How a Complete Pesticide Screening Changes the Assessment ofSurface Water QualityChristoph Moschet,*,†,‡ Irene Wittmer,†,∥ Jelena Simovic,† Marion Junghans,§ Alessandro Piazzoli,†,‡

Heinz Singer,† Christian Stamm,† Christian Leu,∥ and Juliane Hollender*,†,‡

†Eawag, Swiss Federal Institute of Aquatic Science and Technology, 8600 Dubendorf, Switzerland‡Institute of Biogeochemistry and Pollutant Dynamics, ETH Zurich, 8092 Zurich, Switzerland§Swiss Center for Applied Ecotoxicology Eawag/EPFL, 8600 Dubendorf, Switzerland∥Federal Office for the Environment (FOEN), 3000 Bern, Switzerland

*S Supporting Information

ABSTRACT: A comprehensive assessment of pesticides insurface waters is challenging due to the large number of potentialcontaminants. Most scientific studies and routine monitoringprograms include only 15−40 pesticides, which leads to error-prone interpretations. In the present study, an extensive analyticalscreening was carried out using liquid chromatography−high-resolution mass spectrometry, covering 86% of all polar organicpesticides sold in Switzerland and applied to agricultural or urbanland (in total 249 compounds), plus 134 transformation products;each of which could be quantified in the low ng/L range. Fivemedium-sized rivers, containing large areas of diverse crops andurban settlements within the respective catchments, were sampledbetween March and July 2012. More than 100 parent compoundsand 40 transformation products were detected in total, between 30 and 50 parent compounds in each two-week compositesample in concentrations up to 1500 ng/L. The sum of pesticide concentrations was above 1000 ng/L in 78% of samples. Thechronic environmental quality standard was exceeded for 19 single substances; using a mixture toxicity approach, exceedancesoccurred over the whole measurement period in all rivers. With scenario calculations including only 30−40 frequently measuredpesticides, the number of detected substances and the mixture toxicity would be underestimated on average by a factor of 2.Thus, selecting a subset of substances to assess the surface water quality may be sufficient, but a comprehensive screening yieldssubstantially more confidence.

■ INTRODUCTION

Surface waters in agriculturally and urban influenced catch-ments contain a large number of pesticides1,2 and theirtransformation products (TPs),3,4 which can pose risks toaquatic organisms even at low ng/L concentrations.5,6 Theexposure of surface waters to pesticides is heavily dependent onlocal conditions (e.g., land use, pesticide application, weatherconditions, soil type, topography, sewer type) and therefore canbe spatially and temporally variable. In the Swiss and Europeanlegislation, there is a distinction between active ingredients inplant protection products (further referred to as PPPs) andactive ingredients in biocide products (further referred to asbiocides). PPPs are used to protect plants and are thereforeallowed to be used in agriculture and private gardens. Biocidesare used to protect materials, such as wood, building facades,roofs, or for in-house applications, and are therefore used fordomestic, industrial, and/or commercial applications. The maintransport pathways of PPPs are diffuse via surface runoff,leaching to field drains, and spray drift from agricultural fields(e.g., ref 7); the main sources for biocides are effluents from

wastewater treatment plants (WWTPs), rain gutters, andcombined sewer overflows.8 A considerable number of activeingredients can function both as a PPP and as a biocide.Because of the high number of compounds and their high

spatial and temporal variability, it is difficult to design a propermonitoring campaign to assess the exposure and risk to aquaticorganisms in surface waters which will provide a high level ofconfidence. The results of any monitoring program are, to alarge part, dependent on the sampling location, sampling time,and sampling strategy.Moreover, a crucial factor in any monitoring program is the

selection of the substances which are to be investigated.Routine analysis usually focuses on just 15−40 analytes;9

analysis is mostly carried out by liquid chromatography−massspectrometry (LC−MS/MS) using triple quadrupole instru-

Received: January 22, 2014Revised: April 17, 2014Accepted: April 27, 2014Published: May 12, 2014

Article

pubs.acs.org/est

© 2014 American Chemical Society 5423 dx.doi.org/10.1021/es500371t | Environ. Sci. Technol. 2014, 48, 5423−5432

ments. The selection of the substances is normally done basedon expert knowledge and analytical feasibility. Until now,surface water monitoring programs have usually focused onselected herbicides (e.g., refs 1 and 10), in general becauseherbicide concentrations were expected to be higher thaninsecticide or fungicide concentrations. This assumption isbased on the interpretation of previous monitoring (e.g., refs 1and 9) and sales data,11 as well as environmental fateproperties.12 Modern insecticides (e.g., neonicotinoids13,14),fungicides (e.g., azoles15−17), and also TPs18,19 have only rarelybeen included in monitoring programs. However, insecticidesoften show ecotoxicological effects at lower concentrations thanherbicides.12 This implies that, despite their low concentrations,the ecotoxicological relevance of insecticides cannot beoverlooked.6

With continuously shifting pesticide management practicesand authorization, the relevant substances change over time andmay even be regionally different. A full assessment of allregistered pesticides and TPs would be highly desirable. Recentadvances in LC−HRMS/MS and appropriate software toolsallow efficient screening of an almost unlimited number ofpolar organic substances in a single run.20,21 Our hypothesis isthat using a comprehensive analytical screening has a largeinfluence on the exposure assessment, and will therefore changethe risk assessment of surface waters. In fact, we postulate thatonly by conducting a full analytical assessment of all potentiallyoccurring pesticides together with the relevant ecotoxicologicaldata, a complete mixture toxicity estimation can be done. Dueto the advancement of analytical instrumentation, the need formixture toxicity assessments22−25 may become even moreimportant than in the past.To test this hypothesis, we carried out a comprehensive

pesticide screening of almost all authorized, polar, organic PPPs(i.e., herbicides, fungicides, insecticides) and biocides, as well asa large number of pesticide TPs by using LC−HRMS/MS. Theaim was to gain an overview of the diversity of pesticides inmedium-sized Swiss rivers, to determine spatial and temporaldifferences due to land use and application season, and tocomprehensively assess the ecotoxicological risk in the rivers.Compliance with environmental quality standards (EQS) forsingle substances was investigated, as well as mixture toxicityassessments. Finally, it was possible to evaluate how exposureand risk assessments change when pesticide screening is lesscomplete, i.e., by carrying out different assessments with only asubset of analytes.

■ MATERIALS AND METHODSSampling Site. Five river catchments distributed over the

Swiss plateau (see map in Figure 1) were selected in order tomonitor the large diversity of PPPs and biocides occurring inSwitzerland. The catchments were between 38 and 105 km2

and differed widely regarding arable crop densities, densities ofspecial crops, urban areas, and WWTP discharges (see barcharts in Figure 1). Hence, most potential sources of pesticidecontamination in Switzerland were covered. The five riverswere classified as medium-sized (stream order 3−4 afterStrahler26).Sampling. Nine biweekly (March 19−July 23, 2012) time-

proportional composite water samples were collected from eachriver with automatic sampling devices (Isco sampler) with 60-min subsampling intervals. Weekly composites of 1 L weretaken (cooled on-site) and transported to the laboratory. Twoweekly composites were mixed to biweekly samples in the

laboratory, and stored at −20 °C until analysis. During thesampling period, there were 6−10 rainfall events (see Figure 2,top). Overall, a relatively dry spring (discharge in the river was30−60% less than the average of the previous 12 years) wasfollowed by a summer with an average river discharge (seeSupporting Information (SI) Figure SI-1).

Substance Selection. At the outset of substance selectionfor this monitoring, all polar, synthetic, organic PPPs (i.e.,herbicides, fungicides, insecticides) that have been registeredand sold in Switzerland between 2005 and 201128 (220substances) and all polar, synthetic, organic substances listed asbiocides in Switzerland29 (109 substances) were included in thestudy (Table 1, see SI 2.1 for all substances). Forty of theselected compounds were registered both as a PPP and as abiocide. Only substances that were expected in the water phasewere considered. Substances with log Kow > 5, includingpyrethroid insecticides and quaternary ammonia cations, wereexcluded due to their sorption to organic matter. For thebiocides, substances with low half-lives in water (< 1 d12) werealso excluded. In addition to the parent compounds, 134 TPs,

Figure 1. Location and characteristics of the five investigatedcatchments. Green: catchment area, red: sampling location. Bar chartsshow the densities of the most important crops (%) and the number ofinhabitants connected to WWTPs in the catchments.27

Environmental Science & Technology Article

dx.doi.org/10.1021/es500371t | Environ. Sci. Technol. 2014, 48, 5423−54325424

covering a variety of pesticide classes, taken from the footprintdatabase,12 were included in the analysis (see SI 2.1).Analytics. One hundred thirty-six out of 289 measurable

pesticides and 54 out of 136 TPs were analyzed with an offlinesolid phase extraction (SPE) LC-electrospray-ionization−HRMS/MS method and quantified with target screeningusing reference standards and isotope-labeled internal stand-ards.21 In brief, 1 L of sample was passed over a multilayeredcartridge containing Oasis HLB, Strata XAW, Strata XCW, andIsolute ENV+ in order to enrich a broad spectrum ofsubstances. Elution was done subsequently by ethyl acetate/methanol (50%/50%) with 0.5% ammonia and ethyl acetate/methanol (50%/50%) with 0.5% formic acid. Combined neutralextracts were evaporated to 0.1 mL and reconstituted with 0.9mL of nanopure water. The chromatographic separation wascarried out with a XBridge C18 column using nanopure andmethanol acidified with 0.1% formic acid as eluents. HRMS andMS/MS data were generated on a QExactive (Thermo FisherScientific Corporation). Full scan with resolution (R) of140 000 and data-dependent MS/MS (R = 17 500, Top 5) withseparate runs for positive and negative ionization wereacquired.The presence of the remaining parent compounds and TPs

was checked by suspect screening using a recently developedapproach.21 In comparison to nontarget screening, theinformation about the chemical structure is available a prioriin the suspect screening.20 Briefly, after an automatic peakpicking of the exact masses of all theoretically measurablesubstances from the MS full scan, different filter criteria (blanksubtraction, peak area, signal-to-noise, peak symmetry, andisotopic pattern) were applied to reduce the number of falsepositives. For confirmation and quantification of the filteredpeaks, reference standards were purchased and the retentiontimes and MS/MS spectra were compared. The automatedsuspect screening method led to slightly higher LOQs than thetarget method and thus, approximately 30% of low-intensitypeaks close to the LOQs were missed. All detected suspectswere confirmed by a reference standard and were subsequentlyquantified in the samples.Risk Assessment. For each detected pesticide, a chronic

environmental quality standard (AA-EQS), derived in line withthe Technical Guidance Document (TGD)30 of the WaterFramework Directive (WFD) of the European Union, wassearched for in the literature (values are listed in SI 2.2). For 22substances, AA-EQS values were derived in-house.31 For 14substances, where no AA-EQS value was available, an ad hocEQS value was derived using a limited data set (see SI 2.3).Only parent compounds were considered in the ecotoxico-

logical assessment because not enough is known about thetoxicity of TPs.Risk quotients (RQs, i.e., measured concentrations in the

composite samples divided by the AA-EQS)32 were calculatedfor each substance in each sample. Initially, risk assessment wasbased on single substances. Subsequently, mixture toxicity wasassessed in a tiered process. First, as a worst case estimation ofthe mixture toxicity, the RQ of each detected substance wassummed in a sample (concentration addition model22,24,25).Second, the risk was categorized into the three pesticide classesherbicides, fungicides, and insecticides. Thereby, only RQs ofsubstances from the same pesticide class were summed. Thisrealistic worst-case approach was done in order to determinewhich substance class most affects the total risk.

Scenario Calculations. The results from the comprehen-sive screening were evaluated based on the number of detectedsubstances, the sum of pesticide concentrations, the singlesubstance toxicity, and the mixture toxicity in each sample.Then, three scenarios were created which included only asubset of compounds and the same evaluations were done withthese results. In the first scenario, a “Swiss Monitoring”scenario, only the 32 most frequently measured pesticides bySwiss authorities were included (selected substances, see SI2.1). In the second scenario, 20 studies were selected frominternational scientific literature which included multiresiduemethods for pesticide measurements in surface waters andwhich were from various countries with agricultural practicessimilar to those of Switzerland. The 36 most frequentlyinvestigated pesticides in those studies were used for theevaluation (“International Studies” scenario, see SI 2.1).Although this literature search was not complete, it gives aclear overview of which pesticides are considered to be mostimportant in surface waters based on the knowledge of theinternational scientific community. In the third scenario, thepesticides that are listed as priority substances in the WFD33

and have been registered in Switzerland (15 pesticides) wereconsidered (“WFD Pesticides” scenario, see SI 2.1).

■ RESULTS AND DISCUSSION

Analytical Coverage. With the SPE LC-HRMS/MSanalysis (including target and suspect screening), 91% ofpolar organic PPPs registered in Switzerland and 81% ofpotentially relevant biocides were covered (Table 1). Thus, thiscomprehensive screening allows for a thorough exposure andrisk assessment of pesticides in Swiss surface waters.The advantages of this method are 3-fold: the nonspecific

enrichment on the multilayer cartridge, together with a goodchromatographic separation, and very specific and sensitive

Table 1. Number of Investigated and Analytically Covered Pesticides, Separated into Plant Protection Products (PPPs, Dividedinto the Classes Herbicides, Fungicides, and Insecticides) and Biocides

pesticidesa PPPs herbicides fungicides insecticides biocides

investigated substances 289 220b 105 73 42 109c

measurable 249 (86%) 200 (91%) 99 70 31 88 (81%)targetsd 116 113 63 33 17 27suspects 133 87 36 37 14 61analytically not covered 40 20 6 3 11 21

aSubstances that are registered both as PPP and biocide are displayed in both columns so that the sum of PPPs and biocides is not equal to the totalnumber of pesticides. bPPPs: all organic synthetic pesticides that have been sold in at least one year between 2005 and 2011, excluding nonpolarsubstances (log Kow > 5). cBiocides: all organic synthetic biocides, excluding nonpolar substances (log Kow > 5) and quaternary ammonia cations;excluding fast degradable substances (half-life water < 1 d). dThe 25 substances that were confirmed in the suspect screening are included in thetargets.

Environmental Science & Technology Article

dx.doi.org/10.1021/es500371t | Environ. Sci. Technol. 2014, 48, 5423−54325425

detection on the high-resolution MS. Thereby, substances with

very broad physicochemical properties (e.g., log Kow −2 to 5)

can be detected to the low ng/L range in one run. With the

target method, 125 of 136 target pesticides could be measured

with LOQs below their specified AA-EQS values. Another 158

polar pesticides were covered by the suspect screening, from

which additionally 25 pesticides could be detected in at leastone sample.Substances such as carbamates (aldicarb), organophosphates

(chlorpyrifos, chlorpyrifos-methyl), or amino acid derivatives(glyphosate, glufosinate) were not covered by the analyticalmethod because they are not well-ionized in the massspectrometer due to the lack of a heteroatom or because they

Table 2. Detection Frequencies and Maximum Concentrations of the Most Frequently Detected Pesticides in the FiveInvestigated Rivers and Comparison with International Monitoring Studies#

#Only substances that were detected in more than three rivers are shown. aAlso registered as biocide. bAlso registered as veterinary pharmaceutical.cAA-EQS: annual average environmental quality standard, see SI 2.2 for references. dDicamba, 5-chloro-2-methyl-4-isothiazolin-3-on (CMI), andmetosulam are the only additional substances with EQS exceedances. eLetters indicate literature reference: A,15 B,34 C,35,36 D,2 E,37 F,10 G,16 H,38 I,39

K,40 L,41 M,42 N,43 O,44 P,45 Q,46 R,47 S,5 T,48 U.49 - Indicates not included in the 20 analyzed studies.

Environmental Science & Technology Article

dx.doi.org/10.1021/es500371t | Environ. Sci. Technol. 2014, 48, 5423−54325426

are hydrolyzed during preparation. Further, nonpolar sub-stances with log Kow > 5 could not be detected with thismethod. Due to strong sorption to organic matter, the freelydissolved water concentrations for these compounds areexpected to be very low. Nevertheless, pyrethroid insecticides,for which AA-EQS values in the sub-ng/L range are proposed(e.g., cypermethrin33), can still pose a risk. Analytical methodscovering such low detection limits are very challenging. Therisk from insecticides may therefore be underestimated in thisstudy.Screening ResultsParent Compounds. Concentra-

tions and Detection Frequency. From the 249 measurableparent compounds, 104 substances (42%) were detected in atleast one of the 45 samples (see Table 2 and SI Tables 3.1 and3.2 for detailed results of detected substances in all catch-ments). They consisted of 82 PPPs (not additionally registeredas biocides), 2 biocides (not additionally registered as PPPs),and 20 substances registered as both PPP and biocide. In total,54 herbicides, 31 fungicides, and 17 insecticides were detected.Three main reasons were identified why a substance was notdetected: (i) low sales data, (ii) fast degradation in water orsoil, (iii) high LOQ in the analytical method. As expected,herbicides had the highest detection frequencies (58%,compared to 43% and 34% for fungicides and insecticides,respectively) and highest concentrations (95th-percentileconcentration for all detected substances in the 45 samples:

100 ng/L, compared to 35 and 16 ng/L for fungicides andinsecticides, respectively). This corresponds to the differencesin sales data. Kreuger1 also found herbicides most frequentlyand generally in highest concentrations.Three herbicides had concentrations above 1000 ng/L, while

20 herbicides, 5 fungicides, and 3 biocides had concentrationsabove 100 ng/L (Table 2 and SI Tables 3.1 and 3.2). It has tobe noted that the measured concentrations are averageconcentrations over 2 weeks in medium-sized rivers and thatmaximum concentrations, especially in smaller streams, can bemuch higher. Forty percent of herbicide detections were below10 ng/L, while for fungicides 52% and for insecticides 67% ofthe detections were below 10 ng/L. This shows that it is veryimportant to have low LOQs for all substances, especially forinsecticides where in general the EQS values are also lower.

Catchment Differences and Seasonal Trend. In spite of thedifferent land uses in the five catchments, differences in thenumber of detected substances and concentration ranges wereless pronounced than expected (see Figure 1). Between 64(Salmsacher Aach river) and 76 (Surb river) pesticides weredetected at least once. The most substances were detected inMay and June (45 on average), while numbers were slightlylower in March/April (30) and July (40) (Figure 2A). Thisresult is easily explained as the main agricultural pesticideapplication period is between May and June. In contrast, thereis no clear seasonal trend of the sum of concentrations (Figure

Figure 2. Number of detected substances (A), sum of pesticide concentrations (μg/L) (B), and sum of risk quotients (C) in the five rivers duringthe sampling period in 2012 (parent compounds of all herbicides, insecticides, and fungicides considered; biocides that are not registered as PPPs arenot included). In addition, the discharge during the sampling period is shown (top).

Environmental Science & Technology Article

dx.doi.org/10.1021/es500371t | Environ. Sci. Technol. 2014, 48, 5423−54325427

2B) and often a few herbicides dominated the total pesticideconcentration sum. Interestingly, the substances that had thehighest concentrations did change over the season. In theFurtbach catchment for example, metamitron and metolachlorconcentrations accounted for 75% of the total herbicideconcentration in March. At the end of May, terbuthylazineand metolachlor summed up to more than 60% of the totalherbicide concentration.The sum of pesticide concentrations exceeded 1 μg/L in 78%

of samples and was on average 1.6−2.5 μg/L in each river(Figure 2B). Substantially lower concentrations were measuredin the Salmsacher Aach river, with concentration sums below 2μg/L at all time points (average: 0.8 μg/L). This is due to lowerherbicide concentrations probably caused by smaller relativedensities of arable crops.Frequently Detected Pesticides. In total, 33 pesticides (17

herbicides, 4 fungicides, 2 insecticides, 1 biocide, and 9substances with double registration) were detected in all fiverivers (Table 2). These substances are expected to beubiquitous in Swiss rivers with agricultural and urban influence.A large number of these frequently measured substances werealso included in many of the international studies considered,indicating their known surface water relevance. However, 8 ofthe 53 most frequently detected substances in Table 2 werenever or only once included in any of the investigatedinternational studies and are therefore potentially overlookedcompounds.Herbicides detected in the five rivers are mainly applied to

widespread arable crops such as cereals (e.g., mecoprop,isoproturon), corn (e.g., metolachlor, terbuthylazine), sugarbeet (e.g., chloridazone, metamitrone), or potatoes (e.g.,metribuzine), all of which are present in each of thecatchments. Metamitron, metolachlor, mecoprop, and chlor-idazone were also the substances with the highest maximumconcentrations (Table 2). This corresponds well with sales dataof the substances in Switzerland. The herbicide diuron was alsodetected in all rivers. It is, however, not applied to large fieldcrops but is used in orchards and vineyards and is additionallyregistered as biocide. Wittmer et al.50 frequently detecteddiuron as a result of the biocide application in catchments witha high urban land use. But in the Salmsacher Aach, diurondetections were most likely a result of application to appleorchards, since there is no WWTP located within thecatchment.Fungicides detected in all rivers (Table 2) also mainly

originated from arable crops such as cereals (e.g., cyprocona-zole, tebuconazole) or potatoes (e.g., dimethomorph, prop-amocarb). Three of the seven fungicides are also authorized asbiocides and can therefore be attributed to multiple land usetypes, complicating source identification. Fungicides with thehighest sales numbers, such as chlorothalonil, folpet, captan,and mancozeb, are all nonstable (half-life in water or soil < 1 d)and were accordingly never detected.Four of five insecticides (pirimicarb, diazinon, thiamethox-

ame, and dimethoate) (Table 2) that were detected in all fiverivers have the highest insecticide sales numbers, are permittedto be used in many different crops, and are also applied inprivate gardens. Neonicotinoids, whose relevance has beenheavily discussed in recent years due to their effects on bees51,52

and aquatic organisms,53,54 were also frequently detected.Especially the sprayed PPPs thiamethoxame (applied inorchards, vegetables, and private gardens) and thiacloprid(applied in potatoes, cereals, and oilseed rape) were frequently

found. Seed dressings (imidacloprid, clothianidin) were onlydetected sporadically in relatively low concentrations(< 20 ng/L), although sales data are in ranges comparable tothe sprayed neonicotinoids. The detections of imidacloprid inthe rivers Furtbach and Surb are most likely a result of sprayapplication on vegetables.To summarize, the PPPs that were detected in all rivers were

substances with high sales numbers and either applied in arablecrops which were present in all catchments or substances with avery broad agricultural application.Biocides have been investigated only sporadically in previous

years and the pervasiveness of biocides in surface waters has notyet been fully assessed. Interestingly, in this comprehensivescreening, besides the well-known diethyltoluamide (DEET),diuron, carbendazim, azole fungicides, and diazinon,17,50 noother biocides were frequently detected. The fact that thewastewater amount was variable in the five catchments (from0% in the Salmsacher Aach up to 80% in the Furtbach river atbaseline discharge) leads to the conclusion that there are noother important biocides in Swiss surface waters.Some substances were frequently detected in just one or two

catchments. Examples of such site-specific substances aremyclobutanil and methoxyfenozide (both registered fororchards and vineyards), which were only detected in theorchard-dense Salmsacher Aach catchment. Benthiavalicarb-isopropyl (applied on potatoes) and dimefuron (applied onoilseed rape) were only found in the Mentue catchment. Thisshows that a list of relevant compounds on a national scale is, inmany cases, not sufficient. Either very detailed usage data forthe investigation site needs to be available, or a completescreening for all potentially present pesticides has to be carriedout in order to get the full exposure picture.

Screening ResultsTransformation products (TPs).Of the 134 investigated TPs, 40 were detected at least once (31herbicide, 4 fungicide, 4 insecticide, and 1 biocide TPs) (see SITable 3.3). In particular, TPs of chloroacetanilide herbicides(e.g., metolachlor-ESA and metazachlor-ESA), chloridazone,atrazine, and azoxystrobin were detected in nearly every sample.These TPs were already frequently detected in other studies inSwitzerland4 and Germany.3 Between 15 and 25 TPs weredetected in each sample. Fifteen percent of all TP detectionswere above 100 ng/L. The concentration sum of the TPsexceeded 1 μg/L in 35% of samples and was dominated byherbicide TPs. Interestingly, the median concentration sum ofthe detected TPs in this study was 860 ng/L, which is 2.6 timeshigher than the median concentration found in German rivers.3

For six substances, only the TP was detected but not the parentcompound (bifenox, chlorothalonil, dichlobenil, dichlofluanid,fluazifop-butyl, prothioconazole). Either the parent compoundof these TPs has a short half-life in water or soil (e.g., bifenox,prothioconazole), or the parent compound could not bemeasured by the analytical method (chlorothalonil anddichlobenil). Information about the application of the parentcompounds can consequently be inferred from TP monitoring.It can be concluded that herbicide TPs are present inconcentrations comparable to the parent compounds, whilefungicide and insecticide TPs seem to have less relevance insurface waters.

Risk Assessment. Single Substances. For each of the 104detected parent substances, an AA-EQS value could be found inor derived from the literature (see SI Table 2.2). This meansthat in addition to the broad analytical coverage, a full riskassessment could be performed for all detected pesticides.

Environmental Science & Technology Article

dx.doi.org/10.1021/es500371t | Environ. Sci. Technol. 2014, 48, 5423−54325428

First, the risk of single substances was investigated, whichwas considered the least conservative risk assessment scenario.In total, 19 substances contributed at least once to anexceedance of an AA-EQS in the 45 composite samples(Table 2). These substances consisted of 13 herbicides, 4insecticides, 1 fungicide, and 1 biocide (not registered as aPPP). The most critical substances were diuron (13 exceed-ances), metazachlor (12), metolachlor (9), diazinon (8),terbuthylazine (6), and thiacloprid (6). Six substances (diuron,metolachlor, foramsulfuron, diazinon, terbuthylazine, andcarbofuran) had exceedances in more than three rivers; thus,they are probably relevant substances on a national scale.In 31 of 45 surface water samples (70%), at least one

exceedance was registered. In nearly half of the samples, morethan one exceedance was found (14 times 2−3 exceedances, 6times 5−7 exceedances). The most exceedances (6−7) were inthe Furtbach and Surb rivers in the beginning of June. Thus,when applying the least conservative scenario, already two-thirds of the water samples exceeded critical concentrations.Mixture Toxicity. The fact that 104 different pesticides were

measured in the five rivers shows that in the future more focusshould be on mixture toxicity approaches. When applying theworst-case scenario (summation of all RQs), 44 out of 45surface water samples exceeded the RQ of 1, up to an RQ of 25(Figure 2C, black line). It can be clearly seen that the Furtbachand Surb rivers show the highest risks (average exceedances bya factor of 11 and 9, respectively), while the other three riversalso had average exceedances of a factor of 4−6. Highest riskswere found at the end of May and beginning of June in all rivers(the main agricultural pesticide application season). However,samples in March already had mixture RQs of more than two innearly all catchments.Herbicides (Figure 2C, green line) were responsible for the

largest part of the mixture toxicity risk, accounting for 60−80%(median) of the total risk in the catchments. Secondcontributor to the risk were insecticides (6−20%, blue line).It has to be noted that the risk of the highly toxic pyrethroidswas not included. Although expected concentrations are verylow (sub-ng/L), these substances could still substantiallycontribute to the overall risk from insecticides. In contrast,the risk from fungicides and biocides was very low. In classicalrisk assessment, however, only three organism groups (i.e.,plants, vertebrates, and invertebrates) are considered. Fungi-cides obviously target fungi, which are normally not consideredin risk assessments. Thus, there might be a blind spot in thecurrent risk assessment and the risk of fungicides may beunderestimated.16

Scenario Analysis. A comprehensive screening asdescribed here has not been feasible until now, especially forroutine monitoring, due to the labor-intensive analysis and theneed for a high-resolution mass spectrometer. Therefore, weanalyzed how the exposure and risk assessment is different ifonly a subset of substances is considered as compared to acomplete screening. The three scenarios “Swiss Monitoring”,“International Studies”, and “WFD Pesticides” were considered(see Materials and Methods). In all scenarios, a much lowernumber of substances would be detected per sample: 15−20substances in the “Swiss Monitoring” and “InternationalStudies” scenarios, and only 4 substances in the “WFDPesticides” scenario, compared to roughly 40 substances inthe complete screening (see Figure 3A and SI Figure 4.1A forthe temporal dynamic). Thus, about half to two-thirds of thedetected substances were missed in the two scenarios

investigating 35−40 substances. When looking at theconcentration sums, the two scenarios “Swiss Monitoring”and “International Studies” covered 50−60% (Figure 3B and SIFigure 4.1 B). The results of the concentration sums matchedbetter with the complete screening than the results of thenumber of detections because the pesticides accounting for thehighest concentrations were included in these scenarios (e.g.,metamitron, metolachlor, terbuthylazin).With regard to single substance toxicity, in the two scenarios

“Swiss Monitoring” and “International Studies”, 30−35% of theAA-EQS exceedances of single substances were missed, mainlybecause the relevant substances flufenacet, foramsulfuron,nicosulfuron, prosulfocarb, thiacloprid, and carbofuran werenot included. In the “WFD Pesticides” scenario, more than 80%of the exceedances were missed. The only pesticide in the“WFD Pesticides” scenario that had multiple EQS exceedancesin the current screening study was diuron. In the WFD, it iswritten that river-basin specific substances should be selected inaddition to the priority list.33 This study confirms that for aproper risk assessment, this is essential.When considering mixture toxicity, it can be seen that on

average 55−65% of the total risk (65−70% of the herbicide risk,10−50% of the fungicide risk, and 35−55% of the insecticide

Figure 3. Number of detected pesticides (A), sum of pesticideconcentrations (B), sum of risk quotients (RQs) for pesticides (C), forall herbicides (D), for all fungicides (E), and for all insecticides (F) inthe 45 samples in the 5 catchments. Left boxplot (gray): completescreening; right boxplots (gray shaded area): three scenarios for whichonly a subset of substances were selected. Numbers in parenthesescorrespond to the number of investigated substances in each scenario.

Environmental Science & Technology Article

dx.doi.org/10.1021/es500371t | Environ. Sci. Technol. 2014, 48, 5423−54325429

risk) can be detected by the scenarios “Swiss Monitoring” and“International Studies” (Figure 3C−F and SIC Figure 4.1).There are, however, large differences between the catchments.In the Furtbach river, nearly all of the risk was captured by thescenarios (80−85%), while in the Limpach river it was only40−50% (SI Figure 4.2).In 4−8 of the 45 river samples, it would have been concluded

that there was no risk from herbicides when only investigatingthe substances from the scenarios “Swiss Monitoring” and“International Studies”, although a risk was detected by thecomplete screening. For insecticides, 14−18 of the samples(30−40%) would have been interpreted incorrectly. This showsthat especially insecticides are still underrepresented inanalytical methods and consequently in routine pesticidemonitoring.Implications for Routine Monitoring. The results of this

study show that the most frequently measured pesticides by thescientific community and by Swiss authorities may allow areasonable assessment of surface water quality in medium-sizedSwiss rivers. The risk to surface waters from the mixture toxicityassessment was in most cases underestimated by a factor of 2,and in extreme cases the underestimation was up to a factor of10. The addition of some substances (e.g., flufenacet,foramsulfuron, thiacloprid, carbofuran) to the current substanceselection would improve the surface water assessment.Nevertheless, when not all substances are monitored, there isalways the probability that an important substance has beenmissed. This is especially true for monitoring of rivers in smallercatchments where a site-specific substance (that is not relevanton the broader scale) can be the dominant pesticide. Without acomprehensive screening, the only way to overcome thisproblem is detailed knowledge of the applied substances in thecatchment, which is often hard to achieve in practice.The risk assessment based on the complete screening

showed that herbicides and insecticides dominate the risk inSwiss surface waters and exceedances of critical concentrationswere found over the whole investigation period in allcatchments. The fact that over 100 pesticides were detectedin the five rivers demonstrates the importance of a refinedmixture toxicity approach. It is also critical that relevantinsecticides (e.g., neonicotinoids) are included in routinemonitoring programs carried out by authorities. Furthermore,it is essential to use analytical methods with LOQs below 10ng/L in order to agree with EQS values. Finally, practicallyaccessible methods are needed with which very toxicpyrethroids can be detected down to the sub-ng/L range.Most of the relevant pesticides defined here can be measured

after proper sample extraction on low-resolution LC-MS/MS(e.g., triple quadrupoles). Nevertheless, in the near future,sensitive high-resolution mass spectrometers and moreautomated software tools will most likely become accessiblefor routine analysis, too. Thereby, screening of the wholepesticide spectrum may become possible and perhaps evenmore cost-effective than defining the most relevant substancesbeforehand and analyzing them specifically. As the relevantpesticides will change over time and can be regionally different,a comprehensive screening is in either case the optimal way todo a proper pesticide exposure and risk assessment in surfacewaters.

■ ASSOCIATED CONTENT*S Supporting Information(1) Additional study site information, (2) additional substanceinformation, (3) detailed information on measured concen-trations, (4) risk assessment and scenario analysis. This materialis available free of charge via the Internet at http://pubs.acs.org.

■ AUTHOR INFORMATIONCorresponding Authors*Tel: +41 58 765 50 98; e-mail: christoph.moschet@eawag.ch.*Tel: +41 58 765 54 93; e-mail: juliane.hollender@eawag.ch.

Author ContributionsThe manuscript was written through contributions of allauthors. All authors have given approval to the final version ofthe manuscript.

NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThis study was funded by the Swiss Federal Office for theEnvironment (FOEN). The sampling by the cantonal author-ities of the Canton Thurgau, Aargau, Solothurn, Waadt, andZurich is gratefully acknowledged. We thank Philipp Longree,Sebastian Huntscha, and Tobias Doppler (all Eawag) for thehelp in the laboratory and in planning the field study, and wethank Devon Wemyss and Jennifer Schollee (both Eawag) forimproving the manuscript.

■ REFERENCES(1) Kreuger, J. Pesticides in stream water within an agriculturalcatchment in southern Sweden, 1990−1996. Sci. Total Environ. 1998,216 (3), 227−251.(2) Herrero-Hernandez, E.; Andrades, M. S.; Alvarez-Martín, A.;Pose-Juan, E.; Rodríguez-Cruz, M. S.; Sanchez-Martín, M. J.Occurrence of pesticides and some of their degradation products inwaters in a Spanish wine region. J. Hydrol. 2013, 486, 234−245.(3) Reemtsma, T.; Alder, L.; Banasiak, U. Emerging pesticidemetabolites in groundwater and surface water as determined by theapplication of a multimethod for 150 pesticide metabolites. Water Res.2013, 47 (15), 5535−5545.(4) Kern, S.; Singer, H.; Hollender, J.; Schwarzenbach, R. P.; Fenner,K. Assessing exposure to transformation products of soil-appliedorganic contaminants in surface water: Comparison of modelpredictions and field data. Environ. Sci. Technol. 2011, 45 (7), 2833−2841.(5) Schafer, R. B.; Caquet, T.; Siimes, K.; Mueller, R.; Lagadic, L.;Liess, M. Effects of pesticides on community structure and ecosystemfunctions in agricultural streams of three biogeographical regions inEurope. Sci. Total Environ. 2007, 382 (2−3), 272−285.(6) Schulz, R. Field studies on exposure, effects, and risk mitigationof aquatic nonpoint-source insecticide pollution: A review. J. Environ.Qual. 2004, 33 (2), 419−448.(7) Brown, C. D.; van Beinum, W. Pesticide transport via sub-surfacedrains in Europe. Environ. Pollut. 2009, 157 (12), 3314−3324.(8) Wittmer, I. K.; Bader, H. P.; Scheidegger, R.; Singer, H.; Luck, A.;Hanke, I.; Carlsson, C.; Stamm, C. Significance of urban andagricultural land use for biocide and pesticide dynamics in surfacewaters. Water Res. 2010, 44 (9), 2850−2862.(9) Munz, N.; Wittmer, I. Pestizidmessungen in Fliessgewassern.Schweizweite Auswertung. Aqua & Gas 2012, 11, 10.(10) Huntscha, S.; Singer, H. P.; McArdell, C. S.; Frank, C. E.;Hollender, J. Multiresidue analysis of 88 polar organic micropollutantsin ground, surface and wastewater using online mixed-bed multilayersolid-phase extraction coupled to high performance liquid chromatog-

Environmental Science & Technology Article

dx.doi.org/10.1021/es500371t | Environ. Sci. Technol. 2014, 48, 5423−54325430

raphy−tandem mass spectrometry. J. Chromatogr. A 2012, 1268, 74−83.(11) BLW. Agrarbericht 2010; Bern, 2010.(12) University of Hertfordshire. The Pesticide Properties DataBase(PPDB) developed by the Agriculture & Environment Research Unit(AERU); University of Hertfordshire: Hertfordshire, U.K.; 2013.(13) Starner, K.; Goh, K. S. Detections of the neonicotinoidinsecticide imidacloprid in surface waters of three agricultural regionsof California, USA, 2010−2011. Bull. Environ. Contam. Toxicol. 2012,88 (3), 316−321.(14) Yamamoto, A.; Terao, T.; Hisatomi, H.; Kawasaki, H.; Arakawa,R. Evaluation of river pollution of neonicotinoids in Osaka City(Japan) by LC/MS with dopant-assisted photoionisation. J. Environ.Monit. 2012, 14 (8), 2189−2194.(15) Battaglin, W. A.; Sandstrom, M. W.; Kuivila, K. M.; Kolpin, D.W.; Meyer, M. T. Occurrence of azoxystrobin, propiconazole, andselected other fungicides in US streams, 2005−2006. Water, Air, SoilPollut. 2011, 218 (1−4), 307−322.(16) Reilly, T. J.; Smalling, K. L.; Orlando, J. L.; Kuivila, K. M.Occurrence of boscalid and other selected fungicides in surface waterand groundwater in three targeted use areas in the United States.Chemosphere 2012, 89 (3), 228−234.(17) Kahle, M.; Buerge, I. J.; Hauser, A.; Muller, M. D.; Poiger, T.Azole fungicides: Occurrence and fate in wastewater and surfacewaters. Environ. Sci. Technol. 2008, 42 (19), 7193−7200.(18) Kern, S.; Fenner, K.; Singer, H. P.; Schwarzenbach, R. P.;Hollender, J. Identification of transformation products of organiccontaminants in natural waters by computer-aided prediction andhigh-resolution mass spectrometry. Environ. Sci. Technol. 2009, 43(18), 7039−7046.(19) Reemtsma, T.; Alder, L.; Banasiak, U. A multimethod for thedetermination of 150 pesticide metabolites in surface water andgroundwater using direct injection liquid chromatography−massspectrometry. J. Chromatogr. A 2013, 1271 (1), 95−104.(20) Krauss, M.; Singer, H.; Hollender, J. LC−high resolution MS inenvironmental analysis: From target screening to the identification ofunknowns. Anal. Bioanal. Chem. 2010, 397 (3), 943−951.(21) Moschet, C.; Piazzoli, A.; Singer, H.; Hollender, J. Alleviatingthe reference standard dilemma using a systematic exact mass suspectscreening approach with liquid chromatography-high resolution massspectrometry. Anal. Chem. 2013, 85 (21), 10312−10320.(22) Price, P.; Dhein, E.; Hamer, M.; Han, X.; Heneweer, M.;Junghans, M.; Kunz, P.; Magyar, C.; Penning, H.; Rodriguez, C. Adecision tree for assessing effects from exposures to multiplesubstances. Environ. Sci. Eur. 2012, 24 (1), 26.(23) Junghans, M.; Backhaus, T.; Faust, M.; Scholze, M.; Grimme, L.H. Application and validation of approaches for the predictive hazardassessment of realistic pesticide mixtures. Aquat. Toxicol. 2006, 76 (2),93−110.(24) Backhaus, T.; Faust, M. Predictive environmental riskassessment of chemical mixtures: A conceptual framework. Environ.Sci. Technol. 2012, 46 (5), 2564−2573.(25) Belden, J. B.; Gilliom, R. J.; Lydy, M. J. How well can we predictthe toxicity of pesticide mixtures to aquatic life? Integr. Environ. Assess.Manage. 2007, 3 (3), 364−372.(26) Strahler, A. N. Hypsometric (area-altitude) analysis of erosionaltopography. Bull. Geol. Soc. Am. 1952, 63, 1117−1132.(27) BAFU. Gewa sserabschnittsbasierte Einzugsgebietsgliederung derSchweiz; GAB-EZGG-CH, 2013.(28) BLW. Pflanzenschutzmittelverzeichnis. Federal Office for theAgriculture. 2012. http://www.blw.admin.ch/psm/produkte/index.html?lang=de. Accessed 13/12/12.(29) BAG. Biocidal Active Substances. Swiss Federal Office forHealth. 2012.http://www.bag.admin.ch/anmeldestelle/13604/13869/13883/index.html?lang=en. Accessed 13/12/12.(30) European Chemicals Bureau Technical Guidance Document on RiskAssessment. In Support of Commission Directive 93/67/EEC on RiskAssessment for new notified substances, Commission Regulation (EC) No1488/94 on Risk Assessment for existing substances, directive 98/8/EC of

the European Parliament and of the Council concerning the placing ofbiocidal products on the market. Part II . 2003. http://ihcp.jrc.ec.europa.eu/our_activities/public-health/risk_assessment_of_Biocides/doc/tgd/tgdpart2_2ed.pdf. Accessed 13/12/12.(31) Swiss Center for Applied Ecotoxicology Eawag/EPFL. Proposalsfor Acute and Chronic Quality Criteria. 2013. http://www.oekotoxzentrum.ch/expertenservice/qualitaetskriterien/vorschlaege/index. Accessed 13/12/12.(32) EFSA.. Guidance on tiered risk assessment for plant protectionproducts for aquatic organisms in edge-of-field surface waters.Guidance of the PPR Panel Scientific Opinion of the EFSA Panelon Plant Protection Products and their Residues (PPR). EFSA J. 2013,11 (7), 3290 http://www.efsa.europa.eu/de/efsajournal/pub/3290.htm.(33) EU. Directive 2013/39/EU of the European Parliament and of theCouncil of 12 August 2013 amending Directives 2000/60/EC and 2008/105/EC as regards priority substances in the field of water policy Text withEEA relevance. http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2013:226:0001:0017:EN:PDF. Accessed 14/03/11.(34) Dujakovic, N.; Grujic, S.; Radisic, M.; Vasiljevic, T.; Lausevic, M.Determination of pesticides in surface and ground waters by liquidchromatography-electrospray-tandem mass spectrometry. Anal. Chim.Acta 2010, 678 (1), 63−72.(35) Gomez, M. J.; Gomez-Ramos, M. M.; Malato, O.; Mezcua, M.;Fernandez-Alba, A. R. Rapid automated screening, identification andquantification of organic micro-contaminants and their main trans-formation products in wastewater and river waters using liquidchromatography−quadrupole-time-of-flight mass spectrometry with anaccurate-mass database. J. Chromatogr. A 2010, 1217 (45), 7038−7054.(36) Mezcua, M.; Malato, O.; García-Reyes, J. F.; Molina-Díaz, A.;Fernandez-Alba, A. R. Accurate-mass databases for comprehensivescreening of pesticide residues in food by fast liquid chromatographytime-of-flight mass spectrometry. Anal. Chem. 2009, 81 (3), 913−929.(37) Jansson, C.; Kreuger, J. Multiresidue analysis of 95 pesticides atlow nanogram/liter levels in surface waters using online preconcentra-tion and high performance liquid chromatography/tandem massspectrometry. J. AOAC Int. 2010, 93 (6), 1732−1747.(38) Schafer, R. B.; Von Der Ohe, P. C.; Kuhne, R.; Schuurmann, G.;Liess, M. Occurrence and toxicity of 331 organic pollutants in largerivers of north Germany over a decade (1994 to 2004). Environ. Sci.Technol. 2011, 45 (14), 6167−6174.(39) Tanabe, A.; Mitobe, H.; Kawata, K.; Yasuhara, A.; Shibamoto, T.Seasonal and spatial studies on pesticide residues in surface waters ofthe Shinano River in Japan. J. Agric. Food Chem. 2001, 49 (8), 3847−3852.(40) Belmonte Vega, A.; Garrido Frenich, A.; Martínez Vidal, J. L.Monitoring of pesticides in agricultural water and soil samples fromAndalusia by liquid chromatography coupled to mass spectrometry.Anal. Chim. Acta 2005, 538 (1−2), 117−127.(41) Vryzas, Z.; Alexoudis, C.; Vassiliou, G.; Galanis, K.;Papadopoulou-Mourkidou, E. Determination and aquatic risk assess-ment of pesticide residues in riparian drainage canals in northeasternGreece. Ecotoxicol. Environ. Saf. 2011, 74 (2), 174−181.(42) Hladik, M. L.; Smalling, K. L.; Kuivila, K. M. A multi-residuemethod for the analysis of pesticides and pesticide degradates in waterusing HLB solid-phase extraction and gas chromatography-ion trapmass spectrometry. Bull. Environ. Contam. Toxicol. 2008, 80 (2), 139−144.(43) Kampioti, A.; Borba da Cunha, A.; Lopez de Alda, M.; Barcelo,D. Fully automated multianalyte determination of different classes ofpesticides, at picogram per litre levels in water, by on-line solid-phaseextraction−liquid chromatography−electrospray−tandem mass spec-trometry. Anal. Bioanal. Chem. 2005, 382 (8), 1815−1825.(44) Rodrigues, A. M.; Ferreira, V.; Cardoso, V. V.; Ferreira, E.;Benoliel, M. J. Determination of several pesticides in water by solid-phase extraction, liquid chromatography and electrospray tandem massspectrometry. J. Chromatogr. A 2007, 1150 (1−2), 267−278.

Environmental Science & Technology Article

dx.doi.org/10.1021/es500371t | Environ. Sci. Technol. 2014, 48, 5423−54325431

(45) Wode, F.; Reilich, C.; van Baar, P.; Dunnbier, U.; Jekel, M.;Reemtsma, T. Multiresidue analytical method for the simultaneousdetermination of 72 micropollutants in aqueous samples with ultrahigh performance liquid chromatography−high resolution massspectrometry. J. Chromatogr. A 2012, 1270, 118−126.(46) Tankiewicz, M.; Morrison, C.; Biziuk, M. Multi-residue methodfor the determination of 16 recently used pesticides from variouschemical groups in aqueous samples by using DI-SPME coupled withGC−MS. Talanta 2013, 107, 1−10.(47) Finizio, A.; Azimonti, G.; Villa, S. Occurrence of pesticides insurface water bodies: A critical analysis of the Italian national pesticidesurvey programs. J. Environ. Monit. 2011, 13 (1), 49−57.(48) Lissalde, S.; Mazzella, N.; Fauvelle, V.; Delmas, F.; Mazellier, P.;Legube, B. Liquid chromatography coupled with tandem massspectrometry method for thirty-three pesticides in natural water andcomparison of performance between classical solid phase extractionand passive sampling approaches. J. Chromatogr. A 2011, 1218 (11),1492−1502.(49) Phillips, P. J.; Bode, R. W. Pesticides in surface water runoff insouth-eastern New York State, USA: Seasonal and stormflow effectson concentrations. Pest Manage. Sci. 2004, 60 (6), 531−543.(50) Wittmer, I. K.; Scheidegger, R.; Bader, H.-P.; Singer, H.; Stamm,C. Loss rates of urban biocides can exceed those of agriculturalpesticides. Sci. Total Environ. 2011, 409 (5), 920−932.(51) Henry, M.; Beguin, M.; Requier, F.; Rollin, O.; Odoux, J. F.;Aupinel, P.; Aptel, J.; Tchamitchian, S.; Decourtye, A. A commonpesticide decreases foraging success and survival in honey bees. Science2012, 336 (6079), 348−350.(52) Whitehorn, P. R.; O’Connor, S.; Wackers, F. L.; Goulson, D.Neonicotinoid pesticide reduces bumble bee colony growth and queenproduction. Science 2012, 336 (6079), 351−352.(53) Van Dijk, T. C.; Van Staalduinen, M. A.; Van der Sluijs, J. P.Macro-invertebrate decline in surface water polluted with imidacloprid.PLoS One 2013, 8 (5), e62374.(54) Goulson, D. Review: An overview of the environmental risksposed by neonicotinoid insecticides. J. Appl. Ecol. 2013, 50 (4), 977−987.

Environmental Science & Technology Article

dx.doi.org/10.1021/es500371t | Environ. Sci. Technol. 2014, 48, 5423−54325432