7. An applied ozone dose of 0.5 mg/L was sufficient to convert 1 mg/ L carbamazepine spiked in surface water after flocculation (DOC ¼ 1.3 mg/L) at pH 7.8 (Ternes et al., 2002). This is consistent with our findings for an ozone dose of 1 mg/L and a DOC concentration of 4.5 mg/L. The removal of caffeine, primidone, and DEET exhibited a high degree of variation, ranging from no removal to 100% removal. This may be attributable to variations in the ozone dose, which ranged from 0.75 to 2.0 mg/L with an average value of 1 mg/L, or to variations in the influent water quality to the ozonation chamber. Trimethoprim, sulfamethoxazole, and carbamazepine all have amino groups, which are susceptible to chemical attack by ozone. The rate constants for the reaction of molecular O3 with these PhACs are 2.7  105, 2.5  106, and 3  105 M1s1, respectively (Dodd et al., 2006; Huber et al., 2003), as shown in Table 1 and Fig. 5. The rate constants of primidone and caffeine toward molecular O3 are low (1.0 and 0.8 M1s1, respectively) (Benitez et al., 2008; Rosal et al., 2009). The NH functional groups in primidone and the NeC double bonds in caffeine are the likely sites of ozone attack. Dodd et al. (2006) suggest that compounds with kO3/kOH (see Table 1) ratios less than 105 will generally be transformed to a large extent by OH radicals rather than by molecular ozone during wastewater ozonation. This corresponds to a rate constant for molecular ozone (kO3) of less than 104 M1s1, since kOH values tend to be on the order of 109 M1s1. While reactions with OH radicals could be a major mechanism contributing to the removal of primidone, DEET and caffeine by ozone, there are a large number of competing reactants for the OH radical in treated wastewater.

4.

Conclusions

Sixteen out of the nineteen targeted compounds were detected in the primary effluent in the wastewater reclamation plant. Most of compounds were found at concentrations on the order of hundreds of ng/L except caffeine and acetaminophen (w105 ng/L), ibuprofen (w104 ng/L), sulfamethoxazole and DEET (w103 ng/L). Only sulfamethoxazole, primidone, caffeine and DEET were frequently detected in the final effluent, but at concentrations on the order of 10e100 ng/L. After activated sludge treatment and membrane filtration, the concentrations of caffeine, acetaminophen, ibuprofen, DEET, tetracycline, and 17a-ethynylestradiol (EE2) had decreased by more than 90%. Erythromycin and carbamazepine, which

were resistant to biological treatment, were eliminated by 74 and 88%, on average, by GAC. Ozonation oxidized most of the remaining compounds by >60%. Primidone, DEET, and caffeine were not amenable to adsorption by GAC and ozonation. The results obtained in this full-scale plant assessment were generally consistent with previous findings from laboratory studies or in grab samples taken from full-scale wastewater treatment facilities, and with expectations based on physical-chemical characteristics of the compounds.

Acknowledgments We thank the management and operations personnel at the F. Wayne Hill Water Resources Center and the Gwinnett County Department of Public Utilities for financial support of this project and for their assistance with sample collection. We extend appreciation to Kate Bronstein and Shannon Weston who helped with the analytical methodology at the beginning of the project.

Appendix. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.07.026.

references

American Water Works Association, 1981. An assessment of microbial activity on GAC. Journal of American Water Works Association 73, 447e454. Andreozzi, R., Canterino, M., Marotta, R., Paxeus, N., 2005. Antibiotic removal from wastewaters: the ozonation of amoxicillin. Journal of Hazardous Materials 122, 243e250. Batt, A.L., Kim, S., Aga, D.S., 2006. Enhanced biodegradation of iopromide and trimethoprim in nitrifying activated sludge. Environmental Science & Technology 40, 7367e7373. Benitez, F.J., Real, F.J., Acero, J.L., Sagasti, J.J., 2008. Oxidation of the Pharmaceutical Primidone by Single Oxidants (UV radiation and ozone) and Advanced Oxidation Processes. 236th ACS National Meeting, Philadelphia, PA. Benotti, M.J., Brownawell, B.J., 2007. Distributions of Pharmaceuticals in an Urban Estuary During both Dry- and Wet-Weather Conditions, pp. 5795e5802. Buerge, I.J., Poiger, T., Muller, M.D., Buser, H.R., 2003. Caffeine, an anthropogenic marker for wastewater contamination of surface waters. Environmental Science & Technology 37, 691e700. Carballa, M., Omil, F., Lema, J.M., 2007. Calculation methods to Perform mass Balances of micropollutants in sewage treatment plants. Application to pharmaceutical and personal care products (PPCPs). Environmental Science & Technology 41, 884e890. Clara, M., Strenn, B., Ausserleitner, M., Kreuzinger, N., 2004. Comparison of the behaviour of selected micropollutants in a membrane bioreactor and a conventional wastewater treatment plant. Water Science & Technology 50, 29e36. Dickenson, E.R.V., Drewes, J.E., Sedlak, D.L., Wert, E.C., Snyder, S. A., 2009. Applying surrogates and Indicators to Assess removal efficiency of Trace organic chemicals during chemical

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 1 8 e5 2 2 8

oxidation of wastewaters. Environmental Science & Technology 43, 6242e6247. Dodd, M.C., Buffle, M.O., vonGunten, U., 2006. Oxidation of antibacterial Molecules by aqueous ozone: moiety-specific reaction kinetics and application to ozone-based wastewater treatment. Environmental Science & Technology 40, 1969e1977. Federle, T.W., Kaiser, S.K., Nuck, B.A., 2002. Fate and effects of triclosan in activated sludge. Environmental Toxicology and Chemistry 21, 1330e1337. Gobel, A., McArdell, C.S., Joss, A., Siegrist, H., Giger, W., 2007. Fate of sulfonamides, macrolides, and trimethoprim in different wastewater treatment technologies. Science of the Total Environment 372, 361e371. Gobel, A., Thomsen, A., McArdell, C.S., Joss, A., Giger, W., 2005. Occurrence and sorption behavior of sulfonamides, macrolides, and trimethoprim in activated sludge treatment. Environmental Science & Technology 39, 3981e3989. Golet, E.M., Xifra, I., Siegrist, H., Alder, A.C., Giger, W., 2003. Environmental exposure assessment of fluoroquinolone antibacterial agents from sewage to soil. Environmental Science & Technology 37, 3243e3249. Hartig, C., Storm, T., Jekel, M., 1999. Detection and identification of sulphonamide drugs in municipal waste water by liquid chromatography coupled with electrospray ionisation tandem mass spectrometry. Journal of Chromatography A 854, 163e173. Hollender, J., Zimmermann, S.G., Koepke, S., Krauss, M., McArdell, C.S., Ort, C., Singer, H., von Gunten, U., Siegrist, H., 2009. Elimination of organic micropollutants in a municipal wastewater treatment plant Upgraded with a full-scale Postozonation followed by sand filtration. Environmental Science & Technology 43, 7862e7869. Howard, P.H., Meylan, W.M., 1997. Handbook of Physical Properties of Orgnanic Chemicals. Lewis Publishers. Huber, M.M., Canonica, S., Park, G.Y., von Gunten, U., 2003. Oxidation of pharmaceuticals during ozonation and advanced oxidation processes. Environmental Science & Technology 37, 1016e1024. Ikehata, K., El-Din, M.G., Snyder, S.A., 2008. Ozonation and advanced oxidation treatment of Emerging organic Pollutants in water and wastewater. Ozone: Science & Engineering 30, 21e26. Joss, A., Keller, E., Alder, A.C., Gobel, A., McArdell, C.S., Ternes, T., Siegrist, H., 2005. Removal of pharmaceuticals and fragrances in biological wastewater treatment. Water Research 39, 3139e3152. Joss, A., Zabczynski, S., Gobel, A., Hoffmann, B., Loffler, D., McArdell, C.S., Ternes, T.A., Thomsen, A., Siegrist, H., 2006. Biological degradation of pharmaceuticals in municipal wastewater treatment: proposing a classification scheme. Water Research 40, 1686e1696. Kasprzyk-Hordern, B., Dinsdale, R.M., Guwy, A.J., 2008. The occurrence of pharmaceuticals, personal care products, endocrine disruptors and illicit drugs in surface water in South Wales, UK. Water Research 42, 3498e3518. Kasprzyk-Hordern, B., Dinsdale, R.M., Guwy, A.J., 2009. The removal of pharmaceuticals, personal care products, endocrine disruptors and illicit drugs during wastewater treatment and its impact on the quality of receiving waters. Water Research 43, 363e380. Kesavan, P.C., Powers, E.L., 1985. Differential modification of oxic and anoxic components of radiation damage in Bacillus megaterium spores by caffeine. International Journal of Radiation Biology 48, 223e233. Kim, S., Eichhorn, P., Jensen, J.N., Weber, S., Aga, D.S., 2005. Removal of antibiotics in wastewater: effect of Hydraulic and solid Retention times on the Fate of tetracycline in the

5227

activated sludge process. Environmental Science & Technology 39, 5816e5823. Kimura, K., Hara, H., Watanabe, Y., 2007. Elimination of selected acidic pharmaceuticals from municipal wastewater by an activated sludge system and membrane Bioreactors. Environmental Science & Technology 41, 3708e3714. Kolpin, D.W., Furlong, E.T., Meyer, M.T., Thurman, E.M., Zaugg, S. D., Barber, L.B., Buxton, H.T., 2002. Pharmaceuticals, Hormones, and other organic wastewater contaminants in U. S. Streams, 1999e2000: a National Reconnaissance. Environmental Science & Technology 36, 1202e1211. Lindberg, R.H., Wennberg, P., Johansson, M.I., Tysklind, Mats, Andersson, B.A.V., 2005. Screening of human antibiotic Substances and Determination of Weekly mass flows in five sewage treatment plants in Sweden. Environmental Science & Technology 39, 3421e3429. Lindqvist, N., Tuhkanen, T., Kronberg, L., 2005. Occurrence of acidic pharmaceuticals in raw and treated sewages and in receiving waters. Water Research 39, 2219e2228. Lindsey, M.E., Meyer, M., Thurman, E.M., 2001. Analysis of Trace levels of sulfonamide and tetracycline antimicrobials in groundwater and surface water using solid-phase extraction and liquid Chromatography/Mass spectrometry. Analytical Chemistry 73, 4640e4646. Miao, X.S., Yang, J.J., Metcalfe, C.D., 2005. Carbamazepine and its metabolites in wastewater and in biosolids in a municipal wastewater treatment plant. Environmental Science & Technology 39, 7469e7475. Nakada, N., Shinohara, H., Murata, A., Kiri, K., Managaki, S., Sato, N., Takada, H., 2007. Removal of selected pharmaceuticals and personal care products (PPCPs) and endocrine-disrupting chemicals (EDCs) during sand filtration and ozonation at a municipal sewage treatment plant. Water Research 41, 4373e4382. Qiang, Z., Adams, C., Surampalli, R., 2004. Determination of ozonation rate constants for lincomycin and Spectinomycin. Ozone: Science & Engineering 26, 525e537. Reungoat, J., Macova, M., Escher, B.I., Carswell, S., Mueller, J.F., Keller, J., 2010. Removal of micropollutants and reduction of biological activity in a full scale reclamation plant using ozonation and activated carbon filtration. Water Research 44, 625e637. Rivera-Cancel, G., Bocioaga, D., Hay, A.G., 2007. Bacterial degradation of N, N-Diethyl-m-Toluamide (DEET): cloning and Heterologous Expression of DEET Hydrolase. Applied Environmental Microbiology 73, 3105e3108. Rosal, R., Rodrı´guez, A., Perdigo´n-Melo´n, J.A., Petre, A., Garcı´aCalvo, E., Gomez, M.J., Agu¨era, A., Ferna´ndez-Alba, A.R., 2009. Degradation of caffeine and identification of the transformation products generated by ozonation. Chemosphere 74, 825e831. Singer, H., Mu¨ller, S., Tixier, C., Pillonel, L., 2002. Triclosan: occurrence and Fate of a widely used Biocide in the aquatic environment: field measurements in wastewater treatment plants, surface waters, and lake sediments. Environmental Science & Technology 36, 4998e5004. Sithole, B.B., Guy, R.D., 1987. Models for tetracycline in aquatic environments. Water, Air, & Soil Pollution 32, 303e314. Snyder, S.A., Adham, S., Redding, A.M., Cannon, F.S., DeCarolis, J., Oppenheimer, J., Wert, E.C., Yoon, Y., 2007. Role of membranes and activated carbon in the removal of endocrine disruptors and pharmaceuticals. Desalination 202, 156e181. Song, W., Cooper, W.J., Peake, B.M., Mezyk, S.P., Nickelsen, M.G., O’Shea, K.E., 2009. Free-radical-induced oxidative and reductive degradation of N, N’-diethyl-m-toluamide (DEET): kinetic studies and degradation pathway. Water Research 43, 635e642.

5228

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 1 8 e5 2 2 8

Stanford, B.D., Weinberg, H.S., 2007. Isotope dilution for quantitation of steroid estrogens and nonylphenols by gas chromatography with tandem mass spectrometry in septic, soil, and groundwater matrices. Journal of Chromatography A 1176, 26e36. Suarez, S., Carballa, M., Omil, F., Lema, J.M., 2008. How are pharmaceutical and personal care products (PPCPs) removed from urban wastewaters? Reviews in Environmental Science and Biotechnology 7, 125e138. Suarez, S., Dodd, M.C., Omil, F., von Gunten, U., 2007. Kinetics of triclosan oxidation by aqueous ozone and consequent loss of antibacterial activity: relevance to municipal wastewater ozonation. Water Research 41, 2481e2490. Ternes, T.A., Herrmann, N., Bonerz, M., Knacker, T., Siegrist, H., Joss, A., 2004. A rapid method to measure the solid-water distribution coefficient (Kd) for pharmaceuticals and musk fragrances in sewage sludge. Water Research 38, 4075e4084. Ternes, T.A., Meisenheimer, M., McDowell, D., Sacher, F., Brauch, H.J., Gulde, B.H., Preuss, G., Wilme, U., Seibert, N.Z., 2002. Removal of pharmaceuticals during drinking water treatment. Environmental Science & Technology 36, 3855e3863. Ternes, T.A., Stuber, J., Herrmann, N., McDowell, D., Ried, A., Kampmann, M., Teiser, B., 2003. Ozonation: a tool for removal of pharmaceuticals, contrast media and musk fragrances from wastewater? Water Research 37, 1976e1982.

Vieno, N.M., Tuhkanen, T., Kronberg, L., 2005. Seasonal variation in the occurrence of pharmaceuticals in effluents from a sewage treatment plant and in the Recipient water. Environmental Science & Technology 39, 8220e8226. Voogt, P., Janex-Habibi, M.L., Sacher, F., Puijker, L., Mons, M., 2008. Development of a Common Priority List of Pharmaceuticals Relevant for the Water Cycle. IWA World Water Congress and Exhibition, Vienna, Austria. Westerhoff, P., Yoon, Y., Snyder, S., Wert, E., 2005. Fate of endocrine-Disruptor, pharmaceutical, and personal care product chemicals during Simulated drinking water treatment processes. Environmental Science & Technology 39, 6649e6663. Ye, Z., Weinberg, H.S., Meyer, M.T., 2007. Trace analysis of trimethoprim and sulfonamide, macrolide, quinolone, and tetracycline antibiotics in chlorinated drinking water using liquid chromatography electrospray tandem mass spectrometry. Analytical Chemistry 79, 1135e1144. Yu, J.T., Bouwer, E.J., Coelhan, M., 2006. Occurrence and biodegradability studies of selected pharmaceuticals and personal care products in sewage effluent. Agricultural Water Management 86, 72e80. Zwiener, C., Frimmel, F.H., 2003. Short-term tests with a pilot sewage plant and biofilm reactors for the biological degradation of the pharmaceutical compounds clofibric acid, ibuprofen, and diclofenac. The Science of the Total Environment 309, 201e211.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 2 9 e5 2 4 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Contrasts in concentrations and loads of conventional and alternative indicators of fecal contamination in coastal stormwater Reagan R. Converse*, Michael F. Piehler, Rachel T. Noble University of North Carolina at Chapel Hill, Institute of Marine Sciences, 3431 Arendell St, Morehead City, NC 28557, USA

article info

abstract

Article history:

Fecal contamination in stormwater is often complex. Because conventional fecal indicator

Received 19 March 2011

bacteria (FIB) cannot be used to ascertain source of fecal contamination, alternative indi-

Received in revised form

cators are being explored to partition these sources. As they are assessed for future use, it

6 May 2011

is critical to compare alternative indicators to conventional FIB under a range of storm-

Accepted 22 July 2011

water delivery conditions. In this study, conventional FIB and fecal Bacteroides spp. were

Available online 29 July 2011

monitored throughout the duration of five storm events from coastal stormwater outfalls in Dare County, North Carolina, USA to characterize relationships among FIB concentra-

Keywords:

tions, alternative fecal markers, and loading of contaminants. Water samples were

Fecal indicator bacteria

collected multiple times during each storm and analyzed for Enterococcus sp. and Escherichia

Fecal Bacteroides spp.

coli using enzymatic tests and fecal Bacteroides spp. by QPCR. Both conventional FIB and

Stormwater

fecal Bacteroides spp. concentrations in stormwater were generally high and extremely

Hydrologic factors

variable over the course of the storm events. Over the very short distances between sites,

Source partitioning

we observed statistically significant spatial and temporal variability, indicating that stormwater monitoring based on single grab-samples is inappropriate. Loading of FIB and fecal Bacteroides spp. appeared to be affected differently by various hydrologic factors. Specifically, Spearman correlations between fecal Bacteroides spp. and drainage area and antecedent rainfall were lower than those between conventional FIB and these hydrologic factors. Furthermore, the patterns of fecal Bacteroides spp. concentrations generally increased over the duration of the storms, whereas E. coli and Enterococcus sp. concentrations generally followed the patterns of the hydrograph, peaking early and tailing off. Given the greater source-specificity and limited persistence of fecal Bacteroides spp. in oxygenated environments, differences in these patterns suggest multiple delivery modes of fecal contamination (i.e. landscape scouring versus groundwater discharge). ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Stormwater runoff is generally characterized as non-point source (NPS) pollution despite often being conveyed through a pipe or conduit to receiving waters. Stormwater

runoff can carry with it fecal contamination from a range of sources, including human, pet, livestock, wildlife, and waterfowl. Land use can influence the type of contamination carried in stormwater runoff, but given the heterogeneity of land use in most watersheds, this relationship can be quite

* Corresponding author. Tel.: þ1 252 726 6841x141. E-mail address: [email protected] (R.R. Converse). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.029

5230

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 2 9 e5 2 4 0

complex (Mallin et al., 2001; Schwab, 2007; Stumpf et al., 2010). Because stormwater entrains fecal indicator bacteria (FIB) that have accumulated on the land since the preceding precipitation event, the load of FIB from stormwater runoff to recreational waters during storms can be over 1000 times higher than loads during baseflow conditions (Krometis et al., 2007). However, the high magnitude of FIB concentrations contributed to receiving waters during a storm event is not necessarily indicative of human fecal contamination. Because animal feces can contain similarly high concentrations of EPA currently-approved FIB (Layton et al., 2010), fecal coliforms (or the dominant subset, Escherichia coli) and Enterococcus sp., elevated levels of conventional FIB have equivocal meanings. Furthermore, conventional FIB have also been shown to survive and grow in sediments (SoloGabriele et al., 2000; Desmarais et al., 2002; Byappanahalli and Fujioka, 2004; Anderson et al., 2005), and this extraenteric persistence could be another source of FIB to receiving waters. While there is a demonstrated public health risk associated with recreating in water affected by NPS fecal contamination, studies have been unable to demonstrate an association between conventional FIB concentrations and human health outcomes or pathogens of concern (Haile et al., 1999; Colford et al., 2007). Fecal contamination in stormwater must be identified and quantified in order to better approximate public health risk associated with recreation in receiving waters. Quantification includes both accurate assessment of the abundance of the indicator of interest and measurement of stormwater discharge. Alternative indicators of fecal contamination with increased human specificity have been recommended for this sort of quantification in order to distinguish between sources of contamination, which may have different relationships with human health outcomes (Savichtcheva and Okabe, 2006). Furthermore, if stormwater inputs are to be mitigated, an understanding of the delivery patterns of both the water and the microbial contaminants must be fully understood. For example, it has been accepted along the North Carolina (NC) coast, that the delivery patterns of microbial contaminants do not match the “first flush” dynamics of delivery observed in other areas and for other types of contaminants (Stumpf et al., 2010; Parker et al., 2010). Fecal Bacteroides spp. have been suggested as an alternative indicator because they are obligate anaerobes, they are expected to have limited persistence outside the human gastrointestinal tract, and molecular markers have been developed for the genus that are generally more concentrated in human fecal material than animal (Fiksdal et al., 1985; Kreader, 1998; Carson et al., 2005; Layton et al., 2006; Okabe and Shimazu, 2007; Converse et al., 2009). Given these ecological differences from conventional FIB, we hypothesized that the concentration and loading of fecal Bacteroides spp. and conventional FIB would be distinctive in stormwater. The objectives of this study were to: 1) examine patterns of conventional FIB concentrations and loads in stormwater from coastal outfalls throughout the duration of storms; 2) examine hydrological characteristics of the storms in order to assess storm variability; 3) use quantitative polymerase chain reaction (QPCR) based methods to assess patterns of the

alternative indicator fecal Bacteroides spp. in stormwater; and 4) connect the hydrologic conditions with numerical results for bacterial concentrations to assess patterns and deduce information about the sources of contamination.

2.

Methods

2.1.

Sampling location

Located on a barrier island in Dare County, NC, Kill Devil Hills and Nags Head are popular resort areas with a peak visitor population of up to 50,000 (Esnard et al., 2001). Bordered on the east by the Atlantic Ocean and the west by Roanoke Sound, Kill Devil Hills and Nags Head span approximately 20 km from north to south. Swim advisories due to elevated FIB levels have occurred frequently in the area, some lasting longer than 30 days (EPA BEACH Report, 2008). Stormwater is a particular problem: 12 extreme storm events categorized as either hurricanes or tropical storms came within 100 km of the towns between 2000 and 2009 (National Hurricane Center, 2011). Five stormwater ocean outfalls maintained by the NC Department of Transportation (NCDOT) in Nags Head and Kill Devil Hills were chosen as study sites (Fig. 1). These outfalls were identified as Sites 1 through 5 (Table 1). At each site, the outfall catch basin was outfitted with a Doppler flow meter (Teledyne ISCO 750 Area Velocity Module, Teledyne ISCO, Lincoln, NE), which measured water velocity, and an ISCO automated water sampler (Model 6712, Lincoln, NE).

2.2.

Field sampling

Water samples were collected during five storms from June 2007 to November 2008. Storm dates and descriptions are given in Table 2. Sampling was initiated when storms were forecasted to produce at least 2.5 cm (1 in) of rainfall according to weather forecasts. Unfortunately, the automated water samplers proved unreliable when programmed to sample on either velocity or level triggers, and samplers had to be programmed to sample at discrete time intervals for each storm. Sampling times and frequency were thus based on qualitative assessments of weather forecasts and flow data in order to capture important parts of the storm hydrographs. Three to six samples were collected at each site over the course of each storm, depending upon the storm duration. Samples were collected from the outfall catch basins following standard methods sampling techniques (APHA, 2005, 2009) and stored at 4  C until they arrived in the laboratory.

2.3. Enumeration of E. coli and Enterococcus sp. in stormwater Enterococcus sp. and E. coli concentrations were measured in duplicate for each sample using defined substrate technology kits, Enterolert and Colilert-18 (IDEXX Laboratories, Westbrooke, ME) following manufacturer’s instructions. Positive wells were converted to a most probable number (MPN) using manufacturer-supplied MPN tables.

5231

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 2 9 e5 2 4 0

Fig. 1 e Map of study area and monitored stormwater outfalls.

2.4.

Preparation of stormwater samples for QPCR

One-hundred milliliter of water samples were filtered in triplicate onto 47 mm diameter, 0.45 mm pore-size polycarbonate filters (type HTTP, Millipore, Bedford, MA) and stored at 80  C until DNA extraction.

2.5.

diluting, fixing in formalin (1% v/v final) and counting cells using SYBR Green (Invitrogen, Carlsbad, CA) following the protocol of Noble and Fuhrman (1998). One-hundred thousand cells were filtered onto 47 mm diameter, 0.45 mm pore-size polycarbonate filters (type HTPP, Milliopore), and filters were stored at 80  C until extraction.

QPCR calibration standards

Bacteroides thetaiotamicron (ATCC 29148) was used as a calibration standard for the fecal Bacteroides spp. QPCR assay. Cells were grown anaerobically in an overnight culture at 37  C in cooked meat medium. A portion of the cell suspension was removed and centrifuged for 5 min at 6000  g. The supernatant fluid was removed and aliquoted for use as a cell standard. Aliquots were frozen at 20  C. Cell counts were obtained by removing a portion of the cell suspension, serially

Table 1 e Site descriptions. Site # (site name) 1 2 3 4 5

(Baum Street) (Martin Street) (Gallery Row) (Conch Street) (Soundside Road)

Drainage area (acres)

Latitude

Longitude

192 366 214 53 31

36 000 3700 N 36 000 2100 N 35 590 1900 N 35 570 5300 N 35 570 2500 N

75 390 1800 W 75 390 0800 W 75 380 3200 W 75 370 4200 W 75 370 2600 W

5232

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 2 9 e5 2 4 0

Table 2 e Storm descriptions. Date June 3e4, 2007 Dec 15e16, 2007 April 21e23, 2008 Sept 5e10, 2008 Nov 4e6, 2008

2.6.

# Days since last rainfall

Total rainfall (cm)

93 49 15 146 51

3.91 5.87 10.85 10.59 22.02

Specimen processing control

A specimen processing control (SPC) was added to each sample in order to estimate PCR inhibition. Salmon sperm DNA (Sigma, St Louis, MO) was added at the beginning of the extraction step at a concentration of 100 ng per 500 mL for each sample, calibration standard, and blank. The QPCR assay targeting the SPC was developed by Haugland et al. (2005). The primers and probe target a segment of the ribosomal RNA gene operon, internal transcribed spacer region 2 of chum salmon, Oncorhynchus keta. All SPC QPCR sample reactions with a quantification threshold value 1.5 greater than that of the calibrators and blanks were considered inhibited. Inhibited samples were diluted 1:10 and 1:100 with sterile water and reanalyzed.

2.7.

DNA extraction

Briefly, 25 mL reactions were prepared using OmniMix beads (a lyophilized premix with 1.5 units of TaKaRa hot start Taq polymerase, 200 mM of dNTPs, 4 mM of MgCl2, and 25 mM HEPES with a pH of 8; Cepheid, Sunnyvale, CA), 1 mM each of forward and reverse primers, 0.1 mM of the TaqMan probe, and 5 mL of sample DNA extract. Reactions were thermal cycled and monitored in a SmartCycler II (Cepheid). Thermal cycling occurred in two stages: first, 2 min at 95  C, followed by 45 cycles of 15 s at 94  C and 30 s at 60  C. All analyses of unknowns were run in duplicate. After manually adjusting the threshold on the SmartCycler II to 30, quantification threshold was determined automatically by the instrument. A duplicate standard curve was run concurrently with samples using the calibrator and three serial 10-fold dilutions. Amplification efficiency (E ) was calculated using the slope of the log standard curve given by the SmartCycler software: E ¼ 10^(slope). Efficiencies ranged from 92 to 100% and averaged 97%, with r2 values between 0.990 and 0.999. The ratio of target DNA in the samples was multiplied by the amount of target DNA in the calibrator to get the sample quantities in number of cells following Pfaffl (2001). Results from the SPC QPCR indicated that all samples were inhibited and required 1:100 dilution of the DNA extracts to alleviate inhibition. To compensate for this dilution, sample quantities were also multiplied by the dilution factor (100) to get a corrected quantity.

2.9.

Data analysis

DNA was extracted from the polycarbonate filters using the DNA-EZ RW04 Extraction Kit (GeneRite, Brunswick, NJ). Filters were transferred into 2 mL semiconical screw-cap microcentrifuge tubes pre-loaded with 0.3 g of sterile 0.1 mm zirconia/ silica beads. Four hundred and ninety microliters of AE Buffer (Qiagen, Valencia, CA) and 10 mL of the SPC (10 mg/mL salmon sperm DNA) were dispensed into each tube. Tubes were bead milled in an eight-position mini bead beater (Bio-Spec Corp, Bartlesville, OK) for 2 min, followed by centrifugation in an Eppendorf Microfuge for 1 min at 12,000  g. Supernatants were transferred to sterile 1.7 mL microcentrifuge tubes and centrifuged at 12,000  g for 5 min. Supernatant was transferred to a sterile 1.7 mL microcentrifuge tube, mixed by vortexing with 500 mL of Binding Buffer, and applied to a DNAsure column. Columns were inserted into collection tubes and centrifuged at 12,000  g for 1 min. Columns were transferred to new collection tubes, and 500 mL of EZ-Wash Buffer was applied to the columns. Columns were again centrifuged at 12,000  g for 1 min. After transferring to new collection tubes, columns were washed again with 500 mL of Wash Buffer and 1 min of centrifugation. Columns were transferred to sterile, low retention 1.7 mL microcentrifuge tubes (Genemate, ISC Bioexpress, Kaysville, UT), and 50 mL of DNA Elution Buffer was applied to the center of the columns and allowed to incubate at room temperature for 1 min. Columns were again centrifuged for 1 min at 12,000  g. Extracted DNA in the microcentrifuge tube was stored at 20  C.

Fecal Bacteroides spp. samples that yielded a “non-detect” QPCR result were assigned a concentration of 5 cells per 100 mL. Enterolert and Colilert results above the detection limit were assigned the highest value within the limits of detection, 24,196 MPN/100 mL; results below the detection limit were assigned the lowest value within the limits of detection, 10 MPN/100 mL. Instantaneous loads were calculated by multiplying the instantaneous flow velocity (in 100 mL/s) by indicator concentrations (in MPN or CE per 100 mL). Event loads were not calculated because of the complexity of the storms and the resultant difficulties with programming the autosamplers. FIB and fecal Bacteroides spp. concentrations and instantaneous loads did not have normal distributions and were logtransformed to reduce but not eliminate the skewness. Because data were not normally distributed, all statistical correlations and differences were tested in SPSS Version 18.0 statistical analysis software (Chicago, IL) using nonparametric statistical tests: Spearman-Rank analyses, the KruskaleWallis Test, and the ManneWhitney U-Test.

2.8.

Our goal was to identify spatial and temporal dynamics in concentrations and instantaneous loads of conventional FIB and an alternative indicator using an intensive sampling approach to study entire storm duration. Flow, conventional FIB concentrations, fecal Bacteroides spp. concentrations, and

QPCR analyses and quantification of target DNA

The SPC QPCR assays were conducted following Haugland et al. (2005). Fecal Bacteroides spp. primers and probes and the assay used for QPCR are described in Converse et al. (2009).

3.

Results and discussion

3.1. Concentrations and loading patterns over the course of storms

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 2 9 e5 2 4 0

instantaneous FIB loads exhibited considerable variability over the course of each storm event at each site. Flow and conventional FIB and fecal Bacteroides spp. concentration data from the April 2008 storm are shown in Fig. 2 to demonstrate this variability.

3.1.1.

5233

results of this study are suggestive of a wash-off effect and likely differ from Surbeck’s findings because Dare County, NC is much less urbanized and populous than the southern California site used in Surbeck’s study. It is also common for rainfall to occur in sporadic pulses along the NC coast, allowing for a difference in observed delivery patterns.

E. coli and Enterococcus sp.

Stormwater concentrations of Enterococcus sp. surpassed the NC single sample recreational water quality standard of 104 MPN per 100 mL during every storm event and at all sites. Even though NC does not manage recreational water quality using E. coli, the average E. coli stormwater concentrations exceeded the single sample fecal coliform marine recreational water quality standard of 400 MPN per 100 mL during 4 of the 5 storms (Fig. 3) and at 4 of the 5 sites (Fig. 4).Over the course of each storm, average E. coli concentrations in stormwater ranged from 1144 MPN per 100 mL to 5413 MPN per 100 mL during the five study storms. Average Enterococcus sp. concentrations were similarly high, ranging from 648 MPN per 100 mL to 8025 MPN per 100 mL. These high concentrations of FIB have important ramifications because children recreate in areas of pooled stormwater (JD Potts, NC Department of Environment and Natural Resources, Shellfish Sanitation and Recreational Water Quality Section, personal communication) in eastern NC and are more susceptible to waterborne disease than their adult counterparts (Wade et al., 2008). Previous research has demonstrated that on the scale of several hours E. coli and Enterococcus sp. concentrations in stormwater remained high with little variability (Parker et al., 2010). Similar analyses were performed in the current study with storms of longer duration (on the scale of days rather than hours). Because the autosampler programming using level and velocity pacing failed to capture such complex storms, the time intervals between samples varied from storm to storm. In order to standardize results for temporal comparison, average concentrations and loads were calculated for the first 24 h of a storm event (beginning when flow was first observed to be greater than 0 at a site), the last 24 h of a storm event, and the entire storm. When the mean concentrations or loads using data from the entire storm were higher than the means during the first 24 or last 24 h, it was deduced that the concentration was highest during the middle of the storm. Table 3 tracks these concentrations at each site for the storms that lasted longer than 48 h (3 of the 5 storms). There was a great deal of variability observed in the timing of the highest FIB concentrations over the course of the storms: highest in the first 24 h 54% of the time; highest in the last 24 h 23% of the time, and highest in the middle 23% of the time. In general, however, conventional FIB concentrations were more than an order of magnitude higher than the single sample recreational water quality standards throughout the entire duration of the storms. Despite low flow speeds early in the storms, the highest E. coli and Enterococcus sp. loads were often observed during the first 24 h of a storm (62% of cases) and never during the last 24 h. This is in direct opposition with the findings of Surbeck et al. (2006), who hypothesized that stormwater loads of conventional FIB ubiquitous in urban environments will remain constant and high. Surbeck’s hypothesis assumes that because concentrations of conventional FIB are uniformly high there is no visible wash-off as FIB are entrained by stormwater. The

3.1.2.

Fecal Bacteroides spp.

Because conventional FIB can be found at consistently high concentrations in soil and water (Solo-Gabriele et al., 2000; Desmarais et al., 2002; Byappanahalli and Fujioka, 2004; Anderson et al., 2005), an alternative FIB with greater hostspecificity and shorter persistence in the environment was used to further understand the dynamics of load and possible fecal source. Fecal Bacteroides spp. are expected to persist in the environment for much shorter periods than FIB (Rolfe et al., 1977; Kreader, 1998; Walters and Field, 2009) and have relatively lower concentrations in animal scat than human waste (Converse et al., 2009). Given this, in an area that receives intense periods of precipitation, we did not expect fecal Bacteroides spp. concentrations to behave like E. coli and Enterococcus sp. over the course of a storm event. Fecal Bacteroides spp. concentrations were often quite high and were highest in the last 24 h of the storms in 50% of cases (Table 3). As a result, fecal Bacteroides spp. loads were highest in the middle or last 24 h of storms, generally opposite the patterns observed for E. coli and Enterococcus sp. Low fecal Bacteroides spp. loads in the first 24 h of storms were the result of not only low concentrations but compounded by low flow. It is important to note, however, that this low flow did not prevent E. coli and Enterococcus sp. loads from often being highest at the beginning of storms. Because surface fecal depositions come almost exclusively from animals with two to five orders of magnitude lower fecal Bacteroides spp. concentrations than humans (Converse et al., 2009), these results suggest that septic leachate is contributing significantly to the fecal Bacteroides spp. load in Dare County’s stormwater. Septic system usage is high on NC’s Outer Banks with 85% of properties on Nags Head and Kill Devil Hills using septic systems, compared with the national average of roughly 20% (KE Conn, UNC Institute of Marine Sciences, personal communication). Results from this study suggest that as a storm continues, the water table rises, bringing fecal Bacteroides spp. from septic leachate to the surface and into stormwater. This same pattern of increase over the course of a storm was not discernible when measuring conventional FIB because pet waste contributes higher conventional FIB loads than septic sources by up to three orders of magnitude (Kelsey, 2004). As a result, the septic signal was overwhelmed by the signal from animal waste. It is feasible that in watersheds with very large agricultural operations, where animal fecal production is orders of magnitude higher than human fecal pollution, stormwater fecal Bacteroides spp. concentrations may be as high or higher than those concentrations found in stormwater affected by septic leachate. In these situations, it will be important to consider the context of the high fecal Bacteroides spp. concentrations before assuming that they are suggestive of human fecal contamination. In our study area, however, there were not significant livestock operations to confuse fecal Bacteroides spp. interpretations.

5234

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 2 9 e5 2 4 0

Fig. 2 e Flow and log10 indicator concentrations during the April 2008 storm. Flow is in L/s; Enterococcus sp. and E. coli concentrations are in MPN/100 mL; fecal Bacteroides spp. concentrations are in CE/100 mL. Time on the x-axis indicates the number of hours since the storm began.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 2 9 e5 2 4 0

5235

Fig. 3 e Average log10 indicator concentrations (MPN/100 mL or CE/100 mL) and log10 instantaneous loads (MPN/s or CE/s) for each storm. Data from all sites were pooled in the analyses. Error bars represent standard error. The solid line represents the NC recreational water single sample standard for Enterococcus sp., 104 MPN per 100 mL. The dotted line represents a single sample E. coli standard of 400 MPN per 100 mL.

3.2.

Differences between storms and sites

Fig. 3 summarizes average conventional and alternative FIB concentrations and loads for each storm. Results from KruskaleWallis tests reveal significant differences in E. coli, Enterococcus sp., and fecal Bacteroides spp. concentrations and flow across storm events ( p ¼ 0.007, <0.001, <0.001, 0.015, respectively). Specifically, fecal Bacteroides spp. concentrations were significantly higher during the December 2007 and November 2008 storms, according the Wilcoxon test (using the Bonferroni-corrected significance level of 0.005). Enterococcus sp. concentrations were significantly lower during the December 2007 and April 2008 storms, using the same Bonferroni-corrected significance level, and E. coli concentrations were higher with marginal significance during the September and November 2008 storms. There were also

significant differences in loads of E. coli, Enterococcus sp., and fecal Bacteroides spp. between storms when using the KruskaleWallis test ( p ¼ 0.001, <0.001, and <0.001, respectively). The November 2008 storm was most often significantly different from the other storms in conventional and alternative indicator loads according to the Wilcoxon test and using the Bonferroni-corrected significance level of 0.005. This storm had the highest loads by orders of magnitude, which was expected given that this storm had the highest rainfall totals. Rainfall amounts have been shown to be significantly correlated with conventional FIB loads (Stumpf et al., 2010). Magnitude of contamination was also significantly different among stormwater outfalls even though the sites were geographically close (all sites within 10 km of one another). Fig. 4 shows average conventional and alternative FIB concentrations and drainage-area-normalized loads for

5236

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 2 9 e5 2 4 0

Fig. 4 e Average log10 indicator concentrations (MPN/100 mL or CE/100 mL) and drainage-area-normalized instantaneous loads (MPN/s or CE/s) for each site. Data from all storms were pooled in the analyses. Error bars represent standard error. The solid line represents the NC recreational water single sample standard for Enterococcus sp., 104 MPN per 100 mL. The dotted line represents a single sample E. coli standard of 400 MPN per 100 mL.

each site. Enterococcus sp. concentrations, fecal Bacteroides spp. concentrations, and flow were significantly different among sites: p ¼ 0.003 for Enterococcus sp.; p ¼ 0.002 for fecal Bacteroides spp.; and p < 0.001 for flow. There was not a significant difference in E. coli concentrations between sites. However, E. coli, Enterococcus sp., and fecal Bacteroides spp. loads (normalized by drainage area) were significantly different between sites ( p < 0.001, <0.001, and 0.018, respectively). In general, Site 5 had the lowest concentrations and normalized loads. This site is the most commercialized and least residential; industrial and commercial areas have less stormwater fecal contamination than residential areas (Selvakumar and Borst, 2006; Surbeck, 2009).

3.3. Correlations between loads, concentrations, and hydrographic conditions When data were pooled across sites and storms, E. coli, Enterococcus sp., and fecal Bacteroides spp. concentrations in stormwater were all significantly correlated with outfall

drainage area and antecedent rainfall (Table 4). Flow was significantly correlated with storm rainfall totals (r ¼ 0.294, p ¼ 0.002, n ¼ 111), though none of the indicator concentrations were. As a result, conventional FIB and fecal Bacteroides spp. loads (which depend upon both concentration and flow) were all significantly (or marginally significantly) correlated with total rainfall, antecedent rainfall, and drainage area (Table 4). The correlations between drainage area and concentrations of FIB or fecal Bacteroides spp. were expected: with greater area, there is greater opportunity for fecal contamination, and hence heightened concentration of FIB and other markers in outfalls. The significant correlation between antecedent rainfall and E. coli and Enterococcus sp. concentrations is not consistent with what has been observed in rural NC tidal creeks (Stumpf et al., 2010). However, Kelsey (2004) and Hathaway et al. (2010) demonstrated relationships between antecedent rainfall levels and concentrations of fecal coliforms and Enterococcus sp. in urbanized watersheds. The correlations observed in this study, Kelsey (2004) and Hathaway et al. (2010) suggest that

Table 3 e Average indicator concentrations and instantaneous loads during the first 24 h of storms, last 24 h of storms, and entire storms. Only storms longer than 48 h are included. N/A indicates that the site was not sampled during a particular storm. Site

Storm

E. coli concentration (MPN/100 mL) 1st 24 h

Last 24 h

Entire storm

Enterococcus concentration (MPN/100 mL) 1st 24 h

Last 24 h

Bacteroides concentration (CE/100 mL)

Entire storm

1st 24 h

E. coli load (MPN/s)

Last 24 h

Entire storm

1st 24 h

Bacteroides load (CE/s)

Enterococcus load (MPN/s)

Last 24 h

Entire storm

1st 24 h

Last 24 h

Entire storm

1st 24 h

Last 24 h

Entire storm

April 2008 Sept. 2008 Nov. 2008

8285, n¼1 397, n¼2 8063, n¼2

935, n¼2 121, n¼1 286, n¼2

3678, n¼6 325, n¼5 3176, n¼7

1540, n¼1 1.39e4, n¼2 2679, n¼2

415, n¼2 1201, n¼1 327, n¼2

1088, n¼6 7475, n¼5 1897, n¼7

1.21e5, n¼1 1.33e5, n¼2 3.78e5, n¼2

1.99e5, n¼2 1.99e5, n¼1 3.73e5, n¼2

1.36e5, n¼6 1.33e5, n¼5 4.61e5, n¼7

5.73e5, n¼1 7.73e4, n¼2 1.99e6, n¼2

2.34e5, n¼2 1.16e4, n¼1 5.55e4, n¼2

6.06e5, n¼6 8.23e4, n¼5 8.46e5, n¼7

1.3e5, n¼1 2.73e6, n¼2 1.23e6, n¼2

1.32e5, n¼2 1.15e5, n¼1 6.45e4, n¼2

3.31e5, n¼6 1.67e6, n¼5 8.51e5, n¼7

1.07e7, n¼1 2.6e7, n¼2 1.99e7, n¼2

4.27e7, n¼2 1.91e7, n¼1 6.93e6, n¼2

2.76e7, n¼6 2.87e7, n¼5 1.78e7, n¼7

2

April 2008 Sept. 2008 Nov. 2008

121, n¼1 6488, n¼1 3618, n¼2

667, n¼1 1333, n¼1 1919, n¼2

1004, n¼6 4169, n¼5 2889, n¼7

200, n¼1 6867, n¼1 4950, n¼2

373, n¼1 691, n¼1 6772, n¼2

1107, n¼6 7253, n¼5 2002, n¼7

N/A

3.51e4, n¼5 1.14e5, n¼5 5.16e4, n¼7

2.98e4, n¼1 8.94e6, n¼1 6.55e6, n¼2

2.58e5, n¼1 8.33e5, n¼1 2.51e6, n¼2

6.33e5, n¼6 5.02e6, n¼5 4.71e6, n¼7

5.7e4, n¼1 1.17e7, n¼1 9.49e6, n¼2

5.49e4, n¼1 4.32e5, n¼1 2.62e6, n¼2

5.02e5, n¼6 8.06e6, n¼5 1.16e7, n¼7

N/A

3.31e5, n¼1 1.07e5, n¼2

4.44e4, n¼1 2.55e4, n¼1 2.37e4, n¼2

3.04e8, n¼1 2.21e8, n¼2

2.89e7, n¼1 1.6e7, n¼1 3.11e7, n¼2

1.18e7, n¼5 1.68e8, n¼5 9.59e7, n¼7

April 2008 Sept. 2008 Nov. 2008

575, n¼2 N/A

710, n¼2 N/A

640, n¼5 N/A

925, n¼2 N/A

150, n¼2 N/A

556, n¼5 N/A

6.26e4, n¼2 N/A

4.49e5, n¼2 N/A

2.08e5, n¼5 N/A

1.31e5, n¼2 N/A

1.14e5, n¼2 N/A

1.27e5, n¼5 N/A

1.31e5, n¼2 N/A

3.66e4, n¼2 N/A

9.57e4, n¼5 N/A

0, n¼2 N/A

6.8e7, n¼2 N/A

3.12e7, n¼5 N/A

4144, n¼2

237, n¼1

1649, n¼7

4276, n¼2

405, n¼1

2951, n¼7

1.83e5, n¼2

1.03e5, n¼1

1.21e5, n¼7

8.63e6, n¼2

6.08e5, n¼1

3.19e6, n¼7

1.18e7, n¼2

1.04e6, n¼1

5.19e6, n¼7

4.86e8, n¼2

2.66e8, n¼1

2.63e8, n¼7

April 2008 Sept. 2008 Nov. 2008

20, n¼1 1607, n¼1 3483, n¼2

148, n¼2 6131, n¼1 625, n¼2

214, n¼6 2985, n¼5 1771, n¼6

520, n¼1 1968, n¼1 2076, n¼2

167, n¼2 1.47e4, n¼1 1083, n¼2

327, n¼6 7652, n¼5 181, n¼6

N/A

1.03e6, n¼5 6.98e4, n¼5 3.40e4, n¼6

1.07e5, n¼1 9.22e5, n¼1 2.86e6, n¼2

1.16e5, n¼2 3.87e5, n¼1 4.32e5, n¼2

2.25e5, n¼6 5.33e5, n¼5 1.81e6, n¼6

1.40e5, n¼1 1.19e6, n¼1 1.62e6, n¼2

1.30e5, n¼2 9.26e5, n¼1 1.88e5, n¼2

3.04e5, n¼6 1.17e6, n¼5 1.2e6, n¼6

N/A

5, n¼1 1.48e4, n¼2

1.53e5, n¼2 8.20e4, n¼1 5.39e4, n¼2

2.73e3, n¼1 1.26e7, n¼2

1.77e8, n¼2 5.18e6, n¼1 2.65e7, n¼2

8.11e7, n¼5 1.42e7, n¼5 2.41e7, n¼6

April 2008 Sept. 2008 Nov. 2008

100, n¼1 1.05e4, n¼2 N/A

20, n¼1 2.15e4, n¼2 N/A

227, n¼6 1.42e4, n¼5 N/A

100, n¼1 8322, n¼2 N/A

150, n¼1 1.40e4, n¼2 N/A

275, n¼6 9719, n¼5 N/A

5, n¼1 2.59e5, n¼2 N/A

5, n¼1 1.11e5, n¼2 N/A

5, n¼6 1.48e5, n¼5 N/A

630, n¼1 1.42e6, n¼2 N/A

660, n¼1 0, n¼2 N/A

3.67e4, n¼6 5.67e5, n¼5 N/A

63, n¼1 6.0e5, n¼2 N/A

4420, n¼1 0, n¼2 N/A

3.84e4, n¼6 2.40e5, n¼5 N/A

0, n¼1 1.62e8, n¼2 N/A

0, n¼1 0, n¼2 N/A

189, n¼6 6.48e7, n¼5 N/A

4

6

7

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 2 9 e5 2 4 0

1

5237

5238

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 2 9 e5 2 4 0

Table 4 e Significant Spearman correlation coefficients between indicator concentrations and instantaneous loads (pooled across sites and storms) and hydrologic conditions. NS indicates that there was not a significant correlation between factors. Measurement Concentration

Load

Indicator

Flow

Total rainfall

Antecedent rainfall 0.301, p ¼ 0.001, n ¼ 106 0.341, p ¼ 0.001, n ¼ 106 0.586, p < 0.001, n ¼ 106

0.272, p ¼ 0.009, n ¼ 106 0.309, p ¼ 0.001, n ¼ 106 0.512, p < 0.001, n ¼ 106

0.241, p ¼ 0.013, n ¼ 91 0.172, p ¼ 0.079, n ¼ 105 0.284, p ¼ 0.003, n ¼ 105

0.299, p ¼ 0.002, n ¼ 91 0.508, p < 0.001, n ¼ 105 0.517, p < 0.001, n ¼ 105

Bacteroides

NS

NS

E. coli

NS

NS

Enterococcus

NS

NS

Bacteroides

0.803, p < 0.001, n ¼ 91

E. coli

0.863, p < 0.001, n ¼ 105

Enterococcus

0.846, p < 0.001, n ¼ 105

0.283, p ¼ 0.003, n ¼ 91 0.349, p < 0.001, n ¼ 105 0.300, p ¼ 0.002, n ¼ 105

conventional FIB become increasingly concentrated on land in developed areas when there has been no rain to wash the FIB downstream. The landscape of NC’s Outer Banks likely strengthened the correlations between conventional FIB concentrations and antecedent rainfall conditions: sites are flat but for deep stormwater ditches; and conventional FIB deposited on the flat land will not be easily transferred to the ditches without a rain event. E. coli and Enterococcus sp. survival on land are affected by a range of meteorological factors such as temperature, humidity, and exposure to sunlight (Hathaway et al., 2010), and differences in the strength of the correlations are likely due to differential die-off of the bacteria. Enterococcus sp. have been shown to survive longer than E. coli in most sediments (Howell et al., 1996), though the concentration of neither is likely to decline quickly; research on fecal coliform densities in dog feces has shown no attenuation of concentration for up to 30 days after deposition (Weiskel et al., 1996). The significant correlation between fecal Bacteroides spp. and antecedent rainfall is more difficult to explain; the limited persistence of Bacteroides spp. (and Bacteroides spp. genetic markers) in oxygenated environments has been demonstrated in numerous studies (Kreader, 1998; Okabe and Shimazu, 2007; Walters et al., 2009). However, when our data were separated by site, there were no significant correlations between fecal Bacteroides spp. and antecedent rainfall at four of the five sites (though the significant correlations between antecedent rainfall and Enterococcus sp. and E. coli were still observed at 4 of the 5 sites). A significant correlation between fecal Bacteroides spp. and antecedent rainfall (r ¼ 0.629, p ¼ 0.003, n ¼ 20) was only observed at Site 5. Though this correlation is still difficult to explain and will require further examination in the future, it does not appear that antecedent rainfall is always an important factor in fecal Bacteroides spp. concentration and load in stormwater. Several studies have found significant correlations between conventional FIB concentrations and stormwater flow (e.g. Davis et al., 1977; Stumpf et al., 2010). However, these studies have been performed in tidal creeks and compared baseflow to stormflow concentrations. The current study monitored flow in stormwater outfalls that experienced no baseflow (i.e. the junction boxes were dry during dry weather), and no significant correlations were found between

Area

conventional FIB concentrations or fecal Bacteroides spp. concentrations and flow when data were pooled across sites and storms. Even when data were separated by site and storm, significant correlations were observed less than 25% of the time and with no clear trends. These results were consistent with Surbeck et al. (2006), which found that fecal pollutant concentrations were insensitive to stormwater flow. The differences in FIB indicator correlations with flow between these studies is likely due to: 1) the urbanized nature of the study areas in this and Surbeck’s studies; and 2) the inclusion of baseflow conditions in the former studies. Results from this study suggest that the relationship between flow and FIB concentrations is not constant in stormwater and that FIB concentrations are likely influenced more strongly by other factors. The finding of extreme variability of FIB concentrations and loading across sites and time is consistent with McCarthy et al. (2007). They found that E. coli loads at four different sites in Melbourne, Australia correlated differently with meteorological and antecedent climatic conditions with little consistency among sites. Though some differences among sites and storms can be attributed to measurement-related uncertainties (McCarthy et al., 2008), data from this study highlights the temporally- and spatially-variable patterns of fecal pollution loading in stormwater.

4.

Conclusions and implications

This study demonstrates that stormwater conveyed to important bathing beaches in coastal NC is contaminated with high levels of FIB. It also illustrates differences in patterns of delivery of different contaminants and highlights the variability observed across time and space in stormwater delivery. Patterns of concentrations and loads were effectively used to deduce that conventional FIB and fecal Bacteroides spp. signals originate from different sources, making it likely that a complex mixture of fecal contamination is being delivered to the coast. Fecal Bacteroides spp. and conventional FIB concentration and loading patterns were also found to be highly variable over the course of single storms. Fecal Bacteroides spp.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 2 9 e5 2 4 0

concentrations in particular were often found to be highest at the end of storms. The unique characteristics of fecal Bacteroides spp., limited persistence as well as higher concentrations in human fecal material than most animal waste, suggest septic contributions to stormwater in Dare County that would be neglected with exclusive monitoring of conventional FIB. Such late peaks in fecal Bacteroides spp. concentrations and loads mean that current stormwater sampling strategies that rely on a single sample taken early in the storm may underestimate fecal pollution from relevant sources and that grabsample “snapshots” are not temporally representative of an entire storm. Accurate total maximum daily load (TMDL) calculations will require multiple sampling during each storm with sample times ideally corresponding with important changes in the hydrograph. Furthermore, stormwater mitigation strategies involving reactors intended to collect and treat only the first flush of stormwater are unlikely to be beneficial in areas prone to discharge of contaminated groundwater through hydrological connection to shallow septic systems.

Acknowledgments We thank all at the UNC CSI for their assistance with sample and data collection, including Dr. Nancy White, Corey Adams, Matt Lusk, and Karen Fisher. Thanks to UNC CSI personnel also for watershed parameter and on-site reconnaissance during storms. We are also appreciative of Dr. Kathy Conn for her insights and revisions. We would like to thank the NC Department of Environmental and Natural Resources (NCDENR), the NC Department of Transportation, the Town of Nags Head, and Moffatt & Nichol for logistical and project support. Funding has been provided by NCDENR contract W07035S-2-Moffatt & Nichol-NCDENR Ocean Outfall StudySCO #07-07127-01.

references

American Public Health Association, 2005. Standard Methods for the Examination of Water and Wastewater, 21st ed. American Public Health Association, Washington, DC. American Public Health Association, 2009. Standard Methods for the Examination of Water and Wastewater. Revised online ed.. American Public Health Association, Washington, DC. Anderson, K.L., Whitlock, J.E., Harwood, V.J., 2005. Persistence and differential survival of fecal indicator bacteria in subtropical waters and sediments. Applied and Environmental Microbiology 71, 3041e3048. Byappanahalli, M., Fujioka, R., 2004. Indigenous soil bacteria and low moisture may limit but allow faecal bacteria to multiply and become a minor population in tropical soils. Water Science and Technology 50, 27e32. Carson, C.A., et al., 2005. Specificity of a Bacteroides thetaiotaomicron marker for human feces. Applied and Environmental Microbiology 71, 4945e4949. Colford, J.M., et al., 2007. Water quality indicators and the risk of illness at beaches with nonpoint sources of fecal contamination. Epidemiology 18, 27e35.

5239

Converse, R.R., Blackwood, A.D., Kirs, M., Griffith, J.F., Noble, R.T., 2009. Rapid QPCR-based assay for fecal Bacteroides spp. as a tool for assessing fecal contamination in recreational waters. Water Research 43, 4828e4837. Davis, E., Casserly, D.M., Moore, J.D., 1977. Bacterial relationships in stormwaters. Water Research Bulletin 13, 895e905. Desmarais, T.R., Solo-Gabriele, H.M., Palmer, C.J., 2002. Influence of soil on fecal indicator organisms in a tidally influenced subtropical environment. Applied and Environmental Microbiology 68, 1165e1172. Esnard, A., Brower, D., Bortz, B., 2001. Coastal hazards and the built environment on barrier islands: a retrospective view of Nags Head in the late 1990s. Coastal Management 29, 53e72. Fiksdal, L., Maki, J.S., LaCroix, S.J., Staley, J.T., 1985. Survival and detection of Bacteroides spp., prospective indicator. Applied and Environmental Microbiology 49, 148e150. Haile, R.W., Witte, J.S., Gold, M., Cressey, R., McGee, C., Millikan, R.C., Glasser, A., Harawa, N., Ervin, C., Harmon, P., Harper, J., Dermand, J., Alamillo, J., Barrett, K., Nides, M., Wang, G., 1999. The health effects of swimming in ocean water contaminated by storm drain runoff. Epidemiology 10, 355e363. Hathaway, J., Hunt, W.F., Simmons, O.D., 2010. Statistical evaluation of factors affecting indicator bacteria in urban storm-water runoff. Journal of Environmental Engineering, 1360e1367. Haugland, R., Siefring, S., Wymer, L., Brenner, K., Dufour, A., 2005. Comparison of measurements in freshwater at two recreational beaches by quantitative polymerase chain reaction and membrane filter culture analysis. Water Research 39, 559e568. Howell, J., Coyne, M.S., Cornelius, P.L., 1996. Effect of sediment particle size and temperature on fecal bacteria mortality rates and the fecal coliform/fecal Streptococci ratio. Journal of Environmental Quality 25, 1216e1220. Kelsey, H., 2004. Using geographic information systems and regression analysis to evaluate relationships between land use and fecal coliform bacterial pollution. Journal of Experimental Marine Biology and Ecology 298, 197e209. Kreader, C.A., 1998. Persistence of PCR-detectable Bacteroides distasonis from human feces in river water. Applied and Environmental Microbiology 64, 4103e4105. Krometis, L., et al., 2007. Intra-storm variability in microbial partitioning and microbial loading rates. Water Research 41, 506e516. Layton, A., et al., 2006. Development of Bacteroides 16S rRNA gene TaqMan-based real-time PCR assays for estimation of total, human, and bovine fecal pollution in water. Applied and Environmental Microbiology 72, 4214e4224. Layton, B.A., Walters, S.P., Lam, L.H., Boehm, A.B., 2010. Enterococcus species distribution among human and animal hosts using multiplex PCR. Journal of Applied Microbiology 109, 539e547. Mallin, M., Ensign, S.H., McIver, M.R., Shank, G.C., Fowler, P.K., 2001. Demographic, landscape, and meteorological factors controlling the microbial pollution of coastal waters. Hydrobiologia 460, 185e193. McCarthy, D., Mitchell, V.G., Deletic, A., Diaper, C., 2007. Escherichia coli in urban stormwater: explaining their variability. Water Science and Technology 56, 27e34. McCarthy, D.T., Deletic, A., Mitchell, V.G., Fletcher, T.D., Diaper, C., 2008. Uncertainties in stormwater E. coli levels. Water Research 42, 1812e1824. National Hurricane Center, 2011. NHC Archive of Hurricane Seasons. National Oceanic and Atmospheric Administration, Washington, DC. http://www.nhc.noaa.gov/pastall.shtml. Noble, R.T., Fuhrman, J.A., 1998. Use of SYBR Green I for rapid epifluorescence counts of marine viruses and bacteria. Aquatic Microbial Ecology 14, 113e118.

5240

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 2 9 e5 2 4 0

Okabe, S., Shimazu, Y., 2007. Persistence of host-specific BacteroidesePrevotella 16S rRNA genetic markers in environmental waters: effects of temperature and salinity. Applied Microbiology and Biotechnology 76, 935e944. Parker, J.K., McIntyre, D., Noble, R.T., 2010. Characterizing fecal contamination in stormwater runoff in coastal North Carolina, USA. Water Research 44, 4186e4194. Pfaffl, M.W., 2001. A new mathematical model for relative quantification in real-time RT-PCR. Nucleic Acids Research 29 (9), e45. Rolfe, R.D., Hentges, D.J., Barrett, J.T., Campbell, B.J., 1977. Oxygen tolerance of human intestinal anaerobes. American Journal of Clinical Nutrition 30, 1762e1769. Savichtcheva, O., Okabe, S., 2006. Alternative indicators of fecal pollution: relations with pathogens and conventional indicators, current methodologies for direct pathogen monitoring and future application perspectives. Water Research 40, 2463e2476. Schwab, K.J., 2007. Are existing bacterial indicators adequate for determining recreational water illness in waters impacted by nonpoint pollution? Epidemiology 18, 21e22. Selvakumar, A., Borst, M., 2006. Variation of microorganism concentrations in urban stormwater runoff with land use and seasons. Journal of Water and Health 4, 109e124. Solo-Gabriele, H., Wolfert, M.A., Desmarais, T.R., Palmer, C.J., 2000. Sources of Escherichia coli in a coastal subtropical estuary. Applied and Environmental Microbiology 66, 230e237.

Stumpf, C.H., Piehler, M.F., Thompson, S., Noble, R.T., 2010. Loading of fecal indicator bacteria in North Carolina tidal creek headwaters: hydrographic patterns and terrestrial runoff relationships. Water Research 44, 4704e4715. Surbeck, C., Jiang, S.C., Ahn, J.H., Grant, S.B., 2006. Flow fingerprinting fecal pollution and suspended solids in stormwater runoff from an urban coastal watershed. Environmental Science & Technology 40, 4435e4441. Surbeck, C.Q., 2009. Factors influencing the challenges of modelling and treating fecal indicator bacteria in surface waters. Ecohydrology 2, 399e403. US EPA, 2008. BEACH Report: North Carolina’s 2007 Swimming Season. Environmental Protection Agency, Washington, DC. http://water.epa.gov/type/oceb/beaches/2007_index.cfm. Wade, T.J., et al., 2008. High sensitivity of children to swimmingassociated gastrointestinal illness. Epidemiology 19, 375e383. Walters, S.P., Field, K.G., 2009. Survival and persistence of human and ruminant-specific faecal Bacteroidales in freshwater microcosms. Environmental Microbiology 11, 1410e1421. Walters, S.P., Yamahara, K.M., Boehm, A.B., 2009. Persistence of nucleic acid markers of health-relevant organisms in seawater microcosms: implications for their use in assessing risk in recreational waters. Water Research 43, 4929e4939. Weiskel, P., Howes, B.L., Heufelder, G.R., 1996. Coliform contamination of a coastal embayment: sources and transport pathways. Environmental Science & Technology 30.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 4 1 e5 2 4 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Disinfection by-product dynamics in a chlorinated, indoor swimming pool under conditions of heavy use: National swimming competition ShihChi Weng, Ernest R. Blatchley III* School of Civil Engineering and Division of Environmental & Ecological Engineering, Purdue University, 550 Stadium Mall Drive, West Lafayette, IN 47907-2051, USA

article info

abstract

Article history:

Anecdotal evidence suggests that water quality in chlorinated, indoor pools deteriorates

Received 23 March 2011

under conditions of heavy use. However, data to define these dynamics have not been

Received in revised form

reported. To address this issue, a study was performed in which water chemistry was

3 June 2011

monitored in a chlorinated, indoor pool before and during a national swimming compe-

Accepted 24 July 2011

tition, a period of heavy, intense use. NCl3 concentration was observed to double after the

Available online 9 August 2011

first day, and increased by a factor of 3e4 over the 4 days of competition. CNCHCl2 and CH3NCl2 concentrations both increased by a factor of 2e3 during the course of the meet,

Keywords:

while CHCl3 concentration showed only a modest increase during this same period.

Chlorine

Diurnal patterns of NCl3, CH3NCl2 and CHCl3 concentrations were observed, and these

DBPs

patterns appeared to depend on the Henry’s law constant of the compound.

Swimming

Urea concentration showed a diurnal pattern, superimposed on a trend of steady increase during each day of the competition; however, the diurnal pattern of urea behavior could not be explained by reactions with chlorine, as the urea-free chlorine reaction is relatively slow. It is more likely that the overnight decrease in urea concentration was attributable to mixing of surface water with water in the deeper parts of the pool. The findings of this study provide an indication of the changes in pool water chemistry that take place in a chlorinated, indoor pool under heavy use conditions. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Swimming is widely recognized as an activity that contributes to good health and competitive swimmers are generally viewed as being healthy people. However, swimmers and swimming activities adversely affect water and air quality in chlorinated, indoor pools, thereby introducing health risks to swimmers and pool employees (Weng et al., 2011). Chlorination of pool water is commonly practiced for control of microbial pathogens and oxidation of reduced compounds in

swimming pools. Chlorine also reacts with compounds “left behind” by swimmers (primarily the constituents of human urine and sweat) to yield a large number of disinfection byproducts (DBPs). Richardson et al. (2010) identified more than 100 DBPs in pool water samples. Two well-known DBP groups in chlorinated pool water are trihalomethanes (THMs) and chloramines. All four THMs (chloroform, bromodichloromethane, dibromochloromethane, and bromoform) have been shown to cause cancer in laboratory animals (Richardson et al., 2007), and they are regulated in the

* Corresponding author. Tel.: þ1 765 494 0316; fax: þ1 765 494 0395. E-mail addresses: [email protected] (ShihChi Weng), [email protected] (E.R. Blatchley). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.027

5242

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 4 1 e5 2 4 8

United States at a maximum contaminant level (MCL) of 80 mg/L (total THMs) in drinking water (U.S. EPA, 2006). THMs are common DBPs in pools. Weaver et al. (2009) observed TTHM concentrations in pools that ranged from (approximately) 4e400 mg/L. Lindstrom et al. (1997) estimated 80% of THM uptake for swimmers is via dermal absorption. Collectively, these factors indicate that THM exposure in pools can be more substantial than from drinking water. On the other hand, trichloramine (NCl3), which is often identified as being responsible for the “chlorine odor” of swimming pools, has been identified as a byproduct of chlorination of many N-organics that are common to swimming pools, including: urea, creatinine, arginine, and histidine (Li and Blatchley, 2007; Blatchley and Cheng, 2010). Although there is uncertainty regarding the role of NCl3 exposure in pools relative to childhood asthma (Weisel et al., 2009; Uyan et al., 2009), literature evidence indicates that NCl3 is an irritant of several human tissues. Gagnaire et al. (1994) demonstrated NCl3 to be a respiratory irritant to mice, and more recent studies have indicated NCl3 to contribute to acute ocular and respiratory irritation symptoms in lifeguards, swimming pool workers, and competitive swimmers (Jacobs et al., 2007; Dang et al., 2010; Clearie et al., 2010). Previous retrospective studies have shown positive correlations between irritation symptoms among swimmers and patrons and high gas-phase NCl3 concentration at indoor pool facilities (Kaydos-Daniels et al., 2008; Bowen et al., 2007). Among the DBPs that are formed in pools, eleven volatile compounds have been identified that could adversely affect air quality in chlorinated, indoor pool facilities: monochloramine (NH2Cl), dichloramine (NHCl2), trichloramine (NCl3), cyanogen chloride (CNCl), cyanogen bromide (CNBr), chloroform (CHCl3), bromodichloromethane (CHBrCl2), dibromochloromethane (CHBr2Cl), bromoform (CHBr3), dichloromethylamine (CH3NCl2), and dichloroacetonitrile (CNCHCl2) (Li and Blatchley, 2007; Weaver et al., 2009). These compounds have the ability to adversely affect water and air quality, particularly within chlorinated, indoor swimming pools. A recent study suggested that asthma is more likely to occur among elite swimmers than among other high-level athletes (Goodman and Hays, 2008). Moreover, anecdotal evidence suggests that water and air quality in a chlorinated, indoor pool both deteriorate under conditions of heavy use. Zwiener et al. (2007) demonstrated a positive relationship between THM concentrations in water and bather loading within several swimming pool facilities. However, relatively little information is available in the literature to characterize this behavior in terms of other DBPs or their precursors. To address this issue, an experiment was conducted to define changes in pool water chemistry that took place during a period of heavy use at a natatorium.

2.

held at a natatorium facility that houses three pools: a competition pool (50 m long  22.86 m wide  1.22e3.66 m deep, 2790 m3); a diving well (19.20 m long  22.86 m wide  4.42 m deep, 1940 m3); and a spa (2.44 m long  3.66 m wide  0.91 m deep, 8.1 m3). The three pools are hydraulically independent, and have separate treatment/recirculation systems; however, the pools share a common air space. The target pool (competition pool) was operated to accomplish 4 turnovers per day, and the recirculation system for the pool includes a balance tank, filtration, and chlorination. The chlorine concentration of the pools was controlled by an automated chemical feed system. During the meet, the competition pool was configured into two zones: a warm-up zone and a competition zone (Fig. 1); these two zones were separated by a moveable bulkhead. Because the vast majority of swimming activity took place in the warm-up zone, samples were collected from there. The sample and event schedule is shown in Table 1. Prior to the competition (March 11e12), samples were collected under conditions of normal use, where the typical bather load was 10e20 swimmers at any given time. The precompetition sampling data were used to represent baseline conditions of water quality. After teams arrived (March 16), swimmer activity in the pool increased markedly; at times, as many as 200 swimmers were in the pool for practice or competition. In addition, the nature of the swimming activity during this period was generally more intense than the activity of recreational/lap swimmers that usually use this facility. Competition started on March 18, and most swimmers remained at the facility until the conclusion of competition on March 20. Measurements of water chemistry were monitored 3 or 4 times per day at 6:30 AM, 8:30 AM, 2:00 PM, and 9:30 PM. These sample times were selected to catch the impacts of intense use and to characterize the diurnal cycle of water quality. Grab samples were collected from the warm-up zone, and were immediately transported (about 10 min) to the laboratory where they were analyzed for volatile DBPs by membrane introduction mass spectrometry (MIMS, Weaver et al., 2009), for urea by a colorimetric digestion method (Prescott and Jones, 1969), and for residual chlorine and inorganic chloramines by MIMS and the DPD/KI Colorimetric Method (APHAAWWA-WEF, 1998).

Methods

Water samples were collected without headspace from roughly 30 cm below the water surface of a chlorinated, indoor swimming pool facility before and during a national swimming competition in March 2010. The competition was

Fig. 1 e Scheme of swimming pool configuration and sample collection location.

5243

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 4 1 e5 2 4 8

Table 1 e Sampling and event schedule during experiment period. Time

Date 3/11 (Thurs)

06:30 06:30e08:30 08:30 08:30e11:00 11:30e13:30 14:00 14:00e19:00 19:00e21:30 21:30 Note

3.

3/12 (Fri)

3/13e3/15

3/16 (Tue)

3/17 (Wed)

3/18 (Thurs)

Free practice Sample Free practice

Sample Free practice Sample Free practice

Sample Free practice

Sample Free practice

Sample Sample Warm up Warm up Sample Sample Warm up Warm up Competition Competition Sample Sample Warm up Warm up Competition Competition Sample Sample Competition Days

Diving Competition Sample

Sample

Sample

Sample

Sample Sample Regular use, open for public

Sample Sample Natatorium closed to public; available to for competitors only

Results and discussion

3.1. Dynamic behavior of volatile DBPs in indoor, chlorinated swimming pools Several processes are available that have the potential to influence the concentrations of chemicals in pool water samples collected from a near-surface location, including:     

Reactions that produce constituents, Reactions that consume constituents, Liquid-gas transfer, Mixing, Uptake by swimmers.

The results of previous experiments involving organic-N precursor chemicals that are common in human sweat and urine have demonstrated that some of these compounds (e.g., amino acids and creatinine) react rapidly with free chlorine (Li and Blatchley, 2007). These same reactions lead to formation of DBPs, including the volatile compounds that were the target of MIMS measurements. Among the DBPs that are formed in pools, some are chemically stable (e.g. THMs, CNCHCl2, NO-3), while others may undergo additional reactions with chlorine, or other compounds, and therefore are unstable intermediates (e.g., CNCl, inorganic chloramines). The MIMS configuration used in this research is selective for volatile chemicals. Specifically, the membrane interface is used to allow only those chemicals that are sufficiently volatile to pervaporate through the membrane to be introduced to the mass spectrometer. The ability to detect compounds in liquid water by MIMS is determined by the product of aqueous-phase concentration and Henry’s law constant (i.e., escaping potential). Previous experiments have demonstrated that 11 volatile DBPs are consistently found in chlorinated pool water samples (Weaver et al., 2009); however, the Henry’s law constants for these compounds vary by more than two orders of magnitude. Volatilization is a relevant process for only those DBPs with the highest Henry’s law constants. Moreover, since volatilization will take place through an air:water interface, it is reasonable to expect that this process will lead to depletion of volatile chemicals from near-surface water.

3/19 (Fri)

3/20 (Sat) Warm up Sample Warm up Competition Sample Warm up Competition Sample

Mixing processes will influence the distribution of chemicals in any body of water, including swimming pools. In a general sense, mixing within a pool may be attributed to recirculation of water from the pool through the treatment system, as well as mechanical mixing induced by swimmers themselves. Given the wide range of pool geometries and uses, it is difficult to generalize further. The pool that was the target of this research is a tank with a rectangular plan and a sloped bottom (1.22e3.66 m depth). Water leaves the pool by gentle overflow in to a gutter system that spans the entire perimeter of the pool. This water is then pumped though a surge tank and a treatment system. The treated water is re-introduced to the pool through a series of 56 10-cm diameter circular diffusers spaced in a uniform, staggered pattern across the bottom of the pool. This pattern of circulation within the pool is intended to yield gentle current within the pool that is essentially vertical (bottom to top). The pool is used almost exclusively for lap swimming. During the course of this investigation, pool users were actively swimming while in the water. As such virtually all mechanical mixing and precursor introduction would have taken place in the near-surface area.

3.2.

Chlorine and chloramines

The competition brought intense use to the natatorium; bather loads were typically 150e200 swimmers during warmup sessions, which is roughly 5e10 times the bather load that is normally present at this facility. A corresponding increase in chlorine demand was observed when the competition began. In order to maintain acceptable free chlorine concentration (NSPF (2006) have suggested a minimum chlorine concentration of 1 mg/L as Cl2), the operator of the swimming pool doubled the free chlorine application rate (in the form of calcium hypochlorite), relative to normal operating conditions. Figure S1 illustrates the mass of Ca(OCl)2 added each day during the competition, as well as the free chlorine concentration measured at the chlorine controller by continuous titration. Water losses from the pool by splashing and evaporation required replacement. Roughly 40,000 L (1.4% of total volume)

5244

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 4 1 e5 2 4 8

of water was added to the pool during the competition to maintain water level. Fig. 2a illustrates the time-course behavior of residual chlorine measured by the DPD/KI colorimetric method. Free chlorine dropped rapidly after the competitors arrived, presumably because of introduction of reduced compounds that reacted with free chlorine. Note that the free chlorine concentration measured at the controller (Figure S1) was substantially higher than the corresponding free chlorine concentration measured in the pool water samples (Fig. 2a). It is likely that the difference in free chlorine concentrations between these two locations was attributable to chlorine demand that was expressed in the pool. Measurements of the concentration of chloramines by the DPD/KI method did not change substantially over the sampling period. However, previous research has demonstrated that the DPD/KI method is susceptible to interference (Harp, 2002). Organic chloramines are believed to be largely responsible for this interference in pools (Weaver et al., 2009). In contrast, MIMS measurements demonstrated a rapid

increase in NCl3, and a smaller increase in NHCl2. DPD/KI measurements were consistently in excess of the MIMS measurements, presumably because of the presence of organic chloramines that interfere with the colorimetric method (Harp, 2002). MIMS has been demonstrated to provide an accurate measurement of inorganic chloramines (Shang and Blatchley, 1999). The MIMS data indicated that NCl3 concentration increased by a factor of 3e4 over the course of the event. Aqueous NCl3 concentration displayed a consistent diurnal trend during the competition: decrease in the middle of the day and increase overnight. Three major mechanisms are believed to have influenced this diurnal pattern: loss by volatilization, formation by reactions between free chlorine and organic-N compounds, and mixing within the pool. The mid-day decrease was believed to be attributable to intense physical activity of the swimmers, which typically takes place at or near the surface of the pool. This activity has been demonstrated to promote liquid-gas transfer of volatile chemicals such as NCl3 (Weng et al., 2011). This “loss” of NCl3 would be expected to be most important for water located near the free surface, where all water samples were collected in this work. During the overnight hours, mechanical mixing of near-surface water would be substantially reduced, thereby reducing liquid-gas transfer. However, the general circulation pattern in the pool would have allowed for near-surface water to be removed, subjected to treatment (filtration and chlorination), and re-introduced to the pool through the diffusers located in the bottom of the pool. The 9 h (overnight) period between sampling events would also have allowed for DBP formation by reactions between chlorine and precursors.

3.3.

Fig. 2 e Dynamics of free chlorine and inorganic chloramine concentration measured by (a) DPD/KI method, and (b) MIMS. The vertical, dashed lines represent midnight of each day.

Volatile disinfection by-products

Volatile DBPs were analyzed by MIMS; Fig. 3 illustrates the results of these measurements. CHCl3 concentration trended upward during the competition and showed a consistent diurnal pattern: decrease in the daytime and increase overnight (see Fig. 3a). A similar diurnal pattern of CHCl3 concentration was reported by Kristensen et al. (2010). The loss during the day (while the pool was being used) is believed to be attributable to volatilization, which is known to be promoted by mixing associated with swimming activities (Weng et al., 2011). Losses due to volatilization would take place from near-surface locations, where samples were collected, and where competitive swimmers spend most of their time in the pool. As with NCl3, the overnight increase is believed to be attributable to reactions that yield CHCl3. A positive relationship between THM concentration and bather loading was proposed by Zwiener et al. (2007), suggesting that bathers may introduce precursors that react with chlorine to yield THMs. The diurnal pattern for CHBr2Cl was less distinct than the pattern for CHCl3; CHBr2Cl concentration generally decreased during the event. The overall decrease of CHBr2Cl concentration suggests that the loss rate (e.g., volatilization) was higher than its formation rate. The concentrations of the other THMs were consistently low; in the case of CHBrCl2 and CHBr3, concentrations were below the limit of detection for the MIMS instrument used in these experiments.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 4 1 e5 2 4 8

CNCHCl2 and CH3NCl2 concentrations both increased by a factor of 2e3 during the course of the meet. Li and Blatchley (2007) identified two constituents in human urine and sweat, creatinine and L-histidine, that function as precursors to formation of CH3NCl2 and CNCHCl2, respectively, by chlorination. The data in Fig. 3b indicate a steady increase in the concentration of CNCHCl2, with no diurnal variations. On the other hand, time-course CH3NCl2 concentration data showed a pattern that was similar to the behavior of CHCl3. Specifically, the CH3NCl2 concentration data were characterized by a diurnal pattern with lowest daily concentrations being observed near mid-day and increases in concentration overnight, superimposed on a steady increase of concentration. No consistent trend for CNCl concentration was observed during the experiment period. Free chlorine is required for CNCl formation; however, Na and Olson (2004) demonstrated that CNCl is oxidized by an OCl--catalyzed reaction. Therefore, the concentration of CNCl in pools will be strongly influenced by maintenance practices that influence residual (free) chlorine concentration in pools.

Fig. 3 e The concentration of volatile DBPs sampling during sampling period, as analyzed by MIMS: (a) THMs (CHCl2Br and CHBr3 are not shown because their concentrations were below the detection limits) (b) CNCl, CNCHCL2 and CH3NCl2. The vertical, dashed lines represent midnight of each day.

3.4.

5245

Urea

By mass, urea is likely to represent the greatest source of organic-N introduced to pools by swimmers. Urea is introduced to pools by swimmers via at least three sources: urine, sweat, and extraction of “natural moisturizing factor”, a group of chemicals that is produced in skin tissues to maintain skin moisture (Gunkel and Jessen, 1986; Erdinger et al., 1997; ISRM, 2009). De Laat et al. (2011) reported urea concentrations in swimming pools ranging from 0.12 mg/L to 3.6 mg/L. Urea concentrations measured in this study were at the low end of this range. Fig. 4 illustrates the time-course pattern of urea concentration in the pool. During the competition, urea concentration in near-surface water showed a pattern of steady increase during each day of the competition, with a decrease at night. Superimposed on this diurnal pattern was an overall trend of increasing concentration during the entire event. This diurnal pattern of urea behavior cannot be explained by reactions with chlorine, as the urea-free chlorine reaction is slow relative to the time-scale of sampling in this study (Blatchley and Cheng, 2010; De Laat et al., 2011). Given the lack of volatility of urea, it is more likely that the decrease in urea concentration for water samples collected at the surface was attributable to overnight mixing of surface water with water in the deeper parts of the pool. Although urea is slow to react with free chlorine, it is an effective precursor to NCl3 formation (Samples, 1959; Blatchley and Cheng, 2010). Among the inorganic chloramines, NCl3 concentration doubled after the first day, and increased by a factor of 3e4 over the 4 days of competition, as shown in Fig. 2. The data illustrated in Fig. 4 indicate that a large mass of urea was introduced by competitors. In turn, this urea contributed to NCl3 formation during the competition. The data in Fig. 4 were used to estimate urea contributions per swimmer, based on the following assumptions:

Fig. 4 e Time-course behavior of urea concentration during the sampling period. The vertical, dashed lines represent midnight of each day.

5246

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 4 1 e5 2 4 8

Table 2 e Estimated contribution of urea from sweat and urine per swimmer. Date

Daily difference (mg/L)a

Total urea introduction (g)b

Urea introduced per swimmer (g/person)c

Sweat introduced per swimmer (mL/swimmer)d

Urine introduced per swimmer (mL/swimmer)

117 72.4 155 120 99.2

138 83.9 180 139 115

0.92 0.56 1.20 0.93 0.77

1350 823 1760 1360 1130

89.6 54.7 117 90.5 74.9

3/16 3/17 3/18 3/19 3/20 a b c d

[urea concentration in the evening]-[urea concentration in the morning]. Assume volume of top layer is half of the total volume in the pool (1.16  106 L) and was well-mixed. Assume there were 150 swimmers per day. 0.68 g urea/L of sweat and 10.245 g urea/L of urine(WHO, 2006).

 The pool was stratified into two layers during the daytime with equal depth where the top layer was well-mixed (because of the disturbance of swimmers).  The urea measurements were representative of the top layer.  No exchange (diffusion or recirculation effects) took place between the two layers while swimmers were in the pool.  Consumption of urea by reaction with chlorine and uptake by swimmers was negligible.  Bather loading was 150 persons each day during the competition. The daily urea concentration increases ranged from 72.4 mg/L to 155 mg/L during the five-day period from March 16e20, as shown in Table 2. Based on the pool dimensions, the water volume in the pool was 2.33  106 L, and the water volume of the top layer was equal to half of the total volume. Based on these assumptions, the total mass of urea introduced by swimmers was 83.9e180 g (1.16  106 L  daily difference (mg/L)), which implies daily urea introduction ranged from 0.56 to 1.2 g/swimmer/day (total urea introduction (g)/150 swimmers/day). Reported values of urea concentration in human sweat and urine are 0.68 g/L and 10.24 g/L, respectively (WHO, 2006). If all urea introduced to the pool is assumed to have originated from sweat, this implies a daily sweat introduction rate that ranged from 823 to 1760 mL/swimmer (urea introduced per person (g)/0.68 (g/L)). Similarly, if all urea is assumed to have

been introduced from urine, this implies a urine introduction rate that ranged from 54.7 to 117 mL/swimmer (urea introduced per person (g)/10.24 (g/L)). Previous studies have presented estimates of sweat introduction from 200 to 1000 mL/ person; similarly, daily urine introduction estimates have ranged from 25 to 80 mL/person (Gunkel and Jessen, 1986; Erdinger et al., 1997; Seux et al., 1985). Although these calculations were based on several unvalidated assumptions, the results are consistent with previously published estimates of sweat and urine introduction rates. More generally, the data presented in Fig. 4 and Table 2 indicate that heavy use of a pool (such as a large competition) will result in substantial introduction of pollutants to a pool, and corresponding adverse effects on water quality.

2.5.

Diurnal pattern of volatile DBPs and DBP precursors

Among DBPs, diurnal patterns of NCl3, CH3NCl2, and CHCl3 were evident, with concentration decreasing during the day, and increasing overnight, presumably due to the combined effects of swimmer-induced volatilization, DBP production, and mixing within the pool. As indicated by the Henry’s law constants in Table 3, these chemicals are the most volatile DBPs detected in this study. This suggests that liquid-gas transfer contributes to diurnal behavior of these compounds in near-surface water. CNCl has the next highest Henry’s law constant among those listed in Table 3. CNCl is a common byproduct from chlorination of amino acids (Lee et al., 2006).

Table 3 e Henry’s Law constants and liquid-phase behavior among volatile DBPs and urea. Compound NH2Cl NHCl2 NCl3 CHCl3 CHBr2Cl CHBr3b CNCl CNCHCl2 CH3NCl2 Urea

Henry’s Law Constant (atm, 20  C)a 0.45 1.52 432 185 57.3 21.5 108 0.21 154 e

Diurnal Pattern

Overall Concentration Changes

No No Yes Yes Yes e No No Yes Yes

e Slight Increase Increase Increase Decrease e e e Increase Increase

a Holzwarth et al. (1984); Krasner and Wright (2005); Cimetiere and De Laat (2009). b Concentration of CHBr3 was close to detection limit.

Description e e Volatilization and formation Volatilization and formation Volatilization e Controlled by free chlorine Steady increase Volatilization and formation Mixing behavior between top and bottom layer

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 4 1 e5 2 4 8

CNCl is also known to be oxidized via an OCl--catalyzed process (Na and Olson, 2004). The pattern of CNCl behavior observed over period of these experiments was erratic, possibly due to the combination of processes that are known to influence its behavior, and the fact that measured values of CNCl concentration were consistently near the limit of detection. NH2Cl and NHCl2 generally increased during the study, but trends in their behavior were less obvious. CNCHCl2 increased steadily over the course of the competition. No pattern of diurnal behavior was evident for this compound, probably because of its relatively low volatility and stability in the presence of free chlorine. The diurnal pattern of behavior for urea was essentially a mirror image of behavior observed with the volatile DBPs (CHCl3, NCl3, and CH3NCl2). Urea is slow to react with chlorine and is essentially non-volatile. The increases in urea concentration in near-surface samples that were observed during each day of competition were consistent with a pattern of steady urea input. The decrease that was observed overnight is believed to be attributable to mixing of near-surface water with water from other parts of the pool (i.e., deeper water), where urea input would be negligible, and daytime concentration was likely to be lower.

3.

Conclusions

This study provided evidence of the influence of swimmers on swimming pool water quality in terms of free chlorine concentration, volatile DBPs, and urea. NCl3 concentration increased by a factor of 3e4 during the competition; CHCl3, CNCHCl2 and CH3NCl2 concentrations increased during the competition, presumably due to introduction of precursor chemicals. Diurnal patterns of CHCl3, CH3NCl2, and NCl3 were observed which were attributed to the combined effects of volatilization enhanced by aggressive swimmer activity, reactions to produce these DBPs, and mixing. Urea concentration increased at the water surface during the period of heavy use, and decreased overnight. A diurnal pattern in nearsurface urea concentration was also observed. Water quality deteriorated during the course of the competition, as represented by increases in the concentrations of volatile DBPs, thereby leading to enhanced DBP exposure by swimmers, pool employees, and spectators at the competition.

Acknowledgments We are grateful for the cooperation and support of pool operators and facility managers involved in this study.

Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.07.027.

5247

references

APHA-AWWA-WEF, 1998. Standard Methods for the Examination of Water and Wastewater, twentieth ed. (Washington. DC). Blatchley III, E.R., Cheng, M., 2010. Reaction mechanism for chlorination of urea. Environmental Science & Technology 44 (22), 8529e8534. Bowen, A.B., Kile, J.C., Otto, C., Kazerouni, N., Austin, C., Blount, B. C., Wong, H.N., Beach, M.J., Fry, A.M., 2007. Outbreaks of shortincubation ocular and respiratory illness following exposure to indoor swimming pools. Environmental Health Perspectives 115 (2), 267e271. Cimetiere, N., De Laat, J., 2009. Henry’s law constant of N, Ndichloromethylamine: application to the contamination of the atmosphere of indoor swimming pools. Chemosphere 77 (4), 465e470. Clearie, K.L., Vaidyanathan, S., Williamson, P.A., Goudie, A., Short, P., Schembri, S., Lipworth, B.J., 2010. Effects of chlorine and exercise on the unified airway in adolescent elite Scottish swimmers. Allergy 65 (2), 269e273. Dang, B., Chen, L.L., Mueller, C., Dunn, K.H., Almaguer, D., Roberts, J.L., Otto, C.S., 2010. Ocular and respiratory symptoms among lifeguards at a hotel indoor waterpark resort. Journal of Occupational and Environmental Medicine 52 (2), 207e213. De Laat, J., Feng, W., Freyfer, D.A., Dossier-Berne, F., 2011. Concentration levels of urea in swimming pool water and reactivity of chlorine with urea. Water Research 45 (3), 1139e1146. Erdinger, L., Kirsch, F., Sonntag, H.G., 1997. Potassium as an indicator of anthropogenic contamination of swimming pool water. Zentralbl Hyg Umweltmed 200 (5e6), 297e308. Gagnaire, F., Azim, S., Bonnet, P., Hecht, G., Hery, M., 1994. Comparison of the sensory irritation response in mice to chlorine and nitrogen trichloride. Journal of Applied Toxicology 14 (6), 405e409. Goodman, M., Hays, S., 2008. Asthma and swimming: a metaanalysis. Journal of Asthma 45 (8), 639e647. Gunkel, K., Jessen, H.J., 1986. Investigations on introduction of urea into bating water. Acta gydrochimica et hydroblologica 14 (5), 451e461. Harp, D.L., 2002. Current Technology of Chlorine Analysis for Water and Wastewater. Retrieved from: http://www.hach. com/fmmimghach?/CODE:L70191473%7C1//true. Holzwarth, G., Balmer, R.G., Soni, L., 1984. Fate of chlorine and chloramines in cooling towers; Henry’s law constants for flashoff. Water Research 18 (11), 1421e1427. Institute of Sport and Recreation Management (ISRM), 2009. Managing Trichloramine in Indoor Pools. Loughborough University, Loughborough, UK. Information Notes Ref-349: 01/09. Jacobs, J.H., Spaan, S., Van Rooy, G.B.G.J., Meliefste, C., Zaat, V.A. C., Rooyackers, J.M., Heederik, D., 2007. Exposure to trichloramine and respiratory symptoms in indoor swimming pool workers. European Respiratory Journal 29 (4), 690e698. Kaydos-Daniels, S.C., Beach, M.J., Shwe, T., Magri, J., Bixler, D., 2008. Health effects associated with indoor swimming pools: a suspected toxic chloramine exposure. Public Health 122 (2), 195e200. Krasner, S.W., Wright, J.M., 2005. The effect of boiling water on disinfection by-product exposure. Water Research 39 (5), 855e864. Kristensen, G.H., Klausen, M.M., Hansen, V.A., Lauritsen, F.R., 2010. On-line monitoring of the dynamics of trihalomethane concentrations in a warm public swimming pool using an unsupervised membrane inlet mass spectrometry system with off-site real-time surveillance. Rapid Communications in Mass Spectrometry 24 (1), 30e34.

5248

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 4 1 e5 2 4 8

Lee, J.H., Na, C., Ramirez, R.L., Olson, T.M., 2006. Cyanogen chloride precursor analysis in chlorinated river water. Environmental Science & Technology 40 (5), 1478e1484. Li, J., Blatchley III, E.R., 2007. Volatile disinfection byproduct formation resulting from chlorination of organic-nitrogen precursors in swimming pools. Environmental Science & Technology 41 (19), 6732e6739. Lindstrom, A.B., Pleil, J.D., Berkoff, D.C., 1997. Alveolar breath sampling and analysis to assess trihalomethane exposures during competitive swimming training. Environmental Health Perspectives 105 (6), 636e642. Na, C., Olson, T.M., 2004. Stability of cyanogen chloride in the presence of free chlorine and monochloramine. Environmental Science & Technology 38 (22), 6037e6043. National Swimming Pool Foundation (NSPF), 2006. Certified PoolSpa Operator Handbook: National Swimming Pool Foundation, 2006 ed. Colorado Springs, CO. Prescott, L.M., Jones, M.E., 1969. Modified methods for the determination of carbamyl aspartate. Analytical Biochemistry 32 (3), 408e419. Richardson, S.D., DeMarini, D.M., Kogevinas, M., Fernandez, P., Marco, E., Lourencetti, C., Balleste, C., Heederik, D., Meliefste, K., McKague, A.B., Marcos, R., Font-Ribera, L., Grimalt, J.O., Villanueva, C.M., 2010. What’s in the pool? A comprehensive identification of disinfection by-products and assessment of mutagenicity of chlorinated and brominated swimming pool water. Environmental Health Perspectives 118 (11), 1523e1530. Richardson, S.D., Plewa, M.J., Wagner, E.D., Schoeny, R., DeMarini, D.M., 2007. Occurrence, genotoxicity, and carcinogenicity of regulated and emerging disinfection byproducts in drinking water: a review and roadmap for research. Mutation Research 636, 178e242. Samples, W. R., (1959) A study on the chlorination of urea. Ph.D. Dissertation, Harvard University, Cambridge, MA. Seux, R., Weicherding, J., Besse, P., Alouini, Z., Cle´ment, M., 1985. Etude qualitative et quantitative de la pollution apporte´e par

les baigneurs. In: Actes du colloque national Piscines et Sante´. ENSP Rennes, pp. 92e103 (17e19 June 1985). Shang, C., Blatchley III, E.R., 1999. Differentiation and quantification of free chlorine and inorganic chloramines in aqueous solution by MIMS. Environmental Science & Technology 33 (13), 2218e2223. U.S. Environmental Protection Agency, 2006. National primary drinking water regulations: stage 2 disinfectants and disinfection by-products rule. Federal Register 71, 387e493. Uyan, Z.S., Carraro, S., Piacentini, G., Baraldi, E., 2009. Swimming pool, respiratory health, and childhood asthma: should we change our beliefs? Pediatric Pulmonology 44 (1), 31e37. Weaver, W.A., Li, J., Wen, Y., Johnston, J., Blatchley, M.R., Blatchley III, E.R., 2009. Volatile disinfection by-product analysis from chlorinated indoor swimming pools. Water Research 43 (13), 3308e3318. Weisel, C.P., Richardson, S.D., Nemery, B., Aggazzotti, G., Baraldi, E., Blatchley III, E.R., Blount, B.C., Carlsen, K., Eggleston, P., Frimmel, F.H., Goodman, M., Gordon, G., Grinshpun, S.A., Heederik, D., Kogevinas, M., LaKind, J.S., Nieuwenhuijsen, M.J., Piper, F.C., Sattar, S.A., 2009. Childhood asthma and environmental exposures at swimming pools: state of the science and research recommendations. Environmental Health Perspectives 117 (4), 500e507. Weng, S.C., Weaver, W.A., Afifi, M.Z., Blatchley, T.N., Cramer, J.S., Chen, J., Blatchley III, E.R., 2011. Dynamics of gas-phase trichloramine (NCl3) in chlorinated, indoor swimming pool facilities. Indoor Air. doi:10.1111/j.1600-0668.2011.00710.x. World Health Organization (WHO), 2006. Guidelines for Safe Recreational Water Use, vol. 2. Swimming Pools and Similar Environments, Geneva. Retrieved on November 24th, 2010, from: http://www.who.int/water_sanitation_health/bathing/ srwe2chap4.pdf. Zwiener, C., Richardson, S.D., De Marini, D.M., Grummt, T., Glauner, T., Frimmel, F.H., 2007. Drowning in disinfection byproducts? Assessing swimming pool water. Environmental Science & Technology 41 (2), 363e372.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 4 9 e5 2 5 6

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Methanogen community structure-activity relationship and bioaugmentation of overloaded anaerobic digesters V.P. Tale a, J.S. Maki b, C.A. Struble c, D.H. Zitomer a,* a

Department of Civil and Environmental Engineering, Marquette University, P.O. Box 1881, Milwaukee, WI 53201, United States Department of Biological Sciences, Marquette University, Milwaukee, WI 53201, United States c Department of Mathematics, Statistics and Computer Science, Marquette University, Milwaukee, WI 53201, United States b

article info

abstract

Article history:

Accumulation of acids in anaerobic digesters after organic overload can inhibit or stop CH4

Received 26 March 2011

production. Therefore, methods to reduce acid concentrations would be helpful. One

Received in revised form

potential method to improve recovery involves bioaugmentation, addition of specific

7 July 2011

microorganisms to improve performance. In this study, transiently overloaded digesters

Accepted 25 July 2011

were bioaugmented with a propionate-degrading enrichment culture in an effort to

Available online 31 July 2011

decrease recovery time. Biomass samples from 14 different, full-scale anaerobic digesters were screened for specific methanogenic activity (SMA) against propionate; the microbial

Keywords:

communities were also compared. SMA values spanned two orders of magnitude. Principal

Anaerobic digestion

component analysis of denaturing gradient gel electrophoresis (DGGE) banding patterns for

Bioaugmentation

a functional gene (mcrA) suggested an underlying community structure-activity relation-

Microbial community analysis

ship; the presence of hydrogenotrophic methanogens closely related to Methanospirillum

Organic overload

hungatei and Methanobacterium beijingense was associated with high propionate SMA values.

Propionate

The biomass sample demonstrating the highest SMA was enriched for propionate

Specific methanogenic activity

degrading activity and then used to bioaugment overloaded digesters. Bioaugmented digesters recovered more rapidly following the organic overload, requiring approximately 25 days (2.5 solids retention times (SRTs)) less to recover compared to non-bioaugmented digesters. Benefits of bioaugmentation continued for more than 12 SRTs after organic overload. Bioaugmentation is a promising approach to decrease recovery time after organic overload. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

In anaerobic digestion, the bioconversion of fatty acids, such as propionate, to acetate and H2 is not thermodynamically spontaneous under standard conditions. H2-consuming reactions, such as CH4 production, are required to reduce the H2 concentration and drive the bioconversion of propionate in the forward direction. In this regard, degradation of propionate stops when the H2 concentration is above 10
4 atm; higher H2

concentrations can result in increased concentrations of propionic acid and other carboxylic acids in the digester (McCarty and Smith, 1986). Increased acid concentration can cause the pH to decrease and inhibit or stop CH4 production. Propionate accumulation is an indicator of anaerobic digester organic overload or process imbalance and propionate-utilizing microbial consortia play an important role when anaerobic digesters are subjected to organic overload. Smith and McCarty (1990) studied the effect of substrate

* Corresponding author. Tel.: þ1 414 288 5733; fax: þ1 414 288 6149. E-mail addresses: [email protected] (V.P. Tale), [email protected] (J.S. Maki), [email protected] (C.A. Struble), [email protected] (D.H. Zitomer). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.035

5250

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 4 9 e5 2 5 6

overloading in a continuous-stirred tank reactor (CSTR) fed propionate and ethanol. When the reactor was overloaded with an increased pulse dose of ethanol, the effluent propionate concentration remained high for over 18 days (3.7 hydraulic residence times (HRTs)), whereas the ethanol concentration decreased to a low value after 4 days. Therefore, propionate concentrations can remain chronically elevated for a significant time after a process overload. Strategies to reduce propionate and soluble chemical oxygen demand (SCOD) concentrations in organically overloaded digesters would be helpful to consistently meet effluent SCOD requirements of full-scale applications. One possible strategy to reduce propionate and SCOD concentrations is bioaugmentation, defined as the addition of specialized microorganisms to biological systems to improve process performance. Bioaugmentation has typically been considered to remediate hazardous waste sites and improve aerobic bioprocesses, such as nitrification (Rittmann and Whiteman, 1994). It has also been studied in anaerobic systems to degrade specific organics (Tawfiki Hajji et al., 2000; Guiot et al., 2002; Lenz et al., 2009), lipid-rich wastes, cellulose, and cellulosic material present in manure (Nielsen et al., 2007) and to reduce the recovery time of digesters exposed to toxicants (Schauer-Gimenez et al., 2010). In this study, we screened biomass samples from various full-scale anaerobic reactors by measuring maximum propionate utilization rates. The most rapid propionate-utilizing biomass was enriched and used to bioaugment organically overloaded digesters in an effort to decrease the recovery time, decrease effluent SCOD and increase methane production.

2.

Methodology

2.1. SMA tests of anaerobic cultures against calcium propionate Specific methanogenic activity (SMA) values of biomass samples from full-scale anaerobic reactors were determined using propionate as the substrate by a standard protocol (Sorensen and Ahring, 1993). All biomass samples were diluted to <2 g volatile suspended solids (VSS)/L using nutrient medium containing calcium propionate (3.8 g/L), with 50 mL of diluted biomass placed in 160-mL serum bottles and sparged with oxygen-free gas (3:7 v/v CO2:N2). Serum bottles were maintained in an incubator/shaker at 35  C and 150 rpm. Granular biomass was disrupted in a glass beaker using light pressure from a latex-gloved finger before testing. Gas generation was monitored for 30 days using a glass syringe with a wetted glass plunger, and CH4 concentration was measured as described below. The maximum CH4 production rate was determined by linear regression using the initial points on a graph of cumulative CH4 production volume versus time.

2.2.

Methanogenic community analysis

Methyl coenzyme-M reductase (MCR) is the terminal enzyme complex in the biological methane generation pathway (Woese and Fox, 1977). This enzyme complex is thought to be

unique to and ubiquitous in methanogens (Thauer, 1998), making it a tool for the detection and identification of methanogens. The MCR operon exists in two forms, MCRI and MCRII. One peptide of the MCRI complex that encodes for the mcrA gene has been selected as a suitable descriptor in the phylogenetic analysis of methanogenic communities (Luton et al., 2002). Studies have highlighted the use of the mcrA gene as a target for the detection of methanogens in a wide range of environments including rice paddies (Lueders et al., 2001), peat bogs (Hales et al., 1996; Lloyd et al., 1998; Nercessian et al., 1999; Juottonen et al., 2006), termite gut (Ohkuma et al., 1995), anaerobic digesters (Rastogi et al., 2008), polluted water (Ufnar et al., 2007), hypereutrophic lakes (Earl et al., 2003), hydrothermal sediments (Dhillon et al., 2005), subsurfaces of tidal flats (Wilms et al., 2007) and marine environments (Bidle et al., 1999). Methanogenic communities in biomass samples were characterized using primers that amplify the mcrA gene (Luton et al., 2002). DNA was extracted using the PowerSoil DNA Isolation Sample Kit (MoBio Laboratories, Inc., Carlsbad, CA). DNA was amplified by conducting PCR for GCmcrA1f (50 GCclamp-GGTGGTGTMGGATTCACACARTAYGCWACAGC-30 ) and mcrA500r (50 e TTCATTGCRTAGTTWGGRTAGTT e 30 ) primers (Luton et al., 2002). These primers were expected to generate 460- to 490-bp-long amplified products (Luton et al., 2002). The PCR conditions were as follows: initial denaturation at 95  C (5 min), 35 cycles of 95  C (1 min), 58  C (1 min), and 72  C (3 min), and a final extension of 10 min at 72  C. The program included a slow ramp in temperature (0.1  C s
1) between the annealing and extension steps of the first 5 cycles of the protocol to assist in the initial formation of product due to the degenerate nature of the primers, as recommended (Luton et al., 2002). The size of the expected PCR products was confirmed using a 1% w/v Tris-acetate-EDTA buffer (Sambrook and Russell, 2001) agarose gel and a l(Hind III digest) fX174 (Hae III digest) DNA ladder stained with ethidium bromide (0.01%, v/v). Gels were visualized using a transilluminator (Model M-20, UVP, Upland, CA). The amplified products were separated using denaturing gradient gel electrophoresis (DGGE). For a general review of DGGE, refer to Muyzer (1999). The DGGE technique has been extensively used in the field of microbial ecology to compare microbial communities and use of DGGE with mcrA as a target gene have been reported (Wilms et al., 2007; Galand et al., 2002). A 8% polyacrylamide gel with a linear gradient (40e70% v/v denaturant concentration from top to bottom of gel) was used. Forty mL of the amplified DNA product (equivalent to approximately 75 ng of DNA) was added to each lane of the polyacrylamide gel. Amplified products separated on the gel were stained using SYBR Green (Invitrogen, CA USA) dye and the gel was visualized and photographed using an ultra violet transilluminator as described above. The gel image was analyzed using Lab Works software (v. 4.6.00.0). Densitometric data obtained from the DGGE image were used to perform principal component analysis (PCA) using the MATLAB (v.7.6 (R2008a)) software package. Clustering of biomass samples on the plot of the first two principal components was performed using the farthest neighbor algorithm. Two DGGE bands having the most significant effect on the clustering pattern were identified by PCA analysis and

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 4 9 e5 2 5 6

excised from the DGGE gel. The excised bands were eluted from the gel and PCR amplified for GCmcrA1f and mcrA500r using protocol stated above. The obtained PCR product was cloned using the TOPO TA Cloning Kit according to the manufacturer’s instructions (Invitrogen, Carlsbad, CA). Transformants containing plasmids with amplified product were screened via blue/white selection (Sambrook and Russell, 2001). Twelve light-colored colonies were picked per DGGE band and PCR amplified with PucF (50 -GGA ATT GTG AGC GGA TAA CA- 30 ) and PucR (50 - GGC GAT TAA GTT GGG TAA CG - 30 ) primers. The PCR conditions for the PUC primers were as follows: denaturing temperature of 94  C (1 min), annealing temperature of 55  C (1 min), elongation temperature of 72  C (1 min), and a final extension of 10 min at 72  C. The size of the PUC-amplified PCR products were confirmed using an agarose gel as described above. The amplified DNA was cleaned using the UltraClean PCR Clean-up Kit (MoBio Laboratories, Carlsbad, CA) according to the manufacturer’s protocol. The amplified products were sequenced at the University of Chicago Cancer Research Center DNA sequencing facility using a capillary automated DNA sequencer (Applied Biosystems 3730XL, Foster City, CA). Contiguous sequences were assembled for each clone using forward and reverse sequences. Vector segments from the contiguous sequences were removed using a tailor-made computer program which utilized the UniVec database of the National Center for Biotechnology Information (NCBI) (http://www.ncbi.nlm.nih.gov/VecScreen/UniVec.html) using the Basic Local Alignment Search Tool (BLAST) (Altschul et al., 1997) for cloning vectors. Clone sequences were submitted for BLASTn (Altschul et al., 1997) query on the NCBI database (http://www.ncbi.nlm.nih.gov).

2.3. Enrichment of most rapid propionate-utilizing culture The biomass sample demonstrating the highest SMA value was enriched for propionate utilization. Enrichment was accomplished in a 750-mL serum bottle containing 150 mL of active volume and operated in daily fill-and-draw mode. The serum bottle was initially sparged with a N2:CO2 gas mixture (mixed in 7:3 ratio v/v), then fed 0.17 g propionate/L-day (0.25 g COD/ L-day) with nutrient medium (see below), and shaken continuously at 150 rpm and 35  2  C at a 15-day SRT by removing 10 mL of culture and replacing it with 10 mL of feed once every day. The SMA of the enrichment culture was measured at the start of enrichment and again after 580 days (38.6 HRTs).

2.4.

Bioaugmented digesters

The effectiveness of bioaugmentation was evaluated by organically overloading small-scale, non-fat-dry-milk-fed anaerobic digesters; one digester set was bioaugmented, whereas another set was not. In addition, some digesters were neither overloaded, nor bioaugmented. Each digester was a 160-mL serum bottle containing 50 mL of active volume maintained in a shaker incubator at 35  2  C and 150 rpm. Digesters were seeded with biomass from a laboratory-scale digester fed non-fat dry milk. Initial VSS concentration of the seed biomass in the digester was 4.4  0.48 g/L. Digesters

5251

were initially sparged with a N2:CO2 gas mixture (mixed in 7:3 ratio v/v) and then operated at a 10-day SRT by daily wasting and feeding. Digesters received nutrient medium and non-fat dry milk (2.7 g COD/L-day). A shock overload of non-fat dry milk (32 g COD/L digester volume) was given for one day to all the digesters except the non-overloaded set. This shock load dose was identified during preliminary studies using 8, 16 and 32 g COD/L doses (data not shown); digesters receiving the lower doses recovered within one week. Therefore, the highest dose was selected for subsequent testing to produce a longer-term upset condition. Following the organic overload, bioaugmented digesters were provided with 1.7 mL/day (70 mgVSS/L-day) of the enrichment culture whereas nonbioaugmented digesters received 1.7 mL/day of an autoclaved, abiotic version of the enrichment culture substituted for 1.7 mL of nutrient medium.

2.5.

Nutrient medium

The nutrient medium, as suggested by Speece (2008), contained the following [mg/L]: NH4Cl [400]; MgSO46H2O [250]; KCl [400]; CaCl22H2O [120]; (NH4)2HPO4 [80]; FeCl36H2O [55]; CoCl26H2O [10]; KI [10]; the trace metal salts MnCl24H2O, CuCl22H2O, Zn(C2H3O2)22H2O, AlCl36H2O, NH4VO3, NaMoO42H2O, H3BO3, NiCl26H2O, NaWO42H2O, and Na2SeO3) [each at 0.5]; NaHCO3 [5000]; and resazurin [1].

2.6.

Analytical methods

SCOD was measured by filtering the sample through a 0.45-mm filter and measuring the filtrate COD concentration by standard methods (APHA et al., 1998). The pH was measured using a bench-top pH meter and a general-purpose pH electrode. The volume of biogas produced was measured using a water-lubricated glass syringe via the plunger displacement method. Biogas CH4 concentration and volatile fatty acids concentrations were determined by standard methods using gas chromatography (APHA et al., 1998). Volatile suspended solids (VSS) concentration was measured using standard methods (APHA et al., 1998). For statistical comparison between SMA values, Student’s t statistic for unequal population variances was used.

3.

Results

3.1.

SMA of biomass for various anaerobic reactors

Average SMA values for the biomass samples from full-scale anaerobic reactors varied over two orders of magnitude (Fig. 1). The SMA values were grouped using farthest neighbor algorithm and Student’s t statistic for unequal population variances. A probability threshold of 0.05 (95% confidence interval) clustered the biomass samples into six statistically distinct groups based on their SMA values (Fig. 1). All the brewery biomass samples were represented by higher SMA values, whereas, with the exception of the Municipal WWTP - 1 sample, all municipal wastewater biomass samples demonstrated lower SMA values. The average SMA values of brewery and municipal wastewater treatment plant biomass samples

5252

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 4 9 e5 2 5 6

Fig. 1 e SMA of various full-scale anaerobic reactors against propionate groups 1 to 6 represent biomass samples having statistically distinct SMA values. Error bars represent standard deviation among three replicates.

were 6.9  0.2 and 1.7  0.2 mL CH4/gVSS-hr, respectively. Statistically, the brewery biomass samples represented higher SMA values than municipal wastewater biomass samples ( p > 0.99). DGGE provides a genetic fingerprint of the microbial diversity based on the physical separation of unique nucleic acid sequences (Muyzer, 1999). DGGE banding patterns for the mcrA gene were generated for biomass samples (Fig. 2) and compared to SMA results using PCA. The first two principal components based on DGGE band intensities were employed

as x- and y-coordinates, respectively, and explained 81.6% of the total densitometric data variation of banding patterns (Fig. 3). SMA results were superimposed as a third dimension by representing biomass samples with higher SMA with larger-diameter circles in Fig. 3. The two samples represented by ‘’ symbols exhibited negligible SMA. PCA indicated an underlying relationship between SMA and DGGE banding pattern. Biomass samples clustered in three groups based on principal component coordinates (Fig. 3). Cluster 1 represented biomass samples with statistically higher SMA values compared to biomass within Cluster 2 ( p > 0.99). The average SMA of biomass samples within Clusters 1 and 2 were 5.8  0.4 and 1.7  0.2 mL CH4/gVSS-hr, (average  standard deviation) respectively. Interestingly, all the brewery biomass samples were in Cluster 1. Cluster 3 contained biomass samples having intermediate SMA values and the average SMA of the samples represented in this cluster was 3.5  0.4 mL CH4/gVSS-hr. Projections of the five bands with the highest contribution to DGGE data variability (B1, B2, B3, B7 and B8) are also shown in Fig. 3. Bands B1 and B2 contributed to the high-SMA cluster (Cluster 1). Organisms represented by these bands may have an important metabolic function leading to higher SMA values. Clones extracted from band B1 shared 88e89% sequence similarity to Methanospirillum hungatei (GenBank accession no. AF313805). Clones from band B2 shared 93e98% sequence similarity to Methanobacterium beijingense (GenBank accession no. EF465106). Others have reported average mcrA gene sequence similarities within a genus and family to be 88.9 and 79%, respectively (Steinberg and Regan, 2008). Biomass from an upflow anaerobic sludge blanket (UASB) reactor treating brewery wastewater (Brewery-1) demonstrated the most rapid specific propionate conversion to CH4 (10.7  0.4 mL CH4/gVSS-hr). This culture was enriched for propionate utilization and demonstrated an SMA value of 10.7  3.3 mL CH4/gVSS-hr after 580 days of enrichment. The granular biomass was initially disrupted, and was not granular in SMA tests. In addition, the biomass was not granular after enrichment. This culture was used in the subsequent investigation of bioaugmentation for recovery of organically overloaded digesters.

3.2.

Fig. 2 e DGGE banding pattern obtained for mcrA gene analysis B1 through B10 represent individual, distinct DGGE bands.

Recovery of organically overloaded digesters

Before the organic overload, all digesters required about 40 days (4 SRTs) to attain an average quasi steady-state effluent SCOD concentration of 290  150 mg/L (Fig. 4). The shock organic overload was administrated on Day 57 and resulted in an increase in effluent SCOD concentrations to 5000  750 mg/ L for all overloaded digesters (Fig. 4). Following the organic overload, the effluent SCOD started to decrease. Approximately 6 SRTs after the overload, the influence of bioaugmentation was apparent in terms of the lower effluent SCOD concentration of bioaugmented digesters compared to non-bioaugmented digesters (Fig. 4). In addition, volatile fatty acids concentrations were measured on Days 180 and 300. On Day 180, the non-bioaugmented digesters exhibited elevated acetic, propionic and butyric acid concentrations of approximately 870, 360 and 200 mg/L, respectively, whereas bioaugmented and non-overloaded digesters averaged <50, <170

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 4 9 e5 2 5 6

5253

Fig. 3 e Principal component analysis using DGGE band intensities. Points represent individual biomass samples and are clustered according to their methanogenic community structure (i.e., DGGE band intensities). The diameter of each point is proportional to the measured biomass activity (i.e., SMA values). Vectors B1, B2, B3, B7 and B8 represent DGGE bands having the greatest influence on differences in community structure. Clones from bands B1 and B2 were most similar to Methanospirillum hungatei and Methanobacterium beijingense, respectively.

and <30 mg/L of these acids, respectively. On Day 300, the concentrations of all volatile fatty acids were <20 mg/L in all digesters. The time required for organically overloaded digester effluent SCOD to decrease below 1000 mg/L was used as a measure of recovery time. Bioaugmented digesters required 25 days less to attain 1000 mg SCOD/L compared to non-bioaugmented digesters ( p > 0.97). Before organic overload, all digesters reached an average quasi steady-state CH4 production rate of 32  4 mLCH4/day after 40 days (4 SRTs). Following the organic overload on Day 57, CH4 production from the overloaded digesters decreased to

18  13 mLCH4/day (Fig. 5). Daily CH4 production began to recover after the organic overload. The time required to attain 75% of the quasi steady-state CH4 production rate (i.e. 24 mLCH4/day) was chosen to define the CH4 production recovery period. The bioaugmented digesters required 28 days less to achieve 24 mLCH4/d after overload compared to nonbioaugmented digesters ( p > 0.99). The percent difference between bioaugmented and nonbioaugmented digester effluent SCOD concentrations was

Fig. 4 e Effluent SCOD of digesters. Error bars represent standard deviation among four replicates. Most error bars are small and not visible.

Fig. 5 e Average daily CH4production from digesters. Points represent average methane production form four replicates. Error bars not shown for visual clarity.

5254

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 4 9 e5 2 5 6

evaluated at 6, 9, 12 and 24 SRTs following the organic overload (see Fig. 6a). The influence of bioaugmentation persisted for more than 12 SRTs following the shock overload. For example, the effluent SCOD of bioaugmented digesters was still lower than that of controls after Day 210. However, bioaugmented and non-bioaugmented effluent SCOD concentrations were similar 24 SRTs (i.e, >7 months) after the time of organic overload (Fig. 6a). The percent difference between bioaugmented and nonbioaugmented digester CH4 production rates was evaluated at 6, 9, 12 and 24 SRTs following the organic overload (see Fig. 6b). Bioaugmented digesters demonstrated higher average CH4 production rates than non-bioaugmented digesters for as long as 12 SRTs (i.e., 4 months) following the organic overload. However, bioaugmented and non-bioaugmented CH4 production rates were similar 24 SRTs (i.e, >7 months) after the time of organic overload (Fig. 6b).

4.

fraction reduces the apparent SMA value calculated in units of methane production rate per gram of total VSS. Therefore, VSS is a poor surrogate measure of active biomass concentration. In the future, better methods to quantify active biomass concentration, such as total DNA or adenosine triphosphate (ATP) concentration measurements could be employed. An improved active biomass measurement could then be used to calculate more appropriate specific rates for comparison among different biomass samples. Even with the complication of VSS as a poor surrogate for active biomass quantification, a relationship between microbial community structure and SMA value was discerned. Principal components analysis of the DGGE banding data of methanogens indicated an underlying relationship between propionate SMA and methanogenic community profile. In the future, statistically significant quantitative structure-activity relationships (QSARs) may be developed to relate microbial community structure and biomass activities, such as SMA. These relationships will be very useful when attempting to improve the operation of existing anaerobic digesters. The two DGGE bands that positively influenced clustering of the samples with higher SMA values represented organisms closely related to M. hungatei and M. beijingense. These two hydrogenotrophic methanogens appear to be distinctively important for the rapid metabolism of propionate in full-scale anaerobic bioprocesses. Both M. hungatei and M. beijingense demonstrate maximum growth at 37  C and their optimum specific growth rates are 0.053 (Koster and Koomen, 1988; Robinson and Tiedje, 1984) and 0.049 h
1 (Ma et al., 2005) respectively. These growth rate values are in the low to intermediate range compared to published growth rates of other hydrogenetrophic methanogens that vary from approximately 0.029e0.14 h
1. For example, a high growth rate (0.14 h
1) was reported for both Methanomicrobium paynteri (Koster and Koomen, 1988) and Methanobacterium strain AZ (Koster and Koomen, 1988; Zehnder and Wuhrmann, 1977), whereas a low growth rate (0.029 h
1) was reported for Methanobaeterium bryantii M.o.H. (Koster and Koomen, 1988). Strains of M. beijingense were originally isolated from a UASB digester treating brewery wastewater (Ma et al., 2005). In the current study, all the brewery biomass samples were within Cluster 1 having higher SMA values and organisms sharing 93e98% sequence similarity with M. beijingense had a significant effect on the clustering pattern. Also, it is important to note that both bands that significantly contributed to the highSMA cluster represented hydrogenotrophic methanogens.

Discussion

The SMA values for biomass samples from fourteen different anaerobic bioprocesses varied over two orders of magnitude. Statistically, the measured SMA values represented six different groups, and there was significant variability among anaerobic cultures in terms of their maximum propionate degradation rate. In addition, the SMA against propionate of a biomass sample did not significantly increase after over 500 days of enrichment feeding propionate. Therefore, the SMA of biomass used to seed or reseed full-scale anaerobic bioprocesses may have a significant impact on the start-up time and steady-state operation. All the brewery biomass samples demonstrated high-SMA values, whereas municipal digester biomass demonstrated lower SMA values. The SMA values for biomass samples from processes treating other types of waste generally were between these extremes. At least two phenomena may explain the variation in biomass SMA values: (1) variability in the active VSS fraction and (2) variability in microbial community structure among samples. Undigested, inactive VSS was ostensibly present in the biomass samples used for testing, and could have been more prevalent in biomass from municipal digesters that treat sludges containing a high concentration of inactive VSS. This inactive fraction of VSS represents solids that are not live, active microorganisms, and do not contribute to biomass activity. The inactive VSS

b

50 40 30 20 10 0 6 9 12 24 SRT following the organic overload

Increase in methene production, %

Decrease in effluent SCOD, %

a

140 120 100 80 60 40 20 0 6 9 12 24 SRTs following the organic overload

Fig. 6 e (a) Decrease in effluent SCOD concentration from bioaugmented digesters compared to non-bioaugmented digesters. (b) Increase in CH4 production rate from bioaugmented digesters compared to non-bioaugmented digesters. Error bars represent standard error among three replicates.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 4 9 e5 2 5 6

This finding underlines the fact that hydrogenotrophic methanogens play a very important role in anaerobic metabolism of propionate (McCarty and Smith, 1986). Under the conditions studied, bioaugmentation using an anaerobic culture having a relatively high propionate utilization rate reduced the recovery time of organically overloaded anaerobic digesters by decreasing effluent SCOD concentration and increasing methane production. The conditions included small-scale reactors with excellent mixing and flocculant suspended growth with no effluent recycle at an SRT of 10 days. The results may not be indicative of other systems that include larger volumes, granular biomass, or recycle flows. The benefits of bioaugmentation, in terms of decreased SCOD and increased methane production, were apparent under the studied conditions throughout the recovery and even after 12 SRTs following the organic overload. This observation is in contrast to the findings of others that bioaugmenting an anaerobic filter with methanogenic cultures enriched for propionate and butyrate did not speed up the recovery following an organic overload (Lynch et al., 1987). The reason for the difference may be because our bioaugmentation culture was supplied on a daily basis throughout the recovery period, whereas the previous researchers (Lynch et al., 1987) supplied the bioaugmentation culture only once, possibly leading to washout.

5.

Conclusions

The SMA values of biomass samples from various full-scale anaerobic reactors can differ greatly. The SMA values against propionate for biomass samples tested herein varied over two orders of magnitude. Therefore, seed biomass for new reactors should be chosen carefully, and activity testing is recommended when selecting seed biomass. Comparison of DGGE banding patterns for the mcrA gene with SMA values for 14 biomass samples indicate an underlying relation between methanogenic community structure and activity. The presence of hydrogenotrophic methanogens closely related to M. hungatei and M. beijingense was related to high-SMA values against propionate for anaerobic biomass. However, more research is required to establish a refined structure-activity relationship (SAR) to predict biomass activity from microbial community structure. In the future, SARs or quantitative SARs (QSARs) may be developed, and more highly-defined microbial communities may be employed to improve specific aspects of anaerobic digester performance. Bioaugmentation is a promising approach to improve digester operations after transient upsets. Under the conditions studied, bioaugmentation with a propionate-utilizing enrichment was successful, and decreased the recovery time of digesters following an organic overload. Biogas and CH4 production rates and effluent SCOD values returned to pre-upset values more quickly when bioaugmentation was implemented. In addition, the benefits of bioaugmentation (lower effluent SCOD, higher CH4 production) were evident for an unexpectedly long time after upset (i.e., 12 SRTs). Future research should address the changes in microbial community structure and longer-term system performance that may occur in anaerobic systems when bioaugmentation is practiced.

5255

Acknowledgments The authors thank WE Energies for providing funding for this research.

references

Altschul, S.F., Madden, T.L., Schaffer, A.A., Zhang, J., Zhang, Z., Miller, W., Lipman, D.J., 1997. Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Research 25, 3389e3402. American Public Health Association (APHA), American Water Works Association (AWWA), Water Environment Federation (WEF), et al., 1998. Standard Methods for the Examination of Water and Wastewater, twentieth ed. Bidle, K.A., Kastner, M., Bartlett, D.H., 1999. A Plylogenetic analysis of microbial communities associated with methane Hydrate containing marine Fluids and sediments in the Cascadia margin (ODP site 829B). FMES Microbiology Letters 177 (1), 101e108. Dhillon, A., Lever, M., Lloyd, K.G., Albert, D.B., Sogin, M.L., Teske, A., 2005. Methanogen diversity evidenced by molecular characterization of methyl coenzyme M reductase A (mcrA) genes in hydrothermal sediments of the Guaymas basin. Applied Environmental Microbiology 71 (8), 4592e4601. Earl, J., Hall, G., Pickup, R.W., Ritchie, D.A., Edwards, C., 2003. Analysis of methanogen diversity in a hypereutrophic lake using PCR-RFLP analysis of mcr sequences. Microbial Ecology 46 (2), 270e278. Galand, P.E., Saarnio, S., Fritze, H., Yrja¨la¨, K., 2002. Depth related diversity of methanogen archaea in Finnish oligotrophic Fen. FEMS Microbiology Ecology 42 (3), 441e449. Guiot, S.R., Tartakovsky, B., Lanthier, M., Levesque, M.J., Manuel, M.F., Beaudet, R., Greer, C.W., Villemur, R., 2002. Strategies for augmenting the pentachlorophenol degradation potential of UASB anaerobic granules. Water Science and Technology 45 (10), 35e41. Hales, B.A., Edwards, C., Ritchie, D.A., Hall, G., Pickup, R.W., Saunders, J.R., 1996. Isolation and identification of methanogen-specific DNA from blanket bog peat by PCR amplification and sequence analysis. Applied Environmental Microbiology 62 (2), 668e675. Juottonen, H., Galand, P.E., Yrjala, K., 2006. Detection of methenogenic archaea in peat: comparison of PCR primers targeting the mcrA gene. Research in Microbiology 157 (10), 914e921. Koster, I.W., Koomen, E., 1988. Ammonia Inhibition of the maximum growth rate (mm) of hydrogenotrophic methanogens at various pH-Levels and temperatures. Applied Microbial Biotechnology 28, 500e505. Lenz, M., Enright, A.M., O’Flaherty, V., van Aest, A.C., Lens, P.N.L., 2009. Bioaugmentation of UASB reactors with Immobilized Sulfurospirillum barnesii for simultaneous selenate and nitrate removal. Applied Microbiology and Biotechnology 83 (2), 377e388. Lloyd, D., Thomas, K.L., Hayes, A., Hill, B., Hales, B.A., Edwards, C., Saunders, J.R., Ritchie, D.A., Upton, M., 1998. Micro-ecology of peat: minimally invasive analysis using confocal laser scanning microscopy, membrane inlet mass spectrometry and PCR amplification of methanogen-specific gene sequences. FMES Microbiology Ecology 25 (2), 179e188. Lueders, T., Chin, K.J., Conrad, R., Friedrich, M., 2001. Molecular analysis of methyl-coenzyme M reductase a-Subunit (mcrA) gene in rice field Soil and enrichment cultures Reveal the

5256

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 4 9 e5 2 5 6

methanogenic phenotype of a novel archaeal lineage. Environmental Microbiology 3 (3), 194e204. Luton, P.E., Wayne, J.M., Sharp, R.J., Riley, P.W., 2002. The mcrA gene as an alternative to 16S rRNA in the phylogenetic analysis of methanogen populations in landfill. Microbiology 148, 3521e3530. Lynch, N., Daniels, L., Parkin, G.F., 1987. Bioaugmentation of Stressed Anaerobic Filters with Methanogenic Enrichment Cultures. In: Proceedings of the 42nd Industrial Waste Conference. Purdue University, West Lafayette, Indiana, pp. 285e296. Ma, K., Liu, X., Dong, X., 2005. Methanobacterium beijingense sp. nov., a novel methanogen isolated from anaerobic digesters. International Journal of Systematic and Evolutionary Microbiology 55, 325e329. McCarty, P.L., Smith, D.P., 1986. Anaerobic wastewater treatment. Environmental Science and Technology 20 (12), 1200e1206. Muyzer, G., 1999. DGGE/TGGE a method for identifying genes from natural ecosystems. Current Opinion in Microbiology 2, 317e322. Nercessian, D., Upton, M., Lloyd, D., Edwards, C., 1999. Phylogenetic analysis of peat bog methanogen population. FMES Microbiology Letters 173 (2), 425e429. Nielsen, H.B., Mladenovska, Z., Ahring, B.K., 2007. Bioaugmentation of a two-stage thermophilic (68 C/55 C) anaerobic digestion concept for improvement of the methane yield from Cattle manure. Biotechnology and Bioengineering 97 (6), 1638e1643. Ohkuma, M., Noda, S., Horikoshi, K., Kudo, T., 1995. Phylogeny of symbiotic methanogens in the gut of the termite reticulitermes speratur. FMES Microbiology Letters 134 (1), 45e50. Rastogi, G., Ranade, D.R., Yeole, T.Y., Patole, M.S., Shouche, Y.S., 2008. Investigation of methanogen population structure in biogas reactor by molecular characterization of methylcoenzyme M reductase A (mcrA) genes. Bioresource Technology 99 (13), 5317e5326. Rittmann, B.E., Whiteman, R., 1994. Bioaugmentation: a coming of age. Water Quality International 1, 12e16. Robinson, J.A., Tiedje, J.M., 1984. Competition between sulfatereducing and methanogenic bacteria for H2 under resting and growing conditions. Archives of Microbiology 137 (1), 26e32.

Sambrook, J., Russell, D.W., 2001. Molecular Cloning: A Laboratory Manual, 3rd ed. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Schauer-Gimenez, A.E., Zitomer, D.H., Maki, J.S., Struble, C.A., 2010. Bioaugmentation for improved recovery of anaerobic digesters after toxicant exposure. Water Research 44 (12), 3555e3564. Smith, D.P., McCarty, P.L., 1990. Factors governing methane fluctuations following shock loading of digesters. Research Journal of the Water Pollution Control Federation 62 (1), 58e64. Sorensen, A.H., Ahring, B.K., 1993. Measurements of the specific methanogenic activity of anaerobic digestor biomass. Applied Microbiology and Biotechnology 40, 427e443. Speece, R.E., 2008. Anaerobic Biotechnology and Odor/Corrosion Control for Municipalities and Industries. Archae Press, Nashville, TN. Steinberg, L.M., Regan, J.M., 2008. Phylogenetic comparison of the methanogenic communities from an acidic, oligotrophic Fen and an anaerobic digester treating municipal wastewater sludge. Applied and Environmental Microbiology 74 (21), 6663e6671. Tawfiki Hajji, K., Lepine, F., Bisaillon, J.G., Beaudet, R., Hawari, J., Guiot, S.R., 2000. Effects of bioaugmentation strategies in UASB reactors with a methanogenic consortium for removal of phenolic compounds. Biotechnology and Bioengineering 67 (4), 417e423. Thauer, R.K., 1998. Biochemistry of methanogenesis: a tribute to Marjory Stephenson. Microbiology 144, 2377e2406. Ufnar, J.A., Ufnar, D.F., Wang, S.Y., Ellender, R.D., 2007. Development of a swine-specific fecal pollution marker based on host differences in methanogen mcrA genes. Applied and Environmental Microbiology 73 (6), 5209e5217. Wilms, R., Sass, H., Ko¨pke, B., Cypionka, H., Engelen, B., 2007. Methane and Sulfate Profiles within the Subsurface of a Tidal Flat are Reflected by the Distribution of Sulfate-Reducing B. Woese, C.R., Fox, G.E., 1977. Phylogenetic structure of the prokaryotic domain: the primary kingdoms. Proceedings of the National Academy of Sciences of the United States 74 (11), 5088e5090. Zehnder, A.J.B., Wuhrmann, K., 1977. Physiology of a Methanobacterium strain AZ. Archives of Microbiology 111, 199e205.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 5 7 e5 2 6 5

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Ammonium adsorption in aerobic granular sludge, activated sludge and anammox granules J.P. Bassin a,b, M. Pronk a, R. Kraan c, R. Kleerebezem a, M.C.M. van Loosdrecht a,* a

Department of Biotechnology, Delft University of Technology, Julianalaan 67, 2628 BC Delft, The Netherlands Chemical Engineering Program, Federal University of Rio de Janeiro, Rio de Janeiro, Brazil c DHV Water b.v., Laan 1914 nr. 35, 3818 EX Amersfoort, The Netherlands b

article info

abstract

Article history:

The ammonium adsorption properties of aerobic granular sludge, activated sludge and

Received 12 April 2011

anammox granules have been investigated. During operation of a pilot-scale aerobic

Received in revised form

granular sludge reactor, a positive relation between the influent ammonium concentration

4 July 2011

and the ammonium adsorbed was observed. Aerobic granular sludge exhibited much

Accepted 25 July 2011

higher adsorption capacity compared to activated sludge and anammox granules. At an

Available online 30 July 2011

equilibrium ammonium concentration of 30 mg N/L, adsorption obtained with activated sludge and anammox granules was around 0.2 mg NH4-N/g VSS, while aerobic granular

Keywords:

sludge from lab- and pilot-scale exhibited an adsorption of 1.7 and 0.9 mg NH4-N/g VSS,

Ammonium adsorption

respectively. No difference in the ammonium adsorption was observed in lab-scale reac-

Granular sludge

tors operated at different temperatures (20 and 30
C). In a lab-scale reactor fed with saline

Activated sludge

wastewater, we observed that the amount of ammonium adsorbed considerably decreased

Anammox

when the salt concentration increased. The results indicate that adsorption or better ion

Isotherms

exchange of ammonium should be incorporated into models for nitrification/denitrifica-

Kinetics

tion, certainly when aerobic granular sludge is used. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

The increasing amounts of nitrogen compounds in water and wastewater contribute to the occurrence of eutrophication of surface waters. Nitrogen removal is generally accomplished by nitrification and denitrification, two important processes involved in wastewater treatment successfully applied for many decades. Nitrification is the microbial oxidation of NHþ 4  to NO 2 and further to NO3 . The last compound is reduced to nitrogen gas during denitrification in a multi-step reaction  (NO 3 / NO2 / NO / N2O / N2). Calculation of nitrogen conversion and mass balances in full scale treatment systems or batch activity tests is a complex task due to the numerous parallel conversions

involved in nitrogen removal processes. For instance, regu larly the measured production of NO 2 and NO3 by nitrification þ is higher than the NH4 removed from solution. This is generally attributed to simultaneous nitrification of the ammonium generated by ammonification, biomass decay or to analytical problems (Nielsen, 1996). There are however indications in literature suggesting that other phenomenon should be taken into account to track the flow of the nitrogen compounds. Among them, adsorption of ammonium to biomass seems to be an important process. The extracellular polymeric substances (EPS) and microbial cell surfaces carry a negative electric charge (Wilkinson, 1958). Therefore, the EPS matrix can function as an ion exchanger for cations (e.g. Ca2þ, Mg2þ and NHþ 4 ) and heavy metals. The binding of heavy metals

* Corresponding author. Tel.: þ31 15 2781618; fax: þ31 15 2782355. E-mail address: [email protected] (M.C.M. van Loosdrecht). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.034

5258

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 5 7 e5 2 6 5

Nomenclature Ceq Cinf Gads Gmax ads G30 ads

equilibrium ammonium concentration (mg N/L) influent ammonium concentration (mg N/L) ammonium adsorbed (mg N/g VSS) maximum adsorption constant (mg N/g VSS) ammonium adsorbed at a Ceq concentration of 30 mg NH4-N/L (mg N/g VSS)

(Liu et al., 2001; Fukushi et al., 1996; Guibaud et al., 2003; Comte et al., 2006) and some cations such as Ca2þ and Mg2þ (Dupraz et al., 2004; Dupraz et al., 2009) to EPS has been well studied. However only a limited number of references point to ammonium adsorption in activated sludge systems or biofilms. Nielsen (1996) studied the extent of adsorption of NHþ 4 to activated sludge from full scale wastewater treatment plants (WWTP) where nitrification and denitrification were occurring. In that study, it was observed that the percentage of ammonium adsorbed to the sludge flocs was between 20 and 25% at dissolved ammonium concentrations of 1e6 mg NH4N/L. When the bulk concentration was around 15 mg NH4-N/L, the equivalent of 2 mg NH4-N/L was absorbed. The maximum adsorption capacity reported was in the range of 0.3e0.4 mg NH4-N/g VSS. Wik (1999) estimated an ammonium adsorption of 2.7 mg NH4-N/m2 in a trickling filter at an influent ammonium concentration of 15 mg N/L. During the treatment of municipal wastewater by the BIOFIX-process, Temmink et al. (2001) observed that 9 or 21% of the influent ammonium load was adsorbed by the biofilm when the influent ammonium concentration was 52  20 mg/L or 37  20 mg/L, respectively. Schwitalla et al. (2008) found that the adsorption to activated sludge flocs was within a range of 0.07e0.20 mg NH4-N/g VSS. Neglecting the ammonium adsorption could therefore in cases lead to underestimations of 10e25% of the ammonium available for nitrification. In the experiments with lab- and pilot-scale aerobic granular sludge reactors with alternate anaerobic/aerobic phases carried out in this study, ammonium concentrations after anaerobic feeding were found to be lower than expected based on the influent concentration and dilution in the reactor. This fact was associated with a possible ammonium adsorption phenomenon to the aerobic granules. Therefore we decided to perform a study on the ammonium adsorption properties of aerobic granular sludge in comparison with activated sludge and anammox granular sludge. Adsorption kinetics and adsorption isotherms were determined in order to provide a better insight in the ammonium adsorption process and for potential future inclusion in mathematical process models.

2.

Materials and methods

2.1. Lab- and pilot-scale aerobic granular sludge reactors Two lab-scale aerobic granular sludge sequencing batch reactors with a working volume of 2.6 L were operated at different temperatures (20 and 30
C). The cycle time of both reactors

K X kads VSS VER

half saturation constant (mg N/L) total biomass concentration (g VSS/L) adsorption rate constant (L/g VSS/h) volatile suspended solids (g/L) volume exchange ratio

was 3 h and comprised 60 min anaerobic feeding from the bottom of the reactor in a plug-flow regime through the settled bed, 112 min aeration provided by an air diffuser, 3 min settling and 5 min effluent withdraw. The volume exchange ratio (VER) was 0.57 (57%), resulting in a hydraulic retention time of 5.2 h. DO concentration during aeration phase was kept constant at 20% air saturation by mixing air and nitrogen in the inlet gas by a mass flow controller. The reactors were fed with synthetic wastewater with the following composition: (A) NaAc 63 mM, MgSO4$7H2O 3.6 mM, KCl 4.7 mM and (B) NH4Cl 35.4 mM, K2HPO4 4.2 mM, KH2PO4 2.1 mM and 10 mL/L trace element solution (Vishniac and Santer, 1957). In each cycle, 150 mL from both media together with 1300 mL of tap water was mixed. This resulted in COD of 400 mg/L and ammonium concentration of 60 mg N/L in the feeding media. The initial ammonium concentration for each cycle was considered to be the expected concentration after anaerobic feeding (i.e. 34 mg N/L), taking into account the dilution of the feeding media with the remaining liquid in the reactor (volume exchange ratio of 0.57) and assuming that all ammonium was depleted during the previous cycle. The reactor operated at 20
C was previously fed with the same wastewater containing different salt concentrations (0e30 g NaCl/L). The pilot-scale aerobic granular sludge reactor had a volume of around 1.5 m3. Domestic sewage and wastewater from a slaughterhouse contributed to 75% and 25% of the incoming COD, respectively. Influent COD and ammonium concentrations were around 600 mg/L and 50e100 mg NH4-N/L, respectively. The cycle profile was largely comparable of to the lab-scale reactors.

2.2.

Adsorption batch tests

Adsorption tests were carried out using granular sludge that was collected from lab- and pilot-scale sequencing batch reactors at the end of their operational cycle. Anammox granules were collected from Dokhaven WWTP (Rotterdam, The Netherlands). Activated sludge was taken from the outflow of the nitrification tank from two WWTP (Harnaschpolder (biological P-removal) and Kralingseveer (chemical Premoval), The Netherlands). Aerobic granular sludge and activated sludge were aerated for 1 h to minimize residual ammonium that could be present. For the anammox granules, nitrogen gas was supplied instead of compressed air. Two types of adsorption batch experiments were conducted: one varying the ammonium concentration and keeping biomass concentration constant and the other varying biomass concentration and keeping the initial ammonium concentration the same. In the experiments with a constant biomass

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 5 7 e5 2 6 5

concentration, 250 mL flasks were filled with a fixed amount of biomass (either aerobic granules, activated sludge or anammox granules) and with 0.1 M TriseHCl buffer (pH 7). In the beginning of the experiment, pulses of ammonium were added to the flasks in order to have different initial ammonium concentrations. For the second type of experiment (same initial ammonium concentration), the flasks were filled with different amounts of biomass and with 0.1 M TriseHCl buffer (pH 7). An ammonium pulse was added to have a similar final concentration in each flask. Nitrogen gas was supplied to all flasks to ensure anaerobic conditions. Samples were taken in different time intervals in order to have an overview of the adsorption kinetics and maximum adsorption capacity of aerobic granules, activated sludge and anammox granules.

2.3.

Desorption batch tests

Desorption batch tests were conducted to observe the reversibility of ammonium adsorption. Firstly, a normal adsorption test was performed as described in Section 2.2, in which a certain amount of biomass (8.4 g VSS/L) was used. A pulse of ammonium was added in the beginning of the experiment to reach 50 mg N/L. The equilibrium ammonium concentration was measured after the test and the amount of ammonium adsorbed in the granules (mg NH4-N/g VSS) was estimated based on the removed ammonium. These granules with known amount of adsorbed ammonium were sieved (to remove all the bulk liquid) and transferred to batch flasks (250 mL) filled with 0.1 M TriseHCl buffer (pH 7). After 80 min the ammonium concentration was measured and the amount of ammonium remaining adsorbed was estimated. The theoretical ammonium concentration expected in the liquid phase after desorption was calculated from the previous obtained adsorption isotherm with the granular sludge used in the desorption tests.

2.4. Analytical measurements and calculation procedures In order to be able to detect even small differences in the ammonium concentrations in the adsorption batch tests, ammonium was measured by a flow-injection analysis system (QuikChem 8500, Lachat Instruments, Inc.). The detection limit was 0.1 mg NH4-N/L. Biomass concentration was determined according to Standard Methods (APHA, 1998). The importance of adsorption in the aerobic granular sludge reactors, expressed as percentage, was calculated based on the expected ammonium concentration at the end of anaerobic feeding without adsorption and the measured ammonium concentration at the end of the feeding phase.

2.5.

Modelling ammonium adsorption

A mathematical tool was developed to predict ammonium adsorption in an aerobic granular sludge reactor as a function of the expected ammonium concentration at the end of the feeding phase, which was estimated based on the ammonium influent concentration (Cinf) and the dilution in the reactor

5259

taking into account the volume exchange ratio (VER). Ammonium adsorption can be described by an ammonium mass balance (1), and a Langmuir adsorption isotherm (2): Cinf $VER ¼ Gads $X þ Ceq Gads ¼

Ceq
$Gmax ads Ceq þ K

(1)

(2)

The equations consist of three known variables (Cexp, VER and X ), two unknown variables (Ceq and Gads) and two parameters (Gmax ads and K ). Parameter values characterizing the adsorption capacity of the biomass were estimated by fitting the measured Ceq-values to the modelled values that can be obtained from a quadratic solution of equations (1) and (2). For easy comparison of the adsorption capacity of different types of biomass, a characteristic value of the adsorption capacity at an equilibrium concentration of 30 mg N/L is defined and calculated from equations (1) and (2) (G30 ads , mg N/g VSS). Ammonium adsorption kinetics were characterized using a simple model that assumes that the biomass specific ammonium adsorption is first order in the driving force for adsorption:
dCðtÞ ¼ kads $X$ CðtÞ  Ceq dt

(3)

In this equation kads is the biomass specific kinetic constant for ammonium adsorption (L/g VSS/h), and C(t) is the time dependent ammonium concentration in the liquid. Integration of this equation allows for description of C(t) as a function of time:
CðtÞ ¼ Ceq þ C0  Ceq $ekads $X$t

(4)

where C0 is the initial liquid concentration of ammonium. Values for kads were estimated by minimizing the sum of the square of the errors between measured and calculated liquid concentrations of ammonium.

3.

Results

3.1. Adsorption in pilot-scale aerobic granular sludge reactor During operation of the pilot-scale aerobic granular sludge reactor, we observed a positive relation between the influent ammonium concentration and the ammonium adsorbed. Based on the equilibrium ammonium concentration at the end of the anaerobic feeding phase (Ceq) and the ammonium concentration that was expected (represented as Cinf*VER) based on the influent ammonium concentration (Cinf) and the dilution in the reactor considering the volume exchange ratio (VER), the amount of adsorbed ammonium (Gads*VSS) was estimated (Fig. 1). Dotted lines in Fig. 1 represent the model fitted to the data and will be discussed in Discussion section. In general, the higher the influent ammonium concentration (and therefore the expected ammonium at the end of the anaerobic feeding), the greater the amount of ammonium adsorbed. The biomass concentration in the pilot-scale

5260

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 5 7 e5 2 6 5

Fig. 1 e Amount of ammonium adsorbed (Gads*VSS) and ammonium concentration in the equilibrium (Ceq) at different expected ammonium concentrations after anaerobic feeding (Cinf*ER). The expected ammonium concentration was estimated based on the influent ammonium concentration and the dilution in the reactor considering the volume exchange ratio (VER). Dotted lines represent the model fitted to the experimental data.

reactor was roughly constant and equal to 8 g VSS/L. Around 18e24% of the expected ammonium concentration at the end of the feeding phase was adsorbed when the ammonium concentration in the incoming wastewater ranged from 50 to 100 mg NH4-N/L. Since some ammonification may have taken place in the anaerobic period, the data in Fig. 1 do not provide a true adsorption isotherm.

3.2. Adsorption in lab-scale aerobic granular sludge reactor In two lab-scale reactors operated at different temperatures (20 and 30
C) and fed with the same medium, the adsorption of ammonium was quite similar to the pilot-scale system ðG30 ads w1 mg NH4  N=g VSSÞ. The adsorption varied from 23 to 36% (for 20
C) and from 27 to 37% (for 30
C) of the ammonium concentration expected after anaerobic feeding (34 mg NH4-N) in the aerobic granular sludge reactor. Biomass concentration was kept roughly constant in both reactors (around 12 g VSS/ L). The temperature seems not to influence the adsorption in granules in the range studied (20e30
C). During operation of the aerobic granular sludge lab-scale reactors, it was also observed that the amount of ammonium adsorbed during anaerobic feeding was reversely proportional to the ammonium concentration remaining in the end of the previous cycle. Therefore, when incomplete nitrification occurred, a smaller amount of influent ammonium was adsorbed. This observation emphasises the importance of nitrification for the extent of adsorption. In the reactor fed with synthetic wastewater containing salt (NaCl), it was observed that the amount of ammonium adsorbed considerably decreased when the salt concentration was increased (Fig. 2). At 10 g NaCl/L, the ammonium adsorption was approximately half of that obtained when no salt was added to the reactor. Moreover, no adsorption was observed at 30 g NaCl/L.

Fig. 2 e Ammonium adsorption in a lab-scale aerobic granular sludge reactor operated at different salt concentrations.

3.3.

Adsorption batch tests

3.3.1.

Determination of adsorption kinetics

Experiments were conducted in order to determine the kinetics of the adsorption taking place in aerobic granular sludge. No adsorption kinetics study was performed for activated sludge, since the experiments clearly showed that adsorption was very fast (within 5 min). Adsorption rates with granular sludge are significantly lower probably due to mass transfer limitations in the biofilm. The data from the adsorption kinetics experiments at variable biomass or initial ammonium concentrations are shown in Fig. 3. In both experiments, we observed a rapid ammonium adsorption in the beginning of the experiment. The adsorption rate gradually decreased until the equilibrium concentration (Ceq) was reached. In general, Ceq was reached within 60 min of experiment. The anaerobic feeding phase in the aerobic granular sludge reactors has a similar length, suggesting that in this period equilibrium adsorption is reached in the reactors. A kinetic model described in Materials and methods section was used to characterize the kinetic properties of the adsorption process. The average biomass specific kinetic constant for ammonium adsorption obtained in the experiments with variable biomass and initial ammonium concentrations was comparable and amounted 0.33  0.06 L/g VSS/h and 0.31  0.14 L/g VSS/h, respectively.

3.3.2.

Adsorption isotherms

The ammonium adsorption isotherms obtained in the experiments with activated sludge, in which either ammonium concentration was varied and the amount of biomass was kept constant or the other way around are shown in Fig. 4. The experimental data were fitted to a Langmuir adsorption isotherm. Parameters such as maximum adsorption capacity ðGmax ads Þ and the half saturation constant (K ) are not reported here since measurements were conducted at ammonium concentrations that were too low for identification of the maximum adsorption capacity. At ammonium concentrations lower than 40 mg N/L, the amount of ammonium adsorbed was directly proportional to the equilibrium concentration, independent of the activated sludge used. Ammonium adsorption obtained at 30 mg NH4-N/L (G30) for the activated

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 5 7 e5 2 6 5

5261

Fig. 3 e Experimental data and model fitted to the adsorption kinetics for the experiment keeping biomass (a) and initial ammonium (b) concentrations constant.

sludge collected from Harnaschpolder and Kralingseveer WWTP was around 0.18 and 0.16 mg NH4-N/g VSS, respectively. The sludges of Harnaschpolder and Kralingseveer showed a similar ammonium adsorption behaviour. Since typical sludge concentrations are lower than 4 g VSS/L in activated sludge systems, less than 2% of the ammonium will be present in an adsorbed form. This amount can, as is usually done, be neglected. The adsorption of ammonium to aerobic granular sludge from both lab- and pilot-scale reactors is illustrated in Fig. 5. As for activated sludge at lower ammonium concentrations the amount of ammonium adsorbed is linearly proportional to the adsorbed amount. The G30 ads  values were equal to 1.7 and 0.9 mg NH4-N/g VSS for the lab-scale and pilot-scale reactor granules. Clearly these values are an order of magnitude higher than for activated sludge. At sludge concentrations of 8 g VSS/L and higher in aerobic granular sludge reactors, this means that a very significant fraction of ammonium is adsorbed to the granular sludge. A Langmuir-type isotherm was fitted through the experimental data (non-linear regression). Values obtained for the maximum adsorption constant ðGmax ads Þ were 10 and 1.65 mg N/VSS for lab-scale granules and pilot-scale granules, respectively. Half saturation constants (K ) amounted 175 and 28 mg N/L for lab-scale granules and pilot-scale granules, respectively.

For comparison, adsorption tests with anammox granular sludge were conducted as well (Fig. 6). The G30 ads was around 0.20 mg NH4-N/g VSS. Anammox granules showed therefore also a low ammonium adsorption capacity. However due to much higher biomass content in these anammox granular sludge reactors (up to 20e30 g VSS/L) the total mass of adsorbed ammonium can be significant in a reactor system.

3.4.

Desorption batch tests

The reversibility of the ammonium adsorption process was investigated by performing desorption batch experiments. First, ammonium was allowed to adsorb to the biomass by incubation at an initial concentration of ammonium of 50 mg N/L. After adsorption 37 mg N/L remained, giving an adsorbed amount of 1.8 mg N/g VSS. In the subsequent desorption test, the granules were incubated in ammonium free medium. The ammonium concentration in the bulk after desorption reached 8.5 mg N/L. Based on the corresponding ammonium adsorption isotherm it was possible to estimate the ammonium concentration in the bulk assuming full desorption. From this test was concluded that more than 90% of the adsorbed ammonium could be desorbed, indicating full reversibility of the adsorption.

Fig. 4 e Ammonium adsorption isotherms in activated sludge from Harnaschpolder (a) and from Kralingseveer (b) obtained in the batch experiments either keeping biomass concentration constant and varying ammonium concentration (-) or keeping initial ammonium concentration constant and varying biomass concentration (,).

5262

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 5 7 e5 2 6 5

Fig. 5 e Ammonium adsorption isotherms in aerobic granular sludge from lab-scale reactor (a) and pilot-scale reactor (b) obtained in the batch experiments keeping biomass concentration constant and varying ammonium concentration (-) and keeping ammonium concentration constant and varying biomass concentration (,). Gmax: 10 mg N/g VSS; K: 175 mg N/L for lab-scale granules; Gmax: 1.65 mg N/gVSS; K: 28 mg N/L for pilot-scale granules.

4.

Discussion

From the literature it is known that ammonium can be adsorbed to activated sludge flocs (Nielsen, 1996; Schwitalla et al., 2008) and to biofilms (Wik, 1999; Temmink et al., 2001). In the lab- and pilot-scale aerobic granular sludge reactors with alternate anaerobic/aerobic phases, the ammonium concentration after the anaerobic feeding was observed to be lower than expected based on the influent concentration, which suggested the occurrence of a possible ammonium adsorption phenomenon inside the granules. Table 1 summarizes the different investigations on ammonium adsorption obtained from literature and from this research. Some results are difficult to interpret in a proper way due to the lack of information. For instance, the ammonium adsorption found by Temmink et al. (2001) was higher when the influent ammonium concentration was lower, which was possibly caused by different biomass concentrations in the reactor. Unfortunately, the biomass amount or biofilm amount was not mentioned in the publication. The same is

Fig. 6 e Ammonium adsorption isotherms in anammox sludge from a pilot-scale reactor obtained in the batch experiments either keeping biomass concentration constant and varying ammonium concentration (-) or keeping ammonium concentration constant and varying biomass concentration (,). Gmax: 0.46 mg N/g VSS; K: 35 mg N/L.

true for the work of Wik (1999), who only mentioned the ammonium adsorption per m2 or m3 of biofilm, without a reference to the biofilm thickness. Also in general the actual equilibrium concentrations at a certain amount of ammonium adsorbed is not presented. During operation of a pilot-scale aerobic granular sludge reactor, a positive relation between the influent ammonium concentration and the ammonium adsorbed was observed. This is consistent with the results obtained from adsorption batch experiments that showed higher adsorption when Ceq was higher. Temperature variation in the lab-scale reactors of 10
C did not influence the extent of adsorption, which is in line with expectations for adsorption processes. The results obtained in this study also showed that salt (NaCl) concentrations are also a key factor for the amount of ammonium that will be adsorbed. The fact that the adsorption significantly decreased as the salt concentration was increased can be explained by the competition between Naþ and NHþ 4 for binding to the negatively charged groups in the EPS or microbial cell walls. The adsorption of the ammonium to the biomass can best be seen as an ion exchange process and the presence of other cations will directly influence the amount of ammonium adsorbed. For experimental determination it is therefore possible to add salt (like Naþ or Kþ) to sludge and measure the amount of ammonium desorbing as e.g. suggested by Nielsen (1996). Since an adsorption saturation behaviour was observed at high ammonium concentration a Langmuir isotherm was used to describe the adsorption process, an ion exchange model would likely be more appropriate. Values obtained with a Langmuir model for ammonium can only be used for estimating adsorption at the medium composition for which the isotherm is determined. The adsorption isotherms obtained in the batch experiments clearly demonstrated that the adsorption in aerobic granular sludge is considerable higher than the one achieved with activated sludge and anammox granules. The estimated G30 ads  values for activated sludge and anammox granules were 0.16 and 0.20 mg NH4-N/g VSS, respectively. The adsorption to aerobic granular sludge was characterized by G30 ads  values for lab- and pilot-scale granules of 1.6 and 0.8 mg NH4-N/g VSS, respectively. From the comparison

5263

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 5 7 e5 2 6 5

Table 1 e Ammonium adsorption in different wastewater treatment systems. Author(s) Nielsen (1996) Wik (1999) Temmink et al. (2001) Schwitalla et al. (2008) Valdivia et al. (2007) This research This research This research This research This research

System Activated sludge Trickling filter BIOFIX-process Activated sludge Biofilm SBR Activated sludge Pilot-scale AGS Lab-scale AGS (T ¼ 20
C) Lab-scale AGS (T ¼ 30 7degC) Lab-scale AGS (0e30 g NaCl/L)

Adsorption %

mg N/g VSS at 30 mg N/L

NA NA 9e21%a NA 14e27%b NA 18e24%c 23e36%c 27e37%c 0e30%c

d

0.3e0.4 2.7e NA 0.07e0.20 NA 0.16e0.18 0.9 1.7 NA NA

NH4 influent (mg N/L) NA w16 30e50 NA 22 20e50 50e100 60 60 60

NA: not available. a Percentage is relative to ammonium influent load. b Percentage is relative to ammonium influent concentration. c Percentage is relative to the expected ammonium concentration after anaerobic feeding. d Total adsorption capacity at an ammonium concentration of approximately 5 mg N/L. e mg N/m2 of biofilm.

between anammox granules and aerobic granular sludge it is clear that granular sludge as such does not lead to a higher ammonium adsorption capacity. The pilot-scale granular sludge as well as the sludge from Harnaschpolder were grown on wastewater that employed biological P-removal conditions. From the comparison it is therefore clear that the higher adsorption to granular sludge is not directly related to the presence of phosphate accumulating organisms in the sludge. In the adsorption batch tests where biomass was varied and initial ammonium concentration was kept constant, we observed that the ammonium concentration in equilibrium hardly varied, especially in the experiments using activated sludge and anammox granules. This observation can be partly related to the substantially low amount of ammonia adsorbed at low biomass concentrations which increases the measurement error. Still, the higher adsorption per unit of biomass at lower biomass concentrations is to date not fully understood. Actually these experiments at low biomass concentrations do not reflect the biomass content in the biological systems from where sludge samples were collected. As an ion exchange process, the amount of ammonium adsorbed into aerobic granules will be directly related to the compounds functioning as ion exchanger for cations. Among them, extracellular polymeric substances (EPS) likely play a dominant role in the adsorption of ammonium. Sludge contains a mixture of microbial species, which can promote the synthesis of several types of EPS. These polymers can significantly vary in their composition and therefore in their chemical and physical properties. While some are neutral, others are either polyanionic (due to the presence of uronic acids and ketal-linked pyruvate) or polycationic macromolecules (Sutherland, 2001). In this respect it is interesting to note that recent studies have described that aerobic granules from different treatment systems can produce exopolysaccharides with a unique composition (Seviour et al., 2010; Lin et al., 2010; Adav and Lee, 2008). Lin et al. (2010) observed that the dry weight of aerobic granules from the pilot plant reactor treating municipal wastewater (same reactor as in this study however samples at a different time) contained more than 10% of alginate-like exopolysaccharides. These authors pointed out

that this specific exopolysaccharide is only one present in granules, although the amount reported was even higher than the total EPS content reported by Adav and Lee (2008) and Wang et al. (2006). Seviour et al. (2010) also mentioned that granules are characterized by an over-production of a single gel-forming EPS, although the structure of exopolysaccharides was different from that described by Lin et al. (2010). Nevertheless, some similarities in the granules exopolymers are described in these two publications (e.g. presence of uronic acids). Unfortunately there is no information on the ammonium adsorption to the granules. The fact that ammonium adsorption in granular sludge is considerably higher than that observed for activated sludge and anammox granules reinforce the importance of taking this ion exchange process into consideration especially when working with aerobic granules. For mass balancing over a treatment plant this is no real problem. Under stationary conditions, adsorption does not make a difference in the ammonium effluent concentrations. However, when doing kinetic studies in batch experiments or evaluating conversions from dynamic changes in for instance sequencing batch reactor (SBR) processes or plug-flow systems, a significant error can be introduced in the calculations when ammonium adsorption is neglected. Since the adsorption strongly depends on the type of EPS produced and currently this cannot be predicted from basic principles, adsorption of ammonium has to be considered in each experimental evaluation of nitrogen conversion processes. For example the adsorption of ammonium to the moving-bed biofilm in the BIOFIX-process made the process in practise not feasible due to excessive release of ammonium in the anoxic phase of the process (Temmink et al., 2001). Also the ammonium dynamics in the trickling filters described by Wik (1999) was heavily influenced by the ammonium adsorption to the biofilm matrix. Yilmaz et al. (2008) observed a decrease of 30 mg NH4-N/L during anaerobic feeding of an aerobic granular sludge reactor, but they neglected the occurrence of ammonium adsorption due to the relatively long time to reach the equilibrium concentration. The assumption of these authors was referred to the publication of Nielsen (1996), who worked with activated sludge. As

5264

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 5 7 e5 2 6 5

indicated by the experiments performed in this study, adsorption in granules takes longer and is not a fast process like in activated sludge. Therefore, ammonium adsorption should not have been neglected.

4.1.

Modeling ammonium adsorption

Ammonium adsorption was well described using a Langmuir isotherm. Due to the low ammonium concentrations applied, the maximum adsorption capacity ðGmax ads Þ could only be estimated for the aerobic granular sludges. The biomass specific adsorption characteristics as determined in the batch experiment were extrapolated to the granular sludge bioreactor conditions to validate the impact of ammonium adsorption on the ammonium concentration after feeding. Fig. 1 demonstrates that using the Gmax ads and K values obtained in the batch experiments and the biomass concentration (X, 8 g VSS/L) and the influent ammonium concentration (Cinf) in the pilot could very well be used to predict the ammonium concentration after feeding. The data clearly suggest that biomass characterization in terms of Gmax ads and K values is strictly required to describe nitrogen conversion processes in dynamic processes, like SBRs. Thus, knowing the influent ammonium concentration (and therefore the ammonium expected after feeding in aerobic granular sludge SBRs) and the biomass concentration, it is possible to predict the amount of ammonium that will be adsorbed and the amount that will remain in the bulk solution. The developed approach is valid when no ammonium is present at the end of a SBR cycle (complete nitrification). In the case where ammonium is not completely removed during the cycle, the amount of ammonium adsorbed would be lower and depends on the residual ammonium concentration in the bulk. We are assuming that granules have maximum adsorption capacity when nitrification is complete. This assumption is based on the desorption tests performed in this study, which showed that >90% of the ammonium adsorbed in the biomass can be desorbed and will therefore be available for nitrification. However, it is possible that even when ammonium is completely depleted in the liquid phase, a fraction of the adsorbed ammonium remains inside the granules. From the experiments of Nielsen (1996), using activated sludge, it was observed that 0.5e0.6 mg NH4-N/L was still adsorbed to the sludge flocs even when the dissolved ammonium was almost completely removed by microbial oxidation. Nielsen (1996) also pointed out that ammonium desorption rate can be quite slow, based on the experiment where the biomass containing 0.5 mg NH4-N/L adsorbed was further oxidized in a vigorously shaken flask for 50 min, and still 0.3 mg NH4-N/L was adsorbed. Therefore, desorption kinetics seems to be important to predict the amount of ammonium that can be exchanged and further oxidized, and should be studied in more details.

5.

Conclusions

Adsorption tests have shown that ammonium adsorption in aerobic granular sludge can be considerably higher than that occurring in activated sludge and anammox granules. Kinetic

experiments with granules showed furthermore that adsorption in granules is much slower than for activated sludge. Ammonium adsorption cannot be neglected in granular sludge bioreactor systems that are characterized by strongly variable ammonium concentrations as a function of place (plug-flow systems) or time (batch systems). A method for description of ammonium adsorption in computational models for nitrification/denitrification in biofilm and granular sludge systems was proposed.

references

Adav, S.S., Lee, D.-J., 2008. Extraction of extracellular polymeric substances from aerobic granule with compact interior structure. J. Hazard. Mater. 154, 1120e1126. APHA, 1998. Standard Methods for the Examination of Water and Wastewater, 20th ed. American Public Health Association, Washington D.C., USA. Comte, S., Guibaud, G., Baudu, M., 2006. Biosorption properties of extracellular polymeric substances (EPS) resulting from activated sludge according to their type: soluble or bound. Process Biochem. 41, 815e823. Dupraz, C., Reid, R.P., Braissant, O., Decho, A.W., Norman, R.S., Visscher, P.T., 2009. Processes of carbonate precipitation in modern microbial mats. Earth-Sci. Rev., 141e162. Dupraz, C., Visscher, P.T., Baumgartner, L.K., Reid, R.P., 2004. Microbeemineral interactions: early carbonate precipitation in a hypersaline lake (Eleuthera Island, Bahamas). Sedimentology 51, 745e765. Fukushi, K., Chang, D., Ghosh, S., 1996. Enhanced heavy metal uptake by activated sludge cultures grown in the presence of biopolymer stimulators. Water Sci. Technol. 34, 267e272. Guibaud, G., Tixier, N., Bouju, A., Baudu, M., 2003. Relation between extracellular polymers composition and its ability to complex Cd, Cu and Pb. Chemosphere 52, 1701e1710. Lin, Y., de Kreuk, M., van Loosdrecht, M.C.M., Adin, A., 2010. Characterization of alginate-like exopolysaccharides isolated from aerobic granular sludge in pilot-plant. Water Res. 44, 3355e3364. Liu, Y., Lam, M.C., Fang, H.H.P., 2001. Adsorption of heavy metals by EPS of activated sludge. Water Sci. Technol. 43, 59e66. Nielsen, P.H., 1996. Adsorption of ammonium to activated sludge. Water Res. 30, 762e764. Schwitalla, P., Meneerich, A., Austermann-Haun, U., Mu¨ller, A., Dorninger, C., Daims, H., Holm, N.C., Ro¨nner-Holm, S.G.E., 2008. NHþ 4 ad-/desorption in sequencing batch reactors: simulation, laboratory and full-scale studies. Water Sci. Technol. 58, 345e350. Seviour, T.W., Lambert, L.K., Pijuan, M., Yuan, Z., 2010. Structural determination of a key exopolysaccharide in mixed culture aerobic sludge granules using NMR spectroscopy. In: Proceedings of the IWA World Water Congress and Exhibition, Montreal, Canada. Sutherland, I.W., 2001. Biofilm exopolysaccharides: a strong and sticky framework. Microbiology 147, 3e9. Temmink, H., Klapwijk, A., de Korte, K.F., 2001. Feasibility of the BIOFIX-process for the treatment of municipal wastewater. Water Sci. Technol. 43, 241e249. Valdivia, A., Gonza´lez-Martı´nez, S., Wilderer, P.A., 2007. Biological nitrogen removal with three different SBBR. Water Sci. Technol. 55, 245e254. Vishniac, W., Santer, M., 1957. The thiobacilli. Bacteriol. Rev. 21, 195e213. Wang, Z.W., Li, Y., Zhou, J.Q., Liu, Y., 2006. The influence of shortterm starvation on aerobic granules. Process Biochem. 41, 2373e2378.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 5 7 e5 2 6 5

Wik, T., 1999. Adsorption and denitrification in nitrifying trickling filters. Water. Res. 33, 1500e1508. Wilkinson, J.F., 1958. The extracellular polysaccharides of bacteria. Bacteriol. Rev. 22, 46e73.

5265

Yilmaz, G., Lemaire, R., Keller, J., Yuan, Z., 2008. Simultaneous nitrification, denitrification, and phosphorus removal from nutrient-rich industrial wastewater using granular sludge. Biotechnol. Bioeng. 100, 529e541.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 6 6 e5 2 7 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Improving the efficiencies of simultaneous organic substance and nitrogen removal in a multi-stage loop membrane bioreactor-based PWWTP using an on-line Knowledge-Based Expert System Zhao-Bo Chen a,b, Shu-Kai Nie a, Nan-Qi Ren b,*, Zhi-Qiang Chen b, Hong-Cheng Wang a, Min-Hua Cui a a

School of Materials Science & Chemical Engineering, Harbin Engineering University, 145 Nantong Street, Harbin 150001, China State Key Laboratory of Urban Water Resource and Environment, Harbin Institute of Technology, P.O. Box 2614, 202 Haihe Road, Harbin 150090, China b

article info

abstract

Article history:

The results of the use of an expert system (ES) to control a novel multi-stage loop

Received 6 April 2011

membrane bioreactor (MLMBR) for the simultaneous removal of organic substances and

Received in revised form

nutrients are reported. The study was conducted at a bench-scale plant for the purpose of

23 July 2011

meeting new discharge standards (GB21904-2008) for the treatment of chemical synthesis-

Accepted 25 July 2011

based pharmaceutical wastewater (1200e9600 mg/L COD, 500e2500 mg/L BOD5,

Available online 3 August 2011

50e200 mg/L NHþ 4 -N and 105e400 mg/L TN in the influent water) by developing a distributed control system. The system allows various expert operational approaches to be

Keywords:

deployed with the goal of minimizing organic substances and nitrogen levels in the outlet

Multi-stage loop membrane

while using the minimum amount of energy. The proposed distributed control system,

bioreactor (MLMBR)

which is supervised by a Knowledge-Based Expert System (KBES) constructed with G2 (a tool for expert system development) and a back propagation BP artificial neural network,

Pharmaceutical wastewater Simultaneous

nitrification

and

permits the on-line implementation of every operating strategy of the experimental

denitrification (SND)

system. A support vector machine (SVM) is applied to achieve pattern recognition. A set of

BP artificial neural network

experiments involving variable sludge retention time (SRT), hydraulic retention time (HRT) and dissolved oxygen (DO) was carried out. Using the proposed system, the amounts of

Real time Knowledge-Based

Expert

System

COD, TN and NHþ 4 -N in the effluent decreased by 55%, 62% and 38%, respectively, compared to the usual operating conditions. These improvements were achieved with little energy

(KBES)

cost because the performance of the treatment plant was optimized using operating rules implemented in real time. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

The Ministry of Environmental Protection of the People’s Republic of China has required most existing pharmaceutical

wastewater treatment plants (PWWTPs) to meet new, stricter conditions, particularly with regards to the presence of organic substances and nitrogen in the effluent. Chemical synthesis-based pharmaceutical wastewaters contain

* Corresponding author. Tel.: þ86 451 8628 2195. E-mail address: [email protected] (N.-Q. Ren). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.032

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 6 6 e5 2 7 8

a variety of organic and inorganic constituents, including spent solvents, catalysts, additives, reactants and small amounts of intermediates and products, and may therefore be high in COD (Magureanu et al., 2010; Graaff et al., 2011) and nitrogen (Tang et al., 2011). Thus, organic matter and excess nutrients (particularly nitrogen) in wastewater must be removed prior to discharging effluent into particularly sensitive media. These new requirements entail redesigning former removal procedures and developing new ones to meet the regulations. The new goals can be reached in various ways that involve alterations of existing civil works (extensions, purchases of new equipment), changes in operating procedures (e.g., the development of new treatments) or the use of control systems to optimize processes. Due to the high COD and nitrogen concentrations in such pharmaceutical wastewaters, attempts have been made to work with MBR processes as an innovative and promising wastewater treatment approach (Reif et al., 2008; Chen et al., 2008; Radjenovic et al., 2009; Tambosi et al., 2010) in recent decades. The MBR allows for the retention of a higher level of biomass and the enrichment of slow-growing microorganisms such as nitrifiers, which improve biochemical reaction rates and system performance as well as nutrient removal (Silva et al., 1998). The simultaneous removal of organic substances and nitrogen can be accomplished in the MBR through SND. Many factors, such as hydraulic retention time (HRT) (Jeff et al., 2000), sludge retention time (SRT) (Sun et al., 2007), mixed liquor suspended solids (MLSSs) and dissolved oxygen (DO), might affect the SND process. As nitrogen removal requires both aerobic and anoxic conditions, the ambient DO concentration is one of the important parameters affecting SND. Approaches to testing SND have included changing the oxygen setpoint according to the ammonium concentration, controlling the DO concentration as a function of the nitrate and ammonium concentrations and altering recirculation according to the nitrate concentration in the aerobic membrane bioreactor. A large number of control approaches therefore exist in these contexts. These approaches, however, have usually been assessed under simulated conditions. In fact, there are few references to the experimental validation of such approaches with real MBR systems. One example of the use of control systems that allow plants to be adapted to variable conditions is the STAR system (Isaacs et al., 1999), which has been successfully applied to various full-scale WWTPs. MBR is a convenient process for automation. However, the control systems applied to most MBR processes have been restricted to environmental variables, such as temperature, pH and others. The most reliable information for the control of MBR is a combination of available off-line measurements, on-line data and a detailed knowledge of the process (Huyskens et al., 2008). Control of the process mainly depends on conditions such as influent quality, temperature, HRT and DO. As most MBRs are different from one another from operational, biological and hydrodynamic standpoints, the use of control strategies requires expertise. The differences between MBRs and the complexity of the process make the supervision of the treatment difficult. Fenu et al. reported the activated sludge model (ASM) for the modeling of MBRs (Fenu et al., 2010). However, ASM has its shortcomings in processing

5267

activated sludge, as the process is not a linear system and has many operational parameters. Recently, different procedures based on artificial intelligence (e.g., artificial neural networks, fuzzy logic or expert systems) have been used to control operation. Expert systems (ESs) are intelligence computer programs that have a wide base of knowledge in a specified domain and use inferential reasoning for some problems as a human expert would (Wagner et al., 2002). During the last decade, more attention has been paid to the study and development of the monitoring, diagnosis and control of wastewater treatment systems using ES (Henze, 1997; Steyer et al., 1997, 1999). Several studies have been done in this area related to the activated sludge process. Serra et al. designed an ES for the supervision and control of activated sludge processes for a municipal wastewater treatment plant (Serra et al., 1993). Ladiges developed on-line and off-line ES to control municipal wastewater treatment plants (Ladiges and Mennerich, 1996). The on-line ES recognized faults based on signals obtained from the process, making diagnosis of the plant possible. The off-line ES was able to respond to diagnoses because it could determine the appropriate setpoints for control elements based on knowledge acquired. From the structural point of view, these two studies are very helpful in the creation of an ES for MBR. The present work was aimed at improving the efficiency of the removal of organic substances and nitrogen under dynamic and strictly controlled influent conditions. By using a control system suited to the specific operating conditions of a novel multi-level loop membrane bioreactor (MLMBR), organic substances and nitrogen can be removed with substantially higher efficiency simply by adapting the process conditions to the requirements in real time. The study was conducted by performing a series of experiments at a benchscale plant. The response of the system to changes in influent composition and flow rate was examined, and various operational and control approaches were used with the goal of maximizing COD and nitrogen removal while reducing energy costs. Although the usage of knowledge-based systems in wastewater treatment continues to grow (Maeda, 1989; Watts and Knight, 1991; Barnett, 1992; Szafnicki et al., 1998), such systems are rarely designed for MBRs. Because of their high sludge retention times (SRTs), the accumulation of soluble microbial products (SMPs) rejected by membrane filtration, high mixed liquor concentrations and high aeration rates, it is hard to build an appropriate knowledge base for MBRs. In this work, a KBES (knowledge-based expert system) for monitoring and diagnosis of the operation of an MLMBR treating chemical synthesis-based pharmaceutical (CSP) wastewater is presented. Some improvements have been made on a G2 expert system and have allowed for real-time control. In order to evaluate the efficiency and reliability of the KBES, several groups of controlled experiments were carried out.

2.

Material and methods

2.1.

Experimental apparatus and operation

A schematic representation of the MLMBR system is shown in Fig. 1. The MLMBR had a working volume of approximately

5268

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 6 6 e5 2 7 8

Fig. 1 e Layout of the KBES-controlled system.

138 L and included two submerged hollow-fiber PVDF microfiltration (MF) membrane modules (Tianjin Motian Membrane Engineering and Technology Co. Ltd., China). The MF membrane modules were characterized with a pore size of 0.22 mm and an effective surface area of 2.0 m2. The MLMBR combining the multi-stage loop reactor and membrane bioreactor was composed of an outside cylinder, an inside cylinder, a membrane module, a worm-packed bed, an aerated system, inlet and outlet systems and a backwash system (Fig. 1). The MLMBR was equipped with a feeding flow-meter (Electromagnetic flow-meter MAG 3100 P Simens, Germany), a standard pH meter (Ingold M700), an MLSS meter (Flow Electronic LS-100) and a DO meter (Flow Electronic FLX201). The wastewater was fed continuously by a peristaltic pump 1 (Fig. 1). A level controller and an automatic vacuum effluent system (Fig. 1) were used to control the reactor working volume at a constant value. The automatic vacuum effluent system consisted of a vacuum reservoir, a vacuum pump, a gasewater segregator, a level sensor, a peristaltic pump (Pump 2), an electromagnetic valve and a power control device (Fig. 1). When the vacuum pump was operated, a negative-pressure condition was created as gas stored in the vacuum reservoir was pumped out, which pushed wastewater to flow into the vacuum reservoir from the bioreactor through the membrane module. As the liquid surface in the vacuum reservoir reached 80% of the maximum height, the effluent stored in the reservoir was drained out by Pump 2. The vacuum pump and Pump 2 worked alternately so that the MLMBR effluent was continuously discharged. Data from the MLMBR were measured every 10 min. The sludge pump, air pump and influent pump were used as final control elements. Their operational values followed the recommendations from the KBES and were implemented through the PC. The MLMBR system was equipped with an airewaterechemicals backwashing system. Fig. 1 shows the

electromagnetic valve that controls water level and air flow. The cleaning scheme of the membrane modules proceeded as follows: the membrane modules were first cleaned with air for 10 min at a flow rate of 500 L/min; next, the modules were subjected to a simultaneous air wash (flow rate 500 L/h) and water wash (flow rate 20 L/min) for 10 min; the modules were further cleaned with chemicals (flow rate 5 L/min) and air (flow rate 500 L/h) for 30 min simultaneously; finally, the membrane modules were backwashed with water for 20 min at a flow rate of 20 L/min. The chemical wash was a mixed solution containing 95% ethanol and 5% sodium hypochlorite. Membrane permeate was used as washing and backwashing water. The water temperature was 20e25  C when cleaning was performed. The process studied was biological wastewater treatment using an MLMBR with simultaneous nitrification and denitrification and enhanced COD removal (Fig. 1). The MLMBR included three zones: anaerobic, anoxic and aerobic. In the aerobic zone, COD removal and nitrification were performed. The alternating anaerobic/anoxic conditions improved the efficiency of nitrate removal by favoring the growth of denitrifying bacteria (DNB). Seed sludge for starting up the MLMBR was obtained from the wastewater treatment facilities (secondary settling tank) of a pharmacy company and a distant noodle factory in Harbin, China. The initial concentrations of seed sludge were approximately 2400 mg/L MLSS in the MLMBR. The raw wastewater was initially stored in a buffer tank and then pumped at a fixed-flow rate of 17.3 and 9.9 L/h through the MLMBR, resulting in an HRT of 8.0 and 14.0 h, respectively. The MLMBR reactor was operated at a temperature of 20  5  C, a dissolved oxygen (DO) content of 1.0e6.0 mg/L, and a pH of 5.5e9.0. Sewage parameters of the influent and the effluent,   including pH, COD, BOD5, TN, NHþ 4 -N, NO3 -N and NO2 -N, were determined according to the Standard Methods for

5269

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 6 6 e5 2 7 8

Examination of Water and Wastewater (American Public Health Association, 1998). Influent and effluent samples were obtained by filtering the mixed liquid through filter paper  (0.45 mm pore size). The concentrations of NHþ 4 -N, NO3 -N and  NO2 -N were quantified using a Hach DR 2400 spectrophotometer with Hach chemical kits.

2.2.

Wastewater characteristics

CSP wastewater was obtained from a local chemical synthesis-based pharmacy company that manufactures intermediates (6-APA, 7-ACA, GCLE), cefazolin, cefoperazone sodium, cephalosporins ampicillin, penicillin G sylvite, amoxicillin, ampicillin sodium and polyethylene oxide third ethylene glycerin. The CSP wastewater was mainly generated from product manufacturing and equipment cleaning and contained a variety of organic and inorganic constituents, such as spent solvents, catalysts, reactants and a small amount of intermediates or products. The CSP wastewater had been treated in an anaerobic digester before it entered the experimental setup. The characteristics of the CSP wastewater are shown in Table 1. In order to obtain the required inflow, the CSP wastewater was diluted with tap water in variable proportions. Compounds with different rates of biodegradability were included in the CSP wastewater. The CSP wastewater containing carbon sources and nitrogen sources was used in order to minimize microbial contamination. Table 1 shows the exact composition of each concentrate. The bench-scale plant was fed with an influent of variable composition and flow rate to simulate the conditions of a fullscale plant. The influent COD, BOD, NHþ 4 -N, and TN profiles are shown in Fig. 2. These resulted in a period of low load involving a low concentration and flow rate and a period of high load. During the experiments, samples were periodically collected from the influent in order to check whether the dosing system was operating as scheduled. The volume of concentrate used was compared to the expected value in each experiment; tests revealed this volume to be highly reproducible, a necessary condition if reliable conclusions are to be derived from a bench-scale plant.

2.3. Architecture of the Knowledge-Based Expert System (KBES) The hardware architecture is shown in Fig. 3. The software for the bench-scale plant control computer, developed in C language, includes graphic monitoring, data backup, PLC supervision and control of key process parameters (DO, HRT and SRT). The Knowledge-Based Expert System (KBES) is at the top of the system architecture. The KBES was developed using G2 (the Real-Time Expert System, Gensym Corporation, 1992) as a tool to develop the real-time expert system, although implementation is also possible using other development tools. The implementation of the ES took advantage of the shell. All functions and features of the KBES were developed using the built-in tools. Knowledge about the process was systematized based on available scientific knowledge and on the practice acquired by the particular system. The KBES acts as the master in a supervisory setpoint control scheme based on a distributed architecture integrated by a supervisor. The KBES is fed with monitored on-line data (pH, DO, aeration, and flow rates) using the data server. Qualitative data (microbiological observations) and discrete data from off-line analyses (COD, BOD, MLVSS, TN and NHþ 4 -N) can also be fed to the KBES. Using rules based on available data, the KBES continuously determines the optimum control required to achieve the required nitrogen and organic matter removal efficiency. Finally, control actions are transmitted to the process computers that actuate on each element of the plant. In the KBES, a set of 397 rules and 52 procedures intended to assist in fault detection, plant maintenance and SND was implemented and validated on the MLMBR. Knowledge was organized in several modules. The utilization of this modular configuration offers some advantages, for example, reusability of the modules and extendibility of the system, making the system able to manage the increasing complexity of artificial intelligence systems. Each module acts as an independent agent by applying numerical algorithms and rules when the situation is normal and a single process parameter is being controlled. Process parameters are usually data-driven through forward chaining when new data reach the real-time database. The main modules used are the Preprocessing, Bioreactor, Pumping System, Settler, Feeding System, COD Removal, Nutrient Removal, Sludge Age Control and Process Simulation

Table 1 e Composition of feed concentrates. Composition COD BOD5 pH NHþ 4 -N TN TP Abamectin GCLE 7-ACA Cefazolin Ceftriaxone

mg/L

Composition

mg/L

Composition

mg/L

1200e9600 500e2500 6e8 50e200 105e400 35e72 25e50 15e58 8e74 32e53 6e17

Cefotaxime 6-APA 7-ADCA PenicillinG sylvite Sporins pull set Amoxicillin Ampicillin sodium Acetone Butyl acetate Metacresol Dichloromethane

11e26 12e45 9e34 17e35 9e24 12e37 7e23 65e136 132e236 78e167 154e257

Toluene Isopropanol Acetic Propionic acid Butyrate Ethanol Butanol Ethyl benzene O-xylene Cl SO2 4

241e350 178e345 35e64 18e36 5e16 31e45 95e137 86e129 45e97 621e896 1253e1789

5270

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 6 6 e5 2 7 8

Fig. 2 e Influent COD, BOD, total nitrogen and ammonia nitrogen.

modules. A brief description of the different modules is summarized below. In the Preprocessing module, the support vector machine (SVM) is used for pattern recognition while the MLMBR is in malfunction. In the Bioreactor module, some rules for the supervision of the operation of these units are established (especially the aeration and probes subsystem behavior). The possible deactivation of local control loops, such as oxygen control, is included in this module. In the Settler module, rules for settler supervision are implemented. These rules mainly monitor the hydraulic load and microorganism load to this unit, the sludge settling ability and the TSS concentration in the effluent to detect early problems in the wateremicroorganism separation. The Pumping System module contains rules to supervise the operation and maintenance schedule of bench-scale plant pumps. It can work with different strategies, such as fixed-flow or fixed-time. The Feeding System module supervises the automatic feeding system, checking the total volume and the concentrations of compounds fed to the pilot plant. The COD Removal module estimates the ratio of food to microorganisms and warns the operator when this ratio may cause problems with plant operation. Some predefined common situations, such as overload or low load, can be detected by the rules, and then the programmed control actions can be applied to minimize future problems. In the Nutrient Removal module, some control algorithms to improve nitrogen removal without affecting phosphorous removal are implemented. In the Process Simulation module, the BP artificial neural network is used to provide quantitative information on the MLMBR. Other modules include rules to maintain a predefined sludge age or to monitor the microbial community for changes in the sludge to detect bulking problems. In these rules and procedures, every measure is checked using different criteria. The measures must be within a predefined range, the rate of change must not be too fast or too slow and the measures must not be in contradiction with

other measures. If the data are considered unreliable, the KBES can deactivate local control loops and establish constant actuation based on normal values. Due to its auto-control and further development in industrial environments, a programmable logic controller (Siemens LOGO! 230RC) was used for data acquisition and for the final control. The PLC collects and sends the data to the computer through RS-232, which makes the exchange possible. The data are pretreated by the support vector machine (SVM) Preprocessing module, which decides whether the MLMBR is malfunctioning. Concerning the optimization of the system, two types of tasks were applied to the MLMBR. The parameters not directly affecting the performance of the MLMBR (T, tank levels, and so on) are auto-controlled by the closed loop of the PLC. The parameters concerning the MLMBR (feed, nutrients, SRT and so on) are introduced by the experimenter. In case of the problem of PC or communication failure, the PLC has an automatic program. It utilizes the last setpoints sent by the PC to manage the MLMBR.

2.4. Support vector machine (SVM) preprocessing module The support vector machine (SVM) is a particular classifier that is based on the margin-maximization principle. The SVM performs the classification by using the separating hyperplane (Amari and Wu, 1999; Ayat et al., 2005; Lv et al., 2005; Shieh and Yang, 2008; Wang et al., 2008; Adankon and Cheriet, 2009). The MLMBR problems are nonlinear, and SVM uses the kernel trick to solve these problems and produce nonlinear boundaries. The essential function of the kernels is to map training nonlinear data into a High Dimensional Space (HDS) via a mapping function. These nonlinear data are represented with a kernel function which describes the inner product in HDS.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 6 6 e5 2 7 8

5271

Fig. 3 e Architecture of the software.

The kernel function is restricted with the Mercer condition (Smola et al., 1998; Wisniewski et al., 1999; Vapnik and Vashist, 2009). The kernel function can be described as below: Kðx; yÞ ¼ fðxÞfðyÞ

(1)

where fðxÞ is a function in low dimension input space to an HDS and x are data points. The decision function is expressed as follows: f ðxÞ ¼

N X

ak yk kðx; yÞ þ bk

(2)

k¼1

where N is the number of data, yk are the target values and ak are the Lagrangian multipliers, which can be found with linear constraints by solving a quadratic programming problem (Yao et al., 2003). The SVMs attempt to find an optimal hyperplane that correctly classifies input data points by dividing them into two classes, negative and positive (Frias-Martinez et al., 2006). In this research, the support vector machine classification was simulated with the LIBSVM tool.

2.5.

Knowledge base (KB)

The most important part of the KBES is the knowledge base. The knowledge base of the KBES comes from two sources: the operator’s experience and mathematical models. Knowledge that comes from the operator’s experience is called heuristic knowledge. The following rules are some examples of heuristic knowledge: Rule 1 (mud pump start-MLSS): If mixture sludge concentration is higher than 13 g/L (MLSS> 13 g/L) Then start mud pump automatically Rule 2 (mud pump shut-MLSS): If sludge load rises to 0.35 kg/(kg MLSS$d)

Then shut down mud pump automatically Rule 3 (mud pump start-sludge load): If sludge load in 5 h is less than 0.3 kg COD/(kg MLSS$d) or in 24 h less than 0.3 kg COD/(kg MLSS$d) of 5 times Then start mud pump automatically Rule 4 (mud pump shut-sludge load): If sludge load rises to 0.35 kg COD/(kg MLSS$d) Then shut down mud pump automatically Rule 5 (DO): If DO<1 mg/L Then increase aeration Rule 6 (operation pressure) If operation pressure >0.045 MPa Then close suction pump, feed pump and open backwash pump 30 min Heuristic knowledge can provide qualitative diagnoses, but it has minimal quantitative information. Process simulation is based on mathematical models of activated sludge processes, so it can provide quantitative predictions. An intelligent model was developed for the MLMBR. The intelligent computer program, based on the back propagation neural network (BPNN) (Fig. 4), can accurately simulate the major process dynamics of the MLMBR. The BPNN model in the knowledge base was established using the Matlab platform. The topological architecture of BPNN illustrated in Fig. 4 shows a three-level network for effluent COD, BOD, total nitrogen and ammonia nitrogen. The input layer includes eight parameters: influent COD, BOD, total nitrogen, ammonia nitrogen, DO, pH, SRT and HRT. The eight parameters played a key control role and were easily quantified. The adopted input parameters were sufficient to describe the system and gave a satisfactory fit with later experimental observations. This is because the input parameters adopted for the BPNN were derived according to the correlation coefficients developed from the SVM. The parameters chosen influence the

5272

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 6 6 e5 2 7 8

Fig. 4 e Topological architecture of the BPNN model.

output maximum. The hidden layer adopted monolayer structure and included seven neurons to ensure that the uncertain weights occupied half of the training sample. This ensured that the networks would have a moderate scale and compact structure and avoided the phenomena of over-fit. The output parameters of the whole network were effluent COD, BOD, total nitrogen and ammonia nitrogen. The hidden layer and the output layer adopted tansig and purelin as activation functions, respectively. Traingdx was utilized to train the network. The algorithm, based on the classical BPNN algorithm, was able to modulate the learning rate and incidental momentum automatically. This avoided the local minimum and accelerated the convergence rate significantly.

To study the efficiency of the MLMBR in different operational modes, two groups of experiments were carried out. In experiments IeIIX, the plant operation was maintained in a fixed way and different operational conditions were tested using an openloop scheme. Data obtained in this group of experiments resulted in the final rules implemented in the KBES. In the second group of experiments, different strategies were implemented and checked using the previously developed KBES. The KBES supervision allowed for the pilot plant to be maintained under proper operation conditions during the experiments. The KBES was able to detect some problems (pump failure, feeding problems, probe malfunctions, equipment maintenance problems, analyzer control and maintenance problems) and carry out correcting procedures when possible.

rate in order to optimize SND and COD removal. Such an HRT is altered in response to the concentrations of TN, NHþ 4 -N, COD and BOD in the effluent. In fact, the HRT used is a function of the amount of TN, NHþ 4 -N, COD and BOD present in the effluent. The highest HRT used was what is recommended for this type of system, 14.0 h, and the lowest was 8.0 h. Changing the HRT also improves the homogeneous distribution of microbial communities in the reactors and increases the efficiency of COD and BOD removal by denitrifying bacteria, allowing for SND and COD removal. In experiments involving a constant HRT, HRTs of 14.0 and 8.0 h usually provided a proper compromise between removal efficiency and costs in the MLMBR, as checked by our system. As in one of the experiments shown, using a fixed SRT is also possible; however, it is more common to use a changing SRT based on the loads in the influent. The SRTs used in our experiments were 10 and 20 days. If necessary, however, the SRT can be increased without limit to respond to high loads in the influent. Thus, if the TN and COD concentrations in the effluent exceed 45 and 50 mg/L, respectively, the SRT increases in order to allow any solids to accumulate in the settler, temporarily increasing their concentration. Under typical operating conditions, the MLMBR is an aerobic reactor that uses a dissolved oxygen setpoint of 2.0e6.0 mg/L. As revealed by some experiments, when the DO setpoint exceeds 4 mg/L, it may result in the increase of the presence of NHþ 4 -N in the MLMBR and therefore exert an inhibitory effect on denitrification. When DO is below 1.0 mg/ L, it may result in SND and COD removal. This problem can be minimized by using a changing dissolved oxygen setpoint of 4.0 and 1.0 mg/L. The rules used in the experiments are summarized in Table 2.

3.1.

3.2.

3.

Results and discussion

Control strategies used

The operational and control strategies used by the KBES were dictated by the specific operational parameters that can be modified at the MLMBR, namely, the HRT, the SRT and the dissolved oxygen setpoint. Each variable was adapted based on the factors described below. Some treatment plants based on the same operational scheme studied in this work use a fixed HRT. However, a preset HRT is established as a function of the influent flow

Experimental results

The efficiency of the MLMBR in different operational modes was studied. Initially, the response of the system in each mode was examined by using an open-loop scheme. If the operational mode in question provided favorable results with regard to removal efficiency, it was then implemented as a control strategy in a close-loop scheme and adopted for subsequent experiments. Table 3 summarizes the operational parameters studied in each experiment.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 6 6 e5 2 7 8

Table 2 e Rules used in the experiments. Modifiable parameter

Supervised variable

Modified setpoint Setpoint

HRT

Effluent TN and COD

HRT

SRT

Effluent TN and COD

SRT

DO

Effluent COD and NHþ 4 -N DO

Values HRT ¼ 8.0 h HRT ¼ 14.0 h SRT ¼ 10 d SRT ¼ 20 d DO ¼ 1.0 mg/L DO ¼ 4.0 mg/L

Fig. 5 shows the proportion of nitrogen, BOD and COD removed from the wastewater and the nitrogen, COD, BOD in the effluent; the contributions of ammonium and other forms of nitrogen to residual nitrogen are also shown. The amount of nitrogen in the effluent was calculated from the results of the analyses for ammonium performed by the CFA and FIA instruments. The integrated combination of each nitrogen form was obtained from the results of all the analyses available (one datum every 10 min), which were subjected to numerical integration. Although increasing SND and COD removal efficiencies was the principal aim of this work, it should be noted that COD was removed virtually completely in all instances e the removal efficiency always exceeded 82%, no matter which control strategy was used.

3.3.

Discussion

The fact that the amount of nitrite and nitrate in the effluent exceeds 50% of the initial nitrite and nitrate in practically all experiments is especially important, confirming that SND and COD removal should be considered in any wastewater treatment scheme. In fact, the measured amounts were significant relative to those of nitrite and nitrate, so any control system to be implemented should take account of this variable. The results obtained from each operational strategy used are discussed below.

3.3.1. Effect of hydraulic retention time (HRT) and sludge residence time (SRT) Experiments II, IIII, VI and VIII were used to examine the response of the MLMBR to an influent of variable composition and flow rate by comparing the removal efficiencies obtained

Table 3 e Operational conditions used in every experiment. Experiment

DO setpoint (mg/L)

HRT setpoint (h)

SRT setpoint (d)

I II III IV V VI VII VIII IX

1.0 1.0 1.0 1.0 4.0 4.0 4.0 4.0 4.0/1.0/4.0/1.0

8.0 8.0 14.0 14.0 8.0 8.0 14.0 14.0 8.0/14.0/8.0/14.0

10 20 10 20 10 20 10 20 20/10/20

5273

with a constant DO and SRT. The MLMBR accomplished the separation of HRT and SRT. When HRT changes and a fixed SRT is used, which is realized by control of the inflow rate, the settler can change the MLSS. This can alter the feed/microorganism ratio and decrease the removal efficiency. Experiment II, which used the least restrictive control conditions, provided a high nitrogen removal efficiency of 74% and COD and BOD removal efficiencies of 81% and 82.7% respectively. Increasing the flow rate and concentration of the influent caused the system to be unable to transform all nitrogen in the influent during the highest load period. Overall, nitrogen, BOD and COD removals were quite efficient; however, the highest concentrations of nitrogen, BOD and COD in the effluent were too high, at 46.8, 393.24, and 147.08 mg/L, respectively. These levels exceed applicable legal limits. Experiment IIII used an additional control strategy based on maintaining a constant SRT. This resulted in a slightly decreased nitrogen removal efficiency (ca. 72%); however, the highest concentration of ammonium nitrogen in the effluent decreased to 9.8 mg/L. BOD and COD removal efficiencies also increased (86.3% and 90.1%, respectively). Experiments V and VI were used to examine the effect of SRT on SND and COD removal. SRT was changed by controlling the rate of raw sludge. The results of Experiment V were compared with those of VI, where a control strategy that altered the SRT as a function of the combined concentration of nitrification and denitrification and COD in the effluent was implemented. The DO setpoint in the MLMBR was 4.0 mg/L. This prevented the anoxic zones from being supplied with an excessive amount of oxygen and hence avoided an inhibitory effect on denitrification. The proportion of nitrogen removed increased from 65% in Experiment V to 68% in Experiment VI, mainly as a result of the decreased SRT. Two effects (HRT and SRT) enhancing SND and COD removal were thus combined.

3.3.2.

Effect of the oxygen setpoint in the MLMBR

Experiments III and VII were used to assess the influence of the dissolved oxygen setpoint in the MLMBR reactors on SND and COD removal efficiencies when all other experimental variables are constant. Experiments III and VII were conducted with a DO setpoint of 1.0 and 4.0 mg/L, respectively, and total nitrogen removal efficiencies of 68% and 59%, respectively, were observed. The experiment using the lowest DO setpoint resulted in the highest removal efficiency; however, the effluent contained increased ammonium nitrogen. These conditions hinder nitrification but substantially favor denitrification. In fact, as can be seen from Fig. 5, ammonium nitrogen contributes strongly to residual nitrogen in the effluent, at the expense of other forms of nitrogen. The total COD removal rate increased with the increase in DO concentration from 1.0 to 4.0 mg/L, but it did not vary as DO concentration increased to 5.0 mg/L. To remove nitrogen from wastewater, an aerobic nitrification followed by anaerobic denitrification is commonly used. The different requirements of nitrifiers and denitrifiers result in the necessity of keeping the processes separate, either in two reactors or in alternating aerobic and anaerobic zones. Previous studies revealed that these two important steps can occur concurrently in a single reactor. In a single reactor with continuous aeration, it is generally recognized that SND occurs

5274

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 6 6 e5 2 7 8

Fig. 5 e Proportion of nitrogen, COD and BOD removed in each experiment and the properties of the effluent.

as a result of a DO concentration gradient distribution within microbial flocs, granules or biofilms due to diffusion limitations. According to microenvironment theory, oxygen may be depleted at significant rates within the granules so that the DO cannot penetrate the entire depth of flocs, granules or biofilms. That is, nitrifiers will preferentially be active on the surface of the flocs, granules or biofilms, whereas the anoxic microzones in the center of flocs, granules or biofilms allow heterotrophic denitrifiers to produce nitrogen gas. It is postulated that a similar mechanism functions in the MLMBR system; as ambient DO concentration was controlled in a certain range, the formation of an anoxic micro-zone inside the flocs allowed for denitrification. In the present study, total nitrogen removal was significantly limited when DO concentration was higher than 4.0 mg/L, presumably due to the absence of an anoxic environment. Under low DO conditions, diffusion limitations may create an anoxic zone within the biological flocs where denitrification can take place. Fig. 5 shows that the proportion of ammonium nitrogen in the effluent was lower with the decrease of DO concentration. Overall, nitrification improved as the oxygen setpoint in the MLMBR was increased; the improvement, however, was insignificant above a DO setpoint of 4.0 mg/L. Denitrification

was favored by a low oxygen concentration in the MLMBR reactors. Based on the results of these three experiments, the DO setpoint used in the aerobic reactors should not exceed 4.0 mg/L.

3.3.3.

Overall improvements obtained with the control system

The overall effect of using all of the control strategies tested can be assessed by comparing the results of Experiments VIII and IX, as the former involved minimal control actions whereas the latter used all the strategies studied. Fig. 6 illustrates the behavior of different variables monitored during Experiments VIII and IX (influent pH, concentrations of nitrogen, ammonium nitrogen, BOD and COD in the effluent, dissolved oxygen). While the characteristics of the influent were identical in both experiments, influent flow rates were different for the different HRTs. The ammonium nitrogen concentration varied markedly between the two experiments. The changes in influent TN, the removal rate and effluent TN with time are illustrated in Fig. 6. The TN of mixed wastewater was 105e400 mg/L. The results of trials showed that the average value of MLMBR effluent TN decreased gradually to 65.75 mg/L. The interception by microfiltration membranes played a central role in the

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 6 6 e5 2 7 8

5275

Fig. 6 e Values for selected monitored variables in experiments VIII and IX and compared treatment efficiencies.

elimination of TN, and the removal efficiency did not vary greatly with influent quality. Fig. 6 shows the changes in þ influent NHþ 4 -N, the removal rate and effluent NH4 -N with þ time. The effluent NH4 -N gradually decreased to 13.93 mg/L in the MLMBR. However, in the KBES-controlled system, the

effluent TN was 45.84 mg/L. And the effluent of NHþ 4 -N was 9.13 mg/L. NHþ 4 -N was one of the most important oxygenconsuming contaminants. The high removal rate showed that biological conversion played an important role in the elimination of NHþ 4 -N. The efficient nitrification in the KBES-

5276

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 6 6 e5 2 7 8

controlled system was due to the proper oxygen level, an appropriate pH value and the relatively long sludge retention time. In the present study, TN removal was significantly limited at DO concentrations higher than 4.0 mg/L, presumably due to the absence of an anoxic environment. In the KBES-controlled system, DO varied between 1.0 and 4.0 mg/L and the aeration rate changed with changes in the influent. The system maintained a relatively high MLSS, leading to a sufficient anoxic environment to maintain a higher denitrification rate. The pH remained fairly constant, no matter what the load, in Experiment VIII. However, pH was strongly influenced by the aeration cycles used in the MLMBR during Experiment IX; it increased coinciding with the aeration periods, possibly as a result of CO2 stripping. SND is effective in maintaining a neutral pH level in the MLMBR. This is important because a narrow optimal pH around 7.5 is known to exist for the nitrifying bacteria. The optimal pH lies between 7.0 and 7.8 for denitrification in the MLMBR studied. With regard to dissolved oxygen, the control strategies used in Experiment VIII ensured proper actuation on the aeration valves to maintain the pre-established conditions. The amount of air needed under high load conditions was significantly greater. The conditions of Experiment VIII were different as a result of the different setpoints prescribed by the supervisory control system. In fact, the MLMBR used large amounts of air in the aerobic zone; this was offset by the low air consumption in the anaerobic and anoxic zones. The KBES prescribed the appropriate DO setpoint to achieve SND and COD removal. To achieve effective SND, it is necessary to use simultaneous and alternating nitrification/denitrification processes in the same reactor and to maintain DO in an appropriate range to keep the microorganisms in direct contact with the wastewater. Finally, the MLMBR used very little air with both oxygen setpoints tested: 1.0 and 4.0 mg/L DO. There were no obvious changes in COD removal and SND around 15e21 h at a DO level of 1.0 mg/L, even though low DO levels may inhibit the activity of bacteria to a certain extent. However, several factors, such as pH, SRT and HRT, influence COD removal and SND. Alternating between anoxic/anaerobic and aerobic conditions is necessary to take advantage of the lag-time between the nitrite oxidizers and the ammonium oxidizers. The lag-time is different for each species of denitrifying bacteria. The durations of the periods of high and low DO should be optimal for effective COD removal, nitrification and denitrification. Fig. 6 shows the ideal DO level curve for one cycle in the MLMBR. Because this test wastewater contains cephalosporin intermediates (6-APA, 7-ACA, GCLE and others), it is difficult to treat. The changes of influent COD, effluent COD and COD removal rates compared with time are illustrated in Fig. 6. The COD concentration in the mixed wastewater varied from 1200 to 9600 mg/L. The average removal rate remained constant at the rate of 90.9% throughout the experiment. Fig. 6 illustrates the changes in BOD, which varied from 500 to 2500 mg/L. The results of trials showed that the average value of MLMBR effluent BOD decreased gradually to 25 mg/L. It has been noted that the KBES-controlled system has a strong anti-shock loading capacity. The convenience and endurance of the system during practical operation and the security of treated

water quality should also be noted. The control strategies selected were beneficial, leading to an increase of autotrophic bacteria and heterogeneous growth and flora specialized for degradation efficiency. The system reacted accurately to changes in influent load to guarantee effluent quality. To fully estimate the performance of the model, both the apparent experimental observations and some of the environmental changes were considered. The performance of the MLMBR is subject to uncertainties, including influent loads, design and operational parameters and receiving-water conditions. In this paper, the SVM and the ES method were combined to assess and analyze the operation of the MLMBR. The general goal of this paper was to develop design procedures that explicitly consider uncertainty and variability, thereby minimizing the need for empirical safety factors and providing quantitative estimates of MLMBR performance. The results showed that the constructed model was able to produce system dynamics in response to environmental changes, such as sudden changes in influent concentrations. At times 14:00e16:00 and 22:00e0:00, the influent had a sudden change and the constructed model provided stable results. In addition to the overall enhancements described above, the profiles for the different nitrogen forms in the effluent reveal a substantial decrease in ammonium nitrogen concentration from Experiment I (10.9 mg/L) to Experiment IX (1.1 mg/L). This may be crucial in meeting legal regulations regarding allowable levels of this species. Off-line measures of phosphorus were analyzed in all the experiments. With the implemented strategies, phosphorus removal efficiency was not significantly modified. In terms of extending the model to other reactors, there are no parameters in the model concerning geometry except for the volume of the reactor, so the reactor geometry does not affect the model output in this study. However, a particular reactor has specific bacterial flora, so the optimized parameters are different. The chosen parameters for the MLMBR were optimized ones. If future researchers want to apply the system to their own system, they need to change the knowledge base and train their own SVM systems. The MLMBR process combines the complete mixing of the activated sludge process for removing dissolved constituents and microfiltration (MF) or ultra-filtration (UF) for removing suspended solids. Some of the optimized parameters, such as HRT, SRT, DO and pH, have some significance for other reactors, such as complete mixing activated sludge processes and biofilm reactors.

4.

Conclusions

The implementation of the KBES in the MLMBR has resulted in the transformation of a classical control system with a fixed behavior into a system adaptable to different problems that could appear in a WWTP. The capacity to deal with these problems is useful for controlling abnormal situations and maintaining the effluent within legal restrictions. Moreover, some situations that could cause long-term problems can be prevented by this system. Although corrective actions could be implemented in a classical control system, they require greater effort and have less flexibility. Furthermore, it is possible to implement classical

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 6 6 e5 2 7 8

algorithms for numerical control using few rules or procedures. KBES real-time implementation also allows the improvement of the overall process performance, adapting the process to meet current conditions. Changes in the control strategies do not require a deep knowledge of the physical implementation of the control system. The developed system allows the final user to implement control strategies and adapt the system to conditions not envisaged in the original design. Overall, the removal efficiency of the MLMBR was substantially increased with available resources while avoiding wasting energy on aeration. It should be noted that the body of control strategies developed for the MLMBR studied may not be suitable for others; in any case, the control system can always adapt the system’s behavior to particular influent conditions. Based on the results of this work, the following conclusions can be drawn: (a) SND and COD and BOD removal rules for a KBES have been established by analyzing the experimental results. (b) A significant improvement in SND and COD and BOD removal efficiency was obtained using the amount of energy needed in each case by adapting the process in real time. The amount of nitrogen removed was increased by 19% compared with common operating conditions. (c) The proposed modular architecture permits the implementation of the KBES on a conventional control system or an SCADA (Supervisory Control and Data Acquisition) e the type of system most frequently used by control systems at full-scale treatment plants. The system also allows for the easy implementation of different operational strategies in order to adapt the system to the actuation variables involved.

Acknowledgments This work was supported by the National Natural Science Foundation of China (No. 50908062), the State Key Laboratory of Urban Water Resource and Environment (No. HITQAK201103), the Special Fund for Basic Scientific Research of Central Colleges (HEUCP101020) and Harbin innovation Science Foundation for Youths (2008RFQXS023). The authors are grateful to the Research Center of Environmental Biotechnology in Harbin Institute of Technology for their technical and logistic assistance.

references

Adankon, M.M., Cheriet, M., 2009. Model selection for the LS-SVM. Application to handwriting recognition. Pattern Recognition 42, 3264e3270. Amari, S., Wu, S., 1999. Improving support vector machine classifiers by modifying kernel functions. Neural Networks 12, 783e789.

5277

American Public Health Association (APHA), 1998. Standard Methods for the Examination of Water and Wastewater, nineteenth ed. Washington, DC. Ayat, N.E., Cheriet, M., Suen, C.Y., 2005. Automatic model selection for the optimization of SVM kernels. Pattern Recognition 38, 1733e1745. Barnett, M.W., 1992. Knowledge-based expert system applications in waste treatment operation and control. ISA Transactions 31, 53e60. Chen, Z.B., Ren, N.Q., Wang, A.J., Zhang, Z.P., Shi, Y., 2008. A novel application of TPAD-MBR system to the pilot treatment of chemical synthesis-based pharmaceutical wastewater. Water Research 42, 3385e3392. Fenu, A., Guglielmi, G., Jimenez, J., Spe`randio, M., Saroj, D., Lesjean, B., et al., 2010. Activated sludge model (ASM) based modelling of membrane bioreactor (MBR) processes: a critical review with special regard to MBR specificities. Water Research 44, 4272e4294. Frias-Martinez, E., Sanchez, A., Velez, J., 2006. Support vector machines versus multi-layer perceptrons for efficient off-line signature recognition. Engineering Applications of Artificial Intelligence 19, 693e704. Graaff, M.S., Vieno, N.M., Kujawa-Roeleveld, K., Zeeman, G., Temmink, H., Buisman, C.J.N., 2011. Fate of hormones and pharmaceuticals during combined anaerobic treatment and nitrogen removal by partial nitritation-anammox in vacuum collected black water. Water Research 45, 375e383. Henze, M., 1997. Trends in advanced wastewater treatment. Water Science and Technology 35, 1e4. Huyskens, C., Brauns, E., Van Hoof, H., Wever, De, 2008. A new method for the evaluation of the reversible and irreversible fouling propensity of MBR mixed liquor. Journal of Membrane Science 323, 185e192. Isaacs, S., Nielsen, M., Hansen, N.P., 1999. STAR solution to variable loads. Water Quality International, 42e44. Jeff, A.R., Paul, M.S., Prakash, N.M., 2000. Application of the membrane biological reactor system for combined sanitary and industrial wastewater treatment. International Biodeterioration and Biodegradation 46, 37e42. Ladiges, G., Mennerich, A., 1996. Application and experience with expert systems for the operation of waste water treatment plants. Water Science and Technology 33, 265e268. Lv, G., Cheng, H., Zhai, H., Dong, L., 2005. Fault diagnosis of power transformer based on multi-layer SVM classifier. Electric Power Systems Research 75, 9e15. Maeda, K., 1989. A knowledge-based system for the wastewater treatment plant. Future Generation Computer Systems 5, 29e32. Magureanu, M., Piroi, D., Mandache, N.B., David, V., Medvedovici, A., Parvulescu, V.I., 2010. Degradation of pharmaceutical compound pentoxifylline in water by nonthermal plasma treatment. Water Research 44, 3445e3453. , J., Petrovic , M., Barcelo, D., 2009. Fate and Radjenovic distribution of pharmaceuticals in wastewater and sewage sludge of the conventional activated sludge (CAS) and advanced membrane bioreactor (MBR) treatment. Water Research 43, 831e841. Reif, R., Omil, S.S.F., Lema, J.M., 2008. Fate of pharmaceuticals and cosmetic ingredients during the operation of a MBR treating sewage. Desalination 221, 511e517. Serra, P., Lafuente, J., Moreno, R., Prada, C., Poch, M., 1993. Development of a real-time expert system for wastewater treatment plants control. Control Engineering Practice 1, 329e335. Shieh, M., Yang, C., 2008. Multiclass SVM-RFE for product form feature selection. Expert Systems with Applications 35, 531e541.

5278

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 6 6 e5 2 7 8

Silva, D.G.V., Urbain, V., Abeysinghe, D.H., Rittmann, B.E., 1998. Advanced analysis of membrane bioreactor performance with aerobiceanoxic cycling. Water Science and Technology 38 (4e5), 505e512. Smola, A.J., Scho¨lkopf, B., Mu¨ller, K., 1998. The connection between regularization operators and support vector kernels. Neural Networks 11, 637e649. Steyer, J., Rolland, D., Bouvier, J., Moletta, R., 1997. Hybrid fuzzy neural network for diagnosis e Application to the anaerobic treatment of wine distillery wastewater in a fluidized bed reactor. Water Science and Technology 36, 209e217. Steyer, J., Buffie`re, P., Rolland, D., Moletta, R., 1999. Advanced control of anaerobic digestion processes through disturbances monitoring. Water Research 33, 2059e2068. Sun, D.D., Khor, S.L., Hay, C.T., Leckie, J.O., 2007. Impact of prolonged sludge retention time on the performance of a submerged membrane bioreactor. Desalination 208 (1e3), 101e112. Szafnicki, K., Bourgois, J., Graillot, D., Benedetto, D.D., Breuil, P., Poyet, J., 1998. Real-time supervision of industrial waste-water treatment plants applied to the surface treatment industries. Water Research 32, 2480e2490. Tambosi, J.L., Sena, R.F., Favier, M., Gebhardt, W., Jose´, H.J., Schro¨der, H.F., Moreira, R.F., 2010. Removal of pharmaceutical

compounds in membrane bioreactors (MBR) applying submerged membranes. Desalination 261, 148e156. Tang, C.J., Zheng, P., Chen, T.T., Zhang, J.Q., Mahmood, Q., Ding, S., Chen, X.G., Chen, J.W., Wu, D.T., 2011. Enhanced nitrogen removal from pharmaceutical wastewater using SBA-ANAMMOX process. Water Research 45, 201e210. Vapnik, V., Vashist, A., 2009. A new learning paradigm: learning using privileged information. Neural Networks 22, 544e557. Wagner, W.P., Otto, J., Chung, Q.B., 2002. Knowledge acquisition for expert systems in accounting and financial problem domains. Knowledge-Based Systems 15, 439e447. Wang, X., Yang, H., Cui, C., 2008. An SVM-based robust digital image watermarking against desynchronization attacks. Signal Processing 88, 2193e2205. Watts, P., Knight, B., 1991. Fault diagnosis in ASPEX: an expert system for the control of the activated sludge process. Engineering Applications of Artificial Intelligence 4, 151e155. Wisniewski, C., Cruz, A.L., Grasmick, A., 1999. Kinetics of organic carbon removal by a mixed culture in a membrane bioreactor. Biochemical Engineering Journal 3, 61e69. Yao, Y., Marcialis, G.L., Pontil, M., Frasconi, P., Roli, F., 2003. Combining flat and structured representations for fingerprint classification with recursive neural networks and support vector machines. Pattern Recognition 36, 397e406.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 7 9 e5 2 8 6

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Removal of organics and nutrients from food wastewater using combined thermophilic two-phase anaerobic digestion and shortcut biological nitrogen removal Fenghao Cui, Seungho Lee, Moonil Kim* Department of Civil & Environmental Engineering, Hanyang University, Sa 3-Dong, Ansan, Gyeonggi-Do, Republic of Korea

article info

abstract

Article history:

A process combining pilot-scale two-phase anaerobic digestion and shortcut biological

Received 30 April 2011

nitrogen removal (SBNR) was developed to treat organics and nutrients (nitrogen and

Received in revised form

phosphorus) from food wastewater. The thermophilic two-phase anaerobic digestion

18 July 2011

process was investigated without adjusting the pH of the wastewater for the pre-

Accepted 25 July 2011

acidification process. The digested food wastewater was treated using the SBNR process

Available online 30 July 2011

without supplemental carbon sources or alkalinity. Under these circumstances, the combined system was able to remove about 99% of COD, 88% of TN, and 97% of TP.

Keywords:

However, considerable amounts of nutrients were removed due to chemical precipitation

Two-phase anaerobic digestion

processes between the anaerobic digestion and SBNR. The average TN removal efficiency of

Shortcut biological nitrogen removal

the SBNR process was about 74% at very low C/N (TCOD/TN) ratio of 2. The SBNR process

Food wastewater

removed about 39% of TP from the digested food wastewater. Conclusively, application of

Pilot plant

the combined system improved organic removal efficiency while producing valuable

C/N ratio

energy (biogas), removed nitrogen at a low C/N ratio, and conserved additional resources (carbon and alkalinity). ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Food waste is one of the main components of municipal solid waste, which induces the emission of odorous compounds, adversely affects the quality of leachate from landfills, and is a source of greenhouse gas (CH4 and CO2) emissions from decomposing food. Anaerobic digestion of organic matter has been used as suitable method for waste treatment, since food waste is highly biodegradable and may be used for the production of energy in the form of biogas (Fehr et al., 2002; Speece, 1996; Verstraete et al., 1996). Nevertheless, the effluent of the anaerobic digester contains high concentration of nitrogen, mostly in the form of ammonia nitrogen, and inorganic phosphorous due to the biotransformation of

proteins and solids (Demirer and Chen, 2005; Cheng and Liu, 2002; Noike et al., 2004). Anaerobic digestion generally has two temperature ranges, mesophilic (30e40
C) and thermophilic (50e60
C). Thermophilic digestion is much faster than mesophilic digestion because biochemical reaction rates increase with temperature. Kim et al. (2002) reported that thermophilic two-phase digestion showed slightly better performance than mesophilic digestion during both the start-up and the long-term periods in their lab-scale study. Furthermore, two-phase anaerobic digestion has several advantages over conventional single-phase systems, such as higher organic degradation rates, methane production rates, and process stability, as well as reduced risk of digester overloading (Yilmazer and

* Corresponding author. Tel.: þ82 31 400 5142; fax: þ82 31 502 5142. E-mail addresses: [email protected] (F. Cui), [email protected] (S. Lee), [email protected] (M. Kim). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.030

5280

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 7 9 e5 2 8 6

Yenigun, 1999; Pohland and Ghosh, 1971; Zhang and Noike, 1991; Wang and Banks, 2003). A two-phase process permits selection and enrichment of different bacteria in each digester by independently controlling the digester operating conditions. Thus, the first phase (acidogenesis) can be operated to optimize acidogenic growth and the second (methanogenesis) to optimize methanogenic growth (Ince, 1998). Conventional biological nitrogen removal processes could be economical for removing high-strength ammonium wastewater (Teichgraeber and Stein, 1994). Moreover, shortcut biological nitrogen removal (SBNR), which utilizes the concept of direct nitrite reduction to nitrogen gas, is an option to enhance cost-effectiveness and removal efficiency at low C/N ratios because theoretically it has lower oxygen demand of about 25% and conserves up to 40% of carbon sources (Abeling and Seyfried, 1992; Surmacz-Gorska et al., 1997). The anaerobic digested food wastewater could offer a critical benefit for SBNR process because of high alkalinity of the wastewater itself which reduces need for alkaline supplements. It has been reported that combined systems (anaerobic digestion and biological nitrogen removal) are economically attractive, as they produce renewable energy and reduce oxygen requirements (Rajagopal et al., 2011). In addition, the SBNR process will achieve several benefits by using digested food wastewater, such as conservation of carbon sources and energy required for aeration. However, there have been few experimental studies of combined processes and hence, our study of a pilot-scale operation is significant. Accordingly, a pilot-scale combined process that includes thermophilic twophase anaerobic digestion and SBNR was used in this study. The primary purpose of this study is to investigate the removal efficiencies of organics and nutrients in the combined system without supplementation with additional sources or pH adjustment. There are several hypothetical

benefits of combined systems: high energy consumption can be compensated by producing methane and saving additional resources (carbon and alkalinity); organic matter and nutrient removal efficiency can be improved; and the digested food wastewater will be appropriate for the SBNR process because the effluent from anaerobic digestion contains biodegradable organic matter and high-strength ammonium.

2.

Materials and methods

2.1.

Process descriptions

2.1.1.

Thermophilic two-phase anaerobic digestion process

A schematic illustration of the combined pilot plant system is shown in Fig. 1. A thermophilic two-phase anaerobic digestion reactor and shortcut biological nitrogen removal reactor were combined in this study. The anaerobic digester consisted of an acid phase for pre-fermentation and a gas phase for anaerobic digestion. Both are continuous stirred-tank reactor (CSTR) tanks without recycling; hence, the hydraulic retention time (HRT) is equal to the solid retention time (SRT). Thermophilic temperatures of 55
C in both the acid and gas phase reactors were maintained by heat exchangers throughout the study. Complete mixing conditions were achieved by mechanical pumping systems. The effective capacities of the prefermentation and anaerobic digestion reactors were 1.2 m3 and 4.2 m3, respectively. Effluent from the anaerobic digestion reactor was first passed through a mixing tank to mix the wastewater with a coagulant (PraestolAlum). A commercial polymer known as Praestol Alum was used with constant dosage of 90 mg/L. The volume of chemical coagulation reactor was 0.28 m3. Rapid stirring at 150 rpm for 5 min was followed by gentle mixing at 30 rpm for 25 min, and the flocs formed were left to settle for 30e60 min. The wastewater and

Fig. 1 e Schematic diagram of the pilot-scale combined thermophilic two-phase anaerobic digestion and SBNR process.

5281

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 7 9 e5 2 8 6

coagulant mixture was delivered to the subsequent storage tank, and most of the suspended material settled out there.

2.1.2.

Shortcut biological nitrogen removal process

After the precipitation process, the digested wastewater was pumped to the shortcut biological nitrogen removal (SBNR) reactor. The SBNR reactor used a post-denitrification process (a four-stage process) in order to enhance nutrient (TN and TP) removal efficiency. This process was comprised of two anoxic reactors (anoxic-1 and anoxic-2) mixed by stirrers at 60 rpm and two aerobic reactors (aerobic-1 and aerobic-2) mixed by aeration. The effective volumes of the anoxic and aerobic reactors were 0.24 m3 and 0.48 m3, respectively. The internal recycles were used between the anoxic and aerobic reactors with a recycle ratio of 200% based on the reactor flow. Nitrite accumulation was implemented in the aerobic-1 reactor. Temperature was maintained at 35
C in the aerobic-1 reactor. Previous researchers have proposed that mesophilic temperatures (between 35
C and 40
C) and pH levels between 7 and 8 are more effective for growth of ammonia oxidizers compared to the growth of nitrite oxidizers, permitting the accumulation of nitrite (Mosquera-Corral et al., 2005; Hellinga et al., 1998). It was difficult to maintain the dissolved oxygen (DO) at concentrations lower than 2 mg/L because it was difficult to achieve complete mixing by aeration in the reactor at such a low level; hence, DO concentration varied between 2 mg/L and 4 mg/L. A three-day HRT was used for the SBNR reactor. A settling tank was placed after the SBNR reactors. Through the settling tank, suspended solids were captured to produce clear effluent and to provide concentrated sludge underflow for recycling and wasting. The sludge recycle ratio was 100% based on influent flow rate. Eight days of SRT were used for the SBNR process. More details about the operating conditions are given in Table 1.

2.2.

Feed and seed

Wastewater from a food waste treatment facility in Ansan, Korea was used as feed throughout the study, after the removal of solid waste from it. The influent characteristics of the food wastewater are presented in Table 2. The SBNR process was seeded with activated sludge taken from a biological nitrogen removal reactor. The acclimation period for the SBNR process was one month by seeding digested food wastewater. Both digested and activated sludge were obtained from a municipal wastewater treatment plant in Ansan, Korea.

Table 1 e Operating conditions for combined two-phase anaerobic and SBNR processes. Reactor Acid phase Gas phase Anoxic-1 Aerobic-1 Anoxic-2 Aerobic-2

Liquid volume (m3)

HRT (day)

DO (mg/L)

1.2 4.2 0.24 0.48 0.24 0.48

3w5 15w20 0.53 1.1 0.53 1.1

<0.1 <0.1 <0.1 2.0w4.0 <0.1 2.0w4.0

Temperature (
C) 55 55 28 35 26 25

2 2 1 1 1 1

Table 2 e Characteristics of the food wastewater used in this study. Characteristic

Average  S.D

pH Temperature (
C) BOD (mg/L) COD (mg/L) SS (mg/L) TS (mg/L) VS (mg/L) T-N (mg/L) T-P(mg/L) NHþ 4  Nðmg=LÞ

3.99  23.6  62136  121950  39848  90266  76905  2595  456  334 

0.1 2.0 14903 23197 11895 15518 17049 589 60 116

S.D: standard deviation.

2.3.

Experimental design

In the two-phase anaerobic digestion process, the organic loading rate (OLR) was increased from 2.5 to 5.2 kg VS/m3/d by changing the HRT. Operational conditions were separated into three different periods (Table 3). In the SBNR process, the temperature was adjusted to 35
C for the aerobic-1 reactor. We did not supplement with any additional carbon sources or alkalinity for denitrification. The C/N ratio was expressed as COD to TN, and the influent C/N ratio fluctuated between 1.2 and 2.64.

2.4.

Analytical methods

The samples analyzed in this study were as follows: food wastewater (storage tank before anaerobic reactor), acid phase (mixed liquid in the pre-fermentation reactor), gas phase (mixed liquid from the anaerobic digestion reactor), digested wastewater (supernatant of storage tank before the SBNR reactor) and final treated water (effluent from settling tank). Total solids (TS) and volatile solids (VS), total suspended solids (SS), biochemical oxygen demand (BOD), and chemical oxygen demand (COD) were analyzed according to standard methods (APHA, 2005). Ammonia, NO2eN, and NO3eN concentrations were determined by the phenate method 4500-NH3 F, method 4500eNO2 B, and method 4500-NO3 B, respectively, from standard methods (APHA, 2005). Total nitrogen (TN) and total phosphorus (TP) were measured using a spectrophotometer (DR/2500, Hach Co., USA). Biogas volume was measured automatically by a gas flow meter every day. Free ammonia (NH3eN) and free nitrous acid (HNO2eN) concentrations were calculated according to the formula which was given by Anthonisen et al (Anthonisen et al., 1976):

3.

Results and discussion

3.1. Performance of thermophilic two-phase anaerobic digestion 3.1.1.

Fates of organics

The corresponding OLR, biogas production, VS, and TCOD are represented in Fig. 2. Biogas production gradually increased during the initial period due to the progress of acclimation,

5282

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 7 9 e5 2 8 6

Table 3 e Operational conditions tested for the two-phase anaerobic digestion process. Period

Initial period Acid phase

HRT (days) OLR (kg VS/m3/d) Operating days

5

Middle period

Gas phase

Acid phase

Gas phase

Acid phase

Gas phase

20

4

17

3

15

2.5 1w43

3.7 44w121

whereas it did not reveal an obvious increase when OLR increased again in the final period. Overall VS were increased during the acid phase, which corresponds with the increased OLR, whereas the VS of the gas phase did not increase clearly when OLR increased. The TCOD concentrations in the acid phase were hardly changed from those of food wastewater, while the gas phase showed very low TCOD concentration during all 3 periods, implying that the acid phase converted organics in food wastewater to readily-biodegradable ones, thus, methanogens converted them to biogas very effectively. Failure problems did not occur in this study. According to these results, we concluded that the performance of the two-phase anaerobic digestion was steady and very efficient (average VS removal was about 81%) up to an OLR of 5.2 kg VS/m3/d. Average solid (TS, VS and SS) concentrations and removal efficiencies during the middle period are shown in Fig. 3. The fermentation process is essential to hydrolyze complex organic matter into simple carbohydrates and fatty acids before methane production. Therefore, the pre-fermentation

Initial period

Biogas production, L/d

40000

Middle period

Final period

30000

20000

10000

0 120 Food wastewater (Average) Acid phase Gas phase

100

Final period

5.2 122w149

reactor, which leads to hydrolysis and fermentation, is theoretically critical. However, the pre-fermentation reactor did not have a notable effect in this study. SS reduction mainly occurred in the anaerobic digestion reactor (54% of removal efficiency) and showed poor removal efficiency for the prefermentation reactor (18% of removal efficiency). VS removal was also mostly achieved by the anaerobic digestion reactor. Uncontrolled acidity, which could cause low pH in the prefermentation reactor, may have impeded the hydrolysis process. The pH of the acid phase was maintained between 3.5 and 4.0. However, the pH increased to 7.7 in the gas phase (Fig. 4). The pH strongly affects production of soluble organics and volatile fatty acid distribution during the fermentation process (Eastman and Ferguson, 1981). In pH sensitivity studies of anaerobic digesters, a pH range of 5.7e6.0 for the acid reactor was recommended to provide a stable and most favorable substrate for the methane reactor (Zoetemeyer et al., 1982; Elefsinotis and Oldham, 1994; Ren et al., 1997). According to the results in this study, the pH of the acid phase was much lower than the optimal conditions. A recycle line was not used for the two-phase anaerobic digestion reactor in this study. It has been reported that recycling the effluent of a methanogenic phase into an acid phase provided a better pH condition for acidogens because the effluent of the methanogenic phase has a considerable amount of alkalinity. Thus, volatile suspended solids (VSS) removal efficiency could be improved in a pH-controlled reactor (Kim, 2001; Gomec et al., 2002). Therefore, it was proposed that more favorable pH conditions could be addressed by the control of the recycle ratio between the acid phase and gas phase to improve pre-fermentation

Acid phase removal efficiency Gas phase removal efficiency

60

100

100 Food wastewater Acid phase Gas phase

40 20

TS, VS and SS, g/L

0 120 Food wastewater (Average)

TCOD, g/L

100

Acid phase Gas phase

80 60 40

80

80

60

60

52% 40

54%

50%

40

20

20

20

21%

15%

0

18% 0

0

0

20

40

60

80

100

120

140

Time, days

Fig. 2 e Effluent (acid phase and gas phase) of the twophase digestion process for treating food wastewater.

160

TS

VS

SS

Fig. 3 e Average TS, VS and SS concentrations and their removal efficiencies for the acid phase and gas phase during the middle period.

Removal efficiency, %

VS, g/L

80

5283

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 7 9 e5 2 8 6

OLR= 2.5 kg/m3/day

OLR= 3.7 kg/m3/day

OLR= 5.2 kg/m3/day 600

4000

500 3000

300

2000

TP, mg/L

TN, mg/L

400

200 1000 100

0

0 Acid phase

Gas phase

Food wastewater Acid phase

Gas phase

2500

10

2000

8

1500

6

1000

4

500

2

pH

NH4+-N, mg/L

Food wastewater

0

0 Food wastewater

Acid phase

Gas phase

Food wastewater Acid phase

Gas phase

Fig. 4 e TN, TP, NHD 4 LN, and pH variations during two-phase anaerobic digestion.

performance without alkalinity supplementation. Nevertheless, the performance of VS removal (higher than 70% in the middle period) was estimated to be better than that of conventional anaerobic digestion because food wastewater is highly biodegradable and has higher soluble organic content than other wastewaters.

the previous section, we noted that most SS were destroyed in the anaerobic digestion reactor, which could be the most probable reason for increased ammonium accumulation. Undesirably acidic conditions may also have restricted the biodegradation of complex organics, thereby generating a limited amount of ammonium in the acid phase.

3.1.2.

3.2.

Performance of the SBNR process

3.2.1.

Nitrite accumulation

Fates of nutrients

TN and TP were investigated as nutrient sources in this study, as were relationships to ammonium and pH variations (Fig. 4). Food wastewater contains high nitrogen and phosphorus levels as well as high organic loading. TN and TP concentrations were increased according to the OLR increase. There were no change on concentrations of TN and TP after both acid and gas phases. The ammonium concentration of the wastewater was increased after the two-phase anaerobic digestion reactor, because the anaerobic degradation of proteinaceous wastes produces ammonia. In particular, the amount of ammonium from the anaerobic digestion reactor was much higher than that from the pre-fermentation reactor. The maximum NHþ 4  N concentration was 2380 mg/L in the anaerobic digestion reactor, implying that most of the ammonium was generated in the anaerobic digestion reactor. The inhibition concentration of NHþ 4  N was much higher than 2380 mg/L, about 3000 mg/L (McCarty and McKinney, 1961). In this experiment, there seemed to be no inhibition by NH3 until OLR increased up to 5.2 kg VS/m3/d because the reactor produced stable biogas and maintained reasonable pH range in gas phase. In

The oxidation of ammonia to nitrite is the first step of nitrification, and is performed by ammonia-oxidizing bacteria. The oxidation of nitrite to nitrate is the second step of nitrification, 250

5

200

4

150

3

100

2

50

1

0

0

Fig. 5 e Concentrations of nitrogen compounds in aerobic1 reactor (all three OLR periods).

5284

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 7 9 e5 2 8 6

3.2.2.

Nitrogen removal

Coagulants were used after the digesting process to remove high loads of SS from the digested effluent. The masses of organics and nutrients were reduced by settling abundant biomass. Specifically, the average TP concentration was 20 mg/L after precipitation. However, organics (mean BOD of 1029 mg/L, mean COD of 2250 mg/L) and TN (mean of 1150 mg/ L) were still at a high level (Table 4). Most of the nitrogen was ammoniumenitrogen (mean of 864 mg/L) with smaller quantities of nitrite and nitrate (mean NO 2  N concentration of 1.89 mg/L, mean NO 3  N concentration of 0.08 mg/L). Table 4 shows the influent and effluent concentrations and overall removal efficiencies for the SBNR process. The removal efficiency of COD was 71%, indicating that the influent includes a certain amount of non-biodegradable COD. The average removal efficiencies of TN and TP could reach 74% and 39%, respectively, through the SBNR process. The average COD to TN ratio was determined to be very low in the digested food wastewater (average C/N ratio of 2). The influent C/N ratio is one of the most critical parameters of the biological nitrogen removal process due to its direct influence on the growth competition between autotrophic and heterotrophic bacteria (Carrera et al., 2004; Van et al., 1993).

Influent TN

Effluent TN

TN removal efficiency 100

1800 1600

1200 60 1000 800 40 600 400

Removal efficiency, %

80

1400

TN, mg/L

which is performed by nitrite-oxidizing bacteria. A key operating factor in the SBNR process is to suppress nitrite oxidation without inhibiting the ammonia oxidation reaction (Schmidt et al., 2003). The proper manipulation of the operating parameters includes maintaining a high temperature (35e40
C), pH levels of 7.5e8.5, and DO at < 1.5 mg/Ldthereby affecting the free ammonia concentration to help in the accumulation of nitrite in the system (Sinha and Annachhatre, 2007). The aerobic-1 reactor was used to accumulate nitrite by maintaining a temperature of 35
C. Since digested food wastewater has a high alkalinity itself, the pH of the wastewater could be maintained between 7.0 and 7.5 without supplemental alkalinity. Fig. 5 shows variations in the nitrogen compounds in the aerobic-1 reactor. Nitrite accumulation was a little higher than that of nitrate, with the ratio   ½NO 2  N=ðNO2  N þ NO3  NÞ of 0.51. It was determined that an average concentration of about 180 mg/L of NO2eN could be accumulated when the average NH3eN was about 2.84 mg/L in this study. Therefore, nitrite accumulation is feasible for the SBNR process of this study.

20

200 0

0 0

20

40

60

80

100

120

140

160

180

Time, days

Fig. 6 e Nitrogen removal efficiencies in the SBNR process.

The variations of influent and effluent nitrogen concentrations and removal efficiencies are showed in Fig. 6. The C/N ratio showed inconsistent variations within the range of 0.8e4 for the digested food wastewater. Influent organics from digested food wastewater increased because the OLR had been raised. Overall TN removal efficiency was increased when our investigation of SBNR process was finished. We concluded that the SBNR process could remove significant nitrogen from digested food wastewater without supplementation with additional carbon and alkalinity at a low mean C/N ratio of 2.

3.3.

Performance of the combined system

The first hypothesis of this study was that the economization of additional supplements besides energy, complemented with biogas production, could offset the energy consumption of the processes. Accordingly, the combined system was estimated without supplementing any additional resources for biological treatment. Under these circumstances, the two-phase anaerobic digestion process could convert 80% of VS to valuable biogas and the SBNR process removed 74% of nitrogen and 39% of phosphorus, respectively, without any supplementation of additional resources (carbon and alkalinity). The second hypothesis addressed the improvement of organic and nutrient removal efficiencies. Fig. 7 shows the

Table 4 e Influent and effluent concentrations and removal efficiencies for the SBNR process (all three OLR periods). Parameter (units: mg/L) pH Alkalinity (as CaCO3) BOD COD SS TN TP NH 4 N NO 3 N NO 2 N

Influent

Effluent

Average

Range

7.65 3102.5 1059 2250 94 1150 20 864 0.08 1.89

7.38e7.85 2994e3195 247e1860 880e4420 59e138 728e1675 3e34 581e1316 0.00e0.60 0.48e4.12

Average e e 88 660 17 304 12.2 15.0 113 220

Removal Range e e 44e160 481e966 3e41 198e441 3.9e24.1 4.5e36.7 72e224 129e310

% e e 92 71 82 74 39 98 e e

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 7 9 e5 2 8 6

Overall removal efficiency, %

100

80

60

40

20

0 SS

COD

TN

Pre-fermentation (Acid phase) Anaerobic digestion (Gas phase)

TP

Under these circumstances, about 80% of VS was removed by the two-phase anaerobic digestion process. Furthermore, 70% of COD was removed by the SBNR process. TN removal efficiency was about 74% when the influent C/N (TCOD/TN) ratio was about 2 in the SBNR process. The application of the combined system improved the removal of organics and nutrients from food wastewater because (1) anaerobic digestion is very effective for high concentrations of organic matter, (2) digested food wastewater produces significant biodegradable organics and high concentrations of ammonium, which provides an optimal influent for the SBNR process, and (3) the SBNR process could remove high-strength ammonium at a low C/N ratio, which eliminates the need for additional carbon sources and alkalinity.

Precipitation SBNR

Fig. 7 e Overall removal efficiencies of SS, COD, TN and TP, including the removal portions of each process in the combined system.

Acknowledgments

overall removal performance for the combined system, including the removal portions of each process. About 99% of organics (COD) were removed, and nitrogen and phosphorus removal were about 88% and 97%, respectively. The combined system had excellent organic removal performance because three processes were involved and food wastewater contains significant biodegradable organics. The anaerobic digestion process converted most of the organic nitrogen into ammoniumenitrogen, which is an easier substrate for biological nitrogen removal. Hoverer, considerable nutrients were removed by precipitation in this study. After anaerobic digestion, a liquid/biosolids mixture with highly concentrated nitrogen and phosphorus is produced and could be considered as valuable fertilizer ingredients. Recovery of these nutrients from the treated wastes is a potential source of revenue, partially offsetting the cost of treatment. The precipitation could be the best way to achieve these objectives at full scale operation. The final hypothesis was that digested food wastewater would be appropriate for the SBNR process. The characteristics of digested food wastewater were highly biodegradable organic contents, high concentrations of ammoniumenitrogen, a high alkalinity level and a low C/N ratio. These characteristics are well suited to the SBNR process, which was designed to remove high-strength ammonium at a low C/N ratio. Moreover, pH conditions can be satisfied without the need to add alkalinity, since digested food wastewater already has a high alkalinity.

references

4.

5285

Conclusions

The removal of organic matters and nutrients (nitrogen and phosphorus) from food wastewater by combining thermophilic two-phase anaerobic digestion and the SBNR process was investigated through pilot-scale experiments in this study. Their removal efficiencies were examined without the supplementation of additional resources or adjustment of pH.

The Authors thank AETeC for the research fund (10-8/8-WT81/1) and Greennes Co., Ltd. for the technical cooperation.

Abeling, U., Seyfried, C.F., 1992. Anaerobic-aerobic treatment of high-strength ammonium wastewater nitrogen removal via nitrite. Water Sci. Technol. 26, 1007e1015. Anthonisen, A.C., Loehr, R.C., Prakasam, T.B.S., Srinath, E.G., 1976. Inhibition of nitrification by ammonia and nitrous acid. J. WPCF 48 (5), 835e852. APHA, 2005. Standard Methods for the Examination of Water and Wastewater, twenty first ed. American Public Health Association, Washington D.C. Cheng, J., Liu, B., 2002. Swine wastewater treatment in anaerobic digesters with floating medium. Trans. ASME 45 (3), 799e805. Carrera, J., Vicent, T., Lafuente, J., 2004. Effect of influent COD/TN ratio on biological nitrogen removal (BNR) from high-strength ammonium industrial wastewater. Process Biochem. 39, 2035e2041. Demirer, G.N., Chen, S., 2005. Two-phase anaerobic digestion of unscreened dairy manure. Process Biochem. 40, 3542e3549. Eastman, J.A., Ferguson, J.F., 1981. Solubilization of particulate organic carbon during the acid phase of anaerobic digestion. J. WPCF 53, 352e366. Elefsinotis, P., Oldham, W.K., 1994. Influence of pH on the acidphase anaerobic digestion of primary sludge. Chem. Technol. Biotechnol. 60, 89e96. Fehr, M., Calcado, M.D.R., Romao, D.C., 2002. The basis of a policy for minimizing and recycling food waste. Environ. Sci. Policy 5, 247e253. Gomec, C.Y., Kim, M., Ahn, Y., Speece, R.E., 2002. The role of pH in mesophilic anaerobic sludge solubilization. J. Environ. Sci. Health A Tox. Hazard. Subst. Environ. Eng. 37 (10), 1871e1878. Hellinga, C., Schellen, A.A.J.C., Mulder, J.W., van Loosdrecht, M.C. M., Heijnen, J.J., 1998. The sharon process: an innovative method for nitrogen removal from ammonium-rich waste water. Water Sci. Technol. 37 (9), 135e142. Ince, O., 1998. Performance of a two-phase anaerobic digestion system when treating dairy wastewater. Water Res. 32 (9), 2707e2713. Kim, M., 2001. Comparative process stability and efficiency of mesophilic and thermophilic anaerobic digestion. Ph.D. dissertation, Vanderbilt University, USA.

5286

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 7 9 e5 2 8 6

Kim, M., Ahn, Y., Speece, R.E., 2002. Comparative process stability and efficiency of anaerobic digestion; mesophilic vs. thermophilic. Water Res. 36, 4369e4385. McCarty, P.L., McKinney, R.E., 1961. Salt toxicity in anaerobic digestion. J. Water Pollut. Control Fed. 33, 399e414. Mosquera-Corral, A., Gonzalez, F., Campos, J.L., Mendez, R., 2005. Partial nitrification in a SHARON reactor in the presence of salts and organic carbon compounds. Process Biochem. 40, 3109e3118. Noike, T., Goo, I.S., Matsumoto, H., Miyahara, T., 2004. Development of a new type of anaerobic digestion process equipped with the function of nitrogen removal. Water Sci. Technol. 49, 173e179. Pohland, F.G., Ghosh, S., 1971. Developments in anaerobic stabilization of organic wastes-the two-phase concept. Envir Lett. 1 (4), 255e266. Rajagopal, R., Rousseau, P., Bernet, N., Be´line, F., 2011. Combined anaerobic and activated sludge anoxic/oxic treatment for piggery wastewater. Bioresour. Technol. 102 (3), 2185e2192. Ren, N.Q., Wang, B.Z., Huang, J.C., 1997. Ethanol-Type fermentation from carbohydrate in high-rate acidogenic reactor. Biotechnol. Bioeng. 54, 428e433. Speece, R.E., 1996. Anaerobic Biotechnology for Industrial Wastewaters. Archae Press, Tennessee, USA. Surmacz-Gorska, J., Cichon, A., Miksch, K., 1997. Nitrogen removal from wastewater with high ammonia nitrogen concentration via shorter nitrification and denitrification. Water Sci. Technol. 36 (10), 73e78.

Schmidt, I., Sliekers, O., Schmidt, M., Bock, E., Fuerst, J., Kuenen, J. G., Jetten, M.S.M., Strous, M., 2003. New concepts of microbial treatment process for the nitrogen removal in wastewater. FEMS Microbiol. Rev. 27, 481e492. Sinha, B., Annachhatre, P.A., 2007. Partial nitrificationoperational parameters and microorganisms involved. Rev. Environ. Sci. Biotechnol. 6, 285e313. Teichgraeber, B., Stein, A., 1994. Nitrogen elimination from sludge treatment reject water-comparison of the steam-stripping and denitrification processes. Water Sci. Technol. 30 (6), 41e51. Van, E.W.J., Arts, P.A.M., Weeselink, B.J., Roberston, L.A., Kuenen, J.G., 1993. Competition between heterotrophic and autotrophic nitrifiers for ammonia in chemostat cultures. FEMS Microbiol. Ecol. 102, 109e118. Verstraete, W., Beer, D., Pena, M., Lettinga, G., Lens, P., 1996. Anaerobic bioprocessing of organic wastes. World J. Microbiol. Biotechnol. 12, 221e238. Wang, Z., Banks, C.J., 2003. Evaluation of a two stage anaerobic digester for the treatment of mixed abattoir wastes. Process Biochem. 38, 1267e1273. Yilmazer, G., Yenigun, O., 1999. Two-phase anaerobic treatment of cheese whey. Water Sci. Technol. 40, 289e295. Zhang, T.C., Noike, T., 1991. Comparison of one-phase and twophase anaerobic digestion in characteristics of substrate degradation and bacterial population level. Water Sci. Technol. 23, 1157e1166. Zoetemeyer, R.J., Van den Heuvel, J.C., Cohen, A., 1982. pH influence on acidogenic dissimilation of glucose in an anaerobic digester. Water Res. 16, 303e313.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 8 7 e5 2 9 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Protein fouling behavior of carbon nanotube/polyethersulfone composite membranes during water filtration Evrim Celik 1, Lei Liu, Heechul Choi* School of Environmental Science and Engineering, Gwangju Institute of Science and Technology (GIST), 261 Cheomdan-gwagiro, 1 Oryong-dong, Buk-gu, Gwangju 500-712, Republic of Korea

article info

abstract

Article history:

The protein fouling of membranes can be related to the hydrophobic and electrostatic

Received 11 November 2010

interactions between proteins and the membrane material; i.e., protein fouling can be

Received in revised form

reduced by changing the membrane properties. In this study, multi-walled carbon nano-

3 June 2011

tube/polyethersulfone (C/P) composite membranes were prepared via the phase inversion

Accepted 28 July 2011

method in order to investigate protein fouling, with bovine serum albumin (BSA) and

Available online 4 August 2011

ovalbumin (OVA) used as the model protein for assessing the protein fouling behavior. The results show that the C/P composite membranes were fouled less compared to the bare

Keywords:

polyethersulfone (PES) membrane at 4 h of static protein adsorption at neutral pH. More-

Composite membrane

over, the irreversible fouling ratio of the C/P composite membranes was less than the bare

Multi-walled carbon nanotubes

PES membrane after 1 h of protein ultrafiltration, and the flux recovery ratio of the C/P

Polyethersulfone

composite membranes was higher than the bare PES membrane after 20 min of DI water

Ultrafiltration

filtration. Based on these results, C/P composite membranes were shown to have the

Protein

potential to alleviate the effects of protein fouling, thereby enabling C/P composite membranes to be used for several runs of protein filtration after simple washing with water. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Ultrafiltration membranes are used extensively as separation technique for macromolecules such as proteins due to their low energy consumption, compact design, and absence of phase change (Li et al., 2004). However, flux decline due to membrane fouling is a major limitation in the use of ultrafiltration membranes, as the nonspecific adsorption and deposition of proteins onto the membrane surface or into pores increases the hydraulic resistance to the flow, severely reducing the permeation flux (Shi et al., 2007). The membrane properties, feed solution chemistry, and process conditions are the main factors influencing membrane fouling (Mo et al., 2008).

The protein fouling of ultrafiltration membranes consists of two stages: (i) a monolayer of protein adsorption on the membrane surface, and (ii) deposition of macromolecules onto the adsorbed monolayer. The magnitude of the adsorbed monolayer is referred to as the irreversible fouling ratio due to the strong adhesion of the monolayer to the membrane, which cannot be removed by cleaning with water. The further deposition of proteins on the monolayer is referred to as the reversible fouling ratio since these proteins are loosely attached and can be removed by cleaning with water (Beier et al., 2007). As such, the protein fouling behavior of ultrafiltration membranes can be related to the hydrophobic and electrostatic interactions between the protein and the membrane

* Corresponding author. Tel.: þ82 62 715 2441; fax: þ82 62 715 2434. E-mail address: [email protected] (H. Choi). 1 Present Address: Department of Environmental Engineering, Suleyman Demirel University, Isparta 32260, Turkey. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.036

5288

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 8 7 e5 2 9 4

(Huisman et al., 2000). Membrane surface chemistry is an important factor affecting the performance of ultrafiltration operations (Reddy and Patel, 2008); it was also found that the hydrophobic nature of membranes induces macromolecules which have hydrophobic regions to be deposited easily (Blanco et al., 2006). Moreover, protein fouling generally occurs on the surface of the ultrafiltration membranes since most proteins are large compared to the pore size of the ultrafiltration membranes (Palacio et al., 2003). For these reasons, it is expected that changing the membrane properties can reduce protein fouling. The primary materials used in membrane technology are polymers, due to their good flexibility, toughness, and separation properties. However, their limitations in terms of chemical and thermal resistances and their relatively short lifetime restrict their wider application (Yang et al., 2006). Hence, research into producing organiceinorganic composite membranes with better separation performance and the ability to adapt to rigorous environmental conditions has been attracting a great deal of interest (Yang et al., 2006). Also, carbon nanotubes (CNTs) have an exceptionally high aspect ratio and high strength and stiffness; their unique properties make them attractive candidates for polymer composites (Gojny et al., 2004). To date, several authors have shown the successful preparation of CNT blended polymeric membranes (Choi et al., 2006; Qui et al., 2009; Celik et al., 2011); though the protein fouling behavior of these membranes has yet to be determined. In an attempt to characterize them, Qui et al. (2009) have shown that CNT blended polysulfone membranes have a lower protein adsorption than polysulfone membranes. And to our knowledge, even though the CNT blended polymeric membranes have lower protein adsorption than polymeric membranes, the protein fouling behavior of CNT blended polymeric membranes has yet to be reported. Based on these considerations and the body of previous research, the objective of this work is to determine the protein fouling behavior of multi-walled carbon nanotube (MWCNT)/polyethersulfone (PES) composite membranes. To determine the protein fouling resistance of the composite membranes, cross-flow permeation tests with bovine serum albumin (BSA) and ovalbumin (OVA) as a model proteins and BSA adsorption onto membrane surfaces were conducted.

2.

Materials and methods

2.1.

Materials

Polyethersulfone (PES; Radel H2000) was kindly supplied by Solvay Korea Co.; Commercial ultrafiltration membranes (UFcommercial), MWCNTs and N-methyl-2-pyrrolidinone (NMP) were purchased from Starlitech Corporation (USA), Hanwha Nanotech. Co. Ltd. (Korea) and SigmaeAldrich (USA), respectively. In addition, BSA having an approximate molecular weight of 68 kDa, molecular size of 14 nm  4 nm  4 nm, and an isoelectric point (IEP) at pH 4.7e4.9 (Nakamura and Matsumoto, 2006) was purchased from Roche (Switzerland). OVA having an approximate molecular weight of 47 kDA and IEP at 4.6 (Morefield et al., 2005) was purchased from Sigma (USA).

Deionized (DI) water was obtained from a water purification system (Synergy, Millipore, USA), having a resistivity of 18.2 mU cm. Then, for the protein adsorption and protein fouling tests, 1.0 mg/mL BSA solution and 0.5 mg/mL OVA solution were prepared using a 10 mM phosphate buffer solution (PBS) at pH 7, unless otherwise stated.

2.2.

CNT functionalization

The full solubility of CNTs in some solvents can be obtained by chemical functionalization. In addition, the chemical functionalization of CNTs improves the interfacial bonding between the polymer matrix and CNTs (Yang et al., 2004). Note that the chemical modification of CNTs in concentrated sulfuric and nitric acids is described elsewhere (Liu et al., 1998). In brief, MWCNTs were refluxed in a 1:3 concentrated sulfuric acid and nitric acid mixture at 100  C for 3 h, followed by washing with DI water and overnight drying at room temperature. The dried MWCNTs were then ultrasonicated in a similar acid mixture at 70  C for 9 h, before being washed until attaining a neutral pH. Finally, the MWCNTs were dried in a vacuum oven at 100  C overnight. The morphology of the CNTs was then analyzed by transmission electron microscopy (TEM; JEM-2100, JEOL, Japan). In brief, a drop of a CNT suspension in ethanol was placed on a copper grid after 10 min ultrasonication and dried under ambient conditions for TEM examination. A BrunauereEmmeteTeller (BET; ASAP 2020, Micromeritics, USA) surface area analysis was then used to investigate modifications in the surface area density and the pore volume of CNTs.

2.3.

Membrane preparation and characterization

The preparation and characterization of the composite membranes were described in our previous publication (Celik et al., 2011). In brief, functionalized MWCNTs were dispersed homogenously in NMP prior to dissolving 20% PES in the blend solution while continuously stirring and heating at 60  C until the solution became completely homogenous. The resultant polymer solution was ultrasonicated to allow a complete release of air bubbles. The blend solution was then casted on a glass plate using a casting knife at room temperature, and the glass plate was immersed in a coagulation bath of DI water. The pristine membranes were peeled off and subsequently rinsed with and stored in DI water until use. Note that membranes marked as C/P-0% refer to the original polyethersulfone membrane and C/P-4% refers to the membranes prepared in a casting solution in which the amount of MWNTs with respect to polyethersulfone was 4% by weight. The hydrophilicity of the composite membranes on the top and bottom surfaces was determined based on their captive bubble contact angle. Contact angles of the membranes were measured using a contact angle goniometer (Model 100, Rame-Hart, USA) by floating a 2 mL air bubble under the surface of the membrane and measuring the contact angle. A minimum of seven water-contact angles at different locations on one surface was averaged to ensure a reliable value was obtained. In addition, 1 g/L PEG or PVP solution was used to determine the molecular weight cut-off (MWCO) of the blend

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 8 7 e5 2 9 4

membranes by solute rejection measurements as described by Mulder (1997). In order to determine the porosity, the membranes stored in DI water were weighed after wiping excessive water with filter papers. Then, the membranes were dried in a vacuum oven at 80  C for 24 h prior to being weighed. The porosity of each membrane was calculated as follows (Zheng et al., 2006): Pð%Þ ¼

Ww  Wd  1000 Ah

(1)

where, P is the membrane porosity, A is the membrane surface area (cm2), h is the membrane thickness (mm), and Ww and Wd are the weights of the wet and dry membranes (g), respectively. In order to minimize the experimental errors, the membrane porosity was measured at least two times and average values were reported.

2.4.

Protein adsorption

BSA was chosen as the protein source for evaluating the adsorptive fouling property of the membranes. In the BSA adsorption experiment, membranes were cut into small pieces and immersed into 1 mg/mL BSA solution and gently shaken at 25  C for 4 h. The pH value of the solution was kept at 7 using PBS or at 3 by using NaOH and HCl, as needed. Then, the membrane coupons were washed with and ultrasonicated in DI water for 2 min. Thereafter, the amount of adsorbed BSA was directly measured using a UVevis spectrometer (UV-mini 1240, Shimadzu, Japan) via its absorbance at 280 nm. The average of at least two measurements was reported.

2.5.

Ultrafiltration experiments

A cross-flow membrane test unit (Fig. 1) with an effective membrane area of 18.56 cm2 was used during the permeation tests. All ultrafiltration experiments were performed at a temperature of 22  1  C, and the permeate flux of the membranes was measured by weighing the permeate on an electronic balance at fixed time intervals. Adsorptive fouling behaviors of the membranes were then determined by filtration of 1 mg/mL BSA or 0.5 mg/mL OVA aqueous solutions in 10 mM PBS at pH 7.0. Note that membranes were initially operated at a 0.41 MPa trans-membrane pressure (TMP) for 4 h using DI water for stabilization; after that, the pressure was reduced to 0.35 MPa and the pure water flux of the virgin

5289

membranes (Jwv) was determined. The permeation flux was calculated using J¼

V ADt

(2)

where, V is the volume of the permeated water (L), A is the effective membrane area (m2), and Dt is the permeation time (h). Next, BSA or OVA filtration was performed at 0.35 MPa for 1 h, and the initial (Jpi) and final protein flux (Jpf) after 1 h of BSA or OVA filtration was determined. The BSA or OVA rejection ratio (R) of the membranes was subsequently calculated using Rð%Þ ¼

  Cp  100 1 Cf

(3)

where, Cp and Cf (mg/L) are the BSA or OVA concentrations of the permeate and feed solutions measured with a UVeVis spectrophotometer at 280 nm, respectively. Finally, the membranes were flushed with DI water with a cross-flow velocity of 28.7 cm/s for 20 min and the pure water flux of the cleaned membranes (Jwp) after BSA or OVA filtration was measured. In order to evaluate the antifouling property of the membranes, the flux recovery ratio (FRR) was calculated using the expression:  FRRð%Þ ¼

Jwp Jwv

  100

(4)

Each experiment was repeated at least two times. In the experiments, all composite membrane samples were prepared from at least two replicate syntheses sets.

3.

Results and discussion

3.1.

CNT functionalization

CNTs are held together as bundles and have very low solubility in most solvents because of their intrinsic van der Waals forces. Moreover, due to the atomically smooth surface of CNTs, interfacial bonding with polymers is limited (Zhu et al., 2004). Hence, effective utilization of the CNTs in polymer composites strongly depends on both the ability to homogeneously disperse CNTs as well as good interfacial bonding (Qian et al., 2000). Hence, chemical modification of the CNT surface and good dispersion is necessary. For this task, CNTs can be opened at their ends and the terminal carbons can be

Fig. 1 e Schematic of the membrane filtration unit.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 8 7 e5 2 9 4

converted to carboxylic groups via oxidation in concentrated sulfuric and nitric acids (Kim et al., 2006). As shown in Fig. 2A and B, raw MWCNTs were 1e3.5 mm in length and most of the tips of MWCNTs were closed, though treatment with a strong acid mixture shortened the lengths of MWCNTs to 50e800 nm. In addition, treatment with the strong acid mixture successfully opened their tips (Fig. 2C,D). BET surface area of the raw MWCNTs and the functionalized MWCNTs were 181.54 m2/g and 174.91 m2/g, respectively. The total pore volume of MWCNTs increased from 0.59 cm3/g to 0.89 cm3/g by functionalization, also indicating the successful tip opening.

3.2.

Characterization of composite membranes

70

bottom top

Contact angle (°)

5290

60

50

40 C/P-0%

The functionalization of CNTs simplifies the fabrication process of polymer nanocomposites and improves the interfacial bonding between the polymer matrix and CNTs (Yang et al., 2004). Note that the possible formations of interfacial bonding by hydrogen bonding interactions between the sulfonic groups of PES and carboxylic groups of functionalized MWCNTs was shown in our previous publication (Celik et al., 2011). The hydrophilicity of the polymer membranes is best determined by the captive bubble method (Ridgway et al., 1999); artifacts could be introduced into the membrane surface during air-drying, and water droplets on a dry membrane surface may cause membrane swelling. Furthermore, the captive bubble method is more relevant for the actual operation conditions because polymer membranes generally operate in a fully hydrated condition (Ridgway et al., 1999). Contact angles of both the top and bottom surfaces of the composite membranes are shown in Fig. 3. The contact angles of the top surface of the composite membranes decreased gradually up to the C/P-2% membrane and did not change between the C/P-2% and C/P-4% membranes (based on

C/P-0.5%

C/P-2%

C/P- 4%

Membrane type

Fig. 3 e Contact angles of the top and bottom surfaces of the C/P composite membranes (average contact angle and standard error of seven replicates are reported).

a one-way ANOVA test). However, the contact angles of the bottom surface of the composite membranes showed a gradual decrease between the C/P-2% and C/P-4% membranes. This result suggests that hydrophilic MWCNTs migrated spontaneously to the membrane surface to reduce the interface energy during the phase inversion process, making the membrane surface hydrophilic (Choi et al., 2006). Moreover, between the C/P-2% and C/P-4% membranes the MWCNTs started positioning mostly in the membrane structure and on the bottom surface of the membrane instead of the top surface of the membrane. The improved hydrophilicity of the membrane can enhance the water permeability by attracting water molecules inside the membrane matrix and facilitate them to pass

Fig. 2 e [A] and [B] transmission electron micrographs of raw MWCNTs; and [C] and [D] functionalized MWCNTs.

5291

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 8 7 e5 2 9 4

through the membrane (Yang et al., 2006). As shown in Table 1, the permeability of the composite membranes increases with 0.5% MWCNT content, though higher amounts of MWCNTs lead to a decrease in the membrane permeability. The increase of hydrophilicity and porosity of the membranes enhances the permeability of the membrane up to 0.5% MWCNT content. Higher MWCNT concentrations form a denser structure in the sublayer of the membrane because of the delayed phase separation with increased viscosity (Han and Nam, 2002), which results in a negative effect on the pore size and permeability. Even though the MWCO of the C/P4% membrane is smaller than the bare PES membrane, the porosities of the two membranes are quite similar and the permeability of the C/P-4% membrane is higher than the bare PES membrane. This higher permeability is probably due to the increased hydrophilicity of the C/P-4% membrane.

3.3.

Protein adsorption of composite membranes

The adsorption of BSA onto C/P composite membranes is demonstrated in Fig. 4. At pH 3, electrostatic attraction occurred since the protein and all composite membranes had opposite charges. Moreover, at pH values below the IEP, the protein denaturation is relatively high (Huisman et al., 2000). Hence, at pH 3 a higher amount of protein adsorption was observed on the membranes. At pH 7, both BSA and composite membrane surfaces were negatively charged, and electrostatic repulsion was dominant between the membrane surfaces and the protein. In this case, protein adsorption decreased with an increasing amount of MWCNTs in the blend solution, at both pH 3 and pH 7, due to the increased hydrophilicity of the composite membranes. And even though the protein adsorption decreased with increasing amounts of MWCNTs in the membrane structure, the protein adsorptions on the C/P-2% and C/P-4% membranes were similar at a 95% confidence level, based on one-way ANOVA tests, probably due to the similar hydrophilicity of these two membranes (Fig. 3).

3.4.

Protein fouling tests

The BSA and OVA rejections of the composite membranes are shown in Table 2. Since BSA is a very big molecule compared to the pores of the composite membranes (Table 1), the BSA rejection rates for all composite membranes were higher than 95%. Even though OVA is a comparatively smaller molecule than BSA, the OVA rejection rates for all composite membranes were also higher than 95%. Note that there is

240 pH 3 pH 7

200

160

120

80

40

0 C/P- 0%

C/P-0.5%

C/P-2%

C/P-4%

Membrane type

Fig. 4 e BSA adsorption on the C/P composite membranes at pH 3 and pH 7 (average BSA adsorption of two replicates are reported).

a slight increase in the BSA and OVA rejection with C/P composite membranes due to the increased hydrophilicity and the higher negative charge density of the composite membranes with MWCNTs. Fluxes of the C/P composite membranes during protein ultrafiltration were used to analyze the protein fouling resistance of the C/P composite membranes (Fig. 5). In the figure, all C/P composite membranes have higher protein filtration fluxes than the bare PES membrane for both BSA and OVA filtration. It is also known that membrane fouling can be influenced by hydrodynamic conditions and the chemical interactions between the foulants and the membranes (Hua et al., 2008). Since all protein fouling tests were conducted under the same hydrodynamic conditions, the different flux profiles of the composite membranes shown here were due to the different surface properties of the composite membranes. The bare PES membrane displayed the lowest flux during the protein fouling tests because the hydrophobic membranes exhibit a higher flux decline than hydrophilic membranes (Fan et al., 2001). Here, the protein permeation fluxes of the C/ P composite membranes showed a similar trend for both BSA and OVA as the hydrophilicity trend of the composite membranes (Fig. 3). Hughes et al. (2006) showed a very sharp flux decline for OVA filtration which is consistent with our findings as the absolute flux for OVA filtration was lower than BSA filtration.

Table 1 e Properties of the C/P composite membranes and commercial UF membrane. Membrane type UF-commercial C/P-0% C/P-0.5% C/P-2% C/P-4%

PWF (L/m2 h) (Celik et al., 2011) 39 12 93 70 58

PWF: Pure water flux; MWCO: Molecular weight cut-off. a Value obtained from company.

MWCO (kDa) (Celik et al., 2011) a

20 26.5 33.1 22.6 24.7

Porosity (%) NA 42 58 52 38

5292

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 8 7 e5 2 9 4

A 100

BSA rejection (%) 95.8  96.6  98.2  97.5 

C/P-0% C/P-0.5% C/P-2% C/P-4%

OVA rejection (%)

1.6 1.9 1.2 1.1

95.2 95.8 98.1 97.0

   

1.8 3.2 3.1 1.2

The total flux loss after 1 h of protein filtration and flux recovery after 20 min of DI water filtration is shown in Fig. 6. The membranes were cleaned by DI water to remove adsorbed BSA or OVA from the membrane surfaces, thereby demonstrating the recovery ability of the membranes fouled by protein. Note that there is a significant difference in the flux recovery after the 20 min DI water cleaning step for both proteins; the flux recovery ratio of the composite membranes increased from 32% and 40% for bare PES membranes to 70% and 80% for the C/P-2% membrane for OVA and BSA filtrations, respectively. The total flux loss, defined as the total protein fouling on the membranes, was calculated using the following equation:   Jwv  Jpf  100 Rt ¼ Jwv

(5)

The C/P composite membranes displayed an antifouling property toward protein fouling, which is shown by the decrease of the total flux loss from 82% and 72% for the bare PES membrane to 66% and 50% for the C/P-2% membrane for OVA and BSA filtrations respectively (Fig. 6). However, the C/P composite membranes did not show a significant difference for total flux loss or flux recovery ratio, at a 95% confidence level by one-way ANOVA test. Hence, reversible (Rr) and irreversible fouling ratios (Rir) were then determined using the following equations in order to analyze the fouling behavior of the composite membranes in further detail.

OVA BSA

40

2

Flux (L/m h)

40 20

C/P-0%

C/P-0.5%

C/P-2%

C/P-4%

C/P-0%

C/P-0.5%

C/P-2%

C/P-4%

B 100

Average and standard error of three replicates are reported.

50

60

0

Flux recovery ratio (%)

Membrane type

OVA BSA

80

Total flux loss (%)

Table 2 e BSA and OVA rejections of the C/P composite membranes.

80 60 40 20 0

Membrane type

Fig. 6 e Flux changes of the C/P composite membranes during BSA and OVA filtration (average flux change and standard error of three replicates are reported).

Rr ¼

  Jwp  Jpf  100 Jwv

(6)

Rir ¼

  Jwv  Jwp  100 Jwv

(7)

The reversible protein fouling ratios increased significantly by increasing the amount of MWCNTs in the membrane structure to 2%, which led to a reduction in the irreversible protein fouling ratio (Table 3) for both of the proteins. The lower irreversible fouling ratio of the C/P-2% membrane is probably due to increased hydrophilicity and good dispersion of the MWCNTs which prevented the direct contact between the protein molecules and the membrane and the protein molecules removed easily by water flushing. However, the irreversible protein fouling ratio increased again for the C/P-4% membrane, which can be explained by the hydrophilicity of the top and bottom surfaces of the composite membranes (Fig. 3). The hydrophilicity of the bottom surface of the C/P-4% membrane increased dramatically, possibly due to the positioning of the MWCNTs on the

30

20

10

Table 3 e Fouling ratios of the C/P composite membranes (average fouling ratios and standard error of three replicates are reported).

0

Membrane Fouling ratios for BSA Fouling ratios for OVA type Rr Rir Rr Rir 0

0.5

2

4

MWCNTs content of the composite membranes (wt%)

Fig. 5 e Absolute fluxes of the C/P composite membranes during protein ultrafiltration.

C/P-0% C/P-0.5% C/P-2% C/P-4%

12.5  15.0  29.4  18.7 

2.2 3.6 1.0 2.9

59.2  40.5  20.6  32.7 

3.0 11.0 4.0 6.3

13.9 19.6 36.7 29.9

   

4.7 5.2 4.4 8.3

68.7 51.6 29.7 39.3

   

6.6 14.8 4.7 1.6

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 8 7 e5 2 9 4

bottom surface and in the membrane structure of the C/P-4% membrane. This irregular positioning of MWCNTs led to a degradation of the surface properties of the C/P-4% membrane compared to C/P-2% membrane. Wang et al. (2005) stated that the reversible protein fouling can be easily removed by water flushing. As such, the lowest total and irreversible fouling ratio of the C/P-2% membrane indicates that protein fouling can be easily removed by simple water washing, thereby enabling the membrane to be used for several runs of protein filtration.

3.5.

Comparison with commercial membrane

Even though the MWCO of the C/P-2% and UF-commercial membranes were similar (Table 1), the pure water flux of the C/P-2% membrane is 77% higher than the pure water flux of the UF-commercial membrane probably due to the hydrophilic MWCNTs on the surface of the C/P-2% membrane. The OVA and BSA rejections was 100% with UFcommercial membrane and 98% with C/P-2% membrane. Comparison of the fouling ratios and flux recovery ratios of UF-commercial membrane and C/P-2% membrane is shown in Fig. 7. Even though the total fouling decreased from 46% and 65% for C/P-2% membrane to 31% and 41% for UFcommercial membrane for BSA and OVA filtrations, respectively, the irreversible fouling ratio of UF-commercial membrane is higher than C/P-2% membrane for both BSA and OVA. The flux recovery ratio increased from 65% and 77% for UF-commercial membrane to 72% and 83% for C/P-2% membrane for OVA and BSA filtrations, respectively. The lower irreversible fouling ratio and higher flux recovery ratio of C/P-2% membrane compared to UF-commercial membrane shows that the protein fouling can be removed easily and C/P-2% membrane can be used for longer periods of protein filtration.

4.

5293

Conclusions

In this paper, the protein fouling of the C/P composite membranes was investigated. From the findings of this study the following conclusions can be drawn: (1) BSA adsorption on C/P composite membranes was 58% less at pH 3 and 72% less at pH 7 compared to the bare PES membrane. (2) The total protein fouling and the irreversible fouling ratio of the C/P composite membranes were less than the bare PES membrane for both BSA and OVA, which indicates the ease of removal of protein fouling by simple water washing for the C/P composite membranes. (3) The flux recovery ratio of the C/P composite membranes was higher than both the bare PES membrane and the commercial ultrafiltration membrane after simple DI water washing for both BSA and OVA, which indicates the possibility of using C/P composite membranes for several runs of protein filtration. However, additional studies including filtration tests with complex water matrix and calcium ion effects on the fouling behavior of the composite membranes will need to be performed in the future.

Acknowledgment This work was supported by the Gwangju Institute of Science and Technology, Korea, through the Basic Research Project and partially supported by the Korean Ministry of Environment, through the Converging technology project.

references OVA BSA

80

(%)

60

40

FRR

20

Rt Rir

UF

-c

om

Rr

C/ P2%

UF -c om

C/ P2%

0

Fig. 7 e Summary of the flux recovery ratio (FRR), the total fouling ratio (Rt), the reversible fouling ratio (Rr), and the irreversible fouling ratio (Rir) of commercial PES ultrafiltration membrane (UF-com) and C/P-2% composite membrane (average fouling ratios of two replicates are reported).

Beier, S.P., Enevoldsen, A.D., Kontogeorgis, G.M., Hansen, E.B., Jonsson, G., 2007. Adsorption of amylase enzyme on ultrafiltration membranes. Langmuir 23, 9341e9351. Blanco, J.F., Sublet, J., Nguyen, Q.T., Schaetzel, P., 2006. Formation and morphology studies of different polysulfone-based membranes made by wet phase inversion process. J. Membr. Sci. 283, 27e37. Celik, E., Park, H., Choi, H., Choi, H., 2011. Carbon nanotube blended polyethersulfone membranes for fouling control in water treatment. Water Res. 45, 274e282. Choi, J.H., Jegal, J., Kim, W.N., 2006. Fabrication and characterization of multi-walled carbon nanotubes/polymer blend membranes. J. Membr. Sci. 284, 406e415. Fan, L., Harris, J.L., Roddick, F.A., Booker, N.A., 2001. Influence of the characteristics of natural organic matter on the fouling of microfiltration membranes. Water Res. 35 (18), 4455e4463. Gojny, F.H., Wichmann, M.H.G., Kopke, U., Fiedler, B., Schulte, K., 2004. Carbon nanotube-reinforced epoxy-composites: enhanced stiffness and fracture toughness at low nanotube content. Compos. Sci. Technol. 64, 2363e2371. Han, M.-J., Nam, S.-T., 2002. Thermodynamic and rheological variation in polysulfone solution by PVP and its effect in the preparation of phase inversion membrane. J. Membr. Sci. 202, 55e61.

5294

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 8 7 e5 2 9 4

Hua, H., Li, N., Wu, L., Zhong, H., Wu, G., Yuan, Z., Lin, X., Tang, L., 2008. Anti-fouling ultrafiltration membrane prepared from polysulfone-graft-methyl acrylate copolymers by UV-induced grafting method. J. Environ. Sci. 20, 565e570. Hughes, D.J., Cui, Z., Field, R.W., Tirlapur, U.K., 2006. In situ threedimensional characterization of membrane fouling by protein suspensions using multiphoton microscopy. Langmuir 22, 6266e6272. Huisman, I.H., Pradanos, P., Hernandez, A., 2000. The effect of proteineprotein and proteinemembrane interactions on membrane fouling in ultrafiltration. J. Membr. Sci. 179, 79e90. Kim, J.A., Seong, D.G., Kang, T.J., Youn, J.R., 2006. Effects of surface modification on rheological and mechanical properties of CNT/epoxy composites. Carbon 44, 1898e1905. Li, X., Zhang, Y., Fu, X., 2004. Adsorption of glutamicum onto polysulphone membrane. Sep. Purif. Technol. 37, 187e198. Liu, J., Rinzler, A.G., Dai, H., Hafner, J.H., Bradley, R.K., Boul, P.J., Lu, A., Iverson, T., Shelimov, K., Huffman, C.B., RodriguezMacias, F., Shon, Y.S., Lee, T.R., Colbert, D.T., Smalley, R.E., 1998. Fullerene pipes. Science 280, 1253e1256. Mo, H., Tay, K.G., Ng, H.Y., 2008. Fouling of reverse osmosis membrane by protein (BSA): effects of pH, calcium, magnesium, ionic strength and temperature. J. Membr. Sci. 315, 28e35. Morefield, G.L., Jiang, D., Romero-Mendez, I.Z., Geahlen, R.L., HogenEschc, H., Hem, S.L., 2005. Effect of phosphorylation of ovalbumin on adsorption by aluminum-containing adjuvants and elution upon exposure to interstitial fluid. Vaccine 23 (12), 1502e1506. Mulder, M., 1997. Basic Principles of Membrane Technology, second ed. Kluwer Academic Publishers, The Netherlands. Nakamura, K., Matsumoto, K., 2006. Properties of protein adsorption onto pore surface during microfiltration: effects of solution environment and membrane hydrophobicity. J. Membr. Sci. 280, 363e374. Palacio, L., Ho, C.-C., Pradanos, P., Hernandez, A., Zydney, A.L., 2003. Fouling with protein mixtures in microfiltration: BSAelysozyme and BSAepepsin. J. Membr. Sci. 222, 41e51.

Qian, D., Dickey, E.C., Andrews, R., Rantell, T., 2000. Load transfer and deformation mechanisms in carbon nanotubepolystyrene composites. Appl. Phys. Lett. 76 (20), 2868e2870. Qui, S., Wu, L., Pan, X., Zhang, L., Chen, H., Gao, C., 2009. Preparation and properties of functionalized carbon nanotube/PSF blend ultrafiltration membranes. J. Membr. Sci. 342, 165e172. Reddy, A.V.R., Patel, H.R., 2008. Chemically treated polyethersulfone/ polyacrylonitrile blend ultrafiltration membranes for better fouling resistance. Desalination 221, 318e323. Ridgway, H., Ishida, K., Rodriguez, G., Safarik, J., Knoell, T., Bold, R., 1999. Biofouling of membranes: membrane preparation, characterization, and analysis of bacterial adhesion. Methods Enzymol. 310, 463e494. Shi, Q., Su, Y., Zhu, S., Li, C., Zhao, Y., Jiang, Z., 2007. A facile method for synthesis of pegylated polyethersulfone and its application in fabrication of antifouling ultrafiltration membrane. J. Membr. Sci. 303, 204e212. Wang, Y., Wang, T., Su, Y., Peng, F., Wu, H., Jiang, Z., 2005. Remarkable reduction of irreversible fouling and improvement of the permeation properties of poly(ether sulfone) ultrafiltration membranes by blending with pluronic F127. Langmuir 21, 11856e11862. Yang, J., Hu, J., Wang, C., Qin, Y., Guo, Z., 2004. Fabrication and characterization of soluble multi-walled carbon nanotubes reinforced P(MMA-co-EMA) composites. Macromol. Mater. Eng. 289, 828e832. Yang, Y., Wang, P., Zheng, Q., 2006. Preparation and properties of polysulfone/TiO2 composite ultrafiltration membranes. J. Polym. Sci. Pol. Phys. 44, 879e887. Zheng, Q.-Z., Wang, P., Yang, Y.-N., Cui, D.-J., 2006. The relationship between porosity and kinetics parameter of membrane formation in PSF ultrafiltration membrane. J. Membr. Sci. 286, 7e11. Zhu, J., Peng, H., Rodriguez-Macias, R.F., Margrave, J.L., Khabashesku, V.N., Imam, A.M., Lozano, K., Barrera, E.V., 2004. Reinforcing epoxy polymer composites through covalent integration of functionalized nanotubes. Adv. Funct. Mater. 14 (7), 643e648.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 9 5 e5 3 0 1

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Supplementation of inorganic phosphate enhancing the removal efficiency of tannery sludge-borne Cr through bioleaching Guanyu Zheng, Lixiang Zhou* Department of Environmental Engineering, College of Resources and Environmental Sciences, Nanjing Agricultural University, Nanjing, Jiangsu Province 210095, China

article info

abstract

Article history:

þ 2þ Four inorganic mineral nutrients including NHþ and soluble inorganic phosphate 4 , K , Mg

Received 27 March 2011

(Pi) were investigated to reveal the potential limiting nutrients for tannery sludge biol-

Received in revised form

eaching process driven by Acidithiobacillus species, and the feasibility of supplementing the

14 July 2011

limiting nutrients to accelerate tannery sludge bioleaching was studied in the present

Accepted 25 July 2011

study. It was found that the concentration of Pi was lower than 3.5 mg/L throughout the

Available online 2 August 2011

whole bioleaching process, which is the most probable restricting nutrient for tannery sludge bioleaching. Further experiments revealed that the deficiency of Pi could seriously

Keywords:

influence the growth of Acidithiobacillus thiooxidans and lower its oxidization capacity for S0,

Bioleaching

and the limiting concentration of Pi for the growth of A. thiooxidans was 6 mg/L. The low

Tannery sludge

concentration of soluble Pi in sludge matrix was resulted from the extremely strong

Nutrients

sorbing/binding capacity of tannery sludge for phosphate. The supplementation of more

Inorganic phosphate

than 1.6 g/L KH2PO4 into tannery sludge bioleaching system could effectively stimulate the growth of Acidithiobacillus species, enhance Cr removal rate and further shorten tannery sludge bioleaching period from 10 days to 7 days. Therefore, inorganic phosphate supplementation is an effective and feasible method to accelerate tannery sludge bioleaching process, and the optimum dosage of KH2PO4 was 1.6 g/L for tannery sludge with 5.1% of total solids. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Cr(III) compounds are extensively used in tanning process to protect leather against microbial degradation, moisture, sweat and so on (Erdem, 2006; Zheng et al., 2009). However, a large amount of chromium still remains in the tannery effluent after tanning process because of low reaction efficiency of Cr(III) with hides and consequently goes into the tannery sludge during sewage treatment process (Esmaeili et al., 2005). As a result, tannery sludge usually contains

high content of Cr(III) (1e4%, wt/wt) and is classified as a hazardous waste by many nations (Zheng et al., 2009). It is therefore urgent to find a suitable method to dispose tannery sludge economically and safely to avoid chromium accumulated being released to environment (Chuan and Liu, 1996) and thus threatening animal and human health (Bartlett, 1991). It is found recently that bioleaching technique is a convincing way to remove Cr(III) from tannery sludge over other physical or chemical methods (Zhou et al., 2005, 2006; Fang and Zhou, 2007; Wang et al., 2007; Zheng et al., 2009).

* Corresponding author. Tel./fax: þ86 25 84395160. E-mail address: [email protected] (L. Zhou). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.031

5296

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 9 5 e5 3 0 1

In this process, Cr(III) can be dissoluted through sludge acidification processes driven by Acidithiobacillus species (Sand et al., 2001; Rohwerder et al., 2003; Rawlings, 2005; Zheng et al., 2009; Wang et al., 2010), mainly Acidithiobacillus ferrooxidans and/or Acidithiobacillus thiooxidans, and the resulting sludge could perfectly maintain its soil conditioning and fertilizing properties for the subsequent land application (Couillard and Mercier, 1993; Fang and Zhou, 2007). To date, although many parameters of tannery sludge bioleaching have been extensively studied to optimize the growth conditions of applied bacteria and thus achieve the maximum efficiency of metal solubilization, such as type of substrates, sulfur concentration, solid content and dissolved carbon dioxide concentration (Zhou et al., 2005, 2006; Fang and Zhou, 2007; Wang et al., 2007; Zheng et al., 2009), litter information is available for the effect of inorganic mineral nutrients on the tannery sludge bioleaching process and the feasibility of enhancing bioleaching efficacy via adjusting inorganic mineral nutrients in sludge. Previous studies focusing on the bioleaching of ores have revealed that other than energy sources and environmental factors including temperature, pH, redox potential and composition of the leaching medium inorganic mineral nutrients also play important roles in regulating the growth of Acidithiobacillus species. For instance, Niemela¨ et al. (1994) found that ammonium amendment (6 mM) could significantly enhance Fe2þ oxidation during bioleaching of a sulfidic black-schist ore, while it is reported by Hugues et al. (2008) that available ammonium ion was critical to both bacterial growth and bioleaching efficiency during continuous bioleaching of a pyrite concentrate in stirred reactors, which resulted from a combination of factors such as less precipitate formation and decreased bacterial attachment to the pyrite surface. Besides, the lack of inorganic phosphate (Pi) may also greatly influence the bioleaching of minerals (Rawlings, 2002), since phosphorus plays an essential part of cell structure and metabolism, forming part of nucleic acids, phospholipids, lipopolysaccharides, nucleotide cofactors, and some proteins, where it is incorporated through posttranslational modification (Vera et al., 2008). Other studies demonstrated that under phosphate starvation circumstances A. ferrooxidans reduced its growth rate and capacity of oxidizing ferrous iron and fixing CO2 (Seeger and Jerez, 1993a, 1993b; Varela et al., 1998), and polyphosphate (poly P) which exists in A. ferrooxidans cells is involved in a functional heavy metal tolerance mechanism for the bacterium (Alvarez and Jezez, 2004). All these results illuminated that the major or minor nutrient requirements, e.g., N, P, K and Mg, have the possibilities of affecting bioleaching processes driven by Acidithiobacillus species, and nutrients amendment might be a feasible approach to enhance bioleaching efficiency. There is a high content of nutrients including N, P and trace metals existing in sludge matrix (Couillard and Mercier, 1993), so it is usually deemed that Acidithiobacillus species could easily meet their nutrients requirement without any extra inorganic nutrients supplements during sludge bioleaching processes. However, not all nutrients in sludge can be readily utilized by Acidithiobacillus species since some nutrients are not present in dissolved inorganic form which can be used by the obligate chemolithotrophic autotrophs (Varela et al., 1998).

In fact, Choi et al. (2009) have revealed that most phosphorus was either biologically bound to microorganisms or physicochemically bound to metals such as Fe and Al in sewage sludge, and soluble phosphorus was very less. Also, Maurer and Boller (1999) reported that very less inorganic phosphate was present in effluent of wastewater treatment plant since most of phosphorus consisting of particulate phosphorus, polyphosphates and organic phosphorus was removed as sludge. Therefore, it is presumed that the amount of some soluble nutrients, such as inorganic phosphate, in tannery sludge might be so low that restrict, to some extent, the growth of Acidithiobacillus species during tannery sludge bioleaching process. Therefore, the objectives of the present study are to (1) investigate the changes of nutrients levels including Pi, NHþ 4, Kþ and Mg2þ during tannery sludge bioleaching processes to identify the potential restricting nutrients for the growth of Acidithiobacillus species, (2) study the behaviors of the restricting nutrients in tannery sludge matrix, and (3) explore the feasibility of enhancing tannery sludge bioleaching efficiency through supplementing the restricting inorganic nutrients into bioleaching systems.

2.

Materials and methods

2.1.

Sludge sample

The tannery sludge was collected from Tannery Sewage Treatment Plant from Fubang Leather Co. Ltd., Zhejing, China and stored at 4  C until use. Freezed-dried sub-sample was first digested according to standard methods and then measured for heavy metals by inductively coupled plasma (ICP) method, total N and total P by persulfate method, total S by turbidimetric method, and organic matter content by hightemperature combustion method (APHA, 2005). Selected physicochemical properties are shown in Table 1.

2.2. Iron (sulfur)-oxidizing bacterium and bioleaching inoculum preparation A. ferrooxidans LX5 (CGMCC No. 0727) and A. thiooxidans TS6 (CGMCC No.0759) obtained from China General Microbiological Culture Collection Center (CGMCC) were cultivated in modified 9 K medium (Silverman and Lundgren, 1959; Zhou et al., 2006; Vera et al., 2008) containing 3.0 g/L of (NH4)2SO4, 0.1 g/L of KCl, 0.5 g/L of K2HPO4 and MgSO4$7H2O, and

Table 1 e Selected physicochemical properties of tannery sludge. pH

Solids Organic Total Total Total Total Total content matter N (%) P (%) S (%) Fe (%) Cr(III) (%) (%) (%)

7.92 5.10 (0.06) (0.21)

43.11 (1.94)

1.68 0.47 5.45 2.11 2.34 (0.06) (0.02) (0.29) (0.13) (0.07)

The mean value of triplicate samples is shown; values within brackets denote standard deviation.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 9 5 e5 3 0 1

Starkey’s medium (Suzuki et al., 1990; Takeuchi and Suzuki, 1994; Fang and Zhou, 2006; Zhou et al., 2006) containing 0.3 g/L of (NH4)2SO4, 3.0 g/L of KH2PO4, 0.5 g/L of MgSO4$7H2O and 0.25 g/L of CaCl2, respectively. The modified 9 K medium and Starkey’s medium autoclaved at 121  C for 15 min were adjusted to pH 2.5 and 3.0 with sulfuric acid, and then spiked with 44.2 g/L of 0.22 mm membrane-filtered FeSO4$7H2O or 10 g/L of elemental sulfur as the energy source, respectively. The inoculums were prepared by culturing these bacteria in 500 mL conical flasks each containing 250 mL of these modified 9 K or Starkey’s medium on a gyratory shaker at 200 rpm and 28  C.

2.3. Nutrients changes during the bioleaching process of tannery sludge Bioleaching of tannery sludge was conducted in three parallel 500 mL conical flasks, each containing 285 mL of tannery sludge, 1.2 g of S0 (Zhou et al., 2005; Zheng et al., 2009), and 15 mL of viable cultures of A. ferrooxidans LX5 and A. thiooxidans TS6 (1:1, v/v) (Zhou et al., 2006). These flasks were incubated in a gyratory shaker at 28  C and 180 rpm. During the incubation, 10 mL of sludge samples were withdrawn from each flask at two days intervals, centrifuged at 19,784  g for 15 min and filtered through 0.45 mm membrane filter. The filtrates were analyzed for Mg2þ and Kþ using inductively coupled plasma-atomic emission spectrometry (ICP-AES, Optival200, USA); inorganic phosphate (Pi) and NHþ 4 in the filtrates were also determined using molybdenum blue method and Nessler’s reagent method, respectively. All experiments were performed in triplicate throughout the present study unless otherwise noted, and the data presented are the mean values of the triplicate samples with standard deviation.

2.4. Effect of inorganic phosphate on the growth of A. thiooxidans TS6 in liquid medium Previous studies have revealed that phosphate starvation could seriously restrict the growth of A. ferrooxidans (Seeger and Jerez, 1993a, 1993b; Varela et al., 1998), while litter information is available for its influence on the growth of A. thiooxidans. Therefore, the present study focuses only on investigating Pi starvation on the growth of A. thiooxidans TS6 in liquid medium. 50 ml of logarithmic growth phase A. thiooxidans TS6 cells cultivated as described above was filtered through 0.45 mm membrane filter to remove elemental sulfur particles and the filtrate was subsequently centrifuged at 8793  g for 10 min to collect bacterial cells which was then resuspend in 100 mL of acidified distilled water (pH ¼ 3.0). This washing procedure was repeated continuously for three times to remove any inorganic phosphate in the medium, after which cells were stored at 4  C for less than 2 h before inoculation. Effect of Pi on the growth of A. thiooxidans TS6 was studied in 250 mL conical flasks containing 0.4 mL (0.4%, v/v) of A. thiooxidans TS6 cells prepared above, 1 g (1%, w/w) of elemental sulfur and 97.6 mL of Pi-deficient Starkey’s medium. Then, 2 mL of autoclaved phosphate stock solutions prepared with KH2PO4 and with the concentration of Pi at 100, 200, 300, 500, 4000 and 8000 mg/L were added to these flasks to

5297

make the final concentration of Pi in the medium at 2, 4, 6, 10, 80, 160 mg/L, respectively. The controls were also performed through adding autoclaved distilled water instead of phosphate stock solutions to make the medium without phosphate. All flasks were adjusted to pH 3.0 with sulfuric acid and then incubated in a gyratory shaker at 28  C and 180 rpm. The loss of water in each flask due to the evaporation was compensated by adding distilled water based on weight loss. During the incubation, samples were withdrawn everyday from flasks and measured for pH and SO24 concentration according to the standard methods (APHA, 2005). The S0 oxidation rate was calculated as the ratio of oxidized S0 in the form of SO24 to the total S (10 g/L) present in the medium.

2.5. Sorption behavior of inorganic phosphate in tannery sludge and effect of inorganic phosphate supplementation on the removal of Cr during the bioleaching of tannery sludge The sorption behavior of Pi on the tannery sludge was studied in a series of 250 mL conical flasks, each containing 100 mL of tannery sludge. Inorganic phosphate in the form of KH2PO4 covering a wide range from 0.4 to 3.2 g/L was added into the flasks. 0.05 g of NaN3 was added to the mixture as microbial growth inhibitor. Then, all flasks were incubated on gyratory shaker at 25  C and 250 rpm for an equilibration period of 24 h as determined from a preliminary study, after which 10 mL of sludge sample was collected from each flask and centrifuged at 13,739  g for 10 min. The supernatant was filtered through 0.22 mm membrane filter and subjected to determine inorganic phosphate using molybdenum blue method. Bioleaching of tannery sludge was also conducted in 500 mL conical flasks as described above. Except 285 mL of tannery sludge, 1.2 g of S0 and 15 mL of viable cultures of A. ferrooxidans LX5 and A. thiooxidans TS6 (1:1, v/v), KH2PO4 in the concentration range of 0.4e3.2 g/L was also added into the flasks. Then all flasks were incubated in a gyratory shaker at 28  C and 180 rpm as described before, during which 7.5 mL of sludge samples were collected at 24 h intervals, centrifuged at 19,784  g for 15 min and filtered through 0.45 mm membrane filter. The filtrates were used to determine pH value and solubilized Cr concentration using inductively coupled plasma-atomic emission spectrometry (ICP-AES, Optival200, USA). Cr solubilization efficiency was calculated as the ratio of the solubilized Cr in the sludge to the total Cr present in sludge.

3.

Results and discussion

3.1. Changes of inorganic nutrients during the bioleaching process of tannery sludge Previous studies have revealed that the essential inorganic nutrients for the growth of A. ferrooxidans and A. thiooxidans included ammonium-nitrogen, phosphate, potassium, magnesium and sulfate, while excess nitrate and chloride could inhibit the growth of the two obligate chemolithotrophic autotrophs (Tuovinen et al., 1971; Niemela¨ et al., 1994; Suzuki et al., 1999). Since elemental sulfur was

5298

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 9 5 e5 3 0 1

employed as the energy source in tannery sludge bioleaching process, sulfate concentration in the system is thus much higher than the limiting concentration (Zhou et al., 2005; Zheng et al., 2009; Wang et al., 2010). Therefore, mainly four þ 2þ and soluble inorganic phosphate ions including NHþ 4 , K , Mg (Pi) were studied to reveal the potential limiting nutrients for tannery sludge bioleaching process. As shown in Table 2, the concentration of NHþ 4 changed in the range from 216.8 mg/L to 392.9 mg/L, higher than 84.8 mg/L which is its concentration in Starkey’s medium, indicating that NHþ 4 in sludge is enough to support the growth of Acidithiobacillus species during tannery sludge bioleaching. The concentrations of Kþ and Mg2þ were 82.4e90.5 mg/L and 112.2e124.7 mg/L, respectively, within the incubation period, both of which are higher than either the limiting concentrations for A. ferrooxidans (Tuovinen et al., 1971) or their concentrations in Starkey’s medium. Among the four inorganic nutrients investigated, the concentration of soluble Pi is the least, which was always less than 0.94 mg/L during the initial 6 days of incubation. Although it slightly increased to 3.50 mg/L in the rest period of bioleaching process probably due to the decrease of sludge pH, its concentration was still far lower than both the concentration present in Starkey’s medium and the concentration range (0.2e4.9 mM) during bioleaching of Black-Schist ore (Niemela¨ et al., 1994). Therefore, inorganic phosphate (Pi) is most probably to be the restricting nutrient for the growth of Acidithiobacillus species during tannery sludge bioleaching process.

3.2. Effect of inorganic phosphate on the growth of A. thiooxidans in Pi-deficient Starkey’s medium The changes of medium pH and S0 oxidation rate during the growth of A. thiooxidans in Starkey’s medium supplemented with different concentration of Pi are displayed in Fig. 1 and Fig. 2, respectively. It was found that when Pi was lower than 6 mg/L the growth of A. thiooxidans was severely and negatively influenced, as exhibiting that both medium pH decrease and S0 oxidation rate were lowered by Pi deficiency. For example, when Pi concentration in the Starkey’s medium was decreased from 6 mg/L to 4 mg/L, S0 oxidation rate achieved by A. thiooxidans within 72 h of incubation was lowered from 25.6% to 18.6%, and the final medium pH correspondingly increased from 0.71 to 1.06. When Pi was completely removed from the Starkey’s medium, the growth of A. thiooxidans was not totally inhibited probably due to the fact that Acidithiobacillus species could assimilate some phosphonates to meet their phosphorus requirements (Vera et al., 2008). However, S0 oxidization rate within the same incubation

period was only 4.2% which is less than 16% of that achieved when more than 6 mg/L of Pi was incorporated into the medium, indicating that the oxidizing capacity of A. thiooxidans for S0 was decreased by more than 84% by the Pi deficiency, and the medium pH was only 1.74 at the end of incubation. On the other hand, the insignificant differences between the treatments amended with more than 6 mg/L of Pi implied that even more Pi supplementation could not further enhance the growth of A. thiooxidans. All these results revealed that the limiting concentration of Pi for the growth of A. thiooxidans was 6 mg/L, below which the bacterial growth would be significantly influenced. Combining the present results with Pi change during tannery sludge bioleaching, it is reasonable to presume that inorganic phosphate is the restricting nutrient for the growth of Acidithiobacillus species during tannery sludge bioleaching process, and the supplementation of Pi might have the potential of effectively enhancing the overall bioleaching efficiency.

3.3. Sorption behavior of inorganic phosphate in tannery sludge matrix Soluble Pi concentration as a function of KH2PO4 added into sludge matrix was plotted in Fig. 3. Obviously, soluble Pi detected in sludge was not linearly increased with the increase of KH2PO4 added until the amount of KH2PO4 exceeded 1.2 g/L, below which no Pi could be detected in sludge supernatant. The batch adsorption isotherm of Pi in tannery sludge matrix was given in Fig. 4. It was found that the amount of Pi adsorbed on tannery sludge increased with an increase in equilibrium Pi concentration and eventually attained a plateau value at high equilibrium Pi concentrations. The adsorption data were plotted according to the Langmuir isotherm model as demonstrated in Fig. 4. Obviously the Pi adsorption isotherms conformed better to the Langmuir equation (R2 ¼ 0.9977), and the maximum binding/sorping capacity of tannery sludge for Pi was 11,177 mg/kg dry sludge. It should also be noted that sludge solid content could influence the equilibrium sorption amount of Pi onto tannery sludge, and the equilibrium sorption amount of Pi would decrease with the reduction of sludge solid content or even increase with the increase of sludge solid content. Previous studies have found that chemical removal of P during wastewater treatment processes results from both precipitation of phosphates with Fe(II), Fe (III), Al(III) and Ca(II) esalts in the forms of Fe(PO4)$2H2O (strengite), Al(PO4)$2H2O (variscite) and Cax(PO4)y(OH)z (apatite), and the adsorption of phosphates on metal-hydroxide/oxide precipitates

Table 2 e Changes of inorganic mineral nutrients during tannery sludge bioleaching. Time (D) 0 2 4 6 8 10

6.92 4.90 3.78 2.57 2.16 1.81

pH

Mg2þ (mg/L)

Kþ (mg/L)

(0.06) (0.11) (0.07) (0.09) (0.02) (0.04)

114.37 (3.87) 122.91 (2.14) 123.34 (5.69) 112.21 (1.37) 127.35 (2.72) 115.34 (3.18)

86.66 (2.95) 90.91 (4.10) 89.92 (4.21) 91.34 (3.95) 91.51 (2.78) 92.19 (2.67)

The mean value of triplicate samples is shown; values within brackets denote standard deviation.

NHþ 4 eN (mg/L) 216.83 392.86 364.24 320.95 344.21 308.79

(11.47) (17.89) (15.63) (18.17) (16.24) (15.38)

Pi (mg/L) 0.34 0.55 0.69 0.94 1.68 3.50

(0.02) (0.04) (0.04) (0.03) (0.05) (0.16)

5299

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 9 5 e5 3 0 1

3.5

200

3

Soluble Pi (mg/L) h

160

pH

2.5 2

Control 2 mg/L 4 mg/L 6 mg/L 10 mg/L 80 mg/L 160 mg/L

1.5 1 0.5

0 0

12

24

36 48 Time (h)

60

72

84

Fig. 1 e Change of medium pH with the time during the growth of A. thiooxidans in synthetic Starkey’s medium with the presence of different concentration of inorganic phosphate (Pi).

mechanisms (Luedecke et al., 1989). Iron precipitation is extensively used in tannery sewage treatment processes (Esmaeili et al., 2005; Erdem, 2006), thus there is usually high content of iron-hydroxide/oxide and iron sulfide precipitates which possess high binding/sorping capacity for Pi in tannery sludge (Smolders et al., 2001; Liao et al., 2009; Choi et al., 2009; Wang et al., 2010), which might be responsible for the strong sorption behavior of Pi onto tannery sludge particles in the present sorption study. However, during tannery sludge bioleaching process, Fe2þ was steadily solubilized out from sludge particles because of the decreased sludge pH values and oxidization effect of Acidithiobacillus species. Then the resulting Fe2þ was oxidized to Fe3þ readily by A. ferrooxidans, which could form iron hydroxysulfate precipitates such as schwertmannite very easily in bioleach solutions (Liao et al., 2009; Wang et al., 2010). Therefore, the soluble phosphates in tannery sludge matrix might be both precipitated with Fe2þ

35

20 15

5 12

24

1.6 2.4 3.2 Addition of KH PO (g/L)

4

and/or Fe3þ and sorbed onto the surface of these alreadyformed secondary iron precipitates, resulting in the low concentration of soluble Pi during tannery sludge bioleaching processes as revealed previously. Furthermore, when the sludge pH value was below 2, the solubility of Fe(PO4)$2H2O steadily increased with pH decrease (Maurer and Boller, 1999), which probably resulted in the increased Pi concentration in sludge matrix at the last several days of tannery sludge bioleaching. Considering the limiting concentration of Pi for the growth of A. thiooxidans was 6 mg/L, the amount of KH2PO4 added to enhance bioleaching efficiency during tannery bioleaching process was calculated to be higher than 1.6 g/L, above which the concentration of Pi in tannery sludge could be higher than the limiting concentration for A. thiooxidans growth. In other words, the dosage of KH2PO4 has to exceed 64% of the maximum binding/sorping capacity of tannery sludge for Pi.

3.4. Effect of inorganic phosphate supplementation on the tannery sludge bioleaching process

12000

10

0

0.8

Fig. 3 e Change of soluble Pi in the sludge after supplementing different amounts of KH2PO4 into tannery sludge.

Sorbed amount (mg/kg)h

25

0

As shown in Fig. 5 and Fig. 6, it took about 10 days by bioleaching systems either without inorganic phosphate

Control 2 mg/L 4 mg/L 6 mg/L 10 mg/L 80 mg/L 160 mg/L

30 S oxidization rate (%) h

80 40

0

0

120

36 48 Time (h)

60

72

84

Fig. 2 e Change of S0 oxidization rate achieved by A. thiooxidans grown in synthetic Starkey’s medium with the presence of different concentration of inorganic phosphate (Pi).

10000 Qe = 4715Ce / (1+0.4233Ce)

8000 6000 4000 2000 0

0

50 100 150 Aqueous concentration (mg/L)

200

Fig. 4 e Fit of adsorption of Pi onto tannery sludge with Langmuir isotherm.

5300

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 9 5 e5 3 0 1

the three treatments supplemented with more than 1.6 g/L of KH2PO4 were not significant. Therefore, inorganic phosphate supplementation is an effective and feasible method to stimulate the growth of Acidithiobacillus species and further enhance Cr removal efficiency during tannery sludge bioleaching process, and the optimum dosage of KH2PO4 was 1.6 g/L for tannery sludge with 5.1% of total solids.

8 Control

7

0.8 g/L

6

pH

1.6 g/L

5

2.4 g/L

4

3.2 g/L

3

4.

2 1 0 0

2

4

6 8 Time (Days)

10

12

Fig. 5 e Change of pH during tannery sludge (with 2.34% of total Cr present) bioleaching process with/without the supplementation of KH2PO4 at 0.8, 1.6, 2.4 and 3.2 g/L.

amendment or with 0.8 g/L KH2PO4 supplementation to decrease sludge pH from 6.94e7.10 to 1.86e1.94 and solubilize more than 90% of tannery sludge-borne Cr, which is usually considered as the end of tannery sludge bioleaching (Zhou et al., 2005). These results were consistent with our previous studies that the sludge-borne Cr would be released or removed drastically when sludge pH declines to 2 or below (Zhou et al., 2005; Fang and Zhou, 2007; Zheng et al., 2009). However, the supplementation of more than 1.6 g/L of KH2PO4 to bioleaching system could significantly enhance bioleaching efficiency, as exhibiting that only 7 days were needed by the three treatments amended with more than 1.6 g/L of KH2PO4 to achieve pH decrease from initial 7.1e7.2 to 1.69e1.86 and more than 91.6% of Cr solubilization. Undoubtedly, three days of bioleaching time was shorten by more than 1.6 g/L of KH2PO4 supplementation in comparison with systems without inorganic phosphate amendment or with only 0.8 g/L of KH2PO4 supplementation, although the differences among

Conclusion

Inorganic mineral nutrients were investigated to reveal the potential limiting nutrients for tannery sludge bioleaching process driven by Acidithiobacillus species. It was found that inorganic phosphate (Pi) was the most probable restricting nutrient for the growth of Acidithiobacillus species due to its low concentration throughout the whole bioleaching process. Further experiments revealed that the deficiency of inorganic phosphate could seriously influence the growth of A. thiooxidans and lower its oxidization capacity for S0, and the limiting concentration of Pi for the growth of A. thiooxidans was 6 mg/L. Sorption test indicated that the low concentration of inorganic phosphate in sludge matrix was resulted from the extremely strong sorping/binding capacity of tannery sludge for phosphate. The supplementation of more than 1.6 g/L of KH2PO4 into tannery sludge bioleaching system could effectively stimulate the growth of Acidithiobacillus species and further enhance the removal rate of sludge-borne Cr. Therefore, inorganic phosphate supplementation is an effective and feasible method to accelerate tannery sludge bioleaching process, and the optimum dosage of KH2PO4 was 1.6 g/L for tannery sludge with 5.1% of total solids.

Acknowledgments This study was supported jointly by the 863 Program of China (2009AA06Z317) and National Natural Scientific Foundation of China (40930738).

Solubilization of Cr (%)g

100

references

80 60 Control

40

0.8 g/L 1.6 g/L

20

2.4 g/L 3.2 g/L

0 0

2

4

6 8 Time (Days)

10

12

Fig. 6 e Change of Cr solubilization during tannery sludge (with 2.34% of total Cr present) bioleaching process with/ without the supplementation of KH2PO4 at 0.8, 1.6, 2.4 and 3.2 g/L.

Alvarez, S., Jezez, C.A., 2004. Copper ions stimulate polyphosphate degradation and phosphate efflux in Acidithiobacillus ferrooxidans. Appl. Environ. Microbiol. 70 (9), 5177e5182. APHA, 2005. Standard Methods for the Examination of Water and Wastewater, 21st ed. Amer. Public Health Assoc., Washington, D.C. Bartlett, R.J., 1991. Chromium cycling in soils and water: links, gaps, and methods. Environ. Health Persp. 92, 17e24. Choi, H., Choi, C., Lee, S., 2009. Analyses of phosphorus in sewage by fraction method. J. Hazard. Mater. 167 (1e3), 345e350. Chuan, M.C., Liu, J.C., 1996. Release behavior of chromium from tannery sludge. Water Res. 30 (4), 932e938. Couillard, D., Mercier, G., 1993. Removal of metals and fate of N and P in the bacterial leaching of aerobically digested sewage sludge. Water Res. 27 (7), 1227e1235.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 2 9 5 e5 3 0 1

Erdem, M., 2006. Chromium recovery from chrome shaving generated in tanning process. J. Hazard. Mater. 129 (1e3), 143e146. Esmaeili, A., Mesdaghi, A., Vazirinejad, R., 2005. Chromium (III) removal and recovery from tannery wastewater by precipitation process. Am. J. Appl. Sci. 2 (10), 1471e1473. Fang, D., Zhou, L.X., 2007. Enhanced Cr bioleaching efficiency from tannery sludge with coinoculation of Acidithiobacillus thiooxidans TS6 and Brettanomyces B65 in an air-lift reactor. Chemosphere 69 (2), 303e310. Fang, D., Zhou, L.X., 2006. Effect of sludge dissolved organic matter on oxidation of ferrous iron and sulfur by Acidithiobacillus ferrooxidans and Acidithiobacillus thiooxidans. Water Air Soil Pollut. 171 (1e4), 81e94. Hugues, P., Joulian, C., Spolaore, P., Michel, C., Garrido, F., Morin, D., 2008. Continuous bioleaching of a pyrite concentrate in stirred reactors: population dynamics and exopolysaccharide production vs. bioleaching performance. Hydrometallurgy 94 (1e4), 34e41. Liao, Y., Zhou, L., Bai, S., Liang, J., Wang, S., 2009. Occurrence of biogenic schwertmannite in sludge bioleaching environments and its adverse effect on solubilization of sludge-borne metals. Appl. Geochem. 24 (9), 1739e1746. Luedecke, C., Hermanowicz, S.W., Jenkins, D., 1989. Precipitation of ferric phosphate in activated sludge: a chemical model and its verification. Water Sci. Technol. 21 (4e5), 325e337. Maurer, M., Boller, M., 1999. Modelling of phosphorus precipitation in wastewater treatment plants with enhanced biological phosphorus removal. Water Sci. Technol. 39 (1), 147e163. Niemela¨, S.I., Riekkola-Vanhanen, M., Sivela¨, C., Viguera, F., Tuovinen, O.H., 1994. Nutrient effect on the biological leaching of a black-schist ore. Appl. Environ. Microbiol. 60 (4), 1287e1291. Rawlings, D.E., 2002. Heavy metal mining using microbes. Annu. Rev. Microbiol. 56, 65e91. Rawlings, D.E., 2005. Characteristics and adaptability or iron- and sulfur-oxidizing microorganisms used for the recovery of metals from minerals and their concentrates. Microb. Cell Factories 4, 1e13. Rohwerder, T., Gehrke, T., Kinzler, K., Sand, W., 2003. Bioleaching review part A: progress in bioleaching: fundamentals and mechanisms of bacterial metal sulfide oxidation. Appl. Microbiol. Biotechnol. 63 (3), 239e248. Sand, W., Gehrke, T., Jozsa, P.-G., Schippers, A., 2001. (Bio) chemistry of bacterial leaching - direct vs. indirect bioleaching. Hydrometallurgy 59 (2e3), 159e175. Seeger, M., Jerez, C.A., 1993a. Response of Thiobacillus ferrooxidans to phosphate limitation. FEMS Microbiol. Rev. 11 (1e3), 37e41. Seeger, M., Jerez, C.A., 1993b. Phosphate-starvation induced changes in Thiobacillus ferrooxidans. FEMS Microbiol. Lett. 108 (1), 35e41.

5301

Silverman, M.P., Lundgren, D.G., 1959. Studies on the chemoautotrophic iron bacterium Ferrobacillus ferrooxidans. I. An improved medium and a harvesting procedure for securing high cell yields. J. Bacterial 77, 642e647. Smolders, A.J.P., Lamers, L.P.M., Moonen, M., Zwaga, K., Roelofs, J. G.M., 2001. Controlling phosphate release from phosphateenriched sediments by adding various iron compounds. Biogeochemistry 54 (2), 219e228. Suzuki, I., Takeuchi, T.L., Yuthasastrakosol, T.D., Oh, J.K., 1990. Ferrous iron and sulfur oxidation and ferric iron reduction activities of Thiobacillus ferrooxidans are affected by growth on ferrous iron, sulfur, or a sulfide ore. Appl. Environ. Microbiol. 56 (6), 1620e1626. Suzuki, I., Lee, D., Mackay, B., Harahuac, L., Oh, J., 1999. Effect of various Ions, pH, and osmotic pressure on oxidation of elemental sulfur by Thiobacillus thiooxidans. Appl. Environ. Microbiol. 65 (11), 5163e5168. Takeuchi, T.L., Suzuki, I., 1994. Effect of pH on sulfite oxidation by Thiobacillus thiooxidans cells with sulfurous acid or sulfur dioxide as a possible substrate. J. Bacteriol. 176 (3), 913e916. Tuovinen, O.H., Niemela, S.I., Gyllenberg, H.G., 1971. Effect of mineral nutrients and organic substances on the development of Thiobacillus ferrooxidans. Biotechnol. Bioeng. 13 (4), 517e527. Varela, P., Levica´n, G., Rivera, F., Jerez, C.A., 1998. An immunological strategy to monitor in situ the phosphate starvation state in Thiobacillus ferrooxidans. Appl. Environ. Microbiol. 64 (12), 4990e4993. Vera, M., Pagliai, F., Guiliani, N., Jerez, C.A., 2008. The chemolithoautotroph Acidithiobacillus ferrooxidans can survive under phosphate limiting conditions by the expression of a CP lyase operon that allows it to grow on phosphonates. Appl. Environ. Microbiol. 74 (6), 1829e1835. Wang, Y.S., Pan, Z.Y., Lang, J.M., Xu, J.M., Zheng, Y.G., 2007. Bioleaching of chromium from tannery sludge by indigenous Acidithiobacillus thiooxidans. J. Hazard. Mater 147 (1-2), 319e324. Wang, S., Zheng, G., Zhou, L., 2010. Heterotrophic microorganism Rhodotorula mucilaginosa R30 improves tannery sludge bioleaching through elevating dissolved CO2 and extracellular polymeric substances levels in bioleach solution as well as scavenging toxic DOM to Acidithiobacillus species. Water Res. 44 (18), 5423e5431. Zheng, G., Zhou, L., Wang, S., 2009. An acid-tolerant heterotrophic microorganism role in improving tannery sludge bioleaching conducted in successive multibatch reaction system. Environ. Sci. Technol. 43 (11), 4151e4156. Zhou, L.X., Fang, D., Wang, S.M., Wong, J.W.C., Wang, D.Z., 2005. Bioleaching of Cr from tannery sludge: the effects of initial acid addition and recycling of acidified bioleached sludge. Environ. Technol. 26 (3), 277e284. Zhou, S.G., Zhou, L.X., Wang, S.M., Fang, D., 2006. Removal of Cr from tannery sludge by bioleaching method. J. Environ. Sci. 18 (5), 885e890.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 0 2 e5 3 1 2

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Rapid free chlorine decay in the presence of Cu(OH)2: Chemistry and practical implications Caroline K. Nguyen a,*, Kim A. Powers a, Meredith A. Raetz a, Jeffrey L. Parks a, Marc A. Edwards b a b

Virginia Tech, 418 Durham Hall, Blacksburg, VA 24061, USA Virginia Tech, 407 Durham Hall, Blacksburg, VA 24061, USA

article info

abstract

Article history:

A rapid reaction between free chlorine and the cupric hydroxide [Cu(OH)2] solids

Received 24 March 2011

commonly found on pipe walls in premise plumbing can convert free chlorine to chloride

Received in revised form

and rapidly age Cu(OH)2 to tenorite (CuO). This reaction has important practical implica-

5 July 2011

tions for maintaining free chlorine residuals in premise plumbing, commissioning of new

Accepted 29 July 2011

copper pipe systems, and maintaining low levels of copper in potable water. The reaction

Available online 5 August 2011

stoichiometry between chlorine and Cu(OH)2 is consistent with formation of CuO through a metastable Cu(III) intermediate, although definitive mechanistic understanding requires

Keywords:

future research. Natural levels of silica in water (0e30 mg/L), orthophosphate, and higher

Copper

pH interfere with the rate of this reaction. ª 2011 Elsevier Ltd. All rights reserved.

Chlorine Decay Aging New construction

1.

Introduction

1.1.

Copper aging and interfering ions

The presence of cupric hydroxide (Cu(OH)2) on pipe walls in relatively new plumbing can maintain high levels of soluble copper in potable water (Edwards et al., 1996, 2001; Lagos et al., 2001; Schock et al., 1995) and in rare cases can contribute to gastrointestinal upset from ingesting elevated copper (Craun and Calderon, 2001; Pizarro et al., 2001). For this reason,

copper concentrations in drinking water are regulated by the United States Environmental Protection Agency’s (USEPA) Lead and Copper Rule (LCR) (USEPA, 1991). As the relatively soluble Cu(OH)2 is converted to less soluble tenorite (CuO) and other solids, soluble copper concentrations can decrease markedly (Edwards et al., 1996, 2001; Hidmi and Edwards, 1999; Lagos et al., 2001; Patterson et al., 1991; Schock et al., 1995). The transition from Cu(OH)2 to CuO proceeds more rapidly at higher pH or in warmer water (Hidmi and Edwards, 1999). However, naturally occurring

Abbreviations: AWWA, American Water Works Association; Al(OH)3, aluminum hydroxide; CaCO3, calcium carbonate; CSMR, chloride-to-sulfate mass ratio; Cl2, chlorine; CuO, tenorite; Cu(OH)2, cupric hydroxide; DI, deionized; Fe(OH)3, ferric hydroxide; HCl, hydrochloric acid; HNO3, nitric acid; HOCl, hypochlorous acid or free chlorine; ICP-ES, inductively coupled plasma emission spectroscopy; NOM, natural organic matter; Na2SiO3, sodium silicate; QA/QC, quality assurance/quality compliance; USEPA, United States Environmental Protection Agency; SiO2, silicon dioxide; XRD, X-ray diffraction. * Corresponding author. Tel.: þ1 301 206 8141; fax: þ1 301 206 8057. E-mail addresses: [email protected] (C.K. Nguyen), [email protected] (M.A. Raetz), [email protected] (J.L. Parks), [email protected] (M.A. Edwards). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.039

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 0 2 e5 3 1 2

organic materials (NOM) and other constituents in water can stop the beneficial transition from Cu(OH)2 to CuO (Edwards and Sprague, 2001; Edwards et al., 2001). Silica is another important constituent in water that can sorb to hydroxide solid surfaces and potentially interfere with transitions from one oxide phase to another (Anderson and Benjamin, 1985; Davis et al., 2002; Hingston and Raupach, 1967; Sigg and Stumm, 1981), although prior research has not examined implications of silica for aging of Cu(OH)2.

1.2.

5303

(2e4 mg/L Cl2) have been shown to initiate copper pitting (Marshall, 2004; Rushing, 2002). These and other practical observations make it important to study possible interactions between free chlorine and the Cu(OH)2 solids that can be present as scale on the wall of new copper pipes. The objectives of this research were to 1) examine practical aspects of chemical reactions between free chlorine and Cu(OH)2 solids, and 2) determine the role of silica, pH, and orthophosphate on chlorine decay rates in the presence of Cu(OH)2 solids and in new copper pipe.

Chlorine and control of microbes in copper systems

Free chlorine disinfectant is commonly added to drinking water at low levels to inactivate bacteria, and in some localities, high levels may be dosed (i.e., “superchlorination” or “shock chlorination”) to commission new plumbing systems (Table 1). Occasionally, it has been reported that the presence of chlorine can dramatically reduce copper release to water, presumably due to inactivation of corrosion inducing microbes (Bremer et al., 2001; Cantor, 2009; Edwards et al., 2000), although it has also been suspected that abiotic reactions between copper pipe and chlorine might also be occurring (Edwards et al., 2000). These abiotic reactions might also occur during superchlorination for commissioning and disinfection of pipe systems before use, resulting in rapid chlorine decay in copper tube. This has prevented some building plumbing systems from passing requirements to maintain chlorine residuals above target levels following stagnation in the pipe (BOCA, 1997; Edwards et al., 2011; International Association of Plumbing and Mechanical Officials, 1997; International Code Council, 2000; SBCCI, 1988). In at least one case, replacement of the copper plumbing with plastic pipe was required before the plumbing system could pass the test. Even for copper pipe systems that pass the superchlorination procedure, there are concerns that the very high chlorine concentrations (25e300 mg/L Cl2) could damage pipes, especially given that much lower chlorine levels

Table 1 e Overview of practical issues related to copper pipe and chlorine interactions. Practical Issue Microbes/ Legionella Copper pitting Copper leaching

Commissioning

Role of Chlorine in Issue High chlorine inactivates Legionella and other harmful bacteria (including those involved in some types of corrosion) Higher chlorine can sometimes initiate copper pittinga Higher chlorine can increase copper leachingb, but in other cases dramatically decrease copper concentrations even after a single dosec High chlorine decay rates can cause a failure to meet minimum residual requirements to pass new building commissioning tests

a LeChevallier et al., 1990. b Boulay and Edwards, 2000; Atlas et al., 1982; Cantor et al., 2003. c Edwards et al., 2000; Schock et al., 1995.

2.

Materials and methods

2.1.

Experimental protocol and water chemistry

2.1.1.

Formation of Cu(OH)2 solids

Cu(OH)2 solids were formed by addition of sodium hydroxide (50 mL of 1 M NaOH every 5 min for 45 min) to stirred 0.5 L solutions containing 0.5 mM cupric nitrate (Cu(NO3)2) and 1 mM sodium nitrate (NaNO3). The resulting solution contained Cu(OH)2 solids, final pH of 7.0, and soluble copper of 9  0.8 mg/L (Hidmi and Edwards, 1999).

2.1.2.

Beaker testing with pre-formed Cu(OH)2 solids

Aliquots of pre-formed Cu(OH)2 solids, aged for 2 h at pH 7 and 20  C, were dosed with a desired level of silica (Table 2) from a fresh stock solution of sodium silicate (Na2SiO3), which contained 8000 mg/L as silicon dioxide (SiO2) at pH 12.8. A predetermined dose of 1 M hydrochloric acid (HCl) was also added when the solids and silica were mixed to maintain a constant pH of 7  0.1. This approach minimizes but does not eliminate the likelihood of forming polymeric silicates and maintains monomeric silica thought to be present in natural waters (Davis et al., 2001). The pH and free chlorine concentration were then adjusted to target values between 7e10.5 and 0e75 mg/L Cl2 (Table 2), respectively. The solution was mixed on an orbital stir plate at 100 RPM. The pH was maintained within 0.20 units of the target during the first 8 h and 0.3 pH units during the remainder of the experiment (up to 40 days). Plastic containers were closed to the atmosphere. The sorption density (mole Si per mole Cu in the solids) was calculated by collecting solids on a 0.45 mm pore size nitrocellulose filter, and then dissolving the filter and solids in 20 mL distilled and deionized water containing 2% (v/v) nitric acid (HNO3). Visual observations and mass balances confirmed that dissolution of captured solids was complete for copper but not always for silica, which can be resistant to dissolution in strong acid.

2.1.3.

Copper polarization experiment

Distilled and deionized water containing 103 M NaCl at pH 7.0 was recirculated at a flow rate of 3.8 L/min through three 2.2-cm diameter copper pipe (4.1-cm length). Each copper pipe was polarized to 1.5 V vs. Ag/AgCl using a potentiostat.

2.1.4.

Practical case study of rapid chlorine decay

The reaction between free chlorine and copper pipe was tested using a plumbing commissioning procedure. Specific protocols vary but the American Water Works Association

5304

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 0 2 e5 3 1 2

Table 2 e Water chemistry. Parameter

Calcium (mg/L) Magnesium (mg/L) Alkalinity (mg/L as CaCO3) Sulfate (mg/L SO4) Chloride (mg/L) Initial Cu (mg/L) Si (mg/L as SiO2) pH Free chlorine dose (mg/L as Cl2)

Test Beaker tests with pre-formed Cu(OH)2

Case Study, low alkalinity water

Case Study, high alkalinity water

0 0 0 0 0 32 (0.5 mM) 0, 2.5, 5, 7.5, 10, or 30 7, 8, 9, 9.5, 10, or 10.5 0e75

17 0 35 14 20 0 0 6, 7, 8, 9, or 10 50 or 200

73 23 123 70 2230 0 5.4 8.1 or 9.4 150

(AWWA) C651 standard is typically accepted. The AWWA C651 standard includes three approaches using an initial chlorine concentration of 25e300 mg/L Cl2. Although the maximum allowable chlorine decay varies among the approaches, one method requires that no more than 60% decay of chlorine over 24 h, and the second specifies less than 50% decay after 3 h. The third AWWA C651 method requires that “a chlorine residual” (actual level is unspecified) be detected after 24 h. In this work, a more stringent protocol used by Brantford, Ontario (City of Brantford, 2010) was examined at bench-scale. The Brantford procedure requires initial chlorine residual of 100e150 mg/L Cl2. After a 24-h holding time, the chlorine residual must be greater than 25 mg/L Cl2 and must not decrease more than 30% from the initial concentration. To test this procedure, new 1.3-cm, 1.9-cm, or 5.1-cm diameter copper pipes (Type M) were obtained from a local hardware store in Blacksburg, VA, and cut into 61-cm lengths. Low alkalinity water (Table 2) was prepared in the lab, and high alkalinity water from the City of Brantford in Canada was shipped to Virginia Tech (Table 2). Copper pipe samples that had failed commissioning tests in newly constructed buildings in Brantford, Ontario, were also sent to our lab for testing. For all water conditions, the pH, orthophosphate corrosion inhibitor dose, and disinfectant concentrations were adjusted to target values (Table 2) prior to exposure to the plumbing materials. Silicone stoppers were used to hold the water in the pipes. The pH, free chlorine concentration, and orthophosphate concentration were measured at 3, 12, and 24 h of stagnation. After 24 h, the entire water volume remaining inside the copper pipe was poured into a plastic Nalgene bottle, acidified to 2% (v/v) HNO3, and analyzed for total copper. All conditions were tested with three replicate pipes.

2.2.

Analytical methods

A Malverne ZetaSizer 3000HS was used to measure zeta potential and the instrument performance was checked by comparison to standard solutions. Representative solids were collected for analysis using X-ray diffraction (XRD) using a Scintag instrument. Copper and silicon in the total, filtrate, and digested filter samples from the beaker tests (Table 2) were acidified with 2% (v/v) HNO3 and measured using inductively coupled plasma emission spectroscopy (ICP-ES).

Total copper, chloride, and sulfate in the bulk water from the practical case study (Table 2) was quantified by acidifying unfiltered water samples with 2% (v/v) nitric acid for at least 24 h and analyzing with an inductively coupled plasma mass spectrometer (ICP-MS) in accordance with Standard Method 3125-B (APHA, 1998). Concentrations of chloride and sulfate were cross-checked using DIONEX DX-120 ion chromatography according to Standard Method 4110 (APHA, 1998). The pH of the bulk water in all experiments was measured with an Accumet electrode in accordance with Standard Method 4500Hþ B (APHA, 1998). Free chlorine were measured on a Hach DR 2700 spectrophotometer according to Standard Method 4500Cl (APHA, 1998).

3.

Results

3.1.

Reaction of chlorine with cupric hydroxide

The initial expectation was that free chlorine would not react rapidly with pre-formed cupric hydroxide solids because Cu2þ species are generally considered the highest oxidation state of copper in aqueous solution. In the absence of cupric species (Fig. 1a), chlorine decayed very slowly in solution as expected based on practical experience, although it is known that factors such as light and heat could accelerate autodecomposition (APHA, 1998). When cupric ions or solids were present, however, the chlorine decayed very rapidly in solution at pH 7 and pH 9 (Fig. 1a). Indeed, approximately 50% of the chlorine decayed before the first sample could be collected. When silica was added to water containing Cu(OH)2, the rate of free chlorine decay slowed dramatically (Fig. 1b). While 5 mg/L Cl2 disappeared completely within 12 h in water with no silica, 1.7 mg/L Cl2 remained in water with 10 mg/L SiO2 (Fig. 1b). The ability of the solid to catalyze chlorine decay and subsequently form chloride after four consecutive doses of chlorine to the targeted amount was not lost (Fig. 2), although the rate of chlorine decay decreased marginally with each chlorine dose. Specifically, chlorine in the presence of Cu(OH)2 solids decayed during the first and third doses 1.2 h1 and 1.0 h1, respectively, whereas very little decay was observed when no solids were present (0.03 h1). After five consecutive

5305

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 0 2 e5 3 1 2

doses of 9.6 h1 mg/L Cl2 (0.7 mg/L), the first-order (Biswas et al., 1993; Kiene et al., 1998; Rossman et al., 1994; Wable et al., 1991) chlorine decay rate (kd) slowed to 0.3 h1 in the presence of Cu(OH)2; however, 75% of the chlorine decayed in just 4.5 h versus very little decay in pure DI water. Still other tests demonstrated that the Cu(OH)2 solids had a much greater impact on chlorine decay than did soluble cupric species (Fig. 3), and that the chlorine-induced transition of Cu(OH)2 solids to CuO markedly reduced the soluble copper concentration within 2 h (Fig. 4). That is, water dosed with 10 mg/L Cl2 required less than 2 h before soluble Cu decreased below 1 mg/L Cu (from an initial total concentration of 32 mg/L Cu), whereas reducing the chlorine concentration to 2 mg/L Cl2 required 21X more time or approximately 42 h to reach 1 mg/L soluble Cu (Fig. 4).

Effect of silica on cupric hydroxide aging

3.2.1.

Sorption density

Silica sorbed very strongly to the surface of Cu(OH)2 (Fig. 5). The sorption density was determined by measuring the difference between the total copper and silica prior to filtration, and measurements of total copper and silica in the filtrate as per the logic and approach used elsewhere for silica sorption to iron (Davis et al., 2001). Between pH 7e10.5, the sorption density was 0.30e0.77 M Si/M Cu for solutions initially containing 30 mg/L as SiO2. Although, the USEPA recommends a pH range of 6.5e8.5 for drinking water (pH as a secondary standard), water utilities can increase the pH above pH 8.5 as a corrosion control measure (Edwards et al., 1996). Unlike data for iron and aluminum (Davis et al., 2002; Hingston and Raupach, 1967; Sigg and Stumm, 1981), the sorption density was a relatively weak function of pH for initial silica at 30 mg/L as SiO2 (Fig. 5). Consistent with trends reported for Al(OH)3 and Fe(OH)3 surfaces (Davis et al., 2002; Hingston and Raupach, 1967; Sigg and Stumm, 1981), sorption density generally increased with reaction time and SiO2 concentration. Dosing 30 mg/L as SiO2 at pH 9 increased the sorption density from 0 to 0.37 M Si/M Cu after 8 h (Fig. 5). Zeta

Fig. 1 e Total chlorine (A) versus time at pH 7 or 9 for solutions containing chlorine with and without cupric hydroxide, and (B) versus time for cupric hydroxide solutions (Cu(OH)2) at pH 9 with and without silica. The control condition contained no cupric hydroxide solids. The pre-formed Cu(OH)2 solution initially contained 32 mg/ L total Cu and 9 mg/L soluble Cu. Error bars represent 95% confidence intervals.

DI water with no Cu Cu - Cl2 Dose #2 Cu - Cl2 Dose #4

3.2.

Cu - Cl2 Dose #1 Cu - Cl2 Dose #3 Cu - Cl2 Dose #5

OCl- only OCl- + Soluble Cu (2 mg/L) OCl- + Soluble Cu (1 mg/L) + Particulate Cu (1 mg/L) 6

8 Total Chlorine (mg/L Cl2)

Free Chlorine (mg/L Cl2)

10

6 4 2 0 0

1

2

3 4 Decay Time (h)

5

6

Fig. 2 e Chlorine decay over time for consecutive chlorine doses in distilled and deionized water (“DI”) or in water containing pre-formed Cu(OH)2 solids (“Cu”). Chlorine was dosed to 9.6 mg/L Cl2 (±0.7 mg/L) five times in water containing pre-formed Cu(OH)2 solids (initial concentration of 32 mg/L total Cu and 9 mg/L soluble Cu).

5 4 3 2 1

7 0 0

5

10

15

20

25

30

35

40

45

50

Time (h)

Fig. 3 e Free chlorine decay over time in distilled and deionized water with or without 2 mg/L soluble and/or particulate copper at pH 7. Particulate Cu was added as preformed Cu(OH)2 as described in the methods section.

5306

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 0 2 e5 3 1 2

pH 9.2

greater than 1 year Transition time (h)

Soluble Copper (mg/L Cu)

5

A 800

2 mg/L Cl2 0.5 mg/L Cl 2 USEPA Action Level (1.3 mg/L Cu)

0 mg/L Cl 2 10 mg/L Cl 2

4 3 2

600

400

200

1 0 0

0 12

24

36 Time (h)

48

60

72

Fig. 4 e Soluble copper concentration for samples at pH 7 initially containing 0e10 mg/L free chlorine as Cl2 and 0.5 mM total Cu (32 mg/L Cu).

potentials varied from þ45 to 30 mV, depending on the pH and silica concentration.

3.2.2.

Transition time of cupric hydroxide to tenorite

When Cu(OH)2 solids were aged in the laboratory at 20  C, the solids changed color from light blue and soluble Cu(OH)2, to the much less soluble brown CuO phase (Hidmi and Edwards, 1999). This transition typically occurs in a few hours at pH 9 but requires a few days at pH 7 (Edwards et al., 2001; Hidmi and Edwards, 1999). At a fixed pH of 9.2, solutions without silica were brown immediately, but the transition in solutions with 7.5 mg/L as SiO2 took 600 h (Fig. 6a). Even when the pH was raised to 10, the Cu(OH)2 solids remained blue for over one year when more than 10 mg/L as SiO2 was present (Fig. 6a), whereas the transition to CuO was complete in a few hours when silica was absent (Fig. 6b). Analysis of XRD patterns of representative solids confirmed that the transition of blue to brown solids in the experiments without silica corresponded with disappearance of Cu(OH)2 and the appearance of CuO (Table 3).

3.3.

Testing of chlorine and copper aging were extended to a practical case study of commissioning practices with very 8 hours

1 week

3 weeks

[Si]/[Cu]

0.80 0.60 0.40 0.20 0.00 6

7

5 Si (mg/L as SiO2)

7.5

10

0 mg/L as SiO2

75

50

25

0 6

7

8

9

10

11

pH

Fig. 6 e Transition time from blue Cu(OH)2 to brown CuO solids for (A) 0e10 mg/L SiO2 at pH 9.2, and (B) varying pH when SiO2 is absent. Chlorine was not dosed. The transition to CuO did not occur within 1 year in pH 7e10.5 water if the silica concentration was more than 30 mg/L SiO2. At pH 9.2, the time of transition was a strong function of silica concentration.

Table 3 e Solids identified in experiments by X-ray diffraction (XRD).

Practical case study

1.00

2.5

B 100 Transition Time (h)

0

8

9

10

11

pH

Fig. 5 e Sorption density of cupric hydroxide water as a function of pH in water containing 30 mg/L SiO2. Data are shown for 8 h, 1 week, and 3 weeks.

Solid Description

Color

Pre-formed Cu(OH)2, pH 7, aged 0.5 h Slow titration (5 h), pH 7, aged 8 h 0 mg/L as SiO2, pH 9.2, aged 8 h 10 mg/L as SiO2, pH 9.2, aged 8 h 30 mg/L as SiO2, pH 9.2, aged 8 h Pre-formed Cu(OH)2, pH 7, aged 8 h 65 mg/L Cl2, pH 7, aged 8 h 65 mg/L Cl2 þ 10 mg/L as SiO2, pH 7, aged 0.5 h 65 mg/L Cl2 þ 10 mg/L as SiO2, pH 7, aged 8 h

Blue

Cu(OH)2

Blue

no Cu(OH)2, unidentified peaks CuO

Brown Blue

Identification

Brown

no Cu(OH)2, unidentified peaks no Cu(OH)2, unidentified peaks CuO

Brown

CuO

Green

CuO

Brown

CuO

Blue

5307

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 0 2 e5 3 1 2

3.3.1.

Effect of copper pipe and diameter

In low alkalinity water at pH 8 (Table 2), free chlorine in glass pipes decayed slowly over 24 h (Fig. 7). In contrast, 97% of the initial 200 mg/L Cl2 decayed in 24 h while contacting 1.9-cm diameter copper pipe (Fig. 7). Moreover, chlorinated water in contact with smaller diameter (1.3-cm) copper pipe had the fastest decay rate in the low alkalinity water (Fig. 7). Specifically, the chlorine residual decreased from 200 mg/L Cl2 to 0 mg/L Cl2 (100% chlorine decay) after 6 h. Similar chlorine decay trends were observed for tests with an initial chlorine residual of 50 mg/L Cl2. Consistent with previous results, sequential dosing of free chlorine to the new copper pipes did not significantly slow the rate of free chlorine decay (results not shown).

3.3.2.

Effect of pH and orthophosphate in low alkalinity water

After establishing the catalyzing effect of new copper pipe on free chlorine decay, additional testing to evaluate the role of pH and orthophosphate on maintaining the chlorine residual was conducted. Regardless of the pH, the chlorine concentration decreased more than 80% in 12 h in water containing no orthophosphate corrosion inhibitor (Fig. 8). The chlorine decay was fastest in water at pH 6 or 7, whereas higher pH (8e10) maintained chlorine residuals longer (Fig. 8). Specifically, chlorine decayed from 50 mg/L Cl2 to 18e25 mg/L Cl2 (50e64% decay) within 3 h, while only 1e9 mg/L Cl2 (2e18%) decayed in 3 h for pH 8e10 water (Fig. 8). However, the

Glass pipe (1.9-cm diameter) Cu pipe (1.9-cm diameter)

Cu pipe (1.3-cm diameter)

Free Chlorine (mg/L as Cl2)

250 200 150 100 50 0 0

6

12

18

24

Time (h)

Fig. 7 e Residual free chlorine during 24-h holding time of water initially containing 200 mg/L Cl2 for glass and copper pipe. The test water had an alkalinity of 35 mg/L as CaCO3 and a pH of 8. Error bars represent 95% confidence intervals.

A

pH 6

pH 7

pH 8

pH 9

pH 10

30% decrease

60

Free Chlorine (mg/L as Cl2)

0 mg/L P

B Free Chlorine (mg/L as Cl2)

high concentrations of chlorine (typically 25e300 mg/L Cl2) for new copper pipes. Reaction of chlorine with copper can dramatically deplete the chlorine concentration and can cause plumbing to fail commissioning procedures. Although commissioning standards vary, the following case study focuses on a protocol imposed in Brantford, Ontario, which requires an initial chlorine concentration of 100e150 mg/L Cl2 and no more than 30% residual decay over 24 h. Specifically, two levels of chlorine (50 and 200 mg/L Cl2) were evaluated with low alkalinity water, and 150 mg/L Cl2 was tested for high alkalinity water from the City of Brantford, Ontario.

50 40 30 20 10 0 60 1 mg/L P 50 40 30 20 10 0 0

6

12

18

24

Decay Time (h)

Fig. 8 e Residual free chlorine after 50 mg/L Cl2 shock chlorination of water containing (A) 0 mg/L P and (B) 1 mg/L orthophosphate as P in 1.9-cm diameter copper pipe. The alkalinity of the water was 35 mg/L as CaCO3. The dashed line at approximately 36 mg/L Cl2 represents the allowable 30% chlorine decay according to the Brantford, Ontario, commissioning procedure.

chlorine residual disappeared completely after 24 h for all pH values (pH 6e10). Consistent with other research (Edwards et al., 1996; Rahman et al., 2007; Schock et al., 1995), higher pH in this low alkalinity water decreased copper solubility (results not shown). Specifically, the decay rate of chlorine decreased 47% from 0.21 h1 to 0.11 h1 when 1 mg/L P was added to the water at pH 7 (Fig. 8). Analogous to water without orthophosphate, increasing pH increased the stability of the chlorine residual. Similar trends were observed for water initially containing 200 mg/L Cl2. The effect of pH and phosphate on copper release in the low alkalinity water was visible in water exposed to copper pipes for 24 h (Fig. 9). In water with no orthophosphate corrosion inhibitor and high pH (i.e., pH 9 and 10), the copper aged to brown CuO, consistent with prior tests in water with high pH, high chlorine concentrations, and low silica concentration. In contrast, water dosed with orthophosphate corrosion inhibitor emerged after 24 h of exposure to copper pipe as clear solutions (results not shown), consistent with observations that the copper phosphate solids that formed did not react with chlorine to form brown CuO (Fig. 8b).

3.3.3. water

Effect of pH and orthophosphate in high alkalinity

The investigation of the effects of pH and orthophosphate corrosion inhibitor was extended to high alkalinity water

5308

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 0 2 e5 3 1 2

Fig. 9 e Photograph of water after exposure to 200 mg/L Cl2 and 0 mg/L orthophosphate as P.

shipped from Brantford, Ontario (Table 2) using harvested 5.1cm diameter copper pipes that failed in new construction commissioning tests. Increasing the pH from pH 8.1 to pH 9.4 in the bench-scale tests stabilized the chlorine residual in water containing no orthophosphate (Fig. 10). The USEPA typically recommends a pH range of 6.5e8.5 in drinking water, but the pH guideline is unenforceable, and some water utilities increase the pH above the specified range as a corrosion control strategy. That is, the chlorine residual in the pH 9.4 water decreased 28% from 151 mg/L Cl2 to 109 mg/L Cl2 after 24 h, whereas the lower pH water had slightly higher chlorine decay of 34% from 149 mg/L Cl2 to 98 mg/L Cl2 (Fig. 10). While the Brantford water at the typical pH of 8.1 would not have met the commissioning standard, the chlorine decay was slowed enough at pH 9.4 that the copper plumbing could have passed the standard. Similar trends were also observed for brand new copper pipe, although the free chlorine decay was 36% for pH 9.4 water in the new copper pipe after 24 h. Dosing 1 mg/L orthophosphate as P in the higher alkalinity water increased chlorine decay compared to water with no orthophosphate (Fig. 10),

pH 8.1

pH 9.4

pH 8.1 and 1 mg/L P

30% reduction

Free chlorine (mgL Cl2)

160 140 120 100 80 60 0

6

12 Decay Time (h)

18

24

Fig. 10 e Residual free chlorine over time in high alkalinity water (123 mg/L as CaCO3) initially containing 150 mg/L Cl2 and exposed to 5.1-cm diameter copper that was harvested from Brantford, Ontario in Canada. Error bars represent 95% confidence intervals.

4.

Discussion

4.1.

Reaction of chlorine with cupric hydroxide

Free chlorine decay was rapid in the presence of Cu(OH)2 solids, and the “chlorine demand” of the Cu(OH)2 solids seemed to be without practical limits (Fig. 2). Other testing showed complete and rapid chlorine consumption in response to as many as 20 doses of chlorine (data not shown). QA/QC testing confirmed that, while soluble cupric ions had a slight positive influence on the DPD test for chlorine used in this experiment (APHA, 1998), the observed chlorine decay was not a measurement artifact. Moreover, as chlorine decayed, the color of the Cu(OH)2 solids changed within 10 min from blue to brown, and some tenorite (CuO) was detectable via XRD. This transition from Cu(OH)2 to CuO is associated with decrease in copper solubility (Edwards et al., 1996, 2001; Hidmi and Edwards, 1999; Lagos et al., 2001; Schock et al., 1995). In combination, these results indicate that the observed chlorine decay was not simply due to catalysis of chlorine autodecomposition because the Cu(OH)2 solids themselves were obviously altered (Gordon et al., 1995; Powers, 2000; Sneed et al., 1954).

4.2.

Effect of silica on cupric hydroxide aging

The experiments with Cu(OH)2 solids (e.g., Figs. 1a and 2) suggest that it is nearly impossible for new copper pipe to pass targets for chlorine residuals during superchlorination in pure water. Because copper piping sometimes passes commissioning, it was suspected that constituents found in some water conditions could interfere with the reaction between chlorine and Cu(OH)2 solids. Silica is naturally occurring in drinking water and can react with Cu(OH)2 by sorption or through formation of a new solid such as dioptase (CuH2SiO4, Equation [1]): CuðOHÞ2 þSiðOHÞ4 /CuH2 SiO4 ðdioptaseÞ þ 2H2 O

(1)

When reaction products of Cu(OH)2 and silica were analyzed, the molar ratios of Cu:Si ranged between 3.5 and 43, which is well above the 1-to-1 ratio for dioptase, if it formed quantitatively. Thus, a solid such as dioptase did not form exclusively, although both Cu(OH)2 and CuH2SiO4 might be present. Considering the Si:Cu ratios in the solid and assuming that only Cu(OH)2 and dioptase were present, the concentration of dioptase could have been no greater than 2e29% of the total copper. The solids were also analyzed by XRD, and no patterns out of the ten possible Cu:Si solids in the standard database (including dioptase) matched the solid peaks in this system (Table 3). When silica was present, only weak XRD peaks were observed, with no strong signal of either Cu(OH)2 or CuO, which is indirect evidence that silica sorbed to the Cu(OH)2 surface. All further discussion assumes that the solids present were actually Cu(OH)2 with silica sorbed to the surface in a monomeric or polymeric form. As the pH and silica concentration increased in experiments with Cu(OH)2 solids, the zeta potential decreased (results not shown; Powers, 2000). Thus, silica sorption converted positively charged sites on the Cu(OH)2 surface to

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 0 2 e5 3 1 2

neutral and negatively charged sites, presumably through surface complex formation. Observations of color changes in 32 experiments for this work with Cu(OH)2 demonstrated that silica slowed the transition from Cu(OH)2 to CuO. This is analogous to the effect of silica on iron hydroxide solids aging. The transition of ferrihydrite (Fe2O3∙0.5H2O) to goethite (FeOOH) normally completes in 24 h, but can take one to two weeks in the presence of silica (Anderson and Benjamin, 1985). The results in this work suggest that silica sorption to the Cu(OH)2 surface (Fig. 5) also decelerated the chlorine decay rate (Fig. 1b).

4.2.1. Effects on solubility, particle size and deposition to the plasticware It was observed that soluble copper levels were higher at 10e30 mg/L SiO2 in the presence of a Cu(OH)2 solid at pH 7 when compared to 0e5 mg/L SiO2 (results not shown). This might translate to longer times for the copper solubility to drop below 1.3 mg/L Cu (U.S. Environmental Protection Agency’s action level for copper in drinking water) in water contacting copper plumbing systems (Fig. 4). Measurement of free cupric ion (Cu2þ) at pH 7 in standard solutions using an ion specific electrode indicated that the addition of 30 mg/L SiO2 did not significantly change the concentration of free copper in a standard solution at pH 7.0 (results not shown). This further supports the idea that silica was causing higher copper solubility short-term by hindering the transition of Cu(OH)2 to CuO. At pH 9.2, there was little effect of silica on copper solubility (results not shown), despite the fact that the blue cupric solid was maintained for long time periods (Fig. 6a). However, dosing free chlorine caused a rapid drop in soluble copper (e.g., Fig. 4), which is consistent with the formation of a less soluble CuO phase. Particle size, unlike copper solubility, was greatly influenced by the concentration of silica at pH 9.2, consistent with reported effects in experiments starting with pre-formed iron hydroxide (Davis et al., 2001). Although the size of most particles was above the particle sizer’s upper detection limit of 3 mm, the general visual trend was that the higher the silica concentration, the larger the particles. Moreover, the blue solids in systems with silica tended to remain suspended much longer than in the systems where CuO was confirmed to form after 0.5 h settling, and only a small fraction of the solids in the presence of silica attached to the walls of the plastic reactors whereas the majority of the CuO solids did attach. In summary, silica in water, either naturally present or added as a corrosion inhibitor, can sorb to Cu(OH)2 surfaces and dramatically slow transitions to other solid phases such as CuO and decrease the rate of chlorine decay. In the rare practical situations when very high concentrations of particulate copper are released to drinking water near pH 9.2 from copper pipe, the sorption of silica to copper can explain why the copper solids occur as “blue water” from Cu(OH)2 (Edwards et al., 2000) and not as “brown water” from the thermodynamically favored tenorite phase. It is noteworthy that silica was initially hypothesized to be a key contributor in blue water propagation (Page et al., 1974), and this work directly supports the idea that tenorite rapidly forms from Cu(OH)2 at higher pH values if silica is less than about 5 mg/L as SiO2. It remains to be seen whether the silica is involved in helping to

5309

mobilize the high concentration of Cu(OH)2 particles from copper pipe and directly contribute to cases of blue water. Tenorite solids attaching strongly to the plasticware whereas blue Cu(OH)2 coated with silica remaining unattached suggests that meaningful differences in adhesion and durability are likely.

4.3.

Practical case study

The reactions of chlorine with Cu(OH)2 can have implications for commissioning copper pipes. Tests revealed that chlorine decayed rapidly in copper pipes (Figs. 7 and 8, and 10). Although free chlorine decay in low alkalinity water was more rapid with increasing copper surface area relative to water volume (i.e., smaller diameter pipe), no such trend was observed in potable water with higher alkalinity and hardness (results not shown). It is possible that in some cases, chlorine autodecomposition reactions with other species in water would dominate over pipe surface catalyzed reactions (i.e., copper in water). Prior work also found greater chlorine decay rates for smaller diameter pipes made of steel, cast iron, and polyvinyl chloride in water with unspecified alkalinity, although copper was not evaluated in that study (Al-Jasser, 2007). These results suggest that it would be more difficult for smaller diameter copper plumbing in low alkalinity water to pass commissioning tests, and the effects of copper on chlorine residuals persist even after multiple doses of chlorine. The mechanism of rapid free chlorine decay in the presence of copper pipe (Fig. 7) is likely due to Cu(OH)2 on the copper pipe wall reacting with free chlorine, aging the copper scale to CuO and decreasing the free chlorine residual. Likewise, the chlorine residual was more stable in higher pH water, presumably because less Cu(OH)2 (copper corrosion byproduct) was present to catalyze chlorine decay reactions. By adding orthophosphate to the low alkalinity water, the rate of chlorine decay decreased. Perhaps chlorine did not react with the copper phosphate solid that formed, unlike what would occur with Cu(OH)2 solid. In this test, the chlorine residual in the low alkalinity water at pH 10 and with orthophosphate did not drop more than 30% and would have passed the Brantford, Ontario, commissioning criteria (Fig. 8). However, the copper concentration increased with the addition of orthophosphate for pH 8e10. In water with orthophosphate, copper concentration was greater at pH 10 than at pH 8, although the magnitude of copper release was much less at pH 8e10 than in pH 6e7 water (results not shown). These results are consistent with prior work that found orthophosphate is most effective at reducing copper solubility when the pH is between 6.5 and 7.5 (Schock et al., 1995). The net effects of phosphate addition to high alkalinity water are complicated and depend on other constituents (Fig. 10). It is possible that orthophosphate in high alkalinity water interfered with the formation of insoluble copper species (Schock et al., 1995) that would normally protect the copper surface from further corrosion. Hence, relatively more Cu(OH)2 may have been allowed to form compared to high alkalinity conditions with no phosphate, subsequently catalyzing chlorine decay.

5310

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 0 2 e5 3 1 2

4.4. Proposed mechanism of rapid chlorine decay and simultaneous copper aging A thorough review of the literature suggested several possible reactions that might be occurring to age the Cu(OH)2 solids and catalyze chlorine decay, including the following: 1) precipitation reaction between cupric ion and free chlorine formed a cupric hypochlorite solid (Gordon et al., 1995; Sneed et al., 1954), 2) chlorine oxidized Cu(II) to a Cu(III) species that was less soluble (Sneed et al., 1954), 3) Cu(OH)2 or cupric solids catalyzed chlorine autodecomposition (Gordon et al., 1995) and chlorine simultaneously catalyzed Cu(OH)2 aging to another solid such as CuO (Powers, 2000). In the latter case, both proposed catalytic reactions had to occur simultaneously, because both of the reactants, chlorine (Cl2) and soluble Cu(OH)2, were consumed. After examining these possibilities, the only established reaction pathway consistent with the chlorine data was catalysis of chlorine decay by Cu(OH)2 solid (Gray et al., 1977). We speculate that the catalysis proceeds via an unstable Cu(III) species, although the specific pathway is uncertain (Gray et al., 1977; Lister, 1956; Thenard, 1818). A possible reaction pathway consistent with the data and experimental observations is: 

Step 1: CuðOHÞ2 þ0:5HOCl þ 0:5OH /CuðOHÞ3 þ 0:5Cl

(2)

Step 2: CuðOHÞ3 /CuO þ 0:25O2 þ 1:5H2 O

(3) 

Overall : CuðOHÞ2 þ0:5HOCl þ 0:5OH /CuO þ 0:5Cl þ 0:25O2 þ 1:5H2 O

ð4Þ

To further examine the reaction pathway described above via an unstable Cu(OH)3 intermediate (Equation [4]), pH 7.0 water was recirculated through copper pipes. When the copper tubes were polarized to 1.5 V vs. Ag/AgCl, which is above the 0.52 V vs. Ag/AgCl potential required for conversion of Cu(II) to Cu(III) at pH 7 (El Din and El Wahab, 1964; Nahle and Walsh, 1994), the blue Cu(OH)2 solids released to the water from the oxidized Cu metal turned brown (indicative of tenorite or CuO) much more rapidly than when polarized to only 0.3 V (results not shown). Assuming Cu(III) was formed above 0.52 V, this is indirect evidence that unstable Cu(III) intermediates could accelerate the formation of brown tenorite CuO, as was believed to occur for Cu(OH)2 in the presence of free chlorine. Chloride (Cl) is more thermodynamically stable than free chlorine (HOCl or OCl) below about 1.0 V vs. Ag/AgCl (Pourbaix, 1974). In prior work with electrodeposition, an unstable Cu(III) species was thought to form and was immediately reduced to CuO (Siegfried and Choi, 2007). Although the specific mechanism of the reaction between Cu(OH)2 and free chlorine could not be unambiguously defined using techniques in this research, the reaction has obvious implications for maintaining disinfectant residuals, disinfecting biofilms on copper or brass plumbing, and controlling copper leaching (Table 1). In practical circumstances where copper leaching is decreased after dosing free chlorine (Cantor, 2009; Edwards et al., 2000; Schock et al., 1995), the assumption of the impact of microbial inactivation is probably overly simplistic because it is now clear that

chlorine causes important abiotic reactions with pre-formed Cu(OH)2 that control the aging of Cu(OH)2 to less soluble CuO. Other biocides that have been tested such as heat, oxidants, and even antibiotics might also be expected to abiotically influence transitions between solid phases that are important to corrosion. It is also understandable that maintaining a free chlorine residual in new copper plumbing is occasionally difficult while commissioning pipelines because the sources of chlorine demand obviously go well beyond direct oxidation of copper metal to cupric species as highlighted by Reiber (1989) It would be interesting to determine if similar disinfectant-consuming reactions occur between Cu(OH)2 solids and chloramine or chlorine dioxide, which are becoming increasingly popular in drinking water.

5.

Conclusions

1) Free chlorine can catalyze aging of Cu(OH)2 to CuO, which occurs at the same time as rapid free chlorine decay. Although chlorine residuals play important roles in controlling biofilm microbial growth on pipe surfaces and reducing microbial corrosion, chlorine also clearly plays an abiotic role in promoting aging (i.e., passivation) of copper scales on pipe surfaces. 2) Silica, either naturally present in water or added as a corrosion inhibitor, can sorb to Cu(OH)2 surfaces (e.g., in or near copper pipe scale) and dramatically slow transitions to other solid phases such as CuO. Greater concentrations of silica can also slow the rate of free chlorine decay. In the absence of silica, “blue water” (likely Cu(OH)2 solids) could not occur at pH 9.2 because brown CuO would form. It is uncertain if silica plays a direct role in actually causing elevated copper release in practical situations where blue water problems occur. 3) The practice of shock chlorination or “superchlorination” of new copper plumbing can corrode the pipe, producing high concentrations of copper in water and very rapid chlorine decay during the commissioning procedure. The chlorine decay can prevent achievement of targeted residuals to “pass” commissioning. 4) Slower chlorine decay was observed in lower alkalinity water with larger pipe diameters, addition of orthophosphate, and increasing pH. 5) In higher alkalinity water, slower chlorine decay occurred with increasing pH but was not a function of pipe diameter. Orthophosphate addition accelerated chlorine decay in high alkalinity water, perhaps due to orthophosphate interfering with the formation of more insoluble copper solids (e.g., malachite).

Acknowledgments This material is based upon work supported under a National Science Foundation (NSF) Graduate Research Fellowship and with the financial support of the Water Research Foundation (WaterRF) and the Copper Development Association (CDA).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 0 2 e5 3 1 2

Any opinions, findings, conclusions or recommendations expressed in this publication are those of the author(s) and do not necessarily reflect the views of the NSF, WaterRF, or CDA. Portions of this work were presented at the 2001 Australasian Corrosion Association Annual Conference in Newcastle, Australia and appeared in the conference proceedings.

references

Al-Jasser, A.O., 2007. Chlorine decay in drinking-water transmission and distribution systems: pipe service age effect. Water Research 41 (2), 387e396. American Public Health Association (APHA), 1998. Standard Methods for the Examination of Water and Wastewater. APHA, American Water Works Association, and Water Environment Federation, Washington, D.C. Anderson, P.R., Benjamin, M.M., 1985. Effect of silicon on the crystallization and adsorption properties of ferric oxides. Environmental Science & Technology 19 (11), 1048e1053. Atlas, D., Coombs, J., Zajicek, O.T., 1982. The corrosion of copper by chlorinated drinking waters. Water Research 16 (5), 693e698. Biswas, P., Lu, C., Clark, R.M., 1993. A model for chlorine concentration decay in pipes. Water Research 27 (12), 1715e1724. Boulay, N., Edwards, M., 2000. Copper in the urban water cycle. Critical Reviews in Environmental Science and Technology 30 (3), 297e326. Bremer, P.J.W., Barbara, J., Brett, Wells D., 2001. Biocorrosion of copper in potable water. Journal American Water Works Association 93 (8), 82e91. Building Officials and Code Administrators (BOCA), 1997. BOCA National Plumbing Code, nineth ed. BOCA, Country Club Hills, IL. Cantor, A.F., 2009. Water Distribution System Monitoring: A Practical Approach for Evaulating Drinking Water Quality. CRC Press, Boca Raton, FL. Cantor, A.F., Park, J.K., Vaiyavatjamai, P., 2003. Effect of chlorine on corrosion in drinking water systems. Journal American Water Works Association 95 (5), 112e123. City of Brantford, 2010. Section 3: The Corporation of the City of Brantford Specification for Water Main Construction. Craun, G.F., Calderon, R.L., 2001. Waterborne disease outbreaks caused by distribution system deficiencies. Journal American Water Works Association 93 (9), 64e75. Davis, C.C., Chen, H.-W., Edwards, M., 2002. Modeling silica sorption to iron hydroxide. Environmental Science & Technology 36 (4), 582e587. Davis, C.C., Knocke, W.R., Edwards, M., 2001. Implications of aqueous silica sorption to iron hydroxide: mobilization of iron colloids and interference with sorption of arsenate and humic substances. Environmental Science & Technology 35 (15), 3158e3162. Edwards, M., Jacobs, S., Taylor, R., 2000. The blue water phenomenon. Journal American Water Works Association 92 (7), 72e82. Edwards, M., Schock, M.R., Meyer, T.E., 1996. Alkalinity, pH, and copper corrosion by-product release. Journal American Water Works Association 88 (3), 81e94. Edwards, M., Sprague, N., 2001. Organic matter and copper corrosion by-product release: a mechanistic study. Corrosion Science 43 (1), 1e18. Edwards, M.A., Parks, J., Griffin, A., Raetz, M., Martin, A., Scardina, P., Elfland, C., 2011. Lead and Copper Corrosion Control in New Construction. Water Research Foundation, Denver, CO.

5311

Edwards, M.A., Powers, K., Hidmi, L., Schock, M.R., 2001. The role of pipe ageing in copper corrosion by-product release. Water Science & Technology: Water Supply 1 (3), 25e32. El Din, A.M.S., El Wahab, F.M.A., 1964. The behaviour of the copper electrode in alkaline solutions upon alternate anodic and cathodic polarization. Electrochimica Acta 9 (1), 113e121. Gordon, G.A., Adam, L., Bubnis, B., 1995. Minimizing Chlorate Ion Formation in Drinking Water when Hypochlorite Ion is the Chlorinating Agent. American Water Works Association Research Foundation, Denver, CO. Gray, E.T., Taylor, R.W., Margerum, D.W., 1977. Kinetics and mechanisms of the copper-catalyzed decomposition of hypochlorite and hypobromite: properties of a dimeric copper(III) hydroxide intermediate. Inorganic Chemistry 16 (12), 3047e3055. Hidmi, L., Edwards, M., 1999. Role of temperature and pH in Cu(OH)2 solubility. Environmental Science & Technology 33 (15), 2607e2610. Hingston, F., Raupach, M., 1967. The reaction between monosilicic acid and aluminium hydroxide. I. Kinetics of adsorption of silicic acid by aluminium hydroxide. Australian Journal of Soil Research 5 (2), 295e309. International Association of Plumbing and Mechanical Officials, 1997. Uniform Plumbing Code. International Association of Plumbing and Mechanical Officials, Walnut, CA. International Code Council, 2000. International Plumbing Code. International Code Council, Falls Church, VA. Kiene, L., Lu, W., Levi, Y., 1998. Relative importance of the phenomena responsible for chlorine decay in drinking water distribution systems. Water Science and Technology 38 (6), 219. Lagos, G.E., Cuadrado, C.A., Letelier, M.V., 2001. Aging of copper pipes by drinking water. Journal American Water Works Association 93 (11), 94e103. Lister, M.W., 1956. Decomposition of sodium hypochlorite: the uncatalyzed reaction. Canadian Journal of Chemistry 34, 465e478. Marshall, B.J. (2004) Initiation, propagation, and Mitigation of aluminum and chlorine induced pitting corrosion. Master’s Thesis, Virginia Tech, Blacksburg, VA. Nahle, A., Walsh, F., 1994. Ex-situ x-ray diffraction of electrochemically formed films on copper. Bulletin of Electrochemistry 10 (9e10), 401e408. Page, G.G.B., Bailey, P.C.A., Wright, G.A., 1974. Mechanisation of new type of copper corrosion in water. Australasian Corrosion Engineering 18 (11/12), 13e19. Patterson, J.W., Boice, R.E., Marani, D., 1991. Alkaline precipitation and aging of copper from dilute cupric nitrate solution. Environmental Science & Technology 25 (10), 1780e1787. Pizarro, F., Olivares, M., Araya, M., Gidi, V., Uauy, R., 2001. Gastrointestinal effects associated with soluble and insoluble copper in drinking water. Environmental Health Perspectives 109 (9), 949e952. Pourbaix, M., 1974. Atlas of Electrochemical Equilibria in Aqueous Solutions. National Association of Corrosion Engineers (NACE), Houston, TX. Powers, K.A. (2000) Aging and copper corrosion by-product release: role of Common anions, impact of silica and chlorine, and Mitigating release in New pipe. Master’s Thesis, Virginia Tech, Blacksburg, VA. Rahman, S., McDonald, B.C., Gagnon, G.A., 2007. Impact of secondary disinfectants on copper corrosion under stagnant conditions. Journal of Environmental Engineering 133 (2), 180e185. Reiber, S., 1989. Copper plumbing surfaces: an electrochemical study. Journal American Water Works Association 81 (7), 114e122.

5312

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 0 2 e5 3 1 2

Rossman, L.A., Clark, R.M., Grayman, W.M., 1994. Modeling chlorine residuals in drinking water distribution systems. Journal of Environmental Engineering 120 (4), 803e820. Rushing, J.C. (2002) Advancing the understanding of water distribution system corrosion: effects of chlorine and aluminum on copper pitting, temperature Gradients on copper corrosion, and silica on iron release. Master’s Thesis, Virginia Tech, Blacksburg, VA. Schock, M.R., Lytle, D.A., Clement, J.A., 1995. Effect of pH, DIC, Orthophosphate and Sulfate on Drinking Water Cuprosolvency. U.S. Environmental Protection Agency, Office of Research and Development, National Risk Management Research Laboratory, Cincinnati, OH. Siegfried, M.J., Choi, K.-S., 2007. Conditions and mechanism for the anodic deposition of cupric oxide films in slightly acidic aqueous media. Journal of the Electrochemical Society 154 (12), D674eD677.

Sigg, L., Stumm, W., 1981. The interaction of anions and weak acids with the hydrous goethite (a-FeOOH) surface. Colloids and Surfaces 2 (2), 101e117. Sneed, M.C., Maynard, J.L., Brasted, R.C., 1954. Comprehensive inorganic chemistry: volume 1. Soil Science 77 (1), 76. Southern Building Code Congress International (SBCCI), 1988. Standard Plumbing Code. SBCCI, Birmingham, AL. Thenard, L.-J., 1818. Nouvelles observations sur les acides et les oxides oxigenes. Annales de chimie et de physique 9, 51e56. U.S. Environmental Protection Agency, 1991. Code of Federal Regulations 40 Parts 141 and 142. Wable, O., Dumoutier, N., Duguet, J.P., Jarrige, P.A., Gelas, G., Depierre, J.F., 1991. Modelling chlorine concentrations in a network and applications to Paris distribution network. American Water Works Research Foundation, Denver, CO.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 1 3 e5 3 2 2

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Occurrence and fate of TMDD in wastewater treatment plants in Germany Arlen A. Guedez*, Wilhelm Pu¨ttmann Institute of Atmospheric and Environmental Sciences, Department of Analytical Environmental Chemistry, J.W. Goethe University Frankfurt am Main, Altenho¨ferallee 1, 60438 Frankfurt am Main, Germany

article info

abstract

Article history:

The occurrence and fate of 2,4,7,9-tetramethyl-5-decyne-4,7-diol (TMDD) was investigated

Received 25 February 2011

in four wastewater treatment plants (WWTPs) in Germany. The concentration of TMDD in

Received in revised form

influents and effluents in the WWTPs ranged from 134 ng/L to 5846 ng/L and from
11 July 2011

3539 ng/L correspondingly. Loads determined in influents (10.1 g/de1142 g/d) and effluents

Accepted 29 July 2011

(
Available online 5 August 2011

elimination rates varied between 33% and 68%. Based on the load analysis, the TMDD effluent discharge of WWTPs investigated varied from 8.29 kg/a to 52.6 kg/a. Day and week

Keywords:

profiles were recorded and indicated that TMDD is introduced into the sewage through

2,4,7,9-Tetramethyl-5-decyne-4,

household and indirect dischargers with high fluctuations. Seasonal variations in the

7-diol

TMDD loads were also analyzed in three of the studied WWTPs. One of the WWTPs

Wastewater treatment

demonstrated statistically higher TMDD loads during the warm period (164 g/d) than during the cold period (91.3 g/d), for the others WWTPs any differences could not be established. The input of TMDD during weekends and working days was also studied. The results did not show any significant trend of TMDD discharge into the WWTPs. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Wastewater treatment plants (WWTPs) are important sources of organic pollutants in aqueous environments. Many studies have proven that these operating plants are not capable of completely removing all compounds that are present in wastewater. Moreover, during wastewater treatment, substances can undergo transformation reactions generating toxic metabolites as shown is the case of the alkylphenol polyethoxylate surfactants (Ahel et al., 1994). Others compounds, such as carbamazepine pass almost unaffected the WWTPs and are discharged into rivers without a significant depletion (Zhang et al., 2008). This study focuses on the occurrence and fate of 2,4,7,9tetramethyl-5-decyne-4,7-diol (TMDD) in WWTPs. TMDD is

a non-ionic surfactant used as defoamer in many industrial processes. Its commercial name is Surfynol 104 which has been marketed e.g. for waterborne industrial applications in coatings, printing inks and adhesives as well as in agriculture chemicals. It has been incorporated into the high production volume chemical challenge program (HPVC) of the Environmental Protection Agency (EPA) in the United States (USA) for substances, which are produced or imported in the USA in quantities of 1 million pounds or more per year (EPA, 2011). Beside the uses of TMDD in the series Surfynol 104, it is also employed as an ingredient for other products, like Surfynol 502, ZetaSperse 1600, Surfynol PSA204, Surfynol DF-37, Surfynol SE, Surfynol OP-340, Surfynol GA and in the series Surfynol CT-111, CT-121, CT-131, CT-136, CT-211, CT-222 (Air Products and Chemicals, 2008). Furthermore,

* Corresponding author. Tel.: þ49 69 79840228; fax: þ49 69 79840240. E-mail addresses: [email protected], [email protected] (A.A. Guedez). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.038

5314

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 1 3 e5 3 2 2

TMDD is used as a precursor compound for the synthesis of its ethoxylates, Surfynol 420 and Surfynol 440. TMDD is characterized by a high water solubility (1.7 g/L at 20  C) and is expected to remain in water phase because of the low Henry’s Law constant (8.58  107 atm m3/mol) preventing the compound from evaporation into the gas phase. The low octanol/water partition coefficient (Log Kow 2.8) prevents the compound from being adsorbed to particulate matter (Air Products and Chemicals, 2002). Based on these physicochemical properties, TMDD is highly mobile in aqueous systems and there is a potential for this compound to be transferred from domestic, commercial, or industrial dischargers to natural waters. Therefore, information on the possible sources of TMDD and its degradability in WWTPs is required. If TMDD is released from products with relevant water applications, its primary biodegradability must be at least 80% using the OECD-Screening-Test or the OECDConfirmatory-Test, (tenside regulation), (Tensidverordnung, 1977) in accordance with the regulation for the environmental sustainability of detergents and cleaning agents in Germany (WRMG, 2007). No data are available from application of the OECD-Screening-Test or the OECD-ConfirmatoryTest on TMDD. Instead, the biodegradability of TMDD has been tested using the OECD Tests Guide-line 302 A and was found to be 25.4% during 57 days. To date, only a few studies have reported the occurrence of TMDD in river water. In 2004 TMDD was found in Lippe river with concentrations ranging from 0.054 mg/L to 0.427 mg/L (Dsikowitzky et al., 2004a, 2004b). More recently, TMDD has been found in the river Rhine in Worms. The concentration of the compound varied from 0.125 mg/L to 1.33 mg/L and the mean load was 62.8 kg/d (Guedez et al., 2010). The LANUV, (Landesamt fu¨r Natur Umwelt und Verbraucherschutz), state office for nature, environmental and consumer protection of North-Rhine-Westphalia (state in Germany) reported the presence of TMDD in rivers such as Wupper and Ruhr. The concentrations ranged from 0.50 mg/L to 4.25 mg/L (LANUV, 2008). WWTPs have been suggested as possible sources for TMDD in rivers (Dsikowitzky et al., 2004b) but until now detailed studies on the occurrence of TMDD in effluents of wastewater treatment plants and on the removal efficiency of WWTPs with respect to TMDD are scarce. The only published data on concentrations of TMDD in a WWTP influent (0.73 mg/L) are coming from San Diego County, California, USA (Loraine and Pettigrove, 2005).

In contrast to a previous study (Guedez et al., 2010), which was about the presence of TMDD in river water, this research is the first comprehensive report on TMDD in wastewater. Until now, there is no information about, if TMDD can be found in wastewater in Germany or if WWTPs discharge TMDD into the rivers, because they are not able to remove it completely from the wastewater. In addition, this works deal with the monitoring of TMDD in different WWTPs in order to recognize seasonal patterns of the input of TMDD into the sewage (day, working days, weekends or seasons). Therefore, this study will facilitate the understanding the occurrence of TMDD in rivers by assessing the levels of TMDD in WWTPs.

2.

Materials and methods

2.1.

Chemicals and reagents

TMDD with a purity of 98% and squalane (internal standard) were purchased from SigmaeAldrich. Acetone was HPLC grade. The other solvents (dichloromethane and methanol) were analytical grade and were purified using fractional distillation. The distillation column was filled with metal structured packing. Calcium chloride (CaCl2) and magnesium turnings were used as desiccants for dichloromethane and methanol respectively.

2.2.

Description of the wastewater treatment plants

Four WWTPs located in the Rhine-Main area (Germany) were sampled in this study. Their particular characteristics are presented in Table 1. The yearly treatment volumes vary from 16 million m3/a to 92 million m3/a. The general treatment of the sewage in these plants is composed of a mechanical pretreatment by screening coarse particles, an aerated gritremoval tank, a primary clarifier and one or two biological stages with activated sludge and a denitrification stage. After settling of sludge, the water is discharged into the receiving river. All WWTPs remove the phosphorus from the wastewater by chemical flocculation. The WWTPs 1 and 2 are equipped with one biological treatment stage, while the WWTPs 3 and 4 with two biological treatment stages. WWTP 1 has an additional filtration unit before the treated water is discharged into the river. WWTP 3 has the highest capacity; it serves 750,000 inhabitants and handles approximately 92 million m3/a of

Table 1 e Wastewater treatment plants characteristics and sampling events. WWTP 1

WWTP 2

WWTP 3

WWTP 4

Population served Influent volume of sewage (m3/a) % indirect dischargers Sampling events

190,000 23,700,000 24 JaneFeb 2010

750,000 92,600,000 38 Nov 2008eNov 2009

220,000 24,400,000 65 FebeNov 2009

Sampling Mode

Time-prop. Time-prop.

140,000 16,000,000 30 AugeSep 2009 JaneFeb 2010 Flow-prop. Flow-prop.

Time-prop. Flow-prop.

Time-prop. Flow-prop.

Influent Effluent

Time-prop: time-proportional. Flow-prop: flow-proportional.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 1 3 e5 3 2 2

wastewater. Plant 1, has a lower capacity (190,000 inhabitants), plants 2 is the smallest (140,000) and plant 4 cleans the wastewater from a served population of 220,000. All of them treat mixed sewage from indirect dischargers and households. The WWTP 4 has the highest proportion of indirect dischargers 65% (415,000 population equivalents, PE), followed by WWTP 3 with 38% (460,000 PE), WWTP 2 with 30% (60,000 PE) and WWTP 1 with 24% (60,000 PE). These indirect dischargers are non-domestic sources of wastewater in municipal sewer systems and they can be commercial or industrial facilities. Some types of relevant industrial indirect dischargers are for example, the paper and metal processing industry together with the textile and paint industry. WWTP 3 receives two influents A and B, which correspond to different parts of the catchment area. The influent A represents approximately 57% and the influent B approximately 43% of the wastewater. In this study the TMDD concentrations in influent of WWTP 3 are reported for both influents A and B.

2.3.

Sampling

All analyzed samples were 24 h composite samples collected by refrigerated autosamplers from the effluents and influents on the same day in each WWTP. Different composite sampling modes were employed to collect the samples (Table 1). In the so called time proportional mode, samples of equal volume are taken at equal time increments. In the so called flow proportional mode, samples are collected at equal time increments additionally they are composited proportionally to the flow rate at the time of each sampling. 69 samples were collected from the WWTP 3 and 42 samples from the WWTP 4 during a time span of one year. The WWTP 2 was sampled in two sampling campaigns, AugeSep 2009 and JaneFeb 2010, leading to 112 samples overall. The WWTP 1 was sampled over a period of approximately 5 weeks (JaneFeb 2010) to provide a total of 66 samples. An influent of WWTP 3 was sampled on 04/26/10 and 13 samples were collected, 12 were 2 h composite samples and one was a 24 h composite sample. The samples were collected in 500 mL amber bottles, which were cleaned before use with distilled water and methanol followed by drying in the oven at 110  C for 2 h.

2.4.

Analytical methods

Immediately after collection, the samples were filtered in the laboratory to separate coarse and suspended solids from the wastewater samples. Pre-cleaned paper filters used for previous studies of river water (Guedez et al., 2010) provided low blank values but were not suitable for the filtration of wastewater due to the rapid obstruction of the filters with the suspended solids in the wastewater sample. Therefore, a variety of filters available for pressure filtration were evaluated in order to determine blank values and to obtain appropriate filters for filtration of the wastewater samples. Borosilicate glass filter type A/E without binder with a diameter of 142 mm and a pore size of 1 mm from Pall Corporation were finally used due to the high efficiency for the pressure

5315

filtration. Initially blank values of TMDD with the use of selected filters without pre-treatment were in the range of 1500 ng/L. In order to lower the blank values, different types of filter pre-extraction methods were tested. Soxhlet extraction and ultrasonic extraction were tested by use of different solvents such as acetone, methanol and dichloromethane. Despite of pre-extraction, the blank values of TMDD were still high between 300 and 1000 ng/L. Therefore, borosilicate glass filters were cleaned with dichloromethane for 1 h and then heated to 400  C for 2 h. This procedure resulted in a decrease of the mean blank value for TMDD to 54 ng/L (n ¼ 10) with a standard deviation of 5 ng/L. The blank value was subtracted from TMDD concentrations determined in the wastewater samples. Procedural blanks were periodically performed for each sampling set in order to detect changes in the blank value. The black values varied during the time span from Nov 2008 till Jan 2010 between 40 ng/L and 60 ng/L and were continuously adjusted. Solid phase extraction (SPE) was carried out using Bond Elute PPL (1 ml, Varian) cartridges to extract the organic substances from the samples after the filtration. For the extraction, 80 mL and 200 mL of wastewater from the influents and effluents respectively were taken. The cartridges were conditioned using 1 mL methanol, 1 mL methanol/acetone (1:1) and 1 mL ultrapure water. After the extraction of the wastewater samples, the cartridges were dried under a nitrogen stream and then eluted with 3  333 mL of a mixture methanol/acetone (1:1). From the total extract, 1 mg was taken and dissolved in 400 mL acetone. Squalane (2 mg dissolved in 2 mL of hexane) was added as internal standard to each extract before GC/MS analysis. Recoveries, detection limit and determination limit for the method were determined by spiking 1 L distilled water (n ¼ 8) with standard solutions of TMDD varying between 5 ng/L and 1000 ng/L. Detection and quantitation limits were calculated in accordance with the DIN 32,645 (1994) on the basis of a measured calibration curve providing values of 12 ng/L and 36 ng/L respectively. No adjustments were made in regard to the SPE recovery rate (76%, n ¼ 5). TMDD was not detectable in distilled water used for determination of the recovery, detection limit and quantitation limit. The extracts were further analyzed by gas chromatography coupled to mass spectrometry (GC/MS). For GC separation a Thermo-Fisher TR-5MS column (30 m length, 0.25 mm inner diameter, and 0.25 mm film thickness) was used in a TRACE GC Ultra gas chromatograph coupled to an ITQ-900 mass spectrometer operating in the full scan mode (both Thermo Fisher). The samples were injected in the splitless mode with a splitless time of 1 min. The oven temperature program started at 80  C and the temperature was increased with 4  C/ min until the final temperature of 300  C. Helium was used as carrier gas with a constant velocity of 1.1 mL/min. Mass spectra were obtained in the electron ionization mode (EI) using an electron energy of 70eV. Data acquisition, processing and quantification were performed using Xcalibur software. The quantification of TMDD was carried out using the characteristic fragment m/z 109. The correction and response factors were determined as described in Fries and Pu¨ttmann (2003) for each batch of samples.

5316

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 1 3 e5 3 2 2

Table 2 e Influent and effluent concentrations (ng/L) of TMDD in 4 municipal wastewater treatment plants. WWTP 1

No. of samples Min Median Mean Max

WWTP 2

WWTP 3

WWTP 4

Influent

Effluent

Influent

Effluent

Influent

Effluent

Influent

Effluent

33 174 489 572 1747

33
56 250 1030 1266 3195

56 208 726 950 3539

46 205 1398 1877 5846

23 134 663 785 2256

21 134 992 1168 3628

21
3.

Results and discussion

3.1.

TMDD concentrations and loads in WWTPs

3.1.1.

TMDD concentrations in influents and effluents

effluents of WWTPs other important sources for TMDD in river Rhine must be present in order to explain the high TMDD concentrations.

3.1.2. TMDD was detected in all of the analyzed samples, except for two effluent samples. An overview of all influent and effluent concentrations determined at the four WWTPs is given in Table 2. The TMDD influent concentrations varied between 134 ng/L (WWTP 4) and 5846 ng/L (WWTP 3), and the effluent concentrations between
TMDD loads in influents and effluents

Loads of TMDD in all samples (Fig. 1, Table 3) were calculated from the TMDD concentrations and mean water discharges per day. The TMDD loads in the influents ranged from 10.1 g/d (WWTP 4) to 1142 g/d (WWTP 3) and in the effluents from
Fig. 1 e Influent and effluent loads of TMDD in the 4 WWTPs during the whole sampling period.

5317

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 1 3 e5 3 2 2

Table 3 e TMDD loads (g/d) in effluents and influents in 4 municipal wastewater treatment plants. WWTP 1

WWTP 2

WWTP 3

WWTP 4

Influent

Effluent

Influent

Effluent

Influent

Effluent

Influent

Effluent

33 12.8 29.8 42.9 191 10.3 34

33
56 12.6 46.5 51.7 106 12.3 33

56 8.90 29.6 34.8 103

46 77.3 270 432 1142 52.6 67

23 21.9 118 144 425

21 10.1 47.9 67.9 213 8.29 68

21
No. of samples Min Median Mean Max Mean Annual discharge (kg/a) % Elimination

of indirect industrial dischargers. Although WWTP 4 has the highest proportion of sewage water from indirect dischargers (65%), the mean influent load of TMDD is lower (67.9 g/d) than in WWTP 3 (432 g/d), where the proportion of sewage water from indirect dischargers is only 38%. The reason for this might be that most of the sewage handled in this plant (WWTP 4) comes from an indirect discharger which is not TMDD relevant, a sewage sludge dewatering and incineration plant, where the sludge from WWTP 3 and 4 is dewatered and incinerated. The resulting water is returned to the first biological stage of WWTP 4 and it is expected not to contain high TMDD concentrations due to its low tendency to accumulate in sludge. Therefore, the TMDD contribution of this indirect discharger to the wastewater should be negligible. The total discharges per year for each WWTP were estimated assuming that the mean effluent loads remain constant throughout the whole year (Table 3). Considering this approximation, the total discharge of TMDD from all four WWTPs together was 83.5 kg per year. Per-capita loads of TMDD in WWTPs influents were also calculated (Table 4). The mean per-capita load generated in all four WWTPs together was 0.38 mg/capita$day. The per-capita load of 1e3 WWTP provides a linear correlation with the proportion of wastewater contributed from indirect discharges which indicates a major external input of TMDD apart from the households into the sewage system. The correlation is not valid for WWTPs 4 because it has an irrelevant indirect discharge for TMDD generating a mean percapita load lower (0.31 mg/capita$day) than in WWTP 3 (0.59 mg/capita$day) in spite of its higher indirect discharger proportion.

April 26, 2009 (Sunday) at influent A of WWTP 3 in order to analyze the concentration and load variation of TMDD entering this plant over a time span of 24 h (day profile). One additional 24 h composite sample was collected on the same day, at the same sampling point. The sampling was carried out during a dry period to avoid a temporal dilution of the wastewater. Fig. 2 shows TMDD loads obtained for the day profile (n ¼ 12) and the 24 h composite sample. Additionally the volume of wastewater reaching the WWTP during the sampling period is shown with a resolution of 2 h. The TMDD concentrations varied from 383 ng/L (12:00 pme2:00 pm) to 2550 ng/L (8:00 pme10:00 pm) having a maximum of 2550 ng/L between 8:00 and 10:00pm (see Supplementary content). The TMDD loads were also calculated and they also show a high variability over the course of the day from 2.64 g/h to 14.8 g/h. The maximum influent load was also reached between 8:00 and 10:00pm (14.8 g/d). Based only on this information, it is difficult to define at what time TMDD is introduced into the sewage system by the users connected to the sewage system. The transportation time from the point of discharge to the WWTP varies due to the large catchment area of this WWTP. However, the variation of the TMDD loads is largely correlated to the variation of the amount of wastewater influent as indicated by the dotted band in (Fig. 2). There are two exceptions: between 12:00 and 2:00pm the load of TMDD is unexpectedly low and between 8:00 and 10:00pm the load is unexpectedly high. Despite of these inconsistencies the data indicate that private households contribute to the discharge of TMDD into the sewage systems.

3.1.3.2. Weekly variation profiles. To examine the weekly 3.1.3. Temporal variation of concentrations and loads of TMDD in WWTPs 3.1.3.1. Day

variation profiles. Twelve samples, each composed of influent water covering 2 h were collected on

Table 4 e Estimated per-capita input of TMDD (mg/ capita$day) based on data from 4 municipal WWTPs. WWTP

1

2

3

4

Min Median Mean Max

0.07 0.16 0.23 1.00

0.09 0.33 0.37 0.76

0.10 0.37 0.59 1.55

0.05 0.22 0.31 0.95

fluctuations of TMDD concentrations and loads, daily composite samples were collected over three weeks in the influent and in the effluent of WWTP 2. The samples were collected between 08/21/09e08/27/09 (one week) and 02/06/ 10e02/19/10 (two weeks). The concentrations (Supplementary content) varied between 250 ng/L and 2924 ng/L (mean 1111 ng/L) in the influent and between 208 ng/L and 1358 ng/L (mean 625 ng/L) in the effluent. The loads were also calculated, in terms of day to day variations. The influent loads varied from 12.6 g/d to 106 g/d (mean 49.1 g/d) and the effluent loads from 8.9 g/d to 65.2 g/d, (mean 27.2 g/d), (Fig. 3). A correlation between high loads and particular days was not found neither in the influents nor in the effluents. During the last two weeks of sampling, the variation of loads between influent and effluent samples show an offset of one day; high

5318

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 1 3 e5 3 2 2

Fig. 2 e Two-hourly loads of TMDD in sewage influent (3A) on April 26, 2009 (n = 13).

influent loads are followed by high effluent loads one day later (Fig. 3). The time difference between the load waves in influents and effluents can be attributed to the hydraulic retention time in the WWTP 2 and 3 which is approximately one day, in contrast to WWTP 1 and 4 where the retention time is around two days. This observed offset may be caused due to irregular

TMDD discharges to the sewage system by indirect industrial dischargers. In order to study the influence of the weekends on the TMDD input in the WWTPs, the concentrations and loads of TMDD in wastewater influents of WWTPs 2 and 3 were divided into data originating from working days (Monday-Friday) and weekends (Saturday, Sunday). For the

Fig. 3 e Variation of TMDD loads in WWTP 2 during three weeks (n = 21).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 1 3 e5 3 2 2

Table 5 e Concentrations and loads of TMDD in influents of the WWTP 2 and 3 during weekends and working days (WD). 2

2

Weekends WD Influent (ng/L) Load Influent (g/d)

1056 50.9

3A

3B

Weekends

1387 2161 58.8 504

2503

3A

3B

WD 1802 1554 401

effluents the concentrations and loads of WWTPs 2 and 3 were used for the calculation considering an offset of one day in case of working days (Tuesday-Saturday) and in case of weekends (Sunday-Monday) taking in account the retention time of the wastewater in these WWTPs. The results are presented in Table 5. Differences between influent loads during working days and weekends were low for the two WWTPs and the differences were not statistically significant (Welch t-test, a ¼ 0.05) for any of the two WWTPs. The results show that no systematic temporal trend or differences between mean weekends and working days load of TMDD is observed.

3.1.3.3. Seasonal variation profiles. To study seasonal variations of TMDD in WWTPs, influent and effluent samples were taken from the WWTPs 3 and 4 during a whole year and from the WWTP 2 during two periods, AugeSep. 2009 and JaneFeb. 2010. For the evaluation of seasonal variations only effluent loads are considered. The data of TMDD loads in effluents of the WWTPs 3 and 4 were divided into two groups; samples from the cold period (October to February) and samples from the warm period (March to September) and from WWTP 2 between AugeSep 2009 (warm period) and JaneFeb 2010 (cold period), (Table 6). A statistically significant load difference during warm and cold seasons was only observed in the WWTP 3 (Welch t-test, a ¼ 0.05) with a mean effluent TMDD load of 91.3 g/d in the cold season and of 164 g/d during the warm season. In WWTP 2 the load during the warm season (37.5 g/d) was only slightly higher than in the cold season (31.8 g/d). In WWTP 4 almost no difference between warm and cold season was observed (21.5 versus 21.7 g/d). These values do not show a significant discrepancy (t-test, a ¼ 0.05). Only for WWTP 3 these results show a significantly higher TMDD discharge during the warm season compared to the cold season. An annual load profile for TMDD in WWTPs 3 and 4 is showed in Fig. 4. The load for each month represents the mean value obtained from all samples collected during that month. Supplementary content provides information about the number of samples collected during each month and

Table 6 e Mean loads (g/d) in effluents of the WWTPs during cold and warm periods. WWTP Cold period Warm period

2

3

4

31.8 37.5

91.3 164

21.7 21.5

5319

TMDD loads in the samples. Tables 2 and 3 present the mean loads and concentrations of effluent and influent samples during the entire collection period. In the case of WWTP 3, high fluctuations in the influent samples were observed. Fig. 4 shows a pattern for the influent of WWTP 3 with three peaks of very high TMDD loads in March (1132 g/d), May (998 g/d) and October (530 g/d) 2009. The measured loads show an input of TMDD throughout the whole year in both WWTP 3 and 4, but the variation is high particularly for the influent of WWTP 3. Such high temporal load variations were not found in effluents. Lower variations of loads in effluents compared to influents can be expected due to mixing of wastewater during the treatment process in the WWTPs which will eliminate peak concentrations.

3.2.

Removal of TMDD in WWTPs

The influent loads during the whole sampling period were in all four WWTPs higher than the effluent loads (Fig. 1), showing that TMDD is partially degraded during the wastewater treatment process. The elimination rates for each WWTP were calculated with the mean effluent and influent loads from the whole period. The calculated mean elimination of TMDD varied between 33% (WWTP 2) and 68% (WWTP 4). The highest elimination rates for TMDD were measured in the WWTPs 3 and 4, with values of 67% and 68% respectively while the elimination rates of WWTPs 1 and 2 were much lower with values of 34% and 33% respectively. A comparison of other surfactants, the removal rate of TMDD during the wastewater treatment is quite low. Studies about the elimination rate of surfactants reported values of 99% for LAS (linear alkylbenzene sulfonates), (Mungray and Kumar, 2009; Waters and Feijte, 1995) and between 92% and 97% for nonylphenol polyethoxylates (NPE), (Fytianos et al., 1997). However, similar elimination rates of 27% and 59% were measured for the surfactants perfluorooctanoate (PFOA) and perfluorodecanoate (PFDA) respectively. Regarding the differences found between the elimination rates of TMDD in the studied WWTPs, there are factors, which might be influencing the efficiency of the treatment. One factor influencing the high elimination rates in plants 3 and 4 in contrast to WWTP 1 and 2 is that these WWTPs are equipped with two biological stages, which enable a higher biodegradation rate of TMDD, while WWTPs 1 and 2 are composed only of one biological stage. Another factor is the sampling mode which was different in the influents (time-proportional) and in the effluents (flowproportional) for the WWTP 3 and 4, in contrast to the other studied WWTPs, where the same sampling mode was used both in the influent and in the effluent (in WWTP 1, timeproportional and in WWTP 2, flow-proportional). The sampling mode might have an influence on the measured loads of TMDD used to calculate the removal rate. Previous research has showed that the flow-proportional mode is the most appropriate sampling mode because it weights the individual subsamples in the 24 h composite sample according to the flow in the sewer (Ort et al., 2010), while the timeproportional mode tends either to overestimate or underestimate pollutant loads (Ort et al., 2010; Stone et al., 1999). Therefore, the difference of the sampling mode in the influent

5320

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 1 3 e5 3 2 2

Fig. 4 e TMDD loads in influents and effluents during a year in the WWTPs 3 (n = 69) and 4 (n = 42).

and in the effluent of WWTP 3 and 4 might cause a deviation of the calculated loads and elimination rates, which can result in higher removal rates. However, both influent and effluent samples from the WWTP 2 were collected in flow-proportional mode, which is the most appropriate method for calculating loads (Ort et al., 2010). Therefore, the elimination rate of TMDD in the WWTP 2 (33%) shows the most reliable value. Furthermore, the samples from the WWTP 2 were collected over a number of consecutive days decreasing the error resulting during the sampling campaign. The reason is that usually in the effluent sample water packets collected on the same day in the influent sample will be missed due to the hydraulic retention time of the WWTP. However, they will be captured in the effluent sample of the next day by a consecutive sampling. Therefore, water packets will be missed only during the first and last day of sampling. In contrast samples at WWTPs 3 and 4 were not collected on consecutive days. The only reference values for TMDD’s elimination rates in WWTPs were calculated using QSAR (quantitative structureactivity relationship) by Air Products (2002). According to QSAR the total removal of TMDD in WWTPs can be expected to be 4.35%. This value is fairly lower than the degradation rate obtained from the present study. It was also predicted that TMDD will be preferably eliminated from wastewater by sorption to sludge (Air Products and Chemicals, 2002). In order to test the validity of this assumption 6 sludge samples (grab samples) which were collected twice (11/19/09 and 12/01/09) from WWTP 3 were analyzed in the present study. TMDD concentrations were in all 6 samples below the quantitation limit (32 ng/L). This indicates that TMDD shows no tendency to accumulate in sludge, which is consistent with its low log Kow (2.8). Furthermore, its very low Henry’s law constant of

8.58  107 atm m3/mol prevents migration of TMDD from the aquatic phase to air. These results provide evidence that the elimination of TMDD in WWTPs is caused primarily by biodegradation and not by adsorption to sludge or evaporation.

3.3.

Sources of TMDD

The sources of TMDD in the sewage system are still unknown, but due to the permanent input into the WWTPs the concentrations remain constant on a high level. This argues for an origin of TMDD from both household wastewaters and indirect industrial dischargers. The importance of households as source of TMDD and its contribution to the input in the WWTPs has not yet been studied. However, due to the broad applications of TMDD in industrial processes its presence cannot be ruled out from private goods. Canellas et al. (2010) detected TMDD in 2 adhesives, which are used in laminates for food packaging and showed its capacity of migration from the packages into food. Moreover, TMDD was found as a contaminant in water beverage carton packages and it was suggested that TMDD can migrate from the printing ink used for the labels of the package into the water (Kleinschnitz and Schreier, 1998). According to the current knowledge it is not possible to explain the origin of the high amounts of TMDD in the influents of WWTPs only from private household wastewater. TMDD was found in 6 recycling toilet papers manufactured in Germany, Australia and China with concentrations between 0.16 mg/kg and 2.39 mg/kg (Gehring et al., 2009). To calculate the mean influent load of TMDD from toilet paper it can be assumed that the TMDD concentration in the toilet paper is 2.39 mg/kg and that the paper

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 1 3 e5 3 2 2

capital consumption of toilet paper in Germany is 15 kg per year (WWF, 2010). The mean influent loads of TMDD caused only by the toilet paper were calculated to be 18.7 g/d (WWTP 1), 13.8 g/d (WWTP 2), 73.7 g/d (WWTP 3) and 21.6 g/d (WWTP 4). These values are much lower than those found in this study (Table 1) suggesting that the introduction of TMDD into the WWTPs due to leaching of toilet paper cannot be the dominant pathway. Furthermore, TMDD is not used as a detergent in household cleaning products. Therefore the observed high concentrations of TMDD in sewage water (w1000 ng/L) can only partly be attributed to effluents from households. Indirect industrial dischargers have to be considered as additional and possibly dominating sources of TMDD in WWTPs, but these sources have not been analyzed in detail so far. In one study, very high concentrations of TMDD were detected in the sewage of a paper mill using a high proportion of waste paper. The occurrence of TMDD in the sewage is explained by the leaching of TMDD during recycling of paper impregnated with printing color inks in the so called pulping process (Schwarzbauer et al., 2010). Additional studies have to be carried out to further explore direct and indirect discharge of sewage by paper mills. Due to the widespread use of TMDD in colors and inks, industries producing these types of products might have to be investigated as candidates for emission of TMDD into the sewage systems. Apart from printing inks TMDD is also used as wetting agent in waterborne coatings for wood, plastic, etc. and as emulsifier in agricultural chemicals (Air Products and Chemicals, 2002). These applications represent additional possible sources for TMDD in surface waters. Another possible source of TMDD in sewage treatment plants might be the cleavage of the TMDD-ethoxylates (Surfynol 440 and 440) similar to the cleavage of alkylphenol polyethoxylate to form alkylphenols (Ahel et al., 1994). However, the in situ formation of TMDD from the precursor molecules is not expected to be significant due to lower loads found in the effluents compared to the influents (Fig. 1). In contrast to nonylphenols, TMDD will not accumulate in sewage sludge due to its low log Kow value of 2.8. This was confirmed by the analysis of six sewage sludge samples from WWTP 3. In all samples TMDD was not detectable. This confirms that the effluents are the only pathway for the nonbiodegraded proportion of TMDD to leave the WWTPs. Since TMDD is not accumulating in sludge one might expect higher effluent loads than influent loads in case that the cleavage of ethoxylates of TMDD is a process of significance.

4.

Conclusion

This first comprehensive investigation on TMDD in municipal WWTPs has shown that TMDD is found in wastewater and that is not completely removed during the wastewater treatment. The removal efficiency of TMDD varies from 33% to 68% demonstrating that municipal WWTPs are a dominating source of TMDD in rivers. The elimination of TMDD from the wastewater can be attributed to biological degradation and not to adsorption to activated sludge. High temporal variations (daily and weekly) of the influent TMDD concentrations

5321

argue for an input of TMDD into the WWTPs through wastewater from households and indirect industrial dischargers. Therefore, more studies are needed to identify the sources of TMDD in the sewage. Due to the applications of TMDD in coatings, paper industries and producers of printing inks should be considered as possible sources for future research. Furthermore, to better understand the removing mechanisms during the wastewater treatment, a more detailed study should be carried out in the different stages of a WWTP.

Acknowledgments We would like to acknowledge the German Academic Exchange Service (DAAD) for financial support of the PhD project of Arlen A. Guedez. The support of the staff from the four wastewater treatment plants for sampling is acknowledged specially Dr.-Ing. Rolf Go¨tz and Elke Seyffer from WWTPs 3 and 4. The authors thank Werner Haunold and Robert Sitals for their support with transportation of samples.

Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.07.038.

references

Ahel, M., Giger, W., Koch, M., 1994. Behaviour of alkylphenol polyethoxylate surfactants in the aquatic environment I. Occurrence and transformation in sewage treatment. Water Research 28 (5), 1131e1142. Air Products and Chemicals, I, 2002. Robust Summary for 2,4,7,9Tetramethyl-5-decyne-4, 7-diol. Air Products and Chemicals, I, 2008. Material Safety Data Sheet Pigments. Canellas, E., Aznar, M., Nerin, C., Mercea, P., 2010. Partition and diffusion of volatile compounds from acrylic adhesives used for food packaging multilayers manufacturing. Journal of Materials Chemistry 20 (24), 5100e5109. DIN 32645, 1994. Detection Limit, Identification Limit and Determination Limit (Nachweis-, Erfassungs- und Bestimmungsgrenze). Deutsches Institut fu¨r Normierung. Dsikowitzky, L., Schwarzbauer, J., Kronimus, A., Littke, R., 2004a. The anthropogenic contribution to the organic load of the Lippe River (Germany). Part I: qualitative characterisation of low-molecular weight organic compounds. Chemosphere 57 (10), 1275e1288. Dsikowitzky, L., Schwarzbauer, J., Littke, R., 2004b. The anthropogenic contribution to the organic load of the Lippe River (Germany). Part II: quantification of specific organic contaminants. Chemosphere 57 (10), 1289e1300. EPA, 2011. High Production Volume Information System (HPVIS). Fries, E., Puttmann, W., 2003. Occurrence and behaviour of 4nonylphenol in river water of Germany. Journal of Environmental Monitoring 5 (4), 598e603. Fytianos, K., Pegiadou, S., Raikos, N., Eleftheriadis, I., Tsoukali, H., 1997. Determination of non-ionic surfactants (polyethoxylated-nonylphenols) by HPLC in waste waters. Chemosphere 35 (7), 1423e1429.

5322

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 1 3 e5 3 2 2

Gehring, M., Vogel, D., Bern, B., 2009. Concentrations of 2,4,7,9Tetramethyl-5-decin-4,7-diol (TMDD) and the endocrine disrupting substances bisphenol A (BPA), 4-tert-octylphenol and pentachlorophenol in recycling toilet paper from different countries (Belastung von Recycling-Toilettenpapier aus verschiedenen La¨ndern mit 2,4,7,9-Tetramethyl-5-decin-4,7diol (TMDD) und den endokrin aktiven Stoffen Bisphenol A, 4tert-Octylphenol, technischem 4-Nonylphenol und Pentachlorphenol). 4. Dresdner Symposium "Endokrin Aktive Stoffe in Abwasser, Kla¨rschlamm und Abfa¨llen". Schriftenreihe fu¨r Abfallwirtschaft und Altlasten, pp. 91e105. Guedez, A.A., Fro¨mmel, S., Diehl, P., Pu¨ttmann, W., 2010. Occurrence and temporal variations of TMDD in the river Rhine, Germany. Environmental Science and Pollution Research 17 (2), 321e330. HLUG, 2004. Hessian Agency for the Environment and Geology (Hessisches Landesamt fu¨r Umwelt und Geologie). Kleinschnitz, M., Schreier, P., 1998. Identification and semiquantitative determination of a migration contaminant from beverage carton packages into mineral water by on-line solid phase extraction gas chromatography-mass spectrometry (SPE-GC-MS). Chromatographia 48 (7), 581e583. LANUV, 2008. State Office for Nature. Environmental and Consumer Protection of North-Rhine-Westphalia (Landesamt fu¨r Natur Umwelt und Verbraucherschutz). http://luadb.lds. nrw.de/LUA/gues/welcome.htm. Loraine, G.A., Pettigrove, M.E., 2005. Seasonal variations in concentrations of pharmaceuticals and personal care products in drinking water and reclaimed wastewater in southern California. Environmental Science & Technology 40 (3), 687e695. Mungray, A.K., Kumar, P., 2009. Fate of linear alkylbenzene sulfonates in the environment: a review. International Biodeterioration and Biodegradation 63 (8), 981e987.

Ort, C., Lawrence, M.G., Reungoat, J., Mueller, J.F., 2010. Sampling for PPCPs in wastewater systems: comparison of different sampling modes and optimization Strategies. Environmental Science & Technology 44 (16), 6289e6296. Schwarzbauer, J., Heim, S., 2005. Lipophilic organic contaminants in the Rhine river, Germany. Water Research 39 (19), 4735e4748. Schwarzbauer, J., Illlguth, S., Botalova, O., 2010. Organic contaminants in process wastewater from a recycling paper mill (Organische Kontaminanten in Prozess- und Abwasser eines Altpapier-verarbeitenden Betriebs) Umweltwiss. Schadstoff-Forschung 22 (4), 322e323. Stone, K.C., Hunt, P.G., Novak, J.M., Johnson, M.H., Watts, D.W., 1999. Flow-proportional, time-composited, and grab sample estimation of nitrogen export from an eastern coastal plain watershed. Transactions of the ASABE 43 (2), 281e290. Tensidverordnung, 1977. Regulation on the Biodegradability of Anionic and Non-Ionic Surfactants in Detergents and Cleaning Products Verordnung u¨ber die Abbaubarkeit anionischer und nichtionischer grenzfla¨chenaktiver Stoffe in Wash- und Reinigungsmitteln. Tensidverordnung-TensidV. p. 244. Waters, J., Feijte, T.C.J., 1995. AISþ/CESIOþ environmental surfactant monitoring programme: outcome of five national pilot studies on linear alkylbenzene sulphonate (LAS). Chemosphere 30 (10), 1939e1956. WRMG, 2007. Law on the environmental impact of detergents and cleaning products Gesetz u¨ber die Umweltvertra¨glichkeit von Wasch- und Reinigungsmitteln. Wasch- und Reinigungsmittelgesetz - WRMG. WWF, 2010. World Wide Fund for Nature. Zhang, Y., Geißen, S.-U., Gal, C., 2008. Carbamazepine and diclofenac: removal in wastewater treatment plants and occurrence in water bodies. Chemosphere 73 (8), 1151e1161.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 2 3 e5 3 3 3

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Removal of persistent pharmaceutical micropollutants from sewage by addition of PAC in a sequential membrane bioreactor D. Serrano, S. Sua´rez, J.M. Lema, F. Omil* Department of Chemical Engineering, School of Engineering, University of Santiago de Compostela, Rua Lope Gomez de Marzoa s/n E-15782, Spain

article info

abstract

Article history:

The performance of a membrane bioreactor operating in a sequential mode (SMBR) using

Received 4 January 2011

an external flat-plate membrane was investigated. After 200 days of operation, a single

Received in revised form

addition of 1 g L
1 Powdered Activated Carbon (PAC) was added directly into the mixed

20 July 2011

liquor in order to enhance the simultaneous removal of nutrients and pharmaceutical

Accepted 30 July 2011

micropollutants from synthetic urban wastewater. Throughout the entire operation (288

Available online 5 August 2011

days), Chemical Oxygen Demand (COD) removal efficiencies were up to 95%, ammonium

Keywords:

only high values (around 80%) after PAC addition. During the operation of the SMBR

Activated carbon

without PAC addition, micropollutants which exerted a more recalcitrant behaviour were

Pharmaceutical organic

carbamazepine, diazepam, diclofenac and trimethoprim, with no significant removal. On

nitrogen removal was maintained over 70e80%, whereas phosphorus removal achieved

micropollutants

the other hand, moderate removals (42e64%) were observed for naproxen and erythro-

Sequencing batch membrane

mycin, whereas ibuprofen, roxithromycin and fluoxetine were removed in the range of

bioreactor (SMBR)

71e97%. The addition of PAC into the aeration tank was a successful tool to improve the

Sorption

removal of the more recalcitrant compounds up to 85%. The highest removal with PAC was observed for carbamazepine, trimethoprim as well as for roxithromycin, erythromycin and fluoxetine. The latter four compounds have amine groups and pKa in the range 6.7e10.1, thus the interaction between PAC and the positively charged amino groups might be the cause of their comparatively better results. Microbial ecology present in the biomass showed a higher abundance of Accumulibacter phosphatis as well as the ammonium oxidizing bacteria belonging to the genus Nitrosomonas after PAC addition. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Due to the incomplete removal of the most recalcitrant organic micropollutants detected in urban wastewaters during conventional treatment processes (Castiglioni et al.,

2006; Lindqvist et al., 2005) it is necessary to develop new treatment concepts to adequately address this problem. Among the different studies carried out in the last decade, different strategies appear of interest such as the implementation of anoxic-aerobic processes for nitrogen removal;

* Corresponding author. E-mail addresses: [email protected] (D. Serrano), [email protected] (S. Sua´rez), [email protected] (J.M. Lema), [email protected] (F. Omil). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.037

5324

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 2 3 e5 3 3 3

the use of Membrane Biological Reactors (MBRs); or the feeding of additives such as adsorbents directly into the aeration tank of activated sludge processes. Different works have reported that alternating anoxicaerobic processes, as existing in modern Sewage Treatment Plants (STPs) designed for biological nitrogen removal, can be useful to improve the degradation of certain micropollutants (Suarez et al., 2010; Andersen et al., 2003). Among others, Sequential Batch Reactors (SBRs) constitute an interesting alternative in which anoxic and aerobic conditions can be easily combined and controlled in a single stage. Some of their main features are their low footprint (secondary settler not needed) as well as their ability to face shock loads and the operation even under the presence of inhibitory substances (Sirianuntapiboon and Ungkaprasatcha, 2007; Andreottola et al., 2001). Membrane Biological Reactors (MBR) have gained significant popularity in STPs and are nowadays considered as a powerful (and expensive) technology able to produce higher quality effluents in terms of conventional pollutants, which can be appropriate for direct discharge, further postreatment or even reuse purposes. However, since membrane filtration does not enhance the elimination of most micropollutants by means of a size-exclusion mechanism it is still not clear if these systems may effectively enhance the removal of organic micropollutants (Reif et al., 2008). Powdered and Granular Activated Carbon (PAC and GAC) have been commonly used for sorption of organic micropollutants like pesticides or taste and odor compounds (Ternes and Joss, 2006). Previous studies carried out in samples with no or extremely low organic matter content (ultra-pure water or samples from Drinking Water Plants, respectively) reported successful removal of a wide number of complex and recalcitrant micropollutants including Pharmaceutical and Personal Care Products (PPCPs) and Endocrine Disrupting Compounds (EDCs) commonly found in urban wastewaters (Snyder et al., 2007; Westerhoff et al., 2005). However, activated carbon addition is not common in STPs. With activated sludge processes, Ng and Stenstrom (1987) showed that the use of 0.5e4 g L
1 of PAC may enhance nitrification rates by 75e97%, whereas other authors observed an improvement of organic matter removal as well as a significant decrease of toxicity caused by certain inhibitors on the nitrification process (Widjaja et al., 2004). In fact, activated carbon is a suitable support for bacterial attachment, being possible in this way to enhance the retention of the more slowly growing bacteria, such as nitrifiers (Thuy and Visvanathan, 2006; Aktas and Cecen, 2001). Previous studies carried out by our group (Serrano et al., 2010) showed that a GAC addition of 0.5e1 g L
1 directly into the aeration tank of an activated sludge reactor can be a useful tool to increase the removal of the recalcitrant PPCPs carbamazepine, diazepam and diclofenac. Moreover, recent works have shown that activated carbon can be useful to minimise fouling problems in MBRs. In this way, the use of PAC concentrations of 0.5e3 g L
1 inside the aeration tank of a MBR have been used to attain an easier control of membrane fouling (Remy et al., 2009). The aim of this work was to assess the removal of selected pharmaceutical micropollutants contained in synthetic municipal wastewaters using a Sequential Membrane Batch

Reactor (SMBR) which comprises an SBR unit coupled with an external MF membrane. Moreover, the addition of PAC directly into the aeration tank will be assessed as a tool to enhance the removal of the more recalcitrant compounds. In this way, the aim was to combine the potential of PAC (sorption of PPCPs and development of a much more diverse microbial ecology) with the advantages of MBRs (higher SRT and sludge concentrations, complete retention of biomass and PAC) in the removal of the target compounds.

2.

Materials and methods

2.1.

Experimental set-up

The experiments were performed in a membrane bioreactor with a sequential mode of operation (SMBR) using an external flat-plate membrane. The reactor consisted of two stainless steel units: an SBR, with a working volume of 30 L, and an 18 L vessel which contained the MF flat membrane (Kubota). Recirculation of biomass from the external membrane chamber to the SBR was carried out in order to achieve an external recirculation ratio of 0.23. The membrane has an effective filtration area of 0.1 m2 and a nominal pore size of 0.4 mm. Air was supplied by distributors placed at the bottom of the SBR and under the membrane module in the second unit. The air flow rate provided oxygen to both units (>2 mg L
1) and induced shear stress around the membrane to minimise fouling. The membrane was operated intermittently (6.25 min suction and 1.25 min relaxation). A constant flow was continuously maintained. When the transmembrane pressure (TMP) exceeded 15 kPa a physical cleaning using pressurized water was performed. The reactor was operated during 288 days divided in three periods. Since no sludge purge was performed, Sludge Retention Time (SRT) increased with time of operation. The start-up period lasted 104 days (P1), until stable conditions were achieved, during which no PPCPs were added to the SMBR. From day 105 onwards (P2), the feeding solution was spiked with the selected PPCPs (Table 1) and during the following 96 days the performance of the SMBR was monitored. The selection of the compounds has been based on several criteria previously indicated (Sua´rez et al., 2008). At the start of P3, which lasted 86 days, the SMBR received a single addition of PAC at a concentration of 1 g L
1. Commercial PAC QP (code 211237) was purchased from PANREAC and presented the following characteristics: 1.665 g cm
3 real density, 0.25 g cm
3 apparent density, 328.2 m2 g
1 specific surface area. A programmable logic controller was used to operate the reactor using a 6 h cycle with the following phase sequence: filling, 8 min; anoxic reaction, 93 min; aerobic reaction, 236 min; settling, 15 min and effluent withdrawal, 8 min. The SBR unit was operated under anoxic-aerobic conditions, with a Volume Exchange Ratio (VER) of 25% after each cycle. The discharged effluent was fed to the membrane unit, from which a continuous permeate was produced. The SMBR was seeded with 1.40 g VSS L
1 of aerobic sludge obtained from STP located in NW Spain which treats municipal wastewater. No sludge was taken off from the system during the whole experimental periods except for sampling. The reactor was fed with a synthetic mixture with similar

5325

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 2 3 e5 3 3 3

Table 1 e Theoretical concentrations of PPCPs spiked to the SMBR feeding (Cfeed in mg LL1), limits of quantification (LOQ in ng LL1) and percentage recoveries (standard desviations percentages are given in parentheses); and the reported physicochemical characteristics: dissociation constant (pKa), octanol-water partition coefficient (log Kow), solid-water distribution coefficient (Kd in kg LL1) and pseudo first-order degradation constant (kbiol in L gL1SS dL1). Compound

Anti-depressant Fluoxetine (FLX) Anti-inflammatories Ibuprofen (IBP) Naproxen (NPX) Diclofenac (DCF) Anti-epileptic Carbamazepine (CBZ) Antibiotics Trimethoprim (TMP) Roxithromycin (ROX) Erythromycin (ERY) Tranquillizer Diazepam (DZP) a b c d e f

Cfeed

20

LOQ

1.2

Recovery (n ¼ 4)

pKa

log Kow

log Kd

Influent

Effluent

70.4 (8)

71.8 (8)

10.1

4.0

0.7a,b

kbiolf

e

10 10 10

30 30 120

103.2 (8) 91.1 (7) 89.0 (8)

97.4 (8) 83.2 (9) 75.5 (8)

4.9e5.2 4.2 4.1e4.2

3.5e4.5 3.2 4.5e4.8

0.9c 1.1d 1.2c

9e35 0.4e1.9 <0.1

20

480

84.5 (7)

74.2 (8)

<1, >7

2.3e2.5

0.1c

<0.01

98.5 (5) 83.4 (4) 97.0 (4)

107.3 (5) 86.3 (11) 93.6 (5)

6.6e7.2 9.2 8.9

0.9e1.4 2.1e2.8 2.5e3.0

2.3a 2.2e 2.2a

e <0.3 0.5e1

97.3 (12)

87.6 (6)

3.3e3.4

2.5e3.0

1.3c

<0.03

10 10 10 20

6 1.2 1.2 240

Jones et al. (2002). Brooks et al. (2003). Ternes et al. (2004). Urase and Kikuta (2005). Joss et al. (2005). Sua´rez et al. (2008).

characteristics to a high-strength urban wastewater: 1000 mg L
1 COD (as NaCH3CO2), 80 mg L
1 of N
NHþ 4 and and a solution of trace elements (FeCl 8 mg L
1 P
PO
3 3, 4 H3BO3, CuSO4, KI, ZnSO4, CoCl2, MnCl2 at concentrations in the range of 3e150 mg L
1). An inlet flow rate of 30 L d
1 was applied, which corresponds to a hydraulic residence time of 1 d. Samples were taken twice a week from the influent, permeate and mixed liquor in the middle of the aerobic phase of the SMBR in order to analyse conventional parameters. For PPCPs, 4 sampling campaigns were carried out during P2 and 5 during P3, at 2e3 weeks intervals.

2.2.

Analytical methods

Samples collected from the reactor were analysed for conventional physicalechemical parameters (COD, TSS, VSS,

3 Sludge Volume Index (SVI), N
NHþ 4 , N
NO3 and P
PO4 ) according to standard methods (APHA, 1999). Additionally, sampling campaigns of the soluble concentrations of PPCPs in the influent and permeate were performed along P2 and P3. The composition of each sample was time-proportional along two cycles. Immediately after collecting 1 L of samples in aluminium bottles, each sample was filtered through a 0.45 mm glass fiber filter. For the analysis of PPCPs sample extraction based on Solid-Phase Extraction (SPE) was used as pre-concentration technique prior to their quantitative determination. Liquid or Gas Chromatography coupled with Mass Spectrometry (LC-MS or GC-MS, respectively) was used for the final quantification. Analysis of the soluble content of anti-inflammatory compounds, carbamazepine (CBZ) and diazepam (DZP) was performed following the methodology described in Rodrı´guez et al. (2003), which consists of adjusting the pH of the samples to 2.5, adding

meclofenamic acid and dihydrocarbamazepine as surrogate standards, carrying out a SPE of 250 mL samples using 60 mg OASIS HLB cartridges (Waters, Milford, MA, USA) and a final elution from the cartridge using 3 mL of ethyl acetate. This extract was divided into two fractions: one of them was used for direct determination of CBZ and DZP, while the other one was employed for the analysis of anti-inflammatories as their tert-butyldimethylsilyl derivatives. Finally, GC/MS detection was carried out in a Varian CP 3900 chromatograph (Walnut Creek, CA, USA) equipped with a splitesplitless injector and connected to an ion-trap mass spectrometer. The chemical analysis for the determination of fluoxetine (FLX), trimethoprim (TMP), roxithromycin (RXT) and erythromycin (ERY) was performed as published by Vanderford et al. (2003). Solid-Phase Extraction (SPE) was done as described for the previous compounds, although the elution step was performed with a mixture of methanol (1.5 mL) and methyl tert-butyl ether (1.5 mL). Final detection was performed in an LC-MS-MS (analysed in an Agilent Liquid Chromatograph API 400 GI312A equipped with a binary pump and autosampler HTC-PAL and the detection was performed with a triple quadruple Mass Spectrometer) in the positive ESI mode. Quantification limits and recoveries of the analytical procedure are given in Table 1.

2.3.

Morphological observation and FISH analysis

Activated sludge was morphologically characterized by phase-contrast microscopy and several staining procedures. Gram, Neisser, Polyhydroxybutyrate (PHB) and sheath stains techniques were performed as described by Jenkins et al. (2004). Filaments were classified according to their morphology, cell inclusions, motility, staining reactions and

5326

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 2 3 e5 3 3 3

filament abundance was estimated using the criteria suggested by Ka¨mpfer and Wagner (2002). Size distribution of the sludge was analysed by laser diffraction technique. The floc evolution, shape and quality of the biomass developed were monitored by stereo-microscope observation at 100 magnifications. Microbial populations were determined by the Fluorescence In Situ Hybridization (FISH) technique. Biomass samples from the reactor were collected, disrupted and fixed, according to the procedure described by Amann et al. (1995), with 4% paraformaldehyde solution. Hybridization was performed at 46  C for 90 min adjusting formamide concentrations at different percentages shown in Table 2. The used probes for in situ hybridization were 500 labelled with the fluorochromes FITC and Cy3. Fluorescence signals of disrupted samples were recorded with an acquisition system coupled with an Axioskop 2 epifluorescence microscope (Axioskop 2 plus, Zeiss).

3.

Results and discussion

3.1.

Reactor performance evaluation

The SMBR was inoculated with 1.40 g VSS L
1, this inoculum had poor settling properties (SVI around 640 mL g
1 VSS). During the entire SMBR operation, temperature and pH were not controlled but monitored, being around 19e20  C and 8.4e8.7, respectively. The oxygen concentration in the SBR unit was maintained above 2 mg L
1 during aerobic phases and around 0.1 mg L
1 in anoxic stages. At the end of the 104 days of the start-up period (P1), in which no PPCPs were added to the system, COD, N and P removal efficiencies accounted up to 95%, 71% and 36%, respectively. The biomass in this stage was flocculent with spongy, irregular and bad settling properties (maximum SVI values up to 1000 mL g
1 VSS) due to the presence of a variety of filamentous microorganisms. Suspended biomass reached maximum concentration of 2.22 g VSS L
1 in the SMBR. After this initial stage, the feeding solution was spiked with the selected pharmaceuticals from concentrated stock solutions (2000 mg L
1) of individual compounds dissolved in methanol (antibiotics, antiinflammatories and fluoxetine) or acetone (carbamazepine and diazepam). During the following 96 days (P2) N and P removal

efficiencies were not affected, while COD removal efficiencies increased up to 98%. However, an increase of biomass concentration was observed, up to 3.30 g VSS L
1, although with only slightly better settling characteristics (SVI around 580 mL g
1 VSS). In this period, the most common filamentous bacteria were Flexibacter spp. type 1702 which is gram-negative and is responsible for bulking and poor settling properties (Ka¨mpfer and Wagner, 2002). The growth of these bacteria could be attributed to the relative lack of substrate and nutrients in the external filtration chamber, since biological reactions took place preferentially inside the SBR tank. Furthermore, the complete retention of biomass by the membrane and its recirculation into the SBR unit led to poor settling characteristics of biomass, as commonly reported for MBRs (Wang et al., 2010). From day 202 onwards (P3), SMBR was supplemented with a single addition of PAC at a concentration of 1 g L
1 in accordance with previous works (Serrano et al., 2010). Powdered AC was chosen instead of GAC in order to prevent potential erosion of the membrane. PAC addition significantly influenced the performance of the system, leading to increase of N and P removal efficiencies up to 81% and 80%, respectively. However, COD removal efficiency was no affected. Taking into account the operating VER, the maximum theoretical N removal would be 78%. The higher values obtained during P3 are due to the higher biomass growth observed, which represented 7.6% of total N removal. Phosphorus removal with activated carbon in activated sludge systems has not been deeply studied. However, a number of media have been tested as adsorbents for phosphorus removal such as sand in subsurface flow constructed reed beds (Arias et al., 2001), dolomite in pilot vertical-flow constructed wetlands (Prochaska and Zouboulis, 2006) and oyster shells in constructed wetland systems (Park and Polprasert, 2008). These works conclude that the high removal efficiencies observed for P is due to adsorption on these media. A significant increase of biomass concentration was observed during P3 at a constant increasing trend from 3.5 (including 1 g L
1 of PAC on day 200) up to 6 g VSS L
1 (day 250). Additionally, a significant enhancement in the biomass settling properties was measured (SVI always below 150 mL g
1 VSS). Other investigations have reported that activated sludge shows better settling properties after PAC addition, due to a lower compressibility of sludge flocs as well as a lower content of extracellular polymeric substances inside microbial floc

Table 2 e List of FISH probes used in this work. Probe EUB338I Alf1B Bet42a NEU653 Nit3 Ntspa712 Pae997 Nso1225 Nsv443 PAO651 PAO462 PAO0846

Sequence (50 / 30 )

% Formamide

Reference

GTC GCC TCC CGT AGG AGT CGT TCG YTC TGA GCC AG GCC TTC CCA CTT CGT TT CCC CTC TGC TGC ACT CTA CCT GTG CTC CAT GCT CCG CGC CTT CGC CAC CGG CCT TCC TCT GGA AAG TTC TCA GCA CGC CAT TGT ATT ACG TGT GA CCG TGA CCG TTT CGT TCC G CCC TCT GCC AAA CTC CAG CCG TCA TCT ACW CAG GGT ATT AAC GTTAGCTACGGCACTAAAAGG

20 20 35 40 40 50 0 35 30 35 35 35

Amann et al., 1995 Manz et al., 1992 Manz et al., 1992 Wagner et al., 1995 Wagner et al., 1996 Daims et al., 2001 Amann et al., 1996 Mobarry et al., 1996 Mobarry et al., 1996 Crocetti et al., 2000 Crocetti et al., 2000 Crocetti et al., 2000

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 2 3 e5 3 3 3

5327

Fig. 1 e Microscopic images of sludge flocs in different days of operation of SMBR: (A) day 76; (B) day 127 and (C) day 259. Scale: 2 mm.

(Satyawali and Balakrishman, 2009; Kim et al., 1998). The decrease in filamentous bacteria and the appearance of small and regular flocs are other characteristics that typically enhance the settling properties of sludge. As mentioned, the amount of filamentous bacteria detected in P3 was lower and corresponded mainly to the type 0041, which normally appears in systems with high SRT. The biomass developed after PAC addition not only presented better settling properties, but also led to the development of a cake layer which was easier to remove by physical cleaning, in this way minimizing fouling.

3.2.

Morphological observation and FISH analysis

Particle size distribution is an important parameter in MBRs since it affects the characteristics of the cake formed by the

rejected solids, thereby influencing the filtration process (Satyawali and Balakrishman, 2009). The mean floc diameter (d50 based on volume ratios) shows that the floc diameter decreased from 439 mm in P2 to 108 mm in P3. Floc size distribution between 79 and 1179 mm was observed during P2 whereas at the end of P3 ranged 23e302 mm. This shift in particle size distribution of the aerobic biomass to smaller sizes was associated with the addition of PAC. Li et al. (2005) observed the same behaviour in the reduction of floc size and particle size distributions after PAC addition in a submerged membrane bioreactor. Other investigations suggested that the presence of PAC shift the particle size distribution down to the lower region because of the attached growth of microorganisms onto the surface of PAC (Kim et al., 1998). Furthermore, microscopic observations

Fig. 2 e FISH images of samples in P3. (A) Ammonia-oxidizing bacteria are marked in yellow by superposition of red (BET42a) and green (NEU653); (B) Denitrifying bacteria are marked in pink by superposition of blue (DAPI) and red (Pae997); (C) Polyphosphate-accumulating bacteria are marked in pale yellow by superposition of blue (DAPI), green (EUB338I) and red (PAO462) (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

5328

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 2 3 e5 3 3 3

(Fig. 1) of the sludge showed remarkable differences in the biomass, especially before and after the addition of PAC in the aeration tank. The evolution of microbial populations in the SMBR was followed by FISH analysis. A set of general probes, a-Proteobacteria (ALF1b) and b-Proteobacteria (BET42A), were applied in combination with the general eubacteria domain (EUB3381) to detect the main bacteria involved in the process. In P2, bProteobacteria were poorly observed since only few positive results were detected with specific probes for ammoniaoxidizing bacteria Nitrosomonas (Neu653) and Accumulibacter bacteria (PAO651). In the case of a-Proteobacteria, it was evaluated with a specific probe (Nit3) for the identification of nitrite oxidizing bacteria Nitrobacter (NOB), but no positive results were detected. Furthermore, denitrifying bacteria Pseudomonas spp. (Pae997) gave few positive results. In P3, a much more diverse biocenosis was observed: i) an increase in b-proteobacteria population belonging to halophilic and halotolerant Nitrosomonas spp. using probes Neu653 (Fig. 2); ii) an increase of the positive results were found for Accumulibacter bacteria (Fig. 2) with different probes (PAO651, PAO0846 and PAO462); iii) few positive results were detected with probes for NOB, Nit3 and Ntspa712, and finally; iv) a higher density of denitrifying bacteria was observed when using probes PAE997. The results indicate a slight enrichment

in nitrifying and denitrifying bacteria, which allow the system to reach the theoretical maximum of nitrogen removal efficiency. Additionally, the detection of PAO indicates that the enhancement in phosphorus removal during P3 might be attributed not only to sorption onto PAC, but also to the accumulation within the cells of Accumulibacter as polyphosphate. As previously indicated, the enhancement of biomass-solution contact caused by PAC addition appears to be the key point not only for increasing biomass but also to develop a much more complex and diverse microbial ecology in the system.

3.3.

PPCPs removal

3.3.1.

Carbamazepine and diazepam

Considering the concentration profiles for CBZ and DZP in the liquid phase from the influent to the effluent in P2, no significant removals were obtained (<20%). These results are consistent with the low removal efficiencies reported for these compounds when using membrane bioreactors or conventional processes (Suarez et al., 2010; Ying et al., 2009; Reif et al., 2008) and are related to the low lipophilicity (log Kow 2.3e3.0) and the hardly biodegradable character (kbiol < 0.03 L gSS
1 d
1) of CBZ and DZP. However, high removal efficiencies of up to 90% during P3 for both compounds were observed,

Fig. 3 e Concentrations of CBZ and DZP in the influent and effluent of SMBR.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 2 3 e5 3 3 3

which was associated with the PAC addition. In a previous work (Serrano et al., 2010) it was observed that a single addition of 1 g L
1 of GAC into the aeration tank of a CAS system led to removal efficiencies up to 43% for CBZ and around 35% for DZP. Li et al. (2011) also reported that the addition of 1 g L
1 of PAC into an MBR treating sewage was a successful tool to remove CBZ up to 92  15% during three weeks of operation, indicating that hydrophobicity, loading and PAC dosage were the key factors influencing its removal.

5329

Fig. 3 shows that the measured removal efficiencies for CBZ and DZP decreased after around three months of operation with PAC (P3). This result is probably due to saturation of the active pores of PAC by organic matter which competes with PPCPs and also by bacteria that grow on PAC. Other studies already pointed out that site competition and pore blockage are two mechanisms involved in the reduction of sorption capacity of target compounds when organic matter is present (Snyder et al., 2007; Fukuhara et al., 2006).

Fig. 4 e Concentrations of IBP, NPX and DCF in the influent and effluent of SMBR.

5330

3.3.2.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 2 3 e5 3 3 3

Ibuprofen, naproxen and diclofenac

After three weeks of operation with PPCPs, medium removal efficiencies were observed for IBP (51%), which were enhanced to almost complete removals (>90%) after day 140 (Fig. 4). The high removal of IBP during biological treatment has been confirmed in literature (Suarez et al., 2010; Kimura et al., 2007). Due to the low affinity for solids, no sorption onto sludge is expected for this compound, being the main removal mechanism biodegradation (kbiol 9e35 L g
1 SSV d
1). No removal was observed in the case of NPX during the first three weeks of operation with PPCPs, although similar to

IBP, after this initial adaptation period high removal efficiencies were determined (>90%) along P2 and P3 (Fig. 4). Similarly to IBP, the main removal mechanism for NPX is biodegradation. Other researchers observed that the development a wide microbial diversity, as indicated by the gradual increase of amonium biodegradation, might be the cause of those increases (Serrano et al., 2010; Suarez et al., 2010; Kimura et al., 2007). Moreover, no difference in the removal efficiencies was observed for both compounds between P2 and P3, which indicates that PAC has low affinity for these compounds.

Fig. 5 e Concentrations of ERY, ROX and TMP in the influent and effluent of SMBR.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 2 3 e5 3 3 3

5331

Fig. 6 e Concentrations of FLX in the influent and effluent of SMBR.

Similarly to CBZ and DZP, DCF was not significantly removed in the SMBR during P2, whereas PAC addition caused a very high removal, up to 93%. Unlike CBZ and DZP, for which removal efficiencies in the range of 80e90% were maintained for two months, in this case this level was only achieved during 1 month approximately (Fig. 4). These values are in the same range as those obtained previously after the addition of 0.5 g L
1 of GAC directly into the aeration tank of a CAS system (Serrano et al., 2010).

3.3.3.

Antibiotics

During P2 the three antibiotics used exhibited different behaviour: ROX was highly removed (71e86%), removals obtained for ERY were moderate (42e64%) whereas no significant removal was obtained for TMP. These results are considerably higher than those previously reported in fullscale installations (Go¨bel et al., 2007; Joss et al., 2005). Suarez et al. (2010) reported high eliminations for ERY and ROX under aerobic conditions (>89%) in a lab-scale CAS reactor, whereas TMP exhibited a recalcitrant behaviour, as indicated in the literature TMP (Castiglioni et al., 2006). Batt et al. (2006) concluded that TMP can be only degraded at a certain extent in activated sludge systems in which nitrifying bacteria are present. In fact, the lower amount of nitrifying bacteria detected using the FISH technique in P2 would explain the lack of removal detected. On the other hand, PAC addition caused an almost complete removal for the three antibiotics during P3 (Fig. 5), which was maintained during more than 2 months. In fact, PAC saturation rate appears to be a very slow process for these substances.

3.3.4.

Fluoxetine

The operation with the SMBR during P2 has shown to be successful to remove FLX at a high extent, in the range of 82e89%. Similar eliminations were previously reported by Suarez et al. (2010) and Zorita et al. (2009). However, there is a lack of conclusive information about the physicochemical characteristics of this compound which makes it difficult to draw conclusions about removal mechanisms.

According to Kwon and Armbrust (2006), FLX is quite persistent to biodegradation, so it is expected that the removal rates are more related to sorption onto solids. In this sense, the improvement observed during after PAC addition during P3, with removal rates in the liquid phase up to 98%, is very likely the result of the combined interaction between this compound and the biomass and the PAC (Fig. 6).

3.3.5. Comparative affinity for activated carbon and saturation As previously reported, although most of the compounds tested in this work can be significantly removed using activated carbon, their affinity degree for this adsorbent is different. Working with various natural waters, Westerhoff et al. (2005) showed that substances like IBP were poorly adsorbed to PAC (<20%), whereas TMP or FLX were highly removed (>90%). In this work, the highest removal efficiencies obtained after PAC addition corresponded to CBZ, TMP, ROX, ERY and FLX (97e99%). Among these compounds, the latter four have pKa values in the range 6.7e10.1, i.e. they are neutral or moderately alkaline. Taking into account that these compounds have amino groups in their structures that will be  protonated under neutral conditions forming cations (Babic et al., 2007), it is reasonable to explain the higher affinities exerted by these compounds because of the electrostatic interactions between the cations and the PAC. Other substances like DCF were highly removed after PAC addition, although its maximum removal efficiency was 93%, which was clearly decreasing in the following samples, thus showing a lower PAC affinity. The highest removal efficiency (99%) was achieved for the recalcitrant CBZ, which was maintained along more than two months of operation. In fact, saturation was only observed in the sampling carried out 80 days after PAC addition (CBZ removal efficiency of 14%). Saturation was also observed for other compounds such as DZP and TMP although with different kinetics, probably as the result of their different affinity for PAC and the competition for the available PAC adsorption sites with the other substances present in the mixed liquor.

5332

4.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 2 3 e5 3 3 3

Conclusions

A Sequential Membrane Biological Reactor (SMBR) comprising a SBR unit coupled with an external MF flat-plate membrane was operated to study the removal of selected PPCPs as well as conventional pollutants during the treatment of synthetic municipal wastewaters. After 200 days of operation moderate removals (42e64%) were observed for NPX and ERY, whereas IBP, ROX and FLX were removed in the range of 71e97%. During this period biomass showed poor settling characteristics as well as a low microbial diversity. After a single addition of 1 g/L of PAC directly into the aeration tank an immediate and sharp removal increase was observed for the more recalcitrant PPCPs not previously degraded at any significant extent: CBZ, DZP, DCF and TMP, with removal efficiencies in the range of 93e99%. Moreover, other substances which were moderately degraded such as ERY, ROX and FLX were completely removed after PAC addition (97e99%). The kinetics of saturation of PAC was different for each compound. An almost complete removal was maintained for CBZ along more than two months of operation, being saturation only observed 80 days after PAC addition (CBZ removal efficiency of 14%). Saturation was also observed for other compounds such as DZP and TMP although with different kinetics, probably as the result of their different affinity for PAC and the competition for the available PAC adsorption sites with the other substances present in the mixed liquor. Biomass characteristics were strongly affected by the presence of PAC. Physico-chemical properties such as conformation of agglomerates and settling were clearly enhanced (more compact flocs, lower SVI) which led to the development of a cake layer which was easier to remove by physical cleaning, in this way minimizing fouling. Moreover, microbial diversity was also positively affected, since nitrifiers were detected at higher amounts as well as phosphorus accumulating organisms (PAOs), which improved P removal (up to 80%).

Acknowledgements This work was supported by the Spanish Ministry of Education and Science (MICROFARM 2007-2010, CTQ2007-66265/PPQ, and NOVEDAR_Consolider 2007-2012, CSD2007-00055) and by the Galician Government (ESTRAFARM 2008-2011, PGIDIT08MDS005265PR), as well as by the Council of Science and Technology (CONACYT, Me´xico).

references

Aktas, O., Cecen, F., 2001. Addition of activated carbon to batch activated sludge reactors in the treatment of landfill leachate and domestic wastewater. Journal of Chemical Technology and Biotechnology 76 (8), 793e802. Amann, R., Ludwig, W., Schleifer, K.H., 1995. Phylogenetic identification and in situ detection of individual microbial

cells without cultivation. Microbiological Reviews 59 (1), 143e169. Amann, R., Ludwig, W., Schulze, R., Spring, S., Moore, E., Schleifer, K.H., 1996. rRNA-targeted oligonucleotide probes for the identification of genuine and former pseudomonads. Systematic and Applied Microbiology 19 (4), 501e509. Andersen, H., Siegrist, H., Halling-Sorensen, B., Ternes, T.A., 2003. Fate of estrogens in a municipal sewage treatment plant. Environmental Science & Technology 37, 4021e4026. Andreottola, G., Folodori, F., Ragazzi, M., 2001. On-line control for a SBR system for nitrogen removal from industrial wastewater. Water Science and Technology 43 (3), 93e100. APHA, Standard, 1999. Methods for the Examination of Water and Wastewater, twentieth ed. American Public Health Association/American Water Works Association/Water Environment Federation, Washington DC, USA. Arias, C.A., Bubba, M.D., Brix, H., 2001. Phosphorus removal by sands for use as media in surface flow constructed reed beds. Water Research 35 (5), 1159e1168. , S., Horvat, A.J.M., Pavlovic , D.M., Ka Babic stelan-Macan, M., 2007. Determination of pKa values of active pharmaceutical ingredients. Trends in Analytical Chemistry 26 (11), 1043e1061. Batt, A.L., Kim, S., Aga, D.S., 2006. Enhanced biodegradation of iopromide and trimethoprim in nitrifying activated sludge. Environmental Science & Technology 40 (23), 7367e7373. Brooks, B.W., Foran, C.M., Richards, S.M., Weston, J., Turner, P.K., Stanley, J.K., Solomon, K.R., Slattery, M., La Point, T.W., 2003. Aquatic ecotoxicology of fluoxetine. Toxicology Letters 142 (3), 169e183. Castiglioni, S., Zuccato, E., Crisci, E., Chiabrando, C., Fanelli, R., Bagnati, R., 2006. Identification and measurement of illicit drugs and their metabolites in urban wastewater by liquid chromatography-tandem mass spectrometry. Analytical Chemistry 78 (24), 8421e8429. Crocetti, G.R., Hugenholtz, P., Bond, P.L., Schuler, A., Keller, J., Jenkins, D., Blackall, L.L., 2000. Identification of polyphosphate-accumulating organisms and design of 16S rRNA-directed probes for their detection and quantification. Applied and Environmental Microbiology 66 (3), 1175e1182. Daims, H., Nielsen, J.L., Nielsen, P.H., Schleifer, K.H., Wagner, M., 2001. In situ characterization of nitrospira-like nitriteoxidizing bacteria active in wastewater treatment plants. Applied and Environmental Microbiology 67 (11), 5273e5284. Fukuhara, T., Iwasaki, S., Kawashima, M., Shinohara, O., Abe, I., 2006. Adsorbability of estrone and 17b-estradiol in water onto activated carbon. Water Research 40 (2), 241e248. Go¨bel, A., McArdell, C.S., Joss, A., Siegrist, H., Giger, W., 2007. Fate of sulfonamides, macrolides, and trimethoprim in different wastewater treatment technologies. Science of the Total Environment 372 (2e3), 361e371. Jenkins, D., Richard, G.M., Daigger, T.G., 2004. Manual on the Causes and Control of Activated Sludge Bulking and Foaming and Other Solids Separation Problems. Lewis Publishers, Washington, DC. Jones, O.A.H., Voulvoulis, N., Lester, J.N., 2002. Aquatic environmental assessment of the top 25 English prescription pharmaceuticals. Water Research 36 (20), 5013e5022. Joss, A., Keller, E., Alder, A.C., Go¨bel, A., McArdell, C.S., Ternes, T., Siegrist, H., 2005. Removal of pharmaceuticals and fragrances in biological wastewater treatment. Water Research 39 (14), 3139e3152. Ka¨mpfer, P., Wagner, M., 2002. Filamentous bacteria in activated sludge: current taxonomic status and ecology. In: Bitton, G. (Ed.), The Encyclopedia of Environmental Microbiology. Wiley, New York. Kim, J.S., Lee, C.H., Chun, H.D., 1998. Comparison of ultrafiltration characteristics between activated sludge and BAC sludge. Water Research 32 (11), 3443e3451.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 2 3 e5 3 3 3

Kimura, K., Hara, H., Watanabe, Y., 2007. Elimination of selected acidic pharmaceuticals from municipal wastewater by an activated sludge system and membrane bioreactors. Environmental Science & Technology 41 (10), 3708e3714. Kwon, J.W., Armbrust, K.L., 2006. Laboratory persistence and fate of fluoxetine in aquatic environments. Environmental Toxicology & Chemistry 25 (10), 2561e2568. Li, X., Hai, F.I., Nghiem, L.D., 2011. Simultaneous activated carbon adsorption within a membrane bioreactor for an enhanced micropollutant removal. Bioresource Technology 102 (9), 5319e5324. Li, Y.Z., He, Y.L., Liu, Y.H., Yang, S.C., Zhang, G.J., 2005. Comparison of the filtration characteristics between biological powdered activated carbon sludge and activated sludge in submerged membrane bioreactors. Desalination 174 (3), 305e314. Lindqvist, N., Tuhkanen, T., Kronberg, L., 2005. Occurrence of acidic pharmaceuticals in raw and treated sewages and in receiving waters. Water Research 39 (11), 2219e2228. Manz, W., Amann, R., Ludwig, W., Wagner, M., Schleifer, K.H., 1992. Phylogenetic oligodeoxynucleotide probes for the major subclasses of Proteobacteria: problems and solutions. Systematic and Applied Microbiology 15 (4), 593e600. Mobarry, B.K., Wagner, M., Urbain, V., Rittmann, B.E., Stahl, D.A., 1996. Phylogenetic probes for analyzing abundance and spatial organization of nitrifying bacteria. Applied and Environmental Microbiology 62 (6), 2156e2162. Ng, A.S., Stenstrom, M.K., 1987. Nitrification in powdered activated carbon activated sludge process. Journal of Environmental Engineering 113 (6), 1285e1301. Park, W.H., Polprasert, C., 2008. Roles of oyster shells in an integrated constructed wetland system designed for P removal. Ecological Engineering 34 (1), 50e56. Prochaska, C.A., Zouboulis, A.I., 2006. Removal of phosphates by pilot vertical-flow constructed wetlands using a mixture of sand and dolomite as substrate. Ecological Engineering 26 (3), 293e303. Reif, R., Sua´rez, S., Omil, F., Lema, J.M., 2008. Fate of pharmaceuticals and cosmetic ingredients during the operation of a MBR treating sewage. Desalination 221 (1e3), 511e517. Remy, M., Van der Marela, P., Zwijnenburg, A., Rulkens, W., Temmink, H., 2009. Low dose powdered activated carbon addition at high sludge retention times to reduce fouling in membrane bioreactors. Water Research 43 (2), 345e350. Rodrı´guez, I., Quintana, J.B., Carpinteiro, J., Carro, A.M., Lorenz, R. A., Cela, R., 2003. Determination of acidic drugs in sewage water by gas cromatography-mass spectrometry as tert -butyldimethylsilyl derivatives. Journal of Chromatography A 985 (1e2), 265e274. Satyawali, Y., Balakrishman, M., 2009. Effect of PAC addition on sludge properties in an MBR treating high strength wastewater. Water Research 43 (6), 1577e1588. Serrano, D., Lema, J.M., Omil, F., 2010. Influence of the employment of adsorption and coprecipitation agents for the removal of PPCPs in conventional activated sludge (CAS) system. Water Science and Technology 62 (3), 728e735. Sirianuntapiboon, S., Ungkaprasatcha, O., 2007. Removal of Pbþ2 and Niþ2 by bio-sludge in sequencing batch reactor (SBR) and

5333

granular activated carbon-SBR (GAC-SBR) system. Bioresource Technology 98 (14), 2749e2757. Snyder, S.A., Adham, S., Redding, A.M., Cannon, F.S., DeCarolis, J., Oppenheimer, J., 2007. Role of membranes and activated carbon in the removal of endocrine disruptors and pharmaceuticals. Desalination 202 (1e2), 156e181. Suarez, S., Lema, J.M., Omil, F., 2010. Removal of pharmaceutical and personal care products (PPCPs) under nitrifying and denitrigying conditions. Water Research 44 (10), 3214e3224. Sua´rez, S., Carballa, M., Omil, F., Lema, J.M., 2008. How are pharmaceutical and personal care products (PPCPs) removed from urban wasterwaters? Reviews in Environmental Science Biotechnology 7 (2), 125e138. Ternes, T.A., Herrmann, N., Bonerz, M., Knacker, T., Siegrist, H., Joss, A., 2004. Determination of Kd-values for pharmaceuticals and musk fragrances in sewage sludge. Water Research 38 (19), 4075e4084. Ternes, T.A., Joss, A., 2006. Human Pharmaceuticals, Hormones and Fragrances: The Challenge of Micropollutants in Urban Water Management. IWA, London. Thuy, Q.T.T., Visvanathan, C., 2006. Removal of inhibitory phenolic compounds by biological activated carbon coupled membrane bioreactor. Water Science and Technology 53 (11), 89e97. Urase, T., Kikuta, T., 2005. Separate estimation of adsorption and degradation of pharmaceutical substances and estrogens in the activated sludge process. Water Research 39 (7), 1289e1300. Vanderford, B.J., Pearson, R.A., Rexing, D.J., Snyder, S.A., 2003. Analysis of endocrine disruptors, pharmaceuticals, and personal care products in water using liquid chromatography/ tandem mass spectrometry. Analytical Chemistry 75 (22), 6265e6274. Wang, Z., Wang, P., Wang, Q., Wu, Z., Zhou, Q., Yang, D., 2010. Effective control of membrane fouling by filamentous bacteria in a submerged membrane bioreactor. Chemical Engineering Journal 158 (3), 608e615. Wagner, M., Rath, G., Amann, R., 1995. In situ identification of ammonia-oxidizing bacteria. Systematic and Applied Microbiology 18 (2), 251e264. Wagner, M., Rath, G., Koops, H.P., Flood, J., Amann, R., 1996. In situ analysis of nitrifying bacteria in sewage treatment plants. Water Science and Technology 34 (1e2), 237e244. Westerhoff, P., Yoon, Y., Snyder, S., Wert, E., 2005. Fate of endocrine-disruptor, pharmaceutical, and personal care product chemicals during simulated drinking water treatment processes. Environmental Science & Technology 39 (17), 6649e6663. Widjaja, T., Miyata, T., Nakano, Y., Nishijima, W., Okada, M., 2004. Adsorption capacity of powdered activated carbon for 3,5dichlorophenol in activated sludge. Chemosphere 57 (9), 1219e1224. Ying, G., Kookana, R.S., Kolpin, D.W., 2009. Occurrence and removal of pharmaceutically active compounds in sewage treatment plants with different technologies. Journal of Environmental Monitoring 11 (8), 1498e1505. Zorita, S., Martensson, L., Mathiasson, L., 2009. Occurrence and removal of pharmaceuticals in a municipal sewage treatment system in the south of Sweden. Science of the Total Environment 407 (8), 2760e2770.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 3 4 e5 3 4 2

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Fenton-like initiation of a toluene transformation mechanism Scott G. Huling a,*, Sangchul Hwang b,1, Dennis Fine c,2, Saebom Ko d,3 a

U.S. Environmental Protection Agency, Office of Research and Development, National Risk Management Research Laboratory, Robert S. Kerr Environmental Research Center, P.O. Box 1198, Ada, OK 74820, United States b University of Puerto Rico at Mayaguez, Department of Civil Engineering and Surveying, P.O. Box 9041, Call Box 9000, Mayaguez, PR 00681, United States c Shaw Environmental & Infrastructure, Inc., Robert S. Kerr Environmental Research Center, P.O. Box 1198, Ada, OK 74820, United States d National Research Council, Robert S. Kerr Environmental Research Center, P.O. Box 1198, Ada, OK 74820, United States

article info

abstract

Article history:

In Fenton-driven oxidation treatment systems, reaction intermediates derived from parent

Received 13 April 2011

compounds can play a significant role in the overall treatment process. Fenton-like reac-

Received in revised form

tions in the presence of toluene or benzene, involved a transformation mechanism that

13 July 2011

was highly efficient relative to the conventional Fenton-driven mechanism. A delay in

Accepted 2 August 2011

hydrogen peroxide (H2O2) reaction occurred until the complete or near-complete trans-

Available online 7 August 2011

formation of toluene or benzene and involved the simultaneous reaction of dissolved oxygen. This highly efficient transformation mechanism is initiated by Fenton-like reac-

Keywords:

tions, and therefore dependent on conventional Fenton-like parameters. Results indicated

H2O2

that several potential parameters and mechanisms did not play a significant role in the

Oxidation

transformation mechanism including electron shuttles, Fe chelates, high valent oxo-iron

Toluene

complexes, anionic interferences in H2O2 reaction, and H2O2 formation. The Fenton-like

Efficiency

initiation, formation, and propagation of a reaction intermediate species capable of

Radical propagation

transforming toluene, while simultaneously inhibiting H2O2 reaction is the most viable mechanism. Published by Elsevier Ltd.

1.

Introduction

Oxidative transformation of organic contaminants in water involving the Fenton-like reaction is initiated by the reaction of ferric iron (Fe(III)) and hydrogen peroxide (H2O2). Ferrous iron (Fe(II)), the product of this reaction, reacts with H2O2 forming the non-selective, highly reactive hydroxyl radical (
OH).
OH reacts with a wide range of organic contaminants and non-target reactants. In this seemingly simple system,

reaction intermediates derived from the parent compounds can play a significant role in the overall treatment process. In Fenton-driven oxidation treatment systems involving H2O2, an iron (Fe) catalyst and one or more target organic compounds, various parameters and small differences in treatment systems can significantly influence the rate and course of reactions. Specifically, the target organic compound and/or the transformation products play a variety of contrasting roles including: electron shuttles, high-valent

* Corresponding author. Tel.: þ1 580 436 8610; fax: þ1 580 436 8614. E-mail addresses: [email protected] (S.G. Huling), [email protected] (S. Hwang), [email protected] (D. Fine), [email protected] (S. Ko). 1 Tel.: þ1 787 832 40403454. 2 Tel.: þ1 580 436 8669. 3 Tel.: þ1 580 436 8742. 0043-1354/$ e see front matter Published by Elsevier Ltd. doi:10.1016/j.watres.2011.08.001

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 3 4 e5 3 4 2

iron-oxo or ironeperoxo complexes stabilized by ligand complexes, reduction of Feþ3 by neutral organic molecules or radicals, oxidation of Feþ2 by radicals or carbocations (Chen and Pignatello, 1997; Pignatello et al., 2006), consumption of
OH radicals, and reduction of Feþ3 to Feþ2 (Kang et al., 2002). These reactions may accelerate the reaction between Fe and H2O2, while other organics coordinate with Fe, decrease Fe availability, and lower H2O2 reactivity and contaminant transformations (Chen and Pignatello, 1997; Pignatello et al., 2006; Sun and Pignatello, 1992; Kwon et al., 1999). Byproducts from the oxidation of p-chlorophenol were found to deactivate iron by forming stable organo-ferro complexes (Kwon et al., 1999). Specifically, a colored precipitate was filtered and found to contain C, H, O, and Fe at 40, 2.7, 27, and 6.6 wt%. Fourier transform infrared spectroscopy analysis of the precipitate indicated the presence of H bonds and CeO stretching bonds which suggested that organics were complexed in the Fe solids. In the same study, oxalate, a decomposition product of p-chlorophenol, inhibited p-chlorophenol degradation when amended to a Fenton system. Formation of an Feþ3-oxalate complex that inhibited the reaction of H2O2, where the initial concentration of oxalate (30 mM) was 3 times greater than Feþ3 (10 mM) (Zuo and Hoigne, 1992). In another study, two separate mechanisms including (1) the complexation between Fe(III) and an organic acid, and (2) quinone reduction of Feþ3 to Feþ2, were incorporated into a kinetic model to account for Fe availability and reaction with H2O2 (Kang and Hua, 2005). The decomposition rate of p-chlorophenol was sensitive to both mechanisms. Fenton-driven degradation of phenol resulted in a distinct lag-phase in the transformation of phenol (Chen and Pignatello, 1997). Progression from the lag-phase to the reaction phase of phenol was attributed to the slow rate-limiting reduction of Feþ3 to Feþ2 by H2O2 through other more efficient reactions. Such reactions involved quinone intermediates that shuttled electrons to Feþ3, forming Feþ2. Similarly, the oxidation of Orange II, an organic dye compound exhibiting phenolic characteristics, was transformed into hydroquinone-like compounds. Consequently, in a Fenton system including mixed dyes, the oxidation of malachite green (i.e., an organic dye) was accelerated and was attributed to the formation of quinone electron shuttles. Enhanced toluene oxidation in soil resulted from the addition of various chelating agents (L-ascorbic acid, gallic acid, N-(2hydroxyethyl) iminodiacetic acid) (Kang and Hua, 2005). The increase in toluene transformation was attributed to retaining greater concentrations of dissolved Fe. The H2O2 concentration was not reported, so it is unclear whether the reaction was delayed or whether it occurred simultaneously with toluene. In this study, preliminary experiments involving toluene oxidation in Fenton-driven treatment systems revealed nonclassic concentration trends for both toluene and H2O2. Specifically, the simultaneous decline in toluene and H2O2 concentrations, typical in Fenton-like reaction systems, was not observed. Rather, a decline in toluene concentration was measured and H2O2 reaction was delayed until the toluene reaction was complete, or nearly complete. Results suggest a transformation mechanism other than the conventional Fecatalyzed,
OH-mediated transformation mechanism, or

5335

previously-described mechanisms involving organic intermediates. The objective of this study was to independently investigate potential mechanism(s) responsible for these results through variations in chemical and physical conditions of the treatment system.

2. Methods, materials, and analytical procedures 2.1.

Preparation of batch reactors

Batch reactors were prepared using 20-mL borosilicate glass vials with Teflon-coated septa and amended with various solutions. The solutions were added in the following order and the final concentrations of reactants are indicated in parenthesis: iron (Fe), amended as ferric sulfate (Fe2(SO4)3$9H2O) (5 or 30 mM as Fe), NaOH (1 M) (initial pH was 3), toluene (0.08e0.88 mM) or methyl tert-butyl ether (0.12e0.17 mM), and H2O2 (0.6e60 mM). In some cases, ferric sulfate was replaced with ferric nitrate (Fe(NO3)3$9H2O) in the treatment reactors. H2O2- and Fe-free reactors were used as controls. The vial was immediately capped after the H2O2 was amended to prevent volatilization of the toluene. Thermal desorption tubes containing Supelco Chromosorb 106, a styrene divinylbenzene porous polymer, were inserted through the Teflon septa which allowed the O2(g) to vent from the reactor. Volatilization of toluene, over the course of each experiment (120 min), was captured and quantified in the thermal desorption tubes (TDT’s). The total mass of toluene recovered in the TDT’s never exceeded 10% of the initial toluene mass. In the flask reactors, Fe-free and H2O2-free controls indicated that volatile losses were negligible (1%). In some experiments, isopropanol (IPA) was used as a hydroxyl radical (
OH) scavenger (k
OH ¼ 6  109 M1 s1) (Buxton et al., 1988). The addition of IPA to the reactors followed the addition of Fe. Complete-mix batch reactors (1 L glass flasks; stirred with magnetic stir bar; sealed to prevent volatilization) were used to allow a greater number of samples, to minimize variability, and to provide greater resolution in toluene and H2O2 concentration with time. Toluene, the target compound, reacts rapidly with 9 1 1
OH (k
s ) (Buxton et al., 1988). OH ¼ 3  10 M Eight reactors were used in each experiment and were sacrificed at 5, 10, 20, 30, 45, 60, 90, and 120 min after H2O2 was added. The thermal desorption tubes were removed from each reactor, capped, and subsequently analyzed for toluene using automated thermal desorption with a gas chromatograph (GC) and mass spectrometer (MS). Aqueous samples were removed for H2O2 and toluene analysis. The toluene samples were added to a pre-acidified 40-mL VOA vial (0.5 mL H2SO4 conc. acid) to quench the H2O2 reaction. The pH was measured in the remaining solution. The large flask reactors (1 L) were sealed and samples were removed through a top port to be analyzed for H2O2 and toluene. A dissolved oxygen (DO) probe was inserted and sealed to prevent volatile losses and to continuously measure DO. The probe compound, 5,5-dimethyl-1-pyrroline-N-oxide (DMPO) (1 mM), was amended to a treatment system containing toluene (1.65 mM), Feþ3 (5 mM), and H2O2 (6 mM) to assess the potential presence and role of benzyl radicals, a suspected

5336

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 3 4 e5 3 4 2

intermediate. Deuterated toluene (toluene-d8) was added in place of toluene in a separate reactor to confirm the presence of reaction intermediates detected with DMPO. Detailed methods and analytical procedures are presented in the Supporting Information (Section SI-2.1 e DMPO Methods and Analysis).

2.2.

Analytical

The toluene captured on the thermal desorption tubes was analyzed by automated thermal desorption and GC/MS. Analyte calibration was conducted by loading analyte standards onto thermal desorption tubes using a N2(g) gas carrier. The range of the standard curve was 100e17,500 ng (r2 > 0.99; 100 ng quantitation limit) for toluene. An Agilent 5890 GC, a 5972 mass selective detector (MSD), and a Perkin Elmer Turbomatrix Automated Thermal Desorber apparatus were used. Aqueous toluene samples were analyzed by the purge and trap GC method (HewlettePackard, Model 5890 series II). Aqueous samples containing toluene were purged with helium and volatiles were transferred to a K-VOCARB3000Encon trap. The trap was dry purged with helium to remove water vapor, and toluene was thermally desorbed onto the GC column for separation and measurement. Sample transfer was through a heated 1.9 mm  1.0 m Silcosteel (Restek) transfer line that is coupled directly to the analytical column. Following separation on the column, toluene was determined with photoionization and electrolytic conductivity detectors. The standard curve was 1.1  102e3.3 mM L1 (r2 > 0.99; 2.7  103 mM L1 detection limit). The standard error of toluene was 4e12% in duplicate samples. The quality of the data measurements was checked through analysis of quality control samples including method blanks, continuing

H2O2 (0.09 mM Toluene) H2O2 (0.88 mM Toluene) Toluene (0.09 mM)

calibration check standards, a second source quality control standard, lab duplicates, and matrix spikes. H2O2 was measured using a modified peroxytitanic acid colorimetric procedure (Boltz and Howell, 1978; Huling et al., 2001) with a detection limit of 2.9 mM. Aqueous samples collected from test reactors were filtered (0.2 mm) and analyzed in duplicate (standard error ranged between 2 and 3%). The TiSO4 reagent was from Pfaltz and Bauer Inc., and the H2O2 (30 %wt. solution in water, reagent grade) was from Aldrich. Total iron was measured using EPA Phenanthroline Method No. 3500-Fe D (Clesceri et al., 1989). The pH was measured using an Orion Sure-Flow ROSS Combination pH probe, and DO was measured using a Thermo Scientific Orion 3 Star meter.

3.

Results

3.1.

Toluene concentration

In conventional Fenton-driven treatment systems, loss in both H2O2 and the target compound occur simultaneously. In this study, toluene transformation occurred immediately upon H2O2 amendment, however, the H2O2 reaction was delayed until transformation of toluene was complete, or nearly complete (Fig. 1). The length of time for toluene to reach non-detectable concentrations was extended by increasing the initial toluene concentration (0.09, 0.18, 0.22, 0.88 mM), which also extended the delay in the reaction of H2O2. In toluene-free reactors, H2O2 reaction at 6 mM and 60 mM occurred immediately indicating that the presence of toluene played an integral role in the delay of the H2O2 reaction. Reaction byproducts detected in the toluene-amended, Fenton-driven treatment system included 2-, 3-, or 4-methyl

H2O2 (0.18 mM Toluene) H2O2 (6 mM) (Toluene-free) Toluene (0.18 mM)

H2O2 (0.22 mM Toluene) H2O2 (60 mM) (Toluene-free) Toluene (0.22 mM)

Toluene (0.88 mM)

H2O2, Toluene (C(t)/CINITIAL)

1.0

0.8

0.6

0.4

0.2

0.0 0

20

40

60

80

100

120

Reaction time (minutes) Fig. 1 e Influence of initial toluene concentration on Fenton-driven oxidation of toluene and H2O2 reaction. H2O2 reaction (without lag) in a toluene-free batch reactor oxidative system. Initial conditions: 0.09e0.88 mM toluene, 6 mM H2O2, 5 mM Fe3D (Fe2(SO4)3$9H2O), initial and final pH [ 3, 2.7.

5337

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 3 4 e5 3 4 2

phenol (cresols), benzyl alcohol, benzaldehyde, and benzoic acid and are consistent with previous toluene degradation results (Walling and Johnson, 1975). Simultaneous depletion of toluene and increases in cresol and benzyl alcohol were measured (Supporting Information, Section SI-1, Supplemental Figures, Figure SI-1).

3.2.

H2O2 and iron concentration

h ¼ ððD½toluene=D½H2 O2 Þ=0:67Þ  100

The rate of toluene transformation is dependent on the concentration of both H2O2 (0.6e60 mM) and Fe (0.5e30 mM) (Figs. 2 and 3). The sequential depletion of toluene and H2O2 in these treatment systems is consistent with previous results (Fig. 1). The concentration of total Fe was static over the course of the 120 min experiments.

3.3.

Counter-anion of iron

A similar trend of sequential depletion of toluene and H2O2 was also observed by replacing ferric sulfate with ferric nitrate. Toluene and H2O2 transformation were more rapid in the ferric nitrate-amended system relative to the ferric sulfate-amended system (Supporting Information, Section SI-1, Supplemental Figures, Figure SI-2). In a similar study, sulfate interfered in the formation of the ferro-peroxy intermediate species, inhibited H2O2 reaction,
OH formation, and MTBE oxidation (Hwang et al., 2010). Results indicate that sulfate plays a role in limiting the rate of H2O2 reaction but is not responsible for the delay in H2O2 reaction prior to toluene depletion in this study.

3.4.

(3) other reactions which interfere with the reaction of Fe (Feþ2, Feþ3) and
O 2 , 2 mol toluene would be transformed per 3 mol H2O2 per the balancing of reactions R1-R4 (Table 1) (Huling et al., 1998). Therefore, toluene oxidation under theoretical conditions is 0.67 mol toluene/mol H2O2. The reaction efficiency (h) is defined as the ratio of actual and theoretical transformation (Eq (1)):

Competition kinetics analysis

Under theoretical Fenton conditions in the absence of (1) nonproductive H2O2 reactions (i.e., reactions in which H2O2 is consumed but
OH are not produced), (2)
OH scavenging, and

(1)

The main sources of treatment inefficiency under actual conditions include non-productive H2O2 reactions (R5-R6, Table 1) and
OH scavenging (R6-R13, Table 1), which may play a significant role in environmental Fenton-driven treatment systems. The actual oxidative treatment efficiency in environmental systems is much less than theoretical (i.e., <<0.67 mol toluene/mol H2O2). Assuming conventional Fenton-driven reactions, the relative rate of reaction of toluene (RTOLUENE) (Eq (2)) in this treatment system can be used to assess reaction kinetics between
OH and toluene, versus
OH and H O . 2 2 RTOLUENE ¼ ðk4 ½toluene½
OH=ðk4 ½toluene þ k6 ½H2 O2 Þ½
OHÞ

Based on the initial conditions of these experiments ([H2O2] ¼ 6 mM; [toluene] ¼ 0.09e0.88 mM),
OH reaction is predominantly with toluene (i.e., RTOLUENE ¼ 0.63e0.94), at least initially, and therefore, 6e37% of the
OH are projected to react with H2O2 (i.e., RH2 O2  100). As the concentration of toluene declines and H2O2 concentration remains nearly constant during the lag-phase, the rate of reaction between
OH and H2O2 would increase, resulting in a decline in RTOLUENE and an increase in RH2 O2 . Additionally, reactions between
OH and toluene decomposition products and other scavengers would further reduce toluene transformation efficiency. Based on the competition kinetics analysis and the reaction efficiency calculations, results indicate that theoretical

1.0

H2O2 or Toluene (Ct/Co)

0.8

H2O2 (0.6 mM) H2O2 (6 mM) H2O2 (60 mM)

0.6

Toluene (0.6 mM H2O2) Toluene (6 mM H2O2) Toluene (60 mM H2O2)

0.4

0.2

0.0 0

20

40

(2)

60

80

100

120

Reaction time (minutes)

Fig. 2 e Influence of H2O2 concentration on Fenton-driven oxidation of toluene and H2O2 reaction. Initial conditions: 0.6e60 mM H2O2, 0.17 mM toluene, 5 mM Fe3D (Fe2(SO4)3$9H2O), initial and final pH [ 3, 2.7.

5338

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 3 4 e5 3 4 2

H2O2 or Toluene (Ct/Co)

1.0

0.8 H2O2 (0.5 mM Fe) H2O2 (5 mM Fe) H2O2 (30 mM Fe)

0.6

Toluene (0.5 mM Fe) Toluene (5 mM Fe)

0.4

Toluene (30 mM Fe)

0.2

0.0 0

20

40

60

80

100

120

Reaction time (minutes) Fig. 3 e Influence of iron (Fe) concentration on Fenton-driven oxidation of toluene and H2O2 reaction. Initial conditions: 0.5e30 mM Fe3D (Fe2(SO4)3$9H2O), 0.17 mM toluene, 6 mM H2O2, initial and final pH [ 3, 2.7.

treatment efficiency (i.e., 0.67 mol toluene/mol H2O2) would not exist due to the presence of known scavengers. Consequently, under the conditions of these experiments, the oxidation efficiency would be much less (<<100%) than theoretical, yet the actual toluene reaction efficiency was greater than (>100%) theoretical oxidation efficiency in all cases (Table 2). The variability in estimates of overall oxidation efficiency is attributed to measuring small differences in high H2O2 concentration in the sacrificed reactors. For this reason, additional experiments involving large (1 L), completemix, sealed flask reactors were conducted to increase sampling and analysis, to minimize variability, and to provide greater resolution in toluene and H2O2 concentrations. In these complete-mix reactors, an immediate decline in H2O2 was measured in the toluene-free control, the variability in H2O2 concentration was limited (Supporting Information, Section SI-1, Supplemental Figures, Figure SI-3), and estimates of the overall oxidation efficiency were less variable and remained greater than 100% in all cases (Table 2) relative to the vial reactors. The toluene transformation mechanism was initiated by Fenton-like reactions, and despite some variability in estimates of the oxidation efficiency, these data clearly indicate that conventional Fenton-driven oxidative transformation reactions cannot be the predominant mechanism responsible for such efficient toluene transformation.

3.5.

Reaction intermediates

The potential involvement of the benzyl radical intermediate was investigated by amending a spin-trap compound, DMPO, to the treatment system. The DMPO-benzyl radical adduct was detected by mass spectrometry in the treatment system during the H2O2 lag-phase. The presence of the benzyl radical

was confirmed through the use of deuterated toluene and mass spectral detection of a molecule with the unique fragmentation expected for DMPO-benzyl-d7. Detailed data and information regarding these experiments, analysis, and results are included in the Supporting Information (Section SI2.1. DMPO Methods and Analysis; Section SI-2.2, Results; Section SI-2.3 e Benzyl Radical). Analysis via LC/MS/MS of the treatment system solution (i.e., toluene or toluene-d8, Feþ3, H2O2) without DMPO revealed the presence of a stable oxidized product of toluene containing three oxygens, having a molecular formula of C8H8O3. The mass spectra of this compound and its deuterated analog suggests that the molecule is a cyclic hexane ring containing one double bond, an alcohol, a ketone and an epoxy group (Section SI-2.4 Epoxide Intermediate).

Table 1 e Fenton and related reactions. R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17

H2O2 þ Feþ2 / Feþ3 þ OH þ
OH H2O2 þ Feþ3 / Feþ2 þ
HO2 þ Hþ
 O2 þ Feþ3 / Feþ2 þ O2 toluene þ
OH / toluene’ 2þ
H2O2 þ
SO 4 / SO4 þ H þ HO2 H2O2 þ
OH /
HO2 þ H2O Fe2þ þ
OH / Fe3þ þ OH FeOHþ þ
OH / Fe3þ þ 2 OH
HO2 þ
OH / O2 þ H2O
 O2 þ
OH / O2 þ OH
OH þ
OH / H2O2 

HSO 4 þ OH / SO4 þ H2O 3þ 
FeSO4 þ OH / Fe SO24 þ OH Fe3þ þ H2O2 / FeIII(HO2)2þ þ Hþ FeOH2þ þ H2O2 / FeIII(OH)(HO2)þ þ Hþ FeIII(HO2)2þ / Feþ2 þ
HO2 FeIII(OH)(HO2)þ / Fe2þ þ
HO2 þ OH

k1 ¼ 76 M1 s1 k3 ¼ 3.1  105 M1 s1 k4 ¼ 3.0  109 M1 s1 k5 ¼ 1.2  107 M1 s1 k6 ¼ 2.7  107 M1 s1 k7 ¼ 2.7  108 M1 s1 k8 ¼ 2.7  108 M1 s1 k9 ¼ 0.71  1010 M1 s1 k10 ¼ 1.01  1010 M1 s1 k11 ¼ 5.2  109 M1 s1 k12 ¼ 3.5  105 M1 s1 k13 ¼ 2.7  108 M1 s1

k16 ¼ 2.3  103 s1 k17 ¼ 2.3  103 s1

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 3 4 e5 3 4 2

Table 2 e Toluene transformation in Fenton-like treatment systems. Initial conditions: 6 mM H2O2, 5 mM Fe3D (ferric sulfate, unless otherwise noted), pH 3. Vial reactors (10 mL total solution volume) and flask reactors (1 L total solution volume) were used. Initial [toluene] (mM)

H2O2 Toluene/ Overallc Toluene a H2O2b Efficiency transformed reacted (mM) (mM) (mM/mM) (%)

Ferric sulfate-amended system Vial Reactors 0.09 0.077 0.18 0.125 0.22 0.193 0.88 0.711 0.16 0.137 Flask Reactors 0.078 0.076 0.177 0.137 0.192 0.186 0.177 0.218d 0.323 0.28 0.438 0.194 Ferric nitrate-amended systeme Vial Reactors 0.15 0.116 0.19 0.143

0.021 0.015 0.21 0.189 0.58

3.68 8.32 0.919 3.76 0.236

550 1240 140 560 40

0.09 0.12 0.07 0.13 0.36 0.18

0.85 1.14 2.66 1.36 0.78 1.08

130 170 400 200 120 160

0.15 0.127

0.8 1.13

120 170

a Toluene volatilization occurred during the H2O2 reaction period and was captured in thermal desorption tubes and analyzed via GC. The quantity of toluene transformed (column 2) was corrected for both volatile losses and for the concentration of toluene remaining in the aqueous phase. b Toluene transformed/H2O2 reacted (mM/mM). c Overall efficiency ¼ ((toluene/H2O2)/0.67)  100. d [H2O2]INITIAL ¼ 0.63 mM. e 0.15 mM toluene with 0.5 mM Fe3þ; 0.19 mM toluene with 5.0 mM Fe3þ.

The possible role of the aromatic ring structure in the formation of reactive intermediates responsible for the gratuitous transformation (co-oxidation) of other reactants present in a binary mixture of contaminants was investigated (Section SI-2.5 Binary Treatment System). Both the tolueneand benzene-amended Fenton-driven binary treatment systems resulted in a lag-phase in H2O2 reaction and rapid and efficient toluene and benzene transformation. No lag-phase in H2O2 reaction was measured with 2CP indicating that 2CP did not involve a similar mechanism. Relative to single solute treatment systems, greater transformation of the cocontaminants, dichloropropane, an agricultural pesticide, and methyl tert-butyl ether (MTBE), a fuel oxygenate, were measured in all binary systems amended with toluene, benzene, or 2CP (Section SI-2.5 Binary Treatment System).

4.

Discussion

Results support several significant observations including: (1) the delay in H2O2 reaction (lag-phase) until the complete or near-complete transformation of toluene and benzene, (2) highly efficient (>100%) toluene and benzene transformation without significant H2O2 reaction, and (3) no delay in H2O2

5339

reaction involving treatment systems amended with 2CP. Further, the toluene transformation mechanism is initiated by Fenton-like reactions and therefore is dependent on conventional Fenton-like parameters (Fe, H2O2, scavengers, anionic interference in H2O2 reaction, etc.).

4.1.

H2O2 reaction

The delay in H2O2 reaction until complete transformation of toluene provides direct evidence of an alternative, nonclassical transformation pathway. Time-dependent concentrations of H2O2 are often not reported even in elaborate and insightful Fenton studies investigating the role of reaction intermediates. Consequently, it is unclear whether the formation of reaction intermediates (i.e., ferro-organo complexes, electron shuttles, organic radicals, etc.) enhance, inhibit, or delay H2O2 reaction (i.e., a lag-phase) in some Fenton-driven treatment systems. In one study where time-dependent H2O2 concentrations were reported, Fenton-driven oxidation of 4-nitrophenol (4NP) involved an extended delay in both 4NP and H2O2 reaction (De Laat et al., 2004). Subsequently, commencement of both 4NP and H2O2 reaction occurred simultaneously. The delay in 4NP and H2O2 reaction was attributed to the combined effects of the rate-limiting decomposition of the ferro-peroxo complex (De Laat et al., 2004; De Laat and Le, 2005), and the role of organic intermediates (i.e., quinone electron shuttles) that reduce Fe(III) more rapidly than H2O2. The oxidation efficiency was estimated to be 50e60% (i.e., ((Δ4NP/ΔH2O2)/0.67)  100) (De Laat et al., 2004). These results are more consistent with a conventional Fenton-driven treatment mechanism in a clean system, unlike the high overall treatment efficiency of toluene (>100%, Table 2) measured during the H2O2 lag-phase in this study. In another study, it was proposed that H2O2 formation had resulted in a system involving Feþ2, H2O2, and benzene, where dissolved oxygen was consumed faster than H2O2 depletion and where benzene oxidation efficiency was slightly greater than 100% (Kunai et al., 1986). It was reported that initially,
OH was added to the benzene ring, formed the hydroxycyclohexadienyl radical, and reacted with O2 to form the peroxy radical. It was assumed that the peroxy radical was transformed to phenol, liberating
HO2 which was reduced by Feþ2 to form H2O2. In contrast with our treatment system, Fe was added as Feþ3, and very low concentrations of Feþ2 are expected in Fenton-like systems. Furthermore, significant H2O2 reaction occurred immediately after toluene depletion (Fig. 1). Specifically, 0.09 mM of toluene was depleted in the first 30 min of reaction, and over the course of the next 30 min, greater than 3 mM H2O2 was consumed. Assuming stoichiometric quantities of H2O2 were formed (i.e., per toluene destroyed), the quantity of H2O2 that should have been produced during the lag phase was greater than 30 times the toluene destroyed (i.e., 0.09 mM toluene << 3 mM H2O2). Although results from these two studies exhibit general similarities, including the likely role of the hydroxycyclohexadienyl radical, our results appear inconsistent with the H2O2 formation mechanism proposed by Kunai et al. (1986). Several lines of evidence have been provided that document the lag phase in H2O2 reaction. This is predominantly

5340

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 3 4 e5 3 4 2

indicated by the direct measurement of H2O2 in both the small reactors (20 mL) (Figs. 1e3) and in large reactors (1 L) (Figure SI3) where numerous sampling events and sample duplication (triplicates) was performed. The existence of the H2O2 lag phase is most evident when contrasting the immediate decline in H2O2 concentration measured in the absence of toluene with the delay in H2O2 reaction observed in tolueneamended reactors (Fig. 1). Subsequently, rapid reaction of H2O2 is measured upon the complete or near-complete depletion of toluene. During the lag phase, loss of dissolved oxygen (DO), an alternative electron acceptor to
OH, was measured and is discussed further in the Reaction intermediates section, below. Assuming H2O2 reaction was occurring but was masked by the simultaneous formation of H2O2, the DO level would increase, not decline, as observed. Once H2O2 begins to react, it is much faster than can be explained by H2O2 synthesis mechanisms. In general, it is improbable that the rate of H2O2 formation and rate of reaction are nearly identical and would cause the H2O2 concentration to be steady in time. A permanganate titration method was used to measure H2O2 concentration in the test reactors as a supplementary method to the peroxytitanic acid colorimetric method. This was performed for assurance to confirm (1) the H2O2 concentration, and (2) the absence of intermediate oxidizing species that could interfere with H2O2 measurements and could mask the detection of a H2O2 formation mechanism. This test involved similar conditions using large complete-mix flask reactors (5 mM Fe(III), 6 mM H2O2, 0.1 mM Toluene). Results indicated similar H2O2 concentrations using the two methods (data not included) which confirms the validity of the peroxytitanic acid colorimetric method, and the absence of matrix interferences and a potential H2O2 formation mechanism.

4.2. Electron shuttles, iron chelates, high valent oxo-iron complexes Several mechanisms were evaluated to assess their potential role in these experiments. Quinone or hydroquinone compounds, resulting from Fenton-like oxidation of phenol, may function as electron shuttles in Fenton-driven oxidation systems, resulting in an additional pathway for Feþ2 formation,
OH production, and contaminant oxidation (Chen and Pignatello, 1997). However, quinone- or hydroquinone-like compounds were not detected via GC/MS in the test reactors. Coordination stoichiometry for a ligand-Fe complex was 1 mol Fe and 2 mol of p-hydroxybenzoic acid (Rivas et al., 2001). Many organic ligands (n ¼ 50) evaluated in another study required approximately 1:1 (mol:mol) coordination with Fe(III) and resulted in a reduction in Fe reactivity with H2O2 (Sun and Pignatello, 1992). The toluene concentration (0.078e0.88 mM) used in our treatment system was significantly less than the Fe concentration (5 mM). Assuming a toluene-derived ligand was formed in this treatment system, a high number of Fe ions would require complexation (5e55 mol Fe:mol ligand) to completely inhibit the H2O2 reaction and is unlikely. In addition to
OH, potential candidates for reactive species include the high valence ferryl species. Formation of high valent oxo-iron complexes in Fenton systems have been used to describe the transformation of organic compounds via hydroxylation,

ketonization, epoxidation, cleavage, dehydrogenation, and other reactions. However, these reactions are not well established and proposed non-hydroxyl radical schemes for the Fenton reaction have been controversial (Pignatello et al., 2006). For example, a non-radical process involving the oxidation of Feþ3 by H2O2 results in the formation of Fe(V) and, subsequently, the highly reactive Fe(V) ¼ O species (Schuchardt et al., 2001). This species can react with organics to form an organometallic Fe(V) complex which then undergoes a series of possible reactions. Assuming quinone or hydroquinone compounds, Fe chelates, or high valent oxoiron complexes, were present and played a significant role, the toluene oxidation efficiency under the best case theoretical conditions would still be significantly less than 100%. The actual toluene transformation efficiency measured in the treatment system was much greater (Table 2). Therefore, the role of these compounds as potential electron shuttles, Fe chelators, or alternative oxidants, was ruled out as the cause of the highly efficient transformation mechanism.

4.3.

Reaction intermediates

Detection and confirmation of the benzyl radical in the treatment system indicates that it may have played a role in toluene transformation mechanism (Supporting Information, Section SI-2.1e2.3, DMPO Methods, Analysis, Results, and Benzyl Radical). In the absence of the DMPO spin-trap compound, a cyclic hexane ring containing one double bond, an alcohol, a ketone, and an epoxy group was detected via LC/ MS/MS analysis (Supporting Information, Section SI-2.4, Epoxide Intermediate). A similar organic chemical epoxide structure was predicted based on theoretical quantum mechanical calculation by others involving
OH addition and oxidation of aromatics (Bartolotti and Edney, 1995). Depending upon reaction conditions including oxygen, redox, Fe, and probably other reactants, the aromatic transformation can proceed in multiple pathways. It was further predicted to undergo additional transformations by reaction with dissolved oxygen (DO) followed by carbonecarbon scission reactions (Bartolotti and Edney, 1995). Isomers of the hydroxycyclohexadienyl radical are widely accepted as the initial product of
OH addition to aromatic compounds, and are known to reduce O2 and Fe(III) in producing hydroxylated products (Deusterberg et al., 2005). In acidic solutions, the species readily undergoes water elimination to form the phenoxyl radical, a strong oxidizing agent. In these treatment systems, DO serves as an alternate electron acceptor to H2O2 and may play a significant role in Fenton-initiated treatment systems, where the Fenton process can be described as a soup-like radical solution (Utset et al., 2000). In such a system, when H2O2 is replaced through the external supply of O2, as the main oxidizing species, it could attack organic molecules, overcome the need for stoichiometric consumption of H2O2, and eventually improve the economical requirements of the system (Utset et al., 2000). This mechanism was investigated using similar reagents and concentrations (5 mM Feþ3; 6 mM H2O2; 0.17 mM toluene; pH 3), a larger volume (1 L), and simultaneous measurement of DO, H2O2, and toluene (Supporting Information Section SI-2.6 e Reaction of Dissolved Oxygen During the H2O2 Lag-phase).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 3 4 e5 3 4 2

Prior to H2O2 reaction, the DO declined from 7.5 to 5 mg/L until the toluene was fully consumed. Upon complete reaction of toluene, H2O2 began to react and the DO rose sharply reaching saturation conditions. The role of DO as a replacement of H2O2 (Utset et al., 2000) is consistent with the intermediate radical mechanism (Deusterberg et al., 2005) where an oxygen molecule is inserted in the hydroxymethylcyclohexadienyl radical of toluene resulting in an epoxide intermediate.

4.4.

Summary

The toluene transformation mechanism that occurs during the H2O2 lag phase is initiated by Fenton-like reactions and is dependent on conventional Fenton-like parameters such as (1) Fe and H2O2 concentration that enhance toluene oxidation, and (2)
OH scavengers and inhibitors of the ferro-peroxy intermediate (i.e., SO24 ) that limit toluene oxidation. This sequence of reactions explains why these conventional parameters either enhanced or interfered with the Fentondriven initiation reaction, but did not prevent the mechanism from occurring that was ultimately responsible for efficient toluene transformation. Results also suggest that the relative magnitude of the initial Fenton-like reaction, and subsequent burst of
OH, directly impacts the overall mechanism responsible for toluene transformation. For example, toluene transformation is faster, and the lag-phase in H2O2 reaction is shorter when initial conditions are favorable for Fenton-like reactions. Also, at low initial H2O2 concentration (i.e., 0.6 mM, Fig. 2), the reaction of H2O2 was not completely inhibited suggesting that a minimum concentration of H2O2 is required to initiate and perpetuate the mechanism, and/or that both mechanisms occur simultaneously. Through the various experiments conducted in this study, it was determined that several parameters and mechanisms do not play a significant role including stoichiometric formation of H2O2, electron shuttles, Fe chelates, high valent oxo-iron complexes, and interferences in H2O2 reaction by anions such as SO24 . The Fenton-like initiation, formation, and propagation of a reaction intermediate species capable of transforming toluene, while simultaneously inhibiting H2O2 reaction is the most viable mechanism explaining these results. The mechanism investigated in this study may be overlooked in Fenton-driven oxidation studies. Assuming H2O2 concentration is not simultaneously measured with the loss of the target compound, the role of this mechanism may not be recognized and critical analyses of reaction pathways or mechanisms may potentially be misinterpreted. Additional investigation is needed to characterize the mechanism and ultimately optimize this unique mechanism for eventual use in water treatment processes to improve treatment efficiency and reduce treatment cost.

Notice The U.S. Environmental Protection Agency, through its Office of Research and Development, funded and managed the research described here. It has not been subjected to Agency review and therefore does not necessarily reflect the views of the Agency, and no official endorsement should be inferred.

5341

Acknowledgments The authors acknowledge J. Andrews (East Central University, Ada, OK), S. Caldwell, S. Beach, M. Blankenship, D. Kovacs, and Dr. N. Xu (Shaw Environmental Inc., Ada, OK) for their assistance. Additionally, the authors appreciate technical input from Drs. Robert G. Arnold and Wendell P. Ela (University of Arizona, Department of Chemical and Environmental Engineering, Tucson, AZ).

Appendix. Supporting information The supplementary data associated with this article can be found in the on-line version at doi:10.1016/j.watres.2011.08. 001.

references

Bartolotti, L.J., Edney, E.O., 1995. Density functional theory derived from the OH initiated atmospheric oxidation of toluene. Chemical Physics Letters 245 (1), 119e122. Boltz, D.F., Howell, J.A., 1978. In: Colorimetric Determination of Non-metals. Wiley-Interscience Publications, John Wiley and Sons. Buxton, G.V., Greenstock, C., Hellman, W.P., Ross, A.B., 1988. Critical review of rate constants for reactions of hydrated electrons, hydrogen atoms and hydroxyl radicals (
OH/
O) in aqueous solution. Journal of Physical and Chemical Reference Data 17 (2), 513e886. Chen, R., Pignatello, J.J., 1997. Role of quinone intermediates as electron shuttles in Fenton and photoassisted Fenton oxidations of aromatic Compounds. Environmental Science & Technology 31 (8), 2399e2406. Clesceri, L.S., Greenberg, A.E., Trussell, R.R., 1989. Standard Methods for the Examination of Water and Wastewater, seventeenth ed. American Public Health Association, Washington D.C. De Laat, J., Le, G., Legube, B., 2004. A comparative study of the effects of chloride, sulfate, and nitrate ions on the rates of decomposition of H2O2 and organic compounds by Fe(II)/H2O2 and Fe(III)/H2O2. Chemosphere 55 (5), 715e723. De Laat, J., Le, T.G., 2005. Kinetics and modeling of the Fe(III)/H2O2 system in the presence of sulfate in acidic aqueous solutions. Environmental Science & Technology 39 (6), 1811e1818. Deusterberg, C.K., Cooper, W.J., Waite, T.D., 2005. Fentonmediated oxidation in the presence and absence of oxygen. Environmental Science & Technology 39 (13), 5052e5058. Huling, S.G., Arnold, R.G., Sierka, R.A., Miller, M.A., 2001. Influence of peat on Fenton oxidation. Water Research 35 (7), 1687e1694. Huling, S.G., Arnold, R.G., Sierka, R.A., Miller, M.A., 1998. Measurement of hydroxyl radical activity in a soil slurry using the spin trap a-(4-pyridyl-1-oxide)-N-tert-butylnitrone. Environmental Science & Technology 32 (21), 3436e3441. Hwang, S., Huling, S.G., Ko, S., 2010. Fenton-like degradation of MTBE: effects of iron counter anion and radical scavengers. Chemosphere 78 (5), 563e568. Kang, N., Lee, D.S., Yoon, J., 2002. Kinetic modeling of Fenton oxidation of phenol and monochlorophenols. Chemosphere 47 (9), 915e925.

5342

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 3 4 e5 3 4 2

Kang, N., Hua, I., 2005. Enhanced chemical oxidation of aromatic hydrocarbons in soil systems. Chemosphere 61 (7), 909e922. Kunai, A., Hata, S., Ito, S., Sasaki, K., 1986. The role of oxygen in the hydroxylation reaction of benzene with Fenton’s reagent. 18 O tracer study. Journal of the American Chemical Society 108 (19), 6012e6016. Kwon, B.G., Lee, D.S., Kang, N., Yoon, J., 1999. Characteristics of p-chlorophenol oxidation by Fenton’s reagent. Water Research 33 (9), 2110e2118. Pignatello, J.J., Oliveros, E., MacKay, A., 2006. Advanced oxidation processes for organic contaminant destruction based on the Fenton reaction and related chemistry. Critical Reviews in Environmental Science and Technology 36 (1), 1e84. Rivas, F.J., Beltran, F.J., Frades, J., Buxeda, P., 2001. Oxidation of P-hydroxybenzoic acid by Fenton’s reagent. Water Research 35 (2), 387e396.

Schuchardt, U., Jannini, M.J.D.M., Richens, D.T., Guerreiro, M.C., Spinace, E.V., 2001. Gif chemistry: new evidence for a nonradical process. Tetrahedron 57 (14), 2685e2688. Sun, Y., Pignatello, J.J., 1992. Chemical treatment of pesticide wastes. Evaluation of Fe(III) chelates for catalytic hydrogen peroxide oxidation of 2,4-D at circumneutral pH. Journal of Agricultural and Food Chemistry 40 (2), 322e327. Utset, B., Garcia, J., Casado, J., Domenech, X., Peral, J., 2000. Replacement of H2O2 by O2 in Fenton and photo-Fenton reactions. Chemosphere 41 (8), 1187e1192. Walling, C., Johnson, R.A., 1975. Fenton’s reagent. V. hydroxylation and side-chain cleavage of aromatics. Journal of the American Chemical Society 97 (2), 363e367. Zuo, Y., Hoigne, J., 1992. Formation of hydrogen peroxide and depletion of oxalic acid in atmospheric water by photoloysis of iron(III)-oxalato complexes. Environmental Science & Technology 26 (5), 1014e1022.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 4 3 e5 3 4 5

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Comment

Discussion of Arnaldos, M., Pagilla, K., 2010. Effluent dissolved organic nitrogen and dissolved phosphorus removal by enhanced coagulation and microfiltration. Water Research 44, 5306e5315 John Bratby*, Denny S. Parker Brown and Caldwell, 1697 Cole Blvd., Suite 200, Golden, CO 80401, USA

article info

.

Article history: Received 12 April 2011 Received in revised form 4 July 2011 Accepted 12 July 2011 Available online 27 July 2011

The paper is an important contribution and provides valuable information on the methods developed at the Illinois Institute of Technology for measuring low DON and organic phosphorus residuals. However, there are two important issues that the writers consider important to discuss below: The first relates to the importance of final pH after coagulant addition. The second relates to the extraordinarily favorable results presented in the paper, compared to other workers. Fig. 1 in the paper by Arnaldos and Pagilla shows an increase in residual dissolved organic nitrogen (DON) at higher aluminum dosages. This was also demonstrated by Bratby et al. (2008) and reproduced in Fig. 1 below. The importance of pH was clearly shown and it results from the competition between metal hydrolysis products and hydrogen ions for organic ligands and also between hydroxyl ions and organic anions for metal hydrolysis products. A similar effect can be shown with chemical phosphorus removal. Fig. 4.28 in Bratby (2006) demonstrates this.

Notwithstanding the importance of pH in understanding the removal mechanisms, the paper does not provide the effective pH values for coagulation reactions. The initial pH values are provided in Table 3 in the paper by Arnaldos and Pagilla, but the parameter of interest which is the final pH after the addition of aluminum sulfate was not provided in the results of the experiments, although the general range of 6.8e7.4 was indicated. The individual final pH values corresponding with the points on Fig. 1 in the paper by Arnaldos and Pagilla would have been of interest. In this regard, since pH was such an important parameter in the results, fitting a polynomial expression to the DON results would appear to be of limited utility. pH data would also have been useful to possibly better understand the relatively high residual P values shown in Fig. 3 in the paper by Arnaldos and Pagilla. The results in the paper show a dosage of 1.5 M Al per M initial DON-N. This dosage achieved 69-percent DON removal

DOI of original article: 10.1016/j.watres.2010.06.066. * Corresponding author. Tel.: þ1 303 239 5452. E-mail address: [email protected] (J. Bratby). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.07.016

5344

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 4 3 e5 3 4 5

Fig. 1 e Effect of pH correction on DON removal. Adjusting pH to an optimal value maximizes DON removal.

for a final DON concentration of 0.3 mg/L. This translates to an initial DON concentration of approximately 1 mg/L and a DON removal of approximately 0.67 mg/l. The reported dosage can be expressed as 2.25 M Al per M DON-N removed.

The results presented by Bratby et al. (2008) are shown in Fig. 2. The results reported in the paper by Arnaldos and Pagilla appear to be extraordinarily favorable when compared to the results of other workers, including Randtke

Fig. 2 e Compilation of coagulant dosages to achieve treated DON levels. Secondary effluent results are reasonably consistent within a broad band.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 4 3 e5 3 4 5

et al. (1978) and Bratby et al. (2008), and even for the surface waters presented by Westerhoff et al. (2006). Fig. 2 indicates that for an effluent DON of 0.3 mg/L, molar dosages on the order of several hundred M Al per M N removed would be required. Another more recent point of reference is Dwyer et al. (2009) who showed that for two wastewater plant effluents, and a synthetic effluent, for effluent DON concentrations of 1.90, 1.52 and 0.61 mg/L respectively, the corresponding values for M Al per M N removed were 13.2, 14.0 and 50.2, respectively. These results are within the general band of results presented by Bratby et al. (2008) and reproduced in Fig. 2. The significant differences between the results reported in the paper by Arnaldos and Pagilla, and earlier literature results appear to be a significant issue and would have been very important to discuss in their paper.

5345

references

Bratby, J., Jimenez, J., Parker, D., 2008. Dissolved organic nitrogen e is it significant and can it be removed? In: Proceedings of the 81st Annual Water Environment Federation Technical Exhibition and Conference. WEFTEC 08, Chicago, Illinois October. Bratby, J., 2006. Coagulation and Flocculation in Water and Wastewater Treatment, Second Edition. IWA Publishing, London. Dwyer, J., Griffiths, P., Lant, P., 2009. Simultaneous colour and DON removal from sewage treatment plant effluent: alum coagulation of melanoidin. Water Research 43, 553e561. Randtke, S., Parkin, G., Keller, J., Leckie, J., McCarty, P., 1978. Soluble Organic Nitrogen Characteristics and Removal. In: USEPA Environment Protection Technology Series EPA-600/278-030, Cincinnati, Ohio, USA. Westerhoff, P., Lee, W., Croue, J.-P., Gallard, H., Amy, G., 2006. Organic Nitrogen in Drinking Water and Reclaimed Wastewater. AWWA Research Foundation, Denver, Colorado, USA.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 4 5 e5 7 5 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Optimization of capacity and kinetics for a novel bio-based arsenic sorbent, TiO2-impregnated chitosan bead Sarah M. Miller a, Matthew L. Spaulding a, Julie B. Zimmerman a,b,* a b

Department of Chemical and Environmental Engineering, Yale University, United States School of Forestry and Environmental Studies, Yale University, United States

article info

abstract

Article history:

The optimization of TiO2-impregnated chitosan beads (TICB) as an arsenic adsorbent is

Received 2 May 2011

investigated to maximize the capacity and kinetics of arsenic removal. It has been previ-

Received in revised form

ously reported that TICB can 1) remove arsenite, 2) remove arsenate, and 3) oxidize arsenite

24 August 2011

to arsenate in the presence of UV light and oxygen. Herein, it is reported that adsorption

Accepted 25 August 2011

capacity for TICB is controlled by solution pH and TiO2 loading within the bead and

Available online 1 September 2011

enhanced with exposure to UV light. Solution pH is found to be a critical parameter, whereby arsenate is effectively removed below pH 7.25 and arsenite is effectively removed

Keywords:

below pH 9.2. A model to predict TICB capacity, based on TiO2 loading and solution pH, is

Arsenic

presented for arsenite, arsenate, and total arsenic in the presence of UV light. The rate of

Water

removal is increased with reductions in bead size and with exposure to UV light. Phosphate

Chitosan

is found to be a direct competitor with arsenate for adsorption sites on TICB, but other

TiO2

relevant common background groundwater ions do not compete with arsenate for

Bio-based

adsorption sites. TICB can be regenerated with weak NaOH and maintain full adsorption

Sustainable

capacity for at least three adsorption/desorption cycles. ª 2011 Published by Elsevier Ltd.

1.

Introduction

Over 120 million people in Bangladesh and India rely upon groundwater with elevated arsenic levels as their primary source of drinking water (Ratnaike, 2003). This “mass poisoning” has resulted in many negative health outcomes, including toxicity to the liver, skin, kidney, and cardiovascular system, as well as multiple cancers (Agros et al., 2010). As community scale infrastructure for water treatment is not likely in the near future, point-of-use technologies, most relying on adsorption, have been advocated for arsenic removal (Petrusevski et al., 2008). Despite promising lab results for many of these technologies, no technology has adequately demonstrated long-term effectiveness and sustainability in the field (Petrusevski et al., 2008).

Arsenic in groundwater primarily exists as arsenite (As(III)) and arsenate (As(V)) (Bhattacharya et al., 2007). Arsenite is up to 60 times more toxic than arsenate (Ratnaike, 2003; Tien et al., 2004). Because arsenite is uncharged at environmentally relevant pH, it is also more difficult to remove than arsenate, which is negatively charged at environmentally relevant pH. Metal oxide sorption technologies, usually based on Ti oxides, Fe oxides, or Al oxides, have the unique ability to remove both As(III) and As(V). This is a major advantage relative to technologies like ion exchange, which can only remove As(V) and require a pre-oxidation step to achieve total arsenic removal. The synthesis of TiO2-impregnated chitosan bead (TICB), a bio-based adsorbent with promising arsenic removal capacity, has been previously reported (Miller and

* Corresponding author. Department of Chemical and Environmental Engineering, Yale University, United States. Tel.: þ1 203 432 9703. E-mail address: [email protected] (J.B. Zimmerman). 0043-1354/$ e see front matter ª 2011 Published by Elsevier Ltd. doi:10.1016/j.watres.2011.08.040

5746

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 4 5 e5 7 5 4

Zimmerman, 2010). Previously reported results were for TICB comprised of 70% chitosan and 30% TiO2 by weight (Miller and Zimmerman, 2010). Chitosan, a derivative of chitin, has a positive environmental and economic profile because it is renewable, biodegradable (Muzzarelli and Muzzarelli, 2005), and can be isolated from the millions of tons of shellfish waste generated globally per year (Gerente et al., 2007). TiO2 is a nontoxic nanopowder (Deebar et al., 2009) that has shown promise as filtration media in a point-of-entry arsenic removal system (Bang et al., 2005). Like pure TiO2, TICB has demonstrated removal of both arsenite and arsenate and has demonstrated the ability to oxidize arsenite to arsenate in the presence of UV light (Miller and Zimmerman, 2010). Although arsenic removal by TICB is less than TiO2 on a sorbent weight basis, TICB exceeds TiO2 sorption on a sorbent surface area basis (Dutta et al., 2004; Bang et al., 2005; Miller and Zimmerman, 2010), and has the added advantage of selfseparation. The objective of this study was to optimize the arsenic adsorption capacity and kinetics of TICB and develop a predictive model to guide system conditions for implementation. Parameters affecting capacity and kinetics, including solution pH, bead surface area, bead size, TiO2 loading in the bead, UV irradiation, and background ions in the water matrix, are investigated. Optimization of the useful lifetime of the TICB adsorbent is also considered by examining the potential for regeneration and reuse of TICB.

2.

Materials and methods

2.1.

Standards and reagents

Experiments were conducted with either As(III) (arsenite) or As(V) (arsenate) as indicated. Stock solutions of As(III) and As(V) were prepared by dissolving NaAsO2 (Sigma Aldrich) and Na2HAsO.47H2O (Fisher Scientific) in deionized water, respectively. Stock solutions (10,000 mg/L) and appropriate dilutions were prepared daily, immediately before use. Chitosan was purchased from TCI America. TiO2 (anatase nanopowder, 99.7% trace metals basis, <25 nm particle size) was purchased from Sigma Aldrich. All other reagents were of standard laboratory grade. HNO3 (Fisher Scientific, trace metal grade) and HCl solutions were prepared from concentrated stock solutions; NaOH solutions were prepared from pellets.

2.2.

Bead preparation and characterization

TiO2-impregnated chitosan beads were prepared as reported in (Miller and Zimmerman, 2010). In brief, chitosan was dissolved in 0.1 M HCl (1 g/60 mL). TiO2 (0.4242 g TiO2/1 g chitosan) was added and mixed with a magnetic stir bar until a homogenous solution was achieved. Unless otherwise noted, TICB is 30% TiO2 on mass basis. Syringes with 18G1 needles were filled with the homogenous solution and loaded into a syringe pump; bead size was adjusted by varying the needle gage (19G1, 22G1). The syringe pump discharged the homogenous solution into 0.1 M NaOH (20 mL solution/100 mL 0.1 M NaOH), resulting in beads. Beads were rinsed in

deionized water until filtrate reached pH 6. After drying for > 18 h, beads were collected and stored at room temperature in the dark. BET surface area analysis was performed by Micromeritics Analytical Services (Norcross, Georgia). Samples were weighed at room temperature after degassing at 45  C for 16 h. Gas adsorption analysis was conducted with krypton gas using a TriStar II 3020 surface area and porosity system. Bead porosity was measured with Hg intrusion by Micromeritics Analytical Services (Norcross, Georgia). Bead diameter was measured with a Marathon Electronic Digital Micrometer (0e25 mm). Reported diameter values are the average and standard deviation of 10 randomly selected beads from the same batch.

2.3.

Adsorption experiments

Duplicate samples were prepared in 50 mL polypropylene falcon tubes into which 40 mL arsenic solution (either arsenite or arsenate) and the specified amount of TICB was added. Batch adsorption experiments were conducted in a shaking incubator (VWR 1575R), where temperature was maintained at 25  C and samples were agitated at 150 rpm. Unless otherwise indicated, experiments were conducted for >185 h (Miller and Zimmerman, 2010). Irradiated samples were continuously exposed to an 8 W, 365 nm lamp (UVP, UVL-28 EL Series 8) (Ferguson et al., 2005) suspended 1.5 feet from samples in a closed incubator. In identical experimental conditions to the batch experiments, kinetics of As(III) and As(V) were measured in the presence and absence of UV light. These studies were conducted by sacrificing duplicate samples at specified time intervals for up to 317 h. pH experiments were conducted in the presence and absence of UV light, where [As]0 ¼ 500 mg/L pH adjustments were made with 16 M HNO3 or 1 M NaOH. Synthetic groundwater was prepared based on a procedure by (Leupin and Hug, 2005). MgCO3 (112.7 mg), CaCO3 (367 mg), and KH2PO4 (11.0 mg) were added to 900 mL of deionized water and stirred, resulting in pH 9.06. Under rapid mixing, CO2 gas was bubbled through this solution for 42 min, resulting in pH 5.08. A 10 mL solution of Na2SiO.39H2O (205.9 mg) dissolved in deionized water was added to the bulk solution and stirred, resulting in pH 5.39. Compressed air (house air filtered through granular activated carbon) was bubbled through the solution for 15 min, resulting in pH 6.93. 40 mL deionized water and 50 mL of freshly prepared 10 ppm As(III) stock solution was added to the solution to achieve 500 mg As (III)/L initial concentration and pH 7.19. Groundwater was collected from a tubewell at Amhiribad High School in West Bengal, India on December 17, 2010. The tubewell was flushed for 5 min prior to collection. Water composition was analyzed by EnviroCheck Laboratory in West Bengal and is as follows: 62 ppb As, 0.83 ppm Fe, 0.37 ppm Mn, 0.66 ppm P, 30.66 ppm Si, 7.5 ppm SO24 , 33.18 ppm Mg, 101.0 ppm Ca, 408.2 ppm HCO 3 , 48.04 ppm Cl, 2.78 ppm K, and 14.44 ppm Na. Tubewell water was transported to Yale University where batch experiments and analyses were performed as previously described.

5747

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 4 5 e5 7 5 4

2.4.

Analyte concentration

To measure arsenic concentration, all samples were diluted to a maximum of 100 mg As/L and acidified with 16 M HNO3 to reach a final concentration of 1% HNO3. Samples with neat TiO2 nanopowder were filtered through a 0.45 mm filter prior to dilution. Measurement of As, Mg, Ca, Si, and P was conducted on a Perkin Elmer DRC-e ICP-MS. An internal Germanium standard was used, and quality control standards were analyzed every 10 samples to verify instrument performance. Three readings for each sample were performed and an average and standard deviation for each sample was reported. Detection limit was determined to be 0.25 mg As/L. Arsenic speciation was measured using HPLCICP-MS (Perkin Elmer DRC-e) following published procedures (Neubauer et al., 2004).

3.

Results and discussion

3.1. Equilibrium adsorption capacity of TICB for dissolved arsenic 3.1.1.

pH

Solution pH is a critical variable in metal oxide chemistorptive processes. Fig. 1 shows data for arsenite and arsenate removal by TICB without UV irradiation as a function of pH, where arsenite and arsenate both exist. These results, obtained by deliberately altering solution pH, are consistent with a hydroxide exchange mechanism. Above pH 7.25, the pzc of TICB (Miller and Zimmerman, 2010), electrostatic repulsion between arsenate and hydroxyl groups on the bead surface prevents chemisorption. Neuturally charged arsenite is effectively sorbed up to its pKa, 9.2, where it becomes a negative oxyanion. Differences between arsenate and arsenite removal at pH > 9.2 can be attributed to the fact that arsenate (H2AsO2 4 ) is more negatively charged than arsenite (H2AsO 3 ) in this pH

100

As(III)

80

% As removal

For desorption/resorption experiments, TICB (25 mg) was saturated with As(III) or As(V), as indicated, in a batch experiment without the presence of UV light, where [As]0 varied from 100e10,000 mg/L as indicated. Samples were removed after > 185 h in the shaking incubator (25  C, 150 rpm), and the aqueous solution was immediately decanted from the falcon tube and measured for arsenic concentration. Deionized water (10 mL) was added to the falcon tube to rinse the beads. After gentle agitation, the rinse water was decanted, and the beads, contained in uncapped falcon tubes, were allowed to dry in the fumehood for 24 h. Beads were then reweighed and transferred to a new 50 mL polypropylene falcon tube. 40 mL of NaOH (0.07 M, 0.67 M, 1.67 M) was added to these falcon tubes and placed in the shaking incubator, where temperature was maintained at 25  C and samples were agitated at 150 rpm for 24 h. NaOH was then decanted and measured for arsenic concentration. Deionized water (10 mL) was added to the falcon tube to rinse the beads. After gentle agitation, the rinse water was decanted and the beads, contained in uncapped falcon tubes, were allowed to dry in the fumehood for 24 h. This process was repeated for a total of three cycles.

As(V)

60 40 20 0 4

6

8

10

12

Final pH Fig. 1 e Role of final solution pH in arsenite and arsenate removal by TICB, where data for pH < 4.4 has been omitted because of bead dissolution. Experiments conducted in absence of UV light, where [As]0 [ 500 mg/L.

range and, therefore, experiences more repulsion with net negatively charged TICB (Miller and Zimmerman, 2010).

3.1.2.

Bead size

Surface area measurements were performed across adjustable parameters in the synthesis and preparation of TICB, including bead size (Fig. 2a). TICB surface area increases as bead size decreases and with exposure to UV light. Because UV exposure enhances surface area (Miller and Zimmerman, 2010), surface area differences across bead size are more apparent after exposure to UV light. Given that materials with high surface area can achieve high adsorption capacities per unit mass, these results suggest that an optimized design of TICB might incorporate diameter reduction and UV irradiation. To maximize efficacy, that is removal of As(III) and As(V), there is an inherent tradeoff in that smaller particles (with higher surface area) are more effective but require posttreatment filtration. As such, there is likely an optimal bead size range that balances removal efficiency while maintaining density-separation. Fig. 2b shows the percent As(III) and As(V) arsenic removal across a range of bead sizes, where the final pH is <7.6 for As(III) and <7.2 for As(V), and where the dosing of TiO2: As ratio remains constant to isolate the impact of bead size on efficacy. The largest bead size tested (937 um) demonstrates similar removal capacities to neat TiO2 nanopowder. That is, bead size does not have an effect on the removal efficiency in terms of capacity for a given set of system conditions (i.e., arsenic oxidation state; irradiation). This is an unexpected result because the surface area of neat TiO2 powder is two orders of magnitude greater than that of TICB, and the TiO2 in TICB is bound to a chitosan matrix such that it may not all be available for arsenic removal. This suggests that TICB, in terms of capacity, acts like TiO2 powder given sufficient exposure time. However, this also suggests a need to assess the system kinetics, a critical factor for actual implementation in a field setting. Fig. 2 also illustrates improved performance (i.e., sorption of both As(III) and As(V)) for UV-exposed TICB relative to UVunexposed TICB, indicating that UV light is a variable that may be employed to enhance adsorption capacity.

5748

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 4 5 e5 7 5 4

a

b 100 80

0.6

% As Removal

BET surface area (m2/g)

0.8

0.4

0.2

60 40 20

0

0 0

200

400

600

800

1000

0

200

bead diameter (µm)

400

600

800

1000

bead diameter (µm) As(III); 0 h UV irradiation As(III); 220 h UV irradiation As(V); 0 h UV irradiation As(V); 220 h UV irradiation

220 h UV irradiation 0 h UV irradiation

Fig. 2 e Relationship between bead size (where all beads are 30% TiO2 by weight and have been exposed to UV irradiation as noted) and a) BET surface area and b) % arsenic removal from [As]0 [ 1000 mg/L.

3.1.3.

TiO2 loading

Chitosan beads impregnated with increasing amounts of TiO2 were synthesized, where it was empirically determined that 46% TiO2 by mass was the maximum amount of TiO2 that can be incorporated into a homogenous bead formulation. As shown in Fig. 3a, the surface area of these beads increases with increasing concentration of TiO2 in the bead. These beads were tested for As(III) and As(V) removal, in both the presence and absence of UV light to determine the relationship between TiO2 loading and functional performance. Arsenic removal increases as TiO2 loading increases for all experimental conditions, as illustrated with the representative plot in Fig. 3b; results from other experimental

b

0.8

% As removal

BET surface area (m2/g)

a

conditions are included in Supplementary Information (SI.1). This suggests that all TiO2 in the bead contributes to arsenic adsorption, regardless of the mass fraction of TiO2 in the bead. To confirm that all TiO2 present provides adsorption capacity in TICB, regardless of relative TiO2 mass fraction in the bead, neat TiO2 nanopowder, equivalent in mass to the amount of TiO2 impregnated in the beads, was tested across the same experimental conditions. For removal of both As(III) and As(V), in the presence or absence of UV light, arsenic removal by TICB and TiO2 for a given mass of TiO2 are nearly identical (Fig. 3 and SI.1). This supports the finding that all of the TiO2 impregnated in the bead, not just the TiO2 on the

0.6 0.4 0.2

100 80 60 40 20

0

0 0

0.05

0.1

0.15

mmol TiO2 /25 mg TICB

0

0.05

0.1

0.15

mmol TiO2/ 25 mg TICB As(III)_UV As(V)_UV As(III)_UV, TiO2 powder As(V)_UV, TiO2 powder

Fig. 3 e Relationship between mmol TiO2 incorporated into TICB (875 mm) and a) BET surface area and b) % arsenic removal from [As]0 [ 1000 mg/L in the presence of UV light.

5749

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 4 5 e5 7 5 4

bead surface, is forming chemical complexes with the arsenic in solution, given sufficient time to equilibrate.

3.1.4.

UV light

0.8

UV absent UV present

Final pH (<7)

As removal (mg/g) Experimental observation

6.88  0.04 6.42  0.03

1054  74 1397  11

binuclear complex forms between TiO2 and both arsenite and arsenate (Jegadeesan et al., 2010; Jing et al., 2005, 2009; Pena et al., 2006) (Fig. 5b). A similar complex, shown in Fig. 5c, where one of the TiO2 coordinating sites is replaced with a UVinduced COOH group may form. It has been previously reported that TICB can oxidize As(III) to As(V) in the presence of UV light under system conditions of 365 nm and sufficient oxygen (Miller and Zimmerman, 2010). Experiments conducted in sunlight confirm that sunlight can also induce TICB’s photooxidative process. As(III) samples exposed to sunlight showed 100% conversion to As(V) within 8 days, whereas samples not exposed to sunlight showed <16% conversion to As(V) over the course of 12 days (SI.3). As

100 80

0.6

60 0.4 40 0.2

% Porosity

BET surface area (m2/g)

Beads of identical size were exposed to different durations of UV irradiation. As shown in Fig. 4, there is a positive relationship between surface area and UV exposure, and SEM images of these beads (Miller and Zimmerman, 2010) reveal changes in surface morphology after exposure to UV light. This may be a result of a photooxidative process occurring at the bead surface. Although UV irradiation affects the surface of TICB, porosity measurements of beads with varying exposures to UV light are relatively constant and suggest that UV irradiation does not affect pores >3 nm within the dehydrated bead interior (Webb, 2001). The presence of UV light enhances arsenate sorption on TICB (Miller and Zimmerman, 2010). Because the pH of all samples, including controls, is lowered with UV irradiation (SI.2) and because lower solution pH is associated with greater arsenic removal (Fig. 1) (Miller and Zimmerman, 2010), pH was investigated as a potentially confounding variable. When controlling for the pH fluctuations caused by irradiation, enhanced arsenic removal for samples exposed to UV light is still observed. As shown in Table 1, more As(V) is removed when UV is present than when UV is absent, where final pH is <7. Based on a model developed to predict TICB performance in the presence of UV light (Section 3.3), the 7% difference in final pH would account for a 15% difference in final removal capacities, far less than the actual 28% difference observed between 1054 mg/g and 1397 mg/g, suggesting that UV light does in fact enhance sorption independent of its impact on solution pH. UV irradiation can result in scission of linkages in the chitosan backbone, producing carboxyl groups that do not significantly affect the chemical structure of chitosan (Zubieta et al., 2008). One likely oxidation product of chitosan, where a carboxy group is formed in the C6 position, is shown in Fig. 5a (Ahmed et al., 2003). The addition of new carboxyl groups may provide additional coordinating sites for arsenic oxyanions. Previous studies have reported that a bidentate,

Table 1 e % removal from batch experiment where [As(V)]0 [ 1000 mg/L and samples were analyzed at time [ 240 h. UV absent samples were spiked with HNO3 to match pH drop of UV present samples. Values reported are averages of four replicates.

20

0

0 0

100

200

300

h UV irradiation pretreatment Surface area

Porosity

Fig. 4 e Relationship between UV irradiation duration and BET surface area and porosity, where all beads are w875 mm and 30% TiO2 by weight.

Fig. 5 e Molecular structures relevant to the TICB-As system. a) Potential scheme for oxidation of chitosan (Ahmed et al., 2003). b) Bidentate binuclear complexation of arsenate by TiO2 (Jegadeesan et al., 2010; Jing et al., 2005; Jing et al., 2009; Pena et al., 2006). c) Potential bidentate binuclear complexation of arsenate by TiO2 and oxidation product of chitosan.

5750

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 4 5 e5 7 5 4

decentralized treatment is the intended application, this is a significant result because sunlight is a free, non-resource intensive, and readily available source of UV light.

3.1.5.

Groundwater

Adsorbent behavior in a deionized water matrix may not be predictive of behavior in a natural water matrix, where background ions and pH can control adsorption processes (Schwartzenbach et al., 2003). For arsenic chemisorption by TiO2, in particular, researchers have reported that phosphate and silicate can act as competitive background ions (Jing et al., 2009; Mohan and Pittman Jr., 2007; Viraraghavan and Chowdhury, 2007). In a synthetic groundwater matrix, designed to model water from Bangladesh (Leupin and Hug, 2005), removal of As, Ca, Mg, P, and Si with TICB was tested (Fig. 6). These batch TICB adsorption tests were conducted in the presence and absence of UV light. In all cases, concentrations of Ca, Mg, and Si remained constant, and the concentration of As and P decreased. In the UV irradiated system, percentages of arsenic and of phosphate removed, with a range of TICB dosing, were nearly identical, suggesting that phosphate and arsenate are direct competitors for sorption sites. TICB performance was also tested in a natural arsenic-laden groundwater, collected from a tubewell in West Bengal, India. As with the synthetic groundwater, TICB removed nearly identical percentages of arsenic and phosphate in natural groundwater (SI.4). This suggests the TICB intended for use in the field will require enough TiO2 (either through increased mass per bead or an increased number of beads) to account for removal of both P and As.

3.1.6.

Capacity for arsenic desorption and resorption

From ease of use and sustainability perspectives, an adsorbent that can be simply regenerated and reused minimizes resource consumption and offers economic and environmental advantages. TICB was investigated for regeneration and reuse capacity over several adsorption/desorption cycles. A variety of solvents in which chitosan is insoluble, including base and some acids (Pillai et al., 2009), were evaluated for arsenic desorption from TICB. These regeneration solvents were evaluated at various concentrations to minimize the

a

Consistent with other literature reports (Bang et al., 2005; Pena et al., 2005), equilibrium between arsenic and TiO2 nanopowder in our batch system was reached rapidly, within 2 h (SI.6). Equilibrium between our “standard” TICB system (w875 um in diameter, 30% TiO2 loading by mass) and arsenic, however, was not reached for at least seven days. This was the case for As(III) and As(V), in the absence or presence of UV light. The adsorption kinetics of TICB systems with varying bead size and UV exposure (identical to those reported in Section 3.1.2 (Fig. 2) and Section 3.1.4 (Fig. 4), respectively)

% analyte removal

% analyte removal

60

Kinetics of adsorption

UV present 100

As Ca Mg P Si

80

3.2.

b

UV absent 100

amount of resources required for regeneration and to minimize hazard associated with aqueous arsenic-laden waste after regeneration. Of potential desorbing solvents screened, NaOH was the most effective. Kinetics of As(III)- and As(V)- desorption from saturated TICB were tested for a range of NaOH concentrations (SI.5). In all cases, >50% of arsenic adsorbed was desorbed within 2 h. Of the NaOH concentrations tested (0.07 M, 0.67 M, 1.67 M), 0.07 M NaOH demonstrated the most rapid and most complete desorption for both As(III) and As(V). For three cycles of adsorption/desorption, TICB achieved equilibrium with a non-buffered arsenic solution of either 10,000 mg As(III)/L or 10,000 mg As(V)/L. During desorption, Assaturated beads were equilibrated with NaOH (0.07 M, 0.67 M, 1.67 M). For all NaOH concentrations, > 60% of the arsenic adsorbed, whether As(III) or As(V), is desorbed; with 0.07 M NaOH, >85% of As(III) adsorbed and >78% of As(V) adsorbed was desorbed each of the three cycles. A representative plot of these adsorption/desorption cycles is shown in Fig. 7. In addition to investigating bead reuse and resorption from 10,000 mg/L arsenic solutions, other concentrations of greater relevance to source waters were tested. Resorption isotherms for up to three cycles, where initial As concentration ranged from 100 mg/L to 10,000 mg/L, are shown in Fig. 8. These isotherms show sustained and possibly improved resorption performance for As(III) but slightly varied resorption behavior for As(V), where the final solution pH can account for the changes in the observed amounts of As(V) resorbed.

40 20 0

As Ca Mg P Si

80 60 40 20 0

0

0.02

0.04

0.06

0.08

g TICB/ 40 mL groundwater

0.1

0

0.02 0.04 0.06 0.08

0.1

g TICB/ 40 mL groundwater

Fig. 6 e Ion removal in synthetic groundwater matrix by TICB where a) UV is absent, and b) UV is present.

5751

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 4 5 e5 7 5 4

% As(III) Removal

100 80 60 40 20 0 0

100

200

300

400

Time (h) TICB ∅ ~875 μm TICB ∅ ~745 μm TiO2∅ ~0.025 μm Fig. 7 e Mass of As(III) sorbed onto TICB from a 10,000 mg/L stock solution and subsequently desorbed from TICB with 0.07 M NaOH for three cycles.

were evaluated to identify potential strategies to minimize this kinetic limitation which would likely be unacceptable in a field situation. These experiments were conducted without buffering to assess the kinetics of the system without interference from buffering ions. However, because pH can influence the rate of arsenic adsorption by TiO2 (Dutta et al., 2004), solution pH was closely monitored and was found to remain relatively constant throughout kinetic experiments.

3.2.1.

TICB size

The relationship between bead size and rate of removal was investigated. Kinetics of arsenic adsorption by TICB with

a

Fig. 9 e % As(III) removal in the absence of UV light for neat TiO2 nanopowder, 744 um TICB and 875 mm TICB, where the mass of TiO2 per sample is equal and [As]0 [ 500 mg/L.

diameter of 875 um and 744 um as well as neat TiO2 nanopowder are shown in Fig. 9. In these experiments, the mass of TiO2 in the system is constant for the two TICB sizes and the neat powder. Results indicate that a slight reduction (<15%) in the size of standard TICB can significantly increase the rate of arsenic removal, so much so that these slightly smaller TICB directly mimic neat TiO2 powder. Water and free molecules can travel within chitosan, which forms hydrogels in aqueous solution (Berger et al., 2004a, 2004b). Hydration of a hydrogel requires diffusion of water molecules into the polymer network, relaxation of

b

As(III)

µg As resorbed/g TICB

3000

µg As resorbed/g TICB

As(V) 3000

2500 2000 1500 1000 500 0

2500 2000 1500 1000 500 0

0

5000

10000

0

2000 4000 6000 8000 10000

[As]e (µg/L)

[As]e (µg/L)

Sorption cycle 1

Sorption cycle 1

Sorption cycle 2

Sorption cycle 2

Sorption cycle 3

Sorption cycle 3

Fig. 8 e Resorption isotherms over three adsorption/desorption cycles with 0.07 M NaOH where for a) As(III), where final pH [ 9.27, 8.85, 8.88 for sorption cycles 1, 2, and 3 respectively, and b) As(V), where final pH [ 8.74, 7.46, 7.72 for sorption cycles 1, 2, and 3, respectively.

5752

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 4 5 e5 7 5 4

Table 2 e Kinetic calculations for TICB at 25  C, Orpm, where [As]0 [ 10,000 mg/L and bead diameter w875 mm. Pseudo 1st qe, calc (mg/g) As(III) removal As(V) removal

UV UV UV UV

absent present absent present

2861.6 4708.4 1940.8 3807.9

2892.9 3102.6 2540.2 3827.6

k (1/hour) 9.30 1.39 7.54 1.56

   

103 102 103 102

R2 0.904 0.868 0.886 0.946

polymer chains, and expansion of the polymer network into the surrounding bulk water medium (Okano, 1998). Enhanced kinetics observed for smaller beads reflects the fact that less diffusion is required for dissolved arsenic oxyanions to reach potential TiO2 binding sites when bead diameter is reduced.

3.2.2.

UV light

Several kinetic models, including power function, pseudo first order and pseudo second order, were tested to fit experimental data for arsenic sorption by TICB, models previously reported for similar metal-sorbent systems with photooxidation (Jing et al., 2009; Ofomaja et al., 2010). For all systems examined, that is regardless of arsenic species and the presence of UV light, the pseudo first order model best predicted the equilibrium sorption capacity, and best fit parameters for the system at 25  C are shown in Table 2. Not only does UV irradiation result in higher equilibrium capacities, but UV irradiation also results in higher rates of arsenic removal. The increased rate of arsenite and arsenate removal in the presence of UV light can be attributed to the availability of more arsenic binding sites as described above. Identical kinetic testing was conducted outdoors with fluctuating temperature, both with and without exposure to sunlight; a pseudo first order model also best predicts equilibrium sorption capacity for these data, and sunlight resulted in higher removal capacities as well as rates of removal (SI.7).

3.3.

Toward a predictive TICB sorption capacity model

Data presented in Section 3.1 indicate that arsenic removal in a given TICB system is directed by the solution pH and by the TiO2 content in the bead and is enhanced in the presence of

Table 3 e Semi-empirical model to predict TICB performance for pH range 4e11 and % TiO2 (by mass) range 0e46. Results apply to reference experiment, where equilibrium is reached after 220 h between 25 mg TICB and 40 mL 1000 mg As/L solution. a mg As=g TICB ¼ 1 þ ð20:8=%TiO2 Þ  e^ðpH  pKa Þ Experimental conditions As(III) As(V) AsUV(III) þ AsUV(V)

a

pka

599 880 1658

9.2 6.98 6.98

µ g As sorbed / g TICB

qe, exp (mg/g)

model 10% TiO2 model 30% TiO2 10% TiO2; R2=0.87 30% TiO2; R2=0.94

1600

1200

800

400

0 4

6

8

10

12

pH Fig. 10 e Arsenic removal (mg As/g TICB) across pH for beads of varying TiO2 loading. Experimental and model data are compared for beads loaded with 10% TiO2 by weight and 30% TiO2 by weight.

UV light. A semi-empirical model, based on the logistic function, was developed to predict TICB sorption capacity from solution pH and TiO2 content in the bead. Based on our mechanistic understanding that TICB-As complexes form through hydroxide exchange and can only form at pH values where electrostatic repulsion does not occur, a logistic function was chosen. The logistic function depicts the sigmoidal shape of the arsenic removal data across pH, analogous to the shape of a titration curve across pH. Fitting parameters were derived from batch equilibrium experiments between arsenic ([As]0 ¼ 1000 mg/L) and TICB (30% TiO2 by weight, 875 mm in diameter), across the pH range 4e11. This model, shown in Table 3, was optimized for three experimental conditions: As(III), As(V), and AsUV(III) þ AsUV(V). The different adsorption capacities for the experimental conditions are accounted for in parameter a. Different inflection points observed for As(III) and As(V) across pH (Fig. 1) are a result of repulsive charges between negatively charged TICB and negative arsenic oxyanions. These inflection points are determined by the relevant pKas for As(III), As(V), and AsUV(III) þ AsUV(V), where all arsenic exists as As(V). Experimental data is compared to this model for the experimental condition AsUV(III) þ AsUV(V) in Fig. 10; R2 values for TiO2 compositions of 10% and 30% by weight are 0.87 and 0.94, respectively.

4.

Conclusions

The optimization of TiO2-impregnated chitosan bead, TICB, as an arsenic sorbent was performed. As TiO2 loading increases, TICB surface area increases, and arsenic removal capacity of TICB loaded with a given amount of TiO2 is similar to the arsenic removal capacity of the neat TiO2 nanopowder equivalence. pH is a critical variable in the TICB sorption system, where As(III) removal is greatest below 9.2 and As(V) removal is greatest below 7.25. Sorption capacity by TICB is

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 4 5 e5 7 5 4

determined to be a function of pH and % TiO2 loading, and a model was developed to predict arsenic sorption, for given experimental conditions, based on these two variables. Across pH 4e11 in the presence of UV light, this model predicts adsorption capacity with R2 values of 0.87 and 0.94 for 10% TiO2 beads and 30% TiO2 beads, respectively. Reduction in bead diameter does not influence sorption capacity, given sufficient equilibrium time. That is, the largest TICB (800 mme940 mm) results in arsenic removal that is similar to that of an equivalent dose of neat TiO2 nanopowder. However, bead diameter reductions increase bead surface area and increase rate of arsenic removal. Exposure to UV light also increases bead surface area and the rate of arsenic removal but does not affect the porosity of the bead. Long detention times are not practical for point-of-use systems, and future work will aim to minimize detention times through bead diameter reductions and incorporation of UV irradiation. Laboratory tests were performed to predict performance of TICB in a field setting. In synthetic and natural groundwater, phosphate is found to be a direct competitor for arsenic sorption sites on TICB. Other ions tested, including Ca, Mg, and Si, do not compete for sorption sites on TICB. In the presence of sunlight and TICB, dissolved As(III) is oxidized to As(V) within 8 days. Finally, TICB can be regenerated with weak NaOH, and TICB can be reused without losing effectiveness for at least three cycles in batch experiments.

Acknowledgments We are grateful to Jamila Yamani for her laboratory assistance and for her support of this project. Funding was provided by the National Science Foundation Graduate Research Fellowship and National Science Foundation Environmental Sustainability (CBET-0932060).

Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.08.040.

references

Agros, M., Kalra, T., Rathouz, P.J., Chen, Y., Pierce, B., Farvez, F., Islam, T., Ahmed, A., Rakibuz-Zaman, R., Hasan, R., Sarwar, G., Slavkovich, V., van Geen, A., Graziano, J., Ahsan, H., 2010. Arsenic exposure from drinking water, and all-cause and chronic-disease mortalities in Bangladesh (HEALS): a prospective cohort study. The Lancet 376 (9737), 252e258. Ahmed, G.A.-W., Khairou, K.S., Hassan, R.M., 2003. Kinetics and mechanism of oxidation of chitosan polysaccharide by permanganate ion in aqueous perchlorate solutions. Journal of Chemical Research S, 182e183. Bang, S., Patel, M., Lippincott, L., Meng, X., 2005. Removal of arsenic from groundwater by granular titanium dioxide adsorbent. Chemosphere 60, 389e397. Berger, J., Reist, M., Mayer, J.M., Felt, O., Gurny, R., 2004a. Structure and interactions in chitosan hydrogels formed by

5753

complexation or aggregation for biomedical applications. European Journal of Pharmaceutics and Biopharmaceutics 57 (1), 35e52. Berger, J., Reist, M., Mayer, J.M., Felt, O., Peppas, N.A., Gurny, R., 2004b. Structure and interactions in covalently and ionically crosslinked chitosan hydrogels for biomedical applications. European Journal of Pharmaceutics and Biopharmaceutics 57 (1), 19e34. Bhattacharya, P., Mukherjee, A.B., Bundschuh, J., Zevenhoven, R., Loeppert, R.H. (Eds.), 2007. Arsenic in Soil and Groundwater Environment. Elsevier, Amsterdam. Deebar, N., Irfan, A., Ishtiaq, Q.A., 2009. Evaluation of the adsorption potential of titanium dioxide nanoparticles for arsenic removal. Journal of Environmental Sciences 21, 402e408. Dutta, P.K., Ray, A.K., Sharma, V.K., Millero, F.J., 2004. Adsorption of arsenate and arsenite on titanium dioxide suspensions. Journal of Colloid and Interface Science 278, 270e275. Ferguson, M.A., Hoffmann, M.R., Hering, J.G., 2005. TiO2photocatalyzed As(III) oxidation in aqueous suspensions: reaction kinetics and effects of adsorption. Environmental Science & Technology 39, 1880e1886. Gerente, C., Lee, V.K.C., Le Cloirec, P., McKay, G., 2007. Application of chitosan for the removal of metals from wastewaters by adsorption - mechanisms and models review. Critical Reviews in Environmental Science and Technology 37 (1), 41e127. Jegadeesan, G., Al-Abed, S.R., Sundaram, V., Choi, H., Scheckel, K.G. , Dionysiou, D.D., 2010. Arsenic sorption on TiO2 nanoparticles: size and crystallanity effects. Water Research 44, 965e973. Jing, C., Liu, S., Patel, M., Meng, X., 2005. Arsenic leachability in water treatment adsorbents. Environmental Science & Technology 39, 5481e5487. Jing, C., Meng, X., Calvache, E., Jiang, G., 2009. Remediation of organic and inorganic arsenic contaminated groundwater using a nanocrystalline TiO2-based adsorbent. Environmental Pollution 157, 2514e2519. Leupin, O.X., Hug, S.J., 2005. Oxidation and removal of arsenic (III) from aerated groundwater by filtration through sand and zero-valent iron. Water Research 39, 1729e1740. Miller, S.M., Zimmerman, J.B., 2010. Novel, bio-based, photoactive arsenic sorbent: TiO2-impregnated chitosan bead. Water Research 44, 5722e5729. Mohan, D., Pittman Jr., C.U., 2007. Arsenic removal from water/ wastewater using adsorbents - A critical review. Journal of Hazardous Materials 142, 1e53. Muzzarelli, R.A.A., Muzzarelli, C., 2005. Polysaccharides 1: Structure, Characterization and Use, pp. 151e209. Neubauer, K.R., Reuter, W., Perrone, P., Grosser, Z., 2004. In: Services, P.L.a.A. (Ed.), Simultaneous Arsenic and Chromium Speciation by HPLC/ICP-MS in Environmental Waters. PerkinElmer, Shelton, CT. Ofomaja, A.E., Baidoo, E.B., Modise, S.J., 2010. Kinetic and pseudosecond-order modeling of lead biosorption onto pine cone powder. Industrial & Engineering Chemistry Research 49, 2562e2572. Okano, T. (Ed.), 1998. Biorelated Polymers and Gels. Academic Press, San Diego. Pena, M., Meng, X., Korfiatis, G.P., Jing, C., 2006. Adsorption mechanism of arsenic on nanocrystalline titanium dioxide. Environmental Science & Technology 40, 1257e1262. Pena, M.E., Korfiatis, G.P., Patel, M., Lippincott, L., Meng, X., 2005. Adsorption of As(V) and As(III) by nanocrystalline titanium dioxide. Water Research 39, 2327e2337. Petrusevski, B., Sharma, S., van der Meer, W.G., Kruis, F., Khan, M., Barua, M., Schippers, J.C., 2008. Four years of development and field-testing of IHE arsenic removal family filter in rural Bangladesh. Water Science and Technology 58 (1), 53e58.

5754

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 4 5 e5 7 5 4

Pillai, C.K.S., Paul, W., Sharma, C.P., 2009. Chitin and chitosan polymers: Chemistry, solubility, and fiber formation. Progress in Polymer Science 34, 641e678. Ratnaike, R.N., 2003. Acute and chronic arsenic toxicity. Postgraduate Medical Journal 79, 391e396. Schwartzenbach, R.P., Gschwend, P.M., Imboden, D.M., 2003. Environmental Organic Chemistry. John Wiley and Sons, Inc., Hoboken, New Jersey. Tien, V.N., Chaudhary, D.S., Ngo, H.H., Vigneswaran, S., 2004. Arsenic in water: concerns and treatment technologies. Journal of Industrial and Engineering Chemistry 10 (3), 337e348.

Viraraghavan, T., Chowdhury, P., 2007. Recent trends in arsenic removal technologies. Progress in Environmental Science and Technology, 692e694. Webb, P.A., 2001. An Introduction to the Physical Characterization of Materials by Mercury Intrusion Porosimetry with Emphasis on Reduction and Presentation of Experimental Data Norcross, Georgia. Zubieta, C.E., Messina, P.V., Luengo, C., Dennehy, M., Pieroni, O., Schulz, P.C., 2008. Reactive dyes remotion by porous TiO2chitosan materials. Journal of Hazardous Materials 152, 765e777.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 5 5 e5 7 6 3

Available online at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

One year environmental surveillance of rotavirus specie A (RVA) genotypes in circulation after the introduction of the Rotarix
vaccine in Rio de Janeiro, Brazil Tulio Machado Fumian a,*, Jose´ Paulo Gagliardi Leite a, Tatiana Lundgreen Rose a, Tatiana Prado b, Marize Pereira Miagostovich a a

Laboratory of Comparative and Environmental Virology, Oswaldo Cruz Institute, Oswaldo Cruz Foundation (Fiocruz), Av. Brasil 4.365, Manguinhos, CEP 21040-360, Rio de Janeiro (RJ), Brazil b Laboratory of Technological Development in Virology, Oswaldo Cruz Institute, Oswaldo Cruz Foundation (Fiocruz), Av. Brasil 4.365, Manguinhos, CEP 21040-360, Rio de Janeiro (RJ), Brazil

article info

abstract

Article history:

Rotavirus specie A (RVA) infection is the leading cause of severe acute diarrhea among

Received 26 May 2011

young children worldwide. To reduce this major RVA health impact, the Rotarix
vaccine

Received in revised form

(GlaxoSmithKline, Rixensart, Belgium) was introduced in the Brazilian Expanded Immu-

24 August 2011

nization Program in March 2006 and became available to the entire birth cohort. The aim of

Accepted 25 August 2011

this study was to evaluate the spread of RVA in the environment after the introduction of

Available online 1 September 2011

Rotarix
in Brazil. For this purpose, a Wastewater Treatment Plant (WTP) in Rio de Janeiro was monitored for one year to detect, characterize and discriminate RVA genotypes and

Keywords:

identify possible circulation of vaccine strains. Using TaqMan
quantitative PCR (qPCR),

Rotavirus A genotypes

RVA was detected in 100% (mean viral loads from 2.40
105 to 1.16
107 genome copies

Rotarix
vaccine

(GC)/L) of sewage influent samples and 71% (mean viral loads from 1.35
103 to

Wastewater

1.64
105 GC/L) of sewage effluent samples. The most prevalent RVA genotypes were P[4],

Wastewater treatment plant

P[6] and G2, based on VP4 and VP7 classification. Direct nucleotide sequencing (NSP4 fragment) and restriction enzyme digestion (NSP3) analysis did not detect RVA vaccine-like strains from the sewage samples. These data on RVA detection, quantification and molecular characterization highlight the importance of environmental monitoring as a tool to study RVA epidemiology in the surrounding human population and may be useful on ongoing vaccine monitoring programs, since sewage may be a good screening option for a rapid and economical overview of the circulating genotypes. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Rotavirus specie A (RVA) is the main etiological agent of viral gastroenteritis in infants throughout the world and is associated with significant mortality in developing countries, where over 600,000 deaths occur annually (Parashar et al., 2006). In

developed countries, this virus remains a common cause of morbidity with significant economic burden (Charles et al., 2006; Parashar et al., 2006). RVA belongs to the Reoviridae family, Rotavirus genus, and possesses a double-stranded RNA (dsRNA) genome with 11 segments that encode six structural (VP) and six non-

* Corresponding author. Tel.: þ55 21 25621875; fax: 55 21 25621851. E-mail address: [email protected] (T.M. Fumian). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.039

5756

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 5 5 e5 7 6 3

structural proteins (NSP) (Estes and Kapikian, 2007). A widely used binary classification scheme has been established based on the two genes that codify the outer capsid proteins, VP4 and VP7, defining G (from VP7, glycoprotein) and P (from VP4, protease-cleaved protein) genotypes. Currently 27 G and 35 P genotypes are recognized (Abe et al., 2009; Solberg et al., 2009; Ursu et al., 2009; Matthijnssens et al., 2011); however, only five RVA G genotypes (G1eG4 and G9) and two P genotypes (P[8] and P[4]) are prevalent worldwide (Santos and Hoshino, 2005; Ursu et al., 2009). RVA virions are shed in extremely high concentrations (up to 1010 virus/g) in the stool of infected children with acute gastroenteritis and can persist in the environment for long periods of time (Carter, 2005; Bosch et al., 2008). The features of the virions, including stability in aqueous environments and resistance to water treatment, may facilitate their transmission to humans via contaminated water (Ansari et al., 1991; Espinosa et al., 2008). Direct sewage discharge into environmental waters such as lagoons, rivers, beaches and coastal waters represents a public health problem mainly in developing countries. These contaminated waters have been broadly linked to the causation of several waterborne gastroenteritis outbreaks (Kukkula et al., 1997; Villena et al., 2003a; Schmid et al., 2005; Godoy et al., 2006). Despite the difficulty of determining the proportion of gastroenteritis cases due to contaminated water, it has been suggested that a significant percentage of the cases are related to the quality of the water (Bosch et al., 2008). As RVA is one of the most important causes of mortality in infants worldwide, two equally safe and efficacious live oral rotavirus vaccines, G1P[8] RVA vaccine (RV1 e Rotarix
, GlaxoSmithKline, Rixensart, Belgium) and a pentavalent G1G4 and P[8] RVA vaccine (RV5 e RotaTeq
, Merck and Co., Whitehouse Station, NJ, USA), were developed and are licensed for use in more than 100 countries worldwide (Jiang et al., 2010). The first one, RV1, was included in the Brazilian Expanded Immunization Program (PNI) in March 2006 and became available to the entire birth cohort. The impact of this vaccine on the circulating RVA genotypes is unknown and difficult to predict, so continuous genotype surveillance is needed to identify the effects of the vaccine program on circulating strains, particularly on genotype prevalence and the emergence of uncommon strains. The monitoring of the viruses circulating in sewage from a wastewater treatment plant (WTP) has been described as an appropriate model to understand the spread of RVA in the population served by the WTP, as influents may contain viruses shed from patients with sporadic or asymptomatic cases (Haramoto et al., 2006; Bosch et al., 2008). The main goal of this study was to evaluate the spread of RVA in the environment following the introduction of the Rotarix
vaccine in Brazil. For this purpose, a WTP located in Rio de Janeiro was monitored for one year to detect, quantify and characterize RVA genotypes and to investigate the possible presence of the vaccine strain in sewage samples. RVA genomes were investigated in samples collected from raw and treated sewage using Taqman
quantitative PCR (qPCR), and the P (VP4) and G (VP7) genotypes were characterized by nested PCR in a multiplex reaction (Gentsch et al., 1992; Gouvea et al., 1994). Protocols based on direct

nucleotide sequencing (NSP4) and restriction enzyme digestion (NSP3) analysis (Rose et al., 2010) were applied to discriminate between wild-type and vaccine strains.

2.

Materials and methods

2.1. Wastewater treatment plant (WTP) sample collection Sewage samples were collected from an urban WTP located in the metropolitan area of Rio de Janeiro, Brazil. The WTP receives sewage from around 1.5 million inhabitants leaving in both the central and north zone of the city and is one of the largest in Brazil. Sewage treatment employs a secondary treatment (aerobic process: activated sludge) with an inflow mean of 2500 L s1. Initial sewage treatment is composed of grid separation and primary sedimentation (five primary settling tanks with a volume of 7700 m3 each). There are four aeration tanks in parallel (volume: 11,500 m3 per tank) with a capacity to treat 625 L s1 of effluent. Secondary sedimentation is performed in four secondary settling tanks (volume: 8800 m3 per tank) with no chlorination before effluents are discharged into the water environment. A total of 48 sewage samples were collected bi-monthly (15 day interval) from August 2009 to July 2010, 24 of them were collected from raw sewage (influent) and 24 from the final treated sewage (effluent). At each sampling point, 50 ml of sewage was collected in sterile plastic bottles, kept at 4  C and transported to the laboratory for immediate analysis.

2.2.

Virus concentration

Viruses were concentrated using the ultracentrifugation method as described by Pina et al. (1998). To avoid false negative results and to evaluate the presence of inhibitors, sewage samples were inoculated with 500 ml of an internal control (bacteriophage PP7) before the concentration assay, and the extracted RNA was diluted 10-fold (Fumian et al., 2010).

2.3. Nucleic acid extraction, reverse transcription (RT) and quantitative PCR (qPCR) The viral dsRNA was extracted by the glass powder method (Boom et al., 1990), and the synthesis of cDNA was carried out by reverse transcription using a random primer (PdN6 e 50 A260 units e Amersham Biosciences, Chalfont St Giles, Buckinghamshire, UK). Multiplex qPCR to detect RVA and PP7 was performed as described previously (Fumian et al., 2010) using primers described by Zeng et al. (2008) and Rajal et al. (2007). RVA primers and probe were designed to target a highly conserved region of the non-structural protein 3 (NSP3), and the PP7 primers to amplify a region of the PP7 replicase gene. Both were synthesized by Applied Biosystems (CA, USA). qPCR was carried out using an ABI PRISM 7500 Sequence Detection System (Applied Biosystems, CA, USA). A standard curve (SC; 107, 105, 103 and 101 copies per reaction) was generated using 10-fold serial dilutions of a pCR2.1 vector (Invitrogen, USA) containing either the RVA NSP3 gene or the PP7 replicase gene.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 5 5 e5 7 6 3

2.4.

VP4 and VP7 nested PCR amplification

Nested PCR was used for molecular characterization of RVA genotypes G and P, and it partially amplified VP7 and VP4 segments, respectively. In the first-round, RT-PCR was performed with VP7 and VP4 consensus primers 9con1e9con2 (Das et al., 1994) and 4con2e4con3 (Gentsch et al., 1992), respectively. Following the first-round, RVA G genotype classification was performed using specific primers for genotypes G1eG4, G5 and G9 (Das et al., 1994; Gouvea et al., 1994), and P genotype classification was carried out using primers for genotypes P[4], P[6] and P[8]-P[10], described by Gentsch et al. (1992).

2.5.

RVA molecular characterization

To discriminate RVA wild-type G1P[8] from the vaccine strain, the NSP4 gene was amplified according to the protocol described by Cunliffe et al. (1998). Two nucleotide mutations after the first initiation ATG codon at positions 100 and 134 were observed when the NSP4 nucleotide sequence of the Rotarix
vaccine (patent number: PCT/EP2004/009725) was compared to sequences available in GenBank including reference strains. These two nucleotide shifts were used to classify RVA (data not shown). The NSP4 PCR amplicons were purified and sequenced using an ABI Prism BigDye Terminator Cycle Sequencing Ready Reaction Kit
and an ABI Prism 3730 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA). The chromatograms were analyzed using BioEdit
(Hall, 1999). A phylogenetic dendrogram was constructed by the neighbor-joining method using a matrix of genetic distances established under the Kimura-two parameter model (Felsenstein, 1993) using MEGA V. 4.0 (Tamura et al., 2007). The robustness of each node was assessed by bootstrap analysis using 2000 pseudoreplicates. RVA NSP4 isolated from sewage samples was classified according to the most recent full genome-based classification proposed by Matthijnssens et al. (2011).

2.6.

Cloning and restriction endonuclease analysis

To characterize vaccine strains, another protocol based on BspHI restriction endonuclease analysis of the NSP3 gene was

5757

performed as previously described (Rose et al., 2010). Prior to endonuclease restriction analysis, PCR amplicons generated by NSP3 amplification (Matthijnssens et al., 2006) were cloned into the PCR4-TOPO vector (Invitrogen, USA) following the manufacturer’s recommendations.

2.7.

Statistical analysis

The total frequency of detection obtained in WTP, in both influent and effluent samples, using qPCR assay was compared by using a chi-square test and Fisher’s exact test at a significance level of 0.05. The same statistical analysis was performed to determine significant differences between VP4 and VP7 PCR detection in all of the 48 samples collected. Analysis of Variance (ANOVA) was performed to determine differences in mean levels of RVA, present in influent samples, during the four seasons (summer, fall, spring and winter), and a paired t-test was performed to verify differences between the mean levels of RVA in influent and effluent samples throughout the study. Statistical analyses were performed using GraphPad Prism
software version 5.

3.

Results

3.1.

Rotavirus A detection and quantification

The RVA genome levels and genotypes were determined in a one-year monitoring study from influent and effluent streams at a WTP located in Rio de Janeiro city, Brazil. Using qPCR, 41 out of 48 (85%) of the samples were positive, corresponding 100% (24/24) of influent and 71% (17/24) of effluent samples. The difference in the total frequencies of RVA detection (qPCR) in WTP was significant between influent and effluent samples ( p ¼ 0.0042, Chi-square; p ¼ 0.0047, Fisher). Fig. 1 shows the monthly distribution of RVA genome copies (GC/L) and the standard deviation. For sewage influent, RVA concentrations ranged from 2.40
104 to 1.16
107 GC/L, and in effluent, positive sample concentrations, ranged from 1.35
103 to 1.64
105 GC/L. The differences in mean levels of RVA present in influent samples throughout the seasons were not significant (ANOVA/ NewmaneKeuls Multiple

Fig. 1 e Monthly distribution of rotavirus specie A (RVA) in influent (A) and effluent (B) samples from a WTP in Brazil. The plots show the geometric monthly mean values (GC/L). The upper and lower bars show the standard deviations of the mean values.

5758

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 5 5 e5 7 6 3

Comparison Test), on the other hand, difference in the mean levels of RVA was significant between influent and effluent samples ( p ¼ 0.041, Paired t-test). Table 1 summarizes the results obtained from the genomic amplification protocols used for detection, quantification (qPCR) and molecular characterization of RVA genes (RT-PCR for NSP4, VP4 and VP7). The RVA genotype G2 was detected in 100% (24/24) of influent samples, and the genotypes P[4] and P [6] were detected in 33% (8/24) and 25% (6/24), respectively. Effluent samples showed a lower RVA detection rate, with genotypes G2 and P[4] being detected in 25% (6/24) and 4% (1/ 24) of samples, respectively. Genotypes G1 and P[8] were not identified in the samples tested. The difference in the frequency of RVA detection using VP4 and VP7 PCR was significant ( p ¼ 0.002, Chi-square; p ¼ 0.004, Fisher). RT-PCR based on the NSP4 gene was able to detect viruses in 92% (22/24) and 21% (5/24) of influent and effluent samples, respectively. No evidence of inhibitors was observed as bacteriophage PP7, inoculated as an internal control in all 48 sewage samples, was detected in 100% of samples tested using a multiplex qPCR. PP7 viral titers recovered ranged from 3.4
105 to 1.6
104 GC per 500 ml of PP7 suspension.

3.2.

Rotavirus A strain characterization

The sequence of gene segment 10 (encoding NSP4) of Brazilian waste samples were compared with NSP4 segments available

Table 1 e Rotavirus specie A (RVA) detection from influent (24) and effluent (24) sewage samples by quantitative (qPCR) and qualitative (NSP4, VP4, VP7) PCR protocols for genotyping. Year Month

Sewage influent

Aug Sep Oct Nov Dec

2010

Jan Feb Mar Apr May Jun Jul

þ þ þ þ þ þ þ þ þ þ þ þ þ þ þ þ þ þ þ þ þ þ þ þ

 þ þ þ þ þ þ þ þ þ þ þ þ þ þ þ þ þ þ þ  þ þ þ

þ: Positive; and : negative.

 P[4] P[4]  P[6] P[4] P[6] P[4] P[4] P[6] P[6]  P[6]      P[4]   P[4] P[6] P[4]

G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 G2

4.

Discussion

Sewage effluent

qPCR NSP4 VP4 VP7 qPCR NSP4 VP4 VP7 2009

in GenBank. The phylogenetic analysis of this segment classified the samples within two distinct genotypes: E1 genotype, with a single sequence (RJ-VA-550) and E2 genotype, in which 10 sequences clustered. Among the sequences that clustered in genotype E2, the sequence obtained from RJ-VA-575 sample clustered in a separate branch of the tree (Fig. 2). None of the 11 NSP4 sequences showed the two nucleotide mutations as in the vaccine pattern. RVA NSP4 nucleotide sequences obtained in the present study were deposited at the National Center for Biotechnology Information (GenBank, http://www.ncbi.nlm.nih.gov/) under the accession numbers: RVA/Env-wt/BRA/RJ-VA-500/2009/ GXP[X]: JF731369; RVA/Env-wt/BRA/RJ-VA-515/2009/GXP[X]: JF731370; RVA/Env-wt/BRA/RJ-VA-518/2009/GXP[X]: JF731371; RVA/Env-wt/BRA/RJ-VA-521/2009/GXP[X]: JF731372; RVA/Envwt/BRA/RJ-VA-523/2009/GXP[X]: JF731373; RVA/Env-wt/BRA/ RJ-VA-524/2009/GXP[X]: JF731374; RVA/Env-wt/BRA/RJ-VA526/2009/GXP[X]: JF731375; RVA/Env-wt/BRA/RJ-VA-527/2009/ GXP[X]: JF731376; RVA/Env-wt/BRA/RJ-VA-547/2009/GXP[X]: JF731377; RVA/Env-wt/BRA/RJ-VA-550/2009/GXP[X]: JF731378; and RVA/Env-wt/BRA/RJ-VA-575/2010/GXP[X]: JF731379. In order to accurately determine the presence of the vaccine components in the environment, six influent samples with high viral loads (1.7
107e3.7
105 GC/L) were subjected to PCR amplification of the region of the genome encoding NSP3, and the resulting products were cloned. Forty-seven colonies were screened for the appropriate banding pattern after BspHI restriction endonuclease analysis, and none of them demonstrated the vaccine pattern. A Rotarix
NSP3 amplicon was analyzed in the same reaction as a positive control.

 þ þ þ þ þ þ þ þ  þ þ þ   þ þ þ  þ þ   þ

     þ þ         þ þ       

     P[4]                  

    G2 G2 G2         G2 G2 G2      

In this study, an environmental approach was used to evaluate the circulation of RVA genotypes in the city of Rio de Janeiro, Brazil, which has the second largest population in the country. Samples from a large WTP were analyzed using a concentration method (ultracentrifugation) and molecular techniques to detect, quantify and characterize the detected viruses (Pina et al., 1998; Fumian et al., 2010). This type of approach has been extensively employed to obtain information on circulating viruses in populations throughout the world, independently of single reported cases or outbreaks, and to assess virus circulation causing asymptomatic infections (Bosch et al., 2008; Gajardo et al., 1995; Haramoto et al., 2006; Clemente-Casares et al., 2009; Fumian et al., 2010; Kamel et al., 2010; Prado et al., 2011). Besides revealing the predominant genotypes circulating in Rio de Janeiro, this monitoring strategy also aimed to investigate the presence of the attenuated G1P[8] RVA vaccine Rotarix
in the environment. The combination of virus concentration and molecular detection methods was successfully employed, and the results showed a high level of RVA contamination in sewage samples. The high recovery rate of RVA (47%) from sewage samples (Fumian et al., 2010), using the ultracentrifugation method, was fundamental for the success in RVA recovering. Another study using this ultracentrifugation method to

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 5 5 e5 7 6 3

5759

Fig. 2 e Phylogenetic dendrogram based on partial NSP4 nucleotide sequences of rotavirus A strains isolated from sewage samples in this study. All sequences obtained from GenBank are named according to Matthijnssens et al. (2011), and G and P genotypes are indicated at the right. The Brazilian environmental samples are marked with a filled diamond (influent samples) and an unfilled diamond (effluent samples). The scale bar at the bottom of the tree indicates distance. Bootstrap values (2000 replicates) are shown at the branch nodes and values lower than 50% are not shown.

recover RVA from domestic sewage and polluted water river samples demonstrated a high percentage of positive samples: 67% and 83% (Rodrı´guez-Dı´az et al., 2009). Lower RVA detection rates have been observed when membrane-active charged filtration was used as a concentration method associated with organic or inorganic elution (Ferreira et al., 2009; Kamel et al., 2010). A pattern of seasonality of RVA-induced gastroenteritis has been demonstrated in Latin American countries, including Brazil, based on a higher incidence of infection occurring in winter months (Kane et al., 2004; Carvalho-Costa et al., 2011). However, this differential distribution was not observed by the analysis of sewage samples during the monitoring period, suggesting a high level of virus shedding occurring throughout the year.

The average reduction of 2 logarithms in viral load observed in effluent samples demonstrates that WTPs play an important role in reducing environmental contamination. However, as demonstrated in this study, the persistence of such viruses in treated effluents and in other studies from different regions, highlights the importance of evaluating the efficiency of different types of treatments used by WTPs in viral load reduction (Bofill-Mas et al., 2006; Haramoto et al., 2006; da Silva et al., 2007; Meleg et al., 2008; La Rosa et al., 2010). Despite the difficulties in associating virus infection to contact with contaminated water, the environmental dissemination of RVA, demonstrated by the high prevalence and concentration in the treated or untreated sewage samples, poses a risk to human health that must be considered and evaluated. Although the detection of nucleic acid does not

5760

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 5 5 e5 7 6 3

directly indicate the presence of infectious viruses, it is strongly suggestive of an infectious particle (Girones et al., 2010). Different studies have demonstrated that signals generated after RT-PCR amplification of viral genomes correlated well with infectivity or that a great part of viral nucleic acid recovered from environmental samples corresponded to infectious virus particles (Bhattacharya et al., 2004; Espinosa et al., 2008; Barrella et al., 2009). The results obtained in this study regarding RVA dissemination, along with other studies conducted in developing countries, indicate RVA as a possible viral indicator of human fecal contamination in environmental samples, at least in countries where there is a high RVA prevalence (Ferreira et al., 2009; Miagostovich et al., 2008; Rodrı´guez-Dı´az et al., 2009; Prado et al., 2011; Sdiri-Loulizi et al., 2010). Data concerning virus genotyping provide significant epidemiological information necessary for the introduction and ongoing monitoring of vaccination programs (Villena et al., 2003b; Pinto et al., 2007; Bosch et al., 2008). The NSP4 segment analysis showed samples clustered with two genotypes. E1 sequence was close related to NSP4 from genotypes G1P[8], G3P[8] and G12P[6] and was probably associated with P [6] genotype detected. Within E2 genotype, nine samples formed a monophyletic group, and one sequence (RJ-VA-575) clustered in another group, including Brazilian G2P[4], isolated in 2008. The high detection frequency of E2 genotype, with G2 and P[4], is in agreement with trends described by Matthijnssens et al. (2011), where strains with a G2 and P[4] genotype presented an E2 profile and strains with a G1, G3, G12, P[6] and P[8] genotypes demonstrated an E1 profile. The prevalence of G2 and P[4] genotypes in sewage samples is in agreement with the results obtained in a previous survey using clinical samples from acute infantile gastroenteritis cases in the municipality of Rio de Janeiro after Rotarix
introduction (Carvalho-Costa et al., 2009, 2011). RVA P[6] genotype detected in a lower prevalence than P[4] in sewage sample, reflects results obtained from clinical samples, showing that, in Brazil, the major circulation of G2P[4] and in a slight ratio, G2P[6] genotype. Data from the Laboratory of Comparative and Environmental Virology (LVCA), a Brazilian Regional Reference Laboratory for Rotaviruses, from 2009 to 2010, showed a higher percentage (78%) of RVA G2P[4] circulating in Rio de Janeiro when compared with other genotypes characterized as G4P[8] (6%); G9P[X] (6%); G2P[6], G2P[X] and G1P[X] (3%) (data not published). Although an increasing prevalence of genotype G2P[4] has also been reported in countries that have not established Rotarix
vaccination programs (Ferrera et al., 2007; Antunes et al., 2009), it is important to note that in a smaller WTP sewage monitoring program also conducted in Rio de Janeiro in 2005, before RVA vaccine introduction, G1 and P[8] were the most prevalent RVA genotypes detected (Ferreira et al., 2009). This change in the RVA genotypes prevalence profile could be explained by a natural genotypic fluctuation, although the role of the Rotarix
vaccine introduction cannot be ruled out (Go´mez et al., 2011). In Australia, where both vaccine types are used, it was observed a higher prevalence of G2P[4] genotypes in states that used exclusively Rotarix
vaccine when compared with states that used Rotateq
, showing a higher prevalence of G3P[8] strains (Kirkwood et al., 2011). Another important

aspect regarding the RVA vaccine is that Rotarix
prevents around 90% of severe gastroenteritis cases caused by G1P[8], as well as other partially heterotypic strains; however, it is less effective (45%) in preventing diarrhea caused by fully heterotypic G2P[4] strains (Ruiz-Palacios et al., 2006). Linhares et al. (2008), when evaluating Rotarix
efficacy against rotavirus gastroenteritis in a phase III study performed in Latin American infants, demonstrated a vaccine efficacy of 82% and 40% against G1P[8] and G2P[4], respectively. Genotypes G1 and P[8] were not found in these sewage samples, showing that these genotypes are no longer circulating or are circulating at a very low level, reinforcing data obtained via surveillance of clinical specimens (CarvalhoCosta et al., 2009, 2011). The high shedding of RVA antigens (up to 1010 virus/g) from naturally infected individuals may restrict the detection of the vaccine strain that would be less prevalent in the environment. In a study conducted in South Africa during 2003e2004 to evaluate the safety, reactogenicity and immunogenicity of the Rotarix
vaccine, virus shedding was observed in healthy infants, ranging from 31% to 46% depending on the vaccination regimen used (Steele et al., 2010). The cloning of the gene NSP3 followed by restriction enzyme analysis was an alternative attempt to increase the probability of vaccine strain detection, as cloning of the PCR products enables detection of genotypes that are at lower abundance in the environment. This methodology based on NSP3 gene amplification followed by BspHI digestion was previously described for discrimination of the Rotarix
vaccine (Rose et al., 2010). Epidemiological and laboratory surveillances to assess vaccine effectiveness and vaccine impact are currently significant concerns (WHO, 2008). As Brazil was the first Latin American country to introduce universal rotavirus vaccination, the evaluation of vaccine performance to examine possible changing strain patterns of RVA in circulation is a priority in this country. Sentinel RVA surveillance in selected pediatric settings has been recommended as part of the immunization program in Latin America (Carvalho-Costa et al., 2009). Environmental surveillance, as conducted in this study by investigating RVA in sewage samples, could be an alternative approach to support clinical monitoring of RVA infection. This kind of surveillance would allow continuous investigation of the genotypes circulating in the WTP service area, providing an overview of the prevalent genotypes and possibly discriminating between RVA vaccine and wild strains.

5.

Conclusion

(1) The high circulation of RVA in the population was measurable by environmental surveillance coupled with appropriate molecular tools. (2) Wastewater surveillance demonstrated that genotypes G2 and P[4] were the most prevalent, reflecting a natural fluctuation of RVA genotypes or a consequence of Rotarix
vaccine introduction or even both. (3) This is the first study concerning RVA detection and discrimination between vaccine and wild-type strains

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 5 5 e5 7 6 3

from environmental samples carried out in a WTP and may assist clinical epidemiological studies that will be essential in the post-vaccination era.

Acknowledgements This work was financially sponsored by the National Council for Scientific and Technological Development (CNPq e PROSUL 490292/2008-9; CNPq e PAPES V) and by CGVAM/Ministry of Health, Brazil. The authors thank the staff of PDTIS DNA Sequencing Platform at FIOCRUZ (RPT01A) for technical support in sequencing reactions and the WTP staff for supplying the samples, under the agreement between Fiocruz and the Water Company of Rio de Janeiro state (CEDAE). This research study is under the scope of the activities of Fiocruz as a collaborating center of PAHO/WHO of Public and Environmental Health.

references

Abe, M., Ito, N., Morikawa, S., Takasu, M., Murase, T., Kawashima, T., Kawai, Y., Kohara, J., Sugiyama, M., 2009. Molecular epidemiology of rotaviruses among healthy calves in Japan: isolation of a novel bovine rotavirus bearing new P and G genotypes. Virus Res. 144, 250e257. Ansari, S.A., Springthorpe, V.S., Sattar, S.A., 1991. Survival and vehicular spread of human rotaviruses: possible relation to seasonality of outbreaks. Rev. Infect. Dis. 13, 448e461. Antunes, H., Afonso, A., Iturriza, M., Martinho, I., Ribeiro, C., Rocha, S., Magalha˜es, C., Carvalho, L., Branca, F., Gray, J., 2009. G2P[4] the most prevalent rotavirus genotype in 2007 winter season in an European non-vaccinated population. J. Clin. Virol. 45, 76e78. Barrella, K.M., Garrafa, P., Monezi, T.A., Ha´rsi, C.M., Salvi, C., Violante, P.A.B.C., Mehnert, D.U., 2009. Longitudinal study on occurrence of adenoviruses and hepatitis A virus in raw domestic sewage in the city of Limeira, Sa˜o Paulo. Braz. J. Microbiol. 40, 102e107. Bhattacharya, S.S., Kulka, M., Lampel, K.A., Cebula, T.A., Goswami, B.B., 2004. Use of reverse transcription and PCR to discriminate between infectious and non-infectious hepatitis A virus. J. Virol. Methods 116, 181e187. Bofill-Mas, S., Albinana-Gimenez, N., Clemente-Casares, P., Hundesa, A., Rodriguez-Manzano, J., Allard, A., Calvo, M., Girones, R., 2006. Quantification and stability of human adenoviruses and polyomavirus JCPyV in wastewater matrices. Appl. Environ. Microbiol. 72, 7894e7896. Boom, R., Sol, C.J., Salimans, M.M., Jansen, C.L., Wertheim-van Dillen, P.M., van der Noordaa, J., 1990. Rapid and simple method for purification of nucleic acids. J. Clin. Microbiol. 28, 495e503. Bosch, A., Guix, S., Sano, D., Pinto´, R.M., 2008. New tools for the study and direct surveillance of viral pathogens in water. Curr. Opin. Biotechnol. 19, 295e301. Carter, M.J., 2005. Enterically infecting viruses: pathogenicity, transmission and significance for food and waterborne infection. J. Appl. Microbiol. 98, 1354e1380. Carvalho-Costa, F.A., Arau´jo, I.T., Santos de Assis, R.M., Fialho, A. M., de Assis Martins, C.M., Bo´ia, M.N., Leite, J.P., 2009. Rotavirus genotype distribution after vaccine introduction, Rio

5761

de Janeiro, Brazil. Emerg. Infect. Dis. 15, 95e97. Carvalho-Costa, F.A., Volota˜o Ede, M., de Assis, R.M., Fialho, A.M., de Andrade Jda, S., Rocha, L.N., Tort, L.F., da Silva, M.F., Go´mez, M.M., de Souza, P.M., Leite, J.P., 2011. Laboratorybased rotavirus surveillance during the introduction of a vaccination program, Brazil, 2005e2009. Pediatr. Infect. Dis. J. 30, S35eS41. Charles, M.D., Holman, R.C., Curns, A.T., Parashar, U.D., Glass, R.I., Bresee, J.S., 2006. Hospitalizations associated with rotavirus gastroenteritis in the United States, 1993e2002. Pediatr. Infect. Dis. J. 25, 489e493. Clemente-Casares, P., Rodriguez-Manzano, J., Girones, R., 2009. Hepatitis E virus genotype 3 and sporadically also genotype 1 circulate in the population of Catalonia, Spain. J. Water Health 7, 664e673. Cunliffe, N.A., Kilgore, P.E., Bresee, J.S., Steele, A.D., Luo, N., Hart, C.A., Glass, R.I., 1998. Epidemiology of rotavirus diarrhoea in Africa: a review to assess the need for rotavirus immunization. Bull. World Health Organ. 76, 525e537. da Silva, A.K., Le Saux, J.C., Parnaudeau, S., Pommepuy, M., Elimelech, M., Le Guyader, F.S., 2007. Evaluation of removal of noroviruses during wastewater treatment, using real-time reverse transcription-PCR: different behaviors of genogroups I and II. Appl. Environ. Microbiol. 73, 7891e7897. Das, B.K., Gentsch, J.R., Cicirello, H.G., Woods, P.A., Gupta, A., Ramachandran, M., Kumar, R., Bhan, M.K., Glass, R.I., 1994. Characterization of rotavirus strains from newborns in New Delhi, India. J. Clin. Microbiol. 32, 1820e1822. Espinosa, A.C., Mazari-Hiriart, M., Espinosa, R., Maruri-Avidal, L., Me´ndez, E., Arias, C.F., 2008. Infectivity and genome persistence of rotavirus and astrovirus in groundwater and surface water. Water Res. 42, 2618e2628. Estes, M., Kapikian, A.Z., 2007. Rotaviruses. In: Fields Virology, fifth ed. Lippincott Williams &Wilkins, Philadelphia. Felsenstein, J., 1993. Phylogeny Interference Package, Version 3. 5. Department of Genetics, University of Washington, Seattle, USA. Ferreira, F.F., Guimara˜es, F.R., Fumian, T.M., Victoria, M., Vieira, C.B., Luz, S., Shubo, T., Leite, J.P., Miagostovich, M.P., 2009. Environmental dissemination of group A rotavirus: P-type, G-type and subgroup characterization. Water Sci. Technol. 60, 633e642. Ferrera, A., Quan, D., Espinoza, F., 2007. Increased prevalence of genotype G2P(4) among children with rotavirus-associated gastroenteritis in Honduras. In: 17th European Congress of Clinical Microbiology and Infectious Diseases ICC, Munich, Germany. Fumian, T.M., Leite, J.P., Castello, A.A., Gaggero, A., Caillou, M.S., Miagostovich, M.P., 2010. Detection of rotavirus A in sewage samples using multiplex qPCR and an evaluation of the ultracentrifugation and adsorption-elution methods for virus concentration. J. Virol. Methods 170, 42e46. Gajardo, R., Bouchriti, N., Pinto, R.M., Bosch, A., 1995. Genotyping of rotaviruses isolated from sewage. Appl. Environ. Microbiol. 61, 3460e3462. Gentsch, J.R., Glass, R.I., Woods, P., Gouvea, V., Gorziglia, M., Flores, J., Das, B.K., Bhan, M.K., 1992. Identification of group A rotavirus gene 4 types by polymerase chain reaction. J. Clin. Microbiol. 30, 1365e1373. Girones, R., Ferru´s, M.A., Alonso, J.L., Rodriguez-Manzano, J., Calgua, B., Correˆa Ade, A., Hundesa, A., Carratala, A., BofillMas, S., 2010. Molecular detection of pathogens in waterethe pros and cons of molecular techniques. Water Res. 44, 4325e4339. Godoy, P., Nuı´n, C., Alseda`, M., Llovet, T., Mazana, R., Domı´nguez, A., 2006. Waterborne outbreak of gastroenteritis caused by Norovirus transmitted through drinking water. Rev. Clin. Esp. 206, 435e437.

5762

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 5 5 e5 7 6 3

Go´mez, M.M., de Mendonc¸a, M.C., Volota˜o Ede, M., Tort, L.F., da Silva, M.F., Cristina, J., Leite, J.P., 2011. Rotavirus A genotype P [4]G2: genetic diversity and reassortment events among strains circulating in Brazil between 2005 and 2009. J. Med. Virol. 83, 1093e1106. Gouvea, V., de Castro, L., Timenetsky, M.C., Greenberg, H., Santos, N., 1994. Rotavirus serotype G5 associated with diarrhea in Brazilian children. J. Clin. Microbiol. 32, 1408e1409. Hall, T.A., 1999. BioEdit: a user-friendly biological sequence alignment editor and analysis program for Windows 95/98/NT. Nucl. Acids Symp. Ser. 41, 95e98. Haramoto, E., Katayama, H., Oguma, K., Yamashita, H., Tajima, A., Nakajima, H., Ohgaki, S., 2006. Seasonal profiles of human noroviruses and indicator bacteria in a wastewater treatment plant in Tokyo, Japan. Water Sci. Technol. 54, 301e308. Jiang, V., Jiang, B., Tate, J., Parashar, U.D., Patel, M.M., 2010. Performance of rotavirus vaccines in developed and developing countries. Hum. Vaccin. 6, 532e542. Kamel, A.H., Ali, M.A., El-Nady, H.G., Aho, S., Pothier, P., Belliot, G., 2010. Evidence of the co-circulation of enteric viruses in sewage and in the population of Greater Cairo. J. Appl. Microbiol. 108, 1620e1629. Kane, E.M., Turcios, R.M., Arvay, M.L., Garcia, S., Bresee, J.S., Glass, R.I., 2004. The epidemiology of rotavirus diarrhea in Latin America. Anticipating rotavirus vaccines. Rev. Panam. Salud. Publica 16, 371e377. Kirkwood, C.D., Boniface, K., Barnes, G.L., Bishop, R.F., 2011. Distribution of rotavirus genotypes after introduction of rotavirus vaccines, Rotarix
and RotaTeq
, into the National Immunization Program of Australia. Pediatr. Infect. Dis. J. 30, 48e53. Kukkula, M., Arstila, P., Klossner, M.-L., Maunula, L., von Bonsdorff, C.-H., Jaatinen, P., 1997. Waterborne outbreak of viral gastro-enteritis. Scand. J. Infect. Dis 29, 415e418. La Rosa, G., Pourshaban, M., Iaconelli, M., Muscillo, M., 2010. Quantitative real-time PCR of enteric viruses in influent and effluent samples from wastewater treatment plants in Italy. Ann. Ist. Super. Sanita. 46, 266e273. Linhares, A.C., Vela´zquez, F.R., Pe´rez-Schael, I., Sa´ez-Llorens, X., Abate, H., Espinoza, F., Lo´pez, P., Macı´as-Parra, M., OrtegaBarrı´a, E., Rivera-Medina, D.M., Rivera, L., Pavı´a-Ruz, N., Nun˜ez, E., Damaso, S., Ruiz-Palacios, G.M., De Vos, B., O’Ryan, M., Gillard, P., Bouckenooghe, A., 2008. Human Rotavirus Vaccine Study Group. Efficacy and safety of an oral live attenuated human rotavirus vaccine against rotavirus gastroenteritis during the first 2 years of life in Latin American infants: a randomised, double-blind, placebo-controlled phase III study. Lancet 371, 1181e1189. Matthijnssens, J., Rahman, M., Martella, V., Xuelei, Y., De Vos, S., De Leener, K., Ciarlet, M., Buonavoglia, C., Van Ranst, M., 2006. Full genomic analysis of human rotavirus strain B4106 and lapine rotavirus strain 30/96 provides evidence for interspecies transmission. J. Virol. 80, 3801e3810. Matthijnssens, J., Ciarlet, M., McDonald, S.M., Attoui, H., Ba´nyai, K., Brister, J.R., Buesa, J., Esona, M.D., Estes, M.K., Gentsch, J.R., Iturriza-Go´mara, M., Johne, R., Kirkwood, C.D., Martella, V., Mertens, P.P., Nakagomi, O., Parren˜o, V., Rahman, M., Ruggeri, F.M., Saif, L.J., Santos, N., Steyer, A., Taniguchi, K., Patton, J.T., Desselberger, U., Van Ranst, M., 2011. Uniformity of rotavirus strain nomenclature proposed by the Rotavirus Classification Working Group (RCWG). Arch. Virol. 156, 1397e1413. Meleg, E., Ba´nyai, K., Martella, V., Jiang, B., Kocsis, B., Kisfali, P., Melegh, B., Szucs, G., 2008. Detection and quantification of group C rotaviruses in communal sewage. Appl. Environ. Microbiol. 74, 3394e3399. Miagostovich, M.P., Ferreira, F.F., Guimaraes, F.R., Fumian, T.M., Diniz-Mendes, L., Luz, S.L., Silva, L.A., Leite, J.P., 2008. Molecular

detection and characterization of gastroenteritis viruses occurring naturally in the stream waters of Manaus, central Amazonia, Brazil. Appl. Environ. Microbiol. 74, 375e382. Parashar, U.D., Gibson, C.J., Bresse, J.S., Glass, R.I., 2006. Rotavirus and severe childhood diarrhea. Emerg. Infect. Dis. 12, 304e306. Pina, S., Jofre, J., Emerson, S.U., Purcell, R.H., Girones, R., 1998. Characterization of a strain of infectious hepatitis E virus isolated from sewage in an area where hepatitis E is not endemic. Appl. Environ. Microbiol. 64, 4485e4488. Pinto, R.M., Alegre, D., Dominguez, A., El Senousy, W.M., Sanchez, G., Villena, C., Costafreda, M.I., Aragones, L., Bosch, A., 2007. Hepatitis A vı´rus in urban sewage from two Mediterranean countries. Epidemiol. Infect. 135, 270e273. Prado, T., Silva, D.M., Guilayn, W.C., Rose, T.L., Gaspar, A.M., Miagostovich, M.P., 2011. Quantification and molecular characterization of enteric viruses detected in effluents from two hospital wastewater treatment plants. Water Res. 45, 1287e1297. Rajal, V.B., McSwain, B.S., Thompson, D.E., Leutenegger, C.M., Kildare, B.J., Wuertz, S., 2007. Validation of hollow fiber ultrafiltration and real-time PCR using bacteriophage PP7 as surrogate for the quantification of viruses from water samples. Water Res. 41, 1411e1422. Rodrı´guez-Dı´az, J., Querales, L., Caraballo, L., Vizzi, E., Liprandi, F., Takiff, H., Betancourt, W.Q., 2009. Detection and characterization of waterborne gastroenteritis viruses in urban sewage and sewage-polluted river waters in Caracas, Venezuela. Appl. Environ. Microbiol. 75, 387e394. Rose, T.L., Miagostovich, M.P., Leite, J.P., 2010. Rotavirus A genotype G1P[8]: a novel method to distinguish wild-type strains from the Rotarix vaccine strain. Mem. Inst. Oswaldo Cruz 105, 1068e1072. Ruiz-Palacios, G.M., Pe´rez-Schael, I., Vela´zquez, F.R., Abate, H., Breuer, T., Clemens, S.C., Cheuvart, B., Espinoza, F., Gillard, P., Innis, B.L., Cervantes, Y., Linhares, A.C., Lo´pez, P., Macı´asParra, M., Ortega-Barrı´a, E., Richardson, V., Rivera-Medina, D. M., Rivera, L., Salinas, B., Pavı´a-Ruz, N., Salmero´n, J., Ru¨ttimann, R., Tinoco, J.C., Rubio, P., Nun˜ez, E., Guerrero, M.L., Yarza´bal, J.P., Damaso, S., Tornieporth, N., Sa´ez-Llorens, X., Vergara, R.F., Vesikari, T., Bouckenooghe, A., Clemens, R., De Vos, B., O’Ryan, M., 2006. Human rotavirus vaccine study group. Safety and efficacy of an attenuated vaccine against severe rotavirus gastroenteritis. N. Engl. J. Med. 354, 11e22. Santos, N., Hoshino, Y., 2005. Global distribution of rotavirus serotypes/genotypes and its implication for the development and implementation of an effective rotavirus vaccine. Rev. Med. Virol. 15, 29e56. Schmid, D., Lederer, I., Much, P., Pichler, A.M., Allerberger, F., 2005. Outbreak of norovirus infection associated with contaminated flood water, Salzburg, 2005. Euro. Surveill. 16, 10e16. Sdiri-Loulizi, K., Hassine, M., Aouni, Z., Gharbi-Khelifi, H., Chouchane, S., Sakly, N., Neji-Gue´diche, M., Pothier, P., Aouni, M., Ambert-Balay, K., 2010. Detection and molecular characterization of enteric viruses in environmental samples in Monastir, Tunisia between January 2003 and April 2007. J. Appl. Microbiol. 109, 1093e1104. Solberg, O.D., Hasing, M.E., Trueba, G., Eisenberg, J.N., 2009. Characterization of novel VP7, VP4, and VP6 genotypes of a previously untypeable group A rotavirus. Virology 385, 58e67. Steele, A.D., Reynders, J., Scholtz, F., Bos, P., de Beer, M.C., Tumbo, J., Van der Merwe, C.F., Delem, A., De Vos, B., 2010. Comparison of 2 different regimens for reactogenicity, safety, and immunogenicity of the live attenuated oral rotavirus vaccine RIX4414 coadministered with oral polio vaccine in South African infants. J. Infect. Dis. 202, S93e100. Tamura, K., Dudley, J., Nei, M., Kumar, S., 2007. MEGA4: molecular evolutionary genetics analysis (MEGA) software version 4.0. Mol. Biol. Evol. 24, 1596e1599.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 5 5 e5 7 6 3

Ursu, K., Kisfali, P., Rigo, D., Ivanics, E., Erdelyi, K., Dan, A., Melegh, B., Martella, V., Ba´nyai, K., 2009. Molecular analysis of the VP7 gene of pheasant rotaviruses identifies a new genotype, designated G23. Arch. Virol. 154, 1365e1369. Villena, C., El-Senousy, W.M., Abad, F.X., Pinto, R.M., Bosch, A., 2003a. Group A rotavirus in sewage samples from Barcelona and Cairo: emergence of unusual genotypes. Appl. Environ. Microbiol. 69, 3919e3923. Villena, C., Gabrieli, R., Pinto, R.M., Guix, S., Donia, D., Buonomo, E., Palombi, L., Cenko, F., Bino, S., Bosch, A.,

5763

Divizia, M., 2003b. A large infantile gastroenteritis outbreak in Albania caused by multiple emerging rotavirus genotypes. Epidemiol. Infect. 131, 1105e1110. WHO, 2008. Generic Protocol for Monitoring Impact of Rotavirus Vaccination on Gastroenteritis Disease Burden and Viral Strains. Available at:. WHO www.who.int/vaccines-documents/. Zeng, S.Q., Halkosalo, A., Salminen, M., Szakal, E.D., Puustinen, L., Vesikari, T., 2008. One-step quantitative RT-PCR for the detection of rotavirus in acute gastroenteritis. J. Virol. Methods 153, 238e240.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 6 4 e5 7 7 2

Available online at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Managed aquifer recharge of treated wastewater: Water quality changes resulting from infiltration through the vadose zone Elise Bekele a,*, Simon Toze b,c, Bradley Patterson a,d, Simon Higginson e a

CSIRO Water for a Healthy Country Flagship, CSIRO Centre for Environment and Life Sciences, Private Bag No 5, PO Wembley, Western Australia 6913, Australia b CSIRO Water for a Healthy Country Flagship, Ecosciences Precinct, 41 Boggo Road, Dutton Park, QLD 4102, Australia c School of Population Health, University of Queensland, Herston Road, Herston, QLD 4006, Australia d School of Biomedical, Biomolecular and Chemical Sciences, University of Western Australia, Crawley, WA 6009, Australia e Water Corporation of Western Australia, PO Box 100, Leederville, WA 6902, Australia

article info

abstract

Article history:

Secondary treated wastewater was infiltrated through a 9 m-thick calcareous vadose zone

Received 8 June 2011

during a 39 month managed aquifer recharge (MAR) field trial to determine potential

Received in revised form

improvements in the recycled water quality. The water quality improvements of the

23 August 2011

recycled water were based on changes in the chemistry and microbiology of (i) the recycled

Accepted 27 August 2011

water prior to infiltration relative to (ii) groundwater immediately down-gradient from the

Available online 3 September 2011

infiltration gallery. Changes in the average concentrations of several constituents in the recycled water were identified with reductions of 30% for phosphorous, 66% for fluoride,

Keywords:

62% for iron and 51% for total organic carbon when the secondary treated wastewater was

Managed aquifer recharge

infiltrated at an applied rate of 17.5 L per minute with a residence time of approximately

Wastewater infiltration

four days in the vadose zone and less than two days in the aquifer. Reductions were also

Natural attenuation processes

noted for oxazepam and temazepam among the pharmaceuticals tested and for a range of microbial pathogens, but reductions were harder to quantify as their magnitudes varied over time. Total nitrogen and carbamazepine persisted in groundwater down-gradient from the infiltration galleries. Infiltration does potentially offer a range of water quality improvements over direct injection to the water table without passage through the unsaturated zone; however, additional treatment options for the non-potable water may still need to be considered, depending on the receiving environment or the end use of the recovered water. Crown Copyright ª 2011 Published by Elsevier Ltd. All rights reserved.

1.

Introduction

An essential design question for proponents of managed aquifer recharge (MAR) schemes using recycled water in shallow aquifers is whether it is more beneficial to recharge the aquifer via direct well injection or to use infiltration with

either ponds, basins or shallow buried trenches. There is a range of different types of MAR as described in Dillon (2005). Ideally the design of a MAR scheme in an urban setting should have minimal surface footprint and minimal exposure potential for the community. Well injection offers advantages such as no evaporative loss, algae or mosquitoes, and no loss

* Corresponding author. Tel.: þ61 0 8 9333 6718; fax: þ61 0 8 9333 6211. E-mail address: [email protected] (E. Bekele). 0043-1354/$ e see front matter Crown Copyright ª 2011 Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.058

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 6 4 e5 7 7 2

of prime land (Pyne, 2006). Well injection, however, can suffer from clogging issues relating to the initial quality of the recycled water and loses the benefit of potential treatment processes provided by the unsaturated zone. In comparison, surface infiltration offers potential treatment during the migration of recycled water through the vadose zone, which can result in improved water quality via biological, chemical and physical processes before reaching the water table but can have the drawback of having a larger surface footprint and potential for exposed water, none of which is ideal in an urban environment. Improvements to water quality using infiltration have been demonstrated to reduce organic matter (Quanrud et al., 2003; Vanderzalm et al., 2010), trace organic compounds (Montgomery-Brown et al., 2003), nitrogen (Zhang et al., 2005) and bacteria (Schafer et al., 1998; Toze et al., 2004). However, a rigorous determination of the magnitudes of concentration reductions for a range of different chemical and biological contaminants in treated wastewater has not been conducted previously in MAR studies using unsaturated, calcareous sands such as the Spearwood sands of Swan Coastal Plain of Western Australia. The aim of this study was to determine the changes in recycled water quality after infiltrating vertically through a 9 m-thick vadose zone and migrating laterally through 2.3 m of aquifer using infiltration galleries as the MAR method during a 39 month MAR field trial. The focus of this work is on MAR in calcareous sand and limestone due to the prevalence of these deposits, which comprise unconfined aquifers in the Perth metropolitan region of Western Australia where there is also keen interest in enhancing the role of MAR (Scatena and Williamson, 1999; Smith and Pollock, 2010). The evaluation of water quality was based on measured concentrations of major ions, nutrients, trace metals, organic carbon, pharmaceutical compounds (carbamazepine, diazepam, oxazepam, phenytoin, temazepam) and numbers of faecal indicator microorganisms and selected enteric pathogens (thermotolerant coliforms, enterococci, bacteriophage, adenovirus) before and after recycled water passed through the subsurface.

2.

Materials and methods

2.1.

Facilities

The source of secondary treated wastewater was the Subiaco Wastewater Treatment Plant in Western Australia. After passage through a multi-media filtration system (AMIAD), the treated wastewater was pumped to the recharge site (Bekele et al., 2009). Two infiltration galleries were used at the CSIRO Centre for Environment and Life Sciences in Floreat, Western Australia for a pilot-scale investigation of MAR (Bekele et al., 2009). The east gallery was filled with 10 mm graded and washed granite gravel; the west gallery contained a series of modular polypropylene tanks, referred to as the Atlantis system by the manufacturer. The dimensions of each tank were 685 mm  408 mm  450 mm (length by width by height) and the tanks have a modular design so that they can be positioned in the trench and clipped together. The construction of each module resembles a milk crate that has

5765

large void spaces supported by matrix. In this study, each gallery trench was 25 m  1 m  0.5 m (length by width by height) and the width of the Atlantis gallery was 816 mm, consisting of two tank modules positioned side-by-side. Each gallery was buried to a depth of 1 m below ground and covered with 0.5 m of sediment backfill composed of Spearwood Dune sand from digging the trench (Bekele et al., 2009). The recharge site was located in a grassy paddock where sheep were allowed to graze. The infiltration galleries operated almost continuously for 39 months and infiltrated a total of 36.7 ML of treated wastewater to the aquifer supplied at a daily constant rate of 17.5 L per minute (Bekele et al., 2009).

2.2.

Local geology and hydraulic considerations

The geology of the site consisted of a 7 m-thick top layer of Spearwood Dune sand, overlying the Tamala Limestone aquifer. The Tamala Limestone is a calcareous aeolinite which has been weathered to produce the overlying Spearwood sands (Tapsell et al., 2003). The aquifer extends to a depth of 31 m below ground and is underlain by sediments from a regional aquitard. Regional investigations of the Tamala Limestone reveal high porosity zones from fractures and cavities (Davidson, 1995). The sand mineralogy from 0 to 7.5 m below ground was predominantly quartz (>80%), underlain by a more heterogeneous section from 7.5 to 11.6 m below ground with an average composition of quartz (60%), calcite (30%), microcline (5%) and anorthite (5%) from XRD analysis. Further mineral phase characterization using AutoGeoSEM (Robinson et al., 2000) on a sand sample revealed aluminum and iron oxides. The aluminum oxides are silicate weathering products (clay minerals) deposited as coatings on sand grains, which are colored yellow by the presence of hydrated iron oxides (Bastian, 1996). The depth to the water table below the galleries varied seasonally between 10 and 11 m. The regional groundwater flow direction was from east to west. For experimental purposes, an artificial hydraulic gradient was produced by continual pumping from a well located 50 m west of the infiltration galleries. The natural hydraulic gradient coincided with the imposed gradient. A series of monitoring wells were installed with slotted intervals positioned at different depths below ground as shown in Fig. 1. The migration rate of the infiltrated recycled water through the vadose zone was previously determined based on a bromide tracer experiment (Bekele et al., 2009). From this data, a minimum travel time of 3.7 days was estimated through the unsaturated zone. The travel time to BH1 after passage through the vadose zone was estimated to be an additional 0.5 day based on a comparison of the electrical conductivity of the recycled water relative to groundwater from BH1, indicating that the substantial time for processes to occur was in the vadose zone compared to the aquifer.

2.3.

Description of monitoring wells

To assess the potential infiltration benefits, improvements in quality of the recycled water immediately down-gradient from the infiltration galleries were compared to the quality

5766

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 6 4 e5 7 7 2

general accordance with ASTM D 4448-01 Standard Guide for Sampling Ground-Water Monitoring Wells (2001). Wastewater samples were collected from the discharge chamber in the infiltration galleries. Water samples were analyzed for major ions, nutrients, trace metals, dissolved and total organic carbon, trace organics and faecal indicator microorganisms and pathogens. Further details regarding water sampling, analytical procedures and detection limits for the water chemistry are given in Bekele et al. (2009). Faecal indicator microbes E. coli and enterococci were detected using membrane filtration and selective media while the presence of enteric pathogens were determined from a concentrate of 40 L using PCR as described in Toze and Bekele (2009). Sampling water from the galleries and the wells for different chemicals and microbes was conducted during the first 25 months of the MAR trial at different time intervals, but generally on a weekly, fortnightly or monthly basis. Physical water parameters measurements were taken before water sampling occurred. These included pH, dissolved oxygen, electrical conductivity, temperature, oxidationreduction potential and turbidity measurements using a Troll 9000 multisensory meter housed within a flow cell.

2.5. Fig. 1 e The Floreat MAR site showing the positions of the two galleries, the monitoring wells (BH1, BH2 and BH5) and the recovery well in map view (a), and relative to the maximum height of the water table in cross-section (b).

of the infiltrated recycled water. Monitoring wells close to the infiltration galleries (BH1, BH2 and BH5) were selected to evaluate the benefits to water quality of infiltrating recycled water through the vadose zone. Three background groundwater monitoring wells located hydraulically up-gradient of the infiltration galleries (BGRND1, BGRND2 and BGRND3) were also monitored to provide an assessment of changes in groundwater quality due to infiltration. Monitoring well details are given in Table 1.

2.4.

Collection and analysis of water samples

Details of the groundwater sampling procedure are given in Bekele et al. (2009). Groundwater sampling procedures were in

Criteria to assess changes in recycled water quality

To assess the water quality changes as a result of the migration through the vadose zone, the quality of water samples collected from the shallow monitoring wells immediately down-gradient from the infiltration galleries were compared to the quality of the recycled water prior to infiltration. Sampling of the vadose zone water directly below the infiltration galleries and immediately above the water table was not undertaken. Thus, only gross changes after passage over the entire vadose zone could be quantified as changes during infiltration could not be determined. The significance of the effects of migration through the subsurface on recycled water quality was assessed using Student’s two-tailed t-test (unpaired) applied to datasets of measured nutrients, inorganic compounds, pharmaceutical and microbial pathogens, comparing the mean concentrations in recycled water prior to infiltration relative to the mean concentrations in recharged water sampled from BH1. For the Student’s t-test, the null hypothesis was that water quality was unchanged despite passage through the vadose zone. Microsoft Excel was used to

Table 1 e Monitoring well details. Well BH1 BH2 BH5b BGRND1 BGRND2 BGRND3

Ground elevation (m AHDa)

Distance and direction from west gallery

Total depth below ground (m)

Screened depth interval below the maximum height of the water table (m)

12.97 13.01 12.86 15.94 12.00 12.00

2.3 m west 2.5 m west 8.8 m east 185 m northeast 75 m east 75 m east

12.01 12.04 10.81 15.00 12.23 20.11

0.0e2.0 0.0e2.0 0.0e1.0 0.0e2.05 2.23e3.23 10.11e11.11

a Australian Height Datum. b BH5 is located up-gradient relative to the groundwater flow direction and a distance of 4 m east of the east gallery.

5767

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 6 4 e5 7 7 2

calculate the probability (P) and the null hypothesis was rejected if P < 0.05.

3.

Results and discussion

3.1. Characterization of recycled water and ambient groundwater To understand what water quality changes were potentially achievable during passage of the recycled water through the vadose zone, it was also necessary to characterise and compare the ambient groundwater. Both the ambient groundwater and the recycled water were consistently aerobic. The average water temperature, pH, electrical conductivity, dissolved oxygen concentrations and sulfate concentrations of the ambient groundwater were similar to that of the recycled water (Table 2). The concentration of total dissolved solids was highly variable in the recycled water and the range of variation overlapped with that of the ambient groundwater. The major ions defining the water types were CaeNaeCleHCO3 for ambient groundwater and NaeCleHCO3 for recycled water. The recycled water and ambient groundwater had low concentrations of several inorganic species, namely arsenic, boron, cadmium, cobalt, chromium, copper, mercury, manganese, molybdenum, nickel, lead, selenium, uranium, vanadium and zinc. The impact of infiltration on these inorganic chemicals was not investigated as their measured concentrations were either very low or were below detection limits.

3.2. MAR impacts on groundwater quality near the galleries Site groundwater monitoring was undertaken to confirm that the shallow groundwater collected closest to the infiltration galleries did represent infiltrated recycled water and not ambient groundwater. Average concentrations of potassium, chloride and sodium in the recycled water were similar to the average concentrations from water samples collected from the monitoring well BH1 (Table 3). The recycled water had an average potassium concentration that was greater than fourfold more enriched in potassium compared to the ambient groundwater and therefore was an effective tracer for the

Table 2 e Mean values of water quality parameters of the recharge water that were similar to ambient groundwater with standard deviations in parentheses. Parameter (Unit of Measure)

Recycled water

Water temperature ( C) pH Electrical Conductivity (mS m1) Eh (mV-SHE) Dissolved Oxygen (mg L1) Sulfate as S (mg L1) Total Dissolved Solids (mg L1)

24 7.33 143 385 2.15 64.1 755

(4) (1.11) (53) (184) (1.82) (8) (179)

Ambient Groundwater 22 (1) 7.04 (0.88) 124 (63) 321 (189) 4.02 (2.56) 64.5 (18) 644 (25)

infiltrated recycled water as demonstrated by others in previous studies (Rueedi et al., 2009; Wolf et al., 2004). The recharged water quality sampled from monitoring wells BH1 and BH2 was indistinguishable from the recycled water prior to infiltration, but substantially different from ambient groundwater on the basis of potassium and chloride concentrations (Fig. 2). Water samples collected from the monitoring well BH5, 4 m up-gradient from the galleries, showed a greater proportion of ambient groundwater mixed with recycled water compared to the water samples from the monitoring wells down-gradient of the galleries. A theoretical line of mixing between the average concentrations of chloride and potassium in ambient groundwater and that of recycled water was generated (Fig. 2). During the first three months of infiltration, the composition of groundwater from BH5 was predominantly ambient groundwater (80%); thereafter, the composition of groundwater from BH5 had a higher proportion of recycled water relative to ambient groundwater. Since water sampled from well BH1 did not show temporal transition and it was frequently monitored, water collected from BH1 was used for evaluating influent water quality improvements as a result of its migration through the vadose zone.

3.3.

Water quality changes during infiltration

3.3.1.

Inorganic chemicals

The recycled water prior to infiltration had higher average concentrations of most species of nitrogen, phosphorous, and fluoride compared to ambient groundwater (Table 3). Recharged water sampled down-gradient from the galleries from BH1 had average concentrations for these chemicals that were generally between those for ambient groundwater and recycled water or within one standard deviation of these two water sources. To assess improvements due to infiltration, total nitrogen concentrations and a relatively non-reactive analyte (i.e.

Table 3 e Mean concentrations (mg/L) of selected chemicals in recycled water and groundwater sampled down-gradient from the galleries from BH1 compared to ambient groundwater. Standard deviations are provided in parentheses. Analtye

Recycled water

Chloride 245 (62) Potassium 22.9 (3.5) Sodium 194 (44) Ammonia as Nitrogen 0.64 (0.85) Nitrate as Nitrogen 2.16 (1.41) Total Kjeldahl Nitrogen 1.86 (0.98) Total Nitrogen 4.27 (1.9) Soluble reactive 6.31 (3.32) phosphorus as P Total organic carbon 9.98 (3.8) Fluoride 0.73 (0.2) Iron 0.14 (0.20) Calcium 28.6 (9.49) Aluminum 0.018 (0.011)

Ambient groundwater

BH1

162 (27) 4.96 (0.83) 92.6 (15) <0.01 0.16 (0.26) 0.066 (0.030) 0.302 (0.36) 0.0126 (0.007)

248 (50) 21.7 (4.1) 194 (42) 0.037 (0.046) 3.90 (1.61) 0.95 (0.51) 4.93 (1.8) 2.19 (1.53)

2.66 (2.72) 0.15 (0.1) 0.44 (0.72) 98.8 (23.5) 0.22 (0.48)

6.42 (3.57) 0.25 (0.05) 0.053 (0.054) 60.4 (7.28) 0.045 (0.048)

5768

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 6 4 e5 7 7 2

30

Potassium (mg/L)

25 20 15 10

Ambient groundwater BH1 BH2 BH5, samples during first 3 months BH5, samples after first 3 months Recycled water

5 0 0

100

200

300 Chloride (mg/L)

400

500

Fig. 2 e Cross plot showing concentrations of chloride and potassium in ambient groundwater sampled from the background monitoring wells, recycled water and in water recovered from the monitoring wells (BH1, BH2 and BH5). A mixing line is shown between the average composition of recycled water and ambient groundwater.

potassium) were plotted (Fig. 3). No improvement in total nitrogen was demonstrated as shown by the cluster of data for BH1 in Fig. 3, which overlaps with that of recycled water and indicates a higher range of total nitrogen in contrast to ambient groundwater (<1.2 mg/L). Migration of recycled water through the vadose zone and aquifer to reach BH1 did not improve total nitrogen concentrations (P ¼ 0.24, Student’s ttest). Although passage through the vadose zone did not significantly alter the concentration of total nitrogen in the recycled water, there were changes to different nitrogen species (Table 3). The mean concentration of total Kjeldahl nitrogen in the recharged water sampled from BH1 was 49% lower relative to the mean concentrations in recycled water prior to infiltration by (P < 0.01, Student’s t-test), whereas the mean concentration of ammonia was 94% lower (P < 0 0.01). In comparison, the mean concentration of nitrate was 77% higher (P < 0.001) in the recharged water sampled from BH1 relative to the mean concentrations in recycled water prior to infiltration. This data suggests that nitrification was occurring through the vadose

zone with nitrate production from the ammonium and total Kjeldahl nitrogen, and provides a plausible explanation for the lack of change in total nitrogen concentrations. As conditions in the vadose zone were not conducive for reducing nitrate concentrations, potential treatment options to consider include manipulation of the oxygen content to drive denitrification and the addition of carbon amendments to promote nitrate removal within the aquifer (Patterson et al., 2004, 2011). Migration of the recycled water through the vadose zone and 2.3 m of aquifer aided removal of phosphorous (P < 0.0001, Student’s t-test). Further investigation revealed a timedependence for phosphorous removal. During the first 124 days of infiltration, phosphorous in recharged water sampled from the monitoring well BH1 down-gradient from the galleries remained low, similar to ambient groundwater (<0.02 mg/L), whereas the average phosphorous in the influent recycled water during this initial period was 10 mg/L (Fig. 4). After 271 days of infiltration, the average phosphorous concentration in the influent recycled water was 4.7 mg/L and more variable than previously (standard deviation of 2.1 mg/ L). In comparison, the concentration of phosphorous in recharged water collected from BH1 reached a maximum on day 271 and then stabilized with a mean concentration of 3.2 mg/L (standard deviation of 0.68 mg/L). Comparison of average phosphorous concentrations in BH1 and the influent recycled water reveals a reduction of 31% after 271 days of infiltration. Adsorption and precipitation are reported to be the main causes of phosphorous retention in calcareous sands and soils (von Wandruszka, 2006; Whelan, 1988; Whelan and Barrow, 1984). However, calcareous sands have a limited capacity to store phosphorous, thus caution may be warranted in anticipating consistent levels of phosphorous removal, particularly at sites recharging higher volumes of recycled water and after prolonged recharge. Desorption and remobilization of phosphorous should be considered in dealing with long-term MAR systems. Previous studies have demonstrated the transient nature of phosphorous retention in carbonate soils receiving wastewater. Examples include the release of sorbed phosphate caused by dissolution of carbonate, e.g. from the acidification of leachate during nitrification of ammonium

10 14

Ambient groundwater BH1 Recycled water

8 7

Phosphorous (mg/L)

Total Nitrogen (mg/L)

9

6 5 4 3 2 1

12 10 8 6 4 2

0 0

5

10

15 20 Potassium (mg/L)

25

30

35

Ambient groundwater BH1 Recycled water

Fig. 3 e Cross plot showing concentrations of total nitrogen and potassium in recycled water, ambient groundwater and recharged water sampled down-gradient from the galleries from monitoring well BH1.

0 0

100

200

300

400

500

600

700

Time since start of infiltration (days)

Fig. 4 e Temporal changes in concentrations of phosphorus in recycled water, ambient groundwater and water sampled down-gradient from the galleries from monitoring well BH1.

5769

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 6 4 e5 7 7 2

3.3.2.

Organic carbon

Migration of recycled water through the subsurface to reach well BH1 reduced the total organic carbon concentrations (P < 0.01, Student’s t-test). Comparison of average TOC concentrations in BH1 and the influent recycled water reveals a reduction of 51% after 400 days of infiltration (Fig. 5). Probable mechanisms for the removal of organic carbon included filtration, adsorption and biodegradation (Drewes et al., 2003; Fox et al., 2001; Vanderzalm et al., 2010).

30 Total Organic Carbon (mg/L)

(Whelan, 1988), and soil saturated with phosphate becoming a source of phosphate if concentrations in the wastewater recharge decrease (Lin and Banin, 2005). Migration through the subsurface reduced fluoride concentrations in the infiltrated water (P < 0.0001, Student’s ttest). The comparison of average fluoride concentrations in BH1 and the influent recycled water reveals a reduction of 66% (Table 3). Similar to the reduction in phosphorous, reductions in fluoride concentrations were likely due to adsorption and precipitation in calcite-bearing layers (Turner et al., 2005). Adsorption of fluorite was more likely as both the recycled water and ambient groundwater were under-saturated with respect to fluorite. Concentrations of soluble iron in the ambient groundwater sampled from the three background wells were quite variable (mean of 0.44 mg/L; standard deviation of 0.72 mg/L) making it difficult to interpret changes to recycled water quality from passage through the subsurface. Infiltration of the recycled water sustained aerobic conditions favorable for the removal of soluble iron near the gallery. This may explain the nearly three-fold difference between the average concentration of soluble iron in water sampled from well BH1 and recycled water prior to infiltration (P ¼ 0.013, Student’s t-test). Comparison of average iron concentrations in BH1 and the influent recycled water reveals a reduction of 62%. While the pH of the recycled water varied over time during the trial, the pH of the groundwater at BH1 was consistently higher than the recycled water. During infiltration, the average pH of water sampled from well BH1 was 7.61 (standard deviation of 0.80), whereas the average pH of recycled water prior to infiltration was 7.33 (standard deviation of 1.11; Table 2). The increase in pH of the recycled water as it infiltrated through the calcareous sand and limestone was likely due to calcite dissolution in the vadose zone. Calcium and aluminium concentrations in water collected from well BH1 were higher relative to their concentrations in the recycled water; however ambient groundwater had the highest concentrations of these ions (Table 3). The average calcium concentration in the water sampled from BH1 was twice the average concentration in the recycled water prior to infiltration. The recycled water was under-saturated with respect to calcite whereas ambient groundwater was in equilibrium with calcite; hence conditions were favourable for calcite dissolution thereby increasing calcium in water sampled down-gradient from the galleries. The average concentration of aluminium in groundwater sampled from well BH1 was also twice the average aluminium concentration in the recycled water. It is likely that the mobility of aluminium is linked to dissolution of calcium carbonate coated with aluminium oxides.

Ambient groundwater BH1 Recycled water

25 20 15 10 5 0 0

100

200

300

400

500

600

700

Time since start of infiltration (days)

Fig. 5 e Temporal changes in concentrations of total organic carbon in recycled water, ambient groundwater and down-gradient from the galleries in water sampled from monitoring well BH1.

3.3.3.

Pharmaceuticals

All of the pharmaceuticals tested revealed concentrations in ambient groundwater that were below detection limits (i.e. <0.05 mg/L for carbamazepine; <0.01 mg/L for diazepam, oxazepam, phenytoin and temazepam). In comparison, diazepam and phenytoin were the only pharmaceuticals below the detection limit in the recycled water (Table 4). Migration of recycled water through the vadose zone produced reductions in oxazepam and temazepam (P < 0.01, Student’s t-test); however, concentrations of these pharmaceuticals in water collected from BH1 were highly variable. Additional treatment may be required to produce consistent levels of concentration reduction for oxazepam and temazepam, depending on the end use of the extracted water and the potential for large volumes to be consumed intentionally or unintentionally by the community. Sorption and biodegradation are common mechanism for the removal of pharmaceuticals under aerobic conditions in the vadose zone (Amy and Drewes, 2007; Conn et al., 2010; Scheytt et al., 2006; Tiehm et al., 2011), but further investigation of removal mechanisms for the pharmaceuticals was not undertaken in this study. An unexpected observation was that concentrations of carbamazepine in the water collected from well BH1 were consistently higher (80% higher on average; standard deviation of 37%) than carbamazepine detected in the recycled water. The higher carbamazepine concentrations from BH1 relative to recycled water may be due to changes in concentration in the recycled water over time, which varied between 0.13 and 0.33 mg/L (average of 0.21 mg/L, Table 4). In column experiments using aquifer sediment and recycled water from the MAR project and under fully saturated, aerobic conditions, no sorption or degradation were observed for carbamazepine and oxazepam (Patterson et al., 2009). The persistence of carbamazepine in the environment has been demonstrated previously and it has been suggested that it could be used as a potential anthropogenic marker in aquatic environments (Clara et al., 2004). These results may be important for wastewater reuse near wetlands or ecologically-sensitive areas where endocrinedisrupting chemicals could impact negatively on aquatic species in the receiving environment (Lister et al., 2009; Saaristo et al., 2010; Sulleabhain et al., 2009).

5770

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 6 4 e5 7 7 2

Table 4 e Mean concentrations (mg/L) of pharmaceuticals in recycled water, water sampled from BH1, and ambient groundwater. Standard deviations are given in parentheses. Analtye

Carbamazepine Diazepam Oxazepam Phenytoin Temazepam

3.3.4.

Recycled water (n ¼ 25)

Ambient groundwater (n ¼ 33)

BH1 (n ¼ 16)

0.21 (0.067) <0.1 0.31 (0.090) <0.1 0.31 (0.084)

<0.05 <0.1 <0.1 <0.1 <0.1

0.38 (0.089) <0.1 0.21 (0.14) <0.1 0.17 (0.12)

Faecal indicator microorganisms and enteric pathogens

It is worth noting that thermotolerant coliforms and enterococci were detected in ambient groundwater from wells upgradient from the recharge site in the sheep paddock. With regard to thermotolerant coliforms, 13% of ambient groundwater samples tested (n ¼ 38) and 19% of water samples from BH1 (n ¼ 27) had numbers greater than 10 cfu 100 mL1. It is hypothesized that surface contamination was most likely due to excreta from the grazing sheep at the recharge site as the source of these microbes and not the treated wastewater being recharged to the aquifer. Despite the additional source of faecal indicator microorganisms from sheep excreta, migration of recycled water from the galleries through the vadose zone to well BH1 produced a reduction in thermotolerant coliforms (P < 0.0001, Student’s t-test). Thermotolerant coliforms were routinely detected in the recycled water with 80% of the samples tested (n ¼ 46), exceeding 10 colony forming units (cfu) 100 mL1. High numbers of enterococci were present in the recycled water with 79% of the water samples tested (n ¼ 48) exceeding 100 cfu 100 mL1, whereas 19% of ambient water samples (n ¼ 42) exceeded 100 cfu 100 mL1. Despite the high counts in the recycled water, there was some reduction in enterococci numbers comparing recycled water and water from BH1 with only 28% of samples from the BH1 (n ¼ 29) exceeding 100 cfu 100 mL1 (P < 0.001, Student’s t-test). Reductions in microbial pathogen numbers in the recycled water were shown by fewer detections of adenovirus and Fþ bacteriophage after migration through the subsurface. Adenovirus was detected in 68% of the samples of the recycled water (n ¼ 19) prior to infiltration but in only 6% of the samples from BH1 (n ¼ 18) and in none of the wells further down gradient. Adenovirus was not detected in any of the ambient groundwater samples. Fþ bacteriophage, commonly used as a surrogate enteric virus, were detected in 94% of recycled water samples tested (n ¼ 36). In comparison only 4% of the water samples from BH1 (n ¼ 24), and 6% of the ambient groundwater samples (n ¼ 33) were positive for this bacteriophage. Reductions in microbial pathogens in treated wastewater recharged to the Tamala Limestone aquifer over time were demonstrated using survival experiments conducted with selected faecal indicators (Cryptosporidium, adenovirus, rotavirus, coxsackievirus, MS2, E. coli, S. enterica and E. faecalis) in in-situ diffusion chambers (Toze et al., 2010). The reductions in

microbial pathogens were attributed to a combination of physical removal processes during filtration and the activity of indigenous groundwater microorganisms (Gordon et al., 2002; Toze and Hanna, 2002). The aquifer has an active treatment capacity to remove pathogens (Toze and Bekele, 2009), but a longer period of aquifer residence may be needed to allow for more inactivation of microbial pathogens (Toze et al., 2010).

4.

Conclusions

 Water quality benefits were achieved by infiltrating secondary treated wastewater through weathered calcareous sands and limestone of the vadose zone in an urban area using infiltration galleries.  Reductions in the average concentrations of several constituents in the recycled water before and after MAR were identified as follows: 30% for phosphorous, 66% for fluoride, 62% for iron and 51% for total organic carbon. Phosphorous removal declined over time, implying the “maximum” P adsorption capacity was reached. Reductions in the average concentrations of oxazepam and temazepam and the numbers of thermotolerant coliforms and enterococci in the infiltrated water were also identified, but not as easily quantified.  The removal rates determined for chemical and biological species are particular to the MAR conditions in this study. The impact of a thicker unsaturated zone, anoxic conditions and different flow rates and contact times were not investigated and would give rise to different removal rates.  The aerobic conditions present in the vadose zone were not conducive for denitrification to reduce nitrate concentrations in the water recharged to the aquifer, revealing that secondary treated wastewater recharged via infiltration cannot rely entirely upon processes in the vadose zone for nitrate removal.  Geochemical conditions were favorable for calcite dissolution, suggesting that porosity in the calcareous vadose zone may increase over time and could cause faster breakthrough of contaminants to the water table.  Future MAR in this aquifer type should consider the effects of prolonged recharge, higher rates of recharge and concentration-dependency of adsorption and precipitation reactions (e.g. controlling phosphorous mobility) to determine the long-term sustainability of recharging recycled water to a calcareous aquifer in urban environments.

Acknowledgements This research was funded by the Western Australian Government through the Water Foundation (gs1), the Water Corporation of WA and the CSIRO Water for a Healthy Country Flagship Program. Chemical analyses were conducted by the Chemistry Centre of Western Australia. The authors would like to thank Mr. Mark Shackleton and Mr. Sebit Gama from the CSIRO for their assistance with water sampling and microbial analyses. Drs Joanne

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 6 4 e5 7 7 2

Vanderzalm and Grant Douglas (CSIRO) are gratefully acknowledged for discussions on the geochemistry. The content of this paper benefited from internal reviews by Drs Declan Page and Warish Ahmed at the Commonwealth Scientific and Industrial Research Organization.

references

Amy, G., Drewes, J., 2007. Soil aquifer treatment (SAT) as a natural and sustainable wastewater reclamation/reuse technology: fate of wastewater effluent organic matter (EfOM) and trace organic compounds. Environmental Monitoring and Assessment 129 (1e3), 19e26. ASTM, 2001. Standard Guide for Sampling Ground-water Monitoring Wells D 4448-01. edited. American Society for Testing and Materials (ASTM). Bastian, L., 1996. Residual soil mineralogy and dune subdivision, Swan coastal plain, Western Australia. Australian Journal of Earth Sciences 43 (1), 31e44. Bekele, E., Toze, S., Patterson, B., Devine, B., Higginson, S., Fegg, W., Vanderzalm, J., 2009. Design and operation of infiltration galleries and water quality guidelines - Chapter 1. In: Toze, S., Bekele, E. (Eds.), Determining the Requirements for Managed Aquifer Recharge in Western Australia (A Report to the Water Foundation). Clara, M., Strenn, B., Kreuzinger, N., 2004. Carbamazepine as a possible anthropogenic marker in the aquatic environment: investigations on the behaviour of Carbamazepine in wastewater treatment and during groundwater infiltration. Water Research 38 (4), 947e954. Conn, K.E., Siegrist, R.L., Barber, L.B., Meyer, M.T., 2010. Fate of trace organic compounds during vadose zone soil treatment in an onsite wastewater system. Environmental Toxicology and Chemistry 29 (2), 285e293. Davidson, W.A., 1995. Hydrogeology and groundwater resources of the Perth region, Western Australia, Bulletin. Western Australia Geological Survey 142. Dillon, P., 2005. Future management of aquifer recharge. Hydrogeology Journal 13 (1), 313e316. Drewes, J.E., Reinhard, M., Fox, P., 2003. Comparing microfiltrationreverse osmosis and soil-aquifer treatment for indirect potable reuse of water. Water Research 37 (15), 3612e3621. Fox, P., Narayanaswamy, K., Genz, A., Drewes, J., 2001. Water quality transformations during soil aquifer treatment at the Mesa Northwest water reclamation plant, USA. Water Science and Technology 43 (10), 343e350. Gordon, C., Wall, K., Toze, S., O’Hara, G., 2002. Influence of Conditions on the Survival of Enteric Viruses and Indicator Organisms in Groundwater. In: Paper Presented at Proceedings of the 4th International Symposium on Artificial Recharge of Groundwater Isar-4-Management of Aquifer Recharge for Sustainability. Balkema Publishers, Adelaide SA. Lin, C.Y., Banin, A., 2005. Effect of long-term effluent recharge on phosphate sorption by soils in a wastewater reclamation plant. Water Air and Soil Pollution 164 (1e4), 257e273. Lister, A., Regan, C., Van Zwol, J., Van Der Kraak, G., 2009. Inhibition of egg production in zebrafish by fluoxetine and municipal effluents: a mechanistic evaluation. Aquatic Toxicology 95 (4), 320e329. Montgomery-Brown, J., Drewes, J., Fox, P., Reinhard, M., 2003. Behavior of alkylphenol polyethoxylate metabolites during soil aquifer treatment. Water Research 37 (15), 3672e3681. Patterson, B., Pearce, J., Spadek, T., Linge, K., Busetti, F., Shackleton, M., Furness, A., Blair, P., Heitz, A., 2009. Fate of trace organics using laboratory column experiments - chapter

5771

2. In: Toze, S., Bekele, E. (Eds.), Determining the Requirements for Managed Aquifer Recharge in Western Australia (A Report to the Water Foundation). Patterson, B.M., Grassi, M.E., Robertson, B.S., Davis, G.B., Smith, A. J., McKinley, A.J., 2004. Use of polymer mats in series for sequential reactive barrier remediation of ammoniumcontaminated groundwater: field evaluation. Environmental Science & Technology 38 (24), 6846e6854. Patterson, B.M., Shackleton, M., Furness, A.J., Bekele, E., Pearce, J., Linge, K.L., Busetti, F., Spadek, T., Toze, S., 2011. Behaviour and fate of nine recycled water trace organics during managed aquifer recharge in an aerobic aquifer. Journal of Contaminant Hydrology 122 (1e4), 53e62. Pyne, R.D.G., 2006. Groundwater Recharge and Wells: a Guide to Aquifer Storage Recovery, second ed. CRC Press, Boca Raton, Florida, 401 pp. Quanrud, D., Hafer, J., Karpiscak, M., Zhang, H., Lansey, K., Arnold, R., 2003. Fate of organics during soil-aquifer treatment: sustainability of removals in the field. Water Research 37 (14), 3401e3411. Robinson, B.W., Hitchen, G.J., Verrall, M.R. 2000, The AutoGeoSEM: A programmable fully-automatic SEM for rapid grain-counting and heavy mineral characterisation in exploration, paper presented at International Mineralogical Association Commission on Ore mineralogy, Proceeding of the Modern Approaches to Ore and Environmental mineralogy, Short Course, p. 71e74. Rueedi, J., Cronin, A.A., Morris, B.L., 2009. Estimation of sewer leakage to urban groundwater using depth-specific hydrochemistry. Water and Environment Journal 23 (2), 134e144. Saaristo, M., Craft, J.A., Lehtonen, K.K., Lindstrom, K., 2010. An endocrine disrupting chemical changes courtship and parental care in the sand goby. Aquatic Toxicology 97 (4), 285e292. Scatena, M.C., Williamson, D.R. 1999, A potential role for artificial recharge within the Perth region: a pre-feasibility study. Center for Groundwater Studies report No. 84. Schafer, A., Ustohal, P., Harms, H., Stauffer, F., Dracos, T., Zehnder, A., 1998. Transport of bacteria in unsaturated porous media. Journal of Contaminant Hydrology 33 (1e2), 149e169. Scheytt, T.J., Mersmann, P., Heberer, T., 2006. Mobility of pharmaceuticals carbamazepine, diclofenac, ibuprofen, and propyphenazone in miscible-displacement experiments. Journal of Contaminant Hydrology 83 (1e2), 53e69. Smith, A.J., Pollock, D.W., 2010. Artificial Recharge Potential of the Perth Region Superficial Aquifer: Lake Preston to Moore River. CSIRO (Water for a Healthy Country National Research Flagship Report). Sulleabhain, C.O., Gill, L.W., Misstear, B.D.R., Johnston, P.M., 2009. Fate of endocrine-disrupting chemicals in percolating domestic wastewater effluent. Water and Environment Journal 23 (2), 110e118. Tapsell, P., Newsome, D., Bastian, L., 2003. Origin of yellow sand from Tamala limestone on the Swan Coastal Plain, Western Australia. Australian Journal of Earth Sciences 50 (3), 331e342. Tiehm, A., Schmidt, N., Stieber, M., Sacher, F., Wolf, L., Hoetzl, H., 2011. Biodegradation of pharmaceutical compounds and their occurrence in the Jordan Valley. Water Resources Management 25 (4), 1195e1203. Toze, S., Bekele, E., 2009. Determining Requirements for Managed Aquifer Recharge in Western Australia. A Report to the Water Foundation. CSIRO, Canberra, AU. http://www.csiro.au/files/ files/py7j.pdf. Toze, S., Bekele, E., Page, D., Sidhu, J., Shackleton, M., 2010. Use of static Quantitative Microbial Risk Assessment to determine pathogen risks in an unconfined carbonate aquifer used for Managed Aquifer Recharge. Water Research 44 (4), 1038e1049. Toze, S., Hanna, J., 2002. The Survival Potential of Enteric Pathogens in a Reclaimed Water ASR Project. In: Paper

5772

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 6 4 e5 7 7 2

Presented at Proceedings of the 4th International Symposium on Artificial Recharge of Groundwater ISAR-4-Management of Aquifer Recharge for Sustainability. Balkema Publishers, Adelaide, SA, pp. 139e142. Toze, S., Hanna, J., Smith, T., Edmonds, L., McCrow, A., 2004. Determination of water quality improvements due to the artificial recharge of treated effluent. In: Steenvoorden, J., Endreny, T. (Eds.), International Symposium on Wastewater Re-Use and Groundwater Quality. Int Assoc Hydrological Sciences, Wallingford, UK, pp. 53e60. Turner, B., Binning, P., Stipp, S., 2005. Fluoride removal by calcite: evidence for fluorite precipitation and surface adsorption. Environmental Science & Technology 39 (24), 9561e9568. Vanderzalm, J., Page, D., Barry, K., Dillon, P., 2010. A comparison of the geochemical response to different managed aquifer recharge operations for injection of urban stormwater in a carbonate aquifer. Applied Geochemistry 25 (9), 1350e1360.

von Wandruszka, R., 2006. Phosphorus retention in calcareous soils and the effect of organic matter on its mobility. Geochemical Transactions 7 (6). doi:10.1186/14674866-7-6. Whelan, B.R., 1988. Disposal of septic tank effluent in calcareous sand. Journal of Environmental Quality 17 (2), 272e277. Whelan, B.R., Barrow, N.J., 1984. The movement of septic-tank effluent through sandy soils near Perth .2. Movement of phosphorus. Australian Journal of Soil Research 22 (3), 293e302. Wolf, L., Held, I., Eiswirth, M., Hotzl, H., 2004. Impact of leaky sewers on groundwater quality. Acta Hydrochimica Et Hydrobiologica 32 (4e5), 361e373. Zhang, J., Huang, X., Liu, C., Shi, H., Hu, H., 2005. Nitrogen removal enhanced by intermittent operation in a subsurface wastewater infiltration system. Ecological Engineering 25 (4), 419e428.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Removal of mercury (II) by dithiocarbamate surface functionalized magnetite particles: Application to synthetic and natural spiked waters P. Figueira a, C.B. Lopes b,*, A.L. Daniel-da-Silva a, E. Pereira b, A.C. Duarte b, T. Trindade a a b

CICECO & Department of Chemistry, University of Aveiro, Campus de Santiago, 3810-193 Aveiro, Portugal CESAM & Department of Chemistry, University of Aveiro, Campus de Santiago, 3810-193 Aveiro, Portugal

article info

abstract

Article history:

In order to take advantage of the high affinity between mercury and sulphur, magnetite

Received 24 March 2011

(Fe3O4) particles functionalized with dithiocarbamate groups (CS
2 ), were synthesized to be

Received in revised form

used as a new type of sorbent to remove Hg (II) from synthetic and natural spiked waters.

22 July 2011

The effectiveness of this type of sorbent was studied, and its potential as cleanup agent for

Accepted 27 August 2011

contaminated waters was assessed.

Available online 3 September 2011

Batch stirred tank experiments were carried out by contacting a volume of solution with known amounts of functionalized Fe3O4 particles, in order to study the effect of sorbent

Keywords:

dose, salinity, and the kinetics and the equilibrium of this unit operation. A complete Hg (II)

Mercury

removal (ca. 99.8%) was attained with 6 mg/L of magnetic particles for an initial metal

Magnetite particles

concentration of 50 mg/L. It was confirmed that highly complex matrices, such as seawater

Dithiocarbamate functionalization

(ca. 99%) and river water (ca. 97%), do not affect the removal capacity of the functionalized

Isotherms

magnetic particles. Concerning isotherms, no significant differences were observed

Water remediation

between two- and three-parameter models (P ¼ 0.05%); however, Sips isotherm provided the lowest values of SS and Sx/y, predicting a maximum sorption capacity of 206 mg/g, in the range of experimental conditions under study. The solid loadings measured in this essay surmount the majority of the values found in literature for other type of sorbents. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

The increasing awareness for the effects of pollution, has led to strict environmental regulations, such as the creation of a list of 33 priority hazardous substances by European Parliament and the Council of the European Union (EU 2008), and the instigation of the Member States to implement the necessary actions aiming at a progressive reduction of pollutants, and ceasing or phasing out emissions and discharges of priority hazardous substances, as considered in the Framework Directive (EU 2000).

Mercury and its compounds are one of the most dangerous contaminants in the environment, threatening the human health and natural ecosystems. They are included in the list of priority hazardous substances and consequently, the removal of Hg and its compounds, particularly, from aquatic systems is a major goal of wastewater treatment and cleanup technologies. Conventional techniques for Hg removal from aqueous solutions include sulphate or hydrazine precipitation, ionexchange, liquideliquid extraction, adsorption and solid phase extraction via activated carbon adsorption (Starvin and Rao, 2004).

* Corresponding author. Tel.: þ351 234 370 721; fax: þ351 234 370 084. E-mail address: [email protected] (C.B. Lopes). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.057

5774

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4

Nomenclature C V M q k1 k2 kL qm kF n Nt a m kRP

Concentration of Hg (II) in bulk solution, mg/L Volume of solution, L Dry weight of Fe3O4/SiO2/NH/CS
2 particles, mg Amount of Hg (II) in the particles, mg/g First-order rate constant, h
1 Second-order rate constant, g/mg h Langmuir constant, L/g Maximum loading of Fe3O4/SiO2/NH/CS
2 particles, mg/g Freundlich parameter, (mg1
1/n$L1/n/g) Freundlich parameter Total number of binding sites, mg/g Sips constant Heterogeneous index Redlich-Peterson constant, L/g

The high toxicity of Hg to organisms is generally attributed to the blockage of the enzyme binding sites and interference in protein synthesis (Starvin and Rao, 2004), due to the high affinity of Hg to bind with any molecule that has sulphur or a sulphurehydrogen combination in its structure. A promising strategy to achieve high-performance materials able to capture Hg (II) from aqueous solutions is to take advantage of its high affinity to sulphur, developing new materials with sulphur-containing functional groups such as diverse thiolates. This approach has been investigated by Antochshuk et al. (2003) and materials with high surface area have been used as an insoluble matrix for the attachment of sulphurcontaining groups (Antochshuk et al., 2003). Several materials such as mesoporous silica (Antochshuk et al., 2003; Venkatesan et al., 2003; Mattigod et al., 1999; Mercier and Pinnavaia, 1998; Feng et al., 1997; Mureseanu et al., 2010), silica gel (Venkatesan et al., 2002), activated carbon (Starvin and Rao, 2004) and organoceramic composites (Nam et al., 2003) have been used as matrices for the attachment of sulphur-containing groups, such as thiol (Mattigod et al., 1999; Mercier and Pinnavaia, 1998; Nam et al., 2003), dithiocarbamate (Venkatesan et al., 2003, 2002), 1-(2-thiazolylazo)-2naphthol (Starvin and Rao, 2004), mercaptopropylsilane (Feng et al., 1997), 1-furoyl thiourea urea (Mureseanu et al., 2010) and benzoythiourea (Antochshuk et al., 2003). Those materials have surface coverage of ligands typically between 2.5  10
4 and 3.7  10
3 mol/g, and some of them exhibit very high Hg (II) loading capacities (Antochshuk et al., 2003). We have been interested on magnetic particles, in particular iron oxides (e.g. Fe3O4), to develop new materials for environmental and bio-applications (Daniel-da-Silva et al., 2007). Iron oxides such as maghemite and magnetite offer convenient magnetic properties, low toxicity and price, high surface to volume ratios, and possibility for surface chemical modification (Girginova et al., 2010). Keeping in mind the interesting properties of magnetic magnetite and the higher values of the stability constants reported for Hgdithiocarbamate (Dtc) complexes (e.g. 1.2  1038 for Hg(Et2Dtc)2 (Venkatesan et al., 2002)), we have recently reported the synthesis and the preliminary application of silica

aRP b RL ARE SS CV-AFS DF RE RSD Sx/y

Redlich-Peterson constant, (L/mg)b Redlich-Peterson exponent Separation factor Average relative error Sum of squares Cold vapour atomic fluorescence spectroscopy Degrees of freedom Relative Error Relative Standard Deviation Standard Deviation of Residues

Subscripts 0 initial condition of experiment t intermediate condition of the experiment at a certain time e equilibrium condition of experiment

coated Fe3O4 particles functionalized with Dtc groups for the decontamination of synthetic waters with realistic Hg (II) levels (Girginova et al., 2010). However, this work also raised a number of important issues related with the performance of these materials as colloidal adsorbents in natural waters, particularly in strong salinity conditions. We wish to report here our research on the removal of Hg (II) at trace levels found in natural waters, and the effect of sorbent dose and salinity on the sorption efficiency. The results of this study on the sorption equilibrium and kinetics of Hg (II) onto the functionalized magnetic particles (MPs) were fitted to well-known kinetic and equilibrium equations. For the assessment of the real effectiveness of the developed sorbent, the functionalized magnetic NPs were applied in two natural waters (seawater and river water).

2.

Material and methods

2.1.

Chemicals

All chemicals used in this work were of analytical reagent grade, and obtained from commercial chemical suppliers and were used without further purification. The certified standard stock solution of mercury (II) nitrate was purchased from Merck (1000  2 mg/L).

2.2.

Adsorbent material

2.2.1.

Synthesis

Fe3O4/SiO2/NH/CS
2 magnetic particles were investigated for remediation of Hg (II) contaminated waters. The synthesis of these magnetic particles includes three distinct steps: (i) the synthesis of magnetite particles (Fe3O4), and their encapsulation in amorphous silica shells (Fe3O4/SiO2), (ii) further modification with 3-aminopropyltriethoxysilane (Fe3O4/SiO2/ NH2) and (iii) the grafting of dithiocarbamate groups at the surface of amine modified silica coated magnetite (Fe3O4/SiO2/ NH/CS
2 ).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4

The first two steps of the synthesis were performed exactly as described by Girginova et al. (Girginova et al., 2010), while the procedure adopted for the grafting of dithiocarbamates groups was slightly different in order to graft a higher amount of particles. This procedure was performed as follows: the amine modified silica coated magnetite (50 mg) was added to 15 mL 2-propanol under mechanic stirring for 1 h. After, 0.75 mL 1 M NaOH, and 0.12 mL CS2 were added and the suspension was mechanically stirred for 4 h. The powder was then collected magnetically from the suspension formed, washed with 2-propanol and dried at room temperature.

2.2.2.

Structural and chemical characterization

Fourier Transform Infrared (FT-IR) spectra of Fe3O4/SiO2/NH/ CS
2 magnetic particles were recorded using a spectrometer Mattson 7000 with 256 scans and 4 cm
1 resolution, using a horizontal attenuated total reflectance (ATR) cell. Elemental analysis for carbon, nitrogen, hydrogen and sulphur were performed using a LECO CHNS-932 elemental analyzer. The crystalline phase of the particles was identified by X-ray powder diffraction of the dried samples using a Philips X’Pert X-ray diffractometer equipped with a Cu Ka monochromatic radiation source. Transmission electron microscopy (TEM) was performed using a transmission electron microscope JEOL 200CX operating at 300 kV. The specific surface area of the magnetic particles was determined with nitrogen adsorption BET measurements performed with a Gemini Micromeritics instrument.

2.3.

Batch adsorption experiments

The ability of Fe3O4/SiO2/NH/CS
2 particles to capture Hg (II) from water was evaluated by contacting the particles with a Hg (II) solution for a required period of time. The sorption experiments were carried out in 500 mL batch reactors at 295  1 K, under mechanical stirring. Hg (II) solutions were prepared daily by diluting the corresponding standard solution, in high purity water (18 MUcm), to the desired initial concentration (50 mg/L), with a subsequent adjustment of pH to 7 with 0.1 M NaOH. The initial concentration of 50 mg/L was selected, as this is the current limit value for Hg discharges from industrial sectors other than the chloro-alkali electrolysis industry. All glassware used in these experiments was acid-washed prior to use with HNO3 25%, 12 h, and ultra-pure water. Experiments were performed to evaluate the time profile and the effects of sorbent concentration and salinity on the sorption of Hg (II) onto the functionalized magnetic particles. Accurately weighed amounts of Fe3O4/SiO2/NH/CS
2 particles were added to Hg (II) solutions, in glass batch reactors, which were immediately placed in an ultrasonic bath for ca. 10 s, for dispersing the magnetic particles. This time was considered the starting point for each experiment. For every experiment, the Hg (II) solution was continuously stirred, using a glass rod, and samples were withdrawn for analysis at several sampling times. Each sample was analyzed for the Hg (II) concentration after magnetic separation of the particles using a NdFeB magnet (1.48 T), and posterior pH adjustment (lower than 2) with concentrate HNO3 (Hg free, Merck). Mercury analyses were performed by cold vapour atomic fluorescence

5775

spectroscopy (CV-AFS), on a flow-injection-cold vapour atomic fluorescence spectrometer (Hydride/vapour generator PS Analytical Model 10.003, coupled to a PS Analytical Model 10.023 Merlin atomic fluorescence spectrometer; PS Analytical, Orpington, Kent, England) and using SnCl2 as reducing agent. An Hg (II) solution in the absence of magnetic particles was run as a control experiment. The time profile of Hg (II) sorption onto Fe3O4/SiO2/NH/CS
2 particles and the effect of sorbent dose, were determined by using four different amounts of the functionalized particles, namely 0.124, 0.256, 0.501 and 3.063 mg, which correspond to a sorbent dose of 0.2, 0.5, 1 and 6 mg/L. In order to study the effects of salinity changes on the sorption capacity of Fe3O4/SiO2/NH/CS
2 , 6 mg/L of particles were added to three different matrices spiked with 50 mg Hg (II)/L: ultra-pure water, 3 g/L NaCl solution and seawater collected from about 1 nautical mile from the Portuguese coast. In the following sections, those matrices will be denoted in terms of salinity (0% of salinity, 10% of salinity and 100% of salinity, respectively). The efficiency of Hg (II) removal by Fe3O4/SiO2/NH/CS
2 particles from different types of natural waters was also assessed by adding 6 mg/L of particles to river water collected from Vouga River (near the drinking water treatment plant) and spiked with 50 mg Hg (II)/L. Isotherms were obtained varying the amount of Fe3O4/ SiO2/NH/CS
2 particles from 0.124 to 3.063 mg, for an initial Hg (II) concentration of 50 mg/L and a temperature of 295  1 K. The amount of Hg (II) sorbed per unit of particles, at time t, qt (mg/g) was estimated from the mass balance between initial Hg (II) concentration and concentration at time t in solution, qt ¼ ðC0
Ct Þ 

V M

(1)

where V is the volume of solution (L) and M is the dry weight of Fe3O4/SiO2/NH/CS
2 particles (mg), C0 (mg/L) is the initial Hg (II) concentration and Ct (mg/L) is its concentration at time t. The results were also compared by removal percentage, which at equilibrium time is defined by: Removal% ¼

ðC0
Ce Þ  100 C0

(2)

where, Ce (mg/L) is the equilibrium Hg (II) concentration in the solution.

2.4.

Kinetic and equilibrium models

Different theoretical models were applied to experimental data in order to find a model which adequately describes kinetic and equilibrium data.

2.4.1.

Kinetic models

The chemical kinetics gives relevant information about reaction pathways and rates, along with time to reach equilibrium. Moreover, the physical and chemical features of the sorbent have great impact on both sorption kinetics and sorption mechanism (Hasan and Srivastava, 2009).

5776

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4

In order to evaluate the differences in the kinetic rates and to describe the kinetic removal of Hg (II) ions onto Fe3O4/SiO2/ NH/CS
2 particles in three matrices tested (ultra-pure water, NaCl solution and seawater), two simple and widely applied models, the pseudo-first- and pseudo-second-order models were used to describe the removal process. Numerous studies in literature have pointed out that one of those two models is able to describe the majority of the sorption studies (Wang and Chen, 2009; Ho and McKay, 1999).

2.4.1.1. Pseudo-first-order equation or Lagergren model. The first-order equation, firstly applied by Lagergren (Lagergren, 1898), is mathematically expressed by:
 dqt ¼ k1 qe
qt dt

(3)

(4)

2.4.1.2. Pseudo-second-order equation. The pseudo-secondorder equation may be also applicable and, in contrast with the previous model, usually correlates the behaviour over the whole range of sorption (Ho and McKay, 1999). The kinetic rate equation is expressed as:
2 dqt ¼ k2 qe
qt dt

(5)

where k2 (g/mg h) is the rate constant of pseudo-second-order. By applying the boundary conditions t¼0 to t¼t and qt¼0 to qt¼qe, the integrated form of Eq. (5) is: qt ¼

k2 q2e t 1 þ k2 qe t

2.4.2.2. Langmuir isotherm. This two-parameter model developed by Langmuir in 1916, relating the amount of gas sorbed on a surface to the pressure of the gas is probably one of the best known and widely used sorption isotherm (Ho et al., 2002). This model assumes that exist a fixed number of accessible sites available on the sorbent surface, all of them with the same energy and that the sorption is reversible. The Langmuir model is represented by the following equation: qe ¼

where k1 (h
1) is the rate constant of pseudo-first-order and qe (mg/g) is the amount of solute sorbed per gram of sorbent at equilibrium. After integration and application of the boundary condition qt¼0 at t¼0, Equation (3) becomes
 qt ¼ qe 1
e
k1 t

favourable adsorption, and is related to the non-linearity of the model.

(6)

qm kL Ce 1 þ kL Ce

(8)

where kL (L/mg) and qm (mg/g) are the Langmuir sorption equilibrium constant related to the energy of sorption and the maximum sorption capacity corresponding to complete monolayer coverage, respectively.

2.4.2.3. Sips or Langmuir-Freundlich isotherm. As the own name indicates, this three-parameter isotherm is a composite of the former isotherms. At high sorbate concentrations, it predicts a monolayer sorption capacity, characteristic of the Langmuir isotherm and at low sorbate concentrations it reduces to a Freundlich isotherm (Ho et al., 2002). The Sips equation can be describe as follows: qe ¼

Nt aCm e 1 þ aCm e

(9)

where Nt (mg/g) is the total number of binding sites, a is related to the median binding affinity (k), since a¼km and m is the heterogeneous index, which varies from 1 (homogeneous material) to 0 (heterogeneous material for m < 1) (Umpleby et al., 2001).

2.4.2.4. Redlich-Peterson 2.4.2.

Equilibrium models

Sorption equilibrium provides fundamental data for evaluating the applicability of the sorption process. Equilibrium isotherms give the equilibrium relationships between sorbent and sorbate, i.e. the quantity of solute sorbed and remaining in solution at a given temperature, at equilibrium. The equation parameters and the underlying thermodynamic assumptions of the equilibrium models often provide some insight into the sorption mechanism, the surface properties and the capacity and affinity of the sorbent (Ho et al., 2002). In this study, twoparameter isotherms (Langmuir and Freundlich) and threeparameter isotherms (Sips and Redlich-Peterson), in their nonlinear form were chosen to fit the experimental data.

2.4.2.1. Freundlich isotherm. This empirical model developed by Freundlich in 1906 (Freundlich, 1906), can be applied to multilayer sorption as well as non ideal sorption on heterogeneous surfaces and is represented by the following equation: qe ¼ kF Ce1=n

(7)

where kF (mg(1
1/n) L(1/n)/g) and n are the Freundlich parameters. n is usually between 1 and 10, which points out

isotherm. This three-parameter isotherm, which also incorporates characteristics of both the Langmuir and Freundlich isotherms, can be represented as follows: qe ¼

kRP Ce 1 þ aRP Cbe

(10)

where kRP (L/g) and aRP ((L/mg)b) are the Redlich-Peterson constants and b is the Redlich-Peterson exponent.

2.4.3.

Error analysis

The parameters of the kinetic and equilibrium models here considered were obtained by nonlinear regression analysis using GraphPad Prism 5 program, which uses the least-squares as fitting method and the method of Marquardt and Levenberg, which blends two other methods, the method of linear descent and the method of Gauss-Newton for adjusting the variables. In order to confirm which model presents the best fit to experimental data, the adjusted R squared (R0 2), the sum of squares (SS) and the standard deviation of residues (Sx/y) were analyzed. The standard error of the best fit parameters, the average relative error (ARE) of the fittings and the relative error (RE) between the experimental qe value and the models’ estimation value are also presented. The different statistical

5777

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4

parameters were determined using the following mathematical expressions: (11)

^

yi
y i

(13)

(440)

(511)

(400)

2000

(422)

2

Counts (a.u.)

X


(12)

(220)

3000

^ 2 yi
y i R2 ¼ 1
P
2 yi
yi

P


SS ¼

(311)

n
1 n
p
1

(111)


 R02 ¼ 1
1
R2 

1000

(P
Sx=y ¼

^

yi
y i n
2

2 )1=

2

(14)

0 20

 ^  P jyi
y i j yi  100 ARE ¼ n jyi
y i j  100 yi

(15)

Results and discussion

3.1.

Characterisation of Fe3O4/SiO2/NH/CS
2 particles

In the present work, surface modified Fe3O4 particles were firstly prepared using methodologies previously described by us (Girginova et al., 2010). Fig. 1 and Table 1 summarizes relevant results obtained for the characterization of these materials and that have been collected here for monitoring the materials properties. In brief, the powder X-ray diffraction patterns of the materials match those of magnetite,1 whose identity was unequivocally proved by Mo¨ssbauer spectroscopy (Girginova et al., 2010). As expected, TEM analysis (Fig. 1) showed single cubic particles of magnetite coated with amorphous silica shells. As previously reported, the cubic particles exhibited a magnetization hysteresis loop at room temperature with a saturation magnetization of 62 emu/g (Girginova et al., 2010). Finally, the surface functionalization of the silica coated magnetite particles was monitored step by step using infrared spectroscopy and the results are summarized in Table 1. Due to the higher value of the stability constant for the HgDtc complex is expectable that Hg (II) loading onto the functionalized particles will depend of the extent of functionalization carried out on the sorbent. Based on surface area (14.6 m2/g), sulphur content (0.98%) and taking into account that each Dtc group has two sulphur atoms, the amount of Dtc 1

50

60

70

Fig. 1 e X-ray powder diffraction patterns of Fe3O4 and TEM image of silica coated Fe3O4 particles. The Miller indices corresponding to the most intense reflection peaks of Fe3O4 are indicated in brackets.

(16)

where n is the sample size, p is the number of adjustable parameters in the model and yi are the experimental or observed values, yi is the mean of the observed data and y^i are the modelled or predicted values. The absolute error for each experimental point from the kinetics and equilibrium experiments can be found in the complementary file.

3.

40

2 θ (°)

^

RE


30

Joint Committee for Powder Diffraction Studies, JCPDS 19-0629.

groups on the sorbent was found to be 1.53  10
4 mol/g or 1.05  10
5 mol/m2. This value is slightly lower but of the same order of magnitude than the density of Dtc groups found for other sorbents like mesoporous silica grafted with Dtc (Venkatesan et al., 2003) (2.5  10
4 mol/g) and silica gel grafted with Dtc (Venkatesan et al., 2002) (3.7  10
4 mol/g).

3.2.

Hg (II) removal by Fe3O4/SiO2/NH/CS
2 particles

3.2.1.

Time profile of Hg (II) removal

Kinetic curves corresponding to the sorption of Hg (II) onto different amounts of Fe3O4/SiO2/NH/CS
2 particles are shown in Fig. 2. The plots represent the normalized Hg (II) concentration that remained in the liquid phase vs. time in Fig. 2A and the Hg (II) concentration in the MPs vs. time in Fig. 2B, both for an initial concentration of 50 mg Hg (II)/L. For all the amounts of Fe3O4/SiO2/NH/CS
2 particles, a decrease with time on Hg (II) concentration in the liquid phase was observed, even when only 0.2 mg/L were used. The time profile curves show that equilibrium time is attained in 24 h for the highest amount of particles used (6 mg/L), and as expected, it increases as long as the amount of particles decreases. This increase is particularly notorious by comparing the equilibrium time for the lowest and the highest amount of Fe3O4/SiO2/NH/CS
2 ; for the lowest

Table 1 e Assignment of infrared diagnosis bands for Fe3O4 and surface modified Fe3O4 particles. Sample Fe3O4 Fe3O4/SiO2

Fe3O4/SiO2/NH2 Fe3O4/SiO2/NH/CS
2

Wave number (cm
1)

Assignment

540, 310 (s) 1070 (vs) 789 (w) 943 (w) 786 (w) 1438 (s)

n(FeeO) n(SiOeSi) n(SieOH) n(SieOeFe) d(NeH) n(CeN)

(vs-very strong, s-strong, w-weak; n-stretching vibration, d-bending vibration).

5778

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4

amount of MPs (0.2 mg/L) the equilibrium time was achieved only after 96 h, while for the highest amount of MPs (6 mg/L) the equilibrium was achieved in less than 24 h (Fig. 2). This difference suggests that equilibrium time can be drastically affected by small variations in sorbent amount, and it is mainly related with the large availability of the Dtc groups. The time profile curves also reveal that during the first 12 h, the Hg (II) content in the particles increased quickly and then slowed down approaching equilibrium (Fig. 2B).

3.2.2. The effect of Fe3O4/SiO2/NH/CS
2 concentration on Hg (II) removal The concentration of sorbent is an important parameter to obtain quantitative metal removal, since for a given initial concentration of metal, it influences both the contact time necessary to reach equilibrium and the sorption capacity (Lopes et al., 2009). Fig. 3 shows the equilibrium removal percentage and the equilibrium amount of Hg (II) sorbed by gram of Fe3O4/SiO2/
NH/CS
2 vs. concentration of Fe3O4/SiO2/NH/CS2 . Given the obtained results it can be concluded that the particles concentration strongly affects Hg (II) removal and Hg (II) uptake per gram of particles. It is noticeable, that at equilibrium, the removal percentage increases with the increasing of

A

1.0 6 mg/L 1 mg/L

0.8

0.5 mg/L 0.2 mg/L

Ct/C0

0.6

Fe3O4/SiO2/NH/CS
2 concentration, from 53.2  1.3% (for 0.2 mg/L of particles) to 99.8  1.5% (for 6 mg/L of particles) while the amount of Hg (II) sorbed per gram of Fe3O4/SiO2/NH/ CS
2 particles decreases drastically from112  3 mg/g to 9.2  0.1 mg/g. Thus, for an invariable initial Hg (II) concentration, the increase of the sorbent concentration provides greater contact surface area and increases the number of available sorption sites, promoting the removal of Hg (II). As the number of sorption sites increases, the concentration of Hg (II) in the liquid phase and the amount of Hg (II) per mass of sorbent decreases and more sorption sites remain unsaturated during the sorption process. Additionally, the increase of the amount of Hg (II) per mass of MPs will only occur as long as the maximum capacity of the MPs is not fulfilled. This fact, suggest that under the experimental conditions tested, the maximum capacity of the magnetic particles was not achieved. Another remarkable result was the achievement of a considerably low residual concentration (0.10  0.02 mg Hg (II)/L) with only the application of a few milligrams of sorbent (6 mg/L). The residual Hg (II) concentration achieved is ten times lower than the current guideline value of the European Union for water of drinking quality ([Hg (II)]  1 mg/L) (EU 1998). This result clearly evidence the huge potential of this magnetic material to decontaminate waters and to achieve totally Hg free effluents, as is required by the Water Framework Directive (EU 2000). Based on the obtained results, 6 mg of Fe3O4/SiO2/NH/CS
2 particles per litre of Hg (II) solution (50 mg/L) should be used for a completely effectiveness of the decontamination process, since any further increase in the particles dose will not significantly affect the equilibrium between the ions sorbed in the solid phase and those remaining in solution.

0.4

3.2.3. 0.2

0.0

B

The effect of salinity on Hg (II) removal

It is well know that in the presence of chloride ions, Hg forms chloro-complexes of Hg (II) and consequently some sorbents become useless in saline waters (Lopes et al., 2007). The removal of Hg (II) by Fe3O4/SiO2/NH/CS
2 particles was studied in the presence of two different Cl
concentrations (ca. 10 and

120 120

120 Removal.. (%)

100

qe (mg/L)

100

100

80

80

60

60

40

40

20

20

qt , mg/g

80 60 40 20 0 0

24

48

72

96

120

t, h Fig. 2 e Normalized Hg (II) concentration in the liquid phase (A) and sorbed on the particles (B) as a function of time, for different amounts of Fe3O4/SiO2/NH/CSL 2 particles.

0

0 0.2

0.5

1.0

6.0

0.2

0.5

1.0

6.0

Fe 3O4/SiO2/NH/CS2- concentration (mg/L)

Fig. 3 e Effect of Fe3O4/SiO2/NH/CSL 2 concentration on Hg (II) removal (%) and sorption capacity (mg/g).

5779

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4

A

1.0 0% 10% 100%

0.8

Ct /C0

0.6

0.4

0.2

0.0

B

10

qt , mg/g

8

6

4

2

0 0

24

48

72

96

120

t, h Fig. 4 e Normalized Hg (II) concentration in the liquid phase (A) and sorbed on the particles (B) as a function of time, for different percentages of salinity.

100% of salinity) and the results, which are shown in Fig. 4, were also compared with those obtained for ultra-pure water (0% of salinity). The plots in Fig. 4A represent the normalized Hg (II) concentration remaining in the liquid phase vs. time, while the Hg (II) concentration in the MPs vs. time is represented in Fig. 4B. According with the results, the presence of high Cl
concentrations did not change significantly the equilibrium

values of the amount of Hg (II) sorbed per gram of Fe3O4/SiO2/ NH/CS
2 particles (qe values ranged from 9.08 to 9.35 mg/g, with a RSD ca. 1.5%), neither the equilibrium removal percentage, which was higher than 98% for all matrices (0, 10 and 100% of salinity). Even the residual concentration of Hg (II) in solution was lower than the current guideline value of the European Union for water of drinking quality (1 mg/L), for all three systems. For 0% of salinity, i.e. for ultra-pure water the residual concentration achieved was 0.10  0.02 mg Hg (II)/L, for 10% of salinity it was 0.97  0.01 mg Hg (II)/L and for 100% of salinity, i.e. for seawater the residual concentration achieved was 0.82  0.02 mg Hg (II)/L. Conversely, the presence of high Cl
concentrations clearly increased the equilibrium time from 24h (0% of salinity) to ca. 96h (10 and 100% of salinity) (see Fig. 4) and the rate which Hg (II) is removed by the particles decreases with the increasing of the matrix’s complexity, i.e. from ultra-pure water to seawater. These results suggest that besides, the equilibrium time, the matrix of the system Hg (II)/Fe3O4/SiO2/NH/CS
2 particles also influences the kinetic rate of the sorption process, as confirmed by the rate constants (k1 and k2) of the pseudo-firstand pseudo-second-order kinetic models (see Table 2). The kinetic sorption rate constants k1 and k2 decrease with the increasing of the salinity of the matrix, which reinforces that the higher is the ionic strength, the slower equilibrium is reached. The obtained results also suggest that during the first hours not all Hg (II) is available for the active sites of the Fe3O4/ SiO2/NH/CS
2 particles, probably due to the formation of mercury chloro-complexes. However, as the equilibrium values of the amount of Hg (II) removed per gram of functionalized particles (qe) did not change significantly for the different matrices, it is expectable that as the concentration of the Hg (II) available in solution decreases the decomplexation of the chloro-complexes occurs, since they are less stable (Ks 107-1015) (Nam et al., 2003) than the complexes formed between mercury and the Dtc groups (Ks 1038) (Venkatesan et al., 2002), increasing the concentration of Hg (II) available in solution, for complexing with the Dtc groups present at the surface of the magnetic particles. The results also allow to conclude that Naþ do not compete with Hg (II) for the active þ sites of the Fe3O4/SiO2/NH/CS
2 particles, even when the Na concentration was much higher than that of Hg (II). Furthermore, the modelling of the kinetic process by the pseudo-first- and pseudo-second-order models allows to conclude that whatever the kinetic equation used, the description of the sorption kinetics was satisfactory for all

Table 2 e First- and second-order sorption rate constants obtained for the removal of Hg (II) from matrices with different percentage of seawater, together with experimental and fitted qe, and the goodness of the of the fittings. Kinetic model Pseudo - 1st order 0% 10% 100% Pseudo - 2nd order 0% 10% 100%

Model’s parameters k1 h
1 0.397 0.183 0.154 k2 g/mg h 0.0667 0.0259 0.0224

SE 0.037 0.016 0.028 SE 0.0077 0.0032 0.0042

qe mg/g 9.10 8.89 8.78 qe mg/L 9.61 9.67 9.59

Goodness of the fit SE 0.19 0.22 0.45 SE 0.18 0.24 0.37

R0 2 0.98 0.98 0.92 R2 0.99 0.99 0.97

SS 1.53 1.69 6.85 SS 0.90 1.29 2.71

Sx/y 0.44 0.43 0.87 Sx/y 0.33 0.38 0.55

ARE % 4.4 10 18 ARE % 3.7 7.2 12

Experimental values DF 8 9 9 DF 8 9 9

qe mg/g 9.22 9.08 9.35 qe mg/g 9.22 9.08 9.35

RE % 1.3 2.1 6.1 RE % 2.6 2.3 2.2

5780

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4

matrices, as confirmed by the high values of R0 2 (0.92e0.99) and low values of Sx/y (0.33e0.87) (see Table 2). It is clear that both kinetic models describe very well the quantity of Hg (II) removed in the early stages of sorption, although there are slightly deviations between the experimental and the fitted data, particularly in the inflexion zone (Fig. 5). However, both models are able to accurately estimate the qe value of the three systems, which is corroborated by the low values of RE (1.3e6.1%) found between the predicted and the experimental qe values in all matrices. Still, comparing the predicted and experimental qe values it is perceptible that the Langergren model underestimates the qe values while the pseudo-secondorder model overestimates them. Although these slight differences, for a confidence level of 95% there is no significant difference between the goodness of the fit of the two models (F-test), for all tested matrices.

3.3.

The Langmuir equation assumes that adsorption occurs at definite localized sites on the surface, each site being able to bind a single molecule of the adsorbing species. The energy of adsorption is equal for all sites and there are no interaction forces between adjacently adsorbed molecules (Cooney, 1999). Theoretically, a saturation value is reached, beyond which no further sorption can take place, which is represented by a plateau in the equilibrium isotherm and corresponds to the assumption of one complete monomolecular layer of coverage

Sorption equilibrium

An accurate mathematical description of the equilibrium data between the concentration of the sorbate in the liquid and the amount in the solid phase is essential for a consistent prediction of the sorption parameters and for quantitative comparison of the sorption capacity of different sorbents. This mathematical function, called isotherm, is a basic requirement for designing any sorption system (Marin et al., 2009), and is obtained for a specific temperature and initial sorbate concentration. Fig. 6 shows the sorption isotherm of Hg (II) onto Fe3O4/ SiO2/NH/CS
2 particles, as well as the fit to the isotherm models described in the experimental section. The parameters of the isotherm models obtained from the corresponding fittings are presented in Table 3. All isotherms are positive and concave to the concentration axis, and under the experimental conditions here used, the experimental equilibrium values of the amount of Hg (II) sorbed in the functionalized particles increases with the increasing of Hg (II) concentration in solution, without reaching a saturation plateau. The Freundlich isotherm is an empirical equation, which does not assume that the material coverage must approach a constant value corresponding to one complete solute monomolecular layer as Ce gets larger. This model predicts that qe monotonously increases with increasing Ce which, being physically impossible, means that the Freundlich equation should fail to describe the experimental data at high Ce values (Cooney, 1999). However, the concentration values in real sorption processes are considered sufficiently diluted, in order to avoid the process entering the region where the Freundlich equation breaks down (Cooney, 1999). According with the obtained results, the Freundlich model provides a good description of the experimental data (ARE¼5.8%, R0 2¼0.98), since in the range of experimental conditions used, the equilibrium data do not achieved a plateau at a limiting value of Ce, suggesting the existence of heterogeneous surface conditions. Freundlich isotherm allows to calculate two empirical constants, kF and n, which are related to adsorption capacity of the sorbent and sorption intensity, respectively. The magnitudes of kF (601 mg(1
1/n) L(1/n)/g) and n (2.26) indicate easy separation of Hg (II) from liquid phase and favourable sorption (1 < n < 10).

0 ,

Fig. 5 e Sorption kinetics modelling of Hg (II) on the particles, for different percentages of seawater.

5781

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4

monolayer capacity (qm) of Fe3O4/SiO2/NH/CS
2 particles estimated by the Langmuir model is 142 mg/g, with a 95% confidence interval of [115e170 mg/g], and the Langmuir constant (kL), which represents the affinity between the sorbent and sorbate is 145 L/mg. Moreover, the isotherms can be classified as irreversible when the separation factor (RL) is 0; favourable when 0 < RL¼<1, linear when RL ¼ 1 and unfavourable when RL > 1. The separation factor (RL) can be calculated using the follow equation:

120

Two parameter models 100

qe, mg/g

80 60 40 20 0 0.000

RL ¼

Langmuir Freundlich

0.005

0.010

0.015

0.020

0.025

120

Three parameter models 100

qe, mg/g

80 60 40 Sips Redlich-Peterson

20 0 0.000

0.005

0.010

0.015

0.020

0.025

Ce , mg/L Fig. 6 e Equilibrium isotherms of Hg (II) on the particles at 21 ± 1  C.

of the adsorbing species on the adsorbent (Cooney, 1999; Kocaoba, 2007). Langmuir plot shows a higher difference between experimental equilibrium data and the predicted ones (ARE¼14%), although no significant differences were observed between the goodness of the fit of the Freundlich and Langmuir models (F-test for 95% confidence level). The

1 1 þ kL C0

The calculated RL value was 0.12, corroborating that Hg (II) sorption on Fe3O4/SiO2/NH/CS
2 particles is favourable. The Langmuir-Freundlich isotherm, also known as Sips isotherm, is an equation with three fitting coefficients with physical meaning (Nt, a and m), that describes the relationship of the equilibrium concentration of the sorbate between the solid and liquid phase in heterogeneous systems (Umpleby et al., 2001). As the name implies, this isotherm is a combination of the Langmuir and Freundlich isotherms. At low sorbate concentrations it can be reduced to the Freundlich isotherm, while at high sorbate concentrations, it predicts a monolayer sorption capacity characteristic of the Langmuir isotherm (Ho et al., 2002; Marin et al., 2009). This isotherm is capable of modelling both homogeneous and heterogeneous binding surfaces. The value of exponent m on Sips equation was 0.705, which means that Hg (II) sorption onto the functionalized particles is more of Langmuir type than that of Freundlich, since for m¼1 the Sips equation (Eq. (9)) reduces to the Langmuir equation (Eq. (8)) in which the variable a corresponds directly to binding affinity (k). The predicted value of Nt was higher than the corresponding value (qm) of the Langmuir model (Table 3). Redlich-Peterson isotherm also incorporate features of both Langmuir and Freundlich equations, approximating to Henry’s law at low concentrations (b¼0), while at higher concentrations its behaviour approaches that of the Freundlich isotherm (Ho et al., 2002). For b¼1 the RedlichPeterson isotherm reduces to the Langmuir form. In this study, the value of the exponent b approximates to 1 (0.860), suggesting like the Sips equation, that the equilibrium data can preferably be fitted by the Langmuir model rather the Freundlich. Among the three-parameter models, the Sips isotherm provides the highest value of R0 2 and the lowest values of ARE,

Table 3 e Isotherm constants of two- and three-parameter models for Hg (II) sorption on magnetic particles at 21 ± 1  C. Isotherm Freundlich Langmuir Sips Redlich-Peterson

Model’s parameters kF mg(1
1/n)L(1/n)/g 601 kL L/mg 145 a L/g 16.9 kRP L/g 26188

SE 109 SE 29 SE 30.7 SE 13485

n 2.26 qm mg/g 142 Nt mg/g 206 aRP (L/mg)b 114

Goodness of the fit SE 0.21 SE 11 SE 101 SE 44

m 0.705 b 0.860

SE 0.211 SE 0.227

R0 2i 0.98 R2 0.98 R2 0.98 R2 0.98

SS 148 SS 130 SS 102 SS 122

Sx/y 5.43 Sx/y 5.11 Sx/y 4.51 Sx/y 4.93

ARE 5.8 ARE 14 ARE 10 ARE 14

% % % %

DF 5 DF 5 DF 4 DF 4

5782

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4

SS and Sx/y, suggesting that this model is more appropriate to describe the experimental equilibrium data than the RedlichPeterson. Among the two-parameter models, the Langmuir is in agreement with the experimental data and with the variables obtained from the three-parameter models and provides higher value of R0 2 and lower values of SS and Sx/y than the Freundlich. However, the comparison (F-test) between the values of Sx/y obtained for all models indicates that for 95% confidence level there are no significant differences between the goodness of the fit of the different models, in the range of experimental conditions studied (Ce [0e0.025 mg/L]).

3.4. Application of Fe3O4/SiO2/NH/CS
2 particles in natural waters Besides the application of Fe3O4/SiO2/NH/CS
2 particles to seawater, already discussed in a former section, the feasibility of the functionalized particles were also tested in river water. Likewise for the seawater, the time profile curve (plot not shown) reveals a decrease on Hg (II) concentration with time. However, probably due to a higher complexity of the matrix and higher levels of organic matter, the equilibrium time in the river water was attained in 240 h against the 96 h necessary to achieve equilibrium in seawater. The percentage of Hg (II) removal obtained was ca. 97% for river water and ca. 99% for seawater, while the residual Hg (II) concentration in the liquid phase was, respectively, 1.20  0.07 and 0.82  0.02 mg Hg (II)/L. Those results suggest that in the case of river water, the amount of particles employed should be slightly higher than 6 mg/L in order to achieve a completely effectiveness of the decontamination process in this type of natural waters; however, it must be highlight the high effectiveness of Fe3O4/ SiO2/NH/CS
2 particles to remove Hg (II) from water, even from

natural waters, where the high complexity of the matrix could undermine their performance.

3.5.

Comparison with other sorbents

The effectiveness of the Fe3O4/SiO2/NH/CS
2 particles by means of maximum Hg (II) sorption capacities, is quantitatively compared with other sorbents, particularly with algal biomass (Tuzun et al., 2005), Romanian clays (Hristodor et al., 2010), ETS-4 titanosilicate (Lopes et al., 2009) and furfural carbon (Yardim et al., 2003) (Table 4). The maximum Hg (II) sorption capacities of these sorbents range between 122 and 246 mg/g and are of the same order of magnitude of that found for Fe3O4/SiO2/NH/CS
2 particles. Compared to other sorbents containing Dtc ligands, the maximum Hg (II) sorption capacity of Fe3O4/SiO2/NH/CS
2 particles is considerable higher than that found for mesoporous silica (MCM-41-Dtc) (Venkatesan et al., 2003) and silica gel (Si-Dtc) (Venkatesan et al., 2002), despite the lower Dtc surface coverage (respectively 2.5  10
4 and 3.7  10
4 mol/g against 1.5  10
4 mol/g of the magnetic particles). Comparable binding capacities were observed between Fe3O4/SiO2/ NH/CS-2 particles and other sorbents, like mesoporous silica grafted with other sulphur ligands as thiol (Mercier and Pinnavaia, 1998), mercaptan (Mattigod et al., 1999) and 1furoyl thiourea urea (Mureseanu et al., 2010) (see Table 4). While sorbents like benzoylthiourea-modified mesoporous silica (Antochshuk et al., 2003) and thiol functional organoceramic composite (Nam et al., 2003), with higher sulphur surface coverage (ca. 10
3 mol/g) and/or multifunctional ligands, possess several “active” groups toward mercury ions and exhibit considerable higher sorption capacities, on the other hand they do not offer the possibility of magnetic separation.

Table 4 e Residual mercury concentration (Ce) and sorption capacity (qm) of other sorbents for Hg (II).

Biosorbents

Clays and zeolitic materials Carbons

Sorbents containing sulphur ligands

n.a. not available.

Adsorbent

qm (mg/g)

Ce (mg/L)

Ref.

Rice husk ash Bacillus sp. Eucalyptus bark Seaweed biomass Yeast cells Algal biomass Zeolitic mineral Clay ETS-4 titanosilicate Activated carbon Carbon aerogel Activated carbon Furfural carbon Thiol functional organoceramic composite Dithiocarbamate grafted on mesoporous silica Dithiocarbamate grafted on silica gel Thiol functionality grafted on mesoporous silica Mesoporous silica containing mercaptan groups Benzoylthiourea-modified mesoporous silica Mesoporous silica grafted with 1-furoyl thiourea urea Dithiocarbamate grafted on magnetite particles

6.72 7.94 33.1 84.7 93.4 122 10.1 152 246 25.8 34.9 43.8 174 726 40.1 61 110e301 26e270 1000 122 142e206

n.a 20 n.a n.a n.a n.a n.a n.a <1 n.a w5000 n.a n.a <1 n.a. n.a. n.a. 401 n.a. n.a. <1

(Feng et al., 2004) (Green-Ruiz, 2006) (Ghodbane and Hamdaoui, 2008) (Zeroual et al., 2003) (Yavuz et al., 2006) (Tuzun et al., 2005) (Gebremedhin-Haile et al., 2003) (Hristodor et al., 2010) (Lopes et al., 2009) (Rao et al., 2009) (Goel et al., 2005) (Ranganathan, 2003) (Yardim et al., 2003) (Nam et al., 2003) (Venkatesan et al., 2003) (Venkatesan et al., 2002) (Mercier and Pinnavaia, 1998) (Mattigod et al., 1999) (Antochshuk et al., 2003) (Mureseanu et al., 2010) This study

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4

It must also be highlighted that the Fe3O4/SiO2/NH/CS
2 particles together with thiol functional organoceramic composite and ETS-4 titanosilicate were the only sorbents, mentioned in Table 4, able to reduce the initial mercury concentration to values lower than 1 mg/L (Table 4). This fact is due not only to their sorption capacity but also because the majority of the studies use initial Hg concentrations extremely high and nothing realistic of the degree of contamination found in the environment.

4.

Conclusions

The sorption capacity towards Hg (II) of silica coated magnetite particles derivatized with dithiocarbamate groups was studied in batch mode under different experimental conditions. It was confirmed that silica coated magnetite particles functionalized with Dtc groups are effective sorbents for Hg (II) removal from synthetic and natural waters; however the sorption process is strongly dependent on contact time and particles concentration. The presence of higher concentrations of Cl
and Naþ ions did not affect the amount of Hg (II) removed per gram of sorbent at equilibrium but reduced the rate at which Hg ions were removed from solution, as confirmed by pseudo-first- and pseudo-second-order kinetic models. The Sips model provides a good description of the equilibrium data, predicting a maximum sorption capacity of 206 mg/g at 211  C, which is quite high compared with the sorption capacities found in the literature for other materials. Furthermore, this study confirmed the effectiveness of silica coated magnetite particles grafted with Dtc groups in two distinct types of natural waters, seawater and river water. However, the functionalized magnetite particles exhibited a slightly higher performance in seawater than in river water. It must be highlighted that in both ultra-pure water and seawater, only 6 mg/L of functionalized particles was sufficient to achieve a residual concentration lower than 1 mg/L, which is the current acceptable value for drinking water quality. Our results emphasize the advantages of these Dtc functionalized particles, such as high affinity towards mercury ions, selective removal, and large efficiency in high complex matrices. Additionally, the easy separation of the sorbent from solution due to the magnetite ferrimagnetic properties opens new prospects in the design of highperformance sorbents for environment remediation and mercury ions recovery.

Acknowledgements C.B. Lopes thanks Fundac¸a˜o para a Cieˆncia e Tecnologia for a Post-Doc grant (SFRH/BPD/45156/2008).

Appendix. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.08.057.

5783

references

Antochshuk, V., Olkhovyk, O., Jaroniec, M., Park, I.S., Ryoo, R., 2003. Benzoylthiourea-modified mesoporous silica for mercury(II) removal. Langmuir 19 (7), 3031e3034. Cooney, D.O., 1999. Adsorption Design for Wastewater Treatment. Lewis Publishers, Boca Raton (FL). Daniel-da-Silva, A.L., Trindade, T., Goodfellow, B.J., Costa, B.F.O., Correia, R.N., Gil, A.M., 2007. In situ synthesis of magnetite nanoparticles in carrageenan gels. Biomacromolecules 8 (8), 2350e2357. EU, 1998. Council Directive 98/83/EC on the quality of water intended for human consumption. Official Journal 330, 0032e0054. EU, 2000. Directive 2000/60/EC of the European Parliament and of the Council of the European Union, establishing a framework for Community action in the field of water policy. Official Journal of the European Communities 327/1. EU, 2008. Directive 2008/105/EC of the European Parliament and of the Council of the European Union, on environmental quality standards in the field of water policy, amending and subsequently repealing Council Directives 82/176/EEC, 83/513/ EEC, 84/156/EEC, 84/491/EEC, 86/280/EEC and amending Directive 2000/60/EC of the European Parliament and of the Council. Official Journal of the European Communities 348/84. Feng, X., Fryxell, G.E., Wang, L.Q., Kim, A.Y., Liu, J., Kemner, K.M., 1997. Functionalized monolayers on ordered mesoporous supports. Science 276 (5314), 923e926. Feng, Q.G., Lin, Q.Y., Gong, F.Z., Sugita, S., Shoya, M., 2004. Adsorption of lead and mercury by rice husk ash. Journal of Colloid and Interface Science 278 (1), 1e8. Freundlich, H., 1906. Concerning adsorption in solutions. Zeitschrift Fur Physikalische ChemieeStochiometrie Und Verwandtschaftslehre 57 (4), 385e470. Gebremedhin-Haile, T., Olguin, M.T., Solache-Rios, M., 2003. Removal of mercury ions from mixed aqueous metal solutions by natural and modified zeolitic minerals. Water Air and Soil Pollution 148 (1e4), 179e200. Ghodbane, I., Hamdaoui, O., 2008. Removal of mercury(II) from aqueous media using eucalyptus bark: kinetic and equilibrium studies. Journal of Hazardous Materials 160 (2e3), 301e309. Girginova, P.I., Daniel-Da-Silva, A.L., Lopes, C.B., Figueira, P., Otero, M., Amaral, V.S., Pereira, E., Trindade, T., 2010. Silica coated magnetite particles for magnetic removal of Hg2þ from water. Journal of Colloid and Interface Science 345 (2), 234e240. Goel, J., Kadirvelu, K., Rajagopal, C., Garg, V.K., 2005. Investigation of adsorption of lead, mercury and nickel from aqueous solutions onto carbon aerogel. Journal of Chemical Technology and Biotechnology 80 (4), 469e476. Green-Ruiz, C., 2006. Mercury(II) removal from aqueous solutions by nonviable Bacillus sp from a tropical estuary. Bioresource Technology 97 (15), 1907e1911. Hasan, S.H., Srivastava, P., 2009. Batch and continuous biosorption of Cu2þ by immobilized biomass of Arthrobacter sp. Journal of Environmental Management 90 (11), 3313e3321. Ho, Y.S., McKay, G., 1999. Pseudo-second order model for sorption processes. Process Biochemistry 34 (5), 451e465. Ho, Y.S., Porter, J.F., McKay, G., 2002. Equilibrium isotherm studies for the sorption of divalent metal ions onto peat: copper, nickel and lead single component systems. Water Air and Soil Pollution 141 (1e4), 1e33. Hristodor, C., Copcia, V., Lutic, D., Popovici, E., 2010. Thermodynamics and kinetics of Pb(II) and Hg(II) ions removal from aqueous solution by Romanian clays. Revista De Chimie 61 (3), 285e289.

5784

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4

Kocaoba, S., 2007. Comparison of Amberlite IR 120 and dolomite’s performances for removal of heavy metals. Journal of Hazardous Materials 147 (1e2), 488e496. Lagergren, S., 1898. Zur theorie der sogenannten adsorption gelo¨ter stoffe. Kungliga Svenska Vetenskapsakademiens. Handlingar 24 (4), 1e39. Lopes, C.B., Otero, M., Coimbra, J., Pereira, E., Rocha, J., Lin, Z., Duarte, A., 2007. Removal of low concentration Hg2þ from natural waters by microporous and layered titanosilicates. Microporous and Mesoporous Materials 103 (1e3), 325e332. Lopes, C.B., Otero, M., Lin, Z., Silva, C.M., Rocha, J., Pereira, E., Duarte, A.C., 2009. Removal of Hg2þ ions from aqueous solution by ETS-4 microporous titanosilicate-kinetic and equilibrium studies. Chemical Engineering Journal 151 (1e3), 247e254. Marin, A.B.P., Aguilar, M.I., Meseguer, V.F., Ortuno, J.F., Saez, J., Llorens, M., 2009. Biosorption of chromium (III) by orange (Citrus cinensis) waste: batch and continuous studies. Chemical Engineering Journal 155 (1e2), 199e206. Mattigod, S.V., Feng, X.D., Fryxell, G.E., Liu, J., Gong, M.L., 1999. Separation of complexed mercury from aqueous wastes using self-assembled mercaptan on mesoporous silica. Separation Science and Technology 34 (12), 2329e2345. Mercier, L., Pinnavaia, T.J., 1998. Heavy metal lan adsorbents formed by the grafting of a thiol functionality to mesoporous silica molecular sieves: Factors affecting Hg(II) uptake. Environmental Science and Technology 32 (18), 2749e2754. Mureseanu, M., Reiss, A., Cioatera, N., Trandafir, I., Hulea, V., 2010. Mesoporous silica functionalized with 1-furoyl thiourea urea for Hg(II) adsorption from aqueous media. Journal of Hazardous Materials 182 (1e3), 197e203. Nam, K.H., Gomez-Salazar, S., Tavlarides, L.L., 2003. Mercury(II) adsorption from wastewaters using a thiol functional adsorbent. Industrial and Engineering Chemistry Research 42 (9), 1955e1964. Ranganathan, K., 2003. Adsorption of Hg(II) ions from aqueous chloride solutions using powdered activated carbons. Carbon 41 (5), 1087e1092.

Rao, M.M., Reddy, D., Venkateswarlu, P., Seshaiah, K., 2009. Removal of mercury from aqueous solutions using activated carbon prepared from agricultural by-product/waste. Journal of Environmental Management 90 (1), 634e643. Starvin, A.M., Rao, T.P., 2004. Removal and recovery of mercury(II) from hazardous wastes using 1-(2-thiazolylazo)-2-naphthol functionalized activated carbon as solid phase extractant. Journal of Hazardous Materials 113 (1e3), 75e79. Tuzun, I., Bayramoglu, G., Yalcin, E., Basaran, G., Celik, G., Arica, M.Y., 2005. Equilibrium and kinetic studies on biosorption of Hg(II), Cd(II) and Pb(II) ions onto microalgae Chlamydomonas reinhardtii. Journal of Environmental Management 77 (2), 85e92. Umpleby, R.J., Baxter, S.C., Chen, Y.Z., Shah, R.N., Shimizu, K.D., 2001. Characterization of molecularly imprinted polymers with the Langmuir-Freundlich isotherm. Analytical Chemistry 73 (19), 4584e4591. Venkatesan, K.A., Srinivasan, T.G., Rao, P.R.V., 2002. Removal of complexed mercury from aqueous solutions using dithiocarbamate grafted on silica gel. Separation Science and Technology 37 (6), 1417e1429. Venkatesan, K.A., Srinivasan, T.G., Rao, P.R.V., 2003. Removal of complexed mercury by dithiocarbamate grafted on mesoporous silica. Journal of Radioanalytical and Nuclear Chemistry 256 (2), 213e218. Wang, J.L., Chen, C., 2009. Biosorbents for heavy metals removal and their future. Biotechnology Advances 27 (2), 195e226. Yardim, M.F., Budinova, T., Ekinci, E., Petrov, N., Razvigorova, M., Minkova, V., 2003. Removal of mercury (II) from aqueous solution by activated carbon obtained from furfural. Chemosphere 52 (5), 835e841. Yavuz, H., Denizli, A., Gungunes, H., Safarikova, M., Safarik, I., 2006. Biosorption of mercury on magnetically modified yeast cells. Separation and Purification Technology 52 (2), 253e260. Zeroual, Y., Moutaouakkil, A., Dzairi, F.Z., Talbi, M., Chung, P.U., Lee, K., Blaghen, M., 2003. Biosorption of mercury from aqueous solution by Ulva lactuca biomass. Bioresource Technology 90 (3), 349e351.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 8 5 e5 7 9 5

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Influence of operational parameters on nitrogen removal efficiency and microbial communities in a full-scale activated sludge process Young Mo Kim a, Hyun Uk Cho b, Dae Sung Lee c, Donghee Park d,*, Jong Moon Park b,** a

Department of Civil and Environmental Engineering, University of Massachusetts, Amherst, MA 01003, USA Department of Chemical Engineering, School of Environmental Science and Engineering, Division of Advanced Nuclear Engineering, Pohang University of Science and Technology, Pohang 790-784, Republic of Korea c Department of Environmental Engineering, Kyungpook National University, Daegu 702-701, Republic of Korea d Department of Environmental Engineering, Yonsei University, Wonju 220-710, Republic of Korea b

article info

abstract

Article history:

To improve the efficiency of total nitrogen (TN) removal, solid retention time (SRT) and

Received 5 May 2011

internal recycling ratio controls were selected as operating parameters in a full-scale

Received in revised form

activated sludge process treating high strength industrial wastewater. Increased biomass

17 August 2011

concentration via SRT control enhanced TN removal. Also, decreasing the internal recy-

Accepted 29 August 2011

cling ratio restored the nitrification process, which had been inhibited by phenol shock

Available online 3 September 2011

loading. Therefore, physiological alteration of the bacterial populations by application of specific operational strategies may stabilize the activated sludge process. Additionally, two

Keywords:

dominant ammonia oxidizing bacteria (AOB) populations, Nitrosomonas europaea and

Activated sludge

Nitrosomonas nitrosa, were observed in all samples with no change in the community

Operational parameter

composition of AOB. In a nitrification tank, it was observed that the Nitrobacter populations

Nitrogen removal

consistently exceeded those of the Nitrospira within the nitrite oxidizing bacteria (NOB)

qPCR

community. Through using quantitative real-time PCR (qPCR), nirS, the nitrite reducing

Nitrifying bacteria

functional gene, was observed to predominate in the activated sludge of an anoxic tank,

Denitrifying bacteria

whereas there was the least amount of the narG gene, the nitrate reducing functional gene. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Activated sludge process is one of the most widely used biological treatments of wastewaters containing carbon and nitrogen pollutants. Biological nitrogen removal has traditionally been accomplished using autotrophic nitrification and heterotrophic denitrification. Nitrification is carried out in two sequential steps via two distinct groups of bacteria: ammonia oxidizing bacteria (AOB) and nitrite oxidizing bacteria (NOB). Denitrification consists of consecutive

reactions in which nitrate or nitrite is transformed into gaseous forms (N2 or N2O). Although efficient and reliable in treating industrial wastewater, activated sludge process is susceptible to disturbances and toxic loadings (Juliastuti et al., 2003; Mertoglu et al., 2008; Kim et al., 2009). In particular, the activity of nitrifying bacteria in wastewater treatment plants is sensitive to shifts in the process’s pH and temperature, ammonia/ nitrite concentrations, oxygen concentration and the presence of toxic compounds, often leading to process failure

* Corresponding author. Tel.: þ82 33 760 2435; fax: þ82 33 760 2571. ** Corresponding author. Tel.: þ82 54 279 2275; fax: þ82 54 279 2699. E-mail addresses: [email protected] (D. Park), [email protected] (J.M. Park). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.063

5786

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 8 5 e5 7 9 5

(Eighmy and Bishop, 1989; Kim et al., 2007). Fluctuation inflow, organic loads, solid retention time (SRT) and lack of nutrient matter also inhibit the nitrification process (Hallin et al., 2005). Even though nitrifying bacteria have been fairly well studied, nitrification is difficult to maintain stably throughout the activated sludge process (Kim et al., 2009). Actually, many current full-scale activated sludge processes treating industrial wastewater have long experienced trouble with instability, but previous reports have simply focused on monitoring the emergence of the process failure (Kim et al., 2007, 2008, 2009). The causes of the process instability have not been clearly identified; thus appropriate solutions have not yet been suggested (Kim et al., 2011a,b). Moreover, research targeting a full-scale wastewater treatment process (WWTP) has rarely been attempted due to either sampling problems or difficulties in proper WWTP selection. Results from lab scale experiments have proved difficult to extrapolate to real WWTP conditions (Kim et al., 2007, 2009). In addition, little is known about the microbial ecology of nitrifying and denitrifying bacteria in full-scale activated sludge processes treating high strength industrial wastewater containing toxic compounds like thiocyanate, ammonia and phenol. Therefore, this study aimed to identify the causes underlying current difficulties achieving and maintaining the legal discharge of total nitrogen (TN) in the final effluent of a full-scale activated sludge process. To improve the efficiency of TN removal, SRT and internal recycling ratio controls as main operation parameters were attempted in an unstable process. Meanwhile, bacterial populations responsible for biological nitrogen removal were investigated in relation to changes in the efficiency of TN removal of a full-scale activated sludge process treating wastewater from a coke plant. Diversity surveys assessing the relationship between bacterial populations and activity to the overall processing conditions may lead to an understanding of the basis of process instability and be of help in designing better process monitoring while avoiding operational failure.

2.

Materials and methods

2.1.

Wastewater treatment plant operation and samples

nitrified effluent e the pre-denitrification activated sludge process. This system is composed of an anoxic tank (900 m3), an aerobic tank (3600 m3) and a nitrification tank (1800 m3) with a combined volume of 6300 m3 and treats about 300 m3 of cokes wastewater per hour (Fig. 1). Nitrified effluent was recycled from the nitrification tank to the anoxic tank at the rate of about 600 m3 per hour. Return activated sludge collected from the secondary clarifier was pumped back to the anoxic tank at the rate of about 300 m3 per hour. Both the aerobic and nitrification tanks were aerated. In the course of this study, the concentrations of pollutants in the raw wastewater were as follows: 2025e3150 mg/L of chemical oxygen demand (COD), 680e1070 mg/L of biochemical oxygen demand (BOD5), 642e916 mg/L of total organic carbon (TOC), 268e715 mg/L of phenol, 177e236 mg-N/L of total nitrogen (TN), 76e140 mg-N/L of ammonia, 190e297 mgSCN/L of thiocyanate (SCN) and 12.4e17.2 mg-CN/L of total cyanides. The mixed-liquor suspended solids (MLSS) of the system were controlled at about 1800 mg/L, the average hydraulic retention time (HRT) was 0.9 day and the average SRT was 15 days. The SRT was controlled by removal of excess sludge, resulting in different MLSS concentrations of the system. The temperature range of the tanks varied between 33 and 36  C. The pH of the influent, anoxic tank and nitrification tank was maintained at 9.0, 7.5 and 7.0, respectively. The dissolved oxygen (DO) concentration in the aerobic and nitrification tank was more than 4.0 mg/L, while the DO level of the anoxic tank was maintained below 0.3 mg/L. Functional stability of the system was defined and quantified by the effluent concentration of TN. MLSS samples for this study were taken from the last sections of the anoxic and nitrification tanks weekly for 3 months (AugusteOctober). Nitrification and denitrification activity was measured directly on fresh samples. For the DNA based studies, each sample of 1.0 mL was dispensed into a 1.5 mL sterile tube and centrifuged at 13000 g for 10 min. The supernatant was decanted and the pellet was stored at 20  C before being used.

2.2.

A full-scale WWTP of a coke manufacturing plant in Pohang, Korea employs a single sludge along with the recycling of

Process monitoring and chemical analysis

The collected samples were centrifuged at 3500 rpm for 3 min (MF550, Hanil Sci. Ind., Korea), and the supernatants were analyzed as follows: according to standard methods (APHA,

Fig. 1 e Schematic diagram of a full-scale activated sludge process treating high strength industrial wastewater.

5787

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 8 5 e5 7 9 5

1998), chemical oxygen demand (COD), ammonia, phenol and thiocyanate (SCN) were analyzed by the colorimetric method with a spectrophotometer (Genesys TM-5, Spectronic Inc., USA). After distillation, the cyanide (CN) concentration was determined by the pyridine-pyrazolone method. Nitrite and nitrate ions were measured with an ion chromatograph (ICS-1000, Dionex Co., USA). Total oragnic carbon (TOC), inorganic carbon (IC), and TN were measured with a TOC/TN analyzer (TOC-V csu, TNM1, Shimadzu Co., Japan).

2.3.

2.4.

DNA extraction

The pellet was washed with 1 mL of deionized and distilled water (DDW) and centrifuged at 16,000 g for 5 min to ensure a maximal removal of residual medium. The supernatant was carefully removed and the pellet resuspended in 100 mL of DDW. All DNA in the suspension was immediately extracted using an automated nucleic acid extractor (Magtration System 6 GC, PSS, Chiba, Japan). Purified DNA was eluted with 100 mL of TriseHCl buffer (pH 8.0) and stored at 20  C for further analyses.

Microbial activity test 2.5.

To investigate microbial activities of nitrifiers and denitrifiers in each tank at the WWTP, nitrification and denitrification rates were estimated weekly through batch experiments with synthetic medium containing ammonia or nitrate. Batch experiments for nitrification and denitrification activity were carried out in 500 mL Erlenmeyer flasks filled with 100 mL of test solution containing 50 mg-N/L of ammonia and nitrate ion, respectively. Without any pretreatment each flask was inoculated with fresh activated sludge (about 2000 mg/L), which was sampled directly from each full-scale tank, and then agitated on a thermostatic shaker at 200 rpm and 35  C, maintaining the pH at 7.5. The specific nitrification and denitrification rates were calculated applying the equation provided by Kim et al. (2011a).

T-RFLP analysis

T-RFLP was used to analyze the nitrifying bacteria community in the pre-denitrification process reactor based on the known 16S rRNA genes of ammonia oxidizing bacteria (AOB) and nitrite oxidizing bacteria (NOB), as described in the protocol of a previous study (Siripong and Rittmann, 2007). Because of the low concentration of DNA from the nitrifiers, it was amplified through nested PCR, using the universal primers 11f and 1492r (Table 1), followed by the specific amplification of nitrifier genes (Nitrifier-specific reverse primer: Nso1225r, NIT3r, Ntspa685r, Forward primer: Eub338f included phosphoramidite dye 6-FAM (Table 1)) (Siripong and Rittmann, 2007). 2 mL of template DNA was used for the universal amplification step and 1 mL of the universal amplification product as the

Table 1 e Primers and probes used in T-RFLP and qPCR. Target For T-RFLP Bacterial 16S rDNA Bacterial I6S rDNA AOB 16S rDNA Nitrobacter 16S rDNA Nitrospira 16S rDNA For qPCR Bacterial 16S rDNA

AOB 16S rDNA

Nitrospira spp. 16S rDNA

Nitrobacter spp. I6S1DNA

narG gene nirS gene nirK gene nosZ gene

Primer/probe

Sequence (5’-3’)

References

11f 1492r Eub338f Nso1225r NIT3r Ntspa685r

5’-GTTTGATCCTGGCTCAG-3’ 5’-TACCTTGTTACGACTT-3’ 5’-(6-FAM)-ACTCCTACGGGAGGCAGC-3’ 5’-CGCCATTGTATTACGTGTGA-3’ 5’-CCTGTGCTCCATGCTCCG-3’ 5’-CGGGAATTCCGCGCTC-3’

Kane et al., 1993 Lin and Stahl, 1995 Amann et al., 1990 Mobarry et al., 1996 Wagner et al., 1995 Regan et al., 2002

1055f 1392r 16STaq 1115 CTO 189fA/Ba CTO 1891Ca RTlr TMP1 NSR 1113f NSR 1264r NSR 1143Taq Nitro 1198f Nitro 1423r Nitro 1374Taq narG 1960m2f narG 2050m2r nirS If nirS 3r nirK 876 nirK 1040 nosZ 2f nosZ 2r

5’-ATGGCTGTCGTCAGCT-3’ 5’-ACGGGCGGTGTGTAC-3’ 5’-(6-FAM)-CAACGAGCGCAACCC-(TAMRA)-3’ 5’-GGAGRAAAGCAGGGGATCG-3’ 5’-GGAGGAAAGTAGGGGATCG-3’ 5’-CGTCCTCTCAGACCARCTACTG-3’ 5’-(6-FAM)-CAACTAGCTAATCAGRCATCRGCCGCT-(TAMRA)-3’ 5’-CCTGCTTTCAGTTGCTACCG-3’ 5’-GTTTGCAGCGCTTTGTACCG-3’ 5’-(6-FAM)-AGCACTCTGAAAGGACTGCCCAGG-(TAMRA)-3’ 5’-ACCCCTAGCAAATCTCAAAAAACCG-3’ 5’-CTTCACCCCAGTCGCTGACC-3’ 5’-(6-FAM)-AACCCGCAAGGAGGCAGCCGACC-(TAMRA)-3’ 5’-TAYGTSGGGCAGGARAAACTG-3’ 5’-CGTAGAAGAAGCTGGTGCTGTT-3’ 5’-TACCACCCSGARCCGCGCGT-3’ 5’-GCCGCCGTCRTGVAGGAA-3’ 5’-ATYGGCGGVCAYGGCGA-3’ S’-GCCTCGATCAGRTTRTGGTT-S’ 5’-CGCR ACGGC AAS AAGGTSM SSGT-3’ 5’-CAKRTGCAKSGCRTGGCAGAA-3’

Ferris et al., 1996 Ferris et al., 1996 Harms et al., 2003 Hermansson and Lindgren, Hermansson and Lindgren, Hermansson and Lindgren, Hermansson and Lindgren, Dionisi et al., 2002 Dionisi et al., 2002 Harms et al., 2003 Graham et al., 2007 Graham et al., 2007 Graham et al., 2007 Lo`pez-Gutie`rrez et al., 2004 Lo`pez-Gutie`rrez et al., 2004 Braker et al., 1998 Braker et al., 1998 Henry et al., 2004 Henry et al., 2004 Henry et al., 2006 Henry et al., 2006

a A mixture of CTO 189fA/B and CTO 189fC at the weight ratio of 2:1 was used as the forward primer.

2001 2001 2001 2001

5788

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 8 5 e5 7 9 5

template for the nitrifier-specific amplification. Finally, the PCR products were purified and 16S rRNA gene amplicons were digested with MspI restriction endonuclease (Siripong and Rittmann, 2007). Digested PCR products were run through an ABI 3130XL Genetic Analyzer (Applied Biosystems, Foster City, USA) at the SolGent Company (Korea). The peak results were analyzed using the Peak Scanner software v 1.0 (https://products.appliedbiosystems.com; Applied Biosystems, Foster City, USA). Details of the PCR conditions, product purification and restriction digestion are provided elsewhere (Siripong and Rittmann, 2007).

2.6.

qPCR analysis

To investigate the changes in the nitrifying and denitrifying bacteria populations occurring during the process performance, all qPCR assays were performed using a 7300 Real Time PCR system (Applied Biosystems, Foster City, USA). To determine the amount of the nitrifying bacteria, four independent qPCR assays were conducted by quantifying total bacterial 16S rDNA, ammonia oxidizing bacterial 16S rDNA, Nitrospira spp. 16S rDNA, and Nitrobacter spp. 16S rDNA (Table 1). Each capillary tube was separately loaded with 2 mL of template DNA (at 14e26 ng/mL), followed by 4.0 pmol of the forward and reverse primers (1 mL), together with 2.0 pmol of the TaqMan probe (0.5 mL) corresponding to each primer and probe set, 12.5 mL of TaqMan Universal PCR Master Mix (No 4304437 Applied Biosystems, New Jersey, USA), and PCR-grade sterile water for a final volume of 25 mL. The amount of total bacterial 16S rDNA was amplified using primer (1055f and 1392r) (Ferris et al., 1996). The TaqMan probe (16STaq1115) was modified by the 1114f primer (Harms et al., 2003). The PCR program was 2 min at 50  C, 10 min at 95  C; 45 cycles of 30 s at 95  C, 60 s at 50  C, and 40 s at 72  C. To determine the amount of AOB 16S rDNA genes, two forward primers (CTO 189A/B and CTO 189C), one reverse (RT1r), and the TaqMan probe (TMP1) were used as described previously by Hermansson and Lindgren (2001). The PCR program for AOB 16S rDNA quantification included 2 min at 50  C, 10 min at 95  C; 40 cycles of 30 s at 95  C, 60 s at 60  C. The Nitrospira spp. 16S rDNA primers (NSR 1113f/NSR 1264r) (Dionisi et al., 2002) and the TaqMan probe (NSR 1143Taq) (Harms et al., 2003) were tested. PCR amplification consisted of 2 min at 50  C, 10 min at 95  C; 50 cycles of 30 s at 95  C, 60 s at 60  C. Lastly, the amount of Nitrobacter spp. from Graham et al. (2007) was amplified using primer (Nitro 1198f/Nitro 1423r) and TaqMan probe (Nitro 1374Taq). The program used for amplification was 2 min at 50  C, 10 min at 95  C; 50 cycles of 20 s at 94  C, 60 s at 58  C, and 40 s at 72  C. Meanwhile, the denitrifying functional genes were quantified with SYBR Premix Ex Tag (Takara, Japan). Amplification reactions were performed in a total volume of 25 mL containing 2 mL of template DNA (at 17.5e22.5 ng/mL), 4.0 pmol of the forward and reverse primers (1 mL), together with 0.5 mL (1X) of the ROX reference dye (50X), 12.5 mL of SYBR Premix, and PCR-grade sterile water. The qPCR program for 16S rDNA amplification using primer 1055f and 1392r was 30 s at 95  C; 30 cycles of 15 s at 95  C, 20 s at 55  C, and 31 s at 72  C. Primers designed by Lo´pez-Gutie´rrez et al. (2004) were used to determine the amount of narG gene. The PCR program for narG gene

quantification included 30 s at 95  C; 35 cycles of 15 s at 95  C, 30 s at 58  C, and 31 s at 72  C. The nirS gene PCR amplification using primers (nirS 1f and nirS 3r) (Braker et al., 1998) consisted of 30 s at 95  C; 30 cycles of 15 s at 95  C, 20 s at 60  C, and 31 s at 72  C. The PCR condition for nirK gene included 30 s at 95  C; 30 cycles of 15 s at 95  C, 30 s at 58  C, and 31 s at 72  C. Lastly, Primers (nosZ 2f and nosZ 2r) designed by Henry et al. (2006) were used to determine the amount of nosZ gene. The program used for amplification was 30 s at 95  C; 30 cycles of 15 s at 95  C, 30 s at 60  C, and 31 s at 72  C. All experiments were performed in duplicate per sample and all PCR runs included control reactions without the template. The specificity of each PCR assay was confirmed using both melting curve analysis and agarose gel electrophoresis. Gene copy numbers were calculated by comparing threshold cycles obtained in each PCR run with those of known standard DNA concentrations. Standards were prepared using serially diluted plasmid DNA with 103e108 gene copies/mL. Standard curves for the 16S rDNA, AOB, Nitrospira spp., Nitrobacter spp., narG, nirS, nirK, and nosZ assays were generated by plotting the threshold cycle values versus log10 of the gene copy numbers. The amplification efficiency (E ) was estimated using the slope of the standard curve through the following formula: E ¼ (101/slope)  1. The efficiency of PCR amplification for each gene was between 90% and 100%.

2.7.

Cloning and sequencing

Prior to cloning, the amplified unlabeled AOB 16S rRNA genes fragments were purified using the PCR purification kit (SolGent, Korea). Purified PCR products were ligated into pGEM-T Easy cloning vectors (Promega, USA) and transformed into competent Escherichia coli One-Shot Mach 1-T1 (Invitrogen, USA), as described in the manufacturer’s protocol. Transformants were selected by ampicillin resistance and blueewhite screening was performed to identify clones with inserts. Seventy-four white colonies were selected and cultivated. Primers T7 and SP6 were used to perform colony PCR and to verify that the insert size was correct. Following PCR confirmation of insert size, the amplified inserts were run on 2% (wt/vol) agarose gels. The samples containing inserts of the estimated size were used for subsequent sequencing. The 16S rRNA gene inserts were sequenced through an ABI 3130XL Genetic Analyzer (Applied Biosystems, Foster City, USA) at the SolGent Company (Korea). Database homology searches for these sequences were performed using the BLAST program in the National Center for Biotechnology Information (NCBI) database.

3.

Results and discussion

3.1. Functional performance of the full-scale activated sludge process The process performance of the full-scale wastewater treatment system during the study is presented in Fig. 2. Although the pre-denitrification activated sludge process is simple, various microbial reactions occur sequentially under both

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 8 5 e5 7 9 5

5789

Fig. 2 e Variation of influent- and effluent-concentrations of (a) COD; (b) TOC; (c) phenol; (d) SCNL; (e) ammonia; (f) TN in the full-scale process during the study period (solid circles: influent, open circles: effluent, solid triangles: removal efficiency).

anoxic and aerobic conditions. During the study, 2025e3150 mg/L of COD was fed into the anoxic tank where 75e80% of it was subsequentially removed. Residual COD flowed into the aerobic tank and then further removed, resulting in a COD removal efficiency of 80e88%. The removal pattern of TOC was similar to that of COD. There was no further degradation of any residual COD or TOC in the nitrification tank. This implies that the remaining residual organic carbon was non-biodegradable. Phenol flowed into the fullscale process in the range of 268e715 mg/L and increased to be 75% higher than normal concentrations of about 400 mg/L in the ninth week (Fig. 2c). Most phenol was first degraded in the anoxic tank, with remaining phenol being almost completely degraded in the aerobic tank. Transformation of organic carbons such as phenol to inorganic carbon took place in the anoxic tank due to denitrification and fermentation reactions by various heterotorphes (Kim et al., 2009). The inorganic carbon was consumed by autotrophic nitrifiers and thiocyanate-degrading bacteria under aerobic conditions. Almost all SCN, which varied from 190 to 297 mg/L in the influent, was removed in the aerobic tank (Fig. 2d). Various autotrophic bacteria in activated sludge are known to degrade SCN under aerobic conditions ðSCN þ 2O2 þ 2H2 O/ 2 NHþ 4 þ SO4 þ CO2 Þ (Lee et al., 2008). Total cyanides in the range of 12.4e17.2 mg/L flowed into the anoxic tank and 1.7e3.8 mg/L of it flowed into the aerobic tank (data not shown). Additional degradation of cyanides did not take place under aerobic conditions. The incomplete removal of total cyanides was due to the existence of ferric cyanide ðFeðCNÞ3 6 Þ, which is known to undergo very slow biodegradation (Kim et al., 2008). The main role of the nitrification tank was to convert the ammonium ion into nitrite and/or nitrate. The average NHþ 4 N

concentration range in the influent was about 76e90 mg-N/L, except for shock loading of ammonia (Fig. 2e). Some amounts of ammonia were newly generated due to cell lysis and degradation of SCN in the anoxic and aerobic tanks, respectively. During stable operation, the final effluent concentration of NHþ 4  N in the nitrification tank remained less than 10 mg-N/L, while the NO 2  N concentration was in the range of 17e20 mgN/L. Detected NO 3  N concentration values were less than 3.0 mg-N/L. Denitrification was promoted by recycling the nitrite/nitrate formed via nitrification back to the anoxic tank. Complete denitrification was consistently achieved in the anoxic tank until 11 weeks regardless of ammonia and phenol shock loading. Excluding this shock loading of ammonia, the average TN in the influent was 190 mg-N/L (Fig. 2f). The effluent TN concentration remained below 40 mg-N/L, corresponding to an average TN removal efficiency of 85% (note that regulations stipulate a TN concentration less than 60 mg-N/L for discharge into surface water in South Korea). Also, the effluent TN concentration of less than 40 mg-N/L indicated that the system was functionally stable throughout the study.

3.2. Influence of operational parameters on nitrogen removal performance of the full-scale activated sludge process Nitrification of the full-scale activated sludge process treating high strength pollutants targeted in our study has been unstable during the past several summers due to an abnormal influx of pollutants such as phenol, SCN, ammonia and cyanide. Therefore, responsive system operations have been proposed to accomplish discharge level regulations under various situations.

5790

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 8 5 e5 7 9 5

Fig. 3 shows the effect of operational parameters on the effluent nitrogen concentration of the full-scale activated sludge process. In the first week, although TN concentration remained below the legal discharge level (i.e., 60 mg-N/L), NO 2  N concentration, a by-product of nitrification, was only 7 mg-N/L (Fig. 3e). This indicated inadequate nitrification performance. The nitrification tank was operated at an MLSS concentration of about 1800 mg/L at 15 days of SRT (Fig. 3c), which was lower than the 2500e3000 mg/L MLSS level of general activated sludge processes treating cokes wastewater (Manekar et al., 2011). To increase the nitrifying biomass having a slow growth rate, the SRT was lengthened from 15 days to 20 days by reducing excess sludge removal. This resulted in an increase in MLSS concentration. For 2 weeks, the MLSS concentration in the nitrification tank increased

Fig. 3 e Variation of operational parameters: (a) SRT and internal recycling ratio; (b) MLSS concentration and MLVSS/ MLSS ratio in the nitrification tank; (c) MLSS concentration and MLVSS/MLSS ratio in the anoxic tank; (def) effluentconcentrations of ammonia, nitrite nitrate, TN in the nitrification tank during the study period.

from 1800 mg/L to 2000 mg/L (Fig. 3c). In the final effluent of the nitrification tank, the ammonia concentration was below 20 mg-N/L while the NO 2  N concentration gradually increased to 18 mg-N/L (Fig. 3d and e). TN concentration was consistently maintained under the legal discharge level of 60 mg-N/L. It is believed that this rather long SRT may lead to an increase in the slow growing nitrifying bacteria and support their dominance, resulting in better nitrification performance (Teck et al., 2009). At the beginning of the fourth week, however, an abrupt change in the effluent nitrogen concentration of the nitrification tank occurred (Fig. 3f). The ammonia concentration sharply increased to over 40 mg-N/L and the TN concentration exceeded its legal discharge level of 60 mg-N/L. Initially, it was doubted that a decrease in nitrification activity had occurred as a result of inhibition by cyanide or phenol. However, analysis results of the influent and effluent revealed that the shock loading of ammonia was to blame. In the fourth week, the ammonia concentration in the influent sharply increased to 140 mg-N/L, a level more than twice normal loading (data not shown). Meanwhile, the NO 2  N concentration in the effluent did not decrease, implying that the nitrification performance had not been inhibited. For high volumetric ammonia removal, the process was controlled as nearly infinite SRT without removal of excess sludge, resulting in an accumulation of MLSS in the system for 1 week. It is known that high volumetric loading can be achieved by maintaining a high MLSS (Rittmann and McCarty, 2001). At the end of 4 weeks, the MLSS concentration in the system significantly increased to 2800 mg/L (Fig. 3d). This led to a gradual decrease in both the ammonia and TN concentrations in the final effluent. In addition, as soon as the usual nitrogen concentration in the influent flowed into the system after the fifth week, the effluent nitrogen removal efficiency improved more than previously. The ammonia concentration was controlled at less than 10 mg-N/L and the TN concentration was maintained much lower than the legal level. The NO 2 N concentration of about 17 mg-N/L was produced by nitrification. Complete denitrification was consistently achieved in the anoxic tank. As a result, the increased MLSS concentration in the system by long SRT ensured stable nitrification of the full-scale activated sludge process during ammonia shock loading. These results implied that the selection of an adequate biomass concentration by SRT control can be vital in achieving the desired efficiency of the process. However, full-scale process performance took a sudden turn for the worse in the ninth week (Fig. 3f). Ammonia concentration increased to 87 mg-N/L while NO 2 N concentration decreased to 3.5 mg-N/L in the effluent of the nitrification tank (Fig. 3d and e). The decrease of nitrite concentration reflected incomplete nitrification in the nitrification tank. Since nitrification was significantly inhibited, TN concentration abruptly increased from 30 to 120 mg-N/L. The analysis of variations in the pollutants’ concentrations in the influent indicated that shock loading of phenol was one of the causes for the nitrification failure (Fig. 2c). As the influent concentration of phenol increased from 300e400 mg/L to 715 mg/L, more phenol flowed into the aerobic tank. The inhibitory effect of phenol on nitrification is well known. Previous research has shown that periods of nitrification

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 8 5 e5 7 9 5

failure coincide with increased phenol concentration in the wastewater (Figuerola and Erijman, 2010). To prevent deterioration of the nitrification performance from phenol inhibition, the internal recycling ratio was controlled from 2.0 (600 m3/h) to 0.5 (150 m3/h) and long SRT was consistently maintained at 20 days. By reducing the internal recycling ratio, phenol removal performance was enhanced in the anoxic tank, resulting in inflow of less phenol into the nitrification tank. In addition, the decreased recycling ratio increased the retention time of both the biomass and effluent in the nitrification tank. The increased retention time, in turn, provided both adequate nitrification reaction time and contact stabilization, allowing nitrifiers to rebound from any prior inhibition. In addition, it is known that high internal recycling values have a negative effect on the maximum specific growth rate of nitrifiers (Jimenez et al., 2011). Meanwhile, the long SRT prevented the nitrifying biomass from washing out of the system. As a result, as soon as the phenol concentration in the influent decreased to below 500 mg/L after the tenth week (Fig. 2c), the TN removal efficiency improved and a discharge level less than 50 mg-N/L could be achieved (Fig. 3f). The effluent ammonia concentration in the tenth week quickly decreased from 87 mg-N/L to 25 mg-N/L and in the eleventh week was lower than 10 mg-N/L achieving the same functional stability period as between the fifth and eighth week (Fig. 3d). NO 2  N production also recovered from concentration slightly phenol inhibition and NO 3 N increased to 4.1 mg-N/L (Fig. 3e). This meant that nitrification performance had totally recovered. In addition, the increased retention time may be helpful in allowing NOB to generate nitrate ion from nitrite substrate. At the end of the twelfth week, although the effluent ammonia and TN concentrations were stably maintained, NO 2  N production began to decrease from 18 to 9 mg-N/L (Fig. 3e). MLSS concentration climbed to 3400e3600 mg/L in the system in spite of consistent SRT. These results indicated that an increase of organic matter may cause proliferation of heterotrophic microorganisms in the nitrification tank, increasing uptake ammonia for their growth. Contrary to the increase in MLSS concentration, MLVSS/MLSS ratio in the nitrification tank quickly fell to 0.65, resulting in a decrease in nitrification efficiency (Fig. 3b and d). In the anoxic tank, the MLVSS/MLSS ratio sharply decreased to 0.7 with NO 2 N concentration of 6.0 mg/L, resulting in a decrease in denitrification efficiency (data not shown). This decreased VSS concentration may result from an imbalance between feed and biomass, since influent concentrations like organic and nitrogen matter gradually decreased after the ninth week, corresponding to an increase of MLSS concentration by consistently long SRT. Consequently, these results in the twelfth week implied that nitrogen removal performance of the full-scale process in the future could be vulnerable to environmental factors.

3.3.

ammonia, nitrite and nitrate concentrations. As illustrated in Fig. 4, nitrifying bacteria activity was affected by process conditions. Until the third week, increased SRT did not influence the specific nitrification rate, maintaining at about 5.0 mg-N/g-VSS$h. In the fourth week, however, the specific nitrification rate decreased to 3.7 mg-N/g-VSS$h due to a sudden increase in the MLSS concentration. Between the fifth and eighth week, the specific nitrification rate gradually increased from 4.3 to 6.3 mg-N/g-VSS$h, irrespective of a small increase in MLSS concentration. In this period, nitrogen removal efficiency in the full-scale system achieved almost 90%. In the ninth week, nitrification performance of the fullscale process sharply decreased; on the other hand, the batch test revealed only slightly decreased nitrification activity, decreasing from 6.3 to 5.9 mg-N/g-VSS$h. This indicated that there was discordance between the potential rate (batch test) and the actual rate (full-scale performance) regarding nitrification performance. One may conclude that, although nitrifying bacteria possess good potential activity, difficult to identify environmental factors influence their actual performance in the full-scale process, leading to decreased nitrification. During the study, the specific denitrification rate pattern was similar to variations in nitrification activity. The specific denitrification rate remained in the range of 2.8e4.8 mg-N/gVSS h. Like the variation of the nitrification rate, the specific denitrification rate decreased in the fourth week, due to a considerable increase in the MLSS concentration, then gradually increased to 4.8 mg-N/g-VSS h. Even in the ninth week when phenol shock loading incidentally occurred, the specific denitrification rate was not affected and remained consistent. Contrary to the nitrification performance, denitrification in the full-scale process had a stable rate - similar to its batch tests. This implied that denitrification may be less affected than nitrification by environmental factors. Meanwhile, the MLVSS/MLSS ratio in both the anoxic and nitrification tanks gradually decreased. As shown in Fig. 4, however, both the nitrification and denitrification rates in the batch tests were consistent, regardless of the nitrogen removal performance of the full-scale process. As a result, the decrease of MLVSS did not lead to any loss of microorganism activity.

Microbial activity

Batch experiments to observe the variations of nitrification and denitrification activities in the activated sludge process were carried out. The specific nitrification and denitrification rate in each batch test was analyzed through variations of

5791

Fig. 4 e Variation of denitrification and nitrification activities by batch test during the study period.

5792

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 8 5 e5 7 9 5

This may result from a consistent percentage of active bacterial populations among total microorganisms.

3.4.

T-RFLP analysis of AOB and NOB populations

We identified the nitrifying bacterial communities present in an activated sludge process using T-RFLP designed for the identification of AOB and NOB with terminal fragment (TF) lengths (Regan et al., 2002). Fig. 5 shows electropherograms of AOB, Nitrobacter-specific NOB, and Nitrospira-specific NOB present in the nitrification tank of the full-scale process, respectively. As shown in Fig. 5a, AOB-targeted T-RFLP allowed us to differentiate between AOB groups. All samples showed a peak at 164 bp, a signature of Nitrosomonas europaea/ eutropha and Nitrosomonas marina lineage (Table 2). Because the influents originate from industrial wastewater, marine AOB species need not be considered. Besides the major peak at 164 bp, we detected another peak at 273 bp, representing the potential presence of N. europaea/eutropha, Nitrosomonas oligotropha, Nitrosomonas cryotolerans, or Nitrosomonas communis lineage (Table 2). To better understand the AOB community present in the process, AOB 16S rRNA gene based cloning and sequencing was performed using the AOB-target primer (Nso1225r and Eub338f) without fluorescent dye (Table 1). Sixty-eight of total 74 AOB clones from the reactor were closely associated with N. europaea in the N. europaea/eutropha lineage and Nitrosomonas nitrosa in the N. communis lineage, but the AOB clone related to the Nitrosospira lineage was not detected. As a result, based on the 16S rRNA gene sequences, microorganisms corresponding to the peaks at 164 bp and 273 bp could be

identified as N. europaea and N. nitrosa, respectively (Fig. 5a). Thus, the high peak at 164 bp implies the dominance of the N. europaea within AOB in this wastewater treatment system, irrespective of variations in the full-scale process performance; N. nitrosa had a minor presence. N. europaea has been widely observed in WWTPs (Lydmark et al., 2007; Siripong and Rittmann, 2007), while N. nitrosa has previously been detected on occasion in activated sludge treating industrial wastewater (Layton et al., 2005). Despite variations in environmental conditions such as MLSS concentration, internal recycling ratio and influent characteristics for the AOB, the total selective pressure in the full-scale process has been insufficient to induce a population shift (Hallin et al., 2005; Kim et al., 2011b). This finding also indicated that low diversity of AOB populations may be conducive to nitrification failure. Based on Nitrobacter-specific T-RFLP, Fig. 5b shows a prominent peak at 137 bp, characteristic of the Nitrobacter species. Meanwhile, in the fifth week, the peak at 214 bp was dominant along with the 137 bp peak, but disappeared after the ninth week. This Nitrobacter sp. corresponding to 214 bp peak may be affected by phenol. We also found TF sizes at 92, 162, 245 and 273 in the samples. These unexpected peaks could be the result of incomplete digestion, uncharacterized Nitrobacter species or imperfectly matched primer (Siripong and Rittmann, 2007). The results of Nitrospira-specific T-RFLP showed four dominant peaks at 135, 192, 272 and 334 bp (Fig. 5c). The peak at 135, 192 and 272 corresponds to several Nitrospira clones in the database. The 334 TF belongs to one of the Nitrospira moscoviensis strains (Siripong and Rittmann, 2007). The Nitrospira species corresponding to the peak at 272 bp was a consistently dominant population, while

Fig. 5 e T-RFLP profiles of (a) AOB, (b) Nitrobacter, (c) Nitrospira in the nitrification tank during the study period.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 8 5 e5 7 9 5

Table 2 e Expected TF sizes and their corresponding AOB and NOB groups based on T-RFLP of 16S rDNA gene (Siripong and Rittmann, 2007). TF size (bp) 164e166, 276 276 276 166 276 105e107 141, 196 133, 194, 265e267, 277,333

Nitrifying bacteria group Nitrosomonas europaea/eutropha lineage Nitrosomonas oligotropha lineage Nitrosomonas cryotolerans lineage Nitrosomonas marina lineage Nitrosomonas communis lineage Nitrosospira lineage Nitrobacter species Nitrospira species

N. moscoviensis sp. corresponding to the peak at 334 bp was found to be predominant when nitrification performance of the full-scale process began to stabilize in the fifth week. Despite production of small amounts of nitrate by NOB in the nitrification tank, harsh environmental conditions for NOB may stimulate diversity of the NOB species.

3.5. qPCR analysis of nitrifying and denitrifying bacteria populations Fig. 6a shows the changes in the 16S rRNA gene copies for the total bacteria, AOB, Nitrobacter, and Nitrospira, quantified using qPCR assays in the nitrification tank of the full-scale process. In all samples the total bacterial population in the nitrification tank ranged from 4.8  1012 to 3.8  1013 copies/L. These values are the same order of magnitude as those obtained from activated sludge samples of WWTPs (Limpiyakorn et al., 2005). As the MLSS concentration increased as a result of SRT control, the total bacterial population increased to 3.8  1013 copies/L in the seventh week, indicating an eightfold increase compared to that in the first week. The concentration of AOB identified using the AOB 16S rDNA assay gradually increased and the variation was not large, compared to that of the NOB population. In the tenth week, an approximately 9-fold increase was observed in the

5793

number of AOB 16S copies/L over the first week. This result indicated that newly controlled SRT and the internal recycling ratio may ensure continued AOB population. In addition, such a gradual increase in the AOB population may affect the increase in nitrification activity observed in the batch test. In the ninth week when the nitrification performance of the fullscale process was severely inhibited, there was no observed decrease in the AOB number. However, any change in the AOB number generally coincided with the variation of NO 2 N concentration produced in the nitrification tank. Meanwhile, the percentages of the AOB within the total bacteria varied from 1.07 to 3.29% in the nitrification tank. This result is similar to the values for activated sludge samples obtained from systems treating industrial wastewater (0.01e9.3%) (Layton et al., 2005). But we could not identify any correlation with the nitrification activity. We observed coexisting Nitrospira and Nitrobacter genera for NOB. The Nitrospira and Nitrobacter populations in the initial operating condition were similar, at 2.0  109 copies/L and 2.8  109 copies/L, respectively. However, a shift to Nitrobacter sp. in the NOB community was observed throughout the study. The 16S rRNA gene concentration of the Nitrobacter increased to a range of 2.8  109 to 4.8  1010 copies/ L, and the percentages of the Nitrobacter population within the total bacteria also sharply increased from 0.03 to 0.16% in the nitrification tank, as the MLSS concentration increased and more nitrite was produced. Finally, the Nitrobacter populations in the nitrifying system were consistently higher than the Nitrospira populations throughout the study. On the other hand, the number of Nitrospira gradually decreased from 2.0  109 to 1.1  109 copies/L, until the third week. When high concentration of ammonia in the influent flowed into the activated sludge process, a 5-fold increase was observed in the number of Nitrospira 16S copies/L along with an increase in the number of Nitrobacter sp. However, in the ninth week when high concentrations of phenol flowed into the system, a sharp decrease of Nitrospira population was observed. The percentage of the Nitrospira population among all bacteria shrank to 0.01%. Previous research has reported that Nitrospira

Fig. 6 e Changes in copies per liter of (a) the total bacteria, AOB, Nitrobacter, and Nitrospira, (b) the total bacteria, narG, nirS, nirK, nosZ in the full-scale activated sludge process during the study period.

5794

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 8 5 e5 7 9 5

were far more sensitive to toxic compounds than Nitrobacter (Blackburne et al., 2007; Kim et al., 2011a). Meanwhile, despite coexisting Nitrospira and Nitrobacter genera for NOB (Fig. 6a) and generation of NO 3  N as the final product through nitrification activity in the batch test (data not shown), low concentrations of NO 3  N were produced in the nitrification tank of the full-scale process (Fig. 2). This may reflect the short retention time necessary for NOB to react with nitrite ions in the full-scale process. Meanwhile, the abundance of narG, nirS, nirK and nosZ genes of denitrifying bacteria was investigated during the study in the activated sludge process. The narG, nirS, nirK and nosZ target molecules were less abundant than the 16S rDNA gene copies for the total bacteria: total bacteria ranged from 9.9  1012 to 3.9  1013 copies/L; narG ranged from 2.0  109 to 9.2  109 copies/L; nirS ranged from 1.3  1012 to 1.7  1013 copies/L; nirK ranged from 1.1  1010 to 1.0  1011 copies/L; nosZ ranged from 2.1  1011 to 2.4  1012 copies/L (Fig. 6b). In the activated sludge process treating industrial wastewater, gene copy numbers per liter of the nirS gene exceeded those of the narG and nosZ genes for all sampling points. This trend implies that there is a greater abundance of genes for the nitrite reducing genes than for the nitrate and nitrous oxide reducing genes. Meanwhile, for nirS, the copy numbers of genes detected were much higher than those for nirK at all sampling points. It is known that more taxonomically diverse nirK denitrifiers are more sensitive to environmental changes than the nirS denitrifiers; however, the latter are more abundant (Yoshie et al., 2004). To evaluate the ratio of denitrifiers relative to total bacteria, the percentages of denitrification genes in proportion to 16S rDNA were calculated, resulting in proportions of approximately 0.02%, 26.9%, 0.30%, and 4.77% for narG, nirS, nirK, and nosZ genes, respectively. The maximum amount of nirS relative to 16S rDNA was 43%, confirming the high proportion of denitrifiers to total bacteria in this activated sludge process. On the other hand, the smallest amount of the narG gene, the nitrate reducing functional gene, in the anoxic tank for all samplings may result from low concentrations of NO 3  N in the nitrified effluent. Lastly, in the twelfth week, both nitrifying and denitrifying populations decreased along with a decrease in the MLVSS/ MLSS ratio, resulting in a slight decrease in both nitrification and denitrification efficiencies of the full-scale process. However, in the batch tests both the nitrification and denitrification rates remained steady. This may result from little loss of active bacterial populations among total microorganisms. Also, in actual environment conditions, there may be more factors inhibiting the activity of microorganisms in the full-scale process than in the batch test. Consequently, it was very difficult to identify any relationships between the bacterial populations, their activity and the process performance in the fullscale process. However, to prevent failure of the process performance, it is important to monitor any sudden decrease in the populations of important bacteria such as nitrifiers.

4.

Conclusions

In a full-scale activated sludge process, SRT and internal recycling ratio controls were selected as the main operating

parameters to improve TN removal efficiency. Increased biomass concentration via SRT control enhanced TN removal. In addition, decreasing the internal recycling ratio restored nitrification activity which had been inhibited by phenol shock loading. These results indicate that application of specific operational strategies can change the physiological state of the activated sludge process’s bacterial populations (Hallin et al., 2005). Therefore, proper operational strategies should be tailored to accommodate the possibility of erratic changes in the composition of the influent via consistent monitoring of its components to achieve and maintain a stable process.

Acknowledgments The authors thank David Nielsen for assistance during this work. This research was supported by WCU (World Class University) program through the National Research Foundation of Korea funded by the Ministry of Education, Science and Technology (R31-30005) and the Advanced Biomass R&D Center (ABC) of Korea Grant funded by the Ministry of Education, Science and Technology (ABC-2010-0029800). Also, this work was financially supported by the second phase of the Brain Korea 21 Program in 2011 and Korea Ministry of Environment (MOE) as ’Human resource development Project for Energy from Waste & Recycling’.

references

Amann, R.I., Binder, R.J., Olson, S., Chisholm, S.W., Devereux, R., Stahl, D.A., 1990. Combination of 16S ribosomal RNA targeted oligonucleotide probes with flow cytometry for analyzing mixed microbial populations. Applied and Environmental Microbiology 56, 1919e1925. APHA, 1998. Standard Methods for the Examination of Water and Wastewater. APHA, AWWA, WPCF, twentieth ed. American Public Health Association, Washington, DC, USA. Blackburne, R., Vadivelu, V.M., Yuan, Z.G., Keller, J., 2007. Kinetic characterisation of an enriched Nitrospira culture with comparison to Nitrobacter. Water Research 41, 3033e3042. Braker, G., Fesefeldt, A., Witzel, K.P., 1998. Development of PCR primer systems for amplification of nitrite reductase genes (nirK and nirS ) to detect denitrifying bacteria in environmental samples. Applied and Environmental Microbiology 64, 3769e3775. Dionisi, H.M., Layton, A.C., Harms, G., Gregory, I.R., Robinson, K.G. , Sayler, G.S., 2002. Quantification of Nitrosomonas oligotropha-like ammonia-oxidizing bacteria and Nitrospira spp. from full-scale wastewater treatment plants by competitive PCR. Applied and Environmental Microbiology 68, 245e253. Eighmy, T.T., Bishop, P.L., 1989. Distribution and role of bacterial nitrifying populations in nitrogen removal in aquatic treatment systems. Water Research 23, 947e955. Ferris, M.J., Muyzer, G., Ward, D.M., 1996. Denaturing gradient gel electrophoresis profiles of 16S rRNA-defined populations inhabiting a hot spring microbial mat community. Applied and Environmental Microbiology 62, 340e346. Figuerola, E.L., Erijman, L., 2010. Diversity of nitrifying bacteria in a full-scale petroleum refinery wastewater treatment plant

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 8 5 e5 7 9 5

experiencing unstable nitrification. Journal of Hazardous Matererials 181, 281e288. Graham, D.W., Knapp, C.W., Van Vleck, E.S., Bloor, K., Lane, T., Graham, C.E., 2007. Experimental demonstration of chaotic instability in biological nitrification. ISME Journal 1, 385e394. Hallin, S., Lydmark, P., Kokalj, S., Hermansson, M., So¨rensson, F., Jarvis, A., Lindgren, P.E., 2005. Community survey of ammonia-oxidizing bacteria in full-scale activated sludge processes with different solids retention time. Journal of Applied Microbiology 99, 629e640. Harms, G., Layton, A.C., Dionisi, H.M., Gregory, I.R., Garrett, V.M., Hawkins, S.A., Robinson, K.G., Sayler, G.S., 2003. Real-time PCR quantification of nitrifying bacteria in a municipal wastewater treatment plant. Environmental Science and Technology 37, 343e351. Henry, S., Baudouin, E., Lo´pez-Gutie´rrez, J.C., Martin-Laurent, F., Brauman, A., Philippot, L., 2004. Quantification of denitrifying bacteria in soils by nirK gene targeted real-time PCR. Journal of Microbiological Methods 59, 327e335. Henry, S., Bru, D., Stres, B., Hallet, S., Philippot, L., 2006. Quantitative detection of the nosZ gene, encoding nitrous oxide reductase, and comparison of the abundances of 16S rRNA, narG, nirK, and nosZ genes in soils. Applied and Environmental Microbiology 72, 5181e5189. Hermansson, A., Lindgren, P.E., 2001. Quantification of ammoniaoxidizing bacteria in arable soil by real time PCR. Applied and Environmental Microbiology 67, 972e976. Jimenez, J., Melcer, H., Parker, D., Bratby, J., 2011. The effect of degree of recycle on the nitrifier growth rate. Water Environment Research 83, 26e35. Juliastuti, S.R., Baeyens, J., Creemers, C., 2003. Inhibition of nitrification by heavy metals and organic compounds: the ISO 9509 test. Environmental Engineering Science 20, 70e90. Kane, M.D., Poulsen, L.K., Stahl, D.A., 1993. Monitoring the enrichment and isolation of sulfate-reducing bacteria by using oligonucleotide hybridization probes designed from environmentally derived 16s ribosomal RNA sequences. Applied and Environmental Microbiology 59, 682e686. Kim, Y.M., Park, D., Lee, D.S., Park, J.M., 2007. Instability of biological nitrogen removal in a cokes wastewater treatment facility during summer. Journal of Hazardous Matererials 141, 27e32. Kim, Y.M., Park, D., Lee, D.S., Park, J.M., 2008. Inhibitory effects of toxic compounds on nitrification process for cokes wastewater treatment. Journal of Hazardous Matererials 152, 915e921. Kim, Y.M., Park, D., Lee, D.S., Jung, K.A., Park, J.M., 2009. Sudden failure of biological nitrogen and carbon removal in the fullscale pre-denitrification process treating cokes wastewater. Bioresource Technology 100, 4340e4347. Kim, Y.M., Lee, D.S., Park, C., Park, D., Park, J.M., 2011a. Effects of free cyanide on microbial communities and biological carbon and nitrogen removal performance in the industrial activated sludge process. Water Research 45, 1267e1279. Kim, Y.M., Cho, H.U., Lee, D.S., Park, C., Park, D., Park, J.M., 2011b. Response of nitrifying bacterial communities to the increased thiocyanate concentration in pre-denitrification process. Bioresource Technology 102, 913e922.

5795

Layton, A.C., Dionisi, H., Kuo, H.W., Robinson, K.G., Garrett, V.M., Meyers, A., Sayler, G.S., 2005. Emergence of competitive dominant ammonia-oxidizing bacterial populations in a fullscale industrial wastewater treatment plant. Applied and Environmental Microbiology 71, 1105e1108. Lee, C., Kim, J., Do, H., Hwang, S., 2008. Monitoring thiocyanatedegrading microbial community in relation to changes in process performance in mixed culture systems near washout. Water Research 42, 1254e1262. Limpiyakorn, T., Shiohara, Y., Kurisu, F., Yagi, O., 2005. Communities of ammonia-oxidizing bacteria in activated sludge of various sewage treatment plants in Tokyo. FEMS Microbiology Ecology 54, 205e217. Lin, C., Stahl, D.A., 1995. Comparative analyses reveal a highly conserved endoglucanase in the cellulolytic genus Fibrobacter. Journal of Bacteriology 177, 2543e2549. Lo´pez-Gutie´rrez, J.C., Henry, S., Hallet, S., Martin-Laurent, F., Catrou, G., Philippot, L., 2004. Quantification of a novel group of nitrate-reducing bacteria in the environment by real-time PCR. Journal of Microbiological Methods 57, 399e407. Lydmark, P., Almstrand, R., Samuelsson, K., Mattsson, A., So¨rensson, F., Lindgren, P.E., Hermansson, M., 2007. Effects of environmental condition on the nitrifying population dynamics in a pilot wastewater treatment plant. Environmental Microbiology 9, 2220e2233. Manekar, P., Biswas, R., Karthik, M., Nandy, T., 2011. Novel two stage bio-oxidation and chlorination process for high strength hazardous coal carbonization effluent. Journal of Hazardous Matererials. doi:10.1016/j.jhazmat.2011.02.006. Mertoglu, B., Semerci, N., Guler, N., Calli, B., Cecen, B., Saatci, A.M. , 2008. Monitoring of population shift in an enrich nitrifying system under gradually increased cadmium loading. Journal of Hazardous Matererials 160, 495e501. Mobarry, B.K., Wagner, M., Urbain, V., Rittmann, B.E., Stahl, D.A., 1996. Phylogenetic probes for analyzing abundance and spatial organization of nitrifying bacteria. Applied and Environmental Microbiology 62, 2156e2162. Regan, J.M., Harrington, G.W., Noguera, D.R., 2002. Ammonia- and nitrite-oxidizing bacterial communities in a pilot-scale chloraminated drinking water distribution system. Applied and Environmental Microbiology 68, 73e81. Rittmann, B.E., McCarty, P.L., 2001. Environmental Biotechnology: Principles and Applicatons. McGraw-Hill Science. Siripong, S., Rittmann, B.E., 2007. Diversity study of nitrifying bacteria in full-scale municipal wastewater treatment plants. Water Research 41 (5), 1110e1120. Teck, H.C., Loong, K.S., Sun, D.D., Leckie, J.O., 2009. Influence of a prolonged solid retention time environment on nitrification/ denitrification and sludge production in a submerged memebrane bioreactor. Desalination 245, 28e43. Wagner, M., Rath, G., Amann, R., Koops, H.P., Schleifer, K.H., 1995. In situ identification of ammonia-oxidizing bacteria. Systematic and Applied Microbiology 18, 251e264. Yoshie, S., Noda, N., Tsuneda, S., Hirata, A., Inamori, Y., 2004. Salinity decreases nitrite reductase gene diversity in denitrifying bacteria of wastewater treatment systems. Applied and Environmental Microbiology 70, 3152e3157.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 9 6 e5 8 0 4

Available online at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Reduced microbial attachment by D-amino acid-inhibited AI-2 and EPS production Huijuan Xu a, Yu Liu a,b,* a

Division of Environmental and Water Resources Engineering, School of Civil and Environmental Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798, Singapore b Advanced Environmental Biotechnology Centre, Nanyang Environment & Water Research Institute, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798, Singapore

article info

abstract

Article history:

This study investigated the effect of D-tyrosine on microbial attachment to hydrophilic D-tyrosine

did not

Received 21 March 2011

glass and hydrophobic polypropylene surfaces. Results showed that

Received in revised form

influence microbial growth, ATP and substrate utilization, but significantly inhibited the

22 August 2011

synthesis of autoinducer-2 (AI-2), eDNA and extracellular polysaccharides and proteins,

Accepted 29 August 2011

and subsequently reduced microbial attachment onto glass and polypropylene surfaces

Available online 3 September 2011

was observed. It was shown that D-amino acid would be a non-toxic agent for control of microbial attachment.

Keywords:

ª 2011 Elsevier Ltd. All rights reserved.

Microbial attachment D-amino

acid

Autoinducer-2 ATP eDNA EPS

1.

Introduction

It had been considered that D-amino acids are excluded from living systems except for D-amino acids in the cell wall of microorganisms. D-amino acids have been discovered in many physiological processes. The best described of D-amino acids may be their involvement in the formation of the peptidoglycan. Both the thick cell wall of Gram-positive bacteria and much thinner cell wall of Gram-negative bacteria consist of peptidoglycan which contain D-amino acids. Besides components of bacterial cell wall, D-amino acid have been known to regulate bacterial germination and to be incorporated into peptides (Wood et al., 2011). The bacteria begin to synthesize

those D-amino acids in stationary phase, which may regulate the chemistry of the cell wall through slow production of peptidoglycan that is crucial for cell wall (Lam et al., 2009). It has been reported that many bacteria would produce various D-amino acids just before biofilm disassembly and the release of rapid diffused small molecule D-amino acid could be a signal to coordinate the whole population action to different environment (Kolodkin-Gal et al., 2010). During biofilm formation, microorganisms need first attach onto a solid surface, followed by secretion of extracellular polymeric substances (EPS) (Flemming and Wingender, 2010), whereas other factors have also been reported to be essentially involved in biofilm development. For example, both Gram-

* Corresponding author. Division of Environmental and Water Resources Engineering, School of Civil and Environmental Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798, Singapore. Tel.: þ65 67 905 254; fax: þ65 67 910 676. E-mail address: [email protected] (Y. Liu). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.061

5797

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 9 6 e5 8 0 4

negative and Gram-positive bacteria could communicate through small signaling molecules to coordinate population behavior (Bassler, 1999). Autoinducer-2 (AI-2) is a speciesnonspecific signal molecule found in both Gram-negative and Gram-positive bacteria, and can influence a mixed-species biofilm formation between Porphyromonas gingivalis and Streptococcus gordonii, and it was found that production of AI-2 by either species was sufficient for inter-species communication and biofilm formation (McNab et al., 2003). ATP provides a pool of energy for most energy-consuming microbial activities, such as signaling molecules secretion, EPS synthesis, motility and flagella etc. In addition, the role of EPS in the formation of biofilms has been well documented (Flemming and Wingender, 2010). Although the role of D-amino acids in pure culture biofilm dispersal has been reported (Kolodkin-Gal et al., 2010), little is currently known about the effects of exogenous D-amino acids on the formation of mixed-culture biofilm and synthesis of cellular ATP, AI-2, eDNA and EPS, which are all essential for biofilm development on a solid surface. Furthermore, the present study focused more on the interaction between D-amino acid and AI-2-mdeiated cellular communication. For this purpose, a typical D-amino acid, D-tyrosine was used due to its potent activity (Kolodkin-Gal et al., 2010). Therefore, this study aimed to investigate how D-tyrosine could affect attachment of mixed-culture microorganisms onto hydrophobic PP and hydrophilic glass surfaces through determining changes in surface charge, ATP, AI-2, eDNA and EPS. It is expected that this study can offer an alternative approach for biological control of microbial attachment on various solid surfaces including membrane.

2.

Materials and methods

2.1.

Carriers for microbial attachment

Glass slides with the dimension of 24  50 mm (CEP, SPD Scientific, Singapore) and 24  50 mm polypropylene (PP) coupons (Kinary, Singapore) were used as biocarriers in microbial attachment experiments. The PP coupons were cleaned with detergent and rinsed thoroughly with distilled water, whereas the glass slides were cleaned by being soaked in 10% nitric acid for 24 h, and were then thoroughly rinsed with distilled water and dried. The hydrophobicity of the carrier surface was characterized by contact angle that was measured using a contact angle goniometer (dataphysics OCA 20, Filderstadt, Germany). Eight measurements were made on triplicate samples. The average water drop contact angle for clean glass slide was 16.9  0.5 and 99.3  2.2 for PP coupons, i.e. the PP coupons are highly hydrophobic, and hydrophilic for glass slides.

2.2.

Microbial attachment assay

Activated sludge microorganisms were taken from a local wastewater treatment plant and acclimated with a synthetic substrate for one month. The synthetic substrate consisted of 690 mg l1 of sodium acetate and 240 mg l1 ethanol as carbon source, 200 mg l1 NH4Cl, 60 mg l1 K2HPO4, 15 mg l1

CaCl2·2H2O, 12.5 mg l1 MgSO4·7H2O and 20 mg l1 FeSO4·7H2O (Liu et al., 2003). Experiments were designed to investigate the effect of D-tyrosine on microbial growth and attachment potentials. Thus, microorganisms with and without exposure to D-tyrosine for different times were used in 1-h static microbial attachment experiments conducted under the same conditions, as detailed below: (i) two series of batch experiments were conducted: one served as control free of D-tyrosine, while the other was added with 6 mg l1 of D-tyrosine (SigmaeAldrich, St. Louis, MO, USA); (ii) suspended biomass cultivated with and without exposure to D-tyrosine was collected at different exposure times of 1e4 h for 1-h microbial attachment assay and determination of surface charge, cellular ATP, AI-2, eDNA and EPS. The static microbial attachment was conducted in Petri dishes mounted with one PP coupon and one glass slide on the bottom. Suspended microorganisms harvested from the batch reactors at different exposure times were resuspended in 30 ml of 10 mM phosphate buffered saline (PBS) solution with 100 mg dry biomass l1 and were made contact with carriers for 1 h in Petri dishes. After attachment, carriers were gently rinsed three times with distilled water to remove loosely attached microorganisms. Fixed biomass was quantified in terms of TOC by a TOC analyzer (ASI-V, TOCVcsh, Shimadzu, Japan).

2.3.

Surface charge

The colloid titration method was used to determine surface charge of suspended microorganisms with and without exposure to D-tyrosine (Wilen et al., 2003). Polybrene (SigmaeAldrich) was used as positive colloidal reagent and polyvinyl sulfate potassium salt (PVSK) (SigmaeAldrich) as negative reagent. For titrating negatively charged suspended microorganisms, 5 ml of 0.001 N polybrene was added to the sample. The excess polybrene was back titrated with 0.0005 N PVSK using 100 ml of 0.1% toluidine blue (SigmaeAldrich) as the endpoint indicator. Titration was terminated when the color changed from blue to pink, indicating that electrical neutrality was reached. Equal volumes of polybrene in distilled water were used as blanks. The surface charge expressed as mill equivalents per gram of dry biomass can be determined from the equation given below. Charge meq g1 SS




¼

1000ðA  BÞN XV

(1)

Where A is the volume of PVSK added to the sample (ml), B is the volume of PVSK added to the blank (ml), N is the normality of PVSK solution used (0.0005 N), V is the volume of the sample (ml), X is the biomass concentration of the sample (g L1).

2.4.

Determination of cellular ATP

The cellular ATP were extracted from freshly collected biosamples according to the trichloroacetic acid (TCA) method (Chen and Leung, 2000) with some modifications. Five milliliter of the bacterial suspension was added into 5 ml of 5% TCA solution, and was homogenized with an ultrasonic homogenizer (Sonics & Materials, Newton, CT, USA) for 3 min. Aliquots of 0.5 ml homogenized suspension were diluted by ten times with Tris-Acetate-EDTA (TAE) buffer (Bio-Rad,

5798

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 9 6 e5 8 0 4

Singapore) to adjust the pH of about 7.75. The mixture was filtered through 0.2 mm syringe filters and the collected filtrate was stored at 20  C for further use. ATP concentration was determined according to firefly luciferin-luciferase bioluminescence method with the FLAA Adenosine 50 -triphosphate (ATP) Bioluminescent Assay Kit (SigmaeAldrich, St. Louis, MO, USA) as recommended in the instruction. The intensity of luminescence was measured by a TD-20/20 Luminometer (Turner Designs, Sunnyvale, CA, USA).

2.5.

Autoinducer-2 measurement

For determination of AI-2, 10 ml suspended microorganisms were collected and resuspended in fresh autoinducer bioassay (AB) medium (Surette and Bassler, 1998). The resuspended sample was well mixed and filtered through 0.2 mm syringe filter. The filtrate was collected and stored at 20  C. The cellfree culture supernatant was thawed before determination of AI-2 concentration. The amount of AI-2 was measured by Vibrio harveyi BB170 (ATCC BAA 1117) bioluminescence reporter assay (Rickard et al., 2008). The reporter strain V. harveyi BB170 (ATCC, Manassas, VA, USA) was cultured in fresh AB medium for 13e16 h with shaking at 30  C and then diluted 1:5000 with fresh AB medium. One hundred and eighty microliter of the diluted cells was added to the well of 96-well plate containing 20 ml cell-free supernatant to be tested for AI-2 activity. The 96well plate was incubated in a rotary shaker at 30  C. The intensity of luminescence was measured hourly using a microplate reader. The fold induction was converted to the molar concentration of AI-2 by comparing the fold induction and 4,5-dihydroxy-2,3-pentanedione (DPD) concentration, as DPD (Omm Scientific, Dallas, USA) were used as the calibration standard. Each filtrated sample was assayed six times in parallel and the mean values reported.

2.6.

Response of AI-2 reporter strain to D-tyrosine

In order to further investigate the effect of D-tyrosine on AI-2 repression, an AI-2 bioluminescence assay in presence and absence of D-tyrosine was conducted. In this assay, five thousand times diluted reporter strain V. harveyi BB170 as described before was grown in fresh AB medium supplemented with 0.1e0.7 mM of DPD in the wells of a 96-well plate. Two series of experiments were conducted: for control, the wells were free of D-tyrosine; other wells were added with 6 mg l1 D-tyrosine. The 96-well plate was shaken in a rotary shaker at 30  C. The light intensity was assayed over time using a microplate reader until get the maximum fold induction of bioluminescence.

2.7.

Extraction and quantification of eDNA

Extracellular DNA was extracted from suspended microorganisms sample according to (Steinberger and Holden, 2005) with modification. Five milliliter of bacterial suspension was collected and resuspended in 0.9% NaCl solution. The resuspended sample was well mixed with a homogenizer (Sonics & Materials, CT, USA). Treated cell solution was filtered through 0.2 mm syring filter and the collected filtrate was stored at 20  C for determining eDNA concentration. The concentration of

DNA was measured by using PicoGreen dsDNA Quantification Kit (Molecular Probes, Invitrogen, Eugene, OR, USA) following the protocol provided by the kit. The calf thymus DNA was used as the standard. The fluorescence intensity was recorded by a microplate reader (BioTek, sygnergy 2, VT, USA).

2.8. Determination of extracellular polysaccharides and proteins Extracellular polysaccharides (PS) and proteins (PN) were extracted from biosample by modified cold aqueous technique method (Jia et al., 1996). Ten milliliter of suspended microorganisms was washed twice with distilled water and centrifuged at 3500 rpm for 10 min. The settled biomass was recovered and resuspended in 10 ml of 8.5% NaCl and 0.22% formaldehyde solution. The mixture was homogenized for 2 min using an ultrasonic homogenizer (Sonics & Materials, CT, USA) in an ice-water bath, and then was centrifuged at 10,000 rpm for 30 min to remove solid residues. The supernatant was harvested for PS, PN measurement and high performance size exclusion chromatography analysis. PS was determined by the phenol-sulfuric acid method (Dubois et al., 1956), whereas PN was analyzed by the modified Lowry method (Lowry et al., 1951). Glucose and bovine serum albumin (SigmaeAldrich) were used as the standards for PS and PN, respectively. High performance size exclusion chromatography (HPSEC) analysis was carried out with Series 200 HPLC system (Perkin Elmer, Waltham, MA, USA) equipped with a Series 200LC quaternary pump, Series 200 autosampler, a Perkin Elmer 600 interface and a UV/Vis detector (785A). A 300  7.8 mm size exclusion chromatography column BioSep SEC S2000 (Phenomenex, Torrance, CA) was used. The mobile phase consisted of 9.0 mM NaCl and 0.9 mM Na2HPO4 at pH 7.0 (Comte et al., 2007). Extracellular polymeric substances were extracted from suspended microorganisms as mentioned above and all samples were filtered through 0.20 mm filters prior to injection. All measurements were conducted at 25  C, mobile phase flow 1.0 ml min1, the sample injection 100 ml. The detection was carried out with a UV detector at 280 nm.

2.9.

Staining and visualization

In order to visualize microbial attachment, the adherent bacteria on glass slides and PP coupons surfaces were stained with LIVE/DEAD BacLight Bacterial Viability kits (Molecular Probes, Eugene, OR, USA), which consisted of two nucleic acid dyes staining on both live and dead cells: SYTO 9 and propidium iodide (PI). SYTO 9 is a green-fluorescent dye which stains both live and dead bacteria with intact and damaged cell membranes while the red-fluorescing PI only stains dead bacteria with damaged cell membranes. The excitation/emission maxima for these dyes are about 480/500 nm for SYTO 9 stain and 490/635 nm for PI. With an appropriate mixture of both dyes, viable bacteria with intact cell membranes are stained green, whereas bacteria with damaged cell membranes fluoresce red. The color assigned to the live and dead cells follows from the color at which the stained cells fluoresce under laser excitation. The sample staining procedure was carried out following the instructions in the manual. First, two

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 9 6 e5 8 0 4

hundred microliters of the mixed solution (1000 times diluted SYTO 9 and PI from the stock solution) was added to each attachment sample on glass slides and PP coupons. The stained sample was then incubated in the dark at room temperature for 15 min. After that, the sample was gently rinsed two times with DI water to remove unbound dyes. Finally, the sample was covered with cover slip and viewed using an Olympus Fluoview FV300 confocal laser scanning microscopy (CLSM) (Olympus Optical, Tokyo, Japan) with a 100X objective.

2.10.

Statistical analysis

All tests were performed in triplicate otherwise stated. Results were expressed as mean value  absolute deviation. Student ttests were employed for analyzing the significance of results at the level of P < 0.05.

3.

Results

3.1.

Microbial attachment on glass and PP

Fig. 1a shows that attachment of microorganisms exposed to 6 mg l1 D-tyrosine was reduced significantly on glass slides compared to the control free of D-tyrosine (Student’s t-test, P < 0.05). After 2-h culture, attachment of microorganisms without exposure to D-tyrosine was 10.8 mg TOC cm2 on the glass surface, while attachment of microorganisms with exposure to D-tyrosine was 8.5 mg TOC cm2, indicating 22% reduction in microbial attachment caused by D-tyrosine. Similar phenomenon was also observed in microbial attachment on PP surface (Fig. 1b). These results indeed are supported by the microscopic observations (Fig. 2). It should be noted that mixed-culture microorganisms with and without exposure to D-tyrosine were collected at different culture (exposure) times of 1e4 h, and used for 1-h microbial attachment assays, as shown in Fig. 1. According to substrate availability over 4-h culture, 1 h- and 2-h old microorganisms were basically in earlier exponential and post exponential growth phases, while 3 h- and 4 h-old microorganisms already entered into earlier stationary and post-stationary growth phases, respectively. It is reasonable to consider that microorganisms at different growth states would have different attachment abilities as observed in the control assays (Fig. 1). This view is strongly supported by the findings of Fletcher

5799

(1977), showing that the number of attached cells harvested from exponential growth phase was greatest, followed by those from stationary and decay phases. In addition, Fig. 3a shows that addition of D-tyrosine to the culture media had no negative effect on the TOC removal efficiency. It was shown in Fig. 3b that the suspended biomass concentration increased from 450 mg l1 to 630 mg l1 in the cultures supplemented with and without D-tyrosine. These suggest that D-tyrosine is not inhibitory to substrate utilization and microbial growth at the concentration studied. Fig. 4 shows that suspended microorganisms with and without exposure to D-tyrosine both carried negative surface charge, but microorganisms exposed to D-tyrosine carried more negative surface charge compared to that of control free of D-tyrosine.

3.2. Cellular ATP and AI-2 contents of suspended microorganisms Fig. 5a showed the possible effect of D-tyrosine on energy metabolism of suspended microorganisms. It can be seen in Fig. 5a that the cellular ATP content did not change significantly in the presence of D-tyrosine compared to that of control. D-tyrosine thus does not appear to inhibit the ATP synthesis. AI-2 as inter-species signaling molecules coordinates the formation of biofilm by various species (Rickard et al., 2008). To investigate the effect of D-tyrosine on cellular communication, Fig. 5b showed the respective AI-2 content of suspended microorganisms with and without D-tyrosine addition. After 1 h culture, the AI-2 content of suspended microorganisms without exposure to D-tyrosine was about 0.27 nmol mg1 in the control, while it decreased to about 0.19 nmol mg1 for suspended microorganisms with exposure to D-tyrosine, i.e. D-tyrosine could suppress the synthesis or secretion of AI-2. The response of reporter strain V. harveyi BB170 to D-tyrosine was further studied, and results were presented in Fig. 6, showing fold induction of luminescence in presence and absence of D-tyrosine, and obviously luminescence was suppressed significantly when the culture was supplemented with D-tyrosine (Student’s t-test, P < 0.05). For example, for 0.4 mM DPD, the fold induction of luminescence was 23.6 in the control free of D-tyrosine, whereas it decreased to 12.4 in the media with addition of D-tyrosine, indicating 48% reduction as compared to that of control. These imply that Dtyrosine at the concentration studied had an inhibitory effect on AI-2 expression.

Fig. 1 e Attachment of microorganisms with (-) and without (,) treatment by D-tyrosine on glass slides (a); on PP coupons (b). Each point represents the mean of triplicate measurements and error bar is absolute deviation from the mean.

5800

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 9 6 e5 8 0 4

Fig. 2 e CLSM images of attachment of microorganisms without D-tyrosine treatment on glass slides (A) and PP coupons (C) at the exposure time of 1 h; with D-tyrosine treatment on glass slides (B) and PP coupons (D) at the exposure time of 1 h.

3.3.

EPS production of suspended microorganisms

EPS are composed of a variety of organic substances, in which polysaccharides and proteins are two major components, and play an important role in microbial attachment onto a solid surface (Flemming and Wingender, 2010). Fig. 7 shows the respective contents of extracellular polysaccharide (PS) and protein (PN) in microorganisms with and without exposure to D-tyrosine. As compared to the control, a 31% reduction in PN

and 17% decrease in PS were observed in microorganisms after 1 h exposure to D-tyrosine, leading to a lowered PN/PS ratio. eDNA is a unique component of the organic substances in the matrix of suspended microorganisms. Fig. 8 shows eDNA content of suspended microorganisms with and without exposure to D-tyrosine. After 1 h exposure to D-tyrosine, eDNA was reduced to 0.006 mg g1 biomass, i.e. a 68% reduction compared to the control. These results suggest that D-tyrosine would significantly inhibit eDNA secretion.

Fig. 3 e Profiles of TOC removal efficiency (a) and microbial growth (b) in the cultures supplemented with (C) and without (B) D-tyrosine. Each point represents the mean of triplicate measurements and error bar is absolute deviation from the mean.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 9 6 e5 8 0 4

Fig. 4 e Surface charge of suspended microorganisms with (C) and without (B) exposure to D-tyrosine. Each point represents the mean of triplicate measurements and error bar is absolute deviation from the mean.

Fig. 9 shows the HPSEC spectra of the EPS extracted from microorganisms with and without exposure to D-tyrosine. In general, retention time in HPSEC spectrum reflects molecular weight of a target chemical, i.e. the peak of a chemical with higher molecular weight appears quicker. For EPS extracted from microorganisms exposed to D-tyrosine, the peaks showed a significant decrease in area compared with those of the control, indicating a lowered EPS production that is consistent with the results in Fig. 7. Two peaks were observed for the EPS extracted from microorganisms without exposure to D-tyrosine at the retention time of 10e13 min, while only one peak appeared for the EPS from microorganisms exposed to D-tyrosine. This implies that EPS with higher molecular weight was reduced due to exposure to D-tyrosine. These suggest that D-tyrosine would not only inhibit the EPS production, but also can alter the EPS composition.

4.

Discussion

Figs. 1 and 2 show that microbial attachments onto hydrophobic PP and hydrophilic glass surfaces were inhibited by

5801

Fig. 6 e Response of AI-2 reporter strain in the cultures supplemented with (C) and without (B) D-tyrosine. Each point represents the mean of six measurements in parallel and error bar is absolute deviation from the mean.

D-tyrosine. So far, little information is available for the role of Dtyrosine in control of mixed-culture biofilm development. Kolodkin-Gal et al. (2010) reported that D-tyrosine-triggered release of amyloid fibers from cell surface would suppress development of a pure culture biofilm. It has been known that the long amyloid fiber facilitates the anchoring of cells to various surfaces, which is essential for microbial attachment and biofilm formation. In addition, the PP coupons used in this study have a contact angle of 99.3  2.2 , while 16.9  0.5 for glass slides. As observed in Figs. 1 and 2, more microorganisms attached to hydrophobic PP than to hydrophilic glass slide. In fact, it has been well documented that higher hydrophobicity of a solid surface would favor microbial attachment (Liu et al., 2004); on the contrary, coating surfaces with non-charged hydrophilic polymers resulted in reduced cell adsorption on a variety of surfaces (Park et al., 1998). Fig. 3 showed that D-tyrosine did not appear to affect the biomass growth and substrate removal efficiency. Such observation is consistent with the results obtained from pure culture experiments (Kolodkin-Gal et al., 2010). It had been reported that D-amino acids at a concentration higher than 20 mg l1 would inhibit bacterial growth (Teeri and Josselyn,

Fig. 5 e Cellular ATP content (a) and AI-2 concentration (b) in the cultures supplemented with (-) and without (,) D-tyrosine. Each point represents the mean of triplicate measurements and error bar is absolute deviation from the mean.

5802

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 9 6 e5 8 0 4

Fig. 7 e PS (a) and PN (b) contents of suspended microorganisms with (-) and without (,) exposure to D-tyrosine. Each point represents the mean of triplicate measurements and error bar is absolute deviation from the mean.

1953). Peptidoglycan is an important mesh-like polymer component of cell wall. In the peptidoglycan polymer, there are two unique amino acids at the terminal of a peptide side chain of peptidoglycan: D-alanine as opposed to its isomer L-alanine. D-tyrosine can replace D-alanine in the peptide side chain of cell wall (Lam et al., 2009), and further alter the cell wall-building protein so that the peptidoglycan production would be slowed down, i.e. D-amino acid negatively regulate the amount of peptidoglycan production. In the presence of D-methionine, peptidoglycan synthesis could be severely inhibited, whereas biomass continued to grow (Caparros et al., 1992). The amount of peptidoglycan per cell decreased significantly due to the increased biomass and decreased peptidoglycan synthesis. However, it had been reported that Escherichia coli would be able to grow properly with 60% decrease of the normal peptidoglycan content (Prats and De Pedro, 1989). The unchanged cellular ATP synthesis in microorganisms with and without exposure to D-tyrosine (Fig. 5a) strongly supports this. It appears from Fig. 5b that AI-2 content of suspended microorganisms decreased due to D-tyrosine in the culture. To exclude other factors affecting AI-2 quorum sensing, the response of reporter strain V. harveyi BB170 to D-tyrosine

Fig. 8 e eDNA of suspended microorganisms with (C) and without (B) exposure to D-tyrosine. Each point represents the mean of triplicate measurements and error bar is absolute deviation from the mean.

showed that the inhibition of AI-2 regulated bioluminescence was only observed in presence of D-tyrosine (Fig. 6), which further confirmed that AI-2 expression was suppressed due to the presence of D-tyrosine. In study of D-amino acids regulated cell wall remodeling, Lam et al. (2009) found that exogenous Dmethionine produced by Vibrio cholera were incorporated into E. coli at the same position in the peptide even though E. coli bacterium did not produce or release D-amino acids. Thus, these rapid diffused small D-amino acids molecules could regulate cells releasing them and the neighboring cells of different species. In study of biofilm inhibition by D-amino acids, Kolodkin-Gal et al. (2010) hypothesized that D-amino acid may play an important role of chemical signal, but opposite to quorum sensing signal molecule, to mediate interspecies communication for facilitating cell dispersion from biofilm. In addition, AI-2 is known to be a cellular communication signal molecule both for Gram-negative and Grampositive bacteria and has a positive effect on biofilm formation (Federle and Bassler, 2003). Due to the opposite effects of these two signal molecules on biofilm formation, D-amino acid

Fig. 9 e HPSEC chromatograms of EPS extracted from suspended microorganisms with (---) and without (d) exposure to D-tyrosine at the exposure time of 1 h.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 9 6 e5 8 0 4

may co-coordinate AI-2 regulated quorum sensing as shown in Fig. 6. However, how these two signals may co-regulate each other needs further investigation. Furthermore, it is speculated that D-amino acids may play a coordinating role through regulating the gene expression to control biofilm community structure or induce biofilm dispersion. Fig. 7 shows that the PS and PN contents of suspended microorganisms tended to decrease with exposure to D-tyrosine. Tsuruoka et al. (1984) also observed that D-amino acid caused reduction of lipoprotein in study of D-amino acid incorporation into peptidoglycan. It has been reported that an incorporation of D-tyrosine into the cellular proteins of Bacillus subtilis (Champney and Jensen, 1970) and E. coli (Miyamoto et al., 2010). As the D-isomer has a similar shape and size to the L-isomer molecule, the D-analog incorporated into proteins in the place of the natural amino acid would modify the structure of the proteins and the enzymic activity (Richmond, 1962). These would eventually lead to the reduced production of PS and PN. As can be seen in Fig. 9, high molecular-weight EPS at the retention time of 10e13 min disappeared in microorganisms exposed to D-tyrosine. In fact, in study of the effect D-amino acid on structure and synthesis of peptidoglycan, Caparros et al. (1992) also found a direct inhibition of D-methionine on the production of high molecular-weight proteins. In addition, EPS have been believed to play an important role in microbial attachment. The reduced production of PS and PN would result in inhibited microbial attachment (Fig. 1). Oliveira et al. (1994) reported that extracellular polysaccharides could promote a preconditioning of the surface, making attachment more favorable, whereas Flint et al. (1997) found that treatment the cells with trypsin or sodium dodecyl sulfate to remove cell surface proteins resulted in a 100-fold reduction in the attachment of Thermophilic streptococci onto stainless steel. It had been shown that eDNA would play an important role in initial microbial adhesion to hydrophobic and hydrophilic surfaces (Das et al., 2010). Many studies have shown that eDNA is an important component of extracellular network that mediates cellecell and cellesurface interactions (Bockelmann et al., 2006; Das et al., 2010). As can be seen in Fig. 8, the presence of D-tyrosine in the culture media caused reduction of DNA in the extracellular network, as the result, less attachment was observed both on glass and PP surface (Fig. 1). These suggest that eDNA may facilitate microbial attachment onto both hydrophilic glass and hydrophobic PP surfaces. Further study is needed to elucidate how D-amino acid would regulate eDNA production. In study of the role of eDNA in Listeria monocytogenes attachment, Harmsen et al. (2010) observed that peptidoglycan, specifically N-acetylglucosamine, together with eDNA could induce adhesion. Inhibited production of peptidoglycan and subsequently eDNA by D-tyrosine would also be responsible for reduced microbial attachment (Fig. 1). It appears from Fig. 4 that microorganisms carried more negative surface charge in presence of D-tyrosine. Since EPS often have charged functional groups, the higher negative charge density would be associated with changes in the composition and quantity of EPS induced by D-tyrosine. According to DLVO theory, increased negative charge would lead to strong electrostatic repulsion between cell and approaching surface (Zita and Hermansson, 1994). Hence, the extent of attachment

5803

would be less (Fig. 1) as a result of increased surface charge or a greater level of electrostatic repulsion.

5.

Conclusions

This study showed that D-tyrosine as a typical D-amino acid could inhibit microbial attachment on both hydrophilic glass and hydrophobic PP surfaces, while no inhibitory effect on microbial growth, ATP synthesis and substrate utilization was observed. It was further found that the synthesis of AI-2, eDNA and EPS were all reduced in the presence of D-tyrosine in the culture media. These in turn provide a plausible explanation for the D-tyrosine-triggered reduction in microbial attachment and demonstrate a mean for biological control of microbial attachment on a solid surface.

references

Bassler, B.L., 1999. How bacteria talk to each other: regulation of gene expression by quorum sensing. Current Opinion in Microbiology 2 (6), 582e587. Bockelmann, U., Janke, A., Kuhn, R., Neu, T.R., Wecke, J., Lawrence, J.R., Szewzyk, U., 2006. Bacterial extracellular DNA forming a defined network-like structure. FEMS Microbiology Letters 262 (1), 31e38. Caparros, M., Pisabarro, A.G., Depedro, M.A., 1992. Effect of Damino acids on structure and synthesis of peptidoglycan in Escherichia-coli. Journal of Bacteriology 174 (17), 5549e5559. Champney, W.S., Jensen, R.A., 1970. Molecular events in the growth inhibition of Bacillus subtilis by D-tyrosine. Journal of Bacteriology 104 (1), 107e116. Chen, G.H., Leung, D.H.W., 2000. Utilization of oxygen in a sanitary gravity sewer. Water Research 34 (15), 3813e3821. Comte, S., Guibaud, G., Baudu, M., 2007. Effect of extraction method on EPS from activated sludge: an HPSEC investigation. Journal of Hazardous Materials 140 (1e2), 129e137. Das, T., Sharma, P.K., Busscher, H.J., van der Mei, H.C., Krom, B.P., 2010. Role of extracellular DNA in initial bacterial adhesion and surface aggregation. Applied and Environmental Microbiology 76 (10), 3405e3408. Dubois, M., Gilles, K.A., Hamilton, J.K., Rebers, P.A., Smith, F., 1956. Colorimetric method for determination of sugars and related substances. Analytical Chemistry 28 (3), 350e356. Federle, M.J., Bassler, B.L., 2003. Interspecies communication in bacteria. Journal of Clinical Investigation 112 (9), 1291e1299. Flemming, H.C., Wingender, J., 2010. The biofilm matrix. Nature Reviews Microbiology 8 (9), 623e633. Fletcher, M., 1977. Effects of culture concentration and age, time, and temperature on bacterial attachment to polystyrene. Canadian Journal of Microbiology 23 (1), 1e6. Flint, S.H., Brooks, J.D., Bremer, P.J., 1997. The influence of cell surface properties of Thermophilic streptococci on attachment to stainless steel. Journal of Applied Microbiology 83 (4), 508e517. Harmsen, M., Lappann, M., Knochel, S., Molin, S., 2010. Role of extracellular DNA during biofilm formation by Listeria monocytogenes. Applied and Environmental Microbiology 76 (7), 2271e2279. Jia, X.S., Furumai, H., Fang, H.H.P., 1996. Yields of biomass and extracellular polymers in four anaerobic sludges. Environmental Technology 17 (3), 283e291. Kolodkin-Gal, I., Romero, D., Cao, S.G., Clardy, J., Kolter, R., Losick, R., 2010. D-amino acids trigger biofilm disassembly. Science 328 (5978), 627e629.

5804

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 9 6 e5 8 0 4

Lam, H., Oh, D.-C., Cava, F., Takacs, C.N., Clardy, J., de Pedro, M.A., Waldor, M.K., 2009. D-amino acids govern stationary phase cell wall remodeling in bacteria. Science 325 (5947), 1552e1555. Liu, Y., Yang, S.F., Li, Y., Xu, H., Qin, L., Tay, J.H., 2004. The influence of cell and substratum surface hydrophobicities on microbial attachment. Journal of Biotechnology 110 (3), 251e256. Liu, Y., Yang, S.F., Tay, J.H., 2003. Elemental compositions and characteristics of aerobic granules cultivated at different substrate N/C ratios. Applied Microbiology and Biotechnology 61 (5-6), 556e561. Lowry, O.H., Rosebrough, N.J., Farr, A.L., Randall, R.J., 1951. Protein measurement with the folin phenol reagent. Journal of Biological Chemistry 193 (1), 265e275. McNab, R., Ford, S.K., El-Sabaeny, A., Barbieri, B., Cook, G.S., Lamont, R.J., 2003. LuxS-based signaling in Streptococcus gordonii: autoinducer 2 controls carbohydrate metabolism and biofilm formation with Porphyromonas gingivalis. Journal of Bacteriology 185 (1), 274e284. Miyamoto, T., Sekine, M., Ogawa, T., Hidaka, M., Homma, H., Masaki, H., 2010. Detection of D-amino acids in purified proteins synthesized in Escherichia coli. Amino Acids 38 (5), 1377e1385. Oliveira, R., Melo, L., Oliveira, A., Salgueiro, R., 1994. Polysaccharide production and biofilm formation by Pseudomonas fluorescens: effects of pH and surface material. Colloids and Surfaces B-Biointerfaces 2 (1e3), 41e46. Park, K.D., Kim, Y.S., Han, D.K., Kim, Y.H., Lee, E.H.B., Suh, H., Choi, K.S., 1998. Bacterial adhesion on PEG modified polyurethane surfaces. Biomaterials 19 (7e9), 851e859. Prats, R., De Pedro, M.A., 1989. Normal growth and division of Escherichia coli with a reduced amount of murein. Journal of Bacteriology 171 (7), 3740e3745.

Richmond, M.H., 1962. The effect of amino acid analogues on growth and protein synthesis in microorganisms. Bacteriological Reviews 26 (4), 398e420. Rickard, A.H., Campagna, S.R., Kolenbrander, P.E., 2008. Autoinducer-2 is produced in saliva-fed flow conditions relevant to natural oral biofilms. Journal of Applied Microbiology 105 (6), 2096e2103. Steinberger, R.E., Holden, P.A., 2005. Extracellular DNA in singleand multiple-species unsaturated biofilms. Applied and Environmental Microbiology 71 (9), 5404e5410. Surette, M.G., Bassler, B.L., 1998. Quorum sensing in Escherichia coli and Salmonella typhimurium. Proceedings of the National Academy of Sciences of the United States of America 95 (12), 7046e7050. Teeri, A.E., Josselyn, D., 1953. Effect of excess amino acids on growth of certain Lactobacilli. Journal of Bacteriology 66 (1), 72e73. Tsuruoka, T., Tamura, A., Miyata, A., Takei, T., Iwamatsu, K., Inouye, S., Matsuhashi, M., 1984. Penicillin-insensitive incorporation of D-amino acids into cell-wall peptidoglycan influences the amount of bound lipoprotein in Escherichia-coli. Journal of Bacteriology 160 (3), 889e894. Wilen, B.M., Jin, B., Lant, P., 2003. The influence of key chemical constituents in activated sludge on surface and flocculating properties. Water Research 37 (9), 2127e2139. Wood, T.K., Hong, S.H., Ma, Q., 2011. Engineering biofilm formation and dispersal. Trends in Biotechnology 29 (2), 87e94. Zita, A., Hermansson, M., 1994. Effects of ionic strength on bacterial adhesion and stability of flocs in a wastewater activated sludge system. Applied and Environmental Microbiology 60 (9), 3041e3048.

water research 45 (2011) 5805

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Erratum

Erratum to “Bioassays as a tool for evaluating advanced oxidation processes in water and wastewater treatment” [Water Research 45 (2011) 4311e4340] Luigi Rizzo Department of Civil Engineering, University of Salerno, via Ponte don Melillo 1, 84084 Fisciano (SA), Italy

On page 4319, left column, the sentence starting in line 11 from below should read: “According to the results available in scientific literature, AOPs were found to decrease and increase toxicity”. The subsequent sentence, “A decreased toxicity was . ., 2010).”, should be removed.

DOI of original article: 10.1016/j.watres.2011.05.035. E-mail address: [email protected]. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.013