No title

WATER RESEARCH A Journal of the International Water Association

Editor-in-Chief

Associate Editors

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

Andrew Baker University of New South Wales Australia

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

Damien Batstone University of Queensland Australia Nicolas Bernet INRA France Guang-Hao Chen Hong Kong University of Science & Technology China Rob Eldridge Glen Waverley Australia Monica B. Emelko University of Waterloo Canada Tim Fletcher University of Melbourne Australia Rosina Girones University of Barcelona Spain Stephen Gray Victoria University Australia Kate Grudpan Chiang Mai University Thailand Peter Hillis AECOM Australia

Dionisis Mantzavinos Technical University of Crete Greece Yoshihiko Matsui Hokkaido University Japan Armand Maul Université de Metz France Max Maurer Eawag Switzerland How Yong Ng National University of Singapore Singapore Jong Moon Park Pohang University of Science & Technology Korea Susan Richardson US Environmental Protection Agency USA Miguel Salgot University of Barcelona Spain David Sedlak University of California, Berkeley USA Masahiro Takahashi Hokkaido University Japan Nathalie Tufenkji McGill University Canada Vayalam Venugopalan BARC Facilities India

Hong-Ying Hu Tsinghua University China

Hang-Shik Shin Korea Advanced Institute of Science and Technology Korea

Jae-Hong Kim Georgia Institute of Technology USA

Thomas Ternes Bundesanstalt für Gewässerkunde Germany

Günter Langergraber University of Natural Resources and Applied Life Sciences, Vienna (BOKU) Austria

Stefan Wuertz University of California, Davis USA

Shang-Lien Lo National Taiwan University Taiwan

Eduardo von Sperling Federal University of Minas Gerais Brazil Julie Zilles University of Illinois at Urbana-Champaign USA Anastasios Zouboulis Aristotle University of Thessaloniki Greece

Editorial Office

E-mail: [email protected]

Publication information: Water Research (ISSN 0043-1354). For 2012, volume 46 is scheduled for publication. Subscription prices are available upon ­request from the publisher or from the Elsevier Customer Service Department nearest you or from this journal’s website (http://www.elsevier.com/locate/watres). Further information is available on this journal and other Elsevier products through Elsevier’s website (http://www.elsevier.com). Subscriptions are accepted on a prepaid basis only and are entered on a calendar year basis. Issues are sent by standard mail (surface within Europe, air delivery outside Europe). Priority rates are available upon request. Claims for missing issues should be made within six months of the date of dispatch. Author Enquiries: For enquiries relating to the submission of articles (including electronic submission) please visit this journal’s homepage at http://www.elsevier.com/ locate/watres. Contact details for questions arising after acceptance of an article, especially those relating to proofs, will be provided by the publisher. You can track accepted articles at http://www.elsevier.com/trackarticle. You can also check our Author FAQs at http://www.elsevier.com/authorFAQ and/or contact Customer Support via http://support.elsevier.com. Orders, claims, and journal enquiries: Please contact the Elsevier Customer Service Department nearest you: St. Louis: Elsevier Customer Service Department, 3251 Riverport Lane, Maryland Heights, MO 63043, USA; phone: (877) 8397126 [toll free within the USA]; (+1) (314) 4478878 [outside the USA]; fax: (+1) (314) 4478077; e-mail: [email protected] Oxford: Elsevier Customer Service Department, The Boulevard, Langford Lane, Kidlington OX5 1GB, UK; phone: (+44) (1865) 843434; fax: (+44) (1865) 843970; e-mail: [email protected] Tokyo: Elsevier Customer Service Department, 4F Higashi-Azabu, 1-Chome Bldg, 1-9-15 Higashi-Azabu, Minato-ku, Tokyo 106-0044, Japan; phone: (+81) (3) 5561 5037; fax: (+81) (3) 5561 5047; e-mail: [email protected] Singapore: Elsevier Customer Service Department, 3 Killiney Road, #08-01 Winsland House I, Singapore 239519; phone: (+65) 63490222; fax: (+65) 67331510; e-mail: [email protected] Application for membership of International Water Association should be made to: Executive Director, IWA, Alliance House, 12 Caxton Street, London SW1H 0QS, U.K. (Tel.: +44 207 654 5500; Fax: +44 207 654 5555; e-mail: [email protected]; website: http://www.IWAhq.org.uk). Registered Charity (England) No. 289269. Individual membership is available from £30 upwards. For details contact IWA.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 e1 0

Available online at www.sciencedirect.com

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

Corrosion in drinking water pipes: The importance of green rusts a,
Joanna Swietlik *, Urszula Raczyk-Stanisławiak a, Paweł Piszora b, Jacek Nawrocki a a b


, Poland Department of Water Treatment Technology, Faculty of Chemistry, Adam Mickiewicz University, ul. Drzymały 24, 60-613 Poznan
, Poland Department of Chemistry of Materials, Faculty of Chemistry, Adam Mickiewicz University, Grunwaldzka 6, 60-780 Poznan

article info

abstract

Article history:

Complex crystallographic composition of the corrosion products is studied by diffraction

Received 29 April 2011

methods and results obtained after different pre-treatment of samples are compared. The

Received in revised form

green rusts are found to be much more abundant in corrosion scales than it has been

7 October 2011

assumed so far. The characteristic and crystallographic composition of corrosion scales

Accepted 9 October 2011

and deposits suspended in steady waters were analyzed by X-ray diffraction (XRD). The

Available online 25 October 2011

necessity of the examination of corrosion products in the wet conditions is indicated. The drying of the samples before analysis is shown to substantially change the crystallographic

Keywords:

phases originally present in corrosion products. On sample drying the unstable green rusts

Green rust

is converted into more stable phases such as goethite and lepidocrocite, while the content

Corrosion products

of magnetite and siderite decreases. Three types of green rusts in wet materials sampled

Drinking water

from tubercles are identified. Unexpectedly, in almost all corrosion scale samples signifi-

Distribution system

cant amounts of the least stable green rust in chloride form was detected. Analysis of

Steady water

corrosion products suspended in steady water, which remained between tubercles and

Tubercles

possibly in their interiors, revealed complex crystallographic composition of the sampled material. Goethite, lepidocrocite and magnetite as well as low amounts of siderite and quartz were present in all samples. Six different forms of green rusts were identified in the deposits separated from steady waters and the most abundant was carbonate green rust GR(CO2
3 )(I). ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Drinking water distribution systems in several countries have a large amount of iron and steel pipes that are subjected to corrosion causing economic, hydraulic and aesthetic effects, including water leaks, increasing pumping costs, and corrosion products build up. Estimated overall percentage of steel and iron tubes in working networks amounted 53% in Poland (Kwietniewski et al., 2011), 56,6% in USA (AWWA, 2004) and 67,2% in Italy (Veschetti et al., 2010). The highest share of iron

and steel pipes were noted in large cities, e.g. 91% in Warsaw and 93% in Innsbruck (Kwietniewski et al., 2011). Corrosion in steel or cast iron water distribution pipes, is not only responsible for the destruction of pipe material but also for deterioration of potable water quality due to unwanted chemical and biochemical reactions occurring in the distribution systems (McNeill and Edwards, 2001; Edwards, 2004; Hansen et al., 1996; Huck and Gangon, 2004; Ona-Nguema et al., 2002; Chaves, 2005). Corrosion of iron and steel pipes releases iron into distributed water that can re-precipitate

* Corresponding author. Tel.: þ48 618293430; fax: þ48 618293409.
E-mail address: [email protected] (J. Swietlik). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.006

2

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 e1 0

forming corrosion scales, also referred to as tubercles (Gerke et al., 2008). Corrosion scales are quite reactive species; they can actively modify physicochemical parameters of water in the distribution system not only by releasing Fe oxyhydroxides (red water) but also by reactions with e.g. chlorinated disinfection by-products (Chun et al., 2005), nitrates (Hansen et al., 1996) or with natural organic matter (NOM) (Nawrocki et al., 2010). Corrosion scales usually consist of a corroded floor, a porous core containing solid and often also fluid, a hard shell layer and surface layer, present on the top of shell-like layer (Baylis, 1926; Baylis, 1953; Sarin et al., 2004a,b; Gerke et al., 2008; Ray et al., 2010). The core of tubercles may be composed of a mixture of iron oxyhydroxides such as goethite and lepidocrocite, green rusts, magnetite, hematite, ferrous hydroxide, ferric hydroxide and siderite (Tuovinnen et al., 1980; Sontheimer et al., 1981; Sarin et al., 2001, 2004a,b). The hard shell layer is mainly composed of magnetite and goethite while the surface layer is more heterogeneous and can be composed of goethite, lepidocrocite, ferric hydroxide, silicates, phosphates and carbonates (Sontheimer et al., 1981; Sarin et al., 2001). The structures of corrosion scale have been examined by many methods: elemental analysis
_ 1995; Lin et al., 2001; Gerke ´ z, (Rudnicka and Swiderska-Bro et al., 2008), diffraction (Lin et al., 2001; Sarin et al., 2003, 2004a; Tang et al., 2006; Gerke et al., 2008) or microscopic (Sarin et al., 2001; Tang et al., 2006; Gerke et al., 2008). Magnetite, lepidocrocite and goethite have been repeatedly found (Sarin et al., 2001, 2004a) as the main crystallographic forms of corrosion scales. Maghemite, siderite and green rusts (GR) have also been detected in tubercles, however their amounts differ in reports of different authors. The most interesting and unstable fraction of corrosion by-products is made by green rusts (GR). GRs are layered double hydroxides (LDH) composed of positively charged layers and charge balancing anions located in the interlayer region. GRs with brucite-type layers containing Fe(II) and Fe(III) can have sulphate, chloride, carbonate (Refait et al., 2003) as anions and also water filling the interlayer spaces. The crystal structures of GRs depend on the type of interlayer anion (McGill et al., 1976; Hansen, 1989; Evans and Slade, 2006). The anions compete to form GRs but carbonate is considered as forming the most stable green rust e GR(CO2
3 ) (Refait et al., 1997; Reffas et al., 2006). Green rust in sulphate form, the so called GRII, was reported already in 1976 as a part of corrosion scales in drinking water pipe (McGill et al., 1976). The basic research on corrosion has shown that microbial corrosion also leads to the formation of GR(SO2
4 ) or closely coexists with sulphate green rust (Ge´nin et al., 1991, 2002; Refait et al., 2006; Pineou et al., 2008). Microbially-influenced corrosion has been reviewed by Beech and Gaylarde (1999). Green rust in sulphate form is preferentially formed even at the cost of drastic decrease in sulphate concentration in a surrounding solution (Refait et al., 2003). In other words formation of GR(SO2
4 ) may act as a “sulphate pump”. It has been recently shown that the “steady water” surrounding the corrosion scales may sometimes contain fewer sulphates than the water delivered by the treatment plant (Nawrocki et al., 2010). GR(Cl
) are considered as less stable and easily oxidizable than the other (carbonate or sulphate) forms of green rusts (Refait et al., 1997). Moreover, hydroxycarbonate green rust has been also shown to be the

main product of bioreduction of lepidocrocite (Ona-Nguema et al., 2002). Thus the question arises: why green rusts have been rather rarely reported as corrosion scales component? Usually the scales are dried before XRD analysis, no serious precautions are taken to avoid atmospheric oxygen, and this fact probably prevents detection of the phases actually existing in the corrosion scales. Details about preparation of the corrosion products for X-ray diffraction (XRD) measurements have been omitted in many papers (McGill et al., 1976; Badan et al., 1991). Analytical results are very similar when measurements have not been preceded by any special sample treatment, and, in most cases, magnetite, lepidocrocite and goethite are the only observed crystal phases (Badan et al., 1991; Frateur et al., 1999; Lin et al., 2001; Sarin et al., 2001, 2003, 2004a,b; Gismelseed et al., 2004). Similar results have been reported for the air dried samples of corrosion scales (Tang et al., 2006; Gerke et al., 2008). It indicates that observation of the intermediate and metastable corrosion products, such as green rust, requires a special sample treatment. The rapid reaction of GRs with atmospheric oxygen makes it difficult to identify these compounds in natural and engineered systems (Gismelseed et al., 2004; Lin et al., 2001; Zhang et al., 2008). Previously it had been expected that detection of GR phases could be possible by XRD simply after covering a flat sample with glycerol (Hansen, 1989; Zegeye et al., 2007a,b) or freeze-drying (Williams and Scherer, 2001). However, detailed research has shown a progressive sample decomposition during measurement, even for samples covered with glycerol (Mazeina et al., 2008). Unfortunately, also the freeze-drying procedure could alter the structure of corrosion products (Greffie´ et al., 2001). Moreover, sample drying, even in inert atmosphere, can result in product decomposition due to removal of the interlayer water. Therefore, even complex sample treatments before XRD analysis cannot guarantee stabilization of the rust sample. During presented study the corrosion products sampled from working distribution networks were carefully protected against contact with oxygen and drying, and transported to the laboratory directly after sampling. The storage of samples in inert atmosphere enabled to reveal of corrosion products’ complex crystallographic composition by diffraction methods. The comparison of wet and not oxidized deposits and suspensions with those usually analyzed and described in literature (i.e. oxidized and dried) show significant and very important differences in corrosion products structure. The main goal of this paper was to show how sample pretreatment can change the results of XRD analysis. Moreover microscopic images of the scales were taken to confirm the existence of various crystals in the tubercles.

2.

Experimental

2.1.

Sample collection

A large distribution system delivering drinking water produced in three different water treatment plants (WTPs) to about 1.6 million citizens was investigated. Table 1 presents full description of treatment trains in all WTPs as well as the number of pipe samples taken from the network.

3

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 e1 0

Table 1 e Water treatment plants description, chloride concentration in treated water and number of taken pipe samples. Water source

Water treatment train

Average Cl
concentration in treated water [mg L
1]

Number of pipe samples

WTP-1

Wisła River (bottom-infiltration)

130.8

9

WTP-2


skie Lake Zegrzyn

23.8

7

WTP-3

Wisła River (bottom-infiltration)

Coagulation with alum, sand filtration and slow sand filtration (with a layer of active carbon), final disinfection with Cl2/ClO2. Preozonation, alum coagulation, sand filtration, final disinfection with Cl2/ClO2. Aeration, sand filtration and final disinfection with Cl2/ClO2

120.0

3

Treatment plant

During the sampling procedure selected pipes were unearthed in a specially reinforced ditch, and next brushed and rinsed with water to remove soil from the surface. Then, the 30e70 cm fragments of 80e120 mm diameter cast iron pipes were cut off and horizontally removed from the network using supportive strips attached to the shoulders of excavator. In the next step, steady water e the black water surrounding and partly filling the tubercles in the pipes, containing suspension of corrosion products and usually retaining between corrosion scales on the bottom of the corroded pipe (Fig. 1) e was collected (Nawrocki et al., 2010). Steady water present on the bottom of the tubing (usually no more than 150 mL) (Nawrocki et al., 2010), was carefully drained to a glass bottle, transported to the laboratory and kept in a refrigerator at 4  C pending the analysis. Next, the deposits present in steady water samples were separated and analyzed in form of wet pulp. Simultaneously, the pipe segments were protected against drying and the contact with oxygen by placement in an argon saturated container (portable glovebox, Witko, Poland) and transported to the laboratory within 1e10 h. Afterwards, the samples of corrosion scales were taken e one part in wet state, in argon saturated 100 mL plastic containers, and the second part left in opened containers for drying at 25  C with oxygen access. The chemical compositions of corrosion by-products, both deposits and tubercles, were

Fig. 1 e Photograph of iron pipe sample (the steady water samples are taken from the hollows located between tubercles at the bottom of pipe fragment).

analyzed by ICP, and their morphology and structural composition were analyzed by SEM, TEM and XRD. To emphasize the complex composition of liquid containing loose deposits the steady water samples were analyzed by IC and ICP (Table 2).

2.2.

X-ray diffraction (XRD)

The rust deposits were taken from different fragments of pipes sampled from working distribution systems. The pipes with tubercular deposits were cut off and quickly transported to a laboratory. The wet samples were transferred to a mortar, homogenized as sampled (wet) and divided into two parts. One part was loaded as a dense slurry into a flat sample holder and covered with kapton foil; while the other part was dried in air. Then X-ray diffraction measurements were taken for both, dried and wet sample in the standard BraggeBrentano geometry on a Bruker D8 diffractometer using CuKa radiation. For the high-resolution synchrotron X-ray diffraction measurements, small quantities of rust-in-water suspension were ground with a pestle in an agate mortar, and introduced as slurry into 0.5 mm diameter glass capillaries. X-ray powder diffraction data were collected on a high-resolution X-ray powder diffractometer (beamline ID31 at ESRF) selecting Xrays of 0.41274(6)  A wavelength from the white undulator source. Data were collected for half an hour, normalized against monitor counts and detector efficiencies, and rebound into steps of 2q ¼ 0.001 . The samples were introduced into glass capillaries to avoid degradation of the airsensitive phases, and the capillaries were rotated during measurement to minimize the effect of preferential orientation. Rietveld refinement was performed using the Topas package (Bruker AXS, 2003). The starting structure models were taken from the following papers: magnetite (Fe3O4) (Fleet, 1981); goethite (a-FeOOH) (Verdonck et al., 1982); lepidocrocite (g-FeOOH) (Ewing, 1935); green rust II with SO2
4 anion, denoted as GR2_SO4 (Fe6(OH)12(SO4)(H2O)8) (Simon et al., 2003); green rust I with Cl
anion, denoted as GR1_Cl (Fe6(OH)10Cl(H2O)3) (Allmann and Donnay, 1969); siderite (FeCO3) (Graf, 1961); green rust I with CO2
3 anion, denoted as GR1_CO3 (Fe6(OH)12(CO3)(H2O)3) (Aissa et al., 2006); quartz (SiO2) (Glinnemann et al., 1992); akagene´ite (b-FeOOH) (Stahl et al., 2003); chukanovite (Fe2(CO3)(OH)2) (Pekov et al., 2007). For each crystal phase, the atomic displacement parameters

4

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 e1 0

Table 2 e The results characterizing the quality of steady waters. Parameter

Location 1

DOC [mg L
1] pH Fe [mg L
1] Mn [mg L
1] Al [mg L
1] Pb [mg L
1] Zn [mg L
1] K [mg L
1] Na [mg L
1] Mg [mg L
1] Ca [mg L
1]
1 NHþ 4 [mg L ] Ptot [mg L
1] Alkalinity [mval L
1] Cl
[mg L
1] F
[mg L
1] Br
[mg L
1]
1 NO
3 [mg L ]
1 SO2
4 [mg L ] Formate [mg L
1] Acetate [mg L
1] Oxalate [mg L
1]

WTP-1 9.399 6.06 360 8.6 2.5 e 0.6 4.9 39 14 94 e 0.77 0.60 857.5 0.660 1.37 0.00 43.00 0.066 0.052 0.378

Location 2

Location 3

33.307 6.25 1120 8.0 e e e 4.4 40.3 16.1 170 e 0.036 e 1057 e 4.26 0.11 140.53 e e 1.519

3.828 5.84 960 3.5 0.06 e e 3.8 23.0 12.6 170 e 0.00 e 1808.2 0.470 8.10 0.00 267.99 0.323 1.656 0.129

Location 4 WTP-3 3.350 5.81 nm nm nm nm nm nm nm nm nm nm nm e 664.4 0.425 1.11 1.23 55.44 0.193 1.219 0.033

Location 5

Location 6

Location 7

3.100 5.91 nm nm nm nm nm nm nm nm nm nm nm e 847.8 0.421 2.07 1.28 90.84 0.295 0.949 0.111

WTP-2 5.348 6.85 0.4 1.4 e e 0.002 4.3 17 11.3 107 e e 0.30 159.2 0.174 0.17 0.00 132.69 0.068 0.034 0.030

10.503 6.26 250 48 13 0.05 0.3 5.4 16.7 27 240 30 1.7 e 1346.6 0.552 1.75 0.27 80.61 0.044 e 0.421

nm e Not measured due to small sample volume.

were constrained to be equal for all atoms because of the multiphase character of the sample. The peak shape profile parameters were well fitted with a pseudo-Voigt function for the LaB6 standard and as such fixed for all phases of the corrosion sample. Background was corrected using a Chebyshev polynomial function.

2.3.

DOC measurement

DOC was analyzed by means of a TOC 1030 system (I.O. Analytical, USA). Before analysis all samples were filtered through 0.45 mm filters (Fisherbrand, Fisher Scientific). The detection limit (MDL) was 0.01 mg C L
1, while RSD of the method was <3% (Nawrocki et al., 2010).

2.4.

2.5.

Metal ions and total phosphorous determination

The metal ions (Fe, Mn, Ca, Mg,) and total phosphorous were determined by Inductively Coupled Plasma Spectroscopy. In water samples, concentrations of individual ions were analyzed after previous acidification with nitric acid (65%, POCH Gliwice, Poland) while the samples containing the suspension were additionally previously treated with 35% HCl (POCH Gliwice, Poland). The inductively coupled plasma optical emission spectrometry (ICP-OES) was measured on a Varian ICP-OES analyser model Vista-MPX (CCD simultaneous). The inductively coupled plasma mass spectrometry (ICP-MS) was measured on a Varian ICP-MS analyser. The details of the analysis method are described elsewhere (Nawrocki et al., 2010).

Inorganic ions and carboxylic acids determination

2
The inorganic ions (F
, Cl
, Br
, NO
3 , SO4 ) as well as organic acids (acetic, formic and oxalic) were determined by ion chromatography on DIONEX ICS-2500 system with IonPac AS19-HC analytical column (4  250 mm) and IonPac AG19-HC guard column (4  50 mm), connected with a conductivity detector ED 50A (Dionex, USA) e the details of the analytical
method are described elsewhere (Swietlik et al., 2009). Calibration curves were linear from the detection limit 5 mg L
1 to at least 100 mg L
1. RDS was less than 10% for each of the ions monitored. All organic acids standards were obtained from SigmaeAldrich (Poland) and the standard solution of inorganic ions were obtained from Dionex (USA). Before analysis all samples were filtered through 0.45 mm filters (Fisherbrand, Fisher Scientific).

2.6.

SEM

The morphology of corrosion products was examined using a scanning electron microscope Carl ZEISS EVO 40 SEM operating at 20 kV. The samples were dehydrated with acetone and then were mounted on stubs and coated with gold in a BALZERS SCD 050 sputter coater.

2.7.

TEM

Images of some sediments and bacteria were obtained using JEOL JEM 1200 Ex II high-resolution 120 kV transmission electron microscope.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 e1 0

3.

Results and discussion

The heterogeneous composition of corrosion scales has been shown by many authors (Baylis, 1926; Baylis, 1953; Tuovinnen et al., 1980; Sontheimer et al., 1981; Sarin et al., 2001, 2004a,b; Gerke et al., 2008; Ray et al., 2010). Physically, the tubercles are reported as a main form of the scales in steel or unlined cast iron water pipes (Tuovinnen et al., 1980; Gerke et al., 2008). The tubercles are themselves heterogeneous e under relatively hard layer of the yellowish-red material there is a much softer, black interior. The hard shell is porous and enables the exchange between the interior and the flowing water
_ 1995; Gerke et al., 2008). The ´ z, (Rudnicka and Swiderska-Bro scales interior contains usually more or less soft and loose material with very low redox potential (Tuovinnen et al., 1980); and pH measured inside the tubercle interior is lower that that in flowing water (Baylis, 1926, 1953; Tuovinnen et al., 1980). As recently found, a steady water layer exists also in the external part of the tubercles and it could have similar properties to the water situated inside the corrosion scales (Nawrocki et al., 2010). Steady waters usually contain corrosion products in the form of loose-fitting slurry, however the crystallographic composition of suspended material has not yet been analyzed since it would have required a special sample preparation (avoiding oxidation) and analysis while wet. In the present study corrosion scales as well as deposits sampled from steady water were analyzed in the wet state by the use of SEM, TEM and XRD methods.

3.1.

SEM and TEM of the corrosion products

During the study hundreds of SEM images of the corrosion scales scraped from the several fragments of cast iron pipes used for potable water distribution in a large city were taken. Generally, the microscopic images show the variety of crystallographic structures in the scales. The first set of images (Fig. 2) shows the external layer covering the tubercle. According to Gerke et al. (2008) the external layer thickness can be up to several millimetres while in the present study very diverse widths of hard shell were observed e from about 1 mm (Fig. 2A) to even single micrometer (Fig. 2B). The thickness of tubercles’ shell may depend on the age and size of

5

corrosion products as well as on the quality of flowing water and corrosion products composition. SEM images have also indicated that tubercles with thin external layer can coexist with lumps covered with thicker shells. Under the shells a variety of structures can be observed (Fig. 3), some of them amorphous, some crystalline and some of the structures difficult to identify, like the ones shown in Fig. 3A. The crystalline products of corrosion are observed in suspended matter as well as in steady waters (Nawrocki et al., 2010). Fig. 4 shows the SEM and TEM images of the hexagonal crystals which are similar to those of GR1(CO2
3 ) synthesized in laboratory and shown by Ge´nin group (Ge´nin et al., 2005; Bocher et al., 2004). Formation of carbonate green rust requires the presence of alkaline conditions and carbonate ions. In drinking water conditions both the pH value and the concentration of bicarbonates almost exclude the availability of carbonate anions, at least in the actually examined system. Hence, in potable water pipes preferential formation of sulphate form of green rusts is expected. Sulphate form green rusts are often connected with the presence of desulphatizing bacteria (Desulfovibrio). In all of the examined corrosion scales the presence of sulphides was determined qualitatively (Nawrocki et al., 2010). Moreover, the TEM examination of steady water confirmed the presence of bacteria similar to those reported in literature (Beech and Gaylarde, 1999) and in many sources on the Internet (Fig. 5). The bacteria in steady waters are always associated with tiny crystalline suspension of corrosion scales. The presence of various crystalline structures and bacteria cannot be reliably confirmed solely by the SEM and TEM images.

3.2.

XRD analysis of corrosion scales

As mentioned above, the corrosion scales in general are very heterogeneous, covered by hard but often fragile shell. Under the shell the material is wet and seems to be more homogeneous but it contains various crystalline forms of rust as visualized in SEM images. It is thus difficult to sample the material which would be homogeneous. As a result all the quantitative results from e.g. diffraction methods cannot be treated as reliable information on the corrosion scale

Fig. 2 e External layer of the tubercle A) photograph, B) SEM image of a single tubercle (the bar in the lower left corner is equal to 2 mm), C) SEM image of the hard shell (the bar in the lower left corner is equal to 100 mm).

6

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 e1 0

Fig. 3 e SEM images of A) unidentified structure found in the interior of the tubercle; BeD) other types of crystallite forms (iron oxides); E) single crystals of green rust; F) calcite and aragonite like crystals.

contents. Results of determination of the composition of corrosion products can be also distorted by improper sample pre-treatment. For this reason some experiments were conducted to show the influence of drying the samples on the

contents of the corrosion scales. The material for the diffraction measurements was taken from a pipe in which heavy tuberculation was observed. Three types of the material have been separated: M e the black material from the interior of the

Fig. 4 e Images of the hexagonal green rust crystals in suspended matter in steady water: a) SEM (the bar in the upper left corner is equal to 2 mm) and b) TEM (the bar at the bottom of photo is equal to 0.5 mm).

7

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 e1 0

Fig. 5 e a) TEM image of bacteria found in steady water, b) desulfovibrio bacteria (http://www.textbookofbacteriology.net/ structure_2.html; 26.04.2011).

tubercle, G/L e mixed phase containing yellowish-red-brown material, and G e the most outer, yellowish material taken from the shell layer. The sampled wet and homogenized materials were divided into two parts. One part was analyzed in the form of dense slurry and the other one was dried in air. Measurements for the samples as slurry enabled recreation of the conditions close to those in the working water distribution system. The diffraction study revealed the presence of three types green rusts in the wet material (Table 3), while drying was found to convert the unstable green rusts into more stable phases such as goethite and lepidocrocite. Magnetite contents also decreased upon drying. The results shown in Table 3 have to be treated as semi quantitative as i) it is difficult to sample homogeneous material from the corrosion scales and ii) it is almost impossible to avoid a contact with air, particularly when the fragment of the pipe is cut off from the distribution system. However, the results presented in Table 3 confirm the

necessity of examination of corrosion scales in the wet conditions, as drying of the samples before XRD analysis changes substantially the crystallographic phases originally present in the corrosion products. The green rusts are generally unstable and they are often overlooked in water pipe corrosion research (Refait et al., 1997). In the present study, samples of the corrosion scales taken in the wet state from the pipes excavated at different locations (U and N) were compared. Two samples of the corrosion products from each pipe were collected the yellowish and the black material. The black material was sampled from the interior of the tubercles, while the yellow one was collected from the outer shell of the tubercles. The XRD results presented in Table 4 revealed the presence of three types of green rust, in both samples examined. Particularly surprising was the presence of considerable amounts of LDH in chloride stabilized forms. Chloride form of green rust is considered as the least stable (Refait et al., 1998). However the

Table 3 e The influence of sample drying on the corrosion products composition determined by XRD analysis [%].

Table 4 e Dominant crystallographic forms in the corrosion scale [%].

Phase identified

Phase

Wet samples

Dried samples

Mwet G/Lwet Gwet Mdry Quarz Green rust Cl
Green rust CO2
3 Goethite Magnetite Lepidocrocite Green rust II SO2
4 Siderite

e 46.2 1.1 19.6 31.1 0 1.4 0.6

0 9.1 2.4 61.7 17.9 7.4 0.5 1

e 16.8 3.3 45.6 7.5 21.6 2.6 2.5

0 0 0 75.8 11.5 12.7 0 0

G/Ldry

Gdry

0 0 0 60.5 12.7 26.7 0 (Traces)

0 0 0 72.9 8.5 17.1 0 1.5

Quartz Green rust Cl
Green rust CO2
3 Goethite Magnetite Lepidocrocite Green rust II SO2
4 Siderite

Yellow scale

Black scale

N

U

N

U

0 1.2 10.8 56.2 21.4 7.9 1.2 1.3

0 9.1 2.4 61.7 17.9 7.4 0.5 1

e 22.1 4.3 22.4 47.1 0 0.2 3.8

e 46.9 1.1 19.3 31.7 0 1.4 0.6

8

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 e1 0

Table 5 e Composition of steady water suspensions measured by X-ray powder diffraction measurements of the wet samples [%]. Sample

Location 1

Location 2

Location 3

Location 4

Location 5

Location 6

Location 7

Magnetite Goethite Lepidocrocite GR2_SO2
4 _(1) GR2_ SO2
4 _(2) GR1_Cl
_(1) GR1_Cl
_(2) GR1_CO2
3 _(1) GR1_ CO2
3 _(2) Siderite Akaganeite_(w) Chukanovite Quartz

WTP-1 18.05  0.23 26.01  0.67 4.51  0.12 4.27  0.30 1.73  0.16 15.35  0.47 10.05  0.19 6.18  0.26 2.49  0.18 1.242  0.82 3.13  0.11 5.09  0.19 1.897  0.98

6.681  0.035 70.98  0.18 14.47  0.12 e e e e 1.387  0.086 e 1.260  0.025 0.712  0.041 3.559  0.091 0.959  0.047

21.56  0.10 49.45  0.23 0.067  0.036 e e e 0.13  0.14 1.74  0.13 e 4.268  0.031 e 22.62  0.12 0.157  0.046

WTP-3 6.299  0.040 27.57  0.31 1.360  0.043 1.623  0.068 e 3.254  0.042 e 22.37  0.15 11.42  0.31 1.320  0.029 5.812  0.071 18.63  0.16 0.350  0.036

21.13  0.14 48.37  0.32 1.459  0.065 0.200  0.014 e 0.414  0.035 2.287  0.095 7.25  0.31 5.96  0.27 3.326  0.040 1.530  0.053 6.88  0.11 1.196  0.055

WTP-2 31.35  0.20 37.68  0.31 1.298  0.075 e e 0.622  0.035 11.94  0.27 3.64  0.11 4.80  0.27 0.773  0.032 e 2.45  0.12 5.444  0.085

18.96  0.14 40.64  0.35 0.544  0.054 e e e e 2.948  0.089 e 1.031  0.054 1.82  0.11 29.41  0.30 4.655  0.096

presence of GR(Cl
) corresponds very well to huge concentrations of chlorides found in steady waters (Nawrocki et al., 2010). Since much higher amounts of GR(Cl
) are present in interiors of the tubercles it would be reasonable to assume that chloride stabilized green rust plays an important role in the corrosion of cast iron water pipes. Carbonate and sulphate green rusts are also present but in much lower proportions. It has to be emphasized that fragments of pipes were taken from distribution system which is supplied by water relatively rich in chloride with the average concentration of 130.8 mg L
1 (Nawrocki et al., 2010).

3.3.

XRD analysis of steady water suspension

Since green rusts are metastable redox systems (Fe(II)/Fe(III)) it is important to find out if they are present in appreciable amounts in the loose-fitting products of corrosion suspended in water that remains between tubercles and possibly in their interiors. Steady water samples obtained with the fragments of cut off pipes contained corrosion products material in the form of loose deposits as well as heavy black suspension. The solid material from the waters was separated, loaded into glass capillaries as slurry to avoid oxidation, and analyzed by the high-resolution X-ray powder diffractometer. Table 5 shows a variety of structures present in corrosion products. The most oxidized material (goethite, lepidocrocite and magnetite) was present in all samples. Low amounts of siderite and quartz were also found in all samples. Six different forms of green rusts were identified in the suspensions examined. Carbonate green rust GR (CO2
3 )(I) seemed to be the most abundant and was found in all samples. Carbonate green rust is considered as the most stable of the rusts and thus its identification in the corrosion scale should not be surprising. Carbonate green rust is mostly observed in alkaline conditions (Refait et al., 1997) while all the samples of steady water were characterized by a low pH (Table 2). Although the samples of the black suspensions were kept closed in refrigerator before analysis they showed considerable differences in composition. Only in two samples appreciable amounts of GR(Cl
) were found. The general trend

observed for the samples was that the more goethite in the sample, the less of the other unstable (intermediate) forms of corrosion products. Despite GRs, akaganeite and chukanovite were detected in the majority of the samples. Formation of akaganeite has been reported in chloride rich systems (Refait and Ge´nin, 1997; Remazeilles and Refeit, 2007) and thus this would confirm the importance of chloride forms of LDH in water corrosion. It should be emphasized that GR(Cl
) is also present in a sample of the rust from WTP-2 which supplies water with much lower average chloride content (23.8 mg L
1) in water (Nawrocki et al., 2010) than that from WTP-1 (130.8 mg L
1) and WTP-3 (120 mg L
1). It indicates that GR(Cl
) is most likely preferentially formed during corrosion of cast iron. As very unstable compound it can be easily oxidized to e.g. akaganeite or goethite, and chloride ions are released to the solution surrounding corrosion products, i.e. steady water.

4.

Conclusions

The crystallographic composition of corrosion scales and deposits suspended in steady waters were analyzed by XRD, SEM, TEM and ICP. The results revealed the necessity of the examination of corrosion products in the wet conditions. XRD data showed that drying of the samples before XRD analysis substantially changes the crystallographic phases originally present in the corrosion products. The diffraction method revealed the presence of three types of green rusts in the wet materials. The most astonishing was the occurrence of significant amounts of the least stable green rust in chloride form in almost all corrosion scale samples. XRD results also proved that sample drying converts the unstable green rusts into more stable phases such as goethite and lepidocrocite, and leads to a decrease in magnetite and siderite content. The analysis of the corrosion products suspended in steady water evidenced the complex crystallographic composition of the sampled material. Goethite, lepidocrocite and magnetite as well as low amounts of siderite and quartz were present in all samples. Six different forms of green rusts were identified

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 e1 0

in the suspensions examined; and the most abundant was carbonate green rust GR(CO2
3 )(I). This finding was also surprising as the environment of the steady water is rather acidic. Additionally, SEM and TEM images exemplified the complex structure of corrosion products and indicated the presence of sulphate reducing bacteria attached to suspended corrosion deposits.

Acknowledgement The authors thank the Polish Ministry of Science and Higher Education (Grants no. N N204 339337 and no. N N523 418737) and the Municipal Water Supply and Sewerage Company in Warsaw (Poland) for financial support of this project. We also acknowledge the European Synchrotron Radiation Facility for provision of synchrotron radiation facilities and we would like to thank Dr Michela Brunelli for assistance in using beamline ID31.

references

Aissa, R., Francois, M., Ruby, C., Fauth, F., Medjahdi, G., Abdelmoula, M., Genin, J.-M., 2006. Formation and crystallographical structure of hydroxysulphate and hydroxycarbonate green rusts synthetised by coprecipitation. J. Phys. Chem. Solids 67, 1016e1019. Allmann, R., Donnay, J.D.H., 1969. About the structure of iowaite. Am. Mineral. 54, 296e299. AWWA, 2004. Water://Stats 2002 Distribution Survey. AWWA, Denver, CO. Badan, B., Magrini, M., Ramous, E., 1991. A study of the microbiological-corrosion products of steel and cast iron pipes in fresh water. J. Mater. Sci. 26, 1951e1958. Baylis, J.R., 1926. Prevention of corrosion and “red water”. J. AWWA 15, 598e633. Baylis, J.R., 1953. Cast-iron pipe coatings and corrosion. J. AWWA, 807e831. Beech, I.B., Gaylarde, Ch.C., 1999. Recent advances in the study of biocorrosion. An overview. Rev. Microbiol. 30, 177e190. Bocher, F., Ge´hin, A., Ruby, Ch., Ghanbaja, J., Abdelmoula, M., Ge´nin, J.M.R., 2004. Co-precipitation of Fe(IIeIII) hydroxycarbonate green rust stabilised by phosphate adsorption. Solid State Sci. 6, 117e124. Bruker AXS, 2003. TOPAS V2.1: General Profile and Structure Analysis Software for Powder Diffraction Data. User’s Manual. Bruker AXS, Karlsruhe, Germany. Chaves, L.H.G., 2005. The role of green rust in the environment: a review. Rev. Bras. Eng. Agric. Ambient. 9 (2), 284e288. Chun, Ch.L., Hozalski, R.M., Arnold, W.A., 2005. Degradation of drinking water disinfection by-products by synthetic goethite and magnetite. Environ. Sci. Technol. 39, 8525e8532. Edwards, M., 2004. Controlling corrosion in drinking water distribution systems: a grand challenge for the 21st century. Water Sci. Technol. 49 (2), 1e8. Evans, D.G., Slade, R.C.T., 2006. Structural aspects of layered double hydroxides. In: Duan, X., Evans, D.G. (Eds.), Layered Double Hydroxides. Springer, Berlin/Heidelberg. Ewing, F.J., 1935. The crystal structure of lepidocrocite. J. Chem. Phys. 3, 420e424.

9

Fleet, M.E., 1981. The structure of magnetite. Acta Crystallogr. B37, 917e920. Frateur, I., Deslouis, C., Kiene, L., Levi, Y., Tribollet, B., 1999. Free chlorine consumption induced by cast iron corrosion in drinking water distribution systems. Water Res. 33, 1781e1790. Ge´nin, J.M.R., Aı¨ssa, R., Ge´hin, A., Abdelmoula, M., Benali, O., Ernstsen, V., Ona-Nguema, G., Upadhyay, Ch., Ruby, Ch., 2005. Fougerite and Fe IIeIII hydroxycarbonate green rust; ordering, deprotonation and/or cation substitution; structure of hydrotalcite-like compounds and mythic ferrosic hydroxide Fe (OH)(2þx). Solid State Sci. 7, 545e572. Ge´nin, J.M.R., Olowe, A.A., Benbouzid-Rollet, N.D., Prieur, D., Confente, M., Resiak, B., 1991. The simultaneous presence of green rust 2 and sulphate reducing bacteria in the corrosion of steel sheet piles in a harbour area. Hyperfine Interact. 69, 875e878. Ge´nin, J.M.R., Refait, Ph., Abdelmoula, M., 2002. Green rusts and their relationship to iron corrosion: a key role in microbially influenced corrosion. Hyperfine Interact. 139/ 140, 119e131. Gerke, T.L., Maynard, J.B., Schock, M.R., Lytle, D.L., 2008. Physicochemical characterization of five iron tubercles from a single drinking water distribution system: possible new insights on their formation and growth. Corrosion Sci. 50, 2030e2039. Gismelseed, A., Elzain, M., Yousif, A., Al Rawas, A., Al-Omari, I.A., Widatallah, H., Rais, A., 2004. Identification of corrosion products due to seawater and fresh water. Hyperfine Interact. 156/157, 487e492. Glinnemann, J., King, H.E., Shulz, H., Hahn, T., La Placa, S.J., Dacol, F., 1992. Crystal structures of the low-temperature quartz-type phases of SiO2 and GeO2 at elevated pressure. Z. Kristallogr. 198, 177e212. Graf, D.L., 1961. Crystallographic tables for the rhombohedral carbonates. Am. Mineral. 46, 1283e1316. Greffie´, C., Amouric, M., Parron, C., 2001. HRTEM study of freezedried and untreated synthetic ferrihydrites: consequences of sample processing. Clay Miner. 36, 381e387. Hansen, H.C.B., 1989. Composition, stabilization, and light absorption of Fe(II)Fe(III) hydroxy-carbonate (“green rust”). Clay Miner. 24, 663e669. Hansen, H.Chr.B., Koch, Chr.B., Nancke-Krogh, H., Borgaard, O.K., Sørensen, I., 1996. Abiotic nitrate reduction to ammonium: key role of green rust. Environ. Sci. Technol. 30, 2053e2056. Huck, P.M., Gangon, G.A., 2004. Understanding the distribution system as a bioreactor: a framework for managing heterotrophic plate count levels. Int. J. Food Microbiol. 92, 347e353. Kwietniewski, M., Tłoczek, M., Wysocki, L., 2011. Zasady doboru
materiałowo-konstrukcyjnych do budowy rozwia˛zan przewodo´w wodocia˛gowych, Izba Gospodarcza “Wodocia˛gi Polskie” Bydgoszcz. Lin, J., Ellaway, M., Adrien, R., 2001. Study on corrosion material accumulated on the inner wall of steel water pipe. Corrosion Sci. 43, 2065e2081. Mazeina, L., Navrotsky, A., Dyar, D., 2008. Enthalpy of formation of sulfate green rusts. Geochim. Cosmochim. Acta 72, 1143Ae1153A. McGill, I.R., McEnaney, B., Smith, D.C., 1976. Crystal structure of green rust formed by corrosion of cast iron. Nature 259, 200e201. McNeill, L.S., Edwards, M., 2001. Iron pipe corrosion in distribution systems. J. AWWA 88 (7), 88e100.
Nawrocki, J., Raczyk-Stanisławiak, U., Swietlik, J., Olejnik, A., Sroka, M., 2010. Corrosion in a distribution system. Steady water and its composition. Water Res. 44 (6), 1863e1872. Ona-Nguema, G., Abdelmoula, M., Jorand, F., Benali, O., Ge´hin, A., Block, J.-C., Ge´nin, J.-M.R., 2002. Iron (II, III) hydroxycarbonate

10

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 e1 0

green rust formation and stabilization from lepidocrocite bioreduction. Environ. Sci. Technol. 36, 16e20. Pekov, I.V., Perchiazzi, N., Merlino, S., Kalachev, V.N., Merlini, M., Zadov, A.E., 2007. Chukanovite, Fe2(CO3)(OH)2, a new mineral from the weathered iron meteorite Dronino. Eur. J. Mineral. 19, 891e898. Pineou, S., Sabot, R., Quillet, L., Jeannin, M., Caplat, Ch., DupontMorral, I., Refait, Ph., 2008. Formation of Fe(IIeIII) hydroxysulphate green rust during marine corrosion of steel associated to molecular detection of dissimilatory sulphitereductase. Corrosion Sci. 50, 1099e1111. Ray, R.I., Lee, J.S., Little, B.J., Gerke, T.L., 2010. The anatomy of tubercles: a corrosion study in a fresh water estuary. Mater. Corrosion 61 (12), 993e999. Refait, P., Drissi, S.H., Pytkiewicz, J., Ge´nin, J.M.R., 1997. The anionic species competition in iron aqueous corrosion: role of various green rust compounds. Corrosion Sci. 39 (9), 1699e1710. Refait, P., Memet, J.B., Bon, C., Sabot, R., Ge´nin, J.M.R., 2003. Formation of the Fe(II)eFe(III) hydroxysulphate green rust during marine corrosion of steel. Corrosion Sci. 45, 833e845. Refait, Ph., Abdelmoula, M., Ge´nin, J.M.R., Sabot, R., 2006. Green rusts in electrochemical and microbially influenced corrosion of steel. C. R. Geosci. 338, 476e487. Refait, Ph., Abdelmoula, M., Ge´nin, J.M.R., 1998. Mechanisms of formation and structure of green rust one in aqueous corrosion of iron in the presence of chloride ions. Corrosion. Sci. 40, 1547e1560. Refait, Ph., Ge´nin, J.M.R., 1997. The mechanism of oxidation of ferrous hydroxychloride b-Fe2(OH)3Cl in aqueous solution: the formation of akageneite vs. goethite. Corrosion Sci. 39 (3), 539e553. Reffas, M., Sabot, R., Savall, C., Jeannin, M., Creus, J., Refeit, Ph., 2006. Localised corrosion of carbon steel in NaHCO3/NaCl electrolytes; role of Fe(II)-containing compounds. Corrosion Sci. 48, 709e726. Remazeilles, C., Refeit, Ph., 2007. On the formation of b-FeOOH (akaganeite) in chloride containing environments. Corrosion Sci. 49, 844e857.
_ M., 1995. Skład chemiczny ´ z, RudnickaŁ. and Swiderska-Bro osado´w z wrocławskiej sieci wodocia˛gowej. (Chemical composition of deposits in Wroclaw’ distribution system).
Ochrona Srodowiska 3/58, 63e65. Sarin, P., Clement, J.A., Snoeyink, V.L., Kriven, W.M., 2003. Iron release from corroded, unlined cast-iron pipe. J. AWWA 95 (11), 85e96. Sarin, P., Snoeyink, V.L., Bebee, J., Jim, K.K., Beckett, M.A., Kriven, W.M., Clement, J.A., 2004a. Iron release from corroded iron pipes in drinking systems: effect of dissolved oxygen. Water Res. 38, 1259e1269.

Sarin, P., Snoeyink, V.L., Bebee, J., Kriven, W.M., Clement, J.A., 2001. Physico-chemical characteristics of corrosion scales in old iron pipes. Water Res. 35 (12), 2961e2969. Sarin, P., Snoeyink, V.L., Lytle, D.A., Kriven, W.M., 2004b. Iron corrosion scales: model for scale growth, iron release, and colored water formation. J. Environ. Eng. 130 (4), 364e373. Simon, L., Franc¸ois, M., Refait, P., Renaudin, G., Lelaurain, M., Ge´nin, J.M.R., 2003. Structure of the Fe(IIeIII) layered double hydroxysulphate green rust two from Rietveld analysis. Solid State Sci. 5, 327e334. Sontheimer, H., Kolle, W., Snoeyink, V.L., 1981. The siderite model of the formation of corrosion-resistant scales. J. AWWA 73 (11), 572e579. Stahl, K., Nielsen, K., Jiang, J., Lebech, B., Hanson, J.C., Norby, P., van Lanschot, J., 2003. On the akaganeite crystal structure, phase transformations and possible role in post-excavational corrosion of iron artifacts. Corrosion Sci. 45, 2563e2575.
Swietlik, J., Raczyk-Stanisławiak, U., Nawrocki, J., 2009. The influence of disinfection on aquatic biodegradable organic carbon formation. Water Res. 43, 463e473. Tang, Z., Hong, S., Xiao, W., Taylor, J., 2006. Characteristics of iron corrosion scales established under blending of ground, surface and saline waters and their impacts on iron release in the pipe distribution system. Corrosion Sci. 48, 322e342. Tuovinnen, O.H., Button, K.S., Vuorinen, A., Carlson, L., Mair, D.M. , Yut, L.A., 1980. Bacterial, chemical and mineralogical characterization of tubercles in distribution pipelines. J. AWWA 72 (11), 626e635. Veschetti, E., Achene, L., Ferretti, E., Lucentini, L., Citti, G., Ottaviani, M., 2010. Migration of trace metals in Italian drinking waters from distribution networks. Toxicol. Environ. Chem. 92 (3), 521e535. Verdonck, L., Hoste, S., Roelandt, F.F., Van der Celen, G.P., 1982. Normal coordinate analysis of a-FeOOH e a molecular approach. J. Mol. Struct. 79, 273e279. Williams, A.G.B., Scherer, M.M., 2001. Spectroscopic evidence for Fe(II)eFe(III) electron transfer at the iron oxide water interface. Environ. Sci. Technol. 35, 4782e4790. Zegeye, A., Huguet, L., Abdelmoula, M., Carteret, C., Mullet, M., Jorand, F., 2007a. Biogenic hydroxysulfate green rust, a potential electron acceptor for SRB activity. Geochim. Cosmochim. Acta 71, 5450e5462. Zegeye, A., Ruby, Ch, Jorand, F., 2007b. Kinetic and thermodynamic analysis during dissimilatory g-FeOOH reduction: formation of green rust 1 and magnetite. Geomicrobiol. J. 24, 1e14. Zhang, Z., Stout, J.E., Yu, V.L., Vidic, R., 2008. Effect of pipe corrosion scales on chlorine dioxide consumption in drinking water distribution systems. Water Res. 42, 129e136.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 e2 0

Available online at www.sciencedirect.com

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

The impact of bromide/iodide concentration and ratio on iodinated trihalomethane formation and speciation Darryl B. Jones, Aysenur Saglam, Hocheol Song, Tanju Karanfil* Department of Environmental Engineering and Earth Sciences, Clemson University, Anderson, SC 29625, USA

article info

abstract

Article history:

The objective of this study was to evaluate the formation and speciation of iodinated

Received 25 July 2011

trihalomethanes (I-THMs) from preformed chloramination of waters containing bromide

Received in revised form

(Br
) and iodide (I
) at a Br
/I
weight ratio of 10:1. The factors investigated were pH, iodide

6 October 2011

to dissolved organic carbon (I
/DOC) ratio, and NOM characteristics, specifically SUVA254. A

Accepted 9 October 2011

Br
/I
ratio of 1:2 was also evaluated to determine the importance of Br
and I
concen-

Available online 20 October 2011

trations and ratio on I-THM formation and speciation. Regulated triholamethanes (THMs) were measured alongside I-THMs for a more complete understanding of trihalomethane

Keywords:

formation. The results showed that, in general, both I-THM and THM formation increased

Bromide

with decreased pH. Greater formation at lower pH was likely attributed to monochloramine

Iodide

decomposition and the formation of additional oxidants and substituting agents, most

I-THM

notably chlorine. For pH  7.5, I-THM yield increased with increasing I
/DOC ratio and

THM

decreasing specific ultraviolet absorbance (SUVA254) of the water. The Br
/I
, Br
/DOC and

Monochloramine

I
/DOC ratios were important factors for I-THM and THM speciation. At pH 6, dichlor-

Chlorine

oiodomethane (CHCl2I) and bromochloroiodomethane (CHBrClI) were the dominant

Natural organic matter

species at the common bromide and iodide levels. For pH  7.5 and for elevated bromide and iodide levels, iodoform (CHI3) was always the dominant specie regardless of the Br
/I
ratio. The results demonstrated that it is important to examine I-THM formation and speciation at typical Br
/I
ratios (w10) of natural waters, which have often been overlooked in previous investigations, in order to obtain practical and relevant results. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Disinfection byproducts (DBPs) form in drinking waters as a result of reactions between oxidants and natural organic matter (NOM) (Rook, 1974). The NOM characteristics of the source water influence DBP formation and speciation (Kitis et al., 2001; Ates et al., 2007). The most commonly used oxidant in water treatment is chlorine. If bromide (Br
) and iodide (I
) are also present, they can react with oxidants and form halogenated DBPs, often measured as total organic halogen (TOX) (Krasner et al., 1989). Chlorine, bromine, and

iodine may all substitute into NOM to form TOX. In terms of toxicity, it has been shown that iodinated DBPs, as a group, are more cyto- and genotoxic than brominated or chlorinated DBPs (Plewa et al., 2008). Trihalomethanes (THMs) and haloacetic acids (HAAs) include brominated and chlorinated DBP species that are currently regulated by the United States Environmental Agency (USEPA) (USEPA, 1998, 2006). However, iodinated DBPs are not currently regulated. In order to comply with increasingly stringent THM and HAA regulations in the US, many water utilities have switched from chlorine to monochloramine practices (Seidel et al.,

* Corresponding author. Tel.: þ1 864 656 1005; fax: þ1 864 656 0672. E-mail address: [email protected] (T. Karanfil). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.005

12

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 e2 0

2005) because monochloramine forms fewer regulated THMs and HAAs (Hong et al., 2007). However, iodinated DBP formation is favorable during chloramination (Bichsel and von Gunten, 2000; Krasner et al., 2006; Hua and Reckhow, 2007a; Richardson et al., 2008). As a result, utilities that have converted to monochloramine to lower THM and HAA levels may form iodinated DBPs when treating source waters containing inorganic and organic iodine species. Iodinated THM (I-THM) formation is favorable under chloramination conditions because of the competing kinetics of iodine oxidation and NOM substitution. First, monochloramine oxidizes I
to HOI. Then, HOI can either react with NOM
or be oxidized to IO
2 and further to IO3 . However the oxidation pathway is extremely slow compared to the much faster reactions of HOI with NOM. Therefore, unlike ozone and chlorine, monochloramine does not considerably oxidize HOI to IO
3 , a non-toxic end product for iodine (Bichsel and von Gunten, 1999, 2000). When comparing various treatment scenarios (ozone/chlorine, ozone/monochloramine, chlorine, monochloramine, and chlorine dioxide) of waters spiked with elevated concentrations of iodide (200 mg/L), it was confirmed that I-THM formation was greatest from monochloramine (Hua and Reckhow, 2007a). An occurrence study conducted in the US that investigated 12 water treatment plants with “challenged” source waters (i.e., comparatively high TOC, high Br
concentrations) found that the highest I-THM formation was observed at a monochloramine plant that added chlorine and ammonia at the same location in the plant (i.e., little to no free chlorine contact time) (Krasner et al., 2006). The sum of six I-THM species was 81% of the mass of 4 regulated THM species at this particular plant. Additionally, during this study, iodo-acids were also found in drinking waters for the first time. The higher formation of IDBPs during chloramination prompted a follow up study of chloramination plants in 23 cities (Richardson et al., 2008). The source waters had different bromide and iodide concentrations, and the utilities practiced varying chlorine contact times prior to ammonia addition. The researchers found that I-THM formation (2 of 6 I-THM species measured) was greater for plants with shorter free chlorine contact times (<1 min) than plants with longer free chlorine contact times (>45 min). The same occurrence study demonstrated that iodide levels generally varied between 0.4 and 104 mg/L with a median of 10 mg/L, while the bromide concentration varied from 24 to 1120 mg/L with a median of 109 mg/L in these source waters. It is only under extreme circumstances that iodide can be higher than 100 mg/L. For example, the iodide concentration in the Rio Grande in Brownsville, TX was previously recorded as 212 mg/L (Moran et al., 2002). Iodide concentrations can be higher in arid watersheds such as the American Southwest where irrigation is prevalent, and transpiration dominates. Therefore, salts may concentrate in the soil, and agricultural runoff can transport iodide to the rivers. However, this scenario is rare and most of the time, the iodide concentrations are much lower. For confined groundwaters, iodide generally varies between 0.01 and 20 mg/L (Ali-Mohamed and Jamali, 1989). However, if salt water intrusion is occurring, or if the groundwater is adjacent to halide rocks or brines, iodide concentrations can exceed 50 mg/L (Muramatsu and Wedepohl, 1998; Cancho et al., 2000).

Similar to the concentrations of bromide and iodide, the Br
/I
ratio may also vary in source waters. However, the bromide level is almost always higher than the iodide level. In the Richardson et al. (2008) study, the Br
/I
ratio varied widely from 3/1 to 238/1 with an average of 13/1. Such large variations in the Br
/I
ratio in freshwater are probably due to many factors such as salt water intrusion, various salt deposits containing different levels of bromine and iodine, anthropogenic influences, and slightly greater oxidation potential and biological cycling of iodide compared to bromide (Whitehead, 1984; Fuge and Johnson, 1986; Neal et al., 1990; Muramatsu and Wedepohl, 1998; von Gunten, 2003; Steinberg et al., 2008). Bromide and iodide concentrations will likely play an important role in I-THM formation and speciation. Kristiana et al. (2009) showed that the formation of total organic iodine (TOI) from chloramination of the hydrophobic acid fraction from the Loire River, Australia increased linearly with increasing iodide concentration (50e300 mg/L). The bromide level was held constant at 300 mg/L, so varying Br
/I
ratios were evaluated. Since I-THMs were not measured, it is unknown if I-THM formation would follow the same linear trend as TOI formation. Hua and Reckhow (2008) investigated the effect of pH on ITHM formation from preformed monochloramine. A source water was spiked with very high concentrations of iodide (1900 mg/L). No bromide was added, and the ambient level of bromide was much lower (78 mg/L); as a result, the Br
/I
ratio of the spiked source water was 1:25. This was much lower than a more typical Br
/I
mass ratio of 13:1 in natural waters (Richardson et al., 2008). The results showed that (i) I-THM formation increased with pH from 5 to 9, and (ii) iodoform (CHI3) was the most important I-THM specie for all pH values. The NOM characteristics of the source water are also important in iodinated DBP formation. Kristiana et al. (2009) quantified iodine incorporation into different NOM fractions (hydrophobic (HPO), hydrophobic acid (HPOA), and transphilic (TPI)) for multiple source waters at varying iodide levels (50e300 mg/L) and bromide at 300 mg/L. Iodine incorporation varied from 10 to 64% and depended on iodide concentration and the NOM isolate. However, I-THMs were not distinguished from TOI. Hua and Reckhow (2007b) suggested that bromine and iodine were more reactive with low specific ultraviolet absorbance (low-SUVA254) and low molecular weight (MW) NOM to form I-THMs than high-SUVA, high MW NOM. However, this was not necessarily true for all iodinated DBPs (I-DBPs) because high MW components seemed to contain more TOI precursors than low MW NOM. It is unclear if the IDBPs formed were of high or low MW. While this study evaluated the effects of NOM characteristics, specifically SUVA and MW distribution, the concentrations of iodide were elevated (250 and 500 mg/L), while the corresponding bromide levels were much lower (<10 and 85 mg/L). Therefore, the results may not be applicable for the conditions encountered in natural waters. Since the occurrence of iodide and typical bromide to iodide ratios in natural waters have not been extensively characterized until recently, previous research studies have investigated I-THM formation and speciation at very high iodide concentrations (i.e., in excess of 100 mg/L) and without

13

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 e2 0

correspondingly increasing the bromide to levels higher than iodide (Leitner et al., 1998; Bichsel and von Gunten, 2000; Hua et al., 2006, 2007a,b; 2008). Therefore, I-THM formation for lower I
concentrations and higher Br
than I
scenarios is not adequately understood. Studying more realistic concentrations and ratios is important for a better understanding of ITHM formation and speciation in natural waters. The purpose of this study was to determine the impact of representative Br
and I
concentrations and contrasting Br
/ I
ratios on I-THM formation and speciation during chloramination. Two source waters with different natural organic matter characteristics (different SUVA254) were examined, and selected pH values were evaluated (6.0, 7.5, and 9.0). Typical Br
/I
concentrations and ratios were compared with some exaggerated cases. The four regulated THMs were also measured in addition to the six I-THMs for a more complete understanding of trihalomethane formation in drinking water treatment.

2.

Materials and methods

2.1.

Water treatment plant samples

Source waters and treated water samples were collected from two different water treatment plants (WTPs). Treated water samples were collected after coagulation, flocculation, and sedimentation. Both plants did not use any preoxidant until filtration, so the collected samples had never been exposed to any oxidant. The Hanahan WTP which utilizes a higherSUVA254 source water is located in Charleston, SC, while the SJWD WTP utilizing a slightly lower-SUVA254 source water is located in Lyman, SC. The plants were selected because of probable differences in their NOM characteristics, due to different anthropogenic and natural influences as a result of their location in the state (inland vs. coastal). Two different batches of water were utilized in this study. The first batch fulfilled the first two parts of this study (pH effect, and NOM characteristics), while the second batch was used to evaluate the importance of Br
/I
ratio. The selected characteristics of the waters are provided in Table 1.

2.2.

Preformed monochloramine

Preformed monochloramine was prepared by slowly (drop/ sec) combining a HOCl solution with (NH4)2SO4 solution (both at pH 8.5) to achieve a Cl2/NH3 weight ratio of 3.5, which is within the range of 3:1e5:1 used in practice. The resulting

solution did not have measurable free chlorine, and therefore was composed only of combined chlorine (chlorine combined with ammonia). Monochloramine doses that resulted in a residual concentration of 2.0 (þ/
0.4) mg/L NH2Cl after 24 h of contact time were applied. The concentrations of stock solutions and samples were measured using the N, N-diethylp-phenylenediamine (DPD) method (Standard Method 4500).

2.3.

Water preparation and monochloramine addition

Water samples were buffered with NaHCO3 (4 mM) and adjusted to pH 6, 7.5, or 9 with dilute HCl and NaOH. Suitable volumes of bromide and iodide stock solutions were spiked into the waters to achieve the target concentrations. In order to eliminate any slight deviations in bromide and iodide spiking, all replicates for one tested condition (e.g., 800/80) came from one spiked stock solution. The prepared samples were then added to 65 mL glass bottles and capped with no headspace. Small volumes of monochloramine stock solution were then spiked to the bottom of 65 mL reactors (no headspace) with long needle syringes to achieve the desired concentrations. The bottles were then capped and stirred vigorously on magnetic stir plates for 5 min. After 24 h at ambient room temperature, the samples’ pH and monochloramine residuals were measured. All samples were prepared in duplicate, and the error bars in the figures encompass the range for two duplicates.

2.4.

I-THM and THM analyses

Ten mL samples from the 65 mL reactors were collected for DBP analysis. A stoichiometric amount of sodium sulfite (8 mg/L) was added to quench monochloramine and prevent further unintended reactions between monochloramine and methyl-tert-butyl-ether (MTBE), the extraction solvent. Since quenching agents can also affect DBP measurements, the stability of THMs and I-THMs was evaluated. The results showed that THM and I-THM standards quenched with 8 mg/L sulfite were stable for at least 48 h, which was the longest holding time evaluated. Regardless, all samples were extracted within 1 h of quenching. I-THM and THM standards were prepared by spiking in a calculated volume of 104 mg/L THM and I-THM mixtures into 10 mL of distilled and deionized water. Individual I-THM species were purchased from Orchid Cellmark (Westminster, BC, Canada), and their purity was determined using gas chromatography (GC) equipped with a flame ionization

Table 1 e Selected characteristics of Charleston and SJWD raw and treated waters. Parameter DOC (mg/L) SUVA254 (L/mg-m) I
(mg/L) Br
(mg/L) a values for first batch. b values for second batch.

Charleston raw a

6.0 4.2 <2 100

(7.4) (3.7) (<2) (78)

b

Charleston treated 2.6 2.0 <2 100

(3.1) (2.0) (<2) (78)

SJWD raw

SJWD treated

1.7 (1.9) 2.5 (2.4) 5 (<2) 28 (22)

1.2 (1.4) 1.8 (1.7) 5 (<2) 28 (22)

14

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 e2 0

Results and discussion

3.1.

Effects of bromide/iodide concentrations and pH

When preformed monochloramine was added to Charleston raw and treated waters containing various levels of bromide (mg/L) and iodide (mg/L) (e.g., Br
/I
¼ ambient, 100/10, 200/20, and 800/80), I-THM formation was most favorable at pH 6 (Fig. 1). At this pH, I-THM formation depended on Br
/I
level, and it increased with increasing concentrations of bromide and iodide. When the same experimental conditions were evaluated at pH 7.5 and 9, I-THM formation was substantially less than pH 6, and often below the detection limit. This was especially true for the raw water because only minimal levels were detected at pH 7.5 for even the highest Br
/I
level (800/80) tested; ITHMs were not detected at pH 9. For the treated water, I-THM formation was observed at both pH 7.5 and 9, but again the concentrations were significantly less than for the corresponding conditions at pH 6. These results for Charleston water showed that I-THM formation from preformed monochloramine decreased with increasing pH. The formation of the regulated THMs followed the same trends as the iodinated ones. THM concentrations increased with increasing bromide concentration and decreasing pH (see Supplementary Information). THM formation at lower pH was likely due to the decomposition of monochloramine to chlorine (HOCl), which was then available to oxidize Br
to HOBr, an additional substituting agent, in addition to HOCl, involved in THM formation. At pH 6, I-THMs could still form in the presence of HOCl due to the competing reactions. While HOCl oxidizes HOI further to IO3
, HOI can also react with NOM to form I-THMs. In fact, I-THM formation from chlorine has been observed in

Raw water chloraminated 80

I-THMs (µg/L)

3.

practice and in the lab (Bunn et al., 1975; Brass et al., 1977, Bichsel and von Gunten, 2000; Hua et al., 2006; Hua and Reckhow, 2007a). Perhaps, the reactivity of HOCl with NOM may also play a role in I-THM formation by (i) altering the NOM and making it more reactive with bromine and iodine for substitution, (ii) diverting some of the HOCl from oxidizing HOI to IO3
, (iii) or a combination of both. While the findings for THMs were as expected, the results for I-THMs disagreed with previous research. In this study, ITHM formation decreased with increasing pH. Hua and Reckhow (2008) concluded the opposite for chloramination of a source water whose DOC and SUVA (DOC ¼ 5.0 mg/L, SUVA254 ¼ 3.4 L/mg-m) were relatively similar to those of Charleston raw water (DOC ¼ 6.0 mg/L, SUVA254 ¼ 4.2 L/mgm). In their study, I-THM formation increased with increasing pH. The primary differences between the two studies were the I
concentrations and the Br
/I
ratios. In their study, the concentration of I
and Br
were 1900 mg/L and 78 mg/L, respectively, (i.e., Br
/I
¼ 0.04). This was very different from 80 to 800 mg/L (i.e., Br
/I
¼ 10) evaluated in this study. Since the iodide concentration tested in Hua and Reckhow (2008) was elevated, HOI probably competed effectively with HOCl and HOBr at pH 6. Consequently, iodoform was the dominant I-THM species for all pH values (5, 7, 10). In contrast, in this study, iodoform was never the dominant species at pH 6 for 800 mg/L Br
and 80 mg/L I
(see Section 3.3.1). Increased I-THM

60

ambient 100/10 200/20 800/80

40

20

0 6

7.5 pH

9

Treated water chloraminated 80

60 I-THMs (µg/L)

detector (FID). THM standards were purchased from Sigma Aldrich. The samples and standards were then extracted and analyzed using USEPA Method 551.1 with minor modifications. 10 mL of MTBE was added to the 10 mL samples, and this was immediately followed by the addition of 3 g of sodium sulfate (salting out effect) and 1 g of cupric sulfate (visual phase separation). The extraction vials were then placed on a shaker table at 300 rpm for 30 min to dissolve the salts completely. The MTBE extract was analyzed on an Agilent 6890 GC equipped with a DB-1 column (J&W Scientific 30 m  0.25 mm  0.001 mm), and an electron capture detector (ECD). The GC temperature program was 35  C for 22 min, 10  C/min to 125  C and hold for 1 min, 30  C/min to 300  C and hold for 4 min. A 2 mL injection volume was used in splitless mode. The make-up and carrier gases were ultra-high purity (UHP) helium and UHP nitrogen. The total run time was 42 min. The injector temperature was set at 250  C, while the detector was set at 290  C. The minimum reporting limit for ITHM and THM measurements was 0.5 mg/L and 1.0 mg/L, respectively, for all species. Additional details about the materials and methods can be found elsewhere (Jones, 2009; Karanfil et al., 2011).

ambient 100/10 200/20 800/80

40

20

0 6

7.5 pH

9

Fig. 1 e I-THM formation in Charleston raw and treated waters. xxx/xx: BrL (mg/L)/IL (mg/L). Source: Karanfil et al., 2011. Formation of Halonitromethanes and IodoTrihalomethanes in Drinking Water ª2011 Water Research Foundation. Reprinted with permission.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 e2 0

formation with increasing pH in Hua and Reckhow (2008) was attributed to increases in iodoform formation. Perhaps its formation increased from pH 6 to 9 because as pH increases, hydrolysis becomes the rate-limiting step instead of enolization in the iodoform reaction (Bichsel and von Gunten, 2000). Briefly, OH
catalyzes the enolization reaction which converts carbonyl compounds to an enol which is reactive with HOI. At higher pH, since the enolization reaction and the iodination reaction are very fast (each iodine added to the enol 30 times faster than the previous one), complete iodination occurs before hydrolysis (conversion of iodinated enol to iodinated methane (Bichsel and von Gunten, 2000)). An increase in iodoform formation with increasing pH was also observed for Charleston treated water in this study (see Supplementary Information). In order to further examine the reasons for differences in the observed pH effect between Hua and Reckhow (2008) and this study, a source water with different NOM characteristics (i.e., lower DOC and lower-SUVA254) was tested. I-THM results for SJWD water (DOC ¼ 1.7 mg/L, SUVA254 ¼ 2.5 L/mg-m) are shown in Fig. 2. I-THM formation generally decreased with increasing pH, which corresponded to the results of Charleston water. However, for the highest Br
/I
concentrations (800/80), the trend did not hold. For the chloraminated

Fig. 2 e I-THM formation in SJWD raw and treated waters. xxx/xx:BrL (mg/L)/IL (mg/L). Source: Karanfil et al., 2011. Formation of Halonitromethanes and IodoTrihalomethanes in Drinking Water ª2011 Water Research Foundation. Reprinted with permission.

15

raw water, I-THM formation was relatively independent of pH, whereas for the treated water, there was higher formation at pH 7.5 and 9 than at 6. As expected, THM formation decreased with increased pH (see Supplementary Information). The results for SJWD treated water at the highest bromide and iodide level somewhat agreed with the findings of Hua and Reckhow (2008). However, the pH effect was more obvious in Hua and Reckhow (2008). Of the four waters spiked with 80 mg/L iodide, SJWD treated water had the highest I
/ DOC ratio (67 mg/mg). Since Hua and Reckhow (2008) tested a higher iodide level (1900 mg/L) and a higher I
/DOC (380 mg/ mg), it was hypothesized that I
/DOC ratio also plays an important role in the impact of pH on I-THM formation. Additional experiments were conducted, and the results will now be discussed.

3.2. Effects of SUVA254 and I
/DOC ratio on I-THM formation In order to more adequately compare Charleston and SJWD water, Charleston water was diluted to the same DOC as SJWD water. This allowed for a comparison of two waters with equivalent DOC levels but different SUVA254 values. Equivalent DOC levels also meant equivalent Br
/DOC and I
/DOC ratios. The results in Fig. 3 provided three important observations. The first observation was that I-THM formation at pH 7.5 and 9 was much more favorable in diluted Charleston water (higher I
/DOC) than the original water (lower I
/DOC). This confirmed that I
/DOC ratio is an important factor in I-THM formation. The second observation was that even after dilution, I-THM formation still decreased with increasing pH. This was different than SJWD water, as there was no decrease with increasing pH for SJWD water (equivalent DOC). Finally, the ITHM concentrations in dilute Charleston water (highSUVA254) (Fig. 3) were still less than the levels formed in SJWD water (low-SUVA254) (Fig. 2). Overall, it seems that I-THM formation for pH  7.5 is dependent on both the I
/DOC ratio and SUVA254 of the source water. In order to examine this idea further, the results for all waters including diluted Charleston raw and treated water were compiled in Fig. 4. The results demonstrated that I-THM yields depended on a combination of the I
/DOC ratio and SUVA254 of the water. From this limited data set, it appears that iodine is more easily incorporated by low-SUVA NOM components than high-SUVA components in forming I-THMs. This is consistent with the findings of Hua and Reckhow (2007b) that bromine and iodine are more reactive with small molecular weight (MW), hydrophilic NOM than high MW, hydrophobic NOM in forming THMs and I-THMs. Perhaps high-SUVA hydrophobic molecules require some preoxidation and breakdown of high MW NOM to form I-THM precursors before I-THM formation occurs. Conversely, perhaps lowSUVA NOM already contains smaller MW hydrophilic precursors, and therefore preoxidation may not be as critical in I-THM formation. This may explain, in part, why I-THM formation was more favorable at lower pH. The hydrolysis of monochloramine forms chlorine, a much stronger oxidant than monochloramine. The chlorine may have oxidized some of the NOM, producing compounds more reactive in forming I-

16

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 e2 0

THMs. This hypothesis is based on previous suggestions that monochloramine is more likely to form high MW than cleaved byproducts (Johnson and Jensen, 1986), while chlorine is more likely to form smaller byproducts such as THMs (Li et al., 2002). However, since only two waters with moderately different SUVA254 were studied, the effect of NOM characteristics on ITHM formation needs further investigation. An additional experiment at pH 7.5 showed that significant I-THM formation can still occur in Charleston raw water (high-SUVA), if sufficient iodide is present. When 800 mg/L of iodide (10 more than highest condition tested) was spiked into Charleston raw water (I
/DOC ¼ 165 mg/mg), approximately 200 mg/L of I-THMs were formed. This suggested that there was an iodide concentration threshold in this particular water that needed to be overcome before any I-THMs were formed. It is important to note that some I-THM formation in practice has been observed at much lower iodide levels (Krasner et al., 2006; Richardson et al., 2008). However, it is important to note that the experimental conditions in the laboratory and practice are not always the same. Nevertheless, perhaps the hypothesized I
/DOC ratio threshold may vary for different waters due to their unique NOM characteristics.

3.3. Effects of bromide/iodide concentration and ratio on I-THM speciation

Fig. 3 e I-THM formation in Charleston raw water and diluted raw water. *diluted to the same DOC as SJWD raw water. xxx/xx: BrL (mg/L)/IL (mg/L). Source: Karanfil et al., 2011. Formation of Halonitromethanes and IodoTrihalomethanes in Drinking Water ª2011 Water Research Foundation. Reprinted with permission.

60 Charleston raw* (SUVA = 4.2 L/mg-m) Charleston treated* (SUVA = 2.0 L/mg-m)

50 I-THM/DOC (µg/mg)

SJWD raw (SUVA = 2.5 L/mg-m) SJWD treated (SUVA = 1.8 L/mg-m)

40

30

20

10

0 0

20

40

60 80 I-/DOC (µg/mg)

100

120

140

Fig. 4 e The impact of SUVA and IL/DOC ratio on I-THM yields at pH 7.5. * e includes both original and diluted waters. Source: Karanfil et al., 2011. Formation of Halonitromethanes and Iodo-Trihalomethanes in Drinking Water ª2011 Water Research Foundation. Reprinted with permission.

To further examine I-THM formation, the speciation of ITHMs (6 species) and THMs (4 species) as a function of both bromide/iodide concentrations (ambient, 50/5 or 100/10, 200/ 20, 800/80 mg/L/mg/L) and bromide/iodide ratio (10:1, 1:2) was investigated.

3.3.1.

Br
/I
concentrations

When considering the four Br
/I
concentrations tested, ITHM formation was more favorable at lower pH. Therefore, the results for pH 6 are presented in Fig. 5. Results for SJWD raw water, Charleston raw water, and diluted Charleston water are shown as this allowed for a comparison between waters with different DOC values (SJWD water, Charleston water), and waters with equivalent DOC values, but different SUVA254 values (SJWD water, diluted Charleston water). For SJWD raw water, dichloroiodomethane (CHCl2I) and bromochloroiodomethane (CHBrClI) were detected at all bromide/iodide concentrations (ambient to 800/80). It was not until the 800/80 Br
/I
level that the more brominated and iodinated I-THM species were observed. Overall, there was not a dominant I-THM species at this level; however, bromodiiodomethane (CHBrI2) was in the highest concentration. Chlorodiiodomethane (CHClI2) was the only species not detected at pH 6. Conversely, for Charleston water, CHBrI2 was not the species in the greatest concentration at the highest Br
/I
level (800/80). CHCl2I and CHBrClI were the dominant species. Furthermore, the formation of dibromoiodomethane (CHBr2I) and iodoform (CHI3) were lower than in SJWD water. However, similarly to SJWD water, CHCl2I was the dominant species for the lower bromide and iodide levels. The speciation for diluted Charleston raw water was more similar to SJWD water than Charleston water. Therefore, it

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 e2 0

Concentration (nM)

150 125 100

CHI3 CHBrI2 CHClI2

75

CHBr2I

50

CHBrClI

25

CHCl2I

0 ambient 50/5 200/20 800/80 SJWD raw water chloraminated 150 Concentration (nM)

125 100 75 50 25 0

ambient 100/10 200/20 800/80 Charleston raw water chloraminated

Concentration (nM)

150 125 100 75 50 25 0

ambient 100/10 200/20 800/80 Diluted Charleston raw water chloraminated

Fig. 5 e I-THM speciation at pH 6 for different BrL/IL concentrations.

appears that dilution, which increased the Br
and I
/DOC ratios, impacted speciation. This suggested that I-THM speciation depends on Br
and I
/DOC ratios. I-THM speciation in these waters resembles the speciation observed in practice. In this study, CHCl2I and CHBrClI were the most important species at lower and more common bromide and iodide levels at pH 6. These were the two species most frequently encountered in a recent DBP occurrence study particularly at a plant that added chlorine and ammonia simultaneously (Krasner et al., 2006). However, some differences should be noted between the two studies. In the Krasner et al. study, chlorine and ammonia were either added (i) simultaneously as separate streams (one plant) or (ii) at separate locations in the plant, but not as preformed monochloramine. In contrast, in this study, preformed monochloramine was always used; additionally, the pH was lower. Nevertheless, some comparisons can be made between the field measurements and this study. Some free chlorine was

17

present at pH 6 in this study, and most WTPs use a short chlorine contact time for chloramination (except for a select few that add chlorine and ammonia simultaneously). Therefore, free chlorine’s presence may explain why CHCl2I and CHBrClI are usually the I-THM species most frequently detected. This would support the findings of another study in which these two I-THM species were the most important ones formed when waters with high concentrations of iodide were chlorinated (Hua et al., 2006). Similar to I-THMs, Br
/DOC was also important for THM speciation. There was a gradual shift from chlorinated to brominated THM species as bromide and iodide were increased (see Supplementary Information). This supports the previous findings in which there was a shift to more brominated species with higher Br
/DOC ratios (Symons et al., 1993). Specifically, in this study, chloroform (CHCl3) and dichlorobromomethane (CHCl2Br) were detected for the two lower Br
/ I
levels, and it was not until the bromide levels were higher that dibromochloromethane (CHBr2Cl) and bromoform (CHBr3) emerged as important species. In addition, dilution of the water resulted in an increase in CHBr3 formation and a decrease in CHCl3 formation.

3.3.2.

Br
/I
ratio

Since the Br
/I
ratio can vary in source waters, it is important to understand how the ratio will affect THM and I-THM speciation. Previous research investigated higher iodide than bromide scenarios, and rarely considered the more realistic higher bromide than iodide scenario. Therefore, results for a more typical Br
/I
ratio (10:1) were compared side-by-side with an opposite, lower Br
than I
scenario (1:2). The Br
/I
concentrations tested were 800/80 and 100/200 (mg/L/mg/L). I-THM speciation for both cases is shown in Fig. 6. For pH 6, when the Br
/I
ratio was 10:1, CHBr2I and CHBrI2 were the dominant species, and CHI3 was a minor species. However, for a Br
/I
ratio of 1:2, CHI3 was clearly the dominant species. There was also CHClI2, which was not observed for the 800/80 scenario. Overall, these results show that CHI3 will not be the dominant species at pH 6, if bromide is in a 10 times higher concentration than iodide. This is an important finding as Hua and Reckhow (2008) in a previous study concluded that iodoform was the dominant species for any pH (5e9). It appears that at lower pH, the decomposition of monochloramine forms chlorine, which can react with Br
(if present) to form HOBr, and brominated I-THM species may form. For pH 7.5, CHI3 was clearly the dominant I-THM specie regardless of ratio. Its formation substantially increased when iodide concentration was increased from 80 to 200 mg/L. Overall, there was considerably less formation of brominated species at pH 7.5 than at pH 6. This was anticipated as monochloramine is more stable at higher pH, and therefore HOBr would not be formed significantly. Additionally, bromamines and bromochloramines that may have formed from reactions with monochloramine are not as reactive with NOM as HOBr (Symons et al., 1998). For 800/80, CHBrI2 was observed, and its concentration was greater than the 100/200 case. The difference was expected as bromide concentration was 700 mg/L higher for the 800/80 scenario. Additionally, CHCl2I and CHClI2 formed for both ratios, but their concentrations were greater for the higher I
than Br
scenario.

18

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 e2 0

200 than 800/80 scenario. THM results were similar for Charleston water. There was an increase in CHCl3 formation as the ratio was changed (bromide was decreased). However, THM formation overall still decreased as bromide was reduced from 800 to 100 mg/L.

Concentration (nM)

400 CHI 3 CHBrI2

300

CHClI 2 CHBr2I

200

CHBrClI 100 0

4.

CHCl 2I

800/80

pH 6

100/200

Concentration (nM)

400 300 200 100 0

800/80

pH 7.5

100/200

Concentration (nM)

400

300

200

100

0

800/80

100/200 pH 9

Fig. 6 e I-THM speciation from chloramination of SJWD raw water for two BrL/IL.

For pH 9, the Br
/I
ratio did not significantly impact speciation because CHI3 was the dominant I-THM species for both ratios. The only other I-THMs detected at very small levels were CHCl2I and CHClI2. No brominated I-THM species were observed for either ratio. The results for Charleston water were similar (see Supplementary Information). I-THM formation was always higher for the opposite ratio (100/200) than the more typical 800/80 ratio. For pH 9, there was an absence of I-THM formation at 800/80, and it was not until the iodide was increased to 200 mg/L that CHI3 was observed. Therefore, it appears that I
/ DOC plays a key role in CHI3 formation. THM speciation was also analyzed, and it was found that there was a considerable effect of Br
/I
ratio on THM speciation at pH 6 (see Supplementary Information). When the bromide to iodide ratio was changed from 10:1 to 1:2, THM formation decreased substantially. For SJWD water, CHBr3 was the dominant species for 800/80, while it was not observed for 100/200. CHBr2Cl formation was less for the 100/

Conclusions

 I-THM formation from preformed monochloramine increased with increasing concentrations of bromide and iodide. It appears that higher I
/DOC ratios are required for significant I-THM formation for pH  7.5 than for pH 6.  Br
/I
concentrations and ratio play a key role in understanding the pH effect on I-THM formation. For an opposite ratio in which I
concentration is much higher than Br
, ITHM formation will likely increase with increasing pH. However, for the more realistic scenario, in which the Br
/I
is 10:1 and more common Br
/I
concentrations are evaluated, I-THM formation decreases with increasing pH. It is hypothesized that the decomposition of monochloramine at lower pH to form additional oxidants (HOCl) and substituting agents (HOCl, HOBr, HOI) is responsible for higher I-THM formation at lower pH.  It appears that SUVA254 may be an important factor in ITHM formation as high-SUVA waters required higher I
/ DOC ratios than low-SUVA waters for equivalent I-THM yields. Therefore, it seems that iodine is more reactive with hydrophilic than hydrophobic NOM components in forming I-THMs, which was shown previously. However, since only two waters were tested, more studies are needed to fully understand the effects of NOM characteristics on I-THM formation.  Br
/I
, Br
/DOC, and I
/DOC ratios are important for I-THM and THM speciation, especially at pH 6.  In this study, when higher I
/DOC ratios were tested, iodoform (CHI3) was not the dominant I-THM species formed from preformed monochloramine at pH 6 as was previously reported in another study. On the contrary, brominated ITHM species were more important. The inconsistent results between the two studies were due to the very different Br
/ I
ratios evaluated.  For lower and more realistic concentration of Br
and I
, CHCl2I and CHBrClI were the most important species formed at pH 6. Although there are some differences between the laboratory and practice (such as chloramination procedure), these two species were the most frequently detected at plants utilizing prechlorination followed by ammonia or the simultaneous addition of chlorine and ammonia to form monochloramine.  Iodoform (CHI3) was always the dominant species for pH  7.5. However, CHI3 formation only occurred for Br
and I
concentrations higher than those generally encountered in practice. This supports the finding of previous occurrence studies in which CHI3 was rarely reported.  These findings demonstrate that it is important to conduct experiments at representative Br
/I
concentrations and ratios (w10:1) to obtain practical results regarding I-THM formation and speciation in drinking waters.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 e2 0

Acknowledgements This study was supported by the Water Research Foundation project #4063. The authors appreciate the comments by the project advisory committee. We would also like to thank SJWD (Lyman, SC) and Hanahan (Charleston, SC) water utilities for providing their assistance in the collection of water samples for this research.

Appendix. Supplementary material Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.10.005.

references

Ali-Mohamed, A.Y., Jamali, S.G.A., 1989. Potentiometric determination of fluoride and iodide in natural and processes waters of Bahrain using ion-selective electrodes. Water Research 23, 659e662. Ates, N., Kitis, M., Yetis, U., 2007. Formation of chlorination byproducts in waters with low SUVA-correlations with SUVA and differential UV spectroscopy. Water Research 41, 4139e4414. Bichsel, Y., von Gunten, U., 1999. Oxidation of iodide and hypoiodous acid in the disinfection of natural waters. Environmental Science & Technology 33, 4040e4045. Bichsel, Y., von Gunten, U., 2000. Formation of iodotrihalomethanes during disinfection and oxidation of iodidecontaining waters. Environmental Science & Technology 34, 2784e2791. Brass, H.J., Feige, A., Halloran, T., Mellow, J.W., Munch, D., Thomas, R.F., 1977. The national organic monitoring survey: samplings and analyses for purgeable organic compounds. In: Pojasek, R.B. (Ed.), Drinking Water Quality Enhancement through Surface Water Protection. Ann Arbor Science Publishers, pp. 393e416. Bunn, W.W., Haas, B.B., Deane, E.R., Kleopfer, R.D., 1975. Formation of trihalomethanes by chlorination of surface water. Environmental Letters 10, 205e213. Cancho, B., Ventura, F., Galceran, M., Diaz, A., Ricart, S., 2000. Determination, synthesis, and survey of iodinated trihalomethanes in water treatment processes. Water Research 34, 3380e3390. Fuge, R., Johnson, C.C., 1986. The geochemistry of iodine: a review. Environmental Geochemistry & Health 8, 31e54. Hong, Y., Liu, S., Song, H., Karanfil, T., 2007. HAA formation during chloramination significance of monochloramine’s direct reaction with DOM. Journal of the American Water Works Association 99, 57e69. Hua, G., Reckhow, D.A., 2007a. Comparison of disinfection byproduct formation from chlorine and alternative disinfectants. Water Research 41, 1667e1678. Hua, G., Reckhow, D.A., 2007b. Characterization of disinfection by-product precursors based on hydrophobicity and molecular size. Environmental Science & Technology 41, 3309e3315. Hua, G., Reckhow, D.A., 2008. DBP formation during chlorination and chloramination: effect of reaction time, pH, dosage, and temperature. Journal of the American Water Works Association 100, 82e95.

19

Hua, G., Reckhow, D.A., Kim, J., 2006. Effect of bromide and iodide ions on the formation and speciation of disinfection byproducts during chlorination. Environmental Science & Technology 40, 3050e3056. Johnson, J.D., Jensen, J.N., 1986. THM and TOX formation-routes, rates, and precursors. Journal of the American Water Works Association 78, 156e162. Jones, D. B., 2009. The formation and control of iodinated trihalomethanes in drinking water treatment. Master’s thesis, Clemson University, Department of Environmental Engineering and Earth Sciences. Clemson, SC, USA. Karanfil, T., Hu, J., Jones, D.B., Addison, J.W., Song, H., 2011. Formation of Halonitromethanes and Iodo-trihalomethanes in Drinking Water. Water Research Foundation, Denver, CO. Kitis, M., Karanfil, T., Kilduff, J.E., Wigton, A., 2001. The reactivity of natural organic matter to disinfection byproducts formation and its relation to specific ultraviolet absorbance. Water Science & Technology 43, 9e16. Krasner, S.W., McGuire, M.J., Jacengelo, J.G., Patania, N.L., Reagan, K.M., Marco Aieta, E., 1989. Occurrence of disinfection by-products in US drinking water. Journal of the American Water Works Association 81, 41e53. Krasner, S.W., Weinberg, H.S., Richardson, S.D., Pastor, S.J., Chinn, R., Sclimenti, M.J., Onstad, G.D., Thruston, A.D., 2006. Occurrence of a new generation of disinfection byproducts. Environmental Science & Technology 40, 7175e7185. Kristiana, I., Gallard, H., Joll, C., Croue, J.P., 2009. The formation of halogen-specific TOX from chlorination and chloramination of natural organic matter isolates. Water Research 43, 4177e4186. Leitner, N.K.V., Vessella, J., Dore, M., Legube, B., 1998. Chlorination and formation of organoiodinated compounds: the important role of ammonia. Environmental Science & Technology 32, 1680e1685. Li, C.W., Benjamin, M.M., Korshin, G.V., 2002. The relationship between TOX formation and spectral changes accompanying chlorination of pre-concentrated or fractionated NOM. Water Research 36, 3265e3272. Muramatsu, Y., Wedepohl, K.H., 1998. The distribution of iodine in the Earth’s crust. Chemical Geology 147, 201e216. Moran, J.E., Oktay, S.D., Santschi, P.H., 2002. Sources of iodine and iodine 129 in rivers. Water Resources Research 38, 1149e1159. Neal, C., Smith, C.J., Walls, J., Billingham, P., Hill, S., Neal, M., 1990. Comments on the hydrochemical regulation of the halogen elements in rainfall, stemflow, throughfall, and stream waters at an acidic forested area in mid-wales. The Science of the Total Environment 91, 1e11. Plewa, M.J., Wagner, E.D., Muellner, M.G., Hsu, K.M., Richardson, S.D., 2008. Comparative mammalian cell toxicity of nitrogen-containing disinfection by-products and carbonaceous disinfection by-products. Washington, D.C. In: Karanfil, T., Krasner, S.W., Westerhoff, P., Xie, Y. (Eds.), Disinfection By-products in Drinking Water: Occurrence, Formation, Health Effects, and Control. American Chemical Society Symposium Series, 995, pp. 36e50. 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, M.J., 2008. Occurrence and mammalian cell toxicity of iodinated disinfection byproducts in drinking water. Environmental Science & Technology 42, 8330e8338. Rook, J.J., 1974. Formation of haloforms during chlorination of natural waters. Water Treatment & Examination 23, 234e243. Seidel, C.J., McGuire, M.J., Summers, R.S., Via, S., 2005. Have utilities switched to chloramines. Journal of the American Water Works Association 97, 87e97.

20

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 e2 0

Steinberg, S.M., Kimble, G.M., Schmett, G.T., Emerson, D.W., Turner, M.F., Rudin, M., 2008. Abiotic reaction of iodate with sphagnum peat and other natural organic matter. Journal of Radioanalytical and Nuclear Chemistry 277, 185e191. Symons, J.M., Krasner, S.W., Simms, L.A., Sclimenti, M.J., 1993. Measurement of THM and precursor concentrations revisited: the effect of bromide ion. Journal of the American Water Works Association 85, 51e62. Symons, J.M., Xia, R., Speitel, G.E., Diehl, A.C., Hwang, C.J., Krasner, S.W., Barrett, S.E., 1998. Factors affecting disinfection by-product formation during chloramination. AWWA Research Foundation, USA.

USEPA, Dec. 16, 1998. National primary drinking water regulations: disinfectants and disinfection byproducts; Final Rule. Federal Register 63 (241), 69390. USEPA, May 6, 2006. National primary drinking water regulations: stage 2 disinfectants and disinfection byproducts Rule. Federal Register 71 FR (388). von Gunten, U., 2003. Ozonation of drinking water: Part II. Disinfection and byproduct formation in presence of bromide, iodide, or chlorine. Water Research 37, 1469e1487. Whitehead, D.C., 1984. The distributions and transformations of iodine in the environment. Environment International 10, 321e339.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 1 e3 2

Available online at www.sciencedirect.com

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

Multi-cycle bioregeneration of spent perchlorate-containing macroporous selective anion-exchange resin Mohamadali Sharbatmaleki a,1, Jacimaria R. Batista b,* a

Institute for Energy and the Environment, New Mexico State University (NMSU), PO Box 30001, MSC WERC, EC III, Suite 300 South, Las Cruces, NM 88003-8001, USA b Department of Civil and Environmental Engineering, University of Nevada Las Vegas (UNLV), 4505 Maryland Parkway, Las Vegas, NV 89154-4015, USA

article info

abstract

Article history:

Ion exchange using perchlorate-selective resin is possibly the most feasible technology for

Received 23 May 2011

perchlorate removal from water. However, in current water treatment applications,

Received in revised form

selective resins are used once and then incinerated, making the ion-exchange process

29 September 2011

economically and environmentally unsustainable. A new concept has been developed

Accepted 11 October 2011

involving the biological regeneration of resin-containing perchlorate. This concept involves

Available online 25 October 2011

directly contacting perchlorate-containing resins with a perchlorate-reducing microbial culture. In this research, the feasibility of multi-cycle loading and bioregeneration of

Keywords:

a macroporous perchlorate-selective resin was investigated. Loading and bioregeneration

Ion exchange

cycles were performed, using a bench-scale fermenter and a fluidized bed reactor followed

Bioregeneration

by fouling removal and disinfection of the resin. The results revealed that selective mac-

Perchlorate

roporous resin can be employed successfully in a consecutive loading-bioregeneration ion-

Biodegradation

exchange process. Loss of resin capacity stabilized after a few cycles of bioregeneration,

Resin reuse

indicating that the number of loading and bioregeneration cycles that can be performed is likely greater than the five cycles tested. The results also revealed that most of the capacity loss in the resin is due to perchlorate buildup from previous regeneration cycles. The results further indicated that as the bioregeneration progresses, clogging of the resin pores results in strong mass transfer limitation in the bioregeneration process. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Perchlorate (ClO
4 ) contamination has been detected in several surface and groundwater sources in 26 states of the United States (U.S.), particularly in the arid Southwestern region (USEPA, 2009; Brandhuber and Clark, 2005), and it has been on the U.S. Environmental Protection Agency’s (EPA) drinking water Contaminant Candidate List (CCL) since 1998 (Brandhuber and Clark, 2005). Most of the perchlorate contamination in the environment is due to rocket fuel

manufacturing and use (Urbansky et al., 2000). Perchlorate in the human body interferes with the natural process of iodine uptake by the thyroid gland, inhibits thyroid hormone production, and causes iodine accumulation in the gland (Wolff, 1998; Kirk, 2006). In the U.S., an Interim Drinking Water Health Advisory level for perchlorate has been issued as 15 micrograms per liter (mg/L) in 2005 (NRC, 2005). Biological reduction and ion exchange (IX) are the most effective technologies to remove perchlorate from water (Logan et al., 2001; Gingras and Batista, 2002; Van Ginkel et al.,

* Corresponding author. Tel.: þ1 702 895 1585; fax: þ1 702 895 3936. E-mail addresses: [email protected] (M. Sharbatmaleki), [email protected], [email protected] (J.R. Batista). 1 Tel.: þ1 575 646 5045; fax: þ1 575 646 5474. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.012

22

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 1 e3 2

2008). Although bioremediation is a cost-effective technology for perchlorate removal, it is not an efficient technology for low concentrations because degradation rates slow down with decreasing perchlorate concentration; in other words, perchlorate biodegradation has relatively high half-saturation constants, as seen in Table 1 (Logan et al., 2001; Dudley et al., 2008). Ion exchange is currently the technology of choice to remove low concentrations of perchlorate from drinking water (Gingras and Batista, 2002; Lehman et al., 2008). Morphologically, perchlorate-selective resins can be categorized into two types: gel and macroporous. The average pore size of gel-type resins is about 0.0005 mm, while macroporous resins have an average pore size of about 0.6 mm (Kun and Kunin, 1968; Dale et al., 2001). The percent crosslinking of the backbone polymer chains varies in gel-type and macroporous resins. Gel-type resins have about 4e10% crosslinking, whereas macroporous resins have 20e25% crosslinking in average (MWH, 2005). Due to their stability, resistance to oxidation, and less vulnerability to fouling, the use of macroporous resins compared to gel-type resins is expanding (Clifford, 1999; Li and SenGupta, 2000). Throughout the U.S., there are many pilot and full-scale water supply treatment plants, both operational and under construction, that use IX technology to remove perchlorate from drinking water. These plants have capacities varying from 23 to 55,000 m3/day (4e10,000 gpm) and influent perchlorate concentrations varying from 7 to 350 mg/L. These plants can remove perchlorate to <4 mg/L (NASA, 2006). In recent years, IX resins with high affinity for perchlorate ion, known as perchlorate-selective resins, have been manufactured (Seidel et al., 2006). Even though these resins can treat a large number of bed volumes (BVs) of water before perchlorate breakthrough occurs, regeneration by the traditional brine desorption technique cannot be employed; as a result, the spent perchlorateselective resin either is incinerated or disposed in a landfill after one time use. Incineration of the resin produces greenhouse gases, and disposal in a landfill presents the potential for re-contamination of the environment with perchlorate. Although a regeneration method has been developed that employs FeCl3eHCl solution as the regenerant for one type of perchlorate-selective gel-type resin (Gu et al., 2001), regeneration of most commercially available perchlorate-selective resins using NaCl brine is not feasible and is not currently practiced.

Table 1 e Half-saturation constant and maximum perchlorate utilization rate for degradation of free perchlorate ion in water. Culture

Kinetic parameters
1

PDX KJ INS ABL1 SN1A RC1 PC 1 HCAP-C Mixed culture

qmax (d )

Ks (mg/L)

0.41 1.32 4.34 5.42 4.60 6.00 3.09 4.39 0.49

12  4 33  9 18 4.8 2.2 12 0.14 76.6 <0.1

Reference

Logan et al., 2001 Logan et al., 2001 Waller et al., 2004 Waller et al., 2004 Waller et al., 2004 Waller et al., 2004 Nerenberg et al., 2006 Dudley et al., 2008 Wang et al., 2008

Resin bioregeneration as a new concept in IX technology has been developed, and a patent has been filed (Batista, 2006). This concept is based on directly contacting IX resin-containing perchlorate with a perchlorate-reducing microbial culture under anoxic/anaerobic conditions. The feasibility of this concept has been reported on perchlorate-selective and non-selective gel-type resins (Batista et al., 2007b; Venkatesan et al., 2010; Venkatesan and Batista, 2011). Although the biodegradation of free perchlorate ions in water has been well studied (Logan, 1998; Coates and Achenbach, 2004), the biological reduction of attachedperchlorate ions to a medium such as IX resins has only recently been initiated (Wang et al., 2008, 2009; Venkatesan et al., 2010; Venkatesan and Batista, 2011). Investigation has shown that biological degradation of resin-attached perchlorate has a slower degradation rate compared to biological degradation of free perchlorate (Venkatesan et al., 2010). A main concern in resin bioregeneration is capacity loss due to biological fouling of the resin beads (Batista et al., 2007a). Macroporous resins are less susceptible to biofouling than gel-type resins. Although resin bioregeneration has several environmental benefits, in order to be economically viable, the cost of purchasing fresh resin must be equal or greater than costs of resin bioregeneration. Therefore, resin bioregeneration is economical only if the process can be repeated for several consecutive exhaustionebioregeneration cycles. Recently, bioregeneration of gel-type resins loaded with perchlorate has been investigated (Venkatesan et al., 2010; Venkatesan and Batista, 2011). To date, however, bioregeneration of macroporous resins has not been fully investigated. In this research, multi-cycle bioregeneration of macroporous perchlorate-selective resin, which is thought to be less susceptible to fouling, has been investigated. The specific objectives of this research were to: (1) evaluate the feasibility of multi-cycle bioregeneration of macroporous IX perchlorate-selective resin, (2) estimate the capacity loss after the multi-cycle bioregeneration process, (3) evaluate the influence of resin fouling on the mass transfer mechanism of the bioregeneration process, and (4) compare the degradation rate of resin-attached perchlorate in the macroporous resin with the degradation rate of free perchlorate in water. The findings of this research can be useful in determining potential capacity loss during bioregeneration. Capacity loss is a major factor influencing economics and full-scale application of bioregeneration as a technology for perchlorate treatment.

2.

Materials and methods

2.1.

Experimental approach

Resin bioregeneration system that includes a bioreactor and a fluidized bed reactor (FBR) was designed and built. The resin was loaded batchwise and placed in the FBR (Fig. 1). A perchlorate-reducing microbial culture present in the bioreactor was passed through the FBR containing the resin. At predetermined time intervals, resin samples were taken from the FBR, and residual perchlorate concentration in the resin was

23

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 1 e3 2

Fig. 1 e Schematic of the resin bioregeneration system technology. Left: loading of resin with perchlorate during water treatment. Right: bioregeneration of perchlorate laden resin using a microbial culture.

determined. An oxygen Parr-bomb method was developed to determine the amount of resin-attached perchlorate remaining with time. Following bioregeneration, the resin was defouled and disinfected with sodium hypochlorite, and the resin capacity was measured.

2.2.

Experimental setup

The bioregeneration system consisted of a 10-gallon (37.9 L) polyethylene bioreactor and a 3-inch diameter (7.6 cm)  50inch (127 cm) tall plexiglass FBR. Three resin-sampling ports were drilled along the FBR column (Fig. 1). The bioreactor for cultivation of perchlorate-reducing bacteria (PRB) was equipped with oxidation reduction potential (ORP), pH, dissolved oxygen (DO), and temperature probes (GF-Signet, El Monte, CA). A stirrer was used to gently mix the bacterial enrichment culture in the bioreactor at 25e30 rpm. Before each bioregeneration cycle, the bioreactor was purged with nitrogen gas for 30 min to establish anaerobic conditions. Perchlorateloaded resin was placed in the FBR column, and the bacterial enrichment culture was fed up-flow by using a peristaltic pump to achieve 30e40% expansion of the resin bed. The microbial culture was then recirculated back to the bioreactor, where acetate (the electron donor), nutrients, and minerals were added to sustain bacterial growth. After the bioregeneration cycle, the resin was rinsed, defouled, and disinfected.

2.3.

a mineral/nutrient/buffer mixture (Table 2). Although resinattached nitrate was available in the FBR as the nitrogen source for PRB, NH4H2PO4 was added to the microbial culture to supply the nitrogen needed throughout the bioregeneration process (Table 2). A 3:1 molar ratio for acetate/perchlorate was used during the enrichment phase. The PRB enrichment culture has been characterized using two different molecular methods, Restriction Fragment Length Polymorphism (RFLP) and 16S rRNA sequencing (Kesterson et al., 2005). Characterization of the PRB enrichment culture revealed that the culture was composed of at least 6 bacterial genera, two of which were able to degrade perchlorate as an electron acceptor. All six isolates were gram-negative and facultative anaerobic bacteria. The bacterial genera that have been identified in the culture include Pseudomonas, Azospira (formerly

Table 2 e Nutrient and buffer stock solution for feeding the culture. Stock solution

Components

Nutrients

MgSO4$7H2O EDTA ZnSO4$7H2O CaCl2$2H2O FeSO4$7H2O Na2MoO4$2H2O CuSO4$5H2O CoCl2$6H2O MnCl2$4H2O NiCl2$6H2O NaSeO3 H3BO3

Buffer

K2HPO4 NaH2PO4$H2O NH4H2PO4

Composition of the enrichment culture

A mixed enrichment bacterial culture was used in this research. The sources of PRB inocula were waters from Lake Mead and the Las Vegas Wash in southern Nevada. These areas have been contaminated with ammonium perchlorate for the past five decades and were presumed to be likely sources of PRB. The PRB were enriched under anoxic conditions in a broth containing acetate, perchlorate, and

Concentration of component (g/L) 5.500 0.300 0.200 0.100 0.400 0.040 0.020 0.040 0.100 0.010 0.010 0.060 155.000 97.783 50.000

24

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 1 e3 2

Dechlorosoma), Dechloromonas, Aeromonas, and Rhizobium, all of which are typically present in soils and waters (Kesterson et al., 2005).

2.4.

Resin loading

The resin used in this research was ResinTech SIR-110MACRO, which was specially manufactured for this research. SIR-1110 MACRO is a macroporous perchlorateselective strong base anion-exchange resin with tri-n-butylamine ((C4H9)3N) functional groups having a nominal resin capacity of 0.6 eq/L (ResinTech, 2008). The resin bead size ranges from 1.18 to 0.30 mm (#16 to #50 U.S. standard mesh size). For the bioregeneration tests, the resin was loaded batchwise. For all five cycles, the composition of the synthetic solution used to load the resin and residual concentrations are given in Table 3. The feed water composition was chosen to simulate waters contaminated with high perchlorate concentrations, typical of industrial sites. A volume of 1300 mL of resin was selected for use through five cycles of loading, bioregeneration, fouling removal, and disinfection. In the loading step, the initial volume of resin was measured and added to a glass bottle. An equal volume of feed solution was prepared and added to the bottle, as indicated in Table 3. The bottle was then placed in a rotary mixer (Associated Design MFG Co., Alexandria, VA) at 30 rpm and 22  2  C. After 24 h, the mixture was allowed to settle for

3 min. The solution was then sampled and analyzed for anions by ion chromatography (IC). To remove residual anions not attached to the resin, the loaded resin was rinsed six times with 2 BVs of de-ionized (DI) water. Conductivity measurements of the rinsate solution showed that rinsing six times with 2 BVs of DI water was effective in removing the excess ions from the loaded resin. After the rinsing, the resin was stored in the refrigerator until the start of the bioregeneration experiments.

2.5.

Resin bioregeneration

Resin bioregeneration in each cycle was performed until the residual perchlorate concentration in the resin reached a constant value. The bioregeneration processes ran for a period of 9e14 days. Daily, 4 mL of resin sample were taken from the ports along the FBR column by using a 20-mL syringe. The resin sample was then rinsed six times with 5 BVs of DI water; then, the sample was labeled and stored in the refrigerator. The residual perchlorate in the samples was measured using an oxygen bomb method developed specifically for this research. During the bioregeneration cycles, the effluent line from FBR to the bioreactor was monitored daily for perchlorate, and no perchlorate was detected. Chemical oxygen demand (COD), suspended solids (SS), and conductivity analysis were performed on a daily basis on the samples taken from the bioreactor. Conductivity measurements were

Table 3 e Resin loading data through the five cycles of the experiment. Cycle #

Anion

Added feed conc.,a mg/L

Residual conc. after 24 h, mg/L

Capacity occupied, g/Lresin

Capacity occupied, meq/Lresin

Percent capacity occupiedb

Cycle 1

ClO
4 NO
3 Cl
SO
2 4 HCO
3

11,061 491 514 532 500

0.26 12.6 >2100 251 800

11.06 0.48 N/A 0.28 N/A

111.16 7.71 N/A 5.85 N/A

17.5% 1.2% N/A 0.9% N/A

Cycle 2

ClO
4 NO
3 Cl
SO
2 4 HCO
3

10,151 499 495 532 500

0.26 15.4 2988 133 750

10.15 0.48 N/A 0.40 N/A

102.01 7.81 N/A 8.31 N/A

16.1% 1.2% N/A 1.3% N/A

Cycle 3

ClO
4 NO
3 Cl
SO
2 4 HCO
3

10,229 468 511 506 500

0.50 12.4 2891 161 690

10.23 0.46 N/A 0.34 N/A

102.79 7.35 N/A 7.17 N/A

16.2% 1.2% N/A 1.1% N/A

Cycle 4

ClO
4 NO
3 Cl
SO
2 4 HCO
3

10,082 464 508 516 500

1.31 12.8 2886 163 635

10.08 0.45 N/A 0.35 N/A

101.31 7.27 N/A 7.35 N/A

16.0% 1.1% N/A 1.2% N/A

Cycle 5

ClO
4 NO
3 Cl
SO
2 4 HCO
3

10,018 458 506 495 500

1.06 7.5 2964 199 610

10.02 0.45 N/A 0.30 N/A

100.67 7.27 N/A 6.16 N/A

15.9% 1.1% N/A 1.0% N/A

a Sodium salts of perchlorate, nitrate, sulfate, chloride, and bicarbonate were used for the resin loading. b Total capacity ¼ 0.64 eq/Lresin (does not account for perchlorate accumulated from cycle to cycle).

25

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 1 e3 2

performed to control the level of salinity accumulated in the reactor as degradation proceeds. Salinity is known to negatively affect perchlorate degradation (Gingras and Batista, 2002). Throughout the bioregeneration cycles, the volume of the microbial solution was 30 L. The SS in the bioreactor was maintained at 1000e2000 mg/L. Whenever the SS dropped too much owing to electron acceptor limitation, PRB cells originating from the stock enrichment culture were concentrated by centrifugation and added to the bioreactor in order to increase the biomass. The amount of electron donor (acetate) in the bioreactor was measured as COD, and was maintained above 1500 mg/L. After the bioregeneration process was complete, the microbial culture in the FBR was transferred from the bottom of FBR to the bioreactor by using a peristaltic pump. To remove microbial cells, the resin was rinsed five times with one BV of DI water. The resin was then submitted to a fouling removal procedure developed for this research.

2.6.

Bio-fouling removal and disinfection

Biological fouling is a consequence of bioregeneration. The procedure used for fouling removal was developed in the Environmental Engineering Laboratory at the University of Nevada, Las Vegas (UNLV) (Batista et al., 2007a). The fouling removal procedure includes four consecutive steps, as shown in Table 4. First, 1.5 BVs of fouling removal reagent (12% NaCl and 2% NaOH mixture) was pumped up-flow to the FBR column containing the resin. After 12 h, the rinsate solution was decanted and sampled. Second, 1.5 BVs of fouling removal reagent (12% NaCl and 2% NaOH mixture) was pumped up-flow to the FBR column; after 4 h, the rinsate solution was decanted and sampled. Third, 1.5 BVs of 12% NaCl was pumped up-flow to the FBR column, and after 2 h, the rinsate solution was decanted and sampled. Last, the resin was rinsed with 3 BVs of DI water to remove an excessive amount of ions. The rinsate solution samples collected after each step of fouling removal were submitted for COD analysis. Following fouling removal, the resin was disinfected using 1.5 BVs of sodium hypochlorite (100 mg/L as free chlorine). After 15e20 min contact time, the solution was decanted. The resin then was rinsed until no residual chlorine was detected in the rinsate. Residual chlorine was measured with a Capital Controls Series 17T2000 amperometric titrator (Steven Trent Services, Ft. Washington, PA). Results showed that rinsing with 2 BVs of DI water six times was sufficient to remove all the residual chlorine from resin. After the disinfection step,

total coliform analysis was performed on the DI water rinsate using the IDEXX Quanti-Tray method (IDEXX Laboratories, Inc., Westbrook, ME). No coliform bacteria were detected in the rinsate solutions collected from the resin after the disinfection procedure. The bioregenerated, defouled, disinfected, and rinsed resin was then submitted to the loading process to initiate the next bioregeneration cycle. In the procedure described above, many of the steps involve the use of NaCl to assist with fouling removal. One may argue that this salt could be used directly to regenerate the resins. However, perchlorate-selective resins are poorly regenerated with NaCl. Bioregeneration may be one of a few alternatives that would allow for reuse of these resins.

2.7.

Residual perchlorate analysis

For this research, a method has been developed to measure the resin-attached residual perchlorate remaining as the bioregeneration progresses. In this method, small samples of resin are ignited in an oxygen Parr bomb (Parr Instruments, Moline, IL) and the perchlorate present is converted to chlo
ride ion (resin-ClO
4 þ O2 / CO2 þ Cl ). The chloride concentration is then measured using IC. The method’s error is about 4%. Prior to ignition in the Parr bomb, 1 mL of resin sample was placed in 100 mL of concentrated nitrate solution and mixed for 24 h in a rotary shaker. This was performed to assure that all chloride ions attached to the resin functional groups were replaced by a nitrate ion. Thus, chloride ions detected after ignition would be associated with perchlorate load. The resin was then separated from the supernatant by filtering through a filter paper and rinsed six times with DI water to remove anions not attached to the resin. Next, the resin was dried in a gravity oven (VWR, West Chester, PA) at 105  C for 1 h. Then, 50e200 mg of the dried resin sample and 400 mg of paraffin oil were weighed and placed in the Parr bomb. Ten mL (10 mL) of 35 mM NaOH and 3 mL of 3% H2O2 were also added to the Parr bomb. The Parr bomb was then capped and pressurized with 500 psi (30e35 atm) of oxygen gas. The resin sample was then ignited using 10 cm of nickel fuse wire. During the ignition, perchlorate was converted to chloride ion. The Parr-bomb cylinder was opened, and its contents were rinsed with small portions of DI water and transferred to a 250 mL volumetric flask. Chloride ion was then determined in the resulting ignition solution using ion chromatography. Because combustion of one mole of perchlorate would produce one mole of chloride, the amount of perchlorate in the resin

Table 4 e Fouling removal and disinfection procedure used after bioregeneration. Fouling removal reagent

Applied volume

Retention time

Fouling removal procedure

12% NaCl þ 2% NaOH 12% NaCl þ 2% NaOH 12% NaCl DI water rinse

1.5 BV 1.5 BV 1.5 BV 3.0 BV

12 h 4h 2h N/A

Disinfection procedure

100 mg/L free chlorine using sodium hypochlorite DI water rinse

1.5 BV

15e30 min

Until no residual chlorine was detected in the rinsate

26

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 1 e3 2

sample could be calculated. For quality assurance, a sample of ultimate coal (Alpha Resources Inc., Stevensville, MI) with a known chloride content was used as the chloride standard. In this case, a known amount of coal was ignited in the Parr bomb, and the chloride concentration in the product solution was measured. The measurements were 99.73% accurate.

2.8.

Oxygen meter (YSI, Inc., Warm Springs, OH). The percentage transmittance was measured using a Hach DR 5000 Spectrophotometer at the wavelength of 600 nm. DI water with a resistivity of 17.5 MU cm was obtained from a Barnstead water purification system (Dubuque, IA) and used in all steps, unless otherwise noted. All the salts used were ACS grade and obtained from VWR (West Chester, PA).

Resin capacity measurement 2.10.

It was expected that bioregeneration would result in decreased resin capacity after each cycle. Therefore, the total capacity of fresh and bioregenerated resin was measured. To measure the capacity, 15 mL of wet resin was placed in a pipette filled with a stopcock. To convert the resin to the chloride form, 1 L of 4.0% HCl was passed through the resin bed. Next, the resin was rinsed with 1 L of DI water to remove interstitial chloride. To replace the chloride ions with nitrate, 1 L of 1.0 N NaNO3 solution was passed through the resin. The effluent from the NaNO3 rinse was collected and titrated with AgNO3 in order to measure the chloride concentration. Theoretically, each mole of detected chloride in the effluent corresponds one mole of nitrate exchanged by the active functional groups. The resin capacity in equivalents/L was calculated using the chloride measurements.

2.9.

Scanning electron microscopy (SEM)

Fresh resin and bioregenerated resin, sampled at the end of Cycle 5, were rinsed with DI water and air-dried for 24 h at 22  2  C. The resins were affixed with glue to a stainless steel disk and allowed to air-dry for 1 h. Then, to expose the inner portion of the beads for imaging, the resin samples were polished using soft sandpaper. The samples were rinsed with DI water and air-dried for 24 h at 22  2  C. Scanning electron microscopy imaging of the resin samples was performed using Jeol JSM-7500F SEM (JEOL Ltd., Tokyo, Japan) and employing a secondary electron detector at 1.00 kV. After SEM imaging, the bioregenerated resin sample was then soaked in 100 mg/L sodium hypochlorite solution for 20 min. Then, the sample was rinsed with DI water, air-dried for 24 h at 22  2  C, and submitted for SEM imaging again.

Chemicals and analyses

All perchlorate concentrations and low concentrations of chloride were measured using a Dionex ICS-2000 IC (Dionex Corporation, Sunnyvale, CA), consisting of an Ion SuppressorULTRA II (4 mm), IonPac AS16 (4 mm) analytical, AG16 (4 mm) guard columns, and an AS16 autosampler. For perchlorate, EPA method 314.0 was used with a current of 100 mA and an NaOH concentration of 35 mM with a flow rate of 1.0 mL/min. A calibration curve was established using perchlorate standard solutions with concentrations between 5 and 100 mg/L. A coefficient of determination of 99.97% was used for calibration. Similarly, to measure low-concentration chloride ions, a current of 100 mA and an NaOH concentration of 35 mM with a flow rate of 1.0 mL/min were used. Calibration curves for low-concentration chloride were plotted with standard solutions having concentrations between 100 and 500 mg/L, using a coefficient of determination of 99.97%. For nitrate, sulfate, and high concentrations of chloride anions, IonPac AS20 (4 mm) analytical and AG16 (4 mm) guard columns were used on the same IC with a current of 110 mA, an NaOH concentration of 30 mM, and a flow rate of 1.0 mL/min. The calibration curve for nitrate, sulfate, and high concentrations of chloride anions measurement was prepared for concentrations between 1 and 10 mg/L and a 99.99% coefficient of determination. Bicarbonate and suspended solids were measured according to Standard Methods 4500-CO2-D and 2540-D, respectively (Eaton et al., 2005). Chemical oxygen demand (COD) was measured using high range (0e1500 mg/L) Hach COD digestion vials (Hach Co., Loveland, CO). The pH values were measured using a Fisher Scientific model AR25 pH meter (Springfield, CO). Conductivity was determined using a YSI Model # 30/10 FT conductivity meter (YSI, Inc., Warm Springs, OH). The DO in the FBR was measured using a YSI Model 58 Dissolved

3.

Results

3.1.

Bioreactor performance

Typical values of COD, SS, pH, and conductivity of the microbial enrichment culture that were present in the bioreactor through the bioregeneration cycles appear in Fig. 2aed. The SS varied between 1000 and 2000 mg/L during the five cycles of bioregeneration. As it is shown in Fig. 2a and c, there was a decrease in the amount of biomass in the fermenter as bioregeneration progressed. The likely reason for this decrease was a shortage of electron acceptors for the microbial culture. The electron donor, acetate, was available in the system. At some points during bioregeneration, for example, Day 2 and Day 3, as shown in Fig. 2a, concentrated PRB cells originating from the master seed cultures were added to the fermenter in order to increase the biomass concentration. The pH of the bioreactor was maintained between 7.1 and 8.0 using a phosphate buffer. There was an increase in the pH values as the bioregeneration proceeded (Fig. 2b and d). The observed increase in pH is likely due to biodegradation of resin-attached nitrate, that is, denitrification. The loaded resin contained several anions, including nitrate, and the bacterial culture used is known to degrade both nitrate and perchlorate. Only perchlorate degradation was monitored in this study. Fig. 2b and d also shows a slight increase in conductivity in the microbial culture. This increase is due to the accumulation of chloride, the product of perchlorate biodegradation. The oxidation reduction potential (ORP) and the temperature of the bioreactor also were monitored. ORP varied from (
)390 mV to (
)470 mV, and the temperature of the system varied between 22  2  C. Dissolved oxygen (DO) concentrations were always below 0.2 mg/L.

27

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 1 e3 2

b

pH

2000

7000

1800

6000

1600

5000

1400

4000

1200 1000 2

4

6

8

10

12

conductivity

8.1

26

8

24

7.9

22

7.8

20

7.7 7.6

18

3000

7.5

16

2000

7.4

14 0

14

2

4

6

d

COD and SS in the fermentor during cycle 4

12

14

SS

pH and conductivity in the fermentor during cycle 4 pH

COD

2000 6000

1400

4000

1200

3000

pH

5000

1600

COD, mg/L

1800 SS, mg/L

10

Day

Day

c

8

conductivity

8.1

26

8

24

7.9

22

7.8

20

7.7

18

7.6

16

7.5

1000

2000 0

1

2

3

4

5

6

7

8

7.4

9

Conductivity, mS/cm

0

pH and conductivity in the fermentor during cycle 2

COD

COD, mg/L

SS, mg/L

SS

Conductivity, mS/cm

COD and SS in the fermentor during cycle 2

pH

a

14 0

1

2

3

4

Day

5

6

7

8

9

Day

Fig. 2 e Typical fermenter operating conditions: (a) and (c) decrease of COD and SS during Cycles 2 and 4, and (b) and (d) increase of pH and conductivity in the fermenter during Cycles 2 and 4.

Resin perchlorate biodegradation for all five cycles of bioregeneration is shown in Fig. 3a and b. Cycles 1, 2, and 3 were run for 14 days, while Cycles 4 and 5 were run for 9 and 8 days, respectively. The stabilization of the residual perchlorate concentration in the resin bed was the criterion to stop the bioregeneration cycle (Fig. 3a). The data show that perchlorate biodegradation is rapid during the first days, and then it slows down and stabilizes. In Fig. 3a, all the data except Day 0 are Parr-bomb measurements of resin-attached perchlorate that remains in the resin as bioregeneration progresses. For Day 0 of each cycle, the initial perchlorate concentration is the amount of perchlorate loaded to the resin at the beginning of that cycle (Table 3) plus the remaining perchlorate in the resin from the previous cycle. For Cycle 1, the initial perchlorate concentration is the amount of perchlorate loaded to the resin at the beginning of the cycle. Notice in Fig. 3a that the initial perchlorate concentration in the resin differs for each cycle. The reason for that is due to perchlorate buildup in the resin after each cycle. For example, in Cycle 2, the initial perchlorate at Day 0 equals the perchlorate remaining at the end of Cycle 1 (3.1 gperchlorate/Lresin) plus the perchlorate loaded to the resin for Cycle 2 (10.15 gperchlorate/Lresin); that equals to 13.25 gperchlorate/Lresin. At the end of bioregeneration, the remaining perchlorate values in Cycles 1e5 are 3.1 gperchlorate/Lresin, 5.4 gperchlorate/Lresin, 7.4 gperchlorate/Lresin, 8.5 gperchlorate/Lresin, and 8.8 gperchlorate/Lresin, respectively. Fig. 3b depicts the residual resin-attached-perchlorate mass as g of ClO
4 per liter of microbial culture. In all five cycles depicted in Fig. 3b, the highest perchlorate biodegradation rate was obtained in the first days of bioregeneration

throughout the rest of the experiment, it was reduced. The initial perchlorate load at the start of Cycles 1e5 were 481 mg/ Lculture, 433 mg/Lculture, 385 mg/Lculture, 314 mg/Lculture, and 231 mg/Lculture, respectively (Fig. 3b). The reason for the

a Residual perchlorate in resin (g/L of resin)

Resin bioregeneration

20

Cycle 1

18

Cycle 2

16

Cycle 3

14

Cycle 4

12

Cycle 5

10 8 6 4 2 0

2

4

6

8

10

12

14

16

Day

b Residual perchlorate per L of culture (mg/L of culture)

3.2.

500

Cycle 1

450

Cycle 2

400

Cycle 3

350 300

Cycle 4

250

Cycle 5

200 150 100 50 0

2

4

6

8

10

12

14

16

Day

Fig. 3 e Residual perchlorate concentration during the five cycles of bioregeneration: (a) residual perchlorate concentration attached to the resin, and (b) residual attached-perchlorate per liter of microbial culture.

28

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 1 e3 2

reduction of the initial perchlorate load was the decrease of resin volume from 1305 mL in Cycle 1 to 375 mL in Cycle 5 due to resin sampling in each cycle. Samples were taken during bioregeneration for Parr bomb and resin capacity measurements while the volume of the microbial culture in the fermenter was kept constant at 30 L for all five cycles. The observed degradation rate for resin-attached perchlorate, expressed as mg of perchlorate biodegraded per mg of SS per day (mgp/mgss/d), for Day 1 of Cycles 1e5 is shown in Table 5. For comparison, the degradation rate for free perchlorate ion in water from other studies is also shown in Table 5. Theoretical degradation rates were calculated using Monod’s kinetics (Table 5) (Rittmann and McCarty, 2001), which is an accepted model for perchlorate biodegradation in water (Waller et al., 2004): dS qmax  S X ¼
dt S þ Ks

(1)

where, S is concentration of perchlorate (mg/Lculture); t is the time (days); qmax is the maximum perchlorate utilization rate (d
1); X is the SS in the system (mg/L); and Ks is the halfsaturation constant for perchlorate (mg/L). Nine pairs of Ks and qmax values published by other researchers (Table 1) were used to calculate the degradation rates, and the lowest and highest theoretical rates were reported in Table 5. A shrinking core model has been successfully used to describe perchlorate desorption from resin beads during bioregeneration (Venkatesan et al., 2010). The model asserts that perchlorate ions in the outer region of the resin are available for bacterial degradation prior to the ions that are located deep in the resin bead. In this research, the results are evaluated assuming that bioregeneration follows a shrinking core model. In all five cycles of bioregeneration, the observed perchlorate biodegradation rate in the first day of the bioregeneration process was one order of magnitude slower than the smallest calculated theoretical biodegradation rate of free perchlorate (Table 5). This observation could be attributed to two potential explanations. The first explanation involves kinetics control, where the perchlorate biodegradation rate is concentration dependent and it slows down for perchlorate concentrations smaller that the half-saturation constant for perchlorate (Cox et al., 2000; Logan et al., 2001; Tan et al., 2004;

Hiremath et al., 2006). The second explanation involves mass transfer control, where the perchlorate biodegradation rate depends on the diffusion of the desorbed perchlorate ions from the original functional groups located in the resin bead to the bacteria located outside the resin bead, as envisioned in the shrinking core model. The observed higher perchlorate biodegradation rate for free perchlorate compared to resinattached perchlorate (Table 5) supports mass transfer as a major factors involved in resin bioregeneration. Venkatesan et al. (2010) shows that the initial phase of resin bioregeneration might be controlled by kinetics while the later phase is mass transfer controlled.

3.3.

Capacity loss evaluation

Fig. 4 depicts for each cycle the amount of loaded perchlorate to the resin; initial perchlorate, which is loaded perchlorate plus perchlorate remaining from the previous cycle; remaining perchlorate in the resin; and biodegraded perchlorate in the resin. In Cycles 1e5, 72%, 59%, 53%, 51%, and 53% of the initial perchlorate was degraded during bioregeneration, respectively (Fig. 4). It is likely that every time the resin was loaded, there was a core in the resin in the center of the bead that was not bioregenerated. It seems that the size of this core from Cycle 1 to Cycle 5 grew; every time the resin was loaded, perchlorate was replaced in the outer portion of the resin bead, and permanent perchlorate in the center of bead was carried over to the next cycle. Notwithstanding the remaining perchlorate in the resin carried over from the previous cycles, in Cycles 1e5, 72%, 77%, 80%, 89%, and 97% of the perchlorate, which was loaded to the resin in each cycle was utilized, respectively. Fig. 5 depicts that the load buildup of undegraded perchlorate was rapid through Cycles 1 and 2, and began to stabilize in Cycles 4 and 5. The reason for this observation could be that perchlorate ions located in the center of resin bead could not be regenerated, and were permanently occupied by perchlorate throughout the bioregeneration cycles. This assumption is supported by the fact that the Parr-bomb measurement of resin samples at the end of bioregeneration process show the presence of perchlorate ions. Bio-fouling as a result of clogging of the resin pores can be a potential reason for a permanent load buildup of perchlorate

Table 5 e Resin-attached perchlorate and free perchlorate biodegradation rates. Cycle #

1 (1st 2 (1st 3 (1st 4 (1st 5 (1st e e e e

day) day) day) day) day)

ClO
4 ion situation

Initial perchlorate concentration (mg/Lculture)

Observed perchlorate biodegradation rate (mgp/mgss/d)

Theoretical (calculated) biodegradation rate (mgp/mgss/d)

Reference

Attached to resin Attached to resin Attached to resin Attached to resin Attached to resin Free in water Free in water Free in water Free in water

481 333 385 314 231 298 915 100 250

0.096 0.066 0.065 0.039 0.066 1.68 2.57 0.36 0.64

0.362ae5.854b 0.357ae5.838b 0.352ae5.819b 0.341ae5.780b 0.321ae5.704b 0.338ae5.768b 0.383ae5.922b 0.353ae5.357b 0.385ae5.725b

This study This study This study This study This study Korenkov et al., 1976 Attaway and Smith, 1993 Logan et al., 2001 Shrout and Parkin, 2006

a Calculated using Ks and qmax from Logan et al. (2001). b Calculated using Ks and qmax from Waller et al. (2004).

Perchlorate load (g/Lresin)

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 1 e3 2

29

20 18 16 14 12 10 8 6 4 2 0 1

2

3 Cycle No.

4

5

ClO4- at the beginning of cycle Loaded ClO4- into the resin Biodegraded ClO4Remaining ClO4- after bioregeneration

Fig. 4 e The legends indicate the initial perchlorate (i.e., loaded perchlorate plus remaining perchlorate from the previous cycle); loaded perchlorate to the resin; biodegraded perchlorate; and remaining perchlorate for each cycle of bioregeneration. Although the loaded perchlorate stayed approximately constant through different cycles, the total initial perchlorate load at the beginning of each cycle increased from Cycle 1 through Cycle 5 due to residual perchlorate leftover from the previous cycle. The amount of biodegraded perchlorate also increased from Cycle 1 to Cycle 5.

Residual ClO4- after bioregeneration (g/Lresin)

in the resin. A SEM analysis of the rinsed and air-dried fresh resin and bioregenerated resin that was sampled after Cycle 5 revealed that some of the pores of the bioregenerated resin were clogged. Although the SEM pictures were not taken in the same location of the bioregenerated and fresh resin beads, microscopic observations of several resin beads revealed that pores in the fresh resin beads were clearly accessible (Fig. 6a). To evaluate the removal of the materials that were clogging the resin pores, the bioregenerated resin sample was soaked in 100 mg/L sodium hypochlorite solution for 20 min, rinsed with DI water, and air-dried. A SEM analysis of the bioregenerated resin sample treated with sodium hypochlorite solution showed that most of the pores were unclogged after treatment (Fig. 6c), suggesting that the clogging materials are likely organics that can be removed using the sodium

10 8

Fig. 6 e SEM images of the (a) fresh resin with normal pores, (b) bioregenerated resin sampled after Cycle 5 (arrows show the clogged pores), and (c) bioregenerated resin treated with 100 mg/L sodium hypochlorite (arrows show the unclogged pores).

6 4 2 0 Cycle 1

Cycle 2

Cycle 3

Cycle 4

Cycle 5

Bioregeneration Cycle

Fig. 5 e Permanent perchlorate load buildup through the five cycles of bioregeneration. The load buildup is stabilized in the last cycles, Cycles 4 and 5.

hypochlorite solution. Specific characterization of the substances that were clogging the pores has not been performed because it was difficult to remove from the pores in significant amounts for an analysis; however, it is likely that the pores were clogged by bacterial debris or biodegradation byproducts. The perchlorate load buildup may be due to slowing down of mass transfer in the resin pores because of clogging. Because resin fouling increases with time of contact between the resin and the microbial culture, mass transfer

30

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 1 e3 2

flux is likely to be greatest at the beginning of the bioregeneration cycle. Therefore, a slower perchlorate biodegradation rate at the end of bioregeneration process can be expected, as shown in Fig. 3a. The reason for this is first, because of pore diffusion; it takes longer for perchlorate located deep into the bead to arrive to the surface (Venkatesan et al., 2010), independent on pore clogging; and second, if clogged pores are present, then the mass transfer slows down even more. Resin was treated for biological fouling after each bioregeneration cycle, as expressed in Table 4, and the COD of the rinsate was measured (Fig. 7). Prior to the fouling removal procedure, the resin was rinsed with 1 BV of DI water five times to remove any microbial cells and organics that might remain in the resin. Transmittance measurements of the rinsate solutions indicated that the employed rinsing procedure effectively removed the remaining microbial cells from the resin. Even though prior to bio-fouling treatment the resin was rinsed with 1 BV of DI water five times, significant amounts of organics were detected in the fouling removal rinsate solutions (Fig. 7). In Cycles 1e5, 3840 mg COD/Lresin, 4778 mg COD/Lresin, 5745 mg COD/Lresin, 4650 mg COD/Lresin, and 3465 mg COD/Lresin of organic substances, respectively, were removed from the resin. The organic substances removed were not investigated, but they are thought to be byproducts of biodegradation. The resin treated by the fouling removal and disinfection procedure had a yellowish brown color, which made the resin look different from fresh resin. Though applying higher concentrations of the disinfectant agent may be able to retrieve the original light color of the resin, treatment with a concentrated disinfectant reagent potentially increases the risk of oxidation of the active functional groups of the resin beads, possibly resulting in a loss of resin capacity. Resin capacity was measured before bioregeneration both before disinfection and after disinfection in each cycle. Table 6 shows the resin capacity measurements from Cycle 1 through Cycle 5 of bioregeneration. The method of resin capacity measurement that was used cannot account for perchlorate accumulation in the resin. This method used Cl
ion to convert the resin functional groups to the chloride form. Because the resin used is highly selective for perchlorate,

COD of the rinsate (mg/L)

2500

Table 6 e Resin capacity measurement and capacity loss for five cycles bioregeneration. Cycle #

Sampling point

Capacity (eq/L)

Capacity loss (%)a

Cycle 1

Pre-bioregeneration Pre-disinfection Post-disinfection

0.64 0.61 NA

0.0 4.7% NA

Cycle 2

Pre-bioregeneration Pre-disinfection Post-disinfection

0.52 NA 0.54

18.8% NA 15.6%

Cycle 3

Pre-bioregeneration Pre-disinfection Post-disinfection

0.54 0.54 0.54

15.6% 15.6% 15.6%

Cycle 4

Pre-bioregeneration Pre-disinfection Post-disinfection

NA NA 0.56

NA NA 12.5%

Cycle 5

Pre-bioregeneration Pre-disinfection Post-disinfection

NA 0.54 0.55

NA 15.6% 14.1%

a Virgin resin capacity was assumed 0.64 eq/L.

chloride cannot replace perchlorate that is attached to the resin. Hence, the observed capacity loss reported in Table 6 is due, in part, to the perchlorate load accumulated in the resin. Capacity measurements post-disinfection of four out of five cycles, show that the capacity loss is not cumulative throughout consecutive cycles of loading and bioregeneration. Capacity loss did not increase with each cycle of bioregeneration. By comparing pre-disinfection and postdisinfection values from Table 6, it appears that the applied disinfection step did not negatively affected resin capacity. Table 6 indicates that after five cycles of loading, bioregeneration, fouling removal, and disinfection of the resin, the total capacity loss was about 15%, which corresponds to 9.93 g ClO
4 /Lresin, for a total resin capacity of 0.64 eq/Lresin.
There is 1.13 g ClO
4 /Lresin difference between 9.93 g ClO4 /Lresin capacity loss (Table 6) and the permanent perchlorate load of 8.80 g of ClO
4 /Lresin (Fig. 4, Cycle 5). A significant finding of this study is that the bioregeneration process can be performed as a multi-cycle process. The capacity loss mostly was due to the remaining perchlorate load. It seems there is a tendency for the capacity loss to stabilize with time, but more run cycles are needed to confirm this claim.

2000 1500

4.

Discussion

1000 500 0 12% NaCl + 2% NaOH 12% NaCl + 2% NaOH (4 Hr) (12 Hr) Cycle 1

Cycle 2

Cycle 3

12% NaCl (2 Hr) Cycle 4

Cycle 5

Fig. 7 e Chemical oxygen demand (COD) of eluate obtained during fouling removal process applied in the resin. 12, 4, and 2 h are the contact periods of the fouling removal treatments.

A main concern in resin bioregeneration is the capacity loss of the resin. There is a perchlorate load buildup, thought to be in the center of resin, through Cycle 1 to Cycle 5. The perchlorate load buildup is a major reason for resin capacity loss through the bioregeneration cycles. The capacity loss resulting from perchlorate load buildup is more significant through the first cycles, and seems to stabilize after a few cycles. Wang et al. (2008) has reported on the bioregenerated macroporous resins using batch tests and sodium chloride as an agent for perchlorate detachment. They found that resin capacity did

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 1 e3 2

not change after bioregeneration and that fouling was minimal. Their results are different from those reported for this research. However, the results cannot be directly compared because different experimental procedures were used in both investigations. In this research, bioregeneration impacted resin capacity; It is thought that there is a region in the center of resin bead where some the functional groups of the resin are unavailable and permanently occupied by perchlorate throughout the bioregeneration cycles. The possible reason for unavailability of the functional groups in that region of resin bead is likely to be clogging of the resin pores, which has been observed by SEM imaging. There exist several challenges in implementing resin bioregeneration including long bioregeneration times due to slower kinetics, perchlorate buildup in the resin, and biofouling. Perchlorate buildup was found to be the major drawback of the bioregeneration process. Evaluation is needed to determine whether bioregenerated resins, containing some residual perchlorate, can be used to produce water that meets the regulatory levels for perchlorate. Clearly, since capacity loss is involved in the bioregeneration process, lower bed volumes of water can be processed to reach the breakthrough point. If perchlorate leakage occurs when using bioregenerated resin, another IX column containing fresh resin can be used to polish the effluent. This polishing IX column can last for a very long time since the leakage of the IX column containing bioregenerated resin is expected to be much less than the influent perchlorate concentration to the water treatment plant. In this research, the perchlorate load in resin was typical of that obtained when treating industrial sites contaminated with perchlorate. The bioregeneration of resins used in drinking water applications, where perchlorate loading is smaller, is expected to face the same challenges. Although the kinetics of perchlorate degradation in water is slower for lower perchlorate concentrations, biodegradation or resin-attached perchlorate is thought to be controlled by mass transfer. Therefore, the bioregeneration of resins used for drinking or industrial applications is expected to have similar limitations. One has to evaluate the economical feasibility of the bioregeneration process. Two factors play main roles in the economical evaluation of the IX resin bioregeneration process. On the one hand, there is the cost of purchasing fresh resin in cases where spent resin is incinerated. On the other hand, there are costs associated with the resin bioregeneration process. Capital costs include building the bioreactor, the FBR, pumps, and pipelines. Operating costs include personnel and energy as well as purchasing electron donors, nutrients and minerals, and buffer chemicals. In addition, there is a reduction in the volume of processed water because of capacity loss of the resin due to bioregeneration. Considering the number of loading-bioregeneration cycles and the 15% capacity loss due to bioregeneration process, bioregeneration is feasible only if its costs are less than the costs of replacing the spent resin with fresh resin. There are also environmental costs to be considered when spent resins are incinerated. Depending on regulations concerning minimizing greenhouse gases emissions, carbon credits could be awarded to water utilities that stop resin incineration. In the long-term, resin bioregeneration can be the technology of choice for water utilities that deal with perchlorate contamination.

5.

31

Conclusions

The objective of this research was to determine whether macroporous perchlorate-selective resins can be bioregenerated for many cycles. The results obtained from this study revealed that macroporous perchlorate-selective IX resins can be bioregenerated and reloaded for five cycles. However, the capacity loss is significant, about 15%, but not cumulative in the multicycle loading-bioregeneration process. Capacity loss mostly is due to permanent load buildup of perchlorate in the resin. This perchlorate load buildup is fast through the first two cycles, and almost stabilizes through Cycles 4 and 5. The following can be concluded from the results of this research: 1) The reason for perchlorate load buildup could be due to the unavailability perchlorate attached to functional groups located in the center of resin as a result of clogging of the pores of the resin. The applied fouling treatment and disinfection methods were effective during the regeneration cycles, and could remove a significant amount of organics from the resin. 2) As bioregeneration progresses, the establishment of fouling causes a lower mass transfer flux in the resin, compared to the mass transfer flux in unfouled resin. 3) Slower perchlorate biodegradation at the end of the bioregeneration process can be attributed to: a) pore diffusion independent of pore clogging, as it takes longer for perchlorate located deep in the bead to arrive at the surface; and b) a combination of slower diffusion due to the location within the bead and to clogging of the pores by bacteria, causing the mass transfer to slow down even more. 4) The rate of perchlorate biodegradation of the resinattached perchlorate is significantly slower than the biodegradation rate of soluble perchlorate.

Acknowledgments This research was partially funded Basin Water (Rancho Cucamonga, CA) and a UNLV College Engineering scholarship to Mohamadali Sharbatmaleki. We thank Dr. Fred Reinhardt from HDR Engineering for his valuable assistance with data evaluation and quality control/quality assurance methods. The support from Dr. William Carlin (Rohm and Haas, Philadelphia, PA) with resin capacity measurements is much appreciated. We thank Resintech (West Berlin, PA) for specially manufacturing the resin for this research.

references

Attaway, H., Smith, M., 1993. Reduction of perchlorate by an anaerobic enrichment culture. Journal of Industrial Microbiology 12, 408e412. Batista, J.R., 2006. U.S. Patent No. 20090047732. U.S. Patent and Trademark Office, Washington, DC.

32

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 1 e3 2

Batista, J.R., Howerton, A., Jensen, P., 2007a. Bioregeneration of Spent Ion Exchange Resin Loaded with Nitrate/Perchlorate and Biofouling Removal. Water Environmental Federation (WEFTEC 07), San Diego, CA. Batista, J.R., Howerton, A., Jensen, P., 2007b. Bioregeneration of perchlorate/nitrate contaminated ion-exchange resins. In: Proceedings of the Ninth International In Situ and On-Site Bioremediation Symposium. Baltimore, MA. Brandhuber, P., Clark, S., 2005. Perchlorate Occurrence Mapping Report submitted to American Water Works Association Research Foundation. Clifford, D.A., 1999. In: Letterman, R.D. (Ed.), Water Quality and Treatment: a Handbook of Community Water-Supplies, fifth ed. American Water Works Association, McGraw Hill, New York. Coates, J.D., Achenbach, L.A., 2004. Microbial perchlorate reduction: rocket-fuelled metabolism. Nature Reviews Microbiology 2, 569e580. Cox, E.E., Edwards, E., Neville, S., 2000. In: Urbansky, E.T. (Ed.), Perchlorate in the Environment. Kluwer Academic/Plenum, New York. Dale, J.A., Tavani, L.M., Golden, L.S., 2001. U.S. Patent No. 632324. U.S. Patent and Trademark Office, Washington, DC. Dudley, M., Salamone, A., Nerenberg, R., 2008. Kinetics of a chlorate-accumulating, perchlorate-reducing bacterium. Water Research 42, 2403e2410. Eaton, A.D., Clesceri, L.S., Rice, E.W., Greenberg, A.E., 2005. Standard Methods for the Examination of Water and Wastewater, 21st ed. American Public Health Association, Washington, DC. Gingras, T.M., Batista, J.R., 2002. Biological reduction of perchlorate in ion-exchange regenerant solutions containing high salinity and high ammonium levels. Environmental Monitoring 4, 96e101. Gu, B.H., Brown, G.M., Maya, L., Lance, M.J., Moyer, B.A.A., 2001. Regeneration of perchlorate (ClO
4 )-loaded anion exchange resins by a novel tetrachloroferrate (FeClm4) displacement technique. Environmental Science and Technology 35 (16), 3363e3368. Hiremath, T., Roberts, D.J., Lin, X., Clifford, D.A., Gillogly, T.E.T., Lehman, S.G., 2006. Biological treatment of perchlorate in spent ISEP ion-exchange brine. Environmental Engineering Science 23 (6), 1009e1016. Kesterson, K.E., Amy, P.S., Batista, J.R., 2005. Limitations to natural bioremediation of perchlorate in a contaminated site. Bioremediation Journal 9, 129e138. Kirk, A.B., 2006. Environmental perchlorate: why it matters. Analytica Chimica Acta 567, 4e12. Korenkov, V.N., Romanenko, V.I., Kuznetsov, S.I., Voronov, J.V., 1976. U.S. Patent No. 3943055. U.S. Patent and Trademark Office, Washington, DC. Kun, K.A., Kunin, R., 1968. Macroreticular resins III: formation of macroreticular styreneedivinylbenzene copolymers. Journal of Polymer Science 6, 2689e2701. Lehman, S.G., Badruzzaman, M., Adham, S., Roberts, D.J., Clifford, D.A., 2008. Perchlorate and nitrate treatment by ion exchange integrated with biological brine treatment. Water Research 42 (4e5), 969e976. Li, P., SenGupta, A.K., 2000. Intraparticle diffusion during selective sorption of trace contaminants: the effect of gel versus macroporous morphology. Environmental Science and Technology 34, 5193e5200. Logan, B.E., 1998. A review of chlorate and perchlorate respiring microorganisms. Bioremediation Journal 2 (2), 69e79.

Logan, B.E., Zhang, H.S., Mulvaney, P., Milner, M.G., Head, I.M., Unz, R.F., 2001. Kinetics of perchlorate- and chlorate-respiring bacteria. Applied Environmental Microbiology 67, 2499e2506. MWH (Montgomery Watson Harza), 2005. Water Treatment: Principles and Design, second ed. John Wiley and Sons, New York. NASA, 2006. Perchlorate (ClO
4 ) treatment technologies literature review operable unit 1 expanded treatability study. At. http:// jplwater.nasa.gov/nmoweb/Docs/ROD/Perchlorate-LitReview.pdf (accessed 12.07.08.). Nerenberg, R., Kawagoshi, Y., Rittmann, B.E., 2006. Kinetics of a hydrogen-oxidizing, perchlorate-reducing bacterium. Water Research 40 (17), 3290e3296. NRC, 2005. Health Implications of Perchlorate Ingestion. National Research Council of the National Academies. National Academies Press, Washington, DC. ResinTech, 2008. Perchlorate removal application. At. http:// www.resintech.com/applications/perchlorate_removal.aspx (accessed 01.17.10.). Rittmann, B.E., McCarty, P.L., 2001. Environmental Biotechnology: Principles and Applications. McGraw-Hill, New York. Seidel, C.N., Blute, M., Mcguire, M., 2006. Mini column and pilot test results for perchlorate selective resins. In: AWWA Inorganic Contaminants Workshop. Austin, TX. Shrout, J.D., Parkin, G.F., 2006. Influence of electron donor, oxygen, and redox potential on bacterial perchlorate degradation. Water Research 40, 1191e1199. Tan, K., Anderson, T.A., Jackson, W.A., 2004. Degradation kinetics of perchlorate in sediments and soils. Water, Air, and Soil Pollution 151 (1e4), 245e259. Urbansky, E.T., Magnuson, M.L., Kelty, C.A., Gu, B., Brown, G.M., 2000. Comment on perchlorate identification in fertilizers and the subsequent addition/correction. Environmental Science and Technology 34, 4452. USEPA, 2009. Interim drinking water health advisory for perchlorate. At. http://www.epa.gov/safewater/contaminants/ unregulated/pdfs/healthadvisory_perchlorate_interim.pdf (accessed 12.19.09.). Van Ginkel, S.W., Ahn, C.H., Badruzzaman, M., Robers, D.J., Lehman, S.G., Adham, S.S., Rittmann, B.E., 2008. Kinetics of nitrate and perchlorate reduction in ion-exchange brine using the membrane biofilm bioreactor (MBfR). Water Research 42, 4197e4205. Venkatesan, A.K., Batista, J.R., 2011. Investigation of factors affecting the bioregeneration process for perchlorate-laden gel-type anion-exchange resin. Bioremediation Journal 15 (1), 1e11. Venkatesan, A.K., Sharbatmaleki, M., Batista, J.R., 2010. Bioregeneration of perchlorate-laden gel-type anion-exchange resin in a fluidized bed reactor. Journal of Hazardous Materials 177, 730e737. Waller, A.S., Cox, E.E., Edwards, E.A., 2004. Perchlorate-reducing microorganisms isolated from contaminated sites. Environmental Microbiology 6 (5), 517e527. Wang, C., Lippincott, L., Meng, X.G., 2008. Feasibility and kinetics study on the direct bio-regeneration of perchlorate laden anion-exchange resin. Water Research 42 (18), 4619e4628. Wang, C., Lippincott, L., Yoon, I.H., Meng, X., 2009. Modeling, ratelimiting step investigation, and enhancement of the direct bio-regeneration of perchlorate laden anion-exchange resin. Water Research 43 (1), 127e136. Wolff, J., 1998. Perchlorate and the thyroid gland. Pharmacological Reviews 50 (1), 89e105.

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

Available online at www.sciencedirect.com

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

Characterisation and application of a novel positively charged nanofiltration membrane for the treatment of textile industry wastewaters Shuying Cheng, Darren L. Oatley*, Paul M. Williams, Chris J. Wright Centre for Water Advanced Technologies and Environmental Research (CWATER), College of Engineering, Swansea University, Singleton Park, Swansea SA2 8PP, UK

article info

abstract

Article history:

The present study demonstrates the high potential for the application of a novel self

Received 2 August 2011

assembled positively charged nanofiltration membrane, PA6DT-C, in processes such as the

Received in revised form

recovery of valuable cationic macromolecules in the bioprocess and pharmaceutical

7 October 2011

industries or removal of multi-valent cations such as dyes and heavy metals in the paper

Accepted 11 October 2011

and pulp, textiles, nuclear, and automotive industries. The nanofiltration membrane,

Available online 25 October 2011

prepared in this laboratory, is further characterised and then tested for the removal and recovery of Methylene Blue from a synthetic dye house wastewater. The characterisation

Keywords:

process involved the construction of a rejection profile for NaCl over a wide range of pH and

Nanofiltration

concentration, which illustrates that the optimal process conditions for the removal of

Positive charge

small cations using this membrane is in the region pH <8.0 and concentration less than

Self assembly

15 mol m
3. The salt rejection data was used to calculate the magnitude of the effective

Characterisation

membrane charge density and this was found to be significantly higher for the PA6DT-C

Dye

membrane than two commercially available membranes (Desal-DK and Nanomax-50).

Wastewater

The membrane flux for this new membrane is also superior to the commercial membranes with an approximate increase of 3e4 fold. The PA6DT-C membrane was successful in removal of Methylene Blue dye from synthetic dye house wastewaters achieving 98% rejection and a membrane flux of w17 LMH bar
1. Thus, this new membrane both adds to and complements the existing short supply of positively charged NF membranes. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Nanofiltration (NF) is a pressure driven membrane separation process with characteristics between reverse osmosis and ultrafiltration. The nominal molecular weight cut off of a NF membrane is in the range 100e1000 Da, indicating that NF membranes have an approximate pore size of 1 nm. Separation of solutes in the NF range is dependent upon the micro-

hydrodynamics and interfacial events occurring at the membrane surface and inside the membrane, for example, rejection may be attributed to a combination of both steric and charge effects. Nanofiltration was first introduced in the early 1980’s and has since gained in popularity due to increased selectivity for mono and multi valent ions, low operating pressures and relatively low capital and operating costs. These advantages have led to nanofiltration being adopted for

* Corresponding author. Tel.: þ44 (0) 1792 606 668. E-mail address: [email protected] (D.L. Oatley). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.011

34

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

liquid phase separations in several major industrial sectors (Schaefer et al. (2005)). The major industrial use of NF membranes remains in aqueous phase desalination processes (Khawajia et al., 2007; Walha et al., 2007), however, there are many other industrial or potentially industrial examples to be found. For example; separation of olefins and paraffins in LPG production (Avgidou et al., 2004), extraction and concentration of biologically active compounds (Tylkowski et al., 2011), harvesting marine flavours from cooking waters (Vandanjon et al., 2002), dairy by-product recovery (Nguyen et al., 2003) and fractionation of whey proteins (Pouliot et al., 1999), recovery of pharmaceuticals and personal care products (Bowen et al., 2004; Oatley et al., 2005; Yoon et al., 2006), and recovery of wastes in the automotive (Holmes, 2002), Beverage (Chmiel et al., 2003), fruit juice (Noronha et al., 2002), tanning (Shaalan et al., 2001), and dairy (Dresch et al., 2001) industries, to name but a few. The majority of commercially available nanofiltration membranes are negatively charged (Bellona and Drewes, 2005; Yaroshchuck, 2001); the membranes have a low value pH isoelectric point and are therefore negatively charged at normal operating conditions. This has limited the number of studies conducted for both the fabrication of positively charged NF membranes and the resulting commercial applications. Positively charged NF membranes can have excellent hydrophilicity and exhibit high rejections for multi valent cations (Yan et al., 2008). These properties are particularly important for recovery of valuable cationic macromolecules in the bioprocess and pharmaceutical industries or removal of multi valent cations such as dyes and heavy metals from effluents in the paper and pulp, nuclear, cosmetic, and automotive industries. Typical examples are the recovery of cobalt and lead (Bouranene et al., 2008) from many industrial wastewaters, the recovery of radioactive heavy metals in the nuclear industry (Zakrzewska-Trznadel, 2003; Hwang et al., 2002), zinc and iron recovery in the steel making industry (Wolters et al., 2008), and copper recovery from brass making and electroplating (Ahmad and Ooi, 2010). One attractive application for positively charged NF membranes is the removal of colour from dilute wastewater effluents in the textile industries. The wastewaters generated from the textile industry are among the most polluting of all the industrial sectors and have been considered as a significant environmental problem for several decades (Avlonitis et al., 2008). Most developed nations recognise the impact and toxicity of dyes that are released into the environment and have legislation in place to prevent such discharges necessitating the clean up of such waste streams. There is great variation in the chemical nature of textile discharge which can vary dramatically depending on the type of product produced and chemicals involved in the particular textile factory (Hessel et al., 2007). Forgacs et al. (2004) have conducted a review of technologies available for the removal of colour from textile wastewaters and concluded that traditional techniques are largely ineffective as a result of the inherent stability of these pollutants. The technologies reviewed were adsorption and physiochemical methods, photo-catalysis and oxidative methods, microbiological and enzymatic decomposition, with membrane technologies not considered. However, membranes have the potential to offer

an improved separation and several cost advantages over these more traditional techniques and various works have reported the use of NF membranes for dye separations either for wastewater treatment (Nataraj et al., 2009; Mo et al., 2008; Yu et al., 2010) or process applications (Levenstein et al., 1996). In most cases, the dye wastewater also contains residual salt from the manufacturing process. The removal of dye should be high, but not necessarily complete and the fate of the salt is of less importance. A review of NF processes in the textile industries has been compiled by Lau and Ismail (2009) and has indicated that most commercial membranes either achieve a maximum separation of the dye or high membrane flux. The desirable ideal membrane should exhibit both characteristics. Erswell et al. (1988) were one of the first to investigate the prospect of a charged membrane for the reuse of reactive dye liquors. Membrane performance was monitored in terms of dye and salt rejection and permeate flux, under varying conditions. They concluded that NF is a technically feasible process for treatment of the dye bath at high water recoveries. However, a maximum flux of only 30 LMH was achieved, even when high pressure was used (up to 40 bar). Van der Bruggen et al. (2001) examined the mechanisms of retention and flux decline of nanofiltration membranes using a synthetic dye bath. The study found that rejection of ions decreased with increasing ion concentration and the rejection of dye remained high irrespective of concentration or ionic strength. However, this study investigated the influence of the divalent electrolyte Na2SO4, not the more commonly used NaCl. The amount of wastewater generated from the textiles industry is significant and sustainability of water resources is also becoming an increasingly recognised issue. Therefore, not only is removal of the dye components necessary, but recovery and recycle of the water is desirable from a modern waste treatment plant. Avlonitis et al. (2008) demonstrated that NF treatment of simulated cotton dye effluents was capable of complete colour removal, 72% salt removal and recycle of 90% of the water from the original waste stream. Koyuncu et al. (2004) reported similar findings for an actual remazol dye bath wastewater and concluded that inclusion of an NF treatment plant would have a capital expenditure pay back period of less than two years. In this work, a novel self-assembled positively charged NF membrane (PA6DT-C), prepared in the laboratory and described previously (Bowen et al., 2005; Cheng et al., 2011), is further characterised and then tested for the removal and recovery of Methylene Blue (MB) from a synthetic dye house wastewater. MB is a positively charged dye and the advantage of employing a positively charged NF membrane to remove or recover this species is that membrane flux should not be sacrificed whilst obtaining the high rejection required to obtain favourable process economics.

2.

Experimental methods

2.1.

Materials and membrane

The structure of methylene blue (BDH Chemicals, Poole, UK, part no. 26132) is shown in Table 1. In order to calculate the magnitude of membrane charge, the diffusion coefficient and

35

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

Table 1 e Physical properties of the materials used in this study. Ion

Crystal radii (nm)a

Stokes radii (nm)a

Hydrated radii (nm)a

Diffusion coefficient (10
9 m2 s
1)b

0.181 0.095 0.290 0.065 0.064 e

0.121 0.184 0.230 0.347 0.412 0.486c

0.332 0.358 0.379 0.428 0.461 e

2.032 1.334 1.065 0.706 0.595 0.505c

Chloride (Cl
) Sodium (Naþ) Sulphate ðSO2
4 Þ Magnesium (Mg2þ) Chromium (Cr3þ) Methylene Blue (MBþ)

Chemical structure of Methylene Blue

a Nightingale, 1959. b CRC Handbook of Chemistry and Physics, 2009. c Determined by experiment.

Stokes radius of each species is required. Accurate literature values were not available for the methylene blue dye and so conductivity measurements were made in order to determine these physical properties. Conductivity measurement can provide an accurate determination of the diffusion coefficient for a given ionic species in aqueous solution. Once the diffusion coefficient is known, the Stokes radius may be calculated from the StokeseEinstein equation. The approach taken in this paper is analogous to that used previously (Oatley et al., 2005). The limiting conductance obtained for methylene blue (as a cation) was 95.2  10
4 m2 S mol
1, which corresponds to a diffusion coefficient at infinite dilution of 0.505  10
9 m2 s
1 and a Stokes radius of 0.486 nm. The value obtained for the diffusion coefficient is comparable to that obtained from a single literature source (Tschirch et al., 2008) and was accepted on this basis. Sodium chloride, sodium hydroxide, hydrochloric acid, magnesium chloride, chromium chloride and sodium sulphate were all supplied in high purity form by FisherScientific UK. All materials were used as obtained and not further purified. Deionised (DI) water was obtained from a RiOs water purification system supplied by Millipore UK Ltd. (Watford, UK). The membrane used in this study was PA6DT-C, a positively charged nanofiltration membrane, fabricated in the Swansea laboratory as described previously (Bowen et al., 2005; Cheng et al., 2011).

2.2.

2011) and found to be at pH 9.3. The rejection of NaCl at the IEP was measured experimentally over a range of pressures in order to determine the dielectric properties within the membrane pore. Salt rejection was measured with mono-, biand tri-valence cations, NaCl, MgCl2, and CrCl3 with a concentration of 1 mol m
3 over a pressure range from 200 to 500 kPa to study both the steric and Donnan effects on rejection. The effects of concentration on rejection of NaCl, MgCl2 and Na2SO4 were measured at the constant pressure of 400 kPa. The rejection of MB at different concentrations from 0.4 to 10.5 mol m
3 was also measured over the pressure range 200e500 kPa. The effects of ionic strength on the rejection of MB was studied by measuring a rejection of 1.1 mol m
3 solution of MB mixed with NaCl with a concentration ranging from 1.3 to 1.8 mol m
3. Rejection was based on measurement of the 15 mL of permeate after the first 4 mL permeate was removed. After each run the membrane was washed using DI water to thoroughly remove the remaining solute. The dialfiltration experiment was carried out using initial concentrations of 10.9 and 1.0 mol m
3 for NaCl and MB respectively. Diafiltration was carried out at a constant pressure of 400 kPa and continued until the NaCl concentration was reduced below 1.0 mol m
3. In the diafiltration experiment, a reservoir of 1.5 L capacity containing DI water was inserted between the gas supply and the stirred cell. Permeate samples were taken periodically in order to conduct a mass balance over the diafiltration tank.

Frontal filtration experiments 2.3.

Frontal filtration was employed to measure rejection at the PA6DT-C membrane using feed solutions under constant pressures with stirring. All experiments were conducted at room temperature (20  2  C) using an Amicon 8050 cell filled with 50 mL of the feed solution. The effective membrane area was 13.4 cm
2 and membrane flux was calculated from the collection of 15 mL permeate. Unless specifically stated otherwise, all experiments were conducted at pH 5.6  0.2, which is the pH of the DI water used. The isoelectric point (IEP) for this membrane was determined previously (Cheng et al.,

Sample analysis

Concentrations of salts (single salt solutions only) were determined by conductivity, measured at 25  0.5  C using a Russell RL 105 conductivity meter and probe (Thermo Russell, Auchtermuchty, Fife, UK). The concentration of MB was analysed using a Spectrophotometer (Philips PU 8625 UV/VIS e Phillips Scientific, Cambridge, UK) at a wavelength of 663 nm. The concentration of NaCl in mixed solutions with MB was analysed by using an ICP-OES spectrometer (Spectro Analytical Ins. GmbH, Germany). Solution pH was determined

36

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

using a Phillips PW9421 pH meter and probe (Phillips Scientific, Cambridge, UK).

pore and bulk dielectric constants respectively. In this study, the pore dielectric constant was calculated using

2.4.

   2 d d þ ðεb
εÞ εp ¼ εb
2ðεb
εÞ rp rp

Theoretical background

The rejection of solute was calculated using the following equation Cp Cb

Robs ¼ 1


(1)

where Cp and Cb are concentration of permeate solution and feed solution respectively. Pore ion transport and modelling has been conducted using the updated Donnan Steric Pore Model. This approach has been discussed extensively elsewhere (Bowen and Welfoot, 2002; Bowen et al., 2004; Oatley et al., 2005; Bowen et al., 2005) and only a brief description will be provided. The extended NernstePlanck equation for pore ion transport is given as ji ¼
Di;p

dci zi ci Di;p dj
F þ Ki;c ci V dx RT dx

(2)

where ji is ion the flux, Di,p is the pore diffusion coefficient, ci is the axial concentration of the ion at location x along the pore length, zi is the ion valence, R is the gas constant, T is temperature, F is the Faraday constant, j is the axial charge potential along the pore, Ki,c is a hinderance factor, and V is the solvent velocity. Equation (2) is normally rearranged to provide the concentration gradient of the ion or solute along the pore  dci V
F dj ¼ Ki;c ci
Ci;p
zi ci dx Di;p RT dx

(3)

Taking into account the condition of electroneutrality and considering all ions in solution, the potential gradient is determined as Pn dj ¼ dx

zi V


Ki;c ci
Ci;p Di;p F Pn 2 z ci RT i¼1 i



i¼1

(4)

and substituted into equation (3) to form the pore ion transport equation. In order to solve the ion transport equations, the concentration at the entrance to the pore must be known and this is obtained from equilibrium partitioning     gi ci zi F
DWi exp ¼ F exp
Dj i D goi Ci kT RT

(5)

where gi and goi are activity coefficients, Fi is the steric partitioning factor, DjD is the Donnan potential, k is the Boltzmann constant, and DWi is the dielectric partitioning energy described here using the Born model. The Born equation for calculating the partitioning energy of an ion moving from a medium of high dielectric to low dielectric is DWi ¼

  z2i e2 1 1
8pεo ai εp εb

(6)

where e is elemental electron charge, εo is the permittivity of free space, ai is the ion hydrodynamic radius, εp and εb are the

(7)

where ε* is the reduced dielectric constant of the oriented solvent layer at the pore wall, d is the diameter of a water molecule and rp is the pore radius. A mass balance over a diafiltration vessel yields  d
VCi;f ¼
Jv ACi;P dt

(8)

where Ci,f and Ci,P are the concentrations of component i in the feed and permeate respectively. V is the vessel volume, A is the membrane area and Jv is the volume flux per unit area of membrane. Expanding the derivative for constant volume gives V

dCi;f ¼
Jv ACi;P dt

(9)

and from the definition of rejection CiP ¼ ð1
Robs ÞCif

(10)

Substituting eq. (10) into eq. (9) yields dCi;f Jv A ¼
ð1
Robs ÞCi;f dt V

(11)

Thus, integrating eq. (5) with respect to time will provide the concentration profile in the diafiltration tank over the time course of the process.

3.

Results and discussion

3.1.

Further membrane characterisation

3.1.1.

Reassessment of pore dielectric constant

Bowen and Welfoot (2002) proposed that the pore dielectric constant is a function of the bulk solvent dielectric properties and the value of the dielectric constant of an orientated solvent layer inside the confines of the membrane pore, ε*. Pore dielectric effects and effective membrane charge density normally exhibit coupled behaviour. Thus, in order to evaluate a single effect their relationship must be decoupled. The membrane IEP provides an opportunity to study only dielectric effects due to the membrane surface charge density being effectively neutralised. Fig. 1 provides the rejection of NaCl over a range of pressure at pH 9.3 (the membrane IEP). At this pH, Donnan effects are negligible allowing investigation of steric and dielectric effects only. The dielectric constant of a single water layer was varied to obtain the best agreement with the experimental rejection by least squares fitting. The dielectric constant of the ordered water layer was calculated by this method to be ε* w 41. Currently, there is a lack of experimental data on the dielectric properties of oriented solvent layers in NF pores and the impact of different membrane materials is also unknown. However, this value is consistent with values obtained for other polyamide NF membranes (Desal-DK ¼ 31 (Bowen and Welfoot, 2002)) and Nanomax-50 ¼ 43 (Oatley et al., 2005) and was therefore used

37

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

a

1.0 Theoretical Fitting 1.0 mol m-3 2.8 mol m-3 -3 9.7 mol m -3 27.7 mol m

Rejection

0.8

0.6

0.4

0.2

0.0 0.0

0.1

0.2

0.3

0.4

0.5

0.4

0.5

0.6

ΔPe / MPa

b

3.0 -3

1.0 mol m 2.8 mol m-3 9.7 mol m-3 27.7 mol m-3 best fit

2.5

as the value for the modified PA6DT-C membrane in subsequent calculations.

2.0

Jv / 10-5 m s-1

Fig. 1 e Reassessment of dielectric constant for the PA6DTC membrane from rejection of 2.8 mol mL3 NaCl at pH 9.3.

1.5

1.0

0.5

3.1.2. Effects of pH and concentration on the membrane rejection for a single mono-valent salt The experimental rejection for different concentrations of NaCl at pH 5.6 for membrane PA6DT-C was measured at varying effective pressure driving force and is shown in Fig. 2. The concentration dependent decrease of the rejection is pronounced, which indicates that NaCl rejection is more dependent on Donnan partitioning (electrostatic interactions) than steric exclusion, a result of the relatively large pore size of the membrane (rp ¼ 1.47 nm (Cheng et al., 2011)). The linearity of the membrane flux data illustrates that membrane compressibility and fouling were not significant and that the pressure dependence is typical for that of porous membranes. At low concentration (1.0 and 2.8 mol m
3), the membrane flux is high (3.75  10
11 ms
1 Pa
1, 13.5 LMH bar
1) and similar in magnitude. However, when the concentration was increased to a maximum of 27.7 mol m
3 the flux declined to 2.3  10
11 ms
1 Pa
1 (8.3 LMH bar
1), indicating that the concentration polarisation effect has increased. The NaCl solution rejection data was then used to calculate the effective membrane charge density, Xd, by the least squares method. The calculated theoretical rejection showed a very good agreement with the experimental rejection, see Fig. 2(a). Having established the values of Xd representative of the rejection data, a charge profile was developed for use in simulations and predictions. Values for the calculated Xd for the different concentrations of NaCl are shown in Fig. 3 and have been represented using a standard power law expression. For comparative purposes, the same information from

0.0 0.0

0.1

0.2

0.3

0.6

Δ Pe / MPa

Fig. 2 e Experimental rejection (a) and membrane flux (b) For NaCl at pH 5.6 vs effective pressure for the PA6DT-C membrane.

two commercially available NF membranes have also been included (Oatley, 2004). Both the Nanomax-50 (Millipore UK Ltd., Watford, UK) and the Desal-DK (GE-Osmonics, France) membranes exhibit a significantly lower magnitude value for Xd indicating that the PA6DT-C membrane is more highly charged in a similar ionic environment. Fig. 4 shows the rejection of NaCl of different concentrations in the range of pH 5.6 to 10.3 at a pressure of 400 kPa. As a general trend, at all pH values studied, the rejection decreases as concentration increases as a result of Donnan interactions; the fixed charge of the membrane surface is screened as ionic concentration increases facilitating transport through the membrane. At fixed concentration, increasing the pH from 5.6 to 9.3 also results in decreasing rejection. Above pH 9.3 the rejection once more increases. This rejection phenomenon corresponds to a minimum at the membrane IEP at pH 9.3 and represents the membrane being positively charged at lower pH values, neutral at the IEP, and negatively charged above the IEP. Fig. 4 illustrates that the

38

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

3.1.3.

Effect of multi valent salts on membrane rejection

Rejection experiments for various single salt solutions of increasing ion valence were conducted and the results are shown in Fig. 5. The rejection data of NaCl, MgCl2, and CrCl3 in Fig. 5(a) demonstrates the rejection of an increasing valence cation from the positively charged membrane. In this instance, as a result of the Donnan exclusion mechanism, one would expect the order of the salt rejection to be R(CrCl3) > R(MgCl2) > R(NaCl). This order is maintained with respect to the Mg2þ and Naþ ions, however, the Cr3þ ion does not exhibit the highest rejection as expected. Moreover, the limiting rejection value is actually comparable with Naþ at approximately 87%. The physical properties of each species is provided in Table 1. The reported Cr3þ ion size is larger than that for the Mg2þ ion for both the Stokes radii and hydrated radii and the bulk diffusion coefficient for Cr3þ is slightly lower than that reported for Mg2þ, confirming that this species should indeed have a larger ionic radii than Mg2þ. Therefore, the Cr3þ species is physically larger and more highly charged than the Mg2þ species, indicating that the Fig. 3 e Effective membrane charge density, Xd, derived for the various membranes. PA6DT-C positive magnitude, Nanomax-50 and Desal-DK both negative magnitude (data for both commercial membranes from previous work (Oatley, 2004)).

a

1.0

optimum rejection performance for small ions using this membrane is in the region below pH 8.0 and at concentrations below 15 mol m
3. Outside of this region the rejection falls below 40% and most industrial processes will not be viable at these recovery values. Note that Fig. 4 would benefit from further data points around the membrane IEP, especially at low concentration, and this will be the focus of further research.

Rejection

0.8

0.6

0.4

NaCl MgCl2 CrCl3

0.2 0.0

0.1

0.2

0.3

0.4

0.5

0.6

Δ Pe / MPa

b

1.0 MgCl2 NaCl Na2SO4

Rejection

0.8

0.6

0.4

0.2

0.0 0

5

10

15

20

Concentration / mol m

Fig. 4 e Rejection of NaCl at the PA6DT-C membrane as a function of pH and concentration.

25

30

-3

Fig. 5 e Rejection of multi-valent ions at the PA6DT-C membrane: (a) Cationic species vs effective pressure at 1 mol mL3 concentration and (b) anionic species vs concentration at fixed pressure [400 kPa].

35

39

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

a

2.5 0.4 mol m-3 -3 1.1 mol m 10.5 mol m-3 best fit

1.5

-5

Jv / 10 m s

-1

2.0

1.0

0.5

0.0 0.0

b

Colour removal from wastewaters

The first industrial application considered in this work was the removal of low concentration pollutants from a watercourse, in this case removal of colouration caused by low level dye contamination. This example was selected in order to fully demonstrate the capabilities of the new highly positively charged and high flux membrane. The retention and flux of different concentrations of the MB dye at constant pH are shown in Fig. 6. The concentration range selected was considered to be representative of medium to concentrated dye house waste streams (typically the dye is 5e10% of the solution, i.e. 0.05e0.1 g L
1 (Koyuncu et al., 2004; Yu et al., 2010)), lower concentrations of dye would simply exhibit similar or higher rejection and flux. The membrane flux, Fig. 6(a), obtained for both the 0.4 and 1.1 mol m
3 solutions was very similar at approximately 4.7  10
11 ms
1 Pa
1 or 16.9 LMH bar
1 and decreased only slightly for 10.5 mol m
3 (4.2  10
11 ms
1 Pa
1 or 15.2 LMH bar
1), indicating that a small degree of concentration polarisation was occurring. The magnitude of this flux value is significantly higher than that normally observed for commercial NF membranes (typically in the range 1e5 LMH bar
1) demonstrating the superior flux rates of this new membrane. The membrane rejection increased with increasing pressure for each of the concentrations. However, the rejection was high (>95% in all cases) and very similar at all pressures, exhibiting a maximum rejection of approximately 98% for the pressure range studied. Hence, under these conditions, the rejection of

0.2

0.3

0.4

0.5

0.6

1.0

0.8

0.6

0.4

0.4 mol m-3 1.1 mol m-3 10.5 mol m-3

0.2

0.0 0.0

3.2.

0.1

Δ Pe / MPa

Rejection

rejection profile should be increased in comparison which is clearly not the case. There has been some debate over which radii should be used when considering NF pore flow (Bowen and Welfoot, 2002), however, in this case the size trend is equivalent for both ions and thus steric effects alone do not explain this unusual phenomenon (for purely steric interactions the limiting rejection for each species is Naþ ¼ 0.216, Mg2þ ¼ 0.291 and Cr3þ ¼ 0.328). Thus, the increased permeation behaviour observed must be attributed to a physicalechemical interaction with the membrane either from adsorption or chelating of the Cr3þ ion either at the membrane surface or within the pore structure. However, the current fundamental lack of understanding of ion hydration and interactions inside an NF pore renders further meaningful discussion impossible. The rejection data for NaCl, MgCl2 and Na2SO4 at constant pH and increasing concentration presented in Fig. 5(b) demonstrates the rejection of an increasing valence anion from the positively charged membrane. In this case, one would expect the order of rejection to be R(MgCl2) > R(NaCl) > R(Na2SO4) as both MgCl2 and NaCl have a common counter ion and Mg2þ should be more highly rejected than Naþ for the reasons explained previously. The SO24 counter ion should experience higher transport across the membrane due to the increased electrostatic attraction when compared with the mono-valent Cl
ion. This is the case observed at all concentrations and the general trend across each of the salts is that rejection falls as concentration increases as would be expected due to screening of the membrane charge as ionic strength increases.

0.1

0.2

0.3

0.4

0.5

0.6

Δ Pe / MPa

Fig. 6 e The (a) flux and (b) retention of methylene blue versus effective pressure for various concentrations at pH 5.6 from the PA6DT-C membrane.

dye from the PA6DT-C membrane surface is predominantly independent of concentration and confirms that the membrane surface is indeed highly positively charged (for purely steric interactions the rejection of MB is 35.3%). The influence of pH on the rejection of 1.1 mol m
3 MB solution is shown in Fig. 7. The rejection exhibits a slight decrease from R ¼ 97.1% to R ¼ 94.2% in the pH range 3.0e8.5 and then decreases sharply when approaching pH 9.5 to R ¼ 56.2%. This rapid decline in rejection corresponds to the membrane IEP. The higher transmission of MB at the membrane IEP demonstrates that steric rejection for this compound is low and the high rejection values observed in Fig. 6 are indeed the result of Donnan exclusion, indicating that the membrane is highly charged and suffers only slight screening at the highest concentration used. Thus, the PA6DT-C membrane is capable of both high resolution removal of low level cationic contaminants from wastewaters and very high flux. Indeed, the high membrane flux rates observed indicate that the process economics of such a clean-up operation should be highly favourable.

40

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

a

1.0

Rejection

0.8

0.6

11.8 mol m-3 4.6 mol m-3 1.0 mol m-3 Pure methylene blue

0.4

0.2 0

100

200

300

400

500

600

-2

Δ Pe / kN m

b Fig. 7 e The rejection of a 1.1 mol mL3 solution of methylene blue versus pH for the PA6DT-C membrane. Rejection

Value added recovery: feasibility for reuse of the dye

The rejection of a synthetic dye:salt mixture was studied by fixing MB concentration at 1.1 mol m
3 whilst varying the NaCl concentration from 1.0 to 11.8 mol m
3 Fig. 8(a) shows the rejection for the MB component of the solution and indicates that at low salt concentration (1.0 mol m
3) the maximum MB rejection is 97%. However, as the salt concentration is increased to 4.6 mol m
3 and then further to 11.8 mol m
3, the MB rejection falls to 94 and 91% respectively at 500 kPa. This slight decline in rejection is attributed to charge screening by the increased NaCl concentration allowing a small quantity of the MB to permeate the membrane. Fig. 8(b) shows the rejection of the NaCl component of the solution and indicates that as the concentration rises the rejection decreases. This is the principle of Donnan exclusion and similarly corresponds to the data presented in Fig. 2(a). The diafiltration experiment was carried out using the initial concentrations of 10.9 and 1.0 mol m
3 for NaCl and MB respectively and used a constant pressure of 400 kPa. The membrane flux was recorded and samples of the permeate were taken periodically in order to conduct a mass balance over the diafiltration tank (eq. (11)), the result of which is shown in Fig. 9. The rejection of MB was greater than 93% at all times and the total recovery of dye was 50.8%. The total diafiltration time was 6.25 h and was equivalent to 8.3 diafiltration volumes. The salt concentration was reduced from 10.9 to 0.83 mol m
3, which is a total removal of 91.3%. The membrane flux during the experiment showed a slight initial decline and then stabilised at 1.38  10
5 ms
1 Pa
1 (12.4 LMH bar
1). In this simulated separation, the degree of recovery is relatively low at 50.8%. However, for a real dye house process, the tolerable level of salt in a recycle stream may be greater than that obtained during this experiment and would facilitate a higher recovery of the MB dye. Also, the

0.8

0.6

0.4 11.8 mol m -3 4.6 mol m -3 1.0 mol m -3 pure NaCl 1.0 mol m -3

0.2

0.0 0

100

200

300

400

500

600

Δ P e / kN m -2

Fig. 8 e The rejection of methylene blue:NaCl mixtures at the PA6DT-C membrane, (a) Methylene blue rejection and (b) NaCl rejection. MB concentration fixed at 1.1 molL3.

degree of recovery which is cost efficient depends heavily on the specific application (De Florio et al., 2005). Therefore, a sensible cost analysis is not possible for this artificial recovery process, however, the example does demonstrate 10

8

Cf / mol m-3

3.2.1.

1.0

NaCl MB

6

4

2

0 0

1

2

3

4

5

6

time / h

Fig. 9 e Concentration of NaCl and methylene blue in the diafiltration tank over time.

7

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

that the membrane is capable of value added recovery of the dye species by salt removal and at high flux rates, which may offer attractive economics.

4.

Conclusion

The novel positively charged PA6DT-C membrane has been shown to be an important development for application in the process industries. The magnitude of the fixed charge of this new membrane is significantly greater than two commercially available membranes and the flux rate of this new membrane was also significantly higher (w17 LMH bar
1 in comparison to w5 LMH bar
1). This demonstrates that the new membrane is capable of achieving both high rejection and high membrane flux for separations involving low concentrations of cationic materials, which is essential for favourable process economics. The positively charged membrane has been characterised with simple salts and an understanding of the change in membrane fixed charge with both pH and concentration has been developed. The optimum processing conditions to achieve high rejections for the recovery of small cations was found to be in the range pH < 8.0 and concentration less than 15 mol m
3. The increased fixed charge and high flux of this new membrane was exploited by considering the potential for industrial applications such as colour removal from dye house effluents. The membrane was employed to remove methylene blue, a positively charged organic dye, from a simulated low concentration effluent typical of the industrial sector. The membrane performed extremely well and a rejection of 98% was achieved at 5 bar with a membrane flux of w17 LMH bar
1. The membrane was then further evaluated for industrial use by considering the value added recovery and reuse of the dye from typical dye house bath water. A diafiltration experiment was conducted and the membrane was successful in desalting the dye and producing a process stream suitable for recycle. Thus, this new membrane both adds to and complements the existing short supply of positively charged NF membranes and is suitable for applications such as the recovery of valuable cationic macromolecules in the bioprocess and pharmaceutical industries or removal of multivalent cations such as dyes and heavy metals in the paper and pulp, textiles, nuclear, and automotive industries.

references

Ahmad, A.L., Ooi, B.S., 2010. A study on acid reclamation and copper recovery using low pressure nanofiltration membrane. Chemical Engineering Journal 156 (2), 257e263. Avgidou, M.S., Kaldis, S.P., Sakellaropoulos, G.P., 2004. Membrane cascade schemes for the separation of LPG olefins and paraffins. Journal of Membrane Science 233 (1e2), 21e37. Avlonitis, S.A., Poulios, I., Sotiriou, D., Pappas, M., Moutesidis, K., 2008. Simulated cotton dye effluents treatment and reuse by nanofiltration. Desalination 221, 259e267. Bellona, C., Drewes, J.E., 2005. The role of membrane surface charge and solute physico-chemical properties in the rejection of organic acids by NF membranes. Journal of Membrane Science 249 (1e2), 227e234.

41

Bouranene, S., Fievet, P., Szymczyk, A., Samar, M.E., Vidonne, A., 2008. Influence of operating conditions on the rejection of cobalt and lead ions in aqueous solutions by a nanofiltration polyamide membrane. Journal of Membrane Science 325, 150e157. Bowen, W.R., Welfoot, J.S., 2002. Modelling the performance of membrane filtration - critical assessment and model development. Chemical Engineering Science 57, 1121. Bowen, W.R., Cassey, B., Jones, P., Oatley, D.L., 2004. Modelling the performance of membrane nanofiltrationeapplication to an industrially relevant separation. Journal of Membrane Science 242, 211e220. Bowen, W.R., Cheng, S.Y., Doneva, T.A., Oatley, D.L., 2005. Manufacture and characterisation of polyetherimide/ sulfonated poly(ether ether ketone) blend membranes. Journal of Membrane Science 250, 1e10. Cheng, S., Oatley, D.L., Williams, P.M., Wright, C., 2011. Positively charged nanofiltration membranes: review of current fabrication methods and introduction of a novel approach. Advances in Colloid and Interface Science 164, 12e20. Chmiel, H., Kaschek, M., Blo¨cher, C., Noronha, M., Mavrov, V., 2003. Concepts for the treatment of spent process water in the food and beverage industries. Desalination 152 (1e3), 307e314. CRC Handbook of Chemistry and Physics, 90th ed., 2009 CRC Press (Taylor & Francis group). De Florio, L., Giordano, A., Mattioli, D., 2005. Nanofiltration of lowcontaminated textile rinsing effluents for on-site treatment and reuse. Desalination 181, 283e292. Dresch, M., Daufin, G., Chaufer, B., 2001. Integrated membrane regeneration process for dairy cleaning-in-place. Separation and Purification Technology 22e23, 181e191. Erswell, A., Brouchaert, C.J., Buckley, C.A., 1988. The reuse of reactive dye liquors using charged ultrafiltration membrane technology. Desalination 70, 157e167. Forgacs, E., Cserhati, T., Oros, G., 2004. Removal of synthetic dyes from wastewaters: a review. Environment International 30, 953e971. Hessel, C., Allegre, C., Maisseu, M., Charbit, F., Moulin, P., 2007. Guidelines and legislation for dye house effluents. Journal of Environmental Management 83, 171e180. Holmes, D.H., October 2002. Water and chemicals recovery in the German automotive industry. Membrane Technology, 6e10. Hwang, E.D., Lee, K.W., Choo, K.H., Choi, S.J., Kim, S.H., Yoon, C.H., Lee, C.H., 2002. Effect of precipitation and complexation on nanofiltration of strontium-containing nuclear wastewater. Desalination 147, 289e294. Khawajia, A.D., Kutubkhanah, I.K., Wie, J.M., 2007. A 13.3 MGD seawater RO desalination plant for Yanbu industrial city. Desalination 203 (1e3), 176e188. Koyuncu, I., Topacik, D., Yuksel, E., 2004. Reuse of reactive dyehouse wastewater by nanofiltration: process water quality and economical implications. Separation and Purification Technology 36, 77e85. Lau, W.J., Ismail, A.F., 2009. Polymeric nanofiltration membranes for textile dye wastewater treatment: preparation, performance evaluation, transport modelling, and fouling control - a review. Desalination 245, 321e348. Levenstein, R., Hasson, D., Semiat, R., 1996. Utilisation of the Donnan effect for improving electrolyte separation with nanofiltration membranes. Journal of Membrane Science 116, 77e92. Mo, J.H., Lee, Y.H., Kim, J., Jeong, J.Y., Jegal, J., 2008. Treatment of dye aqueous solutions using nanofiltration polyamide composite membranes for the dye wastewater reuse. Dyes and Pigments 76, 429e434. Nataraj, S.K., Hosamani, K.M., Aminabhavi, T.M., 2009. Nanofiltration and reverse osmosis thin film composite membrane module for the removal of dye and salts from the simulated mixtures. Desalination 249, 12e17.

42

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

Nguyen, M., Reynolds, N., Vigneswaran, S., 2003. By-product recovery from cottage cheese production by nanofiltration. Journal of Cleaner Production 11 (7), 803e807. Nightingale, E.R., 1959. Phenomenological theory of ion salvation. Journal of Chemical Physics 63, 1381e1387. Noronha, M., Britz, T., Mavrov, V., Janke, H.D., Chmiel, H., 2002. Treatment of spent process water from a fruit juice company for purposes of reuse: hybrid process concept and on-site test operation of a pilot plant. Desalination 143 (2), 183e196. Oatley, D.L. 2004, Characterisation and prediction of membrane separation performance e an industrial assessment, Ph. D thesis, Swansea UK. Oatley, D.L., Cassey, B., Jones, P., Bowen, W.R., 2005. Modelling the performance of membrane nanofiltrationerecovery of a highvalue product from a process waste stream. Chemical Engineering Science 60, 1953e1964. Pouliot, Y., Wijers, M.C., Gauthier, S.F., Nadeau, L., 1999. Fractionation of whey protein hydrolysates using charged UF/ NF membranes. Journal of Membrane Science 158, 105e114. Schaefer, A., Fane, A.G., Waite, T.D., 2005. Nanofiltration: Principles and Applications. Elsevier Advanced Technology, ISBN 1856174050. Shaalan, H.F., Sorour, M.H., Tewfik, S.R., 2001. Simulation and optimization of a membrane system for chromium recovery from tanning wastes. Desalination 141 (3), 315e324. Tschirch, J., Dillert, R., Bahnemann, D., Proft, B., 2008. Photodegradation of methylene blue in water, a standard method to determine the activity of photocatalytic coatings? Research on Chemical Intermediates 34 (4), 381e392. Tylkowski, B., Tsibranska, I., Kochanov, R., Peeva, G., Giamberini, M., 2011. Concentration of biologically active compounds extracted from Sideritis ssp.L. by nanofiltration. Food and Bioproducts Processing 89 (4), 307e314. Vandanjon, L., Cros, S., Jaouen, P., Que´me´neur, F., Bourseau, P., 2002. Recovery by nanofiltration and reverse osmosis of

marine flavours from seafood. Desalination 144 (1e3), 379e385. Van der Bruggen, B., Daems, B., Wilms, D., Vandecasteele, C., 2001. Mechanisms of retention and flux decline for the nanofiltration of dye baths from textile industry. Separation and Purification Technology 22e23, 519e528. Walha, K., Amar, R.B., Firdaous, L., Que´me´neur, F., Jaouen, P., 2007. Brackish groundwater treatment by nanofiltration, reverse osmosis and electrodialysis in Tunisia: performance and cost comparison. Desalination 207 (1e3), 95e106. Wolters, R., Wendler, B., Schmidt, B., Holdinghausen, A., Prade, H., 2008. Rinsing water recovery in the steel industry e a combined UF/NF treatment. Desalination 224, 209e214. Yan, C., Zhang, S., Yang, D., Jian, X., 2008. Preparation and characterisation of chloromethylated/quaternized poly(phthalazinone ether sulfone ketone) for positively charged nanofiltration membranes. Journal of Applied Political Science 107, 1809e1816. Yaroshchuck, A.E., 2001. Non-steric mechanisms of nanofiltration: superposition of Donnan and dielectric exclusion. Separation and Purification Technology 22e33, 143e158. Yoon, Y., Westerhoff, P., Snyder, S.A., Wert, E.C., 2006. Nanofiltration and ultrafiltration of endocrine disrupting compounds, pharmaceuticals and personal care products. Journal of Membrane Science 270, 88e100. Yu, S., Liu, M., Ma, M., Qi, M., Lu¨, Z., Gao, C., 2010. Impacts of membrane properties on reactive dye remove from dye/salt mixtures by asymmetric cellulose acelate and composite polyamide nanofiltration membranes. Journal of Membrane Science 350, 83e91. Zakrzewska-Trznadel, G., 2003. Radioactive solutions treatment by hybrid complexation e UF/NF process. Journal of Membrane Science 225, 25e39.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 3 e5 2

Available online at www.sciencedirect.com

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

Efficient electricity generation from sewage sludge using biocathode microbial fuel cell Guodong Zhang a, Qingliang Zhao a,*, Yan Jiao b, Kun Wang a, Duu-Jong Lee c,d, Nanqi Ren a a

State Key Laboratory of Urban Water Resources and Environments (SKLUWRE), School of Municipal and Environmental Engineering, Harbin Institute of Technology, Harbin 150090, China b Applied Economic Research Institute, Shanxi University of Finance and Economics, Taiyuan 030006, China c Department of Environmental Science and Engineering, Fudan University, Shanghai 200344, China d Department of Chemical Engineering, College of Engineering, National Taiwan University of Science and Technology, Taipei 106, Taiwan

article info

abstract

Article history:

Microbial fuel cells (MFCs) with abiotic cathodes require expensive catalyst (such as Pt) or

Received 8 August 2011

catholyte (such as hexacynoferrate) to facilitate oxidation reactions. This study incorpo-

Received in revised form

rated biocathodes into a three-chamber MFC to yield electricity from sewage sludge at

27 September 2011

maximum power output of 13.2
1.7 W/m3 during polarization, much higher than those

Accepted 15 October 2011

previously reported. After 15 d operation, the total chemical oxygen demand (TCOD)

Available online 25 October 2011

removal and coulombic efficiency (CE) of cell reached 40.8
9.0% and 19.4
4.3%, respectively. The anolyte comprised principally acetate and propionate (minor) as metabolites.

Keywords:

The use of biocathodes produced an internal resistance of 36e46 U, lower than those re-

Electricity generation

ported in literature works, hence yielding higher maximum power density from MFC. The

Sewage sludge

massively parallel sequencing technology, 454 pyrosequencing technique, was adopted to

Microbial fuel cells (MFCs)

probe microbial community on anode biofilm, with dominant phyla belonging to Proteo-

Biocathode

bacteria (45% of total bacteria), Bacteroidetes (19%), Uncultured bacteria (9%), Actinobacteria

454 pyrosequencing technique

(7%), Firmicutes (7%), Chloroflex (7%). At genera level, Rhodoferax, Ferruginibacter, Propioni-

Microbial community

bacterium, Rhodopseudomonas, Ferribacterium, Clostridium, Chlorobaculum, Rhodobacter, Bradyrhizobium were the abundant taxa (relative abundances > 2.0%). ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Sludge management is an expensive practice (Rulkens, 2008). Microbial fuel cell (MFC) can be applied to convert the organic matters in sewage sludge to electricity under ambient temperature, normal pressure, and neutral pH (Dentel et al., 2004). Hu (2008) first utilized a baffle-chamber membraneless MFC to harvest electricity from an anaerobic sludge, but noted that its electricity productivity (0.3 mW/m2) was much lower than the tests using glucose as substrate (161 mW/m2).

Liu et al. (2009) promoted the maximum voltage and maximum power density (Pmax) to of 440.7 mV and 220.7 mW/ m2, respectively, from surplus sludge using single chamber floating-cathode MFCs. Jiang et al. (2009) reported a maximum power density of 8.5 W/m3 with open circuit potential (OCV) of 0.725 V during treatment of sewage sludge using a twochambered MFC with hexacynoferrate as catholytes. Liu et al. (2011) integrated a membrane-less MFC with activated sludge process and yielded the maximum power of the MFC as 2.34 W/m3 with current density up to 14 A/m3. The power

* Corresponding author. Tel.: þ86 451 86283017; fax: þ86 451 86282100. E-mail address: [email protected] (Q. Zhao). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.036

44

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 3 e5 2

density levels for existing sludge-MFC works are far lower than those using pure organic matters such as glucose as anodic substrate (Rabaey et al., 2003; Liu et al., 2005). Reduction rate of oxygen at the cathode presents one of the major factors limiting the oxygen-MFC system performance (Logan and Regan, 2006a; Logan, 2009; Clauwaert et al., 2007a; Rismani-Yazdi et al., 2008). Alternative electron acceptors such as hexacyanoferrate (Rabaey et al., 2005), ferricyanide (Ringeisen et al., 2006), permanganate (You et al., 2006) and H2O2 were applied for improving cathode efficiency; however, the cost of reagent regeneration is high (Logan, 2008). Additionally, the catalysts for abiotic cathodes, such as platinum or other non-precious metals (Zhao et al., 2005; Cheng et al., 2006), are also costly and can be poisoned in long-term applications. The use of biocathode has advantages over conventional cathodes (Franks and Nevin, 2010), including lower construction and operational costs (He and Angenent, 2006), better MFC performance (Clauwaert et al., 2007a), and excess production of useful products or removal of unwanted compounds in operation (Rozendal et al., 2008; Cheng et al., 2009; Clauwaert et al., 2007b). To boost the performance of MFCs, a better understanding is needed on how the operating conditions of MFC affect bacterial community (particularly exoelectrogenic populations), power densities, and recovery of the substrate as current. Though a wide variety of microorganisms is capable of respiring with the help of an electrode (Logan and Regan, 2006b), many of them can only utilize a limited range of substrates (Logan, 2009). Thus the complex mixture of organics such as sewage sludge suggests that diverse microbial communities are needed to convert the chemic energy from these organic matters to electricity. However, no electroactive communities have been reported to evolve to a quasi-axenic culture, even when fed with the same substrate for prolonged periods of time under the same conditions (Aelterman et al., 2006; Clauwaert et al., 2008). Furthermore, in previous MFC studies, bacterial community analysis was limited in that they could not capture information on anode bacterial populations with a relative abundance of less than 1% due to the limitations of the sequencing detection methods they used (Aelterman et al., 2006; Clauwaert et al., 2008; Ishii et al., 2008b; Kim et al., 2004, 2006; Logan and Regan, 2006b). Yet in general, rank abundance plots of microbial populations are long-tailed distributions in which a few highly abundant populations account for most of the organisms in a community, whereas lowerabundance taxa represent a large number of different phylotypes. And in MFC system, a rare microbial population often plays a critical role in the eco-physiology of an entire community (Kim et al., 2006; Pham et al., 2008; Logan and Regan, 2006a). This study aims at investigating the MFC performance when biocathodes were applied in a three-chambered MFC with sewage sludge as the anode substrate, and characterizing the soluble organic component of anolyte during operation to comprehend how these organic matters in sewage sludge are degraded. In addition, to obtain comprehensive microbial information, the 454 pyrosequencing technique was applied for sequencing anodic biofilm sample and phylogenetic analysis gave further insight in their composition.

2.

Materials and methods

2.1.

MFC construction

A three-compartment MFC (one anodic and two cathodic compartments joined with a pipe and cathode electrolyte could freely circulate in between) was constructed (Fig. 1). The MFC made of Plexiglas was consisted of one cylinder (480 mm  100 mm, anode compartment with two symmetrical square view windows (70 mm  70 mm)) and two uniform cubes (70 mm  80 mm  30 mm, cathode compartments, linking with each other by a Plexiglas pipe (420 mm)). Two proton exchange membranes (PEM) (Nafion 117, Dupont Co., Wilmington, USA) with the same cross-sectional area of 4900 mm2 (70 mm  70 mm) were used to separate the anode and the cathode compartments. The graphite fiber brush made of carbon fibers (STS40 24K, 650
17 m2/m3, average fiber diameter of 7.0 mm, Toho Tenax) were cut to a set length and were twisted by two titanium wires. The brushes embedded in graphite granules (diameter of 1e5 mm, 55 m2/ m3, Jiuxin Carbon Goods Co., Jilin, China) were the cathode, with three brushes placed in the anodic compartment as electrodes. Ways of preparing PEM and electrodes were available in Liu and Logan (2004).

2.2.

Operational conditions

The sewage sludge sample was collected from the secondary sedimentation tanks in a municipal wastewater treatment plant of Harbin (China) and was stored at 4  C. The sludge was tested within one week after collection. The sludge sample had water content of 98.8%, pH 6.9, total solids concentration (TS) of 12,765
359 mg/L, volatile solids concentration (VS) of 9872
735 mg/L, TCOD of 15,830
865 mg/L and soluble chemical oxygen demand (SCOD) of 108
4.7 mg/L (triplicate analysis). This sewage sludge was directly used as anodic inocula and substrate. The cathodes were inoculated with topsoil obtained from the turf at Harbin Institute of

Fig. 1 e Schematic drawing of the reactor. (1) Graphite fiber brush; (2) graphite granules; (3) proton exchange membrane (PEM); (4) Ag/AgCl reference; (5) blade stirrer; (6) air; (7) air bubbles; (8) external resistance; (9) inlet; (10) outlet.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 3 e5 2

Technology, Harbin, China. The cathodic medium contained (per L of deionized water): NH4Cl (1.0 g), K2HPO4 (1.2 g), MgSO4 (0.5 g), KCl (0.5 g), KH2PO4 (0.14 g), Fe2(SO4)3$H2O (0.01 g), yeast extract (0.02 g), and trace elements (Rabaey et al., 2005). The cathode chambers were continuously aerated at 200 mL/min to provide dissolved oxygen at the cathode and the anolyte was stirred (300 rpm) for 3 min per h with a blade stirrer. The net anodic compartment (NAC) was 426 mL and that of the total cathodic chambers were 336 mL. Experiments were conducted in fed-batch mode at 25
1  C. The anolyte was replenished with fresh sludge when the voltage of MFC was under 0.3 V over 6 h to initiate a new cycle.

2.3.

Analytics and computations

The voltage difference between two electrodes was recorded across a fixed load (100 U) by a multicenter voltage collection instrument (32-channel data acquisition system, PISO-813, ICP DAS, Co., Ltd, Beijing, China) connected to a personal computer. The anode and cathode potentials were measured against a reference electrode (Ag/AgCl, þ197 mV vs. standard hydrogen electrode, SHE) (model RE-5B, BASi, Ningbo, Jiangsu province, China). Current (I ), power (P ¼ IV), and Coulombic efficiency (CE) were calculated as previously described (Kim et al., 2005). The polarization curves for electrodes were obtained by recording the current response at linear potential decrease at 1 mV/s (Parstat 263A, Princeton Applied Research, Oak Ridge, TN, USA). The gained potential and current values were converted to volumetric power density that was normalized by net liquid volume of anode chamber. Internal resistance (Rint) was determined by the slope of polarization curves (Cheng et al., 2006). The contents of volatile fatty acids (VFA) were determined by gas chromatography (7890A GC-System, Agilent Technologies, USA) equipped with a flame ionization detector and a 15 m capillary column (Innowax; i.d. 0.53 mm; 19095N-121; Agilent Technologies). The oven temperature was programmed to increase from 50 to 170  C at the rate of 10  C/min. The injector and detector temperatures were set at 250  C. Total soluble saccharides and protein contents in sludge were respectively determined by the Phenol-Sulfuric Acid method and Lowry Protein Assay Kit (Mbchem, Shanghai, China). Analyses on pH, SCOD, TS, VS, TCOD, TN, NHþ 4 -N and TP were performed according to Standard Methods (APHA, 1998).

2.4.

Microbial community analysis

Anodic biofilm sample at 96 d operation was collected, whose genomic DNA was extracted using the PowerSoil DNA Isolation Kit (MoBio, Carlsbad, CA, USA) according to the manufacturer’s instructions, quantified the DNA with a nanodrop spectrophotometer, and documented its yield and purity (characterized by 260/280 nm absorbance ratio). We normalized the DNA to the same concentration for amplifying use. Fragments of 16S rRNA genes containing variable V3 regions were amplified from the extracted DNA with primer sets, BSF341 broad-range forward primer 50 -NNNNNNNCCTACGGGAGGCAGCAG-30 and the USR534 universal reverse primer 50 NNNNNNNATTACCGCGGCTGCTGG-30 with 7 unique barcodes to sort each sample from the mixed pyrosequencing outcomes.

45

Sample PCR mixtures were prepared in 50 mL volumes and included 1  High Fidelity PCR buffer (Invitrogen, Carlsbad, CA, USA), 0.2 mM deoxyribonucleoside triphosphates, 0.6 mM each of forward and reverse primers, 1.5 mM MgCl2, 0.4 mg/mL bovine serum albumin, 5 U Platinum Taq DNA Polymerase High Fidelity (Invitrogen, Carlsbad, CA, USA) and 100e200 ng DNA template. Reactions were run on a GenAmp PCR System 9700 (PerkineElmer Applied Biosystems, Foster City, CA, USA) under the following cycling conditions: 5 min initial denaturation at 95  C followed by 20 cycles of denaturing at 94  C for 30 s, annealing at 56  C for 30 s, extension at 72  C for 60 s, and a final extension at 72  C for 7 min. Negative controls (ultrapure water only) were included for the amplification reactions. After PCR amplification, the amplicons were purified by onetime gel electrophoresis/isolation and two-times purifications using a Wizard SV Gel and PCR Clean-Up System (Promega, Madison, Wisconsin, USA). Amplicon pyrosequencing was performed using a 454 Life Sciences GS-FLX sequencer (Roche, NJ, USA). After a sequencing run and basecalling, we sorted the sequences by unique tags using the 454 script (sff file) to separate and group all data and then trimmed the sequences using the 454 script (sffinfo) for downstream analysis. Tag sequences were screened for quality as recommend by Huse et al. (2007). After removing sequences of poor quality, distance matrices, cluster, rarefaction analysis and two indices of diversity (ACE and Chao1) were computed using the program MOTHUR ver. 1.17.0 (Schloss et al., 2009). Representative sequences from each operational taxonomic units (OUT) were phylogenetically assigned with taxonomic classifications obtained from the RDP-II Classifier (Cole et al., 2008; Wang et al., 2007), the National Centre for Biotechnology Information (NCBI) BLAST (Johnson et al., 2008), and the Greengenes databases (DeSantis et al., 2006).

3.

Results and discussion

3.1.

Electricity generation from sewage sludge

After start-up, the reactor revealed a transient lag phase of 24 h (Fig. 2), then a low power production (0.0027 W/m3 net anodic compartment (NAC), at REX ¼ 100 U) was noted since 2 d. On 10.5 d, the anode potential was declined to 400 mV (vs. Ag/AgCl), and the power output was increased to 4.5 W/m3 (cell voltage ¼ 0.529 V). Reproducible cycles of power productions were obtained with new sludge charge after 35 d, giving the highest power density (Pmax) of 8.26
0.4 W/m3 NAC with fixed resistance (REX ¼ 100 U). Take cycle 3 (operation time ¼ 41 d) as demonstration example. The maximum OCV of 0.94 V occurred on 21 d and 31 d (Fig. 3). From the slope analysis, the Rint of cell in cycle 3 was estimated as 36e46 U. The Pmax first slightly increased from 11.8 to 13.2 W/m3 with time, and then dropped to 9.9 W/ m3 on 37 d which was close to the Pmax of 31 d. But compared with those on 31 d (OCV ¼ 0.94 V, Rint ¼ 46 U), both of the OCV and Rint of cell decreased on 37 d (OCV ¼ 0.86 V, Rint ¼ 44 U). The Pmax of 13.2 W/m3 on 21 d was higher than those reported in literature (2.3e8.5 W/m3 with OCV < 0.725 V at Rint > 60 U; air cathode or soluble K3Fe(CN)6 catholytes using sewage

46

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 3 e5 2

6 4 2 0 0

20

40

60

80

100

120

Anode potential (mV vs. Ag/AgCl)

Time (d) 200

Anode potential

B

0 -200 -400 -600 0

20

40

60

80

100

120

Time (d) Fig. 2 e Time courses of power density generation and anode potential of biocathode MFC with sewage sludge as anode substrate (external resistance of 100 U, potentials calculated vs. Ag/AgCl, arrows indicate the points of feeding the anode bacteria with new sludge after the MFC’s operation for 36 and 75 d, respectively). (A) Power density curve, (B) Anode potential curves.

sludge as anodic substrate) (Jiang et al., 2009; Liu et al., 2011). Restated, the present biocathode MFC had low Rint, and hence high Pmax (¼OCV2/4Rint). This experimental finding correlated with the literature reports that the use of biocathodes reduced charge transfer resistance (part of Rint) of the cathode from 188 to 17 U (You et al., 2009) and from 40.2 to 12 U (Chen et al., 2008). Additionally, Bergel et al. (2005), Clauwaert et al. (2007a) and Rabaey et al. (2008) increased the MFC performance by adopting of a wet air cathode inoculated with a consortium of sludge and sediment microbes.

3.2.

Organic matter degradation and CE

Over the cycle 3, the TCOD was removed linearly with time, and reached 88.2
6.1% at the end of cycle. Conversely, the CE

A 1.0

7d 14 d 21 d 31 d 37 d

0.9

Cell voltage (V)

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1

was increased and reached a plateau of 21.3
1.5% since 20 d and onward (Fig. 4). In the first half of cycle (0e20 d), the soluble COD (SCOD) of anolyte was increased and peaked at 509
32 mg/L (Fig. 5A), signaling the occurrence of strong sludge hydrolysis. Passing over the peak, the SCOD was declined. Correspondingly, with certain data fluctuation, the concentrations of soluble saccharides and protein in anodic chamber were first increased, passing over a peak, and subsequently decreased. The peak time for soluble saccharides was on 5 d, while that for soluble proteins was on 20 d, suggesting that the saccharides were more difficult to be hydrolyzed from sludge (Fig. 5A). The total VFAs (acetate, propionate, butyrate and valerate) were increased to 75 mg/L on 4 d, with acetate and propionate (minor) as the principal metabolites (Fig. 5B). The subsequent degradation rate was enhanced (6e10 d) to reduce the VFA to a nearly constant level of around 30 mg/L. Afterwards, the hydrolysis rate was further increased to peak the VFA at around 72 mg/L on 14 d, with acetate, butyrate and valerate as major metabolites. Since 16 d and onward, the degradation overcompeted the hydrolysis that yielded continuous decline in VFA concentration (basically acetate) over time. After 20 d operation in cycle 3, the NHþ 4 -N concentration of anodic solution maximized at 45.2
4.1 mg/L as a hydrolysis product of sludge (Fig. 6). Correspondingly, the pH of anolyte was decreased from 6.9 to around 6.4. In subsequent stage, the NH4-N concentration was declined, while the anolyte pH rapidly reduced to 5.4 on 35 d. As Fig. 3 shows, the OCV and power density on 21 d were the maximum. This coincidence was in accordance with results that high ammonium nitrogen (under 500 mg/L) improved the performance of MFC (Nam et al., 2010a), probably owing to the increased anolyte conductivity with presence of excess NHþ 4 -N and protons, which reduced ohmic resistance and current increased generation in MFCs (Ishii et al., 2008a; Mohan and Das, 2009; Nam et al., 2010b). However, when anolyte pH was reduced from 7 to 5, MFC performance was deteriorated (Gil et al., 2003; He et al., 2008). Martin et al. (2010) suggested that the optimal pH for their MFC ranged 6.25e6.5, and noted critical increase in Rint (and hence drop in electricity production) at pH 5.5. Low pH can limit microbial activities on anodic biofilm, hence

B 14 -3

Power density

A

8

Power density (W m )

Power density -3 (W m )

10

7d 14 d 21 d 31 d 37 d

12 10 8 6 4 2

0.0 0

10

20

30

40

50 -3

Current density (A m )

60

0

10

20

30

40

50

60

-3

Current density (A m )

Fig. 3 e Polarization curves (1 mV/s, 360 min OCV before run, forward sweeps) for biocathode MFC on 7 d, 14 d, 21 d and 31 d after the replacement of new sludge on 75 d. (A) MFC voltage in function of the current (A/m3 projected anode liquid volume). (B) Power output (W/m3 projected anode liquid volume) in function of current (A/m3 projected anode liquid volume).

47

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 3 e5 2

Coulombic efficiency TCOD removal efficiency

45 40

Operation time (d)

35 30 25 20 15 10 5 0

10

20

30 40 50 60 70 80 90 100 TCOD removal and coulombic efficiencies (%)

Fig. 4 e Evolution of TCOD removal and coulombic efficiencies of MFC over time after the new sewage sludge was fed. Error bars of TCOD are ±SD based on samples analyzed in triplicate, and error bars (±SD) of CE are obtained from TCOD by calculation.

reducing MFC outputs. This met with the results of present study: when the pH dropped from 6.4 (21 d) to 5.5 (31 d), the Rint and Pmax observably changed. Continuous sludge feeding or pumping biocathode effluent to anode chamber can mitigate the adverse effects by pH drop.

3.3.

Community analysis

Of 827 total bacterial V3 amplicons (“tags”) sequenced from the anodic biofilm sample, after trimming, sorting, and quality control, 783 or 95% the sequences with an average read length of 177 bp were used in downstream analysis. The sequences could be clustered into 590, 526, 481 and 403 operational taxonomic units (OTUs) respectively at 1%, 3%, 5% and 10% distance thresholds (Table 1). These clusters served as OTUs for generating rarefaction curves (Fig. 7) and for making

600

B 80

SCOD Soluble saccharides Soluble Protein

500 400 300 200 100 0 0

10

20

Time (day)

30

Valeric acid Butirc acid Propionic acid Acetic acid

-1

VFA concentration (mg L )

Organic matter -1 concentration (mg L )

A

calculations with the abundance-based coverage estimator ACE (Chao and Lee, 1992; Chao, et al., 1993), the Chao1 richness estimations (Chao, 1984) and Shannon diversities (Chao and Bunge, 2002). Despite examining nearly 827 tags identified as bacterial, the ACE, Chao1 and Shannon indices indicated that the bacterial community on anode biofilm was high-diversity and our sampling of bacterial richness was far from complete (Table 1). The rarefaction curves at different cutoffs described unprecedented levels of bacterial complexity for anodic biofilm samples, yet none had reached the curvilinear or plateau phases (Fig. 7). The likelihood that they represent underestimates of the number of different kinds of bacteria in anodic sample is supported by observation of significant variation among tags with closest matches to the same sequence in V3 reference database, therefore indicating that deeper sequencing may be required to avoid underestimation of microbial diversity in our samples. Ribotypes were identified phylogenetically and were grouped by phylum or in the case of Proteobacteria, class, using the Global Alignment for Sequence Taxonomy approach as described previously (Sogin et al., 2006). The total frequency for a given phylogenetic group was calculated (Fig. 8). When grouped at the 97% similarity level, there were 153 phylotypes in the data set. Of the classifiable sequences, 12 phyla were identified across the sample set (Fig. 8). The dominant phyla were Proteobacteria (including 11% a-proteobacteria, 28% b-proteobacteria, 1.78% dproteobacteria and 3.95% g-proteobacteria), Bacteroidetes, uncultured bacteria, Actinobacteria, Firmicutes, Chloroflexi, representing approximately 45%, 19%, 9%, 7%, 7% and 7% of the sequences that could be classified below the domain level, respectively. The first 43 most abundant taxa (relative abundances > 0.5%) in the anodic biofilm data set are listed in Table 2. A distinctive feature of the anode community was the predominant (relative abundances  5.0%) groups of the genus Rhodoferax (relative abundance, 19.54%) and Ferruginibacter (5.36%) (Table 2). Subsequently, the subdominant taxa (relative abundances at 1.0e5.0%) included the genus of Erythrobacter, Propionibacterium, Levilinea, Rhodopseudomonas, Ferribacterium Clostridium, Chlorobaculum, Simplicispira, Rhodobacter, Bradyrhizobium, Ornithobacterium, Thermomonas, Desulforhabdus, Deferribacter, Oxalobacteraceae, and Pseudoxanthomonas. Members of these genera including Rhodoferax (Chaudhuri and Lovley, 2003),

40

70 60 50 40 30 20 10 0 5

10

15

20

25

30

35

40

Time (day)

Fig. 5 e Variation of organic matter concentration over operation. (A) Soluble COD, saccharides and proteins. (Error bars are ±SD based on samples analyzed in triplicate); (B) VFA.

48

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 3 e5 2

700 Ammonia pH

45

7.0

unique

600

0.01 40 35 6.0

30

OTUs

6.5

pH

Ammonia concentration -1 mg L

50

500

0.03

400

0.05

300

25 5.5

20 0

10

20

30

200 100

40

Time (day)

0

Fig. 6 e Evolution of NH4-N concentration and pH over time for MFC anolyte using sewage sludge as substrate.

Rhodopseudomonas (Park et al., 2001), Clostridium (Xing et al., 2008), have been shown to produce electricity in an MFC. In a comparison study of microbial communities in eight different MFCs, the phylum Proteobacteria was most frequently detected, followed by phylas Firmicutes and Bacteroidetes (Clauwaert et al., 2008). Chloroflexi were enriched on the anode of a cellulose-fed MFC (Ishii et al., 2008b), whose relevance could be cellulose hydrolyzing and/or electricity production. However, no specific ubiquitous microorganism has been detected in different MFCs, and it is difficult to link such community-structure information to assessment of microbial activity and MFC-reactor performance (Watanabe, 2008). The important role of Actinobacteria in environmental metabolic functioning is well known due to their involvement in decomposition of organic materials, such as cellulose and chitin (Lacey, 1973), thereby playing a crucial role in sludge organic matter turnover and hydrolyzing. Moreover, in present test, the phylas of Chlorobi (1.7%), Planctomycetes (1.4%) and Deferribacteres (1.0%) were important members of the anodic biofilm community. All phyla had been found to be majority of the community on anode of MFCs with different inoculum (Chlorobi, with both of the effluent from an acetatefed MFC and methanogenic culture (Schamphelaire et al., 2010); Planctomycetes, with wastewater or sludge inoculum (Kim et al., 2006); Deferribacteres, with wastewater inoculum (Jong et al., 2006)). A common strategy for isolating electrodereducing microorganisms is to employ Fe(III) as an electron

0

200 400 600 Number of tags sampled

800

Fig. 7 e Rarefaction analysis of anode biofilm sample based on pairwise distance. Rarefaction is shown for OTUs that contain unique sequences and OTUs with differences that do not exceed 1%, 3%, or 5%. OTUs with ‡97% and ‡95% pairwise sequence identity are arbitrarily assumed to form the same species and genus, respectively.

acceptor (Lovley, 2008). Thus we conferred that the populations about Fe(III) reducers such as Ferribacterium, Desulforhabdus, and Deferribacter had been microorganisms producing electrical current on the anodic surface. As mentioned above, the degradation overcompeted the hydrolysis over time (Fig. 5). Thus we could infer that the system had existed not less than two dominating microbial species which were responsible for the hydrolysis and electricity generation (with VFAs as fuel), respectively. When complex organic substrates served as fuel, it was expected that the microorganisms fermented these compounds to simpler substrates would also be components of the anode microbial community (Jung and Regan, 2007). As previously demonstrated in coculture studies (Ren et al., 2007), some

Proteobacteria Bacteroidetes Actinobacteria Firmicutes Chloroflexi Other bactraia Chlorobi

Table 1 e Similarity-based OTUs and species richness estimates. Cluster distance

OUT ACE Chao1 Shannon

Planctomycetes Deferribacteres Unclassified_Bacteria

Unique

0.01

0.03

0.05

0.1

654 14590 7511 6.240

590 7032 3778 6.050

526 6574 2664 5.792

481 4759 2037 5.627

403 2630 1255 5.354

Note: The species richness estimates were determined using the program MOTHUR as described in Methods.

Fig. 8 e Taxonomic breakdown of bacterial V3 tags from the anode biofilm of MFC using sewage sludge as substrate. Pie charts show the Phylum distribution for taxonomically assigned tags that occurred more than 8 times; the remaining tag sequences are grouped into “Other bacteria.”

49

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 3 e5 2

Table 2 e Phylogenetic classification of the clusters (relative abundance > 0.5%) in the anode biofilm. Phylum

Proteobacteria Bacteroidetes Chloroflexi Bacteroidetes Actinobacteria Chloroflexi Proteobacteria Proteobacteria Actinobacteria Bacteroidetes Firmicutes Proteobacteria Chlorobi Proteobacteria Proteobacteria Planctomycetes Bacteroidetes Proteobacteria Proteobacteria Deferribacteres Proteobacteria Proteobacteria Chloroflexi Actinobacteria Proteobacteria Proteobacteria Proteobacteria Proteobacteria Firmicutes Bacteroidetes Actinobacteria Proteobacteria Firmicutes Firmicutes Spirochaetes Proteobacteria Bacteroidetes Proteobacteria Proteobacteria Proteobacteria Firmicutes Proteobacteria Nitrospira

Class

Order

Family

Genus

Relative Abundance (%)

Number of Phylotypes

Betaproteobacteria Sphingobacteria Anaerolineae Sphingobacteria Actinobacteria Anaerolineae Alphaproteobacteria Betaproteobacteria Actinobacteria Flavobacteria Clostridia Alphaproteobacteria Chlorobia Betaproteobacteria Alphaproteobacteria Planctomycetacia Flavobacteria Gammaproteobacteria Deltaproteobacteria Deferribacteres Betaproteobacteria Gammaproteobacteria Anaerolineae Actinobacteria Alphaproteobacteria Alphaproteobacteria Betaproteobacteria Alphaproteobacteria Clostridia Bacteroidia Coriobacteridae Betaproteobacteria Clostridia Clostridia Spirochaetes Alphaproteobacteria Sphingobacteria Betaproteobacteria Betaproteobacteria Alphaproteobacteria Clostridia Alphaproteobacteria Nitrospira

Burkholderiales Sphingobacteriales Anaerolineales Sphingobacteriales Actinomycetales Anaerolineales Rhizobiales Rhodocyclales Actinomycetales Flavobacteriales Clostridiales Rhizobiales Chlorobiales Burkholderiales Rhodobacterales Planctomycetales Flavobacteriales Xanthomonadales Syntrophobacterales Deferribacterales Burkholderiales Xanthomonadales Anaerolineales Actinomycetales Rhodobacterales Sphingomonadales Rhodocyclales Rhodobacterales Clostridiales Bacteroidales Coriobacteriales Burkholderiales Clostridiales Clostridiales Spirochaetales Rhizobiales Sphingobacteriales Rhodocyclales Burkholderiales Rhizobiales Clostridiales Rhizobiales Nitrospirales

Comamonadaceae Chitinophagaceae Anaerolineaceae Erythrobacteraceae Propionibacteriaceae Anaerolineaceae Bradyrhizobiaceae Rhodocyclaceae Propionibacteriaceae Flavobacteriaceae Clostridiaceae Bradyrhizobiaceae Chlorobiaceae Comamonadaceae Rhodobacteraceae Planctomycetaceae Flavobacteriaceae Xanthomonadaceae Syntrophobacteraceae Deferribacteraceae Oxalobacteraceae Xanthomonadaceae Anaerolineaceae Micrococcineae Rhodobacteraceae Sphingomonadaceae Rhodocyclaceae Rhodobacteraceae Eubacteriaceae Porphyromonadaceae Coriobacterineae Comamonadaceae Ruminococcaceae Peptostreptococcaceae Spirochaetaceae Phyllobacteriaceae Chitinophagaceae Rhodocyclaceae Comamonadaceae Bradyrhizobiaceae Clostridiaceae Hyphomicrobiaceae Nitrospiraceae

Rhodoferax Ferruginibacter Unclassified Erythrobacter Propionibacterium Levilinea Rhodopseudomonas Ferribacterium Unclassified Chryseobacterium Clostridium Bradyrhizobium Chlorobaculum Simplicispira Rhodobacter Singulisphaera Ornithobacterium Thermomonas Desulforhabdus Deferribacter Oxalobacteraceae Pseudoxanthomonas Longilinea unclassified Haematobacter Sphingomonas Unclassified Pseudorhodobacter Anaerovorax Unclassified Unclassified Curvibacter Unclassified Sporacetigenium Treponema Mesorhizobium Terrimonas Azoarcus Acidovorax Bosea Anaerobacter Hyphomicrobium Nitrospira

19.54 5.36 3.19 2.55 2.55 2.43 2.43 2.17 2.17 2.17 2.04 2.04 1.79 1.53 1.53 1.40 1.28 1.15 1.15 1.02 1.02 1.02 0.89 0.89 0.89 0.77 0.77 0.77 0.77 0.64 0.64 0.64 0.64 0.64 0.64 0.64 0.51 0.51 0.51 0.51 0.51 0.51 0.51

153 31 25 20 20 19 19 17 17 17 16 16 14 12 12 11 10 9 9 8 8 8 7 7 7 6 6 6 6 5 5 5 5 5 5 5 4 4 4 4 4 4 4

fermentative microorganisms might have little or no capacity for electron transfer to the anode, but their metabolism was key to powering microbial fuel cells. Fermentative bacteria were also able to produce current in microbial fuel cells, however, only one third of the electrons were possibly available for electricity generation whereas two thirds remained in the produced fermentation products such as acetate and butyrate (Logan, 2004); the transfer of electrons to the anode was probably mediated by hydrogenases situated in the membrane surface (McKinlay and Zeikus, 2004). Because of the high bacterial diversity of anode microbial communities, including current producing bacteria and non-current producing microorganisms, the diversity with function of anode biofilms was difficult to be analyzed, and the

environmental factors influencing this competition and the mechanisms had not been elucidated. The microbial community data for the present MFC assessed using 454 pyrosequencing differ from the previously identified by the clone library analysis (Clauwaert et al. 2008). The present study revealed that b and a-subclass of Proteobacteria, Bacteroidetes, Actinobacteria and Clostridia were the major groups on anodic biofilms (Table 2); while g, b, a and dsubclass of Proteobacteria and Bacilli were the major groups by Clauwaert et al. (2008). The 454 pyrosequencing also revealed the populations of Chloroflexi, Deferribacteres, Spirochaetes and Nitrospira were not identified before. Based on the 16S rRNA sequence reads in the present MFC, many strains are novel that need further identification (Fig. 8 and Table 2).

50

4.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 3 e5 2

Conclusions

This study demonstrated an economic and efficient system for simultaneous sewage sludge treatment and electricity generation. For ending a repeated operation cycle, the 88.2
6.1% TCOD of sludge was removed and 21.3
1.5% of that was converted into electricity. Compared with previous study, our biocathode MFC reactor improved the maximum power density more than 55% due to high OCV (0.94 V) and low Rint (36e46 U). During later operation stage for batch-fed, the pH drop as a result of sludge hydrolysis limited the power output from MFC. The 454 pyrosequencing technology was successfully applied to analyze the microbial community in anode biofilm, and provided much more extensive information.

Acknowledgments The authors gratefully acknowledge funding from Project 50776024 and Project 50821002 (National Creative Research Groups) supported by National Nature Science Foundation of China, National Water Pollution Control and Management Key Project (2009ZX07317-008), and partial supports by State Key Laboratory of Urban Water Resource and Environment, Harbin Institute of Technology (No. 2010DX17).

references

Aelterman, P., Rabaey, K., Pham, H.T., Boon, N., Verstraete, W., 2006. Continuous electricity generation at high voltages and currents using stacked microbial fuel cells. Environmental Science & Technology 40 (10), 3388e3394. APHA, 1998. Standards Methods for the Examination of Water and Wastewater, 20th ed. American Public Health Association, Washington, DC, USA. Bergel, A., Fe`ron, D., Mollica, A., 2005. Catalysis of oxygen reduction in PEM fuel cell by seawater biofilm. Electrochemistry Communications 7 (9), 900e904. Chao, A., 1984. Nonparametric estimation of the number of classes in a population. Scandinavian Journal of Statistics 11, 265e270. Chao, A., Bunge, J., 2002. Estimating the number of species in a stochastic abundance model. Biometrics 58 (3), 531e539. Chao, A., Lee, S.-M., 1992. Estimating the number of classes via sample coverage. Journal of the American Statistical Association 87 (417), 210e217. Chao, A., Ma, M.C., Yang, K., 1993. Stopping rules and estimation for recapture debugging with unequal failure rates. Biometrika 80 (1), 193e201. Chaudhuri, S.K., Lovley, D.R., 2003. Electricity generation by direct oxidation of glucose in mediatorless microbial fuel cells. Nature Biotechnology 21 (10), 1229e1232. Chen, G.W., Choi, S.J., Lee, T.H., Lee, G.Y., Cha, J.H., Kim, C.W., 2008. Application of biocathode in microbial fuel cells: cell performance and microbial community. Applied and Microbiology Biotechnology 79 (3), 379e388. 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 (1), 364e369. Cheng, S.A., Xing, D.F., Logan, B.E., 2009. Direct biological conversion of electrical current into methane by

electromethanogenesis. Environmental Science & Technology 43 (10), 3953e3958. Clauwaert, P., van der Ha, D., Boon, N., Verbeken, K., Verhaege, M., Rabaey, K., Verstraete, W., 2007a. Open air biocathode enables effective electricity generation with microbial fuel cells. Environmental Science & Technology 41 (21), 7564e7569. Clauwaert, P., Rabaey, K., Aelterman, P., De Schamphelaire, L., Pham, T.H., Boeckx, P., Boon, N., Verstraete, W., 2007b. Biological denitrification in microbial fuel cells. Environmental Science & Technology 41 (9), 3354e3360. Clauwaert, P., Aelterman, P., Pham, T.H., Schamphelaire, L.D., Carballa, M., Rabaey, K., Verstraete, W., 2008. Minimizing losses in bio-electrochemical systems: the road to applications. Applied Microbiology and Biotechnology 79 (6), 901e913. Cole, J.R., Wang, Q., Cardenas, E., Fish, J., Chai, B., Farris, R.J., Kulam-Syed-Mohideen, D.M., McGarrell, D.M., Marsh, T., Garrity, G.M., Tiedje, J.M., 2008. The ribosomal database project: improved alignments and new tools for rRNA analysis. Nucleic Acids Research 37 (Database issue), D141eD145. Dentel, S.K., Strogen, B., Chiu, P., 2004. Direct generation of electricity from sludges and other liquid wastes. Water Science and Technology 50 (9), 161e168. DeSantis, T.Z., Hugenholtz, P., Larsen, N., Rojas, M., Brodie, E.L., Keller, K., Huber, T., Dalevi, D., Hu, P., Andersen, G.L., 2006. Greengenes, a Chimera-checked 16S rRNA gene database and workbench compatible with ARB. Applied and Environmental Microbiology 72 (7), 5069e5072. Franks, A.E., Nevin, K.P., 2010. Microbial fuel cells, a current review. Energies 3 (5), 899e919. Gil, G.C., Chang, I.S., Kim, B.H., Kim, M., Jang, J.K., Park, H.S., Kim, H.J., 2003. Operational parameters affecting the performance of a mediator-less microbial fuel cell. Biosensors and Bioelectronics 18 (4), 327e334. He, Z., Angenent, L.T., 2006. Application of bacterial biocathodes in microbial fuel cells. Electroanalysis 18 (19e20), 2009e2015. He, Z., Huang, Y.L., Manohar, A.K., Mansfeld, F., 2008. Effect of electrolyte pH on the rate of the anodic and cathodic reactions in an air-cathode microbial fuel cell. Bioelectrochemistry 74 (1), 78e82. Hu, Z., 2008. Electricity generation by a baffle-chamber membraneless microbial fuel cell. Journal of Power Sources 179 (1), 27e33. Huse, S.M., Huber, J.A., Morrison, H.G., Sogin, M.L., Welch, D.M., 2007. Accuracy and quality of massively parallel DNA pyrosequencing. Genome Biology 8 (7), R143. Ishii, S., Hotta, Y., Watanabe, K., 2008a. Methanogenesis versus electrogenesis: morphological and phylogenetic comparisons of microbial communities. Bioscience, Biotechnology and Biochemistry 72 (2), 286e294. Ishii, S., Shimoyama, T., Hotta, Y., Watanabe, K., 2008b. Characterization of a filamentous biofilm community established in a cellulose-fed microbial fuel cell. BMC Microbiology 8 (1), 6e18. Jiang, J.Q., Zhao, Q.L., Zhang, J.N., Zhang, G.D., Lee, D.J., 2009. Electricity generation from bio-treatment of sewage sludge with microbial fuel cell. Bioresource Technology 100 (23), 5808e5812. Johnson, M., Zaretskaya, I., Raytselis, Y., Merezhuk, Y., McGinnis, S. , Madden, T.L., 2008. NCBI BLAST: a better web interface. Nucleic Acids Research 36 (Web Server issue), W5eW9. Jong, B.C., Kim, B.H., Chang, I.S., Liew, P.W.Y., Choo, Y.F., Kang, G. S., 2006. Enrichment, performance, and microbial diversity of a thermophilic mediatorless microbial fuel cell. Environmental Science & Technology 40 (20), 6449e6454. Jung, S., Regan, J.M., 2007. Comparison of anode bacterial communities and performance in microbial fuel cells with

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 3 e5 2

different electron donors. Applied Microbiology and Biotechnology 77 (2), 393e402. Kim, B.H., Park, H.S., Kim, H.J., Kim, G.T., Chang, I.S., Lee, J., Phung, N.T., 2004. Enrichment of microbial community generating electricity using a fuel-cell-type electrochemical cell. Applied Microbiology and Biotechnology 63 (6), 672e681. Kim, J.R., Min, B., Logan, B.E., 2005. Evaluation of procedures to acclimate a microbial fuel cell for electricity production. Applied Microbiology and Biotechnology 68 (1), 23e30. Kim, G.T., Webster, G., Wimpenny, J.W.T., Kim, B.H., Kim, H.J., Weightman, A.J., 2006. Bacterial community structure, compartmentalization and activity in a microbial fuel cell. Journal of Applied Microbiology 101 (3), 698e710. Lacey, J., 1973. In: Sykes, G., Skinner, F.A. (Eds.), Actinomycetales: Characteristics and Practical Importance. Academic Press, New York, pp. 231e251. Liu, H., Logan, B.E., 2004. Electricity generation using an aircathode single chamber microbial fuel cell in the presence and absence of a proton exchange membrane. Environmental Science & Technology 38 (14), 4040e4046. 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 & Technology 39 (2), 658e662. Liu, Z., Li, X., Jia, B., Zheng, Y., Fang, L., Yang, Q., Wang, D., Zeng, G., 2009. Production of electricity from surplus sludge using a single chamber floating-cathode microbial fuel cell. Water Science and Technology 60 (9), 2399e2404. Liu, X.W., Wang, Y.P., Huang, Y.X., Sun, X.F., Sheng, G.P., Zeng, R. J., Li, F., Dong, F., Wang, S.G., Tong, Z.H., Yu, H.Q., 2011. Integration of a microbial fuel cell with activated sludge process for energy-saving wastewater treatment: taking a sequencing batch reactor as an example. Biotechnology and Bioengineering 108 (6), 1260e1267. Logan, B.E., 2004. Extracting hydrogen electricity from renewable resources. Environmental Science & Technology 38 (9), 160Ae167A. Logan, B.E., 2008. Microbial Fuel Cells. Wiley, Hoboken. Logan, B.E., 2009. Exoelectrogenic bacteria that power microbial fuel cells. Nature Reviews Microbiology 7 (5), 375e381. Logan, B.E., Regan, J.M., 2006a. Electricity-producing bacterial communities in microbial fuel cells. Trends in Microbiology 14 (12), 512e518. Logan, B.E., Regan, J.M., 2006b. Microbial fuel cells e challenges and applications. Environmental Science & Technology 40 (17), 5172e5180. Lovley, D.R., 2008. The microbe electric: conversion of organic matter to electricity. Current Opinion in Biotechnology 19 (6), 564e571. McKinlay, J.B., Zeikus, J.G., 2004. Extracellular iron reduction is mediated in part by neutral red and hydrogense in Escherichia coli. Applied and Environmental Microbiology 70 (6), 3467e3474. Martin, E., Savadogo, O., Guiot, S.R., Tartakovsky, B., 2010. The influence of operational conditions on the performance of a microbial fuel cell seeded with mesophilic anaerobic sludge. Biochemical Engineering Journal 51 (3), 132e139. Mohan, Y., Das, D., 2009. Effect of ionic strength, cation exchanger and inoculum age on the performance of microbial fuel cells. International Journal of Hydrogen Energy 34 (17), 7542e7546. Nam, J.Y., Kim, H.W., Shin, H.S., 2010a. Ammonia inhibition of electricity generation in single-chambered microbial fuel cells. Journal of Power Sources 195 (19), 6428e6433. Nam, J.Y., Kim, H.W., Lim, K.H., Shin, H.S., Logan, B.E., 2010b. Variation of power generation at different buffer types and conductivities in single chamber microbial fuel cells. Biosensors and Bioelectronics 25 (5), 1155e1159.

51

Park, H.S., Kim, B.H., Kim, H.S., Kim, H.J., Kim, G.T., Kim, M., Chang, I.S., Park, Y.K., Chang, H.I., 2001. A novel electrochemically active and Fe(III)-reducing bacterium phylogenetically related to Clostridium butyricum isolated from a microbial fuel cell. Anaerobe 7 (6), 297e306. Pham, T.H., Boon, N., Aelterman, P., Clauwaert, P., De Schamphelaire, L., Vanhaecke, L., De Maeyer, K., Hofte, M., Verstraete, W., Rabaey, K., 2008. Metabolites produced by Pseudomonas sp enable a Grampositive bacterium to achieve extracellular electron transfer. Applied Microbiology and Biotechnology 77 (5), 1119e1129. 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. Rabaey, K., Ossieur, W., Verhaege, M., Verstraete, W., 2005. Continuous microbial fuel cells convert carbohydrates to electricity. Water Science and Technology 52 (1e2), 515e523. Rabaey, K., Read, S.T., Clauwaert, P., Freguia, S., Bond, P.L., Blackall, L. L., Keller, J., 2008. Cathodic oxygen reduction catalyzed by bacteria in microbial fuel cells. ISME Journal 2 (5), 519e527. Ren, Z.Y., Ward, T.E., Regan, J.M., 2007. Electricity production from cellulose in a microbial fuel cell using a defined binary culture. Environmental Science & Technology 41 (13), 4781e4786. Ringeisen, B.R., Henderson, E., Wu, P.K., Pietron, J., Ray, R., Little, B., Biffinger, J., Jones-Meehan, J., 2006. High power density from a miniature microbial fuel cell using Shewanella oneidensis DSP10. Environmental Science & Technology 40 (8), 2629e2634. Rismani-Yazdi, H., Carver, S.M., Christy, A.D., Tuovinen, O.H., 2008. Cathodic limitations in microbial fuel cells: an overview. Journal of Power Sources 180 (2), 683e694. Rozendal, R.A., Jeremiasse, A.W., Hamelers, H.V.M., Buisman, C.J. N., 2008. Hydrogen production with a microbial biocathode. Environmental Science & Technology 42 (2), 629e634. Rulkens, W., 2008. Sewage sludge as a biomass resource for the production of energy: overview and assessment of the various options. Energy Fuels 22 (1), 9e15. Schloss, P.D., Westcott, S.L., Ryabin, T., Hall, J.R., Hartmann, M., Hollister, E.B., Lesniewski, R.A., Oakley, B.B., Parks, D.H., Robinson, C.J., Sahl, J.W., Stres, B., Thallinger, G.G., Van Horn, D.J., Weber, C.F., 2009. Introducing MOTHUR: opensource, platform-independent, community supported software for describing and comparing microbial communities. Applied and Environmental Microbiology 75 (23), 7537e7541. Schamphelaire, L.D., Cabezas, A., Marzorati, M., Friedrich, M.W., Boon, N., Verstraete, W., 2010. Microbial community analysis of anodes from sediment microbial fuel cells powered by Rhizodeposits of living rice plants. Applied and Environmental Microbiology 76 (6), 2002e2008. Sogin, M.L., Morrison, H.G., Huber, J.A., Welch, D.M., Huse, S.M., Neal, P.R., Arrieta, J.M., Herndl, G.J., 2006. Microbial diversity in the deep sea and the underexplored ‘rare biosphere’. Proceedings of the National Academy of Sciences of the United States of America 103 (32), 12115e12120. 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. You, S.J., Ren, N.Q., Zhao, Q.L., Wang, J.Y., Yang, F.L., 2009. Power generation and electrochemical analysis of biocathode microbial fuel cell using graphite fiber brush as cathode material. Fuel Cells 9 (5), 588e596. Wang, Q., Garrity, G.M., Tiedje, J.M., Cole, J.R., 2007. Naive Bayesian classifier for rapid assignment of rRNA sequences into the new bacterial taxonomy. Applied and Environmental Microbiology 73 (16), 5261e5267.

52

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 3 e5 2

Watanabe, K., 2008. Recent developments in microbial fuel cell technologies for sustainable bioenergy. Journal of Bioscience and Bioengineering 106 (6), 528e536. Xing, D., Zuo, Y., Cheng, S., Regan, J.M., Logan, B.E., 2008. Electricity generation by Rhodopseudomonas palustris DX-1. Environmental Science & Technology 42 (11), 4146e4151.

Zhao, F., Harnisch, F., Schroder, 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 (12), 1405e1410.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 e6 2

Available online at www.sciencedirect.com

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

Polyaluminum chloride with high Al30 content as removal agent for arsenic-contaminated well water Jasmin Mertens a,b,*, Barbara Casentini c, Armand Masion d, Rosemarie Po¨thig e, Bernhard Wehrli a,b, Gerhard Furrer a a

Institute of Biogeochemistry and Pollutant Dynamics (IBP), ETH Zu¨rich, Universita¨tsstrasse 16, 8092 Zu¨rich, Switzerland Eawag, Swiss Federal Institute of Aquatic Science and Technology, U¨berlandstrasse 133, 8600 Du¨bendorf, Switzerland c Water Research Institute (IRSA), National Research Council (CNR), Area della Ricerca di Roma 1, Via Salaria km 29,300, 00015 Montelibretti, Rome, Italy d CEREGE UMR 6635 CNRS e Aix Marseille University, Europoˆle de l’Arbois e BP80, 13545 Aix-en-Provence, Cedex 4, France e Leibniz-Institute of Freshwater Ecology and Inland Fisheries, Mu¨ggelseedamm 310, 12587 Berlin, Germany b

article info

abstract

Article history:

Polyaluminum chloride (PACl) is a well-established coagulant in water treatment with high

Received 5 June 2011

removal efficiency for arsenic. A high content of Al30 nanoclusters in PACl improves the

Received in revised form

removal efficiency over broader dosage and pH range. In this study we tested PACl with

11 October 2011

75% Al30 nanoclusters (PAClAl30) for the treatment of arsenic-contaminated well water by

Accepted 15 October 2011

laboratory batch experiments and field application in the geothermal area of Chalkidiki,

Available online 25 October 2011

Greece, and in the Pannonian Basin, Romania. The treatment efficiency was studied as a function of dosage and the nanoclusters’ protonation degree. Acidebase titration

Keywords:

revealed increasing deprotonation of PAClAl30 from pH 4.7 to the point of zero charge at pH

Al nanocluster

6.7. The most efficient removal of As(III) and As(V) coincided with optimal aggregation of

Al13

the Al nanoclusters at pH 7e8, a common pH range for groundwater. The application of

Coagulationeflocculatione

PAClAl30 with an Altot concentration of 1e5 mM in laboratory batch experiments success-

sedimentation

fully lowered dissolved As(V) concentrations from 20 to 230 mg/L to less than 5 mg/L. Field

Water treatment

tests confirmed laboratory results, and showed that the WHO threshold value of 10 mg/L

Aggregation

was only slightly exceeded (10.8 mg/L) at initial concentrations as high as 2300 mg/L As(V). However, As(III) removal was less efficient (<40%), therefore oxidation will be crucial before coagulation with PAClAl30. The presence of silica in the well water improved As(III) removal by typically 10%. This study revealed that the Al30 nanoclusters are most efficient for the removal of As(V) from water resources at near-neutral pH. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Arsenic-contaminated groundwater is a major health threat to millions of people worldwide. Specifically in SE Asia arsenic poisoning via drinking water was identified as cause for skin lesions and cancer (Chen et al., 1999; Berg et al., 2001). In

Europe, severe geogenic arsenic contamination of groundwater occurs in sedimentary basins due to reductive dissolution of iron (hydr)oxides, e.g. in the Pannonian Basin spanning the countries Hungary, Romania, Slovakia, and Serbia (Rowland et al., 2011), in the Duero Basin in Spain (Go´mez et al., 2006), in geothermal areas in Ischia and Central Italy

* Corresponding author. Institute of Biogeochemistry and Pollutant Dynamics (IBP), ETH Zu¨rich, Universita¨tsstrasse 16, 8092 Zu¨rich, Switzerland. Tel.: þ41 44 632 5489; fax: þ41 44 633 1193. E-mail address: [email protected] (J. Mertens). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.031

54

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 e6 2

(Daniele, 2004; Aiuppa et al., 2003; Angelone et al., 2009) and in Northern Greece (Kouras et al., 2007). In the Pannonian Basin approximately one million people are exposed to arseniccontaminated water by artesian village wells that serve as source for drinking, household and irrigation water (Gurzau and Gurzau, 2001; Varsa´nyi and Kova´cs, 2006; Lindberg et al., 2006). In the Chalkidiki geothermal area (Northern Greece), arsenic-contaminated groundwater (up to 3000 mg/L) is largely used in agriculture as the sole water resource for irrigation (Casentini et al., 2011; Kouras et al., 2007). The European Directive 98/83/EC lowered the limit of arsenic in drinking water to 10 mg/L following WHO and EU guidelines (WHO, 1998; EC, 1998), and efforts are needed to improve the efficiency of arsenic treatment methods. Coagulation and coprecipitation has been identified as one of the most effective and cheapest treatment technologies (EPA, 2002; Mondal et al., 2006). Polyaluminum chloride (PACl) is widely used as coagulant in water and wastewater treatment due to its high efficiency at low dosage, low cost and convenient application (Duan and Gregory, 2003; Bratby, 2006). The removal of soluble arsenic from solution by coagulation and coprecipitation with PACl includes four main steps: (1) coagulation of aluminum clusters as a result of deprotonation, (2) adsorption of arsenic to the coagulant by outer-sphere or inner-sphere complexation (3) flocculation, i.e. the formation of flocs from coagulates, and (4) precipitation of the flocs resulting in the formation of amorphous solids, including As removal by coprecipitation. The process of precipitation as separation of the aqueous and solid phase is time dependent and can be enhanced by centrifugation and filtration. Compared with conventional aluminum coagulants such as AlCl3 and alum, polyaluminum chlorides have the advantage of coagulating more effectively over a wider pH range with smaller pH change of the solution (Duan and Gregory, 2003). A comparative study showed that removal efficiencies for As(V) with PACl are higher than with polyaluminum sulfate, aluminum chloride or aluminum sulfate, and reached 99% at pH 5.5 at a dosage of 70 mM Altot (Fan et al., 2003). The application of PACl to groundwater in the Pannonian Basin (Serbia) with high organic matter content reduced initial arsenic concentrations of 50 mg/L by 50%, and  et al., 2010). in combination with FeCl3 by 84% (Tubic PACl generally consists out of Al monomers, Al oligomers and Al13 (AlO4Al12(OH)24(H2O)7þ 12 ). The latter is composed of four AlO6 trimers that surround a central AlO4 tetrahedron (Jolivet et al., 2003). Recently, industrial coagulants have been produced containing also the Al nanocluster Al30 (Al2O8Al28(OH)56(H2O)18þ 26 ) immediately after dissolution in water. The larger Al30 was identified to be composed of two Al13 clusters connected to each other by four monomeric aluminum, being 1 nm in diameter and 2 nm length (Allouche et al., 2000; Rowsell and Nazar, 2000). The titration of dissolved Al(III) salts with bases is wellknown to produce Al13 (Equation (1)).

13Al3þ þ 32OH / Al7þ 13

(1)

The formation of Al13 also occurs naturally in streams affected by acidic soils and acidic effluents of mines (Furrer

et al., 2002). Long-term aging (Casey et al., 2001) or heating (127  C for 5 h, or 95  C for 48 h) of concentrated Al13 solutions leads to the conversion of Al13 to Al30 via two pathways, which can contribute in parallel to the overall reaction (Equations (2) and (3)).

18þ 234Al7þ 13 / 91Al30 þ 312Al(OH)3 (s)

(2)

18þ þ 30Al7þ 13 þ 24H / 13Al30

(3)

The dissolution of Locron (aluminochlorohydrate with the general formula Al2(OH)5Cl) in water results in a solution containing a mixture of the nanoclusters Al30 and Al13, as well as non-characterized oligomers. Aging of such a solution shifts the composition of the mixture toward a higher contribution of Al30. The high surface area of Al nanoclusters has the potential to enhance adsorption efficiency. The specific surface area of Al13 was estimated by Bottero and Bersillon (1988) to be between 540 and 1100 m2/g, depending on the pH and the [OH]/[Al] ratio. The presence of Al nanoclusters influences the coagulation and flocculation processes in the polyaluminum solution due to differences in protonation behavior and surface charge. In comparison to Al13 and AlCl3, Al30 developed the strongest floc formation, and showed the highest turbidity removal over wider dosage and pH range than Al13 (Chen et al., 2006). Deprotonation of Al nanoclusters allows for adsorption of As by exchange with Hþ at H2O and OH groups, which are abundant in the structure of Al30 and Al13. In the present study PACl with high Al30 content (PAClAl30) is applied to remove As(III) and As(V) from natural well water. The distribution and transformation of Al species in a polyaluminum solution and precipitated material, and their stability over time is addressed. We show how acidebase properties and concentration of polynuclear Al affect the pH and the aggregation of aluminum, and consequently the removal of arsenic. Finally, the removal efficiencies from artificial well water are compared to field tests in concentration ranges of 70e2300 mg/L in the Pannonian Basin and in the Chalkidiki geothermal area.

2.

Materials and methods

2.1.

Chemicals

High purity (>18 MU) water (Millipore water purification system) was used for the preparation of all stock solutions and synthetic groundwater. All chemicals were reagent grade from Fluka, Merck or Sigma-Aldrich and were used as received. Polyaluminum stock solutions were prepared by dissolving Locron-S powder Al(OH)2.5Cl0.5 (24.4e25.4% m/m Al, Clariant) in 1 L water. Synthetic water with a similar chemical composition than Pannonian Basin groundwater (Table 1) was prepared according to Roberts et al. (2004) in a 5 L volumetric flask using 1 mM CaCO3, 8 mM NaHCO3 and 0.6 mM MgCl2. The pH was adjusted to 7.6  0.4 by addition of CO2 after Roberts et al. (2004). Silicate was added under rapid

55

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 e6 2

Table 1 e Composition of synthetic and natural groundwater. Synthetic groundwater

Hungarian/Romanian groundwatera

8.0  0.4 8.0

8.2  0.2 9.2  5.27

pH Alkalinity (mmol/L)

Chloride Sodium Calcium Magnesium Silica as Na2SiO3 As(tot) As(III)

mg/L 21.3 184.0 40.0 15.0 9.5

mmol/L 0.6 8.0 1.0 0.6 0.3

Average (mg/L) 21.0 198.0 16.0 4.6 9.5

Range (mg/L) 2e159 92e299 4.2e33.7 0.7e13.8 7.3e14.7

mg/L 20e260b

mmol/L 0.3e3.5b

mg/L

mmol/L 23e210 20e202

123 74.8

a Data from Rowland et al. (2011), group 1 general groundwater. b Arsenic was added either as NaAsO2 or as Na2HAsO4.

mixing from an alkaline stock solution prepared daily by dissolving 483 mg Na2SiO3 $ 9H2O in 10 mL water, yielding to 0.34 mM Si in 5 L.

2.2.

Analytical methods

Water samples were stored at 4  C and were measured for anions and cations within four weeks of collection. As(tot), Altot, Si, Fe, K, Na, Ca, Mg, Mn, Sr and U were determined on inductively coupled plasmaemass spectrometry (ICPeMS, Agilent 7500cx). Standards were prepared by dilution of single element standards (Merck), and the detection limit was 0.5 mg/L. As(III) was determined by hydride generation atomic fluorescence spectrometer (HG AFS, PS Analytical Ltd, Kent) with a detection limit of 0.7 mg/L. Total organic carbon (TOC) was measured with a TOC 5000A analyzer (Shimadzu, detection limit 0.5 mg C/L). Chloride was measured with a detection limit of 0.5 mg/L by ion-chromatography (Metrohm 761 Compact IC) and alkalinity has been determined by Gran titration.

2.3.

Ferron method

The Ferron method has been established by Smith and Hem (1972) to differentiate between mononuclear and polynuclear Al species. It is based on the photometrical determination of the time-dependent complex formation of Al species in acetate-buffered solution with Ferron (8-Hydroxy-7iodchinoline-5-sulfonic acid) and has been modified and improved by several research groups. The here employed method is based on the studies by Scho¨nherr et al. (1983, 1987) and Bertram et al. (1994). The buffer was prepared by dissolving 24.2 g NaCl and 26 g concentrated acetic acid in 750 mL distilled water, and adding 100 mL 2 M NaOH. The solution was titrated to pH 5  0.05 with NaOH before it was filled up to 1 L. The Ferron solution was obtained by dissolving 0.5 g Ferron (Sigma-Aldrich) in 250 mL distilled water at 50  C under rapid stirring. Calibration solutions with 10e80 mg Al/mL were prepared by the addition of 0.5e4 mg Al stock solution (1 g/L) to 50 mL water. Sample solutions with a concentration of 15 mmol/L Altot were obtained by dissolving 43.8 mg LocronS in 15 mL distilled water. After 1, 3, 24, 48, and 120 h aging,

samples were diluted to the concentration range of the calibration solutions and measured immediately. A volume of 1 mL of the Al samples was added to 9 mL buffer solution and 3 mL Ferron solution, resulting in a Ferron/Al ratio > 15. The measurements were carried out using a UVeVis-scanning spectrophotometer (UV-2401PC) from Shimadzu. The decomposition of Al polymers by the Ferron reagent was measured in 1 cm cuvettes at l ¼ 368 nm within a time span of 40e3600 s. The kinetic curves were analyzed according to the method of Scho¨nherr et al. (1983, 1987) by computer-aided calculation. The absorbance data of the time curves were converted into the logarithm of non-reacted Al. The concentrations of monomeric Al and polymeric Al were determined from the intersections of the tangents at t ¼ 0 of the initial and the final periods of the converted time curves and described in this text as Ala and Alpoly, respectively. The concentration of Alb was calculated from the difference between Alpoly and monomeric Ala. A comparison with 27Al NMR data showed that Alb measured by the Ferron method is equitable to Al13 (Scho¨nherr et al., 1983; Parker and Bertsch, 1992a). Alpoly described by Scho¨nherr et al. (1983, 1987) and Bertram et al. (1994) correspond to the Al30 polycations first described by Allouche et al. (2000) and Rowsell and Nazar (2000). The fractions of Al species in a solution determined via Ferron kinetic or 27Al NMR have been assessed to be equivalent (Bertram et al., 1996; Parker and Bertsch, 1992b; Changui et al., 1990).

2.4.

27

Al MAS NMR

A PAClAl30 solution with 15 mM Altot and 0.1 M KCl was prepared by dissolving 0.73 g Locron-S and 1.8 g KCl in 250 mL water at room temperature. Samples were obtained by removing 50 mL from PAClAl30 solution after 1, 3, 6, and 24 h. Subsequently, sample pH was increased to 7 with 0.01 M KOH, and the aggregated flocs were separated from solution by centrifugation at 27.7  103 G and freeze-dried for three days. Precipitated Al flocs were analyzed by solid-state 27Al multiquantum magic-angle spinning nuclear magnetic resonance (27Al MQMAS NMR). All 27Al NMR spectra were obtained on a Bruker Avance 400 WB spectrometer operated at 104.3 MHz with a spinning frequency of 12 kHz. For every sample a pulse

56

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 e6 2

of p/3 and a recycle delay of 1 s was used. The free-inductiondecays (FIDs) were Fourier transformed with a line broadening of 30 Hz. Line fitting was performed with Gaussian peaks using the Igor Pro software package. The species at around 7 ppm was assigned to octahedral aluminum (Aloct). The Al concentrations for 61 ppm and at 66e67.6 ppm signals were multiplied by 13 and 15, respectively, to obtain the concentrations for Al13 and Al30. The Al species represented by the peak at 35 ppm is referred to as Ali in this text.

2.5.

Titration experiments

Polyaluminum solutions of 100 mL with total aluminum concentrations of 15 mM and an ionic strength of 0.1 M KCl were titrated with 0.1 M KOH in a 150 mL water jacket vessel at 25  C under continuous magnetic stirring and argon atmosphere. All acidebase titration experiments were carried out by means of a computer-controlled titration device (736 GP Titrino, Metrohm). Titrations were started after 30 min aging and 10 min pH stability control. Aggregation of Al nanoparticles in the titrated PAClAl30 solution with [Altot] of 15 mM was observed using a laser beam (632.8 nm). Surface-specific charge (s) and protonation state (Z ) of the PACl solution were calculated according to Casey et al. (2005) from titration data under the assumption that most Al is present as Al30 with positive charges of þ18 at fully protonated state: s¼

ðZ þ Smax Þ Alsurf

  vB Kw þ Hþ  þ ðv0 þ vÞ ½H  with Z ¼ ½Al30 tot

(4)

(5)

Z: Protonation number; Smax: total charge at fully protonated state; Alsurf: total number of surface Al atoms; v0: initial volume of Al30 in solution (mL); v: volume of the added base (mL); B: concentration of the added base (M); ½Al30 tot : total Al30 concentration as calculated for each titration step (mol/L). For arsenic removal experiments, 1 mL of the PACl solution was replaced by 1 mL of neutral arsenic stock solution (5 mM As) once the desired pH was held constant for 10 min, resulting in an arsenic concentration of 50 mM in the vessel. For kinetic titration experiments, the pH was kept constant for up to four hours and triplicates were taken after 5, 15, 30, 60, 120, 180 and 240 min. Samples were centrifuged at 3.22  103 G and filtered through 0.45 mm filters (Infochroma AG). Above pH 6, filters with 0.2 mm led to instant clogging. In order to use the same separation technique for all experiments it was decided to use 0.45 mm filters.

2.6.

Coprecipitation experiments

Coprecipitation experiments were conducted in 330 mL polyethylene terephthalate bottles with 97e98.9 mL aliquots of synthetic groundwater. A volume of 1 mL of 27, 133, and 347 mmol/L arsenite or arsenate solution was added to yield arsenic concentrations of 0.27, 1.33 and 3.47 mmol/L. The coprecipitation process was initiated by adding PAClAl30 solution from Locron-S stock solutions (5.1 mM, 51.2 mM, and 512.2 mM Altot) to yield Altot concentrations of

0.01e10.3 mM. The bottles were placed in a reclined position on a shaker with horizontal movement at 250 rpm for 30 min at room temperature, and the pH was allowed to equilibrate. After 4 h of settling, the pH was measured and the supernatant solution was filtered through 0.45 mm nylon filters and the filtrate was analyzed for As concentrations and cations. To investigate the arsenic density on Al nanocluster, experiments were carried out in 50 mL sterile polyropylene tubes with 42.5 mL synthetic water at pH 7.5  0.1. To each batch, 5 mL of As(III) or As(V) stock solutions (adjusted to pH 7.6) were added to achieve initial arsenic concentrations of 0.3e66.7 mmol/L. Finally, 2.5 mL of a Locron stock solution (5.1 mM Altot) was added to yield total Al concentrations of 255 mmol/L.

2.7.

Field tests

Arsenic removal has been tested on natural groundwater from reducing Pannonian aquifers in three artesian wells from the towns Avram Iancu, Ciumeghiu and Sepreus (R112, R113, and R164, notation after Rowland et al., 2011) in the counties of Bihor and Arad in SW-Romania ca. 100 km south of Oradea, and from three pumped wells in the Nea Triglia geothermal area of Chalkidiki prefecture, Northern Greece, identified as KL59, KL103, and Pilot. In Romania, artesian wells have been selected for high arsenic concentrations. 330 mL PET bottles were filled with 100 mL groundwater, and the PAClAl30 solution was added in two different concentrations to yield an As:Altot molar ratio of 0.15 and 0.015. The bottle was vigorously shaken for 5 min and left for settling in an upright position over night before the supernatant solution was filtered (0.45 mm). The samples were taken before and after treatment, immediately acidified by adding 2% 1 M HCl, and stored in the dark at 4  C until analysis by HG AFS. In the geothermal area of Nea Triglia, Greece, water was collected from each well in 4 L containers, which were rinsed three times with distilled water before use and filled up to 3.2 L. Locron-S was added in two different concentrations: 0.22 g/L ([Altot] ¼ 1.1 mM) and 0.62 g/L ([Altot] ¼ 3.2 mM). The waters were vigorously shaken three times and left to react for four hours. Samples were collected before and after treatment by filtering through 0.45 mm filters. For ICPeMS analysis 20 mL of sample was acidified with 1% HNO3 and stored at 4  C. For As(III) and As(tot) analysis 15 mL of water was acidified with 250 ml 1 M HNO3 and stored at 4  C in a brown PET bottle until analysis with HG AFS and ICPeMS.

3.

Results and discussion

3.1.

Characterization of PAClAl30

3.1.1.

Acidebase properties of PAClAl30

The chemical behavior of PACl in solution is crucial to understand the removal processes and the best conditions of coagulant application. The initial pH of the PAClAl30 solution is very similar to that of an Al30 solution with an Al30 concentration of 25 mM. Increasing the pH from 4.7 to 6.7 leads to a loss of 18 protons (Z ¼ 18, Fig. 1a). Aggregation of Al nanoparticles in a PAClAl30 solution with [Altot] of 15 mM was

57

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 e6 2

stable afterward (Fig. 2). The initial content of monomeric Al (Almono, Fig. 2) decreased within the entire time of the experiment (five days) from 7.8% to below 1%. However, Al13 increased its share from 13.7% to 25.4% (Fig. 2), partially by polymerization of monomeric Al, and partially from structural transformation of larger aggregates.

3.1.3.

Al floc analysis with

27

Al NMR

27

Fig. 1 e Titration curves of a 1 h altered PAClAl30 solution with total aluminum concentrations of 5.1 mM (corresponding to 392 mM Al13 or 170 mM Al30) in comparison with Al13 (data from Furrer et al., 1992) and Al30 (data from Casey et al., 2005). a) Z Value. The dashed lines indicate the point of complete deprotonation of Al13 and Al30, b) Surface-specific charge. The titration until pH 8 was carried out in 63 min. The pHPZC of the PAClAl30 solution is at 6.7 and corresponds well with the pHPZC of the 25 mM Al30 solution.

observed with a laser (l ¼ 632.8 nm) at pH  5.5 (Z-Value ¼ 4). The pH of the point of zero charge (pHPZC) of the PAClAl30 solution at pH 6.7 correlates with the pHPZC of a solution with lower Al13 and Al30 concentrations, and the obtained values are similar at a surface-specific charge smaller than 0.1 (Fig. 1b). The deprotonation of the main compound of the used PACl, Al30, is known to occur in a broad pH window and takes place at terminal water ligands (Casey et al., 2005). The presented results indicate that the PAClAl30 solution deprotonates over the same large pH window than Al30 (4.7e6.7). Strongest flocculation of the aluminum was found above pHPZC 6.7, where the electrostatic repulsion was smallest. The processes of aluminum coagulation, floc formation and the consequent efficiency of precipitation and liquidesolid separation are enhanced with increasing pH.

3.1.2.

Kinetic analysis with Ferron

Al13 and Al30 solutions are known to change their Al composition with time (Casey et al., 2001), and the formation of monomeric Al could lead to a reduction of As binding sites. The kinetic analysis of Al species in PAClAl30 solutions with the Ferron method showed that the content of Al30 slightly decreased from 78% to 74% within the first 48 h but remained

Solid-state Al NMR analysis of Al flocs from a PAClAl30 solution with 15 mM [Altot] showed one broad peak representing octahedral aluminum (Aloct) centered at ca. 7 ppm (Fig. A.1, supplemental material). Two contributions were distinguished for an asymmetrical peak representing Al(O)4: one at 61 ppm (Al13 ε-Keggin) and another at 66e67.6 ppm. The signal at 66e67.6 ppm falls between the reported chemical shifts for Al30 (68e72 ppm, Allouche et al., 2000) and the Al13 dKeggin (65.5 ppm). Comparison with precipitates of pure Al30 solutions showed the same chemical shift for Al30. It therefore corresponded to an Al30 polymer that was slightly deformed by aggregation. Al13 remained between 15% and 23% without notable trend (Fig. 2). The average of the Al30 share over 24 h was 48  3%, while octahedral aluminum not linked to Al13 and Al30 species varied from 27% to 40%. A small peak at 35 ppm representing 0.4e0.8% of total Al, referred to in this text as Ali (Fig. 2), may be attributed to an intermediate octahedral Al species that was formed during the transformation of Al30. Compared with the Al species distribution in the aqueous solution, about 20e30% of Al30 was transformed into

Fig. 2 e The share of Al species from [Altot] [ 15 mM in % in a freshly prepared aqueous PAClAl30 solution (empty symbols) and in formed Al flocs (filled symbols). Al speciation of the PAClAl30 solution was determined by the Ferron method. The indicated mole fractions Al30 (squares), Al13 (upward triangles) and Almono (empty circles) for the PACl solution are derived from the measured concentrations for Alpoly, Alb, and Ala after Scho¨nherr et al. (1983, 1987). Al flocs from a PAClAl30 solution with 0.1 M KCl were analyzed by solid-state 27Al MAS NMR. Aloct (filled circles) corresponds to the octahedral aluminum species not present in the structure of Al13 or Al30, Ali (downward triangles) is an intermediate, octahedral aluminum species.

58

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 e6 2

octahedrally coordinated Al clusters during aggregation. However, Al13 contents in the PAClAl30 solution were similar to Al13 contents in the precipitated Al flocs. Therefore, aggregation processes did not affect the Al13 content (Fig. 2). In total, two thirds of the total aluminum in the aggregates were represented by the Al nanoclusters Al13 and Al30.

3.2.

Arsenic removal with PAClAl30

The removal of arsenic from aqueous solutions with PAClAl30 includes two processes, which can occur simultaneously or subsequently: (i) the chemical interaction of As species with 2 the polynuclear Al. The As(V) species H2AsO 4 and HAsO4 and the As(III) species H3AsO3 form dissolved AleAs complexes already in the acidic pH range, where no aggregation and precipitation of polynuclear Al can be observed. (ii) The deprotonation of polynuclear Al complexes is dependent on pH: Al nanoclusters fully deprotonate and aggregate strongly at near-neutral or higher pH values. The actual removal of arsenic from solution results as a consequence of aggregation and precipitation of the nanoclusters that might include relevant amounts of bound arsenic. Since the As(V) species have a higher binding affinity for aluminum atoms than the As(III) species (Manning and Goldberg, 1997) the former are removed more efficiently than the latter.

3.2.1.

The effect of pH

The pH value is a key parameter controlling the deprotonation state and the aggregation of polynuclear Al species in PACl. Strong aggregation of Al clusters at pH > pHPZC leads to a pronounced removal of total aluminum from solution (Fig. 3). As(V) was completely removed at pH 7 and 8, whereas As(III) removal reached its maximum of 80% at pH 8 (Fig. 3). Although optimum removal pH of both species was achieved at pH 7e8, the interaction of As(III) and As(V) with Al nanoclusters is different. As(V) was removed in the same proportion than Al over the entire pH range within the margins of error (Fig. 3). Hence, As(V) is uniformly distributed over all Al surface binding sites, including those of dissolved Al species at pH 5, 6 and 6.5 (Fig. 3), and As(V) forms soluble complexes with Al nanoclusters. Hence, a covalent binding by ligand-exchange reactions is taking place independently of the charge of Al clusters. The simultaneous elimination of As(V) with Al was consistent with the precipitation of the As(V)-Al complexes. As(III) removal at all pH values was significantly lower than As(V) removal. This indicates a weaker affinity between As(III) and Al nanoclusters. As shown in Fig. 3, strong flocculation leads to enhanced As(III) removal at pH 7 and 8. This shows that Al aggregation is the most important process for the removal of both arsenic species with PAClAl30. Hence, the pH needs to be kept at the optimum for the removal of Al nanoclusters. Kinetic investigation over 4 h reaction time showed no strong variation of arsenic removal (Fig. 3). Therefore, the reaction of As with Al nanoclusters was completed after at most 5 min. Speciation measurements showed no oxidation of As(III). At optimal pH, 99% of Al was removed by precipitation of Al flocs and filtration. At the total added Al concentration of 405 mg/L this means that 4  0.6 mg/L remained in solution.

Fig. 3 e Removal of As(III), As(V) and Al in dependence of pH. As(V) was added as Na2HAsO4 and As(III) was added as NaAsO2 in the concentration of 45 mM with samples taken in triplicates after target pH was reached. Error bars indicate the standard deviation of measurements over 5, 15, 30, 60, 120, 180, and 240 h. Locron-S solutions with [Altot] [ 15 mM were altered for 30 min and had an ionic strength of 0.1 M KCl. Equilibration pH at t [ 0 was 4.8. For As(V), the five titrations with target pH 5, 6, 6.5, 7, and 8 required 0.53, 4.56, 8.67, 8.09, and 9.47 mL 0.1 M KOH and reached the target pH values after 3.9, 29.7, 55.9, 18, and 15.7 min, respectively. The five As(III) titrations with target pH 5, 6, 6.5, 7, and 8 required 0.39, 3.82, 8.09, 9.47, and 10.8 mL 0.1 M KOH and reached the target pH values after 2.7, 13.8, 52, 18, and 15 min. As(V) and As(III) removal is increasing with higher pH and increasing aggregation of the PACl. Arsenate is removed by 100% above pHPZC of PAClAl30 at 6.7 (0.22 As/coagulated Al), maximum arsenite removal of 80% is achieved at pH 8.

Hence, the maximum contaminant level for aluminum in drinking water of 0.2 mg/L suggested as indicator parameter by the EC Drinking Water Directive 98/83/EC was exceeded by a factor of 20. For practical application at small-scale the Al removal process therefore needs to be optimized by using smaller filters or centrifugation.

3.2.2.

The effect of coagulant dosage in water applications

Al30 is more acidic than Al13 due to the presence of acidic hH2O functional groups (Rustad, 2005). To understand how the application of PACl with high Al30 content affects the water chemistry and the arsenic removal efficiency in natural well water, we used simulated groundwater with the chemical composition of the Pannonian Basin (Table 1) at initial pH of 8  0.4. Independent from the initial arsenic concentration, 98.6e99.4% As(V) were removed with total aluminum concentration of 1e6 mM (Fig. 4b). Aluminum concentrations below 1 mM showed a scattered Al removal and consequently a scattered As removal. Above 6 mmol Al/L the pH dropped to 6.5 or lower. Titration experiments revealed that low pH reduced aggregation and precipitation of aluminum clusters due to charge repulsion (Fig. 3), and this was confirmed by Al removal data at high initial Al concentrations with low and

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 e6 2

Fig. 4 e Effect of increasing aluminum concentrations on a) equilibrium pH, b) arsenate removal with and without Si as 0.34 mM Na2SiO3, the insert zooms to low Al concentrations, and c) silica removal in synthetic water (composition see Table 1). [As(V)]initial [ 0.3 mmol/L, 1.3 mmol/L and 3.47 mmol/L. The pH decreases with increasing aluminum concentration from pH 8 at low Al concentrations to pH 6.8 at high Al concentrations. The dotted line in a) indicates the pHPZC of PAClAl30, above which Al aggregation and removal is at maximum.

high pH for reference (Fig. 4c): highest Al removal was achieved when the pH is above pHPZC, and under these conditions As(V) is removed by >90%. As(III) was removed by 35% at most (Fig. 5b), but it improved in general with increasing Locron-S addition due to the increase in available surface, as long as the pH is not too low to prevent the formation of big aggregates. As(III) removal in titration experiments is about twice as efficient as in batches with synthetic water where pH was at equilibrium (Fig. 5a). PAClAl30 solutions with high Al concentration were more acidic but also showed more coagulation and aggregation when brought to pH 7e8 than solutions with smaller Al

59

Fig. 5 e Effect of increasing aluminum concentrations on a) equilibrium pH, b) arsenite removal with and without Si as 0.34 mM Na2SiO3 and c) silica removal in synthetic water (composition see Table 1). [As(III)]initial [ 0.3 mmol/L, 1.3 mmol/L and 3.47 mmol/L. The pH decreased with increasing aluminum concentration from pH 8 at low Al concentrations to pH 6.8 at high Al concentrations. The two lines in b) show the average removal with and without Si. As(III) removal is enhanced by the presence of Si.

concentrations. Therefore, As(III) can be removed by up to 80% with high Al concentration at optimum pH conditions (Fig. 3). The adsorption of As(V) on Al was investigated by increasing initial arsenic concentrations to 66.7 mmol/L in a solution of 250 mM Altot in water of the same composition than used in batch experiments. The maximum adsorption density for As(V) was determined as 0.2 M/M Al. This is a 60% increase compared to alum coagulation (0.12 M As(V)/Al) reported by Edwards (1994). The As(III) adsorption density is lower by a factor of 100 (Fig. 6).

60

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 e6 2

smaller filter pores is a better alternative. Water treatment facilities in the US using Al2(SO4)3 salts reported similar remaining aluminum concentrations ranging from 0.01 to 2.7 mg/L, with an overall average of 0.16 mg/L after treatment (WHO, 1998).

3.3.

Fig. 6 e Removal of As(III) and As(V) from synthetic water (composition see Table 1) by PAClAl30 (250 mmol/L Altot) with increasing addition of arsenic concentrations (0.3e66.7 mmol/L) at pH 7.5 ± 0.1.

3.2.3.

The effect of silica

Activated silica is used as a coagulation aid for aluminum and ferric-based treatment. It forms negatively charged colloids that increase the size of aluminum hydroxides during coagulation. Water of the Pannonian Basin already contained 9.5 mg Si/L in average (Table 1). We conducted batch experiments with and without adding silica to synthetic water and found initial silica concentrations were reduced after water treatment by up to 50% with 3 mM Altot, while at the lowest and highest PAClAl30 addition no Si was removed (Figs. 4c and 5c). Hence, silica removal depended on the effectiveness of aluminum coagulation and separation. Fan et al. (2003) found an increasing As(V) removal by polyaluminum sulfate and aluminum sulfate with Si dosages of 4e10 mg Si/L. We have no evidence for As(V) removal being effected by the presence of Si, because As(V) was already removed by 98e100% without silica. However, As(III) removal ranged between 10% and 35% with silica and between 0% and 15% without silica (Fig. 5b), unless the pH was at or above 7. This suggests that the presence of Si is favorable to the removal of As(III), either by enhancing the coagulation of aluminum or by forming complexes with Al and As(III). Aluminum was removed by nearly 100% in batch experiments (Figs. 4c and 5c). However, the concentration of Al remaining in solution ranged between 0.03 and 0.8 mg/L with an average of 0.35 mg/L for waters containing no Si concentrations and 0e1.1 mg/L at Si concentrations of 9.5 mg/L with an average value that is slightly below the drinking water limit of 0.2 mg/L suggested by the WHO for drinking water (Fig. A.2, supplemental material). For high and low Al concentrations, the separation of Al from the aqueous phase is scattered and can be as low as 60% due to concentration and pH effects as discussed above (Fig. 4c). The removal of Al can be enhanced by better separation methods than used in this study, e.g. smaller filter pores and ultracentrifugation. Centrifugation is neither cost-effective nor practical for large-scale field operations. It was observed during this study that filters smaller than 0.45 mm clogged easily when filtering Al solutions above pH 6. Therefore, using several filter steps from 0.45 mm to

Field validation

In Romania, groundwater from the artesian wells R112, R113, and R164 has As concentrations in the range of 74e185 mg/L, mainly present as As(III). Samples R112 and R164 had similar arsenic concentrations (80 mg/L and 74.5 mg/L), while R113 groundwater contains 185 mg/L (Table A.1, supplemental material). The three wells generally differed in their water composition, but were within the average groundwater chemistry of the area (group 1, Rowland et al., 2011). The well R113 with the highest As concentration had the highest concentration of TOC (6.1 mg/L), and the groundwater with the lowest As concentration (R 164) had the lowest TOC concentration (0.49 mg/L). The enhanced release of arsenic due to competition with TOC on sorption sites has been proposed for Hungarian groundwater by Varsa´nyi and Kova´cs (2006). Fig. 7a shows that 33e38% of arsenic was removed from As(III)-rich water with an As-to-Al ratio of 0.14, and 12.6e24.1% were removed with an As-to-Al ratio of 0.014. In this study at three well sites, TOC concentrations of up to 6 mg/L seemed to have no effect on As(III) removal. H3AsO3 did not change its oxidation state during treatment (Table A.1, supplemental material). Arsenic removal was tested on water of the geothermal area of Chalkidiki from two open tanks KL59 and KL103 and from a freshly pumped well (Pilot) with initial arsenic concentrations of 193.2 mg/L, 2293 mg/L, and 168.8 mg/L, respectively, mainly present as As(V). The results showed that initial As(V) concentrations of 170 mg/L were reduced to below the drinking water guide value with 1 mM Altot, and are further decreased to less than 5 mg/L with 3.2 mM. High As(V) concentrations of up to 2300 mg/L can be lowered to 10.8 mg/L with the addition of 3.2 mM Altot (Fig. 7b), resulting in a molar ratio of 0.01 M As(V)/M Al. The As removal data from these field trials in two different areas confirm laboratory findings despite different water compositions. For all investigated wells, measured aluminum concentrations were below the detection limit of 0.1 mg/L after the settling of Al flocs and filtration (0.45 mm) of the supernatant solution. These data are in contrast with the results from batch experiments and a result of the reduction of boundary effects during up scaling. Initial silica content was in the range or one third above the concentrations used in lab experiments. For As(III)-rich Romanian groundwater Si was removed by ca. 50% after treatment, while in the case of As(V)rich Greek groundwater Si removal was up to 30% (Table A.1, supplemental material). Manganese concentration was reduced by 25e50%. Other metals present in the waters like copper and iron were eliminated substantially by 40% and 93e99%, respectively. Trace elements like uranium and vanadium showed a tendency to be reduced (Table A.1, supplemental material). The initial pH of the Romanian and Greek groundwater was within the optimal range for removal. During treatment with PAClAl30 the pH decreased by 0.2e1

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 e6 2

61

processes and removal potential. The protonation state of Al30 governs the formation of Al flocs, and consequently the removal of both arsenic species, to occur between pH 6.5 and 8. This is a common pH range for natural groundwater and therefore no pH adjustment has to be undertaken for precipitation to take place. Analysis of the PAClAl30 solution and the precipitate obtained from a commercial coagulant powder verified that polyaluminum clusters remain dominant during flocculation and precipitation. Highly concentrated PAClAl30 (>5 mM) solutions result in pH ranges that are not favorable for flocculation and precipitation (
Acknowledgments

Fig. 7 e As(V) and As(III) remaining in solution before and after applying PAClAl30 to a) three artesian groundwater wells R112, R113 and R164 in Western Romania. PACAl30 solution was applied to each well in the two molar Al:As ratios 70 and 700. b) Three well waters in the geothermal area of Chalkidiki, Northern Greece. Locron-S was added in two different concentrations: 1.1 mM [Altot] and 3.2 mM [Altot]. The groundwater chemical composition before and after treatment is listed in the supplemental information (Table A.1).

units (Table A.1, supplemental material). In Greece, where the water is pumped into open tanks before it is used for irrigation, the coagulation with polyaluminum chlorides in the form of Locron-S powder represents an easy-applicable and effective method to reduce As accumulation in soils due to irrigation practices.

4.

Conclusions

A reliable chemical treatment method requires knowledge of the molecular processes under natural conditions. This study provides insights for an optimal application of PACl with high Al30 content to remove arsenic from contaminated well water. Our findings give evidence that the acidebase properties of a polyaluminum coagulant strongly influence the coagulation

The authors would like to thank Thomas Ru¨ttimann (Eawag) for ICPeMS and arsenic speciation analysis, Bjo¨rn Studer and Rene´ Saladin (ETH Zurich) for laboratory support, and Fabio Ziarelli (Universite´ Aix-Marseille) for 27Al NMR measurements. Prof. Calin Baciu (UBB Cluj-Napoca) is greatly acknowledged for organizational field support. Many thanks go to Stephan Hug, Annette Johnson and Vanessa Sternitzke (Eawag) for invaluable discussions. This is a contribution of the AquaTRAIN Marie-Curie Research training Network (Contract No. MRTN-CT-2006-035420) funded under the European commission Sixth Framework Programme (2002e2006) Marie Curie Actions, Human Resources & Mobility Activity Area e Research Training Networks. It reflects the views of the authors but not necessarily those of the European Community, which is not liable for any use that may be made of the information contained therein.

Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.10.031.

references

Aiuppa, A., D’Alessandro, W., Federico, C., Palumbo, B., Valenza, M., 2003. The aquatic geochemistry of arsenic in volcanic groundwaters from southern Italy. Appl. Geochem. 18, 1283e1296. Allouche, L., Ge´rardin, C., Loiseau, T., Fe´rey, G., Taulelle, F., 2000. Al30: a giant aluminum polycation. Angew. Chem. Int. Ed. 39 (3), 511e514. Angelone, M., Cremisini, C., Piscopo, V., Proposito, M., Spaziani, F., 2009. Influence of hydrostratigraphy and structural setting on the occurrence in groundwater of the Cimino-Vico volcanic area (central Italy). Hydrogeology J. 17, 901e914. Berg, M., Tran, H.C., Nguyen, T.C., Pham, H.V., Schertenleib, R., Giger, W., 2001. Arsenic contamination of groundwater and drinking water in Vietnam: a human health threat. Environ. Sci. Technol. 35, 2621e2626. Bertram, R., Geßner, W., Mu¨ller, D., Danner, M., 1994. Characterization of Al(III) species in basic aluminium chloride flocculants by means of Ferron method and 27Al Nuclear magnetic resonance. Acta Hydrochim. Hydrobiol. 22 (6), 265e269.

62

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 e6 2

Bertram, R., Stieber, E., Geßner, W., 1996. Toxizita¨t von Aluminium e Al Spezies in protolysierten Aluminiumchloridlo¨sungen. Z. ¨ kotox. 8 (2), 78e82. Umweltchem. O Bottero, J.Y., Bersillon, J.L., 1988. Aluminum and Iron(III) chemistry e some implications for organic substance removal. Adv. Chem. 219, 425e442. Bratby, J., 2006. Coagulation and Flocculation in Water and Waste Water Treatment, second ed. IWA Publishing, 407 p. Casentini, B., Nikolaidis, N.P., Hug, S.J., 2011. Arsenic accumulation in irrigated agricultural soils in Northern Greece. Sci. Total Environ. 409 (22), 4802e4810. Casey, W.H., Philips, B.L., Furrer, G., 2001. Aqueous aluminum polynuclear complexes and nanoclusters: a review. Rev. Mineral. Geochem. 44, 167e190. Casey, W.H., Rustad, J.R., Banerjee, D., Furrer, G., 2005. Large molecules as models for small particles in aqueous geochemistry research. J. Nanoparticle Res. 7, 377e387. Changui, C., Stone, W.E.F., Vielvoye, L., Dereppe, J.M., 1990. Characterization by nuclear magnetic resonance spectroscopy, Ferron assay, and acidification of partially neutralized aluminium solutions. J. Chem. Soc., Dalton Trans. 5, 1723e1726. Chen, C.J., Hsu, L.-I., Tseng, C.-H., Hsueh, Y.-M., Chiou, H.-Y., 1999. Emerging epidemics of arseniasis in Asia. Arsenic exposure and health effects. In: Abemathy, C.O., Calderon, R.L. , Chappell, W.R. (Eds.), Arsenic Exposure and Health Effects. Chapman & HaU, London, pp. 113e121. Chen, Z., Fan, B., Peng, X., Zhang, Z., Fan, J., Luan, Z., 2006. Evaluation of Al30 polynuclear species in polyaluminum solutions as coagulant for water treatment. Chemosphere 64, 912e918. Daniele, L., 2004. Distribution of arsenic and other minor tarce elements in the groundwater of Ischia Island (southern Italy). Environ. Geol. 46, 96e103. Duan, J., Gregory, J., 2003. Coagulation by hydrolyzing metal salts. Adv. Colloid Interf. Sci. 100e102, 475e502. Manning, B., Goldberg, S., 1997. Sorption and stability of arsenic(III) at the clay mineralewater interface. Environ. Sci. Technol. 31, 2005e2011. EC, 1998. Council directive 98/83/EC. Official J. Eur. Commun. L330/32. Edwards, M., 1994. Chemistry of As removal during coagulation and FeeMn oxidation. AWWA, 64e78. EPA, 2002. Arsenic treatment technologies for soil, waste and water. EPA-542-R-02-004. Fan, M., Brown, R.C., Sung, S.W., Huang, C.P., Ong, S.K., van Leeuwen, J., 2003. Comparison of polymeric and conventional coagulants in arsenic(V) removal. Water Environ. Res. 75 (4), 308e313. Furrer, G., Phillips, B.L., Ulrich, K.-U., Po¨thig, R., Casey, W.H., 2002. The origin of aluminum flocs in polluted streams. Science 297, 2245e2247. Furrer, G., Ludwig, C., Schindler, P.W., 1992. On the chemistry of the Keggin Al13 polymer. I. Acidebase properties. J. Colloid Interf. Sci. 149 (1), 56e67. Go´mez, J.J., Lillo, J., Sahu´n, B., 2006. Naturally occurring arsenic in groundwater and identification of the geochemical

sources in the Duero Cenozoic Basin, Spain. Environ. Geol. 50, 1151e1170. Gurzau, E.S., Gurzau, A.E., 2001. Groundwater in Transylvania, Romania: an overview. Arsenic Exposure and Health Effects IV, pp. 181e184. Jolivet, J.P., Henry, M., Livage, E., 2003. Metal Oxide Chemistry and Synthesis. John Wiley, Chichester, 321 p. Kouras, A., Katsoyiannis, I., Voutsa, D., 2007. Distribution of arsenic in groundwater in the area of Chalkidiki, Northern Greece. J. Hazard. Mater. 147, 890e899. Lindberg, A.L., Goessler, W., Gurzau, E., Koppova, K., Rudnai, P., Kumar, R., Fletcher, T., Leonardi, G., Slotova, K., Gheorghiu, E., Vahter, M., 2006. Arsenic exposure in Hungary, Romania and Slovakia. J. Environ. Monitor. 8, 203e208. Mondal, P., Majumder, C.B., Mohanty, B., 2006. Laboratory based approaches for arsenic remediation from contaminated water: Recent developments. J. Hazard. Mater. B137, 464e479. Parker, D.R., Bertsch, P.M., 1992a. Identification and quantification of the “Al13” Tridecameric polycation using Ferron. Environ. Sci. Technol. 26 (5), 908e914. Parker, D.R., Bertsch, P.M., 1992b. Formation of the “Al13” tridecameric polycation under diverse synthesis conditions. Environ. Sci. Technol. 26 (5), 914e921. Roberts, L.C., Hug, S.J., Ruettimann, T., Billah, Md M., Khan, A.W., Rahman, M.T., 2004. Arsenic removal with iron(II) and iron(III) in water with high silicate and phosphate concentrations. Environ. Sci. Technol. 38, 307e315. Rowland, H.A.L., Omoregie, E., Millot, R., Jimenez, C., Mertens, J., Baciu, C., Hug, S.J., Berg, M., 2011. Geochemistry and arsenic behaviour in groundwater resources of the Pannonian Basin (Hungary and Romania). Appl. Geochem. 26 (1), 1e17. Rowsell, J., Nazar, L.F., 2000. Speciation and thermal transformation in alumina sols: structures of the polyhydroxyoxoaluminum cluster [AlO(OH)(HO)] and its v-Keggin moiete´. J. Am. Chem. Soc. 122, 3777e3778. Rustad, J.R., 2005. Molecular dynamics simulation of the titrtaion of polyoxocations in aqueous solution. Geochim. Cosmochim. Acta 69 (18), 4397e4410. Scho¨nherr, S., Go¨rz, H., Bertram, R., Mu¨ller, D., Gessner, W., 1983. Vergleichende Untersuchungen an unterschiedlich dargestellten basischen Aluminiumchloridlo¨sungen. Z. Anorg. Allg. Chem. 502, 113e122. Scho¨nherr, S., Go¨rz, H., Bertram, R., 1987. Zur Anwendung der zeitabha¨ngigen Komplexbildung mit Ferron fu¨r die Charakterisierung basische Metallkationen. Wiss Z. Pa¨d. Hochsch. Potsdam 31, 67e74. Smith, R.W., Hem, J.D., 1972. U.S. Geol. Surv. Water-Supply Pap. 1827-D. , A., Agbaba, J., Dalmacija, B., Ivancev-Tumbas, I., Tubic Dalmacija, M., 2010. Removal of arsenic and organic matter from groundwater using ferric and alum salts: a case study of central Banat region (Serbia). J. Environ. Sci. Health A 45, 363e369. Varsa´nyi, I., Kova´cs, L.O., 2006. Arsenic, iron and organic matter in sediments and groundwater in the Pannonian Basin. Appl. Geochem. 21, 949e963. WHO, 1998. Aluminium in Drinking-water. Guidelines for Drinking-water Quality, second ed. WHO, Geneva, pp. 14.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 6 3 e7 2

Available online at www.sciencedirect.com

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

Novel magnetically induced membrane vibration (MMV) for fouling control in membrane bioreactors Muhammad R. Bilad a, Gergo Mezohegyi a, Priscilla Declerck b, Ivo F.J. Vankelecom a,* a

Centre for Surface Chemistry and Catalysis, Faculty of Bioscience Engineering, Katholieke Universiteit Leuven, Kasteelpark Arenberg 23, Box 2461, 3001 Leuven, Belgium b Laboratory of Aquatic Ecology and Evolutionary Biology, Katholieke Universiteit Leuven, Ch. Deberiotstraat 32, 3000 Leuven, Belgium

article info

abstract

Article history:

Conventional submerged membrane bioreactors (MBRs) rely on the coarse bubbles aeration

Received 3 July 2011

to generate shear at the liquidemembrane interface to limit membrane fouling. Unfortu-

Received in revised form

nately, it is a very energy consuming method, still often resulting in a rapid decrease of

14 October 2011

membrane permeability and consequently in higher expenses. In this paper, the feasibility

Accepted 15 October 2011

of a novel magnetically induced membrane vibration (MMV) system was studied in a lab-

Available online 28 October 2011

scale MBR treating synthetic wastewater. The effects on membrane fouling of applied electrical power of different operation strategies, of membrane flux and of the presence of

Keywords:

multiple membranes on one vibrating engine on membrane fouling were investigated.

Magnetically induced vibration

The filtration performance was evaluated by determining the filtration resistance profiles

Membrane bioreactor

and critical flux. The results showed clear advantages of the vibrating system over

Membrane fouling

conventional MBR processes by ensuring higher fluxes at lower fouling rates. Intermittent

Shear rate

vibration was found a promising strategy for both efficient fouling control and significant energy saving. The optimised MMV system is presumed to lead to significant energy and cost reduction in up-scaled MBR operations. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Membrane bioreactors (MBRs) have been widely investigated as an advanced wastewater treatment. Their advantages over the conventional activated sludge systems have been widely cited throughout literature. However, their widespread application is still restricted, mainly due to the membrane fouling and its consequent capital and operational costs. Most traditional approaches for fouling control are based on optimizing operational conditions in favor of fouling mitigation, improving membrane properties and exploiting the hydrodynamics near the membrane surface (Le-Clech et al., 2006; Meng et al., 2009; Drews, 2010). Hydrodynamic control is implemented via the cross-flow velocity in cross-flow MBRs, and as the secondary flow of

the coarse air bubbles in submerged MBRs. Another technique is by moving the membrane itself relative to the feed, or by moving a mass near by the membrane surface. This technique is commonly known as dynamic or shear-enhanced filtration (Beier, 2008; Jaffrin, 2008). In the submerged MBRs, the coarse bubbles aeration generates direct shear on the membrane surface by inducing a secondary flow of liquid that disrupts the mass transfer boundary layer, and promotes local mixing near the membrane surface (Cui et al., 2003). Despite the rather highenergy input, this approach produces relatively weak shear rates. In addition, a “plateau” in terms of flux improvement is reached at a certain air supply (Genkin et al., 2006). Moreover, it is difficult to ensure a homogeneous bubble distribution (Genkin et al., 2006; Wu et al., 2008) and aeration at higher

* Corresponding author. Tel.: þ32 16 321594; fax: þ32 16 321998 . E-mail addresses: [email protected] (M.R. Bilad), [email protected] (I.F.J. Vankelecom). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.026

64

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 6 3 e7 2

velocities can change sludge properties and hence diminish the biomass floc stability (Rosenberger and Kraume, 2003; Drews et al., 2005). Considering the limited efficiency of the coarse bubbles aeration in submerged MBRs, the enhancement of shear rate via mechanical means seems a potential option for fouling control. A couple of studies have investigated the performance of shear-enhanced filtration systems, such as the Vibratory Shear Enhanced Processing (VSEP) and Vibrating Hollow Fiber Modules (VHFM) (Genkin et al., 2006; Beier et al., 2006; Altaee et al., 2010; Beier and Jonsson, 2007, 2009; Low et al., 2009; Kola et al., 2011). In the VHFM systems, the membrane is vibrated by a separated vibrating engine that produces axial oscillations. Membrane and engine are connected via a sliding rod. Although all the referred studies reported a significant improvement on both the critical flux (CF) and the sustainability of operation, they face numerous limitations: (1) the vibrating system is often restricted to a small range of vibration amplitudes and frequencies; (2) because of the membrane unit is separated from the vibration engine, the obtained yield of shear rates is somehow reduced, due to energy loss resulting from the mechanical contacts and their friction; (3) in most cases, filtration is run in continuous vibration mode, without the ability of changing the vibration parameters during the filtration operation, hence not able to adapt to the needs of the mixed liquor that might change over time. In the present study, a novel magnetically induced membrane vibration (MMV) system is proposed as an alternative shear enhancement device for fouling control in MBRs. As the vibrating engine is integrated into the membrane module, and as movement is magnetically induced, it is expected to experience less friction, to consume less energy and to have a very flexible vibration control. One of the main advantages of MMVs compared to the other shear-enhanced filtration systems, is their high flexibility for varying the operation modes in real-time, such as changing vibration amplitudes without interrupting the filtration to allow an online optimisation of the filtration performance. The unsteady fouling behaviour coming from the dynamic changes in feed properties can thus be controlled by real-time manipulation of the vibration parameters. To our best knowledge, no such system has been reported so far. In this study, the efficiency of the MMV system to control membrane fouling was investigated in a lab-scale MBR treating synthetic molasses wastewater. The impact of several operation parameters on fouling was studied, including vibration-related factors (e.g., vibrating power, mode and cycle), membrane flux and presence of multiple membranes arrangement. In addition, the cost efficiency of the lab-scale MMV system was furthermore estimated and discussed.

2.

Materials and methods

2.1.

Activated sludge and wastewater

The activated sludge used to inoculate the lab-scale highthroughput-MBR (HTML, Belgium; Bilad et al., 2011a) was

obtained from a pilot-scale MBR in the Waterleau wastewater laboratory (Wespelaar, Belgium) treating the same molasses wastewater, as the HT-MBR. The feed solutions were prepared by diluting 0.45 ml/l of molasses stock solution. The characteristics of the feed were given in Bilad et al. (2011b). The diluted molasses solution was chosen as feed wastewater, because it does not require pre-fine screening, it has a good COD/N ratio and contains trace elements (Yan et al., 2010). The bioreactor was operated at room temperature (22  C) in a fedbatch mode during the parametric studies and continuous mode during the long-term filtration test.

2.2.

Membrane preparation and cleaning

Two different flat sheet membranes were used during the experiments, a commercial chlorinated polyethylene membrane (KUBOTA, Japan) (PEK) and a commercial polyvinylidene fluoride membrane (Toray) (PVDFT). Both membranes were used to evaluate the effect of vibration on membrane fouling (Sections 3.2.1e3.2.2). Due to the limited amount of membrane material, PVDFT was used to investigate the vibration-related parameters (Sections 3.2.3 and 3.3) and PEK was used for the experiments with multiple membranes (Section 3.4). These membranes were potted as described by Bilad et al. (2011a). The active membrane surface area in each membrane was 0.016 m2. The SEM images of the membranes and their properties are summarized in Fig. 1 and Table 1 respectively. After each experiment, the module was removed from the bioreactor tank. First, physical cleaning was applied by flushing the membrane surface with pressurised tap water for 10 min. Afterwards, permeability of the membrane was measured. Unless the permeability loss was less than 5%, the membrane was chemically cleaned with 0.5 g/L NaOCl solution for 3 h, resulting in quasi-complete membrane permeability recovery in all cases.

2.3.

Experimental setup

A schematic diagram of the lab-scale MBR is shown in Fig. 2. The construction of the reactor system was similar to that of the HT-MBR developed earlier (Bilad et al., 2011a), except for the addition of the MMV system. The reactor had a working volume of 18.6 L and was virtually divided into aerated and non-aerated zones. In the former zone, two different aeration systems giving a total flow rate of 0.6 m3/h were provided: the fine bubbles aeration to provide soluble oxygen for the biological process, while the coarse bubbles aeration had a role in scouring the membrane surfaces as fouling control in a conventional submerged MBR system. In the non-aerated zone, a constant but weak movement of fluid was still present, induced by the air bubbles movement in the aerated zone. In the MMV system, a magnetically induced vibration of the membrane is applied in order to provide shear at the liquidemembrane interface. The module consists of one or more membranes that is integrated in the MMV module. The system includes a vibration driver, an electric wire, a vibration engine and the actual vibrating module. The signal is provided by the vibration driver being installed with a Test

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 6 3 e7 2

65

Fig. 1 e SEM images of surface and cross section of the membranes used in this study.

Tone Generator software (Esser audio, Germany). The vibration itself is created in the vibration engine by magnetic attraction/repulsion forces in a “push and pull” mode. The movement orientation of the vibrating part faces the narrow face of the module in order to both prevent the bumping of liquid onto the membrane and minimise the associated energy loss. The vibration moves the membrane to the left and the right through a sinusoidal pattern. The adjustable vibration parameters are the applied power (determined by combination of vibration amplitude and frequency), the vibration mode and the vibration cycle. The vibration amplitude and frequency can be supplied either with constant or variable values, while vibration can be run either in continuous or intermittent mode. During the filtration experiments, the vibration amplitude was limited to 2 mm at the most and the frequencies were adjusted between 0 and 60 Hz. The real vibration power (PV, W) was later calculated in a first approximation from the electric current and electrostatic potential measured on the electrical wire by two AVO meters (DVM 890-Velleman, Belgium). Heat losses at the engine were thus included in the measurements.

Table 1 e Main characteristics of the membranes used in this study. Parameter Pore size (using imageJ) (mm) Pore size (supplier data) (mm) Surface porosity (%) Thickness (mm) Morphology

PEK 0.4 0.22 11 165 Symmetric

PVDFT 0.08 0.03 0.2 320 Asymmetric

2.4. Determination of sludge characteristics and filtration parameters The mixed liquor suspended solid (MLSS), sludge volume index (SVI) and the feed and effluent quality were measured according to standard methods (APHA, 1992). The filtration performance was evaluated by determining the critical flux (CF) and filtration resistance values. The flux (J, L/m2 h) and membrane permeability (L, L/m2 h bar) were calculated by using Eqs. (1) and (2), respectively: J¼

DV ADt

(1)

L¼

J TMP

(2)

where V is the permeate volume (L), t is the filtration time (h), A is the membrane surface area (m2) and TMP the transmembrane pressure (bar, or Pa for filtration resistance calculation). The filtration resistance (RT, m1) was calculated based on Darcy’s law (Eq. (3)): RT ¼

TMP hJV

RT ¼ RM þ RF

(3)

(4)

where h is the dynamic viscosity of permeate (Pa s), JV is the flow velocity (m/s) calculated from the flux, RM is the intrinsic membrane resistance (m1), and RF the fouling resistance (m1). Two typical types of filtration were performed throughout the study. The first filtration type was the conventional aerated filtration. In this case, the membrane module was placed into the aerated zone and the filtration was performed

66

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 6 3 e7 2

as in conventional submerged MBRs. The second filtration type was the vibrated filtration in which the membrane module was placed into the non-aerated zone, and the filtration was performed as a MMV system. The CF (L/m2 h) was measured using the stepwise method (Le-Clech et al., 2003). The applied initial flux, step height and step duration were 2 L/m2 h, 2 L/m2 h and 15 min, respectively. This method was chosen because of its technical simplicity. To obtain the CF value, the final TMP values after each step were plotted against the fluxes. Below the CF, a linear relationship exists and the CF was determined as the flux at which this linear correlation ceased to exist.

3.

Results and discussion

3.1.

Bioreactor performance

The acclimatization and biological performance of the sludge in the HT-MBR was discussed earlier (Bilad et al., 2011c). To prevent the accumulation of slowly biodegradable substances during the fed-batch operation, a part of the liquid in the reactor was discharged every week. The activated sludge was settled and a part of the supernatant was withdrawn and replaced with tap water. During the test, the MLSS concentrations were kept at 10e12 g/L by partially withdrawing the sludge, while the SVI was in the range of 55e75 mL/g. During the continuous operation, the hydraulic retention time was set at 24 h, to remove more than 98% of the chemical oxygen demand. No suspended solids were detected in the permeates for neither the aerated nor the vibrated modules. Comprehensive analysis of the biological performance was not carried out during this study, since the main objective e at the current stage e was to prove the effectiveness of MMV to control fouling.

3.2.

Effect of vibration parameters on membrane fouling

3.2.1.

Membrane fouling at different filtration modes

In order to observe the impact of the vibration on the filtration performance using MMV system, the filtration of the activated sludge was performed in four different modes.  Mode-1: filtration in the non-aerated zone without vibration. In this mode, the fouling is only controlled by the limited movement of liquid induced by the aerated zone.  Mode-2: subsequent filtration just after Mode-1 without cleaning the membrane. This mode was performed to observe the impact of vibration on cleaning a fouled membrane.  Mode-3: filtration in the aerated zone (‘aerated filtration’). The coarse bubble aeration velocity was set at 0.3 m3/h.  Mode-4: filtration in the non-aerated zone (‘vibrated filtration’). For Modes-2 and -4, the membrane was continuously vibrated at a frequency of 50 Hz, corresponding to a PV of 12.5 W. The purpose of selecting a relatively high PV in this test was to observe the maximum impact of vibration. For all modes, filtrations were run at a fixed flux of 22 L/m2 h for 30 min and the filtration performance was evaluated using the resistance profiles. The virtual division of the MBR zones is schematically shown in Fig. 2. Fig. 3 shows the resistance profiles of the four different filtration modes. Mode-1 shows the highest filtration resistance followed by Mode-3, and -4, for both tested membranes. This order clearly represents the shear rates at the membrane surfaces. In the non-aerated zone, the membrane surfaces experienced very limited shear rates, only from some movement of the bulk liquid. In the aerated zone, a higher degree of shear rates was realized at the membrane surfaces due to both the liquid movement and the air bubbles scouring. In the case

Fig. 2 e Schematic diagram of the (a) HT-MBR setup equipped with the MMV system, (b) MMV module in front view, and (c) MMV module in side view, showing the parallel position on the multiple membranes mounted.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 6 3 e7 2

of the vibrated filtrations, the shear rates were high enough to develop the back transport from the membrane that finally exceeded the fouling rate, thus promoting the removal of colloids, macromolecules and other foulants from the membrane surface. As a consequence, almost no fouling was built-up in Mode-4 by both the PVDFT and PEK membranes. The impact of vibration to clean a fouled membrane is clearly seen from the results of filtration in Mode-2. A significant drop of filtration resistance was immediately obtained just after the vibration started. This result confirms the ability of the developed MMV system for in-situ cleaning of pre-fouled membranes. The impact of vibration might be different on membranes with different properties, as also shown when comparing Fig. 3a with b. This will be the subject of future research. The results confirmed the effectiveness of the MMV system in controlling membrane fouling. The vibration may control fouling via two mechanisms: (1) preventing the convective flow of foulants onto and into the membrane, as seen from the result in Mode-4 and (2) removing the foulants from the prefouled membrane, as seen from the result in Mode-2. It has to be noticed that this experiment was performed in a very short time span. Apart from the attachment of (bio)foulants, physico-chemical interactions with the membrane might also take place (Le-Clech et al., 2006). This type of fouling is generally developed at a slower rate that could not be observed in the experiment above, due to both short duration and the detection limits of the pressure gauges.

3.2.2.

Effect of vibration power

In the MMV system, the membrane movement is closely related to the amount of energy transferred to the system, which is on its turn transferred into shear. It is expected that at higher PV, the membraneeliquid interface experiences increased shear stress. A few studies with vibrated membranes suggested that this shear rate strongly depends on the vibrating parameters and is determined as a function of both the vibration amplitude and frequency (Genkin et al., 2006; Beier et al., 2006; Beier and Jonsson, 2007). A series of filtration tests was performed at different PVs, measured as the electric consumption by the vibration

67

engine, in order to investigate its effect on membrane fouling. PV was varied by adjusting the frequency or amplitude of vibration and was set in the range of 0e13.8 W. The filtrations were performed for 30 min at a fixed flux of 22 L/m2 h, and the final resistances at the end of each filtration are plotted against PV in Fig. 4. Results suggest that the PV significantly affects the filtration performance. In general, the higher the applied electric power, the lower the fouling. A significant reduction of filtration resistance was achieved at a PV of 10.7 W or higher. This PV value beyond which no further reduction of fouling could be achieved is further referred to as the critical power.

3.2.3.

Effect of vibration strategies

3.2.3.1. Continuous vs. intermittent vibration. The effect of intermittent vibration on the filtration performance was tested at a PV of 8 W and a J of 22 L/m2 h. PV was chosen just below its critical value (Section 3.2.2) in order to detect even small changes of the fouling rates among the different filtration tests. The experiments were carried out in three different modes, all in the non-aerated zone: (1) filtration without vibration, (2) filtration with vibration and (3) filtration with intermittent vibration. The intermittent vibration consists of idle and vibration phases to form a cycle. The sum of the idle and vibration times in one cycle is defined as the total cycle time (tC), and the ratio of the vibration time to the tC is defined as the vibration ratio (a). In this particular filtration test, tC and a were set at 60 min and 50%, respectively. The results of different vibration strategies are presented in Fig. 5. As expected, filtration with continuous vibration gave the best performance, while the intermittent vibration resulted in somewhat higher but still acceptable final filtration resistance. A much higher filtration resistance was observed for the filtration without vibration. For instance, the RF values of filtrations in modes (1), (2) and (3) after 300 min of operation were found to be 84%, 13% and 43% of the RT, respectively. A slow rise of filtration resistances for both vibration-assisted filtrations suggests that the vibration in MMV system can be operated in an intermittent mode, which offers an adequate fouling control and a reduced energy consumption.

Fig. 3 e Filtration resistance during filtration under different modes: (-) Mode-1, (,) Mode-2, (:) Mode-3 and (A) Mode-4, using (a) PVDFT and (b) PEK membranes.

68

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 6 3 e7 2

Fig. 4 e Effect of PV on final filtration resistance using (a) PVDFT and (b) PEK membranes.

3.2.3.2. Effect of intermittent cycle time. The performance of the MMV system can be further optimised by varying both the tC and a. In these particular experiments, a was fixed at 50% and tC was varied. The PV and J were fixed at 8 W and 22 L/ m2 h, respectively. Fig. 6 shows that tC strongly affects the filtration performance. For the longest tC (120 min), a sharp rise of fouling resistance occurred during the idle phases and the membrane permeability was only partially recovered during the vibration phases. On the other hand, filtrations at shorter tC resulted in lower filtration resistances. The RF values of filtration with tC of 120, 24 and 4 min after 300 min of operation, were found to be 125%, 75% and 75% of the RM, respectively. The lower fouling rates at shorter tC can be explained by the shorter idle phases, during which less aggressive fouling can be expected. This result confirms that the appropriate choice of tC for the MMV system is

Fig. 5 e Effect of different membrane vibration strategies: (-) filtration with no vibration, (A) with (continuous) vibration and (>) with intermittent vibration (tC: 60 min and a: 50%).

indispensable to ensure both an efficient and an economic operation.

3.3.

Operation flux and critical flux

3.3.1.

Effect of operation flux on membrane fouling

Flux is considered as one of the most important factors affecting membrane fouling in MBRs (Chang et al., 2002). Its appropriate selection is crucial to maintain the fouling rate at a satisfactory level during long-term operations. A series of filtrations was performed to investigate the effect of operational flux on filtration performance. The performances of both the vibrated filtration and the aerated filtration were compared, as shown in Fig. 7. The flux was varied between 14 and 30 L/m2 h, and the MMV system operated in an intermittent mode, with a tC of 4 min, a PV of 8 W and an a of 50%.

Fig. 6 e Effect of intermittent cycle duration: filtration with vibration cycle of (-) 120 min, (>) 24 min and (:) 4 min (a: 50%).

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 6 3 e7 2

As expected, a faster rate of fouling was found at higher fluxes for both systems. The filtration resistance at the corresponding fluxes was always found to be significantly lower for the vibrating module. At J ¼ 30 L/m2 h, a sharp rise of filtration resistance was observed for both the aerated and the vibrated module. However, the MMV system experienced a much lower fouling at the flux ranges of 14e26 L/m2 h. These results suggest that the MMV system could ensure higher operational fluxes compared to conventional submerged MBR systems.

3.3.2.

Effect of vibration on critical flux

Membrane fouling, in general, is managed by operating the system below its CF (Le-Clech et al., 2006). The CF is broadly defined as the maximum flux value, at which (theoretically) no particle deposition on the membrane surface occurs. This critical value depends on several factors such as the feed condition, membrane properties, hydrodynamics and operation conditions (Wu et al., 2008). The enhanced shear upon vibration can facilitate the increase of CF. Its correlation was evaluated by measuring CF values at different PVs (Fig. 8). The experiment was carried out at a ¼ 100% and at a constant vibration frequency of 50 Hz. The PV was adjusted by varying the vibration amplitude. Fig. 8 indicates that the higher the PV, the higher the CF, in accordance with the literature (Genkin et al., 2006; Beier et al., 2006; Altaee et al., 2010). For example, a CF value of 46 L/m2 h was obtained at a PV of 15.4 W, which is about 3 times higher than the CF measured for the similar feed and membrane in a conventional aerated lab-scale MBR (Bilad et al., 2011a). The PV and CF were found to be proportional in the range of the studied PV. The intercept with the y-axis in Fig. 8 gives the CF for the non-vibrating operation. Since the MMV system is operated at a fixed frequency, PV is proportional here to the applied vibration amplitude, and that is in line with earlier findings for the VHFM system (Beier et al., 2006).

3.4. Long-term filtration, multiple membranes operation and energy consumption The applicability of the MMV system to control membrane fouling in short-term filtration duration has been proven in

69

the previous sections. However, a filtration test over an extended time frame is indispensable to provide more convincing results. In the following experiments, the activated sludge filtration was studied from two different aspects: (1) examining the long-term filtration resistance profile and (2) investigating the effect of multiple membranes in the MMV system to allow evaluation of the energy consumption.

3.4.1.

Long-term filtration

Since the membrane used is different from the one used in Sections 3.2.3 and 3.3, preliminary experiments were performed to select the optimum values for tC, a and PV. In order to better represent the full-scale operation, the filtration was now performed with a relaxation time included in the intermittent filtration. The choice of filtration cycle duration and ratio was also based on a preliminary experiment in which both of these parameters were varied. The results of the aforementioned preliminary experiments are provided as Supplementary material. For the long-term filtration, the filtration was operated in a 5 min cycle that consisted of 4.5 min of filtration and 0.5 min of relaxation. The experiment was performed in two sequential runs. Initially, 5 PEK membranes were run in parallel. Two membranes were operated in the aerated zone, and 3 membranes were operated with vibration. The distance between the membranes in the MMV system was about 5 mm. The operational parameters J, PV, tC and a for the MMV were set at 16 L/m2 h, 6.4 W, 5 min and 50%, respectively. The applied flux was selected as the flux generally applied for the particular membrane in full-scale applications, and the PV was obtained as the result of a preliminary test. PV was set to be low enough to reduce energy consumption, but high enough to provide an acceptable fouling control. Fig. 9 shows the profile of the filtration resistances during the long-term filtration. After seven days of operation, fouling was found to be more severe for all modules in the vibrated system compared to the ones in the aerated system. This is an obvious contradiction with all the previous results, but can be explained by the arrangement of the membranes. The strongest resistance increase was observed in the case of the vibrated membrane in the second position (i.e. situated

Fig. 7 e Effect of operational flux on filtration resistance in (a) aerated system and (b) MMV system.

70

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 6 3 e7 2

process, but also suggest the importance of adequate design and arrangement of the membranes in one module.

3.4.2.

Fig. 8 e CF determined at different PVs. The shaded area represents optional conditions at which filtration is enhanced due to the additional shear caused by the membrane vibration.

between the two others), suggesting an inappropriate distance between the membranes. Apparently, the membranes in the MMV system were situated so near to each other that the liquid between the membranes moved in-phase and almost became stagnant, moving together with the membrane. To confirm this hypothesis, the filtration was stopped, the second (middle) vibrated membrane was omitted from the reactor, and the filtration was continued. The remaining membranes were chemically cleaned prior to the filtration re-start. Fig. 9 clearly shows that the two vibrating membranes (now with a distance of 10 mm in between) performed better in terms of fouling than the aerated ones throughout the 15 extra days of operation, even though membrane 3 showed a jump on days 13e16 which cannot be explained. These results not only confirm the efficacy of the MMV system in a long-term filtration

Fig. 9 e The profile of filtration resistance during the longterm filtration. Membranes 1e3 were operated in the MMV system, and membranes 4e5 operated in a normal submerged aerated MBR.

Multiple membrane operation and energy consumption

The reduction of energy consumption associated with fouling control is currently one of the main objectives in MBR research. The energy consumption of submerged MBRs is several times higher than that of conventional activated sludge processes (Cornel et al., 2003), and it mainly comes from the energy associated with the coarse bubble aeration for fouling control (Gander et al., 2000). The use of MMV system might offer a promising alternative as a new approach to control fouling in the MBRs. In most shear-enhanced filtration systems, the energy consumption (ED, kWh/m3) is dominated by the energy that is consumed by the vibration engine. Therefore, the energy consumption associated with the MMV system was monitored during this particular test. However, since the ED is calculated based on the volume of permeate, the scale of the plant becomes very significant, mostly favoring large-scale applications. To evaluate the ED of the MMV system, the filtration with multiple membranes was conducted. The MMV system was loaded with up to 6 membranes, to check if there were any changes in filtration performance when the number of membranes in the module increased. Six filtration runs with activated sludge were performed with the MMV system, with each time a different numbers of membranes attached to the module. One additional filtration with six membranes in one module was also performed without vibration for comparison. The filtration parameters were set at J of 16 L/m2 h and PV of 6.4 W, similar to the values used in the long-term test. The profile of the resistance of filtration with different numbers of membranes on the MMV system is shown in Fig. 10. The filtration resistances are given as the average values and the deviations are represented by the shaded area. The results suggest that the lab-scale MMV system could

Fig. 10 e The profile of the filtration resistance for the filtration with different number of membranes mounted on the MMV system.

71

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 6 3 e7 2

Table 2 e Energy demand of lab-scale MBR and comparison with literature data. Reactor (membrane) Lab-scale MBR (KUBOTA flat sheet) Pilot-scale MBR (KUBOTA flat sheet) Full-scale MBR (Zenon hollow fibre) Full-scale MBR (Zenon hollow fibre) Full-scale MBR (hollow fibre)c

ED (kWh/m3)

J (L /m2 h1)

a

12.12 2.03b 6.06 4.88 0.64 < 0.60 1.07

16 19 25 23 26 20

A (m2)

Reference

0.016 0.096 16

Present study Fenu et al. (2010)

10,160 12,130 63,366

Fatone et al. (2007) Verrecht et al. (2010) Gil et al. (2010)

a Calculated from PV with one vibrating membrane in the reactor. b Calculated from PV with six vibrating modules in the reactor. c Theoretical setup for cost sensitivity analysis.

sustain with at least six membranes (each has 0.016 m2 effective area). The addition of up to six membranes did not significantly affect the filtration performance. This result is in line with the new generation of VSEP system, where increasing the membrane area does not significantly affect the ED (Jaffrin, 2008). The addition of more membranes to the MMV system was not feasible in the current setup, due to the limited space available inside the lab-scale reactor tank. The available information on energy consumption of full- or pilot-scale MBRs in scientific literature is scarce. Table 2 contains some related data of selected publications from the last 5 years and furthermore includes ED data from the labscale MMV system. The ED associated with the MMV system was calculated by using a J, PV and a of 16 L/m2 h, 6.4 W and 50%, respectively, similar to the ones used in the long-term test. Table 2 also confirms that the ED associated with coarse bubble aeration is strongly affected by plant scale. The ED of the pilot-scale MBR (Fenu et al., 2010) is almost 10 times the one of the full-scale ones. This should be considered when analyzing the lab-scale MMV system data. Calculating the ED using six membranes (2.03 kWh/m3) gives a 6 times smaller value than with one membrane (12.12 kWh/m3). Nevertheless, this value is much lower than the ED of a pilot-scale MBR that operates with a 160 time larger membrane area, suggesting a rather economic design of the system, despite being far from optimized yet. It is worth noting that the ED of the lab-scale MMV system is about 3.5 times higher than the ED of the best performing full-scale MBR listed in Table 2. However, direct comparison of these data is not entirely reliable, since the ED a lab-scale setup is not of economical scale and the feed and sludge characteristics of an MBR have a serious influence on the filtration performance. From these comparisons, it can thus be expected that an optimised MMV system (frequency, amplitude, vibration cycle, etc) may lead to a significant cost reduction for fouling control in MBRs. With even membrane moving in one direction and odd membrane in the other, more compact, but still efficient modules could possibly be realized.

conventional MBR processes in terms of realisable flux and fouling control. Significant improvement of CF was obtained due to the enhanced shear at the liquidemembrane interface. The filtration was found sensitive to several operation factors such as the vibration parameters (e.g., vibration power and cycles) and the applied flux. The long-term experiments confirmed the efficacy of the MMV system, but also suggested the importance of an appropriate membrane arrangement in the MBR in the module. The energy demand of vibration, resulting in the highest of all the MBR costs, was found practically constant when the number of modules mounted in the MMV system was increased from one up to six, while increasing treated water volumes 6-fold. The MMV-aided filtration, after process optimisation, is expected to lead to significant cost less membrane area to be installed when taking advantage of the higher CF and energy (especially when expending the number of membranes per module) reduction in (up-scaled) MBRs. This novel membrane fouling limitation method seems very promising in MBRs, but also for the currently progressing anaerobic MBRs where coarse air bubbling is not an option, and possibly also other fouling sensitive ultrafiltration and nanofiltration operations, like algae harvesting.

Acknowledgements The authors gratefully acknowledge the financial support provided by K.U. Leuven (CECAT excellence, GOA and IDO financings), by the Flemish Government (Methusalem funding) and by the Federal Government (IAP grant). Special thanks to Waterleau for providing the sludge seed and giving technical support at start-up of the lab-scale MBR; Wim Ruttens (Intellitech) for his very useful help in the system development; and Toray Membrane Europe for providing the A4 size membrane element samples.

Abbreviations and symbols 4.

Conclusions

Innovative magnetically induced membrane vibration proved very promising in a lab-scale MBR treating synthetic wastewater treatment. Results of both the filtration and the CF measurements showed clear advantages of this system over

A CF ED J JV

membrane surface area (m2) (normalised) critical flux (L/m2 h) energy demand (kWh/m3) permeate flux (L/m2 h) permeate flow velocity (m/s)

72

L MBR MLSS MMV NF PV PEK PVDFT RF RM RT t tC TMP SVI UF DV a h

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 6 3 e7 2

membrane permeability (L/m2 h Pa) membrane bioreactor mixed liquor suspended solid (g/L) magnetically induced membrane vibration nanofiltration vibration power (W) polyethylene (Kubota) polyvinylidene fluoride (Toray) fouling resistance (m1) intrinsic embrane resistance (m1) filtration resistance (m1) filtration time vibration cycle time (min) trans-membrane pressure (Pa) sludge volume index (mL/g) ultrafiltration permeate volume (L) intermittent vibration fraction (%) dynamic viscosity of permeate (Pa s)

Appendix A. Supplementary data Supplementary data associated with this article can be found in the online version at doi:10.1016/j.watres.2011.10.026

references

Altaee, A., Al-Rawajfeh, A.E., Baek, Y.J., 2010. Application of vibratory system to improve the critical flux in submerged hollow fiber MF process. Separation Science and Technology 45 (1), 28e34. American Public Health Association/American Water Works Association (APHA)/Water Environment Federation, 1992. Standard Methods for the Examination of Water and Wastewater, 18th ed. APHA, Washington DC, USA. Beier, S.P., 2008. Dynamic Microfiltration e Critical Flux and Macromollecular Transmission. Ph.D. dissertation, Technical University of Denmark. Beier, S.P., Jonsson, G., 2007. Separation of enzymes and yeast cells with a vibrating hollow fiber membrane module. Separation and Purification Technology 53 (1), 111e118. Beier, S.P., Jonsson, G., 2009. A vibrating membrane bioreactor (VMBR): macromolecular transmission e influence of extracellular polymeric substances. Chemical Engineering Science 64 (7), 1436e1444. Beier, S.P., Guerra, M., Garde, A., Jonsson, G., 2006. Dynamic microfiltration with a vibrating hollow fiber membrane module: filtration of yeast suspensions. Journal of Membrane Science 281 (1e2), 281e287. Bilad, M.R., Declerck, P., Piasecka, A., Vanysacker, L., Yan, X., Vankelecom, I.F.J., 2011a. Development and validation of a high-throughput membrane bioreactor (HT-MBR). Journal of Membrane Science. doi:10.1016/j.memsci.2011.05.052. Bilad, M.R., Declerck, P., Piasecka, A., Vanysacker, L., Yan, X., Vankelecom, I.F.J., 2011b. Treatment of molasses wastewater in a membrane bioreactor: influence of membrane pore size. Separation and Purification Technology 78 (2), 105e112. Bilad, M.R., Westbroek, P., Vankelecom, I.F.J., 2011c. Assessment and optimization of electrospun nanofiber-membranes in a membrane bioreactor (MBR). Journal of Membrane Science 380, 181e191.

Chang, I.-S., Le-Clech, P., Jefferson, B., Judd, S.J., 2002. Membrane fouling in membrane bioreactors for wastewater treatment. Journal of Environmental Engineering 128 (11), 1018e1029. Cornel, P., Wagner, M., Krause, S., 2003. Investigation of oxygen transfer rates in full scale membrane bioreactors. Water Science and Technology 47 (11), 313e319. Cui, Z.F., Chang, S., Fane, A.G., 2003. The use of gas bubbling to enhance membrane processes. Journal of Membrane Science 221, 1e35. Drews, A., 2010. Membrane fouling in membrane bioreactors e characterisation, contradictions, cause and cures. Journal of Membrane Science 363 (1e2), 1e28. Drews, A., Evenblij, H., Rosenberger, S., 2005. Potential and drawbacks of microbiologyemembrane interaction in membrane bioreactors. Environmental Progress 24 (4), 426e433. 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. Industrial & Engineering Chemistry Research 46 (21), 6688e6695. Fenu, A., Roels, J., Wambecq, T., De Gussem, K., Thoeye, C., De Gueldre, G., Van De Steene, B., 2010. Energy audit of a full scale MBR system. Desalination 262 (1e3), 121e128. Gander, M., Jefferson, B., Judd, S., 2000. Aerobic MBRs for domestic wastewater treatment: a review with cost considerations. Separation and Purification Technology 18, 119e130. Genkin, G., Waite, T.D., Fane, A.G., Chang, S., 2006. The effect of vibration and coagulant addition on the filtration performance of submerged hollow fibre membranes. Journal of Membrane Science 281 (1-2), 726e734. Gil, J.A., Tu´a, L., Rueda, A., Montan˜o, B., Rodrı´guez, M., Prats, D., 2010. Monitoring and analysis of the energy cost of an MBR. Desalination 250 (3), 997e1001. Jaffrin, M.Y., 2008. Dynamic shear-enhanced membrane filtration: A review of rotating disks, rotating membranes and vibrating systems. Journal of Membrane Science 324 (1e2), 7e25. Kola, A., Ye, Y., Stuetz, R., Le-Clech, P., Chen, V., 2011 Transverse vibration as a fouling limitation strategy in membrane bioreactors. Oral presentation presented at ICOM 2011, Amsterdam, The Netherlands, July 23e29, 2011. Le-Clech, P., Jefferson, B., Chang, I.S., Judd, S.J., 2003. Critical flux determination by the flux-step method in a submerged membrane bioreactor. Journal of Membrane Science 227 (1e2), 81e93. Le-Clech, P., Chen, V., Fane, T.A.G., 2006. Fouling in membrane bioreactors used in wastewater treatment. Journal of Membrane Science 284 (1e2), 17e53. Low, S.C., Cheong, K.T., Lim, H.L., 2009. A vibration membrane bioreactor. Desalination and Water Treatment 5, 42e47. Meng, F., Chae, S.-R., Drews, A., Kraume, M., Shin, H.-S., Yang, F., 2009. Recent advances in membrane bioreactors (MBRs): membrane fouling and membrane material. Water Research 43 (6), 1489e1512. Rosenberger, S., Kraume, M., 2003. Filterability of activated sludge in membrane bioreactors. Desalination 151, 195e200. Verrecht, B., Maere, T., Nopens, I., Brepols, C., Judd, S., 2010. The cost of a large-scale hollow fibre MBR. Water Research 44 (18), 5274e5283. Wu, Z., Wang, Z., Huang, S., Mai, S., Yang, C., Wang, X., Zhou, Z., 2008. Effects of various factors on critical flux in submerged membrane bioreactors for municipal wastewater treatment. Separation and Purification Technology 62 (1), 56e63. Yan, X., Gerards, R., Vriens, L., Vankelecom, I.F.J., 2010. Hollow fiber membrane fouling and cleaning in a membrane bioreactor for molasses wastewater treatment. Desalination and Water Treatment 18, 192e197.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 7 3 e8 1

Available online at www.sciencedirect.com

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

Effect of moderate pre-oxidation on the removal of Microcystis aeruginosa by KMnO4eFe(II) process: Significance of the in-situ formed Fe(III) Min Ma a,b, Ruiping Liu a, Huijuan Liu a, Jiuhui Qu a,* a b

Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing 100085, China Graduate School of Chinese Academy of Sciences, Beijing 100039, China

article info

abstract

Article history:

This study developed a novel KMnO4eFe(II) process to remove the cells of Microcystis aer-

Received 12 May 2011

uginosa, and the mechanisms involved in have been investigated. At KMnO4 doses of

Received in revised form

0e10.0 mM, the KMnO4eFe(II) process showed 23.4e53.3% higher efficiency than the

21 September 2011

KMnO4eFe(III) process did. This was first attributed to the moderate pre-oxidation of M.

Accepted 15 October 2011

aeruginosa by KMnO4, achieved by dosing Fe(II) after a period of pre-oxidation, to cease the

Available online 25 October 2011

further release of intracellular organic matter (IOM) and the degradation of dissolved

Keywords:

process led to high levels and insufficient molecular weight of DOM, inhibiting the

Microcystis aeruginosa

subsequent Fe(III) coagulation. Additionally, Fe(II) contributed to lower levels of the in-situ

KMnO4eFe(II) process

formed MnO2, the reduction product of KMnO4 which adversely affected algae removal by

organic matter (DOM). The extensive exposure of M. aeruginosa to KMnO4 in KMnO4eFe(III)

Moderate pre-oxidation

Fe(III) coagulation. However, the in-situ formed Fe(III), which was derived from the oxida-

The in-situ formed Fe(III)

tion of Fe(II) by KMnO4, in-situ MnO2, and dissolved oxygen, dominated the remarkably high

MnO2

efficiency of KMnO4eFe(II) process with respect to the removal of M. aeruginosa. On one

Dissolved organic matter

hand, in-situ formed Fe(III) had more reactive surface area than pre-formed Fe(III). On the other hand, the continuous introduction of fresh Fe(III) coagulant showed higher efficiency than one-off dosage of coagulant to destabilize M. aeruginosa cells and to increase the flocs size. Moreover, the MnO2 precipitated on algae cell surfaces and contributed to the formation of in-situ formed Fe(III), which may act as bridges to enhance the removal of M. aeruginosa. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

The removal of algae is a well-concerned issue for drinking water treatment plants (DWTPs) that suffer from algae bloom in water sources. The conventional treatment processes, i.e., coagulation/flocculation, sedimentation, and filtration, cannot effectively remove algae cells. This is ascribed to the nature of algae cells, including the low

specific density, high motility, negatively-charged surface, and diverse morphology (Bernhardt and Clasen, 1991; Pieterse and Cloot, 1997; Ma and Liu, 2002). Consequently, the proliferation of algae often has adverse effects of the increased coagulant demand, the clogging and penetrating of filters, and more frequent backwashing (Schmidt et al., 1998; Plummer and Edzwald, 2001). Additionally, the dramatic growth of algae is also associated with other water quality

* Corresponding author. Tel.: þ86 10 62849160; fax: þ86 10 62923558. E-mail address: [email protected] (J. Qu). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.022

74

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 7 3 e8 1

problems such as unpleasant taste and odor, toxins, and disinfection by-products (DBPs). There have been proposed several strategies to enhance the removal of algae. Generally, air flotation showed better performance than sedimentation with an overall improvement of 15e20% (Henderson et al., 2008a). The dissolved air flotation (DAF) contributed to the removal efficiency of above 94%, with the highest value of 99.8%, toward different algae species of cyanobacteria Microcystis aeruginosa, green C. vulgaris, and diatom Cyclotella. Ultrafiltration (UF) was another effective approach to remove M. aeruginosa with removal efficiencies of above 90% (Tan et al., 2008). Unfortunately, the process innovation from sedimentation to DAF and UF is usually hindered by the heavy investment, increased operating cost, and complicated operation. It is well preferred to achieve the enhanced removal of algae without large-scale reconstruction. The strategies such as increasing coagulant doses, applying coagulants with better efficiency, and introducing inorganic polymers to aid coagulation may be included. Pre-oxidation is another feasible strategy which has been popular all round the world. The addition of various oxidants [i.e., Cl2, O3, KMnO4, ClO2 and Fe(VI)] prior to coagulation has been shown to significantly improve the algae removal by coagulation or filtration. During alum coagulation at different doses, chlorine at 2 mg/L increased the removal of Chlamydomonas and Euglena gracilis by 85% and 95%, respectively (Steynberg et al., 1996). To achieve the same 50% removal of control algae cultures, the addition of ClO2 (1, 3 and 5 mg/L) and O3 (2.6, 4.6 and 8.1 mg/L) may reduce the alum doses by 13e75% and 41e58%, respectively (Sukenik et al., 1987). KMnO4 at 1.7 mg/L increased the algae removal from 70% to 100% when 40 mg/L of alum was dosed (Chen et al., 2009). Ferrate was also reported to significantly enhance the removal of algae and reduce the coagulant demand, even at very short pre-oxidation time (Ma and Liu, 2002). Among these oxidants, KMnO4 is the most convenient to use in case of urgent algae bloom. However, extensive pre-oxidation causes the lysis of algae cells and the release of intracellular organic matters (IOM), and this should be well considered before applying strong oxidants. In addition to the elevated risk of DBPs formation, the released IOM also inhibited coagulation when they were of high levels or insufficient molecular weight (MW) (Bernhardt et al., 1985). Therefore, an ideal pre-oxidation should be “moderate” to achieve the balance between the two goals of avoiding extensive pre-oxidation of algae and improving the removal of algae. To our best knowledge, rare studies have focused on this theme before. As a strong oxidant, the effects of KMnO4 on algae removal varied with its doses (Petrusevski et al., 1996). The appropriate doses benefited the removal of algae by coagulation whereas the excessive doses showed adverse impacts (Chen et al., 2009). Moreover, the overdose of KMnO4 leads to other problems such as the increase in the residual Mn, color, and turbidity. To avoid the aforementioned side effects of extensive pre-oxidation, a novel KMnO4eFe(II) process, i.e., the introduction of Fe(II) after the pre-oxidation of algae cells by KMnO4 for certain period, is brought forward in this study to achieve moderate oxidation of algae cells. Additionally, Fe(III)

will be freshly formed in this KMnO4eFe(II) process due to the oxidation of Fe(II) by KMnO4 and/or oxygen. This kind of Fe(III) formed in the original place of Fe(II) in aqueous solution is defined as the in-situ formed Fe(III), which is more effective than the pre-formed Fe(III) [e.g., Fe2(SO4)3] in removing contaminants such as arsenic (Guan et al., 2009) and phosphate (Lee et al., 2009). Furthermore, the continuous transformation of Fe(II) to Fe(III) may show different coagulation behaviors from the one-off addition of Fe(III). However, there is a lack of studies which have investigated the mechanisms involved in these processes. This study aimed to: (1) compare the removal efficiencies of M. aeruginosa by the KMnO4eFe(II) and KMnO4eFe(III) processes; (2) investigate the roles of different species, i.e., dissolved organic matter (DOM), in-situ formed MnO2, and insitu formed Fe(III), in the removal of algae by KMnO4eFe(II) process, and demonstrate the significance of the in-situ formed Fe(III); (3) propose the interfacial reactions involved in the removal of algae cells by KMnO4eFe(II) process. A laser particle size analyzer was employed to determine the dynamic growth and size distribution of algae flocs; to date, no studies have employed it for this application.

2.

Materials and methods

2.1.

Materials and reagents

An axenic strain of M. aeruginosa (No. FACHB-905) previously described by (Shen and Song, 2007) was used in this study, and was cultured in BG-11 (Rippka et al., 1979) medium. Full details of growth conditions are presented in the Supporting Information. All chemicals used were reagent-grade and all solutions were prepared with deionized water. Ferrous sulfate (FeSO4) or ferric sulfate [Fe2(SO4)3] solutions were prepared just before experiments. Freshly-formed MnO2 was prepared by mixing KMnO4 and MnCl2 at mol ratio of 2:3, immediately before being dosed to algae suspension; therefore, it could be considered as in-situ formed MnO2. Algogenic organic matter (AOM) was derived from algae cells at the exponential growth phase. Full details about AOM extraction procedure are described in the Supporting Information. Source water was collected from Miyun reservoir located in north China. And its qualities were as follows: turbidity 2.8 NTU, pH 7.7e7.9, temperature 23e25  C, DOC 2.60 mg/L, Ca 43.2 mg/L.

2.2.

Jar tests

M. aeruginosa cultures were harvested at the exponential growth phase and then spiked with source water to obtain the cell density of 1.0  106 cells/mL. The pH of this algae suspension was about 8.4. Jar tests were performed with 300 mL sample in 500 mL beaker and conducted on a programmable jar tester (MY3000-6, MeiYu, China). After the dosage of KMnO4 or MnO2, samples were rapidly mixed at 250 rpm for 5 min FeSO4 or Fe2(SO4)3 was dosed after the preoxidation. The coagulation process consisted of the mixing at 200 rpm for 2 min and 40 rpm for 15 min, consecutively. Then,

75

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 7 3 e8 1

2.3.

Analytical methods

2.3.1.

Determination of the levels of Kþ, Fe and Mn

An inductively coupled plasma optical emission spectrometer (OPTIMA 2000, PerkinElmer, USA) and/or an inductively coupled plasma mass spectrometer (5000a, Agilent, USA) was used to determine the concentrations of Kþ, Fe and Mn.

2.3.2.

Measurement of residual KMnO4 and UV254 of DOM

Residual KMnO4 and UV254 of DOM were measured with a UV/ vis spectrophotometer (U-3010, Hitachi Co., Japan) at 530 and 254 nm, respectively.

2.3.3.

located downstream of the laser particle size analyzer and run at a rate of 28 rpm d50 served as the representative floc size.

3.

Results and discussion

3.1. Removal of M. aeruginosa by KMnO4eFe(II) and KMnO4eFe(III) processes Fig. 1 compares the removal efficiency of M. aeruginosa between two processes of KMnO4eFe(II) and KMnO4eFe(III) with KMnO4 doses ranging from 0 to 10 mM KMnO4eFe(II) process exhibited significantly higher removal of algae compared to KMnO4eFe(III) process. Without KMnO4, the coagulation by Fe(III) and Fe(II) contributed to algae removal of 26.5% and 49.8%, respectively. KMnO4 at 1.7 mM increased the removal of algae to 36.5% for KMnO4eFe(III) process and to remarkably higher efficiency of 89.7% for KMnO4eFe(II) process. The elevated KMnO4 doses of 2.3e10.0 mM further enhanced the algae removal, which was 91.0e95.6% in KMnO4eFe(II) process and 45.2e62.6% in KMnO4eFe(III) process, respectively.

3.2. Dynamic growth of flocs in KMnO4eFe(II) and KMnO4eFe(III) processes The dynamic growth of flocs in KMnO4eFe(II) and KMnO4eFe(III) processes were compared at KMnO4 doses of 1.7 and 5.0 mM, and the obviously different trends in flocs growth were observed (Fig. 2). The KMnO4eFe(II) process contributed to smaller flocs than KMnO4eFe(III) process initially; however, there showed larger flocs after a critical contact time. Quantitatively, at 1.7 mM of KMnO4, the d50 values in KMnO4eFe(II) process was below 200 mm in the initial 720 s. After that, d50 values increased steadily to as high as 239 mm at 990 s. Comparatively, the d50 values in KMnO4eFe(III) process were 198 and 204 mm at contact time of 720 and 990 s, respectively. Similar trends were observed at 5.0 mM of KMnO4 except for lower intersecting contact time (330 s) and more significant difference between d50 at 990 s in

100

Algae removal (%)

samples were quiescently settled for 30 min. The equivalent molar ratio of KMnO4 to FeSO4 is 1:3; while the molar ratios of KMnO4 to FeSO4 in this study ranged from 120:1 to 20:1. After the settling, samples were siphoned 2 cm below the water surface and divided in two subsamples: the first sample was measured for residual algae by optical density at 680 nm (OD680) with a U-3100 spectrophotometer (Hitachi Co., Japan). The remaining sample was immediately subjected to filtration through a filter (0.45 mm, glass fiber) and then determined for concentrations of Kþ, and total Fe and Mn. In some tests, the level of residual Fe(II) was measured during the coagulation and settling. Details about residual Fe(II) measurement are presented in Supporting Information. For tests measuring the effects of KMnO4 on algae cells, at each pre-determined time, samples withdrawn from the bulk samples were filtered through a filter (0.45 mm, glass fiber) and divided into two subsamples. One subsample was immediately measured for residual KMnO4 and UV254. The other was for the analysis of Kþ concentration. Details about tests studying the effect of pre-oxidation time on MW of DOM in samples containing M. aeruginosa cells are shown in Supporting Information. All experiments were conducted in at least duplicate. Standard error was calculated by software SPSS 13.0. For the tests studying effects of dissolved organic matter (DOM) and the dosage strategy of coagulant on the algae coagulation, the floc growth was measured with a laser particle size analyzer. Jar tests were conducted with 900 mL sample in 1L beaker. For the tests studying the effect of AOM on algae removal, AOM stock solution was added to algae suspension before any treatment. The mixing procedure was: 250 rpm for 2 min and then 50 rpm for 15 min. For the tests of coagulant dosage strategy, certain volumes of Fe(III) were dosed consecutively at several pre-determined times, with the same total dose of Fe(III). The mixing procedure was: 250 rpm for 15 min and then 50 rpm for 5 min.

75

50

25

KMnO 4 -Fe(II)

Size measurement and structure determination of flocs

In some tests, the floc growth during the coagulation processes was studied with a laser particle size analyzer (Malvern Instruments, UK). Jar tests were conducted with 900 mL sample on a programmable jar tester. Algae suspension was drawn from the jar through a latex tube to the sample cell of the laser particle size analyzer; and then back to the jar by a peristaltic pump (BT00-300M, Longer, China). To avoid disturbing floc before the measurement, the pump was

KMnO 4 -Fe(III) 0 0

2

4

6

8

10

12

KMnO4 dose ( M) Fig. 1 e Comparison of the removal of M. aeruginosa at different KMnO4 doses between the KMnO4eFe(II) and KMnO4eFe(III) processes. Cell density: 1.0 3 106 cells/mL. Fe dose: 197.4 mM. Pre-oxidation time: 5 min.

76

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 7 3 e8 1

400

15 M

a

300 10 % Volume

Floc size d 50 ( m)

KMnO 4 dose: 1.7

200

100

Fe(II) 4 min Fe(II) 17 min Fe(III) 4 min Fe(III) 17 min

a 1.7

M KMnO4

b 5.0

M KMnO4

5

KMnO 4 -Fe(II) KMnO 4 -Fe(III) 0 15

4000

b

M

300 10 % Volume

Floc size d 50 ( m)

KMnO 4 dose: 5.0

200

100

0 0

300

600

900

1200

Time (s) Fig. 2 e Comparison of floc growth during the coagulation processes between the KMnO4eFe(II) and KMnO4eFe(III) processes. Cell density: 1.0 3 106 cells/mL. Fe dose: 197.4 mM. Pre-oxidation time: 5 min.

two processes (148 mm). This may be attributed to the more rapid formation of the in-situ Fe(III) at higher dose of KMnO4. To better understand the difference between KMnO4eFe(II) and KMnO4eFe(III) processes, it is essential to study the change of size distribution of flocs in coagulation. Flocs at 4 and 17 min served as the representative flocs in the floc growth and steady-state phases, respectively. As shown in Fig. 3, for both processes the volume percentage of large flocs increased as coagulation processed. In addition, for KMnO4eFe(II) process there was a substantial decrease in the volume percentage of small flocs whose sizes ranged between 10 and 50 mm (F10e50 mm). The difference between the volume percentage of F10e50 mm at 4 and 17 min was much less significant in the algae coagulation by KMnO4eFe(III).

3.3. Evaluation of different effects on M. aeruginosa removal by KMnO4eFe(II) process 3.3.1.

Contribution of KMnO4 oxidation

KMnO4 in the KMnO4eFe(II) process first served as an oxidant to inactivate algae cells [Eq. (1)]. Fig. 4a illustrates the kinetics of Kþ release at KMnO4 doses from 0.5 to 3.0 mg/L. The release of Kþ, which may indicate the integrity of cell membrane (Peterson et al., 1995), increased with elevated KMnO4 doses and prolonged contact time. Additionally, there was a rapid increase in the level of Kþ within the initial 5e10 min, which was followed by a more gradual increase thereafter. The exposure of M. aeruginosa to 3 mg/L of KMnO4 for 2 min

5

0 10

100

1000

Floc size ( m) Fig. 3 e Particle size distribution for M. aeruginosa flocs after the growth phase in the KMnO4eFe(II) and KMnO4eFe(III) processes. KMnO4 dose: (a) 1.7 mM and (b) 5.0 mM. Fe dose: 197.4 mM. Cell density: 1.0 3 106 cells/mL.

resulted in the Kþ release of as high as 59.6%. There showed consistent trends between Kþ release and KMnO4 decay, and the levels of KMnO4 decreased significantly in the initial 10 min and then decreased slightly (Fig. S1).

KMnO4 þ Cell / Cell* þ MnO2 þ released IOM

(1)

The release of IOM also occurred after the exposure of M. aeruginosa to KMnO4 [Eq. (1)], as indicated by the UV254 analysis (Fig. 4b). The increase in KMnO4 doses resulted in higher UV254 values and thus more significant release of IOM. Interestingly, the levels of IOM decreased to some extent with prolonged contact time, different from that of Kþ release. This is attributed to the adsorption activity of hydrous MnO2 toward DOM (Chen and Yeh, 2005) and to the degradation effect of KMnO4 [Eqs. (2) and (3)].

DOM þ MnO2 / DOM h MnO2

(2)

KMnO4 þ DOM / MnO2 þ DOM*

(3)

The zeta potential of M. aeruginosa suspension was consistently ranging from 2.0 to 4.0 mV before and after KMnO4 oxidation at 3.2e19.0 mM. This trend has also been

77

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 7 3 e8 1

100

120 KMnO4 (mg/L): 0 2.0

1.0

a

Algae removal (%)

75

60

+

K release (%)

90

0.5 3.0

Fe(II) Fe(III)

a

30

50

25

0

3.0

0

b

b

0.10

1.5 Residual Fe (%)

-1

(cm )

0.09

UV

254

0.08 0.07 0.06

0.3 0.2 0.1 0.0 197.3

0.05 0

10

20

30

Pre-oxidation time (min) Fig. 4 e Effect of KMnO4 pre-oxidation on the variation of (a) KD release and (b) UV254 of DOM. Cell density: 1.0 3 106 cells/mL. The concentrations of KD due to KMnO4 addition were subtracted accordingly. Error bars were not shown in (a) for the consideration of readability.

reported before (Chen and Yeh, 2005). The surface charge of Scenedesmus cells may be an indicator on the changes in the outside cell membrane (Plummer and Edzwald, 2002). It is inferred that KMnO4 altered the outside membrane of M. aeruginosa cells to a small extent.

394.7

Fig. 5 e Comparison of (a) algae removal efficiency and (b) Fe in solution after coagulation between Fe(II) and Fe(III) coagulation processes. Cell density: 1.0 3 106 cells/mL.

The effect of settling time on the removal of M. aeruginosa in Fe(III) and Fe(II) coagulation was also investigated (Fig. 6). Despite of the similar removal efficiencies of algae in two processes after a 30-min settling, cells settled more rapidly by Fe(II) coagulation when settling period was prolonged. After a 60-min settling, Fe(II) coagulation contributed to algae removal of 64.0%, which was as low as 36.0% by Fe(III) coagulation. After that, the algae removal by Fe(II) coagulation increased steadily to as high as 82.3% after 150-min settling,

Contribution of the in-situ formed Fe(III)

Levels of residual Fe were higher in KMnO4eFe(II) process as compared to those in KMnO4eFe(III) process (Fig. S2a). The less participation of Fe in algae removal in KMnO4eFe(II) process indicates the different coagulation behaviors of Fe(II) and Fe(III) toward algae cells. To exclusively investigate the contribution of the in-situ formed Fe(III), this study first compared the removal of M. aeruginosa by Fe(III) and Fe(II) coagulation without KMnO4 (Fig. 5a). The algae removal by Fe(II) coagulation was 10%e41% higher than that by Fe(III) coagulation at different Fe doses from 197.4 to 394.7 mM. However, the levels of residual Fe associated with Fe(II) coagulation was 5e8 times higher than those in Fe(III) coagulation, with more than 98.4% of Fe(II) insitu oxidized to Fe(III) by oxygen accordingly (Fig. 5b). Similar trends were also observed in coagulation with KMnO4 preoxidation (Fig. S2a). The discrepancy between the algae removal and the contents of Fe in the precipitates in those two processes indicates a surprisingly higher efficiency of Fe(II) coagulation toward algae removal.

100 Fe(II) Fe(III) 75 Algae removal (%)

3.3.2.

276.3

Coagulant dose (as Fe) ( M)

50

25

0 0

30

60

90

120

150

Settling time (min) Fig. 6 e Comparison of settling rates between Fe(II) and Fe(III) coagulation processes. Cell density: 1.0 3 106 cells/ mL. Fe dose: 197.4 mM.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 7 3 e8 1

and only 4.3% of improvement was achieved by Fe(III) coagulation with the settling time increasing from 30 to 150 min. The different coagulation and settling behaviors of flocs between Fe(II) and Fe(III) coagulation is in support of the better performance of the in-situ formed Fe(III) than the pre-formed Fe(III) with respect to the algae removal. In the KMnO4eFe(II) process, the residual KMnO4 and oxygen, as well as the in-situ formed MnO2 which will be discussed in detail later, showed good activities to oxidize Fe(II) to Fe(III) [Eqs. (4e6)]. Fe(II) acted as the source to continuously provide the in-situ Fe(III) (Fig. S3), which was an important contribution to the remarkably higher removal of algae in KMnO4eFe(II) process.

Fe(II) þ KMnO4 / in-situ Fe(III) þ MnO2þ

(4)

Fe(II) þ MnO2 / in-situ Fe(III) þ Mn2

(5)

Fe(II) þ O2 / in-situ Fe(III)

(6)

First, continuously introduction of the in-situ Fe(III) inhibited the adverse impacts of the coagulant aging. A timedependent mechanism inferred by Flynn et al. depicted a three-step hydrolysis process of Fe(III): 1) formation of low MW complexes, consisting of [Fe(OH)]2þ, [Fe(OH)2]þ and [Fe2(OH)2]4þ; 2) formation and aging of polynuclear polymers, such as [Fep(OH)r(H2O)s](3pr)þ or [FepOr(OH)s](3p2rs)þ; and 3) precipitation of Fe(III) oxides (Fe2O3) and hydroxides [Fe(OH)3 and FeO(OH)] (Flynn, 1984). Over these dissolutionprecipitation processes, the structure of iron oxides became more ordered by in-situ rearrangement (Jolivet et al., 2004). The aging of hydrolyzed Fe(III) also increased the particle size and thus decreased the reactive surface area that was accessible for reaction (Vikesland et al., 2007), reducing the reactivity of Fe(III) oxides as a whole. Due to the slower aging of coagulant as a whole, more active surface areas and higher reaction rates can be achieved by Fe(II) coagulation than by Fe(III) coagulation. Furthermore, hydrolysis intermediates of in-situ formed Fe(III) might immediately attach to and bind with the aged iron hydroxide precipitate to achieve positive surfaces. Thus, the continuous provision of in-situ formed Fe(III), rather than the one-off dosing of Fe(III), advanced the effective agglomerate of tiny particles and the subsequent settling thereafter. The presence of algae cells and DOM with large amount of negative functional groups (i.e., eOH, eCOOH) even complicated the roles of in-situ Fe(III) involved in. The hydrolytically formed iron hydroxo complexes and iron oxide hydroxides can form surface complexes with organic anionic polymers (OAP) which contains eOH and eCOOH groups (Bernhardt et al., 1985; Pivokonsky et al., 2006). The pre-formed Fe(III) promptly hydrolyzed to Fe(OH)3 precipitates soon after its being dosed. In contrast, the in-situ formed Fe(III) and its hydrolysis products may be more active to form these complexes than the pre-formed Fe(III) for the aforementioned reasons. And these interesting characteristics enhanced the interactions between algae cells and Fe(III) to improve algae removal. Although it is difficult to raise the exact reactions, the significance of the in-situ formed Fe(III) can be proposed.

Moreover, the heterogeneous oxidation of Fe(II) on the surfaces of MnO2 also played an important role. KMnO4 rapidly decayed and transformed to the in-situ MnO2 after its interaction with algae suspensions [Eqs. (1) and (2)], and MnO2 may precipitate on the surfaces of algae cells to some extent [Eq. (7)] (Chen and Yeh, 2005). Besides the possibly residual KMnO4 and oxygen, the in-situ MnO2 also contributed to the oxidation of Fe(II). This suggestion is supported by the much higher levels of residual Mn in KMnO4eFe(II) process than those in KMnO4eFe(III) process (Fig. S2b). The heterogeneous oxidation of Fe(II) to Fe(III) on the surfaces of tinny MnO2 particles may continuously provide the in-situ formed Fe(III) to destabilize MnO2 particles, algae cells, and the algae cells with precipitated MnO2 [Eqs. (8) and (9)]. Correspondingly, the role of in-situ formed MnO2 in the coagulation was altered, improving the removal of algae. And the role of in-situ formed MnO2 will be discussed in detail later.

Cell þ MnO2 / Cell h MnO2

(7)

Fe(II) þ DOM-MnO2 / in-situ Fe(III) þ DOM þ Mn2þ

(8)

Fe(II) þ Cell-MnO2 / in-situ Fe(III) þ Cell þ Mn2þ

(9)

To further prove the suggested advantage of continuous provision of in-situ Fe(III) over one-off dosage of pre-formed Fe(III), the dynamic growth of flocs under different dosingFe(III) strategies was investigated (Fig. 7).Despite of the same levels of total Fe, the consecutive dosing of Fe(III) contributed to the formation of larger flocs than the one-off dosing-Fe(III) strategy. After slow-mixing, the d50 value in Stategy-I [i.e., 197.4 mM Fe(III) at 0 min] was 95 mm, which increased to 112 mm in Stategy-IV [i.e., 131.6 mM Fe(III) at 0 min and 32.9 mM Fe(III) at 5 and 10 min]. The higher d50 values of 120 and 125 mm were

150 Strategy I Strategy II Strategy III Strategy IV

120 Floc size d 50 ( m)

78

90 60 30 Rapid mixing 0 0

400

800

Slow mixing 1200

Time (s) Fig. 7 e Floc growth in the coagulation with different dosage strategies of Fe(III). (I) 197.4 mM (0 min); (II) 131.6 mM (0 min) D 65.8 mM (5 min); (III) 65.8 mM (0 min) D 65.8 mM (5 min) D 65.8 mM (10 min); and (IV) 131.6 mM (0 min) D 32.9 mM (5 min) D 32.9 mM (10 min). Cell density: 1.0 3 106 cells/mL. Sample volume: 900 mL. Mixing procedure: 250 rpm, 15 min and then 50 rpm, 5 min.

79

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 7 3 e8 1

observed in Stategy-III [i.e., 131.6 mM Fe(III) at 0 min and 65.8 mM Fe(III) at 5 min] and Stategy-II [i.e., 65.8 mM Fe(III) at 0, 5, and 10 min], respectively.

Contribution of the in-situ formed MnO2

MnO2 may also affect the removal of algae. Fig. 8 compares the effect of KMnO4 and the in-situ formed MnO2 at the same Mn doses on the removal of algae by Fe(III) coagulation. MnO2 showed adverse influences on the algae removal. In the control experiment, Fe(III) coagulation alone removed 23.3% of algae cells. The presence of MnO2 at 0.83e6.67 mM decreased the algae removal to 19.7e21.3%. This might be ascribed to the repulsive forces between the negativelycharged algae cells and MnO2. The isoelectric point of MnO2 was reported to be in the range of 2.8 and 4.5 (Posselt and Anderson, 1968). At pH 8.3 in this study, the negativelycharged MnO2 inhibited the collision and aggregation between algae cells. In contrast, MnO2 at 0.83e6.67 mM improved the algae removal by 2.7e8.0% at pH 3.8 (Fig. S4), which may be attributed to the adsorption of positivelycharged MnO2 onto algae cells and thus the increase in the specific weight of cells (Chen and Yeh, 2005). Different from MnO2, KMnO4 at 0.63e6.67 mM enhanced the removal of algae by 2.0%e30.3%. These different behaviors indicate the significant role of KMnO4 oxidation on the removal of algae cells.

3.3.4.

Effect of DOM on the floc growth

KMnO4 pre-oxidation also contributed to the release of IOM and thus the elevated levels of DOM (Fig. 4b). The impacts of DOM on the dynamic growth of flocs during Fe(III) coagulation varied with its concentration. As shown in Fig. 9, DOM at low levels obviously benefited the formation of larger flocs. The d50 values at 960 s increased from 121 mm to 158 and 212 mm due to the addition of 0.5 and 1 mg/L of DOM. However, DOM at high concentrations of 2 and 3 mg/L negatively decreased d50 to lower values of 130 and 123 mm. There are some previous studies that show the effect of IOM on the coagulation. IOM mainly consists of protein, polysaccharides, and lipids (Henderson et al., 2008b), and these cyanobacterium-derived organics were reported to behave like anionic and non-ionic

60 MnO 2

Algae removal (%)

KMnO 4

AOM addition (mg/L)

Floc size d50 ( m )

3.3.3.

300 0 0.5 1.0 2.0 3.0

200

100

0 0

200

400

600

800

1000

Time (s) Fig. 9 e Effect of DOM on the floc growth during the coagulation process. Cell density: 1.0 3 106 cells/mL. Fe(III) dose: 197.4 mM. Sample volume: 900 mL. Mixing procedure: 250 rpm, 2 min and then 50 rpm, 15 min.

polyelectrolytes, depending on their concentrations and MW (Bernhardt and Clasen, 1991). At low DOM levels, the cyanobacterium-derived AOM species may serve as polymer bridges between algae cells and Fe(III) hydroxides to aid coagulation (Bernhardt et al., 1985). However, the released IOM from M. aeruginosa cells showed a high ratio of protein to DOM (Henderson et al., 2008b), and the eOH and eCOOH groups within cyanobacteria protein can form protein-coagulant complexes with coagulants (Pivokonsky et al., 2006). These interactions inhibited the cross linking and clustering of Fe-hydroxide polymers and increased the coagulant demand accordingly. Thus, DOM showed both positive and adverse effects on the removal of algae by Fe coagulation, depending on the levels of DOM.

3.4. Proposed mechanisms involved in the removal of M. aeruginosa by KMnO4eFe(II) process Fig. 10 illustrates the schematic diagram of the possible mechanisms of KMnO4eFe(II) process on algae removal. KMnO4 pre-oxidation inactivated cells of M. aeruginosa and disturbed their integrity to benefit the algae removal, as indicated by the release of Kþ and IOM. Meanwhile, the MnO2 from KMnO4 reduction may precipitate onto algae surfaces,

45 Mn(VII)

Without Mn

Mn(VII)

1) Inactivation

30

MnOOH Algae

MnOOH Algae

+

0.83

1.67

3.33

6.67

Mn dose ( M) Fig. 8 e Comparison of MnO2 and KMnO4 on the removal of M. aeruginosa by Fe(III) coagulation. Cell density: 1.0 3 106 cells/mL. Fe dose: 197.4 mM. Pre-oxidation time: 5 min.

K , DOM

O2

4) Redox In situ-Fe(III)

5) Hydrolysis

2) Adsorption Mn(II)

3) Release

15

Fe(II)

DOM

Algae

DOM with lower level and higher MW

Moderate pre-oxidation

MnO2

Algae

Mn(II) Algae

Fresh Fe(III) polymer spheres with more active surface areas Continuous formation of In situ-Fe(III)

Fig. 10 e Schematic diagram of the possible mechanisms of KMnO4eFe(II) process on algae removal.

80

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 7 3 e8 1

even though solution pH was as high as 8.3. It is assumed that the co-existing Ca2þ acted as cation bridges to enable the complexation and adsorption of MnO2 onto the surfaces of algae cells, mainly through the eOH and eCOOH groups on the surfaces of their outside membranes. SEM/EDX analysis demonstrated the involvement of Ca2þ in interactions between MnO2 and algae cells (data not shown). After coagulation with the KMnO4 pre-oxidation at 0.83e13.3 mM, the elemental Ca did exist on the surfaces of algae cells with the atomic ratios of Ca to Mn ranging from 1.5:1 to 4:1. The aggregation of MnO2 onto algae surfaces increased the cell density, accelerated the settling of algae, and enabled the heterogeneous oxidation of Fe(II) to in-situ Fe(III) on algae surfaces. On the other hand, the moderate oxidation of algae cells achieved in KMnO4eFe(II) process played important roles. The subsequent introduction of Fe(II) ceased the inactivation of algae cells by KMnO4 to inhibit further IOM release and DOM degradation, as indicated by the less release of Kþ in KMnO4eFe(II) than that in KMnO4eFe(III) process (Table S1). Additionally, the released IOM in KMnO4eFe(II) process was in lower levels with less fractions of small molecular weight (Fig. S5), and this also indicated beneficial effects of moderate oxidation. Furthermore, at KMnO4 doses of 1.7e10.0 mM, the residual Mn2þ concentrations in KMnO4eFe(II) process were 2e18 times higher than those in KMnO4eFe(III) process (Fig. S2b). Mn2þ can also adsorb onto MnO2 (Forrez et al., 2010) and act as cation bridges to improve algae aggregation [Eqs. (10e12)].

Mn2þ þ MnO2 / MnO2 h Mn2þ

(10)

MnO2 h Mn2þ þ Cell / MnO2 h Mn2þ h Cell

(11)

and Duan, 2001) and thus dominates the flocs growth. In the one-off dosing strategy, Fe(III) immediately precipitates and grows to limited size during rapid-mixing with strong shear forces. In contrast, the continuous transformation of Fe(II) provides fresh Fe(III) and increases the discrete number of clusterecluster bond(s) and the magnitude of cohesive force to form stronger and larger flocs. Additionally, in the presence of soluble Fe, the formation of compact flocs and the precipitation of in-situ formed Fe(III) is in progress simultaneously to benefit particles aggregation. This is far to be well illustrated and needs to be studied to give more clear view in the removal of algae by KMnO4eFe(II) process.

4.

Conclusions

KMnO4eFe(II) process showed better capability toward algae removal than KMnO4eFe(III) process. This is attributed to the combined effects of moderate pre-oxidation and the continuous formation of in-situ Fe(III) in KMnO4eFe(II) process.After a period of pre-oxidation, the positive impacts of KMnO4 had been achieved, which would not increased with further prolonged pre-oxidation time. Thus, the introduction of Fe(II) at this time interval would avoid the extensive oxidation of algae cell and the thus the significant release of IOM, achieving moderate pre-oxidation of algae cells and facilitating the coagulation. Besides, the simultaneously formed in-situ Fe(III) was more effective than the pre-formed Fe(III) with respect to algae removal. This is attributable to that 1) in-situ formed Fe(III) had more reactive surface area and 2) that the continuous introduction of fresh coagulant [i.e. in-situ formed Fe(III)] benefited the floc growth than the one-off dosage of preformed Fe(III).

Acknowledgments MnO2 h Mn

2þ

þ DOM / MnO2 h Mn

2þ

h DOM

(12)

Moreover, the continuous formation of fresh Fe(III) and its hydrolysis products dominated in the remarkable efficiency toward algae removal via aforementioned mechanisms. The fresh Fe(III) hydrolysis products may not only form Fe-DOM complexes, but also attach and bind to the negativelycharged surfaces of algae cells. Additionally, the heterogeneous oxidation of Fe(II) by the MnO2 on algae surfaces, and the in-progressing Fe(III) precipitation achieved strong binding of flocs. The continuous provision of fresh Fe(III) also favored the aggregation of destabilized tiny flocs (i.e., algae cells, Fehydroxide, MnO2, DOM, OAP-Fe complexes) to form larger flocs during flocculation. The size distribution analysis of flocs at 4 and 17 min indicates the shift to the right of the major peak (Fig. 3). This supports the entrapment and/or adsorption of tiny flocs onto the surface of large ones and the continuous floc growth during floc growth phase. The significant finding of the present work is that continuous introduction of fresh coagulates is able to promote the growth of floc. Due to the limited solubility of Fe(III), the amorphous Fe(III) hydroxide plays an essential role (Gregory

This work was supported by the National Basic Research Program of China (Grant 2007CB407301), the National Natural Science Foundation of China (Grant No. 51078345), the Funds for the Creative Research Groups of China (Grant No. 50921064) and CAS Major Projects of Knowledge Innovation Program (kzcxl-yw-06-02).

Appendix. Supplementary data Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.10.022.

references

Bernhardt, H., Clasen, J., 1991. Flocculation of microorganisms. Journal of Water Supply: Research & Technology - AQUA 40 (2), 76e87. Bernhardt, H., Hoyer, O., Schell, H., Lusse, B., 1985. Reaction mechanisms involved in the influence of algogenic matter on flocculation. Z. Wasser Abwasser Forsch 18 (1), 18e30.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 7 3 e8 1

Chen, J.J., Yeh, H.H., 2005. The mechanisms of potassium permanganate on algae removal. Water Research 39 (18), 4420e4428. Chen, J.J., Yeh, H.H., Tseng, I.C., 2009. Effect of ozone and permanganate on algae coagulation removal - Pilot and bench scale tests. Chemosphere 74 (6), 840e846. Flynn, C.M.J., 1984. Hydrolysis of inorganic iron (III) salts. Chemical Reviews 84 (1), 31e41. Forrez, I., Carballa, M., Verbeken, K., Vanhaecke, L., Schlu¨sener, M., Ternes, T., Boon, N., Verstraete, W., 2010. Diclofenac oxidation by biogenic manganese oxides. Environmental Science & Technology 44 (9), 3449e3454. Gregory, J., Duan, J.M., 2001. Hydrolysing metal salts as coagulants. Pure and Applied Chemistry 73 (12), 2017e2026. Guan, X., Ma, J., Dong, H., Jiang, L., 2009. Removal of arsenic from water: effect of calcium ions on As(III) removal in the KMnO4eFe(II) process. Water Research 43, 5119e5128. Henderson, R., Parsons, S.A., Jefferson, B., 2008a. The impact of algal properties and pre-oxidation on solid-liquid separation of algae. Water Research 42 (8e9), 1827e1845. Henderson, R.K., Baker, A., Parsons, S.A., Jefferson, B., 2008b. Characterisation of algogenic organic matter extracted from cyanobacteria, green algae and diatoms. Water Research 42 (13), 3435e3445. Jolivet, J.P., Chane´ac, C., Tronc, E., 2004. Iron oxide chemistry. From molecular clusters to extended solid networks. Chemical Communications (5), 481e487. Lee, Y., Zimmermann, S.G., Kieu, A.T., Gunten, U.V., 2009. Ferrate (Fe(VI)) application for Municipal Wastewater treatment: a novel process for Simultaneous Micropollutant oxidation and phosphate removal. Environmental Science & Technology 43 (10), 3831e3838. Ma, J., Liu, W., 2002. Effectiveness and mechanism of potassium ferrate(VI) preoxidation for algae removal by coagulation. Water Research 36 (4), 871e878. Peterson, H.G., Hrudey, S.E., Cantin, I.A., Perley, T.R., Kenefick, S. L., 1995. Physiological toxicity, cell membrane damage and the release of dissolved organic carbon and geosmin by aphanizomenon flos-aquae after exposure to water treatment chemicals. Water Research 29 (6), 1515e1523. Petru sevski, B., van Breemen, A.N., Alaerts, G., 1996. Effect of permanganate pre-treatment and coagulation with dual coagulants on algae removal in direct filtration. Journal of

81

Water Supply: Research & Technology - AQUA 45 (5), 316e326. Pieterse, A.J.H., Cloot, A., 1997. Algal cells and coagulation, flocculation and sedimentation process. Water Science and Technology 36 (4), 111e118. Pivokonsky, M., Kloucek, O., Pivokonska, L., 2006. Evaluation of the production, composition and aluminum and iron complexation of algogenic organic matter. Water Research 40 (16), 3045e3052. Plummer, J.D., Edzwald, J.K., 2001. Effect of ozone on algae as precursors for trihalomethane and haloacetic acid production. Environmental Science & Technology 35 (18), 3661e3668. Plummer, J.D., Edzwald, J.K., 2002. Effects of chlorine and ozone on algal cell properties and removal of algae by coagulation. Journal of Water Supply: Research & Technology - AQUA 51 (6), 307e318. Posselt, H.S., Anderson, F.J., 1968. Cation sorption on colloidal hydrous manganese dioxide. Environmental Science & Technology 2 (12), 1087e1093. Rippka, R., Deruelles, J., Waterbury, J.B., Herdman, M., Stanier, R. Y., 1979. Generic assignments, strain histories and properties of pure cultures of cyanobacteria. Journal of General Microbiology 111, 1e61. Schmidt, W., Hambsch, B., Petzoldt, H., 1998. Classification of algogenic organic matter concerning its contribution to the bacterial regrowth potential and by-products formation. Water Science and Technology 37 (2), 91e96. Shen, H., Song, L.R., 2007. Comparative studies on physiological responses to phosphorus in two phenotypes of bloom-forming Microcystis. Hydrobiologia 592, 475e486. Steynberg, M.C., Pieterse, A.J.H., Geldenhuys, J.C., 1996. Improved coagulation and filtration as a result of morphological and behavioural changes due to pre-oxidation. Journal of Water Supply: Research & Technology - AQUA 45 (6), 292e298. Sukenik, A., Teltch, B., Wachs, A.W., Shelef, G., Nir, I., Levanon, D. , 1987. Effect of oxidants on microalgal flocculation. Water Research 21 (5), 533e539. Tan, S., Li, P., Chen, L., 2008. Test research for algae-removing by ultrafiltration. Journal of Water Resources & Water Engineering 19 (4), 73e75. Vikesland, P.J., Heathcock, A.M., Rebodos, R.L., Makus, K.E., 2007. Particle size and aggregation effects on magnetite reactivity toward carbon tetrachloride. Environmental Science & Technology 41 (15), 5277e5283.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 8 2 e9 2

Available online at www.sciencedirect.com

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

Influence of temperature and salinity on Ostreopsis cf. ovata growth and evaluation of toxin content through HR LC-MS and biological assays Laura Pezzolesi a, Franca Guerrini a, Patrizia Ciminiello b, Carmela Dell’Aversano b, Emma Dello Iacovo b, Ernesto Fattorusso b, Martino Forino b, Luciana Tartaglione b, Rossella Pistocchi a,* a b

Centro Interdipartimentale di Ricerca per le Scienze Ambientali, Universita` di Bologna, Via S’Alberto 163, 48123 Ravenna, Italy Dipartimento di Chimica delle Sostanze Naturali, Universita` degli Studi di Napoli “Federico II”, Via D. Montesano 49, 80131 Napoli, Italy

article info

abstract

Article history:

In the Mediterranean Sea, blooms of Ostreopsis cf. ovata and Ostreopsis siamensis have

Received 9 March 2011

become increasingly frequent in the last decade and O. cf. ovata was found to produce

Received in revised form

palytoxin-like compounds (putative palytoxin, ovatoxin-a, -b, -c, -d and -e), a class of

4 October 2011

highly potent toxins. The environmental conditions seem to play a key role in influencing

Accepted 16 October 2011

the abundance of Ostreopsis spp. High cell densities are generally recorded in concomitance

Available online 25 October 2011

with relatively high temperature and salinity and low hydrodynamics conditions. In this

Keywords:

ovata isolate were investigated. The highest growth rates of the Adriatic strain were

study the effects of temperature and salinity on the growth and toxicity of an Adriatic O. cf. Ostreopsis cf. ovata

recorded for cultures grown at 20
C and at salinity values of 36 and 40, in accordance with

Temperature

natural bloom surveys. Toxicity was affected by growth conditions, with the highest toxin

Salinity

content on a per cell basis being measured at 25
C and salinity 32. However, the highest total toxin content on a per litre basis was recorded at 20
C and

Toxicity Ovatoxins Haemolysis assay

salinity 36, since under such conditions the growth yield was the highest. O. cf. ovata had lethal effects on Artemia nauplii and juvenile sea basses, and produced haemolysis of sheep erythrocytes. A comparison between haemolysis neutralization assay and HR LC-MS results showed a good correlation between haemolytic effect and total toxin content measured through HR LC-MS. Considering the increasing need for rapid and sensitive methods to detect palytoxin in natural samples, the haemolytic assay appears a useful method for preliminary quantification of the whole of palytoxin-like compounds in algal extracts. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Massive blooms of the benthic dinoflagellates Ostreopsis spp. are reported worldwide in many tropical and temperate regions (Faust et al., 1996; Vila et al., 2001; Aligizaki and

Nikolaidis, 2006; Chang et al., 2000; Ciminiello et al., 2008; Mangialajo et al., 2011; Rhodes et al., 2011). In the Mediterranean Sea, blooms of O. cf. ovata and Ostreopsis siamensis have been reported since the late ‘70s (Taylor, 1979; AbboudAbi Saab, 1989) but, in the last decade, they have become

* Corresponding author. Tel.: þ39 (0) 544 937376; fax: þ39 (0) 544 937411. E-mail address: [email protected] (R. Pistocchi). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.029

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 8 2 e9 2

increasingly frequent and resulted in benthic biocenosis sufferings and human health problems. Ostreopsis spp., typically, proliferate in shallow and sheltered waters, with low hydrodynamism; they form a rustybrown coloured mucilaginous film, which covers reefs, rocks (Bottalico et al., 2002), and soft sediments (Vila et al., 2001) as well as seaweeds (Vila et al., 2001; Bottalico et al., 2002; Aligizaki and Nikolaidis, 2006; Totti et al., 2010), marine angiosperms, and invertebrates (Bianco et al., 2007; Totti et al., 2007). The presence of Ostreopsis spp. in coastal waters may pose a real threat to coastal food web and fishery (Aligizaki et al., 2008). Several marine organisms, in particular sea urchins, have lost their spines and died during blooms of O. cf. ovata or O. siamensis (Grane´li et al., 2008; Sansoni et al., 2003; Shears and Ross, 2009); however, the effects on marine organisms and on ecosystem dynamics are still unknown. Ostreopsis spp. are thought to produce palytoxin (or its analogues) (Taniyama et al., 2003), one of the most potential toxic marine compounds, which acts on the Naþ/Kþ pump converting it into an ionic channel and causing the subsequent depletion of the Kþ ions (Habermann, 1989). This hypothesis was later supported by identification of putative palytoxin as the causative toxin of human poisonings which occurred during O. siamensis blooms (Onuma et al., 1999) and, most importantly, by identification of some palytoxin-like compounds from various Ostreopsis spp. Particularly, ostreocin-D was isolated from O. siamensis and structurally elucidated by NMR (Usami et al., 1995; Ukena et al., 2001) while mascarenotoxins were identified, basing only on MS evidence, as palytoxin-like compounds from Ostreopsis mascarenensis (Lenoir et al., 2004). Putative palytoxin and ovatoxin-a were detected in field and cultured samples of O. cf. ovata, collected along the Ligurian coasts (Italy) (Ciminiello et al., 2006; 2008) as well as in O. cf. ovata cultures from the Adriatic and Tyrrhenian Sea (Guerrini et al., 2010) by liquid chromatographymass spectrometry (LC-MS). Recently, several new ovatoxins, namely ovatoxin-b, -c, -d, and -e, were also detected in an Adriatic O. cf. ovata culture through an in-depth high resolution (HR) LC-MS investigation (Ciminiello et al., 2010). Currently, O. cf. ovata blooms occur each year from June to late October at several sites on the Italian coastline, characterized by different environmental conditions, such as seawater temperature in the range 18e30
C and salinity in the range 30e39 (Pistocchi et al., 2011). However, O. cf. ovata has never been detected in the Northwestern Adriatic sea, at sites located close to the Po river delta, where low salinity values occur and a coast-offshore salinity gradient affecting microphytobenthos distribution in the northern Adriatic Sea was observed (Totti, 2003); this suggests that some environmental conditions play a key role in influencing O. cf. ovata growth and/or its geographical dispersal. Several authors indicated seawater temperature is an important factor affecting cell proliferation (Tognetto et al., 1995; Sansoni et al., 2003; Simoni et al., 2004; Aligizaki and Nikolaidis, 2006; Mangialajo et al., 2008). In most studies (as reviewed by Pistocchi et al., 2011), high temperature values (24e29
C) were associated with increases of Ostreopsis cell number in seawater; however, in the Adriatic (Totti et al., 2010) and Catalan seas (Vila et al., 2001) such positive

83

correlation has not been observed. Recently, the influence of temperature on O. cf. ovata growth and toxicity has been also reported by Grane´li et al. (2011) using a Tyrrhenian isolate from the Ligurian coast: the highest toxicity was found in cultures grown at 20
C, whereas the highest algal biomass was recorded at 30
C. In the present study, we report on in-depth investigation on the effect of some environmental parameters on the growth and toxicity of O. cf. ovata. An Adriatic O. cf. ovata isolate, whose growth and toxin profile had been previously characterized at 20
C and salinity 36 during the exponential and stationary phases (Guerrini et al., 2010), was used. Cultures were grown at different temperature (20, 25 and 30
C) and salinity values (26, 32, 36 and 40); HR LC-MS analyses were carried out to determine their toxin profile, including the recently found ovatoxins (Ciminiello et al., 2010), and to evaluate the total toxin amount released in the extracellular medium during the stationary growth phase. A further object of the present study was to compare the total toxin content of algal extracts measured by HR LC-MS with the results obtained through the haemolysis assay (Riobo´ et al., 2008), with the aim of gaining information on the accuracy of the haemolytic test, a rapid and very sensitive biological assay widely employed for palytoxins detection (Riobo´ et al., 2011). Finally, the toxicity of O. cf. ovata cultures on crustaceans and fish was also investigated using Artemia sp. assay and the ichthyotoxicity test with juvenile sea basses (Dicentrarchus labrax) (IRSA-CNR, 2003) to evaluate the potential O. cf. ovata impact on the other marine organisms.

2.

Materials and methods

2.1.

Cultures of Ostreopsis cf. ovata

O. cf. ovata was isolated using the capillary pipette method (Hoshaw and Rosowski, 1973) from water samples collected along the Adriatic coast of Italy (Marche region, Numana sampling site, strain OOAN0601) in October 2006, in proximity to the seaweeds Cystoseira sp. and Alcidium corallinum. After an initial growth in microplates, cells were cultured at 20
C under a 16:8 h L:D cycle from cool white lamp in natural seawater, at salinity 36, adding macronutrients at a five-fold diluted f/2 concentration (Guillard, 1975) and selenium. In order to evaluate the effect of environmental parameters on growth and toxicity of O. cf. ovata, temperature and salinity experiments were carried out. In the salinity experiment, cultures (at 20
C) were established at salinity 26, 32, 36 and 40 in a thermostatic room, maintaining light irradiance at 100e110 mmol m2 s1. Salinity levels 26, 32 and 36 were obtained by diluting seawater (salinity 38) with deionized water, while salinity 40 was obtained by evaporation of the seawater. In the temperature experiment, cultures (salinity 36) were established at 20, 25 and 30
C in water baths kept in the same thermostatic room, thus light irradiance slightly decreased to 90 mmol m2 s1. Phaeodactylum tricornutum (strain PTN0301 from the North Sea, Holland) was cultured using f/2 medium under the same conditions and used in the experiments either for comparisons or control.

84

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 8 2 e9 2

Both temperature and salinity experiments were carried out by using, for each condition, 2 series of batch cultures. One was used to evaluate the growth profile and the other the toxin content.

2.1.1.

Evaluation of growth profile

Since the evaluation of the growth profile of O. cf. ovata in batch cultures was complicated by the presence of mucous aggregates, the sampling method developed by Guerrini et al. (2010) was used for counting. For each temperature/salinity level, 15 Erlenmeyer flasks containing 200 ml of culture were grown in parallel; every other day, two out of the initial flasks were treated with HCl to a final concentration of 4 mM. Acid addition dissolved mucous aggregates and homogenous sampling could be performed. After counting, the two acidified flasks were discarded. Cell counts were made following Utermo¨hl method (Hasle, 1978) and specific growth rate (m, day1) was calculated using the following equation: m¼

ln N1  ln N0 t1  t0

where N0 and N1 are cell density values at time t0 and t1. Calculation of cell volume was made with the assumption of ellipsoid shape using the following equation (Sun and Liu, 2003): 

V ¼ ðp=6Þ a b c where a ¼ dorsoventral diameter (length), b ¼ width, c ¼ mean anteriorposterior diameter (height).

2.1.2.

Evaluation of toxin content

For each temperature and salinity level, a set of four culturing flasks was set up. Due to limitations in the availability of the equipment, the salinity experiment was carried out in a thermostatic room using 1500 mL flasks, while the temperature experiment was carried out by placing the 800 mL flasks in water baths. Cell counting was carried out on one out of the four flasks as described above. Five replicate counts were collected from one of the four flasks for each treatment and used to determine the cell density and to express toxin content on a per cell basis. Culture collection was carried out during the late stationary growth phase by gravity filtration through GF/F Whatman (0.7 mm) filters at day 21st and 25th for the salinity and temperature experiment, respectively. Cell pellets and growth media for each temperature/salinity level were provided for chemical analysis.

2.2.

Chemical analysis

2.2.1.

Chemicals

All organic solvents were of distilled-in-glass grade (Carlo Erba, Milan, Italy). Water was distilled and passed through a MilliQ water purification system (Millipore Ltd., Bedford, MA, USA). Acetic acid (Laboratory grade) was purchased from Carlo Erba. Analytical standard of palytoxin was purchased from Wako Chemicals GmbH (Neuss, Germany).

2.2.2.

Extraction

Cell pellets and growth media for each temperature/salinity level were extracted separately. For each pellet sample 9 mL of a methanol/water (1:1, v/v) solution was added and the solution sonicated for 30 min in pulse mode, while cooling in ice bath. The mixture was centrifuged at 3000 g for 30 min, the supernatant was decanted and the pellet was washed twice with 9 mL of methanol/water (1:1, v/v). The extracts were combined and the volume was adjusted to 30 mL with extracting solvent. The obtained mixture was analyzed directly by HR LC-MS (5_ml injected). Each growth medium was extracted five times with an equal volume of butanol. The butanol layer was evaporated to dryness, dissolved in 5 mL of methanol/water (1:1, v/v) and analyzed directly by HR LC-MS (5 ml injected). Recovery percentage of the above extraction procedures were estimated to be 98% and 75% for the pellet and growth medium extracts, respectively (Ciminiello et al., 2006).

2.2.3. High resolution liquid chromatography-mass spectrometry (HR LC-MS) High resolution (HR) LC-MS experiments were carried out on an Agilent 1100 LC binary system (Palo Alto, CA, USA) coupled to a hybrid linear ion trap LTQ Orbitrap XL Fourier Transform MS (FTMS) equipped with an ESI ION MAX source (ThermoFisher, San Jose`, CA, USA). Chromatographic separation was accomplished by using a 3 mm gemini C18 (150  2.00 mm) column (Phenomenex, Torrance, CA, USA) maintained at room temperature and eluted at 0.2 mL min1 with water (eluent A) and 95% acetonitrile/water (eluent B), both containing 30 mM acetic acid. A slow gradient elution was used: 20e50% B over 20 min, 50e80% B over 10 min, 80e100% B in 1 min, and hold 5 min. This gradient system allowed a sufficient chromatographic separation of most palytoxin-like compounds (Table 1). HR full MS experiments (positive ions) were acquired in the range m/z 800e1400 at a resolving power of 15,000. The

Table 1 e Molecular formulae (M) of ovatoxins, elemental composition of their relevant A and B moieties and most abundant peaks of [M D 2HeH2O]2D and [M D 2H D K]3D ion clusters for each compound. Toxin Palytoxin Ovatoxin-a Ovatoxin-b Ovatoxin-c Ovatoxin-d Ovatoxin-e

Rt (min)

M

A moiety

B moiety

[M þ 2HeH2O]2þ

[M þ 2H þ K]3þ

10.78 11.45 11.28 10.90 11.07 11.07

C129H223N3O54 C129H223N3O52 C131H227N3O53 C131H227N3O54 C129H223N3O53 C129H223N3O53

C16H28N2O6 C16H28N2O6 C18H32N2O7 C18H32N2O7 C16H28N2O6 C16H28N2O7

C113H195NO48 C113H195NO46 C113H195NO46 C113H195NO47 C113H195NO47 C113H195NO46

1331.7436 1315.7498 1337.7623 1345.7584 1323.7456 1323.7456

906.8167 896.1572 910.8318 916.1628 901.4884 901.4884

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 8 2 e9 2

following source settings were used in all HR LC-MS experiments: a spray voltage of 4 kV, a capillary temperature of 290
C, a capillary voltage of 22 V, a sheath gas and an auxiliary gas flow of 35 and 1 (arbitrary units). The tube lens voltage was set at 110 V. Due to commercial availability of the only palytoxin standard, quantitative determination of putative palytoxin, ovatoxin-a, -b, -c, -d, and -e in the extracts was carried out by using a calibration curve (triplicate injection) of palytoxin standards at four levels of concentration (25, 12.5, 6.25, and 3.13 ng mL1) and assuming that their molar responses were similar to that of palytoxin. Extracted ion chromatograms (XIC) for palytoxin and each ovatoxins were obtained by selecting the most abundant ion peaks of both [Mþ2HeH2O]2þ and [Mþ2H þ K]3þ ion clusters (Table 1). A mass tolerance of 5 ppm was used.

2.3.

Toxicity assays

2.3.1.

Artemia sp. assay

The assay was carried out according to the short-term test of the IRSA-CNR (2003) method, consisting in a 24 h exposure of Artemia sp. to the potentially toxic sample. 10 nauplii were incubated in 1 mL of sample in a glass tube for 24 h. Firstly, aliquots of a culture grown at 20
C and salinity 36, containing five increasing concentrations of live cells, lysed cells, algal extracts and growth media, were tested in triplicate. Live cell aliquots were sampled during the stationary phase of the culture. Lysed cell aliquots were obtained by sonicating 10 mL of the culture for 3 min. Algal extracts were obtained as reported above and diluted (1:100 to 1:10,000) with seawater. A palytoxin stock solution (12.5 mg mL1) in methanol/water (1:1, v/v) was diluted with seawater and tested in the concentration range 500e10,000 pg mL1. Growth medium aliquots were obtained by filtering 50 mL of the culture through GF/F Whatman (0.7 mm) filters. The toxicity of O. cf. ovata cultures grown at different temperature/salinity conditions was evaluated by Artemia sp. assay, using only live cells. Five different concentration levels of each sample were obtained through dilution with seawater, and were tested in triplicate. The effects on Artemia sp. of sample exposure were evaluated after 24 h by counting the number of dead organisms. Seawater samples, methanol/ water (1:1, v/v) solution (diluted 1:100, v/v with seawater) and f/2 medium at the investigated salinity levels (diluted 1:5 with seawater) were used as control. EC50 values were calculated (see below Section 2.4).

2.3.2.

Haemolytic assay

Haemolytic assay was carried out following the procedure proposed by Bignami (1993) and modified by Riobo´ et al. (2008). The test is based on photometrical determination of haemoglobin released from sheep erythrocytes following exposure to haemolytic compounds. Sheep blood was kindly provided by the Department of Veterinary Public Health and Animal Pathology (University of Bologna). Erythrocytes were separated from plasma by centrifugation (400  g at 10
C for 10 min) and washed twice with a solution containing phosphate buffered saline (PBS) 0.01 M, pH 7.4, bovine serum albumin (BSA), calcium chloride (CaCl2 2H2O) 1 mM and boric

85

acid (H3BO3) 1 mM. Finally, the erythrocytes solution was diluted with PBS at a final concentration of 1.7  108 red cells mL1. According to the reported method (Riobo´ et al., 2008), two blood solutions, one added of ouabain (2.5 mM) and one ouabain-free, were prepared to a final concentration of 1.7  107 erythrocytes. 1 mL of each blood solution was mixed with 1 mL of the sample diluted in PBS (either pellet extract or palytoxin standard previously dissolved in methanol/water (1:1, v/v)) and incubated at 25
C for 20 h. After the incubation, samples were centrifuged at 400  g for 10 min and the supernatant absorptions were measured at 405 nm. Two replicates of algal extract at different concentration levels, control solutions for blanks (PBS buffer and methanol/water (1:1, v/v) in PBS) and total haemolysis sample were prepared for each experiment. Palytoxin standard at seven concentration levels (4e196 pg mL1) were used for generating the calibration curve. Stock solutions of the algal extracts and palytoxin standard used in the haemolytic assay were quantified by HR LC-MS. The haemolytic effects of the algal extracts were expressed either on cell basis (cell mL1) or on toxin content basis (pg mL1). EC50 values obtained by testing the palytoxin standard and the algal extracts were calculated (see below Section 2.4).

2.3.3.

Fish bioassay

Sea basses (D. labrax) employed in the assay were collected from the hatchery Valle Ca’ Zuliani (Pila di Porto Tolle, Rovigo, Italy). After the transfer, they were kept in a 60-70 L aquarium, aerated by a small dispenser (Hailea) and kept at room temperature and salinity 36 for one month. For the experiments, 2 L aerated tanks containing algal culture were used. Three juveniles (5.0  1.0 g) were put into each tank, kept at 20
C, during a 16:8 h lightedark period and observed for 4 days. Two replicates of O. cf. ovata culture grown for 4e6 days at 20
C and salinity 36 were tested at three concentration levels. An equal volume of P. tricornutum culture was used as control. Fish were considered dead when gill opercular movements ceased.

2.4.

Data analysis

The 50% effect concentration (EC50) of each sample for the Artemia sp. and haemolytic assays was estimated by fitting the experimental concentration-response curves to a logistic model: y¼

a  x b 1þ EC50

Where: y ¼ endpoint value; x ¼ substance concentration; a ¼ expected endpoint value in absence of toxic effect; b ¼ slope parameter. The parameters of the equation, including the EC50, were estimated using the non-linear regression procedures implemented in Statistica (Statsoft, Tulsa, OK, USA). An independent estimate of EC50 was obtained for each of the experiments. Differences in cell biovolume, EC50 value, and toxins concentration among the samples were tested by using the multivariate analysis-of-variance (ANOVA) test, using

86

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 8 2 e9 2

Statistica (StatSoft, Tulsa, OK, USA) software. Whenever a significant difference for the main effect was observed (P < 0.05), a NewmaneKeuls test was also performed.

3.

Results

Batch cultures of an Adriatic strain of O. cf. ovata, collected along the Marche coasts of Italy (Numana sampling site) in October 2006, were established in order to evaluate the effect of salinity and temperature on algal growth and toxin profile. Particularly, in the temperature experiment, cultures were set at 20, 25 and 30
C by maintaining salinity at 36 and light irradiance at 90 mmol m2 s1, while in the salinity experiment cultures were established at salinity 26, 32, 36 and 40, by maintaining temperature at 20
C and light irradiance at 100e110 mmol m2 s1.

3.1.

Growth pattern and cell volume

The growth profiles of O. cf. ovata cultures exposed to different salinity and temperature values were analyzed by measuring the cell density every 2e3 days from the beginning of the exponential phase to the end of the stationary phase (Fig. 1A and B). Under the different growth conditions, O. cf. ovata growth rates in the range 0.34e0.49 day1 were observed. For the temperature experiment, during the first 5 days cells grew better at 25
C; at the end of the exponential phase the maximum growth rate of 0.49 day1 was recorded at 20
C, followed by 0.43 and 0.34 day1 at 25 and 30
C, respectively. For the salinity experiment (carried out at 20
C) growth rate was not significantly affected by the salt concentration (0.43e0.47 day1) (ANOVA, P > 0.05). In the stationary phase, the maximum density was 13,000e16,000 cell mL1 at 20
C and intermediate salinities

(32 and 36), while the cell yield dropped to 7500 cell mL1 both at salinity 26 (temperature 20
C; Fig. 1A) and temperature 30
C (salinity 36; Fig. 1B). In the course of the experiments, we noticed that the culture volume played a key role on the final cell yield: decreasing cell densities were obtained as culture volumes increased from 200 mL to 800 mL up to 1500 mL. Another aspect we considered in the salinity and temperature experiments was the cell biometric measurement. It is to be noted that O. cf. ovata cells appeared highly different both in size and in shape, within each cell culture; therefore, a statistically significant cell number (n  50) was used for estimating the mean biovolumes. In both salinity and temperature experiments, a significant difference was observed between cell volumes measured in the exponential and stationary phases (ANOVA, P < 0.001). For the salinity experiment, the highest difference among biovolumes was observed in the exponential phase where a mean value of 22,000 mm3 was reached at the lowest salinity (26) and resulted significantly higher (Post-hoc SNK test, P < 0.001) than those observed at 36 and 40 (14,000 and 13,000 mm3, respectively). An intermediate biovolume mean value was observed at salinity 32 (17,000 mm3). In the stationary phase, cells were more homogenous in size, with cell volumes in the range 28,000e30,000 mm3; however, the value reported at salinity 32 (22,400 mm3) resulted significantly lower than those observed at the other salinity levels (Posthoc SNK test, P < 0.001). For the temperature experiment, in both growth phases cell volumes decreased as temperature increased, with a maximum biovolume of 22,000 mm3 being reached at 20
C (stationary phase), which was significantly higher (Post-hoc SNK test, P < 0.001) than biovolumes measured at 25 and 30
C (16,000 and 15,000 mm3, respectively).

-1

Abundance (cell mL )

3.2. 100000

26 psu 32 psu 36 psu 40 psu

A

10000 1000

100 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30

-1

Abundance (cell mL )

Time (day)

B

100000

20°C 25°C 30°C

10000

1000

100 0

2

4

6

8 10 12 14 16 18 20 22 24 26

Time (day) Fig. 1 e Growth pattern of O. cf. ovata cultures exposed to different salinity (A) and temperature (B) conditions.

Determination of toxin content by HR LC-MS

Cell pellets and growth media of O. cf. ovata cultures grown at the different temperature and salinity values were collected during the late stationary growth phase. Samples were extracted separately, and the crude extracts were used to evaluate the toxin profile. HR LC-MS experiments were acquired in full MS mode by using an LC method which allowed chromatographic separation of the major components of the toxin profile. The spectra were acquired in the mass range m/z 800e1400 where each palytoxin-like compound (Table 1) produces bi-charged ions due to [M þ H þ K]2þ, [M þ H þ Na]2þ, and [Mþ2H]2þ, tri-charged ions due to [Mþ2H þ K]3þ and [Mþ2H þ Na]3þ, and a number of ions due to multiple water losses from the [Mþ2H]2þ and [Mþ3H]3þ ions. The presence of putative palytoxin and of all the ovatoxins (ovatoxin-a, -b, -c, -d, and -e) recently identified in O. cf. ovata (Ciminiello et al., 2008; 2010) was highlighted in all the analyzed samples. Significant differences were observed in the total toxin content of different algal extracts (ANOVA, P < 0.001), whereas the relative abundance of individual toxins were quite similar: ovatoxin-a was by far the major component of the toxin profile (47e56% of the total toxin content; Post-hoc SNK test,

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 8 2 e9 2

Table 2 e Total toxin content (putative palytoxin, ovatoxin-a, -b, -c, -d, and -e) of O. cf. ovata culture pellet and medium extracts, measured by HR LC-MS in both salinity and temperature experiment. Data are expressed as mg per Litre of culture (mg LL1). Cell density (cell LL1) and extracellular release (%) are also reported. Total toxin content (ug L1) Cell L

1

Extracellular release (%)

Pellet

Medium

Total

3,450,333 4,646,333 4,281,333 5,619,000

57 95 76 68

17 14 12 11

74 109 88 80

23 13 14 14

Temperature 9,869,587 20
C 5,581,677 25
C 4,493,377 30
C

155 129 81

25 25 30

180 154 111

14 16 27

Salinity 26 32 36 40

P < 0.001), followed by ovatoxin-b (24e27%), ovatoxin-d and -e (15e18%), ovatoxin-c (4e8%) and putative palytoxin (0.5e3%) on the basis of their decreasing relative abundance. Total toxin content of pellet and medium extracts expressed as mg L1 culture in both salinity and temperature

A

palytoxin ovatoxin-a ovatoxin-b ovatoxin-c ovatoxin-d,-e tot

20

-1

Toxin content (pg cell )

25

10 5

26

32

36 pellet

40

26

32 36 medium

40

Salinity (psu) 25

-1

experiments are reported in Table 2. Toxin contents were significantly higher in the cell pellet than in the corresponding culture medium (Post-hoc SNK test, P < 0.001), resulting in relatively low release percentages (13e16%) in most of the growth conditions applied; however, the release increased up to 23 and 27% under the most unfavourable growth conditions, namely salinity 26 (temperature 20
C) and temperature 30
C (salinity 36), respectively. Total and individual toxin contents on a per cell basis (pg cell1), are reported in Fig. 2A and B for salinity and temperature experiments, respectively. Small differences in total toxin content were observed between the experiments that should have provided similar results, namely the cultures grown at temperature 20
C and salinity 36. Such differences could be due to the slightly different growth conditions, among which the difference in light intensity and in culture volume could have played a major role. In the salinity experiment, total toxin content reached the highest value in the culture grown at 32 (20 pg cell1) and the lowest at 40 (12 pg cell1). For the temperature experiment, O. cf. ovata grown at 25
C was found to have a total toxin content of 23 pg cell1, while cultures grown at 20 and 30
C produced 16 and 18 pg cell1, respectively. This last finding apparently is not in agreement with the maximum concentration (mg L1) observed at 20
C (salinity 36), which was indeed affected by the high cell yield of the culture (Table 2). Particularly, culture grown at 20
C showed a cell density almost two-fold higher than the others.

3.3.

Haemolytic assay

15

0

Toxin content (pg cell )

87

B

palytoxin ovatoxin-a ovatoxin-b ovatoxin-c ovatoxin-d,-e tot

20 15 10 5 0 20

25 pellet

30

20

25 medium

30

Temperature (°C)

Fig. 2 e Total and individual toxin contents of putative palytoxin, ovatoxin-a, -b, -c, -d, and -e of O. cf. ovata cultures grown under different salinity (A) and temperature (B) conditions. HR LC-MS measurements (pg cellL1) were carried out for both pellet and medium extracts at the end of stationary growth phase.

All the O. cf. ovata culture extracts investigated in the present study were tested by haemolytic assay and the results, expressed as haemolysis percentage versus cell number present in 1 mL of assay solution (cell mL1), are reported in Fig. 3. All showed a strong delayed haemolysis of sheep erythrocytes, which was specifically inhibited by ouabain, even at concentrations corresponding to very low cell densities; however, a percentage of not-specific haemolysis was left over even in the presence of ouabain as shown in Fig. 3 (dotted lines). The haemolytic activity of cultures grown at different salinity levels (Fig. 3A) followed a pattern similar to that measured by HR LC-MS; particularly the highest haemolysis (83%) was observed for the culture grown at salinity 32 (total toxin content ¼ 20 pg cell1) followed by cultures grown at salinity 26 (haemolysis 79%, total toxin content ¼ 16 pg cell1), 36 (haemolysis 76%, total toxin content ¼ 18 pg cell1), and 40 (haemolysis 74%, total toxin content ¼ 12 pg cell1). For the temperature experiment (Fig. 3B) cells grown at 20
C reported the lowest haemolytic activity (haemolysis 76%, total toxin content ¼ 16 pg cell1) in agreement with HR LC-MS results, while an 82% haemolytic effect was observed for both cultures grown at 25 and 30
C, despite the slightly different toxin content of 23 and 18 pg cell1, respectively. All the above data were consistent with HR LC-MS results expressed as pg cell1. A comparison of the results of the haemolytic assay with the quantitative results achieved by HR LC-MS could provide useful information about the haemolytic activity of ovatoxins

Haemolysis (%)

88

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 8 2 e9 2

100 90 80 70 60 50 40 30 20 10 0 1

activity of the overall ovatoxins is quite similar to that of palytoxin standard. Values obtained for the palytoxin standard were interpolated using a non-linear estimation curve, described by the reported equation (f1, Fig. 4). The resulting EC50 values for palytoxin standard and algal extracts were not significantly different (ANOVA, P > 0.05), being 22  2 and 25  8 pg mL1, respectively.

26 psu 26 psu+OUA 32 psu 32 psu+OUA 36 psu 36 psu+OUA 40 psu 40 psu+OUA

A

2

3

4

5

6

7

8

Haemolysis (%)

100 90 80 70 60 50 40 30 20 10 0

B

0

1

20°C 20°C+OUA 25°C 25°C+OUA 30°C 30°C+OUA

2

3

4

5

6

7

8

9

Artemia sp. assay

3.4.

-1

Concentration (cell mL )

10

-1

Concentration (cell mL ) Fig. 3 e Haemolytic activity of O. cf. ovata extracts grown at different salinity (A) and temperature (B) conditions on sheep erythrocytes in absence (solid lines) and in presence (dashed lines) of ouabain (OUA). Data are expressed as haemolysis percentage (%) versus cell number mLL1 assay (cell mLL1).

in comparison with that of palytoxin. The haemolytic activity of the algal extracts from salinity and temperature experiments were also expressed as haemolysis percentage versus concentration of pg total toxin contained in 1 mL assay solution (pg mL1) as measured by HR LC-MS. These data are compared in Fig. 4 with those obtained for the haemolytic activity of palytoxin standard tested at seven different concentrations. This clearly suggests that the haemolytic

Artemia sp. assays were carried out using both live and lysed cells of O. cf. ovata cultures as well as the algal extract and the growth medium of a culture grown at 20
C and salinity 36. O. cf. ovata live cells induced rapid and high mortality of Artemia sp. nauplii, even at low cell concentrations. From the EC50 values of all the samples calculated at 24 h (Table 3), cell toxicity appeared relevant and significantly different (ANOVA, P < 0.001): the growth medium resulted significantly less toxic than the live cells (Post-hoc SNK test, P < 0.001), with an EC50 value of 720 cell mL1 versus 8 cell mL1, respectively. This result confirmed the presence of small amounts of toxins released in the growth medium. The lysed cells induced a similar mortality as the algal extract, as evidenced by the comparable EC50 values (Post-hoc SNK test, P > 0.05). EC50 values of all the tested O. cf. ovata samples were expressed also as pg of toxins per mL assay (pg mL1) (Table 3), basing on the total toxin contents measured by HR LC-MS. A palytoxin standard at five levels of concentrations (500e10,000 pg mL1) was also tested; it presented an EC50 value of 4606 pg mL1, which was significantly higher (Posthoc SNK test, P < 0.001) than that of the algal extract (1146 pg mL1). Administration of live O. cf. ovata cells, grown at different salinity and temperature conditions, to Artemia sp. resulted in no significant differences among the measured EC50 values (ANOVA, P > 0.05). For the temperature experiment the lowest EC50 value was measured for live cells grown at 25
C (EC50 ¼ 6  2 cell mL1) compared to those grown at 20 and palytoxin standard algal extracts palytoxin standard+OUA algal extracts+OUA

100

Haemolysis (%)

90 80

f1

70 60 50 40 30 20 10 0 0

25

50

75

100

125

150

175

200

-1

Concentration (pg mL )

Function (f1) PLTX standard

y = (85.5819)-(85.5819)/(1+(x/(21.473))^(log(9)/(log((254.169)/(21.473)))))

Fig. 4 e Haemolytic activity of O. cf. ovata extracts and palytoxin standard on sheep erythrocytes in absence and in presence of ouabain (OUA). Data are expressed as haemolysis percentage (%) versus concentration of palytoxin equivalent per mL of assay (pg mLL1). Equation f1: non-linear estimation curve obtained for the palytoxin standard.

89

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 8 2 e9 2

et al., 2011). Since very few laboratory studies on the effects of environmental parameters on growth and toxicity of Ostreopsis isolates have been reported (Grane´li et al., 2011; Ashton et al., 2003; Morton et al., 1992), we carried out a detailed study on an Adriatic strain of O. cf. ovata grown at different temperature and salinity conditions.

Table 3 e The 50% mortality on Artemia nauplii (EC50) is expressed both as cells of O. cf. ovata per mL (cell mLL1) and as total toxin content per mL (pg mLL1) basing on the HR LC-MS quantification. Values are reported for O. cf. ovata live and lysed cells, extract, and growth medium. EC50 value obtained for palytoxin standard is reported as pg mLL1. Each value is the mean of three replicates ± standard error. EC50 (cell mL1) 85 96  6 80  7 720  54 e

O. cf. ovata live cells O. cf. ovata lysed cells O. cf. ovata extract O. cf. ovata medium Palytoxin standard

4.1.

EC50 (pg mL1) 115 1376 1146 1822 4606

    

The Adriatic O. cf. ovata strain was tolerant to salinity variation in the range 26e40. Very similar growth rates and yields were observed within the tested salinity range, with the lowest growth yield being recorded at salinity 26. This was in good agreement with field measurements performed during Mediterranean O. cf. ovata blooms (Totti et al., 2010; Monti et al., 2007) as well as with results of a survey of epiphytic dinoflagellates along the Hawaiian coast, (Parsons and Preskitt, 2007) in which O. cf. ovata was the only dinoflagellate to be negatively correlated with salinity. In the temperature experiment, the analyzed Adriatic O. cf. ovata strain reached the highest growth yield at 20
C, whereas the lowest yield was recorded at 30
C. Our results are in good agreement with field surveys in the Adriatic Sea, where O. cf. ovata proliferation occurs from the end of August to October, when water temperature is about 20e22
C (Totti et al., 2010; Monti et al., 2007). On the contrary, Grane´li et al. (2011) indicated, for a Tyrrhenian O. cf. ovata strain, that high water temperatures (26e30
C) increased both growth rate and yield; this is consistent with the field surveys reporting O. cf. ovata blooms in the Tyrrhenian Sea in the middle of the summer. Our results and those observed by Grane´li et al. (2011) suggest that Adriatic and Tyrrhenian strains are differently affected by temperature. For the morphometric characters, in both salinity and temperature experiments, a certain cell size variability was observed; however, the cell volumes reported under the different growth conditions did not show a specific pattern, particularly in the stationary growth phase. A marked variability in the biovolumes of O. cf. ovata cells from the same culture was observed (Guerrini et al., 2010), and is in agreement with field observations (Aligizaki and Nikolaidis, 2006; Bianco et al., 2007).

72 86 272 137 781

30
C (EC50 ¼ 11  3 and 14  1 cell mL1, respectively). For the salinity experiment the lowest EC50 value was obtained at salinity 32 (EC50 ¼ 9  2 cell mL1), while cells grown at the other salinities reported values of 17  3, 24  8, and 17  4 cell mL1 for salinity 26, 36 and 40, respectively. This appears in good agreement with total toxin contents (pg cell1) measured by HR LC-MS (Fig. 4).

3.5.

Fish bioassay

Table 4 shows the results of the ichthyotoxic assay performed with different concentrations of O. cf. ovata live cells. Sea bass mortality occurred after 1 day of exposure, only at the highest O. cf. ovata cell density (2367 cells mL1). Before dying, loss of balance and fish floating was observed. After 45 h from the beginning of the assay, even fish exposed to lower algal concentrations (1138 and 425 cells mL1) began to die and they were all dead after 72 h. Fish exposed to the control diatom P. tricornutum (789, 100 cells mL1) survived and behaved normally till the end of the experiment (96 h).

4.

Growth and cell size pattern

Discussion

Several field surveys have indicated that environmental conditions play a major role in determining Ostreopsis spp. proliferation (reviewed by Mangialajo et al., 2011 and Pistocchi

Table 4 e Toxicity of different concentrations of O. cf. ovata cells on fish (Dicentrarchus labrax). Phaeodactylum tricornutum was used as control and was tested at the reported cell density. Results are reported as n
dead organisms/n
tested organisms, and time is expressed as hours (h). Time (h)

P. tricornutum 1

789 100 cell mL

0 28 30 31 45 51 52 72 96

O. cf. ovata

O. cf. ovata 1

2 367 cell mL

O. cf. ovata 1

1 138 cell mL

425 cell mL1

Tank 1

Tank 2

Tank 1

Tank 2

Tank 1

Tank 2

Tank 1

Tank 2

0/3 0/3 0/3 0/3 0/3 0/3 0/3 0/3 0/3

0/3 0/3 0/3 0/3 0/3 0/3 0/3 0/3 0/3

0/3 1/3 2/3 2/3 3/3

0/3 1/3 2/3 3/3

0/3 0/3 0/3 0/3 3/3

0/3 0/3 0/3 0/3 1/3 1/3 1/3 3/3

0/3 0/3 0/3 0/3 2/3 3/3

0/3 0/3 0/3 0/3 0/3 0/3 1/3 3/3

90

4.2.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 8 2 e9 2

Toxin profile

Putative palytoxin and all the ovatoxins so far known (Ciminiello et al., 2010) were detected in O. cf. ovata extracts. In the cultures grown under different conditions the relative abundance of individual toxins was similar, with ovatoxina and putative palytoxin being the major and the minor component of the toxin profile, respectively. The highest total toxin content on a per cell basis (pg cell1) was recorded in cultures grown at 25
C, while the highest total toxin concentration on a per litre basis was recorded at 20
C, namely under conditions that induced the highest growth yield. A reverse correlation between growth and toxin production has been reported also by Grane´li et al. (2011), as found also for other dinoflagellates (Etheridge and Roesler, 2005; Errera et al., 2010). As for the salinity experiment the highest total toxin content (pg cell1) was measured at salinity 32, while it decreased at lower and higher salinity values. However, no clear correlation between growth and toxin content was observed in the salinity experiment. The extracellular release increased as the temperature increased, with the maximum 27% value being observed at 30
C, the most unfavourable growth condition in the temperature experiment. This suggests that high temperatures favour cell lysis, leading to toxins being released in the growth medium. Similarly, in the salinity experiment, the highest release was also measured at the most unfavourable growth condition (26
C). Comparable results were obtained for Protoceratium reticulatum (Guerrini et al., 2007) and this could represent a response of the cells to the osmotic stress.

4.3. data

Haemolysis results in comparison with HR LC-MS

Palytoxin converts Naþ/Kþ pump into a non-selective cation channel, causing cell lysis; ouabain and other cardiac glycosides are used as indicators for the site of action since these compounds are specific ligands for the Naþ/Kþ-ATPase. The haemolytic assay proposed by Riobo´ et al. (2008) is a rapid and sensitive method to determine palytoxin content. In our study, it was successfully applied to the analyses of O. cf. ovata extracts in order to gain information about the haemolytic activity of ovatoxins. The haemolytic assay resulted highly reproducible even among separate set of experiments and using different blood samples. The haemolytic activity was tested by using O. cf. ovata extracts obtained from cultures set up at different growth conditions. The obtained data showed a good correlation between haemolysis percentage and the total toxin content measured through HR LC-MS. Although comparison of LC-MS and haemolysis assay results has already been done (Rhodes et al., 2010), in this work a detailed and quantitative cross check between biological assay and chemical analysis was applied to palytoxin-like compounds for the first time. Useful information was obtained from haemolytic tests after pretreatment with ouabain. They showed that ovatoxins behave similarly to palytoxin, suggesting a common mechanism of action, which involves a binding to the Naþ/Kþ pump. The haemolytic activity of all the O. cf. ovata extracts was found to be very similar to that of palytoxin, as confirmed also

by the similar EC50 values. These data suggested that ovatoxins, which represent the major components of the O. cf. ovata extracts (99.5e97%), have a similar haemolytic effect as palytoxin standard. It has to be noted that, in our analyses, the total activity of ovatoxins was measured and it has still to be ascertained whether individual components of the ovatoxin profile present different haemolytic activity. This will be possible when each ovatoxin will be isolated as a pure compound and used to evaluate its haemolytic activity. So far, the haemolytic assay appears to be a good method for preliminary quantification of the whole of palytoxin-like compounds in algal extracts: equation (f1, Fig. 4) obtained from the haemolysis curve, indicating the total haemolysis, can be a powerful tool to evaluate total toxin concentration of algal extracts, especially in laboratories where LC-MS is not available. However, some drawbacks of this assay are represented by the interference of other possibly co-occurring haemolytic compounds and its inability to define toxin profile.

4.4.

Toxicity for crustacean and fish

The toxicological assays revealed a marked toxicity of compounds produced by O. cf. ovata on Artemia nauplii and juvenile sea basses. For Artemia sp., the assay performed with the live O. cf. ovata cells reported mortality of nauplii even at very low cell densities and the EC50 value was significantly lower than those obtained for O. cf. ovata lysed cells, algal extract, and growth medium (Table 3). The difference in EC50 values of O. cf. ovata live cells versus both O. cf. ovata lysed cells and algal extract can be related to a different toxin uptake by the Artemia sp. nauplii: live cells were actually ingested by nauplii whereas either lysed cells or algal extract were taken up through filtration. Thus, this latter mechanism of toxin uptake seems to be less powerful than ingestion. This suggests that herbivorous fish, feed on seaweeds where the benthic dinoflagellates proliferates, is the most vulnerable to O. cf. ovata toxicity. The high EC50 value of the O. cf. ovata growth medium also deserves consideration. This can be related to the low toxin extracellular release emerging by HR LC-MS data (Table 2). Despite the apparently low toxicity of O. cf. ovata growth medium on Artemia nauplii, a long lasting bloom could be anyway hazardous to marine crustaceans, particularly considering that cell lyses and toxin extracellular release increase at the end of the stationary phase reached at the end of the bloom. Unlike the haemolytic assay, the Artemia sp. assay was not able to detect differences in the toxin contents of O. cf. ovata cultures grown at different salinity and temperature conditions. This could be due to the extreme sensitivity of Artemia sp. nauplii to O. cf. ovata live cells (EC50 values ranging from 6 to 24 cell mL1), which has not been observed for any other harmful algae so far (Pezzolesi et al., 2010). Thus, Artemia sp. assay is not able to catch relatively small differences among different samples and, therefore, it cannot be used for quantitative purposes. In the ichthyotoxic assay, sea basses exposed to O. cf. ovata live cells died within a few days despite they are known not to feed on microalgal cells. This mortality could be attributed to an haemolytic effect of palytoxin-like compounds on the gills,

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 8 2 e9 2

where Naþ/Kþ ATPase activity is high in the juvenile stage of sea basses (Varsamos et al., 2004). However, we cannot exclude an effect due to accidental ingestion of algal cells, which were contained in the surrounding water at high density.

Acknowledgements This research was supported by MURST PRIN, Rome, Italy. We thank Prof. Poglayen of the Department of Veterinary Public Health and Animal Pathology (University of Bologna) for kindly proving us the sheep blood, and the hatchery of Valle Ca’ Zuliani (Pila di Porto Tolle, Rovigo, Italy) for the juvenile sea basses. We thank Dr. Andrea Pasteris for the advice on the statistical analysis. We are grateful to Dr. Beth Strain for English revision of the manuscript.

references

Abboud-Abi Saab, M., 1989. Les dinoflagelle´s des eaux coˆtie`res libanaises-espe`ces rares ou nouvelles du phytoplancton marin. Lebanese Science Bulletin 5, 5e16. Aligizaki, K., Nikolaidis, G., 2006. The presence of the potentially toxic genera Ostreopsis and Coolia (Dinophyceae) in the North Aegean Sea, Greece. Harmful Algae 5, 717e730. Aligizaki, K., Katikou, P., Nikolaidis, G., Panoub, A., 2008. First episode of shellfish contamination by palytoxin-like compounds from Ostreopsis species (Aegean Sea, Greece). Toxicon 51, 418e427. Ashton, M., Tosteson, T., Tosteson, C., 2003. The effect of elevated temperature on the toxicity of the laboratory cultured dinoflagellate Ostreopsis lenticularis (Dinophyceae). Revista de Biologı´a Tropical 51, 1e6. Bianco, I., Sangiorgi, V., Penna, A., Guerrini, F., Pistocchi, R., Zaottini, E., Congestri, R., 2007. Ostreopsis ovata in benthic aggregates along the Latium Coast (middle Tyrrhenian Sea). International Symposium on Algal Toxins, SITOX, p. 29. Bignami, G.S., 1993. A rapid and sensitive hemolysis neutralization assay for palytoxin. Toxicon 31, 817e820. Bottalico, A., Micella, P., Feliciti, G.P., 2002. Fioritura di Ostreopsis sp. (Dinophyta) nel porto di Otranto. In: Gruppo di Lavoro per l’Algologia. Societa` Botanica Italiana, Chioggia 8e9 November 2002. Chang, F.H., Shimizu, Y., Hay, B., Stewart, R., Mackay, G., Tasker, R., 2000. Three recently recorded Ostreopsis spp. (Dinophyceae) in New Zealand: temporal and regional distribution in the upper North Island from 1995 to 1997. New Zealand Journal of Marine & Freshwater Research 34, 29e39. Ciminiello, P., Dell’Aversano, C., Fattorusso, E., Forino, M., Magno, G.S., Tartaglione, L., Grillo, C., Melchiorre, N., 2006. The Genoa 2005 outbreak. Determination of putative palytoxin in Mediterranean Ostreopsis ovata by a new liquid chromatography tandem mass spectrometry method. Analytical Chemistry 78, 6153e6159. Ciminiello, P., Dell’Aversano, C., Fattorusso, E., Forino, M., Tartaglione, L., Grillo, C., Melchiorre, N., 2008. Putative palytoxin and its new analogue, ovatoxin-a, in Ostreopsis ovata collected along the Ligurian coasts during the 2006 toxic outbreak. Journal of the American Society for Mass Spectrometry 19, 111e120. Ciminiello, P., Dell’Aversano, C., Dello Iacovo, E., Fattorusso, E., Forino, M., Grauso, L., Tartaglione, L., Guerrini, F., Pistocchi, R., 2010. Complex palytoxin-like profile of Ostreopsis ovata.

91

Identification of four new ovatoxins by high-resolution liquid chromatography/mass spectrometry. Rapid Communications in Mass Spectrometry 24, 2735e2744. Errera, R.M., Bourdelais, A., Drennan, M.A., Dodd, E.B., Henrichs, D.W., Campbell, L., 2010. Variation in brevetoxin and brevenal content among clonal cultures of Karenia brevis may influence bloom toxicity. Toxicon 55, 195e203. Etheridge, S.M., Roesler, C.S., 2005. Effects of temperature, irradiance, and salinity on photosynthesis, growth rates, total toxicity, and toxin composition for Alexandrium fundyense isolates from the Gulf of Maine and Bay of Fundy. Deep-Sea Research II 52, 2491e2500. Faust, M.A., Morton, S.L., Quod, J.P., 1996. Further SEM study of the marine dinoflagellates: the genus Ostreopsis (Dinophyceae). Journal of Phycology 32, 1053e1065. Grane´li, E., Vidyarathna, N.K., Funari, E., Cumaranatunga, R., 2008. Climate change and benthic Dinoflagellates - the Ostreopsis ovata case. In: Proceedings of the 13th International Conference on Harmful Algae, p. 42. Hong Kong, 3e7 November, 2008. Grane´li, E., Vidyarathna, N.K., Funari, E., Cumaranatunga, P.R.T., Scenati, R., 2011. Can increases in temperature stimulate blooms of the toxic benthic dinoflagellate Ostreopsis ovata? Harmful Algae 10, 165e172. Guerrini, F., Ciminiello, P., Dell’Aversano, C., Tartaglione, L., Fattorusso, E., Boni, L., Pistocchi, R., 2007. Influence of temperature, salinity and nutrient limitation on yessotoxin production and release by the dinoflagellate Protoceratium reticulatum in batch-cultures. Harmful Algae 6, 707e717. Guerrini, F., Pezzolesi, L., Feller, A., Riccardi, M., Ciminiello, P., Dell’Aversano, C., Tartaglione, L., Dello Iacovo, E., Fattorusso, E., Forino, M., Pistocchi, R., 2010. Comparative growth and toxin profile of cultured Ostreopsis ovata from the Tyrrhenian and Adriatic Seas. Toxicon 55, 211e220. Guillard, R.R.L., 1975. Culture of phytoplankton for feeding marine invertebrates. In: Smith, W.L., Chanley, M.H. (Eds.), Culture of Marine Invertebrates Animals. Plenum Press, New York, pp. 26e60. Habermann, E., 1989. Palytoxin acts through Naþ, Kþ-ATPase. Toxicon 27, 1171e1187. Hasle, G.R., 1978. The inverted microscope method. In: Sournia, A. (Ed.), Phytoplankton Manual. Monographs on Oceanographic Methodology, vol. 6. UNESCO, Paris, pp. 88e96. Hoshaw, R.W., Rosowski, J.R., 1973. Methods for microscopic algae. In: Stein, J.R. (Ed.), Handbook of Phycological Methods. Cambridge University Press, New York, pp. 53e67. IRSA-CNR, 2003. Metodi analitici per le acque, APAT Manuali e Linee Guida 29/2003. In: 8060-Metodo di valutazione della tossicita` acuta con Artemia sp., vol. 3 1043e1050. Lenoir, S., Ten-Hage, L., Turquet, J., Quod, J.P., Bernard, C., Hennion, M.C., 2004. First evidence of palytoxin analogues from an Ostreopsis mascarenensis (Dinophyceae) benthic bloom in Southwestern Indian Ocean. Journal of Phycology 40, 1042e1051. Mangialajo, L., Bertolotto, R., Cattaneo-Vietti, R., Chiantore, M., Grillo, C., Lemee, R., Melchiorre, N., Moretto, P., Povero, P., Ruggirei, N., 2008. The toxic benthic dinoflagellate Ostreopsis ovata: quantification of proliferation along the coastline of Genoa, Italy. Marine Pollution Bulletin 56, 1209e1214. Mangialajo, L., Ganzin, N., Accoroni, S., Asnaghi, V., Blanfune´, A., Cabrini, M., Cattaneo-Vietti, R., Chavanon, F., Chiantore, M., Cohu, S., Costa, E., Fornasaro, D., Grossel, H., MarcoMiralles, F., Maso´, M., Ren˜e´, A., Rossi, A.M., Montserrat, S., Thibaut, T., Totti, C., Vila, M., Leme´e, R., 2011. Trends in Ostreopsis proliferation along the Northern Mediterranean coasts. Toxicon 57, 408e420. Monti, M., Minocci, M., Beran, A., Ivesa, L., 2007. First record of Ostreopsis cfr. ovata on macroalgae in the Northern Adriatic Sea. Marine Pollution Bulletin 54, 598e601.

92

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 8 2 e9 2

Morton, S.L., Norris, D.R., Bomber, J.W., 1992. Effect of temperature, salinity and light-intensity on the growth and seasonality of toxic dinoflagellates associated with ciguatera. Journal of Experimental Marine Biology and Ecology 157, 79e90. Onuma, Y., Satake, M., Ukena, T., Roux, J., Chanteau, S., Rasolofonirina, N., Ratsimaloto, M., Naoki, H., Yasumoto, T., 1999. Identification of putative palytoxin as the cause of clupeotoxism. Toxicon 37, 55e65. Parsons, M.L., Preskitt, L.B., 2007. A survey of epiphytic dinoflagellates from the coastal waters of the island of Hawai‘i. Harmful Algae 6, 658e669. Pezzolesi, L., Cucchiari, E., Guerrini, F., Pasteris, A., Totti, C., Pistocchi, R., 2010. Toxicity evaluation of Fibrocapsa japonica from the Northern Adriatic Sea through a chemical and toxicological approach. Harmful Algae 8, 504e514. Pistocchi, R., Pezzolesi, L., Guerrini, F., Vanucci, S., Dell’Aversano, C., Fattorusso, E., 2011. A review on the effects of environmental conditions on growth and toxin production of Ostreopsis ovata. Toxicon 57, 421e428. Riobo´, P., Paz, B., Franco, J.M., Va´zquez, J.A., Murado, M.A., 2008. Proposal for a simple and sensitive haemolytic assay for palytoxin: toxicological dynamics, kinetics, ouabain inhibition and thermal stability. Harmful Algae 7, 415e429. Riobo´, P., Franco, J.M., 2011. Palytoxins: biological and chemical determination. Toxicon 57, 368e375. Rhodes, L., Smith, K., Munday, R., Briggs, L., Selwood, A., 2010. Ostreopsis ovata isolated from Rarotonga, Cook Islands, 14th International Conference of Harmful Algae, Hersonissos-Crete (Greece), 1e5 November. Conference abstract book, p. 229. Rhodes, L., 2011. World-wide occurrence of the toxic dinoflagellate genus Ostreopsis Schmidt. Toxicon 57, 400e407. Sansoni, G., Borghini, B., Camici, G., Casotti, M., Righini, P., Fustighi, C., 2003. Fioriture algali di Ostreopsis ovata (Gonyaulacales: Dinophyceae): un problema emergente. Biol. Amb 17, 17e23. Shears, N.T., Ross, P.M., 2009. Blooms of benthic dinoflagellates of the genus Ostreopsis; an increasing and ecologically important phenomenon on temperate reefs in New Zealand and worldwide. Harmful Algae 8, 916e925. Simoni, F., Gori, L., Di Paolo, C., Lepri, L., Mancino, A., 2004. Fioriture d’Ostreopsis ovata, Coolia monotis, Prorocentrum lima nelle macroalghe del mar Tirreno settentrionale, Mediterraneo (seconda fase di studio). Biol. Ital. 10, 68e73.

Sun, J., Liu, D., 2003. Geometric models for calculating cell biovolume and surface area for phytoplankton. Journal of Plankton Research 25, 1331e1346. Taniyama, S., Osamu, A., Masamitsu, T., Sachio, N., Tomohiro, T., Yahia, M., Tamao, N., 2003. Ostreopsis sp., a possible origin of palytoxin (PLTX) in parrotfish Scarus ovifrons. Toxicon 42, 29e33. Taylor, F.J.R., 1979. A description of the benthic dinoflagellate associated with maitotoxin and ciguatoxin, including observations on Hawaiian material. In: Taylor, D.L., Seliger, H. H. (Eds.), Toxic Dinoflagellate Blooms. Elsevier/North-Holland, New York, pp. 71e76. Tognetto, L., Bellato, S., Moro, I., Andreoli, C., 1995. Occurrence of Ostreopsis ovata (Dinophyceae) in the Tyrrhenian Sea during summer 1994. Botanica Marina 38, 291e295. Totti, C., 2003. Influence of the plume of the river Po on the distribution of subtidal microphytobenthos in the Northern Adriatic Sea. Botanica Marina 46, 161e178. Totti, C., Cucchiari, E., Romagnoli, T., Penna, A., 2007. Bloom of Ostreopsis ovata on the Conero riviera (NW Adriatic Sea). Harmful Algae News 33, 12e13. Totti, C., Accoroni, S., Cerino, F., Cucchiari, E., Romagnoli, T., 2010. Ostreopsis ovata bloom along the Conero Riviera (northern Adriatic Sea): relationships with environmental conditions and substrata. Harmful Algae 9, 233e239. Usami, M., Satake, M., Ishida, S., Inoue, A., Kan, Y., Yasumoto, T., 1995. Palytoxin analogs from the dinoflagellate Ostreopsis siamensis. Journal of the American Chemical Society 117, 5389e5390. Ukena, T., Satake, M., Usami, M., Oshima, Y., Naoki, H., Fujita, T., Kan, Y., Yasumoto, T., 2001. Structure elucidation of ostreocin D, a palytoxin analog isolated from the dinoflagellate Ostreopsis siamensis. Bioscience, Biotechnology, and Biochemistry 65, 2585e2588. Varsamos, S., Wendelaar Bonga, S.E., Charmantier, G., Flik, G., 2004. Drinking and Naþ/Kþ ATPase activity during early development of European sea bass, Dicentrarchus labrax: Ontogeny and short-term regulation following acute salinity changes. Journal of Experimental Marine Biology and Ecology 311, 189e200. Vila, M., Garce´s, E., Maso`, M., 2001. Potentially toxic epiphytic dinoflagellate assemblages on macroalgae in the NW Mediterranean. Aquatic Microbial Ecology 26, 51e60.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6

Available online at www.sciencedirect.com

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

Recycled water: Potential health risks from volatile organic compounds and use of 1,4-dichlorobenzene as treatment performance indicator Clemencia Rodriguez a,b,*, Kathryn Linge c,1, Palenque Blair d,2, Francesco Busetti c,3, Brian Devine a,4, Paul Van Buynder a,e,5, Philip Weinstein f,6, Angus Cook a,7 a

School of Population Health, Faculty of Medicine, Dentistry and Health Sciences, The University of Western Australia, 35 Stirling Hwy, (M431) Crawley 6009 Western Australia, Australia b Department of Health, Government of Western Australia, Grace Vaughan House 227 Stubbs Terrace, Shenton Park, 6008 Western Australia, Australia c Curtin Water Quality Research Centre, Department of Chemistry, Curtin University, Kent Street, Bentley 6102 Western Australia, Australia d Water Corporation of Western Australia, 629 Newcastle Street Leederville, 6007 Western Australia, Australia e Fraser Health Authority, C200, 9801 King George Boulevard Surrey, BC V3T 5E5, Canada f University of South Australia, City West Campus, GPO Box 2471 Adelaide, 5001 South Australia, Australia

article info

abstract

Article history:

Characterisation of the concentrations and potential health risks of chemicals in recycled

Received 25 May 2011

water is important if this source of water is to be safely used to supplement drinking water

Received in revised form

sources. This research was conducted to: (i) determine the concentration of volatile organic

19 September 2011

compounds (VOCs) in secondary treated effluent (STE) and, post-reverse osmosis (RO)

Accepted 16 October 2011

treatment and to; (ii) assess the health risk associated with VOCs for indirect potable reuse

Available online 25 October 2011

(IPR). Samples were examined pre and post-RO in one full-scale and one pilot plant in Perth, Western Australia. Risk quotients (RQ) were estimated by expressing the maximum

Keywords:

and median concentration as a function of the health value. Of 61 VOCs analysed over

Water recycling

a period of three years, twenty one (21) were detected in STE, with 1,4-dichlorobenzene

Water quality

(94%); tetrachloroethene (88%); carbon disulfide (81%) and; chloromethane (58%) most

Organic pollutants

commonly detected. Median concentrations for these compounds in STE ranged from

Indirect potable reuse

0.81 mg/L for 1,4-dichlorobenzene to 0.02 mg/L for carbon disulphide. After RO, twenty six

Volatile organic compounds

(26) VOCs were detected, of which 1,4-dichlorobenzene (89%); acrylonitrile (83%) chloro-

Reverse osmosis

methane (63%) and carbon disulfide (40%) were the more frequently detected. RQ(max) were all below health values in the STE and after RO. Median removal efficiency for RO was variable, ranging from
77% (dichlorodifluoromethane) to 91.2% (tetrachloroethene).

* Corresponding author. Department of Health, Government of Western Australia, Grace Vaughan House 227 Stubbs Terrace, Shenton Park, 6008 Western Australia, Australia. Tel.: þ8 93884812; fax: þ8 93884910. E-mail addresses: [email protected] (C. Rodriguez), [email protected] (K. Linge), [email protected] (P. Blair), [email protected] (F. Busetti), [email protected] (B. Devine), Paul.VanBuynder@fraserhealth. ca (P. Van Buynder), [email protected] (P. Weinstein), [email protected] (A. Cook). 1 Tel.: þ8 92667534; fax: þ8 92663547. 2 Tel.: þ8 94203328; fax: þ8 94203195. 3 Tel.: þ8 92663273; fax: þ8 92663547. 4 Tel.: þ8 64888667; fax: þ8 64881188. 5 Tel.: þ1 604 587 7621; fax: þ 1 604 587 7625. 6 Tel.: þ61 8 830 25129; fax: þ61 8 830 20828. 7 Tel.: þ8 64887804; fax: þ8 64881188. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.032

94

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6

The results indicate that despite the detection of VOCs in STE and after RO, their human health impact in IPR is negligible due to the low concentrations detected. The results indicate that 1,4-dichlorobenzene is a potential treatment chemical indicator for assessment of VOCs in IPR using RO treatment. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Volatile organic compounds (VOCs) are organic chemicals that have a relatively low boiling point (250  C measured at a standard atmospheric pressure of 101.3 kPa) and high vapour pressure relative to their water solubility. This class of chemicals therefore easily volatilize from water to air at room temperatures and enter the atmosphere upon contact with an airewater interface. Substances that are included in the VOC category are: aliphatic hydrocarbons (e.g. hexane), aromatic hydrocarbons (e.g. benzene, toluene and the xylenes), halogenated hydrocarbons (e.g. tetrachloroethene) and oxygenated compounds (e.g. acetone and similar ketones). VOCs are widely used and comprise an important group of environmental contaminants. They are produced in large volumes and are associated with numerous products and applications, including household cleaners, fuel additives, and commercial and industrial solvents. VOCs dissolve many other substances and are used as cleaning and liquefying agents in fuels, degreasers, adhesives, solvents, polishes, cosmetics, refrigerants, drugs, and dry cleaning solutions (Zogorski et al., 2006). VOCs may be emitted from fabrics, carpets, fibreboard, plastic products, glues, solvents, household cleaners, printed material, methylated spirits, paints and paint products (such as thinners or varnishes), disinfectants, cosmetics, degreasing products, and fuels. They are hence discharged to wastewater treatment plants (WWTP) from a large number of sources including commercial enterprises, industries, and residential households. VOCs have been detected in many water types, including secondary treated effluent (STE). Aliphatic hydrocarbons, aromatic hydrocarbons, halogenated volatiles and dimethyl disulfide account for approximately 70% of all VOCs detected in municipal STE (Koe and Shen, 1997). Although VOCs concentrations in raw wastewater may range from 1 to 150 mg/ L, atmospheric emissions during treatment generally lead to significantly lower dissolved concentrations in STE (Atasoy et al., 2004, Battistoni et al., 2007). Adsorption and biodegradation can reduce the concentration of VOCs in WWTPs. VOC release to the atmosphere during collection and in particular during aeration treatment is considered the most important method of removal of VOCs from STE (Fatone et al., 2011). The aeration that occurs during wastewater treatment and during many sludge treatment processes can achieve more than 90% removal of the VOCs concentration in raw wastewater (NRC, 1996). For example, Wu et al. (2002) reported a 96% decrease in total VOCs in a WWTP during exposure to the atmosphere via air stripping (Wu et al., 2002). In some circumstances, VOCs may also be found in public drinking water supplies as a result of spills, discharges, atmospheric deposition or leaching from contaminated soils.

Tetrachloroethylene, trichloroethene, 1,1-dichloroethene and benzene are examples of VOCs that are occasionally detected (Williams et al., 2002). Industrial discharges may lead to the release of VOCs into groundwater (along with gasoline oxygenates). For example, eighteen (18) of eighty-eight (88) VOCs were detected in twenty eight (28) wells sampled in the San Diego GroundWater Ambient Monitoring and Assessment study (Wright et al., 2005). Groundwater contamination with non-aqueous phase liquids, such as chlorinated solvents and petrol hydrocarbons, may pose a health risk if used as a drinking water source as they can be difficult to remove by treatment (Patterson et al., 1993). The presence of VOCs in drinking water is of concern because some of these compounds have adverse health effects, including potential carcinogenesis, and because they can change the taste and odour of drinking water. The health impact of VOCs varies greatly from those that are highly toxic, to those with no known health effect. As with other organic chemicals, the extent and nature of the health effect will depend on the level of exposure and length of exposure. Some VOCs may adversely affect the liver, kidneys, spleen, and stomach, as well as the nervous, circulatory, reproductive, immune, cardiovascular, and respiratory systems. Some VOCs may affect cognitive abilities, balance, and coordination. At high levels of exposure, VOCs can cause central nervous system depression (Boyes et al., 2000; Brouwer et al., 2005; Herpin et al., 2009) and can be irritating upon contact with the skin, to mucous membranes of the eyes or to the mucous membranes if inhaled (Toccalino et al., 2006; WHO, 2011). Acute symptoms after exposure to some VOCs (mainly from inhalation) include headaches, dizziness, visual disorders, and memory impairment. The chronic health effects to the general public from ingestion of VOCs at low concentrations in drinking water are less well understood but health values are well above offensive taste/odour thresholds and contain significant safety margins.

2.

Regulatory framework

VOCs comprise almost half of the 129 priority pollutants designated by the USEPA for limitation or prevention of introduction to water (USEPA, 1994), and some are also included in the European Commission priority pollutant lists. The EU adopted decision No 2455/2001/EC, which established a list of 33 priority substances in the field of water policy, include the following VOCs: benzene, C10e13-chloroalkanes, 1,2-dichloroethane, dichloromethane, hexachlorobutadiene, trichlorobenzenes and trichloromethane (European Commission, 1998). In Australia, the National Industrial Chemical Notification and Assessment Scheme (NICNAS) is the federal agency

95

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6

assessing the human health risk of industrial chemicals including VOCs introduced into the country. The assessment reports include hazard, exposure assessments and risk characterisation for occupational health and safety, environment and public health. The reports are available on the web and are classified as (i) new chemicals; (ii) priority existing chemicals; and (iii) other assessments (http://www.nicnas.gov.au/ Publications/CAR.asp). National guidelines for sewerage systems e in particular those pertaining to (i) Effluent Management and (ii) Acceptance of Trade Waste (Industrial Waste) e provide a framework for sewerage authorities responsible for the management, monitoring, disposal and implementation of trade waste management programs. Regulations exist for levels of VOCs in various contexts, such as water systems (drinking water, sewage discharges and stormwater disposal), occupational settings and air emissions. Australian Federal and State regulations often limit the quantity of VOCs that are emitted from sources such as industrial facilities, WWTPs and landfills. In Western Australia, the Department of Environment and Conservation regulates the level of contaminants allowable in wastewater streams under the Environmental Protection Act 1986 and may prescribe specific license conditions for VOCs, depending on the type of industry (NWC, 2011). Examples of compounds that have to be reduced or removed from industrial facilities before discharge to sewers include benzene, trichlorethene, vinyl chloride and xylenes. Steam or air stripping, carbon adsorption and solvent extraction are all methods used for removing VOCs from wastewater before secondary treatment for compliance with regulatory requirements. Regulatory agencies and institutions set drinking water values for VOCs based on toxicological and epidemiological assessments. For the majority of VOCs that are assumed to be non-carcinogenic, it is hypothesised that there is a threshold dose below which no adverse health effects will occur. Consequently, drinking water guidelines are calculated based on tolerable daily intake (TDI) values. The TDI values are derived from toxicological studies conducted in animals and epidemiological data when available. Some VOCs are known or suspected carcinogens. For those VOCs classed as carcinogenic (without a threshold dose), guideline values are derived using mathematical models that combine toxicological data with the concept of acceptable levels of risk for lifetime consumption. The WHO guideline values are derived using an acceptable level of risk of one in a hundred thousand excess cancers attributable to a particular VOC consumption at the guideline concentration (WHO, 2011). For the Australian

Drinking Water Guidelines (ADWG), the acceptable level of risk is mainly based on one in a million excess cancers (NHMRC, 2011). In many cases the toxicological data used is the same but the assumptions used to calculate the health value varies. For example, the Canadian standard for 1,4dichlorobenzene is 5 mg/L (FPT Committee on Drinking Water, 2010) while the regulatory value for: USEPA is 75 mg/L (USEPA, 2011); WHO is 300 mg/L (WHO, 2011) and for Australia is 40 mg/L (NHMRC, 2011). The derivation of the health value ¼ 300 mg/L in the WHO guidelines is based on non cancer effects using the lowest observed adverse effect level (LOAEL) of 150 mg/kg for kidney effects in a two-year rat study with an uncertainty factor of 1000. In contrast the Canadian guidelines classified 1,4-dichlorobenzene as Group II e probably carcinogenic to humans based on a National Toxicology Program report and the calculation is based on the slope of a doseeresponse data with linear extrapolation to zero. The estimated ranges of concentrations corresponding to a lifetime risk of one in a hundred thousand are used. The work presented in this paper is part of a larger project investigating the effectiveness of microfiltration/reverse osmosis (MF/RO) to treat STE for indirect potable reuse (IPR), a key water conservation strategy for Western Australia (DOHWA, 2009). The objectives of these study were to (i) evaluate the range and concentration of volatile organic compounds (VOCs) in secondary treated effluent (STE) and post-reverse osmosis (RO) treatment for 61 VOCs; (ii) assess the health risk associated with VOCs for indirect potable reuse (IPR) with post-RO water and, (iii) determine the efficacy of RO to remove VOCs. This study provides the most extensive analysis of VOCs in treated wastewater in Australia published to date.

3.

Methods

3.1.

Sample sites

Six (6) sampling events were conducted from November 2006 to June 2008 (Table 1), with an overall total of 32 sampling days. Typically a single sampling event consisted of between 4 and 6 sampling trips over a week, with sampling focused on STE or and post-RO water. However, on a number of occasions, sampling of post-MF water was also undertaken. During membrane treatment, wastewater undergoes chloramination before MF to prevent RO membrane fouling and samples of post-MF water were analysed to determine the

Table 1 e Measurement of VOCs by event and location. Event 1 2 3 4 5 6 Total

Month

#days

Year

Groundwater

Secondary treated effluent

Post microfiltration

Post-reverse osmosis

Total

November May/June September January April June

4 6 6 6 5 5 32

2006 2007 2007 2008 2008 2008

0 108 0 114 0 0 222

159 432 336 342 392 362 2023

53 162 106 0 56 108 485

158 162 336 342 336 262 1596

370 864 778 798 784 732 4326

96

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6

effect of chloramination during the MF/RO process. Grab STE samples represent the three major WWTPs in Perth and were taken directly from Beenyup WWTP and Subiaco WWTP and at the influent stream of the Kwinana Water Reclamation Plant (KWRP) for Woodman Point WWTP. Samples post-MF and post-RO were collected from two advanced treatment plants at KWRP and Beenyup Pilot Plant (BPP) in order to characterise water quality through the membrane treatment process. Details of each have been previously published (DOHWA, 2009) but briefly, KWRP treats secondary treated wastewater from Woodman Point WWTP by MF/RO to produce approximately 16 ML/day of general process water for neighbouring industrial facilities, reducing Perth’s total demand for drinking water by about 2%. The BPP treats a small volume of STE (approximately 100 kL/day) from Beenyup WWTP by MF/RO and is the first stage of a larger project investigating indirect potable reuse of Beenyup STE. The BPP was commissioned after the May/June 2007 sampling event (sampling event 2 in Table 1). Both plants are owned by the Water Corporation of Western Australia. Woodman Point WWTP receives wastewater primarily from residential areas, but also receives about 6% of wastewater from industrial facilities (DOHWA, 2009), while Beenyup WWTP has a sewer catchment that is mainly residential in nature. Grab samples were also collected from groundwater (GW) during sampling events 2 and 4. The groundwater in this context corresponds to the raw drinking water source. Standard protocols were used to ensure adequate sample preparation, preservation and transportation to the laboratory. Laboratory blanks, trip and field blanks were also analysed and constituted about one third of the samples analysed.

3.2.

Analysis of VOCs

The selection of VOCs was based on their risk profiles and factors. The following criteria were used to guide analytes inclusion in the target list: (i) the VOC is currently or has been registered for use in Australia; (ii) there is a high likelihood of the VOC being detected in wastewater based on known chemical and physical properties; (iii) the VOC has previously been detected in natural waters or wastewater; (iv) there were public perceptions that the chemical may pose a possible public health hazard; and (v) the VOC is listed in the ADWG (2004) (NHMRC, 2004) or other international regulatory agencies as regulated or as part of the USEPA contaminant candidate list. All VOCs except acrolein, acrylonitrile and methyl tertiary butyl ether (MTBE) were measured by purge and trap GCeMS. Acrolein, acrylonitrile, MTBE were extracted and preconcentrated by headspace solid-phase microextraction (SPME) before GCeMS analysis. Quantification was performed by mass spectrometry (MS) with electron ionisation (EI), with peak identification and calculation of recovery was aided by inclusion of surrogate standards. Limits of detection were determined for every analytical run and were calculated using the standard deviation of replicate analyses of a standard solution of appropriate concentration (typically 0.05e0.1 mg/L). Standard deviations were then converted to 95% confidence intervals using the student’s t-test.

Relative standard uncertainties were calculated using an uncertainty budget that incorporated precision, calibration standard preparation, sample volume, and linear regression of the calibration curve. Sample homogeneity was considered a negligible source of uncertainty. Acrolein and acrylonitrile were not validated to the same extent as other VOCs because they were only analysed during sampling event 3. The VOCs analysed, standard relative uncertainty at 0.5 mg/L, and average limits of detection (LODs) are presented in Tables 2 and 3.

3.3.

Data analysis

Unlike other chemical classes of compounds (e.g. dioxins), there is no common toxicological mechanism for VOCs, and therefore the potential human health risk was evaluated for individual compounds. Risk quotients (RQ) were estimated by expressing the maximum and median concentration in STE as a function of the health value for detected VOCs. For VOCs without detections in STE, RQs were calculated as the ratio between the LOD and the health value as a worst case scenario. A three tiered screening health risk assessment approach was used for the derivation of the health values. The basis for the tool has been discussed and applied in previous publications (Rodriguez et al., 2007a, 2007b). Under this system, VOCs were allocated to “tier 1 (regulated contaminants)”; “tier 2 (unregulated contaminants with toxicity information)” or “tier 3 Threshold of Toxicological Concern (unregulated contaminants with no toxicity information)”. For VOCs in tier 1, the order or priority for setting the health benchmark values was the ADWG (2011) (NHMRC, 2011), the ADWG (2004) (NHMRC, 2004), WHO guidelines including the 2nd addendum to the 3rd edition published in 2006 (WHO, 2011), the Drinking water standards and health advisories from the USEPA (USEPA, 2011) and the California Drinking Water Notification Levels and Response Levels (CDPH, 2010), based on the methodology previously described (Rodriguez et al., 2007b). Data were analysed in Stata version 10 (Stata Corp, 2007). Comparison of median concentrations was performed using non-parametric tests. For median comparison between KWRP and BPP the ManneWhitney test was used, while the Kruskal Wallis X2 test was used for comparison of median values of three or more groups. Results are reported at a significance level of 5% ( p < 0.05).

4.

Results

A total of 61 VOCs were analysed in at least one sampling event. A total of 4326 measurements were included in the statistical analysis for VOCs after excluding QA/QC samples. The distribution of sampling by event and location is presented in Table 1.

4.1.

Secondary treated effluent (STE)

Twenty one (21) VOCs (34% of the total) were detected in STE (Table 2). The most frequently detected VOC was 1,4-dichlorobenzene (93.9% of STE samples), followed by

Table 2 e VOCs detected in secondary treated effluent, post-microfiltration and/or post-reverse osmosis water and corresponding RQs. STE samples Parameter

CASR No Mean LOD SRU (%) Tier Health (mg/L) (0.5 mg/L) value (mg/L) 0.099 0.041 0.018 0.026 0.017 0.030 0.056 0.032 0.030 0.046 0.244 0.045 0.033 0.066 0.272 0.091 0.078 1.471 0.031 0.118 0.109 0.029 0.087 0.052 0.042 0.099 0.027 0.033 0.080 0.083 0.017 0.047 0.136 0.096

31.2 20.8 16.4 17.7 8.4 13.2 20.9 15.9 ND 30.4 20.7 53.4 23.2 20.1 21 32.6 37 32 41.7 27.9 30.2 17 48 53.7 18.2 40.2 52.6 13.2 31.7 59 20.1 26.7 37.4 50.4

1 2 1 1 1 1 2 1 1 1 2 2 3 2 2 1 1 1 2 1 1 1 1 2 1 1 1 1 2 2 1 3 1 1

0.057

34.6

3

5 330 40 1500 3 60 600 40 6 1 10 700 0.7 0.7 1000 4 300 40 100 50 800 20 600 260 600 600 30 300 260 330 100 7 30 150 7

Source

n

USEPA 2011 CDPH 2010 WHO 2011 ADWG 2011 ADWG 2011 ADWG 2011 USEPA 2011 ADWG 2011 USEPA 2011 ADWG 2011 USEPA 2011 IRIS 1990 TTC TTC USEPA 2011 ADWG 2011 ADWG 2011 USEPA 2011 USEPA 2011 ADWG 2011 ADWG 2011 WHO 2011 ADWG 2011 CDPH 2010 ADWG 2011 ADWG 2011 ADWG 2011 ADWG 2011 CDPH 2010 CDPH 2010 ADWG 2011 TTC ADWG 2011 CDPH 2008

29 28 29 29 29 29 29 29 6 29 29 15 29 29 24 8

3.5 3.6 10.3 10.3 10.3 34.5 10.3 93.1 50.0 27.6 3.5 80.0 24.1 62.1 4.2 12.5

0.02 0.0001 0.0005 0.00003 0.007 0.0009 0.00003 0.02 0.003 0.08 0.002 0.00002 0.04 0.004 0.0002 0.02

0.04 0.0002 0.005 0.0001 0.02 0.002 0.001 0.08 0.01 0.1 0.02 0.0006 0.8 0.02 0.0005 0.05

26

3.9

0.03

0.13

29 28 29

86.2 14.3 48.3

0.009 0.0002 0.003

0.2 0.0003 0.04

28

17.9

0.00005

0.0001

TTC

RQ (median)

RQ (max)

n

Post-RO samples

% of RQ RQ n % of RQ RQ detection (median) (max) detection (median) (max)

9 9 9

66.7 11.1 55.6

0.0007 0.0005 0.00003

9 9 9

55.6 11.1 100.0

0.002 0.0002 0.03

9 9 3 9 9 7

22.2 11.1 100.0 33.3 88.9 14.3

9

0.009 0.004 0.003

26

7.7

0.0001

0.0002

0.08 0.02 0.0003 0.09 0.006 0.0005

27 27 0.003 27 0.001 27 0.07 27 6 0.08 27 0.02 27 0.001 15 0.8 27 0.03 27 0.0005 22

3.7 3.7 3.7 11.1 88.9 83.3 29.6 7.4 40.0 11.1 63.0 4.6

0.000007 0.003 0.0004 0.00003 0.005 0.01 0.08 0.002 0.00002 0.03 0.003 0.0002

0.0002 0.02 0.0009 0.0002 0.02 0.02 0.1 0.02 0.0002 0.09 0.01 0.0005

66.7

0.0005

0.004

26 24 27 27 26 27 26 26 26 26 27 27 27

9 9 9 9 9 9 9 9

22.2 88.9 88.9 55.6 66.7 22.2 88.9 66.7

0.0002 0.05 0.0004 0.02 0.0003 0.0003 0.0005 0.0004

0.02 0.2 0.008 0.03 0.004 0.003 0.004 0.006

7.7 4.2 7.4 14.8 26.9 7.4 7.7 3.9 15.4 15.4 3.7 11.1 3.7

0.0002 0.03 0.0001 0.003 0.0002 0.001 0.0002 0.0002 0.0001 0.0002 0.0005 0.0001 0.0004

0.0005 0.04 0.002 0.006 0.002 0.003 0.0003 0.0003 0.0001 0.0004 0.004 0.0002 0.0005

9 9 8 9 9

66.7 11.1 25 44.4 11.1

0.0004 0.0001 0.01 0.01 0.001

0.002 0.0003 0.3 0.04 0.002

9

22.2

0.02

0.02

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6

1,1,2-trichloroethane 79-00-5 1,2,4-trimethylbenzene 95-63-6 1,2-dichloropropane 78-87-5 1,2-dichlorobenzene 95-50-1 1,2-dichloroethane 107-06-2 1,2-dichloroethene, cis 156-59-2 1,3-dichlorobenzene 541-73-1 1,4-dichlorobenzene 106-46-7 Acrylonitrile 107-13-1 Benzene 71-43-2 Bromomethane 74-83-9 Carbon disulfide 75-15-0 Chloroethane 75-00-3 Chloromethane 74-87-3 Dichlorodifluoromethane 75-71-8 Dichloromethane 75-09-2 Ethylbenzene 100-41-4 MTBE 1634-04-4 Naphthalene 91-20-3 Tetrachloroethene 127-18-4 Toluene 108-88-3 Trichloroethene 79-01-6 m-xylene 108-38-3 n-butylbenzene 104-51-8 o-xylene 95-47-6 p-xylene 106-42- 3 1,2,3-trichlorobenzene 87-61-6 Chlorobenzene 108-90-7 tert butylbenzene 98-06-6 1,3,5-trimethylbenzene 108-67-8 1,3-dichloropropene 542-75-6 2-propyltoluene 28729-54-6 Styrene 100-42-5 Trichlorofluoromethane 75-69-4 (Freon 11) p-Isopropyltoluene 99-87-6

% of detection

Post-MF samples

LOD, limit of detection; CASR No, registry number for each chemical assigned by the Chemical Abstracts Service, a division of the American Chemical Society; SRU, Standard Relative Uncertainty; n, total number of samples; STE, secondary treated effluent; all values are expressed in mg/L; RQ, risk quotient; ND: not determined; IRIS, Integrated Risk Information System of the USEPA (IRIS, 1990 #64); TTC: Threshold of Toxicological Concern; USEPA Drinking Water Standards and Health Advisories (USEPA, 2011); CDPH: California Department of Public Health (CDPH, 2008, 2010); Australian Drinking Water Guidelines(ADWG) (NHMRC, 2011); TTC: Threshold of Toxicological Concern.

97

98

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6

Table 3 e VOCs without detections in any of the samples and corresponding “worst-case scenario” RQs. Parameter 1,1,1,2-tetrachloroethane 1,1,1-trichloroethane 1,1,2,2-tetrachloroethane 1,1,2-trichloro-1,2, 2-trifluoroethane (Freon 113) 1,1-dichloroethane 1,1-dichloroethene 1,1-dichloropropene 1,2,3-trichloropropane 1,2,4-trichlorobenzene 1,2-dibromomethane 1,2-dibromo-3-chloropropane 1,2-dichloroethene, trans 1,2-dichloropropene 1,3-dichloropropane 2,2-dichloropropane 2-chlorotoluene (ortho) 4-chlorotoluene (para) Acrolein Bromobenzene Carbon tetrachloride Ethylene Dibromide (1,2-dibromoethane) Hexachlorobutadiene Isopropylbenzene Vinyl Chloride n-propylbenzene sec-butylbenzene

CASR No

Mean LOD (mg/L)

630-20-6 71-55-6 79-34-5 76-13-1

0.076 0.037 0.022 0.029

75-34-3 75-35-4 563-58-6 96-18-4 120-82-1 8003-07-4 96-12-8 156-60-5 563-54-2 142-28-9 594-207 95-49-8 106-43-4 107-02-8 108-86-1 56-23-5 106-93-4 87-68-3 98-82-8 75-01-4 103-65-1 135-98-8

SRU (%) (0.5 mg/L)

n

Tier

27.8 14.3 12.2 28.2

73 73 73 24

2 1 1 1

0.065 0.056 0.039 0.057 0.024 0.100 0.048 0.041 0.032 0.077 0.225 0.174 0.260 0.300 0.137 0.052

12.3 20.5 23.7 21.7 55.3 ND 31.3 21.5 19.1 23.8 31.9 41.3 83.6 ND 38.3 24.3

73 73 73 73 73 7 73 73 66 73 73 73 73 12 73 73

1 1 3 2 1 3 1 1 3 3 3 2 2 2 2 1

0.059 0.188 0.156 0.074 0.205 0.025

19.1 46.1 43.8 22.8 64.2 ND

66 73 73 73 73 7

1 1 2 1 3 2

Health value (mg/L)

Source

RQ

USEPA 2011 USEPA 2011 CDPH 2008 CDPH 2008

0.08 0.0002 0.02 0.00002

5 30 0.7 100 30 0.7 1 50 0.7 0.7 0.7 140 140 3.5 70 3

CDPH 2008 ADWG 2011 TTC USEPA 2011 ADWG 2011 TTC WHO 2011 WHO 2011 TTC TTC TCC CDPH 2010 CDPH 2010 OCS 2011 USEPA 2011 ADWG 2011

0.01 0.002 0.06 0.0006 0.0008 0.1 0.05 0.0008 0.05 0.1 0.3 0.001 0.002 0.09 0.002 0.02

0.4 0.7 770 0.3 7 260

WHO 2011 ADWG 2011 CDPH 2010 ADWG 2011 TTC CDPH 2010

0.15 0.3 0.0002 0.3 0.03 0.0001

1 200 1 1200

LOD: limit of detection; CASR No: registry number for each chemical, assigned by the Chemical Abstracts Service, a division of the American Chemical Society; SRU: Standard Relative Uncertainty; n: total number of samples; STE: secondary treated effluent; all values are expressed in mg/L; RQ: risk quotient; USEPA Drinking Water Standards and Health Advisories (USEPA, 2011); CDPH: California Department of Public Health (CDPH, 2008, 2010); Australian Drinking Water Guidelines (NHMRC, 2011); TTC: Threshold of Toxicological Concern; OCS: Office of Chemical Safety, Australian Government (Office of Chemical Safety, 2011); ND: not determined.

tetrachloroethene (87.9%), carbon disulfide (81.2%) and chloromethane (57.6%). Of the 21 VOCs detected, fourteen (14; 67%) were detected in less than 30% of the samples analysed, indicating that the presence of the compounds in STE is not consistent. Median concentrations for these compounds were dominated by non-detects, reported as their correspondent LOD. Seven (7) VOCs (i.e. 1,4-dichlorobenzene, cis-1,2-dichloroethene, carbon disulfide, chloromethane, tetrachloroethene, acrylonitrile and trichloroethene) had a percentage of detections greater than 30% across all samples (Table 2). For these compounds, median concentrations ranged from 0.81 mg/L for 1,4-dichlorobenzene to 0.02 mg/L for carbon disulphide. Comparison for the 7 VOCs with percentage detections greater than 30% showed that median concentrations were generally higher for influent STE from KWRP compared to BPP (ManneWhitney p ¼ 0.0001). For example, the median concentration of chloromethane at KWRP STE influent was double the median concentration at BBP STE influent (KWRP median ¼ 0.12 mg/L; BPP median ¼ 0.06 mg/L). The median concentration of tetrachloroethene in the STE entering KWRP was 5.4 times higher than the corresponding at BPP (KWRP median ¼ 2.2 mg/L; BPP median ¼ 0.41 mg/L). For 4 VOCs the concentrations in KWRP STE influent were statistically higher than BPP STE influent (Fig. 1), i.e. cis-1,2-dichloroethene

(ManneWhitney test, p ¼ 0.001), chloromethane ( p ¼ 0.004), tetrachloroethene ( p ¼ 0.02) and trichloroethene ( p ¼ 0.001). For 1,4-dichlorobenzene and acrylonitrile, the median concentrations were slightly higher at BPP STE influent but the differences were not statistically significant. Samples from Subiaco WWTP were not included in the comparison as fewer samples were taken at this location compared to KWRP and BPP. Seasonal comparison of median VOC concentrations is presented in Fig. 2. Again comparison is only made for compounds with percentage detections greater than 30% (i.e. 1,4-dichlorobenzene, cis-1,2-dichloroethene, carbon disulfide, chloromethane, tetrachloroethene and trichloroethene), except for acrylonitrile which was not included because it was only analysed during sampling event 3 (n ¼ 6, Table 2). Overall median VOC concentrations were higher in spring (0.125 mg/L) and winter (0.12 mg/L) than in summer (0.025 mg/L) and autumn (0.022 mg/L). These differences were statistically significant (Kruskal Wallis X2 p ¼ 0.0001). The VOCs: 1,4-dichlorobenzene, tetrachloroethene, and trichloroethene all had highest median concentrations in spring, whereas the highest median concentrations for cis-1,2dichloroethene and chloromethane were equal in spring and winter. The highest median concentration for carbon disulphide was recorded in winter, although it should be noted that it was not analysed in spring sampling events (Table 1).

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6

99

Fig. 1 e Median VOCs concentration in STE by plant in mg/L. *VOCs with statistically significant differences in concentrations between plants.

4.2.

Post-MF water

Twenty seven (27) VOCs were detected in post-MF samples, a higher number than in STE samples (21). Sixteen (16) of the 27 VOCs in the post-MF samples were also detected in STE (Table 2). Of the 27 VOCs detected in post-MF water, 18 were detected at KWRP only (Fig. 5) and 1 (i.e. 1,3-dichloropropene) was detected at BPP only. Eight (8) VOCs were detected at both locations: 1,2-dichlorobenzene, 1,4-dichlorobenzene, benzene, carbon disulfide, chloromethane, tetrachloroethene, toluene and o-xylene.

detections) followed by acrylonitrile (83.3%), chloromethane (62.9%) and carbon disulfide (40.0%), with respective median concentrations of 0.19 mg/L, 0.13 mg/L, 0.09 mg/L, and 0.02 mg/L. Five VOCs (i.e. ethyl benzene, naphthalene, n-butylbenzene, m-xylene and p-xylene) were detected in post-RO water and post-MF water but not in STE. Three VOCs were only detected in post-RO water (i.e. 1,2,3-trichlorobenzene, chlorobenzene and tert butylbenzene), although all with a percentage of detection below 11%. The percentage detection for all VOCs detected in post-RO water but not in STE was below 10% except for p-xylene (15.4%) (Table 2).

4.3.

4.4.

Post-RO water

A total of 26 VOCs (43%) were detected in post-RO water and 18 of these VOCs were also detected in STE (Table 2). The most commonly detected VOC was 1,4-dichlorobenzene (89.9%

Groundwater

Three (3) VOCs (i.e. 1,4-dichlorobenzene, benzene and toluene) were detected in groundwater. 1,4-dichlorobenzene (0.005 mg/L) was detected once in a sample taken from bore

Fig. 2 e Median VOCs concentration in STE by season in mg/L. * VOCs with statistically significant differences among seasons.

100

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6

VOCs IARC Classification 2, 3%

4, 7%

1: Carcinogenic to humans 10, 16% 27, 44%

2A: Probably carcinogenic to humans 2B: Possibly carcinogenic to humans 3: Unclassifiable as to carcinogenicity in humans NE: Not evaluated

18, 30%

Fig. 3 e Distribution of the VOCs analysed according to the IARC cancer classification.

4.5.

Screening health risk assessment

Of 61 VOCs analysed, 34 (55.7%) were classified in tier 1, 17 (27.9%) were classified in tier 2 and the remaining 10 (16.4%) were classified in tier 3. The list of VOCs analysed, the corresponding tier, health value and calculated RQs are presented in Tables 2 and 3. The VOCs were classified according to the IARC cancer classification and the USEPA cancer classification. Almost half of the VOCs analysed had not been evaluated (27, 44.6%). Of the 34 VOCs evaluated: two (2) are classified as carcinogenic to humans

1, 1 1, ,2-T 2, 4- rich tri lo 1, me roe 2- th D ylb tha 1, ichl en ne 4- or ze di op n c e 1, hlo rop 1, 2-d ro an 2b di ich en e ch lo ze r 1, lor oe ne 3- oe th di th an 1, chlo en e e 2r di ob , c ch en is lo ro zen be e Ac nze ry n lo e ni t r Br B i om en le C om zen ar bo eth e n a C dis ne hl D u ic hl Ch oro lfide e or od loro tha iflu m ne e D oro tha ic ne m hl or eth om an et e Te ha tra ne ch lo MT ro B et E he Tr n ic To e hl l u or oe ene th e o- ne xy le ne

-200

Efficiency (%) 0 -100

100

line A during sampling event 4 (January 2008). A replicate sample from bore line A and a sample from bore line B taken the same day were below the LOD (0.003 mg/L). Benzene was also detected above LOD (0.04 mg/L) in all groundwater samples taken during sampling event 4. Two replicate samples from bore line A were 0.08 mg/L and 0.13 mg/L, while the concentration in the single bore line B sample was 0.1 mg/L. Toluene (0.54 mg/L) was detected once in a sample from bore line B during sampling event 2. The toluene concentration of bore line A sample taken on the same day was below LOD (0.13 mg/L).

Fig. 4 e Reverse osmosis removal efficiency of VOCs detected in secondary treated effluent.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6

101

Fig. 5 e Median concentrations of VOCs in paired secondary treated effluent, post-MF water and post-RO water samples for KWRP and BPP.

(benzene and vinyl chloride); four (4) are classified as probably carcinogenic to humans (tetrachloroethene, trichloroethene, ethylene dibromide and 1,2,3-trichloropropane); ten (10) are classified as possible carcinogenic to humans (group 2B) and; eighteen (18) are unclassifiable as to carcinogenicity in humans (group 3) Fig. 3. No significant differences were observed in the IARC cancer classification distribution of the detected VOCs. Similarly, of the 61 VOCs analysed, 16 (26.2%) had not been evaluated by the USEPA. Two (2) VOCs are classified

as human carcinogens, seven (7) are classified as probable human carcinogens; seven (7) are classified as possible human carcinogens, and sixteen (16) are not classifiable as to human carcinogens. The RQ for the VOCs detected in any STE, post-MF or postRO sample are presented in Table 2. In STE samples, both RQ(max) and RQ(median) were always below 1. Most calculated RQ(max) were between one and three orders of magnitude below 1, whereas all RQ(median) were between one and

102

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6

four orders of magnitude below 1. However, RQ(max) was only slightly below 1 for chloroethane (RQ ¼ 0.8); tetrachloroethene (RQ ¼ 0.2); MTBE (RQ ¼ 0.13) and; benzene (RQ ¼ 0.1). RQs for VOCs detected in post-MF water but not in STE or post-RO water (Table 2) were all below 1 with the highest RQ(max) ¼ 0.3 for 2-propyltoluene. In post-RO water, both RQ(max) and RQ(median) were again consistently below 1. The highest RQ(max) was for benzene (0.14), with all other values between one and three orders of magnitude below 1. For RQ(median), all values were between one and four orders of magnitude below 1. The results indicate that chemical concentrations measured in post-RO water are not of human health concern. A total of 26 VOCs were not detected in any STE, post-MF or post-RO samples (Table 3). RQs for undetected VOCs were calculated using the average LOD as a worst case scenario. For twenty (20) of the undetected VOCs the calculated RQs were between one and four orders of magnitude below 1. For the other 6 undetected VOCs, RQs were slightly higher but all remained below 1 ranging from 0.11 (1,3-Dichloropropane) to 0.32 (2,2-dichloropropane). The human health risk from these VOCs is therefore estimated to be negligible. This screening health risk assessment is based on VOCs concentrations in water for human consumption. However, for a comprehensive risk assessment, it is necessary to consider other routes of exposure, including inhalation and dermal uptake. In the Perth context, VOC concentrations in post-RO water are very low and any potential human risk from inhalation, dermal contact and ingestion will be further minimised by the retention time of the injected recycled water into the confined aquifer. Consequently, no human health risk is anticipated at the VOCs concentrations detected in the postRO water. With respect to other environmental sources, it is likely that diet accounts for some VOC exposure (FlemingJones and Smith, 2003), with inhalation accounting for a larger portion of human intake through VOCs emitted by cigarette smoke, vehicles, household products and industrial pollution. Consequently, risks to human health from VOCs in recycled water for IPR are likely to be negligible with negligible impact on public health compared to other sources of exposure.

5.

Discussion

5.1.

VOC concentrations in STE

A total of 21 VOCs were detected in STE, of which 14 (67%) were detected in less than 30% of the samples analysed. The low concentration and inconsistent occurrence of VOCs in STE in our study may indicate that (i) adequate industrial waste acceptance criteria are in place to limit or prohibit discharge of substances from commercial or industrial premises in Perth, and/or; (ii) WWTPs are able to effectively remove VOCs from raw wastewater. The concentration of VOCs detected in STE depends on the relative importance of removal processes during the wastewater treatment, such as adsorption onto sludge, chemical transformation, volatilization, and biodegradation. The results are consistent with previous studies showing low concentrations of VOCs in STE attributable to

significant reductions during municipal wastewater treatment (more than 90% removal) (Wu et al., 2002). Significant differences in the median concentrations of VOCs were found between the STE influent of KWRP and BPP. This may be related to the differences in level of contamination of the wastewater in the catchments and geographical variability in industrial activities. The higher VOCs concentrations seen in STE samples at KWRP compared to BPP may be related to the fact that KWRP is on the site of an oil refinery. This finding reinforces the importance of wastewater characterisation for projects considering IPR, given the different nature of industry, trade waste agreements/regulations, sewer arrangements and WWTP process in place. Our results are consistent with other studies indicating that the VOCs detected in a WWTP are closely related to the industrial activities in the catchment. For example Cheng et al. (2008) found that the more common VOCs detected in STE were as follows: acetone, acrylonitrile, methylene chloride, and chloroform for the petrochemical districts; acetone, chloroform, and toluene for the science-based districts; and chlorinated and aromatic hydrocarbons for the multiple industrial districts (Cheng et al., 2008). In contrast, Fatone et al. (2011) found that BTEX compounds (excluding benzene) to be the most commonly detected VOCs in five municipal WWTPs, assumed to result from vehicle emissions. Seasonal differences in some VOC concentrations were also observed, with higher concentrations observed in winter and spring compared to summer and autumn. VOCs are more likely to be stable and detectable in cold water because warm temperatures can cause VOCs to volatilize (Metz et al., 2007) and to more readily undergo degradation by the activated sludge process (Martı´nez et al., 2006). This seasonality was consistent when duplicate seasonal sampling was undertaken in summer and winter. The results also correspond with findings from air pollution studies that report higher concentrations of VOCs in air during summer (Millet et al., 2005).

5.2.

VOC concentrations in groundwater

Three VOCs (i.e. 1,4-dichlorobenzene, benzene and toluene) were detected in groundwater, just above the LOD. Toluene and benzene may indicate potential VOC contamination of groundwater, as previously reported in association with landfills and leaking underground petrol storage tanks (Zogorski et al., 2006). Given the limited number of samples and low concentrations detected, further investigation is required to confirm the presence of these compounds in Perth’s groundwater. VOCs have been frequently detected in shallow ground-water beneath urban areas (up to 90% of samples) (Hamilton et al., 2004). Samples taken during this study were a mixture of groundwater from shallow aquifers and deep, confined aquifers. In general, deep aquifers are less vulnerable than shallow aquifers to anthropogenic contaminants that originate on or near the land surface. In other locations, VOC contamination has been observed in public wells which draw on proportionately large volumes of groundwater situated below developed areas (Zogorski et al., 2006). However, in Western Australia, all public drinking water bores are protected by catchment protection reserves.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6

5.3. The effect of MF/RO treatment on VOC concentration In both the BPP and KWRP, wastewater undergoes chloramination before MF to prevent RO membrane fouling. Samples of post-MF water were therefore analysed to determine the effect of chloramination during the MF/RO process. Paired wastewater, post-MF and post-RO samples were taken on 6 occasions at KWRP (1 in sampling event 1, 3 in sampling event 2, 1 in sampling event 3 and 1 in sampling event 6) and on 3 occasions at BPP (sampling events 3, 4 and 6). At KWRP, there were 21 analytes for which the highest median concentration was measured in a post-MF sample: chloromethane, trichlorofluoromethane (freon 11), carbon disulphide, cis-1,2-dichloroethene, trichloroethene, benzene, toluene, ethyl benzene, o-xylene, m-xylene, p-xylene, styrene, 1,3,5-trimethylbenzene, tert butylbenzene, 1,2,4trimethylbenzene 1,3-dichlorobenzene, 2-propyltoluene, pisopropyl toluene, 1,2-dichlorobenzene, n-butyl benzene, and naphthalene. Although the percentage detections for some of these analytes were low, 11 were present in more than 50% of post-MF samples: chloromethane (100%), cis-1,2dichloroethene (83%), toluene (83%), ethyl benzene (100%), oxylene (100%), m-xylene (100%), p-xylene (100%), styrene (67%), 1,3,5-trimethylbenzene (83%), 1,2,4-trimethylbenzene (100%), and naphthalene (83%). During this research project, chloramination was found to increase the concentration of some disinfection by-products (DBPs) during MF/RO treatment, particularly the halomethanes (DOHWA, 2009, Linge et al. in prep). Chloromethane is considered a disinfection by-product (Krasner et al., 2006) and may have also been formed by chloramination. However, cis-1,2-dichloroethene is a chlorinated solvent and would not be expected to be formed by disinfection. All of the other nine VOCs frequently detected in post-MF samples are aromatic compounds associated with gasoline or diesel exhaust (Elbir et al., 2007; Watson et al., 2001) or with oil refinery emissions (Chen et al., 2006; Scheff and Wadden, 1993). The KWRP MF/RO plant is located on the site of an oil refinery that produces petrol, diesel, liquefied petroleum gas, aviation gasoline, jet fuel and bitumen and it is likely that trace concentrations of associated compounds would be found in water samples from KWRP. The low concentrations measured in post-RO water and field blanks compared to post-MF water suggests that this contamination did not occur during sampling, but most likely occurred during the MF treatment where the water is exposed to the atmosphere for about 25 min. A number of the tanks also have air vents that may enable some exposure to the atmosphere both before MF and after RO of up to an hour, although the vents are less likely to be a significant source of exposure compared with the open tanks. At BPP, there were only 4 VOCs for which the highest median concentration was measured in a post-MF sample: i.e. carbon disulphide (66%), toluene (100%), 1,3-dichloropropene (33%), and 1,2-dichlorobenzene (66%). While they are not typically measured as disinfection by-products, dichlorobenzenes are produced from the chlorination of benzene (IARC, 1999) and therefore it is possible that they may form, petroleum-based contamination. However the source of these VOCs is not obvious at this time.

5.4.

103

Treatment performance

Overall treatment efficiency was calculated as a proportion of removal, comparing STE and post-RO samples that were matched for plant and date. For those parameters reported below LOD after RO, the efficiency was calculated assuming a concentration equal to half the LOD. Very high variability in the removal of VOCs was observed, as illustrated in Fig. 4. For ten (10) of the twenty one (21) VOCs detected in STE, the median removal efficiency was above 70%. The median removal efficiency ranged from
77.0% for dichlorodifluoromethane to 91.2% for tetrachloroethene. For 17 samples, corresponding to 6 VOCs (1,4-dichlorobenzene ¼ 1 sample, benzene ¼ 4, carbon disulfide ¼ 3, chloromethane ¼ 7, dichlorodifluoromethane ¼ 1, o-xylene ¼ 1), the concentrations in post-RO samples were higher than their paired STE samples. For 10 (59%) of these paired samples, the concentration in post-RO water was not statistically different from the STE: the difference was within the uncertainty of the analytical method and therefore calculation of removal efficiency using these data is inconclusive. Differences were seen (outside of the limits of uncertainty) for carbon disulphide (1 sample), chloromethane (4 samples), dichlorodifluoromethane (1 sample), and o-xylene (1 sample). As discussed in Section 4.3, elevated concentrations of carbon disulphide and o-xylene occurred in post-MF samples, due to atmospheric exposure to oil refinery emissions during MF treatment at KWRP. While post-RO water did not have similar levels of atmospheric exposure, the volatile nature of the compounds studied mean it is possible that there were individual occasions where trace contamination from atmospheric exposure occurred. Therefore the increased concentrations seen in a few post-RO samples may be a result of similar atmospheric exposure. Chloromethane was also elevated in post-MF samples. Reverse osmosis is the only barrier in the treatment process for chemical removal of VOCs and it is apparent that calculation of RO chemical removal by comparing STE and post-RO samples can be confounded by potential impacts from chloramination and atmospheric contamination, as described above. Thus the treatment efficiency of RO alone was determined using paired post-MF and post-RO samples by plant (Fig. 6). Calculations confirmed that the degree of RO treatment efficiency was higher than overall recycling plant treatment efficiency for all VOCs detected in STE with the exception of chloroethane, tetrachloroethene, toluene and dichlorodifluoromethane. Furthermore, by using post-MF data, treatment efficiency could be calculated for 10 additional VOCs, which were measured in post-MF samples but were not present in STE. As shown in Fig. 6, the majority of positive VOCs in post-MF water occurs at KWRP (n ¼ 26) compared to BPP (n ¼ 9). Variability in RO treatment efficiency (calculated using post-MF and post-RO data) as represented by standard deviation, fell slightly but remained relatively high, although this may relate to the smaller number of paired samples available for analysis. The rejection of VOCs in a range of different RO membranes has been reported as highly variable. For some VOCs (1,1,1-trichloroethane, carbon tetrachloride, p-, m- and o-xylenes, tetrachloroethylene, 1,2-dichlorobenzene),

104

60 40

1, 2, 4t 1, rim 2- et D h 1 ic ylb 1, ,2- hlo en 2- di ro ze 1, dic chlo pro ne 3, hl ro p 5- or b an tr oe e e 1, ime the nze 3- th n n 1, Dic ylb e, e 3- h e ci d lo n s 1, ich ro zen 4- lo pr e di ro op ch be e lo n n 2- ro ze e p r be n e op nz yl en to Br B lue e om e ne C o nz ar m e bo e ne t C n d han hl is e C or ul hl oe fid or t e o h Et me ane hy th l a N ben ne ap z ht en Te ha e tra ch S len lo tyr e ro en Tr et e ic T he hl ric or h To ne of lo lu lu ro e or et ne om he e ne n- m tha b u - ne ty xyl lb en pen e Is op o ze ro -x ne py yle lto n lu e p- en xy e le ne

0

20

Efficiency (%)

80

100

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6

Beenyup Pilot Plant

Kwinana Water Reclamation Plant

Fig. 6 e RO removal efficiency of VOCs detected in post-MF wastewater by plant.

rejection has been reported as higher than 90% (Agenson et al., 2003), but for other VOCs rejection has been reported as much lower (e.g. 6e54% for benzene). Using paired post-MF and post-RO samples, 56% of analytes in this study had rejection efficiency greater than 80% e including 1,4-dichlorobenzene (87%) , while 37% had rejection efficiency greater than 90%. Lower median rejection was associated with benzene (56%), bromomethane (48%) and styrene (63%). It has been found that rejection of VOCs is influenced by solute size, molecular charge, branching of functional groups and Kow (Agenson et al., 2003). VOCs with poorer rejection usually have smaller molecular width and length, and lower Kow. VOCs are not highly hydrophobic (logKow < 3), and Kow has been found to influence rejection, which suggests that there is some degree of interaction between the solute and the membrane.

5.5.

Identification of treatment performance Indicator

The use of chemical indicators in recycled water has been proposed for occurrence monitoring and for assessing treatment process performance assessment (Benotti et al., 2009; Dickenson et al., 2011; Drewes et al., 2008). A treatment performance indicator is an individual chemical occurring at quantifiable level that represents a family of trace constituents with certain physicochemical and biodegradable characteristics that are relevant to fate and transport during treatment, in this case RO. A treatment performance indicator should provide a conservative assessment of removal of represented parameters, and should be sensitive to minor changes in RO treatment performance.

The criteria for selection of a Treatment Performance Indicator are:  Quantifiable using an established and preferably accredited analytical method;  Frequently detected in feed water, preferably 100% detections;  Present in feed water at significant concentrations; generally greater than five times LOD; The selection of a chemical indicator for VOCs is challenging given the diversity of compounds in this chemical group. However, it is neither practical nor feasible to assess for all potential VOCs present in recycled water during routing monitoring. In our study, 1,4-dichlorobenzene was identified as a potential treatment performance indicator for assessment of VOCs in IPR using RO treatment (Blair et al., 2010). 1,4-dichlorobenzene was detected in almost 94% of the STE samples and with a median concentration that was over 30 times the average method LOD, therefore fulfilling the criteria for selection of a Treatment Performance Indicator. While the median treatment efficiency by MF/RO treatment was relatively high (87%), the low method LOD meant that 1,4-dichlorobenzene was still detected in almost 89% of the post-RO water samples, ensuring accurate estimates of treatment performance could be calculated. The high frequency of detection of 1,4-dichlorobenzene in STE is attributed to its long history of domestic use in toilet products, moth repellents, and mildew control agents (Aronson et al., 2007; NICNAS, 2000). Dichlorobenzene has also been

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6

reported as one of the less biodegradable VOCs in wastewater treatment (Koe and Shen, 1997). As a small (molecular weight ¼ 147 g/mol) and uncharged molecule, the RO rejection of 1,4-dichlorobenzene is mostly attributed to its relatively high logKow (3.4) compared to other VOCs. The relevance of 1,4-dichlorobenzene as a chemical indicator of VOCs will be further validated during the Groundwater Replenishment Trial (GWRT) (DOHWA, 2009). During the three year trial, 8 ML/day of STE from the Beenyup WWTP will be treated by MF/RO and ultra violet disinfection before injection into a confined aquifer. If successful, this approach could significantly reduce the analytical cost and complexity of monitoring water treatment systems performance for removal of VOCs.

6.

Conclusions

The screening health risk assessment indicates that the individual VOCs measured in recycled water have a low potential to affect humans from long-term consumption after RO treatment. Detection of VOCs in STE can occur as a result of the widespread use of these compounds. However, the impact on potable supplies through augmentation with recycled water treated by MF/RO is likely to be negligible at the concentrations observed in Perth. Calculated MF/RO treatment removal was variable, with some concentrations in post-RO water higher than in the STE. For some VOCs, this may be due to uncertainty in the analytical method. However, for others it is attributed to industrial contamination during the MF/RO process, or formation during chloramination. For most VOCs, RO treatment efficiency was higher than overall MF/RO treatment efficiency, however more analysis of VOCs before and after RO treatment is recommended to better characterise the RO treatment variability. Management of risks in IPR is dependent on advanced treatment technologies and comprehensive risk management approaches to ensure compliance with drinking water guidelines. Frequent monitoring of a treatment performance indicator such as 1,4-dichlorobenzene (in conjunction with the continuous online monitoring of critical control points) during the MF/RO treatment process is likely to be sufficient to ensure adequate removal of VOCs. If successful, this approach has the potential for significantly reducing the analytical costs and complexity of monitoring water treatment systems performance for removal of VOCs, and other chemical groups.

Acknowledgements This research was conducted as part of the Premiers Collaborative Research Project (PCRP) a collaborative effort between Department of Health, Department of Water, Department of Environment & Conservation, Water Corporation of Western Australia, The University of Western Australia, Curtin University of Technology, Chemistry Centre WA and the National Measurement Institute (NMI). State

105

government funding was provided by the Office of Science & Innovation.

references

Agenson, K., Oh, J.-I., Urase, T., 2003. Retention of a wide variety of organic pollutants by different nanofiltration/reverse osmosis membranes: controlling parameters of process. Journal of Membrane Science 225 (1e2), 91e103. Aronson, D., Bosch, S., Gray, D.A., Howard, P., Guiney, P., 2007. A comparative human health risk assessment of pdichlorobenzene-based toilet rimblock products versus fragrance/surfactant-based alternatives. Journal of Toxicology and Environmental Health Part B: Critical Reviews 10 (7), 467e526. Atasoy, E., D, T., K, S., 2004. The estimation of NMVOC emissions from an urban-scale wastewater treatment plant. Water Research 38 (14e15), 3265e3274. Battistoni, P., Cola, E., Fatone, F., Bolzonella, D., Eusebi, A.L., 2007. Micropollutants removal and operating strategies in ultrafiltration membrane systems for municipal wastewater treatment. Industrial & Engineering Chemistry Research 46 (21), 6716e6723. Benotti, M., Trenholm, R., Vanderford, B., Holady, J., Stanford, B., Snyder, S., 2009. Pharmaceuticals and endocrine disrupting compounds in U.S. drinking water. Environmental Science & Technology 43 (3), 597e603. Blair, P., Linge, K., Rodriguez, C., Busetti, F., Drewes, J., Turner, N., 2010. Development of the System to Manage Chemical Risk During Indirect Potable Reuse. Proceedings Water Reuse & Desalination Water Reuse Association, Sydney. Boyes, W.K., Bushnell, P.J., Crofton, K.M., Evans, M., Simmons, J.E., 2000. Neurotoxic and pharmacokinetic responses to trichloroethylene as a function of exposure scenario. Environmental Health Perspectives 108 (Suppl 2), 317e322. Brouwer, D.H., de Pater, N.A., Zomer, C., Lurvink, M.W., van Hemmen, J.J., 2005. An experimental study to investigate the feasibility to classify paints according to neurotoxicological risks: occupational air requirement (OAR) and indoor use of alkyd paints. Annals of Occupational Hygiene 49 (5), 443e451. CDPH, 2008. Maximum Contaminant Levels and Regulatory Dates for Drinking Water U.S. EPA vs California, California Department of Public Health, California. CDPH, 2010. Drinking Water Notification Levels and Response Levels. California Department of Public Health, California. Chen, C.-L., Shu, C.-M., Fang, H.-Y., 2006. Location and characterization of emission sources for airborne volatile organic compounds inside a refinery in Taiwan. Environmental Monitoring and Assessment 120, 487e498. Cheng, W.-H., Hsu, S.-K., Chou, M.-S., 2008. Volatile organic compound emissions from wastewater treatment plants in Taiwan: legal regulations and costs of control. Journal of Environmental Management 88 (4), 1485e1494. Dickenson, E.R.V., Snyder, S.A., Sedlak, D.L., Drewes, J.E., 2011. Indicator compounds for assessment of wastewater effluent contributions to flow and water quality. Water Research 45 (3), 1199e1212. DOHWA, 2009. Premier’s Collaborative Research Program (20052008). Characterising Treated Wastewater for Drinking Purposes Following Reverse Osmosis Treatment. Department of Health, Western Australia (DOHWA), Perth, WA. Drewes, J.E., Sedlak, D., Snyder, S., Dickenson, E., 2008. Development of Indicators and Surrogates for Chemical Contaminant Removal During Wastewater Treatment and Reclamation. Water Reuse Foundation, Alexandria, VA.

106

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6

Elbir, T., Cetin, B., Cetin, E., Bayram, A., Odabasi, M., 2007. Characterization of volatile organic compounds (VOCs) and their sources in the air of Izmir, Turkey. Environmental Monitoring and Assessment 133 (1e3), 149e160. European Commission, 1998. Guidelines of water intended for human consumption. Official Journal of the European Communities. Fatone, F., Di Fabio, S., Bolzonella, D., Cecchi, F., 2011. Fate of aromatic hydrocarbons in Italian municipal wastewater systems: an overview of wastewater treatment using conventional activated-sludge processes (CASP) and membrane bioreactors (MBRs). Water Research 45 (1), 93e104. Fleming-Jones, M.E., Smith, R., 2003. Volatile organic compounds in foods: a five year study. Journal of Agricultural and Food Chemistry 51, 8120e8127. FPT Committee on Drinking Water, 2010. Guidelines for Canadian Drinking Water Quality. Federal-Provincial-Territorial Committee on Drinking Water, Ontario. Hamilton, P.A., Miller, T.L., Myers, D.N., 2004. Water Quality in the Nation’s Streams and AquiferseOverview of Selected Findings, 1991e2001. U.S.Geologycal Survey, Reston, Va, p. 20. Herpin, G., Gargouri, I., Gauchard, G.C., Nisse, C., Khadhraoui, M., Elleuch, B., Zmirou-Navier, D., Perrin, P.P., 2009. Effect of chronic and subchronic organic solvents exposure on balance control of workers in plant manufacturing adhesive materials. Neurotoxicity Research 15 (2), 179e186. IARC, 1999. Some Chemicals that Cause Tumours of the Kidney or Urinary Bladder in Rodents and Some Other Substances, p. 674. International Agency for Research on Cancer, Lyon, France. IRIS 1,1,2,2-Tetrachloroethane; CASRN 79-34-5; 09/30/2010 USEPA, Integrated Risk Information System. IRIS Carbon disulfide; CASRN 75-15-0; 09/30/2010 USEPA, Integrated Risk Information System. Koe, L.C.C., Shen, W., 1997. High resolution GC e MS analysis of VOCs in wastewater and sludge. Environmental Monitoring and Assessment 44 (1), 549e561. Krasner, S.W., Weinberg, H.S., Richardson, S.D., Pastor, S.J., Chinn, R., Sclimenti, M.J., Onstad, G.D., Thruston Jr., A.D., 2006. Occurrence of a new generation of disinfection byproducts. Environmental Science & Technology 40 (23), 7175e7185. Linge, K.L., Blythe, J.W., Busetti, F., Blair, P. and Heitz, A. (in prep) Formation of halogenated disinfection by-products during wastewater recycling employing microfiltration and reverse osmosis. Water Research. Martı´nez, S.A., Rodrı´guez, M.G., Morales, M.A., 2006. Stability analysis of an activated sludge bioreactor at a petrochemical plant at different temperatures. International Journal of Chemical Reactor Engineering 3. Metz, P.A., Delzer, G.C., Berndt, M.P., Crandall, C.A., Toccalino, P. L., 2007. Anthropogenic Organic Compounds in Ground Water and Finished Water of Community Water Systems in the Northern Tampa Bay Area, Florida, 2002e04, p. 48Report 2006e5267. U.S. Geological Survey Scientific Investigations, Reston, Virginia. Millet, D.B., Donahue, N.M., Pandis, S.N., Polidori, A., Stanier, C.O., Turpin, B.J., Goldstein, A.H., 2005. Atmospheric volatile organic compound measurements during the Pittsburgh Air Quality Study: results, interpretation, and quantification of primary and secondary contributions. Journal of Geophysical Research 110 D07S07. NHMRC, 2004. Australian Drinking Water Guidelines, p. 615. National Health and Medical Research Council and Natural Resource Management Ministerial Council, Artarmon, NSW.

NHMRC, 2011. Australian Drinking Water Guidelines. National Health and Medical Research Council and Natural Resource Management Ministerial Council, Canberra ACT. NICNAS, 2000. In: Australia, C.o. (Ed.), Summary: Priority existing chemical assessment reports. PEC No. 13. paradichlorobenzene. National Industrial Chemicals Notification and Assessment Scheme, Australian Goverment, Canberra ACT. NRC, 1996. Use of Reclaimed Water and Sludge in Food Crop Production. National Academic Press, Washington. NWC, 2011. Environmental Health Management. National Water Commission, Australian Goverment, Canberra ACT. Office of Chemical Safety, 2011. Acceptable Daily Intakes for Agricultural and Veterinary Chemicals. Office of Health Protection, Department of Health and Ageing, Canberra. Patterson, B.M., Pribac, F., Barber, C., Davis, G.B., Gibbs, R., 1993. Biodegradation and retardation of PCE and BTEX compounds in aquifer material from Western Australia using large-scale columns. Journal of Contaminant Hydrology 14 (3e4), 261e278. Rodriguez, C., Cook, A., Van Buynder, P., Devine, B., Weinstein, P., 2007a. Screening health risk assessment of micropollutants for indirect potable reuse schemes: a three-tiered approach. Water Science and Technology 56 (11), 35e42. Rodriguez, C., Weinstein, P., Cook, A., Devine, B., Buynder, P.V., 2007b. A proposed approach for the assessment of chemicals in indirect potable reuse schemes. Journal of Toxicology and Environmental Health, Part A: Current Issues 70 (19), 1654e1663. Scheff, P.A., Wadden, R.A., 1993. Receptor modeling of volatile organic compounds. 1. Emission inventory and validation. Environmental Science & Technology 27 (4), 617e625. Stata Corp, 2007. In: C. (Ed.), Stata Statistical Software Station. TX: StataCorp LP, Texas. Toccalino, P.L., Rowe, B.L., Norman, J.E., 2006. Volatile Organic Compounds in the Nation’s Drinking-water Supply Wells e what Findings May Mean to Human Health. U.S. Geological Survey, Sacramento, California, p. 4. USEPA, 1994. Emerging Technology Report Cross-flow Pervaporation System for Removal of VOCs from Contaminated Wastewater, p. 129. Risk Reduction Engineering Laboratory, Office of Research and Development, U.S. EPA, Cincinnati, Ohio. USEPA, 2011. 2011 Edition of the Drinking Water Standards and Health Advisories. Office of Water U.S. Environmental Protection Agency, Washington, DC. Watson, J.G., Chow, J.C., Fujita, E.M., 2001. Review of volatile organic compound source apportionment by chemical mass balance. Atmospheric Environment 35 (9), 1567e1584. WHO, 2011. Guidelines for Drinking-water Quality. World Health Organization, Geneva, Switzerland. Williams, P., Benton, L., Warmerdam, J., Sheehans, P., 2002. Comparative risk analysis of six volatile organic compounds in California drinking water. Environmental Science & Technology 36 (22), 4721e4728. Wright, M., Belitz, K., Burton, C., 2005. California GAMA Program: Ground-water Quality Data in the San Diego Drainages Hydrogeologic Province. U.S. Geological Survey, Sacramento, California, p. 102. Wu, C., Lu, J., Lo, J., 2002. Analysis of volatile organic compounds in wastewater during various stages of treatment for hightech industries. Chromatographia 56 (1), 91e98. Zogorski, J., Carter, J.M., Ivahnenko, T., Lapham, W.W., Moran, M.J., Rowe, B.L., Squillace, P.J., Toccalino, P.L., 2006. Report on Volatile Organic Compounds in the Nation’s Ground Water and Drinking-water Supply Wells, p. 112. U.S. Department of the Interior, U.S. Geological Survey, Reston, Virginia.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 0 7 e1 1 4

Available online at www.sciencedirect.com

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

Evaluation of the flocculation performance of carboxymethyl chitosan-graft-polyacrylamide, a novel amphoteric chemically bonded composite flocculant Zhen Yang a, Bo Yuan a, Xin Huang a, Junyu Zhou a, Jun Cai a, Hu Yang a,*, Aimin Li b, Rongshi Cheng a a

Key Laboratory for Mesoscopic Chemistry of MOE, Department of Polymer Science & Technology, School of Chemistry & Chemical Engineering, Nanjing University, Nanjing 210093, PR China b State Key Laboratory of Pollution Control and Resource Reuse, School of the Environment, Nanjing University, Nanjing 210093, PR China

article info

abstract

Article history:

In the present work, a novel amphoteric chemically bonded composite flocculant (car-

Received 14 July 2011

boxymethyl chitosan-graft-polyacrylamide, denoted as CMC-g-PAM) was successfully

Received in revised form

prepared and used to flocculate the kaolin suspension. The flocculation performance of

22 September 2011

CMC-g-PAM in acidic, neutral, and alkaline conditions was systematically evaluated by

Accepted 16 October 2011

light scattering in combination with fractal theory, as well as by traditional turbidity and

Available online 25 October 2011

zeta potential measurements. Based on the experimental facts from in situ size and fractal dimension measurements, different flocculation mechanisms play key roles at various pH levels, resulting in substantially varied flocculation kinetic processes under three pH

Keywords: Amphoteric

chemically

bonded

conditions. In acidic condition, patching was the main mechanism involved in the opposite

composite flocculant

zeta potential between CMC-g-PAM and the kaolin suspension. A flat configuration was

Flocculation kinetics

favored when the polymeric flocculant was adsorbed onto the particle surface, leading to

Flocculation mechanism

a slower initial floc growth rate but larger and denser flocs. Bridging was the dominant

Fractal dimension

mechanism in neutral and alkaline conditions. A faster initial rate of bridging resulted in

Light scattering

smaller and more open floc structures. A rearrangement process in neutral pH subsequently led to more compact flocs, whereas no restructuration of flocs occurred in alkaline conditions because of the electrostatic repulsion of the same negative charges on the flocculant and particles. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Flocculation is an important industrial process for solideliquid separation during the primary purification of wastewater. Attention has been increasingly paid to natural polymer flocculants owing to their biodegradability, wide availability, and environment-friendliness (Renault et al., 2009). Chitosan, poly-b(1 / 4)-2-amino-2-deoxy-D-glucose, obtained from the deacetylation of chitin (the second most abundant natural polymer), is

one of the most outstanding candidates among these flocculants (Guibal et al., 2006; Renault et al., 2009). This polymer presents an abundance of free amino groups along its backbone chain that are positively charged in acidic media and show prominent flocculation performance. However, the poor water solubility (Wang et al., 2009) of chitosan at neutral or higher pH remains a limitation in its practical application. Given that this polymer has high tailorability for an abundant number of functional groups on its

* Corresponding author. Tel.: þ86 25 83686350; fax: þ86 25 83317761. E-mail address: [email protected] (H. Yang). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.024

108

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 0 7 e1 1 4

backbone, chemically modified chitosan materials have been manufactured to overcome this limitation. Among these materials, amphoteric chitosan (Yang et al., 2011a, b), containing both cations and anions on the backbone chain, can easily convert chitosan into a water-soluble form over wide pH ranges. Considering that chitosan itself is a cationic polyelectrolyte, the simplest amphoteric chitosan is carboxyalkyl chitosan (Chen and Park, 2003). However, at high pH, carboxyalkylated chitosan also suffers from a drawback: enhanced electrostatic repulsion against negatively charged particles that commonly exist in natural turbid water. Grafting provides an effective method (Sand et al., 2010; Wang et al., 2008) for overcoming this limitation. Synthetic polymeric flocculants, such as polyacrylamide (PAM), have already been proven to be useful in water purification, although they exhibit defects such as poor shear resistance and nonbiodegradability (Biswal and Singh, 2004; Sen and Pal, 2009). Therefore, amphoteric chemically bonded composite flocculants of natural and synthetic polymers are expected to combine the advantages of both variants (Biswal and Singh, 2004; Sen et al., 2009) through graft copolymerization. On the one hand, flexible synthetic polymer chains on comb-like flocculants (Sen et al., 2009) have better approachability (Tripathy et al., 2009) than carboxyalkylated chitosan to suspended particles in water. On the other hand, compared with synthetic polymers, new tailor-made chitosan-based materials can be shear stable and controllably biodegradable (Biswal and Singh, 2004). Generally, for the aforementioned amphoteric flocculants, different pH levels result in various changes in the densities and morphologies of the polymer chains (Yang et al., 2011b). Hence, similar to the fact that various Al(III) species of aluminum-containing flocculants at various pH levels lead to different flocculation mechanisms (Lin et al., 2008), amphoteric flocculants at various pH levels also undergo diverse flocculation mechanisms. Thus, a good understanding of these mechanisms is important for industrial applications to optimize the flocculation process (Yu et al., 2010). Traditional approaches for studying flocculation mechanisms are usually indirect (Zhou and Franks, 2006), and use finalresult-based methods (Yu et al., 2006), such as turbidity, sedimentation rate, or zeta potential (ZP). In these methods, only the overall parameters are obtained, and no direct information about floc properties is measured. This is despite the fact that floc properties are crucial physical properties that should be studied to obtain a better understanding of the flocculation mechanism as well as control the separation process (Fabrizi et al., 2010; Wei et al., 2009). In past decades, several microscopic approaches (Thomas et al., 1999) to model the floc properties and flocculation process have been formulated based on Smoluchowski’s collision theory (von Smoluchowski, 1917). However, considering that the principal assumptions of this theory cannot be satisfied in a real flocculation process, these approaches present a number of shortcomings (Thomas et al., 1999). Therefore, an alternative way is to focus on macroscopic measurements. One such approach is the fractal theory, which considers that irregular and porous flocs have been found to be geometrically fractal during flocculation (Mandelbrot, 1983). In the fractal concept, the most important and powerful quantitative parameter is the fractal dimension (Zhou and Franks, 2006), which indicates the space-

filling capacity (Thomas et al., 1999), that is, the compactness, of the floc. Larger fractal dimensions signify more compact flocs, which are usually preferred in most situations in water treatment to yield lower sludge volumes and easier sedimentation. Fractal dimensions can be expressed depending on the space dimension. The two commonly used fractal dimensions are two-(D2) and three-dimensional (i.e., mass fractal dimension, or DF) presentations. D2 is defined by the power law relationship between the projected area (A) and the characteristic length (l ), which is usually measured by image analysis (IA) (Liao et al., 2006). DF is the power law relationship between mass (m) and l, which can be obtained by light scattering (LS). Compared with IA, LS is more time-saving, simpler, and more accurate. LS tests can also easily supply average size and size distribution, as well as DF, in a continuous manner, if the equipment is pre-programmed for continuous data acquisition (Rasteiro et al., 2008). Therefore, the flocculation kinetics (i.e., the complicated stages of a flocculation process), which has been recognized as more deeply related to flocculation mechanism (Yu et al., 2006; Zhou and Franks, 2006), can be evaluated together in a single experiment. In the present study, a novel amphoteric chemically bonded composite flocculant, carboxymethyl chitosan-graftpolyacrylamide (CMC-g-PAM), is prepared. Given that the new composite flocculant has better solubility over a wide pH range than other flocculants, the flocculation performance of CMC-g-PAM is investigated using turbidity and ZP methods at various pH levels. Considering that few studies on the application of fractal theory on biodegradable natural polymerbased flocculants have been reported, IA and LS methods are introduced to evaluate the flocculation performance of CMCg-PAM. Using online data acquisition of LS, the flocculation kinetics and mechanisms at various pH levels are studied in detail, and various stages during the flocculation process are finally proposed.

2.

Materials and methods

2.1.

Materials

Chitosan, with 85.2% degree of deacetylation and 8.34  105 g/mol viscosity-average molecular weight as calculated from the intrinsic viscosity (Wang et al., 1991), was purchased from Shangdong Aokang Biological Co., Ltd. Monochloroacetic acid (Zibo Lushuo Economic Trade Co., Ltd.), acrylamide (Nanjing Chemical Reagent Co., Ltd.), and ceric ammonium nitrate (CAN) (Sinopharm Chemical Reagent Co., Ltd.) were used without further purification. Kaolin, with a mean particle size of 7.4 mm, was purchased from Sinopharm Chemical Reagent Co., Ltd. Polyaluminium chloride (PAC) and PAM (weight-average molecular weight: 1.20  107 g/mol) were purchased from Guangzhou Yurun Chemical Technology Co., Ltd. All other chemicals were purchased from Nanjing Chemical Reagent Co., Ltd.

2.2.

Preparation of CMC-g-PAM

The preparation route of CMC-g-PAM is shown in Fig. S1 of the supporting information. CMC was synthesized using the same method as described in the authors’ previous work (Yang et al.,

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 0 7 e1 1 4

2011b), and the substitution degree of the carboxymethyl groups was 48.3%, as calculated from 1H nuclear magnetic resonance (1H NMR) spectrum. Graft copolymerization for the preparation of CMC-g-PAM was performed as follows: A specific amount of solid CMC was dissolved in 400 mL of 1 wt% hydrochloric acid. After stirring under N2 for 30 min, a certain amount of CAN as the initiator was added into the CMC solution. The solution was then kept for 5 min for pretreatment of the CMC by the initiator, followed by the dropwise addition of an acrylamide monomer aqueous solution into the bottle for 20 min. The pretreatment of CMC by the initiator could efficiently suppress the formation of the PAM homopolymer (Sonmez et al., 2002). After 3 h of reaction under N2, copolymerization was stopped, and the samples were precipitated in acetone. The solid product was filtered, washed, extracted using acetone as the solvent in a Soxhlet apparatus for 72 h to remove impurities in the sample (Ceresa, 1973; Sonmez et al., 2002; da Silva et al., 2007; Ali and Singh, 2009) and then finally vacuum dried at room temperature. The final product was CMC-g-PAM, with a weight-average molecular weight of 3.10  106 g/mol, as calculated from its 1H NMR spectrum.

2.3.

Characterization

Fourier transform infrared (FTIR) spectra were recorded using a Bruker Model IFS 66/S FTIR spectrometer. The interval of tested wave numbers was 650e4000 cm1. 1H NMR spectra were recorded on a Bruker AVANCE Model DRX-500 spectrometer, operating at 500 MHz, in a mixed solvent composed of CF3COOD and D2O with a mass ratio of 1:1. ZP was measured using a Malvern Model Nano-Z Zetasizer. Hydrodynamic radius (Rh), a parameter commonly used to characterize the size of the polymer chains in solution, was determined by a Brookheaven Model BI200SM dynamic light scattering apparatus.

2.4.

Flocculation experiment

2.4.1.

Flocculation procedure

Floc size measurement

Floc sizes were analyzed by LS using a Malvern Mastersizer 2000 system. Jar tests were conducted in the same process as described in Section 2.4.1. The suspension was measured by continuous recycling of water flowing through the sample cell of the instrument. The apparatus was settled as described in the work of Bushell and Amal (2000). Floc size was defined by volume-weighted mean diameter (Dv), calculated and provided by the data processing software, P 4 d Dv ¼ P 3 (2) d where d is the diameter of the equivalent sphere of each single floc, according to the Mie scattering theory.

2.4.3.

Determination of fractal dimension

D2 was measured by IA and determined from the slope of the logelog plot of A and l (Chakraborti et al., 2000, 2003) AflD2

(3)

In the present study, the largest projection length was taken as l. Afterward, the flocs were carefully withdrawn from the jar and placed into a glass dish with water. An optical microscope (model: Leica DMLP) equipped with a digital camera (Victor Company of Japan, Ltd.) was used to take photos. A and l were derived from the photos using Imagepro Plus 6.0 software. DF was measured by LS. The theory of the LS technique has been reported in the literature (Rasteiro et al., 2008; Yu et al., 2006; Zhou and Franks, 2006) and was used in the present study. In this technique, two parameters, light intensity (I ) and scatter vector (Q), were measured. Q is given by the following equation (Schaefer et al., 1984): Q¼

4pnsinðq=2Þ l

(4)

where n, q, and l are the refractive indices of the medium, scattered angle, and wavelength of radiation, respectively. DF was then determined from the negative slope of logelog plot of I and Q (Jarvis et al., 2005),

Kaolin (1 g) was added into 1 L of distilled water. After 3 min of ultrasonic stirring at 200 rpm, the suspension was used as synthetic water. The pH levels were modified by adding HCl or NaOH solution. The stock solutions of the flocculants were always freshly prepared in distilled water before each flocculation test. Jar tests were conducted using 1-L jars and a six-place programmed paddle mixer model of TA6 (Wuhan Hengling Tech. Co. Ltd.) at room temperature. The detailed procedure consisted of an initial period of rapid mixing for 5 min at 200 rpm, followed by 15 min of slow mixing at 50 rpm, and finally settling for 40 min. After this procedure, samples collected from the supernate were analyzed for turbidity and ZP. Turbidity was measured by the Turbidity Indicator model of ATZ-A22 (Wuxi Guangming Instrument Factory). The residual turbidity percentage (%) is expressed as Residual turbidity percentage ð%Þ ¼ ðTtreated =Traw Þ  100%

2.4.2.

109

(1)

where Traw and Ttreated are the turbidities of raw and treated water, respectively.

IfQ DF

(5)

3.

Results and discussion

3.1.

Characterization of CMC-g-PAM

The FTIR and 1H NMR spectra of chitosan, CMC, and CMC-gPAM are shown in Figs. S2 and S3 of the supporting information, respectively. Aside from the characteristic FTIR peaks of chitosan in Fig. S2(i), the new peak at 1584 cm1 for COO (Chen and Park, 2003) in Fig. S2(ii) indicates that CMC is successfully prepared. In Fig. S2(iii), the new strong band around 1651 cm1 is assigned to amide-I (CeO stretching) on the PAM chain (Yuan et al., 2010). The bands of COO on the CMC backbone and amide-II (NeH bending) on the PAM chain overlap with each other and result in a peak at 1601 cm1 (Biswal and Singh, 2004; Yuan et al., 2010). The new peak at approximately 1447 cm1 is due to CeN stretching in the graft copolymer (da Silva et al., 2007; Yuan et al., 2010). Since the

110

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 0 7 e1 1 4

homopolymer of PAM is removed during the preparation process as mentioned above, the detected PAM chain should be chemically bonded to the chain backbone of CMC. The 1H NMR spectra (Fig. S3) also confirm the final structure of the amphoteric flocculant product, where PAM is successfully grafted onto the CMC. Aside from the signal of protons from chitosan in Fig. S3(i), the intense resonance at 3.93 ppm in Fig. S3(ii) is attributed to the protons on the carboxymethyl groups (Chen and Park, 2003). The new strong peaks at around 2.10 and 1.50 ppm in Fig. S3(iii) are the resonances of methylene and methine protons in the PAM chain (Yuan et al., 2010), respectively. According to the 1H NMR spectrum (Fig. S3(iii)), the grafting ratio (G) of CMC-g-PAM is also calculated to be around 212%, where G is defined as (Yuan et al., 2010) Gð%Þ ¼ ðWPAM =WCMC Þ  100%

flocculation performance of CMC-g-PAM is systematically studied at different dosages and various pH levels. First, the turbidity measurement, a commonly used method, is adopted to evaluate the optimal dosages of the flocculant (Fig. 2). From the turbidityedosage profiles, at pH 4, 7, and 11, the optimal dosages are 2, 5, and 16 mg/L, respectively. With increasing pH, the best residual turbidity percentage corresponding to the optimal dosage also increases, and the flocculation window becomes narrower.

(6)

where WPAM and WCMC are the weights of the PAM branches and the CMC backbone, respectively. As previously mentioned, CMC-based flocculants have enhanced solubility, which is beneficial for a wider application range. Therefore, the solubility of CMC-g-PAM in aqueous solution with various pH levels is tested according to the method of Chen and Park (2003) and shown in Table S1 of the supporting information. As expected, CMC-g-PAM reveals a remarkable improvement in solubility. It is soluble over the wide pH range from 1 to 13. The pH dependence of the ZP of kaolin and CMC-g-PAM were also measured (Fig. 1). The results demonstrate that kaolin particles are negatively charged over the whole pH range, whereas CMC-g-PAM has an isoelectric point at around pH 5 and shows a positive ZP below the isoelectric point. In addition, the ZP of CMC-g-PAM is close to zero at pH levels ranging from 5 to 7 because of nonionic PAM branch chains (Song et al., 2009).

3.2. Evaluation of the optimal dosages of the flocculant at various pH levels A variety of external parameters can affect the flocculation efficiency of flocculants, of which dosage and pH are the two most important ones (Rojas-Reyna et al., 2010). Hence, the

Zeta potential (mV)

0

-10

-20

Kaolin CMC-g-PAM

-30

2

4

6

8

10

12

pH Fig. 1 e ZPepH profiles of the kaolin suspension (---) and ). CMC-g-PAM solution (

Fig. 2 e Residual turbidity percentage (-) and ZP ( ) of the supernate as a function of CMC-g-PAM dosage at various pH: 4 (a), 7 (b), and 11 (c). Insert: The chain structures of CMC-g-PAM in solution at various pH.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 0 7 e1 1 4

Poorer flocculation efficiency with increasing pH has also been reported for a non-grafted chitosan-based amphoteric flocculant (Yang et al., 2011b). Since the conformation of the flocculant molecule in solution has great effects on its flocculation performance, the pH dependence of Rh is measured and shown in Fig. S4 of the supporting information. At lower pH values, the positive charges of the flocculant lead to high Rh because of the effects of intra-chain electrostatic repulsion (Radeva, 2001). Both the extensive conformation and positive charges of the flocculant are beneficial for flocculation efficiency and result in lower optimal dosage requirements. At pH 7, the decrease in Rh indicates that the polymer chains collapse owing to the loss of the net charge based on Fig. 1, which reduces the number of active flocculation sites and leads to a higher optimal dosage. At a higher pH level, the negative charges on the CMC-g-PAM increase, therefore, the polymer chain with a high Rh again becomes extensive. However, negative charges also result in greater electrostatic repulsion between the flocculants and the suspended particles, finally leading to poorer flocculation performance. Compared with non-grafted amphoteric flocculants (Yang et al., 2011b), CMC-g-PAM exhibits a more extended flocculation window and less sensitivity to dosage because of the better approachability of the grafted PAM chain to the suspended particles. Although CMC-g-PAM shows the poorest performance at pH 11, the best residual turbidity percentage can also reach a value of less than 5%. The flocculation performance of several conventional flocculants, such as alum, PAC, and PAM, are investigated at various pH levels and shown in Fig. S5 of the supporting information for comparison. Based on Fig. S5, the optimal dosages and the corresponding residual turbidity percentages of different flocculants are summarized in Table 1. Based on Table 1 and Fig. S5, the three conventional flocculants also show effective turbidity removal effects similar to CMC-gPAM. However, after further comparison, CMC-g-PAM consistently reveals the best flocculation performance amongst all four flocculants, and shows both the lowest optimal dosage and highest turbidity removal efficiency at pH 4 and 7. While PAM has the lowest optimal dosage at pH 11, its residual turbidity is much higher than that of the new flocculant. These flocculation behaviors demonstrate that CMC-gPAM is applicable for primary water purification, whether under acidic, neutral, or alkaline conditions.

Table 1 e The optimal dosages and corresponding residual turbidity percentages of various flocculants. Flocculants pH ¼ 4

CMC-g-PAM Alum PAC PAM a

Dosageoptimal (mg/L) RTlowest b (%) pH ¼ 7 Dosageoptimal (mg/L) RTlowest (%) pH ¼ 11 Dosageoptimal (mg/L) RTlowest (%)

2 0.86 5 2.74 16 4.65

8 4.70 12 4.56 25 4.63

4 3.19 8 3.73 20 4.71

2 2.28 6 5.08 12 9.54

a Dosageoptimal is the optimal dosage where the lowest residual turbidity percentage occurs. b RTlowest is the residual turbidity percentage at the optimal dosage.

111

The ZP of the supernates after flocculation is also provided along with the turbidity in Fig. 2. Under acidic conditions (Fig. 2(a)), ZP increases with increasing dosage owing to the opposite charge between the flocculant and the kaolin particles (Fig. 1). The net charge is nearly equal to zero at the optimal dosage. However, the flocculation mechanism cannot be explained simply by charge neutralization. This argument is based on the following: (i) No significant restabilization of suspended particles occurs at excess dosages (Yang et al., 2011b); (ii) ZP does not reach an extraordinarily high positive value, but shows a plateau of almost zero when excess flocculant is added; (iii) floc structures and flocculation kinetics are distinctively different from those that obey charge neutralization mechanisms (Rasteiro et al., 2008; Yu et al., 2006), which will be studied in the following section. Therefore, multimechanisms, including electrostatic patching and/or bridging, are probably involved in the flocculation process of CMC-g-PAM under acidic conditions. Under neutral or alkaline conditions (Fig. 2(b) and (c)), ZP also increases with increasing dosage. Although there is no opposite charge on the flocculant relative to kaolin based on Fig. 1, the increased ZP is due to the adsorption of a layer of polymer chains on the particle surface; these cause the shear plane to shift farther from the particle surface and the increase in the magnitude of ZP. Moreover, regardless of the pH condition, ZP cannot be used to evaluate the optimal dosage of CMC-g-PAM (Fig. 2). This result again demonstrates the limitation of the ZP method (Rasteiro et al., 2008; Yu et al., 2006). This method can only be used to predict the optimal dosage of flocculants when simple charge neutralization is the predominant flocculation mechanism. In addition to the traditional methods, as an alternative, Blanco et al. (1996) developed another method that optimizes the flocculation process based on monitoring of the mean floc size. Thus, in the present work, the dependence of floc size on the flocculant dosage is measured, and other crucial floc structure properties, such as fractal dimensions, are analyzed. The results are all illustrated in Fig. 3. In Fig. 3(a), larger flocs are obtained at the optimal dosage under each pH, which confirms the feasibility of this method for predicting the optimal dosage. The floc sizes grow until the dosages reach the optimal one, after which aggregation no longer takes place. The floc size even decreases at excess dosages because of steric stabilization and electrostatic repulsion (Rasteiro et al., 2008; Yu et al., 2006). Extraordinarily large floc sizes are obtained under acidic conditions. This result further denies the simple charge neutralization mechanism, given that the floc produced by charge neutralization is always smaller than that by bridging (Zhou and Franks, 2006). The microscopic photos shown in Fig. S6 of the supporting information also agree well with the results. In Fig. 3(b), DF values are calculated by linear fitting the logelog plot of I dependence on Q. Examples of the linear fitting are shown in Fig. S7 of the supporting information. For each pH investigated, the largest DF (i.e., the densest floc), as well as floc size, always occurs at the optimal dosages. However, when the flocculant is in excess, steric stabilization and electrostatic repulsion results in looser floc structures.

112

a

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 0 7 e1 1 4

300

pH=4 pH=7 pH=11

Floc size (µm)

250 200 150 100

particles. In Fig. 3(c), the D2 dependence on the flocculant dosage is in accordance with the results of DF, where D2 is calculated from Figs. S6 and S8 of the supporting information. However, considering that both DF and D2 are calculated through linear fitting, the data correlated for DF (correlation coefficient R2 > 0.99) are higher than those for D2 (R2 in the range of 0.80e0.97), which is attributed to the higher accuracy of the LS method than that of IA.

3.3. Monitoring of the flocculation process: flocculation kinetics

50 0

0

5

10

15

20

Dosage (mg/L)

b

2.4

pH4 pH7 pH11

2.2 2.0

DF

1.8 1.6 1.4 1.2 1.0 0

5

10

15

20

Dosage (mg/L)

c

2.0 1.8

D2

1.6

1.4

pH=4 pH=7 pH=11

1.2

1.0

0

5

10

15

20

Dosage (mg/L) Fig. 3 e The floc size (a), DF (b), and D2 (c) as functions of CMC-g-PAM dosage at various pH: 4 ( ), 7 (-), and 11 ( ).

Two points should be noted here: (i) DF at pH 4 is always larger than that in the two other pH levels, implying that patching mechanism is dominant under acidic conditions according to the literature (Rasteiro et al., 2008), and (ii) the extremely loose floc structure obtained under alkaline conditions is mainly due to the repulsive forces between negative charges of the adsorbed polymer and the kaolin

In the aforementioned discussion, results obtained from LS are consistent with the traditional turbidity measurement. From the variations in size, as well as D2 and DF as functions of flocculant dosage (Fig. 3), different flocculation mechanisms at various pH levels are primarily derived. Previous studies (Yu et al., 2006; Zhou and Franks, 2006) have reported that flocculation kinetics is highly related to the flocculation mechanism. In addition, flocculation kinetics is very complicated, and various stages, such as particlepolymer mixing, attachment of polymer onto the surface particles, rearrangement of adsorbed polymer chain, collisions of destabilized particles, breakage of floc due to shearing, and so on, occur continuously or simultaneously in the process. As mentioned in Section 1, LS can provide realtime floc structural information. Hence, from the aspects of size and fractal dimensions of flocs, flocculation processes at various pH levels are monitored in situ in the present work. The floc size and DF dependence on time of CMC-g-PAM at each optimal dosage under pH levels of 4, 7, and 11 are shown in Fig. 4, where DF is calculated based on Fig. S9 of the supporting information. In Fig. 4, the following results are obtained: (i) The floc in the acidic solution has the largest final size, whereas it has the slowest initial growth rate, and (ii) the size and DF of flocs become nearly constant after a brief period of time under both neutral and alkaline conditions, whereas those under acidic conditions continue growing through multiple steps over the measured time range. To explain these phenomena, feasible stages in the flocculation processes at pH levels of 4, 7, and 11 are proposed as follows. At pH 4, given that the amino groups on CMC backbone are protonated, the polymer chains are cationic and show extensive morphology with a large Rh of 171 nm in solution because of intra-chain electrostatic repulsion (Radeva, 2001) as shown in Fig. S4. There is a strong attraction between the flocculant and kaolin particles for the opposite charges. Therefore, at the initial stage of flocculation, a positively charged flocculant with large Rh tends to be adsorbed onto the particle in a flat configuration (Yu et al., 2006; Zhou and Franks, 2006). In this type of adsorption, the size of the kaolin particles coated by the polymer is almost unchanged (Rasteiro et al., 2008; Yu et al., 2006), leading to a slow floc growth rate. However, the coverage on the surface of the suspended particles by the polymeric flocculant is not full and uniform. Negative charges on bare parts of the surface of one particle easily attach to the excess net residual positive charges of the polymer coating of another particle (Mihai and Dragan, 2009), summarized as the patching mechanism. Through patching, different primary flocs aggregate to form

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 0 7 e1 1 4

a

300

Floc size (µm)

Under the alkaline condition, CMC-g-PAM again achieves a large Rh of 221 nm in solution. Owing to the better approachability of the PAM branches, bridging is the main flocculation mechanism. Therefore, the initial floc growth rate is also rapid. However, for aggregates, electrostatic repulsion between polymeric chains and suspended particles prevents particles from further entering into the interior of the floc. Thus, no rearrangement of floc occurs, causing the floc to reach the earliest equilibrium and achieve the most open structure (Fig. 4).

pH=4 dosage=2 mg/L pH=7 dosage=5 mg/L pH=11 dosage=16 mg/L

250

113

200 150 100 50 0

4. 0

5

10

15

20

Time (min)

b

2.4 2.2 2.0

DF

1.8 1.6 1.4 pH=4 dosage=2 mg/L pH=7 dosage=5 mg/L pH=11 dosage=16 mg/L

1.2 0

5

10

15

Conclusion

20

Time (min) Fig. 4 e Time dependence of floc size (a) and DF (b) at the optimal dosage of CMC-g-PAM at various pH: 4 ( ), 7 (-) and 11 ( ).

larger flocs, resulting in a continuously increased floc size. Rearrangements of the polymeric flocculant and original particles in the floc (Rasteiro et al., 2008; Yu et al., 2006; Zhou and Franks, 2006) always occur, accompanied by newly formed larger flocs, thereby inducing increases in DF and size. Despite the dominant mechanism of patching, bridging due to the better approachability of PAM chains to suspended particles (Sen et al., 2009) may also have valid effects and contribute to the formation of large flocs. At pH 7, when the ZP of CMC-g-PAM is close to zero, the bridging flocculation mechanism plays the predominant role. At the beginning of the flocculation process, different segments (PAM branches or the main backbone) of the same polymer chain can bind more than one particle very rapidly (Rojas-Reyna et al., 2010), that is, link multiple particles together. Thus, the initial floc growth rate is high. However, the primary aggregates are unstable and can further rearrange. The flocculant polymer chains with an originally small Rh of 125 nm transform into more extended and flatter configurations that stick onto the surface of suspended particles (Rasteiro et al., 2008; Yu et al., 2006), resulting in more compact but slightly smaller-sized flocs.

A novel amphoteric chemically bonded composite flocculant, CMC-g-PAM, is designed and successfully prepared. The new flocculant demonstrates improved solubility. Flocculation tests at various pH levels prove the applicability of the designed flocculant over a wide range of pH for primary water treatment. The optimal dosages of the flocculant obtained from turbidity, size, and fractal dimension test are consistent; ZP is not applicable under a more complicated flocculation mechanism. By measuring the floc size and fractal dimension, different flocculation mechanisms are also found at various pH levels. Patching is predominant under acidic conditions and causes larger and denser flocs, whereas bridging plays a key role under neutral and alkaline conditions, producing smaller and looser flocs. Moreover, three diverse kinetic processes of flocculation at various pH levels are proposed, according to the online monitoring results of LS. CMC-g-PAM favors adsorption onto the surface of suspended particles under a platform at acidic conditions, resulting in a slow initial growth rate but continuous floc growth. Under neutral or alkaline conditions, the flocculant prefers bridging particles together, causing a rapid initial rate and requiring a shorter period of time to reach equilibrium. In addition, under alkaline conditions, no further rearrangement of polymeric flocculant and particles in flocs occurs because of electrostatic repulsion.

Funding Supported by the Natural Science Foundation of China (Grant No. 51073077, 50633030, 50938004 and 50825802) and the Fundamental Research Funds for the Central Universities (Grant No. 1105020504 and 1116020510).

Appendix. Supplementary material Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.10.024.

references

Ali, S.K.A., Singh, R.P., 2009. An investigation of the flocculation characteristics of polyacrylamide-grafted chitosan. Journal of Applied Polymer Science 114 (4), 2410e2414.

114

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 0 7 e1 1 4

Biswal, D.R., Singh, R.P., 2004. Characterisation of carboxymethyl cellulose and polyacrylamide graft copolymer. Carbohydrate Polymers 57 (4), 379e387. Blanco, A., Negro, C., Hooimeijer, A., Tijero, J., 1996. Polymer optimization in paper mills by means of a particle size analyser: an alternative to zeta potential measurements. Appita Journal 49 (2), 113e116. Bushell, G., Amal, R., 2000. Measurement of fractal aggregates of polydisperse particles using small-angle light scattering. Journal of Colloid and Interface Science 221 (2), 186e194. Ceresa, R.J., 1973. Block and Graft Copolymerization. Wiley, New York. Chakraborti, R.K., Atkinson, J.F., Van Benschoten, J.E., 2000. Characterization of alum floc by image analysis. Environmental Science & Technology 34 (18), 3969e3976. Chakraborti, R.K., Gardner, K.H., Atkinson, J.F., Van Benschoten, J. E., 2003. Changes in fractal dimension during aggregation. Water Research 37 (4), 873e883. Chen, X.G., Park, H.J., 2003. Chemical characteristics of Ocarboxymethyl chitosans related to the preparation conditions. Carbohydrate Polymers 53 (4), 355e359. da Silva, D.A., de Paula, R.C.M., Feitosa, J.P.A., 2007. Graft copolymerisation of acrylamide onto cashew gum. European Polymer Journal 43 (6), 2620e2629. Fabrizi, L., Jefferson, B., Parsons, S.A., Wetherill, A., Jarvis, P., 2010. The role of polymer in improving floc strength for filtration. Environmental Science & Technology 44 (16), 6443e6449. Guibal, E., Van Vooren, M., Dempsey, B.A., Roussy, J., 2006. A review of the use of chitosan for the removal of particulate and dissolved contaminants. Separation Science and Technology 41 (11), 2487e2514. Jarvis, P., Jefferson, B., Parsons, S.A., 2005. Breakage, regrowth, and fractal mature of natural organic matter flocs. Environmental Science & Technology 39 (7), 2307e2314. Liao, J.Y.H., Selomulya, C., Bushell, G., Bickert, G., Amal, R., 2006. On different approaches to estimate the mass fractal dimension of coal aggregates. Particle & Particle Systems Characterization 22 (5), 299e309. Lin, J.L., Huang, C.P., Chin, C.J.M., Pan, J.R., 2008. Coagulation dynamics of fractal flocs induced by enmeshment and electrostatic patch mechanisms. Water Research 42 (17), 4457e4466. Mandelbrot, B.B., 1983. The Fractal Geometry of Nature. Freeman, New York. Mihai, M., Dragan, E.S., 2009. Chitosan based nonstoichiometric polyelectrolyte complexes as specialized flocculants. Colloids and Surfaces A e Physicochemical and Engineering Aspects 346 (1e3), 39e46. Radeva, T., 2001. Physical Chemistry of Polyelectrolytes. Marcel Dekker Inc., New York. Rasteiro, M.G., Garcia, F.A.P., Ferreira, P., Blanco, A., Negro, C., Antunes, E., 2008. The use of LDS as a tool to evaluate flocculation mechanisms. Chemical Engineering and Processing 47 (8), 1329e1338. Renault, F., Sancey, B., Badot, P.M., Crini, G., 2009. Chitosan for coagulation/flocculation processes e an eco-friendly approach. European Polymer Journal 45 (5), 1337e1348. Rojas-Reyna, R., Schwarz, S., Heinrich, G., Petzold, G., Schutze, S., Bohrisch, J., 2010. Flocculation efficiency of modified water soluble chitosan versus commonly used commercial polyelectrolytes. Carbohydrate Polymers 81 (2), 317e322. Sand, A., Yadav, M., Behari, K., 2010. Preparation and characterization of modified sodium carboxymethyl cellulose via free radical graft copolymerization of vinyl sulfonic acid in aqueous media. Carbohydrate Polymers 81 (1), 97e103.

Schaefer, D.W., Martin, J.E., Wiltzius, P., Cannell, D.S., 1984. Fractal geometry of colloidal aggregates. Physical Review Letters 52 (26), 2371e2374. Sen, G., Pal, S., 2009. Polyacrylamide grafted carboxymethyl tamarind (CMT-g-PAM): development and application of a novel polymeric flocculant. Macromolecular Symposia 277, 100e111. Sen, G., Kumar, R., Ghosh, S., Pal, S., 2009. A novel polymeric flocculant based on polyacrylamide grafted carboxymethylstarch. Carbohydrate Polymers 77 (4), 822e831. Song, H., Wu, D., Zhang, R.Q., Qiao, L.Y., Zhang, S.H., Lin, S., Ye, J., 2009. Synthesis and application of amphoteric starch graft polymer. Carbohydrate Polymers 78 (2), 253e257. Sonmez, H.B., Senkal, B.F., Bicak, N., 2002. Poly(acrylamide) grafts on spherical bead polymers for extremely selective removal of mercuric ions from aqueous solutions. Journal of Polymer Science Part A e Polymer Chemistry 40 (17), 3068e3078. Thomas, D.N., Judd, S.J., Fawcett, N., 1999. Flocculation modelling: a review. Water Research 33 (7), 1579e1592. Tripathy, J., Mishra, D.K., Yadav, M., Sand, A., Behari, K., 2009. Modification of kappa-Carrageenan by graft copolymerization of methacrylic acid: synthesis and applications. Journal of Applied Polymer Science 114 (6), 3896e3905. von Smoluchowski, M., 1917. Experiments on a mathematical theory of kinetic coagulation of coloid solutions. Zeitschrift fu¨r Physikalische Chemie e Sto¨chiometrie und Verwandtschaftslehre 92 (2), 129e168. Wang, J.P., Chen, Y.Z., Zhang, S.J., Yu, H.Q., 2008. A chitosanbased flocculant prepared with gamma-irradiation-induced grafting. Bioresource Technology 99 (9), 3397e3402. Wang, J.P., Chen, Y.Z., Yuan, S.J., Sheng, G.P., Yu, H.Q., 2009. Synthesis and characterization of a novel cationic chitosanbased flocculant with a high water-solubility for pulp mill wastewater treatment. Water Research 43 (20), 5267e5275. Wang, W., Qin, W., Bo, S.Q., 1991. Influence of the degree of deacetylation of chitosan on its mark-houwink equation parameters. Makromolekulare Chemie-Rapid Communications 12 (9), 559e561. Wei, J.C., Gao, B.Y., Yue, Q.Y., Wang, Y., Li, W.W., Zhu, X.B., 2009. Comparison of coagulation behavior and floc structure characteristic of different polyferric-cationic polymer dualcoagulants in humic acid solution. Water Research 43 (3), 724e732. Yang, H., Lu, Y.B., Cheng, R.S., 2011a. In: Davis, S.P. (Ed.), Chitosan: Manufacture, Properties, and Usage. Nova Science Publishers, New York. Yang, Z., Shang, Y.B., Lu, Y.B., Chen, Y.C., Huang, X., Chen, A.M., Jiang, Y.X., Gu, W., Qian, X.Z., Yang, H., Cheng, R.S., 2011b. The flocculation properties of biodegradable amphoteric chitosanbased flocculants. Chemical Engineering Journal 172 (1), 287e295. Yu, J.F., Wang, D.S., Ge, X.P., Yan, M.Q., Yang, M., 2006. Flocculation of kaolin particles by two typical polyelectrolytes: a comparative study on the kinetics and floc structures. Colloids and Surfaces A e Physicochemical and Engineering Aspects 290 (1e3), 288e294. Yu, W.Z., Gregory, J., Campos, L., 2010. Breakage and regrowth of Al-humic flocs e effect of additional coagulant dosage. Environmental Science & Technology 44 (16), 6371e6376. Yuan, B., Shang, Y.B., Lu, Y.B., Qin, Z.Q., Jiang, Y.X., Chen, A.M., Qian, X.Z., Wang, G.Z., Yang, H., Cheng, R.S., 2010. The flocculating properties of chitosan-graft-polyacrylamide flocculants (I) e effect of the grafting ratio. Journal of Applied Polymer Science 117 (4), 1876e1882. Zhou, Y., Franks, G.V., 2006. Flocculation mechanism induced by cationic polymers investigated by light scattering. Langmuir 22 (16), 6775e6786.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 5 e1 2 6

Available online at www.sciencedirect.com

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

A model for predicting resuspension of Escherichia coli from streambed sediments Pramod K. Pandey a, Michelle L. Soupir a, Chris R. Rehmann b,* a b

Dept. of Agricultural and Biosystems Engineering, Iowa State University, Ames, IA 50011, USA Dept. of Civil, Construction, and Environmental Engineering, Iowa State University, Ames, IA 50011, USA

article info

abstract

Article history:

To improve the modeling of water quality in watersheds, a model is developed to predict

Received 8 February 2011

resuspension of Escherichia coli from sediment beds in streams. The resuspension rate is

Received in revised form

expressed as the product of the concentration of E. coli attached to sediment particles and

14 October 2011

an erosion rate adapted from work on sediment transport. The model uses parameter

Accepted 16 October 2011

values mostly taken from previous work, and it accounts for properties of the flow through

Available online 23 October 2011

the bottom shear stress and properties of the sediment through the critical shear stresses for cohesive and non-cohesive sediment. Predictions were compared to resuspension rates

Keywords:

inferred from a steady mass balance applied to measurements at sixteen locations in

Resuspension

a watershed. The model’s predictions matched the inferred rates well, especially when the

E. coli

diameter of particles to which E. coli attach was allowed to depend on the bottom shear

Sediment

stress. The model’s sensitivity to the parameters depends on the contributions of particle

Microbial transport

packing and binding effects of clay to the critical shear stress. For the current data set, the

Watershed modeling

uncertainty in the predictions is controlled by the concentration of E. coli attached to sediment particles and the slope used to estimate the bottom shear stress. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Pathogens impair 480,000 km of rivers and shorelines and 2 million ha of lakes in the U.S., and the cost to implement total maximum daily load (TMDL) plans is estimated as $0.9 to $4.3 billion per year (USEPA, 2010). To predict the risk of bacteria to public health and allocate load reductions fairly, models that include accurate representations of the key processes of fate and transport are required (Fries et al., 2008). For example, the high concentrations of bacteria in suspended sediment and bed sediment suggest that the understanding of interactions between pathogens and sediment must be improved (Droppo et al., 2009). Sediment disturbance can account for the majority of total bacterial contamination (Nagels et al., 2002), and a one-dimensional

model applied to the field measurements of Jamieson et al. (2005) showed that including interactions with the sediment improved the predictions of Escherichia coli concentrations in the stream (Rehmann and Soupir, 2009). However, models that U.S. regulatory agencies use to determine pollutant load reductions usually do not include resuspension of bacteria as a source. Even when resuspension is included in models, how to predict it is uncertain. Many researchers have either specified the resuspension rate (e.g., Petersen et al., 2009) or expressed it mainly as a function of flow (Wilkinson et al., 1995; Tian et al., 2002; Collins and Rutherford, 2004). Kim et al. (2010) added a model of resuspension of E. coli to the Soil and Water Assessment Tool (SWAT); resuspension was estimated using a simplified version of Bagnold’s stream power

* Corresponding author. Tel.: þ1 515 294 1203; fax: þ1 515 294 8216. E-mail address: [email protected] (C.R. Rehmann). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.019

116

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 5 e1 2 6

Nomenclature a a1 b b1 C1 C2 c3 c5 Ca d E E0 E0a fa g H2 kn2 N n na ns

coefficient for the effects of particle packing on the critical shear stress sc [L2] coefficient in the alternative model of the resuspension rate (Eq. (14)) [L13b1 Tb1 1 ] coefficient for the effects of particle packing on the critical shear stress sc [M1 L3] exponent in the alternative model of the resuspension rate (Eq. 14) [-] concentration of E. coli in the water column [#L3] concentration of E. coli in the sediment [#L3] prg(s  1)/6, coefficient for the effect of clay on the critical stress sc [M L2 T2] coefficient for the effect of clay on the critical stress sc [M L1 T2] concentration of E. coli attached to sediment in the bed [#L3] diameter of sediment particles to which E. coli attach [L] erosion rate for sediment [L T1] erosion rate at the threshold of erosion [L T1] coefficient in the predicted resuspension rate [L T1] fraction of E. coli in the water column that are attached to sediment [-] acceleration of gravity [L T2] depth of the sediment containing E. coli [L] net growth rate in the sediment [T1] number of parameters [-] Manning roughness coefficient [-] exponent in the predicted resuspension rate [-] exponent in the erosion rate for sediment [-]

equation, which has been criticized for not including the effect of grain size on sediment transport (Ferguson, 2005). Hipsey et al. (2008) accounted for properties of the sediment by including a critical shear stress computed from the Shields criterion, but although the Shields criterion holds for noncohesive sediment, its validity for cohesive sediment is questionable (Mehta and Lee, 1993). Sanders et al. (2005) assumed resuspension to be proportional to the shear stress, while Bai and Lung (2005) expressed resuspension as a nonlinear function of the difference between the shear stress and a critical shear stress. As Rehmann and Soupir (2009) noted, resuspension of microorganisms from a sediment bed depends in general on properties of the flow and sediment (e.g., Lick, 2009, ch. 3), the type of microorganism (Hipsey et al., 2008), and the presence of biofilms (e.g., Droppo et al., 2001). To predict E. coli resuspension reliably, theory for transport of cohesive sediment must be considered because most bacteria attach to cohesive particles (Black et al., 2002). For non-cohesive sediments such as sands, which typically have particle diameters greater than 62 mm, the main forces to consider are the dislodging tendency of the fluid shear stress and the submerged weight of a particle. For cohesive sediment, however, inter-grain forces complicate the predictions. Clay and very fine silt (<8 mm) exhibit strong cohesion, while

Q R Ra Rai Rai S Syi s t T vr ws yi a Dyi r rb s sb sc scn F fb fc j

discharge [L3 T1] hydraulic radius [L] predicted resuspension rate [#L2 T1] inferred resuspension rate [#L2 T1] average inferred resuspension rate [#L2 T1] slope [-] relative sensitivity to yi [-] specific gravity of sediment particles [-] time [T] temperature [Q] resuspension velocity [L T1] settling velocity [L T1] generic parameter coefficient relating particle diameter to shear stress [M1 L2 T2] uncertainty in yi water density [M L3] bulk density of the sediment [M L3] sum of the squares of the differences between the logarithms of resuspension rates bottom shear stress [M L1 T2] critical shear stress for cohesive sediment [M L1 T2] critical shear stress for non-cohesive sediment [M L1 T2] kn2H2C2/fawsC1, parameter measuring the importance of settling and net growth [-] aexp(brb)/d2, contribution of bulk density to the critical shear stress sc [-] c5/c3d, contribution of binding effects of clay to the critical shear stress sc [-] Shields parameter [-]

larger silt particles (8e62 mm) are more weakly cohesive (van Rijn, 2007). Furthermore, because cohesion of deposited flocs renders the critical shear stress for erosion higher than that for deposition, the conditions under which the bed is deposited and the time for consolidation can be expected to affect erosion (Krishnappan, 2007), as well as the properties of the eroded flocs (Stone et al., 2008). For example, sediment beds formed in a sheared flow eroded at higher shear stresses than those formed in quiescent conditions (Droppo et al., 2001). Also, biofilms increase the critical shear stress by forming a coating that protects the sediment against erosion (Droppo et al., 2001). The coating contains extracellular polymeric substances that increase cohesiveness between particles (Paterson, 1989) and strengthens the surficial sediment (e.g., Sutherland et al., 1998). To improve predictions of in-stream transport of E. coli, we develop a formulation for E. coli resuspension that accounts for properties of the flow and properties of both cohesive and non-cohesive sediment. The objectives of this study are 1) to develop a model by assuming that the E. coli resuspension rate is proportional to the erosion rate of sediment, 2) to compare the predictions from the model to resuspension rates inferred from mass balances at several locations in a watershed, 3) to evaluate the model’s predictive skill, and 4) to assess the sensitivity and uncertainty of the resuspension rate to the

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 5 e1 2 6

117

input parameters so that measurements and modeling can be improved. The model is developed in Section 2, and the measurements and calculations used in our application of the model to predict resuspension in a creek are described in Section 3. Objectives 2, 3, and 4 are addressed in Section 4, and the main conclusions are listed in Section 5.

2.

Model

We hypothesize that the rate of resuspension of attached E. coli can be estimated as the product of the concentration of attached E. coli in the sediment bed and an erosion rate similar to that for sediment. Several researchers have proposed formulas to predict erosion (Partheniades, 1965; Mehta, 1989), but we use the formulation of Lick (2009), in which the erosion rate E depends on the bottom shear stress sb, the critical shear stress sc for cohesive sediment, and the critical shear stress scn for non-cohesive sediment. For sb > scn, E ¼ E0


n sb  scn s ; sc  scn

(1)

where E0 ¼ 106 m/s is the erosion rate at the threshold of erosion (Lick, 2009). Lick (2009) found the exponent ns to be approximately equal to 2 for small and intermediate particles, while others expressed the erosion rate as linearly proportional (i.e., ns ¼ 1) to the difference between the bottom stress and a critical stress (Amos et al., 1996). A compilation of data shows that the critical shear stress for non-cohesive sediment depends on the particle diameter d: scn ¼ 414d;

(2) 2

where scn is in N/m and d is in m (Lick, 2009). For cohesive sediment, the packing of the particles, which is quantified by the bulk density rb, and extra binding forces caused by clay must be considered. Combining the work of Roberts et al. (1998) and Lick et al. (2004), Lick (2009) proposed
 aebrb c5 ; sc ¼ scn 1 þ 2 þ d c3 d

(3)

where a and b are coefficients that Lick (2009) specified as 8.5  1016 m2 and 9.07 cm3/g, respectively. The coefficient c3 is given by p c3 ¼ rgðs  1Þ; 6

(4)

where r is the density of water, s is the specific gravity of the sediment particle, and g is the acceleration of gravity. The coefficient c5 depends on the clay fraction; for 2% bentonite added to quartz particles, c5 ¼ 7 N/m2 (Lick et al., 2004). Lick (2009) proposed Eq. (1) as a uniformly valid formulation for erosion. For large particles and no clay fraction, the bulk density does not affect the critical shear stress (Fig. 1), and Eq. (1) follows a form that applies to fine-grained, coarsegrained, cohesive sediments, and non-cohesive sediment (Lick, 2009). As the particle size decreases, effects of cohesion dominate (i.e., sc [ scn), and Eq. (1) reduces to a form similar to that of Roberts et al. (1998) for cohesive sediment. When the binding effects of clay provide the main resistance to particle motion, the critical shear stress depends only weakly on the

Fig. 1 e Dependence of critical shear stress on particle diameter. The dotted line is the critical stress for noncohesive sediment. The dashed line is the critical stress for cohesive sediment with a bulk density rb of 1.26 g/cm3 and no effects of clay (c5 [ 0 N/m2), and the solid line is the critical stress for cohesive sediment with rb [ 1.26 g/cm3 and c5 [ 21 N/m2.

particle diameter (Fig. 1) because both the mobilizing force and the resisting force depend on the surface area of the particle. The erosion rate in Eq. (1) can be adapted to predict the rate of resuspension of E. coli. Bacteria attach to and bioflocculate around solid particles (Black et al., 2002) and deposit to the bottom sediments; attached fractions for streams ranges from 55% during storms (Characklis et al., 2005; Krometis et al., 2007) to between 80 and 100% (Auer and Niehaus, 1993; Hipsey et al., 2008). When sediment is resuspended, an influx of E. coli from the streambed results (Whitman et al., 2006). Therefore, we predict the E. coli resuspension rate Ra by multiplying the erosion rate by the concentration Ca of E. coli attached to sediment in the bed: Ra ¼ Ca E0a


n sb  scn a : sc  scn

(5)

The coefficient and exponent are changed to E0a and na to allow for possible differences from E0 and ns in Eq. (1). The resuspension rate is nonzero only when the bottom shear stress exceeds the critical shear stress for non-cohesive sediment. We expect Eq. (5) to be useful in predicting resuspension rates because it accounts for effects of the flow and sediment, as well as the concentration of E. coli in the streambed.

3.

Methods

We applied the model to the Squaw Creek watershed to predict the resuspension of E. coli attached to stream bottom sediments into the water column. The model was evaluated using data collected from sixteen sites in the watershed. At

118

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 5 e1 2 6

each site, flow geometry was measured, and the concentrations of E. coli in streambed sediment and the overlying water column were determined. The resuspension rates predicted with Eq. (5) were compared to values inferred from the onedimensional model of Rehmann and Soupir (2009). A sensitivity analysis was conducted to assess the influence of certain parameters on model output, and the parameters controlling the uncertainty in the resuspension rate were identified. Table 1 lists the parameters used in computing the predicted and inferred resuspension rates, and it indicates whether the parameters were measured, estimated, taken from previous work, or calibrated.

3.1.

Study area

Squaw Creek passes through four counties of central Iowa, U.S.A. before discharging into the South Skunk River near Ames (Fig. 2). The total area of the Squaw Creek watershed, as defined by the 10-digit hydrologic unit code, is 59,327 ha, and the average basin slope is 2%. Soils consist of loamy Wisconsin glacial till and clayey lacustrine deposits, including loam, silty clay, clay loam, and silty clay loam (Iowa NRCS, 2011); about 87% of the soil is fine, and another 8% is sandy. The study area has a humid climate with an average yearly rainfall of 865.4 mm and average annual high and low temperature of 15.6 and 3.3  C, respectively. The stream network of Squaw Creek watershed was generated using 30 m digital elevation maps from the U.S. Geological Survey’s Earth Resources Observation and Science Center and the geographic information systems software ArcGIS 9 (ArcMap version 9.3.1) to identify the tributaries and main stream. Land cover was determined with a 2002 map for Iowa obtained from the Natural Resources Geographic Information System library, a repository developed by the Iowa Department of Natural Resources. About 74% of the watershed was under

agricultural management (corn 41%, soybean 33%, and row crops 0.4%), 17% of the watershed was under grassland (ungrazed grass 11%, grazed grass 2.5%, CRP grass 1.7%, and alfalfa 1.8%), and 2.7% was deciduous forest. Additionally, 5.4% of the watershed land cover was road, residential, and commercial and 0.3% was water and wetlands. The watershed has 20 listed confined feeding operation units, and hogs are the major livestock.

3.2.

Measurements

Data were collected from the sampling locations to predict E. coli resuspension rates. Water temperature and cross section geometry were measured at sixteen locations on 17 July 2009. The mean air temperature during the sampling was 18.4  C, and although the sky was mostly overcast, precipitation was zero. The mean discharge for the day was reported to be 3.6 m3/ s at the U.S. Geological Survey gaging station 05470500 on Squaw Creek in Ames, which is at the same cross section as our sampling location 16 (Fig. 2); the discharge varied by less than 2% during the sampling. The bulk density of the streambed sediments, expressed as weight per unit volume, was determined from wet and dry weight (dried in the oven at approximately 75  C for 2 days) of sediments (Roberts et al., 1998). The Manning roughness coefficient n was taken to be 0.036 using information for natural streams in Chow (1959, pp. 112e123). Water samples were collected from the center of the stream by lowering a Horizontal Polycarbonate Water Bottle Sampler (2.2 L, Forestry Suppliers Inc., Mississippi, U.S) from a bridge into the center of the stream at all the locations. Sediment samples were collected from the top 2e3 cm of the streambed using a Shallow Water Bottom Dredge Sampler (15 cm  15 cm opening, Forestry Suppliers Inc., Mississippi, U.S) at the same location as water samples. Three replicates of water and sediment samples were used in microbial analyses. Immediately after collection, samples were stored at 4  C and

Table 1 e Parameters used to compute the predicted and inferred resuspension rates. The second column indicates whether the parameter was used in the predicted rate (P), inferred rate (I), or both (B). The fourth column lists the uncertainty, expressed as a percentage of the parameter’s value, assumed in the analysis in Section 4. Parameter C1 ¼ conc. of E. coli in water (CFU/100 ml) C2 ¼ conc. of E. coli in sed. (CFU/100 ml) R ¼ hydraulic radius (m) A ¼ cross sectional area (m2) rb ¼ bulk density of the sediment (g/cm3) T ¼ temperature ( C) Q ¼ discharge (m3/s) E0a ¼ coefficient (m/s) a ¼ coefficient for bulk density effect (m2) b ¼ coefficient for bulk density effect (cm3/g) c3 ¼ coefficient for clay effect (N/m3) H2 ¼ depth of sediment containing E. coli (m) n ¼ Manning roughness coefficient S ¼ slope (m/m) fa ¼ attached fraction na ¼ exponent c5 ¼ coefficient for clay effect (N/m2) d ¼ particle diameter (mm)

Rates

Value

Uncty (%)

Source

I B P P P B P P P P P I P P B P P B

2.25  102  5.47  103 1.63  103  3.43  104 0.10e0.76 0.5e3.5 1.26 17.0e24.6 3.6 1  106 8.5  1016 9.07 8.46  103 0.02 0.036 2.5  104 1.0 1.0 21 1.0; 0.5e3.5

15 15 10 10 5 5 5 0 0 0 0 50 15 20 15 10 10 50

Measured Measured Measured Measured Measured Measured Measured at station 16 Lick (2009) Lick (2009) Lick (2009) Lick (2009), computed with s ¼ 2.65 Estimated from sediment sampler Estimated from Chow (1959) Estimated with Manning’s eq. at station 16 Estimated using range in Hipsey et al. (2008) Estimated from Amos et al. (1996) Calibrated/estimated from Lick (2009) Calibrated using ranges in Oliver et al. (2007) by fitting d ¼ asb with a ¼ 1.9 mm/Pa

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 5 e1 2 6

119

Fig. 2 e Squaw Creek watershed and sampling locations (1e16). Discharge was measured at the U.S. Geological Survey gaging station near station 16. The land cover is also shown.

analyzed within 24 h. The E. coli attached to particles were detached by stirring the mixture of sediment and purified water (ratio 1:1) for 15 min at approximately 200 rpm using a magnetic stir bar. The resulting solution was used to enumerate E. coli in the sediment. The values of C1 and C2, the E. coli concentrations in the water and the sediment, respectively, were determined by membrane filtration techniques (APHA, 1999) on modified mTEC agar (EPA, method 1603).

3.3. rates

Calculation of predicted and inferred resuspension

Resuspension rates were predicted with Eq. (5). All E. coli were assumed to be attached to sediment grains; that is, the attached fraction fa ¼ 1 and Ca ¼ C2. This choice is consistent with the assumptions and work reviewed in Hipsey et al. (2008), which showed attachment between 80 and 100%. The

bottom shear stress was computed from a force balance for steady, uniform flow: sb ¼ rgRS;

(6)

where r is the water density, g is the acceleration of gravity, and R is the hydraulic radius. The slope S was estimated from Manning’s equation to be 2.5  104. Values of the coefficients a and b from Lick (2009) were used, and the coefficient E0a was assumed to be equal to E0 given by Lick (2009). The coefficient c5 was calibrated, and the exponent na was taken to be 1, as suggested by Amos et al. (1996), who found the erosion rate to be linearly proportional to the difference between the shear stress and a critical shear stress. The critical shear stresses scn and sc require an estimate of the diameter d of the particles to which the E. coli attach. A constant value of the diameter and a diameter that is linearly proportional to the bottom shear stress were

120

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 5 e1 2 6

used. The merits of these approaches are discussed in Section 4.1. To evaluate the predictions, resuspension rates were inferred from the mass balance model of Rehmann and Soupir (2009). Considering settling, resuspension, and net growth, they determined that a steady-state mass balance for the sediment yields C2 fa ws ¼ ; C1 vr  kn2 H2

Rai ¼ vr C2 ¼ fa ws C1 þ kn2 H2 C2 ¼ fa ws C1 ð1 þ FÞ;

(8)

where F ¼ kn2 H2 C2 =fa ws C1 indicates the relative importance of settling and net growth in the mass balance for E. coli in the sediment. For example, when F [ 1, settling is unimportant, and net growth balances resuspension. Once the exponent na was chosen, the parameters to determine or calibrate were the coefficient c5 and the diameter of the particles to which E. coli attach. The optimal values of the parameters for the predictions using Eq. (5) were chosen by minimizing s, the sum of the squares of the differences between the logarithms of the inferred and predicted resuspension rates: X

log10 Ra  log10 Rai




DRa Ra

2 ¼

2 X 2 N
N
X 1 vRa Dyi Syi Dyi ¼ ; yi Ra vyi i¼1 i¼1

(12)

where N is the number of parameters and Dyi is the uncertainty in yi.

(7)

where ws is the settling velocity, vr is the resuspension velocity, kn2 is the net growth rate in the sediment, and H2 is the depth of sediment containing E. coli, which is estimated to be about 2 cm for our experiments. The settling velocity ws was estimated with Stokes’s law. The net growth rate is the difference between the growth rate and the natural mortality rate, which were computed as functions of water temperature using the relations in Hipsey et al. (2008). The resuspension velocity at each sampling location was computed from Eq. (7), and the inferred resuspension rate was computed as

s¼

individual parameters, which were assumed to be independent, with the formula of Taylor (1997, p. 75):

2

4.

Results and discussion

4.1. rates

Concentrations, critical stresses, and resuspension

E. coli concentrations were large. The concentration C1 of E. coli in the water column ranged from 225 to 5467 CFU/100 ml with a mean of 789 CFU/100 ml and standard deviation of 1255 CFU/ 100 ml (Fig. 3). All but one of the concentrations exceeded the U.S. water quality standards (USEPA, 2001), which state that the geometric mean of at least five samples during a 30-day period must not exceed 126 CFU/100 ml and that a singlesample must not exceed 235 CFU/100 ml. The concentration C2 of E.coli in the sediment ranged from 1633 to 34,257 CFU/100 ml with a mean of 13,597 CFU/100 ml and standard deviation of 10,463 CFU/100 ml. Concentrations in the sediment were 2e90 times (mean ¼ 30, s.d. ¼ 29) higher than concentrations in the water column. Previous studies reported the ratio C2/C1 to be 10e10,000 (Buckley et al., 1998; Davies and Bavor, 2000; Bai and Lung, 2005). Resuspension rates inferred using Eq. (8) ranged from 11 to 167 CFU/m2s, depending on whether the diameter d of the particles to which E. coli attach was set to a constant value (Fig. 4a) or allowed to vary with hydraulic conditions (Fig. 4b). The inferred rates are smaller than those of

(9)

Values of s were computed with the resuspension rates expressed in CFU/m2s. A quantitative measure of predictive skill (Willmott, 1981) was computed to assess the agreement between predicted and inferred E.coli resuspension rates: P jRa  Rai j2 ; Skill ¼ 1  P ðjRa  Rai j þ jRa  Rai jÞ2

(10)

where the overbar denotes an average over all sampling locations. A skill of 1 indicates perfect agreement between predicted and inferred resuspension rates, while a skill of zero indicates poor performance. To understand the dependence of the resuspension rate on the parameters and help in applying the model, the sensitivity and uncertainty were computed. The relative sensitivity of the resuspension rate to each parameter yi was computed with Syi ¼

yi vRa Ra vyi

(11)

(Haan, 2002). The relative uncertainty in the resuspension rate was computed by propagating the uncertainties of the

Fig. 3 e Concentrations of E. coli in the water column and sediment. Squares denote measurements from the main channel, and circles denote measurements from the tributaries. The solid vertical line is set at the USEPA’s single-sample standard for E. coli, and the dotted lines are contours of C2/C1.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 5 e1 2 6

a

b

clay, the critical stress used for the predictions in Fig. 4 did not depend strongly on the particle diameter (Fig. 1). Estimates of critical stress in other cases vary widely because of the characteristics of the sediment, the presence of biofilms, the depositional history of the bed, and the approach used to define the critical stress. For field measurements in a stream in which 32% of the sediment was finer than 75 mm, Jamieson et al. (2005) computed critical shear stresses of 1.5e1.7 N/m2 using the Manning’s roughness coefficient and the discharge at which E. coli NAR first appeared in discrete samples during a storm. El Ganaoui et al. (2004) analyzed sediment samples from a field site and differentiated between the fluff layer, a surface layer of fine sized particles and organics with a mean particle diameter of 10e20 mm and critical shear stresses of 0.025e0.05 N/m2, and a layer with coarser particles, which had critical shear stresses that were 10e20 times larger. Droppo et al. (2001) conducted laboratory experiments on kaolinite clay with a mean diameter of 5 mm and contaminated sediment from a field site that had particle sizes less than 63 mm. The critical stress increased from 0.024 N/m2 to 0.325 N/m2 when a biofilm was allowed to grow, and critical stresses of 0.100e0.135 N/m2 for beds deposited under shear exceeded the stresses of 0.047e0.054 N/m2 for beds deposited under quiescent conditions. The critical stress estimated for Squaw Creek had a magnitude representative of natural sediment beds with biofilms and a realistic deposition history.

4.2.

Fig. 4 e Comparison between predicted and inferred resuspension rates. The solid line indicates perfect agreement, and the dashed lines indicate difference by a factor of 2. Squares denote measurements from the main channel, and circles denote measurements from the tributaries: (a) Constant value of the particle diameter: d [ 1.0 mm, s [ 1.04, and skill [ 0.82. (b) Particle diameter linearly related to bottom shear stress: d [ asb with a [ 1.9 mm/Pa, s [ 0.40, and skill [ 0.85.

Jamieson et al. (2005), who measured resuspension rates of 8200e15,000 CFU/m2s in a stream during storms. One cause of the discrepancy is that Jamieson et al. (2005) artificially seeded the bed with E. coli NAR; concentrations of E. coli in the sediment corresponding to the three storms highlighted by Jamieson et al. (2005) were between 1.2  105 and 5.5  105 CFU/100 ml, or about 3e300 times larger than in our experiments. In fact, the resuspension velocities vr ¼ Rai/ C2 in our experiments (4  107e1  106 m/s) were only 2e25 times smaller than the values (2  106e1  105 m/s) Jamieson et al. (2005) observed. The critical shear stress for cohesive sediment was about 1.1 N/m2 at all sampling stations. Because of the effects of

121

Predicting resuspension

Using values of parameters from previous work and a constant value of the particle size (Table 1), the model predicted thirteen of the resuspension rates within a factor of 2 and all within a factor of 5 (Fig. 4a). The model predicted the resuspension rates from the main channel and tributaries about equally well. As noted in Section 2, most of the parameters in Table 1 were either measured or taken from Lick (2009). The coefficient c5 was set to 21 N/m2; this value is three times that used by Lick et al. (2004) for quartz particles with 2% bentonite. Because grain size analyses showed that the sediment samples consisted of between 1 and 7% clay (i.e., particle sizes < 8 mm), a larger value of c5 is reasonable. The diameter d of particles to which E. coli attach must be specified to compute both the predicted and inferred resuspension rates. A single value of 1 mm used for all sites yielded s ¼ 1.04 and a skill of 0.82. Attachment of E. coli to small particles is consistent with previous findings. For example, Oliver et al. (2007) observed that 65% of E. coli attached to particles smaller than 2 mm. Because of the uncertainty in the diameter of particles to which E. coli attach, the diameter was allowed to depend on the bottom shear stress as d ¼ asb, where a is a coefficient. The optimal value of a of 1.9 mm/Pa yielded diameters between 0.5 and 3.5 mm, which fall within previously observed ranges (Oliver et al., 2007), and it changed the range of inferred resuspension rates because of the dependence of the settling velocity on particle diameter. The model predicted all of the within a factor of 2 (Fig. 4b), and it yielded s ¼ 0.40 and a skill of 0.85. Again, the model predicted the resuspension rates from the main channel and tributaries about equally well.

122

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 5 e1 2 6

Table 2 e Relative sensitivity of the predicted resuspension rate to the various parameters. As noted in the text, the parameters fb [aexpðbrb Þ=d2 and fc [ c5/c3d represent the contributions of bulk density and clay content, respectively, to the critical shear stress sc. Parameter

Relative sensitivity

Parameter

Relative sensitivity

Hydraulic radius R

na

sb sb  scn

Exponent na


 s  scn na ln b sc  scn

Slope S

na

sb sb  scn

Bulk density rb

na

brb fb fb þ fc br f na b b fb þ fc

Concentration Ca

1

Coefficient b

Coefficient E0a

1

Coefficient a

na

Diameter d

na

Coefficient c5

na




fc fb þ fc

This version of the model involves no more calibration parameters, and the relationship between the diameter and shear stress appeals to physical intuition. Once the exponent na and coefficient c5 were specified using information from Amos et al. (1996) and Lick (2009), the only parameter to adjust in the first application (Fig. 4a) was the diameter d. The second application (Fig. 4b) also had only one parameter to adjust, the coefficient a. The assumed relationship d ¼ asb implies that as the bottom shear stress increases, larger particles can be resuspended. Also, the coefficient a can be related to the Shields parameter, which is used to determine conditions under which non-cohesive sediment will start moving: j¼

sb a ¼ : rgðs  1Þd rgðs  1Þ

(13)

With a ¼ 1.9 mm/Pa Eq. (13) yields a Shields parameter of about 33. This value is much larger than the critical Shields parameter for initiation of motion of 2 mm quartz particles (Cao et al., 2006). The larger value we obtained is reasonable because it deals with suspended sediment instead of initiation of motion and because the Shields criterion does not account for the cohesive effects involved in the transport of small particles.

4.3.

fb fb þ fc fb scn  fb þ fc sb  scn



sediment (sb [ scn) and effects of clay control the critical shear stress for cohesive sediment (fc [ fb). For similar reasons, the sensitivity to the bulk density, particle diameter, and coefficients a and b in Eq. (3) is smaller. When the effects of bulk density outweigh those of the clay content (fb [ fc)d for example, for a soil with greater bulk density or smaller clay content (i.e., reduced c5), most of the sensitivities change little, but the resuspension rate becomes most sensitive to the bulk density and the coefficient b because of the exponential dependence on both. Although the predicted resuspension rate is not sensitive to the particle diameter, the inferred resuspension rate can be. With a settling velocity computed with Stokes’s law, the inferred rate is always twice as sensitive to the diameter as it is to the concentration of E. coli in the water column and the attached fraction (Table 3). When settling is more important than net growth in the mass balance for E. coli in the sediment (F < 1) as at station 14, the inferred resuspension rate is most

Sensitivity and uncertainty

Calculating the sensitivity can help in determining the parameters for other situations. The relative sensitivity can be computed analytically (Table 2); the parameters fb ¼ aexpðbrb Þ=d2 and fc ¼ c5/c3d represent the contributions of bulk density and clay content, respectively, to the critical shear stress defined in Eq. (3). Because the resuspension rate is linearly proportional to both the coefficient E0a and the concentration Ca of attached E. coli in the sediment, the relative sensitivity to those parameters is always 1. All other sensitivities depend on the parameter values (Fig. 5). For the parameter set used in Fig. 4b, the resuspension rate is most sensitive to the slope, hydraulic radius (which for these measurements is approximately equal to the water depth), the concentration of attached E. coli, and the coefficients E0a and c5. The magnitude of these sensitivities is approximately equal to the exponent na, or 1, because the bottom shear stress is much greater than the critical shear stress for non-cohesive

Fig. 5 e Absolute value of the relative sensitivity of the predicted resuspension rate to the parameters listed in Table 2. Sensitivity is computed for station 13. Black bars are computed for the parameter set used to compute the rates in Fig. 4b. Gray bars use the base parameter set with c5 [ 2.5 N/m2, and white bars use the base parameter set with rb [ 1.45 g/cm3.

123

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 5 e1 2 6

Table 3 e Relative sensitivity of the inferred resuspension rate to the various parameters. As noted in the text, the parameter F[kn2 H2 C2 =fa ws C1 measures the relative importance of settling and net growth in the mass balance for E. coli in the sediment. The net growth rate kn2 and its derivative with respect to temperature are taken from Hipsey et al. (2008). Parameter Diameter d Conc. C1 in water column Attached fraction fa

Relative sensitivity

Parameter

2 1þF 1 1þF 1 1þF

sensitive to the diameter and less sensitive to the concentration of E. coli in the sediment, the depth of sediment containing E. coli, and the water temperature T (Fig. 6). For larger values of F, when net growth is more important that settling, the inferred rate is most sensitive to temperature except for temperatures corresponding to the peak in the net growth rate (i.e., vkn2/vT ¼ 0). The growth and decay relations in Hipsey et al. (2008) suggest that sensitivity to temperature will be small around temperatures of 22.6  C. The differences between these two cases is illustrated by the sensitivities for stations 6 and 11 (Fig. 6). Uncertainty is 30% in the predicted resuspension rate and 47e77% in the inferred resuspension rate. For the individual uncertainties in the parameters listed in Table 1, about 75% of the uncertainty in the predicted rate comes from the slope and the concentration Ca of E. coli attached to sediment. Another 20% comes from the coefficient c5 and the hydraulic radius, which is approximately equal to the water depth in most of our cases, and the remaining uncertainty comes from the exponent na. Efforts to reduce uncertainty in the predictions should involve better estimates of Ca and either measuring the slope more accurately or measuring the bottom shear stress with another method, such as one based on velocity measurements at a cross section (e.g., Kim et al., 2000).

Conc. C2 in sediment Depth H2 with E. coli Temperature T

Relative sensitivity F 1þF F 1þF F T vkn2 1 þ F kn2 vT

The main contributions to the uncertainty in the inferred resuspension rate depend on F. When net growth is more important than settling (large F), the depth of sediment containing E. coli controls the uncertainty, and when settling is more important than net growth (small F), the particle diameter controls the uncertainty. In the former case, the uncertainty can be reduced by measuring E. coli concentrations at different depths in the sediment; such measurements would also allow the assumption of uniform concentration to be assessed and revised. In the latter case, the uncertainty occurs because of the dependence on particle diameter through the settling velocity. Reducing the uncertainty in the settling velocitydand thus the resuspension ratedis difficult for several reasons. As Rehmann and Soupir (2009) reviewed in detail, flocculation, which can control the deposition of cohesive sediment (Droppo, 2001), can cause Stokes’s law to overestimate the settling velocity (Burban et al., 1990). Even without flocculation, settling velocities in a flowing stream fall below those from Stokes’s law far from the bed and exceed them near the bed (Cuthbertson and Ervine, 2007). Further uncertainty is introduced by the range of particle sizes present in stream sediment and tendency of E. coli to attach to particles of different sizes (Oliver et al., 2007).

4.4.

Model assessment

The key advantages of our model are that its parameters are related to observable physical quantities and that it accounts for the properties of the flow, sediment, and organisms. Alternative models for computing resuspension rates include those that assume a constant resuspension velocity vr (Chapra, 1997) and those that relate resuspension to the discharge Q using a formula of the form Ra ¼ a1 Ca Q b1 ;

Fig. 6 e Absolute value of the relative sensitivity of the inferred resuspension rate to the parameters listed in Table 3: Black bars are computed for station 14 (F [ 0.4, T [ 17.7  C), gray bars are computed for station 6 (F [ 43, T [ 19.0  C), and white bars are computed for station 11 (F [ 16, T [ 22.7  C).

(14)

where a1 and b1 are coefficients. Examples of models like (14) include those of Wu et al. (2009), who computed the concentration of resuspended organisms, and Collins and Rutherford (2004), who computed the number of E. coli resuspended per unit time. Predictions with a constant resuspension velocity of 5.2  107 m/s and Eq. (14) with a1 ¼ 8  107 and b1 ¼ 0.29 (with discharge expressed in m3/s) also provide good fits to the inferred resuspension rates in Fig. 4b (Table 4). However, choosing the parameters in these two models is difficult in situations without measured or inferred resuspension rates to be used for calibration. For example, to specify the resuspension rate in their model, Petersen et al. (2009) used the average of the resuspension rates reported

124

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 5 e1 2 6

Table 4 e Comparison of methods of predicting resuspension rates. The last two models were evaluated with the dataset computed with variable particle diameter. The last two columns show the number of predictions within factors of 2 and 5 of the measured values. Model

Eq. (5), constant d Eq. (5), constant j Constant vr Eq. (14)

s

1.04 0.40 0.45 0.27

Skill

0.82 0.85 0.98 0.95

Number within a factor of. 2

5

13 16 14 15

16 16 16 16

by Jamieson et al. (2005). As noted in Section 4.1, that rate is much higher than the inferred rates from our study. The ranges of resuspension velocity are closer, but even with the smallest value of vr from Jamieson et al. (2005)dwhich is four times larger than the optimal value, the predictions using constant resuspension velocity are worse than all four cases in Table 4. The coefficients a1 and b1 in (14) are even harder to specify: Collins and Rutherford (2004) did not report their values of the coefficients. Wu et al. (2009) related the concentration of resuspended organisms to Q4.5; this exponent is much larger than b1, and the strong dependence on flow was not reflected in our measurements of E. coli concentrations in the water column. In contrast, the parameters in Eq. (5) can be measured, observed, or estimated from previous work; the most challenging parameters to specify are the exponent na, which was taken from the work of Amos et al. (1996); the coefficient c5, which was estimated from the clay fraction and the results in Lick (2009); and the particle diameter, which was discussed in detail in Section 4.1. The ability of Eq. (5) to account for sediment properties gives it wider applicability than Eq. (14). For the data in Fig. 4b, the bottom shear stress is much larger than the critical shear stress for non-cohesive sediment, and the binding effects of clay make the critical shear stress for cohesive sediment depend only weakly on the particle diameter (Fig. 1). With sc approximately constant, the resuspension rate from (5) is proportional to Ca snba , and if the bottom shear stress can be expressed as a function of the discharge raised to some power, then Eq. (14) should work well. However, in streams with sediment that has a larger bulk density or a smaller clay fraction, the critical shear stress for cohesive sediment will not be constant, and predictions with Eq. (14) will not be as successful. The proposed formula (5) for predicting resuspension rates can in principle be applied in unsteady flow. In contrast, a model with specified resuspension velocity would be difficult to apply because the velocity would have to vary in time. The ability to use (5) to predict resuspension in unsteady flows is important because resuspension typically is largest during the rising limb of storm hydrographs (Jamieson et al., 2005). To apply Eq. (5), estimates of the shear stress would need to be obtained by modifying the force balance in Eq. (6) by considering effects of unsteadiness and nonuniformity or showing that they are negligible, as in Jamieson et al. (2005). Our study also demonstrates the challenge of estimating resuspension from field measurements. The inferred

resuspension rate from Eq. (8) was computed from a steadystate mass balance in Eq. (7). An analysis similar to that of Rehmann and Soupir (2009) shows that the flow in Squaw Creek was approximately steady: The time scale of unsteadinessdestimated as Q/(dQ/dt) using discharge measured at the U.S. Geological Survey’s gaging station at our station 16dwas about 11.5 h. This time scale is about 6 times larger than the time scale for settling (C2H2/C1faws), 20 times larger than the time scale for net growth (k1 n2 ), and 30 times larger than the time scale for resuspension (H2/vr). Therefore, the mass balance in Eq. (7) should hold approximately. Still, as discussed in Section 4.3, resuspension rates inferred with Eq. (8) are uncertain because they require estimates of the settling velocity and depth of sediment containing E. coli, and the various processes contributing to growth and decay of E. coli (Hipsey et al., 2008) are difficult to quantify in the field. Future work involves incorporating the resuspension rate in Eq. (5) in watershed-scale models such as SWAT. Such models provide discharge and channel geometry, from which shear stresses can be estimated. Spatial variations in quantities such as sediment properties can pose a challenge, especially in cases in which the resuspension rate is sensitive to the bulk density. However, because our model shows that variations in sediment properties become less important when the binding effects of clay control the critical shear stress sc, in those casesdas in the case of the Squaw Creek watersheddthe model should be easier to apply. Also, our use of the model for Squaw Creek, as well as future comparisons with the performance of other watershed-scale simulations including resuspension (Collins and Rutherford, 2004; Wu et al., 2009; Kim et al., 2010), will guide users in selecting the model’s parameters. The resulting model should help in creating plans to improve water quality in areas affected by E. coli contamination.

5.

Conclusions

We predicted resuspension of E. coli from sediment beds in streams by expressing the resuspension rate as the product of the concentration of E. coli attached to sediment particles and an erosion rate adapted from work on sediment transport. The model accounts for properties of the flow through the bottom shear stress and properties of the sediment through the critical shear stresses for cohesive and non-cohesive sediment. To evaluate the model’s predictive skill, its predictions were compared to resuspension rates inferred from a steady mass balance applied to measurements at sixteen locations in a watershed. Sensitivity and uncertainty were computed to determine the parameters that affect the predictions most strongly and to identify ways to improve the model. The main conclusions of this study are as follows: 1. The model performed well using parameter values mostly taken from previous work. The coefficient representing the binding effects of clay was increased from a previously reported value because of the higher clay content in the sediment in our study. The application of the model in which the particle diameter was linearly proportional to the bottom shear stress (i.e., constant Shields parameter)

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 5 e1 2 6

performed better than an application with constant particle diameter, while maintaining the same number of model coefficients. 2. Although two simpler models also performed well, the proposed model can be applied more easily in situations without measured or inferred resuspension rates because its parameters are related to observable physical quantities and it accounts for properties of the flow, sediment, and organisms. Furthermore, its ability to be applied in unsteady flow is important because resuspension is often largest during the rising limb of a storm hydrograph. 3. When the binding effects of clay control the critical shear stress, the predicted resuspension rate is more sensitive to properties of the flow, and when the bulk density controls the critical shear stress, the predicted resuspension rate is more sensitive to properties of the sediment. For the current data set, the uncertainty in the predictions would be reduced by reducing uncertainty in the concentration of attached E. coli and the slope used to compute the bottom shear stress.

Acknowledgments The authors thank the U.S. Environmental Protection Agency Region 7 (contract no. X7-97703701-1) for generous support of this work and Kendal Agee, Andrew Paxson, Charles Velasquez, and Ray Sims for assistance with sample collection and analysis. The first two authors acknowledge support from the National Science Foundation under grant 0967845; any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.

references

American Public Health Association (APHA), 1999. Standard Methods for the Examination of Water and Wastewater. AWWA, Water Environment Federation. Amos, C.L., Sutherland, T.F., Zevenhuizen, J., 1996. The stability of sublittoral, fine-grained sediments in a subarctic estuary. Sedimentology 43 (1), 1e19. Auer, M.T., Niehaus, S.L., 1993. Modeling fecal coliform bacteria. I. Field and laboratory determination of loss kinetics. Water Research 27 (4), 693e701. Bai, S., Lung, W.-S., 2005. Modeling sediment impact on the transport of fecal bacteria. Water Research 39 (20), 5232e5240. Black, K.S., Tolhurst, T.J., Paterson, D.M., Hagerthey, S.E., 2002. Working with natural cohesive sediments. Journal of Hydraulic Engineering 128 (1), 2e8. Buckley, R., Clough, E., Warnken, W., Wild, C., 1998. Coliform bacteria in streambed sediments in a subtropical rainforest conservation reserve. Water Research 32 (6), 1852e1856. Burban, P.Y., Xu, Y.J., McNeil, J., Lick, W., 1990. Settling speeds of flocs in freshwater and seawater. Journal of Geophysical Research 95 (C10), 18213e18220. Cao, Z.X., Pender, G., Meng, J., 2006. Explicit formulation of the Shields diagram for incipient motion of sediment. Journal of Hydraulic Engineering 132 (10), 1097e1099.

125

Chapra, S.C., 1997. Surface Water-Quality Modeling. McGraw-Hill, New York, NY, p. 510. ISBN 0-07-011364-5. Characklis, G.W., Dilts, M.J., Simmons III, O.D., Likirdopulos, C.A., Krometis, L.-A.H., Sobsey, M.D., 2005. Microbial partitioning to settleable particles in stormwater. Water Research 39 (9), 1773e1782. Chow, V.T., 1959. Open-channel Hydraulics. McGraw-Hill, New York, NY, p. 112. ISBN 07-010776-9. Collins, R., Rutherford, K., 2004. Modelling bacterial water quality in streams draining pastoral land. Water Research 38 (3), 700e712. Cuthbertson, A.J.S., Ervine, D.A., 2007. Experimental study of fine sand particle settling in turbulent open channel flows over rough porous beds. Journal of Hydraulic Engineering 133 (8), 905e916. Davies, C.M., Bavor, H.J., 2000. The fate of stormwater-associated bacteria in constructed wetland and water pollution control pond systems. Journal of Applied Microbiology 89 (2), 349e360. Droppo, I.G., 2001. Rethinking what constitutes suspended sediment. Hydrological Processes 15 (9), 1551e1564. Droppo, I.G., Lau, Y.L., Mitchell, C., 2001. The effect of depositional history on contaminated bed sediment stability. Science of the Total Environment 266 (1e3), 7e13. Droppo, I.G., Liss, S.N., Williams, D., Nelson, T., Jaskot, C., Trapp, B., 2009. Dynamic existence of waterborne pathogens within river sediment compartments: implications for water quality regulatory affairs. Environmental Science & Technology 43 (6), 1737e1743. El Ganaoui, O., Schaaff, E., Boyer, P., Amielh, M., Anselmet, F., Grenz, C., 2004. The deposition and erosion of cohesive sediments determined by a multi-class model. Estuarine Coastal and Shelf Science 60 (3), 457e475. Ferguson, R.I., 2005. Estimating critical stream power for bedload transport calculations in gravel-bed rivers. Geomorphology 70 (1e2), 33e41. Fries, J.S., Characklis, G.W., Noble, R.T., 2008. Sediment-water exchange of Vibrio sp. and fecal indicator bacteria: Implications for persistence and transport in the Neuse River Estuary, North Carolina, USA. Water Research 42 (4e5), 941e950. Haan, C.T., 2002. Statistical Methods in Hydrology. Iowa State University Press, USA, p. 391. ISBN 978e0813815107. Hipsey, M.R., Antenucci, J.P., Brookes, J.D., 2008. A generic, process-based model of microbial pollution in aquatic systems. Water Resources Research 44 (7), W07408. doi:10.1029/02007wr006395. Iowa Natural Resources Conservation Service, 2011. Information about Soils. http://www.ia.nrcs.usda.gov/soils.html Accessed 15.06.11. Jamieson, R.C., Joy, D.M., Lee, H., Kostaschuk, R., Gordon, R.J., 2005. Resuspension of sediment-associated Escherichia coli in a natural stream. Journal of Environmental Quality 34 (2), 581e589. Kim, J.W., Pachepsky, Y.A., Shelton, D.R., Coppock, C., 2010. Effect of streambed bacteria release on E. coli concentrations: monitoring and modeling with the modified SWAT. Ecological Modelling 221 (12), 1592e1604. Kim, S.C., Friedrichs, C.T., Maa, J.P.Y., Wright, L.D., 2000. Estimating bottom stress in tidal boundary layer from acoustic Doppler velocimeter data. Journal of Hydraulic Engineering 126 (6), 399e406. Krishnappan, B.G., 2007. Recent advances in basic and applied research in cohesive sediment transport in aquatic systems. Canadian Journal of Civil Engineering 34 (6), 731e743. Krometis, L.A.H., Characklis, G.W., Simmons, O.D., Dilts, M.J., Likirdopulos, C.A., Sobsey, M.D., 2007. Intra-storm variability in microbial partitioning and microbial loading rates. Water Research 41 (2), 506e516.

126

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 5 e1 2 6

Lick, W., 2009. Sediment and Contaminant Transport in Surface Waters. CRC Press, pp. 81e93. ISBN 978-1-4200-5987-8. Lick, W., Jin, L.J., Gailani, J., 2004. Initiation of movement of quartz particles. Journal of Hydraulic Engineering 130 (8), 755e761. Mehta, A., Lee, S.C., 1993. Problems in linking the threshold condition for the transport of cohesionless and cohesive sediment transport. Journal of Coastal Research 10 (1), 170e177. Mehta, A.J., 1989. On estuarine cohesive sediment suspension behavior. Journal of Geophysical Research-Oceans 94 (C10), 14303e14314. Nagels, J.W., Davies-Colley, R.J., Donnison, A.M., Muirhead, R.W., 2002. Faecal contamination over flood events in a pastoral agricultural stream in New Zealand. Water Science and Technology 45 (12), 45e52. Oliver, D.M., Clegg, C.D., Heathwaite, A.L., Haygarth, P.M., 2007. Preferential attachment of Escherichia coli to different particle size fractions of an agricultural grassland soil. Water Air and Soil Pollution 185 (1e4), 369e375. Partheniades, E., 1965. Erosion and deposition of cohesive soils. Journal of the Hydraulics Division-American Society of Civil Engineers 91, 105e139. Paterson, D.M., 1989. Short-term changes in the erodibility of intertidal cohesive sediments related to the migratory behavior of epipelic diatoms. Limnology and Oceanography 34 (1), 223e234. Petersen, C.M., Rifai, H.S., Stein, R., 2009. Bacteria load estimator spreadsheet tool for modeling spatial Escherichia coli loads to an urban bayou. Journal of Environmental Engineering 135 (4), 203e217. Rehmann, C.R., Soupir, M.L., 2009. Importance of interactions between the water column and the sediment for microbial concentrations in streams. Water Research 43 (18), 4579e4589. Roberts, J., Jepsen, R., Gotthard, D., Lick, W., 1998. Effects of particle size and bulk density on erosion of quartz particles. Journal of Hydraulic Engineering 124 (12), 1261e1267. Sanders, B.F., Arega, F., Sutula, M., 2005. Modeling the dryweather tidal cycling of fecal indicator bacteria in surface waters of an intertidal wetland. Water Research 39 (14), 3394e3408.

Stone, M., Krishnappan, B.G., Emelko, M.B., 2008. The effect of bed age and shear stress on the particle morphology of eroded cohesive river sediment in an annular flume. Water Research 42 (15), 4179e4187. Sutherland, T.F., Amos, C.L., Grant, J., 1998. The effect of buoyant biofilms on the erodibility of sublittoral sediments of a temperate microtidal estuary. Limnology and Oceanography 43 (2), 225e235. Taylor, J.R., 1997. An Introduction to Error Analysis: The Study of Uncertainties in Physical Measurements. University Science Books, Sausalito, CA, p. 75. ISBN 0-93570242-3. Tian, Y.Q., Gong, P., Radke, J.D., Scarborough, J., 2002. Spatial and temporal modeling of microbial contaminants on grazing farmlands. Journal of Environmental Quality 31 (3), 860e869. U.S. Environmental Protection Agency, 2001. Protocol for Developing Pathogen TMDLs. EPA 841-R-00e002, Office of Water (4503F) Washington, D.C, 6-4. U.S. Environmental Protection Agency, 2010. WATERS (Watershed Assessment, Tracking & Environmental ResultS) Washington, D.C. van Rijn, L.C., 2007. Unified view of sediment transport by currents and waves. I: initiation of motion, bed roughness, and bed-load transport. Journal of Hydraulic Engineering 133 (6), 649e667. Whitman, R.L., Nevers, M.B., Byappanahalli, M.N., 2006. Examination of the watershed-wide distribution of Escherichia coli along southern Lake Michigan: an integrated approach. Applied and Environmental Microbiology 72 (11), 7301e7310. Wilkinson, J., Jenkins, A., Wyer, M., Kay, D., 1995. Modelling faecal coliform dynamics in streams and rivers. Water Research 29 (3), 847e855. Willmott, C.J., 1981. On the validation of models. Physical Geography 2, 184e194. Wu, J.Y., Rees, P., Storrer, S., Alderisio, K., Dorner, S., 2009. Fate and transport modeling of potential pathogens: the contribution from sediments. Journal of the American Water Resources Association 45 (1), 35e44.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 2 7 e1 3 5

Available online at www.sciencedirect.com

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

Novel ferromagnetic nanoparticle composited PACls and their coagulation characteristics M. Zhang, F. Xiao, X.Z. Xu, D.S. Wang* State Key Laboratory of Environmental Aquatic Chemistry, Research Center for Eco-Environmental Sciences, CAS, POB 2871, Beijing 100085, China

article info

abstract

Article history:

Effects of magnetic nanoparticles on inorganic coagulants and their coagulation perfor-

Received 14 February 2011

mances were studied in the present work. The Fe3O4eSiO2 core-shell particle (FSCSP) and

Received in revised form

superfine iron (SI), were compounded with polyaluminium chloride of basicity 2.0 (PACl2.0),

9 September 2011

providing magnetic PACl2.0s (MPACl2.0s). The physiochemical properties of ferromagnetic

Accepted 16 October 2011

nanoparticles were investigated using transmission electron microscopy (TEM), the BET

Available online 25 October 2011

method and a zeta potentiometric analyzer. The Al species distributions of the MPACl2.0s

Keywords:

coagulation performances. Floc properties were assessed by use of the electromotive

Magnetic composited PACl

microscope (EM) and small angle laser light scattering (SALLS). The results showed that

Magnetic coagulation

modified layers of nanoparticles mitigated agglomeration. FSCSP had a larger specific area

Ferromagnetic nanoparticles

and pore volume than SI. The addition of ferromagnetic nanoparticles obviously increased

and PACl2.0 were examined by liquid

27

Al NMR. Jar tests were employed to evaluate the

Coagulation characteristics

the content of Alun. MPACl2.0s performed better than PACl2.0 in turbidity removal and DOC

Floc property

removal when dosed less than 0.06 mmol/L as Al. Generally, PACl2.0 þ FSCSP (50 mg/L) performed best. Large, loose and weak flocs were produced by MPACl2.0s, which were preferred for the magnetic powder recycling. A plausible structure, Al species-nanoparticles cluster, contributing to the unique properties of MPACl2.0 flocs, was proposed. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Magnetic coagulation, the combination of magnetic matericals and coagulants, has been tested for its effectiveness. The presence of ferromagnetic materials and magnetic fields makes magnetic collection and separation possible, leading to higher efficiency, larger handling capacity, easier manipulation and comparatively lower energy consumption (Ambashta and Sillanpa¨a¨, 2010). Hence, investigations have been mainly focused on several aspects as follows: (I) Seeding of magnetic particles in the treating water to produce magnetic flocs that can be rapidaly collected in a magnetic separator (Li et al., 2010); (II) Integration of magnetic ion exchange (MIEX)

hybrid systems (Singer and Bilyk, 2002) to improve treatment efficiency (Zhang et al., 2007; Korbutowicz et al., 2008); (III) Development and improvement of the magnetic separation system/instrument (Svoboda and Fujita, 2003); (IV) Development of magnetic coagulants by compounding magnetic materials with traditional coagulants. Although magnetic seeding coagulation and MIEX resin adsorption can greatly improve removal efficiency, the expense of the magnetic powder and MIEX resin have restricted their practical application. Nanosorbents, nanocatalysts and so on have been evaluated (Savage and Diallo, 2005). In aqueous systems, iron oxide particles are hydrated, and FeeOH groups can completely cover their surface.

* Corresponding author. Tel./fax: þ86 10 62849138. E-mail addresses: [email protected] (M. Zhang), [email protected] (F. Xiao), [email protected] (X.Z. Xu), [email protected] (D.S. Wang). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.025

128

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 2 7 e1 3 5

Hydrous iron oxides have an amphoteric character. The FeeOH sites on the surface can react with Hþ or OH ions from dissolved acids or bases. A positive (FeeOH2þ) or negative (FeeO) charges can be formed on the surface by protolytic reactions depending on the pH of the electrolyte solution (Ille´s and Tomba´cz, 2006). Magnetic nanoparticles exhibit a finitesize effect and/or high surface-to-volume ratio, resulting in a higher adsorption capacity. Additionally, the magnetic flocculants can be recovered by an external magnetic field. Therefore, there is high interest in the application of magnetic nanomaterials to water treatment (Shen et al., 2009). Composite magnetic coagulants are considered to have advantages over traditional coagulants since their magnetic components may optimize coagulation behaviors and facilitate the magnetic separation following the coagulation process. Jiang et al. (2010) successfully synthesized magnetic polyferric chloride (MPFC) by combining Fe3O4 nanoparticles with PFC. The MPFC, a physical mixture, gave a synergistic improvement in Microcystis aeruginosa removal efficiency compared to only PFC. Coagulated flocs showed a higher settling velocity by wrapping up Fe3O4. Liu et al. (2006) developed a special flocculant to control freshwater cyanobacterial blooms. In their study, hydrochloride acid was firstly mixed with flyash in a proper ratio for modification, and the mixture and the magnetic powder were then put into the algae contaminated water; the aquatic hazardous substances (algal toxins) were thereby effectively adsorbed and removed. Polyaluminum chloride (PACl) has been widely used and claimed to be superior to other traditional coagulants since it had the advantages of less alkalinity consumption and wider temperature and pH tolerance (Odegaard et al., 1990; Van Benschoten and Edzwald, 1990). The hydrolysis products of PACl comprise a series of Al species, especially Al13 polycation, [AlO4Al12(OH)24(H2O)12]7þ, which has been widely accepted as the critical species in particle aggregation by strong charge neutralization (Johansson, 1960). In addition, PACl can be temporarily refractory to hydrolysis before adsorption onto particle surfaces (Hu et al., 2006). Thus, combining ferromagnetic nanoparticles with PACl should produce favorable results. However, almost no research has addressed this to date, especially the magnetic coagulation mechanism and the characterizations of the formed flocs. In the present work, ferromagnetic nanoparticles were mixed with PACl2.0 to develop a novel composite magnetic coagulant MAPCl2.0. The developed coagulant was characterized in terms of Al species distributions by a liquid 27Al NMR method. In addition, jar tests were conducted to evaluate the MPACl2.0’s performance by examining turbidity and DOC removal. The formation, breakage and regrowth of MPACl2.0 flocs were subsequently investigated. Finally, a model was proposed to elucidate the coagulation mechanisms of MPACl2.0s with special emphasis on the role of the ferromagnetic nanoparticles.

2.

Material and methods

2.1.

Chemicals

FeCl3$6H2O, FeCl2$4H2O, NaHCO3, tetraethyl orthosilicate (TEOS) and absolute ethanol were obtained from the

National Medicines Corporation Ltd. of China. NH3$H2O, NaOH and hydrochloric acid were provided by Beijing Chemical Plant. AlCl3$6H2O， NaNO3 and Kaolin were obtained from the Xilong Chemical Corporation Ltd. of Shantou, the Jinke Fine Chemical Institute of Tianjin, and the Dongxu Chemical Plant of Beijing, China, respectively. Humic acid was produced by SigmaeAldrich Corporation Ltd. of USA. All reagents were of analytical grade without further purification except for being specified, and deionized water was used in preparing all solutions.

2.2.

Preparation of the ferromagnetic nanoparticles

Fe3O4eSiO2 core-shell particles (FSCSP) were prepared in the laboratory. Firstly, Fe3O4 particles were synthesized by coprecipitation (Hong et al., 2009): FeCl3 and FeCl2 with a molar ratio of 1.7 were prepared in an N2 atmosphere. Excess ammonia aqueous solution was then quickly added into the Fe3þ/Fe2þ solution with vigorous stirring until no precipitates could be seen in the solution, and then another 1 h’s stirring was conducted. Finally, precipitates were collected and washed for several times with deionized water and ethanol until the pH value decreased to 7. A SiO2 layer was built up by directly introducing Fe3O4 particles into the primary silica particles by the Sto¨ber process (Lu et al., 2008): a certain amount of ethanol, deionized water, aqueous ammonia and TEOS were added in a three neck flask in a 40  C water bath. Fe3O4 particles were added into the flask at different time under mechanical stirring. The final product was washed, dried at 50  C and conserved in the dryer. The other ferromagnetic nanoparticle used in this study is the superfine iron (SI), which was provided by the Research Center for Nano Technology of the Chinese Iron and Steel Institute. Its physiochemical properties may be found in the Master’s thesis of Chunfeng Hou (2009).

2.3.

Synthesis of PACl2.0 and of MPACl2.0s

The PACl2.0 was prepared by slow base titration of 0.5 mol/L AlCl3 solution with 0.2 mol/L NaHCO3 solution under rapid stirring to achieve the target molar OH/Al ratio (basicity value) of 2.0 (Tang, 2006). MPACl2.0s were synthesized by blending PACl2.0 with ferromagnetic nanoparticles on a shaking table at room temperature (25  C) for 2 h with a speed of 250 rpm. The final composited MPACl2.0s were aged for 24 h, and conserved at 4  C. FSCSPs, two loading concentrations of 50 and 100 mg/L, were respectively added into PACl2.0 solutions. Hence the formed MPACl2.0s can be described as PACl2.0 þ FSCSP (50 mg/L) and PACl2.0 þ FSCSP (100 mg/L) respectively. Loading concentrations were determined by means of apparent dispersibility. It may be noted that FSCSP aggregated more easily than the SI. Therefore, in order to make the ferromagnetic nanoparticles fully distributed at the surface of PACl2.0, only 25 mg/L was chosen as the loading concentration of the SI (PACl2.0 þ SI (25 mg/L)).

129

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 2 7 e1 3 5

2.4. Characterization of the ferromagnetic nanoparticles, PACl2.0, and MPACl2.0s The particle size and morphology of FSCSP and SI were determined by transmission electron microscopy (TEM) (S-570, Hitachi, Japan). The surface charges of the two nanoparticles were analyzed with a zeta potential meter (Zetasizer, Malvern, UK) by adjusting the pH of their suspensions from 3 to 11 with 1.0 and 0.1 mol/L NaOH and HCl solutions. Moreover, zeta potentials of MPACl2.0s and PACl2.0 were also measured. The specific surface area and pore size distribution were measured by a Brunauer, Emmett, and Teller (BET) surface area analyzer (ASAP-2000, Micromeritics, US) for nitrogen adsorption. The BET method was carried out under relatively high vacuum and measured primarily the external area of the particles and aggregates. Al species of MPACl2.0s and PACl2.0 were characterized by 500 MHz 27Al NMR, and instrumental settings and experimental conditions were addressed elsewhere (Xu et al., 2003). The internal standard was 0.05 mol/L NaAlO2 with its chemical shift at 80 ppm downfield. The signals near 0 and 62.5 ppm represent mononuclear Al (Alm) and Al13, respectively. Alun (the amount of undetectable species) was obtained by: Alun ¼ AlT  Alm  Al13

2.5.

(1)

Preparation of synthetic water

A synthetic water was prepared by dissolving the HA and kaolin stock suspensions with 1 mmol/L NaNO3 and 0.8 mmol/L NaHCO3 so that the ionic strength and alkalinity of the solution could be kept at 1.0 mmol/L and 80 mg/L respectively. The pH was adjusted to 7.00  0.02 using 0.1 mol/L, 0.01 mol/L NaOH or/ and 0.1 mol/L, 0.01 mol/L HCl. The synthetic water had average measured zeta potential, turbidity, UV254 and DOC values of 21.4 mV, 75.6 NTU, 0.254 cm1 and 2.318 mg/L, respectively. All measurements were obtained at room temperature.

2.6.

Coagulation jar tests

range of 0.02e0.20 mmol/L as Al. Before dosing, the three kinds of MPACl2.0s were shaken well in order to disperse the composited systems as uniformly as possible. The jar-test procedure consisted of a 30 s premix (250 rpm), a 1 min rapid mix (200 rpm), a 15 min slow mix (30 rpm) and a 20 min settling period. A small amount of sample was taken immediately to measure the zeta potential (Malvern, Zetasizer 2000, UK) after a 30 s rapid mix. After settling, the supernatants were analyzed by a TOC analyzer (TOC-Vcph, Shimadzu, Japan) and UV254 (UVeVis 8500, Hitachi, Japan) after filtration through a 0.45 mm membrane. The turbidity was measured by a turbidimeter (Hach 2100P Turbidimeter, USA).

2.7.

Floc characterization

Flocs were immediately collected from beakers after the slow stirring phase of the jar test with great care to avoid unnecessary “breakage”. Image observation was done under a low resolution of 1 mm (10  objective) using an electromotive microscope (Axioskop 2 mot plus, Carl Zeiss Co., Germany). To investigate and compare the growth, breakage and regrowth of flocs produced by MPACl2.0s and PACl2.0, experiments similar to those of Zhu et al. (2009) were carried out. Briefly, coagulation tests were performed under the optimal dosage of 0.08 mmol/L as Al. After the slow stirring phase, flocs were suddenly exposed to a 1 min strong stirring of 400 rpm for breakage, and then regrowth was undertaken at 40 rpm for 15 min. Dynamic floc size was monitored in the whole procedure using SALLS (Mastersizer 2000, Malvern, UK). Size measurements were taken every 40 s. The inter-particle bonds that hold aggregate flocs together are considered as the cohesive strength of the flocs. A size ratio method is used here with an index (strength factor) to express the strength of particle flocs, i.e., Strength Factor ¼

Jar tests were performed on a programmable jar test apparatus (Daiyuan Jar Test Instruments, China) to investigate the coagulation performance of MPACl2.0s within the dosage

d2  100% d1

(2)

where d1 and d2 are the mean sizes of the flocs before and after the shear breakage, respectively. A higher value of the

Fig. 1 e TEM images of two ferromagnetic nanoparticles: (a) FSCSP, (b) SI.

130

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 2 7 e1 3 5

d3  d2  100% Recovery Factor ¼ d3  d2

Results and discussion

3.1.

Properties of ferromagnetic nanomaterials

3.1.1.

Size and morphologies of FSCSP and SI

TEM images of FSCSP and SI are shown in Fig. 1. In Fig. 1(a), FSCSPs can be easily observed in the spherical shape with an average size of 20 nm. The non-uniform-sized nanospheres form submicron aggregates with a visible 10 nm shell. The agglomeration is due to the magnetic dipoleedipole attraction and the large specific surface area of the magnetic particles (Liu et al., 2004), in spite of fierce stirring during the silica coating process. Fig. 1(b) confirms the core-shell structure of SI. The passivated layer is Fe3O4 according to the product description. The TEM results reveal that the sizes of the ferromagnetic nanoparticles are of nanoscale before agglomeration. It has been reported that Fe3O4 is superparamagnetic if its diameter is around 16 nm (Hong, 2008). Unlike ordinary ferromagnetic particles, superparamagnetic materials have no hysteresis effect e once the external magnetic field is cut off, the remnant magnetism will disappear. This characteristic is favored the magnetic separation process since the magnetic sludge or flocs may be easily scraped when the magnetic field is off (Ngomsik et al., 2005).

3.1.2.

Specific surface area and pore volume

Nanoparticles have been proven to have relatively large specific surface areas, leading to greater interface reaction rates (Li et al., 2006). The specific surface areas and pore diameters were 78.76 m2/g and 153.24 A for FSCSP and 22.51 m2/g and 63.48 A for SI respectively. They were determined by BET surface measurement. Apparently, compared with the commercially available SI, the lab-prepared FSCSP has a larger surface area with more sites to react with Al species or water pollutants.

3.1.3.

FSCSP SI

20

10

0 2

3

4

5

6

7

8

9

10

11

12

pH value -10

(3)

where d3 is the mean size of the particle flocs after reflocculation at the original shear rate. A higher Recovery Factor suggests a greater flocculation and regrowth capability of the flocs after the shear breakage.

3.

30

Zeta Potential/mV

strength factor indicates a higher ability of the flocs to resist breakage when exposed to an elevated fluid shear. When the shear intensity was reduced after the breakage phase, re-flocculation of the particles could take place. A reversibility factor is used here to measure the reflocculation potential of the particles when the shear is returned to its original level. A modified size ratio approach may be applied to calculate the reversibility (Recovery Factor) by

-20

-30

Fig. 2 e Variation of the zeta potentials of the ferromagnetic nanoparticles with pH.

point (IEP) of FSCSP is at pH 5.2 while that of SI is at pH 6.6. It is worth noting that magnetite can produce charges in the protonation and deprotonation reactions of ^FeeOH surface sites as expected for an amorphous solid (Hajdu´ et al., 2009). Therefore, both particles can be positively charged in composited coagulants since the pH value of PACl2.0 is around 4.02, lower than IEPs of SI and FSCSP (Table 1). Zeta potentials of the MPACl2.0s apparently rise with the addition of magnetic nanoparticles, and the pH values of PACl2.0 and MPACl2.0s are very close around 4.0 Among the three coagulants, PACl2.0 þ FSCSP (100 mg/L) has a highest zeta potential of 21.5 mV PACl2.0 þ FSCSP (50 mg/L) has a lowest zeta potential. The zeta potential of PACl2.0 þ SI (25 mg/L) is 20.5 mV.

3.2.

Al species distributions of MPACl2.0s and PACl2.0

All the coagulants were aged for 1 week. The liquid 27Al NMR method was used to characterize Al species distributions in MPACl2.0s and the original PACl2.0. The results are shown in Fig. 3. The Alm, Al13 and Alun contents of PACl2.0 were 12.8%, 63.0% and 24.2%, respectively. It should be noted that the Alm and Al13 contents decline whereas the Alun content increases after compounding with ferromagnetic nanoparticles. Such change is more obvious for the two PACl2.0 þ FSCSPs. When the loading concentration is 50 mg/ L, Alm and Al13 contents decrease to 9.9% and 54.0%, respectively. For 100 mg/L, they further decrease to 9.3% and 52.5%, respectively.

Table 1 e Characteristics of electric charges for PACl2.0 and MPACl2.0s. Types of coagulants

pH

Zeta potential (mV)

PACl2.0 PACl2.0 þ FSCSP (50 mg/L) PACl2.0 þ FSCSP (100 mg/L) PACl2.0 þ SI (25 mg/L)

4.02 3.99 3.96 3.95

3.6 15.0 21.5 20.5

Surface electric charge

The electric charge characteristics of both nanoparticles are shown in Fig. 2. Zeta potentials of FSCSP and SI decrease with increase in pH values, changing from þ18.7 mV to þ26.0 mV down to 13.0 mV and 27.7 mV, respectively. The isoelectric

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 2 7 e1 3 5

3.3. Coagulation performances of MPACl2.0s and PACl2.0 with different dosages

100

Al species distributions/%

131

90 80 70 Alm

60

Al13

50

Alun

40 30 20 10 0 PACl2.0

PACl2.0+ FSCSP (50 mg/L)

PACl2.0+ FSCSP (100 mg/L)

PACl2.0+ SI (25 mg/L)

Fig. 3 e Comparison of Al species distributions.

All MPACl2.0s are in the form of solideliquid mixure. The high surface energy, strong adsorption and magnetic attraction of nanoparticles are apt to play a significant role in their combination with PACl2.0. The high density of reactive surface sites and the great intrinsic reactivity of ferromagnetic nanoparticles will enhance the reactivity of the nanoscale particles (Li et al., 2006), and allow Al species to adsorb on their surface. In the process of compositing and aging the MPACl2.0, it seems that Alm and Al13 may firstly aggregate into [Al13]n or [Alm]n on the surface ofthe nanoparticles, then transform to a sol or amorphous solid phase, and finally turn to other higher order aggregates. As a result, Al species-nanoparticle clusters form (shown in Fig. 7(a)), being identified as Alun. They are not only large in size and easy to precipitate but also have a crucial positive effect on bridging, enmeshment and sorption flocculation capacity. MPACl2.0s with different nanoparticles and loadings lead to different Al species distributions. According to the BET results, FSCSP has a larger surface area and pore volume, offering more adsorption sites for Al species. Therefore, the content of Al species-nanoparticle clusters in PACl2.0 þ FSCSP, in the presence of Alun, will rise. As for MPACl2.0 with SI, since SI was more positively charged in PACl2.0 (Table 1), the electrostatic repulsion between Al species and SI particles may hinder their approach and aggregation, resulting in an inconspicuous increase of Alun.

The HA and Kaolin particles removal efficiencies of MPACl2.0s and PACl2.0 and the zeta potential of the coagulated suspensions were comparatively investigated, as shown in Fig. 4. As anticipated, the addition of the MPACl2.0s and PACl2.0 effectively reduced the surface charge of the particles in the synthetic water, resulting in particle destabilization (Fig. 4). As the dosage increased, the particle zeta potentials approached zero. Further increase in flocculant dose caused a certain extent of charge reversal of the particles. The isoelectric dosage (IED) of PACl2.0 þ FSCSP (50 mg/L) is close to 0.0 mmo/L as Al, slightly lower than that of the others which were higher than 0.10 mmol/L as Al. The whole dosage range can be divided into three parts according to the change of zeta potentials and turbidity removals: (I) No significant flocculation from 30 to 10 mV. Dosages of coagulants are too low for MPACl2.0s and PACl2.0 to effectively flocculate/coagulate with water pollutants. (II) Obvious flocculation from 10 to 0 mV. As the dosage increases, big flocs are formed and zeta potentials are close to the IEP. Optimal turbidity removals are achieved in this range. (III) Restability from 0 to 10 mV. In the high dosage range, the coagulation performance is ineffective. The jar-test flocculation and sedimentation results for turbidity and DOC removals from the synthetic waters are in general agreement with the zeta potential analysis. For all four flocculants, it is obvious that the optimal dosage range for MPACl2.0s and PACl2.0 lies in Part II, and the optimal dosage is 0.08 mmol/L as Al. At the optimal dose, 90% of DOC and more than 90% of turbidity can be removed. Especially for PACl2.0 þ FSCSP (50 mg/L), the DOC and turbidity removal efficiencies can reach 92% and 98%, respectively. In comparison, PACl showed a slightly worse water treatment performance. The results indicated that the nanoscale magnetic particles did play a significant role in DOC and particle removal. Firstly, ferromagnetic nanoparticles could affect the hydrolyzed Al species by increasing Al species-nanoparticle clusters, which may increase effective collision rates. Secondly, the addition of ferromagnetic nanoparticles would enhance the adsorption capability due to their large specific surface areas and magnetic dipoleedipole attraction, resulting in a higher removal rate of water pollutants (Li et al., 2006). Therefore,

Fig. 4 e Variation of turbidity and DOC removal with coagulant dosages.

132

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 2 7 e1 3 5

Fig. 5 e Representative images of flocs formed with (a) PACl2.0; (b) PACl2.0 D FSCSP (50 mg/L); (c) PACl2.0 D FSCSP (100 mg/L) and (d) PACl2.0 D SI (25 mg/L).

3.4.

Floc properties

3.4.1.

Morphology of flocs

Microscopic examination in this study (Fig. 5) suggested the formation of large and fractal-like flocs, resulting in the removal of turbidity and DOC in the suspension. Representative images of flocs coagulated by PACl2.0, PACl2.0 þ FSCSP (50 mg/L), PACl2.0 þ FSCSP (100 mg/L), and SI (PACl2.0 þ SI (25 mg/L) are shown in Fig. 5. It may be noted the black dots were observed in the flocs formed by MPACl, revealing that nanoscale magnetic particles were successfully composited with the PACl. This can also imply the formation of Al speciesnanoparticle clusters. Those clusters influence not only the kinetics of aggregation, but also the structure and resulting fractal dimension of the aggregates.

3.4.2.

Floc formation, breakage and regrowth

The coagulation kinetics of PACl2.0 and MPACl2.0s are demonstrated in Fig. 6. It shows the variation in the average floc size during the hydrodynamic sequencing. In the jar tests, all the four coagulants were dosed of 0.08 mmol/L as Al. After the slow mix phase, the coagulation for four suspensions achieved a steady-state with the mean sizes denoted as d1 in Table 2. Then floc size was immediately reduced following an increase in shear. 1 min later, it can be assumed the flocs reached the size d2. A reversibility phenomenon in terms of

floc size was observed, as mentioned in previous studies (Yukselen and Gregory, 2002; Zhu et al., 2009). When the shear was reduced to its initial value, the broken particles could collide with each other again to form the larger ones. However, it may be noticed that the formed flocs could not regrow to anywhere near their previous size. This may be attributed to the different coagulation mechanisms. Flocs formed by charge neutralization should give total recoverability. Thus, the irreversible breakage of the flocs was considered as evidence that the flocs formed in these systems were not dominated by pure charge neutralization mechanisms and were therefore held together by chemical rather than physical bonds, such as the combination of entrapment

200

PACl2.0 PACl2.0+FSCSP (50 mg/L) PACl2.0+FSCSP (100 mg/L) PACl2.0+SI (25 mg/L)

160

Floc size/um

adsorption, enmeshment and sweep would have simultaneously taken place by adding MPACl2.0. That is the primary reason why the performance of MPACl2.0 coagulation is superior to that of PACl2.0 coagulation.

120

80

40

0 0

200

400

600

800

1000

1200

1400

1600

1800

2000

Time/s

Fig. 6 e Comparison of the floc formation kinetics curves.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 2 7 e1 3 5

Table 2 e Strength and recovery factors of MPACl2.0s and PACl2.0. Suspensions

d1 (mm)

d2 (mm)

d3 (mm)

Strength factor (%)

Recovery factor (%)

PACl2.0 PACl2.0 þ FSCSP (50 mg/L) PACl2.0 þ FSCSP (100 mg/L) PACl2.0 þ SI (25 mg/L)

136.5 165.4

43.6 42.9

73.9 51.3

31.9 25.9

32.6 6.9

139.0

46.7

77.9

33.7

33.8

173.3

46.4

61.5

26.8

11.9

bridging and complexation with coagulant metal hydrolysis species (Yukselen and Gregory, 2004). Table 2 summarizes the values of the strength factor and recovery factor for the suspensions with different coagulants. PACl2.0 þ SI (25 mg/L) and PACl2.0 þ FSCSP (50 mg/L) are less able to resist the shearing condition, and their strength factor values are 26.8% and 25.9%, respectively, lower than those of 33.7% for PACl2.0 þ FSCSP (100 mg/L) and 31.9% for PACl2.0. Meanwhile, recovery factors of PACl2.0 þ SI (25 mg/L) and PACl2.0 þ FSCSP (50 mg/L) are only as low as 6.9% and 11.9%. Compared with PACl2.0, the small addition of ferromagnetic nanoparticles slightly reduced the floc strength but severely weakened the recovery ability of the flocs. This can be explained according to the variation of the Al species shown in Fig. 3. The floc strength is dependent upon the interparticle bonds between the components of the aggregates. As reported by Wang et al. (2009), the Alm species can complex with HA to form large and strong flocs while the Al13 species react with HA to form small and unstable flocs. Those two HA-

133

Al flocs can be joined together by adsorption and bridging due to larger polymer and solid-phase Al(OH)3, resulting in larger flocs. In our study, the increase of Alun in MPACl2.0s, containing the probably existing Al species-nanoparticle clusters, may have helped to form larger flocs than PACl2.0. The decrease of Alm in MAPCl2.0s and the possible loose conformation of Al species-nanoparticle clusters may lead to weaker flocs. Moreover, the results also indicate that flocs formed by MPACl2.0s show lower strength and reflocculating ability after breakage. As required for the magnetic particle recycling, the smash of magnetic flocs/sludge is a prerequisite, after which the separation of magnetic particles from the flocs/sludge is easier to achieve. Therefore, the incompact flocs, induced by PACl2.0 þ FSCSP (50 mg/L) and PACl2.0 þ SI (25 mg/L), are favored for practical purposes because of the low energy consumption for flocs/sludge breakage.

3.5.

Mechanisms of nanoscale magnetic coagulants

The results of this work indicate that the ferromagnetic nanoparticles could play a significant role in coagulation. A plausible MPACl2.0 coagulation mechanism has been described schematically in Fig. 7. It can be noted that Al species-nanoparticle clusters will be formed when ferromagnetic nanoparticles are added into the PACl (as shown in Fig. 7(a)). The formed clusters would increase the proportion of Alun, which has a great potential to enhance charge neutralization, enmeshment and adsorption when aggregated with the pollutants. The flocs produced by those clusters exhibited different characteristics as shown in Fig. 4.

Fig. 7 e Coagulation mechanism of MPACl2.0: (a) formation of Al species-nanoparticle clusters in MPACl2.0, (b) coagulating process of MPACl2.0.

134

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 2 7 e1 3 5

As illustrated as Fig. 7(b), the flocs formed by PACl2.0 þ FSCSP (50 mg/L) and PACl2.0 þ SI (25 mg/L) are larger but weaker. As stated previously, ferromagnetic nanoparticles exhibited a preference to form Al species-nanoparticle clusters. These clusters enhance the bridge and adsorption effects, leading to a larger floc size. Moreover, the recovery factors of the flocs show marked differences. The recovery factors of PACl2.0 þ FSCSP (50 mg/L) and PACl2.0 þ SI (25 mg/L) are very low (6.9% and 11.9% respectively). This implies that, in a suitable range of ferromagnetic nanoparticle addition, the connections between the clusters and pollutants are unique, being based on chemical bonds rather than physical ones. Flocs formed by PACl2.0 þ FSCSP (100 mg/L) showed a different situation. They were stronger and had a comparable size with those formed by PACl. The difference is in the concentration of ferromagnetic nanoparticles. Apparently, PACl2.0 þ FSCSP (100 mg/L) had the highest concentration. Aggregation of nanoparticles happened easily, and thus Al species-nanoparticle clusters were reduced. Therefore, the bridge and adsorption effects of Alun were limited to some degree, and the floc size did not obviously increase. Additionally, the formed flocs were considerably porous and fractal. The excess nanoparticles could penetrate into flocs. The embodiment of those nanoparticles would increase the strength of the flocs as indicated in Table 2.

4.

Conclusions

Three novel ferromagnetic nanoparticle composites PACl2.0s (MPACl2.0s) were synthesized by compounding FSCSP and SI respectively with PACl2.0. They are described as PACl2.0 þ FSCSP (50 mg/L), PACl2.0 þ FSCSP (100 mg/L) and PACl2.0 þ SI (25 mg/L). A series of characterizations in terms of physiochemical properties illustrated that the ferromagnetic nanoparticles used in the study were all core-shell particles with slight agglomeration and positive charge PACl2.0. FSCSP had a larger surface area and pore volume than SI did. All MPACl2.0s were in the form of solideliquid mixing. Liquid 27Al NMR measurements indicated that, compared with PACl2.0, Alun content increased whereas Al13 and Alm declined with increase of ferromagnetic nanoparticle loadings. Al speciesnanoparticle clusters, attributed to Alun, are suggested to form. Coagulation jar tests revealed that the MPACl2.0s performed better than PACl2.0 in both turbidity removal, and DOC removal. Among the three MPACl2.0s, the coagulation behavior of PACl2.0 þ FSCSP (50 mg/L) was the best. The improved coagulation efficiency of MPACl2.0s may be considered to be due to the co-effect of Al species, ferromagnetic nanoparticles and the possible Al species-nanoparticle clusters. Investigation of floc properties revealed that the key factor that influenced floc growth, re-flocculation, size and strength was the loading of ferromagnetic nanoparticles. Moderate nanoparticle loadings induced large but weak flocs that were hard to rearrange. Such flocs are favored for magnetic particle recovery. The formation of Al species-nanoparticle clusters might have greatly contributed to the unique Al species distributions, the coagulation performance and the floc properties of

MPACl2.0. The additions of FSCSP and SI appeared to enhance the adsorption, enmeshment, and sweep abilities of MPACl2.0s, and to further strengthen the coagulated flocs by entering into the floc pores during flocculation. When the ferromagnetic nanoparticle loadings are controlled within a certain range, the connections between clusters and water pollutants were presumed to be chemical bonds since they had a considerably low recovery factors.

Acknowledgments This research was funded by the Natural Science Foundation of China under 50921064, 51025830 and 51008293. The authors are very grateful to technical support from the Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences.

references

Ambashta, R.D., Sillanpa¨a¨, M., 2010. Water purification using magnetic assistance: a review. Journal of Hazardous Materials 180, 38e49. Hajdu´, A., Ille´s, E., Tomba´cz, E., Borba´th, I., 2009. Surface charging, polyanionic coating and colloid stability of magnetite nanoparticles. Colloids and Surfaces A: Physicochemical and Engineering Aspects 347, 104e108. Hong, R.Y., 2008. Preparation and Application of Magnetic Nanoparticles and Fluid, first ed. Chemical Industry Press, China. 11e12. Hong, R.Y., Li, J.H., Zhang, S.Z., Li, H.Z., Zheng, Y., 2009. Preparation and characterization of silica-coated Fe3O4 nanoparticles used as precursor of ferrofluids. Applied Surface Science 255, 3485e3492. Hou, C. F., 2009. The characterization of nanoscale-iron and synthetic resins and the test of their activity. MD thesis, Water Pollution Control Group of Municipal Engineering, Northern JiaoTong University, China. Hu, C.Z., Liu, H.J., Qu, J.H., Wang, D.S., Ru, J., 2006. Coagulation behavior of aluminum salts in eutrophic water: significance of Al13 species and pH control. Environmental Science and Technology 40, 325e331. Ille´s, E., Tomba´cz, E., 2006. The effect of humic acid adsorption on pH-dependent surface charging and aggregation of magnetite nanoparticles. Journal of Colloid and Interface Science 295, 115e123. Jiang, C., Wang, R., Ma, W., 2010. The effect of magnetic nanoparticles on Microcystis aeruginosa removal by a composite coagulant. Colloids and Surfaces A: Physicochemical and Engineering Aspects 369, 260e267. Johansson, G., 1960. On the crystal structures of some basic aluminum salts. Acta Chemica Scandinavica 14, 771e773. Korbutowicz, M.K., Nowak, K.M., Winnicki, T., 2008. Water treatment using MIEXDOC/ultrafiltration process. Desalination 221, 338e344. Li, L., Pan, M.H., Brown, R.C., Leeuwen, J.V., 2006. Synthesis, properties, and environmental applications of nanoscale ironbased materials: a review. Critical Reviews in Environmental Science and Technology 36, 405e431. Li, Y.R., Wang, J., Zhao, Y., Luan, Z.K., 2010. Research on magnetic seeding flocculation for arsenic removal by superconducting magnetic separation. Separation and Purification Technology 73, 264e270.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 2 7 e1 3 5

Liu, D., Cai, D.Q., Wang, X.Q., Yu, Z.L., 2006. Removal of algal blooms in freshwater by magnetic polymer. China Environmental Science 26 (6), 677e679. Liu, Z.L., Wang, H.B., Lu, Q.H., Du, G.H., Peng, L., Du, Y.Q., Zhang, S.M., Yao, K.L., 2004. Synthesis and characterization of ultrafine well-dispersed magnetic nanoparticles. Journal of Magnetism and Magnetic Materials 283, 258e262. Lu, Z.Y., Dai, J., Song, X.N., Wang, G., Yang, W.S., 2008. Facile synthesis of Fe3O4/SiO2 composite nanoparticles from primary silica particles. Colloids and Surfaces A: Physicochemical and Engineering Aspects 317, 450e456. Ngomsik, A.F., Bee, A., Draye, M., Cote, G., Cabuil, V., 2005. Magnetic nano- and microparticles for metal removal and environmental applications: a review. Comptes Rendus Chimie 8, 963e970. Odegaard, H., Fettig, J., Ratnaweera, H.C., 1990. Coagulation with prepolymerized metal salts. In: Hahn, H.H., Klute, R. (Eds.), Chemical Water and Wastewater Treatment. Springer-Verlag, New York, pp. 189e220. Savage, N., Diallo, M.S., 2005. Nanomaterials and water purification: opportunities and challenges. Journal of Nanoparticle Research 7, 331e342. Shen, Y.F., Tang, J., Nie, Z.H., Wang, Y.D., Ren, Y., Zuo, L., 2009. Preparation and application of magnetic Fe3O4 nanoparticles for wastewater purification. Separation and Purification Technology 68, 312e319. Singer, P.C., Bilyk, K., 2002. Enhanced coagulation using a magnetic ion exchange resin. Water Research 36, 4009e4022.

135

Svoboda, J., Fujita, T., 2003. Recent developments in magnetic methods of material separation. Minerals Engineering 16, 785e792. Tang, H.X. (Ed.), 2006. Inorganic Polymer Flocculation Theory and Flocculants, first ed. China Architecture and Building Press, China, pp. 11e12. Van Benschoten, J.E., Edzwald, J.K., 1990. Chemical aspects of coagulation using aluminum salts - II. Coagulation of fulvic acid using alum and polyaluminum chloride. Water Research 24, 1527e1536. Wang, Y., Gao, B.Y., Xu, X.M., Xu, W.Y., Xu, G.Y., 2009. Characterization of floc size, strength and structure in various aluminum coagulants treatment. Journal of Colloid and Interface Science 332, 354e359. Xu, Y., Wang, D.S., Liu, H., Lu, Y.Q., Tang, H.X., 2003. Optimization of the separation and purification of Al13. Colloids and Surfaces A: Physicochemical and Engineering Aspects 231, 1e9. Yukselen, M.A., Gregory, J., 2002. Breakage and re-formation of alum flocs. Environmental Engineering Science 19 (4), 229e236. Yukselen, M.A., Gregory, J., 2004. The reversibility of floc breakage. International Journal of Mineral Processing 73, 251e259. Zhang, R., Vigneswaran, S., Ngo, H., Nguyen, H., 2007. A submerged membrane hybrid system coupled with magnetic ion exchange (MIEX) and flocculation in wastewater treatment. Desalination 216, 325e333. Zhu, Z., Li, T., Lu, J.J., Wang, D.S., Yao, C.H., 2009. Characterization of kaolin flocs formed by polyacrylamide as flocculation aids. International Journal of Mineral Processing 91, 94e99.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 3 6 e1 4 4

Available online at www.sciencedirect.com

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

Integration of anammox into the aerobic granular sludge process for main stream wastewater treatment at ambient temperatures M.-K.H. Winkler, R. Kleerebezem, M.C.M. van Loosdrecht* Delft University of Technology, Department of Biotechnology, Julianalaan 67, 2628 BC Delft, The Netherlands

article info

abstract

Article history:

Anaerobic ammonium oxidation, nitrification and removal of COD was studied at ambient

Received 25 May 2011

temperature (18
C  3) in an anoxic/aerobic granular sludge reactor during 390 days. The

Received in revised form

reactor was operated in a sequencing fed batch mode and was fed with acetate and

10 October 2011

ammonium containing medium with a COD/N ratio of 0.5 [g COD/gN]. During influent

Accepted 17 October 2011

addition, the medium was mixed with recycled effluent which contained nitrate in order to

Available online 6 November 2011

allow acetate oxidation and nitrate reduction by anammox bacteria. In the remainder of the operational cycle the reactor was aerated and controlled at a dissolved oxygen

Keywords:

concentration of 1.5 mg O2/l in order to establish simultaneous nitritation and Anammox.

Anammox

Fluorescent in-situ hybridization (FISH) revealed that the dominant Anammox bacterial

Granular sludge

population shifted toward Candidatus “Brocadia fulgida” which is known to be capable of

AOB

organotrophic nitrate reduction. The reactor achieved stable volumetric removal rates of

Acetate

900 [g N2eN/m3/day] and 600 [g COD/m3/day]. During the total experimental period

Ambient temperature

Anammox bacteria remained dominant and the sludge production was 5 fold lower than what was expected by heterotrophic growth suggesting that consumed acetate was not used by heterotrophs. These observations show that Anammox bacteria can effectively compete for COD at ambient temperatures and can remove effectively nitrate with a limited amount of acetate. This study indicates a potential successful route toward application of Anammox in granular sludge reactors on municipal wastewater with a limited amount of COD. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Anaerobic ammonium oxidizing bacteria (Anammox) are capable of autotrophic ammonium oxidation with nitrite as electron acceptor (Strous et al., 1999). After the discovery of Anammox by Mulder in 1985, Anammox bacteria have successfully been implemented in full scale wastewater treatment systems to treat ammonium rich wastewater cost effectively (Abma et al., 2010; Sliekers et al., 2003; van der Star

et al., 2007; Wett, 2007). Currently Anammox is applied at mesophilic temperatures and on wastewater containing high concentrations of ammonium. In order to supply Anammox with nitrite for the oxidation of ammonium two different systems are proposed. Nitrite can be either produced in a separated tank by partial nitrification and in turn be fed into an non-aerated Anammox reactor (Sharon-Anammox) or produced in an oxygen limited one stage system (CANON) (Sliekers et al., 2003; van Dongen et al., 2001). In the latter

* Corresponding author. E-mail address: [email protected] (M.C.M. van Loosdrecht). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.034

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 3 6 e1 4 4

system ammonium oxidizing bacteria (AOB) and Anammox grow together in one granule in which AOBs are located on the outer oxygen penetrated shell, where they oxidize ammonium to nitrite. Anammox can grow in the oxygen shielded inner core were ammonium and nitrite are available (Hao et al., 2005). Currently the nitritation/anammox processes are applied predominantly for treatment of sludge digester rejection water and effluent from industrial anaerobic wastewater treatment that both do not contain any or only limited amounts of organic carbon. If Anammox can be applied at lower temperatures and nitrogen concentrations, its application potential could be extended to municipal sewage treatment (Jetten et al., 1997). In order to implement Anammox in sewage treatment, pre-removal of COD is normally required. Heterotrophic growth results in a decrease of the SRT and a very high SRT is essential for successful cultivation of Anammox at ambient temperatures. Given the low yield and growth rate of Anammox bacteria, heterotrophic growth should be minimized, to maintain a high fraction of Anammox bacteria in the sludge. Nitrate produced either by Anammox or nitrite oxidizing bacteria would have to be removed by nitrate reduction processes. Pre-removal of organic carbon can be established after the A-stage in an A/B process, after physio-chemical pretreatment, or after anaerobic digestion (Jetten et al., 1997; Joss et al., 2009; Kartal et al., 2010; Wett, 2007). Simultaneous partial nitrification, anammox and denitrification has been reported for treating wastewater with an approximate COD/N ratio of 0.5 [g COD/gN] at temperatures of 30e36
C under constantly aerated conditions (Chen et al., 2009; Lan et al., 2011; Xu et al., 2010). In these studies the COD was removed by regular heterotrophic bacteria. Recently, it was reported that certain Anammox species have the capacity to oxidize volatile fatty acids with nitrate as electron acceptor, while forming ammonium with nitrite as intermediate (Gueven et al., 2005; Kartal et al., 2007a, 2007b). Anammox does not incorporate the fatty acids into biomass, but completely converts them into CO2 (Kartal et al., 2007a). Why Anammox remains growing autotrophically while oxidizing acetate is not well understood but the low biomass yield associated with autotrophic growth is beneficial for wastewater treatment since sludge production is minimized. When COD oxidation with nitrate can be catalyzed by Anammox, nitrate produced by Anammox bacteria (or nitrite oxidizing bacteria) can be removed resulting in a lower nitrogen effluent concentration. Previous research in anoxic reactors showed that heterotrophs will win the competition for nitrate if the COD/N ratio exceeds 1 (Gueven et al., 2005). If oxygen and acetate are available at the same time Anammox bacteria do not only need to compete with NOBs for nitrite but also general heterotrophs will get an advantage over Anammox since Anammox is inhibited by oxygen (Strous et al., 1997). A better strategy would be therefore to proceed a nitritation/anammox period with an anoxic COD oxidation period. This allows treating wastewater with easy degradable soluble compounds such as acetate to promote the oxidation of acetate with nitrate by Anammox bacteria. In this study we aimed to explore the possibility to convert COD (in the form of acetate) and ammonium at a COD/N ratio of 0.5 by Anammox and AOB in a granular sludge reactor at ambient temperature

137

(18
C  3). Hereto we operated a granular sludge SBR. During one fourth of the cycle time the ammonium and acetate were mixed with reactor effluent containing nitrate from the previous cycle. In this way nitrate reduction can be catalyzed anoxically by Anammox with acetate. During the remaining three-fourth of the cycle the reactor was aerated for ammonium removal by nitritation and Anammox.

2.

Material and methods

2.1.

Long term operation

A lab-scale anoxic/aerobic bubble column reactor with a total volume of 2.9 L was run for 390 days at ambient temperature (18  3
C) in sequencing fed batch mode. The reactor was inoculated with granular Anammox sludge from the Rotterdam Dokhaven Anammox reactor. The reactor was operated in two phases (Fig. 3). In phase I (day 0e260), the reactor was fed with medium containg ammonium and nitrite (115 mg NO2eN/l (NaNO2), 190 mg NH4eN/l (NH4Cl)) and low concentrations of COD (25e90 mg COD/l (C2H3OONa)) to establish

Fig. 1 e Shows the cycle operation in experimental set-up of 1) mixed anaerobic fed batch period with nitrate from aerobic period and ammonium and COD in the influent (60 min) 2) aerobic period (172 min) 3) settling period (3 min) 4) discharge period (5 min).

138

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 3 6 e1 4 4

Fig. 2 e Nitrogen compounds and COD (in form of acetate) during one cycle of operation in A) phase I and B) phase II. Graph displays nitrate (-), nitrite (C) ammonium (:) total nitrogen (3) and acetate (A) profiles over time. Note that opposed to normal cycle operation all media was fed at once into reactor in order to follow trends over time during anoxic period.

a stable system for combined nitritation and Anammox and organic carbon removal. During phase I the COD/N ratio was kept at 0.1 until day 170 after which it was raised gradually by decreasing nitrite and increasing COD in the influent until a ratio of 0.4 (day 260). In phase II the medium consisted of a nitrite free and acetate rich medium (100 mg COD/l) reaching a COD/N ratio of 0.5 (phase II; day 260e390). The mineral medium consisted of 0.2 mM MgSO4∙7H2O, 0.2 mM KCl, 2 mM NaHCO3, 0.2 mM K2HPO4, 0.1 mM KH2PO4. ‘Visniac and Santer’ solution was used to provide trace elements (Visniak and Santer, 1957). The pH was maintained at 7.2  0.2 during the aerobic period by dosage of hydrochloric acid and sodium hydroxide. The dissolved oxygen (DO) concentration was controlled at 1.5 mg O2/l. The DO was set by recirculating the off-gas and blending with fresh air. In this way the DO could be regulated while maintaing a constant superficial air velocity of 2 l/min (Mosquera-Corral et al., 2005). Samples were taken on a weekly basis and analyzed for N-compounds

by the use of standard test kits (Hach-Lange). Sludge bed height remained constant over time (1 cm). Biomass production was monitored over time by catching effluent from one cycle and determining dry weight and ash content.

2.2.

Cycle operation

The reactor was operated in a sequencing fed batch mode and the different periods are displayed in Fig. 1. During the mixed anoxic feeding period (60 min) nitrate produced during the previous cycle was mixed with medium containing acetate and ammonium. After the feeding period an aerated period for partial nitrification was introduced lasting 172 min, followed by a settling period (3 min), and an effluent withdrawal period (5 min). During the effluent removal period half of the reactor liquid volume (1.5 l) was discharged and half (1.4 l) remained in the system.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 3 6 e1 4 4

139

Fig. 3 e A) Volumetric conversion rates of ammonium (A) and nitrite (:) as well as production of nitrate (3) B) biomass production rate (C) and the COD/N ratio (-) over time.

The reactor performance during one cycle of operation was analyzed for phase I (nitrite in the feed) and phase II (no nitrite in the feed) (Fig. 2). Samples were taken every 10e20 min to measure ammonium, nitrite and nitrate by means of flow injection analysis (Quick Chem8500, Lachat instruments). Acetate was measured by using a High Performance Liquid Chromatography (HPLC).

2.3.

Microscopic characterization of granules

Granules were taken for microscopic analysis in order to assess their morphology and microbiological composition. Slicing and FISH was accomplished by the method proposed by (Winkler et al., 2011b) (Fig. 4) to see the spatial distribution of bacteria as a function of depth within the granule. FISH was performed for determination of general Anammox bacteria (Cy3) general bacteria Eub (Cy5) and Candidatus “Brocadia

fulgida” (Fluos). For AOB a mix of two probes was prepared and labeled with (Cy5). Probe sequences are listed in Table 1.

2.4.

Biomass yields

The estimated community composition was based on produced biomass per consumed acetate (Yx/HAC) and ammonium ðYX=NHþ4 Þ, respectively (Table 2). It was assumed that all acetate which was fed into the reactor (100 mg COD/l) would be metabolized by either Anammox or heterotrophic bacteria. In case of Anammox being the only active bacteria half of the consumed ammonium was supposed to be used for partial nitrification (AOB) and the other half by anaerobic ammonium oxidation. Since Anammox does not incorporate acetate into biomass no growth on acetate for Anammox was assumed. For the conversion from COD to VSS a factor of 1.4 was used (Scherer et al., 1983).

140

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 3 6 e1 4 4

Fig. 4 e Microbial images from phase I (a,c) and II (b,d) a) FISH on sliced granules with general anammox bacteria (red), Candidatus “Brocadia fulgida” (green) and general AOB (blue) b) FISH on sliced granules with general anammox bacteria (red), Candidatus “Brocadia fulgida” (green) and general Eub (blue) c, d) Light microscopic images of granules in phase I (c) and II (d). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

3.

Results

3.1.

Long term operation

The reactor was operated in two phases over a time period of 390 days. The volumetric conversion rates are displayed in Fig. 3 and are based on the difference between the total soluble nitrogen compounds in the influent and effluent of the reactor. In phase I when the reactor influent contained nitrite (115 mg NO2eN/l) and small amounts of acetate (25 mg COD/l)

the nitrogen removal rate reached 1200 [g N2eN/m3/day] and gradually decreased upon decreasing the nitrite and increasing the COD concentration in the medium (Fig. 3A). Phase I started with a low COD/N ratio of 0.1 and reached a ratio of 0.5 in phase II. In phase II no nitrite was supplied and the influent COD concentration was kept constant at 100 mg COD/l reaching an average volumetric N conversion rate of 900 [g N2eN/m3/day] as well as a COD removal rate of 600 [g COD/m3/day] (Fig. 3A). The measured nitrate in the effluent during phase I was 71  16 [mg NO3/m3/day] (Fig. 3A day phase I (day0e260)), and decreased when acetate was increased

Table 1 e Oligonucleotide probes and primers target microorganisms, and references used in this study. Probes Amx 368 Bfu613 EUB 338 EUB 338 III NSO190 NSO1225

Sequence (from ‘5 to ‘3)

Specificity

Reference

CCTTTCGGGCATTGCGAA GGATGCCGTTCTTCCGTTAAGCGG GCTGCCTCCCGTAGGAGT GCTGCCACCCGTAGGTGT CGATCCCCTGCTTTTCTCC CGCCATTGTATTACGTGTGA

All Anammox bacteria Candidatus Brocadia fulgida Most bacteria Verrucomicrobiales All AOB All AOB

(Schmid et al., 2003) (Kartal et al., 2008) (Amann et al., 1990) (Daims et al., 1999) (Mobarry et al., 1997) (Mobarry et al., 1997)

Probes for Anammox fulgida were tagged with the fluorescent dye Fluos (green) general Anammox with Cy3 (red) and Eubs as well as AOBs with Cy5 (blue). For analysis probes of one target group were mixed.

141

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 3 6 e1 4 4

Table 2 e Biomass yields for anammox, AOB and heterotrophic bacteria and the corresponding biomass concentration and relative community composition according to consumed substrate (100 mg COD/l; 150 mg NH4eN/l) in reactor during one day of operation in phase II.

Anammox AOB Heterotrophs

YCx/Hac

YCx=NHþ4

g VSS/m3/day

e e 0.4

0.07 0.13 e

30 63 336

Community composition [%] 32a 68a e

7b 15b 78b

Reference for yields (van der Star, 2008) (Blackburne et al., 2007) (Beun et al., 2001)

a Calculations based on assumption that all acetate was converted by anammox. b Calculations based on assumption that all acetate was converted by heterotrophic bacteria.

down to 40  5 [mg NO3/m3/day] (Fig. 3A day phase II (day 260e380)). The biomass production in phase I was 61  14 g VSS/m3/day when a COD/N ratio of 0.1 was applied and 126  49 g VSS/m3/day when the COD/N ratio was elevated until 0.5. During phase II when COD/N ratio was kept constant at 0.5 the biomass production decreased from initial values of approximately 150 g VSS/m3/day to values around 55 g VSS/ m3/day.

3.2.

Reactor performance during one cycle of operation

A cycle measurement was conducted and changes in nitrogen concentrations during one cycle of operation during phase I (influent COD/N ratio 0.1) and phase II (influent COD/N ratio 0.5) are depicted in Fig. 2. Note that for a cycle measurement influent was fed at once into the reactor at t ¼ 0 min to be able to measure concentration profiles over time. During normal operation the mixed feeding period lasted 60 min. In phase I, when nitrite was present during anoxic feeding (0e60 min) nitrite and ammonium decreased while nitrate was formed as is can be expected to common Anammox stoichiometry. During the aerobic period (60e232 min) ammonium was oxidized by AOB and Anammox. Nitrate formation was due to Anammox activity as well the oxidation of nitrite by a small proportion of NOB, this gave a nitrate accumulation of 42 mg NO3eN/l nitrate at the end of the cycle in phase I. In the phase II, acetate removal occurred simultaneously with nitrate removal until complete depletion of electron acceptor and donor (0e60 min) (Fig. 2B). During the aeration period (60e232 min) reactor performance in phase II (Fig. 2B) was similar to the aerated period observed in phase I (Fig. 2A) with the difference that nitrate accumulation was lowered from average values of 42 (phase I) to average values 20 mg NO3eN/l (phase II). Ammonium measurements in the liquid are not completely indicative for the conversions since significant ammonium adsorption occurred (Bassin et al., 2011). Upon feeding ammonium is adsorbed to the granular sludge, while during conversion it gradually desorbs again (maintaining an adsorption equilibrium) during conversion.

3.3.

Microscopic analyses of granular sludge samples

During the reactor operation granules of red granulated biomass originating from the full scale Anammox reactor system in Rotterdam (Fig. 4c) developed over time into lightreddish granules (Fig. 4d) indicating a change in community composition. During phase I the Eubacterial microbial population mainly consisted of AOB and Anammox (data not

shown) whereas no Candidatus “Brocadia fulgida” was detected (Fig. 4a). Candidatus “Brocadia fulgida” (Fig. 4b) accumulated in the sludge once the nitrite was omitted from the feed during phase II, this organism was neither detected in the seed sludge nor during phase I (Fig. 4a). During phase I typical nitrite-anammox architecture of granular sludge was observed (Vlaeminck et al., 2010), with nitrifiers (here EUB) forming an outside coating of the Anammox granule. During phase II this structure essentially remained but now also a small fraction of heterotrophs got intermixed with the Anammox population (Fig. 4b).

3.4.

Biomass yields

In order to investigate the conversion of acetate in the reactor we evaluated the biomass production and microbial community composition on calculated biomass production based on two conversion routes. If only Anammox bacteria and AOB would dominate the system and all acetate and ammonium would be consumed by Anammox bacteria and AOB a sludge productivity of approximately 93 mg VSS/m3/day is expected and hence a system strongly dominated by Anammox bacteria (32%) and AOB (68%). If all acetate was assumed to be oxidized by heterotrophs and all ammonium assumed to be converted by partial nitrification in combination with anaerobic ammonium oxidation the sludge production was calculated to be 429 g VSS/m3/day. If acetate was fully converted by heterotrophs it can be expected that the community was dominated by heterotrophic bacteria (78%), with only 7% of Anammox bacteria and 15% AOB in the community (Table 2).

4.

Discussion

This research showed a possible new application of Anammox for wastewater treatment containing COD and ammonium with a COD/N ratio up to 0.5. It was shown that despite feeding the reactor with acetate, Anammox activity could be maintained and acetate oxidation was combined with anaerobic ammonium oxidation while keeping sludge production low. Removal rates were similar to those reported in system based on one-reactor nitration Anammox processes (Abma et al., 2010; Siegrist et al., 2008; Wett, 2007). Earlier research has shown that Anammox bacteria are capable of using organic acids like propionate and acetate (Kartal et al., 2007a, 2008). These authors demonstrated that Anammox bacteria did not incorporate acetate in biomass leading to low sludge production. Their studies were conducted in flocculent sludge

142

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 3 6 e1 4 4

reactors at a higher temperature and with constant feeding under non-aerated conditions. Our study here showed that Anammox bacteria could outcompete normal heterotrophic denitrifying bacteria for acetate at ambient temperatures. We moreover combined the operation of a SBR similar to aerobic granular sludge systems (Winkler et al., 2011a) with the capability of Anammox to use acetate as a second electron donor for nitrate and nitrite reduction besides ammonium. Excess nitrate from the aerated period was used as electron acceptor to oxidize acetate present in the influent leading to significantly lower nitrate in the effluent in phase II compared to phase I (Figs. 2 and 3). FISH pictures on sliced granules showed a higher amount of Anammox bacteria as expected from calculations assuming that all acetate is consumed by normal heterotrophic bacteria (Table 2). According to this calculation the microbial population would consist of 78% heterotrophs and only a minor fraction of Anammox bacteria (7%) with a total expected biomass production of 429 g VSS/m3/day (Table 2). This is 5 fold higher than what is expected for the pure autotrophic Anammox based acetate oxidation (93 g VSS/m3/day). The measured biomass production remained low during the last 100 days of phase 2 with values of 65  14 g VSS/m3/day which is close to calculated value of 95 g VSS/m3/day which assumed acetate oxidation by Anammox bacteria only (Table 2). FISH images showed a significantly higher fraction of Anammox bacteria than what could be expected if heterotrophic denitrification would have outcompeted Anammox bacteria for actetate (Table 2, Fig. 4b). In addition, the color of the biomass remained red after raising the COD concentration to 600 g COD/m3/day in phase II (Fig. 4cd). A clear shift of the Anammox population toward Candidatus “Brocadia fulgida”, known for its capability to use acetate (Kartal et al., 2008), was detected by FISH (Fig. 4a, b). Earlier studies on Anammox bacteria using granular sludge have shown an increased growth of nitrifiers in smaller granules. This is due to the fact that smaller granules (or flocks) have a larger aerobic volume fraction than larger granules thus favoring the growth of aerobic bacteria (Volcke et al., 2010; Winkler et al., 2011b). Moreover, slow growing organisms (here Anammox bacteria) are expected to grow in the inner part of a biofilm whereas faster growing organisms (here AOB or general heterotrophs) are pushed toward the rim and in turn out of the biofilm (Picireanu et al., 2004). In the granular sludge process flocks and small granules are easier washed-out. This gives a smaller SRT for the population that dominates these smaller granules and flocks (mainly the aerobic organisms and not Anammox bacteria) which is similar to studies as conducted in a nitrifying e denitrifying biofilm airlift suspension reactor (Van Benthum et al., 1997). Visual observation of microbial community composition within a granule confirmed that Anammox bacteria are located in the middle of the granule and a smaller fraction of AOB and heterotrophs are located on the outer shell of the granule (Fig. 4a, b). Therefore also due to erosion the latter organisms will have a shorter retention time. This will lead to a higher SRT for granules and hence to an enrichment of Anammox bacteria in the sludge. Measurements on removal capacities showed a total volumetric nitrogen removal of 900 [g N2eN/m3/day] when the COD/N ratio was 0.5 [mg/mg].

Nitrate reduction to ammonium coupled to acetate oxidation proceeds theoretically in a 4.6 mg COD/mg N-ratio which is below the measured ratio (7  0.7) in the system indicating that also a fraction of COD was converted by traditional heterotrophic COD oxidation. Likely the combination of COD removal by anammox bacteria and a selective washout of heterotrophic bacteria led to the high accumulation of anammox bacteria in the studied system. Previous studies in which Anammox bacteria were exposed to organic acids reported successful operation at low COD/N ratios around 0.5 (Chen et al., 2009; Lan et al., 2011; Xu et al., 2010), an increase in heterotrophic growth at COD/N ratio around 1 (Udert et al., 2008) and a loss in Anammox activity at a constant COD/N ratio above 1 (Gueven et al., 2005). A wastewater stream containing high loads of COD seems to be unsuitable for Anammox bacteria although this likely depends on the actual ammonium load and hence the COD/N ratio. Here we showed an option to increase removal capacity of a one stage nitration Anammox process at ambient temperatures by reducing the excess nitrate from the aeration period by mixing it with the influent containing acetate. Herewith the growth of heterotrophic bacteria can be minimized because excess nitrate can be reduced under nonaerated conditions via the organotrophic pathway of Anammox bacteria. In a continuously aerated reactor heterotrophs would have the availability of a strong electron acceptor and donor (oxygen and acetate), while Anammox bacteria are inhibited by the oxygen (Strous et al., 1997), which would hence give heterotrophs an extra advantage over Anammox bacteria to oxidize acetate. Current research suggests the usage of Anammox based treatment systems after either an A/B process, after physiochemical pretreatment, or anaerobic digestion (Jetten et al., 1997; Joss et al., 2009; Kartal et al., 2010; Wett, 2007). If after an A-stage or a pretreatment some soluble COD is left and ammonium levels are high then the here presented treatment strategy could be applied. This forms an option to use Anammox in the main stream while obtaining a stable nitrogen removal process at ambient temperatures and low COD/N ratios. The approach chosen here is close to the operational conditions of an aerobic granular sludge system (De Bruin et al., 2004) allowing potential future combination of this novel high rate technology and Anammox processes. The advantage of Anammox bacteria being able to remove COD as well as nitrate makes the implementation of the Anammox technology indeed easier. Nitrate is produced in the regular Anammox conversion due to the coupling of CO2 reduction for biomass synthesis to oxidation of nitrite to nitrate. Moreover at lower temperatures it might be more difficult to prevent nitrite oxidizing bacteria to grow in the system. This could lead to too high nitrate concentrations in the effluent. Integrating anoxic periods in the sequencing fed batch cycles of aerobic granular sludge reactors would indeed give the option for nitrate removal by organotrophic Anammox bacteria. In the current perspective it is not easy to speculate on potential effluent nitrate levels within Anammox systems but we believe it will not be problematic to reach similar levels of around 60e75% of total nitrogen removal in a municipal treatment plant. The organic composition of wastewater does not only contain acetate and despite the fact that Anammox is

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 3 6 e1 4 4

reported to use other C sources such as propionate (Gueven et al., 2005; Kartal et al., 2007b) it remains unclear how competitive Anammox can be in an environment containing a variety of different carbon sources and how sensitive Anammox bacteria are to fluctuations in COD/N ratios.

5.

Conclusions

Here we used the acetate oxidizing capacity of Anammox bacteria at ambient temperature conditions and low COD/N ratios in a nitritation/anammox granular sludge system. Nitrate produced by Anammox bacteria or due to the presence of a fraction of nitrite oxidizing bacteria was reduced significantly. Sludge production remained low suggesting that Anammox successfully competed with general heterotrophs for acetate. This is a first step toward the application of Anammox bacteria in the main stream of municipal wastewater treatment processes.

Acknowledgements This study is partly funded by DHV and STOWA in the framework of the Dutch national Nereda research programme.

references

Abma, W.R., Driessen, W., Haarhuis, R., van Loosdrecht, M.C.M., 2010. Upgrading of sewage treatment plant by sustainable and cost-effective separate treatment of industrial wastewater. Water Science and Technology 61, 1715e1722. Amann, R.I., Krumholz, L., Stahl, D.A., 1990. Fluorescentoligonucleotide probing of whole cells for determinative phylogenetic, and environmental-studies in microbiology. Journal of Bacteriology 172, 762e770. Bassin, J.P., Pronk, M., Kraan, R., Kleerebezem, R., v. L., M.C.M., 2011. Ammonium adsorption in aerobic granular sludge, activated sludge and anammox granules. Water Research 45, 5257e5265. Beun, J.J., Heijnen, J.J., van Loosdrecht, M.C.M., 2001. N-removal in a granular sludge sequencing batch airlift reactor. Biotechnology and Bioengineering 75, 82e92. Blackburne, R., Vadivelu, V.M., Yuan, Z.G., Keller, J., 2007. Determination of growth rate and yield of nitrifying bacteria by measuring carbon dioxide uptake rate. Water Environment Research 79, 2437e2445. Chen, H.H., Liu, S.T., Yang, F.L., Xue, Y., Wang, T., 2009. The development of simultaneous partial nitrification, ANAMMOX and denitrification (SNAD) process in a single reactor for nitrogen removal. Bioresource Technology 100, 1548e1554. Daims, H., Bruhl, A., Amann, R., Schleifer, K.H., Wagner, M., 1999. The domain-specific probe EUB338 is insufficient for the detection of all bacteria: development and evaluation of a more comprehensive probe set. Systematic and Applied Microbiology 22, 434e444. De Bruin, L.M.M., De Kreuk, M.K., van der Roest, H.F.R., Van Loosdrecht, M.C.M., Uijterlinde, C., 2004. Aerobic granular sludge technology, alternative for activated sludge technology? Water Science and Technology 49, 1e9.

143

Gueven, D., Dapena, A., Kartal, B., Schmid, M.C., Maas, B., Van de Pas-Schoonen, K., Sozen, S., Mendez, R., Op den Camp, H.J.M., Jetten, M.S.M., Strous, M., Schmidt, I., 2005. Propionate oxidation by and methanol inhibition of anaerobic ammonium-oxidizing bacteria. Applied and Environmental Microbiology 71, 1066e1071. Hao, X.D., Cao, X.Q., Picioreanu, C., van Loosdrecht, M.C.M., 2005. Model-based evaluation of oxygen consumption in a partial nitrification-Anammox biofilm process. Water Science and Technology 52, 155e160. Jetten, M.S.M., Horn, S.J., v. Loosdrecht, M.M., 1997. Towards a more sustainable municipal wastewater treatment system. Water Science and Technology 35, 171e180. Joss, A., Salzberger, D., Eugster, J., Knig, 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. Kartal, B., Kuenen, J.G., van Loosdrecht, M.C.M., 2010. Sewage treatment with anammox. Science 328, 702e703. Kartal, B., Kuypers, M.M.M., Lavik, G., Schalk, J., Op den Camp, H.J.M., Jetten, M.S.M., Strous, M., 2007a. Anammox bacteria disguised as denitrifiers: nitrate reduction to dinitrogen gas via nitrite and ammonium. Environmental Microbiology 9, 635e642. 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., 2007b. Candidatus "Anammoxoglobus propionicus" a new propionate oxidizing species of anaerobic ammonium oxidizing bacteria. Systematic and Applied Microbiology 30, 39e49. Kartal, B., van Niftrik, L., Rattray, J., de Vossenberg, J., Schmid, M.C., Damste, J.S.S., Jetten, M.S.M., Strous, M., 2008. Candidatus ‘Brocadia fulgida’: an autofluorescent anaerobic ammonium oxidizing bacterium. Fems Microbiology Ecology 63, 46e55. Lan, C.J., Kumar, M., Wang, C.C., Lin, J.G., 2011. Development of simultaneous partial nitrification, anammox and denitrification (SNAD) process in a sequential batch reactor. Bioresource Technology 102, 5514e5519. Mobarry, B.K., Wagner, M., Urbain, V., Rittmann, B.E., Stahl, D.A., 1997. Phylogenetic probes for analyzing abundance and spatial organization of nitrifying bacteria (vol 62, pg 2157, 1996). Applied and Environmental Microbiology 63, 815. Mosquera-Corral, A., de Kreuk, M.K., Heijnen, J.J., van Loosdrecht, M.C.M., 2005. Effects of oxygen concentration on N-removal in an aerobic granular sludge reactor. Water Research 39, 2676e2686. Picioreanu, C., Kreft, J.U., van Loosdrecht, M.C.M., 2004. Particlebased multidimensional multispecies Biofilm model. Applied and Environmental Microbiology 70, 3024e3040. Scherer, P., Lippert, H., Wolff, G., 1983. Composition of the major elements and trace elements of 10 methanogenic bacteria determined by inductively coupled plasma emission spectrometry. Biological Trace Element Research 5, 149e163. 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., Damste, J.S.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. Siegrist, H., Salzgeber, D., Eugster, J., Joss, A., 2008. Anammox brings WWTP closer to energy autarky due to increased biogas production and reduced aeration energy for N-removal. Water Science and Technology 57, 383e388.

144

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 3 6 e1 4 4

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 Microbiology Letters 218, 339e344. Strous, M., van Gerven, E., Kuenen, J.G., Jetten, M., 1997. Effects of aerobic and microaerobic conditions on anaerobic ammonium-oxidizing (anammox) sludge. Applied and Environmental Microbiology 63, 2446e2448. Strous, M., Kuenen, J.G., Jetten, M.S.M., 1999. Key physiology of anaerobic ammonium oxidation. Applied and Environmental Microbiology 65, 3248e3250. 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, 277e284. Van Benthum, W.A.J., Van Loosdrecht, M.C.M., Heijnen, J.J., 1997. Process design for nitrogen removal using nitrifying biofilm and denitrifying suspended growth in a biofilm airlift suspension reactor. Water Science and Technology 36, 119e128. van der Star, W., 2008. Growth and Metabolism of Anammox Bacteria: PhD thesis, Technical Univertity Delft. 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, 4149e4163. van Dongen, U., Jetten, M.S.M., Van Loosdrecht, M.C.M., 2001. The SHARON-Anammox process for treatment of ammonium rich wastewater. Water Science and Technology 44, 153e160.

Visniak, W., Santer, M., 1957. The thiobacilli. Bacteriological Reviews 26, 168e175. Vlaeminck, S.E., Terada, A., Smets, B.F., De Clippeleir, H., Schaubroeck, T., Bolca, S., Demeestere, L., Mast, J., Boon, N., Carballa, M., Verstraete, W., 2010. Aggregate size and architecture determine microbial activity balance for onestage partial nitritation and anammox. Applied and Environmental Microbiology 76, 900e909. Volcke, E.I.P., Picioreanu, C., De Baets, B., van Loosdrecht, M.C.M., 2010. Effect of granule size on autotrophic nitrogen removal in a granular sludge reactor. Environental Technology 31, 1271e1280. Wett, B., 2007. Development and implementation of a robust deammonification process. Water Science and Technology 56, 81e88. Winkler, M.K.H., Bassin, J.P., Kleerebezem, R., van Loosdrecht, M.C.M., van den Brand, T.P.H., 2011a. Selective sludge removal in a segregated aerobic granular biomass system as a strategy to control PAO-GAO competition at high temperatures. Water Research 45, 3291e3299. Winkler, M.K.H., Kleerebezem, R., Kuenen, J.G., Yang, J., van Loosdrecht, M.C.M., 2011b. Segregation of biomass in cyclic anaerobic/aerobic granular sludge allows the enrichment of anaerobic ammonium oxidizing bacteria at low temperatures. Environmental Science and Technology. Xu, Z.-Y., Zeng, G.-M., Yang, Z.-H., Xiao, Y., Cao, M., Sun, H.-S., Ji, L.-L., Chen, Y., 2010. Biological treatment of landfill leachate with the integration of partial nitrification, anaerobic ammonium oxidation and heterotrophic denitrification. Bioresource Technology 101, 79e86.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 4 5 e1 5 1

Available online at www.sciencedirect.com

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

Aerobic degradation of sulfanilic acid using activated sludge Gang Chen a,b, Ka Yu Cheng a,*, Maneesha P. Ginige a, Anna H. Kaksonen a a b

CSIRO Land and Water, CSIRO, Floreat, WA 6014, Australia College of Environmental Science and Engineering, Donghua University, Shanghai 201620, China

article info

abstract

Article history:

This paper evaluates the aerobic degradation of sulfanilic acid (SA) by an acclimatized

Received 9 June 2011

activated sludge. The sludge was enriched for over three months with SA (>500 mg/L) as

Received in revised form

the sole carbon and energy source and dissolved oxygen (DO, >5 mg/L) as the primary

15 September 2011

electron acceptor. Effects of aeration rate (0e1.74 L/min), DO concentration (0e7 mg/L) and

Accepted 18 October 2011

initial SA concentration (104e1085 mg/L) on SA biodegradation were quantified. A modified

Available online 28 October 2011

Haldane substrate inhibition model was used to obtain kinetic parameters of SA biodegradation and oxygen uptake rate (OUR). Positive linear correlations were obtained between

Keywords:

OUR and SA degradation rate (R2
0.91). Over time, the culture consumed more oxygen per

Azo dye

SA degraded, signifying a gradual improvement in SA mineralization (mass ratio of O2: SA

Sulfonated aromatic amines

at day 30, 60 and 120 were 0.44, 0.51 and 0.78, respectively). The concomitant release of

Haldane kinetics

near stoichiometric quantity of sulphate (3.2 mmol SO2 4 released from 3.3 mmol SA) and

Oxygen uptake rate

the high chemical oxygen demand (COD) removal efficacy (97.1%) indicated that the

Wastewater treatment

enriched microbial consortia could drive the overall SA oxidation close to a complete mineralization. In contrast to other pure-culture systems, the ammonium released from the SA oxidation was predominately converted into nitrate, revealing the presence of ammonium-oxidizing bacteria (AOB) in the mixed culture. No apparent inhibitory effect of SA on the nitrification was noted. This work also indicates that aerobic SA biodegradation could be monitored by real-time DO measurement. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Wastewaters generated from textile factories often contain sulfonated aromatic amines, which are primarily originated from the reductive cleavage of sulfonated azo dyes (10e50% of the applied dyes remain in the wastewater) (O’Neill et al., 1999). Sulfanilic acid (4-aminobenzenesulfonic acid, SA) is one of the most representative sulfonated aromatic amines (Perei et al., 2001). Due to environmental and health concerns, SA-contaminated wastewaters need to be treated prior to its discharge into the environment (Chung and Cerniglia, 1992; Oh et al., 1997; Topac et al., 2009). Currently, aerobic biodegradation is considered as the most effective and

environmentally benign approach to treat SA-contaminated wastewaters (Tan et al., 2005). However, the negatively charged sulfonyl group of SA molecule is known to suppress biodegradation by most heterotrophic microbial communities due to the low permeability of SA through bacterial membranes (Hwang et al., 1989). An acclimatization period is often required to enrich an efficient SA-degrading microbial community in the treatment process (Tan and Field, 2005; Tan et al., 2005). Earlier research has investigated the aerobic degradation of SA by using microbial enrichments. For example, the pioneering works by Feigel and Knackmuss (1988, 1993) have shown that a defined co-culture of Hydrogenophaga palleroni S1

* Corresponding author. Tel.: þ61 8 9333 6158; fax: þ61 8 9336211. E-mail addresses: [email protected], [email protected] (K.Y. Cheng). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.043

146

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 4 5 e1 5 1

and Agrobacterium radiobacter S2 could effectively degrade SA under aerobic conditions. Some other SA-degrading pure strains such as Pseudomonas paucimobilis (Perei et al., 2001), Agrobacterium sp. PNS-1 (Singh et al., 2004), Pannonibacter sp.W1 (Wang et al., 2009), Ralstonia sp. PBA and Hydrogenophaga sp. PBC (Gan et al., 2011) have also been reported. Arguably, these studies have only limited practical implication as they mainly focused on the use of either pure or co-cultures. To practically treat SA-contaminated wastewaters, mixed microbial cultures such as activated sludge could be of significance. Activated sludge obtained from municipal wastewater treatment plants has been demonstrated as an effective mixed culture inoculum for starting up SA-contaminated wastewater treatment processes. For instances, Tan et al. (2005) reported that SA could be aerobically degraded using activated sludge previously contaminated with a mixture of sulfonated aromatic compounds. In a separate study, efficient acclimatization of a SA-degrading mixed culture was achieved even using activated sludge that has not been previously exposed to SA (Carvalho et al., 2008). Nonetheless, the relationship between oxygen consumption and SA degradation in these aerobic mixed culture processes has not been properly defined. Since oxygen is the primary electron acceptor for these microbial processes, understanding the relationship between the kinetics of SA degradation and oxygen consumption would facilitate process optimization of aerobic SA wastewater treatment. Therefore, the objectives of this study were (i) to enrich an aerobic SA-degrading mixed culture using activated sludge that has not been previously exposed to SA as microbial inoculum, and (ii) to evaluate the kinetics of SA degradation and oxygen consumption of the enriched culture. The relationship between oxygen consumption and SA degradation was elucidated, and the feasibility of using realtime dissolved oxygen (DO) measurement as a strategy to obtain SA degradation kinetics during the treatment process was explored. To our knowledge, these issues have not been previously addressed particularly in a mixed culture environment. This work would shed light on our fundamental understanding of biological removal of sulfonated aromatic amines in wastewaters.

2.

Materials and methods

2.1.

Chemicals

Sulfanilic acid (SA) (4-aminobenzenesulfonic acid, CAS number 121-57-3) was purchased from SigmaeAldrich (Australia). Some selected physicalechemical properties of SA are: density 1.485 g/mL (25  C), water solubility in 10 g/L (20  C) and Henry’s constant 8.89  1013 m3/mol (25  C). A working stock solution of SA (5 g/L) was prepared by dissolving the SA in deionized water. All chemicals used in this study were of analytical grade.

2.2. Bacterial inoculum and SA containing synthetic wastewater Activated sludge was obtained from a domestic municipal wastewater treatment plant in Perth, Australia, and was

stored at 4  C prior to use. Unless specified otherwise, the basal medium used in this work had a composition of (mg/L): NH4Cl 125, NaHCO3 125, MgSO4$7H2O 51, CaCl2$2H2O 300, FeSO4$7H2O 6.25, and 1.25 mL L1 of trace element solution, which contained (g/L): ethylenediamine tetraacetic acid (EDTA) 15, ZnSO4$7H2O 0.43, CoCl2$6H2O 0.24, MnCl2$4H2O 0.99, CuSO4$5H2O 0.25, NaMoSO4$2H2O 0.22, NiCl2$6H2O 0.19, NaSeO4$10H2O 0.21, H3BO4 0.014, and NaWO4$2H2O 0.05 (Cheng et al., 2010). In some experiments where the effects of the background ammonium and sulphate were tested, NH4Cl and FeSO4$7H2O were omitted in the medium (i.e. Fig. 6). Defined SA concentration in the synthetic wastewater was prepared by adding a known volume of the SA stock solution to the basal medium. pH was adjusted to 7.0 by using phosphate buffer (KH2PO4 and K2HPO4, 15e30 mM).

2.3.

Reactor configuration and general operation

A 2-L glass continuously-stirred tank bioreactor was used in this study to aerobically acclimatize the SA-degrading culture. The working volume of the culture medium was 1.5 L. An adjustable aeration pump was used to supply oxygen to the suspended culture medium at an air flow rate that was varied from 0 to 1.74 L/min. The suspension liquor was continuously stirred by using an overhanging turbine impeller stirrer to maximize mass transfer. A DO sensor and process monitor (TPS Pty. Ltd., Australia) was used to measure the DO concentration in the suspension liquor. The DO data was periodically recorded into an excel spreadsheet using a LabVIEW computer program. The reactor was operated in batch mode at room temperature (25  2  C).

2.4.

Experimental procedures

2.4.1.

Reactor start-up

The enrichment process was initiated by mixing the activated sludge (10%, v/v) with the medium to obtain an initial mixed liquor suspended solids (MLSS) concentration of 2000 mg/L. The initial SA concentration in the medium was 500 mg/L during the first two weeks of operation (weeks 1 and 2). During this period, medium renewal was performed every 2 days. For medium renewal, the aeration pump and the overhanging mixer were switched off to allow a complete sludge settlement (ca. 30 min) and only the supernatant was decanted. From week 2 to 8 (i.e. two months after the initial start-up), the initial SA concentration was gradually increased to about 1000 mg/L and medium renewal was performed daily. No sludge wastage was performed during the initial two months to prevent washout of microorganisms relevant to the process. After the initial two months until the end of the acclimatization (i.e. 5 months), 75 mL of the mixed liquor was wasted daily to maintain a steady range of a MLSS of 5000e6000 mg/L (sludge age ¼ 20 days). During the entire acclimatization process, DO concentration in the reactor was maintained at over 5 mg O2/L to ensure sufficient supply of DO.

2.4.2. SA degradation of the acclimatized culture at different initial SA concentrations and different aeration rates SA degradation kinetics of the acclimatized sludge was quantified after the initial 2 months of acclimatization. SA

147

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 4 5 e1 5 1

degradation profiles were obtained with different initial SA concentrations. The relationship between SA degradation rate and initial SA concentration was established using a modified Haldane inhibition model as described by Maeda et al. (2005) (Equation (1)). rSA ¼ 

ds 1 1 mmax SX km SX ¼ ¼ mX ¼ dt Y Y KS þ S þ S2 =KI KS þ S þ S2 =KI

(1)

where rSA (mg/L h) is the SA degradation rate, ds/dt is the change in SA concentration over time, S (mg/L) is the SA concentration, m (1/h) is the specific microbial growth rate, km (1/h) is the product of the maximum specific microbial growth rate (mmax) and the inverse of the biomass yield on the substrate (Y, mg/mg), KS (mg/L) is the saturation constant, KI (mg/L) is the substrate inhibition constant, and X is the biomass concentration (MLSS, mg/L). Since in this study the constants mmax and Y were not measured, the constant km ¼ mmax/Y was determined instead (Maeda et al., 2005). The MLSS concentration was maintained at 5800 mg/L and the aeration rate was maintained at 1.74 L/min during this period of study. To evaluate the effects of different aeration rates (0e1.74 L/ min) on SA degradation rate, a series of batch experiments were conducted with an initial SA concentrations in all runs ranged from 310 to 340 mg/L, and the MLSS concentration was maintained at 5600 mg/L. Changes in SA concentration over time at different aeration rates were compared, and used to derive the specific SA oxidation rates which were then plotted against (i) aeration rate and (ii) average DO concentration in the mixed liquor.

2.4.3. Correlation between oxygen uptake rate and SA degradation rate At day 30, 60 and 120, experiments were conducted to obtain the correlation between SA degradation rate and the oxygen uptake rate (OUR) of the enriched culture. The 2-L bioreactor was operated with an initial SA concentration of 500e1000 mg/ L, MLSS of 5800 mg/L and aeration rate of 1.74 L/min. Over the course of each kinetic experiment, multiple mixed liquor subsamples (50 mL) were periodically taken from the bioreactor and immediately transferred into a separate, 50-mL vessel equipped with a DO probe for the determination of OUR using a dynamic method as described by Garcia-Ochoa et al. (2010). Once transferred in the vessel, the mixed liquor was continuously mixed with a small magnetic stirrer bar and was purged with air until DO reached >7 mg O2/L. Thereafter, the air supply was terminated and the decline in DO was recorded to calculate the OUR. The specific OUR values were obtained by normalizing the OUR value with the MLSS concentration and plotted against the corresponding specific SA degradation rate to obtain a linear correlation. The ratio of mg O2 consumed per mg SA degraded was obtained from the slope. The correlation between oxygen uptake rate (OUR) and initial SA concentration was also established by using the modified Haldane model equation. The specific OUR ðqO2 Þ, is calculated as below (Equation (2)): qO2 ¼

1 YX=O2

m þ m0 ¼

mmax k0 þ m0 ¼ þ m0 2 KS þ S þ S2 =KI YX=O2 KS þ S þ S =KI

where YX=O2 (mg/mg) is the yield of biomass on the oxygen, KS (mg/L) is the saturation constant, KI (mg/L) is the substrate inhibition constant, S (mg/L) is the SA concentration, k0 (1/h) is the product of the maximum specific growth rate of the microorganism (mmax) and the inverse of the yield of biomass on the oxygen ðYX=O2 Þ and m0 is the maintenance constant. Since in this study, the constants mmax and YX=O2 were not measured separately, the constant k0 ¼ mmax =YX=O2 was determined instead.

2.4.4. Real-time OUR determination in a continuously aerated SA-degrading culture The feasibility of using real-time DO measurement for obtaining in-situ SA-dependent OUR data was explored with the continuously aerated culture (MLSS ¼ 5800 mg/L; constant aeration rate of 0.31 L/min). Four aliquots of SA solution (final SA concentration ¼ 50e300 mg/L) were sequentially added to trigger oxygen consumption activity of the culture. Samples (2 mL) were periodically collected to establish concentration  profiles of SA, ammonium (NHþ 4 eN), nitrite (NO2 eN), nitrate  2 (NO3 eN) and sulphate (SO4 ). Real-time OUR (OURRT) was computed from the DO profile according to Equation (3) (Garcia-Ochoa et al., 2010).  OURRT ¼ KLa ðCs  CÞ 

dC dt

 (3)

where KLa is oxygen mass transfer coefficient (1/h) of the reactor under the constant aeration and stirring rates (KLa ¼ 17.6 h1); dC/dt is the rate of DO change; Cs is saturation concentration of DO (mg O2/L); C is the real-time DO concentration (mg O2/L). The amount of oxygen consumed per each SA addition was obtained by integrating the peak areas in the OURRT profile.

2.5.

Chemical analysis

Mixed liquor samples taken from the bioreactor were immediately filtered using a 0.2 mm sterile filter (0.8/0.2 mm Supor Membrane, PALL Life Sciences) to remove biomass. SA concentration of the filtrate was quantified by using a UVspectrophotometer (Cary 50, Varian, USA) at a maximal SA specific adsorption wavelength of 250 nm (determined experimentally in this study by scanning the adsorption at various wavelengths). The samples were diluted (if required) to obtain a SA concentration between 0 and 25 mg/L, which resulted in a linear standard curve (R2 > 0.999) with absorbance. MLSS were quantified according to a standard method (APHA, 1995). Chemical oxygen demand (COD) was measured using a closed reflux dichromate COD method (HACH Method þ   8000, HACH Ltd). SO2 4 , NH4 eN, NO3 eN and NO2 eN were quantified by using Ion Chromatography (ICS-3000, DIONEX).

3.

Results and discussion

3.1. SA biodegradation kinetics of the aerobically enriched activated sludge

1

(2)

After approximately two months of acclimatization, a kinetic experiment was conducted to quantify the SA degradation

148

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 4 5 e1 5 1

2004). This result suggests that the SA degradation behaviour of the acclimatized culture is similar to these pure and coculture studies.

a SA (mg L-1)

1000 800 600

3.2.

400

b

0

Specific SA degradation rate (mg/ g MLSS ·h)

200

5

5

Time (h) 20

10

4

25

30

Model Simulation Experimental Value

3 2 1 0

600 400 800 SA (mg L-1)

200

1000

Fig. 1 e Effect of initial SA concentration on SA degradation by enriched activated sludge culture. (a) SA concentration profiles at different initial SA concentrations; (b) relationship between specific SA degradation rate and the initial SA concentrations (R2 [ 0.97). Notes: all experiments were conducted with the same MLSS concentration of 5800 mg/L and aeration rate of 1.74 L/min.

capacity of the activated sludge at different initial SA concentrations (100e1100 mg/L) (Fig. 1). As expected, a longer time was required for the degradation of a higher initial SA concentration (Fig. 1a). To better quantify the relationship between the specific SA degradation rate and the initial SA concentration, a modified Haldane model was used to fit the obtained data (Equation (1)). The model fits the experimental data well with km, KS and KI determined as 1.04  106 h1, 98.7 mg/L and 3424 mg/L, respectively (R2 ¼ 0.97) (Fig. 1b). Both the KS and KI values for the mixed culture are similar to those reported for other highly acclimatized SA-degrading pure and co-cultures (Gan et al., 2011; Wang et al., 2009; Singh et al.,

a

Aeration facilitates mass transfer of oxygen from the atmosphere into the mixed liquor where the bacteria can degrade the SA with DO as electron acceptor. To account for any possible abiotic loss of SA via aeration (e.g. volatilization), an abiotic test was conducted and the result indicated that the loss was negligible (data not shown). No apparent SA removal was noted with the SA enriched culture when aeration was not provided (data not shown). This agrees with the general consensus that SA can only be biodegraded under aerobic conditions (Pereira et al., 2011; Kuhn and Suflita, 1989; RazoFlores et al., 1996). With the exposure to aerobic conditions a significant increase in SA removal was observed (Fig. 2a and b). At aeration rate of 0.13 L/min (which resulted in an average apparent DO concentration of 1.3 mg O2/L), SA degradation rate was 20.3 mg/L h. When the aeration rate was further increased to 0.31 L/min (average DO w 3 mg O2/L), the culture reached a maximal SA degradation rate of 43.5 mg/L h. A further increase in aeration rate (also resulting in a DO increase) could no longer increase the SA degradation rate. This suggests that beyond a DO of w3 mg O2/L, factor(s) other than oxygen contributes towards further increases in SA degradation rates. Hence, for a cost effective treatment of SA containing wastewaters using activated sludge, aeration should only be provided to maintain a DO of
3 mg O2/L.

3.3. Oxygen consumption by the enriched SA-degrading culture The kinetics of oxygen consumption by the enriched SAdegrading culture was evaluated to establish the relationship between specific OUR and initial SA concentration. Increase in SA concentration from 0 to 200 mg/L resulted in a linear increase in the specific OUR, and a maximum specific OUR (6 mg O2/g MLSS h) was obtained with an initial SA concentration of about 180 mg/L. Further increase in SA concentration did not increase the oxygen demand of the culture.

b

40

(mg L-1·h-1)

SA oxidation rate

50

Effect of aeration rate on SA degradation

30 20 10 0

0.5 1 1.5 Aeration rate (L min-1)

0

2 4 6 -1 DO (mg O2 L )

Fig. 2 e Effect of aeration rate on SA degradation by enriched activated sludge culture. (a) SA degradation rates at different aeration rates; (b) SA degradation rates at different DO concentrations. Notes: all experiments were conducted with the same MLSS concentration of 5800 mg/L, and SA concentrations were maintained between 310 and 340 mg/L.

149

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 4 5 e1 5 1

Specific OUR (mg O2 / g MLSS ·h)

6

4 Model Simulation Experimental Data

2

0

200

400 600 800 1000 SA (mg L-1)

Fig. 3 e The relationship between specific oxygen uptake rate (OUR) by the enriched activated sludge culture and initial SA concentrations (MLSS [ 7000 mg/L, day 130).

120) (Fig. 4). This indicates that the mixed culture had become more acclimatized for a more complete mineralization of SA. Since the microbial metabolism of SA degradation consists of both catabolism (i.e. oxidation of SA and its metabolites) and anabolism (cell growth), of the theoretical 1.29 mg O2/ mg SA the culture appeared to have spent only 60.5% for the complete SA degradation (Fig. 4, day 120). The y intercepts in Fig. 4 suggest that the background oxygen uptake rate due to both the endogenous respiration and the background nitrification by the mixed culture to be 1.31  0.36 mg O2/g MLSS h. Although the described OUR-based method has an advantage of its simplicity and could give a clear trend of an over-time improvement of the microbial activity in mineralizing SA, it only indirectly indicates the degree of mineralization of SA by the mixed culture.

3.4. Close to complete SA mineralization by the acclimatized mixed culture

3.3.1. Gradual increase in oxygen demand indicates enhanced SA mineralization The amount of oxygen consumed by the mixed microbial culture to oxidize a definite amount of SA can be obtained from the slope of a linear regression between oxygen uptake rate (OUR) and SA degradation rate (i.e. oxygen consumed (mg) per SA degraded (mg)). In principle, the maximal value of this slope is denoted by the theoretical COD value of SA (i.e. 1.29 mg O2/mg SA, without nitrification). In this study, the SA degradation rate increased over the 3 months of acclimatization (a 1.8-fold increase in the oxygen demand from day 30 to

To clearly verify the degree of SA mineralization by the mixed microbial culture, a detailed mass balance experiment was conducted taking the COD and sulphate concentrations into account over the course of SA degradation (Fig. 5). The result clearly indicated that the enriched microbial consortia could drive the overall SA oxidation close to a complete mineralization. The evidences are: 1. A high COD removal efficacy of 97.1% (note that the theoretical COD values of SA are close to the experimental COD values), indicating that the degradation intermediates of SA had also been oxidized by the mixed culture. 2. A near stoichiometric quantity of sulphate (1 mole of SO2 4 released per mole of SA oxidized) was produced in proportion to the amount of SA degraded (i.e. 3.2 mmol released from 3.3 mmol SA), corroborating with SO2 4 a similar observation as reported by Tan et al. (2005). The result also reveals that in contrast to other pureculture studies where SA degradation coincided with the

SA COD Actual COD Theoretical SO42NH4+-N 300 NO2--N NO3--N

150

Day 30

(1.29 mg O2 / mg SA) Theoretical Curve

Day 60 Day 120 100

50

y = 0.78x + 6.93 R2 = 0.98

SA / COD (mg/L)

Oxygen uptake rate (mg O2/L· h)

800

600

200

400

100

200

NH4+ / NO2- / NO3- / SO42Concentration (mg/L)

Instead, a gradual decline in the specific OUR was obtained, indicating substrate inhibition of the microbial culture by the SA (Fig. 3). The relationship was also obtained by using a modified Haldane model (Equation (2)), for which a satisfactory fit was obtained with the measured data (Fig. 3). By using a non-linear regression procedure, the values of k0 and m0 were found to be 0.00714 and 0.000846 L/h with a correlation factor of 0.91. A similar relationship between specific OUR and initial substrate concentration was reported with activated sludge that used o-cresol (a toxic aromatic organic pollutant) as the sole source of carbon and energy instead of SA (Maeda et al., 2005).

y = 0.51x + 9.91 (R2 = 0.91) y = 0.44x + 5.94 (R2 = 0.95) 0

0 25 50 75 100 125 SA degradedation rate (mg/L· h)

Fig. 4 e The relationship between oxygen consumed and SA degraded by the enriched activated sludge culture over time (MLSS [ 5800 mg/L, aeration rate [ 1.74 L/min).

2

4 Time (h)

6

0

Fig. 5 e Degradation of SA by the enriched activated sludge culture and release of sulphate, nitrite and nitrate over time (MLSS [ 5800 mg/L, aeration rate [ 1.74 L/min, day 130).

150

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 4 5 e1 5 1

Peak 4

O2 Consumed (mg)

300

200

y = 0.901x - 11.91 R² = 0.998 Peak 3

100 Peak 1 0

Peak 2 200 300 100 SA Degraded (mg)

Fig. 7 e The relationship between oxygen consumed and SA degraded by the enriched activated sludge culture obtained from the SA-spiking experiment in Fig. 6. The amounts of oxygen consumed were calculated from the oxygen consumption rate data in Fig. 6b.

Fig. 6 e (a) Effect of SA additions on dissolved oxygen (DO) concentration and concentrations of ammonium, sulphate, nitrite and nitrate in the SA limited culture (MLSS [ 5800 mg/L, aeration rate [ 0.31 L/min; arrows indicate additions of SA; Dotted line indicates saturation concentration of DO (7.7 mg O2/L)); (b) real-time oxygen uptake rate profile in the SA-degrading culture.

accumulation of both ammonium and sulphate in the medium (Wang et al., 2009; Singh et al., 2006), the ammonium released from the SA oxidation by the enriched mixed culture was predominately converted into nitrate (Fig. 5). This suggests that the culture contained an active population of ammonium-oxidizing bacteria (AOB). It is worth mentioning that as opposed to other studies where non-sulfonated aromatic amines such as aniline was found to have a much severe inhibitory effect on AOB (i.e. nitrification) even at a magnitude lower concentration (
5 mg/L, 125 times lower), the presence of SA appeared to have no inhibitory effect on the AOB as the release of nitrate was in phase with the decline in SA, indicating that SA oxidation proceeded simultaneously with nitrification (Kumar et al., 1984; Than et al., 2002). Although exploring the impact of AOB on SA degradation is beyond the scope of this paper, further studies are currently underway to address this fundamental issue.

DO in a continuously aerated culture was investigated (Fig. 6a), and the changes in OUR during the relevant period of exposure is shown in Fig. 6b. OUR profile was closely linked to the SA concentration profile (Fig. 7). In the absence of SA (0 h), close to saturation concentration of DO (7.7 mg O2/L) was recorded (Fig. 6a). Incrementally dosing the SA into the culture (as denoted by the solid arrows) resulted in an immediate increase in the OUR (Fig. 6b). The more SA added (as denoted by the increased SA concentration), the higher the OUR and the larger the OUR peak area (Peaks 1e4, Fig. 6b). Integrating these OUR peaks with time (i.e. the peak area) gave the amount of oxygen that was consumed by the culture to degrade the spiked SA (Fig. 7). It is noteworthy that although Figs. 4 and 7 report a similar linear regression curve for oxygen consumption and SA degradation (0.44e0.78 and 0.90 mg O2/mg SA, respectively), the experimental approaches taken were completely different. Overall, these results suggest that for the treatment of wastewater primarily contaminated with SA, online DO measurement might be used to continuously monitor the contaminant removal process (i.e. SA biodegradation). It should be noted that the anaerobic treatment of a complex azo dye containing wastewater would generally lead to a complex mixture of sulfonated (or non-sulfonated) aromatic amines residues, which would collectively contribute to the COD content in the final effluent (O’Neill et al., 1999). As we have only tested SA as the model sulfonated aromatic amine, the effectiveness of the proposed DO-based monitoring approach for practical treatment of anaerobically treated azo dye effluent would certainly warrant further studies.

4. 3.5. In-situ DO monitoring as a real-time indicator of SA biodegradation Online DO monitoring is widely used in activated sludge wastewater treatment processes to reveal oxygen demand and to control aeration. To explore the possibility of using DO as a real-time indicator for SA degradation using the acclimatized activated sludge, the impact of SA concentration on

Conclusions

Due to the complexities of using pure or co-cultures to treat wastewaters in general, mixed microbial consortia are favoured even to treat SA-contaminated wastewaters. The novelty of this work is that it describes for a first time that SA biodegradation could be monitored through online measurement of oxygen consumption in a mixed culture system (here activated sludge).

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 4 5 e1 5 1

Based on the results, we conclude the following:  An activated sludge that has not been previously exposed to SA could be acclimatised under aerobic condition to oxidize SA close to a complete mineralization.  The enriched activated sludge exhibited kinetics of both SA degradation and oxygen consumption that could be described by Haldane substrate inhibition model.  In contrast to other pure or co-culture, the ammonium released from SA was almost completely converted into nitrate, indicating a co-enrichment of ammonium-oxidizing bacteria (AOB).  SA had no apparent inhibitory effect on the AOB activity (nitrification). To what extent the AOB would affect the SA degradation is beyond the scope of this study and hence warrants further research.  Oxygen consumption was directly proportional to SA degradation, and hence the continuous DO measurement enabled a real-time monitoring of SA degradation by the mixed culture.

Acknowledgements This work was funded by the CSIRO Water for a Healthy Country Flagship. CG was supported by a scholarship from the China Scholarship Council. We thank Dr. Trevor Bastow and Ms. Yasuko Geste (CSIRO Land and Water) for assistance in the ion chromatography. We are also grateful to Drs. Bradley Patterson and Carlos Descourvie´res (CSIRO Land and Water) for their valuable comments on this work.

references

APHA (American Public Health Association), 1995. In: Eaton, A., Clesceri, L., Greenberg, A. (Eds.), Standard Methods for the Examination of Water and Wastewater, 19th ed., pp. 5-15e517. Washington, DC. Carvalho, M.C., Pereira, C., Goncalves, 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. Cheng, K.Y., Ho, G., Cord-Ruwisch, R., 2010. Anodophilic biofilm catalyses cathodic oxygen reduction. Environ. Sci. Technol. 44, 518e525. Chung, K.T., Cerniglia, C.E., 1992. Mutagenicity of azo dyes: structureeactivity relationships. Mutat. Res. 277, 201e220. Feigel, B.J., Knackmuss, H.J., 1988. Bacterial catabolism of sulfanilic acid via catecho-4-sulfonic. FEMS Microbiol. Lett. 55, 113e118. Feigel, B.J., Knackmuss, H.J., 1993. Syntrophic interactions during degradation of 4-aminobenzenesulfonic acid by a two species bacterial culture. Arch. Microbiol. 159, 124e130.

151

Gan, H.M., Shahir, S., Ibrahim, Z., Yahya, A., 2011. Biodegradation of 4-aminobenzenesulfonate by Ralstonia sp. PBA and Hydrogenophaga sp. PBC isolated from textile wastewater treatment plant. Chemosphere 82 (4), 507e513. Garcia-Ochoa, F., Gomez, E., Santos, V.E., Merchuk, J.C., 2010. Oxygen uptake rate in microbial process: an overview. Biochem. Eng. J. 49, 289e307. Hwang, S.Y., Berges, D.A., Taggart, J.J., Gilvarg, C., 1989. Portage transport of sulfanilamide and sulfanilic acid. J. Med. Chem. 32, 694e698. Kuhn, E.P., Suflita, J.M., 1989. Anaerobic biodegradation of nitrogen-substituted and sulfonated benzene aquifer contaminants. Hazard. Waste Hazard. Mater. 6, 121e134. Kumar, N.J., Krishnamoorthi, K.P., Swaminthan, T., 1984. Studies on nitrification of aniline with acclimated activated sludge. Biotechnol. Bioeng. 26, 197e202. Maeda, M., Itoh, A., Kawase, Y., 2005. Kinetics for aerobic biological treatment of o-cresol containing wastewaters in a slurry bioreactor: biodegradation by utilizing waste activated sludge. Biochem. Eng. J. 22, 97e103. Oh, S.W., Kang, M.N., Cho, C.W., Lee, M.W., 1997. Detection of carcinogenic amines from dyestuffs or dyed substrate. Dyes Pigm. 33, 119e135. O’Neill, C., Hawkes, F.R., Lourenco, N.D., Pinheiro, H.M., Delee, W., 1999. Colour in textile effluents-sources, measurements, discharge, contents and simulation: a review. J. Chem. Technol. Biotechnol. 74, 1009e1018. Perei, K., Rakhely, G., Kiss, I., Polyak, B., Kovacs, K.L., 2001. Biodegradation of sulfanilic acid by Pseudomonas paucimobilis. Appl. Microbiol. Biotechnol. 55, 101e107. Pereira, R., Pereira, L., van der Zee, F.R., Alves, M.M., 2011. Fate of aniline and sulfanilic acid in UASB bioreactors under denitrifying conditions. Water Res. 45, 191e200. Razo-Flores, E., Donlon, B., Field, J., Lettinga, G., 1996. Biodegradability of N-substituted aromatics and alkylphenols under methanogenic conditions using granular sludge. Water Sci. Technol. 33, 47e57. Singh, P., Birkeland, N.K., Iyengar, L., Gurunath, R., 2006. Mineralization of 4-aminobenzenesulfonate (4-ABS) by Agrobacterium sp. strain PNS-1. Biodegradation 17, 495e520. Singh, P., Mishra, L.C., Iyengar, L., 2004. Biodegradation of 4aminobenzensulfonate by a newly isolated bacterial strain PNS-1. World J. Microbiol. Biotechnol. 20, 845e849. 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. Than, K., Gheewala, S.H., Annachhatre, A.P., 2002. Modeling of nitrification inhibition with aniline in suspended-growth process. Water Environ. Res. 6, 531e540. Topac, F.Q., Dindar, E., Ucaroglu, S., Baskaya, H.S., 2009. Effect of a sulfonated azo dye and sulfanilic acid on nitrogen transformation. J. Hazard. Mater. 170, 1006e1013. Wang, Y.Q., Zhang, J.S., Zhou, J.T., Zhang, Z.P., 2009. Biodegradation of 4-aminobenzenesulfonate by a novel Pannonibacter sp. W1 isolated from activated sludge. J. Hazard. Mater. 169, 1163e1167.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6

Available online at www.sciencedirect.com

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

TiO2 and Fe (III) photocatalytic ozonation processes of a mixture of emergent contaminants of water Eva M. Rodrı´guez, Guadalupe Ferna´ndez, Pedro M. Alvarez, Fernando J. Beltra´n* Departamento de Ingenierı´a Quı´mica y Quı´mica Fı´sica, Universidad de Extremadura, Avenida de Elvas S/N, 06006 Badajoz, Spain

article info

abstract

Article history:

A mixture of three emergent contaminants: testosterone (TST), bisphenol A (BPA) and

Received 29 July 2011

acetaminophen (AAP) has been treated with different photocatalytic oxidation systems.

Received in revised form

Homogeneous catalysts as Fe(III) alone or complexed with oxalate or citrate ions, hetero-

14 October 2011

geneous catalysts as titania, and oxidants such as hydrogen peroxide and/or ozone have

Accepted 18 October 2011

been used to constitute the oxidation systems. For the radiation type, black light lamps

Available online 28 October 2011

mainly emitting at 365 nm have been used. The effects of pH (3 and 6.5) have been investigated due to the importance of this variable both in ozone and Fe(III) systems.

Keywords:

Removal of initial compounds and mineralization (total organic carbon: TOC) were fol-

Titania

lowed among other parameters. For the initial compounds removal ozonation alone, in

Iron

many cases, allows the highest elimination rates, regardless of the presence or absence of

Ozone

UVA light and catalyst. For mineralization, however, ozone photocatalytic processes

Testosterone

clearly leads to the highest oxidation rates. ª 2011 Elsevier Ltd. All rights reserved.

Bisphenol A Acetominophen Photocatalytic oxidation Water treatment

1.

Introduction

Nowadays it is well established that advanced oxidation processes are recommended technologies for the removal of the so called emergent pollutants of the water. These compounds are mainly those from pharmaceutical origin (antibiotics, analgesics, etc) or dedicated to personal care but also they belong to other families such as plasticizers, pesticides, phenols, etc (Kolpin et al., 2002; Daughton and Ternes, 2000). Many of these compounds show a potential disrupting character for the endocrine system of living beings (Escher et al., 2011) and are frequently found in influents and, also, effluents of wastewater treatment plants that usually do not apply advanced chemical oxidation processes (Ternes, 1998; Nelson et al., 2011).

Photocatalytic oxidation processes (POP) are well known advanced oxidation where hydroxyl radicals are formed from the synergic effects of radiation, a catalyst and an oxidant. There are two types of POP depending on the nature of the catalyst. The homogeneous POP where metal ions as catalysts are used such as in the photo-Fenton process (Zepp et al., 1992) and heterogeneous POP where metal oxides play the role of semiconductors such as titania (Legrini et al., 1993; Bhatkhande et al., 2001). Another feature of POP is the nature of the oxidant used. In homogeneous POP hydrogen peroxide is a very used oxidant while in heterogeneous POP oxygen is the classical oxidant. The main role of these oxidants is rather different depending on the POP type. Thus, hydrogen peroxide in homogeneous POP directly acts to produce hydroxyl radicals by reacting with the catalyst, for

* Corresponding author. Tel.: þ34 924289387; fax: þ34 924289385. E-mail address: [email protected] (F.J. Beltra´n). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.038

153

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6

example Fe(II) in the Fenton processes (Fenton, 1876). In heterogeneous POP the oxidant, mainly oxygen, acts to capture electrons arriving to the conduction band of the semiconductor to minimize electron-hole recombination (Turchi and Ollis, 1990). Radiation type is also another variable in these processes with UVC radiation as the classical one especially in heterogeneous POP (Matthews and McEvoy, 1992). However new radiation sources are investigated to have POP implying radiation closer to the visible spectrum of light. In this sense, black light lamps that emit radiation in the 350e390 nm range with a maximum at 365 nm are of interest since titania semiconductor can be excited with radiation energy of up to 387 nm (Bhatkhande et al., 2001). Another oxidant of high interest in POP is ozone that is recently investigated in a process called photocatalytic ozonation. Due to its higher oxidizing character and reactivity ozone can improve the formation of hydroxyl radicals in a POP throughout several mechanisms (Agustina et al., 2005). In this work, some POP (homogeneous and heterogeneous) have been investigated to remove a mixture of emergent water pollutants and the subsequent remained organic carbon (TOC). Testosterone (TST), a steroid hormone from the androgen group which is prescribed at sexual functional disorders, vascular disorders and for therapy of tumors including cancer tumors, acetominophen (AAP), very used to reduce pain and fever but can cause serious liver and gastrointestinal side effects, and bisphenol A (BPA), used primarily to make polycarbonate plastic and epoxy resins, have been chosen as model compounds because of their frequent presence in water with different organic content (Kim et al., 2007; Stackelberg et al., 2007). Although these pollutants are present in real water at concentrations below some tenths of mg L1, in this work, however, concentrations of about 1e3 mg L1 have been applied to assure accurate measurements of concentrations, follow the TOC and check possible synergic effects between oxidants, light and catalysts to make predictions about possible mechanisms of photocatalytic ozonation reactions. In any case and as result of oxidation processes applied, concentrations of compounds studied in the order of hundreds of mg L1 (that is, just one order of magnitude higher than in real water) were followed. Also, some estimation about the importance of oxidation ways (direct ozonation and hydroxyl radical oxidation) is also made for mg L1 concentrations of pollutants.

2.

Experimental

2.1.

Products, experimental set-up and procedure

BPA, TST, AAP (see Fig. 1 for molecular structures) and citric acid were obtained from Aldrich (Spain), oxalic and perchloric acids from Merck (Spain) and powdered P25 TiO2 was directly obtained from the manufacturer, Degussa AG (Germany). Other chemicals used were at least reagent grade and used as received. Two 15 W black light lamps (HQ Power Lamp15TBL) emitting mainly 365 nm radiation were used. Photocatalytic oxidation experiments were carried out in a 4 L cylindrical borosilicate glass reactor that was provided with magnetic agitation, air feeding system and devices for

CH3 HO

OH CH3

BPA NH O

HO

AAP OH

O

TST Fig. 1 e Moleular structures of pharmaceuticals studied.

temperature and pH measurements. Two black lamps were installed on opposing walls outside the reactor and the overall system was placed in a closed box to avoid the disturbing effect of direct sunlight irradiation. To start the photocatalytic experiments, 3 L of a buffered BPA, AAP and TST aqueous solution, about 105 M/each (1.5e3 mg L1/each), at pH 3 or pH 6.5 (perchloric/perchlorate ionic strength 0.03 M) were fed to the reactor, air bubbling was set at 40 L h1 and the lamps were turned on. About 20 min later, time enough to allow the lamps work steadily, a known volume of catalyst or free radical promoting agent was added: aqueous Fe(III) solution, carboxylic acid, hydrogen peroxide and/or TiO2. More details of this photoreactor and the procedure followed to complete photocatalytic experiments can be seen in a previous work (Rodrı´guez et al., 2009a). Ozonation and photocatalytic ozonation experiments were performed in the same reactor supplied in these cases with inlets for gas feeding (through a diffuser), sampling, catalyst addition and temperature measurement and outlet for the non absorbing gas (see Figure 1SI of supplementary information). A Sander Labor ozone generator was used. Gas flow rate was always 36 L h1. In photocatalytic ozonation experiments, after the time needed for the lamp emission to reach stationary conditions has elapsed, catalysts were added and the ozone-oxygen gas was fed. In all cases, at regular time intervals, samples were withdrawn from the reactor to be analyzed. Experimental conditions applied were as follows: CBPA0 ¼ CTST0 ¼ CAAP0 ¼ 105 M (1.5e3 mg L1); CTP0 as BPA ¼ 1.6  105 M (3.6 mg L1); TiO2 dose: 0.1 g L1; CFe(III) 5 M (2.8 mg L11) (as ferric perchlorate); 0 ¼ 5x10 4 CH2O20 ¼ 5  10 M (17 mg L1); COxalic0 ¼ 4  104 M (36 mg L1);

154

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6

Ccitric0 ¼ 4  104 M (77 mg L1); T ¼ 22e25  C; CO3g0 ¼ 2.5  104 M (12 mg L1) (except in some cases with 8.3x105 M (4 mg L1).

Chemical analysis

TST, AAP and BPA were analyzed by high-performance liquid chromatography with an HPLC-UV (Agilent 1100) system in a 15 cm long, 0.4 cm i.d. Kromasil C18 column with acetonitrile-water with 0.1% phosphoric acid as mobile phase (15/85 (v/v) for BPA and AAP, and 50/50(v/v) for TST), with a flow rate of 1 mL min1. Detection was made at 280, 220 and 244 nm for BPA, TST and AAP, respectively. Total iron concentration was determined by the ferrozine method (Stookey, 1970), Fe(II) concentration by the phenantroline method (Zuo, 1995), total polyphenol content (TP) by the Folin Ciocalteau method (Singleton and Rossi, 1965) and hydrogen peroxide by the method of Eisenberg (1943), following for all the determinations the procedure detailed in a previous work (Rodrı´guez et al., 2011). Bader and Hoigne´ (1981) method, based on the decoloration of the 5,5,7 indigotrisulphonate, was used to measure the dissolved ozone concentration. Ozone in the gas phase was monitored by means of an Anseros Ozomat ozone analyzer, the analysis based on the absorbance at 254 nm. The intensity of UVA radiation coming from the black light lamps into the aqueous solution was obtained by ferrioxalate actinometry (Hatchard and Parker, 1956). A photon flow of 7.32  107 E s1 was determined with both lamps simultaneously working. Total organic carbon was measured with a TOC-VSCH Shimadzu analyzer. Samples for HPLC and TOC analyses were mixed with Na2S2O3 to ensure the destruction of any oxidizing agent remaining in solution. In ozonation systems, remaining dissolved ozone in samples was eliminated by helium stripping except in cases for analyzing the dissolved ozone concentration. When needed samples were filtered before analysis (MachereyeNagel PET 0.45 mm filters).

3.

3.1.

3.1.1.

Experiments at pH 3

Fig. 2 presents, as example, the evolution of TST remaining dimensionless concentration with time for experiments carried out with ozone alone, ozone/Fe(III) in the dark and POP studied here: O2/Fe/III/UVA, O3/Fe(III)/UVA, O2/Fe(III)/Oxalate/ UVA, O3/Fe(III)/oxalate/UVA, O2/Fe(III)/H2O2/UVA, O3/Fe(III)/ H2O2/UVA and O3/H2O2/UVA. A blank experiment of ozone alone (not shown) did not lead to any difference from that of ozone/UVA radiation thus confirming the absence of ozone photolysis when radiations in the 350e390 nm range are applied. It can be seen from Fig. 2 that Fe(III) photocatalytic oxidation leads to partial removal of TST, about 63%, in 2 h reaction. Approximately, 40% removal was also reached for BPA and AAP at these conditions (see Figures 2SI and 3SI in the supplementary part). These results are due to hydroxyl free radical reactions, the free radicals coming from the photolysis of Fe(III) that at the pH of work is forming the species Fe(OH)2þ (Faust and Hoigne´, 1990). When oxalate is added, the

1.0

0.8

Results and discussion

Water POP can be classified as homogeneous or heterogeneous processes according to the liquid or solid phase nature of the catalyst used. For homogeneous and heterogeneous POP studied here, the systems O2/Fe/III/UVA and O2/TiO2/UVA can be considered as the basic ones around which any other POP involving changes of oxidant and catalyst type can be compared. Homogeneous and heterogeneous POP are studied here in this order changing the type of catalyst and/or oxidant (oxygen, hydrogen peroxide and ozone). However, for comparative reasons a number of blank experiments were also made involving ozone alone or combined with a catalyst. POP studied have been investigated at two pH values: 3 and 6.5. Although water at pH 3 could represent an unrealistic situation, POP at acidic conditions were investigated since these are the optimum ones for Fenton-type processes. Furthermore, efficiency improvement of advanced oxidation systems can be observed when hydroxyl radical scavengers such as carbonate/bicarbonate ions are removed from water. This step could be particularly convenient when a Fenton POP is immediately after applied. In a following step, addition of some

Homogeneous POP

Combinations of Fe(III) or complexed Fe(III) (with oxalate or citrate) with hydrogen peroxide or ozone were the homogeneous POP investigated.

0.6

CTST/CTST0

2.2.

alkali to raise the pH could not necessarily be an expensive step given the possible benefits obtained in the Fenton oxidation process. In any case, regardless of pH, UVA radiation experiments were firstly carried out with no photolysis results. This was expected since the compounds tested do not absorb radiation in the 350e390 nm interval (black light lamp emission range). Also, for heterogeneous experiments, TST, AAP and BPA did not show any appreciable adsorption on TiO2.

0.4

0.2

0.0 0

20

40

60

80

100

120

t, min

Fig. 2 e Evolution of dimensionless remaining concentration of testosterone with time during homogeneous oxidation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 3. Symbols and systems: , O3/UVA; C Fe(III)/UVA; B O3/Fe(III)/UVA; > O3/Fe(III)/UVA*; : Fe(III)/Oxal/UVA; 6 O3/ Fe(III)/Oxal/UVA; ; Photo-Fenton; 7 O3/Photo-Fenton; ✯ O3/H2O2/UVA. See Section 2.1 for experimental conditions except when noted. *Ozone gas concentration: 4 mg LL1.

155

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6

ferrioxalate complex is formed. This complex photolytically decomposes faster than Fe(OH)2þ producing higher concentration of hydroxyl free radicals (Hatchard and Parker, 1956; Safarzadeh-Amiri et al., 1997). Then, it is not surprising the increase of oxidation rate observed with this system compared to the previous one. Also, for the photo-Fenton system, addition of hydrogen peroxide to the Fe(III) photocatalytic system, the reaction rate also increases compared to the latter one and yields approximately similar conversions of the compounds studied. Nonetheless, comparison of this system with the H2O2-free photocatalytic oxidation is a rather difficult task since the oxidation rate is dependent on the H2O2 concentration effect that has not been investigated. Regarding the presence of ozone, it has to be noted that ozone alone, ozone/ UVA and ozone/Fe(III) processes lead to similar experimental results (only O3/UVA results are shown) since black light does not decomposes ozone and ozone-Fe(III) is an inert system regarding hydroxyl radical formation from this catalystoxidant combination. Very different results were obtained, however, when Fe(III) and black light were simultaneously applied. As observed from Fig. 2 TST conversions achieved with the ozone-free systems are much lower than those obtained when ozone processes were applied (similar results were observed in the case of BPA and AAP, see Figures 2SI and 3SI). For example, for the ozone alone process, after 20 min complete disappearance of the organics studied was noticed. This result was also similar to those of other ozone processes, regardless of the type of accompanying agent (UVA light, Fe(III), ferrioxalate, etc). As indicated above, intermediate conversion results, as far as the organic removal rates are concerned, were obtained when iron was accompanied with oxalic acid to form ferrioxalate complex or hydrogen peroxide. Thus, in between 60 and 80 min reaction, TST totally disappeared while 93 and 80e90% BPA and AAP removal, respectively, were observed in these POP at the same reaction times. In these systems hydroxyl radical oxidation is the only way of compound removal. Here, the photo-Fenton process results in the higher oxidation rates likely due to the continuous formation of radicals through the redox system: pH3

2þ

hn

FeðIIIÞ þ H2 O



!FeðOHÞ

! hnFeðIIÞ þ HO

(1)

FeðIIÞ þ H2 O2

!FeðIIIÞ þ OH þ HO

(2)

Thus, the synergic effect of this system is evident.

The fact that differences between ozone processes are negligible is likely due to the importance of the direct ozone reactions compared to the contribution of hydroxyl radicals. The kinetic regime of these ozone reactions can be established from their corresponding Hatta values determined with Equation (3): Ha ¼

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi kD DO3 CM kL

(3)

where kD, kL, DO3 are the rate constant of the direct reaction between ozone and compound M, the individual mass transfer coefficient and ozone diffusivity in water, respectively, while CM is the concentration of the organic compound. Values of these rate constants can be obtained from literature (Andreozzi et al., 2003; Deborde et al., 2005). Table 1 shows the Ha results for the ozone reactions studied. As can be seen Ha is always lower than 0.3 which means the ozone reactions are slow. At these conditions and according to film theory (Beltra´n, 2004) dissolved ozone diffuses throughout the water film close to the gaseliquid interface and reaches the bulk water where it can react with target compounds (such as TST, BPA and AAP) and/or decompose, through different ways, in hydroxyl radicals. Thus, compounds can also be simultaneously consumed through reactions with hydroxyl radicals. If ozone target compounds direct reactions would have been fast (Ha>3) ozone would only be consumed through these reactions in the proximity of the gaseliquid interface. In this situation dissolved ozone would not reach the bulk water where ozone decomposition reactions to yield free radicals take place (Beltra´n, 2004). The fact that ozone can reach the bulk water, then, explains why in the different ozone processes the organic reaction rates are not exactly the same but differences, although low, are also observed (Beltra´n, 2004). These differences increase with the decrease of the direct reaction rate constant, which supports previous comments. In the case of TST, the direct ozone reaction rate constant is not known but Barron et al. (2006) and Brose´us et al. (2009) have reported for the ozone-progesterone reaction values of 480  30 M1 s1 in the 2e8 pH range and 601  9 M1 s1 at pH 8.1, respectively. Since both molecules, testosterone and progesterone, have similar molecular structure and nucleophilic points where ozone can attack, specifically their double carbon bond, it is expected a similar value for the rate constant of the ozone-TST reaction. The rate

Table 1 e Values of the rate constant of the direct reaction between ozone and compounds studied and corresponding Hatta numbers at different pH and reaction time.a Reaction

pH 3 kD, M1 s1

O3 þ BPA O3 þ TST O3 þ AAP a b c d

1.71  10 4.80  102 1.84  103

c d

0.298 0.051 0.097

t ¼ 5 min 0.196 0.037 0.044

pH 6.5 kD, M1 s1

Hatta values t ¼ 0 min

4 b

Reaction

t ¼ 10 min 0.083 0.027 0.025

O3 þ BPA O3 þ TST O3 þ AAP

6 b

1.08  10 4.80  102 1.37  106

c d

Hatta values t ¼ 0 min

t ¼ 5 min

t ¼ 10 min

2.296 0.051 2.613

1.241 0.034 1.388

0.459 0.02

Results correspond to ozonation alone experiments (similar results are obtained from data of other ozone processes). Deborde et al. (2005). Andreozzi et al., 2003. Value of the ozone-progesterone reaction according to Barron et al. (2006).

e

156

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6

FeðIIÞ þ O3 /FeO2þ þ O2 FeO2þ /FeðIIIÞ þ HO þ HO

(4) 

2FeO2þ /2FeðIIIÞ þ HO þ HO 2

(5) (6)

In reaction (4), Fe(II) formed from Fe(III) photolysis shown in reaction (1) reacts with ozone to yields FeO2þ that eventually gives hydroxyl radicals and hydrogen peroxide in the ionic form through reactions (5) and (6) (Logager et al., 1992). Also, Fe(III) will be again photoreduced through reaction (1) to yield more hydroxyl radicals. At pH acid, the ionic form of hydrogen peroxide will pass to its molecular form H2O2 that can react with Fe(II) in another Fenton process to yield more hydroxyl radicals. In addition, formation of hydrogen peroxide from direct ozone reactions gives rise to another photo-Fenton process. Other possible reactions such as the ozonehydrogen peroxide reaction to yield free radicals are negligible at pH 3 (Staehelin and Hoigne´, 1982). Then, development of these reactions confirm the synergic effect between ozonation and photocatalytic oxidation. In the cases studied here both O3/Fe(III)/UVA and O3/photo-Fenton systems show clearly this synergism. However, the results also show that this synergic effect is highly dependent on the relative importance between the rates of ozone-organic direct reactions and ozone decomposition reactions in hydroxyl radicals. Given the molecular structure of organics studied, polyphenol compounds are possible first intermediates of oxidation (Decoret et al., 1984). Then, total concentrations of these compounds were also followed with time as shown in Fig. 3. As it can be seen, results are significantly different depending on the presence or absence of ozone. Thus, in ozone-free oxidation, polyphenol compound concentration clearly increases with time and, in some cases, after reaching

2.4x10

2.0x10

1.6x10

CTP as BPA, M

constant value is much lower, however, than the ones corresponding to the ozone-BPA and ozone-AAP reactions. Because of TST-ozone reaction is the slowest one among those studied here, TST removal rate slightly increases when ozone and Fe(III)/UVA or ozone/photo-Fenton processes are simultaneously applied compared to the single ozonation process. This is undoubtedly due to the contribution of the reaction with hydroxyl radicals coming from the different mechanism in these POP. However, in the case of BPA and AAP differences among ozone processes in organic removal rates are even much lower, especially, in the case of BPA that result to be negligible. Since BPA and AAP are dissociating organic compounds the apparent rate constants of their direct reactions with ozone are pH dependent (Hoigne´ and Bader, 1983). At pH 3 these values are 17140 and 1840 M1 s1 for BPA and AAP, respectively (Andreozzi et al., 2003; Deborde et al., 2005). These results justify the higher reactivity of these compounds with ozone compared to TST. In Fig. 2 the effect of ozone concentration on the ozone/ UVA/Fe(III) process can also be observed at two concentration levels: 2.5x104 and 8.3x105 M (12 and 4 mg L1). It is clearly seen that the increasing concentration of ozone accelerates the organic removal rate which supports the action of ozone direct reaction but also the possible participation of ozone to increase hydroxyl radical formation through reaction (4)e(6):

1.2x10

8.0x10

4.0x10

0.0 0

20

40

60

80

100

120

t, min

Fig. 3 e Evolution of total polyphenol concentration measured as BPA with time during homogeneous oxidation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 3. See Fig. 2 for symbols and experimental conditions.

a maximum value, it decreases. In these latter cases, it is clear that a more intensive oxidation process is being applied. Thus, the simple Fe(III) photocatalytic process does not lead to any decrease of these compounds (after 2 h, formation of polyphenols still continues) while Fe(III)/Oxalate/UVA system seems to be a stronger oxidizing process since, in this case, polyphenols reach a maximum concentration after 60 min and then slowly decrease. The most important process among ozone-free systems is the photo-Fenton system. Here, maximum polyphenol concentration is reached in 30 min and then the concentration slows down to 8  106 M as BPA (1.83 mg L1) in 2 h reaction. When ozone is applied, regardless of the ozonation system, polyphenol concentration diminishes since the beginning of the process to completely disappear in just 80 min reaction. These results are also expected because of the direct reactions between ozone and phenols are of the order of 103e104 M1 s1 at acid pH (Hoigne´ and Bader, 1983). The importance of ozone gas concentration applied is also seen in Fig. 3 since ozonation with 8.3  105 M (4 mg L1) retards complete polyphenol conversion up to 2 h reaction. Since Fe(III) is applied to photolytically decompose and yield Fe(II) and free radicals, the system becomes a redox one especially when ozone or hydrogen peroxide is also present. Then, it is interesting to follow the Fe(II) concentration with time. This can be seen in Fig. 4. As observed, in ozone-free systems, Fe(II) concentration increases with time specially in Fe(III) and Fe(III)/oxalate photocatalytic processes where Fe(II) is accumulated in water. When hydrogen peroxide is also present, there is a photo-Fenton process, Fe(II) concentration slowly increases and keeps a low stationary value after 60 min. This value 5  106 M (0.28 mg L1), in any case, is much lower than the one Fe(II) reaches after 2 h with the other two ozone-free systems (about 4  105 M (2.2 mg L1)). For the same reasons, in ozone systems, the trend Fe(II) concentration follows is similar to the one observed for the photo-Fenton process.

157

5x10

1.0

4x10

0.8

3x10

0.6

CTOC/CTOC0

CFe(II), M

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6

2x10

1x10

0.4

0.2

0

0.0 0

20

40

60

80

100

120

0

20

40

t, min

Fig. 4 e Evolution of Fe(II) concentration with time during homogeneous oxidation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 3. See Fig. 2 for symbols and experimental conditions.

The reactivity of ozone oxidation systems can also be deduced from the evolution of the dissolved ozone concentration (details are shown in Figure 4SI in supplementary data). Thus, the O3/Fe(III)/UVA system was observed as the most reactive one regarding ozone consumption since the concentration of ozone is close to the detection limit of the method applied (about 50 mg L1) for more than 1 h. Then, concentration starts to increase. Notice that polyphenol concentration starts to be negligible after 60 min reaction time (see Fig. 3). In the rest of ozone systems, dissolved ozone concentration follows a similar trend, low or null concentration during the first minutes and then an increase of concentration (for more details see Figure 4SI). These trends are likely due to the contribution of direct ozone reactions during the beginning of the process where BPA, AAP and TST were present and then to the decomposition reactions of ozone in free radicals that start to be important after fast ozone reacting compounds have disappeared. For example, in the ozonation alone system, the concentration of dissolved ozone was negligible during the first 20 min (notice that BPA, AAP and TST needed about 20 min for total conversion). After this time, the concentration of ozone started to rapidly increase to reach a stationary value after 50 min reaction (the time needed to remove polyphenols) (See also Figure 4SI for more details). Therefore, the low values of dissolved ozone concentration observed with the ozone photocatalytic systems are undoubtedly due to the higher ozone consumption in the mechanism of free radical reactions and Fe(II)ozone reaction. Finally, in Fig. 5 the evolution of total organic carbon (TOC) with time corresponding to the experiments commented is shown. Regarding TOC, Fe(III) photocatalytic oxidation gives no mineralization at all while ozonation alone only leads to 15% TOC removal after 2 h reaction. When some agent is added to this system (i.e. hydrogen peroxide or oxalate to have a photo-Fenton or Fe(III)/oxalate/UVA systems, respectively) or ozone combined with other agent is applied, TOC clearly

60

80

100

120

t, min

Fig. 5 e Evolution of dimensionless remaining TOC with time during homogeneous oxidation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 3. See Fig. 2 for symbols and experimental conditions.

diminishes though with different rates depending on the oxidation system applied. The most efficient system is the O3/ Fe(III)/UVA oxidation that allows about 80% TOC conversion in 2 h. Also, the ferrioxalate systems lead to similar TOC conversions but in these cases TOC contribution from oxalate have to be considered. Given the fact that after 20e40 min dissolved ozone is already appreciable in water and that main ozone reacting compounds have disappeared (polyphenols) TOC consumption are undoubtedly due to hydroxyl free radical reactions. TOC results show in Fig. 5, then, clearly confirm a high synergic effect between ozonation and photocatalytic oxidation (Fe(III)/UVA) since none of these processes individually applied allow TOC conversions higher than 15% after 2 h reaction.

3.1.2.

Experiments at pH 6.5

A series of experiments of homogeneous POP were also made at pH 6.5. Since Fe(III) is unstable at this pH, experiments were carried out in the presence of oxalate or citrate to form the corresponding ferricarboxylate complexes. These complexes photolytically decompose to yield free radicals (Faust and Zepp, 1993; Zhang, 2000; Rodrı´guez et al., 2009a). In fact, ferrioxalate was already used in the experiments at pH 3 for comparative reasons. At pH 6.5, ferricitrate complex is also used because of its higher stability compared to ferrioxalate at ´ lvarez et al., 2010; these conditions (Nansheng et al., 1998; A Rodrı´guez et al., 2009b). Also, as at pH 3, some ozonation experiments were carried out. Thus, the following oxidation systems were studied at pH 6.5: Simple ozonation, O3/UVA, O3/Fe(III)/oxalate/UVA, O2/Fe(III)/ O2/Fe(III)/Oxalate/UVA, Citrate/UVA, O3/Fe(III)/Citrate/UVA and O3/H2O2/UVA. The evolution of BPA remaining dimensionless concentrations with time corresponding to the experiments at pH 6.5 is shown in Fig. 6 as an example (information of the changes observed in the concentrations of the other two model

158

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6

1.0

0.8

CBPA/CBPA0

0.6

0.4

0.2

0.0 0

20

40

60

80

100

120

t, min

Fig. 6 e Evolution of dimensionless remaining concentration of BPA with time during homogeneous oxidation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 6.5. Symbols and systems: , O3/UVA; C Fe(III)/Oxal/UVA; B O3/Fe(III)/Oxal/UVA; : Fe(III)/Cit/UVA; 6 O3/Fe(III)/Cit/UVA; ✯ O3/H2O2/UVA. See Section 2.1 for experimental conditions.

compounds, TST and AAP, is shown in Figures 5SI and 6SI of supplementary data, respectively). As can be seen from these figures, Ferrioxalate/UVA system only allows compound conversions between 20 and 25% in 2 h reaction. In fact, the reaction is stopped after about 20 min likely due to the unstability of the complex at pH 6.5. On the contrary, the higher stability of ferricitrate complex allows its application to reach between 90 and 95% compound conversion in 2 h. However, the best POP are those involving ozone. Thus, ozone processes lead to total conversion of initial compounds in between 15 and 20 min with reaction rates faster than those observed for the same compounds in experiments at pH 3 (see Figures 2, 2SI and 3SI for details). In this case, differences between ozone processes are even lower than those observed at pH 3 including the case of TST. The main reason of the increase of oxidation rates can be attributed to different mechanisms depending on the compound tested. Thus, BPA and AAP react with ozone at pH 6.5 faster than at pH 3 since both are dissociating compounds. For example, the rate constants of the direct reactions between ozone and the nondissociating and first dissociating forms of BPA are, according to Deborde et al. (2005), 1.68x104 and 1.06  109 M1 s1, respectively, the pKa being 9.6. With this information the apparent second order rate constant of the ozone-BPA reaction at pH 6.5 can easily be determined as 1.08  106 M1 s1. In the case of AAP (pKa ¼ 9.36), according to data of Andreozzi et al. (2003), similar calculations lead to a value of 1.37  106 M1 s1 at pH 6.5. In the case of TST, a nondissociating compound in water, the rate constant remains as at pH 3. Then, Hatta numbers of the ozone direct reactions are calculated with Equation (3) to yield values between 2.6 and 0.5 for ozone-BPA and ozone-AAP reactions and lower than 0.3 for the ozone-TST reaction during the first 10 min of

reaction (see Table 1). Then, according to what is commented in Section 3.3.1 it is reasonable to admit that both BPA and AAP are exclusively consumed through their direct reactions with ozone during this initial period (Beltra´n, 2004). Regarding TST, however, the ozone direct reaction is slow and competes with the ozone decomposition reactions to yield free radicals as at pH 3. Furthermore, the main route of reaction rate for TST is likely the hydroxyl radical oxidation since it is well known that ozone decomposition in free radicals is triggered with the increasing pH if direct reactions are not fast (Beltra´n, 2004) as it is this case. For example, one possible route is the reaction between ozone and the ionic form of hydrogen peroxide that at pH 6.5, even at very low concentrations, decomposes ozone at a high rate (the rate constant of the ozone-hydroperoxide ion reaction is 2.8  106 M1 s1 as Staehelin and Hoigne´ (1982) reported). This also explains the high rates observed with the O3/H2O2/UVA system but also in any of ozone involving oxidation since the ozone reactions with the organic present yields hydrogen peroxide through cycloaddition reactions to the aromatic ring (Mvula and von Sonntag, 2003). Although concentration of H2O2 was not followed in ozone experiments, but in the O3/H2O2/UVA system, it is well known the formation of hydrogen peroxide in ozonation processes (Mvula and von Sonntag, 2003; Leitzke and von Sonntag, 2009). In fact, in ozone processes, hydrogen peroxide formed in direct ozone reactions can lead to hydroxyl radical formation by reacting with ozone: M þ O3 /P þ H2 O2 pK¼9:3

(7)

þ H2 O2 %HO 2 þH

(8)

  HO 2 þ O3 /HO2 þ O3

(9)

pK¼4:8

þ  HO2 %O 2 þH

(10)

  O 2 þ O3 /O3 þ O2

(11)

þ   O 3 þ H /HO þ O2

(12)

Also, hydroxyl radicals are generated in a modified photoFenton process formed with ozone, Fe(III) and UVA through reactions (1), (7) and (2). The higher reactivity of ozone systems has also been checked by following the total polyphenol concentration with time (not shown). At pH 6.5, there is no variation of polyphenol concentration when Fe(III)/oxalate/UVA process is applied which corroborates the low organic conversion achieved with this system. With the Fe(III)/citrate/UVA system an increase of total polyphenol concentration is observed during the first 100 min of reaction to eventually slowly diminish. In the case of ozone systems, regardless of the type of ozonation process applied, total polyphenol content continuously decrease with time to become practically negligible after 2 h reaction. Also, the absence of dissolved ozone during the first minutes of reaction confirmed the development of ozone fast reactions in water (Beltra´n, 2004). Measurements of Fe(II) concentration in water also confirm the reactivity of ozone systems and the redox process formed since the

159

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6

concentrations remain below 105 M (0.55 mg L1) while in the ferricitrate UVA system Fe (II) concentration were twice higher and remained constant after approximately 20 min of reaction. No appreciable formation of Fe(II) was noticed in the ferrioxalate/UVA system due to the absence of oxidation reactions. Finally, in Fig. 7 the variation of remaining dimensionless TOC with time corresponding to POP experiments at pH 6.5 is shown. As it can be seen, after 2 h reaction, less than 10% TOC conversion is observed with ozonation alone (or O3/UVA) and ferricitrate/UVA systems while no variation at all is seen with the ferrioxalate/UVA system which is in agreement with the results previously shown at pH 6.5 for the individual pharmaceuticals. The other three ozone advanced oxidation systems, however, lead to about 60% TOC conversion in 2 h although at different reaction rates. The fastest process is the combination between ozone and Fe(III)/oxalate/UVA that allows 50% TOC conversion to be reached in 90 min. However, this oxidation system is nearly inhibited after 2 h of reaction. On the contrary, the O3/H2O2/UVA and, specially, the O3/ Fe(III)/citrate/UVA systems present, after 2 h, faster oxidation rates and the processes do not seem to become inhibited. For a practical case, however, systems containing carboxylic acids are not recommended due to their contribution to increase TOC of the water. For example, in the O3/Fe(III)/citrate/UVA experiments applied here, citric acid was added well in excess of Fe(III) (see experimental part for concentrations) to assure total complexation of Fe(III). As a result, the organic carbon of water from citrate represented about 85% of measured TOC at the start of the experiment. Since the start of the reaction, a decrease of the concentration of total citrate (sum of free and complexed) in water was observed so that after about 70 min citrate had completely reacted with hydroxyl radicals (ozone does not react directly with citrate). At this reaction time, total dissolved iron concentration was about half the initially measured at the start of the experiment which means that some other iron complexes were formed since Fe(II)

concentration was very low. After about 2 h reaction, total dissolved iron was not detected in water. Also, at 70 min reaction only about 30% TOC was removed which means that some intermediates formed from citrate oxidation were in water. In any case, hydroxyl free radical oxidation continued until total iron complexes conversion was achieved in 120 min. Apart from the practical point of view, there is a clear synergism in the O3/Fe(III)/citrate/UVA system to remove TOC from water if results are compared to those obtained with ozonation or Fe(III)/citrate/UVA systems.

3.2.

Heterogeneous POP

In a second part of this work, heterogeneous catalytic photooxidation processes were studied. The experiments were also carried out at pH 3 and 6.5 for comparative reasons with the homogeneous ones. Now, TiO2 was used as solid catalyst or semiconductor since its good efficiency in photocatalytic oxidation processes has long been extensively tested (Legrini et al., 1993). Therefore, the O2/TiO2/UVA system can be considered as the basic one in the heterogeneous POP studied here.

3.2.1.

Experiments at pH 3

At pH 3 the following oxidation systems were investigated: O2/ TiO2/UVA, O3/TiO2/UVA, O2/TiO2/Fe(III)/UVA and O3/TiO2/ Fe(III)/UVA. As an example, the changes of organic remaining dimensionless concentration with time in the heterogeneous POP studied are shown in Fig. 8 for the case of BPA. As it can be seen, TiO2 photocatalytic oxidation is not as an efficient process as homogeneous Fe(III) photocatalytic oxidation since compound conversions achieved in 2 h are between 30 and 40% much lower (For information on AAP and TST changes see Figures 7SI and 8SI of supplementary part). Addition of Fe(III) to the heterogeneous TiO2 photocatalytic system allows 1.0

1.0

0.8

0.8

CBPA/CBPA0

CTOC/CTOC0

0.6

0.6

0.4

0.4

0.2

0.2

0.0 0

20

40

60

80

100

120

t, min 0.0 0

20

40

60

80

100

120

t, min

Fig. 7 e Evolution of dimensionless remaining TOC with time during homogeneous oxidation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 6.5. See Fig. 6 for symbols and experimental conditions.

Fig. 8 e Evolution of dimensionless remaining concentration of BPA with time during heterogeneous oxidation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 3. Symbols and systems: , O3/UVA; C TiO2/UVA; B O3/TiO2/UVA; : TiO2/Fe(III)/UVA; 6 O3/TiO2/Fe(III)/UVA. See Section 2.1 for experimental conditions.

160

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6

a significant increase of the conversion that, at these conditions, reaches values between 85 and 97% which are already of the order of those achieved with homogeneous ferrioxalate photocatalytic oxidation. The increase observed is due to the sum of the contributions of both photocatalytic systems to generate hydroxyl free radicals although some synergic effect should not be disregarded. Thus, it is known that TiO2 photocatalytic oxidation generates small amounts of hydrogen peroxide from possible recombination of superoxide ion radicals formed when oxygen captures electrons in the conduction band (Wang et al., 2002): þ

hn

TiO2

!h þ e

(13)

þ

h þ OH /HO

(14)

 O2 þ e /O 2

(15)

 2O 2 þ H2 O/H2 O2 þ 2O2

(16)

Then, hydrogen peroxide in the presence of Fe(II) formed from Fe(III) photolysis and UVA radiation forms a new photoFenton process that can enhance the formation of hydroxyl radicals. However, by comparing Figs. 9 and 4 it is deduced that formation rate of Fe(II) is higher in the TiO2 photocatalytic process than in the Fe(III)/UVA process which would imply the action of Fe(III) to capture electrons of the conduction band to avoid electron-hole recombination (Quici et al., 2007; Mesta´nkova´ et al., 2005; Rodrı´guez et al., 2009a): FeðIIIÞ þ e /FeðIIÞ

(17)

Then, it is unlike that a photo-Fenton process develops in a great extension. In any case, ozone involving systems are the most efficient to simultaneously eliminate the organics BPA, TST and AAP as observed in Fig. 8 for the BPA case (more information is given in Figures 7SI and 8SI for the AAP and TST cases, respectively). In fact, oxidation rates are even something higher than those

obtained in the homogeneous ozone oxidation and photocatalytic oxidation at pH 3 and similar to those obtained in ozone systems at pH 6.5 as shown in Fig. 6 (for more information see also Figures 5SI and 6SI of supplementary data). However, on the contrary to what is observed during homogeneous ozonation processes at pH 3, where there was not practically differences between ozone systems (see especially Fig. 8 for BPA), the heterogeneous O3/TiO2/Fe(III)/UVA system presents the highest reaction rates which confirms the synergic effect between processes since in the presence of ozone-hydrogen peroxide is clearly formed as explained before and a photo-Fenton process is established. The highest reactivity of this system can also be deduced from data of dissolved ozone concentration shown in Fig. 10. In Fig. 10 it can be seen that heterogeneous O3/TiO2/Fe(III)/UVA system gives rise to very low concentrations of ozone during the first minutes of reaction (first 20 min, just the time needed to complete disappearance of initial compounds) and then an increase to reach stationary values, that depend on the ozone process applied. Thus, after the first 20 min during ozonation alone a stationary ozone concentration of about 4x105 M (1.9 mg L1) is reached in less than 40 min while in the case of photocatalytic ozonation, O3/TiO2/UVA, the concentration of dissolved ozone stays nearly constant at about 3x106 M (0.14 mg L1) for the whole reaction period and, finally, during O3/TiO2/Fe(III)/UVA oxidation ozone starts to accumulate in water from 43 min reaction to continuously increase to reach 3.5x105 M (1.7 mg L1) after 2 h. These results, however, cannot be interpreted by only considering the initial organic conversions but all other possible reactions developed as commented later. The evolution of Fe(II) concentration with time (see Fig. 9) supports the reactivity of the O3/TiO2/Fe(III)/UVA system and the possible development of a photo-Fenton process. Thus, Fe(II) concentration rapidly increases during the first 5 min to reach 105 M (0.56 mg L1) and then it slowly decreases during the next 40 min. Finally, it increases again to reach a stationary value something higher than 105 M (0.56 mg L1).

-5

5x10

5x10

4x10

4x10

3x10

3x10

-5

CO3d, M

CFe(II), M

-5

2x10

-5

2x10

-5

1x10

1x10

0

0 0

20

40

60

80

100

120

t, min

Fig. 9 e Evolution of Fe(II) concentration with time during heterogeneous photocatalytic oxidation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 3. See Fig. 8 for symbols and experimental conditions.

0

20

40

60

80

100

120

t, min

Fig. 10 e Evolution of dissolved ozone concentration with time during heterogeneous ozonation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 3. See Fig. 8 for symbols and experimental conditions.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6

It has to note that during the first minutes hydrogen peroxide concentration is expected to also increase due to ozone direct reactions. This was also observed in this work where hydrogen peroxide reached a maximum value, usually after 15e20 min and then it decreased as previously reported for similar ozone systems (Beltra´n et al., 2010). This trend seems to corroborate the development of the photo-Fenton process up to 45 reaction minutes. On the other hand, during TiO2/ Fe(III)/UVA system, Fe(II) accumulates in water just since the start of oxidation to reach, after 30 min, a stationary concentration of about 5  105 M (2.8 mg L1). In other words, all added Fe(III) is in the form of Fe(II) since the start of oxidation. Then, the absence of Fe(II) consumption suggests that in this oxidizing system photo-Fenton processes do not develop which also confirm the lower reactivity of the TiO2/Fe(III)/UVA system compared to the O3/TiO2/Fe(III)/UVA system as, also, shown later for the case of TOC reduction. Regarding the evolution of total polyphenols, compounds of high reactivity with ozone, it was observed (not shown) high degradation rates similar for the ozone systems applied (total conversion was reached in 60 min) and null or negligible decrease concentration in the ozone-free systems which supports the lower efficiency already shown to remove the initial organics and TOC (as commented below). Finally, as observed in Fig. 11, the evolution of TOC with time during the processes investigated not only confirm the higher efficiency of ozone combined systems but the results of Fe(II) and dissolved ozone concentrations shown in Figs. 9 and 10, respectively. Thus, from Fig. 11 it is seen that ozone-free photocatalytic oxidation and ozonation alone hardly allow TOC conversions lower than 10% in 2 h reaction. Results of ozone-free oxidation also support the low reactivity of polyphenols with these systems and, specifically the rapid formation of Fe(II) in the TiO2/Fe(III)/UVA system. During ozonation alone TOC is reduced about 10% during the first 30 min to reach a stationary value which explains why the dissolved ozone concentration starts to increase at about the same reaction time (see Fig. 10) to also reach a stationary value

1.0

CTOC/CTOC0

0.8

0.6

0.4

0.2

0.0 0

20

40

60

80

100

120

t, min

Fig. 11 e Evolution of dimensionless remaining TOC with time during heterogeneous oxidation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 3. See Fig. 8 for symbols and experimental conditions.

161

from 40 min reaction suggesting the inhibition of oxidation. In addition, the absence of Fe(III) in this system (O3/TiO2/UVA) makes it not possible any photo-Fenton process development. In the case of heterogeneous photocatalytic ozonation, TOC degradation continuously takes place for the whole reaction period to reach a conversion of more than 90% after 2 h. This means that the process was not inhibited and explains the low concentration of dissolved ozone for the 2 h reaction. In the system, however, heterogeneous O3/TiO2/Fe(III)/UVA although TOC removal rates are the highest among oxidation systems applied during the first 40e50 min, the process is finally stopped so that maximum TOC conversion achieved, in 2 h, was practically the same as after 50 min, about 82%. This process inhibition is also deduced from data of dissolved ozone concentration shown in Fig. 10. Thus, in this figure it is also seen negligible ozone concentrations measured during the first 43 min, the time period where reactions did develop, but a sudden increase of ozone concentration to reach, after 2 h, a value close to that measured for the ozonation alone process. This undoubtedly means that some sort of inhibition of the process occurs after the first 40 min. A priori, two situations explain the inhibition observed in Fig. 11: a) a possible reduction of TiO2 activity for the fixation of Fe(III) atoms on the catalyst surface or b) the development of a photo-Fenton process during the first 40e50 min when TOC reduction was appreciable. The first explanation, however, is not possible since total dissolved Fe concentration kept practically constant for the whole reaction period, so that iron was not deposited on the TiO2 surface. The second explanation matches the results of the changes observed in dissolved ozone, hydrogen peroxide and Fe(II) concentrations with time as indicated above. Thus, it is evident the synergic effect of the O3/TiO2/Fe(III)/UVA system.

3.2.2.

Experiments at pH 6.5

Finally, a series of experiments of heterogeneous POPs were also carried out a pH 6.5. Now, to avoid Fe(III) precipitation, oxalic acid or citric acid was also added to form ferricarboxylates complexes, known for their photolytic activity (Faust and Zepp, 1993; Zhang, 2000; Rodrı´guez et al., 2009a). Fig. 12 shows, as example, the evolution with time of the remaining dimensionless concentrations of AAP in heterogeneous POPs. As it can be observed, typical TiO2 photocatalytic oxidation leads to slightly better conversion results than at pH 3 but organic conversions remain between 50 and 60% after 2 h reaction (more information is given in Figures 9SI and 10SI for the BPA and TST cases, respectively). When Fe(III) and oxalic acid are added conversions even diminish to about 20% likely due to the unstable ferrioxalate complex at this pH value something also observed in the corresponding homogeneous system (see Fig. 6). However, when citric acid was charged instead of oxalic acid, the higher stability of the ferricitrate complex allows organic conversions to increase to 75% for AAP, 87% for BPA and 90% for TST after 2 h. Nonetheless, when ozonation results are considered, once more, total conversion of compounds studied is reached in less than 20 min. In fact, with the heterogeneous ozone POP applied total conversion of BPA, AAP and TST can be achieved in just 5 min for the first two compounds and 10 min for the third (in the O3/TiO2/Ferrioxalate/UVA system). Contrary to

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6

1.0

1.0

0.8

0.8

0.6

0.6

CTOC/CTOC0

CAAP/CAAP0

162

0.4

0.2

0.4

0.2

0.0

0.0

0

20

40

60

80

100

120

0

20

t, min

the results of ozonation processes in homogeneous (pH 3 and 6.5) and heterogeneous (pH 3) systems, differences of the time needed to reach total conversion are higher for different heterogeneous ozone POP at pH 6.5 as observed, for example, in Fig. 12 for AAP. Thus, for ozonation alone between 20 and 25 min are necessary for total conversions of BPA and AAP and TST, respectively. Thus, it is evident that in heterogeneous ozone POP removal of compounds are not only due to ozone direct reactions but also to possible surface and hydroxyl radical reactions. For example, photo-Fenton processes are likely to develop as also shown in experiments at pH 3. Results obtained with total polyphenol concentration (not shown) indicate high reaction rates in ozone processes faster than those achieved at pH 3 (total polyphenol disappearance is now observed after 40 min reaction). On the contrary, ozone-free heterogeneous POP show again a null or even an increase of polyphenol concentration (case of systems with ferricarboxylates) or a slight decrease (case of classical TiO2 photocatalytic oxidation). In the case of TOC evolution (as seen in Fig. 13) only combined heterogeneous ozone POP allow significant decrease of this parameter. Regarding the O3/TiO2/ UVA process, TOC conversion results at pH 6.5 are lower than those at pH 3 with 70% and 90% TOC reductions, respectively, after 2 h reaction. The highest efficiency of this system at acid conditions could likely be due to the reactions (18) and (12): O3ðadÞ þ e

 /O 3

60

80

100

120

t, min

Fig. 12 e Evolution of dimensionless remaining concentration of AAP with time during heterogeneous oxidation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 6.5. Symbols and systems: , O3/UVA; C TiO2/UVA; B O3/TiO2/ UVA; : TiO2/Fe(III)/Oxal/UVA; 6 O3/TiO2/Fe(III)/Oxal/UVA; ; TiO2/Fe(III)/Cit/UVA; 7 O3/TiO2/Fe(III)/Cit/UVA. See Section 2.1 for experimental conditions.



40

(18)

Regarding the O3/TiO2/Ferrioxalate/UVA system no inhibition is observed as a difference of what is seen at pH 3 for the O3/TiO2/Fe(III)/UVA system. In any case, when carboxylates are present an important fraction of TOC reductions achieved are likely due to the removal of these compounds which would also agree the decrease observed in total iron

Fig. 13 e Evolution of dimensionless remaining TOC with time during heterogeneous oxidation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 6.5. See Fig. 12 for symbols and experimental conditions.

concentration with time as indicated below. Then, TOC is not a good parameter to follow the efficiency of carboxylate POP systems. For systems with ozone and ferricarboxylates a decrease of total dissolved iron was observed since the start of reaction in the case of ferrioxalate and after about 30 min for the case of ferricitrate, which can be attributed to the lower stability of the former at pH 6.5 (Abrahamson et al., 1994; Nansheng et al., ´ lvarez et al., 2010). On the other hand, Fe(II) stayed at 1998; A negligible levels during TiO2 photooxidation in the presence of ferrioxalate while about 2  106 M (0.11 mg L1) concentrations were achieved in the presence of ferricitrate systems, regardless of the presence of ozone in both cases. This confirms the higher reactivity of ferricitrate system likely due to the higher stability of the complex compared to that of ferrioxalate. Concentrations of dissolved ozone follow typical trends of ozonation systems, that is, very low concentration during the first 20 min reactions, as a consequence of fast direct reactions of ozone with initial compounds and, likely, first intermediates (measured as polyphenols) and an increase to reach a stationary concentration. The lowest dissolved stationary concentration was observed in the O3/TiO2/UVA system which suggests this system presented the highest reactivity.

3.3.

Kinetic aspects

Although there are basically one or two main kinetic contributions to the organic removal rate in the POP studied: the hydroxyl radical reaction and/or the direct ozone reaction, POP are complex systems involving different reaction mechanisms (direct reactions of ozone, reactions with hydroxyl radicals generated through different ways, possible direct oxidation with positive holes on the TiO2 surface, photoFenton processes, etc). Thus, the kinetic study has been limited to some aspects related to the direct ozonation

163

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6

contribution in some of the ozone processes and the estimation of the hydroxyl radical concentration to remove the initial compounds and TOC.

3.3.1.

Kinetics of initial organics removal

In ozone-free systems, the organics at any time are removed through free radical oxidation so that a mass balance of any organic compound in water in the semibatch photocatalytic reactor used is as follows: dCM ¼ kHO CHO CM  dt

(19)

where kHO, CHO and CM are the rate constant of the hydroxyl radical-M reaction and the concentrations of hydroxyl radical and compound M, respectively. From experimental values of the concentration and accumulation rate of M, left side of equation (19), determined in this work and the rate constant kHO obtained from literature data the concentration of hydroxyl radicals can be estimated. Since the three compounds studied were simultaneously treated it is expected that values of CHO determined from equation (19) be similar when applied to reaction rate and concentration data of BPA, AAP or TST. Literature only gives values of kHO for BPA and AAP reactions with hydroxyl radicals (Bisby and Tabassum, 1988; Gozmen et al., 2003) so that from data of these compounds CHO was easily found by applying equation (19) (See Table 2 for results). In the case of TST, the rate constant of the hydroxyl radical reaction and the corresponding CHO were determined by a trial and error method as follows. First, a value of kHO was assumed to get CHO from equation (19) at different conditions and oxidation process and then this value was compared to those already determined from data of BPA and AAP for the same oxidation process. The best value of kHO was finally the one minimizing the sum of the squares differences between CHO from TST data and the assumed kHO and the average values of CHO from data of BPA and AAP directly determined from Equation (19) at different reaction time and oxidation process. The optimum kHO value was found to be 1.5  1010 M1 s1. Table 2 also shows the results of CHO from TST data. As it can be seen, in most of cases, at a given time, pH and oxidizing system, similar results of CHO were obtained from Equation (19) for the three initial compounds. This was expected since the three

compounds were simultaneously treated. Also, it is seen that CHO values are nearly constant with time with variations lower than one order of magnitude. At pH 3, the highest concentration of hydroxyl radicals are produced in photoFenton and ferrioxalate/UVA and TiO2/Fe(III)/UVA systems for homogeneous and heterogeneous processes, respectively, while, at pH 6.5, the best processes for generating hydroxyl radicals are the ferricitrate/UVA systems with or without TiO2 for homogeneous and heterogeneous processes, respectively. These results coincide with those commented about the changes in the concentration of compounds with time. In ozonation processes, there is another contribution to the organic reaction rate: the direct ozone-M reaction, so that Equation (19) is now: dCM ¼ kD CO3 CM þ kHO CHO CM  dt

(20)

In these cases, however, the estimation of hydroxyl radical concentration from Equation (20) presents a high uncertainty due to the low values measured for the dissolved ozone concentration. Nonetheless, in ozone systems an important objective is to know the relative effect of the direct and hydroxyl radical reactions. Here, the contribution of the direct ozone reaction to the removal of the organics studied, first term of the right side of Equation (20), has been estimated in a few cases, also because of the low values of the concentration of ozone (about 106 M (0.05 mg L1)) and possible interferences in determining the ozone concentration with some ozone systems (i.e. those where a Fenton reaction develops). This contribution, DR%, is given by Equation (21): DR% ¼

kD CO3 CM  100 dCM =dt

(21)

Thus, only for experiments at pH 3 in ozonation alone and photocatalytic ozonation in the presence of Fe(III), contribution of the direct reaction was estimated after 10 min reaction (for lower times dissolved ozone concentration was very close to the detection limit of the method). Both in ozonation alone and ozone/Fe(III)/UVA systems, DR%, in the case of BPA and AAP, was always something higher than 100% while in the case of TST was between 80 and 60%. The first result clearly suggests BPA and AAP were exclusively removed by ozone direct reaction (the values of DR% obtained higher than 100%

Table 2 e Estimated hydroxyl radical concentration from organic removal in ozone-free UVA photocatalytic processes (CHO 3 1014, M).a System

pH 3 t ¼ 5 min

System

t ¼ 30 min

t ¼ 60 min

t ¼ 5 min

BPA TST AAP BPA TST AAP BPA TST AAP Fe(III) TiO2 Ferrioxalate Photo-Fenton TiO2/Fe(III)

1.65 0.84 5.08 4.65 1.89

1.68 0.63 5.8 4.64 2.15

1.31 0.77 2.99 4.26 1.96

0.99 1.26 0.68 0.61 6.73 13.7 6.72 6.51 2.26 2.49

0.89 0.76 0.57 0.54 5.23 9.00 6.31 10.2 2.19 2.79

0.97 0.53 e 7.93 3.16

0.70 0.48 4.42 8.59 2.36

pH 6.5 t ¼ 30 min

t ¼ 60 min

BPA TST AAP BPA TST AAP BPA TST AAP Ferrioxalate Ferricitrate TiO2 TiO2/Ferrioxalate TiO2/Ferricitrate

1.10 3.03 0.82 0.85 2.46

0.82 2.49 0.58 0.58 1.87

0.90 2.48 0.83 0.75 1.81

0.30 3.41 1.05 0.30 2.43

0.20 2.51 0.64 0.23 1.81

0.25 2.64 0.90 0.25 1.80

0.17 3.97 1.28 0.20 2.46

0.11 2.76 0.69 0.16 1.77

0.15 2.98 0.95 0.16 1.89

a All systems with UVA black light radiation. Results from equation (19). Values of kHO for BPA and AAP are 1.1  1010 and 9.8  109 M1 s1, respectively (Bisby and Tabassum, 1988; Gozmen et al., 2003). The value of kHO for TST-OH reaction, 1.5  1010 M1 s1 was determined in this work as detailed in Section 3.3.1.

164

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6

Table 3 e Estimated hydroxyl radical concentration from TOC results in different UVA POP (CHO 3 1014, M).a System

O3/Fe(III) O3/TiO2 O3/H2O2 O3/Fe(III)/TiO2 Photo-Fenton O3/Photo-Fenton Ferrioxalate O3/ferrioxalate

pH 3

System

t ¼ 15 min

t ¼ 60 min

t ¼ 120 min

4.14 5.14 1.30 7.74 3.04 4.62 5.78 5.45

5.71 6.65 1.34 3.97 7.04 3.08 2.06 8.61

3.60 2.28 3.83

O3/TiO2 O3/H2O2 O3/Ferrioxalate O3/TiO2/Ferrioxalate Ferricitrate O3/ferricitrate TiO2/ferricitrate O3/TiO2/ferricitrate

e 4.69 2.11 1.04 1.75

pH 6.5 t ¼ 15 min

t ¼ 60 min

t ¼ 120 min

2.71 1.52 3.91 3.57 0 1.31 0.29 1.46

2.91 1.71 2.04 4.31 2.14 2.73 0.25 3.12

2.91 2.25 0.58 5.54 0.32 1.93 0.22 4.17

a All systems with UVA black light radiation. Results from Equation (19) with TOC instead of CM. kHO ¼ 5  109 M1 s1.

can be attributed to some deviations of the reported data on kD from the actual ones). In the case of TST, contribution of free radicals is evident since values of DR% agree the results commented while discussing the organic concentration evolution (see Fig. 2).

3.3.2.

Kinetics of TOC oxidation

For any POP studied, once the initial compounds are eliminated hydroxyl radical oxidation is the main way of removal. Thus, if TOC is considered as lumped parameter of the organic concentration in water, the hydroxyl radical concentration can be estimated with equation (19), taken TOC as CM. In this case, for kHO a value of 5  109 M1 s1 was applied. This value is the average between 1010 and 5  107 M1 s1 values of rate constant for first reacting products, such as BPA, and typical end products of ozonation reactions, such as oxalic acid, respectively (Buxton et al., 1988). Thus, in Table 3 calculated values of CHO estimated from the modified Equation (19) are given. As it can be seen, in most of cases the order of magnitude was the same as in Table 2 for initial compounds (1014 M) with ozone processes showing the higher efficiency in producing hydroxyl radicals.

3.3.3. Kinetic considerations for low concentrations of pollutants In a practical case, real water contains concentrations of pollutants such as BPA, AAP and TST, as high as tenths of mg L1 so that it is important to make estimations of the contributions of ozone direct and hydroxyl radical reactions in this case. Since no experimental data was available Equation (19) cannot be applied but DR% can be determined from theoretical Equation (22): kD CO3  100 DR% ¼ kD CO3 þ kHO CHO

(22)

According to Equation (22) DR% is not dependent on pollutant concentration but on the concentration of hydroxyl radicals, dissolved ozone and rate constants of ozone direct and hydroxyl radical reactions. Also, from data of Table 2 or 3, it is deduced that concentration of hydroxyl radicals does not depend either on pollutant concentration but on the experimental conditions applied for ozone and catalyst concentrations and intensity of light, depending on the oxidation process. This particularly holds when dealing with pollutant concentrations of mg L1 level since any possible initiating,

promoting or scavenging effect of pollutants on hydroxyl radical formation would be negligible (Staehelin and Hoigne´, 1985). From Table 3 values of hydroxyl radical concentration between 1014 and 9  1014 M can then be used in equation (22). On the other hand, in water containing BPA, AAP and TST at mg L1 level, concentrations of dissolved ozone should be higher than those observed in this work just once concentrations of BPA, AAP and TST were below and close to the detection limit of the analytical method used here (100 mg L1). These low concentrations were reached at 15e25 min reaction times depending on the ozonation system as seen in Figs. 2, 6 and 8 and 12 (see also Figures 2SI, 3SI, 5SI, etc, of supplementary information). Thus, concentrations of ozone between 106 (0.05 mg L1) and 2  105 M (0.96 mg L1) can be taken as seen for example in Fig. 10. Then, with data of dissolved ozone and hydroxyl radical concentrations Equation (22) was used to estimate DR% for low pollutant concentrations. Application of Equation (22) leads to DR% values close to 100% in the case of BPA and AAP for all cases. Regarding TST, only when dissolved ozone concentration is  5  106 M (0.25 mg L1) and hydroxyl radical concentration is 1015 M, DR% is also close to 100%. In the rest of cases, contribution of hydroxyl radical oxidation varies from 70 to 20%. As a consequence, for removal of mg L1 concentrations of BPA, AAP and TST from water, photocatalytic ozonation systems are mainly recommended in the case of TST.

4.

Conclusions

Major conclusions reached in this study are: For homogeneous POP:  The basic Fe(III) photocatalytic oxidation only allows partial removal of organics during 2 h reaction. Addition of oxalic acid or citric acid, necessary at pH 6.5 to avoid Fe(III) precipitation, or hydrogen peroxide (photo-Fenton process) can lead to total removal but more than 60 min are needed. On the contrary, ozonation processes lead to total removal of organics in less than 20 min.  Ozonation alone and Fe(III)/UVA leads to very low TOC removal. At pH 3 photo-Fenton, ozone/Fe(III)/UVA and ozone/ferrioxalate/UVA systems lead to the highest TOC

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6

removal in 2 h (about 80%). TOC reductions are higher in POP at pH 3, especially in Fe based photocatalytic processes.  Total polyphenols are only completely removed at a significant rate in ozonation systems. In ozone-free POP, total polyphenol concentration remains constant or even increase except in the photo-Fenton process where some decrease in polyphenol concentration is also observed.  A synergic effect between ozone and photocatalytic oxidation is clearly noticed, especially for TOC removal. For heterogeneous POP  At pH 6.5 citric acid is better than oxalic acid to remove the organic compounds in reactions initiated by the corresponding ferricarboxylate photolysis. Ozonation processes allow again total conversion of organic in less than 15 min.  Regarding TOC, basic photocatalytic ozonation (O3/TiO2/ UVA) and also O3/Fe(III)/carboxylic acid/UVA systems leads to the highest TOC reductions (about 90% at pH 3 and 80% at pH 6.5). However, TOC reduction is not a good parameter to measure the efficiency of carboxylic acid based Fe(III) photocatalytic oxidation because the TOC contributing fraction of added carboxylic acids.  As in homogeneous systems, feeding ozone to a basic photocatalytic oxidation yields significant increases of TOC removal due to the synergic effects between both processes. About mechanism and kinetic aspects:  Regardless of any type of ozonation process and pH, initial compounds, at 1e3 mg L1 to some hundreds mg L1, are mainly removed through direct ozone reactions, especially for the cases of BPA and AAP. Then, for the removal of initial organics no advanced oxidation is needed.  In ozone-free systems, hydroxyl radical reactions are the way of oxidation. Concentration of hydroxyl radicals are of the order of 1014 M, regardless of the type of process. However, the highest values, at pH 3, correspond to photo-Fenton and ferrioxalate UVA and TiO2/Fe(III)/UVA systems for homogeneous and heterogeneous processes, respectively, while, at pH 6.5, the best processes for generating hydroxyl radicals are ferricitrate/UVA systems with or without TiO2 for homogeneous and heterogeneous processes, respectively.  For TOC removal, however, advanced oxidation processes are needed, especially, ozone combined processes. In some ozone systems a modified photo-Fenton mechanism develops.  Ozone photocatalytic processes show the higher efficiency in producing hydroxyl radicals.  For concentrations of pollutants in the mg L1 level, photocatalytic ozonation is recommended for TST removal since ozonation alone would allow total removal of BPA and AAP.

Acknowledgments This work has been supported by the CICYT of Spain and the European Region Development Funds of the European Commission (Project CTQ2009/13459/C05/05).

165

Appendix. Supplementary data Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.10.038.

references

Abrahamson, H.B., Rezvani, A.B., Brushmiller, J.G., 1994. Photochemical and spectroscopic studies of complexes of iron(III) with citric acid and other carboxylic acids. Inorganica Chimica Acta 226, 117e127. Agustina, T.E., Ang, H.M., Vareek, V.K., 2005. A review of synergistic effect of photocatalysis and ozonation on wastewater treatment. Journal of Photochemistry and Photobiology 6, 264e273. ´ lvarez, P.M., Rodrı´guez, E.M., Ferna´ndez, G., Beltra´n, F.J., 2010. A Degradation of bisphenol A in water by Fe(III)/UVA and Fe(III)/ polycarboxylate/UVA photocatalysis. Water Science and Technology 61.11, 2717e2722. Andreozzi, R., Caprio, V., Marotta, V., Vogna, D., 2003. Paracetamol oxidation from aqueous solutions by means of ozonation and H2O2/UV system. Water Research 37, 993e1004. Bader, H., Hoigne´, J., 1981. Determination of ozone in water by the indigo method. Water Research 15, 449e456. Barron, E., Deborde, M., Rabouan, S., Mazellier, P., Legube, B., 2006. Kinetic and mechanistic investigations of progesterone reaction with ozone. Water Research 40, 2181e2189. Beltra´n, F.J., Aguinaco, A., Garcı´a-Araya, J.F., 2010. Kinetic modelling of TOC removal in the photocatalytic ozonation of diclofenac aqueous solutions. Applied Catalysis B: Environmental 100, 289e298. Beltra´n, F.J., 2004. Ozone Reaction Kinetics for Water and Wastewater Systems. Lewis Publ., Boca Raton, Florida. Bhatkhande, D.S., Pagarkar, V.G., Beenackers, A.C.M., 2001. Photocatalytic degradation for environmental applicationsa review. Journal of Chemical Technology and Biotechnology 77, 102e116. Bisby, R.H., Tabassum, N., 1988. Properties of the radicals formed by one-electron oxidation of acetaminophen - a pulse radiolysis study. Biochemical Pharmacology 37, 2731e2738. Brose´us, R., Vincent, S., Aboulfadl, K., Daneshvar, A., Sauve´, S., Barbeau, B., Pre´vost, M., 2009. Ozone oxidation of pharmaceuticals, endocrine disruptors and pesticides during drinking water treatment. Water Research 43, 4707e4717. Buxton, G.V., Greenstock, C.L., Helman, W.P., Ross, A.B., 1988. Critical review of data constants for reactions of hydrated electrons, hydrogen atoms and hydroxyl radicals (OH/O) in aqueous solution. Journal of Physical and Chemical Reference Data 17, 513e886. Daughton, C.H., Ternes, T.A., 2000. Special report: pharmaceuticals and personal care products in the environment: agents of subtle change? Environmental Health Perspectives 108, 598e602. Deborde, M., Rabouan, S.D., Duguet, J.P., Legube, B., 2005. Kinetics of aqueous ozone-induced oxidation of some endocrine disruptors. Environmental Science and Technology 39, 6086e6092. Decoret, C., Royer, J., Legube, B., Dore´, M., 1984. Experimental and theoretical studies of the mechanism of the initial attack of ozone on some aromatics in aqueous medium. Environmental Technology Letters 5, 207e218. Eisenberg, G.M., 1943. Colorimetric determination of hydrogen peroxide. Industrial and Engineering Chemistry 15, 327e328. Escher, B.I., Baumgartner, R., Koller, M., Treller, K., Lienert, J., McCardell, C.S., 2011. Environmental toxicology and risk

166

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6

assessment of pharmaceuticals from hospital wastewater. Water Research 45, 75e92. Faust, B.C., Hoigne´, J., 1990. Photolysis of Fe(III)-hydroxy complexes as sources of OH radicals in clouds, fog and rain. Atmospheric Environment 24A, 79e89. Faust, B.C., Zepp, R.G., 1993. Photochemistry of aqueus iron(III) polycarboxylate complexes- roles in the chemistry of atmospheric and surface water. Environmental Science and Technology 27, 2517e2522. Fenton, H.J.H., 1876. On a new reaction of tartaric acid. Chemical News 33, 190. Gozmen, B., Oturan, M.A., Oturan, N., Erbatur, O., 2003. Indirect electrochemical treatment of bisphenol A in water via electrochemically generated Fenton’s reagent. Environmental Science and Technology 37, 3716e3723. Hatchard, C.G., Parker, C.A., 1956. A new sensitive chemical actinometer. II. Potassiumferrioxalate as a standard actinometer. Proceedings of the Royal Society London Service A 235, 518e536. Hoigne´, J., Bader, H., 1983. Rate constants of the reactions of ozone with organic and inorganic compounds. II. Dissociating organic compounds. Water Research 17, 185e194. Kim, S.D., Cho, J., Kim, I.S., Vanderford, B.J., Snyder, S.A., 2007. Occurrence and removal of pharmaceuticals and endocrine disruptors in South Korean surface, drinking and waste waters. Water Research 41, 1013e1021. Kolpin, D.W., Furlong, E.T., Meyer, M.T., Thurman, E.M., Zaugg, S. D., Barber, L.B., Buxton, H.T., 2002. Pharmaceuticals, hormones, and other organic wastewater contaminants in U. S. streams, 1999e2000: a national reconnaissance. Environmental Science and Technology 36, 1202e1211. Legrini, O., Oliveros, E., Braun, A.M., 1993. Photochemical processes for water treatment. Chemical Reviews 93, 671e698. Leitzke, A., von Sonntag, C., 2009. Ozonolysis of unsaturated acids in aqueous solution: acrylic, methacrylic, maleic, fumaric and muconic acids. Ozone Science & Engineering 31, 301e308. Logager, T., Holcman, J., Sehested, K., Pedersen, T., 1992. Oxidation of ferrous ions by ozone in acidic solutions. Inorganic Chemistry 31, 3523e3529. Matthews, R.W., McEvoy, S.R., 1992. A comparison of 254 nm and 350 nm excitation of TiO2 in simple photocatalytic reactors. Journal of Photochemistry and Photobiology A Chemistry 66, 355e366. Mesta´nkova´, H., Mailhot, G., Jirkovsky´, J., Kry´sa, J., Bolte, M., 2005. Mechanistic approach of the combined (iron-TiO2) photocatalytic system for the degradation of pollutants in aqueous solution: an attempt of rationalisation. Applied Catalysis B: Environmetal 57, 257e265. Mvula, E., von Sonntag, C., 2003. Ozonolysis of phenols in aqueous solutions. Organic and Biomolecular Chemistry 1, 1749e1756. Nansheng, D., Feng, W., Fan, L., Mei, X., 1998. Ferric citrateinduced photodegradation of dyes in aqueous solutions. Chemosphere 36, 3101e3112. Nelson, E.D., Do, H., Lewis, R.S., Carr, S.A., 2011. Diurnal variability of pharmaceutical, personal care product, estrogen and alkylphenol concentrations in effluent from a tertiary wastewater treatment facility. Environmental Science & Technology 45, 1228e1234.

Quici, N., Morgada, G.E., Gettar, R.T., Bolte, M., Litter, M.I., 2007. Photocatalytic degradation of citric acid under different conditions: TiO2 heterogeneous photocatalysis against homogeneous photolytic processes promoted by Fe(III) and H2O2. Applied Catalysis B: Environmental 71, 117e124. Rodrı´guez, E.M., Nu´n˜ez, B., Ferna´ndez, G., Beltra´n, F.J., 2009a. Effects of some carboxylic acids on the Fe(III)/UVA photocatalytic oxidation of muconic acid in water. Applied Catalysis B: Environmental 89, 214e222. ´ lvarez, P.M., Rodrı´guez, E.M., Ferna´ndez, G., Ledesma, B., A Beltra´n, F.J., 2009b. Photocatalytic degradation of organics in water in the presence of iron oxides: influence of carboxylic acids. Applied Catalysis B: Environmental 92, 240e249. ´ lvarez, P.M., Herna´ndez, R., Rodrı´guez, E.M., Ferna´ndez, G., A Beltra´n, F.J., 2011. Photocatalytic degradation of organics in water in the presence of iron oxides: effects of pH and light source. Applied Catalysis B: Environmental 102, 572e583. Safarzadeh-Amiri, A., Bolton, J., Cater, S., 1997. Ferrioxalatemediated photodegradation of organic pollutants in contaminated water. Water Research 31, 787e798. Singleton, V.L., Rossi, J.A., 1965. Colorimetry of total phenolic with phosphomolybdic-phosphotungstic acid reagents. American Journal of Enology and Viticulture 16, 144e158. Stackelberg, P.E., Gibs, J., Furlong, E.T., Meyer, M.T., Zaugg, S.D., Lippincott, R.L., 2007. Efficiency of conventional drinkigwater-treatment processes in removal of pharmaceuticals and other organic compounds. Science of the Total Environment 377, 255e272. Staehelin, S., Hoigne´, J., 1982. Decomposition of ozone in water: rate of initiation by hydroxyde ions and hydrogen peroxide. Environmental Science & Technology 16, 676e681. Staehelin, S., Hoigne´, J., 1985. Decomposition of ozone in water in the presence of organic solutes acting as promoters and inhibitors of radical chain reactions. Environmental Science & Technology 19, 1206e1212. Stookey, L.L., 1970. Ferrozine- a new spectrophotometric reagent for iron. Analytical Chemistry 42, 779e781. Ternes, T.A., 1998. Occurrence of drugs in German sewage treatment plants and rivers. Chemosphere 32, 3245e3260. Turchi, C.S., Ollis, D.F., 1990. Photocatalytic degradation of organic water contaminants: mechanisms involving hydroxyl radical attack. Journal of Catalysis 122, 178e192. Wang, S., Shirahisi, F., Nakano, K., 2002. A synergistic effect of photocatalysis and ozonation on depcomposition of formic acid in an aqueous solution. Chemical Engineering Journal 87, 262e271. Zepp, R.G., Faust, B.C., Hoigne´, J., 1992. Hydroxyl radical formation in aqueous reactions (pH 3-8) of iron (II) with hydrogen peroxide: the photo-Fenton reaction. Environmental Science & Technology 26, 313e319. Zhang, Y., 2000. Light and iron(III)-induced oxidation of chromium (III) in the presence of organic acids and manganese (II) in simulated atmospheric water. Atmospheric Environment 34, 1633e1640. Zuo, Y., 1995. Kinetics of photochemical/chemical cycling of iron coupled with organic substances in cloud and fog droplets. Journal of Hazardous Materials 59, 3123e3130.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 6 7 e1 7 5

Available online at www.sciencedirect.com

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

Boron bioremoval by a newly isolated Chlorella sp. and its stimulation by growth stimulators Burcu Ertit Tas¸tan, Ergin Duygu, Go¨nu¨l Do¨nmez* Department of Biology, Faculty of Science, Ankara University, 06100 Bes¸evler, Ankara, Turkey

article info

abstract

Article history:

It has been well documented that excess concentrations of boron (B) causes toxic effects on

Received 3 August 2011

many of the environmental systems. Although Chlorella sp. has been studied to remove

Received in revised form

pollutants from water, its capacity to remove B has not been investigated yet. Boron

11 October 2011

removal levels of newly isolated Chlorella sp. were investigated in BG 11 media with

Accepted 19 October 2011

stimulators as triacontanol (TRIA) and/or sodium bicarbonate (NaHCO3) and without them,

Available online 28 October 2011

to test if they could increase the removal efficiency by increasing biomass. The assays were performed to determine the effect of different medial compositions, B concentrations, pH

Keywords:

and biomass concentrations onto removal efficiency. Boron removal was investigated at

Microalgae

5e10 mg/L range at pH 8 in different medial compositions and maximum removal yield

Boron

was found as 32.95% at 5.45 mg/L B in media with TRIA and NaHCO3. The effect of different

Bioremoval

pH values on the maximum removal yield was investigated at pH 5e9, and the optimum pH

Growth stimulators

was found again 8. The interactive effect of biomass concentration and B removal yield was

Wastewater treatment

also investigated at 0.386e1.061 g wet weight/L biomass. The highest removal yield was found as 38.03% at the highest biomass range. This study highlights the importance of using new isolate Chlorella sp. as a new biomaterial for B removal process of waters containing B. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

As known very well, boron (B) is one of the essential trace elements (Waggott, 1969) and its biological role has still been studied by a number of researchers (Lee et al., 2009; Sheng et al., 2009). In spite of its biologically important role in metabolism at its low concentrations (Frick, 1985; Wojcik et al., 2008; Lee et al., 2009), excess amounts of B is harmful and causes toxic effects (Davis et al., 2002; Gunes et al., 2006; Del-Campo Marı´n and Oron, 2007; Sasmaz and Obek, 2009). Boron is mainly released to the environment by discharged industrial wastewaters (Coughlin, 1998). Manufacturing facilities of heat resistant materials (Morioka et al., 2007), storage and distribution of solar energy systems (Abu-Hamed et al.,

2007), catalysts (Xu et al., 2009), ceramics (Christogerou et al., 2009) and glass (Crawford et al., 2007) can be presented as main examples of important B sources. On the other hand, Turkey has the largest B reserves (60%) in the world (Okay et al., 1985) and therefore, toxicity of B is come into more prominence. As Cervilla et al. (2009) attracted attention to the fact that B toxicity has become important especially in areas close to the Mediterranean Sea, where intensive agriculture has been developed. They evidenced that excess level of B in cultivated soils lead to B toxicity which caused inhibition of nitrate reduction and consequence increase in ammonium assimilation in tomato plants, accompanied with the loss of leaf biomass and disorders in organic nitrogen metabolism. Nable et al. (1997) in fact, previously stated that B-rich soils were

* Corresponding author. Tel.: þ90 312 212 67 20; fax: þ90 312 223 23 95. E-mail address: [email protected] (G. Do¨nmez). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.045

168

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 6 7 e1 7 5

causing B toxicity in the field where crop yields decreased. They also mentioned that various anthropogenic sources of excess B might increase amount of B in soil to the toxic levels for plants, such as wastes from surface mining, fly ash, and industrial chemicals, and the most important source was irrigation water. As a result once the B level is higher than required it is needed removing excess of B in order to decrease the effects of B toxicity. According to the literature, some different methods have been tried to reduce B pollution in environment, including adsorptioneflocculation (Chong et al., 2009; Kavak, 2009), elec¨ ztu¨rk tro coagulation (Yılmaz et al., 2008), reverse osmosis (O et al., 2008), precipitation (Itakura et al., 2005), ion-exchange (Okay et al., 1985), use of B-selective resins (Simonnot et al., 2000) and some biological materials, such as duckweeds (Sasmaz and Obek, 2009; Del-Campo Marı´n and Oron, 2007). On the other hand, a number of studies have focused on the use of microalgae in removal of several pollutants from the culture media or wastewaters (de-Bashan and Bashan, 2010; Karacakaya et al., 2009; El-Sheekh et al., 2005). Chlorella sp. is known as one of the most useful microalgae for different purposes. This genus has been investigated in numerous studies considering its high growth rates under effect of different conditions (Lee et al., 2002; Valderrama et al., 2002; Sung et al., 1999) and it was shown that Chlorella sp. could remove pollutants with a high capacity and in an efficient way compared to many other aquatic organisms (Ruangsomboon and Wongrat, 2006; Gonza´lez et al., 1997; Hanagata et al., 1992). Its capacity to remove B has not been investigated previously; therefore this study aims to fill this gap in literature. In B removal studies, investigators did not add any growth stimulators to the culture media. However, it might be possible to enhance the growth of microorganism by enrichment of media with adding them. Another purpose of the present study was to investigate the effects of some growth stimulators effecting B removal process and to see whether they would enhance the growth of Chlorella sp. and increase the efficiency of bioremoval process. One of the growth stimulator used current study is Triacontanol (TRIA) and the other one is sodium bicarbonate (NaHCO3). TRIA, a long chain 30-carbon primary alcohol, (C30H61OH) is a well known plant hormone and growth regulator (Ries and Houtz, 1983). Stimulatory effects of TRIA on the photosynthesis, growth and net bioproductivity of some green algae and cyanobacteria species have also been reported (Karacakaya et al., 2009; Houtz et al., 1985a, 1985b). On the other hand, carbon is fixed by microalgae and is produced biomass and energy. Wang et al. (2008) reviewed that microalgae can fix carbon dioxide (CO2) from soluble carbonate forms such as NaHCO3 and use it into photosynthesis. It was shown that inexpensive carbon source of NaHCO3 was affected to growth and chlorophyll concentrations of microalgae by the way dissolving into the culture media and replacing the atmospheric CO2 and it was a product resulting from CO2 capture with alkali method and was a suitable C source for the growth of Chlorella vulgaris (Wang et al., 2010). It may be feasible to increase the growth of Chlorella sp. by adding TRIA and NaHCO3 into media and the efficiency of B removal processes can be increased. The aim of this study was to examine the hypothesis that (i) to determine the best B removal conditions in detail by changing experimental parameters, (ii) to examine if the

biomass production and B removal capacity of Chlorella sp. could be increased by adding TRIA and NaHCO3 into medium at different parameters, and (iii) to check if there was a potential offer by Chlorella sp. as an effective and eco-friendly biomaterial for developing B removal procedure. According to our best knowledge this is the first report about B removal by a new isolated microalgae Chlorella sp., with proposed target.

2.

Materials and methods

2.1.

Isolation and culture conditions

The microalgal culture was isolated from water supply located in Sorgun, Yozgat, Turkey. Samples were spread on the Petri plates containing BG 11 medium (Rippka, 1988) with 50.000 penicillin and were incubated at 25  2  C under continuous illumination (cool-white fluorescent, 48 mmol/m2s (2400 l)). The pH of the growth medium was adjusted to 8 by adding diluted (0.01 M) and concentrated (1 M) sulfuric acid or sodium hydroxide solutions. Cells from microcolonies on these plates were isolated by micromanipulation. The microalgal cells were purified at aseptic conditions by streaking the cells repeatedly on the BG 11 medium agar plate. At the final step, the purified microalgal cells were transferred to liquid media. In order to validate the axecinity, these liquid cultures were also tested for bacterial contamination by plating on bacteriological media. A series of batch culture experiments in unshaken flasks illuminated by cool-white fluorescent lamps were carried out at 48 mmol/m2s (2400 lx) light intensity. The microalgal cultures were transferred into 100 mL BG 11 medium in 250 mL Erlenmeyer flasks and incubated at 30  C under continuous illumination for 20 days. For the experiments, to give an initial concentration of about 0.1 g/L dry weight biomass inoculated in culture media.

2.2.

PCR and sequencing

Whole cells from an exponentially growing culture of the isolate were used for 18S rRNA gene amplification. 18S rRNA region was amplified with primers forward F1: 50 WACCTGGTTGATCCTGCCAGT-30 and reverse R1798: 50 GATCCTTCYGCAGGTTCACCTAC-30 are used as described by Luo et al. (2006). PCR reaction is carried out in 50 mL reaction and reaction mix included 0.2 mM of each primer, 0.2 mM of each dNTP, 1.5 mM of MgCl2 and 30 ng of template DNA. Super-HotTaq Taq DNA polymerase (Bioron GmbH, Germany) is used in the amplification. Amplification by PCR technique was carried out by an initial denaturation at 95  C for 10 min, followed by 35 cycles of denaturation at 95  C for 45 s, annealing at 60  C for 45 s, and elongation at 72  C for 45 s, with a final extension of 72  C for 10 min. Four primers are used to sequence the amplified product: 2 forward primers, F371: 50 AGGGTTCGATTCCGGAG-30 and F1132:50 -GAAACTTAAAKGAATTG-30 and 2 reverse primers, R584: 50 -GWATTACCGCGGCKGCTG-30 and R1283: 50 -CGGCCATGCACCACC-30 . BigDye 3.1 (Applied Biosystems Inc, USA) chemistry and ABI 3130 genetic analyzer (Applied Biosystems Inc, USA) is used in sequencing.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 6 7 e1 7 5

2.3.

TRIA, NaHCO3 and boron solutions

Stock TRIA (96%, w/v; Aldrich) solution was prepared by dissolving 0.5 g of the chemical in chloroform. Sodium bicarbonate solution (Merck) was prepared by dissolving 17.2 g/L of the NaHCO3 in distilled water. Stock solution of B was prepared by dilution of boric acid (H3BO3) (Carlo Erba) to a final concentration of 10 g/L of B. Appropriate volumes of the stock solutions were added to the BG 11 media.

2.4. Effects of different media compositions on boron removal To determine the effects of the different media compositions on B removal, the microalgae were cultivated in media containing increasing concentrations of B (5, 7.5 and 10 mg/L) in; (1) BG 11 control medium without any contents; (2) BG 11 medium with 1 mg/L TRIA; (3) BG 11 medium with 34 mg/L NaHCO3 and (4) BG 11 medium with 1 mg/L TRIA and 34 mg/L

169

NaHCO3 solutions (Tas¸tan et al., unpublished results) at pH 8 for 20 days of incubation period. For the experiments, 100 ml BG 11 media was inoculated with 0.1 g/L dry weight biomass.

2.5.

Effect of initial pH on boron removal

In order to highlight the effect of pH on B removal process by Chlorella sp., the experiments were also conducted at pH 5, 6, 7, 8 and 9 at 10 mg/L B concentration in BG 11 media with 1 mg/L TRIA and 34 mg/L NaHCO3 for 15 days incubation period. For the experiments, 100 mL BG 11 media was inoculated with 0.1 g/L dry weight biomass.

2.6. Effects of the biomass concentration on boron removal The effect of microalgal biomass concentrations on B removal was also examined at four different biomass concentrations by the fresh wet weight method. The experiments were performed

Fig. 1 e Comparison of the effect of different media compositions on removal yield of different boron concentrations by Chlorella sp. during the incubation period. (TRIA concentration, 1 mg/L; NaHCO3 concentration, 34 mg/L; pH 8; T 25 ± 2  C; illumination, 48 mmol/m2s (2400 l3)).

170

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 6 7 e1 7 5

in media with 1 mg/L TRIA and 34 mg/L NaHCO3 at pH 8 for 20 days of incubation period. Four different fresh wet weights of Chlorella sp. biomass were added to each treatment to give an initial biomass concentration of 0.386, 0.450, 0.655 and 1.061 g/L.

2.7.

Analytical methods

During the incubation period, 3 mL samples were taken at 5, 10, 15 and 20 days from each of the flasks. The B concentration was determined by measuring the absorbance at 585 nm with a Shimadzu UV 2001 model spectrophotometer by using carmine as the complexing reagent (Adams, 1990). The percentage removal of B and qm (the maximum specific B uptake) was calculated from equations used before by other researches (Tas¸tan et al., 2010). In the study, qm represents the maximum amount of B removal per unit dry weight of microalgal cells (mg/g), X maximum dried cell mass (g/L), and C0 the initial concentration of the B (mg/L), respectively. Cell growth of Chlorella sp. was determined by measuring optic density, maximum dried cell mass and specific growth rate parameters for any set of growth conditions. Optic density was measured at 600 nm with the spectrophotometer. The maximum dried cell mass was saved by the measurement of the pellets, which were dried at 80  C for overnight (Nu¨ve FN 400 model sterilizator) after centrifugation step (3421 g ¼ 5000 rpm for 100 ). Specific growth rate (m) was calculated

according to the equation m ¼ (ln X2  ln X1)/(t2t1), X2 and X1: dry cell weight concentrations (g/L) at time t2 and t1, respectively (Ip and Chen, 2005). The chlorophyll (a þ b) concentrations were also determined for chlorophyll a at 646.6 nm and chlorophyll b at 663.6 nm and calculated with the method developed by Porra et al. (1989). The chlorophyll concentrations were expressed in mg of chlorophyll per milliliter. All of the experiments were performed in triplicate.

3.

Results and discussion

Pairwise distance analysis carried out by using Mega4 software (Tamura et al., 2004, 2007) revealed that our species is in a close relationship to microalgal species Chlorella zofingiensis at % 97.09, Chlorella ellipsoidea at % 95.76, Chlorella sorokiniana at % 95.62, C. vulgaris at % 95.55 and Chlorella lobophora at % 95.33 identity, for 18S rRNA. As a result of these findings new isolate was identified as Chlorella sp.

3.1. Effect of different media compositions and boron concentrations on removal It is important to detect an interactive effect between media contents and B removal in order to optimize the removal

Fig. 2 e Interactive effect of boron removal yield and optic density (OD600) of Chlorella sp. during the incubation period. (TRIA concentration, 1 mg/L; NaHCO3 concentration, 34 mg/L; pH 8; T 25 ± 2  C; illumination, 48 mmol/m2s (2400 l3)).

171

8.37 27.56  1.74 3.923  0.21 7.08 30.57  1.94 2.580  0.11 5.45 32.95  1.39 1.836  0.15 8.47 26.06  1.85 2.524  0.30 6.28 27.39  1.73 1.834  0.03 Co (mg/L) Removal (%) qm (mg/g)

5.03 23.36  1.31 1.773  0.29

7.05 20.66  0.55 1.694  0.08

7.32 19.88  1.33 1.564  0.02

5.66 24.13  0.83 1.481  0.29

6.64 28.53  0.88 2.147  0.03

6.98 24.24  0.93 2.172  0.13

5.55 29.31  0.28 2.153  0.19

BG 11 þ TRIA þ NaHCO3 BG 11 þ NaHCO3 BG 11 þ TRIA BG 11 without any content

process. Previous studies showed that the initial concentration of B was an important parameter in B bioremoval efficiency that was measured in the culture media (Del-Campo Marı´n and Oron, 2007). Therefore, in the first series of the experiments this parameter was tested at three different B concentrations (5, 7.5 and 10 mg/L), which were designed also for measurement of the effect of different media compositions. In our previous study, we tested microalgal biomass amount was more remarkable in media with TRIA and/or NaHCO3 when compared with biomass amount in media without them (Tas¸tan et al., unpublished results). The aim was to test if TRIA and/or NaHCO3 could increase the removal efficiency by increasing biomass. Therefore, four different BG 11 media compositions were investigated for this purpose: (i) BG 11 without any content (control samples), (ii) BG 11 þ TRIA, (iii) BG 11 þ NaHCO3, (iv) BG 11 þ TRIA þ NaHCO3. The results presented in Fig. 1. show that the percentage of B removal decreased when initial B concentrations were increased from 5 to 10 mg/L. As presented in the Fig. 1aed, all of the removal yields showed significant increases only after 10 days of the incubation period. The removal capacity of Chlorella sp. in BG 11 media without any content is of significance important to comparison its efficiency in media with presence of TRIA and/or NaHCO3. As it can be seen clearly in Fig. 1a, Chlorella sp. removed 13.04% B at 5.03 mg/L concentration within 10 days of incubation period in media without any content. The maximum B removal was observed at the same concentration, i.e., 23.36% within 20 days. The removal capacity of Chlorella sp. was also not affected significantly by TRIA plus in media (Fig. 1b). For example, the highest B removal yield was 24.13% at 5.66 mg/L B in the media with TRIA. These results can be interpreted as a sign of ineffectiveness of TRIA when applied alone in low B concentrations. Boron removal capacity of Chlorella sp. in media with TRIA was higher than yields found in media without TRIA, at higher B concentrations. For example, the removal yield was 19.88% at 7.32 mg/ L B concentration in control media and it increased to 24.24% at 6.98 mg/L B concentration, when TRIA added to media. There are some of the other studies in the literature, describing removal of industrial pollutants in an effective way by including TRIA hormone in BG 11 culture media by freshwater cyanobacterium Synechocystis sp. (Karacakaya et al., 2009). The removing yield was about 22.0% at 58.5 mg/L Remazol Blue dye in media without TRIA and increased to about 25.0% in media with 10 mg/L TRIA. They have explained that increasing in dye removal to be due to TRIA growth hormone. Another observation was the considerable increase in B removal capacity of Chlorella sp. when NaHCO3 added into media. The maximum B removal yield was 29.31% at 5.55 mg/L B concentration within 20 days of incubation period. The removal capacity of Chlorella sp. in media with NaHCO3 was also higher than media with TRIA at higher B concentrations (Fig. 1c). On the other hand, another result, which is more important than above all, was obtained by adding of TRIA and NaHCO3 into culture media together where, enhanced the growth and removal capacity of Chlorella sp. As it can be seen clearly in Fig. 1d, TRIA stimulated all of the parameters

Table 1 e Comparison of the removal yields and qm values at different boron concentrations in different media compositions by Chlorella sp. (Co, boron concentrations; qm, the maximum specific boron uptake; TRIA concentration, 1 mg/L; NaHCO3 concentration, 34 mg/L; incubation period, 20 days; pH 8; T, 25 ± 2  C; illumination, 48 mmol/m2s (2400 l3)).

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 6 7 e1 7 5

172

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 6 7 e1 7 5

Table 2 e Effect of pH on boron removal and maximum specific boron uptake (qm) of Chlorella sp. (Boron concentration, 10 mg/L; TRIA concentration, 1 mg/L; NaHCO3 concentration, 34 mg/L; incubation period, 15 days; T, 25 ± 2  C; illumination, 48 mmol/m2s (2400 l3)). pH Removal (%) qm (mg/g)

5

6

7

8

9

13.64  0.50 2.429  0.188

15.24  1.33 2.640  0.281

18.71  0.78 2.702  0.215

23.06  2.78 3.268  0.196

13.50  0.43 1.592  0.074

measured in the presence of NaHCO3. Oxidation of the HCO3 to CO2 and its assimilation was increased the primary production in terms of dry weight. Therefore, microorganism tolerated increasing B concentrations easier than all other media contents. Increasing B concentrations up to 8.37 mg/L did not affect the efficiency in terms of yield, and the maximum B removal yield was 32.95% at 5.45 mg/L and was 27.56% at 8.37 mg/L B concentrations within 20 days of incubation period. For these results, further experiments were performed in media with TRIA and NaHCO3. The findings reported by Del-Campo Marı´n and Oron (2007), who used duckweed Lemna gibba in their study, bioremoval of B by L. gibba decreased when its initial concentrations in the media were increased, and the most efficient removal measured at below 2 mg/L B concentrations. They added that there was no removal at about 10 mg/L B concentration even on 12th day of cultivation. In the present study Chlorella sp. showed significant removal efficiency at 10 mg/L B concentration at the four different media compositions tested at 10th days of incubation period. These data demonstrate Chlorella sp. is a good bioaccumulator when compared other aquatic microorganisms. Fig. 2. gives more information about microalgal growth, in order to confirm the stimulation effect of TRIA and/or NaHCO3 and the relationship between algal growth and B removal. As seen in Fig. 2a, Chlorella sp. showed its minimum growth at 5 mg/L B concentration in media without any content (OD600 1.692 at 20th days). The growth of microorganism was increased by adding TRIA and/or NaHCO3 in media. The OD600 was 1.398, 1.484 and 1.693 at 0 mg/L B concentration at 15th days in media with TRIA, media with NaHCO3 and media with TRIA and/or NaHCO3, respectively. When TRIA and NaHCO3 added to media, indicating the presence of stimulators, Chlorella sp. was less affected at 5 mg/L B concentrations due to high growth rates. The comparisons of the maximum amount of B removal per unit dry weight of microalgal cells (qm) is shown in Table 1. In general, the maximum specific B removal values increased with increasing B concentrations up to a certain level due to stimulators. At the lowest B concentration (5 mg/L), the maximum B amount per unit dry weight of Chlorella sp. was 1.773, 1.481, 2.153 and 1.836 mg/g in media without any contents, media with TRIA, media with NaHCO3 and media with TRIA and NaHCO3, respectively. Comparison of qm values for B removal in media with TRIA and/or NaHCO3 and without them showed that there were slight difference, indicating the presence of hormonal stimulation. When B concentration increased to 7.32 mg/L, qm also decreased to 1.564 mg/g in media without any content. The qm values in media with TRIA were found higher than media without TRIA at higher B concentrations due to higher removal yields. In media with

TRIA and NaHCO3, qm increased nearly 2 times (3.923 mg/g) and considerably higher than other media experiments at 8.37 mg/L B concentration. The reason why the qm values in media with TRIA and NaHCO3 were higher than in media without them was due to the much higher growth rates and higher removal yields of Chlorella sp. The media with TRIA and NaHCO3 was the selected media for further experiments because of the results of above mentioned.

3.2.

Effect of initial pH on boron removal

To find a suitable pH for effective B removal by Chlorella sp., trials were performed at media with five different initial pH values (5e9) including nearly 10 mg/L B, and with TRIA and NaHCO3. The effect of pH on B removal after incubation for 15 days is exhibited in Table 2. As shown in the table, B removal was 13.64%, 15.24%, and 18.71% at pH 5, 6 and 7 respectively. The maximum specific B uptakes were 2.429, 2.640 and 2.702 mg/g at the same pH values, respectively. It was observed that the removal yield of B and maximum specific B uptake increased with an increase in pH values up to pH 8. Chlorella sp. removed B with the highest yield of 23.06% and maximum specific B uptake was 3.268 mg/g at pH 8. Further increase in pH to 9 resulted in decrease percentage of removal yield (13.50%) and maximum specific B uptake (1.592 mg/g). As these experiments demonstrated that Chlorella sp. showed its highest B removal yields at pH 8, the following experiments were conducted at this pH value. Some researchers also marked basic pH values for efficient B removal by the macroalgae Caulerpa racemosa var. cylindracea (Bursali et al., 2009) and by the electro coagulation in a batch reactor (Yılmaz et al., 2008).

Fig. 3 e Effect of different biomass concentrations on boron removal by Chlorella sp. at the end of 20 days. (Boron concentration, 10 mg/L; TRIA concentration, 1 mg/L; NaHCO3 concentration, 34 mg/L; pH 8; T 25 ± 2  C; illumination, 48 mmol/m2s (2400 l3)).

173

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 6 7 e1 7 5

Table 3 e Results of the removal yields, qm values, chl (aþb) (mg/mL) and m (1/d) at different biomass concentrations (g wet weight/L) by Chlorella sp. (Co, boron concentrations; qm, the maximum specific boron uptake; m, specific growth rate; TRIA concentration, 1 mg/L; NaHCO3 concentration, 34 mg/L; incubation period, 20 days; pH 8; T, 25 ± 2  C; illumination, 48 mmol/ m2s (2400 l3)). Biomass concentration (g wet weight/L) 0.386 0.450 0.655 1.061

3.3.

Co (mg/L) 8.30 8.37 8.93 9.19

qm (mg/g)

Removal (%) 19.59 27.56 35.52 38.03

 1.23  1.74  2.36  3.83

Effect of biomass concentrations on boron removal

It was necessary to also determine the effect of biomass concentration for effective B removal. To find a suitable biomass concentration of Chlorella sp., experiments were performed at 10 mg/L B concentrations, in selected optimum media (BG 11 þ TRIA þ NaHCO3) at selected optimum pH level (8), with four different initial biomass concentrations (0.386e1.061 g wet weight/L). It is evident from Fig. 3. that B removal was really linked to the amount of biomass concentrations. The B removal yield was 4.48% at 0.386 g wet weight/L biomass concentration at 5th days of incubation period, and increased to 19.59% at the end of 20 days. When biomass concentration was increased from 0.450 to 0.655 g wet weight/L, the B removal yield was increased from 27.56 to 35.52%. Although 35.52% of 8.93 mg/L B was removed at 0.655 g wet weight/L biomass within 20 days, same removal yield (35.08%) was achieved at 9.19 mg/L B concentration within 15 days by increasing biomass concentration to 1.061 g wet weight/L. It also took only 5 days to remove 9.19 mg/L B with a yield of 18.11% at highest biomass concentration, instead of 20 days to remove 8.30 mg/L B with a yield of 19.59% at lowest biomass concentrations. The highest B removal yield was 38.03% at 1.061 g wet weight/L biomass concentration within 20 days. It is clearly seen in Fig. 3 that the relationship between these variables was linear. This linearity can be interpreted as the indication of the relationship between higher biomass concentrations and higher tolerance capability to B by Chlorella sp. The maximum amount of B removal per unit dry weight of microalgal cells (qm), chlorophyll (a þ b) concentrations and specific growth rate (m) at increasing biomass concentrations after 20 days of incubation period has also been determined and presented in Table 3. The effect of biomass concentrations on qm showed that the microalgal uptake capacity and removal yield of B at 0.450 g wet weight/L biomass concentration was higher than 0. 386 g wet weight/L. The qm was 3.268 mg/g at 8.30 mg/L B concentration at 0.386 g wet weight/L biomass concentration and increased to 4.204 mg/g when the initial biomass concentration increased up to 0.655 g wet weight/L. Although B removal yield was highest (38.03%) at the highest biomass concentration (1.061 g wet weight/L), qm value was lowest (3.174 mg/g) related to the high amount of biomass. When biomass had high amount, the removal of B per 1 g of the dry weight of the microalgae would have low value. The data given in Table 3 also shows the changes in chlorophyll (a þ b) concentrations under effect of a known B

3.268 3.923 4.204 3.174

Chl (a þ b) (mg/mL)

 0.66  0.21  0.73  0.13

0.382 0.720 0.925 1.349

 0.040  0.034  0.143  0.087

m (1/d) 0.088  0.0046 0.106  0.0023 0.095  0.0092 0.094  0.0035

concentration. The chlorophyll (a þ b) concentrations increased from 0.382 to 1.349 mg/mL in parallel to the increasing biomass concentrations. There was also a little difference between the specific growth rates; lowest m was 0.088 1/d at 0.386 g wet weight/L biomass concentration. As expected, when biomass had fewer amounts at a known concentration of B, specific growth rate would have low. On the other hand the maximum specific growth rate was 0.106 1/d at 0.450 g wet weight/L biomass concentration and it decreased to 0.095 1/d at 0.655 g wet weight/L biomass concentration. The specific growth rates of biomass concentrations between 0.655 and 1.061 g wet weight/L were very similar. But, when compared all of the biomass concentrations, and take into consideration of standard declinations, the initial biomass concentration of 0.655 g wet weight/L was the most suitable one. This biomass concentration had higher B removal yield and specific growth rate with a highest qm value. The reason why highest biomass concentration was not the best one due to the already in place much higher biomass of microalgae in media would not have an excellent growth rate.

4.

Conclusions

The present study describes a remarkable topic that is very popular and hard to study around the world. Boron removal in aqueous solutions by using efficient and economical methods is still a challenging problem. Our effort is contribution to the studies finding a sufficiently economical and efficient method which can be taken as a promising alternative technique offering an applicable solution to the problem. Compared to current physical or chemical B removal techniques, microalgal B removal process can be supported as an easily and is more environment friendly method. The ultimate aim of this study was to investigate if Chlorella sp. could serve as a biomaterial for B removal from water and its potential could be increased by the stimulatory effect of some growth stimulating agents. Increasing biomass concentrations in the presence of TRIA hormone and NaHCO3 at pH 8 resulted more prominent effects on B removal process in our experiments. The maximum B removal yield measured was 38.03% at 9.19 mg/L B concentration in the present study which was the highest yield that was obtained with the use of aquatic organisms, to the best of our knowledge. The results showed that Chlorella sp. can be used as a potential bioaccumulator for B removal process up to high B concentrations. Reports on the use of biomaterial for B removal are limited. Some investigators in fact, have used Chlorella sp. as

174

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 6 7 e1 7 5

a biomaterial for removal of some pollutants (Lim et al., 2010; Valderrama et al., 2002), but unfortunately it has not been used in such studies for B bioremoval. The present study not is only the first report that offered Chlorella sp. as an effective biomaterial, but also detected the most efficient way on B removal process, comprehensively.

Acknowledgments Financial support was gratefully acknowledged by the Scien_ ¨ BITAKtific and Technological Research Council of Turkey (TU _ BIDEB). The authors also wish to express their gratitude to reviewers for their valuable comments.

references

Abu-Hamed, T., Karni, J., Epstein, M., 2007. The use of boron for thermochemical storage and distribution of solar energy. Solar Energy 81, 93e101. Adams, V.D., 1990. Water and Wastewater Examination Manual, vol. 247. CRC Press LLC, Florida. Bursali, E.A., Cavas, L., Seki, Y., Bozkurt, S.S., Yurdakoc, M., 2009. Sorption of boron by invasive marine seaweed: Caulerpa racemosa var. cylindracea. Chemical Engineering Journal 150, 385e390. Cervilla, L.M., Blasco, B., Rı´os, J.J., Rosales, M.A., RubioWilhelmi, M.M., Sa´nchez-Rodrı´guez, E., Romero, L., Ruiz, J.M., 2009. Response of nitrogen metabolism to boron toxicity in tomato plants. Plant Biology 11, 671e677. doi:10.1111/j.14388677.2008.00167.x. Chong, M.F., Lee, K.P., Chieng, H.J., Ramli, I.I.S.B., 2009. Removal of boron from ceramic industry wastewater by adsorptione flocculation mechanism using palm oil mill boiler (POMB) bottom ash and polymer. Water Research 43, 3326e3334. Christogerou, A., Kavas, T., Pontikes, Y., Koyas, S., Tabak, Y., Angelopoulos, G.N., 2009. Use of boron wastes in the production of heavy clay ceramics. Ceramics International 35, 447e452. Coughlin, J.R., 1998. Sources of human exposure: overview of water supplies as sources of boron. Biological Trace Element Research 66, 87e100. Crawford, C.L., Marra, J.C., Bibler, N.E., 2007. Glass fabrication and product consistency testing of lanthanide borosilicate glass for plutonium disposition. Journal of Alloys and Compounds 444e445, 569e579. Davis, S.M., Drake, K.D., Maier, K.J., 2002. Toxicity of boron to the duckweed, Spirodella polyrrhiza. Chemosphere 48, 615e620. de-Bashan, L.E., Bashan, Y., 2010. Immobilized microalgae for removing pollutants: review of practical aspects. Bioresource Technology 101, 1611e1627. Del-Campo Marı´n, C.M., Oron, G., 2007. Boron removal by the duckweed Lemna gibba: a potential method for the remediation of boron-polluted waters. Water Research 41, 4579e4584. El-Sheekh, M.M., El-Shouny, W.A., Osman, M.E.H., El-Gammal, E. W.E., 2005. Growth and heavy metals removal efficiency of Nostoc muscorum and Anabaena subcylindrica in sewage and industrial wastewater effluents. Environmental Toxicology and Pharmacology 19, 357e365. Frick, H., 1985. Boron tolerance and accumulation in the duckweed, Lemna minor. Journal of Plant Nutrition 8, 1123e1129.

Gonza´lez, 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. Bioresource Technology 60, 259e262. Gunes, A., Soylemezoglu, G., Inal, A., Bagci, E.G., Coban, S., Sahin, O., 2006. Antioxidant and stomatal responses of grapevine (Vitis vinifera L.) to boron toxicity. Scientia Horticulturae 110, 279e284. Hanagata, N., Takeuchi, T., Fukuju, Y., Barnes, D.J., Karube, I., 1992. Tolerance of microalgae to high CO2 and high temperature. Phytochemistry 31, 3345e3348. Houtz, R.L., Ries, S.K., Tolbert, N.E., 1985a. Effect of triacontanol on Chlamydomonas’. Plant Physiology 79, 357e364. Houtz, R.L., Ries, S.K., Tolbert, N.E., 1985b. Effect of triacontanol on Chlamydomonas: II. Specific activity of ribulosebisphosphate carboxylase/oxygenase, ribulose-bisphosphate concentration, and characteristics of photorespiration. Plant Physiology 79, 365e370. Ip, P.-F., Chen, F., 2005. Production of astaxanthin by the green microalga Chlorella zofingiensis in the dark. Process Biochemistry 40, 733e738. Itakura, T., Sasai, R., Itoh, H., 2005. Precipitation recovery of boron from wastewater by hydrothermal mineralization. Water Research 39, 2543e2548. Karacakaya, P., Kılıc¸, N.K., Duygu, E., Do¨nmez, G., 2009. Stimulation of reactive dye removal by cyanobacteria in media containing triacontanol hormone. Journal of Hazardous Materials 172, 1635e1639. Kavak, D., 2009. Removal of boron from aqueous solutions by batch adsorption on calcined alunite using experimental design. Journal of Hazardous Materials 163, 308e314. Lee, J.-S., Kim, D.-K., Lee, J.-P., Park, S.-C., Koh, J.-H., Cho, H.-S., Kim S-, W., 2002. Effects of SO2 and NO on growth of Chlorella sp. KR-1. Bioresource Technology 82, 1e4. Lee, S.H., Kim, W.S., Han, T.H., 2009. Effects of post-harvest foliar boron and calcium applications on subsequent season’s pollen germination and pollen tube growth of pear (Pyrus pyrifolia). Scientia Horticulturae 122, 77e82. Lim, S.-L., Chu, W.-L., Phang, S.-M., 2010. Use of Chlorella vulgaris for bioremediation of textile wastewater. Bioresource Technology 101, 7314e7322. Luo, W., Pflugmacher, S., Pro¨schold, T., Walz, N., Krienitz, L., 2006. Genotype versus phenotype variability in Chlorella and Micractinium (Chlorophyta, Trebouxiophyceae). Protist 157, 315e333. Morioka, A., Sakurai, S., Okuno, K., Sato, S., Verzirov, Y., Kaminaga, A., Nishitani, T., Tamai, H., Kudo, Y., Yoshida, S., Matsukawa, M., 2007. Development of 300  C heat resistant boron-loaded resin for neutron shielding. Journal of Nuclear Materials 367-370, 1085e1089. Nable, R.O., Ban˜uelos, G.S., Paull, J.G., 1997. Boron toxicity. Plant and Soil 193, 181e198. Okay, O., Gu¨c¸lu¨, H., Soner, E., Balkas¸, T., 1985. Boron pollution in the Simav River, Turkey and various methods of boron removal. Water Research 19, 857e862. ¨ ztu¨rk, N., Kavak, D., Ko¨se, T.E., 2008. Boron removal from O aqueous solution by reverse osmosis. Desalination 223, 1e9. Porra, R.J., Thompson, W.A., Kreidemann, P.E., 1989. Determination of accurate extinction coefficients and simultaneous equations for assaying chlorophylls a and b extracted with four different solvents: verification of the concentration of chlorophyll standards by atomic absorption spectroscopy. Biochimica et Biophysica Acta 975, 384e394. Ries, S.K., Houtz, R., 1983. Triacontanol as a plant growth regulator. Horticultural Science 18, 654e662. Rippka, R., 1988. Recognition and identification of cyanobacteria. Methods in Enzymology 167, 28e67. Ruangsomboon, S., Wongrat, L., 2006. Bioaccumulation of cadmium in an experimental aquatic food chain involving

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 6 7 e1 7 5

phytoplankton (Chlorella vulgaris), zooplankton (Moina macrocopa), and the predatory catfish Clarias macrocephalus  C. gariepinus. Aquatic Toxicology 78, 15e20. Sasmaz, A., Obek, E., 2009. The accumulation of arsenic, uranium, and boron in Lemna gibba L. exposed to secondary effluents. Ecological Engineering 35, 1564e1567. Sheng, O., Song, S., Peng, S., Deng, X., 2009. The effects of low boron on growth, gas exchange, boron concentration and distribution of ‘Newhall’ navel orange (Citrus sinensis Osb.) plants grafted on two rootstocks. Scientia Horticulturae 121, 278e283. Simonnot, M.O., Castel, C., Nicolaı¨, M., Rosin, C., Sardin, M., Jauffret, H., 2000. Boron removal from drinking water with a boron selective resin: is the treatment really selective? Water Research 34, 109e116. Sung, K.-D., Lee, J.-S., Shin, C.-S., Park, S.-C., Choi, M.-J., 1999. CO2 fixation by Chlorella sp. KR-1 and its cultural characteristics. Bioresource Technology 68, 269e273. Tamura, K., Nei, M., Kumar, S., 2004. Prospects for inferring very large phylogenies by using the neighbor-joining method. Proceedings of the National Academy of Sciences (USA) 101, 11030e11035. Tamura, K., Dudley, J., Nei, M., Kumar, S., 2007. MEGA4: molecular evolutionary genetics analysis (MEGA) software version 4.0. Molecular Biology and Evolution 24, 1596e1599.  rul, S., Do¨nmez, G., 2010. Effective bioremoval Tas¸tan, B.E., Ertug of reactive dye and heavy metals by Aspergillus versicolor. Bioresource Technology 101, 870e876.

175

Valderrama, L.T., Del Campo, C.M., Rodriguez, C.M., de- Bashan, L.E., Bashan, Y., 2002. Treatment of recalcitrant wastewater from ethanol and citric acid production using the microalga Chlorella vulgaris and the macrophyte Lemna minuscula. Water Research 36, 4185e4192. Waggott, A., 1969. An investigation of the potential problem of increasing boron concentrations in rivers and water courses. Water Research 3, 749e765. Wang, B., Li, Y., Wu, N., Lan, C.Q., 2008. CO2 bio-mitigation using microalgae. Applied Microbiology and Biotechnology 79, 707e718. Wang, H.-M., Pan, J.-L., Chen, C.-Y., Chiu, C.-C., Yang, M.-H., Chang, H.-W., Chang, J.-S., 2010. Identification of anti-lung cancer extract from Chlorella vulgaris C-C by antioxidant property using supercritical carbon dioxide extraction. Process Biochemistry 45, 1865e1872. Wojcik, P., Wojcik, M., Klamkowski, K., 2008. Response of apple trees to boron fertilization under conditions of low soil boron availability. Scientia Horticulturae 116, 58e64. Xu, J., Chen, L., Tan, K.F., Borgna, A., Saeys, M., 2009. Effect of boron on the stability of Ni catalysts during steam methane reforming. Journal of Catalysis 261, 158e165.  lu, R., Kocakerim, M.M., Kocadag  istan, E., Yılmaz, A.E., Boncukcuog 2008. An empirical model for kinetics of boron removal from boron-containing wastewaters by the electrocoagulation method in a batch reactor. Desalination 230, 288e297.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 7 6 e1 8 6

Available online at www.sciencedirect.com

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

Survival dynamics of fecal bacteria in ponds in agricultural watersheds of the Piedmont and Coastal Plain of Georgia Michael B. Jenkins a,*, Dinku M. Endale a, Dwight S. Fisher a, M. Paige Adams b, Richard Lowrance c, G. Larry Newton d, George Vellidis b a

USDA-ARS J. Phil Campbell, Sr., Natural Resource Conservation Center, Watkinsville, GA 30677, USA Department of Biological and Agricultural Engineering, University of Georgia, Tifton, GA 31793, USA c USDA-ARS Southeast Watershed Research Laboratory, Tifton, GA, USA d Department of Animal Science, University of Georgia, Tifton, GA 31793, USA b

article info

abstract

Article history:

Animal agriculture in watersheds produces manure bacteria that may contaminate surface

Received 14 July 2011

waters and put public health at risk. We measured fecal indicator bacteria (commensal

Received in revised form

Escherichia coli and fecal enterococci) and manure pathogens (Salmonella and E. coli 0157:H7),

18 October 2011

and physicalechemical parameters in pond inflow, within pond, pond outflow, and pond

Accepted 21 October 2011

sediments in three ponds in agricultural watersheds. Bishop Pond with perennial inflow

Available online 2 November 2011

and outflow is located in the Piedmont, and Ponds A and C with ephemeral inflow and outflow in the Coastal Plain of Georgia. Bromide and chloride tracer experiments at Bishop

Keywords:

Pond reflected a residence time much greater than that estimated by two models, and

E. coli 0157:H7

indicated that complete mixing within Bishop Pond was never obtained. The long resi-

Fecal indicator bacteria

dence time meant that fecal bacteria were exposed to solar UV-radiation and microbial

Natural disinfection

predation. At Bishop Pond outflow concentrations of fecal indicator bacteria were signifi-

Ponds

cantly less than inflow concentrations; such was not observed at Ponds A and C. Both

Salmonella

Salmonella and E. coli 0157:H7 were measured when concomitant concentrations of

Watersheds

commensal E. coli were below the criterion for surface water impairment indicating problems with the effectiveness of indicator organisms. Bishop Pond improved down stream water quality; whereas, Ponds A and C with ephemeral inflow and outflow and possibly greater nutrient concentrations within the two ponds appeared to be less effective in improving down stream water quality. Published by Elsevier Ltd.

1.

Introduction

Watersheds with dairies, beef cattle, swine, and poultry operations together with wildlife and watersheds with crops receiving manure as a soil amendment are potential nonpoint sources of fecal bacteria and zoonotic pathogens such as Salmonella and Escherichia coli 0157:H7 (Ferguson et al., 2003). Salmonella is known to cause gastroenteritis and is a leading

cause of food related deaths (Mead et al., 1999). Several outbreaks of haemorrhagic colitis and haemolytic uremic syndrome have been caused by E. coli 0157:H7 infections traced to surface water (Olsen et al., 2002), groundwater (Hrudey et al., 2003), and from fresh spinach (Jay et al., 2007). Riparian filter strips have been developed and tested as measures to attenuate the movement of manure-borne microorganisms into surface waters (Coyne et al., 1995;

* Corresponding author. Tel.: þ1 706 769 5631x228; fax: þ1 706 769 8962. E-mail address: [email protected] (M.B. Jenkins). 0043-1354/$ e see front matter Published by Elsevier Ltd. doi:10.1016/j.watres.2011.10.049

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 7 6 e1 8 6

Entry et al., 2000). Contamination, nevertheless, occurs as evidenced by the enumeration of Salmonella and E. coli 0157:H7 under base flow conditions in surface waters of an agricultural watershed when concentrations of fecal indicator bacteria, commensal E. coli, and fecal enterococci were below impairment criteria (Jenkins et al., 2008; 2009). Given that an infective dose of Salmonella can be as low as 100 cells (Bitten, 1984), and the infective dose of E. coli 0157:H7 can range between 10 and 100 cells (Jones, 1999), dilute concentrations of these two bacterial pathogens in recreational waters may pose a risk to public health. Analogous to waste stabilization ponds associated with domestic waste water treatment facilities (Davies-Colley et al., 2000), and wet ponds designed to treat urban runoff (Hathaway et al., 2009), ponds in agricultural watersheds may provide mechanisms for inactivating fecal bacteria. Natural mechanisms of disinfection in impoundments in agricultural watersheds would be similar, such as solar radiation (DaviesColley et al., 2000) and predation (Barcina et al., 1997). Jenkins et al. (2011) demonstrated that both insolation and predation were factors in die-off of commensal E. coli and fecal entercocci in an impoundment in a first-order watershed containing beef cattle, wildlife, and cropland fertilized with poultry litter. In a study of paired watersheds with significant agricultural land use and in which one watershed had a greater percentage of ponds than the other, Lowrance et al. (2007) showed a significant reduction in sediment and nutrient loads (NH4eN, NO3eN, total N, inorganic N, ortho P, Cl, and total suspended solids) from the watershed with the greater area of ponds. With the expansion of the interface between urban and agricultural land, the need for abatement of nonpoint sources of fecal pollution is an immediate concern. Fisher et al. (2000) reported that placement of grazing cattle upstream from a pond in the landscape was an effective means of reducing the down stream loads of fecal indicator bacteria. Their observations suggested that ponds situated in watersheds with stream inflows and outflows attenuate the load of fecal indicator bacteria and presumably pathogenic bacteria associated with animal agriculture. Comparison of the efficacy of ponds in pollution abatement is difficult because of site-specific differences in pond size, shape, storm flow characteristics, quality of inflow, and hydraulic loading rate (Thackston et al., 1987; Van Buren et al., 1996; Persson and Wittgren, 2003). Water and contaminant residence times and pond bathymetry are essential hydrologic information for determining inputeoutput relationships within small ponds. Hydrology and hydraulics, representing temporal distribution of the inflows and flow patterns that develop in ponds during events, respectively, are two primary factors influencing residence time (Walker, 1998). The theoretical residence time (nominal retention time) is computed as the ratio of impoundment volume to the discharge rate. The actual travel time distribution is, however, complicated by varying inflow rate, evaporation, seepage, wind, changes in storage, thermal stratification, and bypass flow through the impoundment. Sedimentation and biological factors also affect nutrient abatement. Small impoundments typically are built to fit across an existing creek or stream thus limiting ability for a hydraulically optimal design (Walker, 1998).

177

Most residence time determinations and resulting models are made from an analysis based on assumptions such as steady-state, plug flow, complete mixing, single inlet and outlet, and homogeneous systems, borrowed from principles used to design chemical reactors (Kadlec, 1994; Walker, 1998; Nauman, 2008). Many investigators (Nix, 1985; Kadlec, 1994; Werner and Kadlec, 1996; Walker, 1998; Persson and Wittgren, 2003) have pointed out that steady-state analyses and designs based on these assumptions are not suitable for ponds. Despite such complications the usual method employed to measure residence time is a tracer test whereby a quantity of a conservative tracer is injected at the inlet and concentration is monitored as a function of time at the outlet, until background levels are achieved again; thus, a retention time distribution (RTD) curve is developed. A number of numerical descriptors are then derived from the RTD to which hydraulic efficiency as proxy for pond performance can be correlated. Using such approach on shallow basins 60 to 600,000 m3 in size, Thackston et al. (1987) developed a model (Eq. (1)) that they recommended to estimate the hydraulic efficiency of similar basins in the absence of site-specific data needed to develop a more rigorous model.   Tm =T ¼ 0:84 1  eð0:59ðL=WÞÞ

(1)

Where T is the theoretical volumetric resident time computed as V/Q, with V taken as pond volume and Q is flow rate; Tm mean residence time; L is length and W is width. Thackston et al. (1987) referred to the parameter Tm/T as “hydraulic efficiency.” The model neglects wind and depth effects. Konyha et al. (1995) noted that the definition of T implicitly assumes total mixing, and, therefore, gives only an approximation of the residence time where total mixing does not occur. Our objective was to undertake a systematic study of impoundments in watersheds in two physiographic regions of the USA (Piedmont and Coastal Plain) in which animal agriculture occurs, and determine the reduction on the load of fecal indicator bacteria, Salmonella and E. coli 0157:H7. This paper will focus on results of testing the hypothesis that these ponds under baseflow conditions reduce the concentrations and fluxes of fecal indicator bacteria. The relation of fecal indicator bacteria to dynamics of microbial communities and nutrient characteristics were examined. Because exposure time to natural mechanisms of inactivation such as solar UVradiation and microbial predation (Davies et al., 1995; SchulzFademrecht et al., 2008; Jenkins et al., 2011) are important for the elimination of fecal bacteria, retention time was determined at the Piedmont pond to better understand the processes associated with fecal bacterial inactivation.

2.

Materials and methods

2.1.

Study sites

One pond located in the Piedmont and two ponds in the Coastal Plain of Georgia were used in this study. A man-made impoundment, Bishop Pond, located in the Southern

178

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 7 6 e1 8 6

Piedmont, has been described in detail (Jenkins et al., 2008). Briefly, it is approximately 1.6 ha, contains approximately 2.4(10)10 L, and is located in a 100 ha first-order watershed consisting of grazed pastures, a cropped field that is amended with poultry litter annually, and a wooded riparian zone with wildlife and from which cattle are excluded. A perennial firstorder stream fed by a series of springs flows into and out of Bishop Pond which captures base and storm flow from approximately 60% of the 100-ha watershed. It has an approximate mean length of 235 m and mean width of 70.5 m (a w3.3 length to width ratio recommended as a minimum ratio for good pond performance e Van Buren et al. (1996)). The deepest part is about 4-m from permanent pool level and occupies an approximately 35  80 m2 area close to the outlet. The bed level then gradually rises toward the edges where it is about 0.4 m from permanent pool level. The pond holds approximately 24(10)10 L at pool level. Both Ponds A and C are man-made impoundments located in a sub-watershed of the Little River known as the University of Georgia Animal and Dairy Science Farm watershed (ADS watershed). Livestock in the 240 ha ADS watershed include 100 head of beef cows and calves. These are pastured in the watershed year round; there is a 250-cow free stall dairy and dairy heifer and dry cow feeding and grazing area. Pond A is approximately 0.7 ha, holds approximately 4.4(10)6 L and is at the edge of a pasture and is accessible to grazing cattle. Pond C is approximately 2.0 ha and holds approximately 2.5(10)6 L. It captures water draining from liquid manure application areas (spray fields), a restored riparian wetland, and paddocks where replacement dairy heifers are fed with limited grazing. Pond inflow and outflow at Bishop Pond were measured with standard structures (Brakensiek et al., 1979) including a 120 V-notch weir at inflow and a 0.46-m H-flume at the outflow. Sensors and data logging equipment were programmed to store average 5-min flow rates. These 5-min flow data were then integrated over time to produce flows at appropriate time intervals. Both Ponds A and C are outfitted with H-flumes and refrigerated ISCO samplers (Teledyne Isco, Inc., Lincoln, NE) to measure and sample outflow.

2.2.

Tracer studies

Two tracer studies were conducted using sodium bromide (NaBr) in summer 2009 and sodium chloride (NaCl) in early spring 2010 to determine residence time at Bishop Pond in the Piedmont. The greater expense of sodium bromide could not be justified and it was replaced with sodium chloride for the second tracer test. Commensal E. coli was applied as a microbial tracer at the spring 2010 tracer experiment.

2.2.1.

Bromide (Br)

In July 2009, 95.0 kg of NaBr equivalent to 73.8 kg of Br- was mixed with 200 L of inflow water in a polyethylene drum and drained through a rubber hose to the discharge point at approximately 0.11 L/s. The release occurred from 10:30 am to 11:00 am. The stream inflow rate at the time of spiking was 1.27 L/s. The spike point was approximately 5-m upstream of where the stream enters the pond. A Sigma 900 Max portable sampler (American Sigma, Loveland CO) was used to collect outflow samples in 24 glass bottles. Initially the sampler was

programmed to collect 150 mL samples at the hour and half past the hour into one 300 mL glass sampling bottle (1-hr sampling). This protocol was used from the initiation of the tracer study (July 14, 10:00 am) until August 6, 9:30 am. The program was then modified to collect 75 mL samples at the hour and half past the hour to composite over each 2-hr interval (2-hr sampling) from August 6, 10:00 am to September 4, 1:30 pm. Then, until 10:00 am on October 2, 75 mL samples were collected every hour and composited over each 4-hr interval (4-hr sampling). The sampler bottle rack containing 24 bottles was transported to the laboratory once all bottles were full. Duplicate subsamples from each bottle were poured into 125 mL specimen cups and kept at room temperature. A Br ion selective electrode sensor (Model WQBR Sensor; Nexsens Technology Inc. Beavercreek, OH) was initially used to measure Br- concentration in the samples. However, the probe was unstable and unreliable and its use was discontinued. Instead, duplicate 20-mL subsamples were taken from the specimen cup samples at 3 to 4 subsamples per day, time-spaced approximately equally, and sent to the USEPA lab in Athens, GA for analysis of Br on a Metrohm Ion Chromatograph (Metrohm IC Systems AG, Herisau, Switzerland). The typical CV between replicates was 2% or less. A total of 194 samples were analyzed for Br.

2.2.2.

Chloride (Cl)

In March 2010, 386 kg of NaCl equivalent to 234 kg of Cl was mixed with 3780 L of inflow water in a polyethylene drum and drained through a rubber hose to the same discharge point as the Br at approximately 0.42 L/s. The release occurred from 11:00 am to 13:30 pm. The stream inflow rate at the time of spiking was 3.68 L/s. The sampler program was similar to that used for bromide sampling except that only the 1-hr (8:00 March 4, to 9:00 March 29, 2010) and 4-hr composite samples (10:00 March 29 to 13:00 April 27, 2010) were used. A total of 481 bottle samples were collected. Subsamples were transferred to 125 mL specimen cups and kept at room temperature. From the specimen cups a total of 245 subsamples in 20-mL viles, made up of approximately 12 samples/day from spike day to March 15, 2010, 6 samples/day to March 29, 2010, and 1 sample/day thereafter through to April 27, 2010, were sent to the University of Georgia’s Soil and Water Analyses Laboratory (Athens-GA) for determination of Cl on an ion chromatography instrument with precision similar to that of the bromide assay.

2.2.3.

E. coli

A wild-type commensal E. coli strain isolated from fresh calf feces (Jenkins et al., 2003) was grown in five 100 mL volumes of Brain Heart Infusion (BHI) broth. Cells were harvested by centrifugation, washed in phosphate buffered saline (PBS), and each volume resuspended in 100 mL PBS. The total of 500 mL of commensal E. coli was poured and mixed into 100 L of pond water from the inflow site. From five subsamples of inoculated pond water E. coli concentrations were measured as previously described (Jenkins et al., 2008). Immediately after the introduction of the chloride tracer, the E. coli (at a load of 1.6 (10)13 cells) were placed in the pond inflow over approximately10 min. The discharge flux of E. coli from Bishop Pond was measured with an automated system that compiled

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 7 6 e1 8 6

179

total outflow volume through the flume in 5-min intervals. The total flow for this period was then computed from the 5min flow records. The 5-min flow amount multiplied by the corresponding concentration, assuming linear interpolation of concentration between sampling times, produced the flux for that period. This was repeated for all samplings. Cumulative addition of these fluxes produced the total flux of E. coli through the flume over 664 h (w28 d) of monitoring after which E. coli concentrations were at baseflow concentrations or not detectable.

(Hatch Company, Loveland, CO) and colorimetric techniques (Clesceri et al., 1998). Total suspended solids (TSS) were determined by standard methods (Clesceri et al., 1998). Concentrations of chlorophyll a were determined on subsamples following EPA Method 445.0 (Arar and Collins, 1997 - www.epa.gov/microbes/m450_0pdf e accessed 3/2/11) with fluorescence analysis on a Turner-Designs TD-700 Fluorometer (Turner Designs, Sunnyvale, CA).

2.3.

Analysis for fecal indicator bacteria E. coli and fecal enterococci, and total direct microbial counts have been described (Jenkins et al., 2008). Analyses for Salmonella and E. coli 0157:H7 have been described (Jenkins et al., 2008, 2009; 2011).

Sampling scheme

Inflow and outflow samples were taken at all three ponds. At stations along a transect across each pond both surface (5 cm) and deep (50 cm) samples were taken. Deep samples were taken at the extinction depth of solar UV-radiation. In addition, at the Bishop Pond site, two sites up stream from the inflow site were sampled. The samples at these sites were analyzed for commensal E. coli, fecal enterococci, and pond water chemical parameters (NH4eN, NO3eN, total N, inorganic N, ortho P, Cl, and total suspended solids). For pathogen analysis at the Bishop Pond site, for logistical considerations, water samples were taken at the inflow, and outflow sites, and one within pond surface sample. In addition, Salmonella and E. coli 0157:H7 were sampled at separate months during 2006, 2007 and part of 2008. At Bishop Pond Salmonella was sampled starting in August 2006 and followed by sampling in August and December 2006, February, March, April, May, July, October, and December 2007, February, March, April, May June, July, and August 2008. E. coli 0157:H7 was sampled starting in July 2006, followed by sampling in September, and November 2006, February, May, June, August, September, and November 2007, January, March, April, May June July, and August 2008. At Ponds A and C pathogen analysis for both Salmonella and E. coli 0157:H7 was undertaken after the protocols for each pathogen were established, and because of the many months of no inflow or outflow, within pond samples only were taken. Total direct microbial counts were undertaken at Bishop Pond; because of technical and logistic difficulties, total direct microbial counts were not performed on samples from Ponds A and C. At each within pond sampling site sediment samples were obtained with a small benthic dredge (Mighty Grab Dredge, Ben Meadows, Inc., Janesville, WI) and analyzed for commensal E. coli and fecal enterococci. Samples were taken monthly beginning February 2006 and ending August 2008. Because in- and outflows for Ponds A and C were ephemeral and non-existent for several months, we report results on fecal indicator bacteria and nutrients from those months for which inflow and outflow samples were taken.

2.4.

2.5.

2.6.

Microbiological analysis

Data analysis

Significant differences at P  0.05 between inflow, outflow, and within pond sites of concentrations of microorganisms (natural log-transformed) and nutrients were analyzed with Proc Mixed of SAS (Version 9.1; SAS Institute, Cary, NC) and the program’s repeated-measures option and treating time as the repeated measure (Littell et al., 1996). Means of bacterial concentrations were back transformed and presented as log10 MPN mL1 or L1 for the fecal indicator bacteria and pathogens, respectively. Linear relations between microorganisms and chemical characteristics were analyzed with the regression procedure Proc Reg of SAS (version 9.1).

3.

Results

3.1.

Bishop Pond retention time experiments

3.1.1.

Theoretical and Thackston model-based residence time

Base flow variability of the stream flow feeding Bishop Pond meant that the theoretical and Thackston model-based

Chemical analyses

Total N and P analyses were performed on a Technicon Autoanalyzer II (SEAL Analytical, Mequon, WI) using standard colorimetric techniques (Clesceri et al., 1998). Subsamples were filtered through prewashed Whatman 934AH glass fiber filters and analyzed for NO3eN, NH4eN, dissolved reactive P (ortho-P), and Cl on a Lachet 8000 Flow Injection Analyzer

Fig. 1 e Theoretical residence time (RT) and mean RT based on Thackston’s model as a function of flow rate at Bishop Pond.

180

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 7 6 e1 8 6

residence times also varied (Fig. 1). Both show an exponential drop with increasing flow rate. Mean flow rate during the JulyeSeptember 2009 bromide tracer test was 8.7 L s1 and was influenced by storm flows during the latter half of September when a peak of 85 L s1 was observed. For most of July and August base flow was at 3 L s1 or less. The theoretical and Thackston model-based residence times for a 3 L s1 flow are approximately 92 and 67 days, respectively. Mean flow rate during the chloride tracer test in MarcheApril 2010 was approximately 15 L s1 and was influenced by nine storm flows with peaks reaching 70 L s1. Base flow varied from 9 to 13 L s1. The theoretical and Thackston model-based residence times for a 15 L s1 flow are approximately 18 and 13 days, respectively.

3.1.2.

3.1.3.

Chloride tracer test

The outflow Cl concentration for the first 41 days was steady around 8 mg L1, and then decreased during the next 13 days (Fig. 3A). The corresponding 5-min average discharge rates fluctuated over this sampling period (Fig 3b). Base flow stayed around 10 L s1 while 11 storm flows occurred with peak flows ranging from 45 to 80 L s1. Chloride concentration had not decreased to background levels (w4.85 mg L1) by day 54 after spiking. Therefore, a non-linear regression was used to extend estimates of concentration to background levels and determination of Cl mass balance (Fig. 3A). Based on this model the 234 kg of chloride would have exited the pond in 69 days (w1664 h). The outflow Cl concentration (Fig. 3A) showed much scatter for the first 25 days after spiking (w600 h) indicating a spatially non-uniform mixing.

Bromide tracer test

Bromide concentration peaked at 1.41 mg L1 on day 8 then went down uniformly to 0.93 mg L1 by day 64 (Fig. 2A). Because of storms in September and increased discharge (Fig. 2B) Br concentrations dropped to 0.25 mg L1 by day 70 and then to background levels of 0.2 mg L1 by day 74. Cumulative Br mass exiting the pond amounted to 2.0 kg by day 8, 16.9 kg by day 64 (0.266 kg day1 for day 8e64), and 35.8 kg by day 70 (3.16 kg day1 for day 64e70). Total Br mass accounted for in the 76 days after spike when concentration returned to background levels was only 39.5 kg (54.3%) of the 72.8 kg added to the pond.

3.1.4.

Fig. 2 e Bromide tracer experiment: A. Bromide concentrations; B. Total daily discharge.

Fig. 3 e Chloride tracer experiment: A. Chloride concentrations; B. Total daily discharge.

E. coli tracer test

The time between samplings at the outflow after spiking with chloride and E. coli at the inflow ranged from 4 to 48 h and had a mean of 20.4 h. Assuming linear interpolation between sampling times, the flux of E. coli leaving the pond was estimated using 5-min mean flow rates and concentrations. To account for total flux of E. coli exiting Bishop Pond at each sampling time the 5-min fluxes were cumulatively summed. Two sources of E. coli were considered: the E. coli spike and the continuous input from the creek during this experiment. The initial E. coli concentration in the creek water (before spiking)

181

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 7 6 e1 8 6

the highest concentrations of total P and chlorophyll. In contrast to the continuous inflow and outflow fluxes at Bishop Pond, inflow and outflow fluxes of Ponds A and C were ephemeral and retention times for them could not be determined. The concentrations of fecal indicator bacteria at Bishop Pond were significantly greater in the upstream and inflow sites than within pond and pond outflow. No differences were observed between surface (5 cm) and deep (50 cm) samples within Bishop Pond. The total direct microbial counts were, contrary to the fecal indicator bacteria, significantly greater within the pond than upstream and pond inflow (Fig. 5). Concentrations of Salmonella within pond and pond outflow were significantly less than inflow concentrations. In contrast, differences between inflow, within pond and outflow concentrations of E. coli 0157:H7 were not observed (Fig. 6). Of the 17 sampling times, Salmonella was detected in pond inflow, within pond, and pond outflow nine, five, and three times, respectively. Of those times of detecting Salmonella, concentrations of commensal E. coli were below the impairment level of 126 cfu 100 mL1 (USEPA, 1986) one time in pond inflow, one time within pond, and three times in pond outflow. Similarly out of 17 sampling times, E. coli 0157:H7 was detected in pond inflow, within pond, and pond outflow nine, eight and six times, respectively. Of those times of detecting E. coli 0157:H7, concentrations of commensal E. coli were below impairment levels three times in pond inflow, eight times within pond, and five times in pond outflfow. In- and outflow fluxes of commensal E. coli, fecal enterococci, and the Pond’s microbial community paralleled their respective concentration patterns (Fig. 7). In the sediments of Bishop Pond concentrations of commensal E. coli were significantly less than concentrations of fecal enterococci (Fig. 8). Concentration of both E. coli and fecal enterococci correlated with pond water nitrate and ammonia concentrations (Table 2). Concentrations of the two pathogens, Salmonella (ranging between below the detection limit of 0.001 and 120 MPN L1) and E. coli 0157:H7 (ranging between below detection limit and 946 MPN L1), were several log10 orders of magnitude less than either of the fecal indicator bacteria (Fig. 6). No differences in concentrations of commensal E. coli and fecal enterococci were observed between pond inflow, within pond and pond outflow of Pond A. In contrast to Pond A, concentrations of commensal E. coli and fecal enterococci within Pond C increased compared to inflow concentrations (Fig. 9). Concentrations of E. coli and fecal enterococci were ten-times greater in Ponds A and C than in Bishop Pond. Differences between shallow and deep samples were not

Fig. 4 e Cumulative mass of E. coli cells in Bishop Pond outflow and hypothetical mass of E. coli remaining in Bishop Pond based on a T99-value of 20 days (Jenkins et al., 2011).

was 200 MPN mL1; this concentration we assumed to continue throughout the period of the experiment. Assuming these two sources, total input after the spike was 1.601(10)13 MPN of E. coli. Total flux of E. coli measured at the outflow 27 days after the spiking event was 3.69(10)10 cells which represented 0.23% of the total influx of E. coli. A comparison of the accumulative load of spiked E. coli exiting Bishop Pond with the load of E. coli remaining in the pond based on a hypothetical 20 days for E. coli to reach 99% inactivation (T99 ¼ 20 days) (Jenkins et al., 2011) indicated a convergence of loads exiting with that remaining in the pond (Fig. 4). Based on the die-off model (T99 ¼ 20 days), at 27 days less than 1% of the spiked load of E. coli would be remaining in the pond. The convergence of the model data and measured accumulative load at the outflow, thus, indicates die-off of E. coli.

3.2. Pond characteristics, fecal indicator bacteria, and pathogens Differences in nutrient and chemical concentrations between Bishop Pond and Ponds A and C were evident (Table 1). With the exception of NO3, Bishop Pond had the lowest concentrations of NH4, total N, ortho P, total P, TSS, and chlorophyll compared to Ponds A and C. Pond C had the highest concentrations of NO3, NH4, total N, and ortho P; whereas, Pond A had

Table 1 e Mean nutrient and chemical concentrations within Bishop Pond (BP), and ponds A and C. The mean of each within pond sampling site is treated as a replicate per pond (n [ 8). Different letters by means of each chemical indicates a difference at P < 0.05. Pond

NO3

NH4

Total N

Ortho P mg L

Bishop Pond A Pond C

0.264b 0.167a 1.712c

0.158a 0.079a 1.418b

0.897a 3.042b 4.315c

Total P

TSS

1

Chlorophyll mg L1

0.009a 0.111b 0.279c

0.051a 0.743c 0.501b

7.57a 118.10c 24.17b

42.4a 1379.3c 120.6b

182

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 7 6 e1 8 6

observed at either Pond A or Pond C (Fig. 9). Differences in within pond mean concentrations of Salmonella (ranging between detection limit and 112.3 MPN 100 mL1) were observed between ponds; whereas, no differences in concentrations of E. coli 0157:H7 (ranging between limit of detection and 7.3 MPN 100 mL1), commensal E. coli, and fecal enterococci were observed between Ponds A and C (Fig. 10). Each time Salmonella or E. coli 0157:H7 was detected in Pond A, the concentrations of commensal E. coli were greater than the impairment level. In contrast, at two sampling times at Pond C concentrations of commensal E. coli were below the impairment level when Salmonella was detected. E. coli 0157:H7

Fig. 5 e Mean concentrations (n [ 31) of commensal E. coli (Ec), and fecal enterococci (FE) at log10 MPN 100 mLL1, and total direct microbial counts (TC) at log10 cells mLL1 for each sampling site of stream inflow, within pond, and outflow at Bishop Pond. Different letters above means of each bacterial type and community indicates differences within each community at P £ 0.05.

Fig. 6 e Mean concentrations of Salmonella (n [ 17), E. coli 0157:H7 (Ec0157H7) (n [ 17) at MPN LL1, and commensal E. coli, and fecal enterococci (n [ 31) at log10 MPN 100 mLL1 associated with pond inflow, within pond, and pond outflow at Bishop Pond. Different letters over means between each of the bacterial communities indicate differences at P £ 0.05.

Fig. 7 e Flux rates of commensal E. coli, fecal enterococci, and total direct microbial counts for 32 months.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 7 6 e1 8 6

Fig. 8 e Mean sediment concentrations of commensal E. coli and fecal enterococci at Bishop Pond (n [ 30), and ponds A (n [ 24) and C (n [ 27). Means with different letters indicates a difference between organisms and ponds at P £ 0.05.

183

observed was 1.4 mg 1 (Fig. 2A) indicating very little mixing of Br in the pond. The September storms increased the Br flushing rate almost 12-times. In addition, only 53.5% of the added bromide exited the pond after 74 days indicating a resident time greater than the 92 and 67 days based on the theoretical or Thackston model, respectively. Since there were no storms soon after spiking, early flushing of the Br did not occur. Nearly half the added Br stayed trapped in the pond. Substances entering the pond under base flow would be retained for an extended period of time. A ‘well-mixed-vessel’ criterion for determining residence time does not appear to hold for Bishop Pond. Like the Br data, the Cl data indicated that a ‘well-mixedvessel’ criterion for determining residence time did not appear to hold for Bishop Pond. Assuming complete mixing in 24(10)10 L of pond water Cl concentration in the pond would have been approximately 9.79 mg L1 above background Cl concentration which was at 4.85 mg L1. But only 3.1 mg L1 above background was accounted for. This Cl imbalance suggests that the regression model used to extend the estimated concentration was likely invalid. The results of both

Table 2 e Coefficients of determinations (R2) from linear regression analysis of commensal E. coli (Ec), fecal enterococci (FE) against Bishop Pond water chemical parameters: nitrate (Nitr), ammonia (Amm). (N [ 327). Organisms

Variable

R2

Pr > F

Nitr Amm Nitr Amm

0.3096 0.3497 0.2550 0.2738

<0.0001 <0.0001 <0.0001 0.004

Ec FE

detection in Pond C occurred when the concentration of commensal E. coli was below the impairment level. As observed at Bishop Pond, concentrations of commensal E. coli in sediment samples at Pond A and Pond C were significantly smaller than sediment concentrations of fecal enterococci (Fig. 8). Significant linear correlations were observed between commensal E. coli and fecal enterococci and chemical parameters at Ponds A and C (Table 3). Concentrations of commensal E. coli in Pond A correlated with total P, and concentrations of E. coli in Pond C correlated with ortho P, total P, and chlorophyll concentrations. Concentrations of fecal enterococci in Pond A correlated with ammonia, total N, ortho P, TSS, and chlorophyll. No correlation between fecal enterococci concentrations and pond water chemical parameters were observed at Pond C.

4.

Discussion

The Br and Cl tracer and E. coli spiking experiments highlighted the difficulty of estimating residence time under the experimental conditions for ponds like Bishop Pond. Assuming complete mixing of the added Br in 24(10)10 L of pond water, Br concentration in the pond would have been 3.09 mg L1 above background Br concentration which was steady around 0.2 mg L1. The maximum concentration

Fig. 9 e Mean concentrations (n [ 9) of commensal E. coli and fecal enterococci for inflow (A1, C1), within pond sites (A2-A5, C2-C5), and outflow (A6, C6) of ponds A and C. Lower case a and b indicate shallow and deep samples, respectively. Different letters above means of each bacterial community indicates differences within each community at P [ 0.05.

184

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 7 6 e1 8 6

Fig. 10 e Mean concentrations (n [ 7) of Salmonella, E .coli 0157:H7 at MPN LL1, and commensal E. coli and fecal enterococci at log10 MPN 100 mLL1 within ponds A and C for seven months of sampling. Different letters above means indicate differences between bacterial communities at P £ 0.05.

the Br and Cl tracer experiments characterized a pond with very little mixing and one in which the time to effectively flush substance out from the pond was greater than the duration of the tracer experiments, and greater than the theoretical residence time or the residence time determined by the Thackston model. The flux of E. coli showed that a significant amount of the spiked E. coli (w99.7%) was either residing in various parts of the pond 27 days after the initial spike or died off. Like the results of the Br and Cl tracer experiments the E. coli spike appeared to indicate that Bishop Pond is not well-mixed, and fecal bacteria entering the pond under base flow were retained for periods greater than two to three months. Assuming the results of the Br and Cl tracer experiments to be valid, a resident time of several weeks for commensal E. coli in Bishop Pond would imply extended exposure to solar UV-radiation and microbial predation. The convergence of the cumulative load of commensal E. coli in the pond discharge,

Table 3 e Statistically significant coefficients of determination (R2) from linear regression analyses of commensal E. coli(Ec) and fecal enterococci (FE) concentrations against chemical parameters at Tifton site’s ponds A and C: ammonia (Amm), total N (TotN), ortho P (OrtP), total P (TotP), total suspended solids (TSS), and chlorophyll (Chl). The number of samples analyzed (n) is given. Organisms

Variable

n

R2

Pr > F

A C

Ec

A

FE

TotP OrtP TotP Chl Amm TotN OrtP TSS Chl

44 81 73 60 59 52 59 58 39

0.1848 0.1743 0.1057 0.2113 0.1733 0.3303 0.1758 0.1502 0.2821

0.0036 0.0001 0.0050 0.0002 0.0010 <0.0001 0.0009 0.0027 0.0005

Pond

and the hypothetical mass of spiked E. coli cells remaining in the pond based on a T99-value of 20 days (Fig. 4) (Jenkins et al., 2011), indicates that die-off of E. coli likely occurred, and at a rate commensurate with the hypothetical T99-value. Furthermore, given the mean in- and outflow fluxes of E. coli at Bishop Pond and assuming a T99-value of 20 days, days to decrease the inflow load to the outflow load (Fig. 7) was calculated to be around 27 days. Considering the in- and outflow flux of fecal enterococci at Bishop Pond, and an assumed T99-value of 10 days (Jenkins et al., 2011), days to decrease the inflow flux to outflow flux was calculated to be around 6.5 days. Based on the time of year and variable intensity of solar UV-radiation, and variability in predation dynamics, the T99-values appear to be close to those experimentally determined in Bishop Pond (Jenkins et al., 2011). These results at Bishop Pond corroborate observations that fecal enterococci are inactivated more quickly than commensal E. coli (Schultz-Fademrecht et al., 2008; Jenkins et al., 2011). Not meeting the criterion of being ‘well-mixed’, as the tracer experiments indicate, may also be observed in the differences and breadth of the ranges between mean concentrations of fecal indicator bacteria at different sampling sites within each pond. Because exposure to solar UV-radiation (Davies et al., 1995; Schultz-Fademrecht et al., 2008; Jenkins et al., 2011) and predation by a microbial community (Davies et al., 1995; Jenkins et al., 2011) can reduce the concentrations of fecal indicator bacteria, retention time within an impoundment is an important factor in their elimination as they move through ponds in agricultural watersheds. This kind of reasoning was supported by data from Bishop Pond as concentrations of fecal indicator bacteria were consistently lower in outflow than inflow samples. On the other hand, Ponds A and C did not appear to function like Bishop Pond. The combination of photochemical breakdown of dissolved organic matter that can be consumed by the autochthonous microbial community, and primary photosynthetic production may account for greater concentration of total microbial counts within Bishop Pond than the total microbial count of the allochthonous inflow microbial community (Cole, 1999). Unlike Bishop Pond whose source is a first-order stream with continuous inflow, the sources of Ponds A and C were ephemeral. Although resident time in Ponds A and C were likely long, neither solar radiation nor predation appeared to decrease influx loads of fecal indicator bacteria. In addition to the ephemeral nature of base flow fluxes, both Ponds A and C had significantly greater concentrations of total N, ortho P, and total P than Bishop Pond. These higher levels of nutrients could support a microbial community of greater density than Bishop Pond, and may be connected to the greater concentration of chlorophyll. Based on studies that Davies-Colley et al. (2000) reported, results of interactions between solar UV-radiation, pond organic matter, and the algal sources of chlorophyll could lead to photo-oxidative inactivation of the fecal indicator bacteria. On the other hand, Davies-Colley et al. (2000) also maintained that a greater algal biomass, as the greater chlorophyll concentrations in Ponds A and C indicate, can reduce pond water clarity and effectiveness of inactivating solar radiation. The within Pond increase in concentrations of fecal indicator bacteria in Pond C may also be

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 7 6 e1 8 6

attributable to non-point runoff of dairy lagoon effluent applied to the spray field contiguous to it. As Mallin et al. (2000) pointed out, nutrient loading in Pond C could stimulate and prolong survival of fecal bacteria. Both Salmonella and E. coli 0157:H7 were observed in all three ponds. The occurrences in which the concentration of either pathogen was determined and concomitant determination of commensal E. coli concentrations were below the criterion for surface water impairment further supports previously reported observations (Jenkins et al., 2008, 2009; Duris et al., 2009) on the lack of connection between the presence of commensal E. coli and these manure pathogens. The apparent decrease in concentration of Salmonella within pond and pond outflow at Bishop Pond (Fig. 6) may indicate the effectiveness of solar radiation and predation on Salmonella as it transited the pond with the fecal indicator bacteria. On the other hand, the lack of differences in mean concentration of E. coli 0157:H7 between pond inflow, within pond, and pond outflow at Bishop Pond may indicate that this zoonotic pathogen may be a persistent member of fecal coliforms as Duris et al. (2009) have suggested, and some strains of E. coli 0157:H7 may be more resistant to exposure to natural disinfection mechanisms of ponds as Jenkins et al. (2011) have observed. The seven months of data collected on the concentrations of Salmonella and E. coli 0157:H7 at Ponds A and C indicated significant differences between the two ponds. Concentrations of Salmonella were greater in Pond C than in Pond A; this difference between ponds may indicate management differences between bovine sources of fecal pollution at the ponds. The quantification of Salmonella in Pond C when the concurrent concentrations of commensal E. coli were below levels of impairment further underscores the apparent disjunction between the impairment criterion and the indicator organism’s effectiveness at providing evidence on the presence of zoonotic pathogens in surface waters. Although mean concentrations of commensal E. coli and fecal enterococci were not significantly different in the water column within each of the three ponds, and die-off rate of the fecal enterococci was more rapid than the die-off rate of commensal E. coli (Jenkins et al., 2011) the sediment concentrations of commensal E. coli were significantly less than concentrations of fecal enterococci in all three ponds. This trend has been observed in estuarine sediments (Roslev et al., 2008); and Liu et al. (2006) reported that fecal enterococci survived longer in lake sediments than commensal E. coli. Davies et al. (1995), however, observed a faster decline in commensal E. coli than fecal Streptococci in freshwater sediments. This pattern of survival in pond sediments that we observed would indicate that the sediments of the three ponds in our study appear more likely to be secondary sources of fecal enterococci than commensal E. coli under conditions of sediment resuspension as during storm flow events. Sediment concentrations of fecal entercocci were greater than water column concentrations at Bishop Pond and Pond A and reflect similar observations that Davies and Bavor (2000) reported for pollution control ponds. Differences in sediment and water column concentrations of fecal enterococci were not observed at Pond C. The correlations observed between commensal E. coli and chemical characteristics differed between the ponds. Whereas

185

E. coli concentrations correlated with NO3 and NH4 in Bishop Pond; in Ponds A and C, E. coli correlated with total P. Only in Pond C did E. coli correlate with ortho P. Mallin et al. (2000) observed correlations between E. coli and NO3 and ortho P both of which had agricultural sources. The sources of chemical nutrients and fecal bacteria at the three pond study sites were from animal agriculture and wildlife. All of these nutrients were present, and as we have observed, not all correlated with the presence of either fecal indicator organism.

5.

Conclusions

This paper focused on the survival dynamics of fecal indicator bacteria and two zoonotic bacterial pathogens in three different ponds in two watersheds impacted by animal agriculture. We examined the interactions between pond hydrologic, nutrient, and biotic characteristics under baseflow conditions and survival of commensal E. coli, fecal enterococci, Salmonella, and E. coli 0157:H7 as they were transported into the pond, traversed and exited the pond.  Tracer experiments indicated that models like Thackston’s residence time model appeared not to fit ponds like Bishop Pond. For the three ponds complete mixing did not occur and residence times appeared to be around three months and long enough for solar UV-radiation and microbial predation to decrease the concentrations of fecal bacteria. Reduction of fecal bacteria occurred in Bishop Pond with continuous in- and outflow fluxes; without continuous inand outflow fluxes at the two ponds in the Coastal Plain, reduction of fecal bacteria was not observed.  The quantification of Salmonella and E. coli 0157:H7 when concentrations of the fecal indicator bacterium E. coli was below the criterion of surface water impairment underscores the concern that fecal indicators in agricultural watersheds can lead to false negatives that can put public health at risk.

Acknowledgments The authors are grateful for the expert technical support of Shaheen Humayoun, Stephen Norris, Anthony Dillard, Stephanie Steed, Jessica Sterling, Michael Martin, Gregory Surratt, Brooke Powell, Wynn Page, Debbie Coker, and Chris Clegg. This research was supported in part by a grant from the USDA-CSREES NRI Competitive Grants Program.

references

Arar, E.J., Collins, G.B., 1997. In vitro determination of Chlorophyll a and Pheophytin a in Marine and Freshwater Algae by Fluorescence. Method 445.0. National Exposure Research Laboratory, Office of Research and Development, U.S. Environment Protection Agency, Cincinnati, OH.

186

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 7 6 e1 8 6

Barcina, I., Lebaron, P., Vivies-Rego, J., 1997. Survival of allochthonous bacteria in aquatic systems: a biological approach. FEMS Microbiology Ecology 23, 1e9. Bitten, G., 1984. Pathogens and parasites in domestic wastewater. In: Bitten, G. (Ed.), Wastewater Microbiology. John Wiley & Sons, New York, pp. 77e100. Brakensiek, D.L., Osborn, H.B., Rawls, W.J., 1979. Field Manual for Research in Agricultural Hydrology. Agricultural Handbook 224. USDA, Washington, DC. Clesceri, L.S., Greenberg, A.E., Eaton, A.D., 1998. Standard Methods for Examination of Water and Wastewater, twentieth ed. APHA, AWWA, WEF, Washington, DC. Cole, J.J., 1999. Aquatic microbiology for ecosystem scientists: new and recycled paradigms in ecological microbiology. Ecosystems 2, 215e225. Coyne, M.S., Gilfifillen, R.A., Rhodes, R.W., Blevins, R.L., 1995. Soil and fecal coliform trapping by grass filter strips during simulated rain. Journal of Soil and Water Conservation 50, 405e408. Davies, C.M., Long, J.H.A., Donald, M., Ashbolt, N.J., 1995. Survival of fecal microorganisms in marine and freshwater sediments. Applied Environmental Microbiology 61, 1888e1896. Davies, C.M., Bavor, H.J., 2000. The fate of stormwater associated bacteria in constructed wetland and water pollution control pond systems. Journal of Applied Microbiology 89, 349e360. Davies-Colley, R.J., Donnison, A.M., Speed, D.J., 2000. Towards a mechanistic understanding of pond disinfection. Water Science and Technology 42, 149e158. Duris, J., Haack, S.K., Fogarty, L.R., 2009. Gene and antigen markers of shiga-toxin producing E. coli in Michigan and Indiana river water: occurrence and relation to recreational water quality criteria. Journal of Environmental Quality 38, 1878e1886. Entry, J.A., Hubbard, R.K., Thies, J.E., Fuhrmann, J.J., 2000. The influence of vegetation in riparian filterstrips on coliform bacteria: I. Movement and survival in water. Journal of Environmental Quality 29, 1206e1214. Ferguson, C., de Roda Husman, A.M., Atavilla, N., Deere, D., Ashbolt, N., 2003. Fate and transport of surface pathogens in watersheds. Critical Reviews in Environmental Science and Technology 33, 299e361. Fisher, D.S., Steiner, J.L., Endale, D.M., Stuedemann, J.A., Schomberg, H.H., Franzluebbers, A.J., Wilkinson, S.R., 2000. The relationship of land use practices to surface water quality in the Upper Oconee watershed of Georgia. Forest Ecology and Management 128, 39e48. Hathaway, J.M., Hunt, W.F., Jadlocki, S., 2009. Indicator bacterial removal in storm-water best management practices in Charlotte, North Carolina. Journal of Environmental Engineering 135, 1275e1283. Hrudey, S.E., Payment, P., Huck, P.M., Gillham, R.W., Hrudey, E.J., 2003. A fatal waterborne disease epidemic in Walkerton, Ontario: comparison with other waterborne outbreaks in the developed world. Water Science and Technology 47, 7e14. Jay, M.T., Cooley, M., Carychao, D., Wiscomb, G.W., Sweitzer, R.A., Crawford-Miksza, L., Asmundson, R.V., Atwill, E.R., Mandrell, R.E., 2007. Escherichia coli O157:H7 in feral swine near spinach fields and cattle, Central California Coast. Emerging Infectious Disease 13, 1908e1911. Jenkins, M.B., Endale, D.M., Fisher, D.S., 2008. Most probable number methodology for quantifying dilute concentration and fluxes of Salmonella in surface waters. Journal of Applied Microbiology 104, 1562e1568. Jenkins, M.B., Endale, D.M., Fisher, D.S., Gay, P.A., 2009. Most probable number methodology for quantifying dilute concentration and fluxes of Escherichia coli 0157:H7 in surface waters. Journal of Applied Microbiology 106, 572e579. Jenkins, M.B., Endale, D.M., Fisher, D.S., Gay, P.A., 2011. Comparative die-off of Escherichia coli 0157:H7 and fecal

indicator bacteria in pond water. Environmental Science & Technology 45, 1853e1858. Jenkins, M.B., Hartel, P.G., Olexa, T.J., Stuedemann, J.A., 2003. Putative temporal variability of Escherichia coli ribotypes from yearling steers. Journal of Environmental Quality 32, 305e309. Jones, D.L., 1999. Potential health risks associated with the persistence of Escherichia coli 0157 in agricultural environments. Soil Use and Management 15, 76e83. Kadlec, R.H., 1994. Detention and mixing in free water wetlands. Ecological Engineering 42, 175e186. Konyha, K.D., Shaw, D.T., Weiler, K.W., 1995. Hydrologic design of a wetland: advantages of continuous modeling. Ecological Engineering 4, 99e116. Littell, R.C., Milliken, G.A., Stroup, W.W., Wolfinger, W.R., 1996. SAS Systems for Mixed Models. SAS Institute, Cary, NC. Liu, P., Phanikumar, M.S., Molloy, S.L., Whitman, R.L., Shively, D.A., Nevers, M.B., Schwab, D.J., Rose, J.B., 2006. Modeling the transport and inactivation of E. coli and enterococci in the near-shore region of Lake Michigan. Environmental Science and Technology 40, 5022e5028. Lowrance, R., Sheridan, J.M., Williams, R.G., Bosch, D.D., Sullivan, D.G., Blanchett, D.R., Hargett, L.M., Clegg, C.M., 2007. Water quality and hydrology in farm-scale coastal plain watersheds: effects of agriculture, impoundments, and riparian zones. Journal of Soil and Water Conservation 62, 65e76. Mallin, M.A., Williams, K.E., Esham, E.C., Lowe, R.P., 2000. Effect of human development on bacteriological water quality in coastal watersheds. Ecological Applications 10, 1047e1056. Mead, P.S., Slutsker, L., Dietz, V., McCaig, L.F., Bresee, J.S., Shapiro, C., Griffin, P.M., Tauxe, R.F., 1999. Food-related illness and death in the United States. Emerging Infectious Disease 5, 607e625. Nauman, E.B., 2008. Residence time theory. 100th anniversary review. Industrial & Engineering Chemistry Research 47, 3752e3766. Nix, S.J., 1985. Residence time in stormwater detention basins. Journal of Environmental Engineering 111, 95e100. Olsen, S.J., Miller, G., Breuer, T., Kennedy, M., Higgins, C., Walford, J., McKee, G., Fox, K., Bibb, W., Mead, P., 2002. A waterborne outbreak of Escherichia coli O157:H7 infections and hemolytic uremic syndrome: implications for rural water systems. Emerging Infectious Disease 8, 370e375. Persson, J., Wittgren, H.B., 2003. How hydrological and hydraulic conditions affect performance of ponds. Ecological Engineering 21, 259e269. Roslev, P., Bastholm, S., Iversen, N., 2008. Relationship between fecal indicators in sediment and recreational waters in a Danish estuary. Water, Air, & Soil Pollution 194, 13e21. Schultz-Fademrecht, C., Wichern, M., Horn, H., 2008. The impact of sunlight on inactivation of indicator microorganisms both in river water and benthic biofilms. Water Research 42, 4771e4779. Thackston, E.L., Shields Jr., F.D., Schroeder, P.R., 1987. Residence time distribution of shallow basins. Journal of Environmental Engineering 113, 1319e1332. USEPA, 1986. Ambient Water Quality Criteria for Bacteriad1986. EPA440/5-84-002.9. Office of Water Regulations and Standards, Criteria and Standards Division, Washington, DC. Van Buren, M.A., Watt, W.E., Marsalek, J., 1996. Enhancing removal of pollutants by an on-stream pond. Water Science and Technology 33, 325e332. Walker, D.J., 1998. Modelling residence time in stormwater ponds. Ecological Engineering 10, 247e262. Werner, T.M., Kadlec, R.H., 1996. Application of residence time distribution to stormwater treatment systems. Ecological Engineering 7, 213e234.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 8 7 e1 9 4

Available online at www.sciencedirect.com

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

Quantitative detection of Cryptosporidium oocyst in water source based on 18S rRNA by alternately binding probe competitive reverse transcription polymerase chain reaction (ABC-RT-PCR) Naohiro Kishida a,*, Ryo Miyata b, Atsushi Furuta b,c, Shinji Izumiyama d, Satoshi Tsuneda c, Yuji Sekiguchi b, Naohiro Noda b, Michihiro Akiba a a

Division of Water Management, Department of Environmental Health, National Institute of Public Health, 2-3-6 Minami, Wako, Saitama 351 0197, Japan b Biomedical Research Institute, National Institute of Advanced Industrial Science and Technology (AIST), Central 6, Higashi 1-1-1, Tsukuba, Ibaraki 305 8566, Japan c Department of Life Science and Medical Bioscience, Waseda University, 2-2 Wakamatsu-cho, Shinjuku-ku, Tokyo 162 8480, Japan d Department of Parasitology, National Institute of Infectious Diseases, Toyama 1-23-1, Shinjuku, Tokyo 162 8640, Japan

article info

abstract

Article history:

We describe an assay for simple and cost-effective quantification of Cryptosporidium

Received 11 August 2011

oocysts in water samples using a recently developed quantification method named alter-

Received in revised form

nately binding probe competitive PCR (ABC-PCR). The assay is based on the detection of 18S

19 October 2011

rRNA specific for Cryptosporidium oocysts. The standard curve of the ABC-PCR assay had

Accepted 21 October 2011

a good fitting to a rectangular hyperbola with a correlation coefficient (R) of 0.9997.

Available online 29 October 2011

Concentrations of Cryptosporidium oocysts in real river water samples were successfully quantified by the ABC-reverse transcription (RT)-PCR assay. The quantified values by the

Keywords:

ABC-RT-PCR assay very closely resemble those by the real-time RT-PCR assay. In addition,

Cryptosporidiosis

the quantified concentration in most water samples by the ABC-RT-PCR assay was

Fluorescence quenching

comparable to that by conventional microscopic observation. Thus, Cryptosporidium

Health-related water microbiology

oocysts in water samples can be accurately and specifically determined by the ABC-RT-PCR

Real-time PCR

assay. As the only equipment that is needed for this end-point fluorescence assay is a simple fluorometer and a relatively inexpensive thermal cycler, this method can markedly reduce time and cost to quantify Cryptosporidium oocysts and other health-related water microorganisms. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Members of the genus Cryptosporidium are protozoan parasites that can cause the gastrointestinal disease cryptosporidiosis (O’Donoghue, 1995). Cryptosporidiosis remains a public health

concern, as demonstrated by continued outbreaks of this disease (Nichols, 2008). Waterborne cryptosporidiosis is particularly important because Cryptosporidium oocysts are resistant to disinfectants such as chlorine commonly used for water treatment (Peeters et al., 1989; Carpenter et al., 1999).

* Corresponding author. Tel.: þ81 48 458 6274; fax: þ81 48 458 6275. E-mail address: [email protected] (N. Kishida). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.048

188

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 8 7 e1 9 4

Detection of Cryptosporidium oocysts in water samples has been carried out by direct microscopic visualization or enzyme immunoassays of obtained and/or cultured oocysts with or without fluorescent antibodies. However, these detection methods are labor-intensive, require a large number of oocysts for positive detection and are not suitable for high-throughput processing of samples (Ramirez and Sreevatsan, 2006). Instead of these conventional methods, molecular (nucleic acid-based) techniques have been developed for rapid detection of Cryptosporidium oocysts from water samples. In particular, real-time polymerase chain reaction (PCR) is widely used for the quantification of Cryptosporidium oocysts (Keegan et al., 2003; King et al., 2005; Miller et al., 2006; Garce´s-Sanchez et al., 2009) because of its high rapidity, sensitivity and reproducibility. However, the real-time PCR requires highly specialized and expensive devices that combine a fluorometer and a thermal cycler for real-time fluorescence measurement. Therefore, other simpler molecular quantification techniques are strongly needed for the routine detection of Cryptosporidium oocysts in drinking water treatment plants, bathing facilities, and wherever water quality monitoring is needed. In this study, we focused our attention on a recently reported novel molecular method for the quantification of nucleic acid sequences named alternately binding probe competitive PCR (ABC-PCR) (Tani et al., 2007) as an improved method for quantification of Cryptosporidium oocysts. This method is cost-effective because it does not require an expensive real-time fluorescence measurement device but requires a simple end-point fluorescence measurement device. We developed this ABC-reverse transcription (RT)PCR assay for cost-effective quantification of Cryptosporidium oocysts, and applied the assay to the detection from real water samples collected from Japanese rivers. To investigate the validity of the ABC-RT-PCR assay, real-time RT-PCR assay and conventional microscopic observation were also performed, and quantification data were compared with each other.

the end-point of PCR. Therefore, this method does not require expensive real-time fluorescence measurement device, but only requires end-point fluorescence measurement using a simple fluorometer, so that the measurement cost can be markedly reduced. Moreover, false-negative results can be easily confirmed by measuring the fluorescence intensity of red dye at the end-point of PCR, because the intensity does not change if PCR reaction itself does not proceed and therefore neither target nor competitor DNAs are amplified.

3.

Materials and methods

3.1.

Sample collection and pretreatment

In total, 14 water samples were collected at 7 sites in tributary rivers of the Tone River basin in Japan where the surface water is utilized for the main production of drinking water for people living in the Tokyo metropolitan area from November to December 2009. Ten liters of water samples were concentrated by vacuum filtration with polytetrafluoroethylene membrane filters (pore size: 1 mm; diameter: 142 mm, Advantec Toyo, Tokyo, Japan), and concentrated samples were eluted from the filter using a vortex mixer (Oda et al., 2002). Then they were purified through immunomagnetic separation (Dynabeads GC Combo, Life Technologies, California, USA). While the manufacturer recommended using 50 mL of 0.1 N HCl in the oocyst dissociation step, in our investigation the step was repeated once more to recover the oocysts. The neutralization procedure was performed in a 1.5 mL tube with 10 mL of 1 N NaOH. The final volume of concentrated sample solution was 110 mL. Half the volume of concentrated sample solution was used for the ABC-PCR and real-time PCR assays after nucleic acid extraction and reverse transcription, and the other half volume was used for a conventional assay based on microscopic observation.

3.2.

2.

Principle and characteristics of ABC-PCR

The principle of ABC-PCR is as follows: the alternately binding probe (hereafter called “ABProbe”) is labeled at the 50 end with a green dye (BODIPY FL) and at the 30 end with a red dye (6carboxytetramethylrhodamine [TAMRA]) as shown in Fig. 1. These fluorescent dyes have a property of being notably quenched by an electron transfer to guanine at a particular position (Kurata et al., 2001; Torimura et al., 2001), and the green dye is quenched via the fluorescence resonance energy transfer (FRET) to the red dye when two fluorescent dyes are in proximity. The ABProbe hybridizes with the target and competitor in perfect match. The green fluorescence intensity reflects the ratio of the target and competitor (internal standard) DNAs, and the red fluorescence intensity reflects the ratio of the hybridized probe to the unbound probe. At the normal PCR using a thermal cycler, the target and competitor DNAs are coamplified with the same efficiency in the presence of the ABProbe. The number of copies of the target DNA can be calculated from the fluorescence intensity of the ABProbe at

RNA extraction and reverse transcription

Immediately after five freeze (80  C) and thaw (37  C) cycles, the sample solution including Cryptosporidium oocysts was incubated at 60  C for 30 min using a heat block with the solution for nucleic acids extraction containing 10 mM Tris (tris-hydroxymethyl-aminomethane)-HCl (pH 7.6), 1 mM EDTA (ethylenediaminetetraacetic acid), 2 mM NaCl, 0.1% TritonX-100 (t-octylphenoxy-polyethoxyethanol), 2 mM DTT (dithiothreitol) and 1.5 mAnson-U/mL Proteinase K. The sample solution was sonicated for 2 min and incubated at 75  C for 10 min. Then the solution was incubated at 95  C for 5 min to deactivate Proteinase K, and immediately cooled down on ice. Next, extracted RNA solution was subjected to reverse transcription using PrimeScript RT reagent Kit (Takara Bio, Shiga, Japan) with 2.5 mM of Crypt-374r primer shown in Table 1 to obtain cDNA. The reverse transcription reaction was performed using GeneAmp PCR system 9700 (Life Technologies, California, USA) with the following program: reverse transcription at 37  C for 15 min and inactivation of enzyme at 85  C for 5 s. Synthesized cDNA was used as template DNA in the ABC-PCR and real-time PCR assays.

189

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 8 7 e1 9 4

Fig. 1 e Schematic presentation of a novel ABC-PCR assay for calculating the ratio of the target to the competitor in competitive PCR. The target shown here is an example that possesses a T base at the 30 end of the probe binding site; however, the presence of A is also acceptable in this method. The competitor has the same sequence as the target, except that the underlined TA bases are replaced by GG bases. Fluorescence intensities reflect the ratio of target to competitor.

3.3.

Oligonucleotide sequences

The sequences of primers and probes of the ABC-PCR and real-time PCR assays for the quantification of 18S rRNA of Cryptosporidium spp. are listed in Table 1. Primers with the same sequences were used for the ABC-PCR and real-time PCR assays, while the sequences of the ABProbe and TaqMan probe were a little different. A sequence of ABProbe was newly designed based on the sequence of the TaqMan probe used for the real-time PCR assay. The position of binding the ABProbe with 18S rRNA genes of Cryptosporidium spp. is displaced by 2 nt as shown in Table 1, because the nucleotide of ABProbe at the 30 end must be cytosine in order to quench the red fluorescence (TAMRA) as shown in Fig. 1. Specificity of real-time PCR and ABC-PCR assays was confirmed in silico predictions using a basic local alignment search tool (Altschul et al., 1990). The chemically synthesized oligo-DNA with a length of 192 bp was used for standard DNA for the ABC-PCR and realtime PCR assays. The sequence of the standard DNA was a partial sequence of the 18S rRNA genes of Cryptosporidium hominis (Accession No. AF093489; nucleotide position: 188e379). The internal standard (competitor) DNA was also chemically synthesized. The sequence of the internal

standard DNA was different from the standard DNA sequence by 2 nt, i.e., AT to GG at the outside of the 50 end position at which the ABProbe annealed (nucleotide position: 272e273) in order to quench the green fluorescence (BODIPY FL) only when the ABProbe is annealed to the sequence of the internal standard DNA, as shown in Fig. 1.

3.4.

ABC-PCR assay

The PCR mixture (20 mL) consisted of 0.4 mL of 50  TITANIUM Taq polymerase (Takara Bio, Tokyo, Japan), 2.0 mL of 10  TITANIUM buffer, 200 mM each of dATP, dCTP and dGTP, 600 mM dUTP (Roche Diagnostics Japan, Tokyo, Japan), 0.25 unit of uracil-DNA glycosylase (heat-labile; Roche Diagnostics Japan, Tokyo, Japan), 75 nM Crypt-193f, 250 nM Crypt-374r, 200 nM Crypt-274Abp, 1 mL of competitor DNA (internal standard) and 1 mL of standard DNA or cDNA from water samples. PCR amplification was performed using Mastercycler (Eppendorf, Hamburg, Germany) with the following program: an initial denaturation at 95  C for 2 min, 50 cycles of denaturation at 95  C for 20 s, annealing at 60  C for 20 s, extension at 72  C for 20 s and final extension at 72  C for 2 min. After the PCR amplification, the fluorescence intensity of each aliquot was measured at 95 and 60  C using Light Cycler 480 (Roche

Table 1 e Nucleotide sequences of PCR primers and probes used to detect Cryptosporidium spp. Name Primer Crypt-193f Crypt-374r TaqMan Probe Crypt-276pb ABProbe Crypt-274Abpc

Sequence (50 / 30 )

Locationa

Used for

References

GGAAGGGTTGTATTTATTAGATAAAGAACCA CTCCCTCTCCGGAATCGAA

193 374

Real-time PCR and ABC-PCR Real-time PCR, ABC-PCR and reverse transcription

Miller et al., 2006

CATTCAAGTTTCTGACCTATCAGCTTTAGACGG

276

Real-time PCR

Miller et al., 2006

ATCATTCAAGTTTCTGACCTATCAGCTTTAGAC

274

ABC-PCR

This study

a Corresponding nucleotide position of Cryptosporidium hominis 18S rRNA gene (Accession No. AF093489) of the 50 end. b TaqMan probe oligonucleotides were labeled with 6-FAM at the 50 end and BHQ at the 30 end. c Alternatively binding competitive probe oligonucleotides were labeled with BODIPY FL at the 50 end and TAMRA at the 30 end.

190

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 8 7 e1 9 4

Diagnostics Japan, Tokyo, Japan). Excitation was performed at 450e495 and 522e555 nm, with 505e537 and 565e605 nm emission filters for green fluorescence and red fluorescence, respectively. Although a real-time fluorescence measurement device was used in order to analyze the fluorescence intensity at the end-point of PCR quickly in this study, a simple fluorometer can be also used for the same accuracy as has been reported in previous papers (Tani et al., 2007; Noda et al., 2008). Serial dilutions of the standard DNA (range: 101e105 copies per reaction) were used to make a standard curve. In addition, internal standard DNA with the concentration of 103 copies were added to each PCR mixture.

3.5.

Data analysis of ABC-PCR assay

As shown in Fig. 1, the fluorescence intensity of the ABProbe changes as follows: (i) When the ABProbe is free in solution, its green fluorescence (G1) is notably quenched by FRET to the red dye. (ii) When the ABProbe hybridizes with a target, its green fluorescence (G2) is bright because the two fluorescent dyes are separate. (iii) When the ABProbe hybridizes with a competitor (internal standard), its green fluorescence (G3) is darker than G2 because its green fluorescence is quenched by guanine bases. (iv) When the ABProbe is free in solution, its red fluorescence (R1) is bright. (v) When the ABProbe hybridizes with the target or competitor, its red fluorescence is similarly quenched by the guanine complementary to the modified cytosine (R2 ¼ R3). Therefore, the red fluorescence intensity reflects the ratio of the unbound probe to the hybridized probe, and the green fluorescence intensity reflects the ratio of target to the competitor in the PCR products. The fluorescence intensities of the green dye (G60) and red dye (R60) at 60  C represent the intensity after hybridization, whereas those of the green dye (G95) and red dye (R95) at 95  C represent the intensity before hybridization. These fluorescence intensities fluctuate slightly from tube to tube because of differences in the reagent volume and/or unknown factors. Therefore, G60 and R60 were divided by G95 and R95 to normalize the non-PCR-related fluorescence fluctuations occurring from tube to tube. The fluorescence intensities measured from the green dye (G60/ G95) and red dye (R60/R95) are the sum of the fluorescence intensities from the unbound probe (GU or RU), the hybridized probe with the target (GT or RT), and the hybridized probe with the competitor (GC or RC), expressed, respectively, as G60 =G95 ¼ GU ð1  yÞ þ GT ½X=ðX þ CÞy þ GC ½C=ðX þ CÞy;

(1)

where y is the ratio of the bound probe to the total probe, and R60 =R95 ¼ RU ð1  yÞ þ RT ½X=ðX þ CÞy þ RC ½C=ðX þ CÞy:

(2)

X and C are the starting quantities of the target and competitor, respectively. At this point, RT ¼ RC :

(3)

Therefore, Eq. (2) converts to y ¼ ½RU  ðR60 =R95 Þ=ðRU  RC Þ: Under this condition, Eq. (1) converts to

(4)

½ðG60 =G95 Þ  GU =½RU  ðR60 =R95 Þ ¼ ½ðCGC  CGT Þ=ðRU  RC Þ=  ðX þ CÞ þ ðGT  GU Þ=  ðRU  RC Þ:

ð5Þ

This equation shows that [(G60/G95)  GU]/[RU  (R60/R95)] and X have a relationship indicated by a rectangular hyperbola. Therefore, the standard curves were obtained by fitting the data points to a rectangular hyperbola, and the number of copies of 18S rRNA of Cryptosporidium spp. in each sample was calculated using the standard curves.

3.6.

Real-time PCR assay

The PCR mixture (20 mL) consisted of 400 nM of each primer, 80 nM of the TaqMan probe, commercially available PCR mastermix (LightCycler FastStart DNA MasterPLUS HybProbe; Roche Diagnostics Japan, Tokyo, Japan) and 2 mL of standard DNA or cDNA from water samples. PCR amplification and fluorescence measurement was performed using LightCycler 2.0 (Roche Diagnostics Japan, Tokyo, Japan) with the following program: an initial denaturation at 95  C for 10 min, 50 cycles of denaturation at 95  C for 10 s, annealing and extension at 60  C for 30 s. Crossing point (Cp) values were determined automatically by LightCycler software using the second derivative maximum method.

3.7. Conversion of number of 18S rRNA molecules of Cryptosporidium spp. to number of Cryptosporidium oocysts The number of 18S rRNA molecules in a standard Cryptosporidium oocyst was investigated by real-time RT-PCR, and the number was used for conversion of the number of copies of 18S rRNA molecules in water samples as determined by the ABC-RT-PCR and real-time RT-PCR assays to the number of Cryptosporidium oocysts. Cryptosporidium parvum oocysts (H8 strain, Yagita et al., 2001) which were maintained in our laboratory by passages in infected mice were used for the determination. Cryptosporidium oocysts were purified from the feces from the mice by a combination of discontinuous density sucrose gradient centrifugation and cesium chloride gradient centrifugation and enumerated with a hemacytometer. The RNA of purified and enumerated oocysts was extracted, and used as template RNA. The template RNA was prepared in 10-fold serial dilutions. Then cDNA was synthesized through reverse transcription. Next, the real-time PCR assay was performed in triplicate for each diluted cDNA solution using chemically synthesized oligo-DNA as standard DNA.

3.8.

Microscopic observation

The conventional microscopic observation was performed by Japanese standard method for detection of Cryptosporidium oocysts in water supply systems (Ministry of Health, Labor and Welfare, 2007). After purification through immunomagnetic separation, Cryptosporidium oocysts were fixed on a membrane filter. The fixed sample was stained with EasyStain antibody stain (BTF, North Ryde, Australia), and

191

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 8 7 e1 9 4

observed using an epifluorescent microscope (IX71; Olympus, Tokyo, Japan).

4.1.

Standard curve of ABC-PCR assay

We constructed a standard curve for the quantification of the 18S rRNA of Cryptosporidium spp. Various amounts of the DNA fragment encoding the partial 18S rRNA genes of C. hominis as target standard DNA, ranging from 101 to 105 copies in the presence of 103 copies of competitor DNA, were prepared. The DNA mixtures were amplified by PCR followed by measuring the fluorescence of the DNA mixtures. [(G60/G95)  GU]/ [RU  (R60/R95)] values were plotted against the starting quantity of the target DNA (Fig. 2). The results show that [(G60/ G95)  GU]/[RU  (R60/R95)] values obtained from triplicate samples closely match a rectangular hyperbola with a correlation coefficient (R) of 0.9997. The standard deviation (SD) of triplicate determinations was less than 0.008 for 101 to 105 copies. These results indicate that the ABC-PCR assay was quantitative and reproducible for the measurement of the number of copies of 18S rRNA of Cryptosporidium spp. Although the quantification range of the ABC-PCR assay is narrower than that of the real-time PCR assay, we can adjust the range by varying the addition quantity of internal standard DNA according to sample concentration, as written in our previous paper (Miyata et al., 2010).

4.2. The number of 18S rRNA molecules in a Cryptosporidium oocyst Fig. 3 shows the relationship between the number of Cryptosporidium oocysts and the number of copies of 18S rRNA molecules of C. parvum as determined by the real-time RTPCR assay. The ratio of the oocyst number to copy number was independent of the initial concentration of template RNA because reverse transcription and cDNA amplification proceeded effectively under this experimental condition. It was found that one standard Cryptosporidium oocyst has approximately 32,700 copies of 18S rRNA molecules. This value was used for the calculation (conversion) of the number of Cryptosporidium oocysts from the number of gene copies in water samples as determined by the ABC-RT-PCR and real-time RT-PCR assays although this conversion ratio may vary a little depending on the activity of Cryptosporidium oocysts. Meanwhile, it has been reported that each Cryptosporidium genome in one oocyst has 20 copies of 18S rRNA genes (Abrahamsen et al., 2004). Therefore, it is possible to detect Cryptosporidium oocysts without reverse transcription. However, the number of copies of rRNA genes is much lower than that of rRNA molecules. If we try to quantify Cryptosporidium oocysts without reverse transcription, labor-intensive high condensation of water samples may be needed. In addition, duplicate or triplicate examination is almost impossible if there is only one Cryptosporidium oocyst in the water sample. Hence, we added reverse transcription process to increase the sensitivity.

[(G60/G95) – GU]/[RU – (R60/R95)]

Results and discussion

0.1

0

-0.1

-0.2

2

3

4

10 10 10 Starting quantity of target DNA (copies)

10

5

10

Fig. 2 e Typical standard curve. A fixed amount of the internal standard (103 copies) was coamplified with various amounts of the target standard DNA in triplicate. The standard curve was obtained by fitting the data points to a rectangular hyperbola. y [ L559.8/ (x D 1602) D 0.1583, and the correlation coefficient (R) was calculated to be 0.9997. The error bars represent the standard deviation (SD) of triplicate determinations per dilution.

4.3. Quantification of Cryptosporidium oocysts in water samples by ABC-RT-PCR assay The number of copies of the 18S rRNA molecules of Cryptosporidium spp. naturally present in 14 water samples were quantified using ABC-RT-PCR and real-time RT-PCR. Then the copy number was converted to the number of Cryptosporidium oocysts using a conversion coefficient determined by the above mentioned experiment. As shown in Table 2 and Fig. 4, concentrations of Cryptosporidium oocysts in river water samples were successfully quantified by the ABC-RT-PCR assay. The quantified values by the ABC-RT-PCR assay are very closely resembling those by the real-time RT-PCR assay.

Copies of 18S rRNA molecules of Cryptosporidium parvum (copies/tube)

4.

0.2

3000 2500 2000

y = 32700x 2 R = 0.9989

1500 1000 500 0

0

0.02 0.04 0.06 Cryptosporidium oocyst (oocysts/tube)

0.08

Fig. 3 e Relationship between the number of Cryptosporidium oocysts and the number of copies of 18S rRNA molecules of Cryptosporidium spp. as determined by real-time RT-PCR.

192

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 8 7 e1 9 4

Table 2 e Three methods of quantifying Cryptosporidium oocysts in surface water samples. Sample A B C D E F G H I J K L M N

ABC-PCR (oocysts/10 L)

Real-time PCR (oocysts/10 L)

Microscopic observation (oocysts/10 L)

0.23 6.6 0.076 6.6 e e 0.67 38 0.13 19 0.57 38 0.98 49

0.33 7.3 e 11 e e 1.4 30 0.17 26 0.036 30 0.36 71

2 e 2 4 e e 4 12 0 14 10 30 16 40

Cryptosporidiun oocysts in water samples as determined by the realtime-RT-PCR assay (oocysts/10L)

The coefficient of determination (R2) value at the linear least squares method between the two PCR assays was 0.89. In addition, the quantified concentration in most water samples by PCR assays was comparable to that by conventional microscopic observation although the principle of quantification is different, and the conversion ratio of the number of

80

a

60 y = 1.1175x - 0.0174 2 R = 0.8925

40

20

0

0 20 40 60 Cryptosporidiun oocysts in water samples as determined by the ABC-RT-PCR assay (oocysts/10L)

Cryptosporidiun oocysts in water samples as determined by microscopic observation (oocysts/10L)

60

40

b y = 0.6008x + 2.7279 R2 = 0.7196

20

0 0 20 40 60 Cryptosporidiun oocysts in water samples as determined by the ABC-RT-PCR assay (oocysts/10L)

Fig. 4 e Relationship between the number of Cryptosporidium oocysts in water samples as determined by (a) ABC-RT-PCR and real-time RT-PCR (b) ABC-RT-PCR and microscopic observation.

copies of 18S rRNA molecules to that of Cryptosporidium oocysts may vary a little depending on the activity of Cryptosporidium oocysts in river water samples. Therefore, it is considered that Cryptosporidium oocysts can be specifically determined by the ABC-RT-PCR assay. In addition, the fluorescence intensity of red dye at the end-point of PCR was significantly decreased in all samples (data not shown), which indicates false-negative results were not obtained in this study. Thus, Cryptosporidium oocysts in real water samples can be simply and accurately quantified using the ABC-RT-PCR assay. As the only equipment that is needed for this endpoint fluorescence assay is a simple fluorometer and a relatively inexpensive thermal cycler, this method can markedly reduce time and cost to quantify Cryptosporidium oocysts. Moreover, this method can be easily applied to quantify many other health-related water microorganisms because it is very easy to design primers and ABProbes. If primers and probes of the real-time (TaqMan) PCR assay are published, in the ABCPCR assay, the same primers can be used. The only thing that the users need to do is to modify the sequence of ABProbe a little bit from the TaqMan probe in the same manner in this study. Meanwhile, a few data (Sample K and M in Table 2) quantified by two PCR assays were much smaller than those by microscopic observation. It is well known that environmental samples contain PCR inhibitors such as humic acids, and the number of target DNA is underestimated by quantitative PCR assays due to the existence of inhibitors. However, in the ABCPCR assay, accurate quantification is attained regardless of the existence of PCR inhibitors because internal standard DNA is used, and the effect of PCR inhibition on quantification is canceled (Tani et al., 2007). In addition, sample purification using immunomagnetic separation was carefully performed. Therefore, PCR inhibition must not be the main reason for the underestimate. Another suspected reason is that not all Cryptosporidium oocysts in river water samples contain the same amount of the 18S rRNA molecules. Although it has been reported that the 18S rRNA molecules of Cryptosporidium oocysts are more stable than mRNA molecules like b-tubulin (Widmer et al., 1999); reportedly, the number of copies of 18S rRNA molecules in Cryptosporidium oocysts significantly decreased when those were inactivated by heat shock (95  C

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 8 7 e1 9 4

for 4 h) (Fontaine and Guillot, 2003), which indicates that the number of copies of 18S rRNA molecules would reflect the viability of Cryptosporidium oocysts. Therefore, if dead or damaged Cryptosporidium oocysts existed in river water samples, both the number of copies of 18S rRNA molecules and the converted values of Cryptosporidium oocysts as determined by two RT-PCR assays would be small. It was reported that there were many dead (non-viable) Cryptosporidium oocysts in the effluent discharged from many swine wastewater treatment facilities that were one of the main pollution sources in river water (Jenkins et al., 2010), which indicates that dead Cryptosporidium oocysts exist in river water depending on the place and time. Hence, a few data (Sample K and M in Table 2) quantified by two RT-PCR assays would be much lower than those by microscopic observation.

5.

Conclusion

In this study, we developed an ABC-RT-PCR assay for simple and cost-effective quantification of Cryptosporidium oocysts in water samples. The following are the main outcomes of this study. (1) The standard curve of the ABC-PCR targeting 18S rRNA of Cryptosporidium spp. had a good fitting to a rectangular hyperbola with high correlation coefficient, and standard deviation of triplicate determinations was low. The ABCPCR assay was quantitative and reproducible for the measurement of number of copies of 18S rRNA genes of Cryptosporidium spp. (2) Concentrations of Cryptosporidium oocysts in river water samples were successfully quantified by the developed ABC-RT-PCR assay. The quantified values by the ABC-RTPCR assay very closely resembled those by the real-time RT-PCR assay. (3) Most quantified values by the ABC-RT-PCR assay were comparable to those by microscopic observation. However, a few values quantified by the ABC-RT-PCR assay were much smaller than those by microscopic observation, which indicates that dead or damaged Cryptosporidium oocysts, which have only a small number of copies of 18S rRNA molecules, exist in river water depending on the place and time. In this case, the sensitivity of the ABCRT-PCR assay would be lower than that by microscopic observation. (4) As the only equipment that is needed for the ABC-RT-PCR assay is a simple fluorometer and a relatively inexpensive thermal cycler, this method can markedly reduce time and cost to quantify Cryptosporidium oocysts in water samples.

Acknowledgments This study was financially supported by a Grant-in-Aid for Research Activity Start-up (21860094), Japan Society for the Promotion of Science (JSPS).

193

references

Abrahamsen, M.S., Templeton, T.J., Enomoto, S., Abrahante, J.E., Zhu, G., Lancto, C.A., Deng, M., Liu, C., Widmer, G., Tzipori, S., Buck, G.A., Xu, P., Bankier, A.T., Dear, P.H., Konfortov, B.A., Spriggs, H.F., Iyer, L., Anantharaman, V., Aravind, L., Kapur, V., 2004. Complete genome sequence of the apicomplexan, Cryptosporidium parvum. Science 304 (5669), 441e445. Altschul, S.F., Gish, W., Miller, W., Myers, E.W., Lipman, D.J., 1990. Basic local alignment search tool. Journal of Molecular Biology 215 (3), 403e410. Carpenter, C., Fayer, R., Trout, J., Beach, M.J., 1999. Chlorine disinfection of recreational water for Cryptosporidium parvum. Emerging Infectious Diseases 5 (4), 579e584. Fontaine, M., Guillot, E., 2003. Study of 18S rRNA and rDNA stability by real-time RT-PCR in heat-inactivated Cryptosporidium parvum oocysts. FEMS Microbiology Letters 226 (2), 237e243. Garce´s-Sanchez, G., Wilderer, P.A., Munch, J.C., Horn, H., Lebuhn, M., 2009. Evaluation of two methods for quantification of hsp70 mRNA from the waterborne pathogen Cryptosporidium parvum by reverse transcription real-time PCR in environmental samples. Water Research 43 (10), 2669e2678. Jenkins, M.B., Liotta, J.L., Lucio-Forster, A., Bowman, D.D., 2010. Concentrations, viability, and distribution of Cryptosporidium genotypes in lagoons of swine facilities in the Southern Piedmont and in coastal plain watersheds of Georgia. Applied and Environmental Microbiology 76 (17), 5757e5763. Keegan, A.R., Fanok, S., Monis, P.T., Saint, C.P., 2003. Cell cultureTaqman PCR assay for evaluation of Cryptosporidium parvum disinfection. Applied and Environmental Microbiology 69 (5), 2505e2511. King, B.J., Keegan, A.R., Monis, P.T., Saint, C.P., 2005. Environmental temperature controls Cryptosporidium oocyst metabolic rate and associated retention of infectivity. Applied and Environmental Microbiology 71 (7), 3848e3857. Kurata, S., Kanagawa, T., Yamada, K., Torimura, M., Yokomaku, T., Kamagata, Y., Kurane, R., 2001. Fluorescent quenching-based quantitative detection of specific DNA/RNA using a BODIPY((R)) FL-labeled probe or primer. Nucleic Acids Research 29 (6), e34. Miller, W.A., Gardner, I.A., Atwill, E.R., Leutenegger, C.M., Miller, M.A., Hedrick, R.P., Melli, A.C., Barnes, N.M., Conrad, P. A., 2006. Evaluation of methods for improved detection of Cryptosporidium spp. in mussels (Mytilus californianus). Journal of Microbiological Methods 65 (3), 367e379. Ministry of Health, Labor and Welfare, Japan, 2007. Methods for the Detection of Cryptosporidium Oocysts, Giardia Cysts and Indicator Microorganisms in Water Supply Systems Tokyo, Japan (in Japanese). Miyata, R., Adachi, K., Tani, H., Kurata, S., Nakamura, K., Tsuneda, S., Sekiguchi, Y., Noda, N., 2010. Quantitative detection of chloroethene-reductive bacteria Dehalococcoides spp. using alternately binding probe competitive Polymerase Chain Reaction. Molecular and Cellular Probes 24 (3), 131e137. Nichols, G., 2008. Epidemiology. In: Fayer, R., Xiao, L. (Eds.), Cryptosporidium and Cryptosporidiosis, second ed. IWA Publishing, London, UK, pp. 79e118. Noda, N., Tani, H., Morita, N., Kurata, S., Nakamura, K., Kanagawa, T., Tsuneda, S., Sekiguchi, Y., 2008. Estimation of single-nucleotide polymorphism allele frequency by alternately binding probe competitive polymerase chain reaction. Analytica Chimica Acta 608 (2), 211e216. Oda, T., Ito, H., Yano, H., Rai, S.K., Kawabata, M., Inoue, M., Uga, S., 2002. Use of hydrophilic polytetrafluoroethylene (HPTFE) membrane in detecting Cryptosporidium oocysts from source

194

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 8 7 e1 9 4

water. Japanese Journal of Water Treatment Biology 38 (3), 153e158. O’Donoghue, P.J., 1995. Cryptosporidium and cryptosporidiosis in man and animals. International Journal of Parasitology 25 (2), 139e195. Peeters, J.E., Maza´s, E.A., Masschelein, W.J., Villacorta Martiez de Maturana, I., Debacker, E., 1989. Effect of disinfection of drinking water with ozone or chlorine dioxide on survival of Cryptosporidium parvum oocysts. Applied and Environmental Microbiology 55 (6), 1519e1522. Ramirez, N.E., Sreevatsan, S., 2006. Development of a sensitive detection system for Cryptosporidium in environmental samples. Veterinary Parasitology 136 (3e4), 201e213. Tani, H., Kanagawa, T., Kurata, S., Teramura, T., Nakamura, K., Tsuneda, S., Noda, N., 2007. Quantitative method for specific nucleic acid sequences using competitive polymerase chain

reaction with an alternately binding probe. Analytical Chemistry 79 (3), 974e979. Torimura, M., Kurata, S., Yamada, K., Yokomaku, T., Kamagata, Y., Kanagawa, T., Kurane, R., 2001. Fluorescencequenching phenomenon by photoinduced electron transfer between a fluorescent dye and a nucleotide base. Analytical Science 17 (1), 155e160. Widmer, G., Orbacz, E.A., Tzipori, S., 1999. b-tubulin mRNA as a marker of Cryptosporidium parvum oocyst viability. Applied and Environmental Microbiology 65 (4), 1584e1588. Yagita, K., Izumiyama, S., Tachibana, H., Masuda, G., Iseki, M., Furuya, K., Kameoka, Y., Kuroki, T., Itagaki, T., Endo, T., 2001. Molecular characterization of Cryptosporidium isolates obtained from human and bovine infections in Japan. Parasitology Research 87 (11), 950e955.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 9 5 e2 0 4

Available online at www.sciencedirect.com

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

Flux patterns and membrane fouling propensity during desalination of seawater by forward osmosis Zhen-Yu Li, Victor Yangali-Quintanilla, Rodrigo Valladares-Linares, Qingyu Li, Tong Zhan, Gary Amy* Water Desalination and Reuse Center (WDRC), King Abdullah University of Science and Technology (KAUST), Thuwal 23955-6900, Saudi Arabia

article info

abstract

Article history:

The membrane fouling propensity of natural seawater during forward osmosis was

Received 16 August 2011

studied. Seawater from the Red Sea was used as the feed in a forward osmosis process

Received in revised form

while a 2 M sodium chloride solution was used as the draw solution. The process was

18 October 2011

conducted in a semi-batch mode under two crossflow velocities, 16.7 cm/s and 4.2 cm/s.

Accepted 22 October 2011

For the first time reported, silica scaling was found to be the dominant inorganic fouling

Available online 31 October 2011

(scaling) on the surface of membrane active layer during seawater forward osmosis. Polymerization of dissolved silica was the major mechanism for the formation of silica

Keywords:

scaling. After ten batches of seawater forward osmosis, the membrane surface was covered

Forward osmosis

by a fouling layer of assorted polymerized silica clusters and natural organic matter,

Seawater

especially biopolymers. Moreover, the absorbed biopolymers also provided bacterial

Fouling

attachment sites. The accumulated organic fouling could be partially removed by water

Scaling

flushing while the polymerized silica was difficult to remove. The rate of flux decline was

Natural organic matter

about 53% with a crossflow velocity of 16.7 cm/s while reaching more than 70% with a crossflow velocity of 4.2 cm/s. Both concentration polarization and fouling played roles in the decrease of flux. The salt rejection was stable at about 98% during seawater forward osmosis. In addition, an almost complete rejection of natural organic matter was attained. The results from this study are valuable for the design and development of a successful protocol for a pretreatment process before seawater forward osmosis and a cleaning method for fouled membranes. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Due to increasing population, industry, agriculture, economic growth and urbanization, water supply will undoubtedly become more scarce in the years to come. New technologies are being explored to recycle and reuse treated wastewater and/or desalinate seawater or brackish water. Membranes have been largely used for specialized applications for several decades, largely for water treatment to supply the growing demand for fresh water. In addition to

thermal technologies, such as Multi Stage Flash, seawater (SW) desalination with membrane technologies including reverse osmosis (RO) and nanofiltration (NF) have been successfully practiced and commercialized in areas proximate to the ocean. However, both RO and NF are processes driven by a high hydraulic pressure. These processes require an intensive energy input and the water recovery is often limited. Membrane fouling, brine disposal and environmental impacts are also critical constraints to RO and NF systems.

* Corresponding author. Tel.: þ966 2 8082385. E-mail addresses: [email protected], [email protected] (G. Amy). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.051

196

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 9 5 e2 0 4

Forward osmosis (FO) is a novel membrane process that can potentially be used as an energy-saving alternative to conventional membrane processes, such as RO in water treatment. The driving force in the FO process is an osmotic pressure created by the osmotic gradient when a high concentration of a saline solution flows along one side of the membrane versus a low concentration of feed along the other side. Since the FO process utilizes natural osmosis as the pressure source, the energy consumption can be significantly reduced compared to other pressure driven membrane processes (e.g., RO). FO has been considered as a possible technology for water treatment and related processes. Lab and pilot scale research of FO have been conducted for many applications, such as desalination (McCutcheon et al., 2005, 2006), concentration of brine (Tang and Ng, 2008), hybrid process for water reuse (Cath et al., 2010), power generation (Loeb, 2002; Achilli et al., 2009a), osmotic bioreactor (Achilli et al., 2009b), liquid food concentration (Petrotos and Lazarides, 2001), and medical and pharmaceutical applications (Wright et al., 2003; Yang et al., 2009). Although FO has been increasingly studied for many applications, there are several research areas needed to move the FO process forward including innovation of favorable FO membranes and modules, understanding of FO membrane fouling, and identification of an appropriate draw solution (DS). Most work has been focused on developing new FO membrane (such as well-constructed cellulose acetate FO membrane (Zhang et al., 2010), high performance thin film membrane (Yip et al., 2010) and thin film composite FO hollow fiber membrane (R. Wang et al., 2010)) and draw solutes (such as ammoniaecarbon dioxide (McCutcheon et al., 2005), water soluble magnetic nanoparticles (Ling et al., 2010)). However, very few studies on FO membrane fouling under natural water conditions have been reported. In one study, Mi and Elimelech (2008) researched organic fouling on the FO membrane by using some model foulants including humic acid, alginate and protein, all foulants caused flux decline during the FO process. Another study from the same team (Mi and Elimelech, 2010a) showed that alginate fouling on an FO membrane was almost fully reversible. Y. Wang et al. (2010) used a latex microparticle as a model foulant and observed the surface coverage of the FO membrane by latex particles. Lay et al. (2010) employed a silicon dioxide nanoparticle as a model foulant and explained a slow flux decline phenomenon in FO. Gypsum scaling was also able to form on the membrane during the FO process, but was fully reversible by water rinsing (Mi and Elimelech, 2010b). However scaling was not observed on the FO membrane in an osmotic membrane bioreactor with synthetic wastewater as the feed (Lay et al., in press). When activated sludge was used as the feed in an osmotic membrane bioreactor, both reversible and irreversible fouling were not found on the FO membrane under selected operational condition (Cornelissen et al., 2008). So far, most studies of the FO process have only been based on some simple model foulants for fouling analysis in shortterm tests. Although FO is considered to be a promising technology for SW desalination, there have been no published studies on fouling propensity of FO with real SW, particularly in long-term tests. In the present study, we investigated flux patterns and membrane fouling propensity of FO with natural

SW as the feed (SWFO). SWFO experiments were performed for 18e30 day periods. To the best of our knowledge, these are the first results about fouling on an FO membrane surface with real SW as the feed. The results reported in this study can be valuable for developing a successful protocol of pretreatment and membrane cleaning during SWFO.

2.

Materials and methods

2.1.

Feed and draw solution

Natural SW from the Red Sea was used as the feed in this study. SW was collected from the line that provides natural SW to a commercial RO desalination plant at the King Abdullah University of Science and Technology (KAUST). Before the SWFO process, natural SW was pre-filtered by a capsule filter (pore size 10 mm, Whatman, US) to remove large particles. The quality of SW pre-filtered by a 10 mm capsule filter is given in Table 1. In order to prevent the growth of bacteria and algae which may change the properties of the feed, 0.02% NaN3 was added to the pre-filtered SW and SWFO was operated under dark condition to inhibit bioactivity. In addition, 4% NaCl solution (with 0.02% NaN3) was used as a simulated seawater (SSW) for a baseline measurement. The conductivity of the SSW was adjusted to 54  1 mS/cm which was equal to SW used as the feed in this study. NaCl is also the most commonly used draw solute in FO because it has high solubility, low cost and is easily to reconcentrate to high concentration without scaling problems by conventional methods downstream of FO, such as low pressure RO, or possibly, membrane distillation (MD). The draw solution (DS) was prepared by dissolving NaCl in deionized water obtained from a Milli-Q ultrapure water purification system (Millipore, MA) at a concentration of 2 M.

Table 1 e Characterization of SW pre-filtered by a 10 mm capsule filter. Analyte Conductivity (mS/cm) pH DOC (mg/L) UVA254 (1/cm) SUVA (L/mg m) TDS (mg/L) Sodium (mg/L) Potassium (mg/L) Calcium (mg/L) Magnesium (mg/L) Silicon (mg/L) Iron (mg/L) Strontium (mg/L) Boron (mg/L) Bicarbonate (mg/L) Carbonate (mg/L) Chloride (mg/L) Fluoride (mg/L) Sulfate (mg/L)

SW 54  1 8.2  0.1 0.98 0.016 1.63 37,400 11,340 477 557 1427 0.24 0.12 7.0 2.2 137 7.2 21,473 1.7 2312

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 9 5 e2 0 4

2.2.

Forward osmosis set-up and fouling tests

A commercial flat sheet FO membrane was used in this study. The membrane is made of cellulose triacetate (CTA) embedded about a polyester (PES) screen mesh (HTI, LLC, Albany, OR). The active layer (AL) is the shiny side of the membrane (Fig. 1). Before using, the membrane was soaked in deionized (DI) water for 24 h to remove the glycerin which was used to replace the water during shipment. A bench-scale FO membrane system was employed in this study (see Supplementary content). The membrane cell was structured with equal channels on both sides providing an effective membrane area of 2  10 cm2 and a channel height of 0.2 cm for filtration. Two peristaltic pumps (Core-Parmer) were applied for the circulation of feed and draw solutions, respectively, to generate equal fluid flows with the same crossflow velocity in closed loops on each side of membrane. The AL of the membrane was set against the feed. It is smoother than the back side (support layer) and can avoid the effect of surface roughness on the formation of fouling layer. A digital scale (Mettler Toledo) connected to a computer was used to monitor the weight loss of the feed tank at intervals of 20 min during FO, and the weight change of the feed was converted to the permeate flux. All experiments were performed at a temperature of 21  1  C under dark conditions. During FO process, a low crossflow rate is preferred to reduce energy consumption. In order to study the effect of crossflow on the performance of the FO process, the crossflow was varied from a low velocity (4.2 cm/s, i.e. crossflow rate of 100 mL/min) to a high velocity (16.7 cm/s, i.e. crossflow rate of 400 mL/min). The FO process was conducted in a semi-batch mode. Each batch was started with 1 L of feed and 1 L of DS and interrupted when 300 mL of permeate was extracted from the feed to the DS. Then concentrated feed and diluted DS were replaced by fresh aliquots and the SWFO process was continued to the next batch. Before SW was used as the feed, a baseline measurement with SSW was performed. During SWFO, the decline of flux is related to two factors, i.e., formation of fouling and loss of osmotic pressure caused by external concentration polarization (ECP) and internal concentration polarization (ICP). Since SSW contained only NaCl and NaN3 in DI water, there was not an effect from other foulants such as natural organic matter (NOM). The decline of flux was only due to the combined effects from the concentrative ECP (CECP) on the feed side,

197

dilutive ICP (DICP), and dilutive ECP (DECP) on the DS side (Fig. 2a). After the baseline measurement, SW was employed as the feed. The fouling layer would form on the AL of the membrane. A fouling resistance and a concentrative ICP (CICP) in the fouling layer (Fig. 2b) were created. The fouling related CICP could cause an enhanced osmotic pressure in the fouling layer on the feed side. This enhanced osmotic pressure is due to not only the accumulation of salt from the feed bulk into the fouling layer on the membrane surface but also the revise diffusion of the salt from DS to the feed side (Lee et al., 2010). The salt diffused from DS to the feed was trapped by the fouling layer and also contributed to the enhanced osmosis pressure on the feed side. The enhanced osmotic pressure within fouling layer reduced the net driving force, consequently caused the significant decline of flux. When SW was used as the feed (i.e. SWFO), a total of 10 batches (B1 to B10) were operated with each crossflow velocity before cleaning or replacing the membrane. In other words,3 L of permeate passed through the membrane under each crossflow velocity. In each batch, the first 40 min was applied to stabilize the filtration system. The initial flux of each batch was measured from the filtration time of 40 min. The average flux of each batch was calculated by Equation (1): Jave ¼

DV At

(1)

where Jave is the average flux of each batch of FO; DV is the volume of permeate extracted from the feed to DS in each batch (300 mL in this study); A is the effective area of the membrane (20 cm2 in this study); t is the filtration time of each batch. In order to explore the reversibility of the fouling, the fouled membrane was collected after B10 of SWFO with the crossflow velocity of 4.2 cm/s. The fouled membrane was flushed in situ by DI water at a crossflow velocity of 16.7 cm/s for 10 min. Then, the feed was switched to SW again and one more batch of SWFO (i.e. B11) was processed with a crossflow velocity of 4.2 cm/s.

2.3.

Analytical methods

The cations and anions in SW were analyzed by Inductively Coupled Plasma/Mass Spectrometry (ICP-MS 7500, Agilent, US) and Ion Chromatography (ICS 1600, Dionex, US). Total organic carbon (TOC) and dissolved organic carbon (DOC) were

Fig. 1 e HTI CTA FO membrane (a, top view of the membrane under optical microscopy; b, cross-section of the membrane under optical microscopy; c, cross-section of the membrane under scanning electron microscopy; 1, CTA layer for filtration and salt rejection; 2, embedded PES mesh screen for mechanical support; scale bar 100 mm in a and c, 50 mm in b).

198

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 9 5 e2 0 4

a

b

Salt leakage C5

Water flux

Salt leakage C5

Water flux

C4 SSW

C4 SW

C3

Δπbulk Δπeff

C1

C3 Δπeff

DS

C2

(The Baseline Measurement)

DS

C2 C1

Support layer Rejection layer

Δπbulk

C2 Support layer Rejection layer Fouling layer

(SWFO)

Fig. 2 e Illustrations of driving force profiles, expressed as the gradient of salt concentration during the baseline measurement and SWFO (C1 < C2 due to CECP; C2 < C2f due to fouling related CICP; C3 < C4 due to DICP; C4 < C5 due to DECP; Dpbulk is the theoretical osmotic pressure; Dpeff is the effective osmotic pressure).

analyzed by an Organic Carbon Analyzer (TOC-VCPH, Shimadzu, Japan). Total dissolved solid (TDS) was measured by the standard methods from Eaton et al. (2005). NOM in SW and DS was characterized by Liquid ChromatographyeOrganic Carbon Detection (LCeOCD, Model 8, DOCLabor, Germany). The conductivity of the feed and DS was monitored by a digital conductivity meter (Lab960, Schott, Germany) and converted to the salt concentration for the calculation of salt rejection. The osmotic pressure of feed and DS was measured by a Cryoscopic Osmometer (Osmomat 030, Gonotec, Germany) and expressed as Osmolality (Osmol/kg). The fouling on the membrane surface was analyzed by Scanning Electron MicroscopyeEnergy Dispersive X-ray Apparatus (SEMeEDX, Magellan, FEI). The biopolymer (extracellular polymeric substances, EPS) on the AL of the membrane was stained by Alcian Blue 8GX and 40 ,6-diamidine-2-phenylindole (DAPI) solution (Bar-Zeev et al., 2009; Villacorte et al., 2009), then observed and photographed by an optical microscope (DP72, Olympus, Japan) with bright illumination and UV epifluorescence, respectively.

3.

Results and discussion

3.1.

Flux patterns during FO

The osmotic pressure of SSW and SW was 1.17  0.03 Osmol/ kg. When the same DS (i.e. 2 M NaCl solution) was used for all FO processes in this study, the approximate osmotic pressures of SSW and SW made the flux comparable between the baseline measurement and SWFO. The operational time for collecting 300 mL of permeate in each batch varied from 39 to 41 h with a crossflow velocity of 16.7 cm/s and from 48 to 64 h with a crossflow velocity of 4.2 cm/s. The flux patterns of the baseline measurement and SWFO were presented in Figs. 3 and 4. The reproducibility of the process was demonstrated by the data collected from the different batches of FO process. The flux decreased significantly in each batch of FO. The rate of flux decline was about 53% with a crossflow velocity of

16.7 cm/s while it reached more than 70% with crossflow velocity of 4.2 cm/s. In addition, the flux patterns were similar from Bbaseline (the batch of the baseline measurement) to B10 (the tenth batch of SWFO) with a crossflow velocity of 16.7 cm/ s while there were some changes during the initial several batches with a crossflow velocity of 4.2 cm/s. In this study, the AL of the membrane was set against the feed. According to the structure and separation capacity of the FO membrane, fouling on the feed side of the membrane could be classified as external types (such as cake layer or gel layer) which is similar to RO membrane fouling, rather than internal types (such as pore blocking) in other membrane processes like ultrafiltration or microfiltration. The flux decline during SWFO was attributed to a number of factors including diluting the feed, CECP, fouling related CICP and the fouling related resistance on the feed side, and concentrating the DS, DICP and DECP on the DS side. The combined effects from these factors caused the significant decrease of flux during SWFO. Higher crossflow velocity provided greater shear stress and vortex, that not only inhibiting the formation of a fouling layer on the membrane surface but also reduced ECP on both sides of the membrane by enhancing the back diffusion and convection of salts from the surface of the membrane or fouling layer to the bulk feed or DS. Therefore, a higher flux was obtained when a crossflow velocity of 16.7 cm/s was applied.

Fig. 3 e Flux patterns of the baseline measurement and SWFO with the crossflow velocity of 16.7 cm/s.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 9 5 e2 0 4

Fig. 4 e Flux patterns of the baseline measurement and SWFO with the crossflow velocity of 4.2 cm/s.

The variation of Jave of the baseline measurement and different stages of SWFO is represented in Fig. 5. The Jave of Bbaseline decreased about 19%when crossflow velocity was varied from 16.7 to 4.2 cm/s. During the baseline measurement, NaCl solutions at different concentrations were employed as the feed (SSW) and DS, respectively. Thus, there was not fouling related CICP and resistance in the baseline measurement. In addition, the HTI FO membrane uses an embedded PES mesh screen to provide mechanical strength of the membrane. The thickness and porosity of the support layer in this membrane is not as distinct as other thin film membranes. The membrane related DICP could be minimized. The decrease of Jave during the baseline measurement with low crossflow velocity mainly arose from promoted CECP on the feed side and DECP on the DS side. When SWFO was run, a minor decrease of Jave with a crossflow velocity of 16.7 cm/s was also observed between Bbaseline and B1. However, the decrease of Jave was much more significant and continued to B2 with the crossflow velocity of 4.2 cm/s. The decrease of Jave between baseline measurement and SWFO was probably due to the formation of fouling on the membrane surface when SW was used as the feed. Once SW contacted AL of the membrane, the foulants would start to adsorb onto the membrane surface by the interaction between foulants and 4.5 4

Jave (L/m2.hr)

3.5 3 2.5 2 1.5 1 0.5 0 baseline B1

B2

B3

B4

B5

B6

B7

B8

B9

B10

B11

Batch

Fig. 5 e Average flux of the baseline measurement and SWFO (,, the crossflow velocity of 16.7 cm/s; >, the crossflow velocity of 4.2 cm/s).

199

the membrane. For example, the biopolymeric fouling layer was observed on the surface of glass slides after soaking the slides in SW (Bar-Zeev et al., 2009). The adhesion of biopolymer on the membrane surface during SWFO could be enhanced by the permeate flow which can promote the transport of solutes and foulants from the feed bulk to the membrane surface. On the other hand, crossflow velocity had a greater impact on ECP rather than ICP. The formation of fouling layer on the membrane was more severe with low crossflow velocity while it could be reduced, but not eliminated, by high crossflow velocity. Therefore fouling related resistance and CICP negatively affected the flux with both high and low crossflow velocities, and this negative effect was more significant with low crossflow velocity. After initial batches of SWFO (B1 for the crossflow velocity of 16.7 cm/s, B1 and B2 for the crossflow velocity of 4.2 cm/s), Jave was nearly constant afterward to B10. Similar to other membrane processes, it can be hypothesized that the initial stage of SWFO was the main period for the build-up of fouling on the membrane surface, particularly for the semi-batch operation adopted for SWFO in this study. The fouling layer on the membrane surface formed in the initial batches of SWFO and became stable afterward. The fresh feed used in following batches, especially with high crossflow velocity, partially flushed the fouling layer rather than making it thicker or compacter. Consequently, the Jave of SWFO was stable in the later batches of SWFO. After B10 of SWFO with a crossflow velocity of 4.2 cm/s, the membrane was flushed in situ by DI water. The result (B11 with the crossflow velocity of 4.2 cm/s in Fig. 5) indicates that Jave could be partially recovered and almost equal to that of B1 with the same crossflow velocity (i.e., 4.2 cm/s) after flushing by DI water. The recovery of the flux by DI water flushing may be attributed to the removal of the fouling on the membrane surface.

3.2.

Identification of major foulants

The surfaces of the original virgin membrane and fouled membranes collected from the different batches of SWFO were characterized by SEM images combined with EDX spectrometry graphs (Figs. 6 and 7). Only the weak peaks of carbon and oxygen were detected on the surface of the original virgin membrane (Fig. 6a). They were from the raw materials (CTA and PES) of the membrane and probably the residual glycerin from manufacture process of the membrane. After 10 batches of SWFO with a crossflow velocity of 16.7 cm/s, the membrane surface was completely covered by a fouling layer (Fig. 6b). The fouling layer was made up of the scaling-like foulants surrounded by biopolymeric substances. EDX graphs reveal some strong peaks from the fouled membrane surface. The strong peaks of sodium and chlorine were definitely due to the salinity of SW. The enhanced peaks of oxygen and carbon were due to the NOM in the fouling layer. Although the amount of silica in SW is much lower than other inorganic elements (e.g. calcium and magnesium), a strong peak of silicon was observed while other inorganic elements were negligible on the surface of the fouled membrane. Thus, the scaling-like foulants could be due to the deposition of silica.

200

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 9 5 e2 0 4

Fig. 6 e SEMeEDX graphs of the membrane surface (a, the original virgin membrane; b, the fouled membrane after B10 of SWFO with a crossflow velocity of 16.7 cm/s).

Silica is generally found in water supplies in three different forms, namely, dissolved silica (also known as ‘reactive silica’, or monosilicic acid), colloidal silica and particulate silica (Comb, 1996). In order to identify the major form of silica in SW used in this study, a parallel analysis was conducted by measuring silicon concentration in SW pre-filtered by 10 and 0.45 mm capsule filters, respectively. There was not a significant difference in silicon concentration in the 10 mm filtered SW versus 0.45 mm filtered SW. It could be concluded that the dominant form of silica in SW used as the feed in this study was dissolved silica as monosilicic acid (H4SiO4). Thus, the formation of silica scaling was likely due to the polymerization of dissolved silica on the membrane surface rather than the deposition of colloidal or particulate silica. According to the flux results, the fouling should be more severe with the low crossflow velocity. Fig. 7a and b were obtained from the membranes collected from B1 and B11 of SWFO with the crossflow velocity of 4.2 cm/s. After the B1 of SWFO, some particle-like foulants were deposited on the membrane (Fig. 7a). Except carbon and oxygen, the EDX graph shows the strong peaks of sodium and chlorine and the weak peak of silica. At some sites, the polymerization of silica (marked by the arrows in Fig. 7a) on the membrane surface was initiated. Fig. 7a also reveals that the biopolymeric fouling was not severe during B1 of SWFO with the crossflow velocity of 4.2 m/s. A distinct polymerized silica cluster was observed in Fig. 7b. B11 could be considered to be the SWFO process with a pre-fouled

membrane. The distinct polymerized silica in Fig. 7b was formed in previous batches of SWFO (i.e. B2eB10) before B11. It was found that the biopolymeric foulants were flushed out by DI water after B10 of SWFO. After water flushing, the biopolymeric substances disappeared and the residual silica fouling showed a typical pattern of polymerized silica (Sahachaiyunta et al., 2002) rather than deposition of colloidal or particulate silica (Fig. 7b). This indicates the reversibility of the organic fouling layer and explains the increase of permeate flux in B11 (Figs. 4 and 5) as well. The biopolymeric fouling related CICP and resistance could be the major reasons for the decrease of flux in SWFO with a low crossflow velocity. The results reveal that the organic fouling could be reduced by maintaining the crossflow velocity at a certain level during SWFO or water flushing without any chemical treatment after SWFO. The polymerization of silica during SWFO could be initiated by associating the dissolved silica onto the membrane surface when silica was concentrated by CECP during the initial period of SWFO. The hydroxyl group in CTA could play a role in attracting the silica to the membrane surface (Sheikholeslami and Tan, 1999) while metals such as magnesium, calcium and iron in SW can facilitate the polymerization of silica (Sahachaiyunta et al., 2002; Sheikholeslami et al., 2002). During SWFO, CECP led to an increase of salts (such as CaCl2, MgCl2) and silica concentration near the membrane surface. Polymerization of silica would be significantly accelerated by increased concentration of silica with the presence

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 9 5 e2 0 4

201

Fig. 7 e SEMeEDX graphs of the membrane surface (a, the fouled membrane after B1 of SWFO with crossflow velocity of 4.2 cm/s; b, the fouled membrane after B11 of SWFO with crossflow velocity of 4.2 cm/s).

of calcium and magnesium (Sheikholeslami and Tan, 1999). In addition, in the form of monosilicic acid, silica is relatively unionized at most natural pH levels, with about 10% of monosilicic acid ionized at a pH of 8.5 which is very close to the pH of SW. The polymerization of monosilicic acid is favored by partial ionization (Sheikholeslami et al., 2002). The monosilicic acid could initially and rapidly form polysilicic acid of low molecular weight and then large polymeric species (Comb, 1996). Once the polymerized silica fouling forms on the membrane, it is difficult to remove while deposition of colloidal silica and particulate silica can be washed away by physical flushing (Koo et al., 2001). The hypothesized process for the polymerization of silica during SWFO is demonstrated in Fig. 8. During the polymerization process, monosilicic acid and polysilicic acid also reacts with organic compound to form

OH HO

HO

OH

O O

OH

O O

OH

O

O

FO membrane Fig. 8 e Polymerization of dissolved silica on the membrane surface during SWFO.

anhydrides (Ning, 2002) that promotes NOM fouling. Furthermore, the polymerized silica clusters change the morphology of the membrane surface and work as a scaffold in the fouling layer. The space inside the polymerized silica clusters could provide favorable sites without strong hydrodynamic vortex and mixing for the accumulation and adhesion of NOM, especially biopolymers. Therefore, a fouling matrix comprising polymerized silica and adhered NOM can form on the membrane surface during the SWFO process. Silica fouling is one of the major unresolved problems during brackish water RO due to the high concentration of silica in brackish water. Open ocean seawater contains much less silica than brackish water. Therefore this compound is not considered as a typical cause of mineral fouling during seawater RO (SWRO). Calcium carbonate and magnesium hydroxide are the most common causes for SWRO membrane scaling. However, polymerized silica was found to be dominant inorganic foulant while scaling caused by calcium and magnesium was not found during SWFO in this study. The combined effects from the CTA membrane properties (containing hydroxyl group), SW properties (pH, salt content) and hydrodynamics of the FO process (CECP and fouling related CICP) led to a rapid and prevailing formation of polymerized silica, creating a fouling layer complemented with NOM on the surface of the FO membrane. The silica-NOM fouling layer and its related CICP may hinder the accumulation and deposition of other mineral scalants on the membrane surface. In addition,

202

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 9 5 e2 0 4

3.3.

Salt and foulant rejection

The salt rejection was about 98% and almost constant for both the baseline measurement and all batches of SWFO (Fig. 10).

100 98

Salt rejection (%)

the driving force during SWFO was only osmotic pressure. The lack of high applied hydraulic pressure during SWFO also prevented, or retarded, the formation of other mineral scaling since calcium or magnesium scaling are favored at high operating pressure during RO and NF processes. The results suggest that a pretreatment such as coagulation (Sheikholeslami et al., 2002), chemical precipitation (Al-Rehaili, 2003), silica gel seeding (Bremere et al., 2000), or hybrid process involving ultrafiltration (Cheng et al., 2009) to remove silica from SW could improve the performance during the SWFO process. The presence of the biopolymers surrounding the polymerized silica clusters is visually demonstrated in Fig. 9. The membrane sample was the same as that in Fig. 6b. EPS was stained by Alcian Blue and visualized by microscopy. EPS (the blue substances in Fig. 9a and b) showed a sticky and gel-like form, and could play a critical role in the fouling related CICP. In addition, bacteria tended to accumulate at sites where EPS adhered to the membrane. Fig. 9b and c were photographed from a same membrane sample under bright illumination and UV epifluorescence, respectively. The fluorescence in Fig. 9c indicates the presence of bacteria. The position of bacteria in Fig. 9c is in accordance with the position of the EPS mass in Fig. 9b. EPS are mainly polysaccharides, proteins, nucleic acids and lipids. They provide the mechanical stability of biofilms, immobilize bacteria, and also serve as a nutrient source for bacteria (Flemming and Wingender, 2010). Considering the energy-saving and water recovery rate, it is desirable to maintain a FO process for long-term operation with a minimum, or even without crossflow. Therefore, biofouling may be a big challenge during SWFO since it is easy to form and adhere to the membrane surface. In addition to negative effects including reduction of membrane lifetime and increase of fouling resistance, biofouling layer is a favorable structure to hinder the back diffusion of salt from the membrane surface to the feed bulk, thus cause a significant drop of the effective osmotic pressure which is the only driving force during SWFO.

96 94 92 90 88 86 84 82 80

Batch Fig. 10 e Salt rejection during SWFO (-, crossflow velocity of 16.7 cm/s; ,, crossflow velocity of 4.2 cm/s).

Crossflow velocity did not show a significant effect on the salt rejection. It has been reported that the HTI CTA membrane can provide a salt rejection up to about 95% (McCutcheon et al., 2006; Cornelissen et al., 2008; Yip et al., 2010). During the RO process, the fouling layer, especially a biofilm on the membrane surface, leads to a reduction in salt rejection. The reason is known as biofilm enhanced osmotic pressure or cake enhanced concentration polarization (Hoek and Elimelech, 2003; Herzberg et al., 2009; Herzberg, 2010). In the case of FO, it has been proven that reverse NaCl flux (i.e., salt leakage) from the DS to the feed side may increase with increasing NaCl concentration in the DS solution when DI water is used as the feed (Phillip et al., 2010). It can be hypothesized that the principle of salt transport through an FO membrane is similar to an RO membrane. The driving force leading to the transport of NaCl through the membrane is the chemical potential gradient across the membrane and relates to the difference in concentration (chemical activity) across the membrane (Bhattacharyya and Williams, 1992). The salt transport comprises the diffusional salt flux across the dense (or active) layer and convective salt flux in the porous layer. The salt passage is enhanced by increasing concentration gradient of salt. In addition, convective salt flux is entrained by water flux in RO. During SWFO, the convective salt flux is opposed by the permeate flux. CECP, CICP, DECP and DICP on the feed and DS side, respectively, reduced the concentration

Fig. 9 e Visualization of the biopolymers on the membrane surface after B10 of SWFO with the crossflow velocity of 16.7 cm/s (a and b, EPS viewed with Alcian Blue stain under bright illumination; c, bacteria accumulated in EPS mass viewed with DAPI stain under UV epifluorescence; scale bar 200 mm in a, 20 mm in b and c).

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 9 5 e2 0 4

203

about 98% and almost all NOM foulants were rejected by the membrane during SWFO. Further research is needed to understand factors affecting the formation and composition of the fouling layer on the membrane during SWFO and to develop methods for controlling fouling.

Acknowledgments

Fig. 11 e NOM rejection during SWFO (blue line is SW; brown line is the DS collected from B1 of SWFO with the crossflow velocity of 4.2 cm/s). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

gradient between the feed and DS. The effective concentration gradient of salt from the DS side to the feed side was restricted by salt concentration at the interfaces of the fouling layer-AL and AL-support layer (C2f and C3 in Fig. 2b). Both C2f and C3 were dominated by ICP which could not be reduced by the crossflow of the feed and DS. Reduced chemical potential gradient combined with a water flux opposed to the salt transport led to the high and stable salt rejection during SWFO in this study. NOM in the feed and DS was characterized by LCeOCD (Fig. 11). The DS sample was collected from B1 of SWFO with the crossflow velocity of 4.2 cm/s. LCeOCD is able to fractionate NOM into about 10 classes of components that are detected by an organic carbon detector. An early complete rejection of NOM was attained during SWFO. A weak peak was detected from the DS at the elution time corresponding to humic-like substances while the biopolymer peak is completely absent. The absence of NOM, especially in the form of biopolymers in the DS indicates their retention and possible role in fouling.

4.

Conclusions

Flux patterns and membrane fouling propensity during SWFO were investigated in a semi-batch mode of the FO process. Flux decreased significantly in each batch during SWFO. The decrease of flux could be mainly attributed to ECP, fouling related resistance, and fouling related CICP. The membrane surface was covered by a fouling layer after SWFO for 18e30 day periods. The fouling layer was made up of scaling-like foulants surrounded by biopolymeric substances. The membrane scaling was caused by the polymerization of dissolved silica. The polymerized silica cluster facilitated the accumulation and deposition of NOM, especially biopolymers. A higher crossflow velocity could improve the flux by reducing ECP and fouling. Most NOM foulants in the fouling matrix could be removed by water flushing while silica scaling was difficult to hydraulically clean. The salt rejection was stable at

The authors acknowledge Dr. Lan Zhao, Mr. Qingxiao Wang, Mr. Guangchao Wang and Mr. Jian Ren in the Imaging and Characterization Laboratory of KAUST for support of imaging techniques.

Appendix. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.10.051.

references

Achilli, A., Cath, T.Y., Childress, A.E., 2009a. Power generation with pressure retarded osmosis: an experimental and theoretical investigation. Journal of Membrane Science 343, 42e52. Achilli, A., Cath, T.Y., Marchand, E.A., Childress, A.E., 2009b. The forward osmosis membrane bioreactor: a low fouling alternative to MBR processes. Desalination 239, 10e21. Al-Rehaili, A.M., 2003. Comparative chemical clarification for silica removal from RO groundwater feed. Desalination 159, 21e31. Bar-Zeev, E., Berman-Frank, I., Liberman, B., Rahav, E., Passow, U., Berman, T., 2009. Transparent exopolymer particles: potential agents for organic fouling and biofilm formation in desalination and water treatment plants. Desalination and Water Research 3, 136e142. Bhattacharyya, D., Williams, M.E., 1992. Reverse osmosis. In: Ho, W.S.W., Sirkar, K.K. (Eds.), Membrane Handbook. Van Nostrand Reinhold, New York, pp. 265e268. Bremere, I., Kennedy, M., Mhyio, S., Jaljuli, A., Witkamp, G.J., Schippers, J., 2000. Prevention of silica scale in membrane systems: removal of monomer and polymer silica. Desalination 132, 89e100. Cath, T.Y., Hancock, N.T., Lundin, C.D., Hoppe-Jones, C., Drewes, J.E., 2010. A multi-barrier osmotic dilution process for simultaneous desalination and purification of impaired water. Journal of Membrane Science 362, 417e426. Cheng, H.H., Chen, S.S., Yang, S.R., 2009. In-line coagulation/ ultrafiltration for silica removal from brackish water as RO membrane pretreatment. Separation and Purification Technology 70, 112e117. Comb, L.F., 1996. Silica: chemistry and reverse osmosis. Ultrapure Water 13, 41e43. Cornelissen, E.R., Harmsen, D., Korte, K.F.D., Ruiken, C.J., Qin, J.J., Oo, H., Wessels, L.P., 2008. Membrane fouling and process performance of forward osmosis membranes on activated sludge. Journal of Membrane Science 319, 158e168. Eaton, A.D., Clesceri, L.S., Rice, E.W., Greenberg, A.E., 2005. Standard Methods for the Examination of Water and Wastewater, 21st ed. Port City Press, Baltimore, Maryland, US. Flemming, H.C., Wingender, J., 2010. The biofilm matrix. Nature Reviews (Microbiology) 8, 623e633.

204

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 9 5 e2 0 4

Herzberg, M., 2010. Osmotic effects of biofouling in reverse osmosis (RO) processes: physical and physiological measurements and mechanisms. Desalination and Water Treatment 15, 287e297. Herzberg, M., Kang, S., Elimelech, M., 2009. Role of extracellular polymeric substances (EPS) in biofouling of reverse osmosis membrane. Environmental Science and Technology 43, 4393e4398. Hoek, E.M.V., Elimelech, M., 2003. Cake-enhanced concentration polarization: a new fouling mechanism for salt-rejecting membranes. Environmental Science and Technology 37 (24), 5581e5588. Koo, T., Lee, Y.J., Sheikholeslami, R., 2001. Silica fouling and cleaning of reverse osmosis membranes. Desalination 139, 43e56. Lay, W.C.L., Chong, T.H., Tang, C.Y., Fang, A.G., Zhang, J., Liu, Y., 2010. Fouling propensity of forward osmosis: investigation of the slower flux decline phenomenon. Water Science and Technology 61, 927e936. Lay, W.C.L., Zhang, Q., Zhang, J., McDougald, D., Tang, C.Y., Wang, R., Liu, Y., Fane, A.G. Study of integration of forward osmosis and biological process: membrane performance under elevated salt environment. Desalination, in press. Lee, S., Boo, C., Elimelech, M., Hong, S., 2010. Comparison of fouling behavior in forward osmosis (FO) and reverse osmosis (RO). Journal of Membrane Science 365, 34e39. Ling, M.M., Wang, K.Y., Chung, T.S., 2010. Highly water-soluble magnetic nanoparticles as novel draw solutes in forward osmosis for water reuse. Industrial and Engineering Chemistry Research, 5869e5876. Loeb, S., 2002. Large-scale power production by pressure-retarded osmosis using river water and sea water passing through spiral modules. Desalination 143, 115e122. McCutcheon, J.R., McGinnis, R.L., Elimelech, M., 2005. A novel ammoniaecarbon dioxide forward (direct) osmosis desalination process. Desalination 174, 1e11. McCutcheon, J.R., McGinnis, R.L., Elimelech, M., 2006. Desalination by ammoniaecarbon dioxide forward osmosis: influence of draw and feed solution concentrations on process performance. Journal of Membrane Science 278, 114e123. Mi, B.X., Elimelech, M., 2008. Chemical and physical aspects of organic fouling of forward osmosis membranes. Journal of Membrane Science 320, 292e302. Mi, B.X., Elimelech, M., 2010a. Organic fouling of forward osmosis membranes: fouling reversibility and cleaning without chemical reagents. Journal of Membrane Science 384, 337e345. Mi, B.X., Elimelech, M., 2010b. Gypsum scaling and cleaning in forward osmosis: measurements and mechanisms. Environmental Science and Technology 44, 2022e2028.

Ning, R.Y., 2002. Discussion of silica speciation, fouling, control and maximum reduction. Desalination 151, 67e73. Petrotos, K.B., Lazarides, H.N., 2001. Osmotic concentration of liquid foods. Journal of Food Engineering 49, 201e206. Phillip, W., Yong, J.S., Elimelech, M., 2010. Reverse draw solute permeation in forward osmosis: modeling and experiments. Environmental Science and Technology 44, 5170e5176. Sahachaiyunta, P., Koo, T., Sheikholeslami, R., 2002. Effect of several inorganic species on silica fouling in RO membranes. Desalination 144, 373e378. Sheikholeslami, R., Al-Mutaz, I.S., Tan, S., Tan, S.D., 2002. Some aspects of silica polymerization and fouling and its pretreatment by sodium aluminate, lime and soda ash. Desalination 150, 85e92. Sheikholeslami, R., Tan, S., 1999. Effects of water quality on silica fouling of desalination plants. Desalination 126, 267e280. Tang, W., Ng, H.Y., 2008. Concentration of brine by forward osmosis: performance and influence of membrane structure. Desalination 224, 143e153. Villacorte, L.O., Kennedy, M.D., Amy, G.L., Schippers, J.G., 2009. The fate of transparent exopolymer particles (TEP) in integrated membrane systems: removal through pretreatment process and deposition on reverse osmosis membranes. Water Research 43, 5039e5052. Wang, R., Shi, L., Tang, C.Y., Chou, S., Qiu, C., Fane, A.G., 2010. Characterization of novel forward osmosis hollow fiber membranes. Journal of Membrane Science 355, 158e167. Wang, Y., Wicaksana, F., Tang, C.Y., Fane, A.G., 2010. Direct microscopic observation of forward osmosis membrane fouling. Environmental Science and Technology 44, 7102e7109. Wright, J.C., Johnson, R.M., Yum, S.I., January/February 2003. Duros osmotic pharmaceutical systems for parenteral and sitedirected therapy. Drug Delivery Technology (e-journal) 3 (1). Yang, Q., Wang, K.Y., Chung, T.S., 2009. A novel dual-layer forward osmosis membrane for protein enrichment and concentration. Separation and Purification Technology 69, 269e274. Yip, N.Y., Tiraferri, A., Phillip, W.A., Schiffman, J.D., Elimelech, M., 2010. High performance thin-film composite forward osmosis membrane. Environmental Science and Technology 44, 3812e3818. Zhang, S., Wang, K.Y., Chung, T.S., Chen, H.M., Jean, Y.C., Amy, G., 2010. Well-constructed cellulose acetate membranes for forward osmosis: minimized internal concentration polarization with an ultra-thin selective layer. Journal of Membrane Science 360, 522e535.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 0 5 e2 1 7

Available online at www.sciencedirect.com

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

What are the costs and benefits of biodiversity recovery in a highly polluted estuary? M. Pascual a,b,*, A. Borja a,*, J. Franco a, D. Burdon b, J.P. Atkins c, M. Elliott b a

AZTI-Tecnalia, Marine Research Division, Herrera Kaia, Portualdea s/n, 20110 Pasaia, Spain IECS, Institute of Estuarine and Coastal Studies, University of Hull, Hull HU6 7RX, UK c Centre for Economic Policy, The Business School, University of Hull, Hull HU6 7RX, UK b

article info

abstract

Article history:

Biodiversity recovery measures have often been ignored when dealing with the restoration

Received 20 June 2011

of degraded aquatic systems. Furthermore, biological valuation methods have been applied

Received in revised form

only spatially in previous studies, and not jointly on a temporal and spatial scale. The

24 October 2011

intense monitoring efforts carried out in a highly polluted estuary, in northern Spain

Accepted 24 October 2011

(Nervio´n estuary), allowed for the economic valuation of the costs and the biological

Available online 2 November 2011

valuation of the benefits associated with a 21 years sewage scheme application. The analysis show that the total amount of money invested into the sewage scheme has

Keywords:

contributed to the estuary’s improvement of both environmental and biological features,

Biodiversity valuation

as well as to an increase in the uses and services provided by the estuary. However, the

Water treatment investment

inner and outer parts of the estuary showed different responses. An understanding of the

Recovery of aquatic systems

costs and trajectories of the environmental recovery of degraded aquatic systems is

Nervio´n estuary

increasingly necessary to allow policy makers and regulators to formulate robust, cost-

Basque country

efficient and feasible management decisions. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Due to continued habitat degradation (Halpern et al., 2008), there is an increasing need to restore degraded ecosystems (Lotze, 2010). Legislation worldwide, such as the Clean Water Act (USA) or the Water Framework Directive (WFD; 2000/60/ EC) and the Marine Strategy Framework Directive (MSFD) (Europe), include restoration of degraded aquatic habitats as one of their primary goals (Apitz et al., 2006; Borja et al., 2008a, 2010). Estuaries offer a wide range of societal benefits, some are commercial and can be readily valued in monetary terms, for example fish harvested for human consumption, whilst others are non-commercial, for example intrinsic biological

diversity and filtration functions (America, 2008). Thus, an estuary’s ecological restoration may be a worthwhile investment for society as it can lead to improvements or enhancements in the supply and quality of ecosystem services to society (Aronson et al., 2010). Additionally, the ecosystem services approach ensures that all the values of the ecosystem are captured, and these can be valued using a variety of means, including monetary, non-monetary and biological (see Atkins et al., 2011; Beaumont et al., 2007; Derous et al., 2007a). Most studies use the recovery of the ecosystems structure and functioning or the recovery of a specific biological ecosystem component/species as indicators of restoration (Borja et al., 2006b; Elliott et al., 2007; Gorostiaga et al., 2004; Mialet et al., 2010; Whitfield and Elliott, 2002) rather than

* Corresponding authors. AZTI-Tecnalia, Marine Research Division, Herrera Kaia, Portualdea s/n, 20110 Pasaia, Spain. Tel.: þ34 946574000. E-mail addresses: [email protected] (M. Pascual), [email protected] (A. Borja). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.053

206

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 0 5 e2 1 7

societal benefits. There are many examples of estuarine condition improvements due to sewage abatement actions worldwide (Aslan-Yilmaz et al., 2004; Brosnan and O’Shea, 1996; Conley and Josefson, 2001; Hawkins et al., 2002) including the notable example from the Thames estuary (Andrews, 1984; Attrill, 1998). However, until now, no studies have looked at a comparison between investment and biodiversity valuation. Many methods have recently been developed which value biodiversity directly (Balvanera et al., 2006; Beaumont et al., 2007; Derous et al., 2007a; Nijkamp et al., 2008; Pascual et al., 2011) or indirectly (using contingent valuation; choice experiments; etc.). Despite this, most measures of biodiversity valuation look at the spatial biodiversity, and no studies have approached biodiversity valuation on both a spatial and temporal scale. Furthermore, there are very few examples of long-term monitoring, that include different biological and physico-chemical data from both water and sediments, which would be useful for showing the recovery trajectories after remediation or restoration processes (Borja et al., 2010). One of these examples occurs at the Nervio´n estuary (northern Spain), where the intense monitoring system carried out for many years, gives the opportunity for observing the development of the ecosystem as water quality improves; providing a valuable record of the status of the different ecosystem components. The estuary of the Nervio´n was one of the most polluted areas on the northern coast of Spain (Cearreta et al., 2004; Borja et al., 2006a). This estuary, which harbours one of the most important ports in Spain, Bilbao, has suffered from serious environmental degradation as a result of many pollutant discharges (both industrial and domestic), since the 19th century, together with the development of the iron, steel and ship building industries and mining activities (Cearreta et al., 2004). This industrialisation led to a sharp increase in population, with the consequent intensification of domestic untreated waste water discharges into the estuary (Garcı´aBarcina et al., 2006). Furthermore, the original morphology of the estuary was strongly modified, with the consequent loss of wetlands and sand dunes. In recent times only 68% of its original extension remains, due to the channelling and straightening of its course, the diking of large intertidal areas, intense dredging activities to maintain navigation in the channel, etc. (Ferna´ndez Pe´rez, 2005). Hence, the Nervio´n estuary represents a good example of man induced alteration and is regarded as a heavily modified water body, according to the European WFD (Borja et al., 2009a). In 1979 the Sewage Scheme for the area was approved by the competent local water management authority (Consorcio de Aguas Bilbao-Bizkaia (CABB)), establishing as the overall objective the restoration of good aesthetic, sanitary and ecological conditions along the estuary and fixing a water quality standard of 60% dissolved oxygen saturation. The environmental clean-up of the catchment waters began in 1990 which included physical and chemical treatments whereas the biological treatment began in 2001 (Garcı´aBarcina et al., 2006). The latter includes the sewage treatment scheme, consisting of more than 300 km of sewer network, where the waters of more than one million

inhabitants are conveyed into a central waste water treatment plant (WWTP), with a biological treatment capacity of 6 m3 s1 (Fig. 1). As a consequence of this sewage scheme start-up, a change in the condition of the Nervio´n estuary from an ‘open-navigable-sewer’ to an aerobic tideway, supporting many ecosystem components, has occurred over the last 21 years (Garcı´a-Barcina et al., 2006). The water quality improvement of the Nervio´n estuary has been described elsewhere (Borja et al., 2006b, 2010; Garcı´aBarcina et al., 2006; Gonza´lez-Oreja and Sa´iz Salinas, 1999) throughout the different phases of the water treatment. There has been a total public investment of around V600 million, including support from national, Basque, and provincial governments, as well as through higher water service user charges (Barreiro and Aguirre, 2005). However, to date, the investment undertaken by private companies to the sewage scheme remains unknown. Hence, the aims of this study are: (i) to value the costs, economically, and the benefits, biologically, associated with the sewage scheme over the last 21 years and the subsequent ecological recovery that has taken place in the waters of the Nervio´n estuary, and (ii) to highlight the benefits of using these biological and economic valuation techniques as instruments to formulate robust, cost-efficient and feasible water treatment decisions as required by policy makers and regulators.

2.

Materials and methods

2.1.

Study area

The Nervio´n estuary is located on the northern coast of Spain (Fig. 1) and drains a watershed of about 1700 km2, which provides an annual average freshwater inflow of 25 m3 s1. The estuary has two areas: a narrow, relatively shallow and highly stratified channel of about 15 km in length, that crosses the metropolitan area of the city of Bilbao (hereafter, ‘inner part’) and a semi-enclosed coastal embayment, with an area of about 30 km2 and an average water depth of about 25 m (hereafter, ‘outer part’) (Fig. 1). Extensive monitoring of the area has taken place since 1989 (Franco et al., 2010), within the framework of local projects. A synthesis of the methods used and the ecosystem components sampled (which included zooplankton, macroalgae, macrobenthos and demersal fish) are given in Franco et al. (2010), Borja et al. (2006b, 2010) and Dı´ez et al. (2009).

2.2.

Databases

Economic, abiotic and biological temporal trends were analysed from 1989 to 2010. All the public and private economic environmental expenditure information, publicly available in the annual economic reports of local businesses and environmental incentive investments announcements in the official bulletins of the regional area, were gathered into an economic input database. However, although the economic investment data gathered were the best available, the total of the abatement costs should be treated with caution as the difficulties faced when accessing businesses pollution

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 0 5 e2 1 7

207

Fig. 1 e Study area: Nervio´n estuary location at the western part of the Basque Coast together with a scheme of both inner and outer parts of the estuary. The localization of the main Galindo’s Waste Water Treatment Plant (WWTP) is also highlighted.

abatement economic data allowed us to identify particular types of capital expenditure and operating costs only for some of the businesses, but not for others. Besides, a filter was applied in order to gather only the information for the sewage scheme actions from 1989 to 2010 and, following the Pollution Abatement Costs and Expenditures (PACE) 2005 survey guidelines (U.S. Census Bureau, 2008), expenditure actions were divided into four types of activities: treatment/capture, disposal, recycling, and pollution prevention (Table 1). Although certain expenditures may have had multiple benefits, for the purposes of this survey, consideration was only given to those for treatment/capture, for which pollution abatement was the primary purpose. When pollution abatement capital expenditures included any installation or equipment for the treatment/capture activities, only incremental capital expenditures and incremental operating costs additional to annual operating investments or maintenance costs were taken into account (following U.S. Census Bureau, 2008). These expenditures were budgeted and adjusted, where possible, as a time-frame cost according to the estimation of the average life time of the equipment. By doing so, we avoided referring to capital expenditures as one-off costs, which could also lead to an overestimation of the investment efforts at the beginning of each installation commissioning. In order to determine the aggregate economic effort being put

into abatement over time, annual investment expenditures were deflated by the Gross Domestic Product (GDP) deflator (World Bank). As a proxy for the water treatment results, the temporal trends in annual load of biochemical oxygen demand (BOD) and nutrient load discharges, evaluated as ammonia (NH3), were studied. The BOD was computed from the main sources: domestic, industrial, WWTP effluent and river pollution. These data were obtained from Garcı´a-Barcina et al. (2006) and were updated using CABB unpublished data. The biological information of the ecosystem components, for which detailed spatial distribution data were available (zooplankton, macroalgae, macrobenthos and demersal fish), were included and integrated in a database, in order to obtain a Biological Valuation Map (BVM) of the Nervio´n estuary for each of the 21 years of the sampling period. The zooplankton relative abundance database covered a total of 16 sampling years (1994e2009) (Villate et al., 2004; Aravena et al., 2009). Macroalgae sampling period database (2003, 2004 and 2006), with percentages of spatial cover, was only available for the hard-bottom substrata (Dı´ez et al., 2009). The macrobenthos was intensively sampled and studied during the period 1989e2010 for the soft-bottom substrata (Borja et al., 2006b, 2010) and only for three years (2004, 2006 and 2008) of the sampling period for hard-bottom substrata (Pagola-Carte and

208

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 0 5 e2 1 7

Table 1 e Division criteria table for the total sewage scheme abatement actions applied at the Nervio´n estuary (modified from U.S. Census Bureau, 2008). Activity category Pollution abatement capital expenditures

Treatment/capture

Recycling

Disposal Pollution prevention

Examples Purchase, installation and start-up costs of “end of pipe” pollution abatement equipment: e Absorption and filtering systems. e Emissions capture systems. e Oil/water separating systems. e Dewatering systems. e Loads control and capture systems. e Decontamination systems. e Treatment and purification systems. e Pollutant substances elimination processes and systems. e Pouring-off systems. e Distillation columns; compactors, etc. e Reuse, recovery and recirculation systems. e Water consumption reducing systems. e Recycling equipment. e Recycling facilities installations. Construction of waste storage facilities or retention ponds e Set-up; fit-up and renewal actuations. e Sewage management decisions. e Pollution prevention; disposal; efficiency; environmental impact and measures to emplace reporting. e Environmental appointment establishment (ISO 14001 and others).

Sa´iz Salinas, 2001). The soft-bottom database consisted of a set of sample sites where species abundances (per sampled surface area) were known, while the hard-bottom database samples only provided presence/absence data. Fish abundance data, from 1989 to 2010, were obtained from trawling surveys (Uriarte and Borja, 2009). Trawl data covering multiple grid cells were treated so that every grid cell visited by the trawl was given the abundance value of the entire trawl. Although there has been an important increase in the use of the Nervio´n estuary by seabirds (Soler et al., 2008; Borja et al., 2010), due to the lack of spatial seabird point data, the changes in this ecosystem component were excluded. Finally, a general data quality check was undertaken (geographical coordinates, dates, time, and taxonomy). The taxonomy was checked against ERMS (European Register of Marine Species), in order to avoid the use of synonymous taxa that could overestimate the number of species (Pascual et al., 2011).

2.3.

Methodology

The Biological Value (BV) of the four ecosystem components was analysed according to the Biological Valuation Methodology developed for the Belgian part of the North Sea by Derous et al. (2007a). The same methodology has already been applied by Pascual et al. (2011) to the entire Basque continental shelf. This methodology allows for the valuation of the overall improvement of the ecosystem biological components identified in the Nervio´n estuary. Due to the inherent differences of the two parts of the estuary (inner and outer), regarding the pollutant discharges and human pressures, the changes in the BV in both areas were assessed independently. In fact, these differences in human pressures were the reasons for dividing the estuary into separate inner

and outer water bodies for the implementation of the European WFD (Borja et al., 2006a). The minimum width of the inner part of the estuary is 50 m, and so it was decided to divide the whole study area into 25  25 m grid cells for the valuation of the ecosystem components. The aim of the BV methodology is to provide an integrated view on nature’s intrinsic non-anthropogenic value of the subzones, relative to each other, within a study area (Derous, 2007). For this study subzones equalled the 25  25 m division grid cells. By answering a set of assessment questions (Table 2a), within the database, through mathematical algorithms, it is possible to visualize all the biodiversity aspects linked to the biological and ecological valuation. These questions were determined at a European workshop, involving expert judgement, and focus on rarity and aggregation-fitness consequences (Derous et al., 2007b). As this methodology seeks to determine the costs and benefits of restoration, it is essential to address the question posed by both Elliott et al. (2007) and Simenstad et al. (2006), that is, ‘what are we restoring to?’. In this case, a literature review was undertaken to determine the different criteria for each assessment question and ecosystem component to reach reference condition values sensu the European WFD, as included in Borja et al. (2009a, b). When this information was not available (hard substratum macrobenthos and zooplankton), the maximum value was determined by the highest value for each of the assessment questions for each of the ecosystem components throughout the whole period considered in this study. A summary of the application criteria is shown in Table 2b. Questions 5 and 6 in Table 2a address the occurrence and quantity of Habitat-Forming (HF) or Ecologically Significant (ES) species. The selection of the species in each of these categories was the result of both a large and detailed literature

209

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 0 5 e2 1 7

Table 2 e a) Set of assessment questions, related to the different structure and processes of biodiversity (Derous et al., 2007a), b) Set of assessment questions criteria for each of the ecosystem components assessed. a) Question Code

Assessment questions (following Derous et al., 2007a) High counts/presence/covertures of many species?a The mean density values?b the abundance/relative abundance/coverture/presence of a commonc species very high in the subzone? the subzone characterized by mean biodiversity values? the subzone characterized by the presence of many rare species? the abundance/relative abundance/coverture of rare species high in the subzone? the abundance/relative abundance/coverture of habitat forming species high in the subzone? the abundance/relative abundance/coverture of ecologically significant species high in the subzone? the species richness in the subzone high?

Q1

Is the subzone characterized by

Q2

Is Is Is Is Is Is Is

Q3 Q4 Q5 Q6 Q7

b) Ecosystem component Data availability Zooplankton MB_Soft MB_Hard Macroalgae_Hard Dem. Fish

1994e2009 1989e2010 2004/2006/2008 2003/2004/2006 1989e2010

Q1

Q2

a

(count) Relative abund. Mean biodivd a (pres.) Presence a (cov.) Coverture e e b d

Q3

Q4 c

Rarity Rarityc Rarityc Rarityc Rarityc

Q5

Q6

Q7 

Relative abund. e 80% of Relative abund. n sp. Abund. Abund. 50% of abund.e n sp.d e Au or All C; H; DF; FF n sp. Cov. 75% of Cov. % of Cov.g n sp.f Abund. e %h %h

a When only abundance/presence/covertures data was available. b When only mean density values data was available, as occurring with the macrobenthos_hard ecosystem component data. c Rarity and Commonness defined as those <0.05% of the total number of individuals and as those >0.5% of total number of individuals, respectively (Gering et al., 2003). d According to Borja and Collins, 2004 (Table 18.1). e According to AMBI’s defined Ecological Groups IeV. f According to Borja et al., 2004. g According to Wells et al., 2007 (Table 3). h According to Uriarte and Borja, 2009.

review and local expert judgement on each ecosystem component. Where possible, this analysis is as objective as possible although subjectivity cannot always be excluded in this BV method and, therefore, the selection should be regarded as expert judgment assessment choices (Derous et al., 2007a). A detailed classification of the selected species, per ecosystem component, for each category and the criteria followed is shown in Table 3. Due to the lack of subzone specific data, quantitative scoring is often not possible and the subzones are weighted qualitatively, scored against each other, or semiquantitatively, ranking subzones in categories of high, medium or low value (Derous et al., 2007a). In this study, each of the ecosystem components was valued separately by averaging the scores of those assessment questions used, giving each assessment question an equal weight over the total score. The integrated BV of each of the subzones was then determined by averaging the values obtained for the different ecosystem components (when values were available). Five equally distanced biological value classes were used in the proposed scoring system (comprising very low, low, medium, high and very high) as this allowed for a better detection of patterns in values without losing too much detail (Pascual et al., 2011). In order to avoid possible bias, which could occur when the amount of information for each subzone was not equal (Breeze, 2004), and so improve interpretation of the results, ‘reliability’ and ‘sampling effort’ labels were attached to

each of the BV. Reliability and sampling effort labels display the quality and amount of data (respectively) used to assess the BV. In order to determine the significance of the relationships between variables, correlation analyses were performed using Statgraphics Plus 5.0 (Statsoft, Inc. 2000).

2.4.

Interpolation

The BV for each of the ecosystem components cannot be calculated for all cells of the Nervio´n estuary, but only at those locations sampled. However, values can be interpolated to give estimates at sites where no samples are available; GIS-aided inter- and extrapolation methods can be used to convert point data to surface data. This approach allows the creation of a full-coverage BV for each of the ecosystem components for the entire Nervio´n estuary (Pascual et al., 2011).

3.

Results

3.1.

Costs

The economic data (Fig. 2a) show that, almost V658 million have been spent to date on the sewage scheme on actions directed towards quality improvements in the Nervion’s estuarine conditions (99% of this expenditure came from

210

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 0 5 e2 1 7

Table 3 e List of selected a) Habitat-Forming (HF) and b) Ecologically-Significant (ES) species for each of the ecosystem components. a) Habitat-Forming (HF) species MB_Softb

Zooplankton e

Abra alba (5) Abra prismatica (5) Abra sp. (4) Cerastoderma edule (4) Scrobicularia plana (5) Tellina tenuis (5) Venus sp. (5)

MB_Hardb Balanus amphitrite (All) Balanus crenatus (All) Balanus perforatus (All) Balanus sp. (Au) Balanus trigonus (All) Chthamalus sp. (Au) Chthamalus stellatus(Au) Mytilaster minimus (All)

Mytilaster solidus (All) Mytilus galloprovincialis (Au) Ostrea edulis (All) Ostreidae (Au) Pollicipes pollicipes (All) Sabellaria spinulosa (All) Sabellidae (Au) Spongia caudigera (Au) Verruca stroemia (Au)

Macroalgae_Hardb

Dem. Fish

Bifurcaria bifurcata (5 e Corallina elongata (5) Corallina officinalis (5) Cystoseira baccata (5) Cystoseira tamariscifolia (5) Gelidium corneum (5) Gelidium spinosum (5) Halopteris filicina (5) Lichina pygmaea (5) Lithophyllum byssoides (5) Mesophyllum lichenoides (5) Stypocaulon scoparium (5) Verrucaria maura (5)

b) Ecologically-Significant (ES) species Zooplanktona

MB_Softc

MB_Hardb

Phyllum Cnidaria (5) Ecological Group I (5) Phyllum Rotifera (1) Ecological Group II (4)

Actinia equina (C) Amphiglena mediterranea (FF) Class Gastropoda (5) Ecological Group III (3) Apherusa jurinei (FF) Class Maxillopoda (5) Ecological Group IV (2) Aplysia punctata (H) Class Polychaeta (1) Ecological Group V (1) Bittium reticulatum (H) Order Tintinnida (5) Campecopea hirsuta (H) Genus Acartia (5) Caprella danilevskii (H) Caprella penaltis (C) Cymodoce truncata (DF) Dynamene bidentata (H) Eulalia viridis (C) Gastrochaena dubia (FF) Hyale perieri (H) Hyale spinidactyla (H) Hyale stebbingi (H) Ischyromene lacazei (H) Jassa falcata (FF)

Jassa marmorata (FF) Lasaea adansoni (FF) Melarhaphe neritoides (DF) Modiolula phaseolina (H) Modiolus barbatus (H) Musculus costulatus (H) Paracentrotus lividus (H) Patella depressa (H) Patella rustica (H) Patella ulyssiponensis (H) Patella vulgata (H) Platynereis dumerilii (H) Polydora sp. (DF) Syllis amica (C) Syllis gracilis (C) Tanais dulongii (FF)

Macroalgae_Hardd

Dem. Fishe

Phyllum Chlorophyta (1) Omnivorous (5) Phyllum Rodophyta (5) Piscivorous (5) Flat fish (5)

Key: 5 ¼ Very High Biological Value; 4 ¼ High Biological Value; Au ¼ Autogenic; All ¼ Allogenic; C ¼ Carnivores; H ¼ Herbivores; FF ¼ Filter feeders; DF ¼ Deposit feeders. a According to Uriarte and Villate, 2004. b According to Pascual et al. (2011). c According to AMBI’s defined Ecological Groups IeV. d According to Wells et al., 2007 (Table 3). e According to Uriarte and Borja, 2009.

CABB and 1% from annually totalling over years 1989, 1990, 1993, cided with the years equipment acquisitions ment facilities, etc.).

3.2.

private businesses). Investments V30 million were applied in the 1997, 1998 and 2000, which coinof major treatment and capture (storage tanks, waste water treat-

Benefits

The BOD of the industrial and domestic waste water inputs into the estuary has decreased since 1990, with a notable decrease in the BOD input by the WWTP from 2001 onwards (Fig. 2b). Uncontrolled BOD river pollution loads continued to fluctuate between these years. Ammonia loads also show an overall decreasing trend from 1990 onwards.

The BV of both inner and outer parts of the Nervio´n estuary (Fig. 2c) shows some fluctuations throughout the CABB sewage scheme period (1989e2010). Decreases in the inner part occurred for the 1992e1993, 1999e2000 and 2001e2002 periods, while decreases in the outer part occurred in the 1989e1990 and 2001e2002 periods. However, there was, a significant overall improvement (Rs ¼ 0.86; p ¼ 0.0001) in the BV, from low (2.00) to very high (4.77), in the inner part of the Nervio´n estuary. In turn, for the same period the outer part of the estuary did not show significant improvement (Rs ¼ 0.13; p ¼ 0.55) except for the period between 1989 and 2001 (Rs ¼ 0.59; p ¼ 0.04) (Fig. 2c and Table A.1 in the supplementary material). BV clearly increases with time in the inner part of the estuary, while no temporal trend is observed in the outer area (Fig. 3).

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 0 5 e2 1 7

211

Fig. 2 e Investment and responses: (a) Cumulative Economic Investment (Cum. V); (b) Annual loads of Biochemical Oxygen Demand (BOD) and Ammonia nitrogen (NH3) (t yrL1), and (c) Averaged Total Biological Values and Standard Deviations (BV) per year; being 1 [ low and 5 [ very high. Main facts that occurred in the Nervio´n (Key: Phase I & II [ External port widening phases; WWTP [ Waste Water Treatment Plant; AHV [ Altos Hornos de Vizcaya).

212

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 0 5 e2 1 7

Fig. 3 e Biological value evolution mapping: BV changes along the estuary throughout the sewage scheme period.

213

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 0 5 e2 1 7

The performance of the reliability and sampling effort values (Table A.1 in the supplementary material), in general show that the reliabilities of both inner and outer areas increased along the sewage scheme period from lower reliabilities to medium reliabilities, with some high reliabilities at specific years. The low reliabilities obtained at both areas for 2010 are explained by having obtained the latest data for only two of the four studied ecosystem components (soft substratum macrobenthos and demersal fish). Sampling effort remained constant throughout time in the inner part of the Nervio´n estuary, whilst they increased from medium to high in the outer part of the estuary.

3.3.

Cost-benefit analysis

When comparing the improvements in the total ecosystem component BV with the cumulative investment expenditures in abatement actions (V), BOD and NH3 annual loads into the estuary, there is a clear increase in the BV together with a cumulative increase in the total abatement action investments and the decrease in the BOD and NH3 into the estuary (Fig. 2). There are significant direct response correlations between BOD/Cumulative Investment/NH3 with the changes to the BV of the inner part, as well as between the inner BV and the outer BV, shown in Table 4. There is a highly significant negative correlation between the cumulative investment expenditures and the loads of BOD and NH3 into the estuary (Table 4), i.e. an increase of funding reduced the inputs; while the correlation between BOD and NH3 is also significant but positive, i.e. the higher BOD and NH3 loadings occurred in the same years. In addition, there is a significant positive correlation between the cumulative investments and the inner part BV, showing an increasing ecological response to the investments, although the same relationship is not significant for the outer part of the estuary. However, if we look at the 1989e2001 period, a significant positive correlation is seen between the outer part BV and the cumulative investments (Table 4). The correlation between the BOD and NH3 and the BV at the inner part of the estuary is significant but negative and there is no statistical significance in their correlation at the outer part of the estuary even when applying different time lags (Table 4). There is a significant positive correlation between the BV of the inner and outer areas (Rs ¼ 0.45; p ¼ 0.03), i.e. the BV of the two areas increase and decrease in tandem.

4.

Discussion

Management interventions attempt, as one of their main objectives, to restore the biodiversity of degraded ecosystems, together with the functions and processes of those systems (Elliott et al., 2007). The application of the BV Method allows us to collate all of the observed improvements in biodiversity into a single value whose evolution can be studied throughout the restoration period. This approach has previously been used elsewhere (Derous et al., 2007a; Forero, 2007; Rego, 2007; Vanden Eede, unpublished; Weslawski et al., 2009; Pascual et al., 2011) although most authors have concentrated on the spatial biodiversity and have not addressed its valuation on both a spatial and temporal scale. The Nervio´n estuary provides the opportunity for observing the response of the ecosystem to sewage abatement investments as water quality improves, providing a valuable record of the status of the different ecosystem components throughout time. Fig. 4 summarizes the overall findings of our study and allows us to conclude that the total amount of investment expenditures into the sewage scheme of the Nervio´n estuary has contributed to the improvement of: firstly, the abiotic factors (NH3 and BOD) and secondly, the biotic factors at the inner part of the estuary. Knowing the total amount of money invested during the 21 year period of the sewage scheme allows us to obtain the cost of the pollution abatement actions, both for public and private investments. The observed direct negative correlation between BOD and NH3 loads and the investment expenditure (Fig. 4) was also identified by Garcı´a-Barcina et al. (2006), who reported that water quality showed statistically significant increases in dissolved oxygen saturation and decreases in ammonia nitrogen, which could primarily be attributed to the pollution abatement measures undertaken by the local water authority. The overall BV change (Fig. 2c and 3 and Table A.1 in the supplementary material) show a clear biological improvement in the Nervio´n estuary (see also Borja et al., 2010). This improvement is more apparent in the inner part of the estuary especially as this area suffered the worst environmental conditions, such as a notable oxygen depletion, loss of fauna and flora species, and aesthetic problems (Sa´iz Salinas and Gonza´lez-Oreja, 2000). The improved oxygen level in the

Table 4 e Correlation analysis results between the BV of both inner and outer parts of the estuary with the total annual loads of Biochemical oxygen demand (BOD); with the total cumulative economic investments and with the total annual loads of ammonia nitrogen (NH3). Significant correlations are highlighted in bold. *p < 0.05; **p < 0.01 or ***p < 0.001; this is because “a´” refers to the Type I error, quantified through the probability “p” (Key: ƞ [ number of samples; r [ correlation coefficient; p-value [ probability value). BOD

BOD Cum. Invest. BV_inner BV_outer BV_outer(1989e2001)

Cum. investment

ƞ

r

p-value

21 21 21 12

0.9156 0.6870 0.0650 0.4553

p < 0.0001 0.0021** 0.7714 0.1310

ƞ

21 21 13

r

0.8429 0.0013 0.5887

NH3

p-value

ƞ

r

p-value

0.0002*** 0.9954 0.0414*

21 21 21 21 12

0.8792 0.8987 0.6260 0.1000 0.2662

0.0001*** 0.0001*** 0.0051** 0.6546 0.3773

214

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 0 5 e2 1 7

Fig. 4 e Investment and responses significant correlation results summary: Cumulative Economic Investment (Cum. V); Annual loads of Biochemical Oxygen Demand (BOD); Ammonia nitrogen (NH3) and Averaged Total Biological Values (BV) in inner and outer estuary (Key: 4 positive significant correlation; . negative significant correlation).

water also supported the increased penetration upstream of species which are sensitive to pollutants, allowing for the improvement of the BV at this inner part (Borja et al., 2006b; Leorri et al., 2008). A strong response of the BV to specific management actions can be identified (Fig. 2c). As such, in 1990, as a consequence of the commissioning of Galindo WWTP, the data show an immediate increase in the BV in the outer part of the estuary, as well as a 2 year time-lag response in the inner part of the estuary. The further closure in 1996 of one of the main iron and steel industries, Altos Hornos de Vizcaya (AHV), allowed for a further BV improvement observed both in the inner and outer parts (Borja et al., 2006b, 2008b, 2010). The implementation of the biological treatment at the Galindo WWTP (in 2001), which provides organic matter and nitrogen removal, greatly reduced the contribution of the plant effluent to the overall load (Garcı´aBarcina et al., 2006). As a result, a primary decrease was observed in the BV of both areas subsequent to the treatment start-up and a posterior increase and stabilization of the BV in the inner and outer parts, respectively. Furthermore, the decrease in the BV in the outer part of the Nervio´n estuary could be related to Phases I & II of major port and dock building works at the harbor of Bilbao and with major dredging in the area (2001), which could be responsible for this decrease of almost half a unit in the BV, as detected in the benthic component (Borja et al., 2009b). It must be highlighted that the inner and outer parts of the estuary react differently to the investment expenditures and abiotic improvements. While the correlation of the inner part BV with cumulative investment and BOD and NH3 loads are significant and direct, the BV in the outer part only seems to correlate directly with the inner BV as well as with cumulative investments only for the 1989e2001 time frame, unlike for the rest of the sewage scheme. This allows us to further corroborate the fact that pressures, and therefore responses, differ between the inner and outer parts of the Nervio´n estuary and

that, as stated in Borja et al. (2006a), these should be regarded as two different water bodies with different management approaches and different times of recovery. When comparing the recovery times obtained by Andrews (1984), for the Thames estuary, with the ones for the Nervio´n estuary, similar time spans of 10e11 years are obtained, which coincides with the recovery time boundaries stated in Borja et al. (2010), for most of the aquatic ecosystem components. Knowing the recovery times and applying the conceptual model of the trajectories of change of Elliott et al. (2007) and Borja et al. (2010) to the state of the Nervio´n estuary, allows us to determine that the Nervio´n estuary system falls into a state of good recovery (phase four) at which biological communities start influencing the physico-chemical system through bioturbation, biosedimentation, vegetative growth, etc. This would help to formulate robust, cost-efficient and feasible water treatment decisions as is required from regulators and policy makers. Furthermore, the 2009 and 2010 BVs show that there is still potential for further improvements in the BV of the estuary (especially at the outer part). These could be achieved following the forthcoming sewage scheme activities announced by the CABB which include the establishment of another WWTP, a possible submarine outfall (to avoid the development of harmful algae blooms within the inner estuary (Ferna´ndez Pe´rez, 2005), and the renewal of the pipelines. The better control and limiting of dredging activities, and diffuse and riverine pollution (i.e. through the building of storm tanks (Ferna´ndez Pe´rez, 2005), would further reduce the loads of uncontrolled inputs into the estuary, allowing for its continued recovery. Despite of the already cited difficulties faced when accessing the economic data, categorising some of the capital expenditures as one-off costs could lead to an overestimate of the investment effort being made at the beginning of all installation commissioning, whilst an incorrect division of capital expenditure over the average life time of installations could underestimate the investment being applied to the sewage scheme. As stated above, estuaries offer a wide range of economically quantifiable goods and services to society (America, 2008; Atkins et al., 2011). Together with the estuary’s improvement, new uses and services have developed in the Nervio´n estuary: European bathing water quality standards were met at local beaches (Garcı´a-Barcina et al., 2006) and new recreational activities are being established on and around the estuary (rowing competitions, recreational fishing competitions, canoeing, boat-cruises, etc.). As this study was unable to analyse the further link between the environmental improvement of the estuary and the increase in its service provision, as there is not a unique response path from which they might have evolved, the authors highlight this as a possible area for further research in the Nervio´n estuary.

5.

Conclusions

Our approach successfully related the ecological and economic data allowing us to produce a cost/benefit analysis of the Nervio´n estuary sewage scheme plan.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 0 5 e2 1 7

Pollution abatement actions carried out in the Nervio´n estuary, which reduced the BOD and total NH3 waste input values, resulted in a significant increase in the BV for the inner part of the estuary. Furthermore, variations in BV, along the sewage scheme, respond to the different human impacts and actions that have occurred along the Nervio´n estuary and the findings are complicated by interactions between different management measures at different places. Progress towards the restoration of the Nervio´n estuary shows that it is possible to recover even an extremely polluted aquatic ecosystem. The Nervio´n estuary acts as an example for decision makers on how, and by how much, the recovery of a highly polluted estuary is possible.

Acknowledgements Data for this study were obtained from different projects funded by the Consorcio de Aguas Bilbao-Bizkaia. Data of annual loads to the estuary were provided by Jose´ Marı´a Garcı´a Barcina (CABB). This paper is a result of the project WISER (Water bodies in Europe: Integrative Systems to assess Ecological status and Recovery) funded by the European Union under the 7th Framework Programme, Theme 6 (Environment including Climate Change) (contract No. 226273), www.wiser.eu. M. Pascual was supported by a grant from the Technological Centres Foundation of the Basque Country. We wish to thank the Institute of Estuarine and Coastal Studies and the University of Hull, UK, for receiving M. Pascual during an internship. Ainhize Uriarte (AZTI-Tecnalia) kindly provided us with fish data. The Department of Applied Economy, at the University of the Basque Country, is also thanked for kindly advising us on some details of this contribution. We are grateful to three anonymous reviewers for their constructive comments. This is paper number 556 from the Marine Research Division (AZTI-Tecnalia).

Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.10.053.

references

America, R., 2008. The Economic and Market Value of Coast and Estuaries: What’s at Stake? Population English Edition. Restore America’s Estuaries. Andrews, M.J., 1984. Thames estuary: pollution and recovery. In: Sheehan, P.J., Miller, D.R., Butler, G.C., Bourdeau, Ph (Eds.), Effects of Pollutants at the Ecosystem Level. SCOPE, Wiley, New York, pp. 195e227. Apitz, S.E., Elliott, M., Fountain, M., Galloway, T.S., 2006. European environmental management: moving to an ecosystem approach. Integrated Environmental Assessment and Management 2 (1), 80e85.

215

Aravena, G., Villate, F., Uriarte, I., Iriarte, A., Ibanez, B., 2009. Response of Acartia populations to environmental variability and effects of invasive congenerics in the estuary of Bilbao, Bay of Biscay. Estuarine, Coastal and Shelf Science 83 (4), 621e628. Aronson, J., Blignaut, J.N., Milton, S.J., Le Maitre, D., Esler, K.J., Limouzin, A., Fontaine, C., de Wit, P.M., Mugido, W., Prinsloo, P., Van der Elst, L., Lederer, N., 2010. Are socioeconomic benefits of restoration adequately quantified? A meta-analysis of recent papers (2000e2008) in restoration ecology and 12 other scientific journals. Restoration Ecology 18, 143e154. Aslan-Yilmaz, A., Okus, E., Ovez, S., 2004. Bacteriological indicators of antropogenic impact prior to and during the recovery of water quality in an extremely polluted estuary, Golden Horn, Turkey. Marine Pollution Bulletin 49, 951e958. Atkins, J.P., Burdon, D., Elliott, M., Gregory, A.J., 2011. Management of the marine environment: integrating ecosystem services and societal benefits with the DPSIR framework in a systems approach. Marine Pollution Bulletin 62 (2), 215e226. Attrill, M.J., 1998. A Rehabilitated Estuarine Ecosystem: The Environment and Ecology of the Thames Estuary. Kluwer, Dordrecht. Balvanera, P., Pfisterer, A.B., Buchmann, N., He, J.S., Nakashizuka, T., Raffaelli, D., Schmid, B., 2006. Quantifying the evidence for biodiversity effects on ecosystem functioning and services. Ecology Letters 9, 1146e1156. Barreiro, P., Aguirre, J., 2005. 25 years of the integral sewage plan of the Rı´a de Bilbao (25 an˜os del plan integral de saneamiento de la Rı´a de Bilbao). DYNA, 25e30 (JanuaryeFebruary). Beaumont, N.J., Austen, M.C., Atkins, J.P., Burdon, D., Degraer, S., Dentinho, T.P., Derous, S., Holm, P., Horton, T., van Ierland, E., 2007. Identification, definition and quantification of goods and services provided by marine biodiversity: implications for the ecosystem approach. Marine Pollution Bulletin 54 (3), 253e265. Borja, A., Collins, M., 2004. Oceanography and marine environment of the basque country. Elsevier Oceanography Series 70, 616. Borja, A., Franco, J., Valencia, V., Bald, J., Muxika, I., Belzunce, M.J., Solaun, O., 2004. Implementation of the European water framework directive from the Basque country (northern Spain): a methodological approach. Marine Pollution Bulletin 48 (3e4), 209e218. Borja, A., Galparsoro, I., Solaun, O., Muxika, I., Tello, E.M., Uriarte, A., Valencia, V., 2006a. The European water framework directive and the DPSIR, a methodological approach to assess the risk of failing to achieve good ecological status. Estuarine, Coastal and Shelf Science 66 (1e2), 84e96. Borja, A., Muxika, I., Franco, J., 2006b. Long-term recovery of softbottom benthos following urban and industrial sewage treatment in the Nervio´n estuary (southern Bay of Biscay). Marine Ecology Progress Series 313, 43e55. Borja, A., Bricker, S.B., Dauer, D., Demetriades, N.T., Ferreira, J.G., Forbes, A.T., Hutchings, P., Jia, X., Kenchington, R., Marques, J. C., Zhu, C., 2008a. Overview of integrative tools and methods in assessing ecological integrity in estuarine and coastal systems worldwide. Marine Pollution Bulletin 56, 1519e1537. Borja, A., Tueros, I., Belzunce, M.J., Galparsoro, I., Garmendia, J.M., Revilla, M., Solaun, O., Valencia, V., 2008b. Investigative monitoring within the European Water Framework Directive: a coastal blast furnace slag disposal, as an example. Journal of Environmental Monitoring 10, 453e462. Borja, A., Bald, J., Franco, J., Larreta, J., Muxika, I., Revilla, M., Rodrı´guez, J.G., Solaun, O., Uriarte, A., Valencia, V., 2009a. Using multiple ecosystem components, in assessing ecological status in Spanish (Basque Country) Atlantic marine waters. Marine Pollution Bulletin 59 (1e3), 54e64. Borja, A., Muxika, I., Rodrı´guez, J.G., 2009b. Paradigmatic responses of marine benthic communities to different

216

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 0 5 e2 1 7

anthropogenic pressures, using M-AMBI, within the European water framework directive. Marine Ecology 30, 214e227. Borja, A., Dauer, D., Elliott, M., Simenstad, C.A., 2010. Medium and long term recovery of estuarine and coastal ecosystems: patterns, rates and restoration effectiveness. Estuaries and Coasts 33, 1249e1260. Breeze, H., 2004. Review of Criteria for Selecting Ecologically Significant Areas of the Scotian Shelf and Slope: A Discussion Paper. Ocean and Coastal Management Report 2004e04, prepared for Oceans and Coastal Management Division. Oceans and Habitat Branch, Maritimes Region, Fisheries, and Oceans Canada, Bedford Institute of Oceanography, p. 96. Brosnan, T.M., O’Shea, M.L., 1996. Sewage abatement and coliform bacteria trends in the lower Hudson-Raritan estuary since passage of the clean water act. Water Environmental Research 68 (1), 25e35. Cearreta, A., Irabien, M.J., Pascual, A., 2004. Human activities along the Basque coast during the last two centuries: geological perspective of recent anthropogenic impact on the coast and its environmental consequences. In: Borja, A., Collins, M. (Eds.), Oceanography and Marine Environment of the Basque Country. Elsevier, Amsterdam, pp. 27e50. Conley, D.J., Josefson, A.B., 2001. Hypoxia, nutrient management and restoration in Danish waters. In: Rabalais, N.N., Turner, R. E. (Eds.), Coastal Hypoxia: Consequences for Living Resources and Ecosystems. Coastal and Estuarine Studies, vol. 58. American Geophysical Union, Washington, D.C, pp. 425e434. Derous, S. 2007. Marine biological valuation as a decision support tool for marine management. Ph.D. thesis, University of Ghent, Belgium., 298. Derous, S., Agardy, T., Hillewaert, H., Hostens, K., Jamieson, G., Lieberknecht, L., Mees, J., Moulaert, I., Olenin, S., Paelinckx, D., Rabaut, M., Rachor, E., Roff, J., Stienen, E.W.M., van der Wal, J. T., Van Lancker, V., Verfaillie, E., Vincx, M., Weslawski, J.M., Degraer, S., 2007a. A concept for biological valuation in the marine environment. Oceanologia 49 (1), 99e128. Derous, S., Austen, M., Claus, S., Daan, N., Dauvin, J.-C., Deneudt, K. , Depestele, J., Desroy, N., Heessen, H., Hostens, K., Marboe, A.H., Lescrauwaet, A.-K., Moreno, M., Moulaert, I., Paelinckx, D., Rabaut, M., Rees, H., Ressureic¸a˜o, A., Roff, J., Santos, P.T., Speybroeck, J., Stienen, E.W.M., Tatarek, A., Ter Hofstede, R., Vincx, M., Zarzycki, T., Degraer, S., 2007b. Building on the concept for marine biological valuation with respect to translating it to a practical protocol: viewpoints derived from a joint ENCORA-MARBEF initiative. Oceanologia 49 (4), 579e586. Dı´ez, I., Santolaria, A., Secilla, A., Gorostiaga, J.M., 2009. Recovery stages over long-term monitoring of the intertidal vegetation in the ’Abra de Bilbao’ area and on the adjacent coast (N. Spain). European Journal of Phycology 44 (1), 1e14. Elliott, M., Burdon, D., Hemingway, K.L., Apitz, S.E., 2007. Estuarine, coastal and marine ecosystem restoration: confusing management and science - A revision of concepts. Estuarine, Coastal and Shelf Science 74 (3), 349e366. Ferna´ndez Pe´rez, D., 2005. Sewage abatement infrastructures in a metropolitana area: the case of urban Bilbao (Las infraestructuras de saneamiento en un a´rea metropolitana: el caso de la aglomeracio´n urbana de Bilbao). Ingenierı´a y territorio 71, 56e65. Forero, C.E. 2007 Biological Valuation of the Marine Environment: The Netherlands. ECOMAMA (Ecological Marine Management Programme). Master thesis Dissertation. Vrije Universiteit Brussel. Belgium. Franco, J., Borja, A., Castro, R., Larreta, J., Muxika, I., Revilla, M., Uriarte, A., Rodrı´guez, G., Villate, F., Orive, E., Seoane, S., Laza, A., 2010. Environmental Monitoring of the estuaries of Nervio´n, Barbadu´n and Butro´n during 2009 (Seguimiento ambiental de los estuarios del Nervio´n, Barbadu´n y Butro´n

durante 2009). Consorcio de Aguas Bilbao Bizkaia. 311 þ annexes (Unpublished report). Garcı´a-Barcina, J.M., Gonzalez-Oreja, J.A., De la Sota, A., 2006. Assessing the improvement of the Bilbao estuary water quality in response to pollution abatement measures. Water Research 40 (5), 951e960. Gering, J.C., Crist, T.O., Veech, J.A., 2003. Additive partitioning of species diversity across multiple spatial scales: implications for regional conservation of biodiversity. Conservation Biology 17 (2), 488e499. Gonza´lez-Oreja, J.A., Sa´iz Salinas, J.I., 1999. Fitting predictive models of the biological recovery in the Bilbao Estuary. Acta Oecologica 20 (4), 471e477. Gorostiaga, J.M., Borja, A., Diez, I., France´s, G., Pagola-Carte, S., Sa´iz Salinas, J.I., 2004. Recovery of benthic communities, in polluted systems. In: Borja, A., Collins, M. (Eds.), Oceanography and Marine Environment of the Basque Country. Elsevier, Amsterdam, pp. 549e578. Halpern, B.S., Walbridge, S., Selkoe, K.A., Kappel, C.V., Micheli, F., D’Agrosa, C., Bruno, J.F., Casey, K.S., Ebert, C., Fox, H.E., Frujita, R., Heinemann, D., Lenihan, H.S., Madin, E.M.P., Perry, M.T., Selig, E. R., Spalding, M., Steneck, R., Watson, R., 2008. A global map of human impact on marine ecosystems. Science 319, 948e952. Hawkins, S.J., Gibbs, P.E., Pope, N.D., Burt, G.R., Chesman, B.S., Bray, S., Proud, S.V., Spence, S.K., Southward, A.J., Langston, W.J., 2002. Recovery of polluted ecosystems: the case for long term studies. Marine Environmental Research 54, 215e222. Leorri, E., Cearreta, A., Irabien, M.J., Yusta, I., 2008. Geochemical and microfaunal proxies to assess environmental quality conditions during the recovery process of a heavily polluted estuary: the Bilbao estuary case (N. Spain). Science of the Total Environment 396 (1), 12e27. Lotze, H.K., 2010. Historical reconstruction of human-induced changes in U.S. estuaries. Oceanography and Marine Biology: an Annual Review 48, 267e338. Mialet, B., Azemar, F., Maris, T., Sossou, C., Ruiz, P., Lionard, M., Van Damme, S., Lecerf, A., Muylaert, K., Toumi, N., Meire, P., Tackx, M., 2010. Spatial spring distribution of the copepod Eurytemora affinis (Copepoda, Calanoida) in a restoring estuary, the Scheldt (Belgium). Estuarine, Coastal and Shelf Science 88 (1), 116e124. Nijkamp, P., Vindigni, G., Nunes, P.A.L.D., 2008. Economic valuation of biodiversity: a comparative study. Ecological Economics 67 (2), 217e231. Pagola-Carte, S., Sa´iz Salinas, J.I., 2001. Cambios en el macrozoobentos de sustrato rocoso del abra de Bilbao: 14 an˜os de seguimiento de la recuperacio´n biolo´gica. Boletı´n del Instituto Espan˜ol de Oceanografı´a 17, 163e177. Pascual, M., Borja, A., Vanden Eede, S., Deneudt, K., Vincx, M., Galparsoro, I., Legorburu, I., 2011. Marine Biological Valuation Mapping of the Basque continental shelf (Bay of Biscay), within the context of Marine Spatial Planning. Estuarine, Coastal and Shelf Science 95 (1), 186e198. Rego, T. 2007 A Biological Valuation of the Pico - Faial Channel. Master thesis Dissertation, University of Azores. Sa´iz Salinas, J.I., Gonza´lez-Oreja, J.A., 2000. Stress in estuarine communities: lessons from the highly-impacted Bilbao estuary (Spain). Journal of Aquatic Ecosystem Stress and Recovery 7, 43e45. Simenstad, C., Reed, D., Ford, M., 2006. When is restoration not? Incorporating landscape-scale processes to restore selfsustaining ecosystems in coastal wetland restoration. Ecological Engineering 26, 27e39. Soler, J.M., Zorrakin, B., Franco, J., 2008. La Bola and La Vega de Lamiako Beaches: two interest places for the birds on the Rı´a of Bilbao (La playa de La Bola y la Vega de Lamiako: Dos enclaves de intere´s para las aves en la Rı´a de Bilbao). Artadi 3, 56e59.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 0 5 e2 1 7

U.S. Census Bureau, 2008. Pollution Abatement Cost and Expenditures: 2005, MA200(05). U.S. Government Printing Office, Washington, DC. Uriarte, A., Borja, A., 2009. Assessing fish quality status in transitional waters, within the European Water Framework Directive: setting boundary classes and responding to anthropogenic pressures. Estuarine, Coastal and Shelf Science 82 (2), 214e224. Uriarte, I., Villate, F., 2004. Effects of pollution on zooplankton abundance and distribution in two estuaries of the Basque coast (Bay of Biscay). Marine Pollution Bulletin 49 (3), 220e228. Vanden Eede, S. unpublished. Marine Biological Valuation of the Isles of Scilly Archipelago. Master in Marine and Lacustrine Sciences Dissertation. University of Ghent, Belgium, p. 61. Villate, F., Uriarte, I., Irigoien, X., Beaugrand, G., Cotano, U., 2004. Zooplankton communities. In: Borja, A., Collins, M.

217

(Eds.), Oceanography and Marine Environment of the Basque Country, vol. 70. Elsevier Oceanography Series, pp. 395e423. Wells, E., Wilkinson, M., Wood, P., Scanlan, C., 2007. The use of macroalgal species richness and composition on intertidal rocky seashores in the assessment of ecological quality under the European Water Framework Directive. Marine Pollution Bulletin 55 (1e6), 151e161. Weslawski, J.M., Warzocha, J., Wiktor, J., Urban˜ski, J., Bradtke, K., Kryla, L., Tatarek, A., Kotwicki, L., Piwowarcyk, J. , 2009. Biological valorisation of the southern Baltic Sea (Polish Exclusive economic Zone). Oceanologia 51 (3), 415e435. Whitfield, A.K., Elliott, M., 2002. Fishes as indicators of environmental and ecological changes within estuaries: a review of progress and some suggestions for the future. Journal of Fish Biology 61 (Suppl. A), 229e250.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 1 8 e2 2 6

Available online at www.sciencedirect.com

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

Reverse osmosis concentrate treatment via a PAC-MF accumulative countercurrent adsorption process Chunxia Zhao, Ping Gu, Hangyu Cui, Guanghui Zhang* School of Environmental Science and Engineering, Tianjin University, Tianjin 300072, China

article info

abstract

Article history:

Organic pollutants in reverse osmosis (RO) concentrates from wastewater reclamation are

Received 6 May 2011

mainly comprised of low molecular weight biorefractory compounds. Generally, advanced

Received in revised form

oxidation methods for oxidizing these organics require a relatively high level of energy

14 October 2011

consumption. In addition, conventional adsorption removal methods require a large dose

Accepted 24 October 2011

of activated carbon. However, the dose can be reduced if its full adsorption capacity can be

Available online 31 October 2011

used. Therefore, the combined technology of powdered activated carbon (PAC) adsorption and microfiltration (MF) membrane filtration was studied to develop a countercurrent two-

Keywords:

stage adsorption process. A PAC accumulative adsorption prediction method was proposed

Wastewater reclamation

based on the verification of a PAC multi-stage adsorption capacity equation. Moreover, the

Reverse osmosis concentrate

prediction method was amended for a more accurate prediction of the effluent quality

Powdered activated carbon

because adsorption isotherm constants were affected by the initial adsorbate concentra-

Microfiltration

tion. The required PAC dose for the accumulative countercurrent two-stage adsorption

Countercurrent adsorption

system was 0.6 g/L, whereas that of the conventional adsorption process was 1.05 g/L when the dilution factor(F ) was 0.1 and the COD and DOC removal rates were set to 70% and 68.1%, respectively. Organic pollutants were satisfactorily removed with less consumption of PAC. Effluent from this combined technology can be further reclaimed by an RO process to improve the overall recovery rate to between 91.0% and 93.8% with both economic and environmental benefits. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Various solutes in water can be almost entirely retained by reverse osmosis (RO) processes. RO is a simple and efficient process and has played an important role in wastewater reclamation (Shannon et al., 2008; Tam et al., 2007). Influent for RO processes usually is derived from the effluent of municipal/industrial wastewater treatment. The organics in the effluent generally comprise biopolymers and humic-like substances, which are resistant to further biological treatment after extensive biodegradation and microfiltration/ ultrafiltration processes (Jarusutthirak et al., 2002; Laabs et al.,

2006; Zheng et al., 2010; Li et al., 2006). Therefore, the main constituents of RO concentrate are dissolved inorganics salts and low molecular weight soluble refractory organics (Ozaki and Li, 2002), which are closely related to the influent quality (e.g., the presence of petrochemicals, pharmaceutical products, pesticides, endocrine disruptors, anti-scaling chemicals, disinfection byproducts, personal care products, soluble microbial products, bacteria, pathogens, or cell debris) (Benner et al., 2008; Comerton et al., 2005). Typically, the recovery rate of RO processes in wastewater reclamation is about 70%e75%. Therefore, the pollutant concentration in RO concentrate can be 2e3 times higher than that in RO influent.

* Corresponding author. Tel./fax: þ86 21 2740 5059. E-mail address: [email protected] (G. Zhang). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.050

219

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 1 8 e2 2 6

RO concentrate is not suitable for discharge into the environment without a proper treatment or further concentration by RO processes due to a high potential for RO membrane fouling. One solution to these problems is to partially remove organic pollutants from RO concentrate and then reclaim the treated concentrate through further RO processes. In this way, the overall recovery rate could be increased, and the final cost of concentrate disposal could be decreased. Currently, methods for removing organic pollutants from RO concentrate mainly include advanced oxidation processes (AOPs) and adsorption process. AOPs include photocatalytic oxidation, ozone oxidation (Benner et al., 2008; Paraskeva and Graham, 2005) and electrochemical oxidation (Badruzzaman et al., 2009). In a study by Westerhoff et al., 2009. RO concentrate was treated by UV-TiO2, and the energy consumption was 9.6 kW h/m3 when DOC decreased from 40 mg/L to 10 mg/L (Westerhoff et al., 2009). N-Nitrosodimethylamine in RO concentrate was degraded using boron-doped diamond (BDD) film electrodes, and 6.9 kW h/m3 was required to achieve 75% DOC conversion (Chaplin et al., 2010). RO concentrate generated from tertiary treatment has been oxidized using BDD electrodes; total COD removal was achieved after 8 h of treatment, and the energy consumption was calculated to be 59 kW h/kg COD or 6.4 kW h/m3 (Perez et al., 2010). In adsorption technologies, granular activated carbon (GAC) and powdered activated carbon (PAC) are the most commonly used adsorbents for dissolved organic matter removal, and the corresponding adsorption process is mature, simple, and cost-effective for wastewater reclamation (Gur-Reznik et al., 2008) compared to AOPs. PAC-membrane adsorption processes have several advantages compared with GAC columns, although all the adsorption capacity of a traditional GAC column could be exhausted when the adsorbate concentration of the effluent equals that of raw water. A research showed that a smaller PAC particles size and larger pore size made PAC more accessible to acrylonitrile by diffusion from the solution than that for GAC. Acrylonitrile removal by PAC and GAC was 93% and 84%, respectively, after 1 h of adsorption, because diffusion phenomena had a greater limiting effect on the adsorption rate for GAC (Kumar et al., 2008). In another investigation, sorption kinetics results showed that the adsorbent size could greatly influence the sorption velocity; GAC required over 168 h to achieve equilibrium, which was much longer than the 4 h required for perfluorooctane sulfonate and perfluorooctanoate removal using PAC (Yu et al., 2009). A pilot-scale hollow-fiber ultrafiltration (UF) unit followed by a PAC or GAC adsorption unit was studied for DOC removal. The removal rate of the PAC-UF combination was 60% compared with 36% for GAC-UF in the municipal wastewater treatment plant of Crete, Greece (Dialynas and Diamadopoulos, 2008). Moreover, PAC could control irreversible membrane fouling and minimize the frequency of chemical cleaning (Campinas and Rosa, 2010). In addition, the GAC adsorption capacity was dependent on its particle size when dissolved organic matter was present; larger GAC particle sizes were directly correlated with lower adsorption capacities per mass of adsorbent (Corwin and Summers, 2010). Multi-stage countercurrent adsorption has been applied as a separating processes in chemical engineering applications. The maximum adsorption capacity of an adsorbent can be obtained by this process (Tseng and Wu, 2009; Wu and Tseng,

2008). However, a countercurrent adsorption process cannot be achieved exactly using conventional filtration and sedimentation methods, whereas it can be achieved using microfiltration (MF) with a proper design. To use the full adsorption capacity of PAC, countercurrent two-stage adsorption was investigated in this paper. A PAC multi-stage adsorption capacity equation was verified using a lab-scale experiment. Based on this equation and adsorption isotherm constants, methods for calculating PAC dose and effluent quality were studied. A significant quantity of adsorbent was saved using accumulative countercurrent two-stage adsorption systems compared with the dose required by a single-stage adsorption system. A smaller PAC does was used to satisfactorily remove organic matter from RO concentrate to meet the RO influent quality requirement for further reclamation.

2.

Experiments

2.1.

Materials and analysis

The RO concentrate was derived from an ultrafiltration membrane bioreactor (MBR)-RO process in a refinery wastewater treatment plant. The average RO concentrate quality is shown in Table 1. Fluctuations in water quality were monitored during the research period. Industrial grade coal-based PAC (200 mesh) was purchased from Luda CO. LTD. in Zunhua, China. Prior to use, PAC was baked for 1 h at 105  C to remove moisture. The BET surface area was 867.3 m2/g, the pore size range was 0.36e4.5 nm, which were obtained from 77 K N2 isotherms using a sorptiometer (NOVA2200e, Quantachrome Instruments, USA). The minimum iodine value was 680 mg/g, the methylene blue value was 147 mg/g and the maximum ash content was 11.2% referred to Granular Activated Carbon Test (AQSIQ and SAC, 2008). COD and DOC were considered to be adsorbates in the experiments. COD was analyzed using the potassium dichromate oxidation method 5220  C (APHA et al., 2005). Commercially available analytical grade chemicals were used. DOC was determined using a TOC analyzer (TOC-VCPH, Shimadzu, Japan).

2.2.

Methods

PAC and effluent were separated by vacuum filtration using flat MF membranes in conical flask adsorption experiments.

Table 1 e Characteristics of RO concentrate. Parameters COD DOC Conductivity pH value Total hardness (CaCO3) Cl SO2 4 NO 3  F

Unit

Average value

mg/L mg/L mS/cm

149.0 (138.3  159.6) 41.43 (39.66  43.20) 3.58 7.95 870 901 337 44.9 5.50

mg/L mg/L mg/L mg/L mg/L

220

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 1 8 e2 2 6

The shaker speed was set at 200 rpm, and the temperature was set at 20  C. The MF membranes were cleaned according to the operating instructions of manufacturers.

2.2.1.

Determination of isotherm and kinetics parameters

RO concentrate (0.100 L) and pre-weighed fresh PAC (0.1e2.0 g/ L) were separately added into 250-mL conical flasks that were placed on a shaker for 30 min (or 20 min) in isothermal experiments. For the determination of kinetics parameters, RO concentrate (0.100 L) contacted with a constant amount of fresh PAC (0.9 g/L or 0.3 g/L) over a range of time (10 se70 min) in a series of shaking conical flasks. Samples containing PAC were filtered through 0.45 mm MF membranes (0.45 mm; mixed cellulose ester; China). The isothermal experiments were performed twice. The kinetics experiments were carried out twice at first, and then a third experiment was performed to demonstrate the variation from 20 min to 60 min at a dose of 0.9 g/L. All of the results were reported as the average values.

2.2.2.

Determination of the multi-stage adsorption capacity

Fresh PAC (0.09 g) was placed in contact with RO concentrate (0.100 L) for 30 min in a 250-mL conical flask. Loaded PAC was then separated using MF membranes (0.45 mm; mixed cellulose ester; China) and was contacted with RO concentrate (0.100 L) in another conical flask for further adsorption. The average values for triplicate experiments were reported.

2.2.3.

produce a mixed influent concentration of C00 . This mixture contacted the remained loaded PAC for 20 min to obtain an equilibrium solution with an adsorbate concentration of Ce1, which was the same concentration as in the initial solution of the first adsorption process. These two procedures were repeated ten times to verify the calculation method. The value of V was 0.200 L in this experiment. The average values for two parallel experiments were reported.

2.2.5.

Amendment prediction experiment

A response surface methodology (RSM) was applied to examine the effect of the initial adsorbate concentration on isotherm constants. The experiment was designed using Design-Expert 7.1 software (Korbahti and Tanyolac, 2009; Sahu et al., 2009). Central composite design, the most commonly used design method, was adopted. A volume of 0.100 L RO concentrate was contacted with fresh PAC at a series of doses (0.1e0.58 g/L) for 20 min in shaking conical flasks to obtain samples with different adsorbate concentrations. Then, different doses of fresh PAC (0.15e0.87 g/L) were added into a series of conical flasks and contacted with samples of the same initial adsorbate concentration for 20 min with shaking. The samples were obtained as described above in this section. Experimental results from the final effluent were incorporated into the design table for model establishment.

Desorption of loaded PAC

Specified amounts of loaded PAC and ultrapure water (0.100 L; 18.2 MU; 25  C; Millipore) were added into 250-mL conical flasks for 60 min with shaking to examine the organic matter that desorbed from loaded PAC. The average values for two parallel experiments were reported.

2.2.4. Accumulative countercurrent two-stage adsorption experiments A combined immersion MF and PAC process was adopted to achieve accumulative countercurrent two-stage adsorption through batch operation. A portion of the equilibrium effluent remained in the reactor to ensure MF membrane immersion. Therefore, the dilution factor (F ) was defined as the ratio of the retention and total solution volumes in the reactor. A magnetic stirring pressure cup with MF membrane disc was used as the adsorption device. During the initiating stage, fresh PAC was contacted with RO concentrate at a dose of m0/ V in the pressure cup reactor for 20 min with stirring at room temperature (19e21  C) to achieve the adsorbate concentration Ce1. Afterwards, the repeated operational stage experiments were carried out. There were two adsorption processes in the operational stage as shown in Fig. 1. The first process, designated as the effluent cycle, involved the contact of the solution, with adsorbate concentration of Ce1, with fresh PAC at a dose of m/V for 20 min to obtain effluent with an adsorbate concentration of Ce2. At the moment Ce2 was achieved, nitrogen was injected into the reactor to force the equilibrium effluent, with a volume of (1F )V, outward via the MF membrane (0.45 mm; regenerated cellulose; Millipore; USA). The second process, designated as the influent cycle, the remaining effluent was mixed with injected RO concentrate at a volume of (1F )V to

3.

Results and discussion

3.1.

Kinetics and isotherm parameters

The results of adsorption kinetics experiments are shown in Fig. 2. The PAC was largely saturated after 60 min of adsorption. However, an adsorption time of 30 min was selected to reduce the reactor volume. When the PAC dose was 0.9 g/L, 65.3% and 71.1% COD removal was observed with adsorption times of 30 min and 60 min, respectively. Fitting results indicated that the adsorption process occurred according to a pseudo second-order kinetics equation, shown in Eq. (1), which was similar to another study (Wu et al., 2009):
2 dqt ¼ k2 qs  qt dt

(1)

where qt and qs are the adsorption capacity of PAC at time t and at saturation (mg/g), respectively; and k2 is the pseudo second-order kinetics constant (g/(mg min)). Fitted isotherms and pollutant removal results after 30 min of adsorption are presented in Fig. 3a. The adsorption isotherm equation suitable for expressing COD or DOC removal in this study was the Freundlich isotherm equation shown in Eq. (2). The same trend was observed by Kumar et al. (2010). qe ¼

C0  Ce ¼ KF Ce1=n m=V

(2)

In Eq. (2), qe is the equilibrium adsorption capacity of PAC (mg/g); C0 and Ce are the adsorbate concentrations of RO concentrate and equilibrium effluent (mg/L), respectively; KF

221

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 1 8 e2 2 6

Fig. 1 e Schematic diagram of the countercurrent two-stage adsorption system. During effluent cycle, the equilibrium solution contacted with fresh PAC to obtain effluent. During influent cycle, the remaining effluent was mixed with injected RO concentrate at a volume of (1 L F )V to produce a mixed influent. Then the mixture contacted the remained loaded PAC to obtain an equilibrium solution.

(mg(11/n)$L1/n/g) and 1/n are the Freundlich constants; and m/ V is the PAC dose (g/L). Although the maximum adsorption capacity of PAC would when the liquid-phase adsorbate equilibrium be KFC1/n 0 concentration approached C0, the equilibrium adsorption in a single-stage adsorption capacity of PAC was only KFC1/n e process. When PAC dose was 0.9 g/L, the values of C0, Ce and 1/ n for COD removal were 159.6 mg/L, 55.2 mg/L and 0.7997, respectively (Fig. 3a). The ratio of the maximum and equilibrium adsorption capacity of PAC was (C0/Ce)1/n, which reached up to 2.34. This result indicated the PAC was underutilized.

3.2. Method for calculating the multi-stage adsorption capacity The PAC adsorption capacity was used most effectively when PAC was continuously contacted with RO concentrate with an adsorbate concentration higher than that of the equilibrium effluent until the PAC was essentially saturated, and the overall adsorption capacity was designated as the multi-stage adsorption capacity. An adsorption experiment was carried out to verify that the PAC multi-stage adsorption capacity can be calculated using an equation derived from cumulative principles. Similar research was described by Zhang et al.

(2009) for cesium removal using potassium zinc hexacyanoferrate. The average results of three parallel experiments are depicted in Fig. 4. Rates of organic matter removal decreased with increasing PAC loading cycles, as shown in Fig. 4. If PAC desorption did not occur, then the adsorption capacity would be cumulative in accordance with the isotherm equation (Eq. (2)). The principles mentioned above can be described as Eq. (3). 1 C0  Cej n ¼ KF Cej qej ¼ qeðj1Þ þ m=V

(3)

In Eq. (3), qej and qe( j1) are the PAC adsorption capacities for j stages and j-1 stages (mg/g), respectively; and C0 and Cej are the adsorbate concentration (mg/L) of RO concentrate and equilibrium effluent, respectively, at the jth stage. Theoretical equilibrium adsorbate concentrations calculated using Eq. (3) and isotherm constants (Fig. 3a) were compared with the experimental values shown in Fig. 4. The calculated COD and DOC values coincided well with the experimental values in the first three stages. The experimental COD removal was 66.1%, 40.0% and 23.3%, respectively. After these stages, the calculated values were higher than the experimental results. This occurred because the 30 min isotherm constants (Fig. 3a) were used, and the unsaturated PAC still could adsorb organics until achieving the adsorption equilibrium after 60 min (Fig. 2). In summary, the multi-stage adsorption capacity of PAC can be calculated appropriately using Eq. (3) when less than 4 PAC loading cycles are used.

3.3.

Fig. 2 e Effect of adsorption time on effluent COD concentration.

ðj > 1Þ

Desorption of loaded PAC

Effluent quality would be affected if organic pollutant desorption occurred together with PAC accumulation in reactor. The loaded PAC was derived from the previous experiments, and the desorption results are listed in Table 2. The desorption of COD (3.17%e8.56%) and DOC (2.29%e9.32%) can be ignored for the small amount returning to solution. This phenomenon can be explained by the fact that the adsorbed organics were mainly hydrophobic DOM (GurReznik et al., 2008), and pore-blocking organic matter can

222

90

DOC q (mg/g)

60 45

300

30 15 0 0.0

qe=4.916Ce

200

R =0.9657

Item

0.6752

75

qe=5.509Ce

60

R =0.9764

2

45

COD

30 8

2

16

24

32

DOC C (mg/L)

150

DOC

100 40

60

80 100 120 140

0.3

0.6

0.9 1.2 1.5 PAC dose (g/L)

20min COD 20min DOC

45 300

30 15

250 200 150

1.8

2.1

2 .4

0.3

0.7057

qe=6.022Ce

80 0.7235 70 q =3.761C e e 60 2 R =0.9564 50 40 30 20 8 12 16 20 24 28 32 36

2

R =0.9661

DOC C (mg/ L )

100 50 20

0 0.0

Parameters Total mass of PAC Volume of treated RO concentrate Total adsorption mass Desorption mass Desorption ratio Total adsorption mass Desorption mass Desorption ratio

Unit g L mg mg % mg mg %

Value 0.09 0.10

0.54 0.91

0.84 1.45

10.98 0.94 8.56 3.39 0.316 9.32

95.06 3.01 3.17 24.43 0.736 3.01

154.5 6.02 3.90 39.19 0.899 2.29

COD C (mg/L)

75 60

Removal rate (%)

250

50 20

C OD q (mg/g)

b

0.7997

DOC q (mg/g)

Removal rate (%)

75

Table 2 e Results of the PAC desorption experiment.

30min COD 30min DOC

C OD q (mg/g)

a

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 1 8 e2 2 6

0.6

40

60

80 100 120 140

COD C (mg/L)

0.9

1.2 1.5 PAC dose (g/L)

1.8

2.1

2.4

Fig. 3 e Effect of PAC dose on COD and DOC removal at (a) 30 min, (b) 20 min.

strongly hinder competing organic matter uptake and micropollutant release (To et al., 2008). Thus, the accumulation of PAC was a benefit to the removal of COD and DOC. In addition, the period of loaded PAC discharge in practical applications should be determined by the MF membrane resistance.

3.4. Predicting PAC-MF accumulative countercurrent two-stage adsorption Generally, less adsorbent is required with additional countercurrent adsorption stages, although higher equipment and

Fig. 4 e Effect of PAC loading cycles on COD and DOC removal (0.9 g/L).

RO concentrate COD ¼ 154.9 mg/L, DOC ¼ 40.49 mg/L.

operating costs would be necessary. When more than three stages are used, the adsorption process is not economical (Tseng and Wu, 2009). The COD removal rate was 64% when the PAC dose was 0.9 g/L and the adsorption time was 20 min, which was slightly less than the 65.3% removal rate at 30 min adsorption (Fig. 3a). Therefore, the adsorption time can be shortened to 20 min to further reduce the reactor volume for engineering considerations. Isotherm constants for the 20-min adsorption are shown in Fig. 3b. In PAC-MF countercurrent two-stage adsorption experiments, the volume of the retention solution would be smaller and the initial adsorbate concentration of the mixed influent would be higher if a lower dilution factor value is used. However, the accumulation of PAC would result in substantial concentration polarization that will increase the membrane filtration resistance in the reactor when the volume of the remaining solution is too small. Therefore, the experiments were carried out only when F ¼ 0.3 and F ¼ 0.1. The calculation process of the prediction method begins with COD removal. The COD removal rate was set as 70%, which allowed the treated effluent meet the RO process requirement. Eq. (4) was obtained using the 20-min adsorption isotherm constants (Fig. 3b) in Eq. (2). qe ¼

C0  Ce ¼ 6:022C0:7057 e m=V

(4)

During the initiation stage, fresh PAC was put into reactor at a dose of m0/V to obtain the equilibrium concentration of Ce1. The adsorption capacity of this loaded PAC is expressed as qie1 in Eq. (5), which was derived from Eq. (4) when C0 is 140.0 mg/ L. During the first effluent cycle at operational stage defined in Section 2.2.4, Ce2 should be 42.00 mg/L if the COD removal rate is 70%. The adsorption capacity of this portion of loaded PAC is defined as qe1 in Eq. (6), which was also derived from Eq. (4). qie1 ¼

140:0  Ce1 ¼ 6:022  C0:7057 e1 m0 =V

(5)

qe1 ¼

Ce1  42:00 ¼ 6:022  42:000:7057 ¼ 84:19ðmg=gÞ m=V

(6)

During the first influent cycle, when the dilution factor was 0.1, the influent adsorbate concentration approached C00 (0.9  140.0 þ 0.1  42.00 ¼ 130.2 mg/L). The mixing influent

223

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 1 8 e2 2 6

that contacted two kinds of loaded PAC remained in the reactor to obtain an equilibrium concentration of Ce1. Because qie1 was much higher than qe1 according to Eqs. (5) and (6), the surplus adsorption capacity of the loaded PAC with qe1 was higher than that with qie1. Therefore, the adsorption processes that adsorbate concentration varied from C00 to Ce1 was regarded only as the contribution of the loaded PAC with an adsorption capacity of qe1. Therefore, the total adsorption capacity of the PAC at a dose of m/V in this stage was calculated by Eq. (7), which was deduced from Eq. (3). qe2 ¼

Ce1  42:00 130:2  Ce1 ¼ 6:022  C0:7057 þ e1 m=V m=V

(7)

Subsequently, during the second effluent cycle, the surplus adsorption capacities of the remaining loaded or dual-loaded PAC mentioned above were ignored when compared to the fresh PAC. Then, during the second influent cycle, the adsorbate in mixed influent was adsorbed by this portion of loaded PAC using its surplus adsorption capacity to obtain equilibrium adsorbate concentration of Ce1. The total adsorption capacity of this portion of PAC also was calculated by Eq. (7). When solving Eqs. (6) and (7), the values of Ce1 and m/V were 92.55 mg/L and0.5999 g/L, respectively. The value of m0/V was 0.3227 g/L, which was obtained by entering the value of Ce1 into Eq. (5). When the dilution factor was 0.2 or 0.3, the calculation method was similar to that mentioned above. In the calculated process of DOC removal, the PAC dose of m/V for the countercurrent two-stage system was 0.5999 g/L as mentioned above. DOC isotherm constants, RO concentrate DOC with an adsorbate concentration of C0 and the dilution factor were regarded as known parameters. These values were substituted into Eqs. (6) and (7) to obtain the values of Ce1 and Ce2. Two sequencing batch experiments run 1 (F ¼ 0.3) and run 2 (F ¼ 0.1) were carried out to compare the degree of coincidence between the calculated and experimental values, and the calculated doses were listed in Table 3. COD and DOC concentrations of RO concentrate were 140.0 mg/L and 40.00 mg/L, respectively, and the COD removal rate was set as 70%. According to the results shown in Table 3, a PAC savings of 42.8% was obtained when a conventional single-stage system was converted into a countercurrent two-stage system with

a dilution factor of 0.1. In addition, the amount of PAC saved increased with a decrease in the dilution factor. Similar results were obtained by Tseng and Wu (2009) who showed that the consumption of activated carbon from plum kernels for a single-stage system was 1.37 g/L, and the consumption level for a countercurrent two-stage system was 0.848 g/L as measured by methylene blue adsorption. The effluent adsorbate concentrations calculated using Eqs. (6) and (7) were represented as a stable straight horizontal line, as shown in Fig. 5a and b. Along with the accumulation of PAC, the duration of loaded PAC contact with the influent increased, which allowed the surplus adsorption capacity of the PAC to be fully used. Therefore, experimental effluent values gradually decreased during the first 4 effluent cycles. Moreover, the experimental effluent adsorbate concentrations were shown as wavy horizontal lines when the overall adsorbate concentration achieved equilibrium with the PAC adsorption capacity after 4 effluent cycles. During runs 1 and 2 of 10 total effluent cycles, 1.46 L and 1.82 L of RO concentrate were treated, respectively. COD and DOC removal rates were greater than 70% when the effluent adsorbate concentration was stable. Excluding the possibility of experimental error, the deviation between the calculated and experimental values in the first effluent cycle (Fig. 5a and b) indicated that effluent adsorbate concentrations could not be calculated accurately using Eq. (6) due to the variable isotherm constants. It is generally accepted that isotherm constants are affected by adsorption initial concentrations in multi-component systems (Knappe et al., 1998; Lu and Sorial, 2009; Li et al., 2003; Campos et al., 2000; Chang et al., 2004). Three different initial concentrations (i.e., C0, C00 and Ce1) are typically evaluated in this cumulative countercurrent two-stage adsorption experiments. Therefore, the isotherm constants for RO concentrate of C0 shown in Eq. (4) were not suitable for further calculations when the initial adsorbate concentration was C00 or Ce1.

3.5.

Amendment of the prediction method

A response surface experiment was developed using a central composite design method. Effluents with different adsorbate concentrations were obtained using the first PAC dose. The second dose of PAC was used to obtain the adsorption

Table 3 e Comparison of calculated doses in different adsorption process. F

Items

0.3a 0.2 0.1b

a b c d

COD DOC COD DOC COD DOC

Removal rate %

70.0 68.1 70.0 68.1 70.0 68.1

Adsorbate concentration Influent C00 mg/L

Equilibrium Ce1 mg/L

Effluent Ce2 mg/L

110.6 31.83 120.4 34.55 130.2 37.28

84.05 24.61 88.40 25.85 92.55 27.03

42.00 12.76 42.00 12.77 42.00 12.77

Run 1. Run 2. PAC dose for a conventional single-stage adsorption system after dilution. PAC dose for a countercurrent two-stage adsorption system.

Single-stage dose m/V g/L

0.815 0.804 0.9312 0.9177 1.048 1.032

Two-stage dose Initiate m0/V g/L

Operation m/V g/L

0.4073 0.4031 0.3625 0.3577 0.3227 0.3175

0.4994 0.4994 0.5508 0.5508 0.5999 0.5999

mc/md

1.632 1.609 1.691 1.666 1.746 1.719

224

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 1 8 e2 2 6

isotherms with a specific initial adsorbate concentration. Experimental results were entered into Design-Expert software to obtain models expressed as Eqs. (8) and (9). Ce2ðCODÞ ¼ 126:47  114:63x1  130:62x2 þ 68:229x1 x2 þ 38:872x21 þ 48:989x22 (8) Ce2ðDOCÞ ¼ 36:981  40:199x1  35:650x2 þ 22:245x1 x2 þ 16:497x21 þ 11:487x22 (9) Ce2(COD) and Ce2(DOC) represent the effluent adsorbate equilibrium concentration of COD and DOC (mg/L), and x1 and x2 represent the first and the second PAC doses (g/L), respectively. The p-values for the two models were less than 0.0001, which indicated that Eqs. (8) and (9) were significant. The Fvalues implied that the lack of fit for these equations was not significant, and their signal to noise ratios were 63.8 and 48.7, which indicates that their signals were adequate according to analyses of variance. Therefore, Eqs. (8) and (9) could be used to satisfactorily predict the relationship between COD or DOC effluent values and the values of x1 and x2 in design space.

The specified initial adsorbate concentration Ce1 can be calculated using the 20-min isotherm constants from Fig. 3b and the value x1. A series of effluent adsorbate concentrations were calculated by entering a specific x1 value and different x2 values into Eq. (8) or Eq. (9). Thus, these adsorption isotherms and corresponding constants with different initial adsorbate concentrations were calculated by entering the values of x2, Ce2(COD) or Ce2(DOC) into Eq. (2) as shown in Fig. 6a and b. Fig. 6a shows that the COD adsorption capacity of the PAC was reduced with decreasing COD initial concentrations at a certain Ce1 value. However, the prediction method mentioned in Section 3.4 ignored this change, and therefore these PAC doses were inadequate because the actual COD initial concentration was C00 or Ce1 but not C0. Therefore, experimental effluent COD values during the first 4 effluent cycles were higher than the values calculated in Fig. 5a and b. However, the effect of different initial adsorbate concentrations on DOC adsorption isotherms can be neglected due to the same isotherm trends shown in Fig. 6b. This effect can be ignored because different values of COD correspond with different types of organics due to their reducibility, whereas DOC values do not have this relationship with the type of organics. For example, two types of organics with the same DOC values can in fact have different COD values. The values of m/V were calculated based on two sets of isotherm constants when the dilution factor was 0.1. The

a

180

Cal. ( 123.7 mg/L) Cal. ( 107.2 mg/L) Cal. ( 92.73 mg/L) Cal. ( 86.30 mg/L) Cal. ( 72.19 mg/L) Exp. data ( RSM)

C OD qe (mg/g)

160 140 120 100 80 60 40 30

40

50

60

70

80

90

100

Ce (mg/L )

b

40

DOC qe (mg/g)

35 30

Cal. ( 35.11 mg/L) Cal. ( 30.94 mg/L) Cal. ( 27.48 mg/L)

25 20

Cal. ( 24.97 mg/L) Cal. ( 19.32 mg/L) Exp. data( RSM)

15 10 5

Fig. 5 e Comparison of experimental and calculated values in (a) Run 1, (b) Run 2. Run 1: F [ 0.3, m0/V [ 0.4073 g/L, m/ V [ 0.4994 g/L; run 2: F [ 0.1, m0/V [ 0.3227 g/L, m/ V [ 0.5999 g/L.

10

15

20

25

30

Ce (mg/L ) Fig. 6 e Calculated isotherms of (a) COD and (b) DOC with different initial adsorbate concentrations. RSM represents response surface methodology.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 1 8 e2 2 6

required PAC doses were 0.600 g/L and 0.680 g/L for 70% COD removal when the COD initial concentrations were 140.0 mg/L and 92.55 mg/L, respectively. Therefore, the dose of m/V was insufficient during the first 4 effluent cycles in the experiments (Fig. 5a and b). If the PAC dose was 0.680 g/L during the first 4 effluent cycles and 0.600 g/L in the later cycles, the calculated values of effluent COD better coincided with the experimental values.

4.

Conclusions

 A PAC accumulative countercurrent two-stage adsorption process can be achieved using submerged membrane technology, whereas desorption was minor and could be neglected. When the dilution factor was 0.1, the PAC dose was 0.6 g/L, and the stable experimental COD and DOC removal rates were 70% and 71%, respectively. The consumption of PAC was 42.8% lower than that of conventional processes.  A simple prediction method for effluent quality was established based on the verification of PAC multi-stage adsorption capacity equation for unknown component solution. When isotherm constants were amended with an initial concentration of adsorbate in a multi-component adsorption system, the PAC dose should be adjusted accordingly.  Greater than 70% of the pollutants in RO concentrate were removed via a PAC-MF accumulative countercurrent twostage adsorption process such that the effluent quality met the requirements of the RO process. Therefore, the effluent can be reclaimed through additional RO processes to improve the overall recovery rate to 91.0%e93.8%. This level of recovery is significant for engineering applications.

Acknowledgements The authors are grateful for financial support from the National Natural Science Foundation of China (50908158) and the research fund for the Doctoral Program of Higher Education of China (20090032120040).

Appendix. Supplementary data Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.10.050.

references

APHA, AWWA and WEF, 2005. Standard Methods for the Examination of Water and Wastewater, 21st ed. APHA, Washington, DC. 5-16e5-18. AQSIQ and SAC, 2008. Test Method for Granular/Powdered Activated Carbon from Coal (National Standards GB/T), 7702.1, 7702.2, 7702.6, 7702.7, 7702.15, State General Administration of Quality Supervision Inspection and Quarantine,

225

Standardization Administration of the People’s Republic of China Beijing, China. Badruzzaman, M., Oppenheimer, J., Adham, S., Kumar, M., 2009. Innovative beneficial reuse of reverse osmosis concentrate using bipolar membrane electrodialysis and electrochlorination processes. Journal of Membrane Science 326 (2), 392e399. Benner, J., Salhi, E., Ternes, T., von Gunten, U., 2008. Ozonation of reverse osmosis concentrate: kinetics and efficiency of beta blocker oxidation. Water Research 42 (12), 3003e3012. Campinas, M., Rosa, M.J., 2010. Assessing PAC contribution to the NOM fouling control in PAC/UF systems. Water Research 44 (5), 1636e1644. Campos, C., Snoeyink, V.L., Marin˜as, B., Baudin, I., Laine, J.M., 2000. Atrazine removal by powdered activated carbon in floc blanket reactors. Water Research 34 (16), 4070e4080. Chang, C.F., Chang, C.Y., Ho¨ll, W., Ulmer, M., Chen, Y.H., Groß, H. J., 2004. Adsorption kinetics of polyethylene glycol from aqueous solution onto activated carbon. Water Research 38 (10), 2559e2570. Chaplin, B.P., Schrader, G., Farrell, J., 2010. Electrochemical destruction of N-Nitrosodimethylamine in reverse osmosis concentrates using boron-doped diamond film electrodes. Environmental Science & Technology 44 (11), 4264e4269. Comerton, A.M., Andrews, R.C., Bagley, D.M., 2005. Evaluation of an MBR-RO system to produce high quality reuse water: microbial control, DBP formation and nitrate. Water Research 39 (16), 3982e3990. Corwin, C.J., Summers, R.S., 2010. Scaling trace organic contaminant adsorption capacity by granular activated carbon. Environmental Science & Technology 44 (14), 5403e5408. Dialynas, E., Diamadopoulos, E., 2008. Integration of immersed membrane ultrafiltration with coagulation and activated carbon adsorption for advanced treatment of municipal wastewater. Desalination 230 (1e3), 113e127. Gur-Reznik, S., Katz, I., Dosoretz, C.G., 2008. Removal of dissolved organic matter by granular-activated carbon adsorption as a pretreatment to reverse osmosis of membrane bioreactor effluents. Water Research 42 (6e7), 1595e1605. Jarusutthirak, C., Amy, G., Croue´, J.P., 2002. Fouling characteristics of wastewater effluent organic matter (EfOM) isolates on NF and UF membranes. Desalination 145 (1e3), 247e255. Knappe, D.R.U., Matsui, Y., Snoeyink, V.L., 1998. Predicting the capacity of powdered activated carbon for trace organic compounds in natural waters. Environmental Science & Technology 32 (11), 1694e1698. Ko¨rbahti, B.K., Tanyolac¸, A., 2009. Electrochemical treatment of simulated industrial paint wastewater in a continuous tubular reactor. Chemical Engineering Journal 148 (2e3), 444e451. Kumar, A., Prasad, B., Mishra, I.M., 2008. Adsorptive removal of acrylonitrile by commercial grade activated carbon: kinetics, equilibrium and thermodynamics. Journal of Hazardous Materials 152 (2), 589e600. Kumar, A., Prasad, B., Mishra, I.M., 2010. Isotherm and kinetics study for acrylic acid removal using powdered activated carbon. Journal of Hazardous Materials 176 (1e3), 774e783. Laabs, C.N., Amy, G.L., Jekel, M., 2006. Understanding the size and character of fouling-causing substances from effluent organic matter (EfOM) in low-pressure membrane filtration. Environmental Science & Technology 40 (14), 4495e4499. Li, Q.L., Marin˜as, B.J., Snoeyink, V.L., Campos, C., 2003. Threecomponent competitive adsorption model for flow-through PAC systems. 1. model development and verification with a PAC/membrane system. Environmental Science & Technology 37 (13), 2997e3004. Li, L., Zhu, W., Zhang, P., Zhang, Z., Wu, H., Han, W., 2006. Comparison of AC/O3-BAC and O3-BAC processes for removing

226

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 1 8 e2 2 6

organic pollutants in secondary effluent. Chemosphere 62 (9), 1514e1522. Lu, Q.L., Sorial, G.A., 2009. A comparative study of multicomponent adsorption of phenolic compounds on GAC and ACFs. Journal of Hazardous Materials 167 (1e3), 89e96. Ozaki, H., Li, H.F., 2002. Rejection of organic compounds by ultralow pressure reverse osmosis membrane. Water Research 36 (1), 123e130. Paraskeva, P., Graham, N.J.D., 2005. Treatment of a secondary municipal effluent by ozone, UV and microfiltration: microbial reduction and effect on effluent quality. Desalination 186 (1e3), 47e56. 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 Research 44 (9), 2763e2772. Sahu, J.N., Acharya, J., Meikap, B.C., 2009. Response surface modeling and optimization of chromium(VI) removal from aqueous solution using Tamarind wood activated carbon in batch process. Journal of Hazardous Materials 172 (2e3), 818e825. 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 (7185), 301e310. Tam, L.S., Tang, T.W., Lau, G.N., Sharma, K.R., Chen, G.H., 2007. A pilot study for wastewater reclamation and reuse with MBR/ RO and MF/RO systems. Desalination 202 (1e3), 106e113. To, P.C., Marin˜as, B.J., Snoeyink, V.L., Ng, W.J., 2008. Effect of poreblocking background compounds on the kinetics of trace

organic contaminant desorption from activated carbon. Environmental Science & Technology 42 (13), 4825e4830. Tseng, R.L., Wu, F.C., 2009. Analyzing a liquid e solid phase countercurrent two- and three-stage adsorption process with the Freundlich equation. Journal of Hazardous Materials 162 (1), 237e248. Westerhoff, P., Moon, H., Minakata, D., Crittenden, J., 2009. Oxidation of organics in retentates from reverse osmosis wastewater reuse facilities. Water Research 43 (16Sp. Iss. SI), 3992e3998. Wu, F.C., Tseng, R.L., 2008. Liquid-solid phase countercurrent multi-stage adsorption process for using the Langmuir equation. Journal of Hazardous Materials 155 (3), 449e458. Wu, F.C., Tseng, R.L., Huang, S.C., Juang, R.S., 2009. Characteristics of pseudo-second-order kinetic model for liquid-phase adsorption: a mini-review. Chemical Engineering Journal 151 (1e3), 1e9. Yu, Q., Zhang, R.Q., Deng, S.B., Huang, J., Yu, G., 2009. Sorption of perfluorooctane sulfonate and perfluorooctanoate on activated carbons and resin: kinetic and isotherm study. Water Research 43 (4), 1150e1158. Zhang, C.P., Gu, P., Zhao, J., Zhang, D., Deng, Y., 2009. Research on the treatment of liquid waste containing cesium by an adsorption-microfiltration process with potassium zinc hexacyanoferrate. Journal of Hazardous Materials 167 (1e3), 1057e1062. Zheng, X., Ernst, M., Huck, P.M., Jekel, M., 2010. Biopolymer fouling in dead-end ultrafiltration of treated domestic wastewater. Water Research 44 (18), 5212e5221.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 2 7 e2 3 4

Available online at www.sciencedirect.com

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

Microscale geochemical gradients in Hanford 300 Area sediment biofilms and influence of uranium Hung Duc Nguyen a, Bin Cao a,b,1, Bhoopesh Mishra c, Maxim I. Boyanov c, Kenneth M. Kemner c, Jim K. Fredrickson b, Haluk Beyenal a,* a

The Gene and Linda Voiland School of Chemical Engineering and Bioengineering, Washington State University, Pullman, WA, USA Pacific Northwest National Laboratory, Richland, WA, USA c Argonne National Laboratory, Argonne, IL, USA b

article info

abstract

Article history:

The presence and importance of microenvironments in the subsurface at contaminated

Received 2 August 2011

sites were suggested by previous geochemical studies. However, no direct quantitative

Received in revised form

characterization of the geochemical microenvironments had been reported. We quanti-

13 October 2011

tatively characterized microscale geochemical gradients (dissolved oxygen (DO), H2, pH,

Accepted 25 October 2011

and redox potential) in Hanford 300A subsurface sediment biofilms. Our results revealed

Available online 31 October 2011

significant differences in geochemical parameters across the sediment biofilm/water interface in the presence and absence of U(VI) under oxic and anoxic conditions. While the

Keywords:

pH was relatively constant within the sediment biofilm, the redox potential and the DO and

Sediment

H2 concentrations were heterogeneous at the microscale (<500e1000 mm). We found

Biofilm

microenvironments with high DO levels (DO hotspots) when the sediment biofilm was

Uranium

exposed to U(VI). On the other hand, we found hotspots (high concentrations) of H2 under

Microelectrode

anoxic conditions both in the presence and in the absence of U(VI). The presence of anoxic

Hotspot

microenvironments inside the sediment biofilms suggests that U(VI) reduction proceeds

Microenvironment

under bulk oxic conditions. To test this, we operated our biofilm reactor under airsaturated conditions in the presence of U(VI) and characterized U speciation in the sediment biofilm. U LIII-edge X-ray absorption spectroscopy (XANES and EXAFS) showed that 80e85% of the U was in the U(IV) valence state. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

The Hanford Site (approximately 1500 km2), located north of the city of Richland in southeastern Washington State, was once home to weapons grade plutonium (Pu) production for World War II and the Cold War (Brown et al., 2010; McKinley et al., 2007; Stubbs et al., 2009). The 300 Area (300A), in the southeastern corner of the Hanford Site, is the location where uranium processing and fuel fabrication took place from 1943

to 1988 (Gerber, 1992; Stubbs et al., 2009). As a consequence of these activities, waste streams including uranyl nitrate hexahydrate and various organic and inorganic chemicals containing dissolved metals were discharged to cribs, ponds and trenches (Gerber, 1992). Total uranium (U) inventories discharged to the 300A were approximately 70,100 kg, resulting in significant U contamination in the sediments and groundwater beneath and near the 300A disposal facilities (Gerber, 1992; Stubbs et al., 2009). A plume of U-contaminated

* Corresponding author. E-mail address: [email protected] (H. Beyenal). 1 Present address: School of Civil & Environmental Engineering, and Singapore Centre on Environmental Life Sciences Engineering, Nanyang Technological University, Singapore. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.054

228

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 2 7 e2 3 4

groundwater due to chemical leaching from disposal ponds at the 300A has existed for more than 30 years and persists today, continuing to discharge U into the adjacent Columbia River, particularly through the interface between the groundwater and the riverbed (McKinley et al., 2007; Peterson et al., 2008; Stubbs et al., 2009). The Hanford formation sediments in the vadose zone are the likely source of U in the groundwater (McKinley et al., 2007). In natural environments U exists in two predominant oxidation states: U(VI) and U(IV). U(VI) exists in the form of uranyl ðUO2þ 2 Þ complexing with other inorganic and organic ions (Beyenal et al., 2004; Fredrickson et al., 2000). U biogeochemistry is relatively complicated, and a range of variables, including redox conditions and pH, affect U speciation and therefore mobility in the subsurface environment (Cao et al., 2010; Suzuki and Suko, 2006; Wall and Krumholz, 2006). For example, the reduction of U(VI) to U(IV), by either microbial or abiotic means, generally results in the formation of uraninite (UO2) or nonuraninite products such as U(VI)orthophosphates that are sparingly soluble under neutral pH in the absence of strong U(IV)-complexing ligands (Fredrickson et al., 2000; Lovley and Phillips, 1992). In several recent studies, microscale environments that exhibited chemical characteristics distinct from those of the bulk phase were an important factor controlling the fate of U (Brown et al., 2010; McKinley et al., 2006). McKinley et al. (2006) reported that uranyl silicates, phosphates, and carbonates were observed in microscale sediment fractures, even though the average concentration of U measured in the bulk sediment samples was so low that no uranyl precipitates were predicted by geochemical models (Brown et al., 2010). These results demonstrate that the geochemical conditions in microscale environments can be significantly different from those in the bulk phase. In the subsurface, biofilms can form at the interface between sediment and groundwater, generating microenvironments where geochemistry can vary significantly over short distances (Bennett et al., 2000). Therefore, the characterization of microscale geochemical conditions inside biofilms is essential for a comprehensive understanding of U mobility and prediction of the fate of U in the subsurface. The goals of this study were to characterize, quantitatively, the microscale geochemical gradients in Hanford 300A subsurface sediment biofilm and the influence of U(VI) on these gradients. We quantified the dissolved oxygen (DO) and H2 concentration, pH, and redox potential profiles in sediment biofilms grown in laboratory reactors fed with synthetic groundwater amended with organics. The microscale geochemical gradients were measured under oxic and anoxic conditions. Last, average uranium speciation in the sediment biofilms was characterized using synchrotron X-ray absorption spectroscopy (XANES and EXAFS).

the Department of Energy’s (DOE) Hanford 300 Area Integrated Field Research Challenge (IFRC) site (http://ifchanford.pnl. gov). Our sediment materials were from a depth of 10e10.3 m in the Hanford formation, where the organic carbon content is relatively high (w100 mmol/g) (Lin et al., unpublished). At this depth, groundwater contains dissolved oxygen at a near-saturation level and microbial diversity is high, with aerobes, fermenters and denitrifier as the dominant microorganisms (Lin et al., unpublished). Each sediment sample was repacked into a polycarbonate column reactor (internal diameter 2.54 cm and length 40 cm) fitted with flow distributors made of 0.3-cm glass beads entrapped between two plastic sieves for uniform flow distribution and placed in the inlet and outlet of the column reactor. The column was filled with approximately 350 g of sediment, and the total porosity of the repacked sediment column was 38.7  5.0%. The column was fed with synthetic groundwater (SGW) amended with organic compounds (2 mM lactate, 2 mM malate, 2 mM succinate and 2 mM fumarate). The SGW consisted of KHCO3, 7.0 mg/L; MgSO4$7H2O, 51.7 mg/L; Ca(NO3)2$4H2O, 42.4 mg/L; CaCl2$2H2O, 61.7 mg/L; Na2SO4, 19.9 mg/ L; and NaHCO3, 92.4 mg/L. This formulation, based on the groundwater composition at the Hanford 300A IFRC site, was provided courtesy of J. McKinley, Pacific Northwest National Laboratory (PNNL). The feed flow for the column was in the upward direction and was maintained at a flow rate of 115 mL/day.

2.2.

Flat plate reactor for microscale measurements

2.

Materials and methods

After sediment biofilm growth in the column had been stimulated for five months, sediment (around 2 g) from the column reactor was placed into a flat plate reactor (Fig. 1), which allowed us to conduct microelectrode measurements. This reactor was operated with the same medium that was used to stimulate the growth of sediment biofilm. Where noted, uranyl chloride was added to obtain a final concentration of 98 mM. It should be noted that, although the U concentration in the groundwater at the Hanford 300A is reported to be lower than 1 mM (Miley et al., 2007), we used a higher U concentration in this study in order to i) investigate detectable metabolic responses of the sediment biofilm to U addition and ii) accumulate a sufficient mass of U in the sediment biofilm for XAFS analysis within a short period of time (w100 h). Throughout the experiment the recirculation rate was 200 mL/h with a residence time of 1.25 h. The flat plate reactor and medium were purged with continuously pumped filter-sterilized air or N2 to keep the bulk environment oxic or anoxic, respectively. DO microelectrodes were used to measure DO concentrations in the sediment biofilm under bulk oxic conditions. Redox potential, H2 concentration, and pH measurements were performed under both oxic and anoxic conditions. A schematic illustration of the experimental setup and a photograph of the sediment biofilm, along with the measurement locations, are shown in Fig. 1.

2.1.

Sediment biofilm preparation

2.3.

A core sample (well ID #C-6190, 2-13/C/330 -340 ) was collected, using resonant sonic drilling, from the Hanford formation at

Microelectrodes

We followed the procedures described by Lewandowski and Beyenal (Lewandowski and Beyenal, 2007) to construct and

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 2 7 e2 3 4

Electrometer/Picoammeter

ADC board

Sediment biofilms Stepper motor controller Stepper motor

Microelectrode Medium

N2 or air

Rubber stopper Vent

2.54 cm 6.35 cm Outlet 1.20 cm Recirculation 12.70 cm 2.54 cm

2.54 cm

Fig. 1 e Schematic illustration of the flat plate biofilm reactor and the experimental setup for microelectrodebased measurements. The rubber stopper was removed during measurements. A positive pressure of N2 prevented oxygen intrusion into the system for measurements under anoxic conditions. The top left photograph shows the sediment biofilm and measurement locations (marked with red stars). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

calibrate the redox potential, DO, and pH microelectrodes and the procedure of Ebert and Brune (Ebert and Brune, 1997) for the H2 microelectrode. The redox potential microelectrode is essentially a platinum (Pt) wire with a porous platinized Pt bulb formed by electroplating Pt on Pt wire (w20 mm in diameter). Redox potential microelectrodes were calibrated using YSI 3682 Zobell Solution (YSI Incorporated, Yellow Springs, Ohio, USA, 937-767-7241). The sensitivity of the redox potential microelectrode was 5 mV. DO and H2 microelectrodes are amperometric sensors with gold or platinized platinum, respectively, as the working electrode and an Ag/ AgCl half-cell as the counter electrode. The detection limits of the DO, H2, and pH microelectrodes were 0.01 mg/L, 50 nM and 0.02 units, respectively.

2.4.

Norton, MA). We measured profiles at seven points along the sediment biofilm/synthetic groundwater interface over a 1.2cm horizontal distance with approximately 0.2 cm between each two consecutive points (Fig. 1). The profiles were used to generate 2D contour maps as a function of horizontal relative distance (x-axis) and depth of measurement (y-axis) using SigmaPlot (Systat Software Inc., Version 11.0).

2.5.

2 mm

Microelectrode measurements and data analysis

The movement of microelectrodes was controlled by a Mercury Step motor controller PI M-230.10S Part No. M23010SX (Physik Instrumente, Auburn, MA) controlled by custom Microprofiler software. In each measurement, the microelectrode was moved downwards from the bulk synthetic groundwater to the sediment biofilm by a step of 10 mm. Data were recorded on a laptop using an Analog/Digital Converter (ADC, Measurement Computing USB-1608FS,

229

The flat plate reactor operation

The reactor was operated under both oxic and anoxic conditions in the presence or absence of U(VI). The operation timeline and the conditions are shown in Fig. S1. For the first set of experiments, the reactor was operated under airsaturated conditions without U, and DO and H2 concentrations, redox potential, and pH profiles were measured in the sediment biofilm. Then, the medium with U was fed into the reactor, and approximately 100 h later, profiles were measured again. Another set of experiments were performed on the same sediment biofilms where identical conditions were employed; however, the reactor was purged with N2 instead of air. Then, H2 concentration, redox potential, and pH profiles were measured. After that the reactor was switched to air-saturated conditions and the reactor was operated for another 100 h. Finally, the reactor was stopped and sediment biofilm samples were prepared for XANES and EXAFS analyses in an anaerobic chamber. The samples were preserved under anaerobic conditions.

2.6.

Scanning electron microscopy

Sediment biofilm samples were taken from the flat plate reactor and imaged using a FEI 200F scanning electron microscope (SEM) (FEI Company, Hillsboro, Oregon) without further preparation in low-vacuum imaging mode. The images were taken with a spot size of 3.5 nm at an acceleration voltage of 10 kV and a vacuum pressure of 120 Pa.

2.7.

XAFS measurement and data analysis

The hydrated sediment biofilm sample was mounted in a drilled plexiglass slide and sealed with Kapton film windows under anoxic conditions. The redox integrity of the sample was maintained during data collection by purging the sample chamber with N2. U LIII-edge (17,166-eV) XAFS measurements were performed at room temperature in sector 10-ID of the Advanced Photon Source (Segre et al., 2000). The beamline undulator was tapered, and the energy of the incident X rays was scanned using a Si(111) cryogenically cooled doublecrystal monochromator. Harmonic content was removed by reflection from a Rh-coated harmonic rejection mirror. EXAFS scans were collected in quick scanning mode (3 min each). Energy calibration was maintained by the simultaneous collection of data from a hydrogen uranyl phosphate standard. The final spectrum was produced by averaging 30 quick EXAFS scans. Processing of the raw data was done using ATHENA (Ravel and Newville, 2005). The EXAFS data were refined using theoretical models. The crystal structure of uraninite, UO2, was used to generate theoretical EXAFS spectra with programs ATOMS and FEFF8 (Ankudinov et al.,

230

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 2 7 e2 3 4

1998; Ravel and Newville, 2005; Wyckoff, 1960). Data were analyzed using the UWXAFS package (Stern et al., 1995) implemented in ARTEMIS (Ravel and Newville, 2005). The Fourier-transformed c(R) spectra were simultaneously fitted at three different k-weights (k1, k2, k3).

3.

Results and discussion

3.1.

Stimulated sediment biofilm

The SEM images indicated surface-associated cells and extracellular polymeric substances on the mineral surfaces in the organic stimulated sediment biofilm (Fig. S2). As biofilms develop on surfaces, the mineral surface area that is directly exposed to the groundwater decreases, decreasing the mass transfer of both water and solute from the groundwater to the surface (Renslow et al., 2010). As a result of decreased mass transfer, the geochemical conditions within the biofilms and near the mineral surfaces become different from those in the bulk groundwater filling the interstitial spaces between particles. These differences will generate microenvironments in the sediment biofilm. In the presence of an appropriate electron donor, microbial metabolism will consume O2 more rapidly than it is supplied by diffusion from the bulk solution, resulting in the development of anoxia and anaerobic microbial metabolism in the deeper parts of the biofilm.

3.2.

Geochemical conditions in the sediment biofilm

Geochemical parameters in the sediment biofilmdredox potential, DO, H2, and pH under oxic/anoxic conditions with and without uraniumdwere quantified. All geochemical conditions inside the biofilm were different from those in the bulk solution.

3.2.1.

Redox potentials

Fig. 2 shows the redox potentials in the bulk aqueous phase and those inside the sediment biofilm. As expected, the redox potentials in the bulk aqueous phase were constant and relatively high (i.e., oxidizing). However, inside the biofilm, there was a significant decrease in redox potential as a function of depth. The redox potentials under anoxic conditions in the biofilm were approximately 200 mV lower than those in the same positions under air-saturated conditions. Interestingly, even in the presence of air-saturated groundwater, redox potentials inside the biofilm were significantly lower (for example, w200 mV against a standard hydrogen electrode (SHE) at 500 mm below the biofilm/groundwater interface) than those in the bulk aqueous phase (w400 mV against SHE). This suggests that even under bulk oxic conditions redox-variable microenvironments develop inside the sediment biofilm. Depending on the Ca2þ and CO2 3 concentrations as well as factors such as the pH and the concentrations of complexing ligands, the mid-point potential of the U(IV)/U(VI) couple in natural environments is between 42 mV and 86 mVSHE (Brooks et al., 2003; Hazen and Tabak, 2005). Under anoxic conditions, the redox potential in the sediment biofilm decreased from þ270 mV at the interface to 0 mV at w500 mm below the sediment biofilm/groundwater interface. Therefore

Fig. 2 e Redox potential profiles near the sediment biofilm/ synthetic groundwater interface under (A) oxic (airsaturated) and (B) anoxic conditions. This interface is at depth [ 0.

we expect that the redox environment inside the sediment biofilm is favorable for U reduction.

3.2.2.

Dissolved oxygen

Consistent with the decrease in redox potential DO concentrations inside the sediment biofilm decreased with depth (Fig. 3). In the bulk aqueous phase, DO was as high as 3.12e4.68 mg/L, while at a depth of 400 mm inside the sediment biofilm DO levels decreased to approximately 0.62e0.94 mg/L. These results suggest that, even in the presence of high DO concentrations in the bulk aqueous phase, there are suboxic and perhaps anoxic zones in the sediment biofilm. Interestingly, a trend of decreasing DO concentration inside the bulk aqueous phase near the interface was observed. The sharp DO gradient into the sediment biofilm indicates that there was significant oxygen consumption in the biofilm, likely due to a combination of direct (aerobic respiration) and indirect (abiotic reactions with Fe(II), H2S, and other products of anaerobic metabolism) microbially driven processes.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 2 7 e2 3 4

3.2.3.

Fig. 3 e DO concentrations near the sediment biofilm/ synthetic groundwater interface. The sediment biofilm/ bulk water phase and bulk water/air phase interfaces are at depths 0 and 3750 mm, respectively.

In the Hanford 300A IFRC site, the groundwater in the upper unconfined aquifer within the Hanford formation is oxic. The presence of anoxic microenvironments inside the sediment biofilms suggests that microbial U reduction may proceed under global oxic conditions. In the presence of either endogenous or exogenous electron donors in mass transfer limited environments, such as biofilms, microenvironments in the aquifer could exist but remain undetected in measurements of groundwater collected from wells (bulk phases). In situ microbial populations can be readily stimulated to form biofilms, and these can force microscale redox potential and DO gradients that promote anoxia and anaerobic metabolism within biofilms, as demonstrated here.

231

H2

Fig. 4 shows the dissolved H2 profiles in the sediment biofilms under anoxic conditions. Under oxic conditions, H2 was below our detection limit (<50 nM) in both the aqueous bulk phase and the sediment biofilm. Under anoxic conditions, H2 was also below the detection limit (<50 nM) in the bulk aqueous phase. Interestingly, a steep H2 gradient within sediment biofilm was measured with concentrations as high as 70 mM in certain locations inside the sediment biofilm. These local concentrations are considerably higher than those measured in sediments (Lovley and Goodwin, 1988). The concentrations measured in our experiments suggest that the H2 production could have been driven by the fermentation of organic compounds supplied during stimulation. H2 concentrations up to 100 mM have been reported in termite guts (Ebert and Brune, 1997), where microbial metabolism is driven largely by organic matter fermentation. To the best of our knowledge, these are the first microscale measurement of H2 concentration in sediment biofilm. H2 can be used as an electron donor by many organisms, including those that reduce metals (Lovley, 1993). Such microenvironment hotspots may contribute to U reduction in the subsurface even under aerobic conditions in the bulk solution. The sharp decrease in H2 concentration towards the bulk solution could be due to a combination of physical effects such as the removal of H2 from the solution (since N2 gas was pumped continuously) and the microbial consumption of H2 at locations in the gradient where electron acceptors are available.

3.2.4.

pH

The pH profiles in the sediment biofilm are shown in Fig. S3. In both anoxic and oxic conditions, the pH inside the sediment biofilm was relatively constant, with a variation of w0.3 units, suggesting that the pH did not vary significantly inside the biofilm. Interestingly, the slightly lower pH at the bottom of the sediment was consistent with the generation of acidic fermentation products and H2 at the base of the biofilm.

3.3. Effects of uranium on biogeochemical microenvironments

Fig. 4 e H2 concentration inside the sediment biofilm under anoxic conditions. The sediment biofilm/bulk synthetic groundwater interface is at depth [ 0.

U(VI) can serve as an electron acceptor for the anaerobic respiration of certain metal-reducing organisms (Fredrickson et al., 2000; Lovley, 1993; Sanford et al., 2007), but it can also be toxic and inhibit microbial growth and metabolism. To investigate the effects of uranium exposure on the biogeochemical microenvironments within sediment biofilm, uranyl chloride at a final concentration of 98 mM was infused into the sediment biofilm. The profiles of redox potential, DO, H2 and pH in the presence of U(VI) are available in the Supporting Information (Fig. S4eS7). Redox potentials in the sediment biofilm increased with the addition of U(VI) under oxic conditions but were not significantly affected under anoxic conditions (Fig. S4). Under oxic conditions the DO levels within and near the surface of the biofilm decreased from 2.81 to 6.24 mg/L in the absence of U(VI) to 1.56e5.93 mg/L in the presence of U(VI). Most interestingly, although in general DO decreased with depth, we

232

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 2 7 e2 3 4

observed a region of high oxygen concentration in the sediment biofilm when U(VI) was present (Fig. S5). In addition, under anoxic conditions the H2 concentrations inside the sediment biofilm with U(VI) (Fig. S6) were much lower than those without U(VI) (Fig. 4). The pH inside the sediment biofilm was not significantly affected by U(VI) (Fig. S7). At least two main processes could be affected when U(VI) is introduced into sediment reactors: abiotic geochemical reactions and microbial metabolism (Finneran et al., 2002; Fredrickson et al., 2000; Gorby and Lovley, 1992; Lovley and Phillips, 1992; Lovley et al., 1991; O’Loughlin et al., 2003; Suzuki et al., 2002). U(VI) likely undergoes re-speciation, potentially forming complexes with naturally occurring organic compounds (humic compounds, cells and extracellular polymer substances, etc.) and inorganic ions (e.g., CO2 3 ). U(VI) may also be reduced abiotically in the sediment biofilm by reducing products of microbial metabolism. On the other hand, U(VI) may inhibit microbial metabolism. It has been reported that a U(VI)- and sulfate-reducing consortium from a field site at the Oak Ridge Field Research Center was inhibited by U(VI) when the concentration was increased to 224 mM from 49 mM (Nyman et al., 2007). Similarly, U(VI) at a concentration of 60 mM and higher inhibited anaerobic metabolism in Hanford aquifer sediments (J. Fredrickson, unpublished data). A high DO microenvironment developed in the sediment biofilm after U was introduced, suggesting that the microbial activities and oxygen respiration in that location were inhibited by U, which in turn may have disrupted the DO gradient within the biofilm. Similarly, microbial fermentation processes may have been inhibited in the presence of U(VI), resulting in the significantly lower H2 concentrations that were observed inside the sediment biofilm.

3.4.

U speciation

Since we found anoxic microenvironments inside the sediment biofilm, we hypothesized that U reduction can proceed

under bulk oxic conditions. To test this hypothesis, we operated our biofilm reactor under air-saturated conditions and characterized U speciation. Fig. 5 compares the U LIIIedge XANES data from the stimulated sediment biofilm sample to spectra from nanoparticulate uraninite (UO2) and from aqueous uranyl carbonate, U(VI). Linear combination analysis of the XANES data indicates that 80e85% of the uranium in the stimulated sediment biofilm sample is in the U(IV) valence state. The Fourier-transformed (FT) EXAFS data from the stimulated sediment biofilm sample (Fig. S8) show significant similarity to those of previously characterized uraninite nanoparticles (Burgos et al., 2008). The smaller amplitudes of the UeO and UeU peaks in the stimulated sediment biofilm sample relative to those of the completely reduced nanoparticulate uraninite could be due in part to spectral interferences with the small oxidized component, but more likely arise from the uraninite in the stimulated sediment biofilm sample being of a smaller particle size than the standard (O’Loughlin et al., 2003; Suzuki et al., 2002). The best fit to the data using an EXAFS model similar to that described in (Burgos et al., 2008) is shown in Fig. S8. The model includes contributions from nearest-neighbor O1 and next-nearest-neighbor U atoms in uraninite, together with an Oax contribution to account for an oxidized component. The best fit values obtained with this model are given in Table S1. To avoid the large parameter correlations of the overlapping Oax and O1 shells, the Debye-Waller parameter (s2) for the smaller Oax signal was fixed to the value reported in (Burgos et al., 2008). The UeOax coordination number of 0.44  0.18 suggests a 78  9% U(IV) content in the sediment biofilm, which is consistent with the 80e85% U(IV) content determined by XANES analysis. The UeO1 coordination number of 6.1  0.8 is also consistent with the fraction of reduced uranium in the sediment biofilm. The UeU coordination number of 5.6  1.7 corresponds to an average particle diameter of 1.0  0.5 nm for completely reduced uraninite particles (Boyanov et al., 2007).

3.5. Practical implications of microenvironments in sediment biofilms

Fig. 5 e U LIII-edge XANES spectrum of the stimulated sediment biofilm sample from the microelectrode experiment setup after w100-h exposure to U(VI). The U(IV) and U(VI) standards used for this analysis are biogenically reduced nanoparticulate uraninite and aqueous uranyl carbonate, respectively.

The microscale conditions inside the sediment biofilms from Hanford 300A exhibited high heterogeneity, and they were distinct from those measured in the overlying bulk fluid. These results also support the importance of microscale measurements for understanding potential biogeochemical transformations of contaminants in the subsurface. In the presence of oxygen-saturated synthetic groundwater, the redox potential and DO concentration inside the sediment biofilm were much lower than those in the bulk fluid, suggesting that even under oxic conditions U immobilization is still viable and that reoxidation by O2 occurs only at the sediment biofilm/groundwater interface and not throughout the sediment biofilm. Although the H2 concentration in the bulk fluid was below detection (<50 nM), there were hotspots with concentrations up to 70 mM inside the biofilm. H2 can serve as an electron donor for microbial metal reduction, facilitating the biotransformation of contaminants inside the sediment biofilm. Finally,

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 2 7 e2 3 4

microenvironments within the Hanford 300A sediment biofilms may be critically important to describe the fate and transport of contaminants.

4.

Conclusions

The presence and importance of microenvironments in the subsurface at contaminated sites were suggested by previous geochemical studies. However, no direct quantitative characterization of geochemical microenvironments had been reported. We have now quantitatively characterized microscale geochemical gradients (DO and H2 concentrations, pH, and redox potential) which create microenvironments in Hanford 300A subsurface sediment biofilms. We found significant differences in geochemical parameters across the sediment biofilm/water interface in the presence and absence of U(VI) under oxic and anoxic conditions. While the pH was relatively constant within the biofilm, the redox potential and the DO and H2 concentrations were heterogeneous at the microscale. Our results, for the first time, directly reveal the presence of geochemical hotspots, i.e., microenvironments with relatively high or low biogeochemical activities in sediment biofilms, which has important practical implications for bioremediation processes.

Acknowledgments This research was supported by the U.S. Department of Energy (DOE) Office of Biological and Environmental Research (BER) under the Subsurface Biogeochemical Research (SBR) Program (grant DE-FG92-08ER64560). PNNL contributions to this research were supported in part by the PNNL Scientific Focus Area, and Hanford 300A IFRC projects and ANL contributions were supported in part by the ANL Scientific Focus Area project. They are part of the SBR Program of the Office of Biological and Environmental Research (BER), U.S. DOE under contracts DE-AC05-76RLO and DE-AC02-06CH11357, respectively. Use of the Advanced Photon Source (APS) was supported by the DOE-SC Office of Basic Energy Sciences, under contract DE-AC02-06CH11357. MRCAT/EnviroCAT operations are supported by DOE and the MRCAT/EnviroCAT member institutions. The U.S. government retains for itself and others acting on its behalf a paid-up, non-exclusive, irrevocable worldwide license in said article to reproduce, prepare derivative works, distribute copies to the public and perform publicly and display publicly, by or on behalf of the Government.

Appendices Eight figures and one table are intended to supplement the material present in the manuscript. The figures display the experimental setup; images of sediment biofilms; pH profiles in the absence of U(VI); profiles of redox potential, DO, H2 and pH in the presence of U(VI); and the Fourier transform magnitude of the stimulated sediment biofilm EXAFS data. A table lists the detailed results of the fitting analysis of the EXAFS data.

233

Appendix. Supplementary information Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.10.054.

references

Ankudinov, A., Ravel, B., Rehr, J., Conradson, S., 1998. Real-space multiple-scattering calculation and interpretation of x-rayadsorption near-edge structure. Physical Review B 58, 7565e7576. Bennett, P., Hiebert, F., Rogers, J., 2000. Microbial control of mineral-groundwater equilibria: miacroscale to microscale. Journal of Hydrology 8, 47e62. Beyenal, H., Sani, R., Peyton, B., Dohnalkova, A., Amonette, J., Lewandowski, Z., 2004. Uranium immobilization by sulfatereducing biofilms. Environmental Science & Technology 38, 2067e2074. Boyanov, M., O’Loughlin, E., Roden, E., Fein, J., Kemner, K., 2007. Adsorption of Fe(II) and U(VI) to carboxyl-functionalized microspheres: the influence of speciation on uranyl reduction studies by titration and XAFS. Geochimica et Cosmochimica Acta 71, 1898e1912. Brooks, S., Fredrickson, J., Carroll, S., Kennedy, D., Zachara, J., Plymale, A., Kelly, S., Kemner, K., Fendorf, S., 2003. Inhibition of bacterial U(VI) reduction by calcium. Environmental Science & Technology 37, 1850e1858. Brown, C., Serne, R., Catalano, J., Krupka, K., Icenhower, J., 2010. Mineralization of contaminant uranium and leach rates in sediments from Hanford Washington. Applied Geochemistry 25, 97e104. Burgos, W., McDonough, J., Senko, J., Zhang, G., Dohnalkova, A., Kelly, S., Gorby, Y., Kemner, K., 2008. Characterization of uraninite nanoparticles produced by Shewanella oneidensis MR1. Geochimica et Cosmochimica Acta 72, 4901e4915. Cao, B., Ahmed, B., Beyenal, H., 2010. Immobilization of uranium in groundwater using biofilms. In: Shah, V. (Ed.), Emerging Environmental Technologies. Springer. Ebert, A., Brune, A., 1997. Hydrogen concentration profiles at the oxic-anoxic interface: a microsensor study of the hindgut of the wood-feeding lower termite Reticulitermes flavipes (Kollar). Applied and Environmental Microbiology 63, 4039e4046. Finneran, K., Anderson, R., Nevin, K., Lovley, D., 2002. Potential for bioremediation of uranium-contaminated aquifers with microbial U(VI) reduction. Soil and Sediment Contamination 11, 339e357. Fredrickson, J., Zachara, J., Kennedy, D., Duff, M., Gorby, Y., Li, S., Krupka, K., 2000. Reduction of U(VI) in geothite (alpha-FeOOH) suspensions by a dissimilatory metal-reducing bacterium. Geochimica et Cosmochimica Acta 64, 3085e3098. Gerber, M., 1992. Past Practices Technical Characterization Study: 300 Area, Hanford Site WHC-MR-0388 (Richland, WA). Gorby, Y., Lovley, D., 1992. Enzymatic uranium precipitation. Environmental Science & Technology 26, 205e207. Hazen, T., Tabak, H., 2005. Developments in bioremediation of soils and sediments polluted with metals and radionuclides: 2. Field research on bioremediation of metals and radionuclides. Reviews in Environmental Science Biology/Technology 4, 157e183. Lewandowski, Z., Beyenal, H., 2007. Foundamentals of Biofilms Research. CRC Press. Lovley, D., 1993. Dissimilatory metal reduction. Annual Review of Microbiology 47, 263e290. Lovley, D., Goodwin, S., 1988. Hydrogen concentrations as an indicator of the predominant terminal electron-accepting reactions in aquatic sediments. Geochimica et Cosmochimica Acta 52, 2993e3003.

234

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 2 7 e2 3 4

Lovley, D., Phillips, E., 1992. Reduction of uranium by Desulfovibrio desulfuricans. Applied and Environmental Microbiology 58, 850e856. Lovley, D., Phillips, E., Gorby, Y., Landa, E., 1991. Microbial reduction of uranium. Nature 350, 413e416. McKinley, J., Zachara, J., Liu, C., Heald, S., Prenitzer, B., Kempshall, B., 2006. Microscale controls on the fate of contaminant uranium in the vadose zone, Hanford Site, Washington. Geochimica et Cosmochimica Acta 70, 1873e1887. McKinley, J., Zachara, J., Wan, J., McCready, D., 2007. Geochemical controls on contaminant uranium in vadose Hanford formation sediments at the 200 area and 300 area, Hanford Site, Washington. Vadose Zone Journal 6, 1004e1017. Miley, T., Bunn, A., Napier, B., Peterson, R., Becker, J., 2007. Current Conditions Risk Assessment for the 300-FF-5 Groundwater Operable Unit. Pacific Northwest National Laboratory, Richland. Nyman, J., Wu, H.-I., Gentile, M., Kitanidis, P., Criddle, C., 2007. Inhibition of a U(VI)- and sulfate-reducing consortia by U(VI). Environmental Science & Technology 41, 6528e6533. O’Loughlin, E., Kelly, S., Cook, R., Csencsits, R., Kemner, K., 2003. Reduction of uranium(VI) by mixed iron(II)/iron(III) hydroxide (green rust): formation of UO2 nanoparticles. Environmental Science & Technology 37, 721e727. Peterson, R., Rockhold, M., Serne, J., Thorne, P., William, M., 2008. Uranium Contamination in the Subsurfaces beneath the 300 Area PNNL-17034 (Richland, WA, US-DOE). Ravel, B., Newville, M., 2005. ATHENA, ARTEMIS, HEPHAESTUS: data analysis for x-ray absorption spectroscopy using IFEFFIT. Journal of Synchrotron Radiation 12, 537e541.

Renslow, R., Majors, P., McLean, J., Fredrickson, J., Ahmed, B., Beyenal, H., 2010. In situ effective diffusion coefficient profiles in live biofilms using pulsed-field gradient nuclear magnetic resonance. Biotechnology and Bioengineering 106, 928e937. Sanford, R., Wu, Q., Sung, Y., Thomas, S., Amos, B., Prince, E., Loffler, F., 2007. Hexavalent uranium supports growth of Anaeromyxobacter dehalogenans and Geobacter spp. with lower than predicted biomass yields. Environmental Microbiology 9, 2885e2893. Segre, C., Leyarovska, N., Chapman, L., Lavender, W., Plag, P., King, A., Kropf, A., Bunker, B., Kemner, K., Dutta, P., Duran, R., Kaduk, J., 2000. The MRCAT inserton device beamline at the advanced photon source. In: Sync. Rad. Instrum.: Eleventh US Conference, pp. 419e422. Stern, E., Newville, M., Ravel, B., Yacoby, Y., Haskel, D., 1995. The UWXAFS analysis package- philosophy and details. Physica B 209, 117e120. Stubbs, J., Veblen, L., Elbert, D., Zachara, J., Davis, J., Veblen, D., 2009. Newly recognized hosts for uranium in the Hanford Site vadose zone. Geochimica et Cosmochimica Acta 73, 1563e1576. Suzuki, Y., Suko, T., 2006. Geomicrobiological factors that control uranium mobility in the environment: Update on recent advances in the bioremediation of uranium-contaminated site. Journal of Mineralogical and Petrological Sciences 101, 299e307. Suzuki, Y., Kelly, S., Kemner, K., Banfield, J., 2002. Radionuclide contamination: nanometer-size products of uranium bioreduction. Nature 134. Wall, J., Krumholz, L., 2006. Uranium reduction. Annual Review of Microbiology 60, 149e166. Wyckoff, R., 1960. Crystal Structures. Interscience, New York.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 3 5 e2 4 6

Available online at www.sciencedirect.com

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

In-pipe water quality monitoring in water supply systems under steady and unsteady state flow conditions: A quantitative assessment Angeliki Aisopou a,*, Ivan Stoianov b, Nigel J.D. Graham b a b

Department of Civil and Environmental Engineering, Imperial College London, South Kensington, London SW7 2AZ, UK Imperial College, London

article info

abstract

Article history:

Monitoring the quality of drinking water from the treatment plant to the consumers tap is

Received 16 February 2011

critical to ensure compliance with national standards and/or WHO guideline levels. There

Received in revised form

are a number of processes and factors affecting the water quality during transmission and

27 October 2011

distribution which are little understood. A significant obstacle for gaining a detailed

Accepted 31 October 2011

knowledge of various physical and chemical processes and the effect of the hydraulic

Available online 6 November 2011

conditions on the water quality deterioration within water supply systems is the lack of reliable and low-cost (both capital and O & M) water quality sensors for continuous

Keywords:

monitoring. This paper has two objectives. The first one is to present a detailed evaluation

Water supply

of the performance of a novel in-pipe multi-parameter sensor probe for reagent- and

Water quality

membrane-free continuous water quality monitoring in water supply systems. The second

Sensors

objective is to describe the results from experimental research which was conducted to

In-pipe

acquire continuous water quality and high-frequency hydraulic data for the quantitative

Monitoring

assessment of the water quality changes occurring under steady and unsteady-state flow

Multi-parameter

conditions. The laboratory and field evaluation of the multi-parameter sensor probe showed that the sensors have a rapid dynamic response, average repeatability and unreliable accuracy. The uncertainties in the sensor data present significant challenges for the analysis and interpretation of the acquired data and their use for water quality modelling, decision support and control in operational systems. Notwithstanding these uncertainties, the unique data sets acquired from transmission and distribution systems demonstrated the deleterious effect of unsteady state flow conditions on various water quality parameters. These studies demonstrate: (i) the significant impact of the unsteady-state hydraulic conditions on the disinfectant residual, turbidity and colour caused by the re-suspension of sediments, scouring of biofilms and tubercles from the pipe and increased mixing, and the need for further experimental research to investigate these interactions; (ii) important advances in sensor technologies which provide unique opportunities to study both the dynamic hydraulic conditions and water quality changes in operational systems. The research in these two areas is critical to better understand and manage the water quality deterioration in ageing water transmission and distribution systems. ª 2011 Elsevier Ltd. All rights reserved.

* Corresponding author. Present address: DTU Environment Department of Environmental Engineering Technical University of Denmark, Miljoevej Building 113, Lyngby 2800, Denmark. Tel.: þ45 28199952. E-mail addresses: [email protected], [email protected] (A. Aisopou), [email protected] (I. Stoianov), [email protected] (N.J.D. Graham). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.058

236

1.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 3 5 e2 4 6

Introduction

Significant spatial variations and temporal water quality changes can occur in water distribution systems. While considerable effort has been applied to the development of contaminant transport network models, the various processes and factors which affect water quality during transmission and distribution remain poorly understood and with limited experimental verification. The published experimental studies emphasise the need for continuous water quality monitoring with sufficient spatial and temporal resolution and demonstrate that the state of art sensing technologies fail to satisfy this requirement (e.g., Vreeburg, 2007; Yang et al., 2009; Panguluri et al., 2009). For example, Van den Hoven and Vreeburg (1992) and Vreeburg (2007) reported experimental results from a water distribution network in the Netherlands which included the measurement of pH, oxygen, turbidity, conductivity, temperature, pressure and flow, at three monitoring sites. The monitoring stations used an assembly of conventional reagent-based water quality sensors with a complex off-line flow extraction, a flow cell and drainage. The results demonstrated that continuous water quality data provide important information on sources of turbid water, residence times and pipe conditions. Woodward et al. (1995) conducted experiments using a large experimental pipe facility (TORUS rig, 1.3km long) to study the decay rates of monochloramine under laminar and turbulent flows. Monochloramine, pH, temperature, DO, redox and particle counts were measured at 10 min time intervals. The chlorine concentrations were continuously monitored using analysers installed in flow cells, but the performance of these analyzers and the fluctuations in the acquired chlorine measurements made data interpretation problematic. In addition, Maier et al. (2000) monitored disinfectant concentrations at two ends of the TORUS rig to estimate the wall demand coefficient during different hydraulic conditions. Total chlorine was measured using on-line meters, but limitations in the sensing technology presented challenges for the data analysis. The American Water Works Association carried out an experimental study which investigates the effects of water velocity, water quality parameters (pH, alkalinity and dissolved oxygen) and pipe materials on corrosion and disinfectant decay rate (AWWARF, 2006). Although general trends were observed, a relationship between corrosion and wall reaction rate was not defined due to the lack of accuracy and precision in the measurements. A recent study by the Greater Cincinnati Water Works, which to the best of our knowledge is the largest reported study of investigating chlorine decay in water distribution systems, used 38 temporary monitors to measure free chlorine and pH for a period of one week (Lee et al., 2010). The deployed monitors included reagent and membrane based sensor technologies installed in a flow cell attached to fire hydrants with slip stream drainage to discharge the water sample. This measurement set-up is expensive to deploy and not practical for long term monitoring. The data may also not be representative of the water quality in the pipes due to the volume of the by-pass/flow cell, the water sample and the long sampling interval. In addition, sub-zero temperatures prevent the use of this setup throughout the year.

The EPA has carried out an extensive experimental evaluation of online water quality sensors to assess whether they could provide a dual use for early warning of intentional contamination, as for monitoring general water quality in water supply networks (Panguluri et al., 2009). One of the main conclusions of the EPA’s experimental study was that free chlorine and total organic carbon (TOC) sensors were the most successful in detecting a wide range of chemical and biological contaminants. The sudden and significant change in free chlorine levels was used as an indicator for the presence of various contaminants while TOC sensors were successful in detecting carbon containing contaminants or carrier liquids. The detection method requires the acquisition of stable or predictable baseline water quality data and variability for each location. Having extensively evaluated the performance and the capital and operational and maintenance (O & M) costs of existing water quality sensing technologies, EPA recommended that sensor manufacturers need to develop reagent-free sensors that result in lower labor and consumable costs. These experimental studies outline a common challenge which is the availability of reliable and low-cost sensing technologies for accurate and continuous in-pipe water quality monitoring. The state of art in remote water quality sensing in distribution networks includes reagent and membrane based sensors adapted from technologies used in water treatment plants. The sensors are installed in a flow cell located in street bollards/cabinets and equipped with slip stream drainage to sewers or ground. These limitations encourage the development and use of reagent-free electrochemical sensors for in-pipe water quality monitoring. Unfortunately, no data are available to demonstrate the reliable operation of these sensors under controlled laboratory and operational conditions. The use of water quality sensors for general monitoring, operational decision support and potentially for early warning of contamination, can only be beneficial if the sensors’ characteristics such as resolution, repeatability and accuracy are within specified uncertainty limits over a defined period of time. Interactions between the authors and various water utilities indicated that the UK water industry is particularly cautious in the adoption of such technologies due to concerns regarding reliability, accuracy, O & M costs and the impact of false positive data. There is considerable concern that inherent limitations in the sensing technologies and the complex operational environment and processes such as bio-fouling, may result in rapid deterioration and poor quality data. As a result, signal drifting and noise will trigger frequent re-calibration, sensor maintenance and/or false positive alarms. The dynamic hydraulic conditions in operational systems are another factor which is rarely assessed and commonly ignored due to the lack of suitable instrumentation for continuous and long-term monitoring. The dynamic hydraulic conditions can adversely affect the water quality by increasing the amplitude and rate of change of the shear stress acting on the pipe wall which could lead to the re-suspension of sediments and the detachment of biofilms and corrosion products (Karney and Brunone, 1999; Vreeburg et al., 2004) and enhanced leaching

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 3 5 e2 4 6

of corrosion byproducts and copper/lead (Calle et al., 2007). Rapid and gradual flow changes occur as a result of stochastic changes in demand, pressure control, routine operational procedures or accidental events. In order to experimentally investigate the dynamic hydraulic conditions in operational water supply systems, the authors have been working on the design of novel technologies for the real-time and highfrequency monitoring of pressure and flow (Stoianov et al., 2006). A new generation of ultra-low-power remote telemetry units (RTUs) was developed at Imperial College London (InfraSense RTU, www.imperial.ac.uk/infrasense). These RTUs have been deployed at over 250 locations in the last three years to continuously monitor the dynamic pressure and flow conditions in transmission and distribution systems with sampling rates from 1S/s (flow) to 200S/s (pressure). The experimental results demonstrate that the assumption water supply systems operate under steady-state conditions is frequently violated. To the best of our knowledge, there have been no published experimental data which describe the impact of the unsteady state hydraulic conditions on the deterioration of water quality. Such experimental research requires the availability of time synchronised hydraulic and water quality data with sufficient spatial and temporal resolution. In view of the foregoing, the main motivation for the experimental research described in this paper was to study the water quality deterioration in water supply systems under unsteady-state hydraulic conditions. This was done by developing and applying technologies for time synchronised highfrequency sampling of the dynamic hydraulic conditions in operational systems and integrating these technologies with novel electrochemical reagent-free water quality sensors. Thus, this paper provides firstly, the results of a detailed evaluation of the performance of a novel multi-parameter sensor probe, and secondly, presents the results of extensive experimental research which was conducted to provide a quantitative assessment of the dynamic changes in water quality during steady and unsteady-state flow conditions. The results of the experimental research currently support the development of a disinfectant decay model which takes into account the increased shear stress during sudden and gradual acceleration and deceleration of the pipe flow (Aisopou et al., 2010).

2.

Experimental methods

Following a detailed review of existing sensor technologies, four multi-parameter probes for continuous water quality monitoring were identified as suitable candidates (YSI 6920DW & 600DW-B, Hydraclam, Censar and the Intellisonde). The Intellisonde (Intellitect Water, UK), was selected based on the method and ease of installation (e.g. in-pipe rather than in a flow cell), range of parameters and the accuracy provided by the manufacturer. The Intellisonde uses electrochemical and optical technologies to monitor conductivity, temperature, pH, free and total chlorine, dissolved oxygen, ORP, turbidity and colour. The sensors are reagent-free and the probe is inserted in pressurized water mains using a 50 mm valve. As this was novel technology with little information regarding the performance of the individual sensors and the probe, a detailed evaluation was

237

carried out under controlled conditions at the Roger Perry Environmental Engineering Laboratory (Imperial College London). The water quality sensor probe was evaluated in terms of accuracy, sensitivity, repeatability, dynamic response and calibration requirements. The following section describes the evaluation procedures and results.

2.1.

Evaluation under controlled laboratory conditions

The initial evaluation of the Intellisonde multi-parameter sensor probe was carried out under controlled laboratory conditions. The accuracy, sensitivity, repeatability, response time and device-to-device reproducibility were assessed with 1min sampling intervals over a period of three months.

2.1.1.

Testing procedure

The testing process followed procedures and recommendations outlined in ISO 5725-1 (1994), BS 1427 (2009), BS/EN ISO 15839 (2006). Two probes were simultaneously tested. The probes were placed in a 5 L glass beaker initially filled with tap water. The tap water had the following initial characteristics: conductivity 650 mS/cm, pH 8, turbidity 0.2NTU, colour 3oHazen, DO 10 mg/L, free Cl 0.02e0.05 mg/L and total Cl 0.16e0.2 mg/L. These values were obtained using standard laboratory reference methods (APHA, 2005). The solution was continuously stirred to achieve uniform aqueous conditions. Experiments were conducted to evaluate each electrochemical and optical sensor using stepwise increase and decrease in the concentration of specific additives for the tested parameters. Fig. 1a shows the sensors and their configuration. The testing procedure is summarised as following: 1) The probe was allowed to stabilise in the water sample and then calibrated using standard laboratory equipment (Table 1). The calibration started with temperature since the performance of the electrochemical sensors is temperature dependent. 2) The value of the parameter under evaluation was increased stepwise by adding the same amount of a specific reagent every 10 min. The reagents included NaOH and NaCl to increase pH and conductivity while humic acid was used to increase colour. A formazin solution was added to increase turbidity and diluted commercial sodium hypochlorite to increase the total residual chlorine. 3) The sensitivity and the dynamic response of each sensor were assessed by defining the smallest absolute amount of change that can be detected as well as the sensor response to a variable input of a reagent. The range of the parameters was within the values expected in water supply systems based on information from grab sampling programmes. 4) The tests were monitored using laboratory reference instruments which had accuracies and response times better than the evaluated sensors (Table 1). The reference instruments were periodically calibrated against standard solutions. 5) Once the test range of a parameter had been reached, the performance of the probe was further evaluated by stepwise decreasing the parameter value at equal time intervals (e.g. 10 min). This was achieved by adding HCl and RO

238

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 3 5 e2 4 6

Fig. 1 e Intellitect sensor probe and measurement setup for Case Study 1: a) Intellisonde, b) Measurement setup and longitudinal section for Case Study 1; c) InfraSenseLP measuring flow and pressure at F & P1,2,3; d) InfraSenseLP integrated with Intellisonde.

2.2. Experimental study in water transmission and distribution systems

water to decrease the pH and conductivity, respectively. Chlorine, colour and turbidity were decreased by dilution with tap. These tests aimed to assess hysteresis and reproducibility.

The experimental study included two field trials that were carried out involving operational water distribution networks and transmission mains. While the main objective of the experimental study was to provide a quantitative assessment of the changes in water quality during steady and unsteadystate flow conditions, the trials were also used to evaluate

The testing procedure was repeated five times under the same conditions. This allowed the estimation of repeatability and reproducibility (R & R) using the average and range method described by Montgomery and Runger (1993).

Table 1 e Specifications of sensor probe and reference meters. Characteristics obtained from laboratory experiments Accuracy

Sensitivity

Noise

R & Rc

pH Conductivity [mS/cm] Colour [oHazen]

0.4 10%

0.2 15

No No

3% 4%

4a

4

4

Turbidity [NTU]

1a

1e2

1

a dependent on cleanness. b dependent on water quality and flow. c Repeatability & Reproducibility.

Accuracy given by supplier (Intellitect, 2008)

Reference meters Meter

Accuracy

Sensitivity

0.2 5%

Hydrus 500 bechtop Jenway 470

0.1 0.5%, 2 digits

0.1 1 mS/cm

2

1b

0.1

0.1

1

0.5

Shimadzu UV-2401PC (450nm) Hach 2100A

0.05 (range 0e1) 0.5 (range 1e10)

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 3 5 e2 4 6

239

Fig. 2 e Dynamic response of sensor probe and reference meters in the laboratory tests for, a) pH, b) conductivity, c) colour, and d) turbidity (sample interval is 1 min; dashed line in c and d indicates the quality threshold for customer complaints).

the robustness and long-term performance of the electrochemical and optical sensors. Each field trial included three water quality probes and five InfraSenseLP RTUs for the longterm time-synchronized (10 m) and high-speed acquisition of hydraulic data. The logging interval for the water quality probes was 5min which was the minimum we could achieve without affecting the performance of the probe. The pressure was sampled continuously at 200S/s using a pressure transducer with a dynamic response of 1kHz and flow was sampled at 1S/s (ABB Aquaprobe v2). The pressure and flow signals were digitized and stored by the InfraSenseLP RTU. The InfraSenseLP RTU is an ultra-low power logger/RTU (22mW, 16-bit ADC, excluding the communication). The time reference is provided by a 60kHz receiver using the UK MSF radio signal. The Aquaprobe is interfaced with the InfraSenseLP RTU so that the flow velocity is sampled at 1 s intervals. During the installation procedure, each water quality sensor probe was

placed in a concentrated chlorine solution (16%, diluted 1:4) for 30 s to accelerate its stabilization, in view of the low concentration of chlorine in drinking water. Field visits were made once per week to take grab samples, calibrate the sensor probes if necessary, check their state and download the data. The first field trial, Case Study 1, was carried out on a 1.22 m diameter water transmission main over a period of 6 weeks. The pipe topology and the layout of the monitoring points are shown in Fig. 1b. There were five monitoring locations in total: high-frequency hydraulic data (flow rate at 1S/s and pressure at 200S/s) were collected at the three sites labelled F & P1,2,3 (Fig. 1c) and the water quality data were collected at the two sites labelled WQ1,2 (Fig. 1d). The F & P sites included an InfraSenseLP RTU interfaced with a pressure transducer and an insertion type electromagnetic flow meter (ABB Aquaprobe V2). The WQ sites included a multiparameter in-pipe water quality probe and an InfraSenseLP

Fig. 3 e Dynamic response of sensor probe for chlorine and ORP in a) tap and b) deionised water (free chlorine is measured by two electrochemical sensors, chlorine 1 and chlorine 2, one of which also measures the total chlorine; free and total chlorine grab sample data are also plotted e ‘palintest’).

240

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 3 5 e2 4 6

RTU measuring pressure continuously at a sampling rate of 200 S/s. The high speed pressure data at all five locations was time synchronized. The second field trial, Case Study 2, was carried out in a water distribution system and included three district meter areas (DMAs). The trial took three months. The water distribution system had a complex pipe topology with a mixture of cast iron and asbestos-cement pipes and it used two sources of supply. The water quality was monitored at three sites. Hydraulic data were also acquired from the district area meters and by flow meters at the outlet of the treatment

plants. These were setup at logging intervals of 1min for the duration of the trial.

3.

Results and data analysis

3.1.

Controlled laboratory tests

The results of the tests undertaken to evaluate the performance of the chemical and optical sensors are summarised as following (see also Table 1).

Fig. 4 e Water quality data obtained by the sensor probe at site WQ1 during a one month period after the installation of the probe (grab sample data are also plotted).

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 3 5 e2 4 6

3.1.1.

pH and conductivity

The performance of the pH and conductivity sensors (Fig. 1a) is illustrated in Fig. 2a and b respectively. The sensors responded instantaneously to step changes with zero noise and no hysteresis was observed. The accuracy of the pH sensor was 0.4 pH units while the sensitivity was 0.2 within the expected pH range in potable water. The accuracy and sensitivity of the conductivity sensor were 10% and 15 mS/cm respectively (Table 1). The pH and conductivity sensors showed good reproducibility and repeatability and the estimated R & R values were 3% for pH and 4% for the conductivity sensor. The uncertainty in the absolute value exceeded the range provided by the manufacturer.

3.1.2.

Colour and turbidity

The sensor probe uses two optical sensors to measure colour and turbidity (Fig. 1a). The ‘apparent’ colour (turbidity effects are not eliminated) is determined by measuring the light absorbed by the sample. Turbidity is determined by measuring the infrared light scattered at a 90 angle by the water sample (ISO 7027). These tests were carried out within a light-tight enclosure to reduce ambient light interference. The surface of the optical sensors required a settling time of one week before calibration so that the optics could become uniformly wet. Fig. 2c and d show a typical dynamic response of the colour and turbidity sensors. The corresponding reference measurements from a spectrophotometer and a turbidity meter are also plotted. The specifications obtained for the colour and turbidity sensors are shown in Table 1. The accuracy of the sensors depends on their calibration and condition of the optics, since interference by deposits adversely affect their performance. The sensors showed good dynamic response and were successful in capturing abrupt changes in colour and turbidity of 4oHazen and 1.5NTU, respectively. The estimated R & R variability of the sensors were 2 for the colour and 1 for the turbidity sensors. However, the absolute values of colour and turbidity varied randomly from the reference data with the presence of measurement noise within 4 oHazen and 1 NTU. Since the colour and turbidity of clean tap water is around 5oHazen and 2 NTU, the noise should not present a problem detecting rapid changes and values that cause discolouration complaints (e.g. 15oHazen and 5NTU, Fig. 2c and d).

3.1.3.

241

The free chlorine electrochemical sensors measure only HOCl, and electrochemical buffering is used to convert ClO to HOCl before sensing in a process known as “Redox cycling”. The total chlorine sensor measures the sum of free chlorine and monochloramine, which are the chlorine species with useful disinfection effects. Chlorinated organic substances, disinfection-by products and other chemicals with no disinfection effect are not measured by the sensor, but these compounds are measured by the DPD method. In addition, constituents in the tap water may include different chlorinespecies and organic and inorganic substances which can affect the test procedures for evaluating the sensitivity and response of the chlorine sensors under reproducible conditions. As a result, the direct comparison of the total chlorine values acquired by the DPD method and the electrochemical sensors might differ depending on the nature of the test water. Fig. 3a presents the dynamic response of the free (free chlorine 1 and free chlorine 2) and total chlorine sensors of one of the probes, when the chlorine dose was added every 10 min to the tap water. A similar response was observed by the second probe. The ORP values, which are a strong indicator of the reactions occurring in the solution, are also plotted. No chlorine was captured by the sensors before the chlorination breakpoint was reached as the added chlorine was consumed in converting monochloramines to dichloramines, which are not measured by the total chlorine electrochemical sensor. The dichloramines are measured by the DPD method and this causes the discrepancy between the electrochemical sensors and the DPD method data shown in Fig. 3a. Once the chlorination breakpoint was reached, the sensor probe captured changes in true residual chlorine. Chloramines are slow reacting so no increase in ORP is observed when adding disinfectant is used to produce chloramines. After the breakpoint is reached, free chlorine is formed and larger step increases in ORP are observed. The chlorine sensors responded rapidly, and their sensitivity was found to be 0.2 mg/L, with a signal noise of 0.1 mg/L. In order to create more stable and reproducible test conditions to control the increase of free and total chlorine, tests were also carried out using deionised water, with salt added in order to increase the conductivity of deionised water

Free and total chlorine

The probe has two sensors to measure free chlorine; one of which can be configured to measure total chlorine (Fig. 1a). To evaluate the performance of the chlorine electrochemical sensors, 3 mL doses of diluted commercial sodium hypochlorite (16% free chlorine, diluted to 1:500, 0.96 mg of chlorine) were added to the tap water and the response of the free and total chlorine sensors were compared to DPD reference measurements. The tap water contained monochlormanine as a residual disinfectant. The ion exchange process is accelerated by placing the probe’s head in highly concentrated chlorine solution (16% diluted 1:4) for approximately 30s, before installing the probe. Once in the water the chlorine sensors need to settle for a period of two weeks, and are then calibrated in situ versus a reference chlorine value taken from a grab sample.

Fig. 5 e Variation of flow rate (unsteady state flow conditions) associated with a discoloration event captured by the high-frequency hydraulic logger at site F & P 1.

242

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 3 5 e2 4 6

to 300 mS/cm for the operation of the electrochemical sensors. When chlorine is added it remains as free chlorine and it equals the total chlorine. In this case, the value measured by the DPD is directly comparable with the values measured by the electrochemical sensors. Fig. 3b presents the dynamic response of the free and total chlorine sensors when the chlorine dose was added every 10 min to the deionised water. The good dynamic response and repeatability under controlled laboratory conditions led to the conclusion that the probe can successfully capture gradual and sudden changes in the chlorine concentrations and could be used for the field

experimental programme in combination with frequent grab sampling to compensate for the low accuracy.

3.2. Operational water transmission system e Case Study 1 The water quality data acquired in Case Study 1 (a water transmission main, see Fig. 1) over a period of one month are presented in Fig. 4. Grab samples data are also plotted for comparison. The data during the first two weeks include the

Fig. 6 e Water quality data obtained by the sensor probes at the upstream (WQ1) and downstream (WQ2) sites during the discolouration event (dashed line in colour plots indicates the quality threshold for customer complaints).

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 3 5 e2 4 6

Fig. 7 e Free chlorine at a district meter in the DMA (district metered area) and the corresponding flow rate data at the treatment plant supplying the DMA (2 h data,).

settling period when the surface of the optical sensors hydrate and the electrochemical processes stabilise.

3.2.1.

Steady flow conditions

The following observations summarise the general performance of the sensor probe under steady-state and gradually varying hydraulic conditions: 1) The sensor probe stabilised within two weeks and successfully captured patterns and daily variations in the monitored parameters.

2) The temperature measurements were stable (Fig. 4a) with diurnal fluctuations. Temperature is a critical parameter to measure correctly because the measurements of the other sensors are temperature dependent. For example, conductivity varies as much as 3% per 1  C. 3) The total chlorine concentration varied with the flow rate as the chlorine is a function of the residence time. The dependence of chlorine kinetics on the average flow velocity has studied extensively (Ozdemir and Ger, 1999; Clark and Haught, 2005). 4) The pH exhibited diurnal fluctuations which correlate with changes in temperature. This is partially attributed to the pH/temperature compensation effect (Nernst effect). pH changes which were not temperature related also occurred in the system; the pH and temperature measurements were not always in phase. 5) The conductivity measurements also exhibited diurnal fluctuations correlated with the temperature data. The dependence of ion mobility, and conductivity, on the temperature varies with the chemistry of the water (e.g. concentration of KCl, NaCl, CaCO3 etc). The detected conductivity variations are small (10 mS/cm) and within the sensitivity range for the conductivity sensor (w15 mS/ cm, based on the laboratory experiments). 6) There was a relatively rapid drift in the measurements which had to be compensated with frequent site visits and grab samples. The acquired data showed that the pH drifted at a rate of 1 pH unit/10 days (Fig. 4b), and the drift of the chlorine sensors was w0.4 mg/L/10 days (Fig. 4g and h). The observed fouling of the sensors after a period of four weeks is shown in Fig. 9a, even though the average flow velocity was high (0.4e0.8 m/s). The level of fouling was system and site specific and this seems to have a significant effect on the drift in the measured parameters. The integrated stirrer which provides additional mixing around the sensor head to reduce the risk of fouling had no impact on the sensors’ drift. 7) Regular daily peaks are observed for colour and turbidity, which are correlated with the daily flow peaks. This may be due to the impact of the flow velocity on the re-suspension of loose particles and sediments in the upstream pipework. Similar turbidity data have been previously reported by KIWA (Amsterdam Water Supply) for the network of the Water Supply Company South-Kennemerland (Van den Hoven and Vreeburg, 1992).

3.2.2.

Fig. 8 e Variation of turbidity (a), and colour (b), during approximately 30 h at three sites in the DMA (site Q1 and Q2 are supplied from works A, and site Q3 from works B).

243

Unsteady-state flow conditions

Unsteady-state hydraulic events can impact the disinfection processes and compromise the critical disinfectant residual due to the increase in the shear stress. This results in the resuspension of sediments, scouring of biofilms and tubercles from the pipe and increased mixing which impact the reaction and disinfectant decay rates. As an illustration, Fig. 5 presents flow data at site F & P1 after the pumps at the nearby pumping station were switched off and then switched on again 24 min later. During these events, the flow reduced abruptly from approximately 650 l/s to 170 l/s at 2:50 PM, and then it rapidly increased to 800 l/s at 3:14 PM before stabilizing at its steadystate value of 650 l/s. The water quality data obtained from the

244

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 3 5 e2 4 6

Fig. 9 e Deterioration of the water quality sensors: a) bio-fouling; b) limescale built-up; c) free and total chlorine data corresponding to the deteriorating performance.

sensor probes at sites WQ1 and WQ2 are shown in Fig. 6. While the water quality probe had a low level of accuracy which had to be compensated with frequent calibration, the acquired data indicated that the water quality significantly deteriorated during the unsteady-state flow conditions. The most critical changes were the significant drop in disinfectant residual and the increase in discolouration above the threshold for customer complaints (see Fig. 6). The acquired data during the unsteady-state flow events are summarized as following:  A rapid change in colour was observed both at WQ1 & WQ2 following the sudden flow increase. The change in colour at WQ2 was in the range of 20oHazen which is above the threshold of 10oHazen associated with discolouration complaints. The change in colour at WQ1 was significantly less than WQ2 which was most likely due to the short length of pipe between the pumping station and WQ1 resulting in a lower concentration of sediments and biofilm. The rapid change in the flow creates high shear stresses which cause particle mobilisation from sediments and biofilms along the pipe. These particles are then advected downstream leading to an increase in colour. No change in turbidity was observed at either site due to faulty turbidity sensors in both probes.  The abrupt increase in flow (3:14 PM) was followed by a significant decay in the disinfectant (monochloramine) residual below a threshold of 0.5 mg/L which is commonly used to prevent bacterial regrowth in water supply systems. The observed change in the disinfectant residual is attributed to the mobilisation and entrainment of particles and biofilms which affect the bulk and wall reactions during the unsteady state hydraulic conditions. While there is uncertainty in the absolute values for the concentration of monochloramines, the good repeatability of the sensors provides a clear indication of the rate and magnitude of change in the observed disinfectant residual.  The colour increase was accompanied by small changes in the pH and conductivity. This is most likely caused by the mobilised particulate material which changes the ionic concentration within the pipe and as a result affects the pH and conductivity. These changes were independent of the

water temperature, which was constant during the event; therefore, the release of wall material is the most likely cause for these changes. The magnitude of the changes in the measured water quality parameters and their correlation in time clearly demonstrate the deleterious effect of the unsteady state hydraulic conditions. The distinct and relatively sudden changes in the measured water quality parameters were also in contrast with the more gradual and smooth behaviour observed during steady state flow conditions.

3.3.

Operational water distribution system e Case Study 2

A second set of trials using the same instrumentation (Intellisonde & InfraSense RTU) was carried out in a water distribution system (Case Study 2, see Section 2.2). This section presents some of the data that have been acquired to demonstrate the performance of the sensor probe and the observed water quality changes. The results of this study are as following:  The free chlorine data for a 24 h period measured at one of the three WQ monitoring stations (SR 1S/5min), and corresponding flow data measured at the nearby pumping station/reservoir (SR 1S/1min), are shown in Fig. 7. The distribution system has two sources of water; one of them is the pumping station/reservoir. Significant and abrupt changes in the free chlorine were observed which were affected by the operation of the booster pump. This illustrates that mixing water from two or more sources might have a detrimental effect on water quality when the sources have a significantly different composition.  Fig. 8 shows one of the discolouration events which were captured and the acquired colour and turbidity data at the three monitoring sites. A significant increase in the turbidity and colour values was measured by the probes at sites Q1 and Q2 which are supplied from the same pumping station. No change was detected at site Q3 which is supplied by a different source. In this case, the discolouration event followed a sudden change in flow at the pumping station

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 3 5 e2 4 6

due to a power failure. It is most likely that the resulting unsteady state flow caused particles and biofilms to mobilize as propagated across the DMAs. Further work is being carried out to model the dynamic hydraulic behaviour for this study area.  The trials in Case Study 2 also exhibited problems with low accuracy and constant drift for the water quality sensors (Fig. 8b). Frequent site visits and calibration based on grab samples were undertaken to correct these problems. For example, the colour sensors drifted significantly with time with an increasing apparent absorbance caused by a gradual build-up of deposits on the optics. This was less evident with the turbidity sensors which had a minor drift in their measurements. Similar to Case Study 1, the observed drift varied with the site location, flow characteristics and pipe material. While the frequent recalibration was justified for an experimental study, it will significantly increase the O & M costs for a permanent installation.

4.

Discussion and conclusions

The continuous monitoring of water quality within water transmission and distribution systems has been extremely limited and rare in practice due to the high capital and O & M costs and the low level of reliability and accuracy of the generated data. Grab sampling remains the main method for assessing the deterioration of water quality between a treatment plant and the end users. The experimental research presented in this paper aimed to improve our understanding of the changes in water quality due to the inherent unsteadystate hydraulic conditions in operational systems. The acquired hydraulic and water quality data show that unsteady-state hydraulic conditions can significantly affect the deterioration of important water quality parameters. A key component of this research was to integrate, develop and evaluate novel technologies for the acquisition of hydraulic and water quality data. These technologies provide a unique insight in the dynamic hydraulic and water quality processes occurring in large scale water transmission and distribution systems which to the best of our knowledge has not been previously demonstrated. Existing technologies for continuous water quality sensing in distribution networks include reagent and membrane based sensors adapted from technologies used in water treatment plants. Complex and costly installation setup, the reliability of the sensors and the accuracy of the acquired data have significantly hindered their application for proactive water quality management of water supply systems. These limitations encourage the development of novel sensor technologies. The use of reagent-free electrochemical sensors for in-pipe water quality monitoring is one of these developments. A detailed evaluation of electrochemical and optical sensors under controlled laboratory conditions and in operational systems, which this paper presents, is a decisive prerequisite for their use and integration with network models for near real-time operational control and contamination risk assessment. The multi-parameter sensor probe which has been used for this experimental study (Intellisonde)

245

offers an improved application of electrochemical and optical sensors for in-pipe, reagent and membrane-free water quality monitoring. The Intellisonde was also the only commercially available instrument at the time of the study which measures nine important water quality parameters via a direct pipe insertion. The analysis of the data from the controlled laboratory experiments and the trials in two operational systems (a transmission main and a distribution network) led to the following conclusions regarding the performance of the multiparameter water quality probe:  The detection limits for the sensors were within the range of relevant EPA and EU standards.  While the sensors displayed a good dynamic response and repeatability for capturing trends and sudden changes, they had continuous problems with low accuracy and constant but non-deterministic drift.  The performance of the sensors was consistent based on the results of the laboratory and field trials. The field trials were designed to be carried out in two different operational systems with regards to hydraulic conditions (e.g. flow velocities), pipe material and diameter, residual disinfectant (free chlorine and monochloramine) and water quality characteristics. The acquisition of reliable data with regards to sudden changes, daily fluctuations and long-term trends required frequent and complex calibration. The sensors performance was significantly affected by bio-fouling and deposits on the electrochemical and optical sensors (Table 1, Fig. 9a). The sensitivity and repeatability also deteriorated within a relatively short period of time (e.g. three weeks). As an illustration, Fig. 9b shows excessive deposits of calcium carbonate on the chlorine sensors over a period of two weeks which affect the measured free and total chlorine and DO measurements (Fig. 9c).  The low accuracy and uncertainties in the sensor data present significant challenges for the analysis and interpretation of the acquired data and its integration with network models for operational control and mitigation of the consequences of water quality related failures.  A multi-parameter water quality probe for in-pipe reagent free and membrane free water quality monitoring is an extremely attractive and important technology for the optimal management of large scale water supply systems. However, further and fundamental developments are necessary before water utilities can confidently use and deploy these sensors for large scale water quality monitoring.  Notwithstanding these limitations with the sensors, unique data sets were acquired from transmission and distribution systems to demonstrate the deleterious effect of unsteady state flow conditions on various water quality parameters. This experimental research required the development of instruments for the continuous and long-term monitoring of the dynamic hydraulic conditions (InfraSense RTU) and their integration with in-pipe electrochemical sensors for monitoring important water quality parameters.  The data acquired provided compelling evidence that unsteady state flow conditions can significantly affect the water quality in water supply system and have a major impact on changes in the disinfectant decay, turbidity, colour, chlorine, and pH. The presented experimental

246

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 3 5 e2 4 6

research has given a unique insight into the water quality processes occurring in water supply systems, the significance of shear stress, flow residence time and sources of discolouration. The wider application is that recognising the impact of the unsteady state hydraulic conditions on the disinfection kinetics can improve the detection of contamination events by reducing false positive alarms. While the recent EPA study (Panguluri et al., 2009) demonstrated that the measurement of chlorine and TOC can be used successfully as trigger parameters for the detection of contamination events, the results presented in this paper show that the EPA approach can be significantly enhanced by continuously monitoring the dynamic hydraulic conditions and recognizing their impact on sudden changes in the measured parameters.  The presented integration of sensor technologies (both hydraulic and water quality) provide unique opportunities to study and better understand both the dynamic hydraulic conditions and water quality changes in operational systems with complex pipe topologies, control elements and loading conditions. This can ultimately lead to optimal control of the booster chlorination and the development of computationally efficient hydraulic and water quality models which successfully simulate both the steady and unsteady-state hydraulic and water quality behaviour.

Acknowledgements The authors wish to acknowledge the financial support of the Engineering and Physical Science Research Council (EPSRC).

references

Aisopou, A., Stoianov, I., Graham, N., 2010. Modelling Chlorine Transport under Unsteady-state Hydraulic Conditions. In: Proceeding of 12th Water Distribution System Analysis Sept. 12e15, 2010, Tucson, Arizona. APHA, 2005. Standard Methods for the Examination of Water and Wastewater. American Public Health Association, American Water Works Association, Water Environment Federation, Baltimore, Maryland, USA. AWWARF, 2006. Disinfectant Decay and Corrosion. AWWA research foundation, American Water Works Association, Denver, USA. BS 1427, 2009. Guide to On-site Test Methods for the Analysis of Waters. BSI, UK. BS/EN ISO 15839, 2006. Water Quality. On-line Sensors/Analysing Equipment for Water. Specifications and Performance Tests. BSI, UK.

Calle, G.R., Vargas, I.T., Alsina, M.A., Pasten, P.A., Pizarro, G.E., 2007. Enhanced copper release from pipes by alternating stagnation and flow events. Environmental Science & Technology 41 (21), 7430e7436. Clark, M.R., Haught, R.C., 2005. Characterizing pipe wall demand: Implications for water quality modelling. Journal of Water Resources Planning and Management 131 (3), 208e217. Intellitect, 2008. Intellisonde: Customer Training Manual. Intellitect Water, Romsey, UK. ISO 5725-1, 1994. Accuracy (Trueness and Precision) of Measurement Methods and Results e Part 1: General Principles and Definitions. Karney, B., Brunone, B., 1999. In: Savic, D., Walters, G.A. (Eds.), Water Hammer in Pipe Network: Two Case Studies. Water Industry Systems: Modelling and Optimization Applications, vol. 1. Research Studies Press Ltd, Baldock, UK, pp. 363e376. Lee, Y., Sinha, E., Piao, H., Stillman, J., Hartman, D., Bush, C., 2010. “Chlorine Wall Decay Coefficient to Calibrate the GCWW AllPipes Distribution System Model.” World Environmental and Water Resources Congress, Challenges of Change. ASCE. Maier, S.H., Powell, R.S., Woodward, C.A., 2000. Calibration and comparison of chlorine decay models for a test water distribution system. Water Resources Research 34 (8), 2301e2309. Montgomery, D.C., Runger, G.C., 1993. Gauge capability and designed experiments. Part 1: basic methods. Quality Engineering 6 (1), 115e135. Ozdemir, O.N., Ger, A.M., 1999. Unsteady 2-D chlorine transport in water supply pipes. Water Research 33 (17), 3637e3645. Panguluri, S., Meiners, G., Hall, J., Szabo, J.G., 2009. Distribution System Water Quality Monitoring: Sensor Technology Evaluation Methodology and Results. U.S. Environmental Protection Agency, Washington, DC. EPA/600/R-09/076, 2009. Stoianov, I., Nachman, L., Whittle, A., Madden, S., Kling, R., 2006. Sensor Networks for Monitoring Water Supply and Sewer Systems: Lessons from Boston. In: Proceedings 8th Water Distribution System Analysis Symposium (WDSA 2006). ASCE, USA 27e30 August. Van den Hoven, T.J.J., Vreeburg, J.H.G., 1992. Distribution system analysis by continuous monitoring and network calculations. Water Supply 20 (1), 117e124. Vreeburg, J. H. G., 2007. Discoloration in drinking water systems: a particular approach. PhD thesis, Technische Universiteit Delft, The Netherlands. Vreeburg, J.H.G., Shchaap, P.G., van Dijk, J.C., 2004. Particles in the drinking water system: from source to discolouration. Water Science and Technology: Water Supply 4 (5), 431e438. WHO, 2006. Guidelines for Drinking-water Quality. In: Recommendations, third ed., vol. 1. World Health Organisation, Geneva. Woodward, C.A., Ta, C.T., Holt, D.M., Colbourne, J.S., 1995. Relationship between Observed Monochloramine Decay Rates and Other Physical and Chemical Parameters. In: Proceedings of the Water Quality Technology Conference. American Water Works Association, New Orleans. Yang, Y.J., Haugh, R., Goodrich, J., 2009. Real-time contaminant detection and classification in a drinking water pipe using conventional water quality sensors: techniques and experimental results. Journal of Environmental Management 90 (8), 2494e2506.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 4 7 e2 5 7

Available online at www.sciencedirect.com

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

High diversity and differential persistence of fecal Bacteroidales population spiked into freshwater microcosm Zhanbei Liang a,1, Zhenli He a,*, Xuxia Zhou a,2, Charles A. Powell a, Yuangen Yang a, Michael G. Roberts b, Peter J. Stoffella a a b

Indian River Research and Education Center, Institute of Food and Agricultural Sciences, University of Florida, Fort Pierce, FL 34945, USA US EPA, Natl Risk Management Res Lab, 919 Kerr Res Dr, Ada, OK 74820, USA

article info

abstract

Article history:

Bacteroidales markers are promising indicators of fecal pollution and are now widely used in

Received 20 June 2011

microbial source tracking (MST) studies. However, a thorough understanding of the

Received in revised form

persistence of Bacteroidales population after being released into environmental waters is

21 October 2011

lacking. We investigated the persistence of two host specific markers (HF183 and CF193)

Accepted 1 November 2011

and temporal change of Bacteroidales population over 14 days in freshwater microcosms

Available online 9 November 2011

seeded with human or bovine feces. The concentrations of HF183/CF193 and Escherichia coli were determined using quantitative polymerase chain reaction (qPCR) and standard

Keywords:

cultivation method, respectively. Shifts in the Bacteroidales population structure were fin-

Bacteroidales population

gerprinted using PCR-denaturing gradient gel electrophoresis (DGGE) and subsequent

Persistence

sequencing analysis targeting both 16S rDNA and rRNA-transcribed cDNA. Both HF183 and

Diversity

CF193 decayed significantly faster than E. coli but the decay curves fit poorly with first-order

Fecal pollution

model. High diversity of Bacteroidales population was observed for both microcosms, and

Freshwater

persistence of different species in the population varied. Sequence analysis indicated that

Denaturing gradient gel

most of the bovine Bacteroidales populations in our study are unexplored. DGGE and decay

electrophoresis

curve indicated that RNA decayed faster than DNA, further supporting the use of rRNA as indicator of metabolically active Bacteroidales population. Evaluations with more realistic scenarios are warranted prior to extending the results of this study to real field settings. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Bacteroidales, primarily from the family Prevotellaceae and Bacteroidaceae, are strict anaerobes. Members of the Bacteroidales group have been suggested as an alternative indicator of recent fecal contamination (Kreader, 1995) because of i) their higher abundance in fecal populations than traditional fecal indicator bacteria (FIB) (Eckburg et al., 2005), ii) the low

potential to grow outside of the host due to their anaerobic metabolism (Kreader, 1995, 1998), and iii) the correlation with the presence of fecal pathogens (Walters et al., 2007). In addition, certain fecal Bacteroidales lineages are host specific (Bernhard and Field, 2000), the presence of which enables identification of the source of fecal pollution. Polymerase chain reaction (PCR) based microbial source tracking (MST) methods targeting host specific bacteroides

* Corresponding author. Tel.: þ1 772 468 3922; fax: þ1 772 468 5668. E-mail addresses: [email protected] (Z. Liang), [email protected] (Z. He), [email protected] (X. Zhou), [email protected] (C.A. Powell), [email protected] (Y. Yang), [email protected] (M.G. Roberts), [email protected] (P.J. Stoffella). 1 Present address: National Research Council at US EPA, Natl Risk Management Res Lab, 919 Kerr Res Dr, Ada, OK 74820, USA. 2 Present address: College of Biological and Environ. Eng., Zhejiang University of Technology, Hangzhou 310014, China. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.11.004

248

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 4 7 e2 5 7

gene markers have been developed and increasingly applied to evaluate the source of fecal pollution in environmental water in different geographical locations (Fremaux et al., 2009; Gawler et al., 2007; Gourmelon et al., 2007). Quantitative PCR methods have also been utilized to assess the abundance of host specific markers in natural waters (Kildare et al., 2007; Layton et al., 2006). Before qPCR data can be reliably used to quantify the contribution of fecal contamination from each specific source, two properties concerning Bacteroidales markers need to be considered (Schulz and Childers, 2011): namely the marker distribution and abundance within different host species, and the relative persistence/survival of markers and/or marker cells in the non-intestinal environments under the interaction of various environmental factors. Understanding the survival of Bacteroidales spp are also important in the development of proper MST models for health risk predication (Balleste and Blanch, 2010). Previous studies indicated that certain fecal Bacteroides spp cells could survive from only a few hours to a few days in environmental water (Bae and Wuertz, 2009; Balleste and Blanch, 2010; Kreader, 1998), while the host specific Bacteroides DNA marker could persist up to weeks (Bae and Wuertz, 2009; Kreader, 1998; Walters and Field, 2006, 2009). Environmental factors, including temperature, the presence of predators, salinity, and sunlight, have been reported to affect the survival of Bacteroidales species and persistence of Bacteroidales DNA (Bell et al., 2009; Kreader, 1998; Okabe and Shimazu, 2007; Savichtcheva et al., 2005). Bacteroidales are a deeply divergent and diverse group of microorganisms based on 16S rRNA phylogeny. A high diversity of Bacteroidales from animal feces was also reported (Jeter et al., 2009). In a realistic scenario for MST study, the clades that the host specific Bacteroidales markers detect may include a variety of species, genera, and families from a wide range of sources (individuals), which may not necessarily share the same survival profiles in the environment (Schulz and Childers, 2011). In order to accurately attribute the fecal pollution to different sources, persistence of the whole Bacteroidales population should be considered. But the few studies investigating the persistence of Bacteroidales spp. were confined to only one to limited species (Balleste and Blanch, 2010; Kreader, 1998). Divergence in the response of different Bacteroidales species to environmental stresses has been detected (Balleste and Blanch, 2010; Wilkins et al., 1978); it is therefore improper to infer the environmental persistence of the Bacteroidales population in feces from the pattern of only a few species. Thus, there is a gap of information on the ecology and environmental behavior of fecal Bacteroidales populations. Using clone library analysis, Schulz and Childers (Schulz and Childers, 2011) indicated that diverse lineages of Bacteroidales survived similarly, but the use of DNA in their study failed to provide information on the survival of live cells under those conditions. Our objective in this study was to determine the persistence pattern of a fecal Bacteroidales population once they are released into environmental waters. Freshwater microcosms inoculated with human or bovine feces were incubated for a period of 14 days; population compositions of Bacteroidales spp. in the microcosm were fingerprinted with both a DNA and RNA based DGGE approach targeting 16S rRNA sequence.

The persistence of the fecal Bacteroidales population was compared with persistence of two host specific Bacteroidales markers. The persistence of the standard fecal indicator, Escherichia coli, was also monitored during the entire incubation, using standard culture methods, for comparison. This may be the first study to evaluate the persistence pattern of metabolically active fecal Bacteroidales population in the extraintestinal environment.

2.

Materials and methods

2.1.

Microcosm setup and sampling

Fresh human fecal samples were collected within 24 h of use from 4 healthy adult volunteers in different families with a sterile utensil and stored in a sterile 50 ml tube. On the day of experimental setup fresh cattle fecal samples were collected immediately after excretion from 4 bovines in 2 cattle farms. Fecal samples were transported in the dark on ice to the laboratory immediately after the collection. Two types of microcosm were constructed: the human microcosms in which human fecal samples were spiked into 28 L water, and the bovine microcosm in which cattle fecal samples were spiked. For each host, 40 g (wet weight) of fecal sample (10 g from each individual) was homogenized by mixing and suspended into 300 ml sterile PBS solution. Large fecal particles were removed and the fecal slurry was stored on ice until use. Fresh water was collected in autoclaved containers from a canal near the Indian River Research and Education Center in Fort Pierce, FL and stored overnight in the dark at room temperature before the start of the experiments. The initial turbidity of the fresh river water was 31 nephelometric turbidity units (NTU). To check for the presence of Bacteroidales spp. prior to the microcosm experiment, 500 ml of freshwater samples were concentrated by membrane filtration and subjected to direct DNA extraction in triplicates using PowerSoil DNA extraction kit as described in Section 2.3. The presence/ absence of host specific Bacteroidales markers HF183 and CF193 in this source water was checked by qPCR as described in Section 2.5. The microcosm was designed to simulate the environmental water conditions fecal microbiota may encounter after being released from the host, such as radiation, fluctuating ambient temperature, and the possible presence of bacterivorous predators. Microcosms were set up in a roofed greenhouse without sides. The polycarbonate roofing blocks most of the UV but allows 90% visible light transmission (http://www.usgr.com/greenhouse-coverings/polycarbonate. php). Sterile glass aquariums with dimensions of 25.4 by 50.8 by 25.4 cm (W/L/H) were used as containers with an air stone installed at the bottom of each aquarium for aeration and constant mixing of the water column. Human fecal Bacteroidales microcosms were established in triplicates by inoculating 100 ml of the human fecal slurry into the aquarium filled with 28 L of freshwater. The bovine microcosms were conducted using the same method. The final concentration of feces in the microcosm was 46 mg wet weight per 100 ml water. The outside of the side wall and bottom of the

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 4 7 e2 5 7

aquarium was wrapped with opaque material to ensure light penetrated only from the top opening of the microcosm. FEPTeflon film (1/20 mm thick) which allows 96% of natural light penetration was used to cover the opening (top) of the microcosms to reduce water evaporation during the incubation. Slight evaporation could still be observed on each sampling event. We calculated daily evaporation loss of water using the following equation: v ¼ [(L  W  DH)S]/14, where v is the daily evaporation loss of water; L and W are the length and width of the microcosm, respectively; DH is the difference between the initial water level and the final water level, and S is the total volume of water sampled during the 14 days incubation. The density of targets in the microcosms was corrected accordingly to account for the loss of evaporation. Sampling was conducted at day 0 (1 h after the inoculation), 2, 4, 6, 8, 10, 12 and 14 for marker concentration analysis, DGGE profiling of Bacteroidales population and E. coli enumeration. Sampling was conducted at approximately the same time on each sampling day. For each microcosm, around 150 ml of water was collected in a sterile container for nucleic acid extraction and subsequent DNA/RNA analysis. Water (10 to 100 ml) was collected for E. coli enumeration. Water samples were transferred back to the laboratory on ice and processed immediately.

2.2.

E. coli enumeration

E. coli densities were determined, in triplicate, using standard membrane filtration methods (USEPA method 1623). A preliminary experiment was conducted to determine the initial concentration of E. coli in the microcosm. Dilutions ranging from one hundreds to one tenth were made for accurate enumeration. The concentration of E. coli was expressed as CFU per 100 ml sample.

2.3.

2.4.

249

Reverse transcription

Reverse transcription was conducted with an Eppendorf Mastercycler (Eppendorf) using High Capacity RNA-to-cDNA master mix (Applied Biosystems, Foster City, CA) according to the manufacturer’s instructions. Two ml of RNA was mixed with 1  RNA-to-cDNA master mix in a 20 ml reaction volume. The reactions were performed by incubating the sample at 25  C for 5 min, 42  C for 30 min, and 85  C for 5 min. Contamination of genomic DNA was tested for using master mix without reverse transcriptase ()RT included in the kit. cDNAs transcribed were stored at 80  C. cDNAs were subjected to qPCR and PCR-DGGE analysis.

2.5. Quantification of host specific Bacteroidales marker DNA and cDNA Human specific primer HF183 and bovine specific primer CF193 (Bernhard and Field, 2000) were paired with 265R (Seurinck et al., 2005) for human and bovine specific marker quantification, respectively. Quantitative PCR was performed with a CFX96 Real-Time PCR Detection System (Bio-rad, Hercules, CA) in a 20 ml reaction volume containing 10 ml iQ SYBR Green Supermix (Bio-rad, Hercules, CA), 0.2 mM of each forward and reverse primers and 2 ml template DNA/cDNA. A universal thermal cycling protocol (initial denaturation step of 94  C for 2 min, followed by 40 cycles of 94  C for 15 s and 60  C for 60 s) was used for the quantification of both markers. A final melt curve analysis was performed to determine the presence or absence of nonspecific amplification products. All samples were run in triplicate and a negative control was included for each analysis. Standard curves were generated using a serial dilution of purified plasmid DNA containing each host specific sequence.

Nucleic acid extraction 2.6.

Canal fresh water: Five hundred ml of source water was concentrated by membrane filtration using nitrocellulose filters (0.45 mm pore size, 47 mm diameter). The filter was aseptically folded, cut and transferred into the PowerBead tube included in the PowerSoil DNA kits (MO BIO Laboratories, Inc., Carlsbad, CA). DNA from the microorganisms on the filters was extracted using PowerSoil DNA kits following the manufacturer’s instruction with some modification as described previously (Liang et al., 2008). Eluted DNA was stored at 80  C. Microcosm sample: A total of 120 ml water sample was centrifuged at 15000 rpm for 20 min at 4  C using an Eppendorf 5430R microcentrifuge in 50 ml Nalgene Oak Ridge polypropylene centrifuge tubes (Fisher Scientific). DNA/RNA from each sample was immediately extracted from the pellet using AllPrep RNA/DNA Mini kit (Qiagen, Valencia, CA) following the manufacturer’s instruction. Extraction control using double stilled water was included each day the extraction was conducted. DNA eluted in 50 ml buffer EB was stored at 80  C. RNA eluted in 30 ml RNase free water were immediately subjected to DNase treatment using TURBO DNA-free Kits (Ambion, Foster City, CA) following the manufacturer’s suggestions to remove co-eluted trace DNA. RNA supernatants were transferred to RNase free 0.2 ml vial and subjected to reverse transcription.

PCR-DGGE and sequence analysis

DNA extracted and cDNA transcribed were amplified by PCR using primers Bfr-F/Bfr-R-GC (Liu et al., 2003) and then subjected to DGGE analysis. This primer pair was originally designed for the identification of the group Bacteroidales fragilis, which included the species Bacteroides caccae, Bacteroides distasonis, Bacteroidales eggerthii, Bacteroidales fragilis, Bacteroidales merdae, Bacteroidales ovatus, Bacteroidales stercoris, Bacteroides thetaiotaomicron, Bacteroides uniformis, and Bacteroidales vulgates (Liu et al., 2003). Since then it has been widely used for the detection of Bacteroidales. spp from human and animal fecal microbiota (Li et al., 2007, 2009; Pang et al., 2005; Yuan et al., 2011). A 40 bp GC-clamp was attached to the 50 end of Bfr-R to facilitate band separation on DGGE. All PCR reactions were conducted in an Eppendorf Mastercycler (Eppendorf). Two mL of DNA extract or 5 ml cDNA was used as template in a reaction volume of 25 mL containing 1.5U Taq DNA polymerase and 1 PCR buffer (20 mM TriseHCl, 50 mM KCl, pH 8.4) with 0.3 mg ml1 BSA. The PCR protocol consisted of denaturation at 95  C for 2 min followed by 35 cycles of 95  C for 30 s, amplification at 54  C for 30 s and elongation at 72  C for 30 s and a final elongation step at 72  C for 30 min. Amplification products were checked for size (270 bp) and

250

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 4 7 e2 5 7

yield by standard 2% (w/v) agarose-0.5 Tris-borate-EDTA (TBE) gel electrophoresis with GelStar staining. Twenty mL of PCR products were subjected to DGGE in a SE 600 Ruby Standard Dual Cooled Vertical Unit (GE healthcare, Piscataway, NJ). Denaturant gradient was pre-tested to sufficiently separate bands on the gel. The 0.75 mm thick gel containing 8% (w/v) polyacrylamide (37.5:1 acrylamide/bisacrylamide) and a 30e60% denaturant gradient (100% denaturant is defined as 7 M urea and 40% (v/v) formamide) were electrophoresed for 16 h 30 min at 80 V and 60  C in 1  TAE buffer. Gels were stained with GelStar nucleic acid stain and gel images were captured using the Gel Doc XR imaging system (Bio-rad, Hercules, CA). Bands of interest were excised from the DGGE gel and stored overnight in 100 mL of water at 4  C to allow the DNA to passively diffuse out of the gel strips before re-amplification. Ten microliter of eluted DNA was then used as template for PCR amplification with primer pair BfrF/Bfr-R-GC as described above, and the products were subjected to DGGE to check for their migration. Confirmed DNA fragments were re-amplified using primer pair BfrF/Bfr-R and purified from a 2% (w/v) agarose gel using Qiaquick Gel extraction Kit (Qiagen, Valencia, CA) and cloned into pCR2.1-TOPO vector (Invitrogen, Carlsbad, CA). The ligation products were chemically transformed into E. coli (strain TOP10) using the TOPO TA Cloning Kit (Invitrogen, Carlsbad, CA) following manufacturer instructions. Recombinant colonies were identified on LB medium containing ampicillin (50 mg ml1) and X-Gal (5-bromo-4-chloro-3indolyl-b-D-galgacto-pyranoside: 0.1 mM). One insert with the right size confirmed with PCR from each clone library was purified using the Wizard Plus Minipreps DNA purification system (Promega, Madison, WI) and sequenced with the BigDye Terminator Cycle Sequencing Kit v 3.1 on the AB 3730xl DNA analyzer (Applied Biosystems, Foster City, CA). Sequences were edited by Chromas (version 2.33), checked for chimera by the RDP CHECK-CHERIMA program, and compared with the NCBI database using BLASTN to obtain estimates on closest phylogenetic affiliates. Sequences obtained in this study were deposited in the European Molecular Biology Laboratory (EMBL) database under accession numbers JF831366-JF831420.

2.7.

Data analysis

To normalize the data, the concentrations of markers and E. coli were transformed to natural logarithm (Ln). Decay rate of each host specific Bacteroidales marker and E. coli were calculated using the following standard exponential growth/ decline equation: k ¼ ½lnðNt Þ  lnðN0 Þ=t, where k ¼ the decay rate with a unit of d1, Nt ¼ the geometric mean concentration of Bacteroides marker or E. coli at time t, N0 ¼ the geometric mean concentration of Bacteroides marker or E. coli at time zero, t ¼ time (days), Time t was determined by the days when the last sampling event occurs (14 day) or when the markers could not be detected by qPCR. Only data points above the assay limit of detection were used in the regression model. Time required for 99% of the initial population to decay (t99 value) was calculated as: t99 ¼ ð2=kÞ. Analysis of variance (ANOVA) and a Turkey’s post-hoc test were performed using SAS software (SAS Institute Inc, 2009) to evaluate the difference among the mean decay values.

DGGE gel images were analyzed using a GelCompar II package (Applied Maths, St-Martens-Latem, Belgium). We defined a band as “dominant” if the intensity of this band comprised of 5% or more of the total intensity in a sample lane. A threshold value of 5% relative to the maximum value in a lane was set when doing band search, band with value below the threshold value was defined “invisible”. The Shannon-weaver index was calculated according to the P following equation: H ¼  ½Pi  lnðPi Þ where Pi is the relative probability of the band in a gel lane as calculated as Pi ¼ ni/N, in which ni is the band intensity of a particular band and N is the sum of the intensities of all the bands in a lane as judged by peak height in the densitometric curves. The richness (R) is a count of the number of bands in a community. Evenness was calculated as: E ¼ H=lnR.

3.

Results

Human specific (HF183) and ruminant specific Bacteroidales markers (CF193) were not detected by qPCR in the freshwater collected from a canal. The average salinity in the microcosms throughout the study was 0.10& and 0.12& for human feces treatment and bovine feces treatment, respectively. Average pH was 7.3 and average dissolved oxygen (DO) concentration was 6.3 mg l1 for both treatments. Daily temperature in this area of Florida, as obtained from Florida climate center, fluctuates between 4 and 26  C with an average of 15  C during the experiment period (Feb, 2010). The average solar irradiation in this area is 4.12 kWh/m2day (http://www.solarpanelsplus. com/solar-calculator), with an average irradiation level of 3.7 kWh/m2day inside the greenhouse.

3.1. Decay dynamic of host specific Bacteroidales markers DNA/RNA and E. coli in the microcosms As shown in Fig. 1A, after remaining relatively unchanged in the first two days of inoculation, the natural log transformed concentration of CF193 DNA decreased linearly from day 2 to day 8 before declining sharply from day 8 to day 10 and become undetected thereafter. The persistence of CF193 RNA largely resembled that of the DNA markers: its log transformed concentration declined only slightly in the first two days, subsequently linearly decreased to day 8, and then was lower than the limit of detection. Persistence profile of the human marker HF183 differed greatly from CF193 (Fig. 1B). Changes in the slope of curve indicated that HF183 decayed in a biphasic mode. From day 0 to day 6 HF183 DNA declined gradually, and the decay tended to become slower from day 6 to day 12. HF183 was still detectable at day 14. Concentration of HF183 RNA declined substantially in the first two days. It then decayed at a lower rate until it dropped below the detection limit after day 10. During the entire sampling period the concentration of E. coli ranged from 16,300 to 500 CFU/100 ml for the human microcosms and from 1080 to 133 CFU/100 ml for the cattle microcosms. E. coli in both microcosms were above the water quality standard through the study. In the human microcosm natural Log transformed concentration of E. coli increased slightly at the first two days before it declined steadily until

251

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 4 7 e2 5 7

A

Table 1 e Decay rates and t99 values from regression lines of inactivation models for each individual genetic markers and fecal indicator bacteria (E. coli) in the freshwater microcosms. Markers CF193 DNA CF193 RNA Bovine E. coli HF183 DNA HF183 RNA Human E. coli

K (d1)

t99 in day

0.73 0.94 0.24 0.73 0.96 0.35

2.75 2.12 8.32 2.72 2.08 5.65

B t99 values for E. coli were 8.32 day in the bovine microcosms and 5.65 day in the human microcosms. The t99 value for both marker DNAs were surprisingly similar: 2.75 day for CF193 and 2.72 day for HF183. As expected, RNAs were less persistent than DNA with a t99 value of 2.12 day for CF193 and 2.08 day for HF183, respectively. However, no significant difference occurred between the k value of the DNA marker and RNA marker ( p ¼ 0.13).

Fig. 1 e Persistence of (A) bovine specific Bacteroidales marker CF193 and (B) human specific Bacteroidales marker HF183 DNA and RNA in microcosms. Solid line: DNA marker, dotted line: RNA marker.

the end of the experiment. In contrast no post-inoculation growth phase was observed for E. coli in the bovine microcosm: it declined exponentially at a much lower rate of 0.24 d1 throughout the incubation. Concentration of E. coli dropped by 3 Lns in the bovine microcosms by the end of the experiment (Fig. 2). To make direct comparison between the persistence of host specific markers DNA/RNA and that of E. coli, in spite of the poor fitting, a first-order decay model was used to describe the decay process of all targets. The mean decay rates over the entire incubation period were calculated (Table 1), which in general were consistent with previous studies (Schulz and Childers, 2011). E. coli persistence tended to be greater than Bacteroidales marker DNA and RNA in the freshwater microcosms, as the decay rates of E. coli were significantly smaller than those of the Bacteroidales markers ( p ¼ 0.023). Average

Fig. 2 e Survival of standard indicator E. coli in microcosms inoculated with bovine (dot line) and human feces (solid line).

3.2. DGGE analysis of the Bacteroidales DNA and RNA in the microcosms Because three replicates sampled at the same time point produced nearly identical DGGE profile, only one replicate was used for analysis. A total of 30 bands of distinct mobility were observed on DGGE profile from bovine microcosms (marked from 1 to 30 in Fig. 3), reflecting a high diversity of Bacteroidales community structure. For the DGGE fingerprint generated from DNA, predominant band profiles did not change significantly between day 0 and day 8 with similar Shannon’s diversity (H) (Fig. 5): only 2 faint bands lost at day 4 (band No 11 and 26), 3 faint bands lost at day 6 (band No 2, 3 and 24) and 3 bands lost at day 8 (band no 21, 25 and 28). H value declined abruptly from day 8 to day 10: the number of bands decreased from 21 at day 8 to 8 at day 10; the relative intensities of the 8 bands were greatly reduced although they remained detectable till the end of the experiment (day 14). RNA DGGE fingerprints at day 0 to day 4 were very similar to those of DNA with H value slightly lower than that of DNA at the same sampling point. At day 6, 16 bands were lost on RNA DGGE profile, as compared to the DNA DGGE pattern, resulting in a significantly lower H value. Band number and intensity declined significantly from day 8 and no band is detectable on day 12. BLAST search in the NCBI database showed that the majority of the sequences obtained from bovine microcosms had closest matches to sequences from uncultured Bacteroidales lineages that were derived from various environments such as bovine manure, human intestine, human skin, gull feces, and chicken feces (26 out of 30 in Table 2), indicating the largely unexplored nature of the bovine Bacteroidales population. The only 4 sequences related to cultured Bacteroidales species were band 13 (98% similarity to Bacteroides dorei), band 16 (95% similarity to B. fragilis), band 20 (96% similarity to B. fragilis) and Band 27 (97% similarity to B. uniformis). Similar to the bovine microcosms, a high diversity of Bacteroidales was detected from the human microcosms, with a total of 25 discrete band profiles (marked from 1 to 25 in

252

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 4 7 e2 5 7

Fig. 5 e Change in Shannon’s diversity (H) with time as calculated from DGGE profiles of Bacteroidales 16S rDNA and rRNA.

Fig. 3 e DGGE profile of Bacteroidale 16S rDNA fragments from DNA (D) and rRNA (R) based community in microcosms inoculated with bovine fecal material. Bands were numbered from top to bottom: 1-30. A indicates band from number 1 to number 10, B denotes bands from number 11 to number 20, C denotes bands number 21e30.

Fig. 4). Clear time dependency in DGGE pattern was observed. The DNA DGGE band profile was relatively unchanged from day 0 to day 8, with only minor reduction in H value at day 8 (Fig. 5). Band number declined from 17 at day 10 to 6 at day 14, accompanied by the lowering of band intensity, indicating fast decay of DNA during this period. The RNA DGGE profile resembled that of DNA in the first two days with a slightly

Fig. 4 e DGGE profile of Bacteroidale 16S rDNA fragments DNA (D) and RNA (R) based community in microcosms inoculated with human fecal material. Bands were number from top to bottom 1e25. A indicates band from number 1 to number 10, B denotes bands from number 11 to number 20, C denotes bands number 21e25.

lower H value, after which the H value experienced a marked reduction. Number of bands dropped from 9 at day 4 to 4 at day 10 and became completely invisible at day 12. Unlike bovine microcosm, BLAST similarity search in the database revealed that a much larger percentage of the sequences (15 out of a total of 25) obtained from human microcosms was affiliated to cultured Bacteroidales genus, including B. thetaiotaomicron, B. fragilis, B. vulgates, B. xylanisolvens, B. caccae, B. uniformis and B. dorei, with similarity levels ranging from 96% to 100%. Two fifth of the sequences were closely related to uncultured Bacteroidales species from different environments, including human, turkey and chicken feces (Table 3).

4.

Discussion

Our primary objective was to investigate the persistence of phylogenetically different Bacteroidales populations with time in a setting that incorporates various environmental stresses that may influence their survival, including the fluctuating ambient temperature, sunlight and the possible presence of bacterivorous predators. This information is not only necessary to proportionate the contribution of fecal pollution from different sources; it is also helpful to bridge the gap in the ecology and environmental behavior of fecal Bacteroidales population. In our study we targeted both rDNA and rRNA to detect Bacteroidales markers and to determine temporal change of Bacteroidales population. The diversities of Bacteroidales communities from human and bovine microcosm declined as time elapsed. Decay of host specific markers DNA was much slower than the decay of marker cells that were metabolically active as represented by rRNA. Both targets decayed faster than the standard fecal indicator bacteria, E. coli, suggesting that Bacteroidales was a less conservative indicator of human health risk as compared with E. coli.

4.1.

DNA vs RNA

Minimal information on the persistence and survival of host specific Bacteroidales marker cell in the environment is available (Bae and Wuertz, 2009; Balleste and Blanch, 2010; Walters and Field, 2009). Molecular methods relying on the detection

253

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 4 7 e2 5 7

Table 2 e Sequence affiliation of bands obtained from Bacteroidales 16S rDNA/rRNA DGGE profile in microcosms seeded with bovine feces and the last day of band detection during the 14 days incubation. Band position

Bovine Bacteroidales DGGE band identification

Similarity %

Closest matches in NCBI database 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

GU939540.1 Uncultured Bacteroidaceae bacterium clone R_247.13-1 HQ993031.1 Uncultured bacterium clone HuMC-B44 JF160912.1 Uncultured bacterium clone ncd1853f02c1 AY978897.1 Uncultured bacterium clone NF59 GU939318.1 Uncultured Bacteroidaceae bacterium clone HC_811.32-1 GU107994.1 Uncultured bacterium clone HFV09_555 EU533564.1 Uncultured bacterium clone Ch6b_1633 EU913596.1 Uncultured Bacteroidales bacterium clone Fhc77 GU939593.1 Uncultured Bacteroidaceae bacterium clone R_347.41-1 GU939318.1 Uncultured Bacteroidaceae bacterium clone HC_811.32-1 FJ220013.1 Uncultured Bacteroides sp. clone RacWA-23 AY978418.1 Uncultured bacterium clone KO17 JF298878.1 Bacteroides dorei strain EBA14-8 GU939494.1 Uncultured Bacteroidaceae bacterium clone PA_525.16-2 GU362577.1 Uncultured bacterium clone Bac.I-11m.h435 HM007585.1 Bacteroides fragilis strain DSM 9671 HQ992999.1 Uncultured bacterium clone HuMC-A3 HQ992999.1 Uncultured bacterium clone HuMC-A3 DQ905531.1j Uncultured bacterium clone 014-g8 JF298885.1 Bacteroides fragilis strain EBA21-17 DQ826965.1 Unidentified bacterium clone H2-20 HQ992999.1 Uncultured bacterium clone HuMC-A3 GU939318.1 Uncultured Bacteroidaceae bacterium clone HC_811.32-1 GQ047288.1 Uncultured bacterium clone nbw1011a03c1 HQ992999.1 Uncultured bacterium clone HuMC-A3 HQ896759.1 Uncultured Bacteroidetes bacterium clone 31 JF298891.1 Bacteroides uniformis strain EBA25-2 GU939540.1 Uncultured Bacteroidaceae bacterium clone R_247.13e1 JF160912.1 Uncultured bacterium clone ncd1853f02c1 HQ896738.1 Uncultured Bacteroidetes bacterium clone 2

of DNA directly from environmental samples lack the ability to differentiate live and dead cells because DNA from various origins, including live/dead cells, cell in viable but nonculturable status, and extracellular sources, will be detected. Research on the use of propidium monoazide (PMA) indicated promise to distinguish DNA from live cells (Bae and Wuertz, 2009), but the high turbidity in some environmental waters may seriously affect the penetration of PMA into live cells and thus the reliability of these methods. Analysis targeting rRNA, on the other hand, offers an alternative for the detection of metabolically active cells because the number of ribosomes per cell is a good indicator of the overall activity of the cell (Schaechter et al., 1958) and rRNA usually degraded quickly when the cell becomes less active. In addition, studies targeting both rDNA and rRNA have the potential to elucidate not only the presence of the markers but also the activity of the species to which the markers are related (Eichler et al., 2006; Walters and Field, 2009). In this study both the decay rate estimation and the decay curve analysis indicated that rRNA decayed faster than rDNA. This tendency was also clearly evidenced by DGGE profile analysis, further supporting the use of rRNA as indicator of metabolically active Bacteroidales population. As compared to other studies that indicated Bacteroidales spp. die off fast in oxygenated waters (Balleste and Blanch, 2010; Kreader, 1998), the prolonged persistence of Bacteroidales RNA observed in this study, however, was in

97 97 98 95 98 97 99 97 97 98 98 99 98 95 100 95 97 97 98 96 98 98 96 98 99 99 97 96 98 98

Last day of detection DNA

RNA

8 4 4 8 8 8 8 14 8 8 2 8 14 14 2 14 8 8 14 14 6 8 8 4 6 2 14 6 14 8

6 N/A N/A 6 4 4 4 8 4 2 0 2 8 10 6 8 6 6 8 8 0 0 4 0 0 2 12 2 8 2

contrast to the assumption that RNA degraded fast upon the death of cell. This discrepancy may be caused by a number of contributing factors. It is possible that some Bacteroidales can still grow or remain alive for an extended time in low oxygen micro-niches or a portion of the Bacteroidales population enters into viable but none cultivable (VBNC) status (Bell et al., 2009) in which certain level of rRNA was still maintained. Alternately, the complex secondary structure of the RNA may prevent the fast decay after cell death (Rodgers, 1970); or similar to the behavior of extracellular DNA (Lorenz and Wackernagel, 1994) some rRNA molecules may bind on suspended solid particles and thus protect it from enzymatic degradation in the microcosms. We centrifuged the water samples to collect sufficient Bacteroidales biomass for DNA/RNA extraction. Although often used to concentrate microbial biomass for environmental water sample (Dick et al., 2010; Okabe and Shimazu, 2007; Walters and Field, 2009), this gravity based step may result in the loss of considerable amounts of the extracellular DNA/ RNA because only those in intact cells or associated with solid particles will be captured and subsequently extracted. Considering the extended persistence of extracellular DNA in the environment compared to intact cells, the loss of the free DNA in this form may result in overestimation of decay rates. When alternative procedures, such as membrane filtration, are used to collect biomass, it is possible that different decay

254

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 4 7 e2 5 7

Table 3 e Sequence affiliation of bands obtained from Bacteroidales 16S rDNA/rRNA DGGE profile in microcosms seeded with human feces and the last day of band detection during the 14 days incubation. Band position

Human Bacteroidales DGGE band identification

Similarity %

Closest matches in NCBI database 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

EF709128.1 Uncultured Bacteroidales bacterium clone MS194A1_F09 HQ896759.1 Uncultured Bacteroidetes bacterium clone 31 JF298890.1 Bacteroides thetaiotaomicron strain EBA25-1 DQ456039.1 Uncultured bacterium clone CFT112F7 GU939215.1 Uncultured Bacteroidaceae bacterium clone FS_133.66-1 DQ827205.1 Unidentified bacterium clone H5-59 FJ210103.1 Uncultured Bacteroides sp. clone BAC_B2_53.T7 JF298885.1 Bacteroides fragilis strain EBA21-17 JF298890.1 Bacteroides thetaiotaomicron strain EBA25-1 AB510697.1 Bacteroides caccae gen JF298873.1 Bacteroides vulgatus strain EBA12-12 JF298887.1 Bacteroides xylanisolvens strain EBA22-11 JF298877.1 Bacteroides vulgatus strain EBA7-36 JF298887.1 Bacteroides xylanisolvens strain EBA22-11 JF298878.1 Bacteroides dorei strain EBA14-8 FJ509956.1 Uncultured bacterium clone 16slp74-02b01.p1k JF298890.1 Bacteroides thetaiotaomicron strain EBA25-1 JF298882.1 Bacteroides caccae strain EBA18-15 JF298878.1 Bacteroides dorei strain EBA14-8 JF298885.1 Bacteroides fragilis strain EBA21-17 EU913596.1 Uncultured Bacteroidales bacterium clone Fhc77 HQ992999.1 Uncultured bacterium clone HuMC-A3 JF298891.1 Bacteroides uniformis strain EBA25-2 JF298882.1 Bacteroides caccae strain EBA18-15 HM443005.1 Uncultured Bacteroidetes bacterium clone SS506

dynamic may be obtained. We considered this in the data interpretation.

4.2.

PCR-DGGE analysis of Bacteroidales populations

The high diversity of Bacteroidales observed in this study is generally in agreement with previous studies (Jeter et al., 2009). A total of 30 band types were observed from bovine microcosms and 25 band types from human microcosms. This may be an underestimation of the Bacteroidales because it is possible that some bands are common to different individuals (Pang et al., 2005). Most sequences from the bovine microcosms were closely affiliated with uncultured Bacteroidales from various sources, suggesting that bovine Bacteroidales is a very deep order and its diversity is largely under-explored (Dick et al., 2005; Lamendella et al., 2007). The detection of B. thetaiotaomicron affiliated sequences from human microcosm was in agreement with previous studies that this species dominates in human Bacteroidales mirobiota. B. fragilis , B. uniformis and B. dorei were detected from both microcosms, indicating they are shared by both hosts. Noticeably, bands affiliated with B. fragilis from human (band 20) and bovine (band 16 and band 20) DGGE profiles persisted until the end of the experiment. This is consistent with the finding that this species has a relatively high oxygen tolerance and can survive in aerobic conditions for extended period of time (Baughn and Malamy, 2004; Rocha et al., 2003). As indicated by DGGE analysis, complicated population trends of persistence were observed, possibly caused by the heterogeneity of Bacteroidales populations present in the

98 98 100 98 100 99 98 98 99 99 98 98 99 99 96 97 99 97 96 99 98 99 99 97 97

Last day of detection DNA

RNA

10 10 12 4 12 10 10 4 4 8 4 10 2 14 12 12 14 10 12 14 12 4 14 14 14

2 2 4 2 2 2 2 2 2 2 2 2 2 6 4 4 6 2 4 10 0 4 10 6 10

microcosms. It is reasonable to claim that Bacteroidales population in the feces used as inoculum contained different subpopulations at various physiological status, and different strains respond differently to a range of environmental conditions (Balleste and Blanch, 2010). Although Bacteroidales population from human and bovine microcosm exhibited differential diversity and band patterns, some common phenomena were still evident: 1) diversity of both microcosms decreased with time, which is expected because of the decline in richness attributed to the limited potential of most Bacteroidales to grow in aerobic conditions; 2) most samples collected from both microcosms in the first 8 days had a richness of more than 10, implying a uniform persistence pattern for most genotypes of Bacteroidales, which is consistent with the finding of Blanch (Balleste and Blanch, 2010) that DNA from all three species of Bacteroidales decays similarly in freshwater microcosm; and 3) several bands were present on the DGGE profile throughout the experiment, revealing the ability of some members of the Bacteroidales population to persist over a long period of time, possibly in a VBNC status or attached to suspended solid to prevent the fast degradation of DNA. Our results were generally in agreement with the findings of a recent study investigating the persistence of Bacteroidales populations in microcosms spiked with sewage and manure influent (Schulz and Childers, 2011).

4.3.

Decay curve analysis

A close examination of the decay curve indicated that neither host specific Bacteroidales markers nor E. coli decayed linearly

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 4 7 e2 5 7

and the patterns fit poorly with a first-order model. The difference in the shape of the decay curve was possibly caused by heterogeneity of marker cells present in the microcosms. Each indicator may have its specific persistence pattern as determined by its intrinsic properties including cell wall composition, sensitivity to oxygen and ability to grow extraintestinally. Previous studies have also shown the insufficiency of using single-parameter model in the description of bacterial decay (Bae and Wuertz, 2009; Crane and Moore, 1986; Gonzalez, 1995). The deviation of decay curves from linearity observed in this study suggested that comparing decay rate from different studies without reference to the raw data should be conducted with caution because large variation may be introduced if inappropriate methods are used. This is clearly illustrated by the persistence profile of HF183 in this study: decay rate estimated over the course of the experiment was 0.7 for DNA and 0.9 for RNA, which were much different from the rates derived from the exponential stage: 1.2 for DNA and 3.1 for RNA (data not shown). One specific need for future studies, as advocated by Schulz and Childers (2011), is to have raw data available for reliable comparison between studies and the development of models. The decay rates of HF183 (0.7) and CF193 (0.7) DNA markers during the entire incubation period were lower than those observed in freshwater in similar studies (HF813 at rate of 1.7 and CF193 at rate of 1.0) (Walters and Field, 2009). This discrepancy may be attributed to the difference of environmental parameters during the incubation such as temperature, sunlight exposure, and the level of predators: strong correlations between temperature and the inactivation of Bacteroidales spp. have been reported (Balleste and Blanch, 2010; Bell et al., 2009; Kreader, 1998; Okabe and Shimazu, 2007). Microcosms in our study experienced a fluctuation of temperature ranging from 4  C to 18  C with an average of 11  C during most of the incubation time. The highest temperature (26  C) was observed in only one day at the late stage of the test. It is possible that temperature variation over such a wide range may be conducive to the persistence of marker DNA. The effects of sunlight on the die-off of microorganisms are variable (Bae and Wuertz, 2009; Dick et al., 2010; Walters et al., 2009). We established the microcosms in a polycarbonate roofed greenhouse, which prevented the direct exposure to sunlight and reduced the UV radiation penetrated into the microcosms, thus delaying the decomposition of marker DNA. Moreover, the presence of predators in the water significantly influences the persistence of Bacteroidales spp. (Bell et al., 2009; Kreader, 1998; Okabe and Shimazu, 2007). The water collected from the canal during the winter time (Feb 2010) in this study might contain lower levels of bacterivorous predators. The strain specific differences as discovered above may also contribute to the long persistence of marker DNA. Other factors, such as the type of inocula, the starting concentration of markers, and the time of the year when the experiment was carried out may also dramatically affected the decay of Bacteroidales, making direct comparison between the two studies difficult. In this study, although qPCR monitoring of the decay of host specific markers and PCR-DGGE profiling the persistence of Bacteroidales population targeted different region of 16S rRNA, some common patterns could still be perceived. For instance, PCR-DGGE profiles of DNA indicated an abrupt

255

decline of bovine Bacteroidales populations after day 8, and the DGGE profile of RNA showed a sharp decrease after day 6. Decay curves of the bovine marker revealed the same trend. This similarity possibly indicates the ubiquity of bovine markers in different lineages of Bacteroidales populations. Although this conclusion is in conflict with the finding of Silkie and Nelson (2009) that bovine specific markers comprised only 4% of the total Bacteroidales, other studies revealed the high variability of human specific Bacteroidales marker HF183 in different individual host, with concentrations ranging from non-detectable to >109 markers g1 wet feces (Seurinck et al., 2005). It is highly probable that the same level of viability may occur in the case of bovine markers.

4.4.

Implication for future research

Although we designed the microcosms to simulate environmental conditions, because environmental parameters that Bacteroidales may encounter are too complex and unpredictable to replicate in the laboratory, further evaluations are warranted before extending the findings of this study to real field settings. For instance, a dialysis tube that provides a continuous flow-through system may be positioned in situ in natural scenarios to monitor the decay of Bacteroidales (Bae and Wuertz, 2009; Balleste and Blanch, 2010). Mixed source samples such as sewage and wastewater that are more representative of the natural pollution scenarios (Schulz and Childers, 2011) should be used in future studies. In addition, sediment resuspension contributed a significant portion of bacteria pollution to water columns (Fries et al., 2008) and, compared to water, the survival of indicator bacteria was enhanced in sediments (Dick et al., 2010). So the presence of sediment may also significantly affect the survival pattern of Bacteroidales populations. Further, because high levels of fecal indicators were detected in runoff water from agricultural land (Howell et al., 1995), it is essential to determine the influence of agricultural practices, such as the application of chemical fertilizers and pesticides, that may lead to the poor water quality on the persistence of Bacteroidales population in agricultural runoff.

5.

Conclusion

 Using DGGE analysis of 16S rRNA genes, we observed a high diversity of Bacteroidales populations from both human and bovine microcosms, indicating the complexity of Bacteroidales spp. Most of the sequences from bovine microcosm are affiliated with uncultured B species, suggesting the largely unexplored nature of Bacteroidales.  Diversity of Bacteroidales decreased with time for both DNA and RNA based analysis because of the decline of Bacteroidales richness due to its anaerobic metabolism.  Persistence patterns of the dominant Bacteroidale species are uniform, some members of the Bacteroidales population persisted over the course of the 14 day incubation.  As indicated by previous studies, a first-order model is not sufficient to describe the decay pattern of both markers and E. coli because of the presence of shoulder and tail time.

256

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 4 7 e2 5 7

 Difference in the decay rates in this study from previous work may be caused by the difference in environmental parameters.  Further investigations are needed to determine the persistence pattern of Bacteroidales populations in more natural scenarios.

Acknowledgments This study was in part supported by a grant (contract #4600001774) from South Florida Water Management District. We thank Dr. V. Harwood in University of South Florida for technical assistance and valuable suggestion in this project. Special thanks are extended to the two anonymous reviewers whose comments improved the manuscript.

references

Bae, S., Wuertz, S., 2009. Rapid decay of host-specific fecal Bacteroidales cells in seawater as measured by quantitative PCR with propidium monoazide. Water Research 43 (19), 4850e4859. Balleste, E., Blanch, A.R., 2010. Persistence of Bacteroides species populations in a river as measured by molecular and culture techniques. Applied Environment Microbiology 76 (22), 7608e7616. Baughn, A.D., Malamy, M.H., 2004. The strict anaerobe Bacteroides fragilis grows in and benefits from nanomolar concentrations of oxygen. Nature 427 (6973), 441e444. Bell, A., Layton, A.C., McKay, L., Williams, D., Gentry, R., Sayler, G.S., 2009. Factors influencing the persistence of fecal Bacteroides in stream water. Journal of Environmental Quality 38 (3), 1224e1232. Bernhard, A.E., Field, K.G., 2000. A PCR assay to discriminate human and ruminant feces on the basis of host differences in Bacteroides-Prevotella genes encoding 16S rRNA. Applied Environment Microbiology 66 (10), 4571e4574. Crane, S.R., Moore, J.A., 1986. Modeling enteric bacterial die-off a review. Water, Air, & Soil Pollution 27 (3e4), 411e439. Dick, L.K., Bernhard, A.E., Brodeur, T.J., Domingo, J.W.S., Simpson, J.M., Walters, S.P., Field, K.G., 2005. Host distributions of uncultivated fecal Bacteroidales bacteria reveal genetic markers for fecal source identification. Applied Environment Microbiology 71 (6), 3184e3191. Dick, L.K., Stelzer, E.A., Bertke, E.E., Fong, D.L., Stoeckel, D.M., 2010. Relative decay of Bacteroidales microbial source tracking markers and cultivated Escherichia coli in freshwater microcosms. Applied Environment Microbiology 76 (10), 3255e3262. Eckburg, P.B., Bik, E.M., Bernstein, C.N., Purdom, E., Dethlefsen, L., Sargent, M., Gill, S.R., Nelson, K.E., Relman, D.A., 2005. Diversity of the human intestinal microbial flora. Science 308 (5728), 1635e1638. Eichler, S., Christen, R., Holtje, C., Westphal, P., Botel, J., Brettar, I., Mehling, A., Hofle, M.G., 2006. Composition and dynamics of bacterial communities of a drinking water supply system as assessed by RNA- and DNA-based 16S rRNA gene fingerprinting. Applied Environment Microbiology 72 (3), 1858e1872. Fremaux, B., Gritzfeld, J., Boa, T., Yost, C.K., 2009. Evaluation of host-specific Bacteroidales 16S rRNA gene markers as

a complementary tool for detecting fecal pollution in a prairie watershed. Water Research 43 (19), 4838e4849. Fries, J.S., Characklis, G.W., Noble, R.T., 2008. Sediment-water exchange of Vibrio sp and fecal indicator bacteria: implications for persistence and transport in the Neuse River Estuary, North Carolina, USA. Water Research 42 (4e5), 941e950. Gawler, A.H., Beecher, J.E., Brandao, J., Carroll, N.M., Falcao, L., Gourmelon, M., Masterson, B., Nunes, B., Porter, J., Rince, A., Rodrigues, R., Thorp, M., Walters, J.M., Meijer, W.G., 2007. Validation of host-specific bacteriodales 16S rRNA genes as markers to determine the origin of faecal pollution in Atlantic rim countries of the European union. Water Research 41 (16), 3780e3784. Gonzalez, J.M., 1995. Modelling enteric bacteria survival in aquatic systems. Hydrobiologia 316 (2), 109e116. Gourmelon, M., Caprais, M.P., Segura, R., Le Mennec, C., Lozach, S., Piriou, J.Y., Rince, A., 2007. Evaluation of two library-independent microbial source tracking methods to identify sources of fecal contamination in French estuaries. Applied Environment Microbiology 73 (15), 4857e4866. Howell, J.M., Coyne, M.S., Cornelius, P., 1995. Fecal bacteria in agricultural waters of the bluegrass region of Kentucky. Journal of Environmental Quality 24 (3), 411e419. Jeter, S.N., McDermott, C.M., Bower, P.A., Kinzelman, J.L., Bootsma, M.J., Goetz, G.W., McLellan, S.L., 2009. Bacteroidales diversity in ring-billed gulls (Laurus delawarensis) residing at lake Michigan beaches. Applied Environment Microbiology 75 (6), 1525e1533. Kildare, B.J., Leutenegger, C.M., McSwain, B.S., Bambic, D.G., Rajal, V.B., Wuertz, S., 2007. 16S rRNA-based assays for quantitative detection of universal, human-, cow-, and dogspecific fecal Bacteroidales: a Bayesian approach. Water Research 41 (16), 3701e3715. Kreader, C.A., 1995. Design and evaluation of Bacteroides DNA probes for the specific detection of human fecal pollution. Applied Environment Microbiology 61 (4), 1171e1179. Kreader, C.A., 1998. Persistence of PCR-detectable Bacteroides distasonis from human feces in river water. Applied Environment Microbiology 64 (10), 4103e4105. Lamendella, R., Santo Domingo, J.W., Oerther, D.B., Vogel, J.R., Stoeckel, D.M., 2007. Assessment of fecal pollution sources in a small northern-plains watershed using PCR and phylogenetic analyses of Bacteroidetes 16S rRNA gene. FEMS Microbiology Ecology 59 (3), 651e660. Layton, A., McKay, L., Williams, D., Garrett, V., Gentry, R., Sayler, G., 2006. Development of Bacteroides 16S rRNA gene TaqMan-based real-time PCR assays for estimation of total, human, and bovine fecal pollution in water. Applied Environment Microbiology 72 (6), 4214e4224. Li, K., Guan, W., Wei, G., Liu, B., Xu, J., Zhao, L., Zhang, Y., 2007. Phylogenetic analysis of intestinal bacteria in the Chinese mitten crab (Eriocheir sinensis). Journal of Applied Microbiology 103 (3), 675e682. Li, M., Zhou, H., Hua, W., Wang, B., Wang, S., Zhao, G., Li, L., Zhao, L., Pang, X., 2009. Molecular diversity of Bacteroides spp. in human fecal microbiota as determined by group-specific 16S rRNA gene clone library analysis. Systematic and Applied Microbiology 32 (3), 193e200. Liang, Z.B., Drijber, R.A., Lee, D.J., Dwiekat, I.M., Harris, S.D., Wedin, D.A., 2008. A DGGE-cloning method to characterize arbuscular mycorrhizal community structure in soil. Soil Biology and Biochemistry 40 (4), 956e966. Liu, C.X., Song, Y.L., McTeague, M., Vu, A.W., Wexler, H., Finegold, S.M., 2003. Rapid identification of the species of the Bacteroides fragilis group by multiplex PCR assays using group- and species-specific primers. FEMS Microbiology Letters 222 (1), 9e16.

w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 4 7 e2 5 7

Lorenz, M.G., Wackernagel, W., 1994. Bacterial gene-transfer by natural genetic-transformation in the environment. Microbiology Reviews 58 (3), 563e602. Okabe, S., Shimazu, Y., 2007. Persistence of host-specific Bacteroides-Prevotella 16S rRNA genetic markers in environmental waters: effects of temperature and salinity. Applied Microbiology and Biotechnology 76 (4), 935e944. Pang, X.Y., Ding, D.Z., Wei, G.F., Zhang, M.L., Wang, L.H., Zhao, L.P., 2005. Molecular profiling of Bacteroides spp. in human feces by PCR-temperature gradient gel electrophoresis. Journal of Microbiological Methods 61 (3), 413e417. Rocha, E.R., Herren, C.D., Smalley, D.J., Smith, C.J., 2003. The complex oxidative stress response of Bacteroides fragilis: the role of OxyR in control of gene expression. Anaerobe 9 (4), 165e173. Rodgers, A., 1970. Effect of secondary structure on degradation of ribosomal RNA. Biochimica et Biophysica Acta 217 (2), 540. Savichtcheva, O., Okayama, N., Ito, T., Okabe, S., 2005. Application of a direct fluorescence-based live/dead staining combined with fluorescence in situ hybridization for assessment of survival rate of Bacteroides spp. in drinking water. Biotechnology and Bioengineering 92 (3), 356e363. Schaechter, M., Maaloe, O., Kjeldgaard, N.O., 1958. Dependency on medium and temperature of cell size and chemical composition during balanced grown of Salmonella typhimurium. Journal of General Microbiology 19 (3), 592e606. Schulz, C.J., Childers, G.W., 2011. Fecal Bacteroidales diversity and decay in response to variations in temperature and salinity. Applied Environment Microbiology 77 (8), 2563e2572.

257

Seurinck, S., Defoirdt, T., Verstraete, W., Siciliano, S.D., 2005. Detection and quantification of the human-specific HF183 Bacteroides 16S rRNA genetic marker with real-time PCR for assessment of human faecal pollution in freshwater. Environmental Microbiology 7 (2), 249e259. Silkie, S.S., Nelson, K.L., 2009. Concentrations of host-specific and generic fecal markers measured by quantitative PCR in raw sewage and fresh animal feces. Water Research 43 (19), 4860e4871. Walters, S.P., Field, K.G., 2006. Persistence and growth of fecal Bacteroidales assessed by bromodeoxyuridine immunocapture. Applied Environment Microbiology 72 (7), 4532e4539. Walters, S.P., Field, K.G., 2009. Survival and persistence of human and ruminant-specific faecal Bacteroidales in freshwater microcosms. Environmental Microbiology 11 (6), 1410e1421. Walters, S.P., Gannon, V.P.J., Field, K.G., 2007. Detection of Bacteroidales fecal indicators and the zoonotic pathogens EColi O157: H7, Salmonella, and Campylobacter in river water. Environmental Science & Technology 41 (6), 1856e1862. Walters, S.P., Yamahara, K.M., Boehm, A.B., 2009. Persistence of nucleic acid markers of health-relevant organisms in seawater microcosms: implications for their use in assessing risk in recreational waters. Water Research 43 (19), 4929e4939. Wilkins, T.D., Wagner, D.L., Veltri, B.J., Gregory, E.M., 1978. Factors affecting production of catalase by bacteroides. Journal of Clinical Microbiology 8 (5), 553e557. Yuan, J., Zeng, B., Niu, R., Tang, H., Li, W., Zhang, Z., Wei, H., 2011. The development and stability of the genus Bacteriodes from human gut microbiota in HFA Mice model. Current Microbiology 62 (4), 1107e1112.