Removal of Micropollutants, Facultative Pathogenic and Antibiotic Resistant Bacteria in a Full-scale Retention Soil Filter Receiving Combined Sewer Overflow † Environmental Impact

Combined sewer systems collect surface runoff as well as wastewater of industrial and domestic origin. During periods of heavy rainfall the capacity of the sewer system is exceeded and the overflow is discharged into receiving waters without any treatment. Consequently, combined sewer overflow (CSO) is considered as a major source of water pollution. This study investigates the effectiveness of a retention soil filter (RSF) for the removal of micropollutants as well as facultative pathogenic and antibiotic resistant bacteria from CSO. The removal of organic group parameters like total organic carbon was excellent and the removal efficiency for micropollutants of the RSF and the wastewater treatment plant (WWTP), which treats wastewater of the same origin during dry and normal weather conditions, was comparable. Compounds of high environmental concern like estrogens or certain pharmaceuticals, e.g. diclofenac, were completely eliminated or removed to a high degree during RSF passage. RSF treatment also reduced the number of E. coli, enterococci and staphylococci by 2.7, 2.2 and 2.4 log-units (median values), respectively. Obviously, some Staphylococcus species can better adapt to the conditions of the RSF than others as a shift of the abundance of the different species was observed when comparing the diversity of staphylococci obtained from the RSF influent and effluent. RSF treatment also decreased the absolute number of antibiotic resistant bacteria. The percentage of antibiotic resistant E. coli and staphylococci isolates also decreased during passage of the RSF, whereas the percentage of resistant enterococci did not change. For E. coli ampicillin and for enterococci and staphylococci erythromycin determined the antibiotic resistance level. The results demonstrate that RSFs can be considered as an adequate treatment option for CSO. The performance for the removal of micropollutants is comparable with a medium sized WWTP with conventional activated sludge treatment. The number of facultative pathogenic and antibiotic resistant bacteria was considerably decreased during RSF passage. However, as RSF effluents still contained antibiotic resistance genes and traces of micropollutants; receiving waters may still be at risk from negative environmental impacts. During periods of heavy rainfall the capacity of combined sewer systems can be exceeded and the overow is discharged without treatment into receiving waters. Consequently, water treatment options for such events are highly desirable as combined sewer overow is considered as a major source of surface water pollution. The use of retention soil lters is a growing technology and rst studies showed a great potential for the removal …


Introduction
Pollution of surface waters poses a threat for the aquatic environment.For this reason the European Union Water Framework Directive (EUWFD) was passed in 2000 to reduce acute and chronic toxicity of aquatic organisms, to minimize the accumulation of pollutants in ecosystems and to protect biological diversity and human health. 1 Environmental quality standards were established for an initial set of 33 priority substances, which must not be exceeded for a good surface water chemical status.In 2013 the list was extended by twelve new compounds.Additionally, three compounds (diclofenac, 17-b-estradiol and 17-a-ethinylestradiol) were put on a watch list, for which monitoring data are to be collected to support a decision on potential prioritization in the future. 2Some chemicals listed in the EUWFD are supposed to be mainly discharged by point sources like wastewater treatment plants (WWTPs).This holds true not only for compounds on the EU list of priority substances, but also for numerous other chemicals as well as facultative pathogenic and antibiotic resistant bacteria which might also affect the aquatic environment or human health.
One successful approach to minimize the discharge of micropollutants and microorganisms (bacteria, viruses, protozoan pathogens) by point sources is upgrading WWTPs by additional treatment technologies like ozonation or powdered activated carbon ltration. 3In Germany, however, two thirds of the sewer systems collect wastewater with domestic and industrial origin together with storm water runoff. 4,5During heavy rainfall up to seven times more water is collected in combined sewer systems than under dry weather conditions. 6s a consequence, the capacity of the sewer system and of WWTPs can be insufficient during heavy rainfalls as most facilities have a maximum treatment capacity equivalent to twice the dry weather discharge. 4t is therefore assumed that 30% to 50% of the annual storm water runoff is discharged as untreated combined sewer overow (CSO) into receiving waters, 5 including readily biodegradable or nonpolar micropollutants which would have been removed by microbial degradation or by sorption onto the sewage sludge under dry weather conditions.One measure to cope with high ows in the sewer system are storm water overow basins (SOBs), which provide an interim storage capacity for excess water, that is transferred to the WWTP aer the rain event.However, no microbial degradation of nutrients or micropollutants is achieved in the SOBs and when their storage capacity is exceeded, untreated storm water overow is also a source for contamination with micropollutants and potentially pathogenic microorganisms.This chemical and bacterial pollution can be a threat to human health when water is used for drinking water production, recreational activities or irrigation.Antibiotic resistance genes within the released microorganisms even increase this health threat and may contribute to their dissemination to the autochthonous microbial community.Therefore, CSO is an important source of pollution for receiving waters as it may represent large loads of microbial and chemical contaminants. 7,8r combined sewer systems retention soil lters (RSFs) are a treatment technology that helps to safeguard WWTPs and receiving waters from hydraulic, chemical and hygienic stress caused by the feed of excess water. 6RSFs are soil lters with a vertical ow-through passage that also provide additional intermediate storage and retention capacity.Basically, RSFs are constructed wetlands that only have an intermittent inow and receive overows from combined sewers, storm water or highway runoff and not predominantly wastewater. 9They provide water treatment by ltration, sorption and biological processes.In previous studies, RSFs showed excellent retention or removal efficiencies for nutrients like ammonium or phosphate as well as for suspended solids and oxidizable organic matter, expressed e.g. as chemical oxygen demand (COD) and biochemical oxygen demand aer 5 days (BOD 5 ). 4,6he interest in RSF technology is growing, but with many newly deployed lters there still is little experience with regard to long term performance.A survey among RSF operators in Germany (n ¼ 83) revealed that only 21% of all lters have been in service for more than ve years. 10he RSF analyzed in this study treats CSO of two SOBs and discharges into the river Schussen, a densely populated catchment area in Southwest Germany.The Schussen is a tributary of Lake Constance, which is an important drinking water reservoir and a popular recreation area. 11The elimination of nutrients, organic trace pollutants and facultative pathogenic bacteria by a RSF downstream of two SOBs is presented.For numerous micropollutants, facultative pathogenic and antibiotic resistant bacteria this study presents the rst baseline data for their retention by RSF.

Study site and sampling protocol
The RSF was put into operation in 2002.It receives the overow of two SOBs, which serve the city of Tettnang in Baden-Württemberg, Germany (Fig. 1).The effluent of the SOBs passes a cross-ow screen with a gap size of 5 mm.The wastewater solely originates from the two residential catchments of Tettnang (2450 population equivalents) as the wastewater from the local hospital is separately discharged to the WWTP.The effluent of the RSF is discharged into the Tobelbach, which ows into the Schussen river, one of the main tributaries of Lake Constance.The RSF has a surface area of 2000 m 2 which is planted with common reed (Phragmites communis).The formation of a colmation layer and the risk of surface clogging are reduced by the reed's stable rootstock growth.At the study site no further measures against clogging are necessary.The rooting and intermittent ooding also enables the penetration of oxygen into the subsurface, which stimulates the microbial activity.Once a year the reed is mown and removed from the lter surface.The lter layer has a thickness of about 80 cm and rests on a gravel drainage system (40-60 cm).At its base the RSF is furnished with a 2 mm non-permeable lining.On average, 40 to 60 impounding events occur per year.In more than one third of the events the inuent volume exceeds the capacity of the RSF of 2000 m 3 .Excess wastewater (storm water) is discharged to the Tobelbach without treatment.The RSF's effluent discharge point is controlled by a throttle valve limiting the outow to 10 L s À1 .Complete drainage takes about 60 h.The removal of group parameters, micropollutants, facultative pathogenic and antibiotic resistant bacteria presented here is based on the difference between concentration values determined for the inuent of the RSF aer SOBs and for the effluent aer the throttle valve.
The WWTP in Eriskirch which serves the two catchment areas depicted in Fig. 2 was used for comparison of treatment efficiencies.This WWTP treats a population equivalent of 50 000 per day and the maximum dry weather ow is 17 300 m 3 per day.Treatment consists of a screen, sand trap and a primary clarier, followed by biological treatment comprised of zones of denitrication and nitrication combined with simultaneous phosphorus elimination.The water is then directed to a secondary sedimentation basin and iron(III) chloride sulfate is added as a precipitating agent.The ocs are removed by ltration through layers of sand and anthracite before the water is discharged into the Schussen river.Results presented here are based on values determined for the inuent of the WWTP aer mechanical treatment and for the wastewater aer activated sludge treatment.
At the RSF there were ve sampling events between June 2012 and November 2013.The WWTP Eriskirch was sampled up to 11 times between May 2012 and August 2014.Though 187 micropollutants were measured, only a small number were dened as indicative parameters based on their nearly ubiquitous presence in WWTP inuents and CSO and were measured more frequently.Some more hydrophobic micropollutants like polybrominated diphenyl ethers (PBDE) and polycylic aromatic hydrocarbons (PAHs) were only screened at the RSF site.Furthermore, heavy metals and group parameters were measured.
During the sampling events, 250 mL of RSF inuent were sampled every 3 min and for the effluent 250 mL were sampled every 10 m 3 until the required total sample volume for all analyses was collected.In case of the WWTP samples, 24 h composite samples were analyzed.All samples were cooled down immediately with icepacks and transported to the laboratories, where they were stored at 4 C until sample preparation and analysis.

Analysis of micropollutants, heavy metals and waterchemical group parameters
For the analyses of micropollutants water samples were spiked with internal standards prior to extraction.Solid-phase extraction (SPE) or liquid-liquid-extraction (LLE) was used for preconcentration.Gas chromatography or liquid chromatography were coupled to different detector techniques ((tandem) mass spectrometry, nitrogen-phosphorus detection, diode array detection).Concentrations of micropollutants, heavy metals and group parameters were analyzed in inuent and effluent waters of the RSF.A complete list of all parameters with their respective limit of quantication (LOQ) can be found in the ESI (Table S1 †).Standard methods used for analysis are listed in Table 1.Non-standard methods are described briey in the text below.
Pharmaceuticals and some of their metabolites, articial sweeteners, iodinated X-ray contrast agents, benzotriazoles and antibiotics were pre-concentrated by SPE using SDB (J.T. Baker, Philipsburg, USA), Strata-X (Phenomenex, Aschaffenburg, Germany) or PPL Bond Elut (Agilent Technologies, Santa Clara, USA) polymeric adsorbent materials.The very polar antidiabetic drug metformin and its degradation product guanylurea were enriched with a cationic exchange adsorbent material (Strata-X-CW from Phenomenex) as described in Scheurer et al. 12 Sample pH, the water volume used for pre-concentration, elution solvents and the established liquid chromatographic methods were optimized for each substance group.The analytes were quantied using a 1290 HPLC system (Agilent Technologies) coupled to an API 5500 mass spectrometer (AB Sciex, Framingham, USA).
Polychlorinated biphenyls (PCBs), polycyclic aromatic hydrocarbons (PAHs) and chlorinated insecticides were extracted by LLE with 20 mL cyclohexane.The organic solvent was evaporated to a nal volume of about 0.5 mL and the compounds were measured by gas chromatography coupled to a tandem mass spectrometer (GC/MS/MS).
PBDEs were extracted with 25 mL cyclohexane.Residual water was removed with a sodium sulfate lled cartridge.The extract was evaporated to a volume of 0.2 mL and measured with GC/MS with negative chemical ionization.
Trialkylphosphates were enriched with a polymeric adsorbent (SDB) and cartridges were eluted with dichloromethane.GC/MS-MS was performed for separation and quantication using a TRACE GC Ultra gas chromatograph coupled to a TSQ Quantum XLS Ultra mass spectrometer (both Thermo Scientic, Waltham, USA).
For the SPE of phthalates self-packed glass SPE cartridges lled with Chromabond C18 Hydra material (Macherey Nagel, Düren, Germany) were used.Phthalates were analyzed using a Autosystem XL GC coupled to a Turbo Mass Gold MS (both Perkin Elmer, Waltham, USA).
Endocrine disrupting chemicals were also pre-concentrated by SPE with a polymeric adsorbent material (Strata-X from Phenomenex, Aschaffenburg, Germany).Aer elution of the analytes with acetone the extracts were evaporated to dryness and reconstituted with a derivatization mixture (MSTFA (Nmethyl-N-(trimethylsilyl) triuoroacetamide), trimethylchlorosilane and pyridine).Aer silylation (80 C for 45 min) a keeper was added and the derivatization reagent was removed in a gentle stream of nitrogen.The residue was reconstituted in cyclohexane and measured by a Trace GC TSQ Quantum XLS Ultra GC-MS/MS (Thermo Scientic).
Pesticides were enriched using 1 g IST Isolute C18 adsorbent (Biotage, Uppsala, Sweden) and analyzed aer elution with acetone by GC-MS using a 6890 5973 GC-MS system (Agilent Technologies).
Aliphatic amines were derivatized in the water sample with uorenylmethyloxycarbonyl chloride (FMOC) and pre-concentrated using 200 mg LiChrolut EN adsorbent material (Merck, Darmstadt, Germany).Measurements were performed with LC coupled to a uorescence detector (both Agilent Technologies).

Analysis of microbiological parametersenumeration, isolation and identication of E. coli, enterococci and staphylococci
Depending on the probable contamination level, either the membrane ltration method (according to ISO 7704H) or direct plating for E. coli, enterococci and staphylococci was used. 19. coli were grown on ECD-agar (Merck, Darmstadt, Germany).Aer incubation at 37 C for 20-24 h, under UV light blue uorescent colonies on the membrane (cellulose nitrate, pore size 0.45 mm, Ø 50 mm (Sartorius, Göttingen, Germany)) or on the agar surface (resulting from the beta-glucuronidase activity that hydrolyzes MUG present in the media) were counted as E. coli.Presumptive E. coli isolates were further tested for tryptophanase activity with Kovac's reagent.20 For Enterococcus spec., membrane lters with appropriate dilutions were incubated on azide nutrient pads (Sartorius) for 40-48 h at 37 C according to Slanetz and Bartley, followed by incubation on kanamycin-aesculin-agar (Merck) for 1 h at 44 C. 21 Red, pink and reddish brown colored colonies with positive aesculin reaction were counted based on ISO 7899-2.

View Article Online
For analysis of staphylococci, Chapman-Stone agar containing 0.05 g L À1 sodium azide was used.Aer incubation for 48 h at 37 C, colonies were counted and streaked on Mannitol-Salt agar.

Antibiotic susceptibility testing
The agar diffusion test according to the German standard method DIN 58940 was performed with 50 enterococci to test the antibiotics ampicillin (10 mg), chloramphenicol (30 mg), ciprooxacin (5 mg), and erythromycin (15 mg). 23To detect vancomycin resistant E. faecium and E. faecalis (VRE), the susceptibility of these isolates against vancomycin was tested as described by the Clinical and Laboratory Standards Institute (CLSI). 24Intrinsic low-level vancomycin resistance of E. gallinarum and E. casseliavus (vanC1-and vanC2-type) were not considered.
Susceptibility of 76 E. coli isolates was tested against the antibiotics ampicillin (10 mg), ciprooxacin (5 mg), cotrimoxazol (1.25 mg trimethoprim/23.75mg sulfamethoxazol) with agar diffusion tests according to DIN 58940 and against cefotaxim (5 mg) using the clinical breakpoints listed by the European Committee on Antimicrobial Susceptibility Testing (EUCAST). 23,25All ampicillin resistant isolates were additionally tested against ceazidim (30 mg) and cefpodoxim (10 mg) according to CLSI; 24 the inhibitory effect of clavulanic acid on the b-lactamase was checked as described by Bradford. 26ccording to the denition of the Robert-Koch-Institute, extended-spectrum-b-lactamase producer phenotypically showed resistance against cefpodoxim as well as ceazidim and/or cefotaxim and the inhibition of the enzyme by clavulanic acid.
Susceptibility of the 144 obtained Staphylococcus isolates against oxacillin (5 mg), ciprooxacin (5 mg) and erythromycin (15 mg) was tested by the disc-diffusion test with Mueller-Hinton agar according to DIN 58940 and to clindamycin (2 mg) according to CLSI. 23,24

Data analysis
Changes in the number of living E. coli, enterococci and staphylococci and in the percentages of antibiotic resistant isolates during passing the RSF were tested respective their statistical signicance using bilateral t-test.Bacterial numbers represent the median values of ve parallels.Removal of micropollutants was determined by calculating the average removal of all single sampling events.The variability of the data is expressed by the standard deviation.In the case of effluent values below the LOQ, the value half of the LOQ was used for calculation.Data are only presented if for at least half of the sampling events concentrations above the LOQ were measured.

Results and discussion
In the years 2012 and 2013, the volume of CSO treated in the RSF per month varied in a wide range between less than 100 m 3 up to more than 12 000 m 3 .The same holds for the number of days when a measureable amount of drainage could be observed in the RSF's effluent.The total number of drainage events at the RSF in Tettnang can be considered as rather high.A survey by Roth-Kleyer et al. showed that almost 60% of RSFs in Germany had less than twelve events per year. 10However, it has to be pointed out that such small numbers of events have to be regarded as contradictory to the establishment of reed vegetation.The intermittent loading of RSFs is important for the aeration of the subsurface.It has been demonstrated that sufficient oxygen supply in the soil lter enhances the removal of COD, ammonia and total nitrogen and that other factors like vegetation or temperature variations between 2 C and 20 C had only little effect on removal rates. 271 Group parameters DOC represents the total carbon concentration of all dissolved organic compounds.It is especially important for the depletion of oxygen and thus affects prevailing redox conditions and the degradation of inorganic nitrogen compounds like ammonia.The UV absorption of dissolved organic matter is measured by the parameter SAC 254 nm .As mainly unsaturated organic compounds absorb radiation at 254 nm, the parameter is used as a proxy for aromatic dissolved organic matter.Both, DOC and SAC 254 nm can be used to calculate the specic UV absorption (SUVA) coefficient (SAC 254 nm divided by DOC).A high SUVA indicates a high portion of aromatic compounds in the water.The good DOC-removal of more than 80% (Fig. 3) during a short residence time in the lter reects a high portion of readily biodegradable dissolved organic matter in the inuent of the lter.The average DOC concentration in the RSF inuent of 6.9 mg L À1 is in a concentration range typically measured in the effluent of municipal WWTPs, where readily biodegradable DOC has already been assimilated during activated sludge treatment.The SAC 254 nm is reduced in the RSF by about 50%.The more pronounced reduction of the DOC relative to the SAC 254 nm is expressed by an increase of the SUVA from 1.7 L (mg m) À1 to 3.1 L (mg m) À1 .Therefore, it can be concluded that less UV light absorbing saturated organic compounds are preferentially degraded in the RSF.
In similar subsurface treatment processes like soil aquifer treatment (SAT) the same observations have been made. 28,29For a SAT eld in Israel a preferential degradation of non-aromatic organic compounds was reported as an increase of SUVA from the percolation basin to a vertically located observation well (detention time 1.5 months) was observed. 30In contrast to the SAC 254 nm , the SSC 254 nm takes the scattering of light due to suspended matter into account. 31As expected, a higher removal rate of SSC 254 nm in the RSF Tettnang could be observed (Fig. 3) due to the removal of suspended particles in the RSF by physical ltration processes.Higher removal rates of TOC compared to DOC and the pronounced improvement in turbidity can also be attributed to the removal of suspended solids by the RSF.
The COD removal of 80 AE 10% is in the range found in literature for other RSFs in Germany.Mean removal rates from 40% up to more than 90% have been reported. 4,32

Micropollutants
Most of the 187 micropollutants analyzed were not detected in the RSF inuent or effluent on a regular basis.Micropollutants with wastewater origin are diluted by storm water and consequently if present their concentrations oen decreased below the respective LOQ.
Pharmaceuticals and their metabolites.From a total of 79 pharmaceuticals (including antibiotics and some metabolites and excluding X-ray contrast media) only 24 were regularly found in more than half of the WWTP inuent samples analyzed within this study.Only nine out of those 24 compounds were detected in the RSF inuent in more than 50% of the samples.Generally, the pharmaceutical residues with the highest concentrations in the WWTP inuent were also detected in the RSF inuent and are depicted in Fig. 4. Pharmaceuticals additionally found in the inuent of the WWTP Eriskirch belonged to several therapeutic classes with betablockers (atenolol, bisoprolol, sotalol) and antibiotics (clarithromycin, sulfamethoxazol, trimethoprim, ciprooxacin, ooxacin) as the most prevailing compound groups.
For most pharmaceuticals removal efficiencies were similar for activated sludge treatment in the WWTP and in the RSF.Pharmaceutically active compounds which are commonly readily biodegradable in WWTPs were also removed in the RSF to a high degree.Examples are paracetamol (98%) and ibuprofen (94%).For the latter substance good removal from 0.5 mg L À1 to a concentration below the LOQ has been reported for another RSF site. 33or diclofenac, metoprolol and 10,11-dihydro-10,11-dihydroxycarbamazepine, a metabolite of the antiepileptic drug carbamazepine, removal rates were slightly higher than in the WWTP.5][36][37] The fact that diclofenac is not readily biodegradable is attributed to its chlorine group.Wastewater characteristics as well as treatment conditions also inuence removal rates in WWTPs. 37The mean removal of diclofenac in the WWTP Eriskirch was 14% (n ¼ 11) aer activated sludge treatment and 37% post sand lter at the end of the treatment train, whereas the RSF removed 89% on average (all effluent values < LOQ).This is in agreement with good diclofenac removal rates reported for other RSFs.In Altendorf (Germany) the mean concentration of diclofenac (0.14 mg L À1 ) was reduced to below the LOQ. 33Tondera et al. reported a diclofenac removal of more than 70% for two RSF in North-Rhine-Westphalia (Germany). 9,380][41] Sorption in water-sediment systems revealed negligible sorption tendency for metoprolol. 42The removal of 36% and 84% in the WWTP Eriskirch and the RSF Tettnang, respectively should be attributed mainly to biodegradation although metoprolol is present in its cationic form at environmental pH values and might also be retained by negatively charged solid particles.
The negative elimination of carbamazepine (see Fig. 4) can be presumably explained by the cleavage of glucuronide conjugates, which are formed in the human body, back to carbamazepine.The fact that carbamazepine is not removed at all may reect the prevailing redox conditions in the RSF, as the compound is persistent in an aerobic environment but can be degraded under anaerobic conditions. 43or the antidiabetic drug metformin concentrations in WWTP inuents up to 100 mg L À1 have been reported and as a rule removal rates are higher than 90%. 12,44This was also the case for the WWTP Eriskirch, but due to high inuent concentrations still several hundred ng L À1 could be detected aer biological wastewater treatment.However, as for most micropollutants no complete mineralization takes place during biological treatment, but degradation products are formed.In the case of metformin, transformation into guanylurea (diaminomethylideneurea) takes place. 44,45Consequently, a maximum concentration of 96 mg L À1 guanylurea was measured in the effluent of the WWTP Eriskirch.Removal of metformin was high and comparable to the WWTP Eriskirch, but again traces of metformin were detected in the RSF effluent.Interestingly, no guanlyurea was detected in the effluent of the RSF.
The residence time and biological activity of the RSF appear to be sufficient to degrade or retain both metformin and its transformation product and therefore, the RSF has a better performance compared to conventional activated sludge treatment in WWTPs.
Benzotriazoles.The corrosion inhibitor 1H-benzotriazole is used in dishwashing liquids, anti-freezing and deicing uids.Its elimination in the WWTP Eriskirch was incomplete (mean value 42%) and an average value of 5.1 mg L À1 could still be measured in the treated wastewater.7][48] The mean removal of 1H-benzotriazole in the RSF was better (66 AE 20%) but not signicantly different.The compound was found in production wells at river bank ltration sites aer a residence of several months in the saturated subsurface. 49However, sorption onto soils is negligible 50 and the decrease in concentration during RSF passage can most likely be attributed to a slow biodegradation.Greater persistence in aquatic environments has been reported for the 4-methyl analogue 4-methylbenzotriazole. 48A decreasing 1H-benzotriazole/4-methylbenzotriazole ratio with increasing residence time was observed in the partially closed water cycle in Berlin, Germany. 49In comparison to 1H-benzotriazole, the removal for 4-methylbenzotriazole was lower at both sites, the WWTP and the RSF.
Estrogens.The main sources for estrogenic activity in wastewater are natural hormones and synthetic compounds like 17a-ethinylestradiol (EE2) and, to a smaller extend, alkylphenols.The removal of such compounds during conventional wastewater treatment is usually good, but oen incomplete.Both, biodegradation and sorption onto sludge contribute to the elimination of these compounds. 51A good compilation of estrogenic compounds, their effects and fate during wastewater treatment can be found in Teske and Arnold. 52Estrone (E1), 17b-estradiol (E2) and estriol (E3) were detected in every inuent sample of the WWTP Eriskirch with mean concentrations of 140 AE 130 ng L À1 , 37 AE 16 ng L À1 and 180 AE 48 ng L À1 , respectively.The removal during wastewater treatment was in almost all cases complete and only once E1 was detected with 2.1 ng L À1 above the limit of quantication (LOQ ¼ 0.5 ng L À1 ) aer activated sludge treatment.Only E1 and E3 were detected on a regular base in the RSF inuent but due to dilution with lower concentrations of 22 AE 19 ng L À1 and 25 AE 17 ng L À1 .In Fig. 4 the removal of both compounds appears to be slightly lower than in the WWTP, but concentrations were lower than the detection limit.The apparent lower removal rate is simply due to lower inuent concentrations, which were compared with the same effluent values as above (half of the LOQ).
Articial sweeteners.For the articial sweeteners saccharin and cyclamate, good removals were observed for both the WWTP and RSF.This is in accordance with high removal rates for both compounds during wastewater treatment reported in literature. 53Acesulfame was retained with a mean value of 46% in the WWTP Eriskirch.However, the removal rate varied over a wide range from À5% to 90% with no obvious trend or seasonal correlation like it was reported for some pharmaceuticals in the past. 54The results are against expectation as in previous publications acesulfame was reported to be recalcitrant and was only poorly removed in WWTPs. 53,55An adaptation of microorganisms cannot be the only reason for the degradation as aer sampling events with high removal a pronounced stability of acesulfame was observed again.However, recent studies have questioned the compound's stability and report a partial degradation in certain environmental compartments like river bank ltration sites. 56The reason for the heterogeneous stability of acesulfame is yet unknown.Tran et al. reported a poor removal of acesulfame in batch tests with nitrifying sludge, but removal was increased with increasing ammonium concentrations. 57Maximum acesulfame concentrations aer activated sludge treatment or in the effluents of the RSF were 19 mg L À1 and 1.3 mg L À1 respectively and thus among the highest values observed for micropollutants.
Antibiotics.Within this study 37 antibiotics comprising different classes (macrolides, tetracyclines, penicillins, sulfonamides, uoroquinolones) were analyzed.Only two of them (sulfamethoxazole and trimethoprim) were detected once in the inuent of RSF Tettnang with a maximum concentration of 0.12 mg L À1 for sulfamethoxazole.Beside these two antibiotics azithromycin, clarithromycin, roxithromycin, ciprooxacin and ooxacin were detected in the inuent of the WWTP.With two exceptions the concentration levels were below 1 mg L À1 , but all antibiotics mentioned before could be found at least once above the LOQ aer activated sludge treatment.No conclusions about the RSF performance in terms of the removal of antibiotics can be drawn as values below LOQ can be a matter of dilution and are no proof for the complete absence of a certain compound.Assuming similar elimination processes and removal rates in the WWTP and the RSF, traces of antibiotics might end up in receiving waters.Other studies have pointed out that presence and long-term exposure even to rather low concentrations of antibiotics might be a reason for the selection of resistant bacteria and their spreading in the environment. 58,59he source of a compound is an important factor for its presence in combined sewers.If compounds which derive mainly from domestic wastewater are biodegradable, their absolute amount released to the environment will increase due to the absence of treatment of CSO overow.However, there will be no change in the absolute amount of recalcitrant compounds.
In contrast, the absolute amount of micropollutants which are only re-mobilized, washed off or introduced by surface runoff during rain events will increase in any case. 60Biodegradable micropollutants are discharged by CSO overow, while recalcitrant substances are released by both CSO overow and WWTP effluent to receiving waters.
One example for the appearance of a micropollutant in runoff is mecoprop, which is used as pesticide but also as biocide, e.g. in roof sealings.As a consequence it can be washed off and released into the environment during rainfall events. 61he compound was not detected in samples from the WWTP inuent in Eriskirch.In contrast, in two RSF inuent samples mecoprop was measured with a maximum concentration of 1.4 mg L À1 .However, mecoprop was efficiently removed during RSF passage to a nal concentration below the LOQ.
Another compound group typical for the phenomenon of remobilization during rainfall events are polycyclic aromatic hydrocarbons (PAHs).Since their sorption tendency to particulate matter is high, it is of special importance for their elimination that the concentration of solids is reduced to a minimum during treatment.Filterable solids were reduced by more than 97% in the RSF Tettnang.A 93% retention was reported by Tondera et al. for a RSF six years in operation. 38AHs were analyzed only for the RSF site.When detected in the RSF inuent removal efficiency was between 70-90%.
In summary the performance of the RSF for the removal of micropollutants can be most likely attributed to a combination of sorption and biodegradation.Compounds not discussed above were not detected in the RSF inuent or the calculation of their removal rate was not feasible due to the lack of frequent detection.

Removal of microorganisms
Concentrations of all microorganisms studied were signicantly lower in the RSF effluent than in the inuent (Fig. 5).Removal rates were variable but comparable with those in WWTPs.The number of E. coli and enterococci decreased by 2.7 (median value; values for individual samples ranged from 2.1-3.2) and 2.2 log-units (median value; values for individual samples ranged from 0.9-2.8),respectively.The maximum elimination achieved was in a similar range as reported for other constructed wetlands planted with Phragmites spec. 33,62,63he median elimination rate of fecal indicator bacteria (FIB) exceeded the observations made by Tondera et al., who observed removal rates of 1.1 log-units for E. coli and enterococci. 9Comparing the number of staphylococci in the inuent and effluent of the RSF, their number signicantly decreased by about 2.4 log-units (median value; values for individual samples ranged from 2.2-3.0 log-units).Faria et al. reported a removal of about 2 log-units in a conventional activated sludge treatment plant. 64A higher retention of up to 4 log-units (median value) was detected by Heß and Gallert. 65Comparing elimination efficiencies of WWTP and RSF, it has to be considered that the cell count of staphylococci was already lower in the inuent of the RSF (10 3 CFU per 100 mL, median value) compared to the inuent of a WWTP (Faria et al.: 10 5 CFU per 100 mL; Heß and Gallert: 2.7 Â 10 4 CFU per 100 mL (median value)). 64,65In consequence, the removal expressed in log-decades has to be lower.Remarkably, the concentration of staphylococci in the effluent of the RSF (median value: 3.8 CFU per 100 mL) was in the same order of magnitude or lower (Faria et al.: 10 2 CFU per 100 mL; Heß and Gallert: 2.3 CFU per 100 mL) 64,65 compared to the WWTP effluents.
Factors like retention time, characteristics of the lter material, organic matter concentration and biocenosis composition differ from site to site, however many authors suggested predation in the colmation layer or the subsurface of the RSF and adsorption as main factors for the retention of bacteria in RSFs. 63,66,67Bacteria and viruses are attached to solids or in the free water phase 9 and like for hydrophobic micropollutants which are preferentially bound to particles, Passerat et al. observed a rst-ush effect for E. coli in a CSO discharge caused by particle attached cells. 7Based on the analysis of TOC and SSC, a considerable removal of suspended matter can be assumed and at least partial removal of bacteria attached to particles by ltration is likely.However, Tondera et al. observed no correlation between the removal of total suspended solids and the removal of E. coli or enterococci. 9Key factors and mechanisms governing removal processes comprise a complex combination of chemical, physical and biological factors, which was not accounted for in this study and should be investigated by further in situ studies.In this context, it would also be interesting to perform long-term experiments investigating whether the RSF sediment is likely to change from sink to source.
Nola et al. studied different soil columns with regard to their retention efficiencies. 68The removal of staphylococci was in the order of 99.99%.Comparing their results with those of the RSF in Tettnang is difficult because of higher staphylococci concentrations in the inuent of the columns (2.9 Â 10 7 CFU per 100 mL; RSF, median value: 10 3 CFU per 100 mL) with a retention efficiency of 6.7 log-units (mean value) and different construction and operating factors.Overall, the efficiency of the retention of particles depends on the construction and the operating factors of the RSF.Sidrach-Cardona and Bécares identied a subsurface-ow constructed wetland planted with Phragmites spec.as the best combination among several types of construction and reported a removal of total coliforms, E. coli and Enterococcus spec.between 2.5 and 3 log-units (mean value). 63The construction of the treatment facility and the removal efficiency are similar to the results obtained in this study.View Article Online

Species diversity of staphylococci
Comparing the diversity of staphylococci obtained from the inuent (71 isolates) and the effluent (73 isolates) of the RSF, a shi of the abundance of the different species could be observed: whereas the percentage of isolates belonging to the Sciuri-group decreased from 59.3% in the inuent to 36.0% in the effluent, the percentage of isolates clustered to the Saprophyticus-group remained stable (inuent: 37.3%; effluent: 36.0%).However, within the Saprophyticus-group the abundance of the respective species changed: the percentage of Staphylococcus xylosus on isolates belonging to the Saprophyticus-group increased from 10.9% in the inuent to 66.7% in the effluent.Obviously, some Staphylococcus-species can better adapt to the conditions of the RSF than others.Coinciding with Faria et al., Heß and Gallert reported that species of the Saprophyticus-group were dominant in raw sewage (Faria et al.: 78.7% of the isolates were representatives of the mentioned group; Heß and Gallert: 51.6% of the isolates were species of the Saprophyticus-group and 31.8% of them were identied as S. xylosus). 64,65Furthermore, the study of Heß and Gallert showed that species of the Sciuri-group were more abundant in river water than in sewage and the percentage of S. saprophyticus ssp.saprophyticus decreased in favor of S. xylosus in receiving water bodies. 65Using the percentage of isolates belonging to the Sciuri-group as well as the percentage of S. xylosus on the abundance of isolates belonging to the Saprophyticus-group as "less-sewage derived" markers, a further characterization of the inuent and the effluent of the RSF is possible: based on diversity and abundance of the staphylococci, the inuent of the RSF was a mixture of raw sewage diluted with a signicant volume of surface runoff.The Staphylococcus-species diversity found in the effluent of the RSF was dominated by species rather associated with river water indicating that these species are better adapted to the condition of the RSF.

Antibiotic resistant E. coli, enterococci and staphylococci
The percentage of antibiotic resistant E. coli isolates (35 isolates obtained from the inuent and 41 strains obtained from the effluent were tested) signicantly decreased during passing the RSF from 17% in the inuent to 10% in the effluent (level of signicance 0.05).If an isolate was resistant, it was resistant against ampicillin.In the inuent two isolates were additionally resistant against cotrimoxazole, one of them also against cefotaxim classifying it as an extended spectrum b-lactamase (ESBL) producer.The fact that ESBL-producers can be detected by culture based approaches points out that they are present in respective numbers in sewage.In the effluent, one ampicillin resistant E. coli was additionally resistant to cotrimoxazole.Luczkiewicz et al. found a comparable resistance pattern for E. coli isolated from a municipal wastewater treatment plant, but overall the resistance level was lower. 69imilar to E. coli, the percentage of antibiotic resistant staphylococci (71 isolates obtained from the inuent and 73 strains obtained from the effluent were tested) was signicantly lower in the effluent of the RSF (6.8%) than in the inuent (16.9%, level of signicance 0.05).The percentage of isolates resistant to erythromycin determined the resistance level.None of the obtained staphylococci was constitutively resistant against clindamycin and/or ciprooxacin.The fact that 5 isolates obtained from the inuent were resistant against oxacillin showed that the antibiotic resistance gene mecA is present in sewage.The observed resistance pattern was comparable to that described for raw sewage, 64,65 but the overall percentage of resistant isolates was lower.A possible explanation for this observation is based on the species diversity.Some species were more frequently resistant than others: for instance, 24.7% of the S. saprophyticus ssp.saprophyticus isolates, a species dominating in raw sewage, was resistant against one of the tested antibiotics, whereas "only" 10.8% of S. xylosus isolates, which rather could be isolated from river water, was resistant. 65Consequently, based on the found resistance levels of E. coli and staphylococci as well as on the described Staphylococcus-species diversity, the inuent of the RSF cannot be directly compared to the inuent of a WWTP under dry weather conditions.
The percentage of resistant enterococci (32 isolates obtained from the inuent and 18 strains obtained from the effluent were tested) did not signicantly change during passing the RSF (inuent: 34.4%; effluent: 38.9%).This observation might be an effect of the small number of isolates; a higher number is needed to conrm this result.Erythromycin resistance determined the resistance level of enterococci in the respective sample.Two isolates in the inuent and one isolate obtained from the effluent were additionally resistant against cipro-oxacin.One Enterococcus isolate of the effluent was resistant against the last resort antibiotic chloramphenicol.None of the obtained isolates were high-level resistant against vancomycin and/or ampicillin.Luczekiewicz et al., who isolated enterococci from sewage of a municipal WWTP, also found the described resistance pattern, but as already seen for E. coli and staphylococci the overall percentage of resistant isolates was lower. 69ven if the resistance level of enterococci did not decrease, the absolute concentrations of antibiotic resistant E. coli, enterococci and staphylococci in the effluent of the RSF were about 2.1 and 2.9 log-units lower than in the inuent.Nevertheless, antibiotic resistant bacteria and antibiotic resistance genes were released into the receiving Tobelbach.At the moment, the remaining risk and the effect on autochthonous microorganisms in receiving waters cannot be predicted.Some authors discussed the long hydraulic-retention time as well as the exposure to low (not inhibitory) concentrations of different antibiotics as promoting factors for the spread of antibiotic resistance genes. 58,59Comparing the detected resistance level of the inuent and the effluent of the RSF in Tettnang, there were no indices at present for the spread of resistance determinants.

Conclusions
The effectiveness of a retention soil lter (RSF) for the removal of micropollutants, facultative pathogenic and antibiotic resistant bacteria from CSO was investigated.The removal efficiency for micropollutans was comparable to activated sludge treatment in a nearby WWTP.WWTP effluent still is an important source of water pollution with micropollutants.The conventional treatment in WWTPs is especially important for biodegradable mircopollutants or compounds which tend to adsorb onto suspended and settleable solids.First measures for upgrading WWTPs with ozonation or activated carbon treatment units on a full-scale are taken and will improve the overall removal efficiency.However, the whole treatment train is only successfully applied during dry weather conditions.Therefore, a complementary treatment option with an active microbial environment and a reliable adsorption capacity is needed for events of heavy rainfall, when WWTPs and sewage overow basins cannot cope with the additional amount of water.The results of this study support the extension of the RSF technology as a low cost addition to overowing combined sewer overow basins to control the discharge of standard pollutants and micropollutants, and the number of facultative pathogenic and antibiotic resistant bacteria.Detailed investigation of the subsurface of full-scale RSFs should be conducted for a better understanding of the basic processes involved in the removal of micropollutants, facultative pathogenic and antibiotic resistant bacteria.

Fig. 1
Fig. 1 Map of Germany and scheme of the study area in the southwestern federal state Baden-W ürttemberg (by courtesy of Katharina Peschke, University of T übingen, Germany.The map of the study area is based on OpenStreetMap.Mapdata: ©OpenStreetMap contributors, license: http://opendatacommons.org/licenses/dbcl/1.0/).

Fig. 2
Fig. 2 Flow diagram of the study site (adopted from Dr.-Ing.Jedele and Partner GmbH, Stuttgart, Germany).

Fig. 3
Fig. 3 Mean removal of physical-chemical and group parameters by RSF treatment.Error bars represent standard deviation.

Fig. 4
Fig. 4 Comparison of removal rates of organic trace pollutants in a retention soil filter and a wastewater treatment plant after activated sludge treatment, n.a.¼ not analyzed.Error bars represent standard deviation of removal.

Fig. 5
Fig. 5 Colony forming units (cfu) (mean values) of total and resistant E. coli, enterococci and staphylococci per 100 mL in the influent and effluent of the RSF Tettnang.