lable at ScienceDirect

Journal of Environmental Management 193 (2017) 491e502

Contents lists avai

Journal of Environmental Management

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

Research article

Pilot-scale removal of pharmaceuticals in municipal wastewater:Comparison of granular and powdered activated carbon treatment atthree wastewater treatment plants

Victor Kårelid*, Gen Larsson, Berndt Bj€orleniusDivision of Industrial Biotechnology, KTH Royal Institute of Technology, AlbaNova University Center, SE-106 91 Stockholm, Sweden

a r t i c l e i n f o

Article history:Received 17 October 2016Received in revised form15 February 2017Accepted 16 February 2017Available online 28 February 2017

Keywords:Advanced wastewater treatmentAdsorptionGACPACPharmaceutical removal

* Corresponding author.E-mail addresses: karelid@kth.se (V. Kårelid), gen

berndtb@kth.se (B. Bj€orlenius).

http://dx.doi.org/10.1016/j.jenvman.2017.02.0420301-4797/© 2017 Elsevier Ltd. All rights reserved.

a b s t r a c t

Adsorption with activated carbon is widely suggested as an option for the removal of organic micro-pollutants including pharmaceutically active compounds (PhACs) in wastewater. In this study adsorptionwith granular activated carbon (GAC) and powdered activated carbon (PAC) was analyzed and comparedin parallel operation at three Swedish wastewater treatment plants with the goal to achieve a 95% PhACremoval. Initially, mapping of the prevalence of over 100 substances was performed at each plant anddue to low concentrations a final 22 were selected for further evaluation. These include carbamazepine,clarithromycin and diclofenac, which currently are discussed for regulation internationally. A number ofcommercially available activated carbon products were initially screened using effluent wastewater. Ofthese, a reduced set was selected based on adsorption characteristics and cost. Experiments designedwith the selected carbons in pilot-scale showed that most products could indeed remove PhACs to thetarget level, both on total and individual basis. In a setup using internal recirculation the PAC systemachieved a 95% removal applying a fresh dose of 15e20 mg/L, while carbon usage rates for the GACapplication were much broader and ranged from <28 to 230 mg/L depending on the carbon product. Theperformance of the PAC products generally gave better results for individual PhACs in regards to carbonavailability. All carbon products showed a specific adsorption for a specific PhAC meaning that knowl-edge of the target pollutants must be acquired before successful design of a treatment system. In spite ofdifferent configurations and operating conditions of the different wastewater treatment plants noconsiderable differences regarding pharmaceutical removal were observed.

© 2017 Elsevier Ltd. All rights reserved.

1. Introduction

The removal of organic micropollutants within wastewatertreatment has become a matter of urgent concern during recentyears. This is a result of comprehensive research on the effects thatenvironmentally relevant concentrations may pose on living spe-cies occupying the receiving waters. Residues of pharmaceuticallyactive compounds (PhACs) are arguably of specific concern sincethey primarily end up in the municipal wastewater treatmentplants (WWTPs) where a large share passes relatively undisturbedthrough the treatment processes (Joss et al., 2006). Adverse effectsin fish have previously been reported e.g. on organ tissue (Hoeger

@biotech.kth.se (G. Larsson),

et al., 2005) and gene expression (Cuklev et al., 2011) caused bydiclofenac and on natural behavior (Brodin et al., 2013) caused byoxazepam. Moreover, the emergence of antibiotic resistance pre-sents a matter of global concern (Larsson, 2014).

During the quest to achieve a reliable and efficient removal oforganic micropollutants in general, adsorption onto activated car-bon and chemical transformation with ozone have emerged as themain alternatives, and a few pioneering studies in pilot and fullscale have been performed on effluent wastewater proving theefficacy (Boehler et al., 2012;Mailler et al., 2015; Margot et al., 2013;Meinel et al., 2015). These convincing results suggest that adsorp-tion is a first choice for high-level removal and is favorablecompared to oxidation by e.g. ozone. The presentation of activatedcarbon in purification systems is however subjected to specificconstraints. Conventionally, adsorption is performed using eitherthe granular (GAC) or the powdered (PAC) form. Promising resultsregarding adsorption of organic compounds, e.g. dyes and

V. Kårelid et al. / Journal of Environmental Management 193 (2017) 491e502492

surfactants have also been observed using activated carbon cloths(Ayranci and Duman, 2009; Duman and Ayranci, 2010). GAC iscontained in a column, which facilitates regeneration and reac-tivation while specific measures are required for PAC. On the otherhand, PAC has a larger available specific surface area resulting infaster adsorption kinetics (Meinel et al., 2015). The stationaryapplication of GAC ensures that saturation of the carbon can bereached, while for PAC applications this can generally only beachieved through internal recirculation due to the long contacttimes required to reach adsorption equilibrium (Meinel et al., 2015;Nowotny et al., 2007). Previous research has shown that a com-parable removal of organic micropollutants has been achieved us-ing separation and recirculation with significant reductions of thePAC dose in both bench-scale (Meinel et al., 2016) and in pilot andfull scale (Boehler et al., 2012). In Sweden, however, the desire torecycle sludge for agricultural purposes restricts the direct dosingof PAC into the biological treatment step and forces the design of afinal, separate process. The choice between adsorption with GACand PAC currently remains uncertain due to the advantages of therespective methods, thus invoking a demand for an increasedknowledge base before widespread full-scale implementation andfor site-specific guidance.

The aim of this studywas to achieve a 95% removal, in relation tothe effluent, of a selected set of 22 PhACs frequently occurring inmunicipal wastewater of representative Swedish WWTPs throughadsorption with activated carbon. This degree of removal is chosento surpass future effluent quality standard requirements by a safemargin. To appropriately target the removal of these substances,sets of activated carbons were selected through screening where arelevant load of PhACs was determined after mapping the effluentwastewater with respect to over 100 substances. The initial selec-tion covered a broad range of adsorption specific properties, e.g.specific surface area, iodine number and particle size. To accountfor site-specific conditions three different WWTPs were chosen forevaluation. These plants differed regarding wastewater character-istics and plant configuration to give a diverse representation ofSwedish wastewater effluents. To accommodate operation atdifferent locations a mobile pilot plant, specifically designed forcomparison of PAC and GAC purification performance, wasconstructed.

2. Materials and methods

2.1. Mobile pilot plant

A mobile pilot plant was constructed in-house for application atthe three WWTPs. Treatment tanks, and equipment for samplingand process control were installed into a 20-foot shippingcontainer allowing for operation at outdoor temperatures down tofreezing. The temperature of the wastewater treated in the pilotplant varied with the seasons according to the WWTP effluenttemperature. The pilot plant consisted of eleven treatment lines;three designed for GAC application, another three for PAC, twoozonation lines, two lines using biofilm (MBBR) and finally one linewith sand filtration after ozonation. This work, however, onlycovers operation with the GAC and PAC treatment lines. Duringoperation, effluent wastewater was pumped by a submergedimpeller pump via coarse particle filters (2 mm perforation) to aleveling tank located outside the container. The PAC and GAC lineswere fed from the leveling tank by separate screw pumps.

GAC filtration was performed in two identical treatment lines.The design was based on both literature (Tchobanoglous et al.,2003) and previous in-house tests (Wahlberg et al., 2010). Eachline consisted of two stainless steel tubes of 2m height in series andan inner diameter of 0.15 m, corresponding to a horizontal filter

area of 0.018 m2 and an operational volume for each line of ~60 L.The filters were filled with 1 m of GAC supported by filter bottomswith four nozzles in each filter. The operation was performed indown-flow configuration. A fixed, but adjustable, water level waskept in the GAC filter by a control valve connected beneath the filterbottom in each column. The GAC filters were supplied with manualvalves to enable backwashing with tap water.

PAC treatment was likewise performed using two identicaltreatment lines. The design of the PAC treatment system wasadapted from previous technical studies, primarily Metzger andKapp (2008) and Abegglen and Siegrist (2012). Each line con-sisted of an initial 1.7 L tank for mixing of effluent wastewater anddosed PAC, followed by three sequential aerated contact tanks, asedimentation tank and a final sand filter. After exiting the thirdcontact tank, wastewater enters the sedimentation tank 0.7 mbelow the surface level. The five main tanks were all made ofstainless steel, while the mixing tank was made from plastic filterhousing. Each contact tank was 0.9 m in height (the operatingwater level was 0.70e0.75 m) with an inner diameter of 0.25 m,amounting to a total operating volume of ~100 L. The sedimenta-tion tank measured 1.46 m in height (of which the bottom coneheight is 0.46 m) and had an inner diameter of 0.5 m (operatingvolume of ~180 L). The sand filter tank was 1.75 m in height with aninner diameter of 0.25 m (empty bed volume of 66 L) supported bya filter bottom with four nozzles. The sand filter was composed oftwo different media: a bottom layer constituting one third of thevolume of filter sand (1.2e2 mm particle size, Rådasand AB) and atop layer constituting two thirds of the volume of Filtralite MC2.5e4 (2e4mmparticle size, Saint Gobain Byggevarer AS). The totalsand filter height was maintained at 0.8e1.0 m during the experi-ments. Recirculation was accomplished by pumping with an airliftpump from the bottom of the sedimentation tank back to the firstcontact tank. Schematic setups of the treatment systems are pre-sented in Fig. 1.

2.2. Selected wastewater treatment plants

Three plants were chosen due to considerable differences inattached populations, disparities between their tertiary treatmentprocesses, as well as differences regarding treatment load such asthe total hydraulic retention time (HRT) and sludge loading rate.

K€appalaverket (K€appala) is the second largest WWTP in theStockholm region and treats 149 000 m3 wastewater per day, cor-responding to 425 000 population equivalents. The treatmentconsists of pre-treatment (screening and grit removal), primarysedimentation, biological treatment and sand filtration. Two thirdsof the wastewater is treated in a conventional activated sludge pre-denitrification setup using simultaneous chemical precipitation ofphosphorous with ferrous sulfate. One third of the wastewater istreated in the UCT setup (University of Cape Town; Ekama et al.,1983), which allows for enhanced biological phosphorous removal.

Kungs€angsverket (Uppsala) treats 50 000 m3 wastewater perday, corresponding to 148 000 population equivalents. The treat-ment consists of pre-treatment (screening and grit removal), pri-mary sedimentation, biological treatment, concluding withflocculation and lamella separation. Approximately 40% of thewastewater is treated in a conventional activated sludge pre-denitrification setup, while the remaining 60% is treated usingactivated sludge in a step-feed pre-denitrification setup. Both pre-and post-precipitation are used to achieve phosphorous removalthrough addition of ferric chloride. In Uppsala, the mobile pilotplant was extended with a pretreatment step in the form of twoshallow sand filter lines, after early observations of a fast cloggingprocess in the upper surfaces of the GAC filters. The filters wereinstalled before the leveling tank in the pilot plant.

Fig. 1. Schematics of the GAC (A) and PAC (B) treatment systems. The GAC system is composed of two sequential columns using down-flow operation. The PAC system is composedof an initial mixing tank (denoted “M”), three sequential contact tanks, a sedimentation tank and a sand filter. Sampling points are denoted “s”.

Table 1Selected activated carbons.

Denotation Product/supplier Specificsurface area,BET N2 (m2/g)

PAC A Pulsorb C/Chemviron 900PAC B Aquasorb MP20/Jacobi 700PAC C Aquasorb 5000P/Jacobi 1200GAC A Aquacarb 207C/Chemviron 1100GAC B Aquasorb 5000/Jacobi 1150GAC C Filtrasorb 400/Chemviron 1050GAC D GPP-20/Chemviron 725GAC E Carbsorb 30/Chemviron 900

V. Kårelid et al. / Journal of Environmental Management 193 (2017) 491e502 493

Kungs€angens WWTP (V€asterås) treats 48 000 m3 wastewaterper day, corresponding to 102 000 population equivalents. Thetreatment consists of pre-treatment (screening and grit removal),primary sedimentation and biological treatment in the form ofconventional activated sludge pre-denitrification setup. Methanoland ethylene glycol are used as carbon sources to improve the ni-trogen removal process. Pre-precipitation is used to achieve phos-phorous removal through addition of ferrous sulfate. Polymericcoagulants are added to the secondary sedimentation tanks toimprove particle separation before the effluent. The shallow sandfilters installed in Uppsala were in operation in V€asterås as well.

The experimental campaign took place during 2014 and wasconducted as follows: JanuaryeMay in K€appala (number of exper-imental weeks, GAC: 19, PAC: 15), June and AugusteSeptember inUppsala (GAC/PAC: 12), and OctobereDecember in V€asterås (GAC/PAC: 8). To fully evaluate the GAC treatment, those treatment lineswere in operation for an additional sevenweeks in the beginning of2015 at K€appala.

2.3. Pharmaceuticals

The selection of PhACs started with a larger set of over 100substances that has shown high potency and potential for bio-accumulation in fish according to Grabic et al. (2012). A subset ofthese was then selected based on 50% or higher occurrence in theeffluent samples at all three WWTPs. The properties of these 22selected substances are given in the supplementary material e

Table S1. Some substances that previously have shown highoccurrence as well as considerably high effluent concentrations inEuropean WWTPs (Loos et al., 2013) were excluded from the se-lection based on low occurrence in this study. These were notably:ciprofloxacin (occurrence: 24%, limit of quantification [LOQ]: 10 ng/L), sulfamethoxazole (occurrence: 6%, LOQ: 5 ng/L) and telmisartan(occurrence: 6%, LOQ: 50 ng/L). In case of detection below the LOQthe concentration was set to LOQ/2 when determining the removalof individual substances. However, when determining the overallremoval rates, i.e. for the sum of the 22 PhACs, detection below theLOQ was set to 0 ng/L. Overall removal rates that were determinedto be 99.5e100% are presented as >99%.

2.4. Activated carbon characteristics and screening

Activated carbon products used during the pilot-scale experi-ments were selected from an initially larger set of products bysuggestion from three different suppliers. Eight PAC products werescreened in dose response experiments in bench-scale carried outas follows: Wastewater was collected as grab samples from theeffluent at K€appala. A stock solution (20 g/L) was prepared for each

PAC product using deionized water. Experiments were performedin 250 mL baffled shake flasks (liquid volume: 120 mL) into whichthe stock solution was pipetted to reach the desired dose. Dosesused were 10, 20, 30, 40, 50, 100 mg/L for all PACs. Mixing wasmaintained using a horizontal shaker set to 150 min�1. The contacttime was 60 min, after which samples were immediately filtered toremove residual PAC with 0.45 mm cellulose acetate filters (What-man). Filtered samples were frozen awaiting PhAC analysis.

The GAC products were chosen based on previous pilot tests, butalso based on product characteristics, e.g. granule size, size distri-bution, specific surface area, iodine and methylene blue number(MBN), as well as cost. During the pilot tests three lines wereoperated in parallel, each with two filters in series and a total filtervolume of 40 L. The empty bed contact time (EBCT) was set to60 min in two of the lines and to 10 min in the third line. In total3500 bed volumes (BV) were treated in the first two lines and16 000 BV in the third line. The performance was evaluated mainlybased on removal rate. Four GAC products were selected due toremoval of over 95% in the pre-trials, methylene blue number�230and specific surface area �900 m2/g. One additional product, GACD, was chosen despite a specific surface area <900 m2/g due to 70%lower cost than the average of the four other products. Thus, in totalfive GAC products and three PAC products were eventually chosen(Table 1). Particle sizes are typical for the respective variants, whilea wide range of porosities is covered as conveyed by the specificsurface areas. More detailed characteristics according to the sup-pliers can be found in the supplementary material e Table S2.

2.5. GAC operation

The GAC filters were in continuous, 24/7, operation after startupat each WWTP and the individual lines were operated usingdifferent carbon products. Every product was however not used ateach WWTP (Tables S10 and S11). The water level was normally

V. Kårelid et al. / Journal of Environmental Management 193 (2017) 491e502494

maintained at 0.4e0.5 m above the GAC surface. On average, thesurface load for the two lines was 6.1/5.1 m/h in K€appala, 2.6/2.4 m/h in Uppsala and 3.3/5.1 m/h in V€asterås. The discrepancy in surfaceload is a result of different tendencies of clogging of the filters; theincoming flow was regulated accordingly. Surface loads werecalculated from the flow through the filter. The correspondingEBCTs for the two lines were 21/30 min in K€appala, 57/61 min inUppsala and 46/29 min in V€asterås. The GAC filters were back-washed with tap water two times per week in K€appala and everyworking day in Uppsala and V€asterås for 30 min at a backwashvelocity of 19e23 m/h. Due to the presumed risk of blending theactivated carbon column, no air was used for backwashing.

Before operation, the activated carbon was soaked in tap waterfor 5e7 days degassing air to prevent floating material. Large dif-ferences in floating property were noted for the products duringthe first days of soaking. At the time of GAC product exchange theused carbon was removed by a vacuum cleaner and replaced withanother product to the same height, 1 m. After soaking in tap water,the newGACwas backwashed, first at a low velocity, then graduallyup to the plant nominal velocity of 20 m/h. The loss of GAC con-sisted only of fine particles, with the exception for GAC B, wheresome grains still floated and were wasted during the first back-wash, before regular operation. The GAC products were exchangedafter indications of 25% overall breakthrough or after longer periodsof frequent filter clogging which prevented operation at adequatesurface loads.

2.6. PAC operation

The PAC treatment was in continuous, 24/7, operation afterstartup at eachWWTP and the individual lines were operated usingdifferent carbon products. Nominal contact times during the pilotexperiments were 55e73 min (design value: 60 min) correspond-ing to an incoming flow of 86e113 L/h and surface loads on thesedimentation tank and sand filter of 0.45e0.59 m/h and1.8e2.4 m/h, respectively during operation. A stock solution (20 g/L) of PAC was prepared with tap water and dispensed into themixing tank using peristaltic pumps (Ismatec Reglo). Mixing of thestock solution was maintained with aeration. A minor accumula-tion of settled PAC occurred over time that has been accounted forin the stated dose values. Mixing in the contact tanks was per-formed with aeration and was maintained at an airflow of ~10 L/min in each tank. Recirculation occurred intermittently for 10 severy minute and was maintained at an average rate of 49e66% ofthe incoming flow, corresponding to 53e59 L/h.

Air-assisted backwashing (16e22 m/h backwashvelocity þ 12 m/h air) of the sand filters was generally performedweekly using tap water and compressed air. Aeration was termi-nated when the turbidity of the backwashed surface water was ator below 10 NTU (measured with a Nanocolor VIS spectropho-tometer, Macherey-Nagel). The relation between turbidity andsuspended solids in the backwash water is given in Fig. S1. Back-washing without aeration was then continued for another fiveminutes to clear the sand filter from trapped air. Before and afterbackwashing the maximum surface load in the sand filters wastested by draining the clear water to ensure that the filter capacitywas restored. Backwashed water was not recycled to the treatmentlines.

Whenever a PAC product was replaced or when experimentswere concluded at a WWTP the treatment tanks were thoroughlycleaned. On a few occasions when the sand filter capacity wasvisibly near exhaustion during continuous operation, PAC wasremoved by opening a valve at the bottom of the sedimentationtank in connection to backwashing. Fresh and accumulated PACdoses as well as amounts for the determination of carbon solids

retention time (SRT) were calculated by setting up mass balances,further detailed in the supplementary material e section S1. Theaccumulated dose was defined as the average amount of PAC in thesystem divided by the treated wastewater volume during eachexperiment.

2.7. Analytical procedures

Samples for pharmaceutical analysis were collected duringseven days to match the sampling procedure at the selectedWWTPs. Samples were collected from effluent wastewater andafter each treatment line (Fig. 1) into refrigerated bottles using aperistaltic pump (Ismatec IP) and then stored at �18 �C. Beforestorage, samples were measured for conductivity and pH using aportable multiparameter meter (Orion Star A329, Thermo Scienti-fic). Pharmaceutical concentrations were quantified using a multi-residue method based on online solid-phase extraction and liquidchromatography tandem mass spectrometry, LC-MS/MS (Lindberget al., 2014), by Umeå University (Department of Chemistry).

At K€appala and V€asterås the total organic carbon (TOC) wasdetermined using cuvette tests (LCK 385, Hach Lange) andmeasured using a Hach Lange DR 3900 spectrophotometer. TOCanalysis at Uppsala was performed according to the standardmethod (SS EN 1484). Dissolved organic carbon (DOC) was deter-mined on effluent samples using the cuvette tests after filtrationthrough 0.45 mm membrane filters (Puradisc AQUA, Whatman).Absorbance measurements at 254 nm (UVA254) were performedwith a Hitachi U-2900 UV spectrophotometer, using quartz cu-vettes with a 50 mm path length (VWR). Suspended solids analysiswas performed according to wastewater standards (ESS Method340.2), using glass-micro fiber filters of grade GF/A (Whatman andMunktell).

The incoming wastewater flow to each treatment line wasrecorded by water meters supplied with internal registers andpulse outputs for external reading (Z-System, Qn 1.5, Systeme-sh).The pulses were optically read by a counter and transformed intodigital values which were stored in a logger 2020 system (WebIQAB). Temperatures were measured by digital sensors (1-wire PRO,Dallas) every minute and stored in the logger 2020 system.

3. Results and discussion

3.1. Characterization of pharmaceuticals in the effluentwastewaters

The effluent wastewaters were characterized from multiplesamples (N in Table 2) with regard to the occurrence and concen-trations of over 100 PhACs. From these PhACs, the 22 that showed50% or higher occurrence were selected for further evaluation.Effluent concentrations of these substances as well as sum con-centrations are presented in Table 2 for the respective WWTPs.Both averages and medians (data not shown) of the overall PhACconcentration show the highest values for V€asterås, followed byUppsala and K€appala. Furthermore, the same order is confirmed ifpopulation normalized discharges of the selected 22 PhACs areconsidered. The discharge in V€asterås was 2.9 mg pe�1d�1 andvalues for Uppsala and K€appala were 1.6 and 1.3 mg pe�1d�1. Theconsistently higher values in V€asterås could either be explained byan overall higher consumption of PhACs or by a poorer degradationof these substances in the wastewater treatment process. The latterwas indicated by the significantly lower sludge age and HRT inV€asterås (Table S3).

Average concentrations above 100 ng/L were observed for 12, 14and 15 substances, respectively of the individual substances andaverage concentrations above 1000 ng/L only for metoprolol in

Table 2Effluent pharmaceutical concentrations for the three WWTPs during pilot-scale operation, as well as values from literature (Loos et al., 2013).

Pharmaceutical LOQ K€appala Uppsala V€asterås Loos et al. (2013)

A (ng/L) B (ng/L) Na Effluent concentrationb

(ng/L)Na Effluent concentrationb

(ng/L)Na Effluent concentrationb

(ng/L)Effluentconcentrationc

(ng/L)

Atenolol 5 15 19/19 (5) 250 ± 124 (93e613) 12/12 (10) 220 ± 102 (123e459) 8/8 (2) 729 ± 239 (318e1093) e

Bisoprolol 0.1 4 19/19 (5) 46 ± 38 (7e126) 12/12 (10) 52 ± 34 (6e116) 8/8 (2) 301 ± 151 (106e644) 42/16 (423)Bupropion 0.1 4 14/19 (4) 8 ± 10 (<LOQ-38) 11/12 (9) 15 ± 11 (<LOQ-40) 8/8 (2) 12 ± 4 (7e20) 1/1 (5)Carbamazepine 1 7.5 18/19 (5) 221 ± 125 (<LOQ-452) 12/12 (10) 379 ± 92 (267e524) 8/8 (2) 202 ± 64 (131e348) 832/752 (4609)Citalopram 5 20 19/19 (5) 92 ± 96 (11e355) 12/12 (10) 179 ± 105 (21e357) 8/8 (2) 206 ± 115 (69e471) 34/21 (189)Clarithromycin 1 3 16/19 (5) 54 ± 45 (<LOQ-197) 12/12 (10) 37 ± 19 (20e78) 8/8 (2) 38 ± 17 (22e77) e

Clindamycin 1 3 19/19 (5) 80 ± 52 (26e222) 12/12 (10) 145 ± 65 (81e289) 8/8 (2) 149 ± 78 (46e335) 70/46 (277)Codeine 0.5 20 16/19 (4) 85 ± 104 (<LOQ-471) 10/12 (9) 74 ± 39 (<LOQ-153) 8/8 (2) 498 ± 615 (102e2079) 71/21 (826)Diclofenac 10 15 19/19 (5) 287 ± 163 (59e688) 10/12 (10) 200 ± 145 (<LOQ-592) 8/8 (2) 690 ± 447 (309e1776) 50/43 (174)Diltiazem 0.5 2 18/19 (4) 10 ± 11 (<LOQ-42) 11/12 (9) 6 ± 5 (<LOQ-16) 8/8 (2) 12 ± 4 (7e19) 11/6 (64)Fexofenadine 5 10 19/19 (5) 183 ± 202 (32e795) 12/12 (10) 280 ± 189 (17e624) 8/8 (2) 172 ± 93 (71e354) 165/59 (1287)Flecainide 0.1 2 18/19 (5) 67 ± 61 (8e215) 12/12 (10) 115 ± 75 (13e264) 8/8 (2) 125 ± 96 (54e368) 46/11 (553)Fluconazole 0.5 7.5 19/19 (5) 142 ± 92 (31e412) 12/12 (10) 190 ± 74 (122e372) 8/8 (2) 89 ± 30 (39e148) 108/68 (598)Irbesartan 0.5 3 19/19 (5) 213 ± 108 (80e488) 12/12 (10) 95 ± 37 (47e171) 8/8 (2) 87 ± 24 (62e132) 480/85 (17 900)Memantine 0.5 4 17/19 (5) 17 ± 22 (<LOQ-71) 12/12 (10) 22 ± 14 (2e45) 8/8 (2) 30 ± 13 (14e52) 23/4 (1312)Metoprolol 5 15 19/19 (5) 1203 ± 662 (322e2624) 12/12 (10) 1095 ± 241 (780e1598) 8/8 (2) 684 ± 200 (333e902) e

Mirtazapine 10 20 16/19 (5) 119 ± 139 (<LOQ-520) 12/12 (10) 189 ± 209 (<LOQ-554) 8/8 (2) 120 ± 204 (23e657) e

Oxazepam 5 10 19/19 (5) 178 ± 165 (10e523) 12/12 (10) 330 ± 72 (257e519) 8/8 (2) 741 ± 1248 (117e4037) 162/64 (1766)Sotalol 0.5 20 19/19 (5) 132 ± 110 (16e389) 12/12 (10) 246 ± 181 (27e600) 8/8 (2) 110 ± 74 (36e295) e

Tramadol 50 20 18/19 (5) 258 ± 178 (<LOQ-648) 12/12 (10) 563 ± 122 (394e837) 8/8 (2) 824 ± 463 (361e1966) 256/218 (1166)Trimethoprim 0.1 4 13/19 (2) 22 ± 29 (<LOQ-129) 12/12 (10) 77 ± 18 (60e128) 8/8 (2) 62 ± 17 (38e91) 229/178 (800)Venlafaxine 0.5 20 18/19 (5) 121 ± 112 (<LOQ-382) 12/12 (10) 370 ± 229 (49e813) 8/8 (2) 329 ± 212 (125e844) 119/97 (548)S Pharmaceuticals e e e 3784 ± 2025

(1127e8545)e 4882 ± 1074

(3715e7382)e 6208 ± 3048

(2455e12 466)e

a Number of experimental weeks detection above LOQ/All experimental weeks. Parenthesized values indicate experimental weeks when LOQ Awas used for quantification;LOQ B was used for the rest.

b Average ± standard deviation (min-max).c Average/median (max).

V. Kårelid et al. / Journal of Environmental Management 193 (2017) 491e502 495

K€appala and Uppsala. The comparatively lower presence of meto-prolol in V€asterås might be a result of the higher presence ofatenolol and bisoprolol belonging to the same pharmaceutical class.Singular recordings above 1000 ng/L in V€asterås were howeverindeed seen for atenolol, codeine, diclofenac, oxazepam and tra-madol. Regarding some internationally highlighted substances, theantiepileptic drug carbamazepine was detected at ~50% higherconcentration in Uppsala compared to the other plants; concen-trations of diclofenac in V€asterås were 2e3 times higher comparedto the other two, while K€appala showed the highest average con-centration among the plants for clarithromycin.

Compared with an extensive European study (Loos et al., 2013),most of the detected substances in this studywould place above thereported median concentrations (Table 2). In fact, only two out ofthe seventeen overlapping substances between the studies, car-bamazepine and trimethoprim are consistently found below thereported medians in this study. In a few cases, specifically forbupropion and diclofenac at all plants and citalopram in V€asterås,the observed concentrations in this study surpassed the previouslyreported maximum concentrations. For some of the evaluatedPhACs in this study similar concentrations have however previ-ously been reported for effluent wastewater in the same geographicarea (Wahlberg et al., 2011).

3.2. Pharmaceutical removal

The only current effluent standard regarding organic micro-pollutants in effluent wastewater that to our knowledge has beenintroduced is for a selection of WWTPs in Switzerland. Afterimplementation of the treatment processes a removal of 80% inrelation to influent wastewater is required for a set of indicatorsubstances, including carbamazepine and diclofenac (Sch€arer andBleny, 2015). Diclofenac has also been highlighted by the EU

Water framework directive as a prioritized substance for moni-toring, as is also the case for azithromycin, clarithromycin anderythromycin (EC, 2015). In the following evaluation carbamaze-pine, clarithromycin and diclofenac are denoted as indicator PhACs.To safely pass future expected discharge limits in Sweden we setout to reach an overall PhAC removal of 95% for both treatmentmethods in relation to the effluent wastewaters.

3.2.1. PAC screeningPrior to the pilot-scale campaign, screening in bench-scale was

performed on eight different PAC products selected due to diversitywith regard to specific surface area, density and raw material. Theexperiments were designed to investigate the dose response; initialPhAC concentrations are given in Table S4. A large variationregarding the overall removal between the products was observedat low doses (Fig. 2) with a large distribution up to 100mg/L. At thisdose, all products gave a 95% or higher removal. A better estimationof the dose required for a 95% removal was achieved with a log plotof the dose (Table S5). Three main contenders emerged with regardto dose and cost efficiency: PAC A (Pulsorb C), PAC B and PAC C (bothAquasorb), and were thus selected for pilot-scale evaluation. Pul-sorb FG4 also showed a good result but was not available in bulk.The dose for 95% removal was ~30 mg/L for PAC B and C and 85 mg/L for PAC A, thus 30 mg/L was set as initial operational dose for theinitial pilot experiments.

3.2.2. PAC pilot-scale experimentsThe overall PhAC removal was above 95% for a majority of the

experiments in spite of different PAC doses (Table 3). Only inV€asterås, where the lowest average dose was applied, did theaverage removal fall below this goal. The overall performance be-tween the PAC products showed no considerable differences(Table S6 and Fig. S2). Both PAC A and B showed consistently higher

Table 3Summary of the PAC pilot experiments.

Parameter Unit K€appalaa Uppsalaa V€asteråsa

Overall PhAC removal % 99 ± 1 99 ± 1 94 ± 4Fresh dose mg/L 32 ± 12 26 ± 6 15 ± 4Accumulated dose mg/L 150 ± 100 40 ± 13 22 ± 10DOC-normalized fresh dose mg/mg

DOC3.4 ± 1.1 3.1 ± 1.0 1.7 ± 0.5

Nominal contact time min 63 ± 12 70 ± 7 65 ± 7Total system HRT h 3.6 ± 0.7 4.0 ± 0.4 3.7 ± 0.4Wastewater temperature �C 13.1 ± 0.8 19.3 ± 1.1 15.7 ± 1.0Number of experiments e 28 24 15

a Average ± standard deviation.

V. Kårelid et al. / Journal of Environmental Management 193 (2017) 491e502496

than 95% removal in K€appala, and the same could be seen for allthree products in Uppsala. PAC B and C that were used in V€asteråsshowed average removals just below 95%. The continuously addedfresh dose of 30 mg/L was proven to be excessive since one third ofthe 30 performed experiments showed complete removal (allPhACs below the LOQ) during operation in K€appala. In Uppsala thefresh dose was thus lowered and although the removal was still onaverage 99%, complete removal was only observed during oneexperiment. In V€asterås, and average removal of just below 95%wasachieved with a fresh dose of 15 mg/L.

The design of the system, including the recirculation, leads toaccumulation of PAC in the tanks and we thus propose the use ofthe “accumulated dose” to better depict the carbon accessibilityand moreover best relate to the carbon usage rate during GACoperation. A third dose concept used in literature is the “DOC-normalized dose” which accounts for the organic matter in theeffluent and this was shown to predict the removal better than thevolumetric dose in some cases (Altmann et al., 2014). However,after statistical evaluation of our data all dose measurements wereshown to predict the overall PhAC removal equally well (Table S7).All the process parameters of the WWTPs are subjected to seasonalvariation, and since these experiments were consecutively per-formed, in particular the temperature could have a major influenceon PAC adsorption. The difference in temperature between theplants was however seemingly moderate and could at most have amarginal effect on the adsorption behavior.

Considering all PAC pilot experiments the individual substancesshowed a 95% or higher removal or were detected below the LOQ in86% of the cases. In order to evaluate the dose dependence for theremoval of individual substances all experiments were sorted intodose intervals between which a significant effect on the overallPhAC removal could be statistically determined (Tables S8 and S9).At an average accumulated dose of 18 mg/L, i.e. at the lowest doseinterval (<30mg/L), 7 of the PhACs showed a 95% or higher removal(Fig. 3a) and at the higher intervals this increased to 17 substances.If 100% detection below the LOQ is added as a measure of sufficientremoval 9,19 and 20 substances respectivelymeet this qualification(Fig. 3b). At the highest interval only fluconazole (92% removal/89%below LOQ) and memantine (83% removal/93% below LOQ) showinsufficient removal. Regarding the indicator PhACs, the lowest

Fig. 2. Dose response curves from the PAC screening in bench-scale. The experiments were p100 mg/L.

interval was adequate to reach the target removal for clari-thromycin, while 30e100 mg/L was required for carbamazepineand diclofenac.

If instead intervals of the fresh dose are considered (Table S9), anaverage dose of 15 mg/L (<20 mg/L) was sufficient for a 95%removal of clarithromycin and diclofenac, while carbamazepinerequired a dose of 20e25 mg/L (Fig. S3). In a considerably largerpilot system that was fed with effluent wastewater and using anextended solid retention time and an average fresh PAC dose of14 mg/L, carbamazepine and diclofenac was reported to beremoved by an average of 73% and 85% (Mailler et al., 2015).

When comparing the results from the screening and the pilotexperiments there is an apparent increase of the removal at thesame (fresh) dose in the latter case. The higher availability of un-saturated PAC due to accumulation in the treatment lines couldlargely explain this increase. To consider is also the effectivelycontact times used in pilot-scale, due to the sedimentation tank andthe sand filter. To evaluate the improvement in performance due tothese aspects the recorded fresh doses from pilot operation werecomparedwith corresponding doses estimated from screening datato achieve the same overall removal. This was achieved by linearinterpolation of log dose against the removal. PAC B was used asscreening reference due to the ubiquitous use in pilot-scale at allthree plants. Furthermore, eventual overdosing was accounted forby excluding experiments showing complete removal. The K€appalaexperiments would have required a dose of 37 mg/L to achieve in

erformed with a 60 min contact time and applied PAC doses were 10, 20, 30, 40, 50 and

V. Kårelid et al. / Journal of Environmental Management 193 (2017) 491e502 497

the bench-scale tests compared with the actual average dose of29 mg/L, an improvement of 22%. Corresponding results fromUppsala and V€asterås were 33% and 52% with dose reductions of 39to 26 mg/L and 31 to 15 mg/L, respectively. The results fromV€asterås are in line with what Boehler et al. (2012) have shown infull scalewhere 15mg/L PAC in a recirculation setup showed similarremoval to 30 mg/L without recirculation. However, there seems tobe room for improvement, since Meinel et al. (2016) recentlyshowed that a 70% or higher improvement is possible in a labora-tory setup. The increasing improvement shown over the course ofthe pilot-scale campaign affirms that the effect of recirculation ishigher for lower doses.

3.2.3. GAC pilot-scale experimentsIn general, the overall PhAC removal by GAC filtration exceeded

90%. A comparison of the adsorption on the individual carbons,represented by the overall PhAC uptake (Fig. S4), showed theexceptional accommodation of the products to occasional very highloads at the different WWTPs, except for GAC Awhich lost capacityquite early on. Table 4 shows the overall performance from the pilotexperiments at the individual WWTPs. Concerning the individualGAC products, three of the five showedmore than 95% overall PhACremoval throughout the experiments (Tables S10 and S11). Thelargest deviation from this performance was the 17% drop inadsorption capacity for GAC E at the higher BV in V€asterås (data tothe right in Table 4), likely due to being a coarser product. In orderto facilitate the comparison of treatment results, removal at 1600and 3200 bed volumes is presented due to being the shortest length

Fig. 3. Substance removability in the PAC pilot experiments displayed for the accumulated dnumber of experiments where detection was below the LOQ.

of operation at one single WWTP. No difference in overall PhACremoval was indicated after treatment of 1600 bed volumes,however, after 3200 treated bed volumes in Uppsala and V€asteråsthe removal was lower, although with a considerable variationbetween the GAC products.

To remove 95% of the selected PhACs the average carbon usagerate was nearly the same for all the three WWTPs, 110 mg/L. Indi-vidually the values for the three best products were: GACB < 43 mg/L (at 6440 BV), GAC C < 41 mg/L (at 10210 BV) and GACD< 28mg/L (at 16130 BV). Absolute values could not be determinedsince the test periods were too short for these high-performingGACs to reach 5% breakthrough. The other two products showedmuch higher carbon usage rates: GAC A 230 mg/L (at 2290 BV) andGAC E 170 mg/L (at 2360 BV). No general explanation for the dif-ferences in carbon usage rates and adsorption capacity has beenfound from carbon characteristics and lack of raw material data forGAC B and GAC D rules out a full comparison. However, in this studya higher methylene blue number (MBN) corresponded well to ahigher adsorption of PhACs, i.e. products GAC A and E both have aMBN of 230 and showed lower removal than GAC B and C with aMBN of 260. MBN could thus be suggested as a guiding parameterwhen selecting GAC products for removal of PhACs.

Neither of the process conditions: empty bed contact time(EBCT), overall PhAC concentration and DOC, suggested a strongcorrelation to product adsorption capacity and thus overall PhACremoval. This is evident if a period of common process conditions(here: EBCT ~20 min) is evaluated for different products. For 11weeks of parallel operation of GAC A and D in K€appala, e.g. the

ose intervals. (A) Individual substance removal for all experiments. (B) Share of the total

Table 4Summary of the GAC pilot experiments.

Parameter Unit K€appalaa Uppsalaa V€asteråsa V€asterås, GAC Ca V€asterås, GAC Ea

Overall PhAC removal, 1600 BV % 99 ± 1 >99 >99 >99 >99Overall PhAC removal, 3200 BV % 96 ± 6 93 ± 7 91 ± 9 >99 83 ± 3Carbon usage rate, 95% removal mg/L 110 ± 100 110 ± 78 110 ± 72 41 180Filter surface load m/h 5.8 ± 1.8 2.5 ± 0.51 4.2 ± 1.2 3.3 ± 0.83 5.1 ± 0.75Total system HRT min 35 ± 20 75 ± 17 46 ± 16 58 ± 15 34 ± 4.7Suspended solids in load mg/L 0.6 ± 0.1 4.6 ± 0.9 3.6 ± 1.7 3.6 ± 1.7 3.6 ± 1.7Filter sludge storage kg/m2 0.29 ± 0.17 0.44 ± 0.15 0.57 ± 0.31 0.45 ± 0.27 0.69 ± 0.29Share of backwash water of filtered water % 3.5 ± 1.9 16 ± 7.2 9.7 ± 3.8 13 ± 2.5 6.4 ± 0.81Wastewater temperature �C 13.4 ± 1.2 19.2 ± 1.1 15.5 ± 1.0 15.5 ± 1.0 15.5 ± 1.0Number of experiments e 38 24 16 8 8

a Average ± standard deviation.

V. Kårelid et al. / Journal of Environmental Management 193 (2017) 491e502498

corresponding removal rates were 82 and > 99%. This suggests thatperformance is dependent on the carbon product under suchconditions.

The individual WWTPs have different concentrations of sus-pended solids in the effluent wastewater as a result of the perfor-mance of their final treatment step (Table 4). This influences thehydraulic capacity, surface load and retention time for wastewaterin the filters. In Uppsala and V€asterås the hydraulic capacities werelower than in K€appala. This had a negative impact on the operationperformance of the GAC products in particular since they havedifferent capacity for sludge storage. Sludge storage is here definedas the amount of particles accumulated on top of the GAC surface inthe first GAC filter in each line. The calculated sludge storage

Fig. 4. Substance removability in the GAC pilot experiments displayed for the carbon usage rnumber of experiments where detection was below the LOQ.

correlated naturally to the required backwash frequency and vol-ume and in Uppsala and V€asterås this increased three to four-foldcompared to in K€appala. Two approaches were used to controlthe situation: installation of shallow sand filters before GAC treat-ment and change of carbon characteristics. By changing to a coarserproduct, GAC E, the hydraulic capacity was nearly recovered inV€asterås although the adsorption capacity was reduced. Comparingthe two measures to restore the hydraulic capacity, a coarse GACproduct wasmore successful than the shallow sand filters that wereinstalled in this study.

The individual removal of PhACs showed that 8 out of 22 sub-stances were removed�95% at a carbon usage rate of 30e100 mg/Lfor all experiments and GAC products (Fig. 4a). Regarding the

ate intervals. (A) Individual substance removal for all experiments. (B) Share of the total

V. Kårelid et al. / Journal of Environmental Management 193 (2017) 491e502 499

suggested indicator PhACs, carbamazepine and clarithromycinwere removed >95% and diclofenac to 93% at this interval. A carbonusage rate exceeding 100 mg/L did not result in a considerableimprovement, which can partly be explained by the poor adsorp-tion capacity of GAC A and E since these products never reached acarbon usage rate <100 mg/L before they were exchanged. For theindividual substances, memantine posed a particular concern sinceonly a 72% removal was achieved at the lower carbon usage rateinterval. A removal of >80% was however achieved exceeding100 mg/L. Some substances were present at very low concentra-tions, in particular bupropion, citalopram and mirtazapine. Despiteshowing removal rates lower than 90%, these substances were notdetected in any sample following GAC treatment. Fig. 4b is there-fore used to illustrate the fact that to a large extent many sub-stances were removed below the LOQ during the experiments.

The removal of individual PhACs varied for different GACproducts, which is illustrated by breakthrough curves for foursubstances (Fig. 5). The substances selected are the three indicatorPhACs as well as irbesartan that showed relatively poor removalwith both PAC and GAC treatment. GAC A and E showed the lowestremoval capacity i.e. breakthrough occurred earlier than for theother products that hadmore than five times the capacity to adsorbclarithromycin and irbesartan. Furthermore, GAC E reached rela-tively early breakthrough of several PhACs, remarkably for diclo-fenac, tramadol, carbamazepine, venlafaxine, oxazepam andfluconazole. The ratio C/C0 scattered mainly due to variations inWWTP effluent PhAC concentration (individual data not shown).

From the breakthrough curves of all 22 substances the specificadsorption capacity at C/C0 ¼ 0.05 was determined for all products

Fig. 5. Breakthrough curves from the GAC pilot experiments. Breakthrough, C/C0, is plotted airbesartan. Dashed lines are added to improve visual interpretation.

(Table S12). Again, early breakthrough and lower adsorption of GACA and E confirms a general lower specific adsorption capacitycomparedwith the other products. To compensate for varying PhACconcentrations into the GAC columns the data of Table S12 werenormalized using these values (Table S13). From these data it isclear that each individual product adsorbs each individual PhACvery differently even though most contain one or more aromaticgroups which lead to the conclusion that not only aromaticitycontributes to adsorption, but also other properties. For GAC B, e.g.the maximum variation of the specific adsorption capacity of in-dividual PhACs was 150-fold, while the average for all productsshowed a 50-fold maximum variation. On the other hand, if thecalculation is based on the sum of the 22 PhACs, this tendency is notas obvious since it only results in a 22-fold variation. Only for car-bamazepine and trimethoprim was the performance in the samerange for the five different products.

3.2.4. Use of UVA254 to predict PhAC removalIt has previously been reported that the reduction of UV ab-

sorption at 254 nm (UVA254) is a strong indicator for the removal oforganic micropollutants, including PhACs (Altmann et al., 2014;Zietzschmann et al., 2014). Presence of aromatic substances,which are highly prevalent among PhACs, is well represented byUVA254 (Westerhoff et al., 1999). Out of the 22 substances evaluatedin this study 19 contain aromatic groups (Table S1), thus a corre-lation was expected in our experiments. Fig. 6 shows this relationfor all pilot-scale experiments, including data points after both thefirst and the second column for GAC. Overall, the two parametersshow good correlation (rspearman ¼ 0.75, p < 0.01). Specifically the

gainst treated bed volumes displayed for carbamazepine, clarithromycin, diclofenac and

Fig. 6. The relation between UV254 reduction and overall removal of PhACs displayed for all pilot experiments. GAC and PAC experiments are presented separately.

V. Kårelid et al. / Journal of Environmental Management 193 (2017) 491e502500

GAC experiments are well correlated (r ¼ 0.80, p < 0.01), while thePAC experiments show considerably weaker correlation (r ¼ 0.56,p < 0.01). Up to a removal of 80% a linear relation has been reported(Zietzschmann et al., 2015), which could however not be consoli-dated by our results (rpearson ¼ 0.34, p ¼ 0.10).

3.2.5. Treatment methods comparisonAs shown in Figs. 3 and 4, in general, the PAC operation resulted

in higher removal rates using the same range of carbon consump-tion compared to GAC operation. At the lowest common interval,30e100 mg/L, PAC showed considerably better performance for allsubstances except for atenolol, clarithromycin and metoprolol,where the removal was comparable or higher for GAC. To clarify the

Fig. 7. Comparison of the specific adsorption cap

differences in specific adsorption capacity for the carbon productsGAC data, determined at 5% breakthrough (Table S12), was plottedagainst the PAC results for individual PhACs (Fig. 7). The PAC datawas determined from all experiments showing a removal of 95 ± 3%(Table S14). It should be noted however that these values have notbeen normalized with respect to WWTP effluent PhAC concentra-tions. In general, the results from Figs. 3 and 4 are further confirmedin the way that all plotted PhACs show a higher specific adsorptionfor PAC. However, this comparison includes the low-performingGAC A and E and if only GAC B, C and D are considered, the dif-ference is less pronounced (data not shown). For some individualsubstances the difference is specifically distinct between GAC andPAC, in particular for tramadol, while almost equivalent results are

acity for the GAC and PAC pilot experiments.

V. Kårelid et al. / Journal of Environmental Management 193 (2017) 491e502 501

observed for e.g. diclofenac.As an alternative way of discerning the capability for individual

PhAC removal of the two methods a ranking based on a combina-tion of removal rate and degree of complete removal (percentagebelow LOQ) was developed (Table S15). Striking differences be-tween the methods was seen for fexofenadine and oxazepam,where GAC showed a better performance with respect to removalof the former and the opposite for the latter. Regarding the indi-cator PhACs both clarithromycin and diclofenac showed relativelybetter removability with PAC, while carbamazepine had a slightlybetter rank for GAC. Removability of PhACs is in Table S15 rankedbased on a combination of both methods showing the best per-formance at the top. It is notable that the first nine substances areall positively charged at normal wastewater pH which is in linewith what has previously been reported (Mailler et al., 2015;Margot et al., 2013). The two poorest ranked substances arememantine and fluconazole. Memantine has a significantly lowermolecular weight as well as a more compact molecular structurethan any of the other monitored substances. Furthermore it is notaromatic, which has proven to be a disadvantageous adsorptioncharacteristic (de Ridder et al., 2010). Fluconazole is neutrallycharged at wastewater pH and is therefore more dependent onhydrophobic interactions than positively charged substances (deRidder et al., 2011; Margot et al., 2013). The low removabilitycompared to the other neutrally charged substances might there-fore be explained by the more hydrophilic nature of this substance,as displayed by the low log D (Table S1).

Regarding the operation of the two treatment methods no sig-nificant difference was perceived regarding the required man-hours. A total of 30e60 min per working day was required foreither method to perform backwashing and ensure continuousdosing in the PAC systems. Although designed for similar incomingflows the hydraulic capacity was in reality lower in the GAC sys-tems, primarily due to higher loads of suspended solids in theeffluent wastewater in Uppsala and V€asterås. In spite of theinstallation of the shallow sand filters, which principally enabledGAC operation, more frequent backwashing and lower incomingflows were required compared with K€appala to avoid flooding ofthe columns. The total HRT differed significantly between the sys-tems. A four to six times longer total HRT was used for the PACsystems revealing the significantly higher demand for tank space.

4. Conclusions

In general, the degree of pharmaceutical pollution in the efflu-ents of three major WWTPs in Sweden was low with regard to thenumber of substances that exceeded the limits of quantification outof the over 100 selected for evaluation. Over 80% of these were onlypresent in less than half of the sample pool and thus did not qualifyfor further investigation. However, among the remaining 22 sub-stances, concentrations were above general values reported in theliterature. In Sweden the design of treatment for removal of phar-maceuticals should most likely consider these specific substances.Although our data represents a restricted set of wastewater efflu-ents all types of seasonal variation is covered.

Activated carbon was shown to be an efficient method forremoval of PhACs in the effluents of municipal wastewater treat-ment plants and the goal of 95% removal was reached for almost alltested substances by design of a relevant purification system.Regarding the choice of an individual carbon product there arelarge variations in the performance with respect to individualsubstances, thus if the goal is to remove a specific substance avariation of both GAC and PAC qualities should be screened. In thecase of GAC, screening with respect to hydraulic capacity and car-bon usage rate is a necessity if the specific product has not

previously been tested on similar wastewater. If the specific task isthe removal of as much as possible of pharmaceutical residues inuncharted wastewater the recommendation is also here to use PAC,however in terms of carbon usage well-performing GAC productscould approach the performance of PAC, as indicated by our results.

Acknowledgements

The study was funded by the Swedish Foundation for StrategicEnvironmental Research (MISTRA), through the MistraPharmaresearch program. We want to thank Richard Lindberg and JerkerFick at Umeå University for their extensive work performing thepharmaceutical analyses. Furthermore, we want to thank the staffsat K€appalaverket, Uppsala Vatten and M€alarenergi for help withlogistics, technical support and for their great hospitality. Finally,we want to thank Anna Hjalmars and Pia Trygg for their invaluablework during operation at the pilot plant and Niklas Dahl�en forassistance during the PAC screening.

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.jenvman.2017.02.042.

References

Abegglen, C., Siegrist, H., 2012. Mikroverunreinigungen aus kommunalem Abwasser- Verfahren zur weitergehenden Elimination auf Kl€aranlagen. In: Micro-pollutants in Municipal Wastewater - Processes for Advanced Removal inWastewater Treatment Plants. Federal Office for the Environment FOEN, Bern.

Altmann, J., Ruhl, A.S., Zietzschmann, F., Jekel, M., 2014. Direct comparison ofozonation and adsorption onto powdered activated carbon for micropollutantremoval in advanced wastewater treatment. Water Res. 55, 185e193. http://dx.doi.org/10.1016/j.watres.2014.02.025.

Ayranci, E., Duman, O., 2009. In-situ UV-visible spectroscopic study on theadsorption of some dyes onto activated carbon cloth. Sep. Sci. Technol. 44,3735e3752. http://dx.doi.org/10.1080/01496390903182891.

Boehler, M., Zwickenpflug, B., Hollender, J., Ternes, T., Joss, A., Siegrist, H., 2012.Removal of micropollutants in municipal wastewater treatment plants bypowder-activated carbon. Water Sci. Technol. 66, 2115e2121. http://dx.doi.org/10.2166/wst.2012.353.

Brodin, T., Fick, J., Jonsson, M., Klaminder, J., 2013. Dilute concentrations of a psy-chiatric drug alter behavior of fish from natural populations. Science 339,814e815. http://dx.doi.org/10.1126/science.1226850.

Cuklev, F., Kristiansson, E., Fick, J., Asker, N., F€orlin, L., Larsson, D.G.J., 2011. Diclo-fenac in fish: blood plasma levels similar to human therapeutic levels affectglobal hepatic gene expression. Environ. Toxicol. Chem. 30, 2126e2134. http://dx.doi.org/10.1002/etc.599.

de Ridder, D.J., Verliefde, A.R.D., Heijman, S.G.J., Verberk, J.Q.J.C., Rietveld, L.C., vander Aa, L.T.J., Amy, G.L., van Dijk, J.C., 2011. Influence of natural organic matteron equilibrium adsorption of neutral and charged pharmaceuticals onto acti-vated carbon. Water Sci. Technol. 63, 416e423. http://dx.doi.org/10.2166/wst.2011.237.

de Ridder, D.J., Villacorte, L., Verliefde, A.R.D., Verberk, J.Q.J.C., Heijman, S.G.J.,Amy, G.L., van Dijk, J.C., 2010. Modeling equilibrium adsorption of organicmicropollutants onto activated carbon. Water Res. 44, 3077e3086. http://dx.doi.org/10.1016/j.watres.2010.02.034.

Duman, O., Ayranci, E., 2010. Adsorptive removal of cationic surfactants fromaqueous solutions onto high-area activated carbon cloth monitored by in situUV spectroscopy. J. Hazard. Mater. 174, 359e367. http://dx.doi.org/10.1016/j.jhazmat.2009.09.058.

Ekama, G.A., Siebritz, I.P., Marais, G.V.R., 1983. Considerations in the process designof nutrient removal activated sludge processes. Water Sci. Technol. 15, 283e318.

Grabic, R., Fick, J., Lindberg, R.H., Fedorova, G., Tysklind, M., 2012. Multi-residuemethod for trace level determination of pharmaceuticals in environmentalsamples using liquid chromatography coupled to triple quadrupole massspectrometry. Talanta 100, 183e195. http://dx.doi.org/10.1016/j.talanta.2012.08.032.

Hoeger, B., K€ollner, B., Dietrich, D.R., Hitzfeld, B., 2005. Water-borne diclofenac af-fects kidney and gill integrity and selected immune parameters in brown trout(Salmo trutta f. fario). Aquat. Toxicol. 75, 53e64. http://dx.doi.org/10.1016/j.aquatox.2005.07.006.

Joss, A., Zabczynski, S., G€obel, A., Hoffmann, B., L€offler, D., McArdell, C.S., Ternes, T.A.,Thomsen, A., Siegrist, H., 2006. Biological degradation of pharmaceuticals inmunicipal wastewater treatment: proposing a classification scheme. Water Res.40, 1686e1696. http://dx.doi.org/10.1016/j.watres.2006.02.014.

Larsson, D.G.J., 2014. Antibiotics in the environment. Ups. J. Med. Sci. 119, 108e112.

V. Kårelid et al. / Journal of Environmental Management 193 (2017) 491e502502

http://dx.doi.org/10.3109/03009734.2014.896438.Lindberg, R.H., €Ostman, M., Olofsson, U., Grabic, R., Fick, J., 2014. Occurrence and

behaviour of 105 active pharmaceutical ingredients in sewage waters of amunicipal sewer collection system. Water Res. 58, 221e229. http://dx.doi.org/10.1016/j.watres.2014.03.076.

Loos, R., Carvalho, R., Ant�onio, D.C., Comero, S., Locoro, G., Tavazzi, S., Paracchini, B.,Ghiani, M., Lettieri, T., Blaha, L., Jarosova, B., Voorspoels, S., Servaes, K.,Haglund, P., Fick, J., Lindberg, R.H., Schwesig, D., Gawlik, B.M., 2013. EU-widemonitoring survey on emerging polar organic contaminants in wastewatertreatment plant effluents. Water Res. 47, 6475e6487. http://dx.doi.org/10.1016/j.watres.2013.08.024.

Mailler, R., Gasperi, J., Coquet, Y., Deshayes, S., Zedek, S., Cren-Oliv�e, C., Cartiser, N.,Eudes, V., Bressy, A., Caupos, E., Moilleron, R., Chebbo, G., Rocher, V., 2015. Studyof a large scale powdered activated carbon pilot: removals of a wide range ofemerging and priority micropollutants from wastewater treatment plant ef-fluents. Water Res. 72, 315e330. http://dx.doi.org/10.1016/j.watres.2014.10.047.

Margot, J., Kienle, C., Magnet, A., Weil, M., Rossi, L., de Alencastro, L.F., Abegglen, C.,Thonney, D., Ch�evre, N., Sch€arer, M., Barry, D.A., 2013. Treatment of micro-pollutants in municipal wastewater: ozone or powdered activated carbon? Sci.Total Environ. 461e462, 480e498. http://dx.doi.org/10.1016/j.scitotenv.2013.05.034.

Meinel, F., Ruhl, A.S., Sperlich, A., Zietzschmann, F., Jekel, M., 2015. Pilot-Scaleinvestigation of micropollutant removal with granular and powdered activatedcarbon. Water. Air. Soil Pollut. 226, 2260. http://dx.doi.org/10.1007/s11270-014-2260-y.

Meinel, F., Zietzschmann, F., Ruhl, A.S., Sperlich, A., Jekel, M., 2016. The benefits ofpowdered activated carbon recirculation for micropollutant removal inadvanced wastewater treatment. Water Res. 91, 97e103. http://dx.doi.org/10.1016/j.watres.2016.01.009.

Metzger, S., Kapp, H., 2008. Technische Verfahren zur Entfernung von Mikrover-unreinigungen aus kommunalem Abwasser - Einsatz von Pulveraktivkohle zurElimination von Mikroverunreinigungen. In: Industrial Processes for theRemoval of Micropollutants from Municipal Wastewater - Application of

Powdered Activated Carbon for the Removal of Micropollutants. VerbandSchweizer Abwasser- und Gew€asserschutzfachleute, Glattbrugg.

Nowotny, N., Epp, B., von Sonntag, C., Fahlenkamp, H., 2007. Quantification andmodeling of the elimination behavior of ecologically problematic wastewatermicropollutants by adsorption on powdered and granulated activated carbon.Environ. Sci. Technol. 41, 2050e2055. http://dx.doi.org/10.1021/es0618595.

Sch€arer, M., Bleny, H., 2015. Elimination of micropollutants - the Swiss approach. In:Final Conference, Transnational Action Program on Emerging Substances(TAPES) (Brussels).

Tchobanoglous, G., Burton, F.L., Stensel, H.D., 2003. Wastewater Engineering:Treatment and Reuse, fourth ed. McGraw-Hill, Boston.

The European Commision, 2015. Development of the First Watch List under theEnvironmental Quality Standards Directive. http://dx.doi.org/10.2788/101376.

Wahlberg, C., Bj€orlenius, B., Pax�eus, N., 2011. Fluxes of 13 selected pharmaceuticalsin the water cycle of Stockholm. Swed. Water Sci. Technol. 63, 1772e1780.http://dx.doi.org/10.2166/wst.2011.124.

Wahlberg, C., Bj€orlenius, B., Paxeus, N., 2010. L€akemedelsrester i StockholmsVattenmilj€o - F€orekomst, F€orebyggande Åtg€arder och Rening av Avloppsvatten.In: Pharmaceuticals in Aquatic Stockholm - Presence, Prevention and Treatmentof Wastewater. Stockholm Vatten, Stockholm.

Westerhoff, P., Aiken, G., Amy, G., Debroux, J., 1999. Relationships between thestructure of natural organic matter and its reactivity towards molecular ozoneand hydroxyl radicals. Water Res. 33, 2265e2276. http://dx.doi.org/10.1016/S0043-1354(98)00447-3.

Zietzschmann, F., Altmann, J., Hannemann, C., Jekel, M., 2015. Lab-testing, predict-ing, and modeling multi-stage activated carbon adsorption of organic micro-pollutants from treated wastewater. Water Res. 83, 52e60. http://dx.doi.org/10.1016/j.watres.2015.06.017.

Zietzschmann, F., Altmann, J., Ruhl, A.S., Dünnbier, U., Dommisch, I., Sperlich, A.,Meinel, F., Jekel, M., 2014. Estimating organic micro-pollutant removal potentialof activated carbons using UV absorption and carbon characteristics. Water Res.56, 48e55. http://dx.doi.org/10.1016/j.watres.2014.02.044.