Occurrence of metolachlor and trifluralin losses in the Save river agricultural catchment during floods

To link to this article:

DOI: 10.1016/j.jhazmat.2011.09.012

http://dx.doi.org/10.1016/j.jhazmat.2011.09.012

This is an author-deposited version published in:

http://oatao.univ-toulouse.fr/

Eprints ID: 5435

To cite this version:

Boithias, Laurie and Sauvage, Sabine and Taghavi, Lobat and Merlina, Georges

and Probst, Jean-Luc and Sánchez Pérez, José Miguel Occurrence of

metolachlor and trifluralin losses in the Save river agricultural catchment

during floods. (2011) Journal of Hazardous Materials, vol. 196 . pp. 210-219.

ISSN 0304-3894

O

pen

A

rchive

T

oulouse

A

rchive

O

uverte (

OATAO

)

OATAO is an open access repository that collects the work of Toulouse researchers and

makes it freely available over the web where possible.

Any correspondence concerning this service should be sent to the repository

Occurrence

of

metolachlor

and

trifluralin

losses

in

the

Save

river

agricultural

catchment

during

floods

Laurie

Boithias

_,

_Sabine

_Sauvage

_,

_Lobat

_Taghavi

_,

_Georges

_Merlina

_,

_Jean-Luc

_Probst

_,

José

Miguel

Sánchez

Pérez

a_{UniversityofToulouse;INP,UPS;Laboratoired’EcologieFonctionnelleetEnvironnement(EcoLab);ENSAT,Avenuedel’Agrobiopole,31326CastanetTolosanCedex,France} b_{CNRS,EcoLab,31326CastanetTolosanCedex,France}

a

r

t

i

c

l

e

i

n

f

o

a

b

s

t

r

a

c

t

Risingpesticidelevelsinstreamsdrainingintensivelymanagedagriculturallandhaveadetrimentaleffect onaquaticecosystemsandrenderwaterunfitforhumanconsumption.TheSoilandWaterAssessment Tool(SWAT)wasappliedtosimulatedailypesticidetransferattheoutletfromanagriculturally inten-sivecatchmentof1110km2_{(Saveriver,south-westernFrance).SWATreliablysimulatedbothdissolved}

andsorbedmetolachlorandtrifluralinloadsandconcentrationsatthecatchmentoutletfrom1998to 2009.Onaverage,17kgofmetolachlorand1kgoftrifluralinwereexportedatoutleteachyear,with annualrainfallvariationsconsidered.Surfacerunoffwasidentifiedasthepreferredpathwayfor pesti-cidetransfer,relatedtothegoodcorrelationbetweensuspendedsedimentexportationandpesticide,in bothsolubleandsorbedphases.Pesticideexportationratesatcatchmentoutletwerelessthan0.1%of theappliedamount.Atoutlet,SWAThindcastedthat(i)61%ofmetolachlorand52%oftrifluralinwere exportedduringhighflowsand(ii)metolachlorandtrifluralinconcentrationsexceededEuropean drink-ingwaterstandardsof0.1␮gL−1forindividualpesticidesduring149(3.6%)and17(0.4%)daysofthe 1998–2009periodrespectively.SWATwasshowntobeapromisingtoolforassessinglargecatchment rivernetworkpesticidecontaminationintheeventoffloodsbutfurtherusefuldevelopmentsofpesticide transfersandpartitioncoefficientprocesseswouldneedtobeinvestigated.

1. Introduction

Risingpesticidelevels in streamwatersdraining intensively managedagriculturalland have becomea widespread problem throughout Europe in recent decades. Intensive agriculture is knowntohave adetrimentaleffect onsoils,surfacewater and groundwaterquality,leadingtoacuteproblemssuchassoilerosion andwatercontamination[e.g.1–4].Excessiveloadingofpesticides, transferredintotheenvironmentthroughvariouspathways(e.g. surfacerunoff,subsurfaceandgroundwaterflows)eitherin solu-tionorsorbedontoparticles,maybeharmfultoterrestrialand aquaticecosystems[5–8],renderingstreamwaterand groundwa-terunfitforhumanconsumption.

∗ Correspondingauthorsat:UniversityofToulouse,INPT,UPS,Laboratoire Ecolo-gieFonctionnelleetEnvironnement(EcoLab),ENSAT,Avenuedel’Agrobiopole, 31326CastanetTolosanCedex,France.Tel.:+33534323920;

fax:+33534323901.

E-mailaddresses:l.boithias@gmail.com(L.Boithias),sanchez@cict.fr

(J.M.SánchezPérez).

1 _{Presentaddress:DepartmentofEnvironmentandEnergy,ScienceandResearch} Branch,IslamicAzadUniversity,Tehran,Iran.

InEurope,pesticidesareconsideredhazardoussubstancesin accordancewithcurrentdirectivesregardingwater[9,10]. Drink-ing water qualitystandard should not exceed0.1␮gL−1 foran individualpesticideconcentrationand0.5␮gL−1 forallpesticide concentration[11].Riverbasinswereadoptedasterritorial man-agementunitsandthescientificcommunitywasaskedtoprovide reliablemodellingtoolstoevaluatepesticidesourcecontributionto waterpollutionandtoquantifypesticideriverloads.Tomodel pes-ticidefateatcatchmentscale,spatiallyvariablelandmanagement andlandscapecharacteristics,temporallyvariableclimatologyand hydrologyaswellasdissipationprocessesintheriverneedtobe takenintoaccount.Therefore,thecombinationofwatershed mod-elsandriverwaterqualitymodelsisneededtocalculatepesticide fluxestotheriverandtransformationprocessesintheriverchannel

[12].

Variousmodelsdescribethepesticidefate,allowinga better understandingof the processesinvolved. Amongthe veryfirst, theChemicalMigrationandRiskAssessment(CMRA)methodology

[13]includedtheAgriculturalRunoffManagement(ARM)[14]and Chemicals,RunoffandErosionfromAgriculturalManagement Sys-tems(CREAMS)[15]models.Itsimulatesthetransportandfateof bothdissolvedandsediment-sorbedcontaminantsandby predict-ingacuteandchronicimpacts,providesriskassessmentonaquatic

biota.Manyone-dimension,riverandcatchmentscalemodels sim-ulatingpesticidefatehavebeenthendeveloped[e.g.16–19].The SoilandWaterAssessmentTool(SWAT,[20])isasemi-distributed modelthatprovideslong-termcontinuouspredictions,including hydrology,plantgrowth,nutrientsandsuspendedsedimentsfrom thefieldtothecatchmentoutletatdailytime-step.Twomain pro-cessesdescribethepesticidetransferinboththesolubleandthe sorbedphases:thepesticideloadgeneratedinthehydrologicalunit andthefateoftheloadintheriver.

Fewworkshavebeenpublishedsofaronpesticidefate mod-ellingusingtheSWATmodel.Moleculesofawiderangeofsolubility weresimulated(e.g.atrazine,metolachlor,trifluralin,diazinonand chlorpyrifos)incatchmentsrangingfrom30to15,000km2_[21–24]. Toourknowledge,noworkhasbeenpublishedonpesticide mod-ellinginboththedissolvedandsorbedphasesatflood-eventscale, i.e.duringafewdaysofhighflow.

Thisstudyhadfourobjectives:(i)toassesstheperformanceof theSWATmodelintheSavecatchment(1110km2_{)intheGascogne} region,anagriculturallyintensiveareaofsouth-westernFrance, inpredictingdailypesticideriverloadsandconcentrationsatthe catchmentoutlet;(ii)totest thesensitivity ofSWATlong-term response,intermsofpesticideexportations,tointerannual hydro-logicalconstraintbyusingan11-yearconstantpesticidesupply; (iii)tohindcastearlierpesticidedatainordertomakethemodel reliableforpredictingrivernetworkcontamination(e.g.exceeding drinkingwaterstandards)dependingontheclimaticcontextand possiblefloodevents;(iv)toidentifyfactorscontrolling exporta-tionsandpreferredpathways.

2. Materialandmethods 2.1. Studyarea

TheRiver Save drainsan areaof 1110km2 _{which is mostly} farmedwithintensiveagriculture.ItislocatedintheCoteauxde Gascogneregion(south-westernFrance)nearToulouse(Fig.1).The RiverSave hasitssourceinthePyreneespiedmont.It joinsthe RiverGaronneaftera140kmcourseata0.4%averageslope. Alti-tudesrangefrom663minthepiedmontto92mattheGaronne confluence.TheLarragaugingstationelevationis114m(Fig.1).

The climate is oceanic. The Save river hydrological regime is mainly pluvial with a maximum discharge in May and low flowsduringthesummer(July–September).Annualprecipitationis 600–900mmandannualevaporationis500–600mm(1998–2008). Thehydrologyiscomplexandsubjecttolargeclimaticvariations: annualaveragerainfallis721mmwitha99mmstandarddeviation. Thecatchmentliesondetritalsediments.Calcicsoilsrepresentover 90%ofthewholecatchmentwithaclaycontentrangingfrom40% to50%.Non-calcicsiltysoilsrepresentlessthan10%ofthesoilin thisarea(50–60%silt)[25].

Becauseofitshighclaycontent,thecatchmentsubstratumis relativelyimpermeable.Riverdischargeisconsequentlysupplied mainlybysurfaceandsubsurfacerunoffandgroundwaterislimited toalluvialandcolluvialphreaticaquifers.Themeaninterannual discharge(1965–2006)fortheRiverSaveis6.1m3_s−1_.Theannual dischargeof‘dry’yearsis approximately4.1m3_s−1 _{whereas for} ‘wet’years itis about8.1m3_s−1_{.Lowwater dischargeis about} 1.3m3_s−1_{(datafromtheCompagnied’AménagementdesCoteaux} deGascogne(CACG)attheLarragaugingstation).Duringlowflows, riverflowissustainedupstreambytheNestecanal(about1m3_s−1_). 90%ofthecatchmentareaisusedforagriculture.Theupstream partofthecatchmentisahillyagriculturalareamainlycovered withpasture–a5-yearrotationincludingoneyearofcornand4 yearsofgrazedfescue– andsometimesforest.Thelowerpartis devotedtointensiveagriculturewithmainlya2-yearcroprotation

Table1

Spatialandtemporal3-yearsurveyaveragedmanagementpracticesforsunflower grownontheSavecatchmentincludingmetolachlorandtrifluralinspreading.

Typeofoperation Dateof

operation

Quantity (kgha−1₎

Pesticidesspreading:trifluralin 05April 0.874

Fertilizer:15-15-15 05April 193.3

Sowing 10April –

Pesticidesspreading:metolachlor 15April 1.12

Fertilizer:15-15-15 16May 193.3

Harvest 01October –

ofsunflowerand winterwheat.Fertilizersaregenerallyapplied fromlatewintertospring.Averagenitrogenoussupply through-outthecatchmentisapproximately72kgNha−1,i.e.320kgha−1 nitrate-equivalents.About150mmofwaterissuppliedby irriga-tionofcorn.

Variouspesticidesareappliedinthecatchmentthroughoutthe yeardependingonthecrops.Ourstudyfocusesonthemostapplied pesticides: each year, 23tons of metolachlor, a highly soluble chemical(Sw=488mgL−1,logKow=2.9),and18tonsoftrifluralin,

a poorlysoluble chemical (Sw=0.221mgL−1, logKow=4.83),are

appliedonthecatchment.Bothpesticidesareherbicides.Theyare appliedeachyearonsunflowerinearlyApril(Table1).Onaverage, sunflowerfieldscover18.4%ofthecatchment(20,600ha). 2.2. Observeddischargedata

TheRiverSavehasbeenmonitoredfordischargesince1965.At theLarrahydrometricstation,hourlydischarges(Q)wereobtained fromCACG.Thehourlydischargewasplottedbytheratingcurve H(Q)inwhichthewaterlevel(H)wasmeasuredcontinuouslyand thenaveragedforeachday.

2.3. Observedwaterqualitydata

2.3.1. Nitrateandsuspendedsedimentmonitoring

Nitrate loads and suspended sediment concentrations were monitored continuously from January 2007 to March 2009 at theLarra gauging station, both manually and automatically, as describedpreviouslyinOeurngetal.[26–28]:anautomaticwater sampler,connectedtoaprobe,wasprogrammedtoactivate pump-ingwateronthebasisofwaterlevelvariationsrangingfrom10cm (duringlowflows)to30cm(duringhighflows)fortherisingand fallingstages.Grabsamplingwasalsoundertakenneartheprobe positionatweeklyintervals.

2.3.2. Pesticidemonitoring

PesticidesweremonitoredfromMarch2008toMarch2009at theLarrastationwithweeklygrabsamplingduringlowflowand dailygrabsamplingduringfloodevents.Laboratoryanalyseswere performedasdescribedinTaghavietal.[29,30].Additionaldata fromAgencedel’EauAdour-Garonne(AEAG)wereusedfor long-termtotal pesticideconcentrationcomparison (Source:Système d’Informationsurl’EauduBassinAdour-Garonne,dataexportedin 2009).

2.3.3. Loadcalculation

Basedonthehighfrequencyofdatacollection,alinear interpo-lationmethodwasappliedbetweentwoneighbouringsampling pointstoconstructthecontinuousnitrate,suspended sediment andpesticideconcentrationseriesandthuscalculatecontinuous dailyloadsthroughtheproductofconcentrationandwatervolume. Yearlyloadswerecalculatedbytotallingdailyloads.

Fig.1.LocalisationofSavecatchment,Larragaugingstationandmeteorologicalstations.

2.4. Modellingapproach 2.4.1. TheSWATmodel

SWATis a physically basedagro-hydrological model [20]. It operatesatadailytime-stepandwasdesignedtopredicttheimpact of management practices onwater quality in ungauged catch-ments.Itallowstheadditionofflowsbyincludingmeasureddata frompointsources.SWATdiscretisescatchmentsintosub-basins. Sub-basinsarethensubdividedintoHydrologicalResponseUnits (HRUs).HRUsareareasofhomogenouslanduse,soiltypeandslope. HRUsoutputsareinputsfortheconnectedstreamnetwork.One sub-basinisdrainedbyonereach.AuthorsrefertoNeitschetal.

[31]fordetaileddescriptionofthemodel.

2.4.2. ThepesticidecomponentinSWAT

PesticideprocessesinSWATaredividedintothreecomponents: (i)pesticideprocessesinlandareas,(ii)transportofpesticidesfrom landareastothestreamnetwork,and(iii)instreampesticide pro-cesses.

SWATuses algorithms from GLEAMS (Groundwater Loading EffectsonAgriculturalManagementSystems)[16]tomodel pesti-cidemovementandfateinlandareas.Thepartitioningofapesticide between the dissolved and sorbed phases is defined by a soil adsorptioncoefficient.Algorithmsgoverningmovementof solu-bleandsorbedformsofpesticidefromlandareastothestream networkweretakenfromtheEPIC (Erosion-ProductivityImpact Calculator)model [32].The SWATmodelincorporates a simple mass-balancemethod[33]tomodelthetransformationand trans-portof pesticidesin streams. Onlyone pesticidecanberouted throughthestreamnetworkina givensimulation. Thefraction ofpesticideineachphaseisafunctionofthepesticide’spartition coefficientandthereachsegment’ssuspendedsolidconcentration. Degradationisbasedonhalf-life.AuthorsrefertoNeitschetal.[31]

forfurtherdetails.

2.5. SWATdatainputs

Spatialiseddatausedinthisstudywere:

- DigitalElevationModel witha resolutionof25m×25mfrom InstitutGéographiqueNational(IGN)France(BDTOPOR).

-Soildataonthescaleof1:80,000fromCACGanddigitisedby CemagrefdeBordeaux[34]andsoilpropertiesfortheSWATsoil database[35].

-Landusedata[34]fromLandsat2005withassociated manage-mentpractices:spatialandtemporalaverageofplanting/seedling dates,amounts,typeanddateoffertilisation,pesticide applica-tionandirrigation,grazing,tillageandharvestoperationsdates from a 3 year survey (2003–2005) with catchment farmers, appliedforeachyearofsimulation.

-Meteorologicaldatafrom5stations(Fig.1)withdaily precipita-tionfromMétéo-France.Missingdataweregeneratedbylinear regression equationfromdata fromthenearest stations with completemeasurements.Twostationsintheupstreamsection hadacompletesetofmeasurementsofdailyminimumand max-imumairtemperature,windspeed,solarradiationandrelative humiditythatwereusedtosimulatethereference evapotranspi-rationinthemodelbyPenman–Monteith[36,37]method. -Pointsourcedata:theSaverivernetworkisconnectedupstream

totheNestecanal.DailydischargewasgivenbyCACG.Sinceitis waterfromamountainousagriculturalextensivearea, concen-trationsofnitrateandpesticidesweresetconstantandequalto2 and0mgL−1respectively.Sincewaterisderivedbyadamwhere sedimentsaretrapped,suspendedsedimentconcentrationwas setconstantandequalto10mgL−1.

Inthisstudy,version2009.93.3ofArcSWATwasused.The catch-mentwasdiscretisedinto73sub-basinswithaminimalareaof 500ha.1642HRUsweregeneratedintegrating8landusesclasses, 23soilclassesand5slopeclasses(%:0–1,1–3,3–5,5–8and8and over).

WholesimulationwascarriedoutdailyfromJanuary1998to March 2009(excluding 4 years’ warm-upfrom 1994 to1997). Thesensitivityof15parametersgoverningdischarge,nitrateand suspendedsedimentdynamicwastestedusingtheArcSWAT2009 sensitivityanalysistool[38].Calibrationofdischarge,nitrate, sus-pendedsedimentandpesticideatdailytime-stepwasperformed manually.PesticideinputvaluesaregiveninTable2.

2.6. Modelevaluation

The performance of the model was evaluated using the Nash–Sutcliffeefficiency (ENS)index [39] and thecoefficientof

Table2

Manuallycalibratedvaluesofpesticidesparameters:half-lifes,KocandCHPSTKOC anddegradationratesinthechannelwaterandinthesedimentbed(respectively CHPSTREAandSEDPSTREA).

Parameters Inputfile Metolachlor Aclonifen

Soilhalf-lifeMet/Tri(days) pest.dat 90 60

KocMet/Tri(mgkg−1/mgL−1) pest.dat 667 13,196

CHPSTKOCMet/Tri(m3_g−1_{) .swq 0.1 2.6}

CHPST/SEDPSTREA(days−1) .swq 0.025 0.025

determination(R2_).E

NSrangesfromnegativeinfinityto1whereas

R2_{rangesbetween0and1.WerefertoKrauseetal.[40]forfurther}

discussionontheseevaluationcriteria.

DailyENSandR2wereappliedondailydischargeforlowflows (belowthe 6.1m3_s−1 _{mean annual discharge), high flow (over} 6.1m3_s−1_{)andtotal(1998–2006 forcalibrationand2007–2009} forvalidation).Theywerealsocalculatedfornitrateloadsand sus-pendedsedimentconcentrationsforthe2007–2009period.Daily andmonthlyR2_{werecalculatedfordissolved,sorbedandtotal} pes-ticidesconcentrationsforcalibrationperiod(2008–2009).Dueto datalimitation,ENSwasnotcalculatedonpesticideconcentration. InthisstudywedeemedENSsatisfactorywhenhigherthan0.36

[41]andR2_{satisfactorywhenhigherthan0.5[42].} 2.7. Waterqualitysimulation

2.7.1. Discharge,nitrateandsuspendedsedimentsimulation DailySWATinterpolatedrainfallandsimulatedwateryield(i.e. theamountofwaterflowingdowntheoutlet)weretotalledfor eachyear.Dailysimulatednitrateandsuspendedsedimentloads attheLarraoutletweretotalledforeach yearandforlow flow andhighflow(usingthe6.1m3_s−1_{threshold).Incomingsuspended} sedimentinagivenreachisthetotaloftheamountofsuspended sedimentscomingfromtheHRUsrelatedtothereach,addedtothe amountenteringfromtheupstreamreach.

2.7.2. Pesticidesimulation

Dailypesticideloadsweretotalledforeachyear,forAprilflood (11/04/08–30/04/08)andJuneflood(14/05/08–18/06/08),andfor lowflowandhighflow(usingthe6.1m3_s−1_{threshold).Total} pesti-cideconcentrationistheconcentrationofpesticideasmeasuredin unfilteredwater.Simulatedtotalconcentrationiscalculatedasthe totalofdissolvedandsorbedpesticideconcentration.The particu-latefractionisthefractionofpesticideinthesorbedphaseandis calculatedasthesorbedpesticideloaddividedbythetotalpesticide load.Theexportationrateiscalculatedastheratiofrompesticide loadexportedatoutletandtheamountofpesticideappliedonthe catchment.Aspesticidetoxicityismorerelevantasconcentration thanas load,long-termsimulation concentrations (1998–2009) werecomparedtoEuropeandrinkingwaterqualitystandardsof 0.1␮gL−1forasinglepesticideandof0.5␮gL−1forbothpesticides beingmodelled.

Correlation analyses were performed on 1998–2008 interannual-averaged metolachlor and trifluralin (both dis-solvedandsorbed)loadsfromreachandHRUstorelatethemto reachandHRUsvariablessuchasslope,soilclasses,rainfall,water, suspendedsedimentandnitrateyields.

3. Results

3.1. Discharge,nitrateandsuspendedsedimentsimulation

According to sensitivity analysis, parameters gov-erning discharge and nitrate were mostly parameters governing runoff and groundwater transfer (CN2, RCHRG DP, GWQMN). Parameters governing suspended sediment were

Table3

Goodness-of-fitindicesfordailydischarge,nitrateloadandsuspendedsediment simulation(p<0.05). Periods R2 _E NS Discharge–calibration 1998–2006 0.52 0.50 Discharge– validation 2007–2009 0.58 0.56 Lowflow 1998–2009 0.02 0.02 Highflow 1998–2009 0.51 0.54 Nitrateload 2007–2009 0.46 0.37

Suspendedsedimentsconcentration 2007–2009 0.36 0.27

mostlyparametersgoverningrunoff(CN2)andin-streamprocesses (CHN2,SPCON,SPEXP).

Fig.2focusesondischargesimulateddailyfromJanuary2007to March2009.Thegoodness-of-fitindicesfordailydischargewere satisfactoryduringbothcalibrationandvalidationperiod(Table3). Theywerealsosatisfactoryforhighflowbutunsatisfactoryforlow flow(Table3).

Nitrate daily load predictions (Fig. 3(a)) were correlated to observationsfor the2007–2009period(Table3).Observedand simulated cumulated nitrate loads in 2007 were 2514 and 2388tons respectively, they were 3047 and 3018 tons respec-tively in 2008(Fig.3(b) and (c)).Considering daily and annual loadspredictionon2007–2009,themodelwasconsideredto hind-castpastdailyandannualnitrateloadsbackto1998withlittle error.Annualnitrateloadswerecorrelatedtoannualwateryield (R2_{=0.84,p<0.05).}

Daily simulated suspended sediment concentrations fitted observations(Fig.4(a))althoughENSand R2 wereunsatisfactory (Table 3). Observedand simulated annualsuspended sediment loadswere9000and15,000tonsrespectivelyin2007and58,000 and64,000tonsrespectivelyin2008(Fig.4(b)and(c)).Considering concentrationandloadpredictionon2007–2009,themodelwas consideredtoreconstructpastannualloadsbackto1998withlittle error.Annualsuspendedsedimentloadswerecorrelatedtoannual wateryield(R2_{=0.59,p<0.05).Onaverageacrossthe73reaches,} theannualratioofdeposited/incomingsuspendedsedimentinthe reachisof54%.

3.2. Pesticidesimulation

Simulationresultswereshowntobepoorlysensitiveto applica-tiondatechangeforbothmolecules(resultsnotshown)although pesticidelossesareknowntobedeterminedmainlybytheperiod oftimebetweenapplicationandthefirstrainfalleventandbythe applicationdose[43,44].SWATpesticidecomponentparameters, includingthesoiladsorptioncoefficient(Koc),werealsopoorly sen-sitive(resultsnotshown)exceptthechannelpartitioncoefficient betweenwaterandsuspendedsediment(CHPSTKOC).

3.2.1. SWATperformances

Therangeofdailysimulatedconcentrationsofmetolachlorand trifluralinfollowedtherangeofrespectivemeasurementsduring boththecalibration2008–2009period(Fig.5)andthelong-term validation1998–2009period(Fig.6).Duringthecalibrationperiod, simulateddissolvedandsorbedmoleculeconcentrationsduring low flow matched respectiveobservations. Floodconcentration peaks of dissolved metolachlor were predicted although over-estimated duringApril and June flood events. Sorbed pesticide concentrationpeaksduringthesameperiodwereunderestimated. TrifluralinconcentrationpeaksinMaywereskipped.Themodel didnotsimulatethetrifluralinconcentrationpeakinearly2009. Simulatedpartitionbetweensoluble andsorbedphasesof both moleculesroughlyfollowedtheobservedpartition.

Intermsofloads,averagesimulatedannualloadsatoutletwere in therangeof observed annualloads (Table 4).Thesimulated

Fig.2. Observedandsimulateddailydischarge(m3_s−1_{)attheLarragaugingstation(January2007–March2009).}

Fig.3.Observedandsimulateddaily(a)nitrateload(tons)attheLarragaugingstation(January2007–March2009),(b)2007accumulation(tons)and(c)2008accumulation (tons).

metolachlorandtrifluralinparticulatefractionsatcatchment out-letfittedtheobservedparticulatefractionsand wereconsistent withthefractions simulated in surface runoff out ofthe HRUs (Table4).Atfloodscale,simulationfollowedthemeasuredrangeof

values(Table5).However,metolachlorloadswereunderestimated whereastrifluralinloadswereoverestimatedduringtheAprilflood and underestimatedduringtheJuneflood. R2 _{for monthly} con-centrationsofmetolachlorandtrifluralinwereover0.38except

Fig.4. Observedandsimulateddaily(a)suspendedsedimentconcentration(mgL−1_{)attheLarragaugingstation(January2007–March2009),(b)2007accumulation(tons)} and(c)2008accumulation(tons).

Fig.5.Observedandsimulateddailypesticideconcentrations(␮gL−1_{)attheLarragaugingstation(2008–2009):(a)metolachlorand(b)trifluralin.}

Table4

Observed(2008–2009)andsimulated(1998–2008)averageannualloadsofmetolachlorandtrifluralinineachcatchmentcompartment(mgha−1_yr−1_{)andparticulate} fractionoutofHRUsandatcatchmentoutlet.

Metolachlor(mgha−1yr−1) Trifluralin(mgha−1yr−1)

Observed Simulated Observed Simulated

HRUs Runoff Dissolved:1653 Sorbed:317 Dissolved:475 Sorbed:1237 Partition:0.16 Partition:0.72

Lateralflow Dissolved:200 Dissolved:10

Outlet

Flowwater Dissolved:204 Dissolved:135 Dissolved:60 Dissolved:2

Sorbed:27 Sorbed:14 Sorbed:195 Sorbed:5

Partition:0.12 Partition:0.09 Partition:0.77 Partition:0.72

Table5

Totalanddissolvedmetolachlorandtrifluralinloads(g)duringAprilandJune2008floodsatLarraoutlet.

Metolachlor Trifluralin

2008floods April(11/04–30/04) June(14/05–18/06) April(11/04–30/04) June(14/05–18/06)

Measuredtotalload(g) 5015 16,692 66 3285

Simulatedtotalload(g) 3588 11,283 92 422

Measureddissolvedload(g) 4591 16,178 13 186

Simulateddissolvedload(g) 3262 10,257 26 117

Table6

DailyandmonthlyR2_{formetolachlorandtrifluralinconcentration(dissolved,sorbedandtotal)atLarraoutlet(2008–2009).}

Metolachlor Trifluralin

Daily Monthly Daily Monthly

Dissolved 0.25 0.43 0.21 0.60

Sorbed 0.01 0.38 0.01 0.15

Fig.6.Observedandsimulateddailypesticidetotalconcentrations(␮gL−1_{)attheLarragaugingstation(1998–2009):(a)metolachlorand(b)trifluralin.}

forsorbedandtotaltrifluralin.R2_{fordailyconcentrationsdidnot}

exceed0.26(Table6).

3.2.2. Long-termpesticideexportationbalancesatoutlet

During the 1998–2008 period, high flows represented 17% ofthetime consideringthe6.1m3_s−1 _{threshold. 50%of nitrate} loadand57%ofsuspendedsedimentloadwereexportedduring highflow. Annualpesticideloads areshownin Fig.7.Thetotal metolachlorloadvariedbetween0.1kgyr−1(2003)and80kgyr−1 (2000), whereas trifluralin varied between 0.01kgyr−1 (2003) and2.3kgyr−1 (2000).Averagetotal metolachlorand trifluralin annualloadswere16.7kg(SD=23kg)and0.8kg(SD=1kg) respec-tively.Thetotalmetolachlorandtrifluralinexportedwerearound 0.072%and0.005%oftheappliedamountrespectively (exporta-tionratewas1%outoftheHRUs forbothmolecules).Atoutlet 61%ofthetotalmetolachlorand52%ofthetotaltrifluralinwere exportedduringhighflows(Table7).Outofthe4108simulated days(1998–2009),metolachloratLarraoutletexceededthe Euro-peanstandardthresholdof0.1_␮gL−1for149daysandtrifluralin exceededthis same threshold for 17 days(3.6 and 0.4% ofthe time period respectively). Maximummetolachlor concentration was5.4mgL−1,predictedinJuly2001whereasmaximum triflu-ralinconcentrationwas0.2mgL−1,predictedinApril1998(Fig.6). Consideringthesumoftotalmetolachlorandtotaltrifluralin con-centrations,thethresholdof0.5␮gL−1 wasexceededduring24 daysatcatchmentoutlet.

3.2.3. Pesticidetransfercontrollingfactors

Atcatchmentscale,thesimulatedpreferredpathwayof pesti-cidetransferwassurfacerunoff(Table4)withassociatednitrate andsuspendedsedimentexportations.Atsub-basinscale, triflu-ralin and metolachlor in the sorbed phase were correlated to suspendedsedimentloads(R2_{=0.69and0.64,respectively).} Triflu-ralinandmetolachlorinthedissolvedphasewerepoorlycorrelated tonitrateloads(R2_{=0.21and0.16,respectively).Eventually,} triflu-ralinandmetolachlorinthedissolvedphasewerebettercorrelated tosuspendedsedimentloads(R2_{=0.33and0.67respectively).At} HRUscale,thecorrelationanalysisdidnotshowanycorrelation betweensuspended sediment loads,metolachlor and trifluralin loadsinthedissolvedphaseandcatchmentvariables.Metolachlor andtrifluralinloadsinthesorbedphasecorrelatedweaklyto sus-pendedsedimentyields(R2_{=0.24and0.43respectively).}

4. Discussion

4.1. Discharge,nitrateandsuspendedsedimentsimulation

OverallENSandR2fordailydischargewereoverthesatisfactory threshold.However,gapsbetweenobservedandsimulatedvalues areexplainedbyerrorsinobservedandsimulatedvalues.Errors inobservedvaluescanstemfromtheprecisionofthesensorand fromtheuseofaratingcurve.Errorsinsimulatedvaluescanbe attributedto(i)actuallocalrainfallstormsthatwerenotwell rep-resentedbytheSWATrainfalldatainterpolationand(ii)theflow

Fig.7. Simulatedannualdissolvedandsorbedpesticideloads(kgyr−1_{)atLarraoutlet(1998–2008):(a)metolachlorand(b)trifluralin.}

uncertaintyinthedensenetworkofcanalsdivertedfromtheriver networktobringpartofriverflowtomanywatermills.ENSforlow flowwasbelowthesatisfactorythreshold.LowENSduringlowflow hastoberelatedtoitsgenerallypoorperformanceinperiodsoflow flow:withonlyminorsimulationerrorsthedenominatorofthe equationtendstowardszeroandENSapproachesnegative infin-ity.LowENS ishoweverofminorconcernsincepesticideswere showntobemostlyexportedduringfloods.Saveriverdischarge simulationqualityishowevercomparabletoOeurngetal.[45].

Dailynitratepredictionisrelatedtodailydischargeprediction, asitisaverysolublenutrient.Biasisthereforelinkedtothe inac-curacyofdischargepredictionsmentionedabove.Uncertaintyin point-sourcenutrientinputalsoexplainsbias.Inaddition, aver-agedlanduseandassociatedmanagementpracticeinputsmaynot reflectwellenoughtheactualandlocallanduseandmanagement practices.Theyalsocanevolveoverthemodelledperioddepending onagriculturalpolicytrends.

ENSandR2 fordailysuspendedsedimentsconcentrationwere belowthesatisfactorythreshold.Calibrationofsedimentis diffi-cultintheSavecatchmentconsideringthedensenetworkofcanals divertedfromtherivernetwork.Thesuccessionofdamsandgates trapssedimentstilltheirrandomemptyingbacktorivernetwork. Uncertaintyinpoint-sourcesuspendedsedimentsinputmayalso

explainbias.Suspendedsedimentresultswerehoweverconsistent withOeurngetal.[45].

4.2. Pesticidesimulation 4.2.1. SWATperformances

R2_{ofdailypesticideconcentrationwasbelowthesatisfactory} threshold.AshighlightedbyLuoetal.[24]possibletimeshiftsin theprecipitation,agriculturalactivity,andmeasurementsforflow andwaterqualitydatarenderdailypesticidestatisticspoor.Errors inpesticideconcentrationandloadpredictionsmayberelatedto(i) aninadequatecalibrationoftheparametersgoverningflowand sol-uble(e.g.nitrate)andparticulate(e.g.suspendedsediment)phases transportandto(ii)aninadequatecalibrationofpesticide compo-nentparameters.

Sorbedpesticideunderestimationhastoberelatedtothehigh depositionrateof suspended sediment in thechannel. Also, as Kocwas showntobe poorlysensitiveand asCHPSTKOC isset asaconstantvalue,bothpartitioncoefficientsmodelledinHRUs andinreachesvarydependingonsuspendedsediment concen-tration.Theymaynotreflectactualvariations: metolachlorand trifluralin measurementsat outlet showvarious inversions(i.e. [soluble]<[sorbed]formetolachlorand[soluble]>[sorbed]for

tri-Table7

Averagemetolachlorandtrifluralindailyloads(gd−1_{)duringhighflowandlowflowindissolvedandsorbedphasesandpercentageofpesticideloadexportedduringhigh} flowatLarraoutlet(1998–2008).

Dissolved Sorbed

Lowflow(gd−1) Highflow(gd−1) Floodlosses(%) Lowflow(gd−1) Highflow(gd−1) Floodlosses(%)

Metolachlor 19 137,118 61 2 13,712 61

fluralin)ofthepartitioncoefficientKd,asdefinedbyTaghavietal.

[30]atcatchmentoutlet,thatwerenotmodelled.

Regarding agricultural activity, i.e. land use and associated managementpractices,themodeldidnotsimulateanytrifluralin concentrationriseinearly2009.Thelanduseinputmapwasbased ona2005landcoversatelliteimagethatmaynotreflectlong-term actualoperations: e.g.canola,representing1%ofthecatchment landusein2005,hasbeengrownincreasinglyoverthepastfive yearsinthenorthernpartofthecatchment.Upto2009,canola wasmanagedwithanaveragespreadof1kgha−1oftrifluralinin August.ThiswasnottakenintoaccountinthemodelbecauseSWAT landuseapproximationwassettoskiplanduserepresentingless than10%ofthesub-basinarea.Reliablepesticidesupplyinputdata arethereforeanecessaryconditiontoachievesatisfactory simula-tionoutputs.

Alastsourceoferroristhemodelledtransferpathways:SWAT simulatespesticidetransfersthroughsurfacerunoffand subsur-facelateralflowbutnotthroughgroundwaterflow,drainflownor atmosphericdeposition[31].Thismayleadtoadditionalerrorsin solublepesticidesimulationalthoughleachingofpesticidesinto deep groundwater and a possible inputof pesticides into sur-facewatersbyoutflowinggroundwaterisknowntobenegligible

[46,47].Inaddition,nopoint-source,suchasthecleaningofthe equipment,wasmodelledinthisstudy[12,48].

Finally,dailypesticidetotalconcentrationsR2_{wereinthelower} rangeofthevaluesmentionedbyNeitschetal.[21].Theyreported R2 _{rangingfrom0.41to0.28fordailymetolachlortotal} concen-trationandR2_{rangingfrom0.51to0.02fordailytrifluralintotal} concentration.

4.2.2. Long-termpesticideexportationbalanceatoutlet

Simulationshowedthatpesticideswereexportedmainlyduring floodevents.Roleoffloodsinpesticideexportationwaspreviously shownattheSavecatchmentoutlet[29].Asinterannualaverage, exportationratesofbothpesticideshavetoberelatedtotheir appli-cationtime(April)andthemonthofmaximumdischarge(May). However,simulatedvalues ofexportationduringa similarhigh flowperiodwerelessthanmeasuredvaluesinasmallcatchment oratfieldscale[49,50].Thesizeofthedrainedarea,butalsothe soilandthelandusemaymodulatetheexportation.

Pesticideexportationsfromlandtooutlet werelessthan1% ofappliedamount.Suchavaluewasreportedbyvariousstudies onmetolachlorinFrance,SwitzerlandandQuébec[51–53]andon otherpesticides[50,53,54].About93%and99%ofmetolachlorand trifluralinrespectivelyenteringthestreamnetworkdoesnotreach theoutlet,suggestinghighdeposition(discussedabove)and degra-dationinstreamwater.Thelatterwouldbeconsistentwiththe carbonconsumptionbyriverbiotashownbySánchez-Pérezetal.

[55]onasimilarcatchmentofGascogne.

Trifluralin concentration exceeded the 0.1␮gL−1 maximum permissiblelevellessthanmetolachlorconcentration.Inthemodel, trifluralinwaslessappliedthanmetolachloramongthecatchment, itssoilhalf-lifewas60days(insteadof90daysformetolachlor)and itstransportwasmorelikelydependingonlandandriverbed ero-sion.ItisworthnotingthattheEuropeanstandardof0.5mgL−1 wasexceededbyapoolofonlytwomoleculesduring0.6%ofthe simulationtime.

4.2.3. Pesticidetransfercontrollingfactors

Runoffwasshowntobethepreferredsimulatedpathwayfor pesticideexportation.Luoetal.[24,56]alreadyreportedthe con-trolofrunoffonpesticidetransfersimulation.Intheenvironment, pesticidesmaybesorbedontomineralsuspendedmatter, Partic-ulateOrganicCarbon(POC)andcomplexedbyDissolvedOrganic Carbon(DOC)[29].Bettercorrelationwasfoundbetweendissolved phasesandsuspendedsedimentthanbetweendissolvedphases

andnitrate.Althoughpesticidetransferthroughgroundwateris notyetmodelledinSWAT,contrarilytonitrate,therelationships highlightedacontrolofbothsurfaceandsub-surfacerunoffs dur-ingfloodsondissolvedphaseexportation.Thisisconsistentwith thepesticides’actualabilitytosorbontoDOC,i.e.smallerthanthe 0.45␮mmeshfilter,anditstransferthroughsub-surfaceflow[29]. 5. Conclusions

TheSWATmodelwasappliedtosimulatepesticidetransferat theoutletofalargeintensiveagriculturalcatchment.Simulation resultsweredeemedtobesatisfactory,takingintoaccountthat themodelledtransferprocessesweresimplifiedanddonot repre-sentallactualtransfers.FurtherimprovementsoftheSWATmodel maybeinvestigated.Thetransferofpesticideinthedissolvedphase fromlandtoriverthroughgroundwatercouldbetestedtoassess possible water-tableeffect. Also, investigating the variationsof thepartitioncoefficientbetweenobserveddissolvedandsorbed phasesduringfloodsinvariouspointsintheriver(areasofrapid, deep,etc.)wouldhelptoassesstheaccuracyoftheSWATmodelled partitionatflood-eventscale.

However,extrapolationtootherchemicalsisconceivableand SWATwasshowntobepromisingforprovidingarobustdecision toolforwaterqualitymanagers.Floodsarequickevents.Suchatool wouldhelp(i)totarget‘whatandwhen’tomonitor,(ii)tohighlight pesticideconcentrationpeakswithoutcost-intensivefield mea-surementsandpredictfuturepeaksand(iii)toevaluateexported loadsascontaminationindicator.Suggestionoflocalised mitiga-tionpracticestoreachwaterpolicyobjectivessuchastheEuropean WaterFrameworkDirectiveismadepossible.

Acknowledgements

ThisworkwasperformedaspartoftheEUInterregSUDOEIVB program(SOE1/P2/F146AguaFlashproject, http://www.aguaflash-sudoe.eu)andfundedbyERDFandtheMidi-PyrénéesRegion.

WesincerelythanktheCACGfordischargedata,Météo-France formeteorologicaldata,AEAGforlong-termpesticide concentra-tion measurements,Arnaud Mansat and Erwan Motte for their data-processingexpertise,andHuguesAlexandreforITsupport. References

[1] C.Soulsby,A.F.Youngson,H.J.Moir,I.A.Malcolm,Finesedimentinfluenceon salmonidspawninghabitatinalowlandagriculturalstream:apreliminary assessment,Sci.TotalEnviron.265(2001)295–307.

[2]G.Zalidis,S.Stamatiadis,V.Takavakoglou,K.Eskridge,N.Misopolinos,Impacts ofagriculturalpracticesonsoilandwaterqualityintheMediterraneanregion andproposedassessmentmethodology,Agric.Ecosyst.Environ.88(2002) 137–146.

[3]M.Probst,N.Berenzen,A.Lentzen-Godding,R.Schulz,M.Liess,Linkinglanduse variablesandinvertebratetaxonrichnessinsmallandmedium-sized agricul-turalstreamsonalandscapelevel,Ecotoxicol.Environ.Safe.60(2005)140–146. [4]M.Zeiger,N.Fohrer,Impactoforganicfarmingsystemsonrunoffformation processes—along-termsequentialrainfallexperiment,SoilTillageRes.102 (2009)45–54.

[5]J.G.M.Cuppen,P.J.VandenBrink,E.Camps,K.F.Uil,T.Brock,Impactofthe fungicidecarbendaziminfreshwatermicrocosms.I.Waterquality,breakdown ofparticulateorganicmatterandresponsesofmacroinvertebrates,Aquat. Tox-icol.48(2000)233–250.

[6]P.J.VandenBrink,J.Hattink,F.Bransen,E.VanDonk,T.Brock,Impactof thefungicidecarbendaziminfreshwatermicrocosms.II.Zooplankton,primary producersandfinalconclusions,Aquat.Toxicol.48(2000)251–264. [7]R.M.Niemi,I.Heiskanen,J.H.Ahtiainen,A.Rahkonen,K.Mäntykoski,L.Welling,

etal.,Microbialtoxicityandimpactsonsoilenzymeactivitiesofpesticidesused inpotatocultivation,Appl.SoilEcol.41(2009)293–304.

[8]A.G.Becker,B.S.Moraes,C.C.Menezes,V.L.Loro,D.R.Santos,J.M.Reichert,etal., Pesticidecontaminationofwateraltersthemetabolismofjuvenile silvercat-fish,Rhamdiaquelen,Ecotoxicol.Environ.Safe.72(2009)1734–1739. [9] EC,Directive2006/11/ECoftheEuropeanParliamentandoftheCouncilof15

February2006onpollutioncausedbycertaindangeroussubstancesdischarged intotheaquaticenvironmentoftheCommunity,2006.

[10]EC,Directive2000/60/ECoftheEuropeanParliamentandoftheCouncilof23 October2000establishingaframeworkforCommunityactioninthefieldof waterpolicy,2000.

[11]EC,Directive98/83/ECof3November1998onthequalityofwaterintended forhumanconsumption,1998.

[12]K.Holvoet,P.Seuntjens,P.A.Vanrolleghem,Monitoringandmodeling pesti-cidefateinsurfacewatersatthecatchmentscale,Ecol.Model.209(2007) 53–64.

[13]Y.Onishi,A.R.Olsen,M.A.Parkhurst,G.Whelan,Computer-based environ-mentalexposureandriskassessmentmethodologyforhazardousmaterials, J.Hazard.Mater.10(1985)389–417.

[14] A.S.Donigian,N.H.Crawford,ModelingPesticidesandNutrientsonAgricultural Lands,EnvironmentalProtectionAgency,Washington,DC,1976.

[15] W.G.Knisel,CREAMS:AField-scaleModelforChemicals,RunoffandErosion fromAgriculturalManagementSystems,Report,U.S.Departmentof Agricul-ture,Washington,DC,1980.

[16]R.A.Leonard,W.G.Knisel,D.A.Still,GLEAMS:GroundwaterLoadingEffectsof AgriculturalManagementSystems,Trans.ASAE30(1987)1403–1918. [17]L.C.Brown,T.O.Barnwell,TheEnhancedWaterQualityModelsQUAL2Eand

QUAL2E-UNCAS:DocumentationandUserManual,1987.

[18] C.D.Brown,J.M.Hollis,SWAT—asemi-empiricalmodeltopredict concentra-tionsofpesticidesenteringsurfacewatersfromagriculturalland,Pestic.Sci. (1996)41–50.

[19]R.Haverkamp,POWERanewapproachofwatershedmodelling,in:Conference InvitedbyGriffinUniversity,Brisbane,Australia,6–12February,2002. [20]J.G.Arnold,R.Srinivasan,R.S.Muttiah,J.R.Williams,Largeareahydrologic

mod-elingandassessment.I.Modeldevelopment,J.Am.WaterResour.Assoc.34 (1998)73–89.

[21]S.L.Neitsch,J.G.Arnold,R.Srinivasan,PesticidesFateandTransportPredicted bytheSoilandWaterAssessmentTool(SWAT)—Atrazine,Metolachlorand TrifluralinintheSugarCreekWatershed,2002.

[22] K.Holvoet,A.vanGriensven,P.Seuntjens,P.A.Vanrolleghem,Sensitivity anal-ysisforhydrologyandpesticidesupplytowardstheriverinSWAT,Phys.Chem. Earth30(2005)518–526.

[23]M.Larose,G.C.Heathman,L.D.Norton,B.Engel,Hydrologicandatrazine sim-ulationoftheCedarCreekwatershedusingtheSWATmodel,J.Environ.Qual. 36(2007)521–531.

[24]Y.Luo,X.Zhang,X.Liu,D.Ficklin,M.Zhang,Dynamicmodelingof organophos-phatepesticideloadinsurfacewaterinthenorthernSan JoaquinValley watershedofCalifornia,Environ.Pollut.156(2008)1171–1181.

[25]J.C.Revel,M.Guiresse,Erosionduetocultivationofcalcareousclaysoilsonthe hillsidesofSouthwestFrance.I.Effectofformerfarmingpractices,SoilTillage Res.35(1995)147–155.

[26] C.Oeurng,S.Sauvage,J.M.Sánchez-Pérez,Dynamicsofsuspendedsediment transportandyieldinalargeagriculturalcatchment,south-westFrance,Earth Surf.Process.Landforms35(2010)1289–1301.

[27]C. Oeurng,S.Sauvage,J.M.Sánchez-Pérez,Temporalvariabilityofnitrate transportthroughhydrologicalresponseduringfloodeventswithinalarge agriculturalcatchmentinsouth-westFrance,Sci.TotalEnviron.409(2010) 140–149.

[28]C.Oeurng,S.Sauvage,A.Coynel,E.Maneux,H.Etcheber,J.-M.Sánchez-Pérez, Fluvialtransportofsuspendedsedimentandorganiccarbonduring flood eventsinalargeagriculturalcatchmentinsouthwestFrance,Hydrol.Process. 25(2011)2365–2378.

[29]L.Taghavi,G.Merlina,J.-L.Probst,Theroleofstormflowsinconcentration ofpesticidesassociatedwithparticulateanddissolvedfractionsasathreat toaquaticecosystems.Casestudy:theagriculturalwatershedofSaveriver (SouthwestofFrance),Knowl.Manage.Aquat.Ecosyst.400(2011)11. [30]L.Taghavi,J.L.Probst,G.Merlina,A.L.Marchand,G.Durbe,A.Probst,Flood

eventimpactonpesticidetransferinasmallagriculturalcatchment (Mon-tousseatAurade,southwestFrance),Int.J.Environ.Anal.Chem.90(2010) 390–405.

[31]S.L.Neitsch,J.G.Arnold,J.R.Kiniry,J.R.Williams,SoilandWaterAssessment ToolTheoreticalDocumentation—Version2005,2005.

[32]J.R.Williams,A.D.Nicks,J.G.Arnold,Simulatorforwaterresourcesinrural basins,J.Hydraul.Eng.111(1985)970–986.

[33]S.C.Chapra,SurfaceWater-QualityModeling,McGrawHill,Boston,USA,1997.

[34]F. Macary,E.Lavie,G.Lucas,O.Riglos,Méthodedechangementd’échelle pourl’estimationdupotentieldecontaminationdeseauxdesurfaceparl’azote, Ingénieries-EAT46(2006)35–49.

[35]J.M.Lescot,P.Bordenave,Adecisionsupporttochoosebetweenchangesof agriculturalpractices:aspatiallydistributedcost-effectivenessassessment framework,in:IntegratedAssessmentofAgricultureandSustainable Develop-ment,SettingtheAgendaforScienceandPolicy,EgmondaanZee,NLD,2009, pp.452–453.

[36]H.L.Penman,Naturalevaporationfromopenwater,baresoil,andgrass,Proc. Roy.Soc.Lond.A193(1948)120–146.

[37]J.L.Monteith,Evaporationandenvironment,in:TheStateandMovementof WaterinLivingOrganisms,CambridgeUniversityPress,Swansea,1965,pp. 205–234.

[38] A.VanGriensven,T.Meixner,S.Grunwald,T.Bishop,M.Diluzio,R.Srinivasan,A globalsensitivityanalysistoolfortheparametersofmulti-variablecatchment models,J.Hydrol.324(2006)10–23.

[39]J.E.Nash,V.Sutcliffe,Riverflowforecastingthroughconceptualmodels.PartI. Adiscussionofprinciples,J.Hydrol.10(1970)282–290.

[40]P.Krause,D.P.Boyle,F.Bäse,Comparisonofdifferentefficiencycriteriafor hydrologicalmodelassessment,Adv.Geosci.5(2005)89–97.

[41] M.W.VanLiew,J.Garbrecht,HydrologicsimulationofthelittleWashitariver experimentalwatershedusingSWAT,J.Am.WaterResour.Assoc.39(2003) 413–426.

[42]C.H.Green,M.D.Tomer,M.DiLuzio,J.G.Arnold,Hydrologicevaluationofthe soilandwaterassessmenttoolforalargetile-drainedwatershedinIowa,Trans. ASABE49(2006)413–422.

[43]J.S.Martins,L.B.Owens,Atrazine,deethylatrazine,anddeisopropylatrazinein surfacerunoffformconservationtilledwatersheds,Environ.Sci.Technol.37 (2003)944–950.

[44]K.Müller,M.Deurer,H.Hartmann,M.Bach,M.Spiteller,H.G.Frede, Hydrologi-calcharacterisationofpesticideloadsusinghydrographseparationatdifferent scalesinaGermancatchment,J.Hydrol.272(2003)1–17.

[45]C.Oeurng,S.Sauvage,J.M.Sanchez-Perez,Assessmentofhydrology,sediment andparticulateorganiccarbonyieldinalargeagriculturalcatchmentusingthe SWATmodel,J.Hydrol.401(2011)145–153.

[46] B.Röpke,M.Bach,H.G.Frede,DRIPS—aDSSforestimatingtheinputquantityof pesticidesforGermanriverbasins,Environ.Model.Soft.19(2004)1021–1028. [47]R.P. Richards,D.B.Baker, Pesticide concentrationpatterns in agricultural drainagenetworksintheLakeErieBasin,Environ.Toxicol.Chem.12(1993) 13–26.

[48]S.Reichenberger,M.Bach,A.Skitschak,H.G.Frede,Mitigationstrategiesto reducepesticideinputsintoground-andsurfacewaterandtheireffectiveness: areview,Sci.TotalEnviron.384(2007)1–35.

[49] X.Louchart,M.Voltz,P.Andrieux,R.Moussa,Herbicidetransporttosurface watersatfieldandwatershedscalesinaMediterraneanvineyardarea,J. Envi-ron.Qual.30(2001)982–991.

[50]M.Rabiet,C.Margoum,V.Gouy,N.Carluer,M.Coquery,Assessingpesticide concentrationsandfluxesinthestreamofasmallvineyardcatchment—effect ofsamplingfrequency,Environ.Pollut.158(2010)737–748.

[51]S.M.Novak,J.M.Portal,M.Schiavon,Effectsofsoiltypeuponmetolachlorlosses insubsurfacedrainage,Chemosphere42(2001)235–244.

[52] C.Leu,H.Singer,C.Stamm,S.R.Muller,R.P.Schwarzenbach,Variabilityof her-bicidelossesfrom13fieldstosurfacewaterwithinasmallcatchmentaftera controlledherbicideapplication,Environ.Sci.Technol.38(2004)3835–3841. [53]L.Poissant,C.Beauvais,P.Lafrance,C.Deblois,Pesticidesinfluvialwetlands

catchments underintensiveagricultural activities,Sci.Total Environ.404 (2008)182–195.

[54]N.Berenzen,A.Lentzen-Godding,M.Probst,H.Schulz,R.Schulz,M.Liess,A comparisonofpredictedandmeasuredlevelsofrunoff-relatedpesticide con-centrationsinsmalllowlandstreamsonalandscapelevel,Chemosphere58 (2005)683–691.

[55]J.M.Sánchez-Pérez,M.Gerino,S.Sauvage,P.Dumas,E.Maneux,F.Julien,etal., Effectsofwastewatertreatmentplantpollutiononin-streamecosystems func-tionsinanagriculturalwatershed,Int.J.Lim.45(2009)79–92.

[56]Y.Luo,M.Zhang,Management-orientedsensitivityanalysisforpesticide trans-portinwatershed-scalewaterqualitymodelingusingSWAT,Environ.Pollut. 157(2009)3370–3378.