Treatment of micropollutaOzone or powdered activa

Jonas Margot a,, Cornelia Kienle b

ThonEngineerEPFL, berreaux 334, 65439echnologiversityter Divisi

by both

Both treatments signicantly reduced the toxicity of WWTP efuent.

inking water resources ored to be themost suitableut forming problematic

Science of the Total Environment 461462 (2013) 480498

Contents lists available at SciVerse ScienceDirect

Science of the Total Environment

j ourna l homepage: www.e lsev ie r .com/ locate /sc i totenv 2013 Elsevier B.V. All rights reserved.disposal, disinfection), operational feasibility and cost. For sensitive receiving waters (drrecreational waters), the PAC-UF treatment, despite its current higher cost, was consideroption, enabling good removal of most micropollutants and macropollutants withoby-products, the strongest decrease in toxicity and a total disinfection of the efuent.a b s t r a c ta r t i c l e i n f o

Article history:Received 12 February 2013Received in revised form 9 May 2013Accepted 9 May 2013Available online 8 June 2013

Editor: Simon James Pollard

Keywords:Organic micropollutantPharmaceuticalWastewater treatmentOzonePowdered activated carbonEfuent toxicity

Many organic micropollutants present in wastewater, such as pharmaceuticals and pesticides, are poorlyremoved in conventional wastewater treatment plants (WWTPs). To reduce the release of these substancesinto the aquatic environment, advanced wastewater treatments are necessary. In this context, two large-scale pilot advanced treatments were tested in parallel over more than one year at the municipal WWTP ofLausanne, Switzerland. The treatments were: i) oxidation by ozone followed by sand ltration (SF) andii) powdered activated carbon (PAC) adsorption followed by either ultraltration (UF) or sand ltration.More than 70 potentially problematic substances (pharmaceuticals, pesticides, endocrine disruptors, drugmetabolites and other common chemicals) were regularly measured at different stages of treatment.Additionally, several ecotoxicological tests such as the Yeast Estrogen Screen, a combined algae bioassayand a sh early life stage test were performed to evaluate efuent toxicity. Both treatments signicantlyimproved the efuent quality. Micropollutants were removed on average over 80% compared with rawwaste-water, with an average ozone dose of 5.7 mg O3 l1 or a PAC dose between 10 and 20 mg l1. Depending onthe chemical properties of the substances (presence of electron-richmoieties, charge and hydrophobicity), eitherozone or PAC performed better. Both advanced treatments led to a clear reduction in toxicity of the efuents,with PAC-UF performing slightly better overall. As both treatments had, on average, relatively similar efciency,further criteria relevant to their implementation were considered, including local constraints (e.g., safety, sludge Both treatments are feasible for use in municipal WWTPs. Corresponding author at: EPFL ENAC IIE ECOL, StatiE-mail addresses: jonas.margot@ep.ch (J. Margot),

luca.rossi@ep.ch (L. Rossi), felippe.dealencastro@ep.cnathalie.chevre@unil.ch (N. Chvre), michael.schaerer@

1 Present address: Swiss Wastewater Association (VS2 Present address: Technical Centre for Sanitation and

0048-9697/$ see front matter 2013 Elsevier B.V. Allhttp://dx.doi.org/10.1016/j.scitotenv.2013.05.034tly by ozone.a wider range of pollutants. Specic substances were removed more efcien Powdered activated carbon effectively removedChristian Abegglen e,1, Denisa School of Architecture, Civil and Environmentalb Swiss Centre for Applied Ecotoxicology, Eawag/c Sanitation Service, City of Lausanne, Rue des ted ECT Oekotoxikologie GmbH, Boettgerstrasse 2-1e Swiss Federal Institute of Aquatic Science and Tf Faculty of Geosciences and the Environment, Ung Federal Ofce for the Environment (FOEN), Wa

H I G H L I G H T S

Micropollutants are efciently removednts in municipal wastewater:ted carbon?

, Anos Magnet c, Mirco Weil d, Luca Rossi a, Luiz Felippe de Alencastro a,ney c,2, Nathalie Chvre f, Michael Schrer g, D.A. Barry a

ing (ENAC), Ecole Polytechnique Fdrale de Lausanne (EPFL), Station 2, 1015 Lausanne, Switzerlandrlandstrasse 133, 8600 Dbendorf, Switzerland, 1002 Lausanne, SwitzerlandFloersheim/Main, Germanyy (Eawag), berlandstrasse 133, 8600 Dbendorf, Switzerlandof Lausanne, 1015 Lausanne, Switzerlandon, 3003 Bern, Switzerland

ozone and powdered activated carbon.on 2, 1015 Lausanne, Switzerland. Tel.: +41 21 693 8086; fax: +41 21 693 8035.cornelia.kienle@oekotoxzentrum.ch (C. Kienle), anoys.magnet@lausanne.ch (A. Magnet), m.weil@ect.de (M. Weil),h (L.F. de Alencastro), christian.abegglen@vsa.ch (C. Abegglen), denis.thonney@sige.ch (D. Thonney),bafu.admin.ch (M. Schrer), andrew.barry@ep.ch (D.A. Barry).A), Postfach, 8152 Glattbrugg, Switzerland.WWTP, Inter-Municipal Management Service (SIGE), Quai Maria-Belgia 18, 1800 Vevey, Switzerland.

rights reserved.

481J. Margot et al. / Science of the Total Environment 461462 (2013) 4804981. Introduction

About 3000 pharmaceutical compounds and more than 300 pesti-cides and biocides are commercially available in Switzerland (OPBio,2005; OPPh, 2010; Swissmedic, 2012). They can enter urban sewersystems via human excretion in urine and feces, by improper disposal,or through leaching of pesticides and biocides from urban areas duringrain events. In conventional wastewater treatment plants (WWTPs),many of these hydrophilic organic compounds are poorly removed(Choubert et al., 2011; Deblonde et al., 2011), and are thus characterizedby a relatively constant input at low concentrations (ng l1 to g l1)into the aquatic environment. As most of these substances are designedto be biologically active, they can affect sensitive aquatic organismseven at very low concentrations (Santos et al., 2010), hence the namemicropollutant. For instance, endocrine effects on sh and musselpopulations such as intersex, reproductive disruption or feminizationof males have been observed in rivers downstream of municipalWWTP outfalls (Alan et al., 2008; Gagn et al., 2011; Tetreault et al.,2011; Tyler and Jobling, 2008; Vethaak et al., 2005; Woodling et al.,2006). These effects were attributed to the release of endocrine-activechemicals such as the synthetic estrogen 17-ethinylestradiol (foundin contraceptive pills), natural estrogens estrone and 17-estradiol ornonylphenol. Furthermore, as lakes and rivers are used in many placesfor drinkingwater supply, pharmaceuticals and pesticides can thereforebe found in tap water at very low concentrations, even after drinkingwater treatment (Huerta-Fontela et al., 2011; Mompelat et al., 2009;Stackelberg et al., 2007). Acute human health effects are not expected(Webb et al., 2003), but effects of long term exposure are unknownand, therefore, the release of these compounds into the environmentshould be avoided.

Efuents of WWTPs are the main source of pharmaceuticals in theaquatic environment (Bartelt-Hunt et al., 2009; da Silva et al., 2011).Since it is unrealistic to limit the consumption of pharmaceuticals,additional steps during wastewater treatment are one of the best op-tions to reduce the release of these compounds into surface waters.Currently, two main technologies with a potential for large-scaleapplication in terms of efciency, costs and energy requirementshave been identied (Abegglen and Siegrist, 2012; Joss et al., 2008):oxidation of micropollutants with ozone or adsorption onto activatedcarbon.

Through the strong oxidative properties of ozone and of thehydroxyl radicals produced spontaneously in its decomposition,ozonation was found to degrade efciently most micropollutantspresent in treated wastewater with a dose of 38 mg O3 l1

(Hollender et al., 2009; Lee et al., 2012; Nakada et al., 2007; Reungoatet al., 2010, 2012; Rosal et al., 2010). A potential disadvantage of thisprocess is the formation of unknown reactive by-products due to partialoxidation of the compounds and reaction with matrix components(von Gunten, 2003a). For example, undesirable toxic oxidation by-products such as nitrosamines N-Nitrosodimethylamine (NDMA),bromate or formaldehyde can be formed (Hollender et al., 2009;Richardson, 2003;Wert et al., 2007), potentially increasing the toxic-ity compared to non-ozonated wastewater (Petala et al., 2006, 2008;Stalter et al., 2010a, 2010b). These oxidation products are usuallymore easily biodegradable and can be partially removed duringbiological post-ltration (Hollender et al., 2009; Richardson et al.,1999; Stalter et al., 2010a, 2010b).

Activated carbon allows removal of a broad spectrum of micro-pollutants via adsorption to its high specic surface area and isthus widely used in drinking water treatment (Snyder et al., 2007;Westerhoff et al., 2005). As organic matter present in wastewater efu-ent can compete for adsorption sites, larger amounts of activated carbonare required. The efciency of granular activated carbon (GAC)ltrationto remove micropollutants has been studied in someWWTPs, showinga mitigated efciency depending on the compound and the frequency

of GAC regeneration/replacement (Grover et al., 2011; Nguyen et al.,2012; Reungoat et al., 2010, 2012; Snyder et al., 2007). Powdered acti-vated carbon (PAC) adsorption, with a dosage of 1020 mg l1, hasbeen proposed as a more efcient alternative compared to GAC treat-ment (Boehler et al., 2012; Metzger et al., 2005; Nowotny et al., 2007;Serrano et al., 2011). However, to date, very few large scale studiesevaluating the efciency of micropollutant removal via PAC treatmentin municipal wastewater have been reported.

In order to nd a feasible and efcient solution for the removal ofpharmaceuticals and pesticides in wastewater, a global pilot studywas conducted at the municipal WWTP of Lausanne, Switzerland.The goals were to evaluate and compare the efciency of ozonationand PAC adsorption (i) to remove a broad range of micropollutantsinWWTP efuents, and (ii) to reduce ecological impacts of the efuent.Finally, we aimed to determine the feasibility of these advanced treat-ments at the WWTP scale in terms of operation, energy consumptionand costs.

2. Materials and methods

2.1. Lausanne wastewater treatment plant

The municipal WWTP of Lausanne, Switzerland, is the largest inthe Lake Geneva watershed and treats on average 95,000 m3 d1 ofwastewater representing a population equivalent (PE) of 220,000individuals. The sewer system is only partially separated, collecting asignicant amount of urban runoff during rain events. The watershedincludes amajor hospital and several clinics, which are a potential sourceof specic pharmaceuticals. The wastewater treatment consists ofpre-treatments (grit removal and screening at 1 cm), primary clariers,biological activated sludge treatment (AS, sludge age of 2 d) withoutnitrication, or, for 5% of the ow, a moving bed bioreactor (MBBR)with partial to complete nitrication (b1 mg N-NH4 l1). In both treat-ments, phosphorus is removed by precipitation with iron chloride.Treated wastewater (WWTP efuent) is then discharged in LakeGeneva, which is the main drinking water reservoir for more than600,000 inhabitants (www.cipel.org, last accessed 7 May 2013).

2.1.1. Ozonation pilot plantThe pilot plant for ozonationwas designed to treat a maximum ow

rate of 100 l s1 (13,000 PE) and consisted of a plug ow reactor(volume of 129 m3) separated into four chambers (nine compartments)in series (Fig. 1a) to assure optimal hydraulic conditions and a minimalreaction time of 20 min. Characteristics of the feed water (efuent ofthe conventional WWTP) are presented in Table 1. Ozone-containinggas (214% w/w) was continuously produced by an ozone generator(Efzon SMO 600 from ITT Wedeco, Wallisellen, Switzerland) fed withpure oxygen. 60% of the gas was injected counter currently into the1st or 2nd chamber depending of the water ow rate and 40% in the3rd chamber. The reaction time in the reactor ranged between 20 and60 min. The ozone dosagewas automatically adjusted to thewater qual-ity (oxidative demand) by varying the gas ow to maintain a constantresidual concentration of dissolved ozone (around 0.1 mg O3 l1),measured with an online sensor (AMI codes II, from Swan, Hinwill,Switzerland), and conrmed with a second probe (AquaTector fromMesin, Winterthur, Switzerland) at the outlet of the 3rd chamber.Corresponding initial ozone doses varied between 2 and 13 mg O3 l1,with on average 5.7 mg O3 l1. Ozone concentrations in the feed andoff gas were continuously measured with BMT 964 probes (Berlin,Germany). The transfer efciency of ozone into the dissolved phasewas from 70 to over 90% depending on the gas ow. In this paper, theozone dose refers to the amount of gaseous ozone injected and not tothe ozone dissolved into the water. The remaining gaseous ozone wascatalytically converted to oxygen before its release into the atmosphere.The efuent of the ozone reactor was then ltered through a rapid sandlter (ux of 8 m h1, characteristics described in the next section)

with biological activity to remove reactive oxidation products.

482 J. Margot et al. / Science of the Total Environment 461462 (2013) 4804982.1.2. Powdered activated carbon treatment pilot plantThe pilot plant for PAC treatment was designed to treat WWTP ef-

uent, in parallel to the ozonation, at a maximum ow of 1015 l s1

(ca. 1700 PE). Based on bench-scale batch adsorption tests on ve dif-ferent PACs (Omlin and Chesaux, 2010), two PACs were selected forthe pilot study to assess if the treatment efciency was inuencedby the type of PAC: Norit SAE SUPER (Norit Activated Carbon, TheNetherlands), with grain size d50 of 15 m, specic surface area of1150 m2 g1, pH of point of zero charge pHPZC > 7.3, and ash contentof 12%; and SORBOPORMV-125 (Enviro Link SA, Switzerland) with

Fig. 1. (a) Ozonation installation. (b) Powdered activated carbon (PAC) insta

Table 1Characteristics of the efuent of the biological treatments (feed water for the advancedtreatments). Average and standard deviation of 33 24-h composite samples taken afterthe biological treatment with low to complete nitrication depending on the campaigns.

Conventional parameters

Total suspended solids (TSS) [mg l1] 14.8 (5.3)Dissolved organic carbon (DOC) [mg l1] 7.3 (1.9)Chemical oxygen demand (COD) [mg l1] 24.4 (12)Biochemical oxygen demand (BOD5) [mg l1] 11.2 (10.2)N-NH4 [mg l1] 7.7 (7.7)N-NO3 [mg l1] 9.9 (5.6)N-NO2 [mg l1] 0.4 (0.3)Ptotal [mg l1] 0.7 (0.6)Psoluble [mg l1] 0.09 (0.08)pH [] 7.2 (0.4)Temperature [C] 17.1 (3.5)Conductivity [S cm1] 914 (96)grain size d80 b 45 m, specic surface area of 1100 m2 g1, pHPZC of911, and ash content b6%. Norit SAE SUPER and SORBOPOR MV-125were used during the rst and the second half of the study respectivelywith ultraltration separation. The installation was composed of awell-mixed contact reactor of 30 m3 where PAC slurry (35 g l1)was added continuously in proportion to the wastewater ow toreach a nal dosage of 10 to 20 mg PAC l1. A coagulant (FeCl3 at415 mg l1) was added to improve the subsequent separation of thePAC. Treated water was then ltrated in low transmembrane-pressure(0.10.3 bar) cross-ow hollow ber ultraltration (UF) membranes(Norit AirLift, in PVDF, molecular weight cut-off of 100300 kDa,total ltration surface of 660 m2) to remove the PAC (Fig. 1b). Thetangential ltration process allowed increasing the concentration (upto 12 g PAC l1) and the residence time of the PAC in the system.Every 4 h, the system was partially drained (volume removed propor-tional to the PAC dose) to maintain a constant PAC concentration inthe reactor and to remove excess old PAC, which was then incineratedwith the sewage sludge from the conventional treatment. The hydraulicresidence time in the contact reactor varied between 40 and 170 min,depending on the ow rate. The corresponding solid (PAC) residencetime was between 2 and 17 d in order to reach adsorption equilibrium.UF membranes were backwashed every 10 min for 10 s, and chemicalcleaning with citric acid and sodium hypochlorite was performedevery month to avoid fast clogging of the membranes. The PAC separa-tionwas also studied over a 5-month periodwith a pilot sand lter (SF)without concentration and recirculation of the PAC (with the PACSORBOPOR MV-125). The lter, also used after ozonation, consisted of

llation with ultraltration separation (after Margot and Magnet, 2011).

483J. Margot et al. / Science of the Total Environment 461462 (2013) 4804981.2 m of expanded shale (grain size 1.62.4 mm), and 60 cm of quartzsand (grain size 0.71.2 mm), with a ltration ux of 816 m h1

and one backwash per day. Supplementary information concerningthe operation of the ozonation and PAC-UF pilot plants can be foundin Margot et al. (2011).

2.2. Sampling campaigns

The pilot systems were operated continuously for more than oneyear. To monitor long term efciency and to optimize the treatments,25 sampling campaigns of one day (23 per month) and four seasonalcampaigns of oneweekwere performedbetween June2009 andOctober2010. During the campaigns, 24-h to 72-h composite samples (takentime proportional every 15 min) were collected with refrigerated auto-matic samplers (ISCO 6712 FR, Teledyne, USA, and WS 316, Watersam,Germany) at 5 locations: 1. Inuent of the WWTP after grit removaland screening (Inuent), 2. Efuent of the biological activated sludgetreatment (only the rst seasonal campaign) or efuent of the biologicalMBBR with nitrication (BIO), 3. Efuent of the ozone reactor (OZ), 4.Efuent of the sand lter following the ozonation (SF), and 5. Efuentof the PAC with ultraltration (PAC-UF) or with sand lter (PAC-SF)treatment (the last seven one-day campaigns). Composite sampleswere stored at 4 C and transferred in less than 12 h (or 24 h for thebioassays) to the laboratories performing analyses.

2.3. Chemicals and reagents

High purity micropollutants, deuterated standards and reagentsused for micropollutant analysis have been listed previously (Moraschet al., 2010).

2.4. Analyses of micropollutants

Upon arrival in the laboratory, samples were immediately acidiedto pH 2.5 with 5 N HCl and ltered at 0.7 m through glass ber lters(type GF/F, Whatman). Analysis of 58 hydrophilic micropollutants(36 pharmaceuticals, 13 biocides and pesticides, 2 corrosion inhibitorsand 7 endocrine compounds, Table S1, Supporting information (SI)),identied in Switzerland as priority micropollutants (Morasch et al.,2010; Perazzolo et al., 2010), was conducted on the ltrate by solidphase extraction (SPE) followed by ultra-performance liquid chroma-tography coupled to tandem quadrupole mass spectrometer (UPLCMS/MS). The analytical method, described in Morasch et al. (2010),was developed and validated for wastewater matrix (further detailscan be found in the SI). Uncertainties of the micropollutant analyses,including recovery and repeatability uncertainties, were compound-and concentration-dependent with a decreased reproducibility closeto the limit of detection (LOD). For the largemajority of the compounds,the relative standard deviation was b30% (Bonvin et al., 2011). Com-pounds detected with this method are presented in Table 2 (analyticalmethod A) with their respective LODs.

Chemical properties of these 58 micropollutants are reported inTable S1, SI. Hydrophobicity was expressed by the log Dow at pH 7, acorrected form of the octanolwater partition coefcient (log Kow)determined for non-ionic substances, to account for the molecule dis-sociation or protonation at pH 7 (de Ridder et al., 2010). The log Dowvalues were calculated from the corresponding pKa values followingSchwarzenbach et al. (2003). For neutral molecules, log Dow = logKow, for ionic compounds, log Dow b log Kow.

During two seasonal campaigns, a broader range of 120 micro-pollutants, including human pharmaceutical metabolites, were ana-lyzed on ltered 7-d composite samples (glass ber lter APFD09050,Millipore) following Hollender et al. (2009) and Kern et al. (2009).The method consists of SPE, with the same cartridges as for method A,followed by LCMS/MS with an XBridge C-18 column (Waters) and

Linear Trap Quadrupole orbitrap mass spectrometer with electrosprayionization (Thermo Fisher Scientic Corporation, USA). Compoundsdetected with this method are presented in Table 2 (analyticalmethod B) with their respective LODs.

Analyses of the endocrine disrupting compounds 17-estradiol (E2)and 17-ethinylestradiol (EE2) (analytical method C in Table 2) weredone on ltered 7-d composite samples during two seasonal campaignsby solid phase extraction (LiChrolut EN-RP18 cartridge, Merck,Germany) followed by LCMS/MS detection (API 4000 LCMS/MS,Applied Biosystems, USA). The method used is described in TablesS2 and S3, SI.

In the case when the efuent concentration was below the LOD ofthe compound, the removal rate was calculated as a minimum valueusing the LOD as efuent concentration. These minimum removalrates were not integrated into the global removal average unlessthey were above 80%, in order not to bias the results.

2.5. Bioassays

In this pilot study, a broad range of bioassays was performed,showing that most acute toxicity bioassays were not sensitive enoughto detect the effects of low micropollutant concentrations in waste-water. An overview on these bioassays can be found in Kienle et al.(2011). Two kinds of assays were therefore selected based on theirsensitivity: i) bioassays on enriched samples and ii) chronic toxicitytests in the whole efuent. These two approaches can be seen ascomplementary for evaluating the effects of the efuents: the rstmentioned assays are very sensitive and focus on the effects producedby specic pollutants, while the second assays evaluate the long-termtoxicity of the efuent, including the effect of very polar compoundsnot well extracted during the enrichment process, such as ozonationby-products (Stalter et al., 2011). For the rst approach, two bioassayswere performed on enriched samples: the Yeast Estrogen Screen (YES) toevaluate the estrogenicity (Routledge and Sumpter, 1996) and theCombined Algae Assay to evaluate the global toxicity and the presenceof photosynthesis inhibitors (Escher et al., 2008b). For the secondapproach, a sh early life stage test (FELST, OECD, 1992b) with rainbowtrout was performed by exposing the sh for 67 d to the efuent fromthe different treatments under ow-through conditions.

2.5.1. Sample enrichment (YES and Combined Algae Assay)The sample enrichment was done by solid phase extraction (SPE),

which allows for increased pollutant concentrations in the extractsand thereby enables a better detection in the bioassays. It also limitsthe impact of the matrix components and metals, which are partiallyseparated during the extraction (Macova et al., 2010). 7-D compositesamples were enriched using SPE as described in Escher et al. (2008b)and as presented in Table S2, SI. Briey, 200 ml (inuent samples) or500 ml (all others) was enriched 200 and 500 times respectivelyusing LiChrolut EN-RP18 cartridges (Merck, Germany), and thenstored in 1 ml of a solvent mixture (~50% ethanol, ~50% acetoneand methanol) at 20 C until analysis.

2.5.2. Yeast Estrogen Screen (YES)TheYeast Estrogen Screenwith the recombinant yeast Saccharomyces

cerevisiaewas performed according to Routledge and Sumpter (1996) in96-well microtiter plates as described in the SI. The estrogenic activity inthe wastewater samples was assessed relative to a reference substance(17-estradiol, a potent estrogen) and expressed as 17-estradiolequivalent concentrations (EEQ). Both, the reference substance andthewastewater samples, were tested in triplicates in a 1:2 dilution series.The highest tested concentration of 17-estradiol was 1.25 109 M(340 ng l1, in ethanol) and the maximum enrichment factors of thewastewater samples were 5 for the WWTP inuent and 50 for all addi-tional treatment steps. Ethanol served as solvent control (50 l/well, 8

wells/plate).

Table 2Concentrations of 70 micropollutants in raw wastewater and after biological treatment (WWTP efuent), and removal rate obtained with the conventional (with low to com te nitrication) or the advanced treatments (in reference to theconcentration in the efuent of the biological treatment) (ozone doses between 2.3 and 9 mg l1 (median 5.9 mg l1) and PAC doses between 10 and 20 mg l1 (median mg l1)). Average with standard deviation of n analyses (24-hcomposite samples) conducted between June 2009 and October 2010. Compounds with analytical method A were regularly analyzed to monitor the efciency of the treat nts. Data for compounds with analytical methods B and C cor-respond to one or two analyses of a 7-d composite sample taken for a larger screening campaign (with partial nitrication, 6 mg O3 l1, or 12 mg PAC l1). Comparison w removal rates obtained in other studies in similar conditions ispresented for the two advanced treatments.

Compound Compound class LOD(ng l1)

Analyticalmethod

Number ofanalysis (n)

Inuent concentration(ng l1)

(n) Efuent concentration(ng l1)

( WWTPremoval (%)

(n) Ozoneremoval (%)

(n) PAC-UFremoval (%)

PharmaceuticalsAtenolol Beta blocker 1.2 A 37 1274 (436) 37 682 (267) 3 42 (27) 28 85 (14)a 21 88 (9)e

Azithromycin Antibiotic 75.6 A 19 2272 (1472) 19 935 (333) 1 44 (26) 12 74 (10)d 8 76 (8)c

Bezabrate Lipid regulator 1.5 A 37 953 (262) 37 595 (314) 3 38 (26) 27 81 (8)a 21 79 (12)e

Carbamazepine Anticonvulsant 0.1 A 37 482 (586) 37 461 (292) 3 7.6 (18) 28 97 (4)a 21 90 (9)e

Ciprooxacin Antibiotic 36.5 A 19 2291 (600) 19 779 (372) 1 63 (18) 12 53 (29)b8 8 63 (32)f

Clarithromycin Antibiotic 0.4 A 37 709 (418) 37 440 (302) 3 37 (26) 28 93 (4)a 21 92 (5)e

Clindamycin Antibiotic 0.2 A 19 65 (33) 19 115 (69) 1 0 (0) 12 99 (1)a 8 82 (13)c

Diatrizoic and iothalamic acid Iodinated contrast medium 32.8 A 17 597 (628) 19 370 (366) 1 28 (25) 12 16 (16)b2 8 15 (13)e

Diclofenac Analgesic/anti-inammatory 1.2 A 37 1197 (497) 37 1187 (389) 3 9 (14) 28 94 (3)a 21 69 (19)e

Eprosartan Antihypertensive 20 B 2 1055 (488) 1 880 37 1 98c 1 65c

Fluconazole Antifungal 20 B 2 120 (14) 1 110 15 1 27d 1 >64c

Gabapentin Anticonvulsant 1.8 A 37 3867 (1339) 37 3692 (1456) 3 9.2 (12) 28 38 (16)b5 21 11.8 (11)f

Gembrozil Lipid regulator 2.9 A 19 411 (128) 19 265 (159) 1 36 (32) 12 94 (5)b11 8 76 (16)d

Ibuprofen Analgesic/anti-inammatory 13.4 A 19 4101 (2465) 19 952 (759) 1 57 (46) 11 63 (12)b11 6 83 (7)e

Iohexol Iodinated contrast medium 2177.3 A 35 21,275 (6975) 34 15,191 (7294) 3 31 (27) 26 38 (16)a 19 57 (25)e

Iomeprol Iodinated contrast medium 306.9 A 35 14,467 (9657) 35 10,534 (6338) 3 25 (24) 28 43 (12)b2 20 54 (21)c

Iopamidol Iodinated contrast medium 145.4 A 30 3360 (2574) 30 2535 (1587) 3 21 (20) 24 42 (13)a 16 49 (21)e

Iopromide Iodinated contrast medium 2044.6 A 22 6408 (2663) 23 4141 (2086) 2 29 (27) 15 34 (19)a 11 47 (30)c

Irbesartan Antihypertensive 20 B 2 4700 (4808) 1 1700 79 1 51b7 1 98c

Ketoprofen Analgesic/anti-inammatory 6.0 A 19 1119 (1328) 19 669 (757) 1 32 (21) 12 63 (16)a 8 81 (9)c

Levetiracetam Anticonvulsant 10 B 2 2100 (566) 1 330 87 1 18a 1 >97c

Losartan Antihypertensive 20 B 2 2405 (2256) 1 510 87 1 >96b7 1 80c

Mefenamic acid Analgesic/anti-inammatory 2.6 A 19 946 (455) 19 581 (299) 1 33 (29) 12 98 (2)a 8 93 (2)e

Metformin Antidiabetic b1000 B 2 >10,000 1 >4000 0 1 >55c

Metoprolol Beta blocker 4.4 A 19 561 (299) 19 653 (400) 1 4.6 (13) 12 88 (8)a 8 95 (4)f

Metronidazole Antibiotic 21.0 A 19 1168 (866) 19 567 (497) 1 45 (34) 12 64 (12)b6 5 79 (17)c

Morphine Analgesic/anti-inammatory 20 B 1 270 1 190 30 1 >90c 1 >90c

Naproxen Analgesic/anti-inammatory 9.4 A 37 697 (249) 37 380 (110) 3 41 (23) 28 90 (8)a 21 81 (12)e

Noroxacin Antibiotic 1.9 A 19 334 (167) 19 59 (35) 1 76 (19) 12 75 (29)b9 8 82 (21)c

Ooxacin Antibiotic 0.4 A 19 234 (60) 19 84 (36) 1 61 (17) 12 85 (20)c 8 83 (24)c

Oxazepam Anxiolytic 20 B 2 305 (134) 1 350 13 1 9d 1 69c

Paracetamol Analgesic/anti-inammatory 7.9 A 18 51,438 (31,884) 18 b7.9 1 100 (0) 1 >85b11 0 Primidone Anticonvulsant 0.7 A 37 114 (39) 37 97 (21) 3 16 (15) 28 57 (11)a 21 51 (19)f

Propranolol Beta blocker 0.3 A 19 127 (37) 19 114 (17) 1 13 (17) 12 99 (1)a 8 99 (1)c

Ritonavir Antiretroviral 20 B 2 110 (14) 1 90 25 1 >78c 1 >56c

Simvastatin Lipid regulator 29.7 A 14 736 (503) 14 98 (96) 14 1 77 (23) 8 >70c 4 >65c

Sotalol Beta blocker 0.5 A 37 337 (175) 37 247 (63) 3 23 (20) 28 99 (1)a 21 81 (15)c

Sulfamethoxazole Antibiotic 0.2 A 37 340 (261) 37 171 (127) 3 38 (30) 25 93 (7)a 20 64 (25)e

Trimethoprim Antibiotic 0.2 A 37 235 (52) 37 158 (73) 3 35 (23) 28 99 (2)a 21 94 (4)f

Valsartan Antihypertensive 5 B 2 2250 (354) 1 2100 16 1 61b7 1 65c

Venlafaxine Antidepressant 10 B 2 235 (21) 1 150 40 1 75d 1 46d

484J.M

argotet

al./Science

oftheTotalEnvironm

ent461

462(2013)

480498ple12meith

n)

79779797711799250119119199179919791477711

Pharmaceutical metabolites10,11-Dihydro-10,11-dihydroxycarbamazepine

Drug metabolite 10 B 2 975 (106) 1 1000 1 0 1 47b10 1 52c

Atenolol acid Drug metabolite 10 B 2 1550 (212) 1 1700 1 0 1 72d 1 >99c

Fenobric acid Drug metabolite 20 B 2 390 (57) 1 490 1 0 1 57b1 1 78c

Formyl-4-aminoantipyrine Drug metabolite 10 B 2 445 (92) 1 700 1 0 1 >99b6 1 59c

N,N-didesvenlafaxine Drug metabolite 10 B 1 250 1 330 1 0 1 >97c 1 61c

N-acetyl sulfamethoxazole Drug metabolite 20 B 2 570 (156) 1 50 1 93 1 50b2 1 >20c

N-acetyl-4-aminoantipyrine Drug metabolite 20 B 2 920 (28) 1 1200 1 0 1 >98b6 1 34c

Valsartan acid Drug metabolite 10 B 2 125 (21) 1 150 1 0 1 39c 1 43c

Endocrine disrupting compounds17-Ethinylestradiol Hormonal contraceptive 1.9 C 2 5.3 (4.3) 1 b1.9 1 >18 0 1 17-Estradiol Hormone 0.5 C 2 14 (1) 1 1.3 1 91 1 >61b2 1 >61c

Bisphenol A Plastic component 48.9 A 18 834 (460) 18 338 (311) 18 50 (36) 3 >95b4 3 >83c

Estriol Hormone 97.5 A 12 306 (140) 12 b97.5 11 >75 (12) 0 0 Estrone Hormone 15.6 A 12 134 (87) 12 71 (83) 12 58 (31) 3 >90b2 3 >92c

Biocides pesticidesAtrazine Herbicide 0.2 A 37 21 (16) 37 14 (8) 37 20 (24) 28 34 (13)a 21 74 (17)c

Carbendazim Fungicide 16.1 A 19 106 (92) 19 132 (79) 19 1.5 (3.5) 12 79 (17)c 5 >93e

Diuron Herbicide 13.7 A 9 69 (49) 9 70 (41) 9 10 (16) 7 73 (16)a 3 >82f

Irgarol Algicide 1.0 A 19 16 (14) 19 7.5 (6.2) 19 34 (29) 10 32 (21)d 5 0 to >60c

Isoproturon Herbicide 16.9 A 16 62 (67) 16 39 (32) 16 27 (22) 3 68 (26)b3 2 75 (12)e

Mecoprop Herbicide 9.6 A 37 386 (408) 37 245 (239) 37 29 (25) 28 60 (22)a 21 48 (27)e

Propiconazole Fungicide 6.9 A 19 59 (28) 19 40 (17) 19 28 (16) 12 32 (14)c 7 66 (15)c

Terbutryn Algicide 0.1 A 37 38 (21) 37 19 (16) 37 49 (25) 28 85 (10)a 20 80 (13)c

Other common chemicalsAspartame Sweetener b100 B 2 >10,000 1 >4000 1 0 1 Benzothiazole Industrial additive 400 B 2 6500 (566) 1 1400 1 80 1 7d 1 >71c

Benzotriazole Corrosion inhibitor 4.1 A 37 9224 (3112) 37 6948 (1846) 37 24 (22) 28 64 (14)a 21 90 (7)e

Caffeine Food component b50 B 2 >10,000 1 820 1 >92 1 >92b11 1 65f

Galaxolidone Fragrance (HHCB) metabolite 40 B 2 335 (177) 1 220 1 52 1 0d 1 77c

Methylbenzotriazole Corrosion inhibitor 48.5 A 19 5720 (2810) 19 4201 (2488) 19 29 (24) 12 80 (15)a 8 96 (2)e

N,N-diethyl-3-methylbenzamide (DEET) insect repellent b50 B 2 805 (445) 1 290 1 74 1 48b8 1 66f

Oxybenzone UV lter 20 B 2 425 (290) 1 60 1 90 1 >67a 1 50c

aSimilar removal (b10% difference) obtained with about 0.6 g O3 g1DOC by Hollender et al. (2009). bSimilar range of removal obtained in other studies (1Ternes et al., 2003, 2Huber et al., 2005, 3Ormad et al., 2008, 4Wert et al., 2009,5Reungoat et al., 2010, 6Rosal et al., 2010, 7Huerta-Fontela et al., 2011, 8Yang et al., 2011, 9Senta et al., 2011, 10Bundschuh et al., 2011, 11Sudhakaran et al., 2012). cNot reported in other studies. dContradictory to other studies (>10%lower removal) (Hollender et al., 2009, Reungoat et al., 2010, 2012). eSimilar removal (b10% difference) obtained with 10 to 20 mg PAC l1 by Zwickenpug et al. 2010 (in Abegglen and Siegrist, 2012). fSimilar range of removalobtained with granular activated carbon (GAC) lters (Reungoat et al., 2010, 2012, Yang et al., 2011).

485J.M

argotet

al./Science

oftheTotalEnvironm

ent461

462(2013)

480498

486 J. Margot et al. / Science of the Total Environment 461462 (2013) 4804982.5.3. Combined Algae AssayThe Combined Algae Assay on the green algae Pseudokirchneriella

subcapitata was conducted as described by Escher et al. (2008a) andin the SI. Photosynthesis inhibition wasmeasured bymeans of effectivequantum yield (after 2 h of exposure) and algae growth by means ofabsorbance at 685 nm (after 24 h of exposure). The herbicide diuronserved as the reference substance and ethanol as the solvent control(50 l/well, 8wells/plate). Both, the reference substance and thewaste-water samples, were tested in triplicates in a 1:2 dilution series, withthe highest concentration of diuron being 3 107 M (69.9 g l1,in ethanol) and maximum enrichment factors of the wastewater sam-ples of 20 (WWTP inuent) and 83.3 (all additional treatment steps).The toxicity of the wastewater samples was expressed as diuron-equivalent concentrations (DEQs) for the endpoint inhibition ofPhotosystem II and toxic equivalent concentrations (TEQs, virtualbaseline toxicant) for growth inhibition (Escher et al., 2008a).

Comparison of the measured photosynthesis inhibition with theconcentration of photosynthesis inhibitors was based on the conceptof concentration addition for substances with similar mode of actionaccording to Chvre et al. (2006). The concentrations of the fourmost abundant photosynthesis inhibitors included in the analyticallist, the herbicides atrazine, diuron and isoproturon, and the algicideterbutryn were converted to diuron-equivalents based on their rela-tive potency (HC50EC50: hazardous concentration affecting 50% ofthe species with 50% effect, Chvre et al., 2006) and then summedup. One g l1 of atrazine, diuron, isoproturon and terbutryn corre-sponds to 0.084, 1, 0.559 and 0.881 g l1 DEQs respectively.

2.5.4. Fish early life stage test with rainbow trout (Oncorhynchus mykiss)This test was performed according to OECD guideline 210 (OECD,

1992b). Details of the methodology are described by Stalter et al.(2010b) (further information can be found in the SI). In brief, freshlyfertilized eggs (b1 h, 40 eggs per replicate) of rainbow trout wereexposed to the test waters in 8-l stainless steel vessels in a ow-through system. Reconstituted water (OECD guideline 203, OECD,1992a) served as the control medium. In total, four controls andthree replicate treatments for each wastewater were assessed. Duringthe test period several endpoints were determined daily, namely:hatching, mortality, swim up, malformations and abnormal behavior.After the end of the test sh were humanely killed with an overdoseof MS222 (tricaine methanesulfonate, Sigma-Aldrich, St. Louis, USA).Afterwards, individual sh were blotted dry and fresh weight andlength were measured. Plasma vitellogenin concentration was deter-mined in whole body homogenates of 20 sh per control and waste-water as described by Holbech et al. (2006) using a vitellogeninELISA test kit for rainbow trout (Biosense, Bergen, Norway) in a1:20 dilution.

The signicance of the difference in the response between thetreatments was assessed with the Tukey's test for single-stepmultiplecomparison. Signicant differences are reported for p values b 0.05.All calculations were performed using R (R Foundation for StatisticalComputing, Vienna, Austria).

2.6. Laboratory-scale batch adsorption experiment

The inuence of wastewater dissolved organic carbon (DOC) con-centration on PAC removal efciency was assessed in laboratory-scalebatch adsorption experiments. Adsorption tests were conducted intriplicates on 24-h composite wastewater samples collected at theLausanne WWTP after either simple coagulationprecipitation treat-ment (DOC of 17 mg l1), activated sludge treatment without nitri-cation (DOC of 11 mg l1), or moving-bed bioreactor treatmentwith full nitrication (three composite samples with DOC of 5, 7and 8 mg l1). PAC (10 mg of SORBOPOR MV-125, Enviro Link SA,Switzerland) was added to 1 l of the different types of wastewater

and agitated at 140 rpm during 24 h in the dark at 20 C. Analysesof initial and nal sample concentrations, after ltration at 0.45 m,of carbamazepine, diclofenac, benzotriazole, mecoprop and iopamidolwere done by SPE followed by UPLCMS/MS as described above. DOCwas analyzed by catalytic combustion oxidation method (ShimadzuTNM1 device).

2.7. Other analyses

Standard wastewater quality parameters (TSS, DOC, COD, BOD5, NH4,NO3, NO2, Ptotal, and Psoluble) were regularly analyzed on 24 h-compositesamples by standard methods recommended in Switzerland (DFI,1983). Temperature, pH and conductivity were continuously ana-lyzed on-line with E + Hmeasurement systems (Endress + Hauser,Switzerland). Indicator bacteria (Escherichia coli, intestinal enterococciand total viable bacteria) and coliphages (F-specic (RNA) and somaticphages) were analyzed by standard plate count methods. Bromide andbromate were analyzed by High Performance Ion Chromatography(HPIC) with a post column-reaction, with UV-detection for bromateand suppressed conductivity detection for bromide.

3. Results and discussion

3.1. Micropollutant concentrations in WWTP inuent and efuent

3.1.1. Raw wastewaterMost of the micropollutants analyzed were detected in the raw

wastewater, with 70 compounds quantied in at least one sample(Table 2). The highest average concentrations of pharmaceuticals wereobserved for the analgesics paracetamol (51 g l1) and ibuprofen(4.1 g l1), the iodinated contrast media family (3.3 to 21 g l1),the antidiabetic metformin (>10 g l1) and the antihypertensiveirbesartan (4.7 g l1). High concentrations (>5 g l1) were alsodetected in raw wastewater for food components (aspartame andcaffeine), corrosion inhibitors (benzotriazole and methylbenzotriazole)and an industrial additive (benzothiazole). On average, 25 compoundsreached an inuent concentration >1 g l1. A similar range of concen-trations was observed in other Swiss municipal wastewater (Hollenderet al., 2009), with the exception of the contrast media. In the presentstudy, these showed higher concentrations probably due to the pres-ence of many hospitals and clinics in the watershed.

High variations of the inuent daily average concentration of thesame compound were observed between the different campaigns.A factor >4 in the 1090 percentile range of the concentrations wasobserved for half of the compounds due, inter alia, to variations ofthe consumption of these compounds (Coutu et al., 2013). These var-iations highlight the importance of long term sampling campaigns,lasting at least one year, to cover the different consumption habitsof the respective substances. During rain events, no noticeable differ-ent variations of the inuent concentration could be detected com-pared to the background variability despite dilution by runoff water.Only the pesticides isoproturon, carbendazim and terbutryn showeda signicant concentration increase duringwetweather (p-value b 0.05for the correlation with the dilution factor, Fig. S1, SI), presumably dueto the leaching of facades and runoff of pesticides used in gardens inthe urban area (Burkhardt et al., 2007; Coutu et al., 2012).

3.1.2. Biological treatmentAs presented in Table 2, most of the micropollutants were not

well removed in the conventional biological wastewater treatment.Average removals of less than 50% were found for 50 (i.e., 71%) ofthe 70 compounds detected, with 16 having an average concentra-tion in the efuent above 1 g l1, and 52 a concentration above100 ng l1. Only the analgesic paracetamol was completely elimi-nated in all the campaigns. The most persistent micropollutants(less than 10% removal on average) were the pharmaceuticals carba-

mazepine, clindamycin, diclofenac, gabapentin and metoprolol, the

487J. Margot et al. / Science of the Total Environment 461462 (2013) 480498pesticides carbendazim and diuron, and most of the pharmaceuticalmetabolites. All these compounds have been reported as persistentin many studies (Kupper et al., 2006; Oulton et al., 2010; Singeret al., 2010; Verlicchi et al., 2012). Some compounds such as theantibiotic clindamycin, the beta blocker metoprolol and most of thepharmaceutical metabolites were found at higher concentrations(in the dissolved phase) in the efuent of the biological treatmentthan in the inuent. Similar observations in other studies wereattributed to (i) release during the treatment of compounds trappedin feces particles (Gbel et al., 2007), (ii) biological cleavage in thetreatment of pharmaceutical conjugates (human metabolites) pro-ducing again the parent compound (Onesios et al., 2009), (iii) forma-tion of bacterial metabolites during the biological treatment or(iv) analytical uncertainties.

Large variations of the removal rate in the biological treatmentwereobserved among the different campaigns (Table 2 and Fig. S2a, SI).For 24 of the 42 regularly studied micropollutants, these removal ef-ciency variations could be explained in part by the different levels ofnitrication reached in the biological treatment. Indeed, signicantpositive correlationswere observed between the removal of those com-pounds and the degree of nitrication of ammonium (Table S4, SI), withespecially strong correlations (r > 0.8) for 11 substances and mediumcorrelations (0.6 b r b 0.8) for 7 others (Fig. S3, SI). Less than 30%removal in a non-nitrifying sludge compared with more than 60%elimination in a treatment with complete nitrication was observedfor instance for atenolol, bezabrate, bisphenol A, gembrozil, methyl-benzotriazole or metronidazole (Fig. S3, SI and Margot and Magnet,2011). Similar observations were reported for some of these sub-stances by Clara et al. (2005). The higher micropollutant removalobserved at high nitrication levels is presumably due to the longerhydraulic residence time in the reactor, leading to a higher timeavailable for biodegradation processes, as well as to the presenceof a more diverse microbial population with different metabolismsand a higher activity of nitrifying bacteria. These bacteria have theability to degrade many micropollutants, probably by cometabolicoxidation by the ammonium monooxygenase enzyme (Fernandez-Fontaina et al., 2012). But, even for themost efcient biological treat-ment with complete nitrication (b1 mg N-NH4 l1), less than 50%removal was observed for 21 out of 43 compounds, with an averageremoval of only 50%. These results conrm the need for advancedtreatments.

3.2. Removal of micropollutants with advanced treatments

Both advanced treatments were able to reduce the micropollutantconcentrations in the efuent signicantly (Table 2). The number ofmicropollutants with an average concentration above 1 g l1 in theefuent of the advanced treatments was reduced from 16 in the bio-logically treated wastewater to nine after ozonation and to sevenafter PAC-UF. Substances with concentrations >1 g l1 after bothtreatments were the contrast media iohexol, iomeprol, iopamidoland iopromide, the pharmaceuticals gabapentin and metformin andthe sweetener aspartame, and after ozonation additionally the chemicalsbenzotriazole and benzothiazole. The number of micropollutants withan average concentration above 100 ng l1 was reduced from 52(out of 70) in the biologically treated wastewater to 30 after bothadvanced treatments.

3.2.1. OzonationThe removal percentages during the ozonation of the 40 micro-

pollutants routinely analyzed are presented in Table 2 (method A)and Fig. S2b, SI.

3.2.1.1. Substances with high ozone reactivity. Twelve substanceswere eliminated to over 90% even with the lowest ozone dose

1 1(2.3 mg O3 l , eq. 0.3 g O3 g DOC), including 4 antibiotics(trimethoprim, clindamycin, sulfamethoxazole and clarithromycin), 2beta-blockers (sotalol and propranolol), 2 anti-inammatory drugs(mefenamic acid and diclofenac), carbamazepine, gembrozil, estroneand bisphenol A. All these compounds contain electron-rich moietiessuch as phenols, anilines, olens or amines (except gembrozil witha benzene derivate), which are known to have high ozone reactiv-ity (second-order rate constant kO3 > 104 M1 s1) (Lee and vonGunten, 2012). The removal of substances with lower reactivity wasmore dependent on the operational conditions, such as the ozonedose (from 2.3 to 9.1 mg O3 l1) and thewastewater quality (presenceof ozone and hydroxyl radical scavengers or competitors, pH, etc.), lead-ing to higher variations in the transformation rate between the differentsampling campaigns.

The macrolide azithromycin showed lower removals (averageof 74%) than expected based on its reported high ozone reactivity(kO3 6 106 M1 s1) (Lee and von Gunten, 2012). One potentialexplanation is the possible sorption (up to 15% in WWTP efuent)of this substance to colloid particles (1 nm to 1 m, considered asbeing part of the aqueous phase) (Maskaoui et al., 2007; Worms et al.,2010), which could protect it against ozone attack (Zimmermann et al.,2011). Potential short-circuiting of a small water fraction through thereactor, which could reduce the exposure to ozone, may also explain theincomplete removal of very reactive substances such as azithromycin,diclofenac or carbamazepine.

3.2.1.2. Substanceswith low ozone but high OH radical reactivity. Ibuprofen,ketoprofen, metronidazole, primidone, mecoprop and benzotriazole,which have low reactivity with ozone (kO3 b 350 M1 s1, Beltrn etal., 1994; Real et al., 2009; Rosal et al., 2010; Zimmermann et al., 2011)showed moderate average removals (around 60%), mainly due to reac-tion with the strong and unselective oxidant OH hydroxyl radical origi-nating from reaction of ozone with the organic wastewater matrix(Huber et al., 2003; Rosal et al., 2010). As the OH radical formation inwastewater is mainly due to the reaction of ozone with specic moietiesof the efuent organic matter (EfOM), variation in the composition ofEfOM, for instance by addition of coagulant, can lead to differentamounts of OH radical formed per unit of ozone (Gonzales et al., 2012;Wert et al., 2011). Moreover, OH radical exposure varies with theconcentration of HO scavengers (such as carbonate) and pH (Bufe etal., 2006). Reactions of micropollutants with OH radicals are thus moreaffected by the quality of the wastewater than direct ozone oxidation(Wert et al., 2011), which could explain the high removal variationobserved for compounds with low ozone reactivity.

3.2.1.3. Substances with low ozone and low OH radical reactivity. Underthe applied ozone doses (average 5.7 mg O3 l1 or 0.8 g O3 g1

DOC), only low removals (average of 34 to 43%) of the iodinatedcontrast media iohexol, iopromide, iomeprol, and iopamidol wereobtained, with particularly low elimination (16%) of diatrizoic andiothalamic acids. Diatrizoate, the anionic form of diatrizoic acid, isone of the most ozone-resistant pharmaceuticals, having, as othercontrast media, very low ozone reactivity, but also a low OH radicalreactivity (Huber et al., 2005; Real et al., 2009). Low removals of atra-zine (34%), gabapentin (38%), irgarol (32%) and propiconazole (32%)were also observed. Atrazine is reported to have low reactivity withozone andOH radicals (Acero et al., 2000). Poor oxidation of gabapentinwas also obtained in other studies (Reungoat et al., 2010). Low removalsof the pesticides irgarol and propiconazole during ozonation were alsoobserved by Bundschuh et al. (2011), but for irgarol, higher removalswere reported by Hollender et al. (2009). Irgarol is expected to havelow ozone reactivity due to its triazine ring, which is very resistantto oxidation (Chen et al., 2008). The very low concentration of thissubstance inWWTP efuent (2 to 17 ng l1), close to the limit of quan-tication, leads however to high analytical uncertainties and could bethe cause of the divergences observed. Resistance to ozonation is partic-

ularly of concern for the contrast media and gabapentin due to their

high concentration in wastewater (above 1 g l1) and their persis-tence even for efcient biological treatment.

3.2.1.4. Removal efciency with higher ozone doses. A higher ozonedose, 17.6 mg O3 l1, equivalent to 2.6 g O3 g1 DOC, was testedduring one campaign (data not illustrated). At this dosage, much betterremoval of the recalcitrant micropollutants was found, with 88%gabapentin elimination, 66% atrazine, and 84, 82 and 81%, respectively,of iopamidol, iohexol and iomeprol. Higher doses lead however tohigher costs and a higher risk of forming bromate, a toxic by-product(see below Formation of toxic oxidation by-products), and thereforewere not further tested.

be explained by the protective effect of the hydroxyl or acetyl group

3.2.1.7. Relation between ozone dose and micropollutant removal rate.Efuent organic matter containing electron-rich organic moietiesand nitrite react rapidly with ozone, contributing to the ozone de-mand with 0.20.6 mg O3 mg1 C and 3.4 mg O3 mg1 N-NO2 re-spectively (Wert et al., 2009, 2011). Thus, in order to have enoughresidual ozone for the oxidation of micropollutants and to assure a suf-cient and relatively constant ozone exposure, the ozone dosage wasregulated to maintain a constant ozone residual concentration nearthe end of the reactor. During the campaigns, the dosage varied from2.3 to 9.1 mg O3 l1 depending mainly on the DOC (0.38 g O3 g1 C)and NO2 (3.4 g O3 g1 N) concentrations (Fig. S5, SI), but also due tothe residence time of the water in the reactor and the choice of thechamber in which ozone was injected. No clear relation betweenozone dose (in mg O3 l1) and micropollutant transformation rate

Log1

tiv

Log Dow [-]

lutaatio

488 J. Margot et al. / Science of the Total Environment 461462 (2013) 480498Fig. 2. Removal of 35 micropollutants with PAC-UF treatment as a function of micropolcharged, and (c) neutral. Median removal of eight 4872 h composite samples. Correl-1 0

b Nega

-2 -1 0 1 2 3

Rem

oval

with

PAC

[%]

0

20

40

60

80

100a Positively chargedon the reactive moiety, which changes the electron density and thusslows down the reaction (Huber et al., 2005).

3.2.1.6. Inuence of the pH on the oxidation process. Reactivity of a sub-strate with ozone is strongly inuenced (up to 4 orders of magnitude)by the protonation of the reactive amine or phenol (Lee and vonGunten, 2012). Dissociated moieties have a higher electron densityand thus are more reactive towards ozone (Lee and von Gunten,2012). Due to their two pKa values close to the pH of wastewater(Table S1, SI), the reactivity of uoroquinolone antibiotics is particu-larly susceptible to pH variations. The variations of pH measured inthe wastewater, from 6.3 to 8, can thus increase the reactivity ofciprooxacin, noroxacin and ooxacin by 1 or 2 orders of magnitude(Dodd et al., 2006), explaining partially the high variation in the re-moval rates of these compounds during the different campaigns.This assumption is supported by the signicant positive correlationsobserved between the pH and the removal rate of these three com-pounds (Fig. S4, SI).3.2.1.5. Removal of other micropollutants and human drug metabolitesmeasured in a screening campaign. Table 2 presents also the removalof 23 other micropollutants (analytical method B), which were ana-lyzed only once on a 7-d composite sample (with 6 mg O3 l1). Halfof them were removed at a rate of over 70%. One can notice howeverthe lower efciency (b51% removal) of ozone for the antihypertensiveirbesartan, the anticonvulsant levetiracetam, the anxiolytic oxazepamand the insect repellent DEET. These substances contain amide func-tions that exhibit low reactivity with ozone (Lee and von Gunten,2012). The human pharmaceutical metabolites, which are mainlyhydroxylated, hydrolyzed or conjugated forms of the parent compound(Ikehata et al., 2006), were mostly not as well removed as the parentcompound. This is especially the case for the 10,11-dihydro-10,11-dihydroxy carbamazepine and N-acetyl sulfamethoxazole, with anelimination of only 50% compared to >90% for carbamazepine andsulfamethoxazole. The lower ozone reactivity of the metabolites cancharged compounds and zwitterions, and signicant for negatively charged (r = 0.743, p =was evident. However, when the ozone dose was normalized by theconcentration of scavenger equivalent, a weighted sum of DOC andNO2 concentrations (3.4 [N-NO2] + 0.38 [DOC]), higher doses (in g O3g1 scavenger equivalent) tended to lead to higher removal ratesfor most micropollutants (Fig. S6, SI). An average ozone dose of5.7 mg O3 l1, corresponding to 1.6 g O3 g1 scavenger equivalentor around 0.85 g O3 g1 DOC in the case of 0.3 mg N-NO2 l1, wassufcient to achieve an average reduction of 80% of the 65 studiedmicropollutants in the WWTP (compared with raw wastewater).

3.2.1.8. Effect of the sand lter on micropollutant removal. The sandlterfollowing the ozonation had only a limited effect on micropollutantremoval, with a slight improvement in the average removal of 36compounds from 73.2% for ozone alone to 75.8% for ozone combinedwith the sand lter. Higher removals (>10%) were observed mainlyfor compounds thatwerewell eliminated in an efcient biological treat-ment, such as ibuprofen, metronidazole and ciprooxacin, and for twopesticides carbendazim and propiconazole, possibly due to sorptionon the biolm (Fig. S7, SI).

3.2.1.9. Formation of toxic oxidation by-products. Formation of toxicoxidation by-products can occur during ozonation of wastewater,such as carcinogenic bromate, nitrosamines or formaldehyde (Wert etal., 2007; Zimmermann et al., 2011). High concentrations of bromide(350 g l1) measured in a 7-d composite sample in the wastewatersuggested that excessive bromate formation could occur during ozona-tion (von Gunten, 2003b). The concentration of bromate was below thedetection limit (1 g l1) in the efuents of the biological treatmentand PAC-UF. After ozonation (6 mg O3 l1, equal to 0.8 g O3 g1

DOC) and sand ltration, the bromate concentration increased to 3.7and 5.1 g l1 respectively. These concentrations remained howeverbelow the Swiss drinking water standard of 10 g l1 (OSEC, 1995)and far below the proposed ecotoxicologically relevant concentrationof 3 mg l1 (Hutchinson et al., 1997). The formation of bromate wasdependent on the ozone dose applied, exceeding the drinking waterstandard for an ozone dose above 1.4 mg O3 mg1 DOC (7 mg O3 l1,

Dow [-]2 3 4

ely charged

Log Dow [-]-4 -2 0 2 4

c Neutral

nt hydrophobicity (log Dow) and charge at pH 7. (a) Positively charged, (b) negativelyn r between PAC removal and log Dow not signicant (p-value > 0.05) for positively

0.014) and neutral compounds (r = 0.648, p = 0.005).

with 70 g l1 bromide), as shown in a laboratory scale experiment(Fig. S8, SI). Unlike nitrosamines that can be partially removed in asand lter (Hollender et al., 2009), the bromate concentration was not

InfluentBIO OZ SF PAC-UFPho

tosy

nthe

sis in

hibi

tion

[ng l-

1 di

uron

equ

ivale

nt]

0

100

200

300

400

InfluentBIO

Alga

e gr

owth

inhi

bitio

n[m

g l-1

toxi

c eq

uiva

lent

]

0

10

20

30

a b

Fig. 3. (a) Inhibition of photosynthetic activity (diuron-equivalent concentration) and (b) inhsubcapitata. (c) Estrogenic activity (YES, estradiol-equivalent concentration). Average results (and in the efuents of the biological treatment (BIO), the ozonation (OZ), the sand lter6.7 mg O3 l1 (eq. 0.76, 0.91, 0.92 g O3 g1 DOC), and PAC doses of 10, 12 and 20 mg l1 for

489J. Margot et al. / Science of the Total Environment 461462 (2013) 480498reduced during the sand ltration, and therefore a high ozone doseshould be avoided to ensure lowbromate concentrations in the efuent.

3.2.2. Powdered activated carbon treatmentThe removal percentage during the PAC-UF treatment of the 40

micropollutants routinely analyzed is presented in Table 2 (methodA) and Fig. S2c, SI. High variations in the removal rate, especiallyfor compounds with lower PAC afnity, were observed among the dif-ferent campaigns. Indeed, to optimize the treatment the PAC dosewas increased from 10 to 20 mg l1 during the study. Moreover,the DOC concentration in the feed water was not constant, leadingto variable competition for the adsorption sites between EfOM andmicropollutants. As the type of PAC (Norit SAE SUPER and SORBOPORMV-125) did not signicantly inuence the removal rate compared toother variables, results are presented for both PAC types together.

3.2.2.1. Substances with high PAC afnity. Seven substances wereremoved at a rate of more than 90% in almost all the campaigns,including the beta-blockers propranolol and metoprolol, as wellas methylbenzotriazole, trimethoprim, mefenamic acid, estroneand carbendazim. In 50% of the campaigns, over 90% of the follow-ing compounds were removed as well: clarithromycin, carbamaze-pine, benzotriazole, ooxacin, noroxacin and atenolol. These 13

ition

400Photosynthesis inhibitors concentration [ng l-1 DEQ]

0 50 100 150 200 250

Phot

osyn

thes

is in

hib

[ng l-1

D

EQ]

0

100

200

300

InfluentBiologyOzonationSand filterPAC-UF

Fig. 4. Comparison of the green algae photosynthesis inhibition (in diuron-concentrationequivalent DEQ) with the sum of the wastewater concentrations of the four mostabundant photosynthesis inhibitors included in the analytical list (atrazine, diuron,isoproturon, and terbutryn), converted to DEQ based on their relative potency. Resultsof 19 analyses on 7-d composite samples taken after the different treatments. Dashedline: linear regression.micropollutants have a very good afnity for PAC, with high elimina-tion rates even with 10 mg PAC l1. Apart from the hydrophobicmefenamic acid, all those compounds were either positively charged(ve substances) or neutral (seven substances) at the pH of the waste-water, covering a broad range of hydrophobicity (log Dow from 1.3to 3.7).

3.2.2.2. Substances with medium PAC afnity. A second group of 15substances (from metronidazole to azithromycin in Fig. S2c, SI) had,on average, between 70 and 90% removal, including six neutral andsix negatively charged compounds. The medium PAC afnity fordiclofenac and gembrozil was reported elsewhere (Snyder et al.,2007; Westerhoff et al., 2005), but better removal of ibuprofen wasobserved in our case, either due to different PAC characteristics orto biodegradation phenomena in the reactor.

3.2.2.3. Substances with variable or low PAC afnity. The 12 remainingsubstances (from sulfamethoxazole to diatrizoic acid in Fig. S2c, SI),composed of neutral or negatively charged compounds (includingall the hydrophilic contrast media), showed poor or very variableafnity for PAC with an average removal between 11 and 66%. Thehigh removal variation observed for sulfamethoxazole, ciprooxa-cin, mecoprop, primidone and the contrast media was partly due tothe different PAC doses applied, with increasing removal when thedose increased from 10 to 20 mg l1. High variations (b20% to>60% removal) occurred also within the same PAC dose (mainly at10 mg l1), which could not be explained by the different parame-ters monitored (water quality and operational parameters such asresidence time, PAC concentration, PAC type, etc.). These high varia-tions may be due to different EfOM contents and compositions,

OZ SF PAC-UF InfluentBIO OZ SF PAC-UF

Estro

geni

c ac

tivity

[ng l-1

est

radi

ol e

quiva

lent

]

0

5

1060

80

100

c

ibition of growth (toxic-equivalent concentration) of the green algae Pseudokirchneriellastandard deviation) of three campaigns of one week in the raw wastewater (inuent)following the ozonation (SF) and the PAC-UF treatment. Ozone doses of 3.5, 6.0 and, respectively, campaigns 1, 2 and 3.as discussed below.The anionic contrast media diatrizoic and iothalamic acids and the

anticonvulsant gabapentin showed less than 20% removal by PAC-UF.The low PAC afnity of these hydrophilic (log Dow of 1.2 to 0.4)and charged substances was reported by Reungoat et al. (2010) andBoehler et al. (2012). Low adsorption of gabapentin could be causedby the absence of an aromatic ring (de Ridder et al., 2010). The vari-able elimination of irgarol (0% to >60%), despite its hydrophobicity(log Dow of 4), is probably due to its very low concentration inthe feed water, leading to high uncertainties in estimates of theremoval rate.

3.2.2.4. Removal efciency with higher PAC dose. A higher PAC dose of60 mg l1 was tested during one campaign, leading to more than 90%removal of substances with a low PAC afnity (e.g., sulfamethoxazole,mecoprop, primidone and the contrast media iohexol, iomeprol andiopromide). Even this high dose was unable to remove gabapentin

satisfactorily (56% removal, data not illustrated). Higher doses of PAClead however to higher costs and larger amounts of sludge produced.

3.2.2.5. Removal of other micropollutants and human drug metabolitesmeasured in a screening campaign. Table 2 shows the removal of 24other micropollutants (analytical method B), which were analyzedonce on a 7-d composite sample (12 mg PAC l1). About half of themwere removed at a rate of over 70%. We observe, however, a lower ef-ciency (b60% removal) for most of the human pharmaceutical metabo-lites. Indeed, pharmaceutical compounds are usually transformed in the

liver or kidney tomore polar and hydrophilic metabolites in order to bereadily excreted in the urine or bile (Ikehata et al., 2006). For instance,the metabolite 10,11-dihydro-10,11-dihydroxycarbamazepine has alog Kow of 0.13 compared to 2.45 for the parent compound carbamaze-pine (Miao et al., 2005). Therefore, due to the low hydrophobicity ofhuman metabolites, a lower PAC afnity is expected. The low removalof the UV lter oxybenzone and the antidepressant venlafaxine is notexplained, however, given the good PAC afnity for those substancesreported in the literature (Reungoat et al., 2012; Snyder et al., 2007).

3.2.2.6. Possible inuence of efuent organic matter on removal efciency.The adsorption process in complex matrix is not yet fully understoodand can be inuenced by many parameters, the main one being thecompetitive effect of the EfOM, either by direct competition for the ad-sorption sites or by pore blockage/constriction (Delgado et al., 2012).EfOM characteristics, mainly the concentration of low molecularweight and hydrophobic molecules, determine the competitivenessof the organic matter (de Ridder et al., 2011; Newcombe et al.,1997). Variation in the concentration and composition of the EfOM,due to different treatments of the wastewater (biodegradation,chemical coagulation, etc.) can thus lead to different micropollutantremoval rates at the same PAC dose. This issue was investigatedwith laboratory batch adsorption experiments. Five micropollutantsin Lausanne wastewater treated to different levels (coagulation/precipitation, biological treatment without nitrication or with fullnitrication) were examined. A strong inuence of the feed waterDOC (from 5 to 17 mg l1) on the substance removal with PAC wasobserved for all the compounds (Fig. S9, SI), conrming the highcompetitive effect of EfOM for the adsorption sites. The highestPAC efciency was observed in the efuent of the biological treat-ment with full nitrication (DOC of 5 mg l1), signicantly higherthan in wastewater coming from a treatment without nitrication

enin

con

cent

ratio

n[ng

/ml]

40

60

80

100

Control BIO OZ SF PAC-UF

Indi

vidua

l leng

th[m

m]

0

10

20

30

40

Control BIO OZ SF PAC-UF

Indi

vidua

l fre

sh w

eigh

t[m

g]

0

100

200

300

400

Control BIO OZ SF PAC-UF

Ove

rall s

urviv

al[%

of eg

gs in

trodu

ced]

0

20

40

60

80

100a

b

c

d

***

*** *** ***

*** *** ***

490 J. Margot et al. / Science of the Total Environment 461462 (2013) 480498Control BIO OZ SF PAC-UF

Vite

llog

0

20***

Fig. 5. Results of the sh early life stage test (FELST) with (a) the overall survival (averageof three replicates per treatment), (b) the individual fresh weight (average of 69 to 152sh per treatment), (c) the individual length (average of 69 to 152 sh per treatment)and (d) the vitellogenin concentration (average of 20sh per treatment) of the sh larvaeat the end of the test (after 69 d). Signicant differenceswith the controls are representedby * (p value b 0.05), ** (p b 0.01), *** (p b 0.001). All the endpoints for the control, OZ,SF and PAC-UF were signicantly different from the endpoints of BIO. Ozone dose:

1 14.7 1.5 mg O3 l . PAC dose: 13.1 2.6 mg l .(DOC of 11 mg l1). Wastewater treated only with coagulation/

Time [d]26 28 30 32 34 36 38 40 42 44 46 48 50

Hat

chin

g su

cces

s[%

of in

trodu

ced e

ggs]

0

20

40

60

80

100

ControlBiologyOzonationSand filterPAC-UF

Time [d]50 52 54 56 58 60 62 64 66 68 70

Swim

-up

[% of

hatch

ed eg

gs]

0

20

40

60

80

100b

a

Fig. 6. Hatching success (a) and swim-up (b) of the eggs and larvae of the rainbowtrout in the efuent of the different treatments. Average and standard deviation of 3

1 1replicates. Ozone dose: 4.7 1.5 mg O3 l . PAC dose: 13.1 2.6 mg l .

with less than 13 mg TSS l1 in the efuent. Similar micropollutantremoval rates were measured with both separation systems (UF andsand lter), on average around 80%, indicating that the PAC, and notthe ultraltration, was responsible for micropollutant removal.

3.3. Ecotoxicological evaluation

In addition to chemical analysis, the results of bioassays providedinformation on potential effects of the mixture of compounds. Bothadvanced treatments were able to reduce signicantly the toxicityof the biological treatment efuent, both in bioassays with algae onenriched samples and in a chronic test on sh with continuous expo-sure to the raw efuent.

3.3.1. Combined Algae Assay on enriched samples

491J. Margot et al. / Science of the Total Environment 461462 (2013) 480498precipitation (DOC of 17 mg l1) led to a strong reduction of the PACadsorption capacity, probably due to the persistence of smaller biode-gradable molecules. Thus, different degrees of secondary treatment canlead to variable adsorption rates. Consequently, the PAC dose necessaryto achieve an average overall micropollutant removal above 80%(whole treatment) in wastewater with a DOC of 5 to 10 mg l1, wasvariable: 10 mg l1 was sometimes sufcient but in most cases20 mg l1 was required. These minimum doses were noted in otherstudies as well (Boehler et al., 2012; Nowotny et al., 2007).

3.2.2.7. Role of electrostatic and hydrophobic interactions in the adsorptionprocess. Electrostatic and hydrophobic interactions seem to play animportant role in the adsorption process. As presented in Fig. 2, onaverage more than 80% (most more than 90%) of all the positivelycharged molecules were removed, independently of their hydropho-bicity. Only the large molecule azithromycin, diprotonated at pH 7,was eliminated to a lower extent despite its higher hydrophobicity,possibly by size exclusion in the micropores of the PAC (Ji et al.,2010). The removal of the negatively charged and neutral substanceswas more dependent of their hydrophobicity, the most hydrophiliccompounds being eliminated to a lesser extent. For the same log Dow,neutral and especially negatively charged compounds were on averageless adsorbed than those that were positively charged.

The two PACs studied have a point of zero charge pHPZC > 7.3,thus the fresh PAC is expected to be neutral or slightly positivelycharged at the pH tested. However, in wastewater, the adsorption ofEfOM, negatively charged at neutral pH, leads to a decrease in thePAC pHPZC due to the EfOM coverage, resulting from a net negativesurface charge on the loaded PAC (Newcombe, 1994; Yu et al.,2012). As both EfOM and micropollutant adsorption occurred simul-taneously, electrostatic attraction between the cationic compoundsand the negatively charged surface of the loaded PAC is expected,even for hydrophilic substances. Conversely, charge repulsion shouldoccur for the anionic substances. These electrostatic repulsions can beoffset by hydrophobic partitioning (expulsion in the solute-watersystem) at high log Dow. Thus, in wastewater, hydrophobic interac-tion is expected to be more signicant for negatively charged andneutral compounds than for positively charged substances, as ob-served in our results. This assumption was tested in batch tests byde Ridder et al. (2011) where very similar behavior was observed,conrming that both log Dow and charge interaction have a signicantinuence on micropollutant adsorption in wastewater. But, for neu-tral or negatively charged substances, log Dow was not by itselfsufcient to explain the observed removals. Although hydrophobicpartitioning has been reported as the dominant mechanism leadingto PAC adsorption for compounds with log Dow > 3.7, other adsorptionmechanisms such as hydrogen bond formation and pipi interactionbetween micropollutants and the PAC surface have been reported tobe more prominent as log Dow decreases (de Ridder et al., 2010).Thus, for hydrophilic compounds with the same log Dow, very differentPAC afnities can be expected depending of the characteristics of themolecules.

3.2.2.8. Separation of PAC with ultra- or sand ltration inuence onmicropollutant removal. As observed in other studies (Snyder et al.,2007; Yoon et al., 2007), the inuence of ultraltration on the re-moval of hydrophilic micropollutants (log Kow b 2.8) is expected tobe negligible due to the relatively high molecular weight cut-off ofthe membrane (100300 kDa) compared to the molecular mass ofmicropollutants (b1 kDa). For more hydrophobic compounds, signif-icant adsorption on the membranes can occur (Yoon et al., 2007),but at a much lower level than on PAC. Therefore, PAC adsorption isconsidered to be by far the main removal process in the PAC-UFsystem. To check this assumption and to evaluate another (cheaper)separation system, a sand lter was used instead of the UF membrane

during seven campaigns. Good PAC retention (>90%) was observed3.3.1.1. Photosynthesis inhibition. As presented in Fig. 3a, raw waste-water induced photosynthesis inhibition equivalent to 253 92 ng l1

of diuron. This specic effect of substances acting on the photosys-tem II (Escher et al., 2008a) was not strongly reduced during thebiological treatment (14 37%, 228 155 ng DEQ l1), suggestinglow biodegradability of these compounds. However, both advancedtreatments led to a clear decrease in this effect with 82 8% re-moval (32 9 ng DEQ l1) during ozonation and 87 11% removal(18 11 ng DEQ l1) during PAC-UF treatment. The residual toxicitywas signicantly lower (p b 0.05) after PAC-UF compared to OZ in cam-paigns 1 and 3 (no signicant difference in campaign 2). Photosynthesisinhibition was not signicantly reduced after the sand lter followingozonation (27 5 ng DEQ l1), presumably due to the lowbiodegrad-ability of those compounds. The overall removal in the WWTP was87 4% with ozonation followed by sand ltration and 92 9% withPAC-UF treatment, showing the ability of these two treatments toimprove the quality of the WWTP efuent.

The herbicides atrazine, diuron and isoproturon, and the algicideterbutryn act as photosystem II inhibitors in plants and algae andcan have a cumulative effect when present in a mixture (Brust et al.,2001; Knauert et al., 2010; Nystrm et al., 2002). A clear relation(correlation r = 0.909, p b 0.001) between inhibition of the photo-system II and the concentration of relevant pesticides measured inthe samples was observed (Fig. 4). The sum of the relative potencyof these four compounds, expressed as diuron equivalents, couldexplain, on average, 56% of the total inhibition observed. The other(unmeasured) compounds participating in the remaining photosyn-thesis inhibition are expected to be eliminated to the same extentas these four inhibitors. Indeed, a reduction of the concentrations ofthese inhibitors in advanced treatments led to a similar reductionof the photosynthesis inhibition. Similar effects were observed for

Table 3Costs and energy needs for construction and operation of the pilot plants. Costs are givenexcluding VAT, based on local (Swiss) prices in 2010 (0.17 kWh1 of electricity,0.25 Nm3 O2, 2 kg1 PAC, 66 h1 staff costs) for an average removal of 80% ofthe 65 studied micropollutants (compared to raw wastewater). Investment costs arecalculated with an interest rate of 4.5% y1, with amortization periods of 10, 20 and30 y for, respectively, electromechanical, mechanical and structural equipment.

Dosage Ozonationwith sand lter

PAC withsand lter

PAC withultraltration

5.7 mg O3 l1 15 mg PAC l1 15 mg PAC l1

Capacity(average ow)

[ls1] 60 15 5

Electricityconsumption

[kWhm3] 0.117 0.08 0.9

Operating costs [m3] 0.043 0.054 0.404Investment costs [m3] 0.133 0.107 0.399Total costs(excluding VAT)

[m3] 0.176 0.161 0.803

ozonation in a previous study at the Regensdorf WWTP, Switzerland(Escher et al., 2009).

3.3.1.2. Algae growth inhibition. A relatively high algae growth inhibitionwas observed in the rawwastewater (Fig. 3b), with a non-specic toxic-ity of 26 7.3 mg l1 (baseline toxic equivalent concentration, Escheret al., 2008a). This was clearly reduced (73 6%, 6.9 1 mg l1)during the biological treatment. This non-specic toxicity, contrary tothe photosynthesis inhibition, can thus be partially attributed to biode-gradable or adsorbable compounds that were removed in this treat-ment. The advanced treatments were able to reduce the residualtoxicity (attributed to non-readily biodegradable micropollutants) by75 7% during ozonation (1.67 0.45 mg l1) and 84 5% duringPAC-UF treatment (1.07 0.17 mg l1). This toxicity was signicantlylower after PAC-UF compared to OZ in campaigns 2 and 3 (no signicantdifference in campaign 1). The sand ltration following ozonation wasalso able to reduce the growth inhibition from 10 to 46% (mean:1.28 0.16 mg l1), the highest improvement being observed whenthe biological treatment was not effective, meaning that biodegradabletoxic compounds remained in the ozone efuent. This resulted in amean overall elimination (compared to WWTP inuent) of 96 1%with ozonation followed by sand ltration and 97 0.1% with PAC-UFtreatment. In a comparable study, Escher et al. (2009) detected a highermaximum reduction of non-specic toxicity during biological treatment(7099.5%) at the RegensdorfWWTP and a subsequently lower removalefciency during ozonation.

3.3.2. Estrogenic activity on enriched samplesHigh estrogenic activity was detected with the YES in rawwastewa-

however, be sufcient to affect the fertility of sensitive sh species(Lahnsteiner et al., 2006), as shown also with the sh test (cf. 3.3.3).Estrogenic activity was further signicantly diminished by 89 4%during ozonation and 77 17%with PAC-UF,which is similar to resultsobtained by Stalter et al. (2011) and Escher et al. (2009). This resulted ina mean overall elimination (compared to WWTP inuent) of 99 1%with both advanced treatments. The residual estrogenicity observedin the efuents, signicantly lower after OZ (0.10.65 ng l1 EEQ)than after PAC-UF (0.291.32 ng l1 EEQ) in campaigns 1 and 2(no signicant difference in campaign 3), was in most cases belowthe environmental quality standard of 0.4 ng l1 proposed for17--estradiol (Kase et al., 2011). Therefore, advanced treatmentsor biological treatment with full nitrication is efcient means toreduce the release of endocrine compounds, and thus to reduce therisk of feminization of sh and mussel populations. As the estrogenicactivity was already very low after the ozonation, there was no im-provement due to the sand lter. During one campaign an increasein estrogenicity was observed, presumably due to contamination ofthe new sand by estrogenic compounds. Indeed, an unexplainedincrease in bisphenol A concentration was measured after the sandlter for this case.

3.3.3. Fish early life stage toxicityBoth advanced treatments signicantly decreased the toxicity of the

WWTP efuent on the development of rainbow trout embryos for allendpoints measured: the overall survival of the sh, the hatching suc-cess, the swim up, the individual development (weight and size) andthe induction of estrogenic effects.

me

prom

eco

PA

(do

492 J. Margot et al. / Science of the Total Environment 461462 (2013) 480498ter (37100 ng l1 estradiol equivalents, EEQ), whichwas then strong-ly reduced (88 10%) during the biological treatment (Fig. 3c). Theremoval of estrogenic activity was dependent on the level of nitrica-tion, from 75% without nitrication to 99% with full nitrication(b1 mg N-NH4 l1) (Fig. S10, SI). The low estrogenicity levelmeasuredin the efuent of the biological treatment (0.78.3 ng l1 EEQ) could,

0 20 400

20

40

60

80

100

Gabapentin

Sulfa

Diatrizoic + iothalamic acid

Iopamidol

Io

M

Irgarol

Removal with

Rem

oval

with

ozo

ne [%

]

Fig. 7. Comparison of the average removal of 40 micropollutants with PAC-UF treatment1 1median 5.9 mg O3 l or 0.83 g O3 g DOC) during one year of operation (3 to 28 analyse3.3.3.1. Overall survival. The overall survival of the rainbow trout after69 d of continuous exposure in the efuent of the biological treatment(BIO) was relatively low, with only 58 6.6% survival (Fig. 5a). Thesurvival was signicantly improved after the ozonation (OZ) (85 6.6% survival) and the PAC-UF treatment (93 3.8%), reaching a levelstatistically similar to the control (95 2% survival). The subsequent

60 80 100

Bezafibrate

Gemfibrozil CarbamazepineDiclofenac

IbuprofenKetoprofen

Mefenamic acid

Naproxen

Primidone

AtenololMetoprolol

PropranololSotalol

Azithromycin

Ciprofloxacin

Clarithromycin

Clindamycin

Metronidazole

Norfloxacin

Ofloxacinthoxazole

Trimethoprim

Iohexol

Iomeprol

ide

Benzotriazole

Methylbenzotriazole

Atrazine

DiuronIsoproturon

prop

Carbendazim

Propiconazole

Terbutryn

EstroneBisphenol A

C-UF [%]

se of 1020 mg PAC l1, median 12 mg l1), or ozonation (dose of 2.39.1 mg O3 l1,

s depending on the substance).

493J. Margot et al. / Science of the Total Environment 461462 (2013) 480498sand ltration (SF) step did not improve the survival of the shcompared to the ozonation alone.

3.3.3.2. Hatching success. The hatching success of the sh reached 80 5% in BIO efuent, which was signicantly lower than in the control(100% success). Both advanced treatments improved the hatching suc-cess to a level statistically similar to the control, with 97 3.8% for OZ,98 2.9% for SF and 100 0% for PAC-UF. However, the hatchingprogress was on average delayed for 2 d in OZ and SF efuents com-pared to PAC-UF or the control, and delayed for one week in the BIOefuent (Fig. 6a). Delay in hatching after ozonation was also observedby Stalter et al. (2010b), and not notable in the sand lter efuent,as discussed below.

3.3.3.3. Swim-up. The swim-up, which is the developmental transitionfrom larval stage to juvenile sh stage, appeared after 60 d in BIOefuent, delayed by 8 d compared to the control (Fig. 6b). Both ad-vanced treatments reduced the delay in the swim-up. The beginningof the swim-up appeared simultaneously in PAC-UF efuent and inthe control, but was delayed by 3 d compared to the control in OZand SF efuents. A notable delay in the swim-up was also observedafter ozonation by Stalter et al. (2010b), possibly due, in their case,to the presence of toxic oxidation by-products. 28% of the sh diedduring the larvae stage in BIO efuent, with only 45 9% of thelarvae reaching the juvenile stage at the end of the test. This wasmuch improved after the advanced treatments, with 93.3 3.8% ofthe larvae in PAC-UF efuent, 88.2 10% in OZ, and 85.5 10.4%in SF that swam up, showing no signicant difference with the control(93.1 3.1%).

3.3.3.4. Weight and length of the sh. Weight and length of the sh atthe end of the test was relatively low in BIO efuent and increasedsignicantly after the advanced treatments. Those parameters werehowever still signicantly lower in OZ and SF efuents compared tothe control, while no difference was observed in PAC-UF efuent(Fig. 5b and c). The sh were on average 6.7% longer and 22% heavierin PAC-UF efuent than in OZ or SF efuents, and 32% longer andtwice as heavy as in BIO efuent. The sand lter did not improvegrowth of the sh compared to the ozonation alone.

3.3.3.5. Vitellogenin concentration. The vitellogenin (VTG) concen-tration in the juvenile sh was signicantly higher in the BIO efu-ent (63.1 33.2 ng ml1) compared to the sh in the control(10.6 4.7 ng ml1) (Fig. 5d). Similar VTG concentrations (67.3 26.9 ng ml1) were found by Stalter et al. (2010b) in juvenile rainbowtrout exposed to secondary efuent. VTG, an egg yolk precursor nor-mally produced by mature female sh, can be used as a biomarker forexposure to exogenous estrogens for juvenile and male sh (Jobling etal., 2006; Thorpe et al., 2000). The increase of VTG content in juvenilesh in BIO efuent indicates the presence of environmentally-relevantconcentrations of estrogenic compounds. This effect was not observedafter both advanced treatments, the VTG content in the sh beingon par with the control in PAC-UF (10.2 5.8 ng ml1), OZ (9.9 7.1 ng ml1) and SF efuent (14.1 9.1 ng ml1). These results con-rm the ability of ozonation and PAC-UF to eliminate the estrogenicityin wastewater, as presented in Fig. 3c. The minor increase of VTG inthe sh exposed to SF efuent compared to OZ efuent, also observedin the YES, is probably due to contamination of the new sand by endo-crine active compounds.

3.3.3.6. Toxicity of the biologically treated efuent and possible inuenceof nitrite and ammonia. As presented above, the efuent of the biolog-ical treatment impaired the survival and the development of rainbowtrout, delaying their swim-up and their growth as expressed by lowerbiomass and body length. Besides the mortality observed (43%),

a delay in the development can, for instance, increase the risk forpredation in natural systems since larvae are unable to escape beforethe swim-up (Stalter et al., 2010b). Moreover, changes in VTG con-centrations in sh can be an indicator for an effect on their reproduc-tion system (Miller et al., 2007; Thorpe et al., 2007). Therefore,efuents from conventional WWTPs can have a signicant impacton salmonid sh in natural environments in the case of low dilu-tion of the efuent. Besides the estrogenic substances and othermicropollutants present in the efuent, macropollutants such asnitrite and ammonia or bacterial contamination could also affectthe sh.

Rainbow trout are sensitive to nitrite (NO2), with lowergrowth rates observed at 0.3 mg N-NO2 l1 and 65% mortality at0.91 mg N-NO2 l1 (with 10 mg Cl l1) (Kroupova et al., 2008).The toxicity can, however, be strongly inhibited by chloride ions(Lewis and Morris, 1986). The relatively high concentration of chlo-ride in the investigated wastewater (80170 mg l1) could havetherefore drastically reduced (up to a factor of 10) the toxic effectof nitrite. In the present study, NO2 concentrations varied between0.04 and 0.55 mg N-NO2 l1 in BIO and OZ efuents and around0.22 mg N-NO2 l1 in PAC-UF efuent. Those concentrations arevery unlikely to have induced signicant lethal and sub-lethal effectson the sh.

Embryos and larvae of rainbow trout are additionally very sensitiveto ammonia (NH3), the unionized form of ammonium NH4+. Sub-lethaleffects such as a decrease in the larvae weight were observed after20 d of exposure at 0.006 to 0.18 mg N-NH3 l1 (Vosylien andKazlauskien, 2004) and a delay in development to the swim-up stageappeared at concentrations above 0.01 mg N-NH3 l1 (Brinkmanet al., 2009). Lethal effects were reported for concentrations above0.022 to 0.13 mg N-NH3 l1 (Brinkman et al., 2009; Solb andShurben, 1989). In the present study, the concentrations of union-ized ammonia, calculated according to Armstrong et al. (2012),were relatively high in the BIO, OZ and SF efuents, varying between0.02 and 0.06 mg N-NH3 l1 during the rst 10 d, decreasingthen below 0.01 mg N-NH3 l1 in all efuents until the end of thetest. The NH3 concentration in the PAC-UF efuent was alwaysb0.01 mg N-NH3 l1 due to further nitrication in the reactor. Am-monia concentrations in BIO, OZ and SF efuents at the beginningof the test were therefore high enough to induce sub-lethal effectsand even mortality. Ammonia could be thus partly responsible forthe lower weight and length of the sh exposed to OZ and SF efu-ents, as well as for their delay in reaching the swim-up stage. Theclear impact on sh development and the high mortality observedin the BIO efuent is, however, not attributable to ammonia toxicityalone as much smaller impacts and mortality rates were observedwith the same ammonia concentration in OZ efuent. Therefore,the toxicity observed in the BIO efuent can presumably be relatedto compounds oxidized during ozonation, such as pharmaceuticalsand pesticides. This demonstrates that several compounds inuenc-ing rainbow trout development and survival in the BIO efuent wereremoved in the advanced treatment. Ozonation and activated carbonare therefore efcient techniques to reduce effects of micropollutantson sh.

3.3.3.7. Ozonation inuence on sh toxicity. Stalter et al. (2010b)reported that sh toxicity increased during the ozonation process,probably due to the formation of labile oxidative by-products suchas toxic aldehydes or metabolites. These adverse effects were reducedafter the sand ltration, probably due to biodegradation or spontane-ous degradation of the reactive products. Unlike Stalter et al. (2010b),in our case ozonation clearly reduced the sh toxicity compared tothe BIO efuent to a level close to the control. Moreover, the sandlter did not affect the residual toxicity of the OZ efuent. These con-tradictory results are likely due to different ozone reactor congura-tions and/or different water compositions. Indeed, the reactor used

in Stalter et al. (2010b) contained 3 chambers with an HRT of only 3

494 J. Margot et al. / Science of the Total Environment 461462 (2013) 480498to 15 min (Zimmermann et al., 2011), risking release of toxic reactiveproducts or even residual ozone in the efuent. In our case, the fourthlarge contact chamber (Fig. 1a) ensured complete reaction of ozoneand reactive products within the reactor, with an overall hydraulicresidence time (HRT) of 20 to 60 min depending on the ow.

3.4. Costs and energy needs

The costs of the construction and the operation of the pilot plantsare presented in Table 3 for an average micropollutant removal of 80%compared to raw wastewater. An average ozone dose around5.7 mg O3 l1 and a PAC dose around 15 mg l1 (between 10 and20 mg l1) were needed to reach this level, remembering that thedoses required varied according to the feed water quality. Althoughsome substances were poorly eliminated with those doses, an aver-age removal of 80%, as recommended by Swiss authorities, is a goodcompromise to reduce the load of micropollutants signicantlywhile keeping the cost of the treatment in an acceptable range.Ozone-SF and PAC-SF had a similar cost (0.160.18 m3) with asimilar average removal rate. Compared to the averag