WATER RESEARCH A Journal of the International Water Association
Editor-in-Chief Mark van Loosdrecht Delft University of Technology Department of Biochemical Engineering Julianalaan 67 2628 BC Delft The Netherlands Tel: +31 15 27 81618 E-mail:
[email protected]
Editors J. Block Université H. Poincaré, Nancy I France David Dixon University of Melbourne Australia Hiroaki Furumai The 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 Swiss Federal Institute of Aquatic Science and Technology (EAWAG) Switzerland W. Rauch University 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 Univ. of California, Davis USA
Associate Editors Andrew Baker University of New South Wales Australia
Damien Batstone The University of Queensland Australia G-H. Chen The Hong Kong University of Science & Technology Hong Kong China Tom Curtis Univ. of Newcastle upon Tyne UK Ana Deletic Monash University USA Francis de los Reyes III North Carolina State University USA Rob Eldridge The University of Melbourne Australia Rosina Girones University of Barcelona Spain Stephen Gray Victoria University Australia Kate Grudpan Chiang Mai University Thailand E.E. Herricks University of Illinois - Urbana USA Peter Hillis United Utilities Plc UK H-Y. Hu Tsinghua University China P.M. Huck University of Waterloo Canada Bruce Jefferson Cranfield University UK Ulf Jeppsson Lund University Sweden Sergey Kalyuzhnyi Moscow State University Russian Federation Jaehong Kim Georgia Institute of Technology USA Jes La Cour Jansen Lund Institute of Technology Sweden G. Langergraber BOKU/Univ. of Natural Res. and Applied Life Scs. Austria S-L. Lo National Taiwan University Taiwan Dionisis Mantzavinos Technical University of Crete Greece
Y. Matsui Hokkaido University Japan A. Maul Université Paul Verlaine-Metz France Max Maurer EAWAG Switzerland How Yong Ng National University of Singapore Singapore Satoshi Okabe Hokkaido University Japan S.L. Ong National University of Singapore Singapore Jong M. Park Pohang University of Science & Technology Korea Susan Richardson U.S. Environmental Protection Agency USA Miguel Salgot University of Barcelona Spain David Sedlak University of California, Berkeley USA Jean-Philippe Steyer LBE-INRA France M. Takahashi Hokkaido University Japan Kai Udert EAWAG Switzerland V.P. Venugopalan BARC Facilities India E. von Sperling Federal University of Minas Gerais Brazil Hanqing Yu University of Science & Technology of China China J. Zilles University of Illinois Urbana USA A.I. Zouboulis Aristotle University Greece
Editorial Office Mailing address Water Research Radarweg 29 1043 NX Amsterdam The Netherlands Tel: 31 20 485 2701 Fax: 31 20 485 2521; 31 20 485 3264 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: Please visit this journal’s homepage at http://www.elsevier.com/locate/watres. You can track accepted articles at http://www.elsevier.com/ trackarticle and set up e-mail alerts to inform you of when an article’s status has changed. Also accessible from here is information on copyright, frequently asked questions and more. Contact details for questions arising after acceptance of an article, especially those relating to proofs, will be provided by the publisher. Orders, claims, and journal enquiries: Please contact the Elsevier Customer Service Department nearest you: St. Louis: Elsevier Customer Service Department, 11830 Westline Industrial Drive, St. Louis, MO 63146, USA; phone: (877) 8397126 [toll free within the USA]; (+1) (314) 4537076 [outside the USA]; fax: (+1) (314) 5235153; 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 ) 1 e1 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Review
Control of start-up and operation of anaerobic biofilm reactors: An overview of 15 years of research Renaud Escudie´, Romain Cresson, Jean-Philippe Delgene`s, Nicolas Bernet* INRA, UR50, Laboratoire de Biotechnologie de l’Environnement, Avenue des Etangs, Narbonne, F-11100, France
article info
abstract
Article history:
Anaerobic biofilm reactors have to be operated in a way that optimizes on one hand the
Received 1 April 2010
start-up period by a quick growth of an active biofilm, on the other hand the regular
Received in revised form
operation by an active control of the biofilm to avoid diffusion limitations and clogging.
28 July 2010
This article is an overview of the research carried out at INRA-LBE for the last 15 years. The
Accepted 28 July 2010
start-up of anaerobic biofilm reactors may be considerably shortened by applying a short
Available online 5 August 2010
inoculation period (i.e. contact between the inoculum and the support media). Then, the increase of the organic loading rate should be operated at a short hydraulic retention time
Keywords:
and low hydrodynamic constraints in order to favor biofilm growth. After the start-up
Anaerobic digestion
period, biofilm growth should be controlled to maintain a high specific activity and prevent
Biofilm
clogging. This can be done in particulate biofilm systems by using hydrodynamics to
Hydrodynamics
increase or decrease shear forces and attrition but is much more difficult in anaerobic fixed
Methane yield
bed reactors.
Start-up
ª 2010 Elsevier Ltd. All rights reserved.
Contents 1. 2.
3.
4.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Start-up: how to grow an active biofilm as fast as possible? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 2.1. Inoculation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 2.2. Increase of the organic loading rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 2.2.1. Short HRT to out compete planktonic microorganisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 2.2.2. Low shear forces to favor biofilm accumulation and minimizing detachment . . . . . . . . . . . . . . . . . . . . . . . . . 3 2.3. The methane yield: a specific indicator to monitor the biofilm installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Normal operation: how to control the process to maintain an active biofilm? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 3.1. Fixed bed reactors: problem of clogging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 3.2. Moving bed reactors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
* Corresponding author. Tel.: þ33 468 425174; fax: þ33 468 425160. E-mail address:
[email protected] (N. Bernet). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.07.081
2
1.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 e1 0
Introduction
Anaerobic digestion is a feasible alternative to aerobic processes for the treatment of wastewater with a high organic carbon concentration, as in anaerobic digestion, comparably less sludge is produced while methane is produced for the generation of heat- or electricity. The most widely used anaerobic wastewater treatment processes are high-rate systems in which hydraulic and solid retention times are uncoupled to achieve high biomass retention (Rajeshwari et al., 2000). These processes may use self-immobilized biomass as granules in the up-flow anaerobic sludge blanket (UASB), expanded granular sludge blanket (EGSB) and Internal Circulation (IC) processes (Lettinga, 1995; Nicolella et al., 2000; McHugh et al., 2003; Liu and Tay, 2004). In other cases, biomass is immobilized on and between inert packing material in anaerobic filter (AF) (Rajeshwari et al., 2000), on inert mobile particles in fluidized bed (FB) (Nicolella et al., 2000), turbulent bed reactors (Buffie`re et al., 2000; Buffie`re and Moletta, 2000b; Arnaiz et al., 2003, 2007) or hybrid reactors (Rajinikanth et al., 2008, 2009). Anaerobic fixed-film reactors are believed to require more sophisticated process control and operating conditions as well as a long start-up period (Salkinoja-Salonen et al., 1983; Hickey et al., 1991; Weiland and Rozzi, 1991; Pun˜al et al., 2000). Startup periods for anaerobic processes have been reported to take between 2 and 9 months (Lauwers et al., 1990). During the startup period an active biofilm is formed. The maturation of this biofilm largely determines the later performance of the system. In contrast, aerobic systems can be started-up much faster. At INRA-LBE we have been working on anaerobic biofilm reactors for about 15 years. This paper presents and discusses a synthesis of the results obtained on control of anaerobic biofilm reactors during both start-up period and normal operation of fixed and mobile bed systems.
2. Start-up: how to grow an active biofilm as fast as possible?
adapted inoculation sludge are brought in contact inside the reactor. The length of the contact time is chosen empirically and can vary from a few days up-to more than one month (Marin et al., 1999; Ye et al., 2005). It is generally believed that a long contact time between a concentrated inoculum and the carrier is necessary and will favor biofilm growth in batch conditions. Cresson et al. (2007a) studied the initial adhesion of bacteria from an anaerobic sludge on mineral particles in an inverse turbulent bed reactor. Using the molecular fingerprinting tool PCR-SSCP (Dabert et al., 2002), Cresson et al. (2007a) showed that, after only 12 h of contact time, microorganisms were attached on the carrier particles. The bacterial molecular fingerprint of this early biofilm was very close to the inoculum. Compared to traditional inoculation strategies, only a very short period is necessary to obtain adhesion of microorganisms on the support media and to initiate biofilm formation. Consequently, it is possible to shorten considerably the duration of the inoculation period. This result has been confirmed by us in the later studies, during which an inoculation time as short as a few hours was chosen (Cresson et al., 2006, 2007b, 2008). An inoculation time of 24 h or even less has also been successfully applied during start-up of a pilot-scale anaerobic fixed bed reactor (1 m3) using the commercial carrier material Cloisonyl (Cresson, 2006). The influence of global properties of carrier materials on the performance of anaerobic biofilm reactor has been the topic of many studies (Garcı´a-Caldero´n et al., 1996; Picanc¸o et al., 2001; Yang et al., 2004). More recently, Habouzit et al. (2009) studied early adhesion (2 h) of a methanogenic consortium on different carrier materials. They showed that the nature and physico-chemical properties of the carrier significantly influenced early adhesion of bacteria and Archaea, not only quantitatively but also qualitatively. The ratio Archaea/Bacteria of the adhered microbial communities, determined by qPCR, was strongly dependent on the nature of the support material.
2.2. The aim of the start-up is to develop an active biofilm on the carrier and to reach the nominal organic loading rate (OLR) with a satisfying treatment performance. In many cases, start-up of an anaerobic reactor takes 4 months or more than a year for thermophilic processes before a steady state with respect to removal efficiency is reached (Kim and Speece, 2002). Shortening the start-up time is a key point to increase the economical competitiveness of the anaerobic processes (Weiland and Rozzi, 1991). In the following, we discuss two distinct steps during startup: (1) the inoculation period during which the carrier is put in close contact with an inoculating sludge to initiate biofilm attachment and (2) the progressive increase of the organic loading rate to stimulate microbial growth of the biofilm. We show how both steps can be shortened and optimized.
2.1.
Inoculation
In most cases, anaerobic reactors are inoculated as a batch. During inoculation the carrier material and an active and well-
Increase of the organic loading rate
After the inoculation period, the organic loading rate is normally increased progressively and continuously. Anaerobic digestion is the result of synergistically interacting microbes with the limiting step being methanogenesis. The increase of the organic loading rate must carefully monitored to avoid overloading of the system which could lead to an inhibition of methanogens and consequently to failure of the start-up process. In our research, we aim for shortening this period by favoring the growth of an active biofilm without inhibiting the system. The main parameters to tune during this period are the hydraulic retention time (HRT) and the hydrodynamic conditions in the reactor.
2.2.1.
Short HRT to out compete planktonic microorganisms
A conventional way to operate increase of the organic loading rate is to feed the reactor at a progressively increasing influent flow rate while keeping the influent COD concentration constant. The flow rate is increased stepwise when a minimum performance (for example, 80% COD removal) is
3
2.2.2. Low shear forces to favor biofilm accumulation and minimizing detachment Shear forces control biofilm formation, but their influence during the start-up phase of bioreactors has not been well characterized. Biofilm accumulation in the reactor results from a balance between growth and detachment mainly due to shear (van Loosdrecht et al., 1995). Biofilm detachment occurs when local shear forces exceed the cohesiveness of the biofilm. At steady state, the balance between growth and detachment determines the physical structure of the biofilm, and thereby the settling and fluidization characteristics in the case of particulate biofilms (van Loosdrecht et al., 2002; Stoodley et al., 2002). Nevertheless, high shear lead to the formation of thin, dense and active biofilm but they are suspected to slow down biofilm formation. Cresson et al. (2007b) studied the influence of hydrodynamic conditions on the start-up phase of an inverse turbulent bed bioreactor. Two identical reactors, differing only by the gas velocity ensuring the carrier fluidization and generating the main hydrodynamic strengths (attrition), were
48
3
36
2
24
1
12
0
AVS (g.L carrier-1)
4
-1
reached. This conservative strategy is frequently successful but needs several months to reach steady state with respect to treatment performance. Such a strategy enhances competition between suspended and biofilm biomass for the organic substrate. This point has been focused on by Tijhuis et al. (1994) for the start-up of an aerobic airlift biofilm reactor. They showed that the dilution rate of the system must be lower than the maximum growth rate of the microorganisms to enhance growth of biofilms. For an anaerobic process, Michaud et al. (2005) showed that six days of initial HRT were less favorable than an HRT of one day for the colonization of a particulate carrier by a methanogenic inoculum. Cresson et al. (2008) confirmed the influence of the hydraulic retention time on the start-up phase of a methanogenic inverse turbulent bed bioreactor. Two identical reactors were monitored, the only differing parameter being the hydraulic HRT: one of the reactors was fed with a diluted wastewater at a constant HRT of one day, the organic loading rate being increased by decreasing the substrate dilution; the second reactor was fed at a constant influent concentration of 20 kgCOD m3, the organic loading rate being increased by decreasing the HRT from 40 to 1 day. After 45 days, both reactors were operated at an organic loading rate of 20 kgCOD m3 d1 and an HRT of 1 day. Strong differences were observed on biofilm growth as shown in Fig. 1: in the reactor operated at a constant short HRT, biofilm concentration was 4.5 higher than in the reactor operated at an increasing HRT. This difference was attributed to the competition between planktonic and biofilm microorganisms in the reactor that started at a long HRT. Suspended biomass was quickly washed out in other reactor because of the low HRT. Najafpour et al. (2006) applied a constant HRT of 1.5 days during the start-up of a hybrid anaerobic reactor inoculated with a granular sludge. They obtained an organic loading rate of 23 kgCOD m3 d1 after 26 days. More recently, Alvarado-Lassman et al. (2010) studied the start-up period of anaerobic inverse fluidized bed reactors. They obtained better results in continuous than in batch mode in which a string competition between planktonic organisms and biofilm occurs.
VSS (g.L reactor )
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 e1 0
0 0
10
20
30
40
Time (days)
Fig. 1 e Comparison of attached volatile solids (AVS, black symbols) and volatile suspended solids (VSS, white symbols) evolution at a constant short HRT in the reactor A (square) and at a constant influent concentration and a decreasing HRT in the reactor B (circle) (Cresson et al., 2007b).
monitored. Over the first 96 days, a faster start-up in terms of substrate removal was observed in the reactor with the lower hydrodynamic strength. This result was correlated to lower attached biomass and higher specific removal rate for the reactor subjected to high hydrodynamic strength. In Fig. 2, we present pictures of the colonized particles at the end of the start-up period. A clear difference can be seen between the thin and smooth biofilm grown at high hydrodynamic strengths (Fig. 2A) and the fluffy biofilm grown at low shear forces (Fig. 2B). Similar results were obtained in aerobic airlift reactors by Tijhuis et al. (1996). Once the start-up was completed and the reactor stabilized at an organic loading rate of 6 kgCOD m3 d1, the same hydrodynamic strengths were applied by adjusting the gas velocity. In Fig. 3, we present the changes observed in both reactors in terms of biofilm production, detachment and global accumulation. The experimental results demonstrate that for this type of process, biofilm development highly depends on the gas velocity which regulates the hydrodynamic strengths and controls growth and detachment rates. The continuation of biofilm growth for the system that was initially subjected to low gas velocity demonstrated that it was possible to strongly increase the hydrodynamic strengths without causing a massive detachment of the biofilm and loss of active biomass leading to a potential failure of the reactor. Thus it seems possible to control biofilm growth in this type of bioreactor via the gas velocity. Detachment of biomass is a useful tool to avoid the loss of carrier material, resulting from the density change due to the over-accumulation of biomass on the carrier. Hydrodynamic conditions can be successful to control biofilm characteristics. The implementation of low hydrodynamic strengths during the start-up of an inverse turbulent bed bioreactor made it possible to treat a more important amount of pollution, and enabled to reach higher volumetric conversion capacities more quickly. Regarding the biofilm development, the implementation of low hydrodynamic shear forces allows a faster accumulation of biomass fixed on the carrier. In contrast, strong hydrodynamic shear forces
4
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 e1 0
Fig. 2 e Microscopic observation of colonized carrier at day 80 for the reactor subjected to high hydrodynamic strengths (A) and low hydrodynamic strengths (B); magnifying 310 (1) and 340 (2) (Cresson et al., 2007b).
generated by high gas velocity slowed down biofilm growth but lead to high specific removal rates. We recommend starting up a bioreactor by applying minimal shear forces in order to enhance the biofilm growth during the early phase of biofilm development. Then hydrodynamic shear forces can be increased after a sufficient amount of well-adapted biomass has accumulated on the carrier.
2.3. The methane yield: a specific indicator to monitor the biofilm installation One of the difficulties to follow and optimize the start-up period is the lack of appropriate measurements, including online or offline data. It is generally difficult or even impossible in the case of fixed bed processes to sample colonized carrier and measure attached biomass. Michaud et al. (2002, 2005) showed that the methane yield YCH4, can be used as an indirect parameter for evaluating the start-up operation of an anaerobic biofilm reactor. YCH4 is defined as the amount of methane produced for a given quantity of organic matter removed. This parameter is the result of the balance between the flows of organic carbon to catabolism and anabolism in methanogenic ecosystems. At the beginning of the start-up period, YCH4 is very low, indicating an important anabolic activity of the microorganisms to build the biofilm. Then, its value increases up-to a stable level close to the theoretical value of 0.35 LCH4 g1 COD, corresponding to the final establishment of a stable methanogenic ecosystem. The time course of
YCH4 describes the three phases of biofilm formation with the induction (decrease of YCH4 showing the washing out of the inoculum sludge), growth (increase of YCH4) and steady state (stabilization of YCH4 close to the theoretical value). Fig. 4 shows an example of the evolution of the methane yield during the start-up of an anaerobic inverse turbulent bed reactor (Michaud et al., 2002). It should be pointed out that YCH4 gives information on the ratio between the catabolic activity (methane production) and the anabolic activity (biofilm production). It is therefore very similar to the model proposed by Liu et al. (2003) to describe the distribution between catabolism and anabolism in aerobic biofilms. In this case, dissolved organic carbon (DOC) distribution between catabolism and anabolism can be described by a ratio of the DOC channeled into carbon dioxide (SCO2) to the DOC converted into biomass (Sg). The methane yield on anaerobic biofilms is equivalent to this SCO2/Sg ratio.
3. Normal operation: how to control the process to maintain an active biofilm? When the start-up period is completed, anaerobic fixed-film reactors are able to treat the nominal organic loading rate with treatment efficiencies that comply with most effluent requirements. However, even though reactor performance (e.g., carbon removal efficiency, effluent quality, biogas production) is stable, the biofilm continues to grow. When biofilm reactors treating wastewater are operated for a long
5
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 e1 0
a
14
Gas velocity equalization
12
-1
AVS ( g. L )
10
8
6
4
2 Time (d) 0 150
160
170
180
190
210
220
AVS - Low strengths
AVS - High strengths
b
200
Gas velocity equalization 0.25
-1
-1
Biomass (g.L .d )
0.15
0.05 0 -0.05
-0.15
Time (d) -0.25 150
160
170
180
190
200
210
220
Detachment rate - High strengths
Accumulation rate - High strengths
Production rate - High strengths
Detachement rate - Low strengths
Accumulation rate - Low strengths
Production rate - Low strengths
Fig. 3 e Biofilm characteristics for the two experimental setups before and after gas velocity equalization: a) Variation of the AVS, b) Biomass detachment rate, biomass accumulation rate on the carrier and biomass production rate (Cresson et al., 2007b).
period, the purpose is to control both biofilm quantity and activity in order to maintain the global reactor performances. Depending on the process used (fixed or mobile bed), the ways to control the biofilm are different.
3.1.
Fixed bed reactors: problem of clogging
When working over a long period under high load conditions, the main problem of fixed bed reactors is excess biomass accumulation leading to clogging of the bed. Consequently, the active volume (i.e., the liquid volume) decreases, thus limiting the treatment capacity of the process. As a result, part of the reactor volume may then be operated as a “dead zone” or “stagnant water zone”, with the liquid flowing through preferential pathways, decreasing the retention time of substrates in the reactor as well as the degree of contact between incoming substrates and the viable microbial populations. At
the industrial scale, mechanical solutions are used to reduce clogging (e.g., high liquid velocity, gas injection). Depending on the carrier media characteristics (size, morphology, surface texture, porosity) and their arrangement inside the reactor, a fixed bed can mainly operate as an anaerobic filter: a large part of biomass immobilized within this reactor is constituted not only by an attached biofilm adhered to the media surfaces, but also by suspended biomass trapped within the interstitial void spaces (Rajeshwari et al., 2000). Such bioreactors are therefore particularly sensible to clogging. In many studies, the focus was thus on packed beds and the influence of the media characteristics on biomass performances, hydrodynamics and biomass retention (Tay and Show, 1998; Show and Tay, 1999). The use of carrier media with large porosity may reduce the extent of shortcircuiting, leading to better treatment performance. In other words, for anaerobic fixed bed, increasing the specific surface
6
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 e1 0
0,4
6
0,3
4,5
0,2
3
0,1
1,5
0
C O D r e mo v a l r a te ( g / l . d )
Y CH4 ( l CH4 / gCODrem)
YCH4 Y CH 4 COD removal rate
0 0
10
20
30
40
50
60
70
80
90
which a gas flow was generated. The plots are very close to each other, and they looks like the RTD curve expected from a continuous stirred tank reactor (CSTR). From Fig. 6b, it is possible to point out that, when biogas is not generated by the biological reaction, the mixing efficiency is reduced and preferential pathways occur. These experimental results suggest the liquid mixing in a fixed bed reactor equipped with Cloisonyl carrier is largely caused by biogas generation.
3.2.
Moving bed reactors
100
Time (days)
Fig. 4 e Time course of methane yield YCH4 and COD removal during the start-up of an ITBR (Michaud et al., 2002).
area of the carrier media at the expense of its porosity may result in lower treatment performances. In this paragraph, we present results obtained on a pilotscale up-flow anaerobic fixed bed process of 0.982 m3 that has been operated for more than 7 years at LBE. The selected media, Cloisonyl tube, is a multi-channel tubular structure as shown in Fig. 5A that limits the filter effect and avoids trapping biomass. Starting from the inoculation in 1997, the anaerobic fixed bed was continuously fed with raw industrial wine distillery vinasses and was fully instrumented for scientific purposes (Steyer et al., 2002a,b). Throughout the 7-year period, no external operations were necessary in order to remove biofilm. During this period, the global treatment performance was maintained whereas in 2004, the total volume of biomass represented 720 L. The liquid volume was about 230 L, which represented only 25% of the initial reactor active volume. Thus, even if 75% of the reactor volume were colonized by a biofilm, using Cloisonyl as a carrier permitted to maintain good liquid mixing and treatment performance during 7 years. Tracer studies were carried out in this colonized reactor to characterize the liquid mixing (Escudie´ et al., 2005). The Retention Time Distribution RTD curves obtained as a response to the pulse input of tracer are shown in Fig. 6a and b. Fig. 6a plots the RTD curves for the five experiments during
In order to avoid clogging of fixed bed reactors, moving bed systems were introduced. We developed reactors using both liquid and gas fluidization technologies for anaerobic wastewater treatment. Liquid fluidization can be usually achieved with an upward liquid flow with particles having a density higher than the liquid (Buffie`re et al., 1995a,b; Garcı´a-Caldero´n et al., 1996), or with downward liquid flow with particles having a density lower than the liquid (Garcı´a-Bernet et al., 1998; Garcı´aCaldero´n et al., 1998a,b): the last configuration is named inverse or down-flow fluidization. Anaerobic down-flow fluidization has been recently applied with good performance at laboratory scale by other research groups (Sowmeyan and Swaminathan, 2008; Alvarado-Lassman et al., 2010). During the start-up phase but also after the stabilization of the organic loading rate, the biofilm thickness increases on the carrier surface and causes modification of these particle characteristics: (i) the growth of biofilm enlarges particle diameter; (ii) because wet density of the biofilm (z1050 kg L1) is different from the specific density of the carrier material, the colonized particle density increases for a conventional fluidized bed or decreases for an inverse fluidized bed. (iii) The shape and the surface roughness of the particles change. The liquid flow rate, used to fluidize the particles needs to be adapted in order to maintain the surface level of the bed and to avoid the wash out of colonized particles from the reactor. For example, in a down-flow anaerobic fluidized bed reactor, Garcı´a-Caldero´n et al. (1998b) determined the bed expansion parameters used by the equation of Richardson and Zaki (1954) to relate the bed porosity (i.e., the bed height) and the superficial liquid velocity according to the biofilm thickness. For conventional fluidization, studies were also carried out in
Fig. 5 e Cloisonyl tube (A) bare, (B) colonized by anaerobic biofilm after 7 years of experiment.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 e1 0
7
Fig. 6 e Experimental curves of Retention Time Distribution in an anaerobic fixed-film reactor (Escudie´ et al., 2005).
order to measure and predict these parameters; however every proposed correlation needs to be used with caution. Further studies are needed in order to determine a model describing this complex phenomenon. In particular, the particle distribution in size, density and shape needs to be considered. A large distribution in particle characteristics can generate their segregation (or stratification) through the bed height (Saravanan and Sreekrishnan, 2006). In an anaerobic up-flow fluidized bed that has been operated for more than 1 year, Buffie`re et al. (1998) reported significant carrier particles stratification. Crushed pozzolana with a density of 1990 kg m3 having a large size distribution was used as carrier media. From the bottom to the top of the bed, the mean diameter increased from 610 to 1057 mm, whereas the mean biofilm thickness increased from 120 to 430 mm. Bed stratification can have negative effects on the reactor performances: thicker biofilms are more likely diffusion limited and may wash out of the system more quickly. To maintain uniform particle size and remove excess biofilm from larger particles, technical solutions have been implemented: external separators, internal screen cleaning devices, operation of an impeller at the top of the bed (Ruggeri et al., 1994; Safferman and Bishop, 1996; Shieh et al., 1981; Trinet et al., 1991). Understanding particle stratification and the prediction of the segregation/mixing regimes may play an important role in the reactor design and scale-up. In order to simplify the problem, studies have focused on mixing and segregation in binary-solids mixtures. Much work has been devoted to investigating the segregation of particles by size (sizing) and by density (sorting). Recently, it was demonstrated that particles of different shapes (sphericity) can segregate in fluidized beds, even when they have the same volume and density (Escudie´ et al., 2006a). In addition, a phenomenon called “layer inversion” appears when a mixture of two particle species is fluidized both in conventional fluidized bed (Escudie´ et al., 2006b) and in inverse fluidized bed systems (Escudie´ et al., 2007). In these cases, the mixed layer manifests a bed contraction. Bed contraction is a reduction of the voidage of the bed containing the binary mixture compare to the theoretical voidage of the bed containing the pure particles. Several schemes for predicting this contraction effect (i.e., the voidage and the bed height for a given operating condition)
have been proposed in the literature. However, they are successful only in some basic operating conditions (Escudie´ and Epstein, 2008, 2009). Further experimental and modeling efforts are required to describe the complexity of these phenomena (combination of solids species of different sizes, shapes, densities, compositions) and thus to develop tools to better design the anaerobic bed reactor fluidized by a liquid flow. Gas fluidization technologies can create higher hydrodynamics forces than liquid fluidization technologies. In particular, in an anaerobic inverse turbulent bed reactor, the intensities of the collisions between particles (called attrition or abrasion) increase drastically when a part of the biogas flow is re-injected, whereas the liquid velocity has only a weaker influence. Investigations by means of a highfrequency-response hydrophone (Buffie`re and Moletta, 2000a) were performed in this type of reactor, and it was demonstrated that the collision frequency and the collisional particle pressure increase with the gas velocity. Since high hydrodynamic forces can control the biofilm growth and thickness, the inverse turbulent bed reactor showed good results as the biofilm thickness remained low with a high specific biofilm activity. Michaud et al. (2003) demonstrated that the gas flow (which directly controls the hydrodynamic force) can be used to control an anaerobic turbulent bed process. The consequences of gas velocity variations on biofilm development and reactor performance were observed in a stabilized process. Short but intensive increases of gas velocity are shown to induce more detachment than a high but constant gas flow rate. Hydrodynamic conditions control the composition of the growing biofilm in terms of microorganisms concentration, estimated from a measurement of the phospholipid content in the biofilm, and exocellular polymeric substances (EPS). The microbial cell fraction within the biofilm was found to be inversely proportional to the gas velocity applied to the process. This could be explained by a denser EPS matrix necessary to maintain biofilm cohesion at higher gas velocities. This has been observed previously by Lazarova et al. (1994) in a similar biofilm process but under aerobic conditions. The specific activity, expressed as methane production or COD removal rate, was higher in biofilms formed under high hydrodynamic stress (Michaud et al., 2003). The control of the
8
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 e1 0
hydrodynamics in a biofilm reactor should make it possible to obtain a dense and active biofilm. This can be easily done in an inverse turbulent bed reactor in which the gas velocity can be changed while keeping the colonized carrier in the reactor. Such a control is more difficult in a conventional fluidized bed reactor in which hydrodynamic conditions are driven by the liquid velocity, and even more difficult in fixedfilm reactors. Chai and Moletta (2007) proposed an intermediate process between fixed- and fluidized bed reactors which is adapted from aerobic moving bed biofilm reactors. In this type of reactor, a floating carrier is sequentially mixed by a submerged pump. This process was successfully applied to the treatment of milk permeate from dairy industry (Wang et al., 2009).
4.
Conclusions
Long start-up periods are difficult to sell when promoting anaerobic reactors for wastewater treatment. Therefore, shortening the time required for starting up anaerobic reactors will greatly enhance the use of this technology. The first challenge for increasing the promotion of anaerobic wastewater treatment reactors is to improve the knowledge of the start-up period in order to reduce its duration. The start-up phase is defined as the period necessary to bring the bioreactor to its nominal load while the treatment efficiency meets the desired requirements. We show for fixed and moving bed reactors at the lab- and pilot-scale that it is possible to considerably shorten the startup period of anaerobic biofilm processes. Within only 30 days, we reached organic loading rates as high as 20 kgCOD m3 d1. A successful start-up strategy included the following two keyfactors: (i) a short initial contact time between the inoculum and the carrier material: only a few hours are necessary to obtain adhesion of microorganisms on the carrier media and to initiate the biofilm formation, (ii) a short hydraulic retention time (e1 day) to wash out suspended biomass from the reactor and to force biofilm growth on the carrier material. For moving bed reactors, applying low hydrodynamic constraints (low fluidization velocity) allows a faster accumulation of biomass fixed on the carrier. Current research is focusing on the affinity of Archaea during initial adhesion towards specific carrier materials. A preference for certain materials could guide the selection of carrier materials in order to establish a quicker onset of methanogenesis in the biofilm. Another challenge is the control of biofilm properties (i.e., thickness and activity) during steady-state operation to maintain optimum performances. This can be operated by an active control of detachment through hydrodynamics: shear stress in fixed bed reactors, attrition in moving bed reactors. For moving bed reactors, the control of biofilm colonization is also a key parameter in order to appropriately operate the particle fluidization and to prevent particle segregation and wash out. For fixed bed reactor, technical solutions (reactor
design or cleaning strategies) need to be implemented in order to limit biomass accumulation leading to reactor clogging.
Acknowledgements We are very grateful to Kim Milferstedt for his critical reading of the manuscript and his suggestions that significantly improved the quality of this article.
references
Alvarado-Lassman, A., Sandoval-Ramos, A., FloresAltamirano, M.G., Vallejo-Cantu, N.A., Mendez-Contreras, J.M., 2010. Strategies for the startup of methanogenic inverse fluidized-bed reactors using colonized particles. Water Environ. Res. 82 (5), 387e391. Arnaiz, C., Buffie`re, P., Elmaleh, S., Lebrato, J., Moletta, R., 2003. Anaerobic digestion of dairy wastewater by inverse fluidization: the inverse fluidized bed and the inverse turbulent bed reactors. Environ. Technol. 24, 1431e1443. Arnaiz, C., Buffie`re, P., Lebrato, J., Moletta, R., 2007. The effect of transient changes in organic load on the performance of an anaerobic inverse turbulent bed reactor. Chem. Eng. Process 46 (12), 1349e1356. Buffie`re, P., Fonade, C., Moletta, R., 1995a. Continuous operation of a fluidized bed bioreactor for anaerobic digestion: residence time influence on degradation kinetics. Biotechnol. Lett. 17 (8), 833e838. Buffie`re, P., Steyer, J.P., Fonade, C., Moletta, R., 1995b. Comprehensive modeling of methanogenic biofilms in fluidized bed systems. Biotechnol. Bioeng. 48 (6), 725e736. Buffie`re, P., Steyer, J.P., Fonade, C., Moletta, R., 1998. Modeling and experiments on the influence of biofilm size and mass transfer in a fluidized bed reactor for anaerobic digestion. Water Res. 32 (3), 657e668. Buffie`re, P., Bergeon, J.P., Moletta, R., 2000. The inverse turbulent bed: a novel bioreactor for anaerobic treatment. Water Res. 34 (2), 673e677. Buffie`re, P., Moletta, R., 2000a. Collision frequency and collisional particle pressure in three-phase fluidized beds. Chem. Eng. Sci. 55, 5555e5563. Buffie`re, P., Moletta, R., 2000b. Relations between carbon removal rates, biofilm size and density of a novel anaerobic reactor: the inverse turbulent bed. Water Sci. Technol. 41 (4e5), 253e260. Chai, S.L., Moletta, R., 2007. Anaerobic treatment of vinasses by a sequentially mixed moving bed biofilm reactor. Water Sci. Technol. 56 (2), 1e7. Cresson, R., 2006. Start-up of intensive anaerobic wastewater treatment processes. Impact hydrodynamic conditions and control strategy of organic load increase on biofilm formation and activity. PhD Thesis, University Montpellier II, (in French). Cresson, R., Carre`re, H., Delgene`s, J.P., Bernet, N., 2006. Biofilm formation during the start-up period of an anaerobic biofilm reactor e impact of nutrient complementation. Biochem. Eng. J. 30 (1), 55e62. Cresson, R., Dabert, P., Delgene`s, J.P., Bernet, N., 2007a. Inoculation of anaerobic biofilm reactors: use of molecular tools to study early biofilm formation. In: CD Proceedings of 11th World Congress on Anaerobic Digestion (AD11), Brisbane, Australia, September 23e27, pp. 6. Cresson, R., Escudie´, R., Carre`re, H., Delgene`s, J.P., Bernet, N., 2007b. Influence of hydrodynamic conditions on the start-up
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 e1 0
of methanogenic inverse turbulent bed bioreactors. Water Res. 41 (3), 603e612. Cresson, R., Escudie´, R., Steyer, J.P., Delgene`s, J.P., Bernet, N., 2008. Competition between planktonic and fixed microorganisms during the start-up of methanogenic biofilm reactors. Water Res. 42 (3), 792e800. Dabert, P., Delgene`s, J.P., Moletta, R., Godon, J.J., 2002. Contribution of molecular microbiology to the study in water pollution removal of microbial community dynamics. Rev. Environ. Sci. Biotechnol. 1 (1), 39e49. Escudie´, R., Epstein, N., 2008. Property-averaging applied to determination of volume contraction in binary-solid liquidfluidized beds. Can. J. Chem. Eng. 86 (3), 261e266. Escudie´, R., Epstein, N., 2009. Voidage at the layer inversion point in binary-solid liquid-fluidized beds. Ind. Eng. Chem. Res. 48, 182e184. Escudie´, R., Conte, T., Steyer, J.P., Delgene`s, J.P., 2005. Hydrodynamic and biokinetic models of an anaerobic fixedbed reactor. Process Biochem. 40, 2311e2323. Escudie´, R., Esptein, N., Grace, J.R., Bi, H.T., 2006a. Effect of particle shape on liquid-fluidized beds of binary (and ternary) solids mixtures: segregation vs. mixing. Chem. Eng. Sci. 61, 1528e1539. Escudie´, R., Esptein, N., Grace, J.R., Bi, H.T., 2006b. Layer inversion phenomenon in a binary-solid liquid fluidised bed: prediction of inversion velocity. Chem. Eng. Sci. 61, 6667e6690. Escudie´, R., Esptein, N., Bi, H.T., Grace, J.R., 2007. Layer inversion and bed contraction in downward binary-solid liquid fluidized beds. Can. J. Chem. Eng. 85, 1e11. Garcı´a-Bernet, D., Buffie`re, P., Moletta, R., Elmaleh, S., 1998. Application of the downflow fluidized bed to the anaerobic treatment of wine distillery wastewater. Water Sci. Technol. 38 (8/9), 393e399. Garcı´a-Caldero´n, D., Buffie`re, P., Moletta, R., Elmaleh, S., 1996. Comparison of three support materials for anaerobic fluidized bed systems. Biotechnol. Lett. 18 (6), 731e736. Garcı´a-Caldero´n, D., Buffie`re, P., Moletta, R., Elmaleh, S., 1998a. Anaerobic digestion of wine distillery wastewater in downflow fluidized bed. Water Res. 32 (12), 3593e3600. Garcı´a-Caldero´n, D., Buffie`re, P., Moletta, R., Elmaleh, S., 1998b. Influence of biomass accumulation on bed expansion characteristics of a down flow anaerobic fluidized bed reactor. Biotechnol. Bioeng. 57 (2), 136e144. Habouzit, F., Ge´vaudan, G., Hamelin, J., Bernet, N., 2009. Influence of support properties on initial adhesion of a methanogenic microbial consortium. In: CD Proceedings of the IWA Specialized Conference on Processes in Biofilms: Fundamentals to Applications, Davis, California, September 13e16. Hickey, R.F., Wu, W.M., Veiga, M.C., Jones, R., 1991. Start-up, operation, monitoring and control of high-rate anaerobic treatment systems. Water Sci. Technol. 24 (8), 207e255. Kim, M., Speece, R.E., 2002. Aerobic waste activated sludge (WAS) for start-up seed of mesophilic and thermophilic anaerobic digestion. Water Res. 36 (15), 3860e3866. Lauwers, A.M., Heinen, W., Gorris, L.G.M., van der Drift, C., 1990. Early stage in biofilm development in methanogenic fluidized bed reactors. Appl. Microbiol. Biotechnol. 33 (3), 352e358. Lazarova, V., Pierzo, V., Fontvielle, D., Manem, J., 1994. Integrated approach for biofilm characterisation and biomass activity control. Water Sci. Technol. 29 (7), 345e354. Lettinga, G., 1995. Anaerobic digestion and wastewater treatment systems. Antonie Van Leeuwenhoek 67, 3e28. Liu, Y., Lin, Y.M., Yang, S.F., Tay, J.H., 2003. A balanced model for biofilms developed at different growth and detachment forces. Process Biochem. 38 (12), 1761e1765. Liu, Y., Tay, J.H., 2004. State of the art of biogranulation technology for wastewater treatment. Biotechnol. Adv. 22 (7), 533e563.
9
van Loosdrecht, M.C.M., Eikelboom, D., Gjaltema, A., Mulder, A., Tijhuis, L., Heijnen, J.J., 1995. Biofilm structures. Water Sci. Technol. 32 (8), 35e43. van Loosdrecht, M.C.M., Heijnen, J.J., Eberl, H., Kreft, J., Picioreanu, C., 2002. Mathematical modelling of biofilm structures. Antonie Van Leeuwenhoek 81 (1e4), 245e256. Marin, P., Alkalay, D., Guerrero, L., Chamy, R., Schiappacasse, M.C., 1999. Design and startup of an anaerobic fluidized bed reactor. Water Sci. Technol. 40 (8), 63e70. McHugh, S., O’Reilly, C., Mahony, T., Colleran, E., O’Flaherty, V., 2003. Anaerobic granular sludge bioreactor technology. Rev. Environ. Sci. Biotechnol. 2 (2e4), 225e245. Michaud, S., Bernet, N., Buffie`re, P., Roustan, M., Moletta, R., 2002. Methane yield as a monitoring parameter for the start-up of anaerobic fixed film reactors. Water Res. 36 (5), 1385e1391. Michaud, S., Bernet, N., Roustan, M., Delgene`s, J.P., 2003. Influence of hydrodynamic conditions on biofilm behaviour in a methanogenic inverse turbulent bed reactor. Biotechnol. Progr. 19 (3), 858e863. Michaud, S., Bernet, N., Buffie`re, P., Delgene`s, J.P., 2005. Use of the methane yield to understand the metabolic behavior of methanogenic biofilms. Process Biochem. 40 (8), 2751e2755. Najafpour, G.D., Zinatizadeh, A.A.L., Mohamed, A.R., Isa, M.H., Nasrollahzadeh, H., 2006. High-rate anaerobic digestion of palm oil mill effluent in an upflow anaerobic sludge-fixed film bioreactor. Process Biochem. 41 (2), 370e379. Nicolella, C., van Loosdrecht, M.C.M., Heijnen, J.J., 2000. Wastewater treatment with particulate biofilm reactors. J. Biotechnol. 80 (1), 1e33. Picanc¸o, A.P., Vallero, M.V.G., Gianotti, E.P., Zalat, M., Blundi, C.E., 2001. Influence of porosity and composition of supports on the methanogenic biofilm characteristics developed in a fixed bed anaerobic reactor. Water Sci. Technol. 44 (4), 197e204. Pun˜al, A., Trevisan, M., Rozzi, A., Lema, J.M., 2000. Influence of C/ N ratio on the start-up of up-flow anaerobic filter reactors. Water Res. 34 (9), 2614e2619. Rajeshwari, K.V., Balakrishnan, M., Kansal, A., Lata, K., Kishore, V. V.N., 2000. State-of-the-art of anaerobic digestion technology for industrial wastewater treatment. Renew. Sust. Energ. Rev. 4 (2), 135e156. Rajinikanth, R., Ramirez, I., Steyer, J.P., Mehrotra, I., Kumar, P., Escudie´, R., Torrijos, M., 2008. Experimental and modeling investigations of a hybrid upflow anaerobic sludge-filter bed (UASFB) reactor. Water Sci. Technol. 58 (1), 109e117. Rajinikanth, R., Ganesh, R., Escudie´, R., Mehrotra, I., Kumar, P., Thanikal, J.V., Torrijos, M., 2009. High rate anaerobic filter with floating supports for the treatment of effluents from smallscale agro-food industries. Desal. Water Treat. 4 (1e3), 183e190. Richardson, J.F., Zaki, W.N., 1954. Sedimentation and fluidisation: part 1. Trans. Inst. Chem. Eng. Part A 32, 35e53. Ruggeri, B., Caire, G., Specchia, V., Sassi, G., Bosco, F., Gianetto, A., 1994. Determination of optimal biofilm activity in a biological fluidized reactor. Water Sci. Technol. 29 (10e11), 347e351. Safferman, S.I., Bishop, P.L., 1996. Aerobic fluidized bed reactor with internal media cleaning. J. Environ. Eng. 122, 284e291. Salkinoja-Salonen, M.S., Nyns, E.J., Sutton, P.M., van den Berg, L., Wheatley, A.D., 1983. Starting-up of an anaerobic fixed-film reactor. Water Sci. Technol. 15 (8e9), 305e308. Saravanan, V., Sreekrishnan, T.R., 2006. Modelling anaerobic biofilm reactors e a review. J. Environ. Manag. 81 (1), 1e18. Shieh, W.K., Sutton, P.M., Kos, P., 1981. Prediction reactor biomass concentration in a fluidized bed system. J. Water Pollut. Control Fed. 53, 1574e1584. Show, K.Y., Tay, J.H., 1999. Influence of support media on biomass growth and retention in anaerobic filters. Water Res. 33 (6), 1471e1481.
10
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 e1 0
Sowmeyan, R., Swaminathan, G., 2008. Evaluation of inverse anaerobic fluidized bed reactor for treating high strength organic wastewater. Bioresour. Technol. 99 (9), 3877e3880. Steyer, J.P., Bouvier, J.C., Conte, T., Gras, P., Harmand, J., Delgene`s, J. P., 2002a. Online measurements of COD, TOC, VFA, total and partial alkalinity in anaerobic digestion processes using infrared spectrometry. Water Sci. Technol. 45 (10), 133e138. Steyer, J.P., Bouvier, J.C., Conte, T., Gras, P., Sousbie, P., 2002b. Evaluation of a four year experience with a fully instrumented anaerobic digestion process. Water Sci. Technol. 45 (4e5), 495e502. Stoodley, P., Sauer, K., Davies, D.G., Costerton, J.W., 2002. Biofilm as complex differentiated communities. Annu. Rev. Microbiol. 56, 187e209. Tay, J.H., Show, K.Y., 1998. Media-induced hydraulic behavior and performance of upflow biofilters. J. Environ. Eng. 124 (8), 720e729. Tijhuis, L., van Loosdrecht, M.C.M., Heijnen, J.J., 1994. Formation and growth of heterotrophic aerobic biofilms on small suspended particles in airlift reactors. Biotechnol. Bioeng. 44 (5), 595e608. Tijhuis, L., Hijman, B., van Loosdrecht, M.C.M., Heijnen, J.J., 1996. Influence of detachment, substrate loading and reactor scale
on the formation of biofilms in airlift reactors. Appl. Microbiol. Biotechnol. 45, 7e17. Trinet, F., Heim, R., Amar, D., Chang, H.T., Rittmann, B.E., 1991. Study of biofilm and fluidisation of bioparticles in a threephase liquid fluidized bed reactor. Water Sci. Technol. 23 (7/9), 1347e1354. Wang, S., Rao, N.C., Qiu, R., Moletta, R., 2009. Performance and kinetic evaluation of anaerobic moving bed biofilm reactor for treating milk permeate from dairy industry. Bioresour. Technol. 100 (23), 5641e5647. Weiland, P., Rozzi, A., 1991. Start-up, operation, monitoring and control of high-rate anaerobic treatment systems: discusser’s report e anaerobic treatment technology for municipal and wastewater. Water Sci. Technol. 24 (8), 257e277. Yang, Y.N., Tada, C., Miah, M.S., Tsukahara, K., Yagishita, T., Sawayama, S., 2004. Influence of bed materials on methanogenic characteristics and immobilized microbes in anaerobic digester. Mater. Sci. Eng. C-Biomimetic Supramol. Syst. 24 (3), 413e419. Ye, F.X., Chen, Y.X., Feng, X.S., 2005. Advanced start-up of anaerobic attached film expanded bed reactor by pre-aeration of biofilm carrier. Bioresour. Technol. 96 (1), 115e119.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Review
Heterotrophic cultures of microalgae: Metabolism and potential products Octavio Perez-Garcia a, Froylan M.E. Escalante a, Luz E. de-Bashan a,b, Yoav Bashan a,b,* a
Environmental Microbiology Group, Northwestern Center for Biological Research (CIBNOR), Mar Bermejo 195, Col. Playa Palo de Santa Rita, La Paz, B.C.S. 23090, Mexico b The Bashan Foundation, 3740 NW Harrison Blvd., Corvallis, OR 97330, USA
article info
abstract
Article history:
This review analyzes the current state of a specific niche of microalgae cultivation;
Received 11 June 2010
heterotrophic growth in the dark supported by a carbon source replacing the traditional
Received in revised form
support of light energy. This unique ability of essentially photosynthetic microorganisms is
9 August 2010
shared by several species of microalgae. Where possible, heterotrophic growth overcomes
Accepted 20 August 2010
major limitations of producing useful products from microalgae: dependency on light
Available online 27 August 2010
which significantly complicates the process, increase costs, and reduced production of potentially useful products. As a general role, and in most cases, heterotrophic cultivation
Keywords:
is far cheaper, simpler to construct facilities, and easier than autotrophic cultivation to
Heterotrophic growth
maintain on a large scale. This capacity allows expansion of useful applications from
Carbon metabolism
diverse species that is now very limited as a result of elevated costs of autotrophy;
Nitrogen metabolism
consequently, exploitation of microalgae is restricted to small volume of high-value
Biofuel
products. Heterotrophic cultivation may allow large volume applications such as waste-
Lipids
water treatment combined, or separated, with production of biofuels. In this review, we
Microalgae
present a general perspective of the field, describing the specific cellular metabolisms
Pigments wastewater treatment
involved and the best-known examples from the literature and analyze the prospect of potential products from heterotrophic cultures. ª 2010 Elsevier Ltd. All rights reserved.
Contents 1. 2.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Nutrient metabolism by microalgae in heterotrophic culture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2.1. Key issues in heterotrophic growth of microalgae . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2.2. Carbon metabolism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2.2.1. Assimilation of glucose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2.2.2. Assimilation of glycerol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 2.2.3. Assimilation of acetate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 2.2.4. Wastewater and other carbon sources for heterotrophic growth of microalgae . . . . . . . . . . . . . . . . . . . . . . . 12
* Corresponding author. Environmental Microbiology Group, Northwestern Center for Biological Research (CIBNOR), Mar Bermejo 195, Col. Playa Palo de Santa Rita, La Paz, B.C.S. 23090, Mexico. Fax: þ52 612 125 4710. E-mail address:
[email protected] (Y. Bashan). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.037
12
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
Metabolism of nitrogen sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 2.3.1. Assimilation of ammonium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 2.3.2. Assimilation of nitrate and nitrite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 2.3.3. Assimilation of urea and organic nitrogen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 Metabolic products and processes using heterotrophic culture of microalgae . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 3.1. Lipids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 3.2. Polyunsaturated fatty acids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 3.3. Biodiesel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 3.4. Pigments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 3.4.1. Carotenoids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 3.5. Wastewater treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Concluding remarks and future prospects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 2.3.
3.
4.
1.
Introduction
Large-scale microalgal production has been studied for decades (Becker, 1994; Lee, 2001), given the wide variety of practical and potential metabolic products, such as food supplements, lipids, enzymes, biomass, polymers, toxins, pigments, tertiary wastewater treatment, and “green energy” products that can be obtained. These products were achieved by cultivating the microalgae on diverse mineral media, organic substrates, and synthetic or real wastewaters (Pulz, 2001; de-Bashan et al., 2002, 2004; Pulz and Gross, 2004; Lebeau and Robert, 2006; Harun et al., 2010). Today, the most common procedure for cultivation of microalgae is autotrophic growth. Because all microalgae are photosynthetic, and many microalgae are especially efficient solar energy convertors, microalgae are cultivated in illuminated environments naturally or artificially. Under autotrophic cultivation, the cells harvest light energy and use CO2 as a carbon source. The introduction of sufficient natural or artificial light to allow massive growth and dense populations is the main objective and a limiting factor of the cultivation: the more light, up to a limit for the species, the better (Mandalam and Palsson, 1998; Yang et al., 2000; Suh and Lee, 2003). Therefore, as practiced with other microbial communities producing economic products, open ponds that mimic natural environments of microalgae are the most common option for mass cultivation (Oswald, 1992; Tredici, 2004). Large open outdoor pond cultivation for mass algal production of single-cell protein, health food, and b-carotene (Borowitzka and Borowitzka, 1989; Wen and Chen, 2003; Carvalho et al., 2006; Chisti, 2007) is one of the oldest industrial systems since the 1950s (Oswald, 1992). Large open ponds can be built of glass, plastic, concrete, bricks, or compacted earth in a variety of shapes and sizes. The most common is the “raceway pond”, an oval form resembling a car-racing circuit (Lee, 2001; Pulz, 2001; Chisti, 2007). These cultivation systems present relatively low construction and operating costs and the large ones can be constructed on degraded and nonagricultural lands that avoid use of high-value lands and crop producing areas (Chen, 1996; Tredici, 2004). All these benefits notwithstanding, open ponds have several inherent disadvantages: (1) Poor light diffusion inside
the pond, decreasing with depth. It is aggravated when cultivation is intensive and causes self-shading. Consequently, shallow depth is required for ponds and they have a low volume to area ratio; (2) Mono-cultivation of the desired microalgae is difficult to maintain for most microalgae species because of constant airborne contamination, except for extremophile species; (3) Environmental growth parameters of cultivation rely primarily on local weather conditions, which may not be controlled and make production seasonal; (4) Harvesting is laborious, costly, and sometimes limited by low cell densities; (5) Continuous and clean water is needed; and (6) Production of pharmaceutical or food ingredients is not feasible or is very limited (Chen, 1996; Tredici, 1999; Molina Grima et al., 1999, 2003; Lee, 2001; Pulz, 2001; Wen and Chen, 2003; Sansawa and Endo, 2004; Carvalho et al., 2006; Chen and Chen, 2006; Chisti, 2007; Patil et al., 2008). To overcome inherent disadvantages of using open, less controlled environments, numerous closed photo-bioreactors (PBR) of various volumes and shapes have been designed (Molina Grima et al., 1999; Tredici, 1999, 2004; Tsygankov, 2000; Zhang et al., 2001; Barbosa, 2003; Suh and Lee, 2003; Zijffers et al., 2008). The principle final goal of any PBR is reduction in biomass production costs. This has been done by improving catalysts, shaping of the PBR, controlling environmental parameters during cultivation, aseptic designs, and operational approaches to overcome rate-limiting of growth, such as pH, temperature, and gas diffusion. Overcoming these limitations make monocultures and production of pharmaceutical and food goods possible (Cooney, 1983; Chen, 1996; Apt and Behrens, 1999; Pulz, 2001; Wen and Chen, 2003; Lebeau and Robert, 2006). Similar to the open-pond concept, large-scale PBRs have three major disadvantages that make them uneconomical for low-cost end-products: At operational volumes of 50e100 L or higher, it is no longer possible to disperse light efficiently and evenly inside the PBR (Chen, 1996; Pulz, 2001); development of algal biofilm fouls PBR surfaces and thereby limiting light penetration into the culture. A high initial investment in infrastructure and continuous maintenance is required (Carvalho et al., 2006). Nonetheless, numerous applications of PBR for microalgae were proposed and were reviewed (Apt and Behrens, 1999; Lebeau and Robert, 2006; Mun˜oz and Guieysse,
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
2006; Moreno-Garrido, 2008; Brennan and Owende, 2010; Harun et al., 2010). Therefore, this approach will not be discussed in this essay. A feasible alternative for phototrophic cultures in PBRs, but restricted to a few microalgal species, is the use of their heterotrophic growth capacity in the absence of light, replacing the fixation of atmospheric CO2 of autotrophic cultures with organic carbon sources dissolved in the culture media. Heterotrophy is defined as the use of organic compounds for growth (Droop, 1974). Heterotrophs are organisms whose substrate and energy needs are derived from organic compounds synthesized by other organisms (Kaplan et al., 1986). The basic culture medium composition for heterotrophic cultures is similar to the autotrophic culture with the sole exception of adding an organic carbon (Tsavalos and Day, 1994). Mixotrophic growth regime is a variant of the heterotrophic growth regime, where CO2 and organic carbon are simultaneously assimilated and both respiratory and photosynthetic metabolism operates concurrently (Kaplan et al., 1986; Lee, 2004). In some open-pond cultivations, organic carbon, such as acetate and glucose, are added continuously in small quantities. This is done to support higher microalgal biomass and simultaneously prevent excessive bacterial growth, which would be the outcome if the organic substrates were added in large quantity. Adding organic carbon substrate is usually done only during daytime hours, otherwise faster growing bacteria would outperform the microalgae under dark heterotrophic conditions. This fed-batch culture process is often limited to only one culture cycle to avoid bacterial contaminants from accumulating to unacceptable levels (Abeliovich and Weisman, 1978; Lee, 2001). Some microalgal species are not truly mixotrophs, but have the ability of switching between phototrophic and heterotrophic metabolisms, depending on environmental conditions (Kaplan et al., 1986). The heterotrophic growth approach eliminates the two major deficiencies of illuminated autotrophic PBR: allowing the use of practically any fermentor as a bioreactor, such as those used for industrial production of medicines, beverages, food additives, and energy and yielding, as a major outcome, a significant reduction in costs for most processes (Gladue and Maxey, 1994; Lee, 1997). Cost effectiveness and relative simplicity of operations and daily maintenance are the main attractions of the heterotrophic growth approach. A side but significant benefit is that it is possible to obtain, heterotrophically, high densities of microalgae cells that provides an economically feasible method for large scale, mass production cultivation (Chen, 1996; Lee, 2004; Behrens, 2005). For example: under some heterotrophic cultures, the growth rate, the dry biomass, ATP generated by the supplied energy, and the effect on ATP yield (mg of biomass generated by each mg of consumed ATP), lipid content and N content are significantly higher than under autotrophic cultures and are mainly dependent on the species and strain used (Martı´nez and Oru´s, 1991; Chen and Johns, 1996a,b; Ogbonna et al., 2000; Shi et al., 2000; Yang et al., 2000, 2002; Behrens, 2005; Boyle and Morgan, 2009). Under some heterotrophic growth conditions, the microalgal biomass yields are consistent and reproducible, reaching cells densities of 50e100 g of dry biomass per liter (Gladue and Maxey, 1994; Radmer and Parker, 1994), much higher than the maximum 30 g l1 of dry cell biomass in autotrophic cultures (Javanmardian and Palsson, 1991) and
13
comparable to the 130 g l1 of yeast dry biomass of commercial fermentors (Chen, 1996). Heterotrophic cultures containing as large as 100,000 l can generate useful biomass reaching hundreds of kilograms. These large volumes and high productivity of cultures make the heterotrophic strategy far less expensive than the autotrophic approach (Radmer and Parker, 1994). For example, in Japan, biomass production of Chlorella spp. use heterotrophic cultures to generate w500 ton of dry biomass, representing w50% of total Japanese production of this algae (Lee, 1997). Mixotrophic cultivation was also shown to be a good strategy to obtain a large biomass and high growth rates (Ogawa and Aiba, 1981; Lee and Lee, 2002), with the additional benefit of producing photosynthetic metabolites (Chen, 1996). Heterotrophic cultures have several major limitations: (1) There is a limited number of microalgal species that can grow heterotrophically; (2) Increasing energy expenses and costs by adding an organic substrate; (3) Contamination and competition with other microorganism; (4) Inhibition of growth by excess organic substrate; and (5) Inability to produce light-induced metabolites (Chen, 1996). Nonetheless, many recent studies show that heterotrophic cultures are gaining increasing interest for producing a wide variety of microalgal metabolites at all scales, from bench experiments to industrial scale (Apt and Behrens, 1999; Yang et al., 2000; Lee, 2001; Sansawa and Endo, 2004; Wen and Chen, 2001a, 2003; Li et al., 2007; Brennan and Owende, 2010). This review critically analyzes the processes and cases solely where heterotrophic cultivation of microalgae is possible to explore the potential and usefulness of this approach. It presents cases of autotrophic growth only for comparison or when similar mechanisms operate under autotrophic and heterotrophic conditions. It focuses on: (1) Basic metabolic processes of the microalgae; (2) Environmental parameters affecting growth and metabolism; (3) Kinetic parameters, such as specific growth rates and biomass production, and (4) Actual and potential end-products and byproducts that can be obtained from heterotrophic microalgal systems. Finally, we discuss some promising avenues of research.
2. Nutrient metabolism by microalgae in heterotrophic culture 2.1.
Key issues in heterotrophic growth of microalgae
Heterotrophic cultivation is inappropriate for most microalgae and more species are obligate autotrophs than facultative heterotrophs (Lee, 2001; Behrens, 2005). Yet, some species are effectively grown in complete darkness and thus their cultures can be grown in conventional dark fermenters. Chen and Chen (2006) listed the required initial characteristics that a microalgae species must have to be useful for heterotrophic cultivation: (a) Faculty of cell division and active metabolisms in absence of light. (b) Ability to grow in culture media with easy-to-sterile organic substrates where energy required for heterotrophic growth must be supplied by oxidation of part of the organic substrate (Droop, 1974). (c) Ability to adapt to fast
14
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
environmental changes, and (d) Capacity to resist hydromechanical stress inside the fermentors. In a broad sense, all organisms, including microalgae, use the same metabolic pathways for respiration. As expected, the metabolism of microalgae generally resembles, with only minor differences, that of higher plants. However, it is impossible to precisely predict which specific substrates can be used or preferred by any given microalgae (Neilson and Lewin, 1974). During respiration, oxygen is consumed and CO2 produced. The respiration rate of any organic substrate is intimately geared to growth and cell division. The rates of endogenous respiration and of O2 uptake vary through the cell cycle (Lloyd, 1974). Dark respiration rates (mol O2 mol carbon1 d1) increase with growth rates. Under optimal conditions, respiration rates are about 20e30% of growth rates (Geider and Osborne, 1989). In microalgae, dark respiration of an organic substrate assimilated from the medium has rates varying from 0.01 to 0.6 d1. This dark respiration plays two major roles in microalgae: (a) It serves as the exclusive source of energy for maintenance and biosynthesis under dark environment and (b) It provides essential carbon skeletons for biosynthesis under any growth condition. Physiological regulation of respiration is assumed to be controlled by demand for the products of respiration metabolism, such as energy in the form of ATP and NADH and carbon skeletons provided by the organic substrate (Geider and Osborne, 1989). Under heterotrophic growth conditions, respiration rates equal or exceed the theoretical minimum cost of biomass synthesis. Values for CO2 evolved per carbon (C) incorporated into new biomass (CO2/C) equaled 0.4e1.4 for several Chlorella species and diatoms. This indicates that biomass synthesis during heterotrophic growth conditions can proceed at nearly the maximal theoretical efficiency, since CO2/C ratios for autotrophic growth are much lower than values for heterotrophic growth (Raven, 1976). Independent of the supplied organic substrate or the microalgae species, growth rates are enhanced by higher levels of aeration (Griffiths et al., 1960). Oxygen supply is a key factor in heterotrophic cultivation of microalgae. For example, the limitation of oxygen in a culture may reduce the specific growth rate of Chlorella spp. and thus lower the productivity of biomass when cell density is high (Wu and Shi, 2007). Species of the genera Chlorella, Tetraselmis, and Nitzschia grew at higher rates under heterotrophic compared to autotrophic systems (Endo et al., 1974; Day et al., 1991; Gladue and Maxey, 1994; Chen and Johns, 1995; Lee, 2001; Shi and Chen, 2002; Boyle and Morgan, 2009). Additionally, under cyclic cultures of autotrophic/ heterotrophic conditions, cell production of biomass of Chlorella is about 5.5 times higher than under autotrophic cycles alone, where cells were producing 16 times more ATP under heterotrophic culture (Yang et al., 2000). In diatoms, heterotrophic growth is linked to their ability to maintain photosynthesis under dark environments using chloro-respiration to protect cells from photo damage after light returns; heterotrophic growth in this case is aided by high lipid accumulation, a product of reduced carbon in the absence of light (Wilhelm et al., 2006). In addition to the initial parameters for heterotrophic cultivation listed earlier (Chen and Chen, 2006), the main practical key issues in large-scale heterotrophic cultures of microalgae are: (a) Good survival of the strain during cultivation, (b) Its robustness, (c) Overall low cultivation costs,
reflected as the ability of the strain to efficiently use inexpensive, common carbon sources, tolerate environmental changes, and generate economical worth in the quantity of the metabolite(s), and (d) At the industrial level, the strains must also be easy to handle; its cell walls must withstand hydrodynamic and mechanical shear occurring in large bioreactors and it should produce high density biomass, all in minimally modified fermentors used for other microorganisms (Day et al., 1991; Gladue and Maxey, 1994; Chen and Chen, 2006). Consequently, these requirements reduce even further the microalgal strains that can be employed and use of available carbon sources. So far, the latter consists of glucose, glycerol, acetate, wastewater, and to a lesser extent, a few other organic carbon sources. Glucose is available to the great majority of heterotrophic algae and galactose and fructose are also somewhat used, but disaccharides are less generally available and, of the polyhydric alcohols, only glycerol is frequently used (Droop, 1974).
2.2.
Carbon metabolism
2.2.1.
Assimilation of glucose
Glucose is the most commonly used carbon source for heterotrophic cultures of microalgae, as is the case for many other microbial species. Far higher rates of growth and respiration are obtained with glucose than with any other substrate, such as sugars, sugar alcohols, sugar phosphates, organic acids, and monohydric alcohols (Griffiths et al., 1960). This may happen because glucose possesses more energy content per mol compared with other substrates. For example, glucose produces w2.8 kJ/mol of energy compared to w0.8 kJ/ mol for acetate (Boyle and Morgan, 2009). Glucose promoted physiological changes in Chlorella vulgaris, which strongly affects the metabolic pathways of carbon assimilation, size of the cells, volume densities of storage materials, such as starch and lipids grains (Martinez et al., 1991) and protein, chlorophyll, RNA, and vitamin contents (Endo et al., 1974). Oxidative assimilation of glucose begins with a phosphorylation of hexose, yielding glucose-6-phosphate, which is readily available for storage, cell synthesis, and respiration. An equivalent of a single phosphate bond is required per mole of glucose assimilated into glucose-6-phosphate. In that process, an additional 30 equivalents of phosphate bonds are generated by aerobic oxidation of a mole of glucose (Droop, 1974). Of the several pathways used by microorganisms for aerobic glycolysis (breakdown of glucose), apparently only two: the EmbdeneMeyerhof Pathway (EMP) and the Pentose Phosphate Pathway (PPP) have been shown in algae (Neilson and Lewin, 1974). Algae cannot metabolize glucose under anaerobic-dark conditions because insufficient energy is liberated during dissimilation of glucose and also retarded by low levels of the enzyme lactate dehydrogenase (EC 1.1.1.27), which is essential to complete the anaerobic fermentation process (Droop, 1974; Neilson and Lewin, 1974). Of 100% of glucose taken up by microalgae, about 1% remains as free glucose. More than 85% of the glucose is assimilated and converted to oligo-(mainly sucrose, w50%) and polysaccharides (mainly starch, w30%) (Tanner, 2000). Some microalgae species, such as Prymnesium parvum and Dunaliella tertiolecta are unable to assimilate glucose even though they possess the enzymes necessary for its metabolism (Neilson and Lewin, 1974).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
Probably the most notable difference in glucose metabolism in heterotrophic growth of microalgae, in comparison with autotrophic glucose metabolism or other non-carbohydrate organic substrates, is that under darkness, glucose is mainly
15
metabolized via the PPP pathway (Fig. 1, Table 1), while the EMP pathway is the main glycolytic process of cells in light conditions (Lloyd, 1974; Neilson and Lewin, 1974; Yang et al., 2000; Hong and Lee, 2007). Both pathways are carried out in the
Fig. 1 e Heterotrophic metabolism in microalgae. The enzymes are indicated as “gene nomenclature” similar to Table 1. Only routes important for heterotrophic growth are shown.
16
Table 1 e Enzymes and proteins regulated in different heterotrophic regimens compared to autotrophic conditions. Enzyme/Protein
E.C. #
EMP pathway (Glycolitic direction) Glucokinase
2.7.1.2
Glucose-6-phosphate isomerase 6-Phosphofructokinase
Glycerol assimilation
e»
Reference
«e
Yang et al. (2000, 2002)
fba1
«e
Yang et al. (2000, 2002)
gap1
e
Hilgarth et al. (1991); Yang et al. (2002)
pfk
e
Yang et al. (2002)
1,3-Biphosphoglycerate þ NADPH þ Hþ ¼> D-Glyceraldehyde-3-phosphate þ Orthophosphate þ NADPþ D-Fructose 1,6-bisphosphate þ H2O <¼> D-Fructose 6-phosphate þ Orthophosphate ATP þ Oxaloacetate <¼> ADP þ Phosphoenolpyruvate þ CO2
gap2
Y
Y
Aubert et al. (1994); Yang et al. (2002)
fbp
e
Y
Aubert et al. (1994); Yang et al. (2002) Boyle and Morgan (2009)
1.1.1.28
(D)-Lactate þ NADþ <¼> Pyruvate þ NADH þ Hþ
dlh
1.1.1.49
D-Glucose
1.2.1.12
6-phosphofructokinase
2.7.1.11
1.2.1.59
Fructose-1,6-bisphosphatase
3.1.3.11
Phosphoenolpyruvate carboxykinase (ATP)
4.1.1.49
6-phosphate þ NADPþ <¼>
»
pckA
Neilson and Lewin (1974); Garcia-Fernandez and Diez (2004).
gld1 or zwf
[»
»
Y
pgl
»
»
gnd
[»
»
Y
rpe or cfxE
e»
»
Y
D-Glucono-1,5-lactone
6-phosphate þ NADPH þ Hþ 6-phosphogluconolactonase
3.1.1.3
D-Glucono-1,5-lactone
6-phosphogluconate dehydrogenase
1.1.1.44
6-phosphate þ H2O <¼> 6-Phospho-D-gluconate 6-Phospho-D-gluconate þ NADPþ <¼> D-Ribulose 5-phosphate þ CO2 þ NADPH þ Hþ
Ribulose-phosphate 3-epimerase
5.1.3.1
D-Ribulose D-Xylulose
5-phosphate <¼> 5-phosphate
Aubert et al. (1994); Yang et al. (2000); Hong and Lee (2007); Boyle and Morgan (2009) Hong and Lee (2007); Boyle and Morgan (2009) Aubert et al. (1994); Yang et al. (2002); Hong and Lee (2007); Boyle and Morgan (2009) Aubert et al. (1994); Yang et al. (2002); Boyle and Morgan (2009)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
glk
Acetate assimilation
pfk2
2.7.1.11
Glyceraldehyde-3-phosphate dehydrogenase-NAD
Pentose phosphate pathway Glucose-6-phosphate dehydrogenase
ATP þ D-Glucose ¼> ADP þ D-Glucose 6-phosphate D-Glucose 6-phosphate <¼> D-Fructose 6-phosphate ATP þ D-Fructose 6-phosphate <¼> ADP þ D-Fructose 1,6-bisphosphate D-Fructose 1,6-bisphosphate <¼> Glycerone phosphate þ DGlyceraldehyde-3-phosphate D-Glyceraldehyde-3-phosphate þ Orthophosphate þ NADþ ¼> 1,3-Biphosphoglycerate þ NADH þ Hþ ATP þ D-Fructose 6-phosphate <¼> ADP þ D-Fructose 1,6-bisphosphate
Glucose assimilation
«e
5.3.1.9
4.1.2.13
Anaerobic fermentation D-lactate dehydrogenase (only present in Prochlorococcus spp.)
Gene
Yang et al. (2000, 2002); Hong and Lee (2007) Yang et al. (2000, 2002)
Fructose-bisphosphate aldolase
EMP pathway (Gluconeogenesis direction) Glyceraldehyde-3-phosphate dehydrogenase-NADP dependent
Reaction/Function
Glycerol metabolism Glycerol kinase sn-glycerol-3-phosphate NADþ oxidoreductase Triose-phosphate isomerase
TCA cycle Citrate synthase
2.7.1.30 1.1.1.8 5.3.1.1
2.3.3.1 4.2.1.3 1.1.1.41
SuccinateeCoA Ligase (ADP forming) Succinate dehydrogenase
6.2.1.5
Fumarate hydratase
4.2.1.2
Malate DehydrogenaseNADP dependent
1.1.1.37
Acetate assimilation and Glyoxylate cycle Acetyl-CoA synthetase
1.1.1.42 1.2.4.2
1.3.5.1
6.2.1.1
glpk
e
[»
Neilson and Lewin (1974)
gpd1
e
[»
Neilson and Lewin (1974)
tpic
e
[»
Neilson and Lewin (1974)
Acetyl-CoA þ H2O þ Oxaloacetate <¼> Citrate þ CoA Citrate <¼> Isocitrate Isocitrate þ NADþ <¼> a-Ketoglutarate þ CO2 þ NADH þ Hþ Isocitrate þ NADPþ <¼> a-Ketoglutarate þ CO2 þ NADPH þ Hþ a-ketoglutarate þ CoA þ NADþ <¼> SuccinyleCoA þ CO2 þ NADH þ Hþ ADP þ Orthophosphate þ Succinyl-CoA <¼> ATP þ Succinate þ CoA Ubiquinone þ Succinate <¼> Ubiquinol þ Fumarate Fumarate þ H2O <¼> (S)-Malate (S)-Malate þ NADPþ <¼> Oxaloacetate þ NADPH þ Hþ
cis
»
Neilson and Lewin (1974); Boyle and Morgan (2009) Boyle and Morgan (2009) Yang et al. (2002).
»
Neilson and Lewin (1974); Boyle and Morgan (2009) Neilson and Lewin (1974); Boyle and Morgan (2009) Neilson and Lewin (1974); Boyle and Morgan (2009) Neilson and Lewin (1974)
ATP þ Acetate þ CoA <¼> AMP þ Diphosphate þ Acetyl-CoA Isocitrate <¼> Succinate þ Glyoxylate
ach1 idh1 or icd
» » e»
idh2
»
ogd1
«
scla1
« »
sdh1 fum1 or citH
e»
»
mdh3
»
acs1
[»
icl
[» [»
»
Neilson and Lewin (1974); Yang et al. (2002) Boyle and Morgan (2009)
de Swaaf et al. (2003); Boyle and Morgan (2009) Neilson and Lewin (1974); Boyle and Morgan (2009) Neilson and Lewin (1974); Boyle and Morgan (2009)
Isocitrate lyase
4.1.3.1
Malate synthetase
2.3.3.9
Acetyl-CoA þ H2O þ Glyoxylate <¼> (S)-Malate þ CoA
mas1
4.1.1.39
D-Ribulose
1,5-bisphosphate þ CO2 þ H2O <¼> (2) 3-phospho-D-glycerate
rbcL
Y
Yang et al. (2002)
4.1.1.39
D-Ribulose
1,5-bisphosphate þ CO2 þ H2O <¼> (2) 3-phospho-D-glycerate
rbcS
Y
Yang et al. (2002)
2.7.1.19
ATP þ D-Ribulose 5-phosphate <¼> ADP þ D-Ribulose 1,5-bisphosphate H2O þ Phosphoenolpyruvate þ CO2 <¼> Orthophosphate þ Oxaloacetate
prk
e
Yang et al. (2002)
ppc
e»
Yang et al. (2002)
CalvineBenson Cycle (Carbon fixation) Ribulose bisophosphate carboxylase/oxygenase large subunit Ribulose bisophosphate carboxylase/oxygenase small subunit Phosphoribulokinase Phosphoenolpyruvate carboxylase
4.1.1.31
17
(continued on next page)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
Aconitate Hydratase Isocitrate dehydrogenaseNADH dependent Isocitrate dehydrogenaseNADPH dependent Ketoglutarate dehydrogenase
ATP þ Glycerol <¼> ADP þ sn-Glycerol 3-phosphate sn-Glycerol 3-phosphate þ NADþ <¼> Glycerone phosphate þ NADH þ Hþ Glycerone phosphate <¼> D-Glyceraldehyde-3-phosphate
18
Table 1 (continued) Enzyme/Protein
Fatty Acids Synthesis Malate dehydrogenase (oxalacetate decarboxylating)NADP dependent Pyruvate formate-lyase
E.C. #
Gene
Glucose assimilation
Acetate assimilation
Glycerol assimilation
Reference
(S)-Malate þ NADPþ <¼> Pyruvate þ CO2 þ NADPH þ Hþ
mme
»
Boyle and Morgan (2009)
2.3.1.54
CoA þ Pyruvate <¼> Acetyl-CoA þ Formate
pfl
»
Boyle and Morgan (2009)
Transport hexoses and protons with a stoichiometry of 1:1 trough the cell membrane investing 1 ATP Transport sugars and protons with a stoichiometry of 1:1 trough the cell membrane investing 1 ATP providing higher spectrum of sugars specificity uptake Transport sugars and protons with a stoichiometry of 1:1 trough the cell membrane investing 1 ATP providing higher spectrum of sugars specificity uptake Mitochondrial membrane hexose transport protein Adenine nucleotide translocator; ATP/ADP translocase Protein that aids transport of monocarboxylic (such as acetate) molecules across the membrane Proteins for ammonium transport across the cellular and chloroplastic membranes belonging to the ammonium transporter family 1(AMT1) Proteins for high affinity nitrate/nitrite transport across the membranes belonging to the ammonium transporter family, also presente in chloroplast Proteins for high affinity nitrate/ nitrite transport across the cellular membranes belonging to the transporter family
hup1
[
Komor and Tanner (1974); Sauer and Tanner (1989); Hilgarth et al. (1991)
hup2
[
Caspari et al. (1994)
hup3
[
Caspari et al. (1994)
hxt1
e
Merchant et al. (2007)
ant or aat
e
Hilgarth et al. (1991)
Hexose/Hþ symport system 2
Hexose/Hþ symport system 3
Hexose transport system ATP/ADP mitochondrial translocator Monocarboxylic/proton transporter Ammonium transporter proteins (AMT1)
Nitrate/nitrate transporter proteins (NAR1)
Nitrate/nitrate transporter proteins (NRT1,2)
6.3.1.2
ATP þ L-Glutamate þ NH3 ¼> ADP þ Orthophosphate þ L-Glutamine
[
mct1
Becker et al. (2005)
e
e
e
Wilhelm et al. (2006); Fernandez and Galvan (2007)
nar1
Y
Y
Y
Kamiya (1995); Fernandez and Galvan (2007)
nar1, nar2
Y
Y
Y
Kamiya (1995); Fernandez and Galvan (2007)
gln
[
e
e
Tischner (1984); Kaplan et al. (1986); Lu et al. (2005); Vanoni and Curti (2005)
amt1 (a,b,c,d,e,f,g,h)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
1.1.1.40
Transports Hexose/Hþ symport system 1
Nitrogen assimilation Glutamine synthetase (GS)
Reaction/Function
1.4.1.14
L-Glutamine
þ a-Ketoglutarate þ NADH þ Hþ ¼> (2) L-Glutamate þ NADþ
gsn1
e
e
e
Glutamate synthasee Ferredoxin dependent (GOGAT)
1.4.7.1
gsf1
e
e
e
Glutamate dehydrogenaseNADH dependent (GDH) Aspartate aminotransferase
1.4.1.3
gdh
e
e
e
ast
e
e
e
Asparagine synthetase
6.3.5.4
asns
Y
Y
Y
Nitrate reductasee NADH dependent Nitrite reductasee ferredoxin dependent
1.7.1.1
þ a-Ketoglutarate þ(2) Reduced ferredoxin þ (2) Hþ ¼> (2) L-Glutamate þ (2) Oxidized ferredoxin a-Ketoglutarate þ NH3 þ NADH þ Hþ <¼> L-Glutamate þ NADþ þ H2O Oxaloacetate þ L-Glutamate <¼> L-Aspartate þ 2-Oxoglutarate ATP þ L-Aspartate þ L-Glutamine þ H2O <¼> AMP þ Diphosphate þ L-Asparagine þ L-Glutamate Nitrate þ NADH þ Hþ ¼> Nitrite þ NADþ þ H2O Nitrite þ (6) Reduced ferredoxin þ (6) Hþ ¼> NH3 þ (2) H2O þ (6) Oxidized ferredoxin Urea þ H2O <¼> CO2 þ (2) NH3
nia2 or nr
Y
Y
Y
nit or nir
«
«
«
ure
e
e
e
2.6.1.1
1.7.7.1
L-Glutamine
Tischner (1984); Kaplan et al. (1986); Fernandez and Galvan (2007); Lu et al. (2005); Vanoni and Curti (2005) Tischner (1984); Lu et al. (2005); Vanoni and Curti (2005); Fernandez and Galvan (2007) Lea and Miflin (2003); Lu et al. (2005) Inokuchi et al. (2002); Coruzzi (2003) Inokuchi et al. (2002); Coruzzi (2003) Kamiya (1995); Kamiya and Saitoh (2002) Morris (1974); Kamiya (1995)
Kaplan et al. (1986); Oh-Hama and Miyachi (1992) Morris (1974); Kaplan et al. (1986); Oh-Hama and Miyachi (1992) Morris (1974)
Urease
3.5.1.5
Urea amidolyase
6.3.4.6
ATP þ Urea þ HCO3 <¼> ADP þ Orthophosphate þ Urea-1-carboxylate
dur
e
e
e
Allophanate hydrolase
3.5.1.54
Urea-1-carboxylate þ H2O <¼> (2) CO2 þ (2) NH3
atzF
e
e
e
3.1.26.5
Endonucleolytic cleavage of RNA, removing 50 -extranucleotides from tRNA precursor Blue-light photoreceptors whit photoinhibitory action over Hup1 gene. ATP þ Protein <¼> ADP þ Phosphoprotein Blue-light photoreceptors to photoinhibition of Hup1 gene
rnpB
e
Yang et al. (2002)
nph1
e»
cry1, 2
e»
Kamiya and Kowallik (1987b); Christie et al. (1998); Kamiya and Saitoh (2002) Kamiya and Kowallik (1987b); Kamiya and Saitoh (2002)
Gene regulation Ribonuclease P
NPH1 Flavoprotein (Autophosphorylating serinee threonine protein kinase action) Cryptochromes 1 and 2 Flavoproteines (ATP binding/protein homodimerization/ protein kinase action)
2.7.11.1
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
Glutamate synthaseeNADH deoendent (GOGAT)
Y Reduction of enzyme concentration and/or gene expression level (mRNA concentration) compared to autotrophic cultures. [ Increase of enzyme concentration and/or gene expression level (mRNA concentration) compared to autotrophic cultures. e Presence of enzyme or it’s mRNA but no changes in their concentration level compared to autotrophic cultures. « Reduction of metabolite flux rate of the reaction compared to autotrophic cultures. » Increase of metabolite flux rate of the reaction compared to autotrophic cultures.
19
20
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
cytosol and are functional in microalgae cells. However, the PPP pathway might have a higher flux rate than the other, depending on light and the presence of glucose. Some examples illustrate the process. In complete darkness and using glucose as sole carbon source, the PPP pathway in Chlorella pyrenoidosa (renamed Chlorella sorokiniana) accounts for 90% of glucose metabolic flux distribution via glucose-6-phosphate dehydrogenase (EC: 1.1.1.49) and the reaction catalyzed by glucose-6-phosphate isomerase (EC: 5.3.1.9) of the EMP pathway is totally “shifted down” (Yang et al., 2000). In heterotrophic culture of the cyanobacteria Synechocystis spp., the PPP was the major pathway of glucose catabolism via glucose-6-phosphate dehydrogenase and 6-phosphogluconate dehydrogenase (EC: 1.1.1.44) (Yang et al., 2002; Hong and Lee, 2007). However, the EMP glycolytic pathway is not completely shifted down, only the key reactions catalyzed by glucose-6phosphate isomerase (EC: 5.3.1.9), 6-phosphofructokinase (EC: 2.7.1.11), and fructose-bisphosphate aldolase (EC: 4.1.2.13) are affected in glucose assimilation, the other reactions of this pathway remain active such as in autotrophic growth (Yang et al., 2000, 2002; Hong and Lee, 2007). The Tricarboxylic Acid Cycle (TCA) and mitochondrial oxidative phosphorylation maintain high activities in cultures of C. pyrenoidosa, such as in illuminated environments, which indicate a minor effect of light on these pathways in this microalga species (Yang et al., 2000; Hong and Lee, 2007). At the same time, the flux through the pentose phosphate pathway during illumination was very small, resulting from light-mediated regulation. Heterotrophic culture of C. pyrenoidosa generated more ATP from the energy supplied as glucose than the autotrophic and mixotrophic cultures with energy supplied as light (Yang et al., 2000). In Synechocystis spp., a multi-level regulatory mechanism of the enzymes required for glucose metabolism depends on the energy source available to the cells and this depends on environmental conditions, transcription rates, metabolites availability, and flux requirements. In heterotrophic cultures, the expression of the genes rbcL (codes for ribulose bisophosphate carboxylase/oxygenase large subunit, EC: 4.1.1.39) and gap2 (glyceraldehyde-3-phosphate dehydrogenase-NADP; EC: 1.2.1.59), were down-regulated about two-fold by a light-regulated transcription mechanism, while the gene gnd (6-phosphogluconate dehydrogenase, EC: 1.1.1.44) was up-regulated about 60% in response to an apparent flux of substrate product of that enzyme because the system requires more of the product of that enzyme. In contrast, the expression of the genes prk, fbp, rnpB, glk, gap1, ppc, pfkA, icd, fum1 (that codes respectively for phosphoribulokinase e EC: 2.7.1.19, fructose1,6-bisphosphatase e EC: 3.1.3.11, ribonuclease P e EC: 3.1.26.5, glucokinase e EC: 2.7.1.2, glyceraldehyde-3-phosphate dehydrogenase e EC: 1.2.1.12, phosphoenolpyruvate carboxylase e EC: 4.1.1.31, 6-phosphofructokinase e EC: 2.7.1.11, isocitrate dehydrogenase e EC: 1.1.1.41, and fumarate hydratase e EC: 4.2.1.2) are not affected by the presence or absence of light or glucose, proving that also in cyanobacteria the TCA and many reactions of the EMP are actively independent of the energy and carbon sources for the culture (Yang et al., 2002). Compared to the mixotrophic condition, the mRNA levels of all the genes were not up or down-regulated significantly during autotrophic growth. The protein expression pattern under the autotrophic condition was very similar to that in the mixotrophic
condition; this means that the presence of glucose under illuminated conditions did not significantly alter the protein expression profiles (Yang et al., 2002). Compared to Synechocystis spp., other marine cyanobacteria, such as Prochlorococcus spp. have an incomplete TCA cycle metabolism lacking key enzyme genes such as those encoding for malate dehydrogenase (EC: 1.1.1.37) and succinyl coAeligase (EC: 6.2.1.5). However some strains of this genus possess alternatively the enzyme D-lactate dehydrogenase (EC: 1.1.1.28) that allows recovery of NADþ produced by glycolysis, while transforming pyruvate to lactate under anoxic environments (GarciaFernandez and Diez, 2004). Chlorella cells possess an inducible active hexose/Hþ symport system responsible for uptake of glucose from the medium (Tanner, 1969; Komor, 1973; Komor and Tanner, 1974, 1976). This mechanism transports sugars and protons with a stoichiometry of 1:1 (Komor et al., 1973) and the cell invests one molecule of ATP per molecule of sugar transported (Tanner, 2000). The transporter is completely inactive for all fluxes: influx, efflux, and exchange flux, between environment and cytosol when the intracellular pH is below 6.0 and is optimally active at an extracellular pH of 6.0 (Komor et al., 1979). In C. vulgaris growing with glucose as the inducer, the minimum time necessary to induce synthesis of the hexose/ Hþ symport system proteins is 15e18 min (Haass and Tanner, 1974; Komor and Tanner, 1974). Induction of the transporter is achieved by D-glucose, D-fructose, and D-galactose, but not by pentoses, sucrose, D-manose, disaccharides, or sugar alcohols (Komor et al., 1985). The gene corresponding to the hexose/Hþ symport system protein is the Hexose Uptake Protein Gene (hup1) (Sauer and Tanner, 1989). The HUP1 protein is a member of the Major Facilitator Superfamily (MFS) of transporter proteins. The structure of the hexose/Hþ symport system was reviewed in detail (Caspari et al., 1994; Tanner, 2000). The mRNA of the hup1 gene, absent in photosynthetically-grown cells, appears within 5 min after sugar is added (Hilgarth et al., 1991). The hup1 gene, the ATP/ADP translocator mitochondrial gene (aat), and the glyceraldehyde-3-phosphate dehydrogenase gene ( gap1) are activated when autotrophically grown Chlorella kessleri cells switch to heterotrophic growth in the presence of D-glucose. Uptake mutants (HUP1 ) do not respond to application of sugars in this way (Hilgarth et al., 1991). Chlorella cells in general possess two more hexose transporter genes (hup2 and hup3), co-induced by D-glucose. The other sugars provide a higher spectrum of sugars specificity uptake (Caspari et al., 1994). In Chlamydomonas reinhardtii, the hxt1 gene is codified for the hexose transporter protein through which glucose is transported through chloroplastic membranes (Merchant et al., 2007). Light plays a major role in glucose uptake. In C. vulgaris cells, light inhibits expression of the hexose/Hþ symport system. The blue end of the visible spectrum is very effective at inhibiting the uptake of hexoses where the red end is only slightly effective (Kamiya and Kowallik, 1987a). Because a similar photo-inhibiting effect occurred in a non-photosynthesizing mutant of C. vulgaris, this suggests that photosensitivity is independent of photosynthesis and is performed by the blue-light photoreceptors flavoproteins NPH1 and cryptochromes 1 and 2 (Kamiya and Kowallik, 1987b; Kamiya
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
and Saitoh, 2002). For Chlorella cells growing with glucose, the presence of blue light controls numerous metabolic reactions, such as inhibiting uptake of glycine, proline, and arginine, and ammonia, but enhancing uptake of oxygen and nitrate through activation of nitrate reductase (EC: 1.7.1.1) (Kamiya, 1995; Kamiya and Saitoh, 2002). Under complete darkness, glucose in the medium induces expression of two transmembrane amino acid transport systems. One is for transport of neutral amino acids, such as alanine, proline, serine, and glycine and the other is for transport of the basic amino acids, arginine and lysine. This induction enhances uptake of these amino acids about 5e10 times faster than the rates of uptake reported for any other plant cell in higher plants or algae (Cho et al., 1981). These enhancing effects were not observed under red or far-red illumination (Kamiya and Saitoh, 2002). When microalgae species and strains are able to grow under mixotrophic regimes, specific growth rates of the mixotrophic cultures is approximately the sum of the growth rates of cells under phototrophic and heterotrophic conditions. Consequently, only algal strains that are not sensitive to photoinhibition are suitable for mixotrophic cultivation (Lee, 2004). Although it is generally agreed that glucose can serve as a common carbon source, the precise effects of glucose on metabolism of microalgae varies greatly and these contradictions may lead to the conclusion that glucose uptake depends mainly on the quantity of light and the species of microalgae that is used. Several examples on the effect of glucose on oxygen, pH, and substrate concentration can illustrate this point. Depletion of glucose and fructose in cultures of Galdieria sulphuraria was accompanied by a rapid increase in concentration of dissolved oxygen in the culture resulting from diminished respiration rate caused by complete depletion of the original carbon source. An intermittent feeding method for the microalga was proposed. After the dissolved oxygen tension increased to >10%, a new batch of substrate was added. The concentration of sugar was kept sufficiently low to serve as the growth-limiting factor, although the total amount of glucose that was added was large (Schmidt et al., 2005). For Schizochytrium limacinum, the concentration of oxygen had no effect on growth. Differences in sugar consumption and dissolved oxygen in the medium could be attributed to the pH of the culture and to the strains that were used (Chi et al., 2007). For C. vulgaris growing on sufficient glucose, the hexose/Hþ symport system is induced to promote the alkalinization of the culture media by a net movement of protons accompanied glucose uptake. Since other sugars can be used, the velocity of the increase in pH depends on the concentration and type of sugar used (Komor and Tanner, 1974). Commonly, under low hexose concentrations, a decrease in pH and sugar consumption occurs. Sugar consumption is apparently linked to a net movement of protons in sugar translocation through the membrane, yielding a pH-shift (Komor and Tanner, 1974, 1976; Komor et al., 1985). Consequently, high concentrations of glucose and glycerol have been shown to inhibit microalgal growth, at least for a considerable period of time. This is the underlying reason of the adoption of the feed-batch configuration of bioreactor operation. This proposed configuration can maintain a constant low substrate concentration and avoid, in practical and easy ways, adverse effects on growth and end-
21
product yields (Tan and Johns, 1991; Cero´n Garcı´a et al., 2000; Wen and Chen, 2000; Schmidt et al., 2005; Xiong et al., 2008). This proposal creates a dilemma: How low is low? It is likely that this depends on the microalgal species and specific growth conditions. For example, to promote cellular growth of C. vulgaris and Scenedesmus acutus, the initial concentration of glucose should be limited to 10 g l1 and 1 g l1, respectively (Ogawa and Aiba, 1981). For optimal growth of Chlorella saccharophila, a concentration of glucose of 2.5 g l1 is required and inhibition occurred at concentrations >25 g l1; inhibition of C. sorokiniana occurs at concentrations >5 g l1 (Tan and Johns, 1991). Chlorella protothecoides has been cultivated at concentrations as high as 85 g l1 to obtain an optimal yield of biomass (Shi et al., 1999). In trials with Nitzschia laevis, yields decreased as concentration increased from 1 to 40 g l1 (Wen and Chen, 2000). G. sulphuraria grown with high concentrations of glucose or fructose up to 166 g l1 (0.9 M) continued to thrive, but higher concentrations inhibited growth (Schmidt et al., 2005). In summary, information on the concentration of glucose required for optimal metabolic growth is too scattered to reach a definite conclusion. The answer may be related to specific combinations of factors, with the microalgal species as the main factor and cultivation and environmental conditions as secondary factor. Consequently, each combination of factors may lead to different consumption levels. From the data published, it appears that glucose might be considered a “preferred substrate” for heterotrophic cultivation of microalgae. Microalgal cells grown on other substrates require a lag period (an acclimation period) to develop the specific transport systems necessary for uptake of other substrates. Consumption of “less preferred” substrates is aborted because the enzymes that catalyze uptake of an alternate substrate cannot be synthesized in the presence of the “preferred” substrate (Lewin and Hellebust, 1978; Ratledge et al., 2001; Narang and Pilyugin, 2005). This lag phenomenon may not always occur because other factors, such as the strain used, bioreactor configuration, and environmental conditions will have a profound impact on uptake of alternative substrates.
2.2.2.
Assimilation of glycerol
Heterotrophic growth using glycerol as a substrate has been demonstrated for several algae, despite the simplicity of glucose metabolism in microalgae (Table 1). Most of these species occur naturally in habitats of somewhat elevated osmolarity, such as seawater and saline pounds (Neilson and Lewin, 1974). Glycerol as an osmoticum (a substance that has the capacity of raising the osmotic strength of the solution and consequently keeps the osmotic equilibrium in cells) is an economical carbon source for an energy supply and carbon requirements and is a very compatible solute for enzymes and membranes, with almost no toxic effects even at high concentrations (Richmond, 1986). It is commonly used for longterm preservation of microorganisms at low temperatures. Microalgae can produce glycerol as part of the glycerolipid metabolism because it is a product of hydrolysis of many lipids that are glyceryl esters of fatty acids (Leo´n and Galva´n, 1999). A few species can assimilate glycerol from the medium, where it increased their growth rate and induced specific biochemical and structural changes in their photosynthetic system, such as reduction of cell phycoerythrin content, degree of tylakoid
22
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
packing, number of tylakoids per cell, and the size of Photo System II particles (Lewitus et al., 1991). When plant cells are growing on a glycerol substrate, it enters the cell by simple diffusion (Neilson and Lewin, 1974). Inside cells, glycerol is used as an osmoregulatory molecule. Glycerol is first phosphorylated using ATP and the glycerophosphate is then oxidized to triose phosphate. Plant cells possess glycerol kinase (EC 2.7.1.30), sn-glycerol-3-phosphate NADþ oxidoreductase (EC 1.1.1.8) and and triose-phosphate (EC: 5.3.1.1) to convert glycerol into glyceraldehyde-3-phosphate and glycerate, which are intermediates in the EMP pathway of glycolysis to form pyruvate that enters the TCA cycle (Neilson and Lewin, 1974). Glyceraldehyde-3-phosphate may also be formed by reduction of 3-phosphoglycerate, a key intermediate of the CalvineBenson cycle of photosynthesis. It has been demonstrated that sn-glycerol 3-phosphate inhibits the reversible glycolytic pathway, as expected in the gluconeogenesis pathway. Aubert et al. (1994) suggest that the pentose phosphate pathway is also inhibited when glycerol is the unique carbon source (Fig. 1). Glycerol can be photometabolized (photoheterotrophy) by some algae species, such as Agmenellum quadruplicatum, Goniotrichium elegans, Navicula pelliculosa, Nostoc sp. These species can only assimilate glycerol as a carbon source, in the presence of light and without an external supply of CO2 (Ingram et al., 1973; Kaplan et al., 1986). Glycerol and light were used as substrates for cultivation of mixotrophic microalgae, yielding significant positive results. For example, in a culture media supplemented with 0.1 M glycerol and 165 mmol photons m2 s1, Phaeodactylum tricornutum increased its growth 74% more compared to autotrophic culture. However, a pronounced lag phase occurred, as explained above for growth on substrates other than glucose (Cero´n Garcı´a et al., 2000). Nannochloropsis sp., Rhodomonas reticulate, and Cyclotella cryptica seem to prefer glycerol over glucose or acetate by using mixotrophic metabolism and positively responding to environmental changes, such as when a nitrate is added to the medium (Wood et al., 1999). When C. vulgaris was immobilized in alginate with Azospirillum brasilense (a microalgae growth-promoting bacteria; MGPB) and grown autotrophically on synthetic wastewater growth medium (SWGM), A. brasilense mitigated environmental stress for the microalgae (de-Bashan et al., 2002). In another study, major cell growth occurred at pH 8 for A. brasilense immobilized with Chlorella, compared to Chlorella grown alone under autotrophic conditions (de-Bashan et al., 2005). In similar experiments using joint immobilization carried out under heterotrophic conditions, an eight-fold increase in the growth of C. vulgaris in SWGM containing 0.17 M glycerol after culturing for 24 h, compared to cultures with C. vulgaris immobilized alone under the same conditions, where there was no growth. Similar growth was obtained at pH 8 compared to pH 6 and 7 (Escalante F.M.E., unpublished data). This suggests that A. brasilense plays a major role in changing the metabolic behavior of Chlorella under autothrophic or heterotrophic conditions. In conclusion, although glycerol can serve as a substrate for heterotrophic growth, knowledge of metabolism under heterotrophic conditions is limited. With a potential for biodiesel production from microalgae (discussed later) where
glycerol is a major by-product and a substrate of the process, this niche probably will be revived.
2.2.3.
Assimilation of acetate
Uptake of dissolved carboxylic acids, such as acetic, citric, fumaric, glycolic, lactic, malic, pyruvic, and succinic under microalgal heterotrophic cultivation has been well known for decades (Bollman and Robinson, 1977). Acetate (or acetic acid) is one of the most common carbon sources for many microbial species, including microalgae (Droop, 1974). Under dark, aerobic conditions, eukaryotic cells uptake acetate using the monocarboxylic/proton transporter protein that aids transport of monocarboxylic molecules across the membrane. This protein is a member of the Major Facilitator Superfamily mentioned earlier (Becker et al., 2005). Once inside microalgal cells in the cytosol, the starting point for acetate assimilation is acetylation of coenzyme A by acetyl-CoA synthetase (EC 6.2.1.1) to form acetyl coenzyme A (acetyl-CoA) in a single-step catalyzed reaction using a single ATP molecule, as shown in Fig. 1 (Droop, 1974; de Swaaf et al., 2003; Boyle and Morgan, 2009). Acetate (carried by coenzyme A) is generally oxidized metabolically through two pathways: (a) the glyoxylate cycle to form malate in glyoxysomes (specialized plastids in the glyoxylate cycle) (Table 1) and (b) through the tricarboxylic acid cycle (TCA) to citrate in the mitochondria, which provides carbon skeletons, energy as ATP, and energy for reduction (NADH). Many of the intermediates of both cycles are the same metabolites (Neilson and Lewin, 1974; Ahmad and Hellebust, 1990; Boyle and Morgan, 2009). In general, microalgae that grow by assimilating acetate must possess a glyoxylate cycle pathway to efficiently incorporate acetyl groups of acetyl-CoA to carbon skeletons. The operation of the glyoxylate cycle requires synthesis of isocitrate lyase (EC 4.1.3.1) and malate synthetase (EC 2.3.3.9). Both enzymes are induced when cells are transferred to media containing acetate (Neilson and Lewin, 1974; Boyle and Morgan, 2009). In C. vulgaris, isocitrate lyase, the key enzyme of the glyoxylate cycle, is synthesized constitutively, but the glyoxylate cycle is functional only during growth on acetate (Harrop and Kornberg, 1966). In Scenedesmus obliquus, activity of isocitrate lyase showed nearly a four-fold increase in activity after 24 h in the dark in the presence of acetate. Under heterotrophic conditions, isocitrate lyase activity increased as a function of increasing acetate concentration (Combres et al., 1994). In Chlorella spp. growing on acetate in the dark, the glyoxylate cycle enzymes are induced but not the TCA cycle enzymes; the latter pathway remains active, but its activity is not enhanced. Light and glucose suppress the formation of isocitrate lyase (Goulding and Merrett, 1967). In assimilation by the glyoxylate cycle, 4 mol of acetate are required per mole of glucose-6-phosphate that is synthesized, one of the four being consumed in the process (Droop, 1974). In C. reinhardtii growing on acetate, the oxidative pentose phosphate pathway is also active, providing reducing power as NADPH for cytosol (Boyle and Morgan, 2009). Additionally, mitochondrial and chloroplastic electron transport chains are active in these cells and have a close interaction through the glycolytic pathway (Rebeille and Gans, 1988). When sodium or potassium salt of acetate is used as a substrate, the pH rises. This happens because the remaining Naþ or Kþ couples with hydroxyl ions (OH) or other
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
anions to form alkalis. This phenomenon also occurs if reactors are pH-neutralized with alkali. Since metallic hydroxides are stronger than organic acids, the media must be neutralized or at least brought to a non-inhibitory pH level by adding an acid, acetic acid for instance, into the cycle (Ratledge et al., 2001). However, acetate does not always promote growth. It could be toxic for many microorganisms at high concentrations, despite its common use for buffering high pH levels in bioreactors, with the exception of Chlamydomonas mundana, which grew rapidly using acetate (Maciasr and Eppley, 1963; Wood et al., 1999). Keeping the concentration of acetate at low levels is useful for the fed-batch configuration in cultures or pH-auxostat (pH is maintained as a constant). In this way, as long as acetate is consumed, more acetic acid is added to the reactor, which avoids the rise in pH at the time that more acetate is added (Wood et al., 1999; Zhang et al., 1999; de Swaaf et al., 2003; Sijtsma et al., 2005). Although acetate fermentation in pH-auxostats is linked to succinic acid production that also inhibits growth, addition of propionate was suggested to provide oxaloacetate to the cells and improve cellular growth (Fig. 1). No explanation was provided on the mechanism by which propionate would alleviate inhibition of acetate in microalgae growing in this substrate. However, since propionate is a precursor of oxaloacetate, a possible explanation might be found in work with the bacterium Pseudomonas citronellolis. One of the requirements for efficient oxidation of the carbon source through the TCA cycle is a proper balance between oxaloacetate and acetyl-CoA. It is likely that elevated concentrations of acetyl-CoA in cells cultured with acetate would inhibit oxaloacetate decarboxylase. Hence, to maintain the supply of oxaloacetate, it is necessary to metabolize the acetate via the TCA cycle. Addition of propionate to the culture medium would lead to oxaloacetate production by alternative pathways (O’Brien and Taylor, 1977). For further details on acetate metabolism, see Sijtsma et al. (2005). Several examples on growing microalgae on acetate are known. Euglena gracilis strain L incorporates acetate efficiently under light but not in the dark (Cook, 1967). On the other hand, E. gracilis var. bacillaris incorporates acetate in the dark when its concentration is below 5 g l1. This strain was able to use a variety of substrates in heterotrophic cultures, such as acetate, sucrose, ethanol, amino acids, butyric acid, among a few other organic substrates (Cook, 1968). Crypthecodinium cohnii is able to grow in heterotrophic cultures with acetate concentrations as high as 1 g l1. No higher concentrations were tested (Vazhappilly and Chen, 1998). Another study with this strain achieved good growth when cultured in a pHcontrolled, pH-auxostat with 8 g l1 of sodium acetate (Ratledge et al., 2001). In a recent study (Perez-Garcia et al., in press), when 10 g l1 sodium acetate was added to municipal wastewater with C. vulgaris for tertiary treatment, significant heterotrophic growth occurred; however, this did not happen when calcium acetate was added. It seems that as long as the level of acetate is low and remains low, several microalgae can use it as its sole carbon source. This is specifically important because acetate is a readily available and inexpensive substrate derived from many industrial applications and its use does not impose severe restrictions on culturing microalgae.
23
2.2.4. Wastewater and other carbon sources for heterotrophic growth of microalgae One commonly proposed application of autotrophically grown microalgae is wastewater treatment (de la Nou¨e and Proulx, 1988; de la Nou¨e et al., 1992; Oswald, 1992; Can˜izares et al., 1994; Gonzalez et al., 1997; Lee and Lee, 2001; de-Bashan et al., 2002, 2004; Hernandez et al., 2006). The major advantages of these treatments are that additional pollution is not generated when the biomass is harvested and efficient recycling of nutrients is possible (de la Nou¨e et al., 1992). To date, this has hardly been tested under heterotrophic conditions (de-Bashan and Bashan, 2010). Nonetheless, Chlorella spp. and strains of Scenedesmus were isolated from wastewaters kept in the dark and in oxidation ponds (Abeliovich and Weisman, 1978; Lalucat et al., 1984; Post et al., 1994). C. pyrenoidosa growing in sterilized sewage were able to use some of the organic matter, as indicated by a decrease in soluble BOD and dissolved volatile solids in cultures of short retention times. Use of organic compounds was influenced by the supply of CO2 to the culture; decrease in the organic matter per unit of cell weight produced was greater when the supply of CO2 was low (Pipes and Gotaas, 1960). Growth characteristics and removal of nutrients from synthetic wastewater with high acetate and propionate concentrations were investigated under heterotrophic and mixotrophic conditions for Rhodobacter sphaeroides, C. sorokiniana, and Spirulina platensis. Heterotrophic cultures of R. sphaeroides and C. sorokiniana produced the best results under dark conditions but S. platensis required light. Neither growth nor removal of nutrients by the cells were affected in synthetic wastewater with as high as 10 000 mg l1 acetate, 1000 mg l1 propionate, 700 mg l1 nitrate and 100 mg l1 phosphate (Ogbonna et al., 2000). Recently, Perez-Garcia et al. (in press) found that adding several carbon sources to municipal wastewater that normally do not support microalgal growth allowed heterotrophic growth of C. vulgaris. Growth effects, in declining order, was Na-acetate, D-glucose, D-fructose ¼ fulvic acid, Nacitrate ¼ lactic acid ¼ acetic acid, malic acid, and L-arabinose. Other carbon sources such as sucrose, lactate, lactose, and ethanol have been tested under heterotrophic microalgae cultures with negative results in growth and metabolite production (Theriault, 1965; Lewin and Hellebust, 1978; Ogbonna et al., 1998; Schmidt et al., 2005; Wang and Peng, 2008). It appears that microalgae do not have invertase to assimilate sucrose. Komor et al. (1985) report that disaccharides connected to carbon 1 (sucrose) or carbon 4 (maltose) are not transported; consequently, sucrose uptake by C. pyrenoidosa (Theriault, 1965) and Chlorella zofingiensis (Wang and Peng, 2008) is poor. Schmidt et al. (2005) show that G. sulphuraria had significant growth on a sucrose substrate, but only at pH 2. It is likely that sucrose was hydrolyzed into glucose and fructose, which are readily assimilated by microalgae. While several carbon sources were proposed for heterotrophic growth of microalgae, practical evaluation of the carbon sources show that only a few substrates are supported by solid evidence. Those include glucose, glycerol, and acetate in wastewater. None of the other carbon sources tested supported sufficient growth. At this juncture, there are no other candidates for additional studies of growth. What is still pending is information about industrial wastes such as
24
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
molasses, vinegar, pharmaceutical by-products, and paper mill effluents. They contain the assimilated low molecular substrates that microalgae can use and can be mixed with water to create a substrate for microalgae. If feasible, these mixtures could be designated “prepared wastewater substrate”.
2.3.
Metabolism of nitrogen sources
After carbon, and apart from hydrogen and oxygen, nitrogen is quantitatively the most important element contributing to the dry matter of microalgal cells, accounting from 1 to 10% dry weight. This excludes diatoms, where silicon plays the more important role instead of nitrogen (for review, Martin-Je´ze´quel et al., 2000) and nitrogen-deficient microalgae that accumulate oils or polysaccharides (Kaplan et al., 1986). Carbon and nitrogen metabolism are linked in microalgae because they share (a) carbon supplied directly from respiration of fixed CO2 (autotrophic growth) or assimilated organic carbon (heterotrophic growth) and (b) the energy generated in the TCA cycle and in the mitochondrial electron transport chain. The primary assimilation of inorganic nitrogen (ammonium) to form amino acids requires carbon skeletons in the form of keto-acids (2oxaloglutarate and oxaloacetate) and energy in the form of ATP and NADPH to synthesize the amino acids glutamate, glutamine, and aspartate. In both autotrophic and heterotrophic growing cells, keto-acids, ATP, and NADPH are provided by the TCA cycle (Huppe and Turpin, 1994; Inokuchi et al., 2002; Lea and Miflin, 2003; Fernandez and Galvan, 2007). Very small quantities of keto-acids were found in Chlorella spp. when grown autothrophically, but the levels were much higher under heterotrophic conditions and nitrogen starvation (Millbank, 1957). Respiration rates appear to be limited indirectly by the supply of inorganic nitrogen through the demand of carbon skeletons. This happens following conditions in which intracellular carbohydrate energy reserves can accumulate, such as under limited nitrogen when carbon is not a limiting factor (Geider and Osborne, 1989). In general, nitrogen has a marked positive effect on growth and a negative effect on lipid accumulation. Microalgae are able to assimilate a variety of nitrogen sources, mainly ammonia (NHþ 4 ), nitrate (NO3 ), and urea, as well as yeast extract, peptone, amino acids, and purines (Oh-Hama and Miyachi, 1992; Armbrust et al., 2004; Chen and Chen, 2006; Wilhelm et al., 2006; Ganuza et al., 2008). The metabolic pathways involved in nitrogen assimilation are depicted in Fig. 1.
2.3.1.
Assimilation of ammonium
Ammonium is the most preferred nitrogen source for algae. It is also the most energetically efficient source, since less energy is required for its uptake (Syreth and Morris, 1963; Goldman, 1976; Kaplan et al., 1986; Shi et al., 2000; Grobbelaar, 2004; Wilhelm et al., 2006). Under autotrophic and heterotrophic conditions, ammonium is transported across the membranes by a group of proteins belonging to the ammonium transporter family (AMT), a group of evolutionarily related proteins commonly found in bacteria, yeast, algae, and higher plants (Wilhelm et al., 2006). Several ammonium transporters, all belonging to the AMT family, have been identified in diatoms (Allen
et al., 2005). An ample array of transporters for inorganic nitrogen compounds have been identified in Chlamydomonas sp., 8 putative ammonium transporters and 13 putative nitrate/nitrite transporters (Fernandez and Galvan, 2007). Ammonium transporters can be divided into two distinct systems: a high affinity system regulated by the nitrogen status of cells and a low-affinity system that exhibits a linear increase in activity in response to increases in ammonium concentration (Howitt and Udvardi, 2000). There are exceptions. Ammonium transporters in Cylindrotheca fusiformis and P. tricornutum are not only up-regulated by nitrogen limitation, but are also expressed at a higher level when grown on nitrate, compared to ammonium (Hildebrand, 2005). Ammonium is present in all compartments of the cell. Its concentration varies, depending on several factors including the concentration of ammonium in the neighboring compartment(s), the diferences in pH, and electrical potential between compartments. In compartments where ammonium is not metabolized, such as the vacuole, the concentration of ammonium may approach its equilibrium value. In compartments in which ammonium is metabolized, such as the cytosol and plastids, the steady-state concentration of ammonium may be much lower than the predicted equilibrium (Howitt and Udvardi, 2000). Dark respiration of nitrogenstarved microalgae cells is correlated with inorganic nitrogen assimilation. Ammonium-enhanced respiration continued until either ammonia concentration in the suspending medium dropped to an undetectable concentration or intracellular carbohydrate energy reserves were almost completely exhausted. Addition of glucose will allow ammonium assimilation to continue, as well as amino acid and protein synthesis (Geider and Osborne, 1989). Assimilation metabolism of ammonium under either authotrophic or heterotrophic conditions is catalyzed by glutamine synthetase (GS; EC 6.3.1.2), which produces glutamine, and glutamate synthase (GOGAT; EC 1.4.1.14), which produces two molecules of glutamate from glutamine plus one molecule of a-ketoglutarate (Tischner, 1984; Kaplan et al., 1986; Lu et al., 2005; Vanoni and Curti, 2005) (Fig. 1). Alternatively, ammonium is incorporated into glutamate by the reversible reductive amination of a-ketoglutarate, which is catalyzed by glutamate dehydrogenase (GDH, EC 1.4.1.2) (Inokuchi et al., 2002). The GS/GOGAT pathway is regarded as the primary pathway for ammonium assimilation, while the GDH pathway plays an insignificant part in the formation of glutamate. However, the evidence suggests an important role for GDH as a catabolic shunt to ensure that nitrogen metabolism does not affect mitochondrial function and to enable synthesis of nitrogen-rich transport compounds during nitrogen remobilization (Lea and Miflin, 2003). Additionally, GDH is believed to be active under conditions of stress (Lu et al., 2005). Glutamine synthetase, known for its high affinity for ammonia and its ability to incorporate ammonia efficiently into amino acids (Miflin and Habash, 2002) is an important enzyme in any photosynthetic organism, even under heterotrophic metabolism. This enzyme plays a dual role by providing glutamine for biosynthesis and by assimilating ammonia (Rahman et al., 1997; de-Bashan et al., 2008). Following incorporation of ammonium into glutamate through either the GS/GOGAT cycle or GDH, nitrogen is
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
distributed to the other amino acids, much of it through transamination with oxaloacetate by aspartate aminotransferase (EC: 2.6.1.1) to yield aspartate. Through an ATPdependent reaction catalyzed by asparagine synthetase (EC: 6.3.5.4), an amino group from glutamine is transferred to a molecule of aspartate to generate a molecule of glutamate and asparagine. Glutamine, glutamate, aspartate, and asparagine provide the building blocks for the synthesis of organic nitrogen compounds, such as amino acids, nucleotides, chlorophylls, polyamines, and alkaloids (Inokuchi et al., 2002; Coruzzi, 2003). Heterotrophic growth conditions do not affect uptake rates of ammonium and the expression of nitrogen assimilation enzymes but mixotrophic regimen does. For example, adding acetate to autotrophic Scenedesmus obliquus affects its rates of ammonium uptake. In autotrophy, uptake is 17.8 mmol cell1 min1 and is similar to that in heterotrophy (17.4 mmol cell1 min1), but this is w4 times lower than occurring under mixotrophy (65.9 mmol cell1 min1) (Combres et al., 1994). R. sphaeroides and C. sorokiniana showed acceptable growth in darkness in synthetic wastewater supplemented with 10 g l1 acetate and containing 400 mg l1 of ammonia, while S. platensis was completely inhibited under these conditions (Ogbonna et al., 2000). The nutritional status of the cells affects ammonium uptake. Nitrogen-limited C. sorokiniana, without organic carbon in the medium, exhibited respiratory oxygen consumption (70%) and photosynthetic oxygen evolution (17%), of cells with sufficient nitrogen. Cells with sufficient nitrogen absorbed NHþ 4 in light at a linear rate, but absorption was totally inhibited by darkness. In contrast, cells with limited nitrogen absorbed NHþ 4 at almost similar rates in light and darkness (Di Martino Rigano et al., 1998). C. kessleri successfully removes high concentrations of ammonium or nitrate from synthetic wastewater that is supplemented with glucose (Lee and Lee, 2002). A preference for ammonium has clearly been demonstrated for Chlorella spp. and Dunalliela spp., which can use a large variety of organic and inorganic nitrogen compounds, mainly ammonium and nitrate salts, and sometimes urea (Morris, 1974; Kaplan et al., 1986). For example, C. sorokiniana has as much as seven ammonium-inducible chloroplastic GDH isozymes composed of varying ratios of a- and bsubunits (Miller et al., 1998), indicating a wide spectrum of adaptation to different environmental conditions. When ammonium and nitrate are supplied together, Chlorella spp. preferentially uses ammonium first, which is incorporated into the organic compounds produced by the microalgae. C. protothecoides, N. laevis, and P. tricornutum exhibit a preference for nitrate or urea over ammonium. This happens when the pH is lowered by consumption of ammonium that induced severe reduction of growth and biomass yields when pH was not controlled (Yongmanitchai and Ward, 1991; Shi et al., 2000; Wen and Chen, 2001a,b; Lee and Lee, 2002). However, when the pH of the culture and other growth conditions are controlled, ammonium is a reliable nitrogen source (de-Bashan et al., 2005). For example, P. tricornutum grew well after adjusting the initial pH to 8 and a fed-batch configuration was used (Cero´n Garcı´a et al., 2000). Another option to control pH and use ammonium as a nitrogen sources is to add a buffer. Using the same species, adding Tris buffer to
25
the culture medium avoided a severe drop in pH; yet, some inhibition of growth occurred even when the buffer alleviated the side effects of ammonium consumption or pH drop (Yongmanitchai and Ward, 1991). However, the acidophilic microalga G. sulphuraria could be efficiently cultured with ammonium at the expected lower pH because of its natural capacity to grow under these conditions (Schmidt et al., 2005). A practical approach to solve the pH problem of avoiding the adverse effects of ammonium consumption of microalgal cultures is the use of a pH-auxostat feed-batch system (Ganuza et al., 2008). In conclusion, regardless of the negative effects on microalgal growth in ammonium-supplemented media, it is still the preferred nitrogen source if the environmental parameters for proper development of the culture are controlled.
2.3.2.
Assimilation of nitrate and nitrite
Nitrate is a major source of nitrogen with a strong impact on metabolism and growth of plants in general. To assimilate nitrate, plant cells transport it across the membrane and then reduce it to ammonia, in the process, consuming large amounts of energy, carbon, and protons (Crawford et al., 2000; Fig. 1). Contrary to the drop in pH observed with ammonium, nitrate consumption causes an increase in pH. Studies of higher plants and microalgae suggest that only two enzymes, nitrate reductase (NR; EC 1.6.6.1-3) and nitrite reductase (NiR; EC 1.7.7.1), work sequentially to catalyze nitrate to ammonium (Kaplan et al., 1986; Fernandez and Galvan, 2007). Assimilatory NR catalyzes the reduction of nitrate to nitrite, using reduced pyridine nucleotides as physiological electron donors (Gewitz et al., 1981; Nakamura and Ikawa, 1993). NiR catalyzes the resulting nitrite; reduction from nitrite to ammonium uses ferredoxin as the electron donor in a reaction that involves the transfer of six electrons (LopezRuiz et al., 1991; Fig. 1). NiR is a chloroplastic enzyme, while NR is located specifically in the cytoplasm and in pyrenoids of green algae (Fernandez and Galvan, 2007; Inokuchi et al., 2002). Environmental variables affect nitrate assimilation. Darkness may have a negative effect on assimilating nitrates. Most algae assimilate nitrate more rapidly in the light than in the dark. A direct photochemical reduction of nitrate and nitrite has been observed in chloroplasts. Light reduces cofactors such as flavoproteins, ferredoxins, and pyridine nucleotides, which then become used as electron donors for nitrate and nitrite reduction (Morris, 1974). Heavy metals affect nitrate assimilation in C. reinhardtii. Consumption of nitrate was not inhibited by metal concentrations below 100 mM. However, concentrations exceeding 150 mM of Cd2þ, Cu2þ, or Zn2þ induced inhibition of 75%, whereas Fe2þ or Co2þ did not significantly affect uptake of nitrate. Among the enzymes of nitrogen assimilatory pathways, exposure of cells for two days to 100 mM Cd2þ did not affect ferredoxin-nitrite reductase (EC 1.7.7.1), ferredoxin-glutamate synthase (EC 1.4.7.1), or NADHglutamate synthase (EC 1.4.1.14) activities, but inhibition of glutamine synthetase activity (EC 6.3.1.2) of 45% occurred (Devriese et al., 2001). In most microalgae, nitrate reductase is fully expressed in cells growing where the sole nitrogen source is nitrate and it is repressed in cells growing in media containing excess ammonium or a mixture of ammonium and nitrate (Gewitz
26
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
et al., 1981; Di Martino Rigano et al., 1982; Sherman and Funkhouser, 1989; Cannons and Pendleton, 1994). This further explains the preference of microalgae species like Chlorella for ammonium and supports the theory that environmental factors must be controlled for proper use of ammonium by microalgae, as explained above.
2.3.3.
Assimilation of urea and organic nitrogen
Consumption of organic nitrogen by microalgae occurs under autotrophic and heterotrophic conditions. All of the organic nitrogen substrates capable of supporting growth under light conditions are also able to do so in the dark. Growth yields with organic nitrogen compounds were generally comparable to those obtained with nitrate or ammonia, although the growth rates varied greatly, depending on the organic nitrogen source, the carbon source, and the strain (Neilson and Larsson, 1980). Growth under heterotrophic conditions with glucose and acetate has been conducted in three microalgae. Selenastrum capricornutum was grown on glucose and urea, glycine, alanine, arginine, asparagines, and glutamine as the organic nitrogen substrate. Chlorella sp. used urea, glycine, glutamate, glutamine, asparagine, ornithine, arginine, and putrescine. E. gracilis grew on acetate, used glycine, alanine, and glutamine. Urea and glutamine are the most widespread organic nitrogen sources that support growth in algae (Morris, 1974; Neilson and Larsson, 1980). Some Chlorella spp. can also use urea as the sole source of nitrogen. It is usually hydrolyzed into ammonia and bicarbonate before its nitrogen is incorporated into cells. In microalgae, two enzymes can metabolize urea, urease (EC: 3.5.1.5) and urea amidolyase (also called urea carboxylase, UALse, EC: 6.3.4.6), but most Chlorella spp. apparently lack urease (Kaplan et al., 1986; Oh-Hama and Miyachi, 1992) and metabolize urea by UALse. The catabolic pathway by UALse is followed by allophanate lyase (EC: 3.5.1.54) that catalyzes hydrolysis of allophanate, resulting in hydrolysis of urea to ammonia and bicarbonate (Morris, 1974). Regardless of the activity, using urea is far less important in the growth cycle of C. vulgaris than ammonium and nitrate. From the data we have so far on nitrogen metabolism under heterotrophic conditions, it is clear that the order of using a nitrogen source by most microalgal species is, in declining order: ammonium > nitrate > nitrite > urea, where special care with the concentration of ammonium is a major consideration.
3. Metabolic products and processes using heterotrophic culture of microalgae The main driving force to grow microalgae commercially is harvesting metabolic products, feed for marine and terrestrial organisms, food supplements for humans, or to use the microalgae for environmental processes, such as wastewater treatment, fertilization of soils, biofuels, and phytoremediation of toxic wastes. The main attractiveness of heterotrophic cultivation is that it is potentially substantially cheaper. Many initiatives have been investigated to produce future products from microalgae, mostly at an experimental stage (Table 2). Ecological uses of microalgae are not discussed in this review.
3.1.
Lipids
Several species of microalgae can be induced to overproduce specific fatty acids through relative simple manipulations of the physical and chemical properties of their culture medium. By manipulating fatty acid content, microalgae represent a significant source of unusual and valuable lipids and fatty acids for numerous industrial applications (Behrens and Kyle, 1996). Accumulation of lipids in the microalgae cells, as well as for other oleaginous microorganisms (high oil producers), depends on diverse factors. These include growth temperature, pH, nutritional imbalances of carbon, nitrogen, phosphorous, and silicate, the growth regime (autotrophic, mixotrophic, or heterotrophic), the age of the culture, and the specific microalgal strain (Ratledge and Wynn, 2002; Wen and Chen, 2003; Chisti, 2007). For example, the lipid content in heterotrophically grown cells of C. protothecoides is as high as 55%, a quantity that is up to four times greater that autotrophically grown cultures under otherwise similar conditions (Xu et al., 2006). In general, accumulation of lipids in yeast and filamentous fungi is associated with exhausting a key nutrient for the microorganisms, usually nitrogen. After the nutrient becomes limited or exhausted, carbon uptake continues and is accumulated as lipids. It might be the same for microalgae (Zhekisheva et al., 2002; Merzlyak et al., 2007). Several proposals to explain the mechanism of accumulating lipids were suggested. In the marine Cryptheconidium conhii and freshwater C. sorokiniana accumulation of lipids may not be dependent on nitrogen exhaustion but on an excess of carbon in the culture media. Hence, in autotrophic or heterotrophic cultures, accumulation could be attributed to consumption of sugars at a rate higher than the rate of cell generation, which would promote conversion of excess sugar into lipids (Chen and Johns, 1991; Ratledge and Wynn, 2002; de Swaaf et al., 2003). This process is often accomplished in two steps: exponential cell division leading to decreased growth from limits of nutrients, thereby leading to accumulation of lipids (Leman, 1997). It might not be only related to higher lipid-synthesizing enzymes under nitrogen starvation, but to the cessation of other enzymes associated with cell growth and proliferation and operation of enzymes specifically related to accumulation of lipids (Ratledge and Wynn, 2002; Ganuza et al., 2008). Another proposed mechanism for accumulating lipids under heterotrophic conditions used E. gracilis as a model. Under nitrogen starvation accumulation of lipids is attributed to mobilization of lipids from chloroplast membranes as chloroplastic nitrogen is relocated by 1,5-biphosphate carboxylase/oxygensae (E.C. 4.1.1.39, Rubisco) (Garcia-Ferris et al., 1996). This proposal is supported by the fact that development of chloroplasts is dependent on nitrogen. Chloroplast breakdown for the internal supply of nitrogen for the cell under nutrient reduction under dark conditions leads to cell survival and growth in the face of prolonged nutrient shortage if an external carbon source is not supplied. But limited nitrogen is not always linked to lipid accumulation. Under nitrogen starvation, the diatoms Achnanthes brevipes and Tetraselmis spp. accumulate carbohydrates (Guerrini et al., 2000; Gladue and Maxey, 1994). This mechanism supports protein synthesis until the nitrogen supply in the medium is restored (Guerrini et al., 2000; Granum et al., 2002;
Table 2 e Potential metabolic products obtainable by heterotrophic cultivation of microalgae. Product
Microalgae species
Significant technical details
Representative references sample
Chlorella vulgaris, C. saccharophila, C. protothecoides, C. sorokiniana, C. pyrenoidosa, Cryptheconidium conhii, Cylindrotheca fusiformis, Euglena gracilis, Navicula incerta, Nitzschia alba, N. laevis, Schizochytrium sp., Skeletonema costatum, Tetraselmis suecica
Up to 4 times higher quantity than under autotrophy; Accumulation probably by similar mechanisms as in autotrophy; Associated with exhausting of a key nutrient for the microalgae, usually nitrogen or silicate (in diatoms); Sugars play a determinant role on the type of lipids accumulated into the cells
Day et al. (1991); Chen and Johns (1991); Tan and Johns (1991, 1996); Gladue and Maxey (1994); Garcia-Ferris et al. (1996); Jiang et al. (1999); Wen and Chen (2000); Ratledge and Wynn (2002); de Swaaf et al. (2003); Wilhelm et al. (2006); Xu et al. (2006); Ganuza et al. (2008)
Polyunsaturated fatty acids
Cryptheconidium conhii, Nitzschia laevis, N. alba, Pavlova lutheri, Schizochytrium limacinum, Tetraselmis suecica
Production high in diatoms; Production control by lowering temperature
Day et al. (1991); Gladue and Maxey (1994); Tatsuzawa and Takizawa (1995); Wen and Chen (2000, 2001a,b); Zhu et al. (2007)
Biodiesel
Chlorella protothecoides
Pigments e phycocianin
Galdieria sulphuraria, Spirulina platensis
Very limiting published data; Comparable to oil-based diesel; Auxiliary pigment to chlorophyll, improve the use of light energy. Can be produced in carbon-limited but nitrogen-sufficient heterotrophic cultures
Wen and Chen (2003); Xu et al. (2006); Chisti (2007, 2008); Xiong et al. (2008) Schmidt et al. (2005); Sloth et al. (2006)
Carotenoids e Xanthophylls
Chlorella pyrenoidosa, Chlorella protothecoides, Chlorella zofingiensis, Haematococcus pluvialis, Dunaliella sp
Pigments that protect chlorophyll against photo damage. Lutein can be produced heterotrophically, with glucose as a C source, and urea as N source. Astaxanthin heterotrophic production is associated to nitrogen starvation at very high C/N ratios
Theriault (1965); Tripathi et al. (1999); Ip and Chen (2005a,b); Wang and Peng (2008); Shi et al. (1997, 1999, 2000)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
Lipids in general
27
28
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
Wilhelm et al., 2006). Accumulation of lipids in diatoms is related to depletion of silicates because of their dependence on silica for growth (Roessler, 1988; Wen and Chen, 2000, 2003; Wilhelm et al., 2006). In any of the above cases, the energy storage molecules, lipids, or carbohydrates, are accumulated. After nitrogen starvation, microalgae, such as C. pyrenoidosa, C. sorokiniana, Nitzschia alba, Skeletonema costatum, C. conhii, accumulate large amounts of lipids, and diatoms respond to depleted silicates by accumulating lipids. In general, this behavior is most probably a survival response until restoration of less nutritionally stressing conditions. As an example in open environments, accumulation of lipids is favored when light is the source of energy. It has been demonstrated that storage products are depleted for energy supply according to their energy content, from lipids to carbohydrates to proteins (Wilhelm et al., 2006). In conclusion, despite the different mechanisms proposed for energy storage compounds, depletion of nitrogen or silicate favors lipid accumulation. Thus, the C:(N or Si) ratio becomes a determining factor in accumulation of lipids and lipid profiles. After nitrogen exhaustion, the remaining sugars play a determining role on the type of lipids accumulated into the cells. Saturation of the fatty acids is directly dependent on the amount of excess sugar and on the autotrophic or heterotrophic conditions (Tan and Johns, 1991; Wood et al., 1999; Wen and Chen, 2000). As the concentration of sugar increases, the fatty acid becomes more saturated (Wood et al., 1999). For example, C. saccharophila, C. vulgaris, N. laevis, Cylindrotheca fusiformis, Navicula incerta, and Tetraselmis suecica accumulate more lipids under heterotrophic than under autotrophic conditions, mainly in the form of triglycerides that provide more energy from oxidation than polyunsaturated fatty acids and therefore, provide superior energy storage (Day et al., 1991; Tan and Johns, 1991, 1996; Gladue and Maxey, 1994). Conversely, autotrophic cultures form more highly unsaturated fatty acids (polar lipids) (Tan and Johns, 1991, 1996; Wen and Chen, 2000) (Fig. 1). In cultures of N. laevis, a cause for variations in accumulating lipids is the source of nitrogen, with ammonia slightly favoring saturated and monounsaturated fatty acids (C14:0, C16:0, C16:1) and nitrate and urea promoting polyunsaturated fatty acids (C20:4 and C20:5) (Wen and Chen, 2001b). Despite varied lipid profiles of specific strains, microalgae mainly accumulate the following fatty acids: (C): 14:0, 14:1, 16:0, 16:1, 18:0, 18:1, 18:2, 18:3, 20:4, 20:5, 22:5, 22:6 (Vazhappilly and Chen, 1998; de-Bashan et al., 2002); therefore, the microalgae are industrially important for medicines and nutritional supplements for humans and animals (polyunsaturated fatty acids), pigments, and lately, biofuels, mainly biodiesel (Wen and Chen, 2003; Kulkarni and Dalai, 2006; Chisti, 2007; Del Campo et al., 2007).
3.2.
Polyunsaturated fatty acids
Long-chain polyunsaturated fatty acids (eicosapentaenoic acid, EPA, u-3, C20:5 and docosahexaenoic acid, DHA, u-3, C22:6) are two important fatty acids in early and old age metabolism in humans. They have been used in prevention and treatment of human diseases such as heart and inflammatory diseases and as nutritional supplements in humans and marine organisms in
aquaculture. Because the common source for EPA and DHA, fish oil, fails to meet the increasing demand for purified EPA and DHA, alternative sources such as microalgae that contain large quantities of high-quality EPA and DHA are considered a potential source of these economically-important fatty acids, especially under heterotrophic conditions that reduced the costs of production (Barclay et al., 1994; Vazhappilly and Chen, 1998; Apt and Behrens, 1999; Wen and Chen, 2003; Sijtsma and de Swaaf, 2004; Sijtsma et al., 2005; Chi et al., 2007). Under autotrophic conditions long-chain fatty acids are assembled from a successive coupling of carbonecarbon bounds from acetate and malonyl-ACP (acyl-carrier protein), beginning with acetyl-CoA as the initial substrate and ending with acyl-ACP. Acetyl-CoA is produced from pyruvate generated during glycolysis or from free acetate taken into plastids, probably activated by acetyl-CoA synthetase in the stroma. For a typical C18 fatty acid, 16 molecules of NAD(P)H are required. In the dark, the pentose phosphate pathway is the producer of the reduced NADPH (Somerville et al., 2000). As mentioned earlier, under heterotrophic conditions, the production of saturated fatty acids is favored, while highly polyunsaturated fatty acids (C16:3 and C18:3) content are mainly produced under autotrophic conditions. However, the production of polyunsaturated fatty acids, EPA and DHA, is higher in dark cultures of the diatoms Tetraselmis spp., N. laevis, and N. alba (Day et al., 1991; Gladue and Maxey, 1994; Wen and Chen, 2000, 2003; Chen et al., 2007). It was further shown that microalgae-based heterotrophic production systems can exhibit u-3 fatty acid productivities that are two to three orders of magnitude greater than those of outdoor autotrophic pond systems. Additionally, long-chain u-3 fatty acid productivities reported for the microalgae fermentation systems are one to two orders of magnitude greater than productivities reported for fungal or bacterial systems (Barclay et al., 1994). The nitrogen source affects production of EPA by the diatom N. laevis in heterotrophic cultures where nitrate and urea are preferred N sources for cell growth and EPA content. Tryptone and yeast extract were found to enhance EPA production (Wen and Chen, 2001a). Temperature also influences the fatty acids profile (Fig. 1). When temperature is below the optimal growth temperature for the microalgae, more unsaturated fatty acids are metabolized, and the reverse effect occurs at higher temperatures. Reducing temperature by 10e15 C leads to a decrease in membrane fluidness. To compensate for decreasing fluidness, over-expression of the genes for desaturases (acyl-CoA desaturases, acyl-ACP desaturases, and acyl-lipid desaturases) promote desaturation of the membrane lipids. However, no change in total fatty acid production occurs (Tatsuzawa and Takizawa, 1995; Quoc and Dubacq, 1997; Sakamoto and Bryant, 1997; Zhu et al., 2007). It was suggested that the lipid 1-oleoyl-2-palmitoylmonogalactosyl-sn-glycerol is involved in the regulation of membrane fluidity during temperature acclimation of the cyanobacteria Anabaena variabilis. This compound increases with increasing temperature and decreases with declining temperature (Sato and Murata, 1986). Another plausible explanation for instaurations is the positive effect of low temperature on increasing the molecular oxygen level in cells and promoting the activity of desaturases and elongases for fatty acids biosynthesis (Richmond, 1986; Wen and Chen, 2003; Chen and Chen, 2006).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
Use of acetic acid as a carbon source for heterotrophic production of DHA in batch-fed cultures of high cell density of C. cohnii resulted in much higher lipid and DHA contents than in cultivation with glucose (Jiang and Chen, 2000; de Swaaf et al., 2003). This difference may be related to the biochemistry and subcellular location of acetyl-CoA metabolism. It is likely that, similar to yeasts, the mitochondrial pyruvateedehydrogenase complex is the main source of acetyl-CoA during growth with glucose. The fatty acid synthetase complex of C. cohnii was shown to be cytosolic (Sonnenborn and Kunau, 1982), which suggests that, similar to yeasts (Ratledge and Evans, 1989), lipid synthesis in this microalga occurs in the cytosol. This implies that, during growth with glucose, export of acetyl-CoA from the mitochondrial matrix to the cytosol is required to make it available for lipid synthesis. In contrast, acetate can be directly activated to acetyl-CoA by the action of acetyl-CoA synthetase (de Swaaf et al., 2003). In summary, studies show that heterotrophic production of EPA and DHA is feasible at larger production scales than in autotrophic regimes, but this can be accomplished with a few microalgae.
3.3.
Biodiesel
Biofuels from microalgae is an attractive option for microalgae biotechnology. Compared to all other applications, it is one of the most attractive, given the high prices of crude oil. Biodiesel is a suitable substitute for petroleum-based diesel fuel because of its multiple advantages for machines and the environment. Currently, biodiesel production depends mainly on vegetable oils, such as canola, soybean, sunflower, and palm containing long-chain fatty acids (LCFA) and to a small extent on animal fat and oil recycling. This implies seasonal availability and large expansion of farmland at the expense of food crops. This is a major limitation and sufficient reason to search for other sources of LCFA. The current type of production of biodiesel is not sustainable because of the inherent conflict with food supply and threat to food security. Biodiesel from microalgae is an attractive, feasible alternative mainly because some microalgae species can significantly increase production of lipids and cultivation of microalgae is now possible through cheaper heterotrophic cultivation. Strains can be genetically engineered to produce the desired fatty acids without negative effects on the environments. Microalgae, potentially in the longer term, offer the greatest opportunities compared to oilseed crops. Productivity of many microalgae exceeds the best producing oil crops where oil content of many microalgae strains under heterotrophic conditions is usually 80% of its dry weight and, their production and processing into biofuels, is economically effective, uses currently available technology, and is environmentally sustainable because their production is not seasonal and the product can be harvested daily. Current mass production of microalgae requires significantly less land area than crop-based biofuels and releases fewer pollutants to the environment. Because biofuels from microalgae was recently reviewed so extensively from every angle (Chisti, 2007, 2008; Li et al., 2008a,b; Sharif Hossain et al., 2008; Song et al., 2008; Khan et al., 2009; Huang et al., 2010; Mata et al., 2010; Sivakumar et al., 2010), this review will briefly present
29
only a few examples, mostly from China with heterotrophic cultivation. While heterotrophic microalgae cultivation represents a good source of LCFA (Wen and Chen, 2003); so far, it is a less popular avenue for biodiesel production from microalgae. C. protothecoides is a suitable microalga for biodiesel production, heterotrophically using organic carbon sources. This species was able to produce quantities of lipids reaching w50% of its dry weight. Enzymatic transesterification (converting lipids to biodiesel) was catalyzed by lipase, and the conversion rate reached close to 100% in several trials. The biodiesel was comparable to oil-based diesel and complies with the US Standard for Biodiesel (Li et al., 2007; Miao and Wu, 2006; Xu et al., 2006; Xiong et al., 2008). Using C. protothecoides biodiesel was produced from hydrolysate of the Jerusalem artichoke tuber under heterotrophic conditions, with significant cost reduction. Accumulated lipid in vivo, with lipid content as high as 44% of dry mass was obtained and converted to biodiesel. Unsaturated fatty acid methyl esters constituted >82% of the total biodiesel content, of which the chief components were cetane acid methyl, linoleic acid methyl, and oleic acid methyl esters (Cheng et al., 2009). One of the potential carbon sources for producing biodiesel heterotrophically is glycerol. Currently, glycerol is an inexpensive and abundant carbon generated as a by-product of biodiesel fuel production. Development of processes to convert this crude glycerol into higher-value products is needed. Given the highly reduced nature of carbon atoms in glycerol, fuel and reduced chemicals can be generated at higher yields than those obtained from common sugars, such as glucose (Yazdani and Gonzalez, 2007; Murarka et al., 2008). For example: Schizochytrium limacinum produced palmitic acid (16:0) as w45e60% of their dry weight when supplied with glucose, fructose, or glycerol (Yokochi et al., 1998; Chi et al., 2007), which could potentially be used for biodiesel production. In summary, production of biodiesel by heterotrophic microalgae is a very new field of research, with little solid information available, apart from commercial promises, to indicate the true commercial potential of this source. Considering metabolism in microalgae, cheap carbon sources yielding promising amounts of long-chain fatty acids make this an attractive venue for future research.
3.4.
Pigments
In addition to the main photosynthesis pigment chlorophyll, microalgae contain auxiliary photosynthetic pigments to improve use of light energy (phycobiliproteins) and protection against solar radiation (carotenoids) (Cohen, 1986; Pulz and Gross, 2004; Del Campo et al., 2007). Naturally, all pigments are produced under autotrophic growth conditions, but surprisingly some are produced, and in large quantities, under heterotrophic dark conditions.
3.4.1.
Carotenoids
Carotenoids from microalgae have been used for commercial purposes. Carotenoids are lipid-soluble pigments composed of isoprene units that are widely distributed in various classes of microalgae. Carotenoids are divided into two groups: those
30
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
containing only hydrocarbons (not oxygenated) and xanthophylls that contain oxygen molecules (Cohen, 1986). The green algae class Chlorophyceae, contain a and b-carotenes and the xanthophylls: lutein, zeaxanthin, violaxanthin, and neoxanthin (Theriault, 1965; Cohen, 1986). For example, autographically grown Dunaliella sp. is the richest source of lutein. It contains up to 14% dry weight and is used as a food supplement. Natural colorants have become increasingly important because of regulations limiting synthetic additives; with microalgae as a source, carotenoids are one of the major fields of exploitation of microalgal biotechnology (Theriault, 1965; Cohen, 1986; Pulz and Gross, 2004; Borowitzka, 2005; Lebeau and Robert, 2006; Del Campo et al., 2007). Among the xanthophylls (zeaxanthin, canthaxantin, and lutein), lutein is considered the principal useful pigment of the group. It has high nutritional value and low toxicity and is used as a pigment for animal tissues (chicken skin and egg yolks coloring), food, cosmetics, and pharmaceutical products, such as an effective agent for prevention and treatment of a variety of degenerative diseases (Shi et al., 1997; Pulz and Gross, 2004). Lutein is an intracellular product of Chlorella. This genus is used for production of lutein, mainly C. protothecoides and, to a lesser extent, C. pyrenoidosa and C. vulgaris. Photoautotrophic systems produce low biomass; hence, heterotrophic cultivation represents an alternative. Increasing glucose concentration increases lutein production, but urea is currently the optimal source of nitrogen (Theriault, 1965; Shi et al., 1997, 1999, 2000). Astaxanthin is a red ketocarotenoid colorant used in the cosmetic, therapeutic, and food industries. In aquaculture, it has been used to increase growth and survival of aquatic animals and as a colorant of tissues (farmed salmon, shrimp, lobster, trout, and fish eggs) to provide a pinkish-red color to the tissue that is appealing to consumers. Its strong antioxidant character makes it a nutraceutical product (food or nutritional supplement that may improve health) and may prevent some cancers (Hagen et al., 2001; Ma and Chen, 2001; Pulz and Gross, 2004; Del Rio et al., 2005; Ip and Chen, 2005a; Wang and Peng, 2008). Astaxanthin production by microalgae increases under stress conditions and is present in the esterified form and stored in lipid bodies outside the chloroplast, which enables green algae to accumulate a considerable amount (Wang and Peng, 2008). Haematococcus pluvialis is the main producer of astaxanthin under autotrophic conditions but C. zofingiensis is superior in yield when heterotrophically cultivated with glucose (Ip and Chen, 2005a). Biosynthesis of astaxanthin in C. zofingiensis starts in the early exponential phase and is a growth-associated metabolite (product that is produced only during active growth); therefore, it depends on assimilation of the carbon source. Its production was associated, similar to lipid accumulation, to nitrogen starvation at very high C/N ratios (Ip and Chen, 2005a; Wang and Peng, 2008) and mainly to the presence of oxidative stress that is essential for promoting the formation of several secondary carotenoids, including astaxanthin. Such oxidative treatments employ the hydroxyl radical (OH supplied by H2O2) (Ip and Chen, 2005b). Addition of reactive nitrogen species, such as peroxynitrite and nitryl chloride induced similar effects (Ip and Chen, 2005c). The blue photosynthetic pigment, phycocyanin, is found in a few cyanobacteria and microalgae; its main source is the
cyanobacterium S. platensis. It is used as a fluorescent marker in diagnostic histochemistry and as dye in food and cosmetics. The red microalgae G. sulphuraria can produce phycocyanin in carbon-limited but nitrogen-sufficient heterotrophic cultures; the content increases in the stationary phase. Although production of phycocyanin in this microalga is lower than in S. platensis, its ability to grow heterotrophically makes it a potential supplier of this pigment (Sloth et al., 2006). Another study found that this microalga produced more phycocyanin in heterotrophic batch-fed cultures of G. sulphuraria than is commonly attained in outdoor, sunlight-dependent cultures of S. platensis (Schmidt et al., 2005). If the heterotrophic process is scaled up, the reduction of cost using G. sulphuraria would be significant. In summary, although pigments are traditionally thought to be the outcome of metabolisms associated with exposure to light, the capacity of some microalgae to produce some of them in the dark under specific growth conditions opens a line of research that is barely explored.
3.5.
Wastewater treatment
As mentioned earlier, tertiary wastewater treatment by microalgae is an old idea that so far has very limited application. This is directly related to the costs involved in treating very large volumes of wastewater in a timely manner under autotrophic conditions (de-Bashan and Bashan, 2010). Heterotrophic wastewater treatment is a novel idea that so far has been studied at the laboratory scale. The most efficient carbon source for using C. vulgaris to treat wastewater heterotrophically was calcium acetate (Perez-Garcia et al., in press). Subjecting the autotrophic, immobilized microalgae-bacteria system for wastewater treatment (de-Bashan et al., 2002, 2004; Hernandez et al., 2006) to heterotrophic conditions revealed even higher potential of the system to eliminate nutrients (Perez-Garcia et al., 2010). The new data cannot provide a solid prediction about the practical potential of this approach.
4.
Concluding remarks and future prospects
Cultivation of microalgae that are primarily photosynthetic under heterotrophic dark conditions for production of economically useful metabolites or technological processes is a tempting option, given significant reductions in complexity of cultivation and costs. Because heterotrophic growth consumes simple, cheap, and available carbon sources (glucose, acetate, glycerol) that are commonly used by fermentation industries for other aims, it is predicted that adoption of this approach is an easy, uncomplicated task. Fortunately, some of the most common and best-studied microalgae, such as Chlorella, are also heterotrophs. This information can jump start research in heterotrophy, which is probably quite prevalent among microalgae (Tuchman, 1996; Hellebust and Lewin, 1977). Furthermore, with current developments in genomics, bioinformatics analyses, and genetic and metabolic engineering, new approaches in microalgae biotechnology, including heterotrophy, have emerged (Hong and Lee, 2007; Boyle and Morgan, 2009).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
As a result of genetic engineering, some obligate photoautotrophs were transformed to heterotrophy through the introduction of sugar transporters. Volvox carteri was one of the first green algae to be transformed with the hexose/Hþ symporter gene derived from Chlorella sp. (Hallmann and Sumper, 1996). Similar trophic conversions have also been carried out in C. reinhardtii (Doebbe et al., 2007) and the diatom P. tricornutum (Zaslavskaia et al., 2001). These examples of a simple genetic transformation of single gene of a sugar transporter in the outer membrane of microalgae show the feasibility to convert microalgae from a photoautotrophic into a heterotrophic organism when sugar is present in the absence of light. Such genetic engineering is probably acceptable by society because microalgae cultivation can be independently managed without risk of environmental contamination; thus, these mutants can be employed in metabolic production of products, such as hydrogen by C. reinhardtii (Doebbe et al., 2007). Adoption of heterotrophy for large-scale industrial processes, such as wastewater treatment and biofuels production, is somewhat more problematic and lies in the more distant future. Because microalgae cultivation alone cannot sustain biofuel production with the current cultivation technologies, perhaps combining two processes, wastewater treatment followed by biofuel production from the residual mass will yield a product that would make development of these technologies economically acceptable. Heterotrophic cultivation of microalgae is a niche of microalgae research field. Yet, the potential of expansion because of the advantages it offers are limitless. Only time will prove if this strategic approach will catch up with the industry.
Acknowledgments This review was mainly supported by Secretaria de Medio Ambiente y Recursos Naturales (SEMARNAT contract #23510) and Consejo Nacional de Ciencia y Tecnologı´a (CONACYT contract 23917) of Mexico and its writing was supported by The Bashan Foundation, USA. FMEE was supported by a postdoctoral fellowship and OPG by a graduate fellowship from CONACYT.
Appendix. Supplementary material Supplementary data related to this article can be found online at doi:10.1016/j.watres.2010.08.037.
references
Abeliovich, A., Weisman, D., 1978. Role of heterotrophic nutrition in growth of the alga Scenedesmus obliquus in high-rate oxidation ponds. Appl. Environ. Microbiol. 35, 32e37. Ahmad, I., Hellebust, J.A., 1990. Regulation of chloroplast development by nitrogen source and growth conditions in a Chlorella protothecoides strain. Plant Physiol. 94, 944e949. Allen, A.E., Ward, B.B., Song, B., 2005. Characterization of diatom (Bacillariophyceae) nitrate reductase genes and their detection in marine phytoplankton communities. J. Phycol. 41, 95e104.
31
Apt, K.E., Behrens, P.W., 1999. Commercial developments in microalgal biotechnology. J. Phycol 35, 215e226. Armbrust, E.V., et al., 2004. The genome of the diatom Thalassiosira pseudonana: ecology, evolution, and metabolism. Science 306, 79e86. Aubert, S., Gout, E., Bligny, R., Douce, R., 1994. Multiple effects of glycerol on plant cell metabolism: phosphorus-31 nuclear magnetic resonance studies. J. Biol. Chem. 269, 21420e21427. Barbosa, M.J.G.V., 2003. Microalgal Photobioreactors: Scale-up and Optimization. Doctoral Thesis, Wageningen University, The Netherlands. Barclay, W.R., Meager, K.M., Abril, J.R., 1994. Heterotrophic production of long chain omega-3 fatty acids utilizing algae and algae-like microorganisms. J. Appl. Phycol. 6, 123e129. Becker, E.W., 1994. Microalgae: Biotechnology and Microbiology. Cambridge University, Cambridge, 293 pp. Becker, H.M., Hirnet, D., Fecher-Trost, C., Sultemeyer, D., Deitmer, J.W., 2005. Transport activity of MCT1 expressed in Xenopus oocytes is increased by interaction with carbonic anhydrase. J. Biol. Chem. 280, 39882e39889. Behrens, P.W., 2005. Photobioreactor and fermentors: the light and the dark sides of the growing algae. In: Andersen, R.A. (Ed.), Algal Culturing Techniques. Elsevier Academic Press, New York, USA, pp. 189e204. Behrens, P.W., Kyle, D.J., 1996. Microalgae as a source of fatty acids. J. Food Lipids 3, 259e272. Bollman, R.C., Robinson, G.G.C., 1977. The kinetics of organic acid uptake by three Chlorophyta in axenic culture. J. Phycol 13, 1e5. Borowitzka, M.A., 2005. Carotenoid production using microorganisms. In: Cohen, Z., Ratledge, C. (Eds.), Single Cell Oils. AOCS Press, Champaign, IL, USA, pp. 124e137. Borowitzka, L.J., Borowitzka, M.A., 1989. Industrial production: methods and economics. In: Cresswell, R.C., Rees, T.A.V., Shah, H. (Eds.), Algal and Cyanobacterial Biotechnology. Longman Scientific & Technical, New York, USA, pp. 294e316. Boyle, N.R., Morgan, J.A., 2009. Flux balance analysis of primary metabolism in Chlamydomonas reinhardtii. BMC Syst. Biol. 3, 4. Brennan, L., Owende, P., 2010. Biofuels from microalgae e a review of technologies for production, processing, and extractions of biofuels and co-products. Renew. Sust. Energ. Rev. 14, 557e577. Cannons, A., Pendleton, L.C., 1994. Possible role for mRNA stability in the ammonium-controlled regulation of nitrate reductase expression. Biochem. J. 297, 561e565. Can˜izares, R.O., Rivas, L., Montes, C., Dominguez, A.R., Travieso, L., Benitez, F., 1994. Aerated swine wastewater treatment with k-carrageenan-immobilized Spirulina maxima. Bioresour. Technol. 47, 89e91. Carvalho, A.P., Meireles, L.A., Malcata, F.X., 2006. Microalgal reactors: a review of enclosed system designs and performances. Biotechnol. Prog. 22, 1490e1506. Caspari, T., Will, A., Opekarova, M., Sauer, N., Tanner, W., 1994. Hexose/Hþ symporters in lower and higher plants. J. Exp. Biol. 196, 483e491. Cero´n Garcı´a, M.C., Ferna´ndez Sevilla, J.M., Acie´n Ferna´ndez, F.G., Molina Grima, E., Garcı´a Camacho, F., 2000. Mixotrophic growth of Phaeodactylum tricornutum on glycerol: growth rate and fatty acid profile. J. Appl. Phycol. 12, 239e248. Chen, F., 1996. High cell density culture of microalgae in heterotrophic growth. Trends Biotechnol. 14, 412e426. Chen, F., Johns, M.R., 1991. Effect of C/N ratio and aeration on the fatty acid composition of heterotrophic Chlorella sorokiniana. J. Appl. Phycol. 3, 203e209. Chen, F., Johns, M.R., 1995. A strategy for high cell density culture of heterotrophic microalgae with inhibitory substrates. J. Appl. Phycol. 7, 43e46.
32
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
Chen, F., Johns, M.R., 1996a. Heterotrophic growth of Chlamydomonas reinhardtii on acetate in chemostat culture. Process Biochem. 31, 601e604. Chen, F., Johns, M.R., 1996b. Relationship between substrate inhibition and maintenance energy of Chlamydomonas reinhardtii in heterotrophic culture. J. Appl. Phycol. 8, 15e19. Chen, G.-Q., Chen, F., 2006. Growing phototrophic cells without light. Biotechnol. Lett. 28, 607e616. Chen, G.-Q., Jiang, Y., Chen, F., 2007. Fatty acid and lipid class composition of the eicosapentaenoic acid-producing microalga, Nitzschia laevis. Food Chem. 104, 1580e1585. Cheng, Y., Zhou, W., Gao, C., Lan, K., Gao, Y., Wu, Q., 2009. Biodiesel production from Jerusalem artichoke (Helianthus tuberosus L.) tuber by heterotrophic microalgae Chlorella protothecoides. J. Chem. Technol. Biotechnol. 84, 777e781. Chi, Z., Pyle, D., Wen, Z., Frear, C., Chen, S., 2007. A laboratory study of producing docosahexaenoic acid from biodieselwaste glycerol by microalgal fermentation. Process Biochem. 42, 1537e1545. Chisti, Y., 2007. Biodiesel from microalgae. Biotechnol. Adv. 25, 294e306. Chisti, Y., 2008. Biodiesel from microalgae beats ethanol. Trends Biotechnol. 26, 126e131. Cho, B.-H., Sauer, N., Komor, E., Tanner, W., 1981. Glucose induces two amino acid transport systems in Chlorella. PNAS 78, 3591e3594. Cohen, Z., 1986. Products from microalgae. In: Richmond, A. (Ed.), Handbook for Microalgal Mass Culture. CRC Press, Boca Raton, FL., USA, pp. 421e454. Combres, C., Laliberte´, G., Sevrin Reyssac, J., de la Nou¨e, J., 1994. Effect of acetate on growth and ammonium uptake in the microalga Scenedesmus obliquus. Physiol. Plantarum 91, 729e734. Cook, J.R., 1967. Photoassimilation of acetate by an obligately phototrophic strain Euglena gracilis. J. Protozool. 14, 382e384. Cook, J.R., 1968. The cultivation and growth of Euglena. In: Buetow, D.E. (Ed.), The Biology of Euglena, vol. 1. Academic Press, New York, USA, pp. 243e314. Cooney, C.L., 1983. Bioreactors: design and operation. Science 219, 728e733. Coruzzi, G.M., 2003. Primary N-assimilation into Amino Acids in Arabidopsis. The Arabidopsis Book. American Society of Plant Biologist. http://www.aspb.org/publications/arabidopsis/. Crawford, N.M., Kahn, M.L., Leustek, T., Long, S.R., 2000. Nitrogen and sulfur. In: Buchanan, B., Gruissem, W., Jones, R. (Eds.), Biochemistry and Molecular Biology of Plants. Am. Soc. Plant Physiol., Rockville, MD, USA, pp. 786e849. Christie, J.M., Reymond, P., Powell, G.K., Bernasconi, P., Raibekas, A.A., Liscum, E., Briggs, W.R., 1998. Arabidopsis NPH1: a flavoprotein with the properties of a photoreceptor for phototropism. Science 282, 1698e1701. Day, J.G., Edwards, A.P., Rodgers, G.A., 1991. Development of an industrial-scale process for the heterotrophic production of a micro-algal mollusc feed. Bioresour. Technol. 38, 245e249. de-Bashan, L.E., Bashan, Y., 2010. Immobilized microalgae for removing pollutants: review of practical aspects. Bioresour. Technol. 101, 1611e1627. de-Bashan, L.E., Bashan, Y., Moreno, M., Lebsky, V.K., Bustillos, J.J., 2002. Increased pigment and lipid content, lipid variety, and cell and population size of the microalgae Chlorella spp. when co-immobilized in alginate beads with the microalgae-growthpromoting bacterium Azospirillum brasilense. Can. J. Microbiol. 48, 514e521. de-Bashan, L.E., Hernandez, J.-P., Morey, T., Bashan, Y., 2004. Microalgae growth-promoting bacteria as “helpers” for microalgae: a novel approach for removing ammonium and phosphorus from municipal wastewater. Water Res. 38, 466e474.
de-Bashan, L.E., Antoun, H., Bashan, Y., 2005. Cultivation factors and population size control uptake of nitrogen by the microalgae Chlorella vulgaris when interacting with the microalgae growth-promoting bacterium Azospirillum brasilense. FEMS Microbiol. Ecol. 54, 197e203. de-Bashan, L.E., Magallon, P., Antoun, H., Bashan, Y., 2008. Role of glutamate dehydrogenase and glutamine synthetase in Chlorella vulgaris during assimilation of ammonium when jointly immobilized with the microalgae-growth-promoting bacterium Azospirillum brasilense. J. Phycol 44, 1188e1196. de la Nou¨e, J., Proulx, D., 1988. Tertiary treatment of urban wastewaters with chitosan immobilized Phormidium. Appl. Microbiol. Biotechnol. 29, 292e297. de la Nou¨e, J., Laliberte´, G., Proulx, D., 1992. Algae and wastewater. J. Appl. Phycol. 4, 247e254. de Swaaf, M.E., Sijtsma, L., Pronk, J.T., 2003. High-cell-density fedbatch cultivation of the docosahexaenoic acid producing marine alga Crypthecodinium cohnii. Biotechnol. Bioeng. 81, 666e672. Del Campo, J.A., Garcı´a-Gonza´lez, M., Guerrero, M.G., 2007. Outdoor cultivation of microalgae for carotenoid production: current state and perspectives. Appl. Microbiol. Biotechnol. 74, 1163e1174. Del Rio, E., Acien, F.G., Garcia-Malea, M.C., Rivas, J., MolinaGrima, E., Guerrero, M.G., 2005. Efficient one-step production of Astaxanthin by the microalga Haematococcus pluviales in continuous culture. Biotechnol. Bioeng. 91, 808e815. Devriese, M., Tsakaloudi, V., Garbayo, I., Leo´n, R., Vı´lchez, C., Vigara, C., 2001. Effect of heavy metals on nitrate assimilation in the eukaryotic microalga Chlamydomonas reinhardtii. Plant Physiol. Biochem. 39, 443e448. Di Martino Rigano, V., Vona, V., Fuggi, V., Di Martino, C., Rigano, C., 1982. Regulation of nitrate reductase in Chlorella nitrate requirement for the appearance of nitrate reductase activity. Plant Sci. Lett. 28, 265e272. Di Martino Rigano, V., Vona, V., Esposito, S., Carillo, P., Carfagna, S., Rigano, C., 1998. The physiological significance of light and dark NHþ 4 metabolism in Chlorella sorokiniana. Phytochemistry 47, 177e181. Doebbe, A., Rupprecht, J., Beckmann, J., Mussgnug, J.H., Hallman, A., Hankamer, B., Kruse, O., 2007. Functional integration of the HUP1 hexose symporter gene into the genome of C. reinhardtii: impacts on biological H2 production. J. Biotechnol. 131, 27e33. Droop, M.R., 1974. Heterotrophy of carbon. In: Stewart, W.D.P. (Ed.), Algal Physiology and Biochemistry. Blackwell Scientific, Oxford, UK, pp. 530e559. Endo, H., Nakajima, K., Chino, R., Shirota, M., 1974. Growth characteristics and cellular components of Chlorella regularis, heterotrophic fast growing strain. Agric. Biol. Chem. 38, 9e18. Fernandez, E., Galvan, A., 2007. Inorganic nitrogen assimilation in Chlamydomonas. J. Exp. Biol. 58, 2279e2287. Ganuza, E., Anderson, A.J., Ratledge, C., 2008. High-cell-density cultivation of Schizochytrium sp. in an ammonium/pHauxostat fed-batch system. Biotechnol. Lett. 30, 1559e1564. Garcia-Fernandez, J.M., Diez, J., 2004. Adaptive mechanism of nitrogen and carbon assimilatory pathways in the marine cyanobacteria Prochlorococcus. Res. Microbiol. 155, 795e802. Garcia-Ferris, C., de los Rios, A., Ascaso, C., Moreno, J., 1996. Correlated biochemical and ultrastructural changes in nitrogen-starved Euglena gracilis. J. Phycol 32, 953e963. Geider, R.J., Osborne, B.A., 1989. Respiration and microalgal growth: a review of the quantitative relationship between dark respiration and growth. New Phytol. 112, 327e341. Gewitz, H.S., Piefke, J., Vennesland, B., 1981. Purification and characterization of demolybdo nitrate reductase (NADH cytocrome c oxido reductase) of Chlorella vulgaris. J. Biol. Chem. 256, 11527e11531.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
Gladue, R.M., Maxey, J.E., 1994. Microalgal feeds for aquaculture. J. Appl. Phycol. 6, 131e141. Goldman, J.C., 1976. Biomass production in mass cultures of marine phytoplankton at continuous culture. J. Exp. Mar. Biol. Ecol. 23, 31e43. Gonzalez, L.E., Can˜izares, R.O., Baena, S., 1997. Efficiency of ammonia and phosphorus removal from a Colombian agroindustrial wastewater by the microalgae Chlorella vulgaris and Scenedesmus dimorphus. Bioresour. Technol. 60, 259e262. Goulding, K.H., Merrett, M.J., 1967. The photo-assimilation of acetate by Pyrobotrys stellata. J. Gen. Microbiol. 48, 127e136. Granum, E., Kirkvold, S., Myklestad, S.M., 2002. Cellular and extracellular production of carbohydrates and amino acids by the marine diatom Skeletonema costatum: duel variations and effects of N depletion. Mar. Ecol. Prog. Ser. 242, 82e93. Griffiths, D.J., Thresher, C.L., Street, H.E., 1960. The heterotrophic nutrition of Chlorella vulgaris. Ann. Bot. 24, 1e11. Grobbelaar, J.U., 2004. Algal nutrition: mineral nutrition. In: Richmond, A. (Ed.), Handbook of Microalgal Culture. Biotechnology and Applied Phycology. Blackwell Publishing, Oxford, UK, pp. 97e115. Guerrini, F., Cangini, M., Boni, L., Trost, P., Pistocchi, R., 2000. Metabolic responses of the diatom Achnanthes brevipes to nutrient limitation. J. Phycol 36, 882e890. Haass, D., Tanner, W., 1974. Regulation of hexose transport in Chlorella vulgaris. Plant Physiol. 53, 14e20. Hagen, C., Gru¨newald, K., Xyla¨nder, M., Rothe, E., 2001. Effect of cultivation parameters on growth and pigment biosynthesis in flagellated cells of Haematococcus pluvialis. J. Appl. Phycol. 13, 79e87. Hallmann, A., Sumper, M., 1996. The Chlorella hexose/Hþ symporter is a useful selectable marker and biochemical reagent when expressed in Volvox. PNAS 93, 669e673. Harrop, L.C., Kornberg, H.L., 1966. The role of isocitrate lyase in the metabolism of algae. Proc. R. Soc. London 166, 11e29. Harun, R., Singh, M., Forde, G.M., Danquah, M.K., 2010. Bioprocess engineering of microalgae to produce a variety of consumer products. Renew. Sust. Energ. Rev. 14, 1037e1047. Hellebust, J.A., Lewin, J., 1977. Heterotrophic nutrition. In: Werner, D. (Ed.), The Biology of Diatoms. Bot. Monogr., vol. 13. University of California Press, Los Angeles, CA, USA, pp. 169e197. Hernandez, J.-P., de-Bashan, L.E., Bashan, Y., 2006. Starvation enhances phosphorus removal from wastewater by the microalgae Chlorella spp. co-immobilized with Azospirillum brasilense. Enzyme Microb. Technol. 38, 190e198. Hildebrand, M., 2005. Cloning and functional characterization of ammonium transporters from the marine diatom Cylindrotheca fusiformis (Bacillariophyceae). J. Phycol 41, 105e113. Hilgarth, C., Sauer, N., Tanner, W., 1991. Glucose increases the expression of the ATP/ADP translocator and the glyceraldehyde-3-phosphate dehydrogenase genes in Chlorella. J. Biol. Chem. 266, 24044e24047. Hong, S.-J., Lee, C.-G., 2007. Evaluation of central metabolism based on a genomic database of Synechocystis PCC6803. Biotechnol. Bioprocess Eng. 12, 165e173. Howitt, S.M., Udvardi, M.K., 2000. Structure, function and regulation of ammonium transporters in plants. Biochem. Biophys. Acta 1465, 152e170. Huang, G.H., Chen, F., Wei, D., Zhang, X.W., Chen, G., 2010. Biodiesel production by microalgal biotechnology. Appl. Energy 87, 38e46. Huppe, H.C., Turpin, D.H., 1994. Integration of carbon and nitrogen metabolism in plant and algal cells. Ann. Rev. Plant Physiol. Plant Mol. Biol. 45, 577e607. Ingram, L.O., Van Baalen, C., Calder, J.A., 1973. Role of reduced exogenous organic compounds in the physiology of the blue-
33
green bacteria (algae): photoheterotrophic growth of an “autotrophic” blue-green bacterium. J. Bacteriol. 114, 701e705. Inokuchi, R., Kuma, K.-I., Miyata, T., Okada, M., 2002. Nitrogenassimilating enzymes in land plants and algae: phylogenic and physiological perspectives. Physiol. Plantarum 116, 1e11. Ip, P.-F., Chen, F., 2005a. Production of astaxanthin by the green microalga Chlorella zofingiensis in the dark. Process Biochem. 40, 733e738. Ip, P.-F., Chen, F., 2005b. Employment of reactive oxygen species to enhance astaxanthin formation in Chlorella zofingiensis in heterotrophic culture. Process Biochem. 40, 3491e3496. Ip, P.-F., Chen, F., 2005c. Peroxynitrite and nitryl chloride enhance astaxanthin production by the green microalga Chlorella zofingiensis in heterotrophic culture. Process Biochem. 40, 3595e3599. Javanmardian, M., Palsson, B.O., 1991. High-density photoautotrophic algal cultures: design, construction, and operation of a novel photobioreactor system. Biotechnol. Bioeng. 38, 1182e1189. Jiang, Y., Chen, F., 2000. Effects of temperature and temperature shift on docosahexaenoic acid production by the marine microalga Crypthecodinium cohnii. J. Am. Oil. Chem. Soc. 77, 613e617. Jiang, Y., Chen, F., Liang, S.Z., 1999. Production potential of docohexaenoic acid by the heterotrophic marine dinoflagellate Crypthecodinium cohnii. Process Biochem. 34, 633e637. Kamiya, A., 1995. Effects of blue light on the uptake of ammonia and nitrate by a colorless mutant of Chlorella. Plant Cell Physiol. 36, 481e485. Kamiya, A., Kowallik, W., 1987a. Photoinhibition of glucose uptake in Chlorella. Plant Cell Physiol. 28, 611e619. Kamiya, A., Kowallik, W., 1987b. The inhibitory effect of light on proton-coupled hexose uptake in Chlorella. Plant Cell Physiol. 28, 621e625. Kamiya, A., Saitoh, T., 2002. Blue-light-control of the uptake of amino acids and of ammonia in Chlorella mutants. Physiol. Plantarum 116, 248e254. Kaplan, D., Richmond, A.E., Dubinsky, Z., Aaronson, S., 1986. Algal nutrition. In: Richmond, A. (Ed.), Handbook for Microalgal Mass Culture. CRC Press, Boca Raton, FL., USA, pp. 147e198. Khan, S., Rashmi, A., Hussain, M.Z., Prasad, S., Banerjee, U.C., 2009. Prospects of biodiesel production from microalgae in India. Renew. Sust. Energ. Rev. 13, 2361e2372. Komor, E., 1973. Proton-coupled hexose transport in Chlorella vulgaris. FEBS Lett. 38, 16e18. Komor, E., Tanner, W., 1974. The hexose-proton symport system of Chlorella vulgaris: specificity, stoichiometry and energetics of sugar-induced proton uptake. Eur. J. Biochem. 44, 219e223. Komor, E., Tanner, W., 1976. The determination of the membrane potential of Chlorella vulgaris: evidence for electrogenic sugar transport. Eur. J. Biochem. 70, 197e204. Komor, E., Schwab, W.G.W., Tanner, W., 1979. The effect of intracellular pH on the role of hexose uptake in Chlorella. Biochem. Biophys. Acta 555, 524e530. Komor, E., Haass, D., Komor, B., Tanner, W., 1973. The active hexose-uptake system of Chlorella vulgaris. Km-values for 6deoxyglucose influx and efflux and their contribution to sugar accumulation. Eur. J. Biochem. 39, 193e200. Komor, E., Schobert, C., Cho, B.H., 1985. Sugar specificity and sugar-proton interaction for the hexose-proton-symport system of Chlorella. Eur. J. Biochem. 146, 649e656. Kulkarni, M.G., Dalai, A.K., 2006. Waste cooking oil e an economical source for biodiesel: a review. Ind. Eng. Chem. Res. 45, 2901e2913. Lalucat, J., Imperial, J., Pare´s, R., 1984. Utilization of light for the assimilation of organic matter in Chlorella sp. VJ79. Biotechnol. Bioeng. 26, 677e681.
34
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
Lea, P.J., Miflin, B.J., 2003. Glutamate synthase and the synthesis of glutamate in plants. Plant Physiol. Biochem. 41, 555e564. Lebeau, T., Robert, J.M., 2006. Biotechnology of immobilized micro algae: a culture technique for the future? In: Rao, S. (Ed.), Algal Cultures, Analogues of Blooms and Applications. Science Publishers, New Hampshire, USA, pp. 801e837. Lee, Y.-K., 1997. Commercial production of microalgae in the AsiaePacific rim. J. Appl. Phycol. 9, 403e411. Lee, Y.-K., 2001. Microalgal mass culture systems and methods: their limitation and potential. J. Appl. Phycol. 13, 307e315. Lee, Y.-K., 2004. Algal nutrition. Heterotrophic carbon nutrition. In: Richmond, A. (Ed.), Handbook of Microalgal Culture. Biotechnology and Applied Phycology. Blackwell Publishing, Oxford, UK, p. 116. Lee, K., Lee, C.G., 2001. Effect of light/dark cycles on wastewater treatment by microalgae. Biotechnol. Bioprocess Eng. 6, 194e199. Lee, K., Lee, C.G., 2002. Nitrogen removal from wastewaters by microalgae without consuming organic carbon sources. J. Microbiol. Biotechnol. 12, 979e985. Leman, J., 1997. Oleaginous microorganisms: an assessment of the potential. Adv. Appl. Microbiol. 43, 195e243. Leo´n, R., Galva´n, F., 1999. Interaction between saline stress and photoinhibition of photosynthesis in the freshwater green algae Chlamydomonas reinhardtii. Implications for glycerol photoproduction. Plant Physiol. Biochem. 37, 623e628. Lewin, J., Hellebust, J.A., 1978. Utilization of glutamate and glucose for heterotrophic growth by the marine pennate diatom Nitzschia laevis. Mar. Biol. 47, 1e7. Lewitus, A.J., Caron, D.A., Miller, K.R., 1991. Effects of light and glycerol on the organization of the photosynthetic apparatus in the facultative heterotroph Pyrenomonas salina (Cryptophyceae). J. Phycol. 27, 578e587. Li, Q., Du, W., Liu, D., 2008a. Perspectives of microbial oils for biodiesel production. Appl. Microbiol. Biotechnol. 80, 749e756. Li, X., Xu, H., Wu, Q., 2007. Large-scale biodiesel production from microalga Chlorella protothecoides through heterotrophic cultivation in bioreactors. Biotechnol. Bioeng. 98, 764e771. Li, Y., Horsman, M., Wu, N., Lan, C.Q., Dubois-Calero, N., 2008b. Biofuels from microalgae. Biotechnol. Prog. 24, 815e820. Lloyd, D., 1974. Dark respiration. In: Stewart, W.D.P. (Ed.), Algal Physiology and Biochemistry. Blackwell Scientific Publications, Oxford, UK, pp. 505e529. Lopez-Ruiz, A., Verleben, J.P., Bocanegra, J.A., Diez, J., 1991. Immunocytochemical localization of nitrite reductase in green algae. Plant Physiol. 96, 699e704. Lu, B., Yuan, Y., Zhang, C., Ou, J., Zhou, W., Lin, Q., 2005. Modulation of key enzymes involved in ammonium assimilation and carbon metabolism by low temperature in rice (Oryza sativa L.) roots. Plant Sci. 169, 295e302. Ma, R.Y.-N., Chen, F., 2001. Enhanced production of free transastaxanthin by oxidative stress in the cultures of the green microalga Chlorococcum sp. Process Biochem. 36, 1175e1179. Maciasr, F.M., Eppley, R.W., 1963. Development of EDTA media for the rapid growth of Chlamydomonas mundana. J. Protozool 10, 243e246. Mandalam, R.K., Palsson, B.Ø., 1998. Elemental balancing of biomass and medium composition enhances growth capacity in high-density Chlorella vulgaris cultures. Biotechnol. Bioeng. 59, 605e611. Martin-Je´ze´quel, V., Hildebrand, M., Brzezinski, M.A., 2000. Silicon metabolism in diatoms: implications for growth. J. Phycol. 36, 821e840. Martı´nez, F., Oru´s, M.I., 1991. Interactions between glucose and inorganic carbon metabolism in Chlorella vulgaris strain UAM101. Plant Physiol. 95, 1150e1155. Martinez, F., Ascaso, C., Orus, M.I., 1991. Morphometric and stereologic analysis of Chlorella vulgaris under heterotrophic growth conditions. Ann. Bot. 67, 239e245.
Mata, T.M., Martins, A.A., Caetano, N.S., 2010. Microalgae for biodiesel production and other applications: a review. Renew. Sust. Energ. Rev. 14, 217e232. Merchant, S.S., et al., 2007. The Chlamydomonas genome reveals the evolution of key animal and plant functions. Science 318, 245e250. Merzlyak, M.N., Chivkunova, O.B., Gorelova, O.A., Reshetnikova, I.V., Solovchenko, A.E., Khozin-Goldberg, I., Cohen, Z., 2007. Effect of nitrogen starvation on optical properties, pigments and arachidonic acid content of the unicellular green alga Parietochloris incisa (Trebouxiophyceae, Chlorophyta). J. Phycol 43, 833e843. Miao, X., Wu, Q., 2006. Biodiesel production from heterotrophic microalgal oil. Bioresour. Technol. 97, 841e846. Miflin, B.J., Habash, D.Z., 2002. The role of glutamine synthetase and glutamate dehydrogenase in nitrogen assimilation and possibilities for improvement in the nitrogen utilization of crops. J. Exp. Bot. 53, 979e987. Miller, P.W., Dunn, W.I., Schmidt, R.R., 1998. Alternative splicing of a precursor-mRNA encoded by the Chlorella sorokiniana NADP-specific glutamate dehydrogenase gene yields mRNAs for precursor proteins of isozyme subunits with different ammonium affinities. Plant Mol. Biol. 37, 243e263. Millbank, J.W., 1957. Keto-acids in the alga Chlorella. Ann. Bot. 21, 23e31. Molina Grima, E., Acie´n Ferna´ndez, F.G., Garcı´a Camacho, F., Chisti, Y., 1999. Photobioreactors: light regime, mass transfer, and scaleup. J. Biotechnol. 70, 231e247. Molina Grima, E., Belarbi, E.H., Acie´n Ferna´ndez, F.G., Robles Medina, A., Chisti, Y., 2003. Recovery of microalgal biomass and metabolites: process options and economics. Biotechnol. Adv. 20, 491e515. Moreno-Garrido, I., 2008. Microalgae immobilization: current techniques and uses. Bioresour. Technol. 99, 3949e3964. Morris, I., 1974. Nitrogen assimilation and protein synthesis. In: Stewart, W.D.P. (Ed.), Algal Physiology and Biochemistry. Blackwell Scientific Publications, Oxford, UK, pp. 513e613. Mun˜oz, R., Guieysse, B., 2006. Algal-bacterial processes for the treatment of hazardous contaminants: a review. Water Res. 40, 2799e2815. Murarka, A., Dharmadi, Y., Yazdani, S.S., Gonzalez, R., 2008. Fermentative utilization of glycerol by Escherichia coli and its implications for the production of fuels and chemicals. Appl. Environ. Microbiol. 74, 1124e1135. Nakamura, Y., Ikawa, T., 1993. Purification and properties of NADH: nitrate reductase from red alga Porphyra yezoensis. Plant Cell Physiol. 34, 1239e1249. Narang, A., Pilyugin, S.S., 2005. Towards an integrated physiological theory of microbial growth: from subcellular variables to population dynamics. Math. Biosci. Eng. 2, 173e210. Neilson, A.H., Larsson, T., 1980. The utilization of organic nitrogen for growth of algae: physiological aspects. Physiol. Plantarum 48, 542e553. Neilson, A.H., Lewin, R.A., 1974. The uptake and utilization of organic carbon by algae: an essay in comparative biochemistry. Phycologia 13, 227e264. O’Brien, R.W., Taylor, B.L., 1977. Formation and dissimilation of oxalacetate and pyruvate in Pseudomonas citronellolis grown on noncarbohydrate substrates. J. Bacteriol. 130, 131e135. Ogawa, T., Aiba, S., 1981. Bioenergetic analysis of mixotrophic growth in Chlorella vulgaris and Scenedesmus acutus. Biotech. Bioeng. 23, 1121e1132. Ogbonna, J.C., Tomiyama, S., Tanaka, H., 1998. Heterotrophic cultivation of Euglena gracilis Z for efficient production of a-tocopherol. J. Appl. Phycol. 10, 67e74. Ogbonna, J.C., Yoshizawa, H., Tanaka, H., 2000. Treatment of high strength organic wastewater by a mixed culture of photosynthetic microorganisms. J. Appl. Phycol. 12, 277e284.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
Oh-Hama, T., Miyachi, S., 1992. Chlorella. In: Borowitzka, M.A., Borowitzka, L.J. (Eds.), Microalgal Biotechnology. Cambridge University Press, Cambridge, UK, pp. 3e26. Oswald, W.J., 1992. Micro-algae and waste-water treatment. In: Borowitzka, M.A., Borowitzka, L.J. (Eds.), Microalgal Biotechnology. Cambridge University Press, Cambridge, pp. 305e328. Patil, V., Tran, K.-Q., Giselrød, H.R., 2008. Towards sustainable production of biofuels from microalgae. Int. J. Mol. Sci. 9, 1188e1195. Perez-Garcia, O., de-Bashan, L.E., Hernandez, J.-P., Bashan, Y., 2010. Efficiency of growth and nutrient uptake from wastewater by heterotrophic, autotrophic, and mixotrophic cultivation of Chlorella vulgaris immobilized with Azospirillum brasilense. J. Phycol 46, 800e812. Perez-Garcia, R.O., Bashan, Y., Puente M.E. Organic carbon supplementation of municipal wastewater is essential for heterotrophic growth and ammonium removing by the microalgae Chlorella vulgaris. J. Phycol., in press. Pipes, W.O., Gotaas, H.B., 1960. Utilization of organic matter by Chlorella grown in seawage. Appl. Microbiol. 8, 163e169. Post, A.F., Cohen, I., Romen, E., 1994. Characterization of two Chlorella vulgaris (Chlorophyceae) strains isolated from wastewater oxidation ponds. J. Phycol. 30, 950e954. Pulz, O., 2001. Photobioreactors: production systems for phototrophic microorganisms. Appl. Microbiol. Biotechnol. 57, 287e293. Pulz, O., Gross, W., 2004. Valuable products from biotechnology microalgae. Appl. Microbiol. Biotechnol. 65, 635e648. Quoc, K.P., Dubacq, J.P., 1997. Effect of growth temperature on the biosynthesis of eukaryotic lipid molecular species by the cyanobacterium Spirulina platensis. Biochem. Biophys. Acta 1346, 237e246. Radmer, R.J., Parker, B.C., 1994. Commercial applications of algae: opportunities and constraints. J. Appl. Phycol 6, 93e98. Rahman, R.N.A., Jongsareejit, B., Fujiwara, S., Imanaka, T., 1997. Characterization of recombinant glutamine synthetase from the hyperthermophilic archeon Pyrococcus sp. strain KOD1. Appl. Environ. Microbiol. 63, 2472e2476. Ratledge, C., Evans, C.T., 1989. Lipids and their metabolism. In: Rose, A.H., Harrison, J.S. (Eds.), The Yeasts. Academic Press, New York, USA, pp. 367e455. Ratledge, C., Wynn, J.P., 2002. The biochemistry and biology of lipid accumulation in oleaginous microorganisms. Adv. Appl. Microbiol. 51, 1e51. Ratledge, C., Kanagachandran, K., Anderson, A.J., Grantham, D.J., Stephenson, J.C., 2001. Production of docosahexaenoic acid by Crypthecodinium cohnii grown in a pH-auxostat culture with acetic acid as principal carbon source. Lipids 36, 1241e1246. Raven, J.A., 1976. The quantitative role of dark respiratory processes in heterotrophic and photolithotrophic plant growth. Ann. Bot. 40, 587e602. Rebeille, F., Gans, P., 1988. Interaction between chloroplasts and mitochondria in microalgae. Plant Physiol. 88, 973e975. Richmond, A., 1986. Cell response to environmental factors. In: Richmond, A. (Ed.), Handbook for Microalgal Mass Culture. CRC Press, Boca Raton, FL., USA, pp. 69e99. Roessler, P.G., 1988. Effects of silicon deficiency on lipid composition and metabolism in the diatom Cyclotella cryptica. J. Phycol 24, 394e400. Sakamoto, T., Bryant, D.A., 1997. Temperature-regulated mRNA accumulation and stabilization for fatty acid desaturase genes in cyanobacterium Synechococcus sp. strain PCC 7002. Mol. Microbiol. 23, 1281e1292. Sansawa, H., Endo, H., 2004. Production of intracellular phytochemicals in Chlorella under heterotrophic conditions. J. Biosci. Bioeng. 98, 437e444.
35
Sato, N., Murata, N., 1986. Temperature shift-induces responses in lipids in the blue green alga, Anabaena variabilis: the central role of dyacylmonogalactosylglycerol in thermo adaptation. Biochim. Biophys. Acta 619, 353e366. Sauer, N., Tanner, W., 1989. The hexose carrier from Chlorella. cDNA cloning of a eukaryotic Hþ-cotransporter. FEBS Lett. 259, 43e46. Schmidt, R.A., Wiebe, M.G., Eriksen, N.T., 2005. Heterotrophic high cell-density fed-batch cultures of the phycocyaninproducing red alga Galdieria sulphuraria. Biotechnol. Bioeng. 90, 77e84. Sharif Hossain, A.B.M., Salleh, A., Boyce, A.N., Chowdhury, P., Naqiuddin, M., 2008. Biodiesel fuel production from algae as renewable energy. Am. J. Biochem. Biotechnol. 4, 250e254. Sherman, T.D., Funkhouser, E.A., 1989. Induction and synthesis of nitrate reductase in Chlorella vulgaris. Arch. Biochem. Biophys. 274, 525e531. Shi, X.-M., Chen, F., 2002. High-yield production of lutein by the green microalga Chlorella protothecoides in heterotrophic fedbatch culture. Biotechnol. Prog. 18, 723e727. Shi, X.-M., Chen, F., Yuan, J.-P., Chen, H., 1997. Heterotrophic production of lutein by selected Chlorella strains. J. Appl. Phycol. 9, 445e450. Shi, X.-M., Liu, H.-J., Zhang, X.-W., Chen, F., 1999. Production of biomass and lutein by Chlorella protothecoides at various glucose concentrations in heterotrophic cultures. Process Biochem. 34, 341e347. Shi, X.-M., Zhang, X.-W., Chen, F., 2000. Heterotrophic production of biomass and lutein by Chlorella protothecoides on varios nitrogen sources. Enzyme Microb. Technol. 27, 312e318. Sijtsma, L., de Swaaf, M.E., 2004. Biotechnological production and applications of the u-3 polyunsaturated fatty acid docosahexaenoic acid. Appl. Microbiol. Biotechnol. 64, 146e153. Sijtsma, L., Anderson, A.J., Ratledge, C., 2005. Alternative carbon sources for heterotrophic production of docosahexaenoic acid by the marine alga Crypthecodinium cohnii. In: Cohen, Z., Ratledge, C. (Eds.), Single Cell Oils. AOCS Press, Champaign, IL, USA, pp. 107e123. Sivakumar, G., Vail, D.R., Xu, J., Burner, D.M., Lay Jr., J.O., Ge, X., Weathers, P.J., 2010. Bioethanol and biodiesel: alternative liquid fuels for future generations. Eng. Life Sci. 10, 8e18. Sloth, J.K., Wiebe, M.G., Eriksen, N.T., 2006. Accumulation of phycocyanin in heterotrophic and mixotrophic cultures of the acidophilic red alga Galdieria sulphuraria. Enzyme Microb. Technol. 38, 168e175. Somerville, C., Browse, J., Jaworski, J.G., Ohlrogge, J.B., 2000. Lipids. In: Buchanan, B., Gruissem, W., Jones, R. (Eds.), Biochemistry and Molecular Biology of Plants. American Society of Plant Physiology, Rockville, MD, USA, pp. 465e527. Song, D., Fu, J., Shi, D., 2008. Exploitation of oil-bearing microalgae for biodiesel. Chin. J. Biotechnol. 24, 341e348. Sonnenborn, U., Kunau, W.H., 1982. Purification and properties of the fatty acid synthetase complex from the marine dinoflagellate, Crypthecodinium cohnii. Biochem. Biophys. Acta 712, 523e534. Suh, I.S., Lee, C.-G., 2003. Photobioreactor engineering: design and performance. Biotechnol. Bioprocess Eng. 8, 313e321. Syreth, P.J., Morris, I., 1963. The inhibition of nitrate assimilation by ammonium in Chlorella. Biochem. Biophys. Acta 67, 566e575. Tan, C.K., Johns, M.R., 1991. Fatty acid production by heterotrophic Chlorella saccharophila. Hydrobiologia 215, 13e19. Tan, C.K., Johns, M.R., 1996. Screening of diatoms for heterotrophic eicosapentaenoic acid production. J. Appl. Phycol. 8, 59e64.
36
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 e3 6
Tanner, W., 1969. Light-driven active uptake of 3-Omethylglucose via an inducible hexose uptake system of Chlorella. Biochem. Biophys. Res. Commun. 36, 278e283. Tanner, W., 2000. The Chlorella hexose/H(þ)-symporters. Int. Rev. Cytol. 200, 101e141. Tatsuzawa, H., Takizawa, E., 1995. Changes in lipid and fatty acid composition of Pavlova lutheri. Phytochemistry 40, 397e400. Theriault, R.J., 1965. Heterotrophic growth and production of xanthophylls by Chlorella pyrenoidosa. Appl. Microbiol. 13, 402e416. Tischner, R., 1984. Evidence for the participation of NADP glutamate dehydrogenase in the ammonium assimilation of Chlorella sorokiniana. Plant Sci. Lett. 34, 73e80. Tredici, M.R., 1999. Bioreactors, photo. In: Flickinger, M.C., Drew, S.W. (Eds.), Encyclopedia of Bioprocess Technology: Fermentation, Biocatalysis and Bioseparation. Wiley, New York, USA, pp. 395e419. Tredici, M.R., 2004. Mass production of microalgae: photobioreactors. In: Richmond, A. (Ed.), Handbook of Microalgal Culture. Biotechnology and Applied Phycology. Blackwell Publishing, Oxford, UK, pp. 178e214. Tripathi, U., Sarada, R., Ramachandra Rao, S., Ravishankar, G.A., 1999. Production of astaxanthin in Haematococcus pluvialis cultured in various media. Bioresour. Technol. 68, 197e199. Tsavalos, A.J., Day, J.G., 1994. Development of media for the mixotrophic/heterotrophic culture of Brachiomonas submarina. J. Appl. Phycol 6, 431e433. Tsygankov, A.A., 2000. Laboratory scale photobioreactors. Appl. Biochem. Microbiol. 37, 333e341. Tuchman, N., 1996. The role of heterotrophy in algae. In: Stevenson, R.J., Bothwell, M., Lowe, R.L. (Eds.), Algal Ecology. Freshwater Benthic Ecosystems. Academic Press, San Diego, CA, USA, pp. 299e319. Vanoni, M.A., Curti, B., 2005. Structure-function studies on the iron-sulfur flavoenzyme glutamate synthase: an unexpectedly complex self-regulated enzyme. Arch. Biochem. Biophys. 433, 193e211. Vazhappilly, R., Chen, F., 1998. Eicosapentaenoic acid and docosahexaenoic acid production potential of microalgae and their heterotrophic growth. J. Am. Oil Chem. Soc. 75, 393e397. Wang, Y., Peng, J., 2008. Growth-associated biosynthesis of astaxanthin in heterotrophic Chlorella zofingiensis (Chlorophyta). World J. Microbiol. Biotechnol. 24, 1915e1922. Wen, Z.-Y., Chen, F., 2000. Heterotrophic production of eicosapentaenoic acid by the diatom Nitzschia laevis: effects of silicate and glucose. J. Ind. Microbiol. Biotechnol. 25, 218e224. Wen, Z.-Y., Chen, F., 2001a. Optimization of nitrogen sources for heterotrophic production of eicosapentaenoic acid by the diatom Nitzschia laevis. Enzyme Microb. Technol. 29, 341e347. Wen, Z.-Y., Chen, F., 2001b. A perfusionecell bleeding culture strategy for enhancing the productivity of eicosapentaenoic acid by Nitzschia laevis. Appl. Microbiol. Biotechnol. 57, 316e322. Wen, Z.-Y., Chen, F., 2003. Heterotrophic production of eicosapentaenoic acid by microalgae. Biotechnol. Adv. 21, 273e294. Wilhelm, C., Bu¨chel, C., Fisahn, J., Goss, R., Jakob, T., LaRoche, J., Lavaud, J., Lohr, M., Riebesell, U., Stehfest, K., Valentin, K., Kroth, P.G., 2006. The regulation of carbon and nutrient
assimilation in diatoms is significantly different from green algae. Protist 157, 91e124. Wood, B.J.B., Grimson, P.H.K., German, J.B., Turner, M., 1999. Photoheterotrophy in the production of phytoplankton organisms. J. Biotechnol. 70, 175e183. Wu, Z., Shi, X., 2007. Optimization for high-density cultivation of heterotrophic Chlorella based on a hybrid neural network model. Lett. Appl. Microbiol. 44, 13e18. Xiong, W., Li, X., Xiang, J., Wu, Q., 2008. High-density fermentation of microalga Chlorella protothecoides in bioreactor for microbio-diesel production. Appl. Microbiol. Biotechnol. 78, 29e36. Xu, H., Miao, X., Wu, Q., 2006. High quality biodiesel production from a microalga Chlorella protothecoides by heterotrophic growth in fermenters. J. Biotechnol. 126, 499e507. Yang, C., Hua, Q., Shimizu, K., 2000. Energetics and carbon metabolism during growth of microalgal cells under photoautotrophic, mixotrophic and cyclic light-autotrophic/ dark-heterotrophic conditions. Biochem. Eng. J. 6, 87e102. Yang, C., Hua, Q., Shimizu, K., 2002. Integration of the information from gene expression and metabolic fluxes for the analysis of the regulatory mechanisms in Synechocystis. Appl. Microbiol. Biotechnol. 58, 813e822. Yazdani, S.S., Gonzalez, R., 2007. Anaerobic fermentation of glycerol: a path to economic viability for the biofuels industry. Curr. Opin. Biotechnol. 18, 213e219. Yokochi, T., Honda, D., Higashihara, T., Nakahara, T., 1998. Optimization of docosahexaenoic acid production by Schizochytrium limacinum SR21. Appl. Microbiol. Biotechnol. 49, 72e76. Yongmanitchai, W., Ward, O.P., 1991. Growth of and omega-3 fatty acid production by Phaeodactylum tricornutum under different culture conditions. Appl. Environ. Microbiol. 57, 419e425. Zaslavskaia, L.A., Lippmeier, J.C., Shih, C., Ehrhardt, D., Grossman, A.R., Apt, K.E., 2001. Trophic conversion of an obligate photoautotrophic organism through metabolic engineering. Science 292, 2073e2075. Zhang, K., Miychi, S., Kurano, N., 2001. Evaluation of a vertical flat-plate photobioreactor for outdoor biomass production and carbon dioxide bio-fixation: effects of reactor dimensions, irradiation and cell concentration on the biomass productivity and irradiation utilization efficiency. Appl. Microbiol. Biotechnol. 55, 428e433. Zhang, X.W., Chen, F., Johns, M.R., 1999. Kinetic models for heterotrophic growth of Chlamydomonas reinhardtii in batch and fed-batch cultures. Process Biochem. 35, 385e389. Zhekisheva, M., Boussiba, S., Khozin-Goldberg, I., Zarka, A., Cohen, Z., 2002. Accumulation of oleic acid in Haematococcus pluvialis (Chlorophyceae) under nitrogen starvation or high light, is correlated with that of astaxanthin esters. J. Phycol. 38, 325e331. Zhu, L., Zhang, X., Ji, L., Song, X., Kuang, C., 2007. Changes of lipid content and fatty acid composition of Schizochytrium limacinum in response to different temperatures and salinities. Process Biochem. 42, 210e214. Zijffers, J.-W.F., Janssen, M., Tramper, J., Wijffels, R.H., 2008. Design process of an area-efficient photobioreactor. Mar. Biotechnol. 10, 404e415.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 e5 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Assessment of the integrated urban water quality model complexity through identifiability analysis Gabriele Freni a, Giorgio Mannina b,*, Gaspare Viviani b a b
Facolta` di Ingegneria ed Architettura, Universita` degli Studi di Enna “Kore”, Cittadella Universitaria, 94100 Enna, Italy Dipartimento di Ingegneria Idraulica ed Applicazioni Ambientali, Universita` di Palermo, Viale delle Scienze, 90128 Palermo, Italy
article info
abstract
Article history:
Urban sources of water pollution have often been cited as the primary cause of poor
Received 8 February 2010
water quality in receiving water bodies (RWB), and recently many studies have been
Received in revised form
conducted to investigate both continuous sources, such as wastewater-treatment plant
29 May 2010
(WWTP) effluents, and intermittent sources, such as combined sewer overflows (CSOs).
Accepted 3 August 2010
An urban drainage system must be considered jointly, i.e., by means of an integrated
Available online 11 August 2010
approach. However, although the benefits of an integrated approach have been widely demonstrated, several aspects have prevented its wide application, such as the scarcity
Keywords:
of field data for not only the input and output variables but also parameters that govern
Uncertainty assessment
intermediate stages of the system, which are useful for robust calibration. These factors,
River water-quality modelling
along with the high complexity level of the currently adopted approaches, introduce
Identifiability analysis
uncertainties in the modelling process that are not always identifiable. In this study, the
Integrated urban drainage
identifiability analysis was applied to a complex integrated catchment: the Nocella basin
modelling
(Italy). This system is characterised by two main urban areas served by two WWTPs and has a small river as the RWB. The system was simulated by employing an integrated model developed in previous studies. The main goal of the study was to assess the right number of parameters that can be estimated on the basis of data-source availability. A preliminary sensitivity analysis was undertaken to reduce the model parameters to the most sensitive ones. Subsequently, the identifiability analysis was carried out by progressively considering new data sources and assessing the added value provided by each of them. In the process, several identifiability methods were compared and some new techniques were proposed for reducing subjectivity of the analysis. The study showed the potential of the identifiability analysis for selecting the most relevant parameters in the model, thus allowing for model simplification, and in assessing the impact of data sources for model reliability, thus guiding the analyst in the design of future monitoring campaigns. Further, the analysis showed some critical points in integrated urban drainage modelling, such as the interaction between water quality processes on the catchment and in the sewer, that can prevent the identifiability of some of the related parameters. ª 2010 Elsevier Ltd. All rights reserved.
* Corresponding author. Tel.: þ39 091 665 7756; fax: þ390916657749. E-mail address:
[email protected] (G. Mannina). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.004
38
1.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 e5 0
Introduction
Integrated modelling of urban wastewater systems is of growing interest, mainly as a result of the recent adoption of the EU Water Framework Directive (WFD) (European Commission, 2000). An integrated modelling approach is also required due to the concurrently growing awareness that optimal management of the individual components of urban wastewater systems (i.e., sewer systems, wastewater-treatment plants and receiving water bodies) does not lead to optimum performance of the entire system (Rauch et al., 2002). One of the main bottlenecks preventing the application of integrated modelling approaches is the complexity of the overall system as well as the lack of field data required for reliable model application. Indeed, in urban drainage water quality assessment, data availability issues are generally quite common in both research and practical applications. Such problems are primarily due to the fact that the required datagathering campaigns can be technically complex and economically demanding. When dealing with complex modelling approaches in the context of insufficient field data, classical calibration approaches may lead to several equally consistent parameter sets and it may thus prove difficult to arrive at sufficient confidence in the obtained results (Kuczera and Parent, 1998; Beven and Binley, 1992). An obvious remedy is model reduction in the sense of restricting the model description to only the observed data (Jakeman and Hornberger, 1993). This theoretical principle has some difficulties in practice related to the definition of an objective procedure for determining the correct model complexity for a specific application. Identifiability analysis enables a response to such an issue, consisting of several mathematical approaches aimed at the investigation of modelling parameters that can be reliably assessed in a specific modelling application and in a specific case study. Model identifiability analysis basically consists of two problems: the problem of model-structure selection and the problem of parameter identification. The model structure is often imposed by physical considerations, especially with large environmental systems involving several processes. For this reason, studies to date have mainly addressed parameter identifiability and the evaluation of related uncertainty (Brun et al., 2001; Campolongo et al., 2007). A distinction has to be made between structural and practical identifiability (De Pauw et al., 2004). The former provides information about the theoretical possibility of obtaining unique values for the parameters once the model structure and the system to be modelled have been established. In contrast, the practical identifiability of parameters is dependent on both model structure and experimental conditions together with the quality and quantity of the measurements. In the past, parameter identifiability issues, although referring to simple models, have been successfully tackled by detailed analysis of sensitivity functions (Holmberg, 1982; Reichert and Vanrolleghem, 2001; Saltelli et al., 2006; Wagener and Kollat, 2007; Campolongo et al., 2007; Gatelli et al., 2009). Holmberg (1982) suggested the use of graphical approaches for sensitivity analysis to enable the evaluation of parameters identifiability. Such approaches are well suited for
small models. Conversely, regarding large models, such as activated-sludge models (ASMs), the previous approach fails due to the fact that it is no longer possible to efficiently analyse the extensive graphical output that is produced (Brun et al., 2002). To cope with such problems, several analytical approaches were presented in literature based on detailed analyses of the sensitivity indices. Morris (1991) and Campolongo et al. (2007) proposed the analysis of Elementary Effects (EEs) of parameters on modelling output based on the statistical analysis of model sensitivities to parameter variations. In such studies, the average value of the EEs is used to rank the parameters in terms of sensitivity. Saltelli et al. (2009) suggested some improvements to the method introducing the concept of Elementary Interaction in order to highlight the interaction among parameters in terms of their impact on modelling outputs. Weijers and Vanrolleghem (1997) and De Pauw (2005), transferring knowledge from the field of control theory, demonstrated the effectiveness as well as the power of FIM-based. The main advantage of such methods is related to the objectivity of identifiability criteria that are not dependent decisions, such as the definition of a threshold in the sensitivity indices to highlight identifiable parameters. In another approach, Brun et al. (2001), adapting methods used in linear regression diagnostics (Belsley, 1991), focused on the analysis of parameter interdependencies and on the exploration of the effects of fixed parameter values on parameter estimates. Both studies showed that the different proposed methods are of variable effectiveness depending on the structure and number of parameters involved in the model; such approaches also have very different computational costs and they are often dependent on user assumptions (Brun et al., 2001). Another study, carried out by Malve et al. (2007), demonstrated that an identifiability analysis based on Bayes’ paradigm could be used for better fitting in environmental modelling and selecting potential measurements. Malve et al. (2007) suggested to use the environmental modelling as a tool for guiding data-gathering campaigns. The methods based on EE demonstrated high computational efficiency, especially after the modifications and the improvements produced in the last decade. The methods based on FIM analysis have the advantage of being less affected by subjective choices of the operator (Freni et al., 2009a; Machado et al., 2009). Finally, Malve et al. (2007) pointed out that Bayesian methods are more data demanding than other identifiability methods and for this reason they are often not readily applicable. For this reasons, methods based on FIM analysis was frequently adopted in integrated urban drainage water-quality modelling for both its simple use and for the low impact of subjective choices. Moreover, Freni et al. (2009a) investigated the reduction of overall modelling uncertainty that can be obtained by fixing some parameters constant (non-identifiable) according to the results of the identifiability analysis. Despite the useful insights gained by Freni et al. (2009a), the effects of the overall data contributions of the different parts of the integrated system were not investigated; the investigation of these effects represents one of the aims of the present study. The Freni et al. (2009a) study was based solely on river flow and water quality data, not including the information coming from the other parts of the integrated system (i.e., the sewer
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 e5 0
system and wastewater-treatment plant). However, in the case of integrated models the analysis of the identifiable parameters on the basis of the whole body of information coming from the different parts of the integrated system is of paramount interest and deserves investigation. Indeed, the uncertainties in parameters and input data propagate through a chain of interacting models running parallel simulations. More control of information transfer between time steps allows for an improved analysis of model-system dynamics. Bearing in mind the considerations discussed above, identifiability analysis is applied to a complex case study in which several data sources are present (i.e., sewer systems, wastewater-treatment plants and a receiving water body) and the related model is characterised by numerous parameters thus increasing response uncertainty. This study attempts to assess the right number of parameters that can be estimated on the basis of data source availability. During the process, several previously published indicators are employed and a novel one is proposed for reducing the subjectivity of the identifiability analysis.
2.
Materials and methods
2.1.
Description of the case study
The analysis was applied to a complex integrated catchment, the Nocella catchment (Fig. 1), which is an urbanised natural catchment located near Palermo in the northwestern part of
39
Sicily (Italy). The entire natural basin is characterised by a surface area of 9970 ha and has two main branches that flow primarily east to west. The two main branches join together 3 km upstream of the river estuary. The southern branch is characterised by a smaller elongated basin and receives water from a large urban area characterised by industrial activities partially served by a WWTP and partially connected directly to the RWB. The northern branch was monitored in the present study. The basin closure is located 9 km upstream of the river mouth; the catchment area is 6660 ha. The catchment end is equipped with a hydrometeorological station (Nocella a Zucco). The northern river reach receives wastewater and stormwater from two urban areas (Montelepre, with a catchment surface of 70 ha, and Giardinello, with a surface of 45 ha) drained by combined sewers. The Montelepre sewer consists of circular and egg-shaped pipes with maximum dimensions of 100 cm 150 cm. The sewer system serves 7000 inhabitants and has an average dry-weather flow of 12.5 L/s and an average dry-weather biological oxygen demand (BOD) of 223 mg/L. The Giardinello sewer consists of circular pipes with a maximum diameter of 80 cm. The served population is 2000 inhabitants and the system has an average dry-weather flow of 2.5 L/s and an average dry-weather BOD concentration of 420 mg/L. Each sewer system is connected to a WWTP protected by combined sewer overflow (CSO) devices. The WWTPs utilise a simplified activated-sludge process for the organic biological carbon removal with preliminary mechanical treatment units, an activated-sludge tank, and a final
Fig. 1 e Nocella catchment.
40
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 e5 0
circular settler. Rainfall was monitored by four rain gauges distributed over the basin area: the Montelepre rain gauge is operated by Palermo University and is characterised by a 0.1-mm tipping bucket and a temporal resolution of 1 min; the other three rain gauges are operated by the Regional Hydrological Service and are characterised by a 0.2-mm tipping bucket and a temporal resolution of 15 min. The hydrometeorological station located at the end of the catchment (“Nocella a Zucco”, operated by the Regional Hydrological Service) is characterised by an ultrasonic level gauge and has a temporal resolution of 15 min. The instruments were integrated by Palermo University technicians by installing, for the quantity data, an area e velocity submerged probe that provides water level and velocity data with a 1-min temporal resolution. An ultrasonic external probe was used to provide a second water-level measurement for validation and as a backup in case the submerged probe failed; an automatic 24-bottle water-quality sampler was used for water-quality data collection. The monitoring was carried out considering both permanent (based on measuring stations already present) and temporary measures (i.e. based on measuring stations on purpose located) (Fig. 2). Flow measurements were carried out using area e velocity probes with a 1-min temporal resolution, which allow the inflow and outflow volumes for each element in the system to be defined. Water-quality sampling was performed using automatic 24-bottle samplers and grab sampling was used for defining pollutant loads and treatment efficiencies. The water-quality parameters monitored were total suspended solids (TSS), BOD, chemical oxygen demand (COD), ammonia (NH4), total Kjeldahl nitrogen (TKN), and phosphorus (P); dissolved oxygen (DO) level was monitored in the river only. All analyses were carried out according to Standard Methods (APHA, 1995). The monitoring campaign began in December 2006 and is still in progress. Rainfall and discharge were monitored continuously, while water quality was measured during specific periods. Further details concerning the case study and monitoring campaign can be found in Freni et al. (2010a) and Candela et al. (2009).
2.2.
The integrated urban drainage model
In the present study, an integrated model developed in previous studies was applied (Mannina et al., 2004; Mannina, 2005). A brief description of the structure of the adopted model follows; the interested reader may refer to the cited literature for a more detailed description of the selected algorithms. The model enables estimation of both the interactions among the three components of the system (sewer system e SS, WWTP and RWB) and the effects, in terms of quality, that urban stormwater causes inside the RWB (Fig. 3). The integrated model is chiefly composed of three sub-models for the simulation of the components; each sub-model is divided into a quantity and quality module for the simulations of the hydrographs and pollutographs. The modelling structure can be adapted to the specific application by removing or adding submodels, such as the stormwater tank (SWT) or CSO (Freni et al., 2010b). The SS sub-model calculates the net rainfall from the measured rainfall by a loss function taking into account both initial and continuous losses (W0 and F, respectively). From the net rainfall, the model simulates the net rainfall-runoff transformation process and the flow propagation with a cascade of one linear reservoir and a channel, representing the catchment, and a linear reservoir representing the sewer network (characterized by the parameters K1, l and K2, respectively). An exponential function is used to simulate water buildup on catchment surfaces (Alley and Smith, 1981). Such an equation depends by two parameters the buildup rate (Accu) and the decay rate (Disp) that control the accumulation of pollutants on the catchment surface. The solid wash-off caused by overland flow during a storm event is simulated using the formulation proposed by Jewell and Adrian (1978) where the wash-off coefficient (Arra) and washoff factor (Wh) are the two parameters that enable one to calculate the washed mass of pollutants from the catchment due to a rainfall event. The solid deposits in the sewers during dry weather are calculated by using an exponential function. Regarding the erosion and transport of sewer sediments, to ensure a realistic approach, particular care is taken regarding
Fig. 2 e Schematic of the urban drainage system monitoring methodology performed on the Montelepre and Giardinello urban areas.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 e5 0
Fig. 3 e Schematic overview of the different submodels, analysed processes, and interconnections.
sediment transformations in sewers due to their semi-cohesive behaviour due to the presence of organic substances and the physicalechemical changes during sewer transport. Specifically, the eroded mass from the sewer bottom is calculated according to the Parchure and Metha’s approach (1985) where M is the key parameter for assessing such a mass. The pollutographs at the outlet of the sewer system are calculated by modelling the complex catchment sewer network as a reservoir and singling out the different types of sewer sediment transport (i.e. suspended and bed load transport). The two types of sediment transport are propagated considering two coefficients: the sewer suspended load linear reservoir constant (Ksusp) and the sewer bed load linear reservoir constant (Kbed). Further, the different types of sewer sediment transport are calculated taking into account the transport capacity of the flow (see, Mannina and Viviani, 2010a). Finally, the WWTP inflow is computed by taking into account the presence of a CSO device; its behaviour was simulated by a rating curve, where CSO efficiency is taken into account by the introduction of two dilution coefficients (rd1 and rd2) (Mannina and Viviani, 2009). The WWTP sub-model simulates the behaviour of the most sensitive part of the plant with respect to storm events; accordingly, the model simulates a plant composed of an activated-sludge tank and a secondary sedimentation tank. In the activated-sludge-tank model, the equations derived from Monod’s theory (Metcalf and Eddy, 2003) are used to describe the removal of BOD and NH4. Specifically, the BOD removal is controlled by: the maximum yield coefficient of heterotrophs (mmax,H), BOD semi saturation constant (Ks), the yield coefficient heterotrophic (YH), the decay velocity of heterotrophs. On the other hand, the NH4 removal is related to the autotroph biomass and accordingly is controlled by the following parameters. The maximum yield coefficient of autotrophs (mmax,A), the yield coefficient autotrophic (YN) and the decay velocity of
41
autotrophs (bA). The sedimentation tank is simulated using the modelling approach of Taka´cs et al. (1991). In particular, the model predicts the solids concentration profile in the settler by dividing the settler into a number of layers of constant thickness and performing a solids balance for each layer. The third sub-model assesses RWB discharges and water quality. More specifically, the modelling approach is focused on rivers characterised by scarce field data and ephemeral characteristics (i.e., rivers characterised by a long dry season and intense flows for short periods following precipitation). This latter aspect is relevant as the phenomena generally involved in the evaluation of the RWB quality state play different roles with respect to the perennial streams commonly presented in the literature (Freni et al., 2009b; Mannina and Viviani, 2010b). Such rivers are also frequently found in Mediterranean areas characterised by semi-arid climates. Due to the highly non-stationary conditions typical of these ephemeral streams, a dynamic model is employed for the propagation of the river flow. Specifically, the simplified form of the Saint Venat equation (cinematic wave) is used for the propagating the flow throughout the river assuming as solely parameter the river bed roughness (ks). On the other hand, for the quality aspects the advectionedispersion equation was implemented to address the water-quality phenomena (Mannina and Viviani, 2010c; Chapra, 1997; Brown and Barnwell, 1987). Specifically, the BOD and DO propagation was assessed considering a longitudinal dispersion coefficient (Kdisp) and kinetic constants for the transformation of the BOD (kd and ksod) and oxygen reaeration (ka).
2.3.
Model identifiability analysis
Most of the techniques designed to find practically identifiable subsets of model parameters are based on an investigation of sensitivity functions. The present study concentrates on numeric criteria based on correlation studies of sensitivity functions (Weijers and Vanrolleghem, 1997; Checchi and Marsili-Libelli, 2005; Saltelli et al., 2006, 2009; Campolongo et al., 2007; Marsili-Libelli and Giusti, 2008; Freni et al., 2009a; Gatelli et al., 2009). Many of the methods briefly discussed in the introduction rely on subjective hypotheses (such as the definition of a sensitivity threshold for defining identifiable parameters). In the present study, the analysis was carried out investigating FIM determinant and eigenvalues because it is less prone to subjectivity and it is successfully applied in the same modelling field in literature. In this section, a brief description of the sensitivity indices and identifiability analysis is presented. We begin with the assumption that a deterministic model can be described by a general set of equations y ¼ f(q), where the vector y ¼ ( y1, y2, .y3) represents the n modelling output variables corresponding to the available measurements y ¼ y1 ; y2 ; .yn and the vector q ¼ (q1, q2, .qm) represents the m model parameters. Independent of the nature of the modelling equations, sensitivity functions can be defined stating the relevance of the dependencies between modelling outputs y and parameters q:
si;j ¼
Dqj vyi ysi vqj
(1)
42
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 e5 0
where Dqj is the variability range of parameter qj (which depends on prior knowledge) and ysi is a reference (or scaling) value for the modelling output variable yi, used for preserving the dimensionless nature of the sensitivity function. The function si;j is useful because it provides information on the raw dependency of the modelling output on the parameters. The parameters Dqj and ysi, the magnitude and the scaling parameter, respectively, of the sensitivity function, can each have a great influence on the results of the sensitivity analysis (Reichert and Vanrolleghem, 2001). In the present study, ysi is defined as the average measured value of the ith model output variable, and Dqj can be taken as the variation range of the jth model parameter obtained according to single-event model calibrations based on each available rainfall in the calibration dataset (Beven and Binley, 1992; Freni et al., 2009aec). With multiple modelling outputs, the analysis of the functions si, j may be only slightly informative and a more aggregated index may be useful. For this reason, a weighted average sensitivity was used for initial parameter evaluation: sj ¼
n si;j 1X n i¼1 max si;j
(2)
where maxðsi;j Þ is the maximum of the n sensitivities derived for the jth model parameter. Scarcely identifiable model parameters may act in two different ways: (i) they can generate small weighted sensitivity function values; or (ii) they can show an approximately linear dependence of sensitivity functions on the parameters. In the first case (the first non-identifiability criterion), the model parameter does not greatly affect the modelling output and thus calibration cannot really assess its value; in the latter case (the second non-identifiability criterion), the model-parameter variability does not clearly affect the modelling output and it can be considered a sort of underlying noise which increases the uncertainty transferred to the model output variable without providing relevant additional information to the model. The identification technique employed here was originally proposed and applied to WWTP models by Weijers and Vanrolleghem (1997) and is based on the elaboration of sensitivity matrices. The technique consists of two phases for the analysis of the two previously discussed causes of non-identifiability. In the preliminary phase, a sensitivity ranking of parameters is accomplished by averaging the sensitivity of different modelling outputs to the parameter (Eq. (2)). The preliminary analysis allows for the reduction of model parameters to the most sensitive ones, i.e., those characterised by model sensitivities higher than a user-defined threshold; model parameters with sensitivities lower than this threshold can be considered non-identifiable according to the first criterion defined above. Such subjective choice is used only for ranking the parameters and for simplifying the following step of the analysis by reducing the number of parameters to be investigated. An inappropriate choice of the threshold may lead to the following consequences: The use of a low threshold leads to the elimination of few parameters, thus increasing the complexity and the computational demands of the following part of the analysis;
The use of an high threshold leads to the initial elimination of an high Qmes number of parameters; in this way, the following phase of the analysis may lead to the identification of all remaining parameters without reaching a nonidentifiability condition. In this case, the analyst can run the analysis again reducing the threshold. The parameters saved in this first elimination phase are passed to the second phase of the identifiability analysis, which is based on elaborations of the Fisher Information Matrix (FIM): 1 ,ST FIM ¼ S,Qmes
(3)
where S is a matrix of n rows and m columns containing the sensitivity indices obtained by Eq. (1) and is the [n n] covariance matrix of the measurement noise. In the cases where measurement noise sources are uncorrelated, the Qmes matrix is diagonal and has a determinant equal to one. Considering a model with m parameters, the FIM is an [m m] matrix. The FIM summarises the importance of each model parameter with respect to the outputs (Dochain and Vanrolleghem, 2001). The FIM provides a lower bound for the parameter error-covariance matrix and its characteristics may then provide information on the shapes and dimensions of the model-confidence regions around the calibration values of the model parameters (So¨derstro¨m and Stoica, 1989). More specifically, as each column of the matrix represents a model parameter, the determinant and the condition number (i.e., the ratio between the highest and lowest matrix eigenvalues) of the FIM provides a reasonable measurement of the correlation of a set of model parameters (Weijers and Vanrolleghem, 1997). The FIM determinant D (the identifiability criterion) is a representation of the importance of the model parameters with respect to model outputs: a higher determinant indicates that the model outputs are more sensitive to the parameters. Conversely, the presence of one insensitive parameter causes a drastic reduction of the FIM determinant, to zero. As the D criterion is dependent on the magnitude of the parameters involved, this criterion was normalised (normD) according to Eq. (4): normD ¼ max D,kqk2
(4)
where kqk2 is the Euclidean norm of the parameter vector evaluated at the mean point of the parameter-variation range. Such normalisation acts as a scaling factor and allows for comparisons among subsets of the same size but with different model parameters. The condition number E (the identifiability criterion) is a representation of the shape of the confidence region (Weijers and Vanrolleghem, 1997; Checchi and Marsili-Libelli, 2005): a value near unity indicates that all parameters are equally important to the model; higher values are obtained in presence of a dominant or insensitive model parameter: modE ¼ min
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi! maxðEV½FIMÞ minðEV½FIMÞ
(5)
where maxðEV½FIMÞ and minðEV½FIMÞ are the maximum and minimum eigenvalues (EV) of the FIM, respectively. From the
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 e5 0
systems-engineering point of view, it is important to include in the parameter subset those parameters that maximise the D criterion and minimise the modE criterion. Both identification criteria have advantages and disadvantages (Freni et al., 2009a): the D criterion represents the size of the confidence region and thus the aggregated impact of parameters can be evaluated but the comparison between parameters in terms of identifiability may be difficult in complex models; the E criterion enables the easy comparison of the impact of each parameter on the model, but an objective approach for evaluating the number of identifiable parameters is missing (the maximum number of identifiable parameters can be detected by a rapid increase in the index value once a new parameter is added to an identifiable parameter subset). For this reason, in the present study, similarly to the method of Machado et al. (2009), a combination of the two criteria was considered. Hence, the ratio between the normD and the modE criteria (the DE criterion) is an interesting index to define subsets of identifiable parameters combining the advantages of both approaches: max D,kqk2 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi! DE ¼ maxðEV½FIMÞ min minðEV½FIMÞ
(6)
Another opportunity can be based on considerations similar to those that generated the modE criterion in an attempt to improve its objectivity. Such an aim can be achieved by comparing the maximum and minimum FIM eigenvalues at different steps of the identifiability process: gradE ¼ max
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi! max EV FIMpþ1 min EV FIMp , max EV FIMp min EV FIMpþ1
(7)
where p is the number of parameters in each step of the identifiability analysis, FIMp is the Fisher Information Matrix of dimension p p and the remaining variables as defined above. At each step of the identifiability analysis, gradE can reach a peak either if a highly sensitive parameter (the first fraction has a peak) or an insensitive parameter (the second fraction has a peak) is included. The number of identifiable parameters is identified by the absolute maximum of the gradE function. Practical identifiability approaches use the discussed criteria for ranking model parameter subsets to find the best combination that can be assessed according to the available data. The identification process is iterative and consists of adding one model parameter at a time to an initial identifiable subset that is usually selected among the most sensitive model parameters. In the subsequent iterative steps, all possible combinations are obtained by adding one parameter to the identifiable subset and evaluating the identification criteria. The combination providing the highest values of the identification criteria is retained and the iteration is repeated until the global maximum of the identification criteria is reached.
2.4.
Methodology application
According to the steps discussed in the previous section, an initial local sensitivity analysis was performed to identify the most sensitive model parameters among the fifty-one
43
characterising the integrated model: twenty-three for each urban drainage system and five for the RWB. Table 1 shows, for each sub-model and each parameter, the symbol, the measurement unit, the variation range and the weighted sensitivity index. Similarly to Beven and Binley (1992), parameter-variation ranges were taken as the intervals strictly including the calibrated values obtained by means of the seven available events. In the present study, sensitivity indices were evaluated by means of 1000 Monte Carlo (MC) simulation runs obtained by varying all parameters simultaneously and assuming a uniform distribution. Sensitivity indices were calculated for thirty modelling outputs for which data were available (Table 2) and neglected parameters were characterised by a sensitivity index lower than 0.015 (shown in grey in Table 1). After the first elimination phase based on weighted sensitivity ranking, the analysis of the Montelepre and Giardinello urban drainage systems was performed in three steps (SS, CSO and WWTP) separately and then the RWB. Such an approach was necessary to avoid the construction of FIMs in which model outputs and parameters are not linked by a cause-and-effect relationship. This approach, as further discussed below, also allowed us to understand the contribution of each data source to the identification process. Regarding the quantity and quality sub-modules, for sake of simplicity we do not considered a step-wise procedure aforementioned as for the three sub-models (i.e. first quantity and thereafter quality modules). For each urban drainage system, the analysis started from the initial subset consisting of the three most sensitive parameters. All the possible combinations of four parameters were considered by adding one model parameter to the initial identifiable set. The FIM was calculated for all the candidate parameter sets and the identifiability indicators were computed. The Qmes matrix was assumed to be diagonal and with determinant equal to one considering that measurement noise sources are uncorrelated. The best set was selected as the one providing the highest value of normD, DE and gradE or the minimum of modE. Therefore, the process was continued considering all possible combinations of parameters obtained by adding one additional parameter to the identifiable set; the parameter providing the best values of the identifiability indicators was added to the identifiable set and the analysis was continued adding a parameter at a time until one of the non-identifiability conditions were reached. The selection of an improper level of complexity in integrated modelling can have significant consequences on model output uncertainty, and non-identifiable parameters contribute to such uncertainty without providing any additional contributions in the representation of real processes. Once such parameters are known, they should be fixed to a default value (for instance the average of the expected variation range) thus neutralising the related uncertainty. To assess the impact of non-identifiable parameters on modelling uncertainty, the Generalised Uncertainty Likelihood Estimation (GLUE by Beven and Binley, 1992) was applied to the model in two scenarios: Considering the variation of all parameters (identifiable and non-identifiable) obtaining the total uncertainty related to the model
44
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 e5 0
Table 1 e Variation range of model parameters and average model sensitivities (parameters neglected after initial sensitivity analysis are greyed). Parameter
Symbol
Unit
Montelepre
Catchment linear channel constant Initial hydrological abstraction Catchment runoff coefficient Catchment linear reservoir constant Sewer linear reservoir constant Build-up rate in the Alley-Smith model Decay rate in the Alley-Smith model Wash-off coefficient in the Alley-Smith model Wash-off factor in the Alley-Smith model Sewer erosion factor Sewer suspended load linear reservoir constant Sewer bed load linear reservoir constant CSO first dilution factor CSO second dilution factor Max yield coefficient of heterotrophs BOD semi saturation constant Yield coefficient heterotrophic Temperature Max yield coefficient of autotrophs Oxygen half saturation constant Yield coefficient autotrophic Decay velocity of heterotrophs Decay velocity of autotrophs
l W0 F K1 K2 Accu Disp Arra Wh M Ksusp Kbed rd1 rd2 mmax,H Ks YH T mmax,A ko YN bH bA
min mm e min min kg/(ha*d) d1 mm-Whh(Wh-1) e kg min min e e h1 g/L e C h1 g/L e d-1 d-1
River bed roughness (GaucklereStrickler) Longitudinal dispersion coefficient De-oxygenation coefficient Sediment oxygen demand coefficient Re-aeration coefficient
ks Kdisp kd ksod ka
m1/3/s m2/s s1 s1 s1
Considering only the identifiable parameters and fixing the others to the average value of the ranges presented in Table 1. In this way, the unavoidable uncertainty can be, i.e. the uncertainty connected to the parameters that can be reliably calibrated. In both cases, the uncertainty bands were obtained by running 10,000 behavioural MC simulations were run assuming that variable model parameters were uniformly distributed in the ranges presented in Table 1. According to the classical application of GLUE, the NasheSutcliffe criterion
Table 2 e Monitored system variables available for the identifiability analysis with the number of data points available for each of them.
Montelepre
Giardinello
DO
System location
Q
TSS
BOD
COD
NH4
SS CSO WWTP SS CSO WWTP RWB
130 316
24 19 14 20 15 15
24 19 14 20 15 15 22
24 19
a
20 15 15
24 19 14 20 15 15
a
a
22
a
314 314 a
118
a
a
a Data not used in the present model application.
a a a a a
Giardinello
Dqj
sj
Dqj
sj
8e30 0.1e04 0.8e09 14e40 15e35 0.1e20 0.01e10 0.01e0.8 0.3e1 0.1e3 0.2e0.8 0.04e0.4 1.2e1.5 2e4 0.6e13.2 0.005e0.15 0.38e0.75 5e30 0.2e0.4 0.1e0.3 0.16e0.18 0.2e0.8 0.2e0.8 Dqj 10e70 1e500 1e100 1e100 1e1000
0.188 0.524 0.540 0.191 0.472 0.307 0.300 0.335 0.240 0.225 0.251 0.002 0.384 0.433 0.081 0.167 0.225 0.130 0.118 0.001 0.428 0.030 0.011
1e10 0.6e1 0.6e0.9 0.1e65 0.1e55 0.1e20 0.01e1 0.01e1 0.1e3.5 0.1e3 0.01e0.6 0.01e1 1.1e1.9 2e2.5 0.6e13.2 0.005e0.15 0.38e0.75 5e30 0.2e0.4 0.1e0.3 0.16e0.18 0.2e0.8 0.2e0.8 sj 0.566 0.001 0.047 0.351 0.894
0.221 0.598 0.462 0.197 0.474 0.284 0.225 0.050 0.437 0.341 0.217 0.004 0.013 0.441 0.003 0.029 0.032 0.014 0.042 0.002 0.226 0.002 0.012
(Nash and Sutcliffe, 1970) was used as likelihood measure and an acceptability threshold equal to zero for the selection of behavioural and non-behavioural simulation runs. The uncertainty bands were computed as the 5% and 95% percentiles of the likelihood distribution. For brevity’s sake, the application details of the uncertainty analysis were not reported in the present paper and they can be found in previous literature (Freni et al., 2009b,c, 2008b).
3.
Analysis of results
The results of the initial weighted sensitivity analysis are presented in Table 1: eleven parameters (all regarding water quality aspects) demonstrated sensitivity indices lower than the threshold and so were neglected in the following part of the study (being non-identifiable by the first non-identifiability criterion). They were mainly related to WWTP processes and to the Giardinello urban area. This fact could be due to several factors such as the lower quality of the Giardinello data, higher uncertainty in the identification of parameter values, or the lower relevance of the Giardinello catchment in determining the quality state of the RWB, thus reducing the related sensitivity indices. According to the initial analysis, six parameters provided higher weighted sensitivities and they were used as initial parameter subsets for the identifiability analysis. More
45
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 e5 0
specifically, the most sensitive parameters were the initial hydrological abstraction (W0), the catchment runoff coefficient (F) and the sewer linear-reservoir constant (K2) for both urban drainage systems. Starting from these parameters the identifiability analysis was carried out according to the procedure described above. All possible combinations of four parameters including the initial identifiable set were analysed selecting the one providing the best values of the identifiability indices. The process was continued considering sets with progressively increasing number of parameters, always adding the parameter that provided the best value of the indices to the previously identifiable set. The best parameter combinations obtained in all performed iterations (increasing the number of parameters in the set) along with computed identifiability indices are reported in Tables 3 and 4. The NormD and DE criteria showed similar results in assessing the same number of identifiable parameters (Fig. 4): eleven parameters for the Montelepre urban drainage system and nine for Giardinello. The DE criterion showed a flat area near the maximum making its assessment quite difficult. This was due to the rapid increase of modE that may have masked the increase of NormD; thus, even if the position of the maximum was preserved in the analysed case, this condition may have led to an incorrect estimation of the identifiable parameter set. The ModE criterion showed some limitations due to the fact that it was constantly growing with the increase in the dimensions of parameter subsets; The ModE criterion is characterised by a jump once the number of identifiable parameters is reached but it is hardly visible in Figures 4a and c and an objective criterion is not easily assessable thus making this criterion difficulty applicable by inexperienced analysts. The gradE index was consistent with the others in the determination of the identifiable number of parameters and it eliminated the subjectivity of the modE criterion. All criteria agree in the composition of the identifiable parameter subsets, which are presented in bold type in Tables 3 and 4. According to the simulation results the following conclusions may be drawn: The first identifiable parameters (i.e., W0, F, and K2) are all connected with water-quantity modules, demonstrating the greater importance of such parameters affecting both water quantity and water-quality modelling outputs; these parameters deeply influence the volume and the shape of
sewer hydrograph thus affecting the behaviour of all the downstream sub-models; this effect is also due to the higher availability of water quantity data with respect to the water quality ones; A group of seven parameters (mostly connected with water quantity sub-models) are identifiable in both urban areas demonstrating their importance in the integrated model; the water quality parameters in this group mainly affect the accumulation of pollutants in the sewer and on the catchment thus indicating that such process affects significantly water quality in all model sub-systems; Conversely, parameters related to water quality processes in the sewers are scarcely identifiable thus showing that they are not relevant or their impact cannot be separated by other water quality parameters according to the available field data; the second possibility is probably the most reliable because water quality at the end of the sewer pipe (where the monitoring station is located) is surely affected by two accumulation/wash-off processes (one taking part on the catchment and the other in the sewer pipe) that are not separable unless a specific campaign is carried out for monitoring water quality at the sewer inlets; Most of the WWTP parameters were non-identifiable (by the second non-identifiability criterion); this behaviour can be explained by their lower variability and by the lower number of affected modelling outputs; many model parameters interact in the same equations so that the variation of one of them may be compensated by the others. From a practical point of view, the previous comment should probably lead to a simplification of the WWTP submodel because it is too complex with respect to the available data; more interestingly, the analysis should take to a deeper field investigation of the WWTP by including additional intermediate monitoring stations in order to identify more parameters; The number of identifiable parameters in the Giardinello urban drainage system remained lower than in the Montelepre system, confirming the initial differences obtained in the preliminary sensitivity analysis; this difference may be related to the different dimensions and characteristics of the two urban areas (with different ratios between dry and wet-weather flows) thus taking to a different relevance of stormwater polluting processes. Giardinello is in fact
Table 3 e Best identifiable model parameter subsets for Montelepre urban drainage systems (SS, CSO and WWTP): the largest identifiable parameter set is indicated in italic; the parameter added at each analysis step is underlined. N
Parameters
normD
modE
DE
gradE
3 4 5 6 7 8 9 10 11 12 13
W0, F, K2 W0, F, K2, rd2 W0, F, K2, rd2, Accu W0, F, K2, rd2, Accu, l W0, F, K2, rd2, Accu, l, K1 W0, F, K2, rd2, Accu, l, K1, M W0, F, K2, rd2, Accu, l, K1, M, YN W0, F, K2, rd2, Accu, l, K1, M, YN, mmax,H W0, F, K2, rd2, Accu, l, K1, M, YN, mmax,H, Disp W0, F, K2, rd2, Accu, l, K1, M, YN, mmax,H, Disp, mmax,A W0, F, K2, rd2, Accu, l, K1, M, YN, mmax,H, Disp, mmax,A, T
5.99E þ 07 1.31E þ 11 2.3E þ 14 4.7E þ 15 4.3E þ 18 3.6E þ 19 3.80E þ 20 1.08E þ 21 2.18E þ 21 1.36E þ 21 1.77E þ 20
6.1054 9.76847 13.3731 18.2585 29.2693 36.9032 65.0466 137.291 327.847 1043.14 1421.82
9.81E þ 06 1.34E þ 10 1.70E þ 13 2.57E þ 14 1.47E þ 17 9.83E þ 17 5.83E þ 18 7.9E þ 18 6.7E þ 18 1.31E þ 18 1.24E þ 17
1.6 1.369 1.365 1.603 1.261 1.763 2.111 2.388 3.182 1.363 e
46
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 e5 0
Table 4 e Best identifiable model parameter subsets for Giardinello urban drainage systems (SS, CSO and WWTP): the largest identifiable parameter set is indicated in italic; the parameter added at each analysis step is underlined. N
Parameters
normD
3 4 5 6 7 8 9 10 11
W0, F, K2 W0, F, K2, rd2 W0, F, K2, rd2, K1 W0, F, K2, rd2, K1, Ksusp W0, F, K2, rd2, K1, Ksusp, YN W0, F, K2, rd2, K1, Ksusp, YN, Wh W0, F, K2, rd2, K1, Ksusp, YN, Wh, YH W0, F, K2, rd2, K1Ksusp,YN, Wh, YH, Accu W0, F, K2, rd2, K1, Ksusp, YN, Wh, YH, Accu, l
2.31E 3.99E 4.61E 8.23E 2.14E 1.05E 6.96E 1.17E 1.88E
characterised by lower dry-weather flows and higher polluting concentrations making the first flush phenomenon less evident than in the Montelepre catchment thus reducing the sensitivity of wet-weather related parameters and their identifiability; The NasheSutcliffe calibration efficiencies (Nash and Sutcliffe, 1970) were w0.85 in the Montelepre urban drainage system and lower than 0.6 in the Giardinello (Freni et al., 2010a, 2008a), thus demonstrating that less information can be derived from the available data;
25
09 12 15 17 20 22 22 22 21
b
2000 normD
1.23 1.25 1.38 1.42 1.62 1.76 1.90 1.28 e
5 DE gradE
20
4
15
1200
15
3
10
800
10
2
5
400
5
1
0
0
0
5
10
gradE
Log(DE)
1600
modE
Log(normD)
3.33E þ 08 4.69E þ 11 4.33E þ 14 5.58E þ 16 1.02E þ 19 3.11E þ 20 1.17E þ 21 1.03E þ 20 1.3E þ 19
20
0
0 0
15
5
10
15
Number of parameters [-]
Number of parameters [-]
25
150
d
normD
25
2.5 DE gradE
modE 120
20
2.0
15
90
15
1.5
10
60
10
1.0
5
30
5
0.5
0
0
0 0
5
10
Number of parameters [-]
15
gradE
Log(DE)
20
modE
Log(normD)
6.94 8.52 10.66 14.73 20.95 33.84 59.56 113.45 145.03
gradE
25
modE
c
DE
impact of additional data sources, the remaining parameters (three parameters for the RWB and fourteen non-identifiable in the previous stage for the two urban drainage systems) were passed through an additional identification step based on available RWB data. The analysis was intended to assess the identifiability of the RWB parameters and to verify if this additional data source would allow for the identification of additional parameters in the upstream submodels. As shown in Table 5, five parameters were assessed as identifiable using the additional data from the RWB. Despite the easily justifiable identification of the initial three RWB parameters (i.e., ks, ka, and ksod), the analysis of this additional data allowed for the identification of two more parameters that were not
Twenty parameters were assessed as identifiable by means of data collected in the SS, CSO and WWTP. To evaluate the
a
þ þ þ þ þ þ þ þ þ
modE
0.0 0
5
10
15
Number of parameters [-]
Fig. 4 e Identifiability criteria for the Montelepre urban drainage system (aeb) and Giardinello urban drainage system (ced).
47
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 e5 0
Table 5 e Additional identifiable model parameter subsets according to RWB data: the largest identifiable parameter set is indicated in italic; the parameter added at each analysis step is underlined. The identifiable parameters. Parameters Montelepre urban drainage system
21 22 23 24 25 26 27
Giardinello urban drainage system
1.31E þ 09
Initial condition: 11 identifiable parameters for Montelepre urban drainage system (Table 3) and 9 for Giardinello urban drainage system (Table 4) ks e e e e ks, ka e e ks, ka, ksod rd1 e ks, ka, ksod Accu ks, ka, ksod rd1 Accu ks, ka, ksod rd1, Wh ks, ka, ksod rd1, Wh Accu, Disp
identifiable by means of the SS, CSO and WWTP data (i.e., rd1 for Montelepre SS and Accu for Giardinello SS). The nonidentifiability of one parameter among ksod and kd was expectable as they both act on RWB BOD concentration once again showing that some processes needs specific monitoring campaigns to be assessable. The identification of additional parameters that were not initially identified in the urban drainage system analysis should stress the importance of interactions in the integrated system that cannot be analysed as the sum of separated compartments. The analysis of RWB identifiability criteria confirmed the good agreement of all adopted indices and the limitations due to the flatness of the DE and the subjective identification of jumps in modE (Fig. 5). This additional step in the identifiability analysis showed the impact that a coordinated monitoring campaign can have on the robustness of the model application. From a qualitative point of view, it would be expected that a larger dataset may satisfy more complex models; the identifiability analysis provides a quantitative response to this consideration by providing the number of parameters (i.e., indirectly providing the proper model complexity) that can be identified with the available dataset and it can suggest an appropriate increase of the number of model parameters effectively assessable when new data become available. Once the non-identifiable parameters were found, the application of uncertainty analysis allowed us to assess
a 16
160
12
120
DE
gradE
1.43E þ 08
1.23
8
80
4
40
þ þ þ þ þ þ þ
10 10 11 11 11 10 09
13.02 28.84 32.35 37.18 42.35 123.10 137.33
2.22E 2.80E 3.95E 7.23E 9.44E 7.41E 5.32E
b 12
3.0
10
2.5
8
2.0
6
1.5
4
1.0 DE
0.5
gradE
15
18
21
24
27
Number of parameters [-]
0 30
09 09 09 09 09 08 07
1.23 1.25 1.34 1.58 1.80 1.12
The results of the uncertainty analysis are dependent on the specific case study and on the subjective hypotheses adopted in the GLUE application. Nevertheless the reduction of uncertainty by fixing the non-identifiable parameters
2 0
þ þ þ þ þ þ þ
Discharge uncertainty bands were reduced by an average of 40% while the impact on water-quality variables was over 60%; The higher impact on water-quality uncertainty was connected with the higher number of non-identifiable waterquality parameters that introduced background noise into the uncertainty analysis; and These reductions were obtained without losing the validity of the modelling hypotheses, as over 90% of the data points remained within the uncertainty bands.
normD modE
2.89E 8.08E 1.28E 2.69E 4.00E 9.12E 7.31E
9.41
the impact of these parameters. The uncertainty bands obtained by varying all the model parameters (i.e., identifiable and non-identifiable parameters) according to the GLUE are displayed in Fig. 6aec, while Fig. 6def shows the uncertainty bands obtained by varying only the twenty-five identifiable parameters (Table 5) and fixing the others to the averages of their initial variation ranges (Table 1). A comparison of the uncertainty bands in Fig. 6 shows that the uncertainty-band width was significantly reduced by neglecting non-identifiable parameters; specifically, the following can be noted:
modE
Log (normD)
modE
RWB
Log(DE)
20
normD
0 20
22
24
26
Number of parameters [-]
Fig. 5 e Identifiability criteria for the RWB.
0.0 28
gradE
N
48
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 e5 0
Fig. 6 e RWB 5th percentile and 95th percentile in terms of discharge, BOD concentration and DO concentration for the total uncertainty [(a), (b), (c)] and for the unavoidable uncertainty [(d), (e), (f)].
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 e5 0
demonstrates the importance of identifiability analysis in the application of complex environmental models.
4.
49
simplification and results reliability. Further research may involve the effect of data availability with respect to parameter identification and the improvements provided by the introduction of new measuring stations in the system.
Conclusions
The present study applied a parameter identifiability analysis to a complex integrated urban drainage model. We proposed the use of identifiability analysis as a tool for assessing the appropriate model complexity to employ for a specific application. In the process, several published identifiability criteria were applied and a new one was proposed for integrating the simplicity of the indices based on FIM eigenvalues and the objectivity of these based on the FIM determinant. The results led to several interesting observations: The normD and DE criteria were unambiguous in the definition of identifiable parameters but DE was characterised by flatness near the maximum making the assessment of the number of identifiable parameter quite difficult; The modE criterion showed some limitations in the definition of identifiable parameters due to its subjectivity; in the presented applications, modE was always consistent with the criteria based on the FIM determinant but inexperienced analysts may misinterpret secondary modE jumps as the consequence of the introduction of a non-identifiable parameter in the analysis; and The gradE criterion solved such subjectivity problems because the number of identifiable parameters is given by the absolute maximum of the function and it maintained the simplicity of identifiability criteria based on eigenvalues estimation. The analysis showed some critical points in integrated urban drainage modelling, such as the interaction between water quality processes on the catchment and in the sewer, that can prevent the identifiability of some of the related parameters. Similar cases may be found the WWTPs, considering the different processes affecting pollutants concentration, or in the RWB, considering, as an example, sediment oxygen demand and the de-oxygenation coefficient. These identifiability issues may be solved either by simplifying the model or by carrying out specific field campaigns including intermediate monitoring stations. Uncertainty analysis carried out according to the GLUE methodology confirmed the effectiveness of the identifiability analysis in selecting the correct model complexity. Indeed, a reduction of the uncertainty in terms of uncertainty bandwidth was shown by fixing the non-identifiable model parameters. As a general conclusion, practical identifiability can be used for guiding the analyst in the selection of the right modelling detail level for a specific application and it is adequately flexible to reapply each time new data sources become available, allowing for modular model complexity adaptable to data availability, minimising “avoidable uncertainty” (i.e., the uncertainty due to the unnecessary complexity of the applied models). The results obtained herein are obviously dependent on the specific case study employed here. Considerations of the advantages provided by identifiability analysis may be generalised, especially with respect to integrated modelling
Acknowledgements Authors wish to thank Mrs R. D’Addelfio and Dr. A. P. Lanza for their valuable assistance during field work. The authors would like also to thank the Editor and the two anonymous reviewers for very helpful and constructive comments that resulted in a much improved manuscript.
references
Alley, W.M., Smith, P.E., 1981. Estimation of accumulation parameters for urban runoff quality modelling. Water Resources Research 17 (6), 1657e1664. APHA., 1995. Standard Methods for Examination of Water and Wastewater. APHA, AWWA and WPCF, Washington DC, USA. Belsley, D.A., 1991. Conditioning Diagnostics e Collinearity and Weak Data in Regression. Wiley, New York. Beven, K.J., Binley, A.M., 1992. The future of distributed models model calibration and uncertainty prediction. Hydrological Processes 6 (3), 279e298. Brown, L.C., Barnwell, T.O., 1987. The Enhanced Stream Water Quality Models QUAL2E and QUAL2E-UNCAS: Documentation and User Manual. USEPA/6003e87/007. USEPA, USA. Brun, R., Kuhni, M., Siegrist, H., Gujer, W., Reichert, P., 2002. Practical identifiability of ASM2d parameters and systematic selection and tuning of parameter subsets. Water Research 36 (16), 4113e4127. Brun, R., Reichert, P., Ku¨nsch, H.R., 2001. Practical identifiability analysis of large environmental simulation models. Water Resources Research 37 (4), 1015e1030. Campolongo, F., Cariboni, J., Saltelli, A., 2007. An effective screening design for sensitivity analysis of large models. Environmental Modelling & Software 22 (10), 1509e1518. 2007. Candela, A., Freni, G., Mannina, G., Viviani, G., 2009. Quantification of diffuse and concentrated pollutant loads at the watershed-scale: an Italian case study. Water Science & Techology 59 (11), 2125e2135. Chapra, S.C., 1997. Surface Water - Quality Modelling. McGrawHill Science/Engineering/Math. Checchi, N., Marsili-Libelli, S., 2005. Reliability of parameter estimation in respirometric models. Water Research 39 (15), 3686e3696. De Pauw, D.J.W. (2005). Optimal Experimental Design for Calibration of Bio-process Models: A Validated Software Toolbox. PhD thesis in Applied Biological Sciences, BIOMATH, University of Gent. Available from:
. De Pauw, D.J.W., Sin, G., Insel, G., Van Hulle, S.W.H., Vandenberghe, V., Vanrolleghem, P., 2004. Discussion of: assessing parameter identifiability of activated sludge model number 1. Journal of Environmental Engineering 130 (1), 111e112. Dochain, D., Vanrolleghem, P.A., 2001. Dynamical Modelling and Estimation in Wastewater Treatment Processes. IWA Publishing, London. European Commission, (2000). Directive 2000/60/EC of the European Parliament and of the Council establishing
50
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 e5 0
a framework for the Community action in the field of water policy. Freni, G., Mannina, G., and Viviani, G. (2008a). Catchment-scale modelling approach for a holistic urban water quality management, Proc. 11 ICUD conference, Edinburgh (Scotland e UK), 31 August e 5 September 2008. Freni, G., Mannina, G., Viviani, G., 2008b. Uncertainty in urban stormwater quality modelling: the effect of acceptability threshold in the GLUE methodology. Water Research 42 (8e9), 2061e2072. Freni, G., Mannina, G., Viviani, G., 2009a. Identifiability analysis for receiving water body quality modelling. Environmental Modelling & Software 24 (1), 54e62. Freni, G., Mannina, G., Viviani, G., 2009b. Uncertainty assessment of an integrated urban drainage model. Journal of Hydrology 373 (3e4), 392e404. Freni, G., Mannina, G., Viviani, G., 2009c. Urban runoff modelling uncertainty: comparison among Bayesian and pseudo-Bayesian methods. Environmental Modelling & Software 24 (9), 1100e1111. Freni, G., Mannina, G., Viviani, G., 2010a. Urban water quality modelling: a parsimonious holistic approach for a complex real case study. Water Science & Technology 61 (2), 521e536. Freni, G., Mannina, G., Viviani, G., 2010b. Urban stormwater quality management: centralized versus source control. Journal of Water Resources Planning and Management e Asce 136 (2), 268e278. Gatelli, D., Kucherenko, S., Ratto, M., Tarantola, S., 2009. Calculating first-order sensitivity measures: a benchmark of some recent methodologies. Reliability Engineering & System Safety 94 (4), 1212e1219. 2009. Holmberg, A., 1982. On the practical identifiability of microbial growth models incorporating MichaeliseMenten type nonlinearities. Mathematical Biosciences 62 (1), 23e43. Jakeman, A.J., Hornberger, G.M., 1993. How much complexity is warranted in a rainfall-runoff model? Water Resource Research 29 (8), 2637e2649. Jewell, T.K., Adrian, D.D., 1978. SWMM storm water pollutant washoff function. Journal of the Environmental Engineering Division 104 (5), 1036e1040. Kuczera, G., Parent, E., 1998. Monte Carlo assessment of parameter inference in catchments models: the Metropolis algorithm. Journal of Hydrology 211 (1e4), 69e85. Machado, V.C., Tapia, G., Gabriel, D., Lafuente, J., Baeza, J.A., 2009. Systematic identifiability study based on the Fisher Information Matrix for reducing the number of parameters calibration of an activated sludge model. Environmental Modelling and Software 24 (11), 1274e1284. Malve, O., Laine, M., Haario, H., Kirkkala, T., Sarvala, J., 2007. Bayesian modelling of algal mass occurrences using adaptive MCMC methods with a lake water quality model. Environmental Modelling and Software 22 (7), 966e977. Mannina, G., Viviani, G., 2010a. An urban drainage stormwater quality model: model development and uncertainty quantification. Journal of Hydrology 381 (3e4), 248e265.
Mannina, G., Viviani, G., 2010b. A parsimonious dynamic model for river water quality assessment. Water Science & Techology 61 (3), 607e618. Mannina, G., Viviani, G., 2010c. A hydrodynamic water quality model for propagation of pollutants in rivers. Water Science & Technology 62 (2), 288e299. Mannina, G., Viviani, G., 2009. Separate and combined sewer systems: a long-term modelling approach. Water Science & Technology 60 (3), 555e565. Mannina, G. (2005). Integrated urban drainage modelling with uncertainty for stormwater pollution management. PhD thesis, Universita` di Catania, (Italy). Mannina, G., Freni, G., Viviani, G., 2004. Modelling the integrated urban drainage systems. In: Bertrand-Krajewski, L., Almeida, M., Matos, J., Abdul-Talib, S. (Eds.), Sewer Networks and Processes within Urban Water Systems (WEMSno.). IWA Publishing, London, UK, pp. 3e12. Marsili-Libelli, S., Giusti, E., 2008. Water quality modelling for small river basins. Environmental Modelling and Software 23 (4), 451e463. Metcalf and Eddy, Inc, 2003. Wastewater Engineering: Treatment and Reuse, fourth ed. McGraw Hill, New York. Morris, M.D., 1991. Factorial sampling plans for preliminary computational experiments. Technometrics 33, 161e174. Nash, J.E., Sutcliffe, J.V., 1970. River flow forecasting through conceptual models. Journal of Hydrology 10 (3), 282e290. Parchure, T.M., Mehta, A.J., 1985. Erosion of soft cohesive sediment deposits. Journal of Hydrology 111 (10), 1308e1326. Rauch, W., Bertrand-Krajewski, J.-L., Krebs, P., Mark, O., Schilling, W., Schuetze, M., Vanrolleghem, P.A., 2002. Deterministic modelling of integrated urban drainage systems. Water Science & Technology 45 (3), 81e94. Reichert, P., Vanrolleghem, P.A., 2001. Identifiability and uncertainty analysis of the River water quality model No. 1 (RWQM1). Water Science & Technology 43 (7), 329e338. Saltelli, A., Ratto, M., Tarantola, S., Campolongo, F., 2006. Sensitivity analysis practices: strategies for model-based inference. Reliability Engineering & System Safety 91 (10e11), 1109e1125. Saltelli, A., Campolongo, F., Cariboni, A., 2009. Screening important inputs in models with strong interaction properties. Reliability Engineering & System Safety 94 (7), 1149e1155. 2009. So¨derstro¨m, T., Stoica, P., 1989. System Identification. PrenticeHall, Englewood Cliffs: New Jersey. Taka´cs, I., Patry, G.G., Nolasco, D., 1991. A dynamic model of the clarification-thickening process. Water Resource 25 (10), 1263e1271. Wagener, T., Kollat, J., 2007. Numerical and visual evaluation of hydrological and environmental models using the Monte Carlo analysis toolbox. Environmental Modelling and Software 22 (7), 1021e1033. Weijers, S.R., Vanrolleghem, P.A., 1997. A procedure for selecting best identifiable parameters in calibrating activated sludge model no. 1 to full-scale plant data. Water Science & Technology 36 (5), 69e79.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 e6 2
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
A Bayesian changepointethreshold model to examine the effect of TMDL implementation on the flowenitrogen concentration relationship in the Neuse River basin Ibrahim Alameddine*, Song S. Qian, Kenneth H. Reckhow Nicholas School of the Environment, Duke University, Durham, NC 27708, USA
article info
abstract
Article history:
In-stream nutrient concentrations are well known to exhibit a strong relationship with
Received 29 April 2010
river flow. The use of flow measurements to predict nutrient concentrations and subse-
Received in revised form
quently nutrient loads is common in water quality modeling. Nevertheless, most adopted
28 July 2010
models assume that the relationship between flow and concentration is fixed across time
Accepted 3 August 2010
as well as across different flow regimes. In this study, we developed a Bayesian change-
Available online 11 August 2010
pointethreshold model that relaxes these constraints and allows for the identification and quantification of any changes in the underlying floweconcentration relationship across
Keywords:
time. The results from our study support the occurrence of a changepoint in time around
TMDL
the year 1999, which coincided with the period of implementing nitrogen control measures
Changepoint
as part of the TMDL program developed for the Neuse Estuary in North Carolina. The
Threshold
occurrence of the changepoint challenges the underlying assumption of temporal invari-
Floweconcentration
ance in the floweconcentrations relationship. The model results also point towards
Neuse River
a transition in the river nitrogen delivery system from a point source dominated loading
Bayesian model
system towards a more complicated nonlinear system, where non-point source nutrient
Water quality
delivery plays a major role. Moreover, we use the developed model to assess the effec-
Nitrogen
tiveness of the nitrogen reduction measures in achieving a 30% drop in loading. The results
Load reduction
indicate that while there is a strong evidence of a load reduction, there still remains a high level of uncertainty associated with the mean nitrogen load reduction. We show that the level of uncertainty around the estimated load reduction is not random but is flow related. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Anthropogenic nitrogen reaching rivers, lakes, estuaries, and coastal areas has been linked to eutrophication, acidification, adverse human health effects, the disruption of ecosystem functions, as well as the lowering of biodiversity in affected water bodies (Kelly, 2008 and references therein). High profile events such as the dramatic fish kills in the Neuse River and Estuary in mid 1980 and early 1990s as well as the
development of extensive hypoxic zones in the Gulf of Mexico have been linked to excessive nitrogen release and delivery (Turner and Rabalais, 1994; Paerl et al., 1995; Paerl, 1997; Alexander et al., 2000, 2008; Scavia et al., 2003; Stow and Borsuk, 2003; Borsuk et al., 2004). Such events have stimulated an impetus towards the implementation of aggressive management and mitigation measures to limit the amount of nitrogen reaching the aquatic environment. While some successes have been made in some water bodies, water
* Corresponding author. Tel.: þ1 919 613 8054; fax: þ1 919 681 5740. E-mail address:
[email protected] (I. Alameddine). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.003
52
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 e6 2
Nomenclature log N Unif Q w b X a bj,k bTemp s d i
natural logarithm normal distribution uniform distribution daily mean flow (m3/s) distributed the mean of parameter X the model intercept slope on the river flow (in natural log scale) slope on water temperature measured in C standard deviation a variable that is added to the slope on river flow to account for the occurrence of a flow threshold index of observation
quality impairment from excessive nitrogen loading continues to be a pressing issue on the national as well as international levels. Currently nutrient impairment ranks fourth on the national impairment list with 6826 water bodies listed as impaired due to excessive nutrient loading (out of a total of 75,677 impairment causes). Of these around 17% have nitrogen explicitly listed as the cause of impairment (Environmental Protection Agency, 2009). The Neuse River and its associated estuary in North Carolina have experienced all the symptoms of eutrophication with extensive algal blooms, fish kills, hypoxia and anoxia that have captured the public attention in the 1980s and 1990s (Paerl et al., 1998). This led to designating the Neuse as a nutrient sensitive water and prompted its listing on the 303 (d) list with nitrogen identified as the main culprit behind eutrophication (Paerl et al., 1995; Stow et al., 2003). In 1998, the USEPA settled a lawsuit brought by the Neuse River Foundation which required North Carolina to establish a Total Maximum Daily Load (TMDL) for nitrogen reaching the estuary. The TMDL was approved by USEPA on August 26, 1999. Meanwhile, the State of North Carolina, through the North Carolina Division of Water Quality (NCDWQ), adopted in 1997 a set of rules that aimed at reducing the amount of nitrogen delivered to the Neuse River Estuary by 30% based on 1991e1995 loads. Despite almost a decade of post-TMDL monitoring, there has been no consensus on whether the TMDL has achieved its stated goal and if the implemented management measures have been successful (Deamer, 2009). To better understand the nitrogen dynamics and the effectiveness of the TMDL program in the Neuse over time, we made use of daily flow measurements to estimate nitrogen concentrations and nitrogen loading rates in the Neuse. The use of flow measurements to predict nutrient concentrations (and thus load) is common in water quality modeling given that in-stream nutrient concentrations have been observed to exhibit a relationship with river/stream flow (Johnson, 1979; Reckhow and Stow, 1990; Stow and Borsuk, 2003). The development of floweconcentration (as well as flow-load) models are often used to draw upon the large databases of daily flow measurements in order to augment infrequent water quality sampling measurements.
j
k
n q Min Max 3
logical operator that returns a value of 1 for pre changepoint observations and a value of 2 for post changepoint observations, where a changepoint is an instance in time when the system changes its behavior logical operator that returns a value of 1 for flow values below the flood threshold and a value of 2 for flow values above the flood threshold number of observations set of model parameters returns the minimum value from a vector of values returns the maximum value from a vector of values error term (assumed to be Gaussian white noise)
The use of regression-based empirical methods to predict daily nutrient loading through the use of daily averaged river flow measurements is one of the more commonly used approaches to determine nutrient concentrations/loads (Cohn et al., 1992; Green and Haggard, 2001; Hooper et al., 2001; Haggard et al., 2003; Runkel et al., 2004; Ide et al., 2007). This approach is based primarily on the work of Cohn et al. (1992) who developed the “rating curve” method that involves a log-linear multivariate regression model linking flow to concentration and load. While refinements have been added to the original “rating curve” method, most of the adopted models assume that the relationship between flow and concentration is fixed over time as well as across the range of river flows. The implementation of environmental management measures at a river basin scale can often result in changes to the underlying relationship linking flow to concentration measurements, and ultimately affect load estimates. With the implementation of basin-scale water quality management plans in different river systems, it is becoming increasingly imperative to evaluate the effects that such basin-scale management plans (like the TMDL program) has on these systems. So far, however, there has been little chance to conduct such as assessment due to the difficulty of finding a river system with both a long monitoring record and that has had TMDL mitigation measures put in place. The Neuse River presents a unique opportunity to study these changes and demonstrate their impacts due to the presence of a monitoring program that stretched for over 30 years, during which a TMDL program has been enforced. The objective of this paper is to assess the dynamics of the relationship of nitrogen concentration and flow between 1979 and 2008 in order to determine whether the relationship has been time invariant or if it has experienced a major changepoint across time. The occurrence of a system changepoint at a specific point in time often signifies an abrupt change in the way a system operates or behaves. These system changes may involve changes to the model parameters, the underlying model structure, or changes to both. We use Total Oxidized Nitrogen (TON) (which is the sum of nitrate (NO 3 ) and nitrite (NO 2 )) concentrations in the Neuse, given its long historical
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 e6 2
53
Fig. 1 e Neuse River Basin showing the major urban areas, the river network, as well as the town of New Bern, where the Neuse River opens to form the Neuse Estuary before discharging in Pamlico Sound. The image in the small frame shows the Neuse River Basin within the context of the State of North Carolina.
data. We then check the identified temporal changepoint against the backdrop of environmental management measures undertaken in the Neuse as well as a set of natural major shocks to the system in order to gain a better understanding of nitrogen delivery. Additionally, we explore whether a flow threshold is discernable, while exploring the implications of its presence/absence. Finally, we use our empirical model to evaluate whether the implemented TMDL mitigation measures have been successful in reducing nitrogen loading by 30%, while taking into account associated model and parameter uncertainties. To our knowledge this is the first attempt to study nutrient delivery in the form of TON in the Neuse River system that spans three decades of data (1979e2008) during which major natural as well as watershedscale management measures have occurred.
2.
Materials and methods
2.1.
Study area and data
The Neuse River at the Fort Barnwell station drains an area of 10,100 km2 (United States Geological Survey, 2009). The basin has a diverse landuse/landcover. Just east of its head waters, an urbanized area e that includes the cities of Raleigh, Durham, and Cary e dominates the basin. Intensive agricultural areas (row crops and Concentrated Animal Feed Operations (CAFOs)) become more prominent towards the lower portions of the Neuse basin, where the river traverses the North Carolina coastal plain (Fig. 1). The major point-source nitrogen emitters in the basin are the 20 major wastewater treatment plants that service the cities and townships in the 19 counties that fall within the Neuse basin. Moreover,
continued landuse changes in the basin have resulted in an increase in the relative importance of non-point nutrient sources to the overall Neuse nutrient budget. TON concentrations and water temperature data between 1979 and 2008 were primarily collected from the Environmental Protection Agency’s (EPA’s) STOrage and RETrieval (STORET) service that publishes ambient monitoring data collected by NCDWQ at Fort Barnwell (Lat ¼ 35.3125 N; Lon ¼ 77.3022 W). TON sampling frequency at Fort Barnwell has changed significantly over the years. Sampling was conducted on a monthly basis between 1979 and 1995, while the years 1996 through 2002 saw a major increase in the sampling frequency (>200 samples per year) before the sampling effort was reduced to weekly post 2002. Water temperature values were included in order to capture the seasonality in TON concentrations that are known to vary by season as a result of changes in the biological and physio-chemical characteristics in the river and its contributing watershed (Malone et al., 1996). Flow data for the same period were collected for the Fort Barnwell Station through the United States Geological Survey’s (USGS’s) National Water Information System (NWIS). The NWIS database also reported data on TON concentrations and ambient river water temperature. These were used to augment data from EPA’s STORET. Missing values for flow measurements at Fort Barnwell were estimated through the regression model that was developed by Stow and Borsuk (2003). Their model predicts flow at Fort Barnwell from two upstream USGS operated gauging stations. In the same context, missing water temperature values were imputed using a linear regression model linking water temperature to ambient air temperature, with a constraint placed to ensure that imputed water temperatures do not drop below freezing (Equation (1)). Both empirical models used to
54
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 e6 2
impute the data resulted in high coefficient of determination (R2 was 0.97 and 0.85 respectively). As such, the uncertainties associated with both models were not included within our modeling framework.
Western and Kleykamp, 2004; Park, 2006); nevertheless far too little attention to our knowledge has been paid to their use in water quality models. The incorporation of a changepoint in a water quality model is relatively easy when the changepoint
Water Temp ð CÞ ¼ 3:26 þ 0:89 Air Temp ð CÞ þ 3; 3wN 0; s2 ; s ¼ 3:02 C d CÞ 0 C; if Water Tempð i 8 d CÞ ¼ Water Tempð CÞ ; if Water Tempð CÞ > 0; > < Water Tempð i i1 i1 d CÞ ¼ Water Tempð CÞ ; if Water Tempð CÞ > 0; then : Water Tempð i iþ1 iþ1 > : d CÞ ¼ MinðWater Tempð CÞÞ else Water Tempð i Fort Barnwell was selected as the point of interest for this study given that 1) it is the most downstream long-term USGS maintained gauge on the Neuse (w20 km upstream from the Neuse Estuary), 2) it drains around 85% of the Neuse watershed, and 3) it has been used in recent Neuse River water quality studies (Stow et al., 2003; Borsuk et al., 2004).
2.2.
Model development
Given the changes that have occurred in the Neuse River Basin we developed a changepointethreshold Bayesian model that is capable of predicting TON concentrations from river flow and water temperature. To avoid confusion, we point out that we will be using the term changepoint to signify a change occurring over time, while the use of the term threshold will be reserved to indicate a change over the range of river flows. Even though the inclusion of a changepoint along with a threshold adds to the complexity of the model, it allows for a better understanding of
is known beforehand. This is usually done through the inclusion of a dummy variable or through fitting separate models for each period. Yet in most cases, we are seldom sure of the occurrence or the exact timing of changepoints. Thus, what we are often more interested in is: 1) to be able to recognize whether a changepoint actually occurs or not, 2) to have a rigorous method to estimate its timing, and 3) to have an associated probability distribution that accounts for the uncertainty in its time of occurrence. The changepointethreshold Bayesian model that we propose is capable of predicting TON concentrations (TON) from flow measurements (Q) and water temperatures (Temp) without constraining the relationship linking flow to TON concentrations to be fixed over time or over the range of observed river flows (Equation (2)). We assume that the random variable TONi¼1,.,TONi¼n, like most water quality concentration variables, follows a lognormal distribution (Ott, 1995). As such we can describe the changepointethreshold model as:
logðTONi Þ ¼ N aj½i þ bj½i;k½i ðlogðQi Þ thresholdÞ þ bTemp Temp; s2j½i;k½i 1; if yeari changepoint < 0 j½i ¼ 2; if yeari changepoint 0 1; if logðQ i Þ threshold < 0 k½i ¼ 2; if logðQ i Þ threshold 0
the system by removing the constraining assumption of invariance both in time and across flows. Accounting for a threshold in the relationship governing nutrient concentration and flow in the Neuse is not new. It was previously suggested by Borsuk et al. (2004) in their regression-based estuarine model for the Neuse Estuary. However, their model e like most other water quality empirical models such as the LOAD ESTimator (LOADEST) (Runkel et al., 2004), and the SPAtially Referenced Regressions On Watershed attributes (SPARROW) (Smith et al., 1997) models e assumes that the floweconcentration relationship is constant over time. In many cases, assuming that a system is statistic over time can be a reasonable assumption; yet in many cases failing to properly account for the temporal dynamics in the system can lead to incorrect parameter estimation, inflated errors, as well as to poor predictive power. This is particularly true when the system under study experiences an external intervention such as a policy change (Congdon, 2006). The incorporation of temporal changepoints in statistical models has been well established in the social sciences (e.g.
(1)
(2)
where aj is the model intercept. It corresponds to the log(TON) concentration when log(Q) is zero in the event that no flow threshold is identified, and to the log(TON) concentration when flow is equal to the mean flow threshold in the event a flow threshold is recognized. bj,k correspond to the slope on log(Q) with different values assigned depending on whether the system has passed the changepoint and/or threshold. Note that in order to improve parameter identifiability when running the Markov-chain Monte Carlo (MCMC) procedures, b2,2 was redefined as b2,1 þ d. bTemp represents the slope on water temperature. We opted to maintain a common slope on water temperature, as there was no reason to suspect that the seasonal patterns captured by temperature values changed over time or flow. Moreover, the model allows for different normally distributed error terms (sj,k) for pre- and post-changepoint as well as for pre- and post-flow threshold. This allows for the possibility that the model may perform better under certain ranges of flow and/or over certain periods of time. The decision to include a changepoint within the model framework permits the model to incorporate any change in the
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 e6 2
floweconcentration relationship over time, while also allowing for the pooling of data into two main groups, namely pre- and post-changepoint. This partial pooling of the data offers a clear advantage over completely combining all measurements into one system or treating each year separately (Gelman et al., 2004; Gelman and Hill, 2007; Qian et al., 2010). For this study the temporal partial-pooling also helps to overcome problems associated with the inter-annual variations in the Neuse hydrograph, whereby some years are wet while others are relatively dry. This ensures that the TON sampling events covered the entire flow range in the Neuse for both pre- and post-changepoint groups. This reduces model bias towards low flow conditions as described by Ide et al. (2007), especially during low sampling frequency periods in the Neuse (1979e1995 and 2003e2008) (Fig. 2). The incorporation of a flow threshold within the model structure allows for the detection of an environmental threshold along the range of observed daily mean flows. The presence of the threshold signifies a change in the manner that TON concentrations respond to flow measurements. For this study the inclusion of the threshold allows for the accommodation of a piecewise-linear relationship linking river flows to TON concentrations. The adopted piecewise-linear model in this framework constrained the two linear pieces to intersect at the identified flow threshold. For a more detailed discussion of piecewise-linear regression models and their uses, refer to Qian and Richardson (1997). Notice that in the event that the data do not support the presence of a threshold and/or of a changepoint, the model will simply collapse back into a simplified floweconcentration linear regression model. Under a Bayesian framework, prior distributions have to be specified on the model parameters q ¼ ðaj ; bj;k ; bTemp ;
si;j ; d; Threshold; Change pointÞ: We defined diffuse prior distributions on all model parameters. A constrained weak prior was specified on the changepoint distribution. This was done by defining a discrete uniform distribution over the range of sampling years i.e. changepointwUnifðMinðYearÞ; MaxðYearÞÞ. All years in that range were given equal prior probabilities of being identified as potential changepoints. The assignment of a weak prior on changepoint lets “the data speak for themselves” (Gelman et al., 2004), and thus assures that any posterior inference on the timing of the changepoint is unaffected by information external to the current data. Similarly the prior on the flow threshold was chosen to be a continuous uniform distribution that is bounded to the range of observed river flows i.e. ThresholdwUnifðMinðlogðQÞÞ; MaxðlogðQÞÞÞ. Meanwhile, the priors on si;j were constrained to be nonnegative and to be upper bound to 55 mg/L i.e. si,j w Unif(0,4), which is more than one order of magnitude larger than the maximum observed TON concentration observed in the past 30 years (max TON ¼ 3.235 mg/L reported on 10-24-2000). Finally, the priors on the model intercepts (aj) and slopes (bj,k, bTemp, d) were given vague normal distribution of the form N(0,103). The motivation behind our choice to adopt vague and non-informative priors for model parameters is primarily due to our desire to reduce subjectivity and let the data drive our inference. The specification and use of weak and non-informative priors in Bayesian analysis is common and is discussed in more detail by Gelman et al. (2004), Gelman (2006), and Van Dongen (2006). We used a Markov-chain Monte Carlo (MCMC) procedure to determine the posterior distributions of these parameters using the Bayesian software package WinBUGS (Lunn et al., 2000). Six chains were initiated at different arbitrary initial values for the parameters and were monitored. The generated posterior distributions were all based on 10,000 MCMC samples. Convergence was assured by monitoring that the b for each parameter was potential scale reduction factor, R, equal to 1.0 (Gelman et al., 2004; Gelman and Hill, 2007). The computer code for our proposed Bayesian model is included in the online Supplementary Material.
2.3.
Fig. 2 e Boxplot of TON concentrations in natural logarithm scale from 1979 till 2008 for the Fort Barnwell station. Grey line indicates a locally-weighted polynomial regression LOWESS curve that traces TON concentrations across time. Sampling frequency for each year is also illustrated in the inverted histogram.
55
Calculating TON load reductions
The posterior predictive distribution for TON concentrations was used to predict the probability that the stipulated 30% TON load reduction has been achieved in the Neuse River basin. This was carried out by first drawing 10,000 simulated daily flow measurements from the 30 year daily hydrograph for Fort Barnwell (Fig. 3a). For each flow value, 2500 corresponding TON concentrations were sampled for both pre- and post-changepoint periods using the posterior predictive distribution of log (TON) (Fig. 3b). The generated pre- and post-changepoint TON concentrations were then compared in order to determine the distribution of the percent reduction in TON loading given a specific flow value (Fig. 3c). The process was repeated over the 10,000 simulated daily flow values in order to construct the posterior predictive distribution of TON percent load reduction over all flow values. Note that since we are using the same simulated flow measurements for both time periods, this ensures that a 30% reduction in concentration translates directly into a 30% reduction in TON load. This methodology ensures that both periods cover the whole range of daily flow
56
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 e6 2
Fig. 3 e The adopted simulation methodology for estimating the probability distribution for the % Total Oxidized Nitrogen (TON) load reduction given flow. In panel (a), a daily flow value is drawn from the Fort Barnwell hydrograph. In this example the drawn flow corresponds to 231.48 m3/s. Panel (b) shows the posterior predictive probability distribution of TON for both pre- and post-1999 that correspond to a flow event of 231.48 m3/s. Panel (c) presents the histogram of the % TON reduction ð1003ð1LðPost 1999 TON=Pre 1999 TONÞÞÞ given the pre-1999 and post-1999 distributions in panel (b).
value. Simulation procedures for both concentration as well as load reductions were conducted using the rv package (Kerman and Gelman, 2007) in the statistical software program R (R Development Core Team, 2009).
3.
Results
Model results indicate that when it comes to TON concentrations in the Neuse, the system has experienced a regime shift in 1999 with respect to the floweTON concentration relationship (Table 1). The posterior distribution around the changepoint sharply peaks at the year 1999 and has almost no associated uncertainty, indicating strong evidence that compels us to conclude that a regime change did indeed occur
in 1999. The identified changepoint coincides with the implementation of one of the key mitigation measures that were stipulated as part of the TMDL action plan for the Neuse, which saw the major point-source discharger, namely the Neuse River Wastewater Treatment Plant for the city of Raleigh (design capacity of 227,125 m3/day; currently treating 138,811 m3/day) reduce nitrogen levels in its discharge by 55% from the 1995 levels. Moreover, the 1999 changepoint also coincides with an active hurricane season in North Carolina that saw the landfall of two major hurricanes, namely Floyd in September 1999 (Category 4) as well as Bonnie (Category 3) in August of 1998. A temporal regime change within the Neuse had been suspected earlier by Paerl (2006), who indicated that the increased tropical storms activity between 1996 and 1999 along with basin-scale management measures could have
Table 1 e Posterior model summary statistics for the floweconcentration changepointethreshold Bayesian model. Parameter Changepoint year a1 (Pre-1999 intercept) a2 (log TON concentration at identified flow threshold for the post-1999 model) b1,1 (Pre-1999 slope on log flow) b2,1 (Post-1999 slope on log flow before flow threshold) bTemp (Slope on water temperature) d (Change in the slope on flow subsequent to the flow threshold in the post-1999 model) s1 (Standard error for the pre-1999 model) s2 (Standard error for the pre-flow threshold & post-1999 model) s3 (Standard error for the post-flow threshold & post-1999 model) Flow threshold for post-1999
Unit Year log mg/L log mg/L
Mean
Standard deviation
Percentiles 2.5e97.5
1999 1.837 0.289
0.000 0.265 0.032
1999e1999 1.318e2.357 0.220e0.350
0.378 1.239 0.028 1.594
0.013 0.149 0.001 0.145
0.410 to 0.350 1.020e1.630 0.030 to 0.025 2.000 to 1.400
log mg/L log mg/L
0.379 0.537
0.008 0.023
0.360e0.400 0.490e0.590
log mg/L
0.399
0.009
0.380e0.420
log m3/s
3.376
0.045
3.288e3.464
e e e e
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 e6 2
resulted in a change in the primary production dynamics in the Neuse estuary. While both point-source reduction and increased flushing from hurricane activity could have played a role in forcing the regime change, we believe that the latter had less of an impact on the system. This is based on the fact that the floweconcentration relationship in the Neuse following the landfall of hurricane Fran (1996) did not deviate appreciably from the pre-1996 relationship. Furthermore, it is also evident from Fig. 4 that the floweconcentration relationships at high river flows are comparable for both pre- and post-1999 models, consequently challenging the idea that the system’s response in terms of TON delivery during extreme flow events has changed over time. The relationship between nutrient concentrations and river flow can be divided into two main behaviors, namely: dilution dominated regimes and flow-driven regimes (Johnson, 1979). Our results indicate that prior to the 1999 changepoint, the TON concentrationeflow relationship is typical of a dilution-dominated system, whereby increased river flows dilutes the overall TON load ð bd 1;1 ¼ 0:378Þ. These findings are consistent with the inverse floweconcentration relationship that Stow and Borsuk (2003) described earlier for the Neuse, using data from 1979 till 2000. Such a relationship is typical of water bodies where point sources play the main role in nutrient delivery. Note that no flow threshold was identified for the pre-1999 period (and as such there is no b1,2). Following the identified 1999 changepoint, our findings point to a transition in the relationship between flow and concentration. This transition involves moving the system away from a dilution dominated system towards a more complicated system where the floweconcentration relationship is no longer continuous or
Fig. 4 e FloweTON relationship both for pre- and post-1999 system changepoint. Reported TON concentrations have been adjusted for the effect of water temperature. The relationship pre-1999 was linear and dilution dominated. Post-1999, the relationship is piecewise linear with a threshold in flow at 30.429 m3/s.
57
monotonic in nature. We believe that this added complexity is a result of implementation aggressive point-source load reduction measures that tightened the discharge of nitrogen from point sources. Given that TON levels at low flows are indicative of mainly point-source discharges e as low flows are associated with very limited overland flow e we observe a significant drop in the contribution of point sources to the TON loads reaching the Neuse at Fort Barnwell post-1999 (Fig. 4). In addition, we observe that at low flow levels (below the flow threshold of 30.429 m3/s) TON concentrations post-1999 tend to increase with flow, as evident by the positive slope on flow ð bd 2;1 ¼ 1:239Þ (Table 1). This positive relationship can be attributed to the delivery of accumulated nutrients from agricultural fields, croplands, as well as from residential areas through overland flow and drainage canals. This behavior suggests that post-1999 non-point sources have come to play a more prominent role in TON delivery to Neuse. The relationship between flow and TON concentration shifts towards a dilution dominated system past the identified flow threshold d b for the post-1999 period ð bd 2;2 ¼ b2;1 þ d < 0Þ (Table 1 and Fig. 4). This shift is most probably due to the fact that past the flow threshold most of the nutrient load on the landscape has already been washed off. With respect to the seasonal component of the model (namely bTemp), we find that the posterior mean for the coefficient on water temperature is negative and is significantly different from zero (Table 1). This illustrates that riverine water temperature has an inverse relationship with respect to the measured TON concentrations. It is worth mentioning that given the sampling frequency bias that is inherent to our data (Fig. 2), we tested for the possible impacts that this bias might have on our model results. For this purpose, a balanced dataset was generated and used to run the developed model. The results from this test indicated that the model results were robust and did not significantly change as the data was balanced over the years. Details on the adopted method to balance sampling frequency over time, as well as the corresponding generated model results are presented in the online Supplementary Material. A comparison of the predictive distribution for both preand post-changepoint periods indicates that the mean predicted TON concentrations post-1999 are lower than the levels pre-1999 over all flows (Fig. 4). These results are a good indicator that the mitigation measures adopted by the State of North Carolina as part of the Neuse TMDL program have had an impact on reducing TON concentrations in the river and ultimately in the estuary too. Moreover, what is more important is to be able to assess whether the drop in concentration is consistent with the mandated 30% decrease in nitrogen loading. Simulation results for the change in TON concentration over the Neuse River hydrograph (and as such also loads) indicate that the predicted mean drop in TON concentration post-1999 is around 32%. Yet, what is more relevant than the mean load reduction in TON is the ability to capture the uncertainty around that reduction. Adopting the previously detailed simulation methodology (Section 2.3), allows us to easily determine the probability distribution for the TON load reduction. What we find is that the probability that TON loading has achieved the stipulated reduction goal is around 49%. Note that even though there is strong evidence of a drop in
58
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 e6 2
TON loading, the model results still indicate that there is a 28% chance that the concentrations (and similarly loads) have either not changed over time or have even increased post-1999. When the change in TON concentrations following the 1999 changepoint is more closely inspected, we find that the percent reduction along with the probability of attaining the 30% reduction goal is flow dependent. Fig. 5 shows that for low flows we are very confident that the reduction goal has been achieved. This can be seen by the fact that the 95% credible intervals around the load reduction are all above the 30% load reduction target. However, as flow levels increase we observe that the confidence in achieving the 30% load reduction diminishes, while the uncertainty in the percent change between pre- and post-1999 TON levels grows. This is consistent with the observed data that indicate that post the identified flow threshold, the floweconcentration relationship for pre- and post-1999 periods are almost identical (Fig. 4). This translates to the fact that at high flows the mean percent load reductions are close to zero and the credible intervals around them are considerably wide.
4.
Discussion
4.1.
Dynamics of the floweconcentration relationship
The results from our study strongly support the occurrence of a changepoint during the year 1999 for the Neuse River. The
change in system behavior in 1999 is most probably attributed to the implementation of the TMDL program that involved the execution of a set of point as well as non-point source nutrient control measures. While the TMDL recommended mitigation measures were executed over a period of time starting as early as 1997, the introduction in 1999 of biological nitrogen removal processes to the Neuse River Wastewater Treatment Plant, which is the major point-source contributor of TON in the Neuse watershed, seems to have culminated in a regime change when it comes to the floweconcentration relationship in the system. Since temporal invariance remains at the heart of many empirical water quality models, such as SPARROW and LOADEST, what our findings highlight is the need to adapt these models so that they can better reflect large-scale management decisions that occur on the watershed level over time. The results from this study also indicate that the classification of the nutrient delivery system in the Neuse River basin by Stow and Borsuk (2003) as point source dominated is only valid for the timeframe prior to the changepoint year of 1999. Post-1999, the system transitioned into a more complex system that is non-monotonic in nature. This added complexity is a direct result of the implementation of successful point-source discharge mitigation measures, while non-point source reductions have lagged behind at best and thus have come to play an increasingly important role in nutrient dynamics. In the pre-1999 period, one would have expected to find the highest TON concentrations at low flows; we now find that low flows post-1999 are associated with low TON concentrations, while the highest concentrations now occur at medium flow levels (around 30.429 m3/s). Moreover, our results indicate that dilution remains the dominating process at high flows in the Neuse as noted earlier by Borsuk et al. (2004). We should note that our developed model tends to over predict TON concentrations during major hurricanes (Fran in 1996 and Floyd in 1999), when extreme flow events are recorded (Fig. 4). This shortcoming of the model may warrant considering the inclusion of a second flow threshold to further govern the floweconcentration relationship at these extreme flows. Yet, we think that the inclusion of another threshold at this stage is not advisable due to the lack of enough data that adequately captures several severe hurricane events in order to ensure that the model is not biased towards a single event.
4.2. The effect of water temperature on TON concentrations
Fig. 5 e Percent reduction in TON levels as a function of flow. Black solid points indicate the predicted mean change in TON levels between pre- and post-1999 levels for the defined flow measurements. The dark grey bands correspond to ±1 standard error credible interval, while the light grey bands indicate ±2 standard error credible intervals. Dashed black line shows the TMDL stipulated reduction in nitrogen loading reaching the Neuse. Gaps towards the high flow values are a direct result of sampling from a right-tailed distribution.
Model results suggest that the dynamics of TON concentrations in the Neuse are seasonal in nature. The negative coefficient on water temperatures in our model indicates that given the same flow values, winter to early spring TON concentrations are expected to be higher than their counterparts during the summer to early fall period. This is consistent with our understanding of the Neuse system and can be explained by three main processes. The first involves the seasonal patterns of fertilizer application, with most fertilizer applied during the late winter and early spring period. The second is associated with the seasonal algal dynamics within the river that affect nutrient uptake, with late spring and summer blooms as described by Borsuk et al. (2004) and Paerl
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 e6 2
(2006). The last process involves an acceleration of the benthic denitrification rate as water temperatures increase (Alperin et al., 2000; Borsuk et al., 2004).
4.3. Assessing load reductions and their management implications The findings from our work have implications when it comes to assessing the effectiveness of the nitrogen mitigation measures executed by the State of North Carolina as part of the Neuse TMDL. The assessment of compliance for a nutrient like TON is complicated by the fact that the applicable standard of interest is set for chlorophyll-a concentrations and not for TON concentrations. With the absence of a TON criterion for the Neuse, compliance assessment has to be conducted through the assessment of load reduction over time. Given that we cannot measure load directly and since load itself is mostly dominated by flow rather than concentration, any adopted load reduction verification methodology has to take into account the underlying flow regime over the assessment period. Several studies have tried to quantify the temporal changes in nitrogen loading in the Neuse River Basin. The results from most of these studies have been inconclusive as to whether the reduction in nitrogen loading has met the 30% reduction goal or not (Stow et al., 2001; Burkholder et al., 2006; Paerl et al., 2006; Rajbhandari, 2007), with some studies even indicating that nitrogen load has actually increased post-TMDL implementation (McNutt, 2007). We think that the major source of this inconclusiveness pertains to three reasons: 1) the inability of the adopted methodologies to account for a changepoint in time, 2) the failure of some studies to properly account for flow, and 3) finally the use of dissimilar and relatively short time periods by different studies. Our methodology has tried to address these three limitations, while also focusing on accounting for the uncertainty in both model parameters as well as model structure. The results from our modeling approach, concerning load reduction in the Neuse, indicate that the TMDL for the Neuse appears to have had some success in reducing TON levels reaching the Neuse Estuary, with a 32% estimated mean drop in TON loading. Nevertheless, the results highlight that load reduction is a function of the observed flow regime in the Neuse. At low flow values, model results very clearly show that the load has been reduced by more than 30% (Fig. 5). Since low flows conditions are associated more closely with point-source loading, the results indicate that point-source mitigation measures targeting load reduction have been successful. Yet, the model results show that at high flow values the estimated load reductions are significantly lower than the 30% mark and are associated with a large degree of variability. Since the contribution of non-point TON sources increases with overland flow, this implies that either the state run non-point source nutrient control programs (the NC Agricultural Cost Share Program, the Conservation Reserve Enhancement Program, as well as USDA’s Environmental Quality Improvement Program) have had little success in reducing nitrogen washoff or that whatever success has been achieved in reducing non-point source nitrogen loading has been offset by an increase in the total load within the Neuse basin. We tend to think that the latter is more plausible, given the proactive engagement that the state has
59
adopted as well as given the findings reported by the state (Neuse Basin Oversight Committee, 2008). Our findings corroborate the conclusions that have been reported in the 2009 Neuse River Basinwide Water Quality Plan, where nutrient delivery through non-point sources was named as the primary cause of impairment to surface water in the Neuse River basin (Deamer, 2009). As such, any future reductions in nitrogen delivery to the Neuse Estuary, particularly TON concentrations, should concentrate on targeting non-point sources in the basin. We strongly believe that in order to increase the efficiency of the currently implemented non-point source nutrient reduction programs more emphasis has to be placed on monitoring and site inspection in order to better address the challenges associated with regulating non-point pollution sources that are diffuse and intermittent in nature, diverse in origin, and hard to identify. Continued efforts towards further load reduction is of particular importance in the Neuse watershed, where denitrification has been shown by Whalen et al. (2008) to be a minor sink for nitrogen removal, with an average removal rate of around 5% of the nitrogen reaching the river. Finally, it should be noted that our analysis has focused solely on TON and not on Total Nitrogen (TN) due to data limitations with TN measurements. TN monitoring limitations are not unique to this study, but are common to many river basins (Lewis et al., 1999). This is partly due to the fact that TON monitoring is often preferred by many monitoring agencies, given that it is instantaneous as compared to TN monitoring, which requires an extra digestion step that adds costs, requires further off-site analysis, and can even add bias (Lewis, 2002; Dodds, 2003). Luckily, the relationship between TON and TN in most river systems is linear in nature, which allows for the utilization of TON to better describe TN dynamics (Dodds, 2003; Turner et al., 2003). This linearity between TON and TN is also evident for the Neuse (Supplementary Material). Furthermore, it has been shown that TON forms the major constituent of the TN pool in major rivers, particularly in developed watersheds (Dodds, 2003; Turner et al., 2003; Sauer et al., 2008). This also holds true for the Neuse River, where TON concentrations contribute around 61% (15%) of the measured TN concentrations. The strength as well as the linearity of the relationship between TON and TN allows us to draw upon the TON results of our model to make conclusions on the TN dynamics in the Neuse, particularly when it comes to TN load reductions. Nevertheless, there has been evidence to indicate that the strength of the linear relationship between TN and TON weakens at low concentrations of TONs, where the balance between biota uptake and remineralization plays a significant role in observed TON concentrations (Dodds, 2003).
4.4. Bayesian framework, adaptive management, and long-term monitoring Our adopted Bayesian methodology provides the ability to quantify the uncertainty in the load reduction and to transparently report the outcomes associated with the success/ failure of the adopted nutrient management policies to the decision makers with a full disclosure of the uncertainties involved. The adoption of such a framework has been
60
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 e6 2
promoted by Reckhow (2003) and helps policy makers better assess their decisions by facilitating the evaluation of tradeoffs between accepting the risk of exceeding the TMDL’s stipulated reductions with a given probability versus incurring additional mitigation and monitoring costs in order to increase the frequency of attainment. Furthermore, the developed model framework removes many of the constraints that are often placed on the relationship linking flow to nutrient concentration. This added flexibility to the model structure allows for a much better understanding of the nutrient transport process and its dynamics in the Neuse River. It also helped resolve a large portion of the uncertainty in the observed data and to reconcile many of the conflicting studies both with respect to the success/failure of the TMDL program for the Neuse as well as the type of the relationship that governs the way flow affects TON concentration in the Neuse River. The fact that we had almost 3 decades of uninterrupted data presented us with an incomparable opportunity to better understand nutrient loading and its sensitivity to mitigation measures across time. This work underscores how our understanding of the system evolves as new data is collected. Our understanding of nutrient delivery in the Neuse River Basin started with a simple linear regression model (Stow and Borsuk, 2003), then it evolved to a piecewise threshold model (Borsuk et al., 2004), and now to a thresholdechangepoint model. We observe this succession as a dynamic learning process, whereby we build on our previous knowledge and incorporate new data to continuously update our models to better our understanding of our riverine systems. We strongly believe that our outlook fits well with the process of adaptive management or as Walters and Holling (1990) called it “learning by doing”, which has become widely accepted for environmental systems in general and has been endorsed by the National Research Council (2001) when it comes to dealing with TMDLs (Reckhow, 2003). Even though our results indicates that there is strong evidence to imply that the year 1999 resulted in a changepoint, there still is a crucial need to follow through this analysis as more data become available to make sure that alternative explanations to our model are formulated and evaluated. Moreover, our findings highlight the need to adopt monitoring programs that operate on long temporal scales that often exceed individual research projects. While the methodology that we presented has been applied to the Neuse River estuary, it can easily generalized to other systems that have had a large-scale management or perturbation occur that may warrant the presumption of the occurrence of a changepoint in time. Even though the developed model in this study has been constrained to identify a single changepoint in time, it can also be easily expanded to identify multiple changepoints if such an assumption is warranted. Furthermore, the model is not constrained to identify either a changepoint or a threshold if the data do not support such a claim.
5.
Conclusions
We have presented a Bayesian changepointethreshold model that allows for a better understanding of the floweconcentration relationship in the Neuse River basin by not
constraining the system to the assumption of temporal invariance. The model identified that the TMDL basin-scale management decisions that mainly targeted point-source reductions were responsible for both the occurrence of a changepoint in the TON nutrient delivery system as well as for the development of a flow threshold. The identified flow threshold following the 1999 changepoint signifies a shift in the Neuse River from a point source dominated system towards a system where non-point sources play an increasingly important role in TON load dynamics. Simulated load reduction estimates provided a realistic assessment of both the success and shortcoming of the existing management measures that have been put in place by the State of North Carolina as part of the Neuse TMDL program. Load reductions were found to be flow dependent, with reductions in TON load exceeding the stipulated 30% reduction benchmark at low flow values. On the other hand as flow levels increased, the average simulated load reductions dropped while the associated degree of uncertainty got amplified. The loading of TON to the Neuse was also found to be seasonal in nature. Higher concentrations of TON for a given flow value were found to occur more frequently in the winter and early spring as compared to the summer and early autumn seasons.
Acknowledgments This study has been supported and funded with funds from the USEPA Office of Research and Development’s Advanced Monitoring Initiative (AMI) Pilot Projects focused on GEOSS (Global Earth Observation System of Systems). Ibrahim Alameddine was partially supported by a scholarship from Anchor QEA. We thank Craig Stow, Jonathan Goodall, and Joseph Jakuta for their help in compiling the data for the Fort Barnwell station, NC.
Appendix. Supplementary material Supplementary data associated with this article can be found in the on-line version, at doi:10.1016/j.watres.2010.08.003.
references
Alexander, R.B., Smith, R.A., Schwarz, G.E., 2000. Effect of stream channel size on the delivery of nitrogen to the Gulf of Mexico. Nature 403, 758e761. Alexander, R.B., Smith, R.A., Schwarz, G.E., Boyer, E.W., Nolan, J. V., Brakebill, J.W., 2008. Differences in phosphorus and nitrogen delivery to the Gulf of Mexico from the Mississippi River Basin. Environmental Science and Technology 42, 822e830. Alperin, M.J., Clesceri, E.J., Wells, J.T., Albert, D.B., McNinch, J.E., Martens, C.S., 2000. Sedimentary processes and benthicepelagic coupling. Neuse River Estuary Modeling and Monitoring Project Final Report: Monitoring Phase, Report, pp. 63e105.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 e6 2
Borsuk, M.E., Stow, C.A., Reckhow, K.H., 2004. Confounding effect of flow on estuarine response to nitrogen loading. Journal of Environmental Engineering 130, 605e614. Burkholder, J.A.M., Dickey, D.A., Kinder, C.A., Reed, R.E., Mallin, M.A., McIver, M.R., Cahoon, L.B., Melia, G., Brownie, C., Smith, J., 2006. Comprehensive trend analysis of nutrients and related variables in a large eutrophic estuary: a decadal study of anthropogenic and climatic influences. Limnology and Oceanography 51, 463e487. Cohn, T.A., Caulder, D.L., Gilroy, E.J., Zynjuk, L.D., Summers, R.M., 1992. The validity of a simple statistical-model for estimating fluvial constituent loads e an empirical-study involving nutrient loads entering Chesapeake Bay. Water Resources Research 28, 2353e2363. Congdon, P., 2006. Bayesian Statistical Modelling. John Wiley & Sons Ltd, Chichester, 573 pp. Deamer, N., 2009. Neuse River Basinwide Water Quality Plan. NC Department of Environment and Natural Resources: Division of Water Quality, Planning Section e Basinwide Planning Unit, Raleigh, NC. Dodds, W.K., 2003. Misuse of inorganic N and soluble reactive P concentrations to indicate nutrient status of surface waters. Journal of the North American Benthological Society 22, 171e181. Environmental Protection Agency, 2009. Watershed Assessment, Tracking & Environmental ResultS (WATERS): National Summary of Impaired Waters and TMDL Information. Gelman, A., 2006. Prior distributions for variance parameters in hierarchical models. Bayesian Analysis 1, 515e533. Gelman, A., Carlin, J.B., Stern, H.S., Rubin, D.B., 2004. Bayesian Data Analysis. Chapman & Hall/CRC, Boca Raton. Gelman, A., Hill, J., 2007. Data Analysis Using Regression and Multilevel/Hierarchical Models. Cambridge University Press, New York, NY. Green, W.R., Haggard, B.E., 2001. Phosphorus and Nitrogen Concentrations and Loads at Illinois River South of Siloam Springs, Arkansas, 1997e1999 No. 01-4217. United States Geological Survey, Little Rock, Arkansas. Haggard, B.E., Soerens, T.S., Green, W.R., Richards, R.P., 2003. Using regression methods to estimate stream phosphorus loads at the Illinois River, Arkansas. Applied Engineering in Agriculture 19, 187e194. Hooper, R.P., Aulenbach, B.T., Kelly, V.J., 2001. The National Stream Quality Accounting Network: a flux-based approach to monitoring the water quality of large rivers. Hydrological Processes 15, 1089e1106. Ide, J.i., Nagafuchi, O., Chiwa, M., Kume, A., Otsuki, K., Ogawa, S., 2007. Effects of discharge level on the load of dissolved and particulate components of stream nitrogen and phosphorus from a small afforested watershed of Japanese cypress (Chamaecyparis obtusa). Journal of Forest Research 12, 45e56. Johnson, A.H., 1979. Estimating solute transport in streams from grab samples. Water Resources Research 15, 1224e1228. Kelly, J.R., 2008. Nitrogen effects on coastal marine ecosystems. In: Hatfield, J.L., Follett, R.F. (Eds.), Nitrogen in the Environment: Sources, Problems, and Management. Academic Press, Amsterdam, pp. 271e332. Kerman, J., Gelman, A., 2007. Manipulating and summarizing posterior simulations using random variable objects. Statistics and Computing 17, 235e244. Lewis, W.M., 2002. Yield of nitrogen from minimally disturbed watersheds of the United States. Biogeochemistry 57e58, 375e385. Lewis, W.M., Melack, J.M., McDowell, W.H., McClain, M., Richey, J. E., 1999. Nitrogen yields from undisturbed watersheds in the Americas. Biogeochemistry 46, 149e162. Lunn, D.J., Thomas, A., Best, N., Spiegelhalter, D., 2000. WinBUGS e a Bayesian modelling framework: concepts, structure, and extensibility. Statistics and Computing 10, 325e337.
61
Malone, T., Conley, D., Fisher, T., Glibert, P., Harding, L., Sellner, K., 1996. Scales of nutrient-limited phytoplankton productivity in Chesapeake Bay. Estuaries and Coasts 19, 371e385. McNutt, C., 2007. “Pre & Post” NSW Strategy Implementation Analysis. Fort Barnwell Ambient Station, North Carolina Division of Water Quality, Raleigh, NC. National Research Council, 2001. Assessing the TMDL Approach to Water Quality Management, Washington, D.C. Neuse Basin Oversight Committee, 2008. Annual Progress Report on the Neuse Agricultural Rule. N.C. Divison of Water Quality, Raleigh, NC. Ott, W.R., 1995. Environmental Statistics and Data Analysis. CRC Press L.L.C, Boca Raton. Paerl, H.W., Pinckney, J.L., Fear, J.M., Peierls, B.L., 1998. Ecosystem responses to internal and watershed organic matter loading: consequences for hypoxia in the eutrophying Neuse River Estuary, North Carolina, USA. Marine Ecology Progress Series 166, 17e25. Paerl, H.W., 1997. Coastal eutrophication and harmful algal blooms: importance of atmospheric deposition and groundwater as “new” nitrogen and other nutrient sources. Limnology and Oceanography 42, 1154e1165. Paerl, H.W., 2006. Assessing and managing nutrient-enhanced eutrophication in estuarine and coastal waters: interactive effects of human and climatic perturbations. Ecological Engineering 26, 40e54. Paerl, H.W., Mallin, M.A., Donahue, C.A., Go, M., Peierls, B.L., 1995. Nitrogen Loading Sources and Eutrophication of the Neuse River Estuary, North Carolina: Direct and Indirect Roles of Atmospheric Deposition No. PBe96e115217/XAB, North Carolina. Water Resources Research Inst., Raleigh, NC. Paerl, H.W., Valdes, L.M., Piehler, M.F., Stow, C.A., 2006. Assessing the effects of nutrient management in an estuary experiencing climatic change: the Neuse River Estuary, North Carolina. Environmental Management 37, 422e436. Park, J.H., 2006. Modeling structural changes: Bayesian estimation of multiple changepoint models and state space models. American Political Science Association Meeting 2006, Philadelphia. Qian, S.S., Cuffney, T.F., Alameddine, I., McMahon, G., Reckhow, K.H., 2010. On the application of multilevel modeling in environmental and ecological studies. Ecology 91, 355e361. Qian, S.S., Richardson, C.J., 1997. Estimating the long-term phosphorus accretion rate in the Everglades: a Bayesian approach with risk assessment. Water Resources Research 33, 1681e1688. R Development Core Team, 2009. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. Rajbhandari, R., 2007. Trend Analysis of Nitrogen and Phosphorus at Fort Barnwell, Neuse River Basin. North Carolina Division of Water Quality, Raleigh, NC. Reckhow, K., Stow, C., 1990. Monitoring design and data analysis for trend detection. Lake and Reservoir Management 6, 49e60. Reckhow, K.H., 2003. On the need for uncertainty assessment in TMDL modeling and implementation. Journal of Water Resources Planning and Management 129, 245. Runkel, R.L., Crawford, C.G., Cohn, T.A., 2004. Load Estimator (LOADEST): a FORTRAN Program for Estimating Constituent Loads in Streams and Rivers. In: Techniques and Methods Book, vol. 4. U.S. Geological Survey, p. 69. Sauer, T.J., Alexander, R.B., Brahana, J.V., Smith, R.A., 2008. The importance and role of watersheds in the transport of nitrogen. In: Hatfield, J.L., Follett, R.F. (Eds.), Nitrogen in the Environment: Sources, Problems, and Management. Academic Press, San Diego, CA. ¨ , D., Wiseman Jr., W. Scavia, D., Rabalais, N.N., Turner, R.E., JustiA J., 2003. Predicting the response of Gulf of Mexico hypoxia to
62
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 1 e6 2
variations in Mississippi River nitrogen load. Limnology and Oceanography 48, 951e956. Smith, R.A., Schwarz, G.E., Alexander, R.B., 1997. Regional interpretation of water-quality monitoring data. Water Resources Research 33, 2781e2798. Stow, C.A., Borsuk, M.E., 2003. Assessing TMDL effectiveness using flow-adjusted concentrations: a case study of the Neuse River, North Carolina. Environmental Science and Technology 37, 2043e2050. Stow, C.A., Borsuk, M.E., Stanley, D.W., 2001. Long-term changes in watershed nutrient inputs and riverine exports in the Neuse River, North Carolina. Water Research 35, 1489e1499. Stow, C.A., Roessler, C., Borsuk, M.E., Bowen, J.D., Reckhow, K.H., 2003. Comparison of estuarine water quality models for total maximum daily load development in Neuse River estuary. Journal of Water Resources Planning and Management 129, 307e314.
Turner, R.E., Rabalais, N.N., 1994. Coastal Eutrophication Near the Mississippi River Delta. Turner, R.E., Rabalais, N.N., Justic, D., Dortch, Q., 2003. Global patterns of dissolved N, P and Si in large rivers. Biogeochemistry 64, 297e317. United States Geological Survey, 2009. National Water Information System. United States Geological Survey, Reston, VA. Van Dongen, S., 2006. Prior specification in Bayesian statistics: three cautionary tales. Journal of Theoretical Biology 242, 90e100. Walters, C.J., Holling, C.S., 1990. Large-scale management experiments and learning by doing. Ecology 71, 2060e2068. Western, B., Kleykamp, M., 2004. A Bayesian change point model for historical time series analysis. Political Analysis 12, 354e374. Whalen, S.C., Alperin, M.J., Nie, Y., Fischer, E.N., 2008. Denitrification in the mainstem Neuse River and tributaries, USA. Fundamental and Applied Limnology/Archiv fur Hydrobiologie 171, 249e261.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 3 e7 4
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Autotrophic nitrogen removal from black water: Calcium addition as a requirement for settleability M.S. de Graaff a,b,*, H. Temmink a,b, G. Zeeman a, M.C.M. van Loosdrecht c, C.J.N. Buisman a,b a
Wetsus, Centre of Excellence for Sustainable Water Technology, Agora 1, P.O. Box 1113, 8900 CC Leeuwarden, The Netherlands Wageningen University, Sub-department of Environmental Technology, P.O. Box 8129, 6700 EV Wageningen, The Netherlands c Delft University of Technology, Department of Biotechnology, Julianalaan 67, 2628BC Delft, The Netherlands b
article info
abstract
Article history:
Black (toilet) water contains half of the organic load in the domestic wastewater, as well as
Received 8 February 2010
the major fraction of the nutrients nitrogen and phosphorus. When collected with vacuum
Received in revised form
toilets, the black water is 25 times more concentrated than the total domestic wastewater
15 June 2010
stream, i.e. including grey water produced by laundry, showers etc. A two-stage nitrita-
Accepted 7 August 2010
tioneanammox process was successfully employed and removed 85%e89% of total
Available online 14 August 2010
nitrogen in anaerobically treated black water. The (free) calcium concentration in black water was too low (42 mg/L) to obtain sufficient granulation of anammox biomass. The
Keywords:
granulation and retention of the biomass was improved considerably by the addition of
Black water
39 mg/L of extra calcium. This resulted in a volumetric nitrogen removal rate of 0.5 gN/L/d,
Separation at source
irrespective of the two temperatures of 35 C and 25 C at which the anammox reactors
Nitrogen removal
were operated. Nitrous oxide, a very strong global warming gas, was produced in situations
Anammox
of an incomplete anammox conversion accompanied by elevated levels of nitrite.
Calcium addition
1.
Introduction
Separation at source of household wastewater results in a concentrated stream from the toilet, called black water, and a relatively diluted stream from the bathroom, kitchen and laundry, called grey water (Otterpohl et al., 1999). A new sanitation concept was proposed (Zeeman et al., 2008), in which energy and nutrients are recovered from black water and clean water is produced from grey water. Concentrated black water can be efficiently treated in a UASB (Upflow Anaerobic Sludge Blanket) reactor at a relatively short hydraulic retention time (HRT) of 8.7 days (de Graaff et al., 2010a). Other options are treatment in a UASB-septic tank at a considerably longer HRT of
ª 2010 Elsevier Ltd. All rights reserved.
29 days (Kujawa-Roeleveld and Zeeman, 2006) or in a CSTR (Continuously Stirred Tank Reactor) at an HRT of 20 days (Wendland et al., 2007). The nutrients nitrogen and phosphorus are largely conserved in the effluent of these anaerobic reactors. Both from energy and cost perspective biological nitrogen removal from this effluent is preferred over nitrogen recovery (Strous et al., 1997; Wilsenach et al., 2003). Because during anaerobic treatment most of the organic material is removed to produce energy from the black water, autotrophic nitrogen removal by the nitritationeanammox process is the only feasible option (Strous et al., 1997). This process consists of partial nitritation where ca. 50% of the ammonium is converted to nitrite, in combination with the anammox process where
* Corresponding author. Current address: KWR Watercycle Research Institute, P.O. Box 1072, 3430BB Nieuwegein, the Netherlands. Tel.: þ31 (0)306069526 E-mail address:
[email protected] (M.S. de Graaff). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.010
64
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 3 e7 4
ammonium and nitrite are converted to dinitrogen gas. Nitrogen removal by nitritationeanammox already is successfully applied on full scale in reject water from municipal (van der Star et al., 2007; Wett, 2007; Joss et al., 2009) and industrial wastewater treatment plants (Abma et al., 2009). Recently, Vlaeminck et al. (2009) demonstrated the technical feasibility of a one-stage nitritationeanammox process to treat digested black water in a rotating contactor. In this study the two-stage nitritationeanammox process was applied to remove nitrogen from anaerobically treated source-separated black water, produced from vacuum toilets with a flushing volume of only 5 L/p/d (Meulman et al., 2008). This wastewater is about 25 times more concentrated, with respect to nitrogen, than the total wastewater stream from Dutch households, which includes grey water and flushing with conventional toilets (124 L/p/d (Kanne, 2005)). During anaerobic treatment, COD in this black water is reduced from 7.7 to 9.7 gCOD/L to 1.2e2.4 gCOD/L, but the liquid effluent still contains readily biodegradable organic material (0.48e0.87 g BOD5/L) for which aerobic post-treatment is required (de Graaff et al., 2010b). Nitrogen and COD concentrations (1e1.5 gN/L and 1.2e2.4 gCOD/L) are considerably higher compared to digested domestic sludge liquors (0.6e1 gN/L and 0.1e0.8 gCOD/L) (e.g. Hellinga et al., 1998; Caffaz et al., 2006). The two-stage nitritationeanammox process was chosen in this study to allow the independent study of the application of the separate processes (van der Star et al., 2007). Also, a separate reactor for partial nitritation may remove biodegradable organic material that otherwise could interfere negatively with the anammox process by stimulating heterotrophic denitrification (Udert et al., 2008). The aerobic conditions in the partial nitritation reactor also may enhance (bio-)flocculation of organic and colloidal material (Wile´n
et al., 2004), which therefore can easily be separated from the black water before it is treated in the anammox reactor. In a sequencing batch reactor (SBR) start up and stable operation of the anammox process was evaluated at 25 C and 35 C. In view of its environmental impact (Kampschreur et al., 2008), also the emission of greenhouse gas nitrous oxide (N2O) was included in this study. Emission of N2O from the new sanitation concept would have a negative impact on its sustainability and therefore should be avoided. Because of its low growth rate, excellent biomass retention is essential for anammox reactors, and the formation of granules therefore is desired (Strous et al., 1998). The presence of sufficient amounts of calcium stimulates granule formation and thus biomass retention. van der Star et al. (2008) reported growth of anammox in free cells rather than in granules at a calcium concentration of only 41 mg/L. In the anaerobically treated black water used in this research, calcium concentrations were similar to van der Star et al. (2008) (41e44 mg/L, results to be published). The effect of calcium concentration and the addition of calcium on granulation of anammox biomass were therefore also studied in this research.
2.
Material and methods
2.1. Combined anaerobic treatment and nitrogen removal from black water The effluent of a UASB reactor treating concentrated black water (de Graaff et al., 2010a) was used as the influent for a two-stage nitritationeanammox process (Fig. 1). Partial nitritation of the anaerobically treated black water took place in a continuously stirred reactor at 25 C. Details and results of
Fig. 1 e Treatment concept for black water: combined anaerobic treatment and nitrogen removal.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 3 e7 4
the partial nitritation reactor can be found elsewhere (de Graaff et al., 2010b). Effluent from the partial nitritation reactor was treated in two anammox reactors. For this purpose sequencing batch reactors (SBRs) were used at two different temperatures. SBR35 was operated at 35 C, i.e. the optimum temperature for anammox growth (Strous et al., 1998) and SBR25 at 25 C as this is the preferred temperature for an energy efficient treatment concept for black water.
2.2.
Influent to the anammox process
Before feeding it to the anammox process, the effluent from the partial nitritation reactor was filtered over a 100 mm sieve to prevent accumulation of sludge from the partial nitritation reactor in the anammox reactor. Na2-EDTA (0.00625 g/L), FeSO4 (0.00625 g/L) and a trace elements solution (1.25 mL/L) were added (van de Graaf et al., 1996; Jetten et al., 2005). Minor amounts of ammonium (0.5 mL/L of 3 M (NH4)2SO4) were added to prevent ammonium limitation in the reactor, because this may lead to accumulation of nitrite towards inhibiting levels. The composition of the influent to the anammox process is shown in Table 1. Calcium was added to SBR35 from day 223 at a concentration of 39 mg/L (0.6 mL/L of 240 g/L CaCl2$2H2O). SBR25 was started up later than SBR35 and additional calcium was added during the whole period of operation at a concentration of 39 mg/L (0.6 mL/L of 240 g/L CaCl2$2H2O). New influent was prepared two times per week.
2.3.
SBR35, operated at 35 C
SBR35 was operated at 35 C for 348 days and seeded with anammox sludge from the fullscale anammox reactor in Sluisjesdijk (Rotterdam, NL) (0.7 L of 26 gVSS/L, maximum load of 10 kgN/m3/d at 33 C (van der Star et al., 2007)). The water jacketed SBR had a total liquid volume of 5 L and was operated in cycles of 12 h. The reactor system was controlled by a PLC system (Siemens-PLC, logo 230RC). Each cycle consisted of 10 min settling, 10 min effluent discharge, 10 min idle time, 10.5 h of feeding and 1 h to complete the conversion processes. The reactor content was flushed with nitrogen gas (9.5 mL/ min) and carbon dioxide (0.5 mL/min) during the feeding phase. The CO2 supply was sufficient to control the pH at a value of 7.8 0.24. To avoid entrance of air during effluent
Table 1 e Characteristics of black water after anaerobic treatment, partial nitritation, sieving over 100 mm and addition of EDTA, FeSO4 and trace elements solution. NHþ 4 -N NO 2 -N NO 3 -N CODtotal CODsoluble TOC
Unit
Average
s.d.
mg/L mg/L mg/L mg/L mg/L mg/L
408 483 6.0 459a 389a 191
49.0 29.3 0.7 72.3 51.4 29.7
a Corrected for the contribution of NO 2 to COD (Nitrite exerts a COD of 1.1 gCOD/gNO 2 -N).
65
discharge, a gas bag (5 L) containing nitrogen was connected to the headspace of the reactor. The reactor content was mixed with a stirrer at a speed of 100 rpm. The reactor was equipped with baffles to prevent concentration gradients in the reactor. Excess sludge was removed daily (60 mL/d), which corresponds with an SRT of 83 days assuming that no biomass was lost with the effluent. The effluent from the partial nitritation reactor initially was diluted by 50%, decreasing to 35%, 20% and 10% and 0% each week with synthetic medium without ammonium and nitrite (Jetten et al., 2005). From day 33 onward the influent no longer was diluted.
2.4.
SBR25, operated at 25 C
SBR25 was operated at 25 C with synthetic influent (NaNO2, (NH4)2SO4, NaNO3), mineral medium and trace element solution (van de Graaf et al., 1996; Jetten et al., 2005) for the first 123 days and was seeded with anammox sludge from SBR35 (0.8 L, several samples mixed, 1.4 gVSS/L). The reactor vessel was surrounded by a silicone-heating blanket. The SBR had a total liquid volume of 4 L and was operated in cycles of 12 h. The reactor system was controlled by an ez-Control system (Applicon Biotechnology, the Netherlands). Each cycle consisted of 10 min settling, 10 min effluent discharge, 15 min idle time and the rest of the cycle was used for feeding until the maximum liquid level was reached. The pH was controlled at 7.7 by CO2 supply. Nitrogen gas was continuously added at a flow rate of 10 mL/min, except during settling and effluent discharge. To avoid the entrance of air during effluent discharge, a gas bag (5 L) containing nitrogen was connected to the headspace of the reactor. The reactor content was mixed with a stainless steel stirrer at a speed of 80 rpm and the reactor was not equipped with baffles. Starting at day 124 the synthetic influent was replaced with effluent from the partial nitritation reactor, increasing the percentage every week from 10%, 50%, 75%, 90% to 100%. SBR25 was operated for a period of 266 days.
2.5.
Sludge analysis
Sludge samples were destructed to determine total concentrations of calcium and phosphorus using the Ethos 1 Advanced Microwave digestion system from Milestone. The following procedure was based on standard methods (APHA, 1998). Two grams of sample was put into a special microwave vessel, 10 mL of HNO3 (68%) was added and milliQ water was added up to a total volume of 30 mL. The vessels were put into the manifold and placed into the microwave. Samples were heated to 180 C in 15 min and this temperature was maintained for another 15 min. After cooling down, the content of the vessels was transferred to a 100 mL flask and diluted to 1% acid for the ICP analysis.
2.6.
Analysis
Liquid samples, taken once or twice a week, were fractionated into suspended, colloidal and soluble compounds by filtering through a black ribbon paper filter (12e25 mm Schleicher & Schuell) and a membrane filter (0.45 mm Cronus filter PTFE). Chemical oxygen demand was determined using DrLange
66
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 3 e7 4
test kits according to standard methods (APHA, 1998). Because nitrite exerts a COD of 1.1 gCOD/gNO 2 -N, the COD values were corrected accordingly. Total ammonium nitrogen concentration (NHþ 4 -N) was determined in membrane filtered samples (0.45 mm) using DrLange test kits according to standard methods (APHA, 1998). Total nitrogen (TN) was determined in the unfiltered sample using DrLange test kits according to standard methods (APHA, 1998). Samples were diluted 5 or 10 times prior to analysis to exclude the possible interfering effect of other ions. 3 Anions (SO2 4 , PO4 , Cl , NO3 and NO2 ) were determined according to standard methods using ion chromatography (Metrohm 761 Compact IC). Calcium concentration was determined by ICP-OES (Perkin Elmer 5300 DV). TOC was analyzed with a Shimadzu TOC analyzer. Total Suspended Solids (TSS) and Volatile Suspended Solids were determined according to standard methods using the ashless black ribbon paper filter (Schleicher & Schuell). Gas samples from the headspace of the reactor were analyzed off-line for N2O on a Varian 8300 Custom Solutions gas chromatograph (Hayesep Q 80/100 Mesh 0.25m 1/ 16“ 1 mm Ultimetal CP1308) in which Helium gas was used as carrier and N2O was detected with an electron capture detector (ECD) at 200 C. The temperature of the injector and the column were respectively 200 C and 50 C (at Delft University of Technology). The second gas chromatograph (at Wageningen University) that was used was a CEinstruments GC8000 Top (Interscience, Breda the Netherlands), because the first GC was not available for analysis for a few samples (SBR35 day 236 and SBR25 day 144). N2O was separated on a Haysep Q 80e100 mesh 3m 1/8” SS column and detected with an ECD. Detection limit for both GCs was 0.5 ppm N2O. Gas samples were taken at the end of the feeding period of the SBR cycles.
2.7.
to methods previously described by Amann et al. (1990). The 16S rRNA-targeting oligonucleotide probes used in this study are AMX368, to detect all anammox organisms, and EUB338, EUB338II, and EUB338III, to detect all bacteria (Schmid et al., 2003). All probes were purchased as Cy3 (AMX368 probe) or 5 (6)-carboxyfluorescein-N-hydroxysuccinimide ester (FLUOS) (EUB probe) labelled derivatives from MWG-Biotech (Ebersberg, Germany) and were diluted to a final concentration of 50 ng/ml. Hybridization was performed at 46 C for 1.5 h followed by washing with pre-warmed (48 C) washing buffer and 15 min incubation at 48 C in washing buffer containing DAPI (0.2 mg/ml). The cells were observed under an epifluorescent microscope Leica DMI 6000B (Leica, Germany) equipped with Leica DFC 350 FX camera.
2.8.
Calculations
The nitrogen load was calculated as the sum of NH4, NO2 and NO3 in the influent and the removal rate as the sum of NH4, NO2 and NO3 removed (gN/L/d). Maximum specific activities in the reactor were determined monthly by increasing the influent flow for about 1 h to such an extent that a slow accumulation of nitrite and ammonium could be observed, up to a maximum nitrite concentration of 20 mgNO 2 -N/L, according to van der Star et al. (2008). The maximum removal rate (gN/L/d) was calculated as the difference between the loading rate and the accumulation rate. After determination of the biomass concentration in the reactor, the maximum specific activity was calculated (gN/gVSS/d). N2O fluxes were calculated using the flux of 10 mL/min (for both reactors), because the total gas production was not measured. The addition of CO2 and the produced nitrogen were assumed to be negligible.
Microbial analysis (SEM and FISH)
Granules from the two SBRs were used for scanning electron microscopical analysis (SEM). Immediately after sampling the samples were washed with phosphate buffer solution (PBS) for 10 min in Eppendorf tubes (those were used for the whole procedure). Samples were centrifuged for 2 min at 13,000 rpm and the supernatant was discarded, after each following step the centrifugation was carried out in the same way to remove the respective liquid. The fixation took place in 3.7% glutaraldehyde solution (SigmaeAldrich, Steinheim, Germany) at room temperature for 2 h or at 4 C for 24 h. Following that, the samples were washed twice in PBS and dehydrated in ascending concentrations of ethanol (30%, 50%, 70%, 90% for 20 min each; 96% for 30 min, twice). Finally they were air dried in a drying chamber (45 C, 30e60 min) and stored in a desiccator until the microscopical investigation. SEM was performed with a JEOL JSM 6480 LV microscope (JOEL Technics Ltd., Tokyo, Japan) in high vacuum mode (emission electrons detection, acceleration voltage 6e10 kV, operating distance 10 mm). The SEM Control V 7.07 software was used for control of the microscope and acquisition of the micrographs. Pictures were stored in bitmap format. FISH (Fluorescent In Situ Hybridization) was used to characterize the bacteria in the sludge. Cell fixation to the gelatinecoated slides and hybridization steps were carried out according
3.
Results
3.1.
SBR35, operated at 35 C
SBR35 was operated for 348 days on partially nitritated anaerobic black water and Fig. 2 shows the removal of ammonium and nitrite, together with the effluent concentration of nitrite. Because an excess of active biomass (2.2 gVSS/L) was present in the reactor at day 1 with an unknown activity (Table 2), the load could be increased relatively fast. For about 70 days (day 33e100) nitrogen removal was stable, and all of the effluent of the partial nitritation reactor could be treated at a removal rate of 0.48 gN/L/d. Nitrite was completely removed (99%) and ammonium was removed for 85% (Table 3). However, after day 100 nitrite started to accumulate and the load had to be decreased to 0.29 gN/L/d. On day 131 the stirrer speed was reduced to 90 rpm, decreasing the shear in the reactor, and excess sludge no longer was removed. As a result, the removal rate stabilized at a value of 0.27 gN/L/d. Effluent VSS concentrations decreased from 18 mgVSS/L to 2.3 mgVSS/ L. The maximum removal rate decreased to 0.43 gN/L/d and the specific activity to 0.32 gN/gVSS/d on day 140 (Table 2). An average SRT of 55 days was estimated for the first 131 days of operation, which due to the loss of biomass with the effluent
67
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 3 e7 4
Fig. 2 e Nitrogen loading and removal, and the effluent NOL 2 -N concentration in SBR35 at 35 C. Extra calcium was added from day 223 onwards.
was shorter than expected. Although an SRT of 55 days still should be long enough to avoid complete wash-out of anammox biomass, growth of other biomass, such as heterotrophs, could have reduced the specific SRT of the anammox bacteria. On day 190 nitrite started to accumulate once more; apparently biomass retention still was not sufficient to sustain a stable removal rate, despite the reduction in stirrer speed and the discontinuation of sludge wasting. Table 2 shows as well that the maximum removal in the reactor decreases from day 70 to a very low value (0.13 gN/L/d) on day 223. From day 223 the calcium concentration in the influent was increased from 42 5.7 mg/L to 73 14 mg/L in an attempt to stimulate the granulation process and in this manner improve biomass retention. As a result, the nitrogen removal increased rapidly (see Fig. 2 and Table 2), e.g. the volumetric removal rate increased from 0.10 gN/L/d on day 223 to 0.40 gN/L/d on day 348. At the same time significant wall
growth was observed after addition of extra calcium and this resulted in better anammox biomass retention. The wall growth had not been observed before, and the biomass concentration no longer could be measured without opening the whole reactor. The maximum nitrogen removal rate in the reactor increased to 0.87 gN/L/d (Table 2) on day 341 and the volumetric reactor removal to 0.40 gN/L/d on day 348 (Fig. 2). At day 348 the reactor was emptied and the total solids concentration and volatile suspended solids concentration were determined to be 6.9 gTSS/L and 2.9 gVSS/L, respectively. About half of the solids were attached to the reactor walls and 58% of the solids was inorganic material. A large fraction of the additional calcium precipitated and the reactor contained an increasing amount of inorganic material; the VSS/TSS ratio of the sludge decreased from 92% to 42% (from day 1 to day 348). Apparently phosphate was removed as well with the addition of calcium, because the concentration decreased
Table 2 e Maximum removal and activity in the anammox SBR35 at 35 C (±standard deviation). Day
1 36 49 70 90 140 154 223 273 341 348
Maximum removal [gN/L/d]
n.d. 1.2 1.3 1.1 0.89 0.43 0.32 0.13 0.48 0.87 Reactor stopped
Biomass concentration [gTSS/L]
[gVSS/L]
VSS/TSS (%)
2.4 1.3 2.1 1.6 2.1 1.8 1.8 1.9 1.2a 2.2a 6.9b
2.2 1.1 1.7 1.5 1.7 1.4 1.3 1.2 0.75a 1.3a 2.9b
92 85 79 90 81 75 75 63 65 59 42
Maximum specific activity [gN/gVSS/d]
n.d. ¼ not determined. a Due to growth on the walls, biomass concentration did not represent the total biomass concentration in the reactor. b Biomass concentration determined after opening the reactor, including biomass on the walls and bottom.
n.d. 1.1 0.78 0.77 0.51 0.32 0.24 0.11 0.64a 0.67a 0.30
68
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 3 e7 4
Table 3 e Overview of the performance in anammox SBR35 at 35 C (±standard deviation).
NO 2 -N/NH4-N NO 3 -N/NH4-N
removed produced
TN removal NH4-N removal NO 2 -N removal CODtotal removed CODsoluble removed
Unit
Day 33e100
Day 100e200
Day 250e348
mol/mol mol/mol % % % % %
1.26 (0.080) 0.079 (0.019) n.d. 85%(8.8%) 99% (0.33%) 13% (11%) 8% (7.4%)
1.35 0.11 87% 94% 99% 11% 1.1%
1.19 (0.092) 0.11 (0.025) 89% (2.6%) 98% (3.1%) 100% (0.20%) 5.1% (13%) 3.2% (11%)
(0.105) (0.022) (2.2%) (3.0%) (0.91%) (18%) (21%)
n.d. ¼ not determined.
from 64 mgPO4-P/L to 44 mgPO4-P/L from day 223 onward. The calcium content of the sludge increased from 116 mgCa/gVSS on day 211 to 409 mgCa/gVSS on day 348 while the phosphorus content increased from 144 mgP/gVSS to 284 mgP/gVSS, strongly indicating the occurrence of calcium phosphate precipitation. Fig. 3 shows the microscopic images of the granules that were formed in SBR35. Granules on day 140 without additional calcium in the influent exhibited filamentous structures. The granules on day 341 with additional calcium were more dense and bigger. Table 3 gives an overview of the removal efficiencies and production of nitrate in SBR35. Nitrite was always limiting and removed for 99e100%. Ammonium was usually present in excess and was removed for 85e98%. Total nitrogen removal was 89% during the last period. Remaining nitrogen in the effluent was 102 mgN/L, mainly consisting of nitrate (50 mgNO 3N/L) and small amounts of ammonium (9.8 mgNHþ 4 -N/L) and nitrite (1.8 mgNO 2 -N/L). Part of the nitrogen is soluble organically bound material and cannot be removed by the anammox process (40 mgN/L). Nitrate production always was lower
þ (0.079e0.11 NO 3 -N/NH4 -N) than stoichiometrically expected at þ maximum anammox growth (0.26 NO 3 -N/NH4 -N (Strous et al., 1998)). Two reasons can explain this lower nitrate production, heterotrophic activity or a lower removal of nitrite. In the first two periods (day 33e100, and day 100e200) nitrite removal was close to what was expected from anammox stoichiometry þ (1.32 NO 2 -N/NH4 -N) and probably nitrate was removed by heterotrophic denitrification explaining the deviation from the expected stoichiometry. In the last period after day 250 the lower nitrate production can be explained by a lower removal of nitrite þ þ (1.19 NO 2 -N/NH4 -N instead of 1.32 NO2 -N/NH4 -N), indicating that anammox was not growing at its maximum growth rate, producing less nitrate.
3.2.
SBR25, operated at 25 C
SBR25 was operated at 25 C for 266 days to investigate the nitrogen removal from black water at the preferred temperature for an energy efficient treatment concept for black water. The first 123 days synthetic medium was used as influent at a load of 0.42 gN/L/d (data not shown). Fig. 4 shows that the
Fig. 3 e Pictures of the granules in anammox SBR35 taken at day 140 (left) and day 341 (right).
69
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 3 e7 4
removal rate did not change after changing the influent to 100% effluent from the partial nitritation reactor and with addition of extra calcium. This demonstrates that the anaerobically treated black water did not contain compounds that may inhibit the anammox process. On day 186 the pH control did not work properly and nitrite accumulated, but the removal recovered quickly and could still be increased to 0.49 gN/L/d at day 266. The maximum specific activity was determined on four occasions and was 0.34 0.091 gN/gVSS/d. On day 279 the reactor was stopped and emptied. The total solids concentration and volatile suspended solids concentrations were determined to be 2.7 gTSS/L and 1.7 gVSS/L, respectively. The amount of solids attached to the walls was negligible. Calcium and phosphate were removed from the influent and similarly to SBR35 precipitated in SBR25. Because of the lower temperature less calcium carbonate may have precipitated than in SBR35, which can also explain the lower amount of inorganic material that was found in SBR25 compared to SBR35. On average SBR25 achieved 85% nitrogen removal (Table 4), with effluent concentrations of 133 mg TN/L, 33 mgNHþ 4 -N/L, 2.2 mgNO 2 -N/L and 65 mgNO3 -N/L. Nitrate production was þ lower (0.15 NO 3 -N/NH4 -N) than stoichiometrically expected þ (0.26 NO3 -N/NH4 -N), possibly due to heterotrophic denitrification (Table 4). Heterotrophic denitrification activity was confirmed in batch tests by the addition of acetate to sludge samples taken from the reactor on day 279. Nitrate and nitrite concentrations during these batch tests showed a small decrease in nitrate concentration and an increase in nitrite concentration (results not shown). A similar nitrogen removal of 85% was achieved in SBR25 compared to 89% in SBR35. Despite the lower temperature in SBR25 also a similar nitrogen removal rate of 0.49 gN/L/d was achieved in SBR25 compared to 0.40 gN/L/d in SBR35.
3.3.
Microbial analysis (SEM and FISH)
The SEM pictures in Fig. 5 show the difference in granules from SBR35 before and after addition of extra calcium.
Table 4 e Overview of the performance in anammox SBR25 at 25 C. Unit NO 2 -N/NH4-N NO 3 -N/NH4-N
removed produced
TN removal NH4-N removal NO 2 -N removal CODtotal removed CODsoluble removed
mol/mol mol/mol % % % % %
Day 151e266 1.27 0.15 85% 92% 100% 5.0% 4.6%
(0.14) (0.028) (2.9%) (4.3%) (0.32%) (15%) (12%)
Granules observed on day 348 showed a more densely populated surface than granules from day 204, which was most probably due to the addition of extra calcium. Microbial analysis by FISH showed that anammox in SBR35 at day 153 was present in a large fraction, although not all bacteria detected with the EUB338 probe hybridized with the AMX368 probe for the anammox (Fig. 6A). In SBR25 anammox was present in an even larger fraction, because almost all bacteria hybridized with the AMX368 probe for the biomass at day 181 (Fig. 6B). This showed that changing the influent to SBR25 from synthetic to anaerobic UASB effluent did not have a negative effect on the anammox bacteria.
3.4.
Nitrous oxide (N2O) emissions
N2O concentrations in the headspace of SBR35 varied from 25 ppm to 1825 ppm, measured in between days 119 and 281 (9 samples). This corresponded to respectively 0.02%e1.0% of the total nitrogen load to SBR35 (Fig. 7). The highest concentration of N2O in the off gas was measured when the nitrite concentration in the reactor was higher than usual. N2O concentrations in the headspace of SBR25 varied from 7.1 ppm to 14 ppm measured between days 144 and 172 (5 samples). This corresponded to approximately 0.01% of the total nitrogen load. On day 189 a high concentration of N2O was detected of 411 ppm, when the nitrite concentration in
Fig. 4 e Nitrogen loading and removal, and the effluent NOL 2 -N concentration in SBR25 at 25 C.
70
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 3 e7 4
Fig. 5 e SEM observations of SBR35 granules (on the left granules in the reactor at day 204 without additional calcium and on the right granules in the reactor at day 348 with additional calcium).
Fig. 6 e FISH pictures (left (red) AMX368, middle (green) EUBmix and right (blue) DAPI); A: SBR35 on day 153; B: SBR25 on day 181. (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 ) 6 3 e7 4
71
Fig. 7 e left: N2O production in SBR35 and the NOL 2 -N concentration in effluent; right: N2O production in SBR35 and SBR25 in relation to the NOL 2 -N concentration in effluent.
the reactor was 6.1 mgNO 2 -N/L, which was much higher than the usual concentration. Fig. 7 clearly shows that the percentage of N2O in the off gas increased with the nitrite concentration in the reactor.
4.
Discussion
In this study the feasibility and operating conditions of the nitritationeanammox process to remove the nitrogen from anaerobically treated black water were investigated and evaluated. The effect of temperature, calcium levels and the emission of N2O are discussed below.
4.1.
Nitrogen removal and the effect of temperature
The two-stage nitritationeanammox process removed 85%e 89% of total nitrogen, which is higher than the nitrogen removal efficiency of 76% reported previously for anaerobic black water treatment in a rotating biological contacter (RBC) (Vlaeminck et al., 2009). This was mainly due to a lower observed nitrate production than in the “standard” anammox stoichiometry (Strous et al., 1998), probably because of the occurrence of heterotrophic denitrification. In the RBC DO control was not applied and to prevent nitrite oxidation, elevated free ammonia concentrations were induced by increasing the pH and effluent ammonium concentrations were relatively high 100 mgNHþ 4 -N/L (Vlaeminck et al., 2009) compared to this research. Vlaeminck et al. (2009) did achieve a higher removal rate of 0.7 gN/L/d, whereas in this study 0.5 gN/L/d was achieved. However, in this study the removal rate was not optimized and several studies show that removal rates of 0.75e10 gN/L/d are possible (e.g. Dapena-Mora et al., 2004; Abma et al., 2007). In the research of Vlaeminck et al. (2009) the digested black water was stored for a long time before application. This may have resulted in a significant decrease of readily degradable organic compounds. In this research the readily degradable organic compounds were removed for a large fraction in the partial nitritation reactor (de Graaff et al., 2010b). In one-stage nitritationeanammox process, like the RBC, these readily degradable organic compounds could interfere negatively
with the anammox process by stimulating heterotrophic denitrification and influencing the sludge retention time of the anammox biomass (Udert et al., 2008). Recently the onestage nitritationeanammox process has been applied at the pilot plant in Sneek (the Netherlands), but results about the removal of organic compounds are not available yet (Meulman, 2010). Effluent total nitrogen concentrations were 102e133 mgN/L of which 25% was organically bound nitrogen. These concentrations are much higher than in effluents of conventional wastewater treatment plants (WWTPs) of 10 mgN/L (CBS-Statline, 2007). On the other hand, when comparing the loads, the discharge load from conventional WWTPs is still higher (1.2 gN/p/d, based on 124 L/p/d and 10 mgN/L in the effluent (CBS-Statline, 2007)) than the load of treated black water, which is 0.7 gN/p/d based on 5 L/p/d. In this research it was demonstrated that the anammox process can be successfully operated, both at 35 C and at 25 C. Similar nitrogen removal efficiencies were obtained and efficient biomass retention is crucial for operation at lower temperatures. This was also shown by Vlaeminck et al. (2009) who operated the one-stage nitritationeanammox at 26 C. Operation at elevated temperatures (35 C) is therefore not necessary (Dosta et al., 2008) and this saves in energy for heating, because the collected black water usually has a lower temperature of about 20 C.
4.2.
Calcium as a requirement for biomass retention
It is known that calcium and other polyvalent cations are important for bioflocculation and granulation, because they can form bridges between bacteria and bioflocs that are negatively charged due to the formation of EPS (Exocellular polymeric substances) (Sobeck and Higgins, 2002). By adding extra calcium to the influent, the ratio of monovalent to polyvalent cations ([M]/[P], equivalents) decreased, from 19 [M]/[P] to 12 [M]/[P] equivalents. Other factors to achieve the formation of anammox granules are a suitable selective pressure for settling using a short settling time, a low growth rate and the presence of inorganic precipitates (van der Star et al., 2008). In this research the settling time was 10 min in
72
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 3 e7 4
both SBRs, to obtain a selective pressure for settling, however this was not enough for sufficient granulation. Operation of SBR35 showed that anammox activity could not be sustained without addition of calcium. The increased removal after day 223 and the growth of a thick biofilm on the reactor walls and bottom in SBR35, indicated that the presence of calcium was important for biomass retention. Part of the calcium precipitated as calcium phosphates because the phosphorus concentration in the sludge and the inorganic fraction of the sludge increased as well after day 223. The amount of extra calcium addition should be optimized, because this accumulation of inorganic material is not desired. In SBR25 extra calcium was already added from the start and in this reactor no decrease in removal rate was observed. Calcium cannot be the only reason for granule formation, as discussed by van der Star et al. (2008). Other research showed that anammox granulation was possible at low calcium concentrations of 5.65 mg/L in membrane SBR (Trigo et al., 2006). In this way a long SRT was achieved and this resulted in a low growth rate favouring granule formation (Trigo et al., 2006; van der Star et al., 2008). Another aspect that should be taken into account is the availability of free calcium. Calcium can only precipitate and bind EPS if it is available in the free Ca2þ form. In black water large amounts of humic acids are present that can form complexes with calcium (van der Stelt et al., 2005) making the calcium less available. To our knowledge and according to Abma (2009) only wastewaters with relatively high calcium concentrations (>80 mg/L) have been treated with the anammox process (e.g. digester effluent from wwtp Dokhoven, Rotterdam, 120 mgCa/ L (analyzed twice), or urine with approximately 177 mgCa/L (Wilsenach, 2006)) and there is no experience yet in applying the anammox process for streams with lower calcium concentrations like the black water in this research. Vlaeminck et al. (2009) operated the one-stage nitritationeanammox process on digested black water for 5 months (155 days) during which no loss of nitrogen conversion was observed. This research (SBR35) also gave stable results for a long period, but showed a decrease in removal rate after 200 days. It is not clear what will happen on the long term in a RBC reactor as used by Vlaeminck et al. (2009). The addition of extra calcium will still be necessary to sustain anammox biomass in the reactor, or the presence of less shear in a biofilm reactor already allows for a satisfactory biomass retention. More research is needed to verify this effect of calcium on granule formation and biomass retention in different reactor configurations. This is also important for application at full scale and the choice between one- and two-stage processes. A two-stage process requires higher construction and maintenance costs because of the two separate reactors, but could achieve higher loading rates than a one-stage process.
4.3. Nitrous oxide (N2O) production in the anammox process The emission of N2O from both anammox SBRs (0.01%e1.0% of N load) were comparable with the emission of N2O in a fullscale anammox reactor treating reject water (0.6%)
(Kampschreur et al., 2008). Because N2O does not play a role in the anammox metabolism (Kartal et al., 2007), Kampschreur et al. (2008) discussed that the main causes for N2O production in an anammox reactor should be attributed to incomplete regular (heterotrophic) denitrification to N2O and the denitrification by ammonium-oxidizing bacteria (AOB). The maximum load of nitrate removed by heterotrophic denitrification on remaining COD from the black water (e.g. maximum 13% of total COD was removed in anammox SBR35 (Table 3)) was estimated to be 2.2% of influent nitrogen load (assuming that all the COD removed was used and 3 gCOD/ gNO 3 -N was needed). COD from decaying biomass could also have contributed to denitrification (Lackner et al., 2008), but this could not be further quantified. Although some denitrification is likely to have occurred, it is not very likely that all N2O was produced by incomplete denitrification. AOBs could have entered the SBRs with the influent, which originated from a partial nitritation reactor and could have produced N2O using nitrite as electron acceptor under oxygen limiting conditions (Kampschreur et al., 2008). Because the anammox influent was sieved, part of the AOBs could have been retained, which could have lead to an underestimation of N2O emission, provided that the AOBs are responsible for the N2O production. Fig. 7 shows that N2O production in the anammox process can be avoided by controlling the process such that nitrite is limiting and the concentration of nitrite in the reactor always is low. In the partial nitritation reactor the emission of N2O cannot be avoided because high nitrite concentrations are always present. In this concept for black water treatment this means that 1.9% of the nitrogen will be emitted as N2O in the partial nitritation process (de Graaff et al., 2010b). The emission of N2O in a full scale one-stage nitritationeanammox process was similar (1.2e1.3% of the nitrogen load) as the two-stage nitritationeanammox process (Kampschreur et al., 2009; Weissenbacher et al., 2010), therefore the choice between a one-stage or two-stage nitritationeanammox process seems not to be determined by the emission of N2O. Another study shows a lower N2O emission of 0.4% in a one-stage nitritationeanammox process (Joss et al., 2009). At conventional WWTPs the N2O emissions vary a lot, depending on the operation and characteristics of the wastewater. More research is needed to determine a quantitative relationship of these factors with the N2O emission (van Voorthuizen et al., 2009). A recent publication showed that N2O was only produced in a nitrifying process during recovery to aerobic conditions after a period of anoxia (Yu et al., 2010).
5.
Conclusions
The two-stage nitritationeanammox process removed 85%e89% of total nitrogen from anaerobically treated black water. The presence of calcium was crucial for granule formation to obtain high biomass retention and therefore an increasing removal of nitrogen in the reactors. The (free) calcium concentration in black water was too low (42 mg/L) to apply granular processes and addition of extra calcium was necessary to obtain a nitrogen removal of 0.5 gN/L/d,
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 3 e7 4
both at 35 C and 25 C. The specific activity of anammox biomass was lower at 25 C, but with an efficient biomass retention, operation at elevated temperatures (35 C) is not necessary and energy for heating can be saved. Nitrous oxide (N2O) was produced in both anammox SBRs, however only when the nitrite concentration increased because of an inefficient anammox conversion process. By preventing nitrite accumulation in the anammox reactor, N2O emissions can be prevented.
Acknowledgements The help of Marta Kamieniak and Mariana Pimenta Machado Braga dos Anjos is highly appreciated. Wouter van der Star and Marlies Kampschreur from Delft University of Technology are thanked for their useful discussions and advice on the manuscript. Wiebe Abma from Paques BV is thanked for providing the anammox biomass to start up the reactors. Astrid Helga Paulitsch is thanked for the help and work on the SEM. 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 theme “Separation at Source” for their interest and financial contribution.
references
Abma, W., 2009. Calcium Concentration in Wastewaters Treated by Anammox, Personal Communication. Paques BV. 7th September 2009. Abma, W.R., Schultz, C.E., Mulder, J.W., van der Star, W.R.L., Strous, M., Tokutomi, T., van Loosdrecht, M.C.M., 2007. Fullscale granular sludge anammox process. Water Science and Technology 55 (8e9), 27e33. Abma, W.R., Driessen, W., Haarhuis, R., Van Loosdrecht, M.C.M., 2009. Upgrading of sewage treatment plant by sustainable & costeffective separate treatment of industrial wastewater. In: 2nd Specialized Conference Nutrient Management in Wastewater Treatment Processes, Krako´w, Poland, 6e9th September 2009. Amann, R.I., Binder, B.J., Olson, R.J., Chisholm, S.W., Devereux, R., Stahl, D.A., 1990. Combination of 16S rRNA-targeted oligonucleotide probes with flow cytometry for analyzing mixed microbial populations. Applied and Environmental Microbiology 56 (6), 1919e1925. APHA, 1998. In: Clesceri, L.S., Greenberg, A.E., Eaton, A.D. (Eds.), Standard Methods for the Examination of Water and Wastewater. American Public Health Association, Washington. Caffaz, S., Lubello, C., Canziani, R., Santianni, D., 2006. Autotrophic nitrogen removal from anaerobic supernatant of Florence’s WWTP digesters. Water Science and Technology 53 (12), 129e137. CBS-Statline, 2007. Zuivering van stedelijk afvalwater; procesgegevens afvalwaterbehandeling (in Dutch). http:// statline.cbs.nl/StatWeb/publication/? DM¼SLNL&PA¼70152ned&D1¼0-1,3,5,18e19,21,23,36-
73
44,51&D2¼0,3,6,12,17&D3¼a&D4¼l&HDR¼T&STB¼G1, G2, G3&VW¼T (accessed 04.09.09.). Dapena-Mora, A., Campos, J.L., Mosquera-Corral, A., Jetten, M.S.M., Me´ndez, R., 2004. Stability of the anammox process in a gas-lift reactor and a SBR. Journal of Biotechnology 110, 159e170. Dosta, J., Ferna´ndez, I., Va´zquez-Padı´n, J.R., Mosquera-Corral, A., ´ lvarez, J., Me´ndez, R., 2008. Short- and Campos, J.L., Mata-A long-term effects of temperature on the anammox process. Journal of Hazardous Materials 154, 688e693. van de Graaf, A.A., de Bruijn, P., Robertson, L.A., Jetten, M.S.M., Kuenen, J.G., 1996. Autotrophic growth of anaerobic ammonium-oxidizing micro-organisms in a fluidized bed reactor. Microbiology 142, 2187e2196. de Graaff, M.S., Zeeman, G., Temmink, H., Buisman, C.J.N., 2010a. Anaerobic treatment of concentrated black water in a UASB reactor at a short HRT. Water 2 (1), 101e119. de Graaff, M.S., Zeeman, G., Temmink, H., van Loosdrecht, M.C.M. , Buisman, C.J.N., 2010b. Long term partial nitritation of anaerobically treated black water and the emission of nitrous oxide. Water Research 44, 2171e2179. 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 Science and Technology 37 (9), 135e145. Jetten, M., Schmid, M., van de Pas-Schoonen, K., Sinninghe Damste, J., Strous, M., 2005. Anammox organisms: enrichment, cultivation, and environmental analysis. Methods in Enzymology 397, 34e57. Joss, A., Salzgeber, D., Eugster, J., Ko¨nig, R., Rottermann, K., Burger, S., Fabijan, P., Leumann, S., Mohn, J., Siegrist, H., 2009. Full-scale nitrogen removal from digester liquid with partial nitritation and anammox in one SBR. Environmental Science and Technology 43, 5301e5306. Kampschreur, M.J., Van der Star, W.R.L., Wielders, H.A., Mulder, J. W., Jetten, M.S.M., Van Loosdrecht, M.C.M., 2008. Dynamics of nitric oxide and nitrous oxide emission during full-scale reject water treatment. Water Research 42, 812e826. Kampschreur, M.J., Poldermans, R., Kleerebezem, R., Van der Star, W.R.L., Haarhuis, R., Abma, W.R., Jetten, M.S.M., Van Loosdrecht, M.C.M., 2009. Emission of nitrous oxide and nitric oxide from a full-scale singlestage nitritation anammox reactor. In: 2nd IWA Specialized Conference Nutrient Management in Wastewater Treatment Processes (L.K. sp., Ed.), Krako´w, Poland, 6e9 September 2009. Kanne, P., 2005. Water Use at Households (Watergebruik thuis 2004, in Dutch), C4628, 31 January 2005. TNS-NIPO, Amsterdam. Kartal, B., Kuypers, M.M.M., Lavik, G., Schalk, J., Op den Camp, H.J. M., Jetten, M.S.M., Strous, M., 2007. Anammox bacteria disguised as denitrifiers: nitrate reduction to dinitrogen gas via nitrite and ammonium. Environmental Microbiology 9 (3), 635e642. Kujawa-Roeleveld, K., Zeeman, G., 2006. Anaerobic treatment in decentralised and source-separation-based sanitation concepts. Reviews in Environmental Science and Bio/ Technology 5, 115e139. Lackner, S., Terada, A., Smets, B.F., 2008. Heterotrophic activity compromises autotrophic nitrogen removal in membraneaerated biofilms: results of a modeling study. Water Research 42, 1102e1112. Meulman, B. Results from the One-stage NitritationeAnammox Process Pilot Plant in Sneek, Landustrie, the Netherlands. Personal Communication, 8th June 2010. Meulman, B., Zeeman, G., Buisman, C.J.N., 2008. Treatment of concentrated black water on pilot scale: options and challenges. In: Sanitation Challenge. Wageningen, the Netherlands 19e21 May 2008. Otterpohl, R., Albold, A., Oldenburg, M., 1999. Source control in urban sanitation and waste management: ten systems with
74
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 3 e7 4
reuse of resources. Water Science and Technology 39 (5), 153e160. Schmid, M., Walsh, K., Webb, R., Rijpstra, W.I.C., Van de PasSchoonen, K., Verbruggen, M.J., Hill, T., Moffett, B., Fuerst, J., Schouten, S., Sinninghe Damste´, J.S., Harris, J., Shaw, P., Jetten, M., Strous, M., 2003. Candidatus “Scalindua brodae”, sp. nov., Candidatus “Scalindua wagneri”, sp. nov., two new species of anaerobic ammonium oxidizing bacteria. Systematic and Applied Microbiology 26, 529e538. Sobeck, D., Higgins, M.J., 2002. Examination of three theories for mechanisms of cation-induced bioflocculation. Water Research 36, 527e538. van der Star, W.R.L., Abma, W.R., Blommers, D., Mulder, J.W., Tokutomi, T., Strous, M., Picioreanu, C., van Loosdrecht, M.C. M., 2007. Startup of reactors for anoxic ammonium oxidation: experiences from the first full-scale anammox reactor in Rotterdam. Water Research 41 (18), 4149e4163. van der Star, W.R.L., Miclea, A.I., Van Dongen, U., Muyzer, G., Picioreanu, C., Van Loosdrecht, M.C.M., 2008. The membrane bioreactor: a novel tool to grow anammox bacteria as free cells. Biotechnology and Bioengineering 101, 286e294. van der Stelt, B., Temminghoff, E.J.M., van Riemsdijk, W.H., 2005. Measurement of ion speciation in animal slurries using the Donnan Membrane Technique. Analytica Chimica Acta 552, 135e140. Strous, M., van Gerven, E., Zheng, P., Kuenen, J.G., Jetten, M.S.M., 1997. Ammonium removal from concentrated waste streams with the anaerobic ammonium oxidation (anammox) process in different reactor configurations. Water Research 31 (8), 1955e1962. Strous, M., Heijnen, J.J., Kuenen, J.G., Jetten, M.S.M., 1998. The sequencing batch reactor as a powerful tool for the study of slowly growing anaerobic ammonium-oxidizing microorganisms. Applied Microbiology and Biotechnology 50, 589e596. Trigo, C., Campos, J.L., Garrido, J.M., Me´ndez, R., 2006. Start-up of the anammox process in a membrane bioreactor. Journal of Biotechnology 126, 475e487. Udert, K.M., Kind, E., Teunissen, M., Jenni, S., Larsen, T.A., 2008. Effect of heterotrophic growth on nitritation/anammox in a single sequencing batch reactor. Water Science and Technology 58 (2), 277e284.
Vlaeminck, S.E., Terada, A., Smets, B.F., van der Linden, D., Boon, N., Verstraete, W., Carballa, M., 2009. Nitrogen removal from digested black water by one-stage partial nitritation and anammox. Environmental Science and Technology 43 (13), 5035e5041. van Voorthuizen, E., van Leusden, M., van Dongen, U., Kampschreur, M.J., van Loosdrecht, M.C.M., 2009. Emissies van broeikasgassen van rwzi’s. Inventarisatie van de emissie van broeikasgassen op Nederlandse rwzi’s. STOWA. concept report, 28 September 2009 (in Dutch). Weissenbacher, N., Takacs, I., Sudhir, M., Fuerhacker, M., Wett, B. , 2010. Gaseous nitrogen and carbon emissions from a fullscale deammonification plant. Water Environment Research 82 (2), 169e175. Wendland, C., Deegener, S., Behrendt, J., Toshev, P., Otterpohl, R., 2007. Anaerobic digestion of blackwater from vacuum toilets and kitchen refuse in a continuous stirred tank reactor (CSTR). Water Science and Technology 55 (7), 187e194. Wett, B., 2007. Development and implementation of a robust deammonification process. Water Science and Technology 56 (7), 81e88. Wile´n, B.-M., Keiding, K., Nielsen, P.H., 2004. Flocculation of activated sludge flocs by stimulation of the aerobic biological activity. Water Research 38, 3909e3919. Wilsenach, J.A., 2006. Treatment of source separated urine and its effect on wastewater systems. PhD Thesis, Department of Biotechnology, Delft University of Technology, June 26, 214 pp. Wilsenach, J.A., Maurer, M., Larsen, T.A., Van Loosdrecht, M.C.M., 2003. From waste treatment to integrated resource management. Water Science and Technology 48 (1), 1. Yu, R., Kamschreur, M.J., van Loosdrecht, M.C.M., Chandran, K., 2010. Directionality of autotrophic nitrous oxide and nitric oxide generation during transient anoxia. Environmental Science and Technology 44, 1313e1319. Zeeman, G., Kujawa, K., De Mes, T., Hernandez, L., de Graaff, M.S., Mels, A., Meulman, B., Temmink, H., Buisman, C., van Lier, J., Lettinga, G., 2008. Anaerobic treatment as a core technology for energy, nutrients and water recovery from source separated domestic waste(water). Water Science and Technology 57 (8), 1207e1212.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 7 5 e9 2
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Environmental toxicology and risk assessment of pharmaceuticals from hospital wastewater Beate I. Escher a,b,*, Rebekka Baumgartner a, Mirjam Koller a, Karin Treyer a, Judit Lienert a, Christa S. McArdell a a b
Eawag, Swiss Federal Institute of Aquatic Science and Technology, 8600 Du¨bendorf, Switzerland The University of Queensland, National Research Centre for Environmental Toxicology (Entox), 39 Kessels Rd, Brisbane, Qld 4108, Australia
article info
abstract
Article history:
In this paper, we evaluated the ecotoxicological potential of the 100 pharmaceuticals
Received 13 February 2010
expected to occur in highest quantities in the wastewater of a general hospital and
Received in revised form
a psychiatric center in Switzerland. We related the toxicity data to predicted concentra-
21 June 2010
tions in different wastewater streams to assess the overall risk potential for different
Accepted 9 August 2010
scenarios, including conventional biological pretreatment in the hospital and urine source
Available online 17 August 2010
separation. The concentrations in wastewater were estimated with pharmaceutical usage information provided by the hospitals and literature data on human excretion into feces
Keywords:
and urine. Environmental concentrations in the effluents of the exposure scenarios were
Pharmaceuticals
predicted by estimating dilution in sewers and with literature data on elimination during
Quantitative structure-activity
wastewater treatment. Effect assessment was performed using quantitative structure-
relationship
activity relationships because experimental ecotoxicity data were only available for less
Predicted no-effect concentration
than 20% of the 100 pharmaceuticals with expected highest loads. As many pharmaceu-
Risk quotient
ticals are acids or bases, a correction for the speciation was implemented in the toxicity
Elimination
prediction model.
Source separation
The lists of Top-100 pharmaceuticals were distinctly different between the two hospital
Wastewater
types with only 37 pharmaceuticals overlapping in both datasets. 31 Pharmaceuticals in the
Hospital
general hospital and 42 pharmaceuticals in the psychiatric center had a risk quotient above 0.01 and thus contributed to the mixture risk quotient. However, together they constituted only 14% (hospital) and 30% (psychiatry) of the load of pharmaceuticals. Hence, medical consumption data alone are insufficient predictors of environmental risk. The risk quotients were dominated by amiodarone, ritonavir, clotrimazole, and diclofenac. Only diclofenac is well researched in ecotoxicology, while amiodarone, ritonavir, and clotrimazole have no or very limited experimental fate or toxicity data available. The presented computational analysis thus helps setting priorities for further testing. Separate treatment of hospital wastewater would reduce the pharmaceutical load of wastewater treatment plants, and the risk from the newly identified priority pharmaceuticals. However, because high-risk pharmaceuticals are excreted mainly with feces, urine source separation is not a viable option for reducing the risk potential from hospital wastewater, while a sorption step could be beneficial. ª 2010 Elsevier Ltd. All rights reserved.
* Corresponding author. The University of Queensland, National Research Centre for Environmental Toxicology (Entox), 39 Kessels Rd, Brisbane, Qld 4108 Australia. Tel.: þ61 7 3274 9180; fax: þ61 7 3274 9003. E-mail address:
[email protected] (B.I. Escher). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.019
76
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 7 5 e9 2
1.
Introduction
1.1.
Environmental effects of pharmaceuticals
Pharmaceuticals are increasingly detected in surface waters, ground waters, and drinking water (Kolpin et al., 2002; Benotti et al., 2009; Watkinson et al., 2009) as not all are removed in conventional wastewater treatment plants (Joss et al., 2008). Often, it is difficult to establish cause-effect relationships of negative consequences for aquatic ecosystems (Ankley et al., 2007). Assessed as single compounds, most pharmaceuticals apparently pose no or moderate environmental risk. A notable exception is the negative effects on fish reproduction after exposure to estrogenic compounds (Routledge et al., 1998; Kidd et al., 2007). Likewise, the adverse effect of diclofenac on vulture populations in Pakistan (Oaks et al., 2004) demonstrates that under specific exposure conditions pharmaceuticals can cause problems. Increasingly, also negative effects of pharmaceuticals that are not related to the pharmacological effect (Owen et al., 2007) or its side effect (Oaks et al., 2004) are found (Tarazona et al., 2010), e.g. specific inhibition of photosynthesis in algae caused by b-blockers (Escher et al., 2006) and fluoxetine (Neuwoehner et al., 2009). Furthermore, in reality, rather than single compounds we find complex mixtures of pharmaceuticals and metabolites that may interact or show concentration additivity (e.g. Altenburger et al., 2004; Brian et al., 2007). Wastewater experts and policy makers are currently discussing whether micropollutants give sufficient rise to concern to justify removal measures from wastewater streams (FOEN, 2009).
1.2.
Removal of pharmaceuticals from wastewater
There are four approaches to remove micropollutants: optimize existing technology at wastewater treatment plants (WWTP), upgrade WWTP with new technology, source control, and source separation (Larsen et al., 2004). The main focus usually lies on end-of-pipe measures, and ozonation of the WWTP effluent or addition of powdered activated carbon were evaluated as promising tertiary treatment step. There was satisfactory removal of most pharmaceuticals by ozonation in a full-scale pilot plant (Hollender et al., 2009; Reungoat et al., 2010). However, removal of iodinated X-ray contrast agents is often not satisfactory. Dosages of 10e20 mg L1 powdered activated carbon also result in a good removal of a broad spectrum of micropollutants (Nowotny et al., 2007; Snyder et al., 2007).
1.3.
Source separation: the example of hospitals
Source control measures include strict prohibition (as for phosphate in detergents), emission standards (as for nutrients from WWTP), or designing pharmaceuticals with improved bio-degradability in cooperation with the industry. Urine source separation with NoMix toilets can contribute to reducing pharmaceuticals from diffuse household sources (Larsen et al., 2009; Lienert and Larsen, 2010, www.novaquatis. eawag.ch), which would on average reduce 60e70% of the mass (Lienert et al., 2007a) and approximately 50% of the
ecotoxicological risk of human pharmaceuticals from wastewater (Lienert et al., 2007b). The type and quantity of pharmaceuticals used in hospitals differs from what is used in the general population (Kummerer, 2001). Therefore, hospitals or homes for the elderly can be considered as point sources, and separate treatment of this wastewater is being discussed (Moser et al., 2007; Heinzmann et al., 2008). To date the contribution of hospitals to the pharmaceutical load in wastewater is unclear, since e.g. contraceptives or painkillers are widely used in the population. Various projects, including a large EU-consortium called “PILLS” (www.pills-project.eu) are currently determining the significance of hospitals as point sources for pharmaceuticals and pathogens, including multi-antibiotic resistant bacteria. In Switzerland, 18% of the total volume of the “most-sold top 100 active compounds list of pharmaceuticals” (IMS, 2004) is being administered in hospitals (Weissbrodt et al., 2009). In mass flow studies in a Swiss hospital, 50% of all X-ray contrast media, but only a few percent of the investigated cytostatics were recovered in the hospital sewer (Weissbrodt et al., 2009). The low recovery is mainly explained by pharmaceuticals consumed in the hospital but excreted at home by out-patients (50% out-patients for X-ray contrast media and 70% for cytostatics in this example). Cytostatics are considered to be especially harmful to the environment and are mainly administered in hospital settings (Lenz et al., 2007). Ort et al. recently determined the fractions of pharmaceuticals stemming from hospitals using a clever sampling design and chemical analytical quantification of 59 pharmaceuticals (Ort et al., 2010). For most pharmaceuticals the contribution of hospitals to overall wastewater was lower than 15%, with exception of two antibiotics (contrast media were not included in this study). These Australian results were consistent with a Norwegian analysis (Langford and Thomas, 2009). Similarly, the load of endocrine-disrupting chemicals did not differ between hospital and general wastewater (Pauwels et al., 2008).
1.4.
The dose makes the poison
Mass fluxes alone are insufficient to evaluate the risk stemming from pharmaceuticals; their ecotoxic potential needs to be considered, what to our knowledge has not been done for hospital wastewater so far. The risk quotient (RQ) is defined as predicted environmental concentration (PEC) divided by the predicted no-effect concentration (PNEC), which is extrapolated preferentially from chronic toxicity data, or, if no chronic data are available, from acute toxicity data (EMEA, 2006, European Parliament and European Council, 2006a). Despite recent large efforts to increase the database on ecotoxicological effects of pharmaceuticals (PhACT Database, 2006), there remain significant data gaps, especially when it comes to chronic effect data (Crane et al., 2006). Data gaps can be closed with predictive models using Quantitative Structure Activity Relationships (QSAR) but again chronic QSARs are less readily available (European Chemicals Agency, 2008; Escher et al., 2009). Therefore, the following analysis is based on acute toxicity data and uses an assessment factor of 1000 to extrapolate the PNEC, which is 100 times higher than the
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 7 5 e9 2
assessment factor of 10 recommended in the EMEA guideline to be used in association with chronic toxicity data. This implicitly accounts for an acute-to-chronic ratio of 100, which should be protective for most modes of toxic action, apart from endocrine effects as is discussed in section 3.3. Previous attempts to predict the risk of large lists of pharmaceuticals (Sanderson et al., 2004) were of limited use because they did not account for the speciation of pharmaceuticals. However, over 60% of pharmaceuticals are acids or bases that are fully or partially dissociated at ambient pH (Avdeef, 2003). Therefore, classical QSAR models cannot be applied without adaption and consideration of speciation of pharmaceuticals (Tarazona et al., 2010). Additionally to the risk from individual pharmaceuticals, also the risk from different mixtures should be estimated. We recently developed a toxicity model for mixtures consisting of an individual pharmaceutical and its metabolic transformation products (Escher et al., 2006; Lienert et al., 2007b). It can also be applied to mixtures of groups of different pharmaceuticals with a common (therapeutic) mode of action using the assumption of concentration addition of mixture toxicity e or for concentration addition of the underlying baseline toxicity for all groups of pharmaceuticals as discussed below.
1.5.
Mixture toxicity of pharmaceuticals
Pharmaceuticals are designed to be bioactive (with exception of contrast agents, which are rather diagnostics than pharmaceuticals). In non-target aquatic life many act as baseline toxicants. However, some exhibit the therapeutic effect also in aquatic life as the unwanted estrogenic effects on fish (Kidd et al., 2007). Others act via a different specific mode of toxic action, as evidenced for fluoxetine effects on algae (Neuwoehner et al., 2009). It is generally accepted that mixtures with components exhibiting the same mode of action act according to the model of concentration addition. If all components act according to a strictly different mode of action they cannot be modeled with concentration addition but act according to the model of independent action (Altenburger et al., 2003). For practical purposes, the concept of concentration addition is usually a realistic worst-case scenario because its prediction is often within an order of magnitude of the experimental findings (Altenburger et al., 2004). The majority of mixture studies with pharmaceuticals was on estrogenic chemicals (Brian et al., 2005, 2007; Thorpe et al., 2006; Kortenkamp, 2002, 2008); with few exceptions on other classes of pharmaceuticals (Escher et al., 2002, 2006; Cleuvers, 2004) and they generally confirmed concentration addition for pharmaceuticals from the same therapeutic class. Also analysis of a large number of pesticide mixtures confirmed that their aquatic mixture toxicity could be predicted by concentration addition in 90% of over 200 mixtures (Deneer, 2000). Furthermore, Hermens and Leeuwangh (1982) put forward the hypothesis that for mixtures of large numbers of chemicals with diverse specific modes of action, where the individual concentrations are well below the threshold of individual effect, the underlying baseline toxicity may add up to a significant mixture effect. All chemicals, regardless of whether they have a specific mode of toxic action, also exert a baseline toxic effect (van
77
Wezel and Opperhuizen, 1995). There is typically a threshold concentration below which the specific mode of toxic action is not observed and above which it is. At the concentrations at which acute toxicity usually occurs, the toxicity of a single pharmaceutical will be predominantly due to the specific mode of toxic action. However, in mixtures the concentration of each single component decreases, while the number of components with various different specific modes of toxic action increases. Therefore, the contribution to the total toxicity by the specific mode of toxic action decreases while that for the non-specific baseline toxicity increases (ECETOC, 2001). Warne and Hawker used this concept to develop the Funnel Hypothesis (Warne and Hawker, 1995). The Funnel Hypothesis argues that the more components an equitoxic mixture (a mixture where each chemical contributes the same to toxicity) contains, the larger the likelihood is that the compounds with specific modes of toxic action will not dominate the mixture toxicity. Thus the components will increasingly act only by their baseline mechanism of action and should be concentration additive. In wastewater, we have a large number of components of varying modes of toxic action. Thus we can assume that the toxicity of a very complex mixture is governed by the underlying baseline toxicity, not the specific mode of toxic action of single components. For risk assessment, if concentration addition can be assumed, the risk quotients of the individual pharmaceuticals can be added up to yield a sum risk quotient (RQmix).
1.6.
Ecotoxicological risk potential in four scenarios
The aim of this study was to estimate the risk potential of wastewater containing pharmaceutical mixtures from two point sources. The 100 active ingredients excreted in the highest amounts in 2007 from two different hospitals, one general hospital and one psychiatric center were compared. To evaluate the elimination of pharmaceuticals in conventional wastewater treatment plants (WWTP) and the effect of dilution of the hospitals’ wastewater in the sewer, we compared the following four scenarios for both hospitals: Scenario 1 HWW: Risk potential of the wastewater of the hospital main wing, before discharge to the sewer (i.e. full risk potential without any degradation or dilution). Scenario 2 WWTP influent: Risk potential at inlet of the WWTP (i.e. reduction of risk potential through dilution in sewers). Scenario 3 WWTP effluent: Risk potential at discharge of the WWTP (i.e. reduction of risk potential through degradation and sorption process during conventional biological treatment; including dilution in sewers). Scenario 4 HWWTP effluent: Risk potential at the hospital main wing after hypothetical conventional biological treatment (i.e. reduction of risk potential through degradation and sorption process in conventional biological treatment without dilution). This scenario thus assumes that some sort of biological treatment would be installed in the main wing of the hospital to deal with the wastewater; in an ideal case, the wastewater might then be directly discharged to surface waters or infiltrated).
78
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 7 5 e9 2
2.
Materials and methods
2.1.
General hospital
The first case is a typical, regionally important general hospital in Switzerland with 338 used beds serving more than 250000 inhabitants. In 2007, there were 122814 “days of care” and 16013 patients leaving the hospital. The whole range of medical services is offered, e.g. internal medicine, oncology, surgery, maternity clinic, nuclear medicine, and radiology, including computer tomography (CT) and magnetic resonance imaging (MRI). In 2008, 11767 CTs were carried out, of which 7490 were with X-ray contrast media; and 5154 MRIs (2691 with X-ray contrast media). Around two thirds of these X-rays were carried out with out-patients. In 2008, 209251 m3 water was used in total, and 115690 m3 in the main hospital wing that hosts patients and where pharmaceuticals are excreted. Wastewater is discharged to a WWTP with conventional biological treatment, which serves 54133 inhabitants. In 2007, the WWTP treated 8641486 m3 wastewater, and discharged 564993 m3 without treatment in combined sewer overflows during rain events. Pharmaceutical concentrations in the hospital wastewater were calculated for the main hospital wing. For the dilution to the WWTP influent, the combined sewer overflow was not considered, resulting in a dilution factor df of 0.013. The hospital kindly provided data of the pharmaceuticals administered in 2007. We additionally purchased Swiss pharmaceuticals sales data for 2004 from IMS (IMS, 2004). The amount of active ingredient in the pharmaceuticals was evaluated from Swiss drug documentations (Documed, 2009) and the sum of each ingredient calculated. Amounts of active compounds excreted unchanged in urine and feces were calculated using excretion rates from literature (Lienert et al., 2007a; Documed, 2009). If excretion was not clearly given, worst case scenarios with highest suggested excretion were taken. For active ingredients been used as cremes, an excretion of 75e100% was assumed, since wash off from the skin is also a source of water contamination without undergoing metabolism in human body. We assumed that all pharmaceuticals were excreted in the hospital, i.e. we neglected pharmaceuticals thrown away, and excretion by out-patients. The 100 active ingredients excreted in the highest amounts (Top-100 pharmaceuticals) were analyzed further in this study. In 2007, 1154 kg of pharmaceuticals were consumed in the hospital, of which 779 kg were excreted. The Top-100 list accounts for 1137 kg consumed pharmaceuticals (777 kg excreted). “Natural” ingredients such as metals, carbohydrates, sugars, enzymes, paraffin oil, herbal medicines etc. were omitted from the analysis. However, we included synthetic laxatives and synthetic sugars. In Swiss households, approximately 74.8 g pharmaceuticals per inhabitant per year were consumed, of which 23.4 g were excreted; based on data from IMS health of the Top-40 pharmaceuticals sold in pharmacies, drug stores, and doctor’s practices. Out of the Top-100 data received, 60 substances belong to the natural ingredients excluded for this study. The amount of pharmaceuticals discharged into this WWTP from
households totals 1267 kg per year or 62% of the total pharmaceutical load in the WWTP (2044 kg per year). Thus, around 38% of the pharmaceuticals at the WWTP in this case study stem from the hospital.
2.2.
Psychiatric center
The psychiatric case study is a regionally important Swiss psychiatric center with 211 used beds, providing stationary and ambulatory services. In 2007, 2008 patients received stationary treatment, with 76855 “days of care”. Besides acute adult psychiatry, there are e.g. wards for psychotherapy, addictive disorders, and geriatric psychiatry. There is also a housing group and working place for long-term psychiatric patients. According to interviews with head physicians and nurses (Lienert and Mosler, in preparation), many patients have acute psychiatric disorders. These are often in an extreme state at admission requiring strong medication. Therefore, there is a focus on pharmaceutical treatment. In 2007, 23250 m3 water was used in the psychiatric hospital. It is discharged to a WWTP, which treats 1742000 m3 raw wastewater with conventional biological treatment and serves 14603 inhabitants, yielding a dilution factor of the wastewater df of 0.013. In 2007, 52 kg of pharmaceuticals were consumed in the psychiatric hospital, of which 17 kg were excreted. As above, these numbers were calculated from the amounts of pharmaceuticals administered, which were kindly provided by the hospital. The Top-100 list, which consists of the 100 active ingredients excreted in the highest amounts, accounts for 50 kg of consumed pharmaceuticals, of which 17 kg were excreted. Again, “natural” ingredients such as metals, carbohydrates, sugars, enzymes, paraffin oil, herbal medicines etc. were discarded, but synthetic laxatives, such as synthetic sugars, included. These 17 kg excreted in the psychiatric center represent approximately 5% of the pharmaceuticals reaching the WWTP (359 kg per year in total) assuming a general excretion of 23.4 g per year per Swiss inhabitant as explained above.
2.3.
Exposure assessment
In the following, the calculation of the predicted environmental concentration PEC for the four scenarios is described. Only parent compounds were regarded and concentrations were corrected for metabolism in the human body. Metabolites were neglected because previous analysis showed that the contribution of metabolites to the overall risk is typically not very high. Moreover, exposure to metabolites is very difficult to assess due to highly variable literature reports on excreted metabolite fractions (Lienert et al., 2007b). In scenario 1, PECHWW was defined as the concentration of active ingredient expected in hospital wastewater. PECHWW was calculated from the amount of each active ingredient consumed in the hospital, M (g), the fraction excreted fexcreted of unchanged active ingredient in urine and feces and the volume of the hospital wastewater in the main wing where pharmaceuticals are consumed VHWW (L). PECHWW ¼
M$f excreted VHWW
(1)
79
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 7 5 e9 2
M was summed up from all amounts mi (g) of active ingredient consumed in the different drug preparations. We derived mi from the units consumed for each drug preparation, Ui, and the amount of active ingredient contained in each unit, mUi (g). M¼
n X
mi ¼
i¼1
n X
Ui mUi
(2)
i¼1
In scenario 2, PECWWTPinfluent was defined to be equivalent to the PECHWW multiplied with the dilution factor df in the sewer and corresponds to the concentration of pharmaceuticals at the inlet of the WWTP. The df was 0.013 for both, the general hospital and the psychiatric center. PECWWTPinfluent ¼ df $PECHWW
(3)
In scenario 3, PECWWTPeffluent refers to the discharge of the WWTP, where the PECWWTPinfluent was reduced by conventional biological secondary treatment with sludge age > 3 days in municipal wastewater treatment, including removal of organic material and denitrification/nitrification. Data on biodegradation were compiled from the literature (Supporting Information, Tables SI-1 and SI-2). The fraction eliminated in the treatment plant felimination in WWTP was assumed to be 0% if no literature data were available. PECWWTPeffluent ¼ f elimination
in WWTP $PECWWTPinfluent
(4)
For scenario 4, the same elimination rates were assumed for the wastewater treatment directly in the hospital (without dilution in the sewer), which yields the PECHWWTPeffluent. PECHWWTPeffluent ¼ f elimination
in WWTP $PECHWW
2.4.
Effect assessment
2.4.1.
Experimental ecotoxicity data
(5)
Literature was screened for ecotoxicity data for all 100 quantitatively most important compounds in each case study. For screening, a straightforward search approach was defined:
4. Search for data with google scholar (http://scholar.google. com.au/) using search terms “compound name”, “EC50”, and “algae”/”daphnia”/“fish”. Whenever possible, toxicity data were chosen that are consistent with the species of the selected QSAR to calculate baseline toxicity (see below). If such data were not available, the lowest acute EC/LC50 of another closely related biological species was chosen. If no acute value was available, also chronic toxicity data were used. However, as the discussion below demonstrates, ecotoxicological literature data on pharmaceuticals remains scarce and there is not enough chronic toxicity data available to base the analysis upon. Therefore toxicity was estimated with QSARs exclusively to avoid inconsistencies between data-rich and data-poor compounds.
2.4.2.
2.4.3. 1. Screen database on ecotoxicity data PhRMA PhACT(R) (PhACT Database, 2006). PhACT database is currently limited to members of PhRMA (US trade association) and was used with permission. 2. Screen the ECOTOX database of the U.S. EPA (http://cfpub. epa.gov/ecotox/) 3. Screen selected reports, books, and papers which compiled ecotoxicity data for pharmaceuticals (Hanisch et al., 2002, BLAC, 2003, Ku¨mmerer, 2004; Besse and Garric, 2007, SRU, 2007).
QSAR model to predict baseline toxicity
To calculate baseline toxicity of the 100 quantitatively most important compounds in each case study, established QSARs for algae-, daphnia-, and fish toxicity were used. The QSARs were selected from the Technical Guidance Document of the EU (European Commission, 2003) because they constitute a well-validated and often applied set. Most published baseline QSAR models were set up for neutral organic molecules and use the octanol-water partition coefficient Kow as hydrophobicity descriptor. However, many pharmaceuticals are acids or bases (Tarazona et al., 2010). For these, Kow is an unsuitable measure of bioaccumulation and surrogate for biomembranes, the target site for baseline toxicants. In pharmaceutical science, the liposome-water distribution coefficient at a defined pH value, e.g. pH 7, Dlipw(pH 7) has replaced the Kow as a descriptor for uptake into biological membranes. More recently, this model was also adapted in environmental science. For a historic overview refer to (Escher and Sigg, 2004). The logarithm of Dlipw(pH 7) was therefore used in the QSARs for baseline toxicity (Table 1) to calculate the toxicity of the compound towards the three aquatic organisms, algae, daphnia, and fish.
Estimating the hydrophobicity descriptor logDlipw(pH 7)
Dlipw(pH 7) is the lipid-water distribution coefficient that corrects for speciation at pH 7 in the case of organic acids and bases, since partitioning into membranes not only depends on the hydrophobicity of a compound but also on its charge and specific interactions with the membrane (Escher et al., 2000). Ideally (but rarely), the experimental Dlipw(pH 7) is available in the literature. If not, the liposome-water partition coefficient of the neutral species Klipw can be used together with an estimate of the speciation derived from the acidity constant pKa. If Klipw is not available it can be estimated from the Kow
Table 1 e Rescaled QSARs used to calculate baseline toxicity (Escher et al., 2009). The original QSAR (based on logKow) were taken from the Technical Guidance Document of the EU (European Commission, 2003). Baseline toxicity QSAR Biological species Green algae Water flea Fish
Scientific Name
Toxicity endpoint
Rescaled QSAR
Pseudokirchneriella subcapitata Daphnia magna Pimephales promelas
72e96h EC50 48h EC50 96h LC50
log(1/EC50(M)) ¼ 0.95$ logDlipw(pH 7) þ 1.53 log(1/EC50(M)) ¼ 0.90$ logDlipw(pH 7) þ 1.61 log(1/LC50(M)) ¼ 0.81$ logDlipw(pH 7) þ 1.65
80
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 7 5 e9 2
(Escher and Schwarzenbach, 2002). For consistency and fair treatment of data-rich and data-poor compounds, we consistently used estimated values of Dlipw(pH 7) derived from the Kow with the algorithms below. To calculate Dlipw(pH 7) from the Kow, following steps were undertaken for each compound: A. logKow-search: The databases of Kowwin v. 1.67 (U.S.EPA, 2008), ChemPlusID (http://chem.sis.nlm.nih.gov/ chemidplus/chemidlite.jsp), and PhysProp (http://www. epa.gov/oppt/exposure/pubs/episuitedl.htm, also accessible via http://www.syrres.com) were checked for an experimentally derived octanol-water distribution coefficient Kow. If no experimental value was found, the value estimated by a program of U.S. EPA (Kowwin v. 1.67, U.S.EPA, 2008) was used. As comparison, the Kow was also calculated using the online prediction program SPARC (Hilal et al., 2005). Contrary to Kowwin, which is based on a database of compounds with known Kow, SPARC calculates Kow values ab-initio from quantum mechanics. B. Selecting Kow and sorting out compounds without baseline toxicity: If the experimental or estimated value by Kowwin was logKow > 0 and less than 10 times greater or smaller than the value estimated by SPARC (logKow 1), the former was used for all further calculations. If both logKow (experimental/Kowwin and SPARC) were negative (i.e. logKow < 0, no accumulation in an organism), the compound was considered to show no baseline toxicity due to its low tendency to partition into biomembranes and insignificant contribution to the mixture toxicity. These compounds were excluded from all further calculations. If the two Kow differed more than an order of magnitude, several more estimation programs were used and the Kow from either Kowwin or SPARC closest to the mean values reported by vcc labs (Virtual Computational Chemistry Laboratory, 2009) was used. C. logKlipw calculation: logKlipw was calculated from the selected logKow using a QSAR for polar compounds (Vaes et al., 1997). logKlipw ¼ 0:905$logKow þ 0:515
(6)
D. Speciation at pH 7: SPARC (Hilal et al., 2005) was used to calculate the fraction that is neutral at pH 7 fneutral. The acidity constants pKa of single functional groups of a compound were also extracted from SPARC and where possible, experimental values from PhysProp database were collected as comparison. E. logDlipw(pH 7) calculation: Calculation of logDlipw(pH 7) based on Klipw of the neutral species and speciation uses the rough assumption that charged species (fraction 1e fneutral), independently whether they are positively or negatively charged, partition one order of magnitude less into organic phases than the corresponding neutral species (fraction fneutral) (equation (7)):
Dmw ¼ log Klipw ðneutral speciesÞ log Klipw ðcharged speciesÞ ¼ 1
ð7Þ
1 logDlipw ðpH7Þ ¼ log f neutral $10logKlipw þ 1 f neutral $10ðlogKlipw (8) We have discussed the limitations of using Δmw of 1 on numerous occasions (Escher and Sigg, 2004; Neuwoehner et al., 2009). Since the database is too limited to generate more precise estimates for Δmw, we kept the generic value of 1. Zwitterionic compounds were treated with a Δmw of 1, too, despite their overall net neutral charge because often the opposite charges are spatially isolated.
2.4.4. Calculating the predicted no-effect concentration (PNEC) To estimate the predicted no-effect concentration (PNEC), the lowest QSAR-based EC50 value (i.e. for the most sensitive species; either fish, daphnia, or algae) of each compound was divided by 1000. The Technical Guidance Document of the European Commission (2003) suggests an assessment factor of 1000 if acute toxicity data (for example EC50i, effect concentration of pharmaceutical i) are available in at least three test systems on three trophic levels: algae, daphnia, fish. PNECi ¼
EC50i 1000
2.5.
Risk analysis
2.5.1.
Calculating the risk quotient (RQ) of single compounds
(9)
For each pharmaceutical i, the risk quotient RQ was calculated as an indicator for ecotoxicological risk. The RQ is the ratio between the predicted concentration in the environment PEC and the concentration at which no effect is expected PNEC (EMEA 2006). RQ i ¼
PECi PNECi
(10)
RQ > 1 indicates an ecotoxicological risk for the aquatic environment. RQ < 1 indicates no ecotoxicological risk for the aquatic environment. Note, while for individual chemicals, the PNEC is derived from the most sensitive species, calculations for mixtures must be based on a common species. Therefore, we assessed the risk for algae, daphnia, and fish individually and then selected the species with the highest resulting RQmix for further analysis. We also point out that for hospital wastewater, cytostatic and antibiotic effects are of particular concern. However, there are only limited and non-standard ecotoxicological data available for these mechanisms.
2.5.2.
Mixture toxicity model and risk quotient of mixtures
The sum of the risk quotients of the Top-100 pharmaceuticals in each hospital was computed to allow comparing drug cocktails of variable compositions. According to the concept of concentration addition, the combined effect of the components is equal to the sum of the concentrations of each chemical expressed as a fraction of its own individual toxicity (Brown, 1968; Sprague, 1970). Concentration addition holds if the components of a mixture exhibit the same mode of toxic action. Since toxicity was estimated using a baseline toxicity
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 7 5 e9 2
QSAR for all compounds, this condition is fulfilled for the QSAR estimates but not necessarily for the experimental toxicity. If individual pharmaceuticals exhibit a specific mode of toxic action (which would be underestimated by the baseline toxicity QSAR), this specific effect would contribute to the mixture toxicity according to independent action, which is a generally lower contribution than one from concentration addition. Thus it is likely that the underestimation of specific toxicity is cancelled out because the contribution of this component is given a higher weight by using the mixture toxicity model with concentration addition instead of the model with independent action for specific toxicity. Hence, to calculate the mixture toxicity RQmix of all 100 quantitatively most important pharmaceuticals using the model of concentration addition, their risk quotients were summed up with eq. (13). RQ mix ¼
n X i¼1
RQ i ¼
n X PECi PNECi i¼1
(13)
By comparing RQi of single compounds to the total risk of the mixture RQmix, the pharmaceuticals or groups of pharmaceuticals of greatest concern can be identified and further assessed.
3.
Results and discussion
3.1.
Mass fluxes in hospital wastewater
The general and psychiatric hospitals showed very different pharmaceutical usage patterns in 2007 (Tables 2 and 3). First, the total amount of pharmaceuticals differed substantially. In the general hospital, 779 kg were excreted, from which we can predict a load excreted from each “bed” of 2.3 kg per year. In the psychiatric hospital only 17 kg were excreted, which gives an excreted load of 0.08 kg per bed. Second, also the types of pharmaceuticals differed significantly. In the general hospital, 58% of the excreted load stemmed from X-ray contrast media, 19% from laxatives, 16% from antibiotics, and 8% from others. In the psychiatric hospital, the main fraction came from laxatives with 36%, followed by analgesics/antiphlogistics to 17%, antidiabetics to 15%, psychotropic pharmaceuticals to 11%, and others to 21%. Even though all these pharmaceuticals were administered in the hospital, it is unclear, which fraction was excreted in the hospital and which fraction was taken home by out-patients. A mass flow study in another hospital showed that only 50% of all X-ray contrast media were excreted there (Weissbrodt et al., 2009). In our case study hospital, two thirds of the patients typically go home after receiving an X-ray, thus a significant fraction of pharmaceuticals will also be excreted at home. Likewise, since many older patients are in hospital, they take a number of pharmaceuticals regularly that they bring into the hospital. Since it is impossible to make an exact mass balance of which pharmaceuticals are excreted where, we assumed the worst case that all pharmaceuticals administered in the hospital would also be excreted there. Likewise, we did not account for the pharmaceuticals brought in by patients. Currently, a mass flow analysis study is performed at the general hospital. The wastewater from the hospital is
81
analyzed and compared to the wastewater of the receiving treatment plant with the aim to elucidate the load fraction of the hospital (Kovalova et al., in preparation).
3.2. High consumption does not always translate to high-risk The ten highest ranked PECHWW, i.e. the concentration of different active ingredients in the hospital wastewater, constituted 5970 mg/L in the general hospital. This equaled 89% of the sum of all Top-100 PECHWW (Table 2 and Table SI-1 in the Supporting Information). However, the mixture toxicity RQmix, i.e. the sum of the risk quotients of these Top-10 pharmaceuticals amounted only to 1.0, equaling 0.4% of the RQmix of the Top-100 pharmaceuticals. The reason is that among the Top-10 pharmaceuticals only two (4-methylaminoantipyrine and amoxillin) showed significant ecotoxicity (logDlipw(pH 7) > 0; Table SI-3). The remainder comprises the polymeric macrogol, which is the laxative polyethylene glycol, and contrast agents such as iodinized and gadalenium compounds of very low hydrophobicity. A similar result on the exposure side was obtained for the psychiatric hospital, where the Top-10 PECHWW summed up to 603 mg/L, which is 81% of the sum of all Top-100 PECHWW (Table 3 and Table SI-2 in the Supporting Information). However, the effect analysis came to a different conclusion than for the general hospital. There were only four pharmaceuticals in the Top-10 list that were not ecotoxic (logDlipw(pH 7) < 0; Table SI-4), namely the laxative macrogol, the antidiabetic metformin, magaldrate, a drug for acid related disorder, and the antiepileptic gabapentin. All others showed substantial ecotoxicity potential (diclofenac, ibuprofen, venlafaxine, amoxicillin, amisulpride, paracetamol). Consequently, the Top-10 pharmaceuticals with respect to their exposure amounted to 23% of RQmix (Table 3). Fig. 1A and B compare the PECHWW with the risk quotients of the different scenarios investigated. The data are ranked with decreasing PECHWW and all data are included, while Tables 2 and 3 only include the results with RQHWW > 0.01. Obviously, there is no correlation between PEC and RQ (Pearson’s R < 0.1). There were only few pharmaceuticals with a RQ > 1 in the hospital wastewater and these mostly had a PECHWW < 10 mg/L. A notable exception is diclofenac, whose risk was equally driven by exposure and effect. For most other compounds the main driver determining the RQHWW was the PNEC (Fig. 2). This observation is substantiated by the fact that the PECHWW varied in our selected dataset by less than four (general hospital, Table 2) and three (psychiatric center, Table 3) orders of magnitude, while the PNEC values covered almost eight orders of magnitude, resulting in an overall range of the RQHWW of more than seven orders of magnitude (Fig. 2). This analysis is relevant to prioritize pharmaceuticals for risk assessment. Generally, those pharmaceuticals with a high consumption are selected for further investigation and risk assessment, which is reflected by many studies on these compounds. However, those pharmaceuticals are not necessarily the most relevant ones with respect to their environmental risk as our present analysis indicates.
82
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 7 5 e9 2
Table 2 e General Hospital: Predicted Environmental Concentration in hospital wastewater PECHWW, Predicted No Effect Concentration PNEC for green algae, and Risk Quotients RQ for all four investigated scenarios.a Ranking according to decreasing RQHWW. Only pharmaceuticals with RQHWW > 0.01 are listed because the contribution of the remainders to the RQ is negligible. In the last row, the summed up risk quotients of the whole mixture of pharmaceuticals RQmix are given for all scenarios.
Amiodarone Clotrimazole Ritonavir Progesterone Meclozine Atorvastatin Isoflurane Tribenoside Ibuprofen Clopidogrel Amoxicillin Diclofenac 4-Methylamino-antipyrine Flucloxacillin/Floxacillin Salicylic acid Paracetamol Azithromycin Thiopental Oxazepam Valsartan Clarithromycin Rifampicin Tramadol Carbamazepine Tetracaine Sevelamer Metoclopramide Dipyridamole Pravastatin Prednisolone Erythromycin RQmix
PECHWW (mg/L)
PNEC (mg/L)
Scenario 1 RQHWW
Scenario 2 RQWWTPinfluent
Scenario 3 RQWWTPeffluent
Scenario 4 RQHWWTPeffluent
0.80 0.90 1 15.85 0.77 0.99 94 0.79 11.4 1.74 499 2.35 161.9 38.9 17.2 64 2.08 21.0 1.84 1.30 5.41 0.59 1.92 0.50 0.48 13.7 3.27 0.47 1.6 2.1 1.4
0.009 0.014 0.028 1.4 0.12 0.16 29.8 0.26 6.6 1.6 625 3.3 961 233 134 583 19 201 32 27 122 16 57 18 18 561 136 21 77 139 132
85.7 64.9 52.6 11.2 6.29 6.13 3.15 3.06 1.73 1.09 0.80 0.71 0.17 0.17 0.13 0.11 0.11 0.10 0.057 0.048 0.044 0.037 0.034 0.028 0.026 0.024 0.024 0.022 0.021 0.015 0.011 239
1.15 0.87 0.70 0.15 0.084 0.082 0.042 0.041 0.023 0.015 0.011 0.0095 0.0023 0.0022 0.0017 0.0015 0.0014 0.0014 0.0008 0.0006 0.0006 0.0005 0.0005 0.0004 0.0003 0.0003 0.0003 0.0003 0.0003 0.0002 0.0001 3.2
1.148 0.17 0.106 0 0.084 0.082 0.042 0.041 0.001 0.015 0.001 0.0063 0.0005 0.0002 0 0 0.0010 0.0014 0.0007 0.0001 0.0005 0.0005 0.0004 0.0004 0.0003 0.0003 0.0003 0.0003 0.0001 0.0002 0.0001 2.4
85.7 13.0 7.89 0 6.29 6.13 3.15 3.06 0.06 1.09 0.06 0.47 0.04 0.01 0 0 0.07 0.10 0.053 0.011 0.035 0.037 0.027 0.028 0.026 0.024 0.024 0.022 0.009 0.015 0.008 179
a The scenarios are: 1 ¼ Risk potential (RQ) of the wastewater from the hospital main wing before discharge to the sewer (i.e., full RQ of hospital wastewater (HWW) without any degradation or dilution); 2 ¼ reduced RQ of scenario 1 by dilution in sewer (i.e., at influent of WWTP); 3 ¼ reduced RQ of scenario 2 by degradation and sorption process during conventional biological treatment (i.e., at discharge of WWTP); 4 ¼ reduced RQ of scenario 1 by conventional biological treatment in hospital main wing (i.e., in effluent of HWW after on-site treatment).
3.3.
How good is the model for effect assessment?
Ideally, chronic toxicity data should be used for the risk assessment of pharmaceuticals (EMEA, 2006). However, data on the chronic toxicity of pharmaceuticals remain scarce (Crane et al., 2006) and the database is not sufficient for the risk analysis attempted here. The use of acute toxicity data is justified in those cases, where the acute-to chronic ratio (ACR) is in the typical range of 10e100 (Roex et al., 2000; Raimondo et al., 2007). However, for pharmaceuticals, the ACR can be much higher, especially for endocrine disruptors in fish such as ethinylestradiol or methyltestosterone, where the ACR may exceed 106 because adverse effects on the endocrine system require very low concentrations (Crane et al., 2006). In other aquatic species the ACR is typically much lower, even for endocrine disruptors (Sanderson and Thomsen, 2009). The top-100 list of pharmaceuticals used in hospitals contains only one sex hormone (progesterone) and three corticosteroids
(prednisolone, betamethasone, dexamethasone). Progesterone has not been tested in fish but its synthetic analogue levonogestrel exhibited chronic effects at the low ng/L range in adult fathead minnows (Zeilinger et al., 2009) and the resulting ACR is >106 (Berninger and Brooks, 2010). However, unlike the synthetic progestins, the natural substrate progesterone is rapidly degraded in wastewater treatment plant and is even not stable in a wastewater sample (Labadie and Budzinski, 2005; Esperanza et al., 2007). Due to its instability no toxicity data exist for progesterone and it is justified to neglect the specific progestagen activity in the risk analysis. As fish have corticosteroid receptors (Prunet et al., 2006), this might translate into a specific effect, but there are no experimental data available for corticosteroids apart from an ACR of 10 for algae (Crane et al., 2006). In addition, it cannot be fully excluded that none of the other pharmaceuticals exhibits a different and more sensitive mode of toxic action in a chronic toxicity study. This would
83
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 7 5 e9 2
Table 3 e Psychiatric hospital: Predicted Environmental Concentration in hospital wastewater PECHWW, Predicted No Effect Concentration PNEC for green algae, and risk quotients for all four investigated scenarios (see footnote Table 2). Ranking according to decreasing RQHWW. Only pharmaceuticals with RQHWW > 0.01 are listed because the contribution of the remainders to the RQ is negligible. In the last row, the summed up risk quotients of the whole mixture of pharmaceuticals RQmix are given for all scenarios.
Ritonavir Clotrimazole Diclofenac Mefenamic acid Lopinavir Nelfinavir Ibuprofen Chlorprothixen Trimipramine Quetiapine Meclozin Nevirapine Venlafaxine Promazine Efavirenz Olanzapine Levomepromazine Clopidogrel Methadone Carbamazepine Atazanavir Oxazepam Hexetidine Candesartan Duloxetine Aripiprazole Buprenorphine Benzoylperoxide Valproate Fluoxetine Lamotrigine Clozapine Diazepam Tramadol Pravastatin Trichlorethanol Amoxicillin Doxepin Citalopram Paracetamol Pantoprazole Clomethiazole RQmix
PECHWW (mg/L)
PNEC (mg/L)
Scenario 1 RQHWW
Scenario 2 RQWWTPinfluent
Scenario 3 RQWWTPeffluent
Scenario 4 RQHWWTPeffluent
0.86 0.39 73.0 5.38 0.26 0.71 26.3 2.53 0.63 7.31 0.11 0.98 24.6 1.67 0.16 8.41 1.15 0.72 3.75 5.00 0.14 7.24 0.21 0.51 0.38 0.11 0.13 0.22 4.05 0.54 0.65 0.97 0.48 2.60 3.39 3.50 22.8 0.17 0.51 9.61 0.72 0.28
0.03 0.01 3.31 0.79 0.05 0.16 6.62 0.91 0.49 7.98 0.12 1.3 35.5 2.7 0.3 14.9 2.4 1.6 10.5 17.7 0.6 32.5 1.0 2.9 2.3 0.7 1.5 2.5 51 6.9 8.7 16 10 57 77 86 625 4.8 17 583 45 23
30.8 28.0 22.1 6.77 5.60 4.47 3.97 2.78 1.28 0.92 0.88 0.75 0.69 0.62 0.58 0.56 0.480 0.452 0.357 0.283 0.251 0.223 0.205 0.177 0.166 0.157 0.089 0.088 0.080 0.078 0.0750 0.0590 0.0472 0.0456 0.0441 0.0407 0.0366 0.0361 0.0310 0.0165 0.0158 0.0122 114
0.41 0.37 0.29 0.09 0.07 0.06 0.05 0.04 0.02 0.012 0.012 0.010 0.009 0.008 0.008 0.008 0.006 0.006 0.005 0.004 0.003 0.003 0.003 0.002 0.002 0.002 0.001 0.001 0.001 0.001 0.0010 0.0008 0.0006 0.0006 0.0006 0.0005 0.0005 0.0005 0.0004 0.0002 0.0002 0.0002 1.5
0.06 0.07 0.19 0.06 0.07 0.06 0.00 0.04 0.02 0.012 0.012 0.010 0.009 0.008 0.008 0.008 0.006 0.006 0.004 0.004 0.003 0.003 0.003 0.002 0.002 0.002 0.001 0.001 0.001 0.001 0.0010 0.0008 0.0006 0.0005 0.0002 0.0005 0.0000 0.0005 0.0004 0.0000 0.0002 0.0002 0.7
4.62 5.60 14.6 4.33 5.60 4.47 0.15 2.78 1.28 0.92 0.88 0.75 0.69 0.62 0.58 0.56 0.480 0.452 0.286 0.283 0.251 0.207 0.205 0.177 0.166 0.157 0.089 0.088 0.080 0.052 0.0750 0.0590 0.0472 0.0362 0.0181 0.0407 0.0026 0.0361 0.0310 0.0000 0.0158 0.0122 52
cause an underestimation of the RQ of the individual components, and if these particular components had a large contribution to the mixture toxicity, the assumptions regarding the mixture toxicity model could be flawed. Nevertheless, based on currently available data, we regard our screening approach as a valuable contribution to risk assessment of hospital pharmaceuticals. Even experimental acute toxicity data were only available for a very limited set of compounds (Tables SI-5 and SI-6 in the Supporting Information). 16/15 (general/psychiatric hospital) acute EC50 values were found for algae (Tables SI-5A and SI6A), 19/21 acute EC50 for Daphnia (Tables SI-5B and SI-6B), and 16/18 acute LC50 for fish (Tables SI-5C and SI-6C). Thus even if
one resigns to acute toxicity data, less than 20% of the pharmaceuticals under investigation actually have experimental toxicity data. This percentage would not be sufficient for the envisaged analysis. Therefore, we had to use the QSAR models for the prediction of toxicity. To evaluate if the experimental toxicity data point to a specific mode of toxic action or if it can be explained by baseline toxicity, we performed a toxic ratio analysis. This analysis helps to decide if the use of baseline toxicity QSARs is justified or if there is a high probability that QSAR predictions lead to underestimation of toxicity as the pharmaceutical analyzed exhibits a specific mode of toxic action to the organism under evaluation.
84
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 7 5 e9 2
Fig. 1 e Risk Quotients RQ of the Top-100 pharmaceuticals ranked with decreasing Predicted Environmental Concentration PEC for A. the general hospital and B. the psychiatric center.
The toxic ratio TR (eq. (14)) is a measure of the specificity of effect (Maeder et al., 2004). If TR > 10, i.e., the experimental toxicity is at least ten times higher than the one predicted from the baseline toxicity QSAR, then the compound is likely to have a specific mode of action (Verhaar et al., 1992). If TR 10, the given compound exhibits merely baseline toxicity. TR ¼
EC50;baseline toxicity EC50;experimental
(14)
The majority of pharmaceuticals with experimental toxicity data could be classified as baseline toxicants with a toxic ratio analysis. Of the 15/16 experimental algae toxicity data, only three antibiotics had a TR exceeding 10 (Tables SI-5 and SI-6). Clarythromycin had a TR of 61165, sulfamethoxazole of 2867, and erythromycin of 6585. Metoprolol had a TR of 71, but another algal species was tested than Pseudokirchneriella subcapitata, which we use for QSAR calculations. For trimethoprim, a TR of 24 was derived from a NOEC value, so no quantitative comparison should be made due to mismatch of endpoints. Out of the 19
experimental Daphnia magna data, two analgesics, tramadol (TR ¼ 814) and paracetamol (TR ¼ 59) indicated specific toxicity. The TR of sulfamethoxazole of 16 was slightly increased but it is uncertain whether it exhibits a specific mode of toxic action as in algae. In fish, only one out of 16/18 experimental data points yielded a TR >10 but this value for tramadol is not reliable, since the fish species tested was not indicated. If we extrapolate the results of the TR analysis of this fraction of pharmaceuticals for which experimental data were available to all pharmaceuticals evaluated in this study, we can safely assume that > 90% of the top-100 pharmaceuticals act as baseline toxicants to the non-target aquatic organisms and that the remainder (<10%) will not dominate the toxicity of the mixture (see Section 1.5). We conclude that the QSAR model for baseline toxicity is valid to predict the toxicity of our mixtures of hospital wastewater. For 54 (general hospital) and 72 (psychiatric hospital) of the Top-100 pharmaceuticals, it was possible to derive a toxicity estimate. For the remainder, the predicted lipophilicity was so small (logDlipw(pH 7) < 0) that independent of the PEC no contribution to the toxicity was expected.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 7 5 e9 2
85
Fig. 2 e Range of values in Hospital Wastewater for Predicted Concentration PECHWW, Predicted No Effect Concentration PNEC, and Risk Quotient RQHWW in A. the general hospital and B. the psychiatric center.
3.4.
On which biological species to base risk evaluation?
In principle, PNEC must be derived from the biological species with the lowest EC50 by extrapolation with an uncertainty factor of 1000 (TGD, European Commission, 2003). This can be a different biological species for each pharmaceutical. However, for the mixture risk quotient, we have to work with a single species and cannot sum up risk quotients from different species. To choose the species for the final risk evaluation, PNECs were defined for each species separately, and risk quotients for all single pharmaceuticals calculated and summed up for each species. In the general hospital, RQmix was 239 for algae, 145 for Daphnia, and 48 for fish. In the psychiatric hospital, RQmix was 114 for algae, 77 for Daphnia, and 31 for fish. Hence, for both hospitals, algae produced the highest RQmix and fish the lowest, with a factor of five between the highest and the lowest RQmix for the general hospital and a factor of 4 for the psychiatric hospital. Thus in all further evaluations, the effect data for algae were used. The PNEC values reported in Tables 2 and 3 are those for green algae but the results for all biological species are given in the Supporting Information (Tables SI-5 and SI-6).
3.5. Mixture risk quotients in undiluted hospital wastewater The risk from the mixture of pharmaceuticals RQmix for scenario 1, i.e., hospital wastewater of the main wing without any dilution in the sewer, was 239 for the general hospital and 114 for the psychiatric hospital (Tables 2 and 3). In the general hospital for 10, 18 and 31 pharmaceuticals, the RQHWW was above 1, 0.1 and 0.01, respectively (Fig. 3A and Table 2), while for the psychiatric center, 9, 26 and 42 pharmaceuticals exceeded an RQHWW of 1, 0.1 and 0.01, respectively (Fig. 3B and Table 3). 31 pharmaceuticals in the general hospital and 42 pharmaceuticals in the psychiatric center made up more than 99% of the RQmix (i.e. had a RQ > 0.01) but together they constituted only 14% (general hospital) and 30% (psychiatric center) of the PECHWW. All of those with RQHWW > 0.01 are depicted in Fig. 3 and are further discussed below.
3.6.
Unexpected “high-risk” pharmaceuticals
Amiodarone, which had the highest ranked risk quotient RQHWW of 86 in the wastewater of the general hospital (Table 2) is an antiarrhythmic agent with numerous severe side effects. It is used in hospitals for cardiac arrest, serious disrhythmias, and other life-threatening situations (see http://www.drugs. com/amiodarone.html, accessed on 30 Nov 2009). It has been demonstrated that amiodarone disrupts the bacterial cell membrane and decreases bacterial growth (Rosa et al., 2000). Amiodarone, whose reported human side effect is cytotoxicity on thyroid follicular cells, also decreased T4 levels in zebra fish larvae (Raldua and Babin, 2009). No classical experimental ecotoxicity data were available for this pharmaceutical. However, the high experimental logKow of 7.8 (Table SI-3) yields a high toxicity prediction despite the fact that the tertiary amine amiodarone is almost completed protonated and thus positively charged at pH 7. Ritonavir dominated the RQmix of the psychiatric hospital with a RQHWW of 31 (Table 3) despite being only ranked 50th with respect to exposure (Table SI-2). In the general hospital, ritonavir was 3rd (RQHWW ¼ 53, Table 2) and 60th (PECHWW; Table SI1). Ritonavir is an antiretroviral drug to treat HIV infections (see http://www.aidsinfonet.org/fact_sheets/view/442, accessed 30 Nov 2009), which is often administered in a hospital setting. Ritonavir is a very large molecule and its hydrophobicity and ecotoxicity had to be estimated due to lack of experimental data. The high logKow of 6.27 (Table SI-3 and SI-4) together with its neutral speciation at pH 7 yields an exceptionally low PNEC of 28 ng/L and consequently a high-risk quotient (Table 3). Ritonavir is definitively a pharmaceutical warranting further attention and experimental investigations into its environmental risk. A search in ISI Web of Knowledge (http://apps. isiknowledge.com, accessed 21 June 2010) revealed not a single entry for the keywords “ritonavir and (ecotox* or environment*)”. This knowledge gap needs to be closed urgently given the high potential environmental risk of ritonavir. Clotrimazole ranked second for the risk quotient in, both, the general (RQHWW ¼ 65) and psychiatric hospital (RQHWW ¼ 28; Tables 2 and 3) despite being ranked only 75th
86
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 7 5 e9 2
Fig. 3 e Risk Quotients of the Top-100 pharmaceuticals ranked with decreasing Risk Quotient in Hospital Wastewater RQHWW for all pharmaceuticals with a RQHWW > 0.01 for A. the general hospital and B. the psychiatric center.
and 67th with respect to exposure (Tables SI-1 and SI-2). Clotrimazole is a widely used over-the-counter antifungal agent. It is very hydrophobic with an experimental logKow of 6.26 (Table SI-3). As imidazole derivative it has a basic function but the acidity constant pKa is low enough that at pH 7, the molecule is predominantly neutral. Both physicochemical properties point
to very high ecotoxicity, although few experimental data are available. Porsbring et al., (2009) recently demonstrated that clotrimazole has sublethal effects on natural marine microalgal communities (periphyton), altering the chlorophyll content and the cycling of photoprotective xanthophyll pigments already at environmentally relevant concentrations
87
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 7 5 e9 2
of 500 pM (0.17 mg/L), which is lower than our PECHWW. Clotrimazole has been found in concentrations of 10e100 ng/L in effluents of Swiss wastewater treatment plants (Kahle et al., 2008) and was also detected in UK surface waters (Roberts and Thomas, 2006). Not unexpectedly, diclofenac ranked third in the psychiatric hospital with a RQHWW of 22 (Table 3) and also third with respect to exposure (PECHWW ¼ 73 mg/L, Table SI-2). This reflects that its risk is equally driven by exposure and effect. However, in the general hospital diclofenac ranked much lower (RQHWW ¼ 0.71, Table 2; exposure: 45th rank, PECHWW ¼ 2.35 mg/L, Table SI-1). Because of the high exposure, diclofenac is well researched in ecotoxicology (Ferrari et al., 2004; Hallare et al., 2004).
3.7.
Comparison of two hospital types
33 Pharmaceuticals were overlapping in the Top-100 set of the general and the psychiatric hospital and 12 of them had a RQHWW > 0.01 in both hospital types. Together they made up 54% (general hospital) and 76% (psychiatric center) of the sum risk quotients. In this overlapping group there were the four commonly in wastewater detected compounds carbamazepine, diclofenac, ibuprofen, and paracetamol. These were also among the highest risk pharmaceuticals for the overall Swiss population including general and hospital use (Lienert et al., 2007b). Four of these overlapping pharmaceuticals, namely clopidrogrel, clotrimazole, meclozine, and ritonavir were in the lower field of exposure ranking (ranked 50th and higher) but exhibit a high ecotoxicity potential. Ritonavir and clotrimazole stick out with their high logKow and have risk quotient RQHWW > 1 in both hospitals as described in Section 3.6. Meclozine and clopidogrel exhibit RQHWW> 1 in the general hospital. The other common four pharmaceuticals, amoxicillin, oxazepam, tramadol, and pravastatin, have 0.01 < RQHWW < 1.
3.8.
Effect of biological treatment on risk quotient
The data on elimination of pharmaceuticals during wastewater treatment were collected from various literature sources (Ternes, 2000; Golet et al., 2003; Loffler and Ternes, 2003; Strenn et al., 2003; Joss et al., 2005; Bernhard et al., 2006; Buerge et al., 2006,, Zuehlke et al., 2006; Gobel et al., 2007; Kimura et al., 2007; Mahnik et al., 2007; Maurer et al., 2007; Nakada et al., 2007; Gulkowska et al., 2008; Kahle et al., 2008; Kasprzyk-Hordern et al., 2009; Radjenovic et al., 2009; Watkinson et al., 2007; Wick et al., 2009) and are listed in Tables SI-1 and SI-2 (Supporting Information). This compilation included values from municipal wastewater treatment
with activated sludge of a sludge age >3 days where denitrification/nitrification occurs. It does not differentiate between actual degradation and sorption to sludge. In Fig. 3, the risk quotients are plotted for all scenarios including those with elimination during wastewater treatment and dilution in the sewer for all pharmaceuticals with RQHWW >0.01 and are ranked according to RQHWW. This analysis is somewhat biased as for 55 of the Top-100 pharmaceuticals in the general hospital and for 66 of the Top-100 pharmaceuticals in the psychiatric center no literature data for biological elimination in wastewater treatment were available and therefore no elimination was assumed (Tables SI-1 and SI-2). As is evident from Fig. 3, dilution in the sewer generally had a larger effect on the decrease of the risk quotient than the actual elimination for most pharmaceuticals. For the pharmaceuticals with RQHWW >1, dilution in the sewer decreased the RQ to around or below 1 (RQWWTPinfluent 1). The RQWWTPeffluent decreased even further for clotrimazole and ritonavir, the Top-2 and Top-3 risk pharmaceuticals for the general hospital, due to high elimination rates in the WWTP. Ibuprofen was the only pharmaceutical in the group of RQHWW > 1 whose risk was reduced due to biological wastewater treatment, yielding a RQHWWTPeffluent < 1. However, for many pharmaceuticals in this group no elimination rates are available. Dilution in the sewer was more effective than removal by biological treatment. This is also evidenced in the psychiatric center where the four highest ranked risk pharmaceuticals (ritonavir, clotrimazole, diclofenac, mefenamic acid) all fell below RQ 1 due to dilution, while biological treatment was beneficial but could not fully compensate for the high ecotoxicity potential (Fig. 3B). A shortcoming of this analysis is that sorption to sewage sludge was not differentiated from actual degradation. Hydrophobic chemicals sorb better to sewage sludge than hydrophilic chemicals. The pharmaceuticals that dominate the RQmix are all very hydrophobic and can therefore be expected to be eliminated through sorption to sewage sludge. Clotrimazole and Ritonavir are eliminated to > 80% during wastewater treatment (Table SI-1). Unfortunately, for other compounds with a high RQHWW (e.g. amiodarone) no literature data are available on the elimination during wastewater treatment.
3.9.
Effect of urine source separation
The potential effect of urine source separation was also evaluated. Urine source separation is considered beneficial because it reduces the nutrient and micropollutant load of wastewater (Larsen et al., 2009; Lienert and Larsen, 2010). The overall pharmaceutical load is mainly expected in the fraction
Table 4 e Influence of source of Top-100 pharmaceuticals from urine or feces on PECHWW and RQHWW.
General hospital Psychiatric center
Sum PECHWW (mg/L)
Sum PECHWW (mg/L) urine
Sum PECHWW (mg/L) feces
Sum RQHWW
Sum RQHWW urine
Sum RQHWW feces
6720 364
4950 238
1770 126
239 114
28 28
210 86
88
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 7 5 e9 2
excreted with urine (e.g. Lienert et al., 2007a,b). Indeed, our survey confirmed that the pharmaceutical load entering wastewater via feces was much lower than that stemming from urine (Table 4). Exceptions were the laxatives, which are not taken up into the circulation. Additionally, the more hydrophobic compounds tend to be rather eliminated through feces than urine. In sum, 74% and 47% of the PECHWW was coming from urine for the general and psychiatric hospital, respectively (Table 4). However, when the RQ is analyzed, the picture looks different: The contribution of the individual pharmaceuticals to the risk quotient is illustrated in Fig. 4, where the RQ in urine and feces, RQHWW(urine) and RQHWW(feces) are plotted against the RQHWW. The dotted line corresponds to one matrix (either urine or feces) dominating the overall risk quotient, all points between the 1:1 line and the bottom points (which indicate no contribution to the RQ) indicate that urine and feces carry a share of the RQ. Despite the higher load of pharmaceuticals in urine in the general hospital, the RQHWW of the top-risk chemicals was generally dominated by the fraction excreted with feces, while for the low-risk pharmaceuticals urine was also a dominant excretory route (Tables SI-1 and SI-2 and dotted line in Fig. 4). The Top-3 pharmaceuticals, amiodarone/diclofenac, clotrimazole, and ritonavir constitute 85% and 71% of RQmix for the general and psychiatric hospital, respectively, and all are excreted predominantly via feces. For ritonavir, urine also plays a minor role, while for the two others urine is negligible as excretory route. As Fig. 4 demonstrates for the example of the general hospital, there is no relationship between the magnitude of RQHWW and its source of excretion from the human body. The three compounds with the highest risk, which dominate the overall RQHWW, all show very high excretion via feces. The fourth ranked pharmaceutical progesterone, in contrast, is predominantly excreted via urine. This analysis clearly demonstrates that urine source separation is a good mean to reduce the overall load of micropollutants, but it does not reduce the high-risk compounds and the risk potential of hospital wastewater. The high-risk
compounds are all very hydrophobic, which makes them intrinsically toxic but also causes excretion via feces because hydrophobicity and water (urine) solubility are inversely correlated (Schwarzenbach et al., 2003). Thus, a sorption step as pretreatment of hospital wastewater would potentially be appropriate before release of hospital wastewater into the communal sewer.
4.
Conclusions
Despite limitations of the toxicity estimation model, the results of the present study give a comprehensive picture on the risk posed by hospital wastewater. It allows setting priorities for further experimental testing. Interestingly (but disturbingly), the pharmaceuticals likely to pose the highest environmental risk have rarely been investigated previously. No or very few experimental data are available for the physicochemical properties and/or ecotoxicity of amiodarone, ritonavir, and clotrimazole, the three top-risk compounds in the general hospital. In the psychiatric center, diclofenac was among the three top-risk compounds, together with ritonavir and clotrimazole. Diclofenac is the only one of these pharmaceuticals that is well researched in ecotoxicology and risk assessment. As this analysis has demonstrated, the PNEC is generally the more important driver for the RQ. The reason is that the variability in the PNEC among all pharmaceuticals investigated is more than seven orders of magnitude while the PEC values cover only three to four orders of magnitude among the group of 100 most used pharmaceuticals. This means that if pharmaceuticals are selected only according to their usage pattern and occurrence, one might miss relevant ones that could pose an environmental risk. Therefore, consumption data are less suited to guide prioritization, but often the only available source for compound identification. Thus hazard identification should precede risk assessment to prioritize according to intrinsic hazard properties such as potential for persistence, bioaccumulation, and toxicity (PBT). The regulation for industrial chemicals in Europe, REACH, has exactly taken this step by using a PBT assessment to identify chemicals to be prioritized for further testing and risk assessment (European Parliament and European Council, 2006b). Following this recommendation, the European Medicines Agency’s guideline also advises to include PBT assessment in the prescreening phase of risk assessment of pharmaceuticals for pharmaceuticals exceeding a logKow of 4.5 complementing the exposure estimate as trigger for refined risk assessment (EMEA, 2006).
Acknowledgements
Fig. 4 e Contribution of urine and feces to the Risk Quotient in Hospital Wastewater, RQHWW for the general hospital.
The PhRMA PhACT(R) database was kindly provided by Vince D’Aco of Quantum Management Group, Inc. We very much thank the general hospital and psychiatric clinic for providing the pharmaceutical consumption data for our case studies and for the very helpful cooperation throughout the project. Funding by Eawag via action field; the State Secretariat for Education and Research SER/COST within the COST Action 636
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 7 5 e9 2
“Xenobiotics in the Urban Water Cycle” (project no. C05.0135); the EU project NEPTUNE (Contract No 036845, SUSTDEV-20053.II.3.2, European Community’s Sixth Framework FP6-2005Global-4), and the Swiss Federal Office for the Environment (FOEN) are acknowledged.
Appendix. Supporting information Supporting information related to this article can be found at doi:10.1016/j.watres.2010.08.019.
references
Altenburger, R., Nendza, M., Schuurmann, G., 2003. Mixture toxicity and its modeling by quantitative structure-activity relationships. Environmental Toxicology and Chemistry 22 (8), 1900e1915. Altenburger, R., Walter, H., Grote, M., 2004. What contributes to the combined effect of a complex mixture? Environmental Science & Technology 38 (23), 6353e6362. Ankley, G.T., Brooks, B.W., Huggett, D.B., Sumpter, J.P., 2007. Repeating history: pharmaceuticals in the environment. Environmental Science & Technology 41 (24), 8211e8217. Avdeef, A., 2003. Absorption and Drug Development: Solubility, Permeability, and Charged State. Wiley, Hoboken, N.J., USA. Benotti, M.J., Trenholm, R.A., Vanderford, Brett J., Holady, J.C., Stanford, B.D., Snyder, S.A., 2009. Pharmaceuticals and endocrine disrupting compounds in U.S. drinking water. Environmental Science & Technology 43, 597e603. Bernhard, M., Muller, J., Knepper, T.R., 2006. Biodegradation of persistent polar pollutants in wastewater: comparison of an optimised lab-scale membrane bioreactor and activated sludge treatment. Water Research 40 (18), 3419e3428. Berninger, J.P., Brooks, B.W., 2010. Leveraging mammalian pharmaceutical toxicology and pharmacology data to predict chronic fish responses to pharmaceuticals. Toxicology Letters 193, 69e78. Besse, J., Garric, J. 2007. Me´dicament a` usage humain: risque d’exposition et effect sur les millieux re´cepteur. Proposition d’une liste de me´dicaments a` usage humain a` surveiller dans les eaux de surface continentale. R.M.C., A.d.l.E. (ed), Cemagref et Agence de l’Eau R.M.C., Lyon, France, Lyon. BLAC, 2003. Arzneimittel in der Umwelt Auswertung der Untersuchungsergebnisse. Chemikaliensicherheit, B.L.f. (ed), Freie und Hansestadt Hamburg, Beho¨rde fu¨r Umwelt und Gesundheit, Institut fu¨r Hygiene und Umwelt, im Auftrag des Bund/ La¨nderausschusses fu¨r Chemikaliensicherheit (BLAC), Hamburg. Brian, J.V., Harris, C.A., Scholze, M., Backhaus, T., Booy, P., Lamoree, M., Pojana, G., Jonkers, N., Runnalls, T., Bonfa, A., Marcomini, A., Sumpter, J.P., 2005. Accurate prediction of the response of freshwater fish to a mixture of estrogenic chemicals. Environmental Health Perspectives 113 (6), 721e728. Brian, J.V., Harris, C.A., Scholze, M., Kortenkamp, A., Booy, P., Lamoree, M., Pojana, G., Jonkers, N., Marcomini, A., Sumpter, J.P., 2007. Evidence of estrogenic mixture effects on the reproductive performance of fish. Environmental Science & Technology 41 (1), 337e344. Brown, V.M., 1968. The calculation of the acute toxicity of mixture of poisons to rainbow trout. Water Research 2, 723e733. Buerge, I.J., Buser, H.R., Poiger, T., Muller, M.D., 2006. Occurrence and fate of the cytostatic drugs cyclophosphamide and ifosfamide in wastewater and surface waters. Environmental Science & Technology 40 (23), 7242e7250.
89
Cleuvers, M., 2004. Mixture toxicity of the anti-inflammatory drugs diclofenac, ibuprofen, naproxen, and acetylsalicylic acid. Environmental Toxicology and Pharmacology 59, 309e315. Crane, M., Watts, C., Boucard, T., 2006. Chronic aquatic environmental risks from exposure to human pharmaceuticals. Science of the Total Environment 367, 23e41. Deneer, J.W., 2000. Toxicity of mixtures of pesticides in aquatic systems. Pest Management Science 56, 516e520. Documed, 2009. Arzneimittelkompendium der Schweiz. Documed AG, Basel, Switzerland. http://www.kompendium.ch/. ECETOC, 2001. Aquatic Toxicity of Mixtures. European Centre for Ecotoxicology and Toxicology of Chemicals, Brussels, Belgium. EMEA, 2006. Guideline on the Environmental Risk Assessment of Medicinal Products for Human Use CHMP/SWP/4447/00. The European Agency for the Evaluation of Medicinal Products, London. Escher, B.I., Schwarzenbach, R.P., Westall, J.W.C., 2000. Evaluation of liposome-water partitioning of organic acids and bases: I. Development of sorption model. Environmental Science & Technology 34, 3954e3961. Escher, B.I., Eggen, R.I.L., Schreiber, U., Schreiber, Z., Vye, E., Wisner, B., Schwarzenbach, R.P., 2002. Baseline toxicity (narcosis) of organic chemicals determined by membrane potential measurements in energy-transducing membranes. Environmental Science & Technology 36, 1971e1979. Escher, B.I., Bramaz, N., Richter, M., Lienert, J., 2006. Comparative ecotoxicological hazard assessment of beta-blockers and their human metabolites using a mode-of-action based test battery and a QSAR approach. Environmental Science & Technology 40, 7402e7408. Escher, B.I., Baumgartner, R., Lienert, J., Fenner, K., 2009. In: Boxall, A.B.A. (Ed.), Reaction and Processes, Part P e Transformation Products of Synthetic Chemicals in the Environment, vol. 2. Springer, Berlin/Heidelberg, pp. 205e244. DOI:210.1007/1978-1003-1540-88273-88272. Escher, B., Schwarzenbach, R.P., 2002. Mechanistic studies on baseline toxicity and uncoupling as a basis for modeling internal lethal concentrations in aquatic organisms. Aquatic Sciences 64, 20e35. Escher, B.I., Sigg, L., 2004. In: Van Leeuwen, H.P., Ko¨ster, W. (Eds.), Physicochemical Kinetics and Transport at Biointerfaces. John Wiley & Sons, Chichester, pp. 205e271. Esperanza, M., Suidan, M.T., Marfil-Vega, R., Gonzalez, C., Sorial, G.A., McCauley, P., Brenner, R., 2007. Fate of sex hormones in two pilot-scale municipal wastewater treatment plants: conventional treatment. Chemosphere 66, 1535e1544. European Chemicals Agency, 2008. Guidance for the implementation of REACH, Guidance on information requirements and chemical safety assessment. Chapter R.6: QSARs and grouping of chemicals. http://echa.europa.eu/ reach_en.aspCorrigendum to Regulation (EC) No 1907/2006 of the European Parliament and of the Council of 18 December 2006 concerning the Registration, Evaluation, Authorisation and Restriction of Chemicals (REACH), establishing a European Chemicals Agency, amending Directive 1999/45/EC and repealing Council Regulation (EEC) No 793/93 and Commission Regulation (EC) No 1488/94 as well as Council Directive 76/769/EEC and Commission Directives 91/155/EEC, 93/67/EEC, 93/105/EC and 2000/21/EC (OJ L 396, 30.12.2006); amended by Council Regulation (EC) No 1354/2007 of 15 November 2007 adapting Regulation (EC) No 1907/2006 of the European Parliament and of the Council on the Registration, Evaluation, Authorisation and Restriction of Chemicals (REACH) by reason of the accession of Bulgaria and Romania (OJ L 304, 22.11.2007). European Commission, 2003. Technical Guidance Document in Support of Commission Directive 93/67/EEC on Risk Assessment for New Notified Substances, Commission
90
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 7 5 e9 2
Regulation (EC) No 1488/94 on Risk Assessment for Existing Substances, and Directive 98/8/EC of the European Parliament and of the Council Concerning the Placing of Biocidal Products on the Market. Office for Official Publications of the European Communities, Luxembourg. European Parliament and European Council, 2006a. REGULATION (EC) No 1907/2006 OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL of 18 December 2006 concerning the Registration, evaluation, Authorisation and Restriction of chemicals (REACH), establishing a European chemicals Agency, amending Directive 1999/45/EC and repealing Council regulation (EEC) No 793/93 and Commission regulation (EC) No 1488/94 as well as Council Directive 76/769/EEC and Commission Directives 91/155/EEC, 93/67/EEC, 93/105/EC and 2000/21/EC. Official Journal of the European Communities. European Parliament and European Council, 2006b. Regulation (EC) No 1907/2006 Reach, Criteria for the identification of persistent, bioaccumulative and toxic substances, and very persistent and very bioaccumulative substances. Official Journal of the European Communities ANNEX XIII. Ferrari, B., Mons, R., Vollat, B., Fraysse, B., Paxeus, N., Lo Giudice, R., Pollio, A., Garric, J., 2004. Environmental risk assessment of six human pharmaceuticals: are the current environmental risk assessment procedures sufficient for the protection of the aquatic environment? Environmental Toxicology and Chemistry 23 (5), 1344e1354. FOEN, 2009. Micropollutants in the Aquatic Environment. Assessment and Reduction of the Pollutant Load Due to Municipal Waste Water. Download. FOEN, Federal Office for the Environment Berne, Switzerland. http://www.bafu.admin. ch/publikationen/publikation/01051/index.html?lang¼en (accessed 05.10.2009). 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 (2e3), 361e371. 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 (15), 3243e3249. Gulkowska, A., Leung, H.W., So, M.K., Taniyasu, S., Yamashita, N., Yeunq, L.W.Y., Richardson, B.J., Lei, A.P., Giesy, J.P., Lam, P.K.S., 2008. Removal of antibiotics from wastewater by sewage treatment facilities in Hong Kong and Shenzhen, China. Water Research 42 (1e2), 395e403. Hallare, A.V., Ko¨hler, H.R., Triebskorn, R., 2004. Developmental toxicity and stress protein responses in zebrafish embryos after exposure to diclofenac and its solvent, DMSO. Chemosphere 56 (7), 659e666. ¨ kotoxikologische Hanisch, B., Abbas, B., Kratz, W., 2002. O Bewertung von Humanarzneimitteln in aquatischen ¨ kosystemen. Landesumweltamt Brandenburg, Potsdam, O Germany. Heinzmann, B., Schwarz, R.J., Schuster, P., Pineau, C., 2008. Decentralized collection of iodinated x-ray contrast media in hospitals e results of the feasibility study and the practice test phase. Water Science Technology 57, 209e215. Hermens, J., Leeuwangh, P., 1982. Joint toxicity of mixtures of 8 and 24 chemicals to the guppy (Poecilia reticulata). Ecotoxicology and Environmental Safety 6, 302e310. Hilal, S.H., Karickhoff, S.W., Carreira, L.A., 2005. SPARC e Performs Automated Reasoning in Chemistry. accessible at. University of Georgia, Athens, GA. http://ibmlc2.chem.uga. edu/sparc/. Hollender, J., Zimmermann, S.G., Koepke, S., Krauss, M., McArdell, C.S., Ort, C., Singer, H., Gunten, U.v., Siegrist, H., 2009. Elimination of organic micropollutants in a municipal wastewater treatment plant Upgraded with a full-scale Post-
ozonation followed by sand filtration. Environmental Science & Technology 43, 7862e7869. IMS, 2004. Hergiswil, Switzerland. Website: http://www.ihaims.ch
[email protected] (accessed 7.10.2009). 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) 3139-3152. Joss, A., Siegrist, H., Ternes, T.A., 2008. Are we about to upgrade wastewater treatment for removing organic micropollutants? Water Science and Technology 57 (2), 251e255. Kahle, M., Buerge, I.J., Hauser, A., Muller, M.D., Poiger, T., 2008. Azole fungicides: occurrence and fate in wastewater and surface waters. Environmental Science & Technology 42 (19), 7193e7200. 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 (2), 363e380. Kidd, K.A., Blanchfield, P.J., Mills, K.H., Palace, V.P., Evans, R.E., Lazorchak, J.M., Flick, R.W., 2007. Collapse of a fish population after exposure to a synthetic estrogen. Proceedings of the National Academy of Sciences of the United States of America 104 (21), 8897e8901. 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. Kolpin, D.W., Furlong, E.T., Meyer, M.T., Thurman, E.M., Zaugg, S. D., 2002. Pharmaceuticals, hormones and other organic contaminants in U.S. streams, 1999e2000: a national reconnaissance. Environmental Science & Technology 36, 1201e1211. Kortenkamp, A., 2002. Toxicology of mixtures. Toxicology 178 (1), 34e35. Kortenkamp, A., 2008. Low dose mixture effects of endocrine disrupters: implications for risk assessment and epidemiology. International Journal of Andrology 31 (2), 233e237. Kovalova, L., McArdell, C.S., Koller, M., Siegrist, H. Mass flow analysis of pharmaceuticals in a Swiss hospital, in preparation. Kummerer, K., 2001. Drugs in the environment: emission of drugs, diagnostic aids and disinfectants into wastewater by hospitals in relation to other sources e a review. Chemosphere 45 (6e7), 957e969. Ku¨mmerer, K., 2004. Pharmaceuticals in the Environment e Sources, Fate, Effects and Risks. Springer, Berlin Heidelberg. Labadie, P., Budzinski, H., 2005. Determination of steroidal hormone profiles along the Jalle d’Eysines River (near Bordeaux, France). Environmental Science & Technology 39, 5113e5120. Langford, K.H., Thomas, K.V., 2009. Determination of pharmaceutical compounds in hospital effluents and their contribution to wastewater treatment works. Environment International 35 (5), 766e770. Larsen, T.A., Lienert, J., Joss, A., Siegrist, H., 2004. How to avoid pharmaceuticals in the aquatic environment. Journal of Biotechnology 113, 295e304. Larsen, T.A., Alder, A.C., Eggen, R.I.L., Maurer, M., Lienert, J., 2009. Source separation: will we see a paradigm shift in wastewater handling? Environmental Science & Technology 43, 6121e6125. Lenz, K., Mahnik, S.N., Weissenbacher, N., Mader, R.M., Krenn, P., Hann, S., Koellensperger, G., Uhl, M., Knasmuller, S., Ferk, F., Bursch, W., Fuerhacker, M., 2007. Monitoring, removal and risk assessment of cytostatic drugs in hospital wastewater. Water Science and Technology 56 (12), 141e149.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 7 5 e9 2
Lienert, J., Bu¨rki, T., Escher, B.I., 2007a. Reducing micropollutants with source control: substance flow analysis of 212 pharmaceuticals in faeces and urine. Water Science and Technology 56 (5), 87e96. Lienert, J., Gu¨del, K., Escher, B.I., 2007b. Screening method for ecotoxicological hazard assessment of 42 pharmaceuticals considering human metabolism and excretory routes. Environmental Science & Technology 41, 4471e4478. Lienert, J., Larsen, T., 2010. High acceptance of urine source separation in seven European countries: a review. Environmental Science & Technology 44, 556e566. Lienert, J., Mosler, H.-J. Understanding stakeholders in a decision to reduce pharmaceuticals in hospital wastewater, in preparation. Loffler, D., Ternes, T.A., 2003. Analytical method for the determination of the aminoglycoside gentamicin in hospital wastewater via liquid chromatography electrospray-tandem mass spectrometry. Journal of Chromatography A 1000 (1e2), 583e588. Maeder, V., Escher, B.I., Scheringer, M., Hungerbu¨hler, K., 2004. Toxic ratio as an indicator of the intrinsic toxicity in the assessment of persistent, bioaccumulative, and toxic chemicals. Environmental Science & Technology 38, 3659e3666. Mahnik, S.N., Lenz, K., Weissenbacher, N., Mader, R.M., Fuerhacker, M., 2007. Fate of 5-fluorouracil, doxorubicin, epirubicin, and daunorubicin in hospital wastewater and their elimination by activated sludge and treatment in a membrane-bio-reactor system. Chemosphere 66 (1), 30e37. Maurer, M., Escher, B.I., Richle, P., Schaffner, C., Alder, A.C., 2007. Elimination of beta-blockers in sewage treatment plants. Water Research 41 (7), 1614e1622. Moser, R., McArdell, C.S., Weissbrodt, D., 2007. Micropollutants from urban drainage: pretreatment of hospital wastewater. GWA Gas, Wasser, Abwasser 11, 869e875. 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 (19), 4373e4382. Neuwoehner, J., Fenner, K., Escher, B.I., 2009. Physiological modes of action of fluoxetine and its human metabolites in algae. Environmental Science & Technology 43, 6830e6837. Nowotny, N., Epp, B., von Sonntag, C., Fahlenkamp, H., 2007. Quantification and modeling of the elimination behavior of ecologically problematic wastewater micropollutants by adsorption on powdered and granulated activated carbon. Environmental Science & Technology 41 (6), 2050e2055. Oaks, J.L., Gilbert, M., Virani, M., Watson, R.T., Meteyer, C.U., Rideout, B.A., Shivaprasad, H.L., Ahmed, S., Chaudry, M.J.I., Khan, A.A., 2004. Diclofenac residues as the cause of vulture population decline in Pakistan. Nature 427, 630e633. Ort, C., Lawrence, M.G., Reungoat, J., Eaglesham, G., Carter, S., Keller, J., 2010. Determining the fraction of pharmaceutical residues in wastewater originating from a hospital. Water Research 44 (2), 605e615. Owen, S.F., Giltrow, E., Huggett, D.B., Hutchinson, T.H., Saye, J., Winter, M.J., Sumpter, J.P., 2007. Comparative physiology, pharmacology and toxicology of beta-blockers: mammals versus fish. Aquatic Toxicology 82 (3), 145e162. Pauwels, B., Noppe, H., De Brabander, H., Verstraete, W., 2008. Comparison of Steroid hormone concentrations in Domestic and hospital wastewater treatment plants. Journal of Environmental Engineering-Asce 134 (11), 933e936. PhACT Database, 2006. Pharmaceutical Research and Manufacturers of America (PhRMA). Porsbring, T., Blanck, H., Tjellstrom, H., Backhaus, T., 2009. Toxicity of the pharmaceutical clotrimazole to marine microalgal communities. Aquatic Toxicology 91 (3), 203e211.
91
Prunet, P., Sturm, A., Milla, S., 2006. Multiple corticosteroid receptors in fish: from old ideas to new concepts. General and Comparative Endocrinology 147, 17e23. Radjenovic, J., Petrovic, M., Barcelo, D., 2009. Fate and distribution of pharmaceuticals in wastewater and sewage sludge of the conventional activated sludge (CAS) and advanced membrane bioreactor (MBR) treatment. Water Research 43 (3), 831e841. Raimondo, S., Montague, B.J., Barron, M.G., 2007. Determinants of variability in acute to chronic toxicity ratios for aquatic invertebrates and fish. Environmental Toxicology and Chemistry 26, 2019e2023. Raldua, D., Babin, P.J., 2009. Simple, Rapid zebrafish Larva bioassay for assessing the potential of chemical pollutants and drugs to disrupt thyroid Gland function. Environmental Science & Technology 43 (17), 6844e6850. Reungoat, J., Macova, M., Escher, B.I., Carswell, S., Mueller, J.F., Keller, J., 2010. Removal of micropollutants and reduction of biological adverse effects in a full scale reclamation plant using ozonation and activated carbon filtration. Water Research 44 (2), 625e637. Roberts, P.H., Thomas, K.V., 2006. The occurrence of selected pharmaceuticals in wastewater effluent and surface waters of the lower Tyne catchment. Science of the Total Environment 356 (1e3), 143e153. Roex, E.W.M., Van Gestel, C.A., Van Wezel, A.P., Van Straalen, N. M., 2000. Ratios between acute aquatic toxicity and effects on population growth rates in relation to toxicant mode of action. Environmental Toxicology and Chemistry 19, 685e693. Rosa, S., Antunes-Madeira, M., Jurado, A., Madeira, V., 2000. Amiodarone interactions with membrane lipids and with growth of Bacillus stearothermophilus used as a model. Applied Biochemistry and Biotechnology 87, 165e175. Routledge, E., Sheahan, D.A., Desbrow, C., Brighty, G.C., Waldock, M., Sumpter, J., 1998. Identification of estrogenic chemicals in STW effluent. 2. In vivo response introut and roach. Environmental Science and Technology 32, 1559e1565. Sanderson, H., Johnson, D.J., Reitsma, T., Brain, R.A., Wilson, C.A., Solomon, K.R., 2004. Ranking and prioritization of environmental risks of pharmaceuticals in surface waters. Regulatory Toxicology and Pharmacology 39, 158e183. Sanderson, H., Thomsen, M., 2009. Comparative analysis of pharmaceuticals versus industrial chemicals acute aquatic toxicity classification according to the United Nations classification system for chemicals. Assessment of the (Q)SAR predictability of pharmaceuticals acute aquatic toxicity and their predominant acute toxic mode-of-action. Toxicology Letters 187, 84e93. Schwarzenbach, R.P., Gschwend, P.M., Imboden, D.M., 2003. Environmental Organic Chemistry, second ed. Wiley, New York, NY, USA. 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. Sprague, J.B., 1970. Measurement of pollutant toxicity to fish. II Utilizing and applying bioassay results. Water Research 4, 3e32. SRU, 2007. Arzneimittel in der Umwelt. Stellungsnahme. Sachversta¨ndigenrat fu¨r Umweltfragen www.umweltrat.de, April 2007. Strenn, B., Clara, M., Gans, O., Kreuzinger, N., 2003. The comportment of selected pharmaceuticals in sewage treatment plants. Water Pollution Vii e Modelling, Measuring and Prediction 9, 273e282. Tarazona, J.V., Escher, B.I., Giltrow, E., Sumpter, J., Knacker, T., 2010. Targeting the environmental risk assessment of pharmaceuticals: facts and fantasies. Integrated
92
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 7 5 e9 2
Environmental Assessment and Management Accepted Article Online: Feb 1 2010, DOI: 2010.1002/ieam.2059. Ternes, T.A., 2000. Ru¨cksta¨nde von Arzneimitteln, Diagnostika und Antiseptika in Abwasser, Flu¨ssen und Grundwasser. Eine neue Herausforderung fu¨r die Wasserwirtschaft., Habilitation. Mainz University, Germany, Mainz. Thorpe, K.L., Gross-Sorokin, M., Johnson, I., Brighty, G., Tyler, C.R., 2006. An assessment of the model of concentration addition for predicting the estrogenic activity of chemical mixtures in wastewater treatment works effluents. Environmental Health Perspectives 114, 90e97. U.S.EPA, 2008. EPISuite Exposure Assessment Tools and Models. http://www.epa.gov/opptintr/exposure/pubs/episuite.htm. Vaes, W.H.J., Urrestarazu-Ramos, E., Hamwick, C., van Holstein, I., Blaauboer, B.J., Seinen, W., Verhaar, H.J.M., Hermens, J.L.M., 1997. Solid phase microextraction as a tool to determine membrane/water partition coefficients and bioavailable concentrations in in-vitro systems. Chemical Research in Toxicology 10, 1067e1072. Verhaar, H.J.M., van Leeuwen, C.J., Hermens, J.L.M., 1992. Classifying environmental pollutants. 1: structure-activity relationships for prediction of aquatic toxicity. Chemosphere 25, 471e491. Virtual Computational Chemistry Laboratory, 2009. ALOGPS2.1. http://www.vcclab.org. Warne, M.S.J., Hawker, D.W., 1995. The number of components in a mixture Determines whether Synergistic and Antagonistic or additive toxicity Predominate e the Funnel hypothesis. Ecotoxicology and Environmental Safety 31 (1), 23e28.
Watkinson, A.J., Murby, E.J., Costanzo, S.D., 2007. Removal of antibiotics in conventional and advanced wastewater treatment: implications for environmental discharge and wastewater recycling. Water Research 41 (18), 4164e4176. Watkinson, A.J., Murby, E.J., Kolpin, D.W., Costanzo, S.D., 2009. The occurrence of antibiotics in an urban watershed: from wastewater to drinking water. Science of the Total Environment 407 (8), 2711e2723. Weissbrodt, D., Kovalova, L., Pazhepurackel, V., Ort, C., Moser, R., Hollender, J., Siegrist, H., McArdell, C.S., 2009. Mass flows of Xrax contrast media and cytostatics in hospital wastewater. Environmental Science & Technology 43, 4810e4817. van Wezel, A.P., Opperhuizen, A., 1995. Narcosis due to environmental pollutants in aquatic organisms: residue-based toxicity, mechanisms, and membrane burdens. Critical Reviews in Toxicology 25, 255e279. Wick, A., Fink, G., Joss, A., Siegrist, H., Ternes, T.A., 2009. Fate of beta blockers and psycho-active drugs in conventional wastewater treatment. Water Research 43 (4), 1060e1074. Zeilinger, J., Steger-Hartmann, T., Maser, E., Goller, S., Vonk, R., Lange, R., 2009. Effecrs of synthetic gestagens in fish reproduction. Environmental Toxicology and Chemistry 28 (12), 2663e2670. Zuehlke, S., Duennbier, U., Lesjean, B., Gnirss, R., Buisson, H., 2006. Long-term comparison of trace organics removal performances between conventional and membrane activated sludge processes. Water Environment Research 78 (13), 2480e2486.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 9 3 e1 0 4
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Fate of aromatic hydrocarbons in Italian municipal wastewater systems: An overview of wastewater treatment using conventional activated-sludge processes (CASP) and membrane bioreactors (MBRs) Francesco Fatone*, Silvia Di Fabio, David Bolzonella, Franco Cecchi Department of Biotechnology, University of Verona, Strada Le Grazie 15, I-37134 Verona, Italy
article info
abstract
Article history:
We studied the occurrence, removal, and fate of 16 polycyclic aromatic hydrocarbons
Received 21 February 2010
(PAHs) and 23 volatile organic compounds (VOCs) in Italian municipal wastewater treat-
Received in revised form
ment systems in terms of their common contents and forms, and their apparent and actual
7 August 2010
removal in both conventional activated-sludge processes (CASP) and membrane bioreac-
Accepted 9 August 2010
tors (MBRs). We studied five representative full-scale CASP treatment plants (design
Available online 14 August 2010
capacities of 12 000 to 700 000 population-equivalent), three of which included MBR systems (one full-scale and two pilot-scale) operating in parallel with the conventional
Keywords:
systems. We studied the solideliquid partitioning and fates of these substances using both
Municipal wastewater systems
conventional samples and a novel membrane-equipped automatic sampler. Among the
Polycyclic aromatic hydrocarbons
VOCs, toluene, ethylbenzene, xylenes, styrene, 1,2,4-trimethylbenzene, and 4-chlor-
Volatile organic compounds
otoluene were ubiquitous, whereas naphthalene, acenaphthene, fluorene, and phenan-
Membrane bioreactor
threne were the most common PAHs. Both PAHs and aromatic VOCs had removal
Solideliquid partitioning
efficiencies of 40e60% in the headworks, even in plants without primary sedimentation. Mainly due to volatilization, aromatic VOCs had comparable removal efficiencies in CASP and MBRs, even for different sludge ages. MBRs did not enhance the retention of PAHs sorbed to suspended particulates compared with CASPs. On the other hand, the specific daily accumulation of PAHs in the MBR’s activated sludge decreased logarithmically with increasing sludge age, indicating enhanced biodegradation of PAHs. The PAH and aromatic VOC contents in the final effluent are not a major driver for widespread municipal adoption of MBRs, but MBRs may enhance the biodegradation of PAHs and their removal from the environment. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
The increasing worldwide contamination of freshwater systems by hundreds of anthropogenic chemicals is a key environmental problem (Schwarzenbach et al., 2006). To face this challenge, two priority research aims have been proposed:
(a) selection of the most serious target compounds based on their actual occurrence and toxicological concerns and (b) definition of the most appropriate water and wastewater treatment technologies to remove these compounds (FattaKassinos et al., 2010). Among the target nonconventional pollutants, oil-derived aromatic substances are hazardous
* Corresponding author. Tel./fax: þ39 045 802 7965. E-mail address:
[email protected] (F. Fatone). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.011
94
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 9 3 e1 0 4
both for the environment and for human health (An, 2004; Luch, 2005; Farhadian et al., 2008; CDC, 2009), and they are ubiquitous, since most originate from nonpoint sources such as internal combustion engines (Fernandez-Martinez et al., 2000). As a result, oil-derived aromatic compounds constitute a major concern for municipal wastewater treatment plants (WTPs), where their removal and final fate have been studied by several investigators who focused on individual WTPs (Blanchard et al., 2001; Busetti et al., 2006; Vogelsang et al., 2006; Manoli and Samara, 1999, 2008). Volatile organic compounds (VOCs) are a related class of organic compounds with a vapor pressure greater than 0.1 mm Hg at 20 C and 1 atm. These compounds are extensively used by many industries, and like PAHs, they can adversely affect both human health and the environment. VOCs are key ingredients in many consumer products such as fuels, paints, aerosols, cosmetics, disinfectants, refrigerants, and pesticides. Thus, they are often abundant in municipal wastewater (Barcelo´, 2004). In fact, the emissions of VOCs from WTPs have been studied since the 1980s, and aromatic VOCs typically account for more than 75% to the total VOC load (Namkung and Rittmann, 1987). The fate of aromatic VOCs at WTPs is a major concern because of their volatilization and the resulting safety risk for the plant’s operators. However, their bioaccumulation and biodegradation are also important factors that define the best approach to waste sludge treatment and disposal. On the other hand, the accumulation of polycyclic aromatic hydrocarbons (PAHs) in sewage sludge is an issue of major concern, together with the potential ecological impacts related to the potential use of these wastes as (for example) soil amendments (Villar et al., 2006; Cai et al., 2007). To date, there is insufficient knowledge to outline clear scenarios for waste treatment in the heterogeneous system of municipal WTPs. In practice, WTPs often receive both urban (mostly combined wastewater and rainfall runoff) wastewater and a number of additional (often variable) waste flows. The current Italian law (decree 152/06) states that liquid waste and industrial wastewater may both be collected by WTPs, so long as total hydrocarbon and aromatic organic solvent contents of lower than 10 and 0.4 mg L1, respectively. In addition, there is insufficient knowledge on whether membrane bioreactors (MBRs) are an appropriate technology for use in WTPs at scales that are representative of municipal WTPs (Cirja et al., 2008), even though this technology is being widely chosen both for treatment of industrial wastewater at the source and for centralized and decentralized treatment of municipal sewage (Fane and Fane, 2005; Judd, 2006; Lesjean and Huisjes, 2008). With particular reference to fuel-derived aromatic hydrocarbons, three recent large industrial references are operating petrochemical and refinery sites in Italy. Therefore, this appears to be a relatively new sector for MBR technology (Lesjean et al., 2009), and the technology is expected to effectively enhance the removal of aromatic contaminants produced by the oil industries. However, a number of recent papers have reported that the effectiveness of MBR technology in the removal of xenobiotics and persistent compounds is not sufficiently pronounced to serve as the sole justification for employing MBRs in municipal wastewater treatment (DeWever et al., 2007; Weiss and
Reemtsma, 2008). Moreover, even volatile compounds could be influenced in different ways by MBRs in which strong coarse-bubble aeration is used to scour the submerged membranes. The present study was part of a national research project aimed at identifying the most commonly occurring nonconventional organic pollutants in Italy and the benefits provided by the application of MBR technology, as the application of this technology is rapidly growing in WTPs. In particular, we present and discuss the occurrence, removal, and phase distribution of PAHs and aromatic VOCs in five full-scale WTPs, of which three include MBRs (one full-scale and two pilot-scale), operating as conventional activated-sludge process (CASP) or MBR plants, that are representative of typical Italian WTP scenarios. First, we evaluated the magnitude of the problem by focusing on the total contents and solideliquid partitioning of the aromatic compounds in sewage influents received by WTPs. Next, we discuss the performance of MBRs and compare this with the performance of CASP-based WTPs to provide insights into the potential advantages of MBR technology in urban wastewater treatment systems, in terms of the ability of MBRs to enhance the removal of PAHs and aromatic VOCs.
2.
Materials and methods
2.1.
The analyzed WTPs and the sampling equipment
To account for the heterogeneity of the municipal WTPs in Italy, we selected five representative WTPs in central and northern Italy (Table 1, Fig. 1) and monitored levels of many nonconventional organic contaminants in their influents and effluents. The criteria used to select the representative WTPs were based on: (1) the design (maximal) treatment capacity (from 12 000 to 700 000 population-equivalent); (2) the types of wastewater collected in the public sewer system; (3) the need for co-treatment of municipal liquid wastes (mainly sewage from septic tanks and municipal landfill leachate); (4) the types
Table 1 e Current influent flowrate and wastewater origin for the five municipal WTPs analyzed. WTP
Average influent flowrate (m3/d)
A
25 000
B Ca D
15 000 4900 þ 15,000 118 000
E
21 000
Rate of Main industrial activities in municipal the catchment area wastewater (%) w30
w100 w100 w60
w90
Chemical-petrochemical, pharmaceutical, agroindustry, metal plating, shipyard e e Chemical-petrochemical, oil refining, metal plating, shipyard, thermoelectric power plant Oil refining, metal plating
a 4900 and 15,000 to the MBR and CASP, respectively.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 9 3 e1 0 4
95
Fig. 1 e Block flow diagrams of the five analyzed publicly owned WTPs.
of biological processes (i.e., predenitrificationenitrification, extended oxidation, multizone biological nutrient removal, intermittent aeration of continuously fed bioreactors, MBRs). In addition, as we discuss in Section 2.2, we studied MBR technology in WTPs B, C, and D, where pilot-, demonstration-, or full-scale MBRs operated in parallel with full-scale CASPs. We analyzed the aromatic compounds by means of gas chromatographyemass spectrometry (GCeMS, Agilent technology
5975 inert Mass selective detector, Agilent technology 6890 N network GCs, Agilent technology 7683B series Injector, e O.I. analytical ECLIPSE 4660) according to the U.S. EPA methods (EPA 8270C/96 and EPA 8260B/96). We selected target aromatic compounds from the widely occurring BTEXS (benzene, toluene, ethylbenzene, xylenes, styrene) VOC group, the 16 PAHs recommended by the US-EPA, plus an additional 17 aromatic VOCs.
96
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 9 3 e1 0 4
To account for variability of the influents, we obtained samples distributed throughout the year (from 2005 to 2008), except in August, when industrial production and municipal inputs were drastically reduced. We obtained at least six daily composite samples of raw wastewater using conventional refrigerated samplers. We paid particular attention to the solideliquid partitioning, since it is well established that the particulate fraction may include the largest portion of the potential pollution load (Ashley et al., 2004). Therefore, it is crucial to determine the levels of pollutants associated with suspended solids in any evaluation of the transport and fate of target nonconventional contaminants in urban WTPs (Buzier et al., 2006). To provide sufficient reliability in the composite sampling operation, we complemented the classical samples by collecting daily composite and concentrated samples using a membraneequipped automatic sampler (Fig. 2) that we designed and engineered for this purpose. This sampler was equipped with a ZeeWeed 10 (GE-Water and Process Technology) submerged membrane module, so as to use the same polyvinylidene fluoride hollow-fiber membrane for this special sampling that was used for the actual wastewater treatment. To collect the composite samples, we used a timer-controlled feeding pump with the timing based on the specific local hourly variation in influent sewage flows. The permeate pump was controlled based on the wastewater level in the tank. The membrane-equipped sampler let us analyze both the composite and concentrated samples (2e12 g SS L1, instead of the 0.1e0.6 g SS L1 commonly analyzed in conventional composite samples from raw urban wastewater) of influent particulates based on a day-long ultrafiltration of some 400 to 700 L of raw wastewater. As a consequence, the calculation of the daily composite soluble fraction was based on sufficiently reliable data to account for the high variability in the characteristics of urban wastewater, including both short- and long-term fluctuations. Prior to analyzing the collected samples, we dried the influent solids and sewage sludge to a constant weight in an oven at 40 C and ground these materials in an agate mortar, followed by sieving to obtain particles smaller than 1 mm in diameter. Higher drying temperatures are not recommended due to possible volatilization of low-molecular-weight PAHs (those with two or three rings). The fraction smaller than 1 mm was stored at 4 C until analysis. As PAHs are easily photodegraded
(Dabestani and Ivanov, 1999), exposure to direct sunlight and other strong light was avoided during all steps of sample preparation, including extraction and storage of the extracts.
2.2.
The MBRs considered in this study
In addition to a full-scale MBR system at plant C, we observed two pilot-scale plants that were operated in parallel with the CASP-based WTPs at plants B and D. The pilot-scale MBRs were stainless-steel tanks with reaction volumes of 11 m3 (MBR-B; Fig. 3a) and 1.4 m3 (MBR-D; Fig. 3b) and. Both were equipped with industrial modules composed of submerged hollow-fiber membranes (manufactured by GE Process and Water Technologies; nominal poresize of 0.04 mm) and had membrane areas of 21.6 and 69.9 m2, respectively, which allowed them to treat real urban wastewater volumes of up to 24 and 75, respectively. Both the pilot MBRs were equipped with on-line meters that measured dissolved oxygen (DO), oxidation-reduction potential (ORP), pH, and mixed-liquor suspended solids (MLSS), and could operate in multizone treatment schemes with intermittent aeration in automatically controlled or sequencing batch reactors. Fatone et al. (2005, 2008) provide a full description of the pilot MBRs, and discuss the occurrence and removal of conventional pollutants and metals by the MBRs. The full-scale MBR (Plant C) was implemented by upgrading an existing municipal WTP, whose original construction dated back to the 1970s. Some of the urban wastewater (up to 6000) is treated in the membrane plant so that it can be reused for irrigation, and the remainder of the inflow (up to 15,000 m3 d1) is diverted to the conventional part of the plant. The ultrafiltration membrane has a membrane area of 12,130 m2; Fatone et al. (2007) provide further details.
3.
Results and discussion
3.1. Occurrence and liquidesolid partitioning of aromatic VOCs in influent Except for benzene, the rest of the BTEXS group (toluene, ethylbenzene, xylene, styrene) was the most commonly Backwashing line
Influent pump (timer-controlled) Pressure gauge Permeate line
Overflow weir ZW10 ultrafiltration membrane Permeate pump
Concentrated sample of suspended particulate
Fig. 2 e Membrane-equipped automatic sampler.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 9 3 e1 0 4
97
Fig. 3 e a) The pilot-scale MBR operating in parallel with the publicly owned WTP B; (b) The pilot MBR operating in parallel with the publicly owned WTP B.
occurring compounds, together with 1,2,4-trimethylbenzene and 4-chlorotoluene (Table 2). In particular, toluene was the most relevant compound because of its high level and because it is widely used as an industrial feedstock and a solvent to replace the more toxic and carcinogenic benzene. High BTEXS concentrations were expected since this is a well-known characteristic of diffuse sources such as vehicle emissions (e.g., exhaust, fuel evaporation). On the other hand, the high concentration of 1,2,4-trimethylbenzene was not expected, even though other researchers have reported that this compound is sometimes unexpectedly present in air at significant levels (Fernandez-Martinez et al., 2000). Along with toluene, 1,2,4-trimethylbenzene occurs naturally in crude oil and is not removed by oil refineries. Refineries pump this and other “unrecovered” substances to other facilities that recover the material and provide it for various uses, such as being added directly to gasoline to improve combustion. As 4chlorotoluene is a high-volume chemical that is widely used, even as a drain pipe solvent, it was found at higher levels in
pure municipal wastewater than in mixed municipal and industrial systems. Because aromatic VOCs are highly mobile, and are not strongly absorbed by various media such as suspended particulates, they are present primarily in the liquid phase of contaminated water (Zytner, 1994). In fact, even in urban wastewater, the fraction of VOCs associated with the suspended particulate matter was always under our detection limit (0.5 mg/kg TS).
3.2. Occurrence and solideliquid partitioning of PAHs in influent PAHs are lipophilic (i.e., hydrophobic) chemicals, and the larger compounds are poorly water-soluble and have lower volatility than smaller compounds. Because of these properties, many studies have reported that PAHs are adsorbed onto organic matter ranging from street particles to the waste activated sludge produced by the WTPs (Dobbs et al., 1989;
98
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 9 3 e1 0 4
Table 2 e Influent concentrations of 23 aromatic VOCs to five Italian municipal WTPs (average over six samples per plant). Parameter (mg/L)
Benzene Toluene m-Xylene þ p-Xylene o-xylene Styrene Ethylbenzene Chlorobenzene Isopropylbenzene Bromobenzene 2-Chlorotoluene n-Propylbenzene 4-Chlorotoluene 1,3,5-Trimethylbenzene 1,2,4-Trimethylbenzene Tert-butylbenzene 1,3-Dichlorobenzene Sec-butylbenzene 1,4-Dichlorobenzene p-Isopropyltoluene 1,2-Dichlorobenzene n-Butilbenzene 1,2,4-Trichlorobenzene
WTP A
WTP B
WTP C
WTP D
WTP E
Gasperi et al. (2008) Influent 100% municipal
Nikolaou al. (2002) Influent 100% municipal
0.21 3.544 0.568 0.035 0.148 0.238 0.020a 0.530a <0.005 0.058 5.567 0.05 0.053 0.168 0.08 0.04 <0.005 0.03 0.02 <0.005 0.020a <0.005
<0.005 4.885 0.15 0.192 0.139 0.093 0.120a 0.063a <0.005 0.208a <0.005 0.301 <0.005 1.018 <0.005 <0.005 <0.005 0.618 <0.005 <0.005 <0.005 <0.005
0.063 3.008 0.193 0.313a 0.226 0.132 0.031a 0.063a <0.005 0.118 0.110a 0.306 <0.005 0.764 <0.005 0.090a <0.005 1.054 0.098 <0.005 0.156 <0.005
0.239 7.169 0.775 0.351 1.116 0.216 0.349a <0.005 <0.005 <0.005 0.149 0.22 0.360a 0.81 <0.005 <0.005 <0.005 <0.005 0.56 <0.005 <0.005 <0.005
0.063a 3.008 0.15 0.035 0.139 0.093 0.020a 0.063a <0.005 0.058a 0.110a 0.05 0.053a 0.168 0.080a 0.040a <0.005 0.030a 0.02 <0.005 0.020a <0.005
<1.0 <1.0e3.2
<0.1e5.90 <0.05e13.70 <0.05e36.40 <0.05e0.60 <0.05e19.80
<0.1e0.8 <0.05e0.6 <0.25e1.60 <0.25e64.80
<0.05e0.20 <0.1e0.30 <0.1e18.10
a only one sample over the detection limit.
Stringfellow and Alvarez-Cohen, 1999). Table 3 shows the contents of the 16 PAHs that we monitored in the composite influent samples at the five urban WTPs. Table 3 shows that the maximum abundances of individual PAH compounds in raw urban wastewater were all well lower than 1.0 mg/L1 and the total values were all less than 2 mg/L1, even for large WTPs where most of the surrounding catchment area is urbanized. Most of the average concentrations were well below 0.5 mg/L1 and were <2 mg/L1 even at the highest level (WTP D). Napthalene was ubiquitous in the analyzed samples, and fluorene and phenanthrene were commonly found above the limit of detection (LOD) in the raw urban wastewaters. Napthalene is a commercially important aromatic hydrocarbon produced from coal tar and petroleum. It is widely used to manufacture phthalic and anthranilic acids, synthetic resins, lubricants, celluloid, lamp black, smokeless gunpowder, and hydronaphthalenes. Napthalene is also used in dusting powders, bathroom products, deodorant discs, wood preservatives, fungicides, and mothballs, and is used as an insecticide. It is possibly carcinogenic to humans (Group 2B) and toxic effects on animals and humans have been demonstrated (IARC, 2002). The available data are inadequate to permit an evaluation of the carcinogenicity of phenanthrene in experimental animals, whereas fluorene does not appear to be a human carcinogen (Group D). It is important to note that the toxicity of the PAHs in Table 3 depends on their structure, with different isomers varying from nontoxic to extremely toxic. PAHs that are well-known for their carcinogenic, mutagenic, and teratogenic properties include benzo[a]anthracene, chrysene, benzo[b]fluoranthene, B[j] FA, benzo[k]fluoranthene, benzo[a]pyrene, benzo[ghi]perylene, coronene, dibenz[a,h]anthracene, indeno[1,2,3-cd]pyrene, and
ovalene(Luch, 2005). These compounds were not the mostly commonly occurring. In fact, they were found with a frequency of occurrence lower than 50%, and their maximum concentrations were all well under 1 mg/L1. In terms of their solideliquid partitioning, PAHs are typically associated with particulate matter that is already present in the sewer system (Lau and Stenstrom, 2005; Mansuy-Huault et al., 2009). In agreement with the results in the literature (except for napthalene), the samples collected by the membrane-equipped sampler showed two things: First, the PAHs were mainly associated (65% or more) with suspended particulate matter in the influent that could be separated by membrane micro- or ultrafiltration (Fig. 4). Second, the higher the KP value of the PAH, which is expressed as a function of the compound octanolewater partitioning coefficient KOW (logKP ¼ 0.58 logKOW þ 1.14; reported by Dobbs et al., 1989) for sorption to solids in municipal sewage, the higher the solidbound fraction (Fig. 5). Taking into account the data from WTP D, where PAH levels were both stable and significant, Fig. 4 shows the typical partitioning of the analyzed compounds between the wastewater solid and liquid phases and the associated partitioning coefficient for the compound (logKP).
3.3. Removal of PAHs and aromatic VOCs in CASPbased WTPs and WTPs with MBRs The target aromatics undergo different removal mechanisms in urban treatment systems: VOCs are subject to volatilization (air stripping) and biodegradation or biotransformation, whereas PAHs are subject to (bio)sorption and biodegradation or biotransformation.
Table 3 e Influent concentrations of 16 PAHs recommended by the US-EPA influent to five Italian municipal WTPs.(average over six samples per plant). Parameter (mg/L)
WTP B
WTP C
WTP D
WTP E
0.25 0.030a 0.18 0.177 0.084 0.014 0.028 0.025 0.023 0.059 0.016a 0.032a 0.016a <0.005 <0.005 0.016a 0.71
0.096 0.017a 0.084a 0.058 0.052 <0.005 0.018 0.018 0.02 <0.005 <0.005 <0.005 <0.005 <0.005 <0.005 <0.005 0.22
0.113 <0.005 0.027a 0.008a 0.047 <0.005 <0.005 <0.005 0.008a <0.005 <0.005 <0.005 <0.005 <0.005 <0.005 <0.005 0.14
0.634 0.022 0.285 0.148 0.188 0.038 0.126 0.107 0.025 0.025 0.014 <0.005 0.014 <0.005 <0.005 0.020a 1.54
0.103 0.011 0.115 0.043 0.039 0.013 0.009 0.01 0.01 0.02 0.02 0.040a 0.020a 0.020a 0.020a 0.020a 0.32
Average standard deviation (mg/L)
0.24 0.02 0.14 0.09 0.08 0.01 0.04 0.03 0.02 0.02 0.01 0.02 0.01 0.01 0.01 0.01 0.76
0.23 0.01 0.10 0.07 0.06 0.01 0.05 0.04 0.01 0.02 0.01 0.02 0.01 0.01 0.01 0.01 0.57
Frequency of occurrence (%) (LOQ, mg/l) 95 32 56 71 85 22 44 44 42 43 22 7 22 4 4 12
(0.005) (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) (0.005)
Kolpin et al., 2002 Influent 100% municipal
Gasperi et al. (2008) Influent 100% municipal
0.02e0.04
<0.05
0.02e0.04 0.02e0.04 0.07 0.02e0.04 0.02e0.04
0.01e0.14 0.02e0.42 0.02e0.06 0.02e0.53 0.02e0.06 0.02e0.08 0.02e0.08 0.02e0.04 0.02e0.06 0.02e0.04 <0.02
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 9 3 e1 0 4
Naphthalene, Np Acenaphthylene, Acy Acenaphthene, Ace Fluorene, F Phenanthrene, Ph Anthracene, An Fluoranthene, Fl Pyrene, Py Benzo[a]anthracene, B[a]An Chrysene, Chry Benzo[b]fluoranthene,B[b]Fl Benzo[k]fluoranthene,B[k]Fl Benzo[a]pyrene, B[a]Py Indeno[1,2,3-cd]pyrene, I[1,2,3-cd]Py Dibenz[a,h]anthracene, dB[a,h]An Benzo[ghi]perylene, B[ghi]Pe P 16 PAHs
WTP A
0.02e0.07
a only one sample over the detection limit.
99
100
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 9 3 e1 0 4
Fig. 4 e Example of solid/liquid partition of the PAHs in a raw urban wastewater and dependence on the logKP.
3.3.1.
Removal and fate of aromatic VOCs
Generally, the removal of aromatic VOCs in the conventional WTPs was almost complete, and the secondary effluents often showed VOC concentrations below the limits of quantification (Table 4). Only toluene was still present in the effluents at significant levels, with concentrations of 0.75e2.7 mg/L and 0.5e1.9 mg/L, after CASP and MBR treatment, respectively. This is probably related to the high concentrations in the raw influent and to the low bioavailability of this compound, which has low volatility and low solubility in water (Nahar et al., 2000). It is important to note that despite the heterogeneity of the WTP framework, the aromatic VOC contents in the effluents were not clearly dependent on CASP or MBR technology. In particular, only the aeration method (microbubble diffusers) was used in all five WTPs. The plants also
Secondary effluent Espo. (Raw wastewater)
differed in their biological processes for the wastewater treatment line (see Fig. 1) and headworks (aerated grit chambers in WTPs A, D, and E, versus vortex-type grit chambers in WTPs B and C). In addition, the submerged MBRs operated with considerable coarse-bubble aeration for membrane scouring. Therefore, the final solideliquid separation might reasonably involve different additional removal mechanisms for aromatic VOCs: biosorption and biodegradation for the CASPs (with gravitational clarifiers) and air stripping and biodegradation for the MBRs (with membrane filtration). As expected, sorption was not a relevant removal mechanism. In fact, the concentration of aromatic VOCs in the sewage sludge was lower than the detection limit (0.5 mg/kg TS). Volatilization was likely to be the major removal mechanism for aromatic VOCs, and has previously been observed as
Primary effluent Espo. (Primary effluent)
Raw wastewater Espo. (Secondary effluent)
100
PAH fraction in the liquid (%)
90
-0,4398x
y = 141,25e 2 R = 0,8092
80 70 60
-1,2108x
y = 227,44e 2 R = 0,9465
50 40 30 20 -1,2973x
10
y = 100,78e 2 R = 0,8787
0 1
1,2
1,4
1,6
1,8
2
2,2
2,4
2,6
2,8
3
PAH classified by logKp @ 25°C
Fig. 5 e PAHs liquidesolid partitioning over the treatment stages in a representative full-scale CASP.
101
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 9 3 e1 0 4
Table 4 e Effluent concentrations and removal efficiencies of the mostly occurring aromatic VOCs in CASPs and MBRs (average over six samples per plant). Parameter (mg/L)
WTP A
WTP B
WTP C
WTP D
WTP E
MBR parallel to B
MBR parallel to C
MBR parallel to D
Benzene (mg/L) Removal (%) Toluene (mg/L) Removal (%) Ethylbenzene (mg/L) Removal (%) m-Xylene þ p-Xylene (mg/L) Removal (%) o-xylene (mg/L) Removal (%) Styrene (mg/L) Removal (%) 1.2.4-Trimethylbenzene (mg/L) Removal (%) 4-Chlorotoluene (mg/L) Removal (%)
0.06 71 <0.005 e 0.03 87 0.04 93 0.02 43 0.005 97 0.02
<0.005 e 2.7 44 <0.005 >95 0.08 47 0.19 1 0.28 na 0.71
<0.005 >92 0.84 72 0.13 2 <0.005 >97 0.17 46 0.005 98 <0.005
<0.005 >98 2.7 62 0.12 44 0.45 42 0.3 15 0.16 100 0.41
<0.005 >92 0.75 75 <0.005 >95 <0.005 >97 <0.005 >86 0.005 96 <0.005
<0.005 e 0.52 89 0.12 na 0.35 na 0.2 na 0.2 na 0.34
<0.005 >92 0.7 77 <0.005 >96 <0.005 >97 0.1 68 0.1 56 0.5
<0.005 >98 1.9 73 <0.005 >98 0.25 68 0.35 na 0.005 100 0.32
88 <0.005 >90
30 <0.005 >98
>99 <0.005 >98
49 0.1 55
>97 <0.005 >90
67 0.27 10
na <0.005 >98
na <0.005 >98
S BTEXS Removal S BTEXS
0.16 82
3.26 37
1.16 68
3.74 60
0.78 90
1.40 45
0.92 81
2.52 73
a consequence of the turbulence in the headworks, together with air stripping in the case of aerated grit chambers (Metcalf and Eddy, 2003). Our multiple sampling along the treatment line, the relative VOC removal effectiveness varied among the different sections of the system: 40e60% for the headworks and primary treatments, and 10e50% of the remaining VOCs (i.e., after removal by the headworks) for the secondary biological treatments. Because off-gassing control strategies are not commonly implemented in urban WTPs, the removal of aromatic VOCs by volatilization does not take advantage of a green technology such as compound transfer from the liquid to the gas phase (Farhadian et al., 2008). Table 4 also shows that when CASP systems were complemented by the addition of MBRs, the MBR technology did not provide any significant advantages in terms of VOC removal. As the concentration of aromatic VOCs in the sludge was always below the detection limit of 0.5 mg/kg TS, mass balance calculations were not possible, and we could not quantify the roles of the different removal mechanisms. However, the xylenes can be used to evaluate the removal mechanisms within the bioprocesses because of their chemical and physical properties. Of these compounds, o-xylene is the least volatile, but p-xylene is the most difficult for microorganisms to detoxify. Given the removal efficiencies of the biological processes, o-xylene was the most persistent VOC. Therefore, volatilization and air stripping were likely the main removal mechanisms in both the CASPs and the MBRs, and no additional significant effect was linked to the major turbulence generated in the strongly aerated filtration chamber.
3.3.2.
Removal and fate of PAHs
Due to their lipophilic and hydrophobic natures, PAHs tend to adsorb on particulate organic matter and their removal from the final effluent should be enhanced by advanced
solideliquid separation mechanisms such as MBRs. However, the WTP effluents analyzed in this study showed PAH contents in the permeates produced by the parallel MBR systems (Table 5) suggesting that these compounds are adsorbed to suspended particles that are already well separated by the conventional gravitational clarifiers. During the primary and secondary treatment stages, the PAHs were removed at rates of 40e60% and 10e90% of the remaining PAHs (i.e., after removal by the headworks), respectively. These results agree with those of other studies of full-scale WTPs (Manoli and Samara, 2008), in which PAH removal ranged from 28 to 67% in the primary treatments, including primary sedimentation. In addition, these results confirmed our expectations, since PAHs adhere significantly to street particles, which are usually removed in the conventional headworks of a WTP. On the other hand, napthalene was almost completely volatilized in the headworks due to the amount of water turbulence. Among the five WTPs we analyzed, only WTP D had sufficiently high PAH contents that we could evaluate the final fate of the PAHs. With a design (maximal) treatment capacity of 400 000 population-equivalent, WTP D0 s conventional flow scheme included headworks (gross and fine sieving, an aerated grit chamber), secondary biological treatment (conventional predenitrification and nitrification, activated sludge reactors, and secondary gravitational clarifiers), and final disinfection with peracetic acid. The solideliquid partitioning of the PAHs in the different treatment stages of a full-scale WTP was significantly linked to the hydrophobicity of each polyaromatic compound, which was given by the KP values (Fig. 5). Since urban wastewater usually takes hours to reach the WTPs, the interactions of PAHs with suspended particulates are likely to be in equilibrium by the time the wastewater reaches the headworks. Therefore, sorption equilibrium is likely to
102
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 9 3 e1 0 4
Table 5 e Effluent concentrations and removal efficiencies of the 16 PAHs recommended by the US-EPA in CASPs and MBRs (average over six samples per plant). Parameter (mg/L) Np (mg/L) Removal (%) Acy (mg/L) Removal (%) Ace (mg/L) Removal (%) F (mg/L) Removal (%) Ph (mg/L) Removal (%) An (mg/L) Removal (%) Fl (mg/L) Removal (%) Py (mg/L) Removal (%) B[a]An (mg/L) Removal (%) Chry (mg/L) Removal (%) B[b]Fl (mg/L) Removal (%) B[k]Fl (mg/L) Removal (%) B[a]Py (mg/L) Removal (%) I[1.2.3-cd]Py (mg/L) Removal (%) dB[a.h]An (mg/L) Removal (%) B[ghi]Pe (mg/L) Removal (%) P 16 PAHs (mg/L) P Removal 16 PAHs (%)
WTP A
WTP B
WTP C
WTP D
WTP E
MBR parallel to B
MBR parallel to C
MBR parallel to D
0.056 78 <0.005 >83 <0.005 >97 0.019 89 0.058 31 <0.005 >64 0.017 39 0.017 32 0.013 43 0.015 75 <0.005 >69 <0.005 >84 <0.005 69 <0.005 e <0.005 e <0.005 >68.75
0.073 24 0.015 12 0.025 70 <0.005 >91 0.014 73 <0.005 e <0.005 >72 0.012 33 <0.005 >75 <0.005 e <0.005 e <0.005 e <0.005 e <0.005 e <0.005 e <0.005 e
0.043 62 <0.005 e <0.005 >81 <0.005 >38 <0.005 >89 <0.005 e <0.005 e <0.005 e 0.007 13 <0.005 e <0.005 e <0.005 e <0.005 e <0.005 e <0.005 e <0.005 e
0.074 88 0.011 50 0.023 92 0.011 93 0.012 94 <0.005 >87 0.011 91 0.015 86 0.011 56 <0.005 >80 <0.005 >64 <0.005 e <0.005 >64 <0.005 e <0.005 e <0.005 >75
0.037 64 <0.005 >55 0.02 83 0.016 63 0.01 74 <0.005 >62 <0.005 >44 <0.005 >50 <0.005 >50 <0.005 >75 <0.005 >75 <0.005 >88 <0.005 >75 <0.005 >75 <0.005 >75 <0.005 >75
0.042 56 <0.005 >71 0.014 83 <0.005 >91 0.012 77 0.007 na 0.048 na 0.048 na 0.007 65 <0.005 e <0.005 e <0.005 e <0.005 e <0.005 e <0.005 e <0.005 e
0.054 52 <0.005 e <0.005 >81 <0.005 >38 0.008 83 <0.005 e <0.005 e <0.005 e <0.005 >38 <0.005 e <0.005 e <0.005 e <0.005 e <0.005 e <0.005 e <0.005 e
0.035 94 <0.005 >77 0.03 89 0.009 94 0.022 88 <0.005 >87 0.017 87 0.013 88 <0.005 >80 <0.005 >80 <0.005 >64 <0.005 e <0.005 >64 <0.005 e <0.005 e <0.005 >75
<0.195 73
<0.139 37
<0.05 64
<0.168 89
<0.083 74
<0.178 19
<0.062 56
<0.126 92
exist within all the analyzed treatment stages, characterized by decreasing bulk PAH concentrations. The correlation between the hydrophobicity of the PAH compounds and their solideliquid partitioning was clear, and values in the liquid fraction were stably higher than 50% only in the secondary P effluent, where the bulk concentration of the PAHs was less than 0.2 mg/L1. In addition, the contents of the most toxic PAHs at
Fig. 6 e PAHs daily specific accumulation (DSA) and solids retention time (SRT) in MBRs (data from the pilot-scale MBR-D).
this treatment stage were below the LOD (data points with logKP greater than 1.8 in Fig. 5). Therefore, the PAH content in the liquid fraction depends on the bulk concentrations and on the particular sorbent. As expected based on the organic content of the sorbents, Fig. 5 confirms that the sorption affinity of the coupled PAH and activated sludge is higher than that of the coupled PAH and suspended particulate matter. It is also important to note that since residual PAHs are dissolved in the secondary effluent, there would be little advantage to incorporating the membrane systems used in MBRs. In fact, the full-scale and pilot MBR systems operating in parallel produced analogous removal levels (80e95%). The two systems operated with similar hydraulic retention times, but the solid retention time was 12 days for the CASP system, and ranged from 200 to more than 500 days for the MBR systems. However, the impact of high sludge age must be considered when evaluating the biodegradation of PAHs (i.e., the actual removal from a contaminated flow). Owing to the high variability of PAH concentrations in the influents and the amount of available data, the concentration of PAHs in the waste activated sludge represents the system’s “historical memory”, and provides
103
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 9 3 e1 0 4
Table 6 e PAHs accumulation and activated sludge characteristics of a CASP and the parallel MBR.
Sludge age MLVSS/MLSS sOUR P PAHsa Daily Specific Accumulation of PAHs (DSAPPAHs)a
days % mgO2/gVSSaL mg/kgTS mgPAHs/kgTSad
CASP
MBR-Run1
MBR-Run2
MBR-Run3
MBR-Run4
12 80 17 1.6e1.8 0.133333e0.150000
12 80 17 1.7 0.141667
200 63 14 1.13 0.00565
1400 53 9 1.2 0.000857
1900 53 8 1.29 0.000679
a Sum of B[a]An, B[a]Py, B[b]Fl, B[k]Fl, B[ghi]Pe, Chry, dB[a,h]An, I[1,2,3-cd]Py.
the most reliable way to evaluate the typical biodegradation potentials of the two systems. Given that the variation in influent PAH levels and hydraulic retention times were equal in the CASP and MBR systems, Fig. 6 shows a logarithmic relationship between the daily specific accumulation of PAHs in the activated sludge and the sludge age (i.e., the solid retention time). This suggests that a high solid retention time improved the elimination of PAHs, but that this was partly compensated for by the lower biological activity. This is demonstrated by two observations (Table 6): (1) the volatile to total solids ratio decreased with increasing retention time in the MBR (from 80 to 52%) and (2) the maximum specific oxygen utilization rate (sOUR; Spanjers et al., 1996) decreased from 17 to 8 mg O2/(g VSS-h).
-
retention times), and showed a logarithmic relationship with sludge age. Both PAHs and aromatic VOCs had removal efficiencies of 40e60% in the headworks, even in plants without primary sedimentation. This demonstrated that PAHs adhere to removed street particles and that VOCs easily volatilize as a result of water turbulence in the headworks.
Acknowledgments This research was supported by the Italian Ministry of University and Research (projects PRIN 2003 and PRIN 2005).
references
4.
Conclusions
In this study, we reviewed five typical Italian scenarios for the occurrence, treatment, and fate of aromatic organic compounds (23 VOCs and 16 PAHs) in municipal conventional activated-sludge plants and membrane bioreactors. Our results suggest the following main conclusions: -
-
Among the aromatic VOCs, toluene, ethylbenzene, xylenes, styrene, 1,2,4-trimethylbenzene, and 4-chlorotoluene were most common. However, toluene showed the highest concentrations, ranging from 3 to 7 mg/L; all the others had concentrations less than 1 mg/L. The removal of aromatic VOCs was comparable in the conventional and membrane systems, despite the strong aeration used for membrane scouring in the submerged MBRs. The biodegradation was not significantly influenced by the high sludge age, probably due to the high water-solubility of these compounds; this differs from the PAHs, which were more efficiently biodegraded. Except for naphthalene (with 95% occurrence in the raw urban wastewater), only acenaphthene, fluorene, and phenanthrene occurred at detectable levels in more than 50% of the samples; however, the concentration of any single compound except napthalene was generally less than 0.3 mg/L, and more than 60% of these substances were associated with the suspended particulate matter. The apparent removal levels of PAHs in conventional CASPbased plants and MBRs were comparable, but the actual removals, which were related to biodegradation of the PAHs, were enhanced by long sludge age (i.e., long solid
An, Y.J., 2004. Toxicity of benzene, toluene, ethylbenzene, and xylene (BTEX) mixtures to Sorghum bicolor and Cucumis sativus. Bull. Environ. Contam. Toxicol. 72, 1006e1011. Ashley, R.M., Bertrand-Krajewsky, J.L., Hvitved-Jacobsen, T., Verbanck, M., 2004. Solids in Sewers: Characteristics, Effects and Control of Sewer Solids and Associated Pollutants. Sci.. and Tech. Rep. No. 14. IWA Publishing, London. Barcelo´, D. (Ed.), 2004, Emerging Organic Pollutants in Waste Waters and Sludge, vol. 1. Springer, London. Blanchard, M., Teil, M.J., Ollivon, D., Garban, B., Cheste´rikoff, C.C., Chevreuil, M., 2001. Origin and distribution of polyaromatic hydrocarbons and polychlorobiphenyls in urban effluents to wastewater treatment plants of the Paris area (France). Water Res. 35, 3679e3687. Busetti, F., Heitz, A., Cuomo, M., Badoer, S., Traverso, P., 2006. Determination of sixteen polycyclic aromatic hydrocarbons in aqueous and solid samples from an Italian wastewater treatment plant. J. Chromatogr. A 1102, 104e115. Buzier, R., Tusseau-Vuillemin, M.H., dit Meriadec, C.M., Rousselot, O., Mouchel, J.M., 2006. Trace metal speciation and fluxes within a major French wastewater treatment plant: impact of the successive treatments stages. Chemosphere 65, 2419e2426. Cai, Q.Y., Mo, C.H., Wu, Q.T., Zeng, Q.Y., Katsoyiannis, A., 2007. Occurrence of organic contaminants in sewage sludges from eleven wastewater treatment plants, China. Chemosphere 68, 1751e1762. CDC (Centers for Disease Control and Prevention e Department of Health and Human Services USA), 2009. Fourth National Report on Human Exposure to Environmental Chemicals. CDC, Druid Hills, GA. Cirja, M., Ivashechkin, P., Schaffer, A., Corvini, P.F.X., 2008. Factors affecting the removal of organic micropollutants from wastewater in conventional treatment plants (CTP) and membrane bioreactors (MBR). Rev. Environ. Sci. Biotechnol. 7, 61e78.
104
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 9 3 e1 0 4
Dabestani, R., Ivanov, I.N., 1999. A compilation of physical, spectroscopic and photophysical properties of polycyclic aromatic hydrocarbons. J. Photochem. Photobiol. 70, 10e34. DeWever, H., Weiss, S., Reemtsma, T., Vereecken, J., Muller, J., Knepper, T., Rorden, O., Gonzalez, S., Barcelo`, D., Hernando, M.D., 2007. Comparison of sulfonated and other micropollutants removal in membrane bioreactor and conventional wastewater treatment. Water Res. 41, 935e945. Dobbs, R.A., Wang, L., Govind, R., 1989. Sorption of toxic organic compounds on wastewater solids: correlation with fundamental properties. Environ. Sci. Technol. 23, 1092e1097. Fane, A.G., Fane, S.A., 2005. The role of membrane technology in sustainable decentralized wastewater systems. Water Sci. Technol. 51, 317e325. Farhadian, M., Duchez, D., Vachelard, C., Larroche, C., 2008. Monoaromatics removal from polluted water through bioreactorsda review. Water Res. 42, 1325e1341. Fatta-Kassinos, D., Bester, K., Ku¨mmerer, K. (Eds.), 2010, Xenobiotics in the Urban Water Cycle: Mass Flows, Environmental Processes, Mitigation and Treatment Strategies, Series Environmental Pollution, vol. 16. Springer Science þ Business Media, London, p. 494. Fatone, F., Bolzonella, D., Battistoni, P., Cecchi, F., 2005. Removal of nutrients and micropollutants treating low loaded wastewaters in a membrane bioreactor operating the automatic alternate-cycles process. Desalination 183, 395e405. Fatone, F., Battistoni, P., Pavan, P., Cecchi, F., 2007. Operation and maintenance of full-scale municipal membrane biological reactors: a detailed overview on a case study. Ind. Eng. Chem. Res. 46, 6688e6695. Fatone, F., Eusebi, A.L., Pavan, P., Battistoni, P., 2008. Exploring the potential of membrane bioreactors to enhance metals removal from wastewater: pilot experiences. Water Sci. Technol. 57, 505e511. Fernandez-Martinez, G., Lopez-Mahia, P., Muniategui-Lorenzo, S., Prada-Rodriguez, D., Fernandez-Fernandez, E., 2000. Measurement of volatile organic compounds in urban air of La Coruna, Spain. Water Air Soil Pollut. 129 (3), 267e288. Gasperi, J., Garnaud, S., Rocher, V., Moilleron, R., 2008. Priority pollutants in wastewater and combined sewer overflow. Sci. Total. Environ. 407, 263e272. IARC (International Agency for Research on Cancer), 2002. Naphthalene. IARC Summary & Evaluation, vol. 82. IARC, Lyon, France. 367. Judd, S.J., 2006. The Mbr Book: Principles and Applications of Membrane Bioreactors in Water and Wastewater Treatment. Elsevier Science, Amsterdam. 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. Environ. Sci. Technol. 36 (6), 1202e1211. Lau, S.L., Stenstrom, M.K., 2005. Metals and PAHs adsorbed to street particles. Water Res. 39, 4083e4092.
Lesjean, B., Huisjes, E.H., 2008. Survey of the European MBR market: trends and perspectives. Desalination 231, 71e81. Lesjean, B., Ferre, V., Vonghia, E., Moeslang, H., 2009. Market and design considerations of the 37 larger MBR plants in Europe. Desalin. Water Treat. 6, 227e233. Luch, A., 2005. The Carcinogenic Effects of Polycyclic Aromatic Hydrocarbons. Imperial College Press, London. Manoli, E., Samara, C., 1999. Occurrence and mass balance of polycyclic aromatic hydrocarbons in the Thessaloniki sewage treatment plant. J. Environ. Qual. 28, 176e187. Manoli, E., Samara, C., 2008. The removal of polycyclic aromatic hydrocarbons in the wastewater treatment process: experimental calculations and model predictions. Environ. Pollut. 151, 477e485. Mansuy-Huault, L., Regier, A., Faure, P., 2009. Analyzing hydrocarbons in sewer to help PAH source apportionment in sewage sludge. Chemosphere 75, 995e1002. Metcalf and Eddy, 2003. Wastewater Engineering: Treatment and Reuse, fourth international ed. McGraw-Hill, New York. Nahar, N., Alauddin, M., Quilty, B., 2000. Toxic effects of toluene on the growth of activated sludge bacteria. World J. Microbiol. Biotechnol. 16, 307e311. Namkung, E., Rittmann, B.E., 1987. Estimating volatile organic compound emissions from publicly owned treatment works. J. Water Pollut. Control Fed. 59 (7), 670e678. Nikolaou, A.D., Golfinopoulos, S.K., Kostopoulou, M.N., Kolokythas, G.A., Lekkas, T.D., 2002. Determination of volatile organic compounds in surface waters and treated wastewater in Greece. Water Res. 36, 2883e2890. Schwarzenbach, R.P., Escher, B.I., Fenner, K., Hofstetter, T.B., Johnson, C.A., von Gunten, U., Bernhard, W., 2006. The challenge of micropollutants in aquatic systems. Science 313, 1072e1077. Spanjers, H., Vanrolleghem, P., Olsson, G., Dold, P., 1996. Respirometry in control of the activated sludge process. Water Sci. Technol. 34 (3e4), 117e126. Stringfellow, W.T., Alvarez-Cohen, L., 1999. Evaluation the relationship between the sorption of PAHs to bacterial biomass and biodegradation. Water Res. 33, 2535e2544. Villar, P., Callejon, M., Alonso, E., Jimenez, J.C., Guiraum, A., 2006. Temporal evolution of polycyclic aromatic hydrocarbons (PAHs) in sludge from wastewater treatment plants: comparison between PAHs and heavy metals. Chemosphere 64, 535e541. Vogelsang, C., Grung, M., Jantsch, T.G., Tollefsen, K.E., Liltved, H., 2006. Occurrence and removal of selected organic micropollutants at mechanical, chemical and advanced wastewater treatment plants in Norway. Water Res. 40, 3559e3570. Weiss, S., Reemtsma, T., 2008. Membrane bioreactors for municipal wastewater treatment e A viable option to reduce the amount of polar pollutants discharged into surface waters? Water Res. 42, 3837e3847. Zytner, R.G., 1994. Sorption of benzene, toluene, ethylbenzene and xylenes to various media. J. Hazard. Mater. 38, 113e126.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 0 5 e1 1 2
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Influence of Ca2D and Suwannee River Humic Acid on aggregation of silicon nanoparticles in aqueous media Xuyang Liu a, Mahmoud Wazne a,*, Tsengming Chou b, Ru Xiao a, Shiyou Xu c a
Keck Geotechnical/Geoenvironmental Laboratory, Center for Environmental Systems, Stevens Institute of Technology, Hoboken, NJ 07030, United States b Laboratory for Multiscale Imaging, Stevens Institute of Technology, Hoboken, NJ 07030, United States c Department of Mechanical Engineering, Stevens Institute of Technology, Hoboken, NJ 07030, United States
article info
abstract
Article history:
Silicon nanoparticles (NPs) have potential applications in many fields including microelec-
Received 20 May 2010
tronics, biomedical imaging, and most recently energetics. Even though silicon NPs are
Received in revised form
thought to be harmless, their full impact on the environment and human health needs
15 July 2010
further investigation due to their potential increased use and recent toxicity data. Various
Accepted 10 August 2010
techniques were used to characterize silicon NPs that are being considered for use in ener-
Available online 19 August 2010
getics. These techniques included dynamic light scattering (DLS), electron microscopy (EM), X-ray diffraction (XRD) and atomic force microscopy (AFM). Experiments were also con-
Keywords:
ducted on the early stage aggregation kinetics of silicon NPs in the presence of Ca2þ and
Silicon nanoparticle
Suwannee River Humic Acid (SRHA). The addition of SRHA in the presence of Ca2þ resulted in
Aggregation
increased attachment efficiencies and decreased critical coagulation concentration (from 0.4
Humic acid
to 0.1 M). The enhanced aggregation was attributed to bridging generated by SRHA aggre-
Bridging effect
gates as evidenced by selected area electron diffraction (SAED) and energy dispersive spec-
Transport
troscopy (EDS). SAED verified the bridging to be amorphous phase comprised of humic
Natural organic matter
substances rather than artifacts of silicon crystallites. Element distribution analyses were
Electron microscopy
also used in the delineation of the silicon NP aggregation mechanisms in the absence and presence of SRHA. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Silicon nanoparticles (NPs) have wide potential application in many fields including microelectronics, biomedical imaging, and most recently energetics. Because of their robust and flexible surface chemistry, silicon NPs facilitate the conjugation of DNA or protein probes in cell biology and medicine (O’Farrell et al., 2006). Moreover, luminescent silicon NPs are used in imaging of tumor and organs (Park et al., 2009). There have been many reports on the use of silicon NPs in electronic devices, secondary batteries, super capacitors and solar cells
for energy devices. For example, silicon ultra thin films were demonstrated to produce large voltage enhancement in solar cells (Kim et al., 2009; Stupca et al., 2007; Tian et al., 2007). Most recently silicon particles are being considered for use in energetics because they are thought to increase metal pushing and blast energy. Even though silicon NPs are thought to be harmless, their full impact on the environment and human health needs further investigation due to their potential increased use and recent toxicity data. For example, particles with a diameter of 3 nm at concentrations >20 mg/L, and particles with diameters
* Corresponding author. Tel.: þ1 201 216 8993; fax: þ1 201 216 8212. E-mail address:
[email protected] (M. Wazne). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.022
106
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 0 5 e1 1 2
between 100 nm and 3000 nm at concentrations >200 mg/L increased cytotoxicity as compared with controls (Choi et al., 2009). Similarly, silicon NPs at a concentration of 112 mg/L induced membrane damage of HeLa cells and a decrease of hepatocytes viability (Fujioka et al., 2008). Due to concerns about the potential release of engineered NPs to the environment, researchers have initiated and conducted many studies on the implications of engineered NPs in the environment. The environmental behavior of carbon based NPs such as C60 fullerene and carbon nanotubes have been extensively studied (Brant et al., 2006; Chen and Elimelech, 2006, 2008; Kang et al., 2009; Saleh et al., 2008; Wang et al., 2008). Titanium oxide, zinc oxide, cerium oxide, silver and gold NPs have also been the interests of environmental scientists due to their availability in the market place and wide application in cosmetics, pharmaceuticals, catalysts, and antimicrobial products. (AlvarezPuebla et al., 2007; Benn and Westerhoff, 2008; Diegoli et al., 2008; Domingos et al., 2009b, Geranio et al., 2009; Keller et al., Limbach et al., 2008; Pallem et al., 2009; Scheckel et al.) However, due to the different behavior of engineered NPs, their transport varies significantly in aqueous environments. For instance, Lecoanet et al. (2004) investigated the deposition behavior of various NPs (including nC60, silica, anatase, and single-walled carbon nanotubes) and observed different breakthrough curves under similar experimental conditions (Lecoanet and Wiesner, 2004). Moreover, various values of NaCl critical coagulation concentrations for multiwalled carbon nanotubes were observed due to the different properties of these carbon nanotubes (including different sources and pretreatment process) (Saleh et al., 2008; Smith et al., 2009). Therefore, a case-by-case study may be more appropriate to assess the aggregation and transport behavior of NPs in the environment (Lecoanet et al., 2004; Nowack and Bucheli, 2007). Of special interest are the effects of natural organic matter (NOM) on the transport behavior of NPs. NOM are ubiquitous in natural environments and are expected to interact with engineered NPs once they are released into the environment. Humic substances (HS) are naturally occurring NOM and comprise the majority of the organic carbon in any freshwater (Steinberg et al., 2008). It was recently reported that HS can make the transport of NPs very complicated (Ghosh et al., 2008; Saleh et al., 2008; Yang et al., 2009), especially in the presence of divalent ions, such as Ca2þ (Chen and Elimelech, ˜ ¡cz, 2007). For 2007, 2008; Liu et al., 2010; Majzik and TombA example, Chen et al. reported enhanced aggregation of fullerene NPs in the presence of Suwannee River Humic Acid (SRHA) and Ca2þ due to bridging (Chen and Elimelech, 2007). HA and Ca2þ were also reported to influence the dispersion and aggregation of montmorillonite particles in aqueous ˜ ¡cz, 2007). Moreover, boron suspensions (Majzik and TombA NPs were observed to stabilize at low Ca2þ concentration in the presence of SRHA, while the aggregation for boron NPs was enhanced in the presence of alginate and high Ca2þ concentration (Liu et al., 2010). However, experimental studies to delineate the mechanisms of NP aggregation in the presence of NOM are lacking. In this paper, we report on the aggregation behavior of silicon NPs that are being considered for use in energetic
applications. Dynamic light scattering (DLS) were used to investigate the aggregation kinetics of silicon NPs in the presence of calcium ions and SRHA, whereas transmission electron microscopy (TEM), selected area electron diffraction (SAED) and energy dispersive spectroscopy (EDS) analyses were used to delineate the aggregation mechanisms. To the best knowledge of the authors’, this paper is the first one to apply these state of the art experimental approaches to discriminate the bridging effect of NOM on the aggregation of NPs in aqueous environment.
2.
Materials and methods
2.1.
Silicon NP dispersion preparation
Silicon NP (American Elements, Los Angeles, CA) dispersions (10 mg/L) were ultrasonicated in an ultrasonic bath (Fisher Scientific) twice for a cycle of 30 min each to breakup aggregates. Between the two sonication periods, the dispersions were mixed with a magnetic stirrer for 15 min to homogenize the dispersion and to improve the efficiency of the subsequent ultrasonication. Silicon NPs were allowed to settle down and sub-samples were removed from the supernatant at various time intervals to assess the stability of the dispersion. Silicon concentration was measured using inductively coupled plasma-atomic emission spectroscopy analysis (ICP-AES, Vista-MPX, Varian, Palo Alto, CA). It was found that silicon concentration in the liquid phase decreased during the first 24 h and then kept stable (Fig. S1, Supplementary data). Therefore, stable supernatant was carefully separated after 96 h for use in the aggregation experiments (Ghosh et al., 2008). All experiments and measurements were conducted at pH 4.2 0.1, except where noted.
2.2.
Solution chemistry
Electrolyte (CaCl2) stock solutions were prepared using analytical reagents (Fisher Scientific) and were filtered through 0.2 mm filters (Whatman Inc., Clifton, NJ) before use. The SRHA (standard II, International Humic Substances Society) solutions were made by dissolving 22.9 mg SRHA standard II into 50 mL DI water and stirred overnight. The solutions were then filtered through 0.2 mm filters and pH was adjusted from the initial value of 3.4 to 10.2 by addition of NaOH. The SRHA solution was stocked in the dark at 4 C. The total organic carbon (TOC) content was measured at 232.76 mg/L (Phoenix 8000 TOC analyzer, Teledyne Tekmar). Key properties of SRHA was reported elsewhere (Hong and Elimelech, 1997) including the molecular weight (range of 1e5 kDa) and composition.
2.3.
Characterization of silicon nanoparticles
The hydrodynamic diameter and the electrophoretic mobility (EPM) measurements for the silicon NPs were obtained by Malvern Zetasizer (Nano ZS, Malvern, UK). A monochromatic coherent HeeNe laser with a fixed wavelength of 633 nm was used as a light source and the intensity of scattered light was measured by a detector at 173 . Each auto-correlation function
107
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 0 5 e1 1 2
was accumulated for 10 s and more than 10 auto-correlations were conducted for each measurement. 1 mL of the suspension was introduced into a disposable polystyrene cuvette (Sarstedt, Germany) to determine particle size. The hydrodynamic diameter (intensity based) is calculated using the StokeseEinstein equation. Volume or number distributions were obtained from the fundamental intensity distribution, using the Mie theory. Folded capillary cells (DTS1060, Malvern) were used in EPM measurement. All DLS measurements were conducted at 25 C, at a minimum, in duplicates. Scanning electron microscope (Zeiss 982 field-emission SEM), transmission electron microscope (Philips FEI CM20 FEG S/TEM) and atomic force microscope (Nano-l, Pacific Nanotechnology) were used to shed light on the DLS measurement. All carbon TEM films (300C-CO, Electron Microscopy Sciences) were treated in a glow discharge system in order to make the carbon surface hydrophilic. This was done to ensure the sample was well dispersed on the carbon surface. For electron microscope sample preparation, one drop of the sample dispersion was placed on a carbon support film. The extra dispersion was blotted off using filter paper from the edge of the film. The remaining dispersion was dried for approximately 5 min under ambient condition before submission to TEM for data collection. For the AFM analysis, a drop of the NP dispersion was placed on a fresh cleaved mica substrate and dried overnight in a clean room. The scanning frequency was conducted at 0.5 Hz. X-ray diffraction (XRD) analyses (Rigaku DXR-3000) were performed at 40 kV and 40 mA using a diffracted beam graphite-monochromator with Cu radiation. The XRD patterns were collected in the 2q range of 5 e85 with a step size of 0.02 and a count time of 3 s per step. The qualitative analyses of the XRD patterns were conducted using the Jade software (Materials Data Inc., California, 2006).
2.4.
Determination of aggregation kinetics
The attachment efficiency (the inverse stability ratio, 1/W ) for the silicon NPs was determined from the increase in the hydrodynamic radius Rh at various Ca2þ concentrations. Assuming that all primary particles within an aggregate are independent and scatter identically, the aggregation rate constant, k11, is determined from the initial relative rate of change of the hydrodynamic radius from the following equation: k11 f
1 drh ðt; qÞ N0 dt t/0
(1)
Here, rh(t,q) is the hydrodynamic radius as a function of t and q; q is a constant of the scattering vector. At a fixed angle, the aggregation rate constant k11 is proportional to the slope of hydrodynamic radius Rh versus time as t / 0 at each salt concentration, divided by the initial nanoparticle number concentration N0 (Chen and Elimelech, 2006). The attachment efficiency a is defined as the aggregation rate constant of interest normalized by the rate constant derived under diffusion-controlled (fast) aggregation conditions (in the absence of an energy barrier) (Chen and Elimelech, 2006, 2007; Saleh et al., 2008):
a ¼ 1=W ¼
k11 ðk11 Þfast
1 drh ðtÞ drh ðtÞ N0 dt dt t/0 t/0 ¼ ¼ drh ðtÞ drh ðtÞ 1 ðN0 Þfast dt dt t/0;fast t/0;fast (2)
In Eq. 2, (k11)fast is the fast aggregation rate constant where all collisions result in aggregation in the absence of energy barriers. In the reaction-controlled regimes, a is smaller than unity where only a fraction of the collisions results in aggregation in the presence of energy barriers. However, in the diffusion-controlled (fast) aggregation every collision results in an aggregate and a is equal to unity. Linear least squares regression was applied to obtain an aggregation rate constant within the initial range of 1.25Rh0 (Rh0 is the initial hydrodynamic radius). For the aggregation experiments in the absence of SRHA, electrolytes were added into 1 mL silicon dispersion in cuvettes. The NP dispersions were then shaken vigorously and were placed into the Zetasizer. For the experiment in the presence of SRHA, 70 mL SRHA stock solution was added to the NP dispersion along with the addition of the electrolytes. The measurement procedure was similar to the one in the absence of SRHA. The value of (k11)fast is obtained from the average values of k11 in the diffusion-controlled regime in the absence of NOM.
3.
Results and discussion
3.1.
Characterization of silicon nanoparticles
The particle size of the silicon NPs, measured at random, was around 100 nm as shown in the SEM image (Fig. 1a). The result of the SEM imaging was similar to the results obtained by the TEM measurements (94.2 nm based on 80 measurements). Conversely, an average diameter of 50.9 nm (based on 7 measurements) was obtained by measuring the z-axis height of NPs, using AFM. The lower value obtained by AFM may be attributed to the limited number of spots measured (only 7 spots measured), compared with the wide size distribution of NPs. On the other hand, the DLS measurements yielded a mean diameter of 138 nm, which is a little greater than that measured by EM (see Fig. S2 for size distribution, Supplementary data). Due to the nature of the DLS technique where larger particles can mask the scattering signal of smaller particles, such deviations towards larger values can be expected (Domingos et al., 2009a). SEM microanalysis using EDS in the solid powders indicated that the silicon NPs are pure silicon (Fig. S3). The silicon NPs were found crystalline based on the XRD (Fig. S4) and selected area electron diffraction (SAED) analysis (Fig. S5). The silicon peaks in XRD spectrum were consistent with the crystalline silicon peaks in the Jade software database. Discrete dots in the SAED image also confirmed the crystalline nature of the silicon NPs. EPM was measured at various pH values as shown in Fig. 2. EPM of the silicon NPs became more negative with the increase of pH from 2.4 to 6.6 decreasing from 1.38 to 4.04 mm cm/V s. When pH value was greater than 6, EPM value did not change significantly. The negative surface
108
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 0 5 e1 1 2
Fig. 1 e Size measurement of silicon NPs using various techniques (a) SEM image, (b) TEM image, (c) AFM image. The scale in (a), (b) and (c) is 200 nm, 200 nm and 133 nm (z-distance), separately.
charge could be attributed to the silanol functional groups. Jarvis et al. reported similar z potential trend with pH for porous silicon microparticles and attributed the negative charge to surface SiOH species by aqueous oxidation (Jarvis et al., 2008). Based on EPM measurements, the negatively charged silicon NPs are not expected to deposit easily on negatively charged soil collectors. In addition, EPM of silicon NPs was more negative at neutral pH compared with other nanomaterials, such as fullerene (Chen and Elimelech, 2006), carbon nanotubes (Saleh et al., 2008), and boron NPs. (Liu et al., 2009) Therefore, silicon NPs are more likely to form stable dispersions in aqueous environment.
3.2.
aggregation curve of silicon NPs at various CaCl2 concentrations appears to adhere qualitatively to the DerjaguineLandaueVerweyeOverbeek (DLVO) model.
3.3.
Influence of SRHA on the aggregation of silicon NPs
The effect of SRHA on the aggregation of silicon NPs was studied in the presence of CaCl2 solutions (Fig. 3b). Silicon NPs started to aggregate at 0.01 M CaCl2 which is at a lower ionic strength than that in the absence of SRHA. It appears that a DLVO-type aggregation curve can be observed in the presence of SRHA; the attachment efficiency increased with
Aggregation of silicon NPs in Ca2þ electrolyte
Aggregation kinetics of silicon NPs were studied in the presence of CaCl2 electrolyte (Fig. 3a). At 0.08 M CaCl2, silicon NPs aggregation was probably induced by compression of the electric double layer. The attachment efficiency increased as CaCl2 concentrations increased from 0.08 M to 0.4 M, which was most likely caused by the decrease in the range and magnitude of the repulsive double layer interaction and hence a decrease in the height of the energy barrier (Elimelech et al., 1995). The regime where the attachment efficiency increases with the increase in ionic strength is defined as the reactioncontrolled regime. With the increase of CaCl2 from 0.4 M to 1 M, the attachment efficiency did not change greatly indicating a diffusion-controlled regime. The critical coagulation concentration (CCC) of Ca2þ is estimated at 0.4 M. The
Fig. 2 e Electrophoretic mobility of silicon NPs at various pH values.
Fig. 3 e A) Aggregation of silicon NPs in CaCl2 solutions in the absence of SRHA (blue diamonds). B) Aggregation of silicon NPs in CaCl2 solutions in the absence and in the presence of 15 ppm TOC SRHA; red squares represent silicon NPs in the presence of 15 ppm TOC SRHA.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 0 5 e1 1 2
109
Fig. 4 e TEM image for silicon nanoparticles in the presence of 0.08 M CaCl2 and 15 mg/L TOC SRHA.
increasing Ca2þ concentration to 0.1 M and then kept constant with further increase in Ca2þ concentration. Moreover, the attachment efficiency in the presence of SRHA was higher than that at the same ionic strength in the absence of SRHA in the reaction-controlled regime. In the diffusion-controlled regime, the attachment efficiency was about twice of that in the absence of SRHA; i.e., greater than unity. In addition, CCC of Ca2þ in the presence of SRHA (0.1 M) was smaller than that in the absence of SRHA (0.4 M). This indicated that SRHA greatly enhanced the aggregation of silicon NPs in the presence of CaCl2. Such enhancement was reported for other nanoparticles, such as fullerene (C60) (Chen and Elimelech, 2007). However, attachment efficiencies with values greater than unity are not consistent with the DLVO model. According to the DLVO model, an attachment efficiency value of unity implies that every collision due to Brownian motion results in an aggregation. For attachment efficiencies greater than unity, it seems that Brownian motion of the particles is not limiting and particles are coming together faster under the influence of another factor. The effects of this factor supersede the effects of the Brownian motion and it is causing the particles to aggregate faster and more efficiently than they would have aggregated on their own due to Brownian motion alone. This factor was attributed to bridging. Similar bridging behavior was also reported for the aggregation of fullerene (C60) NPs in the presence of SRHA and CaCl2 (Chen and Elimelech, 2007)
and for the aggregation of alginate-coated hematite NPs in the presence of CaCl2 (Chen et al., 2006). The authors suggested that suspended polymers undergo bridging with adsorbed polymer on the surface of the particles via calcium complexation, thus enlarging the collision radii of the suspended particles. At the same time, other suspended polymers are attached to each other through complexation with calcium ions and hence creating a gel cluster (Chen et al., 2006). However, the observed alginate-Ca2þ and SRHA-Ca2þ bridges were not well characterized in the aforementioned works. In this research, state of the art experimental approaches were used to verify the presence of the proposed bridging. TEM imaging was used to delineate the aggregation mechanisms of the silicon NPs in the presence and absence of SRHA (Fig. 4). TEM imaging provided evidence for the presence of the gel cluster due to complexation between Ca2þ and SRHA. Fig. 4 is a representative bright field (BF-TEM) micrograph of the silicon nanoparticle/aggregate morphology in the presence of SRHA and 0.08 M CaCl2 which was prepared on a continuous carbon-coated TEM film. Select area electron diffraction (SAED) in TEM was used to characterize the sample structures from a specific area (w50 nm in diameter area). A typical pattern for amorphous materials contains diffused rings and a typical pattern for a crystalline material contains either well-defined diffraction spots (single crystals) or a set of well-defined diffraction rings (poly-crystals). Fig. 4a (marked
110
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 0 5 e1 1 2
Fig. 5 e SEMeEDS images for silicon NPs in CaCl2 solutions in the presence (A) and absence (B) of SRHA.
on the BF-TEM image) shows an area that contains samples in amorphous structure contributed by a combination of the organic part of the sample and the carbon-film substrate. Fig. 4b (marked on the BF-TEM image) shows not only the diffused rings (that was attributed to the amorphous carbonfilm substrate) but also shows well-defined spots that were attributed to the crystalline Si nanoparticles used in the experiment. The crystalline phase marked by the discrete dots in the TEM image (Fig. 4b) was identified as silicon NPs, whereas the amorphous phase marked by the continuous circles was identified as the SRHA-Ca2þ bridge area (Fig. 4a). Bridging was not observed in the absence of SRHA as shown in another TEM image (Fig. S6, Supplementary data). Therefore it can be safely said that the shadowy areas are not silicon NPs artifacts but organic networks. Furthermore, SEMeEDS analysis was used to provide additional evidence about the proposed mechanism by elemental mapping. In the presence of SRHA (Fig. 5A), the bridge features can be observed as the shadowy area between the two silicon aggregates, similar to the TEM imaging results. In silicon mapping, silicon element positions agreed well with the SEM image. However, carbon and oxygen elements cover both the silicon aggregates and the bridge area, indicating that SRHA molecules played a significant role within and between the silicon aggregates. Li et al. also found similar effect of Ca2þ on the adhesion of carboxylate modified latex (CML) colloid, representing HA molecules in solution, to clean membrane surface. This was attributed to intermolecular bridging by Ca2þ, which associated the COO groups on the CML with COO groups on the membrane surface (Li and Elimelech, 2004). Oxygen element presence can be attributed to
functional groups such as carboxylic or phenolic OH in the SRHA molecules, although part of oxygen at silicon area may be attributed to surface SiOH species (Jarvis et al., 2008). There were no domains remarkably enriched in either Ca or Cl, indicating the primary mechanism of aggregation is bridging rather than compression of the double layer by CaCl2. In control experiments, silicon nanoparticles were stable in the presence of SRHA without Ca2þ, which verified the critical role of Ca2þ in the bridging effect. The aggregation of silicon NPs in the absence of SRHA is shown in Fig. 5B. The silicon elemental mapping corresponded with the main parts of SEM silicon position as expected. Carbon background noise is probably caused by the carbon film spread over the imaged area. Oxygen element corresponding with the silicon positions probably resulted from surface SiOH species or silicon oxidation by the high electron energy during the test analyses. However, the correspondence of calcium and chloride elements with the silicon aggregate positions was much more significant than the mapping in the presence of SRHA (Fig. 5A). This indicated that different mechanism predominant in the absence and presence of SRHA. In the absence of SRHA, the adsorption of Ca2þ at surfaces of nanoparticles suppressed the double layer repulsion and induced the aggregation of nanoparticles. Hence, the mapping of Ca was in accordance with the aggregation area of silicon nanoparticles. In contrast, although electrical double layer was also compressed at the same ionic strength of Ca2þ in the presence of SRHA as indicated by electrophoretic mobility measurement (Fig. S7), the bridging interaction of SRHA with Ca2þ was the predominant mechanism in this scenario. This can be verified by the fact that the
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 0 5 e1 1 2
mapping of C and O was more associated with the aggregate positions than Ca and Cl (Fig. 5A). In addition, the much higher aggregation rate (or attachment efficiency) in the presence of SRHA than in the absence of SRHA at the same ionic strength can be only attributed to the bridging effect.
4.
Conclusions
In summary, this paper reported the influence of SRHA on the aggregation behavior of silicon NPs in the presence CaCl2. It appears that SRHA has strong influence on the aggregation process and therefore the transport of silicon NPs in natural aqueous environments. The enhancement of silicon NPs aggregation in the presence of SRHA was attributed to the bridge formation in the Ca2þ solutions. To the best knowledge of the authors’, this paper is the first one to fully characterize and discriminate the NOM bridging by use of diffraction and dispersive energy. The presence of ions and NOMs can make the prediction of the transport of NPs in natural aquatic environments complicated, as evidenced in the literature and in this study. Therefore, more research is needed, on a case-by-case basis, for the assessment of the aggregation and transport behavior of nanomaterials.
Appendix. Supplementary data Supplementary data associated with this article can be found in the online version, at doi:10.1016/j.watres.2010.08.022.
references
Alvarez-Puebla, R.A., Dos Santos Jr., D.S., Aroca, R.F., 2007. SERS detection of environmental pollutants in humic acid-gold nanoparticle composite materials. Analyst 132 (12), 1210e1214. Benn, T.M., Westerhoff, P., 2008. Nanoparticle silver released into water from commercially available sock fabrics. Environmental Science and Technology 42 (11), 4133e4139. Brant, J.A., Labille, J., Bottero, J.Y., Wiesner, M.R., 2006. Characterizing the impact of preparation method on fullerene cluster structure and chemistry. Langmuir 22 (8), 3878e3885. Chen, K.L., Elimelech, M., 2006. Aggregation and deposition kinetics of fullerene (C60) nanoparticles. Langmuir 22 (26), 10994e11001. Chen, K.L., Elimelech, M., 2007. Influence of humic acid on the aggregation kinetics of fullerene (C60) nanoparticles in monovalent and divalent electrolyte solutions. Journal of Colloid and Interface Science 309 (1), 126e134. Chen, K.L., Elimelech, M., 2008. Interaction of fullerene (C60) nanoparticles with humic acid and alginate coated silica surfaces: measurements, mechanisms, and environmental implications. Environmental Science and Technology 42 (20), 7607e7614. Chen, K.L., Mylon, S.E., Elimelech, M., 2006. Aggregation kinetics of alginate-coated hematite nanoparticles in monovalent and divalent electrolytes. Environmental Science and Technology 40 (5), 1516e1523.
111
Choi, J., Zhang, Q., Reipa, V., Wang, N.S., Stratmeyer, M.E., Hitchins, V.M., Goering, P.L., 2009. Comparison of cytotoxic and inflammatory responses of photoluminescent silicon nanoparticles with silicon micron-sized particles in RAW 264. 7 macrophages. Journal of Applied Toxicology 29 (1), 52e60. Diegoli, S., Manciulea, A.L., Begum, S., Jones, I.P., Lead, J.R., Preece, J.A., 2008. Interaction between manufactured gold nanoparticles and naturally occurring organic macromolecules. Science of the Total Environment 402 (1), 51e61. Domingos, R.F., Baalousha, M.A., Ju-Nam, Y., Reid, M.M., Tufenkji, N., Lead, J.R., Leppard, G.G., Wilkinson, K.J., 2009a. Characterizing manufactured nanoparticles in the environment: multimethod determination of particle sizes. Environmental Science and Technology 43 (19), 7277e7284. Domingos, R.F., Tufenkji, N., Wilkinson, K.J., 2009b. Aggregation of titanium dioxide nanoparticles: role of a fulvic acid. Environmental Science and Technology 43 (5), 1282e1286. Elimelech, M., Gregory, J., Jia, X., Richard, A.W., 1995. Particle Deposition and Aggregation: Measurement, Modelling and Simulation. Butterworth-Heinemann, Oxford, U.K. Fujioka, K., Hiruoka, M., Sato, K., Manabe, N., Hoshino, A., Hirakuri, K., Yamamoto, K., 2008. Evidence-based toxicity of probes. Geranio, L., Heuberger, M., Nowack, B., 2009. The behavior of silver nanotextiles during washing. Environmental Science and Technology 43 (21), 8113e8118. Ghosh, S., Mashayekhi, H., Pan, B., Bhowmik, P., Xing, B., 2008. Colloidal behavior of aluminum oxide nanoparticles as affected by pH and natural organic matter. Langmuir 24 (21), 12385e12391. Hong, S., Elimelech, M., 1997. Chemical and physical aspects of natural organic matter (NOM) fouling of nanofiltration membranes. Journal of Membrane Science 132 (2), 159e181. Jarvis, K.L., Barnes, T.J., Prestidge, C.A., 2008. Aqueous and thermal oxidation of porous silicon microparticles: implications on molecular interactions. Langmuir 24 (24), 14222e14226. Kang, S., Mauter, M.S., Elimelech, M., 2009. Microbial cytotoxicity of carbon-based nanomaterials: implications for river water and wastewater effluent. Environmental Science and Technology 43 (7), 2648e2653. Keller, A.A., Wang, H., Zhou, D., Lenihan, H.S., Cherr, G., Cardinale, B.J., Miller, R., Ji, Z., 2010. Stability and aggregation of metal oxide nanoparticles in natural aqueous matrices. Environmental Science and Technology 44 (6), 1962e1967. Kim, K., Park, J.H., Doo, S.G., Nam, J.D., Kim, T., 2009. Generation of size and structure controlled Si nanoparticles using pulse plasma for energy devices. Thin Solid Films 517 (14), 4184e4187. Lecoanet, H.F., Bottero, J.Y., Wiesner, M.R., 2004. Laboratory assessment of the mobility of nanomaterials in porous media. Environmental Science and Technology 38 (19), 5164e5169. Lecoanet, H.F., Wiesner, M.R., 2004. Velocity effects on fullerene and oxide nanoparticle deposition in porous media. Environmental Science and Technology 38 (16), 4377e4382. Li, Q., Elimelech, M., 2004. Organic fouling and chemical cleaning of nanofiltration membranes: measurements and mechanisms. Environmental Science and Technology 38 (17), 4683e4693. Limbach, L.K., Bereiter, R., MuI`^ller, E., Krebs, R., GaI`^lli, R., Stark, W. J., 2008. Removal of oxide nanoparticles in a model wastewater treatment plant: influence of agglomeration and surfactants on clearing efficiency. Environmental Science and Technology 42 (15), 5828e5833. Liu, X., Wazne, M., Christodoulatos, C., Jasinkiewicz, K.L., 2009. Aggregation and deposition behavior of boron nanoparticles in porous media. Journal of Colloid and Interface Science 330 (1), 90e96.
112
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 0 5 e1 1 2
Liu, X., Wazne, M., Han, Y., Christodoulatos, C., Jasinkiewicz, K.L., 2010. Effects of natural organic matter on aggregation kinetics of boron nanoparticles in monovalent and divalent electrolytes. Journal of Colloid and Interface Science 348 (1), 101e107. ˜ ¡cz, E., 2007. Interaction between humic acid Majzik, A., TombA and montmorillonite in the presence of calcium ions II. Colloidal interactions: charge state, dispersing and/or aggregation of particles in suspension. Organic Geochemistry 38 (8), 1330e1340. Nowack, B., Bucheli, T.D., 2007. Occurrence, behavior and effects of nanoparticles in the environment. Environmental Pollution 150 (1), 5e22. O’Farrell, N., Houlton, A., Horrocks, B.R., 2006. Silicon nanoparticles: applications in cell biology and medicine. International Journal of Nanomedicine 1 (4), 451e472. Pallem, V.L., Stretz, H.A., Wells, M.J.M., 2009. Evaluating aggregation of gold nanoparticles and humic substances using fluorescence spectroscopy. Environmental Science and Technology 43 (19), 7531e7535. Park, J.H., Gu, L., von Maltzahn, G., Ruoslahti, E., Bhatia, S.N., Sailor, M.J., 2009. Biodegradable luminescent porous silicon nanoparticles for in vivo applications. Nature Materials 8 (4), 331e336. Saleh, N.B., Pfefferle, L.D., Elimelech, M., 2008. Aggregation kinetics of multiwalled carbon nanotubes in aquatic systems: measurements and environmental implications. Environmental Science and Technology 42 (21), 7963e7969.
Scheckel, K.G., Luxton, T.P., El Badawy, A.M., Impellitteri, C.A., Tolaymat, T.M., 2010. Synchrotron speciation of silver and zinc oxide nanoparticles aged in a kaolin suspension. Environmental Science and Technology 44 (4), 1307e1312. Smith, B., Wepasnick, K., Schrote, K.E., Bertele, A.R., Ball, W.P., O’Melia, C., Fairbrother, D.H., 2009. Colloidal properties of aqueous suspensions of acid-treated, multi-walled carbon nanotubes. Environmental Science and Technology 43 (3), 819e825. Steinberg, C.E.W., Meinelt, T., Timofeyev, M.A., Bittner, M., Menzel, R., 2008. Humic substances e part 2: interactions with organisms. Environmental Science and Pollution Research 15 (2), 128e135. Stupca, M., Alsalhi, M., Al Saud, T., Almuhanna, A., Nayfeh, M.H., 2007. Enhancement of polycrystalline silicon solar cells using ultrathin films of silicon nanoparticle. Applied Physics Letters 91 (6). Tian, B., Zheng, X., Kempa, T.J., Fang, Y., Yu, N., Yu, G., Huang, J., Lieber, C.M., 2007. Coaxial silicon nanowires as solar cells and nanoelectronic power sources. Nature 449 (7164), 885e889. Wang, Y., Li, Y., Pennell, K.D., 2008. Influence of electrolyte species and concentration on the aggregation and transport of fullerene nanoparticles in quartz sands. Environmental Toxicology and Chemistry 27 (9), 1860e1867. Yang, K., Lin, D., Xing, B., 2009. Interactions of humic acid with nanosized inorganic oxides. Langmuir 25 (6), 3571e3576.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 3 e1 2 4
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Asset deterioration and discolouration in water distribution systems P.S. Husband*, J.B. Boxall Pennine Water Group, Department of Civil and Structural Engineering, University of Sheffield, Mappin street, Sheffield S1 3JD, United Kingdom
article info
abstract
Article history:
Water Distribution Systems function to supply treated water safe for human consumption
Received 18 June 2010
and complying with increasingly stringent quality regulations. Considered primarily an
Received in revised form
aesthetic issue, discolouration is the largest cause of customer dissatisfaction associated
4 August 2010
with distribution system water quality. Pro-active measures to prevent discolouration are
Accepted 10 August 2010
sought yet network processes remain insufficiently understood to fully justify and optimise
Available online 17 August 2010
capital or operational strategies to manage discolouration risk. Results are presented from a comprehensive fieldwork programme in UK water distribu-
Keywords:
tion networks that have determined asset deterioration with respect to discolouration.
Hydraulics
This is achieved by quantification of material accumulating as cohesive layers on pipe
Shear stress
surfaces that when mobilised are acknowledged as the primary cause of discolouration. It
Cohesive material
is shown that these material layers develop ubiquitously with defined layer strength
Regeneration
characteristics and at a consistent and repeatable rate dependant on water quality. For UK
Asset deterioration
networks iron concentration in the bulk water is shown as a potential indicator of deterioration rate. With material layer development rates determined, management decisions that balance discolouration risk and expenditure to maintain water quality integrity can be justified. In particular the balance between capital investment such as improving water treatment output or pipe renewal and operational expenditure such as the frequency of network maintenance through flushing may be judged. While the rate of development is shown to be a function of water quality, the magnitude (peak or average turbidity) of discolouration incidents is shown to be dominated by hydraulic conditions. From this it can be proposed that network hydraulic management, such as regular periodic ‘stressing’, is a potential strategy in reducing discolouration risk. The ultimate application of this is the hydraulic design of self-cleaning networks to maintain discolouration risk below acceptable levels. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Discolouration remains the single largest water quality issue facing water companies worldwide. Of the 154,985 customer complaints about drinking water quality in 2007 for England
and Wales, 124,671 (80%) were about discoloured water. Yet with a 99.96% compliance with the European Drinking Water Directive, the overall result for England and Wales is comparable with the best reported in Europe. In 2007, 33% of all incidents investigated by the UK Drinking Water Inspectorate
* Corresponding author. Tel.: þ44 114 2225416; fax: þ44 114 2225700. E-mail addresses:
[email protected] (P.S. Husband),
[email protected] (J.B. Boxall). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.021
114
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 3 e1 2 4
(DWI) were due to discolouration, Fig. 1a. Of significance it was found that 48% of these incidents were a result of ‘planned work’, Fig. 1b (DWI, 2007; DWI, 2008; DWI, 2009). This is a clear indictment that current operational and management strategies are insufficiently informed and that there is inadequate understanding of the processes and mechanisms that lead to discolouration events. In addition the DWI questioned the UK industry’s reactive, failure driven attitude towards the maintenance of assets, which it had aimed to improve through instigating Distribution Operation and Maintenance Strategies, DOMS (DWI 2002). To justify efficient implementation of DOMS, water companies need to balance justifiable capital expenditure against the long term operational costs of maintenance interventions with respect to maintaining water quality. From 2010 UK water supplier’s statutory duties to manage assets will also include discolouration complaints as a quantitative measure reflecting consumers’ expectations set in a framework of maintaining serviceability (OFWAT, 2008). Furthermore implementation of Water Safety Plans (WSP), regarded as the most effective way to ensure a water supply safe for human consumption, require comprehensive risk assessment and risk management of the network (WHO, 2005). To achieve these statutory obligations within the distribution network it is evident that understanding of the processes governing discolouration is essential.
2.
Background
For a discolouration event to occur a three-phase process may be conceptualised consisting of a material supply to the network, a period during which material accumulates and ultimately the mobilisation of material into the bulk flow (Boxall and Saul, 2005). Material sources linked with discolouration include oxidation of iron mains, metal precipitation, biological activity and poor system maintenance or treatment practices (McCoy and Olsen, 1986; Gauthier et al., 2001; Slaats et al., 2002; Sarin et al., 2004). Trials in a plastic only network with particle-free water produced by 0.1 mm ultra-filtration demonstrated that particles responsible for discolouration can originate solely from the supplied water (Vreeburg et al., 2008). Most likely it was proposed through intrinsic processes such as precipitation and biological
activity. With the inevitable presence of material within a distribution system, it could be assumed that localised settling and the subsequent mobilisation of loose deposits may account for discolouration incidents (Barbeau et al., 2005). Research analysing material entrained during flushing operations has however repeatedly demonstrated that it is particulate in nature, with a size predominately around 10 mm and a specific gravity in the range 1e1.3 (Boxall et al., 2001; Gauthier et al., 2001; Seth et al., 2004; Verberk et al., 2006; Vreeburg and Boxall, 2007). Sedimentation of these particles is restricted to quiescent conditions; therefore in general they remain as a permanent wash load or solute within distribution systems (Smith et al., 1999; Boxall et al., 2001; Hossain et al., 2003; Ryan and Jayaratne, 2003). Based on such observations, Boxall et al. (2001) proposed and subsequently validated for field data (Boxall and Saul, 2005; Husband and Boxall, 2010) that it is the generation of cohesive-like material layers that are responsible for discolouration. In their empirical model, Predicting Discolouration in Distribution Systems (PODDS), Boxall et al. (2001) suggest that these material layers behave in a similar way to cohesive estuarine muds and cohesive-like sewer sediments (Parchure and Mehta, 1985; Mehta and Lee, 1994; Skipworth et al., 1999). In the PODDS model however the concept of cohesive layers was specifically applied with the assumption that the layers are conditioned by system hydraulics. The erosion of material layers with distinct shear stress characteristics and their subsequent regeneration has since been demonstrated under laboratory conditions in a full scale pipe loop (Husband et al., 2008). The practically recognised trigger for discolouration event initiation is an increase in hydraulic conditions, for example a rise in flow (Twort et al., 2000; Ackers et al., 2001; Marshall, 2001; Slaats et al., 2002; Ryan and Jayaratne, 2003). It was recognised by Boxall et al. (2001) in the PODDS model that increasing the flow raises the force acting tangentially to the pipe surface, the boundary wall shear stress. If this exceeds the conditioned shear strength of the cohesive layers then mobilisation of discolouration material would occur. An appreciation of the non-linear relationship between shear stress and velocity can explain why some valve operations, rezoning or other network interventions may have resulted in discolouration incidents when planning indicated only marginal increases in flow velocity. By analysing for boundary
Fig. 1 e DWI incidents in England and Wales, 2006.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 3 e1 2 4
shear stress, hydraulically simulating ‘planned works’ as part of a pro-active management strategy, it is likely that such events may be anticipated. In the Netherlands a 0.4 m/s flow velocity has been identified as a design criterion for self-cleaning networks (Slaats et al., 2002). The value of 0.4 m/s is a pragmatic one proposed to prohibit the accumulation of material within Dutch water distribution systems. With progressively more stringent regulations, increasing customer expectations and acceptance that discolouration is a widespread international issue, it is essential that water suppliers have the ability to assess, quantify and proactively manage this risk of water quality failure. The concept of cohesive material layers with a relationship linking shear stress and discolouration potential can accurately describe the mobilisation process and has been verified through modelling of distribution network discolouration events. What is not known however is the rate these layers regenerate following disturbance or the controlling variables, although laboratory and associated field studies suggest that it is a continuous, repeatable and linear accumulation process (Cook, 2007; Husband and Boxall, 2008; Husband et al., 2008). Turbidity profiles from such fieldwork have shown distinct responses from unlined cast iron and plastic pipes and interpretation has indicated internal pipe condition in monitored sections (Husband and Boxall, 2010). However, the significance of upstream pipes, in particular unlined cast iron, and the impact of this on bulk water quality and hence material accumulation and discolouration risk in downstream sections has not been investigated. If the rate at which material layers regenerate was known and the influential factors understood it would be possible to make strategic management decisions that justify cost benefits from capital or operational expenditure. In addition the knowledge gained of system behaviour may be used to support risk assessment methodologies such as WSP and aid prediction of discolouration response from planned works or changes in network operation. The aim of this work is therefore to quantify material layer regeneration and assess the significance of potential influencing factors such as source water type, coagulation treatment processes and pipe material. The work is based on the results of repeated field operations in the UK, spanning a number of years, where turbidity data has been collected from hydraulically induced discolouration events. From this data, regeneration values have been determined and correlation of these with site information has been used to assess influencing factors.
3.
Fieldwork
Collection of field data during uni-directional flushing operations was planned in partnership with nine collaborating water companies who serve over 75% of the registered consumers in England and Wales (40 million customers). Fieldwork objectives involved the implementation and monitoring of incremental hydraulic flushing operations to obtain the temporal and spatial turbidity response of pipes within live distribution systems. The incremental approach of stepped flow increases above typical peak flows has been
115
successfully applied to demonstrate the cohesive nature and variable shear strength properties of material layers shown to cause discolouration and how these layers are conditioned by the hydraulic regime (Husband and Boxall, 2010). The results presented here are for repeated visits to sites such that the amount of material accumulated over time can be investigated and correlated with other system information. The knowledge gained by investigating material quantities and shear strength characteristics of developing layers will help inform management strategies to mitigate discolouration risk. Selection of test sites was planned to facilitate coverage of a range of factors hypothesised as potentially influential in discolouration processes and the development of material layers. Factors included: pipe material, significant as the surface on which the material layers develop, pipe diameter and roughness that determine boundary shear stress identified to condition material layers under hydraulic loading, factors known to influence bulk water quality such as B water source, B water treatment processes and B presence of upstream unlined cast iron pipes, known to contribute to corrosion products to the bulk water. To establish a comparable dataset between sites required a consistent baseline, technically difficult to achieve due to the inherent variables in distribution networks. Ideally the initial flushing operation would mobilise fully developed material layers that represent a maximum discolouration potential. With hydraulics known to condition the material layers, priority was therefore given to sites for which historical records spanning back 5 years or more indicating no hydraulic events in excess of typical demands. Operation dates were typically set by company maintenance programmes yet seasonal variations are an unknown factor in layer development, although work indicated minimal impact for one specific groundwater site (Boxall et al., 2003). To mitigate for possible seasonal effects repeat flushing operations were planned to allow a full year layer development period. For the year between the first and repeat operations, collaborating water companies were requested not to sanction non-critical interventions within the monitored site that may compromise the development of discolouration material. The two operations (initial visit and repeat) were planned to be completed under identical conditions (same time of day, flushing flow rates and duration) but a year apart, thereby producing two sets of comparable turbidity data. From this an annual regeneration of erodable discolouration material, irrespective of operational date or location, could be determined. Subsequent site specific details could then be correlated to identify influencing factors. To capture the required data at the wash out hydrant a specially adapted standpipe was used. This incorporated a valve at the discharge point to control the flow; using the hydrant valve results in a rapid pressure drop in water pressure entering the standpipe causing the dissolution of air into bubbles that critically disrupts turbidity measurements. Prior
116
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 3 e1 2 4
The study utilised 15 locations from across the UK with 67 monitored pipe sections. Site locations are shown in Fig. 2 with the following water supply details: 5 Groundwater supply e no coagulation treatment 2 Groundwater/surface water blended supply e iron coagulation treatment 8 Surface water supply e 5 iron and 3 aluminium coagulation treatment Pipe sections monitored were primarily of typical distribution pipe diameters around 100 mm but did range from 53 mm to 306 mm with the following split in pipe materials: Fig. 2 e Regeneration fieldwork sites.
to the valve a tapping point directed water through a sampling cell housing the turbidity probe, a CENSAR CT, whilst a second probe was placed directly in the standpipe riser. In addition to the two turbidity probes logging at 10 s, manual samples were collected for laboratory water quality analysis and by a HACH Pocket Turbidimeter. This replication of the recorded turbidity data ensured a high degree of confidence with the results. In addition to the turbidity probes the standpipe was also fitted with a Burkert Easyflow flow meter, pressure tapping and pressure gauge to monitor the hydraulic conditions, essential for hydraulic model and pipe roughness calibration (critical in calculating the hydraulic gradient and therefore pipe wall shear stress). In all examples shown upstream pipe sections were flushed to ensure a clean supply and an inlet hydrant with attached sampling standpipe and turbidity monitoring were connected to confirm that material mobilised at the wash out could be solely attributed to the pipe length under investigation.
23 Cast Iron (CI, unlined) 17 PE (or polyethylene derivative) 15 uPVC 12 Asbestos Cement (AC) or cement lined CI
The results reported here generally refer to the annual regeneration and are therefore not directly an indication of the risk or potential magnitude of a discolouration event. The annual regeneration is a percentage value relating material mobilised from the repeat operation to that measured during the initial operation. In addition to an annual regeneration, an average turbidity value is calculated for each flushing operation. As flushing duration and hydraulic steps were variable between pipe lengths, this value is not directly comparable but it may be considered to indicate the possible discolouration potential should a shear force similar to those imposed by the flushing and in excess of typical conditioning hydraulics occur for that pipe length.
4.
Site results
Results from three fieldwork sites are presented that illustrate the turbidity features typical to all flushing studies conducted. Individual field results also demonstrate how pipe length specific characteristics need to be considered when analysing
Fig. 3 e Pipe diagram and hydrant locations, Site 1.
117
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 3 e1 2 4
Fig. 4 e Site 1 initial and 12 month return turbidity results with flushing flow rate.
the flushing turbidity response. Following the examples, overall trends governing the regeneration of discolouration material in distribution systems are explored based on analysis of site characteristics. In order to determine a useful regeneration index, three methods were trialled to establish if there is an optimum for future studies. These calculations have been undertaken for all pipe lengths, but are only presented here for the six example pipe lengths across the three sites. The first method uses the peak turbidity, which is a subjective measure based on manually selecting a turbidity feature that can be clearly identified from both initial and repeat operations (not necessarily the highest single value although typically a dominant turbidity peak). This peak value does however convey a potential risk rating as it is possible that a consumer may be subjected to water with this turbidity as a ‘worst case’. The second measure is the average turbidity, simply the mean of all data within the measured time frame. The third measure is based on integration of the timeeturbidity plots: effectively a step in calculating an amount of material (Gauthier et al., 2001; Boxall et al., 2003). However the material mobilised from sites is not consistent due to differences in water quality
so multiplication by a common conversion factor to suspended solids, a mass of material, is not undertaken. Should site specific conversion factors be determined however, mass of material mobilised would probably be the optimal and operationally pertinent comparison. With the appropriate score determined for each operation, a percentage regeneration value, indicating the rate material returns to a fully developed and maximum discolouration risk, is calculated by dividing the repeat by the initial results (flush 2/flush 1).
4.1.
Example site 1
The first site is shown in Fig. 3 and comprises two sections of polyethylene (PE) pipe of different internal diameters (72 and 89 mm) that form a looped section of an urban District Metered Area (DMA e a term used to denote a management sub-zone of a distribution network in the UK). The site is supplied aluminium coagulation treated river water and has asset records indicating no upstream unlined iron pipe sections. This operation used the planned approach of stepped increases in flushing flow rate to investigate discolouration
Table 1 e Site 1 regeneration flushing results. Flush
A B
Ø mm
89 72
Lm
380 282
Pipe
PE PE
Peak NTU
Average NTU
Amount NTU.s
1
2
%
1
2
%
1
2
%
22.5 25.6
6.1 7.8
27 30
10.5 17.1
3.7 5.3
35 31
28038 33857
9988 10650
36 31
118
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 3 e1 2 4
difference in shear stress (and not velocity) is reflected in more discolouration material mobilised during the initial trial from the smaller diameter and 100 m shorter section of pipe (so less material might have been expected).
4.2.
Fig. 5 e Pipe diagram and hydrant locations, Site 2.
material shear strength properties, in particular the regenerated material mobilised during the repeat flushing operation. Fig. 3 shows the network location of the two monitored sections A and B. Fig. 4 shows the temporal turbidity response (primary y-axis) and the flushing flow (secondary y-axis). Each plot shows two turbidity datasets; the initial field visit flush 1 and the repeat operation 1 year (12 months) later, flush 2. For both plots in Fig. 4 measured turbidity increases are evident in response to the flushing showing discolouration material is eroded and propagates to the wash out. This mobilisation of material is universal to every flushing operation undertaken during the course of these studies and highlights the ubiquitous nature of discolouration. Crucially at every increase in applied shear stress there is a further release of discolouration material, an expected phenomenon with the cohesive layer concept. The same response characteristics to the stepped increases in flow are observed during the initial and repeat operations. This demonstrates that within a year cohesive layers accumulate with a range of shear strengths, not solely weakly or strongly adhered material. The mobilisation of the regenerated material can be seen to produce a scaled version of the initial data, indicating distinct accumulation patterns and a reproducible process. Due to the hydraulic and temporally equivalent consecutive operations the rate at which this material has regenerated can be derived by comparing the two datasets. The percentage regeneration results are shown in Table 1 and demonstrate a consistency between the three calculation methods for these two pipe lengths. A combined average of 30% represents a potential maximum discolouration risk, assuming linear development, in less than 4 years for these plastic pipes. Examination from UK industry standard hydraulic model EPANET software using 24 h extended period simulation with top down demand allocation (pattern derived from DMA flow meter data and distributed according to billing data) of the network shows a peak daily velocity of 0.1 m/s in both pipes with the turbidity observed showing no evidence of self-cleaning, as expected based on the Dutch value of 0.4 m/s. During the flushing the shear stress at each of the three steps is approximately double for the smaller 72 mm pipe at 1/2/8 N/ m2 compared to 0.6/1.3/3.6 N/m2 for the 89 mm pipe (flow velocities 0.6/0.9/1.9 m/s and 0.5/0.7/1.2 m/s respectively). This
Example site 2
The second site presented here is supplied with blended ground/surface water having undergone iron coagulation based treatment. Results are from stepped flushing of two parallel pipe lengths of 79 mm PE that form a loop within a further looped residential DMA, Fig. 5. The turbidity/flow results are shown in Fig. 6. Flush C produced trends of similar magnitude to those from Site 1, Fig. 4 at about 30 NTU, with the exception of a 200 NTU spike on start-up not seen during the repeat operation. Flush D, the second and shorter pipe section however produced turbidity of over 900 NTU (instrument limit), well in excess of values seen elsewhere in plastic pipes. Although most turbidity was mobilised during the first flow step, the two further increases in flow produced turbidity peaks of 80 NTU and 10 NTU respectively. This demonstrates material attachment with low cohesive shear strength and therefore presents a significant discolouration risk. It was suggested that an additional intrusion of material may have been responsible yet the flush a year later produced a scaled replica of the turbidity trend, peaking this time at 300 NTU. This indicates that the accumulation is most likely a chronic process due to a continual material source and favourable hydraulics. No source of material to generate this level of turbidity was identified, but investigations to examine the accountability of the blended ground/surface iron coagulated supply water in Site 2 have been proposed. From Table 2 it can be seen that although the peak turbidity is high for Flush D, all three scores show regeneration of material to be around 30%, indicating over 3 years required for a return to maximum risk assuming linear regeneration. This is akin to the plastic pipes in Site 1. Due to the non-replicated turbidity feature on start-up during the first visit of Flush C having an adverse effect on the material regenerated calculations, the results are non-conclusive. This second trial in a looped network however indicates that the presence of potential tidal points where two flow paths converge (not confirmed due to sensitivity of hydraulic solutions in such low flow situations) may induce unusually high material accumulation, speculatively with a low shear strength and therefore high discolouration risk.
4.3.
Example site 3
The third example site is a groundwater supplied rural DMA with no coagulation treatment; Fig. 7. Operational constraints limited flushing to a single step from the two predominately cast iron (CI) pipe length operations, Flushes E and F. The majority of Flush E (2 km in total), with the exception of the final 594 m dead-end section, and the whole of Flush F (1.5 km) from the supply to a village a further 1.5 km downstream. Peak daily flows exceed 0.4 m/s, the proposed self-cleaning velocity (Slaats et al., 2002). The turbidity results are shown in Fig. 8.
119
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 3 e1 2 4
Fig. 6 e Site 2 initial and 12 month return turbidity results with flushing flow rate.
The turbidity results show that regeneration of discolouration material has occurred in both pipe sections with material accumulation again following distinct and repeatable patterns. Regeneration values for mobilised turbidity are shown in Table 3 and indicate a combined 30% regeneration in 12 months for these predominately CI pipes in this groundwater supplied network. It should be noted however that a discolouration risk may exist long before a ‘full’ accumulation is attained, evident from these operations where turbidity exceeds the UK regulatory limit (at the customers tap) of 4 NTU. From an operational viewpoint the turbidity spikes between 12:20 and 12:30 in Fig. 8b are due to propagation of material entrained during the Flush E. This reinforces the need to plan flushing of all connecting pipe lengths to prevent particulate material, which once suspended, remains as a solute leading to potential customer complaints. By considering travel times, sections from the temporal turbidity plots can be assigned to specific lengths of the flushed pipe sections, such as dead ends and different pipe materials. Segregation of the results for the dead-end CI section of Flush E, the initial turbidity observed in Fig. 8a up to
time 10:05, indicates 48% regeneration or effectively only two years until full discolouration risk, assuming the deterioration is linear. Analysis of the period 10:11 to 10:25, associated to the 900 m CI section of this flush and Flush F Fig. 8b, also a CI pipe, both show lower regeneration. The bulk water quality and pipe materials are the same for these sections; however the normal daily flow is lower in the smaller diameter final deadend section of Flush E. This suggests that the daily hydraulic loading is a factor that influences regeneration. The period 10:05 to 10:11 shown in Fig. 8a represents the turbidity response from the 101 mm asbestos cement (AC) pipe section with a modelled peak daily flow of 0.49 m/s. As little turbidity (less than 4 NTU) is initially mobilised or regenerated this supports the Dutch proposal that a self-cleaning velocity may be a simple and effective management tool to control discolouration in non CI pipes. After this there is a 10 NTU turbidity step that is associated with the 900 m of equivalent diameter CI pipe. Both these sections experience the same daily demand and therefore flow velocity, yet the CI section exhibits a significant discolouration risk. The increased accumulation of material may be anticipated if in-situ
Table 2 e Site 1 regeneration flushing results. Flush
C D
Ø mm
79 79
Lm
254 97
Pipe
PE PE
Peak NTU
Average NTU
Amount NTU.s
1
2
%
1
2
%
1
2
%
34 900
11 313
31 35
27.7 162.3
3.3 51.1
12 31
25186 195392
2991 62275
12 32
120
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 3 e1 2 4
products. All this evidence implicates water quality as the key factor governing regeneration. It is therefore proposed that discolouration material accumulation can be simplified as two processes; primarily an indirect process with material adsorbing to the boundary surfaces from the bulk water and secondly a direct precipitation process due to corrosion of iron assets.
5.
Fig. 7 e Pipe diagram and hydrant locations, Site 3.
corrosion processes are considered contributory but the results also reveal CI pipes, typically rougher than plastic or AC pipes and therefore undergoing greater changes in shear stress, are not self-cleaning at this velocity. Epanet (Rossman, 2000) hydraulic model calibration, determined through the collection of flow and pressure data during the flushing operations, set the DarcyeWeisbach pipe roughness at 0.1 mm and 1 mm in the AC and CI pipes respectively. The majority of sites reported here only underwent one repeat visit. However Husband and Boxall (2008) presented results from a site with multiple repeat visits that demonstrated linear asset deterioration. Combined with the repeatable, and therefore predictable, turbidity trends observed during these trials and a consistent accumulation rate at each site, this indicates a regular, yet site specific, supply of discolouration material. Of the three site examples the lowest average regenerated turbidity came from the groundwater site, even accounting for CI pipes and the likely presence of corrosion by-
Site fieldwork summary
By collating the data from all sites it is possible to explore the relative influence of the factors identified as possibly influencing asset deterioration, in the form of material accumulation leading to potential discolouration. Key factors acknowledged as influencing the water quality include the source water origin, type of coagulation process during treatment and presence of unlined upstream cast iron pipes. In addition the interaction at the boundary of bulk water and the pipe wall is intrinsically important for any study of accumulation processes and hence pipe material is deemed a factor. Due to the number and range of interactions possible in live networks scatter is expected when summarising the data from all the sites. To investigate the statistical significance of this variance the coefficient of variation (Cv), a dimensionless number has been determined where; Cv ¼
s m
(1)
with s the standard deviation and m the mean of the sample groups. Cv values of less than one are low-variance, whereas values greater than one are considered high-variance. From Table 1 to Table 3 it can be seen that the average and integration method for calculating regeneration percentages returns comparable results while the peak method, subject to manual interpretation, is more variable. As the average
Fig. 8 e Site 3 initial and 12 month return turbidity results with flushing flow rate.
121
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 3 e1 2 4
Table 3 e Site 3 regeneration flushing results. Flush
Ø mm
Lm
Pipe
Peak NTU 1
E
F
101 900 101 440 74 594 Combined 101 1617
CI AC CI CI
2
17.2 3.1 No feature 31.4 14.3 Not applicable 13.7 4.1
Average NTU
NTU/s
%
1
2
%
1
2
%
18
10.9 4 14.3 10.4 8.0
2.0 0.9 6.8 2.8 2.4
18 23 48 27 30
9300 1079 6832 21880 24585
1550 255 3298 6020 7660
17 24 48 28 31
46 30
method is simpler and returns turbidity values likely to be indicative of that which may be observed by a consumer should a discolouration event occur, this has been used as the basis to compare factors between sites. The annual percentage and magnitude of regeneration results for all sites tested across the UK are shown in Table 4. An overall appraisal indicates that water quality factors are statistically significant in determining the annual percentage of material regenerated, with an average variance of only 0.5 and no single result greater than 1. Conversely with an average variance of 1.2 this analysis suggests that water quality does not influence the potential discolouration magnitude. From Table 4 information of practical benefit to water suppliers can be derived. For example it indicates CI pipes have the greatest deterioration, presenting a maximum discolouration risk after only 2 years from flushing (52% regeneration). With an average of over 20 NTU mobilised within a year these pipes also pose the greatest potential turbidity risk to customers. Combined with the knowledge that no self-cleaning effect has been identified and the detrimental effect on water quality (and subsequent layer development downstream) due to the release of corrosion by-products, effective management of these pipes is essential. If CI pipes are split between surface water and groundwater supplied sites the regeneration is 66% and 31% respectively, with crucially the former generating significantly higher average turbidities. Non ferrous, or ‘smooth’ pipes with no corrosion by-products, present an average regeneration percentage of 28%, suggesting 4 years after cleaning for full regeneration. After a year the average
turbidity during flushing in these pipes is 5 NTU, just exceeding UK regulatory limits. The results show surface water sites regenerate discolouration material quicker than other water sources, although some blended water sites with looped connections have produced very high turbidity, Fig. 6b. Typically surface waters are regarded as having poorer quality water so this result fits with the understanding that regeneration is governed by water quality. Examination of correlation with treatment processes suggest that sites supplied with iron coagulated water have a greater material layer regeneration yet of lower magnitude than aluminium coagulated waters. However treatment processes are linked to source water characteristics, so this measure in particular may be capturing more than simply treatment process effects. The presence of upstream unlined cast iron as a potential independent source of material was also examined. Not surprisingly sites with this feature present a higher rate of material regeneration, on average 50% or full risk in two years, twice the rate or half the time of sites with no upstream unlined CI pipes. This may explain the anomaly demonstrated between the lower regeneration recorded from the blended water sites compared to the expected higher quality groundwater sites. In these trials the latter sites had significant lengths of upstream CI pipes whereas the blended sites are plastic only networks. This complex interaction of variables is intrinsic when dealing with operational networks and indicates the need for ongoing assessment and a requirement for laboratory trials to study isolated factors. In the UK it is acknowledged that iron is a dominant constituent of discolouration material and it is common for
Table 4 e Coefficient of variation determination of annual regeneration and average regenerated turbidity for identified factors influencing water quality. Factor
CI pipe (all sites) Non-ferrous CI/surface water CI/groundwater Surface water Blended water Groundwater Iron coagulation Aluminium coagulation No coagulation Upstream unlined CI pipes No unlined CI pipes Average Cv
Regeneration per annum, %
Average regenerated turbidity, NTU
m
s
Cv
m
s
Cv
52 28 66 31 55 24 32 49 40 32 54 27
30.7 15.7 32.0 11.4 32.5 14.2 10.5 33.2 29.2 10.5 30.1 14.2
0.6 0.6 0.5 0.4 0.6 0.6 0.3 0.7 0.7 0.3 0.6 0.5 0.5
24 5 36 4 20 13 3 13 27 3 9 9
47.4 3.6 58 2.0 41.8 16.6 2.1 12.6 61.5 2.1 7.8 13.2
2.0 0.8 1.6 0.5 2.1 1.29 0.6 1 2.2 0.6 0.6 1.4 1.2
122
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 3 e1 2 4
Fig. 9 e Bulk water iron concentration against a) annual material regeneration and b) mean annual regenerated turbidity.
water suppliers to receive red water complaints due to precipitated ferric oxides. It is therefore proposed that bulk water iron concentrations could potentially be a single measure capturing the dominant influence of the water quality factors such as source water, coagulation treatment processes and presence of unlined upstream iron assets. To investigate this, bulk water iron concentrations (not the concentrations exiting treatment works) for each pipe length have been determined using data from regulatory sampling at the DMAs studied. These are plotted against the annual percentage regeneration, Fig. 9a, and the average regenerated turbidity, Fig. 9b. Although not conclusive with a proportional variance R2 of 0.57, Fig. 9a indicates a potential relationship exists between bulk water iron concentration and material regeneration. The general trend is for regeneration to increase with iron levels, whilst of operational benefit it appears a greater discolouration risk is likely when concentrations rise above 25e40 mgFe/l. This reinforces the findings that water quality is a key factor governing cohesive layer development and suggests that for the UK bulk water monitoring of iron may be used to justify capital investments such as improvements to treatment to lower iron concentrations versus the necessity of operational interventions such as flushing. Although only statistically indicative, linear regression implies that for every microgram per litre increase in iron concentration the annual regeneration increases by just over 1%. It also suggests that even with negligible iron concentration levels, annual regeneration remains at about 15% or effectively 6 years for material layers to reach their maximum discolouration risk. This indicates that although iron may be a primary source of discolouration, there are other contributory factors requiring investigation, such as the prevalence of other metals or microbiological interactions and biofilm development. If the linear regression is forced through the origin of the graph (with a resulting R2 ¼ 0.44), effectively considering iron as a vital component of discolouration material, an increase of just 1 mgFe/l leads to a 1.5% increase in annual regeneration rate. Compared to the regeneration water quality relationship in Fig. 9a,b indicates no relationship between bulk water iron concentrations and
quantity of material mobilised from the pipe wall, R2 ¼ 0.01, matching the result from Table 4 and indicating water quality does not influence the discolouration magnitude.
6.
Discussion
This research demonstrates that discolouration, or the accumulation of cohesive layers that can lead to discolouration upon mobilisation, is an unavoidable and continuous process that occurs throughout UK water distribution systems. The magnitude or severity of a potential discolouration event however may be controlled through maintenance or as a result of network design and operation. Fortunately in some cases the operational hydraulics under the daily demand patterns may already be sufficient to limit the discolouration risk to acceptable levels and this may have led to the general complacency surrounding discolouration. In time water companies who wish to demonstrate pro-active water quality management will effectively need to assign a risk rating for every pipe and this will necessitate an understanding of the controlling factors. Through work such as this these elements may be defined such that each network and its unique characteristics may be taken into account and suitable management strategies imposed.
Fig. 10 e Factors regulating pipe discolouration potential (asset deterioration).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 3 e1 2 4
Primarily this work demonstrates that water quality defines the rate at which discolouration causing layers regenerate. From the sites tested it has been possible to propose deterioration percentages and even suggest target iron concentrations to limit material accumulation. With this type of data it now becomes a possibility for water companies to optimise costs by balancing capital expenditure to improve water quality justified on a quantified risk and reduction in customer complaints. Not surprisingly, due to the particulate nature of discolouration, it is shown that surface waters with the addition of coagulation treatment processes, supply water that leads to the quickest deterioration. Groundwaters however, generally accepted to be of a higher quality, see a reduced rate of deterioration. The presence of additional material sources, such as unlined cast iron pipes has a negative impact on deterioration. Cast iron pipes, most likely due to the direct corrosion processes are shown to deteriorate quicker than other pipes only subjected to the indirect, bulk water, material source. Results show that water quality does not influence the magnitude of a potential discolouration incident. Therefore if a pipe is known to pose a risk, no improvements to water quality will ameliorate this hazard completely, although it may extend the interval prior to an event occurrence. It is therefore proposed that system hydraulics are crucial in determining the magnitude of mobilised turbidity. A basic understanding then would be that a reduction in the discolouration potential may be achieved by increasing the peak daily flow, and hence shear stress on the pipe walls. Network managers may find it possible to change demand and/or flow patterns such that regular peak boundary shear stress is increased to limit the ultimate accumulation of material and the associated risk to customers. These two overarching findings relating to accumulation and ultimate volume of material are encompassed in the conceptual model shown in Fig. 10. In this model water quality is shown to govern the gradient of deterioration while hydraulic conditions define the ultimate state achieved in a particular pipe. The real significance of hydraulic patterns, the conditioning shear stress regimes, and how they affect the accumulation of discolouration material has yet to be thoroughly investigated. It is possible to make assumptions based on examining hydraulic records, yet the number of variables involved, even within the same network, will always be an issue in comparing results. To enable progress in the understanding of the role of hydraulics in discolouration it is suggested that research needs to be conducted within a controlled environment whereby specific, individual factors may be adjusted and the influence evaluated. It may then be possible to extrapolate results to live systems which could see an improvement in management strategies; reducing the risk of water quality failure and subsequent customer complaints, extending maintenance intervals and reducing operational costs.
7.
Conclusions
Fundamentals: Water quality influences the rate of discolouration material regeneration within water distribution systems but shows no significant correlation to magnitude of potential discolouration
123
Bulk water iron concentrations may effectively be used as an indicator of water quality influencing discolouration regeneration/deterioration System hydraulics are significant in determining the magnitude of potential discolouration Network components: On average CI pipes would present a fully regenerated discolouration risk after 2 years B Surface water sourced CI pipes present a fully regenerated discolouration risk after 1½ years B Groundwater sourced CI pipes present a fully regenerated discolouration risk after 3 years On average plastic pipes present a fully regenerated discolouration risk after 4 years Operational: Lower bulk water iron concentration reduced the rate of asset deterioration, with an apparent lower limit of 25 mg/l after which that data becomes scattered as other factors also become significant Higher normally or routinely experienced pipe wall shear stress reduces the magnitude of potential discolouration Looped networks in some instances may create hydraulic flow patterns that favour the development of material layers leading to an increased discolouration risk Network flow patterns may be managed to reduce discolouration risk The hydraulic design of networks to maintain discolouration risk below acceptable thresholds (self-cleaning) is feasible Issues: The relative significance of peak and variable daily hydraulics are not clear The influence of other system variables are not determined, e.g. hardness, chlorine residuals, organic components, microbiological interactions, seasonal variations, pH, temperature, etc.
references
Ackers, J., Brandt, M., Powell, J., 2001. Hydraulic Characterisation of Deposits and Review of Sediment Modelling. Drinking Water Quality and Health e Distribution Systems e DW/03. UKWIR, pp. 109. Barbeau, B., Julienne, K., Carriere, A., Gauthier, V., 2005. Dead-end flushing of a distribution system: short and longterm effects on water quality. Journal Water SRT e Aqua 54, 371e383. Boxall, J.B., Saul, A.J., Skipworth, P.J., 2001. A novel approach to modelling sediment movement in distribution mains based on particle characteristics. In: Ulanicki, B., Coulbeck, B., Rance, J.P. (Eds.), Water Software Systems: V. 1: Theory and Applications (Water Engineering & Management), vol. 1. Research Studies Press, Hertfordshire, UK, pp. 263e273. Boxall, J.B., Saul, A.J., Gunstead, J.D., Dewis, N., 2003. Regeneration of Discolouration in Distribution Systems. ASCE/EWRI/World Water and Environmental Resources Conference, Philadelphia.
124
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 1 3 e1 2 4
Boxall, J.B., Saul, A.J., 2005. Modelling discolouration in potable water distribution systems. Journal of Environmental Engineering ASCE 131 (5), 716e725. Cook, D.M., 2007. Field Investigation of Discolouration Material Accumulation Rates in Live Drinking Water Distribution Systems. Civil and Structural Engineering. Sheffield, Sheffield. Ph.D. DWI Information Letter 15/02 DWI, 2002. In: Rouse, M. (Ed.), Distribution Operation and Maintenance Strategies (DOMS) e DWI Requirements and Expectations. DWI. DWI, 2007. Drinking Water 2006; Drinking Water in England and Wales 2006. A report by the Chief Inspector Drinking Water Inspectorate. Drinking Water Inspectorate, London, pp. 89. DWI, 2008. Drinking Water 2007; Drinking Water in England and Wales 2007. A report by the Chief Inspector Drinking Water Inspectorate. Drinking Water Inspectorate, London, pp. 83. DWI, 2009. Drinking Water 2008; Drinking Water in England and Wales 2008. A report by the Chief Inspector, Drinking Water Inspectorate. Drinking Water Inspectorate, London. Gauthier, V., Barbeau, B., Millette, R., Block, J.-C., Prevost, M., 2001. Suspended particles in the drinking water of two distribution systems. Water Science and Technology: Water Supply 1 (4), 237e245. Hossain, A., Naser, J., McMannus, K., Ryan, G., 2003. CFD investigation of particle deposition and dispersion in a horizontal pipe. In: Third International Conference on CFD in the Minerals and Process Industries, Melbourne, Australia. Husband, P.S., Boxall, J.B., 2008. Water Distribution System Asset Deterioration and Impact on Water Quality e A Case Study. World Environmental and Water Resources Congress, May 2008. ASCE, Ahupua’a, Hawaii. Husband, P.S., Boxall, J.B., Saul, A.J., 2008. Laboratory studies investigating the processes leading to discolouration in water distribution networks. Water Research 42 (16), 4309e4318. Husband, P.S., Boxall, J.B., 2010. Field studies of discolouration in water distribution systems: model verification and practical implications. Journal of Environmental Engineering 136 (1), 86e94. Marshall, G.P., 2001. Understanding and Preventing Discoloured Water. Drinking Water Quality and Health e Distribution Systems DW-03. UKWIR/Thames Water Utilities Limited. pp. 75. McCoy, W.F., Olsen, B.H., 1986. Relationship among turbidity, particle counts and bacteriological quality within water distribution lines. Water Research 20 (8), 1023e1029. Mehta, A.J., Lee, S.-C., 1994. Problems in linking the threshold condition for the transport of cohesionless and cohesive sediment grain. Journal of Coastal Research 10 (1), 170e177.
OFWAT, 2008. In: Dunn, A. (Ed.), PR09/09 OFWAT’S Review of the OPA and Regulation of Service to Consumers. OFWAT PR09. Parchure, T.M., Mehta, A.J., 1985. Erosion of soft cohesive sediment deposits. Journal of Hydraulic Engineering 111 (10), 1308e1326. Rossman, L.A., 2000. In: U.S.E.P. Agency (Ed.), EPANET 2 Users Manual. EPA, Cincinnati. Ryan, G., Jayaratne, A., 2003. Particles in distribution systems and assessment of discoloured water. In: Maintenance and Assessment of Distribution Sstems to Improve Water Quality. C.G. Workshop. Sydney. Sarin, P., Snoeyink, V.L., Bebee, J., Jim, K.K., Beckett, M.A., Kriven, W.M., Clement, J.A., 2004. Iron release from corroded iron pipes in drinking water distribution systems: effect of dissolved oxygen. Water Research 38, 1259e1269. Seth, A., Bachmann, R., Boxall, J., Saul, A.J., Edyvean, R., 2004. Characterisation of materials causing discolouration in potable water systems. Water Science and Technology 49 (2), 27e32. Skipworth, P.J., Tait, S.J., Saul, A.J., 1999. Erosion of sediment beds in sewers: model development. Journal of Environmental Engineering 125 (6), 566e573. Slaats, N., Rosenthal, L.P.M., Siegers, W.G., Boomen, M.v. d., Beuken, R.H.S., Vreeburg, J.H.G., 2002. Processes Involved in the Generation of Discolored Water. KOA 02.058. Kiwa. American Water Works Association/Kiwa, The Netherlands, pp. 116. Smith, S.E., Holt, D.M., Delanoue, A., Colbourne, J.S., Chamberlain, A.H.L., Lloyd, B.J., 1999. A pipeline testing facility for the examination of pipe wall deposits and red-water events in drinking water. Journal of Chartered Institution of Water and Environmental Management 13, 7e15. Twort, A.C., Ratnayaka, D.D., Brandt, M.J., 2000. Water Supply. Arnold, London. Verberk, J.Q.J.C., Hamilton, L.A., O’Halloran, K.J., Van Der Horst, W., Vreeburg, J., 2006. Analysis of particle numbers, size and composition in drinking water transportation pipelines: results of online measurements. Water Science & Technology: Water Supply 6 (4), 35e43. Vreeburg, J.H.G., Schippers, D., Verberk, J.Q.J.C., van Dijk, J.C., 2008. Impact of particles on sediment accumulation in a drinking water distribution system. Water Research 42 (16), 4233e4242. Vreeburg, J.H.G., Boxall, J.B., 2007. Discolouration in potable water distribution systems: a review. Water Research 41, 519e529. WHO, 2005. In: Davison, A., Howard, G., Stevens, M., et al., (Eds.), Water Safety Plans Managing Drinking-water Quality from Catchment to Consumer. World Health Organization, Geneva, p. 244.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 2 5 e1 3 4
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Kinetics of electro-oxidation of ammonia-N, nitrites and COD from a recirculating aquaculture saline water system using BDD anodes V. Dı´az a, R. Iba´n˜ez a, P. Go´mez b, A.M. Urtiaga a, I. Ortiz a,* a b
Dpto. Ingenierı´a Quı´mica y QI. ETSIIyT, Universidad de Cantabria, Av. de los Castros s/n, 39005 Santander, Spain APRIA Systems S.L., Polı´gono Trascueto s/n, 39600 Camargo, Spain
article info
abstract
Article history:
The viability of the electro-oxidation technology provided with boron doped diamond
Received 15 June 2010
(BDD) electrodes for the treatment and reuse of the seawater used in a Recirculating
Received in revised form
Aquaculture System (RAS) was evaluated in this work.
31 July 2010
The influence of the applied current density (5e50 A m2) in the removal of Total Ammonia
Accepted 10 August 2010
Nitrogen (TAN), nitrite and chemical oxygen demand (COD) was analyzed observing that
Available online 25 August 2010
complete TAN removal together with important reductions of the other considered contaminants could be achieved, thus meeting the requirements for reuse of seawater in
Keywords:
RAS systems.
Aquaculture saline water reuse
TAN removal, mainly due to an indirect oxidation mechanism was described by a second
BDD anode
order kinetics while COD and nitrite removal followed zero-th order kinetics. The values of
Electro-oxidation
the kinetic constants for the anodic oxidation of each compound were obtained as
Nitrogen compounds
a function of the applied current density (kTAN ¼ 7.86 105$exp(6.30 102 J);
COD
kNO2 ¼ 3.43 102 J; kCOD ¼ 1.35 102 J). The formation of free chlorine and oxidation byproducts, i.e., trihalomethanes (THMs) was followed along the electro-oxidation process. Although a maximum concentration of 1.7 mg l1 of total trihalomethanes was detected an integrated process combining electrochemical oxidation in order to eliminate TAN, nitrite and COD and adsorption onto activated carbon to remove the residual chlorine and THMs is proposed, as an efficient alternative to treat and reuse the seawater in fish culture systems. Finally, the energy consumption of the treatment has been evaluated. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Aquaculture is a value-added economic activity worldwide. This sector has become a necessity to meet the demand for fish and seafood. Aquaculture operations require large quantities of feed water that is contaminated with three primary contaminants that are often regulated: organic matter, nutrients, such as phosphorous and nitrogen, in the forms of Total Ammonia
Nitrogen (TAN), nitrite and nitrate and solids. Bacteria and pathogens are also waste products that must be controlled (Davidson et al., 2008). Moreover, legal regulations concerning the discharge of effluents from fish farms have become increasingly strict, therefore, one of the challenges facing the aquaculture industry is the treatment of the culture water. Traditional aquacultural activities are performed in large ponds with low fish density. But this way of culture requires
* Corresponding author. Tel.: þ34 942 20 15 85; fax: þ34 942 20 15 91. E-mail address:
[email protected] (I. Ortiz). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.020
126
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 2 5 e1 3 4
large amounts of land and water resources. Due to increasing cost and limited suitable water supplies, the traditional aquacultural activities are becoming relatively uneconomic. An emerging recent approach is to employ high density aquaculture which requires significantly less amount of land and water resources than conventional methods. High density aquaculture operations that mainly consist of Recirculating Aquaculture Systems (RAS) based fish tanks where the fish tank effluent is biologically treated and the water is recycled back to the rearing tanks, are becoming key solution for the control of contaminants present in the closed culture and keep the quality of water needed for large-scale fish production. In all aquaculture systems a rapid TAN accumulation caused primarily by fish excretion and decomposition of uneaten feed, is found. Being this problem especially important in the RAS as an increase in fish density in a limited aquatic space leads to a more rapid degradation of the water quality. Besides, in the recirculating systems, nitrite is also found as an intermediate of the ammonia nitrification. Moreover, biofiltration is significantly less effective in saline conditions than in freshwater and it is subject to biological upsets and wide fluctuations in performance as a consequence of the dynamic properties of this system. Since unionized ammonia and nitrite are toxic to fish even at low concentrations, the introduction of new technologies is necessary to avoid the accumulation of these toxic compounds and achieve an effective removal, thus allowing the reuse of the treated water in the fish culture systems and reducing environmental problems and operating costs. The treatment of fish culture water for removal of ammonia and nitrite has been investigated by numerous researchers based on the following methods: selective ion exchange (Miladinovic and Weatherley, 2008), flocculation (Ebeling et al., 2005), sand beds (Palacios and Timmons, 2001) and membrane bioreactors (Sharrer et al., 2007; Pulefou et al., 2008), among others. In recent years, oxidation of nitrogen compounds by chemical methods, e.g. ozone (Krumins et al., 2001; Tango and Gagnon, 2003), UV (Bronk et al., 2000) and breakpoint chlorination (Potts and Boyd, 1998) have received significant attention. Although these methods have been proven to be capable of removing nitrite and ammonia, they also have several limitations, such as, high cost of operation and instrumentation and the frequent use of reagents (Sun and Chou, 1999). Recently, electrochemical oxidation has been considered as a promising alternative for the treatment of polluted waters containing nitrogen compounds due to its advantages such as minimal generation of secondary wastes, easy operation and remote control. Besides, this treatment is not subject to failure due to the presence of high salinity and variation in wastewater flow (Vijayaraghavan et al., 2008). In recent literature, relevant works can be found on the application of this technology to the mineralization of organic compounds and ammonia contained in wastewaters (Szpyrkowicz et al., 2001; Szpyrkowicz and Radaelli, 2006; Urtiaga et al., 2009; Li and Liu, 2009). In the context of aquaculture saline water, electrochemical oxidation has not received much attention yet. However, the electrochemical treatment of seawater presents several advantages, as high salinity ensures an excellent electric conductivity that could reduce the energy consumption and the
high chloride concentration improves the indirect oxidation through the electro-generation of strong oxidants like hypochlorous acid. Additionally, the electrochemical treatment avoids the handling of toxic chlorine gas used in the breakpoint chlorination process, since the oxidants are electrolytically generated in situ, and this generation can be controlled by the electrolysis operation conditions (Alfafara et al., 2004). Several authors have studied the influence of the operating conditions of the electrochemical treatment like pH, conductivity, initial concentration or applied current density in the oxidation of nitrite, ammonia and/or organic matter using synthetic freshwater (Lin and Wu, 1996; Abuzaid et al., 1999), prepared seawater (Lin and Wu, 1997; Sun and Chou, 1999; Lee et al., 2002; Wijesekara et al., 2005) or raw wastewater collected from fish farms (Katayose et al., 2007; Vijayaraghavan et al., 2008). This work is focused on the viability and kinetics of the application of electrochemical oxidation to enhance the seawater quality of a Recirculating Aquaculture System by removing TAN, nitrite and COD offering an alternative to the biological filter currently operated. Formation of by-products, total trihalomethanes and free chlorine and the energy consumption have also been evaluated. This study treats seawater collected from the inlet of the biological filter situated in a high density hatchery. A laboratory set-up provided with Boron Doped Diamond (BDD) electrodes has been used. Although, there are no previous references that concern with electrochemical oxidation of seawater from a hatchery using BDD electrodes, this material presents many advantages compared to other anodic materials reported in literature, such as, graphite, titanium alloys or platinum (Rychen et al., 2003; Chen et al., 2003; Cabeza et al., 2007; Chatzisymeon et al., 2009; Polcaro et al., 2009; Pe´rez et al., 2010).
2.
Materials and methods
2.1.
Materials
Seawater from a sea bream hatchery located in Cantabria (Northern coast of Spain) was employed in this work. A previous characterization of the Recirculating Aquaculture System (RAS) installed in this fish farming, allowed to determine that the TAN concentration ranged daily in the interval of 0.05e8.00 mg l1. This wide range of concentration is due to the fish metabolism and the hatchery production requirements. Laboratory scale experiments were performed using the hatchery water doped with ammonium chloride to give an initial TAN concentration of 8 mg l1. The initial physicochemical characteristics (measured according to analytical methods described in the following section) of the seawater used in all experiments are shown in Table 1. Similar characterization was found in the literature about water collected from a high density fish hatchery (Taparhudee et al., 2008). The high conductivity allows the direct application of electrochemical techniques.
2.2.
Laboratory-scale experimental setup
The experiments were performed in a two-compartments electrochemical cell (DiaCell 201 PP) supplied by Adamant
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 2 5 e1 3 4
Table 1 e Initial physicochemical characterization of seawater used in this study. Parameter pH Redox Potential (mV) Conductivity (mS cm1) Salinity (&) Temperature ( C) COD (mg O2 l1) TAN (mg l1) Nitrite (mg l1) Nitrate (mg l1) Chloride (mg l1) Sulphate (mg l1) Total Trihalomethanes (mg l1) Free chlorine (mg l1)
Average value 6.9 31 51.1 32.2 24.7 54.8 8.00 80.00 403.20 26,167 35,000 2.25 0.00
Technologies. Monopolar circular Boron Doped Diamond (BDD) on silicon anode and cathode, with a surface area of 70 cm2 each, an interelectrode gap of 1 mm and a bipolar electrode coated with diamond on both sides and placed between monopolar electrodes were inserted in the cell. Fig. 1 shows the basic layout of the laboratory scale plant. Galvanostatic conditions were applied by using an Agilent power supply 6554A (with a maximum output of 9 A and 60 V). The range of the applied current density, J, was between 5 and 50 A m2. A volume of 2 L of seawater was treated in each experiment. The feed tank was refrigerated in order to maintain the working temperature at 25 2 C. A magnetic pump was used to recirculate the feed from the tank through the cell, at a flow rate of 6 l min1 per compartment. This value of flow rate was selected according to the technical specifications of the electrochemical cell, which depend on the number of compartments and the distance between anode and cathode. The experiments were finished when TAN concentration was lower than 0.06 mg l1 corresponding the detection limit of the analytical method used in this work.
2.3.
127
The pH was measured with a Crison pH 25 pH meter and the conductivity and the salinity were measured with a Crison CM 35 conductivity meter. TCOD was determined by heat of dilution COD procedure (Ruttanagosrigit and Boyd, 1989) employing mercuric sulphate to remove chloride interference. The concentration of TAN, nitrite, nitrate and chloride in solution was measured spectrophotometrically by using a Spectroquant Pharo 100, (Merck Company) according to Standard Methods (APHA, 1998): 4500-NH3-D, 4500-NO2-B, 4500-NO3-B and 4500-Cl-E, respectively. Sulphate was measured using ion chromatography (Dionex 120 IC, with an IonPac AS9-HC Column). Additionally, characterization of electrochemical oxidation by-products, such as, chlorine and trihalomethanes (THMs), was performed. Total and free chlorine were analyzed using a pocket chlorimeter (HI 95734, Hanna Instruments Company) according to DPD (N, N-diethyl-p-phenylenediamine) method. The THMs, chloroform (CHCl3), bromodichloromethane (CHBrCl2), bromoform (CHBr3) and dibromochloromethane (CHBr2Cl) were determined employing Direct Aqueous Injection Gas Chromatography with electron capture detection (HP 6890 Series GC Systems, with Headspace injector 7694 Agilent), following the Standard Methods 6232C (APHA, 1998), LiquideLiquid Extraction gas chromatographic method with modifications of the GC column and temperature heating rate. An HP-1 methyl siloxane column (30 mm 0.53 mm i.d., with 2.65 mm thickness) was selected in order to perform the analysis, and temperature was as follows: 50 C heating at 10 C min1 up to 150 C and isothermal conditions for 5 min. All analytical determinations were performed immediately after sampling and were done by replicate, except THMs measurement. For this determination, samples were stocked at 4 C within 48 h after sampling.
3.
Results and discussion
3.1.
Kinetics of TAN oxidation
Analytical methods
Seawater samples were withdrawn at regular time intervals from the feed tank using a syringe and they were characterized by means of the following procedures.
Fig. 2 shows the influence of the applied current density on TAN removal in the J range 5e50 A m2. For better comparison of the results, dimensionless data have been plotted, since the initial TAN concentration in the feed water slightly varied
Fig. 1 e Experimental Setup: 1, refrigerated feed tank; 2, pump; 3, two-compartments electrochemical cell; 4, power supply.
128
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 2 5 e1 3 4
1.00 0.90
TAN/TAN0
0.80 0.70 0.60 0.50 0.40 0.30 0.20 0.10 0.00 0
10
20
30
40
50
60
70
80
90 100 110 120 130
Time (minutes) Fig. 2 e Influence of the applied current density (A J [ 5 A mL2; - J [ 10 A mL2; : J [ 20 A mL2; C J [ 30 A mL2; > J [ 50 A mL2) on the evolution of TAN/ TAN0; [TAN]0 z 8.0 mg/l.
around the value of 8 mg l1. It can be observed in the Fig. 2 that increasing values of J enhanced the removal of TAN, as also shown in literature (Lin and Wu, 1997; Lee et al., 2002; Cabeza et al., 2007; Anglada et al., 2009; Li and Liu, 2009). Fig. 2 shows that a value of the current density of J ¼ 5 A m2 fails to significantly remove TAN. This fact can be explained according to the works of Cabeza et al. (2007). In that work, the authors demonstrated that if ammonia and COD have to be eliminated simultaneously, current densities higher than the Jlim,COD have to be applied, being Jlim,COD, defined as the current density value that implies null accumulation of oxidizable substances at the surface of the anode. In this work, the value of Jlim,COD is 13.09 A m2. On the other hand, in the range 10e50 A m2 complete TAN removal (remaining concentration below detection limit of 0.06 mg l1 of TAN) was achieved. At J ¼ 10 A m2, a value of current density close to the Jlim,COD, complete TAN removal was achieved after 115 min. As reported by Deng and Englehardt (2007), the rule of competition between the removal of COD and the removal of ammonia seems to be that the removal of ammonia is favored when indirect oxidation is dominant, whereas the removal rate of COD takes priority under direct anodic oxidation. Ammonia oxidation takes place due to an indirect oxidation through the electro-generation of hypochlorous acid, according to a mechanism analogous to the breakpoint chlorination reactions (Szpyrkowicz et al., 2001; Lee et al., 2002; Li and Liu, 2009; Anglada et al., 2010). Besides, some authors (Alfafara et al., 2004; Wijesekara et al., 2005) consider other oxidants, such as hypobromous acid as well as hypochlorous acid, although HBrO is not as efficient as HClO. Taking into account that in the normal pH range of pond water (6e7.5), more than 98% of TAN is in the form of NHþ 4, being HOCl the major component of free chlorine in this pH interval, the overall reaction occurring between HClO and NHþ 4 is shown by equation (1).
þ 2NHþ 4 þ 3HClO/N2 þ 5H þ 3Cl þ 3H2 O
(1)
A second order kinetics has been proposed by Xu et al. (2008) and Anglada et al. (2009) to describe the ammonia oxidation
by free chlorine. Szpyrkowizc and Radaelli also proposed a second order to describe the kinetics of decolourisation via indirect electro-oxidation. A zero-th order kinetics (linear profiles) has been proposed when working with high values of the applied current density, whereas curved evolutions with an initial delay are observed when working at lower values of the applied current density (Cabeza et al., 2007). In the present work, the experimental data regarding TAN disappearance versus time shown in Fig. 2 were fitted to equation (2): d½TAN ¼ k$½TAN$½HOCl dt
(2)
where, k is the rate constant (l mg1 min1) and [HOCl] is the concentration of hypochlorous acid (mg l1). Assuming that the rates of chlorine disappearance due to i) cathodic reduc tion of active chlorine, ii) anodic oxidation to ClO3 and, iii) homogeneous reaction with TAN were much lower than the chlorine production rate in the anode, given by the high chloride concentration present in seawater, the variation of chlorine concentration with time (until saturation) can be described by equation (3): d½HOCl 4 A J ¼ dt nFV
(3)
where F is the Faraday constant (96485 C mol1) and 4 is the current efficiency for chlorine evolution, which depends on the applied current density, mass transport rate coefficient and chloride concentration (Anglada et al., 2009). In this work, it is assumed that the chloride concentration remains constant throughout the oxidation process and that free chlorine evolution is the main anodic reaction. Consequently, the substitution of the integrated form of equation (3) into equation (2), and subsequent integration gives equation (4): ln
½TAN k4AJt2 ¼ k0 t2 ¼ 4FV ½TAN0
(4)
Values of the apparent rate constant, k0 (min2) presented in Table 2 were obtained from the slopes of the logarithms of [TAN]t/[TAN]0 vs. the square of the electrolysis time, for the different applied current densities. The values of the correlation coefficients were high enough in all cases (Table 2). The apparent rate constant of TAN (k0 ) increases exponentially with the applied current density according to equation (5), obtained from the fitting of the data given in Table 2. 2 R ¼ 0:997 k0 ¼ 7:86 105 $exp 6:30 102 J
(5)
Combining equations (4) and (5), equation (6) was obtained:
Table 2 e Values of k0 for TAN oxidation at the different applied current densities. J (A m2) 10 20 30 50
k0 (min2)
R2
1.55 e4 2.78 e4 4.71 e4 1.92 e3
0.991 0.953 0.965 0.979
129
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 2 5 e1 3 4
Equation (6) is the second order kinetic model developed to describe the kinetics of TAN removal along the electrooxidation treatment of seawater from a hatchery as a function of the operation variable, J, in the range 10e50 A m2 and considering indirect oxidation mediated by electrogenerated active chlorine. A parity graph of the dimensionless TAN concentration for all the performed experiments is shown in Fig. 3. Good agreement between experimental and simulated data is observed, because 80% of the results of Csim fall within the Cexp 15% range.
3.2.
Kinetics of nitrite oxidation
Nitrite concentration along electrochemical experiments was measured according to the procedure described in the previous section. The experiments were performed with an average initial nitrite concentration of 80.0 mg l1, as received from the hatchery. Fig. 4 shows the normalized nitrite concentration profiles during the electrochemical treatment of seawater at different applied current densities, in the range 5e50 A m2. More than 90% of the initial nitrite concentration was removed in the whole range of applied current densities, by keeping the adequate operating time. It is observed that the nitrite removal rate is highly affected by the applied current density. Concerning nitrite removal, a mechanism based on the indirect oxidation reaction between nitrite and the oxidant, HClO, generated at the anode, has been reported in literature (Sun and Chou, 1999) as shown by equation (7):
þ NO 2 ðaqÞ þ HClOðaqÞ /NO3 ðaqÞ þ ClðaqÞ þ HðaqÞ
(7)
Few works have been found in literature describing the influence of the applied current density on the kinetics of nitrite electro-oxidation (Lin and Wu, 1996, 1997). Sun and
1.0
0.90 0.80 0.70 0.60 0.50 0.40 0.30 0.20 0.10 0.00
0
40
80
120
160
200
240
0.8
280
Time (minutes) Fig. 4 e Influence of the applied current density (A J [ 5 A mL2; : J [ 20 A mL2; C J [ 30 A mL2; > J [ 50 A mL2) on the evolution of L L ½NOL 2 =½NO2 0 ; ½NO2 z80:0 mg=l.
Chou (1999) used the ButlereVolmer equation to define the effect of current density (J ) on nitrite removal kinetics. In that work, the nitrite oxidation rate was expressed according to equation (8): d NO 2 ¼ dt
! kK NO 2 1 þ K NO 2
(8)
Where k and K are reaction rate constants expressed in mol L1 min1 and L mol1, respectively. In the present study equation (8) has been used to describe the behaviour of nitrite in our system. For an initial nitrite concentration of 80 mg l1 and employing a value of K of 2.48 104 l mol1 obtained by Sun and Chou (1999) for a similar system, the resulting K½NO 2 value is 43.13, being 1 < < K½NO 2 and therefore, equation (8) can be simplified to equation (9). As shown in equation (9), the nitrite oxidation rate is described by a zero-th order expression, in the range of variables under study. Lin and Wu (1997) also described the nitrite oxidation by means of a zero-th order kinetics. d NO 2 ¼k dt
0.9
Simulated Value
1.00
(6)
NO2-/NO2-0
½TAN ln ¼ 7:86 105 $exp 6:30 102 J $t2 ½TAN0
(9)
The integrated form of equation (9) gives equation (10):
0.7
NO k 2 $t ¼ 1 k0 $t ¼ 1 NO NO 2 0 2 0
0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Experimental Value Fig. 3 e Parity graph: dimensionless TAN concentration at the different applied current densities (- J [ 10 A mL2; : J [ 20 A mL2; C J [ 30 A mL2; > J [ 50 A mL2).
(10)
Values of the apparent rate constant, k0 (min1) reported in Table 3 were obtained from the slopes of the dimensionless nitrite concentration vs. the electrolysis time (Fig. 4) for the different applied current densities. In all cases correlation coefficients higher than 0.980 were obtained (Table 3). Table 3 also includes the values of k (mg l1 min) calculated from the values of k0 and the initial nitrite concentration (80 mg l1). It is observed, that an increase of the current density leads to a significant increase in the nitrite removal specific rate. The kinetic constant of nitrite removal (k) increases linearly with the applied current density according to equation (11). Similar relationship between k and J was reported in the works of Lin and Wu (1997).
130
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 2 5 e1 3 4
1.00
Table 3 e Values of k0 and k for nitrite oxidation at the different applied current density. 2
J (A m2)
k (min1)
R
5 20 30 50
3.44 e3 8.99 e3 1.32 e2 2.11 e2
0.995 0.998 0.997 0.987
½NO 2 0 1
(mg l ) 80.0
0.90 0.80
k
(mg l
1
min)
0.272 0.720 1.056 1.688
COD/COD0
0
0.70 0.60 0.50 0.40 0.30 0.20 0.10 0.00
k ¼ 3:43 102 J
R2 ¼ 0:998
(11)
A parity graph of the dimensionless nitrite concentration for all the performed experiments is shown in Fig. 5. Good agreement between experimental data and simulated values obtained from equations (10) and (11) is observed, since more than 90% of the results of Csim fall within the Cexp 10% range. The values of the zero-th kinetics constants of nitrite oxidation working with real seawater from a hatchery are higher than those obtained by Lin and Wu (1997) who oxidized nitrite from synthetic seawater samples, using graphite anodes and operating at current density between 68 and 182 A m2.
3.3.
Kinetics of COD oxidation
Chemical Oxygen Demand (COD) of pond waters may be used as an index of the organic matter concentration. In the presence of organic matter, the oxidants (HOBr and HOCl) generated during the electrolysis process will also be consumed by oxidation of organic matter (Westerhoff et al., 2004; Wijesekara et al., 2005). Fig. 6 shows the influence of the applied current density in the range 5e50 A m2 on COD removal starting from an average initial COD concentration of 54.8 mg l1. COD removal follows a linear trend with time in the whole range of J
40
80
120
160
200
240
280
Time (minutes) Fig. 6 e Influence of the applied current density (A J [ 5 A mL2; : J [ 20 A mL2; C J [ 30 A mL2; > J [ 50 A mL2) on the evolution of COD/COD0 COD0 z 54.8 mg lL1.
considered, concluding that the higher the current density, the faster the removal of COD. Similar behaviour on COD removal with time, operating at galvanostatic conditions has been reported in the works of Vijayaraghavan et al. (2008); treating raw aquaculture wastewater from a shrimp farm with a initial COD concentration of 1730 mg l1 at 372 and 745 A m2 and using a graphite anode, achieved a residual COD concentration at the end of experiments around 50 mg l1. The initial COD concentration (54.8 mg l1) used in the present work is representative of the sea bream production on hatcheries. Comninellis and co-workers (Panizza et al., 2001) developed one of the most cited models to describe the electrochemical oxidation of organic pollutants, COD, on BDD electrodes (Can˜izares et al., 2005; Cabeza et al., 2007). In this model, two different operating regimes are defined, depending on the value of the applied current density, Jappl, and the value of the limiting current density, Jlim: a) If Jappl < Jlim, the electrolysis is under current control and the COD decreases linearly with time. b) If Jappl > Jlim, the electrolysis is under mass transport control and the COD evolution follows an exponential trend with time.
1.0 0.9 0.8
Simulated Value
0
0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Experimental Value Fig. 5 e Parity graph: dimensionless Nitrite concentration at the different applied current densities (: J [ 20 A mL2; C J [ 30 A mL2; > J [ 50 A mL2).
Nevertheless, in those matrices where chloride is the major compound like in seawater, chlorine is also formed during the electrochemical process, thus, consuming a high percentage of the applied current. In the present work where [Cl]0 is around 500 times higher than [COD]0, most of the Jappl is employed in the oxidation of chloride. According to the latter, although the total applied current density in this work (range 5e50 A m2) is higher than Jlim,COD (13.3 A m2), the effective applied current density for COD oxidation is just a fraction of the total applied current density, suggesting an operating regime under current density control. As shown in the experimental data plotted in Fig. 6, a linear decrease in the normalized COD removal with time is observed, confirming the previous hypothesis. Therefore,
131
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 2 5 e1 3 4
equation (12) can be used to describe the evolution of COD during the electrochemical treatment at the different applied current densities.
(13)
Values of the kinetic constant, k, corresponding to the different values of J applied are shown in Table 4. These values were calculated from the slopes (k0 ) of the dimensionless COD concentration vs. the electrolysis time (Fig. 6) for the different values of the applied current density and the initial COD concentration (54.8 mg l1). In all experiments, correlation coefficients higher than 0.980 were obtained. The data in Table 4 show a linear relation between k and J expressed by equation (14): k ¼ 1:35 102 J
R2 ¼ 0:985
(14)
Consequently, the substitution of equation (14) in equation (13) results in equation (15): COD ¼ COD0 1:35 102 J t
(15)
Equation (15) describes the evolution of COD concentration as a function of the current density, in the studied range 5e50 A m2, during the electrochemical oxidation of such a complex matrix as seawater from a hatchery. A parity graph of the dimensionless COD concentration for all the performed experiments is shown in Fig. 7. Good agreement between experimental and simulated data is observed, since more than 90% of the results of Csim fall within the Cexp 10% range.
3.4. Formation of chlorine and organochlorinated compounds During the electrochemical oxidation of seawater, chlorine gas generated on the anode is converted into hypochlorous acid (HOCl), as shown by equations (16) and (17). The sum of the three species: dissolved gas chlorine (Cl2), hypochlorous acid (HOCl) and hypochlorite ion (OCl) is termed free chlorine. In the normal pH range of pond water (6e7.5), HOCl is the major component of free chlorine.
2Cl /Cl2 þ 2e
(16)
Cl2 þ H2 O/HOCl þ Cl þ H
þ
(17)
Table 4 e Values of k for COD oxidation at the different applied current density.
5 20 30 50
k0 (min1) 1.32 5.78 6.91 1.24
e3 e3 e3 e2
½COD0 (mg l1)
k (mg l1 min)
54.8
0.072 0.316 0.379 0.679
Simulated Value
½COD k ¼1 $t ¼ 1 k0 $t ½COD0 ½COD0
0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Experimental Value
Fig. 7 e Parity graph: dimensionless COD concentration at different applied current densities (A J [ 5 A mL2; : J [ 20 A mL2; C J [ 30 A mL2; > J [ 50 A mL2).
The evolution of free chlorine concentration within the electro-oxidation time is shown in Fig. 8, for an applied current density in the range 5e50 A m2. As expected, the results show that chlorine concentration increases with electrolysis time and applied current density, in the range of J values considered. An increase in chlorine formation leads to a higher efficiency on the electrochemical oxidation when the indirect oxidation is the main oxidation mechanism. This behaviour has been also considered in the works of Lee et al. (2002), Katayose et al. (2007), Taparhudee et al. (2008) and Vijayaraghavan et al. (2008). On the other hand, an undesirable effect of free chlorine is the formation of organochlorinated compounds. Trihalomethanes (THM) are the most commonly halogenated byproducts found after water electro-oxidation. HOCl generated
9.00
Free Chlorine (mg l-1)
(12)
The integrated form of equation (12) gives equation (13):
J (A m2)
0.9 0.8
d½COD ¼ k dt
1.0
7.50 6.00 4.50 3.00 1.50 0.00 0
10
20
30
40
50
60
70
80
90 100 110 120
Time (minutes) Fig. 8 e Influence of the applied current density (A J [ 5 A mL2; - J [ 10 A mL2; : J [ 20 A mL2; C J [ 30 A mL2; > J [ 50 A mL2) on the evolution of free chlorine.
132
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 2 5 e1 3 4
a
b
1.40
1.60 1.40
THM (mg l-1)
TTHM (mg l-1)
1.20 1.00 0.80 0.60 0.40 0.20 0.00
CHBr3
1.20 1.00
CHBr2 Cl
0.80 CHBrCl2
0.60 0.40
CHCl3
0.20 0 10 20 30 40 50 60 70 80 90 100
0.00
10
20
30
50
J (A m-2)
Time (minutes)
Fig. 9 e THM formation during electrochemical oxidation: (a) influence of the applied current density (- J [ 10 A mL2; : J [ 20 A mL2; C J [ 30 A mL2; > J [ 50 A mL2) on the evolution of TTHM concentration; (b) Individual THM concentrations at the end of each electro-oxidation experiment.
0.80 0.60 0.40 0.20
Energy consumption
The technical feasibility of the electrochemical oxidation is usually evaluated in terms of the percentage removal of pollutant reached, while the economic feasibility is determined by the energy consumption. The cumulative energy consumption is shown against the percentage of TAN eliminated in Fig. 10a, where higher energy consumption is required to obtain higher TAN removal, for all the J values considered. Operation at 10 A m2 yielded the lowest energy consumption, whereas similar profiles were obtained at 30 and 50 A m2. Fig. 10b shows the time for total TAN removal and the corresponding energy consumption to these removals, at the different applied current densities. As shown in Fig. 10b the energy curve has a clear increase up to a value of J of 20 A m2. This increase shows lower slopes for current densities higher
b
1.00
0.00 0.0
3.5.
Time for total TAN removal (min)
W (kW h m-3)
a
Excess of chlorine and the THMs are harmful to aquatic organisms (Lee et al., 2002; Katayose et al., 2007; Taparhudee et al., 2008). However this problem can be minimized by adsorbing these compounds onto activated carbon. This alternative is currently under investigation in our group.
140 120 100 80 60 40 20 0
20.0
40.0
60.0
80.0
TAN eliminated (%)
100.0
0
10
20
30
40
50
0.90 0.80 0.70 0.60 0.50 0.40 0.30 0.20 0.10 0.00 60
W (kW h m-3)
in the electrochemical oxidation of seawater reacts with dissolved organics resulting in the formation of chlorinated halomethanes. HOCl also oxidizes the bromide anions (Br) present in seawater, that react readily with organic matter to form brominated halomethanes. In the present study four organic halogenated compounds have been detected: bromoform, dibromochloromethane, bromodichloromethane and chloroform. Fig. 9a shows the influence of the applied current density in the range 10e50 A m2 on the evolution of Total Trihalomethanes (TTHM) concentration, resulting from the sum of these four organic halogenated compounds considered. The TTHM concentration increased with the increase in the available chlorine concentration, formed during electrochemical oxidation, whatever the value of the applied current density is. Fig. 9b shows the concentrations of the organic halogenated compounds at the end of each experiment. In all experiments, bromoform was the predominant by-product formed with a mean weight percentage of 85.3%, followed by dibromochloromethane (10.2%) and traces of bromodichloromethane (3.1%) and chloroform (1.4%) were also detected. The high levels of brominated by-products are attributed to the presence of bromide ions in seawater (Allonier et al., 2000; Budziak et al., 2007; Katayose et al., 2007).
J (A m-2)
Fig. 10 e Energy consumption during electrochemical oxidation: (a) Cumulative energy consumption profiles against the percentage of TAN eliminated (- J [ 10 A mL2; : J [ 20 A mL2; C J [ 30 A mL2; > J [ 50 A mL2); (b) Time and energy consumption profiles to achieve total TAN removal.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 2 5 e1 3 4
than 20 A m2. The calculated energy consumption to achieve complete oxidation of TAN operating at 10, 20, 30 and 50 A m2 was 0.33, 0.61, 0.70 and 0.78 kWh m3, respectively. These consumptions are lower than the range from 12 to 21 kWh m3 found by Lin and Wu (1996), who oxidized nitrite using graphite anodes and operating in the range of J from 442 to 1106 A m2. According to these results, a compromise must be reached between maximizing the oxidation rate and minimizing energy consumption.
4.
Conclusions
This work reports the electro-oxidation viability and kinetics of the organic matter and nitrogen compounds contained in the seawater of a Recirculating Aquaculture System (RAS). Seawater collected from the inlet to the biological filter operated in a high density hatchery with average initial concentrations [TAN] ¼ 8.00 mg l1, [NO2] ¼ 80.00 mg l1, [COD] ¼ 54.80 mg l1 and [Cl] ¼ 26167 mg l1 has been oxidized in an electro-oxidation cell working in the range of current densities 5e50 A m2, and observing that this variable exerts strong influence on the removal kinetics of the considered parameters. A second order expression correlated satisfactorily well the kinetic data of TAN removal, whereas the kinetic evolution of nitrite and COD needed zero-th order equations being the kinetic constants for the anodic oxidation dependent on the applied current density (kTAN ¼ 7.86 105$exp(6.30 102 J ); kNO2 ¼ 3.43 102 J; kCOD ¼ 1.35 102 J ). The evaluation of the energy consumption by the oxidation process showed a dependency with the applied current density up to a value of J ¼ 30 A m2; above this value no apparent sensitivity of the variable on the energy consumed was observed. Thus this work reports for the first time the kinetics of oxidation of TAN, nitrites and COD contained in seawater of a high density hatchery that are needed for process design. A comparison of the results with previous works using graphite anodes already referred in this work, showed a considerable improvement both on the oxidation kinetics as well as on the reduction of the consumed energy. Finally the formation of THMs during the electrochemical treatment has been evaluated, detecting a maximum concentration of 1.7 mg l1 of total trihalomethanes. A hybrid process that couples an adsorption step onto activated carbon to the electro-oxidation cell in order to remove the generated THM and the residual chlorine is currently under operation in the hatchery facilities.
Acknowledgements Financial support of projects CTQ2008-03225/PPQ, CTQ200800690/PPQ, Consolider CSD 2006-44 (Spanish Ministry of Science and Innovation (MICINN)), 080/RN08/03.2 (Spanish MARM) and 18-04-2007 (SODERCAN, Cantabria Government) is gratefully acknowledged. The collaboration of Tinamenor S.L.
133
is also acknowledged. V. Dı´az would like to thank the MICINN for an FPI research grant.
references
Abuzaid, N.S., Al-Hamouz, Z., Bukhari, A.A., Essa, M.H., 1999. Electrochemical treatment of nitrite using stainless steel electrodes. Water Air Soil Pollut. 109 (1e4), 429e442. Alfafara, C.G., Kawamori, T., Nomura, N., Kiuchi, M., Matsumura, M., 2004. Electrolytic removal of ammonia from brine wastewater: scale-up, operation and pilot-scale evaluation. J. Chem. Technol. Biotechnol. 79 (3), 291e298. Allonier, A.-S., Khalanski, M., Bermond, A., Camel, V., 2000. Determination of trihalomethanes in chlorinated sea water samples using a purge-and-trap system coupled to gas chromatography. Talanta 51 (3), 467e477. Anglada, A., Iba´n˜ez, R., Urtiaga, A., Ortiz, I., 2010. Electrochemical oxidation of saline industrial wastewaters using boron-doped diamond anodes. Catal. Today 151 (1e2), 178e184. Anglada, A., Urtiaga, A., Ortiz, I., 2009. Pilot scale performance of the electro-oxidation of landfill leachate at boron-doped diamond anodes. Environ. Sci. Technol. 43 (6), 2035e2040. APHA, 1998. Standard Methods for Examination of Water and Wastewater, twentieth ed. American Public Health Association, Washington, DC. Bronk, D.A., Lomas, M.W., Glibert, P.M., Schukert, K.J., Sanderson, M.P., 2000. Total dissolved nitrogen analysis: comparisons between the persulfate, UV and high temperature oxidation methods. Mar. Chem. 69 (1e2), 163e178. Budziak, D., Richard Jr., L., Beltrame, E., Carasek, E., 2007. Monitoring the formation of trihalomethanes in the effluents from a shrimp hatchery. Environ. Monit. Assess. 127 (1e3), 435e444. Cabeza, A., Urtiaga, A.M., Ortiz, I., 2007. Electrochemical treatment of landfill leachates using a boron-doped diamond anode. Ind. Eng. Chem. Res. 46 (5), 1439e1446. Can˜izares, P., Lobato, J., Paz, R., Rodrigo, M.A., Sa´ez, C., 2005. Electrochemical oxidation of phenolic wastes with borondoped diamond anodes. Water Res. 39 (12), 2687e2703. Chatzisymeon, E., Xekoukoulotakis, N.P., Diamadopoulos, E., Katsaounis, A., Mantzavinos, D., 2009. Boron-doped diamond anodic treatment of olive mill wastewaters: statistical analysis, kinetic modeling and biodegradability. Water Res. 43 (16), 3999e4009. Chen, X., Chen, G., Gao, F., Yue, P.L., 2003. High-performance Ti/BDD electrodes for pollutant oxidation. Environ. Sci. Technol. 37 (21), 5021e5026. Davidson, J., Helwig, N., Summerfelt, S.T., 2008. Fluidized sand biofilters used to remove ammonia, biochemical oxygen demand, total coliform bacteria, and suspended solids from an intensive aquaculture effluent. Aquac. Eng. 39 (1), 6e15. Deng, Y., Englehardt, J.D., 2007. Electrochemical oxidation for landfill leachate treatment. Waste Manag. 27 (3), 380e388. Ebeling, J.M., Rishel, K.L., Sibrell, P.L., 2005. Screening and evaluation of polymers as flocculation aids for the treatment of aquacultural effluents. Aquacult. Eng. 33 (4), 235e249. Katayose, M., Yoshida, K., Achiwa, N., Eguchi, M., 2007. Safety of electrolyzed seawater for use in aquaculture. Aquaculture 264 (1e4), 119e129. Krumins, V., Ebeling, J., Wheaton, F., 2001. Part-day ozonation for nitrogen and organic carbon control in recirculating aquaculture systems. Aquacult. Eng. 24 (3), 231e241. Lee, J.-K., Lee, K.-R., Hong, S.-H., Kim, K.-H., Lee, B.-H., Lim, J.-H., 2002. Residual chlorine distribution and disinfection during electrochemical removal of dilute ammonia from an aqueous solution. J. Chem. Eng. Jpn. 35 (3), 285e289.
134
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 2 5 e1 3 4
Li, L., Liu, Y., 2009. Ammonia removal in electrochemical oxidation: mechanism and pseudo-kinetics. J. Hazard. Mater. 161 (2e3), 1010e1016. Lin, S.H., Wu, C.L., 1997. Electrochemical nitrite and ammonia oxidation in sea water. J. Environ. Sci. Health A Tox. Hazard. Subst. Environ. Eng. 32 (8), 2125e2138. Lin, S.H., Wu, C.L., 1996. Electrochemical removal of nitrite and ammonia for aquaculture. Water Res. 30 (3), 715e721. Miladinovic, N., Weatherley, L.R., 2008. Intensification of ammonia removal in a combined ion-exchange and nitrification column. Chem. Eng. J. 135 (1e2), 15e24. Palacios, G.L., Timmons, M.B., 2001. Determining design parameters for recovery of aquaculture wastewater using sand beds. Aquacult. Eng. 24 (4), 289e299. Panizza, M., Michaud, P.A., Cerisola, G., Comninellis, C., 2001. Anodic oxidation of 2-naphthol at boron-doped diamond electrodes. J. Electroanal Chem. 507 (1e2), 206e214. Pe´rez, G., Ferna´ndez-Alba, A.R., Urtiaga, A.M., Ortiz, I., 2010. Electro-oxidation of reverse osmosis concentrates generated in tertiary water treatment. Water Res. 44 (9), 2763e2772. Polcaro, A.M., Vacca, A., Mascia, M., Palmas, S., Rodiguez Ruiz, J., 2009. Electrochemical treatment of waters with BDD anodes: kinetics of the reactions involving chlorides. J. Appl. Electrochem 39 (11), 2083e2092. Potts, A.C., Boyd, C.E., 1998. Chlorination of channel catfish ponds. J. World Aquac. Soc. 29 (4), 432e440. Pulefou, T., Jegatheesan, V., Steicke, C., Kim, S.-H., 2008. Application of submerged membrane bioreactor for aquaculture effluent reuse. Desalination 221 (1e3), 534e542. Ruttanagosrigit, W., Boyd, C.E., 1989. Measurement of chemical oxygen demand in waters of high chloride concentration. J. Aquacult. Soc. 20, 170e172. Rychen, P., Pupunat, L., Haenni, W., Santoli, E., 2003. Water treatment applications with BDD electrodes and the DiaCell concept. New Diam. Front. Carbon Technol. 13 (2), 109e117. Sharrer, M.J., Tal, Y., Ferrier, D., Hankins, J.A., Summerfelt, S.T., 2007. Membrane biological reactor treatment of a saline
backwash flow from a recirculating aquaculture system. Aquacult. Eng. 36 (2), 159e176. Sun, C.-C., Chou, T.-C., 1999. Kinetics of anodic oxidation of nitrite ion using in situ electrogenerated HClO in a NaCl aqueous solution. Ind. Eng. Chem. Res. 38 (12), 4545e4551. Szpyrkowicz, L., Radaelli, M., 2006. Scale-up of an electrochemical reactor for treatment of industrial wastewater with an electrochemically generated redox mediator. J. Appl. Electrochem. 36 (10), 1151e1156. Szpyrkowicz, L., Kelsall, G.H., Kaul, S.N., De Faveri, M., 2001. Performance of electrochemical reactor for treatment of tannery wastewaters. Chem. Eng. Sci. 56 (4), 1579e1586. Tango, M.S., Gagnon, G.A., 2003. Impact of ozonation on water quality in marine recirculation systems. Aquacult. Eng. 29 (3e4), 125e137. Taparhudee, W., Suksamran, S., Srisapoome, P., Chuchird, N., 2008. Application of an electrolytic water treatment technique in a Litopenaeus vannamei (Boone, 1931) closed-hatchery system. Kasetsart J. Nat. Sci. 42 (3), 503e512. Urtiaga, A., Rueda, A., Anglada, A., Ortiz, I., 2009. Integrated treatment of landfill leachates including electro-oxidation at pilot plant scale. J. Hazard. Mater. 166 (2e3), 1530e1534. Vijayaraghavan, K., Ahmad, D., Bin Fadzin, T.S., 2008. In situ hypochlorous acid generation for the treatment of brackish shrimp aquaculture wastewater. Aquac. Res. 39 (5), 449e456. Westerhoff, P., Chao, P., Mash, H., 2004. Reactivity of natural organic matter with aqueous chlorine and bromine. Water Res. 38 (6), 1502e1513. Wijesekara, R.G.S., Nomura, N., Matsumura, M., 2005. Electrochemical removal of ammonia, chemical oxygen demand and energy consumption from aquaculture waters containing different marine algal species. J. Chem. Technol. Biotechnol. 80 (12), 1408e1415. Xu, H., Su, J., Xiang, X.-L., Huang, W.-M., Lin, H.-B., 2008. A kinetics of indirect electro-oxidation reaction of NH3-N with chloride ion in a filter-press type electrochemical cell. Gaodeng Xuexiao Huaxue Xuebao/Chemical Journal of Chinese Universities 29 (7), 1416e1419 (in Chinese).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 3 5 e1 4 4
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Performance of high-loaded ANAMMOX UASB reactors containing granular sludge Chong-Jian Tang a, Ping Zheng a,*, Cai-Hua Wang a, Qaisar Mahmood b, Ji-Qiang Zhang a, Xiao-Guang Chen a, Lei Zhang a, Jian-Wei Chen a a b
Department of Environmental Engineering, Zhejiang University, Hangzhou 310029, China Department of Environmental Sciences, COMSATS University, Abbottabad, Pakistan
article info
abstract
Article history:
The performance of high-loaded anaerobic ammonium oxidizing (ANAMMOX) upflow
Received 1 November 2009
anaerobic sludge bed (UASB) reactors was investigated. Two ANAMMOX reactors (R1 with
Received in revised form
and R2 without effluent recycling, respectively) were fed with relatively low nitrite
28 May 2010
concentration of 240 mg-N L1 with subsequent progressive increase in the nitrogen
Accepted 10 August 2010
loading rate (NLR) by shortening the hydraulic retention time (HRT) till the end of the
Available online 17 August 2010
experiment. A super high-rate performance with nitrogen removal rate (NRR) of 74.3e76.7 kg-N m3 day1 was accomplished in the lab-scale ANAMMOX UASB reactors,
Keywords:
which was 3 times of the highest reported value. The biomass concentrations in the
ANAMMOX
reactors were as high as 42.0e57.7 g-VSS L1 with the specific ANAMMOX activity (SAA)
Granular characteristics
approaching to 5.6 kg-N kg-VSS1 day1. The high SAA and high biomass concentration
Process performance
were regarded as the key factors for the super high-rate performance. ANAMMOX granules
UASB reactor
were observed in the reactors with settling velocities of 73e88 m h1. The ANAMMOX granules were found to contain a plenty of extracellular polymers (ECPs) such as 71.8e112.1 mg g-VSS1 of polysaccharides (PS) and 164.4e298.2 mg g-VSS1 of proteins (PN). High content of hemachrome (6.8e10.3 mmol g-VSS1) was detected in the ANAMMOX granules, which is supposed to be attributed to their unique carmine color. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Anaerobic ammonium oxidation (ANAMMOX) is a promising biotechnology for the treatment of ammonium-rich wastewater (van der Star et al., 2007; Joss et al., 2009). Under anoxic conditions, the ANAMMOX bacteria accomplish autotrophic ammonium oxidation to dinitrogen gas employing nitrite as an electron acceptor (Strous et al., 1998). It offers several advantages over conventional nitrification-denitrification systems including higher nitrogen removal rate, lower operational cost and less space requirement (Jetten et al., 2005; van der Star et al., 2007; Joss et al., 2009). Combined with single reactor high activity ammonium removal over nitrite (SHARON)
process in which half of ammonium is converted to nitrite, the first full-scale ANAMMOX process (70 m3) was applied to treat sludge dewatering effluents in Rotterdam, The Netherlands in 2002 (van Dongen et al., 2001; van der Star et al., 2007). It stably operated achieving nitrogen removal rate (NRR) up to 9.5 kg-N m3 day1 (van der Star et al., 2007). High-rate is one of the prime objectives for ANAMMOX process. The NRR of conventional nitrogen removal biotechnologies was less than 0.5 kg-N m3 day1 (Jin et al., 2008); while for ANAMMOX process, it was higher than 5 kgN m3 day1 as obtained by a number of researchers using different reactors such as upflow biofilter, upflow anaerobic sludge blanket (UASB) reactor and gas-lift reactor (Sliekers
* Corresponding author. Tel./fax: þ86 571 86971709. E-mail addresses:
[email protected] (C.-J. Tang),
[email protected] (P. Zheng). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.018
136
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 3 5 e1 4 4
SAA Specific ANAMMOX activity SEM Scanning electron microscopy SHARON Single reactor high activity ammonium removal over nitrite SVI Sludge volume index TEM Transmission electron microscopy TSS Total suspended solids UASB Upflow anaerobic sludge bed VSS Volatile suspended solids
Nomenclature ANAMMOX Anaerobic ammonium oxidation ECP(s) Extracellular polymer(s) HLR Hydraulic loading rate HRT Hydraulic retention time NLR Nitrogen loading rate NRR Nitrogen removal rate PN Protein PS Polysaccharide
et al., 2003; Imajo et al., 2004; Isaka et al., 2007; van der Star et al., 2007; Tang et al., 2009a). To date, the highest NRR reported was 26.0 kg-N m3 day1 at hydraulic retention time (HRT) of 0.24 h (Tsushima et al., 2007). Previous works on anaerobic processes including anaerobic digestion (Thiele et al., 1990) and denitrifying process (Franco et al., 2006) attributed high volumetric removal rates to three main aspects. Firstly, the reactors should have high-quality sludge retention for sufficient biomass accumulation. Secondly, the microbial communities should aggregate as granular sludge or biofilms for optimum metabolic activity. Finally, the substrate requirements of ANAMMOX bacteria should be satisfied simultaneously avoiding substrate inhibition, especially nitrite inhibition (Strous et al., 1999; Isaka et al., 2007; Tsushima et al., 2007). The granular sludge characterized by good settling property and high activity plays a pivotal role in the performance of high-rate bioreactors (Thiele et al., 1990; Franco et al., 2006; Zhang et al., 2008). The characteristics of granular sludge such as heterotrophic aerobic granules (Beun et al., 1999; Beun et al., 2002; Zheng and Yu, 2007; Adav et al., 2008), anaerobic granules (Hulshoff Pol et al., 2004; Show et al., 2004; Wu et al., 2009), hydrogen-producing granules (Mu and Yu, 2006; Zhang et al., 2008), denitrifying granules (Franco et al., 2006) and autotrophic nitrifying granules (Tsuneda et al., 2003; Liu et al., 2008; Belmonte et al., 2009) have been extensively studied. In case of ANAMMOX granules, the settling property, diameter
Table 1 e Operational parameters of the ANAMMOX sludge and the two UASB reactors before the start of the experiment. Characteristic
R1
A: Characteristics of the sludge Diameter (mm) TSS/VSS (%) SAA (kg-N kg-VSS1 day1)
1.9 82 0.3
distribution and substrate diffusion have been reported (Arrojo et al., 2006; Ni et al., 2009). The characteristics of carmine color of ANAMMOX granules and their associated extracellular polymers (ECPs) have also drawn considerable attention for the process optimization. The hydroxylamine oxidoreductase and hydrazine oxidoreductase are two important enzymes of the ANAMMOX pathway. Both of these enzymes are rich in heme c (Klotz et al., 2008; Schmid et al., 2008), which endows the granular sludge with the carmine color. The extracellular polymers are assumed to be a key factor in the formation of granular sludge, which can be secreted by ANAMMOX bacteria (Cirpus et al., 2006). In the present study, two ANAMMOX UASB reactors were operated to investigate the performance of high-loaded reactors possessing carmine granular sludge.
2.
Material and methods
2.1.
Synthetic wastewater
Ammonium and nitrite were supplemented to mineral medium as required in the form of (NH4)2SO4 and NaNO2, respectively. The composition of the mineral medium was (g L1 except for trace element solution) (Trigo et al., 2006): KH2PO4 0.01, CaCl2$2H2O 0.00565, MgSO4$7H2O 0.3, KHCO3 1.25, FeSO4 0.00625, EDTA 0.00625 and 1.25 mL L1 of trace elements solution. The trace element solution contained (g L1): EDTA 15, H3BO4 0.014, MnCl2$4H2O 0.99, CuSO4$5H2O 0.25, ZnSO4$7H2O 0.43, NiCl2$6H2O 0.19, NaSeO4$10H2O 0.21, NaMoO4$2H2O 0.22 and NaWO4$2H2O 0.050 (adapted from van de Graaf et al. (1996)).
R2
2.2. 2.1 85 0.2
B: Operational characteristics of the two UASB reactors before the start of the experiment 300 200 Influent ammonium concentration (mg-N L1) 360 240 Influent nitrite concentration (mg-N L1) Effluent recycling ratio 0.5 e HRT (h) 6.90 11.7 18.7 26.8 Sludge concentration (g-VSS L1) 6.0 2.9 NRR (kg-N m3 day1)
ANAMMOX bioreactors
The experimental work was carried out in two glass-made UASB reactors of 1.1 L capacity having internal diameter of 50 mm. Both reactors were completely covered with black cloth to avoid the growth of phototrophic organisms and the related oxygen production (van der Star et al., 2008). The reactors were fed with synthetic wastewater which was flushed with 95%Ar-5%CO2 continuously to maintain anoxic conditions. The temperature was set at 35 1 C according to Tsushima et al. (2007) and the influent pH was controlled in the range of 6.8e7.0 (Tang et al., 2009b). The produced gas was
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 3 5 e1 4 4
initially absorbed by NaOH solution and then recorded by a wet gas meter. The two reactors (designated as R1 and R2) were initially inoculated with anaerobic granular sludge taken from a paper mill wastewater treatment plant (100 m3, located in Zhejiang Province, China). The average diameter of the anaerobic granular sludge was about 2.2 mm; and the VSS/TSS content was about 85%. The reactors were successfully started up and operated stably for 214 days before the experiments. The operational characteristics of R1 and R2 are listed in Table 1.
2.3.
Reactor operation
In order to avoid the nitrite inhibition, both reactors were operated at low influent nitrite concentration. R2 was 1 without effluent recyconstantly fed with 240 mg-NO 2 -N L 1 with cling; while, R1 was constantly fed with 360 mg-NO 2 -N L effluent recycling ratio (recycling flow to inflow ratio) about 0.5. Thus, the influent nitrite concentration was about 240 mg-N L1 after dilution. Ammonium was supplemented relatively in excess and was progressively increased during the shortening of HRT in order to gain better nitrite removal efficiency and performance stability (Tsushima et al., 2007). The HRT was progressively shortened after 5 days at each step when the reactor performance was stable. Throughout the operation, no sludge was deliberately removed from the reactors.
2.4.
137
Analytical procedures
The influent and effluent samples were collected on daily basis and analyzed immediately. The determination of pH, ammonium, nitrite, nitrate, 5-min and 30-min sludge volume indices (SVI5 and SVI30), total suspended solids (TSS) and volatile suspended solids (VSS) concentrations were carried out following the Standard Methods (APHA, 1998). The size of granular sludge was measured by an image analysis system (QCOLite) with a Leica DM2LB microscope and a digital camera (Canon S30). ECPs were extracted from sludge by EDTA (Sheng et al., 2005), then the extracellular proteins (PN) were determined by the Lowry method using egg albumin as standard and the polysaccharide (PS) content was analyzed by the anthrone method with glucose as standard (Wu et al., 2009). Heme c content was determined according to Berry and Trumpower (1987) and Sinclair et al. (1999). The nitrogen ligands from protein-bound heme were replaced by pyridine in alkali, and the resultant heme c was quantified through the difference between spectra of the reduced (sodium dithionite crystals) and oxidized (potassium ferricyanide) compounds. The heme c concentration was calculated based on the millimolar extinction coefficient of 23.97 (mM cm1) for the difference in absorption between the peak at 550 nm and the trough at 535 nm. Specific ANAMMOX activity (SAA) was determined following the procedures described by Tang et al. (2009b). The initial substrate concentration was maintained at
Fig. 1 e Profile of nitrogen removal rate and influent flow rate of the two ANAMMOX R1 (A) and R2 (B).
138
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 3 5 e1 4 4
day1 (Fig. 1A). The nitrogen removal performance of R2 also showed a similar trend when the influent flow rate was progressively increased during the first 290 days. During that period, the NRR was 65.4 3.0 (60.9e69.8) kg-N m3 day1 (HRT, 0.18e15 h; NLR, 78.9e97.9 kg-N m3 day1). Further increasing the influent flow rate caused the nitrogen removal performance of both reactors to decline (Fig. 1A and B). Subsequent attempts made to increase influent flow rate of both reactors did not significantly improve NRR. At the end of the operation, the NRR of R1 was recorded to be 74.3 6.7 (66.8e82.8) kg-N m3 day1 (HRT, 0.16 h; influent flow rate, 152.4 L L1 day1; NLR, 125.0 kg-N m3 day1; days 399e412); while for R2, the NRR was up to 76.7 4.5 (69.8e84.6) kg-N m3 day1 (HRT, 0.11 h; influent flow rate, 221.0 L L1 day1; NLR, 137.1 kg-N m3 day1; days 410e417). The NRRs observed for both reactors were three times of the highest reported values (Tsushima et al., 2007).
3.1.2.
Removal efficiency
3.
Results and discussion
3.1.
Nitrogen removal performance
The influent ammonium concentration was gradually increased for both reactors as described in Materials and Methods section. The ammonium removal efficiency was up to 90% when the HRT of R1 was longer than 1.58 h (influent ammonium concentration, 330 mg-N L1; NLR, 10.5 kg-N m3 day1); while it decreased to 70% when the HRT was further shortened to 0.21 h (influent ammonium concentration, 420 mg-N L1; NLR, 89.1 kg-N m3 day1). For R2, the ammonium removal efficiency was around 90% and 70% at the corresponding HRT of 1.29 h (influent ammonium concentration, 220 mg-N L1; NLR, 8.6 kg-N m3 day1) and 0.30 h (influent ammonium concentration, 250 mg-N L1; NLR, 39.7 kg-N m3 day1), respectively. Nitrite removal efficiency of both the reactors at different HRTs is presented in Fig. 2. For R1, the nitrite depleted with the removal efficiency higher than 96.8% and the average effluent nitrite concentration was about 11 mg-N L1 when working at HRTs longer than 0.39 h. But it increased to 124.4 mg-N L1 when HRT was shortened to 0.16 h with a sharp decrease in nitrite removal efficiency to 65.4% (Fig. 2A). Fig. 2B shows the nitrite removal efficiency of R2 at HRT range of 0.21 he0.10 h. The average effluent nitrite concentration increased to 90.8 mg-N L1 with nitrite removal efficiency of 62.2% when HRT was shortened to 0.10 h. The nitrate production did not significantly fluctuate in both reactors and it correlated to the substrate removal. The stoichiometric ratios of ammonium conversion, nitrite removal and nitrate production were 1: (1.31 0.03):(0.23 0.01) (R1) and 1:(1.32 0.06):(0.25 0.02) (R2), both were close to the reported ratios (Strous et al., 1998).
3.1.1.
Volumetric capacity
3.1.3.
Fig. 2 e Effluent nitrite concentration and nitrite removal efficiency of R1 (A) and R2 (B) at different HRTs. 100 mg-N L1 for both ammonium and nitrite and the sludge concentration was kept about 0.6 g-VSS L1. The cellular yield and the specific growth rate were calculated based on the procedures of Chen et al. (2010). Sludge granules settleability (50 in number) was measured during the last 200 mm through a 300 mm water column (Franco et al., 2006). Specific density was measured according to Beun et al. (2002). Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) were performed according to Tang et al. (2009a) and Tang et al. (2009b), respectively. Nitrogen removal rate was calculated as the sum of ammonium and nitrite consumption.
Throughout the reactors’ operation, the influent nitrite concentration was maintained constant whereby the nitrogen loading rate (NLR) was progressively increased by shortening HRT. The nitrogen removal performance of R1 and R2 is depicted in Fig. 1A and B, respectively. During the first 250 days, the hydraulic loading rate (HLR) of R1 was increased from 3.5 L L1 day1 to 114.3e123.8 L L1 day1, that corresponded to HRT of 0.21e0.19 h; the NRR was enhanced to 72.5 2.8 (70.3e78.5) kg-N m3 day1 with the NLR of 89.1e99.0 kg-N m3
Biomass growth
Progressive increase of the ANAMMOX granules was observed throughout the reactors’ operation (Fig. 3). The biomass concentration in R1 enhanced to 57.7 g-VSS L1 after 400 days. While, the biomass concentration in R2 gradually increased to 42.0 g-VSS L1 during that period. The VSS/TSS ratios of the ANAMMOX granules in both reactors were in the range of 0.82e0.90. Relatively low calcium and phosphorous concentrations were included in the mineral medium based on the results of Trigo et al. (2006), which gave rise to the high volatile
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 3 5 e1 4 4
139
Fig. 3 e Apparent characteristics of the sludge in R1 (A) and R2 (B) at different periods, and the washed-out sludge (C).
140
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 3 5 e1 4 4
Table 2 e Diameter, settling velocity (vs), sludge volumetric index (SVI5), density and specific density of the granules in the two UASB reactors at the end of the experiment (average value). Parameter Diameter (mm) vs (m h1) SVI5 (mL g-VSS1) Density (g mL1) Specific density (g-VSS L-granules1)
R1
R2
2.5 88 25 1.0323 108
2.4 73 24 1.0260 94
solids content of the enriched granules in both the UASB reactors (Trigo et al., 2006). The ANAMMOX biomass yield in R1 was 0.23 g-VSS g-NHþ 41 N with the specific growth rate of 0.0047 h1 corresponding to the doubling time of 6.1 days under NRR higher than 70 kgN m3 day1. The corresponding biomass values in R2 were 1 1 with the 0.22 g-VSS g-NHþ 4 -N , growth rate of 0.0060 h doubling time of 4.8 days. The cellular yields in our study were approximately 2e3 times of the previously reported values þ 1 (0.07 g-VSS g-NHþ 4 -N , Trigo et al., 2006; 0.088 g-VSS g-NH4 þ 1 1 N , Strous et al., 1998; and 0.11 g-VSS g-NH4 -N , van Dongen et al., 2001); and the doubling times were shorter than the value reported by Strous et al. (1998) (11 days).
3.1.4.
Specific ANAMMOX activity
The specific activities of the ANAMMOX granules progressively increased with the passage of time. Amazingly, high values of 5.6 0.9 kg-N kg-VSS1 day1 were detected for the carmine granules when the NRR was higher than 70 kg-N m3 day1. The high activity was an important factor leading to the super high NRR of the UASB reactors.
3.1.5.
Sludge washout
The nitrogen gas production rate progressively increased along the development of nitrogen removal performance, resulting in the increased superficial gas upflow velocity. The nitrogen gas production rates for R1 and R2 were 52.42 7.42 and 52.68 5.24 L L1 day1 with the accompanied gas upflow velocity up to 1.11 and 1.17 m h1, respectively, when superficial liquid upflow velocities were gradually increased to 5.78 and 5.24 m h1, respectively.
The high shear force from liquid upflow and gas upflow in both reactors led to the severe sludge washout (Figs. 1 and 3). The effluent VSS as high as 8.3 g day1 (R1) and 7.7 g day1 (R2) were observed during days 433e435. Consequently, the NRR declined (Fig. 1).
3.2.
Characteristics of ANAMMOX granules
3.2.1.
Diameter distribution and settling property
Granulation of ANAMMOX microorganisms resulted in granular diameters of 1.0e6.4 mm in both reactors. The average granule diameter was 2.5 mm (R1) and 2.4 mm (R2). The percentage of granules with diameter larger than 2 mm ranged 68%e71%. The granules in both reactors possessed a high settling velocity (73e88 m h1, Table 2). The SVI5 range decreased from 42e51 mL g-VSS1 to 24e25 mL g-VSS1 (Table 2) with a thickening process verified by an SVI5 to SVI30 ratio of 1 suggesting a fabulous sedimentation property. The density of ANAMMOX granules in both reactors was about 1.03 g mL1; and the specific density of the ANAMMOX granules (91e120 g-VSS L-granules1) was also comparable to aerobic granules (40e70 g-VSS L-granules1, Beun et al., 2002) and high-loaded denitrifying granules (128e136 g-VSS L-granules1, Franco et al., 2006). The formation of well settling granules resulted in accumulation of high concentrations of sludge in both reactors in spite of working at extremely high NLRs and short HRTs. It was another factor contributing to the super high volumetric nitrogen removal rates.
3.2.2.
Extracellular polymers
The microbial ECPs are a rich matrix of polymers, mainly including polysaccharides and proteins (Liu et al., 2009). They are supposed to play a central role in the formation of granules in bioreactors (Liu and Tay, 2002; Hulshoff Pol et al., 2004; Liu et al., 2009). In the present study, the ECP content determined at different nitrogen removal levels is presented in Fig. 4. It was evident that both the polysaccharide and protein contents increased with the increasing NRR. It was obvious that the polysaccharides increased slowly as compared to the proteins. The polysaccharide contents (mg g-VSS1) for R1 and R2 were 71.8 2.3 (69.6e74.2) and 112.1 2.8 (109.8e115.2) at NRR higher than 70 kg-N m3 day1; whereby the extracellular protein contents (mg g-VSS1) increased sharply to 164.4 9.3
Fig. 4 e ECP content of the sludge in R1 (A) and R2 (B) at different nitrogen removal rates.
141
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 3 5 e1 4 4
Table 3 e ECP content in different microbial granules. Extracellular polymers (mg g-VSS1)
Granular sludge
ANAMMOX granules Aerobic granules Phenol-degrading granules Anaerobic granules Hydrogen-producing granules Nitrifying granules Denitrifying granules
Proteins
Polysaccharides
PN/PS
164.4 9.3 z40 240 13 42.7 37.8 70.9 4.5 56 25a N.A.
71.8 2.3 z16 61.0 9.4 17.3 6.8 115.6 5.2 18 1a N.A.
2.29 2.5 3.93 2.5 0.6 3.1 2.2
HRT (h)
Reference
0.16 8 N.A. N.A. 18 2.8 N.A.
This paper Zheng and Yu (2007) Adav et al. (2008) Wu et al. (2009) Mu and Yu (2006) Martı´nez et al. (2004) Franco et al. (2006)
a mg L1.
(153.7e170.1) (R1) and 298.2 8.7 (288.1e303.7) (R2), respectively (Fig. 4). It was amazing to observe that the autotrophic ANAMMOX granules possessed a high ECP content as compared to the heterotrophic granules (Table 3). It was previously reported that the autotrophic bacteria unable to utilize organic compounds would secrete low ECP content (Tsuneda et al., 2003). ECPs could physically bridge neighboring cells to each other by altering the negative charges on bacterial surface (Liu et al., 2004). Thus, granulation may be facilitated by large secretion of ECP. As evident from Fig. 3, complete granulation of the ANAMMOX biomass occurred in both reactors with the average granular diameter above 2 mm when NRR was around 70 kg-N m3 day1. The proteins to polysaccharides ratio (PN/PS) was usually used to evaluate the granular settleability and strength (Quarmby and Forster, 1995; Cuervo-Lo´pez et al., 1999; Batstone and Keller, 2001; Martı´nez et al., 2004; Franco et al., 2006; Wu et al., 2009). The PN/PS ratio of the ANAMMOX granules was also low when compared to other microbial granules (Table 3), suggesting a greater granular stability (Franco et al., 2006). Nevertheless, PN/PS ratios increased during the reactor operation when hydrodynamic shear force increased (Fig. 4). As pointed out by various researchers, the higher PN/PS ratio of microbial granules indicated lower strength and weaker settleability (Quarmby and Forster, 1995; Cuervo-Lo´pez et al., 1999; Batstone and Keller, 2001; Martı´nez et al., 2004); thus, the sludge floating or foaming would be easily accompanied (Franco et al., 2006; Wu et al., 2009). In this study, we determined the ECP contents of the floated granules. We found that extracellular proteins and PN/PS ratios of the floated ANAMMOX granules at high NRRs and HLRs were significantly higher than the counterparts of well-settled ones (Table 4). Wu et al. (2009) reported that the secretion of extracellular protein by heterotrophic anaerobic granules was stimulated under high hydrodynamic shear force in the internal circulation anaerobic reactor. On the contrary, Tay et al. (2001)
pointed out that the hydrodynamic shear force stimulated the production of extracellular polysaccharides in heterotrophic aerobic reactors. In the present study, the secretion of extracellular proteins and polysaccharides were enhanced when hydrodynamic shear force was increased. But extracellular proteins were produced at a higher rate, leading to the increased PN/PS ratio. Moreover, the over-production of extracellular proteins can raise fluid viscosity in the reactor; as a consequence, the shear force in the reactor is intensified in turn based on Newton’s law (Wu et al., 2009). This would increase the risk of sludge disruption due to increasing shear force. So, in the case of violent shear conditions, the disruption of aggregates and sludge washout becomes inevitable (Wu et al., 2009). The over-production of extracellular proteins might be a potential cause resulting in the severe sludge washout from the UASB reactors.
3.2.3.
Heme c content
The morphology of microbial granules is affected by a number of factors such as seed sludge characteristics, substrate composition, loading rate, feeding strategy, reactor design, and hydrodynamic shear force (Liu et al., 2009). The color of aerobic granules, nitrifying granules, denitrifying granules and hydrogen-producing granules is usually yellow, while the color of methanogenic granules is black because of the precipitation of metal sulfides (Hulshoff Pol et al., 2004; Franco et al., 2006; Liu et al., 2009). The color of high-load ANAMMOX granules is uniquely carmine (Fig. 3). Heme c, which is an important cofactor of some ANAMMOX bacterial enzymes, was presumed to play a key role to attribute the carmine color of ANAMMOX sludge. The present study showed that the heme c content significantly increased with the increasing NRR. The heme c content was 0.7e1.4 mmol g-VSS1 when NRR was lower than 10 kg-N m3 day1; and finally reached 6.8 0.9 (5.9e7.6) mmol g-VSS1 (R1) and 10.3 0.6 (9.7e10.8) mmol g-VSS1 (R2) at NRR higher than 70 kg-N m3 day1. The
Table 4 e ECP content of the settled ANAMMOX granules and the floated granules. NRR (kg-N m3 day1) 44.87 2.39 76.68 4.46
Sludge Settled granules Floated granules Settled granules Floated granules
Polysaccharides (mg g-VSS1) 76.77 79.98 112.14 134.23
2.29 3.54 2.77 4.34
Proteins (mg g-VSS1)
PN/PS
2.40 4.09 2.65 3.88
184.39 327.48 298.19 520.78
9.27 35.25 8.72 204.75
142
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 3 5 e1 4 4
Fig. 5 e Transmission electron micrographs of the sludges taken from R1 (A and B) and from R2 (C and D) after the end of the experiment. The innermost compartment, the anammoxosome (A), filled with material of moderate electron density and granular texture, but devoid of ribosome-like particles, is surrounded by a single membrane. The paryphoplasm (P), in this case relatively electron-transparent, surrounds the rim of the cell. The scale bar in A and C [ 2 mm; in B and D [ 0.2 mm.
increase of heme c content was related with the increase of ANAMMOX bacterial numbers, resulting in high SAA. As evident from Fig. 3, the number of red ANAMMOX granules in both reactors increased along the reactor operation period, which was in accordance with the increase of heme c content. On the contrary, the amount of initial seed sludge (grey particles) decreased significantly. The carmine granules dominated in the R2 after 400 days’ operation (Fig. 3B). However, several black zones appeared in R1 after about 310 days’ operation (Fig. 3A). To our visual observation, higher concentration of granular sludge was held in R1 and the black zones may have resulted from the sludge blockage.
3.2.4.
Morphology of granular sludge
The structure of the microbial granules developed in both reactors was observed by means of SEM and TEM. The scanning electron micrographs represented a sample of red-colored mature ANAMMOX granule characterized by a cauliflower-like shape (Arrojo et al., 2006). The granular surface mainly consisted of spherical and elliptical bacteria; few or even no bacilli and filamentous bacteria were observed in the two reactor enrichments, suggesting that the ANAMMOX bacteria dominated after enrichment in both reactors. The shape of dominating cocci in R1 enrichment was like a shrunken ball, while that in R2 enrichment it was like a gaseous ball, which were different from each other. TEM performed on the enriched biomass taken from bottom of the two UASB reactors revealed that the dominant cells in both enrichments displayed typical ultrastructural features of ANAMMOX bacteria; i.e., a single membrane bound anammoxsome containing tubule-like structure (Strous, 2000; Lindsay et al., 2001; Schmid et al., 2003; Kartal et al., 2007)
(Fig. 5B and D). As evident from Fig. 5A and C, both enrichments were dominated by ANAMMOX cells. The morphology of the ANAMMOX cells in the two enrichments showed some differences of the paryphoplasm (relatively electron-transparent, as proposed by Lindsay et al. (2001)). It is evident that the paryphoplasm in R1 cells was larger than that in R2 (Fig. 5). The structural differences of the two reactor enrichments observed by SEM and TEM were interesting. As shown, the seed sludge and the mineral medium used in the study were the same; and the hydrodynamic conditions as well as the substrate concentrations in both reactors were also similar. The major difference was the effluent recycling in R1 while no recycling was done for R2. Thus the recycling of ANAMMOX products (known or unknown) for R1 may be a cause leading to the ultrastructural differences of the two reactor enrichments.
4.
Conclusions
A super high-rate performance with nitrogen removal rate of 74.3e76.7 kg-N m3 day1 was revealed for the lab-scale ANAMMOX UASB reactors, which was 3 times of the previously reported top value. The performance was stable until the HRT was shortened to 0.16e0.11 h with the hydraulic loading rate larger than 152.4e221.0 L L1 day1. The biomass concentrations in the reactors were 42.0e57.7 g-VSS L1 with the specific ANAMMOX activity up to 5.6 kg-N kg-VSS1 day1, both of which were regarded as the key factors leading to the super high-rate performance. The settling velocities of ANAMMOX granules ranged 73e88 m h1 in both the reactors. The ANAMMOX granules
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 3 5 e1 4 4
were found to contain a large amount of extracellular polymers with the polysaccharides and proteins contents of 71.8e112.1 mg g-VSS1 and 164.4e298.2 mg g-VSS1, respectively. Relatively high ECP content and low PN/PS ratios were attributed to the ANAMMOX granulation. High hemachrome content of 6.8e10.3 mmol g-VSS1 were detected in the ANAMMOX granules, which was an important cofactor of some ANAMMOX enzymes and was supposed to be responsible for the unique carmine color.
Acknowledgements Financial support of this work by the National Hightech Research and Development (R&D) Program of China (2009AA06Z311), the National Key Technologies R&D Program of China (2008BADC4B05) and the Natural Science Foundation of China (30770039) is gratefully acknowledged. We wish to thank the anonymous reviewers and editors for their valuable suggestions on revising and improving the work.
Appendix. Supplementary material Supplementary data associated with this article can be found in the on-line version, at doi:10.1016/j.watres.2010.08.018.
references
Adav, S.S., Lee, D.J., Tay, J.H., 2008. Extracellular polymeric substances and structural stability of aerobic granule. Water Res. 42, 1644e1650. APHA, 1998. Standard Methods for the Examination of Water and Wastewater, twentieth ed. American Public Health Association, Washington DC, USA. Arrojo, B., Mosquera-Corral, A., Campos, J.L., Me´ndez, R., 2006. Effects of mechanical stress on Anammox granules in a sequencing batch reactor (SBR). J. Biotechnol. 123, 453e463. Batstone, D.J., Keller, J., 2001. Variation of bulk properties of anaerobic granules with wastewater type. Water Res. 35, 1723e1729. Belmonte, M., Vazquez-Padı´n, J.R., Figueroa, M., Franco, A., Mosquera-Corral, A., Campos, J.L., Me´ndez, R., 2009. Characteristic of nitrifying granules developed in an air pulsing SBR. Process Biochem. 44, 602e606. Berry, E.A., Trumpower, B.L., 1987. Simultaneous determination of hemes a, b, and c from pyridine hemochrome spectra. Anal. Biochem. 161, 1e15. Beun, J.J., Hendriks, A., van Loosdrecht, M.C.M., Morgenroth, E., Wilderer, P.A., Heijnen, J.J., 1999. Aerobic granulation in a sequencing batch reactor. Water Res. 33, 2283e2290. Beun, J.J., van Loosdrecht, M.C.M., Heijnen, J.J., 2002. Aerobic granulation in a sequencing batch airlift reactor. Water Res. 36, 702e712. Chen, J.W., Zheng, P., Yu, Y., Tang, C.J., Mahmood, Q., 2010. Promoting of sludge quantity and activity results in high loading rates in Anammox UBF. Bioresour. Technol. 101, 2700e2705. Cirpus, I.E.Y., Geerts, W., Hermans, J.H.M., Op den Camp, H.J.M., Strous, M., Kuenen, J.G., Jetten, M.S.M., 2006. Challenging protein purification from anammox bacteria. Int. J. Biol. Macromol. 39, 88e94.
143
Cuervo-Lo´pez, F.M., Martinez, F., Gutie´rrez-Rojas, M., Noyola, R.A., Go´mez, J., 1999. Effect of nitrogen loading rate and carbon source on denitrification and sludge settleability in upflow anaerobic sludge blanket (UASB) reactors. Water Sci. Technol. 40 (8), 123e130. van Dongen, U., Jetten, M.S.M., van Loosdrecht, M.C.M., 2001. The SHARON-ANAMMOX process for treatment of ammonium rich wastewater. Water Sci. Technol. 44 (1), 153e160. Franco, A., Roca, E., Lema, J.M., 2006. Granulation in high-load denitrifying upflow sludge bed pulsed reactors. Water Res. 40, 871e880. van de Graaf, A.A., De Bruijn, P., Robertson, L.A., Jetten, M.S.M., Kuenen, J.G., 1996. Autotrophic growth of anaerobic ammonium-oxidizing microorganisms in a fluidized bed reactor. Microbiology 142, 2187e2196. Hulshoff Pol, L.W., de Castro Lopes, S.I., Lettinga, G., Lens, P.N.L., 2004. Anaerobic sludge granulation. Water Res. 38, 1376e1389. Imajo, U., Tokutomi, T., Furukawa, K., 2004. Granulation of Anammox microorganisms in up-flow reactors. Water Sci. Technol. 49 (5e6), 155e163. Isaka, K., Sumino, T., Tsuneda, S., 2007. High nitrogen removal performance at moderately low temperature utilizing anaerobic ammonium oxidation reactions. J. Biosci. Bioeng. 103, 486e490. Jetten, M.S.M., Cirpus, I., Kartal, B., van Niftrik, L., van de PasSchoonen, K.T., Sliekers, O., Haaijer, S., van der Star, W., Schmid, M., van de Vossenberg, J., Schmidt, I., Harhangi, H., van Loosdrecht, M., Kuenen, J.G., Op den Camp, H., Strous, M., 2005. 1994e2004: 10 years of research on the anaerobic oxidation of ammonium. Biochem. Soc. Tran. 33, 119e123. Jin, R.C., Zheng, P., Hu, A.H., Mahmood, Q., Hu, B.L., Jilani, G., 2008. Performance comparison of two Anammox reactors: SBR and UBF. Chem. Eng. J. 138, 224e230. Joss, A., Salzgeber, D., Eugster, J., Ko¨nig, R., Rottermann, P., Burger, S., Fabijan, P., Leumann, S., Mohn, J., Siegrist, H., 2009. Full-scale nitrogen removal from digester liquid with partial nitritation and Anammox in one SBR. Environ. Sci. Technol. 43, 5301e5306. Kartal, B., Rattray, J., van Niftrik, L.A., van de Vossenberg, J., Schmid, M.C., Webb, R.I., Schouten, S., Fuerst, J.A., Damste´, J.S., Jetten, M.S.M., Strous, M., 2007. Candidatus ‘Anammoxoglobus propionicus’ a new propionate oxidizing species of anaerobic ammonium oxidizing bacteria. Syst. Appl. Microbiol. 30, 39e49. Klotz, M.G., Schmid, M.C., Strous, M., op den Camp, H.J.M., Jetten, M.S.M., Hooper, A.B., 2008. Evolution of an octahaem cytochrome c protein family that is key to aerobic and anaerobic ammonia oxidation by bacteria. Environ. Microbiol. 10, 3150e3163. Lindsay, M.R., Webb, R.I., Strous, M., Jetten, M.S.M., Butler, M.K., Forde, R.J., Fuerst, J.A., 2001. Cell compartmentalization in planctomycetes: novel types of structural organization for the bacterial cell. Arch. Microbiol. 175, 413e429. Liu, Y.Q., Liu, Y., Tay, J.H., 2004. The effects of extracellular polymeric substances on the formation and stability of biogranules. Appl. Microbiol. Biotechnol. 65, 143e148. Liu, Y.Q., Wu, W.W., Tay, J.H., Wang, J.L., 2008. Formation and long-term stability of nitrifying granules in a sequencing batch reactor. Bioresour. Technol. 99, 3919e3922. Liu, X.W., Sheng, G.P., Yu, H.Q., 2009. Physicochemical characteristics of microbial granules. Biotechnol. Adv. 27, 1061e1070. Liu, Y., Tay, J.H., 2002. The essential role of hydrodynamic shear force in the formation of biofilm and granular sludge. Water Res. 36, 1653e1665. Martı´nez, F., Lema, J., Me´ndez, R., Cuervo-Lo´pez, F., Go´mez, J., 2004. Role of exopolymeric protein on the settleability of nitrifying sludges. Bioresour. Technol 94, 43e48.
144
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 3 5 e1 4 4
Mu, Y., Yu, H.Q., 2006. Biological hydrogen production in a UASB reactor with granules. I. Physicochemical characteristics of hydrogen-producing granules. Biotechnol. Bioeng. 94, 980e987. Ni, B.J., Chen, Y.P., Liu, S.Y., Fang, F., Xie, W.M., Yu, H.Q., 2009. Modeling a granule-based anaerobic ammonium oxidizing (ANAMMOX) process. Biotechnol. Bioeng. 103, 490e499. Quarmby, J., Forster, C.F., 1995. An examination of the structure of UASB granules. Water Res. 29, 2449e2454. Schmid, M., Walsh, K., Webb, R., Rijpstra, W.I.C., van de PasSchoonen, K., Verbruggen, M.J., Hill, T., Moffett, B., Fuerst, J., Schouten, S., Sinninghe Damste´, J.S., Harris, J., Shaw, P., Jetten, M.S.M., Strous, M., 2003. Candidatus “Scalindua brodae”, sp nov., Candidatus “Scalindua wagneri”, sp nov., two new species of anaerobic ammonium oxidizing bacteria. Syst. Appl. Microbiol. 26, 529e538. Schmid, M.C., Hooper, A.B., Klotz, M.G., Woebken, D., Lam, P., Kuypers, M.M.M., Pommerening-Roeser, A., op den Camp, H.J.M., Jetten, M.S.M., 2008. Environmental detection of octahaem cytochrome c hydroxylamine/hydrazine oxidoreductase genes of aerobic and anaerobic ammonium-oxidizing bacteria. Environ. Microbiol. 10, 3140e3149. Sheng, G.P., Yu, H.Q., Yu, Z., 2005. Extraction of extracellular polymeric substances from the photosynthetic bacterium Rhodopseudomonas acidophila. Appl. Microbiol. Biotechnol. 67, 125e130. Show, K.Y., Wang, Y., Foong, S.F., Tay, J.H., 2004. Accelerated start-up and enhanced granulation in upflow anaerobic sludge blanket reactors. Water Res. 38, 2293e2304. Sinclair, P.R., Gorman, N., Jacobs, J.M., 1999. Measurement of heme concentration. Current Protocols in Toxicology, Unit 8.3, John Wiley &Sons, Inc. Sliekers, A.O., Third, K.A., Abma, W., Kuenen, J.G., Jetten, M.S.M., 2003. CANON and anammox in a gas-lift reactor. FEMS Microbiol. Lett. 218, 339e344. van der Star, W.R.L., Abma, W.R., Bolmmers, D., Mulder, J., Tokutomi, T., Strous, M., Picioreanu, C., van Loosdrecht, M.C.M., 2007. Startup of reactors for anoxic ammonium oxidation: experiences from the first full-scale Anammox reactor in Rotterdam. Water Res. 41, 4149e4163. van der Star, W.R.L., Miclea, A.I., van Dongen, U.G.J.M., Muyzer, G., Picioreanu, C., van Loosdrecht, M.C.M., 2008. The membrane
bioreactor: a novel tool to grow Anammox bacteria as free cells. Biotechnol. Bioeng. 101, 286e294. Strous, M., 2000. Microbiology of anaerobic ammonium oxidation. Ph.D. Thesis, TU Delft. Strous, M., Heijnen, J.J., Kuenen, J.G., Jetten, M.S.M., 1998. The sequencing batch reactor as a powerful tool to study very slowly growing micro-organisms. Appl. Microbiol. Biotechnol. 50, 589e596. Strous, M., Kuenen, J.G., Jetten, M.S.M., 1999. Key physiology of anaerobic ammonium oxidation. Appl. Environ. Microbiol. 65, 3248e3250. Tang, C.J., Zheng, P., Mahmood, Q., 2009a. The shear force amendments on slugging behavior in upflow Anammox sludge bed reactor. Sep. Purif. Technol. 69, 262e268. Tang, C.J., Zheng, P., Mahmood, Q., Chen, J.W., 2009b. Start-up and inhibition analysis of the Anammox process seeded with anaerobic granular sludge. J. Ind. Microbiol. Biotechnol. 36, 1093e1100. Tay, J.H., Liu, Q.S., Liu, Y., 2001. The role of cellular polysaccharides in the formation and stability of aerobic granules. Lett. Appl. Microbiol. 33, 222e226. Thiele, J.H., Wu, W.M., Jain, M.K., Zeikus, J.G., 1990. Ecoengineering high rate anaerobic digestion systems: analysis of improved syntrophic biomethanation catalysts. Biotechnol. Bioengin. 35, 990e999. Trigo, C., Campos, J.L., Garrido, J.M., Me´ndez, R., 2006. Start-up of the Anammox process in a membrane bioreactor. J. Biotechnol. 126, 475e487. Tsuneda, S., Nagano, T., Hoshino, T., Ejiri, Y., Noda, N., Hirata, A., 2003. Characterization of nitrifying granules produced in an aerobic upflow fluidized bed reactor. Water Res. 37, 4965e4973. Tsushima, I., Ogasawara, Y., Kindaichi, T., Okabe, S., 2007. Development of high-rate anaerobic ammonium-oxidizing (anammox) biofilm reactors. Water Res. 41, 1623e1634. Wu, J., Zhou, H.M., Li, H.Z., Zhang, P.C., Jiang, J., 2009. Impacts of hydrodynamic shear force on nucleation of flocculent sludge in anaerobic reactor. Water Res. 43, 3029e3036. Zhang, Z.P., Adav, S.S., Show, K.Y., Tay, J.H., Liang, D.T., Lee, D.J., Su, A., 2008. Characteristics of rapidly formed hydrogen-producing granules and biofilms. Biotechnol. Bioeng. 101, 926e936. Zheng, Y.M., Yu, H.Q., 2007. Determination of the pore size distribution and porosity of aerobic granules using sizeexclusion chromatography. Water Res. 41, 39e46.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 4 5 e1 5 4
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Improvement of metal adsorption onto chitosan/Sargassum sp. composite sorbent by an innovative ion-imprint technology Huijuan Liu a,*, Fan Yang a,c, Yuming Zheng b, Jin Kang a,c, Jiuhui Qu a, J. Paul Chen b,** a
State Key Laboratory of Environmental Aquatic Chemistry, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, 18, Shuangqing Road, Beijing 100085, China b Division of Environmental science and Engineering, National University of Singapore, 10 Kent Ridge Crescent, Republic of Singapore 119260 c Graduate School, Chinese Academy of Sciences, Beijing 100039, China
article info
abstract
Article history:
Technology for immobilization of biomass has attracted a great interest due to the high
Received 16 March 2010
sorption capacity of biomass for sequestration of toxic metals from industrial effluents.
Received in revised form
However, the currently practiced immobilization methods normally reduce the metal
15 July 2010
sorption capacities. In this study, an innovative ion-imprint technology was developed to
Accepted 10 August 2010
overcome the drawback. Copper ion was first imprinted onto the functional groups of
Available online 17 August 2010
chitosan that formed a pellet-typed sorbent through the granulation with Sargassum sp.; the imprinted copper ion was chemically detached from the sorbent, leading to the
Keywords:
formation of a novel copper ion-imprinted chitosan/Sargassum sp. (CICS) composite
Biosorption
adsorbent. The copper sorption on CICS was found to be highly pH-dependent and the
Copper removal
maximum uptake capacity was achieved at pH 4.7e5.5. The adsorption isotherm study
Granulation
showed the maximum sorption capacity of CICS of 1.08 mmol/g, much higher than the
Ion-imprinted chitosan
non-imprinted chitosan/Sargassum sp. sorbent (NICS) (0.49 mmol/g). The used sorbent was
Sargassum sp.
reusable after being regenerated through desorption. The FTIR and XPS studies revealed that the greater sorption of heavy metal was attributed to the large number of primary amine groups available on the surfaces of the ion-imprinted chitosan and the abundant carboxyl groups on Sargassum sp.. Finally, an intraparticle surface diffusion controlled model well described the sorption history of the sorbents. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Biosorption has been considered as a promising technology for the removal of low-level toxic metals from industrial effluents and natural waters (Volesky, 2007; Mehta and Gaur, 2005; Wang and Chen, 2009). Marine algae have received greater attention because of their high metal biosorption
capacity, low cost, and renewable nature. They can effectively remove heavy metal ions with concentrations ranging from few ppm to several hundreds ppm. The maximum metal biosorption capacity ranging from 0.1 to 1.5 mmol/g biosorbent has been reported (Davis et al., 2003; Chen and Yang, 2005, 2006). However, biosorbents often have small size, weak mechanical strength and low density. Leaching of such
* Corresponding author. Tel./fax: þ86 10 62849160. ** Corresponding author. Fax: þ65 6872 5483. E-mail addresses:
[email protected] (H. Liu),
[email protected] (J.P. Chen). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.017
146
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 4 5 e1 5 4
organics as carbohydrates and protein may be serious when some of biosorbents (e.g. raw seaweed) are used. These drawbacks have hindered the applications for treatment of waste metallic streams from various sources. In recent years, various efforts have been focused on the immobilization of alga biomass, which is a potential way to overcome the disadvantages. Such supporting materials as lu and Yakup Arıca, 2009; Mata et al., 2009), alginate (Bayramog silica gel (Rangsayatorn et al., 2004), polyacrylamide and polyurethanes (Mehta and Gaur, 2005) were used for the biomass immobilization. However, metal uptake efficiency of immobilized cells is reportedly often much lower than that of raw biomass. The functional groups on biomass for metal binding may become less available due to the immobilization and thus the sorption becomes normally less. For example, the cadmium sorption capacity drops from 98 mg-Cd/g for the Spirulina platensis to 71 and 37 mg-Cd/g for the alginate and silica immobilized species, respectively (Rangsayatorn et al., 2004). Mata et al. (2009) reported that the sorption capacity of alginate immobilized Fucus vesiculosus for copper was reduced nearly 60% due to immobilization. Synthetic polymers, such as polyacrylamide and polyurethanes, may be used for immobilization of biomass; the high cost and toxicity restrict their applications (Mehta and Gaur, 2005). Chitosan produced by partial deacetylation of chitin is the second most abundant biopolymer next to cellulose in nature. In addition to the high adsorption capacity for various heavy metals (Bassi et al., 2000; Guibal, 2004), chitosan can easily form hydrogel (Zhao et al., 2007). Thus, chitosan may be a good carrier for immobilization of biomass. However, the chitosanbased hydrogel are often poorly resistant to acid and has weak mechanical strength. In order to overcome the weaknesses, a cross-linking approach has commonly been used; the disadvantage is that the sorption capacity becomes reduced after cross-linking (Hsien and Rorrer, 1997; Ruiz et al., 2000). In this study, an innovative ion-imprint technology illustrated in Fig. 1 was developed and used for the immobilization
of alga biomass with the objectives of achievement of the higher sorption capacity of biomass and prevention of leaching of organics. Metal ion is used as an imprint ion to first “occupy” some of the functional groups (adsorption sites) in immobilization agent that can be used for metal adsorption shown as Step (1). The immobilization agent and biomass are solidified into a pellet shown as Step (2). Finally, elution agents (e.g., EDTA and NaOH) are used to strip the “pre-occupied” metal ions from the functional groups (elution of imprinted metal ions) so that they become available for metal sorption illustrated as Step (3). The chitosan was selected as an immobilization agent for Sargassum sp., which has higher metal uptake capacity and is abundant in many parts of the world (Chen and Yang, 2005, 2006; Mehta and Gaur, 2005). More importantly, it originates from natural living organisms and is not toxicity to human beings. A copper imprinted chitosan/Sargassum (CICS) composite sorbent was prepared according to the abovementioned ion-imprint technology. The adsorption properties such as pH effect and adsorption isotherm were studied. The surface characteristics of sorbent and sorption mechanisms were elucidated through Fourier transform infrared spectroscopy (FTIR) and X-ray photoelectron spectroscopy (XPS) analysis.
2.
Materials and methods
2.1.
Materials
The raw biomass of Sargassum sp. was collected from the coast of Singapore. The biomass was first washed with deionized (DI) water, and then dried overnight in an oven at 60 C. The dried seaweed was ground to fine particles with size below 150 mm. Chitosan (90% deacetylation), acetic acid and copper sulfate were purchased from Sinopharm Chemical Reagent Company (China). Sodium pyrophosphate was provided by
Fig. 1 e Demonstration of fabrication of metal imprinted composite sorbents.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 4 5 e1 5 4
Tianjin Fuchen Chemical Reagent Company (China). Potassium oxalate and sodium nitrate were obtained from Beijing Chemical Reagent Company (China). All the chemicals were of analytical grade.
2.2.
Sorbent preparation
0.01 g CuSO4$5H2O was dissolved into 30 mL acetic acid (0.5 mol/L), followed by an addition of 0.5 g chitosan. Ultrasonication was used to promote the dissolution of chitosan. 0.05 mL epichlorohydrin was added to the mixture for 24 h at room temperature for cross-linkage of chitosan. 1.5 g Sargassum sp. was then added into the cross-linked mixture and stirred for 1 h. The mixture was dropwise injected into a sodium pyrophosphate and potassium oxalate solution to form small pellets. The pellets were aged for 30 min in the solution to enhance their mechanical stabilities. They were collected by filtration, and the residual sodium pyrophosphate and potassium oxalate solution were removed with a DI water. Finally, the pellets were eluded by 6.8 mmol/L EDTA for 24 h and then with 0.03 mol/L NaOH for 2 h in a shaking bath to strip the chelated Cu(II) on the chitosan. The CICS composite sorbent was rinsed by DI water and dried at 50 C in an oven. In order to find out the effect of imprinted copper ion on the improvement in metal sorption, two other sorbents were prepared. A non-imprinted chitosan/Sargassum sp. (NICS) sorbent was prepared according to the above approach with an exception that the copper ion was not added. The cross-linked chitosan (CLC) was prepared without copper and Sargassum sp. The specific surface areas of sorbents were measured by nitrogen adsorption/desorption isotherm method at liquid nitrogen temperature using the Accelerated Surface Area and Porosimetry (ASAP 2000, U.S.A., Micromeritics Co.). The BrunauereEmmetteTeller (BET) model was applied to calculate the apparent surface area.
2.3.
Batch biosorption experiments
All experiments were conducted at 25 0.1 C. For kinetics experimental study, 2 g CICS composite sorbent and NICS sorbent were respectively added into 500 mL of 1.7 mmol/L cupric sulfate solution with an ionic strength of 0.01 mol/L NaNO3 at pH 5.0. The mixture was shaken at 130 rpm. The samples were taken at different time interval and analyzed for the copper concentrations by an inductively coupled plasmaoptical emission spectrometer (ICP-OES) (PerkineElmer Optima, 2000, USA). An intraparticle surface diffusion controlled model illustrated as follows was used to simulate the adsorption history: 2 v q 2 vq vq þ ¼ ; 0 r ap ; t > 0 Ds vr2 r vr vt
(1)
The initial and boundary conditions may be specified as Eqs. (2)e(4). vq ¼ 0; r ¼ 0 vr Ds
vq r ¼ kf ðC C Þ; r ¼ ap vr p
(2)
(3)
q ¼ 0; t ¼ 0
147
(4)
where C and q are the concentrations of the metal ions in bulk (mmol/L) and in solid phases (mmol/g), respectively, C* is the aqueous phase concentration at the particle surface (mmol/L), in equilibrium with the corresponding concentration in the solid phase q* (mmol/g), Ds is the surface diffusivity (m2/s), rp is the particle density (g/L), and kf is the external masstransfer coefficient (m/s). Eq. (1) with the initial and the boundary conditions can be numerically solved (Tien, 1994). In the pH effect experiment, 50 mL of 0.262 mmol/L cupric sulfate solutions with an ionic strength of 0.01 mol/L NaNO3 at various initial pH (2.0e5.5) were prepared in conical flasks. 0.25 g of CICS was added into the metal solutions. The flasks were shaken at 130 rpm for 24 h. The solution pH was measured and adjusted accordingly during the experiments by 0.1 mol/L HCl or 0.1 mol/L NaOH. In the sorption isotherm study, 0.1 g of biosorbent was added into a 50 mL cupric solution with different initial concentrations (0.12e2.5 mmol/L). The solution pH and ionic strength were controlled at 5.0 and 0.01 mol/L NaNO3, respectively. Other procedures were the same as those in the above pH effect experiment. The copper uptake at equilibrium was calculated by the following equation qe ¼
VðCi Ce Þ W
(5)
where qe is the equilibrium copper concentration in the sorbent (mmol/g), V is the solution volume (L), W is the amount of sorbent (g), and Ci and Ce are the initial and equilibrium sorbate concentrations in solution (mmol/L), respectively. The sorption data were fitted by the Langmuir isotherm equation as Eq. (6). qe ¼
qmax bCe 1 þ bCe
(6)
where qe and Ce are the same as description in Eq. (5), qmax is the maximum sorption capacity (mmol/g), b is the sorption affinity constant related to the binding energy of sorption (L/mmol).
2.4.
Desorption experiments
The sorption/desorption experiments were conducted to study the potential industrial applications of CICS composite sorbent. EDTA widely used in many studies to remove metal ions from metal loaded absorbent was tested for its suitability for the recovery of sorbent’s biosorption capacity. The experiments were conducted as follows. 400 mL copper solution with a concentration of 1.28 mmol/L and 2 g of CICS composite sorbent was stirred for 24 h at pH 5.0. The copper-loaded sorbent was separated from the solution and washed by using DI water. It was subsequently immerged into 20 mL of 0.1 mol/L EDTA mechanically stirred at 200 rpm for 24 h. The copper ion concentrations in the solution were analyzed. The desorbed adsorbents were then treated by 0.1 mol/L NaOH to neutralize hydrogen ions adhered on the adsorbent surface. The regenerated sorbent was used in the subsequent three-cycle sorption/desorption experiments.
148
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 4 5 e1 5 4
2.5.
Fourier transform infrared spectroscopy
FTIR spectroscopy was used to probe the surface characteristics of CICS composite sorbent for elucidation of the sorption mechanisms of copper ions. Specimens of various biosorbents were first mixed with KBr at an approximate ratio of 1/100 (sorbent/KBr) and then ground in an agate mortar. The resulting mixture was pressed at 10 tons for 5 min to form the pellet, which was characterized using a Nicolet 5700 FTIR spectrometer (Thermo, USA). Thirty-two scans and 2 cm1 resolution were applied in recording the spectra. The background obtained from scan of pure KBr was automatically subtracted from the sample spectra. All the spectra were recorded and plotted in the same scale on the absorbance axis.
2.6.
X-ray photoelectron spectroscopy
The chemical analyses on the virgin and metal loaded CICS composite sorbents were conducted by X-ray photoelectron spectroscopy (Kratos AXIS Ultra, UK). The XPS spectra were obtained by applying the monochromatic Al Ka X-ray energy source (1486.7 eV) operated at 15 kV and 10 mA. The wide scans were conducted from 0 to 1200 eV with pass energy of 160 eV. The high resolution scans were conducted according to the peak being examined with pass energy of 40 eV. To compensate for the charging effect, all spectra were calibrated with graphitic carbon as the reference at a binding energy of 284.6 eV. The software package Vision (PR2.1.3) and CasaXPS (2.3.12Dev7) were used to fit the XPS spectra peaks.
3.
Results and discussion
3.1.
Preparation of CICS composite sorbent
The ratio of chitosan to Sargassum sp. plays an important role during the preparation of CICS composite sorbent. As chitosan is more expensive than Sargassum sp., it is more economical to use more Sargassum sp. in the sorbent preparation. Three mass ratios of chitosan to Sargassum sp. (1:3, 1:4 and 1:5) were used in the preparation. At the mass ratio of 1:4 or 1:5, the slurry of chitosan and Sargassum sp. was highly thick, resulting to difficulty in the production of granulated sorbents. While the mass ratio was reduced to 1:3, the mixture became less viscous. Thus, the ratio was used in the fabrication of
sorbent. The CICS composite sorbent has a spherical shape (Fig. I of Supporting Information) with an average diameter of about 1.5 mm. It has been reported that organic matters can be leached from the marine algae during the sorption operation. Our previous study demonstrated that the organic concentrations of the aqueous solution were 110.9 and 186.3 TOC mg/L after contacting with 1 g/L Sargassum sp. for 24 h at controlled pHs of 5.0 and 2.0, respectively (Chen and Yang, 2005). The organic leaching from our CICS composite sorbent was studied; the sorbent was kept in contact with water for 24 h at a dosage of 1 g/ L. The TOC values were 14.9 and 22.2 mg/L at controlled pHs of 5.0 and 2.0, respectively. The comparison of the organic leaching of raw Sargassum sp. with the CICS composite sorbent clearly indicates that the CICS is more suitable for water treatment. The concentration of metal (copper ion) used as an imprint ion is another important factor, which would determine the sorption capacity. Table 1 shows that the presence of copper concentration as an imprint ion in the sorbent obviously improves the sorption. The content of imprint ion of 1.28 mg Cu/g sorbent is the optimum value to achieve the greatest sorption and thus was selected in the fabrication of the CICS. Under the optimum imprint ion content (1.28 mg Cu/ g), the sorption capacity of CICS composite sorbent is 0.212 mmol/g, 2.3 times of that of NICS (0.092 mmol/g) at an initial copper concentration of 0.56 mmol/L. The measurement of specific surface area of sorbents was conducted. It was found that CICS and NICS had specific surface area of 11.6 m2/g and 7.3 m2/g, respectively. The ionimprint operation has some improvement (enlargement) in the specific surface area, which is essentially beneficial to the adsorption kinetics and capacity.
3.2.
Sorption kinetics
Fig. 2 shows the kinetics of copper biosorption onto CICS and NICS. About 85% of total (ultimate) copper sorption on CICS and NICS rapidly occurs within 2 h, followed by a relatively slow process. The sorption equilibrium can be achieved within about 6 h. Comparison of this finding with those reported in literature shows that the immobilization does not obviously alter the biosorption kinetics (Kratochvil and Volesky, 1998; Chen and Yang, 2005). As the specific surface area of sorbents is relatively low, it can be assumed that the surface diffusion controls the
Table 1 e Effect of the imprinted Cu2D concentration in preparation on sorption capacity of CICS. Co
2þ Cu
(mmol/L)a
Sorption capacity of NICS, q (mmol/g)
Sorption capacity of CICS with different imprinted Cu2þ concentration, q (mmol/g) b CI 2þ Cu (mg/g sorbent)
0.56 1.00 3.00
0.092 0.152 0.325
0.67
1.28
2.64
0.145 0.278 0.593
0.212 0.341 0.749
0.148 0.273 0.582
a Co Cu2þ: the initial Cu2þ concentration in the solution (mmol/L). b CI Cu2þ: the imprinting Cu2þ concentration in the preparation (mg/g sorbent).
149
0.25
0.05
0.20
0.04 q (mmol/g )
q (mmol/g)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 4 5 e1 5 4
0.15
0.10
CICS
0.03
0.02
NICS
0.05
Modeling
0.01
0.00 0
200
400
600
800
1000
1200
1400
0
2
Time (min) Fig. 2 e Kinetics of copper biosorption by CICS and NICS. (m [ 4 g/L; pH [ 5.0; [Cu2D]0 [ 1.17 mmol/L; [NaNO3] [ 0.01 mol/L).
biosorption kinetics. The intraparticle surface diffusion controlled model was used in the simulation of adsorption history. As shown in Fig. 2, the model well describes the data of biosorption kinetics. Since the size of sorbents and stirring speed are the same in the experiments, the kf values of CICS and NICS are 6.4 105 m/s. The diffusivities of CICS and NICS are 1.0 1011 and 9 1012 m2/s, respectively. We previously reported that the kf and Ds of Sargassum sp. were 1.3 104 m/s and 3.7 1012 m2/s, respectively (Chen and Yang, 2005). The diffusivity of CICS is slightly higher than that of Sargassum sp., possibly due to the presence of chitosan used for the immobilization of Sargassum sp.. The external mass-transfer coefficient is lower than that of Sargassum sp. which likely results from the larger size and lower porosity of our sorbent.
5
6
Fig. 3 e pH effect on copper uptake by CICS. (m [ 5.0 g/L; [Cu2D]0 [ 0.262 mmol/L; [NaNO3] [ 0.01 mol/L).
metal ion binding, and hence the biosorption is enhanced. As the optimum pH for copper sorption on CICS was above 4.7, the pH of 5.0 was set in the subsequent sorption isotherm experiments.
3.4.
Sorption isotherms
The sorption isotherms of copper on CICS, NICS, CLC and Sargassum sp. are presented in Fig. 4. The amounts of copper adsorbed (q) increase with an increase equilibrium copper concentration in solution (Ce). The sorption data were fitted by the Langmuir equation and the calculated parameters are summarized in Table 2. It is shown that the Langmuir equation can well describe the
Effect of pH on copper sorption
The distribution of metal species in aqueous solution as a function of pH reveals that copper (maximum concentration: 2.0 mmol/L) ions precipitate in the forms of metal oxides or hydroxides at pH > 6 (Chen et al., 2003; Chen and Yang, 2005). Hence, the pH effect experiments were conducted at pH between 2 and 5.5, so sorption, not precipitation was responsible for the removal of copper. As shown in Fig. 3, the metal uptake increases with the increasing equilibrium solution pH and reaches a plateau at pH > 4.7, which is consistent with what was observed in the sorption of metals by biosorbents, such as prontonated brown alga Sargassum (Fourest and Volesky, 1996; Chen and Wang, 2001; Chen and Yang, 2006; Yang and Chen, 2008). At 2 < pH < 4, the amine groups on the surface of the sorbent could be easily protonated, which induces an electrostatic repulsion of copper ion. Competition between protons and Cu ions for sorption sites greatly decreases the sorption capacity (Chen and Wang, 2001; Lim et al., 2008). As pH is increased, the functional groups become more available for
4 pH
1.2
CICS Sargassum sp. NICS CLC Langmuir model
1.0 0.8 q (mmol/g)
3.3.
3
0.6 0.4 0.2 0.0 0.0
0.5
1.0
1.5 2.0 C ( mmol/L)
2.5
3.0
Fig. 4 e Copper biosorption isotherms of different adsorbents (CICS, NICS, CLC and Sargassum sp.) (m [ 2.0 g/L; pH [ 5.0; [NaNO3] [ 0.01 mol/L).
150
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 4 5 e1 5 4
Table 2 e Langmuir constants for copper sorption by different sorbents. qmax (mmol/g)
b (L/mmol)
r2
1.08 0.49 0.36 1.21
1.49 0.76 17.2 2.53
0.98 0.97 0.95 0.99
CICS NICS CLC Sargassum sp. powder
is as high as 96.8e98.3%. Most importantly, the sorption capacity is almost unaffected by the number of cycles, implying the higher possibility of the recovery of the copper ions and the reuse of the sorbent in industrial-scale applications. More studies are being conducted in our laboratory to further improve the sorbent in order that it will eventually behave like an ion exchange resin and can be used for many cycles (Chen et al., 2002).
3.6. sorption isothermal behaviors with high correlation coefficients (r2 ¼ 0.95e0.99). The maximum copper sorption capacities (qmax) of CICS, NICS, CLC, and Sargassum sp. are 1.08, 0.49, 0.36 and 1.21 mmol/g, respectively. The copper uptake is only reduced by 10% due to our ion-imprint immobilization of the raw biomass (CICS vs. Sargassum sp.). However, it increases by 120% due to the inclusion of raw biomass (CICS vs. NICS). As both CICS and NICS sorbents have chitosan and Sargassium sp., it can be concluded that more sorption sites become available on the CICS than the NICS as a result of ion-imprint modification. It was reported that the sorption capacities of copper on the granular activated carbon (Chen et al., 2003), the algal biomass (Aksu et al., 1992; Davis et al., 2003; Ozer et al., 2004), and the poly (vinyl alcohol) immobilized Sargassum (Sheng et al., 2008), and the calcium alginate-entrapped algae sorbent (Mehta and Gaur, 2001) were 0.2, 0.25e1.14, 0.21 and 0.7 mmol/g, respectively. It is clearly indicated that the ionimprinted composite sorbent developed in this study has a great potential for the treatment of heavy metal wastewater from various industries.
3.5.
Desorption study
Sorption mechanism of CICS
FTIR and XPS analyses were used to investigate the adsorption mechanisms of the CICS composite sorbent. Fig. 5 shows the FTIR spectra of Sargassum sp., CLC, NICS, CICS, copper-loaded CICS, and CICS with one cycle of sorptionedesorption. The broad and strong band ranging from 3200 to 3600 cm1 may be due to the overlapping of eOH and eNH stretching bands. The peak at wavenumber of 1539.5 cm1 in CLC, 1535.0 cm1 in NICS and 1538.4 cm1 in CICS can be assigned to the presence of eNH2 group (Clothup et al., 1990). The absorbance at above two wavenumbers in the CICS is much stronger than the CLC and NICS; this indicates that the modification by the ionimprint in the CICS preparation could preserve the more sorption sites for copper ions uptake. The absorbance at wavenumber of 1640.0 cm1 and 1421.5 cm1 in Sargassum sp. corresponds to stretching vibrations of carbonyl double bond (yC]O) and carboneoxygen single bond (yCeO), respectively (Clothup et al., 1990). The strong absorbance at 1645.4 cm1 and 1410.3 cm1 in CICS should be contributed by the functional groups from Sargassum sp. The vibration intensity of the eNH2 bonds at 1538.4 cm1 and the CeN stretching vibration at 1325.4 cm1 are reduced significantly after the copper sorption onto the CICS (Curve (e))
0.1 mol/L EDTA (wpH 4.5) was used as elusion agents for the desorption of copper ions adsorbed onto the CICS composite sorbent. 95% of desorption efficiency was achieved within 6 h. As the copper ions can strongly be chelated with the functional groups of EDTA, they can be easily released into the aqueous solution from the sorbent. After copper ions are desorbed, more hydrogen ions of EDTA become available and therefore adhered onto the surface of the sorbent. Such important functional groups on sorbent as eNH2 and eCOO are protonated to eNHþ 3 and eCOOH, leading to less metal binding on the sorbent (Su et al., 2003). As a result, 0.1 mol/L NaOH was used to neutralize hydrogen ions in the regeneration. The amount of copper ions adsorbed and the desorption efficiency in three consecutive sorptionedesorption cycles using EDTA are presented in Table 3. The desorption efficiency
Table 3 e Sorption and desorption behaviors of copper ions on Cu(II)eCTSS.
Cycle I Cycle II Cycle III
Uptake (mmol/g)
Desorption efficiency (%)
0.45 0.43 0.43
97.2% 98.3% 96.8%
Fig. 5 e FTIR spectra (a) Sargassum sp.; (b) CLC; (c) NICS; (d) CICS; (e) Copper-loaded CICS; and (f) CICS with one cycle of Cu2D sorptionedesorption.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 4 5 e1 5 4
and are recovered after the desorption (Curve (f)). These amino groups are involved in the copper sorption; the desorption by EDTA could restore the sorption ability of amino groups. After the copper ions are adsorbed onto the CICS, the carbonyl double bond stretching band exhibits a clear shift to a lower frequency at 1640.5 cm1, while the carboneoxygen single bond band shifts to 1408.8 cm1, corresponding to the complexation of copper to C]O and CeO bonds. This indicates the typical carboxylic absorption. The differences of frequencies of C]O and CeO bond stretching (D ¼ yC]OyCeO) in Sargassum sp., CICS, copper-loaded CICS and CICS suffered one sorptionedesorption cycle are 228.5, 235.1, 231.7 and 234.7, respectively. The difference between C]O and CeO bond stretching is related to the relative symmetry of these two carboneoxygen bonds and reflects the nature of carboxyl group binding status. The CICS has larger D value (235.1 cm1) than the metal loaded CICS (231.7 cm1). The change in D in the presence of copper ions clearly indicates more involvement of carboxyl groups forming complexes with copper ions (Chen and Yang, 2006). Another change to be noted in Fig. 5 is that the intensity of the peak representing the eCOO groups after copper sorption becomes much lower than that for the eCOO groups before the sorption (comparing the peak at 1645.4e1640.5 cm1 in Curves (d) and (e) of Fig. 5, respectively). The results indicate that eCOO groups are involved in the metal binding. Chen and Yang (2006) reported that the carboxyl groups of Sargassum sp. were the active functional groups for copper ion uptake and that the copper sorption on Sargassum sp. was mainly through the formation of Cu2þeCOO complex. These further confirm that both chitosan and Sargassum sp. are involved in the cupric sorption. The XPS studies of virgin and copper-loaded of CLC and CICS were conducted. The results of wide scan of samples given in Fig. II of Supporting Information clearly show a small peak around binding energy (BE) of 933 eV after the copper
151
biosorption on both NICS and CICS, indicating the accumulation of copper on the sorbents. The high resolution XPS spectra of N 1s are shown in Fig. 6. The deconvoluted N 1s spectrum of CICS shown in Fig. 6c comprises two peaks with BEs of 399.73 and 401.64 eV. The peak at 399.73 eV is attributed to the N atom in the ReNH2 group. The peak at 401.64 eV is assigned to a high oxidation states of nitrogen with positive charges (ReNHþ 3 ) (Jin and Bai, 2002; Liu and Bai, 2006). The similar N 1s spectrum of CICS as that of NICS confirms that the ion-imprinting process dose not influence the states of N atom. After the copper sorption, the BEs of the two peaks of N 1s shift to 400.15 and 402.14 eV, respectively. This is likely due to the formation of ReNH2Cu2þ complexes as shown in Eqs (7) and (8). A lone pair of electrons in the nitrogen atom is donated to the covalent bond between N and Cu2þ. As a consequence, the electron cloud density of the nitrogen atom is reduced, resulting in a higher BE peak observed. ReNH2 þ Cu2þ /ReNH2Cu2þ
(7)
2þ ReNHþ /ReNH2Cu2þ þ Hþ 3 þ Cu
(8)
The deconvolution of C 1s spectra of CICS in Fig. 7c produces four peaks with BE of 284.80, 286.41, 287.83 and 288.92 eV respectively. These peaks can be assigned to the C atom in forms of CeC, CeO, OeCeO and OeC]O, and the last three can be assigned to alcoholic, ether and carboxylate groups (Moulder et al., 1992; Lim et al., 2008). The carbon atoms in these three respective organic functional groups of the CICS are typical in algae polysaccharides and have different electron densities (Chen and Yang, 2006). After the copper binding onto the CICS, the binding energy of CeO, OeCeO and O¼CeO shifts to 286.57, 288.14 and 289.41 eV,
a
c
b
d
Fig. 6 e XPS N 1s spectra of adsorbent. (a) NICS; (b) Copper-loaded NICS; (c) CICS and (d) Copper-loaded CICS.
152
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 4 5 e1 5 4
a
e
b
f
c
g
d
h
Fig. 7 e XPS spectra of adsorbent. (a) C 1s of NICS; (b) C 1s of Copper-loaded NICS; (c) C 1s of CICS and (d) C 1s of Copperloaded CICS; (e) O 1s of NICS; (f) O 1s of Copper-loaded NICS; (g) O 1s of CICS and (h) O 1s of Copper-loaded CICS.
respectively. It indicates that alcoholic, ether and carboxylate groups in the CICS become involved in the cupric ion sorption, in which oxygen atom donates electrons to cupric ions and thus the electron density at the adjacent carbon atom in C]O and CeO decreases. The similar results can be found for the cases of NICS and copper-loaded NICS; this indicates that the sorption mechanism of CICS is the same as that of NICS. Furthermore, the O 1s spectra can be deconvoluted into two individual component peaks, which come from the different functional groups and overlap on each other, as
shown in Fig. 7eeh. The peaks at binding energy of 531.33 eV and 532.80 eV (for CICS) can be assigned to the oxygen atom in the forms of C]O (carboxyl and or quinine groups) and eOH or CeO groups, respectively. Fig. 7eeh shows that the binding energy of the peaks of the copper-loaded sorbent has a certain degree of shift, which is due to the copper ions bound onto the oxygen atoms and thus the electron density towards the oxygen atoms is decreased (Moulder et al., 1992). The changes in the BE of CeO and C]O indicate that both are involved in the sorption of copper, which are consonant with the FTIR analysis and the observations in C 1s analysis.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 4 5 e1 5 4
4.
Conclusion
A novel ion-imprinted chitosan/Sargassum sp. complex sorbent for effective copper ions removal is developed in this study. Three main steps are used in the fabrication of sorbent as illustrated in Fig. 1: First, the epichlorohydrin as a cross-linker is added to the mixture of copper, acetic acid and chitosan. Second, Sargassum sp. is added into the cross-linked mixture, which is then dropwise injected into a sodium pyrophosphate and potassium oxalate solution to form small pellet-typed sorbent. Third, the particle is eluded by EDTA and NaOH, forming the sorbent for improved sorption of heavy metal ions. The copper sorption on CICS is highly pH-dependent; it increases as the solution pH is increased. The maximum uptake capacity is achieved at pH 4.7e5.5. The sorbent has strong affinity towards copper ions with a maximal sorption capacity of 1.08 mmol/g. Lower organic leaching is found during the sorption. The adsorbed copper ions can be desorbed and the sorption capacity is almost not affected by the desorption process. The high efficiency, low organic leaching and reuseability of CICS make it a promising sorbent for the treatment of heavy metals from aqueous solutions. The FTIR and XPS studies show that the ion-imprinting method in sorbent preparation can shield the amine groups in chitosan in cross-linking process and enhance the sorption capacity. In addition, the carboxyl functional groups on Sargassum sp. play an important role in metal binding; this shows that Sargassum sp. contributes to the sorption capacity. An intraparticle surface diffusion controlled model is used to simulate the adsorption kinetics data. It is demonstrated that the model well describes the sorption history of the CICS and NICS.
Acknowledgements This work was supported by National Natural Science Foundation of China (Grant No. 50728806). The fellowship to YM Zheng funded by Agency for Science, Technology and Research, Singapore is appreciated (Grant No. 092101 0059).
Appendix. Supporting Information Supporting information of this article can be founded in online version at doi:10.1016/j.watres.2010.08.017.
references
Aksu, Z., Sag, Y., Kutsal, T., 1992. The biosorption of Copper(II) by C. vulgaris and Z. ramigera. Environ. Technol. 13, 579e586. Bassi, R., Prasher, S.O., Simpson, B.K., 2000. Removal of selected metal ions from aqueous solutions using chitosan flakes. Sep. Sci. Technol. 35, 547e560. lu, G., Yakup Arıca, M., 2009. Construction a hybrid Bayramog biosorbent using Scenedesmus quadricauda and Ca-alginate for biosorption of Cu(II), Zn(II) and Ni(II): kinetics and equilibrium studies. Bioresour. Technol. 100, 186e193.
153
Chen, J.P., Chua, M.L., Zhang, B.P., 2002. Effects of competitive ions, humic acid, and pH on removal of ammonium and phosphorous from the synthetic industrial effluent by ion exchange resins. Waste Manag. 22, 711e719. Chen, J.P., Wang, L., 2001. Characterization of a Ca-alginate based ion-exchange resin and its applications in lead, copper, and zinc removal. Sep. Purif. Technol. 36, 3617e3637. Chen, J.P., Yang, L., 2005. Chemical modification Sargassum sp. for prevention of organic leaching and enhancement of uptake during metal biosorption. Ind. Eng. Chem. Res. 44, 9931e9942. Chen, J.P., Yang, L., 2006. Study of a heavy metal biosorption onto raw and chemically modified Sargassum sp. via spectroscopic and modeling analysis. Langmuir 22, 8906e8914. Chen, J.P., Yoon, J.T., Yiacoumi, S., 2003. Effects of chemical and physical properties of influent on copper sorption onto activated carbon fixed-bed columns. Carbon 41, 1635e1644. Clothup, N.B., Daly, L.H., Wiberley, S.E., 1990. Introduction to Infrared and Raman Spectroscopy, third ed. Academic Press, London. Davis, T.A., Volesky, B., Mucci, A.A., 2003. Review of the biotechnology of heavy metal biosorption by brown algae. Water Res. 37, 4311e4330. Fourest, E., Volesky, B., 1996. Contribution of sulphonate groups and alginate to heavy metal biosorption by the dry biomass of Sargassum fluitans. Environ. Sci. Technol. 30, 277e282. Guibal, E., 2004. Interactions of metal ions with chitosan-based sorbents; a review. Sep. Purif. Technol. 38, 43e74. Hsien, T.Y., Rorrer, G.L., 1997. Heterogeneous cross-Linking of chitosan gel beads: kinetics, modeling, and influence on cadmium ion adsorption capacity. Ind. Eng. Chem. Res. 36 (9), 3631e3638. Jin, L., Bai, R.B., 2002. Mechanisms of lead adsorption on chitosan/ PVA hydrogel beads. Langmuir 18, 9765e9770. Kratochvil, D., Volesky, B., 1998. Biosorption of Cu from ferruginous wastewater by algal biomass. Water Res. 32, 2760e2768. Lim, S.F., Zheng, Y.M., Zou, S.W., Chen, J.P., 2008. Characterization of copper adsorption onto an alginate encapsulated magnetic sorbent by a combined FTIR, XPS and mathematical modeling study. Environ. Sci. Technol. 42, 2551e2556. Liu, C.X., Bai, R.B., 2006. Adsorptive removal of copper ions with highly porous chitosan/cellulose acetate blend hollow fiber membranes. J. Memb. Sci. 284, 313e322. Mata, Y.N., Bla´zquez, M.L., Ballester, A., Gonza´lez, F., Mun˜oz, J.A., 2009. Biosorption of cadmium, lead and copper with calcium alginate xerogels and immobilized Fucus vesiculosus. J. Hazard. Mater. 163, 555e562. Mehta, S.K., Gaur, J.P., 2001. Removal of Ni and Cu from single and binary metal solutions by free and immobilized Chlorella vulgaris. Eur. J. Protistol. 37, 261e271. Mehta, S.K., Gaur, J.P., 2005. Use of algae for removing heavy metal ions from wastewater: progress and prospects. Crit. Rev. Biotechnol. 25, 113e152. Moulder, J.F., Stickle, W.F., Sobol, P.E., Bomben, K.D., 1992. Handbook of X-ray Photoelectron Spectroscopy: a Reference Book of Standard Spectra for Identification and Interpretation of XPS Data; Perkin-Elmer Corp. Physical Electronics Division, Eden Prairie, MN. Ozer, A., Ozer, D., Ekiz, H.I., 2004. The equilibrium and kinetic modeling of the biosorption of Copper(II) ions on Cladophora crispata. Adsorption 10, 317e326. Rangsayatorn, N., Pokethitiyook, P., Upatham, E.S., Lanza, G.R., 2004. Cadmium biosorption by cells of Spirulina platensis TISTR 8217 immobilized in alginate and silica gel. Environ. Int. 30, 57e63. Ruiz, M., Sastre, A.M., Guibal, E., 2000. Palladium sorption on glutaraldehyde-crosslinked chitosan. React. Funct. Polym. 45, 155e173.
154
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 4 5 e1 5 4
Sheng, P.X., Wee, K.H., Ting, Y.P., Chen, J.P., 2008. Biosorption of copper by immobilized marine algal biomass. Chem. Eng. J. 136, 156e163. Su, H.J., Wang, Z.X., Tan, T.W., 2003. Adsorption of Ni2þ on the surface of molecularly imprinted adsorbent from Penicillium chysogenum mycelium. Biotechnol. Lett. 25, 949e953. Tien, C., 1994. Adsorption Calculations and Modeling. Butterworth-Heinemann, Boston, MA.
Volesky, B., 2007. Biosorption and me. Water Res. 41, 4017e4029. Wang, J.L., Chen, C., 2009. Biosorbents for heavy metals removal and their future. Biotechnol. Adv. 27, 195e226. Yang, L., Chen, J.P., 2008. Biosorption of hexavalent chromium onto raw and chemically modified Sargassum sp. Bioresour. Technol. 99, 297e307. Zhao, F., Yu, B.Y., Yue, Z.R., Wang, T., Wen, X., Liu, Z.B., Zhao, C.S., 2007. Preparation of porous chitosan gel beads for Copper(II) ion adsorption. J. Hazard. Mater. 147, 67e73.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 5 e1 6 2
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Characterization of endotoxic indicative organic matter (2-keto-3deoxyoctulosonic acid) in raw and biologically treated domestic wastewater Mokhtar Guizani a,*, Nogoshi Yusuke a, Mahmoud Dhahbi b, Naoyuki Funamizu a a
Environmental Engineering Department, Graduate School of Engineering, Hokkaido University, Kita 13, Nishi 8, Kita-ku, Sapporo, Hokkaido 060-8628, Japan b Laboratoire Eau et Technologies Membranaires, CERTE, Borj Cedria Technopark, BP 273 Soliman 8020, Tunisia
article info
abstract
Article history:
The aim of this research is to characterize the organic matter showing endotoxicity in
Received 30 April 2010
domestic wastewater. It is assumed that endotoxicity is caused by lipo-polysaccharide
Received in revised form
(LPS), particularly large and hydrophobic molecules. In this study, a batch experiment
5 August 2010
(decay test for 12 h) was conducted to confirm whether LPS is the cause of endotoxicity or
Accepted 10 August 2010
not. 2-keto-3deoxyoctulosonic acid (KDO) was used as an indicator of presence of LPS.A
Available online 17 August 2010
size and structural characterization of several samples from raw and domestic wastewater was also carried out in order know which fractions are causing endotoxicity. Endotoxin
Keywords:
and KDO patterns were found to be similar, peaking at the same time. Thus, organic matter
Endotoxin
showing endotoxicity, such as LPS was released in the decay test. Moreover, the organic
Hydrophobic
matter released from bacteria during decay test was partly biodegradable. Results from size
Lipo-polisaccharide
characterization (Molecular Weight Distribution) showed that the majority of endotoxin
Organic matter
(up to 82%), in domestic sewage and secondary effluents,is composed of molecules larger
Potable reuse
than 100 kDa and less than 0.1 mm. Similarly, structural characterization (hydrophobic and
Wastewater
hydrophilic) showed that the majority of endotoxin, ranging from 59% to 83% of the total endotoxicity, is hydrophobic fractions. Therefore, removing large and hydrophobic molecules from wastewater can be an effective way to achieve a significant decrease in its endotoxicity. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Potable reuse of domestic effluents is becoming widespread because of the increasing stress on water resources (Asano and Levine, 1998; MetCalf and Eddy, 2007). Rising concerns about water reclamation and reuse have led to increasing interest in removing organic micro-pollutants from treated water (Eckenfelder, 1994; Le-Clech et al., 2006; Rivera-Utrilla et al., 2006). Organic compounds found in reclaimed water
originate from three major sources: 1) Organic compounds added by consumers, 2) Natural Organic Matter (NOM), originally existing in water, and 3) treatment’s byeproducts and Soluble Microbial Products (SMP) generated during the wastewater treatment process due to the decay of bacteria. While it is possible to control a number of the contaminants at their sources, SMPs are almost inevitable in the effluent of the activated sludge process since they are associated with the biological reaction (Barker and Stuckey, 1999; Kimura et al., 2009).
* Corresponding author. Tel./fax: þ81 117066270. E-mail address:
[email protected] (M. Guizani). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.013
156
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 5 e1 6 2
In previous studies, lipo-polysaccharide (LPS) endotoxin has been detected in environmental water (rivers, surface water bodies, and aquifers) and treated wastewater. Data from these studies has shown that its concentration has a broad range of concentration levels, ranging from around 10 EU/ml to thousands of EU/ml. For example, LPS concentrations in Finnish waters experiencing cyanobacteria blooms have been measured at levels ranging from 20 EU/ml to 38,000 EU/ml (Rapala et al., 2002). Groundwater has a concentration ranging from 1 to 200 EU/ml (Anderson et al., 2002). Endotoxin concentrations in Yodo river basin has been assessed and found to be ranging from 311 to 2430 EU/ml. The effluents from sewage treatment plant were considered to be the cause for the increase of endotoxic activity in the river water (Ohkouchi et al., 2007). Moreover, tap water in both Japan and United States contains endotoxin levels ranging from 50 EU/ml to 200 EU/ml. Sewage and biologically treated wastewater have an extremely high LPS concentration ranging from around 1000 EU/ml to 9800 EU/ml (Guizani et al., 2009a, 2009b). These chemicals were found in the effluent of conventional wastewater treatment systems. One of the principal emergent issues associated with indirect potable reuse is LPS endotoxin resulting from the SMPs that are abundant in reclaimed wastewater. They are derived from cell membranes of Gram-negative bacteria and present in domestic wastewater and the effluent of treatment plants using biological processes. LPS endotoxins are toxic to most mammals regardless of their bacterial source. The injection of living or killed Gram-negative cells or small doses of purified LPS endotoxin into most mammals induces a wide spectrum of reactions such as: fever, changes in white blood cell counts, disseminated intravascular coagulation, hypotension, and organ dysfunction and may even results in septic shock and death. Moreover, toxicity has already been detected in reclaimed water and the secondary clarifier effluent’s toxicity is higher than that of the influent (Narita et al., 2007a). It was further found that these waters have high endotoxin concentration (Guizani et al., 2009a, 2009b). In addition, endotoxin exposure incidents have been well documented (Bhattacherjee et al., 1983; Ragnar and Birgitta, 2007; Reed and Milton, 2001; Thorn, 2001). It is known that LPS endotoxin has adverse health effects on humans in many ways and under some circumstances such as injection and inhalation (Olivier, 2003 and Williams et al., 2007). Hindman et al., 1975 reported that the presence of endotoxins in drinking water coincided with an epidemic of pyrogenic reactions among kidney dialysis patients. Anderson et al., 2002 reported that endotoxins present in tap water can be ingested but there is insufficient information available to quantify potential health risks. He also noticed that endotoxic pyrogenecity in humans depends on the species that originated the endotoxin. Furthermore, Narita et al. have reported that 220 mg/ml of LPS is sufficient to irreversibly disrupt tight junction permeability in caco-2 cells, known as the first line of defence against ingested toxic substances, in just 30 min (Narita et al., 2007b). Therefore LPS ingestion is considered to be toxic and it has been recently recognized as a water constituent of concern because of the potential human health effects of exposure through water drinking. Hence, as potable
reuse of reclaimed wastewater is gaining widespread acceptance, the removal of endotoxins from treated wastewater is imperative. As far as we know, no previous research has been conducted on the examination of the removal of endotoxic material from reclaimed wastewater. The few available published articles dealing with endotoxin removal from water concern drinking water and are exemplified in work by Jarkko et al., 2002; Yosuke et al., 1986 and Rezaee et al., 2008. Other investigations include descriptions of endotoxin removal from drugs, pharmaceuticals and bio-prcessing industry (Kathleen et al. 1977; Andrew et al., 2007 and Magalha˜es et al., 2007). However, concentration levels and the characteristics of LPS endotoxins in domestic sewage and biologically treated wastewater are not necessarily the same as those in either drinking water or drugs. Endotoxin removal from treated wastewater requires the advanced understanding of the characteristics of the organic matter responsible for endotoxic material. Therefore, the investigation of the characteristics of organic matter (OM) in wastewater, and in turn its endotoxicity, is of great importance for understanding removal possibilities of LPS endotoxin and the associated technologies that will be needed. The removal of endotoxin from reclaimed wastewater is an ultimate goal and to achieve it, one needs to know what chemicals cause this endotoxicity in order to target removal technologies and alternatives. It is known that the term LPS is commonly used interchangeably with endotoxin, but it is assumed that LPS is the main cause of endotoxicity, and thus the characterization of OM in terms of LPS is of great importance. LPS consists of three parts: a Lipid A, an O chain of Oligosaccharide and a core KDO (Williams et al., 2007). The KDO is unique and invariably present in LPS (Kenneth, 2008 and Ling et al., 2005). Therefore KDO can be used as an indicator of LPS found in biologically treated wastewater. In this study, a batch experiment (decay test for 12 h) was conducted to confirm whether LPS is the cause of endotoxicity or not. 2-keto-3deoxyoctulosonic acid (KDO) was used as an indicator of presence of LPS. Further, a size and structural characterization of several samples from raw and domestic wastewater was also carried out. For size characterization, a micro-filter (MF) and ultra-filter (UF) membrane with molecular weight cut-off (MWCO) ranging from 25 kDa to 0.1 mm were used to identify the molecular weight (MW) distribution of wastewater organic matter (OM) in terms of dissolved d-COD and endotoxicity. Moreover, Sep-Pak C18 cartridge were employed to quantify the amounts of the hydrophilic (HPLC)/hydrophobic (HPBC) organic fractions in different MW ranges. This characterization is imperative and will help in selecting appropriate, and advanced, alternative treatments to remove endotoxins from secondary treated wastewater.
2.
Material and methods
2.1.
Chemicals
1,2-Diamino-4,5-methylenedioxybenzene. 2HCl (MDB) and 2keto-3deoxyoctulosonic acid (KDO) and LPS standard were
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 5 e1 6 2
purchased from Wako Corp., Japan. An endotoxin kit was purchased from Seikagaku Corp., Japan.
2.2.
Water sampling and analytical techniques
2.2.1.
Batch tests (sludge)
2.2.2.
Composite samples
The characterization of organic matter showing endotoxicity in terms of structure and size is a preliminary step in the selection of technologies and alternatives for the removal of endotoxin from reclaimed wastewater. To achieve this characterization, fractionation campaigns were carried out on water samples from three different wastewater treatment plants in Sapporo, Japan (Plant A, Plant B and Plant C). An activated sludge process is used for wastewater treatment in all three plants. The plants A and B receive domestic wastewater only, while plant C receives domestic wastewater and rejected water from the sludge treatment facility. The amount of rejected water from the sludge treatment plant is approximately 13% of the total influent of Plant C. Water samples (Table 1) were collected from the influent and effluent of the three treatment plants and all samples were centrifuged and filtered with a 0.45 mm filter and analyzed for d-COD and endotoxin concentrations according to Standard Methods, 1989 and the Limulus Amebocyte Lysate (LAL) testing manual (Seikagaku manual, 2007). LPS endotoxin are part of the outer layer of Gram-negative bacterial cell walls. These are released in large amounts when these cells lyse in general, and during biological treatment of wastewater in particular. LAL is a reagent derived from amoebocytes in the blood cells of the Horseshoe Crab (Limulus polyphemus). This is an extremely sensitive indicator of the presence of bacterial endotoxin.
Table 1 e Samples for characterization. Samples Plant A Plant B Plant C
Influent Effluent Influent Effluent Influent Effluent
The dissolved d-COD and LPS endotoxin concentration of un-fractionated and fractionated samples were determined as shown in the following sections (structural analysis 2.5.1 & size analysis 2.5.2).
2.3.
In this section, we focused on 2-keto-3deoxyoctulosonic acid (KDO) in the LPS as an index of organic matter released from bacteria during the decay process. In order to evaluate the relationship between KDO and endotoxicity to determine whether the endotoxin originated from LPS or not, an aeration test for sludge was conducted. An activated sludge sample was taken from the return line of the secondary clarifier of a municipal wastewater treatment facility we identify as plant B at which a conventional activated sludge process is deployed. In the laboratory, the sludge was aerated for 12 h in absence of substrate and under controlled conditions, and samples were taken every 2 h, with KDO, endotoxin and d-COD levels measured in each sample. Two different trials were performed for each sample.
Dissolved COD mg/l
EC ms/cm
pH
61.3 13.9 75.4 16.7 77.7 18.2
0.57 0.5 0.65 0.81 0.53 0.67
7.2 6.8 7.2 6.6 6.8 6.5
157
Analysis of 2-keto-3-deoxyotulosonic acid (KDO)
A modified reverse phase High Performance Liquid Chromatography method was used to analyze KDO (Hara et al., 1987; Narita et al., 2005a), and afluorescence detector was used for detection of KDO. The column used was symmetric C18 (150 2.1 mm; particle size, 5 mm waters Inc.). The 0.45 mm filtrate was concentrated 10 times using a rotary evaporator (rotavap), and then hydrolyzed with 0.025 N HCl at 80 C in order to isolate KDO sugar from LPS. Fluorescence labelling was performed at 60 C for 150 min using 1,2-diamino-4,5methylenedioxybenzene 2HCl (MDB).
2.4.
Analytical method: sample preparation protocols
Dissolved chemical oxygen demand (d-COD) was measured according to standard methods after the samples were filtered through 0.45 mm filters to remove any suspended matter (Standard Methods, 1989). Endotoxin activity was analyzed using the LAL test kit (Seikagaku Corp., Tokyo, Japan) with the sample (200 ml) and LAL (200 ml) were mixed and incubated at 37 C for 30 min. The absorbance of the developed yellow colour was determined at 545 nm by the chromogenic endpoint approach method. The samples were diluted with LPS-free water (Seikagaku Corp., Tokyo, Japan) until their endotoxin levels were in the range of 0.01e0.1 endotoxin units [EU]/ml relative to the reference endotoxin (E. coli O113:H10) calibration system provided. The LPS units (EU/ml) of the samples were determined using the LAL coagulation data based on the reference endotoxin by linear regression analysis. The pH was adjusted to between 6 and 8 before measuring endotoxin concentration (Williams et al., 2007). Only the released part was measured and that still attached to cells was rejected using a 0.45 mm filter.
2.5.
Determination of OM fractions
2.5.1.
Structural analysis
Pre-filtered wastewater samples (0.45 mm filter) were processed through a Sep-Pak C18 cartridge to isolate the hydrophobic fractions by adsorption. Narita et al., 2005b mentions that, for simplicity, a Sep-Pak C18 cartridge can be used instead of XAD-8, which is more commonly used, as it gives the same fractionation performance (Narita et al., 2005b). As Fig. 1 shows, the division of OM into hydrophobic and hydrophilic fractions was based on the following procedure: samples were filtered through a 0.45 mm filter, then the samples were run through the cartridge, and subsequently, the effluent from the cartridge was collected as the hydrophilic fraction. Endotoxin activity and d-COD of the hydrophobic fraction were calculated from the difference between the d-COD of all dissolved organic matter and that of the hydrophilic fraction. Two independent replicates for each sample were performed, with mean and standard deviation of the data represented in the histograms.
158
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 5 e1 6 2
a
Water sample 0.45µm filter
Fractionation by Ultra-filter
30 COD(mg/ l)
Fractionation by C18 carthridge
COD
40
20
10
0
Hydrophilic fraction
#MW fractions
b
2
4
6 Time (Hrs)
8
10
KDO and endotoxin
0.50
5.4E+03 4.5E+03
Size analysis
The division of OM into different MW was performed using a micro-filter and ultra-filters with different molecular weight cut-offs (see Fig. 1). The pore sizes of the filters were 25 kDa, 100 kDa and 0.1 mm respectively. The filtrates were collected and the d-COD and endotoxin concentration were measured.Two independent replicates for each sample were performed, with mean and standard deviation of the data represented in the histograms.
3.
Results and discussions
3.1. OM released during biological reaction and KDO analysis
3.6E+03 0.30 2.7E+03 0.20 1.8E+03 0.10
9.0E+02 KDO
Endotoxin
0.00
0.0E+00 0
2
4
6
8
10
12
Time (Hours)
c
Ratio KDO to COD
0.05 0.04
KDO/ COD
2.5.2.
KDO (µ g/ ml)
0.40
Fig. 1 e Analytical procedure for the analysis of water sample.
12
Endot ox in Conc entrat ion (EU/ml)
0
0.03 0.02 0.01 0.00
Fig. 2,a b and c show the time course of d-COD, KDO concentration and the ratio KDO/d-COD respectively from two different runs of sludge aeration. Samples were hydrolyzed to release KDO from LPS and the KDO in these hydrolyzed samples gave the total amount of KDO including the KDO in LPS. Narita et al., 2005a reported that KDO after hydrolysis is more than that without treatment because KDO is comprised of two types: 1) liberated form, and 2) LPS and its fragments. From the aforementioned figures, which depict the time course of d-COD and KDO during the 12 h of aeration, the following patterns were observed: d-COD decreased during the first 6 h, whereas KDO increased (peaking after 6 h). This difference is indicative of the non-bacterial organic matter contained in the samples. The decrease in d-COD, shows that the organic matter originally existing in the sludge had been degraded. On the other hand, the increase in KDO reflects the release of the LPS during the decay process of sludge (bacteria cells). Also, the KDO decrease after 6 h shows that some KDO is biodegradable. This increase and then decrease of KDO concentration is consistent with the finding from Narita et al., 2005a showing that KDO increases then decreases during decay test.
0
2
4
6 Time (Hrs)
8
10
12
Fig. 2 e Fate of COD, KDO, endotoxin concentration and the ratio KDO to COD during decay test. (a) COD; (b) KDO and endotoxin; (c) Ratio KDO to COD.
The ratio KDO/d-COD of aerated sludge is plotted in Fig. 2c and characterises the organic matter from bacterial cells. As shown in the figure, the ratio KDO/d-COD increased during the aeration process from a very low value (0.005) at the beginning of the decay process, and reached approximately 0.015 after 12 h of aeration. This indicates that most of the organic matter is LPS, as after 2 h aeration, the d-COD fluctuation is not so high and remains almost constant; the increase of the ratio over time implies that the organic matter released from the decay process is probably biodegradable. In order to examine the relationship between endotoxin and KDO we compared the time course of KDO in Fig. 2b with that of endotoxin depicted also in Fig. 2b during the 12 h of the decay process. Similar patterns were observed: the presence and increases in both the endotoxin and KDO concentrations,
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 5 e1 6 2
159
c. In these figures the average d-COD, endotoxin activity, and their ratio are shown. Fig. 3a illustrates that dissolved organic matter with MW larger than 25 kDa and less than 0.1 mm, formed the largest OM constituent of the samples representing 60%e83% of the total. As for endotoxin concentration, Fig. 3b clearly illustrates that higher concentrations are observed in samples with MW ranging from 100 kDa to 0.1 mm. Thus, 80% of the endotoxin concentration is composed of MW size of 100 kDaw0.1 mm. In brief, the majority of dissolved organic matter is composed of MW size of 25 kDaw0.1 mm, while the majority of endotoxin concentration is composed of MW size of 100 kDaw0.1 mm. Kenneth, 2008 reports that pure LPS samples contain molecules larger than 10 kDa.The size of endotoxin molecules can range from smaller than 10 kDa in a monomeric form to larger than 10,000 kDa in an aggregated form (Belanich, 1996). Therefore, because they are larger than 100 kDa,the endotoxins in domestic wastewater are composed of aggregates. According to Gorbet and Sefton, 2006 and Darkow et al., 1999, endotoxins can self-assemble in a variety of shapes, with diameters up to 0.1 mm and MW of 1000 kDa, depending on the characteristics of the solution in which they are suspended. Considering the ratio of endotoxin to d-COD (shown in Fig. 3c), the highest value is observed in molecules ranging from 0.1 mm to 0.45 mm in size. This indicates that a considerable amount (about 80%) of the organic matter showing endotoxicity is larger in size than 0.1 mm. To summarize, the water samples characterization shows that organic matter with larger sizes (100 kDaw0.1 mm) exhibits higher endotoxin concentration. To remove a significant amount of endotoxins from secondary treated wastewater, it is necessary to target large molecules (>100 kDa).
3.3.
Fig. 3 e Size characterization of Organic matter showing endotoxicity in domestic wastewater. (a) COD; (b) endotoxin; (c) Ratio endotoxin to COD.
followed by a decrease in their concentration. The highest values of endotoxin and KDO concentration were observed at almost the same time (6 h aeration time). Knowing that KDO is released from the core part of lipo-polysaccharide endotoxin after hydrolysis, it can be said that decaying bacteria release endotoxic active material such as LPS.
3.2.
OM size distribution and endotoxin
The results of the size characterization of the organic matter in the samples are shown in the bar diagrams of Fig. 3,a, b and
Organic matter structure and endotoxin
Fig. 4 (a), (b) and (c) represent the OM fractions for sewage and treated water (hydrophobic/hydrophilic). The characterization was performed using Sep-Pack C18 cartridges. As shown in Fig. 4a, the dissolved d-COD concentrations of the hydrophilic fractions are higher than that of the hydrophobic fractions. The hydrophilic fraction of dissolved organic matter represents from 60 to 80% of the total dissolved organic matter. With respect to endotoxin concentration (Fig. 4b), the hydrophobic fractions are the major part, especially in the effluent. They represent at least 60% of the total endotoxin concentration in the influent and are slightly more than 59% in the case of effluent. It was reported that LPS is composed of two major parts, the hydrophobic and the hydrophilic portion (Williams et al., 2007). Yosuke et al., 1986 reported that a hollow fiber hydrophobic member with a pore size of 0.04 mm removed small-size endotoxin (smaller than the pore size of the membrane), indicating a hydrophobic character of endotoxin which is removed by adsorption to the membrane surface. Similarly, experimental findings from this research confirmed the hydrophobic character of LPS endotoxin in domestic wastewater. The endotoxin to d-COD ratio (Fig. 4c), indicates that hydrophobic fractions are much higher than the hydrophilic portions. The hydrophobic parts (HPBC) of dissolved organic matter are mainly endotoxic material. These HPBC fractions
160
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 5 e1 6 2
a
COD 80 HPLC
HPBC
70
COD (m g/ l)
60 50 40 30 20 10 0 Influent A
Effluent A
Influent B
Effluent B
Influent C
Effluent C
samples
b
Endotoxin 4.0E+03
HPLC
HPBC
3.5E+03
Endotoxi n (EU/ m l )
3.0E+03 2.5E+03 2.0E+03
With respect to dissolved organic matter, all fractions significantly decrease after biological treatment (Fig. 3a and b). The endotoxin fractions ranging from 0.1 mm to 0.45 mm and smaller than 100 kDa are initially present at low levels and then decrease slightly. However, despite its significant decrease, the fraction composed of 100 kDaw0.1 mm is still present at significant levels in the effluent. As shown in Fig. 3c, the ratio of organic matter showing endotoxicity increases in all cases with the highest level recorded for the fraction composed of 0.1 mme0.45um. Coagulation is a process of interest when it comes to reducing a significant amount of the remaining endotoxins in the effluent. Similarly, hydrophobic and hydrophilic fractions of endotoxin material are found to be reduced after treatment and the rate of reduction is more significant in the case of hydrophobic fractions. In addition, both hydrophilic and hydrophobic fractions of dissolved organic matter are biodegraded after treatment. Thus, most of the dissolved organic matter showing endotoxicity in the effluent is hydrophobic. Removal of these fractions from effluent is possible.
1.5E+03 1.0E+03
4.
Conclusion
5.0E+02 0.0E+00 Influent A
Effluent A Influent B
Effluent B
Influent C Effluent C
samples
Ratio Endotoxin to COD
c 4.5E+05 Rati o: Endotoxi n/COD (EU/ m g)
HPLC
HPBC
3.6E+05
2.7E+05
1.8E+05
9.0E+04
0.0E+00 Influent A
Effluent A
Influent B
Effluent B
Influent C
Effluent C
samples
Fig. 4 e Structural characterization of Organic matter showing endotoxicity in domestic wastewater. (a) COD; (b) endotoxin; (c) Ratio endotoxin to COD.
can be removed by making use of their tendency to attach to colloids and particulate matter; hence soil treatment can be advantageous. Therefore, removal of a large proportion of the endotoxin contained in wastewater can be through the removal of the hydrophobic organic matter. Hydrophobic membranes or soil treatment are two of the some the alternatives to be investigated in pursuing this goal.
3.4.
We investigated the characteristics of organic matter showing endotoxicity in secondary effluent of domestic wastewater. The conclusions resulting from this study are as follows: The main components of dissolved COD are solutes over 25 kDa and less than 0.1 mm (60e70% of the total d-COD). On the other hand, hydrophilic molecules constituted more than 60% of the COD. The main components of endotoxin material are solutes over 100 kDa and less than 0.1 mm (80% of the total endotoxin concentration); yet on the other hand, endotoxicity is mainly caused by hydrophobic portions of organic matter. The ratio endotoxin to COD showed that a large amount of endotoxin concentration is found in a relatively small amount of hydrophobic organic matter. The similar patterns of KDO and endotoxin concentration make it plausible to consider that the endotoxicity is, in fact, due to LPS. These results lead to the conclusions that removal of endotoxin from treated wastewater can be accomplished by the removal of the large MW and hydrophobic fractions. This can be done through the application of advanced treatment alternatives such as soil treatment and coagulation for hydrophobic fractions, and ultra-filters, nano-filters and reverse osmosis to remove the larger molecules. In light of these results, the removal efficiencies of such alternatives should be investigated in the near future.
Characteristics of endotoxic active organic matter
Acknowledgements Secondary effluent is the input to advanced treatment processes. It is therefore important to discuss the characteristics of the organic matter contained in the effluent of secondary wastewater treatment processes.
This work was supported in part by Sapporo Water Works (SWW) and the Tunisian government. The authors also thank Porter Wayne (Arizona State University-USA) and Nawaz Rab
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 5 e1 6 2
(Asian Institute of Technology-Thailand) for proof reading of the manuscript and linguistic support.
Appendix. Environment impact This is a first study on the characterization of organic matter showing endotoxicity in domestic wastewater. As endotoxin material have various impacts on water in environment and humans, this characterization will help in selecting treatment technologies and alternatives to remove endotoxin from secondary effluents. Such a study is likely to reduce the environment impacts on water sustainability and augmenting safe drinking water supplies.
references
Anderson, W.B., Slawson, R.M., Mayfield, C.I., 2002. A review of drinking water associated endotoxin, including potential routes of human exposure. Canadian Journal of Microbiology 48, 567e587. Andrew, C., James, K., John, L., 2007. Separation and Purification: endotoxin reduction using disposable membrane adsorption technology in cGMP manufacturing. BioPharm International 20 (5). APHA, AWWA, WPCF, et al., 1989. Standard Methods for the Examination of Water and Wastewater, seventeenth ed. American Public Health Association, Washington, DC. Asano, T., Levine, A.D., 1998. Wastewater Reclamation and Reuse. Wastewater Reclamation, Recycling, and Reuse: An Introduction. Technomic Publishing Co. Inc, Lancaster, PA. Barker, D.J., Stuckey, D.C., 1999. A review of soluble microbial products (SMP) in wastewater treatment system. Water Research. 33 (14), 3063e3082. Belanich, M., March 1996. Reduction of endotoxin in a protein mixture using strong anion-exchange membrane absorption. Pharmaceutical Technology, 142e145. Bhattacherjee, P., Williams, R.N., Eakins, K.E., 1983. An evaluation of ocular inflammation following the injection of bacterial endotoxin into the rate foot pad. Association for Research in Vision and Ophthalmology. 24, 196e202. Darkow, R., Groth, Th, Albrecht, W., Lu¨tzon, K., Paul, D., 1999. Functionalized nanoparticles for endotoxin binding in aqueous solutions. Biomaterials 20, 1277e1283. Eckenfelder Jr., W., 1994. Alternative strategies for meeting stringent effluent guidelines. Water Science and Technology 29 (8), 1e7. Gorbet, M.B., Sefton, M.V., 2006. Endotoxin: the uninvited guest. Biomaterials 26, 6811e6817. Guizani, M., Dhahbi, M., Funamizu, N., 2009a. Assessment of endotoxin activity in wastewater treatment plants. Journal of Environmental Monitoring 11, 1421e1427. Guizani, M., Dhahbi, M., Funamizu, N., 2009b. Survey on LPS endotoxin in rejected water from sludge treatment facility. Journal of Environmental Monitoring 11, 1935e1941. Hara, S., Takemori, Y., Nakamura, M., Ohkura, Y., 1987. Fluorometric high performance liquid chromatography of N-acetyl and N-glycolylneuraminic acids and its application to their microdetermination in human and animal sera, glycoproteins, and glycolipids. Analytical Biochemistry 164, 138e145. Hindman, S.H., Favero, M.S., Carson, L.A., Peterson, N.J., Schonberger, L.B., Solano, J.T., 1975. Pyrogenic reactions during haemodialysis caused by extramural endotoxins. Lancet, 732e734.
161
Jarkko, R., Kirsti, L., Leena, A.R., Anna-Liisa, E., Seppo, I.N., Kaarina, S., 2002. Endotoxins associated with cyanobacteria and their removal during drinking water treatment. Water Research 36, 2627e2635. Kathleen, J.S., Fort, M., Nelsen, L.L., 1977. Filtration Removal of Endotoxin (Pyrogens) in Solution in Different States of Aggregation. Applied and Environmental Microbiology 34 (4), 382e385. Kenneth, T., 2008. Bacterial Endotoxin (Online Textbook of Bacteriology). www.textbookofbacteriology.net. Kimura, K., Takuro, N., Yoshimasa, W., 2009. Changes in characteristics of soluble microbial products in membrane bioreactors associated with different solid retention times: relation to membrane fouling. Water Research 43 (4), 1033e1039. Le-Clech, P., Vicki, C., Tony, A.G., 2006. Fouling in membrane bioreactors used in wastewater treatment. Journal of Membrane Science 284 (1e2), 17e53. Ling, L.H., David, J.M., Kenneth, H.D., 2005. Field Guide for the Determination of Biological Contaminants in Environmental Samples, second ed. American Industrial Hygiene Association. Magalha˜es, P.O., Lopes, A.M., Mazzola, P.G., Rangel-Yagui, C., Penna, T.C., Pessoa Jr., A., 2007. Methods of endotoxin removal from biological preparations: a review. Journal of Pharmacy & Pharmaceutical Sciences 10 (3), 388e404. MetCalf, Eddy, 2007. Water Reuse: Issues Technologies and Applications. Mc-Graw Hill, NewYork, USA. Narita, H., Isshiki, I., Funamizu, N., Takakuwa, T., Nakagawa, H., Nishimura, S.I., 2005a. Organic matter released from activated sludge bacteria cells during their decay process. Journal of Environmental Technology, 433e440. Narita, H., Funamizu, N., Takakuwa, T., Kunimoto, M., 2005b. Role of hydrophilic organic matter on developing toxicity in decay process of activated sludge. Water Science and Technology 52 (8), 63e70. Narita, H., Funamizu, N., Takakuwa, T., Kunimoto, M., 2007a. Toxicity assessment of treated wastewater using cultured human cell lines. Environmental Monitoring and Assessement 129, 71e77. Narita, H., Talorete, T.P.N., Han, J., Funamizu, N., Hiroko, I., 2007b. Human intestinal cells incubated with activated sludge and lipopolysaccharide express HSP90b. Environmental Sciences. 14 (1), 35e39. Ohkouchi, Y., Ishikawa, S., Takahashi, K., Itoh, S., 2007. Factors associated with endotoxin fluctuation in aquatic environment and characterization of endotoxin removal in water treatment process. Environmental Engineering Research 44, 247e254. Olivier, M., 2003. Role of lipo-polysaccharide (LPS) in asthma and other pulmonary conditions. Journal of Endotoxin Research 9 (5), 293e300. Ragnar, R., Birgitta, F., 2007. Inflammatory responses by inhalation of endotoxin and (13)ed-glucan. American Journal of Industrial Medicine 25 (1), 101e102. Rapala, J., Lahti, K., Rasanen, L.A., Esala, A.L., Niemela, S.I., Sivonen, K., 2002. Endotoxins associated with cyanobacteria and their removal during drinking water treatment. Water Research 36 (10), 2627e2635. Reed, C.E., Milton, D.K., 2001. Endotoxin-stimulated innate immunity: a contributing factor for asthma. Journal of Allergy and Clinical Immunology 108 (2), 157e166. Rezaee, A., Ghanizadeh, G., Yazdanbakhsh, A.R., Behzadiannejad, G., Ghaneian, M.T., Siyadat, S.D., Hajizadeh, E., 2008. Removal of endotoxinin water using ozonation process. Australian Journal of Basic and Applied Sciences 2 (3), 495e499. Rivera-Utrilla, J.J., Mendaz-Diaz, M., Sanchez-Polo, M., FerroGarcia, M.A., 2006. Removal of the surfactant sodium dodecylbenzenesulphomate from water by simultaneous use of ozone and powder activated carbon: comparison with systems based on O3 and O3/H2O2. Water Research 40, 1717e1725.
162
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 5 e1 6 2
Seikagaku (Anonymous), 2007. Endospecy. In: Manual for Endotoxin Measurement. Seikagaku Corp., Tokyo, Japan. Thorn, J., 2001. The inflammatory response in humans after inhalation of bacterial endotoxin. Journal of Inflammation Research 50 (5). Williams, K.L., Roberts, K., Weary, M., Nnalue, N.A., Jorgensen, James H., F.C, 2007. Endotoxins: Pyrogens, LAL
Testing and Depyrogenation. In: Endotoxin Structure, Function and Activity. Published by CRC (Chapter4). Yosuke, S., Reiko, F., Ikuo, I., Atsushi, K., Teruo, K., Makoto, N., 1986. Removal of endotoxin from water by microfiltration through a microporous polyethylene hollow-fiber membrane. Applied and Environmental Microbiology 51 (4), 813e820.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 e1 7 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Mn oxide coated catalytic membranes for a hybrid ozonationemembrane filtration: Comparison of Ti, Fe and Mn oxide coated membranes for water quality S. Byun a, S.H. Davies a, A.L. Alpatova a, L.M. Corneal b, M.J. Baumann b, V.V. Tarabara a, S.J. Masten a,* a b
Department of Civil and Environmental Engineering, Michigan State University, East Lansing, MI 48824, USA Department of Chemical Engineering and Materials Science, East Lansing, MI 48824, USA
article info
abstract
Article history:
In this study the performance of catalytic membranes in a hybrid ozonationeceramic
Received 21 April 2010
membrane filtration system was investigated. The catalytic membranes were produced by
Received in revised form
coating commercial ceramic ultrafiltration membranes with manganese or iron oxide
23 July 2010
nanoparticles using a layer-by-layer self-assembly technique. A commercial membrane
Accepted 14 August 2010
with a titanium oxide filtration layer was also evaluated. The performance of the coated
Available online 21 August 2010
and uncoated membranes was evaluated using water from a borderline eutrophic lake. The permeate flux and removal of the organic matter was found to depend on the type of the
Keywords:
metal oxide present on the membrane surface. The performance of the manganese oxide
Ceramic membrane
coated membrane was superior to that of the other membranes tested, showing the fastest
Catalytic membrane
recovery in permeate flux when ozone was applied and the greatest reduction in the total
Drinking water treatment
organic carbon (TOC) in the permeate. The removal of trihalomethanes (THMs) and halo-
Ozonation
acetic acids (HAAs) precursors using the membrane coated 20 times with manganese oxide
Membrane filtration
nanoparticles was significantly better than that for the membranes coated with 30 or 40
Iron oxide
times with manganese oxide nanoparticles or 40 times with iron oxide nanoparticles. ª 2010 Elsevier Ltd. All rights reserved.
Manganese oxide Titanium oxide
1.
Introduction
As a result of the demand for high quality water there is a need for more effective, economical and energy efficient processes for the treatment of surface and contaminated water sources (Shannon et al., 2008). Over the last two decades, the use of membrane filtration for water treatment has expanded rapidly (Leiknes, 2009) as costs have decreased (Laine et al., 2000) and performance has increased (Song et al., 2003; Weber et al., 2003).
Although polymeric membranes are used in most water and wastewater applications, the use of ceramic membranes is becoming more common. For example, METAWATER (formerly NGK) has installed ceramic membranes in more than 70 water and wastewater treatment plants (Myers, 2009). One significant advantage of ceramic membranes is their long service life compared to polymeric membranes. In fact, Kru¨ger Inc. offers a 20-year warranty on their ceramic membranes (Myers, 2009). The chemical resistance of ceramic membranes allows for the use of ozone and other strong oxidants to
* Corresponding author. Tel.: þ1 517 355 2254; fax: þ1 517 355 0250. E-mail address:
[email protected] (S.J. Masten). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.031
164
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 e1 7 0
improve contaminant removal and reduce the problems associated with fouling. The use of ozone for pretreatment (Park, 2002; Lee et al., 2004) or in combination with membrane filtration (Schlichter et al., 2003; Karnik et al., 2005a; You et al., 2007; Kim et al., 2008) reduces membrane fouling. The effectiveness of the hybrid ozonationemembrane filtration process depends upon the gaseous ozone concentration; the pH of the feed water (Lee et al., 2004; Karnik et al., 2005a,b; Kim et al., 2008); operational parameters such as transmembrane pressure (TMP) and cross flow velocity (Kim et al., 2008); and the nature of the membrane surface (Karnik et al., 2005c, 2009). With both fluoropolymer and ceramic membranes, the concentration of residual ozone in the reject stream was found to be critical to system performance in terms of permeate flux (e.g., You et al., 2007; Karnik et al., 2005a). With a polyvinylidene fluoride ultrafiltration membrane, You et al. (2007) found that a dissolved ozone concentration of approximately 4.0 mg/L was necessary to maintain a permeate flux of 90% of its initial value. By contrast, with a titania membrane, Karnik et al. (2005a) found that fouling could be significantly reduced at much lower ozone concentrations; at an ozone residual concentration of 0.05 mg/L, in the recirculation feed water, the permeate flux was 95% of the clean water flux. The surface properties of ceramic membranes allow for the use of coating, doping and grafting techniques to improve membrane performance (Karnik et al., 2005c, 2006, 2007, 2009; Zhang et al., 2009; Wei and Li, 2009). Metal oxides promote the decomposition of ozone and the formation of OH or other radicals (for example, see Sa´nchez-Polo et al., 2006; Wu et al., 2008; Zhao et al., 2008, 2009). Consequently, the performance of the hybrid ozonationemembrane filtration system can be enhanced by coating the membrane surface with metal oxide catalysts such as iron oxide (Karnik et al., 2005c, 2006, 2007, 2009) and manganese oxide (Davies et al., 2010). Karnik et al. (2005b) reported the enhanced removal of the THM precursors and ozonation byproducts, such as aldehydes, ketones and ketoacids using hybrid ozonationefiltration and confirmed the OH radical reaction mechanism using salicylic acid as a probe (Karnik et al., 2007). The morphological characteristics related to enhanced catalytic efficiency of the iron oxide coated membranes were described by Karnik et al. (2005c, 2009). In this paper, the performance of ceramic membranes with three different filtration layers (Ti, Fe and Mn oxide), in a hybrid ozonationemembrane filtration system is compared. The permeability of the coated membranes and the removal of TOC, trihalomethanes (THMs) and haloacetic acids (HAAs) precursors by the hybrid ozonationefiltration system were studied to gain a better understanding of how the coating of the membrane affected the performance of the system.
2.
Experimental methods
2.1.
Feed water
The feed water was obtained from Lake Lansing (Haslett, Michigan), a borderline eutrophic lake. Water samples were collected at the boat ramp at the Lake Lansing Park e South, Haslett, Michigan in five-gallon carboys and stored at 4 C in a refrigerator. Water samples were pre-filtered through
a 0.5-mm ceramic cartridge microfilter (Doulton USA, Southfield, MI). Table 1 shows the water quality data obtained for the water used in this study. Simulated distribution system (SDS) THMs and SDS HAAs were measured after chlorination to simulate the formation of these disinfection byproducts (DBPs) in the distribution system. As shown in Table 1, seasonal variations of UV254 and SDS THMs and SDS HAAs were observed. The sample collected in Spring contained more compounds that absorbed UV radiation at 254 nm and had a higher concentration of HAA and THM precursors than that collected in Winter. Other water quality parameters for Lake Lansing have been reported elsewhere (Karnik et al., 2005a,b).
2.2.
Hybrid ozonationefiltration setup
Fig. 1 presents a schematic of the hybrid ozonationemembrane filtration apparatus used in these experiments. The apparatus consists of an ozone injection system, membrane module, recirculation pump, feed tank and data acquisition system. The membrane module housing (TAMI North America, St. Laurent, Que´bec, Canada) is made of stainless-steel. The system was operated in a total (permeate and retentate) recycling mode in which the permeate was, as shown in Fig. 1, recycled using a pump (Masterflex, Cole Parmer Inc., Vernon Hills, IL) into a feed tank at 15 min intervals using a timer. The transmembrane pressure (TMP) and cross flow velocity were controlled using a recirculation pump (Gear Pump Drive, Micropump, Cole Parmer Inc., Vernon Hills, IL) and a back pressure regulator (Swagelok, Solon, OH). The temperature of the water was 22.5 0.5 C during all the experiments. Temperature, cross flow rate and pressure were monitored continuously with a multifunctional sensor (L Series, Alicat Scientific, Tuscon, AZ) installed in the recirculation line and recorded using a LabView data acquisition system. The permeate mass was measured by an electronic balance (Adventure Pro Analytical Balance, Ohaus Corp., Pine Brook, NJ) at intervals of 60 s. The operational conditions for hybrid ozonationemembrane filtration are summarized in Table 2.
2.3.
Ozone injection system
Ozone was generated using a high-pressure ozone generator (Atlas Series, Absolute Ozone Generator, Absolute System Inc., Edmonton, Canada) that can operate as pressures of up to
Table 1 e Water quality characteristics of Lake Lansing (LL) after prefiltration. Water quality parameters pH TOC (mg C/L) UV254 (1/cm) SUVA (L/mg cm) SDS HAA (mg/L) SDS THM (mg/L) Conductivity (mS/cm)
Winter (2009) (Dec. 18)
Spring (2009) (Mar. 17)
8.0 10.4 0.12a 0.186 1.78 347 5 137 3 270
7.9 12.0 0.5 0.368 3.07 455 13 436 1 268
a Standard deviation for analysis.
165
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 e1 7 0
Fig. 1 e Schematic of the hybrid ozonationefiltration system.
8.3 bar (120 psig). To produce ozone gas, pure oxygen from a gas cylinder was dried using a moisture trap containing anhydrous calcium sulfate (21001, Drierite Co., OH, USA) and then fed to the high-pressure ozone generator. Ozone containing gas was injected directly into the pressurized membrane system. The gas pressure was monitored using a pressure gauge (Ashcroft Inc., Stratford, CT, USA) and was maintained at a pressure that was 0.2e0.3 bar greater than the transmembrane pressure (TMP). The flow rate of the ozone gas was regulated using a mass flow controller (Model GFC 17, Aalborg, Orangeburg, NY), which was installed between the ozone generator and the membrane module. The ozone concentration in the gas phase was measured by an ozone gas monitor (Model 450H, Teledyne Instruments, City of Industry, CA) and the dissolved ozone concentration in the retentate in the recirculation loop was continuously monitored using an amperometric ozone microsensor (AMT Analysenmesstechnik GmbH, Rostock, Germany). All data were recorded using the LabView program (National Instruments, Austin, TX). All gas concentrations and flow rates are reported at standard temperature and pressure.
2.4.
Membrane preparation
The membranes (Inside Ce´RAM, TAMI North America, SaintLaurent, Que´bec, Canada) used had a nominal molecular
weight cut-off of 5 kDa, an external diameter of 10 mm and an active length of 25 cm. The uncoated virgin membranes, as supplied by the manufacturer, have a TiO2 and ZrO2 filtration layer on a TiO2 support. The TiO2 was characterized as a rutile phase (Corneal, 2010). As measured from scanning electron micrographs, the thickness of the support layer is approximately 1 mm (Corneal et al., 2010), while the thickness of the filtration layer varies between 110 and 360 nm (Corneal, 2010). The grain size within the support layer and the filtration layer of the uncoated membrane (as received from the supplier) vary between 132 and 296 nm and between 1.05 and 6.64 nm, respectively. The average surface roughness of the uncoated and unsintered membrane is 150 57 nm, as measured by Atomic Force Microscopy (Corneal et al., 2010). The membrane had seven channels and a total filtering surface area of 131.9 cm2. The layer-by-layer method described by Karnik et al. (2005c) was used to coat the membranes with iron oxide nanoparticles. We have previously described the influence of the number of times the membrane was coated with iron oxide nanoparticles on system performance. The best results for the removal of TOC, ozonation byproducts and DBPs were obtained using a membrane that had been coated 40 times with iron oxide nanoparticles (Karnik et al., 2005c). The technique used to coat the membrane with manganese oxide nanoparticles is described by Corneal (2010) and is based on the layer-by-layer coating method developed by Lvov et al.
Table 2 e Operating conditions for the hybrid ozonationefiltration. Conditions Membrane filtration
Ozonation
a With the iron oxide coated membrane.
5 kD Transmembrane pressure (bar) Cross flow velocity (m/s) Initial water flux (L/m2 h) Temp. ( C) Applied ozone dosage (mg O3/s) Ozone inlet pressure (bar) Gas flow rate (mL/min)
a
1.24, 1.86e2.21 0.47 0.1 90 4 22.5 0.5 1.67 1.45,a 2.21e2.41 10
1 kD 2.8 0.1 0.30 0.1 35 1 23.0 0.5 3.67 3.2 1 40
166
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 e1 7 0
(2000) and Espinal et al. (2004). The manganese nanoparticles were synthesized by ozonating 1000 mL of ultrapure water for 20 min, then injecting 100 mL of a 2 mM manganese (II) chloride (Aldrich, 99þ%) solution. The resulting particles were gel-like flakes with an average size of 92 14 nm. They were settled by centrifugation and the supernatant was decanted and discarded. The solids were then mixed with 104 M KNO3, and sonicated using a 58 W sonicator (Model 250 Sonifier, Branson Ultrasonics Corporation, Danbury, CT) for 30 min in an ice bath to disperse the nanoparticles. To coat the membranes, the membrane channels were filled with a 0.2 wt% poly (diallyldimethylammonium chloride) (PDDA) solution (average MW <100 kDa, Aldrich Chemical Co., St. Louis, MO) for 15 min, rinsed with a 0.01 M NaOH solution (pH 12) for 15 s, filled with the suspension of manganese oxide nanoparticles for 15 min and then rinsed with a 0.01 M NaOH solution for 15 s (Corneal, 2010). This sequence deposits a single layer of particles on the membrane surface. The sequence was repeated for the desired number of times. Following the layer-by-layer deposition, the coated membranes were sintered in air at 500 C for 45 min to permanently fuse (sinter) the manganese oxide nanoparticles to the underlying ceramic membrane. After coating and sintering, the microstructure of the coated membranes was characterized by AFM, SEM and TEM. Details of the results for the iron oxide membrane are described by Karnik et al. (2006, 2009) and by Corneal (2010) for Mn oxide coated membrane. The Mn oxide coating was determined to be crystalline Mn2O3, with a thickness varying between 14 and 54 nm for the membrane coated 20 times, between 19 and 110 nm for the membrane coated 30 times and between 20 and 73 nm for the membrane coated 40 times. The iron oxide coated membrane was reported to have an average coating thickness of w46 nm for the membrane coated 40 times (Karnik et al., 2009). After the membrane coating and sintering process, the membrane was soaked in a 0.1 N NaOH solution overnight, then the ends were sealed with commercial polyurethane varnish (Minwax Corp., Upper Saddle River, NJ) by dipping each end of membrane in the polyurethane, then allowing it to dry for about 2 h before redipping the membrane. The membranes were dipped 20e40 times, after which the polyurethane was allowed to dry for 3 d. The coated membranes were labeled using the protocol: the nominal molecular weight cut-off in kDa e the number of coatings e sintering temperature. For example, 5e20 e 500 denotes a membrane with a molecular weight cut-off of 5 kDa, coated 20 times with nanoparticles and sintered in air at 500 C.
SDS THMs and SDS HAAs were measured according to the procedures in Standard Method 2350 (Clesceri et al., 1998) and EPA method 552.3 (US EPA, 2003), respectively. Water samples were dosed with chlorine at a concentration that ensured that the residual chlorine concentration was in the range 0.5e2 mg/L after 48 h incubation at room temperature. The THM compounds, chloroform (CHCl3), bromodichloromethane (CHBrCl2), dibromochloromethane (CHBr2Cl), and bromoform (CHBr3), were extracted from the water samples using hexane and analyzed by gas chromatography (Method 5710, Clesceri et al., 1998). A Perkin Elmer Autosystem gas chromatograph (Perkin Elmer Instruments, Shelton, CT) equipped with an electron capture detector (ECD), an autosampler, and a 30 m 0.25 mm ID, 1 mm DB5ms column (J&W Scientific, Folsom, CA) was used for the analysis. A 1 mL sample was injected onto the column. The oven temperature was ramped from 50 to 150 C at a rate of 10 C/min. The flow rate of the carrier gas (N2) was 12.0 mL/ min. The injector and detector temperatures were 275 C and 350 C, respectively. The concentrations of monobromoacetic acid (MBAA), monochloroacetic acid (MCAA), dichloroacetic acid (DCAA), trichloroacetic acid (TCAA), and dibromoacetic acid (DBAA) were determined by gas chromatography. A Perkin Elmer Autosystem gas chromatograph (Perkin Elmer Instrumens, Shelton, CT) equipped with an ECD, an autosampler, and a 30 m 0.32 mm ID, 3 mm DB-1 column (J & W Scientific, Folosom, CA) was used for the analysis. A 5 mL sample was injected onto the column. The oven temperature was programmed to hold for 15 min at 40 C, then increased to 75 C at a rate of 5 C/min and held for 5 min. The carrier gas flow rate (nitrogen) was 10 mL/min with the injector and detector temperature set to 200 C and 260 C, respectively.
2.5.
2.6.2.
Membrane cleaning and permeability tests
The virgin membranes were cleaned by soaking in a sodium hydroxide solution (0.1 N NaOH) overnight. The membranes were then rinsed with distilled deionized (DDI) water for at least 3 h. The fouled membranes were cleaned with DDI water using ozonationefiltration with 10 g/m3 gaseous ozone at a flow rate of 10 mL/min for 3e5 h until the initial clean water flux was restored. Before each filtration experiment, the
effectiveness of the cleaning procedure was verified by measuring the permeate flux through the membranes using DDI water to ensure that the permeability of the membrane was within (98 2)% of the initial value. The initial flux was set at (115 6) L/m2 h by adjusting the TMP. The TMP varied between 1.9 and 2.2 bar (27e32 psi) for the virgin and Mn oxide coated membrane. For the Fe2O3 coated membrane the TMP was 1.24 bar (18 psi) (see Table 1). The lower operating pressure was used with the Fe2O3 coated membranes, because the permeability of the iron oxide coated membrane was 1.7 times higher than that of the uncoated membrane, as sintering at 900 C leads to coarsening of the grains within the filtration layer (Corneal et al., 2010).
2.6.
Chemical analysis
2.6.1.
Total SDS THMs and HAAs
TOC and UV254
Samples for the measurement of TOC and UV absorbance at 254 nm were collected at the end of the 7 h ozonationefiltration, and purged with nitrogen gas for 1 min to remove any residual ozone. The TOC concentration was measured using a TOC analyzer (Model 1010 Analyzer, OI Analytical, College Station, TX). The UV absorbance at 254 nm (UV254) was measured by means of a UV spectrophotometer (Spectronic Genesys 5, Milton Roy Inc., Ivyland, PA) using a 1 cm quartz cell.
167
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 e1 7 0
3.
Results and discussion
3.1. The effect of the number of manganese oxide coatings on the permeate flux Fig. 2 shows the effect of the number of times the membrane was coated with manganese oxide nanoparticles on the transient behavior of the permeate flux. The results are presented as the normalized permeate flux (J/J0) vs. filtration time. As shown in Fig. 2, the J/J0 profiles with Mn oxide coated membranes show a similar two-stage pattern with an initial stage of rapid decline, followed by a long period of relatively slow increase until the end of the tests. For all three membranes, the flux begins to increase after approximately 1.5 h. The increase in membrane flux occurs after the initial ozone demand of the water is met and the residual ozone concentration increases. At this time sufficient ozone is present to oxidize the foulants present on the membrane and, so the flux increases. As can be also seen in Fig. 2, the permeate flux behavior shows a strong dependence on the presence of ozone as well as the number of times the membrane was coated with manganese oxide nanoparticles. The operational importance of this is that the permeate flux in the hybrid ozonationefiltration system can be controlled by coating the membrane surface with a catalyst and maintaining a sufficiently high ozone concentration in the retentate. Even with the uncoated membrane, the addition of ozone resulted in an increase in the permeate flux. This is consistent with published findings showing that the use of ozonation during membrane filtration ameliorates membrane fouling (Karnik et al., 2005a; You et al., 2007; Kim et al., 2008, 2009). Higher permeate fluxes were obtained using the manganese oxide coated membranes than with the uncoated membrane. With the membrane that had been coated twenty times (5e20 e 500), the permeate flux recovered to greater than 90% of the initial flux after 5 h. It is interesting to note that the recovery of the flux was faster for
the membrane coated 20 times than for those coated 30 or 40 times, although one might intuitively expect that performance would be enhanced by additional coatings. The reason for the better performance for the 20 layer Mn oxide coatings isn’t clear from material characterizations since increasing the number of coating layers did not result in a statistically significant increase in the coating thickness (Corneal, 2010) and there was also no statistically significant change in the roughness and grain size of the surface of the membranes after coating and sintering (Corneal, 2010). For the four membranes tested, the dissolved ozone in the retentate ranged from 0.31 to 0.38 mg/L (see Table 3). Notably, in experiments with the membrane (5e20 e 500) that showed the highest flux improvement, the dissolved ozone concentration was actually slightly lower than in experiments with other (5e30 e 500, 5e40 e 500) membranes. It appears that the small variations in the dissolved ozone concentration that were observed did not have a pronounced effect on the fouling behavior, since at higher dissolved ozone concentrations more, rather than less, fouling was observed.
3.2.
The effect of surface properties on fouling behavior
The effect of the composition of the membrane surface on the permeate flux can be observed in Fig. 3. Experiments were conducted with and without ozone. In all cases, the flux rapidly declined over the first 1e2 h due to fouling. No recovery of the permeate flux was observed without ozone (see Fig. 3a). As shown in Fig. 3b the least degree of fouling is seen with the manganese oxide coated membrane, the most with the iron oxide and the degree of fouling on the titania coated membrane is between these extremes. This behavior can be explained by the electrostatic interaction between the negatively charged NOM and the membrane surface. The manganese oxide has the lowest point of zero charge of the three membranes (pHzpc ¼ 2.8e4.5; Morgan and Stumm, 1964; Bernard et al., 1997), followed by TiO2 (pHzpc ¼ 4.1e6.2; Zhou et al., 2009; Zhang et al., 2009; Kim et al., 2009) and lastly, Fe2O3 (pHzpc ¼ 6e8; Mustafa et al., 2004; Pochard et al., 2002). At the pH of the natural water (about pH 8) virtually all the hydroxyl groups on the Mn oxide surface would be deprotonated, the surface hydroxyl groups on the TiO2 surface would be largely, but not completely, deprotonated and those on the iron oxide would only be partly deprotonated. Thus, the manganese oxide surface is likely to have the highest charge density, and the iron oxide the least charge density. As a result, the repulsive forces between the NOM and the oxide
Table 3 e Dissolved ozone concentration in the retentate. No. of coatings of Mn oxide Fig. 2 e Effect of the number of manganese oxide coatings on the normalized permeate flux (Conditions: Feed TOC [ 10.4 mg C/L, TMP [ 1.93e2.07 bar, J0 [ 92 ± 1 L/m2 h, Cross flow velocity [ 0.47 m/s, Temp. [ 22.5 C, Ozone dose [ 1.67 mg O3/s).
0 20 30 40
Dissolved ozone concentrationa (mg O3/L) 0.31 0.34 0.38 0.38
0.01 0.02 0.03 0.02
a measured during the final 30 min of the filtration experiments.
168
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 e1 7 0
ozonationefiltration. As shown in Fig. 4, in the presence of ozone, the concentration of TOC in the permeate and retentate was lower than that observed with conventional membrane filtration (i.e., without O3). TOC removal was dependent on the type of catalyst on the membrane surface and on the number of times the membrane was coated (in this work this was studied for manganese oxide only, we have studied the effect of the number of times the membrane was coated with iron oxide nanoparticles previously (Karnik et al., 2005c)). The TOC reduction obtained with the manganese oxide membrane coated 20 times was significantly higher at the 95% confidence interval ( p < 0.05) than that observed with the other membranes studied. The manganese oxide coated 30 or 40 times showed no statistical difference in TOC removal as compared to that obtained with the uncoated membranes (TiO2). Interestingly manganese oxide was the most effective catalyst for TOC removal even though the degree of fouling of this membrane was less than that of the iron oxide coated or uncoated membranes (see Fig. 3).
3.4.
Fig. 3 e Effect of the type of coating material on the normalized permeate flux (a) without and (b) with ozonation (Conditions: Feed TOC [ 10.4 mg C/L, TMP [ 1.24e2.28 bar, J0 [ 90 ± 4 L/m2 h, Cross flow velocity [ 0.47 m/s, Temp. [ 22.3 C, Ozone dosage [ 1.67 mg O3/s). Experiments were repeated in triplicate. Consistent results were observed with replicate experiments.
would be expected to be in the order Mn oxide > TiO2 > Fe2O3, which is the opposite to the degree of fouling observed. This is consistent with previous observations that repulsive electrostatic forces reduce the adsorption of organics and limit the membrane pore blocking, which reduces the degree of fouling (Seidel and Elimelech, 2002; Zhang et al., 2009). In the presence of ozone, the permeate flux begins to increase after approximately two hours (see Fig. 3b). This recovery is believed to be due to the catalytic ozonation of the foulants at the membrane surface (Karnik et al., 2005a, 2005c). The increase in permeate flux is in the order Mn oxide > Fe2O3 > TiO2, indicating that Mn oxide is the best catalyst.
3.3.
Effect of surface properties on the removal of TOC
To investigate the effect of coating type and the number of times the membrane was coated with manganese oxide nanoparticles on the oxidation of organic matter in water, TOC was measured in the retentate and permeate after 7 h of
Removal of DBP precursors
The removal of chlorinated DBP precursors was studied by evaluating the formation of THMs and HAAs in the permeate samples under conditions, which simulated those typically found in water distribution systems (see Fig. 5). The addition of ozone clearly enhanced the removal of THM precursors, except in the case of the uncoated (TiO2) membrane. The greatest removal of THM precursors was observed for the membrane coated 20 times with manganese oxide nanoparticles (5e20 e 500). The removal of HAA precursors was greatest for the membrane coated 40 times with manganese oxide nanoparticles.
Fig. 4 e Effect of type of coating material on the TOC concentration in (a) the retentate and (b) the permeate. Experimental conditions are as indicated in the captions for Figs. 2 and 3. Error bars denote the 95% confidence interval for the TOC analyses. The figures shown under bars indicate the number of times the membranes were coated with nanoparticles. The Mn coated membranes were sintered at 500 C and the iron oxide membrane at 900 C.
169
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 e1 7 0
Table 4 e Summary of the statistical analysis of membrane performance. The table indicates if there is a statistically significant difference in removal of TOC, SDS THM or SDS HAA by the hybrid process compared to conventional filtration using the uncoated membrane. Membrane coating
Ti oxide Mn oxide
Fe oxide
Uncoated 5e20 e 500 5e30 e 500 5e40 e 500 5e40 e 900
Water quality parameters TOC
THM
HAA
þ þþ þ þ þ
þ þþ þ þ e
þ þþ þþ þþ þþ
Not significant at 90% confidence interval. þ Significant at 90% confidence interval. þþ Significant at 95% confidence interval.
having a smaller MWCO (i.e., cut-off 1 kDa). In this experiment we achieved a 91 1% reduction in the SDS THMs when treating the same poor quality (Lake Lansing) water at an ozone dosage rate of 6.67 mg O3/s. Further work is ongoing to test the process using several different source waters using membranes with a 1 kDa MWCO.
4.
Fig. 5 e Effect of type of coating material and the number of layers of manganese oxide on the formation of (a) SDS THMs and (b) SDS HAAs in the permeate samples as compared to that observed in raw water (Conditions: TMP [ 1.93e2.07 bar, J0 [ 92 ± 1 L/m2 h, Cross flow velocity [ 0.47 m/s, Temp. [ 22.5 C, Ozone dose [ 1.67 mg O3/s). Error bars of all the values are the 90% confidence interval of triplicate tests.
As noted in Table 4, in all but one case (the removal of THM precursors by the iron oxide coated membrane), the improvement in the removal of TOC and DBP precursors by the hybrid process compared to that of conventional filtration using an uncoated membrane is statistically significant at the 90% confidence interval. However, only for the membrane coated 20 times with manganese oxide nanoparticles, this difference is significant at the 95% confidence level for all three parameters considered. Increasing the number of times the membrane was coated with manganese oxide nanoparticles to either 30 or 40 times did not significantly improve the removal of TOC or the DBP precursors. The levels of DBP precursors in Lake Lansing water are high and in this study, and although up to 39% removal of THM precursors and 55% removal of HAA precursors were achieved, we did not meet the regulatory limits for THMs and HAAs after chlorination using the membrane with a 5 kDa MWCO. Though the results obtained with the hybrid process are promising, further work is needed to optimize the treatment process for poor quality waters. We confirmed that these limits could be met using an uncoated membrane
Conclusions
The permeate flux and the removal of the organic matter by the process were found to depend on the type and the number of times the membrane was coated with the metal oxide nanoparticles. The performance of the manganese oxide coated membrane was superior to that of the other membranes tested. It showed the least degree of fouling and the greatest reduction of TOC in the permeate. The low degree of fouling observed with this membrane is due to the fact that its surface is negatively charged. The reduction in TOC during treatment is due to the fact that Mn oxide is an excellent catalyst for the oxidation of organic material. For the Mn oxide coated membranes, the number of times that the membranes were coated influenced the fouling behavior and TOC removal of the membrane. With the membrane that was coated with 20 layers of manganese oxide nanoparticles significantly better removal of THM and HAA precursors was achieved, than with the other membranes. The degree of fouling exhibited by this membrane was also less than that of the other membranes studied. Increasing the number of times the membrane was coated with manganese oxide coating up to either 30 or 40 did not lead to improved removal of THM and HAA precursors in the permeate.
Acknowledgement This material is based upon work supported by the National Science Foundation under Grant No. CBET e 0506828. Funding from the U.S. Environmental Protection Agency (US EPA) under Grant No. RD830090801 supported our earlier work on iron oxide coated membranes. Mr. Hugh MacDowell is also
170
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 e1 7 0
acknowledged for his assistance in measuring the SDS THMs in the water samples.
references
Bernard, S., Chazal, Ph., Mazet, M., 1997. Removal of organic compounds by adsorption on pyrolusite (b-MnO2). Water. Res. 31 (5), 1216e1222. Clesceri, L.S., Greenberg, A.E., Eaton, A.D. (Eds.), 1998. Standard Methods for Examination of Water & Wastewater, twentieth ed. American Public Health Association. Corneal, L.M., Masten, S.J., Davies, S.H.R., Tarabara, V.V., Byun, S., Baumann, M.J., 2010. AFM/SEM and EDS characterization of manganese oxide coated ceramic water filtration membranes. J. Memb. Sci. 360, 292e302. Corneal, L.M., 2010. The Development of MnO2 Coated Ceramic Membranes For Combined Catalytic Ozonation and Ultrafiltration of Drinking Water. Ph.D. dissertation, MichiganState University. Davies, S.H., Baumann, M.J., Byun, S., Corneal, L.M., Tarabara, V.V., Masten, S.J., 2010. Fabrication of catalytic ceramic membranes for water filtration. Water Sci. Technol. Water Supply 10 (1), 81e86. Espinal, L., Suib, S.L., Rusling, J.F., 2004. Electrochemical catalysis of styrene epoxidation with films of MnO2 nanoparticles and H2O2. J. Am. Chem. Soc. 126, 7676e7682. Karnik, B.S., Davies, S.H.R., Chen, K.C., Jaglowski, D.R., Baumann, M.J., Masten, S.J., 2005a. Effects of ozonation on the permeate flux of nanocrystalline ceramic membranes. Water Res. 39, 728e734. Karnik, B.S., Davies, S.H., Baumann, M.J., Masten, S.J., 2005b. The effects of combined ozonation and filtration on disinfection by-product formation. Water Res. 39, 2839e2850. Karnik, B.S., Davies, S.H.R., Baumann, M.J., Masten, S.J., 2005c. Fabrication of catalytic membranes for the treatment of drinking water using combined ozonation and ultrafiltration. Environ. Sci. Technol. 39, 7656e7661. Karnik, B.S., Baumann, M.J., Masten, S.J., Davies, S.H., 2006. AFM and SEM characterization of iron oxide coated ceramic membranes. J. Mater. Sci. 41, 6861e6870. Karnik, B.S., Davies, S.H., Baumann, M.J., Masten, S.J., 2007. Use of salicylic acid as a model compound to investigate hydroxyl radical reaction in an ozonationemembrane filtration hybrid process. Environ. Eng. Sci. 24 (6), 852e862. Karnik, B.S., Davies, S.H., BaumannCorneal, L.M., Masten, S.J., 2009. TEM characterization of iron-oxide-coated ceramic membranes. J. Mater. Sci. 44, 4148e4154. Kim, J., Davies, S.H.R., Baumann, M.J., Tarabara, V.V., Masten, S.J., 2008. Effect of ozone dosage and hydrodynamic conditions on the permeate flux in a hybrid ozonation e ceramic ultrafiltration system treating natural waters. J. Memb. Sci. 311, 165e172. Kim, J., Shan, W., Davies, S.H.R., Baumann, M.J., Masten, S.J., Tarabara, V.V., 2009. Interactions of aqueous NOM with nanoscale TiO2: implications for ceramic membrane filtration e ozonation hybrid process. Environ. Sci. Technol. 43 (14), 5488e5494. Laine, J.M., Vial, D., Moulart, P., 2000. Status after 10 years of operation e overview of UF technology today. Desalination 131, 17e25. Lee, S., Jang, N., Watanabe, Y., 2004. Effect of residual ozone on membrane fouling reduction in ozone resistance microfiltration (MF) system. Water Sci. Technol. 50 (12), 287e292. Leiknes, T., 2009. The effect of coupling coagulation and flocculation with membrane filtration in water treatment: a review. J. Environ. Sci. 21, 8e12.
Lvov, Y., Munge, B., Giraldo, O., Ichinose, I., Suib, S.L., Rusling, J.F., 2000. Films of manganese oxide nanoparticles with polycations or myoglobin from alternate-layer adsorption. Langmuir 16, 8850e8857. Morgan, J.J., Stumm, W., 1964. Colloid e chemical properties of manganese dioxide. J. Colloid. Sci. 19, 347e359. Mustafa, S., Tasleem, S., Naeem, A., 2004. Surface charge properties of Fe2O3 in aqueous and alcoholic mixed solvents. J. Colloid. Interface. Sci. 275, 523e529. Myers, N., 2009. New treatment plant to feature ceramic membranes. http://www.krugerusa.com/lib/kruger/ 3FEBE49iar679sFW9y1Pp1gM.pdf. Park, Y.G., 2002. Effect of ozonation for reducing membrane e fouling in the UF membrane. Desalination 147, 43e48. Pochard, I., Denoyel, R., Couchot, P., Foissy, A., 2002. Adsorption of barium and calcium chloride onto negatively charged a-Fe2O3 particles. J. Colloid. Interface. Sci. 255, 27e35. Sa´nchez-Polo, M., Ultrilla, J.R., von Gunten, U., 2006. Metal-doped carbon aerogels as catalysts during ozonation process in aqueous solution. Water. Res. 40, 3375e3384. Schlichter, B., Mavorv, V., Chmiel, H., 2003. Study of a hybrid process combining ozonation and membrane filtration e filtration of model solutions. Desalination 156, 257e265. Seidel, A., Elimelech, M., 2002. Coupling between chemical and physical interactions in natural organic matter (NOM) fouling of nanofiltration membranes: implications for fouling control. J. Memb. Sci. 203, 245e255. Shannon, M.A., Bohn, P.W., Elimelech, M., Georgiadis, J.G., Marin˜as, B.J., Mayes, A.M., 2008. Science and technology for water purification in the coming decades. Nature 452 (20), 301e310. Song, Lianfa, Hu, J.Y., Ong, S.L., Ng, W.J., Elimelech, M., Wilf, M., 2003. Performance limitation of the full-scale reverse osmosis process. J. Memb. Sci. 214, 239e244. US EPA (2003) Determination of haloacetic acids and dalapon in drinking water by liquideliquid microextraction, derivatization, and gas chromatography with electron capture detection, 815 e B e 03 e 002. Weber, R., Chmiel, H., Mavrov, V., 2003. Characteristics and application of new ceramic nanofiltration membranes. Desalination 157, 113e125. Wei, C.C., Li, K., 2009. Preparation and characterization of a robust and hydrophobic ceramic membrane via an improved surface grafting technique. Ind. Eng. Chem. Res. 48, 3446e3452. Wu, C.-H., Kuo, C.-Y., Chang, C.-L., 2008. Homogeneous catalytic ozonation of C.I. Reactive Red 2 by metallic ions in a bubble column reactor. J. Hazard. Mater. 154, 748e755. You, S.-H., Tseng, D.-H., Hsu, W.-C., 2007. Effect and mechanism of ultrafiltration membrane fouling removal by ozonation. Desalination 202, 224e230. Zhang, Q., Fan, Y., Xu, N., 2009. Effect of the surface properties on filtration performance of Al2O3eTiO2 composite membrane. Sep. Purif. Technol. 66 (2), 306e312. Zhao, L., Ma, J., Sun, Z., Zhai, X., 2008. Catalytic ozonation for the degradation of nitrobenzene in aqueous solution by ceramic honeycomb-supported manganese. Appl. Catal. B. 83, 256e264. Zhao, L., Sun, Z., Ma, J., 2009. Novel relationship between hydroxyl radical initiation and surface group of ceramic honeycomb supported metals for the catalytic ozonation of nitrobenzene in aqueous solution. Environ. Sci. Technol. 43, 4157e4163. Zhou, J., Zhang, X., Wang, Y., Hu, X., Larbot, A., Persin, M., 2009. Electrokinetic characterization of the Al2O3 ceramic MF membrane by streaming potential measurements. Desalination 235, 102e109.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 e1 7 8
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Biological nitrogen removal with a real-time control strategy using moving slope changes of pH(mV)- and ORP-time profiles S.G. Won a, C.S. Ra b,* a b
Department of Chemical and Biological engineering, University of British Columbia, Vancouver, B.C. Canada Department of Animal Life System, Kangwon National University, Chunchon, 200-701, South Korea
article info
abstract
Article history:
A new real-time control strategy using moving slope changes of oxidationereduction
Received 2 December 2009
potential (ORP)- and pH(mV)-time profiles was designed. Its effectiveness was evaluated
Received in revised form
by operating a farm-scale sequencing batch reactor (SBR) process using the strategy.
5 August 2010
The working volume of the SBR was 18 m3, and the volumetric loading rate of influent
Accepted 14 August 2010
was 1 m3 cycle1. The SBR process comprised six phases: feeding / anoxic /
Available online 25 August 2010
anaerobic / aerobic / settle / discharge. The anoxic and aerobic phases were controlled by the developed real-time control strategy. The nitrogen break point (NBP) in the pH(mV)-
Keywords:
time profile and the nitrate knee point (NKP) in the ORP-time profile were designated as
Oxidationereduction potential
real-time control points for the aerobic and anoxic phases, respectively. Through
pH(mV)
successful real-time control, the duration of the aerobic and anoxic phases could be
Moving slope change (MSC)
optimized and this resulted in very high N removal and a flexible hydraulic retention time.
Swine wastewater
Despite the large variation in the loading rate (0.5e1.8 kg NH4-N m3 cycle1) due to influent strength fluctuation, complete removal of NH4-N (100%) was always achieved. The removal efficiencies of soluble nitrogen (NH4-N plus NOx-N), soluble total organic carbon, and soluble chemical oxygen demand were 98%, 90%, and 82%, respectively. Monitoring the ORP and pH(mV) revealed that pH(mV) is a more reliable control parameter than ORP for the real-time control of the oxic phase. In some cases, a false NBP momentarily appeared in the ORP-time profile but was not observed in the pH(mV)-time profile. In contrast, ORP was more the reliable control parameter for NKP detection in the anoxic phase, since the appearance of NKP in the pH(mV)-time profile was sometimes vague. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Swine wastewater contains a large amount of nitrogen, phosphorus, potassium, and various other minerals. When discharged without proper treatment, these can cause serious environmental problems with severe land and water pollution. To prevent environmental pollution by swine wastewater, various aerobic processes such as activated sludge (Osada et al., 1991), rotating biological contactors (Kameoka
et al., 1986), biofilters (Bortone et al., 1994), oxidation ditches (Ghaly and Kok, 1986), and anoxic/aerobic processes (Fernandes et al., 1991; Pan and Drapcho, 2001) have been used during the last two decades, but their nitrogen and phosphorus removal is not stable and satisfactory. The use of a sequencing batch reactor (SBR) (Su et al., 1997; Ra et al., 1997; Bernet et al., 2000) with intermittent aeration (Bicudo and Svobada, 1995; Cheng and Liu, 2001) has improved nutrient removal efficiency in swine wastewater treatment.
* Corresponding author. Tel.: þ82 33 250 8618; fax: þ82 33 251 7719. E-mail address:
[email protected] (C.S. Ra). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.030
172
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 e1 7 8
The SBR process has also been successfully used to treat various wastewaters such as chemical wastewater and salinity wastewater as well as swine wastewater, since nutrient removal and solid/liquid separation is possible in the same tank (Obaja et al., 2003; Casellas et al., 2006; Deng et al., 2008; Kim et al., 2008). Compared to a continuous process, the SBR process is characterized by a more stable operation, but it requires a higher level of process control and automation (Pavselj et al., 2001). Control parameters commonly used for the biological nutrient removal (BNR) process include ORP, pH, dissolved oxygen (DO), and oxygen uptake rate (OUR). By monitoring the control parameters, it is possible to determine the termination points of the nitrification and denitrification (Plisson et al., 1996; Ra et al., 1997, 1999, 2000; Libelli, 2006). Computers can be used to detect DO, pH, and ORP patterns and provide rapid real-time information on the biological state of the system. Online measurements of DO, ORP, and pH contain characteristic patterns that indicate the end of the biodegradation process and there is thus great potential for online monitoring (Pavselj et al., 2001). The DO profile cannot provide enough information in the anaerobic condition. Moreover, ORP and pH are more stable than DO in the SBR process (Akin and Wgurlu, 2005; Tanwar et al., 2008; Gao et al., 2009) and hence they provide better control of the reaction status in the bioreactor. It is known that the ORP profile is very effective for anoxic phase control, but the pH profile cannot be used effectively because the real-time control point does not always appear (Kishida et al., 2003; Ga and Ra, 2009). A strategy using the absolute value of the control parameters is less reliable because of the signal drift under variable operation conditions. Thus, various real-time control strategies using specific parameter patterns have been proposed (Wareham et al., 1993; Yu et al., 1997; Ra et al., 1997, 1998, 2000; Kim et al., 2004; Ga and Ra, 2009). Control parameters for optimization of the aerobic and anoxic duration of the biological treatment process and dose control of the wastewater chlorination have also been evaluated (Ra et al., 2000; Guo et al., 2007; Ga and Ra, 2009). Ra et al. (1999) stated that the main advantages of real-time control are the achievement of consistent effluent quality and a relatively complete nutrient removal in spite of the fluctuation in influent strength, since the system HRT self-adjusts to system variation. Wang et al. (2009) reported that pH, ORP, and DO variations are closely related to the dynamic variations in nutrient pollutants, and real-time control could enable the treatment to adjust automatically to match the influent wastewater quality. In this study, we operated a farm-scale SBR process with a newly designed real-time control strategy using ORP and pH (mV)-time profiles. We evaluated its practicality and analyzed the operational characteristics of the real-time control process.
2.
Materials and methods
2.1.
System configuration and process operation
A schematic diagram of the pilot-scale SBR system is shown in Fig. 1. The working volumes of the influent storage tank, the
Fig. 1 e Schematic process with online control system.
SBR, and the effluent storage tank were 6 m3, 18 m3, and 6 m3, respectively, and the total volume of the SBR was 24.2 m3 (380 cm 300 cm 230 cm). A submersible aerator (KS, Inc.) and a mixer (KH, Inc.) were set at the bottom of the reactor to ensure aeration and complete mixing. To achieve real-time control and to monitor the biological status, ORP and pH probes were inserted into the reactor and connected to a custom-built amplifier for the accurate measurement of voltage. Electrical signals obtained from the reactor were relayed into a computer through an electric cable ribbon, and the computer was connected to the Internet. Swine wastewater was stored in the influent tank after solid/liquid separation using gravity. The SBR process was operated with the sequence presented in Table 1. After the swine wastewater was fed into the reactor, the mixer was turned on to provide an anoxic condition for the denitrification of NOx-N produced by the nitrification of NH4-N in the aerobic phase. As listed in Table 1, there was an anoxic phase for denitrification after the feeding of influent, but not after the aerobic phase. Thus, the NOx-N produced during the oxidation of NH4-N in the aerobic phase was partially discharged (1/18; the influent feeding volumecycle1and working volume of SBR was 1 and 18 m3, respectively) at the stage of final effluent and the remaining NOx-N in the reactor was denitrified after influent feeding. Because influent contains organic matter, which is required for the biological denitrification of NOx-N, the anoxic phase was introduced after influent feeding. The duration of the anoxic phase was determined by a real-time control strategy using ORP. After the detection of the NKP in the ORP-time profile, the reactor was kept in an anaerobic condition for 3 h before starting of aeration to induce phosphate release from the bacteria. Aeration was then turned on, and the oxic phase was maintained until the detection of the NBP in the pH(mV)-time profile. Sludge settlement and final effluent discharge then followed. The volume of influent loaded into the system was 1 m3 cycle1, and as a supplemental carbon source to enhance the denitrification of NOx-N and biological treatment, 2 L methanol was added to the reactor at the time of influent loading. To test the feasibility of the real-time control strategy in
173
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 e1 7 8
Table 1 e Operation modes for SBR. Sequences Duration (h)
Feeding 0.33
Anoxic RTC
Anaerobic
a
3
Aerobic RTC
b
Settle
Discharge
2
0.33
a Real-time control using ORP. b Real-time control using pH(mV).
a livestock farm situation in which there is no expert to perform regular sludge wastage, sludge wastage for solid retention time (SRT) control was not conducted during the operation.
2.2.
Real-time and remote control strategies
The real-time control strategy is shown in Fig. 2. After feeding the swine wastewater into the SBR, the computerized
Fig. 2 e Real-time control strategy of the process.
174
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 e1 7 8
program turned on the mixer and took readings from the reactor of the ORP and pH(mV) values every second; it calculated an average of 60 data points every minute. The average ORP and pH(mV) values were logged for the monitoring of profiles and further signal processing. The moving slope change (MSC) of the ORP and pH(mV) was calculated every minute, with a sample size of ten values (r ¼ 10). We have previously given the details of the MSC calculation and discussed the advantages of using pH(mV) rather than the unit pH value (Ga and Ra, 2009). To prevent erroneous process control, the control strategy was programmed to recognize each feature on ORP- and pH(mV)-time profiles in sequence (NKP / ASP / NBP) (Figs. 2 and 3). Because the MSC is proportional to the sample size, a value that can consistently be used to detect specific features in the ORP/pH (mV)-time profiles needs to be chosen as a trigger value to monitor the ORP/pH (mV) for a predetermined sample size. In this study, the real-time control point in the anoxic phase, NKP, was detected by tracking the changing pattern of the MSC (r ¼ 10) of ORP. A value for the MSC of the ORP-time profile (ORP-MSC) of 80 was used as the trigger value for the NKP; this indicates the completion of denitrification in the anoxic condition. Aeration started 3 h after the detection of the NKP in the ORP-MSC, and each designated MSC control point in the pH(mV)-time profile (pH(mV)-MSC) was detected
in sequence. The real-time control point (NBP) for the aerobic phase, indicating the termination of the nitrification reaction, was detected in sequence after the recognition of the aeration start point (ASP). With this strategy, real-time control of the aerobic phase began at the onset of the NBP in the pH(mV)MSC. The MSC values þ1.0 and 1.0 were used as trigger values for the ASP and NBP, respectively. At the onset of the NBP in the pH(mV)-MSC (Fig. 3A, B), aeration was terminated, and sludge settlement and final effluent discharge followed. The whole procedure was then reset for the next cycle after the feeding of influent into the reactor. The designated real-time control points, NKP for the anoxic phase and NBP for the aerobic phase, are shown in Fig. 3.
2.3.
The influent and effluent samples were collected and preserved at 4 C until analysis. Solid analyses were performed immediately after sampling and other chemical contents were measured within one week. The parameters studied were soluble total organic carbon (STOC), soluble chemical oxygen demand (SCODcr), ammonium nitrogen (NH4-N), nitrate and nitrite nitrogen (NOx-N), orthophosphate (PO3e 4 ), total solids (TS), total volatile solids (TVS), suspended solids (SS), and volatile suspended solids (VSS). Analyses were performed in accordance with the standard methods (A.P.H.A, 1995). STOC was analyzed with a Shimadzu total organic carbon analyzer (Model TOC-500), while NH4-N, NOx-N, and PO3e 4 were analyzed with an auto analyzer (Quick Chem 8000, LACHAT).
3.
Fig. 3 e Profiles of (A) ORP and pH(mV); (B) MSC.
Analytical methods
Results and discussions
Fig. 3 shows the changes in nitrogen (NH4-N and NOx-N) with the ORP and pH(mV)-time profiles in the anoxic and aerobic phases, as well as the specific change patterns of the MSC (r ¼ 10) for each parameter. Tracking of the NH4-N and NOx-N levels (Fig. 3A) shows that the designated NKP and NBP are the completion points of denitrification and nitrification, respectively. Fig. 3B also shows that the real-time control points can easily be detected by monitoring the specific change patterns of ORP- and pH(mV)-MSC. During the anoxic phase, which followed the feeding of the influent, the ORP value was kept constant, despite an active denitrification of NOx-N occurring in the reactor (Fig. 3A). The ORP value then fell abruptly and the NOx-N level reached zero at the same time, indicating that denitrification of NOx-N in the reactor was complete; this point was the NKP. According to Kishida et al. (2003), the start of sulfate-reducing activity producing sulfides causes the sudden decrease in the ORP value after denitrification. In addition, three different patterns for the ORP-time profile during the denitrification of NOx-N have been well described in a previous study (Ga and Ra, 2009). The NKP could be detected by monitoring the pH(mV)-time profile. The pH(mV) value slowly decreased while the denitrification was occurring and then began to increase after the completion of denitrification. This specific pattern in the pH (mV)-time profile in the anoxic phase might reflect alkalinity changes in the reactor. Since the alkalinity is recovered during
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 e1 7 8
the denitrification of NOx-N to N2 gas, the pH(mV) value would gradually decrease with the occurrence of denitrification; pH (mV) is inverse to unit pH. The observed increase in the pH (mV) value after complete denitrification could be due to the formation of the anaerobic condition and the following hydrolysis of organic matter into small molecules. Tanwar et al. (2008) reported that the fall in pH during the anaerobic and anoxic phases can be attributed to the release of acid by fermentation. As aeration was introduced, increases in the ORP and pH (mV) values were observed. The steep increase in ORP should be attributed to the sensitivity of ORP to O2. The ORP increase gradually became slower and then the ORP value jumped up again. It is well known that the ORP value increases rapidly when nitrification is completed in the reactor (Peddie et al., 1990; Ra et al., 1998, 2000; Hao and Kim, 2000). However, this point was not the termination point of nitrification in this study; it was a false nitrogen break point (FNBP). Ndegwa et al. (2007) stated in their relationship study of pH, DO, and ORP that all three parameters display features defining stabilization of the wastewater and hence all three can be used singly or in combination to monitor and/or to control the treatment process. However, the individual use of pH might pose more technical challenges than the individual use of either ORP or DO. In contrast, however, the FNBP in the ORP-time profile appeared momentarily during this study. The appearance of the FNBP in the ORP and DOtime profiles in some cases was also reported by Kishida et al. (2003). The occurrence of the FNBP could be due to the imbalance between the aeration rate and the OUR of aerobes or to the reactor loading rate. This observation of FNBP in the ORP-time profile surely indicates that erroneous control could result if real-time control is done using the ORP-time profile. The pH(mV) value increased continuously until NH4-N removal was completed in the aerobic condition, and this increase could be due to the consumption of alkalinity during the nitrification of NH4-N to NOx-N. Tanwar et al. (2008) also observed a steep fall in the pH (increase in pH-mV) during the initial aeration phase and reported that it was due to the alkalinity consumption during the nitrification of ammonia. The pH(mV) value began to drop at the completion of the nitrification, and this specific NBP consistently appeared in the pH(mV)-time profile but not in the ORP-time profile (see Fig. 3). Hence, real-time control of the aerobic phase using the pH(mV)-time profile is feasible. The continuous drop in the pH(mV)-time value after the completion of nitrification could be caused by the termination of alkalinity consumption due to the termination of nitrification and CO2 air-stripping by the aeration. The routinely appearing NKP and NBP in the anoxic and aerobic phases were easily detected by tracking the specific change patterns of ORP- and pH(mV)-MSC, respectively (Fig. 3B). The NKP occurring in ORP-MSC could be consistently detected by designating 80 as the trigger value. However, NKP detection using pH(mV)-MSC was not stable since the change pattern was sometimes vague. In contrast, the NBP was easily detected using the specific change pattern of pH (mV)-MSC. Although the NBP was the point where the MSC switched from positive to negative values, 1 was designated
175
as the trigger value to secure a stable process control, and successful real-time control of the aerobic phase was thus achieved. Fig. 4 shows the ORP and pH(mV)-time profiles observed during real-time control of the process using the specific patterns of the ORP- and pH(mV)-MSC changes. ORP- and pH (mV)-time profiles that clearly displayed the NKP and NBP respectively were consistently obtained throughout the experiment, and thereby successful real-time control was achieved using the developed control algorithm. The anaerobic condition was maintained artificially for 3 h after the detection of the NKP in the ORP-MSC-time profile, and the aerobic phase was consecutively begun and terminated when the NBP was detected in the pH(mV)-MSC-time profile. The NBP rarely appeared in the ORP-time profile during the study. With real-time control using the ORP and pH(mV)-time profiles, the duration of the anoxic and aerobic phases were flexible from cycle to cycle (Fig. 5). The time to the appearance of the NKP and NBP varied greatly from cycle to cycle, responding to the physical and biological conditions of the reactor. Analysis of cycles 143e170 revealed that the time taken for the detection of the NBP was much longer than that for NKP in most cases, resulting in the aerobic phase being much longer than the anoxic phase. Three hours should be added to the anoxic duration given in Fig. 5 for the calculation of the air-off duration in this study, since the air-off condition was maintained for 3 h after the detection of the NKP in the ORP-MSC. Therefore, it was found that the duration of the
Fig. 4 e Operational mode and real-time profiles of parameters.
176
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 e1 7 8
Fig. 5 e Variation of consumed time for detection of NKP and NBP.
aerobic phase was sometimes shorter than that of the anoxic phase in particular in cycles 165 and 166. The most evident advantage of real-time process control over a fixed-time cycle mode of operation is the complete removal of contaminants in the wastewater by the achievement of a flexible hydraulic retention time (HRT). This varies with both the wastewater quality and the bacteria activity in the system. A further advantage is the optimization of operating costs and system capacity. It is known that real-time control strategies based on occurrences of the NBP in the parameteretime profile allow the optimization of the energy cost for the aeration (Ra et al., 2000), and real-time control of the aeration improved the efficiency of the reactor (Chen et al., 2002). The average duration of cycles 143 to 170 was 14.6 h, having 4.01 h of air-OFF (mixer-ON) and 8 h of air-ON (Fig. 5). A comparison of these results with those of the fixed-time operation (24 h cycle1) revealed that approximately 39% of the total energy usage could be saved by the application of the real-time control strategy. Yu et al. (1997) also reported that approximately 35% of the overall cycle time could be reduced and 42% of the aeration
energy saved when real-time control was applied to a continuous-flow activated sludge reactor system. Table 2 gives the removal efficiencies of the pollutants and the characteristics of the influent and effluent. The average removal efficiencies of NH4-N, STOC, and SCODcr were approximately 100%, 90%, and 82%, respectively. The 100% removal of NH4-N is likely due to the real-time control of the process, since the aeration duration was self-adjusted or artificially maintained until nitrification of the NH4-N was complete when using the real-time control strategy. As seen in Fig. 6, which presents the tracked loaded and removed values of N cycle1, complete removal of NH4-N occurred under a wide range of loading rate conditions (0.5e1.8 kg cycle1). Therefore, a very high removal of soluble nitrogen (NH4-N plus NOx-N) resulted (approximately 98%). The NOx-N levels in the final effluent were in the range of 0e83 mg L1 (average 23 mg L1) during the operation. The non-complete removal of NOx-N could be due to the operational sequence, in which the final effluent was decanted after the nitrification without post-denitrification; in this study, an anoxic phase for the denitrification was provided only before the aerobic phase.
Table 2 e Removal efficiencies. Parameter (mg L1)
Influent Means
STOC SCODcr NH4-N NOx-N PO34 TS (g L1) TVS (g L1) SS (g L1) VSS (g L1)
1623 2346 1189 1.2 23.5 5.1131 2.5041 0.9589 0.9362
means standard deviation.
642 1205 337 1.2 10.5 1.3435 0.7650 0.3956 0.3229
Effluent
Min.
Max.
617 947 540 0.0 11.4 2.8400 1.2567 0.2900 0.4600
3062 5050 1765 5.2 48.8 6.8233 3.6067 1.7500 1.5600
Means 160 418 0.0 23.1 18.7 2.5293 0.8339 0.7634 0.5459
79 318 0.0 20.4 9.7 1.0295 0.8209 1.4196 1.0348
Removal (%)
Min.
Max.
74 85 0.0 0.0 6.7 1.2167 0.1433 0.0333 0.0001
459 1234 0.0 83.2 42.0 4.6367 1.8400 5.0400 3.3733
90.1 82.1 100.0 e 20.1 50.5 66.7 20.4 41.7
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 e1 7 8
2.
3.
4.
5.
177
chemical nitrogen reactions, were easily detected by monitoring the patterns of the MSC of the parameters. Even when an FNBP appeared in the ORP-time profile and hence real-time control of the aerobic phase using that profile was not possible, the NBP was consistently detected using pH(mV) as a control parameter since no FNBP appeared in the pH(mV)-time profile. Therefore we are convinced that it would be desirable to use pH(mV) rather than ORP as a real-time control parameter for the aerobic phase. However, ORP was more suitable than pH(mV) for the realtime control of the anoxic phase, since the appearance of the NKP in the pH(mV)-time profile was unclear in some cases. The newly designed real-time control strategy using pH (mV)-MSC for the control of the aerobic phase and ORP-MSC for the anoxic phase was very successful, and thus optimization of the duration of each phase was feasible and resulted in flexible HRTs. Through real-time control of the SBR process, a very high nitrogen removal efficiency was achieved (100% NH4-N and 98% soluble N) despite major variations in the N loading rate.
Acknowledgements
Fig. 6 e Removal characteristics of N under real-time control.
Zeng et al. (2009) reported on real-time control of the aerobic duration based on pH “NBP”-inhibited nitrite oxidizing bacteria (NOB) growth. They found that it was possible to prevent the NOB from further oxidizing the accumulated nitrite to nitrate. This resulted in stable performance of the nitrogen removal via nitrite at low DO levels of 0.5e1.0 mg L1 and normal DO levels of 1e2 mg L1. The nitrite accumulation rate was above 95% and the ammonia removal efficiencies were above 97%. The removal efficiency of PO4-P averaged 20% and those of the solids were around 20e50%, and these low removals can probably be attributed to the fact that sludge wastage was not conducted during this study.
4.
Conclusions
An evaluation of a newly designed real-time control strategy using the ORP- and pH(mV)-time profiles was conducted, and the following conclusions are made based on the results obtained: 1. Specific control points appearing in the ORP- and pH(mV)time profiles, which indicate the completion of biological or
This research was funded by ARPC (Agricultural R&D Promotion Center), Korea and supported in part by a grant from the Institute of Animal Resources at Kangwon National University.
references
A.P.H.A, 1995. Standard Methods for the Examination of Water and Wastewater, tenth ed. American Public Health Association, Washington, DC. Akin, B.S., Wgurlu, A., 2005. Monitoring and control of biological nutrient removal in a sequencing batch reactor. Process Biochem. 40, 2873e2878. Bernet, N., Delgenes, N., Akunna, J.C., Delgenes, J.P., Moletta, R., 2000. Combined anaerobiceaerobic SBR for the treatment of piggery wastewater. Water Res. 34, 611e619. Bicudo, J.R., Svobada, I.F., 1995. Intermittent aeration of pig slurry farm scale experiments for carbon and nitrogen removal. Water Sci. Technol. 32, 83e90. Bortone, G., Malaspona, F., Stane, L., Tilche, A., 1994. Biological nitrogen and phosphorus removal in an anaerobic/anoxic sequencing batch reactor with separated biofilm nitrification. Water Sci. Technol. 30, 303e313. Casellas, M., Dagot, C.M., Baudu, M., 2006. Set up and assessment of a control strategy in a SBR in order to enhance nitrogen and phosphorus removal. Process Biochem. 41, 1994e2001. Chen, K.C., Chen, C.Y., Peng, J.W., Houng, J.Y., 2002. Real-time control of an immobilized-cell reactor for wastewater treatment using ORP. Water Res. 36, 230e238. Cheng, J.Y., Liu, B., 2001. Nitrification/denitrification in intermittent aeration process for swine wastewater treatment. J. Environ. Eng. 127, 705e711.
178
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 e1 7 8
Deng, L.W., Zheng, P., Chen, Z., Mahmood, Q., 2008. Improvement in post-treatment of digestion swine wastewater. Bioresour. Technol. 99, 3136e3145. Fernandes, L., Mckyes, E., Warith, M., Barrington, S., 1991. Treatment of liquid swine manure in the sequencing batch reactor under aerobic and anoxic condition. Can. Agr. Eng. 33, 373e379. Ga, C.H., Ra, C.S., 2009. Real-time control of oxic phase using pH (mV)-time profile in swine wastewater treatment. J. Hazard. Mater. 172, 61e67. Gao, D., Peng, Y., Li, B., Liang, H., 2009. Shortcut nitrificationedenitrification by real-time control strategies. Bioresour. Technol. 100, 2298e2300. Ghaly, A.E., Kok, R., 1986. Nitrogen transformation in swine manure in an oxidation ditch under aerobic condition. Trans. ASAE 29, 788e799. Guo, J., Yang, Q., Peng, Y., Yang, A., Wang, S., 2007. Biological nitrogen removal with real-time control using step-feed SBR technology. Enzym. Microb. Tech. 40, 1564e1569. Hao, J.O., Kim, H., 2000. pH and oxidation reduction potential (ORP) strategy for optimization of nutrient removal in an alternating aerobic-anoxic system. Water Environ. Res. 73 (1), 95e102. Kameoka, T., kagi, T., Sakimoto, M., Inno, Y., 1986. Characteristics of concentrated wastewater treatment of swine by the rotating disc system. Jpn. J. Zootechnol. Sci. 57, 209e215. Kim, J.H., Chen, M., Kishida, N., Sudo, R., 2004. Integrated realtime control strategy for nitrogen removal in swine wastewater treatment using sequencing batch reactors. Water Res. 38, 3340e3348. Kim, J.H., Guo, X.J., Park, H.S., 2008. Comparison study of the effects of temperature and free ammonia concentration on nitrification and nitrite accumulation. Process Biochem. 43, 254e260. Kishida, N., Kim, J.H., Chen, M., Sasaki, H., Sudo, R., 2003. Effectiveness of oxidationereduction potential and pH as monitoring and control parameters for nitrogen removal in swine wastewater treatment by sequencing batch reactors. J. Biosci. Bioeng. 96, 285e290. Libelli, S.M., 2006. Control of SBR switching by fuzzy pattern recognition. Water Res. 40, 1095e1107. Ndegwa, P.M., Wang, L., Vaddella, V.K., 2007. Potential strategies for process control and monitoring of stabilization of dairy wastewater in batch aerobic treatment systems. Process Biochem. 42, 1272e1278. Obaja, D., Mace, S., Costa, J., Sans, C., Alvarez, J.M., 2003. Nitrification, denitrification and biological phosphorus removal in piggery wastewater using a sequencing batch reactor. Bioresour. Technol. 87, 103e111. Osada, T., Haga, K., harada, Y., 1991. Removal of nitrogen and phosphorus from swine wastewater by activated sludge units
with the intermittent aeration process. Water Res. 25, 1377e1388. Pan, P., Drapcho, C.M., 2001. Biological anoxic/aerobic treatment of swine waste for reduction of organic carbon, nitrogen and odorous compounds. Trans. ASAE 44, 1789e1796. Pavselj, N.H., Kocijan, N.J., Ro, M., Ubelj, M., Mul, G., Stmnik, S., 2001. Experimental design of an optimal phase duration control strategy used in batch biological wastewater treatment. ISA Trans. 40, 41e56. Peddie, C.C., Mavinic, D.S., Jenkins, C.J., 1990. Use of ORP for monitoring and control of aerobic sludge digestion. J. Environ. Eng. 116, 461e471. Pilsson, S.S., Capdeville, B., Mauret, M., Deguin, A., Baptiste, P., 1996. Real-time control of nitrogen removal using three ORP bending points: signification, control strategy and results. Water Sci. Technol. 33, 275e280. Ra, C.S., Lo, K.V., Mavinic, D.S., 1997. Swine wastewater treatment by a batch-mode 4-stage process: loading rate control using ORP. Environ. Technol. 18, 615e622. Ra, C.S., Lo, K.V., Mavinic, D.S., 1998. Real-time control of two-stage sequencing batch reactor system for the treatment of animal wastewater. Environ. Technol. 19, 343e356. Ra, C.S., Lo, K.V., Mavinic, D.S., 1999. Control of a swine manure treatment process using a specific feature of oxidationereduction potential. Bioresour. Technol. 70, 117e127. Ra, C.S., Lo, K.V., Shin, J.S., Oh, J.S., Hong, B.J., 2000. Biological nutrient removal with internal organic carbon source in piggery wastewater treatment. Water Res. 34, 965e973. Su, J.J., Kung, C.M., Lin, J., Lian, W.C., Wu, J.F., 1997. Utilization of sequencing batch reactor in situ piggery wastewater treatment. J. Environ. Sci. Heal. A 32, 391e405. Tanwar, P., Tapas, N., Pallavi, U., Pravin, M., 2008. Correlating on-line monitoring parameters, pH, DO and ORP with nutrient removal in an intermittent cyclic process bioreactor system. Bioresour. Technol. 99, 7630e7635. Wang, Y., Peng, Y., Stephenson, T., 2009. Effect of influent nutrient ratio and hydraulic retention time (HRT) on simultaneous phosphorus and nitrogen removal in a twosludge sequencing batch reactor process. Bioresour. Technol. 100, 3506e3512. Wareham, D.G., Hall, K.J., Mavinic, D.S., 1993. Real-time control of aerobiceanoxic sludge digestion using ORP. J. Environ. Eng. 119, 120e136. Yu, R.F., Liaw, S.L., Chang, C.N., Lu, H.J., Cheng, W.Y., 1997. Monitoring and control using on-line ORP on the continuousflow activated sludge batch reactor system. Water Sci. Technol. 35, 57e66. Zeng, W., Zhang, Y., Peng, Y.Z., Wang, S.Y., 2009. Control and optimization of nitrifying communities for nitrification from domestic wastewater at room temperatures. Enzym. Microb. Tech. 45, 226e232.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 9 e1 9 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Changes in the Daphnia magna midgut upon ingestion of copper oxide nanoparticles: A transmission electron microscopy study Margit Heinlaan a,c, Anne Kahru a,*, Kaja Kasemets a, Brigitte Arbeille b, Ge´rard Prensier b, Henri-Charles Dubourguier a,c a
Laboratory of Molecular Genetics, National Institute of Chemical Physics and Biophysics, Akadeemia tee 23, 12618 Tallinn, Estonia University of Franc¸ois Rabelais, Microscopy Department, IBISA Electron Microscopy Facility, 10, bd Tonnelle´, 37032, Tours Cedex, France c Estonian University of Life Sciences, Institute of Agricultural and Environmental Sciences, Kreutzwaldi 5, 51014 Tartu, Estonia b
article info
abstract
Article history:
This work is a follow-up of our previous paper (Heinlaan et al., 2008. Chemosphere 71,
Received 26 March 2010
1308e1316) where we showed about 50-fold higher acute toxicity of CuO nanoparticles (NPs)
Received in revised form
compared to bulk CuO to water flea Daphnia magna. In the current work transmission electron
30 June 2010
microscopy (TEM) was used to determine potential time-dependent changes in D. magna midgut
Accepted 14 August 2010
epithelium ultrastructure upon exposure to CuO NPs compared to bulk CuO at their 48 h EC50
Available online 21 August 2010
levels: 4.0 and 175 mg CuO/L, respectively. Special attention was on potential internalization of CuO NPs by midgut epithelial cells. Ingestion of both CuO formulations by daphnids was evident
Keywords:
already after 10 min of exposure. In the midgut lumen CuO NPs were dispersed whereas bulk
Crustaceans
CuO was clumped. By the 48th hour of exposure to CuO NPs (but not to equitoxic concentrations
Nanoparticle uptake
of bulk CuO) the following ultrastructural changes in midgut epithelium of daphnids were
Ultrastructure
observed: protrusion of epithelial cells into the midgut lumen, presence of CuO NPs in circular
Stress
structures analogous to membrane vesicles from holocrine secretion in the midgut lumen.
Metal oxides
Implicit internalization of CuO NPs via D. magna midgut epithelial cells was not evident however CuO NPs were no longer contained within the peritrophic membrane but located between the midgut epithelium microvilli. Interestingly, upon exposure to CuO NPs bacterial colonization of the midgut occurred. Ultrastructural changes in the midgut of D. magna upon exposure to CuO NPs but not to bulk CuO refer to its nanosize-related adverse effects. Time-dependent solubilisation of CuO NPs and bulk CuO in the test medium was quantified by recombinant Cu-sensor bacteria: by the 48th hour of exposure to bulk CuO, the concentration of solubilised copper ions was 0.05 0.01 mg Cu/L that was comparable to the acute EC50 value of Cu-ions to D. magna (48 h CuSO4 EC50 ¼ 0.07 0.01 mg Cu/L). However, in case of CuO NPs, the solubilised Cu-ions 0.01 0.001 mg Cu/L, explained only part of the toxicity. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Compared to many other metal oxide nanoparticles (NPs) (TiO2, ZnO and SiO2), the potential hazardous effects of CuO
NPs are poorly studied (Kahru and Savolainen, 2010) which needs reconsideration since CuO NPs are increasingly used and thus sooner or later end up in natural water bodies. This may cause considerable hazard due to the high toxicity of copper to
* Corresponding author. Tel.: þ372 6 398 373; fax: þ372 6 398 382. E-mail address:
[email protected] (A. Kahru). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.026
180
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 9 e1 9 0
aquatic organisms (Kahru and Dubourguier, 2010). Indeed, various copper compounds have been used as an antifoulants for centuries and extensive research has been performed to understand how copper speciation influences bioavailability and toxicity (Thomas and Brooks, 2010). The free copper ions (Cuþ, Cu2þ) are the most bioavailable and thus toxic to aquatic organisms whereas copper bound to organic matter is widely considered non-bioavailable (Arnold et al., 2005). Concerning copper oxides, for aquatic test organisms CuO NPs are remarkably more toxic than bulk CuO: 51-fold more toxic to Daphnia magna and 48-fold more toxic to bacteria Vibrio fischeri (Heinlaan et al., 2008), 16-fold more toxic to algae Pseudokirchneriella subcapitata (Aruoja et al., 2009) and up to 23fold more toxic to protozoa Tetrahymena thermophila (Mortimer et al., 2010). This remarkably higher toxicity of CuO NPs compared to its bulk formulation has been explained by the increased solubility of CuO nanopowder as according to material safety data sheet information, (MSDS, J.T.Baker, 2008) bulk (microsized) CuO is claimed “insoluble in water”. However, as solubilisation did not fully explain the toxicity of CuO NPs to cell cultures (Karlsson et al., 2008) and yeast Saccharomyces cerevisiae (Kasemets et al., 2009), other mechanisms of toxicity such as the formation of reactive oxygen species (ROS) and DNA damage (Karlsson et al., 2008) by NPs of CuO has to be considered. In addition, ROS formation by CuO NPs may also be caused by solubilised Cu-ions as has been shown by recombinant ROS sensitive sensor bacteria (Ivask et al., in press). In case of particle-ingesting organisms (like crustaceans) ingested CuO NPs may become more bioavailable in the gut. Moreover, due to their small size, large specific surface area and thus high oxidative stress inducing potential that may facilitate the damage of the midgut cell membranes the CuO NPs may be internalised by the gut epithelial cells and release higher damaging concentrations of Cu-ions within the cell. The latter mechanism of toxic action of metallic NPs has been entitled “Trojan horse” mechanism and shown on in vitro cell cultures of human lung epithelial cell by Limbach et al. (2007). Whatever the mechanism of toxicity, direct contact between particles and cell membranes may be a pre-requisite for manifestation of the toxic effect and specially magnification/potentialisation due to the “nano” property. Indeed, several nanoecotoxicological studies have underlined the importance of direct contact between the NP and the cell (Griffitt et al., 2008; Heinlaan et al., 2008; Ku¨hnel et al., 2009; Blinova et al., 2010). Each structure in the organism or cells, e.g., (ultra)structure of the gut epithelium has been developed/ evolutionised to support a certain biological function. Thus, evolution of changes in cellular ultrastructure upon exposure to certain chemical or environmental pollutant may signal on cellular stress that may result in toxic outcome. The changes in ultrastructure of the cells can be followed by transmission electron microscopy (TEM) due to its submicron resolution power. TEM has been successfully used in mechanistic aquatic toxicology already for several decades (Griffiths, 1980; Bodar et al., 1990; Nogueira et al., 2006). In addition, TEM has become a vital technique in nanotoxicology as a tool for i) characterization of NPs as toxic action of synthetic NPs is very much dependent on their size, shape and aggregation and ii) the evaluation of ultrastructural changes in cells and tissues due to the exposure to nanoscale toxicants (Mu¨hlfeld et al., 2007).
The water flea D. magna is considered a keystone species in aquatic toxicology and proposed as a model organism for the ecotoxicological testing of nanomaterials (Baun et al., 2008; Kahru et al., 2008). The aim of the current work was a follow-up to our previous paper where we showed the remarkably higher toxicity of CuO NPs compared to the bulk CuO to D. magna, Thamnocephalus platyurus and V. fischeri. In the current paper we apply transmission electron microscopy (TEM) to evaluate the ultrastructural changes in the D. magna midgut cells that may be related to elevated toxicity of nanosized CuO compared to its bulk size analogue to D. magna by controlling the following hypothesis: does exposure of non-selective particle-ingesting organism D. magna to CuO NPs lead to i) ultrastructural changes of midgut epithelial cells ii) uptake of NPs by the midgut epithelial cells Immobilization of daphnids and solubilised copper in the standard test medium were quantified in parallel. For TEM observations, daphnids were exposed to CuO NPs and to bulk CuO particles from 10 min up to 48 h at the respective 48 h NOEC (0.5 and 50 mg CuO/L) and 48 h EC50 (4.0 and 175 mg CuO/L) values, i.e. at nominal equitoxic concentrations for the both copper oxides. The time-dependent solubilisation of copper oxide particles in the standard test medium with and without daphnids was quantified in parallel, to evaluate the toxic impact of solubilisation.
2.
Materials and methods
2.1.
Test chemicals
CuO NPs were purchased from SigmaeAldrich. The mean particle size as advertised by the manufacturer was e30 nm. Bulk CuO was purchased from Alfa Aesar (Germany). The stock suspensions of copper oxides were prepared immediately prior to the test in the synthetic freshwater (OECD standard test medium, STM) (OECD 202). The STM (pH 7.8 0.2) contained (mg/L MilliQ water): CaCl2$2H2O: 294; MgSO4$7H2O: 123.25; NaHCO3: 64.75; KCl: 5.75. The suspensions of the copper oxides (both nano and bulk formulations) in MilliQ water were characterized by scanning electron microscopy (SEM) (JEOL, JSM-8404). The specific surface area of the powders of both oxides was analysed by Sorptometer Kelvin 1042 (Costech Instruments) at Tallinn University of Technology, Centre of Materials Research and Laboratory of Inorganic Materials. The size distribution of CuO NPs (40 mg CuO/L) was determined by transmission electron microscopy (TEM) in D. magna standard test medium (STM) using a JEOL 1230 at 120 kV. Also, size of the individual particles in the midgut in situ (based on 200 measurements) was measured from TEM micrographs using the freeware ImageJ (NIH, USA).
2.2.
Exposure of D. magna to the test chemicals
The Daphtoxkit F was purchased from MicroBioTests, Inc. (Mariakerke-Gent, Belgium). The crustacean D. magna acute
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 9 e1 9 0
immobilization assays were performed according to the Standard Operational Procedure of Daphtoxkit F (1996) that adheres to OECD 202 guideline (OECD, 2004). Briefly, 5 daphnid neonates per test well (containing 10 ml of standard OECD test medium; STM) in 4 replicates were incubated in the dark at 20 C for up to 48h. Daphnids were exposed to CuO NPs and to bulk CuO particles at their respective NOEC (0.5 and 50 mg CuO/L of STM) and EC50 (4.0 and 175 mg CuO/L of STM) levels according to previous results (Heinlaan et al., 2008). According to the standard test protocol (OECD 202) prior to the testing, the hatched daphnids were fed during 2 h with unicellular alga Spirulina sp. provided also in the Daphtoxkit F. In addition, a control (unexposed daphnids in STM) was included. Quality criteria of the tests were fulfilled as in the control immobilization of daphnids did not exceed 10%. The number of immobilized daphnids (toxicity endpoint) was determined at 10 min, 2, 18, 24, 36 and 48 h instead of only at the end of exposure (at the 48th hour). At these chosen timepoints, also pre-fixation of daphnids was performed: alive and immobilized (presumably dead) daphnids were separately collected and fixed in TRUMP solution (see chapter 2.3). For each time-point, the daphnids were collected from separate wells.
2.3. Sample preparation for transmission electron microscopy (TEM) Examination of the semi-thin sections of daphnids in light microscopy showed that tissues of daphnids started to degrade rapidly upon their death (immobilization). Thus, only the daphnids that were alive (not immobilized) at the moment of fixation were prepared for further TEM examination. As a total, 18 control daphnids (3 per sampling time) were prepared for TEM. For bulk and nano CuO, at least 4 alive daphnids per sampling time were fixed for observation and at least 2 representative preparations per sampling point were studied with TEM. Briefly, at each exposure time, alive daphnids were collected and fixed in the TRUMP fixative (4% paraformaldehyde, 1% glutaraldehyde in 0.15 M phosphate buffer, pH ¼ 7.4) for 2 h (McDowell and Trump, 1976). Samples were stored at þ4 C before further treatment. Daphnids were then washed ten times at room temperature in 0.07 M NaCle0.15 M Sorensen phosphate buffer, followed by post-fixation during 1 h in 1% OsO4 in 0.15 M sodium phosphate buffer at room temperature in the dark. Dehydration was performed at room temperature in graded ethanol (50 , 70 , 90 , 100 ) series and finished by propylene oxide. The samples were embedded in Epon812 and polymerised for 48 h at 60 C. Sectioning of samples was done with an ultramicrotome Ultracut S (Leica) using a diamond knife (Diamant ultra 45 ). Semi-thin sections (0.5e1 mm) for light microscopy (Olympus CX41) were stained with toluidine blue. Ultra-thin sections (70e90 nm) were placed on Cu/Au/Ni Formvar coated 300mesh grids (Oxford Instruments). At least, 5e10 ultra-thin sections of the midgut per animal were stained with uranyl acetate and lead citrate and observed with JEOL 1110 or JEOL 1230 TEM at accelerating voltage of 100 kV and 120 kV, respectively.
2.4.
181
Quantification of dissolved copper
Dissolved copper in D. magna standard test medium (STM) was quantified using two methods: recombinant Cu-sensor bacteria and AAS (atomic absorption spectroscopy). For the analyses, daphnids were incubated with nano and bulk CuO at the respective 48 h NOEC (0.5 mg CuO/L or 50 mg CuO/L, respectively) or 48 h EC50 (4.0 mg CuO/L or 175 mg CuO/L, respectively) levels as described in chapter 2.2. At 2, 18, 24, 36 and 48 h, the number of immobile daphnids in each well was registered, all the daphnids were removed, the remaining suspension was filtered (d ¼ 0.1 mm sterile Minisart filter) and the filtrate was analysed. All the samples were kept in sterile polystyrene tubes at þ4 C and analysed within 2 days from harvesting. Abiotic control, where suspensions of CuO NPs and bulk CuO were incubated without the daphnids was included to determine whether daphnids influenced CuO solubilisation during the testing conditions. For the analysis of Cu-ions solubilised from CuO particles, recombinant Cu-sensor bacteria Escherichia coli MC1061 (pSLcueR/pDNPcopAlux) (Ivask et al., 2009) and constitutively luminescent control strain E. coli MC1061 (pDNlux) (Leedja¨rv et al., 2006) were used. The recombinant Cu-sensor bacteria could be considered specific to Cu-ions as its metal-response elements (transcriptional regulator and its controlled promoter) that originate from resistance system for Cu (cue/cop, both from E. coli chromosome) and reporter system (luminescence genes luxCDABE from Photorhabdus luminescens) were under the control of Cu-inducible response elements (Ivask et al., 2009). The bioluminescence of the sensor strain increases proportionally with the concentration of internalised copper ions and the control strain lacking the Cu-sensing element is incubated in parallel to the sensor strain for taking into account the potential quenching of bacterial luminescence by the turbidity of metal oxide suspensions (even if suspensions were filtered and the resulting filtrates visually transparent) and possible toxic effects of the tested compounds. Briefly, 100 ml of sensor/control bacteria in the analysis medium (0.9% of NaCl, 0.1% of casaminoacids (acid hydrolyzed casein) and 0.1% of glucose) was mixed with 100 ml of sample (on white 96-well microplates; Thermo Labsystem, Finland) and incubated for 2 h in the dark at þ30 C. Luminescence was recorded with Fluoroskan Ascent Luminometer (Thermo Labsystem, Finland). The amount of solubilised Cu-ions was quantified using the CuSO4 calibration curve assuming its 100% bioavailability to the sensor bacteria. The detection limit of the Cu-sensor bacteria was 0.004 mg Cu-ions/L. AAS analysis of <0.1 mm filterable copper takes into account Cu-ions, their complexes as well as CuO NPs. AAS was performed according to standard procedures (AASgraphite furnace; detection limit 0.1 mg Cu/L) in accredited laboratory (EVS-EN ISO/IEC 17025:2005) of the Institute of Chemistry of Tallinn University of Technology, Estonia.
3.
Results and discussion
3.1.
Characterization of copper oxides
Structure of CuO NPs and bulk CuO suspensions in MilliQ water was analysed by SEM (Fig. 1a and b): both suspensions
182
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 9 e1 9 0
contained agglomerates of particles but suspension of bulk form contained considerably larger particles than the suspension of nano CuO. The difference in particle size was confirmed by the BET-analysis (see chapter 2.1.): specific
surface area of CuO NPs was 39-fold bigger than that of the respective bulk form: 25.5 and 0.64 m2/g, respectively. The actual size of CuO NPs in the real exposure conditions, i.e. in the D. magna midgut was measured from the TEM micrographs.
Fig. 1 e Characterization of the nano and bulk CuO: structure and solubilisation. (aeb). Scanning electron microscopy (SEM) images of nano CuO (a) and bulk CuO (b) particles suspended in MilliQ water. (ced). Kinetics of solubilisation of nano CuO (c) and bulk CuO (d) in standard test medium (STM) upon incubation with Daphnia magna up to 48 h. Black bars represent concentration of Cu-ions quantified by Cu-sensor bacteria and grey bars represent <0.1 mm filterable copper concentration analysed by AAS. Immobilization of daphnids (%) is presented on secondary Y-axis. All data are the average of 3 replicates ± standard deviation. The area inbetween the dashed horizontal lines represents the 48 h EC50 value for Cu-ions (Heinlaan et al., 2008).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 9 e1 9 0
183
The average CuO NP size in the midgut of D. magna was 31 12.8 nm, i.e. similar to that advertised by the manufacturer.
3.2. Analysis of the solubilisation of copper oxides: quantification of dissolved copper Time-dependent solubilisation of nano and bulk CuO during 48 h of incubation with and without daphnids at the respective EC50 and NOEC levels was quantified using Cu-sensor bacteria and AAS analysis in parallel (see Materials and methods). The concentrations of dissolved copper were very low: below 0.1 mg/L (Fig. 1c and d). The comparison of the results obtained with Cu-sensor bacteria and chemical analysis showed that in general they were in good agreement (R2 ¼ 0.88). However the AAS values on dissolution of nano CuO suspensions exceeded the values obtained by Cu-sensor bacteria (Fig. 1c). The difference could be explained by the fact that after filtering the nanosuspensions through 100 nm pore-size filters the filtrate contained CuO nanoparticles (<100 nm) that did not induce the bioluminescence of Cu-sensor bacteria but were quantified during AAS analysis. Thus, our results indicate that due to the complexity of separation of nanoparticles from the dissolved copper, the biosensors seem to be more relevant tools for the analysis of Cu-ions than chemical methods.
3.3. The midgut structure of unexposed (control) D. magna The gut of Daphnia is divided into 3 parts: chitin-lined muscular foregut, absorptive midgut and chitin-lined proctodeum or hindgut (Schultz and Kennedy, 1976). Contrarily to the fore- and the hindgut, the midgut features peritrophic membrane (PTM, Chatton, 1920), that is a mesh of chitin, sugars and protein (Avtsyn and Petrova, 1986) secreted by the epithelial cells (Schultz and Kennedy, 1976). PTM protects the gut epithelium and regulates exchange of nutrients and enzymes. It was observable in all control daphnids at all exposure times of the experiment and occurred in several formation stages (Fig. 2a): thinner layer along the microvilli in the initial phase of secretion but more folded and thicker in the gut lumen as previously described by Gaino et al. (1997). PTM of D. magna has been shown to be permeable to inert particles up to 130 nm (Hansen and Peters, 1997/1998). PTM is also likely the reason for the absence of resident bacteria in D. magna gut lumen since it prevents bacterial adhesion on epithelial cells (Schoenberg et al., 1984; Peter and Sommaruga, 2008). Up to 24 h of incubation, a small amount of debris (Fig. 2a) but no bacteria were observed in the midgut lumen of unexposed daphnids. Gut epithelial cells occurred in two different phases (Fig. 2b): electron-dense cells in the absorptive phase and electron-lucent cells in the holocrine phase that according to Schultz and Kennedy (1976) lead to a release of the cellular contents and digestive enzymes into the lumen by cell disruption. Spaces between cells were tight with no significant dilatations analogously to the previous observation by Quaglia et al. (1976). Microvilli were dense and regular (Fig. 2aec).
Fig. 2 e TEM micrographs of the midgut of unexposed (control) Daphnia magna in OECD standard test medium (STM). (a) Incubation time: 2 h. Peritrophic membrane (PTM) in different formation stages. Regular microvilli (MV). Some debris (asterix) present in the gut lumen (L). Scale bar [ 5 mm. (b) Incubation time: 18 h. Holocrine cell (asterix) in the gut epithelium. Nucleus (N), mitochondria (M), regular border of microvilli (MV). Scale bar [ 5 mm. (c) Incubation time: 48 h. Large secondary lysosome (SL), nucleus (N), mitochondria (M), regular border of microvilli (MV) and rough endoplasmatic reticulum (RER). Scale bar [ 1 mm.
184
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 9 e1 9 0
3.4. Effects of bulk CuO particles on the midgut ultrastructure of D. magna Bulk CuO was included in the study to evaluate the sizerelated effects (Fig. 1b and chapter 3.1.). Ingested bulk CuO aggregates were visually observable in the midgut already upon 10 min of exposure (data not shown). TEM analysis
showed that at the 2nd hour of exposure to 175 mg CuO/L (48 h EC50; nominal concentration), the ingested CuO particles in the gut lumen were wrapped in PTM as aggregates (Fig. 3aec and 6b) and single crystals of copper oxide were almost not distinguishable. These clumped bulk CuO particles, partitioned and isolated from the midgut epithelium by PTM were observable
Fig. 3 e Midgut of Daphnia magna, exposed to bulk CuO (48 h EC50 [ 175.0 mg/L). Alive daphnids were sampled for analyses. (a) Light micrograph. Incubation time: 2 h. Clumped bulk CuO (arrow) in the posterior part of the midgut. Holocrine cells (asterix) in the gut epithelium. Scale bar [ 200 mm. (b) TEM micrograph. Incubation time: 2 h. Clumps of bulk CuO (CuO), isolated by peritrophic membrane (PTM). Basal membrane (BM), intercellular spaces (IS). Scale bar [ 5 mm. (c) TEM micrograph. Incubation time: 48 h. Clumped bulk CuO (CuO) isolated by peritrophic membrane (PTM). Microvilli (MV) seem disturbed. Scale bar [ 2 mm. (d) TEM micrograph. Incubation time: 48 h. Clumps of bulk CuO in the gut lumen. Dilatation of intercellular spaces (IS). Scale bar [ 2 mm.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 9 e1 9 0
throughout the experiment for up to 48 h. At 2 h, the midgut epithelium showed no changes compared to the control (Figs. 3b and 2a). Up to 36 h the gut epithelium was comparable to that of 2 h (data not shown). However, at 48 h of exposure extensive disturbance of microvilli was clearly evidenced in all the bulk CuO-exposed samples: they were irregular in shape and diameter (Fig. 3c). In some places, dilatation of intercellular spaces was also observed (Figs. 3a and 6b), refering to the osmoregulatory stress (Bianchini et al., 2004). These ultrastructural changes were observed after the exposure to bulk CuO suspensions at 48 h EC50 (175 mg/L) as well as at 48 h NOEC (50 mg/L) (data not shown) concentrations. No other noticeable time-dependent ultrastructural changes
185
were registered. Quantification of Cu-ions from bulk CuOexposed D. magna culture media (Fig. 1d) showed that although bulk CuO is generally considered insoluble (MSDS, J.T.Baker, 2008), the sensor bacteria analysed Cu-ion concentration at the 48th hour (0.05 0.01 mg Cu/L) was sufficiently high to explain the immobilization of daphnids (48 h CuSO4 EC50 ¼ 0.07 0.01 mg Cu/L). As Cu-ion has been shown to be time- and dose-dependent osmoregulatory toxicant (Bambang et al., 1995; Brooks and Mills, 2003), dilatation of intercellular spaces and the severe disturbance of microvilli could at least partially be the result of disturbed ion transportation caused by Cu-ions released from bulk CuO particles.
Fig. 4 e Midgut of Daphnia magna, exposed to nano CuO (48 h EC50 [ 4.0 mg/L) for 2 h. Alive daphnids were sampled for analyses. (a) Light micrograph. Incubation time: 2 h. Dispersed CuO nanoparticles (arrows) in the gut lumen. Holocrine cells (asterix) in the gut epithelium. Scale bar [ 200 mm. (b) TEM micrograph. Incubation time: 2 h. Dispersed CuO nanoparticles, isolated from the microvilli by the peritrophic membrane (PTM). Scale bar [ 10 mm. (c) TEM micrograph. Incubation time: 2 h. Dispersed CuO nanoparticles in the gut lumen, isolated from the microvilli (MV) by the peritrophic membrane (PTM) delineating peritrophic space (PS). Scale bar [ 1 mm.
186
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 9 e1 9 0
3.5. Effects of CuO nanoparticles on the midgut of D. magna As in the case of bulk CuO, nano CuO particles (exposure concentration of nano CuO was 4.0 mg/L ¼ 48 h EC50; nominal
concentration) were also ingested by daphnids already within 10 min. At the 2nd hour of exposure, the semi- and ultra-thin sections (Fig. 4a and b) showed that the posterior midgut was filled with CuO NPs. Differently from the bulk form (Fig. 3b and c), CuO NPs were dispersed and no large aggregates of nano
Fig. 5 e Midgut of Daphnia magna, exposed to nano CuO (48 h EC50 [ 4.0 mg/L) for 18e48 h. Alive daphnids were sampled for analyses. a) Light micrograph. Incubation time: 18 h. Dispersed CuO nanoparticles and protruded epithelial cells (arrow) in the gut lumen. Scale bar [ 50 mm. b) TEM micrograph. Incubation time: 18 h. Dilatation of intercellular spaces (IS). Lipid-like cytoplasmic inclusions (asterix) observed. Scale bar [ 2 mm. c) TEM micrograph. Incubation time: 18 h. Dispersed CuO nanoparticles (arrow) in the vicinity of microvilli (MV). Scale bar [ 0.5 mm. d) TEM micrograph. Incubation time: 24 h. Protrusion of gut epithelial cells (arrow) and lipid-like cytoplasmic inclusions (asterix) observed. Scale bar [ 10 mm. e) TEM micrograph. Incubation time: 48 h. Cellular debris (asterix) in the gut lumen adsorbing CuO nanoparticles (arrows). Scale bar [ 1 mm.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 9 e1 9 0
CuO were observed in the midgut lumen. At the 2nd hour of exposure there was no direct contact between the NPs and the midgut epithelium microvilli due to the presence of intact PTM (Fig. 4c). At the 18th and 24th hours of exposure, however, protruded midgut epithelial cells (Fig. 5a and d) were observed in the midgut lumen. Such protrusions have also been described in the case of ingestion of cyanobacteria (Nogueira et al., 2006) and of C60-fullerens by D. magna (Yang et al., 2009). In addition, the epithelial cells of the midgut contained abundant lipid-like cytoplasmic inclusions (Fig. 5b and d) and dilatation of intercellular spaces of epithelial cells was observed (Figs. 5b and 6c). No clear PTM was visible and random CuO NPs were noticeable in the lumen in the vicinity of microvilli (Fig. 5c). The 36th hour observations (data not shown) revealed similar effects to that of the 24th hour. At the 48th hour, upon both, sub-lethal (48 h NOEC ¼ 0.5 mg/ L) and half-lethal (48 h EC50 ¼ 4.0 mg/L) exposures, extensive bacterial colonization of the gut (Fig. 6c) was observed. It is important to note that bacterial colonization was not observed in daphnids exposed to the equitoxic value of bulkCuO (48 h EC50 or NOEC) and only a few transient bacteria were noted in the midgut of unexposed daphnids (Fig. 6a) throughout the incubation even though the testing environment was not sterile. Fig. 6a also shows that in unexposed daphnids at 48 h, a large amount of debris was present in the gut lumen, most probably from the turnover of the gut epithelium cells and not the food as the daphnids were not fed during the test (i.e. 48 h). It could be suggested that although the analysed daphnids were alive at the moment of fixation, stress caused by exposure to CuO NPs was manifested by bacterial colonization of the midgut that may be due to immunosuppression and/or by a severe impairment of the synthesis of digestive enzymes by CuO NPs. Indeed, Poynton et al. (2007) reported downregulation of genes coding for proteins similar to b-1,3-glucan binding proteins and lectins involved in the arthropod innate immune response and downregulation of digestive glycanolytic enzymes such as cellulose, endo-b-1,4-mannanase and aamylase as a result to sub-lethal copper exposure. Since we observed no bacteria during exposure of daphnids to the bulk CuO, this could be a specific phenomenon induced by the CuO NPs. By the end (48 h) of CuO NPs exposure, a few single CuO NPs, sorbed to cellular debris (Figs. 5e and 7a) were observed in the midgut lumen, however, the total amount of particles was only a fraction of that observed at the 2nd hour (Fig. 4aec). The CuO NPs were no longer contained within the PTM and had relocated between the microvilli. The brush border was noticeably disturbed (Fig. 7b) however there did not seem to occur “internalization” of CuO NPs by epithelial cells although we observed a few black dots within the microvilli (Fig. 7b) which could theoretically refer to the incident of cellular internalization. Also NPs included in circular structures similar to membrane vesicles were observed in the midgut lumen (Fig. 7c). These structures might have originated from holocrine secretion which releases cell contents as well as membrane debris. Similarly, when D. magna was exposed to gold nanoparticles with a diameter of 17 nm, the NPs were mainly found between the microvilli of the midgut epithelium (Hansen and Peters, 1997/
187
Fig. 6 e TEM micrographs of the midgut of Daphnia magna: exposed versus not exposed organisms. Alive daphnids were sampled for analyses. (a) Unexposed (control) daphnid in OECD standard test medium (STM). Incubation time: 48 h. A few bacteria (arrow) and a lot of debris (asterix) in the gut lumen (L). Scale bar [ 5 mm. (b) D. magna, exposed to bulk CuO (48 h EC50 [ 175.0 mg/L). Incubation time: 48 h. Clumps of bulk CuO (CuO) in the gut lumen (L). Dilatation of intercellular spaces (IS). No bacteria observed. Scale bar [ 5 mm. (c) D. magna, exposed to nano CuO (48 h EC50 [ 4.0 mg/L). Incubation time: 48 h. Single CuO nanoparticles (arrowhead) and a lot of bacteria (arrow) in the gut lumen (L). Dilatation of intercellular spaces (IS). Scale bar [ 5 mm.
188
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 9 e1 9 0
1998). Also, Lovern et al. (2008) showed some gold NPs (17e23 nm) in the vicinity of nanogold-exposed D. magna gut microvilli but the particles still did not appear to be taken up by the microvilli. A number of studies on carbon NPs (Alves de Matos et al., 2009; Petersen et al., 2009; Tervonen et al., 2009) did also not show their uptake by D. magna gut epithelial cells. Rosenkranz et al. (2009) however demonstrated that 20 and 1000 nm polystyrene fluorescent NPs crossed the gut’s epithelial barrier and were relocated into the oil storage droplets of daphnids. Differently from the exposure of daphnids to bulk CuO (Fig. 1d), upon the exposure to the CuO NPs the immobilization of daphnids was much higher than expected by the solubilised Cu-ions (0.01 0.001 mg Cu/L) from CuO NPs at the 48th hour (Fig. 1c). Thus, there were additional toxic effects due to the exposure of daphnids to CuO nanoparticles. As the specific surface area of nano CuO was about 39-fold bigger than that of the bulk CuO (i.e. about equal ratio to the differences in toxicities to D. magna) it is likely that the toxic effect of nano CuO is related to oxidative stress. However, as we also observed dilatation of intercellular spaces (Figs. 5b and 6c), the toxic impact of solubilised Cu-ions (although at a very low level, 0.01 mg Cu/L) may contribute to the overall toxicity via osmoregulatory disturbance. Although information about the mechanism of acute toxicity regarding freshwater crustaceans is scarce (Bianchini et al., 2004), it has been demonstrated that waterborne copper exposure (0.1 mg/L) induced 77% reduction of the sodium uptake in the crustacean amphipod Gammarus pulex due to a reduction in sodium influx by the inhibition of Naþ/Kþ ATPase, primarily arising from interference with eSH groups on Naþ/Kþ ATPase (Brooks and Mills, 2003). Moreover, Sutcliffe (1984) estimated that approximately 11% of the total energy budget of G. pulex was spent on osmoregulation. Hence, even if the reduction in osmoregulatory capacity will not result in death, it is likely that the metabolic costs associated with ion and water regulation considerably increase.
4.
Conclusions
The submicron resolution power of TEM makes it a powerful technique for the nanotoxicological studies. Indeed, it enables observation and characterization of nanosize particles and visualisation of the ultrastructural changes in the tissues. Both are crucial for the mechanistic nanotoxicology.
Acknowledgements Fig. 7 e TEM micrographs of the midgut of Daphnia magna, exposed to nano CuO (48 h NOEC [ 0.5 mg/L). Alive daphnids were sampled for analyses. a) Incubation time: 48 h. CuO nanoparticles (arrows), some being sorbed on cellular debris (asterix). Scale bar [ 1 mm. b) Incubation time: 48 h. CuO nanoparticles (arrows) present inbetween disturbed microvilli (MV). Bacteria (B) are present. Scale bar [ 0.5 mm. c) Incubation time: 48 h. CuO nanoparticles (arrows) present in membrane vesicles-like structures. Scale bar [ 1 mm.
We are deeply grateful for Prof. Henri-Charles Dubourguier, our co-author and mentor of Margit Heinlaan, who initiated this study and contributed a lot to its initial writing and interpretation of the TEM photos but unfortunately did not see that last version of this manuscript as he passed away on March 11, 2010. We thank Prof. Philip S. Rainbow (Natural History Museum, UK) and Dr. Angela Ivask (NICPB, Estonia) for helpful comments. M. Heinlaan was supported by scholarships from Archimedes Foundation (Kristjan Jaak), Estonian Society of Toxicology and Doctoral School of Ecology and
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 9 e1 9 0
Environmental Sciences, Estonia. This research was supported by the Estonian Ministry of Science and Education project SF0690063s08, the Estonian Science Foundation grants ETF6956 and ETF6974 and EU 6th Framework Integrated Project OSIRIS (contract no. GOCE-ET-2007-037017).
references
Alves de Matos, A., Diniz, M.S., Mendonc¸a, E., Peres, I., Silva, L., Brito Correia, J., Picado, A., 2009. Histopathological and ultrastructural observations of Daphnia magna exposed to diamond nanoparticles. In: Abstract of the Joint Congress of the Spanish and Portuguese Microscopy Societies, Segovia, Spain, June 16e19, 2009. Arnold, W.R., Santore, R.C., Cotsifas, J.S., 2005. Predicting copper toxicity in estuarine and marine waters using the biotic ligand model. Marine Pollution Bulletin 50, 1634e1640. Aruoja, V., Dubourguier, H.C., Kasemets, K., Kahru, A., 2009. Toxicity of nanoparticles of CuO, ZnO and TiO2 to microalgae Pseudokirchneriella subcapitata. Science of the Total Environment 407, 1461e1468. Avtsyn, A.P., Petrova, T.P., 1986. Resistance of the alimentary canal of Daphnia magna Straus to enteropathogenic Nag vibrios. Bulletin of Experimental Biology and Medicine 102 (3), 1269e1272. Bambang, Y., Thuet, P., Charmantier-Daures, M., Trilles, J.P., Charmantier, G., 1995. Effect of copper on survival and osmoregulation of various developmental stages of the shrimp Penaeus japonicus bate (crustacea, decapoda). Aquatic Toxicology 33, 125e139. Baun, A., Hartmann, N.B., Grieger, K., 2008. Ecotoxicity of engineered nanoparticles to aquatic invertebrates: a brief review and recommendations for future toxicity testing. Ecotoxicology 17, 387e395. Bianchini, A., Martins, S.E.G., Barcarolli, I.F., 2004. Mechanism of acute copper toxicity in euryhaline crustaceans: implications for the Biotic Ligand Model. International Congress Series 1275, 189e194. Blinova, I., Ivask, A., Heinlaan, M., Mortimer, M., Kahru, A., 2010. Ecotoxicity of nanoparticles of CuO and ZnO in natural water. Environmental Pollution 158 (1), 41e47. Bodar, C.W.M., van Donselaar, E.G., Herwing, H.J., 1990. Cytopathological investigations of digestive tract and storage cells in Daphnia magna exposed to cadmium and tributyltin. Aquatic Toxicology 17, 325e338. Brooks, S.J., Mills, C.L., 2003. The effect of copper on osmoregulation in the freshwater amphipod Gammarus pulex. Comparative Biochemistry and Physiology, Part A 135, 527e537. Chatton, E., 1920. Les membranes pe´ritrophiques des Drosophiles (Dipte`res) et des Daphnies (Cladoce`res); leur gene`se et leur role a’le´gard des parasites intestinaux. Bulletin de la Socie´te´ Zoologique de France 45, 265e280. Daphtoxkit FTMmagna, 1996. Crustacean Toxicity Screening Test for Freshwater, Standard Operational Procedure, Creasel, Deinze, Belgium, p. 16. Gaino, E., Rebora, M., Taddei, A.R., Mazzini, M., 1997. Ultrastructural aspects of the alimentary canal in some mayflies. BiologyeEcologyeSystematics. In: Landolt, P., Sartori, M. (Eds.), Ephemeroptera & Plecoptera. Mauron þ Tinguely & Lachat, SA, Fribourg, pp. 332e337. Griffiths, P.R.E., 1980. Morphological and ultrastructural effects of sublethal cadmium poisoning on Daphnia. Environmental Research 22, 277e284. Griffitt, R.J., Luo, J., Gao, J., Bonzongo, J.-C., Barber, D.S., 2008. Effects of particle composition and species on toxicity of
189
metallic nanomaterials in aquatic organisms. Environmental Toxicology and Chemistry 27 (8), 1972e1978. Hansen, U., Peters, W., 1997/1998. Structure and permeability of the peritrophic membranes of some small crustaceans. Zoologischer Anzeiger eA Journal of Comparative Zoology 236, 103e108. Heinlaan, M., Ivask, A., Blinova, I., Dubourguier, H.-C., Kahru, A., 2008. Toxicity of nanosized and bulk ZnO, CuO and TiO2 to bacteria Vibrio fischeri and crustaceans Daphnia magna and Thamnocephalus platyurus. Chemosphere 71 (7), 1308e1316. Ivask, A., Bondareno, O., Jepihhina, N., Kahru, A. Profiling of the reactive oxygen species related ecotoxicity of CuO, ZnO, TiO2, silver and fullerene nanoparticles using a set of recombinant luminescent Escherichia coli strains: differentiating the impact of particles and solubilised metals. Analytical and Bioanalytical Chemistry, in press, doi:10.1007/s00216-010-3962-7. Ivask, A., Ro˜lova, T., Kahru, A., 2009. A suite of recombinant luminescent bacterial strains for the quantification of bioavailable heavy metals and toxicity testing. BMC Biotechnology 9, 41. doi:10.1186/1472-6750-9-41. Kahru, A., Dubourguier, H.-C., 2010. From ecotoxicology to nanoecotoxicology. Toxicology 269 (2e3), 105e119. Kahru, A., Dubourguier, H.-C., Blinova, I., Ivask, A., Kasemets, K., 2008. Biotests and biosensors for ecotoxicology of metal oxide nanoparticles: a minireview. Sensors 8, 5156e5170. Kahru, A., Savolainen, K., 2010. Potential hazard of nanoparticles: from properties to biological and environmental effects. Toxicology 269 (2e3), 89e91. Karlsson, H.K., Cronholm, P., Gustafsson, J., Mo¨ller, L., 2008. Copper oxide nanoparticles are highly toxic: a comparison between metal oxide nanoparticles and carbon nanotubes. Chemical Research in Toxicology 21 (9), 1726e1732. Kasemets, K., Ivask, A., Dubourguier, H.-C., Kahru, A., 2009. Toxicity of nanoparticles of ZnO, CuO and TiO2 to yeast Saccharomyces cerevisiae. Toxicology In Vitro 23 (6), 1116e1122. Ku¨hnel, D., Busch, W., Meißner, T., Springer, A., Potthoff, A., Richter, V., Gelinsky, M., Scholz, S., Schirmer, K., 2009. Agglomeration of tungsten carbide nanoparticles in exposure medium does not prevent uptake and toxicity toward a rainbow trout gill cell line. Aquatic Toxicology 93, 91e99. Leedja¨rv, A., Ivask, A., Virta, M., Kahru, A., 2006. Analysis of bioavailable phenols from natural samples by recombinant luminescent bacterial sensors. Chemosphere 64, 1910e1919. Limbach, L.K., Wick, P., Manser, P., Grass, R.N., Bruinink, A., Stark, W.J., 2007. Exposure of engineered nanoparticles to human lung epithelial cells: influence of chemical composition and catalytic activity on oxidative stress. Environmental Science and Technology 41, 4158e4163. Lovern, S.B., Owen, H.A., Klaper, R., 2008. Electron microscopy of gold nanoparticle intake in the gut of Daphnia magna. Nanotoxicology 2 (1), 43e48. Material Safety Data Sheet: Cupric Oxide (CAS No.: 1317-38-0). Available at, 2008. http://www.jtbaker.com/msds/ englishhtml/C5885.htm (Last accessed 29.06.10). McDowell, E.M., Trump, B.J., 1976. Histological fixatives suitable for diagnostic light and electron microscopy. Archives of Pathology and Laboratory Medicine 100, 405e414. Mortimer, M., Kasemets, K., Kahru, A., 2010. Toxicity of ZnO and CuO nanoparticles to ciliated protozoa Tetrahymena thermophila. Toxicology 269 (2e3), 182e189. Mu¨hlfeld, C., Rothen-Rutishauser, B., Vanhecke, D., Blank, F., Gehr, P., Ochs, M., 2007. Visualization and quantitative analysis of nanoparticles in the respiratory tract by transmission electron microscopy. Particle and Fibre Toxicology 4, 11. doi:10.1186/1743-8977-4-11. Nogueira, I.C.G., Lobo-da-Cunha, A., Vasconcelos, V.M., 2006. Effects of Cylindrospermopsis raciborskii and Aphanizomenon ovalisporum (cyanobacteria) ingestion on Daphnia magna
190
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 9 e1 9 0
midgut and associated diverticula epithelium. Aquatic Toxicology 80 (2), 194e203. OECD, 2004. OECD Guidelines for the Testing of Chemicals/ Section 2: Effects on Biotic Systems, Test No. 202: Daphnia sp. Acute Immobilisation Test. OECD Publishing. Peter, H., Sommaruga, R., 2008. An evaluation of methods to study the gut bacterial community composition of freshwater zooplankton. Journal of Plankton Research 30 (9), 997e1006. Petersen, E.J., Akkanen, J., Kukkonen, J.V.K., Weber Jr., W.J., 2009. Biological uptake and depuration of carbon nanotubes by Daphnia magna. Environmental Science and Technology 43 (8), 2969e2975. Poynton, H.C., Varshavsky, J.R., Chang, B., Cavigiolio, G., Chan, S., Holman, P.S., Loguinov, A.V., Bauer, D.J., Komachi, K., Theil, E.C., Perkins, E.J., Hughes, O., Vulpe, C.D., 2007. Daphnia magna ecotoxicogenomics provides mechanistic insights into metal toxicity. Environmental Science and Technology 41, 1044e1050. Quaglia, A., Sabelli, B., Villiani, L., 1976. Studies on the intestine of Daphnidae (Crustacea, Cladocera). Ultrastructure of the midgut of Daphnia magna and Daphnia obtusa. Journal of Morphology 150, 711e726. Rosenkranz, P., Chaudhry, Q., Stone, V., Fernandes, T.F., 2009. A comparison of nanoparticle and fine particle uptake by
Daphnia magna. Environmental Toxicology and Chemistry 28 (10), 2142e2149. Schoenberg, S.A., Maccubbin, A.E., Hodson, R.E., 1984. Cellulose digestion by freshwater microrustacea. Limnology and Oceanography 29 (5), 1132e1136. Schultz, T.W., Kennedy, J.R., 1976. The fine structure of the digestive system of Daphnia pulex (crustacea: Cladocera). Tissue and Cell 3, 479e490. Sutcliffe, D.W., 1984. Quantitative aspects of oxygen uptake by Gammarus (Crustacea, Amphipoda): a critical review. Freshwater Biology 14 (5), 443e489. Tervonen, K., Waissi, G., Petersen, E.J., Akkanen, J., Kukkonen, J.V.K., 2009. Analysis of fullerene-C60 and kinetic measurements for its accumulation and depuration in Daphnia magna. Environmental Toxicology and Chemistry 29 (5), 1072e1078. Thomas, K.V., Brooks, S., 2010. The environmental fate and effects of antifouling paint biocides. Biofouling 26 (1), 73e88. Yang, X.Y., Edelmann, R.E., Oris, J.T., 2009. Suspended C60 nanoparticles protect against short-term UV and fluoranthene photo-induced toxicity, but cause long term cellular damage in Daphnia magna. Aquatic Toxicology. doi:10.1016/j.aquatox. 2009.08.011.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 9 1 e2 0 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Fate of aniline and sulfanilic acid in UASB bioreactors under denitrifying conditions Raquel Pereira, Luciana Pereira*, Frank P. van der Zee, M. Madalena Alves IBB-Instituto Biotecnologia e Bioengenharia, Centro Engenharia Biolo´gica, Universidade do Minho, Campus de Gualtar, 4710-057 Braga, Portugal
article info
abstract
Article history:
Two upflow anaerobic sludge blanket (UASB) reactors were operated to investigate the fate
Received 31 March 2010
of aromatic amines under denitrifying conditions. The feed consisted of
Received in revised form
wastewater containing aniline and/or sulfanilic acid and a mixture of volatile fatty acids
22 June 2010
(VFA) as the primary electron donors. Reactor 1 (R1) contained a stoichiometric concen-
Accepted 15 August 2010
tration of nitrate and Reactor 2 (R2) a stoichiometric nitrate and nitrite mixture as terminal
Available online 25 August 2010
electron acceptors. The R1 results demonstrated that aniline could be degraded under
synthetic
denitrifying conditions while sulfanilic acid remains. The presence of nitrite in the influent Keywords:
of R2, caused a chemical reaction that led to immediate disappearance of both aromatic
Biodegradation
amines and the formation of an intense yellow coloured solution. HPLC analysis of the
Aromatic amines
influent solution, revealed the emergence of three product peaks: the major one at
UASB bioreactors
retention time (Rt) 14.3 min and two minor at Rt 17.2 and 21.5 min. In the effluent, the
Denitrification
intensity of the peaks at Rt 14.3 and 17.2 min was very low and of that at Rt 21.5 min increased (w3-fold). Based on the mass spectrometry analysis, we propose the structures of some possible products, mainly azo compounds. Denitrification activity tests suggest that biomass needed to adapt to the new coloured compounds, but after a 3 days lag phase, activity is recovered and the final (N2 þ N2O) is even higher than that of the control. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Aromatic amines are important industrial chemicals as their major sources in the environment include several chemical industry sectors such as oil refining, synthetic polymers, dyes, adhesives and rubbers, pharmaceuticals, pesticides and explosives (Pollution inventory, England and Wales, The environment Agency, 2003). They range from simplest aniline to highly complex molecules with conjugated aromatic or heterocyclic structures and multiple substituents. Aromatic amines are commonly generated during the biodegradation of azo dyes by microorganisms under anaerobic conditions, resulting from the reductive cleavage of azo bonds (eNaNe) (Pinheiro et al., 2004; van der Zee and Villaverde, 2005). As they
are difficult to be removed via traditional wastewater treatment and inevitably tend to persist, the potential toxicity of these compounds should be considered in the treatment process (Bor-Yann et al., 2009). Because many different types of sulfonated azo dyes are currently be utilized, a wide variety of sulfonated aromatic amines will be formed under anaerobic conditions that will not easily be biodegraded and will constitute an important part of untreated COD fraction in azo dyes containing wastewater treatment. Aerobic biodegradation of many aromatic amines has been extensively studied (Brown and Laboureur, 1983; Pinheiro et al., 2004; van der Zee and Villaverde, 2005); however, this may not apply to all aromatic amines. It has been demonstrated that especially sulfonated aromatic amines are often difficult to degrade
* Corresponding author. Tel.: þ351 253 604 420. E-mail address:
[email protected] (L. Pereira). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.027
192
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 9 1 e2 0 0
(Razo-Flores et al., 1996; Tan and Field, 2005; Tan et al., 2005). Aromatic amines are commonly not degraded under anaerobic conditions. Among the many different aromatic amines tested, only a few were degraded. Some of them, substituted with hydroxyl or carboxyl group, were degraded under methanogenic and sulphate reducing conditions (Kalyuzhnyi and Skyler, 2000; Razo-Flores et al., 1999b). A drawback of using aerobic treatment, with the aim of degrading aromatic amines from azo dye cleavage, is that many of them are prone to autoxidation once they are exposed to oxygen. Since autoxidation often involves enlargement of the molecules, their biodegradability may consequently be decreased. Alternatively, nitrate, instead of oxygen, can be used as electron acceptor. Indeed, several ecosystems are characterized by lack of oxygen, such as aquatic sediments, stratified lakes, wetlands and some soil horizons. In those environments, microorganisms can utilize compounds like nitrate, iron, sulphate, manganese and carbonate as electron acceptors. It has been reported that at least some aromatic amines can be degraded coupled to nitrate reduction (Hyung-Yell et al., 2000; Wu et al., 2007; Va´zquez-Rodrı´guez et al., 2008). Moreover, it has been observed in previous research at our laboratory that the presence of nitrate does not lead to autoxidation of reduced azo dyes. A further interesting feature of using nitrate is that the first step of denitrification yields nitrite, a compound that has been found to react with aromatic amines, resulting in deamination, thereby yielding aromatics with a higher biodegradation potential (Seymor et al., 2002). Considering the biodegradability of azo dyes in the environment, it is clear that the fate of aromatic amines under denitrifying conditions is of utmost importance. In this work two UASB bioreactors are operated under denitrifying conditions with nitrate (R1) and a mixture of nitrate and nitrite (R2) as electron acceptors, and the fate of aniline and/or sulfanilic acid is described.
2.
Materials and methods
2.1.
Chemicals
2.2.
Mineral medium
The basal medium contained (mg L1): NH4Cl (4750), KH2PO4 (1300), CaCO3 (270), MgSO4$7H2O (500), FeCl2$4H2O (2000), H3BO3 (50), ZnCl2 (50), CuCL2$2H2O (38), MnCl2$2H2O (409), (NH4)6Mo7O24$4H2O (50), AlCl3 (49), CoCl2$6H2O (2000), NiCl2$6H2O (92), Na2SeO3r5H2O (164), EDTA (1000) and HCl 37% (1 mL L1).
2.3.
Experimental set-up
The experimental installation is schematized in Fig. 2. The two UASB reactors have a diameter and height of 2 and 83 cm, respectively. Reactors were filled with a 5.09 0.21 gVSS L1 of granular biomass, with a useful volume of 0.28 L (Table 1). One dual-channel peristaltic pump was used to initially feed the reactors with a constant flow (0.28 L d1). The influent originated from two 5 L containers, was stored at 4 C during the entire process. The reactor’s recycle flows were made by a second dual-channel peristaltic pump at a flow rate of 9.6 L d1.
2.4.
Aromatic amines biodegradation
The operation conditions of the two lab-scale UASB, used to study the biodegradation of aniline and sulfanilic acid under different redox conditions, are summarized in Table 1. Both reactors were fed with the same synthetic wastewater containing nutrients and yeast extract (6 g L1), aniline and/or
The aromatic amines, Aniline and Sulfanilic Acid (SA), were purchased from Sigma Aldrich at the highest analytic grade purity commercially available (99%); chemical structures are represented in Fig. 1. Sodium nitrate and sodium nitrite were purchased from Riedel-de-Hae¨n and the chemicals used to prepare the macronutrients solution were purchased from Sigma, Fulka and Panreac. Ammonium acetate and Methanol (MeOH, 99.9%) for HPLC analysis were obtained from Sigma and Fisher Scientific, respectively.
NH2
Aniline
HO3S
NH2
Sulfanilic acid (SA)
Fig. 1 e Chemical structure of the aromatic amines (Aniline and Sulfanilic acid).
Fig. 2 e Schematic diagram of the experimental set-up. SI1 and SI2; SE1 and SE2 e sample withdraw from the influent and effluent of reactors 1 and 2, respectively.
193
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 9 1 e2 0 0
Table 1 e Operation conditions of the UASB reactors.
Volume reactor (L) Upflow velocity (m h1) Flow rate (L d1) Aromatic Amines (mgCOD L1) VFA (mgCOD L1) Total COD (mg L1) Organic loading rate (mgCOD L1 d1) HRT (h) Nitrate (mM) Nitrite (mM)
Reactor 1
Reactor 2
0.28 1.00 0.28 100 200 300 300 24 7.5 e
0.28 1.00 0.28 100 200 300 300 24 6.0 2.5
sulfanilic acid (0.22 mM, corresponding to 50 mgCOD mL1) and a mixture of volatile fatty acids (VFA: 200 mgCOD L1 at 1:1:1 COD bases ratio of acetate, propionate and butyrate) as the primary electron donors. The aimed total COD in the influents of both reactors was 300 mgCOD L1. According to the stoichiometric mass balances of the organic matter conversion under nitrification conditions, the total oxidation of this amount of COD requires 7.5 mM nitrate or 12.5 mM of nitrite. The conditions of both reactors were similar, except that the influent of Reactor 1 was prepared with a stoichiometric concentration of nitrate (7.5 mM) and of the Reactor 2 with a stoichiometric nitrate/nitrite mixture (6 and 2.5 mM respectively) rather than with nitrate solely. Nitrite was at lower concentration in reactor 2, for reasons of nitrite toxicity. The different phases of reactor operation (Table 2) reflect the adjustments of three parameters: influent pH, influent VFA concentration and the nature of aromatic amines added. The influent pH of phase I was set to 7, but in the following phases was lowered to pH 4.8 to prevent that the value inside the reactors rises beyond the upper value of the optimum range (pH 7e9) due to the denitrifying process. In phase IIb, the VFA concentration in the influent was increased in response to the relatively high nitrate/nitrite concentration in the reactor effluent during the phases I and IIa. The influent of phase III contained only SA, in order to study the fate of this compound individually. At last, phase IV was characterized by reactors operation in batch mode in the same conditions as of phase III. During the reactor operation, samples of 1 mL were withdrawn from the influent and effluent every two days and filtered by Acrodisc 0.20 mm. Amines concentration was monitored by high performance liquid chromatography (HPLC) and nitrite and nitrate by Ion-Chromatography (IC). COD was measured adapting the standard method SM5220C
(APHA, 1989) with the modification of measuring the absorbance at 600 nm instead of titulation. The pH was daily monitored in an Orion-Model 720 A pH meter.
2.5.
HPLC analysis
HPLC analyses were performed in an HPLC (JASCO AS-2057 Plus) using a reverse phase Nucleosil MNC18 (300 mm 4.6 mm, 5 mM ˚ from Machenerey-Nagel, particle size and pore of 100 A Switzerland) column. The following solvent systems were used as mobile phase: solvent A (MeOH), solvent B (Sodium phosphate buffer, pH 7). Compounds were eluted at a flow rate of 0.8 mL min1 and at room temperature, with a linear gradient of mobile phase from 10% to 80% of solvent A, over 45 min. Aniline elution was monitored at 230 nm, sulfanilic acid at 248 nm and coloured products from R2 at 350 nm. Identification of aromatic amines in the influent and effluent of reactors was achieved by comparing the retention times (Rt) with respective standards (Rt at 3.6 min and 14.6 min for SA and Aniline, respectively).
2.6.
Ion Chromatography analysis
Nitrate and nitrite were monitored using IC-DIONEX with a manual injector with a 25 mL loop, a column DIONEX ION PAC AS4A (4 mm 225 mm) and an acquisition and treatment of data VarianWS-Worstation program. The regenerated solution used was H2SO4 (25 mM). Compounds were eluted with a mixture composed of 1.80 mM Na2CO3 and 1.70 mM NaHCO3 and at a flow rate of 1.5 mL min1; the pressure was 49.21 Kg cm3. Concentration of nitrate and nitrite were calculated by the previous determined calibration curves: C ¼ 7.58 e4 AU and C ¼ 6.51 e4 AU, respectively; where C is the concentration (mM) and AU the chromatogram area in arbitrary units.
2.7.
ESI-ion traps analysis
The ESI-ion trap MS system was an LCQ ion trap mass spectrometer (Thermofinnigan) equipped with electrospray source and run by Xcalibur (Thermofinnigan) version 1.3 software. The following conditions were used in experiments with ESI source in positive and negative mode: temperature of the heated capillary of 350 C and source voltage of 4.5 kV. Nitrogen was used as sheath gas and auxiliary gas. The sheath and auxiliary gas flow rates were 80 and 20 arbitrary units, respectively. MS was performed in the full scan mode from m/ z 50 to 1100.
Table 2 e Characterization of the different phases of UASB reactors operation. Phase
I IIa IIb IIc III IV
Days
0e15 15e34 34e48 48e55 55e90 90e100
pHinfluent
VFA (mgCOD L1)
7.0 200 4.8 200 4.8 300 4.8 400 4.8 300 Operation in batch mode
Aromatic Amines (mM) Aniline
SA
0.22 0.22 0.22 0.22 e
0.22 0.22 0.22 0.22 0.44
COD Total (mgCOD L1)
300 300 400 500 400
194
2.8.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 9 1 e2 0 0
temperatures of the injection port and detector were 110 C and of the column was 35 C. The minimum detection limit was 0.1 mM. The removal percentage of nitrate/nitrite reduction products is expressed as the ratio (%) between the N2 þ N2O produced and NO2 3 þ NO2 reduced as the equation: % N-recovery ¼ (N2 þ N2O)produced/(NO2 3 þ NO2 )reduced 100.
Batch experiments
Activity measurements were supported by batch experiments using 70 mL glass serum bottles with a liquid volume of 60 mL and a headspace volume of 10 mL. Assays were conducted in triplicate. Tests included series containing the influents and effluents of reactors, a substrate-free control and a control with acetate only. The principle was using nitrate as electron acceptor and acetate as electron donor, at concentrations of 14 mM and 8.75 mM respectively. Vials were supplemented with 5 mL denitrifying sludge (5.09 0.21 gVSS L1), closed with crimp caps and the headspace was flushed with 100% He. A pressure transducer was used for measuring the pressure every 90 min. The assay was finished when the pressure remained stable. Gas (N2, N2O, CO2 and CH4) formation was also measured at the end of the test in a Pye Unicam GC-TCD gas chromatograph (Cambridge, England), using a Porapack Q (100e180 mesh) column. Helium was used as carrier gas (30 mL min1) and the
Aromatic amines (mM)
A
I
IIa
3.
Results and discussion
3.1.
Aromatic amine degradation in Reactor 1
In Fig. 3A are represented the results of Aniline and SA concentration in the influent and effluent of reactor 1, during the six process phases. Aniline removal was high in all the operation phases, although a slight increase was obtained with increasing nitrate consumption. Indeed, the concentrations of nitrate in the effluent at phases I and IIb were higher than the expected, 3.93 0.70 mM and 3.77 0.57 mM,
IIb
IIc
40
50
III
IV
0.5 0.4 0.3 0.2 0.1 8
NO3(mM)
B
6 4 2 9
pH
C
8 7 6 5
COD Removal (%)
D
100 80 60 40 20 0 0
10
20
30
60
70
80
90
100
Time (d) Fig. 3 e (A) Aromatic amines concentration: (B, C) Sulfanilic acid; (D, :) Aniline; (B) Nitrate consumption; (C) pH monitorization and (D) Percentage of COD removal. White markers correspond to the results from the influent and the black to the results from the effluent.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 9 1 e2 0 0
respectively (Fig. 3B), probably due to a substrate lack. Aniline removal was 80% in phase I and 88% in phase IIa. In order to increase nitrate consumption, VFA in the influent was increased from 200 to 300 mgCOD L1 (phase IIb). Consequently, nitrate concentration in the effluent dropped to 2.03 0.44 mM, corresponding to w54% of consumption. At that phase, the aniline removal also increased, to w95%. In phase IIc, with 400 mgCOD L1 VFA, the nitrate and aniline removal was almost complete. No SA removal was found in any of the operating reactor phases. An increase of effluent pH to 8.5 was observed in phase I (Fig. 3C), presumably due to denitrification which is a pH-increasing process (Rust et al., 2000; Karim and Gupta, 2003). Once values higher than the optimum (pH 7e8) can be harmful to the bacteria, the pH of the influent solution was lowered to 4.8 in phase IIa and onward. From then, the pH values of the effluent stabilized below pH 8. The removal of COD in reactor 1 was around 60e70% in all the phases, with a sharp decrease to 40% just at the beginning of the phase III (Fig. 3D). The application of high-rate reactors, such as the UASB reactors, has been proved to be capable of treating various wastewaters bearing toxic aromatic compounds with a high degree of efficiency and stability (Donlon et al., 1996; RazoFlores et al., 1997, 1999a,b; Karim and Gupta, 2003). The first report of a pure culture capable of aniline dissimilation belongs to Schnell and Schink (1991); they have demonstrated that Desulfobacterium aniline was able to catabolize aniline via reductive deamination of 4-aminobenzoyl-CoA, in a process which was linked to sulphate reduction. Since then, many other studies about aniline degradation have been reported. De et al. (1994) have found that aniline is able to be degraded markedly in estuarine sediments under denitrification conditions; Hyung-Yell et al. (2000) demonstrated that the strain HY99, a novel microorganism capable of aerobic and anaerobic degradation of aniline, could catabolize aniline under anaerobic conditions linked with nitrate reduction. More recently, Wu et al. (2007) showed the possibility of aniline degradation by microbes in riverbed sediments under denitrifying conditions. As in our work, they have also observed that aniline biodegradation was enhanced by the nitrate consumption. Aniline showed readily biodegradability in denitrifying environments, 60% in a short time period (<28 days), when tested as a model compound to develop a method for measuring the anoxic biodegradability under denitrifying conditions (Va´zquez-Rodrı´guez et al., 2008). Concerning the SA biodegradation less information is available, but it is known that the sulfonic groups in the molecule difficult its degradation; this is due to the, high solubility conferred by those groups preventing the penetration of the aromatic amine through bacterial membranes and, depending on their position relatively to the azo bound, they also confer steric hindrance (Perei et al., 2001; Carvalho et al., 2008). In fact, some reports on aromatic amines degradation have demonstrated that especially sulfonated aromatic amines are often difficult to degrade (Tan et al., 1999, 2000). More recently these authors have found that none of the several sulfonated amines tested was degraded under anaerobic conditions and, under aerobic, only two of them, 2- and 4-aminobenzenesulfonic acid, were mineralized but with inocula sources that were historically polluted with sulfonated
195
aromatic amines (Tan et al., 2005). Successfully cases on SA degradation are those reported by Perei et al. (2001), whom have isolated, from a contaminated site, the aerobic bacterium Pseudomonas paucimobilis, able to degraded it; and by Coughlin et al. (2002), demonstrating the biodegradation of SA by an aerobic bacterial co-culture, during the degradation of the azo dye Acid Orange 7. A recent study about the efficient biodegradation of both aniline and SA is that of Carvalho et al. (2008), the aromatic amines were easily degraded by three different types of aerobic inocula: domestic wastewater and activated sludge from municipal and industrial treatment plant; although, a high lag phase was required for SA, corroborating the idea that SA biodegradation has high specificity.
3.2.
Aromatic amine degradation in Reactor 2
Curiously, the influent solution of reactor 2, containing nitrite, turned yellow and colour development increased with time, even stored at 4 C. The UVevis spectra at all the phases showed a band with lmax at 350 nm (data not shown). HPLC analysis of the influent solution, revealed a decrease of aniline and SA in the first phase of 45% and 27%, respectively. In the following phases, aniline in the influent was totally consumed and SA was present at very low concentration (0.02 0.007 and 0.06 0.003 in phase II and III, respectively). In HPLC chromatograms at 350 nm, three resulting peaks were observed in phase II, the major one at Rt 14.3 min and two minor at Rt 17.2 and 21.5 min (Fig. 4A). In the phase I only the peak at Rt 21.5 min appeared. That difference may be explained by the different pH in this phase. Indeed, colour development of influent solution was more pronounced after changing the pH from 7 to 4.8. Finally, in phase III (absence of aniline), no colour products were obtained. A chemical reaction of single and/or both aromatic amines may occur in the presence of nitrite, justifying the vanishing of the aromatic amines from the influent, before entering in the reactor, and the colour formation. Moreover, once no such reaction was observed in the influent of Reactor 1 (prepared without nitrite) it is clear that the nitrite was responsible for the reaction. Although the SA is also consumed, the fact that no colour formation occurs in its absence leads us to conclude that the presence of aniline is crucial for the chemical reaction. Samples from the effluent were also analyzed by UVevis and HPLC. A decrease of the lmax at 350 nm was observed and the solution was lighter in colour. In HPLC chromatograms, the peaks at Rt 14.3 and 17.2 min almost disappeared and the one at Rt 21.5 min increased w3-fold (Fig. 4B). The production of yellow/orange compounds in the growth media of isolated aniline cultures, during the disappearance of aniline under nitrate-reducing conditions, was also observed by O’neill et al. (2000). Although any attempt to detect the products formed was done, they attributed the colour to the probably formation of azo dye by two ways: 1) partial oxidation of one amino group to a nitroso group followed by condensation with an un-oxidised amine and/or 2) formation of a benzene diazonium salt produced by the reaction of one molecule of aniline with nitrite (obtained from the reduction of nitrate) and its coupling with other molecule of aniline in the para position to give p-aminoazobenzene.
196
0.12
0.08
B 0.12 ■
AU at 350 nm
A AU at 350 nm
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 9 1 e2 0 0
■■ 0.04
●
▲ ▲
●
III
●
▲
II c
0.08
10
15
25
III II c II b
■ ▲
II a
20
●
■ ●▲ ■ ▲
0.04
II
5
IV
●
II b
●
0 0
● ●
II a
0
30
I
0
5
Retention time (min)
10
15
20
25
30
Retention time (min)
Fig. 4 e HPLC results of the influent (A) and effluent (B) in all the phases of reactor 2 operation. (-) Rt at 14.2 min; (:) Rt at 17.3 min and (C) Rt at 21.5 min.
The COD removal in reactor 2 was lower than in the reactor 1, w40%.
MS analysis was done. The mass spectrum in negative mode exhibits peaks with m/z 197, 247, 277, 322 and 387 and, in positive ionization, the peaks at m/z 123, 219, 251, 284, 342, 393, 409, 433, 482, 526, 570, 664, 686 and 764 (Fig. 5). The structures for some of the possible products are proposed (Table 3). The ion m/z 123 may correspond to nitrobenzene. The compound with mass of 218 g mol1 may result from the electrophilic aromatic substitution in the molecule of AS: an
3.3. Identification of the products formed in the influent of Reactor 2 In an attempt to identify the products of the chemical reaction of aniline and sulfanilic acid with nitrite, in the influent of R2,
277.00
100
ESI -
80
60
Relative Intensity (%)
40
322.00
20 197.08
0 100
387.33
247.42
200
300
400
600
500
700
800
900
ESI +
219.25
100
1000
251.25
80
60 409.42 284.42
40
685.67
432.50 393.42 481.50 342.42
764.33
525.58
20
663.67 569.58
122.92
0 100
200
300
400
500
600
700
800
900
1000
m/z Fig. 5 e Mass spectra of the influent of reactor 2 at the different phases of operation, in negative (ESIL) and positive ionization mode (ESID).
197
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 9 1 e2 0 0
Table 3 e Proposed chemical structures for some of the MS detected ions in the influent of reactor 2. m/z (ESI)
m/z (ESIþ)
Molecular weight (g mol1)
123
122
197
198
219
218
247
248
277
278
322
323
342
387
341
388
433
432
Proposed structure
atom of hydrogen from the aromatic system is replaced by a molecule of NO2. The diazotization of this last and of SA can occur (Seymor et al., 2002) to produce the diazonium ions which undergo a coupling reaction with one molecule of aniline to form the compounds with M ¼ 323 g mol1 and 278 g mol1, respectively, producing a coloured azo chromophore (disazo coupling reaction). Other azo compounds proposed may result from other coupling reactions: between two molecules of sulfanilic acid (M ¼ 341 g mol1); one molecule of the compound with mass 218 and one of AS (M ¼ 388 g mol1); and two molecules of the compound M ¼ 218 (product with M ¼ 432 g mol1). Products with mass 278 and 342 can also result as a fragment of the molecules with mass 323 and 388, respectively, by losses of an NO2 group. The presence of azo compounds in the solution is the responsible of the yellow colour obtained. The formation of the product with M ¼ 198 (4-phenylazophenol) may be explained by the loss of SO2 from the azo compound formed from coupling of aniline with SA; other possibility is the coupling between one molecule of aniline and one of 4-aminophenol, this last resulted from SA after losing of SO2. Aromatic sulfonates easily undergo loss of SO2 in the negative mode (Suter et al., 1999). No structures could be attributed to the masses of 250, 283, 392, 408, 481, 525, 569, 663, 685 and 763. The higher masses may correspond to oligomers formed from coupling reactions between the lower masses products. It is worth noting that all the reaction products were soluble in the aqueous influent solution; indeed, except for nitrobenzene and 4-phenylazophenol, all the proposed structures have at least one sulfonic group.
3.4.
Batch denitrification activity experiments
Batch denitrification experiments were performed using the influent (Fig. 6A) and effluent (Fig. 6B) of both reactors. The shape of the curves for the experiments with the influent of R1 is similar to the one of the control with acetate, meaning that no negative effect on the biomass activity was present; nitrate and the aromatic amines at the given concentration in the influent were not inhibitory for the biomass. Kim and Kim (2003) have found that nitrification was not affected by aniline concentrations up to 20 mg L1 (0.215 mM). O’Neill et al. (2000) have grown aerobic bacterial cultures with aniline as substrate at range of 250e1000 mg L1 and, in all cases, it was totally consumed and the specific growth rates were not significantly affected. Due to their poor lipophilicity, sulfonated aromatic amines are normally less toxic than the non-sulfonated (Jung et al., 1992; Chen, 2006). Tan et al. (2005) have concluded that sulfonic aromatic amines at concentrations of 100e200 mg L1 (0.58e1.16 mM) were not toxic to the tested cultures. On the other hand, in batch assays with the influent of R2, a 3 days lag phase is visible, suggesting that products formed by the chemical reaction with nitrite in the influent may exert an initial toxicity. From the results of MS analysis we have proposed the formation of nitro-compounds and it is known that nitro-compounds are more toxic than the corresponding amines (Razo-Flores et al., 1997; Karim and Gupta, 2003). However, we saw that those nitro-compounds coupled to form azo products, probably less toxic. After that initial acclimatization period the activity is recovered and the
A
14
N2+N2O (mmol.L-1)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 9 1 e2 0 0
12 10 8 6 4 2 0
0
2
4
6
8
6
8
Time (d)
B
14
N2+N2O (mmol.L-1)
198
12 10 8 6 4 2 0
0
2
4
Time (d) Fig. 6 e Results from denitrifying activity test with the influent (A) and effluent (B) of both reactors: (C), reactor 1; (-), reactor 2. Controls: (,), only substrate (acetate) and (B), only biomass.
end value of (N2 þ N2O) produced (w12 mmol L1) is even higher than that of using the influent of R1 and of the control with acetate (w9 mmol L1), which indicates that the microorganisms are able to transform/degrade the compounds generated and may use them as a carbon source to grow. The higher activity was also evidenced by the higher values of N-recovery percentage in influent of R2, 67 5%, compared to R1, 49 1% and to the control, 50 1% (Table 4). In the experiments using the effluent, again similar profiles were obtained for the activity with the effluent of R1 and the control with acetate, although in the first the level of (N2 þ N2O) reached was slightly high (w8 mmol L1 and 7 mmol L1, respectively). With the effluent of R2, a lag phase was observed but for a lower period than the obtained with the influent, less than 1 day, and the level of (N2 þ N2O) produced at the end was also lower (w10 mmol L1). The values of Nrecovery are similar for both assays with the effluent (51 2% with R1 effluent and 57 1 with R2 effluent). The results hint
Table 4 e Percentage of N-recovery in the denitrifying activity test. N-recovery %
Reactor 1 Reactor 2 Control with acetate
Influent
Effluent
49 1 67 5 50 1
51 2 57 1
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 9 1 e2 0 0
that some detoxification of the coloured products formed in the influent of R2 occurs upon the treatment in the reactor. No CH4 and CO2 production were obtained confirming that denitrification process prevails.
4.
Conclusions
The results obtained by the continuous UASB reactors operation showed that the proposed system of reactor 1, using nitrate as electron acceptor, works efficiently to totally remove aniline and w60e70% of COD. Sulfanilic acid, with similar structure as aniline but containing an additional sulfonic group, was not degraded in any of the operation phases. The presence of nitrite in the influent of reactor 2 caused a chemical reaction that led to an yellow colour solution development. Both aromatic amines were consumed and 3 new peaks appeared in the HPLC chromatograms at 350 nm: a major one at Rt 14.3 min and two minors at Rt 17.2 and 21.5 min. A decrease of the colour solution was observed after entering the reactor. In the effluent, the peaks at Rt 14.3 and 17.2 min were very low and the one at Rt 21.5 increased, being the main one. The overall COD removal in this reactor was lower (w40%). Some inhibitory effects by the chemically formed products should not be excluded; although, denitrification activity tests suggest that some detoxification occurs upon the treatment in the reactor. Indeed, with the influent solution, a 3 days lag phase was observed, but after that period higher (N2 þ N2O) concentration than in the control was obtained, corroborating the idea that products formed were consumed/degraded after biomass acclimatization.
Acknowledgement This work was supported by the PTDC/AMB/69335/2006 project grants. Luciana Pereira holds a Post-Doc fellowship (SFRH/BPD/20744/2004) and Raquel Pereira holds a fellowship (SFRH/BPD/39086/2007) from Fundac¸a˜o para a Cieˆncia e Tecnologia.
references
APHA, 1989. Standard Methods for the Examination of Water and Wastewater, seventeenth ed. American Public Health Editions Association, Washington DC. Bor-Yann, C., Lin, K.-W., Wang, Y.-M., Yen, C.-Y., 2009. Revealing interactive toxicity of aromatic amines to azo dye decolorizer Aeromonas hudrophila. J. Hazard Mater. 166, 187e194. Brown, D., Laboureur, P., 1983. The aerobic biodegradability of primary aromatic amines. Chemosphere 12, 405e414. Carvalho, M.C., Pereira, C., Gonc¸alves, I.C., Pinheiro, H.M., Santos, A.R., Lopes, A., Ferra, M.I., 2008. Assessment of the biodegradability of a monosulfonated azo dye and aromatic amines. Int. Biodeterior. Biodegrad. 62, 96e103. Chen, B.-Y., 2006. Toxicity assessment of aromatic amines to Pseudomonas luteola: chemostat pulse technique and doseeresponse analysis. Process Biochem., 1529e1538.
199
Coughlin, M.F., Kinkle, B.K., Bishop, P.L., 2002. Degradation of acid orange 7 in aerobic biofilm. Chemosphere 46, 11e19. De, M., O’Connor, O., Kosson, D., 1994. Metabolism of aniline under different anaerobic electron-accepting and nutritional conditions. Environ. Toxicol. Chem. 13, 233e239. Donlon, B.A., Razo-Flores, E., Lettinga, G., Field, J.A., 1996. Continuous detoxification, transformation, and degradation of nitrophenols in Upflow Anaerobic Sludge Blanket (UASB) reactors. Biotechnol. Bioeng. 51, 439e449. Hyung-Yell, K., Jerome, J.K., Kye-Heon, O., 2000. Characterization of a strain HY99, a Nobel microorganism capable of aerobic and anaerobic degradation of aniline. FEMS Microbiol. Lett. 190, 215e221. Jung, R., Steinle, D., Anliker, R., 1992. A compilation of genotoxicity and carcinogenicity data on aromatic aminosulphonic acids. Food Chem. Toxicol. 30, 635e660. Kalyuzhnyi, S., Skyler, V., 2000. Biomineralisation of azo dyes and their breakdown products in anaerobiceaerobic hybrid and UASB reactors. Water Sci. Technol. 41, 23e30. Karim, K., Gupta, S.K., 2003. Continuous biotransformation and removal of nitrophenols under denitrifying conditions. Water Res. 37, 2953e2959. Kim, S.-S., Kim, H.-J., 2003. Impact and Threshold concentration of toxic materials in the stripped gas liquor on nitrification. Korean J. Chem. Eng. 20. O’neill, F.J., Brownley-Challenor, K.C., Greenwood, R.J., Sknapp, J.S. , 2000. Bacterial growth on aniline: implications for the biotreatment of industrial wastewater. Water Res. 34, 4397e4409. Perei, K., Rakhely, G., Kiss, I., Poliak, B., Kovaks, K.L., 2001. Biodegradation of sulfanilic acid by Pseudomonas paucimobilis. Appl. Microbiol. Biotechnol. 55, 101e107. Pinheiro, H.M., Touraud, E., Thomas, O., 2004. Aromatic amines from azo dye reduction: status review with emphasis on direct UV spectrophotometric detection in textile industry wastewaters. Dyes Pigm. 61, 121e139. Razo-Flores, E., Donlon, B.A., Field, J.A., Lettinga, G., 1996. Biodegradability of N-substituted aromatics and alkylphenols under methanogenic conditions using granular sludge. Water Sci. Technol. 33, 47e57. Razo-Flores, E., Donlon, B.A., Lettinga, G., Field, J.A., 1997. Biotransformation and biodegradability of N-substituted aromatics in methanogenic granular sludge. FEMS Microbiol. Rev. 20, 525e538. Razo-Flores, E., Lettinga, G., Field, J.A., 1999a. Biotransformation and biodegradation of selected nitroaromatics under anaerobic conditions. Biotechnol. Prog. 15, 358e365. Razo-Flores, E., Smulders, P., Prenafeta-Bold, F., Lettinga, G., Field, J.A., 1999b. Treatment of anthranilic acid in an anaerobic expanded granular sludge bed reactor at low concentrations. Water Sci. Technol. 40, 187e194. Rust, C.M., Aelion, C.M., Flora, J.R.V., 2000. Control of pH during denitrification in subsurface sediment microcosms using encapsulated phosphate buffer. Water Res. 34, 1447e1454. Schnell, S., Schink, B., 1991. Anaerobic aniline degradation via reductive deamination of 4-aminobenzoyl-CoA in Desulfobacterium anilini. Arch. Microbiol. 155, 183e190. Seymour, E.H., Lawrence, N.S., Pandurangappa, M., Compton, R.G., 2002. Indirect electrochemical detection of nitrite via diazotization of aromatic amines. Microchim. Acta 140, 211e217. Suter, M.J.-F., Riediker, S., Giger, W., 1999. Selective determination of aromatic sulfonates in landfill leachates and groundwater using microbore liquid chromatography coupled with mass spectrometry. Anal. Chem. 71, 897e904. Tan, N.C.G., Prenafeta-Boldu´, F.X., Opsteeg, J.L., Lettinga, G., Field, J.A., 1999. Biodegradation of azo dyes in coculture of anaerobic granular sludge with aerobic aromatic amine
200
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 9 1 e2 0 0
degrading enrichment cultures. Appl. Microbiol. Biotechnol. 51, 865e871. Tan, N.C.G., van Leeuwen, A., van Voorthuinzen, E.M., Slenders, P., Prenafeta-Boldu´, F.X., Temmink, H., Lettinga, G., Field, J.A., 2005. Fate and biodegradability of sulfonated aromatic amines. Biodegradation 16, 527e537. Tan, N.C.G., Field, J.A., 2005. Biodegradation of sulfonated aromatic compounds. In: Environmental Technologies to Treat Sulfur Pollution Principles and Engineering. IWA Publishing, London, pp. 377e392.
Van der Zee, F., Villaverde, S., 2005. Combined anaerobiceaerobic treatment of azo dyes e a short review of bioreactors studies. Water Res. 39, 1425e1440. Va´zquez-Rodrı´guez, G.A., Beltra´n-Herna´ndez, R.I., LuchoConstantino, C.A., Blasco, J.L., 2008. A method for measuring the anoxic biodegradability under denitrifying conditions. Chemosphere 71, 1363e1368. Wu, Y.-G., Hui, L., Li, X., Zhang, Y.-Z., Zhang, W.-C., 2007. Degradation of aniline in Weihe riverbed sediments under denitrification conditions. J. Environ. Sci. Health Part A 42, 413e419.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 0 1 e2 1 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Enhanced nitrogen removal from pharmaceutical wastewater using SBA-ANAMMOX process Chong-Jian Tang a, Ping Zheng a,*, Ting-Ting Chen a, Ji-Qiang Zhang a, Qaisar Mahmood b, Shuang Ding a, Xiao-Guang Chen a, Jian-Wei Chen a, Da-Tian Wu c a
Department of Environmental Engineering, Zhejiang University, Hangzhou 310029, China Department of Environmental Science, COMSATS University, Abbottabad, Pakistan c Tianyi Environmental Protection Group Ltd., Jinhua, Zhejiang Province 321000, China b
article info
abstract
Article history:
Efficient biological nitrogen removal from pharmaceutical wastewater has been focused
Received 14 April 2010
recently. The present study dealt with the treatment of colistin sulfate and kitasamycin
Received in revised form
manufacturing wastewater through anaerobic ammonium oxidation (ANAMMOX). The
13 August 2010
biotoxicity assay on luminescent bacterium Photobacterium phosphoreum (T3 mutation)
Accepted 19 August 2010
showed that the pharmaceutical wastewater imparted severe toxicity with a relative
Available online 27 August 2010
luminosity of 3.46% 0.45%. During long-term operation, the cumulative toxicity from toxic pollutants in wastewater resulted in the performance collapse of conventional
Keywords:
ANAMMOX process. A novel ANAMMOX process with sequential biocatalyst (ANAMMOX
Pharmaceutical wastewater
granules) addition (SBA-ANAMMOX process) was developed by combining high-rate
ANAMMOX
ANAMMOX reactor with sequential biocatalyst addition (SBA). At biocatalyst addition rate
Biological nitrogen removal
of 0.025 g VSS (L wastewater)1 day1, the nitrogen removal rate of the process reached up
Toxicity
to 9.4 kg N m3 day1 in pharmaceutical wastewater treatment. The effluent ammonium
Sequential biocatalyst addition
concentration was lower than 50 mg N L1, which met the Discharge Standard of Water Pollutants for Pharmaceutical Industry in China (GB 21903-2008). The application of SBAANAMMOX process in refractory ammonium-rich wastewater is promising. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
The pharmaceutical industry greatly developed in the last few decades due to enormous demands of life saving drugs (Wang et al., 2005). It is a major source of wastewater containing refractory toxicants which are difficult for wastewater treatment (Wang et al., 2005; Chelliapan et al., 2006). Some pharmaceutical wastewaters (PW) like colistin sulfate and kitasamycin wastewater are characterized by the presence of high concentration of chemical oxygen demand (COD) and ammonium species. The organic pollutants were efficiently removed by anaerobic methanogenic process in the last 20
years (El-Gohary et al., 1995; Jenicek et al., 1996; Wang et al., 2005; Chelliapan et al., 2006), while the ammonium was extensively treated with physicochemical methods or conventional biological methods (Gupta and Sharma, 1996; Peng et al., 2004; Ma et al., 2009).The applications of these technologies were significantly limited due to the high operational cost and low removal efficiency. In many cases, the conventional nitrificationedenitrification process was not feasible for nitrogen removal due to insufficient biodegradable organic matter in the anaerobic effluent (Ferna´ndez et al., 2009). Biological nitrogen removal from refractory pharmaceutical wastewater is very difficult due to its strong biotoxicity
* Corresponding author. Tel./fax: þ86 571 86971709. E-mail addresses:
[email protected] (C.-J. Tang),
[email protected] (P. Zheng). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.036
202
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 0 1 e2 1 0
(Ferna´ndez et al., 2009; Okuda et al., 2009). Most researchers focused their efforts on pre-treatment technologies such as solar photo-Fenton technique and ultrasonic irradiation to remove the pharmaceuticals from wastewater (Naddeo et al., 2009; Sirtori et al., 2009; Stalter et al., 2009). However, their high implementation costs are the serious bottleneck (Sirtori et al., 2009). Therefore, high-rate and cost-effective biotechnologies are imminently required for nitrogen removal from ammonium-rich pharmaceutical wastewater. Anaerobic ammonium oxidation (ANAMMOX) was initially discovered in a denitrifying reactor in 1995 (Mulder et al., 1995). Under anoxic condition, the autotrophic ANAMMOX bacteria oxidize ammonium to nitrogen gas using nitrite as electron acceptor (Strous et al., 1998). ANAMMOX process offers some significant advantages such as no demand for oxygen and organic carbon and low sludge production. The high-rate ANAMMOX reactors with nitrogen removal rate (NRR) above 8 kg N m3 day1 have been frequently reported (Sliekers et al., 2003; Isaka et al., 2007; Tsushima et al., 2007; van der Star et al., 2007; Tang et al., 2010a,c). Several fullscale ANAMMOX plants have been employed for nitrogen removal from ammonium-rich wastewaters with maximum NRR up to 9.5 kg N m3 day1 (van der Star et al., 2007; Joss et al., 2009). However, ANAMMOX process was restricted to treat a few low COD-containing ammonium-rich wastewaters such as sludge digester liquor, tomato processing effluent and landfill leachate (van der Star et al., 2007; Joss et al., 2009). It is susceptible to toxic components in wastewater (Ferna´ndez et al., 2009) because ANAMMOX bacteria grow slowly with a low cellular yield (Strous et al., 1998). The treatment of recalcitrant pharmaceutical wastewater using ANAMMOX process is a great challenge. When exposed to toxic substances during a long-term operation, the activity of ANAMMOX bacteria will be significantly inhibited and thus, the quantity of ANAMMOX bacteria in reactor system will not satisfy the biomass requirement considering the slow growth rate and low cellular yield. Consequently, reactor performance will finally deteriorate. Exogenous addition of high-activity ANAMMOX biomass into the reactor system can supplement the required bacterial amount and thus, enhance the nitrogen removal performance. The objective of this study was to develop an innovative ANAMMOX process (SBA-ANAMMOX process) by combining high-rate ANAMMOX technology with sequential biocatalyst (ANAMMOX biomass) addition (SBA) and then to investigate the performance of SBA-ANAMMOX process in treating nitritated pharmaceutical wastewater.
2. 2.1.
Materials and methods Pharmaceutical wastewater
The pharmaceutical wastewater was collected from a colistin sulfate and kitasamycin manufacturing plant (colistin sulfate manufacturing wastewater was 75 m3 day1 and kitasamycin processing wastewater was 50 m3 day1), located in Zhejiang Province, China. The two types of wastewaters were mixed in an equalization pond. Efficient COD removal was accomplished in the anaerobic methanogenic process prior to the biological
nitrogen removal process. The raw pharmaceutical wastewater taken from the outlet of nitrificationedenitrification process was fulvous in color with unpleasant smell. Partial nitritation was accomplished as pre-treatment in a 10 L laboratory-scale sequential batch reactor (SBR) and its effluent was used as the experimental wastewater for ANAMMOX process. The inoculum for the SBR was taken from a pilot-scale nitritation reactor (2.5 m3) treating landfill leachate for more than 2 years. The biomass concentration in the SBR was set at 8 g VSS L1 (VSS: volatile suspended solids), and the pH was adjusted to 8.2e8.3 by dosing NaHCO3. The dissolved oxygen was controlled in a range of 1.0e1.3 mg L1, and the temperature was set at 35 1 C. After the filling of 5 L raw pharmaceutical wastewater (10 min), the SBR was aerated and stirred for a period of 11.5 h. After the aeration, the sludge was allowed to settle for 10 min. In the remaining 10 min of the total cycle, part of the liquid was discharged. The constituents of the SBR effluent are listed in Table 1. After partial nitritation, the ratio of þ NO 2 eN/NH4 eN in the SBR effluent was in the range of 1.0e1.4 (ideal for subsequent ANAMMOX process), with relatively low biochemical oxygen demand (BOD5) value but containing relatively high nitrogenous compounds concentration.
2.2.
ANAMMOX reactor
The experimental work for pharmaceutical wastewater treatment was carried out in two upflow anaerobic sludge blanket (UASB) reactors (R1 and R2). A high-rate ANAMMOX reactor (HR) was operated simultaneously in order to produce high-activity seed ANAMMOX granules. All the reactors were made of glass with 1.1 L capacity having internal diameter of 50 mm and were completely covered with black cloth to avoid the growth of phototrophic microorganisms and the related oxygen production (van der Star et al., 2008). The reactors were continuously fed with the nitritated pharmaceutical wastewater (R1 and R2) and synthetic wastewater (HR) flushed with 95% Are5% CO2 to maintain anoxic conditions. The temperature was set at 35 1 C. The possible trace elements deficiency in the pharmaceutical wastewater was corrected by adding a trace elements solution (van de Graaf et al., 1996).
2.3.
Seed sludge
Both reactors (R1 and R2) were inoculated with 0.5 L ANAMMOX granules (about 30 g VSS), which were cultivated for about 400 days in laboratory-scale reactors with NRR higher
Table 1 e Characteristics of the nitritation SBR effluent. Parameter Colour pH 1 NHþ 4 eN (mg L ) 1 NO2 eN (mg L ) þ NO 2 eN/NH4 eN NO3 eN (mg L1) COD (mg L1) BOD5 (mg L1)
Range Fulvous 6.8e7.8 123e257 133e264 1.0e1.4 0e32 415e843 0e51
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 0 1 e2 1 0
than 10 kg N m3 day1 (Tang et al., 2010a,b). The dominant ANAMMOX population in seed granules was Brocadia (Hu et al., 2010). The seed ANAMMOX granules were carmine in color having average diameter of 2.51 mm. The VSS and total suspended solids (TSS) were 29.3 and 32.9 g L1, respectively, with VSS/TSS ratio of 0.89. During the operation of both UASB reactors, HR was also operated continuously feeding synthetic medium with high NRR up to 50 kg N m3 day1 to develop high-activity seed ANAMMOX granules, which served as the biocatalyst for R1.
2.4.
Reactor operation
The overall reactor operation was divided into three phases. During phase I (days 1e41), both reactors were continuously fed with the effluent of the nitritation SBR, aimed at testing the feasibility of direct treatment of pharmaceutical wastewater with ANAMMOX process. The hydraulic retention time (HRT) of both reactors was adjusted at 16 h. During Phase II (days 42e76), the SBA-ANAMMOX process was established. During Phase III (days 77e170), SBA-ANAMMOX process was developed to enhance nitrogen removal from the pharmaceutical wastewater. On day 42, 150 mL of seed ANAMMOX granules (about 9 g VSS) was added to both reactors and equal amount of sludge was simultaneously taken out of the two reactors to maintain constant sludge concentrations. Subsequently, R1 was operated in SBA-ANAMMOX mode, while R2 was operated as a check without any biocatalyst addition. During Phases II and III, 5e10 mL ANAMMOX granules (0.3e0.6 g VSS) were added into R1 once the effluent nitrite concentration was higher than 10 mg N L1 to prevent the performance deterioration. During Phase III, the HRT of R1 was progressively shortened to explore the nitrogen removal potential. The SBA experiment was terminated after an operation of 170 days where the effluent ammonium concentration was higher than the ammonium discharge standard (50 mg N L1) for pharmaceutical wastewater (GB 21903-2008). The HRT of R2 was kept constant at 16 h during Phases II and III.
2.5.
Batch test
The feasibility of pharmaceutical wastewater treatment with ANAMMOX process was initially assessed by batch tests. Sealed vials (120 mL capacity) containing 100 mL of liquid volume were used to perform the batch cultivation. At the beginning of experiments, the biomass concentration was set at 2.4 g VSS L1 keeping temperature at 35 1 C. The ammonium and nitrite concentrations of the nitritated pharmaceutical wastewater and simulated synthetic wastewater (SW) used in batch cultivation were 187.4 0.9, 227.8 0.9 and 186.9 2.0, 221.7 1.6 mg N L1, respectively. The pH was set at 7.8. Gas and liquid phases were purged with 95% Are5% CO2 to maintain anaerobic condition. The ammonium and nitrite concentrations were analyzed at regular intervals and each assay was performed in triplicates.
2.6.
Analytical methods
The influent and effluent samples were collected on daily basis and were analyzed immediately. The determination of
203
pH, ammonium, nitrite, nitrate, COD and BOD5 were carried out following the Standard Methods (APHA, 1998). TSS and VSS were measured by the weighing method after being dried at 105 C and burnt to ash at 550 C. The size of granular sludge was measured by image analysis system (QCOLite) with Leica DM2LB microscope and a digital camera (Canon S30). The transmission electron microscopy (TEM) observation of the biomass was performed according to Tang et al. (2009). NRR was calculated as the sum of ammonium and nitrite consumption rate. Extracellular polymeric substances (EPS) were extracted from sludge by EDTA, then the extracellular proteins were determined by Lowry method using egg albumin as standard and the polysaccharide content was analyzed by the anthrone method taking glucose as standard (Sheng et al., 2006; Wu et al., 2009).
2.7.
Acute toxicity assay
The raw pharmaceutical wastewater, nitritated pharmaceutical wastewater and simulated synthetic wastewater samples were used for acute toxicity assay. Prior to acute toxicity assay, the luminescent bacterium Photobacterium phosphoreum (T3 mutation) was reactivated in 1 mL 2.5% NaCl solution and stored in the ice water bath (Jiao et al., 2008). Aliquots of treated sample (0.2 mL) and 10 mL reactivated bacteria were added to 2 mL 3% NaCl solution. After an exposure of 15-min at 20e25 C, the decrease in bioluminescence (indicator of the toxic effect) was measured by Model Toxicity Analyzer (DXY-2, Institute of Soil Science, Chinese Academy of Sciences, Nanjing, China). The relative luminosity (X ) was calculated according to Eq. (1). Xð%Þ ¼
Lsample 100% Lblank
(1)
where, L is the luminosity.
2.8.
Cumulative toxicity assay
ANAMMOX granules (2 g) were harvested from R2 and were filtrated with 0.45 mm filter. Then the sludge liquor was obtained after the centrifugation at 3000 rpm. The cumulative toxicity of the ANAMMOX sludge liquor after long-term operation with the pharmaceutical wastewater was assayed by the decrease in relative luminosity using the seed ANAMMOX sludge as control.
3.
Results
3.1.
Performance of HR
During the overall operation, HR was continuously operated under relatively short HRTs (0.55e0.24 h) and high nitrogen loading rate (NLR, 20e60 kg N m3 day1) to develop highactivity ANAMMOX granules (Fig. 1) which would be used as the biocatalyst in the subsequent real pharmaceutical wastewater treatment. As evident in Fig. 1, during 40e170 days, HR was operated stably with NRR as high as 40e50 kg N m3 day1, ammonium and nitrite removal efficiencies of 77.8% 8.0% and 84.3% 7.3%, respectively. Complete biomass granulation
204
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 0 1 e2 1 0
with the specific activity of 2.0 kg N kg VSS1 day1 suggesting a stable ANAMMOX reaction. The stoichiometric ratio and specific activity were close to the counterparts for simulated þ synthetic wastewater (NO 2 eN/NH4 eN removal ratio, 1.31; specific activity, 1.7 kg N kg VSS1 day1, Fig. 2A). The ammonium to nitrite stoichiometric ratio in batch tests treating pharmaceutical wastewater was close to the reported values for ANAMMOX process (Strous et al., 1998; Tang et al., 2010a) showing that denitrification did not prevail, which was probably due to the lack of biodegradable organic matter. The stable ANAMMOX reaction in pharmaceutical wastewater batch cultivation did not indicate any inhibition during shortterm exposure to the pharmaceutical wastewater and thus, nitrogen removal from the pharmaceutical wastewater through ANAMMOX seemed quite feasible.
3.2.2.
Fig. 1 e Nitrogen removal performance of the high-rate reactor.
occurred in the reactor; and the stoichiometric ratio of ammonium conversion, nitrite removal and nitrate production was 1:(1.24 0.09):(0.25 0.03), which was close to the reported values for ANAMMOX process (Strous et al., 1998; Tsushima et al., 2007; Tang et al., 2010a). The carmine ANAMMOX granules enriched in HR possessed relatively high activity (1.7e2.0 kg N kg VSS1 day1) and high EPS content (210 mg g VSS1).
3.2.
Direct treatment of pharmaceutical wastewater
3.2.1.
Batch tests
Batch cultivation was performed to assess the feasibility of direct treatment of the nitritated pharmaceutical wastewater in ANAMMOX process (Fig. 2). As evident in Fig. 2B, ammonium and nitrite concentrations decreased simultaneously at stoichiometric ratio of 1.33
Reactor performance in Phase I
During Phase I, two ANAMMOX UASB reactors were operated in parallel by feeding with the nitritated pharmaceutical wastewater. The influent and effluent pollutant concentrations for both reactors are shown in Fig. 3. As indicated in Fig. 3A and B, the effluent ammonium and nitrite concentrations of R1 during the initial 10 days were 28.4 7.8 mg N L1 and 5.9 4.8 mg N L1 with the corresponding removal efficiencies of 84.1% and 97.0%, respectively. The ammonium and nitrite concentrations in R2 effluent were close to those of R1 effluent (Fig. 3A and B) attaining ammonium and nitrite removal efficiencies of 85.2% and 95.0%, respectively. The nitrate production was observed in both reactors (Fig. 3C). The stoichiometric ratio of the ammonium removal, nitrite consumption and nitrate production was 1:1.31:0.06. Such results suggested that the performance of both ANAMMOX UASB reactors was stable and efficient. Thereafter, the nitrogen removal efficiency progressively decreased. At the end of phase I, even no ammonium removal was observed, concomitantly the nitrite removal efficiency also decreased to 24.9%e57.1% (Fig. 3B), indicating the operation failure of both ANAMMOX reactors. The influent and effluent COD concentrations and COD removal efficiency did not vary significantly throughout the operation (Fig. 3D), and the effluent BOD5 concentrations kept lower than 10 mg L1 (data not shown).
Fig. 2 e Batch tests for synthetic wastewater (A) and pharmaceutical wastewater (B).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 0 1 e2 1 0
205
The dominant cells were obviously compartmentalized, displaying typical ultrastructural features of ANAMMOX bacteria, i.e., a single membrane bound anammoxosome containing tubule-like structure, and cytoplasm with ribosome-like particles separated from paryphoplasm (electron-transparent in Fig. 4C) at the cell rim by an intracytoplasmic membrane (Lindsay et al., 2001). However, the sludge in both reactors changed to filemot after the operation for 41 days (Fig. 4D and G), suggesting a decrease in hemachrome content and number of ANAMMOX bacteria (Tang et al., 2010c). The diameter of the ANAMMOX granules decreased to 1.21 0.43 mm (R1, Fig. 4E) and 1.49 0.72 mm (R2, Fig. 4H). TEM photographs showed a significant decrease in number of the ANAMMOX cells (Fig. 4F and I). The sludge concentration in both reactors also decreased from 19.0 g VSS L1 to 15.7 g VSS L1 (R1) and 17.3 g VSS L1 (R2). All the parameters mentioned above suggested hydrolysis of the ANAMMOX granules in both reactors (Tang et al., 2010b). The nitrite inhibition could be excluded since the nitrite concentration in the nitritated pharmaceutical wastewater never surpassed 250 mg N L1, which was similar to that of HR. The inhibition caused by pH and free ammonia was also considered out of question since the pH was in the range of 8.01e8.36 in the reactors and the free ammonia was below the toxic limits (Tang et al., 2009). It seemed that some inhibitory substances in the pharmaceutical wastewater were responsible for the operation failure of two ANAMMOX UASB reactors. Furthermore, a meagre amount of organics would be released during sludge hydrolysis (Tang et al., 2010d). Thus, heterotrophic denitrification probably occurred in the reactors using released organic matter as electron donor to reduce nitrite. The nitrite removal above 20% was observed when no ammonium was oxidized, which further confirmed the prevalence of heterotrophic denitrification and sludge hydrolysis in the reactor system directly treating the pharmaceutical wastewater.
3.3.
Fig. 3 e Performance of the two UASB reactors during Phase I.
3.2.3.
Sludge characteristics in Phase I
The sludge characteristics during Phase I are shown in Fig. 4. The seed ANAMMOX granular sludge possessed scarlet color (Fig. 4A) with an average diameter of 2.51 0.91 mm (Fig. 4B).
Toxicity assessment
The biotoxicity of pharmaceutical wastewater was assayed using luminescent bacteria. Fig. 5 shows the acute biotoxicity of pharmaceutical wastewater at different dilution rates. Acute toxicity of the raw as well as the nitritated pharmaceutical wastewater was observed when the dilution rate was lower than 2.5. The relative luminosities were only 3.46% 0.45% when no dilution was applied. While the simulated synthetic wastewater containing the same ammonium and nitrite concentrations did not show acute toxicity (Fig. 5). Therefore, the acute toxic effects did not result from ammonium and nitrite; instead, they might be caused by some refractory toxicants (such as colistin sulfate, kitasamycin and other toxicants) of pharmaceutical wastewater. Such toxicants were possibly discharged into the pharmaceutical wastewater and were inhibitory to microbial activity even under low concentrations (Jordan and Knight, 1984; Jeong et al., 2009). Based on the short-term batch cultivation, long-term continuous cultivation and acute toxicity assays, it seemed that the biotoxicity of pharmaceutical wastewater was accumulative. Thus, the cumulative toxicity of ANAMMOX sludge liquor was assessed (Fig. 6).
206
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 0 1 e2 1 0
Fig. 4 e Apparent characteristic, diameter distribution and TEM observation of the sludges at the beginning (A, B, C) and the end of Phase I (R1: D, E, F; R2: G, H, I).
As shown in Fig. 6, the acute toxicity of the liquor extracted from ANAMMOX sludge of R2 treating pharmaceutical wastewater for 41 days was stronger than that of the extracted liquor from seed ANAMMOX sludge. It suggested that the toxic effects to ANAMMOX sludge during long-term operation were cumulative.
3.4.
Fig. 5 e Acute biotoxicity assays of the pharmaceutical wastewater (PW) using synthetic wastewater (SW) as control.
Establishment of SBA-ANAMMOX process
The performance of the two UASB reactors during initial 10 days of Phase I suggested that the inhibition did not occur during short-term operation. The performance could be improved by modifying the ANAMMOX process with sequential ANAMMOX granules addition. The performance of R1 with sequential biocatalyst addition (SBA-ANAMMOX process) and R2 without ANAMMOX granules addition (conventional ANAMMOX process) are shown in Figs. 7 and 8, respectively. During Phase II, both SBA-ANAMMOX reactor and conventional ANAMMOX reactor displayed efficient performance for
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 0 1 e2 1 0
207
Fig. 6 e Cumulative toxicity of the ANAMMOX granules extracted liquor after treating the pharmaceutical wastewater in long-term operation.
the initial 10e15 days. The effluent ammonium concentration was 6.6e20.7 (R1) and 0e8.9 mg N L1 (R2), respectively. During the following days, the performance of R1 (SBA-ANAMMOX reactor) was good and stable (Fig. 7). The effluent nitrite concentration never surpassed 20 mg N L1 (Fig. 7B), while the effluent ammonium concentration maintained in the range of 5e50 mg N L1 (Fig. 7A) which met the ammonium discharging standard (50 mg N L1) for pharmaceutical industry, China (GB 21903-2008). Correspondingly, the ammonium and nitrite removal efficiencies were 80.4%e86.2% and 92.6%e94.8%, respectively (Fig. 7A and B). As for R2 (conventional ANAMMOX process), the performance significantly deteriorated as both residual ammonium and nitrite concentrations were over 150 mg N L1. Ammonium and nitrite removal efficiencies were as low as 1.9%e13.8% and 14.5%e34.2%, respectively (Fig. 8), which were similar to those in Phase I. The comparison of R1 and R2 performance in Phase II (35 days) showed that SBAANAMMOX process could successfully overcome the toxic effects caused by pharmaceutical wastewater. During Phase II, the nitrite removal to ammonium conversion ratio in R1 progressively decreased to 1.0e1.5 (Fig. 7E), which was close to the value reported by Strous et al. (1998). While, it varied largely from 2.69 to 5.22 for R2. Relatively high nitrite removal suggested the prevalence of heterotrophic denitrification in R2 as described above. During Phase II (35 days), 2.5 g VSS of ANAMMOX sludge was added into R1 by 10 times; consequently, the performance of SBA-ANAMMOX reactor significantly improved.
3.5. Performance of SBA-ANAMMOX process treating pharmaceutical wastewater In Phase III, the HRT of R1 with SBA was progressively shortened to 1.0 h and the corresponding NLR was increased to 8.1e10.4 kg N m3 day1. During that period, the effluent nitrite concentration was 18.2 13.2 mg N L1 and the average effluent ammonium concentration was 40.8 mg N L1, with the
Fig. 7 e Performance of R1 treating pharmaceutical wastewater using sequential ANAMMOX granules addition.
average ammonium and nitrite removal efficiencies of 78.1% and 90.2%, respectively (Fig. 7A and B). As a consequence, the NRR reached 7.2e9.4 kg N m3 day1. To our knowledge, such a high nitrogen removal capacity for pharmaceutical wastewater has not been previously reported. As a control reactor, the HRT of R2 was maintained at 16 h. The ammonium and nitrite removal efficiencies were 4.7% e21.1% and 22.9%e35.9%, respectively (Fig. 8). The NRR was in the range of 0.05e0.178 kg N m3 day1. Obviously, the newlydeveloped SBA-ANAMMOX process was far more efficient.
208
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 0 1 e2 1 0
Fig. 8 e Nitrogen removal performance of R2 during Phase II and III.
The toxicity assessment using luminescent bacteria showed that the cumulative toxicity to the ANAMMOX sludge occurred after long-term exposure to the pharmaceutical wastewater. The ANAMMOX bacteria can secret a large amount of EPS, which contain ample quantities of negatively charged functional groups and have strong capabilities to adsorb toxicants (Guine et al., 2006; Sheng et al., 2006). Such a high EPS content of the ANAMMOX granules in the present study was supposed to be responsible for the adsorption of toxicants from wastewater that resulted in cumulative toxic effects. Moreover, different components in both the antibiotics manufacturing wastewaters might result in an additive effect when they were mixed (Quinn et al., 2009).
4.2. Towards the end of Phase III when the NLR of R1 was higher than 6.0 kg N m3 day1, the biocatalyst addition was applied every day. The biocatalyst addition rate was 0.025 g VSS (L wastewater)1 day1 during days 143e170.
4.
Discussion
4.1.
Biotoxicity of pharmaceutical wastewater
The strong inhibitory effects of antibiotics to bacterial activity have been extensively reported (Flaherty and Dodson, 2005; Gagne´ et al., 2006). Campos et al. (2001) observed 50% inhibition of nitrifying activity when 250 mg L1 of oxytetracycline was continuously fed to a 1 L bioreactor. The inhibition was even stronger to anaerobic bacteria (Sanz et al., 1996; Ferna´ndez et al., 2009). The inhibitory effects of toxicants to ANAMMOX bacteria have been recently reported. van de Graaf et al. (1996) observed a 68% decrease in ANAMMOX activity when exposed to 200 mg L1 of chloramphenicol in batch cultivation. Ferna´ndez et al. (2009) investigated the short-term and long-term effects of two broad-spectrum antibiotics, tetracycline and chloramphenicol on ANAMMOX bacteria. It was found that nitrogen removal efficiency was just 25% over control in the presence of 20 mg L1 of chloramphenicol in a continuously operated SBR; and the ANAMMOX activity decreased from 0.25 to 0.05 kg N kg VSS1 day1 after about 70 days operation. Similarly, tetracycline concentration of 10 mg L1 caused a decreased in ANAMMOX activity around 40% of the initial value. In the present study, the pharmaceutical wastewater from a colistin sulfate and kitasamycin manufacturing plant exhibited strong biotoxicity, but the results of batch cultivation and the short-term continuous operation (the initial 10e15 days) showed that the ANAMMOX activity was not inhibited. A possible reason may be the high concentration of ANAMMOX bacteria enriched in the seed granules since the specific ANAMMOX activity was as high as 1.7e2.0 kg N kg VSS1 day1. Thus, they might have possessed greater resistance to the toxicants during short-term operation (Batchelor et al., 1997; Tang et al., 2010b). Nevertheless, cumulative toxicity to the ANAMMOX sludge was observed after a relatively longer operation (41 days). Furthermore, the color of ANAMMOX granules changed and the ANAMMOX granules dispersed.
Effect of sequential biocatalyst addition
The high-activity ANAMMOX granules were added into the reactor for prevention of performance deterioration caused by the biotoxicity of the pharmaceutical wastewater. As described above, the biotoxicity was effectively overcome by the SBA due to the probable stronger resistance of the highlyenriched ANAMMOX granules to toxic substances. During the progressive shortening of HRT (130 days), the SBA-ANAMMOX process was stably operated without severe inhibition. Another positive effect of the SBA might be the introduction of some unknown growth factors. As known, the autotrophic ANAMMOX bacteria possess very slow growth rate with very long doubling time (11 days, Strous et al., 1998); thus, they were difficult to culture (Strous, 2000). Until now, no pure culture of ANAMMOX bacterium has been isolated and identified. One of the possible explanations seems that the growth factors required for culturing of the ANAMMOX bacteria are still unknown. In the present study, the HR ANAMMOX reactor possessed extremely high nitrogen removal rate and significant biomass production. Thus, the growth factors might be probably included in the HR reactor system. From this viewpoint, deliberate addition of high-activity ANAMMOX seeds into R1 probably led to the addition of some unknown growth factors contained in the granules that enhanced the growth of ANAMMOX bacteria. The granules addition subsequently contributed to the high-rate nitrogen removal performance at the low biocatalyst addition rate.
4.3.
Application of SBA-ANAMMOX process
In this study, the novel SBA-ANAMMOX process has been established and proven to be suitable for high-rate nitrogen removal from refractory pharmaceutical wastewater. The NRR of the SBA-ANAMMOX process was up to 9.4 kg N m3 day1 at the sludge addition rate of 0.025 g VSS (L wastewater)1 day1, suggesting that the application of SBA-ANAMMOX process to pharmaceutical wastewater treatment was quite promising. As proposed by Wett (2006) and van der Star et al. (2007), the pre-enrichment of high-activity ANAMMOX biomass in lab-scale high-rate reactors contributed significantly to the full-scale application of ANAMMOX process. As regards the SBA-ANAMMOX process, it is more important to enrich ANAMMOX biomass in high-rate reactors as the novel process required sequential addition of biocatalyst.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 0 1 e2 1 0
In previous studies, researchers have demonstrated that ANAMMOX process possessed very high nitrogen loading and removal rate under laboratory conditions by applying the optimum temperature (e.g., 35e37 C) and pH value (6.8e7.5) (Isaka et al., 2007; Tsushima et al., 2007; Tang et al., 2010a,c). To date, the highest NRR of ANAMMOX process was developed to 77 kg N m3 day1 when the operation mode of low substrate concentration and high flow-rate was applied (Tang et al., 2010c). The ANAMMOX biomass production was significant under the high loading and removal rates (Tang et al., 2010c). Therefore, the cultivation of high-activity ANAMMOX biomass in lab-scale reactors was not so difficult. Based on the results of the present study, the implementation of SBA-ANAMMOX for full-scale treatment of the antibiotics manufacturing wastewater (125 m3 day1) was estimated under the nitrogen removal rate of 1.0 kg N m3 day1 (the normal value of full-scale ANAMMOX reactors as concluded by van der Star et al. (2007)). It was shown that operating a 216 L HR ANAMMOX reactor with NRR of 50 kg N m3 day1 was necessary to fulfill the biocatalyst requirement for the SBA. As estimated, the additional cost of keeping the high-rate ANAMMOX reactor was around 0.41 Euro kg1 N; and the overall cost was about 1.16 Euro kg1 N which was much lower than that of nitrificationedenitrificaiton process (2e5 Euro kg1 N, Jetten et al., 2005). Thus, the cost-effective and high-rate SBA-ANAMMOX process is a promising biotechnology with significant advantages that can effectively treat refractory ammoniumrich wastewaters from pharmaceutical industry.
5.
Conclusions
The pharmaceutical wastewater possessed severe acute toxicity with a relative luminosity value of 3.46% 0.45%. Because of the cumulative toxicity, the conventional ANAMMOX process was not suitable for nitrogen removal from pharmaceutical wastewater. A novel SBA-ANAMMOX process was developed by combination of high-rate ANAMMOX reactor with sequential biocatalyst addition. With the very low biocatalyst addition rate of 0.025 g VSS (L wastewater)1 day1, the nitrogen removal performance significantly improved with the nitrogen removal rate up to 9.4 kg N m3 day1, which was of high value for nitrogen removal from pharmaceutical wastewater. The effluent ammonium concentration was lower than 50 mg N L1, which met the Discharge Standard of Water Pollutants for Pharmaceutical Industry in China (GB 219032008). The present strategy based on SBA-ANAMMOX process could be further applied for enhancing nitrogen removal from refractory real ammonium-rich wastewaters.
Acknowledgements This work is partially supported by the National High-Tech Research and Development (R&D) Program of China (2009AA06Z311), the Natural Science Foundation of China (30770039) and the National key Technologies R&D Program of China (2008BADC4B05). Dr. C.J. Tang wants to thank the
209
anonymous reviewers and editors for their contributions to improvement of the manuscript.
Appendix. Supplementary material Supplementary data related to this article can be found online at doi:10.1016/j.watres.2010.08.036.
references
APHA, 1998. Standard Methods for the Examination of Water and Wastewater, twentieth ed. American Public Health Association. Batchelor, S.E., Cooper, M., Chhabra, S.R., Glover, L.A., Stewart, G. S.A.B., Williams, P., Prosser, J.I., 1997. Cell density-regulated recovery of starved biofilm populations of ammonia-oxidizing bacteria. Appl. Environ. Microbiol. 63, 2281e2286. Campos, J.L., Garrido, J.M., Me´ndez, R., Lema, J.M., 2001. Effect of two broad-spectrum antibiotics on activity and stability of continuous nitrifying systems. Appl. Biochem. Biotechnol. 95, 1e10. Chelliapan, S., Wilby, T., Sallis, P.J., 2006. Performance of an upflow anaerobic stage reactor (UASB) in the treatment of pharmaceutical wastewater containing macrolide antibiotics. Water Res. 40, 507e516. El-Gohary, F.A., Abou-Elela, S.I., Aly, H.I., 1995. Evaluation of biological technologies for wastewater treatment in the pharmaceutical industry. Water Sci. Technol. 32, 13e20. Ferna´ndez, I., Mosquera-Corral, A., Campos, J.L., Me´ndez, R., 2009. Operation of an Anammox SBR in the presence of two broadspectrum antibiotics. Process Biochem. 44, 494e498. Flaherty, C.M., Dodson, S.I., 2005. Effects of pharmaceuticals on Daphnia survival, growth, and reproduction. Chemosphere 61, 200e207. Gagne´, F., Blaise, C., Fournier, M., Hansen, P.D., 2006. Effects of selected pharmaceutical products on phagocytic activity in Elliptio complanata mussels. Comp. Biochem. Physiol. 43, 179e186. Guine, L., Spadini, G., Sarret, G., Muris, M., Delolme, J.P., Gaudet, J.M.F., 2006. Zinc sorption to three gram-negative bacteria: combined titration, modeling, and EXAFS study. Environ. Sci. Technol. 40, 1086e1813. Gupta, S.K., Sharma, R., 1996. Biological oxidation of high strength nitrogenous wastewater. Water Res. 30, 593e600. Hu, B.L., Zheng, P., Tang, C.J., Chen, J.W., van der Biezen, E., Zhang, L., Ni, B.J., Jetten, M.S.M., Yan, J., Yu, H.Q., Kartal, B., 2010. Identification and quantification of anammox bacteria in eight nitrogen removal reactors. Water Res.. doi:10.1016/j. watres.2010.07.021. Isaka, K., Sumino, T., Tsuneda, S., 2007. High nitrogen removal performance at moderately low temperature utilizing anaerobic ammonium oxidation reactions. J. Biosci. Bioeng. 103, 486e490. Jenicek, P., Zabranska, J., Dohanyos, M., 1996. The influence of anaerobic pretreatment on the nitrogen removal from biosynthetic pharmaceutical wastewaters. Antonie Van Leeuvenhoek 69, 41e46. Jeong, S.H., Song, Y.K., Cho, J.H., 2009. Risk assessment of ciprofloxacin, flavomycin, olaquindox and colistin sulfate based on microbiological impact on human gut biota. Regul. Toxicol. Pharmacol. 53, 209e216. Jetten, M.S.M., Cirpus, I., Kartal, B., van Niftrik, L., van de PasSchoonen, K.T., Sliekers, O., Haaijer, S., van der Star, W., Schmid, M., van de Vossenberg, J., Schmidt, I., Harhangi, H.,
210
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 0 1 e2 1 0
van Loosdrecht, M., Kuenen, J.G., Op den Camp, H., Strous, M., 2005. 1994e2004: 10 years of research on the anaerobic oxidation of ammonium. Biochem. Soc. Trans. 33, 119e123. Jiao, S.J., Zheng, S.R., Yin, D.Q., Wang, L.H., Chen, L.Y., 2008. Aqueous photolysis of tetracycline and toxicity of photolytic products to luminescent bacteria. Chemosphere 73, 377e382. Jordan, F.T., Knight, D., 1984. The minimum inhibitory concentration of kitasamycin, tylosin and tiamulin for Mycoplasma gallisepticum and their protective effect on infected chicks. Avian Pathol. 13, 151e162. Joss, A., Salzgeber, D., Eugster, J., Ko¨nig, R., Rottermann, P., Burger, S., Fabijan, P., Leumann, S., Mohn, J., Siegrist, H., 2009. Full-scale nitrogen removal from digester liquid with partial nitritation and Anammox in one SBR. Environ. Sci. Technol. 43, 5301e5306. Lindsay, M.R., Webb, R.I., Strous, M., Jetten, M.S.M., Butler, M.K., Forde, R.J., Fuerst, J.A., 2001. Cell compartmentalization in planctomycetes: novel types of structural organization for the bacterial cell. Arch. Microbiol. 175, 413e429. Ma, W.L., Qi, R., Zhang, Y., Wang, J., Liang, C.Z., Yang, M., 2009. Performance of a successive hydrolysis, denitrification and nitrification system for simultaneous removal of COD and nitrogen from terramycin production wastewater. Biochem. Eng. J. 45, 30e34. Mulder, A., van de Graaf, A.A., Robertson, L.A., Kuenen, J.G., 1995. Anaerobic ammonium oxidation discovered in a denitrifying fluidized bed reactor. FEMS Microbiol. Ecol. 16, 177e184. Naddeo, V., Meric, S., Kassinos, D., Belgiorno, V., Guida, M., 2009. Fate of pharmaceuticals in contaminated urban wastewater effluent under ultrasonic irradiation. Water Res. 43, 4019e4027. Okuda, T., Yamashita, N., Tanaka, H., Matsukawa, H., Tanabe, K., 2009. Development of extraction method of pharmaceuticals and their occurrences found in Japanese wastewater treatment plants. Environ. Int. 35, 815e820. Peng, Y.Z., Li, Y.Z., Peng, C.Y., Wang, S.Y., 2004. Nitrogen removal from pharmaceutical manufacturing wastewater with high concentration of ammonia and free ammonia via partial nitrification and denitrification. Water Sci. Technol. 50 (6), 31e36. Quinn, B., Gagne´, F., Blaise, C., 2009. Evaluation of the acute, chronic and teratogenic effects of a mixture of eleven pharmaceuticals on the cnidarians, Hydra attenuate. Sci. Total Environ. 407, 1072e1079. Sanz, J.L., Rodrı´guez, N., Amils, R., 1996. The action of antibiotics on the anaerobic digestion process. Appl. Microbiol. Biotechnol. 46, 587e592. Sheng, G.P., Zhang, M.L., Yu, H.Q., 2006. Characterization of adsorption properties of extracellular polymeric substances (EPS) extracted from sludge. Colloids Surf. B. Biointerfaces 62, 83e90. Sirtori, C., Zapata, A., Oller, I., Gernjak, W., Agu¨era, A., Malato, S., 2009. Decontamination industrial pharmaceutical wastewater by combining solar photo-Fenton and biological treatment. Water Res. 43, 661e668. Sliekers, A.O., Third, K.A., Abma, W., Kuenen, J.G., Jetten, M.S.M., 2003. CANON and anammox in a gas-lift reactor. FEMS Microbiol. Lett. 218, 339e344.
Stalter, D., Magdeburg, A., Weil, M., Knacker, T., Oehlmann, J., 2009. Toxication or detoxication? In vivo toxicity assessment of ozonation as advanced wastewater treatment with the rainbow trout. Water Res. 44, 439e449. Strous, M., 2000. Microbiology of anaerobic ammonium oxidation. Ph.D. Thesis, TU Delft. Strous, M., Heijnen, J.J., Kuenen, J.G., Jetten, M.S.M., 1998. The sequencing batch reactor as a powerful tool to study very slowly growing micro-organisms. Appl. Microbiol. Biotechnol. 50, 589e596. Tang, C.J., Zheng, P., Hu, B.L., Chen, J.W., Wang, C.H., 2010a. Influence of substrates on nitrogen removal performance and microbiology of anaerobic ammonium oxidation by operating two UASB reactors fed with different substrate levels. J.Hazard. Mater. 181, 19e26. Tang, C.J., Zheng, P., Mahmood, Q., Chen, J.W., 2009. Start-up and inhibition analysis of the Anammox process seeded with anaerobic granular sludge. J. Ind. Microbiol. Biotechnol. 36, 1093e1100. Tang, C.J., Zheng, P., Wang, C.H., Mahmood, Q., 2010b. Suppression of anaerobic ammonium oxidizers under high organic content in high-rate Anammox UASB reactor. Bioresour. Technol. 101, 1762e1768. Tang, C.J., Zheng, P., Wang, C.H., Mahmood, Q., Zhang, J.Q., Chen, X.G., Zhang, L., Chen, J.W., 2010c. Performance of highloaded ANAMMOX UASB reactors containing granular sludge. Water Res.. doi:10.1016/j.watres.2010.08.018. Tang, C.J., Zheng, P., Zhang, L., Chen, J.W., Mahmood, Q., Chen, X. G., Hu, B.L., Wang, C.H., Yu, Y., 2010d. Enrichment features of anammox consortia from methanogenic granules loaded with high organic and methanol contents. Chemosphere 79, 613e619. Tsushima, I., Ogasawara, Y., Kindaichi, T., Okabe, S., 2007. Development of high-rate anaerobic ammonium-oxidizing (anammox) biofilm reactors. Water Res. 41, 1623e1634. van de Graaf, A.A., De Bruijn, P., Robertson, L.A., Jetten, M.S.M., Kuenen, J.G., 1996. Autotrophic growth of anaerobic ammonium-oxidizing microorganisms in a fluidized bed reactor. Microbiology 142, 2187e2196. van der Star, W.R.L., Abma, W.R., Bolmmers, D., Mulder, J., Tokutomi, T., Strous, M., Picioreanu, C., van Loosdrecht, M.C. M., 2007. Startup of reactors for anoxic ammonium oxidation: experiences from the first full-scale Anammox reactor in Rotterdam. Water Res. 41, 4149e4163. van der Star, W.R.L., Miclea, A.I., van Dongen, U.G.J.M., Muyzer, G., Picioreanu, C., van Loosdrecht, M.C.M., 2008. The membrane bioreactor: a novel tool to grow Anammox bacteria as free cells. Biotechnol. Bioeng. 101, 286e294. Wang, L.K., Hung, Y.T., Lo, H.H., Yapijakis, C., 2005. Waste Treatment in the Process Industries. CRC Press: Taylor & Francis Group. Wett, B., 2006. Sovled upscaling problems for implementing deammonification of rejection water. Water Sci. Technol. 53 (12), 121e128. Wu, J., Zhou, H.M., Li, H.Z., Zhang, P.C., Jiang, J., 2009. Impacts of hydrodynamic shear force on nucleation of flocculent sludge in anaerobic reactor. Water Res. 43, 3029e3036.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 1 1 e2 2 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Long term laboratory column experiments to simulate bank filtration: Factors controlling removal of sulfamethoxazole Benno Baumgarten a,*, Jeannette Ja¨hrig a, Thorsten Reemtsma b, Martin Jekel a a
Chair of Water Quality Control, Department of Environmental Engineering, Technische Universita¨t Berlin e Berlin Institute of Technology, Sekr. KF 4, Strasse des 17 Juni 135, 10623 Berlin, Germany b Federal Institute for Risk Assessment, Safety in the Food Chain, Thielallee 88-92, 14195 Berlin, Germany
article info
abstract
Article history:
Microbial removal of the poorly degradable antibiotic sulfamethoxazole (SMX) from surface
Received 9 March 2010
water was investigated in laboratory columns to identify critical factors for SMX removal
Received in revised form
during bank filtration, such as the substrate concentration, redox conditions and the
5 August 2010
availability of biodegradable DOC. About 60% of SMX at a start concentration of 0.25 mg/L in
Accepted 19 August 2010
surface water were removed within 14 d of column passage under aerobic conditions while
Available online 27 August 2010
no removal occurred under anoxic conditions. The adaptation time was very long and was not completed after 2 years of operation. Adaptation was faster and SMX degradation was
Keywords:
improved at an elevated concentration of SMX (4.5 mg/L) with 90% removal in 3.5 d under
Sulfamethoxazole
aerobic conditions. SMX removal was less effective under anoxic conditions (27% in 14 d)
Sulfonamides
but increased again under anaerobic conditions (51% in 14 d). According to the half-lives for
Antibiotics
SMX determined from the column data (1e9 d aerobic, 49 d anoxic and 16 d anaerobic) it is
Biodegradation
essential to provide several weeks up to months of travel time in bank filtration to allow for
Surface water
the degradation of SMX, and likely, also for other poorly degradable compounds. Thus, the
Groundwater
occurrence of SMX in groundwater samples does not indicate persistency of SMX but
Bank filtration
reflects insufficient residence time or unfavorable respective redox conditions. Adaptation
Trace pollutants
times of years may also be required for new bank filtration sites to develop their full removal potential towards trace pollutants. Long operation time, a comparable concentration level and similar redox conditions as in the field appear to be essential to obtain realistic results with laboratory column experiments that can be transferred to real bank filtration sites. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Bank filtration and aquifer recharge systems are important measures for a safe drinking water supply in constant good quality (Kuehn and Mueller, 2000). Provided that the hydrogeological setting is adequate, the natural purification processes taking place during subsurface travel times from weeks to months (as usual at European bank filtration sites)
lead to the elimination of biodegradable dissolved organic carbon (BDOC), pathogens, and many trace pollutants (Eckert and Irmscher, 2006; Kuehn and Mueller, 2000; Weiss et al., 2003). These elimination processes during subsoil passage, filtration, sorption and biodegradation, are of particular importance, if water cycles are partially closed, holding the risk of micropollutants’ migration from wastewater to waters used for drinking water production. The purification potential
* Corresponding author. Tel.: þ49 30 31424281; fax: þ49 30 31479621. E-mail address:
[email protected] (B. Baumgarten). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.034
212
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 1 1 e2 2 0
of bank filtration has been shown to be strongly dependent on site specific conditions and may vary seasonally (Eckert and Irmscher, 2006). The sulfonamide sulfamethoxazole (SMX) is found in considerable concentrations in wastewater treatment plant effluents (100e400 ng/L (Hirsch et al., 1999; Jekel and Gru¨nheid, 2007)) and also in surface waters (4e480 ng/L (Christian et al., 2003; Heberer et al., 2008; Hirsch et al., 1999; Jekel and Gru¨nheid, 2007; Mompelat et al., 2009; Watkinson et al., 2009)). SMX is also a critical trace pollutant in bank filtration due to its enhanced persistency compared to other antibiotics (Halling-Sorensen et al., 1998; Jjemba, 2002). Correspondingly, SMX has been detected in groundwaters (Heberer et al., 2008; Hirsch et al., 1999; Sacher et al., 2003) as well as in bank filtrate samples (BLAC, 2003, Jekel and Gru¨nheid, 2008; Karthikeyan and Meyer, 2006; Schmidt et al., 2004). Due to the inherent biological activity of antibiotics such as SMX their occurrence in waters used for drinking water production is highly undesirable. Moreover SMX may be considered a marker indicating the possible breakthrough of wastewater constituents into raw waters used for drinking water production. Based on field data it was suggested that SMX is better degradable under strictly anaerobic (Schmidt et al., 2004) and under anoxic (Gru¨nheid et al., 2005; Heberer et al., 2008; Jekel and Gru¨nheid, 2007) conditions, than under aerobic ones. On the contrary in laboratory columns studies SMX was more effectively removed under aerobic conditions (Jekel and Gru¨nheid, 2008). It was further hypothesized that SMX degradation would only take place if a threshold concentration of 0.3 mg/L was exceeded (Gru¨nheid et al., 2007). Thus, contradictory results have been reported from studies at field sites and from laboratory studies. Generally, the complexity of natural systems and the multitude of factors influencing removal processes in the field make it highly unlikely to link certain effects observed in the field to one wellfounded parameter. Laboratory studies are preferred for this purpose because experimental conditions can be much better controlled. However, lab studies may result in erroneous or irrelevant results if designed or operated inappropriately. This study uses laboratory columns to investigate the effect of several parameters that were supposed to influence SMX degradation in bank filtration. Among them are redox conditions, the amount and quality of dissolved biodegradable organic matter and the initial concentration of SMX. Contrary to many other lab studies it therefore covered also low concentrations of SMX, such as those found in surface waters. It was intended to resolve the contradictory results regarding SMX degradation that have been gathered in studies in field site and in the laboratory. Beyond that it was also intended to check under which conditions laboratory column systems are capable of simulating a natural system as complex as bank filtration.
2.
Materials and methods
2.1.
Chemicals
Sulfamethoxazole (SMX), with a molar mass of 253.3 g/mol, pKa of 1.69e1.8 and 5.6, and a log Kow of 0.89 (Drillia et al., 2005; Gao and Pedersen, 2005) was purchased from Dr. Ehrenstorfer
GmbH (Augsburg, Germany). Deuterated SMX (SMX-D4) was received from Toronto Chemicals (North York, ON, Canada) and methanol (MeOH) as well as formic acid (HPLC-grade) from E. Merck GmbH (Darmstadt, Germany). Starch was provided by SigmaeAldrich Laborchemikalien GmbH (Seelze, Germany). Ultrapure water was produced by a water purification system ELGA maxima (High Wycombe Bucks, U.K.).
2.2.
Column studies
Microbial degradation of sulfamethoxazole was investigated in a system of sand columns in laboratory scale (Fig. 1). Eight columns (filter bed height 2 m, diameter 0.155 m) were filled with technical quartz sand of a grain size of 0.7e1.2 mm. The column material had a grain density of 2507 kg/m3 and was filled in the columns with a voidage of 0.4 as found for Lake Tegel sediment at the natural bank filtration site. Sampling outlets were situated in distances of 0.5 m over the column height. The columns were operated with surface water of Lake Tegel at a filter velocity of 0.13 m/d, resulting in a hydraulic retention time of 14 days. The whole system was placed in a cooling chamber at 11 1 C, corresponding to natural aquifer temperature (Massmann and Su¨ltenfuss, 2008). LakeTegel is a larger lake of the Havel river system in Berlin, Germany, and used for drinking water production via bank filtration. Since the Lake is affected by inflow of treated secondary effluent, substances originating from wastewater can be found in it. Dissolved oxygen and pH were measured online at several sampling points along the columns. For further off-line analyses, samples of 0.2 L were taken at the influent, every 0.5 m along the columns and at the column effluent. The samples were taken very slowly to avoid disturbance of the column flow
Fig. 1 e Setup of soil column test system. A1,2 and D1,2 were operated as parallel columns during the whole study. Columns B and C were operated in parallel from month 1e11 (not shown), from month 12 the effluent of column C feed into column B. Below: Structural formula of SMX.
213
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 1 1 e2 2 0
and aeration of the columns. Each sample was filtered through 0.45 mm membrane filters (cellulose nitrate filters for SMX- and DOC-analyses, cellulose acetate for all other analyses). Different redox conditions were established in the column system (Table 1). Half of the sand columns were operated at aerobic conditions with water of Lake Tegel stored in an open tank. The other four columns were fed with the same water, which was degassed by nitrogen and stored in a closed tank. This led to a remaining oxygen concentration of 1.2 mg/L in the column influent, which was consumed already within the first centimeters of the columns, resulting in anoxic conditions in the column (dissolved oxygen concentration of 0 mg/L after 0.5 m of sand passage). To create an even lower redox potential (anaerobic), column F (Fig. 1) was additionally spiked with a concentration of 1 mg/L of starch as easily available carbon source, leading to a faster decrease of dissolved oxygen and of oxo-anions such as nitrate and sulfate. Accordingly, the nitrate concentration was diminished from 1.5 0.4 to 0.0 0.5 mg/L in this column. Two aerobic (SMX0 ¼ 3.9 1.1 mg/L and 3.0 0.7 mg/L respectively), one anoxic (SMX0 ¼ 4.2 1.2 mg/L) and the anaerobic (SMX0 ¼ 3.9 1.3 mg/L) columns were spiked with SMX, while the other four columns received SMX at the concentration level of the Lake Tegel surface water (0.25 0.08 mg/L).
2.2.1.
Influence of BDOC on SMX removal
To examine whether certain biodegradable DOC (BDOC) fractions influence the transformation of SMX an experimental setup was chosen in which the effluent of one aerobic column (C in Tables 1 and 2) was used as influent for a second aerobic column (B in Tables 1 and 2). Between both columns a second spiking was performed, increasing the SMX concentration to 2.95 0.8 mg/L. This experiment was carried out over 14 months. To provide identical microbial development when starting the experiment, both columns (B and C) had previously been operated in parallel with SMX spiking for eleven months.
2.3.
Batch biodegradation experiments
To examine the behavior of SMX and DOC, batch biodegradation tests were performed at 11 1 C in the dark in glass bottles (volume 1 L) filled with 350 mL surface water of Lake Tegel and
350 g of sand from an infiltration site close to the lake, where lakewater is infiltrated for groundwater enrichment. This sand was expected to provide biomass adapted to SMX degradation. The water was either used directly (initial SMX concentration 0.23 mg/L) or after spiking (SMX concentration 0.48, 1, 3 and 10 mg/L). Tests were carried out for 60 d (aerobic) and 75 d (anoxic). 26 aerobic batch tests were stored open in contact with the air. The same number of anoxic batches was kept airtight under nitrogen with septum caps. For comparison batches with 3 mg/L SMX in surface water and pristine (technical) sand (without adapted biomass) were examined. Mixing was performed once a day by a gentle horizontally rotation for 180 , to avoid destruction of biofilms at the sand surface. The extent of sorption was determined by sacrificing a few bottles 3 h after filling.
2.4.
Analytical procedures
2.4.1.
Sulfamethoxazole
Trace analysis of SMX was carried out via HPLC-MS/MS subsequent to a solid phase extraction (SPE) according to the methods proposed by Hartig et al. (1999), Reemtsma and Quintana (2006) and Ternes (2001). SPE was conducted with an AutoTrace SPE Workstation (Zymark, Idstein, Germany). 100 mL of 0.45 mm filtered water samples were extracted at a flow rate of 4 mL/min in cartridges (Bakerbond EN, 200 mg, 6 mL, Mallinckrodt Baker, Phillipsburg, NJ, U.S.A.) conditioned two times with 5 mL of methanol and subsequently two times with 5 mL of ultrapure water. After rinsed with 10 mL of ultrapure water at a flow rate of 20 mL/min the cartridges were dried for 30 min with N2. SMX was eluted with 2 times 2 mL of MeOH. The 4 mL extract was evaporated to dryness at 45 C in a Savant SpeedVac concentrator (GMI, Ramsey, Minnesota/USA) and redissolved in 0.25 mL of MeOH. Before transferring the extract to a HPLC-vial the extract was filled up with ultrapure water to a volume of 0.5 mL. Chromatography was performed with a HP 1100 (Agilent, Bo¨blingen, Germany) HPLC system, using a C18 column (3 mm, 50 2 mm Luna) and a pre column (C18, 4 2.0 mm, both Phenomenex, Aschaffenburg, Germany). Separation was performed at 30 C at a flow rate of 0.25 mL/min using a water (A)/MeOH (B) gradient, each adjusted with 0.5% (v/v) formic
Table 1 e Column experiments: filter parameters and operating conditions for columns influent. spiked [ addition of SMX, natural [ inherent SMX concentration in surface water. Aerobic A1,2 natural* DOC0 [mg/L] SMX0 [mg/L] added starch [mg/L] nitrate [mg/L] pH O2 [mg/L]
7.6 06
a
0.25 0.08f e
Anoxic
B spiked b
6.3 0.3 w/o biopolymers 2.95 0.8g e 2.2 0.3h 8 0.4h 10.6 0.6h (89e100% saturation)
C spiked 7.8 0.7
c
3.93 1.1f e
D 1,2 natural* 7.4 0.7
a
anaerobic E spiked 8.1 0.5
d
0.25 0.08f 4.15 1.2f e e 1.8 0.3h 8.3 0.4h 1.2 0.4h (9e14% saturation)
*A and D: two parallel columns each. aeh medians mean absolute deviation: an ¼ 24 bn ¼ 8 cn ¼ 12 dn ¼ 10 en ¼ 13 fn ¼ 48 gn ¼ 16 hn ¼ 36.
F spiked 8.6 0.7e incl. starch 3.9 1.3c 1.0 1.5 0.4e 8.3 0.3e 0.4 0.4e (0e9% sat.)
214
3.9 1.3c 1.45 0.57c 51 11c 8.6 0.7e 7.7 0.3e 105e 4.15 1.2f 2.83 0.85a 27 11a 8.1 05d 7.6 0.6d 83d 0.27 0.25 11 6.4 5.9 8 0.30 0.12 57 7.3 6.3 13 0.08 0.04 4 0.2 0.1 1 0.36 0.19 46 7.4 6.5 12 0.24 0.19 19 7.6 6.5 13 0.25 0.21 13 8.4 7.6 9
0.02 0.08 33 0.1 0.1 1
0.03 0.04 8 0.1 0.2 2
22 14 9
operation time [months] SMX0 [mg/L] SMXeffluent [mg/L] SMX removal [%] DOC0 [mg/L] DOCeffluent [mg/L] DOC removal [%]
A1,2 natural*
*A and D: two parallel columns each (arithmetic means standard deviation) of 4 samples in month 9, 14 and 22 and of 2 samples in month 27. aei medians mean absolute deviation: an ¼ 27 bn ¼ 8 cn ¼ 12 dn ¼ 10 en ¼ 13 fn ¼ 48 gn ¼ 16 hn ¼ 36. removal ranges for column C and F: month 3e23, column B: month 3e14, column E: month 3e16.
0.01 0.01 2 0.1 0.2 2 3.93 1.1f 0.23 0.06h 951h 7.8 0.7c 6.4 0.7c 134c 2.95 0.7g 0.24 0.08g 955g 6.3 0.3b 5.9 0.4b 62b
0.08 0.02 6 0.2 0.3 2
3e23 3e14 27
B spiked (w/o biopolymers)
C spiked
0.25 0.24 4 8.5 8.0 6
0.24 0.24 0 8.0 7.2 10
14 9
0.01 0.02 1 0.2 0.1 1
0.37 0.36 2 6.9 6.4 8
22
0.06 0.06 15 0.1 0.3 3
27
0.03 0.03 8 0.4 0.9 3
3e23 3e16
F spiked D1,2 natural*
anoxic aerobic
Table 2 e SMX and DOC removal in the column experiments. spiked [ addition of SMX, natural [ inherent SMX concentration in surface water.
E spiked
anaerobic
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 1 1 e2 2 0
acid to pH 2.4 and 3.9, respectively. Elution started with 35% B, increasing to 60% B within 5 min, back to 35% B in 1 min, followed by 4 min equilibration with 35% B. The injection volume was 20 mL. Detection was performed by a Quattro LC triple-stage quadrupole mass spectrometer (Micromass, Manchester, U.K.) with Z-spray interface equipped with the electrospray probe operated in the positive ion mode. Drying and nebulizer gas (890 L h1 and 85 L h1) was nitrogen produced by a nitrogen generator (Whatman, Maidstone, U.K.). Argon (purity 5.0, Messer-Griesheim, Siegen, Germany) was used as collision gas at a pressure of 1.2$103 mbar in the collision cell. Source block temperature was 120 C and the MS was operated with a desolvation temperature of 280 C and an adjusted capillary voltage of 3.2 kV. Sulfamethoxazole was determined by two MRM-transitions, m/z 254 > 156 and m/z 254 > 108. The internal standard SMX-D4 was recorded by m/z 258 > 160 and m/z 258 > 112. External calibration was performed with standards prepared in ultrapure water and recovery was corrected via a deuterated internal standard of SMX-D4 that was added in a concentration of 125 ng/L prior to solid phase extraction. A recovery between 80 and 110% and a limit of quantification (S/N ¼ 10) of 25 ng/L SMX was obtained.
2.4.2.
SEC-OCD
Advanced NOM characterization was performed using SECOCD (size exclusion chromatography e organic carbon detection, DOC-Labor Dr. Huber, Karlsruhe, Germany) following the method of Huber and Frimmel (1996). The analyzer was equipped with a size exclusion column (Toyopearl Gel HW 50 S; 2 25 cm; separation size 105 g mol1, Tosoh Bioscience, Stuttgart, Germany). After SEC separation organic matter in the column effluent was transformed to CO2 using a thin film oxidation reactor with subsequent IR detection of CO2 (WellChrom K 200 UV-detector, Knauer, Berlin, Germany). Depending on molecular size the fractions of the SEC-OCD chromatogram can be classified into biopolymers, humic substances, building blocks, low molecular acids and amphiphilics (Huber and Frimmel, 1996). PEG (polyethylene glycol) standards were used for quantification of each fraction.
2.4.3.
Further measurements
Chemical conditions in the sand columns and batch tests were monitored by the following parameters: dissolved organic carbon (DOC) was analyzed in triplicates by thermocatalytic oxidation using a highTOCII analyzer (elementar Analysensysteme, Hanau, Germany). Online determination of the oxygen content was done optically via optical fibres equipped with an oxygen sensitive layer (Fibox 3, Presens, Regensburg, Germany). Off-line oxygen determination was performed electrochemically by an Oxi 340i (WTW, Weilheim, Germany).
3.
Results and discussion
3.1.
Column studies
The removal of SMX in lab scale columns was investigated over a period of 27 months on columns filled with sand and
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 1 1 e2 2 0
215
Fig. 2 e Relative removal of (a) SMX and (b) DOC in laboratory columns SMX0 [ 0.25 ± 0.08 mg/L, travel time 14 d (arithmetic means with standard deviation of two parallel columns (A1,2 and D1,2 in Table 2)).
operated with surface water with a residence time of 14 d at a temperature of 11 C. Several operational parameters of the column system were varied to investigate their influence on SMX removal in subsurface environment.
3.1.1. Removal of ng/L concentrations of SMX under different redox conditions In columns fed with surface water with its inherent SMX concentration of 0.25 0.08 mg/L no significant removal was observable within the first nine months of operation (Fig. 2). Then, SMX removal became effective and increased to about 57% after 27 months of operation, corresponding to concentrations in the column effluents of 0.12 0.02 mg/L (Table 2). It is not clear from this time course (Fig. 2) whether SMX removal in the aerobic columns may have improved further at longer operation times or whether it had reached its maximum, already. Under anoxic conditions, however, no significant removal of the inherent SMX concentration was observed over the whole column operation time of 27 months (Fig. 2) and effluent concentrations remained at a level of 0.25 0.03 mg/L (Table 2). The 57% removal of SMX achieved on the aerobic column corresponds to the extent of removal in a real bank filtration site (46e64%; Jekel and Gru¨nheid, 2007). Concerning redox conditions, however, previous field studies disagree with the results of this column study, as they suggested that SMX was removed more effectively under anoxic (97%) and anaerobic than under aerobic (46e64%) conditions (Jekel and Gru¨nheid, 2007; Schmidt et al., 2004). The determination of redox conditions in field studies is, however, not an easy task and the integrated conditions determined by chemical bulk parameters may not agree with the actual conditions in micro-environments. Moreover, redox conditions in bank filtration sites have been shown to be astonishingly variable over time (Massmann et al., 2006). Therefore the effect of combinations of anoxic and aerobic soil passages is currently under examination. Obviously, the column system required a remarkably long time of more than two years to be capable of removing SMX effectively. Two factors may account for an increase in SMX removal: (a) general growth of microbial biomass or (b) adaptation of an
existing biomass to SMX, either by the growth of specialists or the development of enzyme systems capable of degrading SMX. Without detailed investigation of the biomass (amount of biomass, microbial activity and community structure) in a given system it is not always obvious which of these factors was most important. As these experiments started with pristine sand, the initial increase in SMX removal in the aerobic columns can be attributed to biomass growth, as reflected in the increasing DOC removal (Fig. 2b). In a second phase (month 14e27, Table 2) DOC removal was relatively stable and, thus, the continuing increase of SMX removal in this phase may be due to ongoing adaptation of the existing biomass to degrade SMX. These experiments suggest that differences between laboratory studies and field studies may occur if laboratory studies are not operated for sufficient time. Hence, very long operation times may be required in laboratory column studies to adequately simulate the removal potential of field sites used for bank filtration. However, such long adaptation times would also have to be considered if new sites for bank filtration or for infiltration are put into operation to become fully effective in trace pollutant removal.
3.1.2. Removal of mg/L concentrations of SMX under different redox conditions Removal on columns operated with elevated SMX concentration of 3.9 1.1 mg/L of SMX was significantly different. In the aerobic columns a constant 95% removal was observable after 3 months of operation, already (Figure SI 1). This led to effluent concentrations of 0.23 0.06 mg/L (Table 2). At the elevated SMX concentration even the anoxic column and the anaerobic column showed a significant SMX removal of 27 and 51% after 3 months of operation, corresponding to effluent concentrations of 2.8 0.85 mg/L and 1.5 0.57 mg/L, respectively (Table 2). The more effective removal of microgram per litre concentrations of SMX under aerobic than under anoxic conditions in this column study agrees to a previous column study, where 95% of SMX were removed under aerobic and 60% under anoxic conditions (Jekel and Gru¨nheid, 2007). This column study further suggests that SMX removal, while it decreases from oxic to anoxic conditions, recovers when the redox potential decreases further to anaerobic
216
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 1 1 e2 2 0
conditions. Therefore one has to clearly differentiate between different reducing redox milieus. A comparison with the low concentration data (Section 3.1.1) implies that elevated SMX concentrations foster the establishment of biomass capable of SMX degradation: the adaptation time is shorter and the relative removal is more extensive than at low concentrations. At these elevated concentrations even the less favourable anoxic and anaerobic conditions allow for some SMX removal. Based on the data of an earlier column study it had been hypothesized, that a threshold concentration would exist below which no transformation of SMX would take place (Gru¨nheid et al., 2007). The results presented here make it likely that this was a matter of insufficient adaptation time rather than of a true threshold concentration.
3.1.3.
Removal kinetics and half-lives
The kinetics of SMX transformation during percolation through the columns could be followed by determining the remaining SMX concentration at different heights of the columns, i.e., at different residence times in the column. The concentrations of SMX along the columns fed with the elevated SMX concentration (3.9 1.1 mg/L SMX) after 14 month of operation are displayed in Fig. 3. In the aerobic column SMX removal occurred quite fast with 90% removal in a travel time of 3.5 d (first 0.5 m); in the following 10 d, however, no further removal occurred. The reason for the lack of removal in deeper zones of the column is not clear. It is, however, not due to a lack of oxygen, as 4 mg/L were still present in a depth of 0.5 m. In the columns operated with low SMX concentration a similar profile was found, with a rapid removal within the first 3.5 d (Fig. 4c). However the concentration did not decrease below 40% of the influent concentration in that experiment.
Fig. 3 e Concentration of SMX (median with mean absolute deviation) along the laboratory columns under aerobic (C in tables, month 3e23, n [ 18), anoxic (E in tables, month 3e16, n [ 6) and anaerobic (F in tables, month 3e23, n [ 8) conditions (SMX0 [ 3.93e4.11 mg/L). Further displayed: aerobic column fed with surface water DOC without biopolymer fraction (SMX0 [ 2.95 mg/L, in month 1e3 (n [ 3) and month 6e14 (n [ 9), B in tables) in contrast to aerobic columns fed with complete DOC.
Both, the onset of SMX degradation as well as the speed of removal was slower in the anoxic and the anaerobic column (Fig. 3). In the anoxic column removal is slow along the whole column. It is not clear if SMX removal would continue on longer columns, i.e., at extended residence times. In the anaerobic column SMX removal started after a travel time of 7 d (1 m, Fig. 3). This suggests that the consumption of the starch added into the influent during the first meter of the anaerobic column leads to a further decrease of the redox potential, which is essential for a more effective SMX transformation. It remains to be investigated in detail, which redox potential is achieved at which depth in the anaerobic column and which milieu was favourable for SMX degradation. These height-resolved kinetic data support the removal data given in Table 2. These results outline that it is essential in laboratory column studies not only to determine the influent and effluent concentrations of such columns but also to follow the time course of degradation along the column height. Likely, the laboratory columns do not only exhibit a gradient in SMX concentration. Rather the amount and composition of the DOC (see below) and of the microbial biomass, the redox conditions and further parameters may vary along the column height. Thus, kinetic parameters gathered from the height/time-resolved concentration data of the columns can only be apparent parameters and would necessarily differ from parameters derived under controlled conditions (if such conditions exist). In this study the column data were used to calculate apparent half-lives (t½) for SMX for the different experiments. After 27 months of operation time a t½ of 9 d is calculated via a first order kinetics approach for columns with inherent SMX concentration under aerobic conditions at 11 C. The modelling was verified by comparison to the experimental data, to which the model fits well with an R2 of 0.9 (more detailed information on kinetic modelling in SI). At the elevated SMX level (c0 ¼ 3.9 1.1 mg/L) t½ was 3 d only, calculated with the same kinetic approach and a R2 of 0.54. A modelling according to a second order kinetics was performed to improve the agreement with the experimental data. Using this approach an even shorter half-life of 1 d with an R2 of 0.85 was obtained for the elevated SMX concentration. For the less favourable anoxic conditions a first order kinetic was appropriate and resulted in a t½ of 49 d (R2 ¼ 0.61) for the elevated SMX concentration. Under these conditions a column length of 8 m would be required for 50% removal. Anaerobic conditions led to an apparent half-life of 16 d (R2 ¼ 0.78), modelled with first order kinetic approach. The half-lives data can be used to estimate the required travel time to obtain a desired removal (e.g., 90%, 99%) of SMX in bank filtration, provided that the kinetic data gathered from the laboratory column experiment can be transferred to field sites. For the case of SMX it is obvious that sufficient adaptation time must have been given to the column system, and that the concentration level of the substrate as well as the redox conditions should reflect the conditions of the respective site to allow transfer of kinetic data. An apparent half-life of 9 d for SMX at low concentrations under aerobic conditions highlights that several weeks of residence time in an aquifer are required for complete removal of this trace pollutant. This example clearly illustrates why
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 1 1 e2 2 0
217
long travel times are preferred in bank filtration, if removal of trace pollutants is intended. Even longer times are required if redox conditions in the subsurface are not optimal, as with anoxic instead of aerobic conditions for SMX.
3.1.4.
Degradation of DOC and its influence on SMX removal
Since DOC is an important bulk parameter for every water matrix, it may be instructive to compare removal of DOC and of SMX in the column experiments (Table 2). Indeed a quite similar trend as for SMX was found for DOC, with a more effective removal under aerobic conditions (9e13%) as compared to anoxic conditions (6e10%) over the whole experimental period of 27 months. A gradual improvement of DOC removal over time in the aerobic columns was visible in the first 9 months of operation time (Table 2, Fig. 2b), reflecting the growth of microbial biomass in the columns. From then on DOC removal remained stable while SMX removal increased continuously (Fig. 2). This suggests that the established microorganisms continued to adapt to SMX degradation. Not only the amount but also the composition of the DOC changed during the column passage, as visible from the SECOCD chromatograms (Fig. 4). The biopolymer fraction is by far the fastest degraded fraction of the DOC of the surface water. About 90% of it (0.42 0.08 mg/L) was removed aerobically (Fig. 4a), thus contributing to approx. 23% of the total DOC removal. Under anoxic conditions only 40% of this fraction was removed (Fig. 4b). Also the humic substances (around 55 min) were removed more effectively under aerobic conditions (Fig. 4a,b). Aerobic and anoxic columns also differ in the removal kinetics of DOC (Fig. 4a,b). While the biopolymer fraction was removed within 3.5 d (first 0.5 m) under aerobic conditions, its removal did not start until day 7 (1 m) of column passage and remained much slower under anoxic conditions. Apparently microbial degradation of the biopolymer fraction of the DOC was delayed and slowed down under anoxic conditions. This effect has also been shown for SMX (Section 3.1.1). Based on the parallel behaviour of SMX and biopolymers in some of the columns (Fig. 4c) it was assumed that SMX transformation may be linked to the degradation of DOC. It could be possible that BDOC may be required to establish and maintain sufficient microbial activity in the column system to enable measurable transformation of SMX. In bank filtration this BDOC would be provided from particulate organic matter of the colmation zone. The improved removal of trace pollutants in the case that BDOC is available would agree to results of another recently published column study (RauchWilliams et al., 2010). Whether or not BDOC supports SMX removal was examined in a column experiment, in which the effluent of one aerobic column (C in Tables 1 and 2) was spiked with SMX and used as influent for a second aerobic column (B in Tables 1 and 2). Both columns had already been operated for eleven months at spiked concentration of SMX (Section 3.1.2). In this experiment the influent to column B contained only 10% of the initial biopolymer fraction (<0.05 mg/L) and a total DOC of 6.3 0.3 mg/L. Of this DOC only 0.4 mg/L could be removed during passage of column B as compared to about 1.4 mg/L in the parallel column C fed with fresh lakewater (Table 2).
Fig. 4 e SEC-OCD chromatograms for the influent and effluent of the laboratory columns (after 3.5, 7 and 14 days of travel time) under (a) aerobic and (b) anoxic conditions (SMX0 [ 0.25 mg/L (A and B in Tables)). Extension: biopolymer fraction at retention time 30e48 min. (c) Concentration of biopolymer fraction (BP, n [ 4) and SMX (n [ 8 aerobic, n [ 3 anoxic) along the laboratory columns (months 23e27, SMX0 [ 0.25 ± 0.08 mg/L) in aerobic (columns A1,2) and anoxic (columns D1,2) non-spiked columns.
During the first three months of operation of column B the SMX removal under aerobic conditions was significantly slower than on column C with the fresh lakewater (Fig. 3). In this phase it took 14 d of sand passage to obtain 90% SMX removal, while 3.5 d were required on column C, only (Table 2, Fig. 3). Thus, the lack of BDOC appeared to have slowed down the biodegradation of SMX.
218
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 1 1 e2 2 0
Fig. 5 e Removal of SMX in biodegradation batch tests. (a) aerobic (60 days test duration) and anoxic (75 days) (arithmetic means of two samples), all tests with sand from the infiltration site and water from Lake Tegel, numbers indicate the initial SMX concentration in mg/L (inherent conc. [ surface water concentration; 3h-sample to examine adsorption effects in a batch spiked with 3 mg/L SMX). (b) Comparison of pristine sand and sand from the infiltration site, SMX0 [ 3 mg/L.
Over 14 months of operating time, however, SMX removal in column B recovered again and finally reached the same velocity as in column C (Fig. 3). Obviously the biomass adapted to the low BDOC conditions and its SMX removal also recovered. Also this finding is in line with results of Rauch-Williams et al. (2010), who have suggested that heterotrophic biomass would account for the effective removal of trace pollutants at very low BDOC concentrations.
3.2.
Batch experiments
In comparison to column test systems batch experiments require a much lower experimental effort and can be performed with a stable water quality. Moreover more parallel experiments can be operated as in the column system. For these reasons it would be attractive to use batch experiments (with water and sand) rather than column systems for assessing the extent of biodegradation of trace pollutants that may occur during infiltration of surface water.
3.2.1.
Influence of redox conditions
Water from Lake Tegel was spiked with different concentrations of SMX (Fig. 5) in four parallel batches for each concentration. Two aerobic batches, spiked with 3 mg/L SMX were analyzed after 3 h, showing that less than 10% of the initial SMX was removed by adsorption onto the particulate material (Fig. 5). This limited sorption is in agreement with the findings of Yu et al. (2009) for soils. Under aerobic conditions in batches with sand from the infiltration site SMX was transformed by more than 89% at 0.23 mg/L initial concentration within 60 days and a concentration near the LOQ of 25 ng/L (but above the LOD) was
reached. For increasing start concentration a relative removal of up to 99% at 3 and 10 mg/L SMX was found (Fig. 5). In the batch spiked with 3 mg/L SMX 85% were removed and 0.45 mg/L were still found after 30 days, showing that SMX transformation did not occur instantaneously (data not shown). As in the column system removal of SMX under anoxic conditions (75 d) was less complete than under aerobic conditions (Fig. 5): only 66% removal were recorded at a start concentration of 0.23 mg/L. In this respect the simple batch tests show the same trends as the more demanding column test systems. Under anoxic conditions, however, the relative removal decreased with increasing start concentration down to 34% at 10 mg/L, which does not agree to the trend observed on the columns. For this batch adaptation or degradation time may have been incomplete.
3.2.2.
Growth and adaptation
Growth and adaptation phenomena were also investigated in batch tests. For this purpose pristine ‘technical’ sand was used in some batches parallel to sand from the infiltration site, where surface water had previously been infiltrated for groundwater enrichment. It was expected that the sand from this site carried substantial microbial biomass which should be adapted to the degradation of trace pollutants found in that surface water, such as SMX. Indeed, degradation in the batch tests over 60e75 d was faster with the sand of the infiltration site for both, aerobic and anoxic conditions (Fig. 5b). With this material 98% removal were obtained under aerobic conditions and 67% under anoxic conditions as compared to 23% and 3% with the pristine sand. This finding of the batch test confirms the importance of sufficient time for bacterial growth and/or microbial adaptation for an effective transformation of poorly degradable trace pollutants such as SMX. The long phase of more than two years required to obtain full elimination potential in the columns (Section 3.1.1, Figs 2 and 3) could be shortened if material from an existing infiltration or bank filtration site was used as filling material, provided that operational conditions of the field site are selected.
4.
Conclusions
C The operation of a laboratory column system to simulate SMX removal in bank filtration with surface water clearly showed that SMX is more effectively degraded under aerobic than under anoxic conditions. At SMX concentration found in surface water of around 0.25 mg/ L, it took about a year of adaptation for degradation to start and removal was still increasing during the second year of operation with a column effluent concentration of 0.12 mg/L after a travel time of 14 d. C At elevated concentrations of SMX (4 mg/L) adaptation proceeded faster (3e12 months) and relative removal was more effective (95%). At this concentration level SMX was removed even under anoxic conditions (27%). Removal increased again under anaerobic conditions (51%).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 1 1 e2 2 0
C Apparent half-lives of SMX under aerobic conditions calculated from the height-resolved column data ranged from 1 to 9 d for high and low concentrations, respectively. Under anaerobic and anoxic conditions apparent SMX half-lives for high concentrations increased to 16 and 49 d respectively; no removal occurred under anoxic conditions at the low concentration level. C The availability of a few mg/L of BDOC supports microbial activity in the subsurface environment and fosters SMX removal. But with sufficient adaptation SMX is also removed if BDOC is lower. However, a fluctuating quality of surface waters (in terms of BDOC) can affect the extent of removal of trace pollutants like SMX. C The results of this study outline that bank filtration under aerobic conditions should be suitable to remove SMX to levels below 0.12 mg/L. It is essential to provide several weeks up to months of travel time in bank filtration to allow for the degradation of SMX, and likely, also for other poorly degradable compounds. C The occurrence of SMX in groundwaters does not indicate a general persistency of SMX but reflects insufficient residence time or unfavorable redox conditions. C Adaptation times of years may also be required for new bank filtration sites to develop their full removal potential towards trace pollutants. C Because the concentration of SMX and the redox conditions strongly influenced the extent of removal in the laboratory column experiments it is essential to select these conditions appropriately to allow for the transfer of results from laboratory columns to real bank filtration sites. Namely, elevated substrate concentrations should be avoided.
Acknowledgements The authors are grateful to Berliner Wasserbetriebe (BWB) for the kind permission to take water and soil samples from Lake Tegel and the infiltration site, respectively. The constructive criticism of to anonymous reviewers is gratefully acknowledged.
Appendix. Supplementary material Supplementary data associated with this article can be found in online version, at doi:10.1016/j.watres.2010.08.034.
references
BLAC, 2003. Pharmaceuticals in the environment e Evaluation of the results. (Arzneimittel in der Umwelt e Auswertung der Untersuchungsergebnisse). Freie und Hansestadt Hamburg, Beho¨rde fu¨r Umwelt und Gesundheit, Umweltuntersuchungen im Auftrag des Bund/La¨nderausschuss fu¨r Chemikaliensicherheit (BLAC), Hamburg. Christian, T., Schneider, R.J., Farber, H.A., Skutlarek, D., Meyer, M.T. , Goldbach, H.E., 2003. Determination of antibiotic residues in
219
manure, soil, and surface waters. Acta Hydrochimica Et Hydrobiologica 31 (1), 36e44. Drillia, P., Stamatelatou, K., Lyberatos, G., 2005. Fate and mobility of pharmaceuticals in solid matrices. Chemosphere 60, 1034e1044. Eckert, P., Irmscher, R., 2006. Over 130 years of experience with Riverbank filtration in Du¨sseldorf, Germany. Journal of Water Supply: Research and Technology e Aqua 55 (4), 283e291. Gao, J.A., Pedersen, J.A., 2005. Adsorption of sulfonamide antimicrobial agents to clay minerals. Environmental Science & Technology 39 (24), 9509e9516. 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. Gru¨nheid, S., Hu¨bner, U., Jekel, M., 2007. Impact of temperature on biodegradation of bulk and trace organics during soil passage in an indirect reuse system. 6th Conference on Wastewater Reclamation and Reuse for Sustainability, 9e12, Oct, 2007, Antwerpen/BE. Halling-Sorensen, B., Nielsen, S.N., Lanzky, P.F., Ingerslev, F., Lutzhoft, H.C.H., Jorgensen, S.E., 1998. Occurrence, fate and effects of pharmaceutical substances in the environment e A review. Chemosphere 36 (2), 357e394. 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 (1e2), 163e173. Heberer, T., Massmann, G., Fanck, B., Taute, T., Dunnbier, U., 2008. Behaviour and redox sensitivity of antimicrobial residues during bank filtration. Chemosphere 73 (4), 451e460. Hirsch, R., Ternes, T., Haberer, K., Kratz, K.L., 1999. Occurrence of antibiotics in the aquatic environment. Science of the Total Environment 225 (1e2), 109e118. Huber, S., Frimmel, F., 1996. Size-exclusion chromatography with organic carbon detection (LC-OCD): a fast and reliable method for the characterization of hydrophilic organic matter in natural waters (in German). Vom Wasser 86, 277e290. Jekel, M., Gru¨nheid, S., 2007. Is bank filtration an effective barrier against organic substances and pharmaceutical residues?. (Ist die Uferfiltration eine effektive Barriere gegen organische Substanzen und Arzneimittelru¨cksta¨nde?) GWF Wasser Abwasser 148 (10), 698e703. Jekel, M., Gru¨nheid, S., 2008. In: Jiminez, B., Asano, T. (Eds.), Water Reuse. An International Survey of Current Practice, Issues and Needs. IWA Publishing, London, pp. 401e413. Jjemba, P.K., 2002. The potential impact of veterinary and human therapeutic agents in manure and biosolids on plants grown on arable land: a review. Agriculture Ecosystems & Environment 93 (1e3), 267e278. Karthikeyan, K.G., Meyer, M.T., 2006. Occurrence of antibiotics in wastewater treatment facilities in Wisconsin, USA. Science of the Total Environment 361 (1e3), 196e207. Kuehn, W., Mueller, U., 2000. Riverbank filtration e an overview. Journal American Water Works Association 92 (12), 60e69. 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., Su¨ltenfuss, J., 2008. Identification of processes affecting excess air formation during natural bank filtration and managed aquifer recharge. Journal of Hydrology 359 (3e4), 235e246. Mompelat, S., LeBot, B., Thomas, O., 2009. Occurrence and fate of pharmaceutical products and by-products, from resource to drinking water. Environment International 35 (5), 803e814.
220
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 1 1 e2 2 0
Rauch-Williams, T., Hoppe-Jones, C., Drewes, J.E., 2010. The role of organic matter in the removal of emerging trace organic chemicals during managed aquifer recharge. Water Research 44 (2), 449e460. Reemtsma, T., Quintana, J.B., 2006. In: Reemtsma, T., Jekel, M. (Eds.), Organic Pollutants in the Water Cycle. Wiley-VCH, Weilheim, pp. 1e63. Sacher, F., Gabriel, S., Metzinger, M., Wenz, M., 2003. Pharmaceuticals in ground and surface waters. Chimia 57 (1e2), 29e31. Schmidt, C.K., Lange, F.T., Brauch, H.-J., 2004. Assessing the Impact of Different Redox Conditions and Residence Times on the Fate of Organic Micropollutants during Riverbank Filtration. 4th International Conference on Pharmaceuticals and Endocrine Disrupting Chemicals in Water, 13e15, Oct, 2004, Minneapolis, Minnesota.
Ternes, T.A., 2001. Analytical methods for the determination of pharmaceuticals in aqueous environmental samples. TracTrends in Analytical Chemistry 20 (8), 419e434. Watkinson, A.J., Murby, E.J., Kolpin, D.W., Costanzo, S.D., 2009. The occurrence of antibiotics in an urban watershed: from wastewater to drinking water. Science of the Total Environment 407 (8), 2711e2723. Weiss, W.J., Bouwer, E.J., Ball, W.P., O’Melia, C.R., Lechevallier, M.W., Arora, H., Speth, T.F., 2003. Riverbank filtration - fate of DBP precursors and selected microorganisms. Journal American Water Works Association 95 (10), 68e81. Yu, L., Fink, G., Wintgens, T., Melin, T., Ternes, T.A., 2009. Sorption behavior of potential organic wastewater indicators with soils. Water Research 43 (4), 951e960.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 2 1 e2 3 1
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Screening of 47 organic microcontaminants in agricultural irrigation waters and their soil loading D. Caldero´n-Preciado a, C. Jime´nez-Cartagena a,b, V. Matamoros c, J.M. Bayona a,* a
Department of Environmental Chemistry, IDAEA-CSIC, Jordi Girona 18-26, E-08034 Barcelona, Spain Diagnosis and Pollution Control Group, Faculty of Engineering, Antioquia University, Medellı´n, Colombia c Department of Chemistry, University of Girona, Campus Montilivi, 17071 Girona, Catalonia, Spain b
article info
abstract
Article history:
Reclaimed water usage for crop irrigation is viewed both as an excellent sustainable water
Received 12 November 2009
source and as a potential entrance for emerging organics into the food chain. This concern
Received in revised form
is backed by the already documented pollutant crop uptake potential. In the present study,
6 July 2010
irrigation waters used in agricultural fields (Torroella de Montgri, NE Spain) were screened
Accepted 16 July 2010
for 47 analytes in a two year study (2007e2008). A total of 26 contaminants belonging to
Available online 27 July 2010
different chemical classes namely, pesticides, pharmaceuticals, personal care products, phenolic estrogens, antioxidants and disinfection by-products, were detected. Marked
Keywords:
differences in concentration trends for the different chemical classes were evidenced from
Reclaimed water
2007 to 2008, and attributed to a persistent drought endured by the region in 2008. Also,
River water
loading mass rates of chemical classes were estimated based on crop irrigation regimes
Pharmaceuticals and personal care
and they ranged from 0.8 to 121.3 g ha1 per crop cycle. These values were contrasted with
products
those obtained for other water sources from countries where crop irrigation is commonly
Disinfection by-products
practiced. Finally, crops grown under these irrigation regimes, namely alfalfa and apple,
Pesticides
were analyzed and 5 anthropogenic compounds were identified and quantitated, whose
Micropollutant soil loadings
concentrations ranged from 13.9 to 532 ng g1 (fresh weight). ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
It is becoming widely accepted that available freshwater resources are very limited and constitute a small fraction of the total freshwater budget (Oki and Kanae, 2006). Moreover, an increase in world population has raised the pressure on natural resources, exemplified by an escalating demand for domestic, industrial and agricultural water (Zimmerman et al., 2008). Furthermore, on a global basis, 70% of freshwater is currently destined to crop irrigation, 20% for industrial purposes and the remaining 10% to domestic use (Zimmerman et al., 2008). Nowadays, 28 countries with a total population exceeding 300 million habitants are faced with water scarcity (Rosegrant and
Cai, 2001). As world demands for water grow, water reclamation and reuse become increasingly important as an indispensable component of the integral water resource management, and are widely regarded as sustainable approaches in agricultural irrigation (Jimenez and Asano, 2008). This aspect is of special significance considering the importance of adequately sustaining agricultural activities in order to ensure food production (Rosegrant and Cai, 2001). In this regard, reclaimed wastewater usage is a sustainable crop irrigation alternative already in use in some countries such as USA, Israel, Australia, and Spain (Jimenez and Asano, 2008). Nevertheless, it is well known that one of the main micropollutant inputs to the hydrological cycle are the WWTP
* Corresponding author. Tel.: þ34 934006119; fax: þ34 932045904. E-mail addresses:
[email protected] (D. Caldero´n-Preciado),
[email protected] (C. Jime´nez-Cartagena), vmmqam@ cid.csic.es (V. Matamoros),
[email protected] (J.M. Bayona). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.07.050
222
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 2 1 e2 3 1
effluents. Indeed, these micropollutants do not need to be persistent to have an effect on ecosystems, since their medium to high removal rates are overshadowed by their continuous introduction into the environment (Petrovic et al., 2003). Many microcontaminant classes have been documented to occur in WWTP effluents including human and veterinary pharmaceuticals and personal care products (PPCPs), pesticides, phenolic estrogens, surfactants, dispersants, biocides and disinfection by-products (Kolpin et al., 2002). Though the potential health and environmental hazards derived from continued exposure to these chemicals are not well understood, the ecotoxicological effects that they could induce remain of great concern. Still estrogenic responses on aquatic organisms exerted by endocrine disruptors (Jobling et al., 1998; Fent et al., 2006), inhibition of multixenobiotic resistance in mussels caused by polycyclic musks (Luckenbach and Epel, 2005), the carcinogenicity and/or mutagenicity elicited by PAHs in humans, and the potential development of antibiotic resistance as a result of low concentration exposure to pharmaceuticals (Boxall et al., 2006) are well documented. These evidences illustrate that the occurrence of some microcontaminants, even at low concentrations, can lead to chronic health effects especially at highly contaminated sites, particularly when a possible synergistic effect of pharmaceutical mixtures could take place (Fent et al., 2006; Jjemba, 2008). In addition, the uptake of veterinary medicines (Migliore et al., 1998; Luckenbach and Epel, 2005; Kong et al., 2007), PAHs (Fismes et al., 2002; Samsoe-Petersen et al., 2002) and pesticides (Gonzalez et al., 2003; Gent et al., 2007) from soil into plants has also been documented but remains largely unknown for a variety of micropollutants (e.g. PPCPs). Therefore, the aim of this study was to survey the occurrence of 16 PPCPs, 15 pesticides, and 9 disinfection by-products, 3 phenolic estrogens, 2 flame retardants and 2 antioxidants, in reclaimed wastewater, river water and their mixture used for crop irrigation in an agricultural community (i.e. Torroella de Montgrı´, Girona, northeastern Spain). Moreover, Principal Component Analysis (PCA) was used to reduce the dimensionality of data set by explaining the correlation among the large number of variables analyzed. Then geographical and yearly trends for target analytes were obtained. Moreover, the loading mass rate of these micropollutants discharged into agricultural soil corresponding to different crops, namely maize, Zea mays, apple, Malus domestica, lettuce, Lactuca sativa and alfalfa, Medicago sativa, was estimated from the crop irrigation regimes and contaminant concentrations. Finally, alfalfa and apple crops grown in the agricultural community under the different irrigation regimes were analyzed for the target microcontaminants.
2.
Materials and methods
2.1.
Materials and reagents
All standards were analytical grade (97e99% purity). Reagent details are provided in Supplementary Information (S1.1). Stock solutions of each individual compound were prepared in methanol or ethyl acetate at a concentration of 5000 mg L1. All prepared standards were stored in the darkness at 20 C
and used to prepare single and mixed working standards solutions. THMs and nitrosamines working solutions were prepared daily from dilutions of stock solutions.
2.2.
Sampling site description
This survey was carried out in agricultural fields located at the Torroella de Montgri municipality, Girona, northeastern Spain. This community has a territorial extension of 6613 ha and due to a high tourism activity, during the summer season undergoes important fluctuations in population size. The agricultural fields subjected to our study have an approximate area of 3000 ha, with apple, corn and alfalfa crop predominance. Field irrigation is carried out through a piping network and water supply is mainly provided by the Ter River. This river (average flow rate at the mouth is 25 m3 s1) receives along its watershed, discharges from metallurgic, pulp mill, textile and tannery industries as well as raw sewage inputs from small adjacent communities. When the water supply needed for agriculture is not met by the river itself, reclaimed water from the local WWTP is injected into the network system, where it combines with riverine freshwater in an uncontrolled manner. Therefore, the reclaimed water use for agricultural irrigation is variable, mostly depending on weather conditions and usually carried out during the dry season (i.e., JuneeAugust). The Torroella de Montgri WWTP treats 2.3 106 m3 of wastewater per year, mainly from urban origin. The WWTP, total surface of 2.2 ha, was designed to treat the wastewater of 68 750 equivalent inhabitants with an inflow of 16 500 m3 d1. The secondary treatment is based on activated sludge, with the corresponding nitrogen removal, followed by disinfection through high intensity-low pressure UV lamps and chlorination as tertiary treatment. In order to evaluate irrigation water quality of the Ter River, the WWTP and their mixture, four sampling points were chosen, one in each input point in the network, namely the Ter River and WWTP, and two more sampling points within the network in order to evaluate the influence of both water origins (Fig. 1). These points were termed as Ter River Influence and WWTP Influence, respectively. Sampling campaigns were carried out in 2007 and 2008, three per year from May to September, and analyzed for the target compounds listed in Table 1.
2.3.
Sample collection
Water samples for THMs and nitrosamines determination were collected in 40 mL amber glass vials and 250 amber glass bottles, respectively containing ascorbic acid to remove residual chlorine in water samples, 100 mg for THMs and 250 mg for nitrosamines. Both vials and bottles were fully filled (headspace free). Water samples for the other micropollutant analysis were collected in 2.5 L glass amber bottles. All samples were kept refrigerated during transport to the laboratory, where they were stored at 4 C until they were analyzed. The total sample holding time in all cases was less than 72 h. Crop samples of alfalfa and apple tree leaves were collected from two irrigation points: Ter Influence and WWTP Influence during the first two campaigns of 2008. Samples were carefully
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 2 1 e2 3 1
223
Fig. 1 e Map of North-eastern Spain representing the sampling spots studied.
wrapped in aluminum foil placed in a plastic bag and stored at 20 C until their analysis.
2.4.
Analytical methodology
2.4.1.
Water analysis
2.4.1.1. THMs. Trihalomethane determination was carried out by solid-phase microextraction (SPME) using an SPME TRIPLUS autosampler from Thermo Fisher, the extraction conditions were adapted from a former report (Cho et al., 2003). Fiber thermal desorption and analysis was carried out in a GC Ultra Trace (Thermo Fisher, Milan, Italy) coupled to an electron capture detector (ECD). Details of the methodology carried out are further explained in the Supplementary Information (S1.2.1.1). The linearity range of the HSeSPME method was evaluated by performing calibration curves from the relative area of the internal standard iodomethane versus the concentration of each analyte. Standard calibration curves were plotted for twelve concentrations levels. Linearity was observed at the concentration range from 0.002 to 10 mgL1 (R2 > 0.99). The HSeSPME GCeECD method sensitivity was evaluated in terms of LODs and LOQs, which were determined from a blank triplicate (reagent water) using the mean background noise plus 3 and 10 times the standard deviation of the background noise, respectively. LOD and LOQ range from 0.001 to 0.350 and 0.002e0.474 mg L1 respectively. 2.4.1.2. N-Nitrosamines. N-nitrosamines extraction was adapted from the EPA 521 method (Munch and Basset, 2004). Their determination was performed in an Ultra Trace GC coupled to a nitrogen phosphorous detector (Thermo Fisher, Milan, Italy). Detailed extraction and determination procedures of these compounds are thoroughly detailed in the Supplementary Information (S1.2.1.2). Furthermore, compound identity and quantification confirmation were performed by GCePI MS by using ammonia as reagent gas (Agilent Technologies 5975C) as reported previously (Charrois et al., 2004). Quantitation was performed on the basis of the internal standard procedure. Method sensitivity was calculated as
reported in the Section 2.4.1. The LODs and LOQs obtained for n-nitrosoamines ranged respectively, from 0.049 to 0.108 and 0.050e0.110 mg L1. Linearity of the SPE GCeNPD method for the n-nitrosamine determination ranged from 0.0057 to 7.5 mg L1 with correlation coefficients higher than 0.99.
2.4.1.3. Pharmaceuticals and personal care products. Wastewater samples were filtered, processed and analyzed as reported previously (Matamoros and Bayona, 2006). Derivatized samples were analyzed in a TRACE GCeMS (Thermo Fisher) in the electron impact mode. Detailed extraction and determination procedures of these compounds are thoroughly detailed in the Supplementary Information (S.1.2.1.3). The LOD and LOQ of the analytical procedure were determined (using reagent water) from the mean background noise plus 3 or 10 times the standard deviation of the background noise, respectively. LOD and LOQ were from 0.002 to 0.28 and 0.003e0.47 mg L1, respectively. Recoveries ranged from 90 to 107%. 2.4.1.4. Suspended particulate matter. Filters were processed and analyzed as previously reported (Matamoros and Bayona, 2006). Further extraction and chromatographic analysis details are provided in Supplementary Information (S1.2.1.4).
2.4.2.
Crop analysis
Alfalfa and apple tree leaves were extracted as previously reported (Calderon-Preciado et al., 2009). Sample analysis was carried out by gas chromatography tandem mass spectrometry (GC/MS/MS), details are provided in Supplementary Information (S1.2.3). LOQ ranged from 0.011 to 0.099 mg Kg1.
2.5.
Statistical analysis
Principal component analysis (PCA) (Jollife, 2002; Pere-Trepat et al., 2006) was performed on the data set using the SPSS 13 package (Chicago, IL). For all compounds concentration values were obtained from the sum of concentrations in the dissolved and particulate phases. Trihalomethanes were included in the analysis as total trihalomethanes rather than in the individual species. Out of the 26 compounds quantitated only 18 were
224
Table 1 e Frequency of detection (FOD), minimum, maximum and mean concentration of the 47 analytes in the four sampling points. Target analyte
WWTP
WWTP Influence
Ter River Influence
Ter River
Conc. (mg/L) (minemax) mean
FOD (%)
Conc. (mg/L) (minemax) mean
FOD (%)
Conc. (mg/L) (minemax) mean
Disinfection by-products Chloroform Dichlorobromomethane Dibromochloromethane Bromoform Nitrosodimethylamine, NDMA Nitrosomethylethylamine, NMEA Nitrosopyrrolidine, NPYR Nitrosopiperidine, NPIP Nitrosodibutylamine, NDBA
4/4(100) 4/4(100) 4/4(100) 4/4(100) 2/2(100) 2/2(100) 2/2(100) 1/2(50) 0/2(0)
(0.67e6.16) 3.36 (3.25e6.09) 4.87 (0.15e14.24) 7.32 (1.08e23.60) 8.71 (0.097e0.101) 0.099 (0.18e0.52) 0.352 (0.10e0.17) 0.133 0.15 <0.056
1/5(20) 3/5(60) 3/5(60) 3/5(60) 2/2(100) 1/2(50) 2/2(100) 0/2(0) 0/2(0)
0.92 (0.04e0.79) 0.42 (0.04e0.06) 0.05 (0.06e0.08) 0.07 (0.12e0.14) 0.128 0.31 (0.188e0.190) 0.190 <0.0108 <0.056
1/4(25) 1/4(25) 3/4(75) 3/4(75) 2/2(100) 1/2(50) 2/2(100) 1/2(50) 0/2(0)
1.04 0.07 (0.01e0.15) (0.02e0.18) (0.08e0.21) 0.51 (0.11e0.18) 0.15 <0.056
Pharmaceuticals Diclofenac Carbamazepine Clofibric acid Caffeine Ibuprofen Flunixin Ketoprofen Acetaminophen Naproxen Irgasan
3/4(75) 3/4(75) 2/4(50) 4/4(100) 3/4(75) 0/4(0) 0/4(0) 0/4(0) 4/4(100) 0/4(0)
(0.072e0.171) (0.203e0.971) (0.118e0.132) (0.113e0.492) (0.017e0.259) <0.092 <0.071 <0.037 (0.035e0.139) <0.022
2/4(50) 4/4(100) 2/4(50) 4/4(100) 3/4(75) 0/4(0) 0/4(0) 0/4(0) 3/4(75) 0/4(0)
(0.134e0.506) 0.320 (0.058e0.160) 0.121 (0.121e0.180) 0.150 (0.083e2.380) 0.789 (0.017e0.245) 0.139 <0.092 <0.071 <0.037 (0.039e0.186) 0.092 <0.022
2/5(40) 5/5(100) 2/5(40) 5/5(100) 4/5(80) 0/4(0) 0/4(0) 0/4(0) 4/5(80) 0/4(0)
Personal care productsa Ambrettolide Methyldihydrojasmonate Galaxolide Tonalide Cashmeran Hidrocinnamic acid
0/4(0) 3/4(75) 4/4(100) 4/4(100) 2/4(50) 4/4(100)
<0.036 (0.073e0.971) (0.986e1.054) (0.143e0.585) (0.061e0.259) (0.028e3.502)
0/4(0) 4/4(100) 4/4(100) 4/4(100) 3/4(75) 0/4(0)
<0.036 (0.065e0.704) (0.159e0.438) (0.035e0.329) (0.218e0.399) <0.025
Phenolic estrogens Bisphenol A Tert-octylphenol Nonylphenol
2/4(50) 0/4(0) 0/4(0)
(0.072e0.171) 0.121 <0.097 <0.081
3/4(75) 0/4(0) 0/4(0)
Antioxidants BHT BHA
3/4(75) 0/4(0)
(0.086e0.621) 0.266 <0.124
Flame retardants Tributylphosphate TCEP
3/4(75) 2/4(50)
(0.126e0.216) 0.160 (0.361e0.376) 0.368
0.121 0.526 0.125 0.238 0.150
0.071
0.512 1.023 0.357 0.160 1.397
FOD (%)
Conc. (mg/L) (minemax) mean
2/4(50) 0/4(0) 0/4(0) 3/4(75) 2/2(100) 0/2(0) 2/2(100) 1/2(50) 0/2(0)
2.80 <0.0092 <0.001 (0.01e0.02) 0.015 (0.07e0.13) 0.102 <0.049 (0.18e0.27) 0.225 0.16 <0.056
(0.053e0.714) 0.383 (0.076e0.193) 0.136 (0.103e0.135) 0.119 (0.055e0.646) 0.295 (0.011e0.192) 0.083 <0.092 <0.071 <0.037 (0.036e0.250) 0.097 <0.022
4/4(100) 4/4(100) 2/4(50) 4/4(100) 4/4(100) 0/4(0) 0/4(0) 0/4(0) 3/4(75) 0/4(0)
(0.064e0.168) 0.113 (0.069e0.274) 0.145 (0.116e0.134) 0.125 (0.149e1.779) 0.791 (0.120e0.303) 0.228 <0.092 <0.071 <0.037 (0.100e0.444) 0.224 <0.022
0/4(0) 4/5(100) 5/5(100) 5/5(100) 4/5(80) 0/5(0)
<0.036 (0.065e0.704) (0.109e0.636) (0.026e0.339) (0.176e0.416) <0.025
0/4(0) 4/4(100) 4/4(100) 3/4(75) 3/4(75) 2/4(50)
<0.036 (0.085e1.027) (0.107e0.190) (0.027e0.176) (0.238e0.533) (0.052e0.137)
(0.037e0.089) 0.067 <0.097 <0.081
2/5(40) 0/4(0) 0/4(0)
(0.055e0.094) 0.075 <0.097 <0.081
3/4(75) 0/4(0) 0/4(0)
(0.069e0.1) 0.081 <0.097 <0.081
4/4(100) 0/4(0)
(0.079e0.525) 0.197 <0.124
4/5(80) 0/4(0)
(0.079e0.499) 0.188 <0.124
3/4(75) 0/4(0)
(0.080e0.738) 0.306 <0.124
2/4(50) 2/4(25)
(0.145e0.229) 0.187 0.284
2/5(40) 2/5(40)
(0.139e0.167) 0.153 (0.241e0.298) 0.269
2/4(50) 2/4(50)
(0.173e0.201) 0.187 (0.222e0.240) 0.231
0.374 0.307 0.152 0.291
0.08 0.10 0.148 0.142
0.374 0.309 0.138 0.283
0.422 0.149 0.119 0.341 0.094
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 2 1 e2 3 1
FOD (%)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 2 1 e2 3 1
<0.081 <0.29 <0.046 <0.031 <0.127 <0.018 < 0.128 <0.058 <0.058 <0.021 < 0.155 <0.002 0.447 (0.050e0.147) 0.098 (0.089e0.103) 0.096
included in the PCA since the inclusion of the remaining 8 compounds lead to failure of Bartlett’s sphericity test, invalidating PCA results. Data were arranged in a matrix in which samples were represented in rows, whereas target analytes (18) and electrical conductivity were in columns; samples were further arranged by sampling year in increasing order. However, concentrations below calculated quantification limits were replaced by convention by half of their limit of quantification (Pere-Trepat et al., 2006; Hildebrandt et al., 2008). Once the data matrix was completed, it was autoscaled to have zero mean and unit variance (correlation matrix) in order to avoid problems arising from different measurement scales and numerical ranges of the original variables (Helena et al., 2000; Hildebrandt et al., 2008). a Concentration of compounds within the Personal Care Product group is expressed as the sum of concentrations of dissolved and particulate phase.
0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 1/4(25) 2/4(50) 2/4(50) <0.081 <0.29 <0.046 <0.031 <0.127 <0.018 < 0.128 <0.058 <0.058 <0.021 < 0.155 <0.002 <0.14 0.059 (0.083e0.135) 0.109 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 1/5(20) 2/5(40) <0.081 <0.29 <0.046 <0.031 <0.127 <0.018 < 0.128 <0.058 <0.058 <0.021 <0.155 <0.002 <0.14 (0.082e0.161) 0.122 (0.117e0.148) 0.133 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 2/4(50) 2/4(50) <0.081 <0.29 <0.046 <0.031 <0.127 <0.018 < 0.128 <0.058 <0.058 <0.021 < 0.155 <0.002 <0.14 0.091 0.186 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 0/4(0) 1/4(25) 1/4(25) Pesticides Dimetoate Chlorpyrifos Atrazine Alochlor Simazine Diclobenil Linuron Lindane Deltamethrin Chlorothalonil Cypermethrin Diuron MCPA Diazinon Mecoprop
225
3.
Results and discussion
3.1. Occurrence of organic micropollutants in irrigation waters Table 1 shows the target analytes of this study, their concentration levels and detection frequencies. These compounds were selected on the basis of their concentration and a high detection frequency in river waters and WWTP effluents (Kolpin et al., 2002; Martı´nez Bueno et al., 2007; Gervais et al., 2008). Overall frequency of detection of target analytes in sampling points is as follows, Ter River (75%) > WWTP (66%) > WWTP influence (62%) > Ter River influence (57%). Even though Ter River exhibits the largest overall frequency of detection among sampling points, it is the WWTP, which displays the largest target analyte concentrations. The overall mean contaminant concentration per sampling points is as follows: WWTP (30.4 mg L1) > Ter River (7.5 mg L1) > WWTP influence (5.8 mg L1) > Ter River influence (5.3 mg L1). Pharmaceuticals were equally detected in WWTP and Ter River with a 56% of frequency of detection (FOD). The concentration ranges of the pharmaceuticals were found to be between 0.017 and 0.971 mg L1 in the WWTP effluent and between 0.064 and 1.779 mg L1 in the Ter River. Fragrances were more frequently detected in the WWTP (FOD ¼ 71%), the concentration ranges for these compounds were found to be between 0.061 and 3.5 mg L1. Out of the 15 pesticides included in the study, only 3 were identified and quantitated, namely mecoprop, MCPA and diazinon. These compounds were detected in the Ter River with a FOD of 8%. Pesticide detection in this sampling point might suggest a direct influence from Ter River since the WWTP exhibited the lowest FOD, 3% for these compounds. The concentration ranges for pesticides in the WWTP influence and Ter River were found to be between 0.082 and 0.161 mg L1 and between 0.050 and 0.147 mg L1, respectively, all values corresponding to diazinon. In the case of other micropollutants, namely trischloroethylphosphate, tributyl phosphate, bisphenol A, and BHT, were equally detected in the WWTP effluents and Ter River with a FOD of 63%. Concentrations of these compounds ranged from 0.072 to 0.621 mg L1 and 0.069e0.738 mg L1, for WWTP effluent and Ter River respectively. THMs were always detected in the WWTP with a 100% of FOD and these compounds displayed concentrations from 11.9 to 46.4 mg L1.
226
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 2 1 e2 3 1
It is worth mentioning, that these concentrations are well below the European limit for total trihalomethanes in drinking water (100 mg L1) effective from January 2009. NDMA and NPYR were the most frequently detected nitrosamines with an FOD of 100% in all sampling sites. Concentration ranges for N-nitrosamines ranged from 0.07 to 0.52 mg L1. Since higher concentrations of these compounds were not found in the WWTP effluent other sources of these compounds are expected (Chen and Valentine, 2007). As it has been explained in Section 2.2, Ter River, is a highly contaminated river receiving discharges from industries which make use of chemical reactions involving alkylamines with nitrogen oxides, nitrous acid or nitrile salts are known to produce and release nitrosamines (EPA, 2008), thus adding to these compounds occurrence and high levels in this area. However WWTP effluent concentration levels are in agreement with levels reported for NDMA, 0.1 mg L1 in other WWTP effluents. It is rather interesting that compounds which exhibit high removal efficiencies in the WWTP, namely ibuprofen, caffeine, methyl dihydrojasmonate (MDHJ), galaxolide and tonalide (i.e., 65e90%) (Petrovic et al., 2003; Chen et al., 2007; Carballa et al., 2008) are the most frequently detected along the irrigation network attributable to untreated sewage discharge into the Ter River. On the other hand, some organic pollutants may occur in WWTP effluents because they are persistent in the activated sludge process or because their biodegradation is not fast enough to be completed during wastewater treatment (Reemtsma et al., 2006). In this regard, the high detection frequency of carbamazepine, naproxen and diclofenac can be explained by their moderately low removal efficiency in a WWTP (Clara et al., 2005; Reemtsma et al., 2006).
3.2. Spatial and inter-annual variability of organic micropollutants in irrigation waters The climatological conditions of the region of study differed greatly from 2007 to 2008. During FebruaryeJuly, 2008, a drought decree was enforced and almost exclusively reclaimed water from the local WWTP to irrigate crops was allowed (Generalitat de Catalunya, 2008). The increase in reclaimed water share throughout the irrigation network is clearly evidenced from the measured electrical conductivity (Table 2) in all the sampling points. Moreover, it is also evident a marked difference between sampling years. In 2007, an almost exclusive influence of the
Ter River can be seen, added by the inactivity of the tertiary treatment of the local WWTP in the last 2007 campaign. In Fig. 2, the interannual and spatial variability of compound concentration are shown. Abundance pattern for the studied compound groups differed from year to year, this difference was attributable to the dominating influence of reclaimed water within the irrigation network due to the 2008 drought. For fragrances, an increase in all sampling points is observed during 2008, the same occurs with flame retardants, this trend is easily explained, since both fragrances and flame retardants are most abundant in the WWTP effluent. Conversely, BHT concentration decreased in all the sampling points, this was expected since BHT is most abundant in the Ter River. Bisphenol A showed no significant interannual differences owing to its similar concentration in both Ter River and WWTP.
3.3.
PCA analysis
The Bartlett’s sphericity test showed a c2 ¼ 318.9, which is greater that the critical value c2 ¼ 146.6 (for 120 degrees of freedom and p ¼ 0.05), thus proving that PCA can achieve a significant reduction of the original data set dimensionality.
3.4.
PCA loadings
The contribution of each variable to every principal component, PC, is shown in Fig. 3a, b and c. Moreover, the explained variance by component is also included. The first five PCs, which explain the 86.2% of the variance contained in the original data set and had an eigenvalue greater than one, are displayed in Table 3. The absolute value of the loadings is an indicator of the participation of the original variables in every PC. However when a complex system such as this one is studied, it is hard to identify the underlying variables in the final PCs. Therefore, we have only retained the first three which altogether explained 67% of the total variance. Fig. 3a shows the variable loadings to each of the three PCs retained. The first PC accounted for the 28% of the total original data variance, and it groups compounds according to sampling year (frequency of detection and abundance). The positive loadings belong to the 2008 year, while the negative ones belong to the 2007. This was clearly explained by the analyte occurrence. Clofibric acid and tonalide dominated the positive loadings of this component, clofibric acid was only detected in the sampling campaigns carried out in 2008 while tonalide mean concentration was two fold that of 2007. The
Table 2 e - Sampling campaigns and their respective electric conductivities in the collected water samples. Campaign Date
WWTPa (mS cm1)
WWTPInfluence (mS cm1)
Ter River (mS cm1)
TerRiver Influence (mS cm1)
July, 2007 September, 2007 September, 2007 May, 2008 July, 2008 Aug, 2008
2820 1151 NS 2587 1455 2188
720 733 720 1666 419 1352
709 736 740 995 414 688
718 730 742 2094 487 734
NS: Not sampled. a The high electrical conductivity observed by WWTP is attributable to saline intrusion in coastal areas covered by Torroella de Montgri WWTP.
227
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 2 1 e2 3 1
1.6
Compound group concentration, g L-1
Table 3 e Explained variance and eigenvalue for each principal component.
Pharmaceuticals Fragances Pesticides Flame retardants BHT Bisphenol A
1.4 1.2
Component
1.0
Explained Variance (%)
Accumulated Variance (%)
Eigenvalue
28.0 21.6 17.2 12.2 7.1
28.0 49.6 66.8 79.1 86.2
4.5 3.5 66.8 79.1 89.2
1 2 3 4 5
0.8 0.6 0.4 0.2
20 08 in flu W en W ce TP 20 07 in flu Te en rR c e2 iv er 00 in 8 flu Te e rR nc e2 iv er 00 in 7 flu en ce 20 Te 08 rR iv er 20 07 Te rR iv er 20 08
W W TP
W W TP
W W TP
20 07
0.0
Fig. 2 e Inter-annual variablity of target compound groups in the different sampling points within the irrigation network.
latter behavior was shared by diclofenac, galaxolide, TBP, ibuprofen and cashmeran. Opposite behavior was displayed by bisphenol A, naproxen and BHT, THMs and carbamazepine, whose mean concentration in 2007 in respect to 2008 was always higher. However, MDHJ and caffeine, in spite of having larger concentrations in 2007 and 2008 respectively, had an inverse loading ratio. Electrical conductivity is positively correlated to the 2008 sampling campaign, which can be easily explained by the large WWTP effluent usage in the irrigation
PC1: 2 8.00%
1.0 0.5 0.0 -0.5 -1.0
PC2: 21.61%
1.0 0.5 0.0 -0.5 -1.0
PC3: 17.23%
1.0
net due to a severe drought endured by the region during the irrigation period of 2008. It is interesting to note that though THMs presence is highly correlated to the WWTP effluent and thus to high electrical conductivity, they appear inversely correlated in PC1. The latter can be explained by higher THMs abundance during 2007, when there was low or no influence of the WWTP effluent on the irrigation network. The second PC accounted for 21.6% of the total variance (Fig. 3b), and it groups compounds according to pollution origin (embodied by abundance). In this case, as explained in the sampling section, the only two water inputs, and thus pollutant input to the irrigation network are the Ter River and the WWTP effluent. The positive loadings in PC2 are related to WWTP effluent while the negative ones are related to Ter River. Accordingly, THMs, carbamazepine and galaxolide, which had the highest weight on PC2, exhibited higher concentrations in the WWTP effluent than in any other sampling point. Electrical conductivity again was as expected correlated positively to the WWTP effluent. Variables with negative loadings in this case such as caffeine, ibuprofen, diazinon, clofibric acid, and cashmeran exhibited their highest concentrations in the Ter River sampling point. Interestingly, compounds that exhibit their highest concentration at the Ter River sampling point are also compounds that are easily degraded in a conventional WWTP, such is the case of caffeine, ibuprofen, cashmeran and diazinon (Buerge et al., 2003; Joss et al., 2005; Gros et al., 2007). The third PC accounted for 17.2% of the total variance of the original data set (Fig. 3c) and it appears to group compounds according to their association to the suspended particulate matter. The variables that appear negatively correlated are the ones that exhibit a direct relation with suspended particulate matter. One of these variables is electrical conductivity, which accordingly displays a larger effect in the negative loadings. The fragrances cashmeran, tonalide and galaxolide, which exhibit large log Kow (4.5e5.9) (Paasivirta et al., 2002), are also strongly related to presence of particulate matter.
0.5
3.4.1.
0.0 -0.5
D i azi non
Cashmeran
MDHJ
Ibuprofen
TBP
Caffeine
Galaxolide
THMs
Carbamazepine
Conductivity
BHT
Diclofenac
Naproxen
Bisphenol A
Tonalide
Clofibric acid
-1.0
Fig. 3 e Loadings for chosen principal components by straight PCA.
PCA scores
Fig. 4a depicts the score plot for PC1 vs. PC2. In this Figure, two large clusters formed along PC1 are rather evident, separating data according to abundance and detection frequency of target analytes during 2007 and 2008 campaigns. On the other hand, the variables that had the highest negative weight in this component are grouped at the left side of PC1 axis, representing the 2007 sampling year. Both clusters spread alongside PC2 axis, managing a separation of pollution source. In the upper quadrants for both years, the WWTP sampling
228
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 2 1 e2 3 1
Fig. 4 e (a) Scores plot for PC1 vs. PC2 for complete data. Clusters show the temporal trend between 2007 (in red) and 2008 (in blue). (b) Scores plot for PC2 vs. PC3. Clusters depict pollution origin (water source).
data appear which is in good accordance with the PC2 loadings, characterized by the positive contribution of variables associated to the WWTP effluent. Consequently, in the lower quadrant for PC2 appear these variables associated to their high abundance in Ter River. Thus sampling data which approach to the PC2 axis origin belongs to the least polluted sampling sites, whereas those points distant from the origin are the most polluted. As expected, the large majority of the target analytes in this study were most abundant in the WWTP tertiary effluent, shown in the PC2 loadings, making this sampling point the most polluted. Fig. 4b displays the score plot for PC2 vs. PC3. Samples are distributed along PC2 in two main clusters indicating, as previously explained, pollution origin. It is evident a significant difference of WWTP samples from the rest. The two data clusters formed spread along the PC3 axis. In the lower quadrants, sampling points which are loaded with compounds related to high particulate matter content appear, we find mostly WWTP and its influence. Consequently WWTP sampling point displays the highest concentrations for galaxolide, tonalide and diclofenac, which are the variables that have the largest influence on the PC3 loadings.
3.5.
Loading mass estimation into agricultural field
Compound load to the studied agricultural fields was calculated from irrigation regimes used on corn, (Zea mays) apple, (Malus domestica) alfalfa, (Medicago sativa) and lettuce (Lactuca sativa) giving an estimate mass input ranging from 0.8 to 121.3 g ha1 per crop cycle, depending on crop irrigation regimes (Table 4). For comparative purposes the latter estimate was also carried out employing pollutant concentrations found in other countries to produce a theoretical scenario in which these waters would be used for agricultural irrigation. In Table 4, it can be seen that the WWTP effluent of our case study was the most suitable for two major compound groups,
namely, pharmaceuticals and flame retardants, this translated in lower pollutant load to agricultural fields. In the case of THMs, concentrations found were comparable to the Israeli WWTP. However in the case of PCPs, its use resulted in almost 3 fold the input to fields when compared to the German WWTP, and 35 fold in the case of nitrosamines when compared to the USA WWTP. Subsequently, when evaluating the suitability of rivers for water irrigation, we could contrast only three compound groups: pharmaceuticals, PCPs and flame retardants. The Ter River displayed the highest mass loading rate for pharmaceuticals and PCPs, among the rivers compared. Nevertheless, it displayed the lowest input for flame retardants: 2.6 fold that of Ho¨je River.
3.6. Occurrence of personal care products and pharmaceuticals in studied crops Non-volatile compounds which displayed higher concentrations and frequency of detection in the irrigation waters analyzed were later selected as target analytes in the screening of crops. The selected compounds were diclofenac, carbamazepine, clofibric acid, caffeine, ibuprofen, naproxen, triclosan, MDHJ, galaxolide, tonalide and hydrocinnamic acid. Table 5 shows the analyzed PPCPs in the sampled crops and their concentration levels. Out of the 11 analytes screened, 6 were detected in alfalfa and 4 in apple tree leaves. However, two of the compounds identified are naturally occurring in plants; salicylic acid found in both crops is present in plants as a signaling molecule in the activation of defence response to pathogen infection and other environmental stresses (Durner et al., 1997), and hydrocinnamic acid, found in apple tree leaves, is known to have a function as a growth inhibitor (Xuan et al., 2009). Caffeine and MDHJ were found in both crops, concentration ranges for these compounds were found to be between <0.011 and 0.016 and 0.041 and 0.532 mg Kg1, respectively, with caffeine being higher in apple leaves while
229
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 2 1 e2 3 1
Table 4 e Estimated loading rates per chemical classes into agricultural field based on the net irrigation regime per crop cycle and water origin. This study, Crop (g ha1)
THMs Nitrosamines Pharmaceuticals PCPs Flame retardants
River WWTP River WWTP River WWTP River WWTP River WWTP
Other countries, Crop (g ha1)
Alfalfaa
Appleb
Cornc
Lettuced
Alfalfaa
Appleb
Cornc
Lettuced
11.0 91.0 1.8 2.5 5.1 7.6 3.5 8.6 0.8 1.2
12.0 103.1 2.3 3.2 6.7 8.7 4.6 9.8 1.0 1.4
14.1 121.3 2.7 3.8 7.9 10.2 5.3 11.5 1.2 1.6
7.05 60.65 1.4 1.9 3.9 5.1 2.7 5.8 0.6 0.8
NA 90.7e NA 0.07f 1.3e2.4g,h 15.5e96.4g,i 0.4e0.5g,h 3.1e3.3g,i 2.1h NA
NA 102.8e NA 0.08f 1.4e2.8g,h 17.6e109g,i 0.4e0.6g,h 3.5e3.7g,i 2.4 h NA
NA 121.0e NA 0.09f 1.7e3.2g,h 20.7e128.5g,i 0.48e0.68g,h 4.15e4.4g,i 2.8h NA
NA 60.5e NA 0.04f 0.8e1.6g,h 10.3e64.3g,i 0.2e0.3g,h 2.1e2.2g,i 1.4h NA
NA: not available. a 375 L m2 (1.5 months). b 425 L m2 (8 months). c 500 L m2 (5 months). d 250 L m2 (3 months). e Israel: Richardson et al. (2003). f Switzerland: Krauss and Hollender (2008). g Romania: Moldovan et al. (2007). h Sweden: Bendz et al. (2005). i Germany: Ternes et al. (2007).
MDHJ was ostensibly higher in alfalfa. In addition to MDHJ, another fragrance was detected and quantified, though only in alfalfa: galaxolide. Interestingly pharmaceuticals were only detected in alfalfa, ibuprofen and naproxen levels ranged from <0.011 to 0.061 mg Kg1. All the studied compounds, with the exception of galaxolide, exhibit log Kow between 0.16 and 3.79, which are within the range reported for possible plant uptake, 1e4 (Briggs et al., 1983). The higher levels and occurrence of anthropogenic compounds in alfalfa can be explained by the amount of irrigation water used in each crop, that being 4 times higher for alfalfa than for apple (Table 4). This behavior could link plant uptake or adsorption to compound amount loaded into the soils. In addition to the latter factor
a much shorter productive cycle, only 1.5 months for alfalfa versus 8 months for apple cultivars which would mean a higher growth rate for alfalfa, would also be reflected in a larger contaminant uptake from the medium. The occurrence of pharmaceuticals and personal care products in agricultural fields irrigated with reclaimed water has already been documented (Pedersen et al., 2005). Moreover, they display a wide range of log Kow and half-lives and for many of the analytes found in this study, these properties may indicate potential crop uptake. Indeed, uptake and translocation of some pharmaceuticals, in addition to the results presented here, have already been described (Migliore et al., 1998; Luckenbach and Epel, 2005; Kong et al., 2007; Redshaw et al., 2008).
Table 5 e Occurrence of PPCPs in irrigated crops with reclaimed (WWTP) and river waters (Ter). Target analyte
Apple tree leaves a
Hydrocinnamic acid Carbamazepine Salicylic acidc Caffeine Ibuprofen Triclosan Naproxen MethylDHJasmonate Tonalide Diclofenac Galaxolide
Alfalfa
WWTP Inf Conc. (mg Kg1)
Ter Inf Conc. (mg Kg1)
WWTP Inf Conc. (mg Kg1)
Ter Inf b Conc. (mg Kg1) (min-max) mean
0.076 <0.011 0.778 0.016 <0.012 <0.011 <0.011 0.041 <0.011 <0.099 <0.015
35.5 <0.011 0.527 15.5 <0.012 <0.011 <0.011 0.041 <0.011 <0.099 <0.015
<0.011 <0.011 0.698 <10.6 0.032 <0.011 <0.011 0.156 <0.011 <0.099 16.9
<0.011 <0.011 (1.709e2.583) 2.146 13.9 (0.025e0.061) 0.043 <0.011 0.014 0.532 <0.011 <0.099 <0.015
Concentrations were calculated in fresh weight. a Reclaimed wastewater influence. b Ter River influence. c Quantitation performed by GCeMS.
b
a
230
4.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 2 1 e2 3 1
Conclusions
26 out of 47 organic compounds were identified and quantitated in a two years study of reclaimed and river water (i.e. Ter) used for agricultural irrigation in the NE Spain. Although the highest overall frequency of detection of target analytes in the study area was found to be in the Ter River, the highest pollutant concentrations were exhibited by the WWTP effluent. Furthermore, as a result of the 2008 drought endured by the region, reclaimed water usage for field irrigation increased markedly, which in turn lead to an increase of pollutant load into agricultural fields. Moreover, compounds which display high removal efficiency in conventional WWTPs (i.e., ibuprofen, caffeine, MDHJ, galaxolide and tonalide) are most abundant in river water than in WWTP effluent. In order to explain underlying factors in compound levels and distribution PCA analysis was applied, and three major factors were identified namely, temporal variability, pollutant origin and compound association with dissolved particulate matter. Moreover, potential mass loading rate of target analytes into agricultural fields according to irrigation regimes was estimated and found to be in the range of g ha1 per crop cycle. The estimated contaminant load as a result of Ter River and Torroella de Montgri WWTP effluent usage in agricultural irrigation was contrasted with other water sources obtaining comparable mass loading rates. Finally, crops grown under these irrigation regimes were analyzed and 5 contaminants were identified and quantitated, namely ibuprofen, naproxen, MDHJ, caffeine and tonalide. A link between compound load into soil and their subsequent translocation or adsorption into crops was found to be plausible, since the studied crop with higher irrigation rate (alfalfa) also exhibited higher occurrence and concentration levels of the target analytes. Since the potential health and environmental hazards derived from continued exposure to these chemicals are not well understood, it becomes of paramount importance the pollutant fate into agricultural fields and the experimental evaluation of their crop uptake potential.
Acknowledgments This research was funded by the Age`ncia Catalana de Seguretat Alimentaria (ACSA) of Catalan Government (Generalitat de Catalunya). M. Sc. D. C. kindly acknowledges a predoctoral fellowship from the CONACYT (Me´xico). C. J. would like to acknowledge a predoctoral fellowship from COLCIENCIAS (Colombia). Dr. V. M. kindly acknowledges a postdoctoral Juan de la Cierva contract from the Spanish Ministry of Science and Innovation. Dr. E. Jover participated actively during early stages of the project and sampling campaigns.
appendix. Supplementary information Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2010.07.050.
references
Bendz, D., Paxeus, N.A., Ginn, T.R., Loge, F.J., 2005. Occurrence and fate of pharmaceutically active compounds in the environment, a case study: Ho¨je River in Sweden. Journal of Hazardous Materials 122, 195e204. Briggs, G.G., Bromilow, R.H., Evans, A.A., Williams, M., 1983. Relationships between lipophicity and root uptake and distribution of non-ionised chemicals in barley shots following uptake by roots. Pesticide Science 14, 492. Boxall, A.B.A., Johnson, P., Smith, E.J., Sinclair, C.J., Stutt, E., Levy, L.S., 2006. Uptake of veterinary medicines from soils into plants. Journal of Agricultural and Food Chemistry 54, 2288e2297. Buerge, I.J., Poiger, T., Muller, M.D., Buser, H.R., 2003. Caffeine, an anthropogenic marker for wastewater contamination of surface waters. Environmental Science and Technology 37, 691e700. Calderon-Preciado, D., Jimenez-Cartagena, C., Pen˜uela, G., Bayona, J.M., 2009. Development of an analytical procedure for the determination of emerging and priority organic pollutants in leafy vegetables by pressurized solvent extraction followed by GCeMS determination. Analytical and Bioanalytical Chemistry 394, 1319e1327. Carballa, M., Omil, F., Lema, J.M., 2008. Comparison of predicted and measured concentrations of selected pharmaceuticals, fragrances and hormones in Spanish sewage. Chemosphere 72, 1118e1123. Clara, M., Strenn, B., Gans, O., Martinez, E., Kreuzinger, N., Kroiss, H., 2005. Removal of selected pharmaceuticals, fragrances and endocrine disrupting compounds in a membrane bioreactor and conventional wastewater treatment plants. Water Research 39, 4797e4807. Charrois, J.W.A., Arend, M.W., Froese, K.L., Hrudey, S.E., 2004. Detecting N-nitrosamines in drinking water at nanogram per liter levels using ammonia positive chemical ionization. Environmental Science and Technology 38, 4835e4841. Chen, D.H., Zeng, X.Y., Sheng, Y.Q., Bi, X.H., Gui, H.Y., Sheng, G.Y., Fu, J.M., 2007. The concentrations and distribution of polycyclic musks in a typical cosmetic plant. Chemosphere 66, 252e258. Chen, Z., Valentine, R.L., 2007. Formation of N-nitrosodimethylamine (NDMA) from humic substances in natural water. Environmental Science and Technology 41, 6059e6065. Cho, D.H., Kong, S.H., Oh, S.G., 2003. Analysis of trihalomethanes in drinking water using headspace-SPME technique with gas chromatography. Water Research 37, 402e408. Durner, J., Shah, J., Klessig, W.R., 1997. Salicylic acid and disease resistance in plants. Trends in Plant Science 2, 266. Environmental Protection Agency (2008). Emerging contaminant -N-Nitrosodimethylamine (NDMA), http://www.epa.gov/nscep/. Fent, K., Weston, A.A., Caminada, D., 2006. Ecotoxicology of human pharmaceuticals. Aquatic Toxicology 76, 122e159. Fismes, J., Perrin-Ganier, C., Empereur-Bissonnet, P., Morel, J.L., 2002. Soil-to-root transfer and translocation of polycyclic aromatic hydrocarbons by vegetables grown on industrial contaminated soils. Journal of Environmental Quality 31, 1649e1656. Generalitat de Catalunya, M.A.i.H., 2008. REAL DECRETO-LEY 3/ 2008, de 21 de abril, de medidas excepcionales y urgentes para garantizar el abastecimiento de poblaciones afectadas por la sequı´a en la provincia de Bar celona., Barcelona. Gent, M.P., White, J.C., Parrish, Z.D., Isleyen, M., Eitzer, B.D., Mattina, M.I., 2007. Uptake and translocation of p, p’dichlorodiphenyldichloroethylene supplied in hydroponics solution to Cucurbita. Environmental Toxicology and Chemistry 26, 2467e2475. Gervais, G., Brosillon, S., Laplanche, A., Helen, C., 2008. Ultrapressure liquid chromatography-electrospray tandem mass
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 2 1 e2 3 1
spectrometry for multiresidue determination of pesticides in water. Journal of Chromatography A 1202, 163e172. Gonzalez, M., Miglioranza, K.S.B., Aizpu´n de Moreno, J.E., Moreno, V.J., 2003. Occurrence and distribution of organochlorine pesticides (OCPs) in tomato (Lycopersicon esculentum) crops from organic production. Journal of Agricultural and Food Chemistry 51, 1353e1359. Gros, M., Petrovic, M., Barcelo, D., 2007. Wastewater treatment plants as a pathway for aquatic contamination by pharmaceuticals in the ebro river basin (northeast spain). Environmental Toxicology and Chemistry 26, 1553e1562. Helena, B., Pardo, R., Vega, M., Barrado, E., Fernandez, J.M., Fernandez, L., 2000. Temporal evolution of groundwater composition in an alluvial aquifer (Pisuerga River, Spain) by principal component analysis. Water Research 34, 807e816. Hildebrandt, A., Guillamon, M., Lacorte, S., Tauler, R., Barcelo, D., 2008. Impact of pesticides used in agriculture and vineyards to surface and groundwater quality (North Spain). Water Research 42, 3315e3326. Jimenez, B., Asano, T., 2008. Water Reuse. An International Survey of Current Practice Issues and Needs. IWA Publishing, London. Jjemba, P.K., 2008. Pharma-Ecology: The Occurrence and Fate of Pharmaceuticals and Personal Care Products in The Environment. John Wiley & Sons, Inc., New Jersey. Jobling, S., Nolan, M., Tyler, C.R., Brighty, G., Sumpter, J.P., 1998. Widespread sexual disruption in wild fish. Environmental Science and Technology 32, 2498e2506. Jollife, I.T., 2002. Principal Component Analysis. Springer-Verlag, New York. 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. 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 US streams, 1999e2000: a national reconnaissance. Environmental Science and Technology 36, 1202e1211. Kong, W.D., Zhu, Y.G., Liang, Y.C., Zhang, J., Smith, F.A., Yang, A., 2007. Uptake of oxytetracycline and its phytotoxicity to alfalfa (Medicago sativa L.). Environmental Pollution 147, 187e193. 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, 834e842. Luckenbach, T., Epel, D., 2005. Nitromusk and polycyclic musk compounds as long-term inhibitors of cellular xenobiotic defence systems mediated by multidrug transporters. Environmental Health Perspectives 113, 17e24. Martı´nez Bueno, J.M., Aguera, A., Gomez, M.J., Hernando, M.D., Garcia-Reyes, J.F., Fernandez-Alba, A.R., 2007. Application of liquid chromatography/quadrupole-linear ion trap mass spectrometry and time-of-flight mass spectrometry to the determination of pharmaceuticals and related contaminants in wastewater. Analytical Chemistry 79, 9372e9384. Matamoros, V., Bayona, J.M., 2006. Elimination of pharmaceuticals and personal care products in subsurface flow constructed wetlands. Environmental Science and Technology 40, 5811e5816. Migliore, L., Civitareale, C., Cozzolino, S., Casoria, P., Brambilla, G., Gaudio, L., 1998. Laboratory models to evaluate phytotoxicity
231
of sulphadimethoxine on terrestrial plants. Chemosphere 37, 2957e2961. Moldovan, Z., Schmutzer, G., Tusa, F., Calin, R., Alder, A.C., 2007. An overview of pharmaceuticals and personal care products contamination along the river somes watershed, Romania. Journal of Environmental Monitoring 9, 986e993. Munch, J.W., Basset, M.V., 2004. Method 521. Determination of nitrosamines in drinking water by solid phase extraction and capillary column gas chromatography with large volume injection and chemical ionization tandem mass spectrometry (MS/MS). In: Laboratory, N.E.R. (Ed.), (Cincinnati). Oki, T., Kanae, S., 2006. Global hydrological cycles and world water resources. Science 313, 1068e1072. Paasivirta, J., Sinkkonen, S., Rantalainen, A.L., Broman, D., Zebuhr, Y., 2002. Temperature dependent properties of environmentally important synthetic musks. Environmental Science and Pollution Research 9, 345e355. Pedersen, J.A., Soliman, M., Suffet, I.H., 2005. Human pharmaceuticals, hormones, and personal care product ingredients in runoff from agricultural fields irrigated with treated wastewater. Journal of Agricultural and Food Chemistry 53, 1625e1632. Pere-Trepat, E., Olivella, L., Ginebreda, A., Caixach, J., Tauler, R., 2006. Chemometrics modelling of organic contaminants in fish and sediment river samples. Science of the Total Environment 371, 223e237. Petrovic, M., Gonzalez, S., Barcelo, D., 2003. Analysis and removal of emerging contaminants in wastewater and drinking water. Trac-Trends in Analytical Chemistry 22, 685e696. Redshaw, C.H., Wootton, V.G., Rowland, S.J., 2008. Uptake of the pharmaceutical Fluoxetine Hydrochloride from growth medium by Brassicaceae. Phytochemistry 69, 2510e2516. Reemtsma, T., Weiss, S., Mueller, J., Petrovic, M., Gonzalez, S., Barcelo, D., Ventura, F., Knepper, T.P., 2006. Polar pollutants entry into the water cycle by municipal wastewater: a European perspective. Environmental Science and Technology 40, 5451e5458. Richardson, S., Thruston, A., Rav-Acha, C., Groisman, L., Popilevsky, I., Juraev, O., Glezer, V., McKague, B., Plewa, M., Wagner, E., 2003. Tribromopyrrole, brominated acids, and other disinfection byproducts produced by disinfection of drinking water rich in bromide. Environmental Science and Technology 37, 3782e3793. Rosegrant, M.W., Cai, X., 2001. Water scarcity and food security: alternative futures for the 21st century. Water Science and Technology 43, 61e70. Samsoe-Petersen, L., Larsen, E.H., Larsen, P.B., Bruun, P., 2002. Uptake of trace elements and PAHs by fruit and vegetables from contaminated soils. Environmental Science and Technology 36, 3057e3063. Ternes, T.A., Bonerz, M., Herrmann, N., Teiser, B., Andersen, H.R., 2007. Irrigation of treated wastewater in Braunschweig, Germany: An option to remove pharmaceuticals and musk fragrances. Chemosphere 66, 894e904. Xuan, T.D., Toyama, T., Fukuta, M., Khanh, T.D., Tawat, S., 2009. Chemical Interaction in the Invasiveness of Cogongrass (Imperata cylindrica (L.) Beauv.). Journal of Agricultural and Food Chemistry 57. Zimmerman, J.B., Mihelcic, J.R., Smith, J., 2008. Global stressors on water quality and quantity. Environmental Science and Technology 42, 4247e4254.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 3 2 e2 4 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
A pH-control model for heterotrophic and hydrogen-based autotrophic denitrification Youneng Tang*, Chen Zhou, Michal Ziv-El, Bruce E. Rittmann Center for Environmental Biotechnology, Biodesign Institute at Arizona State University, 1001 South McAllister Avenue, Tempe, AZ 85287-5701, USA
article info
abstract
Article history:
This work presents a model to predict the alkalinity, pH, and Langelier Saturation Index
Received 15 February 2010
(LSI) in heterotrophic and H2-based autotrophic denitrification systems. The model can
Received in revised form
also be used to estimate the amount of acid, e.g. HCl, added to the influent (method 1) or the
11 May 2010
pH set point in the reactor (method 2: pH can be maintained stable by CO2-sparge using
Accepted 16 July 2010
a pH-control loop) to prevent the pH from exceeding the optimal range for denitrification
Available online 23 July 2010
and to prevent precipitation from occurring. The model was tested with two pilot plants carrying out denitrification of groundwater with high hardness: a heterotrophic system
Keywords:
using ethanol as the electron donor and an H2-based autotrophic system. The measured
Denitrification
alkalinity, pH, and LSI were consistent with the model for both systems. This work also
Groundwater treatment
quantifies: (1) how the alkalinity and pH in Stage-1 significantly differ from those in Stage-
pH control
2; (2) how the pH and LSI differ significantly in the two denitrification systems while the
Precipitation prevention
alkalinity increase is about the same; and (3) why CO2 addition is the preferred method for
Model
autotrophic system, while HCl addition is the preferred method for the heterotrophic system. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Heterotrophic and H2-based autotrophic denitrification processes have been widely tested at the pilot-scale for drinking water treatment, and a few full-scale plants have been built and operated in Europe (Richard, 1989; Rogalla et al., 1990). One important characteristic of denitrification is that it produces approximately one equivalent of strong base for each equivalent of N reduced beyond nitrite (Lee and Rittmann, 2003). One risk from base production is high-pH inhibition. The optimal pH range reported for denitrification is 7e9; values outside this range can retard the denitrification process and lead to accumulation of intermediates: NO 2 , NO2, and N2O (Kurt et al., 1987; Janda et al., 1988; Lee and Rittmann, 2003; Baeseman et al., 2006; Sengupta and Ergas, 2006).
A second risk is precipitation of hardness cations with common basic anions. Common mineral precipitates in biological denitrification processes includes calcium carbonate (CaCO3), calcium hydrogen phosphate (CaHPO4), calcium dihydrogen phosphate (Ca(H2PO4)), hydroxyapatite (Ca5(PO4)3OH), and b-tricalcium phosphate (Ca3(PO4)2) (Snoeyink and Jenkins, 1980; Lee and Rittmann, 2003). For example, CaCO3 precipitation was observed in bench-scale and pilot-scale denitrification reactors using real groundwater (Adham et al., 2004; Ziv-El and Rittmann, 2009). Inorganic precipitates can have negative impacts on biological denitrification processes. For the H2-based membrane biofilm reactor, build-up of mineral solids inside the biofilm and at its interface with the membrane can increase masstransport resistance for H2 diffusion within the biofilm and
* Corresponding author. Tel.: þ1 480 283 5500; fax: þ1 480 727 0889. E-mail addresses:
[email protected] (Y. Tang),
[email protected] (C. Zhou),
[email protected] (M. Ziv-El),
[email protected] (B.E. Rittmann). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.07.049
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 3 2 e2 4 0
out of the biofilm, and it also decreases the efficiency of NO 3 and nutrients transport into the biofilm (Lee and Rittmann, 2003). Another problem is the calcification of fibers, which results from the deposition and accumulation of CaCO3 (up to 25% of the biofilm mass) on the outside of the biofilm and appears to induce fiber breakage as the fiber becomes inflexible (Adham et al., 2004). Though the negative impact of precipitation on heterotrophic denitrification processes has not been studied as systematically, it is susceptible to precipitation-induced increases in mass-transport resistance, medium clogging, and poor flow distribution. When problems associated with base production and high pH are significant, pH control is necessary for a denitrification process. The pH can be controlled using either of two methods. One is to add acid (e.g., HCl) in the influent at a concentration that balances excessive base production from denitrification (method 1); and the other is to sparge CO2 into the reactor to control the pH in the reactor at a set point using a pH-control loop (method 2) (Adham et al., 2004). Though the significance of pH control has been well established in the literature and the two pH-control methods have been proposed (Kurt et al., 1987; Janda et al., 1988; Lee and Rittmann, 2003; Adham et al., 2004; Baeseman et al., 2006; Sengupta and Ergas, 2006; Ziv-El and Rittmann, 2009), pH-control models for denitrification have not been reported to our knowledge. A reliable pH-control model should be able to predict the alkalinity, pH, and precipitation risk within the denitrification reactor. Furthermore, the model should have the ability to estimate the acid concentration in the influent in method 1 and the pH set point in method 2. Therefore, the primary objective of this study is to construct and experimentally test such a model. A second objective is to quantitatively determine the preferred pH-control methods in heterotrophic and H2-based autotrophic systems.
2.
Model development
The alkalinity and pH increase in heterotrophic and H2-based autotrophic denitrification because nitrite reduction consumes protons (Hþ). Proton consumption is illustrated in Eqs. (1)e(4) (based on Rittmann and McCarty (2001)), in which ethanol (CH3CH2OH) and hydrogen gas (H2) are the heterotrophic and autotrophic electron donors, respectively, and biomass synthesis is indicated by C5H7O2N. A typical biomass retention time of 15 days (Rittmann and McCarty, 2001) was used to develop these equations. Heterotrophic denitrification: þ NO 3 þ 0.263CH3CH2OH þ 0.0445H ¼ 0.954NO2
þ 0.0445C5H7O2N þ 0.655H2O þ 0.303CO2
(1)
D NO 2 þ 0.425CH3CH2OH þ H ¼ 0.455 N2 þ 0.0912C5H7O2N
þ 1.457H2O þ 0.393CO2
(2)
Autotrophic denitrification: þ NO 3 þ 1.13H2 þ 0.01H þ 0.05CO2 ¼ 0.99NO2 þ 0.01C5H7O2N
þ 1.1H2O
(3)
233
D NO 2 þ 0.122CO2 þ H þ 1.78H2 ¼ 0.488 N2 þ 0.0244C5H7O2N
þ 2.19H2O
(4)
In both systems, nitrite reduction is the predominant source of alkalinity, consuming 1 Hþ equivalent per N equivalent of NO 2 (highlighted by boldface in Eqs. (2) and (4)). Another factor that affects pH is the net production of CO2 in heterotrophic systems (highlighted by boldface in Eqs. (1) and (2)) and net consumption of CO2 in autotrophic systems (highlighted by boldface in Eqs. (3) and (4)). CO2 is a weak acid, and its addition partially suppresses the pH rise from proton consumption, as well as increases the concentration of total inorganic carbon species. Dissolved oxygen almost always is present in water to be treated by denitrification. While respiration of O2 does not consume significant protons, oxygen respiration can affect the pH by CO2 addition in a heterotrophic system (highlighted by boldface in Eq. (5)) and CO2 consumption in an autotrophic system (highlighted by boldface in Eq. (6)). Heterotrophic O2 respiration: þ O2 þ 0.613CH3CH2OH þ 0.120NO 3 þ 0.120H ¼ 0.120C5H7O2N
þ 0.667CO2 þ 1.48H2O
(5)
Autotrophic O2 respiration: þ O2 þ 2.39H2 þ 0.0282NO 3 þ 0.141CO2 þ 0.0282H
¼ 0.0282C5H7O2N þ 2.31H2O
(6)
The feed water’s alkalinity buffers pH changes and is another factor that affects pH in the reactor. For natural water, the carbonate system dominates the alkalinity due to the common occurrence and dissolution of carbonate minerals and the presence of carbon dioxide in the atmosphere (Snoeyink and Jenkins, 1980). Addition of certain chemicals to the influent or into the reactor can also affect pH. For example, HCl can be added to the influent to lower the alkalinity and pH, while CO2 can be sparged inside the reactor to add a weak acid and increase the buffering capacity of the water. When coupled with an alkalinity mass balance (via the proton condition) in the influent and effluent, the factors mentioned above can be used to predict the effluent pH, alkalinity, and LSI. This constitutes the model whose development is described next in a stepwise manner. First, the following six assumptions are made:. (1) Phosphate species are not considered as a buffer due to two factors. First, the concentration of total phosphorus in most natural groundwater is very low due to its precipitation with calcium (Snoeyink and Jenkins, 1980). Second, phosphate added as a nutrient and dosed at the stoichiometric requirement for P uptake in biomass synthesis provides negligible phosphate species in the reactor, compared to carbonate species. (2) Other natural buffering species (e.g., ammonium) also are neglected, because they are trivial compared to the carbonate species, which account for most of the total alkalinity (Snoeyink and Jenkins, 1980).
234
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 3 2 e2 4 0
(3) Calcium carbonate is the most common mineral precipitate and the only solid species considered. The Langlier Saturation Index (LSI) is used as an indication of precipitation potential of CaCO3 (Snoeyink and Jenkins, 1980). Calcium phosphate species are neglected, since the phosphate concentration is typically low. Mg(OH)2 also is neglected, because it is super-saturated only at pH values that are too high to be relevant for biological treatment. (4) The reactor is a closed system, which means that CO2 does not exchange between the reactor and the atmosphere. (5) Activity coefficients are ignored, since most source water for drinking water treatment has a low salinity. This assumption would need to be removed if denitrification were being carried out with high-salinity water, such as regeneration brine from ion exchange (Van Ginkel et al., 2008). (6) The model assesses the conditions in the bulk liquid based on reactions occurring in the biofilm, which assumes that the pH inside the biofilm does not differ greatly from that in the bulk liquid. Therefore, the model could be expanded to consider the pH gradient in the biofilm and generate more accurate results. Second, the alkalinity in the influent and effluent of the reactor is tabulated by coupling the proton condition, the total concentration of inorganic carbonate species (CT), and the hydrogen-ion concentration ð½Alk ¼ f ðCT ; ½Hþ ÞÞ. The alkalinity equations (Eqs. (7) and (8)) are identical to the proton conditions with H2O and H2CO3 as the reference levels (Van Briesen and Rittmann, 1999). Alkalinity in the influent of the reactor: þ ½Alkin ¼ 2 CO2 3 in þ HCO3 in þ OH in H 1 ¼ 2 CT;in 2 þ 1 þ H in =K2 þ Hþ in =K1 K2 1 1014 þ þ þ þ Hþ in þ CT;in 1 þ H in =K1 þ K2 = H in H in
ð7Þ
Alkalinity in the effluent of the reactor: ½Alkout
þ ¼ 2 CO2 3 out þ HCO3 out þ OH out H out 1 ¼ 2½CT;out 2 1 þ Hþ out =K2 þ Hþ out =K1 K2 þ ½CT;out
(8) in which K1, K2 ¼ acidity constants of H2CO3 and HCO 3 (K1 ¼ 1010.3, K2 ¼ 106.3 at 2 C); CT,in, CT,out ¼ total concentration of inorganic carbon species in the influent and effluent (mole/L). [Alk]in, [Alk]out ¼ alkalinity in the influent and effluent (mole/L). Eq. (7) can be used to obtain CT,in, since [Alk]in and [Hþ]in can be measured. In order to solve Eq. (8) for the effluent pH ([Hþ]out), [Alk]out, and CT,out are calculated using Eqs. (9)e(10):
½Alkout ¼ ½Alkin þ½AlkD
CT;D ¼ NO 3 in NO3 out 0:303 þ NO3 in NO3 out þ NO2 in 2þ NO 2 out 0:393 þ ½O2 in ½O2 out 0:667 þ Ca out 2þ Ca in þ ½CO2 ðheterotrophic denitrificationÞ ð11Þ CT;D ¼
NO NO 3 out NO3 in 0:05 þ 3 out NO3 in þ NO2 out NO Ca2þ out 2 in 0:122 þ ½O2 out ½O2 in 0:141 þ 2þ Ca in þ ½CO2 ðautotrophic denitrificationÞ ð12Þ
[Alk]D ¼ change of alkalinity due to denitrification (Eqs. (1)e(4)), oxygen respiration (Eqs. (5) and (6)), precipitation, and external acid addition (e.g., HCl and H3PO4). ½AlkD ¼
NO NO 3 out NO3 in 0:0445 þ 3 in NO 3 out þ NO2 in NO2 out 1 þ ½O2 out ½O2 in 0:120 þ 2 Ca2þ out Ca2þ in ½Acid ðheterotrophic denitrificationÞ
½AlkD ¼
ð13Þ
NO NO 3 in NO3 out 0:01 þ 3 in NO3 out þ NO 2 in NO2 out 1 þ ½O2 in ½O2 out 0:0282 2þ 2þ þ 2 Ca out Ca in ½Acid ðautotrophic denitrificationÞ
ð14Þ
The concentration of external CO2 addition can be obtained using the CO2 flow rate, or it can be computed from the set pH, as shown below. After substituting in Eqs. (7) and (9)e(14), the only unknown variable in Eq. (8), is [Hþ]out. Thus, we can solve for [Hþ]out and, from that, the pH in the effluent. After that, the effluent alkalinity can be calculated with Eq. (8), and the Langlier Saturation Index (LSI) can be computed with Eq. (15) (Snoeyink and Jenkins, 1980; Ziv-El and Rittmann, 2009). LSI ¼ pH pK2 pKso þ p Ca2þ out þp HCO 3 out log gCa2þ log gHCO3
ð15Þ
where Kso ¼ solubility constant of CaCO3(s) (108.3 at 25 C); gCa2þ ¼ activity coefficient of Ca2þ; gHCO3 ¼ activity coefficient of HCO 3.
1 1014 þ þ Hþ out 1 þ Hþ out =K1 þ K2 = Hþ out H out
CT;out ¼ CT;in þ CT;D
respiration (Eqs. (5) and (6)), precipitation, and external CO2 addition.
(9) (10)
in which CT,D ¼ the change of total concentration of inorganic carbon species due to denitrification (Eqs. (1)e(4)), oxygen
3.
Experimental methods
Two pilot-scale denitrification systems treating hard groundwater were used to test the model: a heterotrophic system (fixed-bed, up-flow reactor with plastic media and ethanol as the electron donor) and an H2-based autotrophic system (H2-based membrane biofilm reactor, or MBfR). The details of both systems and the quality of the feed water have been described in Tang et al. (2010, submitted for publication). Because both were two-stage systems, the raw water, Stage-1 effluent, and Stage-2 effluent were assayed on a regular basis. Temperature, pH, total dissolved solids (TDS), hardness, and alkalinity on a daily basis; nitrate, nitrite, calcium, and phosphate three times per week; and dissolved oxygen once per week. The analysis of anions (nitrate, nitrite, and phosphate) and cation (calcium) followed U.S. Environmental Protection
235
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 3 2 e2 4 0
Table 1 e Phosphoric acid doses. Raw water nitrate concentration (mg-N/L)
Proton addition due to phosphoric acid dosea (mole/L)
Phosphoric acid dose (mg P/L) Autotrophic system Heterotrophic system
11.8 19.6 32.1
0.15 0.48 0.76
Autotrophic system
Heterotrophic system
5
1.5 104 1.9 104 2.9 104
1.5 10 4.6 105 7.4 105
1.5 2 3
3 a Addition of H3PO4 adds 3 Hþ equivalents against the reference level of PO3 4 . Precipitation or uptake of PO4 removes no acid, as it is the reference level.
Agency (1999) Method 300.1 and other measurement followed Standard Methods for the Examination of Water and Wastewater, 20th ed. (Clesceri et al., 1998) (temperature; SM 2550 B; pH: SM 4500 H B; TDS: SM 2510 B; Alkalinity: SM 2320 B; Hardness: SM 2340 C; DO: SM 4500 O G). For each system, three nitrate feed concentrations were tested to give a wide range of potential alkalinity and pH increases: the natural concentration (11.8 mg-N/L) and two spiked concentrations (19.6 and 32.1 mg-N/L). Phosphoric acid was added to the influent as a nutrient at the concentrations listed in Table 1. Since the predicted pH was out of the optimum range for denitrification and the LSI was positive, CO2 was sparged into the autotrophic system to control the pH in the reactor at the set point of 7.0 based on automated feedback control. In the heterotrophic system, pH control by HCl addition to the influent was possible, but not used because the predicted pH was within the optimum range for denitrification and the LSI was negative for most of the experiments.
4.
Results and discussion
alkalinity had an error of less than 1% for all cases, and the LSI deviated by less than 0.1 LSI units. It should be noted that, when the influent nitrate concentration was spiked to elevated levels (19.6 mg-N/L and 32.1 mg-N/L) in the heterotrophic pilot system, calcium precipitation occurred; this was observed as a deficit in effluent soluble calcium concentration compared to the raw water (Table 2). Measures to control the pH were not taken then, since these tests were short term (20 days) and the predicted pH (7.3e8.7) was within the optimum range for denitrification. The implications for longterm operation are discussed in the next section. Table 5 presents a comparison of the measured and modelpredicted values for the autotrophic system. This system required a modification to the model, since the pH in the pilot system had a fixed pH of 7.0; this pH was achieved by adding CO2 with an automated feedback loop. In this case, CO2 dosage (i.e., [CO2] in Eqs. (11) and (12)) was the unknown, and the effluent pH was a model input. The alkalinity in the effluent was obtained using Eqs. (10) and (14) and the LSI with Eq. (15). The deviations between measured and model-predicted values were small for the MBfR: less than 2% for the alkalinity, with the LSI within 0.01 LSI units.
4.1.
Model evaluation
4.2.
The model was evaluated using data from the pilot systems. The model inputs for each system are listed in Tables 2 and 3 (heterotrophic system and autotrophic system, respectively). As described in the next two paragraphs, the model was able to predict the effluent values of pH, alkalinity, and LSI with minimal error. Table 4 presents a comparison for the heterotrophic system of the measured effluent pH, alkalinity, and LSI with the model-predicted values. The model outputs of the pH and
Necessity of pH control
In order to highlight the need to control the influent pH, Figs. 1 and 2 (heterotrophic system and autotrophic system, respectively) present the model simulations of the effluent pH, alkalinity, and LSI with and without controlling the influent pH. The measured raw water values also are displayed. To simulate scenarios where the pH was not controlled, we assumed no acid addition ([CO2] ¼ 0 and [Acid] ¼ 0) and no precipitation ([Ca2þ]out ¼ [Ca2þ]in) when solving Eqs. (11)e(14). The latter is a simplification that yields the maximum effluent
Table 2 e Experimentally measured model inputs for the heterotrophic system.
DO (mg/L) NO 3 (mg-N/L) NO 2 (mg-N/L) Ca2þ (mg/L) PO3 4 (mg-P/L) pH Alkalinity (mg/L as CaCO3)
Raw water nitrate ¼ 11.8 mg-N/L
Raw water nitrate ¼ 19.6 mg-N/L
Raw water nitrate ¼ 32.1 mg-N/L
Raw water
Stage-1
Stage-2
Raw water
Stage-1
Stage-2
Raw water
Stage-1
Stage-2
4.5 1.0 11.8 0.4 <0.01 67.1 0.2 1.5 7.6 0.1 84 4
0 0.34 0.2 0.60 0.3 67.1 0.2 <0.01
0 <0.01 <0.01 67.0 0.3 <0.01
4.5 1.0 19.6 <0.01 65.6 0.1 2 7.6 0.2 84 2
1.0 5.9 1.0 2.3 0.5 65.6 0.1 <0.01
0 <0.01 <0.01 63.6 0.2 <0.01
4.5 1.0 32.1 <0.01 65.7 0.3 3 7.6 0.1 84 3
0 13.1 0.8 2.9 0.2 65.7 0.2 <0.01
0 0.1 2.5 63.6 0.1 <0.01
236
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 3 2 e2 4 0
Table 3 e Experimentally measured model inputs for the autotrophic system. Raw water nitrate ¼ 11.8 mg-N/L a
Raw water DO (mg/L) NO 3 (mg-N/L) NO 2 (mg-N/L) Ca2þ (mg/L) PO3 4 (mg-P/L) pH Alkalinity (mg/L as CaCO3)
0.15
Stage-1
Stage-2
0 <0.01 <0.01 67.1 0.2 <0.01 7.0
0 <0.01 <0.01 67.0 0.3 <0.01 7.0
Raw water nitrate ¼ 19.6 mg-N/L a
Raw water
0.48
Stage-1
Stage-2
1.0 0.7 0.3 1.7 0.5 65.7 0.1 <0.01 7.0
0 <0.01 <0.01 65.5 0.2 <0.01 7.0
Raw water nitrate ¼ 32.1 mg-N/L Raw watera
0.76
Stage-1
Stage-2
0 10.4 7.3 65.6 0.2 <0.01 7.0
0 0.1 2.5 65.7 0.1 <0.01 7.0
a The raw water quality was the same as in Table 2.
precipitation risk displayed as the LSI. For the autotrophic system, the pH in the actual pilot system was controlled, and the modifications to the model are described in the previous section. For the heterotrophic system, to simulate the scenario where the pH was controlled, the concentration of acid required in the influent ([Acid] in Eqs. (13) and (14)) was obtained using a trial-and-error method for the input acid until the predicted LSI equaled 0 and the pH in the effluent was less than 9. The same method could have been used for the autotrophic system, but by varying [CO2] in Eqs. (11) and (12) instead of the input acid. In practice, an LSI above 0.5 leads to noticeably increased scaling (Camerata et al., 2008); thus, LSI ¼ 0 was used here in order to incorporate a safety factor. A pH of 9 was selected, since this is the upper limit for high-rate denitrification, as discussed in the Section 1.
4.2.1.
Heterotrophic system
Looking at the influent nitrate concentration of 11.8 mg N/L, the pH remains within the optimal range in the raw water and both stages, the LSI is negative in Stage-1 and just above zero in Stage-2; thus, pH adjustment was not required in the heterotrophic system when the influent nitrate concentration was 11.8 mg N/L. In this case, the potential to increase in pH by proton consumption during denitrification was mostly balanced by (1) production of CO2 by denitrification and O2 respiration and (2) addition of phosphoric acid. The following calculations quantify the impact of the different factors affecting alkalinity and pH changes. Proton consumption
through denitrification is calculated using Eqs. (1), (2), and (5): 8.4 104 mole/L. External proton addition from phosphoric acid is obtained from Table 1: 1.5 104 mole/L. Thus, the alkalinity increased by 6.9 104 mole/L ¼ 35 mg/L as CaCO3 (Fig. 1). CO2 production through denitrification and O2 respiration is calculated using Eqs. (1), (2) and (5): 6.6 104 mole/L. The CO2 production was slightly lower than the alkalinity increase, resulting in only a slight pH increase. When nitrate was spiked to 19.6 and 32.1 mg-N/L, the scenario of “without HCl addition” shows a pH of around 7.5 in Stage-1, with the LSI remaining negative in Stage-1. However, the effluent pH increases to over 8.5, making the LSI greater than 1 in Stage-2 and indicating a serious precipitation risk that would necessitate pH control for long-term operation. The experimental results for the short-term experiments showed CaCO3 precipitation (Table 2). The following calculations quantify how the alkalinity and pH in Stage-1 significantly differed from those in Stage-2. Taking the scenario of 19.6 mg-N/L for example, alkalinity increased in Stage-1 by 6.1 104 mole/L ¼ 31 mg/L as CaCO3 (Fig. 1). CO2 was produced in Stage-1 at 7.1 104 mole/L through denitrification and O2 respiration. The CO2 production was higher than the alkalinity increase on an equivalent basis, resulting in a slight decrease of pH (Fig. 1) in Stage-1. In Stage-2, compared to the raw water, the alkalinity increase (1.2 103 mole/L ¼ 60 mg/L as CaCO3 (Fig. 1)) was much higher than the CO2 production (1.0 103 mole/L), leading to the large increase in pH.
Table 4 e Comparison of the measured and model-predicted pH, alkalinity, and LSI for the heterotrophic system. Raw water nitrate ¼ 11.8 mg-N/L
Raw water nitrate ¼ 19.6 mg-N/L
Raw water nitrate ¼ 32.1 mg-N/L
Stage-1
Stage-2
Stage-1
Stage-2
Stage-1
Stage-2
pH
Measured Model-predicted Difference (%)
7.7 0.1 7.7 0
7.8 0.1 7.8 0
7.4 0.1 7.4 0
8.2 0.0 8.3 1
7.4 0.1 7.5 1
8.3 0.1 8.4 1
Alkalinity (mg/L as CaCO3)
Measured Model-predicted Difference (%)
116 3 116 0
120 7 119 0.8
114 4 115 0.9
142 2 139 2
127 5 127 0
168 5 164 2
LSI
Measured Model-predicted Difference
0.03 0.07 0.04
0.08 0.09 0.01
0.34 0.29 0.05
0.54 0.64 0.10
0.30 0.22 0.08
0.70 0.78 0.08
237
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 3 2 e2 4 0
Table 5 e Comparison of the measured and model-predicted alkalinity and LSI for the autotrophic system. Raw water nitrate ¼ 11.8 mg-N/L
Raw water nitrate ¼ 19.6 mg-N/L
Raw water nitrate ¼ 32.1 mg-N/L
Stage-1
Stage-2
Stage-1
Stage-2
Stage-1
Stage-2
Alkalinity (mg/L as CaCO3)
Measured Model-predicted Difference (%)
124 3 125 0.08
124 4 125 0.08
140 4 143 2
155 3 152 2
135 4 132 0.2
185 2 186 0.05
LSI
Measured Model-predicted Difference
0.70 0.69 0.01
0.70 0.69 0.01
0.65 0.64 0.01
0.61 0.61 0
0.66 0.67 0.01
0.53 0.52 0.01
Fig. 1 also presents a scenario in which acid addition is administered to the raw water so that the LSI is negative or zero in both stages. The influent pH is 6.5, as the alkalinity drops to w60 mg/L as CaCO3 when enough HCl is added to compensate for alkalinity addition in both stages of the reactor. Since the pH and LSI are within the optimal range in Stage-1 without HCl addition, the pH adjustment could be implemented between Stage-1 and Stage-2, which has an advantage that the pH in the raw water (and the inlet of
Stage-1) would not drop below 7.0, the lower limit of the optimal range for denitrification.
4.2.2.
Autotrophic systems
Fig. 2 presents similar analyses for the autotrophic system. Distinct from the relatively mild pH increase in heterotrophic systems without acid addition, the pH would increase to greater than 10 and the LSI to greater than 2 in both stages and for all three influent nitrate concentrations if CO2 were not
Fig. 1 e Measured and model-predicted pH, alkalinity, and LSI in the heterotrophic system for three influent nitrate concentrations. Two scenarios are considered, with and without HCl addition. To control LSI [ 0 in Stage-2, HCl should be added at 1.7 3 10L4 and 2.3 3 10L4 mole/L when the nitrate concentration at the influent is 19.6 and 32.1 mg-N/L, respectively.
238
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 3 2 e2 4 0
Fig. 2 e Measured and model-predicted pH, alkalinity, and LSI in the autotrophic system for three influent nitrate concentrations. Two scenarios were considered, with and without CO2 addition.
added inside the reactor. This means that the autotrophic denitrification system has greater risks of severe pH and precipitation problems than does the heterotrophic system. The scenario of 11.8 mg-N/L without pH control was analyzed to quantify how the pH and LSI differ significantly in the two denitrification systems, even though the alkalinity increase was about the same. In Stage-2 of the autotrophic system, proton consumption was calculated using Eqs. (3), (4), and (6): 8.4 104 mole/L. External proton addition from phosphoric acid was obtained from Table 1: 1.5 105 mole/L. Thus, the alkalinity increased by 8.2 104 mole/L ¼ 41 mg/L as CaCO3 (Fig. 2), which is close to that in the heterotrophic system (31 mg/L as CaCO3). CO2 consumption was calculated using Eqs. (3), (4), and (6): 1.7 104 mole/L, which contrasts to CO2 production in the heterotrophic system (6.6 104 mole/L). The combined effect of alkalinity increase and CO2 consumption in the autotrophic system led to the much larger increases in pH and, thus, LSI. In the pH-control scenario, the added CO2 offsets the protons consumed by increasing the total concentration of carbonate species, but without changing the alkalinity. The “with CO2” in Fig. 2 indicates that precipitation can be
prevented entirely by sparging CO2 to keep a pH set point at 7, and the experimental results confirm the prediction (Table 3).
4.3.
Preferred pH-control method
Though both methods e CO2 addition and HCl addition e can be used for pH control in heterotrophic and autotrophic systems, CO2 addition is the preferred method for autotrophic system, since it has an additional advantage of providing the inorganic carbon source and increasing the total concentration of the carbonate buffering system. HCl addition is the preferred method for the heterotrophic system, since oxidation of organic matter already is increasing the concentration of the carbonate buffering system. Adding more CO2 by sparging could cause the water to become over-saturated with CO2, which would lead to CO2 escape after the water leaves the denitrification system and a subsequent positive LSI. The scenario of 32.1 mg-N/L was used to identify the better pH-control method for autotrophic versus heterotrophic denitrification. With all nitrate completely reduced to N2 in the H2-based autotrophic system, CO2 consumption due to denitrification and oxygen respiration was 4.1 104 mole/L.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 3 2 e2 4 0
The total inorganic carbon concentration in the influent was calculated using the influent pH and alkalinity: 1.8 103 mole/L. Therefore, around 23% of the total inorganic carbon in the influent was consumed due to denitrification and oxygen respiration; this number would increase if the influent alkalinity were lower or the influent nitrate concentration were higher. Should more than 100% of the influent inorganic carbon be consusmed, CO2 sparging would be the only feasible method. In a heterotrophic system, when all nitrate was completely reduced to N2, CO2 production due to denitrification and oxygen respiration was 1.7 103 mole/L, which is close to the total carbon concentration in the influent. If external CO2 were added to make the LSI 0, it should be added at 4.5 104 mole/L according to the model. When this CO2 super-saturated water leaves the reactor and is open the atmosphere, 3.5 104 mole/L CO2 (78% of the externally added CO2) would escape, the pH would increase to 8.9, and the LSI would increase to 1.3; this follows calculation of carbonate species concentrations in open and closed systems (Snoeyink and Jenkins, 1980). In order to avoid the super-saturation condition in the effluent, HCl addition is a preferred pH-control method for heterotrophic systems.
5.
Conclusions
(1) This work presents a model to predict the pH, alkalinity, and LSI in the effluent of H2-based autotrophic denitrification and heterotrophic systems. If the model outputs a pH value outside the optimal range for the denitrifiers or a high LSI value indicating serious precipitation potential, operators should take measures to control the pH. The pH can be controlled using either of two methods: One is to add acid (e.g., HCl) to balance excessive base production from denitrification; the other is to add acid CO2 into the reactor to hold the pH to an set point using an automated pH feedback loop. The model can be used to estimate the required acid additions for both scenarios. (2) The model was evaluated using data from two pilot denitrification systems: a two-stage heterotrophic system using ethanol as the electron donor and a two-stage H2-based autotrophic system. The model-predicted pH, alkalinity, and LSI matched well with the experimental data in all cases tested. For the heterotrophic system, no acid addition was required for the long-term operation with an input NO 3 concentration of 11.8 mg N/L, since the LSI remained negative. For short-term experiments with higher influent NO 3 concentrations, the LSI was positive, and CaCO3 precipitated. For the autotrophic system, precipitation would have been severe at all input NO 3 concentrations, but precipitation was prevented by sparging CO2 directly into the reactor. (3) The model showed that the autotrophic system was more sensitive to pH increases, even though denitrification directly increased alkalinity about the same amount for heterotrophic and H2-based autotrophic processes. Since CO2 is consumed in the autotrophic process and produced in the heterotrophic process, pH and LSI increase more in the H2-based autotrophic process than in the
239
heterotrophic process, meaning that the autotrophic process is more susceptible to an increase in pH and to CaCO3 precipitation. The actual impact on pH depends on the natural alkalinity (buffering) and the use of pH-control measures. (4) Acid (e.g., HCl) addition is the preferred pH-control method for heterotrophic processes, but CO2 addition is the preferred method for H2-based autotrophic processes.
Acknowledgements This work was supported by Water Research Foundation Project #4131, which is funded by the Water Research Foundation and the City of Glendale. The authors of this paper recognize the City of Glendale and the utility staff for their significant financial and technical contributions for this work. The authors acknowledge that the Foundation is the joint owner of the technical information upon which this paper is based and thank the Foundation for its financial, technical, and administrative assistance in funding and managing the project. The authors of this paper also acknowledge their coworkers e Kerry Meyer, Daniel Candelaria, and Paul Swaim from CH2M HILL, as well as Jung Hun Shin and Chang Hong Ahn from Arizona State University e for their significant contributions to the pilot test.
references
Adham, S., Gillogly, T., Lehman, G., Rittmann, B.E., Nerenberg, R., 2004. Membrane Biofilm Reactor Process for Nitrate and Perchlorate Removal. AWWARF and AWWA, Denver, Colorado. Baeseman, J.L., Smith, R.L., Silverstein, J., 2006. Denitrification potential in stream sediments impacted by acid mine drainage: effects of ph, various electron donors, and iron. Microbial Ecology 51, 232e241. Camerata, J., Pearce, E., Greer, P., White, J., Silva Bueno, M.A., 2008. Active Sludge: Professional Development Course. http:// www.tlch20.com/courses/ActivatedSludge.pdf. Clesceri, L.S., Greenberg, A.E., Eaton, A.D., 1998. Standard Methods for the Examination of Water and Wastewater, 20th ed. American Public Health Association, Washington DC. Janda, V., Rudovsky, J., Wanner, J., Marha, K., 1988. In situ denitrification of drinking water. Water Science and Technology 20 (3), 215e219. Kurt, M., Dunn, I.J., Bourne, J.R., 1987. Biological denitrification of drinking water using autotrophic organisms with H2 in a fluidized-bed biofilm reactor. Biotechnology and Bioengineering 29 (4), 493e501. Lee, K.C., Rittmann, B.E., 2003. Effects of pH and precipitation on autohydrogenotrophic denitrification using the hollow-fiber membrane-biofilm reactor. Water Research 37, 1551e1556. Richard, Y., 1989. Operating experiences of full-scale biological and ion-exchange denitrification plants in France. Water and Environmental Management Journal 3 (2), 154e167. Rittmann, B.E., McCarty, P.L., 2001. Environmental Biotechnology: Principles and Applications. McGraw-Hill Companies, Inc., New York. Rogalla, F., Ravarini, P., de Larminat, G., Couttelle, J., 1990. Largescale biological nitrate and ammonia removal. Journal of the Institution of Water and Environmental Management 4 (4), 319e329.
240
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 3 2 e2 4 0
Sengupta, S., Ergas, S.J., 2006. Autotrophic Biological Denitrification with Elemental Sulfur or Hydrogen for Complete Removal of Nitrate-Nitrogen from a Septic System Wastewater. NOAA/UNH (National Oceanic and Atmospheric Administration/ University of New Hampshire) Cooperative Institute for Coastal and Estuarine Environmental Technology. Final Report. Snoeyink, V.L., Jenkins, D., 1980. Water Chemistry. Wiley, Inc., New York. Tang, Y., Ziv-El, M., Zhou, C., Shin, J.H., Ahn, C.H., Meyer, K., Candelaria, D., Friese, D., Overstreet, R., Scott, R., Rittmann, B. E., 2010. Bioreduction of nitrate in groundwater using a pilotscale hydrogen-based membrane biofilm reactor. Frontiers of Environmental Science & Engineering in China doi:10.1007/ s11783-010-0235-9. Tang, Y., Ziv-El, M., Zhou C., Shin J.H., Ahn C.H., Meyer K., McQuarrie J., Candelaria, D., Swaim P., Scott R., Rittmann, B.E. Using the carrier-surface loading to design heterotrophic
denitrification reactors. Journal AWWA, submitted for publication. U.S. Environmental Protection Agency, 1999. Method 300.1: Determination of Inorganic Inions in Drinking Water by Ion Chromotography. US Environmental Protetion Agency. http:// www.caslab.com/EPA-Method-300_1/. Van Briesen, J.M., Rittmann, B.E., 1999. Modeling speciation effects on biodegradation in mixed metal/chelate systems. Biodegradation 10, 315e330. Van Ginkel, S.W., Ahn, C.H., Badruzzaman, M., Roberts, D.J., Lehman, S.G., Adham, S., Rittmann, B.E., 2008. Kinetics of nitrate and perchlorate reduction in ion exchange brine using the membrane biofilm reactor (MBfR). Water Research 42, 4197e4205. Ziv-El, M., Rittmann, B.E., 2009. Systematic evaluation of nitrate and perchlorate biorecution kinetics in groundwater using a hydrogen-based membrane biofilm reactor. Water Research 43, 173e181.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 1 e2 5 3
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Cost-effectiveness analysis of risk-reduction measures to reach water safety targets Andreas Lindhe a,*, Lars Rose´n a, Tommy Norberg b, Olof Bergstedt a,c, Thomas J.R. Pettersson a a
Department of Civil and Environmental Engineering, Chalmers University of Technology, SE-412 96 Gothenburg, Sweden Department of Mathematical Sciences, University of Gothenburg and Chalmers University of Technology, SE-412 96 Gothenburg, Sweden c Go¨teborg Vatten, Box 123, SE-424 23 Angered, Sweden b
article info
abstract
Article history:
Identifying the most suitable risk-reduction measures in drinking water systems requires
Received 1 March 2010
a thorough analysis of possible alternatives. In addition to the effects on the risk level, also
Received in revised form
the economic aspects of the risk-reduction alternatives are commonly considered impor-
20 May 2010
tant. Drinking water supplies are complex systems and to avoid sub-optimisation of
Accepted 16 July 2010
risk-reduction measures, the entire system from source to tap needs to be considered.
Available online 3 August 2010
There is a lack of methods for quantification of water supply risk reduction in an economic context for entire drinking water systems. The aim of this paper is to present a novel
Keywords:
approach for risk assessment in combination with economic analysis to evaluate
Risk reduction
risk-reduction measures based on a source-to-tap approach. The approach combines
Fault tree analysis
a probabilistic and dynamic fault tree method with cost-effectiveness analysis (CEA). The
Cost-effectiveness
developed approach comprises the following main parts: (1) quantification of risk reduc-
Decision support
tion of alternatives using a probabilistic fault tree model of the entire system; (2) combi-
Water safety plan
nation of the modelling results with CEA; and (3) evaluation of the alternatives with respect to the risk reduction, the probability of not reaching water safety targets and the cost-effectiveness. The fault tree method and CEA enable comparison of risk-reduction measures in the same quantitative unit and consider costs and uncertainties. The approach provides a structured and thorough analysis of risk-reduction measures that facilitates transparency and long-term planning of drinking water systems in order to avoid sub-optimisation of available resources for risk reduction. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Risk management of drinking water systems is becoming increasingly important worldwide. A safe and reliable drinking water supply is vital for public health as well as for economic development and other societal functions. Efficient risk management of drinking water systems requires a thorough evaluation of risk acceptance and identification of the most suitable measures to reduce unacceptable risks.
When analysing risks to a drinking water supply, the entire system needs to be considered to avoid overlooking important interactions between sub-systems and components. The World Health Organisation (WHO, 2008) thus concludes that a holistic risk management approach, including the entire system from catchment to consumer, is necessary to guarantee safe drinking water to consumers. Bartram et al. (2009), the WHO (2008), Davison et al. (2005) and many others advocate the preparation of Water Safety Plans (WSPs), in which
* Corresponding author. Tel.: þ46 31 772 20 60; fax: þ46 31 772 21 07. E-mail address:
[email protected] (A. Lindhe). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.07.048
242
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 1 e2 5 3
risks in source waters, treatment systems and distribution networks should be assessed and managed in an integrated, from source to tap, manner. Implementation of WSPs requires relevant decision support regarding how to manage existing risks. Since drinking water systems are diverse and are exposed to a wide variety of risks, one single method or tool cannot be used universally. A number of strategies and methods for risk assessment and risk management in drinking water supplies have been described, see e.g. MacGillivray et al. (2006), Rose´n et al. (2007) and Pollard (2008). The WHO (2008) suggests that risk assessment in WSPs is made by ranking of hazards using a risk matrix (see also Bartram et al., 2009; Davison et al., 2005). This type of qualitative (or semi-quantitative) method may be useful but have several limitations (e.g. Burgman, 2005; Cox, 2008). Risk ranking methods can only in a limited way consider complex systems with interactions between components. Risk ranking methods also assume a discrete nature of hazards, cannot provide quantitative estimates that can be compared with performance targets, and typically lack procedures for sensitivity and uncertainty analysis. As an alternative to risk ranking methods, Lindhe et al. (2009) presented a quantitative fault tree method for analysing entire drinking water systems. In a risk evaluation and risk management context, this method makes it possible to compare the risk level to acceptable levels in absolute terms and to estimate the effect of risk-reduction measures quantitatively. The primary aims of this paper were to apply the fault tree method presented by Lindhe et al. (2009) to model alternative risk-reduction measures and to combine the results with a cost-effectiveness analysis to provide decision support. Economic resources are often limited and economic evaluation of measures is therefore an important part of the decision-basis, see e.g. Levin and McEwan (2001) and Nas (1996). Simplified cost-benefit calculations were also conducted to allow for a broader economic analysis. Benefits and limitations with the applied approach were identified to evaluate the added value of combining quantitative risk assessment and economic evaluation. Seven alternative risk-reduction measures were analysed for the drinking water system in
Gothenburg, Sweden. The risk associated with each of the seven alternatives was compared to a politically established water safety target.
2.
In order to evaluate and compare risk-reduction measures, fault tree analysis was combined with cost-effectiveness analysis. For each measure, a fault tree model was constructed to represent the entire water supply system structure. The set of fault tree models provided information on the alternatives’ effect on the risk level. A politically established water safety target was used to define an acceptable level of risk. Information on costs was obtained from an economic analysis and the cost-effectiveness of each alternative was calculated. To further evaluate the alternatives, simplified cost-benefit calculations were conducted. The different parts of the applied method are presented in the subsequent sections.
2.1.
Conceptual model
Drinking water systems are composed of a set of sub-systems and components, e.g. raw water source, treatment plant and distribution system. Since these are interconnected, it is important to have a conceptual understanding of how failure events may occur, interact and finally cause problems to the consumers. Two important types of supply failure that may affect the consumers are: (1) quantity failure, i.e. no water is delivered to the consumer; and (2) quality failure, i.e. water is delivered but it does not comply with quality standards (Fig. 1). The applied method can be used to model both types of failure but in this paper only quantity failure was considered, i.e. the upper branch in Fig. 1. The causes of quantity failure may vary but were here categorised into two main types (Fig. 1). Technical components in the system, e.g. pumps or pipes, may break down and interrupt the supply of water. Also quality-related events may occur that cause an unacceptable water quality (raw water or drinking water) and if this is detected the water utility may decide to stop the delivery of water. Consequently, when modelling
Categories of supply failure
Quantity failure (Q = 0) No water is delivered to the co nsum er
Method
Causes Failure of components in the system (e.g. pumps or pipes) Events related to unacceptable water quality causing the water utility to stop the delivery
Supp l y fa ilur e Quality failure (Q > 0, C’) Water is delivered but does not comply with water quality standards
Unacceptable water quality is detected but no action is taken or it is not possible to stop the delivery Unacceptable water quality is not detected and no action is thus possible
Q = Flow (Q = 0, no water is delivered to the consumer; Q > 0, water is delivered) C’ = The drinking water does not comply with water quality standards
Fig. 1 e Categories of supply failure and their main causes (Lindhe et al., 2009).
243
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 1 e2 5 3
quantity failure in this paper, both failures of technical components and the quality-related failures were included. To model the drinking water system, a supply chain was considered including the raw water sources, treatment plants and the distribution system. These sub-systems interact and failure in one part may be compensated for by other parts. For example, if the treatment plant fails to produce drinking water, reservoirs in the distribution system and increased production at an additional treatment plant may prevent quantity failure. However, the ability to compensate is limited in time, due to limited reservoir volume. In some cases the ability to compensate may not exist at all.
2.2.
Fault tree analysis
The fault tree method used here was developed to facilitate integrated and probabilistic risk analysis of entire drinking water systems, see Lindhe et al. (2009). A fault tree is a logic diagram illustrating potential causes of system failure and particularly how different events may interact to cause system failure (e.g. Bedford and Cooke, 2001). The top event in the tree represents the critical system failure, in this case quantity-related supply failure. This event is further developed until a required level of detail is reached. At the lowest level of the tree the basic events are found, where input data are required. The capability of representing interactions between critical system components makes fault tree models powerful for analysis of entire drinking water systems. Traditionally, fault trees are used to calculate the probability of system failure. The applied method also enables calculation of the risk level, see Section 2.3. A Markovian approach was used (Norberg et al., 2009) to represent the dynamic, i.e. timedependent, behaviour of the events (Rausand and Høyland, 2004). Thus, variables failure rate l and mean downtime (duration of failure) 1/m were used to define each basic event in the model. The mean time to failure (uptime) is 1/l, hence the
probability of failure is PF ¼ l/(l þ m). The probability of failure is here identified with the probability that the system is down. This probability is typically time-dependent. However, in an ergodic system, it will tend to l/(l þ m) in the long run as time approaches infinity. In reliability theory, the latter is often referred to as the unavailability of the system. The main reasons for using the failure rate (or mean uptime) and mean downtime instead of direct estimations of the probability of failure, was to facilitate relevant elicitation of expert judgements and enable modelling of the system’s dynamic behaviour. Information on probability of failure as well as failure rate and downtime can be calculated for each event in the fault tree. The two most common types of logic gates are the OR- and AND-gates. The OR-gate describes a system where the occurrence of one input event causes system failure. The AND-gate describes a system where all input events have to occur simultaneously to cause the system to fail. To be able to consider the inherent ability of a system to compensate for failure, Norberg et al. (2009) and Lindhe et al. (2009) formulated two variants of the common type AND-gate. The first variant describes a situation where failure of one component may be compensated for by one or several other components during a limited period. The second variant is similar to the first one, but with the addition that a compensating component that has failed may recover and start to compensate again. Equations for calculating the probability of failure, failure rate and mean downtime at each intermediate level of the fault tree, are given in Table 1. Here, qi denotes the probability of failure on demand where compensating component i fails to start compensating when needed, e.g. a reserve pump that cannot be started when the main pump brakes down. A schematic fault tree illustrating the basic structure of the models used in this paper is shown in Fig. 2. Note that the top event, supply failure, in the fault tree in Fig. 2 includes quantity as well as quality failures, but in this study only quantity
Table 1 e Equations for calculating the output of the logic gates (Norberg et al., 2009). PF denotes the probability of failure, l the mean failure rate and 1/m the mean downtime. For the variants of the AND-gate i [ 1 corresponds to the failure event that may be compensated for by events i [ 2, ., n. For the second variant only one compensating event is considered, i [ 2. OR-gate l¼
n X
AND-gate
li
m¼
n X
i¼1
m¼
n X i¼1
n Q
li $
n Q
mi
i¼1
ðli þ mi Þ
i¼1
PF ¼
n Q
l¼
l¼
PF $m 1 PF
PF ¼
mi $
n Q
li
ðli þ mi Þ
n Q
li
i¼1
n Y l li ¼ l þ m i¼1 li þ mi
Second variant of AND-gate PF ¼
n Y li þ qi m 1
l1 $ l1 þ m1 i¼2 li þ m1
n Q i¼1
i¼1
i¼1
n Y l mi ¼1 l þ mi lþm i¼1 i
m ¼ m1
n X i¼1
mi
First variant of AND-gate
PF ¼
mi
i¼1
l1 l2 þ q2 ðm1 þ m2 Þ $ l1 þ m 1 l2 þ m 1 þ m 2
l¼
m1 l1 q2 ðl2 þ m1 þ m2 Þ þ l1 l2 ð1 q2 Þðm1 þ m2 Þ ðl1 þ m1 Þðl2 þ m1 þ m2 Þð1 PF Þ
m¼
m1 l1 q2 ðl2 þ m1 þ m2 Þ þ l1 l2 ð1 q2 Þðm1 þ m2 Þ ðl1 þ m1 Þðl2 þ m1 þ m2 ÞPF
244
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 1 e2 5 3
Supply failure
Raw water failure
Treatment failure
Raw water quantity failure (Q = 0)
Distribution failure
Treatment quantity failure (Q = 0)
Distribution quantity failure (Q = 0)
Quantity failure
Quantity failure
Treatment fails to compensate
Distribution fails to compensate
Distribution fails to compensate
Distribution quality failure (Q > 0, C')
Treament quality failure (Q > 0, C')
Raw water quality failure (Q > 0, C')
Quality failure
Distribution fails to compensate
Quality failure
Treatment fails to compensate Distribution fails to compensate
OR-gate First variant of AND-gate Q = flow (Q = 0, no water is delivered to the consumer; Q > 0 water is delivered) C' = The drinking water does not comply with water quality standards
Fig. 2 e Schematic fault tree including the main events that can be considered using the method (Lindhe, 2008).
failures were analysed (see Section 2.1). The fault tree models used in this study included approximately 120 basic events and about 100 logic gates.
2.3.
Risk expressed as Customer Minutes Lost (CML)
The risk (R) was estimated in terms of Customer Minutes Lost (CML). In this application CML corresponds to the number of minutes per year the average consumer is not supplied with drinking water. Risk was defined as the estimated value of CML, which is equal to the calculated mean value. A comprehensive description of the risk measure CML is presented by Lindhe et al. (2009). This measure has been used in, for example, the Netherlands as a performance indicator within the water industry and is commonly used in the energy sector (Blokker et al., 2005). It can be shown that the risk should be calculated as R ¼ PF$CF, where PF is the probability of failure and CF is the proportion of all consumers affected (Lindhe et al., 2009). The consequence, i.e. proportion of all consumers affected, was defined for n events in the fault tree where the top event was divided into well-defined categories of events. Consequently, the total risk was calculated as R¼
n X
PFk CFk
(1)
k¼1
2.4.
Uncertainties
To enable uncertainty analysis all input parameters were expressed as probability density functions and the
calculations were performed by Monte Carlo simulations (10,000 iterations), see Section 4.3. The exponential rates l and m were modelled by Gamma densities. The proportion of consumers affected (CF) and the probability of failure on demand (q) were modelled by Beta densities. The Gamma and Beta distributions facilitate a Bayesian approach, where new information, e.g. monitoring data, can be used for a mathematically formal updating of previous knowledge. The Monte Carlo simulations facilitate two important types of analyses: (1) sensitivity analysis of contributions to the total uncertainty from uncertainties in basic events; and (2) analysis of the probability of not meeting established safety targets.
2.5.
Modelling risk reduction
As a basis for this work an existing fault tree model of the drinking water system in Gothenburg was used (Lindhe, 2008; Lindhe et al., 2008). The existing model was reconstructed and updated in order to represent the system with the riskreduction measures implemented. Changes were made with respect to: (1) fault tree structure, i.e. events and logic gates were added and/or removed; (2) input data, i.e. new input data representing the situation as if measures have been implemented. The effect of each measure (Ej), i.e. the achieved risk reduction measured in CML units, was calculated as Ej ¼ R0 Rj
(2)
where R0 is the risk level prior to any measures and Rj is the residual risk after alternative j has been implemented.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 1 e2 5 3
The risk levels R0 and Rj were calculated according to Equation (1).
2.6.
Cost-effectiveness analysis
Cost-effectiveness analysis (CEA) can be described as a tool to identify what alternative meets a specified target to the lowest cost (e.g. Levin and McEwan, 2001). However, a CEA may also provide additional information that can be used in the decision-making process. The main purpose of CEA is to provide a combined, not separate, and proper assessment of both the costs and the effects. The fault tree models were used to estimate the effect of the measures (Ej). To estimate the total cost of each alternative, possible costs for planning, constructing and maintenance were identified and assessed. Costs were assessed considering similar actions within other projects. All alternatives were analysed based on a 100 year time horizon (T) and since costs (Cjt) occur over several years, the present value (Cj) was calculated according to Equation (3). Recommendations regarding discount rates to be used in economic analyses vary. Stern (2007) advocates a discount rate of 1.4% for analyses linked to climate changes and the Swedish Environmental Protection Agency (Lindstedt et al., 2003) recommends a rate of 4% to be used when assessing environmental consequences. In this study a discount rate (r) of 3% was used. A sensitivity analysis of the discount rate was also performed. Cj ¼
T X
Cjt ð1 þ rÞ
A cost-effectiveness ratio (CER) was calculated for each alternative. The ratio corresponds to the cost required to obtain a single unit of effect, i.e. reduction of one CML. The effect (Ej) was calculated according to Equation (2) and was assumed to be constant over time. Therefore, the yearly effect did not need to be discounted or summed up over the chosen time horizon in order to provide the correct relation between the final CER values of the alternatives. The calculation of CER could therefore be simplified to Equation (4). For a discussion regarding discounting effects in a CEA, when the effect is not constant over time, see e.g. Ramsey et al. (2005) and Brouwer and Koopmanschap (2000). CERj ¼
Cj Ej
2.7.
(4)
Cost-benefit analysis
For assessing whether an action is societally beneficial, a costbenefit analysis (CBA) can be conducted. In contrast to CEA, CBA aims to determine the cost as well as the effect and other benefits in monetary units (e.g. Johansson, 1993; Nas, 1996). The result of a CBA may be presented as Fj ¼
T X
1 t1
t¼1
ð1 þ rÞ
discount rate (3%). A positive net benefit implies that the alternative is desirable. In this application a simplified CBA was performed and the only benefit included in the calculations was the alternatives’ effect on the risk level (Ej). Consequently, the benefit was calculated as Bjt ¼ Ej cn ¼ R0 Rj cn
(6)
where R0 is the risk level prior to any measures, Rj is the residual risk for alternative j, c is the economic value of 1 min additional water supply per year and consumer and n is the total number of consumers. The value of c thus represents the cost an average consumer is willing to pay per year to reduce the time of interruption by 1 min, i.e. one CML, per year. In a complete CBA the objective function (Fj) represents the net present benefit value expressed in monetary units. In this application, only the change in risk level was considered as a benefit. No consideration was taken to when in time this change occurs. Furthermore, the benefit was not valued in monetary units, i.e. the value of c was not determined. Instead, the objective function was plotted as a function of c, i.e. Fj(c). This illustrates how the economic valuation of the effect may affect the selection of alternatives and clearly shows break-even points of different alternatives.
3.
Case study site
3.1.
System description
(3)
t1
t¼1
245
Bjt Cjt
The drinking water system in Gothenburg, Sweden, was used to exemplify method application. The supply system is based solely on surface water, includes two treatment plants, and has approximately 500,000 consumers. A schematic description of the raw water system in Gothenburg is given in Fig. 3. ¨ lv River. In addition, The main raw water source is the Go¨ta A two smaller interconnected lakes (main reservoir lakes in Fig. 3) are used for intermediate storage of water and to improve the water quality. Approximately half of the water taken from the river is transferred directly to treatment plant no 1. The other half is transferred via a 12 km rock tunnel to the main reservoir lakes. From the main reservoir lakes, water is pumped to treatment plant no 2. Due to the variable quality of the river water, the intake is closed 100 days a year on average (e.g. ˚ stro¨m et al., 2007). Decisions to close the intake are based on A online monitoring and reports from operating bodies upstream, e.g. industries and municipalities. When the intake from the river is closed, the main reservoir lakes supply both treatment plants. During long periods (weeks) of closure of the intake, water from a reservoir in an additional raw water supply system (lower additional reservoir in Fig. 3) can be pumped to the main reservoir lakes or directly to treatment plant no 2. However, the situation where treatment plant no 1 has no alternative raw water supply of its own makes the system vulnerable.
(5)
where Fj is an objective function representing the net benefit of alternative j, Bjt and Cjt are the streams of benefits and costs over time, T is the time horizon (100 years) and r is the
3.2.
Acceptable risk
The City of Gothenburg has defined quantitative water safety targets as a basis for long-term planning of investments and
246
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 1 e2 5 3
Supply from large lake, alternative 3
Existing water source / reservoir Supply from small lakes, alternative 2
Possible future water source
Water treatment plant Raw water distribution Upper additional reservoir
Possible future raw water distribution
Main reservoir lakes
M ai n
ra w
w at er so ur ce
Watercourse
Water treatment plant no 1
Water treatment plant no 2
Lower additional reservoir
Fig. 3 e Schematic description of the Gothenburg raw water system. Possible future raw water supplies included in the analysis are further described in Section 3.4.
reinvestments (Go¨teborg Vatten, 2006). These targets were compiled and suggested by the water utility organisation and then decided by the politicians in the city. Hence, the targets are confirmed at the political level and can be considered as acceptable levels of risk. The target used in this analysis is: Duration of interruption in delivery to the average consumer shall, irrespective of the reason, be less than a total of 10 days in 100 years. For the average consumer this translates to a target value of 144 annual CML.
3.3.
Current situation
Previous risk analyses have shown high probability of interruption in the supply to one of the two treatment plants due to failures in the raw water system (Lindhe, 2008; Lindhe et al., 2008; Rose´n and Steier, 2006). The maximum capacity of each plant is far below the average demand of the city and drinking water storage is well below the daily need. An additional analysis of the dominating failure scenario showed that increased treatment capacity would reduce the risk of extensive delivery interruptions to an acceptable level, and that the raw water supply would not be a limiting factor in this scenario. As a result, increased treatment capacity is considered as a possible risk-reduction measure. However, there are other scenarios, including pollution of the reservoir lakes, where additional raw water sources would be crucial for sufficient drinking water production. Several options for additional raw water sources have been identified.
3.4.
Risk-reduction alternatives
The three main risk-reduction measures considered relevant for the system and included in this study are increased treatment capacity, supply from small lakes and supply from a large lake. In addition to the three separate measures, also four combinations of these were analysed, see also Rose´n et al. (2010). All alternatives are presented below and the possible raw water supplies are illustrated in Fig. 3.
3.4.1.
Alt. 1: increased treatment capacity
The treatment capacity is increased at both treatment plants so that each plant manages to produce up to the average total water demand. Consequently, each plant will get an enhanced ability to compensate for shortage in raw water supply and failures in the other treatment plant. To model the increased treatment capacity, no additional events or changes in the fault tree structure were needed. However, the input data was changed for four events describing the plants’ ability to compensate for failure. Also the consequences for four events were changed. Based on statistical data on water demand and estimations regarding the reliability of the treatment plants, the time for compensation (i.e. time to failure or uptime, 1/l) was estimated to 3e120 days (90%-interval). The probability of failure on demand was estimated to 0.0025e0.01 (90%-interval). The increased treatment capacity may also reduce the number of consumers affected when a failure occurs. Thus, the number of people affected were estimated to 6,500e33,900 (90%interval) for four events.
3.4.2.
Alt. 2: supply from small lakes
Some fairly pristine lakes are regulated to increase the flow in a small river for transfer to the drinking water system. These lakes contain relatively large water volumes but their watershed is too small for a continuous supply. The lakes will work as reservoir lakes and water from the connected river will be pumped into the existing raw water system (Fig. 3). The lakes were included in the fault tree model as an alternative water source for each treatment plant, modelled using AND-gates. To cause failure, all water sources have to be unavailable simultaneously for the specific treatment plant. For treatment plant no 2 some events had to be restructured to consider that failures in the existing system also may cause the new water source to become unavailable. When only treatment plant no 1 is supplied the uptime was estimated to 25e35 days and for both plants it is 8e18 days (90%-intervals). For both cases the downtime was estimated to 7e60 days (90%-interval).
247
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 1 e2 5 3
3.4.3.
Alt. 3: supply from large lake
The largest lake in the area is regulated and a pipeline is constructed for raw water transfer to the drinking water system (Fig. 3). The lake may provide a high availability to raw water but failures may occur due to: (1) unacceptable water quality, e.g. pollution events; (2) insufficient supply of water due to dry periods and legal restrictions; and (3) failure of technical components making transfer of water to the treatment plants impossible. All three events were assessed to have the same uptimes of 5e15 years and downtimes of 5e30, 1e30 and 0.5e2 days respectively (90%-intervals). The large lake was included in the fault tree in the same way as the small lakes in alternative 2. However, an OR-gate was also used to consider that the lake will become unavailable if one of the three events listed above occurs.
3.4.4.
Alt. 4e7: combinations of measures
Alternatives 4e7 are combinations of the first three measures. Consequently, the fault tree models representing these alternative are combinations of the changes presented above. The following combinations were considered: -
Alt. 4: Combination of alternatives 1 and 2. Alt. 5: Combination of alternatives 1 and 3. Alt. 6: Combination of alternatives 2 and 3. Alt. 7: Combination of alternatives 1, 2 and 3.
4.
Results
4.1.
Current system
Previous studies of the current system structure have shown a significant probability of failures that may cause severe supply interruptions. Furthermore, it has been concluded that the raw water supply contributes most to the risk. A risk level of 612 CML (mean value) was calculated using the fault tree model of the current system. This level is clearly above the specified water safety target of 144 CML per year for
the average consumer. In Fig. 4 an uncertainty distribution of the calculated risk level and its relation to the safety target are presented. The probability of not meeting the target value was calculated to be 0.84. Consequently, the current system most likely has an unacceptable risk level and measures are thus required.
4.2.
Risk reduction
The mean value and percentiles of the empirical distributions of the current risk level and the residual risk levels after implementation of the different measures are presented in Fig. 5. The histograms in Fig. 5 illustrate uncertainties in risk values, originating from uncertainties in input data. Based on the histograms it can be concluded that for some alternatives (alt. 2, 3 and 6) the probability of exceeding the safety target is substantial, whereas for others (alt. 1, 4, 5 and 7) the probability is considerably lower. If the estimated risk (mean value) is used as a criterion to select what alternatives that meet the safety target, only alternatives 1, 4, 5 and 7 should be approved. Although the mean risk level is below the safety target, there may be a probability of exceeding the target value that cannot be ignored. Based on the uncertainties of the risk levels, the probably of exceeding the safety target was calculated for each alternative (Fig. 6). A highest acceptable probability of not meeting the target value should be considered, since it influences what alternatives can be accepted. For example, an acceptable probability of 0.10 makes alternative 1 acceptable, whereas a probability of 0.05 disqualifies it (see also Section 5). The risk reduction, i.e. effect, of the different alternatives can be seen from Fig. 5. Among the tree alternatives that are not combinations of measures (alt. 1e3), increased treatment capacity (alt. 1) has the largest effect on the risk. The risk level is reduced from 612 to 81 CML and the probability of not meeting the safety target is reduced from 0.84 to 0.08. Among the alternative raw water sources (alt. 2 and 3), the large lake (alt. 3) is most effective with respect to both the mean risk level and the probability of not meeting the target. However, 1800 P05 1600
0.05
Mean P95
1400 1200 CML
Probability
0.04 0.03
1000 800 612 600
0.02
364
400
0.01
188
200 81 0
0
0
500
1000 CML
1500
2000
Fig. 4 e Uncertainty distribution of the risk for the current system. The safety target (144 CML per year for the average consumer) is indicated by a solid vertical line (Lindhe et al., 2009).
0
1
2
182 59
52
3 4 Alternative
5
50 6
7
Fig. 5 e Histograms showing mean, 5- and 95-percentiles of simulated risk levels for the current situation (0) and the seven alternatives (1e7). The mean values are given at the mean value bars and the safety target (144 CML) is indicated by a solid horizontal line.
248
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 1 e2 5 3
that when the treatment capacity has been increased the raw water system is no longer the dominating contributor to the total risk level. Thus, after installing increased treatment capacity also measures in the treatment and distribution systems will be important for further risk reduction.
1 0.84
P(R>Rc)
0.8
0.73
0.6 0.42
0.41
4.3.
0.4 0.2 0
0.08
0
1
0.05 0.04 2
3 4 Alternative
5
0.04 6
7
Fig. 6 e Probabilities of exceeding the acceptable risk level (Rc) defined by the water safety target (144 CML) for current situation (0) and the seven alternatives (1e7).
both the small lakes (alt. 2) and the large lake (alt. 3) have risk levels (364 and 188 CML respectively) above the acceptable level and the probability of exceeding it is considerable, 0.73 and 0.42 respectively. Combining increased treatment capacity with either regulation of the small lakes (alt. 4) or the large lake (alt. 5) results in further reduction of the risk level, from 81 CML to 65 and 54 CML respectively. These levels are clearly below the target value. The probability of exceeding the safety target is 0.05 for alternative 4 and 0.04 for alternative 5. Regulation of both the large and small lakes without measures in the treatment plants (alt. 6) results in a risk level of 187 CML. This is above the safety target and the probability of not meeting the target value is 0.41. Combining all measures (alt. 7) provides the greatest risk reduction, however, not substantially larger than for alternatives 4 and 5. The three alternatives (4, 5 and 7) have a probability of 0.04 or 0.05 of not meeting the safety target. The main reason for the small difference between the alternatives is
Risk levels
The fault tree method is probabilistic to include uncertainties in input data, which makes it possible to analyse uncertainties in results. In addition, it is important to evaluate the uncertainties depending on the selected number of iterations in the Monte Carlo simulations. If not enough iterations are made, the standard error of the mean (estimate) will be high, resulting in a low accuracy of the calculations. Non-parametric bootstrapping (e.g. Efron and Tibshirani, 1993) was used to evaluate the uncertainties in the simulation results. This means that statistics were sampled empirically by repeatedly sampling data. Thus, the model data (10,000 iterations) were sampled once and new samples were then created by resampling from this data set. In total 1000 resamplings were made, each including 10,000 samples. Replacement was applied so that each sample can be resampled more than once. By means of bootstrapping the 95 percent upper confidence limit (UCL95) of the mean was calculated for the risk values and for the probabilities of exceeding the safety target (Fig. 7). The bars in both plots in Fig. 7 show small differences between mean and UCL95 values. Thus, simulations based on 10,000 iterations were considered to give results with acceptable uncertainty since using UCL95 instead of mean values when comparing risk levels with the safety target does not affect the results. Values close to target values are especially important to scrutinise. The estimated value of the probability of exceeding the safety target is close to, or equal to, 0.05 for alternatives 4, 5 and 7 (Fig. 6). The UCL95 values are only slightly larger (0.05 for all alternatives) and due to the probabilistic approach applied, with uncertainties in input data, values close to target values should be evaluated with cautiousness.
b
400
Comparison with water safety target 0.8
350
0.7
300
0.6
250
0.5 P(R>Rc)
CML
a
Uncertainty analysis
200
0.4
150
0.3
100
0.2
50
0.1
0
1
2
3
4 5 Alternative
6
7
0
1
2
3
4 5 Alternative
6
7
Fig. 7 e Mean (light colours) and UCL95 values (dark colours) of risk levels (a) and probabilities of exceeding the safety targets (b) for the alternatives.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 1 e2 5 3
900 1% 2% 3% 4% 5%
800 700
249
Combining the cost and effect of each alternative, the costeffectiveness ratio (CER) was calculated (Fig. 9). Based only on the CER, alternative 2 is clearly more cost-effective than the others. However, since also the safety target and the ability of the alternatives to meet the target have to be considered, further evaluation is required (see Section 5).
Cost [MSEK]
600
4.5.
500 400 300 200 100 0
7
5
6
3 4 Alternative
1
2
Fig. 8 e Ranked costs (Cj) for the seven alternatives using different discount rates. The alternatives are ranked from highest to lowest cost. The 3% discount rate was used for the subsequent calculations.
4.4.
Cost-effectiveness
The total cost of each alternative based on different discount rates is presented in Fig. 8 (10 SEK w 1 EUR; w 1.2 USD). In Fig. 8 the costs for the alternatives are showed in ranked order. Ranked from highest to lowest cost the alternatives are sorted as: 7, 5, 6, 3, 4, 1 and 2. It can be seen that the order does not change with respect to the discount rate. It can also be seen in Fig. 8 that a change in discount rate has the largest affect on high costs. In the subsequent calculations the 3% discount rate was used. If the alternative of increased treatment capacity (alt. 1) is chosen, this will probably include additional actions aimed at reducing odour problems and improving the working environment. The costs, and benefits, included in this analysis are only the ones related to actions needed to increase the treatment capacity.
1.5 1.17
CER [MSEK/CML]
1.16 1
0.88
0.87
0.53
The net benefit of all alternatives is presented in Fig. 10 as a function of the economic value of 1 min additional water supply per year and consumer. If the economic value of risk reduction (c) is lower than 0.03 SEK, only alternative 2 provides a positive net benefit. From a strict cost-benefit approach only alternatives resulting in a positive net benefit should be approved. In this application, however, the only benefit included was the risk reduction. Consequently, Fig. 10 should mainly be used to identify and discuss differences between the alternative measures. It can, for example, be seen that whether the value of c is above or below the intersection points at 0.06 and 0.13 clearly affects the relationship between the alternatives. Furthermore, if c is approximately 0.07 SEK or higher, all alternatives result in a positive net benefit (Fig. 10). As an example, the value of c has to be at least 0.033 SEK for alternative 1 to result in a positive net benefit. This alternative reduces the risk by 531 CML, which result in a total cost per year and consumer of 17.5 SEK. This figure can be compared to the annual cost for drinking water, which is approximately 400 SEK per consumer in Gothenburg. In addition to this cost there is a fixed service cost for drinking water and wastewater, which is 2542 SEK per year for a single-family house. Alternatives 2, 3 and 6 do most likely not meet the safety target (Figs. 5 and 6). Among the remaining options, alternatives 1 and 4 as well as 5 and 7 have similar net benefits. The first set of alternatives (alt. 1 and 4) is mainly characterised by the large risk reduction by increased treatment capacity, while the second set (alt. 5 and 7) is characterised by the high cost of regulating the large lake. The difference between these sets of alternatives is fairly similar over the range of c values included in Fig. 10. At c equal to 0.1 SEK the difference is approximately 340 MSEK. Thus, if the large lake provides additional benefits (i.e. not included in this analysis) of more than 340 MSEK, alternatives 5 and 7 may result in the highest net benefits. Furthermore, if c is equal to approximately 2.5 SEK alternative 7 becomes most beneficial.
5.
0.52
Cost-benefit
Discussion
0.5
0.04 0
1
2
3
4 5 Alternative
6
Fig. 9 e Cost-effectiveness ratio (CER) for the seven alternatives.
7
The results clearly show that the risk for the current system is not acceptable and that measures are required. What riskreduction measure to select depends on the criteria for the decision. A summary of the alternatives is presented in Table 2 and it is indicated whether the estimated risks (Rj) exceed the target value of 144 CML (Rc) or not. Furthermore, two examples of a certainty criterion are included, representing the highest acceptable probability of not meeting the target
250
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 1 e2 5 3
300
1500
b
Alt. 4 Alt. 1
Φ Net Benefit [MSEK]
1000
Φ Net Benefit [MSEK]
a
250
200
150 0.05
500
0.055
c [SEK]
0.06
0.065
0.13
0.135
550
Alt. 2
c
Alt. 3 Φ Net Benefit [MSEK]
0 Alt. 1 Alt. 6
Alt. 2 Alt. 3
Alt. 5
Alt. 4
−500
Alt. 5 Alt. 6
Alt. 7
Alt. 7 0
500
0.05
0.1 c [SEK]
0.15
0.2
450 0.12
0.125 c [SEK]
Fig. 10 e Net benefit of alternatives as a function of the economic value of 1 min additional water supply per year and consumer (c). In Figure (a) all alternatives are included and in Figures (b) and (c) the two intersection points have been zoomed in. The same legend is used in all figures.
value, i.e. P(Rj > Rc). The probabilities used in our example are 0.10 and 0.05. It should be emphasised that the decisionmakers have to define the certainty criterion; it is not the task of the risk analysts. As can be seen in Table 2 alternatives 2, 3 and 6 exceed both probabilities. Alternative 1 may be accepted if the certainty criterion is represented by the highest (0.10) probability. Thus, the required level of certainty has an impact on what alternatives that can be accepted. Among the three alternatives meeting all criteria (alt. 4, 5 and 7), increased treatment capacity in combination with regulation of the small lakes (alt. 4) has the lowest cost and CER. Hence, if the certainty criterion is represented by an acceptable probability of 0.05, alternative 4 is the most costeffective alternative. If the certainty criterion is represented by an acceptable probability of 0.10, increased treatment capacity alone (alt. 1) is most cost-effective since it has the
lowest cost. However, to fully understand the results a couple of different aspects have to be considered, see below.
5.1.
Target values and the CER
Although an alternative results in a low CER, such as alternative 2, it cannot be considered cost-effective if the acceptable level of risk is exceeded (Table 2). In the kind of case study performed here, a predefined target value is necessary for a CEA to be useful. If the aim is to achieve the largest risk reduction for a certain amount of money, irrespective of the residual risk, no target value is needed. However, to consider both the cost and effect in a sensible way, a target value is required. If no target value exists and only the CER values are analysed, this may cause (i) insufficient risk reduction as well as (ii) unnecessarily large reduction. The first case (i) is
Table 2 e Summary of risk levels (Rj), probabilities of not meeting the safety target P (Rj > Rc), costs (Cj) and costeffectiveness ratios (CERj) for the alternatives. The safety target value is 144 CML. For the probability of not meeting the safety target two example criteria are used (0.10 and 0.05). Bold values indicate that the criterion is not met. Criterion Alternative 1. 2. 3. 4. 5. 6. 7.
Increased capacity Supply from small lakes Supply from large lake Combination of alt. 1 and 2 Combination of alt. 1 and 3 Combination of alt. 2 and 3 Combination of alt. 1, 2 and 3
144
0.10
0.05
Rj [CML]
P(Rj>Rc)
P(Rj>Rc)
Cj [MSEK]
CERj [MSEK/CML]
81 364 188 59 52 182 50
0.08 0.73 0.42 0.05 0.04 0.41 0.04
0.08 0.73 0.42 0.05 0.04 0.41 0.04
280 9 372 289 652 381 661
0.53 0.04 0.87 0.52 1.16 0.88 1.17
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 1 e2 5 3
Table 3 e Example of two fictitious alternatives to illustrate limitations with the CER. For each alternative the residual risk (Rj), cost (Cj) and cost-effectiveness ratio (CERj) are presented. The current risk level (R0) is assumed to be the same as in the case study (612 CML). Alternative A B
Rj [CML]
Cj [MSEK]
CERj [MSEK/CML]
144 12
420 450
0.90 0.75
exemplified by alternative 2 in this application (Table 2). Although the safety target is not met the CER may seem favourable if the cost, in relation to the effect, is low. The second case (ii) may, for example, occur if one alternative reduces the risk far below the acceptable level and another alternative reduces the risk just to the target value. In Table 3 two alternatives are presented as examples to illustrate this possible problem. Alternative A reduces the risk just to the acceptable level of 144 CML to a cost of 420 MSEK. Alternative B on the other hand, reduces the risk to a level of 12 CML but to a cost of 450 MSEK. Since the safety target is 144 CML, both alternatives provide an acceptable risk. To simplify the example we do not consider uncertainties in risk levels and costs. If the CER values are compared, alternative B appears to be most cost-effective (Table 3). However, alternative A meets the safety target to a lower cost. One may argue that there is no use of reducing the risk below the acceptable risk level. It would be better to use these resources for other purposes. Consequently, the CER should be used to illustrate the cost of reducing the risk one unit but not as the single criterion for evaluating alternatives. An alternative way of calculating the effect (Ej) could be used to avoid the problem with CER values described above. By not considering risk reduction below the safety target (Rc), the effect could be calculated as Ej ¼ min R0 Rj ; R0 Rc
(7)
where R0 is the risk level prior to any measures, Rj the residual risk for alternative j. Thus, the largest possible effect is R0Rc. Using Equation (7) for the two examples in Table 3, results in a CER of 0.96 for alternative B and no changes for alternative A. Based on the alternative approach alternative A is most costeffective (0.90 < 0.96).
5.2.
Aspects included
Considering only the three first alternatives, increased treatment capacity (alt. 1) provides a substantial risk reduction compared to the alternative raw water supplies (alt. 2 and 3). It can be concluded that the capacity of the treatment plants is a bottleneck in the system that cannot be solved only by increased access to raw water. Although increased treatment capacity has a large effect on the risk level, additional measures may be needed depending on the required level of certainty. The combination of increased treatment capacity and both the large and small lakes (alt. 7) provides the largest risk reduction. However, if increased treatment capacity is
251
combined with only the large lake (alt. 5) a risk reduction almost equivalent to alternative 7 is obtained. Increased treatment capacity in combination with the small lakes (alt. 4) also results in similar residual risk. The only combination of measures not meeting the target value is the combination of the large and small lakes (alt. 6). This alternative provides limited additional risk reduction compared to using the large lake only (alt. 3). Thus, including the large lake, solely or combined with increased treatment capacity, the small lakes provide only marginal additional risk reduction. Considering the effect and the cost, alternatives 1 and 4 seems to be the best options. A major advantage of the applied method is the possibility to measure each alternative’s effect on the risk using the same unit (CML). The use of CML as a performance indicator is common in the energy sector and has been used in the drinking water sector in the Netherlands. This study shows that CML is an applicable unit to be used when evaluating and comparing measures for reducing water supply interruptions, i.e. quantity failures. However, alternatives may have other positive effects that are not related to quantity failure. For example, regulation of the large lake will most probably also provide a raw water of better and more stable quality compared to the current main raw water source. The costbenefit calculations (Fig. 10) show that a costly alternative may result in a positive net benefit if the value of risk reduction is high or large benefits are obtained. If a full CBA would have been performed, it may not have provided recommendations similar to those of the CEA, e.g. because of possible health benefits provided by the large lake but not considered in the CEA. The reduced quality risk, i.e. related to human health effects, was not considered in this analysis but could also be assessed using the fault tree method. Another aspect not explicitly included in the CEA is the time for implementing measures. For example, the inclusion of the large lake is expected to be a rather complicated and long process, for various reasons. Assume that increased treatment capacity combined with the large lake (alt. 5) is selected as the most suitable alternative. Furthermore, increased treatment capacity by itself is not associated with an acceptable risk. Since the large lake takes a long time to implement, it might be reasonable to first combine increased treatment capacity with the small lakes (alt. 4). Thus, when the large lake is installed, the small lakes may be excluded. In this way the risk is not unacceptable during the installation of the large lake. The CEA, as applied here, includes one criterion related to the risk level and one to the cost. To further develop the approach, additional criteria can be included and a multicriteria decision analysis (MCDA) applied. A MCDA would make it possible to include additional types of risk, such as the quality risk, and also additional criteria for evaluating the performance of alternative measures. One criterion could be linked to, for example, the time it takes to implement measures. If the aim is to evaluate how well alternative measures contribute to a sustainable supply system, they can be assessed based on criteria linked to environmental, economic and socio-cultural dimensions. In a MCDA it is also possible to take into account that all criteria may not be regarded as equally important.
252
5.3.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 1 e2 5 3
Uncertainties and cost
The uncertainty analysis of the costs was related to the discount rate. By calculating the present value for each alternative using different discount rates, the sensitivity of the results with respect to this variable was analysed. Although uncertainties in input data are not considered, this type of analysis is important to know whether the value of parameters affects the final prioritisation of alternatives.
5.4.
The method
The applied fault tree analysis in combination with CEA illustrates its possibilities as a decision support tool for structured comparison of alternative risk-reduction measures. The fault tree analysis makes it possible to model alternatives that influence different parts of a system and to quantify the risk reduction. The method facilitates two very important analyses: (1) analysis of the risk reduction provided by a specific measure; and (2) analysis of the uncertainty of the risk estimate, which enables an estimation of the probability of not meeting acceptable risk levels. Combination of the quantitative fault tree analysis and economic evaluation provides transparency and a basis for long-term planning of drinking water systems.
6.
Conclusions
The main conclusions of this study are: - The fault tree method enables comparison of risk-reduction alternatives in the same quantitative unit and for an entire drinking water system from source to tap. Interactions between events and components of the system can be modelled in a realistic way. Furthermore, the probabilistic approach enables comparisons with safety targets and the probability of exceeding these target values can be calculated. - A CEA provides useful information by combing the effect and cost. However, it is important to understand existing pitfalls when interpreting the results. For example, alternatives may provide additional benefits not considered in the CEA and alternatives cannot be evaluated solely based on cost-effectiveness ratios. - A CEA requires a safety target representing the acceptable level of risk. In addition to the target value it is also important to consider the highest acceptable probability of not meeting the target value, i.e. a certainty criterion. - A multi-criteria decision analysis could be applied to further develop the applied approach and include additional criteria for evaluating the performance of risk-reduction measures. Combining quantitative risk analysis tools, such as the applied fault tree method and economic evaluation, provides a powerful tool for decision-makers. A combined quantitative risk assessment and economic evaluation can provide a structured and thorough analysis of risk-reduction measures that facilitates transparency and long-term
planning of drinking water systems in order to avoid suboptimisation of available resources for risk reduction.
Acknowledgements This research was sponsored by The Swedish Water & Wastewater Association, the City of Gothenburg and the TECHNEAU project, a EC-funded FP6 project (www.techneau. org). We would like to thank the City of Gothenburg for its valuable and fruitful collaboration. We also appreciate the comments and suggestions given by the anonymous reviewers.
references
˚ stro¨m, J., Pettersson, T.J.R., Stenstro¨m, T.A., 2007. Identification A and management of microbial contaminations in a surface drinking water source. Journal of Water and Health 5 (Suppl. 1), 67e79. Bartram, J., Corrales, L., Davison, A., Deere, D., Drury, D., Gordon, B., Howard, G., Rinehold, A., Stevens, M., 2009. Water Safety Plan Manual: Step-By-Step Risk Management for Drinking-Water Suppliers. World Health Organization, Geneva. Bedford, T., Cooke, R.M., 2001. Probabilistic Risk Analysis: Foundations and Methods. Cambridge University Press, Cambridge. Blokker, M., Ruijg, K., de Kater, H., 2005. Introduction of a substandard supply minutes performance indicator. Water Asset Management International 1 (3), 19e22. Brouwer, W.B.F., Koopmanschap, M.A., 2000. On the economic foundations of CEA. Ladies and gentlemen, take your positions! Journal of Health Economics 19 (4), 439e459. Burgman, M.A., 2005. Risks and Decisions for Conservation and Environmental Management. Cambridge University Press, Cambridge. Cox, A.L., 2008. What’s wrong with risk matrices? Risk Analysis 28 (2), 497e512. Davison, A., Howard, G., Stevens, M., Callan, P., Fewtrell, L., Deere, D., Bartram, J., 2005. Water Safety Plans: Managing Drinking-Water Quality from Catchment to Consumer. WHO/ SDE/WSH/05.06. World Health Organization, Geneva. Efron, B., Tibshirani, R., 1993. An Introduction to the Bootstrap. Chapman & Hall, New York. Go¨teborg Vatten, 2006. Action Plan Water: Long-term Goals for the Water Supply in Gothenburg (In Swedish) City of Gothenburg. Johansson, P.-O., 1993. Cost-benefit Analysis of Environmental Change. Cambridge University Press, Cambridge, New York. Levin, H.M., McEwan, P.J., 2001. Cost-effectiveness Analysis: Methods and Applications, second ed. Sage Publications, Thousand Oaks, California. Lindhe, A., 2008. Integrated and probabilistic risk analysis of drinking water systems, Licentiate Thesis No. 2008:8, Chalmers University of Technology, Go¨teborg. Lindhe, A., Rose´n, L., Norberg, T., Bergstedt, O., 2009. Fault tree analysis for integrated and probabilistic risk analysis of drinking water systems. Water Research 43 (6), 1641e1653. Lindhe, A., Rose´n, L., Norberg, T., Petterson, T.J.R., Bergstedt, O., ˚ stro¨m, J., Bondelind, M., 2008. Integrated risk analysis from A source to tap: Case study Go¨teborg, In: 6th Nordic Drinking Water Conference, pp. 231e241.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 1 e2 5 3
Lindstedt, U., Ingelsson, M., Engleryd, A., 2003. Consequence Analysis Step-By-Step: Guidance on Economic Consequence Analysis for the Environmental Protection Agency (In Swedish). Swedish Environmental Protection Agency, Stockholm. MacGillivray, B.H., Hamilton, P.D., Strutt, J.E., Pollard, S.J.T., 2006. Risk analysis strategies in the water utility sector: an inventory of applications for better and more credible decision making. Critical Reviews in Environmental Science and Technology 36 (2), 85e139. Nas, T.F., 1996. Cost-benefit Analysis: Theory and Application. Sage Publications, Thousand Oaks, Calif. Norberg, T., Rose´n, L., Lindhe, A., et al., 2009. Added value in fault tree analyses. In: Martorell, S. (Ed.), Safety, Reliability and Risk Analysis: Theory, Methods and Applications. Taylor & Francis Group, London, pp. 1041e1048. Pollard, S.J.T., 2008. Risk Management for Water and Wastewater Utilities. IWA Publishing, London. Ramsey, S., Willke, R., Briggs, A., Brown, R., Buxton, M., Chawla, A., Cook, J., Glick, H., Liljas, B., Petitti, D., Reed, S., 2005. Good research practices for cost-effectiveness analysis alongside clinical trials: the ISPOR RCT-CEA task force report. Value in Health 8 (5), 521e533.
253
Rausand, M., Høyland, A., 2004. System Reliability Theory: Models, Statistical Methods, and Applications, second ed. Wiley-Interscience, N.J. Rose´n, L., Hokstad, P., Lindhe, A., Sklet, S., Røstum, J., 2007. Generic FRAMEWork and Methods for Integrated Risk Management in Water Safety Plans. Deliverable no. D 4.1.3, D 4.2.1, D 4.2.2, D 4.2.3. TECHNEAU. Rose´n, L., Lindhe, A., Bergstedt, O., Norberg, T., Pettersson, T.J. R., 2010. Comparing risk-reduction measures to reach water safety targets using an integrated fault tree model. Water Science and Technology: Water Supply 10 (3), 428e436. Rose´n, L., Steier, K., 2006. Risk Analysis: Interruptions in Raw Water Supply to the Gothenburg Drinking Water System (In Swedish), Consulting report, Contract no. 1310786. SWECO VIAK AB. Stern, N., 2007. The Economics of Climate Change: The Stern Review. Cambridge University Press, Cambridge. WHO, 2008. Guidelines for Drinking-Water Quality [Electronic Resource]: Incorporating First and Second Addenda. In: Recommendations, third ed., vol. 1. World Health Organization, Geneva.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 4 e2 6 2
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Aerobic treatment of N-nitrosodimethylamine in a propane-fed membrane bioreactor Paul B. Hatzinger a,*, Charles Condee a, Kevin R. McClay a, A. Paul Togna b a b
Shaw Environmental, Inc., 17 Princess Road, Lawrenceville, NJ 08648, USA Envirogen Technologies, Inc., 250 Phillips Boulevard, Suite 255, Ewing, NJ 08618, USA
article info
abstract
Article history:
N-Nitrosodimethylamine (NDMA) is a suspected human carcinogen that has recently been
Received 18 February 2010
detected in wastewater, groundwater and drinking water. Treatment of this compound to
Received in revised form
low part-per-trillion (ng/L) concentrations is required to mitigate cancer risk. Current
16 June 2010
treatment generally entails UV irradiation, which while effective, is also expensive. The
Accepted 18 July 2010
objective of this research was to explore potential bioremediation strategies as alternatives
Available online 29 July 2010
for treating NDMA to ng/L concentrations. Batch studies revealed that the propanotroph
Keywords:
growth on propane, and that the strain produced metabolites that do not pose a significant
Rhodococcus ruber ENV425 was capable of metabolizing NDMA from 8 mg/L to <2 ng/L after N-Nitrosodimethylamine
risk at the concentrations generated (Fournier et al., 2009). A laboratory-scale membrane
NDMA
bioreactor (MBR) was subsequently constructed to evaluate the potential for long-term ex
Membrane bioreactor
situ treatment of NDMA. The MBR was seeded with ENV425 and received propane as the
Propanotroph
primary growth substrate and oxygen as an electron acceptor. At an average influent
Trichloroethene
NDMA concentration of 7.4 mg/L and a 28.5 h hydraulic residence time, the reactor effluent
TCE
concentration was 3.0 2.3 ng/L (>99.95% removal) over more than 70 days of operation. The addition of trichloroethene (TCE) to the reactor resulted in a significant increase in effluent NDMA concentrations, most likely due to cell toxicity from TCE-epoxide produced during its cometabolic oxidation by ENV425. The data suggest that an MBR system can be a viable treatment option for NDMA in groundwater provided that high concentrations of TCE are not present. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
N-Nitrosodimethylamine (NDMA) is a potent carcinogen and an emerging groundwater and drinking water contaminant in the U.S. (USEPA, 2009). Current concerns with NDMA in groundwater began more than a decade ago when the compound was detected at various military and aerospace facilities that previously handled liquid propellants containing unsymmetrical dimethylhydrazine (UDMH) (Fleming et al., 1996; Girard, 2000; Mitch et al., 2003). UDMH, which is a major component of the propellant Aerozine-50, contains
NDMA as a chemical impurity and has also been observed to rapidly oxidize to NDMA in natural environments. Subsequent testing in California and elsewhere revealed that NDMA was also present in reclaimed wastewater and in numerous drinking water supplies as a disinfection byproduct of chlorination (Mitch and Sedlak, 2002a,b; Mitch et al., 2003; Sedlak et al., 2005). Although various pathways of NDMA formation are possible during chlorination of water, one primary reaction generating this nitrosamine appears to be between naturally occurring secondary amines and monochloramine, which is used as a long acting disinfectant
* Corresponding author. Tel.: þ1 609 895 5356. E-mail address:
[email protected] (P.B. Hatzinger). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.07.056
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 4 e2 6 2
(Choi and Valentine, 2002; Mitch et al., 2003). Dichloramine, which is generated during disproportionation of monochloramine, may also form NDMA in water via reaction with secondary amines (Schreiber and Mitch, 2006, 2007). NDMA is a potent carcinogen and mutagen (ASTDR, 1989; USEPA, 2009, 2010a). Although there is presently no federal maximum contaminant level (MCL) for NDMA in drinking water, the Environmental Protection Agency (EPA) recently added NDMA and four other nitrosamines to its Unregulated Contaminant Monitoring Rule 2 (USEPA, 2010b). As a result, many large water utilities are now required to monitor for these nitrosamines. In addition, California and Massachusetts each have 10 ng/L notification levels for NDMA in drinking water, and the California Office of Environmental Health Hazard Assessment (OEHHA) recently established a more stringent public health goal (PHG) for NDMA in drinking water of 3 ng/L based on risk calculations (OEHHA, 2006). The most widely used treatment technology for removing NDMA from groundwater and drinking water is photolysis via ultraviolet irradiation (UV), which breaks the NeN bond, yielding nitrite and dimethylamine as primary products (USEPA, 2009; Mitch et al., 2003). This approach is effective, but also expensive as the energy input to reduce NDMA concentrations by one order of magnitude is approximately ten times that necessary for standard disinfection of viruses and other water-borne pathogens (Mitch et al., 2003). At some military and aerospace sites, reductions in NDMA concentrations of 3e5 log orders are required to meet treatment objectives. The goal of this study was to evaluate the effectiveness of a biological reactor system as an alternative to UV treatment for NDMA. Recent studies have revealed that bacteria expressing various broad-specificity monoxygenase enzymes, including toluene-4-monoxygenase (Fournier et al., 2006), propane monooxygenase (Sharp et al., 2005, 2007; Fournier et al., 2009) and soluble methane monooxygenase (Yoshinari and Shafer, 1990; Sharp et al., 2005) are capable of degrading NDMA through aerobic cometabolism. In each case, the cells do not grow on NDMA as a carbon and/or energy source but rather biodegrade the nitrosamine fortuitously during growth on the primary substrate (e.g., propane or toluene). However, for in situ or ex situ biological treatment of NDMA to be practical, mg/ L concentrations of the nitrosamine must be reduced to low ng/L concentrations. Few compounds have such stringent treatment requirements, and biodegradation processes are rarely considered for such applications. Batch experiments with the propanotroph Rhodococcus ruber ENV425 revealed the bacterium can biodegrade NDMA from 8 mg/L to <2 ng/L (the practical quantitation limit (PQL) for NDMA) during growth on propane (Fournier et al., 2009). Moreover, the degradation products from the reaction, which include methylamine, nitrite, nitrate, formate, and carbon dioxide do not pose a significant risk at the concentrations generated. Based on the batch culture studies, a laboratory-scale membrane bioreactor (MBR) was constructed to evaluate the potential for sustained NDMA biotreatment. The MBR in this study was seeded with R. ruber ENV425 and received propane as the primary growth substrate and NDMA at concentrations typical at a contaminated military or aerospace site. Although MBRs have been most widely applied for treatment of highstrength municipal and industrial wastewaters (Roberts et al.,
255
2000; Stephenson et al., 2000), there is increasing interest in this design for treatment of xenobiotics and natural compounds requiring very low effluent levels, including endocrine disrupting compounds (EDCs) (Mansell et al., 2005; Hu et al., 2007; De Gusseme et al., 2009). Because the MBR system operates with a high biomass density, has a long solids retention time (SRT), and produces a clean effluent (which passes as permeate through the membrane), it is potentially well suited for removal of NDMA from groundwater or wastewater, followed by direct water re-injection or reuse.
2.
Materials and methods
2.1. Laboratory membrane bioreactor (MBR) design and operation A schematic of the laboratory MBR system is provided in Fig. 1.The reactor consisted of a 3 L glass vessel fitted with a submerged Zenon ZeeWeed 1 (ZW1) hollow-fiber ultra filtration membrane cartridge (General Electric Water and Process Technologies, USA), which has a total membrane surface area of approximately 0.09 m2. The reactor was continually stirred and fed pure oxygen from a cylinder directly through a sparge stone. The influent to the reactor consisted of the following sources mixed at a 1:1 ratio: (1) an artificial groundwater amended with NDMA, and (2) propanesaturated deionized water. The propane (99% purity; Advanced Gas Technologies, Palm, PA) was added to the water from a gas cylinder via diffusion through silicone tubing. Separate peristaltic pumps were used to supply the combined influent (propane- and NDMA-laden water, respectively), and the sources were mixed just prior to entering the MBR. The influent stream also received a dilute solution of inorganic nutrients (1:20 strength basal salts medium; BSM) (Hareland et al., 1975). A third peristaltic pump was connected to the ZW1 cartridge (w0.04 mm pore-size) to pull effluent water (permeate) through the membrane. The reactor volume was controlled by a level sensor such that the maximum volume was 2.94 L and the minimum volume was 2.77 L. The artificial groundwater constituents were based on groundwater chemistry at a site in Maryland (Schaefer et al., 2007), and contained NaSO4 (180 mg/L), NaCl (113 mg/L), NaNO3 (60 mg/L), NaHCO3 (40 mg/L), and MnSO4 H2O (1 mg/ L) in deionized water. The pH of the groundwater was w8.3, and additional pH control in the pilot reactor was unnecessary. The hydraulic residence time (HRT) in the MBR averaged 28.5 h throughout the study, based on a total influent flow averaging 100 mL/h and an average reactor volume of 2.85 L, and the solids retention time (SRT) was infinite (i.e., biomass was not wasted from the reactor). The reactor was initially operated for 35 days without addition of propane or biomass to establish baseline influent and effluent NDMA concentrations. NDMA was added to the reactor at a concentration of w8 mg/L during this period. The reactor was seeded with a liquid culture of the R. ruber ENV425 on Day 36 to achieve an initial optical density of 1.0 (OD550), and then a second time on Day 56 to an OD550 of 0.5. The reactor was operated at a desired influent NDMA concentration of w8 mg/L through Day 107, at which time the influent
256
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 4 e2 6 2
Fig. 1 e Components of the propane-fed membrane bioreactor (MBR).
NDMA was increased tenfold to w80 mg/L. Trichloroethene (TCE) was added to the influent water at 200 mg/L on Day 171 in addition to NDMA (80 mg/L) and it was removed on Day 240. During the final phase of reactor operation from Day 241 to Day 321, only NDMA was added to the reactor. During operation of the laboratory MBR, temperature, pH, dissolved oxygen (DO), and pressure in the silicon tubing used to supply propane gas were generally measured 4 days per week. Reactor temperature was measured using a thermometer, DO was measured with a YSI 550 DO meter (YSI Incorporated, Yellow Springs, OH), and pH was measured with an Orion 720A pH meter fit with an Accumet pH probe (Cole Parmer, Vernon Hills, IL). Cell density was initially measured at OD550 using a spectrophotometer, but this measurement was discontinued on Day 80 because ENV425 was observed to be present primarily as an attached biofilm within the MBR rather than as planktonic cells. The membrane was thoroughly cleaned with a small bottle brush on Day 121 and Day 219 to remove excess biomass. From Day 245 to Day 321, w25% of the membrane was cleaned by gentle scrubbing with a bottle brush on a weekly basis or when necessary based on permeate flow from the reactor. The influent and effluent dissolved propane concentrations were measured twice per week during the initial 6 months of reactor operation. Propane concentrations in solution were not taken during the final 5 months of operation because of a mechanical failure of the instrument used for this purpose. For propane analysis, liquid samples (30 mL) taken from the reactor influent or effluent were diluted 1:1000 in
distilled deionized water in 40-mL VOA vials containing 25% headspace. The dissolved gases were allowed to equilibrate in the headspace, and the headspace samples were analyzed on a Varian 3900 GC equipped with an FID detector and a Restek RTX Alumina column (50 m, 0.53 mm ID). A 4-point standard curve derived from direct injection of varying propane concentrations was used to determine the propane in each sample. Henry’s Law was used to determine the liquid concentration from analysis of headspace gas as detailed previously in Kampbell and Vandegrift (1998).
2.2.
Sample collection and analysis of NDMA and TCE
NDMA influent and effluent samples were generally taken on a weekly basis. Effluent sampling required the collection of 1 L of water, which took w10 h based on the flow rate. The samples were collected into specially-cleaned amber glass bottles (supplied by Maxxam Analytics, Inc., Ontario, Canada) on ice. After the sample jar was full, the effluent water was filtered through a 0.22-mm pore-size cellulose acetate membrane (Corning) for preservation, then placed into a second cleaned amber glass jar. In the absence of cells, multiple tests showed that filtration with the Corning units did not remove any NDMA from solution. In the presence of cells, filtration was observed to remove a maximum of w7% of the added NDMA (i.e., due to adsorption or cell uptake) (Fournier et al., 2009). Because of the membrane within the MBR, however, no cells were expected to pass into the effluent water prior to filtration.
257
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 4 e2 6 2
2.3.
Effect of TCE on propane utilization by ENV425
The effect of TCE on propane utilization by ENV425 was tested in a batch experiment. Strain ENV425 was grown aerobically in BSM with 25% propane in the headspace. The culture was harvested during logarithmic growth phase, washed, and resuspended in BSM to an optical density at 550 nm (OD550) of 2.0. Cells were distributed in 5-mL aliquots into 27-mL serum vials and sealed with Teflon-lined septa. TCE, carried in 5 mL of ethanol, was injected into the vials through the septa to achieve final headspace TCE concentrations ranging from 0 mM to
200 mM. Control samples without TCE received 5 mL of ethanol. The vials were incubated at 30 C with gentle shaking (w70 rpm). The TCE concentrations in the vials that received the highest initial concentrations (100 and 200 mM in headspace, respectively) were monitored periodically via GC analysis of the headspace gas to determine when all of the TCE had been consumed. The analysis was performed by injecting 25 mL of headspace gas withdrawn through the septa onto a Varian 3900 GC (Palo Alto, CA) equipped with an RTX 502.2 column (Restek, Bellefonte, PA) and a flame ionization detector. The vials were then amended with propane gas (1100 uM) and incubated for an additional hour. The propane remaining in the headspace of each vial was then measured essentially as described above for TCE and compared to a standard curve comprised of varying concentrations of propane injected in vials containing only sterile BSM.
3.
Results and discussion
3.1.
NDMA treatment prior to TCE addition
During the initial 35 days of reactor operation prior to seeding with R. ruber ENV425, influent and effluent NDMA concentrations each averaged 10 mg/L, showing that adsorption and/or volatilization were not significant mechanisms of contaminant loss in the laboratory MBR system (Fig. 2). NDMA concentrations in the reactor effluent declined from 9.1 mg/L to 10 ng/L within two days of seeding with ENV425. The effluent concentrations declined further to 3.1 ng/L by Day 45 of operation (9 days after seeding with ENV425), and were below detection (i.e., <2 ng/L) by Day 57. During the initial 76 days of operation after seeding the reactor (Days 37e113), with influent NDMA concentrations averaging 7.4 1.5 mg/L, NDMA
100
10
NDMA (µ g/L)
After filtration, the 1 L samples were stored in the speciallycleaned amber jars at 4 C, and then shipped overnight on ice for extraction and low-level NDMA analysis. The analysis was conducted by Maxxam Analytics, Inc. by High Resolution Mass Spectrometry (HRMS) according to USEPA Method 607/1625 modified for low-level NDMA analysis (USEPA, 2010c,d; Maxxam Analytics, Inc., 2009). The PQL for NDMA by this method is 2 ng/L. The Maxxam laboratory is certified by the California DHS ELAP program to perform trace NDMA analysis as well as by the Standards Council of Canada, and the Ministry of Environment of Ontario Drinking Water PT Program (Maxxam Analytics, Inc., 2009). The concentration of NDMA in the influent to the MBR ranged from w8 mg/L to 80 mg/L, and the samples did not require lowlevel NDMA analysis by HRMS. The influent samples were extracted, concentrated, and analyzed in the Shaw Analytical Testing Laboratory in Lawrenceville, NJ. Initially, 1 L of influent water was collected on ice from a sampling port placed just before the reactor. The water was filtered as detailed for the effluent samples. Extraction of NDMA in the influent samples was conducted by solid phase extraction according to the procedure outlined in EPA Method 521 (USEPA, 2004). The samples (in 1 mL of dichloromethane after extraction and concentration) were then analyzed for NDMA by a modification of EPA method 8270C (USEPA, 1996a) using a Hewlett Packard 6890 Series GC fitted with an XTI-5 chromatography column (Restek, Bellefonte, PA; length ¼ 30 m, ID ¼ 0.25 mm, DF ¼ 0.25 mm). The injector was maintained at 240 C and the detector at 250 C. The column was temperature programmed. The initial temperature was 45 C for 2 min, after which the temperature was increased to 225 C at a rate of 20 C per min and held for 1.0 min. This was followed by a temperature increase to 265 C at a rate of 35 C per min and held for 0.5 min. Under these conditions NDMA eluted at w3.6 min. A Hewlett Packard 5973 Mass Spectrometer was linked to the GC system. The PQL for NDMA using this method was 200 ng/L. The internal standard used for calibration was d14-N-nitrosodi-n-propylamine and d6-NDMA was added as a surrogate prior to solid phase extraction. Data were collected using selective ion monitoring (SIM) to increase sensitivity. For NDMA, the ions monitored were 74, 42, and 43. For the surrogate, ions monitored were 80, 35 and 46, and for the internal standard the ions monitored were 46, 48 and 80. All calibrations and QA/QC follow standard procedures as outlined in EPA Method 8270C. Influent and effluent TCE samples were collected on a weekly basis (Day 171e240) in glass GC vials fitted with Teflon-lined septa. The TCE in solution was analyzed according to EPA Method 8260B (USEPA, 1996b).
TCE removed
1
ENV425 Inoculated
0.1
TCE added
0.01
0.001 0
40
80
120
160
200
240
280
320
Days
Fig. 2 e Influent (,) and Effluent () concentrations of NDMA in the laboratory MBR. The concentrations are given on a logarithmic scale. The reactor was initially inoculated with ENV425 on Day 36. NDMA was fed to the reactor at an average influent concentration of 8.4 mg/L from Day 0 to Day 106. The average influent NDMA concentration from Day 107 to Day 321 was 77 mg/L. TCE was added to the MBR ate200 mg/L from Day 171 to Day 240.
258
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 4 e2 6 2
in the reactor effluent averaged 3.0 2.3 ng/L (with non-detect values averaged conservatively at the PQL of 2 ng/L). Thus, average removal of NDMA exceeded 99.95%. The NDMA degradation rate based on the membrane surface area was 80 mg per h per m2 of membrane area (mg/h m2). The rate was calculated based on membrane surface area because cell biofilm formation precluded rate determination based on cell mass (see Section 3.4). The influent concentrations of NDMA to the MBR were increased to w80 mg/L on Day 107 of reactor operation. There was a single sample in which an effluent concentration of 3.0 mg/L (Day 121) was detected after the influent NDMA concentration was increased. After this time, the effluent concentration again decreased to <3 ng/L, and remained at or below 10 ng/L through Day 162 (126 days after ENV425 was added on Day 36). The 3 mg/L effluent concentration on Day 121 may have been the result of maintenance work on the reactor that caused disruption of biomass adsorbed to the membrane unit within the MBR. The membrane module was thoroughly cleaned by hand with a brush to improve flux on the day prior to sample collection. The initial 4.5 months of operational data with NDMA as the sole contaminant clearly indicate that an MBR can achieve effluent NDMA concentrations in the low ng/L range and that these concentrations can be maintained during typical operation. Moreover, the influent NDMA concentrations tested during this phase are in the upper range of those detected in groundwater at contaminated military and aerospace facilities (i.e., 1e100 mg) (USEPA, 2009; Mitch et al., 2003). Concentrations observed in disinfected wastewater and drinking water in the U.S. and abroad are typically much lower, usually <200 ng/L (Mitch et al., 2003; Zhao et al., 2008; Dillon et al., 2008; Asami et al., 2009).
3.2.
Propane and dissolved oxygen
The dissolved propane concentration entering the reactor averaged 16 4 mg/L for the first two weeks after seeding with 70
Propane in Solution (mg/L)
60
50
40
30
20
10
0 40
50
60
70
80
90
100
110
Days
Fig. 3 e Influent (,) and effluent () concentrations of propane in the laboratory MBR.
120
ENV425 (Fig. 3). This concentration was increased to an average of 45 mg/L beginning on Day 56 (20 days after seeding ENV425) by increasing the pressure of propane in the silicon tubing entering the gassing vessel. The influent propane concentrations averaged 45 8 mg/L between Day 56 and Day 112 of reactor operation, which provided a propane to NDMA mass ratio of approximately 6000 during this period when the feed NDMA averaged 7.4 1.5 mg/L. Propane in the reactor effluent was generally below detection (w1 mg/L), although concentrations of 1e3 mg/L were detected on occasion during the first 4 months of reactor operation. The dissolved oxygen (DO) was supplied in excess using a sparge stone within the reactor, and averaged 14 5 mg/L in the reactor during the course of the 321-day study (data not shown).
3.3.
Temperature and pH
During the course of the entire study, the temperature within the MBR averaged 22.4 1.4 C. The pH of the influent water averaged 8.2 0.5 during the study, while the pH within the MBR was somewhat lower at 6.7 0.2. The MBR system did not have online pH control, but significant alkalinity was present in the artificial groundwater to buffer the reactor system.
3.4.
Biomass
The biomass within the reactor was to be monitored throughout the study using both optical density (OD550) and analysis of total suspended solids (TSS). When growing in batch, ENV425 forms small clumps, but generally remains as a planktonic culture rather than a biofilm. However, when placed within the MBR, the strain rapidly formed a biofilm on the membrane as well as the glass and stainless steel surfaces. The predominant biofilm nature of the culture in the MBR was apparent by visual examination (the culture is orange) and by the rapidly declining OD550 within the MBR after inoculation. Although the MBR was initially inoculated with strain ENV425 to an OD550 of 1.0, the cell density declined to an OD550 of 0.01 within 15 days after inoculation (Day 51 of operation). This occurred despite the extensive biodegradation of NDMA shown during this period (Fig. 2). The reactor was inoculated for a second time on Day 56 to an OD550 of 0.5, but again within one week the OD550 was w0.02 in solution. Because the culture was predominantly present in a biofilm, measurement of TSS was not conducted and measurement of OD550 was discontinued after Day 80 of operation (due to OD550 values consistently <0.02). In addition, because of the thick biofilm, the membrane had to be gently cleaned weekly by hand with a small wire brush to allow adequate flux of permeate. Due to the nature of growth, biomass was not wasted from the reactor during the study, resulting in an infinite SRT. Degradation rates were subsequently calculated as a function of membrane surface area (rather than cell mass) to allow for sizing of a larger-scale fixed film system.
3.5. TCE degradation and influence of TCE on NDMA treatment Trichloroethene (TCE) was introduced to the MBR system as a cocontaminant after the initial 170 days of operation. Prior to
259
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 4 e2 6 2
adding TCE, a batch study was conducted to determine whether ENV425 was capable of oxidizing the compound. As with several other propanotrophs (e.g., Wackett et al., 1989; Tovanabootr and Semprini, 1998), ENV425 was observed to rapidly oxidize TCE after growth on propane (data not shown). The concentration of TCE present in the reactor effluent during this phase of operation was 80 11 mg/L compared to an average influent concentration of 220 120 mg/L (including one anomalously high point at Day 204) (Fig. 4). Thus, an average decline in TCE of w60% was observed. The extent of removal of TCE within the reactor was appreciably lower than that observed for NDMA, for which effluent concentrations averaged 4e5 log orders lower than the influent concentrations during the initial months of operation after seeding with ENV425. In addition, the percentage of TCE lost to sorption or volatilization was not known because the compound was not added prior to seeding the reactor with ENV425 (i.e., to quantify abiotic processes). In contrast, influent and effluent concentrations of NDMA were shown to be the same prior to seeding with ENV425, suggesting that there was no appreciable abiotic loss of the nitrosamine. The addition of TCE to the reactor on Day 171 had a significant effect on effluent levels of NDMA (Fig. 2). Effluent concentrations of NDMA increased from an overall average value of 3.6 ng/L during the initial 162 days of operation, to 3.4 mg/L on Day 204, 17 mg/L on Day 218, and 27 mg/L on Day 239. The addition of TCE to the MBR was discontinued on Day 240. The ability of the reactor to recover from this upset was then evaluated. The data from this experimental phase clearly indicate that NDMA treatment is adversely impacted by the presence of TCE. These are several possible explanations for this impact, the most probable of which include substrate inhibition and/or cell toxicity. Competitive interactions between TCE and NDMA during biodegradation by ENV425 or other propanotrophs have
not been studied in detail. Presuming that a propane monooxygenase is responsible for both reactions in ENV425, as was recently shown for the NDMA-degrader Rhodococcus spp. RHA1 (Sharp et al., 2007), competitive inhibition of TCE on NDMA metabolism is possible. A second potential explanation for the impact of TCE on NDMA metabolism by ENV425 is cell toxicity. During metabolism of TCE by monooxygenase enzymes, including methane monooxygenase and cytochrome-P450 enzymes, a short-lived and highly toxic TCE-epoxide is formed (Alvarez-Cohen and McCarty, 1991; Wackett et al., 1989). A batch study was conducted with R. ruber ENV425 to examine the hypothesis that cell toxicity caused the TCE-induced decline in NDMA treatment efficiency in the reactor. During this study, ENV425 was grown on propane, incubated with varying concentrations of TCE for 3 h, and then tested for propane metabolism after the TCE concentrations had been reduced to below detection. Cultures incubated with TCE headspace concentrations of 50 mM and higher lost their ability to degrade propane (Fig. 5). Conversely, ENV425 cells that were exposed to TCE concentrations in headspace of 25 mM or lower were still able to metabolize propane, although degradation rates were reduced compared to controls without TCE. These data support the hypothesis that cell toxicity accounted for decreased treatment efficiency of NDMA when TCE was present within the MBR. An inability of ENV425 cells to metabolize propane in the absence of TCE (i.e., the TCE was biodegraded before propane was added to vials) would not be expected if competitive inhibition were the cause of reduced treatment in the MBR system. Unfortunately, propane in the effluent of the MBR was not measured during the period that TCE was added to the MBR. If cell toxicity accounted for poor reactor efficiency for NDMA, effluent propane concentrations may have increased significantly. The observation that TCE inhibits NDMA destruction in the MBR does not significantly reduce the potential for this technology in the field. Although it would be beneficial to be able to
500
Propane remaining (µ M in headspace)
1300
TCE (µ g/L)
400
300
200
100
1200
1100
1000
900
800
700 0
0 180
50
100
150
200
250
TCE added (µM in headspace) 190
200
210
220
230
240
250
Days
Fig. 4 e Influent (,) and effluent () concentrations of TCE in the laboratory MBR. TCE was added to the MBR in the influent water from Day 171 to Day 240.
Fig. 5 e Influence of TCE on propane utilization by ENV425. Propane-grown cells were incubated with TCE at the given concentration until all TCE was degraded (w3 h). Propane was then added to all vials ate1100 mM in headspace. After 1 h, the propane remaining in headspace was determined.
260
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 4 e2 6 2
treat both compounds in one unit operation, ex situ aerobic cometabolic treatment of TCE has never proven to be practical or cost-effective. In groundwater streams in which both compounds are present, TCE can be easily removed by either carbon adsorption or air stripping prior to the reactor system. The latter process has the added benefit of oxygenating the groundwater prior to entering an NDMA bioreactor. As an example, Aerojet (Sacramento, CA) has a 19,000 L per minute (LPM) treatment system for combined nitrate, perchlorate, TCE, and NDMA (Girard, 2000; Hatzinger, 2005). The four compounds are removed in sequence by an aerobic biological treatment in fluidized bed reactors (nitrate and perchlorate), air stripping (TCE), and UV treatment (NDMA). In this case, the replacement of the UV system with a more cost-effective approach would be desirable.
3.6. Recovery of NDMA degradation after removal of TCE from influent After TCE was removed from the influent, the concentration of NDMA in the reactor effluent declined rapidly (Fig. 2). The NDMA concentration in the reactor effluent reached a high value of 27.1 mg/L on Day 239, at which time the TCE feed was discontinued. Within w4 weeks (Day 267), the effluent NDMA concentration was 66 ng/L, and after 6 weeks (Day 281) the effluent concentration reached 35 ng/L. Thus, recovery of reactor performance was rapid. However, effluent NDMA concentrations <10 ng/L were not achieved after TCE was removed. Rather, effluent concentrations were 58 24 ng/L during the final six weeks of operation, which still represents 99.93% removal of NDMA from the average influent concentration during this period of 81 5 mg/L, and a removal rate of w86 mg/h m2. The higher effluent concentrations during this time compared to initial operation with the same NDMA load (Days 107e170) may be a residual effect of TCE toxicity or may have been caused by an increased necessity for reactor membrane cleaning, which resulted in reduced efficiency. After operation for more than 10 months, and increased biofilm formation in the reactor, the membrane required cleaning and backwashing on at least a weekly basis to remain operational. The cleaning procedure, which included manually removing biomass from a portion of the membrane, may have affected reactor performance, resulting in slightly elevated effluent NDMA concentrations at the end of the study.
3.7.
Comparison with other reactor tests for NDMA
Overall, the data from this study suggest that a propane-fed MBR system is capable of reducing NDMA in groundwater to low ng/L concentrations, with removal efficiencies exceeding 99.9% at typical groundwater concentrations, provided that TCE is not entering the reactor as a co-contaminant. The impact of other chlorinated solvents and potential cocontaminants was not examined. To our knowledge, this is the first study in which an aerobic, propane-fed bioreactor has been evaluated for NDMA treatment. Previous studies have been conducted to assess the potential for NDMA removal in traditional wastewater treatment plants (WWTP) in the U.S. (Sedlak et al., 2005) and Switzerland (Krauss et al., 2009) and in MBR systems designed for primary
effluent from a WWTP (Mansell et al., 2004, 2005). In 20 WWTPs in Switzerland, removal efficiencies for NDMA during secondary activated sludge treatment were highly variable and ranged from <5% to 95% (average 65%) at influent concentrations ranging from 5 ng/L to 1 mg/L (Krauss et al., 2009). Moreover, the NDMA removal rates in individual plants were shown to be highly variable, ranging from 18% to 96% in one Swiss plant (Krauss et al., 2009), and from 0% to 75% at a U.S. WWTP where influent concentrations varied from <50 ng/L to >350 ng/L (Sedlak et al., 2005). Removal in the MBR systems from an average influent of 100 ng/L averaged 80%, which was similar to performance observed in a sludge water reclamation plant (Mansell et al., 2004). A hydrogen fed membrane biofilm reactor (MBfR; in which the membrane is used to supply gas and support biofilm growth rather than to filter biomass) was also previously tested for NDMA treatment (Chung et al., 2008). This system, which was operated under anoxic conditions (with nitrate- and/or sulfate-reducing strains presumably degrading NDMA), was observed to reduce influent NDMA concentrations from 800 ng/L to 18 6 ng/L (97.7% efficiency). At an average NDMA influent concentration of 7.4 mg/L, which is 1e2 orders of magnitude higher than concentrations tested in a WWTP, traditional MBR, or MBfR, the propane-fed MBR produced consistently lower effluent concentrations (3.0 2.3 ng/ L) and higher removal efficiencies (>99.9%). The laboratory data suggest that a propane-fed MBR is a viable treatment option for NDMA in water. Field studies with a larger-scale system are required to better evaluate the technical and economic viability of this approach. Also, as with other MBR systems (Chang et al., 2002; Miura et al., 2007), membrane fouling is an operation and maintenance issue that must be considered when using this design. The development of propane- and oxygen-fed reactor designs other than the MBR, including a fluidized bed reactor (FBR) and/or an MBfR design may further expand the potential application of this treatment approach. The current results suggest that properly designed high-performance fixed film reactors, such as these, may be capable of achieving similar effluent quality at rates of w80 mg/h m2 of effective surface area.
4.
Conclusions
NDMA-contaminated groundwater and wastewater represents a significant and often expensive treatment challenge. During this project, an aerobic, propane-fed MBR seeded with the culture R. ruber EV425 was evaluated for treatment of NDMA in an artificial groundwater. With influent NDMA at concentrations ranging from w8 mg/L to 80 mg/L, which are typical of some military sites, effluent concentrations from the MBR were typically <10 ng/L for more than 4 months of operation (>99.9% efficiency). The addition of TCE to the influent of the MBR resulted in a rapid and significant increase in NDMA effluent concentrations, most likely due to cell toxicity. However, in a mixed plume, many options are available for TCE removal (e.g., air stripping, carbon adsorption) prior to influent water entering an MBR. Overall, the laboratory data suggest that a propane-fed MBR is a viable treatment option for NDMA in water. Field studies with a larger-scale system are now required to better evaluate the technical and economic viability of this approach.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 4 e2 6 2
Acknowledgements The authors gratefully acknowledge the Strategic Environmental Research and Development Program (SERDP) for funding this research (Project ER-1456). We also thank Anthony Soto and Randi Rothmel for excellent technical assistance. Views, opinions, and/or findings contained herein are those of the authors and should not be construed as an official Department of Defense position or decision unless so designated by other official documentation.
references
Alvarez-Cohen, L., McCarty, P.L., 1991. Product toxicity and cometabolic competitive inhibition modeling of chloroform and trichloroethylene transformation by methanotrophic resting cells. Applied and Environmental Microbiology 57, 1031e1037. 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, 3540e3545. ASTDR, 1989. Toxicological Profile for N-nitrodimethylamine. Agency for Toxic Substances and Disease Registry (ASTDR), U.S. Public Health Service/U.S. Environmental Protection Agency. Chang, I.-S., Le Clech, P., Jefferson, B., Judd, S., 2002. Membrane fouling in membrane bioreactors for wastewater treatment. Journal of Environmental Engineering 128, 1018e1029. Choi, J., Valentine, R.L., 2002. Formation of N-nitrosodimethylamine (NDMA) from reaction of monochloramine: a new disinfection by-product. Water Research 36, 817e824. Chung, J., Ahn, C.-H., Chen, Z., Rittman, B.E., 2008. Bio-reduction of N-nitrosodimethylamine using a hydrogen-based membrane biofilm reactor. Chemosphere 70, 516e520. De Gusseme, B., Pycke, B., Hennebel, T., Marcoen, A., Vlaeminck, S.E., Noppe, H., Bon, N., Verstraete, W., 2009. Biological removal of 17a-ethinylestradiol by a nitrifier enrichment culture in a membrane bioreactor. Water Research 43, 2493e2503. Dillon, G., Blake, S., Rumsby, P., Rockett, L., Hall, T., Jackson, P., Rowlinson, A., 2008. NDMA-concentrations In Drinking Water and Factors Affecting Its Formation. Final Report, Defra. Thames Water Utilities Limited (DEFRA, 7348), 115 pp. Available from:. http://www.dwi.gov.uk/Research/reports/ DWI70_2_210.pdf. Fleming, E.C., Pennington, J.C., Wachob, B.G., Howe, R.A., Hill, D.O., 1996. Removal of N-nitrosodimethylamine from waters using physicalechemical techniques. Journal of Hazardous Materials 51, 151e164. Fournier, D., Hawari, J., Streger, S.H., McClay, K., Hatzinger, P.B., 2006. Biotransformation of N-nitrosodimethylamine (NDMA) by Pseudomonas mendocina KR1. Applied and Environmental Microbiology 72, 6693e6698. Fournier, D., Hawari, J., Streger, S.H., McClay, K., Hatzinger, P.B., 2009. Aerobic biodegradation of N-nitrosodimethylamine (NDMA) by the propanotroph Rhodococcus ruber ENV425. Applied and Environmental Microbiology 75, 5088e5093. Girard, M., 2000. Treatment of perchlorate and other contaminants to produce drinking water. Technical abstract. p. 37. In Environmental Challenges for the Next decade. Partners in Environmental technology technical Symposium and Workshop, Arlington, VA, November 28e30, 2000.
261
Hareland, W., Crawford, R.L., Chapman, P.J., Dagley, S., 1975. Metabolic function and properties of 4-hydrophenylacetic acid 1-hydrolase from Pseudomonas acidovorans. Journal of Bacteriology 121, 272e285. Hatzinger, P.B., 2005. Perchlorate biodegradation for water treatment. Environmental Science and Technology 39, 239Ae247A. Hu, J.Y., Chen, X., Tao, G., Kekred, K., 2007. Fate of endocrine disrupting compounds in membrane bioreactor systems. Environmental Science and Technology 41, 4097e4102. Kampbell, D.H., Vandegrift, S.A., 1998. Analysis of dissolved methane, ethane, and ethene in groundwater by a standard gas chromatographic technique. Journal of Chromatographic Science 36, 253e256. 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. Mansell, B., Kuo, J., Tang, C.C., Thompson, S., Jackson, J., Garcia, A., Gabrielian, E., Horvath, R., Stahl, J., 2004. Comparison of two membrane bioreactors and an activated sludge plant with dual-media filtration: nutrient and priority pollutant removals. In: Proceedings of the Water Environment Federation, WEFTEC 2004, Session 11e20, pp. 749e761. Mansell, B., Kuo, J., Tang, C.C., Huitric, J., Horvath, R.W., Stahl, J.F., 2005. Evaluation of removals of NDMA, hormones, pharmaceuticals, and personal care products in a membrane bioreactor. In: Proceedings of the Water Environment Federation, WEFTEC 2005, Session 81e90, pp. 7411e7742. Maxxam Analytics, Inc., 2009. NDMA (N-nitrosodimethylamine) analysis: analysis by HRMS. Available from: http://www. maxxam.ca/content_man/admin/document/sol_env_NDMA_ 0805.pdf. Mitch, W.A., Sharp, J.O., Trussell, R.R., Valentine, R.L., AlvarezCohen, L., Sedlak, D.L., 2003. N-Nitrosodimethylamine (NDMA) as a drinking water contaminant: a review. Environmental Engineering Science 20, 389e404. Mitch, W.A., Sedlak, D.L., 2002a. Formation of N-nitrosodimethylamine (NDMA) from dimethylamine during chlorination. Environmental Science and Technology 36, 588e595. Mitch, W.A., Sedlak, D.L., 2002b. Factors controlling nitrosamine formation during wastewater chlorination. Water Science and Technology: Water Supply 2, 191e198. Miura, Y., Watanabe, Y., Okabe, S., 2007. Membrane biofouling in pilot-scale membrane bioreactors (MBRs) treating municipal wastewater: impact of biofilm formation. Environmental Science and Technology 41, 632e638. OEHHA, California Office of Environmental Health Hazard Assessment, 2006. Announcement of publication of the final public health goal for N-nitrosodimethylamine and Cadmium in drinking water. Available from: http://www.oehha.ca.gov/ water/phg/cadndma122206.html. Roberts, J.A., Sutton, P.M., Mishra, P.N., 2000. Application of the membrane biological reactor system for combined sanitary and industrial wastewater treatment. International Biodeterioration and Biodegradation 46, 37e42. Schaefer, C.E., Fuller, M.E., Condee, C.W., Lowey, J.M., Hatzinger, P.B., 2007. Comparison of biotic and abiotic treatment approaches for the co-mingled perchlorate, nitrate, and nitramine explosives in groundwater. Journal of Contaminant Hydrology 89, 231e250. Schreiber, M.I., Mitch, W.A., 2006. Nitrosamine formation pathway revisited: the importance of chloramine speciation and dissolved oxygen. Environmental Science and Technology 40, 6007e6014. Schreiber, M.I., Mitch, W.A., 2007. Enhanced nitrogenous disinfection byproduct formation near breakpoint: implications for nitrification control. Environmental Science and Technology 41, 7039e7046.
262
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 4 e2 6 2
Sedlak, D.L., Deeb, R.A., Hawley, E.L., Mitch, W.A., Durbin, T.D., Mowbray, S., Carr, S., 2005. Sources and fate of nitrosodimethylamine and its precursors in municipal wastewater treatment plants. Water Environment Research 77 (1), 32e39. Sharp, J.O., Sales, C.M., LeBlanc, J.C., Liu, J., Wood, T.K., Eltis, L. D., Mohn, W.W., Alvarez-Cohen, L., 2007. An inducible propane monooxygenase is responsible for N-nitrosodimethylamine degradation by Rhodococcus sp. strain RHA1. Applied and Environmental Microbiology 73, 6930e6938. Sharp, J.O., Wood, T.K., Alvarez-Cohen, L., 2005. Aerobic biodegradation of N-Nitrosodimethylamine (NDMA) by axenic bacterial strains. Biotechnology and Bioengineering 89, 608e618. Stephenson, T., Judd, S., Jefferson, B., Brindle, K., 2000. Membrane Bioreactors for Wastewater Treatment. IWA Publishing, Alliance House, London, England, 179 pp. Tovanabootr, A., Semprini, L., 1998. Comparison of TCE transformation abilities of propane- and methane-utilizing microorganisms. Bioremediation Journal 2, 105e124. USEPA, 1996a. SW-846: test methods for evaluating solid waste, physical/chemical methods. Method 8270Cesemivolatile organic compounds by gas chromatography/mass spectrometry (GC/MS). Available from: http://www.caslab. com/EPA-Methods/PDF/8270c.pdf. USEPA, 1996b. SW-846: Test methods for evaluating solid waste, physical/chemical methods. Method 8260B-volatile organic compounds by gas chromatography/mass spectrometry (GC/ MS). Available from: http://www.caslab.com/EPA-Methods/ PDF/8260b.pdf. USEPA, 2004. Determination of Nitrosamines in Drinking Water by Solid Phase Extraction and Capillary Column Gas Chromatography With Large Volume Injection and Chemical Ionization Tandem Mass Spectrometry (MS/MS). EPA600/R-05/
054. National Exposure Research Laboratory, Office of Research and Development, Cincinnati, OH. USEPA, 2009. Emerging Contaminant e N-nitrosodimethylamine (NDMA) Fact Sheet. EPA505-F-09e008. Solid Waste and Emergency Response (5106P). Available from: http://www.clu-in.org/download/contaminantfocus/ epa505f09008.pdf. USEPA, 2010a. N-nitrosodimethylamine (CASRN 62-75-9) Integrated Risk Information Service (IRIS) Substance File. Available from: http://www.epa.gov/iris/subst/0045.htm. USEPA, 2010b. Unregulated contaminant monitoring Rule 2 (UCMR 2). Basic Information. Available from: http://www.epa. gov/safewater/ucmr/ucmr2/basicinformation.html#list. USEPA, 2010c. Clean water act analytical methods: methods for organic analysis. Method 607-nitrosamines. Available from: http://www.epa.gov/waterscience/methods/method/organics/ 607.pdf. USEPA, 2010d. Clean water act analytical methods: methods for organic analysis. Method 1625 revision B-semivolatile organic compounds by isotope dilution GC/MS. Available from: http://www.epa.gov/waterscience/methods/method/ organics /1625.pdf. Wackett, L.P., Brusseau, G.A., Householder, S.R., Hanson, R., 1989. Survey of microbial oxygenases: trichloroethylene degradation by propane-oxidizing bacteria. Applied and Environmental Microbiology 55, 2960e2964. Yoshinari, T., Shafer, D., 1990. Degradation of dimethyl nitrosamine by Methylosinus trichosporium OB3b. Canadian Journal of Microbiology 36, 834e838. Zhao, Y.-Y., Boyd, J.M., Woodbeck, M., Andrews, R.C., Qin, F., Hrudey, S.E., Li, X.-F., 2008. Formation of N-nitrosamines in eleven disinfection treatments of seven different source waters. Environmental Science and Technology 42, 4857e4862.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 3 e2 7 3
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Effect of polymeric substrate on sludge settleability Anto´nio M.P. Martins a,1, O¨zlem Karahan a,b, Mark C.M. van Loosdrecht a,c,* a
Department of Biotechnology, Delft University of Technology, Julianalaan 67, 2628 BC Delft, The Netherlands Istanbul Technical University, Faculty of Civil Engineering, Environmental Engineering Department, 34469 Maslak Istanbul, Turkey c KWR watercycle research institute, Groningenhaven 7, 3422 PE Nieuwegein, The Netherlands b
article info
abstract
Article history:
The study aims to evaluate the role of a polymeric substrate (starch) on sludge settleability.
Received 5 April 2010
Despite being an important COD component of the wastewater, the relationship between
Received in revised form
polymeric substrates and bulking sludge has been hardly studied. The polymers are hydro-
13 July 2010
lysed at a rate smaller than the consumption rate of monomers. This means that the soluble
Accepted 18 July 2010
substrate resulting from hydrolysis is likely to be present at growth rate limiting concen-
Available online 27 July 2010
trations. According to the kinetic selection theory this leads to bulking sludge. However, a recently postulated theory suggests that, strong diffusion limited micro-gradients of
Keywords:
substrate concentration inside flocs lead to bulking sludge, and not a low substrate
Bulking sludge
concentration as such. If the polymeric COD is first incorporated in the sludge floc and
Hydrolysis
afterwards hydrolysed in the sludge floc then there is essentially no substrate gradient inside
Kinetic selection
the biological flocs. The experiments showed that conditions leading to bulking sludge with
Sludge settleability
monomers (glucose) did not lead to bulking when starch was used. A bulking sludge event
Starch
was even cured just by substituting the monomer with starch. These results are clearly in line
Storage
with a diffusion gradient e based theory for bulking sludge. Nevertheless, flocs growing on starch are more open, fluffy and porous than flocs formed on maltose or glucose, most likely because the starch needs to be hydrolysed at the surface of the micro-colonies forming the flocculated sludge. Some additional observations on occurrence of filamentous bacteria in oxygen diffusion limited systems are also discussed in this manuscript. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Particulate substrate, often denoted as slowly biodegradable COD, is an important fraction of the total COD present in the wastewater. For instance, in The Netherlands particulate COD varies usually between 30 and 50% of the total COD (Kruit et al., 1994) while in other countries even higher values are reported: 70e90% in South Africa (Casey et al., 1999) and about 90% in Switzerland (Kappeler and Gujer, 1994). This type of substrate has a high molecular weight and is supposed to
undergo cell external hydrolysis before becoming available for consumption (growth and storage) by bacteria (Gujer et al., 1999). Polymers such as proteins, lipids and polysugars are major components of this COD fraction. Hydrolysis is widely considered the rate-limiting step of the overall COD removal process, and as such included in general activated sludge models (Gujer et al., 1999). This means that hydrolytic products, i.e. soluble substrate, are consumed at a higher rate than they are produced and, likely, will be available for the microorganisms at low (growth rate limiting)
* Corresponding author. Department of Biotechnology, Delft University of Technology, Julianalaan 67, 2628 BC Delft, The Netherlands. Tel.: þ31 15 2781618. ¨ . Karahan), M.C.M.vanLoosdrecht@ E-mail addresses:
[email protected] (A.M.P. Martins),
[email protected] (O tnw.tudelft.nl (M.C.M. van Loosdrecht). 1 ´ guas do Algarve, S.A., Rua do Repouso 10, 8000-302 Faro, Portugal. present address: A 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.07.055
264
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 3 e2 7 3
concentrations. According to the kinetic selection theory (Chudoba et al., 1973) these low soluble substrate concentrations give competitive advantages to filamentous bacteria, leading to bulking. Recently, an alternative hypothesis states that it is not the substrate concentration as such but the microgradients of substrate concentration inside biological flocs that play a more important role in the competition between filamentous and non-filamentous bacteria (Kappeler and Gujer, 1994; Martins et al., 2004a,b). Hydrolysis is generally found to occur at floc level (Dold et al., 1991; San Pedro et al., 1994; Mino et al., 1995; Goel et al., 1998, 1999; Mosquera-Corral et al., 2003; Karahan et al., 2006). Therefore, it can be expected that there will be no real gradients in substrate concentration inside the floc, although the substrate concentration inside the floc will be very low. The objective of this study was to evaluate the effect of polymeric substrate on the development of filamentous bacteria. Hitherto, well-controlled lab-scale systems have been used, as this allow a proper “scale-down” of the conditions bacteria experience in full-scale wastewater treatment plant containing a selector (Martins et al., 2003a, 2004c). For this study fully aerobic systems, fed with a polymeric substrate (potato starch) and easily biodegradable soluble monomers (glucose or maltose) were used as control systems. The use of different feeding periods and dissolved oxygen concentrations allowed to simulate different bulk liquid substrate and oxygen gradients and a variable relative size of the selector.
2.
Material and methods
2.1.
Experimental setup
The experiments were performed in seven different sequencing batch reactors (SBRs) with 2 L working volume. The reactors were controlled and monitored online by a Bioprocessor (Applikon bioprocessor ADI 1030, Schiedam, The Netherlands) connected to the Biodacs data acquisition program (Applikon, Schiedam, The Netherlands). The systems were controlled at a temperature of 20 C, and pH of 7.0 using 1 N HCl or 1 N NaOH.
Each reactor was operated continuously for 60e90 days, in cycles of 4 h with 10 min of aerobic mixing time (0e10 min in the cycle), 3 min (SBR 1, 2, 3, 6 and 7) or 50 min (SBR 4 and 5) of aerobic filling time (10e13 min or 10e60 min in the cycle), 140 min of aerobic reaction time (10e150 min in the cycle), 2 min of sludge withdrawal (148e150), 80 min of settling time (150e230 min in the cycle) and 10 min of effluent discharge time (230e240 min in the cycle). At the end of the cycle 1 L effluent was pumped out of the reactors, resulting in a volumetric exchange ratio of 0.5 and in a hydraulic residence time of 8 h. The applied organic loading rate was 1.2 g chemical oxygen demand (COD) L1 day1 in all experiments. It was tried to keep the solid retention time (SRT) at 10 days but due to a deficient control of the sludge withdrawal pump the actual SRT in pseudo-steady state was considerably different in SBR 2 (7.0 days) and 3 (15.6 days) (Table 1). The length of the aerobic feed phase (or the aerobic fill time ratio, calculated as the quotient of the time for aerobic fill and the total time of one cycle), the type of organic substrate and the mixed liquor dissolved oxygen concentration in the feast phase, when external substrate is present (comparable to the selector), were the operational parameters that were varied in SBR setups (Table 1). Systems with a short feed period (pulse feed in 3 min in SBRs 1, 2, 3, 6, 7, in Table 1) had initially a high substrate concentration as in a process with a plug flow selector with full substrate removal. The other systems, i.e. SBRs 4 and 5, can be compared to a (over-dimensioned) completely mixed selector, with low substrate concentrations, followed by a completely mixed reactor. The reactors were stirred with two standard geometry sixblade turbines. In each cycle two different stirrer speeds were used: 300 rpm during the first hour (allowing a fast mixing and good aeration in the feast phase) and 150 rpm during the remainder famine phase of the cycle (to decrease the turbulence, minimizing the floc break-up). In normal operation, the bulk liquid dissolved oxygen concentration in the feast phase 1 (Sfeast O2 ) was maintained at different levels (i.e., >2.0 mg O2 L 1 in SBR 1e5; <0.2 mg O2 L in SBR 6e7) by adjusting the airflow rate (i.e., 1.0 NL min1 in SBR 1e5 and 0.25 NL min1 in SBR 6e7) and stirring. In the famine phase, when external
Table 1 e Operational parameters e aerobic fill time ratio, organic substrate, dissolved oxygen concentration in the feast phase (Sfeast O2 ), biomass concentration, solids retention time, feast period, sludge loading rate, and floc loading rate e in pseudo-steady state systems. SBR system
Aerobic fill phase Time (min)
1 2 3
10e13 10e13 10e13
4 5
10e60 10e60
6 7
10e13 10e13
Fill time ratio (%) 1.25 1.25 1.25 20.8 20.8 1.25 1.25
Sfeast Sludge Solids Feast Sludge Floc Organic O2 mg O2 L1 content retention period loading rate loading rate substrate in mg TSS L1 time day min kgCOD kgTSS1 kgCOD kgTSS1 normal cyclesa day1 day1 Starch Maltose Starch:maltose (COD basis 1:1) Starch Glucose; starch after 40 days Starch Glucose
>2 >2 >2
4260 3500 5300
11.3 7.0 15.6
>2 >2
4100 3700
8.8 9.5
<0.2 <0.2
5020 4040
11.3 10.3
4.7 3.0 4.6
0.28 0.34 0.22
50 50
0.29 0.32
21 14.5
0.24 0.30
23 27 18 1.4 1.6 19 24
a When the pseudo-steady state was reached transient responses to dump fill of starch, maltose, glucose and mixture starch-maltose on a COD basis 1:1, were measured in each system.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 3 e2 7 3
substrate is depleted, the dissolved oxygen concentration was in all SBR’s always above 2 mg O2 L1 to prevent any effect of oxygen limitation in this phase (Table 1).
2.2.
Feed solution
The synthetic wastewater used as influent was sterilized at 110 C during 40 min. Depending on the SBR system (Table 1) it contained potato starch (filterable through 0.2 mm pore size filters after sterilization, Fulka 85643), maltose (Sigma 5885), or glucose (J. T. Baker 0115) 12.5 CmM (400 mg COD L1) as the carbon source. Glucose was selected as the representative monomer of saccharides, and maltose was used since it is the main compound (a di-mer) generated through the enzymatic hydrolysis of starch. The rest of the nutrients were added by the feed solution containing NH4Cl 1.5 mM (21 mg NL1), KH2PO4 0.48 mM (15 mg PL1), MgSO4$7H2O 0.37 mM, and 1 mL. 1 L1 influent of the following trace solution: EDTA 50 g L , 1 1 ZnSO4$7H2O 22 g L , CaCl2$2H2O 8.18 g L , MnCl2$4H2O 5.06 g L1, FeSO4$7H2O 4.99 g L1, (NH4)6Mo7O24$4H2O 1.1 g L1, CuSO4$5H2O 1.57 g L1, CoCl2$6H2O 1.61 g L1.
2.3.
Activated sludge inoculums
A mixture of fresh sludge from a domestic wastewater treatment plant (1.5 L) and sludge adapted to potato starch (0.5 L), which has been stored at 4 C during 6 months, was used as inoculum in SBRs 1 and 3. The other systems, i.e., SBRs 2, 4, 5, 6 and 7, were inoculated with a mixture of fresh activated sludge (0.3 L) and sludge (1.7 L) coming from SBRs 3, 1, 4, 5 and 2, respectively.
2.4.
Calculation procedures
Achievement of a pseudo-steady state was evaluated from a constant total organic carbon (TOC) and biomass concentration (given by the volatile suspended solids) in the reactor, and when the dynamic pattern in dissolved oxygen during a cycle (i.e., the length of the feast phase) no longer changed. When the organic substrate was consumed the dissolved oxygen concentration rapidly increased. This transition point was used to determine the feast and famine periods. In pseudo-steady state, full SBR cycles were analysed for kinetic characterization of the sludge. This was done through dump filling the SBR systems (the influent was instantaneously added at cycle time of 10 min) with different sugars: starch, maltose, glucose and mixture starchemaltose on a COD basis 1:1. Therefore, kinetics was measured both in normal operational conditions (with the same type of sugar as in pseudosteady state systems) and in transient operational conditions (with a different type of sugar). The experiments were performed in triplicates. The reported data are the average, the standard deviation was always below 7% for clarity this was however not included in the tables. Before each sampling cycle, 100 mg allylthioureum was added to the reactor to prevent interference of nitrification in the measurements. The behaviour of the reactors was characterized by sludge volume index (SVI), specific organic substrate uptake rate in the ), specific poly-glucose storage compounds feast phase (qfeast s production rate and fraction of poly-glucose produced in feast the feast phase (qfeast polyglucose and fpolyglucose ), fraction of sugars
265
consumed and used for synthesis of poly-glucose (Sugarfeast polyglucose ), and specific oxygen consumption rate in ). Sugarfeast the feast phase (qfeast polyglucose was calculated as the O feast feast ratio between qs =qpolyglucose and the maximum yield of polyglucose storage compounds on glucose (0.91 Cmol Cmol1, Dircks et al., 2001). The first samples were taken just before addition of the feed. The last sample was taken just before the settling phase. The feast phase started with the feed phase, and ended when the organic substrate was fully consumed and, simultaneously, internal poly-glucose degradation started. The elemental biomass composition was assumed to be CH1.56O0.59N0.19 (Dircks et al., 2001). The poly-glucose content of the active biomass was calculated as the amount of sugars measured at a certain time in the cycle minus the sugars measured after 24 h of famine aerobic conditions, which were assumed to be mainly intracellular non-stored sugar compounds. All the remaining calculation procedures are described elsewhere (Martins et al., 2003a; Dircks et al., 2001).
2.5.
Hydrolytic activity
The location of hydrolytic activity during starch degradation was determined in batch experiments with (initial ratio substrate/microorganisms of 45 mg starch g1MLVSS) and without biomass. Tests in the absence of biomass were performed using 1 L of mixed liquor collected after 30 min of settling (cycle time of 180 min) and filtered throughout 0.2 mm pore size Gelmann filters. Nitrogen gas was sparged in the bulk liquid during the batch experiments and pH was controlled at 7.0. A spike of potato starch was applied and samples were taken over the time. Starch and TOC (total and soluble) were measured and the hydrolytic activity in the bulk liquid was calculated. In the presence of biomass the hydrolytic activity was detected by staining with iodine. The hydrolytic activity was also evaluated after inactivation of the metabolic activity of the biomass, either by sparging the bulk liquid with nitrogen or adding formaldehyde (1 mL in 100 mL of activated sludge).
2.6.
Analysis
Starch, glucose, poly-glucose storage compounds, soluble and total organic carbon, poly-b-hydroxybutyrate (PHB), ammonium, mixed liquor total and volatile suspended solids, sludge volume index (SVI), pH, dissolved oxygen, oxygen uptake rate, carbon dioxide and hydrolytic activity were measured during the experiments. Carbon dioxide in the incoming air and in the off-gas was measured online with an infrared carbon dioxide analyser. For glycogen determination 4.5 mL of sample was added to 0.5 mL of 6 M HCl. The samples were then digested at 100 C during 5 h. After cooling, the samples were centrifuged and the glucose concentration was measured with highpressure liquid chromatography (HPLC). Starch was determined by the starcheiodine complex (SIC) method (San Pedro et al., 1994). The remaining analytical methods used are as described in (Martins et al., 2003a). Microscopic analysis of the activated sludge was performed according to the reference manuals (Jenkins et al., 1993; Eikelboom, 2000). The method of subjective scoring of filamentous bacteria abundance (Jenkins et al., 1993) was used to
266
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 3 e2 7 3
quantify the abundance of filamentous bacteria present in the samples. Whenever available, fluorescent 16S rRNA probes (Interactiva, Ulm, Germany) were used to identify specific filamentous bacteria. The samples were fixed with paraformaldehyde for Gram-negative bacteria analysis or with ethanol for Gram-positive bacteria analysis. Fluorescent probes were hybridised in situ according to (Manz et al., 1992). Table 2 shows a list of oligonucleotide probes used in this study. Samples for fluorescent in situ hybridisation (FISH) to detect Gram-positive bacteria were treated with 0.1% lysozyme during 15 min at 37 C in 10 mM phosphate buffer, pH 6.5. The effect of the type of substrate, i.e., polymer or monomer, on floc morphology was quantitatively evaluated in the systems SBR 1, 2 and 3 by an image analyser (Galai Cue 2 v. 4.7) coupled with stereo microscopy by a CCD camera. Equivalent diameter, superficial area, aspect ratio and shape factor were the measured parameters. Six hundred particles per sample were analysed in each sample and the reported results are an average of five samples.
Results
3.1.
General observations
The average values of some traditional operating SBR parameters, such as the fill-time ratio, dissolved oxygen concentration in the feast phase, biomass concentration, solids retention
Table 2 e Oligonucleotides probes used in this study.
EUB338
Specificity
Many but not all bacteria EUB338-II Planctomycetales EUB338-III Verrucomicrobiales HGC69a High-GC, Actinobacteria LGC354A, LGC354B, Low-GC, Firmicutes LGC354C ALF968 a-Proteobacteria BET42a b-Proteobacteria GAM42a g-Proteobacteria CF319a Cytophaga-Flavobacterium of Bacteroidetes MPA60, MPA223, ‘Microthrix parvicella’ MPA645 SNA23a ‘Sphaerotillus natans’ PAO462, PAO651, ‘Candidatus PAO846 Accumulibacter phosphatis’ GAOQ431, ‘Candidatus GAOQ989 Competibacter phosphatis’ AMAR839 Genus Amaricoccus ‘G-Bacteria’ TET63 Tetrasphaera spp. ‘G-Bacteria’ actino-1011 A High-GC group closely related to Tetrasphaera sp. NlimII175, Eikeloom morphotype NlimII192 Nostocoida limicola II
Reference Amann et al., 1990 Daims et al., 1999 Daims et al., 1999 Roller et al., 1994 Meier et al., 1999 Manz Manz Manz Manz
et al., et al., et al., et al.,
Sludge characteristics
The activated sludge from the wastewater treatment plant, which was used as partial inoculum in SBRs 1 and 3, contained Microthrix parvicella (confirmed with specific 16S rRNA probe, excessive amounts) and a range of other filamentous bacteria (mainly Types 0041/0675 and Thiothrix sp., very common) in high number. Consequently, the initial SVI in these systems was considerably high (320 mL g1) (Fig. 1). The abundance of these bacteria decreased and after two weeks the SVI was less than 130 mL g1. Except in SBR 5 well settling sludge (SVI < 100 mL g1) was always obtained. Pulse fed systems with monomers (e.g., maltose in SBR 2) produced a slightly better SVI (in the range 30e50 mL g1) than pulse fed systems with starch (SVI in the range 80e100 mL g1, SBR 1 and 6). The only bulking event was registered in the system fed during a long period with glucose (SBR 5, SVI of 340 mL g1 after 40 days). Changing the substrate to starch led to a considerable improvement in the sludge settleability, with SVI decreasing to less than 120 mL g1 after 10 days. Sphaerotilus natans was the filamentous bacteria responsible for the bulking event in SBR5 (Fig. 2c). After starch has been used in the feed solution S. natans was out competed by other bacteria and decreased in number. Higher organisms were also commonly observed in the systems. While rotifers (metazoa) were abundant in systems
1992 1992 1992 1996
Erhart et al., 1997
400 300 -1
Probe name
3.2.
SVI (ml g )
3.
time, feast period, sludge loading rate, and floc loading rate are shown in Table 1. Pseudo-steady state was generally reached one month after inoculation. Except in SBRs 4 and 5 the substrate was always present in excess during the feast phase. The soluble substrate uptake rates were, however, not always the maximum rate due to a slow hydrolysis of polymeric substrate (SBRs 1 and 3) or to oxygen limitation (SBRs 6 and 7). Only in SBR 2 the substrate, maltose, was taken up at the maximum rate. In SBRs 4 and 5 the addition of substrate was relatively slow and the uptake rates were determined by the addition rate. SBR 4 simulates the slow release of soluble substrate due to hydrolysis.
Wagner et al., 1994 Crocetti et al., 2000
Crocetti et al., 2002
Maszenan et al., 2000
200 100 0 0
Kong et al., 2001 Liu et al., 2001
Liu and Seviour, 2001
15
30 45 60 75 Experiment time (days)
90
Fig. 1 e Sludge settleability expressed as SVI during the whole operating period: (B) SBR 1; ( ) SBR 2; (D) SBR 3; (O) SBR 4; (,) SBR 5; (B) SBR 6; (>) SBR 7. The arrow indicates the day (40) when the feed in SBR 5 was changed from glucose to starch.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 3 e2 7 3
267
Fig. 2 e Photomicrographs of fluorescent in situ hybridization (FISH) (a, c, and d) and a phase contrast micrograph (b) of sludge at pseudo-steady state in: a and b pulse fed system with starch (SBR 1); c long fed system with glucose (SBR 5); d pulse fed system with glucose at low dissolved oxygen concentration (SBR 7). The length of the bars corresponds to: a and c 20 mm; b 100 mm; d 10 mm. fed with starch (Fig. 2b), vorticella spp. (protozoa) predominated in the monomers fed systems. The floc morphology was partially quantified in SBRs 1, 2 and 3 (Table 3). Compact, smooth and larger shaped flocs (equivalent diameter usually larger than 250 mm) were observed in the maltose fed system (SBR 2). All the starch fed systems were dominated by smaller and porous flocs (equivalent diameter usually smaller than 200 mm). Although not the dominant microorganisms, short and coiled filaments (trichome length usually smaller than 200 mm) resembling the morphotype Nostocoida limicola II, placed inside the floc, were commonly observed in the starch fed systems. This filamentous bacterium belonged to the Gram-positive bacteria with high DNA G þ C content group (positive signal with the probe HGC69a, Fig. 2a) but it did not hybridize with the specific probes designed for its detection (Liu and Seviour, 2001). In all the systems the dominant bacterial group was the Gram-positive high DNA G þ C content group, followed by the b- and asubclass of Proteobacteria (Bet42a and Alf968 16S rRNA probes),
Table 3 e Morphological characteristics of the flocs in SBR systems 1, 2 and 3. Parameter
Projected area Aspect ratio Shape factor Equivalent diameter
Units
mm2
mm
SBR system 1
2
3
0.02 0.62 0.31 160
0.04 0.64 0.49 350
0.02 0.64 0.33 190
roughly estimated as being respectively 50%, 30% and 5% of the total bacteria population (EUB338 þ II þ III 16S rRNA probes). G-bacteria belonging to the genus Amaricoccus of the a-subclass of Proteobacteria were not detected. In the pulse fed system with glucose at low dissolved oxygen concentration coccoid cells predominantly in tetrad arrangements were clearly dominant. These bacteria also hybridized positively with the HGC69a 16S rRNA probe (Fig. 2d). No hybridization occurred with the oligonucleotide probes TET63 and action-1011, specific respectively for Tetrasphaera spp. ‘G-Bacteria’ and a high-GC group closely related to Tetrasphaera spp.
3.3.
Cycle behaviour
Bulk liquid dissolved oxygen concentration and carbon dioxide profiles had a similar trend throughout the cycles of the different experiments (Fig. 3). In the first 10 min the bulk liquid dissolved oxygen concentration and the stripping rate of carbon dioxide increased due to mixing and aeration of the sludge. When the organic substrate was added the dissolved oxygen concentration decreased due to organic substrate consumption. When the carbon source was consumed the dissolved oxygen concentration rapidly increased, indicating the end of the feast phase. The stripping of carbon dioxide decreased in the feed phase due to low bicarbonate content in the feed solution. Carbon dioxide is produced at high rates in the feast phase. The increase in stripping of carbon dioxide could still be measured in the feast phase. Opposite behaviour was observed when all the substrate was consumed. The decrease in dissolved oxygen concentration and stripping of
268
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 3 e2 7 3
Fig. 3 e Typical dissolved oxygen and carbon dioxide profiles in the pseudo-steady state pulse fed system with maltose and starch (SBR 3) when exposed to a dump fill of glucose (e), maltose (d), and starch ( ). The arrows indicate the decrease of (a) oxygen consumption and (b) carbon dioxide production.
carbon dioxide after 1 h was caused by the decrease in stirring speed. During settling and withdrawal phases the mixing and the aeration were stopped and dissolved oxygen concentration decreased. Carbon dioxide was not measured during these phases because no aeration was applied. All the other monitored parameters showed a typical feastefamine cycle behaviour (Fig. 4). When the substrate was present, a linear decrease of substrate and a linear increase of poly-glucose were detected. The observed growth yield for starch was slightly higher (0.6 Cmol Cmol1) than for maltose and glucose (0.5 Cmol Cmol1), which corresponds to the lower net carbon dioxide production with starch (Fig. 3b). The observed yield of poly-glucose storage compounds on the substrate also changed with the type of oligosaccharide. While maltose gave a yield of 0.84 Cmol Cmol1 (SBR 2) starch pulse fed system gave a storage yield of 0.70 Cmol Cmol1 (SBR 1). When the substrate was present in limiting concentrations (SBR 4 and 5) the storage yield decreased about two times (0.3e0.4 Cmol Cmol1), indicating less storage and relative more substrate being used for growth in these conditions. A very small fraction of PHB was also formed, usually less than 8% of the total carbon present in the influent. In the famine phase the storage compounds were slowly consumed. Phosphate was released in the feast phase in the range of 11 (SBR 1)e44 (SBR 7) Pmmol Cmol1, indicating hydrolysis of
Fig. 4 e Typical change in concentrations during a pulse feed cycle with glucose at (a) high dissolved oxygen concentration (SBR 3) and (b) low dissolved oxygen concentration (SBR 7): (,) glucose; (A) poly-glucose storage 3L compounds; (3) PHB; (B) NHD 4 ; (:) PO4 . poly-P to produce energy for the active transport of saccharides through the cytoplasmic membrane. In the glucose pulse fed system operated at low dissolved oxygen concentration (SBR 7) more phosphate was released in the feast phase (about 0.35 Pmmol L1) than in the remaining systems (usually less than 0.15 Pmmol L1). Phosphorus removal was observed in SBR 7 since the phosphate concentration in the effluent was usually lower than 1 mg PL1. The phosphorous content of the biomass in this system was 5% of the cellular dry weight. Specific gene probes for phosphorus accumulating organisms (PAOs) and glycogen non-polyphosphate accumulating organisms (GAOs) (Crocetti et al., 2000, 2002) were applied but no positive signal was obtained, indicating the absence of these organisms. Neisser staining revealed the presence of positive granules, presumably poly-P, inside the dominant coccoid gram-positive bacteria with high DNA G þ C content, predominantly arranged in packages of tetrads.
4.
Discussion
4.1. Hydrolytic activity, kinetics and storage aspects of starch conversion In pseudo-steady-state systems the starch was quickly and completely adsorbed on the biomass, as observed by the iodine
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 3 e2 7 3
test (blue spots on the floc surface and not in the bulk liquid, indicating the formation of the starcheiodine complex at floc level). This indicates that the hydrolysis was strongly associated with the activated sludge flocs. Starch was hydrolysed at a rate in the range of 500 (SBR 4)e2100 (SBR 1) mg starch L1 h1. Glucose and maltose (analysed by HPLC), and other soluble sugars (by total organic carbon) were not detected in the bulk liquid. This indicates that the hydrolytic products are consumed close to the site of production, i.e. at floc level. The hydrolysis products did not diffuse outside of the floc as hypothesized by others (Ekama and Marais, 1986). The maximum hydrolytic activity measured in the cell-free bulk liquid of SBRs 1, 3, 4 and 6 varied from 5 to 33 mg starch L1 h1, indicating that a very small fraction of starch (less than 7%) could be hydrolysed in the bulk liquid. Starch hydrolysis in the bulk liquid followed first order reaction kinetics, with a reaction rate constant in the range of 0.02 (SBR 1)e0.2 (SBR 3) h1. Inhibition of cell metabolism in the sludge flocs gave a similar hydrolytic activity, which is in line with studies indicating that most of the enzymes are extracellular and immobilized in the exopolymeric substances (EPS) matrix (Frølund et al., 1995). The maximum hydrolysis rate obtained in this study (2100 mg starch L1 h1; 0.47 Cmol Cmol1 h1; 11.9 gCOD gCOD1 d1) is slightly higher than the rates reported in literature (0.30 Cmol Cmol1 h1 in San Pedro et al., 1994 and 3 gCOD gCOD1 d1 in Gujer et al., 1999). The high hydrolysis rate was, however, four to five time times smaller than the maximum soluble monomer uptake rate (2.51 Cmol Cmol1 h1 for maltose in this study, and 2.02 Cmol Cmol1 h1 for glucose in Dircks et al., 2001). A similar trend was observed for the poly-glucose specific production rates. Storage in starch pulse fed systems was slightly less than that of monomer fed systems, indicating that more growth occurred in the former just as if the monomer had been dosed at a slower rate. Even so, the storage capacity with starch was very high (almost 80% of the starch was stored, similar to the result of 75% storage obtained by Karahan et al., 2006), indicating that storage of polymeric substrates can occur as implicitly assumed in the Activated Sludge Model No. 3 (Gujer et al., 1999). Furthermore, these results confirm that storage polymers are an intrinsic part of microbial physiology and ecology of activated sludge processes (van Loosdrecht et al., 1997; Goel et al., 1998). The results of the kinetic characterization of the activated sludge from the different systems are summarized in Table 4. The specific substrate uptake rate in system SBR 1 reflects the rate-limiting step, i.e. hydrolysis of starch. In the long fed systems (SBRs 4 and 5) or with low dissolved oxygen in the feast phase (SBRs 6 and 7) the specific substrate uptake rate reflects mainly the feeding pattern of the system. When to these systems a pulse feed was applied to and oxygen was not limiting the specific substrate uptake rate increased more than two times. The obtained rates were, however, considerably lower than the rates obtained in the continuously pulse fed systems (e.g. 0.44 Cmol1 Cmol1 h1 in SBR 5, where the maximum glucose uptake rate in pulse fed systems was 2.02 Cmol1 Cmol1 h1, in Dircks et al., 2001).
4.2.
Microbiology aspects
Bulking sludge only occurred when the monomer was added at a strong limiting rate (SBR 5, fill time ratio of 25%;
269
q=qmax ¼ 0.10 Cmol Cmol1), similar to observations of previous s reports (Martins et al., 2003a,b). S. natans was the dominant filamentous bacterium in SBR 5 and it was apparently absent at conditions of low dissolved oxygen concentration (SBR 7). These observations are not in agreement with the widely known relationship between filamentous bacteria, causative conditions and control measures (Jenkins et al., 1993), in which S. natans is classified as being a bacterium commonly found in environments with low dissolved oxygen concentration and abundance of soluble substrates. Although they might be correct for some bacteria such relationships have still a high degree of uncertainty and precaution should be a rule whenever they are applied. Surprisingly in strongly diffusion limited oxygen environments, with high concentration of soluble substrate in the bulk liquid (SBR 7), filamentous bacteria were not dominant and well settling sludge was always obtained. These results are conflicting with previous observations in similar conditions, but in which acetate was used as carbon source instead of glucose (Martins et al., 2003b). It might well be that on glucose-fed systems easily a population developed that could use polyphosphate as energy source for glucose uptake and storage; compensating for the lack of oxygen. This aspect was not further investigated but, as shown in other studies (Martins et al., 2004c), the presence of bio-P activity clearly induces a well settling sludge. With the currently available FISH probes we could not identify the involved bacteria. It shows however how complicated the relation between microbial ecology and bulking sludge is.
4.3.
Relation polymeric substrate and sludge settleability
The model polymeric substrate used in this study (potato starch) did not lead to bad settling sludge in any of the experiments. Conditions leading to bulking with soluble monomers (i.e. slowly fed systems simulating a low substrate concentration in the bulk liquid during substrate uptake as occurring in completely mixed systems) did not produce bulking sludge with polymeric substrate. Replacing the soluble monomer (glucose) with starch even cured the bulking event (Fig. 1). In the experiments filamentous organisms belonging to the Gram-positive bacteria with high DNA G þ C content group were commonly present inside the flocs in the systems fed with starch (Fig. 2a). Although not dominant, their presence together with smaller, porous, fluffy and irregularly shaped and, apparently, less compact flocs led to a slightly higher SVI (80e100 mL g1) than in the soluble substrate fed systems (e.g. SBR 2, SVI ¼ 30e50 mL g1). The occurrence of a more open floc structure in the presence of hydrolysable substrates might be the result of the fact that most of the starch adsorbs to the EPS e bulk liquid interface, where it has to be hydrolysed before the monomers can diffuse into the EPS matrix and reach the microbial cells. If the micro-colonies in a floc get too large the bacteria in the interior of the microfloc would not get any substrate and will decay. It could also be that filamentous bacteria create a very open floc matrix like often observed (Wile´n et al., 2003). Since the filamentous bacteria did not extend outside the flocs we propose that the effect is mainly due to the formation of smaller microflocs that flocculate into the larger floc structure. Interestingly in biofilms (Mosquera-Corral et al., 2003) and aerobic granular sludge (de Kreuk et al., submitted for publication) the
270
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 3 e2 7 3
Table 4 e Average specific rates in normal and pulse feed cycles, fraction of poly-glucose storage compounds produced in the feast phase and fraction of sugar consumed and used for synthesis of poly-glucose. All SBR 5 measurements were made during the bulking episode. Values in bold indicate maximum rates or fractions. Values between brackets were calculated. Parameter
Unit
C source
SBR system 1
qfeast s
qfeast polyglucose
feast
qO2
qfeast s qmax s
qfeast polyglucose qfeast s
qfeast polyglucose qmax polyglucose
feast fpolyglucose
f Sugarfeast polyglucose
a b c d e f
Cmol Cmol1 h1
Cmmol Cmol1 h1
O2 mmol Cmol1 h1
Cmol Cmol1
Cmol Cmol1
Cmol Cmol1
Cmol Cmol1
%
Starch Maltose Starch þ maltose Glucose Starch Maltose Starch þ maltose Glucose Starch Maltose Starch þ maltose Glucose Starch Maltose Starch þ maltose Glucosed Starch Maltose Starch þ maltose Glucose Starch Maltose Starch þ maltose Glucosee Starch Maltose Starch þ maltose Glucose Starch Maltose Starch þ maltose Glucose
0.47 0.40 0.44 0.42 0.33 0.32 0.32 0.33 12 27 23 33 1.0 0.16 0.85 0.21 0.70 0.80 0.73 0.79 1.0 0.15 0.84 0.22 33 37 29 36 77 88 80 86
2 2.51 1.01 2.12 0.72 46 50 1.0 0.50 0.84 0.71 1.0 0.48 65 46 93 78
3
4a
0.47 0.59 0.52 0.53 0.35 0.52 0.38 0.37 17 27 24 47 1.0 0.24 1.0 0.26 0.74 0.88 0.73 0.70 1.0 0.25 1.0 0.25 34 41 29 32 82 97 80 77
0.07 (0.18)
5b
6c
7c
0.10 (0.28) 0.39
0.24 0.01 (0.07)
0.10 (0.44)
0.38
0.20 (0.65)
0.17 0.15 27 (25)
0.03 (0.26)
0.13 (0.40) 7 (20)
34 44 0.15 (0.38)
24 (49)
39 0.21 (0.60)
14 (31)
0.16 0.12 0.14 (0.39)
0.05 (0.22)
0.19
0.10 (0.32)
0.44 0.63 0.03 (0.21)
0.30 (0.59)
0.65 (0.62)
0.08 0.10 12 (19)
0.02 (0.17)
0.09 (0.27)
30 21 16 (43)
20 (48)
31 (30)
48 69
33 (65)
71 (68)
after a pulse of potato starch. in cycles with a pulse feed of glucose (bulking sludge event). in cycles with Sfeast > 2.5 mg O2 L1. O2 max qs ¼ 2.02 Cmol Cmol1 h1 (Dircks et al., 2001). 1 1 qmax h (Dircks et al., 2001). polyglucose ¼ 1.49 Cmol Cmol max Ys;polyglucose ¼ 0.91 Cmol Cmol1 (Dircks et al., 2001).
use of starch instead of maltose/sugar lead to a more open and porous biofilm/granule surface. Also here it can be reasoned that the polymeric substrates have to be hydrolysed before the monomers can diffuse in the EPS matrix, a more open structure gives enough sites for the polymer to be adsorbed and hydrolysed (de Beer and Stoodley, 1995). This is further discussed in de Kreuk et al. (submitted for publication). A generalization is not yet possible since biofilms growing on proteins versus amino acids showed a different behaviour (Mosquera-Corral et al., 2003), indicating that more research needs to be done. For instance, the localisation and activity of different enzymes (e.g. glucosidases and proteases) need more investigation. Bulking sludge has been reported in some activated sludge systems fed with complex synthetic solutions or the particulate fraction of municipal wastewater (Lakay et al., 1999). Nothing was mentioned in these studies about the place where hydrolysis occurs. The reported occurrence of bulking events with particulate COD could be the result of an important fraction of
particulate substrate being hydrolysed in the bulk liquid, which certainly gives advantages to filamentous bacteria. For instance, in biofilm systems the hydrolytic activity has been found mainly associated with the bulk liquid in unsteady-state periods (Mosquera-Corral et al., 2003), conditions which are also likely to occur in complex dynamic activated sludge systems. These aspects emphasize the importance of localizing the hydrolytic activity, i.e. whether the enzymes are associated with the biological floc (i.e. cell surface or EPS matrix) or released to the bulk liquid, whenever bulking sludge is reported in systems fed with particulate substrates.
4.4.
Diffusion theory as basis for bulking sludge
The occurrence of concentration gradients over the sludge floc due to diffusion limitation were recently hypothesized as being the most important factor for the development of filamentous bacterial structures, and, eventually, bulking sludge, and not
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 3 e2 7 3
5.
271
Conclusion
Particulate substrates are incorporated in the sludge floc and subsequently hydrolysed. Despite that this will result in a low concentration of soluble substrate (hydrolysed monomers), and the risk of selection of filamentous organisms, bulking sludge did not occur. It is hypothesized that this is due to the fact that the hydrolysis products are uniformly distributed inside the floc. When substrate is taken up at low concentrations form the bulk solution this would lead to gradients over the floc giving advantage to filamentous organisms which extend outside the floc. Consequently occurrence of substrate gradients, easily developing at low substrate concentrations, seem to be more important then a kinetic selection for the proliferation of filamentous bacteria.
Acknowledgment Fig. 5 e Schematic representation of soluble substrate concentration (Cs) in the floc and in the bulk liquid after applying a pulse feed of monomer (maltose or glucose) or polymer (starch). Phase-contrast photomicrograph of typical biological floc at pseudo-steady state in the pulse fed system with starch (SBR 1).
the substrate concentration as such (Kappeler and Gujer, 1994; Martins et al., 2004a,b; Lou and De Los Reyes, 2008). Bacterial morphology and bacterial physiology are also important factors but the trigger for the development of different types of bacterial structures is the presence of gradient-governed environments, typical of substrate limited activated sludge systems. According to this theory in the presence of substrate diffusion limitation inside the biological flocs, filamentous bacterial structures and, thus, filamentous bacteria, have a higher outgrowth velocity because they grow preferentially in one direction, and not in three directions as floc forming bacterial structures (Martins et al., 2004b). If substrate diffusion limitation does not exist, more regularly shaped and compact bacterial structures are expected. The experimental observations in this study are in line with this theory. In pseudo-steady state systems starch hydrolysis takes place inside the flocs, not giving rise to strong gradients of substrate concentration over the floc radius (Fig. 5). The concentration of monomers at floc level is however low and rate limiting, a condition which according the kinetic selection theory would lead to bulking sludge (Chudoba et al., 1973). Filamentous bacteria were detected in the starch fed systems but they remained inside compact micro-colonies and did not show the typical floc morphology behaviour of bulking sludge, similar to observation by Liao et al. (2004). Note that for bulking sludge to occur only a small fraction of filamentous bacteria is sufficient to raise the SVI from 100 to 200 mL/g. Therefore, it is likely that the soluble substrate concentration, although potentially selective for filamentous bacteria, as such is not the most important factor for bulking sludge. The micro-gradients of substrate concentration as hypothesized by the diffusion based theory are a more likely cause of sludge bulking.
The authors gratefully acknowledged the assistance rendered by the analytical and technical staff of the Department of Biochemical Engineering. The technical support given by the Erasmus student Stefania Ortu from University of Cagliari in this study is highly appreciated. Anto´nio Martins received financial support from the Portuguese State in the context of PRAXIS XXI by the Doctoral Scholarship BD/19538/99.
Nomenclature COD chemical oxygen demand, mg L1 FISH fluorescent in situ hybridisation feast fraction of poly-glucose storage compounds fpolyglucose produced in the feast phase, Cmol Cmol1 PHB poly-b-hydroxybutyrate specific oxygen consumption rate in the feast phase, qfeast O O2 mmol Cmol1 h1 qfeast polyglucose specific production rate of poly-glucose storage compounds in the feast phase, Cmol Cmol1 h1 qmax polyglucose maximum specific production rate of poly-glucose storage compounds, Cmol Cmol1 h1 feast qs specific substrate uptake rate in the feast phase, Cmol Cmol1 h1 max qs maximum specific substrate uptake rate, Cmol Cmol1 h1 feast SO2 bulk liquid dissolved oxygen concentration in the feast phase, mg O2 L1 SBR sequencing batch reactor SRT solids retention time, day Sugarfeast polyglucose fraction of sugars consumed and used for synthesis of poly-glucose storage compounds, % SVI sludge volume index, mL g1
references
Amann, R.I., Binder, B.J., Olson, R.J., Chisholm, S.W., Devereux, R., Stahl, D.A., 1990. Combination of 16S rRNA-targeted oligonucleotide probes with flow cytometry for analyzing
272
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 3 e2 7 3
mixed microbial populations. Applied Environmental Microbiology 56 (6), 1919e1925. de Beer, D., Stoodley, P., 1995. Relation between the structure of an aerobic biofilm and transport phenomena. Water Science and Technology 32 (8), 11e18. Casey, T.G., Wentzel, M.C., Ekama, G.A., 1999. Filamentous organism bulking in nutrient removal activated sludge systems. Paper 11: a biochemical/microbiological model for proliferation of anoxic-aerobic (AA) filamentous organisms. Water SA 25 (4), 443e451. Chudoba, J., Grau, P., Ottova´, V., 1973. Control of activated-sludge filamentous bulking-II. Selection of microorganisms by means of a selector. Water Research 7 (10), 1389e1398. 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. Applied Environmental Microbiology 66 (3), 1175e1182. Crocetti, G.R., Banfield, J.F., Keller, J., Bond, P.L., Blackall, L.L., 2002. Glycogen-accumulating organisms in laboratory-scale and full-scale wastewater treatment processes. Microbiology 148 (11), 3353e3364. Daims, H., Bruhl, A., Amann, R., Schleifer, K.-H., 1999. The domain-specific probe EUB338 is insufficient for the detection of all bacteria: development and evaluation of a more comprehensive probe set. Systematic and Applied Microbiology 22, 434e444. Dircks, K., Beun, J.J., Van Loosdrecht, M.C.M., Heijnen, J.J., Henze, M., 2001. Glycogen metabolism in aerobic mixed cultures. Biotechnology Bioengineering 73, 85e94. Dold, P.L., Fleit, E., Han, J., 1991. Hydrolysis of a (1e4) glucan bonds in activated sludge mixed bacterial communities. Environmental Technology 12 (10), 871e879. Eikelboom, D.H., 2000. Process Control of Activated Sludge Plants by Microscopic Investigation. IWA Publishing, London, UK. Ekama, G.A., Marais, G.V.R., 1986. The implications of the IAWPRC hydrolysis hypothesis on low F/M bulking. Water Science and Technology 18 (6), 11e19. Erhart, R., Bradford, D., Seviour, R.J., Amann, R., Blackall, L.L., 1997. Development and use of fluorescent in situ hybridization probes for the detection and identification of ‘Microthrix parvicella’ in activated sludge. Systematic and Applied Microbiology 20, 310e318. Frølund, B., Griebe, T., Nielsen, P.H., 1995. Enzymatic activity in the activated-sludge floc matrix. Applied Microbiology Biotechnology 43, 755e761. Goel, R., Mino, T., Satoh, H., Matsuo, T., 1998. Intracellular storage compounds, oxygen uptake rates and biomass yield with readily and slowly degradable substrates. Water Science and Technology 38 (8e9), 85e93. Goel, R., Mino, T., Satoh, H., Matsuo, T., 1999. Modeling hydrolysis processes considering intracellular storage. Water Science and Technology 39 (1), 97e105. Gujer, W., Henze, M., Mino, T., Van Loosdrecht, M., 1999. Activated sludge model no. 3. Water Science and Technology 39 (1), 183e193. Jenkins, D., Richard, M.G., Daigger, G.T., 1993. Manual on the Causes and Control of Activated Sludge Bulking and Foaming, second ed. Lewis, Boca Raton, Fla. Kappeler, J., Gujer, W., 1994. Verification and applications of a mathematical model for aerobic bulking. Water Research 28 (2), 303e310. Karahan, O., Martins, A., Orhon, D., van Loosdrecht, M.C.M., 2006. Experimental evaluation of starch utilization mechanism by activated sludge. Biotechnology and Bioengineering 93 (5), 964e970. Kong, Y.H., Beer, M., Seviour, R.J., Lindrea, K.C., Rees, G.N., 2001. Structure and functional analysis of the microbial community
in an aerobic: anaerobic sequencing batch reactor (SBR) with no phosphorus removal. Systematic and Applied Microbiology 24, 597e609. de Kreuk, M.K., Kishida, N., Tsuneda, S., van Loosdrecht, M.C.M. Behavior of polymeric substrates in an aerobic granular sludge system. Water Research, submitted for publication. Kruit, J., Boley, F., Jacobs, L.J.A.M., Wouda, T.W.M., 1994. Prediction of the O2 conditions in the selector. Water Science and Technology 29 (7), 229e237. Lakay, M.T., Hulsman, A., Ketley, D., Warburton, C., De Villiers, M., Casey, T.G., Wentzel, M.C., Ekama, G.A., 1999. Filamentous organism bulking in nutrient removal activated sludge systems. Paper 7: exploratory experimental investigations. Water SA 25 (4), 383e396. 1999. Liao, J., Lou, I., De Los Reyes III, F.L., 2004. Relationship of speciesspecific filament levels to filamentous bulking in activated sludge. Applied and Environmental Microbiology 70 (4), 2420e2428. Liu, W.-T., Nielsen, A.T., Wu, J.-H., Tsai, C.-S., Matsuo, Y., Molin, S., 2001. In situ identification of polyphosphate- and polyhydroxyalkanoate-accumulating traits for microbial populations in a biological phosphorus removal process. Environmental Microbiology 3 (2), 110e122. Liu, J.R., Seviour, R., 2001. Design and application of oligonucleotide probes for fluorescent in situ identification of the filamentous bacterial morphotype Nostocoida limicola in activated sludge. Journal of Environmental Microbiology 3 (9), 551e560. van Loosdrecht, M.C.M., Pot, M.A., Heijnen, J.J., 1997. Importance of bacterial storage polymers in bioprocesses. Water Science and Technology 35 (1), 41e47. Lou, I.C., De Los Reyes III, F.L., 2008. Clarifying the roles of kinetics and diffusion in activated sludge filamentous bulking. Biotechnology and Bioengineering 101 (2), 327e336. 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, 593e600. Manz, W., Amann, R., Ludwig, W., Vancanneyt, M., Schleifer, K., 1996. Application of a suite of 16S rRNA-specific oligonucleotide probes designed to investigate bacteria of the phylum cytophaga-flavobacter-bacteroides in the natural environment. Microbiology 142 (5), 1097e1106. 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. 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., Heijnen, J.J., Van Loosdrecht, M.C.M., 2004c. Bulking sludge in biological nutrient removal systems. Biotechnology and Bioengineering 86 (2), 125e135. Maszenan, A.-M., Seviour, R.J., Patel, B.K.C., Wanner, J., 2000. A fluorescently-labelled r-RNA targeted oligonucleotide probe for the in situ detection of G-bacteria of the genus Amaricoccus in activated sludge. Journal Applied Microbiology 88 (5), 826e835. Meier, H., Amann, R., Ludwig, W., Schleifer, K.-H., 1999. Specific oligonucleotide probes for in situ detection of a major group of gram-positive bacteria with low DNA GþC content. Systematic and Applied Microbiology 22, 186e196.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 3 e2 7 3
Mino, T., San Pedro, D.C., Matsuo, T., 1995. Estimation of the rate of slowly biodegradable COD (SBCOD) hydrolysis under anaerobic, anoxic and aerobic conditions by experiments using starch as model substrate. Water Science and Technology 31 (2), 95e103. Mosquera-Corral, A., Montras, A., Heijnen, J.J., van Loosdrecht, M.C.M., 2003. Degradation of polymers in a biofilm airlift suspension reactor. Water Research 37 (3), 485e492. Roller, C., Wagner, M., Amann, R., Ludwig, W., Schleifer, K.-H., 1994. In situ probing of gram-positive bacteria with high DNA G þ C content using 23S rRNA-targeted oligonucleotides. Microbiology 140, 2849e2858.
273
San Pedro, D.C., Mino, T., Matsuo, T., 1994. Evaluation of the rate of hydrolysis of slowly biodegradable COD (SBCOD) using starch as substrate under anaerobic, anoxic and aerobic conditions. Water Science and Technology 30 (11), 191e199. Wagner, M., Amann, R., Kampfer, P., Assmus, B., Hartmann, A., Hutzler, P., Springer, N., Schleifer, K.-H., 1994. Identification and in-situ detection of gram-negative filamentous bacteria in activated sludge. Systematic and Applied Microbiology 17, 405e417. Wile´n, B.-M., Jin, B., Lant, P., 2003. Impacts of structural characteristics on activated sludge floc stability. Water Research 37 (15), 3632e3645.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 7 4 e2 8 2
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Carbon nanotube blended polyethersulfone membranes for fouling control in water treatment Evrim Celik, Hosik Park, Hyeongyu Choi, Heechul Choi* Department of Environmental Science and Engineering, Gwangju Institute of Science and Technology (GIST), 261 Cheomdan-gwagiro, Buk-gu, Gwangju 500-712, South Korea
article info
abstract
Article history:
Multi-walled carbon nanotube/polyethersulfone (C/P) blend membranes were synthesized
Received 3 May 2010
via the phase inversion method. The resultant membranes were then characterized by
Received in revised form
scanning electron microscopy (SEM), Fourier transform infrared spectroscopy (FTIR), and
13 July 2010
contact angle. The C/P blend membranes appeared to be more hydrophilic, with a higher
Accepted 19 July 2010
pure water flux than the polyethersulfone (PES) membranes. It was also found that the
Available online 27 July 2010
amount of multi-walled carbon nanotubes (MWCNTs) in the blend membranes was an important factor affecting the morphology and permeation properties of the membranes.
Keywords:
After 24 h of surface water filtration with 7 mgC/L TOC content, the C/P blend membranes
Ultrafiltration
displayed a higher flux and slower fouling rate than the PES membranes. Subsequent
Fouling control
analyses of the desorbed foulants showed that the amount of foulant on bare PES
Multi-walled carbon nanotubes
membranes was 63% higher than the C/P blend membrane for 2% MWCNTs content. Thus,
Polyethersulfone
the carbon nanotube content of the C/P membranes was shown to alleviate the membrane
Composite membrane
fouling caused by natural water.
Hydrophilicity
1.
Introduction
Membrane technology has been commonly used worldwide for the removal of suspended solids such as microorganisms and a fraction of dissolved solids (Choi et al., 2005). Based on this technology, separation, concentration, and purification have become industrially viable because of the high separation efficiency of these membranes. Moreover, their low energy requirement, low space requirement, and simplicity of operation promote their use in separation processes (Arthanareeswaran et al., 2004). However, fouling is still a major problem limiting the wider application of membrane operations (Cheryan, 1986), which can be defined as the reversible or irreversible deposition of retained solutes such as particles, colloids, emulsions, macromolecules, salts, etc. on or in the membrane (Mulder, 1997). * Corresponding author. Tel.: þ82 62 715 2441; fax: þ82 62 715 2434. E-mail address:
[email protected] (H. Choi). 0043-1354/$ e see front matter ª 2010 Published by Elsevier Ltd. doi:10.1016/j.watres.2010.07.060
ª 2010 Published by Elsevier Ltd.
Membrane surface chemistry is a very important factor determining the performance of ultrafiltration operations (Reddy and Patel, 2008). The nature of these membranes easily induces macromolecules to deposit because of their hydrophobic regions (Blanco et al., 2006). To this end, it is well known that increasing the membrane hydrophilicity can effectively minimize membrane fouling (Wang et al., 2006), though charged membranes can also be used to reduce membrane fouling (Mulder, 1997). As such, methods such as surface graft polymerization, chemical grafting, and radiation induced grafting have been developed in attempts to increase the surface hydrophilicity of membranes (Shi et al., 2007; Wang et al., 2006). In addition, there has been a great deal of interest in the use of organiceinorganic hybrid membranes as a potential next generation membrane material. It is expected that such
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 7 4 e2 8 2
membranes will have the physicochemical stability of inorganic materials and the membrane forming properties of polymers (Peng et al., 2007). In particular, nano-sized inorganic material blended composite membranes are attractive because of their enhanced properties, such as high permselectivity, higher hydrophilicity, and enhanced fouling resistance (Yang et al., 2007). Previous reports have shown that there is a high potential for carbon nanotubes (CNTs) to improve the material properties of polymers. CNTs have an exceptionally high aspect ratio in combination with low density, and high strength and stiffness, which makes them a potential candidate as an effective reinforcing additive in polymeric materials (Gojny et al., 2004). Previous studies by Choi et al. (2006) have shown that multiwalled carbon nanotube (MWCNT) blended polysulfone microfiltration membranes have a slightly higher flux and rejection rates than the polysulfone membranes. These studies have also found that MWCNT blended membranes are more hydrophilic than polysulfone membranes. In a related work, Qui et al. (2009) have also shown that MWCNT blended polysulfone ultrafiltration membranes have a higher flux but much lower rejection than polysulfone membranes. Moreover, this study reported that MWCNT blended membranes have a lower protein adsorption than the polysulfone membranes. To our knowledge, polyethersulfone (PES), which is more suitable for liquid state separations (Barth et al., 2000), has not yet been used to prepare CNT blended membranes for water treatment; in particular, the water or wastewater treatment efficiency 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 synthesize multi-walled carbon nanotube/polyethersulfone (C/P) blend membranes and then to determine their anti-fouling efficiency. To characterize the C/P blend membranes, pure water flux tests, water contact angle, Fourier transform infrared (FTIR) spectroscopy and scanning electron microscopy (SEM) were employed. Then, to determine the fouling resistances of the C/P blend membranes, cross-flow permeation tests with surface water and foulant analyses were conducted.
2.
Materials and methods
2.1.
Materials
Polyethersulfone (Radel H2000) was kindly supplied by Solvay Korea Co; MWCNTs were purchased from Hanwha Nanotech. Co Ltd (Korea), and N-methyl-2-pyrrolidinone (NMP), nitric acid (70%) and sulfuric acid (98%) were purchased from SigmaeAldrich. In addition, polyethylene glycol (PEG) having an approximate molecular weight of 35 kDa and polyvinylpyrrolidone (PVP) having an approximate molecular weight of 55 kDa were purchased from Fluka and Aldrich, respectively. Deionized (DI) water was obtained from a water purification system (Synergy, Millipore, USA), having a resistivity of 18.2 mU cm. For the fouling tests, natural surface water from the Yeongsan River (Gwangju, Korea) was used; the raw water was prefiltered using a 1 mm filter to remove particulate matter and treated with 3 mg/L sodium hypochlorite as a biofouling control.
2.2.
275
Membrane fabrication
MWCNTs were functionalized in a mixture of HNO3 and H2SO4 to increase their dispersion in organic solvents (Liu et al., 1998). The typical approach is as follows; MWCNTs were refluxed in a 3:1 (v/v) HNO3:H2SO4 mixture at 100 C, before being washed with DI and left to dry at room temperature overnight. The dried MWCNTs were then ultrasonicated in 3:1 (v/v) HNO3:H2SO4 mixture at 70 C for 9 h. Finally, the MWCNTs were washed and filtered using a 0.45 mm nylon filter until the pH value of the MWCNT solution reached 7.0 0.2, and again dried in a vacuum oven at 100 C overnight. C/P blend membranes were synthesized via the phase inversion method. In brief, functionalized MWCNTs were ultrasonicated in NMP for good dispersion. After dispersing MWCNTs in solvent, PES (20 wt%) was dissolved in the dope solution by continuous stirring and heating at 60 C until the solution became completely dissolved and homogenous. The resultant polymer solution was ultrasonicated to remove air bubbles; after air bubble removal, the dope solution was casted on a glass plate using a casting knife, and the glass plate was immediately dipped in water. A thin polymeric film separated from the glass within a few minutes. The formed membranes were subsequently washed with and stored in DI water until use. Note that membranes marked as C/P-0.5% refer to membranes prepared in a casting solution in which the content of the MWCNTs with respect to PES was 0.5% by weight.
2.3.
Membrane characterization
The surface and cross-section morphology of the C/P blend membranes were directly observed by SEM (S-4700, Hitachi, Japan), and their structure characterized by FTIR (FTIR-460 plus, JASCO, Japan). The surface hydrophilicity of the blend membranes was evaluated based on dynamic sessile drop and captive bubble methods, using a contact angle goniometer (Model 100, Rame-Hart, USA). For the sessile drop method, 2 mL of DI water was dropped onto a dry membrane surface using a micro-syringe, and the contact angle was measured. For the captive bubble method, a 2 mL air bubble was released from the tip of a U-shaped needle and floated under the surface of the membrane, and this contact angle was also measured. A minimum of at least seven contact angles were averaged to ensure a reliable value was obtained for both methods. Oneway complete statistical analysis of variance (ANOVA) test at a confidence level of 95% applied to the results of contact angle measurements. The molecular weight cut-off (MWCO) of the blend membranes was determined by solute rejection measurements (Mulder, 1997). The typical approach is as follows; 1 g/L PEG or PVP solution was placed in the filtration unit and permeate samples were collected after a period of 1 h. The relative molecular mass distribution of solutes in the feed and the permeate samples was then measured by high performance liquid chromatography (LC600, Shimadzu, Japan), used in conjunction with a PEG separation column (Ultrahydrogel 1000, Waters, USA) and an RI detector (RID-6A, Shimadzu, Japan). Based on the results obtained from the PEG or PVP filtration, the MWCO of the membranes was determined.
276
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 7 4 e2 8 2
2.4.
Permeation measurements
A cross-flow membrane test unit was used for the permeation tests, in which the temperature in the unit was maintained at room temperature (21e23 C) by a circulator (DTRC-620, Jeio Tech., Korea). The test unit was comprised of a pump with a gear type pump head, relief valves (for feed and retentate streams), and a pressure gauge. The effective membrane area used for permeation measurements was 18.56 cm2, and all experiments were conducted in total recycle mode, i.e., permeate and concentrate streams were returned to the feed tank. The trans-membrane pressure was kept constant at 0.35 MPa (50 psi) when determining the pure water flux and during fouling tests with Yeongsan River water. Note that membranes were initially operated for 4 h with DI water at a 0.41 MPa (60 psi) trans-membrane pressure for stabilization. During fouling tests with Yeongsan River water, membrane filtration was performed for 8 h, after which the membranes were cleaned with 1 g/L NaOH aqueous solution for 20 min followed by DI water for 5 min under a high cross-flow velocity (28.7 cm/s). The flux decline was monitored for each step to demonstrate the fouling related flux decline of the blend membranes, and the natural organic matter (NOM) rejection of the membranes was then calculated using the following equation: Rð%Þ ¼
O
OH
O OH
HO
O S
HO
O
O
O O
O
O
S
PES
O
Scheme 1 e The structure of hydrogen bonding interactions between the sulfonic groups of PES and carboxylic groups of functionalized MWCNTs.
2.5.
Water quality analysis
Total organic carbon (TOC) and dissolved organic carbon (DOC) were measured by a TOC analyzer (TOC-VCPH, Shimadzu, Japan). DOC samples were filtered using a 0.45-mm filter paper before analysis, and a UVeVis spectrophotometer (UV-mini 1240, Shimadzu, Japan) was used to measure the absorbance at 254 nm. The specific UV absorbance (SUVA) value was calculated based on the ratio of UVA254 to the DOC concentration, and the humic fraction was estimated from the mass balance between the influent and effluent of XAD-8 resin. A turbidimeter and a pH meter were also used in the analyses.
2.6.
Cf Cp 100 Cf
Fractionation of C/P blend membrane foulants
(1)
where R is the solute rejection, and Cf and Cp are the concentrations of the feed and permeate, respectively. The feed and permeate samples were collected on the last fouling cycle of duplicate tests to assess the water quality. One-way complete statistical analysis of variance (ANOVA) test at a confidence level of 95% applied also to the results of NOM rejection tests.
XAD-8 (Rohm and Hass, USA) and XAD-4 (Supelco, USA) resins were used to determine the mass fractions of hydrophobic NOM (mostly hydrophobic acids; XAD-8 isolate), transphilic NOM (mostly hydrophilic acids; XAD-4 isolate), and hydrophilic NOM (mostly hydrophilic neutral species/bases; effluent from the XAD-4 resin that had initially passed through XAD-8 resin). NOM-fouled C/P blend membranes were removed from the membrane unit after the fouling tests. NOM foulants were then desorbed by soaking the membranes in 0.1 M NaOH solution for 24 h, and the desorbed NOM solution was acidified to pH 2 with HCl, and sequentially passed through the XAD-8 and XAD-4 resins (Aiken et al., 1992).
1384
3440
1633
(a)
advancing (sessile d.) receding (captive b.)
70
(b) 65
Contact angle ( °)
(c) 1404
Transmittance (%)
MWCNT
1404
(d)
60
55
50
45 4000
3500
3000
2500
2000
1500
1000
Wavenumbers (cm-1)
Fig. 1 e The FTIR spectra of the functionalized MWCNTs and the C/P blend membranes having different MWCNT contents: (a) functionalized MWCNTs, (b) C/P-0% membrane, (c) C/P-2% membrane, and (d) C/P-4% membrane.
40 C/P-0%
C/P-0.5%
C/P-2%
C/P-4%
Membrane type
Fig. 2 e Receding and advancing contact angles of the C/P blend membrane surfaces. (Average contact angle and standard deviation of seven replicates are reported).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 7 4 e2 8 2
277
Fig. 3 e SEM micrographs of the surfaces and cross-sections of the C/P blend membranes for different MWCNT content: (a) and (b) C/P-0% membrane; (c), (d) and (g) C/P-2% membrane; and (e), (f) and (h) C/P-4% membrane.
278
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 7 4 e2 8 2
Table 2 e Yeongsan River water characteristics after prefiltration. pH Turbidity, NTU TOC, mg/L DOC, mg/L UVA254, m1 SUVA, L/mg m Humic fraction (%)
Fig. 4 e Pure water flux of the C/P blend membranes as a function of the trans-membrane pressure.
3.
Results and discussion
3.1.
Membrane characterization
Fig. 1 shows the FTIR spectra of the functionalized MWCNTs and the surfaces of the C/P blend membranes with 0e4% MWCNTs. From Fig. 1a, it is seen that the functionalized MWCNTs exhibit three peaks: w3440 cm1 (eOH), w1630 cm1 (>C]O), and w1380 cm1 (eCOOH) (Kim et al., 2005; Kaniyoor et al., 2009). The FTIR spectra of the blend membranes are presented in Fig. 1bed. The C/P blend membranes are noticeably different from the bare PES (C/P-0%) membrane, with a new peak at w1400 cm1, corresponding to the OeH vibration of carboxyl groups (Yuen et al., 2008). Moreover, the carboxylic group density of the functionalized MWCNTs was characterized by Boehm’s titration method (Boehm, 1994; Goertzen et al., 2010) and determined to be 3.93 0.17 carboxyl group/nm2. As such, it is thought that the C/P blend membranes might be formed via hydrogen bonding interactions between the sulfonic groups of PES and the carboxylic groups of functionalized MWCNTs (Rong et al., 2010), as shown in Scheme 1. The contact angles of the surfaces of the blend membranes, as determined by the sessile drop and captive bubble methods, are shown in Fig. 2. The contact angles of the blend membranes gradually decreased up to C/P-2% membrane as the amount of MWCNTs was increased in the blend membranes. This result can be explained to be due to the fact that during the phase inversion process, hydrophilic MWCNTs migrated spontaneously to the membrane/water
Table 1 e MWCO of the C/P blend membranes. Membrane type C/P-0% C/P-0.5% C/P-2% C/P-4%
MWCO (kDa) 26.5 33.1 22.6 24.7
7.7 1.73 7.1 5.5 7.4 1.35 13
interface to reduce the interface energy (Sun et al., 2010). However, increasing the MWCNT amount to more than 2% did not result in further enhancement of the hydrophilicity (by ANOVA tests); this might be explained by the irregular positioning of MWCNTs in the membrane structure at over 2% MWCNT content (Qui et al., 2009). The contact angles measured by the sessile drop and the captive bubble methods are representative of the advancing and the receding contact angles, respectively. The difference between the advancing and receding contact angles is referred to as the contact angle hysteresis (Drelich et al., 1996; Brunet et al., 2008), with the contact angle hysteresis being larger for rougher surfaces (Drelich et al., 1996). As shown in Fig. 2, the contact angle hysteresis is larger for C/P blend membranes than for the bare PES membrane, though it is quite similar for all C/P blend membranes. This result suggests that the roughnesses of the C/P blend membranes are higher than that of bare PES membrane. The surface and the cross-section morphologies of the blend membranes were subsequently characterized by SEM micrographs (Fig. 3). All membranes showed a typical asymmetric membrane structure with a dense top layer, a porous sublayer, and fully developed macropores at the bottom. Nevertheless, the formation of macropores was suppressed by the addition of MWCNTs into the membrane structure (Fig. 3b, d, f). In addition, there was not a distinct difference in the numbers of the MWCNTs positioned on the surface layer of the C/P-2% and C/P-4% membranes (Fig. 3g, h). Moreover, the porosity of the membrane surface initially increased to C/ P-2% membrane and then decreased (Fig. 3a, c, e). This result might be explained by the fact that increasing the amount of MWCNTs in the casting solution to 2% increased the porosity of the synthesized membranes, due to the enhanced phase separation with MWCNTs. Further increases in MWCNT amount led to a denser structure in the sublayer, due to the delayed phase separation with increased viscosity (Han and Nam, 2002; Amirilargani et al., 2010). Fig. 4 shows the pure water flux as a function of the transmembrane pressure. The pure water flux was at maximum when the MWCNT content of the blend membranes was 0.5%; beyond 0.5%, the flux gradually decreased. Moreover, the MWCO of the C/P-0.5% membrane is the maximum and decreasing gradually up to 2% of MWCNTs (Table 1). Beyond 2%, MWCO increased again. This result can be explained by the dual effect of hydrophilic MWCNTs in the phase separation process. When the MWCNT content of the blend membranes was 0.5%, hydrophilic MWCNTs enhanced the phase separation, which resulted in a bigger pore size and a higher pure water flux. When the MWCNT content of the
279
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 7 4 e2 8 2
Fig. 5 e Relative flux profiles of the C/P blend membranes during filtration of Yeongsan River water.
as low-humic waters (Zularisam et al., 2007); the Yeongsan River water can be classified as hydrophilic surface water due to its low SUVA value (1.35 L/mg m). This classification is consistent with its low-humic proportion. Caustic cleaning and oxidizing cleaning agents are the most effective cleaning strategies for the membranes, which are fouled by surface water with high organic content (Zondervan and Roffel, 2007). Moreover, caustic and acidic cleanings are very effective for removing hydrophilic fractions of NOM (Zularisam et al., 2007). Since the Yeongsan River
60
TOC UVA 254
50
NOM Rejection (%)
membranes was above 0.5%, the viscosity of the casting solution increased, thereby delaying the phase separation and resulting in a smaller pore size and lower pure water flux (Han and Nam, 2002). The increase in pore size and reduction in pure water flux above 2% MWCNT content might be explained by the enhanced exchange of solvent and non-solvent during phase inversion, which made the MWCNTs regularly collocate in the membrane at up to 2% MWCNT content. However, at MWCNT content beyond 2% the steric hindrance and electrostatic interactions between the MWCNTs or between the MWCNTs and PES made some portion of the MWCNTs irregularly collocate in the membrane (Qui et al., 2009); this irregular positioning might increase the pore size of the membrane. This irregular positioning of MWCNT above 2% is consistent with the hydrophilicity of the membranes (Fig. 2). The increase in flux with C/P blend membranes is mainly because of the increase in hydrophilicity (Fig. 2) and porosity (Fig. 3). Higher increase in flux might be achieved by aligned CNTs. However, there are difficulties in alignment and anchoring of the CNTs to the surface. In addition, blend membranes with aligned CNT may have high fouling potential due to the hydrophobic nature of the pore walls of CNT (Kim and Bruggen, 2010).
40
30
20
10
0
3.2.
Fouling tests
Prefiltered Yeongsan River water was used to determine the fouling behavior of the C/P blend membranes. Characteristics of the Yeongsan River water are given in Table 2. Note that surface waters with a SUVA less than 3 L/mg m are classified
C/P-0%
C/P-0.5%
C/P-2%
C/P-4%
Membrane type
Fig. 6 e NOM rejections based on TOC and UVA of the C/P blend membranes during Yeongsan River water filtration (Average rejection and standard deviation of two replicates are reported).
280
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 7 4 e2 8 2
Table 3 e Fractionation of the desorbed NOM foulants from bare PES and C/P-2% membranes. Membrane C/P-0% C/P-2%
Desorbed NOM, mg/m2
Hydrophobic NOM, mg/m2
Transphilic NOM, mg/m2
Hydrophilic NOM, mg/m2
159.38 58.81
16.68 5.64
83.83 11.61
58.86 41.57
water has a reasonable amount of organic content and a high hydrophilic content (Table 2), caustic cleaning was used between the cycles of fouling tests. Normalized fluxes were used to analyze the fouling resistance of the C/P blend membranes. Normalized flux is defined as current flux divided by its corresponding initial flux (Fig. 5). The flux of the bare PES membrane decreased by up to 43% with respect to the operating time. However, the flux of the C/ P blend membranes reached a plateau at increased operating time; 20% for C/P-0.5% membrane and 15% for both C/P-2% and C/P-4% membranes. These results may be explained by the increased hydrophilicity of the C/P blend membranes (Fig. 2). Since hydrophobic membranes exhibit higher flux decline than hydrophilic membranes (Fan et al., 2001), the bare PES membrane showed a severe flux decline. The flux decline of the C/P blend membranes decreases for up to 2% MWCNTs content, and does not differ much at higher MWCNT amounts (Fig. 5)da similar trend to that of the contact angle for C/P blend membranes (Fig. 2). As shown in Fig. 5, after the first caustic cleaning cycle, flux recoveries of the bare PES and C/P-0.5% membranes were 85% and 87% respectively, though the flux recoveries of the C/P-2% and C/P-4% membranes were 95%. By increasing the operating time, the flux recoveries of the bare PES and C/P-0.5% membranes were reduced to 77% and 85%, respectively. However, the C/P-2% and C/P-4% membranes continued to show a stable flux recovery of 95% for each cycle. In addition, the removal of the foulant layer on the bare PES membrane was notably more difficult than on the C/P blend membranes, because of the severe fouling on its surface. To this end, Gray et al. (2008) previously reported that the flux recoveries of hydrophilic membranes are more efficient upon backwashing, which is consistent with our results. These results clearly exhibit the superior fouling resistance of the C/P2% and C/P-4% membranes. Even though C/P-0.5% membrane has a bigger pore size than bare PES membrane (Table 1), NOM rejection of C/P-0.5% membrane was not statistically different than that of bare PES membrane by ANOVA test results (Fig. 6). This result might be explained by the fact that the primary removal mechanisms of NOM with bare PES membrane are hydrophobic adsorption and size exclusion, whereas the mechanisms are electrostatic repulsion between the negatively charged NOM and negatively charged carboxylic groups on the C/P-0.5% membrane surface. The NOM rejection of C/P-2% membrane is the highest, probably due to electrostatic repulsion and size exclusion; the hydrophilicities of C/P-2% and C/P-4% membranes are relatively similar (by ANOVA test results), though C/P-4% membrane has a slightly lower NOM rejection than C/P-2% membrane probably because of its bigger pore size. In summary, the C/P-2% membrane has the best performance among the C/P blend membranes based on results of the membrane characterization and Yeongsan River water filtration.
3.3.
Analyses of the NOM foulants on C/P membranes
The desorbed foulants from bare PES and C/P-2% membranes were separated into hydrophobic (HPO), transphilic (TPI), and hydrophilic (HPI) fractions, as presented in Table 3. The total foulant of the bare PES membrane was almost 3 times higher than for the C/P-2% membrane, which is consistent with the results of the flux decline calculations. Since Yeongsan River water is highly hydrophilic (Table 2), the HPO fraction of the foulants on both membranes was the least. HPO fraction of the foulants on bare PES membrane was 3 times higher than C/P-2% membrane, because of the hydrophobic interactions between the hydrophobic PES membrane and hydrophobic NOM. Even though the zeta potential could not be determined for the C/P-2% membrane, because of the electrical conductivity of the MWCNTs, it is hypothesized that C/P-2% membrane is more negatively charged than the bare PES membrane, because of the carboxylic groups on functionalized MWCNTs in the membrane structure, which is confirmed by FTIR (Fig. 1). The HPI fraction of the foulants on the C/P-2% membrane is very similar to the bare PES membrane, because the hydrophilic neutral species and bases (Cho et al., 2000) can easily adsorb onto the more negatively charged C/P-2% membrane. The TPI fraction of NOM is mainly hydrophilic acids (Cho et al., 2000). Hence, the TPI fraction of the foulants on the C/ P-2% membrane is 7 times lower than for the bare PES membrane. In summary, the C/P-2% membrane was superior to the bare PES membrane for natural water filtration due to its higher NOM rejection, lower fouling, higher flux, flux stability, and ease of cleaning.
4.
Conclusion
The preparation and anti-fouling behavior of C/P blend membranes for water treatment were investigated, with several conclusions subsequently drawn. These conclusions include the following. (1) The C/P blend membranes were successfully prepared via a phase inversion method. Prior to membrane synthesis, MWCNTs were functionalized with a strong acid mixture to increase the dispersion in organic solvent. FTIR analysis of the C/P blend membranes showed that there might be hydrogen bonding interactions between the sulfonic groups of PES and carboxylic groups of functionalized MWCNTs. (2) Pure water flux tests and membrane characterization confirmed that the MWCNT content of the blend
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 7 4 e2 8 2
membranes increased the membrane roughness, hydrophilicity, porosity, and pure water flux. Moreover, the morphology and the separation properties of the C/P blend membranes were found to depend on the amount of MWCNTs in the blend. (3) During surface water filtration, C/P blend membranes displayed 42% less flux decline than bare PES membranes. In addition, foulants on the C/P blend membrane surface could be more easily removed by caustic cleaning than on the bare PES membrane, resulting in higher flux recoveries with C/P blend membranes. Analysis of the foulants on the membrane surface further confirmed the superior fouling resistance of the C/P blend membranes.
Acknowledgement The authors gratefully acknowledge the Alternative Water Purification Engineering Laboratory at GIST for their contributions in TOC and the HPLC-RI analyses and for lending XAD resins. Funding was provided by a National Research Laboratory Program grant funded by the Korea Ministry of Science and Technology (M1050000012805-J000012810), and partially provided by a Korea Research Foundation grant funded by the Korea Government (MOHERD) (KRF-2008-211-D00052) and partially provided by the Basic Research Project through a grant provided by the Gwangju Institute of Science and Technology.
references
Aiken, G.R., McKnight, D.M., Thorn, K.A., Thurman, E.M., 1992. Isolation of hydrophilic organic acids from water using nonionic macroporous resins. Org. Geochem. 18 (4), 567e573. Amirilargani, M., Sadrzadeh, M., Mohammadi, T., 2010. Synthesis and characterization of polyethersulfone membranes. J. Polym. Res. 17, 363e377. Arthanareeswaran, G., Thanikaivelan, P., Srinivasn, K., Mohan, D., Rajendran, M., 2004. Synthesis, characterization and thermal studies on cellulose acetate membranes with additive. Euro. Polym. J. 40, 2153e2159. Barth, C., Goncalves, M.C., Pires, A.T.N., Roeder, J., Wolf, B.A., 2000. Asymmetric polysulfone and polyethersulfone membranes: effects of thermodynamic conditions during formation on their performance. J. Membr. Sci. 169, 287e299. Blanco, J.F., Sublet, J., Nguyen, Q.T., Schaetzel, P., 2006. Formation and morphology studies of different polysulfones-based membranes made by wet phase inversion process. J. Membr. Sci. 283, 27e37. Boehm, H.P., 1994. Some aspects of the surface chemistry of carbon blacks and other carbons. Carbon 32 (5), 759e769. Brunet, L., Lyon, D.Y., Zodrow, K., Rouch, J.-C., Caussat, B., Serp, P., Remigy, J.-C., Wiesner, M.R., Alvarez, P.J.J., 2008. Properties of membranes containing semi-dispersed carbon nanotubes. Environ. Eng. Sci. 25 (4), 565e575. Cheryan, M., 1986. Ultrafiltration Handbook. Technomic Publishing Company, Inc., USA. Cho, J., Amy, G., Pellegrino, J., 2000. Membrane filtration of natural organic matter: comparison of flux decline, NOM rejection,
281
and foulants during filtration with three UF membranes. Desalination 127, 283e298. Choi, H., Zhang, K., Dionysiou, D.D., Oerther, D.B., Sorial, G.A., 2005. Effect of permeate flux and tangential flow on membrane fouling for wastewater treatment. Sep. Purif. Technol. 45, 68e78. 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. Drelich, J., Miller, J.D., Good, R.J., 1996. The effect of drop (bubble) size on advancing and receding contact angles for heterogeneous and rough solid surfaces as observed with sessile-drop and captive-bubble techniques. J. Colloid Interf. Sci. 179, 37e50. 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. Goertzen, S.L., Theriault, K.D., Oickle, A.M., Tarasuk, A.C., Andreas, H.A., 2010. Standardization of the Boehm titration. Part I. CO2 expulsion and endpoint determination. Carbon 48, 1252e1261. 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. Gray, S.R., Ritchie, C.B., Tran, T., Bolto, B.A., Greenwood, P., Busetti, F., Allpike, B., 2008. Effect of membrane character and solution chemistry on microfiltration performance. Water Res. 42, 743e753. 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. Kaniyoor, A., Jafri, R.I., Arockiadoss, T., Ramaprabhu, S., 2009. Nanostructured pt decorated graphene and multi walled carbon nanotube based room temperature hydrogen gas sensor. Nanoscale 1, 382e386. Kim, J., Bruggen, B.V., 2010. The use of nanoparticles in polymeric and ceramic membrane structures: review of manufacturing procedures and performance improvement of water treatment. Kim, U.J., Furtado, C.A., Liu, X., Chen, G., Eklund, P.C., 2005. Raman and IR spectroscopy of chemically processed singlewalled carbon nanotubes. J. Am. Chem. Soc. 127, 15437e15445. 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. Mulder, M., 1997. Basic Principles of Membrane Technology, second ed. Kluwer Academic Publishers, The Netherlands. Peng, F., Hu, C., Jiang, Z., 2007. Novel ploy(vinyl alcohol)/carbon nanotube hybrid membranes for pervaporation separation of benzene/cyclohexane mixtures. J. Membr. Sci. 297, 236e242. 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. Rong, C., Ma, G., Zhang, S., Song, L., Chen, Z., Wang, G., Ajayan, P.M., 2010. Effect of carbon nanotubes on the mechanical properties and crystallization behavior of poly(ether ether ketone). Compos. Sci. Technol. 70, 380e386. Shi, Q., Su, Y., Zhu, S., Li, C., Zhao, Y., Jiang, Z., 2007. A facile method for synthesis of pegylated polyethersulfone and its
282
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 7 4 e2 8 2
application in fabrication of antifouling ultrafiltration membrane. J. Membr. Sci. 303, 204e212. Sun, M., Su, Y., Mu, C., Jiang, Z., 2010. Improved antifouling property of PES ultrafiltration membranes using additive of silica-PVP nanocomposite. Ind. Eng. Chem. Res. 49, 790e796. Wang, Y.Q., Su, Y.L., Sun, Q., Ma, X.L., Jiang, Z.Y., 2006. Generation of anti-biofouling ultrafiltration membrane surface by blending novel branched amphiphilic polymers with polyethersulfone. J. Membr. Sci. 286, 228e236. Yang, Y., Zhang, H., Wang, P., Zheng, Q., Li, J., 2007. The influence of nano-sized TiO2 fillers on the morphologies and properties of PSF UF membrane. J. Membr. Sci. 288, 231e238.
Yuen, S.-M., Ma, C.-C.M., Chiang, C.-L., Teng, C.-C., 2008. Morphology and properties of aminosilane grafted MWCNT/ polyimide nanocomposites. J. Nanomater., 15. Article ID 786405. Zondervan, E., Roffel, B., 2007. Evaluation of different cleaning agents used for cleaning ultrafiltration membranes fouled by surface water. J. Membr. Sci. 304, 40e49. Zularisam, A.W., Ismail, A.F., Salim, M.R., Sakinah, M., Hitoaki, O., 2007. Fabrication, fouling and foulant analyses of asymmetric polysulfone (PSF) ultrafiltration membrane fouled with natural organic matter (NOM) source waters. J. Membr. Sci. 299, 97e113.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 8 3 e2 9 1
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Further treatment of decolorization liquid of azo dye coupled with increased power production using microbial fuel cell equipped with an aerobic biocathode Jian Sun a,1, Zhe Bi a,2, Bin Hou a,3, Yun-qing Cao a,4, Yong-you Hu a,b,* a
Ministry of Education Key Laboratory of Pollution Control and Ecological Remediation for Industrial Agglomeration area, Department of Environmental Science and Engineering, South China University of Technology, Guangzhou, 510006, China b The Guangxi Key Laboratory of Environmental Engineering, Protection and Assessment, Guilin 541004, China
article info
abstract
Article history:
A microbial fuel cell (MFC) incorporating a recently developed aerobic biocathode is designed
Received 9 February 2010
and demonstrated. The aerobic biocathode MFC is able to further treat the liquid containing
Received in revised form
decolorization products of active brilliant red X-3B (ABRX3), a respective azo dye, and also
12 July 2010
provides increased power production. Batch test results showed that 24.8% of COD was
Accepted 19 July 2010
removed from the decolorization liquid of ABRX3 (DL) by the biocathode within 12 h.
Available online 3 August 2010
Metabolism-dependent biodegradation of aniline-like compound might be mainly responsible for the decrease of overall COD. Glucose is not necessary in this process and contributes
Keywords:
little to the COD removal of the DL. The similar COD removal rate observed under closed
Microbial fuel cell
circuit condition (500 U) and opened circuit condition indicated that the current had an
Aerobic biocathode
insignificant effect on the degradation of the DL. Addition of the DL to the biocathode resulted
Active brilliant red X-3B
in an almost 150% increase in open cycle potential (OCP) of the cathode accompanied by a 73%
Decolorization liquid
increase in stable voltage output from 0.33 V to 0.57 V and a 300% increase in maximum power
Wastewater treatment
density from 50.74 mW/m2 to 213.93 mW/m2. Cyclic voltammetry indicated that the decol-
Power production
orization products of the ABRX3 contained in the DL play a role as redox mediator for facilitating electron transfer from the cathode to the oxygen. This study demonstrated for the first time that MFC equipped with an aerobic biocathode can be successfully applied to further treatment of effluent from an anaerobic system used to decolorize azo dye, providing both cost savings and high power output. ª 2010 Elsevier Ltd. All rights reserved.
* Corresponding author. Ministry of Education Key Laboratory of Pollution Control and Ecological Remediation for Industrial Agglomeration area, Department of Environmental Science and Engineering, South China University of Technology, Guangzhou, 510006, China. Tel.: þ86 20 39380506; fax: þ86 20 39380508. E-mail addresses:
[email protected] (J. Sun),
[email protected] (Z. Bi),
[email protected] (B. Hou),
[email protected] (Y.-q. Cao),
[email protected] (Y.-y. Hu). 1 Tel.: þ86 13480273800.
[email protected]. 2 Tel.: þ86 15210361625. 3 Tel.: þ86 15914305840. 4 Tel.: þ86 13983696487. 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.07.059
284
1.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 8 3 e2 9 1
Introduction
Microbial fuel cells (MFCs) hold great potential for harvest of electricity directly from various biodegradable materials, ranging from pure compounds to complex substrates present in wastewater (Rabaey et al., 2003; Liu et al., 2004, 2005; Min et al., 2005; Catal et al., 2008). Before it can be used in practical applications, substantial efforts are required to optimize MFC performance, particularly to address the problems of low power output and high cost. Many factors contribute to the overall MFC performance and cost. Among these factors, both kinetic limitations for oxygen reduction at the surface of the cathode and the use of noble metals, such as platinum as catalysts, present a great challenge. Typically, MFCs utilize an abiotic cathode, with oxygen used as an electron acceptor due to its unlimited availability and high standard redox potential. Chemicals other than oxygen are also used, such as ferricyanide (Rabaey et al., 2005; Oh et al., 2004) and permanganate (You et al., 2006), resulting in greater overall potentials. However, the use of these chemicals is obviously unsustainable due to the need for continuous replacement after consumption. Moreover, these chemicals are toxic to anodic microorganisms and potentially pose a high risk of environmental pollution, which directly conflicts with the environmentally friendly nature of MFCs. To accelerate the oxygen reduction rate on the surface of the cathode, platinum is commonly chosen because of its excellent catalytic ability (Oh et al., 2004). However excessive cost limits this application. Certain less expensive metals, such as lead dioxide, manganese dioxide molybdenum/vanadium, cobalt and iron-based materials, have been successfully tested to improve the oxygen reduction efficiency on cathodic electrode (Habermann and Pommer, 1991; Zhao et al., 2005, 2006; Cheng et al., 2006; Morris et al., 2007; Rismani-Yazdi et al., 2008). Metal-based catalysts, however, are generally susceptible to the adverse environmental conditions that may occur in MFCs and cause inactivation (Zhang et al., 2009). A recently developed biocathode that uses microorganisms as catalysts to assist in electron transfer eliminates the use of noble metal, such as platinum, and eliminates the need for replenishment of the electron mediator, resulting in greatly improved MFC sustainability (He and Angenent, 2006). Aerobic and anaerobic biocathodes are classified depending on the terminal electron acceptors adopted in the cathode. In an aerobic biocathode, oxygen is used as the terminal electron acceptors. Microorganisms can transfer electrons directly (Bergel et al., 2005) or indirectly by catalyzing the re-oxidation of redox couples, such as Mn2þ/Mn4þ (Rhoads et al., 2005) and Fe2þ/Fe 3þ (Ter Heijne et al., 2007), from the cathode to oxygen, thus driving oxygen reduction. In the absence of oxygen, other compounds, such as nitrate, sulfate, iron, manganese, selenate, arsenate, urinate, fumarate, and carbon dioxide can function as terminal electron acceptors (Stams et al., 2006). Apart from the sustainability consideration, the biocathode also indicates a potential approach for wastewater treatment due to its variety of terminal electron acceptors. Recently, the biocathode-MFCs have been explored for the removal of nitrogen from water (Clauwaert et al., 2007; Virdas et al., 2008; You et al., 2009).
Azo compounds constitute the largest group of synthetic dyes and are in widespread use in the dye-manufacturing and dye-consuming industries (Stolz, 2001; Pandey et al., 2007). The release of azo dye-containing wastewater represents a serious environmental problem and a public health concern. Thus, it is crucial to remove azo dye completely from effluent before it can enter into the environment. The most logical method removal of azo dyes in biological wastewater treatment systems is based on anaerobic treatment for the reductive cleavage of the azo linkages in the dye in combination with aerobic treatment for further degradation of the products from azo dye cleavage, which are aromatic amines (Pandey et al., 2007; Dos Santos et al., 2007; van der Zee and Villaverde, 2005). As a novel method, MFC technology may be explored to treat azo dye-containing wastewater with a high efficiency and instantaneous cost savings. Previous research in our laboratory has demonstrated that accelerated decolorization of active brilliant red X-3B (ABRX3, a representative azo dye used in textile industry) accompanied by electricity generation from readily biodegradable organic substrates and confectionery wastewater can be achieved using a microfiltration membrane air-cathode single chambered MFC with improved performance (Sun et al., 2009a,b). The decolorization intermediates, however, accumulate at high dye concentrations, and they are resistant to further degradation in the anoxic anodic environment of the MFC. The novel aerobic biocathode was considered in our consecutive study for further degradation of these intermediates, because these intermediates are prone to be biodegradation under aerobic conditions. To date, some research efforts have been reported regarding use of the MFC anode for the removal of recalcitrant contaminants from wastewater (Morris et al., 2009; Luo et al., 2009), but no reports of use of the biocathode have been published. In this paper, we make a first attempt to further treat the decolorization liquid of ABRX3 (DL) while simultaneously obtaining additional power output using a two chambered MFC equipped with an aerobic biocathode. Degradation mechanisms of the DL in the biocathode and the effect of the degradation process on electricity generation are investigated in detail.
2.
Materials and methods
2.1.
Dye
The ABRX3 dye purchased from the Dyeing Chemical Industry in Guangzhou, China was of commercial purity and used without further purification.
2.2.
Aerobic biocathode MFC construction
The MFC was constructed of polycarbonate and contained an anode and a cathode chamber (each approximately 900 ml in volume except for a 100 ml headspace). The chambers were separated by a proton exchange membrane (PEM, Nafion 117, Dupont) with a surface area of 28.26 cm2 (6 cm in diameter) and sealed with a rubber gasket. The PEM was sequentially boiled in H2O2 (30%), deionized water, 0.5 M H2SO4, and deionized water
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 8 3 e2 9 1
before use (each for 1 h). Several ports on the top and one port on the side were installed for each chamber, which were convenient for purging, sampling, draining and introducing electrodes. All of the ports on the anode chamber were sealed with thick rubber stoppers during MFC operation. Porous carbon papers (without waterproofing) with a projected surface area of 6 6 cm2 on each side were used as electrodes. No catalyst was applied on the surface of the cathode. The anode was set parallel to the cathode at a distance of 15 cm and connected by copper wires. All exposed metal surfaces were sealed with a nonconductive epoxy resin. The schematic details of the MFC are shown in Fig. 1.
2.3.
Inoculation and medium
Different types of aerobic and anaerobic sludges and sediment were used as the original anode and cathode inocula in order to obtain sufficient microbial diversity. The aerobic and anaerobic sludges were collected from the Liede domestic wastewater treatment plant, and sediment was collected from a wetland near the South China University of Technology, Guangzhou, China. Before inoculation, the sludge and sediment were washed three times using deionized water to remove soluble carbon sources and then filtered through a 0.25-mm pore size sieve to remove impurities. All sludge and sediment were mixed and added to the anode chamber and cathode chamber to a final concentration of 2 g volatile suspended solids per reactor volume (VSS/L). For the anode chamber, the nutrient medium (containing 2.98 g/L NaH2PO4$2H2O, 9.28 g/L Na2HPO4$12H2O, 0.31 g/L NH4Cl, 0.13 g/L KCl), a mineral solution (12.5 mL), and a vitamin solution (12.5 mL) as reported by Lovley and Phillips (Lovley and Phillips, 1988), was supplemented with 500 mg chemical oxygen demand (COD) of glucose per liter. The biocathode chamber was filled with the same nutrient medium added to the anode during the startup stage, with an additional supply of 5 ml MneFe stock solution (10 mM FeSO4 and 50 mM manganese). Glucose was added or not added according to the experimental design.
2.4.
MFC operation
The aerobic biocathode MFC and an abiotic cathode MFC (abiotic control) were used in this study. Two MFCs were
285
operated in batch-fed mode at a fixed load of 500 U except for power density measurement. Sometimes, the aerobic biocathode MFC was disconnected from the external load to investigate the effect of electric current on the degradation of the DL (open circuit control). For maintenance of anaerobic conditions, the anode chamber was flushed with nitrogen gas for 15 min to remove dissolved oxygen before each batch test, while the cathode was continuously flushed with air at a flow rate of approximately 120 mL/min. Each chamber was mixed using a small magnetic stirrer to enhance mass transfer. The voltage difference across the anode and cathode versus time was monitored. Liquid for the anode and cathode were refreshed in the aerobic biocathode MFC when the voltage decreased below 10 mV and the suspended biomass was reserved. Once stable voltage was observed, the effluent from a high-performance air-cathode single chambered MFC used for decolorization of ABRX3 (1500 mg/L, over 90% of the color removal), was subsequently fed to the biocathode chamber to a final organic load of 210e220 mg COD/L. The effluent consists of anaerobic decolorization products of theABRX3andspent anode medium (voltage lower than 10 mV and glucose was exhausted). The anaerobic decolorization products of the ABRX3 could include the aniline, the phenol-like compound with a naphthalene ring and the compound with a triazine ring due to the cleavage of the azo bond (Lu et al., 2009). The MFC continue to run several batch cycles until the system fully acclimated to the DL and demonstrated repeatable cycles of power generation again. The COD and dissolved oxygen (DO) in the cathode was measured regularly. All experiments were conducted at least in duplicate, in a constant-temperature room (30 1 C), and the average value was reported for all data.
2.5.
Analytics and calculations
The voltage difference between the anode and cathode was recorded every 2 min using a precision multimeter and a data acquisition system (Model 2700, Keithly Instruments, USA). Power density (mW/m2) was calculated according to P ¼ IV=A where I is the current, V is the voltage, and A is the projected cross-sectional area of the anode. Power density curves were used to obtain the maximum power density by varying the external resistance from 5000 U to
Fig. 1 e Schematic of the aerobic biocathode microbial fuel cell.
286
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 8 3 e2 9 1
50 U using a resistor box. Hourly averaged voltages were used to calculate the power density. Two Ag/AgCl reference electrodes (0.195 vs. normal hydrogen electrode; NHE) were inserted in the anode chamber and biocathode chamber, respectively, to measure individual electrode potential. COD was measured according to standard methods (APHA, 1998). All samples were filtered through a 0.22 mm-pore-size syringe filter unit prior to COD measurements. Dissolved oxygen concentration was measured with YSI Model 550A meter (Yellow Springs Instruments, Yellow Springs, USA). The biocathode liquid was analyzed by monitoring the absorbance using a UVevisible scanning spectrophotometer (BECKMANDV640). Test samples (5 mL) were withdrawn from the biocathode chamber at 0, 24, and 48 h and were centrifuged at 10 000 rpm for 5 min to remove suspended biomass from the liquid media. Samples were adequately diluted prior to measurement of absorbance. Cyclic voltammetry (CV) was employed to detect the possible mediator in the DL. CVs were performed under conventional three-electrode mode using an electrochemical workstation (Model 2273, Princeton Applied Research). The working electrode was a new carbon paper and the counter electrode was a platinum electrode. Saturated Ag/AgCl electrode was used as reference electrode. The voltage was changed from 0.8 V to 0.2 V in forward and reverse scans at a scan rate of 25 mV/s for a total of five scans (Bard and Faulkner, 2001). The bacterial morphologies on the surface of the biocathode were determined using an environmental scanning electron microscope (ESEM) (XL-30, Philips, Holland). Before observation, a sample was collected and fixed overnight with paraformaldehyde and glutaraldehyde in a buffer solution (0.1 M cacodylate, pH ¼ 7.5, 4 C), followed by washing and dehydration in water/ethanol. Samples were then coated with Au/Pt before ESEM observation.
3.
Results and discussion
3.1. MFC
Startup and performance of the aerobic biocathode
After 38 days, a repeatable and stable voltage output of approximately 0.35 V (Rex ¼ 500 U) was demonstrated for the aerobic biocathode MFC with additions of 500 mg COD/L glucose (Fig. 2A). The maximum power density increased from 7.68 mW/m2 (16th day) to 50.74 mW/m2, corresponding to a decrease of internal resistance from 1514 U to 987 U (Fig. 2B). These results show that the aerobic biocathode MFC operates smoothly and performs well for electricity generation.
3.2.
Degradation of DL in the aerobic biocathode
To test the ability of the aerobic biocathode to further degrade the decolorization products of ABRX3, the DL from a high-performance air-cathode single chamber MFC for decolorization of ABRX3 (1500 mg/L for dye concentration, over 90% of the color removal) was added to the aerobic biocathode chamber to a final organic load of 210e220 mg
Fig. 2 e Voltage (500 U) (A) and power (B) generation by the aerobic biocathode MFC during startup phase using glucose (500 mg COD/L) as anodic electron donor. (Arrows show the replacement of anodic media solution).
COD/L. As shown by Fig. 3, without the presence of any other carbon resources, the total COD in the liquid of aerobic biocathode was decreased by 24.8% (from 210.3 mg/L to 158.1 mg/L) within 12 h, while the MFC with the abiotic cathode did not show a noticeable COD decrease. Therefore, the decrease of COD was solely dependent upon the microbial metabolism in the aerobic biocathode. Physical adsorption by the biomass could also partially contribute to the COD removal, but this is less likely because the biomass could have reached saturation due to long-term acclimation process using the DL (more than five months). Fig. 3 also demonstrated that the electric current had an insignificant effect on the COD removal of the DL. The COD removal rate under close circuit condition (500 U) was similar to that under open circuit condition. This indicated that the degradation of the DL was not a current-dependent process. In the aerobic cathode, degradation of the decolorization products of the azo dye would consume a portion of DO which was simultaneously served as the terminal electron acceptor for the cathode. Thus, there seems to exist a competition mechanism between the DL and the cathode for the DO. Being operated under open circuit conditions means that there are no electrons transferred to the cathode. The degradation of the DL should be accelerated due to increased supply of the DO. However, similar
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 8 3 e2 9 1
287
Fig. 3 e COD removal of DL by the aerobic biocathode. The aerobic biocathode MFC operated under open circuit condition and an abiotic cathode MFC were used as controls.
degradation rate was observed. The most possible reason might be that the DO supply (about 2.0e2.5 mg/L) in this study is sufficient for the partial degradation of the DL during electricity generation. Although increase of current by incorporating a lower resistor into the electric circuit will result in increased consumption of the DO in the biocathode, the degradation of the DL may still not be significantly affected because the aromatic compounds contained in the DL are even prone to be degraded under anoxic condition (Pandey et al., 2007). Aniline, a complicated phenol-like compound with a naphthalene ring and a compound with a triazine ring are formed after microbial anaerobic reduction of ABRX3 due to the break down of the azo bond. UVevisible spectrophotometry revealed the further degradation of the decolorization products from ABRX3 in the aerobic biocathode (Fig. 4A). In the visible band region, the absence of a peak at 538 nm (lmax for ABRX3) showed that the azo bond of ABRX3 was cleaved completely. The absorbance in the UV region at lmax of 248 nm, which is most likely attributed to the benzene ring, was observed as continuously decreasing (Lu et al., 2009). After 48 h, the extent of degradation of the product at lmax of 248 nm had reached 81.56% according to the change of absorbance value. Compared with the aerobic biocathode, the abiotic cathode did not result in any noticeable decrease in the absorbance at lmax of 248 nm (Fig. 4B), which was consistent with the obtained results for COD removal. As reported in literatures, decolorization product of azo dye will tend to autoxidize, forming colored products when exposed to air (Kudlich et al., 1999; Kalyuzhnyi and Sklyar, 2000; Libra et al., 2004). The observed further color removal of the liquid in the aerobic biocathode suggests that these colored products can partially be removed from the water phase. However, the low removal efficiency of total COD indicated that other breakdown products from ABRX3 were still resistant to further degradation in the aerobic biocathode. It was noted that the additional supply of glucose did not result in the further decrease of total COD of the DL (Fig. 3). Two possible reasons were responsible for this. One is that the microbial consortia, which can utilize the aniline-like compound as a sole carbon source, have developed in the
Fig. 4 e UVevisible absorption spectra for the liquid in the aerobic biocathode (A) and abiotic cathode (B).
aerobic biocathode. The other reason might be that residual glucose contained in the DL can serve as a co-substrate and maybe enough for degradation of aniline in the aerobic biocathode. The first reason is the most likely because glucose was consumed almost completely at the end of each batch decolorization cycle (Sun et al., 2009a).
3.3.
Effect of DL on electricity production
The DL was subsequently fed to the biocathode chamber when the MFC was well developed. Strikingly, this enabled the generation of substantially larger electricity outputs. An almost 73% increase in stable voltage output from 0.33 V to 0.57 V (Fig. 5A) and 300% increase in maximum power density from 50.74 mW/m2 to 213.93 mW/m2, corresponding to a 37% decrease in internal resistance from 987 U to 617 U (Fig. 5B) were observed, respectively. To further examine the effects of DL on electricity generation, the electrode working potentials were measured as a function of current by varying the external resistor from 50 U to 5000 U. Fig. 5C shows that addition of the DL to the biocathode resulted in an almost 1.5 bold increase in the open cycle potential (OCP) of the cathode, while the anode potential was not affected as a whole. Some redox mediators are dye in nature such as neutral red, thionin and methylene blue, which are
288
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 8 3 e2 9 1
Fig. 6 e Cyclic voltammograms at 36 cmL2 carbon paper using DL (line a), spent anode solution obtained from an MFC never used for decolorization of ABRX3 (line b) and fresh anode medium (line c) at 25 mv/s.
Fig. 5 e Effect of DL on voltage generation (A), power density (B) and electrode potential (C) of the aerobicbiocathode MFC (Arrow show the addition of DL).
usually added to MFCs to enhance electron transport from cell to the electrode surface (Park and Zeikus, 2000; Choi et al., 2003; Sund et al., 2007). Even the constituent amines of azo dye can also be used as redox mediators for increasing the decolorization rate of other azo dyes (Me´ndez-Paz et al., 2005). In this study, such a sharp increase of the OCP in the biocathode was most likely due to the role of some byproducts from ABRX3 due to the enhanced electron transfer from the cathode to oxygen possibly via mediation of the transfer of these equivalents. To detect possible mediator in the DL, CV tests were performed using three samples with a clean carbon paper electrode: (1) DL; (2) spent anode medium obtained from an MFC never used for decolorization of ABRX3; (3) fresh anode
medium. The results were presented in Fig. 6. The DL showed a couple of distinguishable oxidationereduction peaks (line a): the oxidation peak in the forward scan at 0.25 V (vs. Ag/AgCl) and the reduction peak in the reverse scan at 0.35 V, this could be the evidence of the mediator contained in the DL. In contrast, there was no peak was found using the spent anode medium obtained from an MFC never used for decolorization of ABRX3 (line b) and the fresh anode medium(line c), excluding the possibility that the mediator was the soluble redox-mediator excreted by the anode microorganism or the components of the anode medium. Thus, it is plausible that the mediator contained in the DL was one of the decolorization products of the ABRX3. Redox-active metabolites secreted by anaerobic communities in the anode chamber (Rabaey et al., 2005; Niessen et al., 2004; Holmes et al., 2004) might partly contribute to the increase of the OCP in the biocathode. However, these mediators were not detected by the CV. This could result from the frequent replacement of the liquid, which is a poor stratagem for the accumulation of a high concentration of mediators (Rabaey et al., 2005). These results demonstrate the contribution of DL to the improvement performance of the biocathode, as consistent with the former hypothesis, some byproducts from the ABRX3 serve as redox mediator in the biocathode to enhance the electron transfer from the cathode to the oxygen. A schematic mechanism for the DL assisted electrochemical reduction of oxygen in the aerobic biocathode was proposed in Fig. 7. These results also indicate that the electrons transfer from the electrode to oxygen in the aerobic biocathode is still a major limiting factor for electricity generation, and the overall performance of the MFC can be improved through cathode optimization.
3.4.
Biofilm growth on the anode and biocathode
ESEM images revealed that a thick, homogeneous biofilm was formed on both the surface of the anode and the cathode
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 8 3 e2 9 1
3.5.
Fig. 7 e Schematic mechanism for the DL assisted electrochemical reduction of oxygen in the aerobic biocathode.
(Fig. 8A and C). The anode surface was covered by individual bacilliform or aggregated bacteria (Fig. 8B), while bacteria on the surface of cathode were clustered in aggregates (Fig. 8D). It was noticeable that the bacteria on the cathode surface produced a large number of nanowires-like long thin filaments that connected to different bacteria aggregates. These filaments could be conductive and may achieve the electron transfer while not directly in contact with the electrode surface (Fig. 8D).
289
Implication
As preliminary research, we have demonstrated the feasibility of using an aerobic biocathode for simultaneous treatment of azo dye decolorization liquid and electricity generation. Further efforts are required to improve the overall performance of this system. First, the microbial diversity and bacterial metabolite-mediated electron transfer mechanisms in the biocathode must be fully understood. The removal of aromatic compounds and stable operation of the MFC indicated the presence of two preponderant groups of bacteria in the aerobic biocathode. One group can utilize the aromatic amines as an electron donor and oxygen as an acceptor. The other group of bacteria can mediate electron transfer from the cathode electrode to oxygen directly through catalysis of the re-oxidation of redox couples (transition metals). It is essential to obtain detailed information about how these interactions work with each other in the mixed culture system. Second, the effect of operation parameters (such as cathode potential, dissolved oxygen and external resistance) on the overall removal efficiency of COD and electricity production need to be systematically investigated. Third, considering the low COD removal efficiency, there is a requirement for acclimating and developing microbial consortia that can rapidly degrade of mixtures of aromatic amines broken down from azo dye. The fate of these aromatic amines must also be addressed. Fourth, the relationship between concentration of aromatic
Fig. 8 e ESEM image of biofilm developed on the surface of anode (A, B) and biocathode (C, D).
290
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 8 3 e2 9 1
amines and the cathode potential will be further investigated and explored for biosensor’s application. Electricity generation may be negatively affected by the degradation of recalcitrant contaminants during use of the MFC anode due to toxicity or potential electron competition with electricity-generating bacteria (Sun et al., 2009a). Recently, You et al demonstrated the enhancement of cathode performance based on the biological nitrification in the aerobic biocathode You et al. (2009). This study also showed that the use of the biocathode for simultaneous removal of recalcitrant contaminants and generation of increased power output can be an attractive option. Oxygen reacts with the aromatic products, reduction products of azo dye, resulting in the formation of undesirable colored oligomers and polymers that may be toxic and mutagenic (Field et al., 1995). Though the autoxidation process eliminates the aromatic amines, the products formed are more recalcitrant to biological degradation. In this study, rapid removal of these colored products indicated that the rate of degradation was much higher than the rate of autoxidation. Partial mineralization coupled with fast further color removal for the autoxidation products in the DL suggests the feasibility of using aerobic biocathode for further treating decolorization liquid of azo dye. The advantage of the MFC described here over a traditional aerobic wastewater system is that it requires little auxiliary power and maintenance. Furthermore, the increased power generated by the MFC will partly offset the energy consumed for aeration in the biocathode and can also potentially be used to power other electrical devices by binding several MFCs for higher voltage output. In combination with our previous study, the sequential process of decolorization of azo dye using an anaerobic anode and further treatment of decolorization liquid using an aerobic biocathode suggests a feasible and novel method for complete removal of azo dyes from azo dye-containing wastewater.
4.
Conclusions
Further treatment of the liquid contain the decolorization products of azo dye coupled with increased power production was demonstrated for the first time using an MFC equipped with an aerobic biocathode. Overall, 24.8% of COD was removed from the DL by the biocathode within 12 h. Metabolism-dependent biodegradation of aniline-like compound might be mainly responsible for the decrease of overall COD. Glucose is not necessary, and it contributes little to the COD removal of the DL. The degradation of the DL was not significantly affected by the electricity generation process. The addition of the DL resulted in a sharp elevation in the OCP of the biocathode accompanied by significant increase in stable voltage output and maximum power density. Decolorization products of the ABRX3 contained in the DL can serve as redox mediator for facilitating electrons transfer from the cathode to the oxygen. Further efforts are required for overall performance improvements in the system.
Acknowledgments This research was supported jointly by the National Natural Science Fund of China (No. 20977032) and the research funds of the Guangxi Key Laboratory of Environmental Engineering, Protection and Assessment (No. 0804Z021).
Appendix. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2010.07.059.
references
APHA, AWWA, WPCF, 1998. Standard Methods for Examination of Water and Wastewater, 20th ed. American Public Health Association, Washington, DC. Bard, A.J., Faulkner, L.R., 2001. Electrochemical Methods: Fundamentals and Applications, second ed. John Wiley, New York. Bergel, A., Fe´ron, D., Mollica, A., 2005. Catalysis of oxygen reduction in PEM fuel cell by seawater biofilm. Electrochemistry Communications 7 (9), 900e904. Catal, T., Xu, S., Li, K., Bermek, H., Liu, H., 2008. Electricity generation from polyalcohols in single-chamber microbial fuel cells. Biosensors and Bioelectronics 24 (4), 849e854. Cheng, S., Liu, H., Logan, B.E., 2006. Power densities using different cathode catalysts (Pt and CoTMPP) and polymer binders (Nafion and PTFE) in single chamber microbial fuel cells. Environmental Science and Technology 40 (1), 364e369. Choi, Y., Kim, N., Kim, S., Jung, S., 2003. Dynamic behaviors of redox mediators within the hydrophobic layers as an important factor for effective microbial fuel cell operation. Bulletin of the Korean Chemical Society 24, 437e440. Clauwaert, P., Rabaey, K., Aelterman, P., De Schamphelaire, L., Pham, T.H., Boeckx, P., Boon, N., Verstraete, W., 2007. Biological denitrification in microbial fuel cells. Environmental Science and Technology 41 (9), 3354e3360. 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 (12), 2369e2385. Field, J.A., Stams, A.J.M., Kato, M., Schraa, G., 1995. Enhanced biodegradation of aromatic pollutant in coculture of anaerobic and aerobic bacterial consortia. Antonie Van Leeuwenhoek 67, 47e77. Habermann, W., Pommer, E.H., 1991. Biological fuel cells with sulphide storage capacity. Applied Microbiology and Biotechnology 35 (1), 128e133. He, Z., Angenent, L.T., 2006. Application of bacterial biocathodes in microbial fuel cells. Electroanalysis 18 (19e20), 2009e2015. Holmes, D.E., Bond, D.R., O’Neill, R.A., Reimers, C.E., Tender, L.R., Lovley, D.R., 2004. Microbial communities associated with electrodes harvesting electricity from a variety of aquatic sediments. Microbial Ecology 48 (2), 178e190. Kalyuzhnyi, S., Sklyar, V., 2000. Biomineralisation of azo dyes and their breakdown products in anaerobiceaerobic hybrid and UASB reactors. Water Science and Technology 41 (12), 23e30. Kudlich, M., Hetheridge, M.J., Knackmuss, H.J., Stolz, A., 1999. Autoxidation reactions of different aromatic o-aminohydroxynaphthalenes that are formed during the anaerobic reduction of sulfonated azo dyes. Environmental Science and Technology 33 (6), 869e901.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 8 3 e2 9 1
Libra, J.A., Borchert, M., Vigelahn, L., Storm, T., 2004. Two stage biological treatment of a diazo reactive textile dye and the fate of the dye metabolites. Chemosphere 56 (2), 167e180. Liu, H., Cheng, S.A., Logan, B.E., 2005. Production of electricity from acetate or butyrate using a single chamber microbial fuel cell. Environmental Science and Technology 39 (2), 658e662. Liu, H., Ramnarayanan, R., Logan, B.E., 2004. Production of electricity during wastewater treatment using a single chamber microbial fuel cell. Environmental Science and Technology 38 (7), 2281e2285. Lovley, D.R., Phillips, E.J.P., 1988. Novel mode of microbial energy metabolism: organism carbon oxidation coupled to dissimilatory reduction of iron and manganese. Applied and Environmental Microbiology 54 (6), 1472e1480. Luo, H.P., Liu, G.L., Zhang, R.D., Jin, S., 2009. Phenol degradation in microbial fuel cells. Chemical Engineering Journal 147 (2), 259e264. Lu, X.J., Yang, B., Chen, J.H., Sun, R., 2009. Treatment of wastewater containing azo dye reactive brilliant red X-3B using sequential ozonation and upflow biological aerated filter process. Journal of Hazardous Materials 161 (1), 241e245. Me´ndez-Paz, D., Omil, F., Lema, J.M., 2005. Anaerobic treatment of azo dye acid orange 7 under fed-batch and continuous conditions. Water Research 39 (5), 771e778. Min, B., Kim, J.R., Oh, S.E., Regan, J.M., Logan, B.E., 2005. Electricity generation from swine wastewater using microbial fuel cells. Water Research 39 (20), 4961e4968. Morris, J.M., Jin, S., Crimi, B., Pruden, A., 2009. Microbial fuel cell in enhancing anaerobic biodegradation of diesel. Chemical Engineering Journal 146 (2), 161e167. Morris, J.M., Jin, S., Wang, J.Q., Zhu, C.Z., Urynowicz, M.A., 2007. Lead dioxide as an alternative catalyst to platinum in microbial fuel cells. Electrochemistry Communication 9 (7), 1730e1734. Niessen, J., Schroder, U., Scholz, F., 2004. Exploiting complex carbohydrates for microbial electricity generation e a bacterial fuel cell operating on starch. Electrochemistry Communication 6 (9), 955e958. Oh, S., Min, B., Logan, B.E., 2004. Cathode performance as a factor in electricity generation in microbial fuel cells. Environmental Science and Technology 38 (18), 4900e4904. Pandey, A., Singh, P., Iyengar, L., 2007. Bacterial decolorization and degradation of azo dyes. International Biodeterioration and Biodegradation 59 (2), 73e84. Park, D.H., Zeikus, J.G., 2000. Electricity generation in microbial fuel cells using neutral red as an electronophore. Applied and Environmental Microbiology 66 (4), 1292e1297. Rabaey, K., Boon, N., Hofte, M., Verstraete, W., 2005. Microbial phenazine production enhances electron transfer in biofuel cells. Environmental Science and Technology 39 (9), 3401e3408. Rabaey, K., Lissens, G., Siciliano, S.D., Verstraete, W., 2003. A microbial fuel cell capable of converting glucose to electricity at high rate and efficiency. Biotechnology Letters 25 (18), 1531e1535. Rhoads, A., Beyenal, H., Lewandowski, Z., 2005. Microbial fuel cell using anaerobic respiration as an anodic reaction and
291
biomineralized manganese as a cathodic reactant. Environmental Science and Technology 39 (12), 4666e4671. Rismani-Yazdi, H., Carver, S.M., Christy, A.D., Tuovinen, I.H., 2008. Cathodic limitations in microbial fuel cells: an overview. Journal of Power Sources 180 (2), 683e694. Stams, A.J.M., de Bok, F.A.M., Plugge, C.M., van Eekert, M.H.A., Dolfing, J., Schraa, G., 2006. Exocellular electron transfer in anaerobic microbial communities. Environmental Microbiology 8 (3), 371e382. Stolz, A., 2001. Basic and applied aspects in the microbial degradation of azo dyes. Applied Microbiology and Biotechnology 56 (1e2), 69e80. Sund, C.J., McMasters, S., Crittenden, S.R., Harrell, L.E., Sumner, J.J., 2007. Effect of electron mediators on current generation and fermentation in a microbial fuel cell. Applied Microbiology and Biotechnology 76 (3), 561e568. Sun, J., Hu, Y.Y., Be, Z., Cao, Y.Q., 2009a. Simultaneous decolorization of azo dye and bioelectricity generation using a microfiltration membrane air-cathode single-chamber microbial fuel cell. Bioresource Technology 100 (13), 3185e3192. Sun, J., Hu, Y.Y., Be, Z., Cao, Y.Q., 2009b. Improved performance of air-cathode single-chamber microbial fuel cell for wastewater treatment using microfiltration membranes and multiple sludge inoculation. Journal of Power Sources 187 (2), 471e479. Ter Heijne, A., Hamelers, H.V.M., Buisman, C.J.N., 2007. Microbial fuel cell operation with continuous biological ferrous iron oxidation of the catholyte. Environmental Science and Technology 41 (11), 4130e4134. Van der Zee, F.P., Villaverde, S., 2005. Combined anaerobiceaerobic treatment of azo dyese a short review of bioreactor studies. Water Research 39 (8), 1425e1440. Virdas, B., Rabaey, K., Yuan, Z., Keller, J., 2008. Microbial fuel cells for simultaneous carbon and nitrogen removal. Water Research 42 (12), 3013e3024. You, S.J., Ren, N.Q., Zhao, Q.L., Kiely, P.D., Wang, J.Y., Yang, F.L., Fu, L., Peng, L., 2009. Improving phosphate buffer-free cathode performance of microbial fuel cell based on biological nitrification. Biosensors and Bioelectronics 24 (12), 3698e3701. You, S.J., Zhao, Q.L., Zhang, J.N., Jiang, J.Q., Zhao, S.Q., 2006. A microbial fuel cell using permanganate as the cathodic electron acceptor. Journal of Power Sources 162 (2), 1409e1415. Zhang, L.X., Liu, C.S., Zhuang, L., Li, W.S., Zhou, S.G., Zhang, J.T., 2009. Manganese dioxide as an alternative cathodic catalyst to platinum in microbial fuel cells. Biosensors and Bioelectronics 24 (9), 2825e2829. Zhao, F., Harnisch, F., Schrorder, U., Scholz, F., Bogdanoff, P., Herrmann, I., 2005. Application of pyrolysed iron(II) phthalocyanine and CoTMPP based oxygen reduction catalysts as cathode materials in microbial fuel cells. Electrochemistry Communication 79 (12), 1405e1410. Zhao, F., Harnisch, F., Schrorder, U., Scholz, F., Bogdanoff, P., Herrmann, I., 2006. Challenges and constraints of using oxygen cathodes in microbial fuel cells. Environmental Science and Technology 40 (17), 5193e5199.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 9 2 e3 0 2
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Behavior, mass inventories and modeling evaluation of xenobiotic endocrine-disrupting chemicals along an urban receiving wastewater river in Henan Province, China Yi-Zhang Zhang a, Xian-Fang Song b, Akihiko Kondoh c, Jun Xia b, Chang-Yuan Tang a,* a
Graduate School of Horticulture, Chiba University, Matsudo 271-8510, Japan Key Laboratory of Water Cycle and Related Land Surface Process, Institute of Geographical Sciences and Natural Resources Research, Chinese Academy of Sciences, Datun Rd. A11, Chaoyang District, Beijing 100101, China c The Center for Environmental Remote Rensing, Chiba Unviersity, Chiba 263-8522, Japan b
article info
abstract
Article history:
Historically, the locations of cities mainly depend on the available water source and the
Received 12 February 2010
urban river not only supplies the fresh water to city but also receives its wastewaters. To
Received in revised form
analyze the influences of urban zone on its receiving water river, the Jialu River in Henan
11 July 2010
Province, China, a typical urban river was chosen. Water and sediment samples were
Accepted 19 July 2010
collected along the river in 2007 to analyze the concentrations of xenobiotic endocrine-
Available online 29 July 2010
disrupting chemicals (XEDCs) including nonylphenol (NP), octylphenol (OP) and bisphenol A (BPA) in surface water and sediment. The results showed that the concentrations of OP,
Keywords:
NP and BPA in surface water were 20.9e63.2 ng L1 (mean 39.8 ng L1), 75.2e1520 ng L1
Alkylphenols
(mean 645 ng L1), 410e2990 ng L1 (mean 1535 ng L1), respectively. The lowest and
Bisphenol A
highest concentrations of XEDCs in surface water were found in the upper stream and
Partition
downstream of Zhengzhou urban zone, which was regarded as the major discharge source
Fugacity
of these chemicals to this river. The concentrations of OP, NP and BPA in the sediment were 15.9e31.1 ng g1, 145e349 ng g1 and 626e3584 ng g1 with the average concentrations of 21.4 ng g1, 257 ng g1 and 2291 ng g1, respectively. The results of in situ sedimentewater partition of XEDCs showed that the partition coefficients (log Koc0 ) in the downstream were higher than that in the upstream, which was mainly caused by the retransfer of surface sediment from the upper stream to the downstream. Comparison of measured and theoretical inventories of XEDCs in sediment indicated that the residual time of XEDCs in sediment in the river was about 5 years, which was in the same order of magnitude with its big flood frequency. In order to predict concentration variances of XEDCs in surface water, a fugacity-hydrodynamic model was developed according to the concept of in series completely stirred tank reactors (CSTR). The model results showed that about 29e65% of XEDCs derived from the urban zone (about 2.0 t yr1) would finally dissipate from aqueous phase in the 170 km downstream of the river. Assuming the discharge amount of XEDCs from the urban zone remaining constant, the predicted concentrations of the total XEDCs in the over 90% river reach would be higher than 1.0 mg L1 under all normal, high water and low water season in 2007. ª 2010 Elsevier Ltd. All rights reserved.
* Corresponding author. Tel./fax: þ81 47 308 8911. E-mail address:
[email protected] (C.-Y. Tang). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.07.057
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 9 2 e3 0 2
1.
Introduction
Xenobiotic endocrine-disrupting chemicals (XEDCs) such as alkylphenols (APs) and bisphenol A (BPA) have attracted serious attention in environmental science research and policy due to theirs ubiquity and estrogenic activities (Sumpter and Johnson, 2005). APs, which mainly contain nonylphenol (NP) and octylphenol (OP), are the degradation products of the widely used non-ionic surfactants alkylphenol ethoxylates (APEs) (Sharma et al., 2009). Bisphenol A is an important intermediate used to synthesis polycarbonate plastic, epoxy resin and other plastic material such as polyvinyl chloride (PVC) (Staples et al., 1998; Klecka et al., 2001). In China, the production and consumption of BPA were estimated at 3.2 104 t and 1.05 105 t in 2001 (Zeng et al., 2006). Most APs and BPA are released from the sewage treatment plants (STPs) effluents, which led to a net accumulation in the aquatic ecosystem of urban rivers (Ying et al., 2002; Ce´spedes et al., 2008; Kang et al., 2007). Once entering the aquatic environment, XEDCs would moderate partition to organic phases such as sediments, soils and aquatic organism in light of their relatively high log octanolewater partition coefficients of 3.4e4.48 (Cousins et al., 2002; Ahel and Giger, 1993). Although they are not considered to be persistent organic pollutants, XEDCs are regularly detected in a variety of environmental media including the surface water and groundwater (Cailleaud et al., 2007; Latorre et al., 2003), sediments (Fu et al., 2007), aquatic organism (Belfroid et al., 2002; Pojana et al., 2007) etc. Concern has increased on these XEDCs such as the 4-tertiary isomers of NP and OP, which was shown to be 103e104 times less estrogenic potent than 17b-oestradiol (Jobling and Sumpter, 1993; Johnson et al., 2005). Thus, these chemicals have already been designated as priority hazardous substances in Water Framework Directive of European Union and they are subject to an environmental risk assessment (ERA) (Soares et al., 2008; Oehlmann et al., 2008). Monitoring of water quality on the concentrations of organic contaminants is costly and difficult to perform due to both temporal and spatial sampling restricts (Gevaert et al., 2009). Mathematical model provides an efficient way to predict the behavior and fate of contaminants in the environment and has been applied to determine levels of organic microcontaminants in STPs effluent and river water such as steroidal estrogens (Johnson and Williams, 2004). Johnson et al. (2008) compared the advantages and weaknesses of modeling and chemical analysis for polar organic chemicals in rivers and indicated that combination of measurements and models would provide the greatest confidence in assessing the risk of organic microcontaminants to the wildlife. In the past, the fugacity models have been successfully applied in predicting the multimedia fate of organic chemicals in the environment (Huang et al., 2007). Although the fugacity model IV can be used for describing unsteady environmental conditions, the model still regards the environment media as a wellmixing box by regardless of the longitudinal, vertical and lateral variances (Deksissa et al., 2004). Kilic and Aral (2009) developed a continuous and dynamic model to simulate the behavior and fate of organic contaminants in all phases, but the model was considered complex due to several non-linear equations were involved. Thus, development of a simple
293
mathematical model for assessing the water quality in an urban receiving river is important. The Huaihe River, one of the five main rivers in China, has attracted national attentions due to its frequent flood disasters and degeneration of water quality. The river has been seriously polluted and becomes more and more vulnerable to water pollution since 1980s. With economic growth and urbanization development, increasing discharges of industrial and domestic wastewaters were observed in the Huaihe River basin (Chen et al., 2005). One of its tributary, the Jialu River, has been regarded as a typical polluted urban receiving river (Lu et al., 2008), where the levels of APs in surface water have also been reported in our previous study (Zhang et al., 2009). As a sequential part of an interdisciplinary research focusing on the assessment of water body quality and hydrological process of the contaminants, the objectives of this paper are 1) to make clear the distribution characterization of XEDCs in surface water and sediment along the river; 2) to estimate the sediment inventory of XEDCs and discuss their behavior and fate along the river; 3) to develop a fugacity-hydrodynamic model for predicting the concentrations of the organic pollutants in surface water.
2.
Materials and methods
2.1.
Site description and sample collection
Originating from Xinmi County, Henan Province, the Jialu River is 256-km long with its basin area of 5896 km2. It flows via Zhengzhou, Zhongmou, Weishi and Xihua then down into the Shaying River near Zhoukou city. The Jialu River basin has been undergoing a rapid economic growth and urbanization, facing massive discharge of wastewater and declining surface water quality. Large numbers of treated and untreated sewage from the alongshore cities, towns and villages were estimated to be 25,124 104 tons per year from 1996 to 1999, 81% of which were discharged from Zhengzhou urban zone (Xiao et al., 1999). Zhengzhou city has a long history for textile and metallurgy industries, and has been listed as one of the six most important industrial cities by “the development of central zones” stratagem of Chinese Government. In 2006, the total population of the city was 7.20 millions, of this 3.06 millions was non-agricultural population (Statistical Bureau of Zhengzhou city, 2007). The concentrations of NH4eN in surface water, which was observed at downstream of the urban zone, varied from 25.8 to 87.8 mg L1 with the average concentration of 63.9 mg L1 in the year 2007 (Li, 2009). Thus, it is reasonable to regard that the Zhengzhou urban zone has a strong impact on the surface water quality of the Jialu River. Surface water samples were collected with pre-rinsed 2-L glass bottles at each site along the river in September 2007. The sampling locations were shown in the Fig. S1 in supplementary information section. After filtering through a prebaked 0.45 mm glass fiber membrane by a portable vacuum pump in situ, the samples were subsequently refrigerated and the extraction was performed within 24 h. Top 5-cm surface sediments were taken using stainless steel small-volume grab sampler and placed in precleaned glass bottles. The samples
294
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 9 2 e3 0 2
were stored in a refrigerator at 20 C until analysis. Repeated measures of dissolved oxygen (DO) and electric conductivity (EC) were performed in situ by portable meters (D-55, Horiba, Japan). The global positioning system (GPS) was used to locate the sampling sites.
2.2.
Analytical procedure
All solvents used for sampling and analysis were HPLC grade. 4-tert-octylphenol and bisphenol A were purchased from Wako Pure Chemical Industries (Japan), 4-nonylphenol (technical grade) corresponding to a mixture of different isomers was obtained from Tokyo Chemical Industry (Japan). Phenanthrene-d10 was purchased from AccuStandard, Inc (USA). Sodium sulfate (analytical reagent grade, Wako Pure Chemical Industries Ltd., Japan) was rinsed with dichloromethane (DCM) and hexane for three times and completely dried in the fume hood. The dried sodium sulfate was baked at 450 C for 4 h and stored in desiccator. The Florisil (60e200 mesh, USA) was activated at 650 C for 3 h and deactivated at 130 C for 12 h, then the Florisil was cooled in desiccators and after it, ultrapure water (5%, w:w) was added. To remove organic contaminants, all the glassware used for organic compounds analysis were baked for 4e5 h at 450 C prior to use. The method of pretreatment for water samples was described by our previous report (Zhang et al., 2009). Briefly, each filtrate was extracted by solid-phase extraction cartridge (InertSep RP-1, GL sciences). These cartridges had been previously washed with 5 mL each of DCM, methanol, and ultrapure water. The analyte was eluted from cartridges with 20 mL dichloromethane after 30 min dryness, and then the solution was concentrated to 0.2 mL under a gentle stream of high purity nitrogen. The extraction of target compounds from sediment was based on the method reported by Arditsoglou and Voutsa (2008). Sediment samples (10.0 g) were ultrasonically extracted in triplicate with 20 ml of acetone: methanol (1:1) for 20 min. And the copper slices were added to remove the sulfur. The extraction was centrifugated for several minutes at >3000 rpm, and then it was reduced to 1e2 ml by using rotary evaporator at 35 C. The Florisil column (1 cm i.d.) was employed and the target compounds were sequentially eluted with hexane, hexane/DCM (7:3, v/v) and ethyl acetate. The last fractions were collected in vials and then were concentrated to 1 ml. Then, an appropriate volume of the internal standard (phenanthrened10) was added into sample prior to GC-MS analysis. The instrumental conditions for analyzing OP, NP and BPA, and the quality assurance/quality control (QA/QC) were described in detailed in supplementary information section. The method of determining organic carbon contents of sediment samples was also provided in supplementary information section. In addition, a non-parametric Spearman rank R correlation test was performed to assess relationships between XEDCs and environmental parameters with the SPSS 13.0 for windows.
3.
Theories and models
To predict the dynamic transport and fate of organic pollutants in a natural river, the model developed in this study
regarded the river as lots of completely stirred tank reactors (CSTR) connected in series. Due to the river has a small width length ratio (<0.3&), one-dimensional advection-disperse St. Venant equation was introduced to solve the hydrodynamic conditions of natural river systems. Without consideration of lateral input along the river, the mass balance equation of XEDCs can be generally described: X vC vC v vC vN þm ¼ E þ rc þ vt vx vx vx vAx
(1)
where C (all symbols were listed in Table 1) are the concentrations of XEDCs in aqueous phase (mg L1); m is the average velocity of the river water (m s1); E denotes longitudinal P r present the sum of all dispersion coefficients (m2 s1); reaction rate constants including biodegradation and photolysis, which are on the basis of assumption of these reaction are subject to first-order kinetics (s1); N represents multimedia mass transfer flux (mg s1) such as airewater diffusion, which is a function of concentration (C ) and varies with the distance x. A is the area of the cross-section (m2). Multimedia fugacity model has been applied to describe the partitioning and transfer of organic pollutants from one phase to the other based on the approach of Mackay (2001). Briefly, the concentration (C ) can be replaced by the product of fugacity (f , Pa) and fugacity capacity (Z, mol m3 Pa1). Fugacity (f ) is identical to the partial pressure of ideal gases related with the chemical potential. Fugacity capacities (Z ) for air, water and sediment can be calculated using the equations (Mackay and Paterson, 1991): ZA ¼
1 1 Sw ZW ¼ ¼ RT H Ps
Zs ¼
r½OC%Koc H
(2)
where R is the universal gas constant (¼8.314 Pa m3 mol1 K1); T is the absolute temperature (K); H is Henry’s law constant (Pa m3 mol1); Sw and Ps are the aqueous solubility (mol m3) and vapor pressure (Pa). Koc is organic carbon-normalized partition coefficient (L kg1); r is the density of sediment (kg L1); OC% is the fraction of total organic carbon. Transfer flux N (mol h1) of the compound between two phases can be calculated with transfer/transformation rate (D, mol h1 Pa1) and fugacity difference between two phases (f 1 f 2) based on the approach described by Mackay (2001): BdZ N ¼ D f 1 f2 ¼ f1 f2 Dy
(3)
where B is molecular diffusion coefficient (m2 h1); Dy represents the diffusion path length (m); d is the interfacial area (m2). The detailed descriptions of parameterization for hydrodynamic and fugacity models were provided in the supplementary information section.
4.
Results and discussion
4.1. Occurrence of XEDCs in surface water and sediment along the river Concentrations of BPA in the surface water of the Jialu River varied from 0.41 to 2.99 mg L1 with mean value of 1.54 mg L1
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 9 2 e3 0 2
295
Table 1 e Symbols used in the context. Symbol
Unit 2
A Bw
m m2 h1
Bs
m2 h1
Cw
mg L1
Cs
mg g1
Ca d Daw Dsw E fa fs fw g h I rbio Kva
ng m3 cm mol h1 Pa1 mol h1 Pa1 m2 s1 Pa Pa Pa m s2 m dimensionless d1 m h1
Kvw Ksw
m h1 m h1
kdil Koc
m1 L kg1
Kp
L kg1
Kow Ntotal
dimensionless mol h1
Ndif-aw Ndif-sw Ps Q R Sw
mol h1 mol h1 Pa m3 s1 P m3 mol1 K mol m3
T OC% m m* Vm Za Zs Zw x әx r q h k d u 4
K dimensionless m s1 m s1 ml mol1 mol m3 h1 mol m3 h1 mol m3 h1 m m kg L1 dimensionless cp dimensionless m2 m kg L1
Definition Area of cross section Molecular diffusion coefficient in water Molecular diffusion coefficient in sediment Concentrations of chemicals in aqueous phase Concentrations of chemicals in sediment Concentrations of chemicals in air Thickness of sediment sampled Airewater transfer parameter Sedimentewater transfer parameter Longitudinal dispersion coefficient Fugacity of air Fugacity of sediment Fugacity of water Gravitational acceleration Depth of river Hydraulic gradient Biodegradation rate constant Air-side mass transfer coefficient over water Water-side mass transfer coefficient Water-side mass transfer coefficient over sediment Dilution factor vs. distance Organic carbon normalized partition coefficient Suspended particles/sediment and water partition coefficient Octanolewater partition coefficient Total mass transfer flux between phases Airewater diffusive flux Sedimentewater diffusive flux Vapor pressure River water discharge Universal gas constant Solubility of chemical in aqueous phase Absolute temperature Organic carbon content in sediment Flow velocity of river water Friction velocity of river water Molar volume of chemical at 20 C Fugacity capacity of air Fugacity capacity of sediment Fugacity capacity of water Simulation distance step Unit river length Sediment density Porosity of sediment Water viscosity coefficient at 20 C Transfer factor Interfacial area Water surface width Suspended particles concentration
(Fig. 1). The lowest and highest concentrations of BPA were found at sampling sites S1 and S2. S1 was allocated in the Jiangang reservoir, 4.5 km upstream of Zhengzhou urban zone, serving as the drinking water source for the urban zone.
Fig. 1 e Concentrations of XEDCs (OP, NP and BPA) in surface water (mg LL1) and sediment (mg gL1) in all sampling sites along the river. Mean of repeated analyses was plotted with one standard deviation.
S2 was allocated in the 2 km downstream of Zhengzhou urban zone, where EC value was the highest (1566 ms cm1). The mean concentration of DO in the surface water was 2.75 mg L1 with the lowest concentration of 0.53 mg L1 at S2, after this site, DO concentrations increased gradually to 3.80 mg L1 at S7 (Zhang et al., 2009). In contrary to DO variances, the concentrations of XEDCs in surface water decreased gradually along the river (S2eS7) and the concentration of BPA declined to 0.92 mg L1 at site S7 near the urban zone of Zhoukou. Several studies reported that high concentrations of XEDCs were detected in the industrial wastewaters such as textile effluents (Fu¨rhacker et al., 2000; Hale et al., 2000; Sole´ et al., 2000). Zhengzhou is one of the important textile industries cities in China and according to Yang et al. (2008), there are at least 11 textile mills, 1 pesticide factory and 1 detergent factory locating in its urban zone with the discharged wastewater of 24e1300 t d1. All the treated/ untreated industrial and domestic wastewaters were finally discharged into the Jialu River, the only wastewater receiver of the urban zone. Furthermore, no more big wastewater source was found along the river after it runs from S2 and therefore, the Zhengzhou urban zone can be regarded as the principal discharge source of XEDCs in the study area (Zhang et al., 2009). The good positive correlations were found between BPA and OP/NP and the relationship coefficients (R) were 0.93 ( p < 0.01) for OP and 0.89 ( p < 0.01) for NP, respectively. The lg (BPA/NP) ratio in surface water increased gradually from 0.29 (site S2) to 1.16 (site S6), whereas the lg (BPA/OP) ratio remained in the range 1.54e1.72 along the river indicating the dissipation of NP from aqueous phase being faster than that of BPA and OP. The explanation for different behaviors of NP, OP and BPA in surface water might be due to their physicalechemical characteristics and biodegradation rates. The octanol-water partition coefficient (log Kow) of NP (4.48) is higher than that of OP (4.12) and BPA (3.40) indicating that compared with OP and BPA, NP should have higher affinity to
296
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 9 2 e3 0 2
SPM and is easier to be adsorbed to SPM or bed sediment (Ahel and Giger, 1993; Staples et al., 1998). Johnson et al. (2000) studied the biodegradation of OP in surface water samples from Calder River and indicated that the biodegradation rate of OP, calculated as half-lives, were ranged from 8 to 13 d in the urban/industrial river water. Compared with the result, the dissipation rates of XEDCs were found to be faster in the Jialu River and the estimated half-lives of OP, NP and BPA removed from aqueous phase were 2.4, 1.6 and 2.5 d, respectively. In addition to biodegradation, sorption to suspended particles and bed sediment will also be an important mechanism for removal of XEDCs from aqueous phase such as OP, which has a strong affinity for bed sediment with Kp of 6e580 L kg1 (Johnson et al., 1998). Sharma et al. (2009) summarized the concentrations of BPA in river water from different regions in the world and reported the concentration range of BPA was from 0.005 to 4.0 mg L1. Jin et al. (2004) reported the levels of XEDCs in surface water near Tianjin city, north China, and gave the highest concentration of BPA was 8.30 mg L1. Dong et al. (2009) found the highest concentration of BPA was 3.92 mg L1 in surface water of the Pearl River estuary with its average concentration of 2.06 mg L1. Thus, it is found that BPA concentration in Jialu River was similar to results of these investigations. The sediment concentrations of XEDCs in the Jialu River were 15.9e31.1 ng g1 for OP, 145e349 ng g1 for NP and 626e3584 ng g1 for BPA and their average concentrations were 21.4, 257 and 2291 ng g1, respectively (Fig. 1). The average sediment concentration of XEDCs in middle and lower reaches (S4eS7) were higher than that in upper stream (S2eS4) with the ratios of 1.56 for OP, 1.22 for NP and 2.15 for BPA, respectively. The lg[OP], lg[NP] and lg[BPA] in all sediment samples were 1.32, 2.39 and 3.30 indicating the BPA and NP were the main contaminants compared with the OP. The correlation coefficients (R) of NP/BPA and OP in sediment were 0.54 ( p > 0.05) for NP and 0.71 ( p > 0.05) for BPA, respectively.
4.2. In situ sedimentewater partition of XEDCs along the river Understanding the degree of sedimentewater partitioning of the organic pollutants is of importance in the risk assessment of river ecosystem, development of sediment quality criteria and modeling the environmental behavior and fate of contaminants (Gobas and Maclean, 2003). Some studies have
reported that the adsorption of organic contaminants to sediment related with sediment’s organic carbon content (Xue et al., 2006). Organic carbon contents (OC%) in all sediment samples from the Jialu River varied from 2.44 to 3.06% with the mean value of 2.72%. The positive correlation between OC% and concentrations of XEDCs were found and correlation coefficients were 0.77 (P > 0.05) for OP, 0.71 (P > 0.05) for NP and 0.43 (P > 0.05) for BPA. The results indicated that sediment with the higher organic carbon was likely to adsorb the more XEDCs, and organic carbon plays an important role in controlling adsorption of XEDCs to sediment in the river. Although the disequilibrium status between sediment and water always occur and organic contaminants always undergo dynamic sorption/desorption in river ecosystem, the distribution of organic contaminants between water and sediment is still good indicator for predicting their behavior and fate in the environment (Schwarzenbach et al., 2003). In situ organic carbon-normalized partition coefficient (Koc0 ) between sediment and water is defined in eq. (4) K0oc ¼ Kp =foc Kp ¼ 1000 Cs =Caq
(4)
where foc is organic carbon fraction in sediment; Cs presents the XEDCs concentration in the sediment (mg g1); Caq denotes the concentration of XEDCs in aqueous phase (mg L1). The unit of partition coefficient (Kp) is L kg1. Table 2 showed the values of equilibrium partition coefficients (log Koc), octanolewater partition coefficient (log Kow) and measured in situ sedimentewater partition coefficients (log Koc0 ) in the Jialu River. Means of measured in situ log Koc0 of XEDCs along the Jialu River were in the range from 4.32 to 4.65 with their standard deviation (SD) of 0.23e0.58. As shown in Table 2, measured in situ log Koc0 values of OP, NP and BPA in the downstream (S5eS7) were greater than that in the upper stream (S2eS4). In situ log Koc0 values in the upstream varied from 4.04 to 4.19 for OP, which were in the range of equilibrium partition coefficients from the literature. And they varied from 3.81 to 3.98 for NP, which were lower than the equilibrium values. Although the values of measured in situ log Koc0 of BPA in the Jialu River were similar with the results of that (4.05e4.23) in Tiber river, Central Italy (Patrolecco et al., 2006), they were one order of magnitude higher than the literature results (Staples et al., 1998; Ying et al., 2003). Some studies showed that comparing with the measured in situ log Koc0 , the lower equilibrium log Koc may be due to the laboratory results were obtained under the minimizing other processes such as
Table 2 e The values of octanolewater partition coefficients (log Kow), equilibrium partition coefficients of XEDCs (log Koc) and the measured sedimentewater partition coefficients (log K0 oc) in the Jialu River. XEDCs
OP NP BPA a b c d
log Kow
a
4.12 4.48a 3.40b
log Koc0
log Koc
c
3.60e4.19 4.13e4.69c 3.02d
S2
S3
S4
S5
S6
S7
Mean log Koc0
SD
4.04 3.88 4.19
4.23 3.98 4.14
4.19 3.81 4.74
4.25 4.36 4.87
4.66 5.13 5.00
4.55 4.96 4.98
4.32 4.35 4.65
0.23 0.58 0.39
Data were obtained from literature Ahel and Giger (1993). Data were obtained from literature Cousins et al. (2002). Data were obtained from literature Navarro et al. (2009). Data were obtained from literature Ying et al. (2003).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 9 2 e3 0 2
297
volatilization, degradation and chemical bonding etc (Du¨ring et al., 2002). However, considering the Zhengzhou urban zone as the principal discharge source of XEDCs in the study area, the higher log Koc0 values occurred in the downstream mainly due to the suspended particles with XEDCs deposited to bottom sediment of the downstream and the retransfer of surface sediment from the upstream to the downstream. Persistence of OP and NP in the anaerobic bed sediment and sewage sludge has been previously reported (Johnson et al., 2000; Ying et al., 2003). Thus, it is reasonable to expect that the long-term accumulation of these chemicals in the downstream also resulted in the high sediment concentrations due to the anoxic conditions in the sediment of the Jialu River.
4.3. Estimated mass inventory of XEDCs in sediment along the river To assess the potential of sediment as a new contaminants source to the downstream and the influences of urban zone on its receiving rivers, the sediment inventory of XEDCs along the Jialu River was estimated. The river was divided into 5 compartments according to the sampling sites. Each of the 5 sampling sites was allocated in the start of one compartment (except S1 and S7) and the average concentration of XEDCs in two adjacent sites was used to represent the entire concentration of the compartment. The inventory (M, in kg) of XEDCs in sediment along the river was calculated by using the following equation: Measured Msediment ¼ krCis Ai d
(5)
where Cis is the average sediment concentration of XEDCs in compartment i (mg g1); Ai is the area of one compartment i (m2); r is the average sediment density of 1.23 g cm3; k is the conversion factor; and d is the thickness of sediment sampled, which was 5 cm in this study. The estimated sediment inventory of XEDCs in the Jialu River was 807 kg with average load of 4.8 kg km1 (Fig. 2). The sediment inventory in the middle and downstream (S4eS7) appeared to be 2.28 times higher than those in the upstream (S2eS4) suggesting that the downstream acts as the net sink of XEDCs to the upper stream. In order to better to understand the accumulation of XEDCs in sediment, we replaced the Cis in eq. (5) with the theoretical Cis0 , which was deduced from respective XEDCs concentration in surface water at site i under the assumption of sedimentewater partition equilibrium, by using the following equation: Theoretical Msediment ¼ krC0is Ai d C0is ¼ Koc foc Caq 0
(6)
where Cis is the theoretical sediment concentration of XEDCs in compartment i under partition equilibrium (mg g1). Koc is the organic carbon-normalized equilibrium partition coefficient. Navarro et al. (2009) showed that the sedimentewater log Koc of XEDCs ranged from 4.13 to 4.69 for NP and 3.60e4.19 for OP, which were in good agreement with the ones obtained from both other sorption experiments and empirical correlations. Therefore, in the present study, the log Koc of XEDCs were derived from the experimental results reported by Navarro et al. (2009) and Ying et al. (2003). As shown in Fig. 2,
Fig. 2 e Measured and theoretical mass inventories of XEDCs in sediment (0e5 cm) along the Jialu River (elevation of each sampling site was plotted in relation to distance from S1).
the theoretical mass inventory from S2 to S4 accounted for 72.6% of the total theoretical mass inventory in sediment (S2eS7). Compared with the measured mass inventory, these results provided useful information on the behavior and fate of XEDCs in sediment along the river: 1) since the Zhengzhou urban zone was regarded as the principal source of XEDCs, the higher accumulation of XEDCs in the middle and downstream derived from the retransfer of suspended surface sediment. And the results were in agreement with hydraulic gradient along the river, which was relatively lower in the middle and downstream (mean 0.15&) than that in upstream of 0.22& (Fig. 2); 2) as for the whole river reach (S2eS7), the measured mass inventory of total XEDCs was about 5.14 times higher than its theoretical values, which can be used as an indicator for reflecting the residual time of XEDCs in the river. That is, the average accumulative time of XEDCs in the river was estimated at about 5 years. In general, a river is a highly complex system and function as a unified whole, the dynamics of which is significantly influenced by several important variables such as discharge, velocity, suspended particles concentrations, and sediment load. Flood disasters frequently occur in the Huaihe River basin and thus, about 5674 reservoirs and 5427 sluices have been constructed in its main stem and tributaries. Heavy rain events in the river catchment basin could cause the floods and change the deposition profile of sediment along the river. Fig. S2 gave the frequency distribution of the rainfall of the Huaihe River basin during the Mei-yu season (mid-Jun to mid-Jul) from 1954 to 2003.The statistic frequency of the rainfall of >300 mm is about 24%, which is in agreement with an average accumulative time of XEDCs in the Jialu River (Fig. S2). Thus, it can be deduced that the big flood, often occurred in the river, would remove the mass loadings of organic contaminants in sediment, which makes the ratio of measured and theoretical mass inventories in sediment was in the one order with the big flood frequency. On the other hand, the measured sediment concentrations of XEDCs in the middle and the downstream were higher than
298
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 9 2 e3 0 2
the respective equilibrium concentrations, which suggested that the non-equilibrium conditions in terms of XEDCs partitioning existed and the sediment may serves as a potential source of XEDCs to river water in the middle and the downstream. Therefore, multimedia mass transfer should be taken into account in predicting the concentration variances of XEDCs and further performing environmental risk assessment in surface water along the river.
4.4. Modeling evaluation of transport of XEDCs in surface water along the river 4.4.1.
Model scheme
There are three STPs in urban zone of Zhengzhou and the treated water (about 70 104 t d1) was finally discharged into the Jialu River (Table S1). The effluents of the three STPs accounted for 39e71% of the river water discharge at S3 in 2007 and the lowest proportion occurred at September, which was used to represent the high flow season in this study. Under the assumption of discharged XEDCs amount and the river water discharge remaining stable during the sampling period (no rain event happened), the river is treated as a steady state system, that is vC=vt ¼ 0. Thus, transport of XEDCs in the river can be described by the one-dimension advectionedispersion equation of stable state. Fugacity models can simplify the analysis of mass fluxes of organic contaminants from one phase to the other such as the airewater and wateresediment exchanges (Kilic and Aral, 2009). The mass transferring processes of different phases mainly include dry/wet deposition, sediment deposition, airewater and wateresediment diffusion and so forth. Because the rain event didn’t happened during the sampling period, the airewater exchange only included the dry deposition and diffusion processes in the present study. Sampling site S1 was allocated in a spare reservoir of Zhengzhou city, a piedmont upper stream of Zhengzhou urban zone, where the water body is relatively still and immune from human activity. With an assumed airewater exchange equilibrium existing at S1, the calculated atmospheric fugacity (f A) of OP, NP and BPA were 5.67 108, 3.75 108 and 1.036 1011 Pa and their respective concentrations in atmosphere were 4.83, 338.7 and 0.001 ng m3, respectively. These values were used as the average atmospheric concentrations of XEDCs in the whole Jialu River basin. Therefore, the dry deposition flux (Ndep) per unit interfacial area can be estimated at 2.52 1010 mol h1 for OP, 1.66 108 mol h1 for NP and 4.82 1013 mol h1 for BPA based on the assumed dry deposition velocity of 10.8 m h1 (Mackay and Paterson, 1991). The total mass flux from air and sediment to aqueous phase (Ntotal, mol h1 m2) through unit interfacial area can be written as following: Ntotal ¼ Ndif aw þ Ndif sw þ Ndep ¼ Daw fA Daw fw þ Dsw fs Dsw fw þ Ndep ¼ Daw fA þ Dsw fs ðDsw þ Daw Þ fw þ Ndep
(7)
where Ndif-aw and Ndif-sw are airewater and wateresediment diffusive flux (mol h1) per unit interfacial area (d, m2), the product of water-surface width (u, m) and unit river length (әx,
m); Daw and Dsw are airewater and wateresediment transfer coefficients (mol h1 Pa1); f A, f w, f s are the fugacities of air, water and sediment. f s was calculated from the average concentrations of the XEDCs in all sediment samples and the values were 1.10 107 Pa, 1.13 105 Pa and 1.11 1010 Pa for OP, NP and BPA, respectively. Conceptually, the whole river was divided into n sections (the section was defined as a tank with its width and length equal to river width and unit river length) and the completely stirred tank reactors (CSTR) model was applied in each section of the river. For section i, the unit concentration variance (DCi) of XEDCs in aqueous phase, which was caused by the multimedia mass transfer, can be written with the increase of mass flux (Dm) per unit water volume: DCi ¼ Dmi =Vi ¼ Ntotali =Qi ¼ dða bC0i Þ=Qi
(8)
Therefore, the final concentration of XEDCs in aqueous phase at each section i can be calculated based on the equation: Ci ¼ ð1 F0 Þ ðC0i þ DCi Þ !# " x ma 1 mb ¼ 1 exp l kdil x F ð1nÞ þ Ci1 ð1nÞ mte Qi Qi " 1=2 # 4rbio E ; Qi ¼ Qi1 exp kdil x l ¼ m=2E 1 1 þ 2 m
(9)
where Qi is the river water discharge at section i (m3 s1); Ci0 is the concentration of XEDCs at section i, which is calculated without consideration of the multimedia mass transfer; Ci1 and Ci are the final concentrations of XEDCs in surface water at sections i 1 and i; Vi is the water column volume in section i, which equal to the product of Q and time (t); x the simulation distance step (m) and d is the interfacial area per unit river length (m2); (1 F0 ) denotes the partition fraction of XEDCs in aqueous phase within distance step; m and n are the relationship coefficients of river water discharge and water-surface width (Parameterization in supplementary information). a and b denote the formulas (Dawf A þ Dswf s þ Ndep) and (Dsw þ Daw)/Zw, respectively. The values of a and b are 3.75 1010 and 0.0012 mol h1 for OP, 1.68 108 and 0.016 mol h1 for NP, 1.20 109 and 5.62 105 mol h1 for BPA.
4.4.2.
Sensitivity analysis and model calibration
The sensitivity of important variables in the model were tested by comparing model results without any changed variable with those with only one variable changed by 10%, which allows us to determine the key parameters and others may remain on default values. Results of sensitivity analysis were shown in Table 3. The results showed that the model was insensitive to the water depth h, water-surface width u, sediment porosity q and water-side mass transfer coefficient Kvw with their changes of <0.1% (except Kvw changes of 1.11% and 0.95% for NP). The estimations of XEDCs were moderately sensitive to river water discharge Q and organic carbonnormalized partition coefficients Koc with their changes of <1.0%, whereas the changes of Q were 3.10% and 3.64% for NP. The model was sensitive to other parameters such as river water velocity m, biodegradation rate constant rbio, dilution factor kdil, and air-side mass transfer coefficient Kva.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 9 2 e3 0 2
299
And 10% changes of these parameters can result in >1.0% changes of the model results. The key model parameters were calibrated using the measured results of BPA in surface water (with 18.6% error). And then, the calibrated parameters were used to run the model for predicting of the concentrations variances of OP and NP in surface water along the river. As shown in Fig. 3, the general trends of the model estimations were in good agreement with the measured results and the relative standard deviation between the simulated and the measured values in all sampling sites were 22.8% for OP and 25.1% for NP. The results indicated that the model fit well with the measured values and can be assumed acceptable.
zone under normal period (average flow of the year 2007), higher water period (the highest water discharge of the year 2007) and lower water period (the lowest water discharge of the year 2007); 2) under the normal period, the amount of discharged XEDCs from urban zone doubles due to the urban development or reduces by half due to improved management of wastewater. Predicted total concentrations of all the three XEDCs in surface water along the river under different scenarios were plotted in Fig. 4. As for scenario of stable discharge of XEDCs, the concentrations of total XEDCs throughout the river were higher than 1.0 mg L1 in the year of 2007. Even though under the high water period, the total XEDCs concentrations in almost half of the river reach were more than 2.0 mg L1 (Fig. 4a). As shown in Fig. 4b, if the discharged XEDCs double, the total XEDCs concentrations in more than 90% river reach would be higher than 3.0 mg L1 under the normal period. Even though the XEDCs discharged from urban zone could be decreased by half, the total XEDCs concentrations of >1.0 mg L1 in surface water would persist about 120 km from the discharge source. Belfroid et al. (2002) reported that the concentrations of BPA in fish were in the range 2e75 mg kg1 dry weight compared with the corresponding concentrations of <0.01e0.33 mg L1 in surface water in The Netherlands. Thus, the potential risk of BPA associated with accumulation in the food chain was expected. Zhao et al. (2008) reported that there were some fish such as cyprinoid living in the Zhoukou reach of the Shaying River, where the Jialu River flowed (near S7). Based on the predicted concentrations of BPA at S7 in the year of 2007 and the average annual flow of the Shaying River in Zhoukou reach (125.5 m3 s1), the estimated concentrations of BPA in surface water of this reach were 0.18e0.28 mg L1 during 2007. And thus, the bioaccumulation of BPA in fish in the Zhoukou reach of the Shangying River could be expected. In addition, based on the results of model, the mass loadings of the total XEDCs in surface water along the river can be calculated and the mass loadings (M, in kg yr1) of XEDCs in surface water vs. distance from discharge source S2 (x, in kilometer) can be described as:
4.4.3.
Mwater ¼ 7:27x þ 2092 R2 ¼ 0:98
Table 3 e Results of sensitivity analyses for variables in the model. Variable
Calibrated values
m
0.71 m s1
h
1.56 m
Q
20.4 m3 s1
u
40.5 m
kdil
3.16 106
Kva
5 m h1
Kvw
0.05 m h1
q
0.3
Koc rbio
Variance (%)
OP
NP
BPA
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
1.26 1.05 0.00 0.00 0.28 0.23 0.00 0.00 2.67 2.58 2.67 2.58 0.01 0.00 0.01 0.01 0.21 0.21 1.26 1.05
1.90 1.59 0.0 0.0 3.64 3.10 0.00 0.00 2.28 2.22 2.28 2.22 1.11 0.95 0.01 0.01 0.14 0.14 1.90 1.59
3.43 2.93 0.0 0.0 0.01 0.01 0.00 0.00 2.52 2.44 2.52 2.44 0.00 0.00 0.01 0.01 0.47 0.46 3.43 2.93
Ratio ¼ 100% (results with one changed variable e results without any changed variable)/results without any changed variable.
Scenario analysis
In this section, two kinds of different scenarios were simulated: 1) constant discharged amount of XEDCs from urban
(10)
The result indicated that the amount of the total XEDCs discharged from Zhengzhou urban zone to the river was about
Fig. 3 e Model calibration results for BPA and model validation results for OP and NP in surface water along the Jialu River. Data are simulated (lines) and measured (symbols) concentrations of XEDCs in aqueous phase. BPA (left); OP (middle); NP (right).
300
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 9 2 e3 0 2
Fig. 4 e Predicted total concentrations of all the three XEDCs in surface water along the river under different scenarios. (a) constant discharge of XEDCs under different hydrological conditions: normal, high water and low water periods in the year of 2007 (left); (b)variable discharge of XEDCs under normal period (right).
2.0 t yr1 and about 29e65% of the discharged XEDCs were finally dissipated from the aqueous phase in its 170 km downstream. Therefore, the influences of Zhengzhou urban zone on its receiving river were significant and the continuous monitoring and prediction of XEDCs in the Jialu River are necessary.
5.
Conclusions
The anthropogenic contaminants XEDCs (OP, NP and BPA) were used as indicators to study the influence of urban zone on its receiving river and thus, the Jialu River seriously polluted by the Zhengzhou urban zone in China was chosen, where the average concentration of NH4eN in surface water (near S3) was 63.9 mg L1 in the year 2007. The concentrations of XEDCs in surface water and sediment were measured and the results suggested that Zhengzhou urban zone can be regarded as the principal discharge source of XEDCs in the catchment. The positive correlation of organic carbon content and XEDCs concentration in sediment were found and the relationship coefficients were 0.77, 0.71 and 0.43 for OP, NP and BPA. The results of in situ sedimentewater partition of XEDCs showed that the log Koc0 values in the downstream was higher than that in the upstream suggesting that the retransfer of suspended surface sediment was one of reasons why the in situ log Koc0 in field was higher than equilibrium value in laboratory. In order to understand the behavior and fate of XEDCs in sediment, measured and theoretical inventories of XEDCs in sediment were estimated with the thickness of sediment sampled. The results showed that mass inventory of XEDCs in sediment (0e5 cm) was 807 kg, about 70% of which would deposit in the middle and lower reaches. Comparison of measured and theoretical sediment inventories of XEDCs in the whole river suggested that the average residual time of XEDCs was about 5 years. The results were in accord with the big flood frequency of the river indicating that the
contaminants in surface sediment would be removed by the big floods in the Jialu River. To predict the concentration variance of XEDCs in surface water along the river, a fugacity-hydrodynamic model was developed. The hydraulic parameters of the model such as the velocity and depth were calibrated with the data of BPA, and then, predicted results of the calibrated model can agree well with the measured values of OP and NP, within 30% error. Compared with its advantage of numerical simplicity, the main limitation of the model is that the discharge sources of the pollutants are regarded as steady system so that the transport of pollutants can be described by one-dimensional advectionedispersion equation of steady state.
Acknowledgements This work was supported from the Grant-in-Aid for Scientific Research of Japan Society for the Promotion of Science (No. 21300334), the Program of Ministry of Science and Technology, China (2008ZX07010-006-1) and National Key Water Project, China (No.2009ZX07210-006). Authors thank Professor F.D. Li, Dr. S.Q. Wang and Dr. J.R. Liu for their assistance in sample collection.
Appendix. Supplementary information Supporting information associated with this article can be found, in the online version, at doi:10.1016/j.watres.2010.07.057.
references
Ahel, M., Giger, W., 1993. Partitionining of alkylphenols and alkylphenol polyethoxylates between between water and organic solvents. Chemosphere 26, 1471e1478.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 9 2 e3 0 2
Arditsoglou, A., Voutsa, D., 2008. Determination of phenolic and steroid endocrine disrupting compounds in environmental matrices. Environ. Sci. Pollut. Res. Int. 15, 228e236. Belfroid, A., Van Velzen, M., Van der Horst, B., Vethaak, D., 2002. Occurrence of bisphenol A in surface water and uptake in fish: evaluation of field measurements. Chemosphere 49, 97e103. Cailleaud, K., Forget, L.J., Souissi, S., Lardy, S., Augagneur, S., Budzinski, H., 2007. Seasonal variation of hydrophobic organic contaminant concentrations in the water-column of the Seine Estuary and their transfer to a planktonic species Eurytemora affinis (Calanoı¨d, copepod). Part 2: alkylphenolpolyethoxylates. Chemosphere 70, 281e287. Ce´spedes, R., Lacorte, S., Ginebreda, A., Barcelo, D., 2008. Occurrence and fate of alkylphenols and alkylphenol ethoxylates in sewage treatment plants and impact on receiving waters along the Ter River (Catalonia, NE Spain). Environ. Pollut. 153, 384e392. Chen, Y.H., Xu, J., Wang, J., Tian, D., Chen, J., Gao, Z.K., 2005. Physical and chemical properties of sediments in Huainan segment of Huaihe River. J. China Univ. Mining Tech. 15, 41e47 (English edition). ka, G.M., Mackay, D., 2002. A Cousins, I.T., Staples, C.A., Klec multimedia assessment of the environmental fate of bisphenol A. Hum. Ecol. Risk Assess. 8, 1107e1135. Deksissa, T., De Pauw, D., Vanrolleghem, P.A., 2004. Dynamic instream fate modeling of xenobiotic organic compounds: a case study of linear alkylbenzene sulfonates in the Lambro River, Italy. Environ. Toxicol. Chem. 23, 2267e2278. Dong, J., Li, X.L., Liang, R.J., 2009. Bisphenol A pollution of surfacewater and its environmental factors. J. Ecol. Rural Environ. 25, 94e97 (in Chinese). Du¨ring, R.A., Krahe, S., Gath, S., 2002. Sorption behavior of nonylphenol in terrestrial soils. Environ. Sci. Technol. 36, 4052e4057. Fu, M.Z., Li, H.Y., Gao, H.W., 2007. Distribution characteristics of nonylphenol in Jiaozhou Bay of Qingdao and its adjacent rivers. Chemosphere 69, 1009e1016. Fu¨rhacker, M., Scharf, S., Weber, H., 2000. Bisphenol A: emissions from point sources. Chemosphere 41, 751e756. Gevaert, V., Verdonck, F., Benedetti, L., De Keyser, W., De Baets, B., 2009. Evaluating the usefulness of dynamic pollutant fate models for implementing the EU water framework directive. Chemosphere 76, 27e35. Gobas, F.A.P.C., Maclean, L.G., 2003. Sedimentewater distribution of organic contaminants in aquatic ecosystems: the role of organic carbon mineralization. Environ. Sci. Technol. 37, 735e741. Hale, R.C., Smith, C.L., De Fur, P.O., Harvey, E., Bush, E.O., La Guardia, M.J., Vadas, G.G., 2000. Nonylphenols in sediments and effluents associated with diverse wastewater outfalls. Environ. Toxicol. Chem. 19, 946e952. Huang, G.L., Hou, S.G., Wang, L., Sun, H.W., 2007. Distribution and fate of nonylphenol in an aquatic microcosm. Water Res., 4630e4638. Jin, X.L., Jiang, G.B., Huang, G.L., Liu, J.F., Zhou, Q.F., 2004. Determination of 4-tert-octylphenol, 4-nonylphenol and bisphenol A in surface waters from the Haihe Riverin Tianjin by gas chromatography-mass spectrometry with selected ion monitoring. Chemosphere 56, 1113e1119. Jobling, S., Sumpter, J.P., 1993. Detergent components in sewage effluent are weakly oestrogenic to fish: an in vitro study using rainbow trout (Oncorhynchus mykiss) hepatocytes. Aquat. Toxicol. 27, 361e372. Johnson, A.C., White, C., Besien, T.J., Jurgens, M.D., 1998. The sorption of octylphenol, a xenobiotic oestrogen, to suspended and bed-sediments collected from industrial and rural reaches of three English river. Sci. Total Environ. 210/211, 271e282.
301
Johnson, A.C., White, C., Bhardwaj, L., Monika, D.J., 2000. Potential for octylphenol to biodegrade in some English rivers. Environ. Toxicol. Chem. 19, 2486e2492. Johnson, A.C., Williams, R.J., 2004. A model to estimate influent and effluent concentrations of estradiol, estrone and ethinyloestradiol at sewage treatment works. Environ. Sci. Technol. 38, 3649e3658. Johnson, A.C., Aerni, H.-R., Gerritsen, A., Gibert, M., Giger, W., Hylland, K., Ju¨rgens, M., Nakari, T., Pickering, A., Suter, M.J.-F., Svenson, A., Wettstein, F.E., 2005. Comparing steroid estrogen, and nonylphenol content across a range of European sewage plants with different treatment and management practices. Water Res. 39, 47e58. Johnson, A.C., Ternes, T., Williams, R.J., Sumpter, J.P., 2008. Assessing the concentrations of polar organic microcontaminants from point sources in the aquatic environment: measure or model? Environ. Sci. Technol. 42, 5390e5399. Kang, J.H., Aasi, D., Katayama, Y., 2007. Bisphenol A in the aquatic environment and its endocrine disruptive effects on aquatic organisms. Crit. Rev. Toxicol. 37, 607e625. Kilic, S.G., Aral, M.M., 2009. A fugacity based continuous and dynamic fate and transport model for river networks and its application to Altamaha River. Sci. Total Environ. 407, 3855e3866. Klecka, G.M., Gonsior, S.J., West, R.J., Goodwin, P.A., Markham, D.A., 2001. Biodegradation of bisphenol A in aquatic environments: river die-away. Environ. Toxicol. Chem. 20, 2725e2735. Latorre, A., Lacorte, S., Barcelo´, D., 2003. Presence of nonylphenol, octyphenol and bisphenol A in two aquifers close to agricultural, industrial and urban Areas. Chromatographia 57, 111e116. Li, F.D., 2009. Study on Groundwater Flow, Its Age, and Hydrogeochemical Evolution in the Alluvial Fans in the North China. Chiba University, Chiba. Lu, D.X., Huang, W.W., Li, X.M., Li, J., 2008. Correlation of Zhengzhou city economic development and Jialu River Zhoumouchenqiao section water quality. Henan Sci. 26, 1117e1120 (in Chinese). Mackay, D., Paterson, S., 1991. Evaluating the multimedia fate of organic chemicals: a level III fugacity model. Environ. Sci. Technol. 25, 427e436. Mackay, D., 2001. Multimedia Environmental Models: The Fugacity Approach, second ed. Lewis Publisher, Boca Raton, FL. Navarro, A., Endo, S., Gocht, T., Barth, J.A.C., Lacorte, S., Barcelo´, D., Grathwohl, P., 2009. Sorption of alkylphenols on Ebro River sediments: comparing isotherms with field observations in river water and sediments. Environ. Pollut. 157, 698e703. Oehlmann, J., Oetken, M., Schulte-Oehlmann, U., 2008. A critical evaluation of the environmental risk assessment for plasticizers in the freshwater environment in Europe, with special emphasis on bisphenol A and endocrine disruption. Environ. Res. 108, 140e149. Patrolecco, L., Capri, S., De Angelis, S., Pagnotta, R., Polesello, S., Valsecchi, S., 2006. Partition of nonylphenol and related compounds among different aquatic compartments in Tiber River (central Italy). Water Air Soil Pollut. 172, 151e166. Pojana, G., Gomiero, A., Jonkers, N., Marcomini, A., 2007. Natural and synthetic endocrine disrupting compounds (EDCs) in water, sediment and biota of a coastal lagoon. Environ. Int. 33, 929e936. Schwarzenbach, R.P., Gschwend, P.M., Imboden, D.M., 2003. Environmental Organic Chemistry, second ed. John Wiley and Sons, Hoboken, NJ.
302
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 9 2 e3 0 2
Sharma, V.K., Anquandah, G.K., Yngard, R., Kim, H., Fekete, J., Bouzek, K., Ray, A., Golovko, D., 2009. Nonylphenol, octylphenol, and bisphenol-A in the aquatic environment: a review on occurrence, fate, and treatment. J. Environ. Sci. Heal A 44, 423e442. Soares, A., Guieysse, B., Jefferson, B., Cartmell, E., Lester, J.N., 2008. Nonylphenol in the environment: a critical review on occurrence, fate, toxicity and treatment in wastewaters. Environ. Int. 34, 1033e1049. Sole´, M., De Alda, M.J.L., Castillo, M., Porte, C., LadegaardPedersen, K., Barcelo´, D., 2000. Estrogenicity determination in sewage treatment plants and surface waters from the catalonian area (NE Spain). Environ. Sci. Technol. 34, 5076e5083. Staples, C.A., Dorn, P.B., Klecka, G.M., O’Block, S.T., Harris, L.R., 1998. A review of the environmental fate, effects and exposures of bisphenol A. Chemosphere 36, 2149e2173. Statistical Bureau of Zhengzhou city, 2007. Zhengzhou Statistical Yearbook 2007. China Statistics Press, Beijing (in Chinese). Sumpter, J.P., Johnson, A.C., 2005. Lessons from endocrine disruption and their application to other issues concerning trace organics in the aquatic environment. Environ. Sci. Technol. 39, 4321e4332.
Xiao, S.Y., Yang, Z.F., Shen, X.H., 1999. Analysis of pollution trend in surface water in the Jialu River. Zhihuai 38 (in Chinese). Xue, N.D., Zhang, D.R., Xu, X.B., 2006. Organochlorinated pesticide multiresidues in surface sediments from Beijing Guanting reservoir. Water Res. 40, 183e194. Yang, L., Jia, B., Zhang, X.H., He, D., Deng, W.N., 2008. Zhengzhou urban water sources and surface water quality analysis of survey. Henan Sci. 26, 1121e1125 (in Chinese). Ying, G.G., Williams, B., Kookana, R., 2002. Environmental fate of alkylphenols and alkylphenol ethoxylatesda review. Environ. Int. 28, 215e226. Ying, G.G., Kookana, R.S., Dillon, P., 2003. Sorption and degradation of selected five endocrine disrupting chemicals in aquifer material. Water Res. 37, 3785e3791. Zeng, G., Zhang, C., Huang, G., Yu, J., Wang, Q., Li, J., Xi, B., 2006. Adsorption behavior of bisphenol A on sediments in Xiangjiang River, Central-south China. Chemosphere 65, 1490e1499. Zhao, C.S., Liu, C.M., Xia, J., Wang, G.S., Liu, Y., Sun, C.L., Wang, R., Ji, X. Y., 2008. In stream ecological flow of dammed river: a case study of Huaihe River. J. Natural Resources 23, 400e411 (in Chinese). Zhang, Y.Z., Tang, C.Y., Song, X.F., Li, F.D., 2009. Behavior and fate of alkylphenols in surface water of the Jialu River, Henan Province, China. Chemosphere 77, 559e565.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 0 3 e3 0 7
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Efficient recovery of nano-sized iron oxide particles from synthetic acid-mine drainage (AMD) water using fuel cell technologies Shaoan Cheng a,b, Je-Hun Jang b,1, Brian A. Dempsey b, Bruce E. Logan b,* a b
State Key Laboratory of Clean Energy Utilization, Department of Energy Engineering, Zhejiang University, Hangzhou 310027, PR China Department of Civil and Environmental Engineering, 212 Sackett Building, Penn State University, University Park, PA 16802, United States
article info
abstract
Article history:
Acid mine drainage (AMD) is an important contributor to surface water pollution due to the
Received 11 April 2010
release of acid and metals. Fe(II) in AMD reacts with dissolved oxygen to produce iron oxide
Received in revised form
precipitates, resulting in further acidification, discoloration of stream beds, and sludge
15 July 2010
deposits in receiving waters. It has recently been shown that new fuel cell technologies,
Accepted 19 July 2010
based on microbial fuel cells, can be used to treat AMD and generate electricity. Here we
Available online 27 July 2010
show that this approach can also be used as a technique to generate spherical nano-
Keywords:
approach therefore provides a relatively straightforward way to generate a product that
AMD
has commercial value. Particle diameters ranged from 120 to 700 nm, with sizes that could
Energy
be controlled by varying the conditions in the fuel cell, especially current density
Iron oxide
(0.04e0.12 mA/cm2), pH (4e7.5), and initial Fe(II) concentration (50e1000 mg/L). The most
particles of iron oxide that, upon drying, are transformed to goethite (a-FeOOH). This
Microbial fuel cell
efficient production of goethite and power occurred with pH ¼ 6.3 and Fe(II) concentrations
Particles
above 200 mg/L. These results show that fuel cell technologies can not only be used for simultaneous AMD treatment and power generation, but that they can generate useful products such as iron oxide particles having sizes appropriate for used as pigments and other applications. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Acid-mine drainage (AMD) and acid rock drainage (ARD) are caused by the oxidation of pyrite and other sulfidic minerals when they are exposed to air and water due to mining or excavation operations. Both types of discharges will be referred to as AMD in this paper. AMD waters contain high concentrations of mineral acid, dissolved Fe(II), and sulfates, as well as a variety of toxic chemicals such as lead, copper, cadmium, and arsenic. AMD can be difficult and costly to
treat. Treatment is usually accomplished with passive systems for low-flow abandoned discharges, and active processes for regulated discharges. Most active treatment involves a rapid increase in pH (lime, caustic, or other base) followed by agitation and energy-intensive forced aeration. In a few instances the iron oxides have been recovered for use as pigments, magnetic materials and catalysts (Kirby et al., 1999; Kairies et al., 2005; Wei and Viadero, 2007) or for fluoride or arsenate removal (Zhao and Stanforth, 2001; Luengo et al., 2007). However, the resulting sludge can be highly variable
* Corresponding author. Tel.: þ1 814 863 7908. E-mail address:
[email protected] (B.E. Logan). 1 Present address: Sandia National Laboratories, Carlsbad Programs Group, 4100 National Parks Highway, Carlsbad, NM 88220, United States 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.07.054
304
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 0 3 e3 0 7
2.
Methods
AMD-FCs were constructed as previously described (Cheng et al., 2007), with two chambers each 14 mL in volume separated by an anion exchange membrane. Both electrodes were made from carbon cloth with a projected surface area of 7 cm2, and the cathode contained a Pt catalyst (0.5 mg/cm2). The anode and cathode chambers were filled with deionized water containing NaCl (0.05 M, to increase solution conductivity) and NaHCO3 (0.06 M, pH buffer), and then sparged with CO2 for 0.5 h to adjust the pH and remove dissolved oxygen. FeSO4 was added to the anode chamber at 400 mg/L in an anaerobic glove box and the pH adjusted to 6.3 (except as noted) using HCl (0.1 M) under conditions of continuous CO2 sparging. The AMDFCs (duplicate reactors) were then placed in a temperature controlled room (30 C), and left in open circuit mode for 0.5 h before connecting the circuit containing a resistor (1000 U in all experiments except as noted). The medium in the reactor was refilled when the cell voltage dropped below e1 mV, which varied depending on conditions from 5 to 25 h. Particles on the anode surface were removed with a plastic plate and added to the anode solution, and then this solution was analyzed for particle sizes. To determine the effect of operating conditions on iron oxide production, the external resistance was varied from 100 to 2000 U, pH from 4 to 7.5, and ferrous iron from 0.9 to 17.9 mM. After each cycle of current generation, the solution was replaced with fresh media. The anode solutions were sonicated to disperse the particles (5 min, 1510R-MTH, Branson Sonicator, USA) and then analyzed for their particle size distribution (Zeta Sizer, Malvern Instruments Limited, UK). The particles were washed (deionized water) and centrifuged (2400g) three times and dried in air
at 40 C. Mo¨ssbauer spectra were obtained at room temperature using constant acceleration and a 57Co/Rh source of 25 mCi initial activity (Jang et al., 2003). The computer program “Recoil” (Lagarec and Rancourt, 1998) was used to analyze the velocity distributions. XRD was performed with an X’PERT Phillips four circle diffractometer using CuKa radiation. The powder sample was re-dispersed in ethanol, sonicated, and settled onto a highly oriented pyrolytic graphite disc for AFM observation (BioScopeAFM, Digital Instrument, Santa Barbara, CA, Nanoscope IIIa control system version 4.32r3).
3.
Results and discussion
Under the baseline conditions of pH ¼ 6.3 and 7 mM of ferrous iron, the AMD-FC produced a current density of 640 mA/m2 (290 mW/m2) as previously reported (Cheng et al., 2007). During the period of electricity production, a red precipitate was observed in the anode solution and on the anode surface, and the columbic efficiency was almost 100% as previously reported (Cheng et al., 2007), indicating that all the particles were likely precipitated through the same mechanism. The Mo¨ssbauer spectrum showed a paramagnetic doublet at room temperature, with parameters indicating presence of ferrihydrite or nano-sized goethite (Fig. 1) (Govaert et al., 1976). XRD patterns indicated the precipitate contained nano-sized goethite (a-FeOOH) (Penn et al., 2006), which is indiscernible from ferrihydrite on room-temperature Mo¨ssbauer spectroscopy because of the lack of magnetic ordering (Fig. 2) (Burleson and Penn, 2006). No other forms of ferric oxides were observed in the precipitate. AFM images showed the goethite particles were highly spherical, with a particle size of 402 20 nm under these conditions (Fig. 3). Goethite particles are usually synthesized through titration using either ferric or ferrous salt solutions. This approach produces particles with a needleshaped morphology (Kosmulski et al., 2004). In some cases, such as in the presence of polymer templates, some goethite particles can be formed that have a spherical shape (Sepulveda-Guzman et al., 2005). In our case, all synthesized particles showed highly spherical shapes across the complete size range obtained here under the studied conditions. This
% absorbance
in composition and can contain magnesium, aluminum, other metals, and organic materials, making resource recovery problematic (Kirby et al., 1999; Matlock et al., 2002). Additional processes such as selective precipitation (Matlock et al., 2002; Wei et al., 2005; Jenke and Diebold, 1983; Tabak et al., 2003) may be required to produce marketable products. An alternative approach for treating AMD was recently proposed based on the use of a new type of fuel cell architecture called an AMD fuel cell (AMD-FC) (Cheng et al., 2007). This device is based on a microbial fuel cell (MFC) that is being developed for power generation from the microbial oxidation of dissolved organic or inorganic matter. The use of an AMDFC has the advantages of generating electricity, while at the same time treating the AMD (Cheng et al., 2007). During treatment, ferrous iron is oxidized in the anode chamber under anoxic conditions, while at the cathode oxygen is reduced to water. Our previous studies showed that ferrous iron was completely removed through oxidation to insoluble Fe(III), and that it formed a precipitate on the bottom of the anode chamber (Cheng et al., 2007). This suggested to us that such a process might be useful for obtaining more controlled conditions for iron oxide production. We therefore examined the composition of the iron oxide formed in this process, and we show here that particle size can be controlled by varying the current density (through variation in the load), pH, and iron concentration.
-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 -12 -10 -8
-6
-4 -2 0 2 4 Source velocity (mm/s)
6
8
10 12
Fig. 1 e Mo¨ssbauer spectrum of the sample measured at room temperature (295 K), sample prepared at 0.05 M NaCl, 0.06 M NaHCO3, 0.007 M Fe2D, pH [ 6.3, and 1 kU external resistance.
305
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 0 3 e3 0 7
Particle size (nm)
A
1000 800 600 400 200 0 0.02
0.04
0.06
0.08
0.1
0.12
Current density (mA/cm2)
Fig. 2 e XRD pattern of the prepared sample under conditions of 0.05 M NaCl, 0.06 M NaHCO3, 0.007 M Fe2D, pH [ 6.3, and 1 kU external resistance.
relatively uniform shape was likely a result of the electrochemical precipitation process used here, compared to more commonly used titration processes. Further investigation of the mechanism of particle formation is needed. Particle sizes could be varied from 120 to 700 nm by changing the operating conditions of the AMD-FCs. Increasing the current from 0.04 to 0.12 mA/cm2 (by varying the external resistance from 2000 to 200 U) decreased the average particle sizes formed in the reactor from 700 50 to 450 60 nm (Fig. 4A), as a result of the increased current which produced a faster rate of nucleation (Kim et al., 2004). The particle size distribution also became narrower with increased current. These two results indicate that current both increased particle size was due to both the increased rate of ferrihydrite nucleation as well as coalescence of smaller particles. The smallest particles of 120 nm were formed at a pH ¼ 4, likely as a result of the slow oxidation rate and low cell voltage under these conditions (Cheng et al., 2007). The particle size distribution at pH ¼ 4 was narrow, with 91% of the particles having an equivalent diameter of 122 nm (Fig. 5B). Increasing the pH from 4 to 7.5 increased the particle sizes by a factor of 5.5, or from 120 20 to 750 50 nm (Fig. 5A). The particle size distribution also became broader with pH. At pH ¼ 7, for
Fig. 3 e AFM image of the particles (in white) (0.003 M Fe2D, pH [ 6.3 and 1 kU external resistance).
Frequency (%)
B
80 459
60 615
531 459
40
396 531 712
712 824
20
615
955 321
1280 1480
0
0.036
0.064
0.089
0.11
Current density (mA/cm2) Fig. 4 e Effect of current on (A) particle size, and (B) particle size distribution at selected current densities (numbers indicate particle sizes in nm) (0.05 M NaCl, 0.06 M NaHCO3, 0.007 M Fe2D, and pH [ 6.3).
example, 12.4% of the particles were 615 nm and increased in size to 955 nm (17.9%) (Fig. 5B). The Fe2þ concentration also significantly affected particle sizes (Fig. 6A). When the Fe2þ concentration increased from 50 to 1000 mg/L, the particle size linearly increased from 402 50 to 932 50 nm. The size distribution at the lower Fe2þ concentrations tended to be less dispersed, with 90% particles 396 nm in equivalent diameter at 50 mg/L. At the very highest Fe2þ concentration of 1000 mg/L, the particle size distribution was broader, with 14.3% of the particles 712 nm and 6.3% of the particles 1280 nm (Fig. 6B). These results show that it is possible to produce nanosized particles of ferrihydrite (transforming to goethite) from AMD while at the same time producing electrical current. The ferric oxide particles were highly spherical, had paramagnetic properties, and particle size could be controlled by varying the operational parameters. Using a low pH or low concentration of Fe2þ (both resulting in low currents due to the lower rate of reaction) produced the smallest particles (120 nm). Under such conditions the maximum voltage was 0.06 V. With pH ¼ 6.3 and 7 mM Fe2þ, a maximum power of 290 mW/m2 was produced (0.64 A/m2) (Cheng et al., 2007). These conditions produce ferric oxide particles rapidly with a size of 402 nm. Iron concentrations for AMD waters could be diluted using reactor effluent, and pH could be adjusted using various alkaline reagents, although the effect of chemical addition on the particles formed would need to be evaluated for specific AMD streams. Practical applications of the technology for AMD treatment will require further development of systems that efficiently
306
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 0 3 e3 0 7
A Particle size (nm)
1000 800 600 400 200 0
100
5
6 pH
7
8
122
80
Frequency (%)
B
4
60
295
40
825 712 459
20
615
531
255 615
531 712
342
106
0
4
5
development of MFCs for wastewater treatment (Logan, 2010a, 2010b). For example, although the cathode contained a Pt catalyst, non-precious metals catalysts such as CoTMPP can be used as a catalyst (Zhao et al., 2005; Cheng et al., 2006). Recently cathodes that have been developed using activated carbon and no metal catalysts have achieved nearly equal performance to cathodes used here (Zhang et al., 2009). Scalable architectures for MFCs have been developed based on using high surface area graphite brush anodes and tubular cathodes (Zuo et al., 2007; Logan et al., 2007), providing increased surface areas per volume of reactor and thus the need for much less treatment times (Logan, 2008). The use of these electrodes and materials in AMD-FC reactors should enable scalable and more efficient systems for iron recovery, power generation, and AMD treatment.
955
825
5.5 pH
615
6.3
4.
7
Fig. 5 e Effect of pH on (A) particle size and (B) and particle size distribution at selected pH conditions (numbers indicate particle sizes in nm) (0.05 M NaCl, 0.06 M NaHCO3, 0.007 M Fe2D, and 1 kU external resistance).
remove particles (for example through continuous centrifugation, that do not use precious metal electrodes, and that have electrodes with scalable architectures). Substantial progress is being made in these latter two topics based on the
Conclusions
It is possible to use AMD types of soluble iron solutions in fuel cell-based technologies to create spherical nano-particles of iron oxide (ferrihydrite) that are transformed to goethite (a-FeOOH) upon drying. Particle diameters were controlled to be in the range of 120e700 nm by varying the conditions in the fuel cell, such as current density (0.04e0.12 mA/cm2), pH (4e7.5), and initial Fe(II) concentration (50e1000 mg/L). These results provide a method to easily produce iron oxide particles that can be used in pigments and other products, although further research will be needed on many aspects of this process before these approaches could be successfully commercialized.
Particle size (nm)
A 1200 Acknowledgements
1000 800
This research was supported by NSF Grant BES-0401885.
600 400
references
200 0 0
0.005 2+
Fe
B
100
0.01
0.015
0.02
concentration (M)
396 615
Frequency (%)
80 295
60 342
40 20
825955 712 459
1110 712 1280
531
0
0.0008 0.004 0.011 Fe2+ concentration (M)
0.018
Fig. 6 e Effect of Fe2D concentration on (A) particle size and (B) particle size distribution at selected Fe2D concentrations (numbers indicate particle sizes in nm) (0.05 M NaCl, 0.06 M NaHCO3, pH [ 6.3, and 1 kU external resistance).
Burleson, D.J., Penn, R.L., 2006. Two-step growth of goethite from ferrihydrite. Langmuir 22, 402e409. Cheng, S., Dempsey, B.A., Logan, B.E., 2007. Electricity generation from synthetic acid-mine drainage (AMD) water using fuel cell technologies. Environmental Science & Technology 41 (23), 8149e8153. Cheng, S., Liu, H., Logan, B.E., 2006. Power densities using different cathode catalysts (Pt and CoTMPP) and polymer binders (Nafion and PTFE) in single chamber microbial fuel cells. Environmental Science & Technology 40, 364e369. Govaert, A., Dauwe, C., Plinke, P., De Grave, E., De Sitter, J., 1976. A classification of goethite minerals based on the MTssbauer behaviour. Journal De Physique 37, C6-825-C6-827. Jenke, D.R., Diebold, F.E., 1983. Recovery of valuable metals from acid-mine drainage by selective titration. Water Research 17, 1585e1590. Jang, J.-H., Dempsey, B.A., Catchen, G.L., Burgos, W.D., 2003. 2 Effects of Zn(II), Cu(II), Mn(II), NO 3 , or SO4 at pH 6.5 and 8.5 on transformation of hydrous ferric oxide (HFO) to more stable ferric oxides. Colloids Surfaces A 221, 55e68.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 0 3 e3 0 7
Kirby, C.S., Decker, S.M., Macander, N.K., 1999. Comparison of color, chemical and mineralogical compositions of mine drainage sediments to pigment. Environmental Geology 37, 243e254. Kairies, C.L., Capo, C.L., Watzlaf, G.R., 2005. Chemical and physical properties of iron hydroxide precipitates associated with passively treated coal mine drainage in the Bituminous regions of Pennsylvania and Maryland. Applied Geochemistry 20, 1445e1460. Kosmulski, M., Serge, D.V., Maczka, E., Rosenholm, J.B., 2004. Morphology of synthetic goethite particles. Journal of Colloid and Interface Science 271, 261e269. Kim, H., Subramanian, N.P., Popov, B.N., 2004. Preparation of PEM fuel cell electrodes using pulse electrodeposition. Journal of Power Sources 138, 14e24. Luengo, C., Brigante, M., Avena, M., 2007. Adsorption kinetics of phosphate and arsenate on goethite. A comparative study. Journal of Colloid and Interface Science 311, 354e360. Lagarec, K. and Rancourt, D. G. (1998) Recoil: Mo¨ssbauer Spectral Analysis Software for Windows, version 1.01998. Logan, B.E., 2010a. In: Rabaey, K., Angenent, L., Schro¨der, U., Keller, J. (Eds.), Bioelectrochemical Systems: From Extracellular Electron Transfer to Biotechnological Applications. IWA Publishing, London, pp. 184e204. Logan, B.E., 2010b. Scaling up microbial fuel cells and other bioelectrochemical systems. Applied Microbiology and Biotechnology. doi:10.1007/s00253-00009-02378-00259. Logan, B.E., Cheng, S., Watson, V., Estadt, G., 2007. Graphite fiber brush anodes for increased power production in air-cathode microbial fuel cells. Environmental Science & Technology 41 (9), 3341e3346. Logan, B.E., 2008. Microbial Fuel Cells. John Wiley & Sons, Inc., Hoboken, NJ. Matlock, M.M., Henke, K.R., Atwood, D.A., 2002. Effectiveness of commercial reagents for heavy metal removal from water
307
with new insights for future chelate designs. Journal of Hazardous Materials 92, 129e142. Penn, R.L., Erbs, J.J., Gulliver, D.M., 2006. Controlled growth of alpha-FeOOH nanorods by exploiting-oriented aggregation. Journal of Crystal Growth 293, 1e4. Sepulveda-Guzman, S., Perez-Camacho, O., Rodrıguez-Fernandez, O., Garcıa-Zamora, M., 2005. In situ preparation of magnetic nanocomposites of goethite in a styreneemaleimide copolymer template. J Magnetism Magnetic Materials 294, e47ee50. Tabak, H.H., Scharp, R., Burckle, J., Kawahara, F.K., Govind, R., 2003. Advances in biotreatment of acid mine drainage and biorecovery of metals: 1. Metal precipitation for recovery and recycle. Biodegradation 14, 423e436. Wei, X.C., Viadero, R.C., 2007. Synthesis of magnetite nanoparticles with ferric iron recovered from acid mine drainage: implications for environmental engineering. Colloids Surfaces A 294, 280e286. Wei, X.C., Viadero, R.C., Buzby, K.M., 2005. Recovery of iron and aluminum from acid mine drainage by selective precipitation. Environmental Engineering Science 22, 745e755. Zhao, H., Stanforth, R.S., 2001. Competitive adsorption of phosphate and arsenate on goethite. Environmental Science & Technology 35, 4753e4757. Zhao, F., Harnisch, F., Schro¨der, U., Scholz, F., Bogdanoff, P., Herrmann, I., 2005. Application of pyrolysed iron (II) phthalocyanine and CoTMPP based oxygen reduction catalysts as cathode materials in microbial fuel cells. Electrochemistry Communications 7, 1405e1410. Zhang, F., Cheng, S., Pant, D., Bogaert, G.V., Logan, B.E., 2009. Power generation using an activated carbon and metal mesh cathode in a microbial fuel cell. Electrochemistry Communications 11 (11), 2177e2179. Zuo, Y., Cheng, S., Call, D., Logan, B.E., 2007. Tubular membrane cathodes for scalable power generation in microbial fuel cells. Environmental Science & Technology 41 (9), 3347e3353.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 0 8 e3 1 4
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Comparative photochemical reactivity of spherical and tubular fullerene nanoparticles in water under ultraviolet (UV) irradiation So-Ryong Chae a,b, Yoshimasa Watanabe c, Mark R. Wiesner a,b,* a
Department of Civil and Environmental Engineering, Pratt School of Engineering, Duke University, Durham, NC 27708, USA Center for the Environmental Implications of NanoTechnology (CEINT), Duke University, Durham, NC 27708, USA c Center for Environmental Nano & Bio Engineering, Hokkaido University, Kita 13, Nishi 8, Kita-ku, Sapporo 060-8628, Japan b
article info
abstract
Article history:
Fullerene nanomaterials are finding an increasing number of applications in energy and
Received 8 May 2010
environmental technologies. However, substantial production and use of fullerenes will
Received in revised form
likely lead to environmental exposure with unknown consequences. In this study, aqueous
19 July 2010
suspensions of three types of fullerenes nanoparticles, C60 fullerene, single-wall (SW) and
Accepted 21 July 2010
multi-wall (MW) carbon nanotubes (CNT) were prepared by sonication and tested for
Available online 3 August 2010
reactive oxygen species (ROS) production and oxidation of benchmark organic compounds under ultraviolet (UV)-A irradiation. All three fullerenes formed colloidal aggregates in
Keywords:
water. SWCNTs showed the highest ROS production and 2-chlorophenol degradation fol-
Fullerene nanoparticles
lowed by MWCNT, and fullerene.
Carbon nanotubes
ª 2010 Elsevier Ltd. All rights reserved.
Reactive oxygen species Colloidal aggregates 2-Chlorophenol degradation
1.
Introduction
Fullerene nanomaterials (FNMs) are a class of molecules composed entirely of carbon that may exist in several geometries. The first of these molecules, C60 was discovered in 1985 and contains 60 carbons in the form of a hollow spherical cage consisting of 12 pentagonal and 20 hexagonal faces (Kroto et al., 1985). Non-spherical fullerenes have also been synthesized including cylinders (carbon nanotubes, CNT), lobed structures, bowls, and dendrimers to name a few. Fullerene (C60), CNT, graphite, and amorphous carbon are structurally different, which is reflected in their reactivity and oxidation resistance. In a single-wall (SW) CNT, all side carbon
atoms are present in hexagonal aromatic rings, except for those at the tips where atoms form pentagonal rings, which tend to be more reactive than regular graphene in graphite (Bhushan, 2007). Multi-wall (MW) CNTs consist of several nested coaxial carbon tubes and are known to be less reactive than the SWCNT, based on their conversion to CO2 when heated in the presence of oxygen (Brukh and Mitra, 2007). Fullerene-based nanomaterials (FNMs) are finding their way into a variety of applications, including cosmetics, flat panel displays, energy production, semiconductors, and medical treatments (Da Ros and Prato, 1999; Kamat et al., 2004; Saran et al., 2004). We have previously shown that aggregates of the poly-hydroxylated fullerene, fullerol (C60(OH)24) present
* Corresponding author. Department of Civil and Environmental Engineering, Pratt School of Engineering, Duke University, Durham, North Carolina 27708, USA. Tel.: þ1 919 660 5292; fax: þ1 919 660 5219. E-mail address:
[email protected] (M.R. Wiesner). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.07.067
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 0 8 e3 1 4
as a aqueous colloidal suspension, produce reactive oxygen species or ROS (primarily singlet oxygen, 1O2) via a type II pathway (Fig. 1) in the presence of light and oxygen (Hotze et al., 2008; Pickering and Wiesner, 2005). Fullerol-generated ROS is capable in turn of degrading 2-chlorophenol (2CP) (Chae et al., 2009), and inactivating viruses (Badireddy et al., 2007). These findings suggest important possible applications of fullerenes as well as cause for concern regarding their chemical and biological activity in aquatic environments. While there is a growing body of literature on the reactive properties of C60 and its derivatized variations such as fullerol, less consideration has been given in the literature to CNT aggregates. Here, we explore the photochemical reactivity of CNT aggregates and compare their properties to those of C60 and fullerol.
2.
Materials and methods
2.1.
Preparation of FNM suspensions
Fullerene and fullerol were purchased from MER (99.9%, Tucson, AZ). SWCNT and MWCNT were obtained from BuckyUSA (þ99.5%, Huston, TX). The initial diameter of the CNTs as reported by the supplier was 0.5e4 mm for the SWCNTs and about 50 mm for the MWCNTs. Aqueous suspensions of the FNMs (i.e., fullerene, SWCNT, and MWCNT) were prepared in 104 M NaHCO3 solution (pH ¼ 7.6) via sonication (35 W, ColeeParmer 8890 sonicator, Vernon Hills, IL) for 10 h without any organic solvent addition. During sonication, sound waves propagate through the liquid medium in alternating high and low pressure cycles at frequencies typically in the (20e40) kHz range. High pressures created due to cavitation produce a localized shock wave that releases high amounts of mechanical and thermal energy, breaking up aggregates and potentially inducing chemical modification to fullerenes
309
surfaces. Sonicated suspensions were filtered through a Nylon microfiltration (MF) membrane of nominal pore size 0.45 mm (Pall Life Science, East Hills, NY) to remove larger settleable particles, producing stable colloidal suspensions of fullerene aggregates (nC60) that varied in size based on sonication time as reported below. Fullerol and Rose Bengal (RB) were prepared by adding the powdered form of these materials to the NaHCO3 buffer solution without sonication (Chae et al., 2009). Fullerol also formed colloidal aggregates (nC60(OH)24).
2.2.
Characterizations of FNM suspensions
Total carbon (TC) concentration of the FNM suspensions was measured by a total organic carbon (TOC) analyzer (TOC5050A, Shimadzu, Columbia, MD). The hydrodynamic diameter of the aggregates was measured by dynamic light scattering (DLS) using an ALV/CGS-3 Compact Goniometer System (ALV-GmbH, Germany). High magnification images of aqueous FNMs were obtained by transmission electron microscope (TEM) (FEI Tecnai G2 Twin, Hillsboro, OR). Image analysis of TEM images was performed by using an Image-Pro version 4.5 (Media Cybermetics, Inc. Bethesda, MD). Elemental analysis of CNT suspensions was done using 10 mL of each suspension, filtered through an aluminum membrane (Anodisc) of nominal pore size 20 nm and diameter of 47 mm (Whatman, Florham Park, NJ) and then examining the membrane with the deposited FNMs by SEM (scanning electron microscopy)eEDX (energy dispersive X-ray spectroscopy) (Superscan SSX-550, Shimadzu Co., Kyoto, Japan).
2.3.
Experimental conditions
ROS production and photo-degradation of organic compounds of the FNM suspensions was studied in a glass beaker (90, O. D. 115 mm, H) with a water jacket connected to a water
Fig. 1 e Schematic diagram of a ROS production mechanism by fullerene aggregates.
310
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 0 8 e3 1 4
circulator for temperature control as described previously (Chae et al., 2009). The light sources was two 15 W fluorescent ultraviolet bulbs (Philips TLD 15W/08) in an UV/Cryo chamber (Electron Microscopy Science, Hatfield, PA). These bulbs had an output spectrum ranging from 310 to 400 nm and a total irradiance of 24.1 W/m2 with a peak at 365 nm (UV-A). The effect of UV-A irradiation on ROS production of the FNM suspensions depends on many factors such as applied dosage, intensity, and reaction dimension. A well-characterized source of UV-A light was used in this study to provide reproducible conditions for illumination in this range and to minimize transfer-limitations of light through the suspension. A volume of 10 mL of each FNM suspension was tested with a stir bar on a magnetic stirrer (Auto Mixer, Fisher Scientific, Hampton, NH) at about 100 rpm. Water samples were collected from the suspension every 5 min for 30 min for further analyses. All experiments were performed in triplicate. Student’s t-test was used to assess the significance of the results employing a 95% confidence interval.
2.4.
Detection of reactive oxygen species (ROS)
aggregates using a high-performance liquid chromatograph (HPLC) (ProStar, Varian, Palo Alto, CA) equipped with a reverse phase column (Ultra aqueous C18, 5 mm, 150 4.6 mm, RESTEK, Bellefonte, PA) and a photodiode array (PDA) detector with a detection wavelength of 210 nm as described previously (Chae et al., 2009). Briefly, the flow rate of the mobile phase consisting of 25 mM KH2PO4 and HPLC-grade acetonitrile was maintained at 1 mL/min. The composition of the mobile phase (KH2PO4/acetonitrile) was 80:20 for the first 2 min and was then adjusted at rate of 10%/min to 50:50 which was maintained for 7 min. The pH in the mobile phase was adjusted to a value of 2.5 using 6 N HCl solution. Under these conditions, the retention times of SA and 2CP were 8.5 and 9.0 min, respectively. Ethanol concentrations were determined by gas chromatography (GC) with a flame ionization detector (GCMS-QP5050, Shimadzu, Columbia, MD) equipped with a fused silica capillary column (DB-1, 5 mm, 30 m 0.32 mm, J&W Scientific, Folsom, CA) using diethyl ether as an internal standard. The oven temperature was ramped from 40 to 160 C at a rate of 10 C/min. The injector temperature and detector temperature were 250 and 300 C, respectively. The flow rate of the carrier gas (He) was 16.2 mL/min. Under these conditions, the
Singlet Oxygen Sensor Green (SOSG) was used to detect singlet oxygen (1O2). SOSG was first pre-diluted in 33 mL methanol and ultrapure water to 165 mM as recommended by the manufacturer and then diluted ten-fold to a final concentration of 16.5 mM prior to measurement. Test suspensions were placed in a 25 mL Petri dish with approximately 20 cm2 surface area. A 1 mL aliquot was taken every 5 min for 30 min in the dark. The fluorescence units (FSU, Ex/Em ¼ 505/525 nm) from this dark measurement were read as background (Modulus Single Tube 9200, Turner Biosystems, Sunnyvale, CA) and subtracted from readings taken at every 5 min for 30 min once the UV lamps were switched on. XTT (2,3-bis(2-methoxy-4-nitro-5sulfophenyl)-2H-tetrazolium-5-carboxanilide) reduction was employed to measure the production of superoxide. The reduction of XTT results in an increase in optical density at 470 nm that can be used to quantify the relative amount of superoxide present (Ukeda et al., 1997). The color change was measured by absorbance at 470 nm using an UV/Visible spectrophotometer (U-2000, Hitachi, Schaumburg, IL). The concentration of superoxide was determined by comparing XTT reduction with and without a quencher for superoxide, superoxide dismutase (SOD), to account for non-superoxide related XTT reactions and possible confirm the role of superoxide in any observed degradation. SOD-containing samples also served to eliminate the influence of background absorbance of suspensions.
2.5. study
Analysis of probe organic compounds and kinetic
Organic compounds were selected for their sensitivity to individual species of reactive oxygen and included salicylic acid (SA) which is known to be degraded by the hydroxyl radical (OH) (Karnik et al., 2007); 2-chlorophenol (2CP) which is degraded by singlet oxygen (1O2) (Gryglik et al., 2007) and, ethanol which is oxidized by superoxide (O 2 ) (Hirakawa et al., 2007). Concentrations of the former two compounds were monitored over time in aqueous suspensions of fullerene
Fig. 2 e Behaviors of total carbon concentration (A) and hydrodynamic diameter (dD) measured by DLS (B) of fullerene, SWCNT, MWCNT suspensions after sonication followed by the MF membrane filtration (0.45 mm), and fullerol in 10L4 M NaHCO3 buffer solution at 25 C (pH [ 7.6).
311
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 0 8 e3 1 4
retention times of ethanol and diethyl ether were 2.7 and 3.8 min, respectively. Kinetics of photosensitized degradation of phenolic compounds has been widely studied. Assuming that the probe substance (for example, 2CP) reacts primarily with singlet oxygen (1O2) and negligibly with the excited sensitizers, a simplified reaction scheme can be used to describe this system (Chae et al., 2009), which was originally described for degradation of phenolic compounds by a photosensitizing dye such as Rose Bengal (Tratnyek and Holgne, 1991): d½2CP ¼ kdp ½2CP dt
(1)
where kdp is the pseudo-first-order rate constant (s1) of chemical degradation by direct photolysis from the experimental results. The net pseudo-first-order rate constant, knet (s1) is calculated by subtracting kdp from kobs. knet ¼ kobs kdp
3.
(2)
Results and discussion
3.1. Characteristics of fullerene suspensions with various sonication times Size (hydrodynamic diameter) of fullerene aggregates (denoted nC60 where n is the number of primary fullerene units in
the aggregate) and total carbon (TC) concentration of FNM suspensions were characterized over a range of sonication times. The size of fullerene aggregates measured by DLS (dD) and obtained from TEM (dT) are shown in Fig. 2 and Table 1, respectively. The TC concentration of the FNM suspensions filtered through the 0.45 mm membrane was about 2 ppm after 1 h sonication and increased to 2.8 ppm (fullerene, C60), 2.5 ppm (SWCNT), and 2.3 ppm (MWCNT) after 10 h sonication (Fig. 2(A)). In contrast, as the sonication time increased from 1 to 10 h, the hydrodynamic diameter (dD) of fullerene, SWCTN, and MWCNT aggregates decreased from 214 56, 383 97, and 409 103 nm to 104 31, 215 47, 341 66 nm, respectively (Fig. 2(B)). Decreases in mean diameter with sonication time are consistent with cumulative break-up of aggregates due to power input and possibly hydroxylation of the fullerene surface that increases affinity of these otherwise hydrophobic materials with the aqueous phase. Consistent with a larger nested coaxial tube structure (Brukh and Mitra, 2007) the MWCNT formed larger aggregates at a given level of sonication than the SWCNT. The TC concentration and size of the fullerol suspension were 2.3 ppm and 437 nm without sonication. TEM images of the various fullerene suspensions qualitatively confirm trends in aggregate diameter (dD) with sonication time (Fig. 3). The mean diameters as calculated from TEM image analysis (dT) were slightly smaller than the hydrodynamic diameters (dD) obtained by DLS but the difference was statistically not distinguishable.
Table 1 e Relationship between structural properties and ROS production of various fullerene nanoparticles. Item
nC60
nSWCNT
nMWCNT
104 31 91.1 50.6 383.4 214.8
215 47 156.6 20.3 861.5 169.9
341 66 214.3 98.6 687.9 92.9
0.005 0.001
0.021 0.004
0.011 0.003
Before sonication (Micro-scale)
After 10 h sonication (nano-scale)
a
dD, mean diameter (nm) by DLS (n ¼ 3) dT, mean diameter (nm) from TEM images (n ¼ 100) a Perimeter (nm) per aggregates from TEM images (n ¼ 100) Projected area (mm2) per aggregates from TEM images (n ¼ 100) a
Relative ROS production
a dD is hydrodynamic diameter of fullerene aggregates measured by DLS (n ¼ 3), dT is the mean diameter of randomly sampled aggregates from TEM images, Perimeter is a length of the object’s outline.
312
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 0 8 e3 1 4
Fig. 3 e TEM images of fullerene, SWCNT, MWCNT aggregates after sonication followed by the membrane filtration (0.45 mm), and fullerol without sonication in 10L4 M NaHCO3 buffer solution (pH [ 7.6).
3.2.
ROS production from fullerene suspensions
ROS production under UV-A irradiation was measured in suspensions that had been sonicated for 10 h. Superoxide production by SWCNT as measured by XTT formazan production was slightly higher than that of the other fullerenes, but overall, superoxide production was negligible (Fig. 4). In contrast, significant singlet oxygen production was produced by most of the fullerenes with differences observed from one fullerene to another. Singlet oxygen production by the fullerenes was benchmarked against RB (1 mM), a well-known singlet oxygen producer (Hotze et al., 2008; Pickering and Wiesner, 2005) that showed the highest extent of singlet oxygen generation
over a 60 min period (about 5400 SOSG fluorescence intensity unit (a.u.)). As shown in Fig. 4, the cumulative singlet oxygen production over this period by fullerol, SWCNT, MWCNT, and C60 measured by fluorescence intensity from SOSG was 31, 16, 14, and 9% respectively of that obtained by RB.
3.3. Degradation of organic compound by the photosensitized fullerene aggregates Among three organic compounds, only the 2CP showed significant degradation underscoring the specificity of the fullerene nanomaterials in producing singlet oxygen. Approximately 15% of initial 2CP concentration (25 mM) was
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 0 8 e3 1 4
Fig. 4 e ROS production of the photosensitized FNM suspensions (TC [ 2.0 mg/L) and Rose Bengal (1 mM) in 10L4 M NaHCO3 buffer solution at 25 C (pH [ 7.6).
degraded by direct photolysis without fullerenes (UV control) following 1 h of UV irradiation. With 2 mg TC/L of fullerenes, approximately 2CP was reduced to 20, 25, 28, and 33%, of the initial concentration by the C60 MWCNT, SWCNT, and fullerol respectively compared with 45% by the 1 mM solution of RB (Fig. 5). Assuming the degradation of organic compounds to be pseudo-first-order, the concentration data in Fig. 5 can be expressed as rate constants for each reaction (Fig. 6). The net degradation rate constant (i.e., net oxidation by the photosensitized FNMs except for direct photolysis by UV irradiation) of SA and ethanol during the reaction was essentially zero for these FNMs. The degradation rate constant for 2CP was highest for RB followed by fullerol, SWCNT, MWCNT, and C60. These results follow the same trends observed for ROS generation by the photosensitized FNMs. Mirroring these results, Brukh and Mitra (2007) reported that SWCNTs were more reactive with oxygen than were MWCNTs followed by amorphous carbon, and graphite. Reactivity in the current study does not appear to follow a trend of aggregate size. However measurements of mean aggregate size as determined by DLS do not necessarily
Fig. 5 e Degradation of 2-chlorophenol (initial concentration [ 25 mM) by the photosensitized FNM suspensions (TC [ 2.0 mg/L) and Rose Bengal (1 mM) in 10L4 M NaHCO3 buffer solution at 25 C (pH [ 7.6).
313
Fig. 6 e Net decay rate constants of organic compounds (initial concentration [ 25 mM) by the photosensitized FNMs (TC [ 2.0 mg/L) and Rose Bengal (1 mM) in 10L4 M NaHCO3 buffer solution at 25 C (pH [ 7.6).
reflect heterodispersivity in the sample and may indeed mask smaller aggregates. As shown in Table 1, as projected area and perimeter (i.e., length of the object’s outline) of the fullerene aggregates increased, ROS production also increased. It indicated that the relative compactness of carbon cages might play a role, reflecting for example the lower reactivity of the MWCNTs compared with that of the SWCNTs. Additional factors that are not accounted for here include the adsorption onto the fullerene aggregates’ surface and the role of residual metal catalysts on the CNTs. In particular, Fe and Ni were observed on the CNT materials by SEMeEDX analysis (data not shown). But it is not clear that those metal compounds played a role in ROS production under UV irradiation and there is a clear need for more detailed study.
4.
Conclusions
Singlet oxygen production following photosensitization is the primary mode of oxidation of chemical compounds by both carbon nanotube and caged C60 fullerenes. For similar carbon concentrations, suspensions of CNTs are observed to be less photo-reactive than fullerol and considerably more photoreactive than aggregates of underivatized C60. The relative reactivity of these suspensions appears to be related to the effective surface area of fullerenes in the suspension and is consistent with our previous theoretical work showing fullerene reactivity related to the structural configuration of carbon cages in an aggregate (Hotze et al., 2010). By this theory, asymmetry of the CNTs may impede the formation of dense aggregates while SWCNTs are more reactive than MWCNTs due to the fraction of carbon that is “buried” or selfquenched within the CNT.
Acknowledgements This material is based upon work supported by the National Science Foundation (NSF) and the Environmental Protection
314
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 0 8 e3 1 4
Agency (EPA) under NSF Cooperative Agreement EF-0830093, Center for the Environmental Implications of NanoTechnology (CEINT). Any opinions, findings, conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the NSF or the EPA. This work has not been subjected to EPA review and no official endorsement should be inferred.
references
Badireddy, A.R., Hotze, E.M., Chellam, S., Alvarez, P., Wiesner, M.R., 2007. Inactivation of bacteriophages via photosensitization of fullerol nanoparticles. Environmental Science & Technology 41 (18), 6627e6632. Bhushan, B., 2007. Handbook of Nanotechnology. Springer. Brukh, R., Mitra, S., 2007. Kinetics of carbon nanotube oxidation. Journal of Materials Chemistry 17 (7), 619e623. Chae, S.R., Hotze, E.M., Wiesner, M.R., 2009. Evaluation of the oxidation of organic compounds by aqueous suspensions of photosensitized hydroxylated-C-60 fullerene aggregates. Environmental Science & Technology 43 (16), 6208e6213. Da Ros, T., Prato, M., 1999. Medicinal chemistry with fullerenes and fullerene derivatives. Chemical Communications 8, 663e669. Gryglik, D., Miller, J.S., Ledakowicz, S., 2007. Singlet molecular oxygen application for 2-chlorophenol removal. Journal of Hazardous Materials 146 (3), 502e507. Hirakawa, T., Daimon, T., Kitazawa, M., Ohguri, N., Koga, C., Negishi, N., Matsuzawa, S., Nosaka, Y., 2007. An approach to estimating photocatalytic activity of TiO2 suspension by monitoring dissolved oxygen and superoxide ion on decomposing organic compounds. Journal of Photochemistry and Photobiology A-Chemistry 190 (1), 58e68.
Hotze, E.M., Bottero, J.-Y., Wiesner, M.R., 2010. Theoretical framework for nanoparticle reactivity as a function of aggregation state. Langmuir 26 (13), 11170e11175. Hotze, E.M., Labille, J., Alvarez, P., Wiesner, M.R., 2008. Mechanisms of photochemistry and reactive oxygen production by fullerene suspensions in water. Environmental Science & Technology 42 (11), 4175e4180. Kamat, P.V., Haria, M., Hotchandani, S., 2004. C60 cluster as an electron shuttle in a Ru(II)-polypyridyl sensitizer-based photochemical solar cell. Journal of Physical Chemistry B 108 (17), 5166e5170. Karnik, B.S., Davies, S.H., Baumann, M.J., Masten, S.J., 2007. Use of salicylic acid as a model compound to investigate hydroxyl radical reaction in an ozonation-membrane filtration hybrid process. Environmental Engineering Science 24 (6), 852e860. Kroto, H.W., Heath, J.R., Obrien, S.C., Curl, R.F., Smalley, R.E., 1985. C-60 e Buckminsterfullerene. Nature 318 (6042), 162e163. Pickering, K.D., Wiesner, M.R., 2005. Fullerol-sensitized production of reactive oxygen species in aqueous solution. Environmental Science & Technology 39 (5), 1359e1365. Saran, N., Parikh, K., Suh, D.S., Mun˜oz, E., Kolla, H., Manohar, S.K., 2004. Fabrication and characterization of thin films of single-walled carbon nanotube bundles on flexible plastic substrates. Journal of the American Chemical Society 126 (14), 4462e4463. Tratnyek, P.G., Holgne, J., 1991. Oxidation of substituted phenols in the environment e a qsar analysis of rate constants for reaction with singlet oxygen. Environmental Science & Technology 25 (9), 1596e1604. Ukeda, H., Maeda, S., Ishii, T., Sawamura, M., 1997. Spectrophotometric assay for superoxide dismutase based on tetrazolium salt 3’-{1-[(phenylamino)-carbonyl]-3,4tetrazolium}-bis(4-methoxy-6-nitro)benzenesulfonic acid hydrate reduction by xanthineexanthine oxidase. Analytical Biochemistry 251 (2), 206e209.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 1 5 e3 2 7
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Dissolved organic matter: Precautions for the study of hydrophilic substances using XAD resins Jerome Labanowski a,*, Genevie`ve Feuillade b a
Laboratoire de Chimie et Microbiologie de l’Eau UMR CNRS 6008, Universite´ de Poitiers e ESIP, 40, avenue du recteur Pineau, 86022 Poitiers Cedex, France b Groupement de Recherche Eau Sol Environnement e ENSIL 16, rue Atlantis Parc Ester, 87068 Limoges Cedex, France
article info
abstract
Article history:
This study concerns the possible changes in the repartition and the molecular character-
Received 1 April 2010
istics of hydrophilic substances (HyS) isolated by XAD resins from the same source of
Received in revised form
organic matter as a function of the distribution coefficient k0 and the DOM concentration.
5 July 2010
We tested that on two different sources of organic matter (a surface water and a landfill
Accepted 22 July 2010
leachate). Breakthrough curves column experiments highlighted a modification of the
Available online 3 August 2010
repartition between hydrophilic and humic substances according to the k0 value applied. But, we find that the composition of HyS is significantly modified between k0 ¼ 50 and 100.
Keywords:
Our observations tend to suggest a higher contribution of humic-like matter (high-
Hydrophilic substances
molecular weight aromatic compounds) with an increase of the k0 value. This results in
Organic matter
a shift of fluorescence bands to longer wavelengths and changing patterns of the SEC
XAD fractionation
profiles and molecular fingerprints performed by flash pyrolysis. Our results show that
Distribution coefficient k0
DOM concentration also influences the composition of HyS while little effect is observed on their quantification at k0 ¼ 50 or 100. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Dissolved organic matter (DOM) has been shown to be one of the most important elements of terrestrial and aquatic environments. That is why for more than one century, scientists try to determine its origin, its formation and its implication in environmental processes (i.e. biogeochemical cycles, fate and transport of pollutants). Such a work is difficult because the relative size, shape, and composition of DOM are very random. DOM can vary greatly, depending on its origin, transformation mode, age, and existing environment, thus its bio-physico-chemical functions and properties vary with different environments (Senesi et al., 2006). Nevertheless, scientists have succeeded in showing that humic substances (HS) are one of the major constituents of DOM (Thurman, 1985). HS (i.e. humic and fulvic acids) are a complex mixture
of aromatic and aliphatic compounds derived from soil and aquatic biota including plants, animals and microorganisms (Leenheer et al., 2000). They have been extensively studied because of their role in humification processes and their reactivity with trace metals (Koopal et al., 2001) or organic pollutants (Schwarzenbach et al., 1993). Recently, other components of DOM have also received an increasing attention: hydrophilic substances (HyS). Industrials, engineers, and scientists working in the water treatment are very interested in these compounds because of their difficult elimination (Sharp et al., 2006) and their potential to form disinfection byproducts (Chang et al., 2001). In environmental sciences, HyS have also aroused a lot of interest because of their implication in biogeochemical cycles (Fearing et al., 2004; Kaushal and Lewis, 2005) and their reactivity with pollutants (Croue´ et al., 2003). Hydrophilic substances include three main types of
* Corresponding author. Tel.: þ33 5 49 45 38 46; fax: þ33 5 49 45 37 68. E-mail address:
[email protected] (J. Labanowski). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.07.071
316
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 1 5 e3 2 7
compounds: non-humic compounds corresponding to simple molecules (e.g. carbohydrates, amino acids.), products of microbial activity, and decaying substances from living organisms (microorganisms, plants, animals) (Thurman, 1985; Croue´, 2004). Despite the large number of studies that have dealt with this type of substances, the level of knowledge about hydrophilic substances is still weak compared to HS. Hydrophilic substances are difficult to study because of the absence of specific methodologies developed and optimized to isolate them in complex samples (i.e. environmental or anthropogenic samples). Much of this difficulty stems from their high affinity for the aqueous phase. Hydrophilic substances are generally obtained after extraction of HS in the sample. Many isolation procedures and quantification methods have been developed to study HS. Two main techniques, resin adsorption and size separation, are commonly used (Leenheer and Croue´, 2003; Belzile and Guo, 2006; Zhao et al., 2006). Among them the procedure using the XAD-8 resin was the first used to define operationally HS. Leenheer (1981), and Thurman and Malcolm (1981) defined the HS as the part of DOM absorbable on a XAD-8 resin at pH 2 and desorbed by NaOH washing. Hydrophilic substances correspond to the non-absorbable part of DOM (n.b. it should be noted here that this non-absorbable part is also called non-humic substances depending on works). These two major references of the humic chemistry, although analogous, present an important difference. They use a different column-capacity factor (k0 ), i.e. the volume of sample passing through a given volume of resin is different. For Thurman and Malcolm (1981), the volume of resin was chosen such that a solute with a k0 factor of 100 is 50% retained by the resin whereas Leenheer (1981) proposed a k0 of 50. Different values of k0 are also reported for more complex procedures involving XAD-8/XAD4 resins in tandem that are generally used for the fractionation of DOM in three fractions: hydrophobic substances, transphilic and hydrophilic substances from DOM (Aiken et al., 1992; Croue´ et al., 1993). Another remarkable feature of XAD procedures is their flexibility. They can be used, without any adjustment, on samples of very different composition. Originally developed for natural waters containing low DOM content (dissolved organic carbon (DOC) ranges from 1 to 15 mg C L1 (Thurman, 1985)), XAD procedures were also successfully applied to DOM-rich samples (DOC 40 mg C L1, e.g. Suwannee river, wastewater) (Serkiz and Perdue, 1990; Imai et al., 2002; Kang et al., 2002). Despite larger DOC concentrations, no accurate information about a modification of the k0 coefficient is mentioned. An explanation may be that the theoretical equations of the fractionation by XAD resin do not consider the DOM concentration for the determination of the volume of sample. However, Leenheer (1981), studying an oil-shale retort wastewater, decreased the k0 value to about 4 in the case of preparative fractionation of HS. This change of k0 value caused a large shift in DOC to the hydrophobic portion of the fractionations. The works of Gadmar et al. (2005) also suggest that the DOC concentration affects the fractionation by XAD resins. Although the International Humic Substances Society (IHSS) proposes an official recommendation (i.e. IHSS recommends to use the Thurman and Malcolm procedure) for the fractionation of DOM (IHSS, 2007), the absence of an official standardized
procedure has given freedoms to scientists. This has contributed to the use of slightly different procedures. However, this variability of procedures for the isolation of HS raises the question to know i) if the isolation HyS may be affected; and ii) if the body of available information in the non-humic substances chemistry, although substantial, may be considered comparable between studies using different k0 value. The aim of the present work was to assess the influence of the distribution coefficient k0 and the DOM concentration on the proportion and the molecular characteristics of hydrophilic substances isolated by XAD resins from the same source of organic matter. We tested that on two different sources of organic matter (a surface water and a landfill leachate). First, we determined the proportions of hydrophilic and humic substances at various k0 values, in particular those reported by Leenheer (1981) and Thurman and Malcolm (1981), from breakthrough curves of column experiments. Second, we determined the effect of DOM concentration on this repartition from the comparison of the same sample concentrated or diluted to various DOM concentrations. Then the hydrophilic substances, isolated in all these conditions, were characterized by a combination of synchronous fluorescence spectroscopy, size exclusion chromatography with UV and fluorescence detection and flash pyrolysis measurements. Finally, observations relative to quantitative or qualitative changes occurring in HyS were discussed.
2.
Material and methods
2.1.
Samples collection and preparation
This work was carried out with different types of sample collected in the Limousin area (France). We selected our samples for their different origin and organic matter content. The water sample was collected from a surface freshwater: the river Glane, and the leachate sample was sampled in a former municipal solid wastes landfill: the landfill of Crezin. The production of leachate is due to rainfall infiltration and waste degradation. A previous study of Labanowski and Feuillade (2009) has shown that the composition of the organic matter is different compared to natural waters or soils. Approximately 60 L of each sample were collected in 30 L polypropylene containers and immediately filtered (0.45 mm, cellulose nitrate filters). Samples were then stored at 4 C before preparation and use. Polypropylene containers were washed several times with acid solutions (HCl, 5 M) and oxidant solutions (NaClO, 1 M at pH ¼ 10) to avoid release of organic material. For each sample, pH (WTW, pH 197) and conductivity (WTW, LF 538) measurements were made in situ. Three parameters were measured to evaluate the content of DOM and its degradability. Dissolved Organic Carbon (DOC) content was estimated with a carbon analyser (Phoenix 8000, Dohrmann). The precision of DOC measurements was 0.2 mg C L1. Chemical Oxygen Demand (COD) and Biological Oxygen Demand after 5 days (BOD5) were determined according to the French norms AFNOR NFT-90101 and NFT-90103 (AFNOR, 1997), respectively. Three replicates were performed for each analyse. Specific UV absorbance (SUVA ¼ UV absorbance at 254 nm O DOC content) (L mg1 cm) was calculated to evaluate
317
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 1 5 e3 2 7
3 Vpore CDOM þ Vpore k0 CDOM ¼ 3 þ k0 Vpore CDOM
the aromaticity of organic matter. For that, UV absorbance was measured at 254 nm by a Safas MC2 spectrophotometer with 1 cm-long quartz cells. The preparation of samples consisted in modifying their DOM concentration without changing the proportions of HyS and HS. For this, many dilutions were performed on the leachate whereas the surface water was concentrated. Table 1 presents the DOC concentrations of the raw and prepared samples. Data correspond to mean values of triplicate measures. Landfill leachate sample was diluted with a sodium nitrate solution having the same ionic strength (i.e. 0.03 M). Ionic strength was calculated from the electrical conductivity of the leachate by the MarioneBabcock equation (Sposito, 1989). The surface water was concentrated on a reverse osmosis unit OI2.5 m2 (Techniques Industrielles Applique´es, France). This reverse osmosis unit uses a spiral polysulfones-polyamides membrane (SW 30-2540, Filmtec). Kilduff et al. (2004) showed that the concentration of natural samples by reverse osmosis had no effect on DOM adsorption on XAD-8 resin. However, we observed about 3% 2% of dissolved organic carbon mass losses on the reverse-osmosis membranes. Kitis et al. (2001) and Lee et al. (2006) ascribed such losses to the sorption of hydrophilic compounds.
where Vpore is the volume of sample in the porosity of the resin. For monitoring this type of experiments, breakthrough curves are generally plotted to follow the retention of OM. At the breakthrough point (equation (4)): VDOM CDOM ¼ 3 þ k0 Vpore CDOM
XAD fractionation
2.2.1.
Theoretical principles
(4)
where VDOM is the volume of sample passed into the column The distribution coefficient is generally much higher than the porosity coefficient (3). Hence, it can be established from equation (4) that k0 is related to the volume of sample responsible of the breakthrough: k0 ¼ VDOM =Vads
(5)
From these theoretical principles, the final relation between the k0 coefficient and the volume of sample to study was given by Leenheer (1981) and Thurman and Malcolm (1981) (equation (6)). This relation is only applicable to the fractionation of DOM by a column of XAD resin. VE ¼ 2VO 1 þ k0
2.2.
(3)
(6)
where VE is the volume of sample that is applied to the column, VO is the void volume of the column (about 60% of the bed volume for XAD resin).
The fractionation of DOM by XAD resins is based on equilibrium between the organic matter (OM) in solution and the resin (R) (equation (1)): K
R þ OM 4 R OM
2.2.2.
(1)
where K is the distribution coefficient defined as the ratio of the amount of DOM retained by gram of resin (Cads) on the amount of DOM (CDOM). The retention of OM by XAD resins involves surface adsorption mechanisms. Consequently the amount of OM adsorbed must be corrected by the porosity (3) (equation (2)): C0ads ¼ Cads 3 ¼ K 3 Cinfluent ¼ k0 Cinfluent
Experimental procedure
The XAD resin procedure provides operationally the fractionation of DOM into humic substances and hydrophilic substances by sorption of HS on the Supelite DAX-8 resin (Supelco). DAX-8 is the substitute for XAD-8 resin (Farnworth, 1995). For this, samples were acidified with HCl to pH 2.0, and then pumped through a column packed with the resin. Our experimental set-up was constituted of a stainless steel column (length: 4.65 cm; diameter: 1.35 cm) and a Gilson Minipuls2 peristaltic pump. The volume of resin inside the column is of 6.64 mL. The column effluent was collected at 5 mL intervals. For the present work, either a given volume of sample (i.e. corresponding to a chosen k0 value) was passed through the column, or experimentations are continued until the saturation of the resin. The retention of OM was followed by breakthrough curves representing the DOC concentration vs. volume of sample passed through the column per volume of resin (Vp/Vr). The proportion of HyS vs. HS was determined by integrating areas over and under the curves, respectively
(2)
where K 3 is defined as the capacity factor (noted k0 ). It should be noted that the fractionation experiments are generally performed with a column. The amount of OM inside the column corresponds to the amount of OM both remaining in solution (i.e. within the porosity) and adsorbed on the resin (equation (3)):
Table 1 e Sample preparation steps and dissolved organic carbon content of the studied samples. Samples collected
Preparation
Samples prepared
Dissolved organic carbon concentration (mg C L1) (0.3 mg C L1)
Landfill leachate
The leachate was diluted with a NaNO3 solution at 0.03 M in ionic strength
Surface freshwater
The water was concentrated by reverse osmosis OI2.5 m2
-
Untreated leachate Leachate diluted twice times Leachate diluted ten times Untreated freshwater Freshwater concentrated ten times
115 1 54 0.5 11 0.2 8 0.2 81 1
318
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 1 5 e3 2 7
(Matlab v6.0). DOC analyses were performed with a Phoenix 8000 carbon analyser (Dohrmann). Results from tests using this analyser are generally accepted to have an error of 0.2 mg C L1. Breakthrough curves presented in this paper correspond to the mean values of triplicate. For studying the characteristics of HyS, the entire sample eluted from column was collected. A separate experiment was conducted for each k0 studied but each experiment was performed in triplicate runs.
2.3.
Characterization of organic matter
In the present study, various analytical techniques were performed to compare and to characterize HyS. Synchronous fluorescence spectroscopy (SF) gives insight into fluorescence properties of OM whereas size exclusion chromatography (SEC) provides a molecular size characterization. Flash pyrolysis-gas chromatography/mass spectrometry (Py-GC/MS) is a method of assessing the molecular fingerprint of a sample. Theses techniques are widely used to study DOM in natural and altered aquatic environments (Peuravuori et al., 2002; Fuentes et al., 2006). They were preferred because of their ability to trace both humic and hydrophilic substances, and hence to appreciate the contamination of one by the other.
2.3.1.
Synchronous fluorescence spectroscopy
Aliquots of hydrophilic substances were studied by synchronous fluorescence. These samples were diluted, if necessary, in 0.05 M NaHCO3 until the DOC concentration reached about 10 mg C L1. Synchronous fluorescence spectra were recorded in these solutions using a PerkineElmer LS 50B luminescence spectrometer applying a wavelength range 280e550 nm with a scan speed of 2 nm s1. This instrumental technique involves the simultaneous scanning of the excitation and emission monochromators, and keeps a constant wavelength between the monochromators (Dl of 18 nm) (Miano and Senesi, 1992). The light-path of the cuvette used was 10 10 mm. Noise was filtered from raw fluorescence readings by smoothing, baseline subtraction. All spectra were normalized by the DOC content of the sample. For determining the accuracy of the synchronous fluorescence measurement, ten aliquots of the same sample were analyzed. Area of curves was used to estimate the accuracy. The accuracy (P0.95) of the spectra was about 3% (including dilution in NaHCO3, synchronous fluorescence measurement and normalization errors) for both landfill leachate and surface freshwater samples.
Whatman) before HPSEC analyses. A double detection was used because UV and fluorescence detections complement each other. The UV detection monitor UV-absorbing organic materials such as humic substances (Nagao et al., 2003) while the fluorescence detection, with our excitation/emission settings, monitor microbial-like materials (Her et al., 2003). Chromatograms presented in the following section correspond to the mean of two analyses by sample. On these chromatograms, different size fractions were distinguished to ease the analysis. The proportion of these fractions (i.e. the area of the fraction to the total area under the chromatogram) was determined with the Millennium 32 software (version 3.05.01, Waters). From these results, percentage distribution graphs were plotted. In the results and discussion section, we applied the following nomenclature to differentiate the fractions: F and F* differentiate between fractions of the surface water and the landfill leachate, respectively; FUV and FFLUO differentiate between the fractions defined by SEC-UV and SEC-fluorescence detections, respectively. It should be noted that in SEC, the first eluted fractions correspond to the highest molecular weight compounds of the sample while the last eluted fractions correspond to the lowest molecular weight compounds.
2.3.3. Flash pyrolysis-gas chromatography/mass spectrometry The experimental procedure used in this study is briefly described below. A quartz reaction tube was packed with 1 mg of dry sample and short quartz wool. The tube is then placed into the platinum filament of the Pyroprobe 2000 pyrolyzer (Chemical Data Systems, Oxford, Pa.). Upon rapid heating at high temperature (50e650 C at a rate of 20 Cm s1), organic matter is degraded into low-molecular weight thermal decomposition products. Product analysis was performed on a HewlettePackard 5890 Series II gas chromatograph with mass spectrometry detector (HP G1800A operating at EI ¼ 70 eV). The system is equipped with a DB-5 MS (J&W Scientific) fused-silica capillary column (30 0.25 mm i.d. with 0.25 mm film thickness). The injector temperature was 280 C, and the detector temperature was 250 C. The oven was programmed with an initial temperature of 50 C, and the temperature was ramped to 300 C at 7 C min1 and held for 10 min. Helium was used as the carrier gas at a constant flow rate of 1.0 mL min1. Py-GC/MS is a powerful analytical tool that reveals a diverse range of chemical constituents in fragments enough to identify probable source biopolymers (i.e. lignins, lipids and proteins) (Bruchet, 1985).
2.3.2. High pressure size exclusion chromatography with UV and fluorescence detection The liquid chromatograph consisted of a high pressure pump Waters 1525 (Waters, MA, USA), an automatic sampler with 200 mL loop, a variable wavelength UV detector Waters 2487 operating at 254 nm, and a waters 2475 fluorescence detector operating at excitation and emission wavelengths of Ex. 278 nm/ Em. 310 nm. The SEC fractionation was carried out with a size exclusion macroporous silica-based Reprosil 200 SEC column (AIT, Houilles, France) (8 mm i.d., 300 mm length) using 0.01 M sodium acetate at pH 7 solution as eluent. The flow rate of the eluent was 1 mL min1 and the injection volume 200 mL. All samples were passed through 0.2 mm filters (cellulose acetate,
3.
Results and discussion
3.1. Physico-chemical properties of organic matter sources Main properties of the studied organic matter sources are presented below. The physico-chemical composition of the surface water and the landfill leachate is very different. The surface water (DOC ¼ 8 0.2 mg C L1) contains much less organic matter than the landfill leachate (DOC ¼ 115 1 mg C L1).SUVA, COD and BOD5 parameters indicate that the characteristics of
319
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 1 5 e3 2 7
less concentrated samples of the leachate and the surface water, the saturation is not observed at Vp/Vr ¼ 700 L L1. These results indicate that at high DOC concentration, a smaller volume of sample is enough to saturate the resin. For the Vp/Vr values of 61.2 L L1 and 121.2 L L1 (corresponding to k0 ¼ 50 and k0 ¼ 100, respectively), the saturation zone is not reached both for the landfill leachate and the surface water samples. The saturation zone must be avoided to determine the proportion of hydrophilic vs. humic substances.
their organic matter are also different. For the surface water, COD ¼ 30 and BOD5 ¼ 20 10 mg O2 L1. Higher COD and BOD5 values are observed for the leachate (380 and 80 10 mg O2 L1, respectively). These findings indicate that much more oxidant is required to oxidize (i.e. to degrade) chemically or biologically the organic matter present in this sample compared to the surface water. The surface water presents a high SUVA value (0.034 L mg1 cm), suggesting a higher aromatic structure content. For the landfill leachate the SUVA value is (0.024 L mg1 cm).
3.3. Cautions to the quantification of hydrophilic substances
3.2. Influence of volume and concentration of DOM on the fractionation by XAD resin
It should be noted that HyS are estimated based on measuring the DOC concentration in the effluent. In order to make a comparison between experiments performed at different DOC concentration, breakthrough curves have been converted to retention curves, representing the percentage of DOC retained as a function of the volume of sample passed through the column per volume of resin. For the leachate samples (Fig. 2b), comparison of the three retention curves shows that the retention of DOM by the XAD resin is independent of the DOC concentration until Vp/Vr ¼ 150 L L1, i.e. k0 ¼ 124. The leachates samples (diluted half and tenth) present similar retention curves until Vp/Vr ¼ 300 L L1, i.e. k0 ¼ 249. For the surface water samples (Fig. 2a), the retention curves are similar until Vp/Vr ¼ 150 L L1. It should be remembered that raw and prepared samples are assumed to contain the same proportion of HyS vs. HS. Hence, our experiments show that the proportion of HyS is strongly influenced by the DOC concentration
The breakthrough curves, resulting of XAD fractionation experiments with the surface water and the landfill leachate samples, are shown Fig. 1a and b, respectively. These curves look like similar of breakthrough curves reported in studies dealing about fractionation of natural waters by XAD resins (Malcolm and Mac Carthy, 1992). Briefly, the shape presents a gradual increase in the DOC breakthrough with increasing volume of sample passing through the column. This finding indicates that the volume of sample influences the residual-DOC content (i.e. the amount organic matter not retained by the resin). The saturation of the resin is observed at high Vp/Vr values. For the landfill leachate samples, the saturation of the resin occurs at Vp/Vr w350 L L1 and w600 L L1 for the undiluted leachate sample and the leachate sample diluted half, respectively (Fig. 1b). For the pre-concentrated surface water, the saturation is observed at about 440 L L1 (Fig. 1a). For the
a
90 initial DOC content in pre-concentrated water
DOC (mg C , L-1)
80 70
pre-concentrated water
60
raw water
50 40 30 20
initial DOC content in untreated water
10 0 0
b
100 200 300 400 500 600 700 Volume sample passed through column per volume of resin (l l-1)
140 initial DOC content in untreated leachate
DOC (mg C. L-1)
120
raw leachate
100
leachate diluted twice
80
leachate diluted tenth initial DOC content in leachate diluted twice
60 40
initial DOC content in leachate diluted tenth
20 0 0
100
200
300
400
500
600
700
Fig. 1 e DOC breakthrough curves for freshwater samples (a) and leachate samples (b) on DAX-8 resin.
320
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 1 5 e3 2 7
% DOC ret ained on DAX-8 res in
a
80
60
90 80
40
70
0
60
20
40
60
80
100
50 40
pre-concentrated water
30
untreated water
20 10 0
50
100 150 200 250 300 350 400 450 500
b 80
% DOC ret ained on DAX-8 res in
60
90 80
40
70
0
20
40
60
80
100
60 50 40
38
untreated leachate
30
leachate diluted twice
20
leachate diluted tenth
10 0
50
100 150 200 250 300 350 400 450 500
Volume of sample passed through column per volume of resin (L.L -1)
0
50
100
150
200
250
300
350
400
k' coefficient
Fig. 2 e Proportion of total DOC retained on the XAD-8 resin. Standard deviation was calculated from three replicates.
a
k' =7
k' =15
k' =50
k' =100
k' =166
k' =7
k' =15
k' =50
k' =100
k' =166
b
Volume of sample passed through column increasing Fig. 3 e Repartition of humic substances vs. hydrophilic substances at various k0 values for the surface water (a) and the landfill leachate (b). -: Humic substances, ,: Hydrophilic substances.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 1 5 e3 2 7
321
Fig. 4 e Synchronous fluorescence spectra of HyS extracted from the surface water (a) and the landfill leachate (b) at various k0 values. (Dl [ 18 nm).
from certain k0 values, especially at high values. According to the study of Gadmar et al. (2005), the fractionation of OM by the XAD resins is influenced (from a quantitative point of view) by the concentration of NOM whatever the volume of sample used. However, their study seems to present large standard deviations and DOM was mainly measured by UV absorbance measurement (i.e. the DOM that does not absorb in the UV domain is not considered). In our study, we found no influence of the DOC concentration on the quantity of HyS for the k0 values of 50 and 100, used in the Leenheer (1981) or Thurman and Malcolm (1981) procedures.
We performed also many fractionation experiments at different k0 values (k0 ¼ 7; 15; 50; 100; 166) in order to determine whether the volume of sample, independently of the DOM concentration, influences the quantification results. The proportion of HyS vs. HS was determined from these experiments. Results are presented in Fig. 3. Results indicate that the proportion changes with the increase of the volume of sample percolated (i.e. with the increase of k0 values). The change of repartition is characterized by a progressive increase of the portion of HyS. This finding is observed for the two types of organic matter (landfill leachate and surface water). At low k0
322
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 1 5 e3 2 7
Fig. 5 e Size exclusion chromatograms of HyS isolated from the surface water (a, b) and the landfill leachate (c, d) at various k0 values and detected by UV or fluorescence detection.
values (7e15) HS seem to be the most abundant component of DOM whereas HyS tend to be more abundant when larger k0 values are applied (k0 > 100). These observations seem to be less remarkable for the surface water than for the landfill leachate. The comparison between the repartitions calculated for k0 ¼ 50 and k0 ¼ 100 highlights a difference of 9 2% and 8 2%, for the landfill leachate and the surface water, respectively. Theses findings show that the proportion of hydrophilic substances, determined from experiments using XAD resins, depend closely of the k0 value applied. From these results we conclude that the k0 value chosen can influence the results, and hence the conclusions of a study. It is therefore necessary to apply the same k0 value whatever the sample studied for having a correct comparison. In addition, such observation points out that the comparison of quantification results obtained by different procedures may be awkward.
3.4. Cautions to the characterization of hydrophilic substances Scientists need robust and repeatable isolation procedures to study hydrophilic substances. This point is essential for reducing artifacts/errors in the determination of their characteristics and properties. For the present work, many fractionation experiments were performed at k0 ¼ 15; 50; 100; 166 and hydrophilic substances were then isolated. After isolation, HyS were characterized by synchronous fluorescence spectroscopy, HPSEC with UV and fluorescence detections and flash pyrolysis. Synchronous fluorescence spectra are presented in Fig. 4a and b. As shown in these figures, most of the HyS studied show similar peaks, but with different intensities. This finding implicates the existence of common fluorophores but in different proportions in the HyS studied. Briefly, for the surface
water the spectra exhibit one main peak, the peak I, with an excitation wavelength of w340 nm, and one additional “shoulder”, the peak II, at w390 nm (Fig. 4a). Several other “little” peaks were observed: the first at 290 nm and others at 450 nm and 465 nm. The peaks observed at these wavelengths have been already reported in works studying algal organic matter (Ferrari and Mingazzini, 1995) or soil organic matter (Kalbitz et al., 2000) by synchronous fluorescence spectroscopy. These authors attribute them to the presence of microbial-like or humic substances-like compounds. For the landfill leachate (Fig. 4b), the shape of spectra of HyS resembles that of HyS isolated in the surface water, except for the HyS extract at k0 ¼ 15. According to Seo et al. (2007), it would be expected that tyrosine-like and tryptophan-like fluorophores (i.e. microbiallike compounds) account for a large portion of the fluorophores in the hydrophilic substances isolated from leachates from municipal solid wastes residues. Synchronous fluorescence results show that k0 value affects the fluorescence properties of HyS. The same changes are observed for the two types of sample studied. Briefly, for the lowest k0 values, spectra presented l and fluorescence intensity values that are systematically lower and larger, respectively, than those corresponding to highest k0 values. Indeed, the fluorescence intensity of the peaks I and II decreases with the increase of the k0 value. The fluorescence bands are shifted to longer wavelengths for the high k0 values. Between k0 ¼ 50 and k0 ¼ 100, we observed a decrease of the DOC-normalized fluorescence intensity of about 5 104. The shift in the maxima of fluorescence intensity from shorter to longer wavelengths may be attributed to an increasing number of highly substituted organic nuclei (Miano and Senesi, 1992). According to Kalbitz et al. (2000), the fluorescence of non-humic substances is more pronounced at shorter than at longer wavelengths. The
323
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 1 5 e3 2 7
60
50
a
50
40
40 30
HyS from surface water at DOC = 8 mg C.L-1
10 0 V1 FU
V2 FU
V3 FU
V4 FU
60 50 40 30
Fluorescence (% of total chromatogram area)
Absorbance (% of total chromatogram area)
30 20
20
20 10 0 Ff
o1 lu
o2 lu Ff
o3 lu Ff
Ff
o4 lu
50 40 30
HyS from landfill leachate at DOC = 115 mg C.L-1
20 10
10 0 1 UV F*
2 UV F*
3 UV F*
0
4 UV F*
fl F*
50
60
40
50
1 uo
F*
o2 luo3 luo4 luo5 luo6 luo7 f f f f f f lu F* F* F* F* F*
b HyS from surface water at DOC = 81 mg C.L-1
40
20 10 0 V1 FU
V2 FU
V3 FU
V4 FU
60 50 40 30
Fluorescence (% of total chromatogram area)
30 Absorbance (% of total chromatogram area)
k' = 15 k' = 50 k' = 100 k' = 166
30 20 10 0 Ff
50
o1 lu
Ff
o2 lu
Ff
o3 lu
o4 lu Ff
k' = 50 k' = 100
40 30
HyS from landfill leachate at DOC = 11 mg C.L-1
20 20 10
10
0
0 1 UV F*
2 UV F*
3 UV F*
4 UV F*
F*
o7 o6 o5 o4 o3 o2 o1 flu f lu f lu flu flu f lu flu F* F* F* F* F* F*
Fig. 6 e Repartition of fractions observed in size exclusion chromatography for the surface water and the landfill leachate. (a) untreated samples, (b) pre-treated samples.
results concerning the influence of the DOM concentration suggest that this parameter (in the range studied) has poor influence on the fluorescence properties of the bulk HyS samples. Indeed, we observed a good overlap of spectra obtained for HyS isolated with the same k0 value (i.e. k0 ¼ 50 or k0 ¼ 100) from samples prepared at different concentrations (Fig. 4a and b). The SEC-UV/fluorescence chromatograms of the hydrophilic fractions are shown in Fig. 5a and b for the surface water and in Fig. 5c and d for the landfill leachate. Based on these chromatograms, different size fractions were distinguished (see Materials and methods section) and their proportions were determined for ease the interpretation. The repartition of size fractions (in percentage of the total chromatogram area) are plotted in Fig. 6a. For the surface water, four fractions were distinguished in the SEC-UV and in the SEC-fluorescence chromatograms. Be`le et al. (2006), using the same analytical
set than us, obtained slightly different chromatograms on DOM fractions isolated from rivers and lakes. This difference might be attributed to the variability of NOM. Our results highlight the modification of the HyS composition with an increase of the k0 value (Fig. 6a). The fraction FUV-1 (which is composed of a group of peaks) and the fraction FUV-2 (which is composed of one peak) tend to increase with the increase of the k0 value while the peak forming the FUV-3 fraction decreases. In the fraction FUV-4, the shape of the double peak change at different k0 values too (Fig. 5a). On the contrary, the SEC-fluorescence chromatograms are very close whatever the k0 value used (Fig. 5b). This finding explains the similar distributions observed in Fig. 6a. For the landfill leachate, the shape of UV (or fluorescence) chromatograms differs greatly with the increase of the k0 value, especially between k0 ¼ 50 and k0 ¼ 100. The peaks overlap at low k0 values while they are better discerned for higher k0
324
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 1 5 e3 2 7
values. The proportions of F* fractions (i.e. relative to landfill leachate) are plotted in Fig. 6a. Several significant modifications are observed with the variation of the k0 value. It should be noted that the fractions distinguished on the leachate chromatograms are different to the fractions distinguished on the surface water chromatograms. Proportions of the F*UV-1 and F*UV-2 fractions increase while that of the F*UV-4 fraction decreases with increasing k0 . These findings reveal some quantitative changes in the composition of the HyS. Moreover, our results also suggest that qualitative changes occur. For example, the F*UV-4 fraction is composed of two overlapping peaks until k0 ¼ 50 whereas the left peaks disappeared at k0 ¼ 100 or 166. The fractions observed on the SEC-fluorescence chromatograms (Fig. 5d) also confirm these observations. The fractions F*FLUO-2, F*FLUO-3 and F*FLUO-4 appear as shoulders at k0 ¼ 15 or 50 and as individualized peaks at higher k0 values. The fraction F*FLUO-5 is composed of one peak that disappeared from k0 ¼ 100 whereas the peak of the fraction F*FLUO-6 appears only at this k0 value. Consequently, the proportion of these fractions is different depending on the k0 value (Fig. 6a). The influence of the DOM concentration was also studied and results are presented in Fig. 6b. The data suggest that modifying the DOM concentration may induce slight changes in
the composition of HyS. About 4e10% of variation was observed between same fractions recovered from HyS isolated at different DOM concentration. For the surface water sample, the FUV-1 and FUV-2 fractions decrease slightly while the FUV-3 and FUV-4 fractions increase. A similar trend is observed for FFLUO fractions. The influence of the DOM concentration seems to be more marked at k0 ¼ 100 than 50. For the landfill leachate, the F*UV-1 and F*UV-2 fractions tend rather to increase and the others fractions to decrease. Fluorescence results exhibit also that many changes in the repartition of fractions occur when the concentration of DOM is modified prior the extraction of HyS. These observations suggest that low DOC concentrations favor the first eluted fractions while high DOC concentrations favor the last eluted fractions. According to Schnitzer (1991), DOM has been known to change its form and microstructure in aqueous solution depending on the pH, concentration and ionic strength. All these results point out significant differences between experiments performed at k0 ¼ 50 and k0 ¼ 100. The k0 value influences the composition (amount and type of compounds) of the HyS whatever the origin of the sample (surface water and landfill leachate). However, the changes seem to be different according to the source of organic matter studied. For the leachate, the proportion of the first fraction increases
Fig. 7 e Pyrochromatograms of HyS isolated from the surface water and the landfill leachate under various conditions of k0 and DOC concentrations. Acetonitrile (A); toluene (B); alkylbenzene (C); unknown (D); styrene (E); acid acetic (F); furfural (G); ethyl hexanol (H); methylfurancarboxaldehyde (I); acetamide (J); methoxyphenol (K); unknown (L); phenol (M); methylphenol (N); phenol-like compound (O); unknown (P); benzoic acid D others (R); ethyl pentane (S); ethyl hexanal (S2); ethylphenol (T); unknown (U); unknown (V); dodecanoic acid (W); benzene (X); phthalate (Y).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 1 5 e3 2 7
and those of the two last fractions decrease with the k0 value while the reverse behavior is observed for the surface water. Considering that in size exclusion chromatography the largest compounds have the fastest the elution, our results suggest that increasing the k0 value favors the presence of higher molecular weight (HMW) compounds in the HyS of the landfill leachate and rather lower molecular weight (LWM) compounds in the HyS of the surface water. The molecular analysis performed by flash pyrolysis is presented in Fig. 7. The “structural fingerprints” gives insight of the composition of HyS. In the present work, we also used Py-GC/MS to provide useful information regarding the origin of the fragments. For that, the fragments observed in pyrochromatograms were compared to the list of typical fragmentation of biopolymers (Bruchet, 1985; Christy et al., 1999). Py-GC/MS highlight several differences according to the k0 value used. For the surface water, A, F and G fragments are very intense at k0 ¼ 50 whereas H and Q compounds are very intense at k0 ¼ 100 (Fig. 7a). The E fragment is only observed at k0 ¼ 50 while the C, D, K and O fragments are only observed at k0 ¼ 100. Some fragments (e.g. N and P) keep relatively the same intensity whatever the k0 used. The HyS isolated from the high or low concentrated samples present close pyrochromatograms (Fig. 7a and c). We observe that the A, F, G, I, J and N fragments are systematically present and intense. Some differences are, however, noticed for the other fragments. The L and P fragments are only observed for the HyS isolated from the sample with the lower DOC content. The M fragment is absent in this condition while it corresponds to the one of the more intense peaks observed of the HyS obtained from rich DOM sample (DOC ¼ 81 mg C L1). A broad examination of the pyrochromatograms reveals that other fragments (e.g. near the P and the D fragments) appear. For the landfill leachate, the use of a higher k0 (i.e. k0 ¼ 100) leads also to a modification of the pyrochromatogram (Fig. 7b). At k0 ¼ 100, several new fragments are observed: S, H, O, V and W. We also note the presence of two intense fragments (M and R) that were weak or absent at lower k0 . By the opposite, the A, B, F, N and U fragments are more intense at k0 ¼ 50. The pyrochromatograms of HyS isolated at k0 ¼ 100 and DOC ¼ 11 mg C L1 show also that concentration of the sample may affect the composition of HyS. Indeed, two new fragments (X and Y fragments) are observed for HyS extracted from a lower DOC sample. However, the Y fragment is a phthalate and hence its presence may be due to a contamination. Four fragments are more intense under these experimental conditions (E, F, O and H fragments). It should be noted that the S fragments seems to be present only at k0 ¼ 100. Regarding to the origin of the fragments, an increase of the k0 seems to lead to a higher humic character for the surface water. Indeed, K and O correspond to polyhydroxyaromaticlike compounds and C to alkylbenzene-like compounds. The significant increase of the phenol (M) suggests that the DOC content of the sample as a similar influence. For the landfill leachate, the fragments observed at k0 ¼ 50 suggest a microbial character because of the presence of protein-like (A, B) and polysaccharide-like (F) compounds. The decrease of these fragments at higher k0 value suggests a progressive loss of the microbial character to the benefit of another character, marked by n-alkane-like, carboxylic acid-like, and polyhydroxyaromatic-like compounds.
4.
325
Conclusion
Although the study of HyS has become a source of great concern in the scientific community throughout the World, there have been problems in comparing studies. Much of this difficulty stems from the isolation procedure. For methodologies using XAD resin, different k0 coefficients and/or DOM matter content may be used. Our results show that the k0 value strongly influences the quantity of HyS isolated. Fortunately, the difference between the two usual k0 values, i.e. k0 ¼ 50 (Leenheer, 1981) and k0 ¼ 100 (Thurman and Malcolm, 1981), is weak (about 8%). But, we find that the composition of HyS is significantly modified between these two k0 values. Our results suggest a contribution of humic-like matter (high-molecular weight aromatic compounds) at k0 ¼ 100. This results in a shift of fluorescence bands to longer wavelengths and changing patterns of the SEC profiles. Molecular fingerprints performed by Py-GC/MS confirm the presence of polyhydroxyaromaticlike and alkylbenzene-like fragments. Our results also show that the concentration of the sample in DOM influence the composition of HyS but not their quantification at the usual k0 values. Several modifications were observed in SEC and Py-GC/MS. These observations suggest that low DOC concentrations favor the presence of HMW structures in the HyS while high DOC concentrations favor the LMW structures. It was found that the impact of the DOM content is higher at k0 ¼ 100 than at k0 ¼ 50. Confrontation of results obtained for the surface water and the landfill leachate suggest that the source of organic matter influences the effects of the distribution coefficient k0 and the DOM concentration. Indeed, even if similar effects are noted, each source of organic matter reacts specifically as shown by SEC-fluorescence and pyrolysis measurements. Finally, the study of HyS appears clearly influenced by the k0 coefficient and the concentration in DOM. This finding points out that precaution should be taken to compare scientific works. Moreover, the pretreatment of sample by concentration or dilution should be also used with precautions. We encourage all the researchers to work with the same experimental condition to produce more comparable results regarding the characteristics of HyS. From our results we suggest that the best conditions to isolate hydrophilic compounds are: low k’ coefficient and low concentration of DOM.
Acknowledgments The authors thank the Pr. Jean-Philippe Croue´ (University of Poitiers, France) and Pr. Jerry A. Leenheer (US Geological Survey Denver, Colorado, USA) for their kind advices to improve this work.
references
AFNOR, 1997. Qualite´ de l’Eau. Recueil des Normes Franc¸aises. AFNOR, Paris, France.
326
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 1 5 e3 2 7
Aiken, G.R., McKnight, D.M., Thorn, K.A., Thurman, E.M., 1992. Isolation of hydrophilic organic acids from water using nonionic macroporous resins. Org. Geochem. 18, 567e573. Be`le, C., Habarou, H., Ambonguilat, S., Croue´, J.-P., Djafer, M., Hei, V., 2006. How to optimize NOM removal on drinking water plants? First step: making a good diagnosis thanks to HPSEC. Water Pract. Technol. 1. Belzile, C., Guo, L., 2006. Optical properties of low molecular weight and colloidal organic matter: application of the ultrafiltration permeation model to DOM absorption and fluorescence. Mar. Chem. 98, 183e196. Bruchet A. Application of the Pyrolysis-GCeMS technique to the study of nonvolatile organic matter of natural waters during treatment, Ph-D thesis, 1985. University of Poitiers, Poitiers, France. Chang, C.-Y., Hsieh, Y.-H., Lin, Y.-M., Hu, P.-Y., Liu, C.-C., Wang, K.-H., 2001. The organic precursors affecting the formation of disinfection by-products with chlorine dioxide. Chemosphere 44, 1153e1158. Christy, A.A., Bruchet, A., Rybacki, D., 1999. Characterization of natural organic matter by pyrolysis-GCeMS. Environ. Int. 25, 181e189. Croue´, J.P., 2004. Isolation of humic and non-humic NOM fractions: structural characterization. Environ. Monit. Assess. 92, 193e207. Croue´, J.P., Martin, B., Deguin, A., Legube, B. Isolation and characterization of dissolved hydrophobic and hydrophilic organic substances of a water reservoir, Proceeding of workshop on NOM in Drinking Water, Chamonix, France, Sept. 19e22, 1993. Croue´, J.P., Benedetti, M., Violleau, D., Leenheer, J.A., 2003. Characterization and copper binding of humic and non-humic organic matter isolated from the South Platte river: evidence of the presence of nitrogenous binding sites. Environ. Sci. Technol. 37, 328e336. Farnworth, J.J., 1995. Comparisons of the sorption from solution of a humic acid by Supelite DAX-8 and by XAD-8 resins. IHSS Newsl. 13, 8e9. Fearing, D., Goslan, E., Banks, J., Wilson, D., Hillis, P., Campbell, A. T., Parsons, S.A., 2004. J. Envir. Eng. 130, 975e982. Ferrari, G.M., Mingazzini, M., 1995. Synchronous fluorescence spectra of dissolved organic matter (DOM) of algal origin in marine coastal waters. Mar. Ecol. Progr. Ser. 125, 305e315. Fuentes, M., Gonzalez-Gaitano, G., Garcia-Mina, J.M., 2006. The usefulness of UVevisible and fluorescence spectroscopies to study the chemical nature of humic substances from soils and composts. Org. Geochem. 37, 1949e1959. Gadmar, T.C., Vogt, R.D., Evje, L., 2005. Artefacts in XAD-8 NOM fractionation. Int. J. Environ. Anal. Chem. 85, 365e376. 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, 4295e4303. Imai, A., Fukushima, T., Matsushige, K., Kim, Y.-H., Choi, K., 2002. Characterization of dissolved organic matter in effluents from wastewater treatment plants. Water Res. 36, 859e870. International Humic Substance Society, 2007. web site. http:// www.ihss.gatech.edu/. Kalbitz, K., Geyer, S., Geyer, W., 2000. A comparative characterization of dissolved organic matter by means of original aqueous samples and isolated humic substances. Chemosphere 40, 1305e1312. Kang, K.-H., Shin, H.S., Park, H., 2002. Characterization of humic substances present in landfill leachates with different landfill ages and its implications. Water Res. 36, 4023e4032. Kaushal, S.S., Lewis, W.M., 2005. Fate and transport of dissolved organic nitrogen in minimally disturbed montane streams of Colorado. Biogeochemistry 74, 303e321.
Kilduff, J.E., Mattaraj, S., Wigton, A., Kitis, M., Karanfil, T., 2004. Effects of reverse osmosis isolation on reactivity of naturally occurring dissolved organic matter in physicochemical processes. Water Res. 38, 1026e1036. Kitis, M., Kilduff, J.E., Karanfil, T., 2001. Isolation of dissolved organic matter (DOM) from surface waters using reverse osmosis and its impact on the reactivity of DOM to formation and speciation of disinfection by-products. Water Res. 35, 2225e2234. Koopal, L.K., van Riemsdijk, W.H., Kinniburgh, D.G., 2001. Humic matter and contaminants. General aspects and modeling metal ion binding. Pure Appl. Chem. 73, 2005e2016. Labanowski, J., Feuillade, G., 2009. Combination of biodegradable organic matter quantification and XAD-fractionation as effective working parameter for the study of biodegradability in environmental and anthropic samples. Chemosphere 74, 605e611. Lee, S., Ang, W.S., Elimelech, M., 2006. Fouling of reverse osmosis membranes by hydrophilic organic matter: implications for water reuse. Desalination 187, 313e321. Leenheer, J.A., 1981. Comprehensive approach to preparative isolation and fractionation of dissolved organic carbon from natural waters and wastewaters. Environ. Sci. Technol. 15, 578e587. Leenheer, J.A., Croue´, J.P., 2003. Characterizing aquatic dissolved organic matter. Environ. Sci. Technol. 37, 18Ae26A. Leenheer, J.A., Croue´, J.P., Benjamin, M., Korshin, G., Hwang, C., Bruchet, A., Aiken, G.R., 2000. Natural Organic Matter and Disinfection By-Products. In: Barrett, S.E., Krasner, S.W., Amy, G.L. (Eds.), ACS Symposium Series 761. American Chemical Society, Washington DC, pp. 68e83. Malcolm, R.L., Mac Carthy, P., 1992. Quantitative evaluation of XAD-8 and XAD-4 resins used in tandem for removing organic solutes from water. Environ. Int. 18, 597e607. Miano, T.M., Senesi, N., 1992. Synchronous excitation fluorescence spectroscopy applied to soil humic substances chemistry. Sci. Total Environ. 117/118, 41e51. Nagao, S., Matsunaga, T., Suzuki, Y., Ueno, T., Amano, H., 2003. Characteristics of humic substances in the Kuji river waters as determined by high-performance size exclusion chromatography with fluorescence detection. Water Res. 37, 4159e4170. Peuravuori, J., Koivikko, R., Pihlaja, K., 2002. Characterization, differentiation and classification of aquatic humic matter separated with different sorbents: synchronous scanning fluorescence spectroscopy. Water Res. 36, 4552e4562. Schnitzer, M., 1991. Soil organic matter. The next 75 years. Soil. Sci. 151, 41e58. Schwarzenbach, R., Gschwend, P.M., Imboden, D.M., 1993. Environmental Organic Chemistry. Wiley, New York, USA. Senesi, N., Xing, B., Huang, P.M. (Eds.), 2006. Biophysico-chemical Processes Involving Natural Nonliving Organic Matter in Environmental Systems. IUPAC, New York, USA. Seo, D.-J., Kim, Y.-J., Ham, S.-Y., Lee, D.-H., 2007. Characterization of dissolved organic matter in leachate discharged from final disposal sites which contained municipal solid waste incineration residues. J. Hazard. Mater. 148, 679e692. Serkiz, S.M., Perdue, E.M., 1990. Isolation of dissolved organic matter from Suwannee river using reverse osmosis. Water Res. 24, 911e916. Sharp, E.L., Jarvis, P., Parsons, S.A., Jefferson, B., 2006. Impact of fractional character on the coagulation of NOM. Colloids Surf. A Physicochem. Eng. Asp. 286, 104e111. Sposito, G., 1989. The Chemistry of Soils. Oxford University Press Inc., New York, USA.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 1 5 e3 2 7
Thurman, E.M., Malcolm, R.L., 1981. Preparative isolation of aquatic humic substances. Environ. Sci. Technol. 15, 463e466. Thurman, E.M. Developments in Biochemistry: Organic Geochemistry of Natural Waters. Nijhoff, M. and Junk, W. (Eds.); Dordrecht, 1985.
327
Zhao, Z.-Y., Gu, J.-D., Fan, X.-J., Li, H.-B., 2006. Molecular size distribution of dissolved organic matter in water of the Pearl River and trihalomethane formation characteristics with chlorine and chlorine dioxide treatments. J. Hazard. Mater. B134, 60e66.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 2 8 e3 3 6
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Effect of imposed anaerobic conditions on metals release from acid-mine drainage contaminated streambed sediments Barbara A. Butler* U.S. EPA Office of Research and Development, National Risk Management Research Laboratory, Land Remediation and Pollution Control Division, 26 W. Martin Luther King Drive, Cincinnati, OH 45268, USA
article info
abstract
Article history:
Remediation of streams influenced by mine-drainage may require removal and burial of
Received 19 March 2010
metal-containing bed sediments. Burial of aerobic sediments into an anaerobic environ-
Received in revised form
ment may release metals, such as through reductive dissolution of metal oxyhydroxides.
23 July 2010
Mining-impacted aerobic streambed sediments collected from North Fork Clear Creek,
Accepted 26 July 2010
Colorado were held under anaerobic conditions for four months. Eh, pH, and concentra-
Available online 4 August 2010
tions of Cd, Cu, Fe, Mn, and Zn (filtered at 1.5 mm, 0.45 mm, and 0.2 mm), sulfate, and dissolved organic carbon (DOC) were monitored in stream water/sediment slurries. Two
Keywords:
sediment size fractions were examined (2 mme63 mm and <63 mm). Sequential extractions
Iron oxyhydroxides
evaluated the mineral phase with which metals were associated in the aerobic sediment.
Manganese oxyhydroxides
Released Cu was re-sequestered within 5 weeks, while Fe and Mn still were present at 16
Metal sulfides
weeks. Mn concentration was lower than in the initial stream water at and beyond 14
Remediation
weeks for the smaller sized sediment. Cd was not released from either sediment size fraction. Zn was released at early times, but concentrations never exceeded those present in the initial stream water and all was re-sequestered over time. The greatest concentrations of Cu, Fe, Mn, and Zn were associated with the Fe/Mn reducible fraction. Sulfate and Fe were strongly correlated (r ¼ 0.90), seeming to indicate anaerobic dissolution of iron oxy-hydroxy-sulfate minerals. DOC and sulfate were strongly correlated (r ¼ 0.81), with iron having a moderately strong correlation with DOC (r ¼ 0.71). Overall concentrations of DOC, sulfate, Cu, Fe, and Zn and pH were significantly higher ( p < 0.05) in the water overlying the small sized sediment samples, while the concentrations of Mn released from the larger sized sediment samples were greater. Published by Elsevier Ltd.
1.
Introduction
Thousands of miles of streams in the United States are contaminated by mine-drainage, resulting predominantly from metal mining in the western U.S. and from coal mining in the eastern U.S. Once this drainage enters a stream,a variety of chemical processes may occur, including
* Tel.: þ1 513 569 7468; fax: þ1 513 569 7620. E-mail address:
[email protected]. 0043-1354/$ e see front matter Published by Elsevier Ltd. doi:10.1016/j.watres.2010.07.077
neutralization of acidity with oxidation and subsequent precipitation of metal oxyhydroxides, such as aluminum oxyhydroxide (ALO), hydrous iron oxide (HFO), and/or hydrous manganese oxide (HMO), each known to sorb and/or co-precipitate metals (Davis and Leckie, 1978; Dong et al., 2003; Jenne, 1968; Karthikeyan et al., 1997; Turner et al., 2004; Webster et al., 1998). Metal oxyhydroxide precipitates
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 2 8 e3 3 6
can be transported downstream in the suspended sediment fraction and/or aggregate and settle to the streambed (Stumm, 1992; Schemel et al., 2000). Remediation of these streams is desired to restore their aquatic ecosystems and might include physical removal of tailings piles and/or streambed sediments. Disposal of streambed sediments might be into an existing tailings impoundment (U.S. EPA, 2001, 2004) or into a repository also containing tailings, waste rock, or other material removed during remediation (U.S. EPA, 2004, 2005a,b). Generally repositories are lined; however, some are unlined (e.g., Bunker Hill Complex e U.S. EPA, 2005a), which might lead to release of metals to groundwater, if they are mobilized by anaerobic conditions and remain mobile. When a metal oxyhydroxide (with or without other sorbed metals) is exposed to anaerobic conditions (as would be expected at depth in a repository or tailings impoundment), it is likely that metals will be released via reductive dissolution (Davranche and Bollinger, 2000; Grybos et al., 2007; Pareuil et al., 2008). Any dissolution of metal oxyhydroxide minerals will release not only the metal that comprised the mineral, but also any metals that were previously sorbed to and/or coprecipitated with it; changes in pH and Eh also are expected to affect any release of metals. Released metals might be precipitated as insoluble metal sulfides, providing sufficient sulfide is present, such as through microbial sulfate-reduction, or they might sorb and/or co-precipitate with other sulfide minerals. In the absence of sulfide or another mineral or organic phase for re-sequestration, released metals might continue to be mobile in the anaerobic environment. Understanding the behavior of previously aerobic streambed sediment-associated metals under anaerobic conditions may assist in evaluating the best remedial disposal strategy. This research assessed the release of Cd, Cu, Fe, Mn, and Zn (filterable at 1.5 mm, 0.45 mm, and 0.2 mm), SO2 4 , and dissolved organic carbon (DOC) from two particle size fractions (63 mm x < 2 mm and x < 63 mm) of initially aerobic streambed sediment over time in an anaerobic chamber. Additionally, pH, Eh, and acid-volatile sulfide (AVS) were measured in the sediment slurries.
2.
Materials and methods
2.1.
Study site
The North Fork of Clear Creek (NFCC) is Operable Unit #4 of the Clear Creek (CC) Superfund Site in the Rocky Mountains of Colorado, about 16 km west of Golden. This region is impacted by acid-mine drainage (AMD) originating from numerous abandoned mines in the watershed from extensive mining in the late 1800s for Au, Ag, Cu, Pb, and Zn in the Colorado Mineral Belt (CMB) (Cunningham et al., 1994; Wildeman et al., 1974). Regional geology includes Precambrian crystalline rocks interlaid with gneiss, granite, schist, and pegmatite with Tertiary intrusives being the source of the sulfide ores containing precious and base metals mined. Fig. 1 shows the sediment sampling area, along with the locations of the primary sources of AMD input to NFCC: the National Tunnel Adit, Gregory Gulch, and Gregory Incline, each originating in
329
Fig. 1 e North Fork Clear Creek Watershed sediment sampling location (adapted from Butler, 2009).
the Black Hawk/Central City region of the CMB. Upon mixing of the AMD with upstream NFCC water, there is transformation from ferrous to ferric iron, resulting in observed precipitation of HFO minerals (Butler et al., 2009). These HFO precipitates scavenge (through coprecipitation and sorption) other metals present in the water column and are transported downstream and also aggregate and settle to the streambed. Additionally, during some parts of the year, HMO is evident on rock and pebble surfaces in the streambed (Butler et al., 2009).
2.2.
Field sediment and water collection
Aerobic sediments were collected in April 2008 from the top 5-cm of the streambed in 50 locations over a distance of approximately 100 m (Fig. 1), using polypropylene scoops. Material observed to be much larger than 2 mm was avoided. Sediment was composited into two new 1-gallon high-density polyethylene (HDPE) jars. Two gallons of water were collected from the central portion of the stream along the 100 m reach for use in wet sieving of the sediments in the laboratory and for use in experiments. The 1-gallon HDPE sampling jars were rinsed three times with stream water and then water was collected by placing the bottles upside down under the water, inverting them, and slowly bringing them upward through the water column. Containers were sealed with duct tape, placed into zip-locked plastic bags, and shipped on ice to the laboratory following EPA chain of custody procedures. Once received at the laboratory, samples were refrigerated at 4 C until use.
2.3.
Laboratory sediment and water processing
The bulk wet sediment was homogenized in a 3-gallon tub by hand mixing with a polypropylene spoon for 30 min. A sub-sample of approximately 150 g wet sediment was removed and the remaining homogenized sediment was wetsieved with stream water into two size fractions: 1) 63-mm
330
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 2 8 e3 3 6
x < 2-mm and 2) x < 63-mm. Overlying water was allowed to settle and then was decanted. Settling and decanting of overlying water was repeated for the smaller sized sediment over 1 week to minimize the amount of water remaining before using the sediment in laboratory experiments; sediments were kept refrigerated at 4 C between manipulations. The sub-sample was wet-sieved at 63-mm x < 2-mm and x < 63 mm and dried at room temperature for 48 h. The dried material was massed on a Mettler PC2200 balance to determine the percentages of each size fraction present in the bulk bed sediment and saved for particle size analysis and sequential extractions. The remaining stream water was filtered through Whatman Nuclepore 0.45-mm filters for use in experiments. This water was then filtered at 1.5 mm, 0.45 mm, and 0.2 mm (in series) for initial concentrations of Cd, Cu, Fe, Mn, and Zn; at 0.45 mm for initial sulfate concentration; and at 1.5 mm for initial DOC concentration, using Target nylon 25 mm disposable syringe filters. Triplicate sub-samples of wet sediment (1e5 g each) from each size fraction were massed on an AND GR-202 AlphaLiberty Company balance, dried at 50 C, and massed again for calculation of the wet/dry ratio to estimate the mass of wet sediment needed for experimental units.
2.4.
Laboratory experiments
Sieved sediment samples were homogenized for 10e30 min before removing aliquots for experiments. For each sediment size, wet sediment was added to each of twenty 250 ml HDPE jars to give a dry mass sediment concentration of 100 g/L (based on previous calculations using the wet/dry ratio), with the exact mass recorded for later calculations. Seventy-five ml of the filtered stream water was added to each of the experimental units and to four additional jars without sediment, which served as blanks. Laboratory replicates also were included; for these, the volume of water and mass of sediment used were doubled but the solid/solution ratio was held at 100 g/L on a dry mass basis. At the beginning, middle, and end of obtaining experimental samples, sub-samples of approximately 1e5 g each were obtained from each size fraction, dried at 50 C and massed. The average wet/dry ratio for each sediment size was used for later conversion of metals, sulfate, and DOC concentrations to a dry sediment mass basis. The experimental units were placed onto a Lab-Line Orbit Shaker Model 3520 in a Coy Laboratory Products (MI) anaerobic chamber (7e10% H2; 90e93% N2, standard method for this chamber). The units were gently agitated at 150 rpm while loosely covered to allow exposure to the anaerobic atmosphere, but to minimize evaporative losses. Individual units were sacrificed at times of 0, 2, 6, and 10 h; 1, 2, 3, and 5 days; 1, 2, 3, 4, 5, 6, 7, 8, 10, 12, 14, and 16 weeks and sampled for Cd, Cu, Fe, Mn, and Zn (serially-filtered at 1.5 mm, 0.45 mm, and 0.2 mm); sulfate (filtered at 0.45 mm), DOC (filtered at 1.5 mm), pH, and Eh. Replicates were obtained at 1 day and 5 weeks for the larger sized sediment and at 5 days and 12 weeks for the smaller sized sediment. Sediments were collected in 15-ml polypropylene cryogenic vials and frozen for acid-volatile sulfide (AVS) and simultaneously-extracted metals (SEM) at
times of 0 and 6 h; 1, 3, 5, and 7 days; and 5, 8, 12, and 16 weeks. Blanks were sacrificed at 2 days and 2, 6, and 16 weeks. Samples collected for metals were acidified to pH < 2 with trace metal grade concentrated nitric acid (U.S. EPA, 1992). DOC samples were filtered at 1.5 mm (to eliminate clogging of the instrument from observed particles) into brown-glass borosilicate vials and acidified to pH < 2 with phosphoric acid (U.S. EPA, 2005c). Dissolved Cd, Cu, Fe, Mn, and Zn were measured on a PerkineElmer Optima 2100DV inductively coupled plasma atomic emission spectrometer (ICP-AES); dissolved sulfate was measured on a Dionex ICS 2000 ion chromatograph (IC) (U.S. EPA, 2000); and DOC was measured on a Phoenix 8000 UVePersulfate TOC Analyzer. The dissolved metals, sulfate, and DOC concentrations were converted to a dry mass basis (mg analyte/kg dry sediment) using the mean wet/dry mass ratios for the given sediment size. Eh and pH were measured using an Oakton pH 110 m with a Thermo Orion 9678BNWP redox probe and an RS232 pH probe, respectively. Eh values were normalized to the standard hydrogen electrode (SHE) by adding the potential of the silver:silver chloride-saturated KCl (Ag:AgCl) redox electrode to the measured values. The temperature during the course of experiments was 23 1 C, with the exception of the timezero samples, which were 18 1 C; resulting in values of 201 and 206 mV being added to the Eh readings, respectively (Nordstrom and Wilde, 2005). Above analyses were conducted following internal standard operating procedures. Measurement of AVS concentrations was conducted via the purge and trap method outlined in U.S. EPA (1991). Particle size analysis was conducted on a Beckman Coulter LS230 Particle Sizer following internal standard operating procedures.
2.5.
Sequential extractions
Sequential extraction of the dried sediment sub-sample was conducted following a modified Tessier et al. (1979) method in triplicate on each of the two sizes of sediment. The method extracts metals in five general fractions: easily exchangeable, carbonate, iron/manganese oxide associated, organic, and residual. A perchloric hood was not available; thus, nitric and hydrochloric acids, following the EPA Method 3051A (U.S. EPA, 2007), were used in lieu of the hydrofluoric and perchloric acids for extraction of residual metals. Iron/manganese oxides are reduced by an excess of hydroxylamine hydrochloride in the third step. Previous extractions on sediments from NFCC have shown there to be greater than 200 g Fe/kg sediment in the <63 mm sized sediment (unpublished data). Assuming this was the maximum concentration that would be present in this study, and using the stoichiometry shown in Eq. (1) and the volumes of reagents suggested by Tessier et al. (1979), a mass of approximately 100 mg dry sediment was chosen to allow for complete extraction without exhaustion of the reagent. For comparison purposes, this mass was used for both size fractions. 2NH2 OH HCl þ 2Fe3þ þ 2OH /2Fe2þ þ N2 þ 4H2 O þ 2Hþ
þ 2Cl
(1)
EPA Method 3051A was conducted on additional triplicate samples to obtain ‘total acid-recoverable’ concentrations of
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 2 8 e3 3 6
metals for mass balance and recovery calculations. The extracts were diluted with DI water and analyzed via ICP-AES for Cd, Cu, Fe, Mn, and Zn.
3.
Results
3.1.
Quality assurance/quality control
DOC and sulfate in the replicated samples were within 20% relative percent deviation (RPD). Replicated metals were within 20% RPD, with the exception of Cu and Zn in the 1.5 mm filtered sample from the 63 mm x < 2 mm sized sediment that was obtained at 35 days. For this sample, one replicate indicated Cu and Zn to be present, while the other did not. Because the measured concentrations were much higher than during any of the surrounding sampling times, as well as being below detection at the other two filtration sizes, it was assumed the sample was contaminated and it was not used in data analysis. Sulfate and DOC control check standards were within 10% of their known values. ICP control check standards were within 10% of their known values, DI water blanks were less than three times the detection limit or below 1% of the lowest measured concentration of analyte, and all reported measurements were within the linear calibration curve, with a reporting limit (RL) of 10 mg/L. The mean experimental blank concentrations for metals, SO2 4 , and DOC over time were subtracted from the measured values for each of the timed samples. The blank metals and SO2 4 concentrations were within 20% of the initial filtered stream water concentrations. The concentrations of DOC in the blanks indicated that there was some release (1e6 mg/l) from the HDPE containers and/or the filters used. Because the amount released was not constant over time and there was no clear trend with time, it is possible that individual sample DOC concentrations reported are either slightly higher or slightly lower than actual; however, the data are useful for assessing trends over time and differences between the sediment sizes, especially where concentrations are much greater than that found in the blanks. QA/QC was acceptable for the AVS measurements with a calibration verification mean value of 111 14% and a spike recovery mean of 92 15%. Only the 4-month small sized sediment sample had an AVS concentration that exceeded the method detection limit of 0.1 mg AVS/g dry sediment.
3.2.
Metals, pH, Eh, SO2 4 , and DOC
Concentrations of dissolved metals released from the sediments to the overlying water (on a dry sediment mass basis) over time are shown in Fig. 2 for the 63-mm x < 2-mm size fraction (larger sized sediment) and in Fig. 3 for the x < 63-mm size fraction (smaller sized sediment). For the Mn and Zn, the initial concentration in the stream water also is plotted (normalized to the mass of dry sediment used for each size fraction); no dissolved Cu or Fe was present in the initial stream water. Sample concentrations above this line represent metal released from the sediments. Eh and pH in the sediment slurries over time and sulfate and DOC released from the sediments (on a dry sediment mass basis) are shown in Fig. 4 for both sediment size fractions.
3.3.
331
Sediments
The 63-mm x < 2-mm size fraction comprised 92% of the total sediment mass, with 8% present in the smaller fraction (<63mm). The mean size of particles was 1-mm and 23-mm for the bulk 63-mm x < 2-mm and x < 63-mm samples, respectively. The mean percentages of each metal present in each of the extraction phases are shown in Figs. 5 and 6 for the 63- mm x < 2-mm and x < 63-mm size fractions, respectively.
3.4.
Statistical analysis
Data were not normally distributed (AndersoneDarling normality test, using Minitabª Release 13.1); thus, non-parametric statistics were used. The KruskaleWallis test evaluated statistically significant differences in dissolved metals concentrations released between the filtration sizes within each sediment size fraction. The p-values are shown in Table 1. The effect of sediment size on each of the chemical parameters and on metals concentrations was assessed using the ManneWhitney test; p-values are shown in Table 2. Statistical p-values of <0.05 indicate that the null hypothesis (no difference) is not true at the 95% significance level and thus, that there is a statistically significant difference between the data sets.
4.
Discussion
4.1.
Metals, pH, Eh, SO2 4 , and DOC
No Cd was released from either of the sediment size fractions at any time point. Cu concentrations released from the larger sized sediment were low and the highest was only about 20% greater than the RL, except for the 28-day sample. In some cases, differing filter sizes were very close to the RL, with some falling above it and some falling below. This resulted in some time points having only 1 or 2 filtration sizes having a plotted value in Fig. 2. After 5 weeks, all Cu was re-sequestered and no additional release was observed. Release of Cu from the smaller sized sediment increased until day 5, after which concentrations began to decrease until no further Cu was present in the overlying water at or beyond 3 weeks. The amount of released Fe increased significantly after 1 week from the smaller sized sediment and after 3 weeks from the larger and then was observed to be re-sequestered over time, but concentrations never reached zero. The 16-week sample obtained from the 63-mm x < 2-mm sized sediment showed an increase in Fe and Mn released. This sample was observed to have an orange color similar to samples obtained at earlier times (e.g., <1 month) and different from the later samples, which had a darker brownish color. Although it is uncertain why this sample appeared this way, it is possible this was an anomaly due to inadvertent introduction of oxygen into the chamber during use, and that time in excess of 16 weeks might show further decreases in both Fe and Mn concentrations. In the smaller sized sediment, the 12-week sample showed a much greater concentration of released Fe, relative to the other samples. This sample also showed greater concentrations of released sulfate, Mn, and Zn, and had
332
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 2 8 e3 3 6
Copper
Iron 160
1.5 μm
Conc. (mg/kg dry sed)
Conc. (mg/kg dry sed)
0.25
0.45 μm
0.2
0.2 μm
0.15 0.1 0.05
1.5 μm
140
0.45 μm 0.2 μm
120 100 80 60 40 20
0
0 0
20
40
60
80
100
120
0
20
40
Time (days)
160 140 120 100 80 60 40 20 0 20
40
60
80
100
120
Zn 1.5 μm 0.45 μm 0.2 μm Initial Stream Water
Conc. (mg/kg dry sed)
Conc. (mg/kg dry sed)
Manganese 180
0
60
Time (days)
80
100
120
Time (days)
10 9 8 7 6 5 4 3 2 1 0
1.5 μm 0.45 μm 0.2 μm Initial Stream Water
0
20
40
60
80
100
120
Time (days)
Fig. 2 e Masses of Cd, Cu, Fe, and Mn (filtered at 0.2 mm, 0.45 mm, and 1.5 mm) released to the overlying water from the 63mm £ x < 2-mm sized sediment; initial amounts present in the water are shown by solid lines for Mn and Zn (converted to a dry mass sediment basis); error bars represent propagated instrumental standard deviations (triplicate measurements) and estimated analytical errors.
a lower pH, relative to surrounding sampling times. It is uncertain why this was the case. Mn and Zn were present in the overlying water. Zn concentrations in both sediment sizes at t0 were less than that in the stream water used, indicating an immediate sequestration of Zn by the sediments; this also was observed for Mn in the larger sized sediment sample. This most likely was due to sorption and/or coprecipitation with metal oxides. After this, there was an increase in concentration indicating a release of Zn, with re-sequestration again occurring after 3 days in the larger sized sediment samples and after 5 days in the smaller. No Zn was present in the overlying water after 5 weeks in the larger sized sediment sample; however, a small amount remained in the smaller sized sample even at 16 weeks. Mn released from the larger sized sediment increased steadily until 28 days, after which there was re-sequestration, although never to levels below what was initially present in the stream water. Mn released from the smaller sized sediment increased until 5 days, with some re-sequestration at 7 days, followed by more release until 21 days. After 21 days, there was again re-sequestration, with a concentration at 16 weeks less than what was initially present in the stream water. The pH decreased initially, but then increased in the water overlying both sediment size fractions after the first week, with values approximately one-half unit higher in the smaller sized sediment samples. The increase in pH over time in both size fractions likely was due to the release of hydroxide ions and/or uptake of protons from the reductive dissolution of HFO (would be similar for HMO) and the reduction of sulfate to bisulfide, as shown in Eqs. (2)e(4).
FeðOHÞ3 þe 4Fe2þ þ 3OH
(2)
2FeðOHÞ3 þ3HS þ 3Hþ 42FeS þ S0 þ 6H2 O
(3)
þ SO2 4 þ 9H þ 8e 4HS þ 4H2 O
(4)
Only the t0 samples had positive Eh values, and after the first 7 days, there was little difference in Eh values between the sediment size fractions. DOC was both released and re-sequestered over time for both sizes of sediment, with re-sequestration beginning after 35 and 28 days for the smaller and larger sized sediment, respectively. Although the concentrations decreased after these times, they never reached zero. Sulfate concentrations increased until 21 and 42 days, for the smaller and larger sized sediment, respectively. Concentrations from both sediment sizes decreased over time; however, it is uncertain at what time past the 16 weeks the values would have reached zero. The mechanism for release of metals likely was reductive dissolution of the oxides present (e.g., HFO and/or HMO, releasing Fe and Mn) with concurrent release of sorbed metals, such as Cu, Zn, and possibly Mn. The release of sulfate and DOC over time likely also was due to their association with the reductively dissolved minerals. There was a strong correlation between concentrations of Fe and SO2 4 (r ¼ 0.90), DOC and sulfate (r ¼ 0.81), and a moderate correlation between concentrations of Fe and DOC (r ¼ 0.71). These correlations suggest the possibility of mineral phases having Fe, SO2 4 , and DOC associated with them e either as part of the
333
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 2 8 e3 3 6
Copper
Iron 700
1.5 μm
0.7
0.45 μm
Conc. (mg/kg dry sed)
Conc. (mg/kg dry sed)
0.8
0.2 μm
0.6 0.5 0.4 0.3 0.2 0.1
1.5 μm 0.45 μm
600
0.2 μm
500 400 300 200 100
0
0 0
20
40
60
80
100
120
0
20
40
Time (days) Manganese
Conc. (mg/kg dry sed)
Conc. (mg/kg dry sed)
120 100 80 60 40 20 0 20
40
60
80
100
120
Zinc 1.5 μm 0.45 μm 0.2 μm Initial Stream Water
140
0
60
Time (days)
80
100
10 9 8 7 6 5 4 3 2 1 0
1.5 μm 0.45 μm 0.2 μm Initial Stream Water
0
120
20
40
60
80
100
120
Time (days)
Time (days)
Fig. 3 e Masses of Cd, Cu, Fe, and Mn (filtered at 0.2 mm, 0.45 mm, and 1.5 mm) released to the overlying water from the x < 63-mm sized sediment; initial amounts present in the water are shown by solid lines for Mn and Zn (converted to a dry mass sediment basis); error bars represent propagated instrumental standard deviations (triplicate measurements) and estimated analytical errors.
Eh 200
8.00
100
≤x < 2-mm 63-μm x < 63-μm
0
7.50
mV
pH
pH 8.50
7.00
-100 -200
6.50
-300
≤x < 2-mm 63-μm x < 63-μm
6.00
-400 0
20
40
60
80
100
120
0
20
40
Time (days) Sulfate
80
100
120
DOC
3000 2500
≤x < 2-mm 63-μm
400
x < 63-μm
350
Conc. (mg/kg dry sed)
Conc. (mg/kg dry sed)
60
Time (days)
2000 1500 1000 500
≤x < 2-mm 63-μm x < 63-μm
300 250 200 150 100 50 0
0 0
20
40
60
Time (days)
80
100
120
0
20
40
60
80
100
120
Time (days)
Fig. 4 e Eh (normalized to the standard hydrogen electrode) and pH in overlying water and masses of sulfate and DOC released from the sediments (diamonds [ 63-mm £ x < 2-mm; squares [ x < 63-mm); error bars represent propagated estimated analytical errors and standard deviations for replicates at days 1 and 35 (63-mm £ x < 2-mm) and days 5 and 84 (x < 63-mm).
334
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 2 8 e3 3 6
Table 1 e Statistical summary (KruskaleWallis) for differences between filtration pore sizes within each sediment size fraction over time (n [ 20; hypothesis test: small [ large vs. small s large; p £ 0.05 indicates a statistically significant difference). Sediment size fraction
Fig. 5 e Percentages of the total extracted Cd, Cu, Fe, Mn, and Zn in each of the five sequential extraction phases for the 63-mm £ x < 2-mm sized.
mineral structure and/or sorbed to the mineral’s surface. Identification of the minerals in the sediment samples was outside the scope of this work; however, HFO minerals that previously have been observed in NFCC include ferrihydrite, goethite, chlorite, and schwertmannite (Butler et al., 2009). As well as being part of the mineral structure of schwertmannite, significant amounts of sulfate also may be found sorbed to its surface (Bigham et al., 1990; Jo¨nsson et al., 2005; Kumpulainen et al., 2008), and dissolved organic carbon may be sorbed to schwertmannite and goethite (Jo¨nsson et al., 2006; Kumpulainen et al., 2008). Mechanisms for re-sequestration might be 1) sorption onto existing solid phases (those not reductively dissolved); 2) precipitation as metal sulfides, metal hydroxides (specifically FeOH2), and/or metal carbonates, in particular Fe and Mn (alkalinity in the initial overlying water was 11.5 mg/l as CaCO3); and/or 3) sorption of the metals to precipitated phases. Based on solubility, the expected formation of metal sulfides over time would follow the series: CuS > ZnS > FeS > MnS (Chang, 1991). Cu and Zn re-sequestration followed this trend, but re-sequestration of Mn occurred at earlier times than for the Fe, especially in the smaller sized sediment samples.
Fig. 6 e Percentages of the total extracted Cd, Cu, Fe, Mn, and Zn in each of the five sequential extraction phases for the x < 63-mm sized.
Metal
p-Value
63-mm x < 2-mm
Cu Fe Mn Zn
0.976 0.999 0.959 0.984
x < 63-mm
Cu Fe Mn Zn
0.982 0.934 0.972 0.963
There are several potential reasons why dissolved Fe and Mn remained in solution much longer than did the Cu and Zn. It is possible that the Cu and Zn were being released as well, but at concentrations in equilibrium with the sulfide being produced, which resulted in no measurable dissolved Cu and/ or Zn and no sulfide remaining for complexation with the Fe and/or Mn. It also is possible that the rates of reductive dissolution of HFO and HMO could be faster than the rates of FeS and MnS formation, or the formation of other possible Fe and Mn precipitates. Another alternative is that the reduction of sulfate to sulfide in the system was kinetically constrained, resulting in insufficient sulfide available at those sampling times for complete metal sulfide formation. Assuming complete reduction of the maximum amounts of sulfate released (3.9 and 2.3 mM from the smaller and the larger sized sediment, respectively) to sulfide (i.e., H2S and/or HS) and precipitation of the maximum concentration of each metal released as metal sulfides, there would be 3.6 and 2.1 mM sulfide remaining in solution overlying the smaller and larger sized sediment, respectively. Considering these numbers, it does not appear that the experimental system was sulfate-limited.
Table 2 e Statistical summary (ManneWhitney) for differences between sediment size fractions over all times sampled and all filtration sizes (n [ 60; p £ 0.05 indicates a statistically significant difference; hypothesis test: small [ large vs. small s large). Parameter
pValue
Observed differences between sediment sizes
Cu Fe Mn
More released from smaller sized sediment More released from smaller sized sediment More released from larger sized sediment
Zn
0.0003 0.0001 0.0571 0.0285a 0.0004
pH
0.0102
Eh
0.3168
Sulfate DOC
0.0193 0.0149
Sequestration of Zn present in initial water was greater in the larger sized sediment slurries (i.e., less in solution) Higher over time in water overlying smaller sized sediment Similar values in water overlying both sediment sizes More released from smaller sized sediment More released from smaller sized sediment
a Alternate hypothesis large > small
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 2 8 e3 3 6
4.2.
Sediments
The sediments developed a darker color over time (initially were reddish-orange, typical of iron hydroxides), with the <63-mm fraction appearing blacker and the 63-mm x < 2mm size fraction appearing browner as time passed. This was suspected to be due to the formation of metal sulfide solids as AVS (generally colored dark brown to black), but because of method detection limitations and high variability between replicate analyses, this could not be confirmed quantitatively except for the x < 63 mm size fraction sample collected at 16 weeks, which had 0.102 mg AVS/g dry sediment. Carbonates are minimal in this system and the carbonate extraction step may remove some metals associated with iron/manganese oxide phases (Gleyzes et al., 2002); a slight orange color was observed in the extract from the carbonate extraction step for the smaller sized sediment, perhaps indicative of this. For both size fractions of sediment, approximately 50% of each metal was associated with the Fe/Mn fraction. Cd differed from the other metals by having approximately equal amounts associated with each fraction for the smaller sized sediment and a significant amount (31%) associated with the organic fraction of the larger sized sediment. Twenty percent more Fe was associated with the residual fraction in the larger sized sediment vs. the smaller. This might have contributed to the observed slower release of Fe into the stream water as these materials were reductively dissolved, combined with the mass transfer limitation of the smaller specific surface area of the more coarse particles. The RPD for recovery (mass balance) of each metal was < 33%, except for Cd in the larger sized sediment, which was not detected by ICP following total acid digestion. The total acid-recoverable samples were darkly colored reddish/orange, indicating a high concentration of ferric iron; thus the samples required dilution (1:20) prior to being analyzed via ICP-AES. Because Cd concentrations were generally low in all extracts (10e60 ppb), this dilution resulted in there being no Cd detected in the triplicate samples.
5.
335
Conclusions
1. This study has shown that metals associated with aerobic sediments are released into the dissolved phase when held under anaerobic conditions. It is believed that this was because of reductive dissolution of metal oxy-hydroxysulfate minerals with the concurrent release of chemically or physically associated metals. 2. Under the observed pH and Eh conditions, coupled with the release and then loss of sulfate over time, it is probable that the metals were re-sequestered as metal sulfides, although some may have precipitated as carbonates and/or hydroxides. The observed darker color of the sediment samples over time was believed to be due to the formation of metal sulfides. It was unfortunate, however, that the measured concentrations of AVS were too low to allow confirmation and quantification of sulfide precipitates. 3. Releases of Fe and Mn did not follow the same trends as Cu and Zn. This is most likely because Fe and Mn were associated with the mineral composition of the sediment and not just associated by sorption and/or ionic exchange processes. Fe and Mn remained in the dissolved phase (presumably in their reduced forms at the measured Eh values) at much higher concentrations and for a much longer time than did the other metals. This might allow them to enter the groundwater, if the repository is not lined and/or capable of capturing leached metals before their reaching the water table. 4. Overall, results suggest that remediation of miningimpacted streambed sediments should include precautions to minimize and/or eliminate the potential for metal release from these sediments if stored under physical and/or chemical conditions differing from their point of origin. A potential remedy would be disposal into a lined repository with capture (and treatment) of any leachate or capping to eliminate infiltration of water into the sediments.
Acknowledgements
4.3.
Statistical interpretation
Although there were some visible differences (generally <15%; see Figs. 2 and 3) in concentrations of metals between filter pore sizes, these were not statistically significant for any of the metals analyzed, with all p-values >0.93 (a ¼ 0.05). That statistically significantly more Cu, Fe, and Zn concentrations were released from the smaller sized sediment may be explained by the fact that smaller particles have a larger surface area and thus, sorb a higher concentration of metals available for later release. Additionally, the solubility of any material is increased with decreasing particle size (Pankow, 1991), especially amorphous minerals. In NFCC, as in many mining-impacted streams, high amounts of amorphous metal oxide precipitates are present in both the suspended solids phase and in streambed sediments.
The author thanks Shaw Environmental, a contractor to the United States Environmental Protection Agency (U.S. EPA), for DOC analyses; Dr. Christopher Impellitteri for sulfate analyses, Dr. Matthew Morrison for technical advice, and Larry Wetzel for sediment mixing and sieving assistance, each from U.S. EPA National Risk Management Research Laboratory; Dr. James Ranville from the Colorado School of Mines for assistance with sediment sampling; Warren Boothman from U.S. EPA National Health and Environmental Effects Laboratory, Atlantic Ecology Division for AVS analyses; and David Reisman from U.S. EPA National Risk Management Research Laboratory and anonymous reviewers for very helpful comments. The U.S. EPA, through its Office of Research and Development, funded the research described in this manuscript. It has been administratively reviewed and approved for publishing. Citations of product, company, or trade names do not constitute endorsement by the U.S. EPA and are provided only for the purpose of better describing information in this manuscript.
336
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 2 8 e3 3 6
references
Bigham, J.M., Schwertmann, U., Carlson, L., Murad, E., 1990. A poorly crystallized oxyhydroxysulfate of iron formed by bacterial oxidation of Fe(II) in acid mine waters. Geochimica et Cosmochimica Acta 54 (10), 2743e2758. Butler, B.A., 2009. Effect of pH, ionic strength, dissolved organic carbon, time, and particle size on metals release from mine drainage impacted streambed sediments. Water Research 43 (5), 1392e1402. Butler, B.A., Ranville, J.F., Ross, P.E., 2009. Spatial variations in the fate and transport of metals in a mining-influenced stream, North Fork Clear Creek, Colorado. Science of the Total Environment 407 (24), 6223e6234. Chang, R., 1991. Chemistry, fourth ed. McGraw-Hill, Inc., New York, New York, 721 pp. Cunningham, C.G., Naeser, C.W., Marvin, R.F., Luedke, R.G., Wallace, A.R., 1994. Ages of selected intrusive rocks and associated ore deposits in the Colorado mineral belt. In: U.S. Geological Survey Bulletin, No. 2109. Davis, J.A., Leckie, J.O., 1978. Surface ionization and complexation at the oxide/water interface: II. Surface properties of amorphous iron oxyhydroxide and adsorption of metal ions. Journal of Colloid and Interface Science 67, 90e107. Davranche, M., Bollinger, J.-C., 2000. Heavy metals desorption from synthesized and natural iron and manganese oxyhydroxides: effect of reductive conditions. Journal of Colloid and Interface Science 227, 531e539. Dong, D., Derry, L.A., Lion, L.W., 2003. Pb scavenging from a freshwater lake by Mn oxides in heterogeneous surface coating materials. Water Research 37, 1662e1666. Gleyzes, C., Tellier, S., Astruc, M., 2002. Fractionation studies of trace elements in contaminated soils and sediments: a review of sequential extraction procedures. Trends in Analytical Chemistry 21, 451e467. Grybos, M., Davranche, M., Gruau, G., Petitjean, P., 2007. Is trace metal release in wetland soils controlled by organic matter mobility or Fe-oxyhydroxide reduction? Journal of Colloid and Interface Science 314, 409e501. Jenne, E.A., 1968. Controls on Mn, Fe, Co, Ni, Cu, and Zn concentrations in soils and water: the significant role of hydrous Mn and Fe oxides. In: Gould, R.F. (Ed.), Trace Inorganics in Water. Advances in Chemistry 73. American Chemical Society, Washington, DC, pp. 337e387. Jo¨nsson, J., Persson, P., Sjo¨berg, S., Lo¨vgren, L., 2005. Schwertmannite precipitated from acid mine drainage: phase transformation, sulphate release and surface properties. Applied Geochemistry 20 (1), 179e191. Jo¨nsson, J., Sjo¨berg, S., Lo¨vgren, L., 2006. Adsorption of Cu(II) to schwertmannite and goethite in presence of dissolved organic matter. Water Research 40, 969e974. Karthikeyan, K.G., Elliott, H.A., Cannon, F.S., 1997. Adsorption and coprecipitation of copper with the hydrous oxides of iron and aluminum. Environmental Science and Technology 31, 2721e2725. Kumpulainen, S., von der Kammer, F., Hofmann, T., 2008. Humic acid adsorption and surface charge effects on schwertmannite and goethite in acid sulphate waters. Water Research 42, 2051e2060. Nordstrom, D.K., Wilde, F.D., September 2005. Reduction Oxidation Potential (Electrode Method) (Version 1.2): U.S. Geological Survey Techniques of Water-Resources Investigations (Book 9, Chapter A6, Section 6.5) [WWW document]. URL: http://pubs.water.usgs.gov/twri9A6 (accessed 12.03.09). Pankow, J.F., 1991. Aquatic Chemistry Concepts. Lewis Publishers, Chelsea, Michigana (Chapter 15).
Pareuil, P., Pe´nilla, S., Ozkan, N., Bordas, Franc¸ois, Bollinger, J.-C., 2008. Influence of reducing conditions on metallic elements released from various contaminated soil samples. Environmental Science and Technology 42 (20), 7615e7621. Schemel, L.E., Kimball, B.A., Bencala, K.E., 2000. Colloid formation and metal transport through two mixing zones affected by acid mine drainage near Silverton, Colorado. Applied Geochemistry 15, 1003e1018. Stumm, W., 1992. Chemistry of the Solidewater Interface: Processes at the Mineralewater and Particleewater Interface in Natural Systems. John Wiley & Sons, New York, New York, 428 pp. Tessier, A., Campbell, P.G.C., Bisson, M., 1979. Sequential extraction procedure for the speciation of particulate trace metals. Analytical Chemistry 51 (7), 844e851. Turner, A., Millward, G.E., Le Roux, S.M., 2004. Significance of oxides and particulate organic matter in controlling trace metal partitioning in a contaminated estuary. Marine Chemistry 88, 179e192. U.S. EPA, 1991. Draft Analytical Method for Determination of Acid Volatile Sulfide in Sediment. U.S. EPA Office of Science and Technology, Washington, DC, EPA-821-R-91-100, 22 pp. U.S. EPA, 1992. Method 3005a. Acid digestion of waters for total recoverable or dissolved metals for analysis by FLAA or ICP spectroscopy. Revision 1 [WWW document]. URL: In: SW-846. Test Methods for Evaluating Solid Waste, Physical/Chemical Methods http://www.epa.gov/osw/hazard/testmethods/ sw846/online/3_series.htm (accessed 3.12.09). U.S. EPA, 2000. Method 300.1. Determination of Inorganic Anions in Drinking Water by Ion Chromatography. Revision 1 [WWW document]. URL: http://www.epa.gov/waterscience/methods/ method/files/300_1.pdf (accessed 3.12.09). U.S. EPA, 2001. Second Five-year Review Report for California Gulch. Prepared for EPA Region 8 by TechLaw, Inc., Denver, CO. Superfund Site ID #COD980717938[WWW document]. URL:. EPA, Washington, D.C. http://www.epa.gov/superfund/ sites/fiveyear (accessed 3.12.09). U.S. EPA, 2004. First Five-year Review Report for Kennecott South Zone. Superfund Site ID #UTD000826404. [WWW document]. URL:. EPA, Washington, D.C. http://www.epa.gov/superfund/ sites/fiveyear (accessed 3.12.09). U.S. EPA, 2005a. Second Five-year Review for the Bunker Hill Mining and Metallurgical Complex Superfund Site Operable Units 1, 2, and 3: Idaho and Washington. Prepared by EPA Region 10, Seattle, WA, Superfund Site ID # IDD048340921 [WWW document]. URL: EPA, Washington, D.C. http://www. epa.gov/superfund/sites/fiveyear (accessed 3.12.09). U.S. EPA, 2005b. Second Five-year Review Report for Silver Bow Creek/Butte Area Superfund Site. Prepared for EPA Region 8 by CDM, Helena, MTSuperfund Site ID # MTD980502777. [WWW document]. URL: EPA, Washington, D.C. http://www.epa.gov/ superfund/sites/fiveyear (accessed 3.12.09). U.S. EPA, 2005c. Method 415.3. Determination of Total Organic Carbon and Specific UV Absorbance at 254 nm in Source Water and Drinking Water, Rev 1.1. EPA-600-R-05-055. [WWW document]. URL: http://www.epa.gov/nerlcwww/m_415_ 3Rev1_1.pdf (accessed 12.12.07). U.S. EPA, 2007. Microwave Assisted Acid Digestion of Sediments, Sludges, Soils, and Oils, Revision 1 [WWW document] URL: http://www.epa.gov/osw/hazard/testmethods/sw846/online/ 3_series.htm (accessed 3.12.09). Webster, J.G., Swedlund, P.J., Webster, K.S., 1998. Trace metal adsorption onto an acid mine drainage iron (III) oxy hydroxy sulfate. Environmental Science and Technology 32, 1361e1368. Wildeman, T.R., Cain, D., Ramiriz, A.J., 1974. The relation between water chemistry and mineral zonation in the Central City Mining District, Colorado. In: Water Resources Problems Related to Mining, Proc. 18. American Water Resources Association, Minneapolis, Mn, pp. 219e229.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 3 7 e3 4 7
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Modeling DBPs formation in drinking water in residential plumbing pipes and hot water tanks Shakhawat Chowdhury a,*, Manuel J. Rodriguez a, Rehan Sadiq b, Jean Serodes c E´cole supe´rieure d’ame´nagement du territoire, Universite´ Laval, 1628 Pavillon Fe´lix-Antoine-Savard, Que´bec City, QC, Canada G1V 0A6 Associate Professor, School of Engineering, University of British Columbia e Okanagan Kelowna, Canada BC V1V 1V7 c De´partement de Ge´nie Civil, Universite´ Laval, 1040 Pavillon Pouliot, Universite´ Laval, Quebec City, QC, Canada G1V 0A6 a
b
article info
abstract
Article history:
Disinfection byproducts (DBPs) in municipal supply water are a concern because of their
Received 30 January 2010
possible risks to human health. Risk assessment studies often use DBP data in water
Received in revised form
distribution systems (WDS). However, DBPs in tap water may be different because of
28 July 2010
stagnation of the water in plumbing pipes (PP) and heating in hot water tanks (HWT). This
Accepted 4 August 2010
study investigated occurrences and developed predictive models for DBPs in the PP and the
Available online 11 August 2010
HWT of six houses from three municipal water systems in Quebec (Canada) in a yearround study. Trihalomethanes (THMs) in PP and HWT were observed to be 1.4e1.8 and 1.9
Keywords:
e2.7 times the THMs in the WDS, respectively. Haloacetic acid (HAAs) in PP and HWT were
Drinking water
observed to be variable (PP/WDS ¼ 0.23e2.24; HWT/WDS ¼ 0.53e2.61). Using DBPs occur-
Disinfection byproducts (DBPs)
rence data from these systems, three types of linear models (main factors; main factors,
DBP changes
interactions and higher orders; logarithmic) and two types of nonlinear models (three
Plumbing pipes
parameters Logistic and four parameters Weibull) were investigated to predict DBPs in the
Hot water tanks
PP and HWT. Significant factors affecting DBPs formation in the PP and HWT were iden-
Modeling
tified through numerical and graphical techniques. The R2 values of the models varied
Exposure to DBPs
between 0.77 and 0.96, indicating excellent predictive ability for THMs and HAAs in the PP and the HWT. The models were found to be statistically significant. The models were validated using additional data. These models can be used to predict DBPs increase from WDS (water entry point of house) to the PP and HWT, and could thereby help gain a better understanding of human exposure to DBPs and their associated risks. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Reactions between natural organic matter (NOM) and chlorine form different types of disinfection byproducts (DBPs) in drinking water. These include trihalomethanes (THMs), haloacetic acids (HAAs), haloacetonitriles (HANs), haloketones (HKs), nitrosamines and other known and unknown byproducts (Singer, 1994; Singer et al., 2002; USEPA, 2006; Richardson et al., 2007, 2008; Health Canada, 2008; Hrudey,
2009). Human exposure to DBPs can occur through ingestion of drinking water as well as inhalation and dermal contact during regular indoor activities (e.g., showering, bathing, swimming in chlorinated pools and cooking). Some of these DBPs can be associated with human bladder cancer and other chronic and sub-chronic effects on human health (Lynch et al., 1989; King and Marrett, 1996; Cantor et al., 1998; Villanueva et al., 2004). A number of past studies have predicted cancer risks from exposure to DBPs in drinking water
* Corresponding author. Tel.: þ1 418 656 2131x6763; fax: þ1 418 656 2018. E-mail address:
[email protected] (S. Chowdhury). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.002
338
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 3 7 e3 4 7
(Jo et al., 1990a, b; Lin and Hoang, 2000; Lee et al., 2004; King et al., 2004; Xu and Weisel, 2005; Uyak, 2006; Nuckols et al., 2005; Savitz et al., 2006; Semerjian and Dennis, 2007; Fristachi and Rice, 2007; Chowdhury and Champagne, 2009). However, most of these studies considered DBP concentrations in the water distribution systems (WDS) for risk assessment purposes. Consumers typically use supply water from taps in houses, while the regulatory and monitoring agencies recommend measuring DBPs at various points within the WDS (e.g., Drinking Water Surveillance Program in Ontario, U.S. Environmental Protection Agency, Health Canada). Depending on the size of the plumbing systems, water may stay in the plumbing pipes for a considerable amount of time before it reaches taps in the house (Baribeau et al., 2004; DionFortier et al., 2009). This stagnation may be even longer during off-peak hours (e.g., midnight to morning; late morning to early evening). The stagnation of water in the plumbing pipes (PP) may allow additional reactions between the residual organics and free residual chlorine (Sadiq and Rodriguez, 2004), which may increase DBP concentrations in the PP. The increased DBPs pose additional risks to human health. Likewise, some previous estimates of exposure and risk to human health during showering and bathing have assumed that the concentrations of THMs in warm water are equivalent to those of cold water. Warm water (35 e45 ) is typically used for showering (Jo et al., 1990a, b; Xu et al., 2002), and warming the supply water during showering can increase THMs formation significantly (Weisel and Chen, 1994; Al-Omari et al., 2004; Chowdhury and Champagne, 2009). Recent research has demonstrated that DBP concentrations can increase significantly between water distribution systems and the consumer’s tap due to stagnation in the plumbing pipes (PP) and hot water tanks (HWT) (Wu et al., 2001; Weinberg et al., 2006; Dion-Fortier et al., 2009). As such, use of supply water (consumption and/or showering purposes) after a long period of stagnation in the PP and/or heating in the HWT may significantly increase risks to human health. Consequently, studies associated with DBPs in the WDS may not adequately represent DBPs exposure and their risks to human health. A number of studies have investigated DBPs increase in the PP and HWT (Weisel and Chen, 1994; Wu et al., 2001; Baribeau et al., 2004; Weinberg et al., 2006; Dion-Fortier et al., 2009). A recent study predicted THMs increases due to warming of water during showering based on limited data on the rates of THMs increases from warming of chlorinated water (Chowdhury and Champagne, 2009). However, this study considered limited data (e.g., less than 10 data points) from a laboratory-scale experiment in Jordan (Al-Omari et al., 2004), while data from the real municipal water systems may be different. In addition, this study did not investigate changes in DBPs due to the stagnation of the supply water in the PP and/ or heating in the HWT. To date, no study has reported any model to predict DBPs formation in the PP and HWT. The availability of such models is critical in understanding human exposure to DBPs, associated risks, epidemiological correlations, status of water treatment and disinfection processes, necessity of chlorine-boosting stations and regulatory guidelines and limitations.
In this study, occurrences of THMs and HAAs in the WDS, PP and HWT are investigated in six houses supplied by three municipal water systems in the Province of Quebec, Canada (Quebec City, Beauport and Danville). Possible effects of water stagnation in the PP and HWT are characterized. Locational and seasonal variability of THMs and HAAs are investigated. Using data for these municipal water systems, three types of linear models (main factors; main factors, interactions and higher orders; logarithmic) and two types of nonlinear models (three parameters Logistic and four parameters Weibull) are investigated to predict THMs and HAAs in the PP and HWT. The significant factors affecting THMs and HAAs formation in the PP and HWT are identified through numerical and graphical techniques. Statistical adequacies of the models are investigated using available approaches. Models were validated using additional data. Finally, implications of DBPs changes in the PP and HWT are discussed.
2.
Material and methods
2.1.
Sampling program
The properties of municipal water entering the house from the WDS and occurrences of DBPs within the WDS, PP and HWT were investigated for three municipal water systems in Quebec, namely, the Quebec City (Q1), Beauport (Q2) and Danville (Q3) municipal water systems. Water samples were collected from February 2007 to January 2008 through eight sampling programs (February, April, May, July, August, October, November and January) for two houses served by each municipality (i.e., six sampling locations). At each sampling location, seven types of samples were collected at three different times during the day. Samples were taken by the order of: (i) water entry points of the houses (WDS) after the last use of water in late evening (WDS-E); (ii) cold water samples in the early morning before the first use of water (PPeMCW); (iii) hot water samples in the morning (HWTMHW); (iv) WDS (water entry points of the houses) in the morning (WDS-M); (v) cold water samples in the afternoon before water use (PPeACW); (vi) hot water samples in the afternoon (HWT-AHW); and (vii) WDS (water entry points of the houses) in the afternoon (WDS-A). For the samples of hot water, a temperature of approximately 55 C was maintained following regulation to prevent risks of burn injuries to children. Further details on the sampling program and/or experimental approaches can be found in Dion-Fortier el al., (2009) and USEPA, (1995a, b).
2.2.
Sample analysis
In this study, THMs, HAAs, free residual chlorine, total chlorine, UV absorbance (UV254), total organic carbon (TOC), temperature, pH, turbidity and conductivity were measured for each water sample. Samples for measuring THMs and HAAs were taken in 40 mL vials containing a dechlorinating agent (ammonium chloride) and the samples for pH, turbidity, UV254, TOC and conductivity were collected in 125 mL plastic bottles. The samples were transported to the laboratory in a cooler (4 C). Free residual chlorine, total chlorine and water
339
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 3 7 e3 4 7
temperature were measured in-situ. THMs were measured by gas chromatography equipped with mass spectroscopy detection (GCeMS) (Varian chromatograph, model 3900 equipped with quadrupole mass spectrometer). The analysis was conducted according to the USEPA method 524.2 (USEPA, 1995a). The analysis of HAAs was conducted according to USEPA method 552.2 (USEPA, 1995b) using a gas chromatography with electron capture detector (GC-ECD) (Perkin Elmer Chromatograph, model AutoSystem XL). The detection limits for THM species were 0.3 mg/L for chloroform, 0.3 mg/L for BDCM (bromodichloromethane), 0.4 mg/L for DBCM (dibromochloromethane) and 0.5 mg/L for bromoform. The detection limits for HAA species were 1.3 mg/L for MCAA (monochloroacetic acid), 0.9 mg/L for DCAA (dichloroacetic acid), 0.4 mg/L for TCAA (trichloroacetic acid), 1.0 mg/L for MBAA (monobromoacetic acid), 0.7 mg/L for DBAA (dibromoacetic acid), 0.8 mg/L for BCAA (bromochloroacetic acid), 4.6 mg/L for CDBAA (chlorodibromoacetic acid), 4.2 mg/L for BDCAA (bromodichloroacetic acid) and 6.4 mg/L for TBAA (tribromoacetic acid). Free residual chlorine and total chlorine were measured using a Hach colorimeter (model DR-820) and a titrimetric method (Standard Methods 4500-Cl G). Turbidity was measured with a turbidimeter (Hach, model 2100N). UV254 was measured with a spectrophotometer (Pharmacia, model 80-2097-62) at 254 nm with a 50 mm optical path quartz cell. TOC was measured with a Shimadzu analyzer (model TOC-5000) following 5310B Standard method.
2.3.
Model development
Three types of linear models (main factors; main factors, interactions and higher orders; logarithmic) and two types of nonlinear models (three parameters Logistic and four parameters Weibull) were investigated to predict THMs and HAAs in the PP and HWT. The main factor linear models are the simplest form of multiple linear models (Montgomery and Runger, 2007). These models are constructed using significant main factors where interaction and higher order terms are ignored. The parameters (e.g., model coefficients) and the factors (predictor variables) are linear in this model. In the main factor, interaction and higher order terms models, the significant main factors, the effect of two factors varying together (e.g., TOC and temperature) and higher order terms (e.g., quadratic, cubic) are incorporated. However, model parameters are linear. In the case of the logarithmic linear model, values of the factors are transformed into a logarithm and then linear regression is performed. The model parameters and logarithm of factors are linear in these models. The fitness and performance of the regression models are generally estimated by the coefficient of determination (R2), F Ratio, root mean square error (RMSE), significance probability, lack of fit test, Normal probability plot of residuals and residuals versus predict and data order plots (Montgomery and Runger, 2007). The nonlinear models are relatively complex compared to the linear models. The parameters of these models are nonlinear. The simplest form of nonlinear model may be constructed using two parameters and a single predictor variable. When constructing nonlinear models, initial models are defined through the observation of data plots and initial
values for the parameters are assigned. Iterations are performed on the initially defined models until desired convergence is obtained. The values of the parameters upon convergence are the model parameter values (SAS Inc., 2009). General formulations of the various models are presented in Eqs. (1)e(5), where Eqs. (1)e(5) represent the main factors’ linear model; main factors, interaction and higher order terms model; logarithmic main factors model; four parameter Weibull model; and three parameter Logistic model, respectively. Weibull and Logistic models are the non linear models. In these equations, y represents the output, b and q represents model parameters, x represents predictor variables, 3 represents residuals and i, j ¼ 1, 2, 3 . n. n X bi xi þ . þ 3 (1) y ¼ b0 þ i¼1
y ¼ b0 þ
n X
bi xi þ
i¼1
LnðyÞ ¼ b0 þ
n X
bi;iþj xi xiþj þ
i;j¼1 n X
n X
bii x2i þ . þ 3
(2)
i¼1
bi ½Lnðxi Þ þ . þ 3
(3)
i¼1
y ¼ q1 q2 Expð Exp½q3 þ q4 LnfxgÞ
(4)
y ¼ q1 =½1 þ q2 Expðq3 xÞ
(5)
Further details regarding these models, their advantages, limitations and applications can be found elsewhere (Montgomery and Runger, 2007; SAS Inc., 2009). For the models in this study, (e.g., THMs at the PP, THMs at the HWT, HAAs at the PP, and HAAs at the HWT), statistically significant main factors, interaction terms and higher order terms were identified through screening test module of JMP, Normal Probability plots of residuals, effects analysis and parameter estimates (SAS Inc., 2009). Using the significant factors, models were developed and the statistical adequacy of the models was tested through numerical and graphical approaches.
3.
Results
3.1.
Occurrences of DBPs
The properties of drinking water entering the houses from the WDS into the PP are summarized in Table 1. The total organic carbon (TOC) in the WDS of Danville (Q3) municipal system was much higher (average: 7.4 mg/L: range: 1.6e12.6 mg/L) than in the WDS of Quebec City (Q1) (average: 1.7 mg/L: range: 1.2e2.7 mg/L) and the WDS of Beauport (Q2) (average: 3.5 mg/L; range: 1.7e4.9 mg/L). The ultraviolet absorbance (UV254) at the WDS of Q3 was observed to be 0.047/cm (0.023e0.069/cm), which was much lower than Q2 (average: 0.087/cm; range: 0.049e0.14/cm). The WDS of Q1 also showed lower UV254 (average: 0.025/cm) than Q2. This study could not establish a strong correlation between UV254 and TOC (Table 1). Similar inconsistencies between UV254 and TOC are reported in HellurGrossman et al. (2001), where a strong correlation between UV254 and TOC could not be established for the Sea of Galilee (Lake Kinneret) water in Israel. The inconsistency between TOC and UV254 in this study may be attributable to the higher
340
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 3 7 e3 4 7
Table 1 e Quality of water entering into houses of the three municipal water systems under study (Values in the brackets are the standard deviations). Q1
TOC (mg/L) Temp ( C)-WDS Temp ( C)-PP Temp ( C)-HWT pH Cl2-Free (mg/L) Cl2-Total (mg/L) Turbidity (NTU) UV254 (/Cm) Conductivity (mS/Cm)
Q2
Average
Range
1.7 (0.6) 9.2 (5.6) 20 (3) 55 (1) 7.8 (0.1) 1.03 (0.3) 1.17 (0.3) 0.33 (0.11) 0.025 (0) 184 (31)
1.2e2.7 2e17.1 17e28 53e56.1 7.61e8.02 0.39e1.38 0.57e1.55 0.192e0.66 0.019e0.034 154e241
Q3
Average 3.5 8.5 19 55 7.4 1.5 1.55 0.76 0.087 76
(1) (5.7) (2) (1) (0.2) (0.4) (0.5) (0.4) (0) (7.1)
Range
Average
Range
1.7e4.9 2e16.1 14e23.1 54e56 7e7.8 0.78e2.34 0.85e2.46 0.37e1.45 0.049e0.14 65e84.2
7.4 (4.1) 9.3 (6.2) 18 (3) 57 (4) 7.4 (0.2) 1.3 (0.2) 1.54 (0.3) 0.26 (0.2) 0.047 (0) 359 (119)
1.6e12.6 2e19.2 11e22.1 54e64 7.3e7.8 0.92e1.7 1.12e1.98 0.08e0.93 0.023e0.069 178e496
Q1: Quebec City; Q2: Beauport; Q3: Danville.
fractions of hydrophilic NOM, while the hydrophilic fractions of NOM may not be adequately represented by the UV254 or specific ultraviolet absorbance (Chowdhury et al., 2010). The averages and variability of THMs and HAAs for these seven sampling scenarios are shown in Table 2. Concentrations of THMs and HAAs for all seven sampling scenarios in the Beauport (Q2) municipal system were observed to be higher than in the Quebec City (Q1) and Danville (Q3) municipal water systems (Table 2). The PP and HWT showed higher THM concentrations than those observed in the WDS in all sampling sites, while HAAs in the PP and WDS were observed to be statistically comparable in most cases (Table 2). THMs and HAAs in the WDS, PP and HWT were observed to be variable for all sampling sites. For Q1, THMs in the WDS, PP and HWT were observed to be in the ranges of 23.0e26.91, 41.52e45.13 and 66.01e67.04 mg/L, respectively, while HAAs were in the ranges of 27.84e30.92, 34.81e36.12 and 41.60e44.53 mg/L, respectively (Table 2). The Q2 and Q3 sampling locations also showed similar variability of THMs and HAAs for the WDS, PP and WT (Table 2). THMs in the PP were 136%e181% of the THMs in the WDS for all sampling locations, while THMs in the hot water tanks were 191%e269% of the THMs in the WDS and 132%e159% of the THMs in the PP in all cases. HAAs in the PP and HWT showed variable results when compared to the HAAs in the WDS. HAA levels in the PP and HWT were 23%e224% and 53%e261% of HAA levels in the
WDS, respectively. The reduction of HAAs from the WDS to the PP and HWT may be due to increased microbiological activity in the PP because of the consumption of free chlorine residuals and the destruction of HAAs within the PP by microorganisms (possible presence of a biofilm). Seasonal variability of THMs and HAAs in the WDS, PP and HWT during different periods from February 2007 to January 2008 is shown in Figs. 1 and 2, respectively. For all of the three sampling sites (Q1, Q2 and Q3), THM concentrations during warm months (e.g., August) were significantly higher than during cold months (e.g., February). For example, in February, THMs in the WDS-E of Q1, Q2 and Q3 were 10.84, 37.51 and 33.82 mg/L, respectively, while in August THMs at those locations were 41.11, 115.0 and 104.13 mg/L, respectively (Fig. 1). In February, THMs in PPeMCW from Q1, Q2 and Q3 were 21.53, 64.91 and 60.71 mg/L, respectively, while in August THMs in the same locations were 67.31, 184.0 and 133.62 mg/L, respectively. Similar results were also obtained for the THMs in the HWT in all sampling sites. These findings demonstrated that THMs formation during the warm months (e.g., August) might be much higher than those in the winter months (e.g., February) (Fig. 1). However, HAAs showed different results due to seasonal variability (Fig. 2). Limited consistency was observed in the patterns of HAAs formation in the WDS, PP and HWT of these sampling sites (Fig. 2). For example, HAAs in the WDS of Q1
Table 2 e Average THM and HAA levels for the various sampling scenarios for three municipal water systems in Quebec. Sampling Site
WDS-E
PPeMCW
HWTeMHW
WDS-M
PPeACW
HWTeAHW
WDS-A
THMs
Quebec City (Q1) Beauport (Q2) Danville (Q3)
23 (8.4) 73.4 (42.5) 62 (26.4)
45.1 (15.2) 121 (46.4) 99 (28.6)
67 (20.4) 166.5 (69.2) 130.8 (41)
24.9 (8.2) 83.8 (43.3) 68.3 (27.8)
41.5 (22.1) 116.4 (49) 87 (28.7)
66 (20.7) 168.4 (64) 124.5 (28.2)
26.9 (14.3) 76.6 (41.3) 64.1 (25.5)
HAAs
Quebec City (Q1) Beauport (Q2) Danville (Q3)
27.8 (10.5) 108.1 (53.5) 54.9 (20.2)
36.1 (12.4) 113.6 (66.9) 72.9 (29.1)
41.6 (10.7) 145.3 (58.3) 76.3 (25)
30.2 (9.8) 103 (42.2) 59 (19.2)
34.8 (12.4) 102 (62.7) 66.2 (25.6)
44.5 (11.1) 147.2 (57.1) 73.9 (23.1)
30.9 (11) 107.0 (59.1) 53.6 (19.9)
WDS-E: samples from WDS after last use of water at the late evening; PPeMCW: cold water samples in the early morning prior to the first water use; HWTeMHW: hot water samples in the morning; WDS-M: samples from the WDS in the morning; PPeACW: cold water samples in the afternoon, before water use; HWTeAHW: hot water samples in the afternoon; WDS-A: samples from the WDS in the afternoon. (Values in the brackets are standard deviations).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 3 7 e3 4 7
341
provide an indication of accelerated microbiological activity in the PP, which may consume free chlorine residuals and compromise water quality before it reaches the consumer’s tap. Microbiological activity may be higher during periods of minimum water use (e.g., midnight to early morning). The changes of THMs and HAAs from the WDS to the PP and HWT may have been affected by the contact time and water temperature within the PP and HWT. The average water temperature in the PP and HWT (for all three municipal water systems: range ¼ 11e28 C; average ¼ 18e20 C in PP and range ¼ 53e64 C and average ¼ 55e57 C in HWT) was observed to be much higher than the WDS (for all three municipal water systems: range ¼ 2e19 C and average ¼ 8.5e9.2 C). Increased water temperature in the PP and HWT may have increased reaction rates between free chlorine residuals and residual organics to form additional THMs and HAAs.
3.2.
Fig. 1 e Seasonal variability of THMs in the three municipal water systems under study (Q1: Quebec City; Q2: Beauport; and Q3: Danville). (WDS-E: samples from WDS after last use of water at the late evening; PPeMCW: cold water samples in the early morning prior to the first water use; HWTeMHW: hot water samples in the morning; WDS-M: samples from the WDS in the morning; PPeACW: cold water samples in the afternoon, before water use; HWTeAHW: hot water samples in the afternoon; WDS-A: samples from the WDS in the afternoon).
were low in February, April, May and January (16.91e23.61 mg/L) and high in July, August, October and November (28.62e42.83 mg/L). In Q2, HAAs in WDS were low between February and April (36.33e51.0 mg/L) and high from May to January (64.42e198.34 mg/L). In the WDS of Q3, HAAs were high in November and January (63.81e74.53 mg/L) and low between February and April (22.52e31.84 mg/L). In these three WDS, January showed higher HAAs for Q2 and Q3, and lower HAAs for Q1. Similar variability was also observed in May. The inconsistencies in the patterns of HAAs formation in PP and HWT were also observed. For PP and HWT of Q1, HAAs were higher between July and January than between February and May (Fig. 2). In Q3, HAAs obtained in January (PP: 96.41e113.92 mg/L; HWT: 95.42e97.31 mg/L) were observed to be the highest during the study period (Fig. 2). HAAs in the PP and HWT were occasionally observed to be less than the HAAs measured in the WDS. For example, HAAs were 39.61 and 58.84 mg/L (Aug.) at the HWT and WDS of Q3, respectively, and 77.92 and 130.33 mg/L (May) at the PP and WDS of Q2, respectively (Fig. 2). The reduction of HAAs in the PP and HWT may
Model adequacies
Statistically significant factors for each of the models were identified through numerical and graphical techniques. The factor selection procedure through effect analysis using JMP is demonstrated in Table 3. Table 3 shows the statistical significance of main factors and interaction terms on THMs formation model in the PP (model 2 in Table 4). THMs at the WDS, free residual chlorine, temperature in the PP and interaction of TOC and temperature were found to be statistically significant for modeling THMs in the PP (Table 3). Despite TOC was not identified to be a significant main factor, TOC was included in the model because the interaction of TOC and temperature was identified to be significant. Inclusion of TOC was necessary to maintain statistical consistency (Montgomery and Runger, 2007). In this sampling program, reaction time was considered as constant in all cases as the samples from the WDS were taken in the late evening after the last use of water during night and the samples from the tap water in house (representing PP) were collected in the early morning of next day prior to the first use. As the reaction time was maintained to be constant, this factor was not included in the model. The effects of parameters were also tested using graphical techniques (Normal plots, interaction plots, etc.: not shown) and the significant factors for modeling were identified. Details about the graphical techniques for the selection of significant factors can be found in Chowdhury et al. (2010). Further details on the screening analysis may be obtained from Montgomery and Runger (2007). The models for predicting THMs in the PP and HWT are summarized in Table 4. The R2 values for the linear models range between 0.77 and 0.91. The four-parameter Weibull and three-parameter Logistic models from the JMP nonlinear models library were tested with the measured data (SAS Inc., 2009). The nonlinear model parameters were estimated through obtained convergence using the Analytic GausseNewton method (SAS Inc., 2009). After fitting, their corresponding RMSE were observed to be 27.2 and 28.1 for the PP and HWT, respectively. In the linear regressions, THMs in the WDS and free chlorine residuals were found to be statistically significant; however, other factors, as well as interaction and higher order terms, were also noted to be statistically significant in a number of models (Table 4). Fig. 3 and Fig. 4
342
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 3 7 e3 4 7
Table 3 e Screening effects of the factors for modeling THMs formation (It is an example of selecting significant factors for model 2 in Table 4). Term Intercept Cl2-F pH TOC Temp-PP THM-WDS-E (Cl2-F-1.3)*(pH-7.5) (Cl2-F-1.3)*(TOC-4.3) (Cl2-F-1.3)*(Temp-PP-18.9) (Cl2-F-1.3)*(THM-WDS-55.8) (pH-7.5)*(TOC-4.276) (pH-7.5)*(Temp-PP-18.9) (pH-7.5)*(THM-WDS-55.8) (TOC-4.3)*(Temp-PP-18.9) (TOC-4.3)*(THM-WDS-55.8) (Temp-PP-18.9)* (THM-WDS-55.8)
Estimate 102.64 30.78 4.13 0.26 4.62 1.12 46.70 5.44 5.78 0.13 4.67 9.81 0.04 1.33 0.02 0.04
Std t Ratio Prob>jtj Error 132.43 8.67 16.10 1.13 1.36 0.17 35.75 4.00 3.60 0.24 5.66 6.51 0.53 0.39 0.04 0.05
0.74 3.61 0.28 0.23 3.47 6.76 1.34 1.37 1.73 0.47 0.80 1.46 0.08 3.56 0.47 0.96
0.4398 0.0017a 0.8012 0.8000 0.0029a <.0001a 0.2217 0.1853 0.1002 0.6130 0.4275 0.1622 0.9467 0.0024a 0.6746 0.3819
a Significant factor.
Fig. 2 e Seasonal variability of HAAs in the three municipal water systems under study (Q1: Quebec City; Q2: Beauport; and Q3: Danville). (WDS-E: samples from WDS after last use of water at the late evening; PPeMCW: cold water samples in the early morning prior to the first water use; HWTeMHW: hot water samples in the morning; WDS-M: samples from the WDS in the morning; PPeACW: cold water samples in the afternoon, before water use; HWTeAHW: hot water samples in the afternoon; WDS-A: samples from the WDS in the afternoon).
show the model predictions and measured data plot for the PP and HWT, respectively. The high and low values of THMs in the PP were reasonably predicted by the models. However, the models showed relatively weak performance at few data points. This may be explained by the fact that these data were from the Danville (Q3) system, which has relatively high concentrations of bromide ions. The presence of bromide ions changes reaction patterns based on water pH, water temperature and type and distribution of natural organic matter (NOM) in water (Sohn et al., 2006; Hellur-Grossman et al., 2001). For the THMs in the HWT, the models predicted most of the peak values consistently, with the exception of the highest value from the Q2 water system (THMs: 345 mg/L). The observation of this single very high value may be associated with an unusual temperature rise in the HWT, excessive residual organics and free chlorine residuals in the HWT, as well as a possible outlier effect resulting from instrument and human errors during preservation and analysis. Table 5 summarizes the models for predicting HAAs in PP and HWT. The R2 values for the linear regression models range
from 0.83 to 0.96. The RMSE for the nonlinear models were observed to be 22.3 and 14.6 for the HAAs in the PP and HWT, respectively. The analysis of variance (ANOVA) showed that the models were statistically significant (P < 0.0001), and residuals were normally distributed and presented no visible trend. Consistent with the THMs, HAAs and free chlorine residuals in the WDS were found be statistically significant in all linear regressions, while temperature in the PP and TOC at the WDS were not found to be significant in any model. This may be explained by the fact that higher temperatures in the PP accelerate microbiological activity, thereby reducing HAAs in the PP. Conversely, higher temperature accelerates reactions between residual NOM and free chlorine residuals in the PP, which may increase HAAs formation in the PP. Due to these combined effects, implications of temperature on HAAs formation in the PP may not be noticeable. A better understanding of the role of temperature on microbiological activity in the PP may be essential to obtain a clearer picture of the scenarios. Figs. 5 and 6 show the plots for the model predictions and measured HAAs in the PP and HWT, respectively. In the case of HAAs in the PP, the models successfully predicted the peak values as well as the low values. For the HAAs in the HWT, models predicted most of the peak values consistently (Fig. 6). Among the models for predicting THMs in the PP, the linear models (models 1e3 in Table 4) had R2 between 0.90 and 0.91, showing that incorporation of interaction term (model 2) increased model performance slightly. The interaction term model showed relatively better performance in predicting THMs for Q3 sites (higher bromide contents) than the other two models (Fig. 3). In case of THMs in the HWT, the linear and log-linear models (models 5 and 7 in Table 4) showed better performance than the others. The visual inspection in Fig. 4 shows that the linear model (model 5) showed better fit with the measured data in most cases. As such, models 2 and 5 in Table 4 can be considered to be the models of choice for
343
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 3 7 e3 4 7
Table 4 e Linear and nonlinear models for predicting THMs in the plumbing pipes and in the hot water tanks. Sl.
Model Type
Output
1
Main factors linear model
THMPP
2
Main factors, Interaction or quadratic terms model
THMPP
3
Logelinear model
THMPP
4
Weibull (nonlinear) model (4 parameters) Main factors linear model
THMPP
6
Main factors and Interaction term model
THMHWT
7
Logelinear model
THMHWT
8
Logistic (nonlinear) model (3 parameters)
THMHWT
5
THMHWT
Prediction Expression
R2
RMSE
F Ratio
p
b0 þ b1THMWDS þ b2Cl2FREE b0 ¼ 14.9; b1 ¼ 0.975; b2 ¼ 39.4 b0 þ b1Cl2FREE þ b2THMWDS þ b3TOC þ b4Tpp þ b5 (TOC 4.1)(Tpp 18.7) b0 ¼ 21.4; b1 ¼ 36.9; b2 ¼ 0.986; b3 ¼ 0.59; b4 ¼ 1.83; b5 ¼ 1.21 Exp(b0 þ b1Ln(Cl2FREE) þ b2Ln(THMWDS)) b0 ¼ 1.735; b1 ¼ 0.448; b2 ¼ 0.67 q1 q2(Exp[q3 þ q4Ln{THMWDS}]) q1 ¼ 228.1; q2 ¼ 190.05; q3 ¼ 6.737; q4 ¼ 1.409 b0 þ b1THMWDS þ b2Cl2FREE þ b3pH b0 ¼ 307.1; b1 ¼ 1.073; b2 ¼ 48.91; b3 ¼ 40.4 b0 þ b1Cl2FREE þ b2THMWDS þ b3(THMWDS 51.57) (THMWDS 51.57)(THMWDS 51.57) b0 ¼ 22.0; b1 ¼ 54.77; b2 ¼ 1.429; b3 ¼ 0.0001 Exp(b0 þ b1Ln(Cl2FREE) þ b2Ln(THMWDS)) b0 ¼ 2.367; b1 ¼ 0.431; b2 ¼ 0.588 q1 ð1 þ q2 Expðq3 THMWDS ÞÞ
0.90
14.2
238.5
<0.0001
0.91
13.4
95.3
<0.0001
217.0
<0.0001
0.90
0.20 27.2
0.81
21.3
77.1
<0.0001
0.77
25.6
55.4
<0.0001
147.2
<0.0001
0.83
0.19 28.1
q1 ¼ 309.1; q2 ¼ 4.907; q3 ¼ 0.021 (THMPP: THMs at the plumbing pipes (mg/L); THMHWT: THMs in the hot water tanks (mg/L); THMWDS: THMs at the water distribution system (mg/L); RMSE: Root Mean Square Error; Cl2FREE: Free chlorine residual at the water entry points (mg/L); TPP: Temperature of water at the plumbing pipe ( C); TOC: Total organic carbon (mg/L); b, q ¼ Model parameters; Ln: Natural Logarithm).
predicting THMs in the PP and HWT respectively. For the HAAs models in PP, the linear models (models 1e3 in Table 5) had R2 between 0.88 and 0.96, indicating that incorporation of interaction terms (model 2 in Table 5) increased model performance significantly. The interaction term model showed much better performance in predicting HAAs all three sites (Q1, Q2, Q3) than the other models (Fig. 5). In case of HAAs in HWT, the main factor linear and interaction term models (models 5 and 6 in Table 5) showed better performance than the others. The visual inspection in Fig. 6 shows that the models (model 5 and 6) had better fit with the measured data in most cases. However, model 6 has higher R2 and lower
Fig. 3 e Measured and modeled THM concentrations in the plumbing pipes (PP) (MFL: Main factors linear model; MFIL: Main factors, interaction and higher order terms model; LMFL: Log main factors linear model; WNL-4P: Weibull nonlinear 4 parameters model; LOEC: Line of equal concentrations).
RMSE than those from the model 5, indicating that the model 6 is statistically better than the model 5 in Table 5. As such, models 2 and 6 in Table 5 can be considered to be the models of choice for predicting HAAs in the PP and HWT respectively. The non linear models did not show better fit than the linear models in any case (Figs. 3e6). This might be due to the fact that the non linear models were developed using one factor (either THMs or HAAs in the WDS), while THMs or HAAs formation in the PP and HWT can be affected be a number of factors and their interactions. These models are simple and easy to use as the models need small number of factors to characterize more than 80% of data variability and no complex mathematical procedure is required. The models were validated using different group of data isolated randomly from the original data generated in this study. Approximately, 1/3rd of the total data from each municipal system were randomly isolated for model validation and these data were not used in the model development. The model predictions and measured values were plotted and correlation coefficients (r) were determined (Table 6). The models for predicting THM in the plumbing pipes and hot water tanks had varying correlations with the measured data (r ¼ 0.58e0.94). For THM in the plumbing pipes and hot water tanks, models 2 (MFIL) and 5 (MFL) showed better correlations with the measured data than the other models (Table 6). The models for predicting HAA in the plumbing pipes and hot water tanks showed relatively better correlations (r ¼ 0.72e0.95). In case of HAA in the plumbing pipes and hot water tanks, models 2 (MFIL) and 6 (MFIL) showed better correlations with the measured data than the other models (Table 6). However, it is to be noted that the data generation for model validation study was limited. In future work,
344
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 3 7 e3 4 7
Fig. 4 e Measured and modeled THM concentrations in the hot water tanks (HWT) (MFL: Main factors linear model; MFIL: Main factors, interaction and higher order terms model; LMFL: Log main factors linear model; LNL-3P: Logistic nonlinear 3 parameters model; LOEC: Line of equal concentrations).
Fig. 5 e Measured and modeled HAA concentrations in the plumbing pipes (PP) (MFL: Main factors linear model; MFIL: Main factors, interaction and higher order terms model; LMFL: Log main factors linear model; WNL-4P: Weibull nonlinear 4 parameters model; LOEC: Line of equal concentrations).
4. extended sampling from other municipal water systems can be performed and validated to better explain the performance of the models. Overall, the models showed moderate to excellent prediction ability for both THMs and HAAs in the PP and HWT. However, data from only three municipal water systems were used in the modeling approach. Comprehensive investigations incorporating additional and diverse municipal water systems (surface water and ground water systems, chlorinated and chloraminated systems, systems with chlorineboosting stations, etc.) as well as PP and HWT from more regions would provide a better understanding of DBPs formation in the PP and HWT. Through these studies, the proposed models can be validated in future.
Discussion
The variability of THMs and HAAs in the WDS, PP and HWT was characterized in this study. Three types of linear models (main factors; main factors, interactions and higher orders; logarithmic) and two types of nonlinear models (threeparameter Logistic and four-parameter Weibull) were investigated to predict THMs and HAAs in the PP and HWT. The R2 values of the models varied between 0.77 and 0.96, indicating good to excellent predictive ability for THMs and HAAs in the PP and HWT. The models were found to be statistically significant. Significant model factors were determined through screening analysis, significance probability and the characterization of factor effects. Free chlorine residuals and THMs or HAAs at the entry points of houses were noted to be
Table 5 e Linear and nonlinear models for predicting HAAs in the plumbing pipes and in the hot water tanks. Sl.
Model Type
Output
1
Main factors linear model
HAAPP
2
Main factors, Interaction or quadratic terms model
HAAPP
3
Logelinear model
HAAPP
4
Weibull (nonlinear) model (4 parameters) Main factors linear model
HAAPP
5
HAAHWT
6
Main factors and Interaction term model
HAAHWT
7
Logelinear model
HAAHWT
8
Weibull (nonlinear) model (4 parameters)
HAAHWT
Prediction Expression
R2
RMSE
F Ratio
p
b0 þ b1HAAWDS þ b2Cl2FREE b0 ¼ 33.69; b1 ¼ 1.057; b2 ¼ 37.0 b0 þ b1Cl2FREE þ b2HAAWDS þ b3(HAAWDS 58.1) (HAAWDS 58.1) þ b4(Cl2FREE 1.25)(Cl2FREE 1.25) b0 ¼ 24.43; b1 ¼ 29.1; b2 ¼ 1.02; b3 ¼ 0.0016; b4 ¼ 15.0 Exp(b0 þ b1Ln(Cl2FREE) þ b2Ln(HAAWDS)) b0 ¼ 1.187; b1 ¼ 0.506; b2 ¼ 0.728 q1 q2Exp(Exp[q3 þ q4Ln{HAAWDS}]) q1 ¼ 774.9; q2 ¼ 759.8; q3 ¼ 7.607; q4 ¼ 1.239 b0 þ b1Cl2FREE þ b2HAAWDS þ b3pH b0 ¼ 158.39; b1 ¼ 11.02; b2 ¼ 0.979; b3 ¼ 19.79 b0 þ b1HAAWDS þ b2(HAAWDS 60.6)(HAAWDS 60.6) (HAAWDS 60.6) b0 ¼ 4.39; b1 ¼ 1.338; b2 ¼ 0.000035; Exp(b0 þ b1Ln(Cl2FREE) þ b2Ln(HAAWDS) þ b3Ln(pH)) b0 ¼ 7.18; b1 ¼ 0.216; b2 ¼ 0.656; b3 ¼ 2.727 q1 q2Exp(Exp[q3 þ q4Ln{HAAWDS}]) q1 ¼ 192.8; q2 ¼ 160.2; q3 ¼ 9.78; q4 ¼ 2.153
0.91
16.4
241.4
<0.0001
0.96
10.3
380.3
<0.0001
189.7
<0.0001
0.88
0.23 22.3
0.92
16.1
154.1
<0.0001
0.93
14.5
274.4
<0.0001
146.6
<0.0001
0.83
0.19 14.6
(HAAPP: HAAs at the plumbing pipes (mg/L); HAAHWT: HAAs in the hot water tanks (mg/L); HAAWDS: HAAs at the water distribution system (mg/L); RMSE: Root Mean Square Error; Cl2FREE: Free chlorine residual at the water entry points (mg/L); b, q ¼ Model parameters; Ln: Natural logarithm).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 3 7 e3 4 7
Fig. 6 e Measured and modeled HAA concentrations in the hot water tanks (HWT) (MFL: Main factors linear model; MFIL: Main factors, interaction and higher order terms model; LMFL: Log main factors linear model; WNL-4P: Weibull nonlinear 4 parameters model; LOEC: Line of equal concentrations).
statistically significant in the linear models, while the nonlinear models were estimated using THMs or HAAs at the entry points of water from the WDS to the houses. In all locations, the average temperature at the WDS was much lower than in the PP and the HWT. Sudden increases in temperature from the WDS to the PP and HWT could result in accelerated THMs and HAAs formation in the PP and HWT. However, temperature was not identified to be a significant factor in 15 of the 16 models (Tables 3,4). This result may be explained by the significance of free chlorine residuals in the models. At the PP, concentrations of THMs were much higher than at the WDS, indicating a possible increase in reaction rates between free chlorine residuals and residual NOM
Table 6 e Correlation coefficients from model validation study using additional data. DBP Type
Model Type
Output
r (Corr. Coef.)
THM
MFL MFIL LMFL WNL-4P MFL MFIL LMFL LNL-3P
THMPP THMPP THMPP THMPP THMHWT THMHWT THMHWT THMHWT
0.90 0.94 0.86 0.64 0.89 0.73 0.86 0.58
HAA
MFL MFIL LMFL WNL-4P MFL MFIL LMFL WNL-4P
HAAPP HAAPP HAAPP HAAPP HAAHWT HAAHWT HAAHWT HAAHWT
0.87 0.95 0.88 0.72 0.86 0.91 0.84 0.73
MFL: Main factors linear model; MFIL: Main factors, interaction and higher order terms model; LMFL: Log main factors linear model; LNL-3P: Logistic nonlinear 3 parameters model; WNL-4P: Weibull nonlinear 4 parameters model; THMPP: THM in the plumbing pipe; THMHWT: THM in the hot water tank; HAAPP: HAA in the plumbing pipe; HAAHWT: HAA in the hot water tank.
345
within the PP. These higher reaction rates may be due to temperature increases in the PP. In the case of HAAs, higher temperatures in the PP could result in accelerated activity of the pipe biofilms, which would reduce HAAs in the PP. Conversely, higher temperature accelerated reactions between residual NOM and free chlorine residuals in the PP, which might have increased HAAs formation in the PP. Due to these combined effects, implications of temperature on HAAs formation in the PP may not be noticeable (temperature not significant in models). A better understanding of the role of temperature on microbiological activity in the PP may be essential to obtain a clearer picture of HAAs behavior in houses. The NOM, which is expressed by surrogate TOC, was not identified as a statistically significant factor in most cases. In fact, the rate of consumption of free chlorine residuals also depends on TOC, while free chlorine residuals were identified as the most significant factor. As such, although TOC and temperature were not identified as significant parameters, these factors may have affected the reactions of free chlorine residuals. The data for the models developed in this study were obtained through a year-round sampling program. These data were collected from the WDS, PP and HWT at the rate of every six weeks (a total of eight sampling programs). The sampling program was designed to understand the effects of several stages, which would occur within the WDS, PP and HWT, on the concentrations of DBPs. Such as, changes of DBPs from the WDS to PP and HWT during overnight stagnation in the PP and HWT were assessed using three types of samples (i, ii and iii), while changes of DBPs in the WDS during overnight stagnation in the WDS were assessed using two types of samples (i and iv). Variation of DBPs in the PP and HWT after morning use and prior to afternoon use were estimated using v and vi, and the variation of DBPs in the WDS from morning to afternoon were estimated using vii. The sampling program assisted in understanding the variability of DBPs in the WDS, PP and HWT in different times of a day. The selection of three municipal systems (Q1, Q2 and Q3) was based on the source water nature and amounts and types of DBPs in those systems. For example, WDS in Q1 reported much lower THMs and HAAs throughout the year, while the WDS of Q2 had much higher THMs and HAAs than the WDS of Q1 and Q3. The source water of Q3 had very high bromide content, which reported much higher brominated THMs and HAAs throughout the year. It was anticipated that these three municipal water systems would provide better understanding at different scenarios of DBPs occurrences and source water quality. This information has been relayed through Tables 1 and 2 in this study. However, DBP concentrations can vary diurnally, and understanding this variability can improve models further. In addition, indoor handling of municipal water, such as storing in the refrigerator, filtration using commercial filters and heating and storing without lids, can also affect DBP exposure concentrations. The effects of these factors must also be investigated in the future. This study proposes a modeling system for DBPs formation in the PP and HWT, which may prove useful in understanding and predicting DBPs variability from the WDS to the PP and HWT as well as population exposure to DBPs and associated risks to human health.
346
5.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 3 7 e3 4 7
Conclusions
This study developed models for predicting THMs and HAAs changes from the WDS to the PP and HWT. Three different municipal water systems having different source water characteristics and DBPs formation potentials were considered. Linear models were found to show better performance in predicting DBPs changes in the PP and HWT than the non linear models. Incorporation of interaction effects of parameters improved model performance significantly in some cases, indicating that better understanding on the interaction effects might be necessary. The modeling systems may be useful in identifying strategies to improve water treatment and disinfection processes, implementing epidemiological correlation analysis (between health outcomes and DBPs in water) considering the point of exposure and establishing regulatory guidelines for DBPs. A future study incorporating more data, data variability and indoor handlings of municipal water might provide a better understanding of DBPs formation in the PP and HWT and their risks to humans.
Acknowledgement The authors acknowledge the National Sciences and Engineering Research Council (NSERC) of Canada and all partners of the Drinking Water Research Chair of Universite´ Laval (Ville de Que´bec, Ville de Le´vis, Dessau and ITF Labs-Avensys). The authors also thank Annick Dion-Fortier for sampling campaigns and management of the three municipal distribution systems under study.
references
Al-Omari, A., Fayyad, M., Qader, A.A., 2004. Modeling trihalomethane formation for Jabal Amman water supply system in Jordan. Environ. Model. Assess. 9, 245e252. Baribeau, H., Boulos, L., Haileselassie, H., Singer, P., Nichols, C., Schlesinger, S., Williams, S., Gullick, R., 2004. Formation and decay of THMs and HAAs in five full-scale distribution systems. In: Proceedings of the Water Quality Technology Conference. American Water Works Association, San Antonio, TX. Cantor, K.P., Lynch, C.F., Hildesheim, M., Dosemeci, M., Lubin, J., Alavanja, M., Craun, G.F., 1998. Drinking water source and chlorination by-products in Iowa: 1. Risk of bladder cancer. Epidemiology 9, 21e28. Chowdhury, S., Champagne, P., 2009. Risk from exposure to trihalomethanes during shower: probabilistic assessment and control. Sci. Total Env. 407 (5), 1570e1578. Chowdhury, S., Champagne, P., McLellan, P.J., 2010. Investigating effects of bromide ions on trihalomethanes and developing model for predicting bromodichloromethane in drinking water. Water Res. 44 (7), 2349e2359. Dion-Fortier, A., Rodriguez, M.J., Serodes, J., Proulx, F., 2009. Impact of water stagnation in residential cold and hot water plumbing on concentrations of trihalomethanes and haloacetic acids. Water Res. 43, 3057e3066. Fristachi, A., Rice, G., 2007. Estimation of the total daily oral intake of NDMA attributable to drinking water. Water and Health 5 (3), 341e355.
Health Canada, 2008. Guidelines for Canadian Drinking Water Quality. Prepared by the. Federal-Provincial-Territorial Committee on Health and the Environment. March. Hellur-Grossman, L., Manka, J., Lamoni-Relis, B., Rebhun, M., 2001. THM, haloacetic acids and other organic DBPs formation in disinfection of bromide rich Sea of Galilee (Lake Kinneret) water. Water Sci. Technol. Water Supply 1 (2), 259e266. Hrudey, S.E., 2009. Chlorination disinfection by-products, public health risk tradeoffs and me. Water Res. 43, 2057e2092. Jo, W.K., Weisel, C.P., Lioy, P.J., 1990a. Routes of chloroform exposure and body burden from showering with chlorinated tap water. Risk Anal. 10 (4), 575e580. Jo, W.K., Weisel, C.P., Lioy, P.J., 1990b. Chloroform exposure and the health risk associated with multiple uses of chlorinated tap water. Risk Anal. 10 (4), 581e585. King, W.D., Marrett, L.D., 1996. Case-control study of bladder cancer and chlorination by products in treated water (Ontario, Canada). Canc. Causes Controls 7, 596e604. King, W.D., Dodds, L., Armson, B.A., 2004. Exposure assessment in epidemiologic studies of adverse pregnancy outcomes and disinfection by-products. J. Exp. Anal. Environ. Epidemiol. 14, 466e472. Lynch, C.F., Woolson, R.F., O’Gorman, T., Cantor, K.P., 1989. Chlorinated drinking water and bladder cancer: effect of misclassification on risk estimates. Arch. Environ. Health 44, 252e259. Lee, S.C., Guo, H., Lam, S.M., Lau, S.L., 2004. Multipathway risk assessment on disinfection by-products of drinking water in Hong Kong. Environ. Res. 94 (1), 47e56. Lin, T.F., Hoang, S.W., 2000. Inhalation exposure to THMs from drinking water in South Taiwan. Sci. Total Environ. 246, 41e49. Montgomery, D.C., Runger, G.C., 2007. Applied Statistics and Probability for Engineers, fourth ed. John Wiley and Sons. Nuckols, J.R., Ashley, D.L., Lyu, C., Gordon, S.M., Hinckley, A.F., Singer, P., 2005. Influence of tap water quality and household water use activities on indoor air and internal dose levels of trihalomethanes. Environ. Health Persp. 113 (7), 863e870. 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. Mutat. Res. 636, 178e242. Richardson, S.D., Fasano, F., Ellington, J.J., Crumley, F.G., Buettner, K.M., Evans, J.J., Blount, B.C., Silva, L.K., Waite, T.J., Luther, G.W., Mckague, A.B., Miltner, R.J., Wagner, E.D., Plewa, P.J., 2008. Occurrences and mammalian cell toxicity of iodinated disinfection byproducts in drinking water. Environ. Sci. Technol. 42 (22), 8330e8338. SAS Inc., 2009. Statistical Programming and Modeling Available at. http://www.jmp.com/. Sadiq, R., Rodriguez, M., 2004. Disinfection byproducts (DBP) in drinking water and predictive models for their occurrences: a review. Sci. Total Env. 321, 21e46. Savitz, D.A., Singer, P.C., Herring, A.H., 2006. Exposure to drinking water disinfection by-products and pregnancy loss. Am. J. Epidemiol. 164 (11), 1043e1051. Semerjian, L., Dennis, J., 2007. Multipathway risk assessment of trihalomethane exposure in drinking water of Lebanon. J. Water Health 5 (4), 511e522. Singer, P.C., 1994. Control of disinfection by-products in drinkingwater. J. Environ. Eng. 120 (4), 727e744. Singer, P.C., Weinberg, H.S., Krasner, S., Arora, H., Najm, I., 2002. Relative Dominance of HAAs and THMs in Treated Drinking Water. American Water Works Association Research Foundation, Denver, Colorado, United States of America, 298 pp. Sohn, J., Amy, G., Cho, J., Yoon, Y., 2006. Bromide ion incorporation into brominated disinfection by-products. Water, Air, and Soil Poll. 174, 265e277.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 3 7 e3 4 7
USEPA, 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, United States of America, 47 pp. USEPA, 1995b. Method 552.2. Determination of Haloacetic Acids in Drinking Water by Liquid Liquid Extraction and Gas Chromatography with Electron Capture Detection. National Exposure Research Laboratory, Office of Research and Development, USEPA, Cincinnati, Ohio, United States of America. USEPA, 2006. National primary drinking water regulations: stage 2 disinfectants and disinfection byproducts rule: final rule. Fed. Reg. 71 (2) January 4. Uyak, V., 2006. Multi-pathway risk assessment of trihalomethanes exposure in Istanbul drinking water supplies. Environ. Int. 32 (1), 12e21. Villanueva, C.M., Cantor, K.P., Cordier, S., Jaakkola, J.J.K., King, W.D., Lynch, C.F., Porru, S., Kogevinas, M., 2004. Disinfection
347
byproducts and bladder cancer: a pooled analysis. Epidemiology 15 (3), 357e367. Weinberg, H.S., Pereira, V.R.P.J., Singer, P.C., Savitz, D.A., 2006. Considerations for improving the accuracy of exposure to disinfection by-products by ingestion in epidemiologic studies. Sci. Total Environ. 354 (1), 35e42. Weisel, C.P., Chen, W.J., 1994. Exposure to chlorination byproducts from hot water use. Risk Anal. 14 (1), 101e106. Wu, W.W., Benjamin, M.M., Korshin, G.V., 2001. Effects of thermal treatment on halogenated disinfection by-products in drinking water. Water Res. 35 (15), 3545e3550. Xu, X., Mariano, T.M., Laskin, J.D., Weisel, C.P., 2002. Percutaneous absorption of trihalomethanes, haloacetic acids and haloketones. Toxicol. Appl. Pharmacol. 184, 19e26. Xu, X., Weisel, C.P., 2005. Human respiratory uptake of chloroform and haloketones during showering. J. Expo. Anal. Environ. Epidemiol. 15, 6e16.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 4 8 e3 5 6
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Development and application of a simulation model for the thermophilic oxic process for treating swine waste Kyoungho Jeon*, Kazunori Nakano, Osamu Nishimura Graduate School of Engineering, Tohoku University, Aoba 6-6-06, Aramaki, Sendai 980-8579, Japan
article info
abstract
Article history:
The thermophilic oxic process (TOP) is a composting process that enables simultaneous
Received 1 December 2009
complete decomposition and evaporation of organic waste under high temperature
Received in revised form
conditions supported by well-balanced calorific value control. To develop the simulation
22 April 2010
model for TOP, three-dimensional relationships among decomposition rate constant,
Accepted 5 August 2010
temperature (20e70 C) and moisture content (30e70%) were determined for swine waste
Available online 12 August 2010
and cooking oil based on the oxygen consumption rate during a thermophilic oxic decomposition reaction.
Keywords:
The decomposition rate of swine waste and cooking oil under various moisture contents
Thermophilic oxic process
was described by the Arrhenius equation. The optimal temperature and moisture content
Swine waste
were 60 C and 60% for swine waste and 60 C and 50% for cooking oil, respectively. The
Cooking oil
simulation model for TOP was constructed on the basis of the carbon, heat, and moisture
Decomposition rate constant
balance. The validation of the simulation model was examined by comparing the measured
Arrhenius equation
temperature in the TOP reactor to that estimated by the simulation. The simulation model
Simulation model
was proven by comparing experimental and calculated values. The relationship between the injection calorific value and the process mechanism of TOP was interpreted by the simulation model. On the basis of their relationship during TOP, the appropriate process conditions were discussed. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Generally, in the swine waste treatment process, the excrement and urine are separated and treated by composting and the activated sludge process. The composting process is used effectively as an aerobic treatment of organic wastes. However, considerable amounts of land and time are necessary for the composting process. Compared to the composting process, the thermophilic oxic process (TOP) is remarkable because it enables efficient processing with minimum land usage and process time (Nakano and Matsumura, 2001, 2002; Liang et al., 2003; Chang et al., 2006).
TOP is a composting process that enables both complete decomposition of organic wastes and complete evaporation of moisture. When the calorific value of the target waste is deficient for the evaporation of moisture contained in the waste, an additional heat source such as cooking oil is supplied. The reactor can be maintained at a high temperature by heat generation accompanied with organic decomposition, which helps to maintain the high decomposition rate of organic substances by thermophilic microorganisms. The complete decomposition of organic substances and evaporation of moisture can be realized under such conditions where the calorific value balance is well controlled (Nakano and Matsumura, 2001, 2002).
* Corresponding author. Tel.: þ81 22 795 7469; fax: þ81 22 795 7471. E-mail address:
[email protected] (K. Jeon). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.005
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 4 8 e3 5 6
TOP can treat high-density organic wastes such as livestock waste, food waste, and excess sludge. The moisture in waste can be evaporated completely when TOP is operated with an additional heat source under a calorific value balance appropriate for the controlled conditions. However, the empirical decision of control parameters such as loading rate of waste and the ratio of additional heat source such as cooking oil limits the application of TOP in various organic waste treatments (Nakano and Matsumura, 2002; Andani et al., 2002). Therefore, it has been necessary to develop a simulation model for TOP, which enables a theoretical optimization of the process conditions. In this study, a simulation model for TOP was developed for the treatment of swine waste. The model was composed of carbon-heat-moisture balance equations considering the organic matters decomposition rate under different temperature and moisture content conditions in the reactor. The decomposition rates for various temperatures and moisture content obtained by oxygen uptake tests were examined and then expressed in a modified Arrhenius equation. The temperature, moisture content, and decomposition rate during the treatment of swine waste by TOP were calculated by the established simulation model, and the results were compared with the experimental results. Finally, a numerical analysis using the simulation model was applied to evaluate the influence of the control conditions such as the additional heat source on treatment performance.
2. Theories and formulation of the TOP simulation model In the TOP, the organic matter is decomposed by a microbial reaction. As a result, heat is generated in the reactor, and this heat is used for moisture evaporation. The temperature in the reactor increases, promoting a microbial decomposition reaction. On the other hand, heat loss by ventilation also occurs in the reactor. During one cycle of treatment, heat generation and consumption occurs, which is then followed by the next feed of organic waste (Miguel et al., 2004a,b, 2005; Chang et al., 2006; Klejment et al., 2008). To explain the process, a model for TOP that consists of equations for carbon balance, heat balance, and moisture balance was proposed in this study. The simulation model for TOP was constructed on the basis of an equation of carbon, heat and moisture balance.
2.1.
Carbon balance
349
Where C is TC density at Dt (mg-TC/g-compost), C0 is initial TC density (mg-TC/g-compost), K is the primary order reaction constant of the decomposition of organic matter by microorganisms in the compost at a certain temperature and moisture content. (1/h), Dt is the unit time (h). The biological reactions by microorganisms are based on the Arrhenius reaction that shows the temperature dependency at the chemical reaction rate. That is, the microbial decomposition rate of organic matter at certain temperature as shown by Eq. (2) (Kosseva et al., 2007; Miguel et al., 2004a,b, 2005). The proliferation rate of the microorganisms is expressed by the primary order equation. Therefore, the decomposition rate, that is, a decrease of carbon in the organic matters by the proliferation of microorganisms, can be expressed by the primary order equation (Miguel et al., 2004a,b, 2005). K ¼ A$EXPð Ea =RTÞðT1 &T&T2 Þ
(2)
Where A is the Arrhenius pre-exponential factor, Ea is activation energy (kJ/mol), R is the ideal gas constant (8.314 J/mol/ K) and T is the absolute temperature (K ).
2.2.
Heat balance
Temperature is one of the important factors affecting microbial growth and biological reactions. Temperature can exert an effect on biological reactions in two ways: by influencing the rates of enzymatically catalyzed reactions and by affecting the rate of the diffusion of substrate into the cells (Miguel et al., 2004a,b, 2005). By using a decomposition rate of organic matter formulated by Eq. (1) at a certain temperature and moisture content, the TC density decomposed at a certain time is obtained by Eq. (3). The generation calorific value from the whole compost is then obtained by Eq. (4). DC ¼ C0 C
(3)
Ei ¼ DC$mc $Hc
(4)
Where DC is TC density decomposed at Dt (mg-TC/g-compost), Ei is the generation calorific value from the whole compost (Kcal), mc is the amount of compost (Kg-compost), Hc is the calorific value content of the organic matter (Kcal/Kg-TC). Since the evaporation of moisture occurs by the heat generated in the microbial decomposition of the organic matters in the reactor, calorific value loss by the evaporation of moisture calculated by Eq. (5) is considered for heat balance. Ew ¼ gw $mew
(5)
The swine waste was input as a processing object and the cooking oil was input as an additional heat source in this study. A constant amount of carbon is present in the organic matter. The total theoretical generating calorific value was calculated under an assumption as to what is required for the organic matter to decompose completely. The carbon balance is the decomposition model of organic matter in which the organic matter is converted into carbon dioxide by microorganisms. The consecutive changes in total carbon (TC) density are shown by Eq.(1).
Where Ew is the calorific value loss from the evaporation of moisture (Kcal), gw is moisture evaporation latent heat (569.1 Kcal kg-H2O1), and mew is the amount of moisture evaporated (kg). Some of the generated calorific values are consumed for the rise of the chip temperature in the reactor. This amount is calculated by Eq. (6), including specific heat for the cedar chip used as a carrier material in the reactor and moisture. The specific heat used for the cedar chip is 0.6 (Ravi et al., 2004).
C ¼ C0 $EXPðK$DtÞ
ES ¼ ðmwr $4w þ mcr $4c Þ$DT
(1)
(6)
350
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 4 8 e3 5 6
Where Es is the calorific value used in the rise of chip temperature in the reactor (Kcal), mwr is the amount of moisture in the reactor (kg), 4w is the specific heat for the moisture (0.999 Kcal kgH2O1 deg.1), mcr is the amount of cedar chip in the reactor (kg), 4c is the specific heat for the cedar chip (0.6 Kcal kg-cedar chip1 deg1), and DT is the temperature variation (deg). Since TOP is an aerobic microbial reaction, a ventilation system run by a compressor is installed in the reactor system. Therefore, heat loss by ventilation and that from the surface of the reactor is considered in the equation. There is no heat loss when the temperature of the reactor is the same as the atmospheric temperature, while the heat loss increases with the temperature rise in the reactor by heat generation with the microbial decomposition of organic matter. The following was done to derive a heat loss equation. First, a sample was used with an input of 300 g of swine waste and 45 g of cooking oil. This experimental condition is likely the stable TOP of swine waste. The average temperature of the reactor in this experiment was then measured. The generation calorific value (Ei), the loss calorific value from evaporation of moisture (Ew), and the calorific value used in the rise of chip temperature in the reactor (Es) were calculated by the theoretical and empirical equation in this condition. The calorific value in which loss of calorific values from evaporation of
moisture and the calorific value used in the rise of chip temperature in the reactor was subtracted from the generation calorific value and was calculated to the loss calorific value from the reactor as in Eq. (7). Ev ¼ Ei Ew Es
(7)
Where Ev is the lost calorific values from the reactor (Kcal). The heat loss calorific value from the reactor used in this study is expressed by Eq. (8). Ev ¼ a$EXPðb$Tr Þ
(8)
Where a, b is heat loss constant determined by experiments (a ¼ 5.9149, b ¼ 0.0598), Tr is the temperature in the reactor (C).
2.3.
Moisture balance
Moisture content is another important factor affecting microbial reactions. The input moisture to the reactor through the organic waste is calculated by Eq. (9). mq ¼ ms $q
(9)
Where mq is the amount of moisture in the swine waste (kg), ms is the amount of organic waste injected into the reactor
Table 1 e Formulation of TOP simulation model. Equation of decomposition of organic matter Carbon balance
Equation
TC density at Dt (C) C ¼ C0$EXP(K$Dt) Primary order reaction constant of decomposition of organic matter (K) K ¼ A$EXP(Ea/RT) TC density decomposed at Dt (DC) DC ¼ C0 C
Generation equation
Heat balance
Heat generation by decomposition of organic matter (Ei) Ei ¼ DC$mc$Hc
Moisture balance
Input of moisture by organic matter (mq) mq ¼ ms$q Moisture generation by decomposition of swine waste and cooking oil (mw, mo) swine waste: mw ¼ msd$0.6 cooking oil: mw ¼ msd$1.0
Loss equation Heat loss by moisture evaporation (Ew) Ew ¼ gw$mew Heat loss by temperature increase in reactor (Es) Es ¼ (mwr$4w + mcr$4c)$T Heat loss by ventilation (Ev) Ev ¼ a$EXP(b$Tr) Moisture loss by evaporation (V) V ¼ Vc$(es e)$A
Here, C is the TC density at Dt (mg-TC/g-compost); C0 is the initial TC density (mg-TC/g-compost); K is the primary order reaction constant of the decomposition of organic matter by microorganisms in the compost at a certain temperature and moisture content (1/h); Dt is the unit time (h); A is the Arrhenius pre-exponential factor; Ea is activation energy (kJ/mol); R is the ideal gas constant (8.314 J/(mol K)); T is the absolute temperature (K); DC is TC density decomposed at Dt (mg-TC/g-compost); Ei is the generation calorific value from the complete compost (kcal); mc is the amount of compost (kg-compost); Hc is the calorific value content of the organic matter (kcal/kg-TC); Ew is the calorific value loss from the evaporation of moisture (kcal); gw is moisture evaporation latent heat (569.1 kcal/kg); mew is the amount of moisture evaporated (kg); Es is the calorific value used to increase the chip temperature in the reactor (kcal); mwr is the moisture content in the reactor (kg); 4w is the specific heat of moisture (0.999 kcal/kg C); mcr is the amount of cedar chip in the reactor (kg); 4c is the specific heat for the cedar chip (0.6 kcal/kg C); DT is the temperature variation ( ); Ev is the lost calorific values (lost heat) from the reactor (kcal); a, b are heat loss constants determined from experiments (a ¼ 5.9149, b ¼ 0.0598); Tr is the temperature in the reactor ( C); mq is the amount of moisture in the swine waste (kg); ms is the amount of organic waste injected into the reactor (kg); q is moisture content in the organic waste being treated (%); mw is the amount of moisture generated from swine waste (kg); mo is the amount of moisture generated from cooking oil (kg); msd is the amount of swine waste decomposed (kg); mod is the amount of cooking oil decomposed (kg); V is the moisture evaporation rate (g/h); Vc is the evaporation rate constant of moisture (0.23 g/h/ cm2/hPa); es is the steam pressure of the cedar chip surface (hPa); e is the atmosphere steam pressure (hPa); and A is the surface area of the cedar chip (cm2).
351
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 4 8 e3 5 6
Input the organic matter (swine waste and cooking oil) Input moisture
(Eq. 9,10,11,12,13) (Eq. 1,2)
Moisture generation
Decomposition of organic
Input temperature
matters by microoganism
Where mw is moisture generation from swine waste (kg), mo is moisture generation from cooking oil (kg), msd is the amount of swine waste decomposed (Kg), and mod is the amount of cooking oil decomposed (Kg). The moisture evaporation in the reactor depends on the Turbulent Natural Convective Flow. The evaporation rate was calculated by Eq. (14). V ¼ Vc $ðes eÞ$A
(Eq. 3,4)
Where V is the moisture evaporation rate (g/h), Vc is the evaporation rate constant of moisture (0.23 g/h/cm2/hPa), es is the steam pressure of the cedar chip surface (hPa), e is the atmosphere steam pressure (hPa), and A is the surface area of the cedar chip (cm2). The steam pressure of surface of the atmosphere is expressed by the function of the temperature and relative humidity. This is shown in Eqs. 15 and 16, respectively.
Heat generation by decomposition (Eq. 14,15,16,17)
Moisture evaporation by generated heat Output temperature
(Eq. 5,7,8)
Heat loss (ventilation and evaporation)
Output (Eq. 6)
moisture
Temperature raise in the reactor
(14)
es ¼ esat $Us
(15)
e ¼ esat $U
(16)
Where esat is the saturated steam pressure (hPa), Us is the surface humidity on the cedar chip (%), and U is the humidity of the atmosphere (%). The saturated steam pressure is calculated by the GoffeGratch equation expressed by Eq. (17) (Walker et al., 2009).
Fig. 1 e Flow chart of the simulation model for organic waste decomposition by thermophilic oxic process.
Lnðesat Þ ¼ 6096:9385,T1 þ 21:2409642 2:711193,102 $T þ 1:673952$105 $T2 þ 2:433502,lnðTÞ (Kg), and q is moisture content in the organic waste being treated (%). It has been reported that 0.6 kg of moisture is generated when 1 kg of swine waste is completely decomposed (Lu¨bken et al., 2007). Thus, moisture generation from swine waste is formulated as Eq. (10). mw ¼ msd $0:6
(10)
Since the principal ingredient of cooking oil used as an additional heat source for TOP is oleic acid and linoleic acid, moisture generation from cooking oil is assumed on the basis of chemical equations Eqs. 11 and 12 (Lu¨bken et al., 2007). Oleic acid : 2C18 H34 O2 þ 51O2 /36CO2 þ 34H2 O
(11)
Linoleic acid : C18 H32 O2 þ 25O2 /18CO2 þ 16H2 O
(12)
As a result, it is assumed that 1.0 kg of moisture is generated from 1.0 kg of cooking oil and moisture generation from cooking oil is formulated as Eq. (13). mo ¼ mod $1:0
(13)
(17)
the equations of the mentioned heat, water, and carbon balance express the reaction that occurred in the TOP. The variables and equations that were used in the simulation model were arranged in Table 1, and a simulation model based on the structure of Fig. 1 was constructed. Fig. 1 expresses the following flow. The decomposition amount of the injected liquid organic matters was calculated using the decomposition equation of the organic matters, a linear equation, depending on the change in the temperature and water content ratio. About 1e2% of the undecomposed inorganic matters are actually contained in swine waste, but it was assumed that all those contained by this model could be decomposed. Further, the generated calorific value was calculated from the calculated decomposition amount, and the amount of moisture generated from the decomposition of organic matters was calculated. In this case, the ratio of swine waste to moisture generated from the decomposition of organic matters was 1:0.6, and the ratio of cooking oil to moisture was 1:1. Next, the value of the heat loss due to the ventilation and the heat loss discharged from the surface of
Table 2 e Characteristics of swine waste and cooking oil.
Swine waste Cooking oil ND:Not Detected.
Calorific value (Kcal/Kg-Solid)
BOD (mg/L)
TOC (mg/L)
T-N (mg/L)
T-P (mg/L)
Moisture content (%)
4500 9000
40,000 e
53,000 728,900
5000 ND
1900 ND
90 0
352
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 4 8 e3 5 6
Fig. 2 e Oxygen consumption rate measurement Equipment (O2 up-tester): a constant temperature water tank, b-O2 up-tester cylinder, c-soda lime, d-compost, e-view-let.
by the swine waste was not sufficient for TOP aiming at complete moisture evaporation and decomposition of organic matter, an additional heat source was required. Cooking oil was used as the additional heat source. Swine waste and cooking oil were analyzed using the sewage experiment method, and TOC was measured using a TOC-5000 analyzer (Shimadzu). The calories of swine waste and cooking oil were measured using a digital calorimeter, and T-N and T-P were measured using an auto analyzerII (SamplerII technicon). The water content of cedar chip in a reactor was calculated from the difference before and after the experiment, after drying an about 3 g sample for approximately 3 h at 105 C. The reaction temperature was measured using a thermocouple data logger attached to the reactor. The weight was measured before and after that organic matters were injected. The characteristics of the swine waste and cooking oil are shown in Table 2.
3.2. Measurement of the organic matter decomposition rate using O2 up-tester the reactor were subtracted from the above calorific value. It is the system in which the values of the water content and temperature after conducting one cycle are added to the calculation of the amount of decomposed organic matters in the next cycle. The analysis of the simulation was done by differential equations using a 1-min unit. This simulation model can predict the reaction of the microorganisms in the reactor and the change in its temperature and water content according to the variable conditions of the characteristics of the organic matters injected (TC concentration, water content, organic matters loaded), the injection amount, the injection interval, the injection ratio of the auxiliary heat source, etc. A program that can theoretically calculate and adjust the condition of the maximum processing load, the optimal condition of the auxiliary heat source, and the optimal control of the TOP reactor, calculating the change in such various experiment conditions, was formulated.
3.
Materials and methods
3.1.
Experimental materials
The swine waste used in this study was a mixture of urine and manure from the piggery. Since the calorific value introduced
The experimental equipment O2 up-tester (Titech Inc., Japan) shown in Fig. 2 was used to measure the decomposition rate of the swine waste and cooking oil at various temperatures (20, 30, 45, 55, 60, 65, and 70 C) and moisture content (30%, 40%, 50%, 60%, 65%, and 70%) were investigated by using 5 g compost, which was actually produced in the TOP reactor. The amount of swine waste and cooking oil used in this experiment was 0.5 and 0.1 g, respectively. The test ranges of temperature and moisture content correspond with the possible conditions in the TOP reactor. The equipment consists of a view-let and a reaction bottle set in a water tank with temperature controlled at an objective value. A known amount of swine waste or cooking oil mixed with compost is put into the reaction bottle. The equipment is then set in the water tank. Swine waste or cooking oil is decomposed under a constant temperature by thermophilic microorganisms in the compost. Decomposition of swine waste and cooking oil contained in the sample within the reaction bottle results in consumption of oxygen and generation of carbon dioxide. The generated carbon dioxide is absorbed with soda lime, leading to a decrease in the atmospheric pressure in the cultivation bottle and an increase in the water level of the view-let. The rising water level shows the amount of oxygen consumption; the amount of oxygen consumption can be measured from the change in water level.
Fig. 3 e Schematic diagram of the bench-scale reactor a-reactor, b-compost, c-drain water, d-balance, e-stirrer, f-thermometer, g-injection entrance, h-air compressor, i-air-flow meter, j-mixing meter, k-temperature recorder.
353
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 4 8 e3 5 6
reactor every 24 h. The stirrer system was operated for approximately 10 min immediately following the feed of waste. To maintain aerobic conditions, air was fed from the bottom of the reactor at a flow rate of 100 L/m3/min. To evaluate the operation of TOP, the change in the weight, temperature, and moisture content in the reactor were monitored.
4.
Results and discussion
4.1. Decomposition rate constant for swine waste and cooking oil Fig. 4 e TC consumption curve swine waste under a condition of 60 C and 60% of moisture content. The amounts of oxygen consumption and carbon dioxide generation are different. Swine waste is composed of carbohydrates, proteins and fats. The principal ingredient of the cooking oil is oleic acid and linoleic acid. The ratio of oxygen consumption and carbon dioxide generation from decomposition of swine waste is approximately 1:1. The ratio of oxygen consumption and carbon dioxide generation from decomposition of swine waste was previously reported as approximately 1:0.7 (Lu¨bken et al., 2007). The decomposition rate of swine waste and cooking oil was calculated by considering the ratio of oxygen consumption and carbon dioxide generation.
3.3.
Reactor experiment
To confirm the results of the numerical analysis, the performance of TOP was examined by a bench-scale reactor. The reactor consisted of a stirrer and aeration system and was covered with a thermal insulating material. The reactor’s effective volume was 19 L. A schematic diagram of the benchscale reactor is shown in Fig. 3. Before the experiment, 300 g of compost as seed and 2.26 kg of cedar chips with a moisture content of 50% were put into the reactor. Then, 300 g of swine waste and a certain amount of cooking oil were fed to the
Since temperature and moisture content exert a significant influence on microorganism proliferation, a numerical estimation of the decomposition rate of swine waste and cooking oil by the microorganism in the compost under the condition of various temperatures and moisture content is necessary for the construction of a simulation model for TOP. Therefore, three-dimensional relationships among the decomposition rate, temperature, and moisture content were measured for swine waste and cooking oil. The decomposition rate under certain temperatures and moisture content was determined by a total carbon (TC) consumption curve that could be measured on the basis of oxygen consumption in the O2 up-tester. Fig. 4 shows an example of the TC consumption curve that was obtained for swine waste under the compost condition of 60 C and 60%. The curve could be approximated by an equation for a first order reaction in which the decomposition rate constant (K ) was expressed as shown in Eq. (1) (Kosseva et al., 2007; Miguel et al., 2004a,b, 2005). The equation shown like Eq. (1) is Eq. (18). C ¼ 5:391$EXPð0:141$DtÞ
(18)
The decomposition rate constant (K ) of swine waste at a compost of 60 C and 60% was determined as 0.141 based on the curve shown in Fig. 4 and Eq. (18). In a similar way, the decomposition rate constant (K ) under a compost of various
Table 3 e Decomposition rate constant k of swine waste and cooking oil obtained at various temperature and moisture content. Temperature
20 C 30 C 45 C 55 C 60 C 65 C 70 C
Items
Swine waste Cooking oil Swine waste Cooking oil Swine waste Cooking oil Swine waste Cooking oil Swine waste Cooking oil Swine waste Cooking oil Swine waste Cooking oil
Moisture content 30%
40%
50%
60%
65%
70%
0.0109 0.0005 0.0195 0.0015 0.0503 0.0032 0.0580 0.0065 0.0880 0.0131 0.0300 0.0056 0.0109 0.0028
0.0200 0.0021 0.0327 0.0045 0.0637 0.0127 0.1020 0.0265 0.1154 0.0257 0.0600 0.0204 0.0200 0.0035
0.0281 0.0045 0.0396 0.0093 0.0845 0.0181 0.1087 0.0314 0.1302 0.0320 0.0767 0.0270 0.0281 0.0043
0.0311 0.0034 0.0476 0.0078 0.0866 0.0197 0.1208 0.0250 0.1410 0.0292 0.0790 0.0257 0.0334 0.0038
0.0168 0.0022 0.0269 0.0047 0.0562 0.0114 0.0677 0.0209 0.0765 0.0228 0.0389 0.0204 0.0211 0.0028
0.0055 0.0005 0.0115 0.0015 0.0346 0.0045 0.0539 0.0097 0.0650 0.0141 0.0316 0.0069 0.0055 0.0021
354
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 4 8 e3 5 6
Fig. 5 e Three-dimensional relationships of decomposition rate constant of organic matters (1/h), temperature and moisture content for (a) swine waste and (b) cooking oil.
temperatures and moisture content was determined (Kosseva et al., 2007; Miguel et al., 2004a,b, 2005) as shown in Table 3. It was found that the decomposition rate for both swine waste and cooking oil increased with temperature until 60 C and decreased when the temperature was over 60 C. However, such critical conditions for moisture content did not correspond
0 R2 = 0.9986
ln (k )
-2 -4 -6 -8 2.9
cooking oil swine waste 3.0
3.1
R2 = 0.9747 3.2
3.3
3.4
3.5
1/T/1000 Fig. 6 e Arrhenius plots for the activation enthalpy of the organic matters decomposition (moisture content: 50%).
Fig. 7 e Change of temperature and moisture content of simulation analysis and experiment when the amount of additional heat source (cooking oil) was set at (a) 15%; (b) 0%; (c) 30% of swine waste (300 g).
between swine waste and cooking oil. The optimal moisture content for swine waste and cooking oil was 60% and 50%, respectively. The decomposition rate constants of swine wastes were higher than that of cooking oil when compared at the same moisture content and temperature. This means that swine wastes will be decomposed faster than cooking oil in TOP. The three-dimensional relationships among decomposition rate constant, temperature, and moisture content for swine waste and cooking oil was compared, as shown in Fig. 5. The decomposition rate constant of the swine waste and cooking oil was shown by the Arrhenius equation. The high regression coefficients shown in Fig. 6 suggest that the Arrhenius equation could adequately describe the relationships between temperature and decomposition of organic matter.
355
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 4 8 e3 5 6
Table 4 e Summary of simulation and experimental results of swine waste treatment by TOP. Additional heat source conditions 15%
0%
30%
Simulation Experiment Simulation Experiment Simulation Experiment Operating conditions
Decomposition efficiency Calorific value Balance obtained by simulation
Loading rate of waste (g/day) Loading rate of cooking oil (g/day) Total input organics (g) Accumulation amount (g) Decomposition efficiency (%) Necessary calorific values (Kcal/day)* Generated calorific values (Kcal/day)** Calorific values balance (Kcal/day)***
300
300
300
45
0
90
4830 600 88 540
4830 480 90 UD
4200 3130 25 540
4200 3360 20 UD
5460 980 82 540
540
32
945
0
508
þ405
5460 1100 80 UD
Necessary calorific values*: Calorific value necessary for temperature rising in the reactor, evaporation of moisture and heat loss by ventilation. Generated calorific values**: calorific value generated by decomposition of organic matters. Calorific values balance***: Generated calorific values Necessary calorific values. UD: Undetectable.
4.2. Numerical analysis of TOP operating with different amounts of an additional heat source The operational conditions of the current TOP were determined empirically. The simulation model constructed in this study, however, can be expected to enable a theoretical determination of the operational conditions constructed by using the decomposition rate of the microorganism, the heat balance, and the moisture balance. The construction and analysis of the TOP simulation model can be done through an understanding of the treatment process. Further, if the simulation analysis is used, it is possible to forecast the treatment process condition. The change in temperature and moisture content in the TOP reactor were estimated and compared with an actual reactor experiment value, and the model was verified. Fig. 7shows the variation in the moisture content and temperature in the TOP reactor operating with different amounts of an additional heat source. Fig. 7 (a) shows the results of a simulation and experiment in which cooking oil of 15%-wt of swine waste was fed as an additional heat source. Both 300 g swine waste and 45 g cooking oil were input to the TOP reactor every 24-h. The temperature was changed up to 60 C and the moisture content was assumed to be stable at around 55%. This indicated that the evaporation of moisture containing swine waste was well realized. The average temperature of the reactor in this experiment was measured. From the average temperature, the heat loss equation was obtained using Eq. (7). The simulation model was constructed using this heat loss equation. The experiment in which 45 g cooking oil was input was calculated theoretically by using the constructed model. The calculation result is shown in Fig. 7(a). The temperature and moisture content variation of the experiment and the simulation corresponded well. Accumulation amount and decomposition efficiency were calculated using the simulation model, and the result was
compared with the experimental value. In simulation, 4830 g of organic matter was input. The accumulation amount was 600 g and the decomposition efficiency was 88%. The experiment result was the accumulation amount of 480 g and the decomposition efficiency of 90%. The calorific value balance of generation and consumption could not be measured from the experiment. However, the calorific value was calculated by the simulation model. The calorific value of generation and consumption was 540 Kcal in the reactor. Complete decomposition and evaporation of the organic matter were possible. The comparison results of the simulation and the experiment are shown in Table 4. The simulation and experiment results of the temperature and moisture content in the TOP reactor operating without an additional heat source are shown in Fig. 7(b). Three hundred grams of swine waste were fed to the TOP reactor every day. The moisture content increased and the temperature never reached 60 C during treatment for 14 days as well as during actual experiment. Decomposition efficiency and calorific value balance during treatment were estimated by a simulation model, as shown in Table 4. Although 4200 g of swine waste were fed to the TOP reactor, the accumulation amount of the reactor was 3130 g in 14 days, revealing that decomposition efficiency was approximately 25% in the simulation. The experiment result showed an accumulation amount of 3360 g and a decomposition efficiency of 20%. The necessary calorific value total to evaporate moisture containing in swine waste was 540 Kcal per day. However, the potential calorific value generated by decomposition of swine waste was only 32 Kcal per day. Thus, it was obvious that the generated calorific value was much less than a necessary calorific value, resulting in moisture accumulation, as shown in Fig. 7(b), and low reactor temperature where the decomposition of swine wastes could not be expected. The behavior of the TOP reactor when the efficiency of cooking oil was set at 30%-wt of swine waste is shown in Fig. 7 (c). Under this condition, 300 g of swine waste and 90 g of cooking oil were fed to the reactor every 24-h. It was assumed
356
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 4 8 e3 5 6
that moisture was evaporated rapidly and the moisture content decreased to approximately 33% after 14 days. The average maximum temperature rose to 65 C immediately after the injection of the organic matter, which could be observed in the experimental result. The amount of organic matter injected in 14 days was 5460 g while the accumulation was 980 g, resulting in a decomposition efficiency of approximately 82% by the simulation. The experiment result was the accumulation amount of 1100 g and the decomposition efficiency of 80%. Since moisture content decreased to 33%, it was expected that decomposition efficiency would become lower if the treatment was continued. As shown in Table 4, the generated calorific value was much higher than a necessary calorific value, leading to a low moisture content of 33%. From these results, 15%-wt of swine waste was confirmed as the proper amount of the additional heat source of cooking oil. The simulation analysis showed that complete decomposition and evaporation of organic matter could be realized by proper calorific value control. Further, the heat loss equation was calculated from this 15%wt of swine waste. The simulation model analysis using this heat loss equation and the experiment result corresponded well in other conditions, too. Therefore, the model’s appropriateness was proven. Moreover, by using this simulation model, a forecast of the process became possible in other experimental conditions along with the appropriate process conditions.
5.
Conclusion
The following conclusions can be derived from the development and application of the TOP simulation model for treating swine waste in this study. A simulation model for the TOP treatment of swine waste was established on the basis of three-dimensional relationships among temperature, moisture content, and decomposition rate constant. The decomposition rate of swine wastes is higher than that of cooking oil. Cooking oil decomposition is more difficult than swine waste because the necessary activation energy for the cooking oil decomposition is higher than that of swine waste. The decomposition rate of swine waste and cooking oil under various moisture contents was described by the Arrhenius equation. The optimal temperature and moisture content were 60 C and 60% for swine waste and 60 C and 50% for cooking oil, respectively. The temperature and moisture content at the simulation value corresponded to the experiment value when the injection amount of supplemental heat source was 0%, 15%, and 30%. Therefore, the validity of the model was proven from these comparisons. Thus, the TOP simulation model could be constructed using the carbon, heat and moisture balance. When the injection amount of the additional heat source changed, the calculation of the temperature-moisture content simulation and the organic matter accumulation amount became possible. Therefore, the relationship between the
injection calorific value and the process mechanism of TOP was interpreted. If there is a high amount of calorific values in the injected organic matter, temperature and moisture evaporation are increased; as a result, the TOP process becomes impossible. In contrast, if the injected organic matter has a low amount of calorific values, temperature and moisture evaporation are decreased, which also inhibits the TOP process. Thus, if the amount of injection organic matter, calorific value, and moisture are known, the interpretation of the swine waste treatment possibility using the TOP model would become possible.
references
Andani, F., Baido, D., Calcaterra, E., Genevini, P., 2002. The influence of biomass temperature on biostabilizationebiodrying of municipal solid waste. Bioresour. Technol 83, 173e179. Chang, J., Tsai, J., Wu, K., 2006. Thermophilic composting of food waste. Bioresour. Technol 97, 116e122. ski, M., 2008. Testing of thermal properties of Klejment, E., Rosin compost from municipal waste with a view to using it as a renewable, low temperature heat source. Bioresour. Technol 99, 8850e8855. Kosseva, M., Fatmawati, A., Palatova, M., Kent, C., 2007. Modelling thermophilic cheese whey bioremediation in a one-stage process. Biochem. Eng. J 35, 281e288. Liang, C., Das, K.C., McClendon, R.W., 2003. The influence of temperature and moisture contents regimes on the aerobic microbial activity of a biosolids composting blend. Bioresour. Technol 86, 131e137. 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 Res 41, 4085e4096. Miguel, A., Funamizu, N., Takakuwa, T., 2004a. Temperature effect on aerobic biodegradation of feces using sawdust as a marix. Water Res 38, 2406e2416. Miguel, A., Funamizu, N., Takakuwa, T., 2004b. Modeling of aerobic biodegradation of feces using sawdust as a matrix. Water Res 38, 1327e1339. Miguel, A., Funamizu, N., Takakuwa, T., 2005. Biological activity in the composting reactor of the bio-toilet system. Bioresour. Technol 96, 805e812. Nakano, K., Matsumura, M., 2001. Improvement of treatment efficiency of thermophilic oxic process for highly concentrated lipid wastes by nutrient supplementation. J. Biosci. Bioeng 92 (6), 532e538. Nakano, K., Matsumura, M., 2002. Utilization of dehydrated sewage sludge as an alternative nutrient to stimulate lipid waste degradation by the thermophilic oxic process. J. Biosci. Bioeng 94 (2), 113e118. Ravi, M., Jhalani, A., Sinha, S., Ray, A., 2004. Development of a semi-empirical model for pyrolysis of an annular sawdust bed. J. Anal. Appl. Pyrolysis 71, 353e374. Walker, M., Zhang, Y., Heaven, S., 2009. Charles Banks Potential errors in the quantitative evaluation of biogas production in anaerobic digestion processes. Bioresour. Technol 100, 6339e6346.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 e3 6 5
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Fouling of microfiltration membranes by organic polymer coagulants and flocculants: Controlling factors and mechanisms Sen Wang a, Charles Liu b, Qilin Li a,* a b
Department of Civil and Environmental Engineering, Rice University, 6100 Main Street, Houston, TX 77005, United States Pall Corporation, 25 Harbor Park Dr., Port Washington, NY 11050, United States
article info
abstract
Article history:
Organic polymers are commonly used as coagulants or flocculants in pretreatment for
Received 10 March 2010
microfiltration (MF). These high molecular weight compounds are potential membrane
Received in revised form
foulants when carried over to the MF filters. This study examined fouling of three MF
4 August 2010
membranes of different materials by three commonly used water treatment polymers:
Accepted 7 August 2010
poly(diallyldimethylammonium) chloride (pDADMAC), polyacrylamide (PAM), and poly
Available online 14 August 2010
(acrylic acid-co-acrylamide (PACA) with a wide range of molecular weights. The effects of polymer molecular characteristics, membrane surface properties, solution condition and
Keywords:
polymer concentration on membrane fouling were investigated. Results showed severe
Polymer flocculant
fouling of microfiltration membranes at very low polymer concentrations, suggesting that
Constant flux
residual polymers carried over from the coagulation/flocculation basin can contribute
Microfiltration
significantly to membrane fouling. The interactions between polymers and membranes
Membrane fouling
depended strongly on the molecular size and charge of the polymer. High molecular weight, positively charged polymers caused the greatest fouling. Blockage of membrane pore openings was identified as the main fouling mechanism with no detectable internal fouling in spite of the small molecular size of the polymers relative to the membrane pore size. Solution conditions (e.g., pH and calcium concentration) that led to larger polymer molecular or aggregate sizes resulted in greater fouling. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Microfiltration (MF) has been increasingly applied to drinking water and wastewater treatment due to the small footprint, superior treated water quality, and high level of automation compared to conventional treatment processes. Major limitations of the MF technology include inefficiency in natural organic matter (NOM) removal (Vickers et al., 1995), and membrane fouling. Many source waters contain significant amount of NOMda precursor of harmful disinfection byproducts. In addition, membranes can be fouled by NOM
over time, leading to loss in water production and requiring more frequent cleaning. As a result, coagulation/flocculation is often used as pretreatment to increase NOM removal and to control membrane fouling. Polymers are widely employed in the coagulation/flocculation process in conventional water treatment systems (Bolto and Gregory, 2007). They are known to improve effluent water quality by increasing floc size and strength, reduce alkalinity consumption, and alleviate sludge handling and disposal problems (Gray and Ritchie, 2006; Jin et al., 2003; Taylor et al., 2002; Zhao, 2004). Therefore, polymers are also adopted in the
* Corresponding author. Tel.: þ1 713 348 2046; fax: þ1 713 348 5268. E-mail address:
[email protected] (Q. Li). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.009
358
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 e3 6 5
coagulation/flocculation pretreatment process for many MF systems (Nozic et al., 2001). However, it is unclear whether the use of polymers in MF systems is beneficial. Firstly, MF membranes can remove significantly smaller particles than those removed by the conventional treatment process; the use of polymers may not have measurable effect on turbidity removal. Secondly, the effect of polymers on NOM removal has been controversial. Although a number of studies reported improved NOM removal when polymers were used (Jarvis et al., 2008; Kim and Walker, 2001; Lee and Westerhoff, 2006), others reported no or negative impact of polymers on NOM removal (Chang et al., 2005; Jarvis et al., 2006). Finally, polymers are high molecular weight (MW) organic compounds. When carried over to membrane filters residual polymers can potentially foul the membranes. The flocs containing polymers may also foul the membrane more than those formed in the absence of polymers. Therefore, the effect of polymers on membrane performance needs to be carefully assessed for system optimization. Fouling of MF or UF membrane by various macromolecules such as proteins and polysaccharides have been intensively studied (de Lara and Benavente, 2009; Guell and Davis, 1996; Kanani et al., 2008; Katsoufidou et al., 2007; Lin et al., 2008; Loh et al., 2009; Susanto and Ulbricht, 2005; Ye et al., 2005; Zator et al., 2009). These macromolecules have been found to foul MF and UF membranes with pores much larger than the macromolecules themselves by accumulating both inside membrane pores and on the membrane surface. Internal fouling, i.e., foulant accumulation in the membrane pores, is usually attributed to macromolecule adsorption due to electrostatic, hydrophobic interactions and hydrogen bonding (Li and Elimelech, 2004; Nakamura and Matsumoto, 2006; Yamamura et al., 2008); it has been demonstrated by flux and hydraulic resistance analysis (Guell and Davis, 1996), measurement of streaming potential across the membrane (de Lara and Benavente, 2009; Nakamura and Matsumoto, 2006), and confocal laser scanning microscopy (CLSM) (Zator et al., 2009). Using CLSM, Zator et al. (2009) showed that MF of ternary or binary solutions of BSA, dextran, and tannic acid resulted in an internal fouling layer 1e3 mm deep from the membrane surface. Macromolecules such as BSA were also found to form aggregates in the feed solution (Maruyama et al., 2001); these aggregates could block membrane pores and subsequently allow deposition of more monomers and aggregates to form a multilayer cake (Kanani et al., 2008). The fouling mechanisms of dextran and other macromolecules were found to be similar to BSA (Guell and Davis, 1996; Katsoufidou et al., 2007; Loh et al., 2009; Susanto and Ulbricht, 2005; Ye et al., 2005). Water treatment polymers are macromolecules with many properties similar to proteins and polysaccharides. However, their membrane fouling potential has not been carefully evaluated. The study reported here systematically investigated the impact of free polymers (i.e., those not bound to a floc) on membrane fouling. The roles of polymer molecular characteristics, MF membrane properties, and solution conditions of feed water were evaluated and the fouling mechanism was elucidated. The use of polymers in conjunction with a hydrolyzing metal salt as the primary coagulant for MF of surface water will be addressed in a separate publication.
2.
Materials and methods
2.1.
Polymers and membranes
Three types of polymers commonly used in water treatment were tested in this study: poly(diallyldimethylammonium) chloride (pDADMAC), poly(acrylic acid-co-acrylamide) (PACA), and polyacrylamide (PAM). To avoid interference from additives and impurities commonly found in commercial water treatment polymers, analytical grade polymers were purchased from SigmaeAldrich (St. Louis, MO). Relevant characteristics of these polymers are listed in Table S1 (Supplementary Information). It is noted that the two PACA polymers have different monomer composition: The 520 kDa PACA contains 80% acrylamide and 20% acrylic acid, while the 200 kDa PACA contains 20% acrylamide and 80% acrylic acid. Electrophoretic mobility and hydrodynamic diameter of each polymer were determined by phase analysis light scattering (PALS) and dynamic light scattering (DLS), respectively, using a Zen3600 Zetasizer (Malvern Ltd., Malvern, UK) under the same solution conditions as those used in the filtration experiments described later. DLS measurements were initially performed with various polymer concentrations to determine an appropriate concentration. A concentration of 1 g/L was found to provide sufficient light scattering for the measurement. Three types of MF membranes, denoted MF-1, MF-2, and MF-3, and made of modified polyvinylidene fluoride (PVDF), polyethersulfone (PES) and polysulfone (PS), respectively (Pall Corporation, East Hills, NY), were employed in this study. Two batches of MF-1 membrane were used. The first batch was named MF-1 and the second batch was named MF-1a. All three types of membranes have a nominal pore size of 0.2 mm. Flat sheet membranes were cut into circular coupons of 2.5 cm diameter and stored at 4 C in ultrapure water generated by a Barnstead Epure purification system (Barnstead Thermolyne, IA, USA). The storage water was changed weekly. Surface zeta potential of the membranes was determined using a ZetaCAD streaming potential analyzer (CAD Instrumentation, Les Essarts-le-Roi, France) under the same solution conditions used in the filtration experiments.
2.2.
Feed water
Synthetic feed water was prepared using the aforementioned ultrapure water. The feed water used in all experiments had an ionic strength of 10 mM made of either 10 mM NaCl or 7 mM NaCl and 1 mM CaCl2. The concentration of polymers in the feed water was 0.05, 0.1, or 0.5 mg/L. These concentrations were chosen to represent the possible concentration range of the free polymer molecules carried over from the coagulation/ flocculation basin to the membrane reactor. 0.1 M HCl, 0.1 M NaOH, and 0.001 M Na2HPO4 were used to adjust the solution pH to 4, 7 and 10.
2.3.
Membrane filtration experiments
Membrane filtration experiments were performed with a bench-scale dead-end filtration system in a constant-flux
359
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 e3 6 5
mode. The schematic of the experimental setup is shown in Fig. S1 (Supplementary Information). A Filtertec peristaltic pump (SciLog, Middleton, WI) fed the water from a 4 L reservoir to a 10 mL Amicon stirred cell (Millipore, Billerica, MA), where the membrane sample was housed. Two pressure sensors, located immediately upstream and downstream of the filtration cell, measured the pressures of the feed and filtrate streams. Membrane filtrate was collected in a container on a bench-top electronic balance, which measured the cumulative mass of filtrate for calculation of membrane flux. The pump and the balance were interfaced with a lab PC to collect data of the measured flux and transmembrane pressure (TMP). In every filtration experiment, ultrapure water was first filtered for 10 min; synthetic feed water spiked with the desired polymer concentration was then filtered through the membrane for 60 min. The initial TMP in all experiments was controlled at 10 psi and the corresponding operating flux was 1.10 103, 2.55 103 and 1.36 103 m/s (3973, 9167 and 4889 LMH) for MF-1, MF-2 and MF-3, respectively. All experiments were run at a feed water temperature of 22 C and were repeated at least once. A new membrane coupon was used for every filtration experiment. TMP at the end of each filtration was normalized with respect to the initial TMP and used as the measure for membrane fouling rate, the normalized TMP was named as NTMP.
the membranes follow the order of MF-3 (PS) < MF-1 (PVDF) < MF-2 (PES) over the pH range tested. Membrane permeability was determined by clean water flux measurement over a pressure range of 5e50 psi and was 1.4 108, 1.21 108, 3.71 108, and 2.09 108 m/s-Pa (347, 300, 921, and 520 LMH/psi) for MF-1, MF-1a, MF-2 and MF-3, respectively. SEM images of the clean membranes are presented in Fig. 5. Polymer molecular size was characterized by hydrodynamic diameter. Fig. 1 shows the intensity based particle size distribution of the polymers. Each distribution is an average of at least 7 measurements. All polymers except the 520 kDa PACA exhibit a bimodal or multi-modal size distribution. The peaks at the larger size positions are attributed to high MW impurities or aggregate formation at the concentration used for size measurement, i.e., 1 g/L. It is noted that the intensity of light scattered by a particle is proportional to the sixth power of its diameter (Xu, 2002). Therefore, the apparently large peaks for the larger sizes represent only a small number of molecules or aggregates. The peak of the smaller size represents the size of individual polymer molecules, which accounts for the majority of the particles detected.
10.0
2.4. Scanning electron microscopy (SEM) characterization of clean and fouled membranes
Intensity, %
8.0
pH=7, 400-500kDa pDADMAC pH=7, 200-350kDa pDADMAC pH=7, 100-200kDa pDADMAC pH=7, <100kDa pDADMAC pH=7, 5000-6000kDa PAM
6.0
4.0
2.0
0.0
20.0
b
pH=4, 200kDa PACA pH=7, 200kDa PACA pH=10, 200kDa PACA
15.0
pH=4, 520kDa PACA pH=7, 520kDa PACA pH=10, 520kDa PACA
Intensity, %
A field Emission scanning electron microscope (FEI Company, Philips XL30) was used to characterize the MF membranes before and after filtration of the ultrapure water and the polymer containing feed water. Membrane samples after filtration were carefully removed from the filtration cell and air dried. Both top surface and cross-section of the membranes were analyzed. For cross-section imaging, the fouled membrane samples were first cut into thin strips. A small incision was made on each long side of the strip along the line where the cross-section was to be imaged. One end of the membrane strip was then dipped into liquid nitrogen for at least 10 s, leaving the incisions just above the liquid nitrogen surface. When taken out from the liquid nitrogen, the frozen membrane sample was immediately cleaved by a gentle tap near the incisions using a pair of clean forceps. This method provides a clean cut exposing the cross-section of the membrane sample while preserving the structure of the fouling layer. Membrane surface porosity and pore size distribution were determined by analyzing surface images using the Image J software (Nation Institute of Health).
a
pH=4, 1.5kDa PAM pH=7, 1.5kDa PAM
10.0
pH=10, 1.5kDa PAM
5.0
3.
Results and discussion
3.1.
Characterization of membranes and polymers
The surface zeta potentials of the three membranes at three different pH values are presented in Fig. S2 (Supplementary Information). All three membranes are negatively charged in the pH range of 4e10, and the negative charge increased with increasing pH. The magnitude of the negative zeta potential of
0.0 0.1
1
10 100 Hydrodynamic diameter, nm
1000
10
4
Fig. 1 e Molecular size distribution: (a) pDADMACs and 5000e6000 kDa PAM at pH 7; (b) 1.5 kDa PAM, 200 kDa and 520 kDa PAA at pH 4, 7, and 10.
360
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 e3 6 5
2.5
3.0
a
a
MF-1 MF-2 MF-3
2.0
2.0
P/ P0
P/ P0
2.5
4 7 10
1.5
1.5
1.0 1.0
0.5
0.5 0.0
0.0 pDADMAC <100kDa
P/ P0
1.5
b
PACA 520kDa
pDADMAC 400-500kDa
PAM 5000 -6000kDa
pDADMAC 400-500kDa
2.5
b
4 7 10
2.0
P / P0
2.0
pDADMAC pDADMAC 100-200kDa 200-350kDa
1.0
4 7 10
1.5
1.0 0.5
0.5 0.0 PAM 1.5kDa
0.0
PAM 5000-6000kDa
Fig. 2 e Effect of polymer molecular weight on membrane fouling by (a) pDADMAC, and (b) PAM. Feed water contains 10 mM NaCl and 0.1 mg/L of the polymer.
The size distributions of the four pDADMACs and the 5000e6000 kDa PAM were independent of solution pH. Therefore, only the distributions at pH 7 were presented in Fig. 1 (a). The particle sizes of the four pDADMACs follow the same order as their MW. The 5000e6000 kDa PAM was much larger than the pDADMACs. Solution pH had significant impact on the molecular and aggregate sizes of PACAs and the 1.5 kDa PAM (Fig. 1(b)). At pH 7 and 10, the 200 kDa PACA showed a primary peak ate20 nm; this peak shifted toe40 nm at pH 4. This was attributed to aggregate formation due to the lower electrostatic repulsion and the formation of hydrogen bond at low pH. The size of the 520 kDa PACA also increased with decreasing pH, suggesting aggregation at lower pH. The molecular size of the 1.5 kDa PAM was much smaller than the rest of the polymers (e2 nm), consistent with its low MW. Fig. S3 shows the electrophoretic mobility of the polymers as a function of pH. All pDADMAC molecules were highly positively charged and the electrophoretic mobility was constant over the pH range tested. This is consistent with their quaternary amine functionality. The 5000e6000 kDa PAM showed near-zero electrophoretic mobility at all measured pH values. The 1.5 kDa PAM, however, was slightly
MF-1, MF-1, PACA MF-2, MF-2, PACA PACA(200kDa) (520kDa) PACA(200kDa) (520kDa) Fig. 3 e Fouling of the three MF membranes by different polymers. (a) Effect of polymer charge; (b) Effect of polymer chemical functionality. The feed water contained 10 mM NaCl and 0.1 mg/L of the polymer. pH [ 7. Filtrate volume for all experiments was 1.95 L.
positively charged at low pH and negatively charged at higher pH, possibly due to the presence of carboxyl and amino functionalities originated from hydrolysis of the amide groups. Both the 200 kDa and the 520 kDa PACA were negatively charged at all pHs measured due to deprotonation of carboxyl groups. The higher content of acrylic acid units in the 200 kDa PACA is responsible for its higher negative electrophoretic mobility than that of the 520 kDa PACA.
3.2.
MF membrane fouling by polymers
In general, all polymers except the 1.5 kDa PAM caused significant fouling of all three MF membranes at a concentration as low as 0.05 mg/L. The effects of the polymer MW, polymer and membrane surface charge, solution condition and polymer concentration are described in details below.
3.2.1.
Effect of polymer MW
Among the various factors investigated, polymer MW had the greatest impact on MF membrane permeability. Fig. 2
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 e3 6 5
361
demonstrates the effect of the polymer MW on fouling of the MF-1 membrane by pDADMAC and PAM at the feed concentration of 0.1 mg/L. The TMP at the end of the 60 min filtration increased significantly with increasing MW of pDADMAC and PAM. Only slight fouling was observed with the <100 kDa pDADMAC and the 1.5 kDa PAM: 6.8e8.7% and 0.7e1.7% increase in TMP respectively. The higher MW polymers, on the other hand, caused much more severe fouling: At the end of the 60 min filtration, the TMP rose by 130% and 60% for the 400e500 kDa pDADMAC and the 5000e6000 kDa PAM, respectively. Similar effects of MW were observed for MF-2 and MF-3 membranes (data not shown).
the anionic PACA in spite of the much higher MW of the PAM. This is attributed to the strong electrostatic attraction between the positively charged pDADMAC (Fig. S3) and the negatively charged membrane surfaces (Fig. S2), which leads to adsorption of pDADMAC molecules on the membrane. On the other hand, the anionic PACA caused the least fouling among the three polymers due to the electrostatic repulsion between PACA molecules and the membrane surfaces as well as the lower MW compared to the PAM. Comparison of the fouling behaviors of the three membranes suggests that mechanisms other than electrostatic interaction are also important. As shown in Fig. S2, the MF-2 membrane had the highest negative surface zeta potential, and the MF-3 membrane the lowest. However, MF-1 experienced the most fouling by all polymers tested. One possible reason is the high surface porosity (Table 1) and roughness of the MF-1 membrane (see Fig. 5(a), Fig. S5(a) and S6(a)), which provides more surface area at the membrane pore openings for polymer accumulation and subsequent pore blockage, the main mechanism responsible for fouling by the polymers as explained later. In addition to charge, polymer chemical functionality also played a role in membrane fouling. The 520 kDa PACA contains 80% acrylamide and 20% acrylic acid while the 200 kDa PACA contains 20% acrylamide and 80% acrylic acid. The difference in acrylic acid content resulted in notable difference in their fouling potential, as depicted in Fig. 3(b). The 200 kDa PACA consistently fouled the MF-1 membrane more than the 520 kDa PACA at all three pHs even though it has much higher negative charge and lower MW. Different results were observed with the MF-2 membrane. The two PACA polymers fouled the MF-2 membrane similarly except at pH 4, when the 200 kDa PACA showed drastically higher fouling potential: The normalized TMP reached as high as 1.78. Apparently, the fouling behavior of the 200 kDa PACA cannot be explained simply by electrostatic interaction. It is speculated that the higher rate of MF-1 fouling by the 200 kDa PACA is due to formation of aggregates through intermolecular hydrogen bonding between carboxyl groups. This is supported by the measured molecular size distribution in Fig. 1. For the MF-2 membrane, whose surface zeta potential is much higher than that of MF-1, the stronger electrostatic repulsion between the highly negatively charged 200 kDa PACA and the membrane partly negates the effect of the aggregates, resulting in fouling potential similar to that of the 520 kDa PACA and overall less fouling than MF-1. We speculate that the severe fouling of MF-2 by the 200 kDa PACA at pH 4 may be due to hydrogen bond formation between the carboxylic acid groups on the polymer and the sulfonyl groups on the MF-2 membrane surface.
3.2.2.
3.2.3.
3.0
a
PAM w/ Ca PAM , w/o Ca PACA w/ Ca PACA w/o Ca pDADMAC w/ Ca pDADMAC w/o Ca
P/ P0
2.5
2.0
1.5
1.0 0 1.8
10
b
30
40
50
60
50
60
10mM NaCl 1mM CaCl2, 7mM NaCl
1.6 P/ P0
20
1.4
1.2
1.0
0.8 0
10
20
30
40
Time, min
Fig. 4 e Effect of calcium ions on fouling of: (a) MF-1 by the 5000-6000 kDa PAM, 520 kDa PACA and 400-500 kDa pDADMAC; (b) MF-1a by 520 kDa PAA. Solutions without Ca contained 10 mM NaCl, and those with Ca contained 7 mM NaCl and 1 mM CaCl2. pH [ 7.
Effect of polymer and membrane surface charge
Charge of the polymer molecules was also found to play an important role in fouling of MF membranes. Fig. 3(a) compares the normalized TMPs of the three membranes after fouled by the 400e500 kDa pDADMAC, the 5000e6000 kDa PAM and the 520 kDa PACA at pH 7. The TMPs at the same permeate volume (1.95 L) were used so that the comparison was based on the same polymer load for all membranes. The cationic pDADMAC caused notably more fouling than the nonionic PAM and
Effect of solution chemistry
3.2.3.1. Effect of pH. Solution pH is usually an important factor in membrane fouling because it affects the charge of ionizable foulants (e.g., polyelectrolytes) and surface charge of the membranes. In our study, however, the effect of feed water pH was small in most cases, as shown in Figs. 2 and 3(b). For the nonionic PAM, this is due to the lack of electrostatic interaction. The charge of the cationic pDADMAC is independent of pH
362
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 e3 6 5
Fig. 5 e SEM images of clean and fouled MF-1: (a) clean membrane surface; (b, c) clean membrane cross sectional; (d, e) fouled membrane surface; (f) fouled membrane cross sectional. The feed water contained 10 mM NaCl, and 0.1 mg/L 400e500 kDa pDADMAC. pH [ 7.
because the quaternary amine on the polymer does not dissociate as pH changes. The negative surface zeta potential of all membranes increases with increasing pH (Fig. S2). In spite of the expected increase in electrostatic attraction caused by higher negative surface charge of membranes at higher pH, no consistent trend in membrane fouling by pDADMAC was observed. A possible explanation is that the attractive interaction between the highly positively charged pDADMAC and the negatively charged membranes is very strong even at the lowest pH tested; further increase in negative membrane
surface charges does not cause notable changes. Similarly, the effect of pH on membrane fouling by PACA is small due to the strong electrostatic repulsion even at the lowest pH. One exception is the fouling of the MF-2 membrane by the 200 kDa PACA at pH 4, which was discussed above.
3.2.3.2. Effect of calcium. Calcium ion has significant influence on membrane surface charge (Saravia et al., 2006), and can effectively neutralize negative charges of polyelectrolytes to form intermolecular bridging, leading to changes in
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 e3 6 5
Table 1 e Surface porosity of Clean and Fouled Membranes.a Parameters 30 3 30/3 DP/DP0
MF-1
MF-2
MF-3
31.2 0.9 18.7 2.9 1.7 2.2
29.6 4.7 15.5 1.5 1.9 2.7
19.6 2.0 12.3 1.3 1.6 1.4
a The feed solution contained 10 mM NaCl, and 0.1 mg/L 400e500 kDa pDADMAC at pH 7.
polymer molecular conformation and aggregation (Hong and Elimelech, 1997; Li and Elimelech, 2004; Yuan and Zydney, 1999). Calcium has been found to aggravate membrane fouling by bolvine serum albumin (BSA) (Palecek et al., 1993) and humic acid (Bouchard et al., 1997; Costa et al., 2006; Hong and Elimelech, 1997; Li and Elimelech, 2004), as well as to increase irreversible membrane fouling by sodium alginate (van de Ven et al., 2008). To investigate the effect of calcium, 1 mM CaCl2 was added to the synthetic feed water and NaCl concentration was reduced to 7 mM to maintain ionic strength of 10 mM. Each experiment was repeated at least five times. The results are presented in Fig. 4(a). The presence of calcium did not change the fouling potential of pDADMAC and PAM. These results are consistent with the molecular structures of pDADMAC and PAM, which do not have functional groups that interact with Ca2þ specifically. The charge screening effect of Ca2þ was not significant enough to cause notable changes in the adsorption of pDADMAC and PAM onto membrane surface. Calcium ions caused a very slight decrease in the fouling of the MF-1 membrane by the 520 kDa PAA. This effect is more evident with the MF-1a membrane (Fig. 4(b)), which has smaller pores than MF-1 as indicated by the lower permeability. The fouling of MF-1a by the 520 kDa PACA was notably
Percentage, %
100.0
80.0
60.0
40.0
20.0 0.01
Clean MF-1 0.1mg/L 5000-6000kDa PAM fouled MF-1 0.1mg/L 400-500kDa pDADMAC fouled MF-1 0.1mg/L 520kDa PACA fouled MF-1 0.1 1 Feret Diameter, µm
10
Fig. 6 e Pore size distribution of clean and fouled MF-1 membranes. The feed solution had a pH of 7 and contained 10 mM NaCl. Feret diameter is the longest distance between any two points along the pore opening circle.
363
less in the presence of Ca2þ, while fouling by PAM or pDADMAC was not affected by Ca2þ (data not shown). This is attributed to the intra-molecular complexation between Ca2þ and carboxyl groups in PACA molecules, which leads to a more coiled and compact conformation of the PACA molecules (Peng and Wu, 1999). This change of molecular configuration was confirmed by molecular size measurement: The number-mean molecular size of the 520 kDa PACA is 12% smaller in the presence of 1 mM Ca2þ than that measured without Ca2þ. As the molecular size decreased, membrane fouling was reduced due to less pore blockage, as discussed in “Fouling mechanism” later.
3.2.4.
Effect of polymer concentration
Three different concentrations of the polymer, 0.05, 0.1 and 0.5 mg/L, were used in the filtration experiments. The results for the MF-1 membrane are shown in Fig. S4. The fouling of the membrane was unexpectedly severe, even when polymer concentration was as low as 0.05 mg/L. The extent of fouling increased greatly with increasing polymer concentration. At 0.5 mg/L, the 400e500 kDa pDADMAC and the 520 kDa PACA resulted in 420 and 243% increase in TMP, respectively.
3.3.
Fouling mechanism
MF membrane fouling is usually attributed to four mechanisms (Hermia, 1982; Hlavacek and Bouchet, 1993): (a) Constriction of membrane pores d restricting flow by the foulants smaller than the membrane pores adsorbed onto the pore walls; (b) complete blockage of membrane pores by the foulant d stopping both the solvent (e.g., water) and the solute (e.g., the foulant) through the blocked pores; (c) intermediate blocking of membrane pores d restricting solute flowing through the blocked pores but allowing the solvent to go through at a lower rate; (d) cake/gel layer formation due to accumulation of foulants on the membrane surface. SEM analysis of the clean and fouled membranes revealed that surface pore blockage was the predominant fouling mechanism. The top and cross-sectional views of the MF-1 membranes fouled by 0.1 mg/L pDADMAC are presented in Fig. 5 together with the clean membranes. pDADMAC formed large aggregates on the membrane surface, and these aggregates blocked openings of the membrane pores, leading to a reduction in membrane surface porosity. Careful inspection of the cross-sectional images (e.g., Fig. 5(f)) did not find any noticeable internal fouling (i.e., accumulation of foulants inside membrane pores), suggesting that pore constriction was not an important mechanism. Fouling of MF-2 and MF-3 was similar as shown in Figs. S5 and S6. Similar phenomena were also observed with PACA and PAM (data not shown). Since the permeate flow is laminar, the flux can be approximated using the Hagen Poiseuille equation (equation (1)), according to which the normalized TMP, DP/DP0, is related to the changes in membrane porosity (equation (2)). J¼
3r2 DP 8hDx
DP 30 ¼ 3 DP0
(1)
(2)
364
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 e3 6 5
where DP0 and DP are initial and final TMP, 30 and 3 are surface porosity of clean and fouled membranes, r is the membrane pore radius and Dx is the effective pore length. Table 1 presents the membrane surface porosity before and after fouling by the 400e500 kDa pDADMAC measured by SEM imaging. The measured porosity change of MF-3 agrees well with the observed TMP increase, confirming that blockage of the pore opening is the main fouling mechanism. The measured porosity changes of the MF-1 and MF-2 membranes, however, predict lower TMP increase than that observed experimentally. This can be partially attributed to artifacts from image processing due to the highly heterogeneous surface of the MF-1 and MF-2 membranes, which makes defining membrane pores difficult. With such artifacts, the possibility of intermediate blocking cannot be excluded. The measured molecular size distributions of the polymers in Fig. 1 suggest that the majority of the polymer molecules exist as individual molecules or small aggregates even at a concentration of 1 g/L, much smaller than the aggregates found in the SEM images. Therefore, it is speculated that the aggregates observed in the SEM images were formed on the membrane surface instead of in the bulk solution. This is supported by an approximate calculation of the foulant mass using the measured aggregate size and surface porosity, which suggests that the aggregates found on the membrane surface accounts for almost all the polymers in the feed solution. It is hypothesized that the polymer molecules and the small aggregates formed in the bulk solution preferably adsorb or deposit on the opening of small pores. They subsequently act as nuclei to catalyze formation of larger aggregates on the membrane surface as filtration proceeds. This hypothesis is supported by the observed changes in MF-1 membrane pore size distribution after fouling (Fig. 6). As shown in Fig. 6, after fouling by the 400e500 kDa pDADMAC, 520 kDa PACA, and 5000e6000 kDa PAM, the pore size distribution of MF-1 shifted towards the larger pore size range, indicating that the smaller pores were preferentially blocked during filtration.
4.
Conclusions
This study demonstrates that carry-over of polymers used in the coagulation/flocculation pretreatment can cause severe fouling of MF membranes even at very low concentrations. MF membrane fouling by polymers strongly depends on the molecular weight, charge, and concentration of the polymer, as well as the membrane surface properties. Cationic polymers tend to cause greater fouling than anionic and nonionic polymers in synthetic feed water due to the strong electrostatic attraction between the positively charged polymer and the negatively charged membrane surface. Among polymers of the same charge, those with higher molecule weight have greater fouling potential. Although changes in electrostatic interaction due to changes in either pH or calcium concentration did not have much impact on MF fouling, pH or calcium concentration can affect fouling by mediating specific foulantefoulant or foulantemembrane interactions. In spite of the small size of the polymers relative to the size of MF membrane pores, surface pore blockage was found to be the
predominant fouling mechanism. Formation of large aggregates on the membrane surface suggests that prediction of fouling mechanisms based on foulant molecular size and membrane pore size can be erroneous sometimes. Membrane surface chemical and physical heterogeneity and specific membraneefoulant interactions may be more important than the physical screening mechanism.
Acknowledgements We thank Pall Corporation for partially funding this project and for providing the membrane samples used in this study.
Appendix. Supplementary data Supplementary data associated with the article can be found in online version, at doi:10.1016/j.watres.2010.08.009
references
Bolto, B., Gregory, J., 2007. Organic polyelectrolytes in water treatment. Water Res. 41 (11), 2301e2324. Bouchard, C.R., Jolicoeur, J., Kouadio, P., Britten, M., 1997. Study of humic acid adsorption on nanofiltration membranes by contact angle measurements. Can. J. Chem. Eng. 75 (2), 339e345. Chang, E.E., Chiang, P.C., Tang, W.Y., Chao, S.H., Hsing, H.J., 2005. Effects of polyelectrolytes on reduction of model compounds via coagulation. Chemosphere 58 (8), 1141e1150. Costa, A.R., de Pinho, M.N., Elimelech, M., 2006. Mechanisms of colloidal natural organic matter fouling in ultrafiltration. J. Memb. Sci. 281 (1e2), 716e725. Gray, S.R., Ritchie, C.B., 2006. Effect of organic polyelectrolyte characteristics on floc strength. Colloids Surf. A Physicochem. Eng. Asp. 273 (1e3), 184e188. Guell, C., Davis, R.H., 1996. Membrane fouling during microfiltration of protein mixtures. J. Memb. Sci. 119 (2), 269e284. Hermia, J., 1982. Constant pressure blocking filtration laws e application to power-law non-Newtonian fluids. Trans. Inst. Chem. Eng. 60 (3), 183e187. Hlavacek, M., Bouchet, F., 1993. Constant flow-rate blocking laws and an example of their application to dead-end microfiltration of protein solutions. J. Memb. Sci. 82 (3), 285e295. Hong, S.K., Elimelech, M., 1997. Chemical and physical aspects of natural organic matter (NOM) fouling of nanofiltration membranes. J. Memb. Sci. 132 (2), 159e181. Jarvis, P., Jefferson, B., Parsons, S.A., 2006. Floc structural characteristics using conventional coagulation for a high doc, low alkalinity surface water source. Water Res. 40 (14), 2727e2737. Jarvis, P., Parsons, S.A., Henderson, R., Nixson, N., Jefferson, B., 2008. The practical application of fractal dimension in water treatment practice e the impact of polymer dosing. Sep. Sci. Technol. 43 (7), 1785e1797. Jin, B., Wilen, B.M., Lant, P., 2003. A comprehensive insight into floc characteristics and their impact on compressibility and settleability of activated sludge. Chem. Eng. J. 95 (1e3), 221e234.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 e3 6 5
Kanani, D.A., Sun, X.H., Ghosh, R., 2008. Reversible and irreversible membrane fouling during in-line microfiltration of concentrated protein solutions. J. Memb. Sci. 315 (1e2), 1e10. Katsoufidou, K., Yiantsios, S.G., Karabelas, A.J., 2007. Experimental study of ultrafiltration membrane fouling by sodium alginate and flux recovery by backwashing. J. Memb. Sci. 300 (1e2), 137e146. Kim, E.K., Walker, H.W., 2001. Effect of cationic polymer additives on the adsorption of humic acid onto iron oxide particles. Colloids Surf. A Physicochem. Eng. Asp. 194 (1e3), 123e131. de Lara, R., Benavente, J., 2009. Use of hydrodynamic and electrical measurements to determine protein fouling mechanisms for microfiltration membranes with different structures and materials. Sep. Purif. Technol. 66 (3), 517e524. Lee, W., Westerhoff, P., 2006. Dissolved organic nitrogen removal during water treatment by aluminum sulfate and cationic polymer coagulation. Water Res. 40 (20), 3767e3774. Li, Q.L., Elimelech, M., 2004. Organic fouling and chemical cleaning of nanofiltration membranes: measurements and mechanisms. Environ. Sci. Technol. 38 (17), 4683e4693. Lin, S.H., Hung, C.L., Juang, R.S., 2008. Applicability of the exponential time dependence of flux decline during dead-end ultrafiltration of binary protein solutions. Chem. Eng. J. 145 (2), 211e217. Loh, S., Beuscher, U., Poddar, T.K., Porter, A.G., Wingard, J.M., Husson, S.M., Wickramasinghe, S.R., 2009. Interplay among membrane properties, protein properties and operating conditions on protein fouling during normal-flow microfiltration. J. Memb. Sci. 332 (1e2), 93e103. Maruyama, T., Katoh, S., Nakajima, M., Nabetani, H., 2001. Mechanism of bovine serum albumin aggregation during ultrafiltration. Biotechnol. Bioeng. 75 (2), 233e238. Nakamura, K., Matsumoto, K., 2006. Properties of protein adsorption onto pore surface during microfiltration: effects of solution environment and membrane hydrophobicity. J. Memb. Sci. 280 (1e2), 363e374. Nozic, D.J., Freese, S.D., Thompson, P., 2001. Long term experience in the use of polymeric coagulants at Umgeni Water. Water Sci. Tech. Water Supply 1 (1), 43e50. Palecek, S.P., Mochizuki, S., Zydney, A.L., 1993. Effect of ionic environment on bsa filtration and the properties of bsa deposits. Desalination 90 (1e3), 147e159.
365
Peng, S.F., Wu, C., 1999. Light scattering study of the formation and structure of partially hydrolyzed poly(acrylamide)/ calcium(II) complexes. Macromolecules 32 (3), 585e589. Saravia, F., Zwiener, C., Frimmel, F.H., 2006. Interactions between membrane surface, dissolved organic substances and ions in submerged membrane filtration. Desalination 192 (1e3), 280e287. Susanto, H., Ulbricht, M., 2005. Influence of ultrafiltration membrane characteristics on adsorptive fouling with dextrans. J. Memb. Sci. 266 (1e2), 132e142. Taylor, M.L., Morris, G.E., Self, P.G., Smart, R.S., 2002. Kinetics of adsorption of high molecular weight anionic polyacrylamide onto kaolinite: the flocculation process. J. Colloid Interface Sci. 250 (1), 28e36. van de Ven, W.J.C., van’t Sant, K., Punt, I.G.M., Zwijnenburg, A., Kemperman, A.J.B., van der Meer, W.G.J., Wessling, M., 2008. Hollow fiber dead-end ultratiltration: influence of ionic environment on filtration of alginates. J. Memb. Sci. 308 (1e2), 218e229. Vickers, J.C., Thompson, M.A., Kelkar, U.G., 1995. The use of membrane filtration in conjunction with coagulation processes for improved nom removal. Desalination 102 (1e3), 57e61. Xu, R., 2002. Particle Characterization: Light Scattering Methods. Kluwer Academic Publishers, New York. Yamamura, H., Kimura, K., Okajima, T., Tokumoto, H., Watanabe, Y., 2008. Affinity of functional groups for membrane surfaces: implications for physically irreversible fouling. Environ. Sci. Technol. 42 (14), 5310e5315. Ye, Y., Le Clech, P., Chen, V., Fane, A.G., 2005. Evolution of fouling during crossflow filtration of model EPS solutions. J. Memb. Sci. 264 (1e2), 190e199. Yuan, W., Zydney, A.L., 1999. Effects of solution environment on humic acid fouling during microfiltration. Desalination 122 (1), 63e76. Zator, M., Ferrando, M., Lopez, F., Guell, C., 2009. Microfiltration of protein/dextran/polyphenol solutions: characterization of fouling and chemical cleaning efficiency using confocal microscopy. J. Memb. Sci. 344 (1e2), 82e91. Zhao, Y.Q., 2004. Settling behaviour of polymer flocculated watertreatment sludge II: effects of floc structure and floc packing. Sep. Purif. Technol. 35 (3), 175e183.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 e3 7 4
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Formation and removal of genotoxic activity during UV/H2O2eGAC treatment of drinking water M.B. Heringa a,*, D.J.H. Harmsen a, E.F. Beerendonk a, A.A. Reus b, C.A.M. Krul b, D.H. Metz c, G.F. IJpelaar a,1 a
KWR Watercycle research institute, P.O. Box 1072, 3430 BB Nieuwegein, The Netherlands TNO Quality of Life, P.O. Box 360, 3700 AJ Zeist, The Netherlands c Greater Cincinnati Water Works, 5651 Kellogg Avenue, Cincinnati, OH 45228, USA b
article info
abstract
Article history:
The objective of this study was to determine the genotoxic activity of water after UV/H2O2
Received 19 February 2010
oxidation and GAC filtration. Pre-treated surface water from three locations was treated
Received in revised form
with UV/H2O2 with medium pressure (MP) lamps and passed through granulated activated
6 August 2010
carbon (GAC). Samples taken before and after each treatment step were extracted and
Accepted 7 August 2010
concentrated by solid phase extraction (SPE) and analyzed for genotoxicity using the Comet
Available online 14 August 2010
assay with HepG2 cells and the Ames II assay. The Comet assay showed no genotoxic response in any of the samples. In the Ames II, no
Keywords:
genotoxic response was obtained with the TAMix (a mix of six strains), but the TA98 strain
AOP
showed an increase in genotoxic activity after MP-UV/H2O2 for all three locations. GAC post
Genotoxicity
treatment effectively reduced the activities to control levels at two of the three locations
Oxidation
and to below the level of the pre-treated water at one site. The results indicate that UV/
UV
H2O2 treatment may lead to the formation of genotoxic by-products, which can be removed
Drinking water
by subsequent GAC filtration. ª 2010 Elsevier Ltd. All rights reserved.
Photolysis
1.
Introduction
With increasing populations and limited availability of groundwater, an increase in the use of surface waters for the preparation of drinking water may be expected. These surface waters carry a large variety of micropollutants (e.g. pesticides, pharmaceuticals and organic solvents), for which the traditional treatment technologies used in direct treatment, i.e. coagulation, rapid sand filtration and granular activated carbon (GAC) filtration, are not a robust barrier (e.g. Kruithof and Schippers, 1994). Especially the more polar emerging substances detected in sources of drinking water (Loos et al., 2009) require more rigorous treatment technologies for removal during drinking water treatment.
During the last decades, many studies have been performed on the applicability of Advanced Oxidation Processes (AOP) for the degradation of contaminants in pre-treated natural water (e.g. Beltra´n et al., 1996). Although a limited number of AOP installations are in operation for drinking water production, UV/H2O2 treatment followed by granular activated carbon (GAC) filtration has proven to be effective in the removal of organic compounds with various chemical characteristics (Kruithof et al., 2007). Typical UV doses applied are in the order of 500e700 mJ/cm2; H2O2 concentrations are typically 5e10 mg/L. In this process, the two mechanisms responsible for contaminant destruction are direct photolysis and oxidation by the in-situ produced hydroxyl radicals (OH). The high oxidation power combined with the aselective
* Corresponding author. Tel.: þ31 30 6069539; fax: þ31 30 6061165. E-mail address:
[email protected] (M.B. Heringa). 1 Present address: Royal Haskoning, P.O. Box 151, 6500 AD Nijmegen, The Netherlands. 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.008
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 e3 7 4
character turns the hydroxyl radical into a highly effective oxidant. It is known and expected that water treatment based on degradation processes may lead to the formation of by-products, e.g. trihalomethanes (THMs) during chlorination and bromate during ozonation of (pre-treated) natural water (e.g. Rook, 1974; Richardson et al., 2007; Von Gunten and Hoigne, 1994). Also UV/H2O2 treatment of water may induce the formation of by-products. It appears that practical UV/ H2O2 process conditions do not fully mineralize contaminants to water and carbon dioxide. Indeed, formation of organic intermediates has been reported (e.g. Lau et al., 2005). Byproducts may result from the direct photolysis or from oxidation of compounds in the water matrix. Known byproducts are nitrite (photolysis of nitrate) and assimilable organic carbon (AOC; photolysis and oxidation of dissolved organic carbon (DOC)). Most organic contaminants strongly absorb light in the UVC range (200e285 nm) of the electromagnetic spectrum and this absorbed energy may lead to changes in the molecular structure of the compounds, resulting in (by)products. Also the omni-present natural compounds in surface water, collectively grouped as natural organic matter (NOM), absorb in the UVC wavelength range and can therefore be degraded into various (by-)products. As the identity of the by-products of UV-oxidation processes is largely unknown, the formation of toxic compounds during UV/H2O2 treatment of natural water should be considered. So far, only a few studies have been conducted on the toxicity of water after UV/H2O2 treatment, including studies on estrogenicity and acute toxicity (e.g. Linden et al., 2004). The formation of genotoxic (i.e. DNA-damaging) by-products by the oxidative reactions of ozone and chlorine is a reason to study the induction of genotoxic activity by AOPs such as UV/ H2O2 treatment. However, although it has been shown that no bromate or THMs are formed (Kruithof et al., 2007; Kashinkunti et al., 2004), no effect-directed genotoxicity studies (detecting any possible genotoxin) have been reported for UV/H2O2 treatment. Quite a few studies have been conducted on the effects of UV-disinfection (without H2O2) on the formation of genotoxicity. Conflicting results have been reported, with some finding an increase in genotoxicity after UV-disinfection and others that do not. These differences might be attributed to the use of different water qualities, applied UV-lamps (medium pressure (MP) vs. low pressure (LP)), UV dose and genotoxicity tests (e.g. Helma et al., 1994; Carnimeo et al., 1995; Haider et al., 2001, 2002). The present study therefore had the following objective: to study the genotoxic activity of surface water before and after treatment with UV/H2O2 AOP and after subsequent GAC. To our knowledge, this is the first submitted study on the formation of genotoxic by-products during UV/H2O2 AOP. Several assays are available for evaluating the genotoxic potential of water extracts. To detect gene mutations, we chose to use the Ames II assay (Gee et al., 1998; Fluckiger-Isler et al., 2004). This is a modified version of the well-known classic Ames test, which demands less sample volume. As complementary assay, detecting chromosomal damage, we chose the Comet assay in HepG2 liver cells. The Comet assay is a sensitive test that can be performed with any cell type and
367
allows rapid detection of chromosomal damage such as single and double DNA strand breaks (Tice et al., 2000). The human HepG2 liver cell line has the advantage of having endogenous metabolic capacity and liver cells are one of the first cell types chemicals encounter after intestinal absorption.
2.
Materials and methods
Three studies were performed: one in October 2007 with pretreated Meuse water from Bergambacht (The Netherlands) in a pilot reactor, one in September 2008 with pre-treated Ohio river water, directly upstream of the Cincinnati metropolitan area (OH, USA), in a pilot reactor, and one in February 2009 with samples taken from the full scale plant of PWN at Andijk (the Netherlands), which treats IJssel Lake water. Experimental details (e.g. materials) can be found in the Supplementary Information.
2.1.
Water treatment and sampling
Fig. 1 shows the general scheme of the three treatment setups and shows at which points samples were taken. Table 1 gives the most important details of the different treatment steps. Further details can be found in the Supplementary Information. Table 2 shows the water quality parameters of the sand filtrate prior to the oxidation step. To all samples of the Meuse and IJssel Lake study, 300 mg Na2SO3/L was added to quench residual H2O2. To all samples of the Ohio River study, 500 mg Na2SO3/L was added, whereafter the samples were frozen and shipped to the Netherlands for analysis. At the full scale plant treating IJssel Lake water, duplicate samples were taken.
2.2.
Sample extraction and concentration
The detailed extraction procedure can be found in the Supplementary Information. In brief, within 24 h after collection or thawing, three replicates of 1 L of every sample were extracted by solid phase extraction (SPE) with 200 mg Oasis HLB cartridges (Waters Corporation, Milford, USA) at pH 2.3. In the studies with Ohio River water and IJssel Lake water, mineral water samples (Evian from glass bottles) were included as procedure controls. Elution was performed with 3 serial additions of 2.5 mL of 20% methanol in acetonitrile. The 7.5 mL eluates were evaporated and taken up in 50 mL of DMSO yielding 20,000-fold concentrated extracts. All extracts were stored at 18 C until analysis.
2.3.
Ames II tests
The Ames II test strains (TA98 and TAMix) and media were purchased from Xenometrix (Basel, Switserland). The test procedure provided by Xenometrix, also described by Fluckiger-Isler et al. (2004), was followed, with minor modifications as described in the Supplementary Information. In brief, the water extracts were diluted to 100 mL (1:1) with DMSO to obtain a sufficient amount of sample for all tests and the bacteria were finally exposed to a 200-fold concentration of the water samples in culture medium. Water extracts were
368
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 e3 7 4
Fig. 1 e Schematic representation of the treatment setup for the studies with Meuse water, Ohio River water, and IJssel Lake water. In the Meuse water study, samples were taken at three different flow velocities at site 2 and for the lowest flow velocity only at site 3. In the Ohio River water study, samples at site 2 were collected both after UV treatment with H2O2 and after UV treatment without H2O2. Samples at site 3 were taken only after UV treatment with H2O2.
tested in triplicate, as well as a triplicate negative control (DMSO only), a triplicate positive control for genotoxicity, and a triplicate positive control for cytotoxicity. Water extracts were tested in both strains (TA98 and TAmix), both with and without the liver enzyme extract S92. A custom cytotoxicity test was performed with subsamples of the exposure cultures in medium with histidine, to check for possible artifacts due to effects on cell survival and growth. Finally, the number of yellow wells per 48 wells of one sample were counted manually as a measure of genotoxicity. Ames test responses follow a binomial distribution (Piegorsch et al., 2000), therefore a sample was considered genotoxic if the response of the sample was different from the response of the negative control with a certainty of 99%, based on a binomial distribution (see Supplementary Information).
2.4.
Comet assay
First, a neutral red uptake assay was performed as described in Borenfreund and Puerner (1985) with minor modifications, to check for cytotoxicity on the HepG2 cells (obtained from dr. B. Knowles). The HepG2 cells were treated for 3 h with 0.25%, 0.5% and 1% of the water extracts in HBSS (v/v). 1% Triton X-100 (v/v) was used as positive control for cytotoxicity. Details can be found in the Supplementary Information. The Comet assay was performed as described by Singh et al. (1988), with minor modifications as fully described in the Supplementary Information. In brief, for the samples from the Meuse water experiment, HepG2 cells were treated for 3 h with HBSS medium containing aliquots of water extract at a concentration of 1% (v/v) in duplicate (exposure to a 200-fold concentration of the water samples). 25 mg/mL methyl methane sulfonate in DMSO was used as positive control for genotoxicity. The Comet assays with the Ohio and IJssel Lake water samples were performed both in presence of S9 (3 h exposure) and in absence of S9 (24 h exposure). The positive control was then 50 mg/mL benzo[a]pyrene. 2
The Ames II assay is performed both with and without S9 liver enzyme extract, in order to detect both direct genotoxic compounds, and indirect genotoxic compounds that need to be converted to a genotoxic metabolite by liver enzymes first.
DNA damage was evaluated by calculation of the mean % tail DNA for a total of 200 cells per water sample (50 cells per slide, two slides per culture and two cultures per water sample). The water extracts were considered positive for genotoxicity when a three-fold increase in tail intensity over the negative control was observed. In addition to the prior neutral red uptake assay, viability was also checked by registering the number of ghost cells, though excluding them from the genotoxicity analysis. The relative proportion of ghost cells had to be less than 30%.
3.
Results and discussion
3.1.
Comet assay
In the neutral red uptake assay, water sample concentrations of up to 1% did not show any cytotoxicity in HepG2 cells. Therefore, water extracts were tested at a concentration of 1% (v/v) in the Comet assay, with certainty that a potential genotoxic response could not be induced by cytotoxicity. The results of the Comet assay are presented in Fig. 2. It shows that all water samples (except for the positive control and one mineral water sample) induced responses comparable to the level of the negative control. Thus, the data of the Comet assay do not show an increase in genotoxicity of the water after MP-UV/H2O2 treatment. It is unclear what caused the positive response in one of the procedure controls, as it was not observed in the duplicate sample. The test system should be sufficiently sensitive to detect a genotoxic potential of water extracts as a genotoxic response has been reported previously in the Comet assay with HepG2 cells with samples of chlorinated drinking water (Buschini et al., 2004; Yuan et al., 2005). However, the compounds involved here can be very different from the compounds formed during UV-oxidation. No data are available in published literature on induction of chromosomal damage by water samples after UV/H2O2 treatment, for comparison with the results of the present study. After UV-disinfection, however, Helma et al. (1994) found an increased response in contaminated groundwater with the micronucleus test with the Tradescantia plant (lamp type and
369
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 e3 7 4
Table 1 e Water treatment conditions. Study
UV/H2O2-treatment
Pretreatment
Meuse
Coagulation ((Fe)2(SO)3) Sedimentation
GAC treatment a
Micro sieves Rapid sand filtration
KWR-design pilot reactor (not optimized) 4 MP lamps: HOK 20/100 2 kW (Philips Lighting; Roosendaal, the Netherlands) 550 mJ/cm2 (chemidosimetry) 10 mg/L H2O2
Virgin Chemviron F400 GAC (Chemviron Carbon; Feluy, Belgium) Column 40 9 cm EBCTb 30 min Flow 5 L/h
Ohio river
Coagulation (Al2(SO4)3 and polyDADMACc) Sedimentation pH correction (CaO) Rapid sand filtration
Aquionics pilot reactor (optimized) 1 MP lamp: 3.5 kW Super TOC (Aquionics; Erlanger, KY, USA) 400 mJ/cm2 (instrument read-out) 10 mg/L H2O2
Reactivated Calgon F400 (Calgon Carbon; Pittsburgh, PA, USA), 325 d serviceColumn 1.73 0.1 m EBCTb 15 min Flow 57 L/h
IJssel Lake
Reservoir (2e6 d residence) Coagulation (FeClSO4) Sedimentation Rapid sand filtration
One of three streets with three of four full scale reactors running Trojan Swift 16L30 reactors with MP lamps (Trojan; London, Ontario, Canada) 0.54 kWh/m3 (e560 mJ/cm2; chemidosimetry) 6 mg/L H2O2
Reactivated Norit ROW 0.8 SUPRA (Norit; Amersfoort, the Netherlands), 2 y service Full scale GAC contactors EBCTb Flow 3000 m3/h
a The KWR reactor was used for comparative research and was therefore not optimally configured for any specific lamp. b EBCT ¼ empty bed contact time. c polyDADMAC ¼ cationic polymer (poly-diallyldimethylammonium chloride).
dose not given) and also after UV treatment of pre-purified contaminated groundwater with an LP lamp (50e150 mJ/cm2). In contrast, Haider et al. (2002) did not find an increase in micronucleus formation with either Tradescantia plants or rat liver cells after treatment of Austrian groundwater with lowpressure lamps (80 mJ/cm2).
3.2.
Ames II tests
In the cytotoxicity tests of the Ames II test, clear effects were obtained with the positive controls. The water samples and other controls showed no significant cytotoxicity for the applied bacterial strains, except for a slight cytotoxicity in a UV-treated sample of the IJssel Lake study (indicated in Fig. 3). This means that absence of genotoxic response in the Ames II test could not have been due to any bacterial death from e.g. cytotoxic compounds. Fig. 3 shows the results of the Ames II test of samples prior to and after UV/H2O2 treatment, for the Meuse river study (Fig. 3A), as well as the Ohio river study (Fig. 3B) and the IJssel Lake study (Fig. 3C). It can be seen that all negative (DMSO) controls present merely the normal, spontaneous mutations as background. With the positive controls, clear increases in
Table 2 e Water quality parameters of pre-treated river Meuse, Ohio and IJssel Lake water (i.e. after coagulation and rapid sand filtration). pH Alkalinity TOCa UV-T254 Nitrate 1 (mg HCO (mg NO 3 /L) 3 /L) (mg C/L) (%, cm ) Meuse Ohio IJssel Lake
5.8 3.1 4.2
7.06 7.7 7.50
151 93.3 130
3.9 1.95 2.2
a Measured as NPOC: Non-Purgeable Organic Carbon.
78 90.6 90.2
mutations were seen, deviating significantly from the background. Therefore, it is concluded that the bacterial strains were functioning normally during the tests and the method was performed correctly. In the TAMix cultures, no significant increase in mutations was detected in any of the water samples, both with and without metabolic activation (S9). This means there was no increase in the type of mutations detected by this strain, compared to the pre-treated water. No other data for Ames tests on water after UV/H2O2 treatment are available for comparison. However, these findings are in agreement with the results of the UV-disinfection research of Haider et al. (2002). They used LP lamps for UV-disinfection and found no response with TA100 (comparable strain to TAMix) in UVdisinfected groundwater samples (80 mJ/cm2). Guzzella et al. (2002) also found no increase in TA100 response after O3/UV and O3/UV/H2O2 treatment (UV dose 40 Vs/cm3). In contrast, significant increases in mutations were measured in the TA98 cultures. The pre-treated water of all three locations showed low genotoxicity either in presence of S9 (Meuse), or in absence of S9 (Meuse, Ohio and IJssel Lake). This has been seen before in the past (e.g. Veenendaal and van Genderen, 1999; Kool and van Kreyl, 1988) and might have a natural origin (Kool and van Kreyl, 1988). UV treatment with MP lamps, both with and without H2O2, resulted in a clear increase in the number of mutations in TA98 for all three water sources. This increase in genotoxicity was stronger in the absence of S9 in all cases, but also in the presence of S9, the increase still is significant for most cases. Interestingly, MP-UV treatment without H2O2 resulted in a higher response than with H2O2 in the one study where this was compared. Subsequent treatment of the UV/H2O2-treated water with GAC filtration removed the genotoxicity to the level of the negative control and of mineral water in two of the three studies (see Fig. 3). This observation is similar to that of
370
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 e3 7 4
Fig. 2 e Results of the Comet assay with water extracts of the Meuse river study (A), Ohio River study (B) and IJssel Lake study (C) for the assay was performed both in presence of S9 (striped bars) and in absence of S9(non-striped bars). Samples tested were a negative control (NC), a positive control (PC), a procedure control (PrC), and extracts of pre-treated water and water after UV treatment alone (UV), UV/H2O2 (UV ox) and after subsequent GAC filtration (UV ox GAC). Bars denote average values, error bars denote standard deviations (n [ 200). Asterisks denote responses determined to show genotoxicity, i.e. deviating from the NC.
Guzzella et al. (2002). Only in the UV/H2O2eGAC-treated Ohio river water a slight genotoxic response, just above the very low significance limit, was observed in TA98eS9. This genotoxic response was lower than that of the initial, SF-treated Ohio river water.
3.3. Difference between strains and responsible compounds The striking difference between the results found with the two bacterial cultures (TA98 and TAMix) can be explained by the differences between the type of mutations detected by these (mixes of) strains. The TAMix is a combination of six strains, detecting six different base-pair substitutions (Gee et al., 1994). The TA98 strain is a single strain, which detects frameshift mutations caused by deletions or additions of base-pairs in the DNA. Although the TAMix culture in theory has 1/6th of the sensitivity of that of the TA98 culture, this cannot explain the absence of TAMix response for samples where the TA98 response is 33 positive wells, i.e. 8 times the detection limit (Fig. 3B). It is more probable that the compounds in the samples simply do not cause base-pair substitutions, but only frameshift mutations. The Comet results are complementary to the results of the Ames II in this respect, as it is known that some base-pair substitutions may lead to chromosomal breaks under the alkaline conditions applied in this Comet assay. This strengthens the presumption that the formed compounds do not cause base-pair substitutions.
3.4. Effect of setup, water type, hydrogen peroxide and sulfite UV/H2O2 advanced oxidation was applied to pre-treated surface water without additional treatment or additions of any kind. Comparisons between the three setups (KWR designed and built pilot reactor, commercially available pilot reactor and a full scale reactor) are difficult to make, as the analyses were performed in different batches. Different batches give some variation in the test response (12e22% in TA98) due to the biological variation of the bacterial cultures. Considering this variation, the genotoxic activity after UV/ H2O2 treatment in the optimized pilot reactor (Ohio river) and a full scale reactor (IJssel Lake) was roughly at the same level, indicating that the scale of the tested system may not matter. The genotoxic activity after treatment in the KWR pilot reactor (Meuse river) was higher, which may be caused by a nonoptimal dose distribution, resulting in a less effective oxidation of (by)products. The increase in genotoxic response after UV/H2O2 treatment is observed in three different waters whose systems are hydrologically not connected. This indicates that the genotoxic compound(s) is/are formed from ubiquitous water components, such as natural organic matter (NOM) or ubiquitous contaminants. It remains to be investigated what these responsible components are, but in the mean time all locations applying UV/ H2O2 treatment should be on the lookout for this effect. The extent to which any observed increase in genotoxic response after UV/H2O2 treatment could be caused by remains
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 e3 7 4
371
Fig. 3 e Results of Ames II tests with water extracts of the Meuse river study (A), the Ohio river study (B) and the IJssel Lake study (C). Results are given for strains TA98 (grey bars) and TAMix (white bars), with (striped bars) and without (non-striped bars) S9. Samples tested were a negative control (NC), positive controls (PC), a procedure control (PrC), and extracts of pretreated water and water after UV treatment alone (UV), UV/H2O2 (UV ox) and after subsequent GAC filtration (UV ox GAC). Bars denote average values, error bars denote standard deviations (n [ 3). Asterisks denote responses determined to show genotoxicity, i.e. deviating from the NC with 99% certainty. Double asterisks denote responses showing an increase in genotoxicity by the treatment, i.e. deviating significantly (99%) from the response of the pre-treated water. The “ct” superscript indicates that the sample was slightly cytotoxic.
of H2O2 or sulfite was verified. Based on the results of earlier, unpublished research it was shown that the excess of sulfite ion as applied in these studies effectively reduces the H2O2 concentration to below the detection limit of 0.06 mg/L. Chemical analysis of the Meuse water after UV/H2O2 treatment in this study confirmed concentrations <0.06 mg/L. Furthermore, according to Aeschbacher et al. (1989), H2O2 appeared not genotoxic in the classic Ames test with TA98 S9 up to 150 mmol per plate (e1650 mg/L, assuming a distribution volume of 3 mL). As control tests with chemical analysis of sulfite levels in the Meuse water extracts showed that the extraction method applied does not extract sulfite ion, the extracts of the water samples will not have contained detectable levels of sulfite. Therefore, any observed increase of the genotoxic response cannot be the result of sulfite or H2O2 remains in the treated water. This is confirmed by the lack of substantial increase of the genotoxic response in pretreated Meuse water samples to which sulfite ion alone or sulfite ion and H2O2 were added (see Fig. 4). In the Ohio River samples, 500 mg/L sulfite was added instead of 300 mg/L, but as sulfite appears not to be extracted, these higher levels are also not expected to have reached the ultimate extracts.
3.5.
Effect of UV-photolysis and UV/H2O2 oxidation
The increase of the genotoxic response in the MP-UV/H2O2treated waters is clearly due to formation of new compounds
during the treatment, as the genotoxic response was absent or lower than before the treatment. To further investigate how these new compounds may have been formed, Ohio River water was treated with both photolysis by UV (without H2O2) and oxidation by UV in combination with H2O2. As can be seen in Fig. 3B, the genotoxic responses after UV were higher in absence of H2O2. This indicates that the formation of genotoxic compounds is (mainly) due to absorption of wavelengthdependent emissions, i.e. direct photolysis. If the formation of genotoxic compounds is indeed due to photolysis, it may be possible that LP lamps show much lower induction of genotoxic activity. In contrast to MP lamps with emissions covering the 200e400 nm UV region, LP lamps only emit at 253.7 nm in the UV region. In the 200e400 nm UV region both organic matter and organic micropollutants absorb UV light. This may lead to photolysis of compounds absorbing UV light at other wavelengths than 253.7 nm which may result in by-products of MP-UV/H2O2 not formed during the LP-UV/H2O2 process. Indeed, preliminary experiments treating Meuse and Ohio River water with LP-UV/H2O2 showed no or a much lower increase in genotoxic response. This remains to be further investigated, applying a method setting an equal, comparable dose for both MP and LP lamps. Another question rises from these results: can an increased genotoxic response be expected during UV-disinfection with MP lamps? If direct photolysis does contribute to genotoxicity, there probably is a relation between the
372
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 e3 7 4
Fig. 4 e Results of Ames II tests with TA98 (grey bars) and TAMix (white bars) with (striped bars) and without (empty bars) S9 on extracts of the pre-treated Meuse water with and without additions of H2O2 (10 mg/L) and Na2SO3 (300 mg/L). Bars denote average values, error bars denote standard deviations (n [ 3). Asterisks denote responses determined to show genotoxicity, i.e. deviating from the NC with 99% certainty.
Fig. 5 e Results of Ames II tests with TA98 (grey bars) and TAMix (white bars) with (striped bars) and without (empty bars) S9 on extracts of Meuse water after UV treatment at different flow velocities, resulting in UV doses of 547, 313 and 180 mJ/cm2, respectively. Bars denote average values, error bars denote standard deviations (n [ 3). Asterisks denote responses determined to show genotoxicity, i.e. deviating from the NC with 99% certainty.
induction of genotoxicity and the applied UV dose. The UV dose during UV-disinfection is typically 10e15 times lower than the UV dose applied in this research. Thus, the induction of genotoxicity would be expected to be much lower or even undetectable during UV-disinfection of water. In fact, the results of earlier research have shown that, although not qualified as significantly genotoxic, there was a slight increase in mutations in TA98 without S9 after MP-UV-disinfection of pre-treated surface water at a biocide UV dose of about 90 mJ/ cm2 in a 300 L/h laboratory-scale UV-disinfection apparatus. Under the same conditions, with the same water quality, an increase in mutations in TA98 without S9 was also found treating the water in a 180 m3/h MP-UV pilot, indicating that the scale of the laboratory research did not give false positive results. Similarly, Haider et al. (2002) found one weak increased response in TA98 without S9 in UV-disinfected groundwater using an LP lamp. Carnimeo et al. (1995) reported negative Ames test results for UV/H2O2-disinfected raw river water using LP lamps. However, when different water flows were applied in the Meuse water tests (resulting in different UV doses), the responses did not show an effect of UV dose on the height of the genotoxic response (Fig. 5) We have no solid explanation for this observation, given the discussion above. A possibility is that there was some kind of saturation in the formation process of the genotoxic compounds. Further studies are necessary to confirm our observation that photolysis seems to be responsible for the observed induction of genotoxicity.
genotoxicity and what doses of these compounds a consumer could tolerate. Additionally, further research is required to determine whether these mutagenic effects also occur in mammalian cells or in mammalian organisms. Most importantly, the increased genotoxic response as observed in treated water was removed to below the detection limit by GAC filtration in all but one study, even with GAC which had been in use for some time. As the present water treatment plants around the world applying UV-oxidation usually also apply subsequent GAC filtration, genotoxic compounds are not expected to be present in the final drinking water. Therefore, we expect no health risk from the final drinking water. But clearly, also new installations should for the mean time include GAC filtration after UV/H2O2 treatment. As the adsorption capacity of GAC decreases in time, and as we have seen one GAC-filtrate exceeds the detection limit, it is important to monitor the effectiveness of GAC filters during their lifetime in removing compounds that contribute to a genotoxic response. In the design of new installations and choice for treatment methods, the UV/H2O2 technique should be compared to other techniques. It is well-known that chlorination, chloramination and ozone-oxidation lead to genotoxic by-products, too (e.g. Rook, 1974; Richardson et al., 2007; Von Gunten and Hoigne, 1994). It is unclear if the induction of the genotoxic response observed in these methods are comparable, as they have been measured in different tests (e.g. classic Ames test vs. Ames II). This remains to be investigated. Furthermore, other factors, such as micropollutant removal, disinfection power and costs of the different methods are also important to consider in a treatment method comparison.
3.6.
Human health
A final question is what these results actually mean in terms of risks for human health. It must be stressed that the applied genotoxicity tests can only indicate a health hazard: namely the presence of genotoxic compounds in that water. It is impossible at this stage to derive a health risk, as it is unknown what compounds are responsible for the measured
4.
Conclusions
The results of these studies show no genotoxic activity after UV/H2O2 treatment in the Comet assay and in the Ames II
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 e3 7 4
TAMix strain with and without S9 under all applied conditions. An increase in genotoxic activity in the Ames II TA98 strain both with and without S9 was measured in three tested waters after MP-UV/H2O2 treatment. The increase in genotoxic activity was also seen after MP-UV treatment without H2O2, to a higher level than with H2O2, indicating that photolysis may be the responsible process for the formation of genotoxic compounds. GAC post treatment effectively reduced this genotoxic activity to control levels for two of the three studies and to below the level of the pre-treated water in one study; no health risks are expected as long as UV/H2O2 is followed by GAC filtration. Further research should primarily include other natural water qualities (e.g. with and without chemical pollutants; low or high DOC), chemical identification of the responsible products, UV doseegenotoxic effect relations, a comparison between LP and MP lamps and comparisons to other treatment methods.
Acknowledgements We thank Paul Baggelaar for the advice on Ames II statistics; John van Genderen for advice on the toxicology; Dunea (formerly known as DZH) for their cooperation with the Meuse water experiments; Maria Meyer of GCWW for the experimental work in Cincinnati; PWN for commissioning the analysis of the treated IJssel Lake water; Rene van Doorn and Hans van Beveren of KWR for the preparation of extracts; Stefan Voost, Ton Braat and Marijan Uytewaal-Aarts of KWR for performing the Ames II tests; and Philips for providing the UV-lamps of the pilot reactor studies. These studies were supported by the Joint Research Programme of the Dutch water utilities (BTO), by a commission of PWN, and by subsidiary from the Dutch Ministry of Economic Affairs.
Appendix. Supplementary data Supplementary data associated with the article can be found in online version, at doi:10.1016/j.watres.2010.08.008.
references
Aeschbacher, H.U., Wolleb, U., Loliger, J., Spadone, J.C., Liardon, R., 1989. Contribution of coffee aroma constituents to the mutagenicity of coffee. Food and Chemical Toxicology 27 (4), 227e232. Beltra´n, F.J., Ovejero, G., Rivas, J., 1996. Oxidation of polynuclear aromatic hydrocarbons in water. 3. UV radiation combined with hydrogen peroxide. Industrial & Engineering Chemical Research 35, 883e890. Borenfreund, E., Puerner, J.A., 1985. A simple quantitative procedure using monolayer cultures for cytotoxicity assays. Journal of Tissue Culture Methods 9, 7e9. Buschini, A., Carboni, P., Frigerio, S., Furlini, M., Marabini, L., Monarca, S., Poli, P., Radice, S., Rossi, C., 2004. Genotoxicity
373
and cytotoxicity assessment in lake drinking water produced in a treatment plant. Mutagenesis 19, 341e347. Carnimeo, D., Di Marino, R., Donaldo, F., Liberti, L., Ranieri, E., Pitzurra, M., Savino, A., 1995. Comparison between H2O2/UV and ClO2 disinfection of drinking water. Water Supply 13, 159e169. Fluckiger-Isler, S., Baumeister, M., Braun, K., Gervais, V., HaslerNguyen, N., Reimann, R., Van Gompel, J., Wunderlich, H.G., Engelhardt, G., 2004. Assessment of the performance of the Ames II assay: a collaborative study with 19 coded compounds. Mutation Research 558 (1e2), 181e197. Gee, P., Maron, D.M., Ames, B.N., 1994. Detection and classification of mutagens: a set of base-specific Salmonella tester strains. Proceedings of the National Academy of Sciences of United States of America 91 (24), 11606e11610. Gee, P., Sommers, C.H., Melick, A.S., Gidrol, X.M., Todd, M.D., Burris, R.B., Nelson, M.E., Klemm, R.C., Zeiger, E., 1998. Comparison of responses of base-specific Salmonella tester strains with the traditional strains for identifying mutagens: the results of a validation study. Mutation Research 412 (2), 115e130. Guzzella, L., Feretti, D., Monarca, S., 2002. Advanced oxidation and adsorption technologies for organic micropollutant removal from lake water used as drinking-water supply. Water Research 36 (17), 4307e4318. Haider, T., Sommer, R., Knasmu¨ller, S., Eckl, P., Pribil, W., Cabaj, A., Kundi, M., 2001. Application of a Test Combination of Three Different Bioassays for the Investigation of Genotoxic Effects Induced by UV-irradiation in Water. First International Congress on Ultraviolet Technologies. International Ultraviolet Association (IUVA), Washington, D.C. Haider, T., Sommer, R., Knasmuller, S., Eckl, P., Pribil, W., Cabaj, A., Kundi, M., 2002. Genotoxic response of Austrian groundwater samples treated under standardized UV (254 nm)edisinfection conditions in a combination of three different bioassays. Water Research 36 (1), 25e32. Helma, C., Sommer, R., Schulte-Hermann, R., Knasmuller, S., 1994. Enhanced clastogenicity of contaminated groundwater following UV irradiation detected by the Tradescantia micronucleus assay. Mutation Research 323 (3), 93e98. Kashinkunti, R.D., Linden, K.G., Shin, G.A., Metz, D.H., Sobsey, M. D., Moran, M.C., Samuelson, A.M., 2004. Investigating multibarrier inactivation for Cincinnati. Journal of the American Water Works Association (AWWA) 96 (6), 114e127. Kool, H.J., van Kreyl, C.F., 1988. Mutagenic activity in drinking water prepared from groundwater: a survey of ten cities in The Netherlands. Science of the Total Environment 77 (1), 51e60. Kruithof, J.C., Schippers, J.C., 1994. Drinking water treatment in the 1990s. Water Supply 12 (1e2), SS5.1eSS5.4. Kruithof, J.C., Kamp, P.C., Martijn, B.J., 2007. UV/H2O2 treatment: a practical solution for organic contaminant control and primary disinfection. Ozone: Science and Engineering 29 (4), 273e280. Lau, T.K., Chu, W., Graham, N., 2005. The degradation of endocrine disruptor di-n-butyl phthalate by UV irradiation: a photolysis and product study. Chemosphere 60, 1045e1053. Linden, K.G., Rosenfeldt, E.J., Chen, P.J., Kullman, S.W., 2004. UV and UV/H2O2 Degradation and Subsequent Toxicity of Endocrine Disrupting Chemicals in Water. European Conference on Ultraviolet Radiation. International Ultraviolet Association, Karlsruhe (Germany). Loos, R., Gawlik, B.M., Locoro, G., Rimaviciute, E., Contini, S., Bidoglio, G., 2009. EU-wide survey of polar organic persistent pollutants in European river waters. Environmental Pollution 157 (2), 561e568. Piegorsch, W.W., Simmons, S.J., Margolin, B.H., Zeiger, E., Gidrol, X.M., Gee, P., 2000. Statistical modeling and analyses of a base-specific Salmonella mutagenicity assay. Mutation Research 467 (1), 11e19. Richardson, S.D., Plewa, M.J., Wagner, E.D., Schoeny, R., Demarini, D.M., 2007. Occurrence, genotoxicity, and
374
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 e3 7 4
carcinogenicity of regulated and emerging disinfection byproducts in drinking water: a review and roadmap for research. Mutation Research 636 (1e3), 178e242. Rook, J.J., 1974. Formation of haloforms during chlorination of natural waters. Water Treatment Examination 23 (2), 234e243. Singh, N.P., McCoy, M.T., Tice, R.R., Schneider, E.L., 1988. A simple technique for quantitation of low levels of DNA damage in individual cells. Experimental Cell Research 175, 184e191. Tice, R.R., Agurell, E., Anderson, D., Burlinson, B., Hartmann, A., Kobayashi, H., Miyamae, Y., Rojas, E., Ryu, J.C., Sasaki, Y.F., 2000. Single cell gel/comet assay: guidelines for in vitro and in vivo genetic toxicology testing. Environmental and Molecular Mutagenesis 35 (3), 206e221.
Veenendaal, H.R., van Genderen, J., 1999. Mutageniteit in Rijn en Maas in 1998. Kiwa, Nieuwegein (the Netherlands), report KOA 99.088. von Gunten, U., Hoigne, J., 1994. Bromate formation during ozonation of bromide-containing waters: interaction of ozone and hydroxyl radical reactions. Environmental Science and Technology 28 (7), 1234e1242. Yuan, J., Wu, X.J., Lu, W.Q., Cheng, X.L., Chen, D., Li, X.Y., Liu, A.L., Wu, J.J., Xie, H., Stahl, T., Mersch-Sundermann, V., 2005. Chlorinated river and lake water extract caused oxidative damage, DNA migration and cytotoxicity in human cells. International Journal of Hygiene and Environmental Health 208, 481e488.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 5 e3 8 3
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Fate of hormones and pharmaceuticals during combined anaerobic treatment and nitrogen removal by partial nitritation-anammox in vacuum collected black water M.S. de Graaff a,b,*, N.M. Vieno a,1, K. Kujawa-Roeleveld b, G. Zeeman b, H. Temmink a,b, C.J.N. Buisman a,b a
Wetsus, Centre of Excellence for Sustainable Water Technology, KWR Watercycle Research Institute, Agora 1, P.O. Box 1113, 8900 CC Leeuwarden, The Netherlands b Wageningen University, Sub-department of Environmental Technology, P.O. Box 8129, 6700 EV Wageningen, The Netherlands
article info
abstract
Article history:
Vacuum collected black (toilet) water contains hormones and pharmaceuticals in relatively
Received 19 February 2010
high concentrations (mg/L to mg/L range) and separate specific treatment has the potential
Received in revised form
of minimizing their discharge to surface waters. In this study, the fate of estrogens (natural
10 August 2010
and synthetical hormones) and pharmaceuticals (paracetamol, metoprolol, propranolol,
Accepted 10 August 2010
cetirizine,
Available online 17 August 2010
ibuprofen and diclofenac) in the anaerobic treatment of vacuum collected black water
doxycycline,
tetracycline,
ciprofloxacin,
trimethoprim,
carbamazepine,
followed by nitrogen removal by partial nitritation-anammox was investigated. A new Keywords:
analytical method was developed to detect the presence of several compounds in the
Anaerobic treatment
complex matrix of concentrated black water. Detected concentrations in black water
Autotrophic nitrogen removal
ranged from 1.1 mg/L for carbamazepine to >1000 mg/L for paracetamol. Anaerobic treat-
Black water
ment was only suitable to remove the majority of paracetamol (>90%). Metoprolol was
Removal of micro-pollutants
partly removed (67%) during aerobic treatment. Deconjugation could have affected the
Separation at source
removal efficiency of ibuprofen as concentrations even increased during anaerobic treatment and only after the anammox treatment 77% of ibuprofen was removed. The presence of persistent micro-pollutants (diclofenac, carbamazepine and cetirizine), which are not susceptible for biodegradation, makes the application of advanced physical and chemical treatment unavoidable. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Separation of domestic wastewater at source results in concentrated black water from the toilet (faeces and urine) and less concentrated grey water from showers, laundry, etc. Black water contains half the load of organic material in domestic wastewater and the major fraction of the nutrients nitrogen
and phosphorus (82% and 68% respectively) (Otterpohl et al., 1999; Kujawa-Roeleveld and Zeeman, 2006). Furthermore, black water contains micro-pollutants such as hormones and pharmaceutical residues. Previous research showed that black water, collected with vacuum toilets, can be treated in a UASB (Upflow Anaerobic Sludge Blanket) reactor to produce biogas at a relatively short hydraulic retention time (HRT) of 8.7 days (de
* Corresponding author. KWR Watercycle Research Institute, P.O. Box 1072, 3430 BB Nieuwegein, The Netherlands. Tel.: þ31 (0)306069526. E-mail address:
[email protected] (M.S. de Graaff). 1 Currently working at Finland WANDER Nordic Water and Materials Institute, Kalliokatu 10 B, FI-26100 Rauma, Finland. 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.023
376
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 5 e3 8 3
Graaff et al., 2010a). A one or two reactor partial nitritationanammox process successfully removed the ammonium from the UASB effluent (Vlaeminck et al., 2009; de Graaff et al., 2010b; de Graaff et al., in press). In conventional wastewater treatment plants (WWTPs) micro-pollutants such as hormones and pharmaceuticals, present in ng/L to mg/L range, are not or only partially removed and present a threat to aquatic life (Schwarzenbach et al., 2006; Ternes and Joss, 2006). The risk those micropollutants may pose is difficult to assess and not well understood (Schwarzenbach et al., 2006; Schirmer and Schirmer, 2008), but several effects have been reported, such as feminization or masculinisation of fishes already at ng/L levels (Harries et al., 1997). In source separation based treatment concepts the micro-pollutants in urine or black water are present at higher concentrations (mg/L to mg/L range) (de Mes, 2007; Winker et al., 2008). Specific treatment of these concentrated streams may reduce their discharge to surface waters compared to WWTPs (Larsen et al., 2004, 2009). The biodegradation potential of a number of selected hormones and pharmaceuticals under different redox conditions has been investigated (e.g. Dytczak et al., 2008), but little is known about their removal and biodegradation during continuous biological treatment of concentrated streams (Kujawa-Roeleveld et al., 2008). Under anaerobic conditions and at long retention times (30 days) only a few compounds (acetylsalicilic acid, ibuprofen and fenofibrate) were removed. Under aerobic conditions the removal rate of those compounds was much higher (Kujawa-Roeleveld et al., 2008). The fate of estrogens during biological treatment of vacuum collected black water was studied in detail by de Mes (2007) and adsorption and biodegradation were the main removal processes. Furthermore it was shown that a significant amount (>70%) of the estrogens was present in conjugated form, showing a limited deconjugation during the treatment. Conjugated compounds are usually the more soluble and inactive forms of the parent compound formed in the human body. In faeces enzymes are present that can hydrolyse conjugates back into their original and therefore active form (Ternes et al., 1999). Additional treatment is necessary to reduce the emission of estrogens to surface waters, because the effluent concentrations of the biological treatment of vacuum collected black water were still in mg/L range (1.4 mg/L E1 and 0.65 mg/L E2, (de Mes, 2007)). In this study, the fate of selected estrogens and pharmaceuticals was investigated during the anaerobic treatment of vacuum collected black water followed by nitrogen removal by two reactor partial nitritation-anammox. Their removal efficiencies under different redox conditions and possibilities for physical-chemical post treatment in new sanitation systems based on separation at source are discussed.
2.
Materials and methods
2.1.
Selection of micro-pollutants
The following compounds were studied, based on their potential harmful effects in the environment, their
consumption (amount and frequency), removal in WWTPs and their therapeutic group: antiphlogistics (paracetamol, diclofenac, ibuprofen), antibiotics (tetracycline, doxycycline, ciprofloxacin, trimethoprim), beta-blockers (metoprolol and propranolol), antihistamine (cetirizine), antiepileptic (carbamazepine), natural hormones (estrone (E1) and 17b-estradiol (E2), birth control pill (17a-ethynylestradiol (EE2))) and conjugated natural estrogens (E2-17G and E2-3S) (Schrap et al., 2003; Clara et al., 2004; Griens and Tinke, 2006; Nakada et al., 2008; Kosonen and Kronberg, 2009). Furthermore, the antibiotics were selected because of the increasing concern about the formation of resistant pathogen strains (Mu¨ckter, 2006). Human estrogens (E1 and E2) and the synthetic estrogen EE2 are mainly responsible for the endocrine disrupting effects seen in the aquatic environment (Desbrow et al., 1998).
2.2.
Treatment of black water
Black water, collected in vacuum toilets, was obtained from the DESAR (Decentralized Sanitation and Reuse) demonstration site in Sneek (Friesland, the Netherlands) (Zeeman et al., 2007). Every two weeks jerry cans were filled with black water from the buffer tank at the demonstration site (hydraulic retention time of 4 h, not cooled), transported to the lab and stored at 4 C. The black water was first anaerobically treated in a UASB reactor. The effluent of the UASB reactor was subsequently treated in a two reactor partial nitritationanammox process to remove the nitrogen. The treatment concept is shown in Fig. 1. The UASB reactor (50 L, transparent Perspex) was operated at 25 C, an HRT of 8.7 day and an SRT of 254 days (de Graaff et al., 2010a). The partial nitritation reactor (6.1 L, glass) was operated at 25 C, an HRT of 1.7 days and an SRT fluctuating between 1 and 17 days (de Graaff et al., 2010b). The anammox SBR35 (5 L, glass) was operated at 35 C and an HRT of 2.0e10 days, depending on the applied hydraulic load (de Graaff et al., 2010c).
2.3.
Samples
Grab samples (15 mL, Greiner tubes of polypropylene) were taken every 3e4 weeks from the black water influent tap (tap 0, Fig. 1) to the UASB reactor (BW) (operation day 672e1070), from the UASB effluent, from the effluent of the partial nitritation reactor (PN effluent) (operation day 425e802) and from the effluent of the anammox SBR35 (AMX effluent) (operation day 96e386) and stored at 20 C. All samples were taken in the same time period (from August 2008 till August 2009), when the reactors were operating in steady state allowing saturation of the reactors’ material and tubing. Adsorption of the compounds to the Greiner sample tubes was tested in duplo and was found to be negligible (data not shown).
2.4. water
Analysis of hormones and pharmaceuticals in black
The analyses were performed according to the method described in the supplementary information. The method consisted of solid phase extraction (SPE), concentration and
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 5 e3 8 3
377
Fig. 1 e Treatment concept for black water.
quantification by liquid chromatography coupled to a tripleequadruple mass spectrometer (LC/MSeMS, Agilent Technologies 1200 and 6410 series). All chemicals were purchased from SigmaeAldrich. All the glassware used for analysis was cleaned and deactivated to prevent interactions of compounds with glass (described in detail in the supplementary information, S.1.1). Black water samples (10 mL) were filtered over a pre-washed Whatman GF6 filter; the effluents (UASB, PN and AMX) were not filtered. For most analytes the adsorption to the GF6 filter was negligible (Table S.4). The relative standard deviation (RSD) of the method was always less than 10%. Internal standards were used to correct responses of the analytes when the RSD was higher than 10%. The recoveries were determined for black water, effluent from the UASB reactor, effluent from the partial nitritation reactor and effluent from anammox SBR35 (BW, UASB, PN and AMX) (Table S.3). Most recoveries were more than 70%, but due to matrix effects some recoveries were lower than 70%, or in some cases higher than 100%. In all cases the obtained recoveries were consistent and no significant differences in recoveries between the four matrices were observed. The linear range between 0.5 mg/L and 1000 mg/L was determined based on the concentrations found in the matrix and a R2 > 0.99 (Lindqvist et al., 2005). For paracetamol, trimethoprim, metoprolol, propranolol, carbamazepine the upper limit of the linear calibration was 1000 mg/L. For tetracycline and ibuprofen the upper limit of the linear calibration was 500 mg/L. For the other analytes the upper limit of the linear calibration was 250 mg/L. The instrumental limit of quantification (LOQinstrumental) and the limit of quantification in the matrices (LOQmatrix) are shown in Table S.5 and range from 0.1 mg/L for most compounds to 10.5 mg/L for E2 in black water. For ciprofloxacin no linear range was found and therefore quantification of this compound was not possible. Because
ciprofloxacin also showed unrealistic recoveries (Table S.3), this compound was excluded from further data analysis.
3.
Results
3.1.
Black water
Table 1 shows the average and median concentrations of the detected compounds in the liquid fraction of black water. Most of the selected pharmaceuticals were detected in all the black water samples that were analyzed. The concentrations of paracetamol in black water were higher than 1000 mg/L in 10 out of 15 samples and because the recovery could not be calculated (see supplementary information), these results should be regarded as semi-quantitative. Tetracycline, propranolol and carbamazepine were only detected in a few black water samples. Metoprolol and ibuprofen were detected in relative high concentrations, average 45 mg/L and 147 mg/L. No significant differences between samples in time were observed, except for tetracycline and diclofenac for which a large difference between the average and median value was observed (Table 1). Estrogens were not detected in the black water and responses were under the LOQ, except for one sample where E1 was found in a concentration of 2.4 mg/L. In Table 1 the average concentrations measured in this study are compared to calculated predicted concentration (PC) based on the defined daily dose (DDD) and to the concentrations found in WWTP influents in the Netherlands. The concentrations of paracetamol, metoprolol and ibuprofen were in the same range as the PC assuming that one person is using that compound. Compared to the WWTP influents the concentrations found were much higher, which was also expected because the black water is about 25 times more concentrated than conventional domestic WWTP
378
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 5 e3 8 3
Table 1 e Concentrations of the detected compounds in black water compared with the calculated predicted concentration (PC) based on the defined daily dose (DDD) (WHO, 2006) and the excretion in unmetabolized form (KNMP, 2006) (all in mg/L). The concentrations found in WWTP influents in the Netherlands are given as well (Schrap et al., 2003). Compound
Paracetamol Trimethoprim Tetracycline Metoprolol Doxycycline Propranolol Carbamazepine Cetirizine Ibuprofen Diclofenac EE2
BW This research (15 samples) Average
Median
Min
Max
>1000 71 107 45 3.9 1.0 1.1 1.4 147 13
>1000 71 12 42 3.3 1.0 0.6 1.2 135 5.1
747 17 2.8 29 1.8 e 0.1 0.2 55 1.9 e
>1000 124 358 68 9.7 e 2.5 3.0 258 66 e
Excretion in unmetabolized form
PCa
WWTP influents
5% 44% 54% 5% 40% 96% 2% 100% 1% 25% 10%
408 478 1467 20 109 417 54 27 33 68 0.0068
0.33e45 0.083e0.51 <0.1 0.25e1.8 <0.1 0.1e0.51 0.5e9.5 e 1.5e17 1.2e13 0.0059a
a PC was calculated assuming that one person in Sneek is using, based on DDD. b (de Mes et al., 2005)
influent. However, the black water was obtained from a community of only 32 houses and therefore the concentrations found in this specific black water may not be representative. The concentrations of tetracycline and doxycycline were lower than the PC. This could be due to sorption to the solids because only the liquid fraction of black water was analyzed, and tetracycline and doxycycline are very hydrophobic (Thomas et al., 2007).
3.2. Removal of the compounds in the treatment concept for black water In Fig. 2 the average concentrations which were found in all the matrices are shown. In the anammox effluent paracetamol, tetracycline and doxycycline were not detected. The other compounds were still present in the effluent of the anammox reactor in the mg/L range, which shows that biological treatment alone is not enough to eliminate the selected compounds from black water. In Table S.6 (supplementary information) the average and median concentrations of the selected compounds are presented, including minimum and maximum values. In Table 2 the removal from the liquid phase for each compound in the treatment concept for black water is shown, together with literature information about the biodegradability under different redox conditions and the tendency to adsorb to sludge. The majority of paracetamol was removed during anaerobic treatment in the UASB reactor, though the removal efficiency could not be calculated because concentrations in black water were above the upper limit of the calibration curve. In the effluent from the anammox reactor no paracetamol was detected and all responses were below the detection limit (0.3 mg/L, Table S.5). For trimethoprim and tetracycline the removal was not calculated because the compounds were detected in only a few samples. However, as both compounds always were detected in samples taken in the same time period, the majority of trimethoprim and tetracycline was removed in the UASB reactor. Metoprolol was detected in almost all the samples, including samples of
the anammox effluent. During anaerobic treatment (both in UASB reactor and anammox reactor) no removal of metoprolol was observed. A removal of 67% of the metoprolol was observed during aerobic treatment in the partial nitritation reactor at an HRT of 1.7 days. Propranolol, carbamazepine and cetirizine were only found at very low concentrations, but were still detected in the anammox effluent (propranolol only once) and not removed in any step of the treatment sequence. For the estrogens only E1 was found in a number of UASB effluent samples in an average concentration of 0.6 mg/L, in the same range as detected by (de Mes, 2007). In the PN and anammox effluent no estrogens were detected. Because the estrogens were not detected in the BW samples, removal efficiencies could not be calculated and therefore the estrogens are not presented in Table 2. Ibuprofen was detected in all black water samples, but was not removed during anaerobic treatment in the UASB reactor and during aerobic treatment in the PN reactor. Only in the anammox reactor a significant removal of ibuprofen of 77% was observed. Diclofenac was still detected in the effluent from the anammox reactor and no significant removal was observed during the black water treatment.
4.
Discussion
4.1.
Analysis of micro-pollutants in black water
In this research a method was developed to analyze a selection of pharmaceutical and hormone residues in concentrated black water (supplementary information). For most compounds the analysis was successful, but problems were observed with the analysis of estrogens and ciprofloxacin. Especially the black water samples contained lots of matrix components that could have interfered with the analyte in the solid phase extraction, chromatographic separation or ionization stage. More discussion on the method can be found in the supplementary information. Methods for the analysis of the selected compounds in the black water sludge were not yet developed.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 5 e3 8 3
379
Fig. 2 e Average concentrations of the selected micro-pollutants in black water (BW), effluent from the UASB reactor (UASB), effluent from the partial nitritation reactor (PN) and effluent from the anammox reactor (AMX). A table is included which shows the number of samples that had a response above LOQ. PAR [ paracetamol, TRI [ trimethoprim, TC [ tetracycline, MET [ metoprolol, DOX [ doxycycline, PRO [ propranolol, CBZ [ carbamazepine, CET [ cetirizine, IBP [ ibuprofen and DCF [ diclofenac.
Analysis of the compounds in the sludge matrix is difficult because the complex matrix will cause significant ionization suppression. Recently, methods were developed to analyze beta-blockers in sludge and it was found that appropriate surrogate standards were needed to compensate the matrixinduced ion suppression (Scheurer et al., 2009). Other examples of methods for the analysis of micro-pollutants have been published previously by Ternes et al. (2002, 2005), which will be used as a basis to develop a method for black water sludge.
4.2. Removal of micro-pollutants during biological treatment This research showed the fate of selected micro-pollutants during anaerobic treatment combined with the two reactor partial nitritation-anammox process. Only paracetamol was removed anaerobically and was not detected in the anammox effluent. This corresponds with
previous studies where paracetamol was found in only a few WWTP effluents and never in surface waters showing that paracetamol is a readily biodegradable compound (e.g. Ternes, 1998; Yu et al., 2006). This research showed that anaerobic treatment alone is not enough to remove micro-pollutants, which is consistent with previous studies (e.g. Ericson, 2007; Kujawa-Roeleveld et al., 2008). Aerobic treatment was successful to remove a large fraction of metoprolol. Maurer et al. (2007) found that the removal of beta-blockers depended on the HRT and estimated that 90% of metoprolol could be removed at an HRT of 1 day and a sludge concentration of 4 gCOD/L (suspended solids) in WWTPs. This lower removal of metoprolol of 67% in this research could be due to a different sludge concentration and sludge characteristics, i.e. activated sludge versus nitrifying sludge mainly containing ammonium oxidizers (de Graaff et al., 2010b). Also the presence of readily degradable compounds can increase the removal of certain compounds, because it is likely that micro-pollutants are
380
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 5 e3 8 3
Table 2 e Removal efficiencies during anaerobic treatment (UASB), aerobic treatment (PN) and anoxic treatment (AMX); literature information about the biodegradability under different redox conditions and the tendency to adsorb to sludge. Compound
UASB
PN
AMX
Overall
Anaerobic/ Anoxic biodegradable?
Aerobic readily biodegradable?
Likely to adsorb to sludge?
Paracetamol Trimethoprim Tetracycline Metoprolol Doxycycline Propranolol Carbamazepine Cetirizine Ibuprofen Diclofenac
>90%a e e 7% 0% e e 20% 23% 22%
x e e 67% 69% e e 23% 16% x
x e e 23% e e e 52% 77% 75%
>90%a e e 56% 69% e e 29% 76% 37%
n.f. n.f. n.f. No/Noh n.f. n.f. Noh,j/Noh n.f. Yes, but limitedh,j/Yesh Yes, but limitedh,j/Noh
Yesb Nod,e Nod Noi Nod Noi Nol Nom Yesh,n Noo
Noc Nof Yesf,g Noi Yesg Noi Nok Nom Nok Nok,o
x: could not be calculated, because the compounds could not be quantified (see supplementary information). u.d.: under detection limit, therefore the removal could not be calculated. e: removal was not calculated because the compounds were detected in only a few samples. n.f.: not found. a estimated, as the concentrations in black water were above the upper limit of the calibration curve. b (Yu et al., 2006). c (Joss et al., 2006a). d (Alexy et al., 2004). e (Halling-Sørensen et al., 2000). f (Kim et al., 2005). g (Thomas et al., 2007). h (Kujawa-Roeleveld et al., 2008). i (Maurer et al., 2007). j (Carballa et al., 2007). k (Carballa et al., 2008). l (Clara et al., 2004). m (Kosonen and Kronberg, 2009). n (Buser et al., 1999). o (Buser et al., 1998).
degraded during co-metabolism (Alexander, 1981; Alexy et al., 2004; Joss et al., 2006a). Because in the UASB reactor most readily biodegradable organic material already is removed, cometabolism of heterotrophs in the aerated partial nitritation reactor may have been limited for the removal of metoprolol. Because several compounds have a relatively slow biodegradation rate, optimization of the biological treatment, for example by increasing the aerobic sludge retention time (SRT) could increase the total removal (Joss et al., 2006b; Jones et al., 2007). In this research the aerobic SRT was relatively short (SRT of 1.0e17 days) and may be increased to remove more ibuprofen and metoprolol. However, the increase in the PN reactor is limited to prevent nitrite oxidation (de Graaff et al., 2010b) and therefore additional biological post treatment might be necessary to remove the remaining biodegradable micro-pollutants. Furthermore, previous research showed the contribution of nitrification to the removal of steroidal compounds, for example estrogens (e.g. Yi and Harper, 2007). Persistent pharmaceuticals such as carbamazepine, cetirizine and diclofenac were still detected in the anammox effluent and were present in the mg/L range, showing their persistency during biological treatment (Buser et al., 1998; Clara et al., 2004; Kosonen and Kronberg, 2009).
4.3.
Risks
As mentioned in the introduction the risk micro-pollutants may pose is difficult to assess and not well understood
(Schwarzenbach et al., 2006; Schirmer and Schirmer, 2008), but several effects have been reported already at ng/L levels. More research is also needed on the combined effect of micropollutants present as mixtures in the environment (Schwarzenbach et al., 2006). To which extent the hormones and pharmaceuticals should be removed is therefore not known and requires further research. Another growing concern is the presence of conjugates and metabolites, because a large fraction of the pharmaceuticals is excreted as conjugates and metabolites and not as the parent compound (Table 1). In most studies these conjugates and metabolites are not analyzed and for example for estrogens it was shown that a large fraction (>70%) of the estrogens was present in the conjugated form after biological treatment of black water (de Mes et al., 2007). The concentrations of ibuprofen in the UASB effluent were even higher than in the black water and this could be due to deconjugation because most ibuprofen is excreted in conjugated or metabolized form (KNMP, 2006). Therefore more research is needed on the presence of conjugates and metabolites and their deconjugation and biodegradation rate. Because some compounds are likely to adsorb to sludge and are not readily biodegradable (e.g. tetracycline and doxycycline), the reuse of the anaerobic sludge in agriculture might be limited (Winker et al., 2009). The conditions during land application may change such that desorption becomes favourable. In this way tetracycline can form a potential risk to the environment (Kim et al., 2005).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 5 e3 8 3
It was found that pharmaceuticals such as propranolol and diclofenac can inhibit methanogenesis, but only in the mg/L range (Fountoulakis et al., 2004), which is much higher than found in this research. Also inhibition effects of antibiotics were found on nitrite oxidizers at concentrations of 10 mg/L (Dokianakis et al., 2004). Although concentrations in black water are higher than in sewage, they are still in a range where inhibition of biological treatment by pharmaceuticals is not expected. Only in specific circumstances where black water of a small community using a large amount of pharmaceuticals is treated, inhibition on biological processes might occur (Fountoulakis et al., 2008).
4.4.
Post treatment options
To eliminate the persistent micro-pollutants additional physical/chemical treatment such as advanced oxidation, activated carbon adsorption or membrane filtration is needed (Joss et al., 2006a). However, the anammox effluent still contains some organic material and humic acids that will interfere with such post treatment techniques. Furthermore the formation of by-products when e.g. ozonation is applied is of growing concern because they can be more toxic than their parent compounds. On the other hand, ozonation is also known to increase the biodegradability and a combination with biological post treatment might increase the total removal of the micro-pollutants (Li et al., 2006). Promising results were achieved with post-ozonation of WWTP effluent followed by sand filtration requiring an additional energy requirement of 12% (Hollender et al., 2009). Most of the persistent micro-pollutants were removed by ozonation and biodegradable compounds formed during ozonation were additionally removed in the sand filter.
5.
Conclusions
The methods developed in this research showed the successful detection of a wide selection of micro-pollutants in a complex matrix of vacuum collected black water and effluents of a sequence of biological reactors operated under different redox conditions. During combined anaerobic treatment and nitrogen removal by partial nitritation-anammox from vacuum collected black water only a few compounds were (partly) removed. Anaerobic treatment was only suitable to remove the majority of paracetamol. Metoprolol and ibuprofen were (partly) removed during aerobic treatment. The application of different redox conditions during biological treatment is not sufficient to fully eliminate hormones and pharmaceuticals in concentrated black water. To eliminate persistent micropollutants (e.g. diclofenac, carbamazepine and cetirizine) additional physical and chemical treatment will be unavoidable in new sanitation concepts.
Acknowledgements Janneke Tempel, Jelmer Dijkstra and Bob van Bijnen are thanked for their help in analyzing the samples on the LCMS.
381
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 theme “Separation at Source” for their interest and financial contribution.
Appendix. Supplementary information Supplementary information associated with this article can be found, in the online version, at doi:10.1016/j.watres.2010.08. 023.
references
Alexander, M., 1981. Biodegradation of chemicals of environmental concern. Science 211 (4478), 132e138 (9 January 1981). Alexy, R., Ku¨mpel, T., Ku¨mmerer, K., 2004. Assessment of degradation of 18 antibiotics in the closed bottle test. Chemosphere 57, 505e512. Buser, H.-R., Poiger, T., Muller, M.D., 1998. Occurrence and fate of the pharmaceutical drug diclofenac in surface waters: rapid photodegradation in a lake. Environmental Science and Technology 32 (22), 3449e3456. Buser, H.-R., Poiger, T., Muller, M.D., 1999. Occurrence and environmental behaviour of the chiral pharmaceutical drug ibuprofen in surface waters and in wastewater. Environmental Science and Technology 33, 2529e2535. Carballa, M., Fink, G., Omil, F., Lema, J.M., Ternes, T.A., 2008. Determination of the solid-water distribution coefficient (Kd) for pharmaceuticals, estrogens and musk fragrances in digested sludge. Water Research 42 (1e2), 287e295. Carballa, M., Omil, F., Ternes, T.A., Lema, J.M., 2007. Fate of pharmaceutical and personal care products (PPCPs) during anaerobic digestion of sewage sludge. Water Research 41, 2139e2150. 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, 947e954. Desbrow, C., Routledge, E.J., Brighty, G.C., Sumpter, J.P., Waldock, M., 1998. Identification of estrogenic chemicals in STW effleunt. 1. Chemical fractionation and in vitro biological screening. Environmental Science and Technology 32 (11), 1549e1558. Dokianakis, S.N., Kornaros, M.E., Lyberatos, G., 2004. On the effect of pharmaceuticals on bacterial nitrite oxidation. Water Science and Technology 50 (5), 341e346. Dytczak, M.A., Londry, K.L., Oleszkiewicz, J.A., 2008. Biotransformation of estrogens in nitrifying activated sludge under aerobic and alternating anoxic/aerobic conditions. Water Environment Research 80 (1), 47e52. Ericson, J.F., 2007. An evaluation of the OECD 308 Water/Sediment Systems for investigating the biodegradation of pharmaceuticals. Environmental Science and Technology 41, 5803e5811. Fountoulakis, M., Drillia, P., Stamatelatou, K., Lyberatos, G., 2004. Toxic effect of pharmaceuticals on methanogenesis. Water Science and Technology 50 (5), 335e340.
382
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 5 e3 8 3
Fountoulakis, M., Stamatelatou, K., Lyberatos, G., 2008. The effect of pharmaceuticals on the kinetics of methanogenesis and acetogenesis. Bioresource Technology 99, 7083e7090. de Graaff, M.S., Zeeman, G., Temmink, H., Buisman, C.J.N., 2010a. Anaerobic treatment of concentrated black water in a UASB reactor at a short HRT. Water 2 (1), 101e119. de Graaff, M.S., Zeeman, G., Temmink, H., van Loosdrecht, M.C.M., Buisman, C.J.N., 2010b. Long term partial nitritation of anaerobically treated black water and the emission of nitrous oxide. Water Research 44, 2171e2179. de Graaff, M.S., Temmink, H., Zeeman, G., van Loosdrecht, M.C. M., and Buisman, C.J.N., 2010c. Autotrophic nitrogen removal from black water: calcium addition as a requirement for settleability. Water Research, in press, doi:10.1016/j.watres. 2010.08.010. Griens, A.M.G.F., Tinke, J.L., 2006. Data and facts 2006 by the organization for pharmaceutical facts and figures. In Dutch: Data en feiten 2006, Stichting Farmaceutische Kengetallen. www.sfk.nl juli 2006. Halling-Sørensen, B., Holten Lu¨tzhøft, H.-C., Andersen, H.R., Ingerslev, F., 2000. Environmental risk assessment of antibiotics: comparison of mecillinam, trimethoprim and ciprofloxacin. Journal of Antimicrobial Chemotherapy 46 (Suppl. S1), 53e58. Harries, J.E., Sheahan, D.A., Jobling, S., Matthiessen, P., Neall, P., Sumpter, J.P., Tylor, T., Zaman, N., 1997. Estrogenic activity in five United Kingdom rivers detected by measurement of vitellogenesis in caged male trout. Environmental Toxicology and Chemistry 16 (3), 534e542. 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 and Technology 43, 7862e7869. Jones, O.A.H., Green, P.A., Voulvoulis, N., Lester, J.N., 2007. Questioning the excessive use of advanced treatment to remove organic micropollutants from wastewater. Environmental Science and Technology 41, 5085e5089. Joss, A., Carballa, M., Kreuzinger, N., Siegrist, H., Zabczynski, S., 2006a. Wastewater treatment. In: Ternes, T.A., Joss, A. (Eds.), Human Pharmaceuticals, Hormones and Fragrances: The Challenge of Micropollutants in Urban Water Management. IWA Publishing, London, UK. Joss, A., Zabczynski, S., Go¨bel, A., Hoffman, B., Lo¨ffler, D., McArdell, C.S., Ternes, T.A., Thomsen, A., Siegrist, H., 2006b. Biological degradation of pharmaceuticals in municipal wastewater treatment: proposing a classification scheme. Water Research 40, 1686e1696. Kim, S., Eichhorn, P., Jensen, J.N., Weber, A.S., Aga, D.S., 2005. Removal of antibiotics in wastewater: effect of hydraulic and solid retention times on the fate of tetracycline in the activated sludge process. Environmental Science and Technology 39, 5816e5823. KNMP, 2006. Informatorium Medicamentorum. Koninklijke Nederlandse Maatschappij ter Bevordering der Pharmacie, Den Haag, The Netherlands. Kosonen, J., Kronberg, L., 2009. The occurence of antihistamines in sewage waters and in recipient rivers. Environmental Science and Pollution Research 16, 555e564. Kujawa-Roeleveld, K., Schuman, E., Grotenhuis, T., Kragic, D., Mels, A., and Zeeman, G. (2008). Biodegradability of human pharmaceutically active compounds (PhAC) in biological systems treating source separated wastewater streams. In “Sanitation Challenge”, Wageningen, the Netherlands, 19e21 May 2008. Kujawa-Roeleveld, K., Zeeman, G., 2006. Anaerobic treatment in decentralised and source-separation-based sanitation
concepts. Reviews in Environmental Science and Bio/ Technology 5, 115e139. Larsen, T.A., Alder, A.C., Eggen, R.I.L., Maurer, M., Lienert, J., 2009. Source separation: will we see a paradigm shift in wastewater handling? Environmental Science and Technology 43 (16), 6121e6125. Larsen, T.A., Lienert, J., Joss, A., Siegrist, H., 2004. How to avoid pharmaceuticals in the aquatic environment. Journal of Biotechnology 113, 295e304. Li, L., Zhu, W., Zhang, P., Zhang, Z., Wu, H., Han, W., 2006. Comparison of AC/O3-BAC and O3-BAC processes for removing organic pollutants in secondary effluent. Chemosphere 92, 1514e1522. Lindqvist, N., Tuhkanen, T., Kronberg, L., 2005. Occurrence of acidic pharmaceuticals in raw and treated sewages and in receiving waters. Water Research 39, 2219e2228. Maurer, M., Escher, B.I., Richle, P.R., Schaffner, C., Alder, A.C., 2007. Elimination of b-blockers in sewage treatment plants. Water Research 41, 1614e1622. de Mes, T., Zeeman, G. and Lettinga, G. (2005). Occurence and fate of estrone, 17b-estradiol and 17a-ethynylestradiol in STPs for domestic wastewater. Reviews in Environmental Science and Bio/Technology (4), 275e311. de Mes, T. (2007). Fate of estrogens in biological treatment of concentrated black water, PhD thesis Environmental Technology, Wageningen University, 152 p. de Mes, T.Z.D., Zeeman, G., Lettinga, G., 2007. Fate of estrogens during anaerobic black water treatment with micro-aerobic post treatment. Water Science and Technology 56 (5), 15e23. Mu¨ckter, H., 2006. Human and animal toxicology of some waterborne pharmaceuticals. In: Ternes, T.A., Joss, A. (Eds.), Human Pharmaceuticals, Hormones and Fragrances: The Challenge of Micropollutants in Urban Water Management. IWA Publishing, London, UK. Nakada, N., Kiri, K., Shinohara, H., Harada, A., Kuroda, K., Takizawa, S., Takada, H., 2008. Evaluation of pharmaceuticals and personal care products as water-soluble molecular markers of sewage. Environmental Science and Technology 42, 6347e6353. Otterpohl, R., Albold, A., Oldenburg, M., 1999. Source control in urban sanitation and waste management: ten systems with reuse of resources. Water Science and Technology 39 (5), 153e160. Scheurer, M., Ramil, M., Metcalfe, C.D., Groh, S., Ternes, T.A., 2009. The challenge of analyzing beta-blocker drugs in sludge and wastewater. Analytical and Bioanalytical Chemistry. doi: 10.1007/s00216-009-3225-7. Schirmer, K., Schirmer, M., 2008. Who is chasing whom? A call for a more integrated approach to reduce the load of micropollutants in the environment. Water Science and Technology 71 (1), 145e150. Schrap, S.M., Rijs, G.B.J., Beek, M.A., Maaskant, J.F.N., Staeb, J., Stroomberg, G., Tiesnitsch, J., 2003. Human and veterinary pharmaceuticals in Dutch surface water and wastewater In Dutch: Humane an veterinaire geneesmiddelen in Nederlands oppervlaktewater en afvalwater. RIZA rapport 2003.023. Schwarzenbach, R.P., Escher, B.I., Fenner, K., Hofstetter, T.B., Johnson, C.A., Von Gunten, U., Wehrli, B., 2006. The challenge of micropollutants in aquatic systems. Science 313 (25 August 2006), 1072e1077. Ternes, T.A., 1998. Occurence of drugs in German sewage treatment plants and rivers. Water Research 32 (11), 3245e3260. Ternes, T.A., Andersen, H., Gilberg, D., Bonerz, M., 2002. Determination of estrogens in sludge and sediments by liquid extraction and GC/MS/MS. Analytical Chemistry 74, 3498e3504. Ternes, T.A., Bonerz, M., Herrmann, N., Lo¨ffler, D., Keller, E., Bago´ Lacida, B., Alder, A.C., 2005. Determination of
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 5 e3 8 3
pharmaceuticals, iodinated contrast media and musk fragrances in sludge by LC/tandem MS and GC/MS. Journal of Chromatography A 1067, 213e223. Ternes, T.A., Joss, A., 2006. Human Pharmaceuticals, Hormones and Fragrances, the Challenge of Micropollutants in Urban Water Management. IWA Publishing, London. Ternes, T.A., Kreckel, P., Mueller, J., 1999. Behaviour and occurrence of estrogens in municipal sewage treatment plants - II. Aerobic batch experiments with activated sludge. The Science of the Total Environment 225, 91e99. Thomas, K.V., Dye, C., Schlabach, M., Langford, K.H., 2007. Source to sink tracking of selected human pharmaceuticals from two Oslo city hospitals and a wastewater treatment works. Journal of Environmental Monitoring 9, 1410e1418. Vlaeminck, S.E., Terada, A., Smets, B.F., van der Linden, D., Boon, N., Verstraete, W., Carballa, M., 2009. Nitrogen removal from digested black water by one-stage partial nitritation and anammox. Environmental Science and Technology 43 (13), 5035e5041. WHO, 2006. ATC (Anatomical Therapeutic Chemical) Index with DDDs. WHO Collaborating Centre for Drug Statistics
383
Methodology. Norwegian Institute of Public Health, Oslo, Norway. Winker, M., Tettenborn, F., Faika, D., Gulyas, H., Otterpohl, R., 2008. Comparison of analytical and theoretical pharmaceutical concentrations in human urine in Germany. Water Research 42, 3633e3640. Winker, M., Vinnera˚s, B., Muskolus, A., Arnold, U., Clemens, J., 2009. Fertiliser products from new sanitation systems: their potential values and risks. Bioresource Technology 100. Yi, T., Harper Jr., W.F., 2007. The link between nitrification and biotransformation of 17a-ethinylestradiol. Environmental Science and Technology 41, 4311e4316. 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. Zeeman, G., Kujawa, K., Meulman, B., and Kwant, F. (2007). Full scale demonstration of vacuum collection, transport & treatment of black water, Poster presentation, Advanced Sanitation Conference, March 2007, Aachen, Germany.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 4 e3 9 2
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Effects of water chemistry on arsenic removal from drinking water by electrocoagulation Wei Wan a, Troy J. Pepping a, Tuhin Banerji b, Sanjeev Chaudhari b, Daniel E. Giammar a,* a
Department of Energy, Environmental and Chemical Engineering and Center for Materials Innovation, Washington University in St. Louis, One Brookings Drive, St. Louis, Missouri 63130, United States b Centre for Environmental Science and Engineering, Indian Institute of Technology Bombay, Powai, Mumbai 400076, India
article info
abstract
Article history:
Exposure to arsenic through drinking water poses a threat to human health. Electro-
Received 17 May 2010
coagulation is a water treatment technology that involves electrolytic oxidation of anode
Received in revised form
materials and in-situ generation of coagulant. The electrochemical generation of coagulant
28 July 2010
is an alternative to using chemical coagulants, and the process can also oxidize As(III) to As
Accepted 10 August 2010
(V). Batch electrocoagulation experiments were performed in the laboratory using iron
Available online 17 August 2010
electrodes. The experiments quantified the effects of pH, initial arsenic concentration and oxidation state, and concentrations of dissolved phosphate, silica and sulfate on the rate
Keywords:
and extent of arsenic removal. The iron generated during electrocoagulation precipitated
Arsenic
as lepidocrocite (g-FeOOH), except when dissolved silica was present, and arsenic was
Water treatment
removed by adsorption to the lepidocrocite. Arsenic removal was slower at higher pH.
Electrocoagulation
When solutions initially contained As(III), a portion of the As(III) was oxidized to As(V)
Phosphate
during electrocoagulation. As(V) removal was faster than As(III) removal. The presence of 1
Adsorption
and 4 mg/L phosphate inhibited arsenic removal, while the presence of 5 and 20 mg/L silica
Lepidocrocite
or 10 and 50 mg/L sulfate had no significant effect on arsenic removal. For most conditions examined in this study, over 99.9% arsenic removal efficiency was achieved. Electrocoagulation was also highly effective at removing arsenic from drinking water in field trials conducted in a village in Eastern India. By using operation times long enough to produce sufficient iron oxide for removal of both phosphate and arsenate, the performance of the systems in field trials was not inhibited by high phosphate concentrations. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Exposure to arsenic through drinking water is a great threat to human health. Arsenic is a carcinogen and its consumption can negatively affect the gastrointestinal tract and cardiac, vascular and central nervous systems. Considering the high toxicity of arsenic, the World Health Organization (WHO) and U.S. Environmental Protection Agency set the maximum acceptable level of arsenic in drinking water at 10 mg/L
(U.S. EPA, 2001, World Health Organization, 1993). Arsenic occurs in groundwater primarily as the result of natural weathering of arsenic-containing rocks, although in certain areas, high arsenic concentrations are caused by industrial waste discharges and application of arsenical herbicides and pesticides (Smedley and Kinniburgh, 2002). Iron oxides have been widely used as sorbents for arsenic removal. They have strong adsorption affinities for arsenic and can have large specific surface areas (Dixit and Hering,
* Corresponding author. Tel.: þ1 314 935 6849; fax: þ1 314 935 7211. E-mail address:
[email protected] (D.E. Giammar). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.016
385
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 4 e3 9 2
2003). Arsenic is present in water mainly in the forms of arsenate (As(V)) and arsenite (As(III)). In the environmentally relevant pH range 4e10, the dominant As(V) species are 2 negatively charged (H2AsO 4 and HAsO4 ), while the dominant As(III) species is neutrally charged (H3AsO3). The negatively charged As(V) species are more likely to be adsorbed and are generally more easily removed than As(III) in treatment systems (Balasubramanian and Madhavan, 2001; Kumar et al., 2004; Parga et al., 2005). Electrocoagulation is an alternative to using chemical coagulants for arsenic removal, and thus may be beneficial for communities with better access to electricity than to chemicals. Electrocoagulation involves electrolytic oxidation of an appropriate anode material and in-situ generation of coagulant (Kumar et al., 2004; Lakshmanan et al., 2009). When a direct current is applied between two electrodes, metal ions such as Fe2þ and Al3þ that can contribute to coagulant formation are released by anode oxidation. With iron electrodes, the Fe2þ released can subsequently be oxidized in solution to produce an Fe(III) hydroxide or oxyhydroxide (Lakshmanan et al., 2009). Several previous studies have reported arsenic removal from water and wastewater by electrocoagulation (Balasubramanian and Madhavan, 2001; Kumar et al., 2004; Parga et al., 2005; Thella et al., 2008). During such processes, arsenic removal by electrocoagulation involved metal oxide formation followed by arsenic removal (Balasubramanian and Madhavan, 2001). Electrocoagulation may also control oxidationereduction reactions; species such as As(III) may be oxidized on the anode and other species may be reduced on the cathode. Several water chemistry factors may affect arsenic removal by electrocoagulation. The pH can affect arsenic species distribution and the surface charge of the metal oxides formed. Arsenic oxidation state can affect arsenic removal, with As(V) being more easily removed than As(III) (Kumar et al., 2004). Phosphate may compete with arsenic for adsorption sites (Meng et al., 2002). Silica can be present at high concentration in groundwater, and in studies with iron oxide-based sorbents it has inhibited arsenic removal (Davis et al., 2001; Zeng et al., 2008a). The objectives of this study were to assess the impact of important water chemistry factors on arsenic removal by electrocoagulation and to examine the performance of electrocoagulation for arsenic removal in laboratory and field settings. Factors studied were the pH, arsenic oxidation state, initial arsenic concentration, and the concentrations of dissolved phosphate, silica, and sulfate.
2.
Materials and methods
2.1.
Laboratory-scale electrocoagulation experiments
The electrocoagulation reactor consisted of a 1 L glass beaker with two iron rods immersed in the aqueous solution. The rods had diameters of 1.75 cm, lengths of 20 cm, and were placed 2 cm apart in the arsenic-containing solution. The total submerged surface area of each electrode was 57 cm2. Before each experiment, the electrodes were abraded with sand paper to remove scales and then cleaned with 1 M HNO3 and
ultrapure water. A voltage of 12 V was applied to the terminal electrodes from a direct current power supply. The electric current was monitored over the course of each two-hour experiment. To provide enough oxygen for the formation of Fe (III) precipitates, the solution was sparged with air at a flow rate of 60 mL/min. The arsenic-containing solution was magnetically-stirred (200 rpm). Duplicate runs were carried out for each set of experimental conditions. The water compositions evaluated are listed in Table 1. All solutions were prepared with ultrapure water and desired volumes of stock solutions. An As(V) stock solution was made from Na2HAsO4$7H2O. The As(III) stock solution was made from NaAsO2. Tests indicated that the stock As(III) solution contained about 15% As(V). The phosphate, silica, and sulfate stock solutions were made from Na2HPO4$7H2O, Na2SiO3$9H2O, and Na2SO4$10H2O, respectively. In order to provide pH buffering, NaHCO3 was added to achieve a concentration of 1 mM. The pH of the As-containing solution in each beaker was periodically readjusted to the target value by adding aliquots of 1 M HNO3 or 1 M NaOH. For the experiments using As(V), every 5, 15 or 30 min, 15 mL of solution was collected from the beaker. Of this amount, 7.5 mL was filtered using a 0.45 mm filter membrane (polyethersulfone), and the filtrate was acidified to 1% HNO3. Another 7.5 mL of unfiltered suspension was acidified to 1% HNO3, which completely dissolved the suspended solids. For the experiments using As(III), an additional sample was collected for arsenic redox speciation (i.e. separation of As(III) and As(V)) using an anion-exchange method (Wilkie, 1997). Before separation, the pH of a 10 mL aliquot of filtered solution was adjusted to around 3.5 and then passed through a column containing anion-exchange resin. During As separation, the first 5 mL of solution were wasted and the remaining 5 mL were collected. In this method, As(III) eluted through the column and As(V) was retained on the resin in the column. At the end of each experiment, 15 mL of unfiltered suspension was collected for zeta potential measurement. The remaining settleable solids were collected and freeze-dried in preparation for solid-phase characterization. The pH of the solution in the beaker was periodically measured over the course of the experiment with a pH electrode and adjusted to the target value if necessary. The current was briefly turned off during the times when the pH was measured, because the current interfered with the pH measurement. It took approximately 1 min to measure and adjust the pH to the desired value. Considering the time used to adjust the pH, each experiment lasted about 130 min, although current was only applied during 120 min of the experiment.
Table 1 e Experimental variables evaluated in the electrocoagulation experiments. Parameters Initial arsenic concentration Arsenic oxidation state pH Phosphate Silica Sulfate
Range of values 100 and 1000 mg/L As(III) and As(V) 5, 6, 7, 8 and 9 0, 1 and 4 mg/L as P 0, 5 and 20 mg/L as SiO2 0, 10 and 50 mg/L as SO2 4
386
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 4 e3 9 2
The filtered and acid-treated samples from the laboratory experiments were analyzed for dissolved and total concentrations of constituents, respectively. The concentrations of As, Fe, P and Si were determined by inductively coupled plasma mass spectrometry (ICP-MS) (7500ce, Agilent Technologies, Santa Clara, CA). The instrument detection limits for As, Fe, P and Si were 0.1 mg/L, 0.05 mg/L, 0.01 mg/L, and 0.03 mg/L, respectively. The specific surface areas (SSA) of the solids were measured by the BET N2-adsorption method (Autosorb-1-C, Quantachrome, U.S.A.). X-ray powder diffraction (XRD) patterns were collected using Cu Ka radiation (D-MAX/A, Rigaku, Japan). Zeta potential was measured by a nanoparticle characterization instrument with zeta potential capability (Nanoseries ZS, Malvern Instruments, U.K.). Dissolved oxygen for selected samples was measured using a Hach Surface Water Test Kit (Fondriest Environmental, Inc.).
2.2.
Materials and methods of the field study
The electrocoagulation system used for the field studies consisted of a 50 L plastic bucket, a 12 V direct current source with a rating of 2 A, two iron plates (10 cm by 15 cm with a submerged depth of 10 cm), and an aquarium pump with a diffuser. The iron plates were separated by a distance of 0.5 cm and were connected to a support that fixed them in place with a non-conductive PVC screw. The electrocoagulation system was followed by a common ceramic candle filter assembly (average pore size of 1 mm) for the removal of suspended particles (Fig. 1). Field trials for the electrocoagulation system were carried out in a village in the Nadia district of West Bengal. The village chosen for the field trials was Ghetugachi, which falls under Chakdaha block. This village has more than 650 families. The average arsenic level was found to be 400 mg/L, and the phosphate levels in the groundwater were also high. Dissolved iron in the groundwater ranges from 0.5 to 4.0 mg/L. Ghetugachi was selected for the field trials because a predominance of As(III) and the high phosphate concentrations presented a challenging scenario for arsenic removal. A total of 17 electrocoagulation systems were distributed in Ghetugachi. After they had been distributed and used by the households for one week, the field systems were evaluated by treating 50 L volumes of water collected in the bucket from the hand pumps located close to each house where an electrocoagulation system had been installed. The aquarium pump was switched 12 V DC + source
air pump
As(III) oxidation As(V)
on for 15 min to allow the dissolved oxygen in the water to increase to over 3 mg/L. The electrode assembly was then immersed in the water and the voltage was applied for 3 h to provide electrocoagulation treatment. The water was then allowed to settle for 4 h. The supernatant remaining after settling was collected and filtered through a ceramic candle filter that was procured locally. For each system, three water samples were collected: (1) raw water prior to aeration, (2) supernatant of the treated water after 4 h of settling, and (3) water after candle filtration. The water samples were preserved by acidification with 0.1 N HCl. The field samples were analyzed by spectroscopic methods. The arsenic concentrations were determined with a rapid spectrophotometric method (Dhar et al., 2004). The iron concentrations were determined using the phenanthroline method (Clesceri et al., 1999). Additional samples were selected randomly for analysis by atomic absorption spectrometry (AA 400-FIAS, Perkin Elmer, USA) to verify the iron and arsenic concentrations determined by the colorimetric methods. The error of the spectrophotometric methods (for arsenic and iron analysis) was only 2.0%. The phosphate concentrations in the field samples were determined using the ascorbic acid method (Clesceri et al., 1999).
3.
Results and discussion
3.1.
Production of iron oxide coagulant
During the electrocoagulation process, the solutions changed from clear and colorless to turbid and reddish brown. The concentrations of the total iron generated during electrocoagulation increased linearly with reaction time. In this study, the reported values are the average plus the standard deviation. For all the electrocoagulation experiments, about 50 mg/L (average value was 50.5 mg/L) total iron was produced over the 2 h experiment duration. The reactor was operated with a current of 22 mA, and the total iron produced was consistent with a value of 52.2 mg/L predicted by Faraday’s law for the oxidation of the iron electrode to dissolved Fe(II). The current was nearly identical for the different pH values studied and did not vary over the course of the experiment, which indicated that the conductivity of the solutions was not affected by the electrocoagulation process. The total electricity used to treat 1 L of water was 0.0005 kW h or 0.5 kW h/m3. Fe was released to solution as Fe(II) (Reaction (1)) and was then oxidized to Fe(III)
iron electrodes water level iron corrosion and oxide precipitation
unfiltered water ceramic candle filter 1 µm pore size
adsorption As water level settling
diffuser
filtered water
Fig. 1 e Domestic electrocoagulation system used for the field study. Water from the electrocoagulation reactor (left) is further treated using a candle filter unit (right) to remove suspended particles.
387
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 4 e3 9 2
by the dissolved oxygen. The Fe(III) then precipitated in the formation of iron oxides (Reaction (2)).
2-
100 ppb As(III), pH = 7, 10 ppm SO4
100 ppb As(III), pH = 7, 5 ppm SiO2
(1)
Fe2þ þ 1:5H2 O þ 0:25O2 /g-FeOOHðsÞ þ 2Hþ
(2)
For the experiment using 100 mg/L As(V) at pH 7, the dissolved oxygen concentration was measured to be about 10 mg/L at 30, 60 and 90 min, which verified that the solution was saturated with dissolved oxygen. In each experiment the dissolved iron concentrations were very low compared with the total iron concentrations, which indicated that nearly all of the iron was present in the solid phases. The Fe(III) solid was identified as lepidocrocite (g-FeOOH) by its XRD pattern (Fig. 2). The identity of the iron oxyhydroxides formed was not affected by As(III) and As(V) or by the presence of sulfate or phosphate. The lepidocrocite particles were 100e200 nm long and about 5e20 nm wide. The XRD pattern at pH 5 had no clear XRD peaks, probably resulting from insufficient collection of solids for XRD characterization. The specific surface area of the solids was 200 m2/g and was independent of the solution compositions. The formation of lepidocrocite is consistent with published synthesis methods involving oxidation of Fe(II) solutions using dissolved oxygen at ambient temperature (Schwertmann and Cornell, 2000). The presence of silica significantly affected the iron oxides formed during electrocoagulation (Fig. 2). Less crystalline iron oxides were formed in the presence of silica. A previous study found that silica influenced the type of iron oxides formed in aerated Fe (II)- and As(III)-containing water by decreasing the degree of corner-sharing linkages of Fe(III)-octahedra during Fe(III) polymerization (Voegelin et al., 2010). Equilibrium calculations indicate that Fe(III) oxyhydroxides should precipitate at the
100 ppb As(III), pH = 7, 1 ppm P
Normalized Intensity
FeðsÞ /Fe2þ þ 2e
100 ppb As(III), pH = 7 100 ppb As(V), pH = 5 100 ppb As(V), pH = 9 1000 ppb As(V), pH = 7 100 ppb As(V), pH = 7 PDF#00-044-1415, Lepidocrocite: FeO(OH)
10
20
30
40
o
2θ( )
50
60
70
80
Fig. 2 e X-ray diffraction patterns of solids generated during electrocoagulation. The reference pattern for lepidocrocite is included for comparison.
experimental conditions and that no arsenic-Fe(III) precipitates (e.g. FeAsO4(s)) were expected to form.
3.2.
Effect of pH on As removal
The removal rate of arsenic was affected by the pH. For experiments using 100 mg/L As(V), it took less than 30 min for the dissolved As(V) concentrations to drop below 1 mg/L at pH 5, 6 and 7 (Table 2). At pH 8, the dissolved As(V) concentration only reached this value after 75 min, and at pH 9 the dissolved
Table 2 e Water compositions in electrocoagulation experiments and the time to achieve dissolved arsenic concentrations below 10 mg/L and 1.0 mg/L. Arsenic oxidation state As(III) As(III) As(III) As(III) As(III) As(III) As(V) As(V)a As(V)a As(V) As(V)a As(V)a As(V) As(V) As(V) As(V) As(V) As(V) As(III) As(III) As(III)
Arsenic (mg/L)
pH
Phosphate as P (mg/L)
Silica as SiO2 (mg/L)
Sulfate as ) So2 4 (mg/L
Time to dissolved As <10 mg/L (min)
Time to dissolved As <1.0 mg/L (min)
100 100 100 100 100 1000 1000 100 100 100 100 100 100 100 100 100 100 100 100 100 100
5 6 7 8 9 7 7 5 6 7 8 9 7 7 7 7 7 7 7 7 7
0 0 0 0 0 0 0 0 0 0 0 0 1 4 0 0 0 0 1 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 20 0 0 0 5 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 10 50 0 0 10
30 30 30 30 45 90 45 15 15 5 75 75 60 90 5 5 5 5 45 30 30
90 90 90 90 90 120 >120 15 30 15 75 >120 >120 >120 60 60 30 45 90 90 90
a The first sample was collected after 15 min in these experiments but after 5 min in all other experiments.
388
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 4 e3 9 2
pH 5 pH 6 pH 7 pH 8 pH 9
80
80
60
40
20
0 0
15
30
45
60
75
90
105
120
Time (min) Fig. 4 e Dissolved concentrations of total As and As(V) during electrocoagulation of solutions initially containing 100 mg/L As(III) at pH 5, 7, and 9.
was modeled as a first-order reaction with respect to arsenic (Balasubramanian et al., 2009). Although such a model did not account for the change in the Fe(III) oxyhydroxide coagulant with time and predicts that the arsenic concentration approaches an unreasonable value of zero with increasing reaction time, this first-order model could still provide a good fit to the experimental data for the treatment times investigated.
3.3.
Effect of oxidation state on As removal
As(V) removal was usually faster than As(III) removal (Figs. 3 and 4). This was in agreement with the results reported by Kumar et al. (2004) for electrocoagulation and by Meng et al. (2002) for arsenic adsorption to iron hydroxides. In all the experiments using As(III), the dissolved As(V) concentration increased first and then decreased with increasing reaction
40 No As 1000 g/L As(V) 100 g/L As(III)
30
Zeta Potential (mV)
D i ss o l v e d A s ( V ) ( g/L)
100
pH 5-As pH 5-As(V) pH 7-As pH 7-As(V) pH 9-As pH 9-As(V)
100
Dissolved As(V) ( g/L)
As(V) concentrations remained at 4 mg/L after 120 min (Fig. 3). For experiments using 100 mg/L As(III), it took 30 min for the dissolved As concentrations to drop below 10 mg/L at pH 5e8, and at pH 9 it took 45 min (Fig. 4 and Table 2). Although As removal was slower at higher pH, the final removal efficiency was independent of pH from 5 to 8. Once sufficient lepidocrocite was produced to provide adsorption sites for As, low dissolved As concentrations could be obtained. Kumar et al. (2004) also reported that the final As removal efficiency was independent of pH with increasing pH from 6 to 8 during electrocoagulation. The influence of pH on adsorption can explain the slower rate of As(V) removal with increasing pH. Previous studies observed less adsorption of As(V) to hydrous ferric oxide and goethite with increasing pH (Dixit and Hering, 2003; Meng et al., 2000). The lepidocrocite produced in the electrocoagulation reactor had an isoelectric pH of about 7.0 (Fig. 5), which is comparable to that reported by Peacock and Sherman (2004) from a potentiometric titration method. Below this pH the surfaces of the particles are positively charged and electrostatic contributions as well as chemical contributions contribute to As(V) adsorption. Above the isoelectric point, both the As(V) species and the lepidocrocite surface are negatively charged and adsorption is less favorable. The rate of arsenic removal depends on the rate of lepidocrocite production and the rate of arsenic adsorption to the lepidocrocite. A recent model combined these two processes to interpret As(V) removal from solution during electrocoagulation (Wan, 2010). The rate of lepidocrocite generation is essentially constant because of the nearly constant release of Fe(II) from the electrode. At the conditions studied Fe(II) oxidation to produce lepidocrocite is fast relative to Fe(II) release from the electrode, so the release of Fe(II) at the electrode is the rate-limiting step in lepidocrocite production. Arsenic adsorption to the lepidocrocite was successfully modeled as a process that was first-order with respect to lepidocrocite, first-order with respect to dissolved arsenic, and that proceeded until equilibrium adsorption was reached. In a separate study, arsenic removal during electrocoagulation
60
40
20
20 10 0 -10 -20 -30
0 0
15
30
45
60
75
90
105
120
Time (min) Fig. 3 e Dissolved As(V) concentrations during electrocoagulation of solutions initially containing 100 mg/L As(V). The data points for pH 5 are partially obscured by those for pH 6.
-40 3
4
5
6
7
8
9
10
11
12
pH Fig. 5 e Effect of pH on zeta potential of the lepidocrocite suspension generated by electrocoagulation in the presence and absence of dissolved arsenic.
389
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 4 e3 9 2
Effect of initial As concentrations on As removal
As(V) and As(III) removal to below 1 mg/L took more time when the solutions had higher initial As concentrations (Fig. 6 and Table 2). When the initial arsenic concentrations were higher, more iron oxides were needed to decrease the dissolved arsenic concentrations. Arsenic removal is consequently limited by the production rate of lepidocrocite. However, the final As removal efficiencies were independent of the initial As (III) concentration and were over 99.9%.
3.5.
[As]/[As]0
Dissolved P(mg/L)
b
0.6 0.4
0.0 0
15
30
45
60
75
15
30
45
60
75
90
105
120
No P 1 mg/L P 4 mg/L P
100 80 60 40 20 0 0
c
60 40
15 , , ,
30
45
60
75
90
105
120
60
75
90
105
120
No P 1 mg/L P 4 mg/L P
20
0
0.2
90
105
120
100 g/L 1000 g/L
1.0 0.8
[As]/[As]0
0
0
100 g/L-As 100 g/L-As(V) 1000 g/L-As 1000 g/L-As(V)
1.0 0.8
b
2
Effect of phosphate on As removal
The presence of 1 mg/L and 4 mg/L phosphate as P inhibited the removal of As (Figs. 7 and 8). The inhibitory effect was more significant at higher phosphate concentrations. Considerable phosphate was also removed during electrocoagulation (Figs. 7a and 8a), which indicates that phosphate can compete with As species for the surface sites of
a
1 mg/L P 4 mg/L P
4
0
Dissolved As(V)(μg/L)
3.4.
a
Fe (mg/L)
time. The increase in As(V) when treating As(III) solutions indicated that at least 25% of the As(III) was oxidized to As(V) during electrocoagulation (Fig. 4). The removal mechanism for As(III) by electrocoagulation was proposed to be the oxidation of As(III) to As(V) followed by adsorption of As(V) to the iron oxides (Kumar et al., 2004). As(III) oxidation to As(V) has previously been proposed to occur with dissolved oxygen and soluble intermediates in Fe(II) oxidation acting as rateenhancing species (Ciardelli et al., 2008; Sahai et al., 2007). As (III) oxidation can also occur when Fe(II) is present with Fe(III) oxyhydroxides, and the mechanism has been proposed to involve the formation of reactive Fe intermediate species (Amstaetter et al., 2010; Bisceglia et al., 2005).
15
30
45
Time (min) Fig. 7 e Dissolved (a) phosphate (in mg P/L), (b) dissolved As (V), and (c) total (closed symbols) and dissolved (open symbols) iron concentrations during electrocoagulation of solutions initially containing 100 mg/L As(V) at pH 7.
lepidocrocite. A competitive adsorption effect of phosphate on As removal agrees with the results of previous studies (Meng et al., 2002; Zeng et al., 2008b). The inhibitory effect of phosphate on As removal during electrocoagulation may also be caused in part by the slower oxidation of Fe(II) to Fe(III) in the presence of phosphate, which can decrease the rate at which the sorbent is formed (Fig. 7c). Recent work observed formation of Fe(III)-phosphate solids during the oxidation of Fe(II) in phosphate-rich solutions (Voegelin et al., 2010); however, in the present study phosphate did not affect the identity of the iron oxide formed during electrocoagulation (Fig. 2). Lepodocrocite was the only phase indicated in XRD patterns, and there were no peaks for Fe(II)- or Fe(III)-phosphate solids.
0.6
3.6.
Effect of dissolved silica on As removal
0.4 0.2 0.0 0
15
30
45
60
75
90
105
120
Time (min)
Fig. 6 e Relative change in dissolved arsenic during electrocoagulation of solutions initially containing 100 mg/L and 1000 mg/L (a) As(III) and (b) As(V) at pH 7. Panel (a) also shows the concentration of dissolved As(V).
The presence of 5 and 20 mg/L dissolved silica had no significant effect on As removal, even though considerable silica was removed during the electrocoagulation process (Fig. 9) and silica prevented the formation of lepidocrocite (Fig. 2). Meng et al. (2002) also observed no significant effects of silica on As(V) adsorption to iron hydroxides when silica was present at concentrations as high as 36 mg/L. In a separate study, Davis et al. (2001) observed silica inhibition of As(V) adsorption to ferric hydroxide, but only when silica and ferric
390
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 4 e3 9 2
a Dissolved SiO (mg/L)
Dissolved P (mg/L)
a 1.0 0.8 0.6 0.4 0.2
5 mg/L SiO
20
20 mg/L SiO
15 10 5 0
15
30
45
60
75
90
105
120
0.0 15
30
45
60
75
Dissolved As ( g/L)
b 100
90
105
120
No P-As No P-As(V) 1 mg/L P-As 1 mg/L P-As(V)
80
b 100 Dissolved As(V)(μg/L)
0
60 40
No SiO2
80
5 mg/L SiO
60
20 mg/L SiO
40 20 0
15
30
45
,
No SiO
,
5 mg/L SiO
,
20 mg/L SiO
15
30
60
75
90
105
120
60
75
90
105
120
20
c
0 15
30
45
60
75
90
105
60
120
Time (min)
Fig. 8 e Dissolved (a) phosphate (in mg P/L) and (b) As concentrations during electrocoagulation treatment of solutions initially containing 100 mg/L As(III) at pH 7.
Fe (mg/L)
0
40 20 0
45
Time (min) hydroxide had been pre-equilibrated for 50 days; at shorter contact times, there was much less inhibition. An inhibitory effect of silica was also observed in a study of arsenic removal using a porous iron oxide-based sorbed in packed columns (Zeng et al., 2008a); the inhibition was likely caused by silica polymerization to physically block access to adsorption sites within internal pores of the solid and not through any competitive adsorption phenomenon.
3.7.
Effect of sulfate on As removal
The presence of 10 and 50 mg/L SO2 4 (data not shown) did not affect the removal of As. Because sulfate did not affect lepidocrocite formation and does not adsorb as strongly as As(V) and phosphate, it is not surprising that sulfate did not affect the performance of the electrocoagulation process. Similar observations were made in As adsorption experiments with sulfate concentrations higher than 200 mg/L (Meng et al., 2000).
3.8.
Arsenic removal in field trials
Almost all of the field units were able to lower arsenic concentrations to below 10 mg/L after electrocoagulation, settling, and candle filtration (Table 3 and Fig. 10). For the small number of systems that were unable to achieve 10 mg/L (systems 15e17), visits to the households determined that the units were not being operated as per the instructions provided. The users of these units were either not aerating the water (i.e. the aquarium pump was not switched on) or were carrying out the electrocoagulation for less than the 3 h period specified in the instructions.
Fig. 9 e Concentrations of (a) dissolved SiO2, (b) dissolved As(V) and (c) total (closed symbols) and dissolved (open symbols) iron concentrations during electrocoagulation of solutions initially containing 100 mg/L As(V) at pH 7.
The untreated waters had very high arsenic concentrations, in the range 400e700 mg/L (Table 3 and Fig. 10), and most also had high phosphate concentrations. Despite the high initial concentrations, most final filtered water samples had arsenic concentrations less than 10 mg/L, and the phosphate concentrations in treated waters were below the detection limit of 0.5 mg/L. While significant removal of arsenic occurred with electrocoagulation followed by settling, the filtration step led to further decrease in the arsenic concentration. In several cases the filtration step was required to lower the arsenic concentrations below the 10 mg/L WHO standard and not just the 50 mg/L Indian standard. Additional arsenic removal probably occurred on the 1 mm pore size candle filters due to the removal of microflocs of iron oxide that adsorbed arsenic but were not separated during the sedimentation phase. No inhibition of treatment performance from phosphate was observed in the field trials. The systems with the highest phosphate concentrations (9 and 10) had settled and filtered water concentrations that were comparable to those from the other properly operated systems. The electrocoagulation treatment time for the field trials was selected to produce sufficient iron oxide coagulant to remove both As and phosphate, so it is not surprising that no effect of phosphate was seen in these trials. Preliminary laboratory studies with the reactors used in the field indicated that when water contained
391
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 4 e3 9 2
Table 3 e Untreated and treated water compositions for field investigation of electrocoagulation systems. System
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
pH
Initial Conc. (mg/L)
7.1 7.1 6.8 6.8 7.0
7.0
7.0
Supernatant Conc. (mg/L)
Astot
As(V)
PO3 4 eP
526 526 484 449 449 483 677 526 675 677 488 449 527 483 527 482 677
125 125
203 203 198 182 182 192 253 198 753 753 144 156 193 178 198 219 753
92 92 103
103
103
Filtrate Conc. (mg/L)
Astot
PO3 4 eP
Astot
PO3 4 eP
13.3 9.4 12.7 13.6 8.2 12.1 16.9 24.1 11.8 11.9 7.9 12.6 14.7 30.7 48.5 50.4 50.9
n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. 3 3 5 21 33 31
0.4 0.2 n.d. n.d. n.d. n.d. n.d. n.d. 3.9 0.5 4.5 n.d. n.d. 5.8 10.3 26.4 25.8
n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. 2 5 4
800
80
700
70
600
60
500
50
400
40
300
30
200
20
100
10
As Conc. in Treated Water (ug/L)
Raw water As Conc. (ug/ L)
*n.d. e not detectable. The detection limits for arsenic and phosphate were 0.2 mg/L and 0.5 mg/L, respectively.
0
0 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
System ID Raw Water
Settled Water
Filtered Water
Fig. 10 e Results of electrocoagulation treatment with domestic systems deployed for the field study. The dashed lines indicate the WHO guideline value for arsenic in drinking water (10 mg/L) and the Indian standards for arsenic in drinking water (50 mg/L). Note that the raw water concentration is shown on a separate axis from the settled and filtered water concentrations.
500 mg/L As(III) and 500 mg P/L phosphate at a pH of approximately 7, the As concentrations could be decreased below 10 mg/L when 20 mg/L of Fe was released to form the coagulant, which occurred in about 1.5 h. To provide a factor of safety, the 3 h treatment time was selected for the field tests, and this reaction time produced 40 mg/L of Fe and overcame the competitive adsorption with phosphate. The iron produced during the treatment time generates about 80 mg of settled solids for each liter of water treated. The electricity required for treatment of 50 L was 0.036e0.039 kW h or 0.72e0.78 kW h/m3. The time required to produce sufficient coagulant for arsenic removal depends on the current, which in turn is affected by the conductivity of the water. The electrode spacing and treatment time can be
adjusted for the conductivity of a specific groundwater. The operating costs of treatment are primarily those of the electricity. Based on typical electricity costs in India, the operating costs are estimated to be 5 Indian Rupees per m3 of water treated ($0.11/m3 with a currency conversion of 46 Rupees per U.S. dollar).
4.
Conclusions
The iron generated during electrocoagulation was present in solid phases as lepidocrocite. Arsenic removal by electrocoagulation involved lepidocrocite formation followed by arsenic adsorption. As removal was slower at higher pH and
392
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 4 e3 9 2
higher initial arsenic concentrations. As(III) was partially oxidized to As(V) during electrocoagulation. As(V) removal was faster than As(III) removal. Phosphate inhibited As removal by acting as a competing adsorbate and possibly by delaying the oxidation of Fe(II) to produce lepidocrocite. Although silica prevented the formation of lepidocrocite, arsenic removal was still very rapid and extensive by adsorption to the amorphous iron oxides that formed. Sulfate had no significant effect on As removal or coagulant formation. Over 99.9% arsenic removal efficiency could be achieved in both the laboratory experiments and the field trials. The field trials indicate that electrocoagulation has the potential to treat arsenic-contaminated water to standards required for drinking water.
Acknowledgments The project was supported by the McDonnell Academy Global Energy and Environment Partnership of Washington University. The field study was performed with the support of the Department of Science and Technology of the Government of India. The authors are grateful to Kate Nelson for assistance with laboratory analyses. The authors thank two anonymous reviewers for their comments that helped improve this paper.
references
Amstaetter, K., Borch, T., Larese-Casanova, P., Kappler, A., 2010. Redox transformation of arsenic by Fe(II)-activated goethite (alpha-FeOOH). Environmental Science & Technology 44 (1), 102e108. Balasubramanian, N., Madhavan, K., 2001. Arsenic removal from industrial effluent through electrocoagulation. Chemical Engineering & Technology 24 (5), 519e521. Balasubramanian, N., Kojima, T., Basha, C.A., Srinivasakannan, C., 2009. Removal of arsenic from aqueous solution using electrocoagulation. Journal of Hazardous Materials 167 (1e3), 966e969. Bisceglia, K.J., Rader, K.J., Carbonaro, R.F., Farley, K.J., Mahony, J.D., Di Toro, D.M., 2005. Iron(II)-catalyzed oxidation of arsenic(III) in a sediment column. Environmental Science & Technology 39 (23), 9217e9222. Ciardelli, M.C., Xu, H.F., Sahai, N., 2008. Role of Fe(II), phosphate, silicate, sulfate, and carbonate in arsenic uptake by coprecipitation in synthetic and natural groundwater. Water Research 42 (3), 615e624. Clesceri, L.S., Greenberg, A.E., Eaton, A.D. (Eds.), 1999. Standard Methods for the Examination of Water and Wastewater. American Public Health Association, American Water Works Association, Water Environment Federation, Washington, DC. 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.
Dhar, R.K., Zheng, Y., Rubenstone, J., van Geen, A., 2004. A rapid colorimetric method for measuring arsenic concentrations in groundwater. Analytica Chimica Acta 526 (2), 203e209. 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. Kumar, P.R., Chaudhari, S., Khilar, K.C., Mahajan, S.P., 2004. Removal of arsenic from water by electrocoagulation. Chemosphere 55 (9), 1245e1252. Lakshmanan, D., Clifford, D.A., Samanta, G., 2009. Ferrous and ferric ion generation during iron electrocoagulation. Environmental Science & Technology 43 (10), 3853e3859. Meng, X.G., Bang, S., Korfiatis, G.P., 2000. Effects of silicate, sulfate, and carbonate on arsenic removal by ferric chloride. Water Research 34 (4), 1255e1261. Meng, X.G., Korfiatis, G.P., Bang, S.B., Bang, K.W., 2002. Combined effects of anions on arsenic removal by iron hydroxides. Toxicology Letters 133 (1), 103e111. Parga, J.R., Cocke, D.L., Valenzuela, J.L., Gomes, J.A., Kesmez, M., Irwin, G., Moreno, H., Weir, M., 2005. Arsenic removal via electrocoagulation from heavy metal contaminated groundwater in La Comarca Lagunera Mexico. Journal of Hazardous Materials 124 (1e3), 247e254. Peacock, C.L., Sherman, D.M., 2004. Copper(II) sorption onto goethite, hematite and lepidocrocite: a surface complexation model based on ab initio molecular geometries and EXAFS spectroscopy. Geochimica Et Cosmochimica Acta 68 (12), 2623e2637. Sahai, N., Lee, Y.J., Xu, H.F., Ciardelli, M., Gaillard, J.F., 2007. Role of Fe(II) and phosphate in arsenic uptake by coprecipitation. Geochimica Et Cosmochimica Acta 71 (13), 3193e3210. Schwertmann, U., Cornell, R.M., 2000. Iron Oxides in the Laboratory. Wiley-VCH, New York, NY. Smedley, P.L., Kinniburgh, D.G., 2002. A review of the source, behaviour and distribution of arsenic in natural waters. Applied Geochemistry 17, 517e568. Thella, K., Verma, B., Srivastava, V.C., Srivastava, K.K., 2008. Electrocoagulation study for the removal of arsenic and chromium from aqueous solution. Journal of Environmental Science and Health Part A-Toxic/Hazardous Substances & Environmental Engineering 43 (5), 554e562. U.S. EPA, 2001. National primary drinking water regulations; arsenic and clarifications to compliance and new source contaminants monitoring; final Rule. Federal Register 66 (14), 6976. Voegelin, A., Kaegi, R., Frommer, J., Vantelon, D., Hug, S.J., 2010. Effect of phosphate, silicate, and Ca on Fe(III)-precipitates formed in aerated Fe(II)- and As(III)-containing water studied by X-ray absorption spectroscopy. Geochimica Et Cosmochimica Acta 74 (1), 164e186. Wan, W., 2010. Arsenic Removal from Drinking Water by Electrocoagulation. M.S. thesis, Washington University, St. Louis. Wilkie, J.A., 1997. Ph.D. dissertation, University of California, Los Angeles, Los Angeles, CA. World Health Organization, 1993. Guidelines for Drinking-water Quality. In: Recommendations, vol. 1. WHO, Geneva. Zeng, H., Arashiro, M., Giammar, D.E., 2008a. Effects of water chemistry and flow rate on arsenate removal by adsorption to an iron oxide-based sorbent. Water Research 42 (18), 4629e4636. Zeng, H., Fisher, B., Giammar, D.E., 2008b. Individual and competitive adsorption of arsenate and phosphate to a highsurface-area iron oxide-based sorbent. Environmental Science & Technology 42 (1), 147e152.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 3 e4 0 3
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Photodegradation of the antibiotics nitroimidazoles in aqueous solution by ultraviolet radiation G. Prados-Joya, M. Sa´nchez-Polo, J. Rivera-Utrilla*, M. Ferro-garcı´a Departamento de Quı´mica Inorga´nica, Facultad de Ciencias, Universidad de Granada, 18071, Granada, Spain
article info
abstract
Article history:
The objective of this study was to analyze the efficacy of ultraviolet (UV) radiation in the
Received 26 May 2010
direct photodegradation of nitroimidazoles. For this purpose, i) a kinetic study was per-
Received in revised form
formed, determining the quantum yield of the process; and ii) the influence of the different
4 August 2010
operational variables was analyzed (initial concentration of antibiotic, pH, presence of
Accepted 10 August 2010
natural organic matter compounds, and chemical composition of water), and the time
Available online 28 September 2010
course of total organic carbon (TOC) concentration and toxicity during nitroimidazole photodegradation was studied. The very low quantum yields obtained for the four nitro-
Keywords:
imidazoles are responsible for the low efficacy of the quantum process during direct
Nitroimidazoles
photon absorption in nitroimidazole phototransformation. The R254 values obtained show
Ultraviolet radiation
that the dose habitually used for water disinfection is not sufficient to remove this type of
Treatment
pharmaceutical; therefore, higher doses of UV irradiation or longer exposure times are
Oxidation
required for their removal. The time course of TOC and toxicity during direct photodegradation (in both ultrapure and real water) shows that oxidation by-products are not oxidized to CO2 to the desired extent, generating oxidation by-products that are more toxic than the initial product. The concentration of nitroimidazoles has a major effect on their photodegradation rate. The study of the influence of pH on the values of parameters 3 (molar absorption coefficient) and k0 E (photodegradation rate constant) showed no general trend in the behavior of nitroimidazoles as a function of the solution pH. The components of natural organic matter, gallic acid (GAL), tannic acid (TAN) and humic acid (HUM), may act as promoters and/or inhibitors of OH$ radicals via photoproduction of H2O2. The effect of GAL on the metronidazole (MNZ) degradation rate markedly differed from that of TAN or HUM, with a higher rate at low GAL concentrations. Differences in MNZ degradation rate among waters with different chemical composition are not very marked, although the rate is slightly lower in wastewaters, mainly due to the UV radiation filter effect of this type of water. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Ultraviolet (UV) radiation is frequently applied to disinfect water intended for human consumption and wastewater. Due to the greater chemical contamination of water, UV radiation (Hijnen et al., 2006) is increasingly proposed as a technology to
remove organic micropollutants, underlining its high efficacy to eliminate certain pesticides and pharmaceuticals from water (Kang et al., 2004; Lazarova and Savoye, 2004). Significant advances have recently been made in our understanding of the photochemical processes undergone by organic contaminants and pharmaceuticals in aqueous
* Corresponding author. Tel.: þ34 958248523; fax: þ34 958248526. E-mail address:
[email protected] (J. Rivera-Utrilla). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.08.015
394
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 3 e4 0 3
medium (Boreen et al., 2004; Latch et al., 2003; Packer et al., 2003). However, fewer data are available on their photochemical transformation (Boreen et al., 2003). The majority of pharmaceuticals are photo-active, i.e. able to absorb light. This is because their structures generally contain aromatic rings, heteroatoms, and other functional groups that make them prone to absorb UVevis radiation (direct photolysis) or to react with photosensitizing species capable of inducing pharmaceutical photodegradation in natural water (indirect photolysis). During direct photolysis, photon absorption gives rise to compounds in excited electronic states that are susceptible to chemical transformation. However, indirect or sensitized photolysis leads to the transformation of contaminants by energy transference or by chemical reactions with transitory species formed by the presence of light, such as hydroxyl radicals (HO), singlet oxygen (1O2), and triplet excited states of natural organic matter (3NOM*) (Schwarzenbach et al., 2003; Canonica et al., 1995; Canonica and Tratnyek, 2003; Gerecke et al., 2001; Zepp et al., 1985). Hence, the efficacy of direct photooxidation is governed by the contaminant absorption spectrum and the quantum yield of the process (V), whereas the dominant mechanism in indirect photolysis is the reaction between OH radicals and the micropollutant. Hence, addition of H2O2 during the photooxidation process accelerates the micropollutant removal rate, reducing the UV radiation required in comparison to direct photooxidation (Rosenfeldt and Linden, 2004); this is due to the generation of highly reactive radicals in H2O2 decomposition (Glaze et al., 1987). Nitroimidazole antibiotics were recently detected in waters at concentrations of 0.1e90.2 mg/L (Lindberg et al., 2004). They are widely used to treat infections caused by anaerobic and
protozoan bacteria (e.g., Trichomonas vaginalis and Giardia lamblia) in humans and animals and are added to chow for fish and fowl, leading to their accumulation in animals, fish-farm waters and, especially, meat industry effluents. Little is yet known about the capacity of current water treatment systems to remove nitroimidazoles (Wennmalm and Gunnarsson, 2005) but it is not expected to be very high given the complex chemical structure of these compounds. The objective of the present study was to analyze the efficacy of UV radiation in the direct photooxidation of nitroimidazoles. For this purpose, i) a kinetic study was conducted to determine the quantum yield of the process; and ii) the influence of the different operational variables (initial concentration of antibiotic, pH, presence of NOM components and chemical composition of water) was analyzed, and the time course of total organic carbon (TOC) concentration and toxicity during nitroimidazole photodegradation was studied.
2.
Experimental
2.1.
Reagents
All chemical reagents used (phosphoric acid, sodium hydroxide, hydrogen peroxide, atrazine, gallic acid, tannic acid, humic acid, acetonitrile, ammonium acetate, and nitroimidazoles) were high purity analytical grade reagents supplied by SigmaeAldrich. Ultrapure water was obtained using Milli-Q equipment (Millipore). Fig. 1 depicts the chemical structure and the acidity constant values of the nitroimidazoles selected for this study: Metronidazole (MNZ), Dimetridazole (DMZ), Tinidazole (TNZ) and Ronidazole (RNZ).
Fig. 1 e Chemical structure of the nitroimidazoles studied.
395
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 3 e4 0 3
natural waters (ground and surface waters) and wastewaters from the Motril (Granada, Spain) drinking water treatment (DWTP) and wastewater treatment (WWTP) plants, respectively. After characterization and filtering, water samples were kept refrigerated (T ¼ 297 K) until use.
1.0
[M]/[M] o
0.8
0.6
0.4
0.2
0.0 0
50
100
150
200
t (min) Fig. 2 e Removal kinetics of the four nitroimidazoles by means of UV photodegradation. [nitroimidazole]o [ 200 mM, pH [ 5e7, T [ 298 K; (>), MNZ; (,), DMZ; (6), TNZ; (B), RNZ.
2.2.
UV irradiation experimental device
The photoreactor used for nitroimidazoles photodegradation was equipped with a low-pressure mercury lamp (Heraeus Noblelight model TNN 15/32, nominal power 15 W). The problem solution is set in 6 quartz tubes with 1 cm diameter and 30 mL placed around and equidistant from the Hg lamp. These tubes are immersed in bidistilled water in continuous recirculation for temperature control, using a Frigiterm ultrathermostat, with a magnetic agitation system in each tube. In each experiment, after stabilizing the lamp and controlling the temperature (298 K), the photoreactor was turned on, and aliquots were withdrawn from the reactor at different time intervals in order to assess: i) nitroimidazole concentration, ii) total organic carbon (TOC), and iii) toxicity of the photodegradation products. The substrate concentration is higher than it was found on environmental samples. This high concentration was used to follow better its analysis by HPLC method. The influence of the presence of organic matter was analyzed by adding gallic acid (GAL), tannic acid (TAN), or humic acid (HUM) to the solution in some experiments. The concentration of these acids (components of natural organic matter) ranged from 10e80 mg/L. Each photooxidation experiment was repeated three times.
2.3.
Water sampling and characterization
The influence of the chemical composition of water on the direct photodegradation of nitroimidazoles was studied in
2.4.
Analytical methods
2.4.1.
Determination of nitroimidazole concentration
Chromatographic follow-up of nitroimidazole concentrations was done using a WATERS ALLIANCE 2690a analytical HPLC with WATERS M-996 photodiode detector and automatic injector with capacity for 120 flasks. The chromatographic column was Nova-Pak C18 Cartridge, particle size 4 mm and 3.9 150 mm inner diameter. The mobile phase used was a buffer solution of pH 4.3 with 96% 5.0 mM ammonic acetate and 4% acetonitrile in isocratic mode at a 1 mL min1 flow.
2.4.2.
Determination of atrazine concentration
Atrazine was used as actinometer to determine the radiant energy of the lamp (Canonica et al., 1995). Atrazine concentration was determined by high performance liquid chromatography (HPLC) using a chromatographic column with the same characteristics as in the nitroimidazole determination. The mobile phase was a buffer solution of pH 4.5, with 50% 2.5 mM ammonium acetate and 50% acetonitrile in isocratic mode and a flow of 1 mL min1.
2.4.3.
Determination of total organic carbon concentration
TOC was determined by using a Shimadzu V-CSH equipment with ASI-V autosampler.
2.4.4.
Toxicity determination
A LUMISTOX 300 system was used to measure toxicity, based on the standardized biotest (DIN/EN/ISO 11348-2) of Vibrio fischeri bacteria inhibition (NRRL B-11177). The measurement is based on inhibition of the luminosity intensity of marine bacteria Vibrio fisheri. Toxicity is expressed as percentage inhibition after 15 min of exposure. The results presented in this manuscript were obtained considering tree measurements of the same sample and the error associated to these data is between 5 and 10%.
3.
Results and discussion
3.1. Direct photodegradation of nitroimidazoles: quantum yields; time course of total organic carbon and toxicity Fig. 2 depicts the kinetics of direct nitroimidazole photodegradation with low-pressure Hg lamp (254 nm). It shows
Table 1 e Parameters obtained from direct irradiation at 254 nm of the four nitroimidazoles. Nitroimidazole MNZ DMZ TNZ RNZ
3 (m2 mol1) 209.72 224.45 233.92 226.13
k 104 (s1) 1.72 1.67 1.09 1.18
0.09 0.20 0.09 0.18
V 104 (mol Eins1) 34.7 31.5 19.6 22.1
1.8 3.9 1.7 3.4
k0 E (m2 Eins1) 1.68 1.63 1.06 1.15
0.09 0.20 0.09 0.18
R254 (%) 0.14 0.14 0.09 0.10
0.01 0.02 0.01 0.01
396
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 3 e4 0 3
a
25
50
b
40
% Inhibition
TOC (mg/L)
20 15 10 5
30 20 10
0
0
0
50
100
150
200
0
t (min)
50
100
150
200
t (min)
Fig. 3 e Time course of the concentration of total organic carbon (a) and toxicity (b) during UV irradiation of the four nitroimidazoles in ultrapure water. [nitroimidazole]o [ 200 mM, pH [ 5e7, T [ 298 K; (>), MNZ; (,), DMZ; (6), TNZ; (B), RNZ.
slightly higher removal rates for MNZ and DMZ than for TNZ and RNZ. At 3 h of treatment, the concentration of all nitroimidazoles was reduced by 70e85%. A pseudo-first order kinetic model was applied to the above results to determine the nitroimidazole photodegradation rate constant. Equation (1) (Sharpless and Linden, 2003) was then used to calculate the quantum yield (V) for each nitroimidazole. Table 1 shows the results obtained. Fl ¼
kl 2:303$El $3l
(1)
where kl is the photodegradation rate constant (s1); El is the rate of energy emitted, corresponding to the photon flow emitted by the lamp (E s1 m2); 3l is the molar absorption coefficient at the wavelength in question (m2 mol1); and Vl is the quantum yield (mol E1).
0.05
-1
Ф (mol·Eins ) · 10
0.04
0.03
0.02
The rate of energy irradiated by the lamp was determined by actimometry, using a solution of 5 mM atrazine as actinometer (Canonica et al., 1995) and obtaining an energy of 1.027$104 E s1 m2 for the lamp used. For this purpose, the quantum yield of atrazine was considered to be 0.046 mol E1 and the molar absorption coefficient to be 386 m2 mol1 at 254 nm wavelength (Hessler et al., 1993). Table 1 shows the values of the study parameters, showing a very low quantum yield for all four nitroimidazoles (values from 34.7$104 for MNZ to 19.6$104 for TNZ), responsible for a low efficacy of the quantum process in nitroimidazole phototransformation. This low efficacy (Table 1), alongside the results in Fig. 2, confirm the need for long irradiation times to achieve their complete removal. Except for MNZ, no data on quantum yields of nitroimidazoles have been reported in the literature; nonetheless, the values obtained are similar to those reported by other authors for pharmaceutical compounds (Rosenfeldt and Linden, 2004; Canonica et al., 2008; Sa´nchez-Polo et al., 2007; Lo´pez et al., 2002; Andreozzi et al., 2003). The value of (34.7 1.8)$104 mol E1 obtained for MNZ is close to the value of 0.0033 mol E1 reported by Shemer et al. (2006). For comparative purposes, it is essential to consider the apparent photodegradation rate constant normalized by the energy of the lamp, k0 E (m2 E1), using Equation (2). This constant is independent of the fluctuations in energy irradiated by the lamp and permits direct comparisons among phototransformation rate constants obtained with different photoreactors (Canonica et al., 2008). 0
kE ¼
0.01
0.00 0
500
1000
1500
2000
[nitroimidazole]o (μM) Fig. 4 e Quantum yield of the photodegradation process as a function of initial nitroimidazole concentration. pH [ 5e7, T [ 298 K; (>), MNZ; (,), DMZ; (6), TNZ; (B), RNZ.
kl El
(2)
where kl (s1) is the photodegradation rate constant of contaminants used in Equation (1), and El (Einstein$s1 m2) is the radiant energy emitted by the lamp at a wavelength of 254 nm, calculated by actinometry (Canonica et al., 1995). The k0 E values obtained are shown in Table 1. Table 1 also shows the percentage nitroimidazole removal for an irradiation dose of 400 J m2 (R254). This parameter determines the applicability of UV radiation in nitroimidazole photodegradation under the real conditions of a water treatment plant. An irradiation dose
397
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 3 e4 0 3
Table 2 e Specific constants of the photodegradation prediction model for the four nitroimidazoles. Nitroimidazole MNZ DMZ TNZ RNZ
m
n
0.3293 1.0143 1.0938 0.1135
0.5857 0.7328 0.8483 0.4742
of 400 J m2 was selected as reference because it is the minimum value recommended by different European bodies ¨ Norm, 2001; Cabaj et al., for water disinfection (DVGW, 1997; O 1996). With the present lamp, the time required to reach an irradiation dose of 400 J m2 is 8.3 s, and the radiant energy is equivalent to 8.49$104 E m2 for a wavelength of 254 nm. Hence, the percentage of nitroimidazole removed with this irradiation dose, R254, can be calculated by means of Equation (3). 0 4 R254 ¼ 1 eðkE 8:49$10 Þ
(3)
0
where k E is the apparent photodegradation rate constant normalized by the energy of the lamp Equation (2). As shown in Table 1, a very low percentage of nitroimidazole removal was achieved at the minimum irradiation dose (400 J m2) (Table 1). Hence, the usual dose for water disinfection is not adequate to remove this type of pharmaceutical, which requires higher doses of UV irradiation or longer exposure times to be eliminated by direct photolysis. Data in Table 1 demonstrates the low performance of the nitroimidazole photodegradation process, decreasing in the order: MNZ > DMZ > RNZ > TNZ.
5.0
-1
ε
3.2. Influence of the different operational variables on nitroimidazole photodegradation This section analyzes the effects on nitroimidazole photodegradation performance of the operational parameters studied: nitroimidazole concentration, medium pH and presence of natural organic matter.
3.2.1.
Influence of nitroimidazole concentration
Fig. 4 depicts the quantum yield of nitroimidazoles as a function of their concentration, showing that the initial concentration
300
5.0
250
4.0
300
ε
250
200
200 3.0
3.0
pka1 150
2
150 2.0
2.0
k’E
1.0
50
0.0
0 0
2
4
6
k’E
100
pka1
8
250
4.0 -1
0 0
300
ε
50
0.0
10
5.0
100
1.0
2
4
6
8
10
5.0
300
ε
4.0
250 200
3.0
2
150
pka1
2.0
k’ E
100 50 0
0.0 2
4
6
pH
150
k’ E
2.0
1.0
0
pka1
8
10
100 1.0
ε (m2·mol-1)
200 3.0
ε (m2·mol-1)
k'E (m ·Eins )
4.0
k'E (m ·Eins )
Two key parameters of the efficacy of any treatment system are: i) the concentration of total organic carbon and ii) the toxicity of the oxidation by-products generated. Fig. 3 shows the time course of TOC and toxicity values during the photodegradation of the four nitroimidazoles under study. According to the results in Fig. 3a, although the nitroimidazole concentration considerably decreases with radiation time (Fig. 2), oxidation by-products do not transform into CO2 to the desired extent. Consequently, they generate fractions of smaller molecular weight than the original nitroimidazole and maintain the TOC concentration constant throughout the treatment time. Moreover, as shown in Fig. 3b, the mixture of by-products generated during nitroimidazole photodegradation sometimes shows higher toxicity than the original product. Hence, nitroimidazole photodegradation may give rise to compounds that are pharmacologically active and have higher toxicity than the original compound. We highlight the results obtained for TNZ, which show a substantial increase in toxicity at 3 h of treatment (Fig. 3b).
50
0.0
0 0
2
4
6
8
10
pH
Fig. 5 e Molar absorption coefficient (3) and apparent photodegradation constant normalized by the energy emitted by the lamp (k0 E) as a function of solution pH for the four nitroimidazoles studied. T [ 298 K (>), MNZ; (,), DMZ; (6), TNZ; (B), RNZ.
398
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 3 e4 0 3
Table 3 e Kinetic parameters of direct phototransformation of ionic (j [ 1) and neutral (j [ 2) species of nitroimidazoles. Nitro imidazole pka1 31 (m2 mol1) k0E1 (m2 Eins1) V1 103 (mol Eins1) 32 (m2 mol1) k0E2 (m2 Eins1) V2 103 (mol Eins1) MNZ DMZ TNZ RNZ
2.58 2.81 2.30 1.32
271.18 288.12 178.88 171.31
3.78 1.33 0.28 0.31
6.05 2.01 0.69 0.79
affects the quantum yield and therefore the photodegradation rate Equation (1). The decrease in photodegradation rate in nitroimidazoles with the increase of their concentration is related to the energy absorbed by each nitroimidazole molecule. Hence, given that the radiation energy deposited in the medium per unit volume is constant, nitroimidazole molecules can accept more radiant energy at lower concentrations, explaining the behavior observed. The equations that predict the quantum yield of each nitroimidazoles are based on the data in Fig. 4. These are potential-type equations Equation (4) and allow the quantum yield of the photochemical process to be obtained at the temperature and pH shown in the footnote of Fig. 4. n
F254 ¼ m$½nitroimidazolo
(4)
where V254 is the quantum yield (mol E1) at a wavelength of 254 nm, m and n are the specific constants obtained from Fig. 4 for each nitroimidazole (Table 2) and [nitroimidazole]o is the initial concentration of the nitroimidazole. Once the quantum yield is known, the photodegradation rate constant of any of the four nitroimidazoles can be determined as a function of their initial concentration by applying Equation (1). These results are of great interest from
209.72 224.45 233.91 226.13
2.10 1.73 1.42 1.80
4.35 3.35 2.64 3.46
an industrial standpoint because they allow pre-treatment prediction of the effectiveness of UV radiation to remove nitroimidazoles.
3.2.2.
Influence of solution pH
The influence of the solution pH in nitroimidazole photodegradation was analyzed in experiments with pH values ranging from 2e9. Based on the pKa of each nitroimidazole (Fig. 1) and the distribution of species that each presents, the four nitroimidazoles are in their cationic and/or neutral form in the pH range studied. Fig. 5 shows the variation in global molar absorption coefficient (3) and global photodegradation rate constant (k0 E) as a function of the solution pH for each nitroimidazole. Interestingly, the changes observed are determined by the pKa of the antibiotic. According to Equation (1), an increase in the molar absorption coefficient (3) would produce an increase in the k0 E value, but the results show no general tendency that describes the influence of pH on the 3 and k0 E values of these nitroimidazoles. In fact, the variation in trends among these nitroimidazoles differentiates four distinct behaviors: a) MNZ: increase in 3 and k0 E at pH < 4
Fig. 6 e Chemical structure of gallic acid (a), humic acid (b), and tannic acid (c).
399
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 3 e4 0 3
60
3.5
[NOM]= 0 mg/L
2.5 -1
k·10 (s )
30
2.0
4
Transmittance %
3.0 45
1.5 1.0
15
0.5 0
0.0 0
20
40
60
80
0
1
[NOM] (mg/L)
b) DMZ: increase in 3 and decrease in k0 E at pH < 5 c) TNZ: decrease in 3 and k0 E at pH < 5 d) RNZ: 3 and k0 E remain virtually unchanged in the studied pH range. The variations in these parameters as a function of the medium pH are similar to those observed for other pharmaceutical compounds (Canonica et al., 2008). This global behavior was studied in greater depth by quantitative analysis of the photodegradation of each nitroimidazole species in the medium as a function of the pH, using Equation (5), which is obtained by combining Equations (1) and (3). 0
kEj ¼ 2:303$3j $Fj
(5)
where k0 Ej is the apparent rate constant for each nitroimidazole species “j” present in the solution. The contribution of each of each species towards total k0 E can be represented by Equation (6): 0
X
0
aj $kEj
(6)
j
Table 4 e MNZ photodegradation rate constants as a function of the concentration of acid present. GAL (mg L1) 0 9.8 18.1 36.0 76.4 0 0 0 0 0 0 0 0
TAN (mg L1)
HUM (mg L1)
0 0 0 0 0 15.9 27.2 43.1 79.0 0 0 0 0
0 0 0 0 0 0 0 0 0 9.5 23.2 36.0 79.2
3
4
[NOM]/[MNZ]o
Fig. 7 e Transmittance of NOM components as a function of concentration at l [ 254 nm. (>), GAL; (,), TAN; (6), HUM.
kE ¼
2
k 104 (s1) 2.36 3.08 2.87 1.97 0.73 1.81 1.50 1.16 0.60 1.84 1.53 1.26 0.66
0.17 0.11 0.15 0.11 0.04 0.05 0.15 0.05 0.03 0.12 0.04 0.04 0.02
Fig. 8 e MNZ photodegradation rate constant as a function of the relationship between NOM and MNZ concentrations. [MNZ]o [ 20 mg/L, T [ 298 K; (>), GAL/MNZ; (,), TAN/ MNZ; (6), HUM/MNZ.
where aj represents the molar fraction of each of the species (Sjaj ¼ 1). Once the species distribution and pKa of each nitroimidazole is known, the rate constant and molar absorption coefficient can be calculated by means of linear regression: 0 0 0 0 0 0 kE ¼ ð1 a2 Þ$kE1 þ a2 $kE2 ¼ kE1 þ kE2 kE1 a2
(7)
3 ¼ ð1 a2 Þ$31 þ a2 $32 ¼ 31 þ ð32 31 Þa2
(8)
0 kEj
After calculating and 3j for the ionic ( j ¼ 1) and neutral ( j ¼ 2) species of each nitroimidazole, the quantum yield of the corresponding species can be obtained by means of Equation (9). 0
Fj ¼
kEj 2:303$3j
(9)
Table 3 shows the results obtained by applying these equations. Using these parameters, the values of k0 E and V can be determined for any pH value in the studied range. As shown in Table 3, the rate constant (k0 E) and quantum yields (V) values obtained for the species of each nitroimidazole vary by only one order of magnitude, with minimum values for the ionic form of TNZ and maximum values for the ionic form of MNZ.
3.2.3.
Influence of the presence of gallic, tannic or humic acid
The influence of natural organic matter (NOM) components during nitroimidazole photodegradation was analyzed with experiments of MNZ photodegradation in the presence of three components of NOM. MNZ was selected for these experiments because it is the most representative nitroimidazole and the one most frequently detected in waters. The concentration of NOM in waters ranges from approximately 0.1 mg to >100 mg of total organic carbon (TOC) per liter, depending on their origin (Artinger et al., 2000). Fig. 6 shows the chemical structures of GAL, TAN, and HUM, the three NOM components selected for the study. GAL is
400
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 3 e4 0 3
Table 5 e Chemical characteristics of the waters used. Water
pH
Ultrapure Surface Ground Waste
6.8 8.1 7.5 7.8
[HCO 3] 1
[SO2 4 ] 1
[NO 3] 1
0.0 5.2 8.9 10.5
0 104 150 257
0.0 1.5 3.7 11.7
a
T (%) TOC (mg L1) (meq L ) (mg L ) (mg L ) 0.0 9.8 17.4 25.2
100.0 97.4 97.8 60.1
Table 6 e Parameters obtained from UV irradiation of MNZ in waters with different chemical compositions. Water
k 104 (s1)
k0 E (m2 Eins1)
Ultrapure Surface Ground Waste
1.72 0.09 1.73 0.09 2.08 0.15 1.66 0.08
1.68 0.09 1.68 0.09 2.03 0.15 1.61 0.08
R254 (%) 0.14 0.14 0.17 0.14
0.01 0.01 0.01 0.01
k0 E: photodegradation rate constant normalized by the energy of the lamp. R254: percentage removal for an irradiation dose of 400 J m2.
a: Transmittance (%) at 254 nm.
considered the basic structural unit of NOM, and HUM is considered its main component. HUM has a high molecular weight and a complex structure with a large number of aromatic rings and oxygenated functional groups (Steelink, 2002; Choudhry, 1984; Cooper et al., 2008). Knowledge of the amount of light absorbed by the different components of NOM is necessary to study the influence of humic matter during UV irradiation of MNZ and its effects on the removal rate. For this purpose, the transmittance of the three acids was determined at the studied concentrations for a wavelength of 254 nm (Fig. 7). According to the transmittance data in Fig. 7, the amount of UV light absorbed by NOM can be known, hence reducing the number of photons that reach the MNZ molecules (Canonica et al., 2008). The little transmittance shown by the three acids at high concentrations will have a very negative effect on the direct photodegradation of MNZ. Table 4 shows the MNZ photodegradation rate constant values in the presence of different concentrations of the selected acids; the MNZ photodegradation rate increases at small concentrations of GAL but decreases at any of the TAN and HUM concentrations studied. Fig. 8 plots the photodegradation rate constant against the relationship between concentrations of NOM constituents and MNZ. The presence of GAL has a markedly different effect on MNZ degradation in comparison to the presence of TAN or HUM. MNZ removal is favored only at a GAL/MNZ
concentration ratio <1.4. The data in Table 4 and Fig. 8 show that low concentrations of GAL in the medium favor MNZ photodegradation, despite the lesser transmittance of GAL versus TAN or HUM (Fig. 7). Therefore, GAL may act as a promoter of OH$ radicals, which oxidizes MNZ molecules. In contrast, the behavior in the presence of TAN and HUM suggests a predominant OH$ radical inhibition effect, due to their complex structure (Fig. 6) and the high reactivity of NOM against OH$ radicals (kOH$ ¼ 108 M1s1) (Basfar et al., 2005; Buxton et al., 1988). An increase in the concentration of all three acids decreases the MNZ removal rate constant to very similar levels, k < 0.73$104 s1, (Table 4 and Fig. 8), due to the little or null transmittance of GAL, TAN and HUM at high concentrations (Fig. 7). The different effects of NOM components on contaminant photodegradation have been studied by various authors (Zepp et al., 1985, 1981a, 1981b; Choudhry, 1984; Simmons and Zepp, 1986; Van Noort et al., 1988; Zheng and Ye, 2001; Zhan et al., 2006; Xu et al., 2007; Garbin et al., 2007). Cooper and Zika (1983) were the first to report that exposure of natural waters to high energy solar radiation gives rise to photoproduction of H2O2 (Cooper et al., 1988), which is an efficient OH radical producer under the action of UVevis light (Glaze et al., 1987).
3.3. Applicability of UV radiation for nitroimidazole degradation in water with different chemical compositions
1.0
[MNZ]/[MNZ] o
0.8
0.6
0.4
0.2
0.0 0
50
100
150
200
t (min) Fig. 9 e Influence of the chemical composition of water on the MNZ photodegradation rate. [MNZ]o [ 200 mM, pH [ 6e8, T [ 298 K; (>), ultrapure water; (B), surface water; (6), groundwater; (,), wastewater.
The applicability of UV radiation to remove nitroimidazole from water was studied by analyzing the influence of water chemical composition during MNZ photodegradation. Experiments were conducted with water of different origin and chemical composition: ultrapure, groundwater and wastewater. Table 5 depicts the results of their chemical characterization. Fig. 9 shows the results obtained from UV irradiation of MNZ in the different waters studied. The differences in MNZ degradation rate among the different waters are not very marked. There is a slight decrease in the wastewater samples, largely to the lower transmittance (T ¼ 60%) in this type of water, causing absorption of UV radiation and considerably reducing the number of photons reaching the nitroimidazole. In this case, the organic matter present in wastewater acts as a filter of UV radiation, reducing the efficacy of the treatment to remove MNZ from the medium. Table 6 shows the kinetic parameters and the percentage removal (R254) obtained by UV irradiation of MNZ in the studied waters. MNZ photodegradation rate constant values
401
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 3 e4 0 3
a
b
50
20
% Inhibition
TOC (mg/L)
40
25
30
20
15
10
5
10
0
0 0
20
40
60
80
100
% MNZ removed
0
20
40
60
80
100
% MNZ removed
Fig. 10 e Time course of total organic carbon (a) and toxicity by Vibrio Fischeri inhibition (b) as a function of %MNZ removed during UV irradiation in waters with different chemical compositions: (,), surface water; (>), groundwater; (6), wastewater; (B), ultrapure. [MNZ]o [ 200 mM, pH [ 6e8, T [ 298 K.
(Table 6) are similar among ultrapure, surface and wastewaters. The rate is slightly higher in groundwater, suggesting a small indirect photodegradation similar to that observed in MNZ photodegradation in the presence of low concentrations of GAL. These results show that the presence of NOM with low humification rate, e.g., GAL and hydrobenzoic acids, favors the indirect photodegradation of contaminants (Fukushima and Tatsumi, 1999). Consequently, there is a need for a wider study to characterize the humic acids present in natural waters and to determine the role of each in contaminant photodegradation. In addition, the presence in this type of water of a high concentration of ions susceptible of transformation into oxidant radicals (sulfates, nitrates) that can react with MNZ, would also explain this increase in MNZ degradation rate in groundwaters. Fig. 10 depicts the time course of TOC values and toxicity during MNZ removal in ultrapure, surface, groundwater and wastewater. According to these results, UV radiation does not significantly reduce the TOC concentration during the 3 h irradiation in any of the studied waters; slight TOC reduction is only detected in wastewater and groundwater. Hence, although there is a considerable reduction in nitroimidazole concentration (Fig. 9), the degradation compounds are not mineralized to the desired extent and remain in the medium. Moreover, Fig. 10b shows that the mixture of by-products generated during MNZ photodegradation has a highly variable toxicity depending on the type of water studied. Interestingly, regardless of the type of water studied, minimum toxicity values are reached when around 40% of the MNZ has been removed.
4.
Conclusions
Quantum yields obtained for the four nitroimidazoles are very low R254 values obtained show that the dose habitually used for the disinfection of waters is not adequate to remove this type of pharmaceutical compounds. The time course of TOC and of toxicity during direct photodegradation of nitroimidazoles, in both ultrapure and real
waters, shows that oxidation by-products do not transform into CO2 to the desired extent. Therefore, they generate fractions of lower molecular weight than the original molecule, maintaining a constant TOC concentration throughout the treatment time and possibly giving rise to pharmacologically active compounds with higher toxicity than the original nitroimidazole. The concentration of nitroimidazole has a major effect on its photodegradation rate. The study of the influence of pH on the values of parameters 3 and k0 E shows no general tendency for the behavior of nitroimidazoles as a function of the pH. The rate constants (k0 E) and quantum yields (V) obtained for the different species of each nitroimidazoles vary by only one order of magnitude, with minimum values for the ionic form of TNZ and maximum values for the ionic form of MNZ. The NOM components GAL, TAN, and HUM may act as promoters and/or inhibitors of OH$ radicals generated by H2O2 photoproduction. Results show that the presence of GAL has a markedly different effect on the MNZ degradation rate from that of TAN or HUM, with an increase in this rate at low GAL concentrations. These results appear to show that, under these conditions, GAL mainly acts as a promoter of OH$ radicals, which oxidize MNZ molecules. In contrast, the presence of TAN or HUM decreases MNZ degradation rate, suggests a predominant effect of their OHradical inhibiting capacity, due to their complex structure and high reactivity against OHradicals. Differences in MNZ degradation rate among the studied waters, which have different chemical compositions, are not very marked, although there is a slight decrease in wastewaters, mainly because of the UV radiation filter effect of this type of water.
Acknowledgments The authors are grateful for the financial support provided by MEC-DGI, FEDER (Project: CTQ2007-67792-C02-01/PPQ), Junta de Andalucı´a (Project: RNM3823).
402
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 3 e4 0 3
references
Andreozzi, R., Raffaele, M., Nicklas, P., 2003. Pharmaceuticals in STP effluents and their solar photodegradation in aquatic environment. Chemosphere 50 (10), 1319e1330. Artinger, R., Buckau, G., Geyer, S., Fritz, P., Wolf, M., Kim, J.I., 2000. Characterization of groundwater humic substances: influence of sedimentary organic carbon. Applied Geochemistry 15 (1), 97e116. Basfar, A.A., Khan, H.M., Al-Shahrani, A.A., Cooper, W.J., 2005. Radiation induced decomposition of methyl tert-butyl ether in water in presence of chloroform: kinetic modelling. Water Research 39 (10), 2085e2095. Boreen, A.L., Arnold, W.A., McNeill, K., 2004. Photochemical fate of sulfa drugs in then aquatic environment: sulfa drugs containing five-membered heterocyclic groups. Environmental Science and Technology 38 (14), 3933e3940. Boreen, A.L., Arnold, W.A., McNeill, K., 2003. Photodegradation of pharmaceuticals in the aquatic environment: a review. Aquatic Sciences 65 (4), 320e341. Buxton, G.V., Greenstock, C.L., Helman, W.P., Ross, A.B., 1988. Critical review of rate constants for reactions of hydrated electrons, hydrogen atoms and hydroxyl radicals in aqueous solution. Journal of Physical Chemistry Reference Data 17 (2), 513e886. Cabaj, A., Sommer, R., Schoenen, D., 1996. Biodosimetry: model calculations for u.v. water disinfection devices with regard to dose distributions. Water Research 30 (4), 1003e1009. Canonica, S., Meunier, L., von Gunten, U., 2008. Phototransformation of selected pharmaceuticals during UV treatment of drinking water. Water Research 42 (12), 121e128. Canonica, S., Tratnyek, P.G., 2003. Quantitative structure-activity relationships for oxidation reactions of organic chemicals in water. Environmental Toxicology and Chemistry 22 (8), 1743e1754. Canonica, S., Jans, U., Stemmler, K., Hoigne´, J., 1995. Transformation kinetics of phenols in water: photosensitization by dissolved natural organic material and aromatic ketones. Environmental Science and Technology 29 (7), 1822e1831. Choudhry, G.G., 1984. Humic substances, structural aspects, and photophysical, photochemical and free radical characteristics. In: The Handbook of Environmental Chemistry: The Natural Environment and the Biogeochemical Cycles, vol. 1C, pp. 1e24. Cooper, W.J., Song, W., Gonsior, M., Kalnina, D., Peake, B.M., Mezyk, S.P., 2008. Recent advances in structure and reactivity of dissolved organic matter in natural waters. Water Science and Technology: Water Supply 8, 615e623. Cooper, W.J., Zika, R.G., Petasne, R.G., Plane, J.M.C., 1988. Photochemical formation of H2O2 in natural waters exposed to sunlight. Environmental Science and Technology 22 (10), 1156e1160. Cooper, W.J., Zika, R.G., 1983. Photochemical formation of hydrogen peroxide in surface and ground waters exposed to sunlight. Science 220 (4598), 711e712. DVGW, 1997. W 294. UVeDesinfektionsanlagen fu¨r die Trinkwasserversorgung e Anforderungen und Pru¨fung. Fukushima, M., Tatsumi, K., 1999. Photosensitized production of hydrogen peroxide in the aqueous system containing dissolved organic carbon. Toxicology Environmental Chemistry 69 (3e4), 435e448. Garbin, J.R., Milori, D.M.B.P., Simo˜es, M.L., da Silva, W.T.L., Neto, L.M., 2007. Influence of humic substances on the photolysis of aqueous pesticide residues. Chemosphere 66 (9), 1692e1698.
Gerecke, A.C., Canonica, S., Muller, S.R., Scharer, M., Schwarzenbach, R.P., 2001. Quantification of dissolved natural organic matter (DOM) mediated phototransformation of phenylurea herbicides in lakes. Environmental Science and Technology 35 (19), 3915e3923. Glaze, W.H., Kang, J., Chapin, D.H., 1987. Chemistry of water treatment processes involving ozone, hydrogen peroxide and ultraviolet radiation. Ozone: Science and Engineering 9, 335e352. Hessler, D.P., Gorenflo, V., Frimmel, F.H., 1993. Degradation of aqueous atrazine and metazachlor solutions by UV and UV/H2O2einfluence of pH and herbicide concentration. Acta Hydrochimica Hydrobiology 21 (4), 209e214. 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 (1), 3e22. Kang, S.J., Allbaugh, T.A., Reynhout, J.W., Erickson, T.L., Olmstead, K.P., Thomas, L., Thomas, P., 2004. Selection of an ultaviolet disinfection system for a municipal wastewater treatment plant. Water Science and Technology 50, 163e169. Latch, D.E., Stender, B.L., Packer, J.L., Arnold, W.A., McNeill, K., 2003. Photochemical fate of pharmaceuticals in the environment: cimetidine and ranitidine. Environmental Science and Technology 37 (15), 3342e3350. Lazarova, V., Savoye, P., 2004. Technical and sanitary aspects of wastewater disinfection by UV irradiation for landscape irrigation. Water Science and Technology 50, 203e209. Lindberg, R., Jarnheimer, P., Olsen, B., Johansson, M., Tysklind, M., 2004. Determination of antibiotic substances in hospital sewage water using solid phase extraction and liquid chromatography/mass spectrometry and group analogue internal standards. Chemosphere 57, 1479e1488. Lo´pez, A., Bozzi, A., Mascolo, G., Ciannarella, R., Passino, R., 2002. UV and H2O2/UV degradation of a pharmaceutical intermediate in aqueous solution. Annali di Chimica 92 (1e2), 41e51. ¨ Norm, 2001. M 5873e1. Anlagen zur Desinfektion von Wasser O mittels Ultraviolett-Strahlen. Packer, J.L., Werner, J.J., Latch, D.E., McNeill, K., Arnold, W.A., 2003. Photochemical fate of pharmaceuticals in the environment: naproxen, diclofenac, clofibric acid, and ibuprofen. Aquatic Sciences 65 (4), 342e351. Rosenfeldt, E.J., Linden, K.G., 2004. Degradation of endocrine disrupting chemicals bisphenol A, ethinyl estradiol, and estradiol during UV photolysis and advanced oxidation processes. Environmental Science and Technology 38 (20), 5476e5483. Sa´nchez-Polo, M., Rivera-Utrilla, J., Me´ndez-Dı´az, J.D., Canonica, S., von Gunten, U., 2007. Photooxidation of naphthalenesulfonic acids: comparison between processes based on O3, O3/activated carbon and UV/H2O2. Chemosphere 68 (10), 1814e1820. Schwarzenbach, R.P., Gschwend, P.M., Imboden, D.M., 2003. Environmental Organic Chemistry. John Wiley & Sons, New York, p. 1311. Sharpless, C.M., Linden, K.G., 2003. Experimental and model comparisons of low- and medium-pressure Hg lamps for the direct and H2O2 assisted UV photodegradation of N-nitrosodimethylamine in simulated drinking water. Environmental Science and Technology 37 (9), 1933e1940. Shemer, H., Kunukcu, Y.K., Linden, K.G., 2006. Degradation of the pharmaceutical metronidazole via UV, fenton and photofenton processes. Chemosphere 63 (2), 269e276. Simmons, M.S., Zepp, R.G., 1986. Influence of humic substances on photolysis of nitroaromatic compounds in aqueous systems. Water Research 20 (7), 899e904. Steelink, C., 2002. Investigating humic acids in soils. Anal. Chem. 74 (11), 327Ae333A.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 3 e4 0 3
Van Noort, P., Lammers, R., Verboom, H., Wondergem, E., 1988. Rates of triplet humic acid sensitized photolysis of hydrophobic compounds. Chemosphere 17 (1), 35e38. Wennmalm, A., Gunnarsson, B., 2005. Public health care management of water pollution with pharmaceuticals: environmental classification and analysis of pharmaceutical residues in sewage water. Journal of Drug Information 39, 291e297. Xu, G., Li, F., Wang, Q., 2007. Effect of humic acids on photodegradation of chloroacetanilide herbicides under UV irradiation. Journal of Environmental Science and Health. Part. B, Pesticides, Food Contaminants, and Agricultural Wastes 42 (2), 165e171. Zepp, R.G., Schlotzhauer, P.F., Sink, R.M., 1985. Photosensitized transformations involving electronic energy transfer in natural waters: role of humic substances. Environmental Science and Technology 19 (1), 74e81.
403
Zepp, R.G., Baughman, G.L., Schlotzhauer, P.F., 1981a. Comparison of photochemical behavior of various humic substances in water: I. Sunlight induced reactions of aquatic pollutants photosensitized by humic substances. Chemosphere 10 (1), 109e117. Zepp, R.G., Baughman, G.L., Schlotzhauer, P.F., 1981b. Comparison of photochemical behavior of various humic substances in water: II. Photosensitized oxygenations. Chemosphere 10 (1), 119e126. Zhan, M., Yang, X., Xian, Q., Kong, L., 2006. Photosensitized degradation of bisphenol A involving reactive oxygen species in the presence of humic substances. Chemosphere 63 (3), 378e386. Zheng, H., Ye, C., 2001. Photodegradation of acetochlor and butachlor in waters containing humic acid and inorganic ion. Bulletin Environmental Contamination Toxicology 67 (4), 601e608.
