Water Research Volume 45 Issue 12

WATER RESEARCH A Journal of the International Water Association

Editor-in-Chief Mark van Loosdrecht Delft University of Technology Department of Biochemical Engineering Julianalaan 67 2628 BC Delft The Netherlands Tel: +31 15 27 81618 E-mail:[email protected]

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

Associate Editors Andrew Baker University of New South Wales Australia

Damien Batstone The University of Queensland Australia G-H. Chen The Hong Kong University of Science & Technology Hong Kong China Tom Curtis Univ. of Newcastle upon Tyne UK Ana Deletic Monash University USA Francis de los Reyes III North Carolina State University USA Rob Eldridge The University of Melbourne Australia Rosina Girones University of Barcelona Spain Stephen Gray Victoria University Australia Kate Grudpan Chiang Mai University Thailand E.E. Herricks University of Illinois - Urbana USA Peter Hillis United Utilities Plc UK H-Y. Hu Tsinghua University China P.M. Huck University of Waterloo Canada Bruce Jefferson Cranfield University UK Ulf Jeppsson Lund University Sweden Sergey Kalyuzhnyi Moscow State University Russian Federation Jaehong Kim Georgia Institute of Technology USA Jes La Cour Jansen Lund Institute of Technology Sweden G. Langergraber BOKU/Univ. of Natural Res. and Applied Life Scs. Austria S-L. Lo National Taiwan University Taiwan Dionisis Mantzavinos Technical University of Crete Greece

Y. Matsui Hokkaido University Japan A. Maul Université Paul Verlaine-Metz France Max Maurer EAWAG Switzerland How Yong Ng National University of Singapore Singapore Satoshi Okabe Hokkaido University Japan S.L. Ong National University of Singapore Singapore Jong M. Park Pohang University of Science & Technology Korea Susan Richardson U.S. Environmental Protection Agency USA Miguel Salgot University of Barcelona Spain David Sedlak University of California, Berkeley USA Jean-Philippe Steyer LBE-INRA France M. Takahashi Hokkaido University Japan Kai Udert EAWAG Switzerland V.P. Venugopalan BARC Facilities India E. von Sperling Federal University of Minas Gerais Brazil Hanqing Yu University of Science & Technology of China China J. Zilles University of Illinois Urbana USA A.I. Zouboulis Aristotle University Greece

Editorial Office Mailing address Water Research Radarweg 29 1043 NX Amsterdam The Netherlands E-mail: [email protected]

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

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 5 1 e3 5 7 0

Available at www.sciencedirect.com

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

Review

Ozone oxidation for the alleviation of membrane fouling by natural organic matter: A review Steven Van Geluwe a,*, Leen Braeken a,b, Bart Van der Bruggen a a

Laboratory of Applied Physical Chemistry and Environmental Technology, Department of Chemical Engineering, K.U. Leuven, W. de Croylaan 46, B-3001 Leuven (Heverlee), Belgium b Department of Industrial Sciences and Technology, KHLim, Universitaire Campus Gebouw B, Bus 3, B-3590 Diepenbeek, Belgium

article info

abstract

Article history:

Membrane fouling by natural organic matter is one of the main problems that slow down

Received 26 January 2011

the application of membrane technology in water treatment. O3 is able to efficiently change

Received in revised form

the physico-chemical characteristics of natural organic matter in order to reduce

1 April 2011

membrane fouling. This paper presents the state-of-the-art knowledge of the reaction

Accepted 8 April 2011

mechanisms between natural organic matter and molecular O3 or OH radicals, together

Available online 15 April 2011

with an in-depth discussion of the interactions between natural organic matter and




membranes that govern membrane fouling, inclusive the effect of O3 oxidation on it. ª 2011 Elsevier Ltd. All rights reserved.

Keywords: Humic acids Hydrophobicity Electrostatic interactions Molecular mass Aggregation Hydrogen peroxide

Contents 1. 2. 3.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The chemical composition of different NOM fractions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The decomposition of NOM by ozone and hydroxyl radicals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1. Ozone reacts selectively with certain functional groups in NOM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2. Which functional groups in the NOM can act as a promoter or inhibitor of O3 decomposition? . . . . . . . . . . . . . 3.3. Reaction mechanisms of NOM with OH radicals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4. The addition of H2O2 slightly improves the mineralization of organic matter during ozonation . . . . . . . . . . . . . 3.5. Guidelines for finding the optimal dose of H2O2 in water treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.6. A short note on the health hazard of O3 in water treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.6.1. Bromate formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.6.2. Trihalomethanes and haloacetic acids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .


3552 3553 3553 3553 3555 3558 3559 3559 3560 3560 3560

* Corresponding author. Tel.: þ32 16 322 341; fax: þ32 16 322 991. E-mail addresses: [email protected] (S. Van Geluwe), [email protected] (L. Braeken), bart.vanderbruggen@cit. kuleuven.be (B. Van der Bruggen). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.016

3552

4.

5.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 5 1 e3 5 7 0

The structural changes of NOM by O3 oxidation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1. Hydrophobic interactions between NOM and membrane surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2. Molecular size and (dis)aggregation of NOM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1. Results obtained by high-performance size exclusion chromatography (HPSEC) . . . . . . . . . . . . . . . . . . . . 4.2.2. Aggregation of humic substances by calcium and magnesium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3. Electrostatic interactions and hydrogen bridges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Abbreviations AOP BOD5 COD DBP DOC FTIR HAA IEP MF MWCO

1.

advanced oxidation process biological oxygen demand after 5 days chemical oxygen demand disinfection by-product dissolved organic carbon Fourier transform infrared haloacetic acid isoelectric point microfiltration molecular weight cut-off

Introduction

Membrane technology has become well established in water treatment, and the demand for membranes increases yearly by 8% (Leiknes, 2009). The most important type of membrane processes are pressure-driven, including microfiltration (MF), ultrafiltration (UF), nanofiltration (NF) and reverse osmosis (RO). Typical values of the main membrane characteristics, i.e. water permeability, operating pressure, pore size and retention characteristics for these four membrane types are listed in Table 1. Because of the large pores of the MF and UF membranes, the water flux is high while the transmembrane pressure is low. MF is used for the removal of suspended particles, turbidity and various micro-organisms (Yuan and Zydney, 1999), while UF removes viruses (van Voorthuizen et al., 2001), colloids and the high-molecular mass fraction of natural organic matter (NOM) as well (Siddiqui et al., 2000; Lee et al., 2005a; Kennedy et al., 2005). NF membranes have smaller pores, but still maintain a fairly high flux at a reasonable pressure. NF is very effective in the removal of the mediumand lower-molecular mass fraction of NOM (Siddiqui et al., 2000; Shon et al., 2004; Meylan et al., 2007; de la Rubia et al., 2008), and emerging micropollutants such as pesticides, pharmaceuticals and endocrine disrupting chemicals (Kimura et al., 2003; Nghiem et al., 2004; Yoon et al., 2006; Verliefde et al., 2007). The retention of inorganic ions by NF membranes is strongly dependent on the charge of the ions. The retention of divalent ions ranges between 50 and 100%. It is much higher than the retention of monovalent ions, which is usually lower than 40%, because of Donnan exclusion (de la Rubia et al., 2008; Ouyang et al., 2008). RO is commonly used for desalting brackish water and seawater, but operates under very high transmembrane pressures and a low permeate flux compared

NF NMR NOM PVDF RO SBH THM TOC UF USEPA UVA

3560 3560 3560 3560 3563 3565 3566 3566 3566

nanofiltration nuclear magnetic resonance natural organic matter polyvinylidene fluoride reverse osmosis Staehelin, Bu¨hler and Hoigne´ trihalomethane total organic carbon ultrafiltration United States Environmental Protection Agency absorbance (optical density) of UV irradiation

to the other pressure-driven membranes. However, RO shares about 45% of the global production capacity of desalinated water, because of its lower energy consumption compared to multistage flash evaporation (Darwish and Al-Najem, 2000; Eltawil et al., 2009). In spite of the excellent retention characteristics of membrane filtration in water treatment, there are still problems that slow down its growth. The best known problem is fouling of the membrane, which results in a reduction in water flux, and thus leads to higher operating costs. Over time, fouling and subsequent cleaning of the membranes causes deterioration of membrane materials, resulting in a compromised permeate water quality and ultimately, a shorter  and Kunst, 2002; Seidel and membrane lifetime (Kosutic Elimelech, 2002; Al-Amoudi and Lovitt, 2007). Membrane fouling is usually minimized by an excessive pretreatment or else a very conservative membrane flux needs to be used. Consequently, the capital cost is high, which makes membrane filtration less competitive against conventional water treatment technologies (such as coagulation or activated carbon) in certain cases (Pianta et al., 2000). The emerging use of O3 oxidation in water treatment offers new opportunities, because O3 is able to decompose certain membrane foulants very efficiently. The present paper is a critical review of literature concerning the fouling potential of NOM in water purification and the use of O3 oxidation for the alleviation of membrane fouling by NOM. The effect of O3 oxidation on membrane fouling is difficult to predict due to the complex nature of NOM, the strong variability of the NOM characteristics and the water matrix with location, season and weather (Lowe and Hossain, 2008), and the major effect of the water matrix on the conformation of NOM and the decomposition of O3. This review paper presents the reaction

3553

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 5 1 e3 5 7 0

Table 1 e Comparison between the four pressure-driven membrane processes with respect to permeability, applied pressure, pore size and rejection characteristics.

2

1

1

Pure water permeability (L m h MPa ) Transmembrane pressure (kPa) Pore size (nm) Molecular weight cut-off (g mol1) Retention: Suspended particles Macromolecules Small organic molecules Multivalent salts Monovalent salts

Microfiltration (MF)

Ultrafiltration (UF)

Nanofiltration (NF)

Reverse osmosis (RO)

>>500 10e100 100-10,000

100e500 100e500 2e100 1000e100,000

15e150 500e2000 0.1e2 150e1000

0.5e15 2000e4000 <1

þ þ /þ þ /þ

þ þ þ þ þ

þ e e e e

mechanisms and its products when NOM solutions are treated by O3 or O3 þ H2O2, and the various interactions that exist between NOM components and the membrane surface. The relation between the complex fouling behavior of NOM, before and after its reaction with O3, is discussed in a systematic and detailed way, in order to better understand the mechanisms behind the fouling reduction in water treatment by O3 treatment.

2. The chemical composition of different NOM fractions NOM is a complex heterogeneous mixture of organic material, such as humic substances, polysaccharides, aminosugars, proteins, peptides, lipids, small hydrophilic acids, and others (Frimmel et al., 2002). In a first approach to separate the different components in NOM, it is divided into two major classes. The first class, i.e. autochthonous NOM, is derived from extracellular macromolecules of micro-organisms in the water body and carbon fixation by algae and aquatic plants. The second class, allochthonous NOM, is derived from the decay of plant and animal residues in the watershed (Frimmel et al., 2002). It is usually referred to as humic substances, and this term will be used in the remaining text. Although the correct chemical structure of humic substances still remains unknown, they consist of a skeleton of alkyl and aromatic units, cross-linked by a variety of functional groups. Humic substances are high in aromatic carbon and have a negative charge. This charge is primarily contributed by their three main functional groups, namely carboxylic acids, methoxyl carbonyls and phenolic functional groups (Thurman, 1986). The chemical properties of humic substances are succinctly explained by McDonald et al. (2004) and Sutzkover-Gutman et al. (2010). The humic substances present in natural waters are traditionally divided into two categories, namely humic and fulvic acids. Humic acids have a higher molecular mass (2000e5000 g mol1) than fulvic acids (500e2000 g mol1) (Her et al., 2003). They have a lower oxygen content and are more hydrophobic than fulvic acids (Thurman, 1986). The knowledge of the structural chemistry of humic substances is somewhat like that of proteins in the middle of the last century. Although the main building blocks are known, there are no conclusive studies about the long-range

þ þ e e e

conformational structure of humic substances. This is probably because of the difficulty of obtaining reliable structural data, due to the size and number of stereochemical isomers (Jansen et al., 1996). In addition, the conformation of humic substances may vary significantly due to changes in pH (humic substances are highly deprotonated is most aquatic environments), cation concentration (humic substances form strong complexes with metals and other cations) and the great stabilization effect on the electrostatic energy by the presence of water molecules (Kubicki and Apitz, 1999). Therefore, the complexity of humic substances in its natural environment is too high for the application of modeling techniques on an atomic scale. Autochthonous NOM includes a large number of relatively simple compounds of known structures: carbohydrates, aminosugars, proteins, peptides, small organic acids, etc. The main functional groups in autochthonous NOM are carboxylic acids, alcohols and amines, which make these compounds hydrophilic, in contrast to the more hydrophobic humic substances. The molecular mass of these hydrophilic compounds shows a great variety. The simple organic molecules have a molecular mass of a few 100 g mol1, and thus have an apparently smaller molecular size than humic substances, while the biopolymers, such as polysaccharides and proteins, have a molecular mass between 10,000 and 30,000 g mol1 (Lee et al., 2004), which is about one order of magnitude higher than the molecular mass of humic substances. Polysaccharides present in surface water have a diameter in the range between 2 and 20 nm. They have a linear molecular structure and can be hundreds of times longer than wide and can be branched (Leppard, 1997).

3. The decomposition of NOM by ozone and hydroxyl radicals 3.1. Ozone reacts selectively with certain functional groups in NOM O3 is a powerful oxidant and its high reactivity can be attributed to the electronic configuration of this molecule. The O3 molecule represents a hybrid, formed by the two possible resonance structures shown in Fig. 1. The positive formal charges on the central oxygen atom in both resonance structures explains the electrophilic character of O3.

3554

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 5 1 e3 5 7 0

Fig. 1 e Resonance forms of the O3 molecule. Adapted from Beltra´n (2004).

Conversely, the excess negative charge present in one of the terminal atoms imparts a nucleophilic character to O3. These properties make O3 an extremely reactive compound (Beltra´n, 2004). The electrophilic character of O3 accounts for the very fast reaction of O3 with unsaturated bonds (von Gunten, 2003a). The fast reaction of O3 with double bonds and aromatic rings present in NOM molecules, is manifested by a sharp decrease of the optical density at 254 nm (UVA254) during ozonation. For instance, Song et al. (2010) reported a UVA254 reduction of 71% in surface water treatment, at a O3 dosage of 3.0 mg L1 (oxidation time: 10 min). Wang and Pai (2001) reported a reduction of 40% for biologically treated wastewater effluents, and Lee et al. (2009) observed a reduction of 55% for RO concentrates in wastewater reclamation, at the same O3 dosage and oxidation time. Although O3 oxidation is able to efficiently remove unsaturated bonds, it shows only a minor dissolved organic carbon (DOC) removal under acceptable economic conditions. Typical reductions of DOC achieved by ozonation in drinking water plants, with O3 doses between 2 and 5 mg L1, is only about 10e20% (Can and Gurol, 2003). The DOC removal in the experiments of Song et al. (2010) was approximately 10%.

Wang and Pai (2001) reported 15% DOC removal, while Lee et al. (2009) observed 5% DOC removal, for the same O3 dosage and oxidation time as mentioned above. von Gunten (2003a) reports that O3 preferentially reacts with unsaturated bonds to oxygenated saturated functional groups, such as aldehydic, ketonic and especially carboxylic groups. This can be demonstrated by the results of different spectroscopic techniques. Nuclear magnetic resonance (NMR) spectroscopy by Westerhoff et al. (1999), used for investigating the oxidation of surface waters by O3, found a depletion of aromatic against aliphatic moieties. The O3 consumption was positively correlated with aromatic carbon content, especially electron enriched aromatics, and inversely correlated with aliphatic carbon content. Fluorescence spectra of NOM solutions, before  and Sikorska, 2004; Zhang et al., and after O3 oxidation (Swietlik 2008), revealed a reduction of the number of aromatic rings and conjugated bonds, and the decomposition of condensed aromatic moieties to smaller molecules. The number of electron withdrawing groups, such as carboxyl, carbonyl, hydroxyl, alkoxyl and amino groups, increased during ozonation. Mass spectrometry analysis of Suwannee River fulvic acids, by These and Reemtsma (2005), showed that O3 removes preferentially molecules with a low oxidation state (low O/C ratio) and a high degree of unsaturation (low H/C ratio). They also observed that molecules with a more extended carbon skeleton and less carboxylate substituents showed higher reactivity, whereas some highly unsaturated molecules did not show measurable removal up to a specific O3 dose of 2.5 mg per mg DOC. The reaction products were characterized by a very high number of carboxylate groups, i.e. the O/C ratio increased from 0.2 to 0.7. Ozonation products can generally contain alcoholic, carbonyl and carboxyl groups (von Gunten, 2003a). The main reaction products after ozonation mainly consist of shortchain (
Fig. 2 e A hypothetical structure of humic substances (a) before and (b) after O3 oxidation. O3 reacts selectively with electronrich sites (double bonds, aromatic rings) by cycloaddition and electrophilic substitution, and can even breakup aromatic rings. O3 transforms them to carbonyl and carboxyl groups, which adsorb less on hydrophobic membranes. Adapted from Song et al. (2004).

3555

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 5 1 e3 5 7 0

acetaldehyde, glyoxal and methylglyoxal (Xiong et al., 1992; Nawrocki et al., 2003; Hammes et al., 2006; Wert et al., 2007). Oxalic acid is formed mainly by the destruction of aromatic rings by O3 (Kusakabe et al., 1990). The amounts of carboxylic acids generated upon ozonation are usually much higher, i.e. approximately one order of magnitude, than those of aldehydes and ketones (Nawrocki et al., 2003; Xie, 2004). Can and Gurol (2003) observed that a high O3 dose results in a decline in the concentration of aldehydes, due to their oxidation to carboxylic acids. The saturated reaction products accumulate in the solution and are not mineralized, even after long oxidation times (Oh et al., 2003; von Gunten, 2003a; Van Geluwe et al., 2010, Van Geluwe et al., in press). The inefficient reaction between O3 and the saturated reaction products can be demonstrated by the low rate constants between O3 and these molecules, which are given in Table 2. With the exception of formate, which reacts relatively well with O3, the rate constants range between 105 and 101 M1 s1, while rate constants with olefins and aromatic rings can reach values of 106 and 109 M1 s1, respectively (Williamson and Cvetanovic, 1970; Hoigne´ and Bader, 1983b). This explains why O3 oxidation can only achieve a small DOC removal, as mentioned above. The vast abundance of unsaturated bonds in humic substances facilitates the efficient decomposition of these compounds by O3 (Van Geluwe et al., 2009; Van Geluwe et al., 2010). The unsaturated bonds in these molecules are transformed to oxygenated saturated bonds. This is schematically represented in Fig. 2, where a model structure of a humic acid molecule is drawn, before and after O3 oxidation. The decomposition of proteins and polysaccharides by O3 was investigated by Cataldo (2003) and Wang et al. (1999), respectively. Cataldo (2003) found that only the aromatic amino acids tryptophan, tyrosine and phenylalanine are oxidized, as well as cysteine. The most reactive amino acid is tryptophan, followed by tyrosine while phenylalanine appears much less reactive toward O3. The reaction schemes between the former two amino acids and O3 are presented in Fig. 3. Histidine and methionine should probably react quite well with O3, but there is no direct spectral evidence for this claim. Concerning cysteine, O3 oxidates the thiol group, with the consequent formation of disulfide bonds and crosslinks between proteins containing cysteine residues (Fig. 3). The polyamide bond of the main chain of the protein is not degraded by the action of O3, even after prolonged exposure. However, O3 causes denaturation of proteins, i.e. introduces changes in their secondary and tertiary structure (Cataldo,

Table 2 e Rate constants (at 298 K) for the reaction of O3 and OH radicals with the main reaction products after O3 oxidation of NOM. Sources: [a] Hoigne´ and Bader, 1983(a), [b] Hoigne´ and Bader, 1983(b), [c] Buxton et al., 1988.


kO3 (M1 s1) [a,b] k
OH (M1 s1) [c] k
OH =kO3 Formaldehyde Acetaldehyde Acetone Formic acid Formate Acetic acid Acetate Oxalate

0.1 1.5 0.032 5 100 <3.0$105 <3.0$105 4.0$102

2.0$109 7.3$108 6.6$107 1.3$108 2.8$109 1.6$107 8.5$107 7.7$106

2.0$1010 4.9$108 2.1$109 2.6$107 2.8$107 >5.3$1011 >2.8$1012 1.9$108

2003). Wang et al. (1999) showed that O3 depolymerizes polysaccharides by reacting with the glycosidic linkages in those molecules. It selectively oxidates b-D-glycosidic linkages to aldonic esters, as shown in Fig. 4. Ozonolysis proceeds under strong stereo-electric control. However, the oxidation of a-D-glycosidic linkages during O3 oxidation is also possible, but slow, and is caused by several side reactions with radicals and acid hydrolysis.

3.2. Which functional groups in the NOM can act as a promoter or inhibitor of O3 decomposition? O3 may react directly with dissolved substances in water, or it may decompose to form radical species, which themselves react with these substances. This corresponds to direct oxidation by molecular O3 and indirect oxidation, respectively. The most important radical species is the OH radical, because of its high standard reduction potential (2.80 V), which is even higher than the standard reduction potential of O3 (2.07 V) (Beltra´n, 2004). In contrast to O3, OH is believed to be a non-selective oxidant, which reacts very fast with the vast majority of organic and inorganic compounds in water (von Gunten, 2003a). The decomposition of O3 in pure water is well-described in literature. The SBH (Staehelin, Bu¨hler, Hoigne´) model is widely used for predicting the lifetime of O3 in natural waters (Beltra´n, 2004). This model is represented in Fig. 5 and a brief discussion is given below. The decomposition of O3 in water is a radical-type chain reaction, where various solutes can act as initiators, promoters or inhibitors (Staehelin and Hoigne´, 1985). Initiation step: The decomposition of O3 is initiated by OH ions (reaction 1), and this leads to the formation of one
 superoxide anion ( O2 ) and one hydroperoxyl radical ( HO2), which are in an acidebase equilibrium (pKa ¼ 4.8):










O3 þ OH / O2 þ HO2


k ¼ 70 M1 s1







(1)

In addition, the reaction of unsaturated bonds in NOM with O3 can lead to the consumption of O3 (reaction 2), or the  production of an ozonide ion radical ( O3 ) by an electron transfer reaction (reaction 3):


O3 þ NOM/NOMOX

(2) 

O3 þ NOM/$ NOMþ þ O3


(3)

The direct reactions (2) and (3) between O3 and NOM are generally attributed to double bonds, electron-rich aromatic systems, amines and sulphides (von Gunten, 2003a). These direct reactions control the decomposition of O3 during the initial phase of ozonation (t < 20 s), in which very high amounts of OH radicals are generated, i.e. the concentration OH is about 106 to 108 times the concentration of O3 (von Gunten, 2003a), and the presence of radical scavengers does not exert any significant effect on the O3 consumption (Xiong et al., 1991). During the second phase (t > 20 s), when the most reactive parts of NOM have already reacted with O3, the O3 decomposition is mostly controlled by radical chain reactions instead of direct reaction with NOM, and the OH concentration is about ten times lower than during the initial phase of ozonation (Buffle and von Gunten, 2006).








3556

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 5 1 e3 5 7 0

Fig. 3 e Mechanism for the reaction between amino acids and O3. Only the aromatic amino acids tryptophan, tyrosine and the much less reactive phenylalanine (not shown) are oxidized, as well as cysteine. Probably histidine and methionine should react quite well with O3, but there was no direct evidence of this. Concerning cysteine, O3 oxidates the thiol groups with the consequent formation of disulfide bonds. Adapted from Cataldo (2003).

3557

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 5 1 e3 5 7 0

Fig. 4 e Mechanism for the reaction between polysaccharides and O3. O3 is very selective in cleaving b-D-aldosidic linkages. Ozonolysis proceeds only if the aldoside can assume a conformation in which the acetal function at the anomeric center has two of its lone-pair electron orbitals oriented trans-antiperiplanar (antiperiplanar: describing an angle between 150 and 210 ) to the alkylidene CeH bond (stereo-electric control). However, non-selective degradation is possible as well, due to the presence of OH radicals and organic acids in the ozonated solution. Adapted from Wang et al. (1999).







Propagation step: O2 is a highly selective catalyst for the decomposition of O3 in water. The rate constant with which
 O2 reacts with O3 molecules is very high and results in the
 formation of O3 :
















k ¼ 1:6$109 M1 s1

O 2 þ O3 / O 3 þ O2  O3

 O3

(4)


decomposes upon protonation into OH radicals:

þH

HO3




þ




/ HO3


k ¼ 5$10

/ OH þ O2

10

1

M

1

s




k ¼ 1:4$105 M1 s1










(5)


Even in solutions where the DOC concentration is low, such as drinking water, the OH radicals react with a solute before they encounter another radical (Staehelin and Hoigne´, 1985). Typical rate constants for reactions of OH radicals with organic solutes are in the range 106e1010 M1 s1 (Buxton et al., 1988; Lal et al., 1988; Mao et al., 1991). Some functional groups present in NOM are known to react with OH, and this can lead to the formation of carboncentered radicals (see Fig. 5, reaction 7). The reaction of these carbon-centered radicals with O2 (see Fig. 5, reaction 8)
 subsequently leads to the elimination of HO2/ O2 in a base

(6)




Fig. 5 e Reactions of aqueous O3 in the presence of NOM, which reacts with O3 or with OH radicals (scavenging or converting OH into HO2). Adapted from Staehelin and Hoigne´ (1985).








3558

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 5 1 e3 5 7 0

Fig. 6 e (a) Effect of cross-flow velocity on the permeate flux of a ceramic UF membrane treating surface water. The transmembrane pressure is 0.136 MPa and the O3 concentration in the gas phase is 1.5 g mL3 (b) Effect of transmembrane pressure on permeate flux at an O3 concentration in the gas phase of 5.5 g mL3 (except if otherwise mentioned) and a crossflow velocity of 0.47 m sL1. Adapted from Kim et al. (2008).

catalyzed reaction (see Fig. 5, reaction 9). Due to the high
 selectivity of O2 for O3, the conversion of the less ozone
 selective OH radical into O2 promotes the chain reaction. This explains how certain functional groups of NOM act as promoters for the O3 decomposition (Staehelin and Hoigne´, 1985).
 Termination step: Many other solutes do not liberate O2 radicals after reaction with OH (see Fig. 5, reaction 10). These substances are called inhibitors or radical scavengers. The question that remains is “which functional groups of NOM promote the decomposition of O3, and which functional groups act as inhibitors?” Primary alcohols (such as methanol, glycol, glycerine, or present in glucose and other carbohydrates) promote the decomposition of O3, although their direct reaction with O3 is rather slow, due to the absence of unsaturated bonds (Staehelin and Hoigne´, 1985). Aromatic rings can also be regarded as promoters for O3 decomposition, because the attack of O3 and OH cleave aromatic rings into olefins, which immediately react with O3, with the subsequent formation of H2O2 (Pi et al., 2005). Glyoxalate can act both as a promoter and initiator (Staehelin and Hoigne´, 1985). Alkyl groups and carboxylic acids (with the exception of formic acid, which is a promoter) are inhibitors (Staehelin and Hoigne´, 1985). OH can also subtract a H atom from higher alkyl alcohols than methanol, such as tert-butyl alcohol, that is not in a position a to the eOH group. These reactions are not followed by the
 formation of O2 radicals, so these compounds act as scavengers. The relative importance of scavenging to promotion for aliphatic alcohols increases with the length of the alkyl chain in these compounds, relative to the amount of H atoms in a position a to the eOH group (Staehelin and Hoigne´, 1985).











3.3.




Reaction mechanisms of NOM with OH radicals


Although the OH radical is thought to be an unselective oxidant, organic molecules with unsaturated bonds react faster with OH radicals than saturated CeC bonds or CeH bonds (Westerhoff et al., 1999). Reactions of OH radicals with olefins and aromatic molecules occur at nearly diffusioncontrolled limits (1010 M1 s1), whereas the slowest reaction rates with OH radicals were observed for aliphatic polyhalogenated compounds such as haloacetic acids (HAA) and trihalomethanes (THM) (107e108 M1 s1) (Buxton et al., 1988; Lal et al., 1988; Mao et al., 1991). OH can be regarded as an electrophilic oxidant like O3, because electron donating groups, such as hydroxyls and amines, enhance the reactivity of adjacent carbon bonds, while electron withdrawing groups, such as carboxyl groups, lessen its reactivity (Song et al., 2008). A few researchers determined the second rate constant of OH reaction with NOM (3.6 $ 108 M1 s1 (Westerhoff et al., 1999)) or humic acids (1.9 $ 107 M1 s1 (Liao and Gurol, 1995), 8.1 $ 108 M1 s1 (Westerhoff et al., 1999), 2.3e3.2 $ 109 M1 s1 (Goldstone et al., 2002)), at which the concentration of organic matter is expressed as mol C per liter. Reactions of OH radicals with organic compounds fall into two basic mechanisms, i.e. addition, generally to an aromatic ring (hydroxylation), or abstraction of a H atom. Both mechanisms can lead to the formation of low-molecular mass acids. Hydroxylation of aromatic moieties in NOM, followed by ring opening, can produce both mono- and dicarboxylic acids (Gopalan and Savage, 1994). Abstraction of H from an unsaturated CeC bond leads to the formation of a carboncentered radical, followed by reaction with O2 to form a peroxyl radical, and subsequent decomposition to a carboxylic acid (Goldstone et al., 2002).




















3559

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 5 1 e3 5 7 0




3.4. The addition of H2O2 slightly improves the mineralization of organic matter during ozonation The contribution of the radical pathway can strongly be enhanced by adding H2O2 to the solution. In this case, the main O3 decay path is the reaction between O3 and the conjugated base of H2O2, i.e. HO2 (pKa ¼ 2.24 $ 1012 (Paillard et al., 1988)), leading to a higher production rate of OH radicals (Staehelin and Hoigne´, 1982):




O3 þ HO 2 / OH þ O2 þ O2





k ¼ 2:8$106 M1 s1




(7)

This leads to an advanced oxidation process (AOP). Indeed, in pure water systems, two OH radicals are formed per three O3 molecules consumed, while in the presence of H2O2, one OH radical is formed per O3 molecule consumed (Acero and von Gunten, 2001). The main advantages of AOPs are the oxidation of ozone-resistant compounds. However, the higher formation rate of OH is at the cost of a higher O3 consumption, i.e. the decomposition of O3 is accelerated in the presence of H2O2 at pH values higher than 5 (Staehelin and Hoigne´, 1982). In addition, OH reacts rather unselective, and thus only a small fraction of these radicals reacts with the target pollutant, which severely reduces the efficiency of AOPs (von Gunten, 2003a). Several researchers reported that O3 þ H2O2 systems remove DOC better, because of the ability of OH radicals to react with the unsaturated reaction products formed during O3 oxidation. The effectiveness of OH radicals for the mineralization of these compounds can be assessed by comparing the rate constants with O3 and OH radicals (see Table 2). The rate constants for the reaction with OH radicals is evidently higher than for the reaction with molecular O3. However, the concentration of OH radicals is typically 107e109 times lower than the dissolved O3 concentration (von Gunten, 2003a). If the proportion of the two rate constants is lower than 108, an increase of the OH concentration will not be able to improve the decomposition of these saturated compounds. This proportion is calculated for each compound in the last column of Table 2, and it can be seen that it is higher than 108 for all compounds, with the exception of formate, which already reacts relatively well with O3. Goldstone et al. (2002) reported that reactions between humic substances and OH radicals produced CO2 with a high efficiency of w0.3 mol CO2 per mole OH. This efficiency remained approximately constant from the early phases of oxidation until complete mineralization of the DOC, and was not significantly altered within the range of pH values found in natural waters (pH 4e10). However, they reported that the CO2 production from OH reaction with lowmolecular mass acids is an insignificant fraction of the overall CO2 formation rate, indicating that most of the mineralization of humic substances does not occur via low-molecular mass acid intermediates. Acero and von Gunten (2001) reported that, when the organic matter content in the water is low (<1 mg L1 DOC), the addition of H2O2 considerably enhances the oxidation capacity by OH radicals. However, if the NOM concentration in natural waters is higher than 3 mg L1 DOC, O3 decomposition is controlled by the promoting or inhibiting effect of certain functional groups in the DOC on the O3 decomposition. Therefore, the effect of









































H2O2 addition in order to raise the OH concentration, will be negligible in that case. This was also observed by the following researchers, who observed that O3 þ H2O2 performed well, but only slightly better than pure O3 oxidation. Irabelli et al. (2008) reported for surface water (DOC > 3 mg L1) a DOC removal of 61% by O3 þ H2O2 (0.7 mg H2O2 was added prior to oxidation) compared to 53% by O3 (O3 dosage: 2.0 mg L1, reaction time: 13.2 min). Van Geluwe et al. (2010) found that the continuous addition of H2O2 between 10 and 30 min oxidation time (0.25 mol H2O2 per mole O3), could enhance the removal of UV absorbing compounds at 280 nm, from 42% (if no H2O2 was added) to 54%, for the treatment of humic acid solutions. If the same experimental conditions were applied to NF concentrates, obtained after filtration of surface water, a reduction of 74% could be achieved by O3 þ H2O2, compared to 68% by O3 (Van Geluwe et al., in press).

3.5. Guidelines for finding the optimal dose of H2O2 in water treatment The dosage of H2O2 has a significant effect on the mineralization of organic compounds. There exists an optimal H2O2 dosage because this compound not only acts as a generator of , OH radicals, but also as a scavenger of OH radicals at high H2O2 concentrations. This is attributed to the formation of
much less powerful HO2 radicals by the reaction between
, H2O2 and OH (reaction 8), and the subsequent reaction of HO2 , with OH (reactions 9 and 10) (Weinstein and Bielski, 1979; Buxton et al., 1988):








H2 O2 þ OH /H2 O þ HO2





HO2 þ OH /H2 O þ O2


2HO2

/H2 O2 þ O2

k ¼ 2:7$107 M1 s1 k ¼ 6:0$109 M1 s1 k ¼ 7:6$105 M1 s1




(8)




(9)




(10)

Multiple additions of H2O2 at different stages is better for DOC removal than a single addition of H2O2 at the inlet of the reactor only (Kosaka et al., 2001). Higher H2O2 doses are required at later stages, in which low reactivity compounds with OH radicals are predominant (Kosaka et al., 2001). It is also important to maintain a low concentration of dissolved O3 for efficient DOC removal, because O3 is required for the rapid formation of OH radicals (Kosaka et al., 2001). Once H2O2 is completely consumed, the AOP comes to a halt, achieving a similar COD conversion as with pure O3 oxidation (Rivas et al., 2009). The optimal H2O2/O3 ratio depends on the composition of the water solution, such as the type and concentration of the solutes (Kosaka et al., 2001), and the alkalinity of the water (Rosal et al., 2009). In most full-scale plants, the optimal ratio of the H2O2 dosage to the O3 dosage, appeared to lie between 0.25 and 1.0. There are no formulas available yet for predicting the optimal value, so it should be determined experimentally for each raw water and O3 installation (Acero and von Gunten, 2001). Paillard et al. (1988) determined the optimal conditions for removal of oxalic acid by O3 þ H2O2. Oxalic acid is a major reaction product that accumulates during O3 oxidation of natural waters. The continuous introduction of H2O2 improved the degradation of oxalic acid, compared to single addition of H2O2 at the start of the experiment. The optimal





3560

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 5 1 e3 5 7 0

pH was equal to 7.5 and the optimal dosage of H2O2 ranged between 0.6 and 0.7 $ 104 M, for an initial oxalic acid concentration of 2 $ 104 M. This corresponded to a ratio of 0.5 mol H2O2 consumed per mole O3. The presence of bicarbonates did not influence the optimal conditions, but if the solution contained bicarbonates, a significant reduction of the degradation rate for oxalic acid was observed.

3.6. A short note on the health hazard of O3 in water treatment 3.6.1.

Bromate formation

The formation of the suspected carcinogenic bromate (BrO 3) in waters containing Br is a serious concern in O3 oxidation. Br is a constituent of all natural waters and its concentration mainly depends on the geochemistry of the watershed (Magazinovic et al., 2004). It ranges from 10 to 1000 mg L1 (von Gunten, 2003b). Currently, the U.S. Environmental Protection Agency (USEPA) and the European Union have set a maximum contaminant level of 10 mg L1 for BrO 3 in drinking water (Agus et al., 2009). At Br concentrations as low as 50 mg L1, excessive BrO 3 formation can already become a problem. Details about the formation mechanisms of bromate and strategies to control it can be found in von Gunten, (2003b). A reduction of the O3 dose by a factor of two could reduce the formation of BrO 3 by a factor of more than ten (Meunier et al., 2006). However, the reduction of O3 doses decreases the inactivation of resistant protozoa, such as Cryptosporidium parvum and Giardia lamblia. The lower residual O3 concentration in AOPs and the reduction of HOBr by H2O2 reduce BrO 3 formation in O3-based AOPs in comparison to conventional O3 oxidation (von Gunten, 2003b). BrO 3 formation is noticeably higher in highly alkaline waters than in waters with lower alkalinity. This is due to the reaction of carbonate radicals with OBr. Therefore, the application of O3 oxidation has to be carefully evaluated for high-alkalinity waters in order to avoid excessive BrO 3 formation (von Gunten, 2003b). Another decisive factor is the temperature of the treated water: BrO 3 formation increases with increasing temperature (von Gunten, 2003b).

3.6.2.

Trihalomethanes and haloacetic acids

The reaction of NOM with chlorine causes the formation of a variety of potentially carcinogenic DBPs such as THMs and HAAs. The USEPA has currently set maximum contamination levels of total THMs and HAAs at 80 mg L1 and 60 mg L1, respectively. The European Union has only a limit for THMs, namely 100 mg L1 (see Agus et al., 2009 for more details). The hydrophobic NOM fraction is the main precursor of DBP formation, because it has the highest tendency to react with chlorine (Chang et al., 2002). In Section 4.1, it will be shown that O3 is able to transform these hydrophobic components into hydrophilic products, so that O3 oxidation will decrease the formation potential of THMs and HAAs quite efficiently (Chang et al., 2002; Gallard and von Gunten, 2002; Chin and Be´rube´, 2005; Li et al., 2008). Kleiser and Frimmel (2000) and Meunier et al. (2006) found that even a small O3 dose of 1.5 and 2.5 mg per mg DOC respectively, reduced the THM formation potential by about 70% in surface water, whereas almost no DOC was removed.

4. The structural changes of NOM by O3 oxidation NOM fouling is attributed to the accumulation of molecules which are retained on the membrane surface, forming a cake or gel layer, and the adsorption of non-retained molecules in the inner pores of the membrane, leading to constriction and blocking of the inner pores of the membrane. The fouling potential of NOM is defined by different types of chemical and physical interactions between the NOM and the membrane surface, such as hydrophobic interactions, hydrogen bridges and electrostatic interactions, besides the size and conformation of the NOM molecules. NOM fouling in membrane filtration is already reviewed by several authors: Zularisam et al., 2006; Al-Amoudi and Lovitt, 2007; Amy, 2008; Al-Amoudi, 2010; Sutzkover-Gutman et al., 2010. Therefore, this section will only discuss the effect of O3 oxidation on the different fouling mechanisms. A summary of recent studies on the effect of O3 oxidation on membrane fouling, both in surface water applications and municipal wastewater treatment, is given in Table 3.

4.1. Hydrophobic interactions between NOM and membrane surfaces The electrophilic nature of O3 results in the transformation of unsaturated bonds in aromatic moieties, into hydrophilic reaction products, such as carboxylic acids, as stated in the previous section. This could be confirmed by several  researchers. Swietlik and Sikorska (2004) treated sand-filtered groundwater with 1.0 mg O3 per mg DOC, for 10 min, and observed that the hydrophobic DOC (i.e. the DOC fraction retained by the Amberlite XAD 8 resin or a similar resin) decreased from 54 to 5% of the total DOC. Song et al. (2010) reported that a O3 dose of 0.5 mg O3 per mg DOC, applied for 10 min, to surface water, reduced the hydrophobic DOC by w45%, while the total DOC dropped by only 10%. Van Geluwe et al. (in press) showed that the hydrophobic fraction of the chemical oxygen demand (COD) decreased by 86%, while the overall COD decreased by only 22%, while treating the concentrate stream after NF of surface water, with 2.0 mg O3 per mg COD for 20 min. The transformation of the hydrophobic part of NOM into hydrophilic reaction products is schematically shown in Fig. 2, where a model structure of a humic acid molecule is drawn, before and after O3 oxidation. Most of the aromatic rings in the humic acid molecule are cleaved by O3, and the molecule is enriched in oxygenated functional groups. The hydrophilic reaction products have a lower propensity for adsorption on membrane surfaces. This is an important reason for the increase of the membrane flux after O3 oxidation.

4.2.

Molecular size and (dis)aggregation of NOM

4.2.1. Results obtained by high-performance size exclusion chromatography (HPSEC) The size of NOM is a critical parameter for membrane fouling. The occurrence of different modes of fouling is related to foulants size relative to membrane pore size (Katsoufidou et al., 2005; Huang et al., 2008; Lin et al., 2009). Substances close to

Table 3 e Experimental results of hybrid ozone-membrane systems. Reference

Raw water composition

Oxidation features 1

Membrane features

Michigan lake water (US) pH: 7.7e8.6 TOC: 8.6e11.6 mg L1 Alkalinity: 145e157 mg L1 CaCO3 UVA (254 nm): 0.160e0.180 cm1

CO3 (g): 1e12.5 mg L Gas flow rate: 6 L h1 Batch volume: 3.5 L

UF (TiO2) MWCO: 15 kDa Constant flux of 1820 L m2 h1 MPa1 Transmembrane pressure: 17 kPa

Kim et al. (2007)

Sewage water TOC: 163 mg L1 BOD5: 85 mg L1 Suspended solids: 203 mg L1

CO3 (g): 58 mg L1 Gas flow rate: 2 or 4 L h1, respectively Backwashing time (per 60 min): 2 or 1 min, respectively

Metallic MF membrane Mean pore size: 1 mm Pure water flux: 5000 L m2 h1 MPa1 Transmembrane pressure: 50 kPa

Kim et al. (2008)

Michigan lake water (US) TOC : 11.8 mg L1

CO3 (g): 1.5/5.5/9.5 mg L1 Gas flow rate: 12 L h1

UF (a-Al2O3 þ TiO2) MWCO: 5 kDa Pure water flux: 1430 L m2 h1 MPa1 Transmembrane pressure: 68/136/204 kPa

Lee et al. (2005b)

Municipal wastewater, influent pH: 6.8e7.2 TOC: 8.89e9.91 mg L1 COD: 105e218 mg L1 BOD5: 50e86 mg L1

For an O3 dose of 2.3 mg L1, there is a residual O3 concentration of 0.1 mg L1 in the permeate

MF (PVDF) Mean pore size: 0.22 mm Pure water flux: 414 L m2 h1

(continued on next page)

3561

Without O3 treatment, flux reduced 40% in 12 h. Stable fluxes of more than 95% of the pure water flux can be maintained if the dissolved O3 concentration at the membrane surface is greater than 0.05 mg L1. Continuous O3 injection was not necessary. Intermittent O3 injection for 1 min per 5 min gave the same results. Improved permeate flux recovery at lower pH due to higher dissolved O3 concentration The effect of air backwashing and O3 backwashing on the filtration flux was compared. The mean membrane flux (for 25 h filtration) was higher in the case of O3 backwashing, i.e. 60 L m2 h1 instead of 20 L m2 h1, in the case of 1 min backwashing per hour, and 75 L m2 h1 instead of 45 L m2 h1, in the case of 2 min backwashing per hour At a gaseous O3 concentration of 1.5 mg L1: no improvement of permeate flux, flux reduced to 40% of initial value after collecting 1 L permeate At an O3 concentration of 5.5 mg L1: flux reduced to 75% of initial value after collecting 1 L permeate and stayed stable afterward Effect of cross-flow velocity and transmembrane pressure is illustrated in Fig. 6. After 20 min, the flux reduced to 20% of the pure water flux without O3 pretreatment against 40% with O3 pretreatment. Due to ozonation, the thickness of the foulant layer was reduced by about 50% (1.4 mm instead of 3.2 mm) TOC increased through O3 pretreatment from 9.6 mg L1 to 14.5 mg L1. TOC of permeate with O3 treatment (15.6 mg L1) was higher than without O3 treatment (12.5 mg L1)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 5 1 e3 5 7 0

Karnik et al. (2005)

Results

3562

Table 3 (continued) Reference Lehman and Liu (2009)

Raw water composition Municipal wastewater, secondary effluent pH: 6.8 TOC: 3.8e5.9 mg L1 Alkalinity: 140 mg L1 CaCO3

Oxidation features O3 doses ranged between 2.7 and 6 mg L1

Membrane features MF (ceramic) Mean pore size: 0.1 mm Constant flux of 170 L m2 h1 Inline direct coagulation (1 mg L1 Al in PACl) to the ozonated water prior to membrane filtration

River water (Thailand) pH: 7.2e9.3 TOC: 10.0e46 mg L1 Conductivity: 263e1781 mS cm1

CO3 (g): 0.3e3.1 mg L1 Gas flow rate: 500 L h1 Ozonation was carried out using a by-pass flow (2 m3 h1) into which gaseous O3 was fed via a venturi injector

MF (ceramic) Mean pore size: 200 nm UF (ceramic) Mean pore size: 80 nm Transmembrane pressure: 30.0 kPa

Schlichter et al. (2004)

Saar river water (Germany) pH: 6.8e7.3 TOC: 3.5e4.5 mg L1 COD: 2.7e3.7 mg L1

O3 dose: 1.7/2.8/4.5 g O3 g1 TOC Raw water feed flow: 1 m3 h1 Gas flow rate: 300 L h1

UF (ceramic) MWCO: 20 kDa Pure water flux: 2100 L m2 h1 MPa1

Song et al. (2010)

Huangpu river water (China) pH: 7.2e7.6 TOC: 5.1e6.6 mg L1 UVA (254 nm): 0.128e0.157 cm1

O3 doses ranged between 0.5 and 3.0 mg L1 Oxidation time: 10 min

MF (PVDF) Pore size: 0.1 mm Hollow fiber

Membrane flux can be stabilized for long-term operation at an O3 dose that is slightly above the instantaneous O3 demand. The results are summarized in Fig. 7. O3 dose (mg L1)

2.7 3.5

5

6

CO3 in permeate (mg L1) <0.05 0.15 0.61 0.89 Normalized permeate flux after 36 70 82 98 5 days (%) The membranes were backflushed by O3 at intervals of 150 min for 2 min, at 0.600 bar (MF) or 0.900 bar (UF) excess pressure. The main results are illustrated in Fig. 8. Without O3 treatment, the fouling layer is tightly fixed to the membrane, which is the reason why backflushing without chemicals is unsuccessful (black points). However, if the surface is continuously ozonated during filtration with an O3 residual in the permeate, simple backflushing is successful in controlling fouling (gray points). An O3 concentration of 0.05 mg L1 (1.7 g O3 g1 TOC) in the permeate is required to maintain stable and high permeate fluxes. 0.0 1.7 2.8 4.5 O3 dose (g O3 g-1TOC) 0.00 0.05 0.10 0.40 CO3 in permeate (mg L1) 80 128 149 194 Membrane permeability (L m2 h1 bar1) Similar results for the ceramic MF membrane (0.1 mm) Maximal membrane flux at a dose of 1.5 mg O3 L1. At the end of the filtration (after collecting 800 mL permeate), the membrane flux was 37% of the pure water flux, with O3 oxidation, while only 20% without O3 oxidation of the feed.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 5 1 e3 5 7 0

Sartor et al. (2008)

Results

3563

UF (PVDF) MWCO: 50 kDa Pure water flux: 675 L m2 h1 Transmembrane pressure: 0.1 MPa Tertiary effluent from industrial wastewater plant in Taiwan (mostly dying and electronic wastewater) pH: 6.53 TOC: 6.51 mg L1 You et al. (2007)

O3 dose: 1.76 mg L1 min1 residual aqueous O3 concentration in permeate: 4.02 mg L1

UF (PVDF) MWCO: 30 kDa Transmembrane pressure: 100 kPa Secondary effluent TOC: 8e10 mg L1 COD: 10.8e41.2 mg L1 BOD5: 6e12 mg L1 UVA (254 nm): 0.106e0.163 cm1 Suspended solids: 10e20 mg L1 Wang et al. (2007)

CO3 (g): 20 mg L1 Gas flow rate: 1 L min1 Oxidation time: 1e20 min Batch volume: 0.5 L

NF concentrate after filtering Dijle river water (Belgium) pH: 7.7e8.4 COD: 43.7  2.2 mg L1 UVA (254 nm): 0.222  0.008 cm1 Van Geluwe et al. (in press)

CO3 (g): 12.2  0.4 mg L1 Gas flow rate: 1 L min1 Oxidation time: 10 min Batch volume: 3 L

Four NF membranes (NF 90, NF 270, Desal 51 HL, NF-PES 10)

The average permeability of the concentrate solution throughout the filtration (40 h), increased from 23 to 74% of the pure water permeability with NF 90 due to O3 oxidation, from 73 to 88% with NF 270, from 63 to 76% with NF-PES 10, and from 69 to 77% with Desal 51 HL. The membrane flux is shown in function of the O3 oxidation time. The membrane flux increased from 80 to 140 L m2 h1, for a O3 contact time of 1 min. When the O3 contact time increased, the membrane flux did not change appreciably. Without O3 oxidation, the permeate flux dropped to 60% after 1 h, but with O3 oxidation, it could be maintained at 90% throughout the test of 2 h. TOC did not decrease during O3 oxidation.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 5 1 e3 5 7 0

the size of the membrane pores can cause pore blocking, which provokes a severe increase of the filtration resistance (Lohwacharin and Takizawa, 2009). This is the case for the high-molecular mass molecules, such as polysaccharides and proteins, constricting and blocking the pores in loose membranes (MF and UF). Substances much larger than the membrane pores lead to cake formation, that is more readily removed. This is the case for the vast majority of the NOM components with tight membranes (NF and RO). In this case, substances with a high-molecular mass that are retained by the membrane, have a small back diffusion rate during crossflow filtration, so that a thicker cake or gel layer is formed on the membrane surface (Ang et al., 2006). Various researchers observed that ozonation is able to decompose NOM molecules into smaller fragments. HPSEC analysis before and after ozonation, shows a clear shift toward longer retention times, corresponding to a decrease in the average molecular mass. This is observed for synthetic solution of humic substances (Kerc et al., 2004; These and Reemtsma, 2005; Van Geluwe et al., 2009), as well as for natural waters and wastewater effluents (Nissinen et al., 2001; Nawrocki et al., 2003; Lehman and Liu, 2009; Song et al., 2010). Similar observations were made in studies where the NOM reacted with OH radicals instead of molecular O3 (Le-Clech et al., 2006; Liu et al., 2008). The cross-flow operation can flush away loose fragments of the cake layer. This decreases the thickness of the foulant layer and thus reduces the flux decline (Karnik et al., 2005; Lee et al., 2005b). Jansen et al. (2006) investigated the molecular mass distribution of concentrated solutions of humic substances with a low polydispersity (1.23) during O3 oxidation. The average molecular mass shifted from 2600 to 1500 g mol1 and the polydispersity of the peak decreased remarkably from 1.23 to 1.09, after a O3 consumption of 0.65 mol O3 per mol C. Another peak emerged at longer retention times, because small compounds were formed during the ozonation process. The size of the reaction products was restricted to only two distinct peaks throughout the reaction. In almost all experiments, the shape of the original peak did not broaden toward longer elution times. This indicates that no random splitting of the humic molecules occurs, since that would result in a very broad range of molecular sizes. Only small molecular fragments are split off from the periphery of the larger humic molecules. The main structure is reduced in size during ozonation, but remains intact. Cleavage occurs mainly at the periphery of the molecules. Steric obstruction related to the coiled structure of the humic substances, prevents O3 and OH radicals from cleaving bonds in the core of the molecules (Jansen et al., 2007; Nawrocki, 2007).





4.2.2. Aggregation of humic substances by calcium and magnesium The size of the humic substances is also influenced by aggregation/disaggregation forces, which in turn depend on the pH of the solution. For a long time, humic substances have been regarded as polymeric molecules, having a relatively highmolecular mass. Molecular masses of several 10,000 g mol1 have been reported (Sutton and Sposito, 2005). Nowadays, humic substances are regarded as supramolecular associations of many, relatively small (<1000 g mol1) and chemically diverse bio-organic molecules, that are hold together by different kinds of weak interactions, such as Van der Waals forces and hydrogen

3564

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 5 1 e3 5 7 0

bridges (Piccolo, 2002; Sutton and Sposito, 2005; Ahn et al., 2008). Therefore, it is appropriate to study the aggregation kinetics of humic substances, and the effect of O3 oxidation on it. Aggregation is promoted by the presence of divalent inorganic cations, such as calcium and magnesium. These ions can bind to carboxylate groups present in certain fractions of the NOM (Hong and Elimelech, 1997; Ahn et al., 2008). Charge screening by the adsorbed cations decreases the net negative charge of the NOM molecules, and thus the electrostatic repulsions with the membrane surface (Engebretson and von Wandruszka, 1998). This phenomena can be observed even at low concentrations of divalent ions, i.e. lower than 0.5 $ 103 M (Shao et al., 2011). If the concentration of divalent ions is higher than 0.5 $ 103 M, these cations act as bridges between carboxylic groups of two adjacent molecules, increasing the attractive forces between them (Shao et al., 2011). This bridging function of divalent cations has been proven with atomic force microscopy (AFM) (Li and Elimelech, 2004; Nguyen and Chen, 2007). This phenomenon was observed for all NOM fractions, but was especially significant in the presence of hydrophobic NOM, i.e. humic substances (Makdissy et al., 2002). Due to the formation of complexes between NOM and divalent ions, the size of the humic aggregates becomes larger (Shao et al., 2011). It is important to note that these conclusions are all obtained without the effect of O3 treatment on NOM. The impact of divalent ions on the complexation of ozonated NOM is not investigated yet, but it is an important issue for further study. An increase of the number of carboxylic acids in NOM after O3 oxidation provides more ligands for complexation, and consequently the negative charge of NOM decreases. A decrease in NOM charge is expected to promote the adsorption of NOM on negatively charged membranes. However, Song et al. (2004) reported that the calcium content of foulants from oxidized surface waters (with UV þ H2O2) is lower than that from raw water (0.3% vs. 0.8%). Presently, there is little agreement on the effect of complexation on membrane fouling. Shao et al. (2011) stated that the porous character of the flocculated aggregates enhances the permeability of the cake layer formed on the membrane surface, and a smaller flux decline was observed as the calcium concentration increased from 0 to 0.5 mM (see Fig. 9). The aggregate porosity (eagg) is related to the primary particle radius (rprim), aggregate radius (ragg), and the fractal dimension D (Jiang and Logan, 1991):  eagg ¼ 1 

ragg rprim

Fig. 7 e Influence of the O3 dose on the flux of a ceramic membrane (nominal pore size: 0.1 mm) treating secondary wastewater effluent. The pure water flux of the membrane is 12,800 L mL2 hL1 MpaL1. The feed water was ozonated and 1e3.5 mg LL1 Al of polyaluminiumchloride was added to the ozonated water before it entered the membrane module. Adapted from Lehman and Liu (2009).

In addition, the aggregates can act as a secondary membrane, which removes a significant portion of the small particles, before they reach the polymeric membrane surface. The removal of the small particles prevents the obstruction of membrane pores or the formation of a tightly packed cake layer on the membrane surface (Kuberkar and Davies, 2000). Membrane fouling will be strongly alleviated if NOM aggregation causes a change of the fouling mechanisms, i.e. from internal pore constriction to the formation of a permeable cake on the membrane surface (Lohwacharin and Takizawa, 2009). Other researchers stated that the complexation causes severe flux decline, because the cross-linked aggregates form a dense, compact cake layer on the membrane surface, with a high fouling resistance (Hong and Elimelech, 1997; Yuan and

D3

Fractal dimensions are related to the aggregate structure, e.g. D ¼ 3 for a sphere, D ¼ 2 for a flat sheet, D z 1.8e2.4 for typical aggregates (Waite et al., 1999). Indeed, the equation shows that, for any other shape than perfect spheres, the aggregate porosity increases as ragg/rprim becomes larger. However, this trend is reversed at higher calcium concentrations (see Fig. 9). Shao et al. (2011) suggested that the higher flux decline is caused by the formation of calcium bridges between the NOM and the membrane surface. This explanation is questionable, because the authors observed the same trend for neutral membranes.

Fig. 8 e Permeate fluxes for a ceramic UF membrane (mean pore size: 80 nm) treating surface water. The pure water flux of the membrane is 1600 L mL2 hL1 MPaL1. Fouling is essentially caused by the formation of a biofilm. The membrane flux can be increased considerably by adding O3 (20e100 mg LL1 O3) prior to the membrane module. The suction pressure is 30.0 kPa. Membrane fouling can be largely removed with each backflushing cycling (gray points), in contrast to the case without O3 pretreatment (black points). Adapted from Sartor et al. (2008).

3565

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 5 1 e3 5 7 0

Fig. 9 e Influence of the calcium concentration on the membrane flux during filtration of surface water. Adapted from Shao et al. (2011).

Zydney, 1999; Cho et al., 1999, 2000; Scha¨fer et al., 2002; Costa et al., 2006; Tang et al., 2007; Katsoufidou et al., 2008; Mattaraj et al., 2008; Zhu et al., 2010), or internal deposition of aggregates reduces the pore size of the membrane (Aoustin et al., 2001).

4.3.

Electrostatic interactions and hydrogen bridges

The surface of many membranes carries a negative charge at neutral pH, because of the presence of carboxylic and amine groups (Childress and Elimelech, 1996). Electrostatic repulsions between the negatively charged groups in NOM and the membrane surface decreases the fouling tendency. For instance, Shao et al. (2011) observed a lower flux decline when humic acids were filtered by a charged regenarated cellulose membrane, in comparison to a neutral membrane of the same material. Due to the absence of ionizable groups in polysaccharides, electrostatic interactions do not play a role for these molecules. Proteins are nearly all negatively charged at neutral pH, because of the preponderance of weakly acid residues in almost all proteins. The isoelectric point (IEP) of natural proteins lies in many cases between 5 and 6 (Scopes, 1994). The presence of carboxylic (pKaw3e5) and phenolic (pKaw8e11) acid groups in humic substances, make them

anionic at neutral pH (Buffle, 1991; Milne et al., 2001). The mole fractions of either carboxylic or phenolic sites in humic substances are usually markedly unequal, with carboxylic acids typically dominating (Buffle, 1991). For instance, Suwannee River humic acids contain 7.85 mol carboxylic groups per kg (dry weight), while only 1.86 mol phenolic groups per kg (IHSS, 2001). The number of carbonyl and carboxyl groups increases after O3 oxidation, as stated previously. Carbonyl compounds carry a negative charge at neutral pH, which is repulsed by the negatively charged membrane. However, carboxyl groups can form strong hydrogen bridges with hydrophilic membranes, such as polyamide and regenerated cellulose membranes. Van Geluwe et al. (in press) recorded Fourier transform infrared (FTIR) spectra of the NF 270 membrane, after 40 h filtration of concentrated surface water. He observed that carbonyl compounds were able to adsorb onto the membrane, if the solution was treated with O3 (O3 dosage: 1.0 mg O3 per mg COD; oxidation time: 10 min). However, the membrane had a significant lower fouling rate after O3 oxidation, although the COD of the solution did not decrease. The mean water flux, during the first 40 h filtration, increased from 73 to 88% of the pure water flux. An estimation of the strength of these hydrogen bridges is provided by Nie et al. (2005). They calculated the strength of hydrogen bridges between the carboxyl group (of butyric acid) and methanol, via ab initio methods. The methanol molecule can be considered as an alcoholic group on the membrane surface. Table 4 shows the dissociation energies of several hydrogen bridges, with water as solvent. The notation, HOeC]O, represents that the hydroxyl O atom of the carboxyl group forms a hydrogen bridge with an alcoholic group on the membrane surface. The strength of the hydrogen bridges is compared to the electrostatic repulsion energy between two negative charges, e.g. a carboxylate group in the NOM and a negative charge on the membrane surface. The calculated hydrogen bridge strengths strongly depend on the specific type of hydrogen bridges. The formation of hydrogen bridges between carboxyl groups and alcoholic groups are all exothermic, with the exception of HOeC]O. The hydrogen bridges are relatively weak, and weaker than those formed between water molecules (11 kJ mol1). The strongest hydrogen bridges are formed with the acid hydrogen of the carboxyl group (HOeC]O). This type of bond does not occur frequently at neutral pH, because most

Table 4 e Dissociation energies, bond lengths and the distance dependence of the dissociation energies, for hydrogen bridges and electrostatic bonds of a COOH group in water. Adapted from Nie et al. (2005). Bond length (pm)

Dependence of the dissociation energy on the separation distance r

0

Infinity

N/A

5.6 1.5 6.7 2.4  2 3.9

294 287 275 265/275 275

1/r3 1/r3 1/r3 1/r3 1/r

Dissociation energy in water (kJ mol1) No interaction Type of H bridge HOeC]O HOeC]O HOeC]O HOeC]O Electrostatic repulsion between two elementary charges

3566

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 5 1 e3 5 7 0

carboxyl groups are dissociated at this pH. The other hydrogen bridges are weaker than the electrostatic repulsion forces between two elementary negative charges. These results show that the hydrogen bridges between the carboxylate groups in the NOM, formed during O3 oxidation, with the membrane surface are in most cases relatively weak, and are of comparable strength as the electrostatic repulsion forces between charges with the same sign. Whether the increase of the number of carboxylic groups by O3 oxidation promotes the adsorption of NOM on the membrane surface or not, is an open question at this moment, and it would be interesting to study this in the future.

5.

Conclusion

Acknowledgments

The electrophilic character of O3 accounts for the fast reaction of O3 molecules with unsaturated bonds. The possible reaction pathways between NOM and O3 were described within the framework of the SBH model (Staehelin, Bu¨hler, Hoigne´). The direct reaction of unsaturated bonds in NOM with O3 can lead to the consumption of O3, or the production of an ozonide
 ion radical ( O3 ), which decomposes upon protonation into a hydroxyl ( OH) radical. The OH radical is a strong oxidizing agent that can react with NOM molecules. This is referred to as the indirect oxidation pathway. Although the OH radical is thought to be an unselective oxidant, it can be regarded as electrophilic oxidant. These direct reactions between NOM and O3 control the decomposition of O3 during the initial phase of ozonation (t < 20 s), in which very high amounts of OH are generated, i.e. the concentration OH is about 106 to 108 times the concentration of O3, and the presence of radical scavengers does not exert any significant effect on the O3 consumption. During the second phase (t > 20 s), the most reactive moieties of NOM have reacted with O3, so that O3 decomposition is mostly controlled by a radical chain reaction, and the OH concentration is about ten times lower than during the initial phase. During the second phase, some functional groups in the NOM, such as aromatic rings and primary alcohols (abundant in polysaccharides), act as promoters for the decomposition of O3. Alkyl groups and carboxylic acids act as scavengers of OH radicals. Although O3 oxidation without addition of H2O2 is not considered as an AOP, the concentration of OH radicals during ozonation of NOM solutions, can be as high as in AOPs at neutral pH (>>1012 M). The contribution of the radical pathway can strongly be enhanced by adding H2O2 to the solution. The addition of H2O2 improves the mineralization of the saturated reaction products, mostly carboxylic acids, that are formed during O3 oxidation. Several researchers proved that the application of O3 oxidation of the feed water prior to membrane filtration, resulted in a significant decrease in membrane fouling, although only a minor DOC removal (10e20%) could be achieved. This is explained by the fact that O3 causes substantial structural changes to the NOM present in the feed water, of which the most important are:


- A significant increase of the number of carboxylic functions, which are repelled by the negative membrane surface. These repulsion forces have a comparable strength as the hydrogen bridges that carboxylic groups can form with the membrane surface. - Decomposition of molecules into smaller fragments, whereby small molecular fragments are split off from the periphery of the larger molecules that remain intact. - A higher propensity for complexation of humic substances with divalent ions, if the concentration of divalent ions is larger than 0.5 $ 103 M. However, it is important to note that this figure is obtained with non-ozonated NOM.

Steven Van Geluwe is grateful to IWT-Vlaanderen (Institute for the Promotion of Innovation by Science and Technology in Flanders) for providing a fellowship.

references






















- An efficient removal of aromatic rings, which decreases the adsorption of NOM by hydrophobic interactions.

Acero, J.L., von Gunten, U., 2001. Characterization of oxidation processes: ozonation and the AOP O3/H2O2. J. Am. Water Works Ass 93 (10), 90e100. Agus, E., Voutchkov, N., Sedlak, D.L., 2009. Disinfection by-products and their potential impact on the quality of water produced by desalination systems: a literature review. Desalination 237, 214e237. Ahn, W.Y., Kalinichev, A.G., Clark, M.M., 2008. Effects of background cations on the fouling of polyethersulfone membranes by natural organic matter: experimental and molecular modeling study. J. Membr. Sci. 309, 128e140. Ang, W.S., Lee, S., Elimelech, M., 2006. Chemical and physical aspects of cleaning of organic-fouled reverse osmosis membranes. J. Membr. Sci. 272, 198e210. Al-Amoudi, A.S., 2010. Factors affecting natural organic matter (NOM) and scaling fouling in NF membranes: a review. Desalination 259, 1e10. Al-Amoudi, A., Lovitt, R.W., 2007. Fouling strategies and the cleaning system of NF membranes and factors affecting cleaning efficiency. J. Membr. Sci. 303, 4e28. Amy, G., 2008. Fundamental understanding of organic matter fouling of membranes. Desalination 231, 44e51. Aoustin, E., Scha¨fer, A.I., Fane, A.G., Waite, T.D., 2001. Ultrafiltration of natural organic matter. Sep. Purif. Technol. 22e23, 63e78. Beltra´n, F.J., 2004. Ozone Reaction Kinetics for Water and Wastewater Systems. CRC Press, Boca Raton (FL, USA). Buffle, J., 1991. Complexation Reactions in Aquatic Systems: An Analytical Approach. Ellis Horwood, Chichester (United Kingdom). Buffle, M.O., von Gunten, U., 2006. Phenols and amine induced HO generation during the initial phase of natural water ozonation. Environ. Sci. Technol. 40, 3057e3063. 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 in aqueous solutions. J. Phys. Chem. Ref. Data 17, 513e886. Can, Z.S., Gurol, M., 2003. Formaldehyde formation during ozonation of drinking water. Ozone Sci. Engin 25 (1), 41e51. Cataldo, F., 2003. On the action of ozone on proteins. Polym. Degrad. Stabil. 82, 105e114.


w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 5 1 e3 5 7 0

Chang, C.N., Ma, Y.S., Zing, F.F., 2002. Reducing the formation of disinfection by-products by pre-ozonation. Chemosphere 46, 21e30. Childress, A.E., Elimelech, M., 1996. Effect of solution chemistry on the surface charge of polymeric reverse osmosis and nanofiltration membranes. J. Membr. Sci. 119, 253e268. Chin, A., Be´rube´, P.R., 2005. Removal of disinfection by-product precursors with ozoneeUV advanced oxidation processes. Water Res. 39, 2136e2144. Cho, J., Amy, G., Pellegrino, J., 1999. Membrane filtration of natural organic matter: initial comparison of rejection and flux decline characteristics with ultrafiltration and nanofiltration membranes. Water Res. 33, 2517e2526. Cho, J., Amy, G., Pellegrino, J., 2000. Membrane filtration of natural organic matter: factors and mechanisms affecting rejection and flux decline with charged ultrafiltration (UF) membrane. J. Membr. Sci. 164, 89e110. Costa, A.R., de Pinho, M.N., Elimelech, M., 2006. Mechanisms of colloidal natural organic matter fouling in ultrafiltration. J. Membr. Sci. 281, 716e725. Darwish, M.A., Al-Najem, N.M., 2000. Energy consumption by multi-stage flash evaporation and reverse osmosis desalters. Appl. Therm. Eng. 20, 399e416. de la Rubia, A., Rodrı´guez, M., Leo´n, V.M., Prats, D., 2008. Removal of natural organic matter and THM formation potential by ultraand nanofiltration of surface water. Water Res. 42, 714e722. Eltawil, M.A., Zhengming, Z., Yuan, L., 2009. A review of renewable energy technologies integrated with desalination systems. Renew. Sust. Energ. Rev. 13, 2245e2262. Engebretson, R.R., von Wandruszka, R., 1998. Kinetic aspects of cation-enhanced aggregation in aqueous humic acids. Environ. Sci. Technol. 32, 488e493. Frimmel, F.H., Abbt-Braun, G., Heumann, K.G., Hock, B., Luedemann, H.D., 2002. Refractory Organic Substances in the Environment. Wiley VCH, Weinheim (Germany). Gallard, H., von Gunten, U., 2002. Chlorination of natural organic matter: kinetics of chlorination and of THM formation. Water Res. 36, 65e74. Goldstone, J.V., Pullin, M.J., Bertilsson, S., Voelker, B.M., 2002. Reactions of hydroxyl radical with humic substances: bleaching, mineralization and production of bioavailable carbon substrates. Environ. Sci. Technol. 36, 364e372. Gopalan, S., Savage, P.E., 1994. Reaction mechanisms for phenol oxidation in supercritical water. J. Phys. Chem. Phys. 98, 12646e12652. Hammes, F., Salhi, E., Ko¨ster, O., Kaiser, H.P., Egli, T., von Gunten, U., 2006. Mechanistic and kinetic evaluation of organic disinfection by-product and assimilable organic carbon (AOC) formation during the ozonation of drinking water. Water Res. 40, 2275e2286. Her, N., Amy, G., McKnight, D., Sohn, J., Yoon, Y., 2003. Characterization of DOM as a function of MW by fluorescence EEM and HPLC-SEC using UVA, DOC and fluorescence detection. Water Res. 37, 4295e4303. Hong, S., Elimelech, M., 1997. Chemical and physical aspects of natural organic matter (NOM) fouling of nanofiltration membranes. J. Membr. Sci. 132, 159e181. Hoigne´, J., Bader, H., 1983a. Rate constants of reactions of ozone with organic and inorganic compounds in water - I: Nondissociating organic compounds. Water Res. 17, 173e183. Hoigne´, J., Bader, H., 1983b. Rate constants of reactions of ozone with organic and inorganic compounds in water - II: Dissociating organic compounds. Water Res. 17, 185e194. Huang, H., Spinette, R., O’Melia, C.R., 2008. Direct-flow microfiltration of aquasols: I. Impacts of particle stabilities and size. J. Membr. Sci. 314, 90e100. I.H.S.S., 2001. Standard and Reference Collection. International Humic Substance Society.

3567

Irabelli, A., Jasim, S., Biswas, N., 2008. Pilot-scale evaluation of ozone vs. peroxone for trihalomethane formation. Ozone Sci. Engin 30, 356e366. Jansen, S.A., Malaty, M., Nwabara, S., Johnson, E., Ghabbour, E., Davies, G., Varnum, J.M., 1996. Structural modeling in humic acids. Mater. Sci. Eng. C 4, 175e179. Jansen, R.H.S., Zwijnenburg, A., van der Meer, W.G.J., Wessling, M., 2006. Outside-in trimming of humic substances during ozonation in a membrane contactor. Environ. Sci. Technol. 40, 6460e6465. Jansen, R.H.S., Zwijnenburg, A., van der Meer, W.G.J., Wessling, M., 2007. Response to comment on “Outside-in trimming of humic substances during ozonation in a membrane contactor”. Environ. Sci. Technol. 41, 5162e5164. Jiang, Q., Logan, B.E., 1991. Fractal dimensions of aggregates determined from steady-state size distributions. Environ. Sci. Technol. 25, 2031e2038. Karnik, B.S., Davies, S.H.R., Chen, K.C., Jaglowski, D.R., Baumann, M.J., Masten, S.J., 2005. Effects of ozonation on the permeate flux of nanocrystalline ceramic membranes. Water Res. 39, 728e734. Katsoufidou, K., Yiantsios, S.G., Karabelas, A.J., 2005. A study of ultrafiltration membrane fouling by humic acids and flux recovery by backwashing: experiments and modeling. J. Membr. Sci. 266, 40e50. Katsoufidou, K., Yiantsios, S.G., Karabelas, A.J., 2008. An experimental study of UF membrane fouling by humic acid and sodium alginate solutions: the effect of backwashing on flux recovery. Desalination 220, 214e227. Kennedy, M.D., Chun, H.K., Quintanilla Yangali, V.A., Heijman, B. G.J., Schippers, J.C., 2005. Natural organic matter (NOM) fouling of ultrafiltration membranes: fractionation of NOM in surface water and characterisation by LC-OCD. Desalination 178, 73e83. Kerc, A., Bekbolet, M., Saatci, A.M., 2004. Effects of oxidative treatment techniques on molecular size distribution of humic acids. Water Sci. Technol. 49 (4), 7e12. Kim, J.O., Jung, J.T., Yeom, I.T., Aoh, G.H., 2007. Effect of fouling reduction by ozone backwashing in a microfiltration system with advanced new membrane material. Desalination 202, 361e368. Kim, J., Davies, S.H.R., Baumann, M.J., Tarabara, V.V., Masten, S.J., 2008. Effect of ozone dosage and hydrodynamic conditions on the permeate flux in a hybrid ozonation-ceramic ultrafiltration system treating natural waters. J. Membr. Sci. 311, 165e172. Kimura, K., Amy, G., Drewes, J.E., Heberer, T., Kim, T.U., Watanabe, Y., 2003. Rejection of organic micropollutants (disinfection by-products, endocrine disrupting compounds, and pharmaceutically active compounds) by NF/RO membranes. J. Membr. Sci. 227, 113e121. Kleiser, G., Frimmel, F.H., 2000. Removal of precursors for disinfection by-products (DBPs) e differences between ozoneand OH-radical-induced oxidation. Sci. Total Environ. 256, 1e9. Kosaka, K., Yamada, H., Shishida, K., Echigo, S., Minear, R.A., Tsuno, H., Matsui, S., 2001. Evaluation of the treatment performance of a multistage ozone/hydrogen peroxide process by decomposition by-products. Water Res. 35, 3587e3594. , K., Kunst, B., 2002. RO and NF membrane fouling and Ko sutic cleaning and pore size distribution variations. Desalination 150, 113e120. Kuberkar, V.T., Davies, R.H., 2000. Modeling of fouling reduction by secondary membranes. J. Membr. Sci. 168, 243e258. Kubicki, J.D., Apitz, S.E., 1999. Models of natural organic matter and interactions with organic contaminants. Org. Geochem. 30, 911e927.

3568

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 5 1 e3 5 7 0

Kusakabe, K., Aso, S., Hayashi, J.I., Isomura, K., Morooka, S., 1990. Decomposition of humic acid and reduction of trihalomethane formation potential in water by ozone with UV irradiation. Water Res. 24, 781e785. Lal, M., Scho¨neich, C., Mo¨nig, J., Asmus, K.D., 1988. Rate constants for the reactions of halogenated organic radicals. Int. J. Radiat. Biol. 54, 773e785. Le-Clech, P., Lee, E.K., Chen, V., 2006. Hybrid photocatalysis/ membrane treatment for surface waters containing low concentrations of natural organic matters. Water Res. 40, 323e330. Lee, N., Amy, G., Croue´, J.P., Buisson, H., 2004. Identification and understanding of fouling in low-pressure membrane (MF/UF) filtration by natural organic matter (NOM). Water Res. 38, 4511e4523. Lee, N., Amy, G., Croue´, J.P., Buisson, H., 2005a. Morphological analysis of natural organic matter (NOM) fouling of lowpressure membranes (MF/UF). J. Membr. Sci. 261, 7e16. Lee, S., Lee, K., Mohtar Wan, W., Choi, Y., 2005b. Comparison of membrane permeability and a fouling mechanism by pre-ozonation followed by membrane filtration and residual ozone in membrane cells. Desalination 178, 287e294. Lee, L.Y., Ng, H.Y., Ong, S.L., Hu, J.Y., Tao, G., Kekre, K., Viswanath, B., Lay, W., Seah, H., 2009. Ozone-biological activated carbon as a pretreatment process for reverse osmosis brine treatment and recovery. Water Res. 43, 3948e3955. Lehman, S.G., Liu, L., 2009. Application of ceramic membranes with pre-ozonation for treatment of secondary wastewater effluent. Water Res. 43, 2020e2028. Leiknes, T., 2009. The effect of coupling coagulation and flocculation with membrane filtration in water treatment: a review. J. Environ. Sci. 21, 8e12. Leppard, G.G., 1997. Colloidal organic fibrils of acid polysaccharides in surface waters: electron-optical characteristics, activities and chemical estimates of abundance. Colloids Surf. A: Physicochem. Eng. Aspects 120, 1e15. Li, Q., Elimelech, M., 2004. Organic fouling and chemical cleaning of nanofiltration membranes: measurements and mechanisms. Environ. Sci. Technol. 38, 4683e4693. Li, J., Liu, H., Zhao, X., Qu, J., Liu, R., Ru, J., 2008. Effect of preozonation on the characteristic transformation of fulvic acid and its subsequent trichloromethane formation potential: presence or absence of bicarbonate. Chemosphere 71, 1639e1645. Liao, C.H., Gurol, M.D., 1995. Chemical oxidation by photolytic decomposition of hydrogen peroxide. Environ. Sci. Technol. 29, 3007e3014. Lin, C.F., Lin, A.Y.C., Sri Chandana, P., Tsai, C.Y., 2009. Effects of mass retention of dissolved organic matter and membrane pore size on membrane fouling and flux decline. Water Res. 43, 389e394. Liu, S., Lim, M., Fabris, R., Chow, C., Chiang, K., Drikas, M., Amal, R., 2008. Removal of humic acid using TiO2 photocatalytic process e fractionation and molecular weight characterisation studies. Chemosphere 72, 263e271. Lohwacharin, J., Takizawa, S., 2009. Effects of nanoparticles on the ultrafiltration of surface water. J. Membr. Sci. 326, 354e362. Lowe, J., Hossain, Md.M, 2008. Application of ultrafiltration membranes for removal of humic acid from drinking water. Desalination 218, 343e354. Magazinovic, B., Nicholson, B.C., Mulcahy, D.E., Davey, D.E., 2004. Bromide levels in natural waters: its relationship to levels of both chloride and total dissolved solids and the implications for water treatment. Chemosphere 57, 329e335. Makdissy, G., Croue´, J.P., Amy, G., Lee, N., Habarou, H., Buisson, H, 2002. Advance in the knowledge of the fouling of an

ultrafiltration membrane. In: Proceedings of the AWWA Water Quality and Technology Conference, Seattle. Mao, Y., Scho¨neich, C., Asmus, K.D., 1991. Identification of organic acids and other intermediates in oxidative degradation of chlorinated ethanes on TiO2 surfaces en route to mineralization. A combined photocatalytic and radiation chemical study. J. Phys. Chem. 95, 10080e10089. Mattaraj, S., Jarusutthirak, C., Jiraratananon, R., 2008. A combined osmotic pressure and cake filtration model for crossflow nanofiltration of natural organic matter. J. Membr. Sci. 322, 475e483. McDonald, S., Bishop, A.G., Prenzler, P.D., Robards, K., 2004. Analytical chemistry of freshwater humic substances. Anal. Chim. Acta 527, 105e124. Meunier, L., Canonica, S., von Gunten, U., 2006. Implications of sequential use of UV and ozone for drinking water quality. Water Res. 40, 1864e1876. Meylan, S., Hammes, F., Traber, J., Salhi, E., von Gunten, U., Pronk, W., 2007. Permeability of low molecular weight organics through nanofiltration membranes. Water Res. 41, 3968e3976. Milne, C.J., Kinniburgh, D.G., Tipping, E., 2001. Generic NICA-Donnan model parameters for proton binding by humic substances. Environ. Sci. Technol. 35, 2049e2059. Nawrocki, J., 2007. Comment on “Outside-in trimming of humic substances during ozonation in a membrane contactor”. Environ. Sci. Technol. 41, 5161.  Nawrocki, J., Swietlik, J., Raczyk-Stanis1awiak, U., Da˛browska, A., Bi1ozor, S., Ilecki, W., 2003. Influence of ozonation conditions on aldehyde and carboxylic acid formation. Ozone Sci. Eng. 25, 53e62. Nghiem, L.D., Scha¨fer, A.I., Elimelech, M., 2004. Removal of natural hormones by nanofiltration membranes: Measurement, modeling, and mechanisms. Environ. Sci. Technol. 38, 1888e1896. Nguyen, T.H., Chen, K.L., 2007. Role of divalent cations in plasmid DNA adsorption to natural organic matter-coated silica surface. Environ. Sci. Technol. 41, 5370e5375. Nie, B., Stutzman, J., Xie, A., 2005. A vibrational spectral maker for probing the hydrogen-bonding status of protonated Asp and Glu residues. Biophys. J. 88, 2833e2847. Nissinen, T.K., Miettinen, I.T., Martikainen, P.J., Vartiainen, T., 2001. Molecular size distribution of natural organic matter in raw and drinking waters. Chemosphere 45, 865e873. Oh, B.S., Park, S.J., Lee, H.G., Kim, K.S., Lee, K.H., Kang, J.W., 2003. Application of ozone/UV process for the reclamation of sewage treatment plant effluent. J. Water Environ. Technol. 1, 141e153. Ouyang, L., Malaisamy, R., Bruening, M.L., 2008. Multilayer polyelectrolyte films as nanofiltration membranes for separating monovalent and divalent cations. J. Membr. Sci. 310, 76e84. Paillard, H., Brunet, R., Dore, M., 1988. Optimal conditions for applying an ozone-hydrogen peroxide oxidizing system. Water Res. 22, 91e103. Pi, Y., Schumacher, J., Jekel, M., 2005. Decomposition of aqueous ozone in the presence of aromatic organic solutes. Water Res. 39, 83e88. Pianta, R., Boller, M., Urfer, D., Chappaz, A., Gmu¨nder, A., 2000. Costs of conventional vs. membrane treatment for karstic spring water. Desalination 131, 245e255. Piccolo, A., 2002. The supramolecular structure of humic substances. A novel understanding of humus chemistry and implications in soil science. Advan. Agron. 75, 57e134. Rivas, J., Gimeno, O., Beltra´n, F., 2009. Wastewater recycling: application of ozone based treatments to secondary effluents. Chemosphere 74, 854e859.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 5 1 e3 5 7 0

Rosal, R., Rodrı´guez, A., Perdigo´n-Melo´n, J.A., Petre, A., Garcı´aCalvo, E., 2009. Oxidation of dissolved organic matter in the effluent of a sewage treatment plant using ozone combined with hydrogen peroxide (O3/H2O2). Chem. Eng. J. 149, 311e318. Sartor, M., Schlichter, B., Gatjal, H., Mavrov, V., 2008. Demonstration of a new hybrid process for the decentralised drinking and service production from surface water in Thailand. Desalination 222, 528e540. Scha¨fer, A.I., Mastrup, M., Lund Jensen, R., 2002. Particle interactions and removal of trace contaminants from water and wastewaters. Desalination 147, 243e250. Schlichter, B., Mavrov, V., Chmiel, H., 2004. Study of a hybrid process combining ozonation and microfiltration/ ultrafiltration for drinking water production from surface water. Desalination 168, 307e317. Scopes, R.K., 1994. Protein Purification: Principles and Practice, third ed. Springer Verlag, New York (NY, USA), p. 73. Seidel, A., Elimelech, M., 2002. Coupling between chemical and physical interactions in natural organic matter (NOM) fouling of nanofiltration membranes: implications for fouling control. J. Membr. Sci. 203, 245e255. Shao, J., Hou, J., Song, H., 2011. Comparison of humic acid rejection and flux decline during filtration with negatively charged and uncharged ultrafiltration membranes. Water Res. 45, 473e482. Shon, H.K., Vigneswaran, S., Kim, I.S., Cho, J., Ngo, H.H., 2004. Effect of pretreatment on the fouling of membranes: application in biologically treated sewage effluent. J. Membr. Sci. 234, 111e120. Siddiqui, M., Amy, G., Ryan, J., Odem, W., 2000. Membranes for the control of natural organic matter from surface waters. Water Res. 34, 3355e3370. Song, W., Ravindran, V., Koel, B.E., Pirbazari, M., 2004. Nanofiltration of natural organic matter with H2O2/UV pretreatment: fouling alleviation and membrane surface characterization. J. Membrane Sci. 241, 143e160. Song, W., Ravindran, V., Pirbazari, M., 2008. Process optimization using a kinetic model for the ultraviolet radiation- hydrogen peroxide decomposition of natural and organic compounds in groundwater. Chem. Eng. Sci. 63, 3249e3270. Song, Y., Dong, B., Gao, N., Xia, S., 2010. Huangpu River water treatment by microfiltration with ozone pretreatment. Desalination 250, 71e75. Staehelin, J., Hoigne´, J., 1982. Decomposition of ozone in water: rate of initiation by hydroxide ions and hydrogen peroxide. Environ. Sci. Technol. 16, 676e681. Staehelin, J., Hoigne´, J., 1985. Decomposition of ozone in water in the presence of organic solutes acting as promoters and inhibitors of radical chain reactions. Environ. Sci. Technol. 19, 1206e1213. Sutton, R., Sposito, G., 2005. Molecular structure in soil humic substances: the new view. Environ. Sci. Technol. 39, 9009e9015. Sutzkover-Gutman, I., Hasson, D., Semiat, R., 2010. Humic substances fouling in ultrafiltration processes. Desalination 261, 218e231.  Swietlik, J., Sikorska, E., 2004. Application of fluorescence spectroscopy in the studies of natural organic matter fractions reactivity with chlorine dioxide and ozone. Water Res. 38, 3791e3799. Tang, C.Y., Kwon, Y.N., Leckie, J.O., 2007. Fouling of reverse osmosis and nanofiltration membranes by humic acid e effects of solution composition and hydrodynamic conditions. J. Membr. Sci. 290, 86e94. These, A., Reemtsma, T., 2005. Structure-dependent reactivity of low molecular weight fulvic acid molecules during ozonation. Environ. Sci. Technol. 39, 8382e8387. Thurman, E.M., 1986. Organic Geochemistry of Natural Waters. Martinus Nijhoff Publishers, Dordrecht (The Netherlands.

3569

Van Geluwe, S., Braeken, L., Vinckier, C., Van der Bruggen, B., 2009. Ozonation and perozonation of humic acids in nanofiltration concentrates. Desalin. Water Treat. 6, 217e221. Van Geluwe, S., Vinckier, C., Bobu, E., Trandafir, C., Vanelslander, J., Braeken, L., Van der Bruggen, B., 2010. Eightfold increased membrane flux of NF 270 by O3 oxidation of natural humic acids without deteriorated permeate quality. J. Chem. Technol. Biot 85, 1480e1488. Van Geluwe, S., Vinckier, C., Braeken, L., Van der Bruggen, B. Ozone oxidation of nanofiltration concentrates alleviates membrane fouling in drinking water industry. J. Membr. Sci., in press. van Voorthuizen, E.M., Ashbolt, N.J., Scha¨fer, A.I., 2001. Role of hydrophobic and electrostatic interactions for initial enteric virus retention by MF membranes. J. Membr. Sci. 194, 69e79. Verliefde, A., Cornelissen, E., Amy, G., Van der Bruggen, B., van Dijk, H., 2007. Priority organic micropollutants in water sources in Flanders and the Netherlands and assessment of removal possibilities with nanofiltration. Environ. Pollut. 146, 281e289. von Gunten, U., 2003a. Ozonation of drinking water: Part I. Oxidation kinetics and product formation. Water Res. 37, 1443e1467. von Gunten, U., 2003b. Ozonation of drinking water: Part II. Desinfection and by-product formation in presence of bromide, iodide and chlorine. Water Res. 37, 1469e1487. Waite, T.D., Scha¨fer, A.I., Fane, A.G., Heuer, A., 1999. Colloidal fouling of ultrafiltration membranes: impact of aggregate structure and size. J. Colloid Interface Sci. 212, 264e274. Wang, G.S., Pai, S.Y., 2001. Ozonation of dissolved organic matter in biologically treated wastewater effluents. Ozone Sci. Eng. 23, 351e358. Wang, Y., Hollingsworth, R.I., Kasper, D.L., Kasper, D.L., 1999. Ozonolytic depolymerization of polysaccharides in aqueous solution. Carbohyd. Res. 319, 141e147. Wang, X., Wang, L., Liu, Y., Duan, W., 2007. Ozonation pretreatment for ultrafiltration of the secondary effluent. J. Membr. Sci. 287, 187e191. Weinstein, J., Bielski, B.H.J., 1979. Kinetics of the interaction of HO2 and O2 radicals with hydrogen peroxide. The HaberWeiss reaction. J. Am. Chem. Soc. 101, 58e62. Wert, E.C., Rosario-Ortiz, F.L., Drury, D.D., Snyder, S.A., 2007. Formation of oxidation byproducts from ozonation of wastewater. Water Res. 41, 1481e1490. Westerhoff, P., Aiken, G., Amy, G., Debroux, J., 1999. Relationships between the structure of natural organic matter and its reactivity towards molecular ozone and hydroxyl radicals. Water Res. 33, 2265e2276. Williamson, D.G., Cvetanovic, R.J., 1970. Rates of ozone-paraffin reactions in carbon tetrachloride solution. J. Am. Chem. Soc. 92, 2949e2952. Xie, Y.F., 2004. Desinfection By-products in Drinking Water. Lewis, Boca Raton (FL, USA). Xiong, F., Legube, B., 1991. Enhancement of radical chain reactions of ozone in water in the presence of an aquatic fulvic acid. Ozone Sci. Eng. 13, 349e363. Xiong, F., Croue´, J.P., Legube, B., 1992. Long-term ozone consumption by aquatic fulvic acids acting as precursors of radical chain reactions. Environ. Sci. Technol. 26, 1059e1064. Yoon, Y., Westerhoff, P., Snyder, S.A., Wert, E.C., 2006. Nanofiltration and ultrafiltration of endocrine disrupting compounds, pharmaceuticals and personal care products. J. Membr. Sci. 270, 88e100. You, S.H., Tseng, D.H., Hsu, W.C., 2007. Effect and mechanism of ultrafiltration membrane fouling removal by ozonation. Desalination 202, 224e230.

3570

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 5 1 e3 5 7 0

Yuan, W., Zydney, A.L., 1999. Humic acid fouling during microfiltration. J. Membrance Sci. 157, 1e12. Zhang, T., Lu, J., Ma, J., Qiang, Z., 2008. Fluorescence spectroscopic characterization of DOM fractions isolated from a filtered river water after ozonation and catalytic ozonation. Chemosphere 71, 911e921.

Zhu, H., Wen, X., Huang, X., 2010. Membrane organic fouling and the effect of pre-ozonation in microfiltration of secondary effluent organic matter. J. Membr. Sci. 352, 213e221. Zularisam, A.W., Ismail, A.F., Salim, R., 2006. Behaviours of natural organic matter in membrane filtration for surface water treatment e a review. Desalination 194, 211e213.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 1 e3 5 7 8

Available at www.sciencedirect.com

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

Performance of granular zirconiumeiron oxide in the removal of fluoride from drinking water Xiaomin Dou a,*, Yansu Zhang a, Hongjie Wang a, Tingjie Wang b, Yili Wang a a b

Department of Environmental Science and Engineering, Beijing Forestry University, P.O. Box 60, Beijing 100083, PR China Department of Chemical Engineering, Tsinghua University, Beijing 100084, PR China

article info

abstract

Article history:

In this study, a granular zirconium-iron oxide (GZI) was successfully prepared using the

Received 8 October 2010

extrusion method, and its defluoridation performance was systematically evaluated. The

Received in revised form

GZI was composed of amorphous and nano-scale oxide particles. The Zr and Fe were evenly

24 February 2011

distributed on its surface, with a Zr/Fe molar ratio of w2.3. The granular adsorbent was

Accepted 3 April 2011

porous with high permeability potential. Moreover, it had excellent mechanical stability and

Available online 12 April 2011

high crushing strength, which ensured less material breakage and mass loss in practical use. In batch tests, the GZI showed a high adsorption capacity of 9.80 mg/g under an equilibrium

Keywords:

concentration of 10 mg/L at pH 7.0, which outperformed many other reported granular

Adsorption

adsorbents. The GZI performed well over a wide pH range, of 3.5e8.0, and especially well at

Defluoridation

pH 6.0e8.0, which was the preferred range for actual application. Fluoride adsorption on GZI

Fluoride

followed pseudo-second-order kinetics and could be well described by the Freundlich

Granular adsorbent

equilibrium model. With the exception of HCO3
, other co-existing anions and HA did not

Zirconium-iron oxide

evidently inhibit fluoride removal by GZI when considering their real concentrations in natural groundwater, which showed that GZI had a high selectivity for fluoride. In column tests using real groundwater as influent, about 370, 239 and 128 bed volumes (BVs) of groundwater were treated before breakthrough was reached under space velocities (SVs) of 0.5, 1 and 3 h
1, respectively. Additionally, the toxicity characteristic leaching procedure (TCLP) results suggested that the spent GZI was inert and could be safely disposed of in landfill. In conclusion, this granular adsorbent showed high potential for fluoride removal from real groundwater, due to its high performance and physicalechemical properties. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Fluoride contamination of drinking water is a worldwide problem, and excess intake of fluoride can cause harmful effects such as dental/skeletal fluorosis, fetal cerebral function, neurotransmitters, etc. (WHO, 2006, Viswanathan et al., 2009). Considering the serious health effects of fluoride, several technologies, including precipitation, adsorption, ion exchange, membrane separation and electrodialysis have been developed

and evaluated for fluoride removal (Meenakshi and Maheshwari, 2006; Mohapatra et al., 2009). Among these methods, adsorption has been considered to be one of the most promising technologies because it was found to be efficient, simple in operation and cost-effective (Mohapatra et al., 2009). A wide range of low-cost adsorbents have been reported for fluoride removal, including granular activated alumina (Ghorai and Pant, 2005), bone char (Leyva-Ramos et al., 2010), zeolite (Onyango et al., 2004), calcite (Turner et al., 2005) and other

* Corresponding author. Tel.: þ86 10 6233 6615; fax: þ86 10 6233 6596. E-mail address: [email protected] (X. Dou). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.002

3572

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 1 e3 5 7 8

low-cost materials (Fan et al., 2003). However, there were several problems associated with their use, for example low adsorption capacity, narrow available pH range and poor mechanical strength. Therefore, frequent regeneration or replacement was needed due to their relatively poor performance. Recently, considerable work was carried out to develop new adsorbents with good performance for fluoride removal. High performance materials, such as zirconium oxide (Blackwell and Carr, 1991), nano-alumina oxide (Wang et al., 2009), nano-hydroxyapatite (Sundaram et al., 2009), cellulosesupported layered double hydroxides (Mandal and Mayadevi, 2008), etc., have been investigated and reported. Among these adsorbents, zirconium-based materials have been paid more attention in recent investigations due to their high binding affinity with F and acceptable cost. In addition, iron oxide-based materials have been reported as showing good F removal performance as well as having favorable characteristics in terms of cost, environmental impact and chemical stability (e.g. resistance to acids and bases, low solubility) (Kumar et al., 2009; Tang et al., 2009; Liu et al., 2010). In order to benefit from the advantages of both of these two kinds of adsorbent, a zirconium-iron composite adsorbent was developed, and the resulting synthetic oxide had hybrid properties and showed promising performance (Biswas et al., 2007). Although Zr-based materials have shown great potential, they were prepared mainly as fine powders, microparticulate (Blackwell and Carr, 1991) or as freshly precipitated suspensions such as hydroxides or gels (Yuchi and Matsuo, 2005). These materials could not be directly used in fixed beds, and showed poor separation, low hydraulic conductivity and unavoidable leaching. To overcome these limitations, the powdered adsorbent needed to be immobilized. Coating, loading, impregnation or entrapment of active components in/on certain carriers to yield granular adsorbents have been attempted in previous studies. Zr(IV)-impregnated carbon (Alagumuthu and Rajan, 2010), Zr(IV)-entrapped chitosan polymeric matrix (Viswanathan and Meenakshi, 2009), chitosan-supported zirconium(IV) tungstophosphate (Viswanathan and Meenakshi, 2010), etc., have shown good performance for fluoride removal. These adsorbents, however, still suffered from several drawbacks. First, binding sufficient quantities of active components in the coating layer or entrapping them on carriers was not easy. Second, the stability and strength was often a problem (e.g., the breaking-off of the coating layer). Among the several granulation technologies reported, including extrusion granulation, high shear granulation, fluidized bed spray granulation, drum granulation, dry granulation and roll pressing, the extrusion method was considered as one of the most effective (Jacob, 2007). This method can fabricate equal-sized granular materials having adequate strength without a carrier core, and has been successfully applied in pharmaceutical production on a large scale (Jacob, 2007). However, the use of such a method to fabricate granular adsorbent, and the defluoridation performance of the resulting adsorbent, have been rarely investigated to date. In this study, granular zirconium-iron oxide (GZI) was fabricated using an extrusion granulation method. Fluoride removal potential of the granular adsorbent was systematically evaluated under various operating conditions such as initial F
concentration, pH, reaction time and co-existing

substances. Desorption behavior of GZI was also explored. In addition, defluoridation performance was evaluated for real F
-containing groundwater samples using column tests, and the leachability potential of the spent adsorbents was tested using the toxicity characteristic leaching procedure (TCLP).

2.

Materials and methods

2.1.

Materials

All chemicals were of analytical reagent grade. The F
stock solution was prepared with deionized water using NaF. F
-bearing solutions were freshly prepared by diluting F
stock solution with distilled water.

2.2.

Granular adsorbent preparation

The GZI adsorbent was prepared in a two-step process. (1) Powder preparation, whereby Zr(SO4)2·4H2O (9 mol) and FeSO4·7H2O (4.5 mol) were dissolved in 30 L of tap water, respectively. Under vigorous mechanical stirring, 6 M NaOH was added at a rate of 12.5 ml/min to slowly raise pH to the range of 7.5e8.0. The pH value was maintained in this range for 1 h with continuous stirring and (if necessary) further addition of NaOH. Next, the suspension was settled and aged at room temperature for 12 h, and then washed repeatedly with tap water 5 times. The suspension was filtered and dried at 65  C for 24 h. The dried material was ground and sieved with a 100-mesh sieve. The process yielded about 1.8 kg of powder. (2) An acrylic-styrene copolymer latex was used as the binder (Wu et al., 2008). The mixture was adequately blended in a dispersion kneader, and then transferred to an extrusion machine to prepare strip-like adsorbent with a diameter of about 1.5 mm, under a pressure of about 5.5 MPa. These strips were oven-dried at 60  C for 24 h, and then manually broken into lengths of 1e1.5 mm.

2.3.

Granular adsorbent characterization

The specific surface area and pore volume of the GZI adsorbent were determined by BET N2 adsorption-desorption analysis using a Micromeritics ASAP2000 surface area analyzer (Norcross, USA). The porosity and pore-size distributions of GZI and the original powder were determined using a mercury intrusion method by a Poremaster 60 GT porisimeter (Quantachrome, USA). XRD patterns of the original powder and the crushed granular material were characterized by an X’Pert PRO MPD diffractometer (PANalytical, the Netherlands) using Cu Ka radiation. Samples were scanned at a speed of 2 /min from 10 to 90 , operated at 40 kV and 40 mA. The bulk density was calculated by weighing a granular sample with a measured volume. The particle strength of GZI in the wet and dry states was characterized using the values of stability loss and crushing strength, respectively. The stability loss of GZI was determined following a previously reported procedure (Wu et al., 2008), in which 0% indicates no loss and the best stability, and 100% indicates the worst stability. The crushing strength was

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 1 e3 5 7 8

determined using a crushing strength tester (KY-20, Jiangyan Keyuan Corp., China), which exerted a steadily increasing force on GZI until it was crushed, with a sensor recording the force (expressed as crushing strength) at the point of failure. The surface morphology was determined by FE-SEM using a HITACHI S-4500 (Japan). The oxidation state of Fe in the GZI was characterized by XPS using a PHI Quantera (USA). The penetration and migration of F
into the granular material was investigated by observations of the transverse section of the adsorbent in column tests and by F
saturation in batch experiments, using an energy-dispersive X-ray spectrometer (KEVEX Level 4, EDAX Inc., USA) connected to the FE-SEM.

2.4.

Batch adsorption experiment

The adsorption isotherms were carried out by varying initial concentrations (10e150 mg/L) of fluoride under a fixed GZI dose of 5 g/L, with a total volume of 100 ml in 250 ml high density polyethylene (HDPE) bottles. These bottles were placed in a thermostatic orbital shaker, at a temperature of 25  1  C and shaking speed of 160 rpm. Sample pH was maintained at 7.0  0.2 by manual adjustment with 0.01 M HCl and NaOH. After a period of 10 h, residual F
in solution was analyzed. Analysis of the effect of solution pH on fluoride removal was performed in 250-ml HDPE bottles containing 100 ml of fluoride solution with pre-selected concentrations, and a GZI dose of 5 g/L. The pH was adjusted and maintained at a specified value in the range 3e11. Temperature was maintained at 25  1  C. After shaking for 10 h, GZI was separated and residual F
was analyzed. Since the rate of fluoride adsorption was instructive for selecting the proper empty bed contact time (EBCT) value, kinetics experiments were performed to determine the reaction time to reach adsorption equilibrium. Fluoride stock solution and deionized water were added into each of a series of 2000-ml HDPE bottles, to reach a pre-selected concentration of fluoride and a total volume of 1500 ml. Then, GZI adsorbent was added at a dose of 5 g/L. The pH of the mixtures was adjusted and maintained at 7.0  0.2 throughout the experiment. The mixtures were stirred at 160 rpm, and maintained at 25  1  C. Approximately 4-mL aliquots were taken from the suspension at predetermined intervals. The samples were immediately filtered through a 0.45-mm membrane, and then residual F
in solution was analyzed. Activated alumina (AA, WHA-104) obtained from Wenzhou Alumina Plant (Zhejiang, China), was also evaluated under similar conditions for comparison. The following quality control measures were taken: blank tests were conducted in which no adsorbent was added; all analytical instruments were calibrated before use; and all measurements were repeated and reported as a mean value. The effects of co-existing anions (Cl
, SO42
, HCO3
, NO3
, PO43
, SiO44
, AsO43
, etc.) and humic acids on fluoride adsorption were also investigated. The initial concentration of F
was fixed at 30 mg/L, with a total volume of 100 ml, and the GZI dose was 5 g/L.

2.5.

Batch desorption experiments

Desorption experiments were carried out by shaking the fluoride-loaded adsorbent in different concentrations of

3573

NaOH solution. Further details of these tests are provided in the Supplementary Material.

2.6.

Column experiments

Column studies on the GZI were performed in three perspex columns with an inner-diameter of 1.9 cm and a length of 40 cm. The height of the packed GZI bed was 20 cm and the volume was 56.7 ml. Groundwater taken from Cangzhou City, Hebei Province, China, was used as the influent and had the characteristics listed in Table S1. The GZI columns were operated at SV values of 0.5, 1 and 3 h
1, respectively. The effluent was collected at regular intervals and the concentrations of F, Zr and Fe were measured.

2.7. test

The toxicity characteristic leaching procedure (TCLP)

The used adsorbents from the column experiment were examined by the TCLP test to determine if the spent granular material was inert or hazardous in terms of the leachability of adsorbed F (EPA, 1999). The concentrations of possibly leached Zr and Fe after TCLP test were also examined. Details of the experimental procedure are reported in the Supplementary Material.

2.8.

Analytical methods

The concentrations of residual F
were analyzed with a fluoride-selective electrode connected to an ion meter (Metrohm 809, Swiss) following reported method (Liu et al., 2010). Fe and Zr in the effluent from the column studies and from the TCLP tests were analyzed using inductively coupled plasma-mass spectrometry (Plasma Quad 3, VG Corporation, UK).

3.

Results and discussion

3.1.

Characterization of the GZI adsorbent

The prepared GZI adsorbent particles were 1e1.5 mm in diameter and 1e1.5 mm in length as shown in Fig. 1 (a). FESEM image (Fig. 1 (b)) revealed that nano-scale particles of size 20e100 nm were aggregated and cohered, and that the surface of the fabricated granular adsorbent was porous. The BET specific surface area of GZI was 95.5 m2/g. The total porosities (MIP method) of the original powder and GZI were 71.3% and 64.2%, with pore-size distributions ranging from 3.6 nm to 226 mm, and 3.6 nm to 218 mm, respectively (Fig. S1 (a)). These results demonstrated the high potential of GZI in terms of permeability and adsorption. Further characterizations are provided in the Supplemental Material. EDX analysis of the GZI surface revealed that Zr and Fe were evenly distributed on the surface with a Zr/Fe molar ratio of w2.3, as well as on the surface of the original powder. This ratio was slightly higher than that in the raw material composition, which might result from the incomplete oxidation and coprecipitation of Fe(II). This result favored fluoride removal since the binding affinity of F to Zr was stronger than that to Fe (Bebeshko, 2004). The oxidation state of Fe on the

3574

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 1 e3 5 7 8

Fig. 1 e (a) Photograph of GZI; (b) FE-SEM image of GZI, 30003 (white scale bar, 2 mm).

surface of the crushed GZI was examined by the Fe2 p spectrum (Fig. S1 (b)). The Fe 2p1/2 and Fe 2p3/3 peak positions (724.8 and 711.1 eV) and peak shapes showed characteristics of Fe(III), indicating that although Fe(II) was used as a raw ingredient, it was oxidized during coprecipitation and oven drying, yielding Fe(III) on the GZI surface, thus leading to the higher stability of the adsorbent. XRD patterns of the original powder and of the powder from the crushed GZI were obtained (Fig. S1 (c)). These revealed a clear amorphous structure and two broad peaks at around 32 and 58 , which were significantly different from the crystalline structure of the FeeZr oxide reported by Biswas et al. (2007). A search/match analysis revealed that the two broad peaks were not attributed to any known Zr/Fe oxides or oxysulphate oxides. Other important characteristics of the granular adsorbent are listed in Tables S2 and S3. Remarkably, the crushing strength was 22.1  2.0 N and the stability loss was 9.1%. These results indicated that the granular adsorbent was resistant to abrasion loss from flow flushing, striking and backwashing, thus ensuring its durability in usage.

Langmuir, Freundlich and Sip isotherm models. The corresponding model results are shown in Table 1. The results indicated that the Freundlich model fitted the experimental data reasonably well, yielding determination coefficients (R2) above 0.980. Curve fitting results obtained using the Freundlich model are presented as solid lines in Fig. 2. As shown in Table 1, KF and 1/n are Freundlich constants corresponding to adsorption capacity and adsorption intensity, respectively. As pH increased from 3.5 to 10.0, KF decreased from 6.714 to 2.567 mg/g and 1/n increased from 0.356 to 0.495, respectively, indicating the adsorption intensity decreased and the surface became less heterogeneous at high pH values (Maliyekkal et al., 2006). It is worthwhile to compare the F
adsorption capacity of GZI with other adsorbents. Considering that high F
concentrations might be found in groundwater, these adsorbents were compared under a fixed equilibrium concentration of 10 mg/L at pH 7.0 (Table 2). Fig. 2 shows that the GZI had a capacity of 9.80 mg/g under such conditions. These results demonstrate the better performance of GZI than other reported adsorbents.

3.2.

3.2.2.

3.2.1.

Batch adsorption experiments

Effect of pH on fluoride removal

Fig. S2 shows the influence of pH on fluoride removal as a function of initial fluoride concentrations. When the initial fluoride concentration was 30 mg L
1, w100% fluoride was removed by GZI at pH < 8.0. The removal of fluoride decreased

Adsorption isotherms

To estimate the adsorption capacity of GZI and AA, the equilibrium data were fitted using several models, including the

Table 1 e Isotherm fitting for fluoride adsorption on the GZI and AA adsorbent at various pH values. Isotherm models

Parameters

qm;L kL Ce Langmiur model qe ¼ 1 þ kL Ce 1=n

Freundlich model qe ¼ kF Ce

Sips model qe ¼

qm;S ðkS Ce Þms 1 þ ðkS Ce Þms

qm,L (mg/g) kL (L/mg) R2 kF (g/mg) 1/n R2 qm,S (mg/g) ks (L/g) ms R2

GZI at investigated pHs

AA

3.5

5

6

7

8

9

10

7

26.54 0.169 0.946 6.714 0.356 0.994 74.58 0.005 0.453 0.996

23.33 0.120 0.965 4.752 0.388 0.988 47.58 0.013 0.535 0.994

22.23 0.120 0.975 4.300 0.400 0.989 38.89 0.022 0.589 0.994

22.82 0.090 0.971 3.484 0.445 0.980 41.02 0.017 0.628 0.982

22.15 0.085 0.978 3.023 0.471 0.986 42.09 0.014 0.652 0.988

23.24 0.085 0.975 3.538 0.439 0.989 43.99 0.014 0.611 0.991

21.35 0.080 0.974 2.567 0.495 0.983 32.12 0.026 0.798 0.993

7.728 0.078 0.992 1.244 0.408 0.990 12.79 0.018 0.625 0.997

3575

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 1 e3 5 7 8

reached within 3 h at an initial fluoride concentration of 10 or 30 mg/L, and most sorption took place within 10 h for initial fluoride concentrations of 50 or 100 mg/L; after 10 h, there was a negligible increase in adsorption rate and residual F
concentration reached an almost constant value. Therefore, a shaking time of 10 h was used in all bath experiments. Also, it was evident that GZI reached equilibrium more rapidly than AA, and had a larger capacity than AA. Kinetic data for fluoride adsorption onto GZI and AA were fitted using different models. Initially, the Langergren pseudofirst-order model (Lagergren, 1898), and Ho’s linearized pseudo-second-order reaction rate models (Ho and McKay, 1999) were tested to describe the kinetic process. The mathematical representations of the two models are given in Equations (1) and (2), respectively. log qe
qt

Fig. 2 e Adsorption isotherm of fluoride on the GZI adsorbent at various pH values.

¼ logqe


k1 t 2:303

(1)

t 1 1 ¼ þ t qt k2 q2e qe

(2)

where qe and qt are the amount of adsorbed fluoride at equilibrium and at any time t (mg/g solid material), k1 (min
1) and k2 (g∙mg
1 min
1) are the equilibrium rate constants for pseudo first- and second-order sorption respectively, and t is the shaking time (min). The calculated rate constants and related parameters are listed in Table 3. The high value of the determination coefficient (R2 > 0.99) of the pseudo-second-order equation indicated that fluoride removal on the adsorbent followed the pseudo-second-order rate law. As the initial fluoride concentration decreased from 100 mg/L to 10 mg/L, the kinetic parameters of the pseudo-second-order rate law significantly increased (Table 3), as would be expected in the treatment of real F
-containing groundwater. In this case, the low fluoride concentration relative to the large dose of adsorbent in fixed bed resulted in a high adsorbent/adsorbate ratio, thus promoting a fast adsorption process and indicating that shorter adsorption times were needed; this result also showed that higher operational flow rates were achievable and that a smaller adsorption bed volume was acceptable. For the granular adsorbent, the intraparticle diffusion in the pores was frequently the rate-determining step of the sorption processes, as also been observed for arsenate

slowly with increasing pH when fluoride was dosed at 50 or 100 mg L
1. However, it was interesting to find that fluoride removal remained almost constant in the pH range of 6.0e8.0, even at high fluoride concentration, consistent with the adsorption isotherm results shown in Fig. 2. This is beneficial to the application of GZI in practice, as the pH of groundwater is often in the range of 6.0e8.0. The fluoride removal by AA as a function of pH is also shown in Fig. S2 for comparison. The GZI was clearly more effective for F
removal than AA, especially at pH 6.0e8.0. Considerable reduction in fluoride removal efficiency of GZI was observed above pH 10, which may be due to the competition between OH
ions in solution and F
for the sorption site (Maliyekkal et al., 2006), and increased electrostatic repulsive force between the deprotonated surface and negatively charged F
at high pH values (Dzombak and Morel, 1990). Overall, a wide pH range was observed to be optimal for fluoride removal using GZI, especially at pH 6.0e8.0.

3.2.3.




Kinetic rate parameters

Fig. 3 shows the time dependence of fluoride sorption onto GZI and AA at various initial concentrations. Equilibrium was

Table 2 e Comparative evaluation of GZI and various adsorbents for fluoride removal under an equilibrium concentration of 10 mg/L Adsorbent GZI granular ferric hydroxide (GFH)

Particle size (mm)

pH

1.0e1.5 7.0 0.32e2.0, 6.0e7.0 media 1.16 siderite 0.074 6.86 zirconium(IV) tungstophosphate 0.004 7.0 zirconium impregnated carbon 0.053 7.0 Manganese-oxide-coated alumina 0.5e0.6 7.0 nano-hydroxyapatite/chitin composite powder neutral pH Cellulose supported Zn/Al LDH powder La-exchanged zeolite F-9 0.15e0.30

Dose Ceq of F
(mg/L) (g/L)

Capacity (mg/g); teq (h); temp. ( C)

Reference

10 10

5 10

9.80; 10; 25 5.84; 24; 25

Present study Kumar et al., 2009

10 10 10 10 10 10 10

20 2 0.015 5 2 4 2

1.60; 12; 25 Liu et al., 2010 2.03; 0.5; 30 Viswanathan and Meenakshi, 2010 1.81; 3; 30 Alagumuthu and Rajan 2010 2.46; 3; 30 Maliyekkal et al., 2006 3.71; 0.5; room temp. Sundaram et al., 2009 9.10; 1; 25 Mandal and Mayadevi 2008 9.52; 24; 30 Onyango et al. 2004

3576

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 1 e3 5 7 8

followed for up to 12 h, which could be described well by the Weber-Morris model (R2 > 0.94); however, the linear fit did not pass through the origin, indicating a complex adsorption process of fluoride onto GZI. Also, increased slopes (Fig. S3 and Table 3) were observed with increasing initial fluoride concentrations, which could be explained by the increasing effect of concentration gradient as a driving force (Choy et al., 2004).

3.2.4.

Fig. 3 e Kinetic data of fluoride adsorption on the GZI adsorbent under different initial fluoride concentrations at pH [ 7.0 ± 0.2.

adsorption on granular ferric hydroxide (GFH) adsorbent (Badruzzaman et al., 2004). In the present study, the role of intraparticle diffusion in the kinetic process was investigated using Equation (3), following Weber and Morris (1963): qt ¼ kp t

0:5

(3)

where qt is the amount of adsorbed fluoride at a time t (mg/g solid material), kp (min
1) is the equilibrium rate constant of intraparticle diffusion and t is the shaking time (min). If intraparticle diffusion is a rate-controlling step, then the plot of qt against t0.5 should be linear and should pass through the origin. If the plot shows multi-linearity, this would indicate further complexity of the adsorption process. The initial curved portion represents boundary layer diffusion, while the following linear portion is attributed to intraparticle diffusion (Weber and Morris, 1963; Choy et al., 2004). In this study, the plots of qt against t0.5 (Fig. S3) showed a short initial linear portion and long successive linear portion, which were attributed to external mass transfer and intraparticle diffusion, respectively (Choy et al., 2004). This implied that the external mass transfer was fast during the initial 0.56 h, and that a stage of intraparticle diffusion

Effects of co-existing substances

Natural groundwater always contains numerous aqueous constituents, which can compete for sorption sites and decrease the removal efficiency of the adsorbent. Fluoride adsorption in the presence of potential co-existing substances was investigated, and the results are shown in Fig. 4. It was clear that the presence of Cl
, NO3
, SO42
, and SiO44
had little effect on F
removal, while HCO3
PO43
, AsO43
and HA competed considerably with F
in the order HA > AsO43
> PO43
> HCO3
(Fig. 4 and Table S4, see Supplementary Material for more detailed discussion). The typical natural concentration ranges in groundwater are: HA, 0e5 mg/L; arsenate, 0.1e5 mg/L; phosphate, 0e5 mg/L; bicarbonate, 0e400 mg/L (Younger, 2007). Considering these natural concentration levels, interference was more likely from HCO3
than from the other three substances.

3.3.

Batch desorption experiments

Regeneration and reuse of the adsorbent was preferred since this allowed full utilization of the capacity of the adsorbent. Therefore, the desorption behavior of GZI was evaluated. When using deionized water as desorption solution, only 3.6% of the fluoride was desorbed, while using 0.01, 0.05, 0.10 and 0.50 M NaOH, respective desorption ratios of 83.71%, 83.81%, 81.39% and 50.61% were obtained. It was found that similarly high efficiencies were yielded using 0.01, 0.05 and 0.10 M NaOH. Considering the negative effect of alkaline residues in the GZI pore structure, as well as the chemical cost, 0.01 M NaOH was considered to be optimal in the regeneration procedure.

3.4.

Column tests using real F
-containing groundwater

Column test results are shown in Fig. 5. With an influent fluoride concentration of 3.59 mg/L, pH of 8.3, and TOC of

Table 3 e Kinetic parameters for fluoride adsorption on the GZI and AA adsorbent under various initial concentrations. Kinetic models

Pseudo first-order

Pseudo second-order

Intraparticle diffusion

Parameters

qe (mg/g) k1 (1/h) R2 qe (mg/g) k2 (g/(mg∙h)) R2 kp (mg/g∙h0.5) R2

GZI

AA

10 mg/L

30 mg/L

50 mg/L

100 mg/L

30 mg/L

1.983 3.913 0.964 2.061 4.043 0.967 1.222  10
5 0.982

5.048 3.080 0.873 5.400 0.905 0.960 2.134  10
4 0.945

8.057 0.668 0.948 9.448 0.086 0.983 7.453  10
4 0.972

14.35 0.464 0.949 17.55 0.029 0.976 1.410  10
3 0.993

3.642 0.441 0.855 4.272 0.127 0.925

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 1 e3 5 7 8

3577

Fig. 4 e Effect of co-existing anions on fluoride adsorption by the GZI adsorbent.

7.6 mg/L, about 370, 239 and 128 BVs of groundwater, respectively corresponding to SVs of 0.5, 1 and 3 h
1, were treated before F
in the effluent reached 1.5 mg/L (WHO standard). A mass balance calculation indicated that the concentrations of total loaded F
in the three columns were about 0.93, 0.66 and 0.33 mg/g, respectively. The cumulative capacities of the column tests were far lower than 22.1 mg/g (pH 8.0) calculated from the Langmuir isotherm, as has been commonly observed and reported in other studies (Ghorai and Pant, 2005; Wu et al., 2007). This result was attributed to the low fluoride concentration, short contact time and intraparticle mass transfer resistance. The contents of Zr and Fe in the effluent were shown to be near zero and met the drinking water standard. To determine whether the entire particle or just its surface was utilized for F
adsorption, a transverse section of GZI following column tests was examined by EDX operated in mapping mode. A negligible fluoride peak was observed, probably because of the low determination sensitivity of this technique. Further, F
-loaded GZI adsorbent with a capacity of 5 mg/g was examined (Fig. S4 (a) and (b)), revealing that

Fig. 5 e Column tests of the GZI adsorbent in the treatment of real FL-containing groundwater (Dot line corresponding to the MCL standard of WHO).

fluoride was distributed quite evenly across the transverse section, suggesting that F
migrated fully into the entire granular adsorbent and that nearly all of the active sites inside the GZI were available.

3.5.

TCLP test

In the TCLP test, the released F concentration from the used GZI in the three columns (operated at SV values of 0.5, 1 and 3 h
1) were 0.24, 0.15 and 0.14 mg/L, respectively; these values are much lower than the established U.S. EPA standard of 48 mg/L. The leached iron and zirconium concentrations were less than 1.48 mg/L and 1.88 mg/L, respectively; these values are much lower than the guidelines of Alberta Environmental Protection of 1000 mg/L and 500 mg/L, respectively. These results indicated that the spent adsorbents can be considered as inert.

4.

Conclusions

Granular zirconium-iron oxide (GZI) adsorbent was fabricated using an extrusion granulation method. GZI exhibited a fluoride capacity of 9.80 mg/g under an equilibrium concentration of 10 mg/L at pH 7.0, which was higher than many reported granular adsorbents. The stability loss of GZI was less than 9.1% and the crushing strength was up to 22.1  2.0 N, demonstrating that GZI had sufficient strength to resist abrasion loss and could withstand the pressures exerted by bed weight. GZI performed well over a wide pH of 3.5e8.0, and especially well at pH 6.0e8.0. The kinetic results revealed that fluoride sorption onto GZI followed a pseudo-second order kinetic model, and that intraparticle diffusion controlled the adsorption process after 0.56 h. With the exception of HCO3
, other co-existing substances had little effect on F
removal by GZI when considering their typical concentration ranges in natural groundwater. Regeneration tests showed that 0.01 M NaOH was optimal for desorbing the F
-loaded GZI. Column studies showed that GZI was efficient in the treatment of real F
-containing groundwater from Cangzhou City. Finally, The TCLP results suggested that spent adsorbents were inert and could be safely disposed of in landfill. Thus, the GZI was

3578

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 1 e3 5 7 8

demonstrated as an effective, pH insensitive, highly selective, high strength and chemically stable granular adsorbent suitable for removing fluoride from water.

Acknowledgments This work was supported by the Fundamental Research Funds for the Central Universities (YX2010-33), the National High-tech R&D Program (No.2007AA06Z301, 2007AA06Z319), and the Beijing Nova Program (No.2008A33). The authors are thankful to Dr. Xiaohong Guan Tongji University and Dr. Gaosheng Zhang (Yantai Institute of Coastal Research for Sustainable Development, Chinese Academy of Sciences, YIC-CAS), for indepth discussions and suggestions.

Appendix. Supplementary data Supplementary data related to this article can be found online, at doi:10.1016/j.watres.2011.04.002.

references

Alagumuthu, G., Rajan, M., 2010. Equilibrium and kinetics of adsorption of fluoride onto zirconium impregnated cashew nut shell carbon. Chem. Eng. J. 158, 451e457. Badruzzaman, M., Westerhoff, P., Knappe, D.R.U., 2004. Intraparticle diffusion and adsorption of arsenate onto granular ferric hydroxide (GFH). Water Res. 38, 4002e4012. Bebeshko, G.I., 2004. Thermodynamic analysis of fluorineemetalewater systems for improving the selectivity of the Potentiometric determination of fluorine in raw minerals. J. Anal. Chem. 59, 528e531. Biswas, K., Bandhoyapadhyay, D., Ghosh, U., 2007. Adsorption kinetics of fluoride on iron(III)-zirconium(IV) hybrid oxide. Adsorption 13, 83e94. Blackwell, J.A., Carr, P.W., 1991. Study of the fluoride adsorption characteristics of porous microparticulate zirconium oxide. J. Chromatogr. A 549, 43e57. Choy, K.K.H., Ko, D.C.K., Cheung, C.W., Porter, J.F., McKay, G., 2004. Film and intraparticle mass transfer during the adsorption of metal ions onto bone char. J. Colloid Interface Sci. 271, 284e295. Dzombak, D.A., Morel, F.M.M., 1990. Surface Complexation Modeling: Hydrous Ferric Oxide. Wiley-Interscience, New York. EPA, 1999. Toxicity Characteristics Leaching Procedure. US Environmental Protection Agency, Fed. Reg, p. 11798. Fan, X., Parker, D.J., Smith, M.D., 2003. Adsorption kinetics of fluoride on low cost materials. Water Res. 37, 4929e4937. Ghorai, S., Pant, K.K., 2005. Equilibrium, kinetics and breakthrough studies for adsorption of fluoride on activated alumina. Sep. Purif. Technol. 42, 265e271. Ho, Y.S., McKay, G., 1999. Pseudo-second-order model for sorption processes. Process Biochem. 34, 451e465. Jacob, M., 2007. In: Salman, A.D., Hounslow, M.J., Seville, J.P.K. (Eds.), Handbook of Powder Technology, Granulation. Elsevier Science B.V, pp. 417e476. Kumar, E., Bhatnagar, A., Ji, M., Jung, W., Lee, S.H., Kim, S.J., Lee, G., Song, H., Choi, J.Y., Yang, J.S., Jeon, B.H., 2009.

Defluoridation from aqueous solutions by granular ferric hydroxide (GFH). Water Res. 43, 490e498. Lagergren, S., 1898. About the theory of so-called adsorption of soluble substances. K. Svenska VetenskapsakadHandl 24, 1e39. Leyva-Ramos, R., Rivera-Utrilla, J., Medellin-Castillo, N.A., Sanchez-Polo, M., 2010. Kinetic modeling of fluoride adsorption from aqueous solution onto bone char. Chem. Eng. J. 158, 458e467. Liu, Q., Guo, H., Shan, Y., 2010. Adsorption of fluoride on synthetic siderite from aqueous solution. J. Fluorine Chem. 131, 635e641. Maliyekkal, S.M., Sharma, A.K., Philip, L., 2006. Manganese-oxidecoated alumina: a promising sorbent for defluoridation of water. Water Res. 40, 3497e3506. Mandal, S., Mayadevi, S., 2008. Cellulose supported layered double hydroxides for the adsorption of fluoride from aqueous solution. Chemosphere 72, 995e998. Meenakshi, Maheshwari, R.C., 2006. Fluoride in drinking water and its removal. J. Hazard. Mater. 137, 456e463. Mohapatra, M., Anand, S., Mishra, B.K., Giles, D.E., Singh, P., 2009. Review of fluoride removal from drinking water. J. Environ. Manage. 91, 67e77. Onyango, M.S., Kojima, Y., Aoyi, O., Bernardo, E.C., Matsuda, H., 2004. Adsorption equilibrium modeling and solution chemistry dependence of fluoride removal from water by trivalent-cation-exchanged zeolite F-9. J. Colloid Interface Sci. 279, 341e350. Sundaram, C.S., Viswanathan, N., Meenakshi, S., 2009. Fluoride sorption by nano-hydroxyapatite/chitin composite. J. Hazard. Mater. 172, 147e151. Tang, Y., Guan, X., Wang, J., Gao, N., McPhail, M.R., Chusuei, C.C., 2009. Fluoride adsorption onto granular ferric hydroxide: effects of ionic strength, pH, surface loading, and major coexisting anions. J. Hazard. Mater. 171, 774e779. Turner, B.D., Binning, P., Stipp, S.L.S., 2005. Fluoride removal by calcite: evidence for fluorite precipitation and surface adsorption. Environ. Sci. Technol. 39, 9561e9568. Viswanathan, G., Jaswanth, A., Gopalakrishnan, S., Siva Ilango, S., Aditya, G., 2009. Determining the optimal fluoride concentration in drinking water for fluoride endemic regions in South India. Sci. Total Environ. 407, 5298e5307. Viswanathan, N., Meenakshi, S., 2009. Synthesis of Zr(IV) entrapped chitosan polymeric matrix for selective fluoride sorption. Colloids Surf. B. 72, 88e93. Viswanathan, N., Meenakshi, S., 2010. Development of chitosan supported zirconium(IV) tungstophosphate composite for fluoride removal. J. Hazard. Mater. 176, 459e465. Wang, S.G., Ma, Y., Shi, Y.J., Gong, W.X., 2009. Defluoridation performance and mechanism of nano-scale aluminum oxide hydroxide in aqueous solution. J. Chem. Technol. Biotechnol. 84, 1043e1050. Weber, W.J., Morris, J.C., 1963. Kinetics of adsorption on carbon solution. J. Sanit. Eng. Div. Am. Soc. Civ. Eng. 89, 31e59. WHO, 2006. Fluoride in Drinking-water (London, UK). Wu, H.X., Wang, T.J., Dou, X.M., Zhao, B., Chen, L., Jin, Y., 2008. Spray coating of adsorbent with polymer latex on sand particles for fluoride removal in drinking water. Ind. Eng. Chem. Res. 47, 4697e4702. Wu, X.M., Zhang, Y., Dou, X.M., Yang, M., 2007. Fluoride removal performance of a novel FeeAleCe trimetal oxide adsorbent. Chemosphere 69, 1758e1764. Younger, P.L., 2007. Groundwater in the Environment: An Introduction. Blackwell Publishing Ltd., Main Street, Malden, MA, USA. Yuchi, A., Matsuo, K., 2005. Adsorption of anions to zirconium(IV) and titanium(IV) chemically immobilized on gel-phase. J. Chromatogr. A 1082, 208e213.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 9 e3 5 8 9

Available at www.sciencedirect.com

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

The effect of primary treatment and flow regime on clogging development in horizontal subsurface flow constructed wetlands: An experimental evaluation Anna Pedescoll, Ange´lica Corzo, Eduardo A´lvarez, Joan Garcı´a, Jaume Puigagut* Environmental Engineering Division, Department of Hydraulic, Maritime and Environmental Engineering, Technical University of Catalonia, c/Jordi Girona 1-3, Building D1, 08034 Barcelona, Spain

article info

abstract

Article history:

The effect of both the type of primary treatment (hydrolitic up-flow sludge blanket (HUSB)

Received 2 December 2010

reactor and conventional settling) and the flow regime (batch and continuous) on clogging

Received in revised form

development in subsurface flow constructed wetlands (SSF CWs) was studied. Clogging

21 March 2011

indicators (such as accumulated solids, hydraulic conductivity and drainable porosity)

Accepted 23 March 2011

were determined in an experimental plant with three treatment lines. Correlations were

Available online 31 March 2011

encountered between the solids accumulated and both saturated hydraulic conductivity and drainable porosity reduction over time (74.5% and 89.2% of correlation, respectively).

Keywords:

SSF CW implemented with a HUSB reactor accumulated ca. 30% lower sludge (1.9 kg DM/

Clogging indicators

m2) than a system with a settler (2.5e2.8 kg DM/m2). However, no significant differences

Treatment wetlands

were recorded among treatment lines concerning hydraulic parameters (such as hydraulic

Flow regime

conductivity or porosity). Root system development contributed to clogging. Accordingly,

Primary treatment

planted wetlands showed between 30% and 40% and 10% lower hydraulic conductivity and

HUSB reactor

porosity reduction, respectively, than non-planted wetlands. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Subsurface flow constructed wetlands (SSF CWs) are extensive systems widely used for the treatment of wastewater generated in small communities (Rousseau et al., 2005). Low energy requirements and non specialized manpower for plant management are among the most important advantages of SSF CWs in comparison to conventional alternatives such as the activated sludge processes (Wallace and Knight, 2006). It is widely accepted that clogging is the worst operational problem of such technology (Cooper et al., 2005; Knowles et al., 2011; Wallace and Knight, 2006). Clogging is a complex phenomenon that involves biological, chemical and physical processes. Accordingly, retention of inorganic and organic influent

particles, biofilm and plant biomass development and decay, and deposition and accumulation of chemical precipitates are among the most important factors promoting a progressive obstruction of the filter media (Knowles et al., 2011). Clogging limits the lifespan of the systems (Caselles-Osorio et al., 2007) and can have negative impacts on treatment efficiency (Rousseau et al., 2005). Because of the drawbacks that clogging may have on SSF CWs (both in treatment and management costs terms) there is great interest in studies aimed at assessing, understanding and preventing development of clogging processes (Caselles-Osorio et al., 2007; Knowles et al., 2010; Mun˜oz et al., 2006; Suliman et al., 2006; Tanner et al., 1998). The quantification of solids accumulation in SSF CWs constitutes a direct measure of clogging, since solids clog the

* Corresponding author. Tel.: þ34 934010898; fax: þ34 934017657. E-mail address: [email protected] (J. Puigagut). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.03.049

3580

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 9 e3 5 8 9

pore space of the filter media (Caselles-Osorio et al., 2007; Tanner et al., 1998). However, this clogging indicator demands an extremely exhaustive sampling effort due to the high heterogeneity of the systems (either in accumulation and solids nature terms) (Caselles-Osorio et al., 2007; Llorens et al., 2009; Tanner and Sukias, 1995). Therefore, indirect clogging indicators have been also widely used, such as tracer tests for hydrodynamic assessment (Mun˜oz et al., 2006) and hydraulic conductivity measurements (Knowles et al., 2010; Pedescoll et al., 2009; Suliman et al., 2006). Furthermore, correlations between clogging indicators have been scarcely addressed in current literature and, whenever performed, relations are not straight forward (Caselles-Osorio et al., 2007; Tanner et al., 1998). Moreover, the solids loading rate have a significant effect on clogging development (Tanner and Sukias, 1995). Therefore, the implementation of improved primary treatments is essential to delay clogging in treatment wetlands (Tchobanoglous, 2003). Conventional primary treatments such as Imhoff or septic tanks are commonly coupled to SSF CWs (Brix and Arias, 2005; Tchobanoglous, 2003). Furthermore, the use of other types of primary treatments in the context of wetland technology (such as low rate anaerobic digesters) has been scarcely investigated ´ lvarez et al., 2008; Barros et al., 2008). Hydrolytic up-flow sludge (A blanket reactors (HUSB) are anaerobic reactors where wastewater suspended solids are trapped within a sludge blanket. In HUSB reactors trapped solids undergo hydrolysis and acid fermentation and methanogenesis is suppressed due to a low ´ lvarez et al., 2008). On hydraulic retention time (from 2 to 5 h) (A the other hand, biomass retention time in HUSB reactors is high (usually more than 15 days) in order to allow a continuous ´ lvarez et al., 2008). growth of acid fermenting bacteria (A In addition to the use of improved primary treatments, alternative operation strategies and design criteria might be of use to avoid rapid clogging (Langergraber et al., 2003; Nguyen, 2000; Zhao et al., 2006). To this regard, horizontal SSF CWs are generally operated under water saturated conditions and, thus, physical oxygen transfer rates from the air to the bulk water are low (<1 g O2/m2d) (Tyroller et al., 2010). Although macrophytes actively transport oxygen from the atmosphere to the bulk water, direct measurements of the amount which is actually released by plant roots is very low (i.e. 0.001e0.004 gO2/m2.d (Bezbaruah and Zhang, 2005)). Therefore, oxygen availability in horizontal SSF CWs (operating at a normal organic loading of 4e6 BOD/m2d (Faulwetter et al., 2009)) might by a limiting factor for the degradation of solids that gradually accumulate in the granular medium. Alternatively, discontinuous or batch feeding strategies have been described to enhance oxygen transfer to wetlands and, therefore, degradation of accumulated solids could be enhanced under these operation conditions (Chazarenc et al., 2009). This study aimed at evaluating the long-term effect of two types of primary treatment (HUSB reactor and conventional settling) as well as two flow regimes (batch and continuous) on clogging development in SSF CWs. To this end, several commonly used clogging indicators, such as solids accumulation, drainable porosity and saturated hydraulic conductivity, were analysed and contrasted. The tested hypothesis was that of the use of a HUSB reactor as primary treatment

and alternating batch-unsaturated and permanently saturated phases (batch feeding mode) could delay the gradual clogging process in SSF CWs. Results here presented belong to the first 3 years of operation of an experimental plant. Although clogging is a long-term phenomena, studies devoted to clogging assessment describe either significant organic matter accumulation during the first 1e2 years of operation (Ruı´z et al., 2010; Tanner and Sukias, 1995) or hydraulic conductivity reduction (Sandford et al., 1995; Suliman et al., 2006). Moreover, evidence of systems in advance state of clogging is also described in literature after just 4 years of operation (Caselles-Osorio et al., 2007). Therefore, authors consider that the time scale devoted to clogging assessment presented in this study (three years) is long enough to address the stated research objectives. However, it is necessary to point out that, even though the wetlands operation conditions applied in this study (organic loading, hydraulic retention time and solids loading, etc.) are comparable to full-scale wetlands, there is obviously an important difference in terms of scale. Therefore, although authors do believe that the conclusions extracted are reliable, the research topics here addressed must by further analysed in full-scale systems.

2.

Materials and methods

2.1.

Experimental plant

The experimental plant used in this study (Fig. 1) was set in operation in February 2007 and treated domestic wastewater. Pre-treatment consisted of coarse screening. After pre-treatment the wastewater was conveyed to a plastic reservoir of 1.2 m3 which was continuously stirred (1380 rpm) in order to avoid solids sedimentation. The effluent of the storage tank was diverted to 3 different treatment lines, which for reasons of comprehension have been named under batch, control and anaerobic lines. Control line consists of a conventional settler as primary treatment followed by a horizontal subsurface flow constructed wetland (HSSF CW) permanently saturated and intermittently fed. The anaerobic line has the same hydraulic regime than the control line but the primary treatment consists of a HUSB reactor. Finally, the batch line has the same

Fig. 1 e Schematic diagram of the experimental plant.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 9 e3 5 8 9

primary treatment than the control line (conventional settler) but it was fed in batch mode (with cycles of saturated and unsaturated phases). Secondary treatment for the three experimental lines consisted of two small wetlands in parallel (0.65 m2 each) and one bigger in series (1.65 m2) planted with common reed (Phragmites australis). Each line also included one unplanted small wetland the effluent of which by-passed the bigger wetland (Fig. 1). Clogging assessment was addressed by comparing planted and unplanted small wetlands of each treatment line without considering the bigger wetlands. The two small wetlands (altogether) had a surface area (1.3 m2), which was approximately the 45% of the total surface area of the treatment line (2.95 m2) (bigger wetlands included). Note that clogging in this type of wetlands is more evident at the inlet zone of the wetlands (Caselles-Osorio et al., 2007), and this is why the total surface of the wetland area was split in two (one big and two small wetlands). Small wetlands (both planted and unplanted) were operated at the same hydraulic loading rate (HLR) (64.5 mm/d). For the batch line the hydraulic loading rate is comparable with the two other experimental lines when a four-day cycle is considered. The HLR for the whole experimental line (bigger wetlands considered) was that of 28.5 mm/d. Wetlands were planted in April 2007 with developed rhizomes of common reed and by July 2007 plants were well established and covered the entire surface of the wetlands. The uniform gravel layer (D60 ¼ 7.3 mm, Cu ¼ 0.83, 40% initial porosity) was 0.3 m deep and the water level was kept 0.05 m below the gravel surface to give a water depth of 0.25 m (shallow horizontal SSF CWs).

3581

Small wetlands of the anaerobic and control lines were fed 6 times a day and remained permanently saturated. On the contrary, small wetlands of the batch line operated alternating batch-unsaturated and permanently saturated phases following a four-days-cycle:  2 days of filling and discharging to the big wetland or to the sewer according whether it was a planted or an unplanted system, respectively. Effluent discharge was carried out when the water level within the wetlands reached 25 cm. During this phase the wetlands remained mostly saturated.  1 day saturated and resting. During this phase small wetlands did not receive influent for one day. At the end of this phase the wetlands were completely emptied by means of an electro valve.  1 day unsaturated and resting. During this phase wetlands did not receive wastewater. Every 4 days each small wetland of the batch line had received the same amount of water than the small wetlands of the anaerobic or the control line but concentrated within the first 2 days of the 4-day-cycle (Fig. 2). Results discussed in this paper consist of a three-yeardata set split in two experimental periods according to the operational conditions of the HUSB reactor. Accordingly, the HUSB reactor was operated at 3 h of hydraulic retention time (HRT) during the first experimental period (from February 2007 to December 2008) whereas during the second period it was operated at 5 h of HRT (from January 2009 to December 2009).

Fig. 2 e Schematic diagram of an operation cycle (4 days) of the batch line compared to the control and the anaerobic line (small wetlands). * For control and anaerobic lines only the operation of one of the two small wetlands is shown.

3582

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 9 e3 5 8 9

Table 1 e Sampling campaigns and clogging factors evaluated. Clogging factors

Small wetlands Campaign*

Solids accumulation Drainable porosity Saturated hydraulic conductivity Campaign Campaign Campaign Campaign Campaign Campaign

2.2.

Sample points

2, 3, 4, 6 3, 4, 5, 6

Gravel samplers The whole wetland in depth fractions of 5 cm 1, 2, 3, 4, 6 2 points at inlet and outlet

1: February 2007 (month 0). 2: September 2007 (month 7). 3: MayeJuly 2008 (month 15). 4: OctobereDecember 2008 (month 20). 5: AugusteSeptember 2009 (month 30). 6: December 2009eJanuary 2010 (month 34).

Clogging indicators

Clogging indicators were monitored during the entire study period (3 years). A total of six sampling campaigns (with a duration ranging from 1 to 3 months each) were carried out in order to assess clogging development at the experimental HSSF CWs. Table 1 summarizes the clogging indicators evaluated at each sampling campaign.

2.2.1.

Solids accumulation

At each small wetland (either planted or unplanted) there were three different types of samplers (gravel, plant and microbial samplers). However, for the purposes of this study only gravel samplers will be considered (Fig. 3). Gravel samplers consisted of two tubes fitted one inside the other, with a diameter of 7 and 8.5 cm, respectively. Tubes were made out of coiled galvanized stainless steel mesh and were inserted into the granular medium to the bottom during the construction of the wetlands. The mesh had a sieve size of 5 mm and, therefore, was able to contain the gravel. The inner tube was filled with gravel and left in the wetland until it was

taken out for sampling. Six gravel samplers were located along the whole width of the wetland and near the inlet (approximately at 15 cm from the water distributor) (Fig. 3). One to two gravel samplers were taken out at each sampling campaign. Samplers were taken out very carefully after complete drain of the wetlands. Gravel samplers were immediately replaced by new ones in order to avoid any disturbance of the filter media (note that only the original samplers were analysed for solids accumulation). Samplers were immediately transported to the laboratory and divided in 3 sections for accumulated solids analyses. Sections considered were: 0.05 m above the water level (which was not analyzed because it corresponded to the non-wetted media), top (0.15 m) and bottom (0.10 m) (Fig. 3). Afterwards, gravel of the considered sections was taken out and processed. Sample processing consisted of removing by hand all recognisable alive and dead roots. Gravel and interstitial solids (sludge) were separated by means of washing (hand shaking) with 1 L of distilled water. Water resulting from the cleaning process was further filtered through a 1 mm metal mesh according to the procedure described by Nguyen (2000). It must be noted that solids strongly adhered to the gravel were not considered because they represent <1% of the total sludge in HSSF CWs (Caselles-Osorio et al., 2007). Water used for gravel washing was analyzed for total solids (TS) and volatile solids (VS) according to APHA-AWWA-WPCF (2001). Plant belowground biomass separated during manual screening was dried at 105  C for 24 h and weighted. Note that, since the sieve for gravel washing had 1 mm pore space and sampler mesh had 5 mm, only roots of 1e5 mm of diameter were considered.

2.2.2.

2.2.3.

Fig. 3 e Schematic plant view of the small wetlands and location of the gravel, root and biofilm samplers (left), and schematic diagram of the gravel samplers (inner tube) and the different sections (right). Note that plant and biofilm samplers were not used for the purposes of the present study. Distances are shown in cm.

Drainable porosity

Drainable porosity represents the pore volume that freely drains in a gravitational field at atmospheric pressure, thus represents the pore volume readily available for wastewater to flow. In order to measure the drainable porosity at the small wetlands the effluent pipe was lowered at intervals of 0.05 m and the water volume discharged was measured (Rowe et al., 2000). For porosity determination the total wetland water volume measured was compared to the total initial volume of the wetland (65.25 L considering an initial porosity of 40%).

Saturated hydraulic conductivity

The measurement of the saturated hydraulic conductivity was carried out following a modification of the falling head method (FHM) Described by Pedescoll et al. (2009, 2010). The FHM consists of measuring the time a column of water that takes to leave a permeameter cell (a tube inserted into the gravel medium). The modification of the FHM described by Pedescoll et al. (2009) consisted of providing the water column to the permeameter cell by a balloon. Accordingly, the balloon was filled with water until it exploded inside the permeameter cell. This modified procedure was applied in order to avoid any flooding of the wetland due to both its high initial conductivity (484  142 m/d) and the smaller size of the system under study. For each wetland two points at the inlet and two at the outlet were considered for hydraulic conductivity measurements. Hydraulic conductivity was measured before the start

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 9 e3 5 8 9

up of the experimental plant, (initial hydraulic conductivity of the wetlands), and approximately every 6 months at the same locations.

2.3.

Physical-chemical analyses

Treatment performance was evaluated by taking water samples 3e4 times per month (from April 2007 to July 2009). Water samples were taken at the influent (storage tank), primary treatments (settlers and HUSB reactor) and effluent (big wetlands). From April to July 2007, samples were weekly analysed for pH, redox potential, turbidity, COD, TSS, ammonium and sulphates. In addition, from October 2007 to July 2009, BOD5, dissolved COD, nitrites, nitrates, TKN and total phosphorus were also analysed approximately once a month. Redox potential values were corrected for the potential of the hydrogen electrode. Analyses were carried out following the methods described in APHA-AWWA-WPCF (2001). Note that the effectiveness of each line was assessed on mass balance basis. A deep analysis on the performance of the systems in terms of pollutants removal is out of the scope of the present paper and will be published elsewhere. However, for the sake of discussion authors will give some details on plant performance (note that plant performance is evaluated at the end of the whole treatment line not at the effluent of the small wetlands).

2.4.

Statistical analysis

Differences between lines, planted and unplanted wetlands and temporal changes were assessed with two factors ANOVA test without replication. Pearson correlations between solids accumulation and porosity and hydraulic conductivity were conducted in order to find relations between clogging indicators. Statistical analyses were conducted using the SPSS 17.0 software package. Differences were considered significant at p < 0.05.

3.

Results

3.1.

Influent characterisation and general effectiveness

The HUSB reactor operated at 5 h of HRT produced effluents with significantly lower redox potential than the settlers (172  103 and 103  100 mV for settlers and HUSB reactor, respectively). TSS clearly decreased from raw wastewater to the effluents of the primary treatments, although significant differences between primary treatments were only observed during the second period (38% and 60% of TSS removal efficiency for settlers and HUSB reactor, respectively). Hydrolysis and solubilisation of particulate matter in the HUSB reactor was clearly favoured during the second period when dissolved COD was significantly higher (dissolved COD/total COD ratio was 75%) than in raw wastewater or settler effluent (50% and 58%, respectively). Average organic loading rates during the whole study period were 8.2  3.3 g COD/m2d and 4.7  1.4 g BOD/m2d for the control and batch lines, respectively (same primary treatment), and 8.8  3.7 g COD/m2d and 4.7  1.5 g BOD/m2d for the anaerobic line (with the HUSB reactor as primary treatment). Average solids loading rate was 2.88  1.35 g TSS/ m2d for the control and batch lines and 2.69  1.61 g TSS/m2d

3583

for the anaerobic line during the first period (HUSB reactor operated at 3 h of HRT). During the second experimental period (HUSB operated at 5 h HRT) the solids loading rate was 2.73  1.38 g TSS/m2d for the control and batch lines and 1.77  0.85 g TSS/m2d for the anaerobic line. The three treatment lines presented good removal efficiencies (on a mass balance basis). More precisely, removal was that of 80% for COD and ammonium and 90% for TSS and BOD5. For COD and ammonium a seasonal pattern was found, with significant higher removal efficiencies in warm periods. Differences between batch and control line were found, mainly in cold periods. Accordingly, the batch line presented 50% higher removal of ammonium than the control line in winter time. Anaerobic line presented the lowest removal efficiencies, mainly due to the reduced state of HUSB’s effluent.

3.2.

Accumulated solids

Accumulated sludge (interstitial solids of the gravel) and belowground plant biomass experienced an important increase through the experimental period, though plant biomass accumulation was especially important during the last year of operation, regardless the experimental line (Fig. 4b and c). Roots accumulation occurred mainly in the last 10 months at the bottom of the wetlands, while roots in the upper layer (the top 15 cm of wetted gravel) remained essentially constant through the experimental period (results not shown). Overall, temporal changes on total accumulated solids (considering both sludge and roots) showed a progressive increase over time for the 3 treatment lines (Fig. 4c). After three years of operation, unplanted wetlands of the anaerobic and the batch line accumulated significantly less sludge (ca. 2 kg DM/m2) than unplanted wetlands of the control line (ca. 5 kg DM/m2) (Fig. 4a). However, for planted wetlands accumulated sludge was of similar extent for the anaerobic and control lines (ca. 2 and ca. 2.5 kg DM/m2, respectively), whereas the batch line accumulated slightly higher sludge with ca. 4 kg DM/m2 (Fig. 4a). Furthermore, accumulated sludge was slightly higher at the bottom of the wetlands (representing from 47% to 67% of the total), regardless the experimental line (Fig. 5). Volatile solids represented, in all cases, less than 50% of total solids without differences between treatment lines, presence of plants or depth. However, the percentage of VS increased from the first to last sampling campaign in 12  8% (data not shown). Roots accumulation was significantly higher at the bottom of the wetlands regardless the experimental line (Fig. 5). Accordingly, roots at the bottom of the wetland (last 10 cm) represented between the 80% and the 90% of the total roots weight. It is important to point out that, even though only roots from 1 to 5 mm are considered, belowground biomass represented a great extent of the total solids accumulated within the wetland. More precisely, root biomass represented the 35% of the total solids of the batch line and around 50%e70% of the accumulated solids in control and anaerobic lines (Fig. 5).

3.3.

Drainable porosity

A progressive reduction of porosity over time was observed (Fig. 6a) in all the wetlands but with no significant differences

3584

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 9 e3 5 8 9

Fig. 4 e Temporal changes on accumulated solids (in terms of dry matter); a) sludge, b) roots and c) its addition, in the small wetlands of the experimental plant. * Data not available due to the lack of sample. between experimental lines. Note that Fig. 6a shows the percentage of porosity reduction considering an initial drainable porosity of 40% (this corresponds to the 100% in the figure). After 3 years of operation wetlands filter media presented a porosity reduction ranging from 15% to 20% for unplanted systems and ca. 30% for planted wetlands, regardless the type of primary treatment or flow regime. Therefore, reduction of drainable porosity was significantly

more important in planted wetlands (between 10% and 15% higher porosity than unplanted wetlands e Fig. 6a).

3.4.

Saturated hydraulic conductivity

Initial hydraulic conductivity was that of 484  142 m/d for all wetlands. Wetlands filter medium experienced an important hydraulic conductivity decrease (especially marked during the

Fig. 5 e Sludge and roots accumulated (in terms of dry matter) after three years of operation in the small wetlands of each treatment. * Data not available due to the lack of sample.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 9 e3 5 8 9

3585

Fig. 6 e Temporal changes on (a) porosity and hydraulic conductivity at inlet (b) and outlet (c) zones in the small wetlands of each treatment line. Note that for the sake of comparison values are plotted as a percentage of the initial ones (P/P0 and K/K0 for porosity and hydraulic conductivity, respectively).

first 2 years of operation). More precisely, hydraulic conductivity in planted wetlands after three years of operation was 44  14 m/d, 72  34 m/d and 43  19 m/d at the inlet zone, and 122  55 m/d 216  65 m/d and 98  23 m/d at the outlet zone of the control, batch and anaerobic lines, respectively. After three years of operation, hydraulic conductivity values at the outlet zone were higher than at the inlet (regardless the presence of plants or the treatment line). However, significant differences were only registered for the batch line (Fig. 6c). Moreover, hydraulic conductivity at the inlet decreased up to a 60% and a 90% for unplanted and planted wetlands, respectively, regardless type of primary treatment or flow regime. At the outlet, values decreased up to a 30% and between 60% and 80% for unplanted and planted beds, respectively. Therefore, hydraulic conductivity was significantly lower for planted than for unplanted wetlands (ca. 30% and ca. 50% lower at the inlet and outlet zones of the systems, respectively), regardless the type of primary treatment or hydraulic regime.

4.

Discussion

In the first part of the discussion section indirect measures of clogging (hydraulic conductivity and porosity) will be compared to accumulated solids in order to describe the

relationship between clogging indicators. Furthermore, the effect of the type of primary treatment and hydraulic regime on clogging development will be addressed in the second and third part of this discussion section, respectively, by comparing planted wetlands of the anaerobic and batch line with the control line. Finally, the role of macrophytes on clogging will be addressed in the last section of the discussion.

4.1.

Clogging indicators

Density properties of accumulated solids in SSF CWs (solids nature) can vary to a great extent according to the sampling location (Llorens et al., 2009). Sludge density and packing properties, in turn, may have a certain influence on both hydraulic conductivity and porosity values, regardless the amount of accumulated sludge (Nguyen, 2000). Despite that the relationship between solids accumulation and indirect measures of clogging (such as hydraulic conductivity) it is not straight forward (Caselles-Osorio et al., 2007; Knowles et al., 2010; Pedescoll et al., 2009; Tanner et al., 1998), Pearson correlation coefficients between clogging indicators measured over time in this study revealed a significant direct (and negative) correlation (Fig. 7). Roots and indirect measures of clogging (such as hydraulic conductivity or porosity) were also correlated (Fig. 7). However,

3586

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 9 e3 5 8 9

the correlation was less powerful than for the accumulated solids. One plausible explanation for the lack of a better correlation between roots and hydraulic conductivity or porosity is the type of roots analysed in this study. Accordingly, plant rhizomes larger than 5 mm may represent more than the 50% of the total belowground biomass (Edwards et al., 2006). Unfortunately, only small roots (diameters between 1 and 5 mm) were considered in this study. Therefore, authors suggest that better correlations could be found between clogging and hydraulic conductivity or porosity if large roots (>5 mm) were considered. Despite the encountered correlation between hydraulic measurements (porosity and hydraulic conductivity) and solids accumulated, no differences were found between treatment lines with regards to reduction of both porosity and hydraulic conductivity (Fig. 6). Drainable porosity might be considered the best strategy for indirect clogging assessment in experimental SSF CWs (namely because its correlation with accumulated solids is better). However, the procedure is of difficult applicability in full-scale facilities (namely because of

the problems coupled to a complete drain of a full-scale wetland). Therefore, and as has been previously described (Pedescoll et al., 2009), in full-scale facilities the measurement of hydraulic conductivity seems to be a more suitable procedure for indirect clogging assessment.

4.2.

Effect of primary treatment on clogging development

HUSB reactors have been described to significantly reduce the solids loading when treating domestic wastewater (Ruı´z et al., 2010). Our results are in agreement with this since the wetlands of the anaerobic line received ca. 20% and ca. 35% less solids than the control line for the whole experimental period and the second experimental period alone, respectively. A reduction of the solids load led to a lower sludge accumulation (roots not included) in the wetlands of the anaerobic line (Fig. 4a), which is in accordance to that described by Caselles-Osorio et al. (2007). According to literature, solids accumulation for wetlands presents a huge variation (even among those operated under similar conditions)

Fig. 7 e Scatter plots showing correlations among clogging indicators measured in this study. r, Pearson coefficient (negative values indicate negative correlation); d.f., degrees of freedom; a, significance level.

3587

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 9 e3 5 8 9

(Table 2). In spite of this variation, solids accumulation rates for the anaerobic line were ca. 30% lower than those of the control line and roughly between 2 and 10 times lower than those generally reported for systems implemented with septic tanks (Table 2). Concerning the analysis of hydraulic measurements, no significant differences were recorded between treatment lines. Accordingly, the hydraulic conductivity method, though to be a reliable tool for on-site clogging assessment (Pedescoll et al., 2009), is not able to discriminate between small hydraulic conductivity differences (close clogging scenarios) (Pedescoll et al., under revision). Therefore, differences between treatment lines on accumulated solids (ca. 2 kg DM/ m2) might still be too small (even after three years of operation) to be detected by both hydraulic conductivity method and drainable porosity (at least under the operational conditions here tested).

4.3.

Effect of operation strategy on clogging development

Author’s hypothesis was that alternating batch-unsaturated and permanently saturated phases may delay clogging due to an increase of oxygen transfer rates and subsequent enhancement of aerobic degradation of accumulated solids (Langergraber et al., 2003). The batch line was obviously under higher redox conditions than the rest of the experimental lines (note that ammonium removal in cold periods was about 50% higher than the control line). However, authors cannot confirm the proposed hypothesis since the amount of accumulated solids for the batch line was not different from the

rest of treatment lines (in some cases it was even slightly higher e Fig. 4a and c for the accumulated sludge and solids, respectively). Accordingly, one plausible explanation is that of, as has been already pointed out, higher oxygen transfer rates may enhance biofilm growth in SSF CWs (Chazarenc et al., 2009) which, in turn, may contribute to a higher sludge accumulation rather than favouring the aerobic degradation of accumulated solids. Although hydraulic conductivity values after 3 years of operation were not statistically different between treatment lines, the batch line presented slightly higher values than the control at the outlet of the system (ca. 10% higher) (Fig. 6c). This result is probably due to a lower root system development for the batch line (Fig. 4b).

4.4.

Effect of vegetation on clogging development

The root system biomass after three years of operation ranged from 2 to 5 kg DM/m2, regardless the type of primary treatment or the flow regime. These values are slightly higher than root biomasses recorded for common reed in gravel-based wetlands (Adcock and Ganf, 1994; Edwards et al., 2006). Furthermore, it has been described that the higher the belowground biomass of common reed, the closer to its optimal plant growth (Adcock and Ganf, 1994; Parr, 1990). Therefore, results suggest that macrophytes were in a very healthy state, regardless the experimental line considered. Roots development took place mainly at the bottom of the systems (root biomass within the last 10 cm accounted for 80%e90% of the total root biomass in the wetland). This is

Table 2 e TSS surface loading rates, accumulated solids, and solids accumulation rates in different SSF CWs studies. Study

Tanner and Sukias (1995) Tanner et al. (1998) Caselles-Osorio et al. (2007)

Chazarenc et al. (2009)

Ruı´z et al. (2010) This studyb

System

Primary treatment

Solids loading rate (g TSS/m2d)

Time (months)

Accum. solidsa (kg DM/m2)

Solids accum. Ratea (kg DM/m2 year)

2.2e7.3 2.2e7.3 2.8e4.5

22 60 48 48 36 48 36 36 60 60 60 60 36

1.9e6.5 8.5e18.6 2.8e12.8 2.3e11.9 2.6e35.1 6.0e57.3 2.3e9.6 2.8e20.3 36.0e44.0 44.0 43.0e44.0 55.5e66.0 3.2 0.90e2.13 1.15e2.25 2.05e4.02 0.83e1.63 1.55e3.13 2.53e4.76 0.79e1.03 0.93e1.41 1.97e2.20

1.5e4.5 1.3e3.0 0.7e3.2 0.6e2.9 0.6e8.8 1.5e14.3 0.8e3.2 0.9e6.8 7.2e8.8 8.2 8.6e8.8 11.1e13.2 1.07 0.32e0.75 0.41e0.79 0.72e1.42 0.29e0.57 0.55e1.11 0.89e1.68 0.28e0.36 0.33e0.50 0.69e0.78

Verdu´ 1 Verdu´ 2 Alfe´s Corbins Almatret north Almatret south Planted aerated Planted non aerated Unplanted aerated Unplanted non aerated Santiago de Compostela

Septic tank Septic tank Septic tank Imhoff tank Septic tank Septic tank

UASB

3.2e4.9 6.5e10 4.5e6.2 2.6e8.0 3.2 3.2 3.2 3.2 5.9  6.5

Batch line

Settler

0.60e7.43

36

Control line

Settler

0.60e7.43

36

Anaerobic line

HUSB reactor

0.67e5.03

36

a Interstitial solids (sludge) are considered. b Planted and unplanted wetlands are considered, thus solids accumulation rates in this study correspond to the first third of the entire planted wetlands. For each line, sludge from the top 15 cm, the bottom 10 cm and the total depth are shown (by order).

3588

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 9 e3 5 8 9

probably favoured by the shallowness of the beds, allowing the roots to penetrate to the bottom, which is only 25 cm deep. It is generally accepted that roots colonise the wetland in the firsts 20 cm from the surface (Parr, 1990; Reed et al., 1995; U.S. EPA, 1993). Furthermore, roots accumulation in planted wetlands represented up to the 70% of the total accumulated solids at the end of the study period (35%, 50% and 70% for the batch, anaerobic and control line, respectively) (Fig. 5). Although the contribution of plant detritus to solids accumulation in SSF CWs has been described to be important (Nguyen, 2000; Tanner et al., 1998), no significant differences on sludge accumulation (roots not considered) between planted and unplanted wetlands were recorded (regardless treatment line) (Fig. 4a). This is probably due to the fact that aboveground biomass was harvested after plant decay took place. Therefore, maintenance tasks are of capital importance for clogging prevention. It would be interesting to study the impact of the plants if they had not been harvested, and the test had run for longer. Perhaps the plants would have had even more damaging impact. Furthermore, the batch line showed a less developed root system when compared to the control line. Accordingly, it has been described that alternating saturated and unsaturated phases (batch operation) may cause a stress to the macrophytes (Tanner et al., 1999) that, eventually, may lead to a lower development of both above and belowground biomass (Sasikala et al., 2009). Although macrophytes production has been related to nutrient uptake (Brix, 1997), we have no empirical evidence that a less developed root system in the batch line is coupled to a reduction of treatment efficiency. In fact, the batch line performed significantly better than the rest of treatment lines for most of the analysed pollutants (especially for ammonium). On the other hand, contrary to earlier believe, the presence of macrophytes do not increase hydraulic conductivity in soilbased constructed wetlands (Brix, 1997) but they can even contribute to a notable extent to the reduction of the effective retention time in reed beds (Tanner and Sukias, 1995). Our results confirm the role of plants (belowground biomass) on hydraulic conductivity decrease. Accordingly, root biomass represented the 10% and 30% of the variation of drainable porosity and hydraulic conductivity, respectively (comparing the reduction between planted and unplanted beds). Therefore, the presence of well developed rizhomes (larger than 5 mm) may cause even a higher reduction of hydraulic conductivity or porosity in SSF CWs. Unfortunately, we cannot determine the extent of the contribution of large roots to hydraulic conductivity or porosity reduction, since only roots smaller than 5 mm were considered in this study. Therefore, the extent of the effect of the root system on hydraulic conductivity reduction in SSF CWs must be further investigated, especially for other macrophyte species whose root system growth is less pronounced than common reed.

5.

Conclusions

Despite the lack of differences in hydraulic parameters between treatment lines, an SSF CW implemented with a HUSB reactor as primary treatment received ca. 30% less

solids than the SSF CW implemented with a settler. As a consequence the HUSB line accumulated ca. 30% lower sludge than a system implemented with a conventional settler. Therefore, the implementation of HUSB reactor is recommended in order to delay solids accumulation in SSF CWs. An SSF CW operated under batch regime accumulates slightly higher sludge than an SSF CW operated under continuous hydraulic regime, probably as a consequence of biofilm growth stimulation. However, because root system in a batch operated systems is less developed than in a continuous operated system (probably as a consequence of great hydric stress on macrophytes), overall differences on solids accumulation between both feeding strategies are not significantly different after three years of operation. The root system of common reed greatly contributed to clogging. Accordingly, after three years of operation roots accumulation in planted wetlands represented 35%e70% of the total accumulated solids. Hydraulic conductivity and drainable porosity are well correlated to accumulated solids (74.5% and 89.2%, respectively). Therefore, they can be considered reliable techniques for indirect measurements of clogging.

Acknowledgements This study was made possible by funding from the Spanish Ministry of Innovation and Science for the NEWWET 2008 Project (CTM2008-06676-C05-01). Ange´lica Corzo and Anna Pedescoll acknowledge predoctoral fellowships of the Management of University and Research Grants Agency (AGAUR) of Catalonia and the Ministry of Science, respectively. The authors are particularly grateful to Enrica Uggetti, Gemma Lesan, Gian Paolo Mottola, Fedro Tapia, Toni Lara, Carlos Soriano, Begon˜a G. Admirable, Roger Samso´, Lina Tyroller, PlazadelosReyes, Leo Vera, Elif Bozdogan, Javier Carretero, Adriana Gallardo and Miriam Planas for their help in this study (contributing in the construction of the experimental plant, maintenance or laboratory tasks).

references

Adcock, P.W., Ganf, G.G., 1994. Growth characteristics of three macrophyte species growing in a natural and constructed wetland system. Water Science and Technology 29 (4), 95e102. ´ lvarez, J.A., Armstrong, E., Go´mez, M., Soto, M., 2008. Anaerobic A treatment of low-strength municipal wastewater by a twostage pilot plant under psychrophilic conditions. Bioresource Technology 99 (15), 7051e7062. APHA-AWWA-WPCF, 2001. Standard Methods for the Examination of Water and Wastewater (2001), twentieth ed. (Washington DC, USA). Barros, P., Ruı´z, I., Soto, M., 2008. Performance of an anaerobic digester-constructed wetland system for a small community. Ecological Engineering 33, 142e149. Bezbaruah, A.N., Zhang, T.C., 2005. Quantification of oxygen release by bulrush (Scirpus validus) in a constructed treatment wetland. Biotechnology and Bioengineering 89 (3), 308e318.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 7 9 e3 5 8 9

Brix, H., 1997. Do macrophytes play a role in constructed treatment wetlands? Water Science and Technology 35 (5), 1e17. Brix, H., Arias, C.A., 2005. The use of vertical flow constructed wetlands for on-site treatment of domestic wastewater: new Danish guidelines. Ecological Engineering 25, 491e500. Caselles-Osorio, A., Puigagut, J., Segu´, E., Vaello, N., Granes, F., Garcı´a, D., Garcı´a, J., 2007. Solids accumulation in six full-scale subsurface flow constructed wetlands. Water Research 41, 1388e1398. Chazarenc, F., Gagnon, V., Comeau, Y., Brisson, J., 2009. Effect of plant and artificial aeration on solids accumulation and biological activities in constructed wetlands. Ecological Engineering 35 (6), 1005e1010. Cooper, D.J., Griffin, P., Cooper, P.F., 2005. Factors affecting the longevity of subsurface horizontal flow systems operating as tertiary treatment for sewage effluent. Water Science and Technology 51 (9), 127e135.  z  kova´, H., 2006. kova´, H., Zemanova´, K., Santr  Edwards, K.R., Ci uc Plant growth and microbial processes in a constructed wetland planted with Phalaris arundinacea. Ecological Engineering 27 (2), 153e165. Faulwetter, J.L., Gagnon, V., Sundberg, C., Chazarenc, F., Burr, M. D., Brisson, J., Camper, A.K., Stein, O.R., 2009. Microbial processes influencing performance of treatment wetlands: a review. Ecological Engineering 35, 987e1004. Knowles, P.R., Griffin, P., Davies, P.A., 2010. Complementary methods to investigate the development of clogging within a horizontal sub-surface flow tertiary treatment wetland. Water Research 44 (1), 320e330. Knowles, P., Nivala, J., Dotro, G., Garcı´a, J., 2011. Clogging in subsurface-flow treatment wetlands: occurrence and contributing factors. Ecological Engineering 37, 99e112. Langergraber, G., Haberl, R., Laber, J., Pressl, A., 2003. Evaluation of substrate clogging processes in vertical flow constructed wetlands. Water Science and Technology 48 (5), 25e34. Llorens, E., Puigagut, J., Garcı´a, J., 2009. Distribution and biodegradability of sludge accumulated in full-scale horizontal subsurface-flow constructed wetland. Desalination and Water Treatment 4, 54e58. Mun˜oz, P., Drizo, A., Cully Hession, W., 2006. Flow patterns of dairy wastewater constructed wetlands in a cold climate. Water Research 40, 3209e3218. Nguyen, L.M., 2000. Organic matter composition, microbial biomass and microbial activity in gravel-bed constructed wetlands treating farm dairy wastewaters. Ecological Engineering 16 (2), 199e221. Parr, T.W., 1990. Factors affecting reed (phragmites australis) growth in UK reed bed treatment systems. Adv. Wat. Pollut. Control no. 11. In: Cooper, P.F., Findlater, B.C. (Eds.), Constructed Wetlands and Water Pollution Control. Proceedings of the International Conference on the Use of Constructed Wetlands in Water Pollution Control. Pergamon Press, Oxford, pp. 383e389. Pedescoll, A., Uggetti, E., Llorens, E., Grane´s, F., Garcia, D., Garcı´a, J., 2009. Practical method based on saturated hydraulic conductivity used to assess clogging in subsurface flow constructed wetlands. Ecological Engineering 35 (8), 1216e1224. Pedescoll, A., Samso´, R., Romero, E., Puigagut, J., Garcı´a, J., 2010. Reliability, accuracy and repeatability of the falling head method for hydraulic conductivity measurements under laboratory conditions. Ecological Engineering. doi:10.1016/j. ecoleng.2010.06.032.

3589

Pedescoll, A., Knowles, P.R., Davies, P., Garcı´a, J., Puigagut, J. A comparison of in situ constant and falling head permeameter tests to assess the distribution of clogging within horizontal subsurface flow constructed wetlands. Journal of Hazardous Materials, under revision.. Reed, S.C., Crites, R., Middlebrooks, E.J., 1995. Natural Systems for Waste Management and Treatment, 2nd 1459 edition. McGraw-Hill, New York. Rousseau, D.P.L., Horton, D., Vanrolleghem, P.A., De Pauw, N., 2005. Impact of operational maintenance on the asset life of storm reed beds. Water Science and Technology 51 (9), 243e250. Rowe, K.R., Armstrong, M.D., Cullimore, D.R., 2000. Particle size and clogging of granular media permeated with leachate. ASCE Journal of Geotechnical and Geoenvironmental Engineering 126 (9), 775e786. Ruı´z, I., Dı´az, M.A., Crujeiras, B., Garcı´a, J., Soto, M., 2010. Solids hydrolysis and accumulation in a hybrid anaerobic digesterconstructed wetlands system. Ecological Engineering 36 (8), 1007e1016. Sandford, W.E., Steenhuis, T.S., Parlange, J.Y., Surface, J.M., Peverly, J. H., 1995. Hydraulic conductivity of gravel sand as substrates in rock-reed filters. Ecological Engineering 4 (4), 321e336. Sasikala, S., Tanaka, N., Wah Wah, H.S.Y., Jinadasa, K.B.S.N., 2009. Effects of water level fluctuation on radial oxygen loss, root porosity, and nitrogen removal in subsurface vertical flow wetland mesocosms. Ecological Engineering 35 (3), 410e417. Suliman, F., French, H.K., Haugen, L.E., Svik, A.K., 2006. Change in flow and transport patterns in horizontal subsurface flow constructed wetlands as a result of biological growth. Ecological Engineering 27, 124e133. Tanner, C.C., Sukias, J.P., 1995. Accumulation of organic solids in gravel bed constructed wetlands. Water Science and Technology 32 (3), 229e239. Tanner, C.C., Sukias, J.P.S., Upsdell, M.P., 1998. Organic matter accumulation and maturation of gravel bed constructed wetlands treating dairy farm wastewaters. Water Research 32 (10), 3046e3054. Tanner, C.C., D’Eugenio, J., McBride, G.B., Sukias, J.P.S., Thompson, K., 1999. Effect of water level fluctuation on nitrogen removal from constructed wetland mesocosms. Ecological Engineering 12 (1e2), 67e92. Tchobanoglous, G., 2003. Preliminary treatment in constructed wetlands. In: Dias, V., Vymazal, J. (Eds.), Proceedings of the 1st International Seminar on the Use of Aquatic Macrophytes for Wastewater Treatment in Constructed Wetlands, 8e10 May 2003. Instituto da Conservacao da Natureza and Instituto da Agua, Lisbon, Portugal, pp. 13e33. Tyroller, L., Rousseau, D.P.L., Santa, S., Garcı´a, J., 2010. Application of the gas tracer method for measuring oxygen transfer rates in subsurface flow constructed wetlands. Water Research 44 (14), 4217e4225. U.S. EPA, 1993. Design Manual No. 74: Subsurface Flow Constructed Wetland for Wastewater Treatment: a Technology Assessment, EPA 832/R-93/008. U.S. EPA Office of Water. Wallace, S.D., Knight, R.L., 2006. Small-scale Constructed Wetland Treatment Systems. In: Feasibility, Design Criteria and O&M Requirements. Final Report, Project 01-CTS-5. Water Environment Research Foundation (WERF), Alexandria, USA. Zhao, Y.Q., Sun, G., Lafferty, C., Allen, S.J., 2006. Optimising the performance of a lab-scale tidal flow reed bed system treating agricultural wastewater. Water Science and Technology 50 (8), 65e72.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 0 e3 5 9 8

Available at www.sciencedirect.com

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

Release of infectious human enteric viruses by full-scale wastewater utilities Fredrick James Simmons, Irene Xagoraraki* Department of Civil and Environmental Engineering, Michigan State University, East Lansing, MI 48824, USA

article info

abstract

Article history:

In the United States, infectious human enteric viruses are introduced daily into the envi-

Received 27 October 2010

ronment through the discharge of treated water and the digested sludge (biosolids). In this

Received in revised form

study, a total of 30 wastewater and 6 biosolids samples were analyzed over five months

17 February 2011

(MayeSeptember 2008e2009) from five full-scale wastewater treatment plants (WWTPs) in

Accepted 3 April 2011

Michigan using real-time PCR and cell culture assays. Samples were collected from four

Available online 19 April 2011

different locations at each WWTP (influent, pre-disinfection, post-disinfection and biosolids) using the 1MDS electropositive cartridge filter. Adenovirus (HAdV), enterovirus (EV) and

Keywords:

norovirus genogroup II (NoV GGII) were detected in 100%, 67% and 10%, respectively of the

Wastewater

wastewater samples using real-time PCR. Cytopathic effect (CPE) was present in 100% of the

Infectious viruses

cell culture samples for influent, pre- and post-disinfection and biosolids with an average log

Human adenovirus

concentration of 4.1 (2.9e4.7, range) 1.1 (0.0e2.3, range) and 0.5 (0.0e1.6, range) MPN/100 L

Human enterovirus

and 2.1 (0.5e4.1) viruses/g, respectively. A significant log reduction in infectious viruses

Human norovirus

throughout the wastewater treatment process was observed at an average 4.2 (1.9e5.0,

Cell culture

range) log units. A significant difference ( p-value <0.05) was observed using real-time PCR

MBR

data for HAdV but not for EV ( p-value >0.05) removal in MBR as compared to conventional

Electropositive cartridge filter

treatment. MBR treatment was able to achieve an additional 2 and 0.5 log reduction of HAdV and EV, respectively. This study has demonstrated the release of infectious enteric viruses in the final effluent and biosolids of wastewater treatment into the environment. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Human enteric viruses are currently listed on the United States Environmental Protection Agency Contaminant Candidate List (USEPA CCL) as emerging contaminants. To this date, no regulations have been implemented into the monitoring of wastewater viral concentration before being discharged into a natural water body. Human Adenovirus (HAdV), Human Enterovirus (EV), Norovirus Genogroups I, II and IV, (NoV GGI) and (NoV GGII), and Hepatitis-A (HAV) are some of the enteric viruses of concern (Gerba et al., 2002; Haramoto et al., 2007; Kittigul et al., 2006). These viruses have been related to

several waterborne diseases, such as acute gastroenteritis, conjunctivitis and respiratory illness in both developed and developing countries world-wide. There are several routes whereby the public can become infected, including direct contact (fecal-oral route or dermal contact) and food borne illness and contamination (Godfree and Farrell, 2005). Virus removal from wastewater continues to receive attention due to the epidemiological significance of viruses as waterborne pathogens and because of the high diversity that is excreted in human waste (Rose et al., 1996). A large number of enteric viruses are excreted in human feces and urine, which makes wastewater one of the most concentrated sources of

* Corresponding author. A124 Engineering Research Complex, Michigan State University, East Lansing, MI 48824, USA. Tel.: þ1 517 353 8539; fax: þ1 517 355 0250. E-mail address: [email protected] (I. Xagoraraki). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.001

3591

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 0 e3 5 9 8

these viruses. During the peak of an infection, it has been reported that enteric viruses are often detected in feces at elevated levels averaging 1011 viruses/gram (Rose et al., 1996). According to the literature, the concentrations of infectious and non-infectious HAdV, EV and NoV detected in untreated and treated wastewater are approximately 104e109 and 102e107 viruses/L, respectively (Bofill-Mas et al., 2006; Carducci et al., 2008; Kuo et al., 2010; Laverick et al., 2004; Rodriguez et al., 2008). To ensure the safety of the public, inactivation of viruses is usually achieved with chlorination and ultraviolet (UV) treatment before discharged. Previous bench-scale studies (Rodriguez et al., 2008; Simonet and Gantzer, 2006a,b) have determined the inactivation of viruses from the influent and final effluent of approximately 1e4 log units and 0.1e1.2 log units for the chlorination and UV disinfection unit processes alone. In addition to the final effluent, a considerable amount of sludge is generated in primary and secondary settling tanks during the treatment process. Depending on the level of sludge treatment, such as mesophilic anaerobic digestion (MAD) and lime stabilization, biosolids are often considered class A or B and depending on the particular crop and its intended use. In the US, approximately 5 million tons of dry biosolids is generated annually and 60% is used for agricultural land application to provide additional nutrients for crops. Class B biosolids are the most commonly produced in the United States by MAD (Gerba et al., 2002; Viau and Peccia, 2009). It has been stated that a variable fraction, as high as 50% of the enteric virus present in the raw sewage, may be associated with the solids (Payment et al., 1986), suggesting that the concentration of viruses in biosolids can be higher than in wastewater. Several studies have reported the occurrence of enteric viruses in biosolids after the digestion process (Monpoeho et al., 2004; Bofill-Mas et al., 2006; Guzman et al., 2007; Viau and Peccia, 2009; Wong et al., 2010). Past studies (Bofill-Mas et al., 2006; Katayama et al., 2008; Rodriguez et al., 2008; Laverick et al., 2004; Carducci et al., 2008; da Silva et al., 2007) have determined the concentration of DNA/RNA viruses using real-time PCR in addition to viral infectivity in wastewater treatment (Aulicino et al., 1995; Petrinca et al., 2009; Rodriguez et al., 2008; Sedmak et al., 2005). However to our knowledge, no studies have looked at the

overall release of enteric viruses from a full-scale wastewater treatment through the final effluent and biosolids using both real-time PCR and cell culture methods. In the current study, we analyzed a total of 30 wastewater and 6 biosolids samples over five different months (MayeSeptember) from five separate full-scale WWTPs to determine the release of enteric viruses. The results in this study provide important information on the overall release of both infectious and noninfectious enteric viruses following treatment. The objectives of this study were to (i) determine the concentration of enteric viruses within the wastewater treatment process using real-time PCR data (ii) determine the release of infectious viruses in the final effluent and biosolids, (iii) compare virus removal efficiency between MBR and conventional wastewater treatment process using real-time PCR data and (iv) compare the effectiveness of two different disinfection processes for virus inactivation. Four different sampling points (raw, pre-disinfection, post-disinfection and biosolids) were chosen to determine how viruses are removed and inactivated during treatment. The viruses studied were Human Adenovirus F40 and F41 (HAdV), Human Enterovirus (EV), Norovirus Genogroup 1 (NoV GI), Norovirus Genogroup 2 (NoV GGII) and Hepatitis-A (HAV).

2.

Methods and materials

2.1.

Wastewater treatment plants

Five different wastewater treatment plants (WWTPs) in Michigan’s Lower Peninsula were sampled from 07/17/2008e09/24/ 2009 in duplicate during separate sampling events. Four different locations were sampled: including influent (raw sewage), pre-disinfection (after secondary biological treatment), post-disinfection (final effluent) and biosolids. Table 1 lists the characteristics of each WWTP that was sampled.

2.2.

Wastewater and biosolids sampling

Thirty wastewater samples were collected using the 1MDS electropositive filter during 10 different sampling events following the procedure explained in the USEPA Manual of

Table 1 e Characteristics of the different WWTPs used in this study. Wastewater Treatment Process (Biological Treatment)

Average Flow (MGD)

Capacity (MGD)

Disinfection

Sludge Treatment

Sludge Production (gal/day)

Disposal of Biosolids

UV Chlorination UV

MAD Dewatering Lime Stabilization Gravity Thickening MAD

4500 15955 ea

Land Application Landfill Land Application

ea

Land Application

1369

Land Application

WWTP 1 2 3

MBR Activated Sludge Activated Sludge

4.0 12.5 17.0

17.0 19.0 20.0

4

Oxidation Ditch

0.2

0.4

UV

5

Rotating Biological Contactors

0.8

2.2

Chlorination

MAD e Mesophilic Anaerobic Digestion. a No biosolids were collected from these utilities.

3592

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 0 e3 5 9 8

Methods for Virology (USEPA, 2001). Approximately 20 L of influent, 375 L of pre-disinfection and 410 L of postdisinfection (final effluent) were sampled at a rate of about 11e12 L/min (3 gal/min). Biosolids samples were collected at three of the five different WWTPs (Table 1). Two L grab samples were collected from the post digestion holding tanks. The dewatered samples were collected from the exiting conveyor belt in the loading bay. All samples collected were then stored on ice and transported to the Water Quality Engineering Laboratory at Michigan State University. Upon arrival, samples were placed in a 4  C cooler before processing.

2.3.

Virus elution process for 1MDS filters and biosolids

All wastewater samples collected were eluted 12e24 h after initial sampling according to the Concentration and Processing of Waterborne Viruses by Positive Charge 1MDS Cartridge Filters and Organic Flocculation (USEPA, 2001) and previously described by Simmons et al., 2011, and Kuo et al., 2010. The biosolids virus elution and concentration were performed according to the ASTM-4994.

2.4.

Nucleic acid extraction

Viral samples were extracted using the MagNa Pure Compact System automatic machine (Roche Applied Sciences, Indianapolis, IN). 1000 mL of sample was extracted and concentrated to a final volume of 100 mL. Immediately following the completion of the extraction all samples were placed in a 80  C freezer. Following extraction the quantity of viral nucleic acid extracts from all samples were checked using the NanoDrop Spectrophotometer (NanoDrop ND-1000, Wilmington, DE).

2.5. Real-time PCR standard curves, sequencing and detection limit The standard curves for sample quantification of HAdV, EV, NoV GGI, NoV GGII and HAV were created using stock cultures

of HAdV 40 (ATCC VR-930), EV Coxsackie virus B5 (ATCC VR1036AS/MK), HAV HM175 (ATCC VR-1402) and NoV GGII stool samples were supplied by the Ingham County Health Department following a confirmed outbreak at Michigan State University. All standard curve assays performed used the LightCycler 1.5 Instrument (Roche Applied Sciences, Indianapolis, IN). Briefly, the PCR amplicons from HAdV, EV, NoV GGII and HAV from pure culture and stool sample extracts were cloned into a plasmid vector (i.e., pCR4-TOPO) which follows the oneshot chemical transformation described in the manufacturer instructions (TOPO TA Cloning Kit for Sequencing, Invitrogen, Carlsbad, CA). The plasmids carrying the cloned HAdV, EV, NoV GGI, NoV GGII and HAV were purified using Wizard Plus SV Minipreps DNA Purification System (Promega, Madison, WI) and sent for sequencing at the Research Technology Support Facility at Michigan State University. All target gene sequences were compared with those published in the National Center for Biotechnology Information (NCBI) database by using the program of Basic Local Alignment Search Tool (BLAST). The concentrations were determined by using the NanoDrop spectrophotometer and then the samples were serial diluted 10-fold (101e108 viruses/reaction) and used for creating the standard curves for all target viruses. The Rsquared values for each standard curve for HAdV, EV, NoV GGI, NoV GGII and HAV are 0.995, 0.996, 0.995, 0.999 and 0.996, respectively. All standard curve reactions were run in triplicate and the detection limit for EV and NoV GGII, is 10 viruses/ reaction and 100 viruses/reaction for HAdV, HAV and NoV GGI.

2.6.

Quantitative real-time PCR assays

The crossing point (Cp) value for each PCR reaction was automatically determined by the LightCycler Software 4.0 and used to calculate the overall viral concentration. The primer and probe sequences and reaction conditions used in this study are summarized in Table 2. Briefly, all real-time PCR reaction mixes included 10 mL of 2X LightCycler 480 TaqMan

Table 2 e List of enteric virus primer and probes, gene regions, reaction conditions and references used in this study. Virus Type HAdV

EV

NoV GGI

NoV GGII

Hep-A

Gene Region Hexon

Primers/Probes

Forward Reverse-1 Rsverse-2 Probe 50 e Untranscribed Forward Region Reverse Probe 50 e Untranscribed Forward Region Reverse Probe Junction 0RF1-0RF2 Forward Reverse Probe Junction 0RF1-0RF2 Forward Reverse Probe

Sequence (50 e30 ) ACCCACGATGTAACCACAGAC ACTTTGTAAGAGTAGGCGGTTTC CACTTTGTAAGAATAAGCGGTGTC CGACKGGCACGAAKCGCAGCGT ACATGGTGTGAAGAGTCTATTGAGCT CCAAAGTAGTCGGTTCCGC TCCGGCCCCTGAATGCGGCTAAT CGCTGGATGCGNTTCCAT CCTTAGACGCCATCATCATTTAC TGGACAGGAGAYOGCRATCT CARGASBCNATGTTYAGRTGGATGAG TCGACGCCATCTTCATTCACA TGGGAGGGCGATCGCAATCT GGTAGGCTACJGGGTGAAAC AACAACTCACCAATATCCGC CTTAGGCTAATACTTCTATG AAGAGATGC

Reaction Condition (temp, time)

Reference

95, 10 s e denaturation Xagoraraki et al., 2007 60, 30 s e annealing Modified from Jiang 72, 12 s e extension et al., 2005 95, 15 s e denaturation Dierssen et al., 2007 60, 60 s e annealing 95, 15 s e denaturation da Silva et al., 2007 60, 60 s e annealing 95, 15 s e denaturation Kageyama et al., 2003 56, 60 s e annealing 95. 10 s e denaturation Jothikumar et al., 2005 55, 20 s e annealing 72,15 s e extension

3593

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 0 e3 5 9 8

Master Mix and the appropriate volume of primers and as previously described (Dierssen et al., 2007; Jothikumar et al., 2005; Kageyama et al., 2003; da Silva et al., 2007; Xagoraraki et al., 2007). The real-time PCR running program (all thermocycles were performed at a temperature transition rate of 20  C/s) was 95  C for 15 min; followed by different cycles of denaturation, annealing, extension and a final cooling step. Each reverse transcription reaction mix for EV, NoV and HAV included 2.5 mL of 10 mM reverse primer, 1 mL of reverse transcriptase (Promega Corporation, Madison, WI), 4 mL of 5X transcriptor reaction buffer, 20 U of protector Rnase inhibitor, and 2 mL of 10 mM deoxynucleotide (Roche Applied Sciences, Indianapolis, IN). Reaction conditions for all three RNA viruses were the same; initial incubation at 55  C for 30 min followed by 85  C for 5 min to inactivate the enzyme. All samples were run in triplicate and included a negative control reaction (PCR grade-H2O without template) and a positive control reaction for all viruses.

2.7.

Inhibition control

To determine if inhibition occurred during viral analysis, the methods previously explained (Viau and Peccia, 2009; Rajal et al., 2007) were used. Bovine Enterovirus was chosen as the virus to spike all samples to determine if inhibition was present, bovine enterovirus was quantified following the methods previously published (Jimenez-Clavero et al., 2005). Prior to the inhibition check, all samples were analyzed for bovine enterovirus using real-time PCR. Next, all extracted samples and molecular grade-H2O were spiked with a concentration of 105 viruses/reaction of bovine enterovirus. Following the analysis, the Cp values of the extracted water and wastewater samples were recorded. If the Cp values of both the spiked water and wastewater samples were within an acceptable level (5%), we assumed that inhibition did not affect our analysis.

2.8.

Calculations of enteric virus concentration

All real-time PCR assays were converted from viruses/reaction to viruses/L or viruses/dry gram, using the following equations: Viruses 1 Reaction 1   100 mL   30; 000 mL Viruses Reaction 5 mL 1000 mL ¼ Initial Sample Volume L (1) Viruses 1 Reaction 1   100 mL  Viruses Reaction 5 mL 25 grams ¼ % Solids dry gram

(2)

In Eq. (1), the 5 mL is the amount of sample per reaction tube, 1000 and 100 mL is the amount of sample extracted and the volume of the extract, respectively. The 30,000 mL is the amount of concentrated eluent after the final filtration through a 0.22 mm syringe filter (Millipore, Billerica, MA). In Eq. (2), the 25 g is the weight of biosolids that were concentrated.

2.9.

Cell culture

In total 15 different wastewater and 3 different biosolids samples were analyzed for virus infectivity and were cultured

using BGM cell line, which were graciously donated by Shay Fout from the USEPA (passage #157). In addition, A549 cell line (ATCC CCL-185, passage #126) was used for the postdisinfection samples to determine the final infectious virus concentration being released. Including all dilutions and replicates, a total of 200 BGM and 50 A549 flasks were used for the final analysis. All samples followed the USEPA Total Culturable Virus Assay detailed in the “Information Collection Rule”.

2.10.

Log removal

Overall log removal achieved by the MBR and conventional WWTPs was calculated using Eq. (3): Log Removal ¼ log10

Influent Concentration Effluent Concentration

(3)

For pre- and post-secondary treatment samples that were below the detection limit, the log removal values were calculated by using the detection limit of the individual assays. This indicates that certain removal values may be greater than reported. However, this will allow for conservative calculations when this value is needed for comparisons.

2.11.

Statistical analysis

Log removal values for each WWTP were analyzed using t-test in Microsoft Excel using an alpha value (a-value of 0.05), showing a 95% confidence interval.

3.

Results

3.1.

Inhibition control

Bovine enterovirus was not initially detected in the 30 wastewater (00/30) and 6 biosolids (00/06) samples. All 36 samples were then spiked with 105 viruses/reaction of bovine enterovirus following extraction including a PCR grade-H2O. The Cp values for the WWTPs and biosolids samples were 26.57 (std 0.14) and for H2O samples, 26.67 (std 0.08) which are within 2% of each other. This indicates that any inhibition that may be present in the extracted samples was not able to suppress the detection of the viruses in this study.

3.2. Quantification of human enteric viruses in wastewater 3.2.1.

HAdV

Fig. 1 shows the average concentration of HAdV from the two different sampling events at each WWTP. HAdV was detected in all 30 samples, with an average influent concentration of 7.3  107 (2.0  104e6.7  108, range) viruses/L, average predisinfection of 8.7  103 (1.4  102e4.5  104, range), and 3.7  103 (1.1  101e2.9  104, range) viruses/L for the postdisinfection samples. WWTP 4 had the lowest average concentration of approximately 8.3  104, 2.6  102 and 1.0  102 viruses/L in the influent, pre- and post-disinfection, respectively. It is possible that the lower concentration could be due to the average daily flow into the plant, which receives the lowest of all 5 plants sampled (Table 1). WWTP 4 is located

3594

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 0 e3 5 9 8

ten influent samples from WWTPs 1 and 2 were positive for NoV GGII with concentrations ranging from 5.2  104 to 1.1  106 (average 4.3  105) viruses/L. However, NoV GGII was not detected in any of the pre- or post-disinfection samples.

3.2.4.

HAV

HAV was not detected in the 30 samples analyzed.

3.3.

Fig. 1 e Average (n [ 2) HAdV real-time PCR virus concentration detected at the five different WWTPs at each sampling point. MBR e Membrane Bioreactor, AS e Activated Sludge, OD e Oxidation Ditch, RBC e Rotating Biological Contactors, UV e ultraviolet, Cl e chlorination.

within a rural city; where most residential homes use individual septic tanks instead of the city sewer system.

3.2.2.

The average HAdV and EV log concentrations are 4.1 (BDL e 7.8, range) and 2.9 (BDL e 4.9, range) viruses/g, respectively. Neither HAdV nor EV was detected in the biosolids at WWTP 1. Interestingly, both NoV GGI and NoV GGII were detected in all biosolids samples but were not detected in the influent, pre- or post-disinfection wastewater samples. NoV GGI was detected in 06/06 biosolids samples at an average log concentration of 4.3 viruses/g (2.4e6.6, range). NoV GGII was also detected in 06/ 06 biosolids samples at an average 5.2 (3.6e7.4, range) viruses/g. HAV was not detected in any of the biosolids samples analyzed.

3.4.

Infectivity of viruses

3.4.1.

Wastewater samples

EV

Fig. 2 shows the average concentration for the two sampling events at each WWTP. In total 10/10 influent, 04/10 predisinfection and 06/10 post-disinfection samples were positive for EV. The influent EV average concentration for all samples is 2.1  105 (3.0  102e1.1  106, range) from viruses/L. The overall concentration detected in the pre-disinfection samples averaged 2.9  102 (below detection limit (BDL) e 2.6  103, range) viruses/L. However, EV was detected in 06/10 post-disinfection (compared to 04/10 in pre-disinfection) samples with an average concentration of 1.6  102 (BDL e 9.4  102, range) viruses/L.

CPE was detected in all five WWTPs at each of the three sampling locations using BGM cell line, indicating the presence of infectious viruses in all types of samples. Fig. 3 shows the overall average log concentration of infectious viruses for all 5 WWTPs monitored in this study at each sampling location. An average log infectious virus concentration of 4.1 (2.9e4.7, range), 1.1 (0.1e2.3, range) and 0.5 (0.1e1.6) MPN/ 100 L was detected in the influent, pre- and post-disinfection samples, respectively.

3.4.2. 3.2.3.

Quantification of human enteric viruses in biosolids

NoV GGI and NoV GGII

NoV GGI was not detected in the 5 WWTPs sampled (00/30). However, 03/30 samples were positive for NoV GGII. Three of

Fig. 2 e Average (n [ 2) EV real-time PCR virus concentration detected at the five different WWTPs at each sampling point. The detection limit was used for the preand post-disinfection for WWTP (3) and pre-disinfection for WWTP (5) samples. MBR e Membrane Bioreactor, AS e Activated Sludge, OD e Oxidation Ditch, RBC e Rotating Biological Contactors, UV e ultraviolet, Cl e chlorination.

Biosolids samples

Fig. 4 shows the average log concentration of infectious viruses found at WWTPs 1, 2 and 5 using both BGM and A549 cell lines. As the results show, an average log infectious concentration of 1.1 (0.5e2.9, range) and 3.2 (2.0e4.1, range) MPN/g was detected using both BGM and A549 cell lines.

Fig. 3 e Virus infectivity distribution using BGM cell line throughout the three different sampling points from all 5 WWTPs (n [ 5). The values are expressed as MPN/100 L.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 0 e3 5 9 8

Fig. 4 e BGM and A549 cell culture data showing the concentration of viruses being released from WWTPs 1, 2 and 5. All results are expressed as MPN/100 L (n [ 3 for influent, pre- and post-disinfection and biosolids).

3.5.

Removal of infectious viruses

From the cell culture results, it was determined that an overall removal of infectious viruses from all WWTPs between the influent and pre-disinfection (after biological treatment) was approximately 4.4, 1.2, 3.2, 4.5 and 1.6 log units for WWTPs 1, 2, 3, 4 and 5, respectively. It was observed that WWTPs 1, 3 and 4 which use MBR, activated sludge and an oxidative ditch, respectively in addition to UV disinfection were able to achieve comparable removal values of 4.4, 3.2 and 4.5 log units, respectively. WWTPs 2 (activated sludge) and 5 (RBCs) which use chlorination achieved lower removal of infectious viruses at a level of 1.2 and 1.6 log units, respectively.

3.6. Inactivation of infectious viruses using UV and chlorination According to Fig. 3, the average log infectious virus concentration for WWTPs 1, 3 and 4 for the pre- and post-UV disinfection samples was approximately 0.5 and 0.2 log units. In WWTPs 2 and 5, an average pre- and post-chlorination disinfection infectious virus concentration is 2.0 and 1.1 log units, respectively. It was observed that the overall average inactivation of infectious viruses for UV and chlorination was about 0.3 and 0.9 log units. In addition to using BGM cell line, this study also analyzed the final effluent (post-disinfection) for the five WWTPs using A549 cell line and for the biosolids samples. HAdV was only positive for WWTPs 1 (UV disinfection) and 5 (chlorination disinfection) and negative for EV for all WWTPs using ICC-PCR for A549 cell line.

3.7.

MBR and conventional WWTP

In total, 8 samples from WWTP 1(MBR treatment) (Kuo et al., 2010) and 8 samples from the current study (conventional treatment) were used to compare the log removal values for both HAdV and EV using real-time PCR data. Due to insufficient data, NoV was not included in the comparison. An

3595

average influent concentration of HAdV at 6.4 (5.6e7.9 range) and 5.6 (4.3e6.7 range) log units was observed in MBR and conventional treatment, respectively. However, the effluent concentration for each process (MBR and conventional treatment) was 2.2 and 3.5, resulting in an overall removal of approximately 4.0 and 2.2 log units, respectively (Fig. 6). This indicates that MBR treatment was able to achieve an extra 2 log reduction as compared to conventional treatment. There was a significant difference ( p-value <0.05) between the two types of secondary treatments log removal values. EV was detected at an average 5.4 (range 4.1e6.1) and 4.8 (range 4.5e5.5) log units in the influent and 1.7 and 1.5 log units in the post-secondary treatment in the MBR and conventional treatment process, respectively. According to Fig. 6, the MBR process achieved an average 3.6 log reduction (2.9e4.3, range). However, conventional treatment achieved an average 2.9 log reduction (2.0e3.7, range) log units. The ttest results for EV, showed a p-value of 0.08 indicating no significant difference.

4.

Discussion

4.1.

Release of viruses by WWTPs

To this date there are no requirements on the level of human enteric viruses that are allowed to be released after wastewater treatment. The results of this study provide conclusive evidence of the levels of infectious viruses, in addition to HAdV, EV and NoV total genomic copies that are being released in the environment. During the current study, the presence of HAdV was detected in 100% of the wastewater samples analyzed with real-time PCR. These results are consistent with past studies (Bofill-Mas et al., 2006: Carducci et al., 2008; Katayama et al., 2008; Pusch et al., 2005; Rodriguez-Diaz et al., 2009) who have reported a presence of HAdV between 55 and 100% of samples with an average of 88%. In addition to HAdV, EV was detected between 65 and 89% with an average of 76% in past studies (Pusch et al., 2005; Rodriguez et al., 2008; Katayama et al., 2008) compared to 67% observed in this study. The presence of NoV was detected in an average of 72% in past studies (Laverick et al., 2004; Nordgren et al., 2009; da Silva et al., 2007) but was only detected in 10% of our wastewater samples (NV1 was not detected). We detected NoV GGII in 3/10 samples and 0/10 effluent samples. Based on our cell culture data we were able to determine that an average concentration of 2.0  104 MPN/100 L enter (raw sewage) the three WWTPs (WWTPs 1, 2 and 5), 1.6  101 MPN/100 L are discharged as final effluent and 1.4  102 MPN/g are retained in the biosolids. It is assumed that there is some virus removal in primary sedimentation (0.1e1.0 log units) as previously reported (Nordgren et al., 2009); however samples were not collected at this particular location. We observed an overall removal of infectious viruses between 1.9 and 5.0 (average 4.2) log units from influent to final effluent. These results are comparable to previous fullscale studies (Aulicino et al., 1995; Sedmak et al., 2005; Petrinca et al., 2009) using cell culture assay reporting removals between 0 and 4.0 log units.

3596

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 0 e3 5 9 8

CPE was detected in 100% of our influent and effluent samples as compared to past studies (Aulicino et al., 1995; Rodriguez et al., 2008; Sedmak et al., 2005) averaging 88% and 45% CPE in the influent and effluent samples, respectively. Furthermore, the reported concentration of infectious viruses in the above studies fluctuated between 1e4 and 0e3 log units in the influent and final effluent, respectively. This shows inconsistencies with determining the concentration of infectious viruses before and after treatment. In the current study, the influent (2.8e4.8 log units) and effluent (0.1e1.6 log units) concentrations only fluctuated by 2 and 1.5 log units, respectively. It is plausible that this difference is due to the low sample volumes used in the above studies which ranged from 0.1e5 to 1e20 L grab samples for the influent and final effluent, respectively. In the current study, a significantly increased volume (20 L, 375 L and 410 L of influent, predisinfection and post-disinfection, respectively) was sampled. High sample volumes increase the chance of virus recovery from source waters with a low concentration of viruses (Sobsey and Glass, 1980). We detected an average 1.1e3.2 log units of infectious viruses in our 6 biosolids samples as compared to previously published (Guzman et al., 2007; Monpoeho et al., 2004) which reported values of 0.4e1.6 log units. The differences in concentrations found in this study could be due to different detention times for the biosolids between WWTPs. Interestingly, NoV GGI and NoV GGII were both detected in 06/06 biosolids samples at an average log concentration of 4.3 and 5.2 viruses/g, respectively but NoV GGI was not detected in any of the wastewater samples and NoV GGII was only detected in 3/30 samples. This could also be due the longer retention times of biosolids in the WWTPs, as compared to the liquid stream.

4.2. Inactivation of infectious viruses between UV and chlorination In this study, WWTPs using UV disinfection achieved an average removal and inactivation of 4.4 log units of infectious viruses as compared to 2.4 log units for the WWTPs using chlorination between the influent and final effluent samples. However, the average log reduction of infectious viruses between the pre- and post-disinfection processes was only 0.3 and 0.9 log units for UV and chlorination, respectively. Our results indicate that chlorination was only able to achieve 0.6 log unit higher inactivation of viruses as compared to UV. This suggests that the given configuration of unit processes in the WWTPs sampled from, are able to achieve a higher inactivation and removal of viruses as opposed to just the disinfection process. As shown in Fig. 5, WWTPs using UV were able to achieve a more consistent final effluent non-infectious virus concentration but no significant difference ( p-value >0.05) was observed between UV and chlorination. Similar results were previously reported (Rodriguez et al., 2008) where it was determined that the inactivation of infectious viruses between pre- and post-disinfection samples in a full-scale WWTP of approximately 0.3e1.3 log units. It was reported that CPE in 03/37 (8%) WWTP samples on BGM cell lines ranging from 1.48 to 1.63 MPN/L were detected. During their study, only 1 L grab samples were analyzed for infectious virus concentration for all sampling points.

Fig. 5 e Infectious virus reduction between WWTPs using UV and Cl disinfection at the Pre- and Post-Disinfection samples. UV (n [ 4), Cl (n [ 2). UV e ultraviolet, Cl e chlorination.

4.3. Comparing the removal of HAdV and EV in MBR and conventional WWTPs HAdV removal results observed in our conventional treatment samples agree with previous studies where an average 1.0e3.0 log reduction was reported (Carducci et al., 2008; Haramoto et al., 2007). However, in the current study average removal efficiency in conventional wastewater treatment of approximately 2.0 log units lower for HAdV was calculated as compared to MBR treatment with real-time PCR data. Interestingly, as shown in Fig. 6, the average removal of EV through MBR treatment was similar to what we observed (average of 0.5 log removal increase with the MBR) during conventional treatment. Our findings are consistent with a previous study (Katayama et al., 2008) in conventional treatment process that reported an average EV log influent and effluent concentration of 4.2 and 1.2 log units, indicating an average removal of 2.6 log units. Viruses have a tendency to attach to solids, and MBR provides better solid separation that the conventional activated sludge WWTPs, which rely on settling.

Fig. 6 e Comparison of HAdV and EV log removal values using real-time PCR data for MBR and conventional wastewater treatment. (MBR removal n [ 8, Con removal n [ 8).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 0 e3 5 9 8

5.

Conclusions

 Based on our cell culture data we were able to determine that an average concentration of 2.0  104 MPN/100 L enter (raw sewage) the three WWTPs (WWTPs 1, 2 and 5), 1.6  101 MPN/100 L are discharged as final effluent and 1.4  102 MPN/g are retained in the biosolids.  It was observed that there is a significant log reduction (1.9e5.0) in infectious viruses throughout the wastewater treatment process before being discharged into natural waterways.  It was observed that the reduction in infectious viruses treated with UV or chlorination can range from 0.1 to 1.2 log units as indicated by cell culture data between pre- and post-disinfection.  Based on real-time PCR data, we concluded that an MBR system is able to achieve approximately 2 log higher reduction of HAdV (average 4.1 log units) as compared with conventional wastewater treatment (average 2.2 log units). However, similar EV log removal values (3.6 for MBR and 2.9 for conventional) were observed between the two types of treatment processes.

Acknowledgments We would like to thank the wastewater utilities personnel for their assistance during this study.

references

Aulicino, F.A., Mastrantonio, A., Orsini, P., Bellucci, C., Muscillo, M., Larosa, G., 1995. Enteric viruses in a wastewater treatment plant in Rome. Water, Air and Soil Pollution 91, 327e334. Bofill-Mas, S., Albinana-Gimenez, N., Clemente-Casares, P., Hundesa, A., Rodriguez-Manzano, J., Allard, A., Calvo, M., Girones, R., 2006. Quantification and stability of human adenoviruses and polyomavirus JCPyV in wastewater matrices. Applied and Environmental Microbiology 72, 7894e7896. Carducci, A., Morici, P., Pizzi, F., Battistini, R., Rovini, E., Verani, M., 2008. Study of the viral removal efficiency in an urban wastewater treatment plant. Water Science and Technology 58, 893e897. da Silva, K., LeSaux, J.C., Parnaudeau, S., Pommepuy, M., Elimelech, M., Le Guyader, F.S., 2007. Evaluation of removal of noroviruses during wastewater treatment, using real-time reverse transcription-PCR: different behaviors of genogroups I and II. Applied and Environmental Microbiology 73, 7891e7897. Dierssen, U., Rehren, F., Henke-Gendo, C., Harste, G., Heim, A., 2007. Rapid routine detection of enterovirus RNA in cerebrospinal fluid by a one-step real-time RT-PCR assay. Journal of Clinical Virology 42, 58e64. Gerba, C.P., Pepper, I.L., Whitehead III, L.F., 2002. A risk assessment of emerging pathogens of concern in the land application of biosolids. Water Science and Technology 46, 225e230.

3597

Godfree, A., Farrell, J., 2005. Processes for managing pathogens. Journal of Applied Environmental Quality 34, 105e113. Guzman, C., Jofre, J., Montemayor, M., Lucena, F., 2007. Occurrence and levels of indicators and selected pathogens in different sludges and biosolids. Journal of Applied Microbiology 103, 2420e2429. Haramoto, E., Katayama, H., Oguma, K., Ohgaki, S., 2007. Quantitative analysis of human enteric adenoviruses in aquatic environments. Journal of Applied Microbiology 103, 2153e2159. Jiang, S., Dezfulian, H., Chu, W., 2005. Real-time quantitative PCR for enteric adenovirus serotype 40 in environmental waters. Canadian Journal of Microbiology 51, 393e398. Jimenez-Clavero, M.A., Escribano-Romero, E., Mansilla, C., Go´mez, N., Co´rdoba, L., Roblas, N., Ponz, F., Ley, V., Sa´iz, J.C., 2005. Survey of bovine enterovirus in biological and environmental samples by a highly sensitive real-time reverse transcription-PCR. Applied and Environmental Microbiology 71, 3543e3536. Jothikumar, N., Cromeans, T.L., Sobsey, M.D., Robertson, B.H., 2005. Development and evaluation of a broadly reactive TaqMan assay for rapid detection of hepatitis A virus. Applied and Environmental Microbiology 71, 3359e3363. Kageyama, T., Kojima, S., Shinohara, M., Uchida, K., Fukushi, F., Hoshino, F.B., Takeda, N., Katayama, K., 2003. Broadly reactive and highly sensitive assay for Norwalk-like viruses based on real-time quantitative reverse transcription-PCR. Journal of Clinical Microbiology 41, 1548e1557. Katayama, H., Haramoto, E., Oguma, K., Yamashita, A., Tajima, H., Nakajima, H., Ohgaki, S., 2008. One-year monthly quantitative survey of noroviruses, enteroviruses, and adenoviruses in wastewater collected from six plants in Japan. Water Research 42, 1441e1448. Kittigul, L., Uthaisin, A., Ekchaloemkiet, S., Utrarachkij, F., Luksamijarulkul, P., 2006. Detection and characterization of hepatitis A virus in water samples in Thailand. Journal Applied Microbiology 100, 1318e1323. Kuo, D., Simmons, F.J., Blair, S., Hart, E., Rose, J.B., Xagoraraki, I., 2010. Assessment of human adenovirus removal in a full-scale membrane bioreactor treating municipal wastewater. Water Research 44, 1520e1530. Laverick, M.A., Wyn-Jones, A.P., Carter, M.J., 2004. Quantitative rtpcr for the enumeration of noroviruses (Norwalk-like viruses) in water and sewage. Letters in Applied Microbiology 39, 127e136. Monpoeho, S., Maul, A., Bonnin, C., Patria, L., Ranarijaona, S., Billaudel, S., Ferre, V., 2004. Clearance of humanpathogenic viruses from sludge: study of four stabilization processes by real-time PCR reverse transcription-PCR and cell culture. Applied and Environmental Microbiology 70, 5434e5440. Nordgren, J., Matussek, A., Mattson, A., Svensson, L., Lindgren, P. E., 2009. Prevalence of norovirus and factors influencing virus concentrations during one year in a full-scale wastewater treatment plant. Water Research 43, 1117e1125. Payment, P., Fortin, S., Trudel, M., 1986. Elimination of human enteric viruses during conventional wastewater treatment by activated sludge. Canadian Journal of Microbiology 32, 922e925. Petrinca, A.R., Donia, D., Pierangeli, A., Gabrieli, R., Degener, A.M., Bonanni, E., Diaco, L., Cecchini, G., Anastasi, P., Divizia, M., 2009. Presence and environmental circulation of enteric viruses in three different wastewater treatment plants. Journal of Applied Microbiology 106, 1608e1617. Pusch, D., Oh, D.Y., Wolf, S., Dumke, R., Schroter-Bobsin, U., Hohne, M., Rsoke, I., Schreier, E., 2005. Detection of enteric viruses and bacterial indicators in German environmental waters. Archives of Virology 150, 929e947.

3598

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 0 e3 5 9 8

Rajal, V.B., McSwain, B.S., Thompson, D.E., Leutenegger, C.M., Kildare, B.J., Wuertz, S., 2007. Validation of hollow fiber ultrafiltration and real-time PCR using bacteriophage PP7 as surrogate for the quantification of viruses from water samples. Water Research 41, 1411e1422. Rodriguez-Diaz, J., Querales, L., Caraballo, L., Vizzi, E., Liprandi, F., Takiff, H., Betancourt, W.Q., 2009. Detection and characterization of waterborne gastroenteritis viruses in urban sewage and sewage-polluted river waters in Caracas, Venezuela. Applied and Environmental Microbiology 75, 387e394. Rodriguez, R., Patricia, A., Gundy, M., Gerba, C.P., 2008. Comparison of BGM and PLC/PRC5 cell lines for total culturable viral assay of treated sewage. Applied and Environmental Microbiology 74, 2583e2587. Rose, J., Dickson, L., Farrah, S., Carnahan, R., 1996. Removal of pathogenic and indicator microorganisms by a fullscale water reclamation facility. Water Research 30, 2785e2797. Sedmak, G., Bina, D., MacDonald, J., Couillard, L., 2005. Nineyear study of the occurrence of culturable viruses in source water for two drinking water treatment plants and the influent and effluent of a wastewater treatment plant in Milwaukee, Wisconsin (August 1994 through July 2003). Applied and Environmental Microbiology 71, 1042e1050. Simmons, F.J., Kuo, D.H.-W., Xagoraraki, I., 2011. Removal of human enteric viruses by a full-scale membrane bioreactor

during municipal wastewater processing. Water Research 45 (9), 2739e2750. Simonet, J., Gantzer, C., 2006a. Degradation of the poliovirus 1 genome by chlorine dioxide. Journal of Applied Microbiology 100, 862e870. Simonet, J., Gantzer, C., 2006b. Inactivation of poliovirus 1 and Fspecific RNA phages and degradation of their genomes by UV irradiation at 254 nanometers. Applied and Environmental Microbiology 72, 7671e7677. Sobsey, M.D., Glass, J.S., 1980. Poliovirus concentration from tap water with electropositive absorbent filters. Applied and Environmental Microbiology 40, 201e210. USEPA, 2001. Manual of Methods for Virology (Chapter 14). EPA 600/4e84/013. Office of Water, U.S. Environmental Protection Agency, Washington, DC. Viau, E., Peccia, J., 2009. Survey of wastewater indicators and human pathogen genomes in biosolids produced by class A and class B stabilization treatments. Applied and Environmental Microbiology 75, 164e174. Wong, K., Onan, B., Xagoraraki, I., 2010. Quantification of enteric viruses, pathogen indicators, and Salmonella bacteria in class B anaerobically digested biosolids by culture and molecular methods. Applied and Environmental Microbiology 76, 6441e6448. Xagoraraki, I., Kuo, D.H.-W., Wong, K., Wong, M., Rose, J.B., 2007. Occurrence of human adenoviruses at two recreational beaches of the Great Lakes. Applied and Environmental Microbiology 73, 7874e7881.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 9 e3 6 1 3

Available at www.sciencedirect.com

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

Multi-biochemical responses of benthic macroinvertebrate species as a complementary tool to diagnose the cause of community impairment in polluted rivers Joana Dama´sio a,b, Maria Ferna´ndez-Sanjuan a, Juan Sa´nchez-Avila a, Silvia Lacorte a, Narcı´s Prat c, Maria Rieradevall c, Amadeu M.V.M. Soares b, Carlos Barata a,* a

Department of Environmental Chemistry (IDAEA-CSIC), Jordi Girona, 18-26, 08034 Barcelona, Spain CESAM & Departamento de Biologia, Universidade de Aveiro, 3810-193 Aveiro, Portugal c Department of Ecology (UB), Av. Diagonal, 645, 08028 Barcelona, Spain b

article info

abstract

Article history:

Biological indexes, based on benthic macroinvertebrate taxa, are currently used worldwide

Received 5 November 2010

to measure river ecological quality. These indexes assign a global ecological status of the

Received in revised form

biotic community, but not necessarily may detect specific effects of water pollutants.

4 April 2011

Conversely a large set of biochemical markers measured in macroinvertebrate benthic

Accepted 5 April 2011

species can detect sublethal effects and inform us about additional environmental factors

Available online 22 April 2011

that are impairing benthic communities. This is especially interesting in moderately polluted sites, where other stressors are already affecting communities but not too strongly

Keywords:

to be detected by biotic indexes. Up to ten different markers belonging to distinct metabolic

Water quality

paths and 42 contaminants measured in sample collections of the caddis fly Hydropsyche

River

exocellata were assessed across a polluted gradient in the industrialized Mediterranean

Hydropsyche

River basins of Beso´s and Llobregat (NE, Spain). Twenty four sample collections were

Biomarker

selected to include macroinvertebrate communities representing the five impairment

Pollution

degrees defined by the Spanish Environmental authorities using the biotic metrics. Results evidenced a clear deterioration of the ecological water quality parameters and benthic communities towards downstream reaches. Biochemical responses varied significantly across the studied samples and were able to differentiate samples within communities having a good and deteriorated ecological stage. Principal Component Analyses indicated that salinity was one of the major stresses affecting macroinvertebrate assemblages, whereas antioxidant and metabolizing enzymes responded differently and were closely related to high and presumably toxic levels of accumulated organic pollutants. Therefore these results indicate that the use of multiple -markers sensitive to water pollution may provide complementary information to diagnose environmental factors that are impairing macroinvertebrate communities. ª 2011 Elsevier Ltd. All rights reserved.

* Corresponding author. Tel.: þ34 93400 6100x1505; fax: þ34 93204 5904. E-mail address: [email protected] (C. Barata). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.006

3600

1.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 9 e3 6 1 3

Introduction

The past thirty years has seen enormous growth in the collection of environmental data, of all types, by European researchers. One of the most active areas of data gathering has been the monitoring of surface waters, and consequently catchment-scale management has a good science basis in Europe. As a result of this monitoring effort, a large quantity of matched hydrochemical and biological information concerning the ecological status of these habitats exists today. In Europe, the appropriate use of such data is at the core of the European Union’s Water Framework Directive (WFD, Directive 2000/60/EC), an ambitious piece of legislation that seeks to achieve good ecological status for European surface waters by 2015. Similar trends have also occurred in Australia, North America, South Africa, New Zealand and elsewhere (for a review see Dole´dec and Statzner, 2010). The approach adopted by these countries is revolutionary in changing the focus from chemical based regulation to ecological effectsbased regulation, and from site-specific controls to the consideration of impacts at a catchment or river basin scale (Hering et al., 2006). Such holistic approaches to integrated surface water management are essential to ensure the sustainable use of ecosystem goods and services (Barbour and Paul, 2010) although care has to be taken in the interpretation of the indicators used (Basset, 2010). One of the particular and unique aspects of this integrated management is the use of ecologically based instruments to assess and predict the ecological impacts of environmental pressures on water quality. In general, ecological status assessment involves sampling the aquatic community, and comparing against a reference prediction for that water body type. This approach is currently used in many countries (see reviews of Bonada et al., 2006; Dole´dec and Statzner, 2010). Various biological metrics exist to quantify change in community composition and these are often combined in multi-metric indices to improve the chances of detecting adverse changes (e.g. Munne´ and Prat, 2009). Among them, those focussing on assemblages of benthic macroinvertebrates are the most widely used (Rosenberg and Resh, 1993; Bonada et al., 2006). However, structural metrics, although can detect the degradation of surface waters, are no reliable indicators of impairment caused by contaminants (Baird and Burton, 2001). Therefore, there is a need to complement the biological metrics actually used with other biological measures that may serve as descriptors of causeeeffect or may inform about further degradation (or improvement) of benthic communities. Recently, the development of biological trait based community indexes have allowed to diagnose effects of pesticides, salinity and certain pollutants (Liess and Von der Ohe, 2005; Kefford et al., 2011), but like the above mentioned structural metrics, these indices cannot respond to pressures other than those they were developed to detect, making diagnosis of the actual relevant pressures difficult. Furthermore, community based indexes can only detect relatively strong effects that usually involve the eradication of one or several species from a particular site, thus they cannot diagnose low levels of ecological impairment cause by sublethal physiological effects. In relation to this there are several studies showing that the use

of physiological responses of macroinvertebrate species may provide additional information to diagnose the cause of any community impairment (Barata et al., 2005, 2007; Dama´sio et al., 2008; Faria et al., 2010). Among the available methods, the integrated use of chemical analyses and biochemical and cellular responses to pollutants is a sound procedure for detecting impact of anthropogenic contaminants in freshwater systems and examine the causeeeffect relationships. Moreover, since in real field situations aquatic organisms are currently being exposed to multiple chemical contaminants involving different toxicity mechanisms, each contributing to a final overall adverse effect, the use of a large set of biochemical responses may allow us to identifying which are the potential hazardous contaminants in the field (van der Oost et al., 2003; Bocchetti et al., 2008) and produce relevant information for Water Authorities to taken actions to prevent further deterioration of ecological status. Recently, several biochemical markers were developed and used in field collected caddis flies of the pollution tolerant species Hydropsyche exocellata and in transplanted Daphnia magna to monitor metallic and organic pollution and ecological quality of water in Mediterranean Rivers (e.g. Barata et al., 2005; Dama´sio et al., 2008; Pue´rtolas et al., 2010). Contrary to macroinvertebrate assemblages, measured biochemical responses of H. exocellata and D. magna did not respond to changes in habitat quality, but were only sensitive to small changes of chemical pollutants in water (Barata et al., 2005; Dama´sio et al., 2008; Pue´rtolas et al., 2010). The main objective of this study was to address if the use of multi-biomarker responses in a benthic macroinvertebrate species may complement existing biotic and multi-metric indices that are used to assess changes in ecological status of river biota. To do so firstly, we assessed and compared biochemical responses of field collected caddis flies of the pollution tolerant species H. exocellata (Bonada et al., 2004) with biological indices calculated from taxa assemblages of benthic macroinvertebrates in seven different sites in one basin (river Beso´s, NE Spain) with a total of ten different samples (in some places summer and spring samples were available when the river didn’t dry up in summer). Samples came from a monitoring programme set up in the Beso´s River since 1994 (Prat and Rieradevall, 2006) and used for the Water Authorities of Catalonia (ACA) in the Water Management Plan and in studies of temporal variation of the different metric values (Munne´ and Prat, 2011), and data from pristine and different levels of polluted sites are available. The tolerant species H. exocellata was selected since it occurs along pristine and degradated benthic communities being the dominant species in the latter ones (Bonada et al., 2004). Biomarkers included the phase II gluthatione-S-transferase activity (GST); glutathione reductase (GR), which aids maintenance of GSH levels recycling oxidized glutathione; antioxidant enzymes involved in detoxifying reactive oxygen species, such as, catalase (CAT), superoxide dismutase (SOD) and glutathione peroxidase (GPX); markers of oxidative tissue damage (lipid peroxidation and DNA strand breaks) and lactate dehydrogenase (LDH) as an indicator of the metabolic state of the animal (Pue´rtolas et al., 2010). The activity of B-esterases was also measured to diagnose exposure to organophosphorous pesticides among other chemicals (Barata et al., 2004). The study also

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 9 e3 6 1 3

attempted to characterize metals and organic contaminants present in H. exocellata larvae, to allow identifying major pollutants affecting the observed responses. This study was conducted in the Beso´s River basin (NE Spain), which is a good example of an intensively used Mediterranean stream system, receiving extensive urban and industrial wastewater discharges from the area of Barcelona (Prat and Munne´, 2000). Finally, the results obtained within the Beso´s River were further tested re-analyzing data obtained in a previous study (Barata et al., 2005), which was conducted in 14 different sample collections (7 different sites and two seasons) in the Llobregat River. The natural resources of both river basins (Beso´s and Llobregat) have been greatly affected by human activities such as agriculture, urbanization, a salinity increase as a result of mining activities (Llobregat) or wastewater treated effluent discharges (Beso´s); and an intensive water use for human consumption (supplying water to many urban areas including Barcelona city) (Prat and Munne´, 2000; Barata et al., 2005; Prat and Rieradevall, 2006).

25e26/7/2003). Selected sites belong to two different river types according to the classification used for the purposes of the WFD by the Water Catalan Agency (from now on abbreviated as ACA) (Munne´ and Prat, 2011). Eleven of the study sites, B3eB7, L1, L2, L4eL7 were located in-streams belonging to the lowland Mediterranean River type with variable and low discharges and two of them, B1, B2 and L3, were situated in small mountain streams. The ecology of the Beso´s and Llobregat Rivers have been extensively studied (e.g.; Prat and Munne´, 2000; Prat and Rieradevall, 2006; Dama´sio et al., 2008) and since 1994 a surveillance monitoring program is being carried out in this river supported by a regional government (Diputacio´ de Barcelona: http://ecobill.diba.cat/) and the ACA (http://mediambient.gencat.net/aca/). According to the previous studies, the selected sites had communities belonging to the five ecological quality types defined in the WFD: very good, good, fair, poor and very poor, and adapted to each river type (see Munne´ and Prat, 2009; 2011 for details).

2.2.

2.

Material and methods

2.1.

Study sites

The study was performed in 14 locations within the Beso´s and Llobregat River basin, NE Spain (Fig. 1) and during two periods (Beso´s; 13e14/4/2005 and 27e28/07/05, spring (sp) and summer (su), respectively; Llobregat; sp, 16e17/4/2003 and su,

3601

Physicochemical water measurements

A set of environmental variables was measured on each sampling occasion. Water physicochemical parameters including flow rate (l/s), temperature (T,  C), pH, conductivity (mS/cm) and dissolved oxygen (O2, mg/l) were measured in situ with a Mini Air probe (Technika-Schiltknecht). Sulphates (SO4, mg/l), chlorides (Cl, mg/l), N-ammonium (NH4, mg/l), N-nitrites (NO2, mg/l), N-nitrates (NO3, mg/l) and P-phosphates (PO4, mg/l) were

Fig. 1 e Map of the sampling sites, which includes the main channel of Beso´s and Llobregat, Beso´s tributaries Congost and Ripoll Rivers, the Llobregat tributaire Cardener River, and the main urban nucleus (La Garriga, Sabadell and Manresa close to sites B6 B7 and L5, respectively). Waste Water Treatment Plants (WWTP) are depicted as squares.

3602

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 9 e3 6 1 3

obtained following well-established methods for water physicochemical parameters ASTM (1998) procedures.

2.3.

Biological conditions and specimens collection

Benthic macroinvertebrates, riparian vegetation and habitat quality at the studied sites were assessed in order to establish the ecological status of the sites using the Guadalmed protocol, which combines three indexes for riparian habitat, fluvial habitat and macroinvertebrate community (Ja´imez-Cue´llar et al., 2002). The quality of the riparian vegetation was measured using the Riparian habitat Ecological Quality Index (QBR) and the in-stream habitat quality was measured by the Fluvial Habitat Index (IHF). Both indexes have been defined elsewhere (Munne´ et al., 2003; Munne´ and Prat, 2009) and score 0 (highly disturbed) and 100 (natural). Regarding the quality of the macroinvertebrate communities, several metrics were assessed including the Iberian Bio-Monitoring Working Party Biological Index (IBMWP); the number of taxons (S) and the Iberian Average Score Per Taxon (IASPT is the result of IBMWP divided by S). The IBMWP is an unimetric index that has been used extensively for biological monitoring and its quite well correlated with the ICM-Star multi-metric index, the index used for the intercalibration purposes in the Med-GIG (Mediterranean group of intercalibration) as demonstrated in Munne´ and Prat (2009). IBMWP establishes five community impairment levels that include very clean waters (very good state), waters with signals of stress (good state), contaminated waters (fair state), very (poor state) and extremely contaminated (very poor state) waters, which for the lowland Mediterranean River type have scores >121, 71e120, 41e70, 40e20, <20, respectively. The IASPT ranges from 0 (no scoring macroinvertebrates present) to 10 (all scoring macroinvertebrates are pollution sensitive). The temporal changes of these metrics in several of the stations used for this paper were recently studied by Munne´ and Prat (2011). Benthic macroinvertebrates were obtained quantitatively by sampling all available habitats with a kick net of 250 mm according to the procedure described in Ja´imez-Cue´llar et al. (2002). Just after collection, 50e200 H. exocellata larvae were separated from other macroinvertebrate species. About half of the individuals were inmediatelly frozen in liquid nitrogen and stored at 80  C until further analysis of biochemical responses. The remaining ones were kept in the field in the same river water for 1e2 h to allow detritus to be removed from their body and the partial clearance of their gut content, rinsed several times with distilled water, frozen in liquid nitrogen and stored at 80  C until further analysis of metals and organic contaminants. The rest of benthic macroinvertebrate species were preserved in formalin (5%) and identified to the family level to calculate the structural metrics. Not in all seasons it was possible to sample enough H. exocellata individuals to fulfil sample requirements, thus the number of communities studied were restricted to ten.

2.4.

Biomarker analysis

Biomarkers were determined in individual larvae of H. exocellata, a tolerant species widely distributed within Beso´s River, whose biomarker responses has been previous applied

(Barata et al., 2005). Samples were homogenized in ice-cold 100 mM phosphate buffer (PBS), pH 7.4, containing 100 mM KCl and 1 mM EDTA. H. exocellata heads were separated from the body, homogenized in the proportion 1 head: 200 ml PBS, centrifuged at 10,000 g for 30 min and supernatants used for acetylcholinesterase (AChE) determination. Insect larvae heads contain tissues rich in neuronal cells (eyes, mouth muscles) and hence have high activities of acetylcholinesterase (Liu, 1993). All the others biomarkers were determined in the body. Bodies were homogenized in 1:8 proportion wet weight: PBS volume and centrifuged at 10,000 g for 30 min. Catalase (CAT) measurements were carried out using a spectrophotometer Cecil-CE 9200 (Cambridge, England), whereas the other biomarkers were determined using a MultiDetection Micro- plate Reader, BioTek (Vermont, USA). Assays were run at least in duplicate. The activities of catalase (CAT, mmol/min/mg protein), superoxide dismutase (SOD, SOD unids/mg protrein), total glutathione peroxidase (GPX, nmol/min/mg protein), glutathione reductase (GR, nmol/min/mg protein), glutathione-Stransferase towards 1-chloro-2,4-dinitrobenzene (GST, nmol/ min/mg protein), lactate dehydrogenase (LDH, mmol/min/mg protein), cholinesterase (ChE, nmol/min/mg protein) and carboxylesterase (CbE, nmol/min/mg protein), and levels of DNA strand breakes (mg DNA/g wet weight) and of lipid peroxidation (LPO, nmol of malondialdehyde/g wet weight) were determined following established procedures as described by Barata et al. (2005) and Pue´rtolas et al. (2010). Proteins were measured by the Bradford method using bovine serum albumin as standard.

2.5.

Contaminant analyses

Metal concentration levels of As, Cu, Pb, Zn, Al, Cd, Cr and Ni were determined in four single or groups of two freeze dried larvae per site (mean  SD; 19.13  5.65 mg dry weight) following Barata et al. (2005) procedures. Dried larvae were digested with 0.75 ml concentrated nitric acid and 0.25 ml hydrogen peroxide (instra quality, Baker) using 60 ml Teflon bombs at 90  C overnight. Within each digestion series, appropriate blanks with no insects were also subject to the same procedure to account for background contamination levels. Cooled digested samples were diluted to a standard volume with deionized water. Trace metal analysis was determined using a Perkin Elmer model Elan 6000 inductively coupled plasma mass spectrometer (ICP-MS). Calibration standards and a reagent blank were analyzed with every ten samples to monitor signal drift. In every instance, the signal typically changed by 3e5% throughout an analytical run. Additionally, rhenium was used as an internal standard to correct for any non-spectral interference. Samples of similar weight of a certified reference material (lobster hepatopancreas, Tor 1, national Council of Canada, Ottawa) were digested during each analytical run; measured trace metal concentrations were within the certified range for the metal. Except for Cd, metals levels exceeded ten fold the analytical detection limit. For Cd, only those levels exceeding two fold the analytical detection limit (0.01 mg/g d.w.) was considered. Up to 35 different organic compounds belonging to six categories were analyzed in freeze-dried larvae. These

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 9 e3 6 1 3

included: alkylphenols (AP), with octylphenol, nonylphenol technical and two nonylphenol ethoxylates (mono- and di-); polychlorinated biphenyls (PCBs), with seven congeners, 28, 52, 101, 118, 138, 153, 180; polycyclic aromatic hydrocarbons (PAHs), that included 14 compounds (naphthalene, acenaphthene, fluorene, fluoranthene, anthracene, phenanthrene, pyrene, chrysene, benzo (a)pyrene, benzo (a)anthracene, benzo (b)fluoranthene, benzo (g,h,i)pyrene, indeno (1,2,3-cd)pyrene, dibenzo (a,h)anthracene); dicloro difenil trichloroethanes (DDTs) with five compounds that included parental and degradation metabolites (2,4 DDD; 2,4 DDT; 4,4 DDD; 4,4 DDE; 4,4 DDT); endosulphan (ENDO) with two isomers (alpha and beta); hexachlorocyclohexane (HCH) with the four isomers (alpha, beta, gamma, delta). Due to analytical constrains pools of 10e20 individuals were used without replication. Surrogate standards (4-n-NP-D8, NP1EO-D2, 4.4 DDD 13C12, DPP D4, PCB 209) were added to a sample aliquot of 0.1 g, to get a final concentration of 1000 ng/g. Sample were homogenized and kept at 4  C overnight and subsequently extracted by sonication (10 min) one time with hexane/dichloromethane (1:1, v/v) in a ratio of 100 ml per gram of dried mass. A twice extraction (same ratio) was performed with a mixture of hexane/acetone (1:1, v/v). After each extraction step, samples were centrifuged for 10 min and the extracts were combined and concentrated to approximately 1 ml under a nitrogen current using a TurboVap at 25  C. Extracts were subsequently cleaned up by solid phase extraction (SPE) cartridges with 5 g of Florisil, previously conditioned with 20 ml of hexane/dichloromethane (1:1,v/v) and 20 ml of hexane/acetone (1:1, v/v). The sample extract was eluted with 20 ml of hexane/dichloromethane (1:1, v/v) and 20 ml of hexene/acetone (1:1, v/v). The eluent was concentrated to a volume of less than 1 ml under a nitrogen-current at room temperature and reconstituted with ethyl acetate to a final volume of 1 ml, with addition of 1 mg/ml of the internal standard Anthracene D10. Final extracts were analyzed by gas chromatography - electronic impact - mass spectrometry in tandem (GC-EI-MS/MS). Analyzes were performed using an Agilent 7890A GC System (Agilent Technologies, Palo Alto, CA, USA) interfaced to a 7000A triple quadrupole mass spectrometer system (Agilent, USA). Separation and detection method parameters were similar to those reported by Sa´nchez-Avila et al. (2010). For increased sensitivity and specificity, peak detection and quantification was performed in Selected Reaction Monitoring (SRM) mode using two transitions per each compound. Internal standard quantification was performed using the surrogate standards method. A nine points calibration curve was constructed at 100, 250, 500, 750, 1000, 2500, 5000, 7500 and 10,000 pg/mL with good linearity over this concentration range (R2 > 0.994). The limits of detection (LOD) of the analytical method were calculated with the minimum concentration of analyte that produced a signal-to-noise ratio (S/N) of 3:1 and 6:1, respectively. Sample blanks were rigorously performed to eliminate any external source of contamination. Blank samples were below detection limits except for phthalates, where blank values were below quantification limits. Finally, accumulation levels of organic contaminants were transformed in Toxic equivalents (TEQ) using eq (1). TEQ ¼

C  100 LR50

(1)

3603

Where LR50 were obtained from reported lethal residue levels on Chironomus and amphipods as follows. LR50 (ng/g d.w.) for lindane, endosulfan and DDT were set to 30234, 21542, 11440, respectively, following Traas et al. (2004) and considering a water-content in H. exocellata larvae of 65%. LR50 for PAH were estimated from the most common congener (pyrene) from Landrum et al. (2003) and was set to 345,000 ng/g d.w. LR50 for tetrachlorobiphenyl (165,000 ng/g d.w) and nonylphenol (244,000 ng/g d.w.) reported by Fay et al. (2000) were used as a surrogate values for PCB and AP, respectively. Contaminant body burdens close to TEQ of 100 should be lethal.

2.6.

Data analysis

Data analyses were performed in two different datasets.

2.6.1.

Beso´s data.

Unbalanced samples across sites and seasons did not allowed to perform a full two way ANOVA, thus biochemical responses and metal levels across sites and seasons were compared using two analyses: A two way ANOVA limited to sites 1, 5 and 6 and one way ANOVA including all data points followed by post-hoc multiple comparison tests (Zar, 1996). Data were log transformed prior to analysis to achieve normality and variance homocedasticity. Due to sample size constrains measurements of organic contaminant levels in H. exocellata samples were not replicated and hence could not be compared statistically. Principal Component Analyses (PCA) on biochemical responses and on the whole dataset of biological and physicochemical measurements was also performed. The former was used considering all replicates within sites to classify the studied sample collections according to their biochemical response patterns. The latter was performed considering only the mean responses and aimed to identify relationships between biological and abiotic variables. Due to the large number of pollutants measured and the existence of empty (non detected) values, only metal levels that were detected in all samples were considered (i.e. As, Cu, Pb, Zn, Al). Organic contaminants were also grouped in the six categories described in the previous section (AP, PCBs, PAHs, DDTs, ENDO, HCH). In both analyses, since variables were very different (physicochemical, quality indices and biochemical responses) and/or they were not measured using the same scale units, the data was auto-scaled prior to analysis (each element was subtracted by its column mean and divided by the standard deviation of its column). The number of PCA components was finally selected according to cross validation leaving one out prediction errors criteria (Wold et al., 2001). Uni and multivariate analyses were performed using the IBM SPSS Statistics ver 19 and the Matlab 6.0 software, respectively.

2.6.2.

Llobregat data

Data from Llobregat River were obtained from a previous study conducted also in H. exocellata larvae (Barata et al., 2005). It included basically the same number and type of ecological, water physicochemical and metals, but biochemical responses were limited to only five oxidative stress markers (SOD, CAT, GST, GPX, LPO). Organic contaminants in water from 2003 were provided by ACA (www.gencat.cat/aca) from its public monitoring database, and correspond to the same

3604

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 9 e3 6 1 3

sampling period and sites. Only 14 contaminants occurred above detection levels in at least one of the studied locations. These were grouped in five categories that included: alkylphenols (AP), with octylphenol and nonylphenol ethoxilates; polycyclic aromatic hydrocarbons (PAH), with pyrene, phenanthrene and fluoranthene; organophosphorous pesticides (OP) with diazinon and chlorpyrifos; organochlorine pesticides (OCl) with hexachlorocyclohexane (alpha, beta, gamma, delta), metolachlor and alachlor; and triazines (TRZ) with atrazine, terbutrine, terbutilazine (all these data is found in appendix 1). Likewise in the previous section Principal Component Analyses (PCA) on the whole dataset of biological and physicochemical measurements were performed to classify the studied sample collections according to their biochemical response patterns and to identify relationships between biological and abiotic variables.

3.

Results

3.1.

Physicochemical water parameters

In general, water flow decreased dramatically in summer and nutrient load and conductivity increased substantially from upper to downstream reaches due to the discharge of effluents coming from wastewater treatment plants (Table 1). The observed quite large variation of temperature, water flow and oxygen levels across sites were related to the inclusion of two river ecotypes, small mountain streams and lowland Mediterranean Rivers with variable and low discharges and data from two different sampling periods. The sites are a good representation of the different levels of pollution present in the area, from pristine reaches (B1, B2) to very polluted (B7).

3.2.

Contaminant levels in organisms

From the eight metals analyzed in biological samples only five were detected in all samples (Table 2). ANOVA analyses restricted to these five metals denoted significant (P < 0.01) differences across sites but not between seasons (Tables 2,3). Organisms of site B6 followed by B5 had the highest levels of metals being from two fold (As, Cu, Zn) to almost two orders of magnitude (Pb) higher than those measured in other locations. PCBs, PAHs, DDTs and ENDO showed higher levels for middle and downstream sites, than upstream ones (Table 2). PCB levels measured at site B6 were over two orders of magnitude greater than those from site B1; those of DDTs were undetected in upstream reaches, reaching levels of 14.0e82.0 ng/g d.w. in downstream locations. PAHs and ENDO showed only a moderate increase from upstream (405.5e507.9 ng/g d.w. and 138.1e241.3 ng/g d.w., respectively) to downstream sites (526.8e689.2 ng/g d.w. and 254.3e420.6 ng/g d.w., respectively). AP and HCH varied little across sites, with no a clear trend towards downstream locations. The greatest toxic equivalents (TEQ close or above 1) were obtained for PCBs and ENDO and estimated TEQ of measured body burdens of organic pollutants increased by almost 4 fold from site B1 to site B6 (Table 2).

3.3.

Ecological status

The quality of the riparian habitat (IHF, QBR) decreased substantially from upper to downstream sites (Table 1) as the physicochemical parameters did. Closely linked with the deterioration of water chemistry and habitat conditions, macroinvertebrate communities were dominated by more diverse (S > 20) and sensitive taxa (values of IBMWP>100; IASPT >5.4) in

Table 1 e Ecological and physicochemical water quality parameters (Mean) at the studied seven sites across seasons. Studied sites

IBMWP S IASPT QBR IHF Flow T pH Cond O2 NH4 NO2 NO3 PO4 SO4 Cl

B1 sp

B1 su

B2 sp

B3 sp

B4 su

B5 sp

B5 su

B6 sp

B6 su

B7 su

205 33 6.2 100 92 12 16.1 8.31 219 10.58 0.33 0.01 0.01 0.01 17.2 9

180 33 5.5 100 88 3 17.5 7.68 215 8.94 0.41 0.03 0.01 0.03 17.3 12

135 23 5.8 72 79 2 8.8 8.11 674 10.25 0.14 0.01 1.23 0.01 102.3 16

101 21 4.8 90 73 3 10.1 7.83 835 10.10 3.30 0.40 1.31 2.70 76.8 141

45 12 3.7 40 60 26 26.5 8.11 2200 9.01 1.15 0.29 0.86 1.88 157 524

35 10 3.5 25 67 56 9.4 8.41 1533 9.24 0.25 0.03 2.21 0.62 119 269

39 10 3.9 20 75 5 23.6 7.72 1510 6.35 0.49 0.03 1.24 0.67 118 294

39 11 3.5 10 68 69 21.2 9.12 887 15.55 0.49 0.02 1.47 0.49 165 343

66 17 3.8 10 53 2 29.7 8.81 2000 12.50 0.41 0.01 0.01 0.51 195 430

19 7 3.4 15 48 66 26.7 8.09 2340 11.90 0.58 0.08 1.54 1.71 282 458

Measured parameters included: Iberian Bio-Monitoring Working Party Biological Index (IBMWP), number of taxons (S), Iberian Average Score Per Taxon (IASPT), Riparian habitat Ecological Quality Index (QBR), Fluvial Habitat Index (IHF), water flow (Flow, l/s), temperature (T,  C), conductivity (Cond, mS/cm), oxygen (O2, mg/l), nutrients (NH4, NO2, NO3, PO4; mg/l), SO4 (mg/l), Cl (mg/l). sp: spring; su: summer.

3605

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 9 e3 6 1 3

Table 2 e Levels of metals (Mean ± SE; mg/g d.w.) and organic pollutants (Mean; ng/g d.w.) measured in H. exocellata larvae collected at the studied sites across seasons. Different letters indicate significant (P < 0.05) differences following ANOVA and Tukey’s post-hoc tests. Within the site column sp and su correspond to spring and summer, respectively. LOD: limit of detection. Summed concentration levels of alkylphenols (AP), polychlorinated biphenyls (PCBs), polycyclic aromatic hydrocarbons (PAHs), dicloro difenil trichloroethanes (DDTs), endosulfans (ENDO) and hexachlorocyclohexane (HCH). Individual and summed (Sum) Toxic equivalents (TEQ) of organic body residue levels. Metals (mg/g d.w.) As Sites Mean 1 sp 1 su 2 sp 3 sp 4 su 5 sp 5 su 6 sp 6 su 7 su LOD

4.4 3.1 4.4 4.3 3.5 5.1 5.6 7.7 8.2 3.6 2.7

Cu SE 0.6 0.8 0.5 0.8 0.5 0.3 0.2 0.7 1.0 0.3

Mean a a a a a a b b b a

17.1 17.3 19.4 14.8 17.4 19.1 19.8 35.8 29.8 20.4 0.6

Pb SE 0.3 0.7 0.2 0.4 0.7 1.7 1.2 1.9 5.1 0.3

a a a a a a a b b a

Zn

Mean

SE

1.8 1.6 1.2 2.4 1.8 4.7 5.9 97.1 102.8 1.7 0.3

0.3 0.1 0.1 0.4 0.0 0.8 0.9 15.1 11.4 0.1

a a a a a a a b b a

Al

Mean

SE

130.5 135.4 144.2 116.8 136.5 130.2 127.4 222.4 232.4 155.4 4.1

1.8 6.1 3.4 5.6 13.9 3.9 1.9 18.5 40.9 6.2

a a a a a a a b b ab

Cd

Mean

SE

333.5 314.6 470.2 196.8 278.0 931.3 1025.0 875.4 991.7 317.5 158.3

28.7 48.4 43.6 17.8 35.4 154.6 79.1 132.9 71.1 24.7

Mean a a a a a b b b b b

0.4 0.2

Cr SE

0.1 0.1

0.1

Mean

8.0 5.8 4.7 1.1

Ni SE

1.0 0.8 0.2

Mean SE

5.0 6.2 2.9 1.1

0.8 1.1 0.1

Organic contaminants (ng/g d.w.) AP 1 sp 1 su 2 sp 3 sp 4 su 5 sp 5 su 6 sp 6 su 7 su LOD

TEQ

740.1 0.3 883.4 0.36 790.3 0.32 854.0 0.35 1031.8 0.42 650.5 0.27 791.2 0.32 831.1 0.34 831.1 0.34 1359.4 0.56 1.5e108.2

PCBs

TEQ

PAHs

TEQ

DDTs

TEQ

ENDO

18.5 25.0 101.3 70.4 260.7 324.9 246.1 2993.7 1500.0 399.1 0.4e1.7

0.01 0.02 0.06 0.04 0.16 0.2 0.15 1.81 0.91 0.24

405.2 420.5 507.9 455.4 633.0 526.8 679.1 619.7 689.2 618.4 0.4e4.3

0.12 0.12 0.15 0.13 0.18 0.15 0.2 0.18 0.2 0.18

1.4 1.4 2.7 1.4 66.6 14.0 28.1 56.4 82.0 19.4 0.2e0.6

0.01 0.01 0.02 0.01 0.58 0.12 0.25 0.49 0.72 0.17

138.1 0.46 140.5 0.46 150.4 0.5 241.3 0.8 470.8 1.56 259.6 0.86 340.6 1.13 254.3 0.84 420.6 1.39 355.7 1.18 9.8e14.1

upper reaches and less diverse (S < 10) and tolerant taxa to pollution (IBMWP<50; IASPT <4) in downstream sites. The exception being site B6 in summer that showed a relative quite high IBMWP score (66) but with an IASPT lower than 4 (which implies that most of the taxa are pollution tolerant). Consequently, the ten samples studied showed different degrees of deterioration and could be classified (according to the ACA classification in its Water Management Plan) as having a very good (B1, B2), good (B3), fair (B4, B6 summer), poor (B5, B6 spring) and very poor (B7) ecological quality.

3.4.

Biochemical responses of H. exocellata

One and two way ANOVA results of the biochemical responses of H. exocellata showed that nine of the ten responses measured varied significantly across the studied samples and also denoted clear responses patterns of H. exocellata to pollution (Fig. 2, Table 3). Significant (P < 0.05) seasonal or interaction effects were only present in three out of the ten biomarkers tested (Table 3), thus denoting a reasonable consistent response pattern across seasons. Activities of ChE and CbE decreased from upper to downstream reaches and the opposite trend was observed for the antioxidant, biotransformation and metabolic enzymes CAT, GST and LDH,

TEQ

HCH

TEQ

4.4 0.02 2.3 0.01 1.2 0.01 6.6 0.03 11.1 0.05 7.3 0.03 4.1 0.02 4.8 0.02 5.6 0.03 12.7 0.06 0.4e3.6

Sum 0.92 0.98 1.06 1.36 2.95 1.63 2.07 3.68 3.59 2.39

respectively, and levels of lipid peroxidation and of DNA strand breaks (Fig. 2). PCA analyses performed on the full biochemical responses defined five interpretable components that explained 87.6% of data variance. The first two components explained 56.6% of data variance and confirmed the findings of univariate ANOVAs. PC1 and PC2 defined a clear stress gradient separating upstream samples having high activities of SOD, ChE and CbE from those of middle and downstream reaches with high activities of CAT, LDH, GST and elevated levels of LPO and DNA strand breaks (Fig. 3). This means that H. exocellata larvae collected from communities with a poor ecological status had lower activities of B-esterases (ChE, CbE), higher activities of antioxidant and metabolizing enzymes, and greater levels of tissue damage. By plotting the mean site scores and their 95% CI it was possible to discriminate six groups with no overlapped 95% CI: two for reference sites with samples having a very good ecological status B1 and B2, one for the sample of site B3, that had a good ecological status, two for samples of sites having a fair and poor ecological status (B5 and B4-B6) and one for the sample of site B7, that had a very poor ecological status. Interestingly, PCA scores of individuals collected from the same site but in different seasons were quite similar indicating that pollution sources responsible for the observed biochemical changes are acting continuously.

3606

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 9 e3 6 1 3

Table 3 e Two and one way ANOVA results comparing log transformed metal levels and biochemical responses measured in H. exocellata larvae collected at the studied sites and seasons. Only the degrees of freedom (df), Fisher’s coefficient (F) and significant levels corrected according to Bonferroni for simultaneous tests are depicted. Due to unbalanced design two way ANOVA was restricted to sites 1, 5 and 6. ns P > 0.05; * 0.01 < P < 0.05 **P < 0.01. Two way ANOVA Site df

F

Metal levels As 2,18 12.1** Cu 2,18 26.9** Pb 2,18 348.8** Zn 2,18 15.8*** Al 2,18 15.9 **

Season

One way ANOVA

Interaction

df

F

df

F

df

F

1,18 1,18 1,18 1,18 1,18

0.1 ns 1.1 ns 1.3 ns 0.2 ns 0.2 ns

2,18 2,18 2,18 2,18 2,18

0.2 1.3 0.5 0.2 0.1

ns ns ns ns ns

9,30 9,30 9,30 9,30 9,30

5.6** 11.9** 46.8** 7.0** 19.0**

0.6 ns 7.8* 1.2 ns 2.2 ns 0.7 ns 7.9* 3.7 ns 2.5 ns 5.2 * 3.7 ns

2,23 2,23 2,23 2,23 2,23 2,23 2,23 2,23 2,23 2,23

3.2 ns 1.6 ns 0.4 ns 3.7 * 0.5 ns 2.3 ns 0.9 ns 0.9 ns 4.9* 2.5 ns

9,42 9,42 9,42 9,42 9,42 9,42 9,42 9,42 9,42 9,42

22.3** 4.5** 15.4** 3.0** 46.8** 21.6** 11.1** 2.0 ns 15.3** 12.2**

Biochemical responses CAT SOD GST GPX CBE CHE LDH GR DNA LPO

2,23 2,23 2,23 2,23 2,23 2,23 2,23 2,23 2,23 2,23

10.6** 12.4** 13.6** 1.2 ns 39.8** 90.4** 18.9*** 3.1 ns 27.8** 28.1**

1,23 1,23 1,23 1,23 1,23 1,23 1,23 1,23 1,23 1,23

3.5. Relationships between biological and environmental factors A PCA performed on up to 38 biological and physicochemical parameters defined three interpretable components that explained 67% of data variance. Salinity related parameters delimited the first component that explained almost 50% of data variance and differentiate the ecological quality of most of the studied samples (Fig. 4A). To remove the effects of salinity all variables except Cond and SO4 were regressed against the most relevant salinity parameter (Cl) and their residuals re-analyzed again. The first two PC explained almost 60% of residual variation and establish the following relationships (Fig. 4B): ecological quality parameters based on macroinvertebrate assemblages were close to the origin and antioxidant and metabolizing enzymes were positively related with organic contaminant levels.

3.6.

Llobregat data

PCA performed on individual biochemical responses defined only two interpretable components that explained 64.6% of data variance but only the first one established a clear stress gradient separating upstream sample collections (L1, L2, L3) having low activities or levels of CAT and LPO from those of middle and downstream reaches with high activities and levels of CAT and LPO. Site discrimination by biomarker responses only partially agreed with that provided for the ecological quality criteria (Fig. 5A). Note, however, that when

Besos biochemical data was re-analyzed using the same biomarkers, similar results were obtained (Fig. 5 A1). PCA conducted with biochemical, ecological and physicochemical data established three components that explained 77.4% of data variance. PC1 and PC2 accounted for almost 60% of variance and were strongly influences by salinity and seasonality parameter, separating upstream samples having good ecological quality from middle and lower reaches with a poor ecological state (Fig. 5B). When salinity effects were accounted for by performing a PCA on regression residuals, seasonality was still present on PC1, which was strongly related with high temperatures and levels of metals in summer. High levels of nutrients, of organic contaminants in water (PAH, TRZ) and of LPO and elevated activities of CAT defined PC2 (Fig. 5B1).

4.

Discussion

Physicochemical water parameters and pollutant concentration levels determined in H. exocellata larvae indicated a clear increase of organic and metal pollution from upper to downstream reaches in the Beso´s River system. These results are characteristic of Mediterranean regions, where intensive water resource use is frequently linked to the lack of water flow due to climatic constrains, and rivers can receive effluents from cities, industries and agriculture with null or scarce dilution in summer. In this situation, water quality is poor and measures that are effective for wet countries, such as the building of wastewater treatment plants, fail to recover river water biological quality (Prat and Munne´, 2000). It is interesting to note that 30 years ago, before the building up of the sewage plants, H. exocellata was not present in sites B4, B5, B6, B7 (Prat and Rieradevall, 2006) indicating the partial recovery of these sites and the possibility to use H. exocellata as a sentinel of community impairment examining its biochemical responses. Measured physicochemical water parameters including salinity and nutrient load along the studied sites (Table 1) were high compared to levels usually found in undisturbed rivers (Chapman and Kimstach, 1996). In particular, conductivity and phosphate levels observed in downstream reaches exceeded 1000 mS/cm and 0.4 mg/l, respectively, thus denoting high organic and saline pollution compared with reference sites in he same basin (B1, B2). It is worth noting that the relatively high (>1000 mS/cm) conductivity levels found in most downstream locations are likely to be related to wastewater effluents coming from treatment plants and industrial activities (Prat and Munne´, 2000). Physicochemical water characteristics thus indicate sub-optimal conditions in many of the studied sites for biological communities. Indeed the IBMWP and IASPT scores obtained for the benthic macroinvertebrate community inhabiting the studied sites denoted a good ecological status for upper reaches and a poor ecological state (and very tolerant taxa) for middle and downstream reaches (Table 1). In this situation, caddis fly assemblages of the river are dominated by the stress tolerant species H. exocellata (Bonada et al., 2004). Except Pb at site B6, measured trace metal concentrations in whole H. exocellata larvae collected from the Beso´s River basin were in the same range of those reported for Hydropsiche sp

3607

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 9 e3 6 1 3

b

b

8

b

bc c

c

4

0

c

d

0,4

d

b

ab ab

200

c 100

cd cd

d

c

d

c c c c

c c

c

nmol/min/mg prot

b

20

1

a a

a bb

b b b

LPO

CbE b b b

200

b

b bb

c

a a

30

a

a ab

20

b

b b

ab

b

10

0

Sites DNA b b b bb a

a

40

aa

B1 B2 B3 B4 B5

600

LDH

0,0

0

0

800

4

0,5

400

40

b

1,0

GR

2

ChE

ab

ab b

8

b

60

ab

a ab

0

0

a a

12

µmol/min/mg prot

300

0

µg DNA /g ww

bb

c

c

3

nmol/min/mg prot

nmol/min/mg prot

b

GST a

nmol/min/mg prot

0,8

a

ab

ab

0,0

400

a

16

a

nmol MDA /g ww

U/mg prot

ab

a

mmol/min/mg prot

a 12

GPX

CAT

1,2

nmol/min/mg prot

SOD

a

16

B6 B7

B1 B2 B3 B4 B5

B6 B7

Sites

c b b

a

400

200

0

B1 B2 B3 B4 B5

B6 B7

Sites Fig. 2 e Biochemical responses (Mean, SE) measured in H. exocellata individuals at the studied communities. Abbreviations are depicted in the text. Different letters indicate significant (P < 0.05) differences following ANOVA and Tukey’s post-hoc tests. GR graph has no letters since there were no significant differences across sites. White and grey bars indicate spring and summer samples, respectively.

species from reference sites (Cain and Luoma, 1998; Cain et al., 2004; Sola` and Prat, 2006). Indeed Sola` and Prat (2006) reported metal body burdens of As, Cu, Zn and Cd in Hydropsiche splarvae, collected from a mine river, one to two orders of magnitude higher than those observed in this study. The quite high levels of Pb measured in site B6 were comparable to those reported by Sola` and Prat (2006) in Hydropshyche sp larvae collected from a mine river (i.e. 100 mg/g d.w.). Therefore, the previous reported

information indicates that the metal levels measured in Beso´s River, except Pb, were far below those challenging their survival. On the other hand organic contaminant levels of PCBs and DDTs, measured in whole H. ecocellataindividuals collected in downstream sites, were quite high compared with reported information in related aquatic insect species. Bartrons et al. (2007) reported levels of PCBs, DDTs and HCH in trichoptera larvae collected from alpine lakes of 20, 5, 2 ng/g d.w.,

3608

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 9 e3 6 1 3

Very good Good Fair

GST LDH

1.0

CAT

GR

B4s

Poor Very poor

PC2 (22.10%)

0.5 B6 B6s

GPX SOD

LPO

0.0 B7s

-0.5

B1

B1s

CHE

CBE

B5s B5

B3

DNA

B2

-1.0 -1.0

-0.5

0.0

0.5

1.0

PC1 (34.5%)

Fig. 3 e Biplots of the first two components of a PCA performed on measured biochemical responses of H. exocellata collected at the studied sites and seasons. Loading abbreviations are depicted in the text. Sample collection scores are depicted as mean values with their 95% CI. Symbols indicate the ecological quality groups of the studied macroinvertebrate communities according to IBMWP classification. Summer samples are identified by “s” after the sample code.

respectively. Kovats and Ciborowski (1993) reported levels of PCBs ranging from 100 to 300 ng/g d.w. in hydropsychidae larvae collected along St Clair, Detroit and Niagara North American Rivers. Bizzotto et al. (2009), in macroinvertebrate taxa with a similar foraging behaviour (i.e. collectors) as Hydropsyche sp, found PCB, DDT and HCH levels of 16e190, 8e210, 1e12 ng/g d.w, respectively, in Alpine River streams from Italy. PAHs levels were of similar magnitude of those reported in larvae of aquatic insects collected from oil-contaminated wetlands (Wayland et al., 2008). Therefore, the organic contaminant levels measured in H. exocellata collected from the Beso´s River were on the top high range reported for this and related species. Estimated lethal toxic equivalent levels for the measured organic pollutant body burdens were in many sites within two orders of magnitude of those causing lethal effects in macroinvertebrate species like amphipods or midges (Table 2). Chronic effects on growth and reproduction usually occur at exposure levels ten to a hundred fold lower than those causing lethality (Roex et al., 2000) and biomarkers may be effected at even lower concentrations (Dama´sio et al., 2008). Therefore, measured body burdens for organic contaminants were likely to challenge H. exocellata physiology. Biomarkers and chemical contaminants measured in wild organisms have been widely used worldwide to biomonitor detrimental effects of pollutants in the field. In few occasions, however, biomarkers have been used to determining the ecological water quality of surface waters according to WFD (Hagger et al., 2006, 2008; Jemec et al., 2010; Sanchez and Porcher, 2009; Solimini et al., 2009; Vighi et al., 2006). The biochemical based classification obtained in this study (Fig. 3) match quite well with the five ecological quality types estimated according to the ACA (Munne´ and Prat, 2009). Our

biochemical based classification was able to differentiate two reference sites having very good (B1, B2), one having a good (B3) quality scores, two more groups of sites with moderate and poor ecological qualities (B4-B6, B5) and other site B7 that had a very poor ecological quality. A further analysis combining biochemical and ecological traits and a broad range of environmental factors and contaminants identified different associations and contributing factors (Fig. 4). The principal one, defined a clear salinity stress gradient and separated macroinvertebrate assemblages and H. exocellata individuals having a good ecological status and high activities of B-esterases (ChE, CbE) versus those having a poor ecological quality and, high levels of DNA strand breaks. In this sense it seems that some biomarkers (e.g DNA strand breaks) may be used as early warning indicators of stress and hence predict future problems in ecological status. It is known that conductivity above 1 mS/cm had measurable detrimental effects on macroinvertebrate communities similar to those studied here (Kefford et al., 2011). Salinity is also known to affect fluctuating asymmetry of H. exocellata larvae, a trait that is physiological linked with ontogenetic effects and hence with DNA integrity (Bonada et al., 2005). Therefore the observed increased levels of DNA strand breaks in H. exocellata larvae exposed to high salinity levels is likely to preclude individual level effects (e.g. mortality, growth, fecundity) as a result of DNA damage. The second source of variability was further studied by performing a PCA on residual variation after regressing all variables against salinity. The results still indicate substantial levels of variability that were explained by high activities of antioxidant and metabolizing enzymes, which were associated with presumable toxic levels of accumulated polycyclic aromatic hydrocarbons (GR vs. PAH), organochlorine pesticides and detergents (GPX, CAT vs. ENDO, AP; SOD vs. HCH). Therefore, these markers were affected differently than biological metrics of macroinvertebrate assemblages to environmental stress and hence they could provide additional information to assess ecological quality of benthic communities, for example for detecting specific detrimental sublethal effects of organic contaminants rather than those associated to high loads of nutrients, salinization and habitat degradation. Furthermore, in a long term the measured detrimental biochemical effects, if persist as it was shown in this study in spring and summer samples, can be lethal and hence affect community structure. It is noticeable that as expected and according to the low levels of metals measured in H. exocellata larvae, most measured biochemical changes were poorly related with metal body burdens, thus the PCA based biochemical-contaminant associations provided here agrees with measured toxic levels of contaminants in Hydropsyche samples. The antioxidant and phase II metabolizing enzymes SOD, CAT, GPX, GR and GST act detoxifying reactive oxygen species and secondary metabolites (Livingstone, 2001) that may also include peroxidated products (Ketterer et al., 1983), thus they are physiologically linked with oxidative stress (Halliwell and Gutteridge, 1999). Accordingly the relationship of these four markers with pro-oxidant factors such as organic pollutants is expected to occur (Di Giulio et al., 1995; Halliwell and Gutteridge, 1999). High activities of lactate dehydrogenase have been associated to increased metabolism under stressful conditions (Menezes et al., 2006; Moreira et al.,

3609

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 9 e3 6 1 3

2006), thus its association with the previous enzymes and pollution is also reasonable. Although limited to just seven sites and ten sample collections, the main conclusions obtained within the Besos River basin were quite similar to those found in seven sites and

14 sample collections from the Llobregat River basin. Firstly, it was clear that the greater and diverse the number of biochemical responses measured the greater was the ability to differentiate H. exocellata sample collections within communities. In both river basins the use of just five

Very good Good Fair

A

1.0

B6

Poor Very poor

Zn

GR

Cu

Al As Pb B6s pH O2

0.5

CHECBE

PCB

IASPTS

PC2 (19.1%)

IBMWP B1 IHF

B1s

GPX

B2

0.0

GST LDH CAT

LPO Flow

B5

SOD

-0.5

DDT PAH

T

SO4 Cl ENDO Cond

B5s

QBR AP

NO3 HCH DNA

B4s

NH4

B3

-1.0

-1.0

B7s PO4

NO2

-0.5

0.0

0.5

1.0

PC1(47.9%)

B

B4s

1.0

PC2 (21.3%)

HCH

CAT APGPX

ENDO

LDH T GST CHE

SOD

0.5

B1s

B7s QBR PO4 NO2

B1

B6s

DDT

IBMWP IASPT S

NH4

0.0

Flow

O2

PAH

Zn

GR

CBE pH DNA

-0.5

B3

B5s

B2 B6

LPO IHF NO3

Cu Pb PCB

Al As

B5

-1.0

-1.0

-0.5

0.0

0.5

1.0

PC1(36.0%) Fig. 4 e Biplots of the first two components of a PCA performed on (A) measured mean ecological (IBMWP, IASTP, S, IHF, QBR), biochemical (SOD, CAT, GST, GPX, GR, CHE, CBE, LPO, DNA), water physicochemical (T, O2, Cond, Cl, SO4, NH4, NO3, NO2, PO4, Flow) and contaminant body burdens (Al, As, Cu, Zn, Pb, PCB, DDT, PAH, ENDO, HCH, AP), measured at the studied sites and seasons. Graph (B) includes the PCA results performed on the residuals of the above mentioned variables except (Cond, SO4) regressed against Cl. Full names of loading abbreviations are depicted in the text. Symbols indicate the ecological quality groups of the studied macroinvertebrate communities according to IBMWP classification. Summer samples are identified by “s” after the sample collection code.

3610

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 9 e3 6 1 3

A1 Very good Good Fair Poor Very poor

GPX

PC2 (23.7%)

A

SOD

SOD

1.0

GST

B5

B1s,B2 B4s

B3

B6 B5s B6s B7s

B1

GST

GPX

CAT

LPO

PC2 (26.5%)

0.5 L3s

0.0

L5 s

L3

L6s

CAT LIPID L6

L4

L1

-0.5

PC1 (39.0%)

L7s

L4s

L2, L2s L1s

L5

L7

-1.0

-1.0

-0.5

0.0

0.5

1.0

PC1 (38.1%)

B1 L7

PC2 (23.7%)

CAT

B 1.0 S IBMWP SOD

IHF

PC2 (17.6%)

Cr

L3s

0.5

QBR

L1s L2

L3

L5s

L5s

L4s

OP

QBR GPX IASTP

SOD

Cu

Al Ni Cr Co T

S IBMWP

IHF

L3s

PC1(27.8%)

NH4 PO4

O2

-0.5

L2s Zn

NO3

PAH

L7

-1.0

L6s

L2

L1s Pb

Pb OCL

Cl

GST

LPO

pH

-1.0

pH

L7s

NO3 NO2 ZnCond

L1

-0.5

GST

O2

TRZ AP Fe

L7s

L6

L4

Flow NO2

L6

L4

L3

L6s

Cu AP Fe OP TRZ SO4

Flow GPX

LPO PAH

L1

Ni

L4s

L2sIASPT

0.0

Co Al T

OCL

L5

PO4 NH4

L5

0.0

CAT

0.5

1.0

PC1(38.9%) Fig. 5 e Biplots of the first two components of PCAs performed on (A, A1) reported biochemical (SOD, CAT, GST, GPX, LPO) responses of H. exocellata collected along the Llobregat River Basin in 2003 (Barata et al., 2005) and (B, B1) ecological (IBMWP, IASTP, S, IHF, QBR), biochemical (SOD, CAT, GST, GPX, LPO), water physicochemical (T, O2, Cond, Cl, SO4, NH4, NO3, NO2, PO4, Flow), metal body burdens (Fe, Al, Zn, Cu, Pb, Cr, Ni, Co) and contaminants in water (PAH, OCL, TRZ, AP, OP), measured at the studied seven sites and seasons. The inlet graph B1 includes the PCA results performed on the residuals of the above mentioned variables except (Cond, SO4) regressed against Cl. For comparison purposes the inlet graph (A1) includes also the PCA results performed on Beso´s data limited to just SOD, CAT, GST, GPX, LPO responses. In graphs A and A1 sample collection scores are depicted as mean values with their 95% CI. Loading abbreviations are explained in the text. Symbols indicate the ecological quality groups of the studied macroinvertebrate communities according to IBMWP classification. Summer samples are identified by “s” after the sample code.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 9 e3 6 1 3

biomarkers belonging to the same detoxication metabolic path (oxidative stress) showed the same power to distinguish samples having a good and poor ecological quality (Fig. 5 A, A1). In both rivers CAT and SOD were the markers contributing most in explaining data variance (Fig. 5 A, A1). The inclusion of five more markers belonging to different metabolic paths in the Beso´s River increased substantially the discrimination power of sample collections within sites (Fig. 3). Secondly, in both river systems salinity was one of the major environmental factor explaining detrimental changes in macroinvertebrate assemblages, whereas most of the studied biomarkers responded differently (Figs. 4 and 5). In the Llobregat River the inability of the studied biomarkers to be associated with organic pollutants is likely to be related to the fact that these pollutants were point measures in water and hence did not necessarily reflect bio-availability levels that may encounter H. exocellalta larvae during their development (Fig. 5B). Note also that contrary to the Beso´s dataset (Figs. 3 and 4), that of Llobregat showed a high seasonality (Fig. 5A,B), which was related to an exceptional warm and dry summer (Barata et al., 2005).

5.

Conclusions

One of the greatest efforts of environmental state agencies for the implementation of the WFD has been to develop robust and harmonized ecological monitoring tools to assess the ecological quality of surface waters across EU countries (Munne´ and Prat, 2009), which implies a considerable effort, time and money. Biomarkers, although not incorporated in the WFD, are among the emerging biological monitoring tools considered for implementation of the WFD (Allan et al., 2006; Mills et al., 2007). By 2020, EU member states will have to improve the quality of their surface waters and report those changes to the WFD. In this sense, the use of markers sensitive to water pollution may provide useful information on small changes in ecological quality specially in the threshold value between moderate and good. Here, the studied biomarkers provided insights of the actual ecological status using biological indexes within communities. Although biomarkers play a great role in ecotoxicology and environmental risk assessment, they are sometimes difficult to interpret (Budka et al., 2010). It is problematic to determine whether a single biomarker response is an indicator of impairment or is a part of the homeostatic response, indicating that an organism is successfully dealing with the exposure (Forbes et al., 2006). In the majority of the studies conducted with invertebrates it is impossible to use the same animal for the whole battery of biochemical and chemical determinations, because of the limited quantity of biological material available. This dramatically reduces the quality of biomarker data for diagnostic purposes using multivariate methods due missing values, high dimensionality (there are often more variables than samples) and the small size of dataset (Budka et al., 2010). The above mentioned problems were minimized in the present study by using a large set of biomarkers, including several reference sites and measurements at different seasons, and by reducing missing values due to the fact that all biochemical responses were measured in the same individual and the use of a species (H. exocellata), which is abundant and widespread in

3611

the area form pristine to polluted sites. Therefore, our approach was specially robust differentiating impacted and reference sample collections within sites (Fig. 3) but not those collected from the same site in different months within the Beso´s River basin. The previous results indicate that populations from the same site within the Beso´s, collected in different seasons maintain the stress due to pollutants, which implies a continuous source of the substances producing such effect. This may create the basis to improve the sewage plants in order to diminish the substances present in the river. Also the determination of biochemical markers in sites that are in good or moderate state will give us good information of the health of their populations. Sites in good ecological status that have some of the biomarkers of stress activated will imply that the site may change to a lower status in the future and therefore measures can be taken to improve the ecological quality. The usefulness of the studied biomarkers as complementary tools to diagnose the cause of community impairment was also tested in a river system impacted by a pesticide. Pue´rtolas et al. (2010) studying the impact of the application of glyphosate in the riverbank of four Llobregat macroinvertebrate communities affected by habitat degradation, salinization and wastewater effluents, reported that biochemical response of H. exocellata larvae but not macronvertebrate asseblages responded negatively to pesticide application. Furrhermore, previous studies performed in different macroinvertebrate species (Daphnia magna, Corbicula fluminea, Procambarus clarki and Dreissena polymorpha) and impacts (pesticides, organochlorine compounds and mercury) have also indicated a consistent biomarker pattern of response across species (Daphnia magna and H exocellata in Pue´rtolas et al., 2010; D. magna and Corbicula fluminea in Barata et al., 2007; Damasio et al., 2010); and most importantly the ability of combining multivariate and multi-biomarker methods to diagnose the cause of community impairments (Dama´sio et al., 2008, 2010; Faria et al., 2010; Barata et al., 2007; Pue´rtolas et al., 2010). Nevertheless, further studies including more river types, environmental stressors, macroinvertebrate species and biochemical or molecular markers are needed to generalize our findings and to establish thresholds of biomarkers that should not be trespassed (particularly in the boundary between good and moderate status).

Acknowledgements This work is supported by the Spanish and Portuguese Ministry of Education and Science (CGL2004-03514/HD, CGL2008-01898; FCOMP-01-0124-FEDER-007069). Joana Dama´sio was supported by FCT PhD grant (SFRH/BD/23269/2005). We thank two anonymous referees, whose comments have improved the manuscript.

references

Allan, I.J., Vrana, B., Greenwood, R., Mills, G.A., Roig, B., Gonzalez, C., 2006. A “toolbox” for biological and chemical monitoring requirements for the European Union’s Water Framework Directive. Talanta 69 (2), 302e322.

3612

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 9 e3 6 1 3

ASTM, 1998. Sandard methods for the examination of water and wastewater, 20th. Edn. American Public Health Association, Washington, D.C. Baird, D.J., Burton, G.A., 2001. Ecological Variability: separating Natural from Antropogeneic Causes of Ecosystem Impairment. SETAC Press, Brussels, Belgium. Barata, C., Solayan, A., Porte, C., 2004. Role of B-esterases in assessing toxicity of organophosphorus (chlorpyrifos, malathion) and carbamate (carbofuran) pesticides to Daphnia magna. Aquatic Toxicology 66 (2), 125e139. Barata, C., Lekumberri, I., Vila-Escale´, M., Prat, N., Porte, C., 2005. Trace metal concentration, antioxidant enzyme activities and susceptibility to oxidative stress in the tricoptera larvae Hydropsyche exocellata from the Llobregat River basin (NE Spain). Aquatic Toxicology 74 (1), 3e19. Barata, C., Dama´sio, J., Lo´pez, M.A., Kuster, M., Lo´pez de Alda, M., Barcelo´, D., Riva, M.C., Raldu´a, D., 2007. Combined use of biomarkers and in situ bioassays in Daphnia magna to monitor environmental hazards of pesticides in the field. Environmental Toxicology and Chemistry 26 (2), 370e379. Barbour, M.T., Paul, M.J., 2010. Adding value to water resource management through biological assessment of rivers. Hydrobiologia 651, 17e24. Bartrons, M., Grimalt, J.O., Catalan, J., 2007. Concentration changes of organochlorine compounds and polybromodiphenyl ethers during metamorphosis of aquatic insects. Environmental Science & Technology 41 (17), 6137e6141. Basset, A., 2010. Aquatic science and the water framework directive: a still open challenge towards ecogovernance of aquatic ecosystems. Aquatic Conservation: Marine and Freshwater Ecosystems 20, 245e249. Bizzotto, E.C., Villa, S., Vighi, M., 2009. POP bioaccumulation in macroinvertebrates of alpine freshwater systems. Environmental Pollution 157 (12), 3192e3198. Bocchetti, R., Fattorini, D., Pisanelli, B., Macchia, S., Oliviero, L., Pilato, F., Pellegrini, D., Regoli, F., 2008. Contaminant accumulation and biomarker responses in caged mussels, Mytilus galloprovincialis, to evaluate bioavailability and toxicological effects of remobilized chemicals during dredging and disposal operations in harbour areas. Aquatic Toxicology 89 (4), 257e266. Bonada, N., Zamora-Mun˜oz, C., Rieradevall, M., Prat, N., 2004. Ecological profiles of caddis fly larvae in Mediterranean streams: implications for bioassessment methods. Environmental Pollution 132 (3), 509e521. Bonada, N., Prat, N., Resh, V.H., Statzner, B., 2006. Developments in aquatic insect biomonitoring: a comparative analysis of recent approaches. Annual Review of Entomology 51, 495e523. Bonada, N., Vives, S., Rieradevall, M., Prat, N., 2005. Relationship between pollution and fluctuating asymmetry in the pollutiontolerant caddis fly Hydropsyche exocellata (Trichoptera, Insecta). Archiv Fur Hydrobiologie 162 (2), 167e185. Budka, M., Gabrys, B., Ravagnan, E., 2010. Robust predictive modelling of water pollution using biomarker data. Water Research 44, 3294e3308. Cain, D.J., Luoma, S.N., 1998. Metal exposures to native populations of the caddis fly Hydropsyche (Trichoptera: Hydropsychidae) determined from cytosolic and whole body metal concentrations. Hydrobiologia 386, 103e117. Cain, D.J., Luoma, S.N., Wallace, W.G., 2004. Linking metal bioaccumulation of aquatic insects to their distribution patterns in a mining-impacted river. Environmental Toxicology and Chemistry 23 (6), 1463e1473. Chapman, D., Kimstach, V., 1996. In: Chapman, D. (Ed.), Water Quality Assessments. A Guide to the Use of Biota, Sediments and Water in Environmental Monitoring. E & FN SPON, London, pp. 59e126.

Damasio, J., Navarro-Ortega, A., Tauler, R., Lacorte, S., Barcelo´, D., Soares, A.M.V.M., Lopez, M.A., Riva, M.C., Barata, C., 2010. Identifying major pesticides affecting bivalve species exposed to agricultural pollution using multibiomarker and multivariate methods. Ecotoxicology 19 (6), 1084e1094. Dama´sio, J., Tauler, R., Teixido´, E., Rieradevall, M., Prat, N., Riva, M.C., Soares, A.M.V.M., Barata, C., 2008. Combined use of Daphnia magna in situ bioassays, biomarkers and biological indices to diagnose and identify environmental pressures on invertebrate communities in two Mediterranean urbanized and industrialized rivers (NE Spain). Aquatic Toxicology 87 (4), 310e320. Di Giulio, R.T., Benson, W.H., Sanders, B.M., Van Veld, P.A., 1995. In: Rand, G. (Ed.), Fundamentals of Aquatic Toxicology, Effects, Environmental Fate, and Risk Assessment. Taylor & Francis, London, pp. 523e561. Dole´dec, S., Statzner, B., 2010. Responses of freshwater biota to human disturbances: contribution of J-NABS to developments in ecological integrity assessments. Journal of the North American Benthological Society 29 (1), 286e311. Faria, M., Huertas, D., Soto, D.X., Grimalt, J.O., Catalan, J., Riva, M.C., Barata, C., 2010. Contaminant accumulation and multibiomarker responses in field collected zebra mussels (Dreissena polymorpha) and crayfish (Procambarus clarkii), to evaluate toxicological effects of industrial hazardous dumps in the Ebro River (NE Spain). Chemosphere 78 (3), 232e240. Fay, A.A., Brownawell, B.J., Elskus, A.A., Mcelroy, A.E., 2000. Critical body residues in the marine amphipod Ampelisca abdita: sediment exposures with nonionic organic contaminants. Environmental Toxicology and Chemistry 19 (4), 1028e1035. Forbes, V.E., Palmqvist, A., Bach, L., 2006. The use and misuse of biomarkers in ecotoxicology. Environmental Toxicology and Chemistry 25 (1), 272e280. Hagger, J.A., Jones, M.B., Leonard, D.R.P., Owen, R., Galloway, T.S., 2006. Biomarkers and integrated environmental risk assessment: are there more questions than answers? Integrated Environmental Assessment and Management 2 (4), 312e329. Hagger, J.A., Jones, M.B., Lowe, D., Leonard, D.R.P., Owen, R., Galloway, T.S., 2008. Application of biomarkers for improving risk assessments of chemicals under the water framework directive: a case study. Marine Pollution Bulletin 56 (6), 1111e1118. Halliwell, B., Gutteridge, J.M.C., 1999. Free Radicals in Biology and Medicine. Oxford University Press, Oxford. Hering, D., Johnson, R.K., Kramm, S., Schmutz, S., Szoszkiewicz, K., Verdonschot, P.F.M., 2006. Assessment of European streams with diatoms, macrophytes, macroinvertebrates and fish: a comparative metric-based analysis of organism response to stress. Freshwater Biology 51, 1757e1785. Ja´imez-Cue´llar, P., Vivas, S., Bonada, N., Robles, S., Mellado, A., ´ lvarez, M., Avile´s, J., Casas, J., Ortega, M., Pardo, I., 2002. A Protocol GUADALMED (PRECE). Limnetica 21, 187e204. Jemec, A., Drobne, D., Tisler, T., Sepcie´, K., 2010. Biochemical biomarkers in environmental studies-lessons learnt from enzymes catalase, glutathione S-transferase and cholinesterase in two crustacean species. Environmental Science and Pollution Research 17 (3), 571e581. Kefford, B.J., Marchant, R., Schafer, R.B., Metzeling, L., Dunlop, J.E., Choy, S.C., Goonan, P., 2011. The definition of species richness used by species sensitivity distributions approximates observed effects of salinity on stream macroinvertebrates. Environmental Pollution 159 (1), 302e310.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 9 9 e3 6 1 3

Ketterer, B., Coles, B., Meyer, D.J., 1983. The role of glutathione in detoxication. Environmental Health Perspectives 49, 59e69 (MAR). Kovats, Z.E., Ciborowski, J.J.H., 1993. Organochlorine contaminant concentrations in caddisfly Adults (Trichoptera) collected from great-lakes connecting channels. Environmental Monitoring and Assessment 27 (2), 135e158. Landrum, P.F., Lotufo, G.R., Gossiaux, D.C., Gedeon, M.L., Lee, J.-H., 2003. Bioaccumulation and critical body residue of PAHs in the amphipod, Diporeia spp.: additional evidence to support toxicity additivity for PAH mixtures. Chemosphere 51, 481e489. Liess, M., Von der Ohe, P.C., 2005. Analyzing effects of pesticides on invertebrate communities in streams. Environmental Toxicology and Chemistry 24 (4), 954e965. Livingstone, D.R., 2001. Contaminant-stimulated reactive oxygen species production and oxidative damage in aquatic organisms. Marine Pollution Bulletin 42 (8), 656e666. Liu, M.Y., 1993. High affinity binding of [3h]imidacloprid in the insect acetylcholine receptor. Pesticide Biochemistry and Physiology 45 (3), 40e46. Menezes, S., Soares, A.M.V.M., Guilhermino, L., Peck, M.R., 2006. Biomarker responses of the estuarine brown shrimp Crangon crangon L. to non-toxic stressors: temperature, salinity and handling stress effects. Journal of Experimental Marine Biology and Ecology 335 (1), 114e122. Mills, G.A., Greenwood, R., Gonzalez, C., 2007. Environmental monitoring within the water framework directive (WFD). Trends in Analytical Chemistry 26, 450e453. Moreira, S.M., Lima, I., Ribeiro, R., Guilhermino, L., 2006. Effects of estuarine sediment contamination on feeding and on key physiological functions of the polychaete Hediste diversicolor: laboratory and in situ assays. Aquatic Toxicology 78 (2), 186e201. Munne´, A., Prat, N., Sola`, C., Bonada, N., Rieradevall, M., 2003. A simple field method for assessing the ecological quality of riparian habitat in rivers and streams: QBR index. Aquatic Conservation-Marine and Freshwater Ecosystems 13 (2), 147e163. Munne´, A., Prat, N., 2009. Use of macroinvertebrate-based multimetric indices for water quality evaluation in Spanish Mediterranean Rivers: an intercalibration approach with the IBMWP index. Hydrobiologia 628 (1), 203e225. Munne´, A., Prat, N., 2011. Effects of Mediterranean climate annual variability on stream biological quality assessment using macroinvertebrate communities. Ecological Indicators 11 (2), 651e662. Prat, N., Munne´, A., 2000. Water use and quality and stream flow in a Mediterranean stream. Water Research 34 (15), 3876e3881. Prat, N., Rieradevall, M., 2006. 25-years of biomonitoring in two Mediterranean streams (Llobregat and Beso´s basins, NE Spain). Limnetica 25 (1e2), 541e550. Pue´rtolas, L., Dama´sio, J., Barata, C., Soares, A.M.V.M., Prat, N., 2010. Evaluation of side-effects of glyphosate mediated

3613

control of giant reed (Arundo donax) on the structure and function of a nearby Mediterranean River ecosystem. Environmental Research 110 (6), 556e564. Roex, E.W.M., Van Gestel, C.A.M., Van Wezel, A.P., Van Straalen, N. M., 2000. ratios between acute aquatic toxicity and effects on population growth rates in relation to toxicant mode of action. Environmental Toxicology and Chemistry 19 (3), 685e693. Rosenberg, D.M., Resh, V.H., 1993. Freshwater Biomonitoring and Benthic Macroinvertebrates. Chapman & Hall, New York. Sanchez, W., Porcher, J.M., 2009. Fish biomarkers for environmental monitoring within the water framework directive of the European Union. Trends in Analytical Chemistry 28, 150e158. Sa´nchez-Avila, J., Quintana, J., Ventura, F., Tauler, R., Duarte, C.M., Lacorte, S., 2010. Stir bar sorptive extraction-thermal desorption-gas chromatography-mass spectrometry: an effective tool for determining persistent organic pollutants and nonylphenol in coastal waters in compliance with existing directives. Marine Pollution Bulletin 60, 103e112. Sola`, C., Prat, N., 2006. Monitoring metal and metalloid bioaccumulation in Hydropsyche (Trichoptera, Hydropsychidae) to evaluate metal pollution in a mining river. Whole body versus tissue content. Science of the Total Environment 359 (1e3), 221e231. Solimini, A.G., Ptacnik, R., Cardoso, A.C., 2009. Towards holistic assessment of the functioning of ecosystems under the water framework directive. Trends in Analytical Chemistry 28 (2), 143e149. Traas, T.P., Van Wezel, A.P., Hermens, J.L.M., Zorn, M., 2004. Environmental quality criteria for organic chemicals predicted from internal effect concentrations and a food web model. Environmental Toxicology and Chemistry 23 (10), 2518e2527. van der Oost, R., Beyer, J., Vermeulen, N.P.E., 2003. Fish bioaccumulation and biomarkers in environmental risk assessment: a review. Environmental Toxicology and Pharmacology 13 (2), 57e149. Vighi, M., Finizio, A., Villa, S., 2006. The evolution of the environmental quality concept: from the US EPA red book to the European Water Framework Directive. Environmental Science and Pollution Research 13 (1), 9e14. Wayland, M., Headley, J.V., Peru, K.M., Crosley, R., Brownlee, B.G., 2008. Levels of polycyclic aromatic hydrocarbons and dibenzothiophenes in wetland sediments and aquatic insects in the oil sands area of Northeastern Alberta, Canada. Environmental Monitoring and Assessment 136, 167e182. Wold, S., Jonsson, J., Sjo¨strom, M., Sandberg, M., Rannar, S., 1993. DNA and peptide sequences and chemical processes multivariately modelled by principal component analysis and partial least squares porjections to Latent Structures. Analytica Chimica Acta 277 (2), 239e253. Zar, J.H., 1996. Bioestatistical Analysis. Prentice-Hall International, Inc, New Jersey.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 1 4 e3 6 2 6

Available at www.sciencedirect.com

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

Preparation, characterisation and application of novel composite coagulants for surface water treatment N.D. Tzoupanos, A.I. Zouboulis* Division of Chemical Technology, Department of Chemistry, Aristotle University of Thessaloniki, GR-54124 Thessaloniki, Greece

article info

abstract

Article history:

The development of the Inorganic Polymeric Flocculants (IPFs) can be regarded as signifi-

Received 3 November 2010

cant progress in the coagulation-flocculation field. However, the IPFs may be less efficient

Received in revised form

when compared to the organic polymers (polyelectrolytes) regarding their aggregation

11 March 2011

abilities. In order to increase further their flocculation efficiency, the combination of

Accepted 5 April 2011

a cationic IPF (polyaluminium chloride, PACl) and an anionic polyelectrolyte in one unique

Available online 21 April 2011

reagent is proposed in this study. During this investigation, several composite coagulants were prepared, which differ on the preparation method and polyelectrolyte content. Major

Keywords:

typical properties of the prepared coagulants were examined, i.e. pH, turbidity, conduc-

Composite coagulants

tivity, Al species distribution. The composition, structure and morphology of the composite

Polyaluminium chloride

coagulants were studied in detail as well, with the application of FT-IR, XRD and SEM

Anionic polyelectrolyte

techniques. Their coagulation performance was investigated in the treatment of a model

Application

water sample (simulating surface water) and compared to the respective coagulation

Photometric dispersion analyser

performance of PACl and the polyelectrolyte applied as separated reagents (common

Surface water treatment

procedure). Finally, the kinetics of coagulation was studied with application of the Photometric Dispersion Analyser (PDA). From the results, it was revealed that interactions take place between the Al species and the polyelectrolyte molecules, which probably lead to the formation of new, “composite” species. The properties of the composite coagulants are significantly affected by these interactions, leading to more effective water treatment. The simplification of the overall treatment process and the cost-effectiveness are considered as the major advantages of the composite coagulants. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

The Inorganic Polymeric Flocculants (IPFs), or pre-polymerised coagulants, such as polyaluminium chloride (PACl, in the case of Al-coagulants) represent a relatively new category of coagulation reagents, which was developed in order to increase the efficiency of coagulation-flocculation process. However, despite the fact that the existence of polymerised metal species in the composition of the IPFs (e.g. keggin-Al13, etc., in the case of Al-coagulants) enabled them to perform more

efficiently than the conventional coagulants such as alum (Sinha et al., 2004; Crittenden et al., 2005), there is still need for further improvement of their properties. The main reason is the insufficient aggregation abilities of the IPFs, which usually imposes the use of a flocculant aid (polyelectrolyte) to increase the efficiency of flocculation process. The main reason for the higher efficiency of organic polymers regarding flocculation is their higher molecular weight (MW), which implies better aggregation properties. Thus, the increase of molecular weight and size of the pre-polymerised

* Corresponding author. Tel./fax: þ30 2310 997794. E-mail address: [email protected] (A.I. Zouboulis). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.009

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 1 4 e3 6 2 6

coagulants ingredients is thought to be the way for further improvement. This increase can be achieved through the combination of a pre-polymerised coagulant and a suitable additive (inorganic or organic) into one reagent. Synthetic polyelectrolytes have been utilized in coagulation/flocculation process for water purification for more than four decades. Their principal uses in water or wastewater treatment are as primary coagulants (cationic polyelectrolytes), as well as in the more traditional flocculation step of further binding the already formed small flocs into larger aggregates (flocculant aids, anionic and non-ionic polyelectrolytes) (Mortimer, 1991; Bratby, 2006). Due to their wide usage, it is not surprisingly to consider them as alternative additives in the pre-polymerised coagulants composition for the production of new modified coagulation reagents. Apart from the expected increase of components size and molecular weight in the composite coagulants, the utilization of polyelectrolytes exhibits several other advantages: the inorganic coagulant (e.g. PACl) and the organic polyelectrolyte will be combined in one reagent, thus avoiding the subsequent addition of a flocculant aid (polyelectrolyte) after coagulant addition (inorganic salt) in order to enhance the flocculation process. In this way, the overall treatment procedure is simplified and the overall cost-effectiveness is also improved, as there will be no need for specific equipment for handling the polyelectrolyte which is usually delivered in solid form (e.g. dissolution and pumping system). Moreover, the introduction of the polyelectrolyte into the structure of the coagulant is expected to reduce to a certain extent the residual toxicity due to remaining un-reacted monomers of the polymer (providing that interactions taking place between the metal species and the polyelectrolyte molecules lead to complexation reactions), a common issue when polyelectrolytes are applied in water treatment (Tzoupanos and Zouboulis, 2008). The aforementioned suggestions make the investigation for the preparation, characterisation and application of composite coagulants containing organic polymers as additives more attractive. Several relevant efforts have been conducted during the past few years by several researchers (Al-coagulants), e.g. Tang and Shi (2002), Gao et al. (2005) and Tzoupanos and Zouboulis (2010). However, these studies were focused mainly on composite coagulants derived from PACl and a cationic polyelectrolyte (p-DADMAC), whereas the combination of PACl and anionic polyelectrolyte has not been studied in detail yet. Considering that in the case of composite coagulants containing cationic polyelectrolytes, the use of a supplementary flocculant aid can not be avoided and that the anionic polymers usually have a higher molecular weight than the non-ionic polymers, thus implying better flocculation properties, the further investigation regarding the combination of inorganic coagulants with anionic polyelectrolytes seems to be promising. Between the anionic polyelectrolytes, the most common are those derived from acrylamide polymerisation (poly-acrylamides). They are commercially available as high molecular weight flocculant agents (Mortimer, 1991; Bolto, 1995). In this study, a commonly used poly-acrylamide (co-polymer of acrylamide, i.e. an anionic polyelectrolyte) used in water or wastewater treatment facilities (especially in Greece) with the commercial name Magnafloc LT-25 was used for the preparation of several new composite coagulants. The new products were derived by

3615

applying the optimum preparation conditions (as determined through preliminary experiments) of PACl and Magnafloc LT25. In total, 8 composite coagulant agents were prepared, with the application of two preparation methods, i.e. copolymerisation and composite polymerisation, and different anionic polyelectrolyte (APE) content (i.e. Al/APE ¼ 5e20 w/w). The impact of APE addition in the major typical properties of obtained products was examined, such as pH, turbidity, conductivity and Al species distribution. Moreover, an extended infrared spectroscopy study was conducted (FT-IR), in order to investigate possible alterations of the chemical bonds in the initial PACl reagent, which could lead to the identification of new, composite species between Al and polyelectrolyte molecules. The morphology of the products was investigated with Scanning Electron Microscopy (SEM) and their composition was further studied by obtaining X-Ray Diffraction patterns (XRD). Their coagulation performance in water treatment was thoroughly studied and compared with the respective coagulation performance of PACl and Magnafloc LT-25, but applied as separated reagents (i.e. following the commonly applied procedure). Finally, the kinetics and dynamics of flocculation ability were studied by using the Photometric Dispersion Analyser (PDA), a technique which allows the (relative) comparison of flocs’ growth rate and extent for the tested coagulants. The aim of the study was the combination of better charge neutralization abilities (coagulation,) as showed by the inorganic coagulant agent and the better aggregation abilities (flocculation), as shown by the polyelectrolyte in one unique reagent. In this way, the efficiency of the whole process of coagulation/flocculation is expected to be improved and additionally, the applications field will be further expanded, thus minimizing the need of simultaneous “optimisation” of both reagents.

2.

Materials and methods

All chemical reagents used were analytically pure chemicals. Deionised water with conductivity lower than 0.5 mS/cm was used to prepare all the solutions, except of the solutions used for the preparation of coagulants. In this case, deionised water made carbonate free by boiling, was used. A poly-acrylamide co-polymer (Magnafloc LT-25, Ciba SC LTD, commercially available) was obtained and used as the organic additive for the synthesis of composite coagulants. The specific polyelectrolyte is commonly used as a flocculant aid in water or wastewater treatment plants, especially in Greece.

2.1. Procedure for the preparation of composite coagulants The synthesis of coagulants took place by the application of two polymerisation methods, i.e. the co-polymerisation, or the composite polymerisation techniques. According to the first procedure, the appropriate amount of 0.135% w/v Magnafloc LT-25 solution was slowly added (addition rate 0.3 mL/min achieved by a peristaltic pump) under magnetic stirring into a predetermined amount of a 0.5 M AlCl3 solution under heating (50 C). Afterwards, the appropriate amount of 0.5 M NaOH solution was slowly added (addition rate

3616

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 1 4 e3 6 2 6

0.1 mL/min) in order to achieve the desired [OH]/[Al] ¼ 2 M ratio. According to the second technique, the basic solution (0.5 M NaOH) was initially added to the Al solution, creating an intermediate PACl solution and then, the appropriate amount of Magnafloc LT-25 solution was introduced under heating (50 C), in order to achieve the desired Al/APE ratio (w/w). The basic addition rate was 0.1 mL/min and the stirring speed was 700e800 rpm. The final volumes of the obtained solutions samples were around 40e65 mL and the final aluminium concentration was fixed for all coagulants at 0.1 M. The aforementioned experimental conditions were optimised after preliminary experiments. The total number of prepared modified composite coagulants was 8, with constant [OH]/[Al] ¼ 2 M ratio, Al/APE ratios (w/w) 5, 10, 15, or 20 and were prepared by using the two polymerisation techniques. Polyaluminium chloride solution was also prepared (PACl with [OH]/[Al] ¼ 2) for comparison reasons under the same conditions, but without the addition of polyelectrolyte. The coagulants prepared with the copolymerisation technique are referred as PAAPEC, whereas the coagulants prepared with the composite polymerisation technique are referred as PACAPE. “PAC” stands for PACl and “APE” was derived from “anionic polyelectrolyte”. According to the basicity and Al/APE ratio, the coagulants are referred as follows: PAAPEC with OH/Al ¼ 2 and Al/APE ¼ 10 as PAAPEC 2/ 10, while PACl with OH/Al ¼ 2 as PACl 2.

2.2.

Characterisation methods

2.2.1.

Aluminium species distribution

Aluminium species distribution was determined with the application of Al-ferron timed spectrophotometric method, which is based on the different reaction time of aluminium species with ferron reagent (8-hydroxy-7-iodoquinoline-5sulphonic acid) to form water soluble complexes in pH 5e5.2. These complexes absorb light with a maximum at 370 nm, hence absorbance measurements at this wavelength allow the calculation of the different species of aluminium. A UVeVis spectrophotometer (Shimadzu) was used for this purpose. The exact procedure was a modification of the method of Parker and Bertsch (1992), as further developed by Zhou et al. (2006).

2.2.2.

Structural and morphological characterisation

2.2.2.1. FT-IR. Specific amounts of samples were placed in glass beakers and kept in an oven at about 40 C for several days, in order to obtain dried powders. FT-IR spectra were recorded in the range of 4000e400 cm1 using a Perkin Elmer Spectrophotometer (Perkin Elmer, Spectrum One). The pellet was prepared by mixing 1 mg of the aforementioned powders with 200 mg KBr. 2.2.2.2. XRD. Specific amounts of samples were freeze dried (for 2e3 d, using a Christ, model Alpha 1-4 apparatus), the obtained solids were ground in a laboratory mortar and a pestle and the powder was kept in a desiccator until analysis. Samples of produced powders were characterised by X-Ray diffraction (XRD) for the determination of crystalline phases, using a Siemens D-500 X-Ray diffractometer with Cu K radiation in the range of 5e65 2q at a scan rate of 1 /min.

2.2.2.3. SEM. Small portions of coagulant powders obtained after drying in the oven (w40 C) were used to observe the morphology of the products, by employing a JEOL, JSM 840 scanning microscope.

2.3.

Coagulation performance

The zeta-potential was measured by using a Laser Zee Meter 501, the pH by using a Metrohm Herisau pH-Metre, the conductivity by using a Crison CM 35 Conductivity Metre and the turbidity measurements were performed by a HACH RATIO/XR Turbidimeter. The UV absorbance at 254 nm, as a convenient indicator of natural organic matter presence, was measured with a Shimadzu UV/Vis spectrophotometer, by using a 1 cm path length quartz cuvette. The residual aluminium concentration was determined with the eriochrome cyanine R standard method (Clesceri et al., 1989). In diluted and pH 6 buffered solutions, aluminium complexes with eriochrome cyanine R dye, resulting in a coloured compound, which absorbs light with maximum at 535 nm.

2.3.1. Jar-tests using contaminated tap water (simulating surface water) The purpose of the coagulation experiments was the evaluation of the coagulation performance of the new composite coagulants and the determination of the most effective properties (i.e. Al/APE ratio and the preparation technique), which lead to the most efficient product. Moreover, the performance of the most effective composite coagulant was compared with the respective coagulation performance of PACl and Magnafloc LT-25, but applied as separated reagents (commonly used procedure). For this purpose, a jar-test apparatus (Aqualytic) with six paddles was used. The treated sample (1 L) was made of tap water, clay (kaolin) suspension (commercially available) and humic acid (Aldrich). The initial concentration of clay suspended particles was 10 mg/L and that of humic acid 5 mg/L Table 1 displays the specific properties of the initial water sample. In the case of composite coagulants no further flocculant aid was used. In the case of PACl, Magnafloc LT-25 was used as flocculant aid in concentrations equal to the 1/10th of the respective concentration of PACl (resulted after preliminary experiments). The jar-test experimental conditions (based upon preliminary relevant experience) are presented in Table 1. 1 min after the addition of the coagulant (rapid mixing stage) about 30 mL of sample was withdrawn for z-potential measurements. The flocculant aid (in the case of PACl) was introduced just 15 s before the initialisation of slow mixing period. After a settling period of 40 min duration about 50 mL of sample was withdrawn 5 cm below the liquid surface for further analytical determinations. The concentrations of coagulants are expressed as mg Al/L in the case of PACl and as mg (Al þ APE)/L in the case of the composite coagulants.

2.3.2.

Study of coagulation kinetics

The extent of aggregation was also examined and the flocculation dynamics was accomplished by using a continuous flow optical flocculation monitor (PDA 2000, Rank Brothers, UK). The test suspension of 1.5 L tap water, containing 5 mg/L clay and

3617

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 1 4 e3 6 2 6

Table 1 e Properties of the model sample to be treated and jar-test conditions. Model sample properties Turbidity

Jar-test experimental conditions

Absorbance at 254 nm

16.5 NTU

0.125

pH

Rapid mixing stage Mixing rate

Duration

Mixing rate

Duration

160 rpm

120 s

45 rpm

10 min

7.65

5 mg/L humic acid, was placed in a 2 L beaker and stirred with the paddle of jar-test apparatus. The suspension flows through the transparent plastic measurement cell (3 mm diameter) using a peristaltic pump, where it was illuminated by a narrow light beam (850 nm wavelength). The pump was placed after the PDA apparatus for preventing the eventual floc breakage caused by the mechanical forces of the pump. The applied flow rate was 30 mL/min in order to have laminar conditions in the sampling tube, hence avoiding flocs breakage. All experiments were conducted at room temperature, without the addition of flocculant aid, to prevent the excessive growth of flocs and the blockage of connecting tubing (diameter 3 mm). The PDA measures the average transmitted light intensity (dc value) and the root mean square (rms value) of the fluctuating component. The RATIO (rms/dc), or Flocculation Index (FI ) provides a sensitive measure for the aggregation of particles. The RATIO value is strongly correlated with the respective floc size and always increases as flocs grow larger, providing a useful (although relative) indication of floc growth, eventually breakage and re-growth, which allows comparisons to be made between the different coagulants and under different shear conditions and coagulant concentrations (Kan et al., 2002; Yukselen and Gregory, 2004).

3.

observed that the addition of an anionic polyelectrolyte in a PACl solution results in increase of its turbidity. This increase is more intensive, when increasing the amount of APE in composite coagulants (i.e. when decreasing Al/APE ratio) and is partially due to solids formation during the preparation of coagulants. Specifically, it was observed that during the addition of APE solution, solids formed in the mixtures when Al/APE was lower than 15. The composite coagulants with Al/APE ¼ 20 were found to be clear solutions with substantially lower turbidity and absence of solids (denoted in Table 2 as appearance: 0). On the other hand, the composite coagulants with the highest content of APE (i.e. with Al/APE ¼ 5) were highly turbid solutions with well distinguishable solids (denoted as appearance: 3). Moreover, the specific reagents (with Al/APE ¼ 5) proved to be quite unstable, as during their storage (at room temperature) the solids formation continued and in less than 15 days from their preparation the turbidity increased suddenly and therefore they were not studied further. A possible reason for the observed turbidity increase in these composite coagulants is the increase of their components size, due to APE addition. This increase could be the result of interactions between APE molecules and the present Al species (e.g. electrostatic interactions, H-bonding etc.). Bridging effect and/or H-bonding can further increase the components size, thus resulting in the formation of well distinguishable solids, as in the case of PACAPE 2/5 and PAAPEC 2/5 and therefore, limiting the amount of APE that can be introduced to PACl. It should be mentioned, that samples of these solids were separated, dried and studied with FT-IR and it was found that they consisted mainly of polyelectrolyte molecules. Therefore, the restriction of limited APE dissolution in PACl solution should also be considered, when preparing these composite reagents.

Results and discussion

3.1. Characterisation of the prepared coagulation reagents 3.1.1.

Slow mixing stage

Physicochemical properties

Table 2 displays the major physicochemical properties of laboratory prepared composite coagulants and of PACl 2 solution with the same molar ratio [OH]/[Al] ¼ 2. It can be

Table 2 e Properties of laboratory prepared coagulants. Coagulant

PACl 2 PACAPE

PAAPEC

Al/APE

e 5 10 15 20 5 10 15 20

pH

4.11 e 4.34 4.39 4.33 e 4.43 4.38 4.40

Turbidity (NTU)

2.0 92 22.6 13.5 8.9 >200 154 89.5 63.5

Appearance (relative scale)

0 3 1 1 0 3 2 1 0

Conductivity (mS/cm)

25.8 e 25.8 25.4 25.3 e 25.5 25.9 24.7

Al species distribution (%) Ala

Alb

Alc

16 e 12 12 14 e 12 13 15

73 e 62 69 73 e 58 67 70

11 e 26 19 13 e 30 20 15

3618

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 1 4 e3 6 2 6

The alteration of Al species distribution is an indication of the existence of interaction between APE molecules and Al species. Particularly, it was observed (Table 2) that the addition of APE results in the decrease of the monomeric Al species (Ala) and of the medium-size Al polymerised species (Alb, where Al13 is included) and in the increase of larger polymerised Al species (Alc). The increase of Alc clearly indicates that the composite coagulants components size increases after the addition of APE, thus explaining the increase of turbidity. Furthermore, the relative impact of APE is becoming more intensive when decreasing the Al/APE ratio. Regarding the pH value of composite coagulants, from Table 2 it can be seen that in all cases it is higher than the pH value of the initial PACl 2 solution. The neutralisation of several positively charged sites of Al species, due to their interaction with APE molecules is the possible explanation for the observed pH increase in the composite coagulants. Conductivity is the parameter that is less affected by the introduction of APE. Particularly, a slight decrease was observed after the addition of APE, comparing to the initial PACl 2 solution. Generally, it was observed that the impact of APE addition on the properties of PACl 2 solution is stronger in the composite coagulants prepared with the co-polymerisation method, denoting that not only the Al/APE ratio affects the properties of the composite coagulants, but the preparation method as well.

disappearance could be an indication of interaction between Al species and polyelectrolyte molecules. Finally, the weak band at 778 cm1 in the PACl spectra, becomes weaker after the polyelectrolyte addition and is slightly shifted at lower wave numbers (i.e. at 774 cm1). Regarding the changes occurring in the spectra of composite coagulants, prepared with the co-polymerisation method (i.e. products denoted as PAAPEC), from Fig. 1b it can be noticed that they are similar as in the case of PACAPEC products (prepared with the composite polymerisation method). Relatively small differences can be noticed in the intensity of the bands and their shift. Summarizing, it can be suggested that the introduction of the anionic polyelectrolyte in PACl composition results in

3.1.2. Composition, chemical bonds and morphology of the composite coagulants Fig. 1 illustrates the FT-IR spectra of PACl 2, composite coagulants and of pure Magnafloc LT-25. In Fig. 1a the spectra of composite coagulants prepared via the composite polymerisation method are presented (i.e. the PACAPE products), whereas in Fig. 1b the spectra of composite coagulants prepared via the co-polymerisation method are presented (i.e. the PAAPEC products). A detailed analysis of PACl spectra can be found in Tzoupanos et al. (2009). Briefly, the bands expected to appear in the IR spectra of PACl are those associated with OH vibrations of water, or bridging hydroxyls and with AleO bond vibrations. Regarding the IR spectra of the polyacrylamide Magnafloc LT-25, details can be found in Samsonova et al. (1975), Murugan et al. (1998) and Moharram et al. (2002). From Fig. 1a it can be observed that the major changes of PACl IR spectra after the addition of polyelectrolyte occur at the region 1200e770 cm1, in which the adsorption bands of oxo-groups or bridges appear. Specifically, the band at 1105 cm1 in PACl spectra gradually degenerates into two distinct bands at 1170 and 1080 cm1. In this region and in the Magnafloc spectra, bands related with the vibrations of CeC bonds appear. The shift of those bands at lower wavelengths in the composite coagulants indicates a possible attenuation of their strength. The alteration of bands in the two initial compounds (PACl and Magnafloc) indicates that both, Al species and Magnafloc molecules, are affected by the combination of two compounds in one reagent. Moreover, at the region 980e890 cm1 the two weak bands appearing at the spectra of PACl gradually disappear. These bands are associated with vibrations of oxo-groups or oxo-bridges and their

Fig. 1 e FT-IR spectra of anionic polyelectrolyte p-DADMAC, PACl 2 and of composite coagulants: (a) composite coagulants prepared with the composite polymerisation method (PACAPE), (b) composite coagulants prepared with the co-polymerisation method (PAAPEC).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 1 4 e3 6 2 6

noticeable alterations of its IR spectra. These alterations could be indicative of the interactions between Al species and Magnafloc LT-25 molecules, such as hydrogen bonding and electrostatic interactions, which result in the formation of new, composite species. It is possible that in these interactions the amino (or amidic) groups of the polyelectrolyte and the eOe or eOHe groups of Al species are involved. Finally, it seems that the different preparation method does not influence at a noticeable degree the nature of the bonds in the composition of composite coagulants. XRD spectroscopy was used for the further investigation of composite coagulants composition. Among the aims was the identification of most active (regarding coagulation) Al species, i.e. the Al13, as its existence is a prerequisite in prepolymerised Al coagulants. Additionally, the possible identification of other existing compounds is desirable as well, such as several other Al species, or even new, composite species formed after the interactions between Al species and polyelectrolyte molecules. Fig. 2 displays the XRD pattern of a PACAPE 2/10 composite coagulant (Fig. 2a), representative for all composite coagulants as all these patterns were quite similar. It should be mentioned that the dried powders from the coagulants were obtained after freeze drying, as this procedure was found to be as the most appropriate method for XRD analysis, according to

a

1500

PACAPE 2/10

Intensity (au)

1000

NaCl 500

0

b

3500

Intensity (au)

3000 2500 2000 1500 1000 500 0

10

20

30

40

50

60

70

80

2θ Fig. 2 e XRD patterns of PACAPE 2/10 sample after freeze drying: (a) without any pre-treatment, (b) after separation of Al13 by sulphate addition.

3619

a previous study (Tzoupanos et al., 2009). Various peaks can be observed, indicating initially a crystalline character of PACAPE 2/10. Specifically, the intense peaks at 27 , 32 , 46 and 57 , 67 and 76 2q are attributed to the presence of NaCl, which is an unavoidable by-product, formed during the neutralisation of the Al solution with the NaOH solution during the preparation of coagulants. It is known, that the presence of Al13 in prepolymerised Al coagulants is characterised by various peaks at low angles (5e15 2q) (Zouboulis and Tzoupanos, 2009). From Fig. 2 it can be seen that in this region instead of well distinguishable peaks, a broad curve appears. Few peaks can be distinguished, but their number and intensities are not sufficient for the identification of Al13. Clearly, there are indications of the existence of Al13, but it seems that the addition of the anionic polyelectrolyte hinders its identification. The interactions of Al species with Magnafloc molecules are thought to be responsible for this behaviour, resulting in a rather amorphous structure. In the case of polyaluminium silicate chloride the identification of Al13 was achieved, therefore it seems that the anionic polyelectrolyte has a stronger effect on the composition and structure of composite coagulants, than the anionic inorganic polysilicates. Possible reason for this is considered the higher molecular weight of polyelectrolyte molecules and the different characteristic groups, which induce hydrogen bonding and enhance bridging effects in a higher extent than the presence of polysilicates, resulting in the formation of larger and amorphous chemical species, especially during freeze drying. Xu et al. (2003) and Shi et al. (2007) suggest the preliminary separation of Al13 through a reaction by the addition of sulphates. This separation technique was applied in the composite coagulants (for the exact procedure see: Tzoupanos et al., 2009) and the determination of Al13 was achieved, as shown in Fig. 2b. The pattern of Fig. 2b is identical with the pattern observed by Shi et al. (2007) for PACl samples, and is corresponding to tetrahedral crystals structure, having one Al13 unit combined with four sulphate units. The reference compound given is the Na[AlO4(OH)24(H2O)].xH2O, establishing the presence of Keggin-Al13 structure. The morphology and surface composition of the dried powders of composite coagulants was also examined by using the SEM technique. Fig. 3 illustrates the SEM images of a dried PACAPE 2/10 sample (Fig. 3aed) and of PAAPEC 2/10 sample (Fig. 3e, f). Several different morphologies were observed, whereas the most predominant were a cubic-like morphology (Fig. 3a) and a separated compact solid surface (Fig. 3c or d). Fig. 3b shows irregular formations, which consisted mainly of Al species and were formed in a relatively low extent. Initially, the cubes were considered to be NaCl crystals, separated from the other constituents, because NaCl was found to be a significant constituent of these products (XRD results). Atomic analysis however, revealed that these cubes consist of the sum of the elements (Na, Al, Si, O, Cl, C); therefore, it can be suggested that they consist of Al polymerised species, of polyelectrolyte molecules and probably, of composite AleAPE species. Regarding the amorphous solid surface, two major configurations were observed. The first one (Fig. 3c) is a rather smooth surface which consists of all the elements except of Al, which was detected only in specific spots on the surface

3620

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 1 4 e3 6 2 6

Fig. 3 e SEM images of composite coagulants, obtained after drying in an oven under constant low temperature (w40 C). (a), (b), (c) and (d) of PACAPE 2/10, (e) and (f) of PAAPEC 2/10.

and in relatively low percentage (i.e. up to 2%), as compared with the other elements. The specific SEM image was obtained by the backscattered technique, which enables the visual discrimination between organic and inorganic phases on the surface of solids, according to the brightness of the image. Particularly, if the surface is dominated by the presence of organic material, the brightness is low, whereas the presence of inorganic materials increases it. Indeed, atomic analysis showed that the darkest regions on the surface consist of O and C, with carbon to be the major element (over 80%) and therefore, it was suggested that these areas consist from polyelectrolyte molecules. The luminous-brightest regions consist of all elements (C in a relatively low percentage, i.e. lower than 8%) except of Al and it was concluded that the surface of Fig. 3c consists of un-reacted polyelectrolyte molecules and of NaCl. On the other side, the surface displayed in Fig. 3d represents the regions which consist of

polyelectrolyte molecules and Al species (“composite regions”). From the variations of the brightness it can be noticed that the organic and inorganic phases overlap each other and the distribution of elements is as following: the darkest regions consist mainly of O, Al, C and Cl, the brightest regions consist mainly of O, Na and Cl and the intermediate regions consist of all elements, except of Na. The existence of those regions can be regarded as an additional evidence of the interactions occurred between Al species and polyelectrolyte molecules, which probably lead to the formation of more complicated, composite species. From the visual comparison of SEM images of PACAPE 2/10 and those of PAAPEC 2/10, it can be seen that the preparation method noticeably affects the morphology of the dried products. Particularly, in the case of the composite coagulants prepared with the co-polymerisation method (products PAAPEC), it was observed that instead of the well distinguishable

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 1 4 e3 6 2 6

cubes and the relatively smooth separate surfaces, amorphous aggregates were formed with extruded rods on their surface (Fig. 3e). Similar formations were not observed in the case of composite coagulants, prepared with the composite polymerisation method (i.e. PACAPE products). After atomic analysis, it was found that the rods consist of all the analysed elements (O, Al, Na, C and Cl), thus it was assumed that Al species, polyelectrolyte molecules and composite species are co-present. Fig. 3f displays the same surface (after zooming) with the backscattered technique. In the whole surface (apart from the rods) O and Cl were detected. In the brightest regions, only Na was additionally detected, whereas in the darkest regions Al and C were also detected.

3.2. Comparison of the prepared coagulants in water treatment Fig. 4 displays the results of coagulation experiments, regarding the treatment of model water sample (simulating surface water) with all laboratory prepared coagulants. The concentration of coagulants ranged between 0.5 and 6 mg/L and the experiments were conducted at the initial pH (7.65) of water sample. In the case of PACl the coagulant concentration

3621

is expressed as mg Al/L and in the case of composite coagulants, as mg (Al þ Magnafloc LT-25)/L. Moreover, in the case of PACl Magnafloc LT-25 was used at concentration equal to 1/10 of the concentration of PACl, determined as the optimum ratio through preliminary experiments (data not shown). Regarding turbidity removal, from Fig. 4a it can be observed that all composite coagulants exhibit similar behaviour, except of PACAPE 2/10, which was found as the most efficient between them. When using PACAPE 2/10 the concentration needed to reduce the final turbidity under 1 NTU (according to the respective legislation limit, EU Directive 98/83/EC) is about 2 mg/ L, whereas with the other composite coagulants the respective concentration is higher than 3 mg/L. The second more efficient composite coagulant was PAAPEC 2/15 and the least efficient composite coagulants are those with the lower polyelectrolyte content (i.e. Al/APE ¼ 20). In comparison with PACl 2, PACAPE 2/ 10 is more efficient as well. Fig. 5 illustrates the % turbidity removal rates of PACAPE 2/10 and PACl 2. PACAPE 2/10 is more efficient than PACl especially for lower coagulant dosages (i.e. 0.5e3 mg/L), whereas by increasing the dosages the efficiencies for both coagulants tend to equalize, reaching up to 98% removal. With PACl, the concentration needed for decreasing the turbidity under 1 NTU is slightly higher than 2 mg/L.

Fig. 4 e Comparative coagulation experiments of all laboratory prepared coagulants. Experimental conditions: concentration of coagulants 0.5e6 mg(Al D APE)/L, pH of water sample to be treated 7.65; (a) residual turbidity (initial turbidity 16.5 NTU), (b) UV absorbance at 254 nm (initial absorbance 0.125), (c) residual aluminum concentration, (d) z-potential measurements.

3622

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 1 4 e3 6 2 6

100 90 80

Re (%)

70 60 50 40 30

PACl + Magn. LT 20 (Turb. Re%) PACAPE 2/10 (Turb. Re%) PACl + Magn. LT 20 (UV Abs. Re%) PAAPEC 2/15 (UV Abs. Re%)

20 10 0 0

1

2

3

4

5

6

Coagulant dose (mg/L) Fig. 5 e Turbidity and UV absorbance at 254 nm % removal rates.

Moreover, it seems that the composite coagulants prepared with the composite polymerisation method (i.e. PACAPE products) are more efficient than those prepared with the co-polymerisation method (i.e. PAAPEC products). UV254 absorbance removal (Fig. 4b) is more efficient by using composite coagulants PACAPE 2/10 and PAAPEC 2/15, although the latter is slightly more efficient, especially for dosages greater than 2.5 mg/L. The least efficient composite coagulants are those with the lower polyelectrolyte content (i.e. Al/APE ¼ 20). PACl 2 at low coagulant dosages (i.e. 0.5e1.5 mg/L) has similar behaviour with PACAPE 2/10 and PAAPEC 2/15, but with increasing dosages it becomes the least efficient between all coagulants. The differences between the two most efficient composite coagulants and PACl are easier to distinguish in Fig. 5. With PAAPEC 2/15 the absorbance removal rate reaches up to 96% for the highest coagulant concentration (i.e. 6 mg/L). The respective removal rate with PACAPE 2/10 was 93.6%, whereas with PACl the highest removal rate achieved was 86.4%. Residual aluminium concentration is very important parameter from health perspective and should be carefully considered, when an aluminium coagulant is applied in water treatment. From Fig. 4c it can be seen that the Al concentration remaining in the sample after treatment varies significantly, according to the initially applied concentration for all examined coagulants. The lowest residual Al concentration was achieved with the addition of 2 mg/L PACAPE 2/10 (corresponding to 140 mg Al/L). This specific coagulant seems to be the most efficient composite coagulant, as for the majority of applied concentrations (i.e. in 6 of the total 9), residual Al concentration remains under the respective legislation limit of 200 mg Al/L (EU Directive 98/83/EC). The highest residual Al concentration with the specific coagulant is 235 mg/L for the initial coagulant concentration of 3 mg/L. The second most efficient composite coagulant is PAAPEC 2/15, as in 4 examined concentrations the residual Al concentration remains under the respective limit. The lowest residual Al concentration is

182 mg/L (for 4 mg/L of initial coagulant concentration) and the highest is 232 mg/L (for 3 mg/L of coagulant). Regarding PACl 2, only for 2 concentrations the residual Al concentration was lower than 200 mg/L. The lowest Al concentration was 169 mg/L (for 1.5 mg/L of PACl 2) and the highest 266 mg/L (for 6 mg/L of PACl 2). It should be mentioned that the results presented correspond to the net effect of coagulation/flocculation processes. In a full-scale treatment plant, the residual Al concentration is expected to be further reduced after optimisation of the “ideal coagulation conditions” and after filtration. Fig. 4d illustrates the z-potential measurements, which can serve as an indirect validation of coagulants components impact on the surface charge of colloids. The differences between the coagulants are quite small, but it is clear that PACl 2 exhibits the biggest impact on the colloids surface charge. Therefore, it can be suggested that the addition of the anionic polyelectrolyte in PACl’s composition reduces at a certain degree the charge neutralisation ability of the composite coagulants. Between the composite coagulants, those with the lowest content of polyelectrolyte (i.e. with Al/APE ¼ 20) exhibit the strongest impact on the colloids surface charge, and the composite coagulant prepared with the composite polymerisation method (PACAPE 2/20) is the most efficient. With PACl 2 charge reversal occurs at coagulant dose 5 mg/L, whereas with the composite coagulants at least 6 mg/L are needed. With the composite coagulants PAAPEC 2/ 15 and PAAPEC 2/10 no charge reversal occurs and it seems that the charge neutralisation ability of the composite coagulants prepared with the co-polymerisation method is inferior, when compared to the respective ability of the composite coagulants prepared with the composite polymerisation method. The weakest effect on the colloids surface charge was exhibited by the coagulant PAAPEC 2/10. In another study (Zouboulis and Tzoupanos, 2009), where the properties and coagulation behaviour of polyaluminium silicate chloride were studied, from the z-potential measurements it was found that the addition of polysilicates resulted in the deterioration of charge neutralisation ability of the initial PACl, but in higher extent than the Magnafloc LT-25. Considering that Magnafloc LT-25 is an anionic polyelectrolyte, the fact that it exhibits a weaker effect on the charge neutralisation ability of the initial PACl than the polysilicates, indicates that probably its molecules participate in the charge neutralisation process. Through adsorption on the surface of the colloids, initially in one binding site (which progressively increase), and the subsequent inter-particle bridging, the polyelectrolyte molecules can also contribute to the surface charge neutralisation of the colloids. Despite the decrease of Al13 content in the composite coagulants (Section 3.1.1) and the weaker charge neutralisation ability, as compared to PACl, the composite coagulants have been proved more efficient in water treatment. One of the main advantages of composite coagulants is the lower residual aluminium concentration that remains in the treated sample. Moreover, the treatment procedure is conducted in one step, i.e. there is no need for the subsequent addition of polyelectrolyte as flocculant aid, as it is already included in the structure of the composite reagent. A further advantage is the additional cost saving, when composite coagulants will be used. Table 3 displays the comparison between the amounts

3623

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 1 4 e3 6 2 6

Table 3 e Comparison of Al and Magnafloc LT-25 amounts used for coagulation in the treatment with composite coagulants, with the respective amounts used, when PACl 2 and Magnafloc are applied as separate reagents (basis: 2 mg/L of coagulants, corresponding to 100%). PACl and Magnafloc LT-25 as separated reagents

PACAPE 2/10 (2 mg/L)

Magnafloc content

Al content

Magnafloc content

Al content

Magnafloc content

2 100

0.2 100

1.82 90.9

0.18 90

1.87 93.8

0.13 62.5

of Al and polyelectrolyte used in the case of the most efficient composite coagulants with the respective amounts when PACl 2 and Magnafloc are applied as separated reagents. In the case of PACAPE 2/10, which was prepared with the composite polymerisation method, more efficient treatment was achieved by using 10% less amount of polyelectrolyte and 9.1% less amount of Al, as compared to the treatment with PACl and Magnafloc LT-25, applied as separated reagents. In the case of PAAPEC 2/15, more efficient treatment was achieved by using 37.5% less amount of polyelectrolyte and 6.25% less amount of Al, as compared to the respective treatment with PACl and Magnafloc LT-25, applied as separated reagents. However, in order to have a complete picture about the possible cost benefits, the excess preparation cost of the composite coagulants should be also considered. It should be mentioned that after the addition of all aluminium coagulants in the conducted experiments, a slight decrease of pH values of the samples was observed (data not shown). The acidic character of Al3þ cation and its hydrolysis products (possessing higher positive charge) are responsible for this decrease. The pre-polymerised coagulants contain already a percentage of these species, resistant to further hydrolysis and show a relatively lower (but not significant) effect to the pH value of samples in comparison with the addition of simple Al salts, such as alum. In the case of composite coagulants, the incorporation of polyelectrolyte molecules into the structure of PACl results in a decrease of coagulants charge density, as shown by the respective z-potential measurements. Furthermore, the possible conjunction of aluminium species with polyelectrolyte in molecular bridging enhances their resistance of them to hydrolysis. As a consequence, the composite coagulants have shown a weaker influence on the pH values of samples, than the PACl with Magnafloc LT-25, applied as separated reagents. Finally, it should be mentioned also that not all composite coagulants are more efficient than PACl. According to the sample to be treated and to the measured parameter, composite coagulants with different properties, such as the polyelectrolyte content and the preparation method, can be noticed as most efficient.

3.3.

PAAPEC 2/15 (2 mg/L)

Al content

Kinetics of coagulation

Selected PDA experiments were conducted in order to compare the extent of aggregation and consequently, the floc size generated with the application of composite coagulants, as well as of PACl. Fig. 6 illustrates the floc growth of contaminated tap water suspension, containing 5 mg/L clay

and 5 mg/L humic acid (initial turbidity 9.9 NTU, UV absorbance at 254 nm 0.116, pH 7.6) with the addition of PACl 2, PAAPEC 2/20, PACAPE 2/15, PAAPEC 2/15 and PACAPE 2/10. The initial 120 s represent the fast mixing period (velocity of mixing paddle 160 rpm, or 200 s1, expressed as velocity gradient), where the destabilization of colloids occurs and the floc growth is not remarkable (lag phase). The aggregation begins during the subsequent slow mixing period (45 rpm) and the respective Ratio values increase. Furthermore, Table 4 displays the calculated values for three different parameters after processing the PDA data for each of the examined coagulants, according to Hopkins and Ducoste (2003), i.e. the slope, which is calculated after the construction of a best-fit line for the linear growth region and is indicative of the flocs growth rate; it is calculated according to the equation: Dratio Dtime

slope ¼

The second parameter is a time-weighted average steadystate ratio value, which represents the average extent of aggregation during the steady-state period and is calculated according to the following equation (data collected from the steady-state region): PN

i¼1 ðratioi  timei Þ PN i¼1 timei

ratio ¼

Finally, the time-weighted average steady-state ratio variance was calculated, which is an indicator of flocs break-up 3.5 3.0 2.5

RATIO

mg/L Comparison (%)

Composite coagulants

2.0 1.5

PACl 2 PAAPEC 2/20 PACAPE 2/15 PAAPEC 2/15 PACAPE 2/10

1.0 0.5 0.0

0

20 0

40 0

600

800

100 0

120 0

1400

Time (s) Fig. 6 e Study of coagulation kinetics by PDA. Experimental conditions: concentration of coagulants 2 mg/L, pH 7.6.

3624

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 1 4 e3 6 2 6

Table 4 e Calculated coagulation kinetics parameters from the PDA data. Coagulant

Slope

ratio

Variance

PACl 2 PAAPEC PACAPE PAAPEC PACAPE

0.067 0.070 0.062 0.088 0.069

1.435 1.687 1.536 1.942 1.769

0.0045 0.0053 0.0060 0.0050 0.0042

2/20 2/15 2/15 2/10

and is related with the flocs size distribution. Particularly, it is suggested that the relatively low values of variance indicate tighter floc size distribution (i.e. more homogeneous, dense and less porous floc structure). The equation used is the following (data collected from the steady-state region): PN h variance ¼

i¼1

2

ðratioi  averageratioÞ  timei PN i¼1 timei

i

From Fig. 6 it can be noticed that the duration of lag phase (i.e. the initial time interval in which coagulation occurs and the Ratio values remain relatively constant and close to zero) varies noticeably according to the specific properties of added coagulants. Particularly, it is slightly shorter for PACl and increases with the increase of polyelectrolyte content in the composite coagulants. Therefore, it can be suggested that the introduction of Magnafloc LT-25 in the composite coagulants results in small retardation of the beginning of flocculation process. This behaviour can be attributed to the deterioration of charge neutralisation ability of the composite coagulants (as compared to PACl), resulting in slower coagulation process. However, after the destabilisation is occurring and the aggregation is taking place (i.e. approximating the linear growth region), from Table 4 it can be seen that the slope of the best-fit line is higher in the case of the composite coagulants, as compared to PACl 2. This is indicative of quicker flocs formation with the composite coagulants, due to the presence of polyelectrolyte molecules, which induce the bridging effect between destabilised particles. The highest slope was achieved with the coagulant PAAPEC 2/15. Observing the extent of aggregation (i.e. the extent of Ratio values increment in the steady-state region), it can be noticed that it is higher for the composite coagulants, as Ratio takes higher values when composite coagulants are used. Despite the fact that the lag phase lasts slightly longer period for the composite coagulants, the flocs growth is quicker and the final size of flocs generated is bigger, than with PACl and therefore, the composite coagulants are more efficient in water treatment. The shortest lag phase between the composite coagulants exhibits the sample PAAPEC 2/15, which is similar to the respective duration of lag phase of PACl. The highest Ratio values are achieved (steady-state period) with the composite coagulant PAAPEC 2/15. The calculated ratio values in Table 4 confirm with this observation. PAAPEC 2/15 and PACAPE 2/10, the most efficient coagulants in water treatment, exhibit the highest ratio values between all examined coagulants. Regarding the impact of polyelectrolyte content in the extent of flocculation, from Fig. 6 it can be observed that independently of the preparation method, the increase of polyelectrolyte content (i.e. the decrease of Al/APE ratio)

increases the size of generated flocs (or increases the ratio value). Additionally, the Ratio (or ratio) with PACAPE 2/15 reaches higher values than with PAAPEC 2/15, and therefore it can be suggested that the composite coagulants prepared with the co-polymerisation method produce flocs of larger size. Despite the decrease of Alb (or Al13) content and the deterioration of charge neutralisation ability of composite coagulants, the bigger flocs generated with the composite coagulants, as compared to PACl, are responsible for their better performance in water treatment. The increased size of composite coagulants components (as shown by ferron study) due to interactions between Al species and polyelectrolyte molecules (as shown by FT-IR study) is responsible for the generation of bigger flocs. Additionally, during the coagulation process the positively charged polymerised Al species can cause excess of positive charge on specific sites on the colloids surface (electrostatic patch) and the negatively charged polyelectrolyte molecules can interact with those sites (bridging effect), increasing the size of generated flocs and at the same time, preventing the colloids from re-stabilisation and the flocs breakage. It is evident that the composite coagulants own their higher efficiency to their improved flocculation abilities. The bridge formation mechanism dominates over the charge neutralisation mechanism, whereas in the case of PACl the main mechanism is charge neutralisation. This is the reason for the higher variance values in the case of the majority of composite coagulants, as compared to PACl (Table 4), which imply a wider range of floc sizes generated, when composite coagulants are used. However, observing the impact of polyelectrolyte content on the variance values of the composite coagulants, it can be seen that in both cases (i.e. coagulants prepared with the co-polymerisation or the composite polymerisation method) the highest variety of floc sizes (i.e. the highest variance values) exhibit the composite coagulants with the relatively lowest polyelectrolyte content (i.e. Al/APE ¼ 20 or 15, in the case of PAAPEC or PACAPE products respectively). Increasing the content of APE in the composite coagulants, narrows the floc size distribution range and it can be observed that PACAPE 2/10 (which represents the most efficient coagulant overall) exhibits slightly lower variance value, even than PACl. Flocs size distribution seems to be significantly affected by the polyelectrolyte content. From the ferron study (Section 3.1.1), it was found that the effective Al species content (i.e. Alb) decreases with the increase of polyelectrolyte content. These species are responsible for the charge neutralisation and for the initial formation and growth of flocs, and their limitation probably results in a restriction of variety of the flocs sizes. Additionally, from the variation values it can be seen that the coagulants prepared with the co-polymerisation method (i.e. the PAAPEC products) generate more dense flocs and result in relatively narrow size distribution. The composite coagulants prepared with the composite polymerisation method result in a more heterogeneous range of floc sizes. The initial composition of coagulants could be responsible for this behaviour, i.e. the Al species distribution, as it was found in the case of PAAPEC products, where the Alb content is lower than in the case of PACAPE products. The statement that larger flocs ensure the better treatment should not be generalised. However, according to the findings

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 1 4 e3 6 2 6

of this study, i.e. that the charge neutralization ability of the composite coagulants is lower than the respective ability of PACl and that the flocs generated with the composite coagulants are bigger than with PACl, it is justified (at least for this case) to attribute the better efficiency of the composite coagulants to the bigger size of flocs. Certainly, other parameters such as the flocs strength play also an important role, which has to be separately examined.

4.

Conclusions

The main conclusions that can be withdrawn from this study aimed to combine an inorganic pre-polymerised coagulant (PACl) and an anionic polyelectrolyte in one unique reagent, are the following:  The incorporation of the anionic polyelectrolyte into PACl’s structure noticeably affects its initial properties, i.e. turbidity, Al species distribution, pH and conductivity.  Between Al species and polyelectrolyte molecules interactions are taking place. Hydrogen bonding and electrostatic interactions are thought to be the primary types of those interactions, which probably result in the formation of new, “composite” species. It is possible that in these interactions the amino (or amidic) groups of the polyelectrolyte and the eOe or eOHe groups of Al species are involved.  Despite the decrease of Al13 content in the composite coagulants and the weaker charge neutralisation ability, as compared to PACl, the composite coagulants have been proved more efficient in water treatment. The main advantage of composite coagulants is the lower residual aluminium concentration that remains in the treated sample.  More efficient treatment was achieved (in terms of turbidity and organic matter removal) by the application of lower amounts of aluminium and polyelectrolyte, as compared to the treatment with PACl and Magnafloc LT-25 when applied as separated reagents.  Additional cost benefits include the avoidance of specific equipment for handling the polyelectrolyte (e.g. dissolution system, pumping system), as there is no need for using it as a flocculant aid.  The better performance of composite coagulants is due to the enhanced flocculation process with those reagents (quicker formation of flocs with bigger size than with PACl). The preparation of composite coagulants with the combination of PACl and Magnafloc LT-25 was successfully accomplished. The development of these new reagents seems to be promising, due to their increased efficiency and costeffectiveness, compared to the IPFs. However, a detailed preparation cost analysis would clarify the exact extent of cost saving, considering also the extra cost for the preparation of composite reagents. Moreover, the specific coagulants were found to be stable for about 2.5 months at room temperature storage. Comparing to PACl, which is stable for more than a year, it seems that their stability should be improved. Therefore, the further research based on the results of this study, seems to be promising.

3625

Acknowledgements Thanks are due to the Greek Ministry of Development (General Secretariat for Research and Technology) which supported this research through the PENED program (25%) and to the European Union, which co-founded this program (75%). This research is part of the Ph.D. Thesis of N.D. Tzoupanos.

references

Bolto, B.A., 1995. Soluble polymers in water treatment. Progress in Polymer Science 20 (6), 987e1041. Bratby, J., 2006. Coagulation and Flocculation in Water and Wastewater Treatment, second ed. IWA Publishing, London. Clesceri, L., Greenberg, A., Trussell, R., 1989. Standard Methods for the Examination of Water and Wastewater, seventeenth ed. APHA-AWWA-WEF, Washington DC. Coagulation, mixing and flocculation. In: Crittenden, J.C., Trussel, R.R., Hand, D.W., Howe, K.J., Tchobanoglous, G. (Eds.), Water Treatment: Principles and Design, second ed. John Wiley & Sons, New Jersey, pp. 643e779. Gao, B.Y., Wang, B.J., Yue, Q.Y., 2005. The chemical species distribution of aluminum in composite flocculants prepared from polyaluminium chloride (PAC) and polydimethyldiallylamonium chloride (PDMDAAC). Acta Hydrochimicha et Hydrobiologica 33 (4), 365e371. Hopkins, D., Ducoste, J., 2003. Characterising flocculation under heterogeneous turbulence. Journal of Environmental Technology and Management 1, 464e471. Kan, C., Huang, C., Pan, J.R., 2002. Time requirement for rapidmixing in coagulation. Colloids and Surfaces A: Physicochemical and Engineering Aspects 203 (1e3), 1e9. Moharram, M.A., Rabie, S.M., El-Gendy, H.M., 2002. Infrared spectra of g-irradiated poly(acrylic acid)epolyacrylamide complex. Journal of Applied Polymer Science 85 (8), 1619e1623. Mortimer, D.A., 1991. Synthetic polyelectrolytes e A review. Polymer International 25 (1), 29e41. Murugan, R., Mohan, S., Bigotto, A., 1998. FTIR and polarised raman spectra of acrylamide and polyacrylamide. Journal of the Korean Physical Society 32 (4), 505e512. Parker, D.R., Bertsch, P.M., 1992. Formation of the ‘Al13’ tridecameric polycation under diverse synthesis conditions. Environmental Science and Technology 26 (5), 914e921. Samsonova, N.S., II’chenko, L.G., Gol’dman, M.M., Ni, L.P., 1975. IR spectroscopic study of the adsorption of polyacrylamide on hematite. Translated from Zhurnal Prikladnoi Spektroskopii 23 (1), 117e121. Plenum Publishing Corporation. Shi, B., Li, G., Wang, D., Tang, H., 2007. Separation of Al13 from polyaluminum chloride by sulfate precipitation and nitrate metathesis. Separation and Purification Technology 54 (1), 88e95. Sinha, S., Yoon, Y., Amy, G., Yoon, J., 2004. Determining the effectiveness of conventional and alternative coagulants through effective characterisation schemes. Chemosphere 57 (9), 1115e1122. Tang, H., Shi, B., 2002. The characteristics of composite flocculants synthesized with inorganic polyaluminum and organic polymers. In: Hahn, H., Hoffmann, E., Odegaard, H. (Eds.), Chemical Water and Wastewater Treatment VII, Proc.. of the 10th Gothenburg Symposium 2002. IWA Publishing, Gothenburg, Sweden, pp. 17e28. Tzoupanos, N.D., Zouboulis, A.I., 2008. Coagulation-flocculation processes in water/wastewater treatment: the application of new generation of chemical reagents. In: 4rd IASME/WSEAS Inter. Conf. 2008, Rhodes (Rodos) Island, Greece.

3626

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 1 4 e3 6 2 6

Tzoupanos, N.D., Zouboulis, A.I., 2010. Novel inorganic-organic composite coagulants based on aluminium. Desalination and Water Treatment 13 (1e3), 340e347. Tzoupanos, N.D., Zouboulis, A.I., Tsoleridis, C.A., 2009. A systematic study for the characterisation of a novel coagulant (polyaluminium silicate chloride). Colloids and Surfaces A: Physicochemical and Engineering Aspects 342 (1e3), 30e39. Xu, Y., Wang, D., Liu, H., Yiqiang, L., Tang, H., 2003. Optimization of the separation and purification of Al13. Colloids and Surfaces A: Physicochemical and Engineering Aspects 231 (1e3), 1e9.

Yukselen, M.A., Gregory, J., 2004. The reversibility of floc breakage. International Journal of Mineral Processing 73 (2e4), 251e259. Zhou, W., Gao, B., Yue, Q., Liu, L., Wang, Y., 2006. Al-Ferron kinetics and quantitative calculation of Al(III) species in polyaluminium chloride coagulants. Colloids and Surfaces A: Physicochemical and Engineering Aspects 278 (1e3), 235e240. Zouboulis, A.I., Tzoupanos, N.D., 2009. Polyaluminium silicate chloride e A systematic study for the preparation and application of an efficient coagulant for water or wastewater treatment. Journal of Hazardous Materials 162 (2e3), 1379e1389.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 2 7 e3 6 3 7

Available at www.sciencedirect.com

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

Modeling the elution of organic chemicals from a melting homogeneous snow pack Torsten Meyer*, Frank Wania Department of Physical and Environmental Sciences, University of Toronto Scarborough, 1265 Military Trail, Toronto, Ontario, Canada M1C 1A4

article info

abstract

Article history:

Organic chemicals are often released in peak concentrations from melting snow packs. A

Received 19 August 2010

simple, mechanistic snowmelt model was developed to simulate and predict the elution of

Received in revised form

organic substances from melting, homogeneous snow, as influenced by chemical proper-

31 March 2011

ties and snow pack characteristics. The model calculates stepwise the chemical transport

Accepted 6 April 2011

along with the melt water flow in a multi-layered snow pack, based on chemical equilib-

Available online 14 April 2011

rium partitioning between the individual bulk snow phases. The model succeeds in reproducing the elution behavior of several organic contaminants observed in previously

Keywords:

conducted cold room experiments. The model aided in identifying four different types of

Snowmelt

enrichment of organic substances during snowmelt. Water soluble substances experience

Model

peak releases early during a melt period (type 1), whereas chemicals that strongly sorb to

Organic chemicals

particulate matter (PM) or snow grain surfaces elute at the end of melting (type 2).

Snow pack

Substances that are somewhat water soluble and at the same time have a high affinity for snow grain surfaces may exhibit increasing concentrations in the melt water (type 3). Finally, elution sequences involving peak loads both at the beginning and the end of melting are simulated for chemicals that are partially dissolved in the aqueous melt water phase and partially sorbed to PM (type 4). The extent of type 1 enrichment mainly depends on the snow depth, whereby deeper snow generates more pronounced concentration peaks. PM influences the elution behavior of organic chemicals strongly because of the very large natural variability in the type and amount of particles present in snow. Urban and road-side snow rich in PM can generate type 2 concentration peaks at the end of the melt period for even relatively water soluble substances. From a clean, melting snow pack typical for remote regions, even fairly hydrophobic chemicals can be released in type 1 mode while being almost completely dissolved in the aqueous melt water phase. The model provides a mechanistic understanding of the processes that lead to chemical peak releases during snowmelt. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Winter-long accumulation of organic pollutants in seasonal snow packs is often followed by a sudden release during the spring melt period. Field studies (Simmleit et al., 1986; Semkin

et al., 1996; Loseto et al., 2004; Lafrenie`re et al., 2006; Bizzotto et al., 2009) and experimental investigations (Scho¨ndorf and Herrmann, 1987; Meyer et al., 2009a,b) indicate a differential release of organic contaminants from melting snow, reflecting their partitioning within the bulk snow. Water soluble chemicals

* Corresponding author. Tel.: þ1 416 287 7506; fax: þ1 416 287 7279. E-mail address: [email protected] (T. Meyer). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.011

3628

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 2 7 e3 6 3 7

tend to be washed out at an early stage of the melt period whereas the bulk of hydrophobic substances is usually released at the end of melting. The former are assumed to accumulate near snow grain surfaces from which they can be taken up by the first downward percolating melt water. Hydrophobic chemicals are likely associated with particulate matter (PM) that accumulates at the snow pack surface. Filter-like retention processes prior and during melting as a result of particle coagulation and snow compaction limit particle transport through melting snow. Both forms of chemical concentration at the beginning and the end of the melt period have previously been referred to as type 1 and type 2 chemical enrichment, respectively (Meyer and Wania, 2008). Laboratory experiments to investigate the snowmelt behavior of organic chemicals are difficult and time consuming (Meyer et al., 2009a,b) and cannot cover the entire range of naturally occurring snowmelt scenarios. Snow characteristics such as snow depth, type and content of PM, and particle transport within melting snow are highly variable in natural snow packs. At the same time they likely have a strong influence on the fate of organic chemicals in melting snow. Those parameters can only be investigated comprehensively using a modeling approach. Wania (1997) calculated the equilibrium partitioning and organic chemical loss in a one-layered, melting snow pack. This model successfully reproduced the observed release of a-HCH, g-HCH, DDT, and PCBs from a melting Arctic snow pack (Semkin et al., 1996). Further development of that model led to the integration of snow into a dynamic multi-media environmental contaminant model (Daly and Wania, 2004). Organic chemical elution from melting snow was also described by Stocker et al. (2007) who included snow and ice into a global multi-media environmental fate model, while including processes such as deposition with snow, degradation in snow, and snow-atmosphere exchange. All these models assume that organic chemical release from a melting snow pack occurs by equilibrium partitioning into the melt water phase. Other modeling studies investigated the release of ionic solutes from melting snow (e.g. Harrington and Bales, 1998; Lee et al., 2008). Those highly water soluble substances are usually released in pulses early during melting. Differences in elution between organic chemicals and ionic solutes from melting snow are discussed in Meyer and Wania (2008). Here, we present a simple, mechanistic model that describes the partitioning and transport of organic chemicals in a multilayered, homogeneous snow pack. Expanding on previous modeling approaches, this model considers the influence of snow depth, the transport of particles along with the melt water, and the variability in PM type and content occurring in natural snow packs. Further, we were able to validate and calibrate our model using an unprecedented set of data from recent laboratory experiments on organic contaminant behavior in melting snow (Meyer et al., 2009a,b). The two types of chemical enrichment introduced in Meyer and Wania (2008) are complemented by two more types, leading to a fuller categorization of possible chemical elution sequences in melting snow. The applicability of the model to various types of natural snow packs and melt conditions is discussed. The methods section (Section 2) describes the structure of the model and its calibration and validation. Section 3 is devoted to the simulation

of the chemical elution observed in the experiments. Section 4 introduces the different types of chemical enrichment behavior, based on the chemical’s partitioning within melting snow. Finally, in Section 4 we discuss the influence of snow pack characteristics on the different types of elution.

2.

Methods

2.1.

Input parameters

The model calculates chemical partitioning between individual bulk snow phases and transport of contaminants with the melt water as a sequence of chemical equilibrium calculations. An organic chemical in a melting snow pack is assumed to be distributed mainly between four phases: aqueous melt water, air pore space, snow grain surface, and PM. Incorporation into the ice lattice of the snow grains is neglected (see Meyer and Wania, 2008), as is sorption to the air-water interface, because the extent of this surface is likely much smaller than that of the snow grain surface (Colbeck, 1979). Also, it is not clear whether chemical sorbed to the air-water interface is subject to transport or retention during melting. A chemical’s distribution between the bulk snow phases is expressed by partition coefficients that are either calculated from poly-parameter linear free energy relationships (LFERs) or taken from the literature. Different mechanisms contribute to the sorption of organic chemicals to PM, such as adsorption to soot or absorption into liquid-like organic matter (Lohmann and Lammel, 2004). Circumventing the need to specify a mechanism of sorption, the partitioning between PM and water is expressed with the generic solid sorption coefficient KD which accounts for the varying types and strengths of sorption. Snow-pore space partitioning is described based on Roth et al. (2004). Detail on the determination of partition coefficients is given in the Supplementary Data. Input parameters describing the snow pack include melt water content, specific snow surface area (SSA), snow density, volume fraction of PM, snow permeability to particles, and snow depth. The melt water content was set to 6% per bulk snow volume for all calculations. This parameter has a small impact on the calculated chemical enrichment, mainly because melt water content in a draining snow pack falls within a fairly small range of approximately 3e6%, and only sporadically reaches 10% (Waldner et al., 2004; Meyer et al., 2009b). The model assumes complete snow grain coverage with melt water, homogeneity of the snow in terms of physical properties, and uniform flow at any stage of melting. These assumptions are likely not justified in natural snow packs (Marsh, 1999), which are almost always highly unsaturated with melt water (Colbeck, 1979), often stratified and characterized by preferential flow. Such flow leads to temporary bypassing of parts of the bulk snow until the background wetting front arrives and generates a more homogeneous flow pattern (Marsh and Woo, 1984). Preferential flow in isothermal snow has a negligible impact on the elution behavior of hydrophobic substances. However, it tends to diminish chemical enrichment of type 1 (Meyer et al., 2009b). In a nonisothermal, melting snow pack, prevalent in colder regions, substantial re-freezing likely has a stronger impact on the

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 2 7 e3 6 3 7

elution behavior of organic substances. Snow density, permeability to melt water, and SSA are assumed to be constant over the course of a melt period. Whereas variations of snow density are comparatively limited (Domine´ et al., 2007), the other two parameters can substantially change during melting. However, for the purpose of this model it was deemed sufficient to apply constant values for both parameters. Permeability to melt water has a relatively small influence on all types of chemical enrichment. The uncertainty of the absolute SSA is likely much larger than the impact that the assumption of a constant SSA has on a chemical’s elution behavior. The influence of different SSAs is discussed in Section 5.4. Although the model is based on a linear system the assumption that the physical snow parameters are invariant in time introduces some uncertainty.

2.2.

Chemical fate processes in melting snow

During each calculation step the current chemical partitioning is determined. Only the chemical fraction that is dissolved in the aqueous phase or sorbed to PM can be subject to downward transport with melt water. The particle-associated fraction either gradually accumulates on top of the snow pack as surface deposit, or percolates with melt water (Fig. 1). The extent to which the latter occurs is parameterized with the percentage snow permeability to particles, which is different from the permeability to melt water (see Section 2.1). The former refers to the particulate chemical fraction that does not accumulate at the snow pack surface. The model applies a constant snow permeability to particles which is likely not justified in natural snow. Often during melting the snow density increases and particles coagulate to a larger extent, which leads to a decrease in permeability. Again, the application of a constant value in the model is justified because the overall parameter uncertainty dwarfs the impact of assuming constancy. A fraction of the chemicals associated with snow grain surfaces also accumulates at the top of the snow pack as surface deposit. This deposit consisting of chemicals with high affinity to PM and snow grain surfaces always constitutes the last sample in the elution sequence. The fraction residing in the gaseous pore space can evaporate from the snow pack. However, snow-atmosphere exchange processes are not considered because parameterization of wind pumping requires a set of relatively uncertain empirical assumptions that would compromise the mechanistic character of the model (Meyer and Wania, 2010). Transport of melt water from one snow layer to the next subjacent layer occurs at a constant volumetric rate (Fig. 1). This flux is enabled only when the layer that releases the melt water contains water at 6% by volume. The melt period is initiated by setting the water content of the surface layer to 6%. The layer beneath the surface layer is gradually filling up with melt water and as soon as its water content also reaches 6%, water and chemical transfer to the next layer commences. Finally, melt water will exit the snow pack at its base containing the first chemical released from the snow pack. In this way chemical concentrations within the downward moving melt water front increase with snow pack depth. Each calculation step decreases the thickness of the surface layer at a constant rate. Upon reaching a defined minimum thickness this layer is merged with the subjacent layer to form the new surface layer. Accordingly, the overall snow depth decreases with each step.

3629

In temperate regions seasonal snow packs are often exposed to small but constant bottom melt over several weeks prior to the spring snowmelt period (Meyer and Wania, 2008). Simulating such scenarios, melting is assumed to occur in both the surface and bottom layers. The bottom layer then also contains 6% water and the water release rate from this layer is arbitrarily set to one tenth of that generated due to surface melt. Previous studies quantifying the extent of bottom melt are scarce. Whereas Waldner et al. (2000) and Lundberg et al. (2004) report bottom snowmelt rates of 0.13 mm/day snow water equivalent (SWE) in the Swiss Alps, and 0.3e1.0 mm/day SWE on Hokkaido, Japan, respectively, Gustafsson et al. (2001) assumed the ground to contribute half of the heat flux to the overall snowmelt. The influence of rain on snow and of meltefreeze cycles on a chemical’s elution behavior is outlined in Meyer and Wania (2008). All elution sequences of organic chemicals presented herein refer to the melt water concentrations normalized to the average concentration of a particular chemical during one simulation.

2.3.

Model calibration and validation

The model was calibrated against the experimental results from Meyer et al. (2009a,b), who used artificially produced and contaminated snow. This type of snow behaves similar as recently deposited snow with respect to the physical properties during wet snow metamorphism. Chemical partitioning within the generated snow was also found to be similar to that of natural snow (Meyer and Wania, 2008; Meyer et al., 2009a). The chemicals that were subjected to melt experiments cover a wide range of environmental partitioning properties (Table S1). Amphiphilic substances, however, are not considered, because they likely partition to the air-water interface to a relatively large extent, rendering their elution behavior somewhat uncertain. Three parameters were adjusted during model calibration. First, the snow layer thickness was set to 10 cm for all calculations and the number of snow layers calculated from the total snow depth (Fig. 1). Second, the KD values were adjusted using a procedure which is described in the Supplementary Data. Third, the snow permeability to particles was set to 5%, which is in the lower range of permeability’s in naturally occurring snow. This parameter can vary widely depending on snow pack characteristics such as the amount and nature of the PM, melt intensity, snow grain size and shape, snow density, and environmental influences such as rainfall. In the calculations presented in Sections 4 and 5 snow permeability to particles was set to 25% in order to reflect conditions likely to be prevalent in natural snow. All other input parameters were based on the experimental measurements. After calibration, the model succeeded in reproducing the type 1 and 2 chemical enrichments that had been observed in the laboratory (see Section 3).

3.

Simulation of snowmelt experiments

3.1.

Relatively water soluble chemicals

The snowmelt elution sequences of the pesticides atrazine and lindane, as well as the polycyclic aromatic hydrocarbons

3630

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 2 7 e3 6 3 7

Fig. 1 e Diagram of the multi-layered snow pack model at an early stage of melting (WC means water content). The surface deposit consists of chemicals with a high affinity to PM and snow grain surfaces.

(PAHs) phenanthrene, pyrene, and benzo(ghi)perylene from four experiments in Meyer et al. (2009b) are compared to simulation results (Figs. 2 and 3). The chemical enrichment at the beginning of melting (Fig. 2) is quantitatively defined as the chemical fraction present within the first quarter of the released melt water (excluding the last sample that refers to the surface deposit). Enrichment at the end of melting (Fig. 3) is quantified as the final snow pack surface deposit (last bar of each plot) divided by the overall amount of chemical. The snow properties used in model scenarios AeD are consistent with those from the experiments (Meyer et al., 2009b e Table 1). The early chemical enrichment of atrazine and lindane for different snow pack thicknesses (scenario A: 16 cm, scenario B: 29 cm) is illustrated in Fig. 2. Enrichment is stronger in the deeper snow pack (see Section 5.1). Fig. 2 shows also the elution sequences of atrazine and lindane that were determined for

aged and coarse-grained snow with high melt water content (scenario C), and for fresh and fine-grained snow containing low melt water content (scenario D) (Fig. 2). The enrichment of lindane is stronger in scenario C compared to scenario D, whereas enrichment of atrazine differs less between scenarios (see Section 5.4). The snowmelt behavior of chemicals with intermediate partition properties such as lindane is more dependent on the varying snow pack properties than those of very water soluble substances such as atrazine (Meyer et al., 2009b). Why the differences between experimentally derived and modeled enrichments are larger in scenario A than in the other scenarios, is not entirely clear. A reason could be that the limited resolution of the model leads to an overestimation of the early enrichment in snow packs of lower depths. E.g., the overestimation of enrichment in scenario A may be related to the need to represent an experimental snow depth of only 16 cm with two layers of 10 cm thickness each.

3631

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 2 7 e3 6 3 7

Melt scenario B

Melt scenario A

deep snow pack

shallow snow pack

relative chemical amount per sample

Experiment

Experiment

Model 76%

ATR 74%

ATR 82%

LIN 51%

LIN 49%

LIN 55%

ATR

ATR

61%

LIN 37%

Melt scenario C - high melt water to surface area ratio

Experiment

Model

Melt scenario D - low melt water to surface area ratio

Model

Experiment

Model

ATR 82%

ATR 83%

ATR 80%

ATR 76%

LIN 55%

LIN 59%

LIN 39%

LIN 41%

sample order Fig. 2 e Comparison of relative elution sequences of atrazine (ATR) and lindane (LIN) in melt scenarios AeD, between experiment and model. Experimental plots: blue columns e dissolved phase; brown columns e particulate fractions; model plots: green columns. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

3.2.

Relatively hydrophobic chemicals

Fig. 3 illustrates the peak release of the more hydrophobic PAHs from melting snow in both model and experiment. Because of its hydrophobicity most benzo(ghi)perylene that is transported through melting snow is sorbed to particles in both model and experiment. Phenanthrene and to a lesser extent pyrene are additionally present within the aqueous melt water phase. Enrichment of phenanthrene and pyrene is larger in scenario D compared to scenario C, whereas in the case of benzo(ghi)perylene enrichments in both scenarios are similar. This behavior reflects the chemicals’ partitioning properties. Somewhat water soluble chemicals such as phenanthrene and pyrene are partially released within the aqueous melt water phase, whereas the extent of this release is larger in scenario C. The respective enrichment at the end of melting is accordingly smaller.

4.

Types of elution behavior

4.1. Chemical phase distribution of organic contaminants in melting snow Chemical properties determine the phase partitioning in a melting snow pack and therefore a chemical’s fate during melting (Meyer and Wania, 2008; Meyer et al., 2009a). Chemical phase partitioning is illustrated using chemical space plots defined by the chemical’s air/water partition coefficient log KAW and the particle/water partition coefficient log KD for a variety of snow surface/air sorption coefficients log (KI/A/m) (Tables S1eS3). Within this partitioning space, regions of predominant presence of a chemical within the bulk snow are illustrated: snow grain surface (bright-blue), PM (olive-green),

3632

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 2 7 e3 6 3 7

Fig. 3 e Comparison of relative elution sequences of phenanthrene (PHE), pyrene (PYR), and benzo(ghi)perylene (BghiP) in melt scenarios C, D, between experiment and model. The length of the last column in each plot was divided by 10 for better view. Experimental plots: blue columns e dissolved phase; brown columns e particulate fractions; model plots: green columns. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

aqueous phase (dark-blue) (Fig. 4). The location of the boundaries between those regions, i.e. the threshold K-values delineating a transition from one phase to another, shifts during snow metamorphism as density, SSA, water content, and PM content change (Meyer and Wania, 2008). The thresholds in Fig. 4 represent one snowmelt scenario, involving aged snow exhibiting a snow density of 0.2 g cm3, an SSA of 200 cm2 g1, a melt water content of 6%, and an intermediate PM content of 50 mg L1. Yellow lines on the maps separate areas of partition properties that lead to different types of chemical enrichment during melting. They will be discussed in Sections 4.2e4.5. The red stripes on the maps in Fig. 4 represent organic contaminants that have previously been measured in snow and comprise the PAHs acenaphthene, fluorene, and phenanthrene (Usenko et al., 2010), the pesticides chlorothalonil and chlorpyrifos (Mast et al., 2007), and the polychlorinated biphenyl congener PCB-180 (Quiroz et al., 2009). The red stripes also reflect the chemicals’ varying KD values, illustrating the variable sorptive capacity of different types of PM (see Section 2.1). For the example simulations presented in Sections 4 and 5 the KD values are assumed to be in the lower part of that range (i.e. the upper end of the stripes in Fig. 4). Further, the snow depth was set to 40 cm, reflecting a natural snow pack of intermediate depth (Environment Canada, 2000).

4.2.

Type 1 enrichment e dissolved chemicals

Chemicals that dissolve appreciably (more than 75%) into the aqueous snow phase, such as chlorothalonil as defined in Section 4.1 (also atrazine e Section 3.1), are released along

with the first downward moving melt water in type 1 enrichment mode: peak chemical concentrations are simulated to occur in the first melt water exiting the snow pack (Type 1A in Fig. 5). Elevated concentrations of chlorothalonil have previously been observed in stream water coinciding with the onset of the spring melt period (Meyer et al., 2011). Based on their predicted phase partitioning a wide variety of relatively water soluble organic contaminants, including several pesticides and halogenated acids, is expected to display type 1A elution behavior in most melt scenarios. If the chemical is sorbed to particles to a relatively small extent (less than 10%), only 25% needs to be dissolved in order to be released in type 1 mode (Fig. 4). The steepness of the type 1 curve in Fig. 5 increases with distance to the yellow line that separates type 1 from the other types of enrichment in Fig. 4. However, when more than 90% of the chemical is dissolved (dark-blue region in Fig. 2) the elution behavior does not change notably anymore. An elution profile of Type 1B (Fig. 5) is simulated for chlorothalonil in snow that is melting both at the surface and at the bottom: the additional bottom melt leads to a less pronounced and delayed type 1 peak. Whereas surface melt leads to chemical enrichment, bottom melt does not (Meyer et al., 2009b). The model suggests that only type 1 elution behavior is notably influenced by bottom melt.

4.3. Type 2 enrichment e particle and snow grain surface associated chemicals Chemicals largely associated with particles, such as PCB-180 (Fig. 4) (see also PAHs in Section 3.2), as well as substances

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 2 7 e3 6 3 7

3633

Fig. 4 e Chemical space plots of the phase distribution of chemicals in a melting snow pack as a function of log (KI/A/m), log KD, and log KAW, at 0  C. Boundaries between the colored regions refer to a chemical presence of 10%, 50%, or 90% within the different bulk snow phases. Yellow lines separate areas in which different types of chemical enrichment prevail. Six real chemicals are placed on the map (red stripes) based on partition coefficients estimated based on Roth et al. (2004), or taken from the literature (Table S3). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

with a very large affinity for snow grain surfaces mostly accumulate at the snow pack surface during melting, and are released at the end of melting in type 2 mode. The elution sequence is characterized by constant, relatively small melt water concentrations over the course of the melt period, followed by a pronounced peak at the very end of melting (Fig. 5). For both particle and snow grain associated chemicals, sorption leads to the retention of the chemical within a melting snow pack. However, both forms of retention are fundamentally different. Whereas particle-associated chemicals are filtered out on top of the snow pack (Meyer and Wania, 2008), snow grain surface associated chemicals accumulate within the upper part of the snow pack itself due to sorption forces and the continuously lowering snow pack surface. When a chemical is less water soluble (<25% presence in the aqueous phase) and at the same time sorbed to PM by more

than 5% or sorbed to the snow grain surface by more than 95%, it elutes in type 2 mode. Time-resolved concentrations of PCBs in melt water receiving river water during spring snowmelt periods have been reported (see Meyer and Wania, 2008). Whereas Semkin et al. (1996) reported increasing concentrations toward the end of the melt period, Lafrenie`re et al. (2006) observed PCBs primarily in the early phase of melting, the latter presumably caused by a low PM content in the stream water (see Section 5.2). Due to their high affinity for PM, other persistent organic pollutants such as polychlorinated dibenzo-p-dioxins and dibenzofurans, and polybrominated diphenyl ethers might display type 2 enrichment as well. Chemicals that are known to exhibit very large affinity for snow grain surfaces and therefore are also predisposed for type 2 enrichment are longer-chained semi-fluorinated alkanes (Plassmann et al., 2010).

3634

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 2 7 e3 6 3 7

Fig. 5 e Simulated chemical elution sequences used to illustrate the four categories of chemical enrichment in snowmelt water, generated by the model while applying the snowmelt scenario defined in Section 4.1.

4.4. Type 3 enrichment e somewhat water soluble, snow grain surface associated chemicals Chemicals with intermediate partitioning between the snow grain surface (75e95% presence) and the aqueous melt water phase (5e25% presence) (Fig. 4) display concentrations in the melt water that increase over the course of the melt (type 3 enrichment in Fig. 3). With progressing melt and increasing chemical concentration at the snow grain surface, chemical equilibrium between the bulk snow phases can only be maintained when an increasing amount of chemical transfers into the aqueous phase, leading to the ascending elution sequence. Such elution behavior is predicted for the pesticide chlorpyrifos in the melt scenario defined in Section 4.1 (Fig. 4). In order to define a boundary between type 2 and type 3 enrichment in Fig. 2, in type 3 mode less than 50% of the initial chemical amount present in the bulk snow accumulates as surface deposit during melting. The elution behavior of chemicals that are located in the type 3 area of the maps in Fig. 4 is very dependent on snow pack properties (see Section 5.4).

4.5. Type 4 enrichment e combination of two types of enrichment Organic substances that are fairly equally distributed between the dissolved phase and the particulate phase (30e70% presence in either of the two phases), such as fluorene in the example melt scenario (Fig. 4) (see also lindane in Section 3.1), shows type 4 elution behavior (Fig. 5). This category combines type 1 and PM related type 2 enrichment, whereby a dissolved fraction is released early during the melt and a particle-bound fraction is retained on top of the snow pack. Interestingly, such a double peak is not observed when the chemical is located at the boundary separating type 1 and type 3 enrichments in Fig. 4. In this case the elution sequence simply shows

uniform elution, with no enrichment at all. Similarly to type 3 enrichment, the elution behavior of chemicals assigned to type 4 in Fig. 4 is very dependent on snow pack properties. A chemical may switch from one type of enrichment to another over the course of a melt period as the snow pack properties change (Meyer and Wania, 2008).

5.

Influence of snow pack properties

The elution behavior of an organic chemical not only depends on its partitioning properties but also on snow pack characteristics. After having established confidence in the model by showing its ability to reproduce experimental observations, we used it to simulate how the different types of elution sequences change in response to variations in snow input parameters. Those variations comprise the range of natural variability, i.e. they cover the properties of most naturally occurring snow packs. Only those cases are presented in which parameter variability causes significant changes in the chemical’s enrichment pattern. The standard snow pack parameterization described in Section 4.1 is used, and only one parameter at a time was varied.

5.1.

Influence of snow depth

The experimental investigations by Meyer et al. (2009b) included only two snow packs of different depth (Section 3.1). Here, we explore the elution behavior of organic chemicals for snow pack depths ranging from 0 to 150 cm (Fig. S1). Only the type 1 elution pattern is notably influenced by snow depth. With increasing depth the chemical fraction in the first quarter of melt water increases very rapidly and eventually approaches 100% asymptotically. The downward percolating melt water front takes up more chemicals from

3635

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 2 7 e3 6 3 7

snow grain surfaces within a deeper snow pack, resulting in more pronounced peak concentrations in the initially released melt water. The net chemical transfer from snow grains to melt water front declines with travel distance and finally ceases in order to maintain equilibrium between the snow phases. In homogeneous snow packs less than approximately 50 cm thick, snow depth is the single most influential parameter for type 1 enrichment of very water soluble organic chemicals.

5.2.

Influence of particle content in snow

The fraction of PM present in snow exhibits the widest natural variability of all investigated parameters, ranging over several orders of magnitude. Particulate organic matter concentrations as low as w50 mg L1 (SWE) (Hagler et al., 2007) and soot concentrations as low as w3 mg L1 (SWE) (Masclet et al., 2000) have been found in Greenland snow. A total particle content of several grams per liter in melt water from urban snow (Viklander, 1999) is the other extreme. Fig. 6A shows the simulated elution sequences of phenanthrene and chlorothalonil as released from “clean” snow with 50 mg PM L1, and “dirty” snow with 500 mg PM L1. In the standard snow pack from Section 4.1, the two chemicals had shown type 1 and 2 enrichment, respectively (Fig. 4). However, in clean snow both substances show type 1 elution behavior. The PM content is so low that the bulk of the relatively hydrophobic phenanthrene becomes dissolved in the melt water phase. Daub et al. (1994) presented observational evidence of such partitioning behavior by reporting a transfer of particulate phenanthrene and pyrene into the dissolved phase during snowmelt as the suspended solid concentration within the melt water dropped. In dirty snow on the other hand even fairly water soluble chemicals such as chlorothalonil can become sorbed to particles to such a large extent to be released in type 2 mode (Fig. 6A). In this case the olive-green region in the lower left of the chemical space plots of Fig. 4 expands significantly, and the

A

partition property combination of chlorothalonil falls into the area designating susceptibility to type 2 chemical enrichment.

5.3.

Influence of snow permeability to particles

The extent to which a snow pack allows particle transport along with the downward percolating melt water influences the snowmelt behavior of particle-associated organic chemicals. To account for a wide range of naturally occurring snow permeability, we simulated the elution of PCB-180 from snow packs exhibiting permeabilities of 0% and 50%. The extent of type 2 enrichment of PCB-180 decreased with increasing permeability (Fig. 6B). When the snow pack allows higher rates of particle transport, less of the particle-bound chemical accumulates at the snow pack surface. Hence, type 2 enrichment decreases. Zero permeability to particles means that only the chemical fraction that is dissolved in the aqueous melt water phase will be eluted during melting (Fig. 6B).

5.4.

Influence of internal snow surface area

The elution behavior of organic chemicals with notable affinity to the snow grain surface is affected by the size of the internal snow surface area, here defined as the product of the SSA and snow density. Surface area in natural snow ranges between approximately 200 and 20,000 m2 m3 (Domine´ et al., 2007). To investigate the influence of surface area, the elution of acenapthene from two snow packs representing the upper and lower range of most commonly encountered surface areas was simulated (Fig. 6C): one fresh and fine-grained (10,000 m2 m3), the other aged and coarse-grained (1000 m2 m3). The elution behavior in the two scenarios reflects the different extent to which acenaphthene is retained in the snow packs during melting. While acenaphthene is released in type 3 mode in snow with a large surface area, it undergoes type 1 enrichment in snow that exhibits a small surface area.

B

C

Fig. 6 e Influence of PM content (A), snow permeability to particles (B), and snow surface area (C), on the simulated elution behavior.

3636

6.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 2 7 e3 6 3 7

Conclusions

A simple, mechanistic snowmelt model was developed and applied to investigate the different types of enrichment of organic chemicals in melt water. The model identifies which chemicals are subject to which type of enrichment and to what extent. It also accounts for the influence of snow properties on enrichment behavior. Thus, assuming the snow pack is homogeneous the organic contaminant elution can be reproduced and explained with a very simple model. The model may be beneficial whenever the differential release of organic chemicals from snow is important. E.g. every winter large amounts of highly polluted snow are deposited at snow dumps in northern urban areas. Upon melt the organic contaminant load often enters surface waters and aquifers untreated. Using the snowmelt model, the timing of contaminant release from the snow piles as well as the extent of enrichment of various substances in the melt water can be estimated, which in turn can inform water management solutions. Another application of the model is to assist in the interpretation of deposition profiles in snow cores from alpine and arctic ice caps, which have been widely used to study current and historical inputs of anthropogenic organic contaminants. Limited melting, which is common in many ice caps, may cause relocation of water soluble organic chemicals deeper into the firn, whereas hydrophobic substances are more likely to remain in place. With knowledge of local conditions, the model may aid in quantifying the possibility and extent of relocation of different contaminants within the ice core. A slightly modified version of the model would further aid in predicting the release of legacy organic pollutants that have been deposited and stored in alpine glaciers for decades. In response to a warming climate melting glaciers now act as a secondary source of contaminants to aquatic and terrestrial environments (Bogdal et al., 2010).

Appendix. Supplementary material The supplementary data associated with this article can be found in the on-line version at doi:10.1016/j.watres.2011.04.011.

references

Bizzotto, E.C., Villa, S., Vaj, C., Vighi, M., 2009. Comparison of glacial and non-glacial-fed streams to evaluate the loading of persistent organic pollutants through seasonal snow/ice melt. Chemosphere 74 (7), 924e930. Bogdal, C., Nikolic, D., Lu¨thi, M.P., Schenker, U., Scheringer, M., Hungerbu¨hler, K., 2010. Release of legacy pollutants from melting glaciers: model evidence and conceptual understanding. Environmental Science and Technology 44 (11), 4063e4069. Colbeck, S.C., 1979. Grain clusters in wet snow. Journal of Colloid and Interface Science 72 (3), 371e384. Daly, G.L., Wania, F., 2004. Simulating the influence of snow on the fate of organic compounds. Environmental Science and Technology 38 (15), 4176e4186.

Daub, J., Fo¨rster, J., Herrmann, R., Robien, A., Striebel, T., 1994. Chemodynamics of trace pollutants during snowmelt on roof and street surfaces. Water Science and Technology 30 (1), 73e85. Domine´, F., Taillandier, A.-S., Simpson, W.R., 2007. A parameterization of the specific surface area of seasonal snow for field use and for models of snow pack evolution. Journal of Geophysical Research 112 (F2), F02031. Environment Canada, MSC, 2000. Canadian Snow Data CD-ROM. CRYSYS Project, Climate Processes and Earth Observation Division, Meteorological Service of Canada, Downsview, Ontario. January 2000. Gustafsson, D., Sta¨hli, M., Jansson, P.-E., 2001. The surface energy balance of a snow cover: comparing measurements to two different simulation models. Theoretical and Applied Climatology 70, 81e96. Hagler, G.S.W., Bergin, M.H., Smith, E.A., Dibb, J.E., 2007. A summer time series of particulate carbon in the air and snow at Summit, Greenland. Journal of Geophysical Research e Atmospheres 112 (D21), D21309. Harrington, R., Bales, R.C., 1998. Modeling ionic solute transport in melting snow. Water Resources Research 34 (7), 1727e1736. Lafrenie`re, M.J., Blais, J.M., Sharp, M.J., Schindler, D.W., 2006. Organochlorine pesticide and polychlorinated biphenyl concentrations in snow, snowmelt, and runoff at Bow Lake, Alberta. Environmental Science and Technology 40 (16), 4909e4915. Lee, J., Feng, X., Posmentier, E.S., Faiia, A.M., Osterhuber, R., Kirchner, J.W., 2008. Modeling of solute transport in snow using conservative tracers and artificial rain-on-snow experiments. Water Resources Research 44 (2), W02411. Lohmann, R., Lammel, G., 2004. Adsorptive and absorptive contributions to the gas-particle partitioning of polycyclic aromatic hydrocarbons: state of knowledge and recommended parametrization for modeling. Environmental Science and Technology 38 (14), 3793e3803. Loseto, L.L., Lean, D.R.S., Siciliano, S.D., 2004. Snowmelt sources of methylmercury to high Arctic ecosystems. Environmental Science and Technology 38 (11), 3004e3010. Lundberg, A., Yuichiro, N., Thunehed, H., Halldin, S., 2004. Snow accumulation in forests from ground and remote-sensing data. Hydrological Processes 18 (10), 1941e1955. Marsh, P., Woo, M.-K., 1984. Wetting front advance and freezing of meltwater within a snow cover: 1. Observations in the Canadian Arctic. Water Resources Research 20 (12), 1853e1864. Marsh, P., 1999. Snowcover formation and melt: recent advances and future prospects. Hydrological Processes 13, 2117e2134. Masclet, P., Hoyau, V., Jaffrezo, J.L., Cachier, H., 2000. Polycyclic aromatic hydrocarbon deposition on the ice sheet of Greenland. Part I: superficial snow. Atmospheric Environment 34, 3195e3207. Mast, M.A., Foreman, W.T., Skaates, S.V., 2007. Current-use pesticides and organochlorine compounds in precipitation and lake sediment from two high-elevation national parks in the Western United States. Archives of Environmental Contamination and Toxicology 52, 294e305. Meyer, T., Wania, F., 2008. Organic contaminant amplification during snowmelt e A review. Water Research 42 (8e9), 1847e1865. Meyer, T., Lei, Y.D., Wania, F., Muradi, I., 2009a. Organic contaminant release from melting snow: influence of chemical partitioning. Environmental Science and Technology 43 (3), 657e662. Meyer, T., Lei, Y.D., Wania, F., Muradi, I., 2009b. Organic contaminant release from melting snow: influence of snow pack and melt characteristics. Environmental Science and Technology 43 (3), 663e668.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 2 7 e3 6 3 7

Meyer, T., Wania, F., 2010. Atmosphere-snow exchange. In: Thibodeaux, L.J., Mackay, D. (Eds.), Environmental Mass Transfer Handbook. CRC Press/Taylor & Francis, pp. 513e530. Meyer, T., Lei, Y.D., Wania, F., 2011. Transport of polycyclic aromatic hydrocarbons and pesticides during snowmelt within an urban watershed. Water Research 45, 1147e1156. Plassmann, M.M., Meyer, T., Lei, Y.D., Wania, F., McLachlan, M.S., Berger, U., 2010. Theoretical and experimental simulation of the fate of semifluorinated n-alkanes during snow melt. Environmental Science and Technology 44, 6692e6697. Quiroz, R., Popp, P., Barra, R., 2009. Analysis of PCB levels in snow from the Aconcagua mountain (Southern Andes) using the stir bar sorptive extraction. Environmental Chemistry Letters 7 (3), 283e288. Roth, C.M., Goss, K.-U., Schwarzenbach, R.P., 2004. Sorption of diverse organic vapors to snow. Environmental Science and Technology 38 (15), 4078e4084. Scho¨ndorf, T., Herrmann, R., 1987. Transport and chemodynamics of organic micropollutants and ions during snowmelt. Nordic Hydrology 18, 259e278. Semkin, R.G., Alaee, M., Fox, J., Franklyn, J., Gray, B., Hanna, M., Healey, P., Mawhinney, M., Neureuther, R., Rowsell, R., Strachan, W., Teixeira, C., Amarualik, P., Diamond, M., Freitas, H., Gregor, D., Kinney, S., Prchalova, M., Koenig, B., 1996. Processes and fluxes of contaminants in aquatic systems e 1994/95. In: Murray, J.L., Shearer, R.G., Han, S.L. (Eds.), Synopsis of research conducted under the 1994/95 northern

3637

contaminants program. Indian and Northern Affairs Canada, Environmental Studies Report No. 73, pp. 105e118. Simmleit, N., Herrmann, R., Thomas, W., 1986. Chemical Behaviour of Hydrophobic Micropollutants During the Melting of Snow, vol. 155. IAHS Publication, pp. 335e346. Stocker, J., Scheringer, M., Wegmann, F., Hungerbu¨hler, K., 2007. Modeling the effect of snow and ice on the global environmental fate and long-range transport potential of semi-volatile organic compounds. Environmental Science & Technology 41 (17), 6192e6198. Usenko, S., Simonich, S.L.M., Hageman, K.J., Schrlau, J.E., Geiser, L., Campbell, D.H., Appleby, P.G., Landers, D.H., 2010. Sources and deposition of polycyclic aromatic hydrocarbons to Western U.S. national parks. Environmental Science & Technology 44 (12), 4512e4518. Viklander, M., 1999. Substances in urban snow. A comparison of the contamination in snow in different parts of the city of Lulea˚, Sweden. Water, Air, and Soil Pollution 114 (3e4), 377e394. Waldner, P.A., Schneebeli, M., Wunderli, H., 2000. Na¨hrstoffaustrag aus einer schmelzenden Schneedecke im Alptal (Kanton Schwyz) am Beispiel von Nitrat. Schweizerische Zeitschrift fu¨r Forstwesen 6, 198e204. Waldner, P.A., Schneebeli, M., Schultze-Zimmermann, U., Flu¨hler, H., 2004. Effect of snow structure on water flow and solute transport. Hydrological Processes 18 (7), 1271e1290. Wania, F., 1997. Modelling the fate of non-polar organic chemicals in an ageing snow pack. Chemosphere 35 (10), 2345e2363.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 3 8 e3 6 5 2

Available at www.sciencedirect.com

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

Sorption of biocides, triazine and phenylurea herbicides, and UV-filters onto secondary sludge Arne Wick a, Olivian Marincas b, Zaharie Moldovan b, Thomas A. Ternes a,* a b

Federal Institute of Hydrology (BfG), D-56068 Koblenz, Am Mainzer Tor 1, 56068 Koblenz, Germany National Institute of Research and Development for Isotopic and Molecular Technology, Donath Str. 71-103, 400 293 Cluj-Napoca, Romania

article info

abstract

Article history:

The sludge-water distribution of a total of 41 organic micropollutants (9 phenylurea herbi-

Received 11 November 2010

cides, 11 triazines, 16 biocides and 5 UV-filters) was investigated in laboratory batch experi-

Received in revised form

ments with fresh secondary sludge taken from a municipal WWTP. Sorption kinetics as well as

29 March 2011

sorption isotherms were examined by analyzing the compound concentration in the aqueous

Accepted 7 April 2011

and solid phase for mass balance control and quality assurance. The sorption kinetic exper-

Available online 19 April 2011

iments revealed a sorption equilibrium time of <2 h and adverse effects of sodium azide on the sludge-water distribution of several compounds. Sorption isotherms were constructed for 6

Keywords:

different spiking levels spanning 3 orders of magnitude (100 ng L
1e30,000 ng L
1) and were

Triazines

well described by the Freundlich model. For some compounds non-linear sorption with

Phenylurea herbicides

Freundlich exponents n < 1 revealed a decreased sorption affinity to the sludge flocs with

Biocides

increasing aqueous phase concentration. Therefore, sludge-water distribution coefficients

UV-filters

(Kd,sec) were calculated from the isotherm data for a constant concentration level of 1 mg L
1.

Secondary sludge

Based on the sludge dry weight (dw), the Kd,sec values of phenylurea herbicides ranged

Sludge-water distribution coefficients Freundlich isotherms


1 (isoproturon) to 320 L kgdw sludge
1 (neburon), those of triazines from sludge
1 (atrazine) to 190 L kgdw sludge
1 (terbutryn), those of biocides from 5 L kgdw sludge 10 L kgdw sludge
1 (N,N-dimethyl-N0 -p-tolylsulfamide) to 40,000 L kgdw sludge
1 (triclocarban) and those of UV-filters from 9 L kgdw sludge
1 (phenylbenzimidazole sulfonic acid) to 720 L kgdw sludge
1 (benzophenone-3). For most compounds Kd,sec values were below 500 L kgdw sludge
1 and thus removal in WWTPs by the withdrawal of excess sludge is expected

from 9 L kgdw

to be negligible (<10%) except for the biocides triclocarban (80e95%), triclosan (55e85%), chlorophene (30e60%), imazalil (25e55%) and fenpropimorph (15e40%) as well as the UV-filter benzophenone-3 (5e20%). A simple linear free-energy relationship (LFER) approach using the logarithmized octanolewater partition coefficient log KOW as single descriptor is discussed for a rough classification of nonionic compounds regarding their potential removal in WWTPs by sorption. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

For many emerging organic micropollutants of ecotoxicological concern such as pharmaceuticals, biocides, UV-filter or perfluorinated flame retardants, WWTPs are the main source

for their release into the environment (Kupper et al., 2006; Richardson, 2009). Sorption on activated sludge is one of the key factors controlling the removal of organic micropollutants in conventional wastewater treatment plants (WWTPs) (Ternes and Joss, 2007). Thus, information on sludge-water

* Corresponding author. Tel.: þ49 261 1306 5560; fax: þ49 261 1306 5363. E-mail address: [email protected] (T.A. Ternes). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.014

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 3 8 e3 6 5 2

distribution is crucial to predict the influence of sorption on their fate in WWTPs and the quantity released into the aquatic environment. However, especially for the often moderately polar emerging micropollutants systematic studies on sludge-water distribution are rare (Carballa et al., 2008; Maurer et al., 2007; Ternes et al., 2004). Many previous studies relied on single point concentrations rather than sorption isotherms, thus assuming sorption being independent from the actual concentration. However, for compounds featuring high hydrophobicity or specific polar and ionic moieties leading to specific interactions with the sludge surface, the linear approach might not be valid due to a limited sorption capacity (Clara et al., 2004; Langford et al., 2005; Thomas et al., 2009). So far, the results suggest that simple empirical model approaches based on the correlation of the effective octanolewater partitioning coefficient (KOW) with the organic carbon based sludge-water distribution coefficient (KOC) are inappropriate for predicting the sludge sorption affinity of polar and ionic compounds (Carballa et al., 2008; Ternes et al., 2004). The sorption behavior can depend on different physicochemical properties of the micropollutant (hydrophobicity, hydrogen bonds, electrostatic interactions, etc.) and the sorbent, e.g. different sludge types (primary, secondary and digested sludge) and sludge ages (Barret et al., 2010; Langford et al., 2005; Ochoa-Herrera and Sierra-Alvarez, 2008). In case of rapid biodegradation of the target compounds (kbiol > 0.1  ksorb), sorption equilibrium might not be achievable and hence, reliable sorption coefficients cannot be determined without inhibiting the microbial activity (Ternes and Joss, 2007). Even if biodegradation kinetics are slow compared to sorption kinetics and a sorption equilibrium can be assumed, calculation of the sorbed quantity from the soluble quantity as proposed by the OECD guideline 106 (2000) can lead to an overestimation of sludge-water distribution coefficients (Ramil et al., 2010; Stein et al., 2008; Wick et al., 2009). Several methods including the addition of inhibitors of microbial activity such as sodium azide (NaN3) (Dobbs et al., 1995), mercury chloride (HgCl2) (Maurer et al., 2007) and mercury sulfate (HgSO4) (Clara et al., 2004) as well as autoclaving (Dobbs et al., 1995; Zhao et al., 2008) and freeze-drying followed by heating (Andersen et al., 2005) have been applied for sludge deactivation. However, freeze-drying and autoclaving alters the texture of the sludge flocs and thus the sorbent structure (Barret et al., 2010; Berns et al., 2008). Chemical deactivation of the sludge might also influence the sorption processes due to inactivation of cells and consecutive cell lysis, or by chemical interaction and reaction with the target compounds (Chefetz et al., 2006; Gaillardon, 1996). The objective of the current study was to elucidate the sorption behavior of several structurally diverse emerging organic micropollutants including biocides, UV-filters as well as triazines and phenylurea herbicides as for most of these compounds this data is still lacking. A special focus was set on (i) the sorption kinetics with and without using NaN3 for inhibition of the microbial activity, (ii) the concentration dependency of sorption (sorption isotherms) and (iii) the correlation of the log KOW as a measure of hydrophobicity with sludge-water distribution coefficients. The selected analytes are listed in Table 1.

2.

Experimental

2.1.

Chemicals

3639

Triazines and phenylurea herbicides were purchased from Dr. Ehrenstorfer (Augsburg, Germany). All other reference standards were received from the same suppliers as described in Wick et al. (2010). Methanol (picograde) and acetonitrile (HPLC grade) were purchased from LGC Promochem (Wesel, Germany). Formic acid and sulfuric acid (both p.a.) were purchased from Merck (Darmstadt, Germany) and ammonium formate (purum grade) from SigmaeAldrich (Schnelldorf, Germany). Ultrapure water was obtained from a Milli-Q system (Millipore, Billerica, MA, USA). Separate standard solutions of all analytes and surrogate standards were prepared in methanol at a concentration of 10 and 1 mg mL
1, respectively, and stored at 4  C in the dark.

2.2.

Sampling

The sludge for the laboratory batch experiments was taken in solvent-rinsed amber glass bottles from the nitrifying zone of the activated sludge treatment of a German municipal WWTP serving 300,000 population equivalents (PE). The activated sludge system is operated with a hydraulic retention time (HRT) and sludge retention time (SRT) of 12 h and 10e12 d, respectively. Details of the complete treatment train can be found elsewhere (Wick et al., 2010). Fresh sludge (concentration of total suspended solids (TSS): 4 gTSS L
1, total organic carbon (TOC) per gram dry weight of sludge: 0.29 gOC gdw sludge
1, pH 6.8) was directly transported to the lab and sorption batch experiments were initiated within 2 h.

2.3. Sorption kinetics: sorption equilibrium and influence of sodium azide A homogenous suspension (190 mL) of the activated sludge was filled into 500 mL centrifuge bottles (polypropylene). For each sampling time point (0, 0.75, 1.5, 3.2, 6 and 24 h) three samples with 0, 0.2 and 1% NaN3 concentrations were prepared by adding 10 mL of Milli-Q water, 4% and 20% NaN3 solution, respectively. Target compounds were spiked to a concentration of 10 mg L
1 and the samples were placed on a horizontal shaker for the different incubation times. The background concentration of the target compounds in the sludge was assessed in non-spiked controls for the first sampling point (t ¼ 0 h). The ambient pH of the sludge was stable (6.8  0.2) throughout the incubation period. After the defined exposure time, the sludge was centrifuged (15 min, 3640 g) and the supernatant was stored at 4  C for solid-phase extraction (SPE). The solid sludge particles were freeze-dried for micropollutant analysis and the water content fwater (L gdw sludge
1) was determined by the difference of the sludge wet and dry weight before and after freeze-drying, respectively.

2.4.

Freundlich sorption isotherms

The experimental set-up and sludge characteristics were similar to the equilibrium experiments described above. The Freundlich isotherms were determined by spiking the target

3640

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 3 8 e3 6 5 2

Table 1 e Chemical structures of the investigated phenylurea and triazine herbicides, biocides, and UV-filters. Analyte Application CAS number

Chemical structure

Analyte Application CAS number

Diuron Herbicide/Biocide CAS: 33-54-1

Isoproturon Herbicide/Biocide CAS: 34123-59-6

Fluometuron Herbicide CAS: 2164-17-2

Chloroxuron Herbicide CAS: 1982-47-4

Linuron Herbicide CAS: 330-55-2

Monolinuron Herbicide CAS: 1746-81-2

Neburon Herbicide CAS: 555-37-3

Monuron Herbicide CAS: 150-68-5

Methabenzthiazuron Herbicide CAS: 18691-97-9

Atrazine Herbicide CAS: 1912-24-9

Simazine Herbicide/Algicide CAS: 122-34-9

Cyanazine Herbicide CAS: 21725-46-2

Propazine Herbicide CAS: 139-40-2

Terbuthylazine Herbicide/Biocide CAS: 5915-41-3

Terbutryn Herbicide/Algicide/ Biocide (Antifouling) CAS: 886-50-0

Irgarol Herbicide/Algicide/ Biocide (Antifouling) CAS: 28159-98-0

M1 Transformation product of irgarol CAS: 30125-65-6

Ametryn Herbicide CAS: 834-12-8

Chemical structure

3641

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 3 8 e3 6 5 2

Table 1 (continued) Analyte Application CAS number

Chemical structure

Analyte Application CAS number

Prometryn Herbicide CAS: 7287-19-6

Prometon Herbicide CAS: 1610-18-0

Mecoprop Herbicide/Biocide CAS: 7085-19-0

Propiconazole Fungicide/Biocide CAS: 60207-90-1

Tebuconazole Fungicide/Biocide CAS: 107534-96-3

Imazalil Fungicide/Biocide CAS: 35554-44-0

Carbendazim Fungicide/Biocide CAS: 10605-21-7

Thiabendazole Fungicide/Biocide CAS: 107534-96-3

Dimethomorph Fungicide CAS: 110488-70-5

Fenpropimorph Fungicide CAS: 67564-91-4

Tridemorph Fungicide CAS: 24602-86-6

1,2-Benzisothiazolin-3-one (BIT) Microbicide/Biocide CAS: 2634-33-5

2-n-Octyl-4-isothiazolin3-one (OIT) Microbicide/Biocide CAS: 26530-20-1

N,N-Dimethyl-N0 -ptolylsulfamide (DMST) Transformation product of tolylfluanide CAS: 668-40-71-9

N,N-Dimethyl-N0 phenylsulfamide (DMSA) (Transformation product of dichlofluanide CAS: 4710-17-2

Triclosan Microbicide/Biocide CAS: 3380-34-5

Triclocarban Microbicide/Biocide CAS: 101-20-2

Chlorophene Microbicide/Biocide CAS: 120-32-1

Chemical structure

(continued on next page)

3642

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 3 8 e3 6 5 2

Table 1 (continued) Analyte Application CAS number

Analyte Application CAS number

Chemical structure

Benzophenone-1 (BZP-1) UV-Filter CAS: 131-56-6

Benzophenone-2 (BZP-2) UV-Filter CAS: 131-55-5

Benzophenone-3 (BZP-3) UV-Filter CAS: 131-57-7

Benzophenone-4 (BZP-4) UV-Filter CAS: 4065-45-6

Chemical structure

Phenylbenzimidazole sulfonic acid (PBSA) UV-Filter CAS: 27503-81-7

compounds into 200 mL of secondary sludge to a concentration of 0.1, 0.3, 1, 3, 10 and 30 mg L
1. All samples were prepared in duplicate and incubated for 1.5 h to ensure sorption equilibrium. Non-spiked samples were analyzed at the beginning and at the end of the incubation period to assess the natural background concentration. Filtered (Whatman, GF6) samples of the supernatant of centrifuged sludge were spiked to a concentration of 10 mg L
1 and served as controls for any compound losses due to sorption to the glass vessels.

2.5.

Analytical methods

The analytical method used for extraction and detection of the target analytes has been described in detail in Wick et al. (2010). Briefly, the supernatant was filtered through glassfiber filters (Whatman, GF6) and 200 mL were adjusted to pH 6 with 3.5 M H2SO4. After addition of 200 ng of an internal standard mix (1 mg mL
1 stock solution), the analytes were enriched by SPE using 200 mg Oasis HLB cartridges (Waters, Milfort, USA) and eluted with 4  2 mL of a mixture of methanol and acetone (60/40, v/v). The detection was performed using reversed-phase liquid chromatography (Synergi FusionRP, 150  3 mm, 4 mm, Phenomenex, Aschaffenburg, Germany) coupled to electrospray mass spectrometry operated in the positive and negative ionization mode (API 4000 tandem MS, Applied Biosystems, Foster City, USA). Extraction of the freeze-dried sludge samples was conducted by pressurized liquid extraction (PLE) (ASE 200 Accelerated Solvent extractor, Dionex, Idstein, Germany). Approximately 0.2 g of the freeze-dried sludge was filled into 22 mL cells and 200 ng of the internal standard mix was added. PLE was accomplished with a mixture of water and methanol (50/50, v/v) at 80  C and 100 bar (4 cycles). The extract was diluted with groundwater to a final volume of 800 mL and further preparation was performed as described for the aqueous samples. All target analytes were quantified using an internal standard calibration and the limit of quantification (LOQ) was

defined as the second lowest calibration point in the regression as long as the calculated signal to noise ratio (S/N) of the analytes in the native sample extracts was >10 for the first transition (MRM 1) used for quantification and >3 for the second transition (MRM 2) used for confirmation. The retention time, MRM transitions and corresponding MS parameters of triazines and phenylurea herbicides not included in the study of Wick et al. (2010), are provided in the Appendix (Table A1).

2.6.

Method validation

The analytical method was entirely validated within the study of Wick et al. (2010). Since the assignment of stable isotopelabeled standards was adapted for some analytes and additional triazines and phenylurea herbicides were included, relative recoveries were examined to confirm the reliability of the method for all analytes within the current study. The relative recoveries were determined as the ratio of the spiked concentration and the quantified concentrations at a spiking level of 2 mg L
1 and 4 mg gdw sludge
1 in the aqueous phase and freeze-dried sludge, respectively (dublicate measurements). The results are reported in the Appendix (Table A2).

2.7.

Calculations

Sorption isotherms were fitted with the Freundlich model which is a commonly applied empirical model to describe environmental sorption processes (Schwarzenbach et al., 2003). According to the Freundlich model, the sorbed concentration Cs (mg kgdw sludge
1) is related to the soluble concentration Cw (mg L
1) at sorption equilibrium based on the following equation: Cs ¼ Kf  Cnw

(1) 1
n

n

kgdw sludge
1)

where Kf (mg L is the Freundlich sorption coefficient and n (dimensionless) the Freundlich affinity constant.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 3 8 e3 6 5 2

If the sorption isotherm is linear (n ¼ 1), sorption will be independent from concentration and can be determined by the sludge-water distribution coefficient Kd,sec (L kgdw sludge
1): Kd;sec ¼

Cs Cw

(2)

For comparison with literature data, Kd,sec values were normalized to the fraction of total organic carbon fOC (kgOC kgdw sludge
1) resulting in the KOC (L kgOC
1): KOC ¼

Kd;sec fOC

(3)

Assuming no significant biodegradation, the removal efficiency via withdrawal of excess sludge hsorb (%) depends on the sludge production SP (kgTSS L
1) as follows:

hsorp ¼

Kf Cw n
1 SP  100 1 þ Kf Cw n
1 SP

(4)

The mass balance, i.e. the recovery Rec. (%) of the spiked compound concentration, was determined according to: Rec: ¼

Cs þ Cw  100 C0;spiked

(5)

The soluble concentration Cw was inserted in eq. (5) after subtraction of the soluble background concentration Cw,background: Cw ¼ Cw;spiked
Cw;background

(6)

For calculation of the sorbed concentration Cs inserted in eq. (5), the sorbed background concentration Cs,background as well as the soluble concentration in the sludge sample after centrifugation derived from the remaining water content fwater (L gdw sludge
1) were considered:
Cs ¼ Cs;spiked
Cw;spiked  fwater
Cs;background
Cw;background
 fwater

(7)

Throughout the study, the log DOW was used instead of the log KOW considering that some target compounds are charged at the ambient pH of 6.8: For acidic compounds the log DOW is defined as log DOW ¼ log KOW þ log

1 1 þ 10pH
pKa

(8)

and for basic compounds as: log DOW ¼ log KOW þ log

1 1 þ 10pKa
pH

3.

Results and discussion

3.1.

Sorption kinetics

3.1.1.

Sorption equilibrium and mass balance

3643

Fig. 1 shows exemplarily the sorption kinetics of the biocides irgarol, thiabendazole and 1,2-benzisothiazolin-3-one (BIT) to secondary sludge with and without adding NaN3 by comparing the ratio of the measured sorbed, soluble and total concentration to the spike concentration of 10 mg L
1. The total analyte concentration is calculated as the sum of the soluble and sorbed concentrations determined according to eq. (6) and (7), respectively. The sorption kinetics of all other examined analytes are shown in the Appendix (Fig. A1). At least in the NaN3 amended samples the sorbed concentration was constant after 1.5 h, indicating that sorption equilibrium was reached except for BIT and 2-n-octyl-4-isothaizolin-3-one (OIT) being subject of significant biodegradation. This is consistent with the results from other studies determining also a fast sorption kinetic for sludge with sorption equilibrium times of less than 2 h (Andersen et al., 2005; Ternes et al., 2004). For 18 of 27 analytes the mass balance in sludge amended samples as well as in filtered control samples without sludge was in an acceptable range of 100  30% indicating a good analytical method performance as well as no significant losses by degradation and volatilization. A significantly lower mass balance after 24 h was observed for the partly (w50%) positively charged analytes tridemorph, fenpropimorph and imazalil as well as the strong sorbing analytes triclosan and triclocarban in the control samples and indicated sorption to the glass vessels. However, due to a higher sorption affinity to secondary sludge, the mass balances of these analytes were also within the acceptable range of 100  30% in the sludge samples. A significant degradation of more than 30% in the sludge samples without NaN3 within 24 h was observed for terbuthylazine, fenpropimorph and tridemorph, of more than 60% for chlorophene, benzophenone-3 (BZP-3) and benzophenone-4 (BZP-4) and of more than 90% for BIT, OIT, benzophenone-1 (BZP-1) and benzophenone-2 (BZP-2) (Table 2). Except for terbuthylazine, dissipation was significantly reduced by addition of NaN3 and can therefore mainly be attributed to biological degradation (Fig. A1). The inhibition of the degradation of BIT and OIT was insufficient for reaching a sorption equilibrium even at a concentration of 1% NaN3. An incomplete inhibition of the microbial activity by NaN3 has also been observed by Ramil et al. (2010) who investigated the sorption of betablockers in soils. However, the authors did not investigate whether the addition of NaN3 also has an influence on the sludge-water distribution.

(9)

However, except for the basic compounds imazalil, fenpropimorph and tridemorph (w50% positively charged) and the acidic compounds mecoprop, BZP-4 and PBSA (w100% negatively charged), the non-charged species dominate at the ambient pH of 6.8 for all target compounds and thus the log DOW equals the log KOW.

3.1.2.

Influence of NaN3 on the sludge-water distribution

For 18 out of 27 analytes the addition of NaN3 had a distinct influence on the sludge-water distribution (Table 2). While for the triazines, such as irgarol (Fig. 1), the ratio of sorbed and soluble concentration was comparable with different NaN3 concentrations, the sorbed concentration was lower for most analytes including thiabendazole (Fig. 1) and diuron (Fig. A1).

3644 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 3 8 e3 6 5 2

Fig. 1 e Ratio [%] of sorbed, soluble and the total concentration to the spike concentration of 10 mg LL1 over time for three selected biocides with and without addition of NaN3 for microbial activity inhibition. The background concentrations as well as the water content of the sludge were assessed and considered for the calculation.

3645

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 3 8 e3 6 5 2

Table 2 e Results of the sorption equilibrium experiments for biocides and UV-filters. The recovery (Rec.) [%] of the spiked compound concentration after 1.5, 3.2 and 6 h of incubation without addition of NaN3 is defined as the ratio of the measured total concentration, i.e. the sum of the sorbed and soluble concentration Cs and Cw, and the spiked concentration according to eq. (5). The average Kd,sec [L kgdw sludgeL1] was calculated from the sludge-water distribution coefficients determined after 1.5, 3.2 and 6 h of incubation. The range indicates the 95% confidence interval. w/o: without; ND: not determined. Analyte

1.5 h

3.2 h

6h

Average

Mass Kd,sec Mass Kd,sec Mass Kd,sec balance [L kgTSS
1] balance [L kgTSS
1] balance [L kgTSS
1] [%] [%] [%] Phenylurea herbicides Diuron 105 Isoproturon 112 Triazines Terbuthylazine 100 Terbutryn 113 Irgarol 104 97 M1c Biocides Mecoprop 96 Propiconazole 135 Tebuconazole 145 Imazalil 141 Carbendazim 136 Thiabendazole 142

Kd,sec [L kgTSS
1]

Influence of NaN3b

Degree of dissipation w/o NaN3a

38 19

105 122

44 10

115 109

32 15

38  7 15  5

o o

lower Cs lower Cs

63 170 150 65

94 106 98 86

43 160 140 59

90 97 90 90

44 150 140 58

50  13 160  10 140  10 60  4

þ o o o

no no no no

15 440 480 3000 25 190

103 109 134 128 117 126

10 330 380 2800 16 210

102 110 138 137 129 133

12 340 580 3200 11 220

12  3 370  70 480  110 3000  200 20  5 210  20

o o o o o o

lower Cs lower Cs and Rec. lower Cs and Rec. lower Cs and Rec. lower Cs lower Cs and higher Cw / Rec. const. higher Cw / Rec. >>100% lower Cs and Rec. lower Cs and Rec. ND ND no no lower Cs and Rec. lower Cs and Rec. lower Cs and Rec.

Dimethomorph

105

72

105

53

98

56

60  12

o

Fenpropimorph Tridemorph BITd OITe DMSTf DMSAg Triclosan Triclocarban Chlorophene UV-filters BZP-1h BZP-2h BZP-3h BZP-4h PBSAi

136 121 10 <5 111 119 166 136 86

1800 24,000 ND ND 28 22 16,000 36,300 2700

103 83 5 <5 134 106 119 126 61

1400 14,000 ND ND 7 14 16,000 39,500 1600

91 72 <5 <5 122 116 119 125 53

1400 18,000 ND ND 17 17 17,000 43,500 1700

1500  300 19,000  6000 ND ND 17  12 18  5 16,000  1000 40,000  4000 2000  700

þ þ þþþ þþþ o o o o þþ

86 133 161 95 103

260 1100 1500 12 7

60 85 137 83 80

320 1300 1200 6 11

45 80 115 83 86

190 1400 1200 7 18

260  70 1300  200 1300  200 83 12  6

þþþ þþþ þþ þþ o

lower Cs and Rec. lower Cs and Rec. lower Cs and Rec. no variable Cw and Rec.

a o: <30%, >30%: þ, >60%: þþ, >90%: þþþ b Details about the influence of NaN3 on the mass balance and sludge-water distribution are provided in Fig. 1 and Fig. A1 in the Appendix. c 2-Methylthio-4-tert-butylamino-6-amino-s-triazine. d 1,2-Benzisothiazolin-3-one. e 2-n-Octyl-4-isothaizolin-3-one. f N,N-Dimethyl-N0 -p-tolylsulfamide. g N,N-Dimethyl-N0 -phenylsulfamide. h BZP: Benzophenone. i Phenylbenzimidazole sulfonic acid.

Similar results of the influence of NaN3 on the sorption of diuron to soil were also reported by Gaillardon (1996). An inhibitory effect of NaN3 on the sorption might be due to physicochemical interactions, such as competition or change of the microbial surface caused by increased cell lysis. However, for analytes such as imazalil and triclosan, the soluble concentration did not increase parallel to the decrease of the sorbed concentration leading to a lower mass balance in the NaN3 amended assays (Fig. A1). Even though stable

isotope-labeled surrogate standards have been used for analysis, this might be due the influence of NaN3 on the extraction efficiency or matrix effects during ionization, which have shown to have a significant influence on the recoveries of the target analytes (Wick et al., 2010). Similarly, the severe overestimation of the soluble concentration of dimethomorph leading to a mass balance of up to 700% in the samples amended with 1% NaN3, might be explained by matrix effects, i.e. strong ion enhancement in the ESI interface

log DOW

Phenylurea herbicides Diuron Isoproturon Fluometuron Chloroxuron Linuron

pKa

Recovery [%]

R2

Kf [mg1
n Ln kg
1]

n

Kd,sec [L kgdw sludge
1] at 1 mg L
1

log KOC

log KOC,LIT

PredictedWWTP removal [%]

2.8e 2.5e 2.2e 3.4f 3.0e

e e e e e

91 96 90 83 95

8 9 16 19 11

0.991 0.994 0.999 0.989 0.999

0.85 0.95 0.90 0.87 0.97

    

0.11 0.22 0.03 0.13 0.03

48  11 10  5 39  2 240  70 73  5

48 9 39 270 74

2.2 1.5 2.1 3.0 2.4

Monolinuron Neburon Monuron Methabenzthiazuron Triazines Atrazine Simazine Cyanazine Propazine Terbuthylazine Terbutryn Irgarol M1 Ametryn Prometryn Prometon Biocides Mecoprop Propiconazole

2.2e 3.8f 1.8f 2.6f

e e e e

100  7 96  10 97  7 101  8

0.996 0.999 0.999 0.999

1.01 0.89 0.96 1.03

   

0.12 0.05 0.05 0.03

82 290  30 17  1 75  5

8 320 17 76

1.4 3.0 1.8 2.4

2.6h 2.1i 1.5e2.3, 3.2h 2.0e2.9, 2.4k 1.4e3.3, 3.3e3.6, 1.4e3.3, 2.7f

2.5f 2.1f 2.1e 2.9g 3.0e 3.7e 3.7g e 2.6e 3.1e 2.9f

1.7e 1.6e 12.9e 1.7e 2.0e 4.3e 4.1f e 10.1e 10.0e 9.7e

95  9 103  7 103  20 101  7 95  8 73  8 88  2 106  7 114  13 117  19 107  18

0.994 0.999 0.867 0.999 0.996 0.968 0.999 0.999 0.999 0.995 0.999

1.21  1.14  1.7d 1.13  0.98  0.63  0.95  0.96  1.06  1.01  1.24 

0.28 0.06 0.07 0.09 0.16 0.04 0.08 0.03 0.10 0.08

53 61 35d 18  2 42  8 190  60 130  10 53  9 32  2 65  13 12  2

5 6 ND 18 42 230 130 53 33 65 (1.5) 13

1.2 1.3 ND 1.8 2.2 2.9 2.7 2.3 2.1 2.4 1.7

2.2f, 1.7k 2.1f 1.6e2.4, B2.3j 2.2f 2.2e2.5f 2.9h 3.0l e 2.6h 2.9h 2.5h

<5 <5 <5 <5 <5 <5e10 <5e5 <5 <5 <5 <5


1.0g 3.7e

3.1e 1.1e

105  10 92  6

NDa 0.999

NDa 0.90  0.04

NDa 350  30

NDa 380

NDa 3.1

<5 <5e15

Tebuconazole Imazalil Carbendazim

3.7e 3.7e 1.5e

e 6.5e 4.2e

106  10 89  5 89  9

0.999 0.990 0.995

0.89  0.04 0.85  0.12 0.73  0.07

390  30 2200  600 24  3

420 3300 23

3.2 4.1 1.9

Thiabendazole Dimethomorph Fenpropimorph Tridemorph BIT OIT DMST DMSA

2.4f 2.7e 4.1e 6.9g 1.2g 2.5e e e

4.7; 12.0e e 7.0 e 6.5e e e e e

105  10 98  10 78  5 43  21 NDc NDc 92  13 125  18

0.990 0.970 0.993 NDb NDc NDc 0.978 NDa

0.86  0.94  0.99  NDb NDc NDc 1.08  NDa

180  50 46  22 1800  500 NDb NDc NDc 98 NDa

190 44 1800 NDb NDc NDc 10 NDa

2.8 2.2 3.8 NDb NDc NDc 1.5 NDa

1.1e1.4f 2.6e3.1, B2.8j log Kd ¼ 2.9m 2.8i 3.5e3.9, B3.6j 2.3e2.9, B2.5j log Kd < 2.7n 3.0e3.5, B3.4j 2.5f 2.9e3.7e 3.4e4.0e e e e e

    

0.49

B2.7j B2.3j B3.4j B2.2j, 1.5k

<5 <5 <5 <5e10 <5 <5 <5e10 <5 <5

<5e15 25e55 <5 <5e10 <5 15e40 NDd NDe NDe <5 <5

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 3 8 e3 6 5 2

0.12 0.23 0.12

B2.7j

3646

Table 3 e Summary of the results of the Freundlich isotherm batch experiments conducted with secondary sludge from the nitrification zone of a biological treatment (TSS: 4 gTSS LL1, TOC: 29%, pH 6.8) using 6 different concentration levels (100 ng LL1 to 30 mg LL1). The Kd,sec values were calculated from the Freundlich parameters and the measured soluble concentration Cw for the 1 mg LL1 spiking level according to eq. (10) (actual total concentrations significantly higher than 1 mg LL1 due to the background concentrations are noted in parenthesis). The WWTP removal [%] by sorption was predicted for the same concentration level and a sludge production of 0.1e0.4 gTSS LL1 according to eq. (4).

Triclosan Triclocarban Chlorophene UV-filters BZP-1 BZP-2 BZP-3 BZP-4 PBSA

4.8g 5.1g 4.1g

8.0b e 9.6b

92  44 95  16 90  28

0.961 0.996 0.988

0.55  0.16 0.83  0.10 0.66  0.10

6400  1400 19,000  7000 2200  300

12,000 (10.4) 40,000 (1.6) 3800 (2.8)

4.6 5.1 4.1

4.3o, 4.7(p,q) 3.8o, 4.7r e

55e85 80e95 30e60

2.9g 2.1g 3.8g
5.9g
1.1g

8.0b 8.0b 8.0b 0.7b 4.9; 0.7b

40  42 55  20 96  19 104 (n ¼ 2) 127  39

0.990 0.996 0.953 NDa 0.892

0.82  0.83  0.88  NDa 0.76 

170  50 230  40 670  260 NDa 14  12

260 (0.2) 300 (0.4) 720 (1.5) NDa 9 (6.1)

3.0 3.0 3.4 NDa 1.5

e e e e e

<5e10 <5e10 5e20 <5 <5

0.37

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 3 8 e3 6 5 2

a Sorption too low for determination of Freundlich isotherms. b Sorption too high for determination of Freundlich isotherms. c Biological degradation too fast for determination of Freundlich isotherms. d The sign “” indicates that the lower limit was below experimental resolution. e Tomlin, 1994. f IUPAC, Pesticide Properties Database, 2010. g VCCLAB, Virtual Computational Chemistry Laboratory, 2005. h Liu and Qian, 1995 (soil). i Berenzen et al., 2005 (soil). j ARS Pesticide Properties Database, 2009 (soil). k Ko¨rdel et al., 1997 (sludge). l Tolosa et al., 1996 (sediment). m Kahle et al., 2008 (sludge). n Kupper et al., 2006 (sludge). o Agyin-Birikorang et al., 2010 (soil, biosolids). p Singer et al., 2002 (sludge). q Orvos et al., 2002 (sludge). r TCC Consortium, 2002 (sludge).

0.12 0.07 0.27

3647

3648

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 3 8 e3 6 5 2

(Fig. A1). For dimethomorph no stable isotope-labeled surrogate was available for analysis. Further investigations are needed to fully understand the mode of action of NaN3 on the sorption kinetics and chemical analysis, but the results confirmed that NaN3 can have a negative impact on the reliability of sorption experiments. However, except for the fast degrading isothiazolones BIT and OIT, the distribution of the analytes between sludge and water phase, i.e. the Kd,sec values, did not significantly change in the period from 1.5 to 6 h of incubation without using NaN3 (Table 2). Thus, the dissipation by biological degradation was considerably slower than the sorption kinetics and reliable results of sludge-water distribution were obtained within this time range. However, analyzing both the sorbed and soluble concentration was crucial to avoid an overestimation of the sludge-water distribution due to a faster dissipation of the soluble amount by biological degradation. If sorption equilibrium is not reached due to rapid biodegradation as observed for BIT and OIT, alternative methods for inactivation of sludge, such as autoclavation or freeze-drying, have to be applied to determine sludge-water distribution coefficients. However, for these rapidly biodegradable compounds the influence of sorption on their overall removal in WWTPs can be expected to be of minor importance.

3.2.

Freundlich isotherms

Isotherm experiments were performed for all analytes for which the sorption kinetics were examined except for BIT and OIT, as for these two isothiazolinones sorption equilibria were not attained due to rapid biodegradation. Moreover, seven additional triazines and seven phenylurea herbicides were included in the isotherm studies. Based on the results from the sorption kinetics, sorption isotherm experiments were conducted without adding NaN3 using an equilibrium time of 1.5 h. Freundlich isotherms were determined by linear regression of the logarithmized sorbed and soluble concentrations for all compounds for which the concentrations were above the limit of quantification for more than 4 spiking levels. This criteria was fulfilled for all analytes except BZP-4, mecoprop and DMSA (sorption too low) and tridemorph (sorption too high) and the Freundlich isotherms are shown in Fig. A2 in the appendix. For most compounds the spiked amount was recovered by more than 75% (Table 3), even though no NaN3 was used for sludge deactivation. Lower recoveries were only determined for terbutryn (73  8%), tridemorph (43  21%), BZP-1 (40  42%) and BZP-2 (55  20%). However, the incomplete mass balance of these compounds is not supposed to significantly influence the sludge-water distribution, since analyte concentrations were determined both in the aqueous and solid phase. In addition, the results from the sorption kinetics showed no significantly influence of biological degradation on the sludge-water distribution coefficients of these compounds within 6 h of incubation (Table 2).

3.2.1.

Concentration dependence

Except for the triazine cyanazine and the UV-filter PBSA, the correlation of determination (R2) was >0.95 and for most analytes even >0.99, indicating a good fit of the Freundlich

model. The Freundlich exponents n and the Freundlich affinity constants Kf together with the corresponding 95% confidence intervals are provided in Table 3. For 14 out of 34 analytes the Freundlich exponents n were significantly smaller than 1, indicating a non-linear sorption to secondary sludge. This implies a decreased sorption affinity to the sludge flocs with an increasing aqueous phase concentration of these compounds. Therefore, Kd,sec values from the linear model determined at relatively high spiking concentrations may underestimate the sorption affinity at lower environmentally relevant concentrations. The effect of the concentration level on the Kd,sec values of analytes with different sorption affinities is illustrated for triclosan, ternutryn and imazalil in Table 4. The Kd,sec values shown in Table 4 were calculated from the measured soluble concentration Cw combining eq. (1) and (2) to: Kd;sec ¼ Kf  Cn
1 w

(10)

It follows from eq. (10) that the effect of a lower Freundlich exponent is more pronounced for analytes with a lower sorption affinity, i.e. a broader soluble concentration range. Consequently, the calculated Kd,sec value of terbutryn is about seven times higher at a spiking level of 100 ng L
1 compared to 30,000 ng L
1, whereas it is less than three times higher for triclosan which has a similar Freundlich exponent. Therefore, possible concentration effects should be taken into account when comparing sludge-water distribution coefficients. However, considering that the WWTP influent concentrations are typically spanning less than two orders of magnitude, the prediction of the removal efficiencies in WWTPs by sorption to excess sludge according to eq. (4) reveals that the bias in the predicted removal efficiency due to non-linear sorption effects is relatively low (Table 4).

3.2.2.

Sludge-water distribution coefficients

For the comparison of the sludge-water distribution, the Kd,sec values shown in Table 3 were calculated for the 1 mg L
1 spiking level based on eq. (10) to consider the non-linear sorption observed for some analytes. The calculated Kd,sec values were in the same order of magnitude as those determined from the sorption kinetics, except for BZP-2 and BZP-3 (Table 2). In accordance with their high log DOW values of 5.1, 4.8 and 4.1, highest distribution coefficients of 40,000, 12,000 and 3800 L kgdw sludge
1 were calculated for the bactericides triclocarban, triclosan and chlorophene, respectively. In addition, values above 500 L kgdw sludge
1 were determined for the partly positively charged (w50%) biocides imazalil (3300 L kgdw sludge
1) and fenpropimorph (1800 L kgdw sludge
1) and the UV-filter BZP-3 (720 L kgdw sludge
1). According to eq. (4), sorption to sludge can be expected to significantly contribute to the removal of these analytes in WWTPs (>10%). Relatively high sorption removal efficiencies of 55e85% and 80e95% predicted for triclosan and triclocarban, respectively, are in accordance with their measured mass balance in fullscale plants (Heidler et al., 2006; Heidler and Halden, 2007). However, for most of the target analytes, the distribution coefficients were below 500 L kgdw sludge
1 and the removal by sorption to secondary sludge in WWTPs can be expected to be negligible (<10%).

3649

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 3 8 e3 6 5 2

Table 4 e Influence of the compound concentration on the sludge-water distribution coefficients Kd,sec [L kgdw sludgeL1] and the predicted removal by activated sludge treatment for three compounds with Freundlich exponents <1. The removal was predicted according to eq. (4) for a sludge production range of 0.1e0.4 gTSS LL1 typical for municipal WWTPs. Spike Conc.

100 300 1000 3000 10,000 30,000

Triclosan (n ¼ 0.55  0.16)

Terbutryn (n ¼ 0.63  0.16)

Kd,sec [L kgdw sludge
1]

Predicted removal range [%]

Kd,sec [L kgdw sludge
1]

Predicted removal range [%]

Kd,sec [L kgdw sludge
1]

Predicted removal range [%]

15,000 14,700 12,400 12,300 8400 5800

60e86 60e86 55e83 55e83 46e77 37e70

480 370 230 160 100 70

5e16 4e13 2e9 2e6 1e4 1e3

5000 3700 3300 2800 2300 1800

32e65 27e60 25e57 22e52 19e48 15e42

The logarithmized organic carbon based distribution coefficients (log KOC), which were determined within the current study for secondary sludge, were mainly within 0.5 log units of those available in literature for soil and sludge. Considering that systematic studies with different soils revealed already standard deviations of 0.3 log units (Schwarzenbach et al., 2003), the results are in good agreement with literature data. However, especially for analytes with a lower sorption affinity (log DOW < 3), the log KOC values determined within the current study were generally in the lower range of those reported in literature for soils. In contrast, the log KOC values of the strongly sorbing compounds such as triclosan, triclocarban, imazalil and fenpropimorph were slightly higher.

3.2.3.

Imazalil (n ¼ 0.85  0.12)

Correlation of log DOW and log KOC

Since the neutral species dominated under the conditions of the batch system, the hydrophobicity of the target analytes was expected to have the strongest influence on their sorption

affinity to secondary sludge, except for the partly (w50%) positively charged compounds imazalil, fenpropimorph and tridemorph and the negatively charged acids mecoprop, PBSA and BZP-4. Therefore, the log DOW as a measure of the compound hydrophobicity was correlated with the log KOC for all compounds for which log KOC values could be determined within the current study, except for the negatively charged UV-filter PBSA (Fig. 2a). This single parameter linear freeenergy relationship (sp-LFER) approach has been widely used to estimate the sorption affinity of certain groups of noncharged pollutants toward soils (Schwarzenbach et al., 2003). However, studies using this approach for sludge are rare (Carballa et al., 2008; Heidler and Halden, 2008; Kim et al., 2009; Urase and Kikuta, 2005). The linear regression in Fig. 2a revealed a best fit linear model for the target analytes of log KOC ¼ 0.958  log DOW - 0.290 featuring a R2 of 0.71 and a root mean square error (RMSE) of 0.51. This confirms a general linear dependency of log DOW and log KOC. However, the variability of the log KOC values of up to one log unit indicates

Fig. 2 e Log KOC values determined in secondary sludge versus log DOW values for (a) 31 target compounds of the current study and (b) 17 compounds different from the target compounds for which KOC or Kd values for sludge were available from literature. The removal ranges were calculated according to eq. (4) assuming a sludge production (SP) in the range of 0.1e0.4 gTSS LL1. (1) Carbamazepine, (2) Estradiol, (3) Estrone, (4) Ethinylestradiol, (5) Tonalide, (6) Galaxolide (aCarballa et al., 2008, digested sludge); (7) Methyl benzoate, (8) 2,5-Dichloroaniline, (9) Phenanthrene (10) Fenthion (bKo¨rdel et al., 1997, freeze-dried sludge); (11) Clarithromycin, (12) Azithromycin (cGo¨bel et al., 2005, estimated from apparent Kd for full-scale samples); (13) Octylphenol, (14) Nonylphenol (dIsobe et al., 2001); (15) Clotrimazole (eKahle et al., 2008, estimated from apparent Kd for full-scale samples); (16) Bisphenol A (fZhao et al., 2008, autoclaved sludge); (17) Loratidine (gRadjenovic et al., 2009, estimated from apparent Kd for full-scale samples).

3650

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 3 8 e3 6 5 2

the influence of further specific sorption mechanisms for certain analytes. For instance, the log KOC of imazalil was significantly higher than of other analytes with comparable log DOW values. This might be attributed to the fact that approximately 50% of the imazalil is positively charged at pH 7 and thus electrostatic interaction with the mainly negatively charged sludge flocs significantly contributed to the overall sorption affinity. In contrast, even though fenpropimorph has a similar pKa as imazalil but a significantly higher log DOW, its log KOC was in the same range as determined for imazalil. Thus, the different sorption behaviors of these two analytes cannot be sufficiently described by the pKa and the log DOW. Further compound specific sorption mechanisms can also be assumed to determine the sorption behavior of atrazine and the UV-filters BZP-1 and BZP-2. Whereas for atrazine with a log DOW of 2.5, the log KOC of 1.2 was unexpectedly low, the log KOC value of 3.0 measured for BZP-1 and BZP-2 was higher than estimated from their log DOW values of 2.9 and 2.1, respectively. In general, a higher variability of the log KOC was observed for compounds with lower log DOW values (<3). This is in accordance with the theory that for polar compounds, capable of H-bonding, various types of interactions can occur which cannot be sufficiently described by only one parameter, i.e. the log DOW (Faria and Young, 2010). The application of multiparameter models with descriptors for both sludge and micropollutant characteristics might be a valuable approach for a more accurate prediction of sludge-water distribution of nonionic compounds. The use of such a multiparameter model to predict the sorption of hydrophobic micropollutants (PAHs and PCBs) to sludge particles by Barret et al. (2010) revealed the importance of several sludge predictors such as the protein content and the mineral density and micropollutant predictors such as the molar mass, the number of chlorines and the log DOW. However, these approaches require an extensive sludge and molecule characterization which was not within the scope of the current study and will need further attention regarding the suitability of these approaches for emerging micropollutants of higher polarity. To check whether the results for the dataset of the current study can be transferred to other compounds being noncharged at pH 7, a correlation of the log DOW, i.e. the log KOW, with the log KOC was also performed with a dataset of 17 structurally diverse compounds with a log DOW between 2 and 6 for which Kd or KOC data were available from literature (Fig. 2b). The comparison of the best fit linear model for this dataset from literature (RSME ¼ 0.50) with that from our dataset (RSME ¼ 0.61) revealed that the latter also performed reasonably well for other compounds and sludge types which were not examined within this study. The results demonstrate that single parameter LFER approaches based on the log DOW can be applied for an estimation of the sorption affinity of non-charged micropollutants to sludge. The DOW prediction approach allows for a classification of non-charged compounds regarding their full-scale removal via withdrawal of excess sludge when other removal processes such as volatilization and biodegradation can be neglected (Fig. 3). Removal is expected to be negligible for nonionic compounds with log DOW < 3.5, whereas, depending on the sludge production, removal by <10e80%, 50e95% and

Fig. 3 e Predicted removal of nonionic compounds via withdrawal of excess sludge versus log DOW for a sludge production of 0.1 and 0.4 gTSS LL1. The removal was predicted using eq. (4) assuming a correlation of log DOW and log KOC according to the best fit linear model determined for the target analytes of the current study (log KOC [ 0.958 3 log DOW L 0.290; fOC [ 0.29 gOC gdw sludgeL1).

>90% can be expected for log DOW values in the range of 3.5e5, 5 to 6 and >6, respectively. For compounds being charged at the ambient pH of the activated sludge treatment (pH 6e8), the sorption affinity and thus the removal by sorption in WWTPs strongly depend on their specific charge and are currently not reasonably predictable (Ternes et al., 2004; Urase and Kikuta, 2005). As observed within this study for imazalil, positively charged compounds tend to have stronger sorption affinities than expected based on their log DOW values due to electrostatic interactions with the mainly negatively charged surfaces of the sludge flocs. Consequently, negatively charged compounds are known to partition predominantly in the aqueous phase due to electrostatic repulsion (Ternes and Joss, 2007). For instance, it has to be considered that the WWTP removal by sorption might be overestimated for triclosan and the UVfilters BZP-1, BZP-2 and BZP-3 using distribution coefficients determined at neutral pH, since their pKa value of 8 is within the range typical for activated sludge (6e8). At pH 7 the noncharged species dominate (90%), whereas the negatively charged species increase to 50% at a sludge pH of 8 and thus most likely reduce the sorption affinity.

4.

Conclusions

Within the present study the sorption of 41 micropollutants to activated sludge was investigated. For 27 contaminants, sludge-water distribution coefficients and their concentration dependency was determined for the first time. Both are major prerequisites for assessing their fate in WWTPs and thus their potential release into the environment via treated wastewater or via sludge amendment to agricultural land. Furthermore, the results of the present study imply important aspects for performing sorption experiments with fresh activated sludge. NaN3 has been shown to be

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 3 8 e3 6 5 2

inappropriate for the deactivation of fresh sludge due to (i) insufficient inhibition of microbial degradation and (ii) its influence on the sludge-water distribution. To the best of our knowledge, this is the first study confirming that increasing NaN3 concentrations can lead to decreasing compound amounts sorbed to activated sludge. In addition, the results show that the deactivation of sludge is not a prerequisite if biodegradation does not hinder the attainment of sorption equilibria. However, quantification of the analytes in both sludge and aqueous phase is crucial if biodegradation cannot be excluded, since analyzing only the dissolved concentration would lead to increased distribution coefficients. The Freundlich model was shown to be appropriate to describe the sorption behavior of various micropollutants in contact to secondary sludge. Since for many compounds the Freundlich exponents were close to 1 and thus Kf z Kd, the linear model can be used in most cases for a rapid determination of sludge-water distribution coefficients. For the linear approach concentrations should be as close as possible to real concentrations in full-scale WWTPs. For compounds being non-charged in the pH range typical for activated sludge treatment (pH 6e8), sludge-water distribution can be reasonable predicted using a simple LFER approach based on the log KOW as a single descriptor. Even though this approach does not consider specific molecular interactions, which are especially important for polar compounds, it is a simple practical tool for a first assessment of the removal of nonionic compounds via withdrawal of excess sludge in WWTPs. Accordingly, nonionic compounds can be classified as not significantly removable (<10%) by sorption for log KOW values <3.5, partly removable (10e95%) for log DOW values of 3.5e6 and almost completely removable (>90%) for log KOW values > 6.

Acknowledgments This study formed part of the EU project NEPTUNE (Project no. 036845). It was financially supported by grants obtained from the EU Commission within the energy, environment and sustainable development program of the 6th framework.

Appendix. Supplementary data Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.04.014.

references

Agyin-Birikorang, S., Miller, M., O’Connor, G.A., 2010. Retentionrelease characteristics of triclocarban and triclosan in biosolids, soils, and biosolids-amended soils. Environ. Toxicol. Chem. 29 (9), 1925e1933. Andersen, H.R., Hansen, M., Kjølholt, J., Stuer-Lauridsen, F., Ternes, T.A., Halling-Sørensen, B., 2005. Assessment of the importance of sorption for steroid estrogens removal during activated sludge treatment. Chemosphere 61, 139e146.

3651

ARS Pesticide Properties Database, 2009. Agricultural Research Service, United States Department of Agriculture (USDA). http:// www.ars.usda.gov/services/docs.htm?docid¼14199 (08/10). Barret, M., Carre`re, H., Latrille, E., Wisniewski, C., Patreau, D., 2010. Micropollutant and sludge characterization for modeling sorption equilibria. Environ. Sci. Technol. 44, 1100e1106. Berenzen, N., Lentzen-Godding, A., Probst, M., Schulz, H., Schulz, R., Liess, M., 2005. A comparison of predicted and measured levels of run-off related pesticide concentrations in small lowland streams on a landscape level. Chemosphere 58, 683e691. Berns, A.E., Philipp, H., Narres, H.-D., Burauel, P., Vereecken, H., Tappe, W., 2008. Effect of gamma-sterilization and autoclaving on soil organic matter structure as studied by solid state NMR, UV and fluorescence spectroscopy. Eur. J. Soil Sci. 59, 540e550. Carballa, M., Fink, G., Omil, F., Lema, J.M., Ternes, T.A., 2008. Determination of the solid-water distribution coefficient (Kd) for pharmaceuticals, estrogens and musk fragrances in digested sludge. Water Res. 42, 287e295. Chefetz, B., Stimler, K., Shechter, M., Drori, Y., 2006. Interactions of sodium azide with triazine herbicides: effect on sorption to soils. Chemosphere 65, 352e357. Clara, M., Strenn, B., Saracevic, E., Kreuzinger, N., 2004. Adsorption of bisphenol-A, 17b-estradiole and 17aethinylestradiole to sewage sludge. Chemosphere 56, 843e851. Dobbs, R.A., Shan, Y., Wang, L., Govind, R., 1995. Sorption on wastewater solids: elimination of biological activity. Water Environ. Res. 67, 327e329. Faria, I.R., Young, T.M., 2010. Comparing linear free energy relationships for organic chemicals in soils: effects of soil and solute properties. Environ. Sci. Technol. 44, 6971e6977. Go¨bel, A., Thomasen, A., McArdell, C.S., Joss, A., Giger, W., 2005. Occurrence and sorption behavior of sulfonamides, macrolides, and trimethoprim in activated sludge treatment. Environ. Sci. Technol. 39, 3981e3989. Gaillardon, P., 1996. Influence of soil moisture on long-term sorption of diuron and isoproturon. Pestic. Sci. 47, 347e354. Heidler, J., Halden, R.U., 2007. Mass balance assessment of triclosan removal during conventional sewage treatment. Chemosphere 66, 362e369. Heidler, J., Halden, R.U., 2008. Meta-analysis of mass balances examining chemical fate during wastewater treatment. Environ. Sci. Technol. 42 (17), 6324e6332. Heidler, J., Sapkota, A., Halden, R.U., 2006. Partitioning, persistence, and accumulation in digested sludge of the topical antiseptic triclocarban during wastewater treatment. Environ. Sci. Technol. 40, 3634e3639. Isobe, T., Nishiyama, H., Nakashima, A., Takada, H., 2001. Distribution and behavior of nonylphenol, octylphenol, and nonylphenol monoethoxylate in Tokyo metropolitan area: their association with aquatic particles and sedimentary distributions. Environ. Sci. Technol. 35, 1041e1049. IUPAC, 2010. Pesticide Properties Database. http://sitem.herts.ac. uk/aeru/iupac/index.htm (08/10). Ko¨rdel, W., Hennecke, D., Franke, C., 1997. Determination of the adsorption-coefficients of organic substances on sewage sludges. Chemosphere 35 (1/2), 107e119. Kahle, M., Buerge, I.J., Hauser, A., Mu¨ller, M.D., Poiger, T., 2008. Azole fungicides: occurrence and fate in wastewater and surface waters. Environ. Sci. Technol. 42, 7193e7200. Kim, H.-J., Lee, H.-J., Lee, D.S., Kwon, J.-H., 2009. Modeling the fate of priority pharmaceuticals in Korea in a conventional sewage treatment plant. Environ. Eng. Res. 14 (3), 186e194. Kupper, T., Plagellat, C., Bra¨ndli, R.C., de Alencastro, L.F., Grandjean, D., Tarradellas, J., 2006. Fate and removal of polycyclic musks, UV filters and biocides during wastewater treatment. Water Res. 40, 2603e2612.

3652

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 3 8 e3 6 5 2

Langford, K.H., Scrimshaw, M.D., Birkett, J.W., Lester, J.N., 2005. The partitioning of alkylphenolic surfactants and polybrominated diphenyl ether flame retardants in activated sludge batch tests. Chemosphere 61, 1221e1230. Liu, J., Qian, C., 1995. Hydrophobic coefficients of S-triazines and phenylurea herbicides. Chemosphere 31 (8), 3951e3959. Maurer, M., Escher, B.I., Richle, P., Schaffner, C., Alder, A.C., 2007. Elimination of b-blockers in sewage treatment plants. Water Res. 41, 1614e1622. Ochoa-Herrera, V., Sierra-Alvarez, R., 2008. Removal of perfluorinated surfactants by sorption onto granular activated carbon, zeolite and sludge. Chemosphere 72, 1588e1593. OECD Guideline 106, 2000. Guideline for the Testing of Chemicals 106: Adsorption/Desorption Using a Batch Equilibrium Method. Organisation for Economic Cooperation and Development (OECD). Orvos, D.R., Versteeg, D.J., Inauen, J., Capdevielle, M., Rothenstein, A., Cunningham, V., 2002. Aquatic toxicity of triclosan. Environ. Toxicol. Chem. 21 (7), 1338e1349. , J., Petrovic , M., Barcelo´, D., 2009. Fate and distribution Radjenovic of pharmaceuticals in wastewater and sewage sludge of the conventional activated sludge (CAS) and advanced membrane (MBR) bioreactor (MBR) treatment. Water Res. 43, 831e841. Ramil, M., El Aref, T., Fink, G., Scheurer, M., Ternes, T.A., 2010. Fate of beta blockers in aquatic-sediment systems: sorption and biotransformation. Environ. Sci. Technol. 44, 962e970. Richardson, S.D., 2009. Water analysis: emerging contaminants and current issues. Anal. Chem. 81, 4645e4677. Schwarzenbach, R.P., Gschwend, P.M., Imboden, D.M., 2003. Environmental Organic Chemistry. Wiley-Interscience, Hoboken, NJ, USA. Singer, H., Mu¨ller, S., Tixier, C., Pillonel, L., 2002. Triclosan: occurrence and fate of a widely used biocide in the aquatic environment: field measurements in wastewater treatment plants, surface waters, and lake sediments. Environ. Sci. Technol. 36, 4998e5004. Stein, K., Ramil, M., Fink, G., Sander, M., Ternes, T.A., 2008. Analysis and sorption of psychoactive drugs onto sediment. Environ. Sci. Technol. 42, 6415e6423.

TCC Consortium, 2002. U.S. High Production Volume (HPV) Chemical Challenge Program Data Availability and Screening Level Assessment for Triclocarban. Report 201-14186A. http:// www.epa.gov/HPV/pubs/summaries/tricloca/c14186.pdf (10. 15.10.). Ternes, T.A., Joss, A. (Eds.), 2007. Human Pharmaceuticals, Hormones and Fragrances: The Challenge of Micropollutants in Urban Water Management. IWA Publishing, London, UK. Ternes, T.A., Herrmann, N., Bonerz, M., Knacker, T., Siegrist, H., Joss, A., 2004. A rapid method to measure the solid-water distribution coefficient (Kd) for pharmaceuticals and musk fragrances in sewage sludge. Water Res. 38, 4075e4084. Thomas, S.M., Bodour, A.A., Murray, K.E., Inniss, E.C., 2009. Sorption behaviour of a synthetic antioxidant, polycyclic musk, and an organophosphate insecticide in wastewater sludge. Water Sci. Technol. 60 (1), 145e154. Tolosa, I., Readman, J.W., Blaevoet, A., Ghilini, S., Bartocci, J., Horvat, M., 1996. Contamination of Mediterranean (Coˆte d’Azur) coastal waters by organotins and irgarol 1051 used in antifouling paints. Mar. Pollut. Bull. 32 (4), 335e341. Tomlin, C. (Ed.), 1994. The Pesticide Manual. British Crop Protection Council (BCPC), Farnham, UK. Urase, T., Kikuta, T., 2005. Separate estimation of adsorption and degradation of pharmaceutical substances and estrogens in the activated sludge process. Water Res. 39, 1289e1300. VCCLAB, 2005. Virtual Computational Chemistry Laboratory. http://www.vcclab.org (08/10). Wick, A., Fink, G., Joss, A., Siegrist, H., Ternes, T.A., 2009. Fate of beta blockers and psycho-active drugs in conventional wastewater treatment. Water Res. 43, 1060e1074. Wick, A., Fink, G., Ternes, T.A., 2010. Comparison of electrospray ionization and atmospheric pressure chemical ionization for multi-residue analysis of biocides, UV-filters and benzothiazoles in aqueous matrices and activated sludge by liquid chromatography-tandem mass spectrometry. J. Chromatogr. A 1217, 2088e2103. Zhao, J., Li, Y., Zhang, C., Zeng, Q., Zhou, Q., 2008. Sorption and degradation of bisphenol A by aerobic activated sludge. J. Hazard. Mater. 155, 305e311.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 5 3 e3 6 6 4

Available at www.sciencedirect.com

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

Identification and chemical characterization of specific organic indicators in the effluents from chemical production sites Oxana Botalova, Jan Schwarzbauer*, Nadia al Sandouk Institute of Geology and Geochemistry of Petroleum and Coal, RWTH Aachen University, Lochnerstrasse 4-20, 52056 Aachen, Germany

article info

abstract

Article history:

The structural diversity of the wastewater composition was described by the use of

Received 3 September 2010

detailed non-target screening analyses of industrial effluents from chemical production

Received in revised form

sites. Determination of the indicative organic compounds acting as potential molecular

4 April 2011

indicators for industrial emissions from chemical production industries has been possible

Accepted 6 April 2011

due to (i) detailed characterisation of industrial contaminants and identification of

Available online 16 April 2011

compounds with high source specificity, (ii) quantitative determination of the organic constituents in the industrial effluents and (iii) the review of their industrial applications.

Keywords:

The determination of potential site-specific markers and industrial molecular indicators

Chemical production plants

corresponding to certain production processes (production of starting materials for

Industrial effluents

manufacturing paper and printing inks, powder coatings as well as epichlorohydrin

GC/MS

production) was performed in this work.

Non-target screening

The results of this study allowed significant contributions to the chemical character-

Industrial molecular indicators

isation of industrial contaminants and isolation of indicators that can act as representa-

Site-specific markers

tives of industrial effluents in the aquatic environment. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Chemical production plants represent a substantial source of contamination introduced into the aquatic environment. They produce a wide range of compounds including pharmaceuticals, plant and crop-protecting agents, solvents, plasticizers, antioxidants, thermal stabilizers, ultraviolet light absorbers, optical brighteners, surfactants and other chemical products. Being emitted at a significant concentration level, organic compounds in the chemical effluents normally possess a high structural diversity and reveal notable ecotoxicological effects. Within a wide range of industrial contaminants associated with synthetic products, and by-products as well as with those formed during wastewater treatment, many of them are unknown as far as structural characteristics and environmental

behaviour are concerned. So far the composition of industrial effluents has been investigated comprising industrial sites of various production branches, i.e. paper, textile, tannery and leather industry as well as chemical manufacturing (Smith, 1990; Knepper, 2002; Morisawa et al., 2003; Rojas and Ojeda, 2005; Lo´pez-Grimau et al., 2006; Soupilas et al., 2008). In a previous study we have made an attempt to identify possible markers for petrochemical industrial activities (Botalova et al., 2009). Besides a high structural diversity of their constituents, industrial effluents have often had pronounced ecotoxicological effects on aquatic organisms. Investigations on the toxicological and ecotoxicological properties of industrial effluents have been performed in a number of studies (Castillo et al., 2001; Hewitt and Marvin, 2005). Phthalate esters used in the production of various plastics (including PVC) are among the most common industrial

* Corresponding author. Tel.: þ49 241 805750; fax: þ49 241 80695750. E-mail address: [email protected] (J. Schwarzbauer). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.012

3654

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 5 3 e3 6 6 4

chemicals. These contaminants together with the food antioxidant butylated hydroxyanisole have revealed estrogenic activity in fish and mammalian tests (Jobling et al., 1995). Genotoxicity of acridine derivatives detected in industrial waters in the Netherlands has been shown by Bobeldijk et al. (2002). The authors reported a first observation in surface water of hexamethoxymethylmelamine, a chemical often used in coating industry. In the effluent from a semi-conductor equipment manufacturer Labunska et al., 2008 found concentrations of 36 mg/L chloroform and 65 mg/L highly toxic and carcinogenic tetrachloromethane, which had been discharged into a river. There are a number of ecotoxicological studies performed on wastewaters from chemical manufacturing sites (Jop et al., 1991; White et al., 1996; Soupilas et al., 2008). Ecotoxicological investigations have been carried out on complex effluents from a chemical plant in Italy producing synthetic rubbers, latexes of synthetic rubber, acrylonitrile butadiene styrene resins, dimethylcarbonate and derivatives (two-ring phenols, phenolic antioxidants, esters of diethylenglycol), fertilizers (ammonium nitrate, compound fertilizers), etc (Guerra, 2001). The evaluation of ecotoxicological response has been performed in the tests with Daphnia magna, Vibrio fisheri, Artemia salina, Brachionus plicatillis. Ecotoxicological investigations were carried out by Swartz et al. (2003) in the wetlands adjacent to an industrial area with a number of chemical factories, i.e. a chlor-alkali plant, located in the Republic of Azerbaijan. The results of the study showed the acute toxicity in fish tests with Russian sturgeon (Acipenser gueldenstaedti) as well as potential mutagenicity in Caspian turtles (Mauremys caspica). Determination of specific molecular indicators in wastewaters from certain industrial branches or production processes can be useful to distinguish various industrial pollution sources. Despite numerous investigations performed on the characterisation of effluents discharged by chemical production plants, there is a gap in information on the specificity of contaminants in terms of their molecular structure. This approach does not consider tracing of synthetic products appearing in industrial effluents, which are likely to be found in municipal sewage wastewater as well. It is rather focused on the identification of industrially related contaminants like synthetic precursors, intermediates and by-products in industrial effluents. These particular molecular structures are specific in relation to a certain industrial branch or production process. The aims of this work are (i) to characterise in detail organic chemical composition in the effluents from a set of industrial sites and (ii) to isolate specific compounds that might act as source indicators (site-specific markers and branch-specific indicators). Some compounds are found in the wastewater exclusively from a certain industrial site discharging its effluents into a river. If detected in the water of the same river along the flow only after the discharge point of this industrial site, they can act as sitespecific markers. As an important precondition, these contaminants appear neither in the effluents from other industries contaminating the river nor in the river water affected by other industrial discharges. Branch-specific indicators are related to the corresponding production processes taking place at the chemical sites. Monitoring and identification of such molecular indicators would substantially contribute to the efficient point source identification.

2.

Methods and materials

2.1.

Samples

Samples of wastewaters subject to biological and chemical treatment before their discharge into a river were collected from five chemical production plants (defined as A, B, C, D, and E) situated in North-Rhine Westphalia, Germany. Information on single grab samples provided for the analyses is given in Table 1. Sampling campaigns at industrial complex A took place four times in summer 2007 (27.06; 19.07; 02.08; 15.08). Effluent samples from industrial sites B, C, D and E were collected in November 2007. The wastewater samples were filled in thoroughly precleaned 1 L aluminium bottles with alumina coated screw caps and stored in the dark at a temperature of 4  C.

2.2.

Chemicals and glassware

Only glass, metal and PTFE equipment were used in the laboratory in order to minimise possible contamination. All glassware was cleaned by ultrasonic agitation in water containing detergent (Extran, Merck, Germany) and rinsed with water followed by high-purity acetone and n-hexane. The solvents were purchased from Merck, Germany, and distilled over a 0.5 m packed column (reflux ratio approximately 1:25). The solvent purity was tested by gas chromatographic analyses. Anhydrous granulated sodium sulphate (Merck, Germany) and hydrochloric acid (Merck, Germany) needed for the analytical procedure were cleaned by their extraction with pure acetone. The results of blank analyses indicated that none of the compounds presented in this study was detected in the blank.

2.3.

Extraction

The extraction method used has been previously described in detail (Dsikowitzky et al., 2002). Briefly, a sequential liquideliquid extraction procedure was applied to approximately 500 mL and 1000 mL aliquots of the wastewater and river water samples, respectively, using n-pentane, dichloromethane and dichloromethane after acidification to pH 2 with hydrochloric acid. 50 mL

Table 1 e Sampling of the wastewater from chemical production sites. Site

Sampling locations

Date of sampling

Industrial site A

eOutflow A

Industrial site B

eOutflow eOutflow eOutflow eOutflow eOutflow eOutflow eOutflow eOutflow

4 sampling campaigns: 27.06.2007 19.07.2007 02.08.2007 15.08.2007 08.11.2007 22.11.2007 22.11.2007 21.11.2007 07.11.2007 19.11.2007

Industrial site C Industrial site D Industrial site E

B1 B3 B4 C1 C2 D1 D2 E

12.11.2007

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 5 3 e3 6 6 4

of solvent were used for each extraction step. The first two extracts were spiked with 50 mL of a surrogate standard solution containing d34-hexadecane (6.0 ng/mL) and decafluoroacetophenone (7.0 ng/mL). 200 mL of surrogate standard containing 4-fluoroacetophenone (7.2 ng/mL) were added to the third extract containing the acidic components. Thereafter, the organic layers were concentrated by rotary evaporation at room temperature under reduced pressure to approx. 1 mL and dried by filtration over 1 g of anhydrous granulated sodium sulphate. Prior to GC- and GC/MS-analyses all the extracts (the third extract after derivatization, see section 2.4) were further concentrated at room temperature under ambient pressure. The final volumes of wastewater extracts were: 50 mL (first two extracts) and 200 mL (the third extract).

2.4.

Derivatization

Prior to chromatographic analysis the acidic compounds in the third extract were derivatized to enhance a successful gas chromatographic separation and detection. For this purpose 1 mL of diazomethane solution was added to approx 0.5 mL of the extract which was simultaneously cooled in an ice bath. The vial was tightly closed and allowed to remain in the icewater for 30 min. Thereafter, sample volume was reduced to approx. 200 mL at room temperature and ambient pressure.

2.5. Gas chromatography (GC) and gas chromatography-mass spectrometry (GC/MS) Gas chromatographic analyses were performed on a GC 8000 gas chromatograph (Fisons Instruments, Germany) equipped with a ZB-1HT fused silica capillary column (30 m length  0.25 mm ID  0.25 mm df, Phenomenex, USA). The GC oven was programed from 60 to 300  C at a rate of 3 C min1 after 3 min at the initial temperature, and was kept at 300  C for 20 min. The injection was carried out on a split/splitless injector at 270  C, splitless time was 60 s. Hydrogen carrier gas velocity was 37 cm/s. Detection was conducted by a flame ionization detector (FID) at 300  C. GC/MS analyses were carried out on a Trace MS mass spectrometer (Thermoquest, Egelsbach, Germany) linked to a Carlo Erba Mega Series 5140 gas chromatograph (CE, Milano, Italy) which was equipped with the same fused silica capillary column of the same dimensions as used for GC analyses. The temperature program and the injection conditions were also the same as described above. Helium carrier gas velocity was set to 30 cm s1. The mass spectrometer was operated in electron impact ionization mode (EIþ, 70 eV) with a source temperature of 200  C. The mass spectrometer was scanning from 35 to 700 amu at a rate of 0.5 s decade1 with an interscan time of 0.1 s.

2.6.

Identification and quantification

Identification of organic compounds was based on comparison of EIþ-mass spectra with those of the reference compounds or mass-spectral databases (NIST/EPA/NIM Mass Spectral Library NIST02, Wiley/NSB Registry of Mass Spectral data, 7th electronic version) and gas chromatographic retention times. For

3655

correction of inaccuracies of retention time, the retention times of the surrogate standard compound were used. Quantitative data were obtained by integration of selected ion chromatograms extracted from the total ion current. Quantification details are presented in the Supplementary Materials (Table S1). For quantification, response factors were determined from four-point linear regression functions based on calibration measurements with different compound concentrations. The concentrations (10e100 ng/mL, injection of 1 mL) ranged within the expected values in the samples and within a linear detection range. Quantification of a compound detected in different extracts was performed by addition of the corresponding concentration values. For correction of injection volume and sample volume inaccuracies, the surrogate standard was used. The detection limit was in the range of 1 ng/L. At concentrations of less than 5 ng/L, no attempts were made to quantify components.

3.

Results and discussion

Industrial effluents from five chemical production sites were subject to the organic analyses. The obtained results demonstrated high structural diversity of organic chemical pollutants in the industrial wastewaters. All qualitative and quantitative results are discussed with respect to the aspects of (i) industrial and technical application of the industrial wastewater constituents, and (ii) possible source specificity of contaminants acting as possible industrial molecular indicators or site-specific markers. It is important to notice that only a restricted number of samples has been investigated as a preliminary evaluation of a potential impact of industrial contaminants from diverse chemical production sites on the aquatic environment. Consequently, due to the analysis of single grab samples and a large number of discontinuous production processes, presented quantitative data cannot be attributed to the mean values and reflect only rough values of wastewater composition depending on fluctuations in technical parameters at the production sites. The chemical industries investigated were different in their production activities. Chemical complex A produces plastic powder and plastic granulates (polypropylene) for further industrial processing as well as PVC. Among many commercial formulations, intermediate products (chlorine, caustic soda solution, dichloroethane, and hydrogen), plant protection agents and intermediates, halogen-free flame-retardants, phosphatefree detergent raw materials and runway deicer are produced at the chemical plant. Production of high temperature super conductors and monochloroacetic acid, used in the manufacturing of special detergents, thickening agents, adhesives, crop protection substances, plastic stabilizers and pharmaceuticals, also takes place. Another direction is the production of phosphorus products, i.e. phosphorus pentasulfide, a raw material for the manufacture of additives that are primarily used in the automobile industry in lubricating materials. On its territory the chemical park also has a waste-toenergy power plant, a combined gas-steam power plant, an air separation plant involved in industrial gase production, biomaterials works and a waste deposit site. Industry B produces a number of chemicals such as alcohols, aldehydes, carboxylic

3656

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 5 3 e3 6 6 4

Table 2 e Selected organic compounds identified by non-target screening in the effluents from chemical production sites. Compound

Industry A

Industry B

27.06.2007 19.07.2007 02.08.2007 15.08.2007 B1 B3 B4 Halogenated compounds 2-(Chloromethyl)-1,3-dioxolaneb Dichlorobenzeneb Dichloroanilineb Trichloroanilineb 3,4-Dichloromethylthiobenzenea 3’-(Trifluoromethyl)acetophenonea Trichloroacetic acid, 2-propenyl esterb Bis(1,3-dichloro-2-propyl)etherb 1,3-Dichloro-2-propyl-2,3-dichloro1-propyletherb Nitrogen-containing compounds 1,4-Dimethylpyrazolb Tributylaminea Tris(2-ethylhexyl)amineb Trioctylamineb Triisooctylaminesb N,N-Dibutyldecanamineb 5-Methyl-1,3-diazaadamantan6-oneb Tetramethylbutanedinitrileb 3,3-Diphenyl-2-propenenitrileb Sulfur-containing compounds Dithiolaneb 1,3-Dithianeb 1,2,4-Trithiolaneb Trithianesb 1,2,4,5-Tetrathianeb 1,2,4,6-Tetrathiepaneb Hexathiepaneb Tetrahydrothiophene-2, 5-dicarboxylic acidb,c Phosphorus-containing compounds Tris(chloropropyl) phosphates (TCPP)a Triethylphosphate (TEP)a Dibutylmethylphosphonate (DBMP)b Oxygen-containing compounds 2,2,4-Trimethyl-1, 3-pentanedioldiisobutyrate (TXIB)a 3-Hydroxy-2,2-dimethylpropanoic acid, 3-hydroxy-2,2dimethylpropyl ester (HPHP)b Triacetinb Polyromatic and other compounds Diisopropylnaphthalene (DIPN)a

Industry C

Industry D

C1

D1

C2

Industry E

D2

þ þþ þ þ 0 þ

þþ þ þ

þ þ 0

þþ 0

þþ

þ

þþ þ þþ þ

þþ þ þ þ þ 0 0 0 0

þ 0

þ 0

þ

þþ þ þþ þ þ þþ 0

þ

þ

þ

þþ

þ

0

þ

0

0

þþ þ

þ þ

0 0

0 þþ

0 0

þ

0

þþ

0 0

þþ high content in a sample, þ low content in a sample, 0 traces in a sample. a Identified by comparison of GC and mass spectral data with those of reference compounds. b Identified of mass spectral data with those of mass spectral databases. c Detected as dimethylester.

acids, esters, amines and other intermediate products. The latter are used by other industries as raw materials in the manufacture of lacquers, paints, high solids and powder coatings, solvents, plasticizers, lubricants, rubber chemicals, diesel fuel and

mineral oil additives, deicers, flavours and fragrances, agrochemicals, surfactants, pharmaceuticals, cosmetics etc. Industry C is a manufacturer of more than 4000 products including industrial chemicals, chlorinated compounds,

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 5 3 e3 6 6 4

polymers and solvents. Steam and electric power plants located on its territory produce the energy needed for the chemical park. The production area of industrial plant D covers the manufacture of synthetic fibres, lacquers and paints, synthetic materials as well as paper. In addition, this chemical plant provides special products for food, cosmetic and pharmaceutical industries, as well as materials for printing inks, chromatography, catalysis and nanotechnology. Industrial site E is involved in the manufacture of soda, sodium bicarbonate, sodium hydroxide, allylchloride, glycerine, epichlorohydrin, PVC, polyarylamide as well as filling materials and other raw materials and substances used in the production of various cleaning and disinfection products, health care and hygiene products and in the paper and food industry.

3.1. Constituents of the effluents from chemical production plants Based on the GC/MS non-target screening analyses of the wastewater effluents from five different chemical production plants, identification of a wide variety of organic pollutants has been performed. Selected environmentally relevant non-

3657

specific and indicative industrial compounds identified in the industrial effluents are presented in Table 2 and are arranged, considering their principal molecular structural properties, in the groups of halogenated, nitrogen, sulphur, phosphorous and oxygen containing compounds and aromatic hydrocarbons. Furthermore, the results from four different sampling campaigns at industrial plant A are presented. The spectrum of organic constituents with respect to structural diversity and quantitative representation is illustrated in Figs. 1e3 with the chromatograms of the outflow wastewater extracts containing various hydrophobic compounds. Some of the substances are well known and ubiquitous pollutants. However, many of the compounds were identified for the first time in industrial effluents. For certain constituents, a lack of information on their industrial and technical applications as well as insufficient knowledge of their ecotoxicological properties were pointed out. Hereafter, the detected compounds are discussed with respect to their ability to act as potential molecular indicators for industrial effluents. The complete screening list and quantitative data as well as the summary of the available information on application and/or formation and ecotoxicological properties of the environmentally relevant wastewater constituents are provided in the Supplementary Materials (Tables S2-S4).

3.1.1.

Non-specific contaminants

Non-specific environmental contaminants are represented by compounds that have been already reported to originate from municipal and industrial waste. For some of them, i.e. dichlorobenzene, triethylphosphate (TEP), tris(chloropropyl) phosphate (TCPP), tributylamine, and benzoic acid, high concentration values (up to 230 mg/L) in the effluents from industries A, B and C have been observed (Table 3). The detailed information on the application and ecotoxicological properties of these substances is summarized in Table S4 of the Supplementary Materials. Major non-specific components identified in high amounts in the effluents of industry D were presented by cyclic polysulfides and 1,4-dimethylpyrazol. These compounds are often found in municipal and various industrial discharges.

3.1.2.

Fig. 1 e Total ion chromatogram of the pentane extract of an effluent from industry A sampled on 27.06.2007 (individual molecular structures corresponding to peak numbers).

Indicative industrial contaminants

A wide range of indicative substances identified in this study (Table S2 of the Supplementary Materials) have been previously reported to stem exclusively from industrial emissions. Those include, for instance, 1,2,3,4-tetrachlorobutane (Santillo et al., 1997), N-methylpyrrolidone (Castillo et al., 1998), N-nitrosomorpholine (Spanggord et al., 1982), Nacetylmorpholine (Botalova et al., 2009), N-formylmorpholine (Brigden et al, 2009), hexathiepane (Gulyas and Reich, 1995; EPA/OTS, 2008), 2,5,8,11-tetraoxadodecane (Beschkov et al., 1997). Some of them represent contaminants that are common for many chemical manufacturing sites and were detected in the effluents of more than one industrial plant presented in this paper. Nevertheless, several groups of chemicals were detected in wastewaters exclusively from a certain chemical industry. These compounds can be distinguished as potential site-specific markers in case they are detected in a river subject to discharges from this

3658

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 5 3 e3 6 6 4

Fig. 2 e Total ion chromatogram of the pentane extract of effluent B3 and DCM extract of effluent C1 (individual molecular structures corresponding to peak numbers).

industrial site. A review of the formation and/or application data is given in Table S4 of the Supplementary Materials. In the following we will focus on the discussion of compounds that so far have not been described as industrial effluent

constituents and those substances that might act as possible industrial molecular indicators. Among known industrial contaminants (e.g. di- and tri-chloroanilines), compounds as 30 -(trifluoromethyl)

3659

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 5 3 e3 6 6 4

C4

O

O

N

N N Trans & cis decalin

C4-Benzene

7

Naphthalene

8

5-Methyl-1,3-diaza-adamantan-6-one 10

9

N,N-Dibutylbutanamide 11

N N

N,N-Dibutyldecanamine

N

Tris-(2-ethylhexyl)amine

12

N

Triisooctylamines

13

Trioctylamine

14

15

Cl O

O

O

O O

Bis(2-ethylhexyl)phthalate 16

O

O

O

H N

O

O

2-(Chloromethyl)-1,3-dioxolane

Octahydro-2,3‘-bifuran

17

Tetramethylcuccinimide

18

19

O O

O

O 2,4,4-Trimethylcyclopentanone 20

OH

O

TXIB

Irgacure 184

21

22

Fig. 2 e (continued).

acetophenone, dichloromethylthiobenzene, 2-(chloromethyl)1,3-dioxolane, 2,2,4-trimethyl-1,3-pentanedioldiisobutyrate (TXIB), dibutylmethylphosphonate (DBMP), 3,3-diphenyl-2propenenitrile, and 5-methyl-1,3-diazaadamantan-6-one have been reported here for the first time as industrial wastewater constituents. Ecotoxicological properties of some of them (e.g., 30 -(trifluoromethyl)acetophenone, 2-(chloromethyl)-1,3-dioxolane) have not been described yet. The data on industrial application of 30 -(trifluoromethyl) acetophenone are rare. The information on its usage for preparation of pesticides and as intermediates in medicine synthesizing is given in several patents. This compound was detected in the effluents of industrial plant A from all sampling campaigns at concentration of up to 5 mg/L. Dichloromethylthiobenzene was detected in the effluent of industry A at a concentration of 0.5 mg/L. It is usually obtained by methylation of dichlorothiophenol (Haraguchi and

Bergman, 1991). 2-(Chloromethyl)-1,3-dioxolane (peak 17, Fig. 2) was found in effluent C1 at a concentration of 0.7 mg/L. To our knowledge, there is no information on industrial applications of these two substances in the literature. TXIB (peak 21, Fig. 2) was observed in effluents B1, C1, and D1 at concentrations of 0.1 mg/L-0.5 mg/L. TXIB is used as a plasticizer for the manufacture of PVC and vinyl plastics (Skjevrak et al., 2003). The occurrence of TXIB in the Elbe river, Rhine river, and Lippe river water has been reported by Franke et al. (1995b), Hendricks et al. (1994), and Dsikowitzky et al. (2004), respectively. Its medium toxicity to fish and Daphnia magna was indicated in the report of Danish EPA (2001). Tetrahydrothiophene-2,5-dicarboxylic acid found in noticeable amounts in the effluents from industry A is an unknown environmental contaminant. Very little data exist on industrial application of this substance in the literature.

3660

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 5 3 e3 6 6 4

Fig. 3 e Total ion chromatogram of the derivatized DCM extract of effluent D2 and pentane extract of effluent E (individual molecular structures corresponding to peak numbers). Tetramethylbutanedinitrile and 3,3-diphenyl-2-propenenitrile (peak 6, Fig. 1) were identified at notable concentrations (up to 1.4 mg/L and 0.5 mg/L, respectively) in the effluents from industry A. Tetramethylbutanedinitrile is a decomposition

product of 2,20 -azobisisobutyronitrile used as a polymerization initiator for monomers in the plastic production. Ishiwata et al., 1994 have reported on trace amounts of this compound released into foods from PVC containers. Therefore, taking

3661

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 5 3 e3 6 6 4

N

S

N

O S

S

1,4-Dimethylpyrazol

1,3-Dithiolane

23

S

1,3,5-Trithiane 28

1,2,4-Trithiolane

O

1,2,4,5-Tetrathiane

27

O

S

O

O

O

O

2,5,8,11,14,17-Hexaoxaoctadecane 30

Cl

O

Cl

Trichloroacetic acid, 2-propenyl ester 31

O

2,5,8,11-Tetraoxadecane

26

S

O

O

S

29

O Cl

S

25

S S

Cl

1,3-Dithiane

24

S S

S

S

O

Cl Cl

Cl Cl

Bis(1,3-dichloro-2-propyl)ether 32

O

Cl

Cl Cl 1,3-Dichloro-2-propyl-2,3-dichloro-1-propylether 33

Fig. 3 e (continued).

into consideration that PVC production is one of the major manufacturing activities at industrial complex A, this substance can act as a possible site-specific marker and industrial indicator for the polymer production process. The group of tertiary amines is presented by tributyl-, trioctyl-, triisooctylamines, tris-(2-ethylhexyl)amine, and N, N-dibutyldecanamime. Tributyl- and trioctylamines, tris(2ethylhexyl)amine and N,N-dibutyldecanamine (Fig. 2) were detected in effluent B3 at concentrations from 0.03 mg/L to 5.1 mg/L. These compounds can be attributed to the sitespecific contaminants for industrial plant B as butyl-, octyland 2-(ethylhexyl)amines and are known to be produced by this chemical manufacturer. They are used as raw materials for rubber chemicals, pesticides, lubricants, detergents, photographic chemicals, corrosion inhibitors, explosives, fuel additives, dyes and pharmaceuticals (Sanders et al., 2001) as well as extraction reagents (Uslu and Kirbas‚lar, 2010). Tertiary amines are known to be toxic to aquatic organisms and humans. In particular, tributylamine and tris-(2-ethylhexyl) amine were found to be toxic towards Daphnia magna (Lyman et al., 1990). Cytotoxic effects of tributylamine in human fibroblasts were observed by Witte et al. (1995). A high content of 3-hydroxy-2,2-dimethylpropyl ester of 3hydroxy-2,2-dimethylpropanoic acid (HPHP) was detected in effluent B3. HPHP is used as an intermediate in the manufacture of binding agents and coil and powder coatings (BASF AG, 2006). The compounds can serve as a possible industrial indicator for this production line as industry B is specialized on the production of raw materials for powder coatings. It is also used as a monomer in the manufacture of polyurethane/acrylic graft copolymers. The compound is relatively persistent in the

aqueous environment and has shown toxicity to fish, aquatic invertebrates and plants, and bacteria (EPA, 2009). Triacetin detected in effluent D1 is mainly used as a cellulosic plasticizer in the manufacture of cigarette filters, plasticizer for laminating resins, vinylidene polymers and copolymers. It is also used as a solvent and carrier in pharmaceutical preparations, in the compounding of perfumes and flavors, as well as an ingredient for printing inks and useful reagents in textile dyeing and the manufacture of photographic films. These applications can explain the appearance of triacetin in the effluent from industrial plant D that is specialized on the manufacture of paper, special cosmetic and pharmaceutical products as well as materials for printing inks. Therefore, it can be suggested to act as a potential indicator for these industrial processes. It is a known contaminant in propellant (Attaway, 1994) and industrial wastewaters (Rivera et al., 1987) as well as in surface water (Franke et al., 1995b; Schwarzbauer and Ricking, 2010). The wastewater from industrial plant E did not show a high molecular diversity of the organic constituents. In detail, a group of chlorinated ethers and related compounds was characterized based on their mass-spectral properties. However, precise identification of many of them was not possible due to the lack of reference materials. Among the constituents of the effluent from industrial plant E, the structures of 2-propenyl ester of trichloroacetic acid, bis(1,3-dichloro-2-propyl)ether and 1,3dichloro-2-propyl-2,3-dichloro-1-propylether (peaks 31, 32, and 33, respectively, Fig. 3) were determined. Two bis(dichloropropyl)ether isomers are by-products of industrial epichlorohydrin production, which is one of the main production activities at plant E. This allows their consideration as industrial indicators for epichlorohydrin

3662

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 5 3 e3 6 6 4

Table 3 e Contaminant concentrations (mg/L) in effluents (outflow) from chemical production sites. Compound

Industry A

Halogenated compounds 2-(Chloromethyl)-1,3-dioxolane Dichlorobenzene Dichloroaniline Trichloroaniline 3’-(Trifluoromethyl)acetophenone 3,4-Dichloromethylthiobenzene Nitrogen-containing compounds Tetramethylbutanedinitrileb 3,3-Diphenyl-2-propenenitrile Tributylamine Tris(2-ethylhexyl)amine Trioctylamine N,N-Dibutyldecanamime Phosphorus-containing compounds TCPP TEP DBMP Oxygen-containing compounds TXIB Polyromatic compounds DIPN

27.06.2007

19.07.2007

02.08.2007

15.08.2007

14 4.4 0.7 0.3 0.5

9.4 4.8

2.4 0.2

5

0.8 0.5 0.2 2.7

0.5 0.2

0.7 0.4

1.4 0.5

0.8

B3

B4

Industry D

C1

D1

C2

D2

0.7

5.1 0.2 0.7 0.03 1.3 9.4 2.9

23 2.3 2.1

Table 4 e Suggested potential industrial indicators and site-specific markers identified in industrial wastewaters.

Industry A, PVC production Branch-specific Epichlorohydrin production

B1

Industry C

0.7

production processes. The formation of C6 chloroethers proceeds as a side reaction in systems containing the allylchloride, epichlorohydrin and chlorine (Beger et al., 1983). First they have been detected in epichlorohydrin and petrochemical plant effluents in the Netherlands and the United States (De Leer, 1985; Dorn and Rodgers, 1989). Since 1990, they have been monitored in European surface water samples with maximum concentrations up to almost 50 mg/L found in the Elbe river in Germany (Franke et al., 1998). Interestingly, the differences in isomeric composition of bis(dichloropropyl) ethers obtained from different sampling locations have been observed in those data. For example, Franke et al., 1995a have

Site-specific Industry B, Raw materials

Industry B

Tributylamine Trioctylamine Tris(2-ethylhexyl)amine Tetramethylbutanedinitrile

Bis(1,3-dichloro-2-propyl)ether 1,3-Dichloro-2-propyl-2,3-dichloro-1propylether

Intermediate for powder 3-Hydroxy-2,2-dimethylpropyl ester of coatings 3-hydroxy-2,2-dimethylpropanoic acid (HPHP) Manufacturing of paper Triacetin and printing inks

1 0.4 0.2

0.5 0.4 11

0.3

0.5

0.5

0.1

0.1

reported the presence of tree isomers of C6 chloroethers in the Elbe River and its tributaries. In 1998 the authors indicated the occurrence in the German Bight of the North Sea receiving waters from the Elbe River of only two isomers (corresponding to peaks 32 and 33; see Fig. 3), with relative amounts of 20e32% and 68e80%, respectively, but in our case the reverse pattern of isomeric composition was observed. In particular, relative abundance of isomer 32 is much higher than that of isomer 33. This can be explained by differences in the control parameters of production processes at various industrial sites. Another reason for this phenomenon are the specific biological treatment methods applied to the wastewaters from different contamination sources. The influence of these bis(dichloropropyl)ether isomers on the environment has not been fully investigated. Residues of these contaminants were determined in fish from the Elbe River (Kruse, 1996). The acute toxicity in marine organisms of a mixture containing chloroethers was determined by Dorn et al. (1991). Mutagenicity was revealed in Salmonella typhimurium strains. Toxicologic investigations on Salmonella typhimurium have exhibited the genotoxic and carcinogenic activities of another bis(dichloropropyl)ether isomer in the 4.5e45.5 mg/mL concentration range (Neurath et al., 2000). 2-Propenyl ester of trichloroacetic acid has been described as a wastewater component in this work for the first time. No information on the physico-chemical properties and industrial application has been reported in the literature so far Based on the available literature and obtained data set, the results of our work allowed the identification of several compounds, which we suggest to be potential molecular indicators for certain industrial processes. These substances have not been found in municipal wastewater effluents so far (e.g. Ellis et al. (1982); Schultz and Kjeldsen (1986); Paxeus (1996); Rudel et al. (1998)). Their source specificity can be

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 5 3 e3 6 6 4

confirmed by comparison with the data on the pollutant distribution in the river systems under discharge from corresponding industrial sites.

4.

Conclusion

The comprehensive pattern of the structural diversity of the wastewater composition was depicted by the use of detailed non-target screening analyses applied to the industrial effluents from chemical manufacturing sites. Determination of organic compounds acting as potential molecular indicators of chemical production industries was possible due to (i) the elucidation of individual molecular structures, (ii) the quantitative characterization of the organic constituents in the industrial effluents and (iii) the literature review of their industrial applications. The determination of possible sitespecific markers and branch-specific indicators corresponding to a certain production process was performed in this work. Moreover, the non-target screening allowed the identification of new environmental contaminants that have never been detected in the industrial wastewaters and compounds, for which the information on their industrial application and ecotoxicological effects is limited. The summary of the possible candidates to act as branchand site-specific markers is presented in Table 4. In conclusion, the results of this study allowed significant contribution to the chemical characterisation of industrial contaminants and isolation of indicators that can act as representatives of industrial effluents in the aquatic environment. Nevertheless, in order to prove the approach proposed in this study it is necessary to trace the distribution of organic pollutants in the river system subject to the discharges from corresponding industrial sites. This will be the main task in our prospective investigations on determination of site-specific industrial markers, which would be valuable for the effective identification of specific emission sources.

Acknowledgements Our acknowledgements go to the Local Environmental Protection Agencies in Cologne and Herten, Germany, for their help and useful advice. This work was financially supported by a grant from the DFG (Deutsche Forschungsgemeinschaft, Schw750/10).

Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.04.012.

references

Attaway, H., 1994. Biodegradation of Nitroglycerin and Perchlorate in Propellant Wastewater. Report. March 1993-July 1994. Advanced Sciences. Inc, Albuquerque. USA.

3663

Basf, A.G., November 28, 2006. Test plan for propanoic acid, 3hydroxy-2,2-dimethyl-, 3-hydroxy-2,2-dimethylpropylester (Cas No. 11 15-20-4). Beger, J., Schiefer, H., Scheller, D., 1983. Three-component reaction. Part XVI. Halogenation of 1-alkenes in the presence of ethylene oxide and propylene oxide. J. Prakt. Chem. 325, 719e728. Beschkov, V., Bardarska, G., Gulyas, H., Sekoulov, I., 1997. Degradation of triethylene glycol dimethyl ether by ozonation combined with UV irradiation or hydrogen peroxide addition. Wat. Sci. Technol. 36, 131e138. Bobeldijk, I., Stoks, P.G.M., Vissers, J.P.C., Emke, E., van Leerdam, J. A., Muilwijk, B., Berbee, R., Noij, Th.H.M, 2002. Surface and wastewater quality monitoring: combination of liquid chromatography with (geno)toxicity detection, diode array detection and tandem mass spectrometry for identification of pollutants. J. Chromatog. A. 970, 167e181. Botalova, O., Schwarzbauer, J., Frauenrath, T., Dsikowitzky, L., 2009. Identification and chemical characterisation of specific organic constituents of petrochemical effluents. Wat. Res. 43, 3797e3812. Brigden K., Labunska I., Santillo D., Johnston P., 2009. Heavy metals and other hazardous chemicals discharged from an industrial wastewater treatment company into the Greater Pearl River Delta, China, 2009. Greenpeace Research Laboratories Technical Note 09/2009. http://greenpeace.to/ publications/discharge-wastewater-treatment-PRD.pdf. Castillo, M., Alonso, M.C., Riu, J., Reinke, M., Klo¨ter, G., Dizer, H., Fischer, B., Hansen, P.D., Barcelo´, D., 2001. Identification of cytotoxic compounds in European wastewaters during a field experiment. Anal. Chim. Acta 426, 265e277. Castillo, M., Oubina, A., Barcelo´, D., 1998. Evaluation of ELISA kits followed by liquid chromatography-atmospheric pressure chemical ionization-mass spectrometry for the determination of organic pollutants in industrial effluents. Environ. Sci. Technol. 32, 2180e2184. Danish E.P.A., 2001. Environmental and health assessment of alternatives to phthalates and to flexible PVC. Environmental Project No. 590. De Leer, E.W.B., 1985. The identification of highly chlorinated ethers and diethers in river sediment near an epichlorohydrin plant. Water Res. 19, 1411e1419. Dorn, P.B., Rodgers, J.H., 1989. Variability associated with identification of toxics in National Pollutant Discharge Elimination System (NPDES) effluent toxicity tests. Environ. Toxicol. Chem. 8, 893e902. Dorn, P.B., van Compernolle, R., Meyer, C.L., Crossland, N.O., 1991. Aquatic hazard assessment of the toxic fraction from the effluent of a petrochemical plant. Environ. Toxicol. Chem. 10, 691e703. Dsikowitzky, L., Schwarzbauer, J., Kronimus, A., Littke, R., 2004. The anthropogenic contribution to the organic load of the Lippe River (Germany). Part I: qualitative characterisation of low-molecular weight organic compounds. Chemosphere 57, 1275e1288. Dsikowitzky, L., Schwarzbauer, J., Littke, R., 2002. Distribution of polycyclic musks in water and particulate matter of the Lippe River. Org. Geochem. 33, 1747e1758. Ellis, D.D., Jone, C.M., Larson, R.A., Schaeffer, D.J., 1982. Organic constituents of mutagenic secondary effluents from wastewater treatment plants. Arch. Environm. Contain. Toxicol. 11, 373e382. EPA, 2009. Comments on Chemical RTK HPV Challenge Submission, pp. 1e4. EPA/OTS, 2008. Acute Oral, Eye, Skin, and Inhalation Toxicity, Preliminary Ground Water Assessment, and Characterization of Plant Effluents of Phenol with Cover Letter Dated 07/27/ 1987. Govt. Reports Announcements & Index (GRA&I) Issue 22, Doc. Nr. 86e870000940. Franke, S., Hildebrandt, S., Francke, W., 1995a. The occurrence of chlorinated bis-(propyl)ethers in the Elbe river and tributaries. Naturwiss 82, 80e83.

3664

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 5 3 e3 6 6 4

Franke, S., Hildebrandt, S., Francke, W., Bester, K., Hu¨hnerfuss, H., Gatermann, R., 1998. Chlorinated bis(propyl)ethers and chlorinated bis(ethyl)formals in the German Bight of the North Sea. Mar. Pollut. Bull. 36, 546e551. Franke, S., Hildebrandt, S., Schwarzbauer, J., Link, M., Francke, W., 1995b. Organic compounds as contaminants of the Elbe River and its tributaries Part II: GC/MS screening for contaminants of the Elbe water. Fresenius J. Anal. Chem. 353, 39e49. Guerra, R., 2001. Ecotoxicological and chemical evaluation of phenolic compounds in industrial effluents. Chemosphere 44, 1737e1747. Gulyas, H., Reich, M., 1995. Organic compounds at different stages of a refinery wastewater treatment plant. Wat. Sci. Tech. 32, 119e126. Haraguchi, K., Bergman, A., 1991. Synthesis of 14C-labelled PCB methyl sulphones. Chemosphere 23, 1837e1843. Hendricks, A.J., Maas-Diepeveen, J.L., Noordsij, A., van der Gaag, M.A., 1994. Monitoring response of XAD-concentrated water in the Rhine delta: a major part of the toxic compounds remains unidentified. Wat. Res. 28, 581e598. Hewitt, L.M., Marvin, C.H., 2005. Analytical methods in environmental effects-directed investigations of effluents. Mut. Res. 589, 208e232. Ishiwata, H., Sugita, T., Yamada, T., 1994. Determination of tetramethylsuccinonitrile in foods by NPD-GC. J. Food Hyg. Soc. Japan 35, 385e389. Jobling, S., Reynolds, T., White, R., Parker, M.G., Sumpter, J.P., 1995. A variety of environmentally persistent chemicals, including some phthalate plasticizers, are weakly estrogenic. Environ. Health Perspec. 103, 582e587. Jop, K.M., Kendall, T.Z., Askew, A.M., Foster, R.B., 1991. Use of fractionation procedures and extensive chemical analysis for toxicity identification of a chemical plant effluent. Environ. Toxicol. Chem. 10, 981e990. Knepper, T.P., 2002. Mass spectrometric strategies for the analysis of polar industrial chemicals and their by-products in wastewater and surface water. J. Chromatog. A. 974, 111e121. Kruse, R., 1996. Ru¨cksta¨nde von Bis-(Dichlorpropyl)ethern in Fischen aus der Elbe. UWSF-Z. Umweltwiss. Schadst.-Forsch 8, 122e124. Labunska, I., Brigden, K., Santillo, D., Kiselev, A., Johnston, P., 2008. PBDEs and other contaminants arising from production, recycling and disposal of electrical and electronic equipment in St-Petersburg area, Russia. Greenpeace Research Laboratories, Technical Note 07/08, p. 51. Lo´pez-Grimau, V., Guadayol, J.M., Griera, J.A., Gutie´rrez, M.C., 2006. Determination of non halogenated solvents in industrial wastewater using solid phase microextraction (SPME) and GCMS. Lat. Am. Appl. Res. 36, p. 12. Lyman, W.J., Reehl, W.F., Rosenblatt, D.H., 1990. Handbook of Chemical Property Estimation Methods. Environmental Behavior of Organic Compounds. American Chemical Society, Washington DC, 960 pp. Morisawa, T., Mizuno, T., Ohe, T., Watanabe, T., Hirayama, T., Nukaya, H., Shiozawa, T., Terao, Y., Sawanishi, H., Wakabayashi, K., 2003. Levels and behavior of 2-phenylbenzotoriazole-type mutagens in the effluent of a sewage treatment plant. Mut. Res. 534, 123e132. Neurath, G., Martin, F.L., Piasecki, A., Ruge, A., Cole, K.J., Franke, S., Francke, W., Marquardt, H., 2000. Cell transformation and genotoxicity induced by bis(2,3-dichloro1-propyl)ether. Environ. Mol. Mutagen 35, 312e318.

Paxeus, N., 1996. Organic pollutants in the effluents of large wastewater treatment plants in Sweden. Wat. Res. 30, 1115e1122. Rivera, J., Ventura, F., Caixach, J., de Torres, M., Figueras, A., Guardlola, J., 1987. GC/MS, HPLC and FAB mass spectrometric analysis of organic micropollutants in Barcelona’s water supply. Intern. J. Environ. Anal. Chem. 29, 15e35. Rojas, F.S., Ojeda, C.B., 2005. Effluent analysis in analytical chemistry: an overview. Annal. Bioanal. Chem. 382, 978e991. Rudel, R.A., Melly, S.J., Geno, P.W., Sun, G., Brody, J.G., 1998. Identification of alkylphenols and other estrogenic phenolic compounds in wastewater, septage, and groundwater on Cape Cod. Massachusetts. Environ. Sci. Technol. 32, 861e869. Sanders, C.A., Rodriguez Jr., M., Greenbaum, E., 2001. Stand-off tissue-based biosensors for the detection of chemical warfare agents using photosynthetic fluorescence induction. Biosens. Bioelectron. 16, 439e446. Santillo, D., Labounskaia, I., Stringer, R., Johnston, P., 1997. Report on the analysis of industrial wastewaters from the Frutarom VCM/PVC plant, near Haifa, Israel, and adjacent shoreline sediments for organic contaminants. Greenpeace Research Laboratories, Technical Note 03/97, p. 25. Schultz, B., Kjeldsen, P., 1986. Screening of organic matter in leachates from sanitary landfills using gas chromatography combined with mass spectrometry. Wat. Res. 20, 965e970. Schwarzbauer, J., Ricking, M., 2010. Non-target screening analysis of river water as compound related base for monitoring measures. Environ. Sci. Pollut. Res. 17, 934e947. Smith, B., 1990. Identification and reduction of toxic pollutants in textile mill effluents. Office of Waste Reduction, North Carolina State University, USA. 108 S. Swartz, C.D., Donnelly, K.C., Islamzadeh, A., Rowe, G.T., Rogers, W.J., Palatnikov, G.M., Mekhtiev, A.A., Kasimov, R., McDonald, T.J., Wickliffe, J.K., Presley, B.J., Bickham, J.W., 2003. Chemical contaminants and their effects in fish and wildlife from the industrial zone of Sumgayit, Republic of Azerbaijan. Ecotoxicology 12, 509e521. Skjevrak, I., Due, A., Gjerstad, K.O., Herikstad, H., 2003. Volatile organic components migrating from plastic pipes (HDPE, PEX and PVC) into drinking water. Wat. Res. 37, 1912e1920. Soupilas, A., Papadimitriou, C.A., Samaras, P., Gudulas, K., Petridis, D., 2008. Monitoring of industrial effluent ecotoxicity in the greater Thessaloniki area. Desalination 224, 261e270. Spanggord, R.J., Glbson, B.W., Keck, R.G., Thomas, D.W., 1982. Effluent analysis of wastewater generated in the manufacture of 2,4,6-trinitrotoluene. 1. Characterization study. Environ. Sci. Technol. 16, 229e232. Uslu, H., Kirbas‚lar, I., 2010. Extraction of aqueous of malic acid by trioctylamine extractant in various diluents. Fluid Phase Equilibria 287, 134e140. White, P.A., Rasmussen, J.B., Blaise, C., 1996. Comparing the presence, potency, and potential hazard of genotoxins extracted from a broad range of industrial effluents. Environ. Mol. Mutagen. 27, 116e139. Witte, I., Jacobi, H., Juhl-Strauss, U., 1995. Correlation of synergistic cytotoxic effects of environmental chemicals in human fibroblasts with their lipophilicity. Chemosphere 31, 4041e4049.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 5 e3 6 8 0

Available at www.sciencedirect.com

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

Modelling Cryptosporidium oocysts transport in small ungauged agricultural catchments Jialiang Tang a,c,*, Stephen McDonald a, Xinhua Peng d, Sukha R. Samadder a,e, Thomas M. Murphy b, Nicholas M. Holden a a

UCD Bioresources Research Centre/Biosystems Engineering, School of Agriculture, Food Science and Veterinary Medicine, University College Dublin, Belfield, Dublin 4, Ireland b Central Veterinary Research Laboratory, Backweston Campus, Young’s Cross, Celbridge, County Kildare, Ireland c Institute of Mountain Hazards and Environment, Chinese Academy of Sciences (CAS), 9, Block 4, Renmin South Road, Chengdu 610041, PR China d State Key Laboratory of Soil and Sustainable Agriculture, Institute of Soil Science, Chinese Academy of Sciences (CAS), Nanjing 210008, PR China e Department of Civil Engineering, MANIT, Bhopal, MP 462051, India

article info

abstract

Article history:

Cryptosporidium is an environmentally robust pathogen that has caused severe waterborne

Received 19 October 2010

disease outbreaks worldwide. The main source of zoonotic Cryptosporidium parvum oocysts

Received in revised form

in human drinking water is likely to be from farm animals via catchment pathways with

6 April 2011

water as the main transport vector. The vast majority of small agricultural catchments are

Accepted 7 April 2011

ungauged therefore it is difficult to use a process model to predict and understand the

Available online 23 April 2011

mechanisms and activities that regulate the risk of surface water contamination from agricultural areas. For this study, two ungauged agricultural catchments in Ireland were

Keywords:

used to model Cryptosporidium oocyst transport using SWAT2005 on a daily basis with

Stream flow

reference data from adjacent catchment gauging stations. The results indicated that

Cryptosporidium oocyst

SWAT2005 could simulate stream flow with good agreement between prediction and

Agricultural catchments

observation on a monthly basis (R2 from 0.94 to 0.83 and E (efficiency) from 0.92 to 0.66), but

SWAT

Cryptosporidium oocyst concentration results were less reliable (R2 from 0.20 to 0.37, P < 0.05; with poor E
0.37 to
2.57). A sensitivity analysis using independent parameter perturbation indicated that temperature was the most important parameter regulating oocyst transport in the study catchments and that the timing of manure application relative to the occurrence of water runoff event was critical. The results also showed that grazing management had little influence on predicted oocyst transport while fields fertilized with manure were the key critical source areas for microbial contaminations in the study catchments. It was concluded that the approach presented could be used to assist with understanding the epidemiology of waterborne cryptosporidiosis outbreaks and to improve catchment management for the safety of the general public health. ª 2011 Elsevier Ltd. All rights reserved.

* Corresponding author. Institute of Mountain Hazards and Environment, Chinese Academy of Sciences (CAS), 9, Block 4, Renmin South Road, Chengdu 610041, PR China. E-mail address: [email protected] (J. Tang). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.013

3666

1.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 5 e3 6 8 0

Introduction

Cryptosporidiosis in humans is characterised by severe watery, non-bloody diarrhoea in otherwise healthy people and is a relatively common gastroenteric disease (Fayer, 2004). It is usually caused by either zoonotic Cryptosporidium parvum or anthroponotic Cryptosporidium hominis (Carey et al., 2004). Ingestion of a small number of Cryptosporidium oocysts of either type can lead to potentially fatal consequences for immuno-suppressed individuals (Rose, 1997), and it is thought that infection can be caused by even a single oocyst (Casemore et al., 1997). In Ireland a loading of 0.05 oocysts per 10 L of treated water is the maximum permissible oocyst concentration based on continuous 24-h sampling for seven consecutive days (HPSC, 2007). Similar maximum loadings apply elsewhere in Europe, Australasia and North America (Haas et al., 1996; Lloyd and Drury, 2002; Bryan et al., 2009). In the last twenty years large numbers of waterborne outbreaks of cryptosporidiosis have been reported worldwide. It is estimated that about half a million human cases have occurred in Europe and North America (Coffey et al., 2007) and between 250 and 500 million infections of C. parvum occur annually in Asia, Africa and Latin America (Current and Garcia, 1991). At present, no specific drug treatment exists for cryptosporidiosis. C. parvum oocysts are 4e6 mm in diameter and are biologically dormant yet resistance to the levels of chlorine routinely used in potable water treatment thus making waterborne transmission of cryptosporidiosis one of the most prominent public health concerns worldwide (Rose et al., 2002; Fayer, 2008). Transmission of Cryptosporidium spp. is normally described as being via robust oocyst excreted in the faeces of the infected host (Smith and Rose, 1998; Tufenkji et al., 2006). Zoonotic cryptosporidiosis occurs through contact with infected neonatal calves or by contaminated drinking water or food (Fayer, 2008). Faeces of neonatal calves infected with C. parvum is the most likely source of environmental and surface water contamination (Goh et al., 2004). On beef and dairy farms, one infected calf may shed up to 10 million oocysts/day (Rose, 1997), and if not contained, these can contaminate any surface water within the precincts of a farm. Transmission of infection to human may be related to current and recent hydrological conditions. For example, storm conditions resulting in large runoff volumes and a failure of water treatment to eliminate contaminating oocysts contributed to the waterborne cryptosporidiosis outbreak that occurred in Milwaukee, Wisconsin, USA, in spring of 1993, during which 403000 illnesses were reported (Coffey et al., 2007). Increase in the parasitic protozoan concentration in surface water has been observed downstream of agricultural areas (Hansen and Ongerth, 1991; Ong et al., 1996; Medema and Schijven, 2001) and regions with high runoff potential and large daily pathogen production are likely to be sources for pathogenic contamination of surface waters (Stewart et al., 1997; Dorner et al., 2004; McDonald et al., in press). Researchers have used various models to study pathogen transport in the natural environment (Medema and Schijven, 2001; Ferguson et al., 2005; Dorner et al., 2006; Oliver et al., 2009), especially at catchment scale (Walker and Stedinger,

1999; Tian et al., 2002; Benham et al., 2006; Chin et al., 2009; Parajuli et al., 2009a; Coffey et al., 2010a,b). A limited number studies on Cryptosporidium spp. transport have been conducted as validation of catchment scale simulations because direct measurements of oocysts are difficult (Park and Huck, 2003; Dorner et al., 2004). Other researchers have used published oocyst concentration data in other areas to validate model results in specific catchment (Coffey et al., 2010b), but since Cryptosporidium oocyst transport is different from Escherichia coli and other fecal bacteria, the parameters of the model need to be further validated by field data. In order to develop more accurate risk analysis tools, an understanding of oocyst transport in the environment is required as this will allow an understanding of the reliability of the model. Despite theoretical limitations associated with hydrological data (Boughton and Chiew, 2007) and microbial dynamics (Oliver et al., 2009), catchment scale models do provide a means of identifying key locations within the hydrological boundary that influence the microbial quality of the surface water. The Soil And Water Assessment Tool (SWAT) was developed to predict the impact of land management practises on water, sediment and agricultural chemical yields in complex watersheds with different soil types, land use and management conditions over long time periods (Neitsch et al., 2004). The SWAT model requires specific information about weather, soil properties, topography, vegetation and land management practises occurring in a watershed. It can be applied to catchments with little or no available hydrological monitoring data (Neitsch et al., 2004; Ndomba et al., 2008), and was designed to be suitable for large watersheds. Some researchers have successfully used SWAT for catchment studies for areas as small as 40 ha (Veith et al., 2003; Feyereisen et al., 2007). In this study the SWAT model outputs were interpreted so as to derive possible recommendations for farmers and policy makers to protect water quality in small agricultural catchments against deterioration caused by C. parvum oocyst contamination. Agriculture in Ireland is based mainly on the production of livestock and livestock products from grass. Around 120,000 farmers are engaged in beef production using 2.5 m ha of grass to rear and fatten the progeny of 1.2 million dairy and 1.1 million beef cows, producing about 2 million animals for either slaughter in Ireland or export for slaughter in another country (Dunne et al., 2004). Weather permits 190e240 days grazing per year and all manure (including slurry) produced is spread on land (Casey and Holden, 2005). Furthermore, most solid geology is overlain by glacial drift from which the soils are formed (Gardiner and Radford, 1980), that give rise to about 3.4 m ha of good and 3.4 m ha of marginal agricultural land (Gardiner and Radford, 1980), both of which contribute to grass production systems. Ireland has a temperate wet maritime climate with mild, moist winters and cool cloudy summers. For the greater part of the year warm maritime air associated with the Gulf Stream helps to moderate the climate. The prevailing winds are westerly to SoutheWesterly. The average relative humidity is high (c. 90%) and annual precipitation, which exceeds evapotranspiration by over 500 mm, is highest on the west coast (>2000 mm) and in inland areas of high relief. As a result of this coincidence of soils, land use and climate, surface waters are perhaps at

3667

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 5 e3 6 8 0

a great risk of C. parvum contamination due to the large potential for source loading by infected calves (McDonald et al., in press), a high probability of free water being available to act as a transport vector (Samadder et al., 2010) and a high density of surface water bodies to act as targets (Toner, 1986). Ireland, because of the high prevalence of small agricultural catchments (Jordan et al., 2005; Boyle, 2009), is a geographical situation where modelling of catchment hydrology would be useful to better understand oocyst transport in agricultural land. Unfortunately, there is a dearth of hydrological data, from the many small catchments, on which to base these models. As a result of this situation, the objectives of this paper were (i) to establish the hydrological conditions for two ungauged small agricultural catchments based on data from adjacent gauging stations; and (ii) to model the Cryptosporidium spp. oocyst transport at the catchment scale using SWAT2005. The results of the modelling were then interpreted to aid epidemiological investigations of cryptosporidiosis outbreaks, to assist in predicting temporal variability of risk to human populations and provide a basis for policy advice for water quality protection.

2.

Methods

2.1.

Research area

Two catchments located in the province of Leinster (Fig. 1) were used for the study. They are referred to as North Leinster (NL), located in County Meath, with an area of 419 ha and elevation ranging from 57 to 158 m and South Leinster (SL), located in Country Kilkenny, with an area of 398 ha and elevation ranging from 129 to 257 m. The gauged catchment used to calibrate SWAT2005 (AVSWAT, modified 12 March 2009) for NL was about 25 km away and has an area of 74 km2, while for SL the gauged catchment was about 17 km away and had an area of 140 km2. Satellite images and topographical maps of the study catchments were obtained from the Ordinance Survey of Ireland (OSI). Field visits during 2007 were used to confirm the land use in each land parcel within the catchments (summarised in Table 1, Fig. 1). Digital Elevation Model (DTM) and soil map data for Ireland were obtained from the Environmental Protection Agency. Co-ordinates (latitude and longitude) of the ground control points for georectification throughout the study area were collected by GPS and processed as reported by Samadder et al. (2010) to create a base map with a pixel resolution of 10 m for each catchment. There are a total 5 working farms located within the NL catchment: one dairy herd and four beef operations (Samadder et al., 2010) with an average stocking rate of 1.35 animals per forage hectare and there are 4 working farms with 2 dairy herds and 2 beef herds in the SL catchment with an average stocking rate of 2.01 animals per forage hectare. In both catchments farm management practises were similar with cattle removed from pasture and housed during the period mid October until April. The animal population in each catchment remained constant throughout the study period. Manure and slurry were collected from housed animals and spread several times during the year.

2.2.

Cryptosporidium modelling theory

The SWAT2005 model (AVSWAT, modified 12 March 2009) with a modified function for microbial survival and a transport sub-model (Neitsch et al., 2004; Parajuli et al., 2009a) was used to model the fate and transport of micro-organisms. It was assumed that Cryptosporidium spp. transport in the environment was adequately captured as “persistent bacteria” in the SWAT2005 model. The model considers microbial organisms in the environment from the following perspectives, wash-off, die-off on foliage, in soil solution and sorbed to soil at the field scale, leaching processes, surface transport via surface runoff, attachment to sediment at the slope scale and decay processes in channels and water bodies. Wash-off will occur when the amount of precipitation on a given day exceeds 2.54 mm. The amount of bacteria/Cryptosporidium oocysts washing off plant foliage during a precipitation event on a day is calculated as: bactp;wsh ¼ frwsh;p  bactp,fol

(1)

where bactp,wsh is the amount of Cryptosporidium oocysts on foliage that is washed off the plant and onto the soil surface on a given day (count/m2), frwsh,p (corresponding to WOF_P in Table 6) is the wash-off fraction for Cryptosporidium oocyst, bactp,fol is the amount of Cryptosporidium oocyst attached to the foliage. Cryptosporidium oocyst that washes off the foliage is assumed to remain in solution in the soil surface layer. Die-off is one of the most important processes at the catchment scale and is governed by first-order kinetics using equations adapted from Reddy et al. (1981) and modified by Crane and Moore (1986) and Moore et al. (1989). The simplified expression is given by the following equation (eq. (2)) without the re-growth processes for Cryptosporidium spp. (Table 7): Ci ¼ Ci
1  e
K20 q

ðT
20Þ


Closs

(2)

where Ci is the Cryptosporidium spp. concentration (count 100 mL
1) on day i, Ci
1 is the Cryptosporidium spp. concentration (count 100 mL
1) on day i
1, K20 (WDPQ, WDPS, WDPF in Table 6) is the first-order die-off rate constant at 20  C (day
1), q (THBACT in Table 6) is the temperature adjustment factor, T is the temperature ( C), Closs is the minimum daily loss of Cryptosporidium spp. (count 100 mL
1). Cryptosporidium oocyst can be transported downward into the soil profile through percolation, only the oocyst present in the soil solution is assumed to leach and die in the deeper soil layers. The leaching process of Cryptosporidium oocyst in SWAT was expressed by the following equation: bactp;perc ¼

bactpsol  wperc;surf 10  rb  depthsurf  kbact;perc

(3)

where bactp, perc is the amount of Cryptosporidium oocyst transported from the top 10 mm into the first soil layer (count/m2), bactpsol is the amount of Cryptosporidium oocyst present in soil solution (count/m2), wperc,surf is the amount of water percolating to the first soil layer from the top 10 mm on a given day (mm H2O), rb is the bulk density of the top 10 mm (Mg/m3), depthsurf is the depth of the “surface” layer (10 mm), kbact,perc (BACTMIX in Table 6) is the Cryptosporidium oocyst percolation coefficient (10 m3/Mg).

3668

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 5 e3 6 8 0

Fig. 1 e The experimental catchments indicating meteorological observation and hydrologic gauging stations. [Note: Rainfall stations 1 [ Clones 2 [ Mullingar, 3 [ Birr, 4 [ Oak Park, 5 [ Kilkenny; Gauged reference stations R1 [ Rosehill and R2 [ Castlecomer]. Due to the low mobility of Cryptosporidium oocyst in soil solution, surface runoff only transport part of the Cryptosporidium oocyst from land surface. The equation is: bactp;surf ¼

bactpsol  Qsurf rb  depthsurf  Kbact;surf

(4)

where bactp,surf is the amount of Cryptosporidium oocyst lost in surface runoff (count/m2), bactp,sol l is the amount of Cryptosporidium oocyst present in soil solution (count/m2), Qsurf is the amount of surface runoff on a given day (mm H2O), rb is the bulk density of the top 10 mm (Mg/m3), depthsurf is the depth of the “surface” layer (10 mm), and Kbact,surf (BACTKDQ in Table 6) is the Cryptosporidium oocyst soil partitioning

Table 1 e Land use distribution in NL and SL catchments. Land use type

Good pasture Poor pasture Tillage Residential and farm buildings

Land area (%) NL

SL

72.63 0.31 26.03 1.02

88.06 0.71 8.59 2.63

Crop

Rye grass Grasses and shrubs Winter wheat Bare land

coefficient (m3/Mg) which is the ratio of the oocyst concentration in the surface 10 mm soil solution to the concentration of oocyst in surface runoff. Another important process considered in SWAT2005 is the attached oocyst to sediment in surface runoff. The amount of Cryptosporidium oocyst transported with sediment to the stream is calculated with a loading function developed by McElroy et al. (1976) and modified by Williams and Hann (1978) for nutrients. bactp;sed ¼ 0:1377 

0:7532  bactp;sorb sed  Q 0:2468 surf rb  depthsurf areahru

(5)

where bactp,sed is the amount of Cryptosporidium oocyst transported with sediment in surface runoff (count/m2), bactp,sorb is the amount of Cryptosporidium oocyst absorbed to the soil (count/m2), rb is the bulk density of the first soil layer (Mg/m3), depthsurf is the depth of the soil surface layer (10 mm), sed is the sediment yield on a given day (metric tons), areahru is the HRU area (ha), Qsurf is the amount of surface runoff on a given day (mm H2O). In the stream or water bodies, die-off process accounted for the only oocyst behaviour by SWAT2005. As the study catchment are both small and sloping, the stream flow was fast and the resident time of water and sediments was short, thus this

3669

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 5 e3 6 8 0

Table 2 e Slurry spreading regimes present in the north and south Leinster catchments. Management options

Application method

Application time

Grazing

Sporadical pat

Dung application Slurry application Slurry application South Leinster catchment

Splash plate Splash plate Splash plate

Grazing period (MarcheOctober) September 23e30 June 18e21 February 18e25

Grazing

Sporadical pat

Dung application Slurry application Slurry application Slurry application

Splash Splash Splash Splash

Application amount (Litres)

Estimated mean oocysts in Manures (Count Litre
1)

North Leinster catchment

Grazing period (MarcheOctober) September 25e30 June 18e30 September 25e30 February 1e8

plate plate plate plate

672,000

3.8  103

35,000 690,000 690,000

1.5  108* 1.91  106 4.47  106 3.8  103

1,008,000 17,500 600,000 100,000 800,000

1.5 1.81 1.31 2.2

   

108* 106 106 106

* Estimated from literatures, eg. (Tate et al., 2000)

situation could ignore the sedimentation effects as resuspension occurred frequently.

2.3.

Data collection and model initialisation

2.3.1.

Climate and hydrologic data

geological and meteorological conditions were identified and taken as reference catchments (Fig. 1). Daily stream flow data for the gauged catchments were obtained from EPA HydroNetOn-Line Hydrometric Data (hydronet.epa.ie) with rating standards of good for NL and fair for SL. In addition to the identical weather conditions, the geology and soils of the gauged catchments were also very similar to the study catchments, belonging to the physiographic division of rolling lowland mostly underlaid by shale and sandstone glacial till (Gardiner and Radford, 1980; Finch et al., 1983). Thus the hydrological conditions in the study catchments and the relative reference catchments were also assumed to be very similar. Due to differences in catchment area, the catchment outflow concentration time in the study catchments were different to the reference catchments. The differences were adjusted by convergence processing with the mass balance equation (6),

Daily rainfall and temperature data from 1 January 2000 to 31 March 2008 for the two catchments were collected for hydrological calibration and validation from five synoptic observation stations (Fig. 1) managed by Met Eireann, the Irish National Meteorological Service. The monthly rainfall and temperature patterns were similar for all five synoptic stations and the climate of the ungauged catchments could be well represented by the neighbouring synoptic stations. It was assumed that there was no difference in meteorological data between the study catchment and the neighbouring reference catchment due to the prevailing weather in Ireland (Jordan et al., 2005). For this study, following the usage of neighbouring gauge data under the similar situation in North American (Parajuli et al., 2009a, b), adjacent gauged catchments with similar

Ss;i ¼

 As   ð1
kÞ  Sr;i þ k  Sr;iþ1 Ar

(6)

where Ss, i is the stream flow of the study catchment on day i, As is the study catchment area and Ar is the reference catchment area, Sr, i is the gauged stream flow of the reference catchment

Table 3 e D Details of the outflow from the north and south Leinster compared to reference data from neighbouring gauged catchments. Date

Rainfall (mm) (24 h)

North Leinster 9/4/2009 10/2/2009 11/11/2009 11/14/2009 2/4/2010 South Leinster 9/15/2009 9/19/2009 10/6/2009 10/7/2009

Rainfall (mm) in previous

Flow (m3 s
1)

Referred data (m3 s
1)

Remarks

5 days

10 days

20 days

0.4 0.2 24.6 3.0 11.2

30.5 5.4 17.2 38 12.4

48.1 6.4 64.9 54.4 18

131 8.4 64.7 105.6 42.8

0.09 0.03 0.75 0.25 0.11

0.12 0.02 0.64 0.35 0.17

Baseflow Baseflow Storm flow Recession Medium Rain

0.0 0.1 25.2 0.1

0.6 0.2 1.4 26.6

10 0.8 1.4 26.6

73.6 57.8 1.8 26.9

0.03 0.02 0.14 0.05

0.03 0.03 0.11 0.06

Baseflow Baseflow Heavy rain After heavy rain

3670

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 5 e3 6 8 0

Table 4 e Sensitivities and chosen values of the main SWAT2005 parameters for the north and south Leinster catchments. Parameters

CN2 SOL_Z SOL_AWC ALPHA_BF ESCO SLOPE SOL_K CANMX SURLAG CH_K2

Brief description

Sensitivitya

Initial SCS runoff curve number for moisture condition II Depth from surface to bottom of soil layer Available water capacity of the soil layer Base flow alpha factor Soil evaporation compensation factor Average slope steepness Saturated hydraulic conductivity Maximum canopy storage Surface runoff lag coefficient Effective hdralulic conductivity in main channel alluvium

Lower bound

Upper bound

Initial value

Method

Value chosen NL

SL

5.51 (7.14)


10

10

Default

Add

5

10

0.68 (0.10) 0.43 (0.23) 0.28 (0.07) 0.24 (0.50) 0.17 (0.24) 0.16 (0.20) 0.15 (0.01) 0.04 (0.02) 0.03 (0.01)

0 75% 0.5 0 0 75% 0 0 0

Default 125% 1 1 1 125% 1 10 100

Default Default 0.9 0.95 Default Default Default 4 0

Unchanged Multiply Replace Replace Unchanged Multiply Unchanged Replace Add

Default 75% 0.99 0.67 Default 124% Default 0.61 18.61

Default 75% 0.98 0.50 Default 125% Default 0.68 4.32

a The number in the column is for North Leinster catchment while the number in the bracket is for South Leinster catchment.

on day i, Sr, iþ1 is the gauged stream flow of the reference catchment on day i þ 1; k is a coefficient of convergence of daily flow curve which represents the faster response and shorter direct flow period for smaller catchment, and k is set by 0.75 for NL catchment and 0.80 for SL catchment based on the relative catchment concentration time (Wanielista et al., 1997). The transformed reference data were treated as observed data in SWAT2005 and were validated using observations of stream flow in the two study catchments manually sampled by flow metre (MJP Geopacks, UK) following the manufacturer’s instructions. The measurement was normally performed consecutively for 6 h to investigate the flow dynamic during storm events and the daily flow was estimated based on the extrapolated curve by the fitted function. The base flow conditions were separated by the base flow program described by Arnold and Allen (1999).

2.3.2.

SWAT set up and oocyst source characterization

Based on the DTM and stream network, SWAT delineates watersheds into sub-basins. The stream channels extracted from the DTM were aligned to match stream coverage digitized from field maps with the burn-in option available within SWAT2005 (AVSWAT, modified 12 March 2009). The stream definition threshold area was set at 20 ha as a critical source area of about 2e5 percent of the watershed was found to be sufficient for modelling purposes as reviewed by Arabi et al. (2006). Thus, 5 sub-basins in the NL catchment and 13 subbasins in the SL catchment (Fig. 1) were defined. Then, subbasins are further subdivided into hydrologic response units (HRUs) which are assumed to be spatially uniform in terms of soil, land use and topographic characteristics. To classify the HRUs which corresponded to critical source area, land use and soil class threshold were both set at 7%, a total of 23 (HRUs) were classified for the NL catchment and 33 for the SL catchment. The dominant soil type in the two catchments was acid brown earth occupying 68% of NL catchment and 74% of SL catchment, other soil types were Gley, Lithosols, Regosols (Gardiner and Radford, 1980; Finch et al., 1983). Soil properties, including soil depth, bulk density, soil carbon content and texture, were estimated from existing descriptions (Gardiner and Radford, 1980; Finch et al., 1983; McGoff et al., 2007).

Generally, the soil properties in Ireland are a result of glacial till with relatively light bulk density of 0.37e1.17 g cm
3 for grass land soils and clay content of 8e15%, silt content of about 20e25% and sand content of more than 60% for most widely distributed acid brown earths. However, since soil depth varies significantly, it was treated as variable parameter in this study as detailed spatial investigation of soil depth was impossible. It was noted that the soils in SL catchment had higher content of clay than those in NL catchment because of the Upper Carboniferous shale origin of soils in SL catchments (Gardiner and Radford, 1980). Other properties of the major soils in Ireland, such as saturated hydraulic conductivity, USLE_K (universal soil loss equation soil erodibility factor), were cited from related literature (Diamond and Sills, 2001; Coffey et al., 2010a). The farm management operations in the study catchments included grazing of livestock, fertilization and tillage, but no point sources were included in the modelling process as no wastewater treatment plant and few sewage existed in the catchments. Wildlife was not included as it is not considered as a major Cryptosporidium spp. source. The grazing rotated among paddocks between late March and early November and the grass on ungrazed paddocks was conserved as silage. The riparian zones were fenced with low-voltage electrical wires except one location in NL catchment where cattle occasionally had access to a stream.

Table 5 e Performance statistics for hydrological simulations for the north and south Leinster catchments. Parameters

North Leinster

South Leinster

Calibration Validation Calibration Validation Monthly slope Monthly R2 Monthly E Monthly RMSE Daily slope Daily R2 Daily E Daily RMSE

0.98 0.74 0.65 0.80 0.67 0.60 0.59 0.050

1.06 0.94 0.92 0.53 0.77 0.71 0.66 0.049

1.07 0.78 0.62 0.52 0.73 0.55 0.48 0.042

1.22 0.83 0.66 0.77 0.88 0.54 0.33 0.049

3671

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 5 e3 6 8 0

Table 6 e The parameters used in SWAT2005 to model the behaviour of Cryptosporidium spp. oocysts in the environment. Parameter

Description

Lower bound

Upper bound

Initial value

Value chosena 1.07 0.2 Measured values 0.05 1.4 0.02 4800 0.5 0.8

THBACT BACTKDDB BACTPDB

Temperature adjustment factor for Cryptosporidium die-off Bacteria partition coefficient Initial concentration of Cryptosporidium in animal waste

0 0 0

10 1 1000,000,000

1.07 0 0

WDPQ WDPS WDPF BACTKDQ FRT_SURFACE WOF_P

Cryptosporidium die-off factor in soil solution 20  C Cryptosporidium die-off factor adsorbed to soil particles at 20  C Cryptosporidium die-off factor on foliage Cryptosporidium soil partitioning coefficient Fraction of manure applied to top 10 mm of soil Wash-off fraction for Cryptosporidium

0 0 0 0 0 0

1 1 1 10000 1 1

0 0 0 0 0 0

a The die-off factors were cited from Dorner et al. (2006); Peng et al. (2008); More detail can be found in Neitsch et al. (2004).

2.3.3.

Collection and processing of field samples

Slurry and dung samples were collected from storage areas just before land application or spreading. Slurry tanks were agitated before spreading ensuring samples were more uniform and representative than without agitation. Water samples were collected once a month at catchment outlets in NL and SL catchments. A portable monitoring unit, PMU7 (Hydraulic Modelling Services, UK) pumped 50 L of raw water through an Idexx Filtamax cartridge. All samples were stored at 4  C and analysed within 72 h of collection. In total, 22 water sampling events in NL and 18 water sampling events in SL were analyzed for Cryptosporidium oocysts. Slurry and dung samples were processed with a Waring stainless blender in a PBS-T solution (Fluka, UK) and filtered to 0.45 mm. Water samples were processed following the Idexx filter washing proceedures and the USEPA 1623 guidelines. All slides from feces and water samples were stained with an FITC monoclonal antibody stain (Cellabs, Australia) and analysed with immunofluoresent microscopy at 400 magnification (McEvoy et al., 2005). Oocyst loading and manure application calendar (Table 2) was obtained by farmer interviews to capture the management in the catchments during the study period. However, in the case of manure stored in dung pits and limited operating machine, spreading often lasted 3e12 days (Table 2). The Cryptosporidium oocyst loadings from the dung and slurry in the study catchments (McDonald et al., in press) were adjusted according to dung heap storage time in the field and slurry storage time before the application. For instance, a die-off factor of about 0.025 (based on the data from Jenkins et al., 1999 and Peng et al., 2008) results in a reduction of about 22% of total oocysts in 10 days. It is estimated that a large proportion of oocysts die during storage. A factor of about 50% reduction had been multiplied to SWAT fertilizer database because on most farms in Ireland dung is stored for more than one month and slurry may be stored for about 10 days before application. In addition, for the purpose of the model one grazing cow/steer was assumed to excrete about 25 kg faeces per day based on the literature (Smith and Frost, 2000).

2.4.

Calibration and parameter estimation

The flow calibration period was 8 April 2004 to 7 April 2006 and the validation period was 8 April 2006 to 7 April 2008 for both catchments. Data from 1 January 2000 to 7 April 2004

were used prior to the calibration period to initialise the model. Initially manual calibration was used followed by automatic calibration (Eckhardt and Arnold, 2001) for secondary hydrologic parameters. SCE-UA method was used in auto-calibration (Eckhardt and Arnold, 2001) and LH-OAT method was used in sensitivity analysis (van Griensven et al., 2006). The best fitted hydrologic parameters for flow simulation were used for Cryptosporidium transport simulation in the microbial survival and transport sub-module. Cryptosporidium spp. transport parameters were manually calibrated in the NL catchment and validated in the SL catchment using observed data from 1 March 2007 to 31 June 2008. The parameters needed for sensitivity analysis included manure application, temperature, oocyst attachment to sediment and oocyst die-off processes associated with different media. Sensitivity analysis was performed manually by changing individual parameters by 10% of the initial value (keeping all other input parameters constant) and examining the resulting effect on the predicted Cryptosporidium spp. oocyst concentration. Based on the sensitivity analysis, parameters were manually calibrated to achieve the best fit between observation and simulation for the NL catchment and validated with the SL catchment.

2.5.

Statistical analysis

The SWAT2005 outputs were evaluated by comparison with observed data. The slope of the regression line (b), the coefficient of determination (R2) the Nash-Sutcliffe model efficiency (E ) (Nash and Sutcliffe, 1970) and root mean squared error (RMSE) were used to evaluate the relationship between measured and predicted flow and Cryptosporidium spp. oocyst concentration. For validation of model results in this study, model performance was classified as excellent for R2 or E > 0.90, very good for R2 or E ¼ 0.75e0.89, good for R2 or E ¼ 0.50e0.74, fair for R2 or E ¼ 0.25e0.49, poor for R2 or E ¼ 0e0.24, and unsatisfactory for R2 or E < 0 (Parajuli et al., 2007). SWAT-CUP (calibration and uncertainty programs) (Abbaspour, 2008) was used to calculate the uncertainties involved. SUFI2 as an optimization algorithm was chosen for this study to deal with the uncertainties of stream flow prediction.

3672

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 5 e3 6 8 0

rainfall

Prediction

Observation 0

30 0.8

40 50

Calibration

0.6

60 0.4

70

-1

20

1.0

Rainfall (mm day )

10

3

-1

Streamflow (m s )

1.2

80

0.2

90 0.0

040408

040708

041008

050108

050408

rainfall

050708

Prediction

051008

060108

Observation 0

30 0.8

40 50

0.6

Validation

60

0.4

70

-1

20

1.0

Rainfall (mm day )

10

3

-1

Streamflow (m s )

1.2

80

0.2

90 0.0 060408

060708

061008

070108

070408

070708

071008

080108

Fig. 2 e Stream flow during the calibration and validation period in North Leinster catchment.

3.

Results and discussion

3.1.

Catchment hydrology

The manually measured storm flow events indicated the stream flows in both catchments had a relative quick response over the storm events (Table 3), for instance on 11 November 2009 in NL catchment and 6 October 2009 in SL catchment. The flow peaks were buffered by the longer concentration time within the larger reference catchments, and after the adjustment, flow peaks for the smaller study catchments were underestimated while base flow conditions were overestimated when compared to flow measurements (Table 3). It is reasonable to expect that peak flow and exiting of the excess water after rainfall occurs earlier in a small catchment than in a larger one. However, the calculated reference daily flow rates were of the same magnitude as the measured daily flow rates for both catchments (Table 3). The difference between calculated daily flow-based on neighbouring gauged catchments and measured daily flow within the study catchments was sufficiently small. Thus, the reference method provided a workable approach to establishing the hydrological conditions of the ungauged catchments using neighbouring catchment records and meteorological data.

The most sensitive model parameters were ranked by the mean variance based on the reference stream flow for the two catchments (Table 4). Initial American Soil Conservation Service (SCS) runoff curve number for moisture condition II (CN2) was the most sensitive parameter, followed by soil depth (SOL_Z), soil evaporation compensation factor (ESCO), available water capacity (SOL_AWC), average slope steepness (SLOPE), saturated hydraulic conductivity (SOL_K), base flow alpha factor (ALPHA_BF) and maximum canopy storage (CANMX) (Table 4). Other parameters related to the surface or subsurface processes in these two catchments were insensitive and ranked as near zero or zero, only ALPHA_BF accounted for the base flow calibration. Runoff curve number for moisture condition II (CN2) was the most important parameter during the surface flow calibration (Kannan et al., 2007; Parajuli et al., 2009a; Chin et al., 2009), and as it is a semi-physical parameter may not represent processes in natural catchments all that well. A higher CN2 value for the SL catchment than NL catchment (Table 4) was due to clay characteristics of the acid brown earth in the SL catchment. The calibrated SWAT model simulation (Table 5) of the monthly stream flow showed good agreement during the calibration validation periods, but was less accurate when used to calculate the daily flow (Figs. 2 and 3). However, with

3673

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 5 e3 6 8 0

rainfall

Prediction

Observation 0

3

-1

30

0.8

40

Calibration

0.6

50 60

0.4

70

-1

Streamflow (m s )

20

Rainfall (mm day )

10 1.0

80

0.2

90 0.0 040708

041008

050108

050408

rainfall

050708

051008

Prediction

060108

Observation

0 10

30 0.8

40 50

Validation

0.6

60 0.4

70

-1

20

1.0

3

-1

Streamflow (m s )

1.2

Rainfall (mm day )

040408

80

0.2

90 0.0

060408

060708

061008

070108

070408

070708

071008

080108

Fig. 3 e Stream flow during the calibration and validation period in South Leinster catchment.

limited availability of data, a monthly time period was considered the most appropriate for comparing and presenting the results. The slightly underestimated flow peaks during extreme storm events (Figs. 2 and 3) were consistent with the buffering effects in calculating reference data from the neighbouring catchments due to catchment area differences. Overall, the SWAT hydrological simulation indicated runoff

coefficients (total simulated flow divided by total rainfall) of 0.54 for the NL catchment and 0.42 for the SL catchment, which was consistent with the results from other Irish agricultural catchments (Jordan et al., 2005). The monthly hydrological characteristics (Fig. 4) showed that the stream flow and runoff coefficient decreased from May to September (summer), suggesting high intensity of evapotranspiration Runoff coefficient

Streamflow 150

North Leinster catchment

120

1.0

60

0.5

30 0.0

150

South Leinster catchment

120

1.0

Runoff coefficient

Monthly streamflow (mm)

90

90 60

0.5

30 0

0.0 1

2

3

4

5

6

7

8

9

10

11

12

Month

Fig. 4 e Monthly predicted stream flow and runoff coefficient in the North Leinster and South Leinster catchments. Note: The error bars indicate the deviations of streamflow during the simulation period (2004e2008).

3674

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 5 e3 6 8 0

-10% at BACTKDDB=0.2 +10% at BACTKDDB=0.2

Cryptosporidium spp. transport in waters. Especially, in regions with small catchments and limited gauging data this approach could be applied provided the catchments have similar climate, geology, soils and land management.

10

3.2.

0

-30

PS

WD

PF KDQ ACT OFP PDB _KG ACE T F B W CT CT FR SUR TH BA BA _ T FR

WD

-40 -50 -60 -70 -80

Fig. 5 e Sensitivity analysis of SWAT2005 when modeling Cryptosporidium oocyst transport.

which may reach about 80 mm (Mills, 2000), which accounted for over 80% of total precipitation and cause water deficit during the months. Compared to flow regimes in summer months, the high runoff coefficients in the winter months suggested relatively low evapotranspiration as about 10 mm under cold conditions (Mills, 2000) and possible soil water recharge. However, the groundwater amount couldn’t be estimated out based on the collected dataset in the small catchments as the there is no available groundwater observation point. The ratio of base flow amount to total stream flow simulated by SWAT2005 was 0.55 for NL catchment and 0.51 for SL catchment during the total simulation period, which was a little (about 0.06) higher than the base flow ratio under observed stream flow conditions. It was concluded that SWAT2005, when run using reference data from a neighbouring catchment could satisfactorily predict the hydrological regime of the two catchments in order to predict

Slurry Dung

The sensitivity analysis indicated that the temperature adjustment factor (THBACT) was the most sensitive parameter in the SWAT microbial module (Fig. 5). The default value of 1.07 achieved a best fit, and was similar to previous reports by modelling of fecal bacteria (Parajuli, 2007). The input parameters, initial concentration of Cryptosporidium oocyst in animal waste (BACTPDB) and manure application amount (FRT_KG) caused proportional changes in output while FRT_SURFACE had no contribution to output variation. The sensitivity analysis also demonstrated distinct output variations due to changes of microbial die-off in soil solution (WDPQ) or absorbed to soil particles (WDPS) and microbial partitioning coefficient (BACTKDQ) at different absorption status indicated by the parameter BACTKDDB. It was shown that BACTKDDB had an interaction with WDPQ, BACTKDQ and WDPS sensitivity. Previous modelling of microbial pathogen transport using SWAT has indicated that the temperature adjustment factor (THBACT) and bacteria partition coefficient (BACTKDQ) partitioning between top 10 mm of soil and surface runoff when combined with a fixed value for BACTKDDB are most important during calibration (Parajuli et al., 2007). This study indicates that BACTKDDB played important roles in simulating Cryptosporidium oocyst transport as a greater portion of oocysts seemed to be absorbed to soil particles and retained in the soil (Mawdsley et al., 1996; Tufenkji et al., 2006; Kim et al., 2010). It can be verified by that when BACTKDDB was at 0.2, meaning more oocysts absorbed by soil particles, WDPS became more sensitive than BACTKDDB was at 0.8 (Fig. 5), indicating more oocysts died off in the form being absorbed by soil particles when BACTKDDB

Observed oocyst Predicted oocyst

Stream flow

0.0

60000 4 3 2

50000 -1

-20

PQ

WD

40000 30000

-1

Streamflow (m s )

0.1

3

20000 -1

-10

Cryptosporidium oocyst transport

0.2

0.3

0.5

Oocysts 100ml

Percentage of output change

200 150 100 50

10000

1500 1000

0.4 500 0

07 04 08 07 05 08 07 06 08 07 07 08 07 08 08 07 09 08 07 10 08 07 11 08 07 12 08 08 01 08 08 02 08 08 03 08

0.0

Manure application in subbasin (L ha )

-10% at BACTKDDB=0.8 +10% at BACTKDDB=0.8

Fig. 6 e Simulation of Cryptospordium spp. oocysts transport in North Leinster catchment.

3675

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 5 e3 6 8 0

Observed oocyst Predicted oocyst

Streamflow 60000

5 4 3 2

0.05

50000 40000

-1

0.00

1.0 20000 -1

0.8

0.15

0.7 0.20

0.6 0.5

0.25

0.4 0.30

Oocysts 100ml

3

30000

0.9

-1

Streamflow (m s )

0.10

10000

1500 1000

0.3 0.2

0.35

500

0.1 0.0

0

07 04 08 07 05 08 07 06 08 07 07 08 07 08 08 07 09 08 07 10 08 07 11 08 07 12 08 08 01 08 08 02 08 08 03 08

0.40

Manure application in subbasin (L ha )

Slurry Dung

Fig. 7 e Simulation of Cryptospordium spp. oocysts transport in South Leinster catchment.

was at 0.2; Vice versa, when BACTKDDB was at 0.8, meaning less oocysts absorbed by soil particles, more oocysts died off in soil solution. In addition, in the studied catchment, both BACTKDQ and WDPS (soil sorbed persistent bacteria die-off factor) exceeded the default setting provided with SWAT2005 (Table 6) for both catchments. BACTKDQ was 4800 and WDPS was set as 1.4 day
1. A higher BACTKDQ value indicates that less oocysts were transported via surface runoff and subsequently the solution in surface 10 mm of soil contained more oocysts and could be potentially sequestrated by soil and degraded. The higher value of WDPS can only represent higher proportion of oocysts died off in the absorbed form by soil particle and more oocysts were probably trapped by macro soil pores and cracks in the catchment. Despite large deviation from other bacteria parameterized modelling (Chin et al., 2009; Coffey et al., 2010a,b; Parajuli et al., 2007), these

parameters about oocyst transport corresponded well with each other and suggested that the soils in the catchment had significant function of retaining and inactivating a large number Cryptosporidium oocyst through coagulation, sedimentation and filtration (Smith and Rose, 1998; Weiss et al., 2005; Tufenkji et al., 2006; Kim et al., 2010). The predicted temporal dynamics of Cryptosporidium oocyst concentration in stream flow at the catchment outlets (Figs. 6 and 7, using parameters in Table 6) showed no distinct seasonal patterns. The highest oocyst concentrations were present after manure applications that coincided with heavy rainfall and runoff events. There is a positive correlation between stream flow and observed oocyst concentration in the NL catchment (R ¼ 0.49, P ¼ 0.02) but no significant relationship for the SL catchment (R ¼ 0.21, P > 0.05). Flowbased transform of predictions for comparison with

Predicted oocyst

South Leinster catchment

North Leinster catchment

0.6 -1

Predicted oocyst (count 100ml )

Fitted line

Y=1.43*X-0.0377 2 R =0.38, E=-2.57 RMSE=0.12, P<0.01

Y=0.54*X+ 0.033 2 R =0.20, E=-0.37 RMSE=0.07, P<0.05

0.5 0.4 0.3 0.2 0.1

0.1

0.2

0.3

0.4

0.5

0.6

0.1

0.2

0.3

0.4

0.5

0.6

-1

Observed oocyst (count 100ml )

Fig. 8 e The relationship between observed and predicted Cryptosporidium spp. oocysts concentrations in the north and south Leinster catchments.

3676

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 5 e3 6 8 0

Table 7 e Simulated of oocysts export in the north and south Leinster catchments during the water year of 2007e2008.

1024 915

Outflow (mm)

Oocyst input (count) 12

observations relies on a positive correlation between microbial concentration and stream flow being observed (Parajuli et al., 2009a; Coffey et al., 2010a). During this study two non-flow related observations were made, one with high Cryptosporidium oocyst concentration during predicted zero flow conditions in NL catchment, which perhaps occurred due to ocassionally direct animal access to the stream, and one with very low concentration during high stream flow. Only one location where no fence existed and allowed animals approach the stream waters might coincidently rotated to graze and could be the possible source area of direct excreted feces. Without flow-based transformation, predicted oocyst concentration at the outflow of both catchments matched observed measurements (Fig. 8) with significant correlation (R2 ¼ 0.20, P < 0.05 for the NL catchment and R2 ¼ 0.38, P < 0.01 for the SL catchment), unsatisfactory efficiency and reasonable deviation (E ¼
0.37, RMSE ¼ 0.07 for NL catchment; E ¼
2.57, RMSE ¼ 0.12 for SL catchment). Although the model results showed negative efficiency for the two catchments, the slopes (b) of 0.54 and 1.43 are acceptable in terms of the challenging difficulties especially with enumeration of Cryptosporidium oocysts (Medema and Schijven, 2001; Dorner et al., 2004, 2006). Fig. 9 shows an oocyst discharge duration curve for the simulation period (2004e2008) and illustrates the exceedance rate of oocysts within the catchment. When the potential infective dose for

6.6  10
5 3.4  10
4

6.4  10 1.9  109

humans (1 oocyst l
1) (Coffey et al., 2010b) is considered in the curve, results show that this rate was exceeded 7.1% of the time in NL catchment and 9.9% of the time in SL catchment. Despite being safe for 90% of the time, the risk of water contamination is still serious in relation to high oocyst concentrations which occurred during and after specific rainfall events. The amount of oocyst load in 10% of the time accounted for 66.1% and 91.7% of the total load exported from the catchment (Fig. 9). Within the studied catchments, the hydrologic response units (HRUs) where manure applied exported higher Cryptosporidium oocyst loads than the units without manure application. As the effluent from grazing fields contained very low concentrations of Cryptosporidium oocysts (Table 2), or even as low as 3.2 oocyst g
1 faeces (Fayer et al., 2000; Atwill et al., 2003), the loading from grazed field was considered negligible in both NL and SL catchments. A higher ratio of oocyst loss to input was predicted in the SL catchment compared to the NL catchment (Table 6). The simulations indicated that these two agricultural catchments could intercept and inactivate large amounts of Cryptosporidium occysts with a removal rate of >3.5 log. This is consistent with removal rates reported in the UK that ranged from 1.55 to 3 log (Anon, 1995). Despite <0.034% of the oocyst loading produced by animal excretion being exported from the catchments, it was still a large amount (about 109 oocysts) and could therefore be regarded as

NL catchment

Concentration Load

30

250 200 150

20

100 10

Oocysts l

-1

50

30

Ratio of loss (100%)

8

9.7  10 5.6  1012

542 365

Oocyst exported (count)

SL catchment

Concentration Load

250 200 150

20

-1

North Leinster South Leinster

Rainfall (mm)

Oocyst Load (1000,000 oocysts d )

Catchment

100 10 50

0.1

0.2

0.3

0.6 0.8

Percent of time that indicated discharge was equaled or exceeded Fig. 9 e Simulated oocyst discharge duration curves in the north and south Leinster catchments during the simulation period.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 5 e3 6 8 0

a potential threat to human health if the water was to be used as a source of potable water.

3.3.

Implications and interpretation

Most microbial transport models have only examined faecal indicators, which may or may not be an indication of contamination with pathogenic micro-organisms such as Crytosporidium spp. oocysts (Lemarchand and Lebaron, 2003). Walker and Stedinger (1999) modelled pathogens such as Cryptosporidium directly, but this study taken together with further studies from the Netherlands and Canada (Medema and Schijven, 2001; Dorner et al., 2006) all indicate highly variable results, perhaps due to limited microbial input data. A major issue with modelling Cryptosporidium spp. in natural ecosystems is the efficiency of the routine isolation techniques used to recover Cryptosporidium oocysts from untreated surface waters. This can vary from no recovery to about 50% oocyst recovery (Hansen and Ongerth, 1991; Nieminski et al., 1995; Watanabe, 1996; Smith and Rose, 1998; Graczyk et al., 2000). The oocyst concentration data used in this study were associated with recoveries of about 25% (unpublished results), thus the relationship between stream flow and oocyst concentration is open to question and may give rise to false negatives (Dorner et al., 2006) and subsequently to the uncertainties when calibrating the model. Furthermore, uncertainties may also originate from the weather data, such as precipitation, which even in very small catchments can give rise to an error of up 5% (Shirmohammadi et al., 2006). About the total uncertainties of hydrological component, the preliminary application of SWAT-CUP showed that 30% of measured data were bracketed within the 95% prediction uncertainty (95PPU) (Abbaspour et al., 2004) setting measured errors of 10 percent, which may be due to the buffered effects on catchment flow caused by the referrence method and the uncertainties derived from currently encountered serious issues (Abbaspour, 2008). However, the statistics still showed that in this study the hydrological component of the model performed reasonably well (Figs. 2 and 3) and the relationship between observations and predictions for oocyst transport (Figs. 6 and 7) was consistent for the known physical processes. Logistics did not allow for constant monitoring of oocyst concentrations so the background level of Cryptosporidium spp. contamination, from point sources such as animal access to water and leaching from dung heaps during times of normal flow was still unknown. Cryptosporidium oocysts are small (4e6 mm diameter) and have a low specific gravity (1.05 g/cm3), so their movement in surface waters is generally not considered to be influenced by gravitational settling (Searcy et al., 2005). However, despite the proportion of oocyst attached to soil particles might reach about 75% of total oocyst in experimental static waters (Medema et al., 1998; Searcy et al., 2005), the turbulence and scouring processes disturbed the sedimentation of attached oocysts in the streams in NL and SL catchments because of the distinct sloping features with average slope gradients about 6% in both catchments. However, the sedimentation in the streams may still bring some uncertainties in the estimation of oocyst transport depending on the sediment type and water chemistry (Searcy et al., 2005). In terms of all the uncertainties existed in

3677

spatially distributed models like SWAT, Beven (1993,1995) suggested the solution of uncertainties inherent in hydrological modelling based on more available field investigation. The sampling protocol, a manual grab sample after storm events (McDonald et al., in press), only allowed one time point determination of oocyst concentration. The possible animal access to stream in few cases also increased modelling uncertainties in the catchments. Despite these limitations, it was shown that SWAT2005 in conjunction with sporadic sampling after heavy rainfall events can be used to better understand the process of Cryptosporidium contamination of surface waters in small agricultural catchments. In the work presented here, simulation data using the microbial module of SWAT2005 were of the same order of magnitude as the observed data and these were validated using a second independent catchment (Fig. 8). The results suggest that a reasonable prediction of Cryptosporidium oocyst contamination may be achieved in ungauged small agricultural catchments using SWAT2005 and reference data from nearby gauged catchments. The most important outcome of this study was the possibility that SWAT2005 can be used to simulate pathogen transport in a small ungauged catchments either to identify source hotspots of contamination, or to confirm the likelihood of a source or vector having occurred. This means that hydrological modelling has the potential to aid in identifying the source of waterborne outbreaks in downstream communities and could become a routine part of subsequent epidemiological investigations. Despite not being precise, the advantage of SWAT model is that it will identify the critical source areas of oocysts and discharging trends and loads. The modelling of Cryptosporidium in two small agricultural catchments indicates that SWAT2005 could be used to assist in predicting the temporal variability of waterborne risk to downstream populations. The model can mechanistically capture the coincidence of sources (animal population and manure/slurry spreading), transport vectors (catchment hydrology) and targets (stream flow) thus allowing a better understanding of when to modify animal husbandry practises in the catchment and where to allow certain activities or protections to be undertaken. For instance, the river bank and the adoption of riparian zone management could prevent access of animals to waterways and allow sediment absorbed Cryptosporidium oocysts to be retained in soil before reaching surface waters (Weiss et al., 2005; Bryan et al., 2009; Kim et al., 2010). The SWAT2005 modelling also showed that manure spreading followed by storm events contributed much more Cryptosporidium oocysts to waterways than grazing animals. The most risky period for Cryptosporidium oocyst contamination in surface water was the first storm event after manure spreading. This means manure application should be restricted to dry periods when surface runoff is limited. Recently, a slurry acceptance map of Northern Ireland was developed by Jordan et al. (2007) in which hydrology of soil types (HOST) classes were combined with slope, rockiness, flood hazard and soil moisture deficit classes. The authors recommended that period during the growing season when the soils were not saturated and there were no predictions of significant rainfall was suitable for slurry application. As the rainfall events may happen at any time during whole year in

3678

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 5 e3 6 8 0

Ireland (Met Eireann, the Irish National Meteorological Service), it is suggested here that more frequent application of smaller quantities of slurry or dung may prevent peaks in Cryptosporidium oocysts concentration and transport in surface water from occurring and might reduce the total numbers of oocysts exported from the catchment.

4.

Conclusions

Based on detailed sampling and investigation at two small agricultural catchments, this study has demonstrated the possiblity of using a spatially distributed model to describe Cryptosporidium oocyst transport within agriculture predominating areas. The SWAT model can provide an order of magnitude reliability based on reference catchment data, regional weather data and can be used to aid epidemiological investigations and plan catchment management for the benefit of public health. However, more attention should be paid on the storm events, especially after manure application to further validate the modelling tools.

Acknowledgement This work was funded by the European Commission Framework Programme 6 Marie Curie Transfer of Knowledge project “Cryptonet.ie” (MTKD-CT-2005-029454). We acknowledge the help given by Drs. X. Wang at Blackland Research Centre, Temple, Texas, F. Olivera from Texas A&M University, N. Sammons and G. Mitchell at Agricultural Research Service, United States Department of Agriculture.

references

Abbaspour, K.C., Johnson, C.A., van Genuchten, M. Th., 2004. Estimating uncertain flow and transport parameters using a sequential uncertainty fitting procedure. VADOSE ZONE Journal 3, 1340e1352. Abbaspour, K.C., 2008. SWAT-CUP, SWAT Calibration and Uncertainty Programs, A User Manual. Swiss Federal Institute of Aquatic Science and Technology (Eawag), Zurich, Switzerland. Anon, 1995. Cryptosporidium in Water Supplies. Second Report of the Group of Experts (Sir John Badenoch, Chairman). Department of the Environment, Department of Health. HMSO, London, UK. Arabi, M., Govindaraju, R.S., Hantush, M.M., Engel, B.A., 2006. Role of watershed subdivision on modelling the effectiveness of best management practices with SWAT. Journal of the American Water Resources Association 42, 513e528 (April). Arnold, J.G., Allen, P.M., 1999. Automated methods for estimating baseflow and ground water recharge from streamflow records. Journal of the American Water Resources Association 35, 411e424. Atwill, E.R., Hoar, B., Pereira, M.D.C., Tate, K.W., Rulofson, F., Nader, G., 2003. Improved quantitative estimates of low environmental loading and sporadic periparturient shedding

of Cryptosporidium parvum in adult beef cattle. Applied and Environmental Microbiology 69, 4604e4610. Benham, B.L., Baffaut, C., Zeckoski, R.W., Mankin, K.R., Pachepsky, Y.A., Sadeghi, A.A., Brannan, K.M., Soupir, M.L., Habersack, M.J., 2006. Modeling bacteria fate and transport in watersheds to support TMDLs. Transactions of the ASABE 49, 987e1002. Beven, K., 1993. Prophecy, reality and uncertainty in distributed hydrological modelling. Advances in Water Resources 16, 41e51. Beven, K., 1995. Linking parameters across scales: subgrid parameterizations and scale dependent hydrological models. Hydrological Processes 9, 507e525. Boughton, W., Chiew, F., 2007. Estimating runoff in ungauged catchments from rainfall, PET and the AWBM model. Environmental Modelling & Software 22, 476e487. Boyle, C., 2009. The Aricultural Catchments ProgrammedInnovation for Farming and the Environment. TResearch. Teagasc, Dublin, pp. 38e39. Bryan, B.A., Kandulu, J., Deere, D.A., White, M., Frizenschaf, J., Crossman, N.D., 2009. Adaptive management for mitigating Cryptosporidium risk in source water: a case study in an agricultural catchment in South Australia. Journal of Environmental Management 90, 3122e3134. Carey, C.M., Lee, H., Trevors, J.T., 2004. Biology, persistence and detection of Cryptosporidium parvum and Cryptosporidium hominis oocyst. Water Research 38, 818e862. Casemore, D.P., Wright, S.E., Coop, R.L., 1997. CryptosporidiosisdHuman and Animal Epidemiology. CRC Press, Boca Raton, FL. Casey, J.W., Holden, N.M., 2005. Analysis of greenhouse gas emissions from the average Irish milk production system. Agricultural Systems 86, 97e114. Chin, D.A., Sakura-Lemessy, D., Bosch, D.D., Gay, P.A., 2009. Watershed-scale fate and transport of bacteria. Transaction of the ASABE 52, 145e154. Coffey, R., Cummins, E., Cormican, M., Flaherty, V.O., Kelly, S., 2007. Microbial exposure assessment of waterborne pathogens. Human and Ecological Risk Assessment 13, 1313e1351. Coffey, R., Cummins, E., Bhreathnach, N., Flaherty, V.O., Cormican, M., 2010a. Development of a pathogen transport model for Irish catchments using SWAT. Agricultural Water Management 97, 101e111. Coffey, R., Cummins, E., Flaherty, V.O., Cormican, M., 2010b. Analysis of the soil and water assessment tool (SWAT) to model Cryptosporidium in surface water sources. Biosystems Engineering 106, 303e314. Crane, S.R., Moore, J.A., 1986. Modeling enteric bacterial die-off: a review. Water Air and Soil Pollution 27, 411e439. Current, W.L., Garcia, L.S., 1991. Cryptosporidiosis. Clinical Microbiology Reviews 4, 325e358. Diamond, J., Sills, P., 2001. Soil Water Regimes. End of project report ARMIS 4479. Teagasc, Johnstown Castle Research Centre, Wexford, ISBN 1841702234. Dorner, S.M., Anderson, W.B., Slawson, R.M., Kouwen, N., Huck, P. M., 2006. Hydrologic modeling of pathogen fate and transport. Environmental Science & Technology 40, 4746e4753. Dorner, S.M., Huck, P.M., Slawson, R.M., 2004. Estimating potential environmental loadings of Cryptosporidium spp. And Campylobacter spp. From livestock in the Grand River watershed, Ontario, Canada. Environmental Science & Technology 38, 3370e3380. Dunne, L., O’Riordan, E., Troy, D., 2004. Foresight for the Beef Industry 2010. Teagasc, Rural Economy Research Centre, 19 Sandymount Avenue, Dublin 4, Ireland. Eckhardt, K., Arnold, J.G., 2001. Automatic calibration of a distributed catchment model. Journal of Hydrology 251, 103e109.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 5 e3 6 8 0

Fayer, R., 2004. Cryptosporidium: a water-borne zoonotic parasite. Veterinary Parasitology 126, 37e56. Fayer, R., 2008. General biology. In: Fayer, R., Xiao, L. (Eds.), Cryptosporidium and Cryptosporidiosis, second ed. CRC Press and IWA Publishing, Boca Raton, FL, pp. 1e42. Fayer, R., Trout, J.M., Graczyk, T.K., Lewis, E.J., 2000. Prevalence of Cryptosporidium, Giardia and Eimeria infections in post-weaned and adult cattle on three Maryland farms. Veterinary Parasitology 93, 103e112. Ferguson, C.M., Croke, B., Ashbolt, N.J., Deere, D.A., 2005. A deterministic model to quantify pathogen loads in drinking water catchments: pathogen budget for the Wingecarribee. Water Science and Technology 52, 191e197. Feyereisen, G.W., Strickland, T.C., Bosch, D.D., Sullivan, D.G., 2007. Evaluation of SWAT manual calibration and input parameter sensitivity in the little river watershed. Transactions of the ASABE 50, 843e855. Finch, T.F., Gardiner, M.J., Comey, A., Radford, T., 1983. Soils of Co. Meath. The Agricultural Institute, Dublin. Gardiner, M.J., Radford, T., 1980. Soil Associations of Ireland and Their Land Use Potential. Mount Salus Press Limited, Dublin. Goh, S., Reacher, M., Casemore, D.P., Verlander, N.Q., Chalmers, R., Knowles, M., Williams, J., Osborn, K., Richards, S., 2004. Sporadic cryptosporidiosis, north Cumbria, England, 1996e2000. Emerging Infectious Diseases 10, 1007e1015. Graczyk, T.K., Evans, B.M., Shiff, C.J., Karreman, H.J., Patz, J.A., 2000. Environmental and geographical factors contributing to watershed contamination with Cryptosporidium parvum oocysts. Environmental Research 82, 263e271. Haas, C.N., Crockett, C.S., Rose, J.B., Gerba, C.P., Fazil, A.M., 1996. Assessing the risk posed by oocysts in drinking water. Journal American Water Works Association 88, 131e136. Hansen, J.S., Ongerth, J.E., 1991. Effects of time and watershed characteristics on the concentration of Cryptosporidium oocysts in River water. Applied and Environmental Microbiology 57, 2790e2795. HPSC, 2007. Cryptosporidiosis Outbreak in the South East, 2005. Epi-Insight. Health Protection Surveillance Centre, Ireland, Ireland, p. 1 & 4. Jenkins, M.B., Walker, M.J., Bowman, D.D., Anthony, L.C., Ghiorse, W.C., 1999. Use of a sentinel system for field measurements of Cryptosporidium parvum oocyst inactivation in soil and animal waste. Applied and Environmental Microbiology 65, 1998e2005. Jordan, C., Higins, A., Wright, P., 2007. Slurry acceptance mapping of Northern Ireland for runoff risk assessment. Soil Use and Management 23, 245e253. Jordan, P., Menary, W., Daly, K., Kiely, G., Morgan, G., Byrne, P., Moles, R., 2005. Patterns and processes of phosphorus transfer from Irish grassland soils to rivers - integration of laboratory and catchment studies. Journal of Hydrology 304, 20e34. Kannan, N., White, S.M., Worrall, F., Whelan, M.J., 2007. Sensitivity analysis and identification of the best evapotranspiration and runoff options for hydrological modelling in SWAT-2000. Journal of Hydrology 332, 456e466. Kim, H.N., Walker, S.L., Bradford, S.A., 2010. Coupled factors influencing the transport and retention of Cryptosporidium parvum oocysts in saturated porous media. Water Research 44, 1213e1223. Lemarchand, K., Lebaron, P., 2003. Occurrence of Salmonella spp. and Cryptosporidium spp. in a French coastal watershed: relationship with fecal indicators. Fems Microbiology Letters 218, 203e209. Lloyd, A., Drury, D., 2002. Continuous monitoring for Cryptosporidiumda novel approach to public health protection. Water Science and Technology 46, 297e301.

3679

Mawdsley, J.L., Brooks, A.E., Merry, R.J., 1996. Movement of the protozoan pathogen Cryptosporidium parvum through three contrasting soil types. Biology and Fertility of Soils 21, 30e36. McDonald, S., Murphy, T., Holden, N.M. Qualitative risk assessment of surface water contamination with Cryptosporidium spp. oocysts: a case study of three agricultural catchments. Human and Ecological Risk Assessment, in press. McElroy, A.D., Chiu, S.Y., Nebgen, J.W., Aleti, A., Bennett, F.W., 1976. Loading Functions for Assessment of Water Pollution from Nonpoint Sources. EPA document EPA 600/2-76-151. USEPA, Athens, GA. McEvoy, J.M., Duffy, G., Moriarty, E.M., Lowery, C.J., Sheridan, J.J., Blair, I.S., McDowell, D.A., 2005. The prevalence and characterization of Cryptosporidium spp. in beef abattoir water supplies. Water Research 39, 3697e3703. McGoff, N., Eaton, J.M., Byrne, K., Leahy, P., Kiely, G., 2007. Measurements of Soil Bulk Density across Differing Soil Types and Land Uses in Ireland. Greenhouse Gas Fluxes in Terrestrial Ecosystems in Ireland. Delgany, Co., Wicklow, Ireland. Medema, G.J., Schets, F.M., Teunis, P.F.M., Havelaar, A.H., 1998. Sedimentation of free and attached Cryptosporidium oocysts and Giardia cysts in water. Applied and Environmental Microbiology 64, 4460e4466. Medema, G.J., Schijven, J.F., 2001. Modelling the sewage discharge and dispersion of Cryptosporidium and Giardia in surface water. Water Research 35, 4307e4316. Mills, G., 2000. Modelling the water budget of Irelandeevapotranspiration and soil moisture. Irish Geography 33, 99e116. Moore, J.A., Smyth, J.D., Baker, E.S., Miner, J.R., Moffit, D.C., 1989. Modeling bacteria movement in livestock manure systems. Transactions of the ASABE 31, 1049e1053. Nash, J.E., Sutcliffe, J.V., 1970. River flow forecasting through conceptual models. Part 1: a discussion of principles. Journal of Hydrology 10, 282e290. Ndomba, P., Mtalo, F., Killingtveit, A., 2008. SWAT model application in a data scarce tropical complex catchment in Tanzania. Physics and Chemistry of the Earth 33, 626e632. Neitsch, S.L., Arnold, J.G., Kiniry, J.R., Srinivasan, R., Williams, J.R., 2004. Soil and water assessment tool, Input/output file documentation, version 2005. Blackland Research Center, Grassland, Soil and Water Research Laboratory, Agricultural Research Service, Temple, TX. Nieminski, E.C., Schaefer, F.W., Ongerth, J.E., 1995. Comparison of 2 methods for detection of Giardia cysts and Cryptosporidium oocysts in water. Applied and Environmental Microbiology 61, 1714e1719. Oliver, D.M., Heathwaite, A.L., Fish, R.D., Chadwick, D.R., Hodgson, C.J., Winter, M., Butler, A.J., 2009. Scale appropriate modelling of diffuse microbial pollution from agriculture. Progress in Physical Geography 33, 358e377. Ong, C., Moorehead, W., Ross, A., IsaacRenton, J., 1996. Studies of Giardia spp and Cryptosporidium spp in two adjacent watersheds. Applied and Environmental Microbiology 62, 2798e2805. Parajuli, P.B., 2007. Swat Bacteria Sub-model Evaluation and Application. Department of Biological and Agricultural Engineering, College of Engineering. Kansas State University, Manhattan, Kansas, USA. Parajuli, P.B., Mankin, K.R., Barnes, P.L., 2009a. Source specific fecal bacteria modeling using soil and water assessment tool model. Bioresource Technology 100, 953e963. Parajuli, P.B., Mankin, K.R., Barnes, P.L., Rossi, C.G., 2009b. Fecal bacteria source characterization and sensitivity analysis of SWAT2005. Transactions of the ASABE 52, 1847e1858. Parajuli, P.B., Mankin, K.R., Barnes, P.L., Green, C.H., 2007. Fecal Bacteria Source Characterization and Sensitivity Analysis of

3680

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 6 5 e3 6 8 0

SWAT2005. In: 2007 ASABE Annual International Meeting, Minneapolis, Minnesota, pp. 1e24. Park, C.H., Huck, P.M., 2003. A conceptual model of Cryptosporidium transport in watersheds. Water Quality Research Journal of Canada 38, 77e113. Peng, X., Murphy, T., Holden, N.M., 2008. Evaluation of the effect of temperature on the die-off rate for Cryptosporidium parvum oocysts in water, soils, and feces. Applied and Environmental Microbiology 74, 7101e7107. Reddy, K.R., Khaleel, R., Overcash, M.R., 1981. Behavior and transport of microbial pathogens and indicator organisms in soils treated with organic wastes. Journal of Environmental Quality 10, 255e266. Rose, J.B., 1997. Environmental ecology of Cryptosporidium and public health implications. Annual Review of Public Health 18, 135e161. Rose, J.B., Huffman, D.E., Gennaccaro, A., 2002. Risk and control of waterborne cryptosporidiosis. Fems Microbiology Reviews 26, 113e123. Samadder, S.R., Ziegler, P., Murphy, T.M., Holden, N.M., 2010. Spatial distribution of risk factors for Cryptosporidium spp. transport in an Irish catchment. Water and Environment Research 82, 750e758. Searcy, K.E., Packman, A.I., Atwill, E.R., Harter, T., 2005. Association of Cryptosporidium parvum with suspended particles: impact on oocyst sedimentation. Applied and Environmental Microbiology 71, 1072e1078. Shirmohammadi, A., Chaubey, I., Harmel, R.D., Bosch, D.D., MunozCarpena, R., Dharmasri, C., Sexton, A., Arabi, M., Wolfe, M.L., Frankenberger, J., Graff, C., Sohrabi, T.M., 2006. Uncertainty in TMDL models. Transactions of the ASABE 49, 1033e1049. Smith, H.V., Rose, J.B., 1998. Waterborne cryptosporidiosis: current status. Parasitology Today 14, 14e22. Smith, K.A., Frost, J.P., 2000. Nitrogen excretion by farm livestock with respect to land spreading requirements and controlling nitrogen losses to ground and surface waters. Part 1: cattle and sheep. Bioresource Technology 71, 173e181. Stewart, M.H., Ferguson, D.M., DeLeon, R., Taylor, W.D., 1997. Monitoring Program to Determine Pathogen Occurrence in Relationship to Storm Events and Watershed Conditions. In:

Proceedings of the WQTC 1997, Am. Water Works Association. Denver, USA. Tate, K.W., Atwill, E.R., George, M.R., McDougald, M.K., Larsen, R. E., 2000. Cryptosporidium parvum transport from cattle fecal deposits on California rangelands. Journal of Range Management 53, 295e299. 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, 860e869. Toner, P.F., 1986. Impact of agriculture on surface-water in Ireland .1. General. Environmental Geology and Water Sciences 9, 3e10. Tufenkji, N., Dixon, D.R., Considine, R., Drummond, C.J., 2006. Multi-scale Cryptosporidium/sand interactions in water treatment. Water Research 40, 3315e3331. van Griensven, A., Meixner, T., Grunwald, S., Bishop, T., Diluzio, A., Srinivasan, R., 2006. A global sensitivity analysis tool for the parameters of multi-variable catchment models. Journal of Hydrology 324, 10e23. Veith, T.L., Srinivasan, M.S., Gburek, W.J., 2003. Process representation in watershed-scale hydrologic models: an evaluation in an experimental watershed. In: Proceedings First Interagency Conference on Research in the Watersheds. Oct. 27e30, 2003, Benson, AZ., pp. 242e247. Walker, F.R., Stedinger, J.R., 1999. Fate and transport model of Cryptosporidium. Journal of Environmental Engineering-Asce 125, 325e333. Wanielista, M., Kersten, R., Eaglin, R., 1997. Hydrology Water Quantity and Quality Control, second ed. John Wiley & Sons. Watanabe, M.E., 1996. New Cryptosporidium testing methods. Environmental Science & Technology 30, A532eA535. Weiss, W.J., Bouwer, E.J., Aboytes, R., LeChevallier, M.W., O’Melia, C.R., Le, B.T., Schwab, K.J., 2005. Riverbank filtration for control of microorganisms: results from field monitoring. Water Research 39, 1990e2001. Williams, J.R., Hann, R.W., 1978. Optimal Operation of Large Agricultural Watersheds with Water Quality Constraints. Technique Report. No. 96. Texas Water Resources Institute, Texas A&M University.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 8 1 e3 6 9 1

Available at www.sciencedirect.com

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

Impact of ozonation on the genotoxic activity of tertiary treated municipal wastewater Miroslav Misı´k a, Siegfried Knasmueller a,*, Franziska Ferk a, Margit Cichna-Markl b, Tamara Grummt c, Heidi Schaar d, Norbert Kreuzinger d a

Institute of Cancer Research, Department of Internal Medicine I, Medical University of Vienna, Borschkegasse 8a, 1090 Vienna, Austria Institute of Analytical Chemistry, Vienna University, Vienna, Austria c Federal Environment Agency, Bad Elster, Germany d Institute for Water Quality, Resources and Waste Management, Vienna University of Technology, Vienna, Austria b

article info

abstract

Article history:

Ozonation is an emerging technology for the removal of micropollutants from treated

Received 28 September 2010

wastewater. Aim of the present study was to investigate the impact of ozone treatment on

Received in revised form

genotoxic and acute toxic effects of tertiary treated municipal wastewater. It is known that

7 April 2011

DNA-damaging chemicals cause adverse effects in the environment and that exposure to

Accepted 7 April 2011

humans leads to cancer and other diseases. Toxicity was tested in organisms from three

Available online 28 April 2011

trophic levels namely in bacteria (Salmonella/microsome assays) which enable the detection of gene mutations, in a plant bioassay (micronucleus assay with root tip cells of Allium

Keywords:

cepa) which reflects clastogenic and aneugenic effects and in single cell gel electrophoresis

Ozonation

(SCGE) tests with mammalian cells which detect DNA migration caused by single-, double

Wastewater treatment

strand breaks and alkali labile sites. In the bacterial tests negative results were obtained

Genotoxicity

with untreated samples but after concentration with C18 cartridges a positive result was

SCGE

found in strains TA1537 and TA98 which are sensitive to frameshift mutagens while no mutations were induced in other tester strains (TA100, TA102 and YG1024). Ozone treatment led to a decrease of the mutagenic activity of the samples. In the SCGE experiments, DNA migration was detected with the unconcentrated effluent of the treatment plant and ozonation led to a substantial decrease of this effect. In the plant bioassays, negative results were obtained with the effluent and ozone treatment did not cause an alteration of the micronucleus frequencies. Also acute toxic effects were monitored in the different indicator organisms under all experimental conditions. The bacteriocidal/bacteriostatic effects which were seen with the concentrated samples were reduced by ozonation. In the experiments with the eukaryotic (plant and animal) cells no acute toxicity was seen with the effluents and ozonation had no impact on their viability. In conclusion findings of this study indicate that ozonation of tertiary effluents of a municipal treatment plant reduces the adverse effects caused by release of mutagens in aquatic ecosystems and does not decrease the viability of bacteria and eukaryotic cells. However, future research is required to find out if, and to which extent these findings can be generalized and which mechanisms account for the detoxification of the wastewater. ª 2011 Elsevier Ltd. All rights reserved.

* Corresponding author. Tel.: þ43 1427765142; fax: þ43 142779651. E-mail address: [email protected] (S. Knasmueller). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.015

3682

1.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 8 1 e3 6 9 1

Introduction

The impact of ozonation on the toxic properties of waters is an important issue. On the one hand ozonation is used as an alternative to other drinking water disinfection procedures such as chlorination (Monarca et al., 2000), on the other it is applied for the removal of micropollutants such as personal care products, pharmaceuticals and endocrine disruptors from treated wastewater (Bozkaya-Schrotter et al., 2009; Hollender et al., 2009; Joss et al., 2008; Schaar et al., 2010; Stalter et al., 2010a). Ozonation of municipal wastewater effluents does not lead to complete degradation of these chemicals (see for example Schaar et al., 2010) and a number of products such as aldehydes, carboxylated compounds and brominated organics which may cause adverse effects in aquatic ecosystems and in humans are formed as a consequence of the treatment (Hu et al., 1999). Furthermore, it is known that ozone decomposes in aqueous solutions and forms reactive radicals which attack macromolecules including DNA and cause mutations in a variety of indicator organisms (for review see Victorin, 1992, 1996). In this context, it is notable, that a number of studies indicate that ozonation of treated wastewater causes toxic effects in invertebrates and affects the development and reproduction of fish (Paraskeva and Graham, 2002; Petala et al., 2006a; Stalter et al., 2010a, 2010b). The aim of the present study was to investigate the impact of ozonation on the genotoxic properties of municipal water treatment plant effluents. It has been shown that release of genotoxins in the environment reduces the fertility of species and affects the stability of ecosystems (Bickham et al., 2000; Medina et al., 2007; Moore et al., 2004; Peru and Doledec, 2010). Furthermore, it is known that damage of somatic cells in humans causes cancer and other diseases and accelerates ageing (Ames, 1989; Shaugnessy and DeMarini, 2009). Most earlier investigations concerning the genotoxic properties of ozonated wastewaters have been carried out with bacteria and the results are quite controversial, i.e. in some an increase of the mutagenic activities was found after ozonation (Collivignarelli et al., 2000; Monarca et al., 2000; Petala et al., 2008), while other findings indicated that the treatment leads to reduction of the mutagenic potencies of the waters (Nakamuro et al., 1989; Ono et al., 1992; Reungoat et al., 2010; Takanashi et al., 2002). The results of bacterial mutagenicity experiments are only partly relevant for humans. The cells are devoid of many drug metabolizing enzymes and the use of liver derived enzyme mixtures does reflects the situation in mammals only partly (Knasmueller et al., 2004). Furthermore, it is known that coliform indicator bacteria possess certain enzymes which may lead to overestimation of the mutagenic potencies of chemicals and complex mixtures (Knasmuller et al., 2002). In order to assess the impact of ozonation in a comprehensive way, we combined in the present study three genotoxicity tests, namely Salmonella/microsome assays, micronucleus (MN) experiments with root tip cells of plants (Allium cepa) and single cell gel electrophoresis (SCGE) assays with primary rat hepatocytes. This battery comprises organisms of all three trophic levels and the three system are complimentary in regard to their sensitivity to different classes of environmentally relevant DNA carcinogens (Helma et al., 1995; Knasmueller et al., 1998).

Bacterial tests detect halogenated compounds as well as organic solvents and pharmaceuticals, nitroaromatics and representatives of different classes of indirect mutagens. They were also included for reasons of comparison as they are the most widely used assays for routine testing of chemicals and complex environmental mixtures (Claxton and Woodall, 2007; Ohe et al., 2004; White and Claxton, 2004). Plant bioassays have a broad detection spectrum and are sensitive towards directly active mutagens such as pesticides and herbicides, radionuclides and heavy metals (Majer et al., 2005). Rat hepatocytes have the advantage that they possess phase I and phase II enzymes which are involved in the activation and detoxification of genotoxins in mammals and detect indirectly acting DNA reactive carcinogens (which require enzymatic activation) such as polycyclic aromatic hydrocarbons, aromatic amines and nitrosamines which have been found in waters (Ohe et al., 2004). The experiments with the liver cells have probably the highest predictive value in regard to the detection of effects that are relevant for humans. The latter two test systems are highly sensitive and concentration procedures which may lead to loss of active compounds are not required for experiments with waters. Also the endpoints of the different test systems are complementary; while bacterial indicators detect gene mutations, MN are formed as a consequence of structural and numerical chromosomal aberrations and SCGE assays reflect single and double strand breaks and apurinic sites (Nersesyan et al., 2009).

2.

Material and methods

2.1.

Chemicals

Bacterial media were from Difco (Detroit, MI) except Nutrient Broth No. 2 (used for overnight cultures) which came from Oxoid (Hampshire, UK). Inorganic salts, dimethylsulfoxide (DMSO), acetone, n-hexane and hydrochloric acid (HCl) were from Merck (Darmstadt, Germany). Nicotinamide adenine dinucleotide phosphate, glucose-6-phosphate, methyl methanesulfonate (MMS), 2-amino-anthracene (2-AA), 2,4,7-trinitro-9-fluorenone (2,4,7-TNF), [6-chloro-9-(3-[2-chloro-ethylamino]propylamino)2-methoxyacridine]-dihydro-chloride (ICR 191), and sodium azide (NaN3) came from Sigma Aldrich (Steinheim, Germany). Aroclor 1254 induced S9 mix was purchased from MP Biomedicals (Illkirch, France). Chemicals for the isolation and cultivation of hepatocytes [N-2-hydroxyethylpiperazine-N0 -2-ethanesulf onic acid, collagenase type IV, modified Eaglesmedium (MEM) and Heparin-Na] were purchased from Sigma Aldrich (Steinheim, Germany). Low melting point agarose and normal melting point agarose were obtained from Gibco (Paisley, UK). Trisma base, Triton X, ethidium bromide, arsenic trioxide and trypan blue were purchased from Sigma Aldrich (Steinheim, Germany), blue rayon came from Funakoshi Inc. (Tokyo, Japan). The male Him-OFA rats for the SCGE experiments were obtained from the breeding facility of the Medical University in Himberg (Austria) and allowed to acclimatize at the Institute of Cancer Research for one week before the experiments. The animals were housed in plastic cages (Macrolon type II) under standard conditions (24  1  C, humidity 50  5%, 12 h

3683

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 8 1 e3 6 9 1

light/dark cycle). The animals were kept on a standard diet (ssniff R/M-H, Ssniff, Soest, Germany).

2.2.

Experimental set-up and sample collection

The ozonation pilot plant was located at a municipal wastewater treatment plant (WWTP) with tertiary treatment (nitrification and denitrification). The plant comprised an ozone generator with a production capacity of 1000 g/h, a storage tank for liquid oxygen (feed gas) supply and a reactor unit. The reactor unit consisted of two reactors (5.0 m3 working volume, each) that were operated in series. Ozone was supplied to the first reactor with fine bubble plate diffusors, the second tank served as a reaction vessel. The effluent of the first reactor contained residual ozone in the range between 0.6 and 0.8 mg O3/L which was removed from the samples by stripping with air, while no ozone was detected in the second reactor. The ozone concentrations were measured with an amperometric ozone probe (Orbisphere Model 31 330.15, Hach Company, Colorado, US). The specific ozone consumption ranged between 0.6 and 0.8 g O3/g DOC during the sampling campaign. A schematic plan of the pilot plant, the treatment and the sampling points are depicted in Fig. 1, further details can be found in the article of Schaar et al. (2010). Sampling points were located at the effluent of the WWTP (corresponding to the influent of the ozonation pilot plant; In O3), the effluent of the first pilot plant reactor (Reactor 1) and the effluent of the second reactor (Reactor 2). The dissolved organic carbon (DOC) in the WWTP effluent ranged between 5.8 and 8.3 mg/L, the pH value was close to neutral (6.8e6.9), NH4-N was between 0.7 and 1.4 mg/L. The hydraulic flow during the campaigns ranged between 32.5 and 36.2 m3/h which corresponds to a total hydraulic retention time of 16.6e18.5 min. More details on the physicochemical properties of the waters and on the concentrations of selected chemicals and their degradation by O3 treatment can be found in Schaar et al. (2010). Immediately after the collection, the samples were transported to the laboratories and concentrated or used for the experiments with plant and liver cells.

2.3.

Salmonella/microsome assays

The indicator strains TA100, TA98, TA1537 and TA102 were obtained from B. Ames (University of Berkeley, CA). Strain YG1024 was provided by B. Majer (Japanese Tobacco Industry, Austria). Deep frozen cultures were prepared as described by

Fig. 1 e Schematic diagram of the pilot plant set-up and of the sampling points.

Maron and Ames (1984) with DMSO and were stored frozen at 80  C. New master plates were made every 4e6 weeks and stored refrigerated at 4  C in the dark. The characteristics of the strains (rfa, uvrB and pKM101) were tested as described by Ames et al. (1973) before the experiments and are listed in Table 1. The tests were carried out as plate incorporation assays (for details see Maron and Ames, 1984). Overnight cultures (ONC) were prepared in Erlenmeyer tubes with 20 mL Nutrient Broth. The medium was inoculated and incubated in the dark at 37  C for 10e12 h. In experiments with exogenous activation mix (i.e. liver homogenate), 0.5 mL of hepatic S9 mix (S9) and 0.1 mL of stationary phase ONC (ca. 1e2 108 cells) were added per plate. Activation mix (S9) which is used to detect the activity of indirectly acting genotoxins (which require enzymatic activation) was prepared according to the standard recipe of Maron and Ames (1984), stored on ice and used within 2 h. In experiment without S9 mix an aliquot amount of PBS was added. Per experimental point, three plates were made in parallel. The wastewater samples were tested with three different protocols. In the first, 100e500 mL of unconcentrated samples of freshly collected waters were added to selective agar plates. In a second series of experiments, the samples were concentrated by solid phase extraction with C18 cartridges according to the protocol of Petala et al. (2008) with and without modifications (see below). All wastewater samples were filtered through a 0.45 mm cellulose membrane filter and acidified to pH 2 with an appropriate volume of 6 N HCl for acidic extraction and subsequently passed through BAKERBOND spe Polar Plus C18 (Octadecyl) cartridges (J.T.Baker, Pittsburgh, US) to reach a 2000-fold concentration. For neutral extraction the wastewater samples (pH 6.8) were used directly

Table 1 e Characteristics of the strains used for the bacterial test. Strain Target LPSb Repairc Plasmidsd OATe Reversion event genea TA98 TA100

hisD hisG

rfa rfa

DuvrB DuvrB

pKM101 pKM101

þ þ

TA102

hisG

rfa

þ

þ

TA1537 YG1024

hisC hisD

rfa rfa

DuvrB DuvrB

pKM101, pAQ1 -nonepKM101, pYG219

þ þþþ

frameshift base pair substitution base pair substitution frameshift base pair substitution

a Type of mutation in the histidine operon. b LPS concerns the composition of the lipopolysacharide membrane, rfa indicates a mutation that causes partial loss of the membrane integrity. c The deletion in the gene encoding for uvrB (DuvrB) leads to increased sensitivity towards certain mutagens. d plasmid pKM101 enhances error prone DNA repair. Plasmid pAQ1 contains an ochre mutation at the hisG428 which increases the sensitivity of the strain towards radicals. pKM101 encodes for ampicilline resistance, pAQ1 for tetracycline resistance. Strain YG1024 is a derivate of TA98 but contains additionally plasmid pYG219 which encodes for O-acetyltransferase. e OAT e bacterial acetyltransferase; þ indicates the wild type gene, þþþ indicate over expression.

3684

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 8 1 e3 6 9 1

after the filtration. 5.0 mL acetone, 5.0 mL methanol and 5.0 mL distilled water were used for the activation, the extraction itself was carried out with 3.0 mL acetone:hexane (20:80) and 2.0 mL hexane. For acidic extraction (pH 2), the activation of each cartridge was conducted with 5.0 mL ethylacetate, 5.0 mL dichlormethane, 10.0 mL methanol and 10.0 mL distilled water and the extraction was carried out with 5.0 mL ethylacetate and 5.0 mL dichlormethane. Subsequently, the samples were either (a) evaporated to dryness under a stream of nitrogen at room temperature, then the material was dissolved in 1.0 mL of DMSO or (b) the same amount of DMSO was added before the concentration step for solvent exchange. The DMSO solutions were stored at 20  C and used in the plate incorporation assays. In a third series of experiments, the waters were concentrated by use of blue rayon a copperephatalocyanin complex which binds chemical carcinogens such as polycyclic aromatic hydrocarbons and heterocyclic aromatic amines. The concentration procedure was carried out as described in the protocol of Sakamoto and Hayatsu (1990) with slight modifications. Eight portions of blue rayon (0.5 g each) were filled in nets and exposed to the effluent before (In O3) and after ozone treatment (Reactor 2) for a period of 24 h. After the exposure, the material was washed and dried at room temperature. Subsequently, the samples were extracted twice with a 50:1 mix of methanol and ammonia at room temperature. The solutions were concentrated under vacuum with rotary evaporation to dryness. The material was dissolved in 1.0 mL DMSO and tested in plate incorporation assays in strains TA98 and YG1024. Strain YG1024 is a derivate of TA98 with increased N-hydroxyarylamine-O-acetyltransferase activity which is highly sensitive towards aromatic amines and nitroarenes (Watanabe et al., 1993), these compounds have been detected in surface waters in earlier studies (Ohe et al., 2004). In all experiments, negative and positive controls were included. DMSO was used as a solvent control. In experiments without S9; 2,4,7-TNF was used for TA98, NaN3 for TA100, MMS for TA102 and ICR 191 for TA1537 as a positive control. 2-AA was used as a positive control for all strains in experiments with S9.

2.4.

Micronucleus (MN) assay with A. cepa L

The experiments were carried out according to the standard protocol published by Ma et al. (1995). Young onion bulbs (diameter 3 cm) from a local market were placed in a glass beaker with 50 mL tap water for 48 h in the dark. Subsequently, the roots (length 1e2 cm) were exposed to unconcentrated wastewater samples for 24 h, then they were transferred to fresh tap water for another 24 h. At the end of the recovery period, the roots were fixed in ethanol and acetic acid (3:1) for 24 h and stored in 70% ethanol. Tap water was used as a negative control, aqueous arsenic trioxide solution (5.0 mM As2O3) was used as positive control. The root tips were hydrolyzed in a mix (1:1) of hydrochloric acid (HCl 5.0 N) and ethanol (99%) for 3 min and washed in tap water before staining with acetocarmine (2% in 45% acetic acid solution). MN were scored according to the criteria described by Ma et al. (1995). For each experimental point, the MN frequencies were determined in five plants. From each bulb, two slides were made and 500 cells were evaluated per slide.

Furthermore, also the mitotic indices (MIs) were determined in 3000 cells per experimental point. The microscopic evaluation was carried out under a light microscope (Nikon YS200, Japan) with 400-fold magnification.

2.5.

Single cell gel electrophoresis (SCGE) assays

SCGE assays were conducted as described in detail by Ferk et al. (2007). The hepatocytes were isolated from the liver of male Him-OFA rats (250 g) with the two step collagenase technique developed by Selgen (1976) with some modifications (Parzefall et al., 1989). The cells were cultivated at 37  C for 60 min in Eagle’s minimal essential medium (MEM, Sigma, Germany) which was either prepared with pure distilled water (control) or with medium which had been prepared with the wastewater samples (100%). The control medium and the media which had been prepared from the samples were mixed in a such way that the final concentrations of the wastewater samples were 11, 33 and 100% respectively. After the incubation, the cells were washed twice with PBS (pH 7.4) and placed on agarose coated slides. After lysis and electrophoresis under standard conditions (20 min, 300 mA, 25 V, at 4  C, pH > 13), the gels were stained with ethidium bromide (20 mg/mL). Acute toxic effects were monitored with the trypan blue dye-exclusion technique which is based on the uptake of the dye due to loss of membrane integrity (Lindl and Bauer, 1994). From each sample, 50 cells were evaluated. Only cultures in which the vitality was 80% were analyzed for DNA migration. The formation of comets (% DNA in tail) was determined with a computer aided comet assay image analysis system (Comet Assay IV, Perceptive Instruments, UK). The experiments were conducted according to the guidelines of Tice, et al. (2000). For each experimental point, three cultures were made in parallel and from each 50 cells were evaluated.

2.6.

Statistical analyses

To evaluate the results of the Salmonella/microsome assay (UKEMs) the “two fold rule” was applied (UKEMS, 1990). In addition, analysis of variance ANOVA with Tukey post tests were performed. For the statistical evaluation of Allium MN test results the ANOVA F-test, followed by Dunnett’s multiple comparisons test was used. In SCGE experiments, statistical significance was tested by ANOVA followed by Dunnett’s multiple comparisons test. In all experiments p-values 0.05 were considered as significant.

3.

Results

With all unconcentrated samples which were tested at three doses in strains TA98, TA100 and TA102 (100, 200 and 500 mL/ plate) in presence and absence of metabolic activation mix consistently negative results were obtained under all experimental conditions (data not shown). On the contrary, a clear positive result was obtained after concentration of the samples with C18 cartridges and acidic extraction (pH 2) in strains TA98 and TA1537. It can be seen in Table 2 that the effects were more pronounced in experiments without

3685

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 8 1 e3 6 9 1

Table 2 e Results of Salmonella/microsome assays after concentration with C18 cartriges and acid elution in absence (LS9) and presence (DS9) of metabolic activation mix. Site

Dose (mL/plate)e

S9 TA98

TA100

þS9 TA102

TA1537

TA98

(mean  S.D.) Negative control In O3 (before ozonisation) Reactor 1

Reactor 2

Positive controld

c

e 100.0 200.0 300.0 100.0 200.0 300.0 100.0 200.0 300.0 e

22.3  2.1 43.3  3.1a 56.0  3.0*,a toxicf 42.3  2.5a 46.0  9.5*,a 51.7  4.0*,a 29.3  1.5 44.7  5.0*,a 54.0  8.5*,a 1483.7  172.6*,a

127.0  20.7 95.7  20.6a 50.3  8.5 toxicf 130.7  6.4 139.3  7.6 104.3  17.0 128.7  4.2 126.0  16.5 137.0  5.6 1367.3  166.2*,a

TA100

TA102

TA1537

(mean  S.D.)

291.7  12.7 272.7  37.9a 128.7  17.8 toxicf 243.7  13.1a 257.7  6.7 245.0  26.3 275.3  19.0 339.7  64.1 291.7  21.5 1496.0  120.1*,a

14.0  2.6 23.7  3.1a 81.7  10.4*,a toxicf 17.7  3.5 26.0  4.4a,b 34.7  2.1*,a 27.3  1.2a 36.0  6.2*,a,b 53.0  6.1*,a 667.3  87.6*,a

36.3  7.1 46.3  6.7 41.0  2.6 toxicf 34.7  5.9 47.0  2.0 60.7  3.2a 34.3  6.7 39.7  4.2 70.0  10.0a 1565.3  196.5*,a

151.0  0.0 161.3  6.4 143.0  41.1 toxicf 134.0  19.3 146.7  26.6 100.3  3.8a 160.7  8.3 155.7  18.0 142.0  17.1 2000.0  132.3*,a

348.0  17.1 322.0  43.3 301.3  26.1 toxicf 340.7  63.1 356.3  20.8 272.7  48.3 340.7  63.5 340.7  63.5 319.7  43.5 1595.0  142.6*,a

10.3  3.5 13.0  1.0 12.3  3.8 16.0  1.7 15.0  1.0 14.0  1.0 15.0  2.6 10.3  3.5 11.7  0.6 15.3  3.2 293.7  55.1*,a

*indicate an effect according to the “two fold rule”. a statistically significant differences ( p  0.05) as compared to the corresponding control (ANOVA). b statistically significant differences ( p  0.05, ANOVA) as compared to the effects seen with corresponding volume of untreated water (analyses were performed only under conditions which did not cause toxic effects). c DMSO (100 mL/plate) was used as a solvent and as a negative control. d Positive controls without metabolic activation mix (S9) were 2,4,7-TNF (0.1 mg/plate) for TA98, NaN3 (2.0 mg/plate) for TA100, MMS (2.0 mL/ plate) for TA102 and ICR 191 (2.0 mg/plate) for TA1537. 2-AA (2.0 mg/plate) was used as a positive control for all strains in presence of metabolic activation mix (þS9). e DMSO extracts correspond to 100.0, 200.0 and 300.0 ml of unconcentrated water per plate. f Toxic effects were diagnosed as dissolved background lawns (bacteria form under non toxic conditions on the selective media plates a dense lawn due to residual divisions, bacteriocidal/bacteriostatic effects lead to formation of “petit point” colonies” which do not represent true revertants, for details see Maron and Ames, 1984).

addition of liver homogenate (S9 mix) and that the number of hisþ revertants was significantly reduced after ozonation. The results shown in Table 2 were obtained after evaporation of the extracts to dryness. Since this may cause loss of volatile compounds (Gustavsson et al., 2004; Stalter et al., 2011), we analyzed also concentrates of the water samples which had been prepared by adding an identical amount of DMSO before complete evaporation. After acidic extraction of untreated wastewater (in O3) the numbers of hisþ revertants obtained with addition of 100 mL of the concentrates/plate were 82.3  10.7 in TA 98 while in TA100 toxic effects were observed (for details see footnote of Table 2). After addition of liver homogenate, the numbers of the hisþ revertants were 55.0  5.0 with TA98 and 134.3  14.7 with TA100, respectively. The numbers of hisþ revertants on control plates were similar as those seen in the experiment shown in Table 1 and ozonation resulted again in a decrease of the revertant frequencies in strain TA98 (data not shown). These comparisons indicate that loss of volatile compounds during the concentration had no strong impact on the results of the experiments. The effects of acidic and neutral extracts were compared in an additional experiment (Fig. 2). Also in these experiments, the extract were not evaporated to dryness. It can be seen that the results confirm the observations of the first experiment. Again, no clear increase of the revertant frequencies was observed with TA100 (Fig. 2b) while in TA98 (Fig. 2a) a significant (approximately 4-fold) increase over the background rate was seen in absence of metabolic activation. This effect declined strongly after ozonation, but the revertant number

was still more than two-fold over the background. When the same extracts were tested in presence of metabolic activation mix (þS9), a less pronounced effect was observed. However, the number of hisþ revertants was still 2-fold over the control value in TA98. After neutral elution of the cartridges no mutagenic effects were detected. The results of the experiment obtained with blue rayon are summarised in Table 3. It can be seen, that no positive results were detected in presence and absence of activation mix in TA98. Also in strain YG1024 no indication of induction of hisþ revertants was found. The results of a representative MN experiment with root tip cells of A. cepa are depicted in Fig. 3. The MN frequencies observed with the effluent before ozonation were higher than the frequencies found in the controls (1.2  1.1 vs. 0.6  0.5&) but the difference was not significant. None of the ozonated samples caused induction of MN frequencies. The MI were not altered after exposure of the plants to the wastewater samples while a clear positive result was obtained with the positive control (As2O3). The lack of a significant effect was confirmed in a second independent experiment (data not shown). The results obtained in a representative SCGE experiment with primary rat hepatocytes are depicted in Fig. 4a and b. Fig. 4a shows that the viability of the cells was not altered after ozone treatment. Incubation of the cells with effluent before ozonation (In O3) caused a clear dose dependent increase of DNA migration while ozonation resulted in a pronounced decline of comet formation (Fig. 4b). With undiluted effluent of Reactor 1, DNA-damage was reduced by 26% as

3686

a

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 8 1 e3 6 9 1

*,a

100

TA98 -S9

+

his revertants

80

Neutral elution (pH 6.8) Acidified elution (pH 2.0)

*,a,b

60

+S9

*,a

*,a,b

*,a

40 20

2

1

R ea ct or

R

ea ct or

O In

3

1

ea ct or

C

TA100 200

Neutral Acidified (pH 2) +S9

-S9 150

100

+

his revertants

-

R

3

In

b

R ea ct or

C

O

-

2

0

2 R

ea ct or

1 R ea ct or

3

C

O

-

In

2 ea ct or R

ea ct or R

In

3

C

O

-

1

0

Toxic

Toxic

50

Fig. 2 e a and b. Induction of hisD revertants by acidic (pH 2.0) and neutral (pH 6.8) extracts in strains TA98 (a) and TA100 (b) in presence and absence of liver homogenate. DMSO extracts which correspond to 200 mL of unconcentrated water were added per plate. Per experimental point three plates were made in parallel. In experiments without S9, NaN3 (1.5 mg/plate for TA100) and 2,4,7-TNF (0.1 mg/plate for TA98) were used as a positive controls and induced 1064 ± 201 and 1733 ± 67 hisD revertants/plate. In experiments with S9, 2-AA (2.0 mg/ plate) was used as a positive control. The compound induced in strain TA98 1523 ± 45 revertants/plate and in strain TA100 1699 ± 135 revertants per plate. *indicate an effect according to the “two fold rule”; astatistically significant differences ( p £ 0.05) as compared to the corresponding control (ANOVA); bstatistically significant different ( p £ 0.05, ANOVA) as compared to the effects seen with untreated water.

compared to the effect seen with the effluent of the plant. With water from Reactor 2, the reduction was even more pronounced (39%) at the highest dose. The difference between the results obtained with samples from the two reactors was not significant but the genotoxic activity was in both samples significantly lower as that found before ozonation.

4.

Discussion

Taken together, the results of the present study indicate that ozonation of tertiary treated municipal wastewater causes a strong reduction of its genotoxic properties. This effect was seen in two different test systems, namely in Salmonella/ microsome assays and in SCGE experiments with primary rat

liver cells. The MN experiments with Allium yielded negative results with all samples. These latter findings do not allow to draw firm conclusions concerning the impact of O3 treatment on the reduction of the genotoxic properties of the effluent but they indicate that O3 does not induce genotoxic effects. As described above, we did not find increased acute toxicity of the tested waters in any of the three models, i.e. O3 treatment did not lead to a decrease of the colony forming ability of the coliform bacteria and of the survival of plant and mammalian cells. In this context, it is notable that in some earlier experiments reduction of the viability was observed after ozonation with different indicator organisms, for example with worms (Lumbricus variegatus), snails (Potmopyrgus antipodarum), different crustacea species and rotifers (Petala et al., 2006a; Stalter et al., 2010a). Also in experiments with fish (early life stage toxicity tests and hatching of larvae) adverse effects were seen (Stalter et al., 2010b). The main objective of the present study was to investigate the impact of O3 on the genotoxic properties of treated wastewater. This issue has been studied in a number of investigations but only few of them were conducted with treatment plant effluents. Out of approximately 80 studies, we found only one with mammalian cells (Itoh and Matsuoka, 1996) and an SCGE study with haemolymph cells of mussels (Stalter et al., 2010a). Furthermore, only in one investigation plant bioassays (Monarca et al., 2000) were included, all other trials were conducted with bacteria. One of the novel aspects of the present investigations is, that we used primary liver cells which possess phase I and phase II enzymes involved in the activation and detoxification of genotoxins (Thruman and Kauffman, 1980). Therefore, it is not necessary to add exogenous metabolic activation mix (which contains enzymes which activates genotoxic carcinogens) as in experiments with bacteria and with stable cell lines which do not reflect the metabolism of xenobiotics in mammals. Investigations concerning the impact of ozonation on the mutagenic activities of surface water which were conducted in the 1970s and 1980s have been reviewed by Noot et al. (1989). Some results obtained in these earlier and also in newer studies with secondary wastewater are in agreement with our observation of decreased mutagenic activity in the frameshift tester strain TA98 without metabolic activation mix (S9) after O3 treatment. For example, Petala et al. (2008) reported on reduction of hisþ revertant induction in treated wastewater under identical experimental conditions after prolonged ozone treatment. Also in experiments with drinking water, a decrease of the mutagenic potency was found in the acidic fraction in TA98 without S9 after O3 treatment (Kool and Vankreijl, 1984). In a study with Rhine river water, a similar effect was seen with a XAD-based concentration procedure (Zoeteman et al., 1982). Furthermore, decreased mutagenicity was found in two studies with ozonated secondary effluents of a sewage treatement plant in umu tests with Salmonella typhimurium (Cao et al., 2009; Reungoat et al., 2010) and Takanashi et al. (2002) reported on reduction of bacterial mutagenicity in wastewater that was treated with ozone before sewage treatment. In some studies increased genotoxic effects were found after ozonation (for review see Noot et al., 1989). For example Monarca et al. (2000) reported on induction of formation of hisþ revertants in TA98 without activation. The only SCGE

3687

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 8 1 e3 6 9 1

Table 3 e Results of Salmonella/microsome assays after concentration with blue rayon in absence (LS9) and presence (DS9) of metabolic activation mix. Dose (gE/plate)c

Site

S9 TA 98

þS9 YG1024

(mean  S.D.) b

e 0.025 0.050 0.100 0.025 0.050 0.100 e

Negative control In O3 (before ozonisation)

Reactor 2

Positive controld

18.3  1.5 17.7  1.5 17.3  3.8 22.3  4.2 20.0  4.6 20.0  3.5 26.3  4.0a 1477  260.9*,a

TA 98

YG1024 (mean  S.D.)

13.3  3.2 13.7  3.8 15.0  5.3 19.7  1.2 16.3  1.5 15.7  4.9 19.0  4.0 1662  301.2*,a

30.3  4.0 31.3  2.5 42. 3  3.2a 38.3  4.9 25.0  7.9 27.7  4.0 31.0  5.0 1413.3  168.0*,a

16.7  1.2 17.3  3.5 22.3  4.5 18.7  3.1 20.0  4.6 22.3  2.1a 18.3  1.2 1734.0  372.1*,a

*indicate an effect according to the “two fold rule”. a statistically significant differences as compared to the corresponding control ( p  0.05, ANOVA). b DMSO (100 mL/plate) was used as a negative control. c Gram equivalent of blue rayon. d 2,4,7-TNF (0.1mg/plate) was used as positive control in absence of metabolic activation mix (-S9). 2-AA (2.0 mg/plate) was used as a positive control in experiments with metabolic acitivation mix (þS9).

assay with ozonated treated wastewater has been published by Stalter et al. (2010a), who used haemolymph cells of a mussel (Dreissena polymorpha) and found increased DNA migration after the treatment. Table 3 shows that we failed to detect mutagenic effects in experiments with blue rayon. This technique has been used successfully to detect mutagens in surface waters (Kataoka et al., 2000; Kira et al., 1997; Kummrow et al., 2006; Ohe et al., 2003). The material binds genotoxic carcinogens with planar structures such as nitroarenes, polycyclic aromatic hydrocarbons and heterocyclic aromatic amines (HAAs, Sakamoto and Hayatsu, 1990). The negative result of the present experiment can be taken as an indication that other contaminants account for the effects. On the basis of the positive results we found in the experiments with the cartridges, it can be concluded that the mutagenic activity is due to directly acting mutagens which induce frameshift mutations. The negative results obtained in TA102 indicate that reactive oxygen species to which this strain is highly sensitive (Maron and Ames, 1984), are not involved in the mutagenic effects of the influent of the pilot plant and that ozonation does not lead to formation of persisting DNA-

b

15

MI (%)

10

* 5

MN/1000 cells

a

damaging oxygen radicals. In this context it is notable that it is well documented that the gas decomposes rapidly in aqueous solutions and forms a number of DNA reactive species (Victorin, 1992). Decreased genotoxic activity after O3 treatment was not only found in bacterial tests but also in the SCGE assays with rat liver cells. This observation shows that the reduction of the mutagenic properties of water is also detectable in mammalian cells and this finding is probably more relevant for vertebrates and possibly also for humans as the results obtained in the other experimental models. A number of investigations which concerned the formation of by-products in waters during O3 treatment has been published (for reviews see Van Hoof, 1983; von Gunten, 2003a, b) and the results of Salmonella/microsome assays show that the majority of these compounds is devoid of mutagenic activity (Cotruvo et al., 1978). It was found, that the genotoxic properties of representatives of different classes of DNA reactive carcinogens such as polycyclic aromatic hydrocarbons, aromatic amines, alkylating agents, aflatoxin and benzidine are strongly reduced by treatment with ozone (Caulfield et al., 1979; Miltner et al., 2008).

14 12

*

10

*

8 6 4 2

0

Control

In O3

Reactor 1 Reactor 2

As2O3

0

Control

In O3

Reactor 1 Reactor 2

As2O3

Fig. 3 e a and b. Impact of water ozonation on the MI of root tip cells of Allium cepa (3a) and on the MN frequencies (3b). As2O3 was used as a positive control (5.0 mM, treatment 2 h, recovery 36 h). Bars represent means ± S.D. Stars indicate statistically significant differences (control vs. test water, p £ 0.05, Dunnett test).

3688

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 8 1 e3 6 9 1

a 100 Viability (%)

80 60 40 20 0

0.0

11

Control

b

33

100

11

33

100

11

Reactor 1

In O3

33

100

Reactor 2

Concentration of the test water in %

*

5

% DNA in Tail

4

*

3

*

*

*

2 1 0

0.0

Control

11

33

100

11

In O3

33

Reactor 1

100

11

33

Reactor 2

100

Concentration of the test water in %

Fig. 4 e a and b. Impact of untreated and ozonated waters on the viability (4a) and DNA-stability (4b) of primary rat liver cells. The comet experiments were carried out under standard conditions. Bars represent means ± S.D. Stars indicate statistically significant differences (control vs. test water, p £ 0.05, Dunnett test).

As mentioned above, the results of earlier studies concerning the impact of ozone treatment on the toxic activities of waters yielded highly controversial results. In some of them adverse effects were seen while others (including ours) indicated reduction of the toxic properties. One reason for these discrepancies may be differences in the sensitivity of the indicator organisms and of the experimental design of the studies. Acute toxic effects were found in some investigations with aquatic invertebrates and also in fish (Petala et al., 2006a; Stalter et al., 2010a, 2010b; 2011) while negative findings were consistently obtained in experiments with higher plants (Crebelli et al., 2005; Monarca et al., 2000; Stalter et al., 2010a) which are apparently less sensitive than animals. Another relevant factor may be the experimental design; intact aquatic organisms can be exposed over long time periods in the native waters, while with isolated cells the treatment periods are in the range of a few hours. Also the specific ozone dose may affect the outcome of such studies. Recent publications (Hollender et al., 2009; Schaar et al., 2010; Stalter et al., 2010a, 2010b) refer to specific ozone doses or consumptions in g O3/g DOC, which takes the organic compounds into account. O3 doses in mg/L without further information on the DOC in treated wastewater do not provide a solid basis for comparisons of the experimental outcomes. Furthermore, the concentrations varied over a broad range in different studies and some findings indicate that toxic effects are dose dependent. For example, Petala et al. (2008) found in toxicity tests with Vibrio fischeri a stronger effect with 2.5 mg/L as with 5.0 mg/L (the latter dose is similar to the one we used) and in a further study even a protective effect with 7.2 mg/L (Petala et al., 2006b). Also in TA98 the mutagenic responses were more pronounced with lower ozone levels (Petala et al., 2008).

The same group tested also the impact of treatment time and observed a decrease of the mutagenic activity with prolonged ozone treatment (>5 min) in strain TA98 (S9). Another relevant criterion is the composition of the treated wastewaters. As shown by Hollender et al. (2009), the amounts of O3 required to eliminate different micropollutants in treated municipal wastewaters depends on their chemical structures and differs strongly. Furthermore, it is known that mutagenic carcinogens such as bromate and dimethylnitrosamine can be formed by ozonation but their yields depend largely on the presence of precursors (for details see Huang et al., 2009; Hollender et al., 2009). Another important factor which may affect the results is the concentration procedure. As stressed by Helma et al. (1998) all currently available methods may lead to loss of potentially active substances. In the case of the solid phase extraction method we used, evaporation to dryness may lead to loss of volatiles and non specific acting toxicans (Gustavsson et al., 2004; Stalter et al., 2011). Furthermore, it was shown that ozonation of certain compounds in water results in formation of oxidation products that are too polar for extraction (Benner and Ternes, 2009; Stalter et al., 2011). Chang et al. (2009) reported in a study concerning removal of endocrine disruptors by ozonation, that different matrices in the columns may have a strong impact on the results. As mentioned above, it is known that the pH of the water samples has a strong impact on the recovery of different groups of chemicals and it was shown by Stalter et al. (2011) that the acute toxic effects of ozonated water (measured in a rat pituary cell line) are lower with samples extracted at pH 7 as compared to those found with pH 2. In the present study, the concentration procedure may have had an impact on the results of the bacterial mutagenicity tests but not on the experiments with Allium and hepatocytes which

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 8 1 e3 6 9 1

were conducted with unconcentrated samples. However, although the experiments were conducted few (6e8 h) after the sampling, it can be not excluded that the storage had an impact on the results. In this context it is notable that Petala et al. (2006a) showed that the acute toxic effects of ozonated samples decline as a function of time. The most relevant bacterial mutagens found in wastewater are polycyclic aromatic hydrocarbons (PAHs), aromatic as well as heterocyclic aromatic amines and nitro compounds including PBTA type mutagens (derived from 4-amino50 ,40 -dinitrophenyl, Houk, 1992; Ohe et al., 2004). However, it is unlikely that these compounds account for the present findings (in which positive results were only seen in TA98 without S9) as their mutational characteristics are different. Aromatic amines as well as PAHs require activation with S9 (Aeschbacher and Turesky, 1991; IACR, 1991, 2010) while nitro compounds are genotoxic without activation. However, representatives of the latter group are not only active in TA98 but also potent mutagens in TA100 (McMahon et al., 1979) and many of them cause more pronounced effects in YG1024 as in other tester strains (Rosenkranz and Mermelstein, 1983; Watanabe et al., 1993, 1994). In several studies with drinking and surface waters mutagenic activity was seen after acidic but not after neutral extraction (Kronberg et al., 1985; Meier, 1988; Monarca et al., 1985; Vartiainen and Liimatainen, 1986; Wigilius et al., 1985). Meier (1988) stressed that these effects may be due to chlorination products such as chlorohydroxyfuranes that are formed as a consequence of reactions with humic acids. However, the fact that these compounds are potent base substitution mutagens which cause effects in strain TA100 distracts from the assumption that they are responsible for the effects we saw in the present study. Many hydrophilic compounds which may be found in wastewaters such as herbicides, endocrine disruptors as well a residues of pharmaceuticals are extracted under acidic conditions (for details see EPA, 1994; Noordsij et al., 1983; Petala et al., 2008) but it is not possible to draw conclusions which of them (or their reaction products) may account for the effects we saw in the present experiments.

5.

Conclusions

The result of this study show that ozonation reduces the genotoxic effects of tertiary treated municipal wastewater which were seen after concentration in bacterial mutagenicity tests and also in SCGE (comet) assays with mammalian cells (i.e. with primary rat liver cells) with unconcentrated samples. In experiments with plants no gentoxic activity was detected with the treated wastewater and ozonation had no impact on the micronucleus frequencies. The decrease of DNA-damage as well as the lack of acute toxic effects after ozonation which was found in all three types of indicator cells can be taken as an indication that the treatment does not cause adverse effects and leads to destruction of DNA reactive micropollutants in the effluent of the plant. Since in some earlier investigations toxic effects were seen after ozonation of water, it remains to be clarified if and to which extent our findings can be generalized and further research will be required to find out which molecular mechanisms account for the effects which we observed.

3689

Acknowledgements The results presented in this work were obtained within the KOMOZON project (GZ A601819), funded by the Austrian Federal Ministry of Agriculture, Forestry, Environment and Water Management, Kommunalkredit Austria AG.

references

Aeschbacher, H.U., Turesky, R.J., 1991. Mammalian cell mutagenicity and metabolism of heterocyclic aromatic amines. Mutat. Res. 259, 235e250. Ames, B.N., 1989. Endogenous DNA damage as related to cancer and aging. Mutat. Res. 214, 41e46. Ames, B.N., Lee, F.D., Durston, W.E., 1973. An improved bacterial assay system for the detecton and classification of mutagens and carcinogens. Proc. Natl. Acad. Sci. 70, 782e786. Benner, J., Ternes, T.A., 2009. Ozonation of propranolol: formation of oxidation products. Environ. Sci. Technol. 43, 5086e5093. Bickham, J.W., Sandhu, S., Hebert, P.D., Chikhi, L., Athwal, R., 2000. Effects of chemical contaminants on genetic diversity in natural populations: implications for biomonitoring and ecotoxicology. Mutat. Res. 463, 33e51. Bozkaya-Schrotter, B., Daines, C., Brunel, A., Schrotter, J.C., Breant, P., 2009. Treatment of nanofiltration membrane concentrates: organic micropollutant and NOM removal. Desalination Water Treat. 9, 36e42. Cao, N., Yang, M., Zhang, Y., Hu, J., Ike, M., Hirotsuji, J., Matsui, H., Inoue, D., Sei, K., 2009. Evaluation of wastewater reclamation technologies based on in vitro and in vivo bioassays. Sci. Total Environ. 407, 1588e1597. Caulfield, M.J., Burleson, G.R., Pollard, M., 1979. Ozonation of mutagenic and carcinogenic alkylating agents, pesticides, aflatoxin B1, and benzidine in water. Cancer Res. 39, 2155e2159. Chang, H.S., Choo, K.H., Lee, B., Choi, S.J., 2009. The methods of identification, analysis, and removal of endocrine disrupting compounds (EDCs) in water. J. Hazard. Mater. 172, 1e12. Claxton, L.D., Woodall Jr., G.M., 2007. A review of the mutagenicity and rodent carcinogenicity of ambient air. Mutat. Res. 636, 36e94. Collivignarelli, C., Bertanza, G., Pedrazzani, R., 2000. A comparison among different wastewater disinfection systems: experimental results. Environ. Technol. 21, 1e16. Cotruvo, J.A., Simmon, V.F., Spanggord, R.J., 1978. Investigation of mutagenic effects of products of ozonation reactions in water. Ann. N. Y Acad. Sci. 298, 124e140. Crebelli, R., Conti, L., Monarca, S., Feretti, D., Zerbini, I., Zani, C., Veschetti, E., Cutilli, D., Ottaviani, M., 2005. Genotoxicity of the disinfection by-products resulting from peracetic acid- or hypochlorite-disinfected sewage wastewater. Water Res. 39, 1105e1113. EPA, 1994. Method 525.2. In: National Exposure Research Laboratory Office of Research and Development (Ed.), Determination of Organic Compounds in Drinking Water by Liquid-Solid Extraction and Capillary Column Gas Chromatography/Mass Spectrometry. US Environmental Protection Agency, Cincinnati, OH. Ferk, F., Misik, M., Hoelzl, C., Uhl, M., Fuerhacker, M., Grillitsch, B., Parzefall, W., Nersesyan, A., Micieta, K., Grummt, T., Ehrlich, V., Knasmueller, S., 2007. Benzalkonium chloride (BAC) and dimethyldioctadecyl-ammonium bromide (DDAB), two common quaternary ammonium compounds, cause genotoxic effects in mammalian and plant cells at

3690

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 8 1 e3 6 9 1

environmentally relevant concentrations. Mutagenesis 22, 363e370. Gustavsson, L.K., Klee, N., Olsman, H., Hollert, H., Engwall, M., 2004. Fate of Ah receptor agonists during biological treatment of an industrial sludge containing explosives and pharmaceutical residues. Environ. Sci. Pollut. Res. Int. 11, 379e387. Helma, C., Eckl, P., Gottmann, E., Kassie, F., Rodinger, W., Steinkellner, H., Windpassinger, C., Schulte-Hermann, R., Knasmueller, S., 1998. Genotoxic and ecotoxic effects of groundwaters and their relation to routinely measured chemical parameters. Environ. Sci. Technol. 32, 1799e1805. Helma, C., Kronberg, L., Ma, T.H., Knasmueller, S., 1995. Genotoxic effects of the chlorinated hydroxyfuranones 3-chloro-4(dichloromethyl)-5-hydroxy-2[5H]-furanone and 3,4-dichloro5-hydroxy-2[5H]-furanone in Tradescantia micronucleus assays. Mutat. Res. 346, 181e186. Hollender, J., Zimmermann, S.G., Koepke, S., Krauss, M., McArdell, C.S., Ort, C., Singer, H., von Gunten, U., Siegrist, H., 2009. Elimination of organic micropollutants in a municipal wastewater treatment plant upgraded with a full-scale postozonation followed by sand filtration. Environ. Sci. Technol. 43, 7862e7869. Houk, V.S., 1992. The genotoxicity of industrial wastes and effluents. Mutat. Res. 277, 91e138. Hu, J.Y., Wang, Z.S., Ng, W.J., Ong, S.L., 1999. Disinfection byproducts in water produced by ozonation and chlorination. Environ. Monit. Assess. 59, 81e93. Huang, W.J., Chang, C.Y., Shih, F.H., 2009. Disinfection by-product formation and mutagenic assay caused by preozonation of groundwater containing bromide. Environ. Monit. Assess. 158, 181e196. IACR, 1991. Some Naturally Occurring Substances: Food Items and Constituents, Heterocyclic Aromatic Amines and Mycotoxins, vol. 56 IACR, Lyon. IACR, 2010. Some Non-heterocyclic Polycyclic Aromatic Hydrocarbons and Some Related Exposures, vol. 92 IACR, Lyon. Itoh, S., Matsuoka, Y., 1996. Contributions of disinfection byproducts to activity inducing chromosomal aberrations of drinking water. Water Res. 30, 1403e1410. Joss, A., Siegrist, H., Ternes, T.A., 2008. Are we about to upgrade wastewater treatment for removing organic micropollutants? Water Sci. Technol. 57, 251e255. Kataoka, H., Hayatsu, T., Hietsch, G., Steinkellner, H., Nishioka, S., Narimatsu, S., Knasmueller, S., Hayatsu, H., 2000. Identification of mutagenic heterocyclic amines (IQ, Trp-P-1 and AalphaC) in the water of the Danube river. Mutat. Res. 466, 27e35. Kira, S., Horiguchi, H., Nogami, Y., Komatsu, T., Fujisawa, K., Ito, T., Hayatsu, H., 1997. Improved blue rayon hanging technique that can measure a time-weighted average concentration of benzo(a)pyrene in sea water. Bull. Environ. Contam. Toxicol. 59, 941e947. Knasmueller, S., Helma, C., Eckl, P., Gottmann, E., Steinkellner, H., Kassie, F., Haider, T., Parzefall, W., Schulte-Hermann, R., 1998. Investigations on genotoxic effects of groundwater from the Mitterndorfer Senke and from the vicinity of Wiener Neustadt. Middle Europ J. Med. 110, 824e833. Knasmueller, S., Mersch-Sundermann, V., Kevekordes, S., Darroudi, F., Huber, W.W., Hoelzl, C., Bichler, J., Majer, B.J., 2004. Use of human-derived liver cell lines for the detection of environmental and dietary genotoxicants; current state of knowledge. Toxicology 198, 315e328. Knasmuller, S., Steinkellner, H., Majer, B.J., Nobis, E.C., Scharf, G., Kassie, F., 2002. Search for dietary antimutagens and anticarcinogens: Methodological aspects and extrapolation problems. Food Chem. Toxicol. 40, 1051e1062.

Kool, H.J., Vankreijl, C.F., 1984. Formation and removal of mutagenic activity during drinking water preparation. Water Res. 18, 1011e1016. Kronberg, L., Holmbom, B., Tikkanen, L., 1985. Fractionation of mutagenic compounds formed during chlorination of humic water. Sci. Total Environ. 47, 343e347. Kummrow, F., Rech, C.M., Coimbrao, C.A., Umbuzeiro, G.A., 2006. Blue rayon-anchored technique/Salmonella microsome microsuspension assay as a tool to monitor for genotoxic polycyclic compounds in Santos estuary. Mutat. Res. 609, 60e67. Lindl, T., Bauer, J., 1994. Zell- und Gewebekultur, third ed.. Gustave Fischer, Stuttgart, Jena, New York. 94e96. Ma, T.H., Xu, Z., Xu, C., McConnell, H., Rabago, E.V., Arreola, G.A., Zhang, H., 1995. The improved Allium/Vicia root tip micronucleus assay for clastogenicity of environmental pollutants. Mutat. Res. 334, 185e195. Majer, B.J., Grummt, T., Uhl, M., Knasmueller, S., 2005. Use of plant bioassays for the detection of genotoxins in the aquatic environment. Acta Hydrochim Hydrobiol 33, 45e55. Maron, D.M., Ames, B.N., 1984. Revised methods for the Salmonella mutagenicity test. In: Kilbey, B.J., Legator, M., Nocols, W., Ramel, C. (Eds.), Handbook of Mutagenicity Test Procedures. Elsevier Science Publishers, BV, New York, pp. 93e140. McMahon, R.E., Cline, J.C., Thompson, C.Z., 1979. Assay of 855 test chemicals in ten tester strains using a new modification of the Ames test for bacterial mutagens. Cancer Res. 39, 682e693. Medina, M.H., Correa, J.A., Barata, C., 2007. Micro-evolution due to pollution: possible consequences for ecosystem responses to toxic stress. Chemosphere 67, 2105e2114. Meier, J.R., 1988. Genotoxic activity of organic chemicals in drinking water. Mutat. Res. 196, 211e245. Miltner, R.J., Speth, T.F., Richardson, S.D., Krasner, S.W., Weinberg, H.S., Simmons, J.E., 2008. Integrated disinfection by-products mixtures research: disinfection of drinking waters by chlorination and ozonation/postchlorination treatment scenarios. J. Toxicol. Environ. Health A 71, 1133e1148. Monarca, S., Feretti, D., Collivignarelli, C., Guzzella, L., Zerbini, I., Bertanza, G., Pedrazzani, R., 2000. The influence of different disinfectants on mutagenicity and toxicity of urban wastewater. Water Res. 34, 4261e4269. Monarca, S., Pasquini, R., Arcaleni, P., 1985. Detction of mutagens in unconcentrated and concentrated drinking water supplies before and after treatment using a microscale fluctuation test. Chemosphere 14, 1069e1080. Moore, M.N., Depledge, M.H., Readman, J.W., Leonard, D.R.P., 2004. An integrated biomarker-based strategy for ecotoxicological evaluation of risk in environmental management. Mutat. Res. 552, 247e268. Nakamuro, K., Ueno, H., Sayato, Y., 1989. Mutagenic activity of organic concentrates from municipal river water and sewage effluent after chlorination or ozonation. Water Sci. Technol. 21, 1895e1898. Nersesyan, A., Misik, M., Knasmueller, S., 2009. Methods used for detection of antimutagens: an overview. In: Knasmueller, S., DeMarini, D.M., Johnson, I., Gerhauser, C. (Eds.), Chemoprevention of Cancer and DNA Damage by Dietary Factors. Wiley-VCH, Weheinim, pp. 211e228. Noordsij, A., Vanbeveren, J., Brandt, A., 1983. Isolation of organic compounds from water for chemical analyses and toxicological testing. Intern. J. Environ. Anal. Chem. 13, 205e217. Noot, D.K., Anderson, W.B., Daignault, S.A., Williams, D.T., Huck, P. M., 1989. Evaluating treatment processes with the Ames mutagenicity assay. J. Am. Water Works Assoc. 81, 87e102. Ohe, T., Watanabe, T., Wakabayashi, K., 2004. Mutagens in surface waters: a review. Mutat. Res. 567, 109e149.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 8 1 e3 6 9 1

Ohe, T., White, P.A., DeMarini, D.M., 2003. Mutagenic characteristics of river waters flowing through large metropolitan areas in North America. Mutat. Res. 534, 101e112. Ono, Y., Somiya, I., Kawamura, M., Uenishi, K., 1992. Genotoxicity of organic substances in municipal sewage and its ozonated products. Water Sci. Technol. 22, 501e517. Paraskeva, P., Graham, N.J., 2002. Ozonation of municipal wastewater effluents. Water Environ. Res. 74, 569e581. Parzefall, W., Monschau, P., Schulte-Hermann, R., 1989. Induction by cyproterone acetate of DNA synthesis and mitosis in primary cultures of adult rat hepatocytes in serum free medium. Arch. Toxicol. 63, 456e461. Peru, N., Doledec, S., 2010. From compositional to functional biodiversity metrics in bioassessment: a case study using stream macroinvertebrate communities. Ecol. Indicat 10, 1025e1036. Petala, M., Samaras, P., Zouboulis, A., Kungolos, A., Sakellaropoulos, G., 2006a. Ecotoxicological properties of wastewater treated using tertiary methods. Environ. Toxicol. 21, 417e424. Petala, M., Samaras, P., Zouboulis, A., Kungolos, A., Sakellaropoulos, G.P., 2008. Influence of ozonation on the in vitro mutagenic and toxic potential of secondary effluents. Water Res. 42, 4929e4940. Petala, M., Tsiridis, V., Samaras, P., Zouboulis, A., Sakellaropoulos, G.P., 2006b. Wastewater reclamation by advanced treatment of secondary effluents. Desalination 195, 109e118. Reungoat, J., Macova, M., Escher, B.I., Carswell, S., Mueller, J.F., Keller, J., 2010. Removal of micropollutants and reduction of biological activity in a full scale reclamation plant using ozonation and activated carbon filtration. Water Res. 44, 625e637. Rosenkranz, H.S., Mermelstein, R., 1983. Mutagenicity and genotoxicity of nitroarenes. All nitro-containing chemicals were not created equal. Mutat. Res. 114, 217e267. Sakamoto, H., Hayatsu, H., 1990. A simple method for monitoring mutagenicity of river water. Mutagens in Yodo river system, KyotoeOsaka. Bull. Environ. Contam. Toxicol. 44, 521e528. Schaar, H., Clara, M., Gans, O., Kreuzinger, N., 2010. Micropollutant removal during biological wastewater treatment and a subsequent ozonation step. Environ. Pollut. 158, 1399e1404. Selgen, O.P., 1976. Preparation of isolated rat liver cells. In: Prescott, D.M. (Ed.), Methods in Cell Biology. Academic Press, New York, NY, pp. 29e81. Shaugnessy, D.T., DeMarini, D.M., 2009. Types and consequence of DNA damage. In: Knasmueller, S., DeMarini, D.M., Johnson, I., Gerhauser, C. (Eds.), Chemoprevention of Cancer and DNA Damage by Dietary Factors. Wiley-VCH, Weinheim, pp. 21e34. Stalter, D., Magdeburg, A., Oehlmann, J., 2010a. Comparative toxicity assessment of ozone and activated carbon treated sewage effluents using an in vivo test battery. Water Res. 44, 2610e2620. Stalter, D., Magdeburg, A., Wagner, M., Oehlmann, J., 2011. Ozonation and activated carbon treatment of sewage

3691

effluents: removal of endocrine activity and cytotoxicity. Water Res. 45, 1015e1024. Stalter, D., Magdeburg, A., Weil, M., Knacker, T., Oehlmann, J., 2010b. Toxication or detoxication? In vivo toxicity assessment of ozonation as advanced wastewater treatment with the rainbow trout. Water Res. 44, 439e448. Takanashi, H., Mayumi, M., Kato, M., Hirata, M., Hano, T., 2002. Removal of mutagen precursor from wastewater by activated sludge and oxidation treatment. Water Sci. Technol. 46, 389e394. Thruman, R., Kauffman, F., 1980. Factors regulating drug metabolism in intact hepatocytes. Pharmacol. Rev. 31, 229e251. Tice, R.R., Agurell, E., Anderson, D., Burlinson, B., Hartmann, A., Kobayashi, H., Miyamae, Y., Rojas, E., Ryu, J.C., Sasaki, Y.F., 2000. Single cell gel/comet assay: guidelines for in vitro and in vivo genetic toxicology testing. Environ. Mol. Mutagen 35, 206e221. UKEMS, 1990. Basic mutagenicity tests: UKEMS recommended procedures. In: Kirkland, D.J. (Ed.), UKEMS Sub-Committee on Guidelines for Mutagenicity Testing: Report, Part 1. Cambridge University Press, Cambridge, New York, Melbourne, Sydney. Van Hoof, F., 1983. Removal and formation of mutagenic activity by ozone, environmental impact and benefit. Proc. ICA Symposium, Brussels, pp. 410e423. Vartiainen, T., Liimatainen, A., 1986. High levels of mutagenic activity in chlorinated drinking water in Finland. Mutat. Res. 169, 29e34. Victorin, K., 1992. Review of the genotoxicity of ozone. Mutat. Res. 277, 221e238. Victorin, K., 1996. Genotoxicity and carcinogenicity of ozone. Scand. J. Work Environ. Health 22 (Suppl. 3), 42e51. von Gunten, U., 2003a. Ozonation of drinking water: part I. Oxidation kinetics and product formation. Water Res. 37, 1443e1467. von Gunten, U., 2003b. Ozonation of drinking water: part II. Disinfection and by-product formation in presence of bromide, iodide or chlorine. Water Res. 37, 1469e1487. Watanabe, M., Igarashi, T., Kaminuma, T., Sofuni, T., Nohmi, T., 1994. N-hydroxyarylamine O-acetyltransferase of Salmonella typhimurium: proposal for a common catalytic mechanism of arylamine acetyltransferase enzymes. Environ. Health Perspect. 102 (Suppl. 6), 83e89. Watanabe, M., Sofuni, T., Nohmi, T., 1993. Comparison of the sensitivity of Salmonella typhimurium strains YG1024 and YG1012 for detecting the mutagenicity of aromatic amines and nitroarenes. Mutat. Res. 301, 7e12. White, P.A., Claxton, L.D., 2004. Mutagens in contaminated soil: a review. Mutat. Res. 567, 227e345. Wigilius, B., Boren, H., Carlberg, G.E., Grimvall, A., Moller, M., 1985. A comparison of methods for concentrating mutagens in drinking water-recovery aspects and their implications for the chemical character of major unidentified mutagens. Sci. Total Environ. 47, 265e272. Zoeteman, B.C.J., Hrubec, J., De Greef, E., Kool, H.J., 1982. Mutagenic activity associated with by products of drinking water disinfection by chlorine, chlorine dioxide, ozone and UV-irradiation. Environ. Health Perspect. 46, 197e205.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 9 2 e3 7 0 3

Available at www.sciencedirect.com

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

Increased sediment oxygen uptake caused by oxygenation-induced hypolimnetic mixing Lee D. Bryant*, Paul A. Gantzer 1, John C. Little Department of Civil and Environmental Engineering, 418 Durham Hall, Virginia Tech, Blacksburg, VA 24061, USA

article info

abstract

Article history:

Hypolimnetic oxygenation systems (HOx) are increasingly used in lakes and reservoirs to

Received 23 December 2010

elevate dissolved oxygen (O2) while preserving stratification, thereby decreasing concen-

Received in revised form

trations of reduced chemical species in the hypolimnion. By maintaining an oxic zone in

1 April 2011

the upper sediment, HOx suppress fluxes of reduced soluble species from the sediment

Accepted 12 April 2011

into the overlying water. However, diminished HOx performance has been observed due to

Available online 19 April 2011

HOx-induced increases in sediment O2 uptake. Based on a series of in situ O2 microprofile

Keywords:

sediment-water interface as a function of HOx operation. These data were used to deter-

Sediment-water flux

mine how sediment O2 uptake rate (JO2 ) and sediment oxic-zone depth (zmax) were affected

and current velocity measurements, this study evaluates the vertical O2 distribution at the

Sediment oxic zone

by applied oxygen-gas flow rate, changes in near-sediment mixing and O2 concentration,

Microprofile

and proximity to the HOx. The vertical sediment-water O2 distribution was found to be

Hypolimnetic oxygenation

strongly influenced by oxygenation on a reservoir-wide basis. Elevated JO2 and an oxic

Lake and reservoir management

sediment zone were maintained during continuous HOx operation, with zmax increasing linearly with HOx flow rate. In contrast, JO2 decreased to zero and the sediment became

In situ

anoxic as the vertical O2 distribution at the sediment-water interface collapsed during periods when the HOx was turned off and near-sediment mixing and O2 concentrations decreased. JO2 and zmax throughout the reservoir were found to be largely governed by HOxinduced mixing rather than O2 levels in the water column. By quantifying how JO2 and zmax vary in response to HOx operations, this work (1) characterizes how hypolimnetic oxygenation affects sediment O2 dynamics, (2) contributes to the optimization of water quality and management of HOx-equipped lakes and reservoirs, and (3) enhances understanding of the effect of mixing and O2 concentrations in other systems. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Dissolved oxygen (O2) has been identified as one of the most critical environmental factors controlling water quality and associated ecological conditions (Hondzo et al., 2005). Aquatic ecosystems, hydropower plants, and water quality are all negatively affected by depleted O2 concentrations (Beutel and

Horne, 1999). Water-quality standards typically require O2 > 150 mmol L
1 to protect aquatic life (EPA, 2000). Hydropower plants are usually required to meet these minimum O2 levels in the water they discharge downstream (Mobley et al., 2000a). Long-term O2 depletion in fish habitats has been shown to cause significant declines in fish populations as a result of endocrine system disruptions and reproductive

* Corresponding author. Present address: Department of Civil and Environmental Engineering, Box 90287, 121 Hudson Hall, Duke University, Durham, NC 27708, USA. Tel.: þ1 919 660 5034; fax: þ1 919 660 5219. E-mail addresses: [email protected] (L.D. Bryant), [email protected] (P.A. Gantzer), [email protected] (J.C. Little). 1 Present address: Gantzer Water Resources LLC, 14816 119th Place NE, Kirkland, WA 98084, USA. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.018

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 9 2 e3 7 0 3

impairment, with low O2 possibly having more of an effect than anthropogenic chemicals (Wu et al., 2003). Organic or nutrient loading of thermally stratified lakes and reservoirs may lead to extensive depletion of O2 in the deeper hypolimnetic water. Hypolimnetic O2 depletion can result in the release of soluble chemical species from the sediment, decreasing water quality and increasing drinking-water treatment costs (Gantzer et al., 2009a). Oxygen depletion in lakes and reservoirs is largely controlled by sediment O2 uptake which is regulated by near-sediment hydrodynamics and the intrinsic sediment O2 demand (Veenstra and Nolen, 1991; O’Connor et al., 2009). Understanding the processes controlling the sediment O2 uptake rate (JO2 ) and other sediment-water fluxes is crucial for optimizing water quality and successfully managing lakes and reservoirs (Zhang et al., 1999; Beutel, 2003). Hypolimnetic oxygenation systems (HOx) are used increasingly by drinking water and hydropower utilities to replenish O2 and decrease concentrations of soluble metals, such as iron (Fe) and manganese (Mn), and other chemical species in source water while preserving stratification (McGinnis and Little, 2002; Beutel et al., 2007; Gantzer et al., 2009b). While several types of systems are used for hypolimnetic oxygenation (Singleton and Little, 2006; Moore and Christensen, 2009), this study is based on the performance of a bubble-plume diffuser HOx. Bubble-plume HOx release oxygen gas from diffusers positioned near the reservoir bottom and impart relatively low levels of mixing within the hypolimnion to maintain overall thermal structure. These HOx thereby increase hypolimnetic O2 concentrations while suppressing mixing of the epilimnion and hypolimnion and preventing destratification (Wu¨est et al., 1992; McGinnis et al., 2004). Ideally, an HOx increases O2 availability both in the water column and in the upper sediment to prevent the release of reduced chemicals into the hypolimnion (Zaw and Chiswell, 1999; Beutel 2003). The balance between the amount of O2 supplied to the sediment via JO2 and the amount consumed via various sediment biogeochemical processes (e.g., benthic mineralization of organic matter and oxidation of reduced chemicals) governs sediment oxic-zone depth (zmax) and sediment O2 availability (Glud et al., 2007). Oxygenation may, however, cause excessive O2 uptake in the hypolimnion and sediment due to HOx-induced increases in hypolimnetic O2 concentrations and turbulent mixing (Moore, 2003; Gantzer et al., 2009b). JO2 is a function of the O2 concentration driving force across the diffusive boundary layer (DBL), a mm-scale laminar layer immediately above the sediment-water interface (SWI; Jørgensen and Revsbech, 1985). Turbulence in the overlying water governs DBL thickness (dDBL; Lorke et al., 2003; Bryant et al., 2010a) and hence directly affects JO2 . Despite the potential influence that HOx may have on sediment O2 uptake, reservoir-specific JO2 measurements are rarely available and HOx are often designed based on O2 depletion rates measured prior to installation of the systems (Moore et al., 1996; Mobley et al., 2000b; Beutel et al., 2007). While the relationship between near-sediment current velocity and JO2 has been investigated (Gundersen and Jorgensen, 1990; Mackenthun and Stefan, 1998), little work has been done to quantify how HOx operations affect JO2 and

3693

the vertical distribution of O2 at the SWI. Most studies performed have been laboratory based (Moore et al., 1996; Beutel, 2003). However, it has been shown that JO2 can be significantly affected by variations in natural turbulence (Lorke et al., 2003; Bryant et al., 2010a) and laboratory studies may not capture actual HOx-driven conditions. The research presented here is therefore based on in situ O2 sediment-water microprofiles and near-sediment current velocity to address five key themes related to the influence of HOx on sediment O2 uptake. This study characterizes how (1) the vertical O2 distribution at the SWI and (2) sediment O2 dynamics (quantified by JO2 ) are affected by HOx operations in a drinking-water-supply reservoir. It also evaluates (3) the sediment oxic zone (quantified by zmax) as a function of HOx oxygen-gas flow rate and (4) spatial variation in the influence of the HOx. Finally, (5) broader impacts on reservoir water quality are assessed. Focus is first placed on a multi-day in situ campaign that shows how JO2 and the vertical O2 distribution at the SWI respond to turning the HOx off for w48 h and then back on. Building on these results, a multi-year data set is then used to quantify the influence of HOx on sediment O2 conditions using a broader range of flow rates. To the authors’ knowledge, this is the first study to assess in situ how HOx-induced variation in nearsediment mixing and O2 concentrations influence JO2 and the sediment oxic zone on a reservoir-wide scale.

2.

Materials and methods

2.1.

Study site

This research focused on Carvins Cove Reservoir (CCR), which is managed by the Western Virginia Water Authority to provide drinking water to the county of Roanoke, Virginia, USA. CCR is a stream-fed lake that has been managed as a drinking-watersupply reservoir since the late 1940s. CCR is eutrophic and has a maximum depth of 23 m, width of w600 m, and length of w8000 m (Fig. 1). In 2005, a bubble-plume line diffuser HOx was installed in the deepest section of the reservoir near the water treatment plant withdrawal (Fig. 1) to replenish O2 depleted during summer stratification and to minimize soluble Fe and Mn in the source water (McGinnis and Little, 2002; Gantzer et al., 2009a). The CCR HOx delivers pure oxygen gas over a wide range of flow rates, providing considerable operational flexibility. Thus, it was possible to vary O2 and artificiallyinduced mixing in the reservoir by changing the applied HOx flow rate. Data were collected from 2005 through 2008 to assess how the HOx affects CCR, with substantial improvement in water quality observed since HOx operations began (Gantzer et al., 2009a,b). Research presented here on HOx-induced variation in the vertical O2 distribution at the SWI extends the work of Gantzer et al. (2009a,b), which focused on watercolumn conditions, by evaluating the influence of HOx on sediment O2 dynamics and the sediment oxic zone.

2.2.

Data collection and analysis

This study is based on in situ data obtained during field campaigns using a microprofiler (MP4; Unisense A/S) to obtain profiles at the SWI and an acoustic Doppler current profiler

3694

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 9 2 e3 7 0 3

end of the HOx; Fig. 1). During May to August 2007, monthly MP4 microprofile measurements were also obtained downstream, alongside, and 1000 m upstream of the HOx at sites CCR-1, CCR-2, and CCR-6, respectively, while the HOx was operated at four different oxygen-gas flow rates increased incrementally on a monthly basis from 17 to 70 m3 h
1. Each day that O2 microprofile measurements were obtained, conductivity, temperature, and O2 as a function of depth (CTD) profiles of the water column were measured at each sampling location using a Seabird Electronics SBE 19plus profiler. The SBE 19plus has a 4-Hz sampling rate and an O2 sensor response time of 1.4 s.

2.2.1.

Fig. 1 e Map of Carvins Cove Reservoir (CCR) showing linear bubble-plume hypolimnetic oxygenation system (HOx) and sampling sites (near-field locations CCR-1 (0 m; relative to start of HOx lines) and CCR-2 (189 m); midreservoir locations CCR-3 thru CCR-6 (683, 1011, 1373, and 1814 m, respectively); back-reservoir location CCR-7 (3000 m)).

(ADCP; Teledyne RDI, Inc.) for current velocities in the water column. During the primary field campaign in 2008, HOx flow was typically maintained at 51 m3 h
1 (or w1580 kg O2 d
1). However, the HOx was turned off for w48 h during two experimental campaigns (each w1e2 weeks in duration) in June and August 2008 to track the response of the vertical O2 distribution at the SWI and corresponding JO2 . The relatively short period that the HOx was turned off was planned to preserve reservoir water quality by minimizing effects on hypolimnetic O2 levels. Previous work by Gantzer et al. (2009b) showed that turning the HOx off for a longer period (e.g., several weeks) can impair water quality substantially due to the accumulation of reduced compounds as the hypolimnion becomes hypoxic. During the first campaign, the MP4 microprofiler and ADCP were deployed alongside the HOx at site CCR-2 (Fig. 1) from June 19 to 26. The HOx was turned off from June 19 to 21. Data were downloaded and batteries for the MP4 and the ADCP were exchanged daily. A similar campaign was performed in August at the mid-reservoir site CCR-6, w1000 m upstream of the end of the HOx, and data were collected almost continuously from August 18 to 30. The HOx was turned off from August 19 to 21. While analogous results were obtained from both campaigns, reported results are based largely on the August CCR-6 campaign due to insufficient background data for the June CCR-2 campaign. In addition to these multi-day campaigns, MP4 and ADCP data were collected monthly from June to September 2008 at CCR-2, CCR-6, and CCR-7 (located w2000 m upstream of the

O2 microprofiles

The in situ autonomous MP4 microprofiler equipped with microsensors (O2 and temperature) was used to obtain microprofiles at the SWI. The O2 microsensor (OX-100; Unisense A/S) had a 100-mm tip diameter and depth resolution, fast response time (90% in <8 s), and negligible stirring sensitivity. The O2 microsensor was a Clark-type sensor with an internal reference and guard cathode. The temperature microsensor (TP-100; Unisense A/S) was a thermo-coupled sensor with a tip diameter and depth resolution of w200 mm, measurement resolution of 0.1 mV per  C, and a 90% response time of <3 s. Profiles were measured nearly continuously (exceptions include while data were downloaded and a period on August 26e27 when a storm prevented use of the boat to download data) during the multi-day campaigns and obtained in duplicate at each CCR sampling location during monthly measurements. Profiles were obtained as follows: 10-mm resolution from 10 cm to 1 cm above the SWI, 1-mm resolution from 1 cm to 0.5 cm above the SWI, and 0.1-mm resolution from 0.5 cm above the SWI to 0.5 cm below the SWI. The SWI location was determined as described by Bryant et al. (2010a,b). A video camera was used periodically to ensure that the MP4 remained stable and did not sink. Ten measurements were typically taken at each depth. For the multi-day campaigns, however, three measurements were obtained per depth due to data-storage limitations during overnight deployments. Following a pause between measurements for equilibrium to be established at each depth, microsensor data were collected at a rate of 1 Hz. Time required to obtain a full profile was w50e70 min and profiles are referenced by the time when the microsensor encountered the SWI. A two-point, linear calibration of the O2 microsensor was performed using zero readings from anoxic sediment and Winkler titration of water sampled immediately above the sediment using a Kemmerer bottle and/or sediment cores. zmax was designated as the depth where O2 drops to <3 mmol L
1.

2.2.2.

Current velocity

Velocity profiles were collected using a 1200 kHz Workhorse Rio Grande ADCP equipped with four transducers in a janus configuration with a beam angle of 20 . The ADCP, positioned adrift alongside the research vessel and facing downward from the water surface, profiled the water-column depth using a 1-m bin size. Samples were obtained in a multi-ping mode with 50 samples per ensemble at a rate of 2 Hz. Accuracy of velocity measurements was 0.25% of water-plus-boat velocity 0.25 cm s
1. ADCP motion relative to the sediment

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 9 2 e3 7 0 3

was measured using bottom tracking. The boat remained stationary during measurements and boat velocity was negligible. Near-sediment velocities were of primary interest; however, the ADCP has a blanking distance near the sediment due to interference from the angled acoustic beams reflecting off the sediment surface, thereby contaminating the acoustic signal returned to the ADCP. The near-sediment blank zone is w6% of the total water-column depth, as defined by RDI, or approximately the bottom 1 m in CCR. Hence, near-sediment velocities were evaluated at w2 m above the sediment.

2.2.3.

O2 microprofile analyses

JO2 and dDBL were evaluated based on O2 microprofile data using Fick’s law (Rasmussen and Jørgensen, 1992):


h i vC
vC
Cbulk
CSWI


1 mmolm
2 d JO2 ¼
4Ds

¼
D

¼D
dDBL vz sed vz water water (1) where 4 is sediment porosity (m3 voids m
3 total volume), Ds is the diffusion coefficient for O2 in sediment (m2 s
1), D is the diffusion coefficient for O2 in water (m2 s
1), vC/vz is the linear O2 concentration gradient in the DBL water above or in the sediment immediately below the SWI (mmol m
4), Cbulk is the O2 concentration in the bulk water (mmol L
1), and CSWI is the O2 concentration at the SWI (mmol L
1). JO2 and depth z are defined positive downwards into the sediment. Sediment cores from the primary sampling locations were used to evaluate 4 following Dalsgaard et al. (2000) and 4 values of 0.95e0.97 were obtained for the upper 1 cm of sediment. Values for D were based on D ¼ 1.97  10
9 m2 s
1 at 20  C, correcting for temperature using the Stokes-Einstein relationship (Li and Gregory, 1974; Arega and Lee, 2005). Ds was defined as Ds ¼ 4D to correct for sediment tortuosity as a function of 4 (Berg et al., 1998; Glud, 2008; Bryant et al., 2010b). The temporal change in O2 concentration (vC/vt) was evaluated for the series of in situ O2 profiles by comparing profiles immediately before and after one another and calculating the rate of change in O2 at each depth. vC/vt was found to be on average <5% of JO2 , establishing that measured profiles were at quasi-steady state. Fick’s law may be applied to either water- or sediment-side data to evaluate JO2 and dDBL (Rasmussen and Jørgensen, 1992; Bryant et al., 2010a,b). Diffusive transport in the sediment is given by the second term in Eq. (1) and in the water by the third and fourth terms. JO2 was evaluated as a function of (vC/ vz)sed from sediment O2 porewater data using the second term in Eq. (1). Because JO2 estimates were based on porewater data, water-side data were used for dDBL estimates to allow for an independent comparison with JO2 . Water-side dDBL was estimated as a function of the measured (vC/vz)water in the DBL using the third and fourth terms in Eq. (1).

2.2.4.

Turbulence estimations

Turbulence has been shown by Lorke et al. (2003) to have a more direct influence on dDBL and JO2 estimates than current velocity. Turbulence is characterized by the dissipation rate of turbulent kinetic energy, e (W kg
1), which is frequently estimated by applying the inertial dissipation method (Grant et al., 1962) to near-sediment velocity data, as performed by Bryant et al. (2010a) and Lorke et al. (2003) using acoustic

3695

Doppler velocimeter (ADV) data measured at 10 cm and 1 m above the sediment, respectively. However, as discussed in Section 2.2.2, obtaining near-sediment velocity measurements was restricted by the limited accuracy of the ADCP approaching the sediment. Estimates of velocity at w2 m above the sediment are thus a relative measure of nearsediment mixing conditions. To quantify turbulence levels in the absence of more precise near-sediment velocity data, a correlation between dDBL, the viscous boundary layer (the cm-scale region immediately above the DBL), and friction velocity (u*) was used to estimate e. According to Wu¨est and Lorke (2003), dDBL and viscous boundary layer thickness (dn) are related by:

dDBL ¼ dv

 1=3 D ½m v

(2)

where v is the kinematic viscosity of water (m2 s
1). In turn, dv is defined as a function of u* (Schlichting, 2000) via: dv ¼

11v ½m u

(3)

Combining Eqs. (2) and (3) allows u* to be related to dDBL:  1=3 D   v m s
1 dDBL

11v u ¼

(4)

The u* values estimate the frictional stress of currents on the sediment and, similar to e, characterize near-sediment turbulence. Estimated u* values obtained via Eq. (4) were used to calculate e using the law-of-the-wall assumption (Lorke et al., 2003): e¼

i u3 h
1 W kg kh

(5)

where k (the von Karman constant) is 0.41 and h is height above the sediment. For these estimates, e was evaluated at an assumed h ¼ 10 cm to obtain near-sediment e predictions and to allow for direct comparison with e based on ADV data measured in a similar system at h ¼ 10 cm (Bryant et al., 2010a). This comparison verified that e estimates for this study were typical for a freshwater lake.

3.

Results and discussion

3.1.

Vertical O2 distribution at SWI

Substantial variation in the vertical O2 distribution on both sides of the SWI was observed in response to halting oxygen flow for w48 h during the 2008 campaigns. Profiles (each of w50 min duration) were collected almost continuously over the course of both multi-day campaigns in 2008. A summary of profile results during the August CCR-6 campaign is presented in Fig. 2. Prior to the point at which the HOx was turned off on August 19 at w15:00, O2 concentrations were relatively high both in the water immediately above the sediment (w125 mmol L
1) and within the sediment porewater (w80 mmol L
1 at the SWI with zmax of 0.8 mm; Fig. 2a). A constant Cbulk and a well-defined DBL are also evident,

3696

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 9 2 e3 7 0 3

signifying active near-sediment turbulence. However, as indicated by the 17:39 profile, O2 started to become depleted from the sediment and overlying water within a few hours of turning the HOx off (Fig. 2b). Approximately 5 h after halting oxygenation, the sediment and overlying water had become completely anoxic (profile 20:02). A transient oxic period occurred on August 21 (profiles 15:11, 15:58; Fig. 2c) when near-sediment O2 and zmax increased immediately after HOx operation resumed, which may be attributed to a large internal wave induced by turning on the HOx. Excluding this brief period, however, conditions remained almost completely anoxic until August 29 (Fig. 2c,d). A relatively rapid increase in O2 is then observed on both sides of the SWI, with the vertical O2 distribution returning to a structure similar to that at the beginning of the campaign (Fig. 2a). Contour plots using a Kriging interpolation scheme were created based on the full 255-profile series of O2 microprofile data obtained during the August CCR-6 campaign, background microprofiles, and corresponding CTD hypolimnetic O2 data (Fig. 3). While O2 concentrations near the SWI changed substantially in response to turning off the HOx for w48 h (Fig. 3a), O2 remained relatively constant (w200 mmol L
1) at 8 cm above the sediment (Fig. 3b) and the bulk water column was affected negligibly (Fig. 3c). The fact that O2 in the overlying water remained largely unaffected while the sediment became anoxic emphasizes how sediment O2 uptake depends on continual operation of the HOx. Although the HOx was turned off for only w48 h, it took w8 days for the vertical O2 distribution at the SWI to be restored during the August CCR-6 campaign (Figs. 2c,d and 3a). This delayed response may be attributed both to time required for a uniform flow pattern in the hypolimnion to be reestablished and to localized sediment resuspension effects (discussed further in Section 3.4). During a six-year study in an HOx-equipped reservoir similar to CCR, it was observed that it typically took w1 week for ‘steady-state’ O2 conditions to return in the water column after resuming oxygenation, with substantially higher initial O2 depletion rates in the hypolimnion (Gantzer, 2002; Gantzer et al., 2009a). The considerable influence that HOx operations had on the vertical sediment-water O2 distribution mid-reservoir at CCR6 (Figs. 2 and 3) was also evident in the near field at CCR-2 (Fig. 4). A strong correlation was found between HOx operation and parameters quantifying the vertical O2 distribution (O2 at 5 cm above the sediment (C5), CSWI, and zmax) during both the June and August 2008 campaigns. At CCR-6, as CSWI dropped from w80 to 0 mmol L
1 and C5 decreased from w200 to 50 mmol L
1 after turning off the HOx on August 19, zmax also rapidly decreased from 1 to 0 mm as O2 was depleted from the sediment (Fig. 4a). Apart from the brief oxic period on August 21 after the HOx was turned back on, sediment porewater and the SWI remained anoxic and C5 remained low until August 29 Fig. 2 e Summary of in situ dissolved oxygen (O2) profile data obtained at the sediment-water interface (SWI; at depth 0 mm) with the MP4 microprofiler during the August 2008 CCR-6 campaign. The HOx was turned off on August 19 (w15:00) and turned back on w48 h later on August 21 (w12:00). Profile data obtained prior to turning the HOx off are shown in (a). Profiles characterizing the periods when

the HOx was turned off and then back on are shown in (b) and (c, d), respectively. (b) Following the halt of HOx operations, O2 rapidly depleted from the sediment and the overlying water column. (c, d) An oxic vertical O2 distribution was not re-established until August 29, 8 days after the HOx was turned back on.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 9 2 e3 7 0 3

3697

Fig. 3 e Contour plots of O2 concentrations at the SWI (a; MP4 microprofile data), in the water overlying the sediment (b; MP4 microprofile data), and in the hypolimnion (c; CTD (conductivity-temperature-O2 as a function of depth) profile data). Data from the full set of 255 MP4 O2 profiles obtained during the August 2008 on/off campaign as well as from monthly profiles measured before and after the campaign are presented in (a, b). Shaded region indicates period when HOx was turned off. While O2 is rapidly depleted from the sediment and water immediately above the SWI after the HOx was turned off (a), O2 concentrations were only minimally affected at w8 cm above the sediment (b) and remained relatively constant in the bulk hypolimnion (c). In (a) and (b), depth represents distance above (L) or below (D) the SWI, which is indicated by the dashed line. Water-column depth in (c) is also characterized by distance above (L) the sediment.

3698

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 9 2 e3 7 0 3

when oxic sediment conditions were re-established. Similar results were observed at CCR-2 after turning off the HOx on June 19 as CSWI and C5 both dropped to 0 mmol L
1 and zmax decreased from 0.5 to 0 mm (Fig. 4b). The response time for the re-establishment of an oxic sediment-water O2 distribution was also delayed at CCR-2, taking w5 days. A transient oxic period after resuming oxygenation was not observed at CCR-2 perhaps due to enhanced sediment resuspension in this nearfield region after turning on the HOx (discussed in Section 3.4). Due to equipment issues and the fact that O2 depletion occurred more rapidly than expected following the halt of HOx operation, data characterizing the initial O2 depletion phase were not obtained during the preliminary June campaign (Fig. 4b). Data measured before and after each multi-day campaign (e.g., August 12 and September 14; Fig. 4a) compare well with average CSWI and zmax values during normal continuous HOx operation (CSWI ¼ 94  38 mmol L
1 and zmax ¼ 1.4  0.6 mm at CCR-6 (n ¼ 22) and 61  45 mmol L
1 and 0.7  0.4 mm (n ¼ 15) at CCR-2).

3.2.

force at the SWI and suppression of the DBL resulting from increased O2 and turbulence levels, respectively, in the lower hypolimnion (Eq. (1); Bryant et al., 2010a,b). Sediment O2 dynamics (as characterized by JO2 ) and water-side controls on diffusive flux (as characterized by CSWI and dDBL) were affected considerably by HOx operations, as shown by CCR-6 data in Fig. 5a. During ongoing oxygenation, dDBL was suppressed while JO2 and CSWI remained elevated. Average summer JO2 and dDBL at CCR-6 were 12.5  7.6 mmol m
2 d
1 and 1.6  0.9 mm (n ¼ 22), respectively. However, in response to turning off the HOx on August 19, dDBL increased from 0.6 mm to the point of becoming undefined (Fig. 5a; no discernable DBL was measurable at dDBL >5 mm; hence, a nominal maximum dDBL ¼ 5 mm was assumed). Simultaneously, JO2 decreased from 12.5 to 0 mmol m
2 d
1 and CSWI dropped from w80 to 0 mmol L
1 as diffusive transport of O2 was restricted and O2 was depleted from the sediment. Per Eq. (1), as dDBL increases to the point of becoming undefined in the absence of turbulent mixing and the O2 concentration driving force

HOx-induced variation in sediment O2 dynamics

While HOx are designed to remediate problems caused by hypolimnetic O2 depletion, conceptually these systems should also increase sediment O2 uptake via enhanced O2 flux into the sediment due to an elevated O2 concentration driving

Fig. 4 e Variations in O2 concentrations at the SWI (CSWI), at 5 cm above the SWI (C5), and within the sediment (as characterized by sediment oxic-zone depth zmax) in response to turning off the HOx. Similar trends are observed in data from (a) the August 2008 campaign at CCR-6 and also (b) the June 2008 campaign at CCR-2. During the period of zero flow when the HOx was not in operation, O2 concentrations dropped both at 5 cm above and directly at the SWI and the sediment became anoxic as O2 was depleted.

Fig. 5 e The response of JO2 and water-side parameters influencing diffusive flux (a; diffusive boundary layer thickness (dDBL) and CSWI) to turning off the HOx corresponds directly to variations in near-sediment mixing (b) as characterized by current velocity at 2 m above the sediment (obtained via ADCP), turbulence (defined by energy dissipation rate (e) estimated as a function of waterside dDBL), and temperature at the SWI (obtained via MP4 temperature microsensor). Results shown are based on in situ data obtained during the August 2008 campaign at CCR-6; similar results were obtained from the June CCR-2 campaign. Data in Fig. 5 are average bi-daily values (standard deviation data provided in Table 1). The period during which the HOx was turned off is designated by the shaded region.

3699

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 9 2 e3 7 0 3

becomes negligible when the sediment and overlying water are anoxic, JO2 goes to zero. Analogous to O2 data presented in Figs. 2c and 4a, JO2 and CSWI increased and dDBL decreased briefly after the HOx was turned back on (Fig. 5a, Table 1). However, JO2 and CSWI remained at approximately zero until August 29 when a steady-state flow pattern and oxic conditions at the SWI returned. As a well-defined and thinner DBL was re-established and diffusive O2 transport to the sediment resumed by August 29, CSWI increased to w70 mmol L
1 and JO2 increased to w15 mmol m
2 d
1 (Fig. 5a, Table 1). Average bidaily values are shown in Fig. 5 with standard deviations provided in Table 1 (variation in e not provided as this parameter is directly related to dDBL). The response of JO2 and the water-side vertical O2 distribution to HOx operations (Fig. 5a) closely parallels variation in near-sediment current velocity (as measured by ADCP), estimated turbulence levels (as characterized by e), and temperature at the SWI (as measured by temperature microsensor; Fig. 5b). Velocities at 2 m above the sediment dropped sharply soon after the HOx was turned off. After the HOx was turned back on, near-sediment velocities are observed to increase though they remained quite variable. Velocities returned to pre-campaign levels (w5 cm s
1; summer average was 6.2 cm s
1) on August 29, which corresponds with the timing of the re-established vertical O2 structure at the SWI (Figs. 2 and 4). Near-sediment velocity is shown to strongly correlate

with e, both of which decreased considerably when the HOx was turned off (Fig. 5b). The correlation between ADCP velocity data and estimated e values, which are based on O2 microprofile data, supports the evaluation of e as a function of dDBL and highlights the influence of mixing on dDBL. Estimates of dDBL-based e for this study are within the same range as ADV-based e values obtained by Bryant et al. (2010a) for windinduced mixing (i.e., seiching) in a freshwater lake. HOxinduced mixing, which has been shown to result in elevated temperatures in the hypolimnion (Gantzer et al., 2009b; Liboriussen et al., 2009), is further confirmed by variations in temperature at the SWI that closely follow trends in e and current velocity (Fig. 5b). A peak in current velocity, e, and temperature on August 21 corresponds to the oxic vertical O2 distribution, decreased dDBL, and elevated JO2 observed briefly at this time (Figs. 4a and 5) and supports the occurrence of a large internal wave following HOx start-up. These results reveal the controlling influence that HOx operation can have on the degree of mixing in the hypolimnion (Fig. 5b) and corresponding sediment O2 uptake (Fig. 5a). O2 concentrations in the lower hypolimnion remained largely unaffected by halting oxygenation (Fig. 3) with O2 levels staying relatively high only w8 cm above the sediment. Despite these high O2 concentrations, as near-sediment mixing decreased, dDBL increased, JO2 and CSWI decreased to zero, and the sediment and overlying water became anoxic (Figs. 3

Table 1 e Bi-daily averages and standard deviations for sediment oxygen uptake rate (JO2 ), dissolved oxygen (O2) at the sediment-water interface (CSWI), temperature at the sediment-water interface, diffusive boundary layer thickness (dDBL), near-sediment current velocity (U; measured at w2 m above the sediment), and hypolimnetic oxygenation system (HOx) flow rate during the August 2008 CCR-6 campaign. Average values for JO2 , CSWI, temperature, and dDBL (based on microsensor data) were estimated using the number (n) of profiles obtained during the designated time period. Corresponding averages for U were estimated using acoustic Doppler current velocity (ADCP) profile data. Standard deviations based on an assumed normal distribution. Date

Flow (m3 h
1)

n

JO2 (mmol m
2 d
1)

CSWI (mmol L
1)

Temperature ( C)

dDBL (mm)

U (cm s
1)

8/12 18:00 8/18 15:00 8/19 19:00 8/20 0:00 8/20 12:00 8/21 0:00 8/21 12:00 8/22 0:00 8/22 12:00 8/23 0:00 8/23 12:00 8/24 0:00 8/24 12:00 8/25 0:00 8/25 12:00 8/26 0:00 8/26 12:00 8/27 0:00 8/27 12:00 8/28 0:00 8/28 12:00 8/29 0:00 8/29 12:00 8/30 0:00 8/30 12:00 9/14 16:00

51 51 51 to 0 0 0 0 0 to 51 51 51 51 51 51 51 51 51 51 51 51 51 51 51 51 51 51 51 59

2 2 6 7 14 13 13 12 14 13 13 13 13 8 15 15 1 12 9 8 10 9 11 11 13 2

9.0  1.3 6.9  1.5 12.5  4.8 0.0  0.0 0.0  0.0 1.6  2.5 0.0  0.0 6.6  10.3 0.0  0.0 0.0  0.0 0.0  0.0 0.0  0.0 0.0  0.0 0.0  0.0 0.0  0.0 0.0  0.0 0.0  0.0 0.0  0.0 0.0  0.0 0.0  0.0 0.0  0.0 9.4  10.4 0.0  0.0 12.8  10.2 14.6  4.6 16.1  2.3

114.2  11.4 78.0  7.8 68.8  6.9 23.7  2.4 0.0  0.0 3.1  0.3 0.0  0.0 18.1  1.8 0.1  0.0 0.1  0.0 0.3  0.0 0.3  0.0 0.0  0.0 0.0  0.0 1.2  0.1 1.2  0.1 0.0  0.0 0.0  0.0 0.0  0.0 0.0  0.0 0.0  0.0 33.8  3.4 11.4  1.1 62.6  6.3 73.2  7.3 117.0  11.5

13.4  0.0 13.1  0.0 12.7  0.1 12.5  0.1 12.1  0.1 12.1  0.2 12.0  0.1 12.3  0.1 12.1  0.1 12.0  0.1 12.1  0.1 12.0  0.0 12.2  0.2 12.2  0.2 12.0  0.0 12.1  0.0 12.2  0.1 12.1  0.2 11.9  0.0 11.9  0.0 11.8  0.2 12.0  0.1 12.0  0.1 12.3  0.1 12.5  0.2 13.2  0.0

0.8  0.1 1.8  0.2 0.6  0.2 5.0  0.0 5.0  0.0 4.3  1.2 5.0  0.0 3.5  2.1 5.0  0.0 5.0  0.0 5.0  0.0 5.0  0.0 5.0  0.0 5.0  0.0 5.0  0.0 5.0  0.0 5.0  0.0 5.0  0.0 5.0  0.0 5.0  0.0 5.0  0.0 2.4  2.1 5.0  0.0 2.1  2.1 1.2  0.4 0.4  0.1

3.1  0.7 3.8  0.9 4.2  0.1 3.6  0.7 1.9  0.2 1.4  0.2 2.3  0.2 2.5  0.3 2.4  0.9 2.5  0.3 2.1  0.5 2.2  0.7 1.2  0.0 2.1  0.4 1.6  0.5 2.0  0.6 3.1  0.8 3.4  0.2 2.1  1.0 2.8  0.4 2.6  0.8 1.3  0.2 3.8  0.8 4.2  0.9 5.1  1.1 4.5  0.0

3700

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 9 2 e3 7 0 3

and 5). The significant role that HOx-induced turbulent mixing plays in driving oxygenated water down to the SWI and facilitating JO2 is thus clearly demonstrated. Brand et al. (2009) and Bryant et al. (2010a) similarly showed that turbulence from seiching was a controlling factor in maintaining oxic sediment conditions regardless of O2 concentrations only a few cm above the SWI. The influence of seiching on sediment O2 uptake (Bryant et al., 2010a) was expected to be minimal in CCR due to relatively mild average wind speeds and irregular basin bathymetry (Fig. 1). Wind speed (based on National Oceanographic and Atmospheric Administration (NOAA) data for the Roanoke airport, located w5 km from CCR) was assessed to determine if seiching was an influence in observed changes in current velocity (NOAA, 2008). As anticipated, a correlation between average wind speed and current velocities (nearsediment, mid-hypolimnion, and mid-epilimnion; data not shown) was not observed which confirms that seiche-induced turbulence was not a significant factor in sediment-water O2 dynamics in CCR.

3.3.

Correlation between zmax and HOx gas flow

While induced JO2 can be problematic if not properly accounted for in HOx design and operation, increased JO2 is beneficial if an enhanced sediment oxic zone inhibits the transport of reduced soluble species into the water column (Lin et al., 2003; Beutel et al., 2007; Gantzer et al., 2009a). To evaluate the relationship between HOx flow rate and zmax, additional data from summer 2007 were used to cover a broader range of flow rates. The 2008 off/on HOx experiments revealed that zmax and the sediment O2 distribution responded directly to changes in HOx operation based on the flow rates of 0 and 51 m3 h
1 (Fig. 4). These results were expanded upon by including O2 microprofile data obtained at CCR-2 and CCR-6 while the HOx was operated at five different flow rates during 2007e2008 (Fig. 6). The sediment oxic zone (quantified by zmax) was found to be linearly related to HOx flow rate in both the near field and mid-reservoir region, with zmax increasing in accordance with flow rate (Fig. 6) and a P-value of 0.04 for CCR-2 and CCR-6 data. The response of zmax to increased flow rate was similar in both regions, although zmax was slightly lower near the HOx (CCR-2) than further upstream (CCR-6). Conversely, average JO2 values were comparable but higher at CCR-2 than at CCR-6 (13.6  7.2 vs. 6.6  2.4 mmol m
2 d
1). These results suggest that increased O2 reached the near-field sediment but was then consumed more rapidly by sediment O2 consumption processes, resulting in decreased zmax (O’Connor et al., 2009). CCR near-field sediment has been found to have considerably higher levels of total organic carbon, Fe, and Mn in the bulk sediment (Bryant et al., unpubl.) which may result from enhanced oxide precipitation and sediment focusing in the deeper region where the HOx is installed (Fig. 1; Schaller and Wehrli, 1997). Thus, sediment near the HOx most likely has increased sources of electron acceptors and subsequently would have a greater capacity for O2 consumption. Variations in mixing resulting from interaction between the HOx bubble plume and CCR bottom topography (Singleton and Little, 2006; Singleton et al., 2010) also likely affected sediment O2 uptake. However, the fact that fairly similar zmax values were observed

at CCR-2 and CCR-6 indicates that oxygenation maintained a balance between sediment O2 supply and consumption processes in both the near- and far-field regions.

3.4.

Spatial variation in influence of HOx

The response of the vertical sediment-water O2 distribution to halting oxygenation was found to be similar both near the HOx at CCR-2 and mid-reservoir at CCR-6 (Fig. 4). Furthermore, ongoing HOx operations maintained a fairly uniform sediment oxic zone throughout most of CCR (Fig. 6) and average JO2 values were comparable in both the near field and midreservoir region (Section 3.3). Results do indicate, though, that proximity to the HOx influenced localized flow patterns and the time required for a vertical O2 distribution to be reestablished at the SWI after resuming oxygenation. Ideally, HOx-induced turbulence establishes a gently mixed hypolimnion (Singleton and Little, 2006) which is supported by ADCP measurements in CCR indicating relatively similar current velocities throughout the reservoir (data not shown). However, short-circuiting of plume-induced flow can occur in the region near the HOx (McGinnis et al., 2004; Singleton et al., 2010). Although the far field is usually less affected by the plume, HOx plume model results by Singleton et al. (2007) showed that the area most directly influenced by the HOx was a detrainment region between the depth of maximum plume rise and the fallback elevation of equal density. CCR-6 is located within this detrainment region and may therefore be subject to more intense turbulence, while CCR-2 is located below the fallback elevation. As shown in Fig. 4, it took w5 days for the vertical O2 distribution to be restored at CCR-2 and w8 days at CCR-6. The time required for suspended particles to settle in the near field and for HOx-induced near-sediment mixing to resume could

Fig. 6 e Variation in zmax as a function of HOx flow rate. A linear relationship between the sediment oxic zone and HOx flow is observed, with zmax increasing in response to elevated HOx flow at both locations (P-value of 0.04 for CCR-2 and CCR-6 data). On average, 2e3 measurements were obtained per flow rate, with the exception of data obtained during the 2008 multi-day campaigns for which bi-daily averages were used.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 9 2 e3 7 0 3

have contributed to this delayed response (Figs. 4 and 5). Sediment resuspension and the introduction of reduced species into the hypolimnion, particularly near the HOx, likely occur immediately after turning on the HOx (Gantzer, 2002). Enhanced sediment resuspension at CCR-2 may have prevented the brief establishment of oxic conditions as observed at CCR-6 (Fig. 4). However, the vertical O2 distribution at the SWI was most likely restored more quickly in the near field due to local mixing effects. At CCR-6, the response was more gradual as it took longer for HOx-induced increases in O2 and near-sediment turbulence to influence sediment O2 conditions up-reservoir. Once an oxic vertical O2 distribution was established, though, O2 microprofiles measured at CCR-6 (Fig. 2) typically maintained a more stable, well-defined structure than those at CCR-2. This may be attributed to minimal sediment disturbance and enhanced turbulence due to plume detrainment mid-reservoir as compared to near the HOx where sediment resuspension and localized flow instabilities are likely more prevalent (Singleton et al., 2010). Less consistent conditions near the HOx are supported by increased variation in CCR-2 data (Fig. 6) especially at high flow rates.

3.5.

Effect on source water quality

While HOx operations were found to increase sediment O2 uptake (Figs. 5 and 6), an O2 balance performed on the CCR water column by Gantzer et al. (2009b) showed that enough O2 can be supplied by the HOx to counteract this increased demand and prevent O2 depletion within the water column. The work presented here on sediment O2 uptake supports that HOx can effectively replenish O2 in thermally stratified watersupply reservoirs such as CCR as long as HOx-induced increases in JO2 are properly accounted for. Though additional factors like external nutrient loading may be important (Ga¨chter and Wehrli, 1998; Liboriussen et al., 2009), previous studies also support the substantial benefits of HOx (Gemza, 1997; Prepas and Burke, 1997; Beutel et al., 2007). Furthermore, the sediment oxic zone was found to be enhanced considerably by oxygenation throughout most of the reservoir (Fig. 6), thereby promoting decreased transport of reduced chemical species to the water column. It has been shown that maintaining oxic conditions in the upper w1e2 mm of sediment can suppress the release of reduced chemical species from the sediment to the hypolimnion (Jørgensen and Boudreau, 2001; Beutel, 2003). By establishing that HOx can facilitate an oxic sediment zone on a reservoir-wide basis, the current study links sediment-water O2 dynamics to water-column studies showing that oxygenation can result in significantly improved water quality (McGinnis et al., 2004; Gantzer et al., 2009a,b).

4.

Conclusions

This research emphasizes the viability of using HOx to maintain a sediment oxic zone on a reservoir-wide scale in order to minimize reduced chemical fluxes from the sediment to the overlying water. Understanding how HOx-induced variations in near-sediment mixing and O2 concentrations affect diffusive transport at the SWI is crucial for accurately quantifying JO2 and other sediment-water fluxes, optimizing

3701

water quality, and effectively managing lakes and reservoirs. Significant conclusions include: 1. The vertical distribution of O2 on both sides of the SWI was strongly controlled by HOx operations. Oxic conditions at the SWI were maintained in both the near field and midreservoir during continuous oxygenation. The influence of the HOx on sediment O2 uptake was emphasized by the onset of anoxia in the sediment and overlying water in response to turning the HOx off for only w48 h. 2. Decreased dDBL and increased JO2 and zmax were observed during oxygenation. While HOx operation increased both near-sediment O2 concentrations and mixing, sediment O2 uptake was more strongly correlated to mixing as opposed to O2 levels in the lower hypolimnion. Regardless of O2 concentration several cm above the sediment, turbulent transport of oxygenated water to the sediment surface governed the vertical O2 distribution above and below the SWI and the corresponding JO2 . 3. A linear relationship between zmax and HOx oxygen-gas flow was established, with zmax increasing with escalating flow rate. 4. Sediment response time for an oxic vertical O2 distribution at the SWI to become re-established after initiating HOx operations was determined to be w1 week in both the near and far field. However, spatial variation in the influence of the HOx was observed as the response time was several days longer and the vertical sediment-water O2 distribution was more stable mid-reservoir than near the HOx. 5. Taking HOx-induced increases in JO2 into account when designing and operating HOx is critical for enhancing source water quality via oxygenation. By evaluating the effect of HOx on sediment O2 uptake, these results enable successful HOx operations that facilitate both elevated source water O2 concentrations and an oxic environment at the SWI. Furthermore, while this study focused on HOx-induced changes in near-sediment mixing and O2 concentrations, variation in these parameters can also be induced naturally (e.g., via fall overturn, wind-induced seiching, and hydraulic inputs during storm events). Results should therefore be more generally applicable.

Acknowledgments The authors thank Elizabeth Rumsey, Kevin Elam, and the staff at Western Virginia Water Authority who offered invaluable assistance in the field and with laboratory analyses. Alfred Wu¨est, Daniel McGinnis, Lorenzo Rovelli, John Petrie, and Peter Berg contributed via beneficial discussion and advice on data interpretation. Feedback from two anonymous reviewers greatly improved the manuscript. Financial support came from the National Science Foundation (NSF IGERT Program) and the Western Virginia Water Authority. The research described in this paper was also partially funded by the United States Environmental Protection Agency (EPA) under the Science to Achieve Results (STAR) Graduate Fellowship Program. EPA has not officially endorsed this publication and the views expressed herein may not reflect the views of the EPA.

3702

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 9 2 e3 7 0 3

references

Arega, F., Lee, J.H.W., 2005. Diffusional mass transfer at sedimentwater interface of cylindrical sediment oxygen demand chamber. Journal of Environmental Engineering e ASCE 131 (5), 755e766. Berg, P., Risgaard-Petersen, N., Rysgaard, S., 1998. Interpretation of measured concentration profiles in sediment pore water. Limnology and Oceanography 43 (7), 1500e1510. Beutel, M.W., 2003. Hypolimnetic anoxia and sediment oxygen demand in California drinking water reservoirs. Lake and Reservoir Management 19 (1), 208e221. Beutel, M.W., Horne, A.J., 1999. A review of the effects of hypolimnetic oxygenation on lake and reservoir water quality. Journal of Lake and Reservoir Management 15 (4), 285e297. Beutel, M.W., Hannoun, I., Pasek, J., Kavanagh, K.B., 2007. Evaluation of hypolimnetic oxygen demand in a large eutrophic raw water reservoir, San Vincente Reservoir, Calif. Journal of Environmental Engineering e ASCE 133 (2), 130e138. Brand, A., Dinkel, C., Wehrli, B., 2009. Influence of the diffusive boundary layer on the solute dynamics in the sediments of a seiche-driven lake: a model study. Journal of Geophysical Research 114, G01010. doi:10.1029/2008JG000755. Bryant, L.D., Lorrai, C., McGinnis, D.F., Brand, A., Wu¨est, A., Little, J.C., 2010a. Variable sediment oxygen uptake in response to dynamic forcing. Limnology and Oceanography 55 (2), 950e964. Bryant, L.D., McGinnis, D.F., Lorrai, C., Brand, A., Little, J.C., Wu¨est, A., 2010b. Evaluating oxygen fluxes using microprofiles from both sides of the sediment-water interface. Limnology and Oceanography: Methods 8, 610e627. Bryant, L. D., Hsu-Kim, H., Gantzer, P. A., Little, J. C. Unpublished results. Using source water as the solution: controlling Mn release at the sediment-water interface via hypolimnetic oxygenation, in preparation. Dalsgaard, T., Nielsen, L.P., Brotas, V., Viaroli, P., Underwood, G., Nedwell, D.B., Sundba¨ck, K., Rysgaard, S., Miles, A., Bartoli, M., Dong, L., Thornton, D.C.O., Ottosen, L.D.M., Castaldelli, G., Risgaard-Petersen, N., 2000. Protocol Handbook for NICE e Nitrogen Cycling in Estuaries: A Project under the EU Research Programme: Marine Science and Technology (MAST III). National Environmental Research Institute, Silkeborg, Denmark. http://www2.dmu.dk/LakeandEstuarineEcology/ nice/NICE_handbook.pdf. EPA (Environmental Protection Agency), 2000. Aquatic life water quality criteria for dissolved oxygen (saltwater): Cape Cod to Cape Hatteras. EPA Report EPA-822-R-00e012. http://www. epa.gov/waterscience/criteria/dissolved/dofacts.html. Ga¨chter, R., Wehrli, B., 1998. Ten years of artificial mixing and oxygenation: no effect on the internal phosphorus loading of two eutrophic lakes. Environmental Science & Technology 32, 3659e3665. Gantzer, P.A., 2002. Diffuser operations at Spring Hollow Reservoir (Master’s thesis). Available from Digital Library and Archives database http://scholar.lib.vt.edu/theses/available/ etd-05212002-163729/. Gantzer, P.A., Bryant, L.D., Little, J.C., 2009a. Controlling soluble iron and manganese in a water-supply reservoir using hypolimnetic oxygenation. Water Research 43 (5), 1285e1294. Gantzer, P.A., Bryant, L.D., Little, J.C., 2009b. Effect of hypolimnetic oxygenation on oxygen depletion rates in two water-supply reservoirs. Water Research 43 (6), 1700e1710. Gemza, A.R., 1997. Water quality improvements during hypolimnetic oxygenation in two Ontario lakes. Water Quality Research Journal of Canada 32, 365e390. Glud, R.N., 2008. Oxygen dynamics of marine sediments. Marine Biology Research 4, 243e289.

Glud, R.N., Berg, P., Fossing, H., Jørgensen, B.B., 2007. Effect of the diffusive boundary layer on benthic mineralization and O2 distribution: a theoretical model analysis. Limnology and Oceanography 52 (2), 547e557. Grant, H.L., Stewart, R.W., Moilliet, A., 1962. Turbulence spectra from a tidal channel. Journal of Fluid Mechanics 12 (2), 241e263. Gundersen, J.K., Jorgensen, B.B., 1990. Microstructure of diffusive boundary layers and the oxygen uptake of the sea floor. Nature 345, 604e607. Hondzo, M., Feyaerts, T., Donovan, R., O’Connor, B.L., 2005. Universal scaling of dissolved oxygen distribution at the sediment-water interface: a power law. Limnology and Oceanography 50 (5), 166e1676. Jørgensen, B.B., Boudreau, B.P., 2001. Diagenesis and Sedimentwater Exchange, P. 211e244. In: Boudreau, B.P., Jørgensen, B.B. (Eds.), The Benthic Boundary Layer: Transport Processes and Biogeochemistry. Oxford University Press. Jørgensen, B.B., Revsbech, N.P., 1985. Diffusive boundary layers and the oxygen uptake of sediments and detritus. Limnology and Oceanography 30 (1), 111e122. Li, Y.-H., Gregory, S., 1974. Diffusion of ions in sea water and in deep-sea sediments. Geochemica et Cosmochimica Acta 38, 703e714. Liboriussen, L., Søndergaard, M., Jeppesen, E., Thorsgaard, I., Gru¨nfeld, S., Jakobsen, T.S., Hansen, K., 2009. Effects of hypolimnetic oxygenation on water quality: results from five Danish lakes. Hydrobiologia 625 (1), 157e172. Lin, C.-H.M., Pedersen, J., Suffet, I.H., 2003. Influence of aeration on hydrophobic organic contaminant distribution and diffusive flux in estuarine sediments. Environmental Science & Technology 37 (16), 3547e3554. Lorke, A., Mu¨ller, B., Maerki, M., Wu¨est, A., 2003. Breathing sediments: the control of diffusive transport across the sediment-water interface by periodic boundary-layer turbulence. Limnology and Oceanography 48 (6), 2077e2085. Mackenthun, A.A., Stefan, H.G., 1998. Effect of flow velocity on sediment oxygen demand: experiments. Journal of Environmental Engineering e ASCE 124 (3), 222e230. McGinnis, D.F., Little, J.C., 2002. Predicting diffused-bubble oxygen transfer rate using the discrete-bubble model. Water Research 36 (18), 4627e4635. McGinnis, D.F., Lorke, A., Wu¨est, A., Sto¨ckli, A., Little, J.C., 2004. Interaction between a bubble plume and the near field in a stratified lake. Water Resources Research 40, W10206. doi:10. 1029/2004WR003038. Mobley, M.H., Ruane, R.J., Harshbarger, E.D., 2000a. And Then It Sank the Development of an Oxygen Diffuser for Hydropower. Hydrovision 2000, Charlotte, North Carolina. http://www. mobleyengineering.com/publications.html. Mobley, M.H., Hauser, G.E., McGinnis, D.F., Ruane, R.J., 2000b. Diffuser system modeling and design for dissolved oxygen enhancement of reservoirs and releases. Mobley Engineering Final Report. Tennessee Valley Authority. http://www. mobleyengineering.com/publications.html. Moore, B.C., 2003. Downflow bubble contact aeration technology (Speece Cone) for sediment oxygenation. In: Pellei, M., Porta, A. (Eds.), Proceedings of the Second International Conference on Remediation of Contaminated Sediments. Moore, B.C., Christensen, D., 2009. Newman Lake restoration: a case study. Part I. Chemical and biological response to phosphorus control. Lake and Reservoir Management 25, 337e350. Moore, B.C., Chen, P.H., Funk, W.H., Yonge, D., 1996. A model for predicting lake sediment oxygen demand following hypolimnetic aeration. Journal of the American Water Resources Association 32 (4), 723e731. NOAA (National Oceanographic and Atmospheric Administration), 2008. Local Climatological Data Report for Roanoke Regional/Woodrum Field Airport (KROA). http://

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 6 9 2 e3 7 0 3

www.ncdc.noaa.gov/oa/ncdc.html June e May 2008. (accessed 10.10.09.). O’Connor, B.L., Hondzo, M., Harvey, J.W., 2009. Incorporating both physical and kinetic limitations in quantifying dissolved oxygen flux to aquatic sediments. Journal of Environmental Engineering e ASCE 135 (12), 1304e1314. Prepas, E.E., Burke, J.M., 1997. Effects of hypolimnetic oxygenation on water quality in Amisk Lake, Alberta, a deep eutrophic lake with high internal phosphorus loading rates. Canadian Journal of Fisheries and Aquatic Sciences 54, 2111e2120. Rasmussen, H., Jørgensen, B.B., 1992. Microelectrode studies of seasonal oxygen uptake in a coastal sediment: role of molecular diffusion. Marine Ecology Progress Series 81, 289e303. Schaller, T., Wehrli, B., 1997. Geochemical-focusing of manganese in lake sediments e an indicator of deep-water oxygen conditions. Aquatic Geochemistry 2 (4), 359e378. Schlichting, H. (Ed.), 2000. Boundary-layer Theory, ninth ed. McGraw-Hill. Singleton, V.L., Little, J.C., 2006. Designing hypolimnetic aeration and oxygentation systems: a review. Environmental Science and Technology 40 (24), 7512e7520. Singleton, V.L., Gantzer, P.A., Little, J.C., 2007. Linear bubble plume model for hypolimnetic oxygenation: full-scale

3703

validation and sensitivity analysis. Water Research 43, W02405. doi:10.1029/2005WR004836. Singleton, V.L., Rueda, F.J., Little, J.C., 2010. A coupled bubble plume reservoir model for hypolimnetic oxygenation. Water Resources Research 46, W12538. doi:10.1029/ 2009WR009012. Veenstra, J.N., Nolen, S.L., 1991. In situ sediment oxygen demand in five southwestern U.S. lakes. Water Resources 25 (3), 351e354. Wu¨est, A., Brooks, N.H., Imboden, D.M., 1992. Bubble plume modeling for lake restoration. Water Resources Research 28 (12), 3235e3250. Wu¨est, A., Lorke, A., 2003. Small-scale hydrodynamics in lakes. Annual Review of Fluid Mechanics 35, 373e412. Wu, R.S.S., Sheng Zhou, B., Randall, D.J., Woo, N.Y.S., Lam, P.K.S., 2003. Aquatic hypoxia is an endocrine disruptor and impairs fish reproduction. Environmental Science and Technology 37 (6), 1137e1141. Zaw, M., Chiswell, B., 1999. Iron and manganese dynamics in lake water. Water Research 33 (8), 1900e1910. Zhang, H., Davison, W., Ottley, C., 1999. Remobilisation of major ions in freshly deposited lacustrine sediment at overturn. Aquatic Sciences 61, 354e361.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 0 4 e3 7 1 1

Available at www.sciencedirect.com

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

A study of titanium sulfate flocculation for water treatment Yi-Fan Wu*, Wen Liu, Nai-Yun Gao, Tao Tao College of Environmental Science and Engineering, Tongji University, Shanghai 200092, China

article info

abstract

Article history:

There are limited studies available on titanium salt flocculation. In this research, coagu-

Received 21 December 2010

lation experiments of titanium sulfate were conducted using both distilled water and

Received in revised form

kaolin clay suspension. Results showed that titanium sulfate flocculation was most

12 April 2011

effective in the pH range 4e6, and negligible concentrations of titanium were found in the

Accepted 15 April 2011

well-flocculated water. The floc isoelectric point (IEP) was found to be near pH 5.

Available online 22 April 2011

Measurements showed that the titanium flocs possessed greater density, diameter and settling velocity than the aluminum flocs. The titanium flocs were composed of TiO(OH)2,

Keywords:

which would change from the amorphous phase into anatase titanium dioxide under

Water treatment

elevated temperatures. Floc images showed the structural similarity of titanium and

Coagulation

aluminum flocs. Laboratory results and a pilot experiment showed that titanium sulfate

Titanium sulfate

could be an alternative coagulant for water and wastewater treatment. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Although the use of aluminum salts for water treatment dates back centuries, there have been disputes over the past two decades on the possible adverse effects of aluminum on human health and the environment (WHO, 1997). People are seeking new materials for water treatment to meet increasingly stringent guidelines which require efficient removal of pollutants from water. Titanium salts used as a chemical for water treatment was first proposed by Upton and Buswell (1937). They reported that coagulation could be effected by dosing titanium sulfate, which flocculated easily at low temperatures and was more efficient for color removal, as compared with alum. For a long time however, probably because of the price of titanium salts, this potential coagulant has been overlooked. Recently, people seem to have a renewed interest in titanium salt for water treatment. Shon et al. (2007, 2009) experimented with titanium tetrachloride (TiCl4) to flocculate a synthetic sewage and they proposed to recover the widely used titanium dioxide (TiO2) from the settled flocs. Okour et al. (2009a,b) also used

TiCl4 and titanium sulfate to treat a synthetic sewage. They reported greater reductions in turbidity, UV254 absorbance and DOC, as compared with the treatment using ferric chloride and aluminum sulfate (Al2(SO4)3). Zhao et al. (2011a) used reservoir water to compare the coagulation effect of TiCl4 with that of poly-aluminum chloride (PAC). They also used synthetic water in other experiments (2011b) to compare the treatment results of TiCl4 with aluminum and iron salts. All these experiments confirmed the flocculation ability of titanium salts. However, the effective pH environment of titanium flocculation was not clearly defined. Titanium exists as one of the most abundant elements on the earth. With the development of titanium industry, titanium salts are gradually becoming comparable in price to the conventional coagulants. Titanium and its compounds are reported to have so little toxicity (WHO, 1982; Lee et al., 2008) that they are rarely included in any water quality guidelines. Also, recovering TiO2 from water treatment sludge could be a good method to solve sludge disposal problems arising from large-scale water treatment. Since experimental data for titanium salt coagulation was still fairly limited and not

* Corresponding author. Tel./fax: þ86 021 65984275. E-mail address: [email protected] (Y.-F. Wu). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.023

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 0 4 e3 7 1 1

readily available, investigation in this field was needed to evaluate the applicability of titanium salts in water treatment. In this study, titanium (IV) sulfate (Ti(SO4)2) was chosen for the coagulation experiments. A jar-test apparatus was used to find the region where effective flocculation occurred. Residual titanium concentrations in the flocculated water were measured. Floc images were provided by scanning electron microscopy (SEM) and transmission electron microscopy (TEM). Floc size, density, settling velocity and isoelectric point (IEP) were investigated. Titanium flocs were also characterized by thermogravimetry (TG), differential scanning calorimetry (DSC), X-ray fluorescence spectroscopy (XRF) and X-ray diffraction analysis (XRD). A preliminary pilot experiment was performed to test the applicability of titanium salt coagulation in potable-water treatment.

2.

Material and methods

2.1.

Coagulation experiments

The coagulation experiments were carried out on an apparatus (Model DC-506, Shanghai Waterworks Company) consisting of six jars each of 1 L volume. Flocculation was conducted in the following steps: rapid mixing for 45 seconds (s), quick stirring for 10 minutes (min), moderate stirring for 10 min, and slow stirring for 10 min. The jar paddle rotations were preset according to water temperature, to provide a step-down velocity gradient (G) of 800 s1, 70 s1, 45 s1, and 20 s1 respectively for the flocculation steps. The flocculated samples were allowed to settle for 30 min before any measurements were taken. This flocculation process was used for all the coagulation experiments under an environment of 10e17  C. Predetermined doses of 1.05 mol/L sodium hydroxide (NaOH) were usually added together with titanium sulfate to provide the pH environment required for flocculation. Distilled water was used to define the region where visible hydrolysis and flocculation happened. Some similar flocculation experiments were also conducted with aluminum sulfate (Al2(SO4)3) for comparison. Turbidity removal experiments were conducted with kaolin clay suspension. Some kaolin clay was mixed into a large volume of water and vigorously agitated. The mixture was then allowed to settle for 6 h and the supernatant was drawn as the stock kaolin clay suspension for the coagulation experiments. The stock suspension was then diluted into two concentration levels, of 138e179 NTU and 17e34 NTU respectively, to simulate moderate and low turbidities. Sodium bicarbonate (NaHCO3) was dosed into the suspension to simulate water with alkalinities ranging from 0 to 120 mg CaCO3/L. If needed, some doses of 0.5 mol/L sulfuric acid (H2SO4) were added together with the coagulant to neutralize the alkalinities, in order to facilitate coagulation.

2.2.

Supernatant measurements

PH and temperature of the samples were measured at the end of the flocculation process, using a pH 3310 portable pH meter (WTW). With a 2100p turbidimeter (HACH), turbidities of the supernatant in the flocculation jars were measured and

3705

recorded as residual turbidities (RT). Titanium ion concentrations in the supernatant were usually determined by the salicyl-fluorone spectrophotometric method (Ministry of Health of the People’s Republic of China, 2007) with a UV spectrophotometer (Model UV765, Shanghai Precision & Scientific Instrument Co.), and concentrations lower than 0.02 mg Ti/L were measured by inductively coupled plasma mass spectrometry (ICP-MS), using an ELAN DRC-e spectrometer (PerkineElmer).

2.3.

Floc property measurements

SEM images of flocs were provided by a Quanta 200F (FEI) electron microscope, and TEM images were obtained with a Tecnai G2 F20 (FEI) electron microscope. Floc size distributions were determined using an Eyetech particle analyzer (Ankersmid). Floc zeta potentials were measured with a Zetasizer Nano Z analyzer (Malvern). Representative settling velocities of flocs were measured in a transparent cylinder by the video recording method. Floc densities were estimated by observing the buoyancy of floc particles released in a series of salt solutions with different densities. Settled flocs were collected and vacuum-filtered onto 0.45 mm pore-size poly-vinylidene fluoride (PVDF) membranes. The filtered flocs were stored under room temperature in a desiccator for one week before further analyses. The TG and DSC analyses were conducted on a Q600 SDT analyzer (TA) with an air flow of 100 mL/min and a temperature ramp of 5  C/ min. Floc composition was analyzed by XRF method using an XRF-1800 spectrometer (Shimadzu). XRD analyses were performed on a D8 Advance (Bruker-AXS) diffractometer.

3.

Results and discussions

3.1. Flocculation region and residual titanium concentration 3.1.1.

Flocculation region

Samples of flocculation experiments were classified into four categories: (a) clear liquid samples with no perceivable hydrolysis; (b) turbid samples with obvious hydrolysis but no visible flocs; (c) supernatant samples with RT > 5 NTU because of unsatisfactory flocculation; and (d) supernatant samples with RT < 5 NTU resulted from good flocculation. The data of good flocculation (category d) defined the flocculation region of Ti(SO4)2 in Fig. 1. The flocculation region of aluminum salts summarized by Amirtharajah and Mills (1982) was also drawn (shaded area) for comparison, using arithmetic scales for the vertical concentration axis. Fig. 1 illustrated that at almost all pH values titanium sulfate hydrolyzed readily when dosages exceeded 0.17 mmol/L, and pH 4e6 was the most effective flocculation range. Good flocculation could even be effected when dosages exceeded 0.03e0.04 mmol/L at about pH 5. The data showed clearly that Ti(SO4)2 flocculated in a narrower and more acidic pH environment than aluminum hydroxides. The flocculation region of titanium sulfate coincided with the 5.3e5.7 pH range where enhanced coagulation (Amirtharajah and O’Melia, 1999; USEPA, 1999) was usually practiced to

3706

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 0 4 e3 7 1 1

Fig. 1 e Flocculation zone of titanium and aluminum hydroxides.

remove natural organic materials (NOM). It was suggested that Ti(SO4)2 might be effective for NOM removal. For kaolin clay coagulation experiments, RT records showed that the minimum Ti(SO4)2 dose required to initiate efficient flocculation increased with the alkalinity, from 0.06 mmol/L for 15 mg CaCO3/L, to about 0.20 mmol/L for 45 mg CaCO3/L alkalinities. Therefore it was concluded that practical flocculation happened only in this alkalinity range. The effect was partly noticed by Mamchenko et al., who declared that titanium salts ‘are expedient to be used for treating water with low alkalinity’ (Mamchenko et al., 2009). Titanium sulfate showed poor flocculation ability with higher alkalinities (>50 mg CaCO3/L), and this was attributed to the strong buffering effect of bicarbonate, which combined with most of the hydrogen ions released from titanium salt hydrolysis. Water pH was thus inhibited from dropping into the acidic region for titanium salt flocculation. With zero or very low alkalinities (0e5 mg CaCO3/L), the hydrogen ions derived from Ti(SO4)2 hydrolysis were so excessive that pH would be depressed to values below 4, also leading to poor flocculation. Further coagulation experiments were conducted for both distilled water and kaolin clay suspension with alkalinities of 80, 100, and 120 mg CaCO3/L. Ti(SO4)2 was added together with some sulfuric acid (H2SO4) to deplete the bicarbonates. The results (Fig. 2) showed that Ti(SO4)2 coagulation was quite dependent on pH, and H2SO4 could be added under high alkalinity conditions (80e120 mg CaCO3/L, experiments so far) to perform excellent flocculation. The flocculation data also showed that when an appropriate pH environment (5.0e5.5) was provided, RT values lower than 3 NTU could be easily obtained with Ti(SO4)2 dosed at about 0.07e0.14 mmol/L. In the experiments of alkalinity and kaolin clay suspension with both concentration levels of 138e179 NTU and 17e34 NTU, results of good flocculation (i.e. RT < 5) were in good accordance with the titanium flocculation region shown in

Fig. 1. Thus the region could be assumed as the flocculation zone suitable for potable-water treatment. Since flocculation performance was strongly dependent on water quality (Shin et al., 2008), the flocculation region of titanium salts might be different for waters with complex impurities.

3.1.2.

Residual titanium concentration

Titanium concentration in the flocculated supernatant was very low. For flocculation experiments with distilled water, titanium concentrations in the supernatant drawn from the well-settled samples (RT 0.11e1.1 NTU, pH 4.3e5.6) ranged

Fig. 2 e Residual turbidity vs pH (Ti(SO4)2 dose 0.07e0.14 mmol/L).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 0 4 e3 7 1 1

between 0.15 and 0.5 mg Ti/L. However, by filtering the supernatant through 0.22 mm pore cellulose acetate (CA) membranes, only trace concentrations lower than 1 mg Ti/L were found in the filtrates. Therefore it was suggested that during coagulation most titanium ions would hydrolyze to produce settleable precipitates and filterable colloids, and could be conveniently removed by conventional water treatment processes like sedimentation and filtration. During the pilot experiment (see Section 3.3) with surface water dosing 30 mg Ti(SO4)2/L (0.125 mmol/L), the final filtered effluent was found to have a concentration around 4 mg Ti/L. The experiments indicated that the titanium concentration in the treated water would easily satisfy any health guidelines. It should be noted that although there is no evidence to indicate the relationship between titanium compounds and human health risks, WHO does estimate a maximum daily intake of 500 mg titanium per individual and a concentration of 15 mg Ti/ L in drinking water (WHO, 1982).

3.2.

Floc properties

3.2.1.

Floc size

The flocs of titanium sulfate (Ti-floc) were generally bigger than the flocs derived from aluminum sulfate (Al-flocs). For the flocculation experiments with distilled water, the mean diameter of Ti-flocs produced at pH 4.4e6 averaged 99 mm (78e113 mm), about three times greater than that of Al-flocs, which were produced at pH 6.0e7.0 and their diameters only averaged 29 mm (21e35 mm). The kaolin clay coagulation with Ti(SO4)2 produced an average floc size of 118 mm (113e127 mm) at pH 5.2e6.2, much greater than the averaged 42 mm (40e44 mm) of kaolin clay-aluminum flocs produced at pH 6.1e6.2. Similar observations were reported by Shon et al. (2009), who measured mean sizes of 47.5 mm for Ti-flocs and 16.9 mm for Al-flocs respectively, when experimenting with a synthetic sewage. Also, Zhao et al. (2011b) reported a Ti-floc size of 800.9 mm and an Al-floc diameter of 404.8 mm in their coagulation experiment with synthetic reservoir water. It was thus suggested that under similar flocculation conditions, titanium flocs might be more resistant to hydraulic shearing forces than aluminum flocs, and would result in efficient flocculation or filtration under higher hydraulic loads in water treatment. Also noticed was that flocs derived from coagulation with kaolin clay were generally bigger, therefore tougher, than flocs produced in the experiments with distilled water. The experimental data did not yet support any meaningful relationship between pH and floc size.

3.2.2.

Fig. 3 e Floc density vs pH.

velocity variations. Also noticed was that within the pH range of 5e6 where enhanced coagulation was usually practiced, the settling velocity of Al-flocs would slow down whereas Ti-flocs would keep a higher settling speed, so efficient sedimentation could be expected in this pH range by titanium flocculation. As shown in Fig. 5, the average settling velocities of both titanium and aluminum flocs in combination with kaolin clay followed the similar variation patterns as their ‘pure’ flocs derived from distilled water coagulation. The kaolin claytitanium flocs reached a mean velocity of 4.0 mm/s at about pH 5.5 whereas at pH 5.1 and pH 5.9, the floc velocities fell to around 3.8 and 2.6 mm/s respectively. The average settling velocities of kaolin clay-aluminum flocs within pH 5e6 were about 2.5e2.8 mm/s, and rose to 3.2e3.5 mm/s when pH reached 7e8. It was noticed that all the kaolin clay combined flocs settled faster than the distilled-water-flocculates, indicating that floc settling speed strongly depended on water impurities.

3.2.3.

Isoelectric point

Zeta potential values of Ti-flocs were measured and plotted against pH, as shown in Fig. 6. The IEP of Ti-flocs was found to be near pH 5, lower than the IEP values of Al(OH)3 flocs

Floc density and settling velocity

Measurements showed that the density of Ti-flocs averaged 1.04 g/cm3 (1.038e1.046 g/cm3), slightly greater than the density of Al-flocs, which averaged 1.02 g/cm3 (1.015e1.031 g/ cm3). As shown in Fig. 3, within each flocculation region, both Ti-floc and Al-floc densities were comparatively constant with respect to pH variations. The average settling velocity of Ti-flocs (ranged 1.01e3.69 mm/s) was a function of pH, as shown in Fig. 4. Considering that floc densities were fairly constant with pH variations (Fig. 3), it was assumed that pH might have some unknown effect on floc agglomeration that led to settling

3707

Fig. 4 e Ti(SO4)2 flocculation RT, and average settling velocities of flocs (coagulation with distilled water).

3708

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 0 4 e3 7 1 1

Fig. 5 e Ti(SO4)2 flocculation RT, and average settling velocities of flocs (coagulation with kaolin clay suspension).

reported as pH 7e9 (Kanna et al., 2005). The measured IEP of Ti-flocs was generally in agreement with the titanium flocculation region depicted in Fig. 1, where good flocculation happened when floc zeta potential approached zero. The IEP value was also partly confirmed by the reservoir water experiment of Zhao (2011a). Their data revealed that when TiCl4 coagulant was near ‘optimal’, floc zeta potential would approach zero, corresponding to a pH value lower than 5.51. As zeta potential was a function of many variables such as water composition and coagulant dose, further research was recommended on defining the IEP range of Ti-flocs.

Fig. 6 e Zeta potential of Ti(SO4)2 flocs vs pH.

3.2.4.

Floc images

SEM and TEM images of aluminum and titanium flocs were provided for comparison. Titanium flocs in appearance were like a kind of white chemical precipitate, which showed stronger light scattering than the aluminum flocs. The SEM

Fig. 7 e SEM images of Ti(SO4)2 and Al2(SO4)3 floc.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 0 4 e3 7 1 1

3709

Fig. 8 e TEM images of Ti(SO4)2 and Al2(SO4)3 floc.

visuals (Fig. 7) showed that titanium flocs were also of a bulky and flocculent nature like aluminum flocs, and the TEM visuals (Fig. 8) showed that the pore in the microstructure of Ti-flocs roughly sized 30e80 nm, similar to that of the Al-flocs. It was thus suggested that titanium flocs might also have adsorptive and enmeshing abilities like aluminum flocculates. Both titanium and aluminum flocs were amorphous since no evidence of electron diffractions were found in the TEM observations.

3.2.5.

Floc characterization

TG and DSC analyses were conducted with Ti-flocs derived from distilled water experiment. Typical sample analyses (Fig. 9) showed that the weight-loss of the flocs from room temperature to 240  C was about 16e30% and there were no apparent weight-loss plateaus within this temperature range. This major weight-loss was attributed to the evaporation of water adsorbed and incorporated in the hydroxides, and it also explained the endothermic peak appeared on the DSC curve. Another small exothermic peak occurred on the DSC

Fig. 9 e Typical TG and DSC curves of Ti(SO4)2 flocs.

curve within the range of 380e470  C was interpreted as the crystallization of the amorphous titanium dioxide under elevated temperatures. Similar findings were reported by Kanna et al. (2005), though they prepared flocs with TiCl4 and diluted ammonia solutions instead. Results of XRD analysis (Fig. 10) further confirmed the amorphous nature of the titanium flocs formed at room temperature (curve a), and clearly showed that under elevated temperatures, titanium hydroxides in the flocs would be gradually transformed into anatase phase (curve b to f). The 700  C heated flocs had an XRD curve (curve f) of pure anatase TiO2. This curve was quite similar to curve g derived from commercial Degussa P-25 TiO2, an efficient photo-catalyst widely used. Similar results were also obtained by researchers (Lee et al., 2008; Okour et al., 2009a,b) experimenting with TiCl4 coagulation for seawater and sewage. The heat transformation experiments not only suggested a method to recover some photo-active catalyst from the sludge derived from titanium salt coagulation, as proposed by Shon et al. (2007, 2009), but also showed a possible process to produce commercial anatase TiO2.

Fig. 10 e XRD curves of Ti(SO4)2 flocs heated under different temperatures, and XRD curve of Degussa P-25 TiO2.

3710

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 0 4 e3 7 1 1

XRF analysis of elements showed that the weight fraction of titanium in Ti-flocs was nearly equal to the fraction of oxygen, and the Ti/O weight ratio was about 1: 0.88e1.08. Therefore it was suggested that flocs originated from titanium ion hydrolysis were mainly composed of TiO(OH)2, which would be transformed into TiO2 when calcined. Apparently floc composition was a function of the impurities in the water being flocculated. For example, Shon et al. had proposed a formula of TiO1.42C0.44P0.14 for the incinerated flocs derived from TiCl4 treated sewage (Shon et al., 2007).

3.3.

Pilot experiment

In order to verify the practicability of titanium salt coagulation, a field experiment was conducted on a pilot potablewater treatment system with a capacity of 160 m3/d. Water flowed sequentially through a baffled flocculator, a tube settler and a sand filter in the system. The flocculator was a zigzag baffled type with a retention time of 20 min and G values ranged 25e80 s1; the tube settler was an up-flow type with a surface loading of 3.5 mm/s; and the filter, which operated under a filtration rate of 12 m/h, consisted of a 1.2 m thick bed of filter sand with grain diameters ranged 0.9e1.2 mm. The pilot plant drew raw water from the Yangtze River, and during the experiment period the influent had turbidities of 40e49 NTU, alkalinities 80e89 mg/L as CaCO3, pH values 7.0e7.8, UV254 absorbance values 0.072e0.074 cm1, and temperatures around 23  C. Titanium sulfate was dosed with sulfuric acid in a concentration of 1.2e2.2% on weight basis. The field application of titanium salts showed effective coagulation results. When the doses were kept as 16e32 mg Ti(SO4)2/L (0.07e0.13 mmol/L) with pH adjusted to 5.4e5.6, residual turbidities of 3.2e5.6 NTU were achieved from the tube settler outlet and the final produced water had turbidities of 0.1e0.2 NTU. The pilot operation also showed an overall UV254 removal of 54e57%, better than the result of about 41% achieved by the comparison runs, in which PAC doses of about 22 mg/L were added for coagulation.

4.

Conclusions

The research studied some aspects of titanium sulfate flocculation and the following implications were drawn from the work: – Titanium sulfate could induce effective flocculation in a slightly acidic environment with doses comparable to that of the conventionally used coagulants, and acids or bases could be added for different waters to provide the appropriate coagulation pH. Efficient NOM removal could be expected by using titanium salt coagulation. The titanium contained in the treated water would cause no health risks. – The settling and filtration properties of titanium flocs were better than that of the aluminum flocs produced in the same hydraulic environment, indicating more efficient performance of treatment facilities with titanium salt coagulation. – It is possible to recover the widely used titanium dioxide from the water treatment sludge derived from titanium salt

coagulation, and solve the problems arising from sludge disposal. To summarize, titanium sulfate seemed quite suitable for enhanced coagulation, for potable-water treatment of a low alkalinity source, or for treatment of certain wastewaters. The application of titanium salt as an alternative coagulant might be promising in the field of water supply and wastewater treatment.

Acknowledgments This work was partly supported by the research fund of Science and Technology Commission of Shanghai Municipality (project# 08230707000). The writers are grateful to the staff in The State Key Laboratory of Pollution Control and Resources Reuse (of Tongji University), for their invaluable instructions in performing the analyses. Special thanks are addressed to Dr. N. Lockington for his cordial help in paper organization. The kind suggestions from the editors and reviewers are deeply appreciated.

references

Amirtharajah, A., Mills, K.M., 1982. Rapid-mix design for mechanisms of alum coagulation. Journal of American Water Works Association 74 (4), 210e216. Amirtharajah, A., O’Melia, C.R., 1999. Water Quality and Treatment, fifth ed. American Water Works Association, Denver, CO. 6.5e6.6. Kanna, M., Wongnawa, S., Sherdshoopongse, P., Boonsin, P., 2005. Adsorption behavior of some metal ions on hydrated amorphous titanium dioxide surface. Songklanakarin Journal of Science and Technology 27 (5), 1017e1026. Lee, B.C., Kim, S., Shon, H.K., Vigneswaran, S., Kim, S.D., Cho, J., Kim, I.S., Choi, K.H., Kim, J.B., Park, H.J., Kim, J.H., 2008. Aquatic toxicity evaluation of TiO2 nanoparticle produced from sludge of TiCl4 flocculation of wastewater and seawater. Journal of Nanoparticle Research 11 (8), 2087e2096. Mamchenko, A.V., Gerasimenko, N.G., Deshko, I.I., Pakhar, T.A., 2009. The investigation of the efficiency of coagulants based on titanium when purifying water. Journal of Water Chemistry and Technology 32 (3), 167e175. Ministry of Health of the People’s Republic of China, 2007. Standard Examination Methods for Drinking WatereMetal Parameters (GB/T 5750.6-2006). Standards Press of China, Beijing. 67-70. Okour, Y., Saliby, I.E., Shon, H.K., Vigneswaran, S., Kim, J.-H., Cho, J., Kim, S., 2009a. Recovery of sludge produced from Tisalt flocculation as pretreatment to seawater reverse osmosis. Desalination 249, 53e63. Okour, Y., Shon, H.K., Saliby, I.E., 2009b. Characterization of titanium tetrachloride and titanium sulfate flocculation in wastewater treatment. Water Science and Technology 59 (12), 2463e2473. Shin, J.Y., Spinette, R.F., O’Melia, C.R., 2008. Stoichiometry of coagulation Revisited. Environment Science and Technology 42 (7), 2582e2589. Shon, H.K., Okour, Y., Saliby, I.E., Park, J., Cho, D., Kim, J.B., Park, H.J., Kim, J., 2009. Preparation and characterization of titanium dioxide produced from Ti-salt flocculated sludge in

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 0 4 e3 7 1 1

water treatment. Journal of the Korean Industrial and Engineering Chemistry 20 (3), 241e250. Shon, H.K., Vigneswaran, S., Kim, I.S., Cho, J., Kim, G.J., Kim, J.B., Kim, J.H., 2007. Preparation of titanium dioxide (TiO2) from sludge produced by titanium tetrachloride (TiCl4) flocculation of wastewater. Environment Science and Technology 41 (4), 1372e1377. Upton, W.V., Buswell, A.M., 1937. Titanium salts in water purification. Industrial and Engineering Chemistry 29 (8), 870e871. USEPA, 1999. Enhanced Coagulation and Enhanced Precipitative Softening Guidance Manual, 2e5. http://www.epa.gov/ safewater/mdbp/coaguide.pdf [accessed 2010].

3711

WHO, 1982. Environmental Health Criteria 24. http://inchem.org/ documents/ehc/ehc/ehc24.htm [accessed 2010]. WHO, 1997. ). Environmental Health Criteria 194. http://www. inchem.org/documents/ehc/ehc/ehc194.htm [accessed 2010]. Zhao, Y.X., Gao, B.Y., Cao, B.C., Yang, Z.L., Yue, Q.Y., Shon, H.K., Kim, J.-H., 2011a. Comparison of coagulation Behavior and floc characteristics of titanium tetrachloride (TiCl4) and polyaluminum chloride (PACl) with surface water treatment. Chemical Engineering Journal 166 (2), 544e550. Zhao, Y.X., Gao, B.Y., Shon, H.K., Cao, B.C., Kim, J.-H., 2011b. Coagulation characteristics of titanium (Ti) salt coagulant compared with aluminum (Al) and iron (Fe) salts. Journal of Hazardous Materials 185 (2e3), 1536e1542.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 1 2 e3 7 2 2

Available at www.sciencedirect.com

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

Designing water supplies: Optimizing drinking water composition for maximum economic benefit M. Rygaard a,*, E. Arvin a, A. Bath b, P.J. Binning a a

DTU Environment, Department of Environmental Engineering, Technical University of Denmark, Miljoevej, Building 113, DK-2800 Kgs. Lyngby, Denmark b Water Corporation, 629 Newcastle Street, Leederville 6007, WA, Australia

article info

abstract

Article history:

It is possible to optimize drinking water composition based on a valuation of the impacts of

Received 6 September 2010

changed water quality. This paper introduces a method for assessing the potential for

Received in revised form

designing an optimum drinking water composition by the use of membrane desalination

11 March 2011

and remineralization. The method includes modeling of possible water quality blends and

Accepted 18 April 2011

an evaluation of corrosion indices. Based on concentration-response relationships a range

Available online 22 April 2011

of impacts on public health, material lifetimes and consumption of soap have been valued for Perth, Western Australia and Copenhagen, Denmark. In addition to water quality

Keywords:

aspects, costs of water production, fresh water abstraction and CO2-emissions are inte-

Corrosion

grated into a holistic economic assessment of the optimum share of desalinated water in

Cost

water supplies. Results show that carefully designed desalination post-treatment can have

Desalination

net benefits up to V0.3
0.2 per delivered m3 for Perth and V0.4(
0.2) for Copenhagen. Costs

Health

of remineralization and green house gas emission mitigation are minor when compared to

Remineralization

the potential benefits of an optimum water composition. Finally, a set of optimum water

Water quality

quality criteria is proposed for the guidance of water supply planning and management. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

In many regions of the world water supply managers are diversifying the resources used for drinking water production. Introduction of new water resources in urban water supply systems often leads to a changed water quality. Advanced water treatment processes such as desalination and water reclamation using membranes and the use of remineralization techniques gives many options for engineering water quality. These techniques and options of blending different water resources make it possible to optimize drinking water quality with many potential benefits to consumers. The method presented in the following can supplement integrated assessments in a decision making process with the target of determining the optimum

water blend based on water production and water quality considerations. Optimization of water quality is often associated with meeting quality criteria (Birnhack and Lahav, 2007) or reducing corrosion potential of the supplied water (Characklis, 2004; Imran et al., 2006). Birnhack and Lahav (2007) proposed a new desalination post-treatment process to improve water quality parameters considering chemical/bio stability, health, economy, taste, and wastewater treatment to meet new Israeli water quality requirements. The main parameters evaluated were alkalinity, Mg2þ, Ca2þ, calcium carbonate saturation, pH, total dissolved solids (TDS), and the Larson Ratio. They recommended a post-treatment process where magnesium partly replaces calcium in the desalinated water to reduce calcium

* Corresponding author. E-mail addresses: [email protected] (M. Rygaard), [email protected] (E. Arvin), [email protected] (A. Bath), [email protected] (P.J. Binning). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.025

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 1 2 e3 7 2 2

content and maintain a minimum level of magnesium. Their process addressed several challenges in the post-treatment of desalinated water by dissolution of calcite, and made it possible to increase alkalinity above 80 mg/l as CaCO3 while keeping Ca levels below 120 mg/l as CaCO3. Imran et al. (2006) simulated how mixing of groundwater, surface water and reverse osmosis desalinated seawater types could lead to problematic blends that increased corrosion rates. Other studies have used lifetime models of materials in contact with water to find optimal percentages of desalination in water supplies with the objective of maximizing net benefits from improved water quality. It has been shown that the benefits of reduced salinities in water supplies can exceed the extra cost of desalinating water at salinities down to 1000 mg/l (Characklis, 2004). Rygaard et al. (2009) considered a broader view of the impact of water quality, including the issues health impacts, bottled water consumption, soap consumption, and mitigation of CO2-emissions for a fixed blending ratio of groundwater and desalinated water. That paper introduced the successive principle to evaluate the large economic uncertainties involved in this type of assessment. These studies show the potential for improving water quality beyond official quality guidelines and requirements. In the following this potential is explored further. This paper aims to provide a new approach for the optimization of drinking water quality using technological possibilities for engineering water, i.e. using membrane and remineralization technologies for water treatment before blending various water sources. The paper updates the method introduced in Rygaard et al. (2009) to include desalination treatment costs and costs of alternatives to desalination. The relative importance of water quality effects is assessed by concurrent calculations of water production costs and estimated costs of fresh water abstraction and green house gas emissions. While Rygaard et al. (2009) considered only fixed blending ratios, this paper assesses the impacts of varying blending ratios and multiple water qualities. Novel insights are provided on the relative economic importance of external effects when compared to production costs. Further, the method can be applied to assess the optimum share of a new water quality (i.e. desalinated water) in urban water supplies. The paper also includes new considerations of water quality. For a better understanding of the corrosion potentials an evaluation of corrosion indices is included. Thus, the impact assessment is based on corrosion indices, water quality guidelines and concentration-response relationships derived from the literature. It is shown that changing blending ratios and varying individual water quality parameters lead to quite different economic effects on society. The results will contribute to the on-going discussion on how we define good drinking water and attempt to answer the question: What is good water quality? An attempt to answer the question is provided here by defining a set of general water quality parameters, that specifies optimum concentration ranges for selected minerals in drinking water. Although the focus is on the use of desalinated water in two specific cases, the results are relevant for all water supply systems that incorporate removal of dissolved solids and remineralization processes. Such systems include softening processes, potable water reclamation and desalination

3713

using nano-filtration or reverse osmosis membranes, and fluoridation.

2.

The cases

The method is applied to two water supply systems: A case from Perth, Western Australia, where desalinated water augments surface water and groundwater since 2006 and a case from Copenhagen, Denmark, based on possible scenarios for a future augmentation with desalinated water. The case from Denmark is provided as Supplementary information (SI1). The two cases illustrate how the method can be applied before and after implementation of changes to water supplies.

2.1.

Case Perth

Perth relies on three main sources for water: groundwater (50e60%), surface water (20e35%) and desalinated water (17%). In reservoirs and the metropolitan distribution system operated by Water Corporation all three sources are mixed in varying ratios. The current population of 1.7 million is expected to increase to 2.3 million by 2030. For the same period climate change is predicted to reduce groundwater and surface water yields by approximately 17% (Water Corporation, 2009). The future demand will be met by increased use of desalinated water and the development of new sources like wastewater reclamation through artificial groundwater recharge. The new sources will add to the complexity of the water supply system and the possible effects on water quality of these initiatives need to be assessed. The analysis is limited to two adjacent subsections of Perth’s water supply named Hamilton Hill and Thompsons Reservoir, which have a combined population of approximately 99,750 persons living in 41,676 households with the majority being detached houses. These areas consume 13.5 million m3 per year (V in equation (2)e(5)) consisting of a blend of all three water sources. The area is typical of the water supply of the Perth Metropolitan Area.

3.

Method and base data

Our method includes two steps. The first step is to predict possible water qualities when mixing the water sources. For each case various possible blends are simulated using water quality modeling software. In the second step concentrationresponse relationships are used to predict the physical and economic impact of water blends. All impacts are calculated as a change from a base scenario. The base scenario is defined as the current average water quality in Hamilton Hill/ Thompsons Reservoir areas.

3.1.

Modeling water blends

Water blends were modeled using PHREEQC with its standard database (Parkhurst and Appelo, 2008). Initial surface and groundwater qualities are based on average values as

3714

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 1 2 e3 7 2 2

measured by the water utility in Perth. The Perth Desalination Plant provided typical water quality data for its product water.

3.2.

Post-treatment and remineralization

Various post-treatment options for remineralization of desalinated water appear in the literature (Withers, 2005). In practice the most common processes are to add alkalinity and calcium by acidification with CO2 combined with dissolution of hydrated lime Ca(OH)2 or limestone CaCO3. In Perth lime stabilization is currently used to achieve a typical alkalinity level of 81 mg/l as CaCO3 and Ca at 65 mg/l (162 mg/l as CaCO3). The Perth Desalination Plant only adds very small amounts of magnesium to the product water where it is a minor component of the lime used for stabilization. To explore the effects of increasing alkalinity and magnesium content in Perth, the analysis includes an evaluation of Perth desalinated water post-treated to a higher alkalinity and remineralized with magnesium. The magnesium is added using an ion-exchange process, described elsewhere (Birnhack and Lahav, 2007). The process takes advantage of the high magnesium content in the pretreated feed water for the reverse osmosis process and provides a low cost alternative to other options like dissolution of dolomite. The impact from low fluoride levels on dental health (see section 3.4) has lead many water supply systems to fluoridate the drinking water. In Perth fluorosilisic acid is used to maintain a stable fluoride level of 0.8e0.9 mg/l as mandated by the state government.

3.3.

Uncertainty assessment

The uncertainty calculations for this paper are based on the successive principle (Lichtenberg, 2000) as adapted by Rygaard et al. (2009). In short, the principle operates with a triple estimate that allows for subjective expert judgment of the economic impact. A “most likely” estimate M, and expected 1% (F.01) and 99% (F.99) quantiles are used to calculate the mean x and variance s2 of a given economic impact (Lichtenberg, 2000): x ¼ ðF:01 þ 2:95$M þ F:99 Þ=4:95 s2 ¼ ðF:99  F:01 Þ2 =21:66 If impacts are assumed to be independent and the largest impact variances are of the same order of magnitude, their sum can be assumed to follow a normal distribution.

3.4.

Effects under investigation

Blending water will affect its corrosive potential (Imran et al., 2006). The following corrosion indicators are considered in the assessment:  Calcium Carbonate Precipitation Potential (CCPP). The CCPP (mg/l) indicates whether water will dissolve or precipitate calcium carbonate. Recommended CCPP values vary widely but are usually within the range of 0e10 mg/l (Lahav and Birnhack, 2007).

 Larson Ratio (LR). The Larson Ratio is defined as LR ¼ 2* [SO42-] þ [Cl]/[HCO3]. Again, different guideline values are considered around the world, but generally values above 0.5 are considered to have a high corrosion potential for steel and cast iron (Delion et al., 2004; McNeill and Edwards, 2001).  pH, alkalinity and SO42-. Both CCPP and Larson Ratio have been shown inadequate to describe copper corrosion (Edwards et al., 1996). High pH (z8) is found to protect against copper corrosion, while increased alkalinity and sulfate has been correlated with increased copper corrosion (Edwards et al., 1996; Xiao et al., 2007). The following economic impacts are evaluated in the assessment. For brevity only key publications are referenced here, while a longer discussion of the rationale for including each effect is presented elsewhere (Rygaard et al., 2009). Water quality related effects:  Dental caries. A strong relationship between fluoride levels in drinking water and dental caries is well proven, and a recent study also indicates a significant correlation between the incidence of caries and calcium levels (Bruvo et al., 2008).  Cardiovascular disease. There has been a long debate on whether drinking water hardness affects cardiovascular health. But recently the World Health Organization concluded that there seems to be a weak protective effect against cardiovascular mortality from drinking water magnesium (Cotruvo and Bartram, 2009).  Atopic eczema. Hardness levels have been associated with commonly occurring atopic eczema among children (McNally et al., 1998).  Corrosion of appliances. High drinking water salinities are known to impair the lifetime of household appliances such as clothes washer, dish washers and water heaters (Ragan et al., 2000).  Corrosion of distribution system. As with the appliances, the lifetime of piping systems has been correlated with drinking water salinity (Characklis, 2004).  Soap consumption. The dosage of detergent for clothes washing is related to local hardness levels (Rygaard et al., 2010).  Bottled water consumption. Danish water supply has a long tradition of simple treatment of unpolluted groundwater resources. The introduction of desalinated water is a significant change and can influence the confidence of Copenhagen’s water supply and eventually lead to increased consumption of bottled water. A change in bottled water consumption is not considered for Perth, since the population is used to receive blends from a diverse range of water resources. Effects from the production of drinking water:  CO2-emissions. Environmental lifecycle assessments show that electricity consumption dominates the environmental impacts of water production due to green house gas emissions from the power production (Munoz and FernandezAlba, 2008).

3715

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 1 2 e3 7 2 2

 Production costs. Decisions in water supply management and planning are based on economic feasibility studies focusing on capital and operation costs of different alternatives (Marsden Jacob Associates, 2007).  Remineralization costs. Potential health effects are dependent on a few minerals in the drinking water (fluoride, magnesium and calcium) that can be re-introduced in remineralized water or increased in natural waters during the water treatment process (Birnhack and Lahav, 2007).

3.5.

Concentration-response relationships

Concentration-response relationships are found or derived from the literature (Table 1). The relationships have been adjusted to provide continuous functions for the varying mineral concentrations found in Perth and Copenhagen. The physical changes (responses, Ei) are calculated as


 f wq1  f wq0
 $E0 Ei ¼ f wq0

(1)

f(wqx) is the concentration-response relationship used for prediction of rate, lifetime or consumption depending on the water quality in the base situation (wq0) and alternative scenario (wq1). E0 is the observed base occurrence, lifetime or consumption. In other words, concentration-response relationships developed for other locations are used to estimate

a relative change, which is then imposed on the observed occurrence, lifetime or consumption in the area in question. An expression employing a relative change from a background level is preferable to a relationship based on water quality alone because other factors, such as general health status of the population, material choices etc. are difficult to include in a direct concentration-response relationship. The predicted physical responses are assigned an annual economic value based on treatment or replacement costs and converted to a unit economic change Ci (V/m3) by dividing by the annual delivered water volume V (m3/yr). All impacts are scaled by a factor Pi, i.e. exposed population (health), number of items (lifetime) or annual usage (consumption). For health the economic impact Ci, health (V/m3) is calculated as: Ci;health ¼

Ei $Ci;y $Pi V

(2)

Where Ci,y (V/case) is the cost per case. According to Ragan (2000) costs of continuously replaced items (e.g. clothes washers) with a steady stage age distribution and uniform prices is calculated as C0/ti, where C0 is investment cost (V/item) and ti is the lifetime of the appliance i. The economic benefit Ci,lifetime (V/item/yr) of changed lifetimes is therefore estimated as  Ci;lifetime ¼

1 1  Ei  E0 E0

 $

C0 $Pi V

(3)

Table 1 e Concentration-response relationships used in the water quality assessments for Perth and Copenhagen. Relationships used to calculate maximum and minimum quantiles (see Section 3.3) are included in the Supplementary material. Impact category Health Caries (Decayed Missing Filled Surfaces (DMF-S)/ person). Cardiovascular diseases (CVD) (RR, Incidents/ person/yr)a Atopic eczema (AE) (RR, persons affected, 1-yr prevalence)a Corrosion Clothes washers (lifetime, yr) Dish washers (lifetime, yr) Water heaters (lifetime, yr) Distribution system (lifetime, yr) Consumption Bottled water (BW) (l/p/yr) (Copenhagen only) Detergents (clothes) (Cost per washing, V)

Concentration-response function    0:11,ðCa83:53Þ DMF  S ¼ exp 1:05  0:18,ðF0:33Þ 0:25 25:63

Relationship range

References

Ca: 31.4e162.3 F: 0.06e1.61 (mg/l)

(Bruvo et al., 2008)

RR ¼ 1  0:0215$ðMgnew  Mgbaseline Þ

Mg: 1.17e10.2 (mg/l)

Derived from (Kousa et al., 2008)

RR ¼ 1 þ 0:0022$ðHardnessnew  Hardnessbaseline Þ

Hardness: 118e314 (as mg/l CaCO3)

Derived from (McNally et al., 1998)

LifetimeCW ¼ 20,e0:0001419$TDS

TDS: 100e3600 (mg/l)

(Ragan et al., 2000)

TDS: 100e3600 (mg/l)

(Ragan et al., 2000)

TDS: 100e3600 (mg/l)

(Ragan et al., 2000)

Unknown

(Characklis, 2004)

No change expected 

LifetimeWH ¼ 26$ 0:412 þ 0:588$e0:00472$TDS



LifetimeDS ¼ 1:1$105 $TDS2  0:04$TDS þ 109:2

DBW ¼ 6:3 (Copenhagen only)

Cwashing ¼ 0:22 þ 0:00056$Hardness

a RR: Relative Risk ¼ New incidence (Ei)/Base incidence (E0).

(Rygaard et al., 2009)

Hardness: <500 (as mg/l CaCO3)

(Rygaard et al., 2010)

3716

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 1 2 e3 7 2 2

For consumption (of soap), the response Ci, consumption is given as a cost and the economic impact is calculated as: Ci;consumption ¼

Ei $Pi V

(4)

For investments that are new at the time of change to the water supply system (e.g. the cost of desalinated water in Copenhagen) the uniform annual cost is found by applying a cost recovery factor: h
i. V Ci;investment ¼ C0 $rð1 þ rÞt = ð1 þ rÞt 1

(5)

where C0 is the initial investment cost, r is the real discount rate (assumed 6% per year) and t is the estimated expected lifetime of the investment. Scaling factors, baseline information (incidences, lifetimes, consumption rates) and costs for the economic impact assessment have been obtained from national statistical agencies, trade organizations and other public available sources (Table 2). The uncertainty associated with each source and value is reflected in the minimum and maximum quantiles used for the triple estimate. All prices are converted to 2009 Euros using exchange rates of June 2009.

3.6.

Production costs

Current costs associated with the abstraction and treatment of water have been obtained from the utilities (Table 3). Costs of future augmentations to the current water supply in Perth are based on Water Corporation’s own projections for development of new drinking water sources for Perth (Water Corporation, 2009). The marginal costs of new sources or increased water efficiency varies from V0 to 4.8/ m3. The long run marginal cost is currently expected to be between V0.7 to 0.96/m3 (Pickering, 2009). However, it is assumed that the future demand-supply gap can be filled for less than V1.2/m3. Minimum costs are assumed to be V0.3/m3. In the

evaluation of the Perth case the long term marginal cost of alternative supply and efficiency is thus introduced as a shadow price for one m3 water abstracted from either groundwater or surface water. This value reflects the avoided cost due to conventional water resources being replaced by desalinated water.

3.6.1.

Shadow costs of green house gas emissions

Treatment and conveyance of water causes green house gas emissions are expressed here in terms of CO2-equivalents. Changed CO2-emissions from the operation of the water supply systems are based on electricity consumption data from the utilities and green house gas emissions officially declared for electricity consumption in the local area (Table 2). Green house gas emissions data for power generation is based on official statements for South West Interconnected System in Western Australia. In Perth it is estimated that 0.8 kg CO2-eq are released per kWh consumed (Department of Climate Change, 2009). There is no consensus on the price of an emitted or avoided ton CO2-eq. However, trading within the European Union Allowances (EUA) scheme indicates a current international market price of CO2-emissions. In 2008-09 the average price was V18 per ton CO2-eq with minimum and maximum at V8 and V29 per ton (BlueNext, 2010). Future marginal prices for avoiding CO2-emissions have been assessed to be in the range of 0 to V74/ton CO2-eq emitted (Sims et al., 2007). Here a price of V18/t with minimum and maximum quantiles of V8 and V74 per ton is employed.

4.

Results

4.1.

Water quality modeling

Groundwater, surface and desalinated water are mixed in the Hamilton Hill and Thompsons Reservoir areas of Perth (Table 4). For aesthetic reasons it is recommended that all

Table 2 e Values used for estimating societal economic impacts in Perth (1% and 99% quantiles used in the triple estimates). Details are found in the Supplementary material. Scaling factor (Pi) Health Population affected Dental caries (DMF-S/ 90,693 person >6 yrs) Cardiovascular diseases 99,750 (cases) Atopic eczema (cases) 15,583 Appliances Items affected Clothes Washer 31,500 Dish Washer 16,670 Heater/heat exchanger 35,000 Distribution 1 Consumption Soap 23 million washings per yr Baseline electricity consumption and CO2-eq emission Mitigation GWh/yr 22.2 (22.1e23.2) CO2-emission

Cost

Baseline occurrence/ lifetime/consumption (E0) Cases per person 1.4 (1.0e2.0)

(V/case/yr) (Ci,y) 5 (2.0e9.7)

0.07 (0.06e0.08)

2547 (2037e3055)

0.11 (0.05e0.17) Lifetime (yrs) 12 (9e15) 11 (8e14) 12 (8e20) 60 (40e100)

1569 (3895e805) Cost V/item (C0) 600 (480e720) 600 (480e720) 627 (313e940) 36 (18e72) Million

0.31 (0.23e0.42) (V/washing)

N/A

kg/kWh 0.84 (0.42e1.3)

V/t 18 (8e74)

3717

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 1 2 e3 7 2 2

Table 3 e Cost of CO2-mitigation, remineralization and production. Details and references in Supplementary material. Values used for estimation of quantiles Most likely Min/max quantile Mitigation CO2 (V/t CO2 emitted) Mg/Ca ion-exchange (V/m3) Production costs Long term marginal, conventional water resources (V/m3) Marginal cost, current desalination (V/m3)

18 0.0045

8e74 0.002e0.009

0.83

0.3e1.2

0.72

0.58e0.86

water blends maintain a total dissolved solids concentration below 500 mg/l and this constrains the possible share of groundwater, since it has an average salinity of 849 mg/l. Notably all three water sources are undersaturated with CaCO3, with CCPP values from 7 to 0.1 mg/l. Also the Larson Ratios range from 1.7 to 4.4 (Table 4) and are much higher than the threshold value of 0.5. It is proposed that the alkalinity of the desalinated water is increased to 100 mg/l as CaCO3 and remineralized with magnesium to the same concentration as in the groundwater (19 mg/l). This can be done by increasing the lime dosage and implementing a Ca:Mg ion-exchange process in the post-treatment of the desalinated water. In the following this water type is called Perth B. Mixing of the three water types in Perth has been done in 10% increments giving 66 blends representing the possible

final water qualities. Mixing desalinated water with only surface water gives a highly aggressive blend (CCPP from 7 to 1.4) while a less aggressive blend is formed from mixing only desalinated water and groundwater (from 1.4 to 0.0) (Fig. 1a). A mix of groundwater and surface water provides the largest span of CCPP values between 7 and 0.1. Mixing all three water resources provides CCPP values that all fall within these extremes as indicated in Fig. 1a. Increasing the lime dosage in the desalination post-treatment (desalination type B) significantly increases the CCPP of blends with desalinated water. From Fig. 1b it can be seen that when desalination provides more than half (53%, read from the graph) of the water, all blends will have a positive CCPP. The Larson Ratio will also be improved by the proposed change in lime dosage at the desalination plant. Although a higher lime dosage in general will lower the Larson Ratio, it will not be reduced below the 0.5 threshold for mild iron and steel corrosion potential (Fig. 2). Also indicators of copper corrosion are affected by the introduction of desalinated water. pH is slightly increased from 7.4 to 7.6 in the case of desalinated water type A replacing surface water from the system and slightly reduced from 7.8 to 7.6 if groundwater is replaced by desalinated water type A (Fig. 3). The proposed post-treatment change to type B will further increase pH above 8 (Fig. 3). Alkalinity will change linearly with the introduction of desalinated water. When desalinated water replaces surface water, alkalinity will increase and when replacing groundwater it will decrease. In the case of Perth type B water, the potential copper corrosive effect of an increase in alkalinity is expected to be damped by the increase in pH, according to studies showing alkalinity to be of less importance for copper corrosion in high pH water (Edwards et al., 1996).

Table 4 e Typical water quality values for groundwater, surface water and desalinated water in Perth. Australian guidelines

Baseline blend

Groundwater

Surface water

30% Groundwater 40% Desalinated water 30% Surface water

Aeration Settling Sand filtration Fluoridation

6.5e8.5 e <200

7.7 83. 132.

e e <1.5 <250 <250 <500 e e

42. 6.7 0.82 27. 155. 375. 3.0 3.5

Treatment

pH Alk (as CaCO3) Hardness (as CaCO3) Ca Mg F SO4 Cl TDS Larson ratio CCPP

Desalination Type A

Desalination Type B Alternative buffering

Chlorination Fluoridation

Pre-treatment Reverse osmosis Acidified with CO2 Lime dissolution Chlorination Fluoridation

7.8 140. 179.

7.4 27. 45.

7.6 81. 162.

Pre-treatment Reverse osmosis Acidified with CO2 Lime dissolution Chlorination Ion-exchange Fluoridation 8.5 100. 150.

41. 19. 0.80 80. 375. 849. 4.4 0.1

12. 3.8 0.8 9.5 63. 142. 3.6 7.0

65. 0. 0.85 0. 60. 195. 1.0 1.4

29. 19. 0.85 0. 60. 189. 0.85 6.8

3718

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 1 2 e3 7 2 2

Fig. 1 e Estimated Calcium Carbonate Precipitation potential for 66 water blends in Perth. Calculations are based on the current desalination post-treatment (a) and proposed desalination post-treatment (b).

4.2.

Economic impacts

Increasing the share of desalinated water (Type A) in Perth from 40% to 70% replacing groundwater will increase the concentration of calcium to 49 mg/l, while fluoride is slightly increased to 0.84 mg/l. Using the most likely relationship for dental caries to estimate DMF-S (Table 1) gives the values 2.4 before and 2.3 after the desalination share is increased. This 4% decrease is then combined with the baseline DMF-S value (1.4) to find a predicted reduction in DMF-S of 0.1e1.3. With an annualized filling cost of V5 per DMF-S the total benefit to Perth is V18,307. The same procedure is used to find the minimum and maximum quintiles of V 4826 and V48,047 respectively. The triple estimation method is used find the mean change to be V21,592 (or V0.002
0.001/m3). The method is repeated for all impacts and blending options. Without desalinated water in Perth, the total economic change from the baseline situation is predicted to be between V0.11
0.1 and 0.25
0.2/m3 delivered

water (Fig. 4) depending on the ratio between groundwater and surface water. With desalinated water in the system the economic benefit is decreased to a minimum value of V0.24
0.2)/m3. Dilution of the conventional sources with desalinated water Perth Type A reduces the magnesium content of the final blends and causes the potential impact from cardiovascular diseases to dominate the economic impact. Supplying the area with 100% desalinated water would lead to a predicted cardiovascular impact of V0.21 (
0.1)/m3 which is equal to 53% of the absolute values of economic change. It is followed by the cost of CO2-mitgation at V0.05 (
0.04) and production costs at V0.05 (
0.2)/m3. Remineralizing with magnesium (Perth Type B, Fig. 4) increases the economic benefit significantly for blends (including desalinated water) for a very small extra cost of at most V0.01(
0.004)/m3 for 100% desalinated water. Remineralization reverses the negative benefit of 100% desalinated water (V0.24) to a positive (V0.36
0.2). For both water types

Fig. 2 e Larson ratio in Perth for various shares of groundwater and surface water mixed with two types of desalinated water.

Fig. 3 e Predicted pH in Perth for various shares of groundwater and surface water mixed with two types of desalinated water.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 1 2 e3 7 2 2

(A and B), Perth benefits from avoided costs of developing alternative water resources and reduced deterioration rates of appliances and the distribution system, especially water heaters.

5.

Discussion

5.1. Perth

Optimal water composition in Copenhagen and

post-treatment and increase the calcium carbonate precipitation potential to a higher level as seen in Fig. 1. With the objectives of increasing CCPP, reducing Larson Ratio (Fig. 2), and achieving a net economic benefit, Perth would benefit from changing the desalination post-treatment to deliver type B water (Fig. 4).

5.2.

The results presented here and in the Supplementary material show that without remineralizing with magnesium and fluoride (Perth A and Copenhagen B), both cities would achieve a net economic benefit by minimizing the share of desalinated water in their respective systems. For Perth the results further indicate that maintaining a high share of groundwater in the system will prevent potential negative impacts of the low magnesium concentration of the surface water. However, in practice any share of groundwater above 40% will violate Australia’s aesthetic drinking water guideline for total dissolved solids (<500 mg/l). Introducing remineralization with fluoride and magnesium will, at a very low cost, make the introduction of desalinated water considerably more beneficial for the two cities. In both cases remineralizing with magnesium to the natural background level in groundwater (both places 19 mg/l) will prevent a potential negative economic outcome. Maintaining the magnesium content in desalinated water will make the optimal blends those with the highest possible share of desalinated water in the water supplies. It should be noted that none of the drinking water guidelines from World Health Organization, Australia, EU, and Denmark currently require water supplies to ensure a minimum magnesium or fluoride content. In Perth the desalinated water (type A) causes blends to be no more or less aggressive with respect to calcium carbonate than the mix of surface water and groundwater resources. Desalinated water is predicted to have a positive effect on corrosion potentials as indicated by reduced Larson Ratio and reduced sulfate levels. It is possible to alter the desalination

3719

General remarks on optimal water composition

The assessment shows significant benefits from a reduction of the mineral content in hard water (hardness >200 mg/l as CaCO3) and with moderate to high total dissolved solids content (>500 mg/l) if fluoride and magnesium is maintained around 0.8 mg/l and 20 mg/l respectively. This conclusion is similar for both cases in Australia and Denmark.. The economic benefits can even outweigh the extra costs associated with changing from a low intensity treatment such as the groundwater treatment in Copenhagen to a high intensity treatment like desalination of brackish water as demonstrated in the Supplementary material (SI1). The example from Copenhagen shows that altering the fluoride and calcium content has a significant impact on dental health expenses (SI1). The economic benefits of fluoridation are also evident from the Perth case. Since Perth fluoridates all water sources the impact on dental health is limited to effects stemming from changed calcium content. The impact on caries from calcium levels in Perth are negligible compared to other economic impacts investigated here (Fig. 4). It is proposed to establish a set of optimum water quality criteria in addition to legal requirements. Based on the results of the economic assessment it is possible to derive a quite narrow span of optimum drinking water quality criteria to be used as a guideline in the planning of drinking water services (Table 5). If a water supply delivers water significantly different from the criteria proposed here, an economic assessment may show that upgrading treatment processes will have an overall economic benefit for society. It is also noteworthy that for 100% desalination, green house gas emission costs accounts for less than 3 and 14% of the total economic impact in Perth and Copenhagen

Fig. 4 e Predicted economic change in Perth for 0e100% desalinated water replacing either groundwater or surface water.

3720

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 1 2 e3 7 2 2

respectively (Fig. 4 and SI1). These results indicate that an energy intensive water treatment process such as desalination can generate other benefits that may outweigh the potential cost of green house gas mitigation. Reverse osmosis desalination in Copenhagen benefits from a low salinity in the Baltic Sea (w1%) whereas Perth relies on ocean water for its desalination (salinity w3%). This partly explains the difference in CO2-equivalent emissions for the two scenarios.

5.3.

Scope and data inventory

It is important to be aware of assumptions made, the scope of the analysis, and to note which impacts have been left out. The concentration-response relationships are directly adopted or based on published studies. Most relationships have been established for other purposes and it can be questioned whether the relationships are transferable to the application made here. This issue has been addressed through the implementation of the successive principle where the reliability of each relationship is reflected in the estimated 1 and 99% quantiles. The reliability of the magnesium-cardiovascular disease relationship is particularly significant. The relationship used is similar to that of Rylander et al. (1991) and is assumed to be a valid description of the drinking water magnesium to cardiovascular disease relationship considered by the World Health Organization (Cotruvo and Bartram, 2009). However, the link is still under debate, with both supporters (Catling et al., 2008) and detractors (Morris et al., 2008) being found amongst the current literature. The studies used for deriving dose-response relationships for health impacts (cardiovascular disease, dental caries, atopic eczema) all state that drinking water quality can only partly explain the observed health effects and this means that changes in rates predicted only by water quality are quite uncertain. Other

factors (e.g. household income) are often the dominant predictor for occurrence rates. The predictions presented in this paper are the best currently available and until further evidence is provided by the research community results must be evaluated in a local context before making decisions on system changes. Details on the concentration-response relationships are included in the Supplementary information. Due to the large economic impact from this particular relationship, the economic analysis would be significantly different if the impact of water quality on cardiovascular diseases is less than presented here. If this effect is removed from the analysis then it becomes economically attractive (up to V0.2
0.1/m3) to increase the share of surface water in Perth (Fig. 5). Increasing the share of desalinated water in Perth has only a minor economic impact, while there is a negative economic impact with increased use of groundwater (up to V0.13
0.1/m3) (Fig. 5). Other impacts not considered here include heat loss in heat exchangers and from heating elements covered with CaCO3-scale, pressure loss in pipes due to scaling, use of softeners in households, and environmental impacts from changed release of household chemicals to recipients. Although evaluated qualitatively here, currently there are no economic relationships between corrosion indicators as those mentioned above and economic effects. For example, reduced copper corrosion and hence reduced failure rates and damage from pipes and appliances after introduction of desalination has not been valued here. There is a well established relation between chlorination disinfection efficiency, byproduct formation and water quality parameters not included here. Natural organic matter is associated with the formation of toxic disinfection by-products (Nikolaou and Lekkas, 2001). For example, the introduction of desalinated water in Perth has reduced the amount of natural organic matter in the

Table 5 e Proposed optimum drinking water criteria for selected parameters in comparison with current official guidelines. Recommendations are based on the economic assessment presented here and cited literature. Parameter

Proposed optimum (mg/l)

Notes

Mg

>10

Ca

40e50

F

0.5e1

Hardness (as CaCO3)

<150

TDS

<200

Maintain magnesium levels as high as possible. The low cost of remineralization justifies a precautionary addition of magnesium up to 20 mg/l in cases with magnesium deprived water sources, see Section 4.1.3. Calcium levels in this range are necessary to maintain a low hardness level and sufficient magnesium. Low levels of calcium are adequate if sufficient fluoride is present to maintain dental health, however F and Ca must be evaluated together. Fluoride has a pronounced beneficial effect (Section 5.1) and according to Bruvo et al. (2008). Here an increase from 0.25 to 0.75 will reduce DMF-S by 1 for calcium levels below 100 mg/l. This ensures a dental protective effect well within the maximum health guideline of 1.5 mg/l. Hardness can be reduced below 150 mg/l (as CaCO3) and ensure low soap consumption and possibly reduce eczema risk (Section 4.1.3). Total dissolved solids deteriorate materials and should be maintained as low as possible (Section 4.1.3).

Guidelines (mg/l) Denmark

Australia

WHO

e

e

e

<200

e

e

<1.5

<1.5

<1.5

89e534

<200

e

<1500

<500

e

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 1 2 e3 7 2 2

Fig. 5 e Predicted economic change assuming magnesium concentration has no impact on cardiovascular diseases. Calculated for 0 and 100% desalinated water in Perth.

system and most likely reduced the formation of undesired disinfection by-products. The impact of temperature on public accept is not considered here. Perth’s Water Corporation has documented (not shown here) many benefits from the introduction of desalinated water into the Perth distribution system not valued in the economic assessment. These include improved disinfection management with stable chlorine residuals, lower customer complaints, less taste and odor issues, and low copper corrosion levels. Most importantly, desalination has also reduced the reliance on rainfall, an uncertain commodity. Finally, the analysis builds on modeled water qualities and assuming complete mixing across the city. In practice blends may vary across the city and by time and this issue is not considered here. A discussion of the uncertainties and the likelihood of a positive economic outcome can be found in the Supplementary material published online with this article (SI2).

3721

 Effects from altering magnesium, calcium and fluoride levels can have economic consequences that outweigh differences in production costs.  Careful design of the desalination post-treatment is necessary to ensure that final water blends not only comply with water quality criteria but also provide a diverse range of indirect economic benefits to society.  Due to the potential risk of increased cardiovascular disease the cost effective option of remineralizing desalinated water (and probably other magnesium deprived water resources) with magnesium should always be considered.  Even relatively small reductions of calcium or fluoride should always lead to consideration of fluoridation or other dental health initiatives.  The results are subject to considerable uncertainties, particularly those related to health impacts, and future water production costs. A decision making process will benefit from improved understanding of these issues. Based on the results presented here a set of optimum water quality parameters have been proposed. Although the proposed criteria are significantly stricter than official guidelines, it is shown that the benefits can in some circumstances outweigh the extra costs of membrane desalination and remineralization.

Acknowledgements The authors would like to thank: Copenhagen Energy; Stewart Burn from CSIRO Land and Water; and Technical University of Denmark for their help during the project.

Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.04.025.

references

6.

Conclusions

It has been shown how engineered water based on membrane treatment, remineralization, and blending can improve drinking water quality beyond the requirements stated in drinking water quality criteria. It is also shown that high production costs of desalination can be justified when the benefits of improved water quality and saved fresh water resources are taken into account. The combination of water quality modeling, concentration-response relationships and the successive principle provides a simple and transparent framework for assessing impacts and the uncertainties associated with subjective judgment. In general it can be concluded that the economic impacts caused by introducing desalinated water to drinking water supplies are very dependent on the post-treatment of the desalinated water:

Birnhack, L., Lahav, O., 2007. A new post-treatment process for attaining Ca2þ, Mg2þ SO42- and alkalinity criteria in desalinated water. Water Research 41 (17), 3989e3997. BlueNext, 2010. Statistics (On line). http://bluenext.eu accessed 3-1-2010. Bruvo, M., Ekstrand, K., Arvin, E., Spliid, H., Moe, D., Kirkeby, S., Bardow, A., 2008. Optimal drinking water composition for caries control in populations. Journal of Dental Research 87 (4), 340e343. Catling, L.A., Abubakar, I., Lake, I.R., Swift, L., Hunter, P.R., 2008. A systematic review of analytical observational studies investigating the association between cardiovascular disease and drinking water hardness. Journal of Water and Health 6 (4), 433e442. Characklis, G.W., 2004. Economic decision making in the use of membrane desalination for brackish supplies. Journal of the American Water Resources Association 40 (3), 615e630.

3722

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 1 2 e3 7 2 2

Cotruvo, J., Bartram, J., 2009. Calcium and Magnesium in Drinking Water: Public Health Significance. World Health Organization, Geneva, Switzerland. Delion, N., Mauguin, G., Corsin, P., 2004. Importance and impact of post treatments on design and operation of SWRO plants. Desalination 165, 323e334. Department of Climate Change, 2009. National Greenhouse Accounts (NGA) Factors. Commonwealth of Australia, Canberra. Edwards, M., Schock, M.R., Meyer, T.E., 1996. Alkalinity, pH, and copper corrosion by-product release. Journal American Water Works Association 88 (3), 81e94. Imran, S.A., Dietz, J.D., Mutoti, G., Xiao, W.Z., Taylor, J.S., Desai, V., 2006. Optimizing source water blends for corrosion and residual control in distribution systems. Journal American Water Works Association 98 (5), 107e115. Kousa, A., Havulinna, A.S., Moltchanova, E., Taskinen, O., Nikkarinen, M., Salomaa, V., Karvonen, M., 2008. Magnesium in well water and the spatial variation of acute myocardial infarction incidence in rural Finland. Applied Geochemistry 23 (4), 632e640. Lahav, O., Birnhack, L., 2007. Quality criteria for desalinated water following post-treatment. Desalination 207 (1e3), 286e303. Lichtenberg, S., 2000. Proactive Management of Uncertainty Using the Successive Principle: a Practical Way to Manage Opportunities and Risks. Polyteknisk Press, Copenhagen, Denmark. Marsden Jacob Associates, 2007. The Cost-effectiveness of Rainwater Tanks in Urban Australia. National Water Commission, Australian Government, Canberra, Australia. McNally, N.J., Williams, H.C., Phillips, D.R., Smallman-Raynor, M., Lewis, S., Venn, A., Britton, J., 1998. Atopic eczema and domestic water hardness. Lancet 352 (9127), 527e531. McNeill, L.S., Edwards, M., 2001. Iron pipe corrosion in distribution systems. Journal American Water Works Association 93 (7), 88e100. Morris, R.W., Walker, M., Lennon, L.T., Shaper, A.G., Whincup, P., 2008. Hard drinking water does not protect against cardiovascular disease: new evidence from the British regional heart study. European Journal of Cardiovascular Prevention & Rehabilitation 15 (2), 185e189. Munoz, I., Fernandez-Alba, A.R., 2008. Reducing the environmental impacts of reverse osmosis desalination by

using brackish groundwater resources. Water Research 42 (3), 801e811. Nikolaou, A.D., Lekkas, T.D., 2001. The role of natural organic matter during formation of chlorination by-products: a review. Acta Hydrochimica et Hydrobiologica 29 (2e3), 63e77. Parkhurst, D.L., Appelo, C.A.J., 2008. PHREEQC for Windows. Build 2.15.07. Pickering, P., 2009. The Cost Effectiveness of Residential Rainwater Tanks in Perth. Marsden Jacobs Associates, Perth, Australia. Ragan, G.E., Young, R.A., Makela, C.J., 2000. New evidence on the economic benefits of controlling salinity in domestic water supplies. Water Resources Research 36 (4), 1087e1095. Rygaard, M., Arvin, E., Binning, P.J., 2009. The valuation of water quality: effects of mixing different drinking water qualities. Water Research 43 (5), 1207e1218. Rygaard, M., Arvin, E., Binning, P.J., 2010. Indirect economic impacts in water supplies augmented with desalinated water. Water Science & Technology: Water Supply 10 (4), 664e671. Rylander, R., Bonevik, H., Rubenowitz, E., 1991. Magnesium and calcium in drinking-water and cardiovascular mortality. Scandinavian Journal of Work Environment & Health 17 (2), 91e94. Sims, R.E.H., Schock, A., Adegbululgbe, A., Fenhann, J., Konstantinaviciute, I., Moomaw, W., Nimir, H.B., Schlamadinger, B., Torres-Martinez, J., Turner, C., Uchiyama, Y., Vuori, S.J.V., Wamukonya, N., Zhang, X., 2007. In: Metz, B., Davidson, O.R., Bosch, P.R., Dave, R., Meyer, L.A. (Eds.), Energy Supply, in Climate Change 2007: Mitigation. Contribution of Working Group III to the Fourth Assessment Report of the Intergovernmental Panel On Climate Change. Cambridge University Press, Cambridge, UK and New York, NY, USA. Water Corporation, 2009. Water forever. Towards Climate Resilience (Perth, Australia). Withers, A., 2005. Options for recarbonation, remineralisation and disinfection for desalination plants. Desalination 179 (1e3), 11e24. Xiao, W.Z., Hong, S.K., Tang, Z.J., Taylor, J.S., 2007. Effects of blending on total copper release in distribution systems. Journal American Water Works Association 99 (1), 78e88.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 2 3 e3 7 3 4

Available at www.sciencedirect.com

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

Survival of prototype strains of somatic coliphage families in environmental waters and when exposed to UV low-pressure monochromatic radiation or heat Hee Suk Lee 1, Mark D. Sobsey* Department of Environmental Sciences and Engineering, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA

article info

abstract

Article history:

The potential use of specific somatic coliphage taxonomic groups as viral indicators based

Received 6 August 2010

on their persistence and prevalence in water was investigated. Representative type strains

Received in revised form

of the 4 major somatic coliphage taxonomic groups were seeded into reagent water and an

14 April 2011

ambient surface water source of drinking water and the survival of the added phages was

Accepted 17 April 2011

measured over 90 days at temperatures of 23e25 and 4

Available online 22 April 2011

PhiX174), Siphoviridae (type strain Lambda), and Myoviridae (type strain T4) viruses were the

Keywords:

and the Siphoviridae (type strain Lambda) were the most resistant viruses to UV radiation

Somatic Coliphage

and the Myoviridae (type strain T4) and the Microviridae (type strain PhiX174) were the most

Survival

resistant viruses to heat. Based on their greater persistence in water over time and their




C. Microviridae (type strain

most persistent in water at the temperatures tested. The Microviridae (type strain PhiX174)

Inactivation

relative resistance to heat and/or UV radiation, the Myoviridae (type strain T4), the Microviridae (type strain PhiX174), and the Siphoviridae (type strain Lambda) were the preferred candidate somatic coliphages as fecal indicator viruses in water, with the Microviridae (type strain PhiX174) the most resistant to these conditions overall. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Fecal indicator microorganisms currently used for water quality monitoring are bacteria, such as total coliforms, fecal (thermotolerant) coliforms, Enterococcus spp. and Escherichia coli. However, many waterborne pathogens are enteric viruses, and bacterial indicators may be inadequate or unreliable indicators of their presence, persistence and concentrations in fecally contaminated environmental waters. Human enteric viruses have been shown to be more persistent in water and more resistant to physical agents such and UV radiation and heat than are the non-spore-forming fecal indicator bacteria. To ensure adequate protection against waterborne disease, there is a need for reliable viral indicators

in water quality monitoring programs. Although there have been a number of previous studies to find reliable viral fecal indicators (Armon, 1993; Havelaar et al., 1993; Jofre et al., 1995; Sobsey et al., 1995), there is no clear evidence of reliability and no clear consensus as to which fecal indicators are most predictive of human enteric viruses. Certain types of bacteriophages, specifically those of E. coli (coliphages) are proposed candidate indicators of human enteric viruses in water. They are present in human and animal feces and some are small, icosahedral and non-enveloped viruses, making them structurally similar to many human enteric viruses. There are two main types of coliphages: somatic and the male-specific (Fþ). The somatic coliphages are DNA viruses that infect E. coli through attachment to specific sites on the outer cell layer,

* Corresponding author. E-mail address: [email protected] (M.D. Sobsey). 1 Present address: Korea Water Resources Corporation, San 6-2, Yeonchuk-dong, Daeduck-gu, Daejeon, Republic of Korea. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.024

3724

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 2 3 e3 7 3 4

such as lipopolysaccharide. The male-specific coliphages are single-stranded RNA and DNA viruses that infect the cell via the pili appendages present on the surface of male strains of the bacterium. Male-specific coliphages have been previously investigated as viral indicators (Chung and Sobsey, 1993; Colford et al., 2007; Dore et al., 2000; Cole et al., 2003; Love and Sobsey, 2007). However, their infrequent presence in human feces, their relative scarcity and their rapid die-off rates in warm waters (Chung and Sobsey, 1993; Love et al., 2007) limit the usefulness of Fþ coliphages as indicator viruses. Somatic coliphages also have been proposed as fecal indicators of human enteric viruses in studies of sewage, source water for drinking water and marine waters (MoceLlivina et al., 2005; Muniesa and Jofre, 2007; Jofre, 2008), and real-time monitoring of somatic coliphages as fecal indicators has been suggested by previous investigators (Araujo et al., 1997; Skraber et al., 2004; Garcia-Aljaro et al., 2008). Although previous studies provide some evidence that somatic coliphages are potentially useful candidates as fecal indicator viruses, their taxonomic diversity and potential heterogeneity within a taxonomic group has not been taken into account. The somatic coliphage group encompasses four distinct virus families, each containing several genera: Myoviridae, Microviridae, Siphoviridae, and Podoviridae. The Microviridae are small, single-stranded DNA viruses, and the other families are double-stranded DNA viruses of varying size, morphology and biophysical properties. It is possible that the survival of these different phages in environmental waters may differ among families and genera, with some being more persistent in water than others. Evidence of their different persistence in the environment comes from previous studies showing different inactivation kinetics of some somatic and male-specific coliphages by UV radiation (Shin et al., 2005). Because some coliphages are already used as biodosimeters for UV radiation calibration and performance evaluation against viruses, data on the responses of representative members the different taxonomic groups of somatic coliphages to UV irradiation has the potential to provide information that informs their further use as biodosimeters. Due to the diversity of the somatic coliphage groups, more data on the comparative persistence and prevalence of representative members of the different somatic coliphage families are needed to identify which ones may serve as reliable viral indicators in water based on their survival and when exposed to physical environmental stressors, such as UV radiation and high temperatures. In order to identify candidate somatic coliphage families for use as indicators of human enteric viruses in surface water, we evaluated the comparative persistence of representative members of each of the 4 taxonomic group(s) using established prototype strains in water incubated at low and intermediate environmental temperatures and when exposed to low pressure, monochromatic UV radiation and high temperatures.

2.

Materials and methods

2.1.

Test waters

Test waters were reagent water and a natural surface water. For the reagent water, Dulbecco’s phosphate buffered saline (PBS)

solution was diluted 10-fold in reagent-grade water. Reagentgrade water was produced from laboratory tap water by a Dracor water purification system (Dracor, Durham, NC) which includes reverse osmosis and ultraviolet light treatment. Ten-fold diluted Dulbecco’s PBS was used to provide salt content (total dissolved solids) in the range of fresh water. Surface water was obtained from University Lake, an impoundment that serves as the source water for the Orange County Water and Sewer Authority (OWASA) drinking water treatment facility serving Chapel Hill, NC. This water had the following physical, chemical and microbiological quality: temperature ¼ 18.2
C (SD: 4.3), turbidity ¼ 4.5NTU (SD: 0.45), pH ¼ 6.6 (SD: 0.13), alkalinity ¼ 25 mg/L (SD: 1.8), hardness ¼ 29 mg/L (SD: 1.4), total organic carbon (TOC) ¼ 6.80 mg/L (SD: 2.6), dissolved organic carbon (DOC) ¼ 6.22 mg/L (SD: 2.5), total coliform bacteria ¼ 1472 colonies/100 mL (SD: 120), E. coli ¼ 66 colonies/100 mL (SD: 8.1) and Heterotrophic Plate Count >5700 CFU/mL (SD: 750).

2.2.

Coliphage propagation

Type strains of somatic coliphages used in this study were PhiX174 (Microviridae), T4 (Myoviridae), T7 (Podoviridae), and T1 and Lambda (Siphoviridae). Phages and their bacterial hosts were obtained from the Felix d’He´relle Reference Center for Bacterial Viruses, University of Laval, Canada. T1, T4, and T7 were propagated in host E. coli B. Hosts for PhiX174 and Lambda were E. coli C and E. coli K12S Lederberg, respectively. Host strains were grown in tryptic soy broth (TSB; Difco). Coliphages were propagated in the appropriate host strain in TSB on a shaker platform (100 RPM) overnight at 36
C. Then, the broth cultures were vigorously mixed with fluorocarbon (Freon) (1:1) for 2 min and then centrifuged at 2600 g for 15 min at 4
C. The semi-purified virus supernatant was retained and further purified by filtering using a 0.22 mm pore size syringe filter, retaining the filtrate for use as test coliphage stock. Host bacteria and coliphage stocks were stored at 80
C. Phage infectivity titers were determined by single agar layer plaque assay (SAL, EPA Method 1602, 2001). For an inactivation experiment, each type strain of somatic coliphage was harvested in PBS as the top agar layer having confluent host cell lysis and the mixture was extracted with an equal volume of Freon. After centrifugation at 2600 g for 15 min at 4
C, the supernatant was archived and dispersed by microporous polycarbonate filtration through 0.2 and 0.08 mm pore size filters in succession. Filters were pretreated with 0.1% Tween 80 solution and then rinsed with reagent water to remove excess Tween 80 prior to filtration of virus suspensions.

2.3. Survival of type strains of coliphage families in seeded water Survival tests in water were conducted using representative type strains of somatic coliphages from each of the 4 major taxonomic groups. For each strain tested, virus stock was spiked into 40 mL of test water to give an initial concentration of about 107e108 PFU/mL. A positive control sample for measuring the initial virus concentration in the spiked water at time 0 was taken and assayed for virus infectivity immediately after spiking. All samples were held in tubes with screw caps to prevent evaporation over the duration of the

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 2 3 e3 7 3 4

3725

experiments. One aliquot of test water was held at room temperature (23e25
C), and one was held at refrigerator temperature (4
C). At each of a series of time intervals from 0 to 90 days, samples were taken and assayed for virus infectivity by SAL using the appropriate E. coli host. Duplicate samples were assayed at each time point.

regression and non-linear regression models were used for estimation of virus inactivation. For the non-linear regression model, R-square (goodness of fit) values were considered to choose the models. An exponential model, a second order polynomial model, and one-phase decay model were used for non-linear regression modeling in this study.

2.4. Inactivation of somatic coliphages by the physical agents of UV radiation and heat

3.

A collimated beam UV apparatus (custom-made) emitting monochromatic UV radiation at 254 nm was used for UV inactivation experiments. This apparatus is wellcharacterized and has been used in previous studies (Shin et al., 2003, 2005). A calibrated radiometer (International Light IL500) was used for measuring UV radiation in a 60  15mm petri-dish located below the lamp source and after UV beam collimation. Purified and dispersed somatic coliphage type strains in 5 ml volumes of PBS containing 105e106 PFU/ml in 60  15-mm petri dishes were irradiated with the UV collimated beam system with slow stirring at room temperature. Average low-pressure UV lamp intensity (mW/cm2) was measured and corrected to provide average incident irradiance. To correct the cross-sectional irradiance distribution of the collimated beam apparatus, low-pressure UV lamp intensity at the surface of the petri dish was measured (approximately 0.125 mW/cm2) to provide average incident irradiance. The measured petri factor was approximately 0.958 and the depth of virus suspension was approximately 0.32 cm. Target UV doses (or fluencies) were calculated by the product of average fluence rate and time. The average irradiance was calculated from the LamberteBeer Law. Samples were taken from the UV irradiation system after calculated exposure times corresponding to specified UV doses and their infectivity was determined. Purified and dispersed somatic coliphage type strains were also investigated for heat inactivation at the following two time and temperature conditions in reagent water (pH ¼ 7.0): Condition 1 ¼ 55
C for 1 h, representing a temperature achieved by solar disinfection in clear plastic bottles, and Condition 2 ¼ 63
C for 40 min, representing a low temperature pasteurization condition achieved by solar disinfection in opaque containers such as solar cookers (Sobsey, 2002). These temperatures were based on the higher one selected because it is considered a standard pasteurization temperature for time periods in the tens of minutes and the lower one being selected because it is considered in the range of thermophilic temperatures used for biological treatment processes and is in the range of temperatures tolerated by thermophilic microbes (thermophiles).

2.5.

Statistical analysis

For each type strain of somatic coliphage, regression analysis was conducted on their data for analysis of survival and inactivation kinetics using several different models in GraphPad Prism 5 (GraphPad, San Diego, CA), SAS (8.2), and Excel 2003 (Microsoft Corp.). The mathematical models to estimate virus inactivation were selected on the basis of best curve fitting by the software used to fit the data. Both linear

Results

3.1. Survival of coliphage type strains of families in seeded test waters Viruses were seeded at initial concentrations sufficient to observe 5e6 log10 reduction over time. Survival of somatic coliphage type strains in seeded waters over 90 days at two temperatures is shown in Figs. 1e4 for reagent and surface waters, respectively. Data are presented with trend lines and their 95% confidence intervals. The same data are presented in supplemental tables with mean data points and their 95% confidence limits for each time point. In reagent water at 4
C (Fig. 1), T1, T4, phiX174 and Lambda showed no infectivity titer reduction and T7 showed approximately 1 log10 reduction over 90 days. At room temperature (Fig. 2), the inactivation rates of somatic coliphage strains differed. Somatic coliphage infectivity titer reductions over 90 days were T1 and T4 < 1 log10, Lambda and PhiX174 ¼ 2 log10, and T7 ¼ 1 log10. Regression analysis on coliphage survival in reagent grade water at 4 and 25
C is shown in Figs. 1 and 2 as trend lines and their 95% confidence bands, with symbols showing the observed survival data of each somatic coliphage type strain over time. Inactivation of somatic coliphage type strains in reagent grade water was log-linear at both temperatures by regression analysis. Regression analysis of reagent grade water at 4
C showed that infectivity of T7 declined by approximately 0.06 log10 per week and the slope was significantly non-zero ( p ¼ 0.0016). However, the slopes of T1, T4, PhiX174, and Lambda were not significantly different from zero by linear regression analysis. Regression analysis on reagent grade water at 25
C showed that T7 and Lambda declined by 0.1 log10 per week and PhiX174 declined by 0.2 log10 per week. The slopes of the regression lines for T7, Lambda and PhiX174 were significantly non-zero ( p < 0.0001). However, the slopes of T1 and T4 were not significantly different from zero. Figs. 3 and 4 show the results for somatic coliphage type strain reductions in surface water. At 4
C (Fig. 3), T4, PhiX174, and Lambda strains showed less than 1 log10 reductions over 90 days, while T1 and T7 showed approximately 6 log10 and 4 log10 reductions, respectively. For T1 and T7 at 4
C, there were progressive declines in infectivity that followed first order kinetics over 90 days. By linear regression analysis of infectivity data in surface water at 4
C, T1 declined by approximately 0.5 log10 per week and T7 declined by approximately 0.4 log10 per week. The slopes of the first-order regression lines for T1 and T7 were significantly non-zero ( p < 0.0001). For PhiX174 and Lambda, regression analysis showed that inactivation rates of each strain were approximately 0.04 and 0.02 log10 per week, respectively. Also, the regression line slopes of PhiX174 ( p < 0.0001) and Lambda ( p ¼ 0.0092) were

3726

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 2 3 e3 7 3 4

T4 0

-2

-2

Log10 (Nt/N0)

Log10 (Nt/N0)

T1 0

-4 -6

-4 -6 -8

-8 0

20

40

60

80

0

100

20

40

T7

80

100

PhiX174 0

Log10 (Nt/N0)

0

Log10 (Nt/N0)

60

day

day

-2

-4

-2

-4

-6

-6 0

20

40

60

80

100

day

0

20

40

60

80

100

day

Lambda

Log10 (Nt/N0)

0

-2

-4

0

20

40

60

80

100

day

Fig. 1 e Log10 reductions (Nt/N0) of somatic coliphages (T1, T4, T7, PhiX174, and Lambda) over 90 days in reagent water at temperatures of 4
C. Points [ observed data; Error bars [ 95% confidence limits for data points.

significantly different from zero. However, the slope of the regression line for T4 was not significantly different from zero. As shown in Fig. 4, there was a decline in the titer of all strains over 90 days at room temperature. After two months, all somatic coliphages tested showed 4e6 log10 reductions. However, compared to the other somatic coliphages tested, T1 and T7 were inactivated most rapidly in surface water at room temperature. T1 experienced 5 log10 reduction after 28 days, and T7 experienced 4 log10 reduction after 14 days. The inactivation kinetics in surface water at 25
C showed that rates of infectivity decline were not first order in all tested strains (Fig. 4). For the results of test conditions that appeared not to be first order, kinetic analysis was done using non-linear regression models for all coliphage strains. For inactivation kinetics of T1 (R-square ¼ 0.9564) and T7 (R-square ¼ 0.9894), regression analysis was conducted using an exponential model with one phase of decline. Also, for T4, regression analysis was conducted using an exponential model with a plateau followed by one phase of exponential decay (R-square ¼ 0.9957). The decline of the infectivity of T4 appeared to begin after one month of incubation at 25
C and

continued gradually during the remaining 2-month duration of the survival test. There was no infectivity decline of T4 in the other test conditions, besides surface water at 25
C. The microbial activity in surface water at 25
C may cause the decline of T4 infectivity and accelerate inactivation, although this mechanism was not specifically investigated in this study. For PhiX174 (R-square ¼ 0.9680) and Lambda (Rsquare ¼ 0.9845), regression analysis was done using a second order polynomial model. Compared to T1 and T7, T4, the infectivity of PhiX174 and Lambda declined more slowly and monotonically over the time. Comparisons of the infectivity reductions by strains for each water type (reagent grade water and surface water) and temperatures (4 and 25
C) according to incubation times (days) are given in Fig. 5. Multiple regression analysis was conducted using SAS (8.2) to determine which variables were significant predictors of inactivation. Regression analysis was performed to determine if there were interaction effects between water type and temperature for inactivation of each virus type. R-square values from regression analysis increased from 0.538 (model without interaction variable) to 0.854

3727

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 2 3 e3 7 3 4

T4

T1 0

Log10 (Nt/N0)

Log10 (Nt/N0)

0

-2

-4

-2

-4

-6

-6 0

20

40

60

80

0

100

20

40

T7

80

100

PhiX174

0

0

Log10 (Nt/N0)

Log10 (Nt/N0)

60

day

day

-2

-4

-6

-2

-4

-6 0

20

40

60

80

100

day

0

20

40

60

80

100

day

Lambda

Log10 (Nt/N0)

0

-2

-4

-6 0

20

40

60

80

100

day

Fig. 2 e Log10 reductions (Nt/N0) of somatic coliphages (T1, T4, T7, PhiX174, and Lambda) over 90 days in reagent water at temperatures of 25
C. Points [ observed data; Error bars [ 95% confidence limits for data points.

(model with interaction variable) confirming that interaction between temperature and water type is a stronger predictor of inactivation than either variable alone. From this analysis, virus type, water type, temperature, and incubation time were significant predictors of inactivation ( p < 0.0001). Although virus type was a significant predictor overall, when each virus type is considered separately, PhiX174, Lambda, T1, and T7 were significant predicative variables ( p < 0.0001), whereas T1 was not significant ( p ¼ 0.3661). When interactions between individual virus types and water types were assessed, the interactions between water type and each individual virus were significant (for PhiX174, Lambda, T4, and T7, P < 0.001 and for T1, p ¼ 0.0286). When interactions between individual virus types and temperatures were assessed, the interactions between temperature and each individual virus were not significant ( p > 0.05). Therefore, water type is a more important predictor than temperature for virus inactivation. Table 1 provides the predicted times for 90%, 99%, 99.9%, and 99.99% (1, 2, 3 and 4 log10) reduction in days of all somatic

coliphage type strains in both water types and at both temperatures by regression analysis. For all tested coliphage strains, the required time for 90% reduction in reagent grade water was longer than in surface water. Also, the time for 90% reduction was longer at 4
C than at 25
C for most coliphage strains in both water types. However, for both T1 and T4 at both temperatures, the predicted time required for 99.99% (4 log10) inactivation in reagent water was over 1 year. At 4
C in surface water, strains T1 and T7 required 68 and 73 days for 99.99% (4 log10) inactivation respectively, while the other three strains were estimated to require over 1 year for such inactivation. Also, there were differences among virus types in predicted times for decimal infectivity reduction. At 25
C in surface water, the predicted times for infectivity reductions differed among virus types but were generally in two groups, with one group, consisting of T4, phiX174 and Lambda, being more thermotolerant and the other group, consisting of T1 and T7, being more thermolabile. At 25
C in surface water, 99.99% (4 log10) infectivity reductions of T4, phiX174, and Lambda required longer times of 47e53 days than T1 and T7, which required only 9 and 10 days, respectively.

3728

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 2 3 e3 7 3 4

T1

T4 0

Log10 (Nt/N0)

Log10 (Nt/N0)

0 -2 -4

-2

-4

-6 -6

-8 0

20

40

60

80

0

100

20

40

day

60

80

100

day

PhiX174

T7 0

Log10 (Nt/N0)

Log10 (Nt/N0)

0 -2 -4

-2

-4

-6 -6

0

20

40

60

80

day

0

20

40

60

80

100

day

Lambda

Log10 (Nt/N0)

0

-2

-4

-6 0

20

40

60

80

100

day

Fig. 3 e Log10 reductions (Nt/N0) of somatic coliphages (T1, T4, T7, PhiX174, and Lambda) over 90 days in surface water at temperatures of 4
C. Points [ observed data; Error bars [ 95% confidence limits for data points.

3.2. Inactivation of somatic coliphages by the physical agents of UV radiation and heat The type strains of the different somatic coliphage families were further tested to determine their inactivation kinetics by UV radiation and those with the slowest inactivation rates in water at 25
C were further tested for inactivation by heat at temperatures of 55 and 63
C. In Fig. 6a are presented the inactivation kinetics of five dispersed coliphage type strains representing different taxonomic groups exposed to monochromatic UV radiation of 254 nm wavelength in a collimated beam apparatus. Using regression analysis, both first order and non-first order models were explored to characterize inactivation kinetics. Due to the limited amount of UV inactivation data and its relatively rapid inactivation, only first order kinetic modeling was applied to coliphage T4. For characterizing the inactivation kinetics of the other coliphages, a one-phase decay model was better, increasing the R-square values from 0.3646 to 0.9998 for T1, from 0.9562 to 0.9941 for T7, from 0.8663 to 0.9435 for PhiX174, and from 0.0901 to 0.9981 for Lambda.

T4 (Myoviridae) showed the highest UV sensitivity, with greater than 6 log10 inactivation at a dose of less than 10 mJ/ cm2. Among tested strains, T4 strain has the largest genome size (w166 kb) compared to the other strains (5e48 kb) and is a double-stranded DNA virus. These physical characteristics of T4 might influence its inactivation rate by UV radiation as it represents the largest nucleic acid “target”. Coliphages phiX174 (Microviridae), T1 (Siphoviridae), Lambda (Siphoviridae), and T7 (Podoviridae) showed non-log-linear regression kinetics (one-phase decay) for inactivation by UV radiation, while T4 (Myoviridae) showed log-linear regression kinetics. Estimated UV doses as mJ/cm2 for 99.9% inactivation were 8 for T1 and T7, 3 for T4, 10 for phiX174, and 18 for Lambda, respectively (Table 2). Both PhiX174 and T7 exhibited retardant inactivation kinetics, with generally declining UV inactivation rates at higher UV doses. As shown in Fig. 6b, inactivation of dispersed preparations of some type strains of somatic coliphages by heat at 55
C for 1 h (condition 1) was modest and differed among them, with log10 inactivation of T7 (0.8) > PhiX174 (0.4) > T4 (0.1). At 63
C for 40 min (condition 2), all viruses were inactivated more

3729

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 2 3 e3 7 3 4

T7 2

0

0

Log10 (Nt/N0)

Log10 (Nt/N0)

T1 2

-2 -4 -6

-2 -4 -6

-8

-8

0

20

40

60

0

10

day T4

30

PhiX174

2

2

0

0

Log10 (Nt/N0)

Log10 (Nt/N0)

20

day

-2 -4 -6

-2 -4 -6

-8

-8 0

20

40

60

80

0

day

20

40

60

80

day

Lambda 2

Log10 (Nt/N0)

0 -2 -4 -6 -8 0

20

40

60

80

day

Fig. 4 e Log10 reductions (Nt/N0) of somatic coliphages (T1, T4, T7, PhiX174, and Lambda) over 90 days in surface water at temperatures of 25
C. Points [ observed data; Error bars [ 95% confidence limits for data points.

extensively than at condition 1, with log10 inactivation of PhiX174 (2.3) > T7 (2.1) > T4 (0.25). T4, T7. Coliphage phiX174 experienced less than 1 log10 reduction at heat condition 1, and T4 showed less than 1 log10 reduction at heat condition 2. Overall, T4 was the most heat-resistant virus at both conditions tested. The results of studies comparing the different strains of somatic coliphages representing different families for their resistance against two physical agents, UV radiation and heat, showed that T4 of the Myoviridae family was least resistant to UV radiation but most resistant to heat. Lambda of the Siphoviridae family and PhiX174 of the Microviridae family showed higher resistance to UV radiation than did the other coliphages tested. It is noteworthy that PhiX174 not only has a very small genome (w5 kb) compared to the other viruses tested (42e166 kb) but is also the only virus of those tested that has circular genomic DNA. The others have linear, doublestranded DNA. Lambda has a relatively small genome size (w48 kb) compared to the other double-stranded DNA strains in this study. However, the extent to which the genetic and structural characteristics of the test viruses contribute to their

resistance to the physical effects of UV radiation and heat are unknown. Nevertheless, the observations that PhiX174, Lambda, and T4 were relatively resistant to UV radiation or heat suggests that these viruses and their families may be among the more persistent somatic coliphages in ambient waters subjected to physical environmental stressors such as UV radiation and high temperature.

4.

Discussion

There are many aspects of somatic coliphage biology and ecology that need to be understood if these viruses are to be used as viral indicators of fecal contamination in water. The results of this study show that for several factors, namely survival in water and response to the physical inactivating agents of heat and UV radiation, different somatic coliphage strains representing different virus families differ in their inactivation responses when exposed to these aquatic conditions and inactivating agents. These results have important implications for the use of these phages and their

3730

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 2 3 e3 7 3 4

T4

0

0

-2

-2

Log10 (Nt/N0)

Log10 (Nt/N0)

T1

-4

-4

-6

-6

-8

-8 0

20

40

60

80

0

100

20

40

T7

80

100

PhiX174

0

0

-2

-2

Log10 (Nt/N0)

Log10 (Nt/N0)

60

Day

Day

-4

-6

-4

-6

-8

-8 0

20

40

60

80

100

Day

0

20

40

60

80

100

Day

Lambda

0

Log10 (Nt/N0)

-2

-4

-6

-8 0

20

40

60

80

100

Day

Fig. 5 e Log10 reductions (Nt/N0) of somatic coliphages (T1, T4, T7, PhiX174, and Lambda) over 90 days (C: Reagent water at 4
C, -: Reagent water at 25
C, :: Surface water at 4
C, ;: Surface water at 25
C).

3731

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 2 3 e3 7 3 4

Table 1 e Predicted times (days) for infectivity reduction of somatic coliphage type strains (T1, T4, T7, PhiX174 (4X174) and Lambda (l)) at 4
C and 25
C in (a) reagent grade water and (b) in surface water. 25
C

Reduction Log10 (Nt/N0))

(a) 1(90%) 2(99%) 3(99.9%) 4(99.99%) (b) 1(90%) 2(99%) 3(99.9%) 4(99.99%)

4
C

T1

T4

T7

4X174

l

T1

T4

T7

4X174

l

>365 >365 >365 >365

>365 >365 >365 >365

68 144 221 297

35 73 111 150

55 104 153 203

>365 >365 >365 >365

>365 >365 >365 >365

124 232 339 >365

>365 >365 >365 >365

>365 >365 >365 >365

2 4 6 9

34 37 42 47

2 4 6 10

15 28 38 48

19 34 44 53

22 37 51 68

>365 >365 >365 >365

22 38 57 73

122 285 >365 >365

288 >365 >365 >365

families as fecal indicator viruses. There are several criteria for the selection of an ideal indicator microorganism for detection of fecal contamination in water. One is that an ideal indicator should have survival characteristics similar to pathogens of interest. The data provided here on the survival and persistence of representatives of different taxonomic groups of somatic coliphages help determine which ones are most likely to persistent in water and wastewater and therefore best indicate the presence of fecal contamination. However, further studies would be needed comparing these representatives of different taxonomic groups of somatic coliphages to specific pathogens or pathogen groups to better characterize their usefulness and applicability as either indicators of the environmental persistence or treatment responses of specific pathogens or pathogen groups or as more general indicators of fecal contamination. The observations of comparative somatic coliphage survival in seeded waters suggest that Myoviridae as represented by T4, Microviridae as represented by phiX174, and Siphoviridae as represented by Lambda are potential candidates to serve as indicators of the more persistent human viruses of fecal origin in water, such as hepatitis A virus, adenoviruses and some enteroviruses. Based on their slower inactivation rates in water over time, the representative strains of these three somatic coliphage families were relatively persistent. At 4
C, their infectivity declined more slowly than did that of T1 (Siphoviridae) and T7 (Podoviridae). For T4 (Myoviridae), PhiX174 (Microviridae) and Lambda (Siphoviridae), there was less than 1 log10 inactivation over 90 days at 4
C in both reagent grade water and surface water. In surface water at room temperature, these three candidate strains representing their families also persisted longer than did T1 and T7. However, these three somatic coliphage strains started to decline in infectivity after one month in surface water, with a continuous infectivity decline over the remaining two months of the experimental period. For the first month, however, these three strains were much more persistent in water than were than T1 and T7. Also, when regression analysis was performed to statistically determine if there were interaction effects between water type and temperature for each virus type, the interactions between water type and each individual virus were significant for PhiX174, Lambda, T4, and T7 ( p < 0.001) and for T1 ( p ¼ 0.0286), while the interactions between temperature and each individual virus

were not significant ( p > 0.05). Therefore, water type was a more important predictor than temperature for estimating reductions of these viruses in this study. The waters used for survival testing in this study were not sterilized, and contained viable cellular microorganisms, primarily bacteria. The concentrations of these microorganisms in test water increased during the survival study (data not shown), suggesting that the microbial activity in test water may affect the rates of viral inactivation. Except for measuring cellular microbe concentrations over time, other water quality parameters were not monitored for changes over time. Therefore, besides the possible effects of increased concentrations of cellular microbes such as bacteria and perhaps their excreted or liberated products in test waters over time, we are unable to attribute differences in virus inactivation to changes or differences in other specific water quality parameters over time. However, the effects of different water quality parameters on virus inactivation in water and other environmental media are well documented from previous studies even though they were not the specific focus of attention for this study (Sattar, 1981; Sobsey, 1986; Gerba, 2007). Although only one prototype strain in each family except one (two Siphoviridae were included) was investigated in this study, if these strains are representative of the survival of other members in their family, their families have potential as candidate somatic coliphage indicators of virological water quality. It is recommended that somatic coliphages be directly compared to Fþ RNA and Fþ DNA coliphages for their survival in water. The Fþ coliphages are more commonly recommended candidate indicators for sewage contamination and enteric viruses in water, and previous studies have characterized their survival in different water matrices in comparison to the survival of human enteric viruses. Chung and Sobsey (1993) showed that Fþ coliphages were inactivated faster than hepatitis A virus, poliovirus, and rotavirus in sea water in warm (summer time) conditions. Yahya et al. (1993) compared the survival of bacteriophages MS-2 (Fþ RNA; Leviviridae family) and PRD-1 (somatic coliphage; Podoviridae family) in groundwater and found that PRD-1 was more persistent than MS-2 at higher water temperatures. Brion et al. (2002) compared the inactivation kinetics of prototype strains of Fþ RNA coliphages in natural surface waters and showed that inactivation rates differed by strain. Of the Fþ RNA

3732

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 2 3 e3 7 3 4

(b)

(a)

3

0

Log10 (Nt/N0)

-2

-4

2.5

Log reduction

T1 T4 T7 PhiX174 Lambda

2 T4 1.5

T7 PhiX174

1

-6

0.5

0 1(55deg.,1hr)

-8 0

10

20

30

40

50

UV Dose (mJ/cm2)

2(63deg.,40min) Condition

Fig. 6 e (a) Log10 inactivation (Nt/N0) of somatic coliphages in buffered water by monochromatic UV radiation. Points [ observed data; lines [ prediction from regression analysis (C: T1, -: T4, :: T7, ;: PhiX174, A: Lambda). (b) Inactivation of selected somatic coliphages by heat treatment at two conditions: (1): 55
C for 1 h and (2) 63
C for 40 min.

phages MS2, GA, GB, F1, and SP, representing genogroups GI, GII, GIII and GIV, MS2 (GI) was inactivated most rapidly, declining by 7 log10 in two weeks (14 days), while the other Fþ RNA coliphages were inactivated completely after 36 days (about 5 weeks) in natural waters. Taken together, previous results suggest that Fþ coliphages are inactivated more quickly than enteric viruses at higher water temperatures. Based on the results of this present study and as compared to the results of other candidate coliphages that were primarily or exclusively Fþ in previous studies, somatic coliphages are likely to be more persistent in water than Fþ RNA coliphages. Therefore, based on their persistence, somatic coliphages may have advantages to Fþ coliphages as more environmentally persistent indicators of enteric viruses in water. In previous studies the Myoviridae family was abundant in human sewage and the Siphoviridae family predominated in surface waters (Ackermann and Nguyen, 1983; Pedroso and Martins, 1995; Muniesa et al., 1999). The results of this present somatic coliphage survival study are consistent with these findings if survival in water and wastewater is an

Table 2 e Predicted UV dose (mJ/cm2) for decimal inactivation of somatic coliphages T1, T4, T7, phiX174 and Lambda. Reduction Log10 (Nt/N0)

T1

T4a

T7

PhiX174

Lambda

1(90%) 2(99%) 3(99.9%) 4(99.99%)

2 5 8 11

<1 2 3 4

2 5 8 12

2 5 10 18

6 12 18 24

a Linear regression analysis was used to predict log10 inactivation. For the other strains, non-linear regression analysis (one-phase decay) was used to predict log10 inactivation.

important factor contributing to their presence. Previous studies have also explored the response of different Fþ and somatic coliphages to disinfection processes, such as chlorination and UV irradiation. Duran et al. (2003) found that isolates belonging to the Siphoviridae family were the most resistant to chlorination compared to enteroviruses and E. coli. However, in this present study, two Siphoviridae family type strains, T1 and Lambda, showed different inactivation kinetics in seeded water, suggesting that the inactivation of Siphoviridae may differ among strains. Because only one type strain of the other somatic coliphages families was investigated in this study, it is recommended that future studies investigate the inactivation kinetics of additional strains from these families to determine if they differ or are the same. In this study, type strains of somatic coliphages showed relatively high UV sensitivity. Regression analysis on predicted UV dose for various log10 reductions (Table 2) showed that all type strains of somatic coliphages were inactivated by 4 log10 at doses lower than 25 mJ/cm2. Retardant UV inactivation kinetics with increasing UV doses was observed for both PhiX174 and T7. The reasons for these kinetics were not investigated in this study, but they could be due to the presence of a small fraction of aggregated viruses despite efforts to disperse the viruses by limited pore size filtration. Based on comparisons to previous work with Fþ RNA coliphages and adenoviruses, all somatic coliphage strains tested showed greater UV sensitivity than does MS2, a male-specific coliphage of the Leviviridae family (approximately 2 log reduction at a dose of 30 mJ/cm2) (Shin et al., 2003), and Adenovirus 2, a member of the Adenoviridae family, (approximately 4 log reduction at a dose of 120 mJ/cm2) (Shin et al., 2005). The taxonomic diversity of the somatic coliphage group makes it challenging to study their usefulness as viral indicator microorganisms. Even though only a few type strains

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 2 3 e3 7 3 4

representative of the different somatic coliphage families were investigated in these studies on survival in water, the results of this study suggest that phages belonging to the Microviridae, Myoviridae, and Siphoviridae family are the most persistent in water. Also, compared to previous studies on the survival of Fþ coliphages (Chung and Sobsey, 1993; Yahya et al., 1993; Brion et al., 2002), these somatic coliphage families also showed slower and less extensive inactivation at higher temperatures in environmental surface water. Further study of comparative survival and inactivation of representative members of individual families of somatic coliphages, including comparisons to human enteric virus survival, would provide better evidence of these somatic coliphages as possible viral indicators of fecal contamination and human enteric viruses in environmental waters and such studies are recommended.

5.

Conclusions

The results of studies on prototype somatic coliphage survival in seeded waters suggest that based on their greater persistence in water over time and their responses to UV radiation and heat, T4 in the Myoviridae, PhiX174 in the Microviridae, and Lambda in the Siphoviridae are possible candidates as fecal indicator viruses in water due to their greater survival compared to T1 in the Siphoviridae and T7 in the Podoviridae. Results of studies of two Siphoviridae family type strains, T1 and Lambda, suggest that members of this family may differ in their survival and persistence in water. By regression analysis, virus type, water type, temperature, and incubation time were significant predictors of inactivation. Although virus type was a significant predictor of virus inactivation overall, when each virus type was considered separately, PhiX174, Lambda, T1, and T7 were significant predictor variables of inactivation, whereas T1 was not significantly predictive. By regression analysis with interaction variables, the interactions between water type and each individual virus were statistically significant, whereas the interactions between temperature and each individual virus were not statistically significant. Therefore, water type was a more important predictor than temperature for estimating virus reductions in water based on relative comparison of water type and temperature by regression analysis with an interaction variable. However, both water type and temperature themselves could be significant predictors for virus inactivation. The water temperatures investigated in this study were not fully representative of those of all ambient waters. Therefore, the role of temperature on coliphage survival over a wider range of temperatures is recommended for future study to better understand and quantify the contribution of this variable to coliphage inactivation in water.

Acknowledgments This work was supported by grants from the National Water Research Institute (NWRI; Fellowship to H. Lee) and K-water (Korea Water Resources Corporation). We are grateful to Doug

3733

Wait for laboratory and technical support and Prof. Lisa Casanova, Georgia State University, for assistance with data and statistical analysis.

Appendix. Supplementary data Supplementary data associated with this article can be found in online version at doi:10.1016/j.watres.2011.04.024.

references

Ackermann, H.W., Nguyen, T.M., 1983. Sewage coliphages studied by electron microscopy. Appl. Environ. Microbiol. 45, 1049e1059. Araujo, R.M., Puig, A., Lasobras, J., Lucena, F., Jofre, J., 1997. Phages of enteric bacteria in fresh water with different levels of faecal pollution. J. Appl. Microbiol. 82, 281e286. Armon, R., 1993. Bacteriophage monitoring in drinking water: do they fulfill the index or indicator function? Water Sci. Technol. 27, 463e470. Brion, G.M., Meschke, J.S., Sobsey, M.D., 2002. F-specific RNA coliphages: occurrence, types, and survival in natural waters. Water Res. 36, 2419e2425. Chung, H., Sobsey, M.D., 1993. Comparative survival of indicator viruses and enteric viruses in seawater and sediment. Water Sci. Technol. 27, 425e428. Cole, D., Long, S.C., Sobsey, M.D., 2003. Evaluation of Fþ RNA and DNA coliphages as source-specific indicators of fecal contamination in surface waters. Appl. Environ. Microbiol. 69 (11), 6507e6514. Colford Jr., J.M., Wade, T.J., Schiff, K.C., Wright, C.C., Griffith, J.G., Sandhu, S.K., Burns, S., Hayes, J., Sobsey, M., Lovelace, G., Weisberg, S.B., 2007. Water quality indicators and the risk of illness in non-point source impacted recreational waters. Epidemiology 18 (1), 27e35. Dore, W.J., Henshilwood, K., Lees, D.N., 2000. Evaluation of Fspecific RNA bacteriophages as a candidate human enteric virus indicator for Bivalve Molluscan shellfish. Appl. Environ. Microbiol. 66 (4), 1280e1285. Duran, A.E., Muniesa, M., Moce-Llivina, L., Campos, C., Jofre, J., Lucena, F., 2003. Usefulness of different groups of bacteriophages as model micro-organisms for evaluation chlorination. J. Appl. Microbiol. 95, 29e37. EPA Method 1602, 2001. Male-Specific (Fþ) and Somatic Coliphage in Water by Two-step Enrichment Procedure EPA-821-R01e029, April, 2001, Washington, DC. Garcia-Aljaro, C., Munoz-Berbel, X., Jenkins, A.T.A., Blanch, A.R., Munoz, F.X., 2008. Surface plasmon resonance assay for realtime monitoring of somatic coliphages in wastewaters. Appl. Environ. Microbiol. 74 (13), 4054e4058. Gerba, C.P., 2007. Virus occurrence and survival in the environmental waters chapter 5. In: Bosch, A. (Ed.), Human Viruses in Water. Perspect. Med. Virol, vol. 17, pp. 91e108. Havelaar, A.H., van Olphen, M., Drost, Y.C., 1993. F-specific RNA bacteriophages are adequate model organisms for enteric viruses in fresh water. Appl. Environ. Microbiol. 59, 2956e2962. Jofre, J., Olle, E., Ribas, F., Vidal, A., Lucena, F., 1995. Potential usefulness of bacteriophages that infect Bacteriodes fragilis as model organisms for monitoring virus removal in drinking water treatment plants. Appl. Environ. Microviol 61, 3227e3231. Jofre, J., 2008. Is the replication of somatic coliphages in water environmental significant? J. Appl. Microbiol. 106, 1059e1069.

3734

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 2 3 e3 7 3 4

Love, D.C., Sobsey, M.D., 2007. Simple and rapid Fþ coliphage culture, latex agglutination, and typing assay to detect and source track fecal contamination. Appl. Environ. Microbiol. 73 (13), 4110e4118. Love, D.C., Vinje, J., Khalil, S.M., Murphy, J., Lovelace, G.L., Sobsey, M.D., 2007. Evaluation of RT-PCR and reverse line blot hybridization for detection and genotyping Fþ RNA coliphages from estuarine waters and molluscan shellfish. J. Appl. Microbiol. 104, 1203e1212. Moce-Llivina, L., Lucena, F., Jofre, J., 2005. Enteroviruses and bacteriophages in bathing waters. Appl. Environ. Microbiol. 71 (11), 6838e6844. Muniesa, M., Lucena, F., Jofre, J., 1999. Study of the potential relationship between the morphology of infectious somatic coliphages and their persistence in the environment. J. Appl. Microbiol. 87 (3), 402e409. Muniesa, M., Jofre, J., 2007. The contribution of induction of temperate phages to the numbers of free somatic coliphages in waters is not significant. FEMS Microbiol. Lett. 270, 272e276. Pedroso, D.M.M., Martins, M.T., 1995. Ultra-morphology of coliphages isolated from water. Water Res. 29, 1199e1202. Sattar, S.A., 1981. Virus survival in receiving waters. In: Goddard, M., Butler, M. (Eds.), Viruses & Wastewater Treatment. Pergamon Press, Toronto, ON, pp. 91e108.

Shin, G.A., Linden, K.G., Arrowood, M.J., Sobsey, M.D., 2003. Lowpressure UV inactivation and DNA repair potential of Cryptosporidium parvum oocysts. Appl. Environ. Microbiol. 67 (7), 3029e3032. Shin, G.A., Linden, K.G., Sobsey, M.D., 2005. Low pressure ultraviolet inactivation of pathogenic enteric viruses and bacteriophages. J. Environ. Eng. Sci. 4, S7eS11. Sobsey, M.D., 1986. Transport and Survival of Microbial Pathogens in On-site Sewage Treatment Systems. Presentation at ’I’hc On-Site Sewage Disposal Conference. State University, N. C. Sobsey, M.D., Battigelli, D., Handzel, T., Schwab, K., 1995. Malespecific coliphages as indicators of viral contamination of drinking water. In: Proceedings of the 1994 Water Quality Technology Conference American Water Works Association, pp. 1847e1862. Denver, Colo. Sobsey, M.D., 2002. Managing Water in the Home: Accelerated Health Gains from Improved Water Supply[Report]. World Health Organization, Geneva, Switzerland. Skraber, S., Gassilloud, B., Schwartzbrod, L., Gantzer, C., 2004. Survival of infectious Poliovirus-1 in river water compared to the persistence of somatic coliphage, thermotolerant coliforms and Poliovirus-1 genome. Water Res. 38, 2927e2933. Yahya, M.T., Galsomies, L., Gerba, C.P., Bales, R.C., 1993. Survival of bacteriophages MS-2 and PRD-1 in ground water. Water Sci. Technol. 27, 409e412.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 3 5 e3 7 4 3

Available at www.sciencedirect.com

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

The strong biocidal effect of free nitrous acid on anaerobic sewer biofilms Guangming Jiang a, Oriol Gutierrez a,b, Zhiguo Yuan a,* a

Advanced Water Management Centre, Gehrmann Building, Research Road, The University of Queensland, St. Lucia, Queensland 4067, Australia b Catalan Institute for Water Research (ICRA), Scientific and Technological Park of the University of Girona, Girona, Spain

article info

abstract

Article history:

Several recent studies showed that nitrite dosage to wastewater results in long-lasting

Received 20 January 2011

reduction of the sulfate-reducing and methanogenic activities of anaerobic sewer bio-

Received in revised form

films. In this study, we revealed that the quick reduction in these activities is due to the

18 April 2011

biocidal effect of free nitrous acid (FNA), the protonated form of nitrite, on biofilm

Accepted 18 April 2011

microorganisms. The microbial viability was assessed after sewer biofilms being exposed

Available online 22 April 2011

to wastewater containing nitrite at concentrations of 0e120 mg-N/L under pH levels of 5e7 for 6e24 h. The viable fraction of microorganisms was found to decrease substantially from

Keywords:

approximately 80% prior to the treatment to 5e15% after 6e24 h treatment at FNA levels

Free nitrous acid

above 0.2 mg-N/L. The level of the biocidal effect has a much stronger correlation with the

Biocidal

FNA concentration, which is well described by an exponential function, than with the

Sewer biofilm

nitrite concentration or with the pH level, suggesting that FNA is the actual biocidal agent.

Inhibition

An increase of the treatment from 6 to 12 and 24 h resulted in only slight decreases in

Nitrite

microbial viability. Physical disrupted biofilm was more susceptible to FNA in comparison

Sulfide

with intact biofilms, indicating that the biocidal effect of FNA on biofilms was somewhat

Methane

reduced by mass transfer limitations. The inability to achieve 2-log killing even in the case of disrupted biofilms suggests that some microorganisms may be more resistant to FNA than others. The recovery of biofilm activities in anaerobic reactors after being exposed to FNA at 0.18 and 0.36 mg-N/L, respectively, resembled the regrowth of residual sulfatereducing bacteria and methanogens, further confirming the biocidal effects of FNA on microorganisms in biofilms. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Hydrogen sulfide production and emission is a ubiquitous problem in sewer systems (Hvitved-Jacobsen, 2002; US EPA, 1974, 1991). Sulfate-reducing bacteria (SRB) proliferate in anaerobic sewers, reducing sulfate to sulfide. Sulfide is a major cause for sewer corrosion and odor, and is also a source of health hazard (Boon, 1995; Pomeroy, 1990; Thistlethwayte, 1972; US EPA, 1991; WERF, 2007). An

increasing trend of sulfide production in sewers has been observed over the last 20 years (WERF, 2007). Methane production in sewers has also been recognized recently. Field measurements conducted by Foley et al. (2009) and Guisasola et al. (2008, 2009) showed that a substantial amount of methane could be formed in pressure sewers. Given the low solubility of methane, it would be released at locations where a gas and liquid interface occurs including pumping stations, gravity sewers and also

* Corresponding author. Tel.: þ61 7 3365 4374; fax: þ61 7 3365 4726. E-mail addresses: [email protected] (G. Jiang), [email protected] (O. Gutierrez), [email protected] (Z. Yuan). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.026

3736

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 3 5 e3 7 4 3

Nomenclature COD FIA FNA IC

chemical oxygen demand flow injection analysis free nitrous acid ion chromatography

the inlet work of a wastewater treatment plant (GWRC, 2011). Methane is a potent greenhouse gas, whose global warming potential is about 21e23 times that of CO2 (IPCC, 2006). Methane emission in sewers could thus contribute significantly to the greenhouse gas emissions from wastewater systems (Guisasola et al., 2009; GWRC, 2011). Besides, methane is also an explosive gas with a lower explosive limit of approximately 5%. To control the production and release of hazardous sulfide in sewers, many chemical dosing strategies have been developed and used in practice (US EPA, 1992; Zhang et al., 2008). Commonly used chemicals include oxygen/air, nitrate, metal salts and alkali. These chemicals either remove dissolved hydrogen sulfide through oxidation (Bentzen et al., 1995; Gutierrez et al., 2008; Ochi et al., 1998), or precipitation reactions (Jameel, 1989; Padival et al., 1995), or reducing the transfer of molecular H2S from wastewater to air through pH elevation (Gutierrez et al., 2009; Yongsiri et al., 2005). Some of these strategies have also been found to reduce methane formation in sewers (Gutierrez et al., 2009; Mohanakrishnan et al., 2009; Zhang et al., 2009a). These traditional dosing strategies involve constant addition of chemicals to remove sulfide already formed. The constant dosing requirement causes large chemical consumption and high operational costs. It is favorable to reduce sulfide and methane production rather than to remove them after their production. To this end, more cost-effective strategies were under investigation using antimicrobial agents, including metabolic inhibitors and broad-spectrum biocides. Several biocidal agents have been reported to be effective in controlling sulfide and methane generation. Gardner and Stewart (2002) reported that a pulse dose of glutaraldehyde at 500 mg/L for 7 h completely stopped sulfide production by a mixed-culture biofilm. However, sulfide production resumed after 60 h Zhang et al. (2009b) found that sulfide generation in sewage was decreased by 90% by formaldehyde at 19 mg/L Greene et al. (2006) combined six broad-spectrum biocides (glutaraldehyde, bronopol, tetrakis hydroxymethyl phosphonium sulfate, benzalkonium chloride, cocodiamine, formaldehyde) with nitrite to inhibit sulfide production. Nitrite was found to be synergistic with biocides and the combination allows more efficient and costeffective control of SRB. As a metabolic inhibitor for SRB, molybdate was employed to control H2S production by a pure culture of SRB, in swine manure treatment, and in anaerobic digestors (Nemati et al., 2001; Predicala et al., 2008; Tanaka and Lee, 1997). Although many of the microbial biocides and inhibitors have been proven to be effective, they may impose adverse impacts on the environment due to their generic toxicity and low degradability. For the application in sewers,

N2O NO SAOB SRB VFA

nitrous oxide nitric oxide sulfide anti-oxidation buffer sulfate-reducing bacteria volatile fatty acids

the residual biocides in sewage may have detrimental effects on the microbial processes in the downstream wastewater treatment plants. Nitrite, known as a specific SRB inhibitor (Greene et al., 2003), has also been shown to be able to suppress both sulfide and methane production in lab-scale sewer reactors (Jiang et al., 2010; Mohanakrishnan et al., 2008). The continuous dosing of nitrite for four weeks was found to achieve lasting (1e2 months) reduction in sulfide and methane production in the sewer reactors. The lasting effectiveness was suggested to be due to the decreased or suppressed growth of sulfate-reducing and methanogenic populations in biofilms during the extended dosing period, and the recovery of these populations required 1e2 months. However, in the field trial conducted by Jiang et al. (2010), a long-lasting reduction in sulfide and methane production (for weeks and months, respectively) was achieved by dosing nitrite at 100 mg-N/L for only 33 h. The recovery of sulfide and methane production resembled the regrowth of sulfatereducing bacteria and methanogens. This could not be explained by the commonly observed inhibiting role of nitrite. Inhibitors like nitrite (sulfite analog) or molybdate (sulfate analog) interrupt the sulfate reduction pathway by blocking the dissimilatory sulfite or sulfate reductase (Greene et al., 2006, 2003). SRB may survive the inhibition if the exposure time is short, and resume their activity when the inhibitor is removed. Long-term application of an inhibitor is thus required in order to decrease the SRB population, through which to achieve long-lasting effectiveness. We hypothesize that the lasting sulfide and methane control effects achieved by a short dosage of nitrite was due to a biocidal rather than an inhibitory effect. The aim of this study is to investigate this potential biocidal effect of nitrite and its derivative free nitrous acid (FNA, the protonated form of nitrite) on anaerobic wastewater biofilms. Laboratory-scale anaerobic reactors were used to grow sewer biofilms with real wastewater. Intact and disrupted biofilm samples taken from the reactors were incubated in wastewater containing nitrite at concentrations of 0, 30, 60, 90, and 120 mg-N/L, at pH levels of 5, 6, 6.5, and 7, and for 6, 12, and 24 h. These combinations of nitrite and pH levels give rise to FNA concentrations of 0e3 mg-N/L. The abundance of viable microorganisms in biofilms both prior to and after the treatment was measured using a LIVE/DEAD staining assay, which assesses microbial viability by verifying cell membrane integrity. Simultaneous dosing of nitrite and acid was also directly conducted with the sewer reactors, during which the suppression and recovery of sulfide and methane production activities of the reactors were measured.

3737

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 3 5 e3 7 4 3

2.

Material and methods

2.1.

Anaerobic sewer biofilm reactors

Three lab-scale sewer reactors, namely R1, R2, and R3, were set up for growing biofilms with real wastewater under anaerobic conditions. These reactors were made of Perspex. Each reactor had a volume of 0.75 L, with a diameter of 80 mm and a height of 149 mm (Fig. S1). Plastic carriers (Anox Kaldnes, Norway) of 1 cm diameter were clustered on four stainless-steel rods inside each reactor to provide additional surfaces for biofilm growth and to allow sampling of intact biofilms (see below). The total volume of the carriers used for each reactor was about 15 mL (2% of the reactor volume). The total biofilm area in each reactor, including both the reactor wall and carrier surfaces, was approximately 0.05 m2. The area to volume ratio (A/V) was therefore 70.9 m2/m3. Domestic wastewater, collected weekly from a wet well in Brisbane, Australia, was stored in a cold room at 4  C. It was used as the feed to all reactors after being heated up to 20  C in a water bath before each pumping event. The sewage typically contained sulfide at concentrations of <3 mg-S/L, sulfate at concentrations between 10 and 25 mg-S/L, and volatile fatty acid (VFA) at 50e120 mg-COD/L. Nitrite was below detection limits in the fresh sewage. The reactors were fed with sewage through a peristaltic pump (Masterflex 7520-47) every 6 h. Every feed pumping event lasted for 2 min, delivering one reactor volume of sewage into each reactor. Mixing (200 rpm) was provided continuously with magnetic stirrers (Heidolph MR3000) to produce a moderate shear force, and also to avoid solids settling at the bottom. Batch tests were carried out regularly (every 1e2 weeks) to measure the sulfide and methane production rates of each reactor. The tests were started by pumping fresh sewage into reactors for 6 min (three hydraulic retention times) to ensure a thorough replacement of liquid in reactors with fresh sewage. Wastewater samples were taken at 0, 30, 60, 90, and 120 min after pumping, for the analysis of dissolved inorganic sulfur (sulfide, sulfite, thiosulfate, and sulfate) and dissolved

methane using methods to be further described below. The sulfide and methane production rates were calculated using linear regression of the sulfide and methane concentrations. The reactors were operated for 9 months to reach stable performance (as indicated by the sulfide and methane production rates) before the FNA tests described below commenced. The three reactors displayed very similar sulfide and methane production rates.

2.2.

Viability tests on the biocidal effect of FNA

2.2.1.

FNA toxicity to intact biofilms

Twelve sets of tests were carried out under conditions summarized in Table 1. In each set, 1 L of wastewater taken from the cold room was heated up to 20  C. The wastewater pH was then adjusted to the designated pH level (Table 1) with 1 M hydrochloric acid. The pH-adjusted wastewater was then used to fill up five 75 mL single-use sterile bottles (Sarstedt, Australia). Pre-determined amounts of a sodium nitrite stock solution (12 g-N/L) were added to the five bottles to achieve the five designated nitrite concentrations, i.e. 0, 30, 60, 90, and 120 mg-N/L (Table 1). For each set of tests, one more bottle was filled with fresh sewage (pH ¼ 7.6) as the control. A plastic carrier with attached biofilm was transferred from the biofilm reactors into each bottle. The bottle was then capped and kept anaerobic by avoiding air bubbles. Gentle mixing was provided by an orbital shaker at 60 rpm. The duration of incubation used in each set of tests is as described in Table 1. The pH level and nitrite concentration in each bottle were measured at the end of each test. For the 6- and 12-h tests, the nitrite concentration decreased marginally (<5%), and pH increased only slightly (<0.2 unit). For the 24-h tests, the wastewater was replaced after 12 h to ensure that the designated pH and nitrite levels were maintained. At the end of each test, the biofilm on the carrier was sampled for LIVE/DEAD staining, with methods to be further described.

2.2.2.

FNA toxicity to disrupted biofilms

In addition to the above tests with intact biofilms, one set of tests was carried out with disrupted biofilms. The biofilm carrier was transferred from a reactor to a 10 mL plastic tube

Table 1 e Experimental conditions used in the FNA toxicity batch tests with intact anaerobic biofilms. Test NO. 1 2 3 4 5 6 7 8 9 10 11 12

Exposure time (hour)

pH

6 6 6 6 12 12 12 12 24 24 24 24

7 6.5 6 5 7 6.5 6 5 7 6.5 6 5

Nitrite (mg-N/L) 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,

30, 60, 30, 60, 30, 60, 30, 60, 30, 60, 30, 60, 30, 60, 30, 60, 30, 60, 30, 60, 30, 60, 30, 60,

90, 120 90, 120 90, 120 90, 120 90, 120 90, 120 90, 120 90, 120 90, 120 90, 120 90, 120 90, 120

FNAa (mg-N/L) 0, 0.008, 0.015, 0.023, 0.031 0, 0.02, 0.05, 0.07, 0.1 0, 0.08, 0.15, 0.23, 0.31 0, 0.77, 1.53, 2.3, 3.07 0, 0.008, 0.015, 0.023, 0.031 0, 0.02, 0.05, 0.07, 0.1 0, 0.08, 0.15, 0.23, 0.31 0, 0.77, 1.53, 2.3, 3.07 0, 0.008, 0.015, 0.023, 0.031 0, 0.02, 0.05, 0.07, 0.1 0, 0.08, 0.15, 0.23, 0.31 0, 0.77, 1.53, 2.3, 3.07

pH a FNA was not measured but calculated from the nitrite and pH levels applied/measured: FNA ¼ NO Þ, where Ka is the ionization 2  N=ðKa  10 constant of the nitrous acid (Anthonisen et al., 1976; Weon et al., 2002). Ka ¼ e2300/(T þ 273), where T is temperature ( C).

3738

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 3 5 e3 7 4 3

filled with wastewater to avoid exposure to air. The biofilm was then detached from the plastic carrier with an ultrasonicator bath (Soniclean 250 HT, 6 L, 120 W pulse swept power) at 45 kHz for 60 s, followed by 30 s of vortexing. The disrupted biofilm was exposed to pH ¼ 6 and nitrite concentrations of 0, 30, 60, 90, and 120 mg-N/L, for 6 and 24 h. Microbial viability was analyzed with LIVE/DEAD staining prior to and after the exposure to nitrite. The particle size distribution of the disrupted biofilms was measured with Mastersizer 2000 (Malvern Instruments).

a KryptoneArgon laser (488 nm) and two HeeNe lasers (543 and 633 nm). Twenty images were taken for randomly chosen areas of each sample. Quantification of live and dead microorganisms was done by determining the relative abundance of green and red pixels. The pixel area counting was conducted with the ImageJ (National Institute of Health, USA). The percentage of green fluorescence to the total fluorescence (red þ green fluorescence) was assumed to be equal to the percentage of viable cells to the total cells (viable þ dead) in the biofilm.

2.3.

2.5.

Direct FNA dosing to sewer biofilm reactors

Upon the completion of the above-described toxicity tests, Reactors R2 and R3 were dosed with nitrite and acid to reach FNA at 0.18 and 0.36 mg-N/L, respectively for 24 h (four pumping cycles). This was achieved by adding nitrite to these two reactors immediately after each pumping event to achieved nitrite concentrations of 70 and 140 mg-N/L, respectively. The pH in both dosed reactors was maintained at 6.0  0.1 by adding 1 M hydrochloric acid. R1 was not dosed with nitrite or acid, and therefore served as a control reactor. This experiment aimed to verify the biocidal effect of FNA revealed in the viability tests through monitoring the loss of biofilm sulfide and methane producing activity after FNA dosage and its subsequent recovery. These dosing conditions (FNA levels and duration) were chosen based on the results of the viability tests. Batch tests as described in a previous section were conducted immediately after the 24 h dosage and continued for 60 days with intervals of 2 days to two weeks, to monitor the sulfide and methane production rates of the three reactors. Biofilm carriers were taken from all reactors prior to and immediately after the dosing events to determine the viability of microorganisms in biofilms.

2.4.

Chemical analysis

For the analyses of dissolved inorganic sulfur species, 1.5 mL wastewater was filtered (0.22 mm membrane) into 0.5 mL preserving solution of sulfide anti-oxidant buffer (SAOB) (Keller-Lehmann et al., 2006). Samples were analyzed within 24 h on an ion chromatograph (IC) with a UV and conductivity detector (Dionex ICS-2000). For the analysis of nitrogen species (nitrite), 1 mL of wastewater was filtered and diluted 10 times. It was analyzed using a Lachat QuikChem 8000 (Milwaukee) flow-injection analyzer (FIA). VFA was measured by gas chromatography (PerkinElmer, Inc.). For the measurement of dissolved methane, 5 mL wastewater was filtered into vacuumed BD vacutainer tubes using a hypodermic needle attached to a plastic syringe. The tubes were allowed to reach gaseliquid equilibrium overnight. Methane in the gas phase was measured by gas chromatography (Shimadzu GC-9A) equipped with a flame ionization detector (FID). Concentrations of methane in wastewater were calculated using mass balance and Henry’s law (Guisasola et al., 2008).

3.

Results

3.1.

Toxicity of FNA to anaerobic sewer biofilm

LIVE/DEAD staining

The viability of microorganisms in biofilms was determined using the LIVE/DEAD BacLight bacterial viability kits (Molecular Probes, L-7012). The viability kits utilize two nucleic acid stains, namely green-fluorescent SYTO-9 and redfluorescent Propidium Iodide (PI) (Invitrogen Molecular Probes, 2004). The SYTO-9 stain generally labels all microorganisms in a population with intact or damaged membranes. In contrast, PI stain penetrates only those microorganisms with damaged membranes, causing a reduction in the SYTO-9 stain fluorescence when both dyes are present. Thus, microorganisms with intact cell membranes (viable cells) are stained green, whereas microorganisms with damaged membranes (dead cells) are stained red. Biofilm on a plastic carrier was detached in filtered (0.22 mm) sewage with vigorous shaking and vortex mixing. Biofilm suspension (125 mL for each test) was transferred into 2-mL plastic centrifuge tubes with 50 mL of SYTO-9 and PI mixture solution. The tubes were incubated in a dark place for 15 min at the room temperature (20  C), allowing the staining reactions to complete. Then, microscope slides with stained biofilm samples were photographed using a confocal laser scanning microscope (Zeiss LSM 510 META), equipped with

Fig. 1A & B shows the percentage of viable microorganisms in biofilm plotted against the nitrite and pH levels, respectively, for the 12-h exposure tests (Tests 5e7). Similar results were obtained with the 6-h and 24-h tests (Figs. S2 and S3). The dependence of viable percentage on FNA concentration for the three exposure times, i.e. 6 (Tests 1e3), 12 (Tests 5e7), and 24 h (Tests 9e11), are shown in Fig. 1CeE respectively. The data obtained in Tests 4, 8 and 12 (i.e. Tests with pH ¼ 5) are not included in these figures and will be presented separately. There is a general negative impact of nitrite on the microbial viability. Higher nitrite concentration induced lower microbial viability. However, microbial viability varied in a wide range for each of the nitrite levels except for the case without nitrite addition, suggesting that nitrite is not the sole factor contributing to the loss of microbial viability and pH likely played a role in conjunction with nitrite. A similar observation can be made on the relationship between microbial viability and pH. Microbial viability was reduced to very different levels at the same pH but different nitrite levels. This suggests that pH within the range of 6.0e7.0 was not the main toxic factor either. Fig. 1CeE shows that the level of microbial viability had a much stronger dependence upon the FNA concentration, indicating that FNA may directly cause the inactivation of

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 3 5 e3 7 4 3

3739

Fig. 1 e The dependency of viable percentages in biofilms (%) on nitrite concentration (A), pH (B) after 12-h of exposure; on FNA concentration after exposure for 6 (C), 12 (D), and 24 h (E). Also presented in subplots C, D, and E are regression lines, which were obtained with a 3-parameter exponential decay model. microorganisms in biofilms. Also shown in the figure is the fit between the experimental data and the predictions by an exponential model ( y ¼ y0 þ aebx). It is seen that the toxicity of FNA to sewer biofilm could be well described by the exponential function. Microbial viability decreased sharply with increased FNA concentration in the range of 0e0.1 mg-N/L. Microbial viability decreased by 50% after being exposed to FNA at a concentration of 0.045 mg-N/L (equivalent to a nitrite concentration of 18 mg-N/L at pH 6) for all three exposure times, i.e. 6, 12 and 24 h. The decrease in microbial viability slowed down when the FNA concentration further increased. The loss of microbial viability also depended on the duration of the FNA treatment. In the exponential model, parameters b and y0 indicate the overall decreasing rate of microbial viability and the residual percentage of viable microorganisms, respectively. When the exposure time increased from 6 to 12 and 24 h, b increased from 17.1 to 19.7 and 22.5, respectively, while y0 decreased from 13.6 to 10.2 and 7.6, respectively. Thus, longer exposure time could increase the microbial killing efficiency. Also, longer exposure resulted in lower residual microbial viability. Fig. 2 shows that a pH level of 5 alone (in the absence of nitrite) can reduce the biofilm viability from 70 to 80% measured prior to the treatment to around 20% after 6e24 h exposure (Tests 4, 8, and 12). The presence of nitrite further enhanced the biocidal effect. It is worthwhile to note, however, that microbial viability was still 2e3% when the FNA concentration was greater than 3 mg-N/L (120 mg-N/L at pH 5). This indicates that a small portion of microorganisms is highly resistant to FNA.

3.2.

Toxicity of FNA to disrupted biofilms

The physical disruption (ultra-sonication and vortex mixing) detached biofilms and broke them down to particles with a median size of 310  5 mm (See Fig. S4 for the particle size distribution). The microbial viability after sonication was at the same level of intact biofilms, i.e. approximately 70% (The leftmost data points in Fig. 3), implying that sonication did not cause significant microbial death by itself. It was found that disruption made the biofilm more susceptible to pH change. In the absence of nitrite, pH change from 7.6 to 6 decreased the viability by 25% in the disrupted biofilms. In comparison, this decrease was negligible for the intact biofilm samples (Fig. 3). Fig. 3 also shows that disrupted biofilm was more sensitive to FNA toxicity. The viable microorganisms in disrupted

Fig. 2 e Microbial viability at pH [ 5 and FNA concentrations of up to 3 mg-N/L, after exposure time of 6, 12, and 24 h.

3740

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 3 5 e3 7 4 3

3.4.

Fig. 3 e The viable percentages (%) in intact (filled symbols) and disrupted (empty symbols) biofilms after being exposed to FNA at pH [ 6 for 6 and 24 h. The leftmost data points are the viability for biofilm samples taken from the reactors directly or sonicated biofilms. Regression lines were obtained with exponential decay equation y [ y0 D aeLbx. The parameters are summarized in Table S2.

biofilm after exposure to FNA for 6 h were 10e20% lower than that in the intact biofilm. It is clear that the FNA toxicity to biofilms can be augmented by physical disruption of biofilms. For the 24-h exposure, the percentage of viable microorganisms (%) in intact and disrupted biofilms reached similar levels at high FNA concentrations. As shown in Fig. 3, about 4e5% of microorganisms survived the exposure to FNA at 0.31 mg-N/L for 24 h. This implies that some microorganisms are resistant to FNA no matter in intact or in disrupted biofilms.

3.3.

Effect of pH on biofilm viability

Biofilms in the reactors were developed with real wastewater, which had a pH around 7.6. Fig. 4 shows the viability test results using intact biofilms at different pH in the absence of nitrite. The viable microorganisms in the anaerobic biofilms incubated with fresh wastewater after 6e24 h were about 70e80% in all tests. No discernable decrease of microbial viability was observed when pH was lowered to 7. For pH adjustments to 6.5 and 6, there was a slight decrease (4e9%) of microbial viability for all exposure times. However, pH level at 5 was found to be strongly biocidal. It reduced the biofilm viability to 20% after 6-h exposure, and even lower for longer exposure times.

Fig. 4 e Effect of pH on microbial viability in biofilms after incubation at the specified pH for 6, 12 and 24 h.

Biofilm activity after FNA dosage to reactors

Fig. 5A shows the sulfide production rates in the FNA dosed reactors, i.e. R2 and R3, relative to that in the control reactor R1. The sulfide and methane production rates of R1 remained stable during the entire experimental period. Both R2 and R3 had sulfide and methane production rates similar to those of R1 before the commencement of FNA dosing on Day 0. Although R2 and R3 were dosed with different levels of FNA, i.e. 0.18 and 0.36 mg-N/L, respectively, sulfide and methane production were both completely suppressed at the end of the 24-h dosing. The viable microorganisms in the R2 and R3 biofilms decreased to 8.0  1.3% and 5.6  2.9%, respectively after FNA dosing, compared to 82.1  2.3% in R1, consistent with the results of viability tests reported above. After the FNA dosing, sulfide and methane production in R2 and R3 started to recover without an obvious lag phase. No prominent differences could be found between R2 and R3 in terms of the rates of recovery. The Gompertz growth equation was employed to fit the recovery phase (Jiang et al., 2010). It is clear that the recovery processes resemble that of microbial regrowth. 50% recovery of the sulfide production rates took 9.2 and 7.7 days, respectively, for R2 and R3. In comparison, the methane production rates took 58.0 and 52.3 days for R2 and R3, respectively, to reach 50% recovery. The recovery of methane production was about 7 times slower compared to the recovery of sulfide production.

4.

Discussion

4.1.

Biocidal effect of FNA

Previous work has demonstrated the inhibitory effects of FNA on a broad range of microbial metabolism. Zhou et al. (2008)

Fig. 5 e Sulfide (A) and methane (B) production rates of R2 and R3, relative to the corresponding R1 rates, prior to and after being exposed to free nitrous acid at 0.18 or 0.36 mgN/L, respectively, on Day 0 (vertical arrows). The solid and dashed lines are regressions with the Gompertz growth equation, see details in Supplementary information.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 3 5 e3 7 4 3

reported that FNA at a concentration of 0.004 mg-N/L or above could completely inhibit the N2O reduction activity of a denitrifying enhanced biological phosphorus removal sludge. Vadivelu et al. (2006a,b) revealed that the anabolic processes (growth) of an enriched Nitrobacter and an enriched Nitrosomonas culture were stopped by FNA at concentrations of 0.023 and 0.4 mg-N/L, respectively. Similarly, Pijuan et al. (2010) and Ye et al. (2010) showed that the growth of Candidatus Accumulibacter phosphatis (a known polyphosphate accumulating organism) and Candidatus Competibacter phosphatis (a known glycogen accumulating organism) was completely suppressed by FNA at 0.006 mg-N/L and 0.007 mg-N/L, respectively. The above reported FNA-caused inhibition on microbial metabolism was found to be reversible, with the microbial activities resuming immediately or within hours after the removal of the inhibitor (Pijuan et al., 2010; Vadivelu et al., 2006a; Ye et al., 2010). It is noteworthy that FNA applied in most of these studies (with the exception of the study on Nitrosomonas) were at parts per billion (ppb) levels. In this study, we revealed for the first time the strong biocidal effect of FNA on microorganisms in anaerobic wastewater biofilms. This was demonstrated through measuring both membrane integrity of microorganisms and biological activities of biofilm reactors before and after FNA treatment. After being exposed to FNA for 6e24 h at concentrations of 0.1e0.3 mg-N/L, viable microorganisms in anaerobic biofilms reduced from approximately 80% to 5e15%. In fact, the biocidal effect initiated at FNA levels far lower than 0.1 mg-N/L, and increased sharply with FNA concentration in the range of 0e0.1 mg-N/L. The biocidal effect was found to be strongly dependent on the FNA concentration rather than the nitrite concentration or the pH level separately (for the pH range of 6.0e7.6). The biocidal effect revealed by microbial viability tests was strongly supported by the loss of biofilm activity after FNA treatment and its subsequent slow recovery following termination of FNA dosages. Mathematical modeling showed that the activity recovery was likely due to the regrowth of the residual viable microorganisms. Our experimental results suggest that acidified conditions promote the formation of FNA, which kills microorganisms in the anaerobic biofilms. While the detailed mechanism is yet to be revealed in future studies, a number of potential contributing factors are discussed below. Once FNA was formed from nitrite under acidic conditions, many reactive derivatives can be generated (Yoon et al., 2006): HNO2 þ HNO2 4N2 O3 þ H2 O4NO þ NO$2 þ H2 O Small molecules such as dinitrogen trioxide (N2O3), nitrogen dioxide (NO2) and nitric oxide (NO) can readily cross cell membranes where they can react with reduced thiols to form nitrosothiols, which are thought to be important in microbial killing (Heaselgrave et al., 2010; Phillips et al., 2004). NO is a well-known antimicrobial agent (Fang, 1997), and has also been demonstrated to be able to cause dispersal of Pseudomonas aeruginosa biofilms and multi-species biofilms from water distribution and treatment systems (Barraud et al., 2006, 2009). The strong nitrosating intermediate, N2O3, is capable of modifying the function of proteins (Yoon et al., 2006). Also, NO$2 can induce lipid peroxidation, resulting in cell membrane damage (Halliwell et al., 1992).

3741

FNA can cause oxidative deamination of the NH2 group of adenine or cytosine to an ether group (Klug et al., 2009; Malling, 2004). It converts adenine to hypoxanthine (which pairs with C), cytosine to uracil (which pairs with A) and guanine to xanthine (which still pairs with C). FNA alters a DNA base pair directly to a “miscoding” form and thus does not require subsequent DNA synthesis for its effect. The change of DNA base pairing (mutagenesis) is lethal to microorganisms. A further possible cause of microbial death could be the metabolic inhibition and the adverse change of their living environment. Interrupted metabolisms combined with the acidic pH can be lethal to microorganisms, especially after being exposed for 6 h or longer. The exposure time (6, 12, and 24 h) used in this study is much longer than those (0.5e3 h) in the FNA inhibition studies previously conducted (Pijuan et al., 2010; Vadivelu et al., 2006b; Ye et al., 2010; Zhou et al., 2008). In the pH 5 tests, 2e3% of microorganisms remained variable after being treated at FNA concentrations of up to 3 mg-N/L. It is possible that some microorganisms residing in deep layers could have been protected against the FNA toxicity to a certain degree due to mass transfer limitations. However, a similar level of residual viable microorganisms was also observed in the disrupted biofilm studies, suggesting that some microbial species may be more tolerant to FNA than others. Indeed, a nitritation reactor treating anaerobic digestion liquor, where ammonia oxidizing bacteria (AOB, e.g. Nitrosomonas) proliferate, nitrite is typically at 500e600 mg-N/L and pH easily reaches 6.0e6.5, giving rise to FNA in the range 0.4e1.5 mg-N/L (Vadivelu et al., 2006a). These results clearly indicate that some AOB can tolerate these levels of FNA. The potentially different abilities of different microbial species to tolerate FNA toxicity require further investigation. Similarly, it is also important to study if microorganisms are able to adapt to high FNA concentrations.

4.2.

Potential applications of FNA toxicity

The biocidal effect of FNA on anaerobic sewer biofilms revealed in this study implies that FNA could be used as a biocidal agent to control the growth and activities of detrimental anaerobic biofilms. One potential application is the control of sulfide and methane formation and emission from sewers, to replace or supplement conventionally used chemicals such as oxygen, nitrate, ferric/ferrous salts and magnesium hydroxide. Based on the experimental results obtained in this study, we propose an FNA dosing strategy comprising a short period of FNA dosing and a longer period of recovery. This will substantially reduce the amount of chemicals required to achieve a desirable control of sulfide production and emission from sewers. There are many other potential benefits associated with the proposed intermittent FNA dosing. For example, continuous oxygen or nitrate addition to sewers causes consumption of organic carbon in sewers thus reduces its availability for nutrient removal at the downstream wastewater treatment plant (Gutierrez et al., 2008; Mohanakrishnan et al., 2009). In contrast, with the intermittent FNA dosing strategy, only a small amount of electron acceptor (i.e. nitrite) is supplied, and consequently the amount of carbon to be oxidized due to nitrite reduction will be small. In addition, given the very slow recovery of methanogenic activities of sewer biofilms, intermittent dosages of FNA will likely permanently suppress

3742

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 3 5 e3 7 4 3

methane production, reducing greenhouse gas emissions from sewers on the one hand, and preserving carbon sources for nutrient removal on the other hand (methane is not available for denitrification or phosphorus removal). Like other broad-spectrum biocides having multiple targets in terms of regulating gene expressions (Lee et al., 2010), FNA was demonstrated to be a biocide to different microorganisms. However, the adverse impacts of FNA to environment and the downstream wastewater treatment processes are expected to be negligible if added appropriately. It is not expected that FNA would be added to an entire sewer network at the same time. Instead, it should be added to different sections of a network at different time. At downstream sewer sections or at the treatment plant, nitrite will be diluted and pH neutralized, leading to non-toxic and non-inhibitory levels of FNA, which can be removed microbiologically. NOx (NO and NO2), form as FNA derivatives at acidic pH in the closed sewer pipes (rising main). Primarily, they can easily cross cell membranes and react with cell DNA, proteins and lipids. The residual NOx will be diluted in downstream pipes of selected dosing sites. Thus, NOx release from sewer system is not expected to be a significant problem. However, it is recognized that further study should be carried out to quantify the impacts of NOx release. The intermittent FNA dosing strategy discussed above is likely applicable to many other anaerobic environments, e.g. for odor management of landfill and souring control of oil fields. The dosing frequency of an intermittent strategy will depend on the microbial growth rate. For faster growing microorganisms, a more frequent dosing of FNA will be required to keep the activity low at all times. Research is required for optimizing the dosing strategy under different environments.

5.

Conclusions

The biocidal effect of free nitrous acid on anaerobic sewer biofilms was investigated through both viability tests and reactor studies. The key findings are:  Free nitrous acid showed strong biocidal effect toward microorganisms residing in anaerobic sewer biofilms. The level of the biocidal effect has a much stronger correlation with the FNA concentration than with the nitrite concentration or with the pH level, suggesting that FNA is the actual biocidal agent.  The biocidal effect of FNA is reduced by the limitation of mass transfer in the thick sewer biofilm. Some microorganisms in the biofilms might be resistant to FNA at concentrations up to 3 mg-N/L for exposure time up to 24 h.  Based on the biocidal effects of FNA, intermittent dosing of nitrite with acid (to form FNA) is potentially a cost-effective strategy to control sulfide and methane production in sewers.

Acknowledgments The study was undertaken as part of the Sewer Corrosion and Odour Research (SCORe) Project LP0882016 funded by the

Australian Research Council and many members of the Australian water industry (for more details see: www.score. org.au). Guangming Jiang receives the Endeavour International Postgraduate Research Scholarship (IPRS) and University of Queensland International Living Allowance Scholarship (UQILAS).

Appendix. Supplementary material Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.04.026.

references

Anthonisen, A.C., Loehr, R.C., Prakasam, T.B.S., Srinath, E.G., 1976. Inhibition of nitrification by ammonia and nitrous acid. Journal Water Pollution Control Federation 48 (5), 835e852. Barraud, N., Hassett, D.J., Hwang, S.H., Rice, S.A., Kjelleberg, S., Webb, J.S., 2006. Involvement of nitric oxide in biofilm dispersal of Pseudomonas aeruginosa. Journal of Bacteriology 188 (21), 7344e7353. Barraud, N., Storey, M.V., Moore, Z.P., Webb, J.S., Rice, S.A., Kjelleberg, S., 2009. Nitric oxide-mediated dispersal in singleand multi-species biofilms of clinically and industrially relevant microorganisms. Microbial Biotechnology 2 (3), 370e378. Bentzen, G., Smith, A.T., Bennett, D., Webster, N.J., Reinholt, F., Sletholt, E., Hobson, J., 1995. Controlled dosing of nitrate for prevention of H2S in a sewer network and the effects on the subsequent treatment processes. Water Science and Technology 31 (7), 293e302. Boon, A.G., 1995. Septicity in sewers: causes, consequences and containment. Water Science and Technology 31 (7), 237e253. Fang, F.C., 1997. Mechanisms of nitric oxide-related antimicrobial activity. Journal of Clinical Investigation 99 (12), 2818e2825. Foley, J., Yuan, Z., Lant, P., 2009. Dissolved methane in rising main sewer systems: field measurements and simple model development for estimating greenhouse gas emissions. Water Science and Technology 60 (11), 2963e2971. Gardner, L.R., Stewart, P.S., 2002. Action of glutaraldehyde and nitrite against sulfate-reducing bacterial biofilms. Journal of Industrial Microbiology & Biotechnology 29 (6), 354e360. Greene, E.A., Brunelle, V., Jenneman, G.E., Voordouw, G., 2006. Synergistic inhibition of microbial sulfide production by combinations of the metabolic inhibitor nitrite and biocides. Applied and Environmental Microbiology 72 (12), 7897e7901. Greene, E.A., Hubert, C., Nemati, M., Jenneman, G.E., Voordouw, G., 2003. Nitrite reductase activity of sulphatereducing bacteria prevents their inhibition by nitratereducing, sulphide-oxidizing bacteria. Environmental Microbiology 5 (7), 607e617. Guisasola, A., de Haas, D., Keller, J., Yuan, Z., 2008. Methane formation in sewer systems. Water Research 42 (6e7), 1421e1430. Guisasola, A., Sharma, K.R., Keller, J., Yuan, Z.Q., 2009. Development of a model for assessing methane formation in rising main sewers. Water Research 43 (11), 2874e2884. Gutierrez, O., Mohanakrishnan, J., Sharma, K.R., Meyer, R.L., Keller, J., Yuan, Z., 2008. Evaluation of oxygen injection as a means of controlling sulfide production in a sewer system. Water Research 42 (17), 4549e4561. Gutierrez, O., Park, D., Sharma, K.R., Yuan, Z., 2009. Effects of long-term pH elevation on the sulfate-reducing and

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 3 5 e3 7 4 3

methanogenic activities of anaerobic sewer biofilms. Water Research 43 (9), 2549e2557. GWRC, 2011. Global Water Research Coalition Report, N2O and CH4 Emission from Wastewater Collection and Treatment Systems: State of the Art Report ISBN: 987-90-77622-23-0. Halliwell, B., Hu, M.L., Louie, S., Duvall, T.R., Tarkington, B.K., Motchnik, P., Cross, C.E., 1992. Interaction of nitrogen-dioxide with human plasma e Antioxidant depletion and oxidative damage. FEBS Letters 313 (1), 62e66. Heaselgrave, W., Andrew, P.W., Kilvington, S., 2010. Acidified nitrite enhances hydrogen peroxide disinfection of Acanthamoeba, bacteria and fungi. Journal of Antimicrobial Chemotherapy dkq075. Hvitved-Jacobsen, T., 2002. Sewer Processes: Microbial and Chemical Process Engineering of Sewer Networks. CRC Press, Boca Raton London New York Washington, D.C. Invitrogen Molecular Probes, 2004. LIVE/DEAD BacLight Bacterial Viability Kits. http://probes.invitrogen.com/media/pis/ mp07007.pdf (accessed on 03.05.11). IPCC, 2006. In: Eggleston, H.S., Buendia, L., Miwa, K., Ngara, T., Tanabe, K. (Eds.), 2006 IPCC Guidelines for National Greenhouse Gas Inventories, Prepared by the National Greenhouse Gas Inventories Programme. IGES, Japan. Jameel, P., 1989. The use of ferrous chloride to control dissolved sulfides in interceptor sewers. Journal Water Pollution Control Federation 61 (2), 230e236. Jiang, G., Gutierrez, O., Sharma, K.R., Yuan, Z., 2010. Effects of nitrite concentration and exposure time on sulfide and methane production in sewer systems. Water Research 44 (14), 4241e4251. Keller-Lehmann, B., Corrie, S., Ravn, R., Yuan, Z., Keller, J., 2006. Preservation and simultaneous analysis of relevant soluble sulfur species in sewage samples. In: Proceedings of the Second International IWA Conference on Sewer Operation and Maintenance Vienna, Austria. Klug, W.S., Cummings, M.R., Ward, S.M., Spencer, C., Palladino, M.A., 2009. Concepts of Genetics. Pearson Benjamin Cummings. Lee, M.H.P., Caffrey, S.M., Voordouw, J.K., Voordouw, G., 2010. Effects of biocides on gene expression in the sulfate-reducing bacterium Desulfovibrio vulgaris Hildenborough. Applied Microbiology and Biotechnology 87 (3), 1109e1118. Malling, H.V., 2004. History of the science of mutagenesis from a personal perspective. Environmental and Molecular Mutagenesis 44 (5), 372e386. Mohanakrishnan, J., Gutierrez, O., Meyer, R.L., Yuan, Z., 2008. Nitrite effectively inhibits sulfide and methane production in a laboratory scale sewer reactor. Water Research 42 (14), 3961e3971. Mohanakrishnan, J., Gutierrez, O., Sharma, K.R., Guisasola, A., Werner, U., Meyer, R.L., Keller, J., Yuan, Z., 2009. Impact of nitrate addition on biofilm properties and activities in rising main sewers. Water Research 43 (17), 4225e4237. Nemati, M., Mazutinec, T.J., Jenneman, G.E., Voordouw, G., 2001. Control of biogenic H2S production with nitrite and molybdate. Journal of Industrial Microbiology & Biotechnology 26 (6), 350e355. Ochi, T., Kitagawa, M., Tanaka, S., 1998. Controlling sulfide generation in force mains by air injection. Water Science and Technology 37 (1), 87e95. Padival, N.A., Kimbell, W.A., Redner, J.A., 1995. Use of iron salts to control dissolved sulfide in trunk sewers. Journal of Analytical Chemistry Environmental Engineering-ASCE 121 (11), 824e829. Phillips, R., Kuijper, S., Benjamin, N., Wansbrough-Jones, M., Wilks, M., Kolk, A.H.J., 2004. In vitro killing of Mycobactetium ulcerans by acidified nitrite. Antimicrobial Agents and Chemotherapy 48 (8), 3130e3132.

3743

Pijuan, M., Ye, L., Yuan, Z., 2010. Free nitrous acid inhibition on the aerobic metabolism of Poly-phosphate Accumulating Organisms. Water Research 44 (20), 6063e6072. Pomeroy, R.D., 1990. The Problem of Hydrogen Sulphide in Sewers. Clay Pipe Development Association Limited, London. Predicala, B., Nemati, M., Stade, S., Lague, C., 2008. Control of H2S emission from swine manure using Na-nitrite and Namolybdate. Journal of Hazardous Materials 154 (1e3), 300e309. Tanaka, S., Lee, Y.H., 1997. Control of sulfate reduction by molybdate in anaerobic digestion. Water Science and Technology 36, 143e150. Thistlethwayte, D., 1972. The Control of Sulphides in Sewerage Systems. Butterworth Pty. Ltd., Sydney. US EPA, 1974. Process Design Manual for Sulfide Control in Sanitary Sewer Systems. US EPA, 1991. Hydrogen Sulphide Corrosion in Wastewater Collection and Treatment System. US EPA, 1992. Detection, Control, and Correction of Hydrogen Sulfide Corrosion in Existing Wastewater System. Office of Wastewater Enforcement and Compliance, Washington, DC, 20460. Vadivelu, V.M., Keller, J., Yuan, Z., 2006a. Effect of free ammonia and free nitrous acid concentration on the anabolic and catabolic processes of an enriched Nitrosomonas culture. Biotechnology and Bioengineering 95 (5), 830e839. Vadivelu, V.M., Yuan, Z., Fux, C., Keller, J., 2006b. The inhibitory effects of free nitrous acid on the Energy generation and growth processes of an enriched Nitrobacter culture. Environmental Science & Technology 40 (14), 4442e4448. Weon, S.-Y., Lee, C.-W., Lee, S.-I., Koopman, B., 2002. Nitrite inhibition of aerobic growth of Acinetobacter sp. Water Research 36 (18), 4471e4476. WERF, 2007. In: Apgar, D., Witherspoon, J., Easter, C., Bassrai, S., Dillon, C., Torres, E., Bowker, R.P.G., Corsi, R., Davidson, S., Wolstenholme, P., Forbes, B., Quigley, C., Ward, M., Joyce, J., Morton, R., Weiss, J., Stuetz, R. (Eds.), Water Environment Research Fundation: Minimization of Odor and Corrosion in Collection Systems Phase 1 London, UK. Ye, L., Pijuan, M., Yuan, Z.G., 2010. The effect of free nitrous acid on the anabolic and catabolic processes of glycogen accumulating organisms. Water Research 44 (9), 2901e2909. Yongsiri, C., Vollertsen, J., Hvitved-Jacobsen, T., 2005. Influence of wastewater constituents on hydrogen sulfide emission in sewer networks. Journal of Environmental Engineering-Asce 131 (12), 1676e1683. Yoon, S.S., Coakley, R., Lau, G.W., Lymar, S.V., Gaston, B., Karabulut, A.C., Hennigan, R.F., Hwang, S.-H., Buettner, G., Schurr, M.J., Mortensen, J.E., Burns, J.L., Speert, D., Boucher, R. C., Hassett, D.J., 2006. Anaerobic killing of mucoid Pseudomonas aeruginosa by acidified nitrite derivatives under cystic fibrosis airway conditions. The Journal of Clinical Investigation 116 (2), 436e446. Zhang, L., De Schryver, P., De Gusseme, B., De Muynck, W., Boon, N., Verstraete, W., 2008. Chemical and biological technologies for hydrogen sulfide emission control in sewer systems: a review. Water Research 42 (1e2), 1e12. Zhang, L., Keller, J., Yuan, Z., 2009a. Inhibition of sulfate-reducing and methanogenic activities of anaerobic sewer biofilms by ferric iron dosing. Water Research 43 (17), 4123e4132. Zhang, L., Mendoza, L., Marzorati, M., Verstraete, W., 2009b. Decreasing sulfide generation in sewage by dosing formaldehyde and its derivatives under anaerobic conditions. Water Science and Technology 59 (6), 1248e1254. Zhou, Y., Pijuan, M., Zeng, R.J., Yuan, Z., 2008. Free nitrous acid inhibition on nitrous oxide reduction by a denitrifyingenhanced biological phosphorus removal Sludge. Environmental Science & Technology 42 (22), 8260e8265.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 4 4 e3 7 5 2

Available at www.sciencedirect.com

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

Modelling biological and chemically induced precipitation of calcium phosphate in enhanced biological phosphorus removal systems R. Barat a,*, T. Montoya a, A. Seco b, J. Ferrer a a b

Instituto de Ingenieria del Agua y Medio Ambiente, Universidad Polite´cnica de Valencia, Camino de Vera s/n. 46022 Valencia, Spain Departmento Ingenierı´a Quı´mica, Universitat de Vale`ncia, C/Dr Moliner 50, 46100 Burjassot, Valencia, Spain

article info

abstract

Article history:

The biologically induced precipitation processes can be important in wastewater treat-

Received 17 December 2010

ment, in particular treating raw wastewater with high calcium concentration combined

Received in revised form

with Enhanced Biological Phosphorus Removal. Currently, there is little information and

12 April 2011

experience in modelling jointly biological and chemical processes. This paper presents

Accepted 16 April 2011

a calcium phosphate precipitation model and its inclusion in the Activated Sludge Model

Available online 22 April 2011

No 2d (ASM2d). The proposed precipitation model considers that aqueous phase reactions quickly achieve the chemical equilibrium and that aqueous-solid change is kinetically governed. The model was calibrated using data from four experiments in a Sequencing

Keywords: Enhanced

biological

phosphorus

Batch Reactor (SBR) operated for EBPR and finally validated with two experiments. The

removal

precipitation model proposed was able to reproduce the dynamics of amorphous calcium

Calcium phosphate precipitation

phosphate (ACP) formation and later crystallization to hydroxyapatite (HAP) under

Mathematical modelling

different scenarios. The model successfully characterised the EBPR performance of the

Wastewater

SBR, including the biological, physical and chemical processes. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Since early days of the use of the Enhanced Biological Phosphorus Removal (EBPR), it was observed that the biologically induced precipitation processes could contribute significantly to the amount of phosphorus removed in the system (Arvin, 1983). This biologically induced precipitation is frequently reported in the literature and can be especially important in wastewaters with high calcium or magnesium concentration and during the sludge treatment (Langerak et al., 1998; Ohlinger et al., 1998; Doyle et al., 2000). Although struvite (magnesium ammonium phosphate) is the most important precipitate reported in wastewater treatment plants, Barat et al. (2009) detected an important precipitation of calcium

phosphate jointly with struvite during the anaerobic digestion (39.2% of the phosphate precipitated as hydroxyapatite). Different studies have reported that this biologically induced phosphorus precipitation cause important operational problems in the anaerobic digestion and downstream processes such as pipe blockage and accumulation on the surfaces of different sludge management devices such as centrifuges and pumps (see Doyle and Parsons, 2002 for a detailed review). Furthermore, recent studies (Barat et al., 2006, 2008) have observed that the phosphorus precipitation during the biological phosphorus removal process could affect the metabolism of Polyphosphate Accumulating Organisms (PAO). The inclusion of chemical processes in biological models has important advantages: to obtain a general model able to

* Corresponding author. Tel.: þ34 963879618; fax: þ34 963877618. E-mail addresses: [email protected] (R. Barat), [email protected] (T. Montoya), [email protected] (A. Seco), [email protected] (J. Ferrer). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.028

3745

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 4 4 e3 7 5 2

accurately predict the whole wastewater treatment performance and to know the optimum operation conditions or plant configuration to minimize the uncontrolled phosphorus precipitation. Nowadays, models for describing biological processes in wastewater treatment are well developed and are widely applied such as ASMs (Henze et al., 2000) and ADM1 (Batstone et al., 2002). Besides the simplified precipitation model included in the ASM2d (Henze et al., 2000), some models regarding the precipitation processes have been developed (Maurer et al., 1999 and Musvoto et al., 2000a,b) and commercial software, such as BioWin, includes a module to calculate the precipitation. However, the current precipitation models calculate the aqueous equilibrium with a set of kinetic expressions instead of solving it with the classical chemical equilibrium that will simplify the aqueous phase speciation calculation. In addition, the Musvoto model (Musvoto et al., 2000a,b) did not consider the dissolution process which could take place when the pH decreases beneath a certain level. Finally, there is very limited information and experience in modelling jointly biological and chemical process. The aim of this paper is to develop a calcium phosphate precipitation model and to couple it with the ASM2d in order to create a model extension able to represent jointly the biological, physical and chemical processes that take place in wastewater treatment. This work is divided into two parts: firstly, the precipitation model and its inclusion in the ASM2d extended with pH calculation and metal cations dynamic is presented; finally, the model was calibrated and validated using data from six experiments carried out in a laboratory scale anaerobiceaerobic (A/O) sequencing batch reactor (SBR) with different influent calcium concentration.

2.

Mathematical model

The model proposed consists of coupling different models: the ASM2d (Henze et al., 2000) modified with the metal cations (magnesium and potassium) dynamics (Barat et al., 2005), pH model (Serralta et al., 2004) and the calcium phosphate precipitation model developed in this study.

2.1.

Calcium phosphate precipitation model

The precipitation model has been developed in order to couple it with any biological model. Conceptually the calculation of the precipitation process firstly consists of the aqueous phase speciation and afterwards in the determination of the precipitate concentration from the speciation previously calculated.

The main model considerations are:  Aqueous phase acidebase and ion pairing reactions quickly achieve the chemical equilibrium (Serralta et al., 2004). This consideration hugely simplifies the aqueous phase speciation calculation, allowing the determination of the chemical equilibrium from a set of algebraic equations. Therefore, this assumption will reduce the simulation time (Rosen et al., 2005). These equations only depend on the equilibrium constants for each reaction whose values are widely available in literature.  Aqueous-solid change is kinetically governed, considering that these solid formation reactions tend to equilibrium limited by the rate of the process.

2.1.1.

Components in the precipitation model

The initial step in the model development was to establish the chemical species involved in the calcium phosphate precipitation process. Two types were distinguished: aqueous phase species and solid phase species. All the weak acidebase, cations and other species formed by ion pairing reactions were included as aqueous phase species. Table 1 shows the set of species included in the model.

2.1.2.

Model development

The precipitation model was divided into two parts according to the previous hypotheses.

2.1.2.1. Aqueous phase model. The aqueous phase reaction model has been performed as a chemical equilibrium problem. Mathematically it consists of the simultaneous resolution of the non-linear mass action expressions and the linear mass-balance relationships. In order to adapt the algebraic equations of the model to the equilibrium software used (MINTEQA2, Allison et al., 1991), it was distinguished between components and species. Chemically, the aqueous phase system is represented by a set of components and a set of species. A species is defined as every chemical entity to be considered. For the set of species selected, a set of components is chosen so that each species can be written as the product of a reaction involving only components, and no component can be written as the product of a reaction involving other components. With this chemical representation, two characteristic variables are defined: the species concentration, Ci (ML3), and the component concentration, Tj (ML3), being Tj the sum of concentration of all the species in which this component participates. The species that can be formed from the

Table 1 e Aqueous and solid phase species included in the precipitation model. Aqueous species Hþ OH CO32 HCO3

H2CO3 PO43 HPO42 H2PO4

Solid species H3PO4 NH4þ NH3 (aq) Mgþ2

Caþ2 Kþ MgOHþ CaOHþ

CaNH3þ2 Ca(NH3)2þ2 MgPO4 MgH2PO4þ

MgHPO4(aq) CaHPO4(aq) CaPO4 CaH2PO4þ

KHPO4 MgCO3(aq) MgHCO3þ CaHCO3þ

Amorphous calcium phosphate (ACP, Ca3(PO4)2  H2O) Hydroxyapatite (Ca5(PO4)3OH)

3746

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 4 4 e3 7 5 2

Table 2 e Stoichiometric matrix componentsespecies. Species

Components H2O

H2O Hþ PO43 NH4þ CO32 Ca2þ Mg2þ Kþ OH MgOHþ CaOHþ NH3 CaNH32þ Ca(NH3)22þ HPO42 H2PO4 H3PO4 MgPO4 MgH2PO4þ MgHPO4 (l) CaHPO4 (l) CaPO4 CaH2PO4þ KHPO4 HCO3 H2CO3 MgCO3 (l) MgHCO3þ CaHCO3þ CaCO3 (l)

þ

H

PO43

NH4þ

a

xj ij

Ca

aij Ci

K

1 1 1 1 1 1 1 1

1 1 1 1 1 2 1 2 3 2 1 1 2 1 1 2

1 1 1 1 2

1 1

1 1 1 1 1 1 1 1 1 1

1 1 1 1 1 1 1 1 1 1 1 1 1

1 1

i ¼ 1; 2; .Nsp

(1)

1 1 1 1

j ¼ 1; 2; .Nc

(2)

i¼1

where xi is the activity of the ith specie, xj is the activity of the jth component, aij is the stoichiometric coefficient of the jth component in the ith specie (see Table 2), Ki is the stability constant of the ith specie corrected for temperature variations with van’t Hoff equation and Ci is the concentration of the ith specie. Nsp is the number of species considered (28 in this case) and Nc is the number of components (8 in this case). The Appendix 1 develops the set of algebraic equations used to

log K 0 0 0 0 0 0 0 0 13.997 11.397 12.697 9.244 9.144 18.788 12.3 19.5 21.7 4.654 21.2561 15.175 15.035 6.460 20.923 13.255 10.329 16.681 2.92 11.339 11.599 3.2

solve the chemical equilibrium. The activity and concentration of every component or species are related through the correspondent activity coefficient that depends on the ionic strength (Davis equation): xi ¼ gi Ci

i ¼ 1; 2; .Nsp

(3)

where gi is the activity coefficient for the ith species. Unlike other precipitation models (Maurer et al., 1999; Musvoto et al., 2000a), which solve the aqueous equilibrium with kinetic expressions, the proposed model solves the equilibrium with a set of algebraic equations. This equilibrium calculation hugely simplifies the numerical solution and the

Table 3 e Calcium phosphate phases. Phase

Tj ¼

Mg

þ

1

j¼1

Nsp X

2þ

1

2.1.2.1.1. Processes. The set of chemical interactions included in the developed model comprises acidebase and ion pairing reactions. Ion pairing effects become significant at high ionic strength and total dissolved solids (Loewenthal et al., 1986). The equilibrium condition assumed for these reactions is described by a set of non-linear algebraic equations including one law of mass action for each species (Eq. (1)) and one massbalance for each component (Eq. (2)) to guarantee the material conservation principle. Nc Y

Constant 2þ

1

components mentioned above are listed on the matrix of componentsespecies shown in Table 2.

xi ¼ Ki

CO32

Brushite (DCPD) Monetite (DCPA) Octacalcium phosphate (OCP) Amorphous calcium phosphate (ACP) Tricalcium phosphate (TCP) Hydroxyapatite (HAP)

Composition

Molar ratio Ca/P

CaHPO4$2H2O

1.00

CaHPO4

1.00

Ca4H(PO4)3$2.5H2O

1.33

Ca3(PO4)2$H2O

1.50

Ca3(PO4)2

1.50

Ca5(PO4)3OH

1.67

3747

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 4 4 e3 7 5 2

model calibration due to the absence of kinetic parameters to solve the aqueous equilibrium and therefore to be calibrated. Furthermore, taking into account the implementation in a biological model, the number of differential equations is significantly reduced. While the kinetic approximation requires one differential equation per each species, in this case 28 equations, the proposed equilibrium consideration requires only 7, that is, one differential equation per each component, thus reducing the computational time.

2.1.2.2. Precipitation model. The precipitation reaction model, based on the chemical equilibrium, supposes the quick solid formation once exceeded the saturation index. Previous attempts carried out by the authors (data not shown) trying to model the precipitation as a process in equilibrium with the aqueous solution, showed that under this hypothesis the model overestimated considerably the precipitation. Therefore, it has been considered that this fast reaction does not occur in many of the heterogeneous equilibriums such as the precipitation processes. In these cases the reaction rate controls the process. Therefore, the precipitation and dissolution processes for the solid components considered in the model (see Table 1) have been modelled with a set of kinetic expressions. Another important aspect to be considered is that the kinetic of these processes are mostly surface controlled (Nancollas, 1979; Musvoto et al., 2000a; Maurer et al., 1999). Then, the kinetic expressions proposed are based on the theory of Koutsoukos et al. (1980). Many authors have studied the precipitation of calcium phosphates from supersaturated aqueous solutions (Abbona et al., 1986, 1988; Kibalczyc et al., 1990; Van Kemenade and de Bruyn, 1987) due to the great importance not only in wastewater treatment field but also in industry and medicine. Calcium phosphate precipitation is a very complex process involving various parameters. In particular, it depends on calcium and phosphate ion concentrations, as well as on supersaturation, ionic strength, temperature, ion types, pH but also on time (solidesolid transformation) as noted in the literature (Van Kemenade and de Bruyn, 1987). The calcium phosphate precipitation has been modelled following the Ostwald rule, which foresees that the least thermodynamically stable phase is the first one formed (see Table 3) and this phase acts as a precursor of the most stable phase, which is hydroxyapatite (HAP) among the different calcium phosphates solids. The precursor considered in this

Table 4 e Components involved in the precipitation model. Component

Unit

SPO4

g P m3

SH SCa

mol H m3 g Ca m3

XACP

mol ACP m3

XHAP

mol HAP m3

Description Inorganic soluble phosphorus Total proton Total soluble calcium Amorphous calcium phosphate Hidroxiapatite

Reference Henze et al. (2000)

Serralta et al. (2004) This study This study

This study

Table 5 e Stoichiometric matrix for the precipitation processes Processes

Components SPO4

ACP precip. ACP diss. HAP precip. Units

230.1 þ230.1 130.1 g m3

SH

SCa

XACP

XHAP

þ1 mol m3

340.1 þ340.1 240.1 g m3

þ1 1 -1 mol m3

þ1 mol m3

model is the amorphous calcium phosphate (ACP) according to the results obtained by Abbona et al. (1986, 1988), Seckler et al. (1996) and Musvoto et al. (2000b). The kinetic expressions proposed for the precipitationedissolution reactions are shown in Eqs. (4)e(6).  Amorphous calcium phosphate precipitation: kp

ACP

3Ca2þ þ 2PO3 4 þ xH2 O/Ca3 ðPO4 Þ2 $xH2 O

1=5 !2   3=5  3 2=5 KACP1 KSP ACP $ Ca2þ  PO4 XACP g3d $g2t KACP1 þ XTSS 1 þ sign ðSIACP Þ $ ð4Þ 2

vXACP ¼ kp ACP $ vt

 Amorphous calcium phosphate dissolution: kd

ACP

Ca3 ðPO4 Þ2 $xH2 O/3Ca2þ þ 2PO3 4 þ xH2 O 0 12 1      vXACP XACP K 3=5 2=5 SP ACP 5 C B ¼kd ACP $ $@ Ca2þ  PO3 A 4 vt KACP2 þ XACP g3d $g2t $

1  signðSIACP Þ 2

ð5Þ

Table 6 e Operational conditions of the experiments. Experiment Operational conditions SRT (days) Inf. P conc. (gP m3) Inf. Acetic acid conc. (gCOD m3) Feed P/COD ratio (gP gCOD1) Inf. Ca conc. (gCa m3) Measured values TSS (g m3) VSS (%) COD (gCOD m3) Total P (gP m3) Precipitated P (gP m3)

1

2

3

4

5

6

9 10

9 10

9 10

10 15

10 15

10 15

218

218

175

268

268

268

0.046

0.046

0.057

0.056

0.056

0.056

10

35

90

65

90

60

1050 72.0 1200

1050 72.0 1200

875 72.0 950

2300 61.0 2100

2438 62.5 2300

1588 65.1 1300

111

111

91

295

308

180

0

0

6.5

27.3

20.9

20.3

3748

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 4 4 e3 7 5 2

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 4 4 e3 7 5 2

 Hydroxyapatite crystallization: kp HAP

 Ca3 ðPO4 Þ2 $xH2 Oþ2Ca2þ þPO3 4 þOH /Ca5 ðPO4 Þ3 OHþxH2 O

 3   2þ  PO4 Ca vXHAP XACP   2þ  $ ¼ kHAP $ $ vt KHAP þ XACP KPO þ PO3 K þ Ca Ca 4

(6)

where XACP and XHAP are ACP and HAP concentration, respectively (mol/l), XTSS is the total suspended solids concentration (g m3), [Ca2þ] and [PO43] are calcium and orthophosphate ions concentration (mol l1) obtained from the liquid phase equilibrium solution, kp ACP and kd ACP are the ACP precipitation and dissolution constant rate, respectively (l mol1 d1), kHAP is the HAP precipitation constant rate (l mol1 d1), KSP ACP is the ACP solubility product (mol l1)5, gd and gt are the activity coefficient of diprotic and triprotic species, respectively, KACP1 is the ACP precipitation hyperbolic inhibition constant (M g1 m3), KACP2 is the ACP dissolution hyperbolic inhibition constant (mol l1), KHAP, KPO and KCa are HAP precipitation hyperbolic inhibition constants (mol l1), sign (x) is the sign of x (1, 0, þ1) and SI the saturation index calculated as:  2þ 3  3 2 Ca PO4 1 SIACP ¼ $log KSP ACP 5 g3d $g2t

(7)

Generally, mineral precipitation is only possible in supersatured solutions. At supersaturation the ion activity product (IAP) of a mineral exceeds its solubility product. The SI provides a non-linear scale for supersaturation. If SI is positive, the mineral can precipitate, while a negative SI indicates dissolution conditions. The last terms of Eqs. (4) and (5) were introduced for preventing precipitation when the solution is not saturated (Eq. (4)) and dissolution when the solution is not under saturated (Eq. (5)) as proposed by Udert et al. (2003). As previously stated, the kinetic of the ACP precipitation is surface controlled processes. This influence is represented including in the kinetic expression the total suspended solid concentration (Eq. (4)), which is assumed to be proportional to the total surface. Similarly to the model proposed by Maurer et al. (1999), a hyperbolic inhibition term takes into account the surface limitation of the ACP precipitation and a hyperbolic term slows down the back reaction at low ACP concentrations. The hydroxyapatite precipitation was modelled as an irreversible crystallization of ACP to HAP (Eq. (6)). Three hyperbolic terms, depending on the ACP, PO43 and Ca2þ concentration, limit the crystallization process.

2.1.3.

Coupling the precipitation model to ASM2d

The precipitation model has been coupled to the ASM2d in order to obtain an extended model able to predict with high accuracy the biological phosphorus removal process and the biologically induced precipitation processes. The structure of

3749

the coupled model is determined by the time scales of the involved model: biochemical (biological reactions) and precipitationedissolution processes are governed by the kinetic and are calculated by mass-balance equations; aqueous phase chemical reactions are assumed to occur instantly and are calculated by equilibrium chemistry-based algorithm. The formulation of ASM2d around the mass-balance for the total concentration of each component gives rise to a set of independent partial differential equations. The time integration of these equations provides the profiles of the total concentration of components (Tj). The profiles of the concentration of the species (Ci) must be obtained from the chemical interaction equations. Thus, the solution procedure involves a sequential iteration among the differential equations obtained from the mass-balance and the chemical module. The total concentration of each component (Tj) needed for applying the chemical model is provided by the solution of mass-balance equations. As a previous step to be able to model jointly the biological and precipitation processes it was necessary to extend the ASM2d with the pH model proposed by Serralta et al. (2004), which includes CO2 mass transfer from the liquid to the gas phase, and with the metal cations dynamic due to biological phosphorus removal process (Barat et al., 2005). Table 4 shows the components involved in the precipitation model. The new processes included in the model correspond to the precipitation and dissolution reactions of the solid components. Table 5 shows the stoichiometry of the precipitation and dissolution processes. The whole model has been included as additional information in the Appendixes 2 (model stoichiometry), 3 (conversion factors), 4 (kinetic equations), 5 (kinetic coefficients) and 6 (stoichiometric parameters). For model calibration purposes the extended model has been implemented in the simulation software DESASS (Ferrer et al., 2008), as shown later.

3.

Model application

Once the model was developed, the next phase was to evaluate the correct model performance and the calibration of the precipitation model parameters in biological phosphorus removal systems at different influent calcium concentrations.

3.1.

Experimental set-up

A laboratory scale sequencing batch reactor (SBR) has been operated under anaerobiceaerobic conditions for biological phosphorus removal. The SBR was operated with four 6-h cycles per day. The SBR description, the operation conditions and wastewater composition are extensively described in Barat et al. (2006). For the model calibration and validation purposes, six experiments were carried out under different operational

Fig. 1 e Simulation results of the global model. Actual measurements are plotted in dots and simulations of the global model are plotted in lines: (a) phosphate, volatile fatty acids and pH; (b) calcium, potassium and magnesium. Experiments 1, 2, 3 and 4 were used for model calibration and experiments 5 and 6 were used for model validation.

3750

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 4 4 e3 7 5 2

conditions (Table 6). The first four experiments were used for calibration and the last two experiments were used for validation. Each experiment consisted in an intensive study throughout one cycle once achieved a pseudo steady state. Acetate, phosphorus, calcium, magnesium, potassium and pH were measured during the cycle. COD, TSS, VSS, total P and precipitated P were measured at the end of the aerobic phase.

3.2.

Analytical methods

The analyses of phosphate (ascorbic acid method), metal cations (atomic absorption spectrophotometry method), nitrate, chemical oxygen demand (COD), total suspended solids (TSS) and volatile suspended solids (VSS) were performed in accordance with Standard Methods (APHA, 1998). Acetic acid and carbonate alkalinity were determined by a method proposed by Moosbrugger et al. (1992). Precipitated phosphorus was determined according to the method proposed by De Haas et al. (2000).

3.3.

Results and discussion

As previously mentioned, the experimental work was focused on an SBR to simultaneously study the biological phosphorus removal and the calcium phosphate precipitation processes under dynamic conditions. The batch reactor operated for EBPR is able to achieve high concentrations of phosphates and metal cations at the end of the anaerobic phase. Hence, this technology was selected to study at the same time both the biological and the chemical processes. Fig. 1a shows the measured values of acetic acid, phosphate and pH. Fig. 1b shows the calcium, potassium and magnesium profiles during each experiment. As pointed out by different authors, Ca profiles show that this cation is not clearly involved in the biological phosphate dynamic as are potassium and magnesium (Barat et al., 2005). Therefore, the Ca variations throughout the SBR operation cycle can be explained by calcium precipitation as discussed below. As can be observed, in all the experiments Ca remains quite constant during the anaerobic phase despite the phosphate increase, which is probably due to the pH decrease during the anaerobic phase. However, during the aerobic phase, Ca concentrations change significantly except in the experiments 1 and 2 which have a low Ca concentration. As experiments 3, 4, 5 and 6 markedly show (Fig. 1b), Ca concentration dropped once the aerobic conditions were established. This could be explained as a calcium phosphate precipitation caused by the high phosphate concentration and the CO2 stripping, which increased the pH value. In experiments 4 and 5, once the concentration of phosphate decreased to a certain level due to the biological phosphate accumulation, the concentration of calcium started to increase until values that were slightly lower than the starting cycle concentrations were achieved. This calcium increase could be due to calcium phosphate dissolution caused by the low phosphate concentrations achieved, changing the equilibrium conditions. Despite this calcium phosphate dissolution, an increase in phosphate concentration was not observed because this additional phosphate was immediately taken up by PAO.

Before model calibration, the Saturation Index (SI) for different precipitates were calculated in order to confirm the formation of amorphous calcium phosphate instead of other precursors reported in literature (Table 3) and to confirm the lack of precipitation of other solids, such as, struvite, newberite and calcium phosphate also reported in literature (Musvoto et al., 2000b). Fig. 2a shows the SI in the experiment 4 for struvite, newberyite, ACP and calcite. The SI profile of calcium carbonate showed that its precipitation was possible throughout the entire aerobic phase. However, the experimental results showed clear calcium dissolution at the middle of the aerobic phase. This dissolution would not be possible for calcite precipitation, so it was considered that calcite did not precipitate in these experiments even though calcite formation was possible thermodynamically. This could be due to crystal formation inhibition by phosphate reported by different authors (Plant and House, 2002). With regard to the calcium phosphate formation, the SI of the possible HAP precursors observed in literature (ACP, DCPD and OCP) was tested to know which mineral better fitted the experimental results (Fig. 2b). As can be seen, OCP and DCPD did not precipitate because their SI predicted precipitation in the anaerobic phase while Ca concentration did not change under anaerobic conditions. The best fit resulted with ACP which was able to thermodynamically predict non-solid formation during the anaerobic phase, precipitation in the first half of the aerobic phase and dissolution at the end of the aerobic phase. The model parameters were calibrated once checked the solids precipitated. The first four experiments were used for calibration purposes and the 5th and 6th experiments were used for model validation. Values of the kinetic parameters of the model were gradually changed to minimize the sum of squared relative deviations of the concentration profiles and those from measurements. Table 7 shows the calibrated values of stoichiometric and kinetic constants. The values used for the rest of the parameters were those proposed by Henze et al. (2000), Serralta et al. (2004) and Barat et al. (2005). As can be seen, only three parameters of the biological model were changed from default values to reproduce the experimental data. Despite that the values of the kinetic and stoichiometric parameters qPHA and YPHA are different from those proposed in the ASM2d, the calibrated values are within the rank of their values proposed in the literature. As can be seen in Table 7, experimental results could not be reproduced with the same value for the parameter YPO4. The values obtained for YPO4 are higher than the default value proposed in the ASM2d (YPO4 ¼ 0.4 gP gCOD1). Nevertheless, this parameter varies hugely in the literature (changing from 0.03 to 0.8 gP gCOD1) due to: pH variations (Filipe et al., 2001), the ratio between phosphorus and chemical oxygen demand (P/COD) in raw wastewater (Schuler and Jenkins, 2003), the ratio between PAO and GAO bacteria present in the sludge (Manga et al., 2001) and a change in the PAO metabolic pathway (Barat et al., 2006). In this case the variation of the YPO4 values could be due to the differences in the P/COD ratio in the influent (see Table 6) and a change in the PAO metabolic pathway. This metabolic change could be induced by the variation in the amount of internal Poly-P available as energy source per unit of PAO bacteria. However, the Poly-P/PAO ratio

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 4 4 e3 7 5 2

3751

Fig. 2 e Experiment 4 saturation index.

is not measured in this work and further research will be needed to confirm this hypothesis. Finally the calibrated model was validated with two experiments (experiments 5 and 6) obtaining good model predictions in both cases. During the validation process it was necessary also to modify the values of the parameter YPO4 (0.56 g P gCOD1 and 0.63 g P gCOD1 for the experiments 5 and 6, respectively). As can be seen in Fig 1a and b, the proposed model successfully characterised the EBPR performance of the SBR. Simulated acetic acid, phosphorus, potassium, magnesium and pH profiles accurately reproduced the experimental data. Also, the model was able to reproduce the calcium phosphate precipitation processes under different situations: without precipitation (experiments 1 and 2), slight precipitation (experiments 3 and 6) and precipitationedissolution (experiments 4 and 5). Therefore, the precipitation model, developed jointly with the biological model (ASM2d) and the pH calculation, were able to reproduce satisfactorily the biological and chemical processes that take place in an SBR operated for Enhanced Biological Phosphorus Removal. The proposed model extended with other precipitation processes (i.e. struvite, calcium carbonate, iron and aluminium phosphate) will be a very useful tool to simulate the whole wastewater treatment performance including the effect of the precipitation on the biological process and to simulate the optimal operation conditions in order to minimize the uncontrolled precipitation problems in WWTPs widely reported in literature (Doyle and Parsons, 2002).

Table 7 e Stoichiometric and kinetic parameter values adjusted during calibration. Parameter

Units

Biological model d1 qPHA g P g COD1 YPO4 g COD g PP1 YPHA Precipitation model (mol l1)5 KSP ACP l mol1 d1 kp ACP l mol1 d1 kd ACP M g1 m3 KACP1 mol l1 KACP2 mol l1 d1 kHAP mol l1 KHAP

Calibrated value (20  C)

ASM2d (20  C)

4.0 0.52e0.52e0.65e0.74 0.32

3.0 0.4 0.2

26.0

10 6.0  10þ8 2.0  10þ8 0.6  109 0.2  103 0.2  102 0.5  103

e e e e e e e

4.

Conclusions

In this paper a calcium phosphate precipitation model coupled with the ASM2d was developed in order to create a model able to represent jointly the biological and chemical processes that take place in biological phosphorus removal systems. The results obtained confirmed the model hypothesis of considering that aqueous phase reactions quickly achieve the chemical equilibrium and that aqueous-solid change is kinetically governed. In contrast to other precipitation models developed, which solve the aqueous speciation with kinetic expressions, the proposed model solved the equilibrium with a set of algebraic equations (mass-balance and law of mass action equations). This equilibrium expressions hugely simplifies the numerical solution and the model calibration due to the absence of kinetic parameters to solve the aqueous equilibrium and therefore to be calibrated. The precipitation model was able to reproduce the amorphous calcium phosphate (ACP) and later crystallization to hydroxyapatite (HAP) under different scenarios. The proposed global model successfully characterised the EBPR performance of SBR, including the biological, physical and chemical processes.

Acknowledgments This research work has been supported by the Spanish Research Foundation (MCYT, project PPQ2002-04043-C02-01), which is gratefully acknowledged.

Appendix A. Supplementary information Supplementary information associated with this article can be found in the online version at doi:10.1016/j.watres.2011.04.028.

references

Abbona, F., Lu¨ndager Madsen, H.E., Boistelle, R., 1986. The initial phases of calcium and magnesium phosphates precipitated from solutions of high to medium concentration. Journal of Crystal Growth 74 (3), 581e590. Abbona, F., Lu¨ndager Madsen, H.E., Boistelle, R., 1988. The final phases of calcium and magnesium phosphates precipitated

3752

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 4 4 e3 7 5 2

from solutions of high to medium concentration. Journal of Crystal Growth 89 (4), 592e602. Allison, J.D., Brown, D.S., Novo-Gradac, K.J., 1991. MINTEQA2/ PRODEFA2, A Geochemical Assessment Model for Environmental Systems: Version 3.0. EPA/600/3-91/021. USEPA, Washington, D.C. Arvin, E., 1983. Observations supporting phosphate removal by biologically mediated chemical precipitation e a review. Water Science and Technology 15 (3e4), 43e63. Barat, R., Montoya, T., Seco, A., Ferrer, J., 2005. The role of potassium, magnesium and calcium in the enhanced biological phosphorus removal treatment plants. Environmental Technology 26 (9), 983e992. Barat, R., Montoya, T., Borras, L., Seco, A., Ferrer, J., 2006. Calcium effect on enhanced biological phosphorus removal. Water Science and Technology 53 (12), 29e37. Barat, R., Montoya, T., Borras, L., Ferrer, J., Seco, A., 2008. Interactions between calcium precipitation and the polyphosphate-accumulating bacteria metabolism. Water Research 42 (13), 3415e3424. Barat, R., Bouzas, A., Marti, N., Ferrer, J., Seco, A., 2009. Precipitation assessment in wastewater treatment plants operated for biological nutrient removal: a case study in Murcia, Spain. Journal of Environmental Management 90 (2), 850e857. Batstone, D.J., Keller, J., Angelidaki, I., Kalyuzhnyi, S.V., Pavlostathis, S.G., Rozzi, A., Sanders, W.T.M., Siegrist, H., Vavilin, V.A., 2002. Anaerobic Digestion Model No 1. Scientific and Technical Report 13. IWA Publishing. De Haas, D.W., Wentzel, M.C., Ekama, G.A., 2000. The use of simultaneous chemical precipitation in modified activated sludge systems exhibiting biological excess phosphate removal. Part 2: method development for fractionation of phosphate compounds in activated sludge. Water SA 26 (4), 453e466. Doyle, J.D., Philp, R., Churchley, J., Parsons, S.A., 2000. Analysis of struvite precipitation in real and synthetic liquors. Process Safety and Environmental Protection 78 (B6), 480e488. Doyle, J.D., Parsons, S.A., 2002. Struvite formation, control and recovery. Water Research 36 (16), 3925e3940. Ferrer, J., Seco, A., Serralta, J., Ribes, J., Manga, J., Asensi, E., Morenilla, J.J., Llavador, F., 2008. DESASS: a software tool for designing, simulating and optimising WWTPs. Environmental Modelling & Software 23 (1), 19e26. Filipe, C.D.M., Daigger, G.T., Grady, C.P.L., 2001. pH as a key factor in the competition between glycogen-accumulating organisms and phosphorus-accumulating organisms. Water Environment Research 73 (2), 223e232. Henze, M., Gujer, W., Mino, T., van Loosdrecht, M., 2000. Activated Sludge Models ASM1, ASM2, ASM2d and ASM3. Scientific and Technical Report 9, IWA Task Group on Mathematical Modelling for Design and Operation of Biological Wastewater Treatment. IWA Publishing, London. Kibalczyc, W., Christoffersen, J., Christoffersen, M.R., Zielenkiewicz, A., Zielenkiewicz, A., 1990. The effect of magnesium ions on the precipitation of calcium phosphates. Journal of Crystal Growth 106 (2-3), 355e366. Koutsoukos, P., Amjad, Z., Tomson, M.B., Nancollas, G.H., 1980. Crystallization of calcium phosphates: a constant composition study. Journal of the American Chemical Society 102 (5), 1553e1557. Langerak, van E.P.A., Gonzales, G.G., Aelst, van A., Lier, van J.B., Hamelers, H.V.M., Lettinga, G., 1998. Effects of high calcium concentrations on the development of methanogenic sludge

in upflow anaerobic sludge bed (UASB) reactors. Water Research 32 (4), 1255e1263. Loewenthal, R.E., Wiechers, H.N.S., Marais, G.V.R., 1986. Softening and Stabilisation of Municipal Waters. Water Research Commission, Pretoria. Manga, J., Ferrer, J., Garcı´a-Usach, F., Seco, A., 2001. A modification of the activated sludge model n 2 based on the competition between polyphosphate accumulating organisms and glycogen accumulating organisms. Water Science and Technology 43 (11), 161e171. Maurer, M., Abramovich, D., Siegrist, H., Gujer, W., 1999. Kinetics of biologically induced phosphorus precipitation in wastewater treatment. Water Research 33 (2), 484e493. Moosbrugger, R.E., Wentzel, M.C., Ekama, G.A. and Marais, G.V.R. (1992) Simple Titration Procedures to Determine H2CO3* Alkalinity and Short-chain Fatty Acids in Aqueous Solutions Containing Known Concentrations of Ammonium, Phosphate and Sulphide Weak Acid/Bases. Water Research Commission, Report No. TT 57/92. University of Cape Town, Research Report W 74. Pretoria, Republic of South Africa. Musvoto, E.V., Wentzel, M.C., Lowenthal, R.E., Ekama, G.A., 2000a. Integrated chemical-physical processes modelling e I. Development of kinetic-based model for mixed weak acid/ base systems. Water Research 34 (6), 1857e1867. Musvoto, E.V., Wentzel, M.C., Ekama, G.A., 2000b. Integrated chemical-physical processes modelling e II. Simulating aeration treatment of anaerobic digester supernatants. Water Research 34 (6), 1868e1880. Nancollas, G.H., 1979. Growth of crystals in solution. Advances in Colloid and Interface Science 10, 215e252. Ohlinger, K.N., Young, T.M., Schroeder, E.D., 1998. Predicting struvite formation in digestion. Water Research 32 (12), 3607e3614. Plant, L.J., House, W.A., 2002. Precipitation of calcite in the presence of inorganic phosphate. Colloids and Surfaces A: Physicochemical and Engineering Aspects 203 (1e3), 143e153. Rosen, C., Vrecko, D., Gernaey, K.V., Jeppsson, U., 2005. Implementing ADM1 for benchmark simulation in Matlab/ Simulink. In: Proceedings of the First International Workshop on the IWA Anaerobic Digestion Model No. 1 (ADM1), September 2e4, Lyngby, Denmark, pp. 11e18. Schuler, A.J., Jenkins, D., 2003. Enhanced biological phosphorus removal from wastewater by biomass with different phosphorus contents, part I: experimental results and comparison with metabolic models. Water Environment Research 75 (6), 485e498. Seckler, M.M., Bruinsma, O.S.L., Van Rosmalen, G.M., 1996. Calcium phosphate precipitation in a fluidized bed in relation to process conditions: a black box approach. Water Research 30 (7), 1677e1685. Serralta, J., Ferrer, J., Borras, L., Seco, A., 2004. An extension of ASM2d including pH calculation. Water Research 38 (19), 4029e4038. Standard Methods for the Examination of Water and Wastewater, 20th ed., 1998 American Public Health Association, American Water Works Association and Water Environmental Federation, Washington DC, USA. Udert, K.M., Larsen, T.A., Biebow, M., Gujer, W., 2003. Urea hydrolysis and precipitation dynamics in a urine-collecting system. Water Research 37 (11), 2571e2582. Van Kemenade, M.J.J.M., de Bruyn, P.L., 1987. A kinetic study of precipitation from supersaturated calcium phosphate solutions. Journal of Colloid and Interface Science 118 (2), 564e585.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 5 3 e3 7 6 4

Available at www.sciencedirect.com

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

Identification of potential nitrogenous organic precursors for C-, N-DBPs and characterization of their DBPs formation Huihsien Chang, Chiayang Chen, Genshuh Wang* Institute of Environmental Health, College of Public Health, National Taiwan University, No. 17, Xuzhou Rd., Taipei 10055, Taiwan

article info

abstract

Article history:

Nitrosamines are a class of emerging disinfection by-products (DBPs), which are mainly

Received 7 January 2011

formed when water is treated by chloramination. Nitrosamines are highly carcinogenic

Received in revised form

and are hence a major concern for drinking water supplies. Although dissolved organic

9 April 2011

nitrogen (DON) compounds such as dimethylamine (DMA) have been recognized as

Accepted 17 April 2011

important precursors of nitrosamines, many of them have not been identified, especially

Available online 22 April 2011

those used in consumer products. In this study, nine representative nitrogenous organic compounds with different DON characteristics and structures were selected to react with

Keywords:

free chlorine, chlorine dioxide and monochloramine, respectively, for their DBP formation

Disinfection by-products

characteristics (nitrosamines, trihalomethanes (THMs) and haloacetic acids (HAAs)). It was

Haloacetic acids

found that in addition to DMA, benzyldimethyltetradecylamine (benzalkonium chloride,

Nitrogenous organic precursors

BKC) and 3-(N,N-dimethyloctyl-ammonio)propanesulfonate (3-N,N-DAPSIS) inner salt were

Nitrosamines

potent precursors for carbonated DBPs (C-DBPs) and nitrogenated DBPs (N-DBPs). The DBP

Trihalomethanes

formation potential (DBPFP) tests showed that 1 mM of BKC formed more than 2
105 ng/L of N-nitrosodimethylamine (NDMA) when treated with monochloramine and high levels of C-DBPs (2713  145 mg/L of THMs and 356  5 mg/L of HAAs) when treated with chlorine. 3-N,N-DAPSIS was a less potent DBP precursor: 1 mM of 3-N,N-DAPSIS generated 1155  7 ng/L of NDMA, 1351  66 mg/L of THMs and 188  1 mg/L of HAAs. DMA, 3-N,NDAPSIS and BKC were examined for their DBPFPs at various pH and temperatures to determine the impact of pH and reaction temperature on DBP yields and their formation mechanisms. The results showed that DBP yields apparently increased with rising temperature. However, no consistent correlations were observed between DBPs yields and pH. Bromide shifted the DBP species into brominated DBPs, and this phenomenon was more apparent when BKC was treated with chloramine. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Nitrosamines include nine major compounds: N-nitrosodimethylamine (NDMA), N-nitrosomethylethylamine (NMEA), N-nitrosodiethylamine (NDEA), N-nitrosodi-n-propylamine (NDPA), N-nitrosomorpholine (NMOR), N-nitroso-pyrrolidine (NPYR), N-nitrosopiperidine (NPIP), N-nitrosodi-n-butylamine

(NDBA) and N-nitrosodiphenylamine (NDPHA). Nitrosamines are a group of mutagenic, teratogenic and carcinogenic chemicals and have been classified as probable human carcinogens (USEPA, 2002). Recently, nitrosamines were shown to be a class of emerging disinfection by-products (DBPs) formed during chlorine-based and chloramine-based drinking water treatment. In comparison with traditional carbonated disinfection

* Corresponding author. Present address: Institute of Environmental Health, College of Public Health, National Taiwan University, No. 17, Xuzhou Rd., Room 734, Taipei 10055, Taiwan. Tel.: þ886 2 33668098; fax: þ886 2 23940612. E-mail address: [email protected] (G. Wang). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.027

3754

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 5 3 e3 7 6 4

by-products (C-DBPs), such as trihalomethanes (THMs) and haloacetic acids (HAAs), nitrosamines are far more carcinogenic. NDMA is the dominant species of nitrosamines in drinking water and has received a large amount of attention since it was first detected in drinking water wells near a rocket engine testing facility in Sacramento County, CA, USA. In 1998, high levels of NDMA (about 4
105 ng/L for on site and 2
104 ng/L for off site) were identified in the groundwater (Macdonald, 2002). Since then, drinking water sources subjected to wastewater effluents have been identified as the major sources of nitrosamines. In several studies, wastewater effluents treated with monochloramine (NH2Cl) had NDMA formation potentials (NDMAFPs) from approximately 800 ng/L to thousands of ng/L (Mitch and Sedlak, 2002; Sedlak et al., 2005; Krasner et al., 2009), which were much higher than those observed in drinking water (about sub-ng/L to 10 ng/L). Furthermore, NDMA precursors were not effectively removed in some wastewater treatment plants. These findings indicate that NH2Cl and certain dissolved organic nitrogenous (DON) materials from wastewater are the two key parameters responsible for NDMA formation. In addition to the effluent organic matter (EfOM) and dimethylamine (DMA) as DON precursors (Mitch et al., 2003; Pehlivanoglu-Mantas and Sedlak, 2006), the associations of NDMA precursors with municipal wastewater effluents suggest that consumer products also play an important role for NDMA formation. However, only about 20% of the DON constituents in secondary wastewater effluents have been identified (Grohmann et al., 1998), and the characteristics and typical concentrations of DON precursors have not yet been determined (Lee et al., 2007a; Schmidt and Brauch, 2008; Guo and Krasner, 2009). DON precursors not only produce nitrosamines but also affect the species distributions of currently regulated DBPs (Richardson et al., 1999; Choi and Valentine, 2002) and other unregulated N-DBPs such as halonitromethanes and haloacetamides (Lee et al., 2007b; Chu et al., 2010). They may affect water treatment efficiency by reacting with free chlorine and inorganic chloramines to form organic chloramines (Donnermair and Blatchley, 2003). Therefore, the compositions and properties of DON require attention. In 2007, a preparative fractionation method utilizing three kinds of resins for separation of DON based on hydrophobicity/ hydrophilicity and acidity/basicity was proposed (Leenheer et al., 2007). These fractions included hydrophobic acids (HPOA), hydrophobic bases/neutrals (HPOB/N), amphiphilic acids (AMPA), amphiphilic bases/neutrals (AMPB/N), hydrophilic bases (HPIB), hydrophilic acids/neutrals (HPIA/N) and amino acids (AA). However, the mechanisms and characteristics of nitrosamines formation from the reaction between different DON isolates and common disinfectants have not been further investigated. Currently, there is a scarcity of information available on the correlations between DON, various disinfectants and C-/nitrogenated DBP (N-DBP) formation. Few studies have shown that chloramination with some nitrogen enriched dissolved organic matter isolates generates significant amounts of N-DBPs especially from treated wastewater (Dotson and Westerhoff, 2009). However, other common disinfectants, such as chlorine dioxide (ClO2), and traditional DBPs like HAAs were not measured. Moreover,

a longer contact time for DBPFP tests may provide additional information for C-, N-DBPs formation characteristics. Free chlorine is a widely used disinfectant due to its efficacy and available residuals in distribution systems. Chloramine is employed as an alternative to free chlorine for its ability to reduce C-DBP formation and for its greater stability. Recently, chlorine dioxide has been also promoted for water treatment because it reduces C-DBPs formation and has selective oxidizing power. Nevertheless, all of these disinfectants can react with some precursors to generate DBPs including nitrosamines. Precursors can be readily oxidized into chlorinated intermediates for THMs or further oxidized into polychlorinated intermediates for HAAs. Then these intermediates undergo hydrolysis reactions to form THMs and HAAs at different pH (Xie, 2003). Amines can form NDMA through unsymmetrical dimethylhydrazine (UDMH) and nitrosation pathways (Mitch and Sedlak, 2002). To date, only individual disinfectant effects have been discussed in literatures. The main objective of this study was to identify the potential nitrogenous organic precursors for C-/N-DBPs when treated with free chlorine, monochloramine and chlorine dioxide. In order to investigate the correlations between the surrogates of different DON characteristics and their DBPFPs (including THMFPs, HAAFPs and nitrosamine-FPs), nine representative nitrogenous organic compounds widely applied to consumer products were selected. The DON precursors used in this study included DMA, triethanolamine (TEA), allantoin, melamine, caffeine, 3-(N,N-dimethyloctylammonio)propanesulfonate (3-N,N-DAPSIS) inner salt, benzyldimethyltetradecylamine (benzalkonium chloride, BKC), L-arginine and glycine. Besides DMA, only limited studies were available concerning the NDMA formation from TEA and BKC (Kemper et al., 2010; Lee et al., 2007a; Mitch and Schreiber, 2008; Schreiber and Mitch, 2006). Chlorination tests for L-arginine and glycine have been conducted to determine the chloroform levels (Hureiki et al., 1994). Although certain surrogates have been studied, most of them reported single DBPs formation with specific disinfectants. In addition, further examinations of the precursors were conducted to characterize the effects of pH, reaction temperature, and bromide on DBP yields and formation mechanisms.

2.

Material and methods

2.1.

Selection of DON surrogate compounds

This study selected nine representative nitrogenous organic compounds (Table 1) according to their chemical structures, their physico-chemical properties and their prevalence in water. DMA is a well-known NDMA precursor and was chosen as the comparison compound in this study. TEA for HPIB is primarily used as an emulsifier and surfactant. The HPIA/N fraction included allantoin, which is widely applied in cosmetics and pharmaceuticals production. The HPOB melamine is a common raw material for the production of melamine resin. The model compound of HPOA was caffeine, which is frequently found in varying quantities in food, drinks and many pharmaceutical and personal care products

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 5 3 e3 7 6 4

(PPCPs). Both 3-N,N-DAPSIS and BKC represented AMPB/N compounds. 3-N,N-DAPSIS is used as a detergent due to its amphiphilicity. Quaternary BKC is widely added in disinfectant formulations, such as skin antiseptics and mouthwash. The AA fraction was represented by L-arginine and glycine because of their dominance in raw water (Dotson and Westerhoff, 2009).

2.2.

Materials

The THMs mix standard (1000 mg/mL each in methanol), HAAs mix standard (1000 mg/mL each in methyl tertiary butyl ether), benzene-d6 (1000 mg/mL, internal standard for THMs), 1,2,3trichloropropane (1000 mg/mL, internal standard for HAAs) and 2-bromobutanoic acid (1000 mg/mL, surrogate standard for HAAs) were obtained from Supelco (Bellefonte, PA, USA). The nitrosamines mix standard (2000 mg/mL each in dichloromethane) and NDMA-d6 (1000 mg/mL, internal standard) were purchased from Chem Service (West Chester, PA, USA). The surrogate compounds were obtained from Supelco and SigmaeAldrich (St. Louis, MO, USA). The stock solutions of chlorine were diluted from 6 to 14% sodium hypochlorite (NaOCl) (SigmaeAldrich). Monochloramine (NH2Cl) solutions were freshly prepared via the reaction of ammonium chloride and NaOCl solutions on the day of chloramination. The chlorine dioxide stock solutions were diluted daily from the 0.3% of ClO2 solution prepared by the reaction of the two powder reagents from TwinOxide (De Tongelreep, Netherlands). All concentrations of disinfectant solutions were confirmed before disinfection processes. Other chemicals (buffers, eluents and solvents, etc.) were reagent grade and used without further purification. All solutions for this study were prepared with Milli-Q water (Millipore, Molsheim, France).

2.3.

DBP formation potentials

The THMFP and HAAFP were determined according to the Standard Method 5710 (APHA, 1998). For nitrosamines-FP, the method described by Mitch and Sedlak (2004) was followed. Two phases of DBPFP tests were performed in this study. In the first phase, all surrogate solutions (0.1 mM and 1 mM) were treated with high levels of disinfectants (NaOCl, ClO2 and NH2Cl) and incubated for 7 days (THMFP and HAAFP) or 10 days (nitrosamines-FP) at pH 7.0 and 25  C in dark. In the second phase, 1 mM of DMA, BKC and 3-N,N-DAPSIS were treated with disinfectants and incubated at different pHs (4, 7 and 10) and temperatures (15  C, 25  C and 35  C) in dark. In order to assess the effects of bromide, 1 mM of BKC was spiked with 0, 0.5 and 2 mg/L of Br, and then reacted with NaOCl and NH2Cl. All of the DBPFP tests were conducted in duplicate to monitor the laboratory performance and data quality.

2.4.

Analysis

An ASI 5000 total organic carbon (TOC) analyzer (Shimadzu, Tokyo, Japan) was used to measure the non-purgeable dissolved organic carbon (NPDOC). The UV254 and other absorbances were determined with a UV 160A UV/Visible spectrophotometer (Shimadzu, Tokyo, Japan). Total dissolved nitrogen (TDN) was measured according to the method

3755

described by Wang et al. (2003). The ammonia concentrations were measured with the Merck Ammonium Cell Test kits (Darmstadt, Germany). A DX-120 ion chromatography (Dionex, Sunnyvale, CA, USA) was used to analyze the concentrations of nitrite and nitrate. DON was obtained by subtracting the dissolved inorganic nitrogen (nitrite, nitrate and ammonia) from TDN. Concentrations of THMs were analyzed by a purge and trap system (Model 4660, OI Analytical, Texas, USA) and a gas chromatography-mass spectrometer (GCeMS) (Agilent 6890 GC/5973 MSD) equipped with an RTX-VOC capillary column (60 m
0.32 mm inner diameter, 1.5 mm film thickness). The method detection limits (MDLs) of chloroform, bromodichloromethane, chlorodibromomethane and bromoform were 0.16, 0.18, 0.14, and 0.12 mg/L, respectively. HAA levels were determined by a liquideliquid extraction according to the USEPA method 552.3 (USEPA, 2003). The extracts were quantified by a gas chromatography-electron capture detector (GCECD) (Agilent 6890 GC/micro ECD) with a DB-1701 column (30 m
0.25 mm inner diameter, 0.25 mm film thickness). The MDLs of chloroacetic acid, bromoacetic acid, dichloroacetic acid, trichloroacetic acid, bromochloroacetic acid, bromodichloroacetic acid, dibromoacetic acid, chlorodibromoacetic acid and tribromoacetic acid were 0.17, 0.10, 0.12, 0.08, 0.09, 0.06, 0.07, 0.05 and 0.10 mg/L, respectively. Nitrosamines were analyzed by a liquideliquid extraction, which was modified from the method described by Raksit and Johri (2001), and the same GCeMS system used for THMs determination. Briefly, 500 mL of samples were adjusted to approximately pH 7 and spiked with an NDMA-d6 internal standard. Then two consecutive extractions were performed using first 70 mL and then 30 mL of dichloromethane. The combined extracts were then concentrated to 1 mL using a rotary evaporator (Buchi Rotavapor R-114, Flawil, Switzerland). Finally, the concentrates were analyzed by a GCeMS (with electron ionization source) in the selective ion monitoring mode for quantification. The MDL of NDMA was 20.4 ng/L. The MDLs of the other eight nitrosamines ranged between 3.9 and 47.7 ng/L.

3.

Results and discussion

3.1. Disinfectant dosages determination of surrogate compounds In order to estimate the disinfectant doses for DBPFP tests, 1 mM solutions of each surrogate were prepared. The initial doses of disinfectants (70 mg/L of ClO2 and 70 mg/L as Cl2 for NaOCl and NH2Cl) were added and incubated for 7 days at 25  C in dark. The disinfectant residuals and NPDOC levels were measured after the 7-day contact time. If the residuals were below 5 mg/L, the disinfectant doses were raised until the residuals after the 7-day incubation were satisfied, and the highest doses were 280 mg/L for allantoin and melamine. The final doses applied for each DON surrogate were shown in Table 1.

3.2.

Disinfectant dosages and surrogate compounds

The results in Table 1 indicated that the disinfectant dosages were mainly affected by the structures and branched chains of

Fraction NO Surrogates

Structure

Molecular weight

pKa

Organic contents in 1 mM of surrogate

Disinfectant dosages for DBPFPs

DOC (mg/L)

DON (mg/L)

DOC/DON (mg/mg)

NaOCl (Cl2 mg/L)

ClO2 (ClO2 mg/L)

NH2Cl (Cl2mg/L)

DMA

45.08

10.77

24

14

1.7

140

70

210

HPIB

2

TEA

149.19

7.92

72

14

5.1

210

210

140

HPIA/N

3

Allantoin

158.12

8.96

48

56

0.9

210

70

280

HPOB/N

4

Melamine

126.12

8.00

36

84

0.4

210

140

280

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 5 3 e3 7 6 4

1

3756

Table 1 e - Characteristics and disinfectant doses of the nine DON surrogate compounds.

5

Caffeine

194.19

1.22e3.6

AMPB/N

6

3-N,NDAPSIS

279.45

7

BKC

8

L-Arginine

9

Glycine

AA

96

56

1.7

210

70

140

8.3

156

14

11.1

70

70

210

365.5

5.12

276

14

19.7

70

70

140

174.2

2.17, 9.04, 12.48

72

56

1.3

210

70

140

2.35, 9.78

24

14

1.7

210

70

140

75.07

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 5 3 e3 7 6 4

HPOA

3757

3758

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 5 3 e3 7 6 4

surrogates. When treated with NaOCl, BKC and 3-N,N-DAPSIS required lower doses (70 mg/L) due to the fewer active sites available, such as the amines and aromatic rings. Higher dosages were applied to L-arginine and glycine, which have primary amines that rapidly react with chlorine to form organic chloramines and further products (Hureiki et al., 1994). For the ClO2 treatments, TEA and melamine required higher dosages for DBPFPs. The nitrogen atoms with lone electron pair and electron-rich aromatic moiety of melamine were also served as reactive sites with ClO2 (Navalon et al., 2009). The other seven precursors had lower reactivities and low NPDOC oxidation efficacy (majority were less than 15%). For the chloramination, more than 140 mg/L as Cl2 of NH2Cl was required to conduct the DBPFP tests with suitable residuals. The highest doses of NH2Cl (up to 280 mg/L as Cl2) were used for allantoin and melamine. The three ketone groups in allantoin (Yang et al., 2008) and the reactive aromatic carbon sites of melamine could account for the high NH2Cl dosages. Besides 3-N,N-DAPSIS, DMA rapidly reacts with chloramines to form organic chloramine, so it required a higher dose of NH2Cl. The other five compounds were added with 140 mg/L as Cl2 for DBPFP tests.

3.3. Identification of potential DBP precursors: correlations between surrogate compounds and DBPFPs In DBPFP tests, 0.1 mM and 1 mM of surrogate solutions were prepared to react with the three disinfectants. The species distributions and concentrations of DBPFPs showed that DMA,

BKC and 3-N,N-DAPSIS were important nitrosamine precursors (Table 2). BKC and 3-N,N-DAPSIS were also significant C-DBP precursors.

3.3.1.

Treatments with NaOCl

Among the nine nitrosamines, NDMA was the most abundant species observed in this study. One mM of DMA and BKC could produce more than 40,000 ng/L and 10,000 ng/L of NDMA after chlorination, respectively. 3-N,N-DAPSIS was also a potent N-DBP precursor after chlorination, 1 mM of 3-N,N-DAPSIS generated approximately 200 ng/L of NDMA. It has been shown that chlorine yielded higher levels of traditional C-DBPs like THMs and HAAs (Nikolaou et al., 1999), but the chlorinated C-DBPFPs of DMA were low (w20 mg/L). Although DMA consumed high levels of NaOCl, it was not completely mineralized to CO2 and N2 based on the NPDOC removals (percentage removals ranged from 30% to 60%). The relatively low degradation rate of DMA and low THMs/HAAs formation after chlorination meant that the intermediate products for C-DBPs formation were not formed. Besides the high NDMA levels, 1 mM of BKC generated the highest levels of C-DBPs (2713  145 mg/L of THMs and 356  5 mg/L of HAAs), which indicated that BKC was also a significant precursor of C-DBPs. Similar to BKC, the chlorination of 3-N,N-DAPSIS yielded significant levels of THMs (1351  66 mg/L) and some HAAs (188  1 mg/L).

3.3.2.

Treatments with NH2Cl

After chloramination, much higher amounts of NDMA were obtained for DMA and BKC. One mM of DMA and BKC could

Table 2 e DBPFPs of DON surrogate compounds. Surrogate

Surrogate concentration ¼ 0.1 mM THMFP (mg/L)

HAAFP (mg/L)

NDMAFP (ng/L)

THMFP (mg/L)

HAAFP (mg/L)

NDMAFP (ng/L)

Avg  SD

Avg  SD

Avg  SD

Avg  SD

Avg  SD

Avg  SD

Treatments with NaOCl DMA 22.6  4.0 TEA 25.3  2.3 Allantoin 18.1  1.8 Melamine 24.34  2.2 Caffeine 27.4  11.2 3-N,N-DAPSIS 73.0  18.1a BKC 297.4  10.1a L-Arginine 69.7  0.2 Glycine 21.1  1.0 Treatments with NH2Cl DMA 2.8  0.4 TEA 1.6  0.3 Allantoin 4.7  0.9 Melamine 1.6  0.2 Caffeine 2.0  0.3 3-N,N-DAPSIS 118.7  2.6a BKC 120.1  6.3 L-Arginine 5.2  0.8 Glycine 0.7  0.1 a b c d

Surrogate concentration ¼ 1 mM

16.8 19.1 14.8 15.6 12.3 9.9 50.5 445.4 12.3

        

0.2 0.7 1.3 1.7 1.4 1.2 7.7 24.9a 1.6

8.9  0.3 8.5  0.5 8.2  0.5 9.0  2.8 5.4  0.2 17.6  0.0 85.2  1.7 28.6  0.4 NDb

130.6  12.6
12.6 2.9 1.3 4.2 3.6 1350.8 2713.3 7.5 1.9

        

3.4 0.6 0.2 0.4 1.1 66.3a 144.6a 1.7 2.0

ND 0.7  0.2 2.7  0.2 1.1  0.1 1.3  0.0 108.2  2.8a 332.2  17.5a 4.0  0.2 0.5  0.1

10.3 19.1 4.2 11.2 10.6 188.2 356.0 41.8 10.6

        

0.8 0.3 0.1 0.3 1.8 0.8a 4.6a 4.1 1.1

6.8  0.4 15.9  11.2 7.8  0.1 5.6  0.1 5.0  1.3 13.6  0.7 87.7  0.4 19.8  0.6 ND

The dominant species of DBPs which had relatively higher levels. ND: not detected. LOQ: limit of quantitation. MDL: method detection limits. The MDL of NDMA was 20.4 ng/L. LOQ was defined as 3.3 times of MDL.

43359.8  1984.5a 87.8  1.0
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 5 3 e3 7 6 4

generate more than 2
106 ng/L and 2
105 ng/L of NDMA, respectively. These NDMAFPs were at least one order of magnitude higher than those obtained after chlorination. Stable benzyl radical and quaternary ammonium groups functioning as n-electron donors could explain these elevated NDMA yields of BKC (Kemper et al., 2010). The nucleophilic substitution reaction between disinfectants and BKC cleaved the bond between the benzyl carbon and the quaternary nitrogen atom, resulted in the formation of dimethylalkylamine and tropylium ion. The dimethylalkylamine intermediate not only further form NDMA but also is more hydrophobic and persistent than BKC. Therefore it may elevate the risk of nitrosamines formation (Tezel and Pavlostathis, 2009). 3-N,N-DAPSIS also formed approximately 1100 ng/L of NDMA. The higher NH2Cl dosage and

3759

lower NDMAFP for 3-N,N-DAPSIS indicated that NH2Cl easily reacted with this compound to form products other than nitrosamines. Furthermore, the chloramination formed less C-DBPs (<332 mg/L) for the three surrogates mentioned in this section. Experimental results showed that the two model compounds from the AMPB/N fraction (3-N,N-DAPSIS and BKC) used in this study were potential precursors for both C-DBPs and N-DBPs. This finding suggested that their homologous compounds may also be considered as potential DBP precursors. Besides the two classifications, other commonly used nitrogenous ingredients of zwitterionic detergents or disinfectants should be considered for their potential contributions to DBPs formation in drinking water through wastewater-impacted source water.

Fig. 1 e Correlations between temperatures and DBP formation for identified precursors (n [ 2). (1 mM solutions of each precursor were prepared and the disinfectant dosages were applied at the same levels as those shown in Table 1.)

3760

3.3.3.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 5 3 e3 7 6 4

Treatments with ClO2

Application of chlorine dioxide as disinfectant can reduce the C-DBPs formation in water treatments (Werdehoff and Singer, 1987; Richardson et al., 1994; Li et al., 1996), and the results of DBPFPs for surrogate compounds treated with ClO2 were lower than 32 mg/L (shown in Supplement Information). However, ClO2 also reacted with the three identified precursors to form NDMA. The NDMAFP of both DMA and BKC were higher than 10,000 ng/L and 3-N,N-DAPSIS generated less amounts of NDMA (<56 ng/L). The majority of C-DBPFPs for the other six compounds were less than 30 mg/L except for the HAAFP of 445  25 mg/L for L-arginine treated with chlorine. For certain precursors, unexpected higher C-DBPFPs were observed from 0.1 mM of precursors than from 1 mM of precursors. It has been shown that different molar ratios between precursors and disinfectants affect not only the speciation of intermediates and DBPs but also the levels of DBPs formation (Kim and Clevenger, 2007; Navalon et al., 2009). Most NDMAFPs of the other six compounds were too low to be accurately determined except for the approximate 200 ng/L of NDMA produced from the chloramination of 1 mM TEA. Other nitrosamine-FPs were also measured and could be detected in this study (e.g. NMEA, NDEA, NPYR, NPIP, NDBA and NDPHA), but majority of these showed no correlation with the surrogates due to the low nitrosamine-FPs.

3.4. Characterization of C-, N-DBPs formation for selected precursors 3.4.1.

Effect of temperature

The DBPFPs of DMA, 3-N,N-DAPSIS and BKC showed a significant upward tendency with increasing temperature, especially for C-DBPs (Fig. 1). The THMFPs and HAAFPs obtained at 35  C were up to 21 times and 19 times higher than the levels at 15  C for 3-N,N-DAPSIS treated with NaOCl. Higher temperature not only increased the yields of C-DBPs but also increased the disinfectant consumptions, as shown in Table 3. The disinfectant consumptions at 35  C increased at least 5e10% over those at 15  C for all disinfectants. For BKC, the chlorine consumption at 35  C was 233% higher than at 15  C. The effects of temperature on C-DBP formations were more apparent for 3-N,N-DAPSIS and BKC.

Higher temperature may change precursor properties, disinfectant stabilities or other parameters, as well as influencing the reactions’ activation energies, altering the positions of equilibrium and the reaction rates, to affect the DBP yields. In general, THMs and HAAs were formed within a few days and thus have characteristics different from the NDMA formation that was affected by other complicated factors, as described later. Temperature variation moderately affected the NDMA formation when NaOCl and ClO2 were used (with 1.2e4.6 times higher levels at 35  C than those at 15  C), but the effects were not so apparent after chloramination, which showed a slight decrease in NDMAFP at 35  C. The differences between the chloramination and treatments with NaOCl and ClO2 can be attributed to the higher oxidizing power of NaOCl and ClO2 than NH2Cl. The much faster reaction rates with NaOCl and ClO2 resulted in higher NDMA formation with increasing temperature except for 3-N,N-DAPSIS treated with NaOCl. Surprisingly, the NDMAFP of BKC for chlorination was highest at 15  C. The debenzylation and stability of BKC, which change with temperature, could explain this result (Suzuki et al., 1989; Ding and Liao, 2001). For NDMA formation, NH2Cl consumption was controlled by its autodecomposition rather than by the disinfectant demand (Jafvert and Valentine, 1992; Vikesland et al., 1998), and the NH2Cl residuals after DBPFP tests showed that NH2Cl decayed faster at higher temperatures. In this study, the contact time for NDMAFP to reach a concentration plateaus was around 10 days. The higher temperature (35  C) may enhance the autodecomposition of NH2Cl and thus reduce the active chloramines before the NDMA concentrations could reach its formation potential. During NDMA formation, higher temperatures not only enhance the rate of UDMH production but also the hydrogenolysis of NeN bonds of UDMH and the production of light impurities (Gorji et al., 2006). Thus, a lower yield of UDMH product forms a lower level of NDMA if the NDMA was formed through this pathway. Those factors may explain why the NDMAFPs after chloramination of the three precursors increased from 15  C to 25  C and slightly decreased at 35  C. Some specific products were identified when the precursors were treated with different disinfectants. N-chlorodimethylamine, an organic chloramine, was identified in all

Table 3 e Disinfectant consumptionsa of identified precursors for three disinfectants at different temperatures and pH. Disinfectant consumption Effects of temperature Reaction Temperature DMA 3-N,N-DAPSIS BKC Effects of pH pH DMA 3-N,N-DAPSIS BKC

ClO2 (mol/mol)

NaOCl (mol/mol)

NH2Cl (mol/mol)

15  C 1.78 <0.01 0.30

25  C 1.98 0.01 0.59

35  C 2.00 0.45 1.00

15  C 1.03 0.49 0.29

25  C 1.04 0.47 0.63

35  C 1.04 0.51 0.85

15  C 2.57 2.75 1.86

25  C 2.62 2.84 1.91

35  C 2.71 2.89 1.95

4.0 1.32 0.14 0.29

7.0 1.92 0.18 0.50

10.0 0.93 0.13 0.34

4.0 0.61 0.09 0.00

7.0 1.04 0.51 0.54

10.0 1.03 0.47 0.46

4.0 2.51 0.61 0.64

7.0 2.60 2.86 1.89

10.0 1.43 1.57 0.29

a 1 mM solutions of each precursor were prepared and the disinfectant dosages were applied at the same levels as those shown in Table 1.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 5 3 e3 7 6 4

samples treated with ClO2. DMA treated with NaOCl also generated N-chlorodimethylamine. In addition, chloramination also produced cyanogen chloride (CNCl), which could be considered as a DBP, to demonstrate that other N-DBPs should be addressed while the NDMAFPs were investigated.

3.4.2.

Effect of pH

The DBPFP results showed that pH did not have consistent effects on DBPs formation and disinfectant consumptions for the three DON surrogate compounds (Fig. 2 and Table 3). After chlorination, the THMFP of DMA increased with increasing pH, and the HAAFP of DMA showed an opposite trend. However, both the THMFPs and HAAFPs of BKC and 3N,N-DAPSIS showed the highest levels at pH 7 when treated with NaOCl and NH2Cl. For example, the amounts of HAAFPs

3761

obtained at pH 4 and pH 10 were 51% and 23% for 3-N,NDAPSIS, and 76% and 28% for BKC, respectively, when compared with those at pH 7. Both HAAFPs and NDMAFPs were affected by the pH after chloramination. It could be seen that the solutions generated higher quantities of HAAs at pH 7 and gave lowest HAA formation at pH 4. The NDMAFPs at different pH showed more significant variation. The highest NDMAFP was obtained at pH 7, and a dramatic decrease was observed when the reaction was conducted at pH 10. The NDMAFPs after chloramination at pH 10 were about 9e19% of those at pH 7. Chlorine dioxide produced minor levels of THMs (mostly less than 32 mg/L) for each precursor at different pH. However, both HAA (from 2 to 9 mg/L) and NDMA yields for each surrogate treated with ClO2 were about the same at all pH values.

Fig. 2 e Correlations between pH values and DBP formation for identified precursors (n [ 2). (1 mM solutions of each precursor were prepared and the disinfectant dosages were applied at the same levels as those shown in Table 1.)

3762

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 5 3 e3 7 6 4

Fig. 3 e Correlations between BrL concentrations and DBP formation for BKC (n [ 2): (a) THMs and (b) HAAs. ([BKC]o [ 1 mM, [NaOCl] o [ 70 mg/L as Cl2, [NH2Cl] o [ 140 mg/L as Cl2, pH [ 7, Temperature [ 25  C.)

Speciation and solubility of precursors, stabilities of intermediates and the oxidizing power of disinfectants could lead to distinct behaviors at different pH values. The three target compounds’ pKa were 10.77, 8.3 and 5.12 for DMA, 3-N,NDAPSIS and BKC, respectively, and the compounds dissociate when the solution’s pH is higher than their pKa. Better dissolutions or deprotonated species that are stronger electron donors could enhance their reaction efficiency with disinfectants and increase the DBP formation. A recent study suggested that NDMA formation was dependent on pH under different ratios of monochloramine and DMA (Kim and Clevenger, 2007). Formation of intermediates is also a function of pH. UDMH had a slower formation rate at pH <8, and other products of UDMH oxidation may be favored at higher pH (Mitch and Sedlak, 2002). Some intermediates could even promote the organocatalyst of ClO2 decomposition and limit the subsequent DBP formation (Navalon et al., 2008). The pH values could also alter the speciation and oxidizing power of disinfectants. Chlorine at lower pHs provides more effective oxidizing ability. Despite the uncertainty of how pH will affect the efficacy of ClO2, pH values have been found to have strong effect on the disproportionation and disinfection efficacy of ClO2. In addition, NH2Cl autodecomposition increases with increasing pH and ultimately results in the oxidation of ammonia and reduction of active chlorine. At pH 10, the less active NH2Cl may limit NDMA formation. At pH 4, the dichloramine predominance contributes to faster and higher NDMA formation compared to monochloramine (Schreiber and Mitch, 2006). Moreover, the variation in the degree of autodecomposition generates different amounts of Cl and affects the DBP speciations via different reaction preferences. The DBPs produced after chloramination are similar to those formed with chlorination. This may be a result of an interaction between precursors and Cl from the hydrolysis of NH2Cl. Conversely, the direct interaction between NH2Cl and precursors generated the DBPs species primarily observed in chloraminated systems (Vikesland et al., 1998).

3.4.3.

Effect of bromide concentration

The presence of bromide in water sources reacts with chlorine or chloramines to form more reactive bromine or bromoamines and thus alter the DBPs speciations. The effects of bromide on THM and HAA formation for BKC were shown in Fig. 3. The C-DBPFPs decreased with increasing bromide concentrations when treated with NaOCl, however, this observation was not shown for NH2Cl treatments. This trend was different from the general observations (Diehl et al., 2000; Liang and Singer, 2003). It is inferred that the more reactive bromine may react with BKC to form other intermediates or products rather than DBPs. Further investigations with target compounds like BKC are necessary. Fig. 3 also showed that the presence of bromide shifted the DBPs species into brominated DBPs. When 2 mg/L of bromide was spiked, the brominated DBPs accounted for about 50% of THMs and HAAs. Moreover, NH2Cl generated a higher proportion of brominated DBPs than those obtained in NaOCl treatments with the same levels of bromide.

4.

Conclusions

In this study, nine representative nitrogenous organic compounds were treated with three widely used disinfectants (free chlorine, chlorine dioxide and monochloramine). Among the nine selected precursors, DMA, BKC and 3-N,N-DAPSIS were found to be potent precursors for both C-DBPs and N-DBPs. NDMA was the dominant species of nitrosamines, and the levels of other nitrosamines were too low to show specific formation characteristics following treatments. BKC was a significant DBP precursor and formed more than 2
105 ng/L of NDMA after chloramination, as well as high levels of C-DBPs (2713  145 mg/L of THMs and 356  5 mg/L of HAAs) after chlorination. 3-N,N-DAPSIS generated fewer DBPs, which contained 1155  7 ng/L of NDMA after chloramination, and 1351  66 mg/L of THMs and 188  1 mg/L of HAAs after chlorination. It was observed that the DBPs yields apparently

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 5 3 e3 7 6 4

showed an upward tendency with increasing temperature. However, no consistent trends of DBPs formation were found for DBPFPs conducted at different pH due to the complex interactions and mechanisms between the surrogate compounds and the disinfectants. The presence of bromide shifted the DBPs species into brominated DBPs, and treatment with NH2Cl generated a higher proportion of brominated DBPs than those obtained with NaOCl. This study attempted to propose an approach to estimate the predominant and simultaneous C-DBPs and N-DBPs formation through the reactions between selected precursors and various disinfectants. Further investigations will be needed to elucidate DBP formation from various groups of surrogate compounds or substances in real water or wastewater samples.

Acknowledgments This work was supported by the National Science Council of Taiwan, R.O.C. (gs1) (NSC99-2221-E-002-066-MY3).

Appendix. Supplementary information Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.04.027.

references

American Public Health Association (APHA), 1998. Standard Methods for the Examination of Water and Wastewater, twentieth ed. American Water Works Association & Water Environment Federation, APHA-AWWA-WEF, Washington, D. C., USA. Choi, J.H., Valentine, R.L., 2002. Formation of Nnitrosodimethylamine (NDMA) from reaction of monochloramine: a new disinfection by-product. Water Res. 36 (4), 817e824. Chu, W.H., Gao, N.Y., Deng, Y., Krasner, S.W., 2010. Precursors of dichloroacetamide, an emerging nitrogenous DBP formed during chlorination or chloramination. Environ. Sci. Technol. 44 (10), 3908e3912. Diehl, A.C., Speitel, G.E., Symons, J.M., Krasner, S.W., Hwang, S.J., Barrett, S.E., 2000. DBP formation during chloramination. J. Am. Water Works Assoc. 92 (6), 76e90. Ding, W.H., Liao, Y.H., 2001. Determination of alkylbenzyldimethylammonium chlorides in river water and sewage effluent by solid phase extraction and gas chromatography mass spectrometry. Anal. Chem. 73 (1), 36e40. Donnermair, M.M., Blatchley III, E.R., 2003. Disinfection efficacy of organic chloramines. Water Res. 37, 1557e1570. Dotson, A., Westerhoff, P., 2009. Occurrence and removal of amino acids during drinking water treatment. J. Am. Water Works Assoc. 101 (9), 101e115. Gorji, M., Kazemeini, M., Bozorgmehri, R., 2006. Kinetic investigation of NDMA to UDMH hydrogenation on a Pd/C catalyst. Iran. J. Sci. Technol. Trans. B-Eng. 30 (B5), 581e593. Grohmann, K., Gilbert, E., Eberle, S.H., 1998. Identification of nitrogen-containing compounds of low molecular weight in effluents of biologically treated municipal wastewater. Acta Hydrochim Hydrobiol 26 (1), 20e30.

3763

Guo, Y.C., Krasner, S.W., 2009. Occurrence of primidone, carbamazepine, caffeine, and precursors for Nnitrosodimethylamine in drinking water sources impacted by wastewater. J. Am. Water Resour. Assoc. 45 (1), 58e67. Hureiki, L., Croue, J.P., Legube, B., 1994. Chlorination studies of free and combined amino-acids. Water Res. 28 (12), 2521e2531. Jafvert, C.T., Valentine, R.L., 1992. Reaction scheme for the chlorination of ammoniacal water. Environ. Sci. Technol. 26 (3), 577e586. Kemper, J.M., Walse, S.S., Mitch, W.A., 2010. Quaternary amines as nitrosamine precursors: a role for consumer products? Environ. Sci. Technol. 44 (4), 1224e1231. Kim, J., Clevenger, T.E., 2007. Prediction of Nnitrosodimethylamine (NDMA) formation as a disinfection byproduct. J. Hazard. Mater. 145 (1e2), 270e276. Krasner, S.W., Westerhoff, P., Chen, B.Y., Rittmann, B.E., Nam, S. N., Amy, G., 2009. Impact of wastewater treatment processes on organic carbon, organic nitrogen, and DBP precursors in effluent organic matter. Environ. Sci. Technol. 43 (8), 2911e2918. Lee, C., Schmidt, C., Yoon, J., von Gunten, U., 2007a. Oxidation of N-nitrosodimethylamine (NDMA) precursors with ozone and chlorine dioxide: kinetics and effect on NDMA formation potential. Environ. Sci. Technol. 41 (6), 2056e2063. Lee, W., Westerhoff, P., Croue, J.P., 2007b. Dissolved organic nitrogen as a precursor for chloroform, dichloroacetonitrile, N-Nitrosodimethylamine, and trichloronitromethane. Environ. Sci. Technol. 41 (15), 5485e5490. Leenheer, J.A., Dotson, A., Paul, W., 2007. Dissolved organic nitrogen fractionation. Ann. Environ. Sci. 1, 45e56. Li, J.W., Yu, Z.B., Cai, X.P., Gao, M., Chao, F.H., 1996. Trihalomethanes formation in water treated with chlorine dioxide. Water Res. 30 (10), 2371e2376. Liang, L., Singer, P.C., 2003. Factors influencing the formation and relative distribution of haloacetic acids and trihalomethanes in drinking water. Environ. Sci. Technol. 37 (13), 2920e2928. Macdonald, A., 2002. Perchlorate and NDMA contamination in the Sacramento area. In: Presentation at the Fourth Symposium in the Series on Groundwater Contamination: Perchlorate and NDMA Contamination in Groundwater: Occurrence, Analysis and Treatment; Groundwater Resources Association of California, April 17, Baldwin Park, C.A. Mitch, W.A., Gerecke, A.C., Sedlak, D.L., 2003. A Nnitrosodimethylamine (NDMA) precursor analysis for chlorination of water and wastewater. Water Res. 37 (15), 3733e3741. Mitch, W.A., Schreiber, I.M., 2008. Degradation of tertiary alkylamines during chlorination/chloramination: implications for formation of aldehydes, nitriles, halonitroalkanes, and nitrosamines. Environ. Sci. Technol. 42 (13), 4811e4817. Mitch, W.A., Sedlak, D.L., 2002. Formation of Nnitrosodimethylamine (NDMA) from dimethylamine during chlorination. Environ. Sci. Technol. 36 (4), 588e595. Mitch, W.A., Sedlak, D.L., 2004. Characterization and fate of Nnitrosodimethylamine precursors in municipal wastewater treatment plants. Environ. Sci. Technol. 38 (5), 1445e1454. Navalon, S., Alvaro, M., Garcia, H., 2008. Reaction of chlorine dioxide with emergent water pollutants: product study of the reaction of three beta-lactam antibiotics with ClO2. Water Res. 42 (8e9), 1935e1942. Navalon, S., Alvaro, M., Garcia, H., 2009. Chlorine dioxide reaction with selected amino acids in water. J. Hazard. Mater. 164 (2e3), 1089e1097. Nikolaou, A.D., Kostopoulou, M.N., Lekkas, T.D., 1999. Organic byproducts of drinking water chlorination. Global Nest 1 (3), 143e156. Pehlivanoglu-Mantas, E., Sedlak, D.L., 2006. Wastewater-derived dissolved organic nitrogen: analytical methods,

3764

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 5 3 e3 7 6 4

characterization, and effects e a review. Crit. Rev. Environ. Sci. Technol. 36 (3), 261e285. Raksit, A., Johri, S., 2001. Determination of Nnitrosodimethylamine in environmental aqueous samples by isotope-dilution GC/MS-SIM. J. AOAC Int. 84 (5), 1413e1419. Richardson, S.D., Thruston, A.D., Collette, T.W., Patterson, K.S., Lykins, B.W., Majetich, G., Zhang, Y., 1994. Multispectral identification of chlorine dioxide disinfection by-products in drinking-water. Environ. Sci. Technol. 28 (4), 592e599. Richardson, S.D., Thruston Jr., A.D., Caughran, T.V., Chen, P.H., Collette, T.W., Floyd, T.L., 1999. Identification of new drinking water disinfection byproducts formed in the presence of bromide. Environ. Sci. Technol. 33 (19), 3378e3383. Schmidt, C.K., Brauch, H.J., 2008. N, N-dimethosulfamide as precursor for N-nitrosodimethylamine (NDMA) formation upon ozonation and its fate during drinking water treatment. Environ. Sci. Technol. 42 (17), 6340e6346. Schreiber, I.M., Mitch, W.A., 2006. Nitrosamine formation pathway revisited: the importance of chloramine speciation and dissolved oxygen. Environ. Sci. Technol. 40 (19), 6007e6014. Sedlak, D.L., Deeb, R.A., Hawley, E.L., Mitch, W.A., Durbin, T.D., Mowbray, S., Carr, S., 2005. Sources and fate of nitrosodimethylamine and its precursors in municipal wastewater treatment plants. Water Environ. Res. 77 (1), 32e39. Suzuki, S., Nakamura, Y., Kaneko, M., Mori, K., Watanabe, Y., 1989. Analysis of benzalkonium chlorides by gaschromatography. J. Chromatogr. 463 (1), 188e191.

Tezel, U., Pavlostathis, S.G., 2009. Transformation of benzalkonium chloride under nitrate reducing conditions. Environ. Sci. Technol. 43 (5), 1342e1348. U.S. Environmental Protection Agency (USEPA), 2002. Integrated Risk Information System. Office of research and development (ORD), National Center for Environmental Assessment. http:// cfpub.epa.gov/ncea/iris/index.cfm?fuseaction¼iris. showSubstanceList&list_type¼alpha&view¼N. U.S. Environmental Protection Agency (USEPA), 2003. Method 552. 3. Determination of Haloacetic Acids and Dalapon in Drinking Water by Liquideliquid Microextraction, Derivatization, and Gas Chromatography with Electron Capture Detection EPA 815-B-03e002. Vikesland, P.J., Ozekin, K., Valentine, R.L., 1998. Effect of natural organic matter on monochloramine decomposition: pathway elucidation through the use of mass and redox balances. Environ. Sci. Technol. 32 (10), 1409e1416. Wang, H., Zang, H., Zhao, X., 2003. Determination of the total nitrogen in waste water with 2,6-dimethylphenol by spectrophotometry. Chem. Anal. Meterage 12 (6), 20e21. Werdehoff, K.S., Singer, P.C., 1987. Chlorine dioxide effects on THMFP, TOXFP, and the formation of inorganic by-products. J. Am. Water Works Assoc. 79 (9), 107e113. Xie, Y.F., 2003. Disinfection Byproducts in Drinking Water: Formation, Analysis, and Control. Lewis Publishers, pp. 24e25. Yang, X., Shang, C., Lee, W., Westerhoff, P., Fan, C., 2008. Correlations between organic matter properties and DBP formation during chloramination. Water Res. 42, 2329e2339.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 6 5 e3 7 7 5

Available at www.sciencedirect.com

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

Monitoring of water quality from roof runoff: Interpretation using multivariate analysis C. Vialle a,b,*, C. Sablayrolles a,b, M. Lovera c, S. Jacob d, M.-C. Huau e, M. Montrejaud-Vignoles a,b a

Universite´ de Toulouse, INP, LCA (Laboratoire de Chimie Agro-Industrielle), ENSIACET, 4 Alle´es Emile Monso, F-31030 Toulouse, France INRA, LCA (Laboratoire de Chimie Agro-Industrielle), F-31029 Toulouse, France c Veolia Water North America, Technical Direction Group, 101, W Washington Street, Suite 1440 East, IN-46204 Indianapolis, USA d Veolia Eau, Direction Technique, Immeuble Giovanni Battista B, 1, rue Giovanni Battista Pirelli, F-94410 Saint-Maurice, France e Veolia Eau, Direction des collectivite´s publiques, 36-38 avenue Kleber, F-75016 Paris, France b

article info

abstract

Article history:

The quality of harvested rainwater used for toilet flushing in a private house in the south-

Received 3 November 2010

west of France was assessed over a one-year period. Temperature, pH, conductivity, colour,

Received in revised form

turbidity, anions, cations, alkalinity, total hardness and total organic carbon were screened

16 March 2011

using standard analytical techniques. Total flora at 22
C and 36
C, total coliforms,

Accepted 17 April 2011

Escherichia coli and enterococci were analysed. Overall, the collected rainwater had good

Available online 23 April 2011

physicochemical quality but did not meet the requirements for drinking water. The stored rainwater is characterised by low conductivity, hardness and alkalinity compared to mains

Keywords:

water. Three widely used bacterial indicators - total coliforms, E. coli and enterococci - were

Rainwater harvesting system

detected in the majority of samples, indicating microbiological contamination of the water.

Physicochemical quality

To elucidate factors affecting the rainwater composition, principal component analysis

Microbiological quality

and cluster analysis were applied to the complete data set of 50 observations. Chemical

Principal component analysis

and microbiological parameters fluctuated during the course of the study, with the highest

Cluster analysis

levels of microbiological contamination observed in roof runoffs collected during the summer. E. coli and enterococci occurred simultaneously, and their presence was linked to precipitation. Runoff quality is also unpredictable because it is sensitive to the weather. Cluster analysis differentiated three clusters: ionic composition, parameters linked with the microbiological load and indicators of faecal contamination. In future surveys, parameters from these three groups will be simultaneously monitored to more accurately characterise roof-collected rainwater. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Currently, the availability of fresh water is one of the major issues facing the human population. A number of complex factors are driving this issue, including population growth, urbanisation, land use transformation, and pollution.

Insufficient availability to drinking water could lead to devastating consequences, such as increasing health problems or social upheaval. Although many solutions have been proposed, there is much interest in the use of roof-collected rainwater. This practice has been used in many countries for thousands of

* Corresponding author. ENSIACET - LCA, 4 Alle´e Emile Monso, BP 44362, F-31030 Toulouse, France. Tel.: þ335 34 32 35 51; fax: þ335 34 32 35 97. E-mail address: [email protected] (C. Vialle). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.029

3766

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 6 5 e3 7 7 5

years (Pinfold et al. (1993); Simmons et al. (2001)). Although this solution is attractive from an ecological point of view, it is necessary to measure the quality of harvested rainwater due to the potential for health risks as a result of chemical and microbiological contaminants. Recently, studies in numerous countries, including Thailand, USA, Nigeria, New-Zealand, India, Zambia, Brazil, Canada, Australia, Jordan, New Guinea and South Korea, have investigated the quality of harvested rainwater (Pinfold et al. (1993); Crabtree et al. (1996); Uba and Aghogho (2000); Simmons et al. (2001); Handia et al. (2003); Kulshrestha et al. (2003); May and Prado (2006); Al-Khashman (2009); Despins et al. (2009); Evans et al. (2009); Horak et al. (2010); Lee et al. (2010)). In Europe, rainwater quality assessment was studied by Fo¨rster (1998), Fo¨rster (1999), Albrechtsen (2002), Polkowska et al. (2002), Fewtrell and Kay (2007), Melidis et al. (2007), Oesterholt et al. (2007), Sazakli et al. (2007), Schriewer et al. (2008), Tsakovski et al. (2010). Although a number of studies have found collected rainwater to be nonpotable, showing unacceptable levels of microbiological contamination and poor physicochemical qualities, “a clear consensus on the quality and health risk associated with roofcollected rainwater has not been reached” (Evans et al. (2006)). This lack of consensus is likely due to the fact that the qualities of harvested and stored rainwater are dependent on a number of factors, including geographical location, catchment area, storage time, handling and management of the water (Vazquez et al. (2003); Chang et al. (2004); Zhu et al. (2004); Evans et al. (2007); Huston et al. (2009); Lye (2009)). Like most countries, France must spare its water resources and, for several years, has taken an interest in harvesting rainwater for domestic use. Despite reluctance from authorities (C.S.H.P.F (2006)), the increasing demand from private customers has prompted authorization of the use of rainwater for certain applications. Currently, French law prohibits the use of harvested rainwater for drinking, showering or bathing, or washing clothes (Decree of August 21 (2008)). In recent years, the application of multivariate analysis to complex datasets has enjoying a high level of scientific interest. One of the main advantages of these techniques, such as principal component analysis (PCA) and cluster analysis (CA), is the ability to analyse large datasets containing many variables and experimental units. PCA and CA identify groups and sets of variables with similar properties and may allow us to simplify our description of observations by finding the structure or

patterns in chaotic or confusing datasets. Additionally, these techniques allow the analysis of data from non-homogeneous variables. Thus, multivariate methods are now used in a variety of scientific disciplines. In the field of rainwater, they have been applied to the study of precipitations composition mainly in ions and sometimes in metals (Zhang et al. (1992); Hu et al. (2003); Simeonov et al. (2003); Vazquez et al. (2003); Zunckel et al. (2003); Astel et al. (2004); Khare et al. (2004); Baez et al. (2007)) or to the study of rainwater monitoring network (Mantovan et al. (1995); Ouyang (2005)). By contrast, our study concerned the characterization of stored roof-collected rainwater and the simultaneous analysis of chemical, physical and microbiological parameters. The purpose of the present study is to monitor physicochemical and microbiological parameters of collected rainwater and to use PCA and CA to further characterise associations present in the complete data set. Rainwater was collected over a one-year period using a commercially available system installed in south-western France.

2.

Material and methods

2.1.

Sampling site

2.1.1.

Rainwater harvesting system

A commercially available domestic rainwater collection system (Sotralentz Habitat) was installed in a rural village in south-western France. The house was occupied by a family consisting of two parents and two children. The average rainfall in this region is 760 mm, and the average temperatures range from 7.9 to 18.3
C. In the system installed, rainwater is first collected from a 204 m2 surface area of tiled roof and then channelled via open zinc gutters through pipes to an underground PEHD storage tank with a 5 m3 capacity. Prior to entering the tank, the water is passed through a mesh. In the event of an overflow, excess water is fed into a nearby canal. A submerged intake with an inlet filter attached to a float is used to pump water inside the house. Prior to use, collected rainwater is treated by being passed through a physical filter (25 mm) and an activated carbon filter. When insufficient water is available in the tank, a probe activates a valve to allow pumping from a backup drinking water tank. The collected rainwater could be used for toilet flushing and watering gardens; it supplied water for two WCs and an outside tap. A schematic of the rainwater collection system is shown in Fig. 1.

Table 1 e Physico-chemical parameters analysed. Parameter

Fig. 1 e Schematic of the rainwater harvesting system installed in south-western France.

pH Conductivity Turbidity Cl, SO42-, NO3, PO43Mg2þ, Ca2þ, Naþ, Kþ, NH4þ Colour Total organic carbon (TOC) Total hardness Alkalinity (AT, TAC)

Norm NF NF NF NF NF NF NF NF NF

T 90-008 EN 27888 EN ISO 7027 EN ISO 10304-1 EN ISO 14911 EN ISO 7887 EN 1484 EN ISO 14911 EN ISO 9963-1

3767

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 6 5 e3 7 7 5

Table 2 e Descriptive statistics for the dataset. Variables

pH Temperature Conductivity Colour Turbidity TOC Hardness AT CAT Cl SO42NO3 PO43Mg2þ Ca2þ Naþ Kþ NH4þ Total coliforms Escherichia coli Enterococci Total flora at 22
C Total flora at 36
C

2.1.2.

Units

Observations

Minimum

Maximum

Mean

Median

Standard deviation

French drinking water guidelines

e C mS.cm1 mg Pt.L1 NTU mg.L1 mmol.L1 mmol.L1 mmol.L1 mg.L1 mg.L1 mg.L1 mg.L1 mg.L1 mg.L1 mg.L1 mg.L1 mg.L1 ufc per 100 mL ufc per 100 mL ufc per 100 mL ufc per mL ufc per mL

55 55 55 55 53 55 55 55 55 54 54 54 54 54 54 54 54 54 40 53 54 52 51

5.6 7.8 13.5 <5 0.50 0.50 <0.01 <0.20 <0.40 0.55 0.50 0.54 <0.10 <0.10 1.0 0.30 0.15 <0.10 <10 <10 <10 10 25

10.4 22.4 235.0 39 6.1 5.1 0.58 0.9 1.1 4.0 6.6 7.8 0.54 0.71 19 2.9 4.9 1.7 >10 000 5 500 >10 000 632 000 368 000

6.5 14.9 56.2 18 2.4 2.3 0.16 0.10 0.30 1.9 1.9 2.8 0.17 0.27 4.4 1.1 1.2 0.58 656 148 322 45 486 26 651

6.2 13.5 38.2 19 2.0 2.2 0.11 <0.20 0.30 1.7 1.8 2.4 0.19 0.24 2.9 0.93 0.78 0.32 40 2 45 9 700 4 500

1.1 4.8 45.5 10 1.4 1.0 0.13 0.20 0.30 0.98 0.92 1.6 0.14 0.15 4.0 0.59 1.1 0.57 2 189 757 1 359 108 954 67 906

6.5 to 9 25 180 to 1000 2 15




Sample collection

Sampling was carried out weekly from January 2009 to January 2010. To monitor water quality, grab samples were taken from the surface water in the tank each week. The sampling was performed using a sampling rod and a beaker. Prior to sampling, the beaker was disinfected with ethanol, rinsed once with ultra high quality water and then rinsed twice with tank water. Samples were stored in polyethylene bottles for chemical analysis or individual sterile bottles for microbiological analysis. Temperature, pH and conductivity were measured in situ, and the sample was stored in a chilled coldbox during transportation to the laboratory, where samples

250 250 50

200 0.10

were stored at 4
C. Microbiological analyses were conducted within 24 h. Samples collected to assess chemical parameters were frozen for later analysis.

2.2.

Analytical determinations

2.2.1.

Chemical analysis

Samples were measured for pH, conductivity, colour, turbidity, total organic carbon (TOC), ionic composition (Cl, SO42, NO3, PO43,Mg2þ, Ca2þ, Naþ, Kþ, NH4þ), total hardness, simple alkalimetric title (AT), complete alkalimetric title (CAT). Samples were analysed in accordance with norms

11

10

pH

9

8

7

6

06/01/09 13/01/09 20/01/09 27/01/09 03/02/09 10/02/09 17/02/09 24/02/09 03/03/09 10/03/09 17/03/09 24/03/09 31/03/09 07/04/09 14/04/09 21/04/09 28/04/09 05/05/09 12/05/09 19/05/09 26/05/09 02/06/09 09/06/09 16/06/09 23/06/09 30/06/09 07/07/09 14/07/09 21/07/09 28/07/09 04/08/09 11/08/09 18/08/09 25/08/09 01/09/09 08/09/09 15/09/09 22/09/09 29/09/09 06/10/09 13/10/09 20/10/09 27/10/09 03/11/09 10/11/09 17/11/09 24/11/09 01/12/09 08/12/09 15/12/09 22/12/09 29/12/09 05/01/10 12/01/10 19/01/10 26/01/10

5

Fig. 2 e pH values observed during the sampling period (January 2009eJanuary 2010). - The highest pH of 10.4 was recorded In January after a violent storm and remained elevated for five weeks before returning to a slightly acidic condition.

3768

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 6 5 e3 7 7 5

1 000 >10 000 and 5 500

Enterococci 900 Concentrations (CFU per 100mL)

Escherichia Coli 800 700 600 500 400 300 200 100 0 06/01/09 13/01/09 20/01/09 27/01/09 03/02/09 10/02/09 17/02/09 24/02/09 03/03/09 10/03/09 17/03/09 24/03/09 31/03/09 07/04/09 14/04/09 21/04/09 28/04/09 05/05/09 12/05/09 19/05/09 26/05/09 02/06/09 09/06/09 16/06/09 23/06/09 30/06/09 07/07/09 14/07/09 21/07/09 28/07/09 04/08/09 11/08/09 18/08/09 25/08/09 01/09/09 08/09/09 15/09/09 22/09/09 29/09/09 06/10/09 13/10/09 20/10/09 27/10/09 03/11/09 10/11/09 17/11/09 24/11/09 01/12/09 08/12/09 15/12/09 22/12/09 29/12/09 05/01/10 12/01/10 19/01/10 26/01/10

LOQ<10

Fig. 3 e Concentrations of E. coli and enterococci observed during the sampling period (January 2009eJanuary 2010). - E. coli and enterococci were simultaneously present in samples, always with enterococci having the higher concentrations.

shown in Table 1. Ionic composition was analysed with ion chromatography (Dionex, AG/AS 18, ICS 2000 for anions and CG/CS 12, ICS 3000 for cations).

corresponding eigenvectors; and 3) the elimination of components that account only for a small proportion of the variation in datasets.

2.2.2.

2.3.2.

Microbiological analysis

All samples were examined for the three widely used bacterial indicators using the relevant ISO (International Organization for Standardization) standards: ISO 9308-1 for total coliforms and Escherichia coli, and ISO 7899-2 for enterococci, as well as ISO 6222 for total flora at 22
C and 36
C.

2.3.

Multivariate data analysis

In this study, multivariate chemometric techniques were performed using the commercial software XL stat.

2.3.1.

Cluster analysis (CA)

Cluster Analysis (CA) nicely complements PCA. It was used to search for natural groupings among objects and discover latent structures present in the data. Analysed parameters

Principal component analysis

Principal Component Analysis (PCA) is one of the most applied approaches in the environmetrics to study data structures. It is aimed at finding and interpreting hidden complex and casually determined relationships between dataset features. This is accomplished by studying the data structure in a reduced dimension while retaining the maximum amount of variability present in the data. To do this, it is necessary to estimate the number of significant components present in the data. More precisely, a matrix of pairwise correlations among parameters is decomposed into eigenvectors, which, in turn, are sorted in descending order of their corresponding eigenvalues. At this point, the raw data are generally unsuitable for statistical analyses due to differences in the sizes of the variables. Mathematically, PCA normally involves three major steps: 1) the standardisation of measurements to ensure that they have equal weights in the analysis by autoscaling the data to produce new variables, where the mean is equal to zero and the standard deviation is equal to the unit; 2) calculation of the covariance matrix by identifying the eigenvalues and their

Table 3 e Loadings of the first three eigenvectors, F1, F2 and F3, in the total data set. Variable

F1

F2

F3

pH Temperature Conductivity Coulour Turbidity TOC Hardness AT CAT Cl SO42NO3 PO43Mg2þ Ca2þ Naþ Kþ NH4þ Escherichia coli Enterococci Total flora at 22
C Total flora at 36
C

0.318 0.104 0.335 0.014 0.178 0.156 0.351 0.115 0.234 0.256 0.158 0.149 0.011 0.225 0.368 0.270 0.358 0.041 0.117 0.113 0.013 0.015

0.138 0.377 0.048 0.235 0.265 0.400 0.062 0.155 0.049 0.083 0.077 0.001 0.384 0.193 0.074 0.102 0.018 0.414 0.110 0.099 0.194 0.276

0.219 0.065 0.089 0.026 0.172 0.062 0.120 0.184 0.190 0.146 0.273 0.316 0.130 0.187 0.102 0.147 0.096 0.085 0.455 0.447 0.284 0.197

Eignevalue (%)

28.47

20.40

11.13

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 6 5 e3 7 7 5

were sorted into groups, or clusters, so that the degree of association is strong between members of different clusters. Prior to CA, the descriptor variables were block standardised by range to avoid effects of scale or units on the distance measurements. Hierarchical agglomerative CA was performed on the normalised data set with the Ward’s method, using Euclidean distances as a measure of similarity.

Results and discussion

3.1.

Descriptive statistics

Prior to multivariate analysis, univariate descriptive statistics were used to compare the measured variables with French drinking water guidelines (Decree of January 11 (2007)). Indeed

Variables (axes F1 et F2 : 48,87 %)

F2 (20,40 %)

a

3.

3769

F1 (28,47 %) Variables (axes F1 et F3 : 39,60 %)

F3 (11,13 %)

b

F1 (28,47 %)

Fig. 4 e The square cosines for all variables in a) components F1 and F2 and b) components F1 and F3. - A variable is increasingly well represented by a component as the corresponding value of the square cosine approaches the unit. Almost all variables are well represented by the first three components.

3770

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 6 5 e3 7 7 5

250

Dissimilarity

200

150

A

C

B

100

3-

PO4

Colour

NH4

+

COT

TF 36°C

Turbidity

TF 22°C

E. Coli

Enterococci

AT

CAT

+

K

Ca

2+

Conductivity

pH

Hardness

+

Na

-

Cl

2-

SO4

2+

Mg

NO3

-

0

Temperature

50

Fig. 5 e A dendrogram obtained by application of the Ward’s method. e Three clusters were identified corresponding to the three principal components identified with PCA.

as it was impossible to present the complete data set which corresponds to 23 parameters for 55 samplings, minimum, maximum, mean, median and standard deviation were used to describe it (Table 2). Values less than the quantification limit were considered to be zero for statistical calculations. The pH range of collected water was 5.6e10.4. Extreme alkaline values were observed after strong weather events. For example, the highest pH of 10.4 was recorded after a violent storm and remained elevated for five weeks before returning to a slightly acidic condition (Fig. 2). Outside of these weatherrelated spikes, the pH range was 5.6e6.9. By comparison, the literature for Europe has reported the following pH ranges for runoff water: 6.0e8.2 (Villarreal and Dixon (2005)), 7.6e8.8 (Sazakli et al. (2007)) and 5.8e8.4 (Schriewer et al. (2008)). Half of

the samples collected in this study exceed the drinking water limits for colour (15 mg Pt/L) and turbidity (2 NTU). Ion concentrations were low, with 89% of conductivity values being below 100 mS cm1. This finding indicates that harvested rainwater had a low level of mineralisation. Concentrations in ion comply with the drinking water guidelines available, except for ammonia, which was often detected at unacceptably high levels. The microbiological composition of the tank water varied over the course of the year. Total flora is a measure of the total bacterial load. At 22
C, bacterial counts ranged from 10 to 6.32  105 organisms per mL. Almost all samples were contaminated with coliform bacteria (i.e., they exceeded zero organisms per 100 mL of water). Two faecal indicators were also monitored and showed varying degrees of

Observations (axes F1 et F2 : 48,87 %) 6

Su 4

Su Su

F2 (20,40 %)

2

Su Su S SSASuSu A A A AS S S A SS W W A S A A A W W A

0

-2

A

Su Su Su Su Su S

A

S W

W

S

W W

W

W W

A

W

-4

-6 -8

-6

-4

-2

0

2

4

6

8

10

F1 (28,47 %)

Fig. 6 e A two-dimensional plot of the 50 observations in F1 and F2. - Su [ Summer; A [ Autumn, W[Winter, S[Spring Observations corresponding to the storm event of January 2009 and summer collection months can be differentiated from the central scatter plot.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 6 5 e3 7 7 5

3771

Fig. 7 e A three-dimensional plot of the 50 observations in F1, F2 and F3. - Su [ Summer; A [ Autumn, W[Winter, S[Spring - Observations corresponding to the storm event of January 2009 and summer collection months can be differentiated from the central scatter plot.

contamination. Roof-collected rainwater often showed high levels of contamination with enterococci, as can be seen from the fact that the maximum value exceeded 10 000 CFU per 100 mL of water. The majority of samples (79%) tested positive for E. coli, an indicator of faecal contamination. E. coli and enterococci were simultaneously present in samples, always with enterococci having the higher concentrations (Fig. 3). Although these bacteria are unable to reproduce in water, enterococci has a better survival ability in water than E. coli. Three widely used bacterial indicators, total coliform, E. coli and enterococci, were detected in the majority of samples. In concordance with previous studies (Simmons et al. (2001; Albrechtsen (2002); Blangis and Legube (2007); Nolde (2007); Sazakli et al. (2007)), our results show that roof rainwater runoff is not suitable for human consumption due to the high levels of microbiological contamination within it. Whereas roof-collected rainwater, in general, meets the classical parameters for drinking water in terms of physicochemistry, the bacterial contamination resent in the collected samples was above acceptable limits. As a result, it must be recommended to use a system equipped of disinfection. In addition, it must be highlighted that no first-flush diversion was used in this study. Now such a system could permit to decrease concentrations of some of the tested water quality parameters (Mendez et al. (2011)).

3.2.

Multivariate methods

A starting data matrix, with columns representing the different samplings (observations) and rows corresponding to the

measured parameters (variables), was constructed. The variability of total coliform load was excluded because at least thirteen results were unusable due to the presence of an interfering flora in the sample. It must be highlighted descriptive statistics of the dataset were already presented in Table 2 but multivariate analysis were performed using primary dataset. A total of 50 complete observations (no missing values across the 22 variables) were thus selected for further analysis with Principal Component Analysis (PCA) and Cluster Analysis (CA). These 50 observations were partitioned into the following groups: 8 observations in winter 2009, 12 observations in spring 2010, 12 observations in summer 2010, 14 observations in autumn 2010 and 4 observations in winter 2010. In PCA, the number of components is equal to the number of variables. A component, however, is comprised not only of a single variable but all of the variables used in the study. The PCA analysis showed that of the 22 components, the first component (F1) accounted for about 28.5% of the total variance, the second component (F2) accounted for about 20.4% of the total variance and the third component (F3) accounted for about 11.1% of the total variance of the dataset. Table 3 gives the loadings for the three first components and square cosines are presented in Fig. 4. A variable is increasingly well represented by a component as the corresponding value of the square cosine approaches the unit. Almost all variables are well represented by the first three components, F1, F2 or F3. Only total flora at 22
C, and AT could have been better represented by a different component. Our discussion, therefore, will focus principally on the three principal components that, collectively, explain 60.0% of the total

3772

Table 4 e Pearson’s correlations between different variables from March 2009 to January 2010. Values in italic are significantly different from 0 (alpha[0.05) Variables

pH

1 0.093 0.361 0.439 0.680 0.121 0215 0.352 0.087 0.209 0.626 0.226 0.095 0.216 0.093 0.642 0.096 0.119 0.153 0.300 0.203

1 0.087 0.373 0.474 0.650 0.278 0.422 0.645 0.440 0.257 0.593 0.660 0.574 0.624 0.316 0.208 0.180 0.020 0.081 0.148

1 0.269 0.353 0.342 0.343 0.059 0.144 0.205 0.432 0.339 0.291 0.182 0.339 0.315 0.267 0.270 0.003 0.076 0.148

1 0.797 0.424 0.217 0.073 0.176 0.133 0.423 0.608 0.292 0.054 0.280 0.684 0.102 0.087 0.217 0.358 0.123

1 0.591 0283 0.088 0.309 0.097 0.713 0.694 0.395 0.068 0.426 0.821 0.200 0.170 0.264 0.430 0.137

1 0.503 0.467 0.276 0.304 0.314 0.752 0.795 0.386 0.664 0.380 0.144 0.131 0.016 0.130 0.088

1 0.145 0.049 0.015 0.214 0.418 0.434 0.092 0.333 0.249 0.095 0.152 0.046 0.169 0.234

Cl

SO42- NO3 PO43- Mg2þ Ca2þ

Naþ

Kþ

NH4þ E. Coli Entero. TF 22
C TF 36
C Pluvio.

1 0.273 1 0.359 0.291 1 0.018 0.198 0.264 1 0.387 0.316 0.335 0.578 1 0.611 0.300 0.508 0.266 0.734 1 0.682 0.697 0.213 0.025 0.286 0.487 1 0.577 0.265 0.623 0.387 0.623 0.883 0.389 1 0.022 0.115 0.145 0.651 0.449 0.244 0.064 0.394 1 0.100 0.252 0.257 0.257 0.314 0.182 0.093 0.188 0.177 1 0.082 0.240 0.208 0.229 0.290 0.175 0.100 0.157 0.152 0.989 0.089 0.017 0.035 0.215 0.063 0.029 0.022 0.175 0.331 0.314 0.082 0.071 0.021 0.314 0.213 0.112 0.023 0.225 0.473 0.007 0.057 0.214 0.001 0.108 0.241 0.142 0.122 0.025 0.041 0.798

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 6 5 e3 7 7 5

pH 1 Temperature 0.035 Conductivity 0.451 Coulour 0.216 Turbidity 0.490 TOC 0.410 Hardness 0.675 CAT 0.307 0.452 Cl 0.376 SO420.156 NO3 0.143 PO430.578 Mg2þ 0.687 Ca2þ 0.490 Naþ 0.546 Kþ 0.211 NH4þ Escherichia coli 0.111 Enterococci 0.101 Total flora at 22
C 0.132 Total flora at 36
C 0.182 Pluviometry 0.049

Temp. Cond. Colour Turb. TOC Hard. CAT

1 0.340 0.007 0.841

1 0.789 0.381

1 0.084

1

3773

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 6 5 e3 7 7 5

1 000

25 Pluviometry Escherichia Coli Enterococci

800

20

700 15

600 500 400

10

Pluviometry (mm)

Concentrations (CFU per 100mL)

900

300 200

5

100 0 03/03/09 10/03/09 24/03/09 07/04/09 14/04/09 21/04/09 28/04/09 05/05/09 19/05/09 26/05/09 02/06/09 09/06/09 16/06/09 23/06/09 30/06/09 15/07/09 21/07/09 28/07/09 04/08/09 11/08/09 18/08/09 25/08/09 01/09/09 08/09/09 15/09/09 22/09/09 29/09/09 06/10/09 13/10/09 20/10/09 27/10/09 03/11/09 10/11/09 17/11/09 24/11/09 01/12/09 08/12/09 15/12/09 21/12/09 05/01/10 13/01/10 19/01/10 26/01/10

0

Fig. 8 e Daily pluviometry and concentrations of E. coli and enterococci observed during the period from March 2009 through January 2010 e The two faecal indicators were significantly correlated with pluviometry.

variance of the dataset. This reduced the dimensionality of the total data from 22 to 3 (an 86.3% reduction) and resulted in a 40.0% loss of information contained in the dimensions. The variables that primarily contributed to the first eigenvector were pH, conductivity, hardness, calcium and potassium. Thus, the first principal component can be interpreted as an ionic component. The second eigenvector was mainly related to organic load, with the most significant variables being temperature, total organic carbon, and ammonium and phosphate values. The third eigenvector represented faecal contamination by enterococci and E. coli (Fig. 4). A variable is increasingly well represented by a component as the corresponding square cosine nears the unit. Graphically, this is represented as the variable nearing the edge of the circle. To confirm the associations between the variables in the total dataset, CA was performed on the measured chemical

variables. The search for natural groupings among variables was a complementary way to study the latent structure of the data and permitted the comparison of CA results to those provided by the PCA. When CA was applied, the dendrogram (Fig. 5) showed three different clusters identified as A, B and C. Cluster A is the ionic component previously described as the first eigenvector in PCA. The level of dissimilarity detected between cluster B and C justified that their association appeared separately in the second and third eigenvectors in PCA. There was adequate agreement between results obtained by unsupervised PCA and CA to confirm the conclusions made over the complete dataset. To elucidate the influence of collection date on stored rainwater composition, different observations were represented in the planes F1 through F2 (Fig. 6). Sample points corresponding to the January 2009 storm event were

180 160 140

Dissimilarity

120 100 80

B

C

A

60 40 20

CAT

Colour

Mg2+

Hardness

+

K

2+

pH

Ca

Na+

SO42-

Conductivity

Cl-

-

NO3

PO43-

Temperature

COT

NH4+

TF 36°C

Turbidity

TF 22°C

Enterococci

E.Coli

Pluviometry

0

Fig. 9 e The dendrogram obtained applying the Ward’s method.- Daily pluviometry is included, and AT is excluded.

3774

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 6 5 e3 7 7 5

differentiated from the central scatter plot (27/01/09, 10/02/09, 17/02/09, 24/02/09). These points are high on F1, the ionic component. In fact, the January 2009 storm strongly affected the collected water parameters, particularly in terms of pH. Samplings that occurred in summer showed high F2 values, corresponding to the bacterial load. Summer months have higher temperatures and less runoff water to refill the tank, thus, explaining the higher levels of biological contamination. Selection of three eigenvectors also permitted direct data evaluation via a three-dimensional plot (Fig. 7). This representation illustrates the variability observed throughout the one year period that the roof runoff was collected. On planes F1 through F3, E. coli and enterococci were well represented and appeared to be positively correlated (Fig. 4). In order to investigate this correlation and to elucidate the origin of these two faecal indicators, an additional ACP was realised using daily pluviometry as extra parameter which was only available beginning in March 2009. Sample points corresponding to the January 2009 storm event were thus excluded. The AT was also removed from analysis because it was generally below the limit of quantification. Thus, the new autoscaled matrix was comprised of 22 parameters and 43 observations. The presence of E. coli and enterococci were significantly correlated with each other (Table 4). The Pearson’s coefficient was equal to 0.989, with a corresponding regression coefficient (R2) of 0.979. In addition, the results show these two faecal indicators were significantly correlated with pluviometry. The linear Pearson’s coefficient (linear regression coefficient) was 0.798 (R2 ¼ 0.637) for E. coli and 0.841(R2 ¼ 0.707) for enterococci. It should be noted that these correlations were strongly influenced by a weather event on March 11th, 2009, during which 39 mm of rain fell in one day. At this time, the bacterial loads were 33 000 CFU per 100 mL for E. coli and 30 000 CFU per 100 mL for enterococci (Fig. 8). As illustrated in the dendrogram, these results were confirmed in the cluster analysis (Fig. 9).

4.

Conclusions

The present work presents results concerning the quality of stored roof runoff. For each sample, we measured pH, conductivity, colour, turbidity, total organic carbon, anions, cations, total hardness, AT, CAT, E. coli, enterococci, and total flora at both 22
C and 36
C. Univariate descriptive statistics for the observed variables were conducted for each sampling event. Several conclusions may be drawn from this study but it must be reminded it is based on a limited data set: the performance of a rainwater collection system was monitored weekly over a period of one year. Although harvested rainwater was found to be of good physicochemical quality, it did not meet drinking water standards. These findings are congruent with a number of other studies indicating that roofcollected rainwater makes poor quality drinking water due to high levels of bacterial contamination. An effort was made to extract more information from the datasets through the use of the multivariate analysis techniques. PCA and CA revealed some specific features of the data structure. Three principal components were identified

which, collectively, accounted for 60.0% of the total variance; the first component was identified as the ionic component, the second component was linked with organic load and the third component represented faecal contamination. PCA results were confirmed with CA. Three clusters of variables were detected, corresponding to the three previously identified components. It is necessary to monitor at least one parameter of each of these three groups to correctly characterise roof-collected water. The great variability of roof runoff quality over the course of the study was illustrated with a three-dimensional plot. We were able to distinguish samples that were influenced by a storm event primarily through the first component, whereas samplings obtained in summer months were discernable due to high levels of microbial contamination. It is known rainfall is characterised by its temporal and spatial distribution and its unpredictability. Most importantly, we found that the quality of roof runoff was also unpredictable over a year at the same location. This great variability is coherent with recommendation of a system equipped with a disinfection. At the same time, E. coli and enterococci were simultaneously present in collected water samples, and their presence was positively correlated with the daily pluviometry. These data suggest that runoff from the roof seeded the tank with faecal contaminants and that bacteriological quality degrades during storage. As a result a first-flush diversion could improve the quality of harvested rainwater.

Acknowledgement The authors would like to thank Sotralentz Habitat for their provision of rainwater-harvesting equipment.

references

Al-Khashman, O.A., 2009. Chemical characteristics of rainwater collected at a western site of Jordan. Atmospheric Research 91 (1), 53e61. Albrechtsen, H.J., 2002. Microbiological investigations of rainwater and graywater collected for toilet flushing. Water Science and Technology 46 (6e7), 311e316. Astel, A., Mazerski, J., Polkowska, Z., Namiesnik, J., 2004. Application of PCA and time series analysis in studies of precipitation in Tricity (Poland). Advances in Environmental Research 8 (3e4), 337e349. Baez, A., Belmont, R., Garcia, R., Padilla, H., Torres, M.C., 2007. Chemical composition of rainwater collected at a southwest site of Mexico City, Mexico. Atmospheric Research 86, 61e75. Blangis, D., Legube, B., 2007. Treatment of rain water by solar photocatalysis. European Journal of Water Quality 38 (2), 121e130. Chang, M., Mcbroom Matthew, W., Scott Beasley, R., 2004. Roofing as a source of nonpoint water pollution. Journal of Environmental Management 73 (4), 307e315. Crabtree, K.D., Ruskin, R.H., Shaw, S.B., Rose, J.B., 1996. The detection of cryptosporidium oocysts and Giardia cysts in cistern water in the U.S. Virgin Islands. Water Research 30 (1), 208e216. C.S.H.P.F, 2006. Position relating to sanitary stakes related to rainwater use for domestic uses, 05/09/06 (Conseil supe´rieur

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 6 5 e3 7 7 5

d’Hygie`ne Publique de France, position relative aux enjeux sanitaires lie´s a` l’utilisation des eaux de pluie pour des usages domestiques) Decree of January 11, 2007. Relating to the limits and rough references of water quality and the water intended for human consumption, JO n
31, 06/02/07, (Arreˆte´ du 11 janvier 2007 relatif aux limites et re´fe´rences de qualite´ des eaux brutes et des eaux destine´es a` la consommation humaine). Decree of August 21, 2008. Relating to rainwater harvesting and their use at the interior and the outside of buildings, JO n
201, 29/08/08, (Arreˆte´ du 21 aouˆt 2008 relatif a` la re´cupe´ration des eaux de pluie et a` leur usage a` l’inte´rieur et a` l’exte´rieur des baˆtiments). Despins, C., Farahbakhsh, K., Leidl, C., 2009. Assessment of rainwater quality from rainwater harvesting systems in Ontario, Canada. Journal of Water Supply Research and Technology-Aqua 58 (2), 117e134. Evans, C.A., Coombes, P.J., Dunstan, R.H., 2006. Wind, rain and bacteria: the effect of weather on the microbial composition of roof-harvested rainwater. Water Research 40 (1), 37e44. Evans, C.A., Coombes, P.J., Dunstan, R.H., Harrison, T., 2007. Identifying the major influences on the microbial composition of roof harvested rainwater and the implications for water quality. Water Science and Technology 55 (4), 245e253. Evans, C.A., Coombes, P.J., Dunstan, R.H., Harrison, T., 2009. Extensive bacterial diversity indicates the potential operation of a dynamic micro-ecology within domestic rainwater storage systems. Science of the Total Environment 407 (19), 5206e5215. Fewtrell, L., Kay, D., 2007. Microbial quality of rainwater supplies in developed countries: a review. Urban Water Journal 4 (4), 253e260. Forster, J., 1998. The influence of location and season on the concentrations of macroions and organic trace pollutants in roof runoff. Water science and technology 38 (10), 83e90. Fo¨rster, J., 1999. Variability of roof runoff quality. Water science and technology 39 (5), 137e144. Handia, L., Tembo, J.M., Carolinemwiindwa, 2003. Potential of rainwater harvesting in urban Zambia. Physics and Chemistry of the Earth 28, 893e896. Horak, H.M., Chynoweth, J.S., Myers, W.P., Davis, J., Fendorf, S., Boehm, A.B., 2010. Microbial and metal water quality in rain catchments compared with traditional drinking water sources in the East Sepik Province, Papua New Guinea. Journal of Water and Health 8 (1), 126e138. Hu, G.P., Balasubramanian, R., Wu, C.D., 2003. Chemical characterization of rainwater in Singapore. Chemosphere 51 (8), 747e755. Huston, R., Chan, Y.C., Gardner, T., Shaw, G., Chapman, H., 2009. Characterisation of atmospheric deposition as a source of contaminants in urban rainwater tanks. Water Research 43 (6), 1630e1640. Khare, O., Goel, A., Patel, D., Behari, J., 2004. Chemical characterization of rainwater at a developing urban habitat of Northern India. Atmospheric Research 69, 135e145. Kulshrestha, U.C., Kulshrestha, M.J., Sekar, R., Sastry, G.S.R., Vairamani, M., 2003. Chemical characteristics of rainwater at an urban site of south-central India. Atmospheric Environment 37, 3019e3026. Lee, J.Y., Yang, J.S., Han, M., Choi, J., 2010. Comparison of the microbiological and chemical characterization of harvested rainwater and reservoir water as alternative water resources. Science of the Total Environment 408 (4), 896e905. Lye, D.J., 2009. Rooftop runoff as a source of contamination: a review. Science of the Total Environment 407 (21), 5429e5434. Mantovan, P., Pastore, A., Szyrkowicz, L., Zilio-Grandi, F., 1995. Characterization of rainwater quality from the Venice region

3775

network using multiway data analysis. The Science of the Total Environment 164, 27e43. May, S., Prado, R.T.A., 2006. Experimental evaluation of rainwater quality for non-potable applications in the city of Sao Paulo, Brazil. Urban Water Journal 3 (3), 145e151. Melidis, P., Akratos, C.S., Tsihrintzis, V.A., Trikilidou, E., 2007. Characterization of rain and roof drainage water quality in Xanthi, Greece. Environmental Monitoring and Assessment 127 (1e3), 15e27. Mendez, C.B., Klenzendorf, J.B., Afshar, B.R., Simmons, M.T., Barrett, M.E., Kinney, K.A., Kirisits, M.J., 2011. The effect of roofing material on the quality of harvested rainwater. Water Research 45 (5), 2049e2059. Nolde, E., 2007. Possibilities of rainwater utilisation in densely populated areas including precipitation runoffs from traffic surfaces. Desalination 215 (1e3), 1e11. Ouyang, Y., 2005. Evaluation of river water quality monitoring stations by principal component analysis. Water Research 39 (12), 2621e2635. Oesterholt, F., Martijnse, G., Medema, G., Van Der Kooij, D., 2007. Health risk assessment of non-potable domestic water supplies in the Netherlands. Journal of Water Supply: Research and Technology e AQUA 56 (3), 171e179. Pinfold, J.V., Horan, N.J., Wirojanagud, W., Mara, D., 1993. The bacteriological quality of rainjar water in rural northeast Thailand. Water Research 27 (2), 297e302. Polkowska, Z., Gorecki, T., Namiesnik, J., 2002. Quality of roof runoff waters from an urban region (Gdansk, Poland). Chemosphere 49 (10), 1275e1283. Sazakli, E., Alexopoulos, A., Leotsinidis, M., 2007. Rainwater harvesting, quality assessment and utilization in Kefalonia Island, Greece. Water Research 41 (9), 2039e2047. Schriewer, A., Horn, H., Helmreich, B., 2008. Time focused measurements of roof runoff quality. Corrosion Science 50 (2), 384e391. Simeonov, V., Kalina, M., Tsakovski, S., Puxboum, H., 2003. Multivariate statistical study of simultaneously monitored cloud water, aerosol and rainwater data from different elevation levels in an alpine valley ( Achenkirch, Tyrol, Austria). Talanta 61, 519e528. Simmons, G., Hope, V., Lewis, G., Whitmore, J., Gao, W., 2001. Contamination of potable roof-collected rainwater in Auckland, New Zealand. Water Research 35 (6), 1518e1524. Tsakovski, S., Tobiszewski, M., Simeonov, V., Polkowska, Z., Namiesnik, J., 2010. Chemical composition of water from roofs in Gdansk, Poland. Environmental Pollution 158 (1), 84e91. Uba, B.N., Aghogho, O., 2000. Rainwater quality from different roof catchments in the Port Harcourt district, rivers state, Nigeria. Journal of Water Supply: Research and Technology - AQUA 49 (5), 281e288. Vazquez, A., Costoya, M., Pena, R.M., Garcia, S., Herrero, C., 2003. A rainwater quality monitoring network: a preliminary study of the composition of rainwater in Galicia (NW Spain). Chemosphere 51 (5), 375e386. Villarreal, E.L., Dixon, A., 2005. Analysis of a rainwater collection system for domestic water supply in Ringdansen, Norrko¨ping, Sweden. Building and Environment 40, 1174e1184. Zhang, P., Dudley, N., Ure, A.M., Littlejohn, D., 1992. Application of principal component analysis to the interpretation of rainwater compositional data. Analytica Chimica Acta 258, 1e10. Zhu, K., Zhang, L., Hart, W., Liu, M., Chen, H., 2004. Quality issues harvested rainwater in arid and semi-arid Loess Plateau of nothern China. Journal of Arid Environments 57, 487e505. Zunckel, M., Saizar, C., Zarauz, J., 2003. Rainwater composition in northeast Uruguay. Atmospheric Environment 37, 1601e1611.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 7 6 e3 7 8 6

Available at www.sciencedirect.com

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

Biologically induced phosphorus precipitation in aerobic granular sludge process Man˜as Angela a,b,c,d, Biscans Be´atrice a,b,c,d, Spe´randio Mathieu a,b,c,* a

Universite´ de Toulouse, INSA, UPS, INP, LISBP, 135 Avenue de Rangueil, F-31077 Toulouse, France INRA, UMR792 Inge´nierie des Syste`mes Biologiques et des Proce´de´s, F-31400 Toulouse, France c CNRS, UMR5504, F-31400 Toulouse, France d CNRS, UMR5503, 4, alle´e Emile Monso BP 84234-F-31432 Toulouse Cedex 4, France b

article info

abstract

Article history:

Aerobic granular sludge is a promising process for nutrient removal in wastewater treat-

Received 1 February 2011

ment. In this work, for the first time, biologically induced precipitation of phosphorus as

Received in revised form

hydroxyl-apatite (Ca5(PO4)3(OH)) in the core of granules is demonstrated by direct spectral

30 March 2011

and optical analysis: Raman spectroscopy, Energy dispersive X-ray (EDX) coupled with

Accepted 17 April 2011

Scanning Electron Microscopy (SEM), and X-ray diffraction analysis are performed simul-

Available online 22 April 2011

taneously on aerobic granules cultivated in a batch airlift reactor for 500 days. Results reveal the presence of mineral clusters in the core of granules, concentrating all the

Keywords:

calcium and considerable amounts of phosphorus. Hydroxyapatite appears as the major

Aerobic granulation

mineral, whereas other minor minerals could be transiently produced but not appreciably

Phosphorus recovery

accumulated. Biologically induced precipitation was responsible for 45% of the overall P

Precipitation

removal in the operating conditions tested, with pH varying from 7.8 to 8.8. Major factors

Hydroxyapatite

influencing this phenomenon (pH, anaerobic phosphate release, nitrification denitrification) need to be investigated as it is an interesting way to immobilize phosphorus in a stable and valuable product. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Phosphorous is a key nutriment for the development of life, constituting one of the major nutrients necessary for agricultural activity. However, the quantities of mineral phosphorus resources (phosphate rock) are decreasing in the world, making phosphorus recovery necessary in the coming century. On the other hand, the high phosphorus and nitrogen content of wastewaters leads to serious problems of eutrophication in ponds, rivers and seas. Therefore, research is now focusing increasingly on combined processes that remove phosphorous from wastewaters and simultaneously recover it in the form of

a valuable product, for example, struvite or hydroxyapatite (De-Bashan and Bashan, 2004; Shu et al., 2006; Suzuki et al., 2006). Phosphorous recovery techniques are particularly suited to high strength wastewaters produced by anaerobic sludge digestion (Demirel et al., 2005; Lemaire, 2007). Calcium or magnesium phosphates can be formed by crystallization and recovered in specific reactors via pH control and chemical dosing (Seckler et al., 1996; Katsuura et al., 1998; Mu¨nch and Barrm, 2001; Giesen, 1999; Baur et al., 2008). The spontaneous phenomenon has been reported to cause economic damage related to pipe clogging when it is not controlled (van Rensburg et al., 2003). In activated sludge systems, biologically induced

* Corresponding author. Universite´ de Toulouse, INSA, UPS, INP, LISBP, 135 Avenue de Rangueil, F-31077 Toulouse, France. Tel.: þ33 0 5 61 55 97 55; fax: þ33 0 5 61 55 97 60. E-mail address: [email protected] (S. Mathieu). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.031

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 7 6 e3 7 8 6

Nomenclature DO PAO VFA EBPR SI STR HAP DCPD ACP HDP U

dissolved oxygen mg/L polyphosphate accumulating organisms volatile fatty acids enhanced biological phosphate removal supersaturation Index Log U struvite MgNH4PO4$6H2O hydroxyapatite Ca5(PO4)3(OH) brushite CaHPO4$2H2O amorphous calcium phosphate Ca3(PO4)2 hydroxy dicalcium phosphate Ca2HPO4(OH)2 supersaturation ratio

phosphate precipitation has also been reported but less investigated (Maurer et al., 1999; Pambrun, 2005; De Kreuk et al., 2005). Calcium phosphate precipitation is thought to contribute to P removal in Enhanced Biological Phosphorous Removal processes (EBPR) and it is considered to enhance biological P removal efficiency (Maurer et al., 1999). Local precipitation is naturally induced when the pH and ion concentrations lead to mineral supersaturation. In the case of calcium or magnesium phosphate, their formation can be caused by phosphate release due to Polyphosphate Accumulating Organisms (PAO) during the anaerobic phase, but also clearly depends on pH. Bioreactions (e.g. nitrification and denitrification) or aeration (CO2 stripping) lead to pH gradients which can be responsible for mineral precipitation in biological sludge (Pambrun, 2005; Bogaert et al., 1997; Saidou et al., 2009; Zhu et al., 2007). These processes still need to be clarified in granular sludge systems. The aerobic granular sludge process is a promising technology for wastewater treatment because of its small footprint and capacity to treat high organic loading rates and its simultaneous nutrient removal through nitrification, denitrification and BioP accumulating processes (Morgenroth et al., 1997; Etterer and Wilderer, 2001; De Kreuk et al., 2005; Lemaire, 2007). The dense-spherical structure of granules leads to transfer limitations (Liu and Tay, 2004; Adav et al., 2008), promoting not only DO gradients but also local pH gradients coming from biological reactions, especially in the case of enhanced denitrification (Wan and Sperandio, 2009; Wan et al., 2009). As phosphate accumulating bacteria can also be present inside the granules (Lemaire, 2007), anaerobic phosphate release can encourage P precipitation within the core of the microorganisms, where subsequent solubilization of the crystals would be more difficult than in the bulk. Phosphate precipitation in a granular sludge process was assumed (but not directly demonstrated) by Yilmaz et al. (2007), De Kreuk et al. (2005) and De Kreuk and van Loosdrech (2007). By estimating its supersaturation index, Yilmaz et al. (2007) suggested that struvite could be transiently formed during the anaerobic phase of the SBR cycle. The contribution of this process to overall P removal was estimated to be less than 10 percent on the basis of a perchloric acid extraction method (Haas et al., 1988; Daumer et al., 2008). Similarly, experimental results by De Kreuk et al. (2005) suggest that P-removal occurs partly by biologically induced

IAP K0 Kp TN MLSS MLVSS CAL MAG WHT OCP ¨ U PCA

3777

ionic activity product conditional solubility product thermodynamic equilibrium of precipitation constant total nitrogen mgN-/L mixed liquor suspended solids g/L mixed liquor volatile suspended solids g/L calcite CaCO3 magnesite MgCO3 whitlockite Ca18Mg2H2(PO4)14 octacalcium phosphate Ca8(HPO4)2(PO4)4 raman shift (cm1) cold perchloric acid

precipitation in granular sludge. Extraction techniques indicated that 2.6% of the sludge mass was due to precipitates (P/VSS), but the whole contribution of this process compared to biological P removal processes was not quantified. For simplicity, precipitation was not included when modelling the process but De Kreuk and van Loosdrech (2007) proposed to increase the maximum fraction of polyphosphate in PAO from 0.35 (Hu et al., 2002) to 0.65 assuming that about 46% of the P removal could be due to P precipitation and 54% due to polyphosphate accumulating bacteria. Recently, Maurer et al. (1999) have proposed a model for naturally induced P precipitation in activated sludge, which is based on the assumption that hydroxyapatite (HAP) and hydroxydicalcium phosphate (HDP) are formed. The model can predict calcium and phosphate concentrations at different pH. However, in all these studies, phosphate minerals formed in biological granules or flocs have never been directly characterized, and the nature of the phosphate precipitate is not demonstrated but only indirectly deduced from stoichiometry of soluble species. The characterization of precipitates inside aerobic granules is still a relatively unexplored field. Minerals involved in phosphorus immobilization have been poorly qualified in biological sludge because traditional techniques (like X-ray diffraction) are difficult to apply directly in such organic matrices (Cloete and Oosthuizen, 2001). SEM-EDX analysis has recently been applied to determine calcite formations in granules and in nacre shells (Ren et al., 2008). However, calcium or magnesium phosphates have not been quantified in previous studies of aerobic granules (Wang et al., 2006; Ren et al., 2008). Therefore, the aim of this study is to reveal the nature of P minerals which can accumulate in EBPR granular sludge systems. In an attempt to determine the chemical composition of precipitates in granules, RAMAN spectroscopy, EDX (Energy Dispersive X-ray) technique coupled with Scanning Electron Microscopy (SEM), and X-ray Diffraction analysis (XRD) are evaluated.

2.

Materials and methods

2.1.

Reactor operating conditions

Aerobic granules were cultivated in a Sequencing Airlift Batch Reactor (SBAR), with a working volume of 17 L, consisting of an

3778

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 7 6 e3 7 8 6

airlift column (D ¼ 15 cm, H/D ratio ¼ 7) with a baffle plate (length/width ¼ 83/15-cm-). An aerating diffuser providing fine bubbles 3 mm in diameter was inserted at the bottom of the reactor at one side of the baffle plate, achieving mixing during both anoxic and aerobic phases (using nitrogen gas for the anoxic phase and air for the aerobic one). Oxygen concentration and pH were measured and recorded online with selective probes (WTW TriOxmatic 701). Temperature was maintained constant at 20  C thanks to a water jacket. Details of the system schematization can be seen in Wan, 2009. Process batch cycles of 4 h length were established as follows: anoxic phase (20 min), aerobic reaction (145 min), idle (30 min), withdraw (30 min) and feed (15 min). Hydraulic Retention Time (HRT) was fixed at 8.5 h, with a volumetric exchange ratio of 50%. The column was fed at the bottom with a synthetic substrate (details in Wan et al., 2009) having the following composition: COD of 1000 mg/L (25% contribution each of glucose, acetate, propionic acid and ethanol); [PO43] ¼ 30 mgP/L, [Ca2þ] ¼ 46 mg/L, [HCO3] ¼ 100 mg/L, [MgSO4$7H2O] ¼ 12 mg/L, [NH4þ] ¼ 50 mgN/L, [NO3] ¼ 100 mgN/L. Therefore, a COD/NNH4þ ratio of 20 was maintained, and nitrate was dosed in order to maintain an anoxic phase after feeding. Influent loading rates coming into the reactor were as follows: 0.14 gN L1 d1 for ammonium, 0.08 gP L1 d1 for ortho-phosphate and 2.82 gCOD L1 d1 for organic substrate.

2.2. Analytical characterization of the liquid and solid phases Chemical analyses were conducted according to standard methods (AFNOR, 1994). COD (NFT 90-101), MLSS (NFT 90-105) and MLVSS (NFT 90-106). NO2, NO3, PO43, NH4þ, Ca2þ, Kþ, Mg2þ concentrations were analyzed by Ion Chromatography (NFT 90-023) after being filtered with 0.2 mm pore-size acetate filters. Microscopic observations over the whole sludge sample were performed with a Biomed-Litz binocular photonic microscope. Particle size distribution was measured with a Malvern 2000 Mastersizer analyser. Granules were sampled at the end of the aerobic phase. Those analyzed by EDX or Raman Spectroscopy had been previously cut into thin slices of 100 mm using a cryo-microtome (Leica CM 30505 Kryostat). Those analysed by XRD had been previously dried and calcined in an oven at 500  C for 2 h, in order to remove the organic fraction. Raman Spectroscopy was performed with an RXN Kaiser Optical Systems INC at a wavelength of 785 nm in the visible range. Two different optical fibres were used for the incident (50 mm) and collected (100 mm) rays. EDX analysis was performed with a photon X analyzer (Quantax Technology Silicon Drift) having a detection limit of 127 eV. It was coupled to a SEM (JEOL 5410 LV) which allowed working in a partial pressure chamber. The reference samples used for comparing the mineral spectra were: struvite (CAS N.13478-16-5), calcite (CAS N. 72608-12-9), magnesite (CAS N. 235-192-7), hydroxyapatite (CAS N. 12167-74-7) and brushite (CAS N. 7789-77-7). XRD analyses were performed with a BRUCKER D5000 diffractometer, with a cobalt tube scattering from 4 to 70 in 2q. Chemical nitrogen and total phosphorus extractions were performed in accordance with standard methods (NFT 90-110

and NFT 90-136 respectively) adapted for the granular samples: first a physical separation was made between flocs and granules by means of a 315 mm shiver, then granules were rinsed with a volume of ultrapure water and the volume of sample extracted was re-established with ultrapure water before analysis. The Supersaturation Index (SI) for each mineral considered (eq. (1)) was calculated as the logarithm of supersaturation ratio:  SI ¼ log U ¼ log

1=j IAP ½K0 

(1)

where IAP is the Ionic Activity Product of the ion concentrations involved in the mineral precipitation; j is the number of ions of the mineral and K0 refers to the conditional solubility product, which includes the thermodynamic mineral precipitation constant at a given temperature, the ionic activity coefficients, and the ionization fractions of each component (Snoeyink and Jenkins, 1980; Burriel et al., 1985). The PHREEQC software (Minteq.v4 database) was used to calculate the chemical equilibrium for each sample collected in the reactor. Ionic Strength was taken into account as well as the ionic activity coefficients by the Davies approach (Parkhurst et al., 1980; Burriel et al., 1985; Mostastruc, 2003). pKp of struvite (STR), hydroxyapatite (HAP), brushite (DCPD), amorphous calcium phosphate (ACP) and hydroxyl dicalcium phosphate (HDP) considered were respectively: 13.26, 57.5, 6.6, 26.52 and 22.6 (Ohlinger et al., 1998). Oversaturating conditions were considered to be achieved when SI > 0.5 (theoretically zero but a security margin is usually given (Burriel et al., 1985; Rahman et al., 2006)).

3.

Results

3.1.

Reactor performance and kinetics assessment

The sequencing batch reactor was operated for 540 days. Mean efficiencies of COD, total Nitrogen and Phosphorous removal, as well as MLSS, MLVSS and SVI30 values are shown in Table 1. An SVI5/SVI30 ratio closer to 1, indicates a major presence of granules in the whole sludge. MLVSS and MLSS of the whole sludge were measured regularly in the reactor with

Table 1 e Mean values measured during 500 days of reactor performance, for Mixed Liquor Suspended Solids; soluble COD, Total Nitrogen and Phosphorous efficiencies; MLVSS/MLSS ratio, Sludge Volume Index after 30 min and SVI5/SVI30 ratio. h TN (%)

hP

(g/L)

h COD (%)

14 13 22 28 34

94 97 93 93 97

92 100 96 96 97

Period of time (days)

MLSS

0e100 100e200 200e300 300e400 400e500

(%)

MLVSS/ MLSS e

(mL/g)

SVI5/ SVI30 e

62 31 56 67 50

78 83 77 67 65

33 34 22 15 15

2.0 1.7 1.5 1.0 1.0

SVI30

3779

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 7 6 e3 7 8 6

time, values of 30e35 and 21e25 g/L being achieved for MLSS and MLVSS respectively at the end of the study. The MLVSS/ MLSS ratio of granules progressively decreased from 80% to 67%, indicating mineral accumulation. Final SVI5 was 15 mL/g. As shown in Fig. 1, the mixed liquor in the reactor was composed of granules and flocs, the latter disappearing progressively with time. Particle size distribution analyses revealed that 800 mm was the most probable diameter for granules. At the end of the 540 days of reactor run, removal efficiencies achieved were 100% for ammonium, 100% for nitrate, 82% for ortho-phosphates and 99% for soluble COD. Kinetic analyses were performed during the batch cycle to assess ammonia, nitrate, COD and phosphate removal rates. Figs. 2 and 3 show typical time-series profiles in the reactor obtained with two different aeration flow rates (160 L/h and 350 L/h respectively). The separation between the anoxic/ aerobic phases is depicted by a dotted vertical line. 50 Ammonium

500

40 P-PO4

35

DCO

3-

400

30 25

300

+

600

50 45

N-NO3-

20 200

15 10

100

N-NH 4 + , N-NO 3 - , P-PO 4 3-(mg/L)

45

N-NH4 , N-NO3 , P-PO4 (mg/L)

b

600

COD (mg/L)

a

Ammonium

500

40

N-NO3-

35

P-PO4

400 DCO

30 25

300

20 200

15 10

COD (mg/L)

Fig. 1 e Granules and flocs in the reactor after 520 days of operation. The bar dimension is 2 mm.

Ammonium was first partially removed during the nonaerated phase and then during the aerobic phase via nitrification. Ammonium consumption during the anoxic phase was due to heterotrophic assimilation but it could also be explained by other, non-biotic processes like adsorption (because of high MLSS) or precipitation (as struvite for example). Nitrate and nitrite concentration remained negligible at all times, confirming that simultaneous nitrification and denitrification occurs in granular sludge. Regarding phosphorus, several mechanisms seemed to take place simultaneously. Kinetics in Figs. 2 and 3, show phosphorus release during the anaerobic phase and P uptake during the aerobic period, Meanwhile, biological staining with sudan black and safranin was carried out according to the method reported in Pandolfi et al. (2007), revealing lipid and PHB accumulations in different granules samples taken during the anoxic phase (results not here shown). Both kinetics and color staining results, suggested the presence and activity of Polyphosphate Accumulating Organisms (PAO). Fig. 2a and b show that phosphate uptake rate was higher for higher aeration rate, because dissolved oxygen was limiting at low aeration rate (DO being maintained at 0.3 mg/L). Final phosphate concentration was thus lower at the high air flow rate (2.5 mgP/L) compared to the low aeration rate (8 mgP/L). Mg2þ and Kþ fluctuations followed those of P (Fig. 3a and b). This was related to polyphosphate synthesis, which general formula is Menþ2PnO3nþ1, where n indicates the chain length, and Me represents a metal cation (Jardin and Po¨pel, 1996). In contrast, Ca2þ concentration showed a very different trend. It decreased rapidly during the non-aerated period following wastewater feeding. This behavior can be explained by a rapid formation of calcium complexes or precipitates, which will be demonstrated in the following section. Total phosphorus was extracted from granular sludge samples collected at the end of the aerobic cycle. It was carried out in triplicate after 520 days of reactor operation (according to the method detailed in section 2.2). Results indicated that P content was 56  7.3 mgP/gSS. This value is not really different from values usually reported for EBPR sludge. In EBPR

100

5

5 0

0 0

20

40

60

80

100

time (minutes)

120

140

160

0

0 0

20

40

60

80

100

120

140

160

time (minutes)

Fig. 2 e Variation of NO3L, PO43L, NH4Dand COD in the reactor bulk during a cycle operation with weak aeration (160 L/h) (2a) and high aeration (350 L/h) (2b) rates.

3780

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 7 6 e3 7 8 6

a

b

9 140

9 140

3 40

O2

6 5

80

4 60 3 40

2 20

1

0

2 20

0 0

20

40

60

80

100

120

140

2D

1

0

160

0 0

time (minutes) D

(mg/L), O2 et pH

(mg/L)

100

2+

+

K , Ca

4 60

7

2+

5

80

120

Calcium Magnessium pH

+

O2

K , Ca

6

pH

2+

(mg/L)

100

2+

Magnessium

(mg/L), O2 et pH

7

Calcium

Mg

120

8

Potassium

Mg

8 Potassium

20

40

60

80

100

120

140

160

time (minutes)

2D

Fig. 3 e Variation of K , Mg , Ca , pH and O2 in the reactor bulk during a cycle operation with weak aeration (160 L/h) (3a) and high aeration (350 L/h) (3b) rates.

systems, this value is related to the polyphosphate content of sludge, which depends on various parameters: fraction of PAO in the sludge, wastewater COD/P ratio and fraction of volatile fatty acids in wastewater. Li and Liu (2005) reported a similar P content at a similar P/COD ratio. However, the following paragraph will show that phosphorus is not only accumulated in polyphosphate form but also as a precipitated mineral compound.

3.2.

Raman analysis

Raman analysis is a non-destructive analytical technique that requires limited sample preparation (Hollas, 1996; Skoog and Nieman, 2003). It was chosen because of its low water background, as well as for providing sharper and clearer bands than IR spectra (Barbillat, 2009). It has already been proved for the characterization and identification of different biological systems since the biologically associated molecules can exhibit a unique spectral pattern (Ivleva et al., 2009). In an attempt to determine the internal structure of the granules, samples were cut into slices of 100 mm width, prepared as described in section (2.2). Then, a central slice was chosen and observed with a binocular microscope before being analyzed by Raman Spectroscopy. As shown in Fig. 4a, microscopic observation of a typical central slice revealed a white crystalline precipitate in the centre of the granule. Spectroscopic analysis was performed at different points of the mineral core (as indicated on Fig. 4a). An initial set of tests (not shown) was also conducted beforehand with different samples: granules taken at different batch cycle times, dehydrated flocs separately, different cuts and thicknesses sliced from the same granule. Finally, some pure minerals used as reference products (Struvite, hydroxyapatite, brushite, calcite and magnesite), were also analyzed with Raman Microscopy and compared to the sample spectra. The following conclusions were drawn: (i) Both flocs and external granule slices showed irregular curved spectra (due to organic matter) with no

remarkable matching peaks (ii) All spectra obtained in the core of granules showed a common and reproducible pattern of 8 peaks of different intensities (see Fig. 4b), considering that a peak is noted when its intensity is three times the mean background noise; (iii) All granule central slices had the same typical peaks regardless of the cycle time. According to Fig. 4b, the most important peaks in the sample were found at the following Raman shifts (cm1): 430, 588, 850, 962, 1072, 1135, 1295, 1448. The spectra of granule core samples were compared with those obtained with reference minerals. After an individual comparison of frequency-intensity coincidence, it was concluded that calcite and magnesite spectra (not shown) did not match the sample at all. In Fig. 5, the most similar mineral spectra have been depicted in order to compare the coincidence of their peaks. Brushite shows two or three peaks not far from those of granule spectra, but most of the major peaks do not coincide (407, 586, 986, 1056 and 1114 cm1). Struvite spectrum indicates five peaks (421, 563, 944, 1053, 1112 cm1) which are very similar to those of the granule sample, but differences in the Raman shifts are statistically significant. The hydroxyapatite spectra show 4 peaks, which all match the granule spectra (427, 589, 962, 1072 cm1) with differences lower than 3 cm1. Globally, among all the different pure mineral patterns compared, the hydroxyapatite spectrum best fitted the sample in intensity and wave number but could not explain all the peaks observed in the granule spectra. These results suggest that hydroxyapatite is a major mineral precipitated in the aerobic granule cores but the presence of other minerals cannot be totally discounted.

3.3.

SEM-EDX analysis

Scanning Electron Microscopy (SEM) coupled with Energy Dispersive X-ray detector (EDX) analyses were carried out on cut mature granules. Typical images are shown in Fig. 6 for two typical central slices of different granules. It was

3781

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 7 6 e3 7 8 6

a

b 20000 19000 18000 17000 In te n s ity

16000 15000 14000 13000 12000

Granule central slice_1

11000

Granule central slice_2 Granule central slice_3

10000 100

300

500

700

900

1100

1300

1500

RamanShift cm-1 Fig. 4 e a: Central slice from a mature granule after 450 days of reactor run. The bar length is 100 mm. b: Spectra of core aerobic granule slices.

found that inorganic precipitates occupied an important fraction of the total volume of the granule, located in different zones, mainly close to the centre. A first scanning map of carbon (Fig. 6c and d) revealed that the central inorganic zones did not contain large amounts of carbon in contrast with peripheral organic biofilm. Similar results were observed for nitrogen and magnesium (not shown). In contrast Ca and P were mainly found together in the central precipitates and comparatively poorly in the organic biofilm (Fig. 6eh). This result again supports the idea that calcium phosphates are formed in the core of the granules. Phosphorus was also detected but with lower concentration in the organic biofilm zone. It probably came from polyphosphates in PAO clusters. Fig. 6i and j, focus on the inorganic precipitate with a higher objective. Fig. 6i reveals porous, ordered holes in the solid mineral phase. This could be related to the mechanism of

precipitate formation around bacterial cells, in relation with gaseous transfers between the microorganisms and the extracellular medium. Another interesting result can be seen in Fig. 8j, where some prismatic structures appear stacked, similar to hydroxyapatite, which crystallizes in the hexagonal system (Morgan et al., 2000;). Furthermore, several localized EDX spectral analyses were made, pointing the probe at different locations of the precipitate. The spectrum obtained was very reproducible in different locations of the central mineral zone (Fig. 6k). Analysis spectra clearly showed that calcium, phosphate and oxygen were the major components observed in the mineral zone whereas magnesium and potassium were definitively absent. Quantitative analysis over 5 different samples showed that the Ca/P mean atomic ratio obtained for the mineral precipitate was 1.63  0.05, which is quite close to the theoretical one for hydroxyapatite (1.67). In parallel, scanning analyses of the flocs and supernatant (not shown) did not reveal any similar calcium phosphates but some sparse mineral particles with high K, Mg and P content were found. These analyses indicated that calcium phosphates were exclusively accumulated in the granules whereas other minerals could be formed in the bulk, e.g. magnesium phosphate, ammonium struvite (MgNH4PO4.6H2O) or potassium struvite (KMgPO4$6H2O). However, they were detected in much smaller amounts than hydroxyapatite.

3.4.

Fig. 5 e Spectra of reference minerals compared to a granule central slice.

XRD analysis

XRD analysis is an efficient tool for distinguishing crystalline minerals from those of amorphous structure, so it was also carried out on some granule samples. Three different preparations of granular samples were tested. Results of two sample analyses (not shown), i.e. a sample of dried pulverised granules and a wet sludge sample, revealed a major peak

3782

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 7 6 e3 7 8 6

coinciding with calcium phosphate patterns. However, high noise due to the organic fraction was present, making any interpretation difficult. Thus, a third granular sludge sample was treated to remove the organic matter (described in 2.2), leading to the diffractogram presented in Fig. 7. A number of distinct rays indicate the presence of crystalline forms. By comparison with reference spectra, most of the peaks, and in particular the bigger ones, coincided with those of the hydroxyapatite spectrum (Ca5(PO4)3(OH)). The remaining minor peaks coincided with whitlockite (Ca18Mg2H2(PO4)14). The large central peaks indicate the possible presence of amorphous mineral species. It should not be forgotten that, for XRD analysis, the sample was heated to 500  C and so some hydroxylation phenomena could have taken place. Considering that hydroxyapatite dehydroxilation does not occur under 800  C (Wang et al., 2004), changes of this mineral in the original sample, due to heating, would not take place. However, in the range of 200e400  C, dehydratation of the lattice and adsorbed water of some other minerals could be possible according to Kohutova et al. (2010). The magnesium and phosphorus initially present in organic polymers (polyphosphate) could precipitate in a new form during heating. Despite the fact that XRD again confirmed the major formation of hydroxyapatite, it is still difficult to know whether other intermediates were present or not and, in the case of whitlockite (WHT), it might have been formed during the heating process.

4.

Discussion

4.1. Hydroxyapatite: a major phosphate mineral in aerobic granules

Fig. 6 e a: SEM image of a granule central slice. b: SEM image of a granule central slice. c: Carbon (red) scanning with EDX. d: Carbon (red) scanning with EDX. e: P (dark blue) scanning with EDX. f: P (dark blue) scanning with EDX. g: Ca (light blue) scanning with EDX. h: Ca (light blue) scanning with EDX. i: granule precipitate SEM images. j: granule precipitate SEM images. k: Punctual analysis with EDX probe of a precipitate

All the results (Raman spectroscopy, SEM-EDX, and XRD) support the same conclusion: hydroxyapatite (HAP) was the major mineral found inside the phosphorus-rich granules in this study. Both Raman and SEM-EDX analysis allowed calcium phosphate mineral to be identified and XRD analysis confirmed its crystalline form, but also suggested the presence of other amorphous minerals. SEM-EDX analysis in Fig. 6e and f, pointed out that P was also present in the organic fraction of the aggregates, probably linked to polyphosphate stored in bacterial biofilm. This last statement may be supported by the fact that Mg and K elements, which are linked to polyphosphate constitution, were also found sparsely in this area. SEM-EDX indicates a very reproducible Ca/P ratio (1.63  0.05) coinciding with that of hydroxyapatite (1.67), and notably different from other calcium phosphates (e.g. amorphous calcium phosphate: 1.50; hydroxydicalcium phosphate: 2; whitlockite: 1.3). XRD analysis finally confirmed the presence of crystalline hydroxyapatite, possibly associated with amorphous forms. Considering that most of the calcium was immobilized with phosphorus (as indicated by SEM-EDX images) with a Ca/P ratio

grown in a biological granule. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

3783

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 7 6 e3 7 8 6

450 granule XRD sample WHITT pattern HAP pattern

400

350

Lin ( counts)

300

250

200

150

100

50

0 5

15

25

35 2-Theta Scale

45

55

65

Fig. 7 e XRD diffractogram of a granule sample compared to HAP and WHT pattern.

of 1.67 (hydroxyapatite), it is possible to estimate the contribution of precipitation to global P removal. As Ca2þ removal yield of 46% was obtained (0.488 mmol/L), it means that about 0.292 mmolP/L was removed via hydroxyapatite precipitation. This represents about 45% of the total P removal in the process (82% of P was removed which represents 0.68 mmol/L), the rest being explained by biological mechanisms. The precipitation contribution is hence much more significant than those estimated in flocculated sludge (Haas et al., 1988). But this is in accordance with the data obtained with granular sludge by De Kreuk et al. (2005) and De Kreuk and van Loosedrech (2007), which suggest that P accumulation in EBPR granules can double the accumulation achieved in flocs, because of precipitation. In the calcium phosphate family, hydroxyapatite (HAP) is commonly considered as the most stable phase and the most insoluble one. According to Ostwald’s ripening theory (Mullen et al., 2001), precursors such as brushite (DCPD), octacalcium phosphate (OCP), and amorphous calcium phosphate (ACP) contribute to its formation, brushite being the most soluble phase. Hydroxyapatite and brushite were both considered in this study as reference samples but brushite was not detected (Raman). In one sense, our results confirm the first assumption of Maurer et al. (1999) who supposed that hydroxyapatite can be accumulated in EBPR systems. However Maurer et al. (1999) also supposed that HDP was formed as an intermediate without any convincing explanation for that choice, and this assumption is difficult to confirm in our case.

Of all the minerals that could be found in wastewater treatment (Musvoto et al., 2000; van Rensburg et al., 2003; Larsdotter et al., 2007), only a few were expected to precipitate in granular sludge, in particular calcium carbonate (Ren et al., 2008; Wang et al., 2006), or struvite (Yilmaz et al., 2007). Calcium carbonate in calcite form, was previously detected in biological aggregates and aerobic granules (Ren et al., 2008; Wang et al., 2006). Due to their competition for calcium (Lin and Singer, 2006) mineral phosphate and carbonate can inhibit each other. According to several authors (Montastruc et al., 2003; Bellier et al., 2006), pH plays an important role, favouring phosphate precipitation at pH 7e8.5, whereas both carbonates and phosphates co-precipitate at pH 9e11. In our case, absence of calcite could be explained simultaneously by low calcium availability due to hydroxyapatite formation and inappropriate pH inside the granules. In contrast with the assumption of Yilmaz et al. (2007), no struvite was detected in the granules and struvite precipitation seems to have played a minor role in phosphate immobilisation in our study. However, samples were taken at the end of the aerobic period, and it is possible that struvite had been transiently formed in the previous anaerobic phase and afterwards solubilized as ammonia was consumed during nitrification.

4.2. Parameters controlling phosphorus precipitation in EBPR granular sludge Supersaturation index calculation (SI) for different minerals, for the supernatant, established for each time of the kinetic

3784

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 7 6 e3 7 8 6

a

b STR_SI

HAP_SI

ACP_SI

HDP_SI

DCPD_SI

2.0

1.5

1.5

1.0

1.0

0.5

0.5

SI

SI

2.0

0.0

0.0

-0.5

-0.5

-1.0

-1.0

STR_SI

HAP_SI

ACP_SI

HDP_SI

DCPD_SI

-1.5

-1.5 0

20 40 60 80 100 120 140 160 time (minutes)

0

20 40 60 80 100 120 140 160 time (minutes)

Fig. 8 e a: Saturation Index for several minerals in the bulk during a cycle with weak aerating conditions. b: Saturation Index for several minerals in the bulk during a cycle with strong aerating conditions.

experiments are shown on Fig. 8. Saturation index was calculated using the Minteq.v4 database (PHREEQC, software) with the pH and concentrations measured in the bulk. Struvite saturation index was negative throughout the batch cycle and lessened progressively with ammonia consumption by nitrification. This indicates that the ammonium and magnesium concentrations were too low to cause struvite precipitation in those conditions. Concerning calcium phosphates, the saturation index for brushite and hydroxydicalcium phosphate were close to zero, i.e. these minerals were not considerably oversaturated. This suggests that the latter were poorly or very briefly formed, only during the feeding period. SI for amorphous calcium phosphate varied from around 1 to less than 0.5. Finally, the highest value of SI (from 1.7 to 1.2) was obtained for hydroxyapatite, i.e. the most stable phase among the calcium phosphates showed oversaturation conditions throughout the experiment. In addition, for all the compounds studied, SI decreased during the aerobic phase, because phosphate concentration lessened (due to P uptake) and pH decreased (due to nitrification). This result confirms that precipitation of calcium phosphate is more probable during the initial anoxic period, which is in accordance with the tendency observed for calcium (Fig. 3). Additionally, hydroxyapatite precipitate was observed in the core of granules. This means that confined conditions were more favorable for hydroxyapatite formation or accumulation than conditions in the bulk. Three explanations can be proposed: (1) higher local phosphate concentration, (2) higher local pH and (3) higher retention time for granules. Firstly higher local phosphate concentration is probably reached during the anaerobic period due to phosphate release by PAO in the internal part of granules. Simultaneously, calcium was also provided during the anaerobic period by means of wastewater feeding. In parallel, observed phosphate release was relatively moderate and would probably have been more significant if precipitation had not occurred. This shows that feeding anaerobically is a method that encourages P precipitation. Secondly, pH is obviously an important parameter controlling phosphate precipitation (high pH increases hydroxyapatite supersaturation). Therefore, another possible

explanation for hydroxyapatite accumulation in the core of granules is the fact that internal pH can be higher than bulk pH, because of denitrification. A last mechanism is the fact that high retention time of granules encourages the formation of the most stable calcium phosphate (HAP) due to low formation rate and low solubilization rate, whereas other calcium phosphates mentioned before can be transiently produced and resolubilized. More generally, the importance of calcium precipitate depends on influent characteristics. In this work, the range of phosphate concentration was similar to those explored by Maurer et al. (1999). It was slightly higher than those found in conventional domestic wastewater but lower than those reported for high strength wastewater like agro-food industry waste (Yilmaz et al., 2007). Calcium, magnesium, and ammonium concentrations were at moderate levels, similar to those found in domestic wastewater. Comparatively to other studies, pH was relatively high in this work, ranging from 7.8 to 8.8 in a typical SBR cycle. This was due either to bicarbonate stripping (pH was higher at high flow rate) and denitrification, and hence, these two processes clearly encourage hydroxyapatite precipitation.

4.3. Advantage of hydroxyapatite accumulation in granular sludge Hydroxyapatite is a phosphorus compound that is much more stable than bacterial polyphosphate. In EBR systems, sludge containing polyphosphates needs to be extracted and removed rapidly from the system in order to avoid problems with secondary P release. This makes it obligatory to restrict the sludge retention time to a reasonable value, and then anaerobic storage is impossible as phosphate would be released in the liquid phase. In contrast, hydroxyapatite accumulation is advantageous because long retention time for granules is possible, as well as storage before agricultural use. In addition, from our experience (data not presented), granules with a mineral HAP core are very stable and can be easily dehydrated. Finally induced precipitation in granules seems to be completely compatible with biological reactions. In

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 7 6 e3 7 8 6

comparison, one of the reported drawbacks of simultaneous phosphorus precipitation in activated sludge process with calcium (lime) is that precipitation occurs at high pH (e.g. 9), which would be out of the optimal pH range for most biological processes (Carlsson et al., 1997; Arvin, 1979). High reactant excess is also necessary to reach very low P concentration at conventional pH. In the Phostrip process, lime addition is thus performed on a side stream anaerobic reactor (Brett et al., 1997). In contrast, due to important gradients within granules, it is possible to maintain conditions in the core (favourable for precipitation) which are different from those in the external zone. In contrast with previous experience with flocculated sludge, it is shown in this study that hydroxyapatite accumulation in granular sludge is perfectly compatible with major biological reactions. Future work will be necessary to find the practical conditions which allow advantage to be taken of this process during the treatment of real wastewater.

5.

Conclusions

For the first time, different analyses (Raman, SEM-EDX, XRD) have revealed the nature of phosphorus precipitates in an EBPR granular sludge process. C

C

C

Raman analysis provided a repetitive pattern over a granule core sample. The four main peaks coincided with those of hydroxyapatite (Ca5(PO4)3(OH)) SEM-EDX demonstrated the presence of mineral clusters in the core of granules. These clusters concentrated most of calcium and phosphorus and EDX revealed that Ca/P ratios (1.63  0.05) were close to the ratio of hydroxyapatite. XRD analysis of the mineral fraction of the sludge confirmed that the major mineral present was a crystalline hydroxyapatite, although it probably coexists with other minor amorphous calcium phosphates.

This work reveals that hydroxyapatite accumulation is an important phenomenon in the EBPR granular sludge process and merits attention in the future. In the conditions tested, it is estimated that about 45% of the P removal was due to biologically induced precipitation. In that sense, future work should focus on the operating conditions which favor hydroxyapatite accumulation as it could become an interesting way of immobilizing and recycling phosphorus.

Acknowledgments The authors would like to thank E. Mengelle, M. Bounouba, D. Delagnes, D. Auban, S. Teychene and S. Julien for their help in this work.

references

Adav, S.S., Lee, D.-J., Show, K.-Y., Tay, J.-H., 2008. Aerobic granular sludge: recent advances. Biotechnology Advances 26 (5), 411e423.

3785

Arvin, E., 1979. The influence of pH and calcium ions upon phosphorus transformations in biological wastewater treatment plants. Progress in Water Technology 1, 19e40. Barbillat, 2009. Spectroscopie Raman. Techniques de l’inge´nieur P2865e1. Baur, R., Ram, Prasad, Ahren, Britton, 2008. Reducing ammonia and phosphorus recycle loads by struvite harvesting. Ostara Nutrient Recovery Processes Brochure for WEFTEC 2008. Bellier, N., Chazarenc, F., Comeau, Y., 2006. Phosphorus removal from wastewater by mineral apatite. Water Research 40 (15), 2965e2972. Bogaert, H., Vanderhasselt, A., Gearney, K., Yuan, Z., Thoeye, C., Verstraete, W., 1997. A new sensor based on pH effect of denitrification process. Journal of Environment Engineering 123, 884e891. Brett, S., Guy, J., Morse, G.K., Lester, J.N., 1997. Phosphorus Removal and Recovery Technologies. Selper Publications, London. Burriel, F., Lucena, F., Arribas, S., Herna´ndez, J., 1985. Quı´mica Analı´tica Cualitativa, eighth ed. Ed. Thompson. Carlsson, H., Aspegren, H., Lee, N., Hilmer, A., 1997. Calcium phosphate precipitation in biological phosphorus removal systems. Water Research 31 (5), 1047e1055. Cloete, T.E., Oosthuizen, D.J., 2001. The role of extracellular polymers in the removal of phosphorous from activated sludge. Water Research 35 (15), 3595e3598. Daumer, M.-L., Be´line, F., Spe´randio, M., Morel, C., 2008. Relevance of a perchloric acid extraction scheme to determine mineral and organic phosphorus in swine slurry. Bioresource Technology 99 (5), 1319e1324. De Kreuk, M.K., van Loosdrech, 2007. Formation of aerobic granules with domestic sewage. Environment Engineering 132 (1), 694e697. De Kreuk, M.K., Heijnen, J.J., Loosdrecht, Van, 2005. Simultaneous COD, nitrogen and phosphate removal by aerobic granular sludge. Biotechnology and Bioengineering 90 (6), 761e769. De-Bashan, L.E., Bashan, Y., 2004. Recent advances in removing phosphorus from wastewater and its future use as fertilizer (1997e2003). Water Research 38 (19), 4222. Demirel, B., Yenigun, O., Onay, T.T., 2005. Anaerobic treatment of dairy wastewaters: a review. Process Biochemistry 40 (8), 2583e2595. Etterer, T., Wilderer, P.A., 2001. Generation and properties of aerobic granular sludge. Water Science and Technology 43 (3), 19e26. Giesen, A., 1999. Crystallization Process enables environment friendly phosphate removal at low cost. Environmental Technology 20, 769e775. Haas, D.W., Lo¨tter, D.W., Dubery, I.A., 1988. An evaluation of the methods used for the determination of ortho-phosphate and total phosphate in activated sludge extracts. Water Science 15 (4), 257e260. Hollas, J.M., (1996). Spectroscopy (3rd ed.). Ed. Dunod. Hu, Z.-r., Wentzel, M.C., Ekama, G.A., 2002. Anoxic growth of phosphate-accumulating organisms (PAOs) in biological nutrient removal activated sludge systems. Water Research 36 (19), 4927e4937. Ivleva, N.P., Wagner, M., Haisch, C., Niessner, R., Horn, H., 2009. Combined use of confocal laser scanning microscopy (CLSM) and Raman microscopy (RM): Investigations on EPS - Matrix. Water Research 43 (1), 63. Jardin, N., Johannes Po¨pel, H., 1996. Influence of the enhanced biological phosphorus removal on the waste activated sludge production. Water Science and Technology 34 (1e2), 17. Katsuura, H., 1998. Phosphate recovery from sewage by granule forming process (full scale struvite recovery from a sewage works at Shimane Prefecture, Japan). In: International Conference on Phosphorus Recovery from Sewage and Animal Waste. Warwick University, UK.

3786

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 7 6 e3 7 8 6

Kohutova, A., Honcova, P., Podzemna, V., Bezdicka, P., Vecernikova, E., Louda, M., Seidel, J., 2010. Thermal analysis of kidney stones and their characterization. Journal of Thermal Analysis and Calorimetry 101 (2), 695e699. Larsdotter, K., La Cour Jansen, J., Dalhammar, G., 2007. Biologically mediated phosphorus precipitation in wastewater treatment with microalgae. Environmental Technology 28 (953), 960. Lemaire R. 2007. Development and fundamental investigations of innovative technologies for biological nutrient removal from abattoir wastewater. Ph.D Thesis. University of Queensland, Australia. Li, Y., Liu, Y., 2005. Diffusion of substrate and oxygen in aerobic granule. Biochemical Engineering Journal 27 (1), 45e52. Lin, Y.P., Singer, P.C., 2006. Inhibition of calcite precipitation by orthophosphate: speciation and thermodynamics considerations. Geochimica et Cosmochimica Acta 70 (2006), 2530e2539. Liu, Y., Tay, J.-H., 2004. State of the art of biogranulation technology for wastewater treatment. Biotechnology Advances 22 (7), 533e563. Mullen, J.W., (2001). Solutions and Solubility. Crystallization (4th ed.). pp. 86e34. Mu¨nch, E.V., Barrm, K., 2001. Controlled struvite crystallisation for removing phosphorus from anaerobic digester sidestreams. Water Research 35 (1), 151e159. Maurer, M., Abramovich, D., Siegrist, H., Gujer, W., 1999. Kinetics of biologically induced phosphorus precipitation in wastewater treatment. Water Research 33 (2), 484e493. Montastruc, L., Azzaro-Pantel, C., Biscans, B., Cabassud, M., Domenech, S., 2003. A thermochemical approach for calcium phosphate precipitation modelling in a pellet reactor. Chemical Engineering Journal 94 (1), 41e50. Morgan, H., Wilson, R.M., Elliott, J.C., Dowker, S.E.P., Anderson, P., 2000. Preparation and characterisation of monoclinic hydroxyapatite and its precipitated carbonate apatite intermediate. Biomaterials 21 (6), 617e627. Morgenroth, E., Sherden, T., van Loosdrecht, M.C.M., Hijnen, J.J., Wilderer, P.A., 1997. Aerobic Granular sludge in a sequencing batch reactor. Water Research 31 (12), 3191e3194. Mostastruc L. 2003. Mode´lisation et optimisation d’un re´acteur en lit fluidise´ de de´phosphatation d’effluents aqueux. Ph.D. Thesis at INPT and UPS Toulouse III. Musvoto, E.V., Wentzel, M.C., Ekama, G.A., 2000. Integrated chemical-physical processes modeling -II. Simulating aeration treatment of anaerobic digester supernatants. Water Research 34 (6), 1868e1880. Ohlinger, K.N., Young, T.M., Shroeder, E.D., 1998. Predicting struvite formation in digestion. Water Research 32 (12), 3607e3614. Pambrun, V., 2005. Analyse et mode´lisation de la nitrification partielle et de la pre´cipitation concomitante du phosphore dans un re´acteur a` alimentation se´quence´. INSA, Toulouse. Pandolfi, D., Pons, M.N., da Motta, M., 2007. Characterization of PHB storage in activated sludge extended filamentous bacteria by automated colour image analysis. Biotechnology Letters 29, 1263e1269. Parkhurst, D.L., Thorstenson, D.C., Plummer, L.N., 1980. PHREEQC - A computer program for geochemical calculations. U.S. Geol. Surv. Water Resour. Invest. Rep. 80 (96), 210e219.

Rahman, R.N., Ghaza, F.M., Salleh, A.B., et al., 2006. Biodegradation of hydrocarbon contamination by immobilized bacterial cells. Journal Microbiology 44 (3), 354e359. Ren, T.-T., Liu, L., Sheng, G.-P., Liu, X.-W., Yu, H.-Q., Zhang, M.-C., Zhu, J.-R., 2008. Calcium spatial distribution in aerobic granules and its effects on granule structure, strength and bioactivity. Water Research 42 (13), 3343e3352. Saidou, H., Korchef, A., Ben Moussa, S., Ben Amor, M., 2009. Struvite precipitation by the dissolved CO2 degasification technique: impact of the airflow rate and pH. Chemosphere 74 (2), 338e343. Seckler, M.M., Bruinsma, O.S.L., van Rosmalen, G.M., 1996. Phosphate Removal in fluidized bed I. Identification of physical processes. Water Research 30 (7), 1585e1588. Shu, L., Schneider, P., Jegatheesan, V., Johnson, J., 2006. An economic evaluation of phosphorus recovery as struvite from digester supernatant. Bioresource Technology 97 (17), 2211e2216. Skoog, Hoeller, Nieman, 2003. In: de Boeck (Ed.), Principes D’analyse Instrumentale, fifth ed. Snoeyink, V.L., Jenkins, D., 1980. Water Chemistry. John Wiley and Sons, New York. Suzuki, K., Tanaka, Y., Kuroda, K., Hanajima, D., Fukumoto, Y., Yasuda, T., 2006. The technology of phosphorus removal and recovery from swine wastewater by struvite crystallization reaction. Bioresource Technology 96 (14), 1544e1550. van Rensburg, P., Musvoto, E.V., Wentzel, M.C., Ekama, G.A., 2003. Modelling multiple mineral precipitation in anaerobic digester liquor. Water Research 37 (13), 3087. Wan J. 2009. Interaction entre l’e´limination des polluants azote´es et la formation des granules ae´robies en re´acteur biologique se´quence´. PhD thesis, INSA (Toulouse). Wan, J., Sperandio, M., 2009. Possible role of denitrification on aerobic granular sludge formation in sequencing batch reactor. Chemosphere 75 (2), 220e227. Wan, J., Bessie`re, Y., Spe´randio, M., 2009. Alternating anoxic feast/aerobic famine condition for improving granular sludge formation in sequencing batch airlift reactor at reduced aeration rate. Water Research 43 (20), 5097e5108. Wang, T., Dorner-Reisel, A., Mu¨ller, E., 2004. .Thermogravimetric and thermokinetic investigation of the dehydroxylation of a hydroxyapatite powder. Journal of the European Ceramic Society 24 (4), 693. Wang, H.L., Yu, G.L., Liu, G.S., Pan, F., 2006. A new way to cultivate aerobic granules in the process of paper-making wastewater in a sequencing batch reactor. Biochemical Engineering 28 (11), 99e103. Yilmaz, G., Lemaire, R., Keller, J., Yuan, Z., 2007. Simultaneous nitrification, denitrification and phosphorus removal from nutrient-rich Industrial wastewater using granular sludge. Biotechnology and Bioengineering 100 (3), 529e541. Zhu, Gui-bing, Yong-zhen, Peng, Shu-yun, Wu, Shu-ying, Wang, Shi-wei, Xu, 2007. Simultaneous nitrification and denitrification in step feeding biological nitrogen removal process. Journal of Environmental Sciences 9 (19), 1043e1048.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 8 7 e3 7 9 6

Available at www.sciencedirect.com

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

Effects of water parameters on the degradation of microcystinLR under visible light-activated TiO2 photocatalyst Miguel Pelaez a, Armah A. de la Cruz b, Kevin O’Shea c, Polycarpos Falaras d, Dionysios D. Dionysiou a,* a

Environmental Engineering and Science Program, University of Cincinnati, Cincinnati, OH 45221-0071, USA Office of Research and Development, U.S. Environmental Protection Agency, Cincinnati, OH 45268, USA c Department of Chemistry and Biochemistry, Florida International University, University Park, Miami, FL 3319, USA d Institute of Physical Chemistry, NCSR Demokritos, 15310 Aghia Paraskevi, Attiki, Greece b

article info

abstract

Article history:

A study was performed to determine the effect of pH, alkalinity, natural organic matter (NOM)

Received 29 January 2011

and dissolved oxygen in the performance of nitrogen and fluorine doped TiO2 (NF-TiO2) for

Received in revised form

the degradation of hepatotoxin microcystin-LR (MC-LR) in synthetic and natural water under

16 April 2011

visible light irradiation. The initial degradation rate of MC-LR was fastest under acidic

Accepted 18 April 2011

conditions (3.50
0.02  103 mM min1 at pH 3.0) and decreased to 2.29
0.07  103 and

Available online 23 April 2011

0.54
0.02  103 mM min1 at pH 5.7 and 7.1, respectively. Attractive forces between the

Keywords:

photocatalytic decomposition of MC-LR resulting from increased interfacial adsorption. For

opposite charged MC-LR and NF-TiO2 are likely responsible for the enhancement in the NF-TiO2

carbonate buffered solutions, the photocatalytic activity of NF-TiO2 was reduced when

NOM

increasing the carbonate concentration up to 150 mg CaCO3 L1. The scavenging of radical

pH

species by the bicarbonate ion at pH 7.1 is discussed. In the presence of NOM, the degradation

Alkalinity

rates decreased as pH and initial concentration of the NOM increased. The inhibition was

Dissolved oxygen

higher with fulvic acid than humic acid under alkaline conditions. Oxygenated solution

Water treatment

yields higher NF-TiO2 photocatalytic degradation of MC-LR compared to nitrogen sparged

MC-LR

solution at pH 5.7. The involvement of specific reactive oxygen species implicated in the photodegradation is proposed. Finally, no significant degradation is observed with various natural waters spiked with MC-LR under visible light (l > 420 nm) but high removal was achieved with simulated solar light. This study provides a better understanding of the interactions and photocatalytic processes initiated by NF-TiO2 under visible and solar light. The results indicate solar photocatalytic oxidation is a promising technology for the treatment of water contaminated with cyanotoxins. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Recently, modifications towards visible light response of titanium dioxide (TiO2) has extended the capacity of TiO2 to utilize a larger portion of the solar spectrum, a renewable

source of energy, for a wide variety of applications including environmental remediation and energy conversion. The use of non-metals (i.e., nitrogen, fluorine, sulfur and/or carbon) has proven to be an effective strategy for doping TiO2 for the degradation of organic contaminants and bacteria in water

* Corresponding author. Tel.: þ1 513 556 0724; fax: þ1 513 556 2599. E-mail address: [email protected] (D.D. Dionysiou). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.036

3788

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 8 7 e3 7 9 6

under visible light irradiation (Asahi et al., 2001; Choi et al., 2007; Czoska et al., 2008; Fu et al., 2006; Goldstein et al., 2008; Kontos et al., 2008; Li et al., 2005; Park et al., 2009; Pelaez et al., 2009, Pelaez et al., 2010a,b; Periyat et al., 2008; RengifoHerrera et al., 2009). An emerging issue in drinking water industry is the presence of cyanotoxins in fresh and brackish water sources. Cyanotoxins are naturally produced metabolites of cyanobacteria (also known as blue-green algae) and they have been positively identified throughout the world and an increase in their occurrence and persistence is evident (Pelaez et al., 2010b). The most frequently detected are microcystins, including microcystin-LR (MC-LR), which exhibit the highest environmental concentrations of any of the cyanotoxins analyzed. Unfortunately, the concentration of these toxins often exceeds the provisional guidelines set by international agencies (i.e., 1 mg L1 by the World Health Organization) (Antoniou et al., 2005, 2008). Established drinking water treatment processes for recalcitrant contaminants have been a challenge for drinking water suppliers (Antoniou et al., 2005; Graham et al., 2010). With this in mind, alternative technologies to ensure the availability of clean potable water are of great interest and non-metal doped TiO2 is a good candidate. MC-LR has been found susceptible to photocatalytic degradation with undoped TiO2 under UV (Antoniou et al., 2009; Feitz et al., 1999; Lawton et al., 2003) and non-metal doped TiO2 under visible light (Choi et al., 2007; Pelaez et al., 2009, Pelaez et al., 2010a,b). However, most of the studies evaluated these materials with ‘clean’ synthetic water. Several matrices and conditions present in natural water can influence the photocatalytic reaction of heterogeneous systems such as TiO2-photomediated processes. The characteristics of the solution matrix (i.e., alkalinity, pH, natural organic matter and dissolved oxygen) are critical to determine the photocatalytic degradation of organic contaminants in water with this so-called ‘green’ technology (Antoniou et al., 2005). Their role on the photocatalytic process needs to be evaluated and understood to develop effective solar-driven technologies for water purification. Specific anions can influence the degradation process by affecting the concentration of reactive oxygen species formed during the photocatalytic process (Dionysiou et al., 2000). For instance, anions can react with the hydroxyl radical to form the corresponding anion radical. Carbonate ions are strong scavengers of hydroxyl radicals resulting in carbonate radical (CO3) (Zhu et al., 2007). Another route to inhibit the photocatalytic degradation of organic compounds with TiO2 is the adsorption of the anions at the active sites of the TiO2 surface. This generates a competition between the anion and the organic compound to interact with the surface where the reactive oxygen species are generated. The inhibition of substrate adsorption depends on the concentration of anions adsorbed which is influenced by solution pH. Zhu et al. (2007) found that the oxidation rates of aqueous ammonia with TiO2 under UV light were inhibited by CO32 but not by other inorganic anions (i.e., SO42, H2PO4/HPO42) at pH w9 and w10. Nevertheless, at lower pH values (w7.5), all anions tested were found to hinder the photocatalytic oxidation of ammonia. A major factor influencing the rate of degradation of organic compounds in TiO2-based photocatalytic processes is pH. Changes in pH could affect the surface charge of TiO2 and

the ionization of compounds present in solution is critical to the adsorption and interaction between catalyst and toxin. MC-LR initial degradation rate increased in acidic pH when employing thin TiO2 films compared to thick films (Antoniou et al., 2009). On the other hand, at neutral pH the interactions between the catalysts and the toxin were weaker and both types of films exhibited similar reaction rate constants. The presence of natural organic matter (NOM) can have a pronounced influence on the photocatalytic degradation of organic molecules via TiO2. NOM, which consists mainly of fulvic and humic substances, can act as photosensitizers (lw290 nm to visible light) and mediate photodegradation. Under natural sunlight, several microcystins showed indirect photolytic degradation when natural dissolved organic matter and fulvic acids were present in natural waters (Welker and Steinberg, 2000). Feitz et al. (1999) suggested that high concentrations of natural dyes secreted by cyanobacteria in natural water scavenged the surface-generated hydroxyl radicals or the organic radicals generated can react among the exudate and lose their oxidizing ability. Kull et al. (2006) studied the influence of NOM during the oxidation of MC-LR by chlorine dioxide and confirmed that the degradstion rate decreases in the presence of NOM. This tendancy was stronger in the case of humic acid than that observed with fulvic acid, probably due to the higher degree of aromaticity of the later (Kull et al., 2006). These NOM constituents also retarded the photocatalytic degradation of carbamazepine, clofibric acid and iomeprol with two commercially available TiO2 materials by competition for active sites and surface deactivation of the catalyst by adsorption (Doll and Frimmel, 2005). The concentration of dissolved oxygen is crucial in the photocatalytic oxidation process. The addition of oxygen in solution will reduce the electron-hole recombination process and increase the degradation rates through the oxidative pathways. However, the effect of oxygen is not always positive for the photocatalytic system. For dye-sensitized metal doped TiO2, the presence of dissolved oxygen can completely inhibit the photoreductive activity or reduced the visible light reactivity, depending on the metal deposited (Bae and Choi, 2003). In this case, oxygen competes for the electrons on the conductive band with the target molecule in the reductive pathway. In this study, the effects of alkalinity, pH, NOM and dissolved oxygen were evaluated using nitrogen and fluorine doped TiO2 (NF-TiO2) nanoparticles for the photocatalytic degradation of MC-LR under visible light. To the best of our knowledge, this is the first study that utilizes visible lightactivated TiO2 photocatalyst to examine the effect of these water parameters during the photocatalytic transformation of contaminants, and in particular of MC-LR, in water. These results provide insight in the area of visible light-activated photocatalysis and the influence of key water parameters.

2.

Materials and methods

2.1.

Reagents and sample preparation

MC-LR standard was obtained from Calbiochem. Suwannee River humic and fulvic acids (SRHA and SRFA, respectively) were purchased from the International Humic Substances

3789

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 8 7 e3 7 9 6

Society (IHSS, University of Minnesota, St. Paul, Minnesota USA) and used as representatives of NOM in water. Stock solutions were prepared in MilliQ-grade water, sonicated for 1 h and filtered with a 0.45 mm glass fiber filter (Acrodisc GHP, Pall Corporation, East Hills, NY USA). From this, specific amounts of solution were transferred to each reactor to achieve initial concentrations of 5 and 10 mg L1 of SRHA and SRFA. Solutions with specific concentrations (50, 100 and 150 mg/L) were prepared by adding different amounts of sodium carbonate (Na2CO3, Fisher) into MilliQ-grade water to evaluate the effect of alkalinity. In solutions in which 50 mg L1 of sodium carbonate were added, the pH was 10.3. Sodium hydroxide (Fisher) and monopotassium phosphate (Fisher) were used to prepare buffer solutions at slightly alkaline pH (i.e., 7.1 and 8.0). Nitric acid was used to adjust the pH at 3.0 in experiments dealing with acidic pH. To evaluate the effect of dissolved oxygen, MilliQ-grade water was sparged with N2 or saturated with O2 for 30 min prior to use without subsequent addition during the experiment. NF-TiO2 nanoparticles were prepared using a modified sol-gel method previously developed and denominated as Particle 1 (Pelaez et al., 2009). In brief, a fluorosurfacant was used as a pore template and fluorine dopant and ethylenediamine as nitrogen precursor. The sol-gel was stirred for 24 h, dried at room temperature and calcined at 400  C in air. Aqueous suspensions of NF-TiO2 nanoparticles were freshly prepared for every experiment and used after 30 min of sonication (2510R-DH, Bransonic).

2.2.

Photocatalytic experiments

A specific solution previously adjusted for the desired experiment (i.e., alkalinity, pH, dissolved oxygen or NOM analysis) with an aqueous NF-TiO2 suspension was spiked with an aliquot of MC-LR standard to reach an initial concentration of 500
10 mg L1. The reactor was sealed with parafilm and mixed continuously during the photocatalytic reaction with

visible light irradiation. To obtain visible light irradiation only, a UV block filter (UV420, Opticology) was mounted below two 15 W fluorescent lamps (ColeeParmer) to eliminate spectral range below 420 nm. To certify that no UV light passed through the filter, a spectroradiometer (Ocean Optics) was employed to obtain the light spectrum after the filter. Fig. 1 shows the lamp spectrum with and without UV block filter. UV wavelengths with peaks at around 310, 356 and 410 nm can be observed when the filter was not mounted. The light intensity of the visible light was 7.81  105 W cm2 and it was determined with a radiant power meter (Newport Corporation). For the solar experiments with the natural water samples, the filter was removed and the light intensity was 9.52  105 W cm2. To prevent evaporation, a fan was positioned behind the reactor. Sampling was done at specific periods of time and the samples were quenched with methanol to stop any further reaction, filtered (L815, Whatman) to remove the suspended nanoparticles, transferred to 0.2 ml glass inserts and placed in sample vials. MC-LR samples were analyzed by liquid chromatography (LC, Agilent Series 1100). The analytical conditions were similar to those reported by Antoniou et al. (2008) but the column employed was a C18 Discovery (Supelco) column (4.6 mm  150 mm, 2.1 mm particle size). The flow rate was 0.2 ml/min and the injection volume was 20 ml. A Spectramax M2 spectrophotometer (Molecular Devices Corp., Sunnyvale, CA) was employed to obtain the absorbance for experiments involving NOM analysis without MC-LR. Samples were filtered prior to analysis. All the experiments were conducted in an Advanced Sterilchemgard III Class II biological safety cabinet (Baker Company, Sanford, ME) with full exhaust since MC-LR is highly toxic and an irritant, so appropriate handling of the toxin is needed. Water samples obtained from Lake Erie and St. John’s River (Florida) were employed to evaluate the photocatalytic efficiency of NF-TiO2 in natural waters. Water quality parameters, including pH, total alkalinity, total hardness and

5000 Lamp Spectrum with UV block filter Lamp Spectrum without UV block filter

Intensity (a.u)

4000

3000

2000

1000

0 200

300

400

500

600

700

Wavelength (nm) Fig. 1 e Spectral distribution of fluorescent lamp after UV cut off filter.

800

3790

Micro solid-phase extraction (Micro-SPE) for NOM

The method was developed based on Rivasseau et al. (1998) work but in smaller scale due to the total volume from each sample. All SPE cartridges (1cc/50 mg, Waters Corporation) were conditioned with 1 ml of acetonitrile, followed by 1 ml of methanol and 1 ml of water. Each sample was diluted to obtain a final concentration of 5% of methanol and then percolated through the cartridge. Afterwards, the cartridge was washed with 1 ml of 20% methanol solution. MC-LR was eluted with 250 ml of 75% methanol solution and then with 100 ml of 85% methanol solution. Finally, each sample was transferred to a 250 ml insert and analyzed with liquid chromatography, as described above.

2.2.2.

Recovery test

A recovery test was performed to determine the recovery percentage of MC-LR after the solid phase extraction and liquid chromatography analysis. Standard solutions with MC-LR only and with SRFA and SRHA were prepared. An average recovery of 85.3
5.2% was obtained. This recovery was verified during the analysis of the samples of each experiment.

3.

Results and discussion

3.1.

Effect of pH

To explore the effect of pH, the photocatalytic degradation of an initial concentration of 500 mg L1 of MC-LR with NF-TiO2 was explored at four pH conditions (3.0, 5.7, 7.1 and 8.0). Fig. 2a shows the toxin degradation profile under visible light for the four pH values. To assure that the observed MC-LR removal occurred mostly when both the catalyst and visible light were present, dark adsorption was studied for MC-LR at all pHs. Under dark conditions, the adsorption of MC-LR was approximately 39.6% at pH 3.0 after 5 h compared to 30.1% at pH 5.7. When increasing the pH solution to 7.1 and 8.0, the adsorbed MC-LR percentage dropped to 9.2% and 7.9%, respectively. The initial degradation rates have been found to be significantly influenced by pH as shown in Fig. 2b. The highest initial reaction rate after 120 min of visible light irradiation was found at pH 3.0 (3.50
0.02  103 mM min1) and decreased at pH 5.7 (2.29
0.07  103 mM min1). When the solution pH was neutral or alkaline (pH 7.1 and 8.0), the degradation rate significantly decreased (0.54
0.02  103 mM min1 and 0.36
0.05  103 mM min1, respectively). This is in agreement with our previous study when employing 1000 mg L1 of initial toxin concentration. We determined the point of zero charge (PZC) for NF-TiO2 to be 6.4 (Pelaez et al., 2009). At a pH higher than the PZC, the TiO2 surface becomes hydroxylated (TiOH) and the overall surface becomes negatively charged. At lower pH, the surface is positively charged and the surface

a

1.0

0.8

0.6

0.4 pH pH pH pH

0.2

3.0 5.7 7.1 8.0

0.0 0

1

2

3

4

5

6

Reaction time (hr)

M min-1)

2.2.1.

MC-LR Reaction Rate x 10-3

conductivity, were measured in the Greater Cincinnati Water Works water quality certified laboratory prior to the experiments. The amount of natural organic matter was estimated as total organic carbon (TOC) and was obtained using a TOC analyzer (Shimadzu). All the experiments were performed in triplicates and the standard error from the mean is represented by vertical error bars on the graphs.

Normalized MC-LR Concentration (C/Co)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 8 7 e3 7 9 6

4

b

Point of Zero Charge of NF-TiO2 -

(TiOH2+)

3

(TiO )

pH 2.2< (MC-LRH )
2

1

0 3

4

5

6

7

8

pH

Fig. 2 e a) Photocatalytic degradation of MC-LR in the presence of NF-TiO2 at pH 3.0, 5.7, 7.1 and 8.0 under visible light b) MC-LR initial degradation rates after 120 min of reaction time with NF-TiO2 at different pH conditions.

becomes protonated (TiO2Hþ). Lawton et al. (2003) found that a net charge for each microcystin can be calculated by identifying the likely species to be protonated or dissociated at different pH values. According to their calculations, MC-LR is positively charged at pH 2.1 and below and negatively charged at higher pH values. As a result, attractive forces between the catalyst and the toxin are expected at the surface where the oxidizing species are formed at pH below the PZC and repulsive forces above the PZC. Nevertheless, a small degradation percentage was observed after 300 min of reaction time (27% at pH 7.1 and 16% at pH 8.0; see Fig. 1). A possible lag phase followed by oxidation of MC-LR through diffusion of the oxidizing species to the bulk of the solution, along with weak van der Waals forces between the catalyst and the toxin, is most likely responsible of this observation. The reaction rates after 120 min with an initial MC-LR concentration of 1000 mg L1 (Pelaez et al., 2009) were relatively higher compared to an initial concentration of 500 mg L1 at pH 3.0 and 5.7, respectively, with NF-TiO2 under visible light. The dependency of the reaction rate on concentration indicates that the photocatalytic degradation of MC-LR follows pseudo-first order reaction during the initial stages of the photocatalytic process.

Normalized MC-LR Concentration (C/Co)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 8 7 e3 7 9 6

3791

1.0

0.9

0.8 pH 7.1 50 mg L-1 Na2CO3 , no phosphate buffer 50 mg L-1 Na2CO3 @ pH 7.1

0.7

100 mg L-1 Na2CO3 @ pH 7.1 150 mg L-1 Na2CO3 @ pH 7.1

0.6 0

1

2

3

4

5

6

Reaction time (hr)

Fig. 3 e The effect of alkalinity at three different Na2CO3 concentrations for the photodecomposition of MC-LR in the presence and absence of phosphate buffered solution.

3.2.

Fig. 4 e Photolysis of SRFA and SRHA only under visible light. Insert shows the UVevis absorbance spectra of each representative natural organic matter.

Effect of alkalinity

As shown in Fig. 3, the addition of 50 mg L1 of Na2CO3 without any phosphate buffer solution (pH 10.3) inhibited completely the degradation of MC-LR with NF-TiO2 under visible light irradiation. In order to differentiate the contribution between the pH and the presence of carbonate/bicarbonate ions to the inhibition of MC-LR degradation, a phosphate buffer solution was employed. In this case, at 50 mg L1 of Na2CO3 with a pH of 7.1, the initial degradation rate of the cyanotoxin (after 120 min of reaction time) was almost identical to that obtained in the buffered solution only (0.54
0.02  103 and 0.54
0.04  103 mM min1, respectively). When the concentration increased to 100 mg L1 of Na2CO3, a decrease in the initial reaction rate was observed with a value of 0.34
0.02  103 mM min1. This represents a reduction of approximately 35% in the same reaction time when doubling the concentration of sodium carbonate. The addition of 150 mg L1 of Na2CO3 was more evident in the decrease of the initial reaction rate (0.09
0.01  103 mM min1) for 120 min of reaction time. Thus, by increasing the Na2CO3 concentration three times, a reduction of 80% in the initial reaction rate was obtained. The main species responsible for alkalinity in water are bicarbonate ion (HCO3), carbonate ion (CO32) and hydroxide ion. Both carbonate and bicarbonate can act as a scavenger of the radical species produced (i.e., hydroxyl radical, superoxide radical anion) and reduce the initial degradation rates of MC-LR. They can also partially inhibit the surface photocatalytic reactions with MC-LR by increasing the negative charge at the surface of NF-TiO2 at pH above the PZC. The equilibrium between all of them is pH dependent. At pH 7.0, the predominant ion is bicarbonate and at pH greater than 10, carbonate ion will mainly be present in solution according to their pKa values (Kumar and Mathur, 2006). Lower initial rates were obtained by increasing the alkalinity concentration at a constant pH value which indicates that the bicarbonate ion concentration increased and therefore the scavenging effect was more pronounced. In the case where no phosphate buffer was employed and 50 mg L1 of Na2CO3 with a pH of 10.3 were

added, carbonate ion was mainly present in solution and the reactivity with the radical species generated is faster than with bicarbonate ion (Nemes et al., 2000) making it a stronger scavenger. Additionally, the increase in the pH value contributed to the repulsive forces between MC-LR and NF-TiO2 (see Section 3.1). Therefore, with 100 mg L1 of Na2CO3 at pH 7.1, there is a reduction on the degradation of MC-LR but with 50 mg L1 of Na2CO3 and no phosphate buffer (pH 10.3) the inhibition was higher. Thus, by adjusting the pH, the inhibiting effect of alkalinity can be altered in the system using NF-TiO2.

3.3.

Effect of NOM

Fig. 4 shows that no photolytic degradation of either SRFA or SRHA was observed under visible light irradiation in the absence of the catalyst. Humic and fulvic substances are known to be photosensitizers in natural waters under solar light; however, it is clearly observed that no significant absorption occurs in the visible region according to the UVevis spectra (see insert in Fig. 4). As expected, no change in the concentration of SRFA and SRHA is observed over time under the conditions tested. To check the possibility of NOM degradation in the presence of NF-TiO2, dark conditions experiments and irradiation under visible light at three different pH values (3.0, 5.7 and 7.1) with SRFA and SRHA were carried out. The residual absorbance at 254 nm (UV254) was monitored as a function of reaction time. This wavelength indicates the absorbance of aromatic structures or conjugated double bonds in the humic and fulvic acid molecules. It was found that under dark conditions, the adsorption of SRFA onto NF-TiO2 was higher than that of SRHA but both were pH dependent. At pH 3.0, the highest adsorption was observed for both SRFA and SRHA at 5 and 10 mg L1; the adsorption capacity decreased when increasing pH. For instance, the average relative UV254 (%) of an initial concentration of 5 mg L1 of SRHA was 30.1%, 22.5% and 16.7% at pH 3.0, 5.7 and 7.0, respectively, after 5 h. In the case of an initial concentration of 5 mg L1 of SRFA, the UV254 (%) in solution was

3792

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 8 7 e3 7 9 6

55.1%, 32.3% and 24.1% at pH 3.0, 5.7 and 7.0, respectively. Organic matter is comprised of a wide variety of protonated and unprotonated moieties (i.e., carboxylic and hydroxyl acids, carbonyls, ketones) (Westerhoff et al., 2007). Under acidic conditions, the organic matter becomes negatively charged overall and the NOM interactions are governed by electrostatic attraction towards the positively charged NF-TiO2. In alkaline solutions, the many acidic functional groups are deprotonated and negatively charged and repulsive forces are dominant. A higher adsorption capacity of SRFA onto NF-TiO2 was obtained and may be due to the more functional groups of acidic nature in the molecule and lower hydrophobicity (Esparza-Soto and Westerhoff, 2003). Few studies have used humic or fulvic acid to evaluate the photocatalytic activity of synthesized doped photocatalysts under visible light. Li et al. (2007) observed a small degradation of humic acid when employing palladium-modified nitrogendoped TiO2. With the optimum palladium loading, they obtained less than 20% reduction of an initial concentration of 100 mg L1 of humic acid after 10 h of irradiation time (Li et al., 2007). However, no adsorption studies were performed to differentiate between both phenomena. In our study, no significant difference was obtained between adsorption under dark conditions and photocatalytic degradation with visible light of SRFA and SRHA (data not shown). This can be attributed to the complexity of each molecule, the slower production of the radical species formed (see also effect of dissolved oxygen) and the competition between possible byproducts produced and these radicals formed. The byproducts can have similar absorbance at 254 nm. This suggests that the reactions rates are not controlled by adsorption but are rather limited by the generation rate of the radical species responsible for the oxidation reactions. The results obtained for the degradation of MC-LR with NFTiO2 under visible light in the presence of SRFA or SRHA in water are depicted in Fig. 5. The presence of SRFA or SRHA reduced the overall MC-LR degradation even though some significant reduction of the toxin can be observed, in particular at pH 3.0. Table 1 lists the initial reaction rates of MC-LR in the presence of two different concentrations of SRFA and SRHA under various pH values. It can be observed that, for most cases, the degradation rates decreased as pH and initial concentration of the NOM increased. Moreover, the inhibition was higher with fulvic acid that humic acid at all pH values. For instance, a reaction rate of 3.48
0.05  103 mM min1 was determined for 10 mg L1 of SRHA compared to a reaction rate of 3.08
0.06  103 mM min1 for 10 mg L1 of SRFA at pH 3.0. This is related to the higher adsorption of SRFA and the direct competition for active sites and surface-generated oxidizing species on NF-TiO2 (Feitz et al., 1999). When comparing to Fig. 2, the degradation of MC-LR is inhibited by the presence of SRFA and SRHA according also to the initial reaction rates values reported in Table 1. Nevertheless, MC-LR is still being degraded under acidic conditions and a considerable reduction of the reaction rates is observed at alkaline pH. The role of natural organic matter may not be simply as a radical scavenger. Paul et al. (2007) suggested the formation of a surface coordination complex between adsorbed fluoroquinolone, a synthetic broad-spectrum antibiotic, and TiO2 that can mediate the electron transfer from the conduction band of

TiO2 to an appropriate electron acceptor. A similar mechanism can occur with SRHA and SRFA when they are highly adsorbed at low pH on the NF-TiO2 surface. At higher pH values, the adsorption of both types of NOM is less and the scavenging effect becomes dominant.

3.4.

Effect of dissolved oxygen

To assess the photoreactivity of NF-TiO2 towards MC-LR in the presence of excess oxygen and limited oxygen under visible light irradiation, the solution was sparged with oxygen or nitrogen for 30 min prior to irradiation and the results are shown in Fig. 6. The photocatalytic conversion of MC-LR was higher when the solution was sparged with oxygen with an initial reaction rate of 3.00
0.08  103 mM min1 compared to the control at pH 5.7 (2.29
0.07  103 mM min1) after 120 min of reaction time. The opposite effect was observed when the oxygen was limited by sparging nitrogen gas before irradiation. This is depicted as a reduction on the degradation efficiency (as seen in Fig. 6) during the same reaction time with a degradation rate of 1.87
0.18  103 mM min1. This implies that the photocatalytic degradation of MC-LR was relatively more efficient under O2 sparged conditions than under N2 sparged conditions. Higher reaction rates at high dissolved oxygen concentrations had been attributed to the scavenger effect of conduction band electrons by the dissolved O2 molecules and the generation of superoxide anion radical that can subsequently lead to the formation of hydroxyl radical and singlet oxygen (Cho et al., 2004). When oxygen concentration is reduced, the photogenerated electron-hole pair’s recombination rate is faster and the rate of generation of reactive oxygen species decreased. In order to identify the dominant radical species formed under the conditions tested, an alcohol quenching agent was employed. In the presence of methanol, a well-known OH free and surface radical scavenger, MC-LR degradation was significantly inhibited and a reaction rate of 0.91
0.11  103 mM min1 was obtained after 120 min of irradiation even though the solution was sparged with oxygen. Nevertheless, even in the presence of methanol, some degradation is observed indicating that other reactive oxygen species that can be formed in such process (Cho et al., 2004) may be present in the system and can play a role in the photocatalytic degradation of MC-LR with NF-TiO2. Fu et al. (2006) reported the presence of hydroxyl and superoxide radicals during the photoactivation of N-doped TiO2 under visible light. The peak intensity of the electron spin resonance signals for each radical was more intense under UV irradiation than with visible light, suggesting a lower production of hydroxyl and superoxide radicals. The authors attributed this to a Lewis acid-base-type of reaction that is responsible for initiating the generation of the reactive oxygen species, which differs from the well established water photoxidation via electron-transfer-type reaction. Moreover, the conjugated double bond at the ADDA moiety of the MC-LR molecule, which plays a critical role in the biological activity (Antoniou et al., 2008), is the most sensitive part of the cyanotoxin molecule with respect to singlet oxygen oxygenation (1O2) (Welker and Steinberg, 2000). The formation mechanism of singlet oxygen is closely related to the presence of superoxide radicals since the latter can undergo further oxidation to

3793

Normalized MC-LR Concentration (C/Co)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 8 7 e3 7 9 6

a 1.0

pH 3.0 5ppm SRHA @ pH 3.0 10ppm SRHA @ pH 3.0 5ppm SRFA @ pH 3.0 10ppm SRFA @ pH 3.0

0.8

0.6

0.4

0.2

Normalized MC-LR Concentration (C/Co)

0.0

b

1.0 pH 5.7 5ppm SRHA @ pH 5.7 10ppm SRHA @ pH 5.7 5ppm SRFA @ pH 5.7 10ppm SRFA @ pH 5.7

0.8

0.6

0.4

0.2

Normalized MC-LR Concentration (C/Co)

0.0

c

1.0

0.8

0.6

0.4 pH 7.1 5ppm SRHA @ pH 7.1 10ppm SRHA @ pH 7.1 5ppm SRFA @ pH 7.1 10ppm SRFA @ pH 7.1

0.2

0.0 0

1

2

3

4

5

6

Reaction time (hr) Fig. 5 e Photocatalytic degradation of MC-LR in the presence of 5 and 10 mg LL1 of SRFA and SRHA at a) pH 3.0, b) 5.7 and c) 7.1 under visible light.

3794

Table 1 e MC-LR initial degradation rates with 5 and 10 mg LL1 of SRFA and SRHA after 120 min of visible light irradiation with NF-TiO2. The confidence interval is based on the standard error of the mean values. 3

NOM Concentration MC-LR Initial Reaction Rate  10 (mM min1) (mg L1) SRHA 5 10 SRFA 5 10 Control

pH 3.0 3.49
0.02 3.48
0.05

pH 5.7 2.31
0.07 2.27
0.07

pH 7.1 0.29
0.03 0.26
0.06

3.20
0.06 3.08
0.06 3.50
0.02

2.12
0.11 2.01
0.07 2.29
0.07

0.46
0.03 0.31
0.06 0.54
0.02

Normalized MC-LR Concentration (C/Co)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 8 7 e3 7 9 6

1.0

0.8

0.6

0.4

WE6 + visible light WE7 + visible light WE9 + visible light GWL + visible light WE7 + simulated solar light GWL + simulated solar light

0.2

0.0 0

1

2

3

4

5

6

Reaction time (hr)

form singlet oxygen. The redox potential of 0.34 V vs NHE indicates that the relation (O2/1O2) is favored thermodynamically (Rengifo-Herrera et al., 2009). This can explain the enhanced reduction on the normalized concentration of MC-LR when the solution was sparged with oxygen and the photocatalytic activity of NF-TiO2 in the presence of methanol. It should be noted that ongoing studies are being carried out in this group to elucidate the mechanism of visible lightactivated photocatalysis including the formation of reactive oxygen species for this particular photocatalyst.

3.5.

Natural water samples

Normalized MC-LR Concentration (C/Co)

Fig. 7 shows the results when the photocatalyst was added to four natural waters from three utilities (over intake) from the western end of Lake Erie (named WE6, WE7 and WE9) and the St. John’s River in Florida (named GWL) spiked with MC-LR. Besides providing a natural-containing aqueous matrix to assess the performance of NF-TiO2 photocatalyst, these samples were collected from regions that have reported increasing episodes of cyanobacterial blooms over the last decade. Several potentially toxin (including MC-LR)-producing cyanobacteria have been monitored and identified in both locations (Pelaez et al., 2010b). Essentially, no degradation of MC-LR occurs when employing visible light only. This observation is related to the surface

1.0

0.8

Fig. 7 e Evaluation of the photocatalytic activity of NF-TiO2 with different natural waters spiked with MC-LR under visible and solar light.

interaction and competitive phenomena involving the water parameters present in solution and discussed in the above sections. As seen in Table 2, all of the waters tested contained high concentrations of alkalinity, had pH values in the alkaline region, and had relatively high natural organic matter content (expressed as TOC), among others. The inhibition of the photocatalytic activity of NF-TiO2 in these natural waters, under visible light irradiation, was likely due to an accumulation of scavenging effects between the water parameters that were evaluated in this study and perhaps other ions that might be present. The contribution of visible light photocatalysis with NF-TiO2 was significantly reduced under these conditions but it is relevant to understand the interactions and photoreactivity of non-metal doped TiO2 materials with inorganic and organic compounds present in water when aiming to solar-driven technologies for drinking water applications. To evaluate the performance of NF-TiO2 under solar light, two natural waters were selected (WE7 and GWL) and the UV block filter was removed from the radiation source set up system (see Fig. 1 for spectrum). As mentioned before, UV radiation with peaks at around 310, 356 and 410 nm can be observed in the absence of filtering. The results obtained using the simulated solar light are shown in Fig. 7. For both cases, a high photocatalytic removal of MC-LR was obtained after 5 h of solar irradiation. This result differs from the data obtained when irradiating WE7 and GWL with visible light only. This

0.6

0.4

Table 2 e Physicochemical properties of western Lake Erie and St. John’s water samples. pH 5.7 N2 purged @ pH 5.7 O2 purged @ pH 5.7 O2 purged + 50mM MeOH

0.2

0.0 0

1

2

3

4

5

6

Reaction time (hr)

Fig. 6 e Performance of NF-TiO2 in oxygen and nitrogen sparged conditions and a scavenger compound (methanol) in the degradation of MC-LR under visible light.

pH Total Alkalinity (mg CaCO3/L) Total Hardness (mg/L) Turbidity (NTU) Conductiviy (mS) Absorbance @ 254 nm (UV254) TOC (mg/L)

WE6

WE7

WE9

GWL

8.5 94.2 104 0.51 691 0.10 4.23

8.25 89.6 94 0.12 572 0.04 2.55

8.60 92.3 92 0.52 657 0.08 3.91

7.40 117.8 110 0.23 873 0.17 9.49

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 8 7 e3 7 9 6

can be attributed to the fact that when using solar light, the photoactivation of NF-TiO2 is enhanced and NOM may act as photosensitizer with the UV portion of the solar spectrum promoting MC-LR degradation. In general, NOM displays increasing absorptivity towards shorter wavelengths and decreases at longer wavelength up to about 400 nm (see insert in Fig. 4 as representative example). When NOM is present in solution, a series of interfacial electron transfers may occur to initiate a photosensitized reaction with MC-LR with the UV portion of the solar light (Welker and Steinberg, 2000). This can lead to a UV light induced degradation mediated by the photosensitization of the NOM present in both natural water samples. Moreover, activation of TiO2 under UV light can increase in the rate of generation of OH radicals and enhance the photocatalytic degradation of MC-LR under the conditions tested.

4.

Conclusions

We systematically investigated several water parameters that influence the visible light photocatalytic process and surface interaction of NF-TiO2 with MC-LR. The effect of selected inorganic and organic substances typically found in natural waters was individually tested. We found that the solution pH is a major factor influencing the degradation rate of MC-LR with NF-TiO2. An enhancement in the photocatalytic activity of NF-TiO2 was observed at pH 3.0 and decreased as pH increased. The speciation of the carbonate ion, which is pH dependent, had an influence in the initial degradation rate of MC-LR. An increase in the Na2CO3 concentration inhibited the degradation of MC-LR at phosphate buffered solutions at pH 7.1. The presence of NOM, represented by SRFA and SRHA, had a relatively negative impact on the performance of NF-TiO2 towards the removal of MC-LR and its role was also influenced by the solution pH. These results can help provide a better understanding of the interactions and mechanisms of degradation of MC-LR in order to improve the viability of visible light-activated TiO2 for the development of solar-driven technologies for the remediation of contaminated water with cyanobacterial toxins or other emerging contaminants of concern.

Disclaimer Although the research described in this article has been funded in part by the U. S. Environmental Protection Agency through grant/cooperative agreement (R833223) to Dionysios D. Dionysiou, it has not been subjected to the Agency’s required peer and policy review and therefore does not necessarily reflect the views of the Agency and no official endorsement should be inferred.

Acknowledgments This work was funded by the U.S. Environmental Protection Agency (R833223), the Ohio State University Research Foundation (OSURF-USGS Project 60021018) and the European

3795

Commission (Clean Water - Grant Agreement number 227017). Clean Water is a Collaborative Project co-funded by the Research DG of the European Commission within the joint RTD activities of the Environment and NMP Thematic Priorities/FP7. We are thankful to Cincinnati Water Works for the assistance provided during the analysis of the water quality analysis. Judy Westrick and GreenWaterLabs are acknowledged for collecting and sending the natural water samples.

references

Antoniou, M.G., de la Cruz, A.A., Dionysiou, D.D., 2005. Cyanotoxins: New generation of water contaminants. J. Environ. Eng. 131 (9), 1239e1243. Antoniou, M.G., Shoemaker, J.A., de la Cruz, A.A., Dionysiou, D.D., 2008. LC/MS/MS structure elucidation of reaction intermediates formed during the TiO2 photocatalysis of microcystin-LR. Toxicon 51 (6), 1103e1118. Antoniou, M.G., Nicolaou, P.A., Shoemaker, J.A., de la Cruz, A.A., Dionysiou, D.D., 2009. Impact of the morphological properties of thin TiO2 photocatalytic films on the detoxification of water contaminated with the cyanotoxin, microcystin-LR. Appl. Catal., B 91 (1e2), 165e173. Asahi, R., Morikawa, T., Aoki, K., Taga, Y., 2001. Visible-light photocatalysis in nitrogen-doped titanium oxides. Science 293, 269e271. Bae, E., Choi, W., 2003. Highly enhanced photoreductive degradation of perchlorinated compounds on dye-sensitized metal/TiO2 under visible light. Environ. Sci. Technol. 37, 147e152. Cho, M., Chung, H., Choi, W., Yoon, J., 2004. Linear correlation between inactivation of E. coli and OH radical concentration in TiO2 photocatalytic disinfection. Water Res. 38 (4), 1069e1077. Choi, H., Antoniou, M.G., Pelaez, M., de la Cruz, A.A., Shoemaker, J.A., Dionysiou, D.D., 2007. Mesoporous nitrogendoped TiO2 for the photocatalytic destruction of the cyanobacterial toxin microcystin-LR under visible light irradiation. Environ. Sci. Technol 41, 7530e7535. Czoska, A.M., Livraghi, S., Chiesa, M., Giamello, E., Agnoli, S., Granozzi, G., Finazzi, E., Di Valentin, C., Pacchioni, G., 2008. The nature of defects in fluorine-doped TiO2. J. Phys. Chem. C 112, 8951. Dionysiou, D.D., Suidan, M.T., Bekoua, E., Baudin, I., Laıˆne´, J.-M., 2000. Effect of ionic strength and hydrogen peroxide on the photocatalytic degradation of 4-chlorobenzoic acid in water. Appl. Catal., B 26, 153e171. Doll, T.E., Frimmel, F.H., 2005. Photocatalytic degradation of carbamazepine, clofibric acid and iomeprol with P25 and Hombikat UV100 in the presence of natural organic matter (NOM) and other organic water constituents. Water Res. 39 (2e3), 403e411. Esparza-Soto, M., Westerhoff, P., 2003. Biosorption of humic and fulvic acids to live activated sludge biomass. Water Res. 37 (10), 2301e2310. Feitz, A.J., Waite, T.D., Jones, G.J., Boyden, B.H., Orr, P.T., 1999. Photocatalytic degradation of the blue green algal toxin microcystin-LR in a natural organic-aqueous matrix. Environ. Sci. Technol. 33, 243e249. Fu, H., Zhang, L., Zhang, S., Zhu, Y., 2006. Electron spin resonance spin-trapping detection of radical intermediates in N-doped TiO2-assisted photodegradation of 4-chlorophenol. J. Phys. Chem. B. 110, 3061e3065. Goldstein, S., Behar, D., Rabani, J., 2008. Mechanism of visible light photocatalytic oxidation of methanol in aerated aqueous suspensions of carbon-doped TiO2. J. Phys. Chem. C 112, 15134e15139.

3796

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 8 7 e3 7 9 6

Graham, D., Kisch, H., Lawton, L.A., Robertson, P.K.J., 2010. The degradation of microcystin-LR using doped visible light absorbing photocatalysts. Chemosphere 78, 1182e1185. Kontos, A.I., Kontos, A.G., Raptis, Y.S., Falaras, P., 2008. Nitrogen modified nanostructured titania: Electronic, structural and visible-light photocatalytic properties. Phys. Stat. Sol. (RRL) 2, 83e85. Kull, T.P.J., Sjo¨vall, O.T., Tammenkoski, M.K., Backlund, P.H., Meriluoto, J.A.O., 2006. Oxidation of the cyanobacterial hepatotoxin microcystin-LR by chlorine dioxide: influence of natural organic matter. Environ. Sci. Technol. 40, 1504e1510. Kumar, A., Mathur, N., 2006. Photocatalytic degradation of aniline at the interface of TiO2 suspensions containing carbonate ions. J. Colloid Interface Sci. 300, 244e252. Lawton, L.A., Robertson, P.K.J., Cornish, B.J.P.A., Marr, I.L., Jaspars, M., 2003. Processes influencing surface interaction and photocatalytic destruction of microcystins on titanium dioxide photocatalysts. J. Catal. 213, 109e113. Li, D., Ohashi, N., Hishita, S., Kolodiazhnyi, T., Haneda, H., 2005. Origin of visible-light-driven photocatalysis: a comparative study on N/F-doped and NeF-codoped TiO2 powders by means of experimental characterizations and theoretical calculations. J. Solid State Chem. 178, 3293e3302. Li, Q., Xie, R., Mintz, E.A., Shang, J.K., 2007. Enhanced visible-light photocatalytic degradation of humic acid by palladiummodified nitrogen-doped titanium oxide. J. Am. Ceram. Soc. 90 (12), 3863e3868. Nemes, A., Fabian, I., van Eldik, R., 2000. Kinetics and mechanism of the carbonate ion inhibited aqueous ozone decomposition. J. Phys. Chem. A 104, 7995e8000. Park, Y., Kim, W., Park, H., Tachikawa, T., Majima, T., Choi, W., 2009. Carbon-doped TiO2 photocatalyst synthesized without using an external carbon precursor and the visible light activity. Appl. Catal., B 91, 355e361. Paul, T., Miller, P.L., Strathmann, T.J., 2007. Visible-light-mediated TiO2 photocatalysis of fluoroquinolone antibacterial agents. Environ. Sci. Technol. 41, 4720e4727. Pelaez, M., de la Cruz, A.A., Stathatos, E., Falaras, P., Dionysiou, D.D., 2009. Visible light-activated N-F-codoped

TiO2 nanoparticles for the photocatalytic degradation of microcystin-LR in water. Catal. Today 144 (1e2), 19e25. Pelaez, M., Falaras, P., Likodimos, V., Kontos, A.G., De la Cruz, A.A., O’Shea, K., Dionysiou, D.D., 2010a. Synthesis, structural characterization and evaluation of sol-gel-based NF-TiO2 films with visible light-photoactivation for the removal of microcystin-LR. Appl. Catal., B 99 (3e4), 378e387. Pelaez, M., Antoniou, M.G., He, X., Dionysiou, D.D., de la Cruz, A.A., Tsimeli, K., Triantis, T., Hiskia, A., Kaloudis, T., Williams, C., Aubel, M., Chapman, A., Foss, A., Khan, U., O’Shea, K.E., Westrick, J., 2010b. Sources and Occurrence of Cyanotoxins Worldwide. In: Fatta-Kassinos, D., Bester, K., Ku¨mmerer, K. (Eds.), Xenobiotics in the Urban Water Cycle. Springer, USA, pp. 101e127. Periyat, P., Pillai, S.C., McCormack, D.E., Colreavy, J., Hinder, S.J., 2008. Improved high-temperature stability and sun-lightdriven photocatalytic activity of sulfur-doped anatase TiO2. J. Phys. Chem. C 112, 7644e7652. Rengifo-Herrera, J.A., Pierzchala, K., Sienkiewicz, A., Forro´, L., Kiwi, J., Pulgarin, C., 2009. Abatement of organics and Escherichia coli by N, S co-doped TiO2 under UV and visible light. Implications of the formation of singlet oxygen (1O2) under visible light. Appl. Catal., B 88, 398e406. Rivasseau, C., Martins, S., Hennion, M.-C., 1998. Determination of some physicochemical parameters of microcystins (cyanobacterial toxins) and trace level analysis in environmental samples using liquid chromatography. J. Chromatogr. A 799 (1e2), 155e169. Welker, M., Steinberg, C., 2000. Rates of humic substance photosensitized degradation of microcystin-LR in natural waters. Environ. Sci. Technol. 34, 3415e3419. Westerhoff, P., Mezyk, S., Cooper, W.J., Minakata, D., 2007. Electron pulse radiolysis determination of hydroxyl radical rate constants with Suwannee River fulvic acid and other dissolved organic matter isolates. Environ. Sci. Technol. 41, 4640e4646. Zhu, X., Nanny, M.A., Butler, E.C., 2007. Effect of inorganic anions on the titanium dioxide-based photocatalytic oxidation of aqueous ammonia and nitrite. J. Photochem. Photobiol. A: Chem. 185, 289e294.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 9 7 e3 8 0 9

Available at www.sciencedirect.com

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

Modelling sediment-microbial dynamics in the South Nation River, Ontario, Canada: Towards the prediction of aquatic and human health risk I.G. Droppo a,*, B.G. Krishnappan a, S.N. Liss b, C. Marvin a, J. Biberhofer a a b

Environment Canada, 867 Lakeshore Road, Burlington, Ontario, Canada Environmental Studies and Chemical Engineering, Queens University, Kingston, Ontario, Canada

article info

abstract

Article history:

Runoff from agricultural watersheds can carry a number of agricultural pollutants and

Received 14 December 2010

pathogens; often associated with the sediment fraction. Deposition of this sediment can

Received in revised form

impact water quality and the ecology of the river, and the re-suspension of such sediment

9 March 2011

can become sources of contamination for reaches downstream. In this paper a modelling

Accepted 18 April 2011

framework to predict sediment and associated microbial erosion, transport and deposition

Available online 23 April 2011

is proposed for the South Nation River, Ontario, Canada. The modelling framework is based on empirical relationships (deposition and re-suspension fluxes), derived from laboratory

Keywords:

experiments in a rotating circular flume using sediment collected from the river bed. The

Sediment deposition

bed shear stress governing the deposition and re-suspension processes in the stream was

Erosion

predicted using a one dimensional mobile boundary flow model called MOBED. Counts of

Transport

live bacteria associated with the suspended and bed sediments were used in conjunction

Modelling

with measured suspended sediment concentration at an upstream section to allow for the

Pollutants

estimation of sediment associated microbial erosion, transport and deposition within the

Bacteria

modelled river reach. Results suggest that the South Nation River is dominated by depo-

Pathogens

sition periods with erosion only occurring at flows above approximately 250 m3 s
1 (above this threshold, all sediment (suspended and eroded) with associated bacteria are transported through the modelled reach). As microbes are often associated with sediments, and can survive for extended periods of time, the river bed is shown to be a possible source of pathogenic organisms for erosion and transport downstream during large storm events. It is clear that, shear levels, bacteria concentrations and suspended sediment are interrelated requiring that these parameters be studied together in order to understand aquatic microbial dynamics. It is important that any management strategies and operational assessments for the protection of human and aquatic health incorporate the sediment compartments (suspended and bed sediment) and the energy dynamics within the system in order to better predict the concentration of indicator organism. Crown Copyright ª 2011 Published by Elsevier Ltd. All rights reserved.

* Corresponding author. E-mail address: [email protected] (I.G. Droppo). 0043-1354/$ e see front matter Crown Copyright ª 2011 Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.032

3798

1.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 9 7 e3 8 0 9

Introduction

Within river systems, it is clear that the microbial population (including pathogens) can be highly associated with the sediment fraction both in the suspended and bed sediment phase (Rehmann and Soupir, 2009; Krometis et al., 2007; Jamieson et al., 2004). This association serves four ecological and/or sedimentological functions; a) the sediment serves as a place for microbial attachment, b) the sediment serves as a food source (DOC, POC) for microbes, c) the sediment floc provides protection from environmental stresses (Cho et al., 2010; Gerba and McLeod, 1976) and d) bacteria and associated extracellular polymeric substances (EPS) (Leppard, 1997) promote sediment flocculation (the building of larger particles) by binding particles together to form a floc matrix, which in turn results in an increased downward flux of sediment (Droppo, 2004), and consequently associated bacteria (with possible pathogens) to the bed sediment (Droppo et al., 2010). Once on the bed, the sediment and microbes can undergo consolidation and biostabilization (biofilm development) (Droppo, 2009). As such, the bed sediment may represent a reservoir of potential pathogenic organisms with possible detrimental impacts in the event of erosion during storm flow (when the critical bed shear stress for erosion is surpassed) (Hirotani and Yoshino, 2010; Rehmann and Soupir, 2009; Wu et al., 2009). Many studies have demonstrated that bed sediment can contain orders of magnitude higher indicator organisms than the overlying water [although the suspended sediment (SS) associated microbes are not accounted for] (Cho et al., 2010; Droppo et al., 2009; Rehmann and Soupir, 2009; Crabill et al., 1999). Muirhead et al. (2004), using a series of artificial storms provided strong evidence that the bed sediment is an in-channel store of Escherichia coli (indicator organism) and that these stores are mobilized into the water column once erosion begins. In their experiments, three successive identical storm flows were generated on consecutive days with each yielding successively less E. coli (cumulative arial concentration of 1.2e1.3  108 cfu m
2). This suggests that the bed contains a limited source of E. coli and further substantiates a transient microbe population whose bed concentrations are dependent on flow conditions (Muirhead et al., 2004). Cho et al. (2010) observed that there were “hot spots” of E. coli concentrations in the bed and that sampling could miss these important areas of potentially pathogenic organisms that may be transported during storms. As indicator organisms and pathogens have been shown to survive for extended periods of time in aquatic sediments (up to months) (Davies et al., 1995; Jamieson et al., 2005), the erosion of sediment and associated pathogens represents significant risk to aquatic and human health, particularly if the fate of the sediment/pathogens is in sensitive areas such as drinking water intakes or swimming areas (Schutze et al., 1998; Donovan et al., 2008). Further, the association of pathogens with suspended sediment (SS) and the erosion of pathogens in association with bed sediments may result in operational sampling protocols for the assessment of human health risk being in error (Hirotani and Yoshino, 2010; Crabill et al., 1999), as such protocols assume the bacteria are planktonic in phase. In

addition, samples collected for health risk assessment may not represent recent pathogenic contamination but may reflect historically deposited microbes that have been resuspended during a high flow/shear event within a river (Ksoll et al., 2007). There has been limited modelling of microbial transport within rivers due to the difficulties of modelling a living, changing, interactive heterogeneous organic population (microbes) (e.g. Rehmann and Soupir, 2009). For example, different microbial species will have different affinities for attachment and flocculation (Characklis et al., 2005) due in part to different quantities and type of EPS generated (Wingender et al., 1999) as well as surface charge and hydrophobicity differences (Liao et al., 2001). In addition, few models have reflected the dynamic association of the microbes with the suspended and bed sediment compartments (Cho et al., 2010; Bai and Lung, 2005; Jamieson et al., 2005). Where modelling has been attempted, generally simplistic approaches have been used which represent bacteria re-suspension and deposition in association with suspended and bed sediment with a single set of parameters (i.e. counts/g of eroded or deposited sediment). Many models require a partition coefficient to estimate the ratio of attached to free-floating bacteria. This has been done with both an irreversible adsorption process (Jamieson et al., 2005) and a reversible linear adsorption process (Bai and Lung, 2005). However, verification of these coefficients is made difficult as there are no standard methods for the differentiation of microbe counts associated with the two sediment compartments (see Droppo and Ongley, 1994; Characklis et al., 2005; Krometis et al., 2007, 2009 for non standardized methods of population separation), and an operationally defined limit of what constitutes particulate material. Cho et al. (2010) has incorporated additional reach specific parameters such as particle size distribution and critical shear stress to better assess sediment-microbe behaviour. Using these parameters, they were able to better predict the re-suspension of E. coli than single parameter models, although the single parameter models were better able to capture the dynamics of bacteria transport. Not accounting for the bed derived source of indicator organisms can make temporal variations in concentrations with flow difficult to model (Hellweger and Masopust, 2008). It is clear that improved model predictions of indicator organism source, transport and fate could be achieved if microbial transport models included sediment dynamics within them (Dorner et al., 2006). In this study we link an experimental evaluation of sediment erosion and deposition fluxes with measured sediment associated live bacteria concentrations to provide an estimate of the potential for sediment to control the erosion, transport and fate of microbes and possibly pathogens in the South Nations River, Ontario, Canada. A one dimensional flow model called MOBED was used to calculate the controlling bed shear stress distribution in the river. The specific objectives of this paper are 1) to improve the prediction of pathogen mobility and fate by pairing field microbial organism counts with a mathematical model of sediment erosion, transport and fate and 2) facilitate a discussion on the need to incorporate the energy dynamics, sediment compartments and sediment

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 9 7 e3 8 0 9

associated pathogens into the management of aquatic and human health risk.

2.

Methods

2.1.

Field program

2.1.1.

Study site

An 8.5 km reach of the main stem of the South Nation River near Ottawa, Ontario, Canada was investigated from St. Albert to Cassleman (Fig. 1). The River drains 3700 km2 with

3799

a very low gradient (0.5 m km
1) over primarily glacial marine clay with some sandy till. The clays (Leda clays) are very unstable and prone to slope failure. The river is deep with a maximum depth measure of 8.5 m and is slow moving (<10 cm s
1 on average). Fig. 1 provides a bathymetric survey of the river (Knudsen 320M echo sounder with a 200 kHz transducer; Knudsen Engineering Limited). The river transports primarily fine-grained cohesive sediment with a d50 of 10 mm and has a bed composition ranging from gravel to cohesive glacial marine clay. Further details on the South Nation River can be found in Chapman and Putnam (1984).

Fig. 1 e Bathymetric survey map of 8.5 km modelled reach of the South Nation River from St. Albert to Casselman, Ontario, Canada (flow from bottom to top of aerial photos). Note the deep sections of the river up to 8.5 m, in depth. Large arrows denote same location on two aerial photos.

3800

2.1.2.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 9 7 e3 8 0 9

SS sampling for microbial assessment

Bulk SS was collected monthly from May to November (3 samples in May) in 2006 at St. Albert using a continuous flow centrifuge (Westfalia Model KA 2-06-075) for microbial assessment and for input of SS-associated microbes into the numerical model. Samples were collected under a variety of flow and SS concentration conditions and, as such, average values are believed representative of the average open water concentrations. In this procedure bulk river water was pumped into a stainless steel bowl with a rotational speed of 9470 rpm and the denser sediment is removed from the water. The duration of pumping is dependent on the concentration of SS within the water. The majority of sediment (>90% recovery) is retained in the bowl which was then removed and placed in sterile polyethylene containers and placed on ice until analysis was performed.

2.1.3.

Bed sediment sampling for microbial assessment

Bed sediment was collected on the same dates as the bulk SS samples at St. Albert for bulk grain size analysis (SedigraphTM) and surficial microbial assessment (model input) with a PONAR bed sediment sampler. Only the top few mm of sediment was scraped off the surface of the grab and placed in autoclaved polyethylene containers. All samples were kept on ice until analysis was performed.

2.1.4.

Enumeration of live/lead bacteria

The microbial population (bulk average cell counts) of bed and bulk SS samples were assessed using the LIVE/DEAD BacLight nucleic acid staining technique (Boulos et al., 1999). Depending on the solids concentration, 0.5 mLe5 mL of raw sample was filtered through a 0.2 mm black polycarbonate filter. One mL of 0.085% NaCl solution was placed on the filtered sample and 30 mL of a 10% dilution of LIVE/DEAD BacLight staining solution (in 0.085% NaCl) was added to the mixture. The apparatus was incubated for 15 min in the dark and the dye mixture was filtered. Filters were mounted on a slide with BacLight mounting oil and bacteria counted (Red ¼ dead bacteria; green ¼ live bacteria) using a Zeiss Axiovert 100 fluorescent microscope equipped with a 520 nm barrier and 470 nme490 nm excitation filter.

2.2.

Transport characteristics of fine sediment

Deposition and erosion characteristics of fine sediments of the South Nations River were studied experimentally using a laboratory rotating circular flume. The details of the experimental program are outlined below:

to stabilize it on the stream bed against the flow. The sampler was deployed at several locations within the river and the deposited sediment was collected together with the river water. About 600 L of sediment-water mixture was collected and transported to the laboratory in a refrigerated transport truck.

2.2.2.

Rotating circular flume

In order to generate model parameters of erosion rate and critical bed shear stress for erosion and deposition, a rotating circular flume measuring 5.0 m in mean diameter, 0.30 m wide and 0.30 m deep was employed (Fig. 2) (Krishnappan, 1993). A counter rotating top cover (ring) fits inside the flume with close tolerance, (w1.5 mm gap on either side) and makes contact with the water surface in the flume. The maximum rotational speeds of the flume and the ring are 3 rpm respectively. The flows generated in the flume are nearly two dimensional and the bed shear stress distribution across the width of the flume is fairly uniform (Krishnappan and Engel, 2004). The flume calibration results of Krishnappan and Engel (2004) were used to predict the relationship between the bed shear stress and the rotational speeds of the flume assembly.

2.2.3.

Deposition experiments

River water was placed in the flume with a known amount of sediment (approximately 350 mg L
1). The sediment-water suspension was then thoroughly mixed in the flume first by mechanical mixing and then by rotating the flume and the lid at relatively high speeds (2.5 rpm for the lid and 2.0 rpm for the flume, which yielded a bed shear stress of 0.6 Pa) for 20 min. Following this mixing period, the flume speed was turned down to a lower rate of shear and maintained for the duration of the experimental run. Shear values used were 0.059, 0.126 and 0.217 Pa. Sediment-water samples were withdrawn from the flume at 5 min intervals during the first hour of the test and every 10 min thereafter until the completion of the test. Each time a sample was withdrawn, the volume removed was replaced by adding an equivalent amount of sediment-water mixture back into the flume, in order to keep the water surface in contact with the lid all the time. A test was considered to be completed after the SS concentration remained nearly constant for about 1 h (generally requiring 5 h for most runs). Sediment-water samples were analysed for concentrations of solids by a gravimetric method which consisted of filtering the sample (0.45 mm pre-weighed Millipore filter), and drying and weighing the residue. Once the test had been completed, the whole procedure was repeated for other flume speeds.

2.2.1. Bed sediment sampling for laboratory flume experimentation

2.2.4.

Sediment-water mixture from the river was collected at St. Albert using a specially designed inverted cone sediment sampler (Krishnappan, 2007). The sampler consists of an inverted cone fitted with a propeller to create sufficiently strong circulation inside the chamber to dislodge the deposited fine sediment, and a submerged pump that delivers the dislodged sediment and the water mixture to a sample container. The sampler is also fitted with a weight

Additional South Nation River bed sediment was added to the flume which was then mixed at a high rate of speed to thoroughly mix the sediment and water. The flume speed was then gradually reduced to a stop to allow the mixture to settle and consolidate/biostabilize. This formed a bed of approximately 2 cm depth. The flume and the lid were then set in motion and their speeds were incrementally increased in 50 min time intervals. At each interval, sediment samples

Erosion experiment

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 9 7 e3 8 0 9

3801

Fig. 2 e Schematic representation of the rotating circular flume.

were collected every 10 min to measure the SS concentration variation as a function of time (as above, sample volumes removed were immediately replaced). When the sediment concentration reached a steady state value, the flume speed was increased to the next level (for our experiment this occurred until the shear reached just over 0.2 Pa). This procedure was carried out over three different consolidation/ biostabilization periods of 22.5, 45 and 92 h.

3.

Results and discussion

flow conditions in the river are such that the bed shear stress in the river is less than the critical shear stress for deposition, then all of the sediment and associated bacteria entering the river would deposit to the river bed. On the other hand, if the bed shear stress reaches 0.217 Pa, then about 83% of the suspended material and associated bacteria would be transported in suspension. Knowing the flow field and the spatial and temporal variation of the bed shear stress, along with the SSassociated bacterial counts, the results from the deposition tests can be used to make quantitative estimates of sediment and bacteria deposition to the river bed.

3.1.

Deposition characteristics

3.2.

Fig. 3 illustrates the SS concentration in the water column as a function of time for three different bed shear stresses during deposition. For all runs the SS concentration decreases rapidly followed by a steady state (equilibrium) concentration for each bed shear stress. For example, for the lowest bed shear stress tested (0.059 Pa), the steady state concentration was about 50 mg L
1 (17% of the initial concentration), whereas for the highest shear stress (0.217 Pa), the steady state concentration was about 250 mg L
1 (83% of the initial concentration). If the bed shear stress was slightly lower than 0.059 Pa, then all the initially SS would have deposited. Such a bed shear stress was defined as the critical shear stress for sediment deposition, which for the South Nation River sediment was considered to be 0.050 Pa. The deposition tests also offer a quantitative measure of the amount of sediment that is likely to be transported through the river for given flow conditions. For example, if the

Erosion characteristics

Fig. 4 illustrates the SS concentration profiles for three different consolidation/biostabilization periods with increasing shear stress. For the more stabilized beds (45 and 92 h consolidation/biostabilization), the sediments are relatively stable at a shear stress of 0.09 Pa with erosion not occurring until the shear reached 0.123 Pa. In this work we use the conservative number of 0.09 Pa to represent the critical shear stress for erosion of the surface sediment layer. Contrary to previous results (Droppo, 2009) the least consolidation/biostabilization sediment (22.5 h) did not show erosion occurring until an increment higher (0.165 Pa). This anomaly is likely related to the higher initial SS concentration in suspension at the start of the test, masking erosion. It should be noted however that, as expected, later in the erosion sequence the weakest bed (22.5 h consolidated/biostabilized) yielded the highest SS concentration, while the oldest, most stabilized sediment bed resulted in the lowest SS concentration. At each

3802

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 9 7 e3 8 0 9

Fig. 3 e Results from deposition experiments.

step in shear stress, SS concentration increased gradually and showed a tendency to approach a steady state value. As such, increasing shear stresses were required to erode deeper material and to continue to increase the SS concentration. Such a trend is suggestive of increasing bed strength with depth and is likely related to a combination of self-weighted consolidation and biostabilization (Droppo, 2009). At the maximum shear stress of 0.21 Pa, not all of the deposited sediment was re-suspended. The maximum concentration reached was only 60 percent of the total concentration that would have resulted from complete re-suspension. The maximum concentration that was attained at the shear stress of 0.21 Pa was about the same as the steady state concentration observed in a deposition test at a much lower shear stress of 0.126 Pa. It is important to note that the critical shear stress for erosion is larger than the critical shear stress for deposition, which is a distinguishing characteristic of cohesive sediment (Lick, 1982; Parchure and Mehta, 1985). The

difference in critical shear stresses has implications for the transport behaviour of cohesive sediments.

3.3.

Derivation of sediment transport functions

The results of the present experimental investigation show that the sediments from the South Nation River behave like cohesive sediments and hence the modelling of sediment transport in the river has to take into account the cohesive sediment transport characteristics. The difference between the cohesive and cohesionless sediment transports results from the fact that the critical shear stresses for erosion and deposition are equal for cohesionless sediments, and hence such sediments undergo simultaneous erosion and deposition when subjected to a constant bed shear stress. On the other hand, for cohesive sediments, these two stresses are not equal and therefore, the sediment flocs do not undergo deposition and erosion simultaneously. A deposited sediment floc will

Fig. 4 e Results from erosion experiments.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 9 7 e3 8 0 9

remain on the bed until it is exposed to a higher shear stress that exceeds the critical shear stress for that sediment floc (Lau and Krishnappan, 1994). A mathematical model reflecting this distinction was developed by Krishnappan (2000) for stream flows and is applied in this study. In the quasi-steady state model proposed by Krishnappan (2000), deposition and erosion functions were determined from the deposition and erosion experiments, respectively, and these functions were used to quantify the vertical exchange of sediment at the sediment/water interface for different bed shear stress conditions. The deposition function (Fig. 5), derived from the results of the deposition experiments as well as from previous experiences with other cohesive sediments tested in the flume (Stone and Krishnappan, 1997; Skafel and Krishnappan, 1999 and Milburn and Krishnappan, 2003), gives the fraction of the deposited sediment as a function of the bed shear stress. The bed shear stress is normalized using the critical shear stress for deposition. This function satisfies the condition that when the bed shear stress is less than the critical shear stress for deposition, then all of the initial SS is deposited, i.e. the fraction deposited equals one. When the normalized bed shear stress is greater than 5.5 (corresponding to a shear stress value of 0.275 Pa), then none of the initial SS deposit, i.e. the fraction deposited is equal to zero. When the bed shear stress is within these two limits, then the fraction of sediment deposited is given by the power function below: 0:543 for f1 < s0 =scd < 5:5g fd ¼ 1:0
0:43ðs0 =scd
1Þ fd ¼ 1:0 for fs0 =scd  1g fd ¼ 0 for fs0 =scd  5:5g

(1)

where fd is the fraction deposited, s0 is the bed shear stress and scd is the critical shear stress for deposition. The erosion function (Fig. 5) reflects the fact that the critical shear stress for erosion is 1.8 times the critical shear stress for deposition and the shear stress that is needed to erode 100% of the deposited sediment is 14 times the critical shear stress for deposition (which corresponds to a shear stress value of 0.70 Pa and obtained by extrapolating the erosion function to a value of 1). Therefore, for the bed shear stress lower than 1.8 times the critical shear stress for deposition, none of the deposited sediment will be eroded (i.e. the fraction of sediment eroded is equal to zero). For shear stresses greater

3803

than 14 times the critical shear stress for deposition, all of the deposited sediment will be eroded (i.e. the fraction of the sediment eroded equals one). The analytical form of the erosion function is as follows: fe ¼ 0:3ðs0 =scd
1:8Þ0:523 for f1:8 < s0 =scd < 14g fe ¼ 0 for fs0 =scd  1:8g fe ¼ 1 for fs0 =scd  14g

(2)

where fe is the erosion function. The above two power law type relationships describing deposition and erosion processes of South Nation River sediment were similar to the ones used to describe the transport of several other cohesive sediments tested in the rotating circular (Krishnappan, 1996; Stone and Krishnappan, 1997; Skafel and Krishnappan, 1999; Milburn and Krishnappan, 2003). Even though the form of the power law remains the same for all these sediments, the coefficients defining the power law, the ratio between the critical shear stress for erosion and deposition and the actual value of the critical shear stress for deposition differ from sediment to sediment. These properties have to be determined empirically using site specific sediments in flumes such as the one used in the present study. Since the bed shear stress is normalized using the critical shear stress for deposition, care should be taken to determine the latter accurately, as any error in its determination will propagate through the calculation of the sediment and bacteria fluxes at the sediment-water interface. In the present measurements, the accuracy of sediment dynamic parameters is in the range of 10e15%, which is well within the accuracy of measurements related to bacteria. Ideally, future models will need to incorporate microbial partitioning, inactivation and growth in order to have dynamic models capable of simulating the complex interaction between microbes, sediment and hydrodynamics. Work by Gao et al. (2011) is progressing towards this end. Using these two functions, sediment and the associated live bacteria transported through the river can be calculated by dividing the river into a number of segments and calculating the sediment and bacteria mass balance through these segments knowing the flow field calculated using the MOBED model. Details of these calculations are given below.

Fig. 5 e Deposition and Erosion functions for the South Nation River sediment.

3804

3.4.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 9 7 e3 8 0 9

Sediment mass balance

The river reach is divided into a number of segments shown schematically in Fig. 6. The flow rate (Q (m3 s
1)) in each segment is considered to be steady. For a varying flow, a quasi-steady state is assumed and the flow hydrograph is approximated by a step function. The average bed shear stress (s0) has to be determined for each segment as a function of the flow rate. This will be done using a mobile boundary flow model, MOBED. For each segment, the range of the flow rates over an average year is established and the corresponding boundaryshear stress range is calculated. This shear stress range is then divided into a fixed number of flow stages. Each flow stage is identified with its own average boundary-shear stress which is the mean value of the shear stresses bounding the range. To be consistent with the sediment transport functions given by Equations (1) And (2), the boundary-shear stress is normalized using the critical shear stress for deposition.

3.5.

Mass balance of sediment in a control segment

3.5.1.

SS inflow

The mass balance of the sediment in a control segment is calculated as follows: SS entering the control segment can be 1) from the upstream segment and 2) from tributary inflows (although for this simulation we assumed no tributaries). The amount of SS (qsu) entering the control segment during a time interval of Dt can be expressed as follows: qsu ¼ QCi
1 Dt þ Qt Ct Dt

(3)

where Ci
1 is the SS concentration in the upstream segment, Ct is the concentration of SS in the tributary inflow to the control segment and Qt is the tributary inflow rate.

3.5.2.

SS depositing to the bed

A portion of the incoming SS will deposit depending on the prevailing flow conditions. The SS quantity transported from the upstream segment will be deposited only when the shear stress in the control segment is lower than that in the upstream segment. If the shear stress in the control segment is equal to or greater than the upstream segment, then the SS arriving from the upstream segment would have gone through

the deposition process already and would have reached the steady state concentration. Therefore, this SS has to be routed straight through the control segment. The amount that would deposit in the control segment, therefore, can be calculated as: qsd ¼ ðQCi
1 DtÞfd þ ðQt Ct Dt Þfd

if sI  sI
1


qsd ¼ ðQt Ct DtÞfd þ ðQCI
1 DtÞ if sI sI
1

(4) (5)

where qsd is the amount deposited during the current time step. The amount remaining in suspension (qss) becomes:   if sI  sI
1 qss ¼ ðQCi
1 Dt þ Qt Ct DtÞ 1
fd

(6)

  qss ¼ Qt Ct Dt 1
fd

(7)

3.5.3.


if sI sI
1

Sediment re-suspension

Sediment can also be re-suspended from deposits that occurred in previous time steps. The mass of re-suspended sediment can be calculated by keeping track of the amount of deposited sediment and the shear stresses at which deposition takes place. This can be done by schematizing the river bed to consist of different compartments and assuming that each compartment holds sediment deposited at a particular shear stress. For example, let us assume that there are N compartments in the control segment and each compartment is identified with an index, say, J. Therefore, J varies from 1 to N. Compartment 1 is assumed to collect sediment deposited at shear stress equal to or less than the critical shear stress for deposition, scd, and compartment 2 collects sediment deposited at shear stresses between scd and 2scd and so on. Let the sediment deposited in each of the compartments from the previous time steps be PJ. For a given shear stress, the sediment re-suspended from various compartments can be calculated by applying Equation (2) for each compartment with the appropriate scd value as follows: qsr ¼

N X J¼1

qsrJ ¼

N X

PJ feJ

(8)

J¼1

where qsr is the total amount of re-suspended sediment and feJ is the erosion function for the compartment J. Knowing the amount of sediment re-suspended from and deposited to the bed, the concentration of the SS and the

Fig. 6 e Schematic representation of sediment mass balance.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 9 7 e3 8 0 9

3805

amount of sediment in the various compartments in the bed of the control segment at the end of the current time step can be calculated as follows:

sediment phases Np (counts L
1 of water) can be derived as follows:

3.5.4.

The above equation can be applied for all segments of the river and for all time steps to compute the transport of bacteria associated with sediment.

SS concentration at the end of the current time step

The SS concentration in the control segment at the end of the current time step is:  CðiÞ ¼

qss þ qsr ðQDtÞ

(9)

The amount of sediment left behind in the control segment at the end of the current time step is: PJ ¼

J¼1

N   X PJ 1
feJ þ dðJ; KÞ  qsd

(10)

J¼1

where PJ* is the updated value of PJ at the end of the current time step. The function d(J,K ) takes a value of unity, when J ¼ K, and zero when JsK. K denotes the compartment which receives the deposited sediment during the current time step. The concentration C(i), as calculated by Equation (9), is routed to the downstream segment and the calculations outlined above for the control segment are repeated for the downstream segment. The process is continued until all the segments are accounted for. The calculations are then repeated for the next time step until the simulation period is covered.

3.6.

(11)



3.5.5. Amount of sediment in various bed compartments at the end of the current time step

N X

  Np ¼ qss Nps þ qsr Npb ðQDtÞ

Sediment associated bacteria mass balance

Knowing average live bacterial concentrations for SS floc and in the bed sediment, an equation for the bacteria concentration associated with sediment in the water column can be derived from the SS concentration equation, i.e. Equation (9). Denoting the bacteria concentration in SS as Nps (counts mg
1 of SS) and the bacteria concentration in bed sediment as Npb (counts mg
1 of bed sediment), the equation for the bacteria concentration contributed to the water column via the

3.7.

Flow computation using MOBED Model

Flow characteristics in the receiving stream were modelled using the MOBED model developed by Krishnappan (1980). MOBED is an unsteady and mobile boundary flow model based on a generalized friction factor relationship and hence is suitable to predict the flow characteristics. Input data to the model include cross sectional geometry of each transect, initial bed and water surface elevations, boundary conditions at the upstream and downstream sections of the study reach and mobile bed roughness parameters. Measured cross sectional shapes at 18 sites were used in the model. The initial bed level along the length of the stream reach was determined from the thalweg at each cross section and the values between the measured sections were linearly interpolated. The initial water surface elevation was obtained by running the model as a steady state model. A 2005 spring melt flow hydrograph from the Cassleman gauging station (02LB013) was used in the modelling exercise to represent the flow entering the upstream boundary at St. Albert (Fig. 7). The model was run for this flow hydrograph and the bed shear stress as a function of distance along the stream and time were calculated. The shear stress variation as a function of time for the peak flow of 600 m3 s
1 and at a lowest model flow of 100 m3 s
1 is shown in Fig. 8 as an example. A full routing of sediment and associated bacteria can be made using the complete solution of the MOBED model. The initial model input variable of SS concentration (Cinitial) for the first segment at St. Albert was calculated by applying the discharge date from Fig. 7 to a rating curve (Equation 12) determined from June SS concentration and discharge

Fig. 7 e Flow hydrograph used at the upstream boundary.

3806

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 9 7 e3 8 0 9

Fig. 8 e Dashed lines are predicted shear stress variation over the modelled reach for the peak flow (600 m3 sL1) and for the lowest model flow (100 m3 sL1). Solid symbols are predicted change in Np (contribution of live bacteria concentrations associated with sediment per L of water) for the first day of simulation (100 m3 sL1) and during the falling limb on day 11 (200 m3 sL1).

measurements from historical SS data (1971e2001 e Plantagenet sediment gauging station 02LB005) (only April data was used for the model as this is the month of our data sampling and the time for which high flows occur in the river e Fig. 7).

the river bed; thus facilitating the bed as a reservoir of pathogens for potential downstream transport with a significant storm or spring melt flood event.

(12)

3.8.1. Implications of sediment associated bacteria to aquatic and human health

Model outputs using Equation 11 are illustrated in Fig. 8 for day 1 and 11 of the simulation over the 8.5 km modelled reach. It was observed that initially when modelled flows are low (100 m3 s
1), the majority of bacteria associated with the SS are delivered to the bed with counts at the 8.5 km transect depleting to 1.3  105 counts L
1 from an initial calculated input value of 6.8  107 at St. Albert. At this point the bed shear stress is very close to the critical bed shear stress for deposition and sediment and associated bacteria are being deposited under the low flow conditions. At flows above approximately 250 m3 s
1, there is no deposition and only erosion takes place routing all sediment associated bacteria through the modelled reach as bed shear levels are 5.5 times greater than the critical bed shear stress for deposition (0.05 Pa) (See Fig. 5). Concentrations increased from 2.3  107 counts L
1 at a flow of 260 m3 s
1 to 4.0  108 counts L
1 at 600 m3 s
1. Only after the falling limb declined to below 250 m3 s
1 did bacteria associated with the SS begin to deposit again. For example, after 11 days, (discharge ¼ 200 m3 s
1), bacteria is still being transported through the modelled reach (Fig. 8), however, deposition is occurring as seen by the decrease in concentration with distance down river. This reduction in concentration is related to the shear level variations allowing partial settling with the ratio of bed shear stress to critical bed shear stress for deposition fluctuating intermediately between 1 and 5.5. Given that the mean annual flow at Casselman (02LB013) is only 29.3 m3 s
1 (flows above 250 m3 s
1 have only been recorded for spring melt periods), it is clear that for the majority of the year the river is depositing microbes associated with the SS to

The model, while simplistic in that tributary inputs and historic sediment deposits are not accounted for (i.e. only the sediment entering the system for transport, deposition and erosion is considered in this model), provides evidence that river sediment dynamics can play a very strong role in controlling the microbial and possibly pathogen dynamics within river systems. The model shows that river bacterial dynamics can be highly influenced by the erosion, transport and deposition of sediment. Bacteria and pathogens have been shown to be orders of magnitude higher in association with sediments (both bed and suspended) with the bed sediment representing a significant reservoir of possible pathogenic organisms (Hirotani and Yoshino, 2010; Droppo et al., 2009; Muirhead et al., 2004; Obiri-Danso and Jones, 2000). Pathogens are known to live for up to several months within bed sediment (Davies et al., 1995; Obiri-Danso and Jones, 2000; Jamieson et al., 2005) and, as such, the determination of indicator organism counts for the assessment of aquatic and human health risk, may not only represent contemporary contamination but also historically deposited microbes (Ksoll et al., 2007). Currently, source area tracking of pathogenic organisms investigate mostly terrestrial sources without due consideration of the river bed as a possible source of pathogens during erosion events (Droppo et al., 2009). Further, a current lack of understanding around the dynamic processes involved with sediment/microbial interactions has lead to current indicator organism monitoring by health departments (for the assessment of human health risk) to assume that bacteria are planktonic in nature. This ignores the fact that a large proportion of the bacteria in suspension can be associated with the SS (Droppo et al., 2010; Crabill et al., 1999). The above assumption is somewhat necessitated by the lack of a standardized test for determining the contribution of

Cinitial ¼ 26:01  Q 0:25

3.8. Sediment associated bacteria transport in South Nation River

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 9 7 e3 8 0 9

3807

Fig. 9 e A schematic diagram demonstrating the relationship between pathogen and sediment dynamics within river systems (modified from Droppo et al., 2010). Erosion, deposition and suspended floc-pathogen dynamics are portrayed in relation to the ambient floc and pathogens populations held in suspension.

microbes from sediment compartments. It is important that such a test be made available and be used across health departments in the future. In addition, typical sampling designs do not take into account the linkage of the energy (shear) regime (e.g. currents, waves) with sediment/pathogen dynamics (i.e. flow dependent concentrations of bacteria), but report counts per volume at a point in time. From the modelling exercise and the above discussion, it is apparent that the dynamics of microbes (including pathogens) within river systems are influenced by 3 domains; 1) flow regimes, 2) sediment properties and 3) microbial community characteristics and behaviour. The interactions of these domains are demonstrated in Fig. 9 (modified from Droppo et al., 2010). This model provides a simplified conceptualization of the numerical model above and is centred on three decision boxes operating within the ambient floc-pathogen population held in suspension within the river. These boxes are responsible for routing pathogens between the bed sediment, SS (floc) and the planktonic phase as influenced by the energy regimes at the sediment-water interface and within the water column. The directional decision boxes of the model are related to: (1) bed shear stress relative to critical bed shear stress which dictates if pathogens will be eroded from the bed in planktonic or floc attached modes, or further consolidated and multiplied in the bed sediment/biofilm; (2) fluid shear relative to floc shear strength which dictates if pathogens will be dissociated from the floc or eventually settle towards the bed in association with the floc; and (3) bed shear stress relative to floc shear strength which dictates if the floc will remain intact and deliver pathogens to the bed, or if it will break up resulting in pathogen dissociation and longer range

transport. The transitory nature of pathogens (i.e. associated with the bed, floc, or planktonic under different energy conditions) will have implications for the potential risk to aquatic and human health. Fig. 9 and the results of the modelling confirm the need to understand the energy regime in relation to both the sediment and associated microbes (pathogens) within the water column. Monitoring programmes intended to assess risk to human health which currently neglect the sediment phase of pathogen existence, and the energy regime they preside in, may provide management decisions which are erroneous with possible detrimental impacts. As such, it is critical that management protocols and policy development for the protection of aquatic and human health include a consideration of both the SS and bed sediment indicator organism population.

4.

Conclusions

Experiments carried out in a rotating circular flume with sediment from the South Nation River indicate that the river sediment exhibits cohesive behaviour. As such, a cohesive sediment transport model in conjunction with a mobile boundary flow model was used to predict SS and associated live bacteria transported within the South Nation River. The modelling provided a first order estimate of the influential impact of suspended and bed sediment on the dynamics of pathogens in a river system. For a given flow condition, the proposed modelling framework predicted the bed shear stress variation along the river reach and thereby facilitated the computation of erosion,

3808

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 9 7 e3 8 0 9

transport and deposition of SS and sediment associated bacteria concentrations and rates along the river reach. Over the modelled hydrograph and 8.5 km reach, live bacteria were found to be eroded from the bed and transported through the system at flows above approximately 250 m3 s
1. For flows below this level (the norm for the South Nation River), there was significant deposition suggesting that the bed of the South Nation River is likely a substantive source of pathogens for subsequent erosion and transport down river to potentially environmentally sensitive areas (e.g. recreational areas and drinking water intakes). While the work provided in this paper is specific to the geometry, flow conditions, sediment dynamics and measured microbial counts of the South Nation River, the modelling framework, proved to be a useful tool to assess the impact of sediment and bacteria transport, erosion and deposition within the South Nation River. The algorithms used in the model are transferable to other rivers, however, the coefficients defining the power law, the ratio between the critical shear stress for erosion and deposition and the actual value of the critical shear stress for deposition differ from sediment to sediment. As such, these variables would need to be determined empirically for each system studied. It is clear that, shear levels, bacteria concentrations and SS have a symbiotic relationship within riverine systems. One can not be studied without accounting for the influence of the others. In this regard, it is important that any management strategies and operational assessments for the protection of human and aquatic health incorporate the sediment compartments (SS and bed sediment) and the energy dynamics within the system in order to better predict the concentration of indicator organism.

Acknowledgements The authors wish to acknowledge M. Hewitt for discussions on the topic and B. Trapp, C. Jaskot, M. Dunnett, T. Nelson and R. Stephens of Environment Canada in Burlington, Ontario, Canada for their technical support. The study was partially supported by the NAESI programme.

references

Bai, S., Lung, W.-S., 2005. Modeling sediment impact on the transport of fecal bacteria. Wat. Res. 39, 5232e5240. Boulos, L., Prevost, M., Barbeau, B., Coallier, J., Desjardins, R., 1999. LIVE/DEAD BacLightTM: application of a new rapid staining method for direct enumeration of viable and total bacteria in drinking water. J. Microb. Meth. 37, 77e86. Chapman, L.J., Putnam, D.F., 1984. The Physiography of Southern Ontario. Government of Ontario, Ontario Ministry of Natural Resources, Toronto, Canada, ISBN 0-7743-9422-6. Characklis, G.W., Dilts, J.M.J., Simmons, O.D., Likirdopulos, C.A., Krometis, L.-A.H., Sobsey, M.D., 2005. Mirobial partitioning to settleable particles in stormwater. Wat. Res. 39, 1773e1782. Cho, K.H., Pachepsky, Y.A., Kim, J.H., Guber, A.K., Shelton, D.R., Rowland, R., 2010. Release of Escherichia coli from the bottom sediment in a first-order creek: experiment and reach-specific modeling. J. Hydrol. 391, 322e332.

Crabill, C., Donald, R., Snelling, J., Foust, R., Southam, G., 1999. The impact of sediment fecal coliform reservoirs on seasonal water quality in Oak Creek, Arizona. Wat. Res. 33, 2163e2171. Davies, C.M., Long, J.A., Donald, M., Ashbolt, N.J., 1995. Survival of fecal microorganisms in marine and freshwater sediments. Appl. Environ. Microbiol. 61, 1888e1896. Donovan, E., Unice, K., Roberts, J.D., Harris, M., Finley, B., 2008. Risk of gastrointestinal disease associated with exposure to pathogens in the water of the lower Passaic River. App. Env. Microb. 74 (4), 994e1003. Dorner, S.M., Anderson, W.B., Slawson, R.M., Kouwen, N., Huck, P. M., 2006. Hydrologic modeling of pathogen fate and transport. Env. Sci. Technol. 40, 4746e4753. Droppo, I.G., 2009. Biofilm structure and bed stability of five contrasting freshwater sediments. Mar. Freshwat. Res. 60, 690e699. Droppo, I.G., 2004. Structural controls on floc strength and transport. Can. J. Civil Eng. 31, 569e578. Droppo, I.G., Ongley, E.D., 1994. Flocculation of suspended sediment in rivers of southeastern Canada. Wat. Res. 28, 1799e1809. Droppo, I.G., King, K., Tirado, S., Sousa, A., Wolfaardt, G., Liss, S.N. , Warren, L., 2010. Assessing riverine sediment-pathogen dynamics: implications for the management of aquatic and human health risk. In: Banasik, K., Horowitz, A.J., Owens, P.N., Stone, M., Walling, D.E. (Eds.), Sediment Dynamics for a Changing Future. IAHS Pub, pp. 245e250. no. 337. Droppo, I.G., Liss, S.N., Williams, D., Nelson, T., Jaskot, J., Trapp, B., 2009. The dynamic existence of waterborne pathogens within river sediment compartments e Implications for water quality regulatory affairs. Env. Sci. Technol. 43 (6), 1737e1743. Gao, G., Falconer, R.A., Lin, B., 2011. Numerical modelling of sediment-bacteria interaction processes in surface waters. Wat. Res. 45, 1951e1960. Gerba, C.P., McLeod, J.S., 1976. Effect of sediments on the survival of Escherichia coli in marine waters. Appl. Environ. Microbiol. 32, 114e120. Hellweger, F.L., Masopust, P., 2008. Investigating the fate and transport of Escherichia coli in the Charles River, Boston, using high-resolution observation and modeling. J. Amer. Wat. Resour. Ass. 44, 509e522. Hirotani, H., Yoshino, M., 2010. Mirobial indicators in natural biofilms developed in the riverbed. Wat. Sci. Technol. 62, 1149e1153. Jamieson, R.C., Joy, D.M., Lee, H., Kostaschuk, R., Gordon, R.J., 2005. Resuspension of sediment-associated Escherichia coli in a natural stream. J. Env. Qual. 34, 581e589. Jamieson, R.C., Lee, J.H., Kostaschuk, R., Gordon, R.J., 2004. Persistence of enteric bacteria in alluvial streams. J. Env. Eng. Sci. 3, 203e212. Ksoll, W.B., Ishii, S., Sadowsky, M.J., Hicks, R.E., 2007. Presence and sources of fecal coliform bacteria in epilithic periphyton communities of Lake Superior. Appl. Environ. Microb. 73, 3771e3778. Krishnappan, B.G., 2007. Recent advances in basic and applied research in cohesive sediment transport in aquatic systems. Can. J. Civ. Eng. 34, 731e743. Krishnappan, B.G., 2000. Modelling cohesive sediment transport in rivers. In: M.Stone (Ed.), The Role of Erosion and Sediment Transport in Nutrient and Contaminant Transfer (Proceedings of a Symposium Held in Waterloo, Canada, July 2000). IAHS Publication no. 263 (Red Book), IAHS Press, Centre for ecology and Hydrology, Wallingford, UK., pp. 269e276. Krishnappan, B.G., 1996. A New Algorithm for Fine Sediment Transport. National Water Research Institute, Environment Canada, Burlington, Ontario, Canada. Contribution No. 96e160. Krishnappan, B.G., 1993. Rotating circular flume. J. Hydraul. Eng. ASCE 119 (6), 758e767.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 7 9 7 e3 8 0 9

Krishnappan, B.G., 1980. Users Manual: Unsteady, Nonuniform Mobile Boundary Flow Model-MOBED. NWRI Contribution No 80-999. National Water Research Institute, Burlington, Ontario, Canada. Krishnappan, B.G., Engel, P., 2004. Distribution of bed shear stress in rotating circular flume. J. Hydraul. Eng. ASCE 130 (4), 324e331. 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. Wat. Res. 41 (2), 506e516. Krometis, L.A.H., Dillaha, T.A., Love, N.G., Mostaghimi, S., 2009. Evaluation of a filtration/dispersion method for Enumeration of particle-associated Escherichia coli. J. Environ. Qual. 38, 980e986. Lau, Y.L., Krishnappan, B.G., 1994. Does reentrainment occur during cohesive sediment settling? J. Hydraul. Eng. ASCE 120 (2), 236e244. Leppard, G.G., 1997. Colloidal organic fibrils of acid polysaccharides in surface waters: electron-optical characteristics, activities and chemical estimates of abundance. Colloids Surf. 120, 1e15. Liao, B.Q., Allen, D.G., Droppo, I.G., Leppard, G.G., Liss, S.N., 2001. Surface properties of activated sludge and their role in bioflocculation and settling. Water Res. 35, 339e350. Lick, W., 1982. Entrainment, deposition and transport of fine grained sediment in lakes. Hydrobiol 91, 31e40. Milburn, D., Krishnappan, B.G., 2003. Modelling erosion and deposition of cohesive sediments from Hay River in Northwest Territories, Canada. Nordic Hydrology vol. 34 (No.1/2), 125e138.

3809

Muirhead, R.W., Davies-Colley, R.J., Donnison, A.M., Nagels, J.W., 2004. Faecal bacteria yields in artificial flood events: quantifying in-stream stores. Wat. Res. 38, 1215e1224. Obiri-Danso, K., Jones, K., 2000. Intertidal sediments as reservoirs for hippurate negative campylobacters, salmonellae and faecal indicators in three EU recognised bathing waters in North West England. Water Res. 34, 519e527. Parchure, T.M., Mehta, A.J., 1985. Erosion of soft cohesive sediment deposits. J. Hydraul. Eng. ASCE 111, 1308e1326. Rehmann, C.R., Soupir, M.L., 2009. Importance of interactions between the water column and the sediment for microbial concentrations in streams. Wat. Res. 43, 4579e4589. Schutze, G.E., Kirby, R.S., Flick, E.L., Stefanova, R., Eisenach, K.D., Cave, M.D., 1998. Epidemiology and molecular identification of Salmonella infections in children. Arch. Pedi. Adolesc. Med. 152, 659e664. Skafel, M.G., Krishnappan, B.G., 1999. A laboratory investigation of depositional characteristics of mud from an inland harbour using a rotating circular flume. Water Air Soil Pollut. vol. 112, 1e19. Stone, M., Krishnappan, B.G., 1997. Transport characteristics of tile-drain sediments from an agricultural watershed. Wat. Air Soil Pollut. 99, 89e103. Wingender, J., Neu, T.R., Flemming, H.-C. (Eds.), 1999. Microbial Extracellular Polymeric Substances. Springer-Verlag, Berlin, Germany, p. 247. Wu, J., Rees, P., Storrer, S., Alderisio, K., Dorner, S., 2009. Fate and transport modeling of potential pathogens: the contribution from sediments. J. Am. Wat. Res. Assoc. 45, 35e44.