Pesticide Science Pestic Sci 55 :175–182 (1999)

Influence of adsorption–desorption phenomenaon pesticide run-off from soil using simulatedrainfallVe� ronique Gouy,1 Jeanne-Chantal Dur,2 Raoul Calvet,3 Rene� Belamie¹1 and

Ve� ronique Chaplain2*1 Divis ion des eaux, Cemagref , 3 bis quai Chauveau 69339 Lyon Cedex Francequalite�

de Phytopharmacie et Chimiques INRA Route de Saint -Cyr 78026 Vers ailles Cedex France2 Unite� Me� diateaursde Science du s ol INRA INAPG 78850 Thiverval-Grignon Cedex France3 Unite�

Abstract : The surface run-oþ of a number of pesticides (diuron, isoproturon, atrazine, alachlor,

aclonifen, triýuralin, lindane and simazine), chosen for their range of adsorption behaviours, was

studied using simulated rainfall applied to small plots over a short time (one hour). Pesticides were

applied together onto bare soil using two diþ erent sandy loam soils from Jailliere and Coet Dan sites.

The surface run-oþ samples were collected throughout the running of the event and concentrations of

pesticides were measured in both liquid and solid phases. Sorption isotherms for isoproturon and

diuron on Jailliere soil as well as eroded particles were measured under equilibrium conditions and

compared to their partitioning during surface run-oþ .

At the rainfall intensity used, both soils generated a large load of eroded particles. The average

run-oþ ýow rate increased with time for the Jallie� re soil, while it remained relatively constant at a

higher level for the Coet Dan soil. The concentrations of each pesticide in the run-oþ samples

decreased as the experiments proceeded. The pesticides were classiüed into two types by their par-

titioning between the solid and liquid phases. Atrazine, simazine, diuron, isoproturon and alachlor

were mainly transported in surface run-oþ water. By contrast, 90% of triýuralin and aclonifen was

adsorbed onto eroded particles. Lindane was intermediate, with a 37% adsorption level. When the

contribution of eroded particles was minor, the agrochemical concentrations were inversely pro-

portional to the water ýow rate. We have proposed a model that describes the mass of chemicals

extracted from soil into surface water during a surface run-oþ event of a given average duration and

ýow rate. This model takes into account the dilution of the soil solution and the desorption of chemi-

cals through two parameters called, respectively, the dilution factor and the extraction retardation

factor. The desorption kinetic was the limiting step in the surface run-oþ of weakly sorbed chemicals,

such as isoproturon.

1999 Society of Chemical Industry(

Keywords: risk assessment ; pesticides ; run-oþ; adsorption; rainfall simulation; soil

1 INTRODUCTION

The use of pesticides is responsible for signiücantbackground contamination of surface waters. Inmany cases, the highest concentrations of pesticidesare correlated with the ürst signiücant rainfall afterapplication.1h3 The processes involved in such rapidpesticide transfer have not yet been elucidated.Better understanding is required to improve riskassessment.

Mechanisms involved in pesticide surface run-oþare very complex and include: (i) water ýow on thesoil surface, which is not homogeneous but rathercharacterised by rills and inter-rills ; (ii) pesticide

extraction from soil to the surface run-oþ stream;(iii) contribution of eroded particles. Hydrodynamicaspects will not be discussed here; we focus rather onpoints (ii) and (iii). Pesticide extraction is a generalterm which simultaneously refers to: (i) diþusive andconvective transport from soil pores to surfacerun-oþ water ; (ii) desorption from soil particles intothe moving liquid boundary ; (iii) dissolution of sta-tionary pesticide particulates ; (iv) transport of pesti-cide particulates and subsequent dissolution.4 Therelative contributions of each process are difficult toevaluate and depend on the soil, the chemicalsinvolved and formulation.

* Corres pondence to : Ve� ronique Chaplain, Unite� de

Phytopharmacie et Me� diateurs Chimiques INRA, Route deSaint. Cyr, 78026 Vers ailles Cedex, France

E-mail : chaplain=vers ailles .inra.fr

¹ Deceas ed.

Contract/ grant s pons or : AIP ‘‘Ecopol’’.(Received 6 March 1998; revis ed vers ion received 4 June 1998;

accepted 17 September 1998)

( 1999 Society of Chemical Industry. Pestic Sci 0031–613X/99/$17.50 175

V Gouy et al

Due to the complexity of these processes, studiesconducted on üeld-size watersheds under naturalrainfall conditions often give large ýuctuations intime of the pesticide concentrations in surface run-oþ. Leonard5 has, however, reported a strong corre-lation between surface run-oþ concentrations,including water and sediments, and pesticide concen-trations in the surface 10mm of watershed soil. Incontrast, experiments conducted on small plots usingsimulated rainfall show more idealised behaviour.The pesticide concentration varies continuously withtime even if these concentrations are higher thanthose measured under natural conditions.6

Run-oþ models are usually based on the deünitionof an ‘active zone’ interacting with run-oþ. Instanta-neous and reversible equilibrium is assumed, so thatthe pesticide concentration in the soil solution can becomputed from the partition coefficient oftenKdused in its normalised form where oc(Koc\ Kd/oc,is the fraction of organic carbon present in the soil).The sorption constant, is roughly estimatedKoc,either from water solubility or, better, octanol/waterpartition coefficient or measured for each soil/pesticide combination using the batch techniqueunder conditions of equilibrium. Once the depth ofthe mixing zone has been conveniently üxed (oftenchosen to be 10mm), an adjustable parameter calledthe ‘extraction ratio’ is introduced in models toapproximate the eþective mass ratio of the inter-acting soil to water in the run-oþ stream. However,this parameter depends on soil properties, rainfallintensity and surface run-oþ rate and thus limits thepredictive feature of models. Furthermore, non-linearity of isotherm curves and hysteresis in desorp-tion experiments have often been demonstrated butare not taken into account in models. The contribu-tion of colloidal particles to pesticide transfer is alsoignored.

As described by Gouy & Belamie,7 theoretical pre-dictions from the CREAMS sub-models8 did notagree with experimental data provided by rainfallsimulation. It was suggested that the value, whichKdwas measured under equilibrium conditions, did notaccurately reýect the dynamic processes occurringduring run-oþ.

The importance of associated colloid transport inpesticide transfer depended on the chemicals andevents. However, contradictory results are reportedin the literature. The impact of sediments appeared

small for pesticides of water solubility below 2mglitre~1.9 Recently, however, Munoz10 has observed alinear relationship between the loads of sedimentsand the concentrations of various chemicals whichwere all more soluble than 2mg litre~1. Otherauthors have reported that strongly adsorbed chemi-cals, characterized by greater than 1000ml g~1,Kdwere mainly transported in association with col-loids.11,12 Fawcett et al13 distinguished betweenthree classes of agrochemicals according to theirvalues of and postulated that molecules withKd , Kdgreater than 100 were mainly transported in run-oþby sorption onto eroded particles.

Further research is thus required to evaluate theimportance of kinetics in adsorption–desorption pro-cesses and the role of colloidal particles in pesticidetransfer. The speciüc goal of our research was toachieve a better understanding of mechanismsinvolved in pesticide extraction from soil into thesurface run-oþ stream. To reach this goal, we fol-lowed the surface run-oþ of various pesticides undercontrolled conditions provided by a rainfall simula-tor. Experiments were performed on very small plotsof two soil types over a short time in order to revealthe mechanism of extraction. Partition of each pesti-cide between the run-oþ water and eroded particleswas measured. The diþerent pesticides were chosento represent a large range of values of water solu-bility and partition coefficient Pesticides wereKoc .simultaneously applied onto the surface of bare soilin order to compare their behaviour under identicalhydrodynamic conditions.

2 MATERIALS AND METHODS

2.1 Soils

Two sandy loam soils over schist were collected nearNantes (La Jailliere) and near Rennes (Coet Dan).They are representative of the north-west quarter ofFrance and contain 2 and 2.5% (w/w) of organicmatter, respectively. The Jailliere soil had sand, siltand clay fractions comprising respectively 50, 20 and2.5% w/w.

2.2 Pesticides applied

Non-ionising pesticides were chosen and their physi-cochemical characteristics are given in Tables 1 and2. Compounds were applied as commercial suspen-

Pes ticide Water s olubility Koc

a Total amount Dis tribution in run-off

(mg litre~1) (ml g~1) in % of appliedOn s ediment In s olution

Atrazine 33 100 3.8 0.1 3.7

Is oproturon 70 155 2.98 0.08 2.9

Diuron 42 480 1.5 0.1 1.4

Aclonifen 1.4–2.5 8800 1.5 1.4 0.1

a AGRITOX data bas e.

Table 1. Total amounts of

pes ticide in run-off s amples

and their dis tribution

between the s ediment

and liquid phas es for the

Jailliere experiment

176 Pestic Sci 55 :175–182 (1999)

Pesticide run-oþ during simulated rainfall

Pes ticide Water s olubility Koc

a Total amount Dis tribution in run-off

(mg litre~1) (ml g~1) in % of appliedOn s ediment In s olution

Atrazine 33 100 13 0.8 12.2

Simazine 6 130 9 0.6 8.4

Alachlor 240 170 14 0.8 13.2

Lindane 7 1100 17 6. 11.

Trifluralin 1 8000 14 12.6 1.4

a AGRITOX data bas e.

Table 2. Total amounts of

pes ticide in run-off s amples

and their dis tribution

between s ediment and

liquid phas es for the

Coet Dan experiment

sion concentrate formulations to Jailliere soil : Novexýo 80 (Calliope) for diuron, Quartz GT (Rhoü nePoulenc) for isoproturon, Atraphyt EL (Sipcam P)for atrazine and Challenge 600 (Rhoü ne Poulenc) foraclonifen. Unformulated triýuralin, atrazine, sima-zine, lindane and alachlor were applied onto CoetDan soil. Analytical-grade compounds (Cluzeau InfoLabo, Ste-Foy-La-Grande, France) with a purity ofat least 98% were used for validation of our analyti-cal methods. Radiolabelled products [carbonyl-14C]diuron and [carbonyl-14C]isoproturon were used forbatch adsorption studies. They were synthesised inthe laboratory with 99% radiochemical purity andspeciüc activities were 309 and 337MBq mmo1~1,respectively.

2.3 Rainfall simulation apparatus

A rainfall simulator was developed to obtain lowrainfall intensity.14,15 This device was composed ofthree parts :

(1) A pierced sprinkling tank equipped with 166capillary rubber tubes (6cm long ] 0.5mminternal diameter). The rain was controlled by avariable volume pump. The rainfall height was3m.

(2) A 0.33m2 receiving tank (receptor) containing a12-cm-thick layer of homogenised soil andinclined at 7%. The lower side was piercedalong a horizontal line at a height of 12cm tocollect run-oþ samples in a receiving gutter.

(3) A percolator tank below the soil-ülled tank con-taining coarse sand to collect percolated watersamples.

Pesticides were applied onto the bare soil surface.The initial soil moisture contents were 30 and 40%(w/w) and the applied doses of active ingredientswere 150mg m~2 and 55mg m~2 for Jailliere andCoet Dan soil, respectively. The simulated rainfall ofintensity 33 and 44mm h~1 respectively, began 20hafter application and surface run-oþ samples werecollected at intervals over the hour of irrigation.After collection, run-oþ samples were stored at 4¡Cfor up to 24h. Separation of solid and liquid phaseswas carried out by means of continuous centrifu-gation (Centrifuge CEPA, Bioblock Scientiüc at45000g for 15min and the concentration of erodedparticles (g litre~1) was measured. The granulo-Cp

metric distribution of eroded particles was thendetermined.

2.4 Extraction and pesticide analysis

Five hundred millilitres of the liquid phase wasextracted using dichloromethane (3] 50ml). Theorganic phase was concentrated under light vacuum,dissolved in toluene (1ml), applied onto a ýorisil car-tridge and eluted by toluene ] ethyl acetate (50] 50by volume, 25ml) for puriücation. After reduction todryness, pesticide was dissolved in isooctane for GCanalysis or in water ] acetonitrile (50] 50 byvolume) for HPLC analysis.

The solid phase was extracted using acetone(2] 100ml). The organic phase was reduced to50ml and transferred to a separating funnel contain-ing pure water (450ml). This mixture was extractedwith dichloromethane (3] 50ml). The organicsolvent was completely evaporated and pesticide wasdissolved in toluene (1ml). The puriücation step wasperformed as described above.

The solution pesticide concentration (lgCslitre~1) and adsorbed pesticide concentration Qs(lg g~1) were measured by CPG or HPLC analysesaccording to multi-residue methods deüned byCemagref Laboratory. Quantitative analyses weremade using an electron-capture detector for tri-ýuralin aclonifen, alachlor and lindane and using athermionic speciüc detector for atrazine in the CoetDan simulation and for simazine analyses. HPLCcoupled with UV detector was used for quantitativeanalyses of diuron and isoproturon as well as foratrazine in the case of the Jailliere simulation.Analytical-grade compounds were used to validateanalytical method. Recovery was close to 100% foreach compound. It was always greater than 90%,except for lindane, for which recovery was 70%.

Gas chromatography analysis used a Varian 3400device and an automatic injector 8200 equippedwith an ECD detector (source 63Ni). Thetemperature detector was 300¡C (injector 250¡C),and the injected volume was 1ll. The column was30m ] 0.32mm ] 0.50lm non-polar (JW, DB608).The oven temperature program was : initial tem-perature 50¡C held for 1min, 20¡Cmin~1 up to150¡C, 10¡Cmin~1 up to 280¡C, held for 20min.The carrier gas was nitrogen with a ýow rate of1.5ml min~1 at 50¡C.

Pestic Sci 55 :175–182 (1999) 177

V Gouy et al

When the thermionic speciüc detector was used,the temperature of the detector was 270¡C, the ýowrate of hydrogen was 4ml min~1 and air175ml min~1. The temperature of the injector was260¡C and the injected volume was 1ll. The columnwas 30m ] 0.32mm ] 0.25lm non-polar (JW,DB1). The oven temperature program was : initialtemperature 100¡C held for 1min, 10¡Cmin~1 up to200¡C, 5¡Cmin~1 up to 250¡C, 10¡Cmin~1 up to320¡C. The carrier gas was nitrogen with a ýow rateof 1.5ml min~1 at 50¡C.

HPLC analysis used a Kontron device ütted witha guard-column of Lichrospher 60 RP Select B type(4mm ] 4mm) and an analytical column of Lich-rospher 60 RP Select B type (250mm ] 4.6mm).Elution with acetonitrile ] water proceeded at1ml min~1 with an isocratic pump; elution beganwith 28] 72 by volume for 28min, followed by anincrease to 68] 32 by volume over 10min, anincrease to 90] 10 by volume over 5min, and a sta-tionary phase during 5min. UV absorbance wasmonitored at 224nm for atrazine analysis and 245nmfor diuron and isoproturon.

The values of the detection limit in liquid and solidphase were respectively 0.03lg litre~1 and10lg kg~1 for aclonifen, 0.52lg litre~1 and3lg kg~1 for atrazine in Jailliere, 0.5lg litre~1 and0.5lg kg~1 for diuron, 0.11lg litre~1 and0.5lg kg~1 for isoproturon, 0.08lg litre~1 and30lg kg~1 for alachlor, 0.5lg litre~1 and 70lg kg~1for atrazine in Coet Dan, 0.02lg litre~1 and20lg kg~1 for lindane, 0.5lg litre~1 and 50lg kg~1for simazine, 0.05lg litre~1 and 20lg kg~1 for tri-ýuralin.

2.5 Adsorption isotherms

Twenty grams of sieved (\2mm) air-dried soil par-ticles were mixed in a glass bottle with 20ml ofdiuron or isoproturon solutions of variable initialconcentrations, in 0.01M The initial con-Ci@ , CaCl2 .centration range was 0.1, 0.5, 1 and 5mg litre~1 forboth compounds. All the pesticides used were radiol-abelled. The initial radiolabelled pesticide solutionwas such that the activity in the batches was200Bq ml~1. Samples were mixed on a rotary shakerat room temperature for 24h. After decantating, thesupernatant solution was centrifuged, radioactivitywas determined by scintillation counting using aWallac 1409 scintillation apparatus and the equi-librium concentration was derived. The amount,(Ce)

(mg g~1) of pesticide sorbed was calculated usingQethe relation

Qe\ (Ci[ Ce) · V/m

where V was the volume of solution (litre) and m themass of soil (g).

Two replicates were used for each sample andchecked with a blank (a volume V of aqueous pesti-cide solution without soil). Experimental error was

estimated to lie within a range of 2 to 5% of quanti-tative results.

The Freudlich model

Qads\ K f Cel@n

was applied to each isotherm curve. Adsorption mea-surements on sediments diþered only in their parti-cle concentration of 7.2g litre~1. Additionaladsorption measurements were performed with aconcentration of Jailliere soil of 10g litre~1, whichwas close to the concentration of sediments inrun-oþ water samples.

3 RESULTS

3.1 Pesticide surface run-off during rainfall

simulation

3.1.1 Pesticide transfer in run-oþ waterAn average ýow rate was derived for each run-oþFwsample from the volume collected during the timestep *t. This average value continuously increasedwith time for the Jailliere experiment and was con-stant for the Coet Dan experiment over the 15min ofsurface run-oþ (Fig 1). Values reached in the lattersoil were relatively large, greater than 200ml min~1corresponding to a intensity of 36mm h~1.

Over the same period, average concentrations, Csof each pesticide measured in the liquid phase ofeach run-oþ sample decreased with time due to adilution eþect (Fig 2). The ratio of the concentrationof the two molecules also varied with time duringeach experiment, indicating that a dynamic process

Figure 1. Variation in the average run-off flow rate left y-axis )(L,and average particle concentration right y-axis ) with time(K,meas ured during rainfall s imulation on (A) Jailliere s oil (full

s ymbols ) and (B) Coet Dan s oil (empty s ymbols ).

Figure 2. Average pes ticide concentration in the liquid phas e of

run-off s amples for each time s tep. (A) Jailliere s oil diuron,(…is oproturon, aclonifen, atrazine and (B) Coet Dan s oil(>) (>) (=)trifluralin, lindane, atrazine, s imazine,(È) (K) ()) (|) (L)

alachlor.

178 Pestic Sci 55 :175–182 (1999)

Pesticide run-oþ during simulated rainfall

was also involved in pesticide transfer. The highestconcentrations (greater than 500lg litre~1) wererecorded at the beginning of the experiment withatrazine, isoproturon and alachlor and to lesserdegree with simazine, diuron and lindane. Triýuralinon Coet Dan soil and aclonifen on Jailliere were dis-tinguished from other pesticides by low concentra-tions (below 100lg litre~1). Both these pesticideshave low solubility in water and a large partitioncoefficient (see Tables 1 and 2). Furthermore, thetemporal variations in the concentration of triýuralinwere small (from 51 to 10lg litre~1) compared tovariations recorded over the same time with otherscompounds. Indeed the concentrations of simazine,alachlor and atrazine decreased by factors of 10, 15and 21, respectively. In contrast, the decrease of the

values of aclonifen, from 20 to 6lg litre~1, wasCsgreater than those recorded for weakly sorbed com-pounds, indicating that our results cannot beextended over diþerent soils but are speciüc to ourexperimental conditions.

3.1.2 Pesticide transfer associated with erodedparticlesSuspended soil particle concentrations, Cp ,decreased with time (Fig 1), and were greater in theCoet Dan experiment than in the Jailliere soil,ranging respectively from 14.5 to 9g litre~1 andfrom 11 to 7g litre~1. This diþerence was attributedto the higher rainfall intensity and, thus, highersurface run-oþ ýow rate in the Coet Dan experi-ment, rather than to any diþerence in the propertiesof the soils.

Average adsorbed pesticide concentrations, onQs ,transported particles were measured for each run-oþsample (Fig 3). We distinguished between twoclasses of pesticides. The ürst was characterised bylow values of (\ 2lg g~1) which were practicallyQsconstant over time; these included atrazine, sima-zine, alachlor, diuron and isoproturon, with the ürstthree behaving very similarly (Fig 3B). The secondclass, including aclonifen and triýuralin, showedhigh values of from 10 to 34lg g~1, and largeQs ,variations with time, especially soon after initiationof run-oþ. These two classes also had diþeringamounts of pesticide transported on colloids, with

Figure 3. Variation with time of the concentration of ads orbed

pes ticide on s ediment particles . (A) Jailliere s oil diuron,(…) (>)is oproturon, aclonifen, atrazine and (B) Coet Dan s oil(=) (+)trifluralin, lindane, atrazine, s imazine,(È) (K) ()) (|) (L)

alachlor.

less than 10% of the total amount transported(atrazine 5.9%, simazine 8%, alachlor 5.8%, diuron9% isoproturon 2% and atrazine 3%) in the ürstclass and close to 90% in the second (triýuralin 89%and aclonifen 91%). Lindane showed an intermediatebehaviour with 37% transported on colloids.

3.1.3 Total pesticide transferThe total amount of herbicide found in run-oþ forthose compounds, and transported mostly in solutionwas correlated to the values. Isoproturon andKdatrazine were transported to a similar degree in theJailliere experiment and atrazine and alachlor in theCoet Dan experiment. Most sorbed or less-solublecompounds, such as diuron and simazine, were stilltransported in solution, but the loss was lower. Thisindicated that the loss was related to the pesticideconcentration of the soil solution. The lower solu-bility in water of simazine is a possible explanation ofsuch a diþerence, but more soil-speciüc adsorptiondata together with replicated simulations arerequired to conürm this assumption.

For compounds transported mostly in associationwith colloids, the partition coefficient was(Koc)insufficient to determine the amount of pesticide lostin run-oþ. Indeed, the amounts of diuron and aclon-ifen transported from Jailliere soil were very similar(equal to 1.5% of applied) even though the two mol-ecules are characterised by very diþerent values for

(Table 1). Similar eþects occurred on the CoetKocDan soil with triýuralin and alachlor.

3.2 Equilibrium reactivity of soil and eroded

particles

3.2.1 Characterisation of eroded particlesParticles eroded from the Jailliere soil were collectedin a single run-oþ sample by applying artiücial rainonto the bare soil surface (without ürst applyingpesticide) for one hour. Their size distribution wasclay 20, silt 78, sand 1.1 and coarse sand 0%. Noparticles with a size greater than 2mm were trans-ported in surface run-oþ waters, and only a smallfraction (1.1% w/w) had a diameter greater than50lm. In contrast, the silt and clay fractions in sedi-ments were increased by factors of 3.9 and 8.3respectively, compared with the parent soil.

Sediment was also enriched in total organic carbonby about 1.8 times for both soils. However, theenrichment progressively increased with time from2.4 to 3.2% w/w in the Jailliere experiment whilst itdecreased from 5.4 to 4.0 in the Coet Dan case. Theconcentration of dissolve organic carbon was small,representing only 1.5% of the organic carbon in thesediments. The pH of run-oþ water was close to 7,corresponding to the pH of the applied water.

3.2.2 Adsorption isothermsThe concentration of particles eroded from Jaillieresoil was large enough (7.2g litre~1) to allow adsorp-tion isotherm measurements. Isoproturon was

Pestic Sci 55 :175–182 (1999) 179

V Gouy et al

adsorbed less than diuron in soil (Fig 4A) asobserved elsewhere.16,17 The Freundlich modelleads to an exponent in soil close to unity, with thepartition coefficients derived from these linearKdisotherms being 6.4 and 1.6g litre~1 for diuron andisoproturon, respectively. In contrast, both mol-ecules were adsorbed to a similar extent on erodedparticles, for which the Freundlich exponent wasclose to 0.7 for both molecules (Fig 4A). As particleconcentration might inýuence adsorption results,18measurements were also performed for diuron on soilsuspensions diluted to 10g litre~1. They showedlittle change in which increased slightly to 8.3gKd ,litre~1. Thus, the discrepancy between soil and sedi-ments observed under conditions of equilibriumreýects a change in the surface reactivity of soil ascompared to eroded particles, which may be attrib-uted to sediment enrichment in üne particles andorganic matter.

3.3 Partition of pesticides during surface run-off

3.3.1 Diuron and isoproturonPartition of diuron and isoproturon between erodedparticles and the liquid phase was described by thevariation in the ratio of the concentration ofQschemical adsorbed to sediments to the aqueous con-centration, measured at each time step duringCs ,the Jailliere experiment (Fig 4B). These values werequite constant (around 1.4lg g~1 and 0.6lg g~1) fordiuron and isoproturon, respectively. Comparingthis observation with the isotherm curves obtainedfor sediments using the batch technique (dotted linesin Fig 4B), adsorption equilibrium of both moleculeson eroded particles was not achieved, particularly atthe beginning of run-oþ when values of wereCslargest. Thus, in the short duration of surface run-oþ, adsorption on eroded particles rather thandesorption into the surface run-oþ stream occurredfor both molecules.

3.3.2 Other moleculesThe concentrations are represented as a functionQsof for all compounds in Fig 5A for the JailliereCssoil and Fig 5B for the Coet Dan soil. Chemicals for

Figure 4. (A) Equilibrium ads orption is otherm curves for (…,L)diuron and is oproturon on Jailliere s oil and(=,K) (…,=) (L,on eroded particles generated by the rainfall s imulation. The=)

s olid and broken lines refer to the Freundlich model applied to

ads orption data for s oil and s ediments res pectively. (B)

Partitioning of diuron and is oproturon between the(LÉ ) (º )water and eroded particles during run-off.

Figure 5. Partitioning of pes ticides between the dis s olved phas e

and trans ported colloids for each time s tep of the rainfall

s imulation with (A) Jailliere s oil, diuron, is oproturon,(…) (>) (=)aclonifen, atrazine and (B) and Coet Dan s oil, trifluralin,(+) (È)

lindane, atrazine, s imazine, alachlor.(K) ()) (|) (L)

which transport essentially proceeded in solution aredistinguished by their rapid decrease in and theirCssmall and constant values of This occurred forQs .diuron and isoproturon on Jailliere soil, for simazineand alachlor on Coet Dan and for atrazine on bothsoils. Due to the low values of for this set ofQschemicals, it appears possible that equilibrium ofadsorption was not being achieved in the run-oþstream, just as in the case of diuron and isoproturon.Adsorption on eroded particles was limited by theshort duration of surface run-oþ, lasting only a fewminutes.

In contrast, triýuralin on Coet Dan soil and aclon-ifen on Jailliere soil were transported mostly on col-loids. The run-oþ of these compounds wascharacterised by large values of and small valuesQsof Figure 5B shows a rapid decrease in forCs . Qstriýuralin along with a slight variation of TheCs .temporal variations in measured for triýuralin andQsaclonifen are dependent on the change in the sorptiveproperties of the sediments during run-oþ. Thisconcentration was related to the organic contentQsmeasured in run-oþ samples. For each molecule,values were quite constant over the time of theexperiment (close to 5lg g~1 for aclonifen and4lg g~1 for triýuralin). Furthermore, these verysimilar values show the importance of the organiccontent in sediments for retention of aclonifen andtriýuralin.

4 DISCUSSION

Results obtained on Jailliere soil have indicated aninverse relationship between the amount of herbicidefound in run-oþ samples and the partition coefficient

for weakly adsorbed molecules of moderate waterKocsolubility (eg atrazine, isoproturon and diuron). Weare now interested in the role played in pesticidetransfer by the kinetics of the extraction process fromsoil to the run-oþ stream. As colloids make only aminor contribution to transport, it is expected thatthe run-oþ ýow rate will have a speciüc eþect oneach compound if their kinetics diþer. A linearrelationship was established between the concentra-tion and the inverse of the water ýow rateCs 1/Fwwith good regression coefficients for diuron, isopro-

180 Pestic Sci 55 :175–182 (1999)

Pesticide run-oþ during simulated rainfall

turon and atrazine on the Jailliere soil. An eþectiveconcentration was deduced by assuming that theCs@amount of adsorbed pesticide was primarily in solu-tion. A linear relationship was also establishedbetween and (see eqn (1) and Fig 6). TheCs@ 1/Fwregression coefficient values were [ 0.99 for diuron,isoproturon and atrazine on the Jailliere soil.

Cs@ \ a@

Fw

[ b@ (1)

Two important points must be made about this plot.Firstly, this linear relationship implies that there wasa continuous extraction process that supplied the soilwater with herbicides. Secondly, the slopes diþeredamongst compounds, due to a kinetic eþect. Twofundamental extraction processes must be involved:(i) initial dissolution of formulated chemicals, (ii)desorption from soil particles. We think that desorp-tion played the dominant role because none of thesecompounds was likely still to be in the formulatedstate.

The average mass of pesticide contained in themisample (i) collected during with an average ýow*tirate can be derived from eqn (1) by the productFwiof the average concentration and the sampleCs(i)volume V(i):

mi(Dt

i)\ C

s(i)V(i)\ a@ *t

i[ b@F

wi*t

i(2)

The ürst term in eqn (2) does not depend on theýow rate and is associated with the dilution of thesoil water. It represents the mass of pesticide thatcould be extracted if the concentration in the soilwater was maintained constant during the run-oþevent. The soil is considered as a reservoir thatwould allow the equilibrium concentration through-out the soil water to be achieved instantaneously. Incontrast, the second term is dependent on ýow rateand represents the efficiency of pesticide extraction.

Figure 6. Variation of the effective concentration of diuron,(…)is oproturon and atrazine vers us the invers e of the(>) (+)

average run-off flow-rate in the Jailliere s imulation. The effective

concentration was derived by as s uming that the whole mas s of

pes ticide was initially in the liquid phas e. Straight lines repres ent

linear regres s ions .

This term gives the mass that must be deductedbecause equilibrium between solids and soil water isnot instantaneous. This delaying eþect is most pro-nounced with the least-adsorbed chemicals. Indeed,this term indicates the capacity of the soil to main-tain a constant concentration. If the amount ofadsorbed chemical is large (and aqueous concentra-tion low), then the desorption of a small fraction ofpesticide is required to regain equilibrium concentra-tion following dilution. Such behaviour wasobserved in desorption studies carried out on undis-turbed soil under static conditions.19 Equilibriumwas reached quickly. It is expected that in this casethe b parameter would be small, as demonstrated fordiuron and isoproturon.

If surface run-oþ proceeds by successive steps ofduration and average ýow rate the total mass*ti Fwi ,M of pesticide collected is :

M \;i

a@ *ti[ ;

ib@V(i)\ a@T [ b@V (3)

where T is the duration of run-oþ and V the totalvolume of run-oþ sample. This quantity M rep-resents the mass of pesticide extracted from the soilinto the surface run-oþ stream. Only a proportion ofM will be transported over a longer distance.

5 CONCLUSION

In this paper we have used rainfall simulation experi-ments to study the mechanism of loss of pesticidesfrom soil into the surface run-oþ stream and the roleof adsorption–desorption phenomena in this dynamicsituation. Results have given us a better understand-ing of the behaviour of weakly sorbed chemicals.Desorption kinetics appeared to be the limiting stepin the surface run-oþ of isoproturon and atrazine.We propose a model that describes the mass of pesti-cide lost from soil by surface run-oþ. This modeldistinguishes between dilution of the soil solutionduring surface run-oþ and desorption of chemicalsfrom soil through two parameters, the dilution factorand the retardation extraction factor. These param-eters are speciüc to each soil/pesticide combination.Additional chemicals should be used to conürm suchresults and to identify any determining role of watersolubility. Further experiments should include repli-cated simulations together with a systematic com-parison of the sorptive properties of soil andsediments in order to study the inýuence of soilproperties on pesticide run-oþ.

ACKNOWLEDGEMENTS

This work received ünancial support from the AIP‘Ecopol’ directed by P. Chassin. The authors wouldlike to acknowledge C Garon-Boucher and B Lailletfrom the Water Quality Division of Cemagref Lyonfor their analysis work. They also thank the labor-

Pestic Sci 55 :175–182 (1999) 181

V Gouy et al

atories of Chemistry and Ecodynamic of Sedimentsof Cemagref Lyon for their help. Dr P Gaillardon isthanked for helpful discussions.

REFERENCES1 Patty L, Re� al B and Gril JJ , The use of grassed buþer strips to

remove pesticides, nitrates and soluble phosphorus com-pounds from runoþ water. Pestic Sci 49 :243–251 (1997).

2 Belamie R and Gouy V, Introduction des polluants dans lemilieu ýuvial. Inýuence du ruissellement des sols. Oceanis

18 :505–521 (1992).3 Carluer N Gouy V and Gril JN, Contamination des eaux de

surface par des produits phytosanitaires et mode� lisation.6 :19–30 (1996).Inge�nie� ries-EAT

4 Bailey GW, Swank AR Jr and Nicholson HP, Predicting pesti-cide runoþ from agricultural land: a conceptual model. J

Environ Qual 3 :95–102 (1974).5 Leonard RA, Movement of pesticides into surface waters, in

Pesticides in the Soil Environment : Processes, Impacts, and

Modeling, Ed by Cheng HH, Soil Science Society ofAmerica, Inc, Madison, Wisconsin, USA, pp 303–349(1990).

6 Karishoþ SW, Semi-empirical estimation of sorption ofhydrophobic pollutants on natural sediments and soils. Che-

mosphere 10 :833–846 (1981).7 Gouy V and Belamie R, Using the CREAMS pesticides trans-

fer sub-model at a rainfall simulation scale. Wat Sci Tech

28 :679–683 (1993).8 Knisel WG, CREAMS, a üeld scale model for chemicals,

runoþ and erosion from agricultural management systems,Ed by Knisel WT, USDA. Conservation Research Report, 26

(1980) 643 pp.9 Wauchope, RD., The pesticide content of surface water drain-

age from agricultural üelds : A review. J Environ. Qual.,7 :459–472 (1978).

10 Munoz JF, Me� thodologie d’etude des produits phytosanitaires.Etude d’un bassin versant viticole : l’Ardieres (Beaujolais).Mise au point de me� thodes analytiques de pesticides. These

de cycle, Universite� Claude Bernard, Lyon, Francetroisieme

(1992) 175 pp.11 Foster GR, Young RA and Neibling WH, Sediment composi-

tion for nonpoint source pollution analyses. Transactions of

the ASAE 133–146 (1985).12 Walter MF, Stennhuis TS and Haith DA, Nonpoint source

pollution control by soil and water practices. Transactions of

the ASAE 77 :834–840 (1979).13 Fawcett RS, Christensen BR and Tierney DP, The impact of

conservation tillage on pesticide runoþ into surface water : areview and analysis. J Soil and Water Conservation 126–135(1994).

14 Patty L, Limitation du transfert par ruissellement vers les eauxsuperücielles de deux herbicides (isoproturon etdiýufe� nicanil). Me� thodologie analytique et e� tude del’efficacite� de bandes enherbe� es. de cycle, Uni-These troisieme

versite� Joseph Fourier, Grenoble, France 217 pp (1997).15 Gouy V, Contribution de la mode� lisation a la simulation du

transfert des produits phytosanitaires de la parcelle agricolevers les eaux superücielles. de cycle, Uni-These troisieme

versite� Louis Pasteur, Strasbourg, France 315 pp (1993).16 Gaillardon P and Dur JC, Inýuence of soil moisture on short-

term adsorption of diuron and isoproturon by soil. Pestic Sci

45 :297–303 (1995).17 Gaillardon P, Inýuence of soil moisture on long-term sorption

of diuron and isoproturon by soil. Pestic Sci 47 :347–354(1996).

18 Jamet P and Roche D, Inýuence of soil–water ratio onadsorption–desorption kinetics of isoxaben in soil, in Fate

and prediction of environmental chemicals in soils, plants and

aquatic systems, Ed by Mansour M, Lewis Publishers, USApp 221–234 (1993).

19 Gaillardon P., Desorption of diuron and isoproturon fromundispersed clay loam soil., Pestic Sci (1998). in press.

182 Pestic Sci 55 :175–182 (1999)