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 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 (121, 71e120, 41e70, 40e20, 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 (IBMWP1000 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 (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