You are on page 1of 25

Pergamon Chemqhwe,Vol. 36, No. 10,pp. 2225-2249.

1998
Q 1998ElsevierScienceLtd
All rightsreserved.printedin GreatBritain
0045-6535198 $19.00+0.00
PII: !soo45-6535(97)10194-1

AN INDICATOR OF PESTICIDE ENVIRONMENTAL IMPACT BASED ON


A FUZZY EXPERT SYSTEM

Hayo M.G. van der We& and Christophe Zimmer

INI& Station d’Agronnmie, BP 507,68021 Colmar, France


*) Current caddress:IN&& Unit6 d’Agronomie de Fhme5Quimpeq ENSAR - 65, rue de Saint Brieuc,
F 35042 Rennes, France
Tel. +33 2 99 28 27 09, fax +33 2 99 28 72 30. Email : Hayo.vanderWert@roazhon.inra.fi

@ceived in Germany20 August1997:accepted28 October1997)

ABSTRACT
Pesticide use options available to fkrmers difkr strongly with respect to the risks they pose to the
environment. This paper proposes a fky expert system to calculate an indicator “Ipest” which reflects an
expert perception of the potential environmental impact of the application of a pesticii in a field crop. We
defined four modules, one reflecting the presence (rate of application) of the pesticide, the other three
~~theriskfortlnee~jor~cOmpartments(groundwater,surfacewater,air).Theinpwt
variabks for these modules are pesticii properties, site-specific conditions and characteristics of the
pesticide application. For each input variable two functions descrii membership to the fkzy subsets
Favourable (F) and UnIiwourable (U) have been defjned. These fimctions are based on criteria drawn from
the krature or on the authors’ expert judgement. The expert system calculates the value of modules
accordingtotbedegreeofmembershipoftheinputvariablestothefuzzysubsetsFaedUandaccordingto
sets of decision rules. The four modules can be considered individually or can bc aggregated (again
accordingtomembershiptofwzysubsetsFandUandasetofdecisionrules)intotheindicatorIpest.The
systemis~~~aodcanbetunedtoe~perception,itcanbeusedasadecisionaidtooltorankor
choose between alternative pesticii application options with respect to their potential environmental
impact. Results of a sensitivity am&ii and module and Ipest scores for some pesticide application cases are
presented. An agro-ecological indicator IPEST, based on the expext system, is proposed as a tool to asws
the em&o& impact of all pesticide applications related to a crop within a year. The practical

implementation of the expert system and its validation arc discussed.


621998 Elsevier Science Ltd. All rights reserved

Key WOKIS:de&ion support, enviromnental impact, expert system, fuxzy logic, i&&or, leaching,
pesticideq ruuoff, toxicity, volatilisation
2225
2226

1. INTRODUCTION

Pesticides are xenobiotic substances which are used in crop production for the control of pests, diseases
and weeds. As field crops are grown outdoors, the application of pesticides to geld crops by defimtion
implies emission to the environmental co- airandsoiLHowever,thereareimnxnsedS&ncesin
the degree to which pesticides are mobile and biologically active in the environment. Hence, the pesticide use
options available to thrmers di&r strongly with respect to the risks they pose to the environment.
Commercial products for the control of pests, diseases or weeds general@ contain active ing&ents,
adjuvants and inert iqmdkmts. The term “pestkGde” is often used as a synonym for “active @edienV e.g.
[ 11, but it is also used as a synonym for “commercial product for the control of pests, diseases or weeda” e.g.
[2]. This is contusing, in this paper we use “pesticide” as a synonym for “active &edienV only. Adjuvants
used in the formulation of the commercial product can change its agronomic effects (e&ctiveness,
phytotoxicity) as well as its environmental impact, as dispersion patterns may be altered and the functional
activity period of the active ingredient may be le@ened or its degradation may be delayed [3].
Unfortunately, very little information on the effects and fate of adjuvants is available in the scientigc
literature and therefore the role of adjuvants will not be taken into account in this paper.
In general, the anticipated effectiveness against the pest, the risk of phytotoxicity to the crop and the
cost of the application are the main factors considered by a thrmer in choosing a particular pesticide. An
increasing number of farmers and other decision-makers would like to be able to take the potential
en~~~oftheapplicationofapesticide~oaccount.Withinthe~twodecadesanumberof
methods have been proposed for estimating this impact (reviewed in [3-51). These methods assess potential
pesticide environmental impact, as perceived by the “experts” that created them.
G&din et al. [ 61 propose a set of “agro-ecological indicators” to evaluate the environmental impact of
farming systems. The term “indicator” has been defined as: a variable which supplies information on other
variables which are difhcult to access [7]. Indicators are valuable tools for evaluation and de&ion making as
they synthesise information and can thus help to tmderstand a complex system [8]. This paper describes au
expert system which is at the core of the agro-ecological indicator reflecting pesticide environmental impact.
The environmental impact of the application of a pesticide in a field crop will depend on characteristics
of: a) the pesticide (e.g. its toxic&y to water organisms), b) the local environment (e.g. soil type), c) the
application (e.g. application in the soil, on the soil or on the crop) [S]. If one accepts the premise that the
“environtmmtal impact’ of a pesticide results from a combination of exposure and toxicity [9,10], obviously
the use of a simulation model to estimate the exposure component is tempting [11,12]. Using pesticii,
environmental and application characteristics as input variables a &nulatiin model can yiekl predicted
environmental concentrations (PEC), which can be related to predicted no-eflbct concent&ons or
maximum admissible concentrations. This approach to estimating pesticide environmental impact is
attractive: its results (at least regarding PEC) can be validated and it provides an elegant way to combine the
2221

threetypesofinput~~which~~becoasidered.However,tbereare~~~tothis
approach.In~~place,noneofthe~lyavailablemodels~pestichie~~e
simuhamuslyrepresentsthemajorrelevautprocesses: behaviour in soiL volatilisation, drift, leachiug,
runoff and degmMotl[13,14]. Secondly, the valid&m status of pesticide fate models genera@ is low [IS-
171. Finally, these models gemrally requite a large amount of data for their pammeMs and input variables
(pesticide properues, soil characteristics, meteorological data). For most pesticide application situations a
major part of these data are unavailable. When available, values otten are impe&ct: they may be imprecise
(e.g. pesticii field halGlilb), or tmmtain (e.g. risk of the occurrence of sutlhce water runoff).
These practical problems probably are the reason why. a review of six methods to assess pesticide
environnxntal impact [S] found only one method using a simulation modeL The other nmthods are simpler,
semiquantitative indexing approaches. They identify variables related to pesticide exposme and toxicity,
which, in three out of iive methods, are aggregated into one or several indicators. The input variables are of
dissimilar type, dimemdons and rauge of possible values (e.g. fiekl half-life, toxicity to aquatic orgamsms,
month of application). Generally, values of input variables are transformed into ratings, which are then
condensed into one or more indicators by a weighted means approach sometimes these calculations also
involve multiplication. These modes of aggregation pose mathematical and concept& problems [3,18].
Considering the inadequacy of the available simulation models aud the shortcomings of current index&
approaches tbr the assessment of pesticide environmental impact, we decided to fillow a d&rent avenue by
setting up a fuzzy expert system [ 191. This technique is robust when uncertain and imprecise data is used and
allows the aggregation of d&in&r input variables in a consistent and reproducible way [20]. The expert
system descrii here regects an expert perception of the potential environmental impact of the application
of a pesticide in a geld crop.

2. THE STRUCTURE OF THE INDICATOR

We consider that pesticide impact on the environment depends on: a) the presence ofacertainamount
of pesticiie, b) the extent to which the pesticide leaves the field on which it is applied by dritl, volatiiin,
runoff or leaching, c) the toxicity of the pesticide. This implies that we do not take into account the soil, or
the organisms in the soil or on the crop. We do consider these to be part of the natural environment, but
there is insu&ient data regarding pesticide impact on soil and terrestrial non-target organisms.
We defined four indicator modules. The module Presence reflects the rate of application of the
pesticide, the modules Risk of slsface water contamination (Rsur), Risk of groundwater contamination
(Rgro) ad Risk of air contamination @air) reflect the risk for three major environmental compartments.
ForeachmoduleavalueonadimensionlessscalebetweenO(noriskofenvironmentalimpact)and1
(msximum risk of environmental impact) is calculated. The value of the module Presence depends on a
2228

Table 1. The imlicator modules Presence, Risk of surjiie water contambtibn (Rsur), Risk of
groundwater contamination (Rgro) and Risk of air contamination @air), and tbir hput va&les.

Inpmt vadabka P- Rmr l&o Rair

GUS X
volatihty X
Aquatic toxicity X
Hunnmtoxici~ X X
v ** ns
DrBpememage X
RunOffrisk X
Leachingrisk X

Rateofapp~n X
Position of appkatkn X X X

single input varbble (Rate of application). The tie of the other three modules depends on four or five
input variables (Table 1) and a set of decision rules. Vahies are calculated according to a procedure which
willbee~lainedindetaiiinthenextsectionofthispaper. Wedi&gukhthreetypesofinputvarkbhxa)
pesticide properties, b) site-specific conditions, c) characteristics of the pesticide application (Table 1). In
selecting these input variables we tirst considered their relevance for impact -ntandsecondlydata
availability, as there is not much point in taking into account input variables for which data are unavailable or
hard to come by for many pesticide application situations.
Thefouriodicatormodulescanbeconsidered~orcanbeaggregatedintoanoverallindicator
estimating the total potential environmental impact of a pesticide application, again on a 0 to 1 scale. This
modular skucture presents several advantages. In the first place users have access both to an indicator
retkxting overall impact, and to each of the modules. Secondly, the mode of aggregation of modules can be
changed and new modules (e.g. risk for soil, impact on beneficial arthropods) can be added, as availability of
data and w of pesticide impact evolves, or dependins on the demands of the users.

3. FUZZY EXPERT SYSTEMS

A kzy expert system was used to aggregate input variables into indicator modules, and to
subsequently aggregate the modules. Fuzzy logic is a superset of conventional (Boolean) logic that has been
extended to handle the concept of partial truth: truth vahies between kompletely true” and “completely
false”. It was proposed by Zadeh [21] to deal with the unc&a@ of natural language. Fuzzy set theory can
be used to cope with vaguely defined classes or categories.
Lnclsssicalsettheory,anelementeaherisorisnotinaset.Forexample,ifsYbsetAconsistsofthe
pesticides with a maximum field h&-life of 20 days, a particular pesticide can be classitied as a member or
2229

IA_/
not a mmber of the subset. If,
degree of
membership however,AisdetInedtobethe
subset of “non-pemktem
peHkkk&thenitismoredifhcuh
todeterm&ifaspeciScpestkhkis
inthesubset.Ifonedecidesthat
onlypestkiiwithamazimum
6ek.Ihahxtbof20daysareinthe
subset,thenapesti&ewitha21
Groundwater ubiquity score day halfxfe cannot be ckssii 85
non-per&tent even though it is
1 B - fuzzy sets 1 “ahnost” non-persistent. The use of
degree
of
memhetship
fuzzy set theory is particularly
compellingbecauseavailabkvahtes
tixtIeldhaHitbaudseveralother
relevant variabks are imprecise
s&or tmc&ain Thus ckmification
based on the conventional
approach where the tmnsitkn
2.8
between classes is abrupt, is
Groundwater ubiquity score
doubtful.
Figurel.Graphicalreprese~knofcrisp(A)amlfUzy~)sets.
FtlZZysettheoryaddnXXSthk
typeofprobkmbyallowingonetodeSnethe”degreeofmembership”ofanelementinasetbymeansofa
Illcmbership t&tion. For classkal or “crisp” sets, the member@ tbnctiin only takes two values: 0 (non-
membersbip)andl(membership).InfUzzysetsthemembershipfunctioncantakeany~~mthe
Merval [O,l]. The vahre 0 mpemnts complete non-men&e&@, the value 1 mpesents complem
membe&p,andvaluesinbetweenareusedtorepresentpartialmembership.
For all input variabks given in Table 1 we de&red two &zzy subsets F (Favourabk) and U
(Unt%vourabk). We based the membership &n&ions on data available in the literature or on our own
“expert” knowkdge. For exampk, GMafIron’s [22] Groundwater Ubiquity Score (GUS) bas three cksses
for pesticide kachabihty (Fig. la). We gave pesticides class&d by Gust&on as “kacher” (i.e. GUS > 2.8)
amembe&@vakeof1forthefbzzysubsetUandamember@val~ofOtbrthe&zysubsetF.
Pesticidesc~as”~~her”(GUS<1.8)ategivenamembershipvalueofOforthe~subsetU
andamemben@valueof1forthefhzzysubsetF.Theclassof%orderlinecompounds”(1.8~GUS~2.8)
fallswithina”tmnsitkninterval”inwhichthemembe&@valuetbrFdecreasest%om1(atGUS=1.8)toO
2230

(at GUS = 24, and the membership value for U &reases5omOto1(thusthe5mctionscharacterisingF


andUarecomplen~ntary,Fig. lb).Accordingtothisapproachwecancham&&etheshapeofthe
membershipfimctionofeach~utvariablebythetwolimasofthe~~ninterval”.Weused
membershipfunctionsthateresinusshapedinthe~n~asthey~vide~~varietionsof
the0utputval~thanmemben&5nlcti0nsthatarelinearinthetra&ionhnervaL
Foreachmodulewetbrmulatedasetofdecisionrulesattt%&gvaluesbetweenOand1toanoutput
variable according to the membe&p of its input variables to the 5uzy subsets F and U. Sugeno’s inference
method [23] was used to compute the modules as well as the indicator. Its princiile is brie5y described in
the following. Sugeno’s i&rence method uses decision rules of the fbrm :
Ifx,isA,,andxzisA,zThenyisB,
Ifx,is&,sndxrisAzzThenyisBz
where xj (i = 1,2) is an input variable (e.g. field haKlife, or volatilisation risk), y is an output variable (e.g.
the. value of the module), Aj a fuzzy subset (Favourable or Unthvourable), and Bi a number called
conclusion of the rule. “q is &j” is called a premise of the i-th rule.
Letx~oandqobethe~takenbyx~and~2,and~(xjo)themenobershipvalueOf~o4oto~fuzzy~j
(givenbythemembershipfUnctionthatde~sAij).Then,onecan~~w~aodw2,tbe~valuesoftbe
rule& as follows: WI = min@,,(& Mxz”))
W2 = min(A31(x*“), A220(1O))

WhereminmeanS”minimum value of’. The first rule infers w,Bi, the second one wz&, and the gkibal
output~ishdbrredby: y” = (w~BI + w2&)4w + ~2)

To i&&rate Sugeno’s infemnce method sn example of a calculation will be given Let’s assume (only to
illustrate the approach and for simplicity’s sake) that the output wuiable Emdronmenfulimpact of the
application of a pesticide depends on two input variables only: Rate of apphation andpesticideField hq”
life. For both input variables membership to fuzzy subsets F (Favourable) and U (Un5woumble) baato be
detiued. Let’s assume that experts tell us that a low Rate of apphcutionad a short Field ha&l@ are
favourable, whet& a high Rate of applicafion and a long Field ha@X# are u~.&voumbIe. To de&e the
shape of the men&e&@ 5mctions we have to be specific: for Rafe of applicationwe ass&u conrplete
membershipto the thzzy subset F if Rate of application -Z 0.051 kg ha“ and complete membership to the
fiuq subset U if Rate of application > 2 kg ti’; for Field hul$&fe we assign umpkte mmbemb@ to F if
Field half-life < 1 day and complete member&p to U if Field ha(f3fe > 120 days (Fig. 2). Consequently,
ratesbetween0.001aml2kgha~‘aml~~halflivesbetweenland120days~withina~~n
interval” as explained above.
Inthisexantple,whichhastwo~utvariablesandtwofirzzy~~for~~~variable,four
situationsmayoccur,as~~t~~bythedecisionruleswhichare ‘tiinTable2.Thesedecision
rules re5ect expert knowledge and/or expert judgement; they read as (e.g. 5rst line of the table): “If Rareof
2231
1

0.9

* 0.8

$ 0.7

2 0.6
E
E 0.5
6
P 0.4

i 0.3

o 0.2

0.1

Field half-life (days)

0.9
0,853
0.8 -
.P
f 0.7 -
n” 0.6 -
E
E 0.5 -
B
8 0.4 -
5 0.3 -
n 0.2 -,

0 0.5 1 1.5 2 2.5

Rate of application (kg/ha)

Fii 2. Member&p to the fkzy sets Favourable and Unkourabk for atmzine (field ha&%% 60 days)
applied at 1.5 kg ha“ as a fimction of Field ha&if (top graph) and Rate of application (bottom graph).

application is hourable and if Field half-l@ is favourable then Environmental impact is 0.“. The decision
rules consist of two premises (if....) linkedby and, followed by a conclusion (then ....). As can be seen thorn
Table 2, h both iqwt variables are F, the value of Environmental impact is 0.0 (no risk of environmental
impact), when both input variables are U the value of Environmental impuct is 1.0 (maximum risk of
environmental impact) and when one input variable is F and the other is U the value of Environmental
impact is 0.5.
Having thus defined the membership fimctions and kmnulated the decision rules, we proceed to
cakulate the value of the output variable Environmental impact for a pesticide appkatiin. Let’s assume that

Table 2. Sumuwy of decision rules descrii the effect of the input variables Rate aud Field half--life on the
hypotktical mod& Em+onmenful impact. F = favourable, U = unfkourabk. For details, see text.

Rate Field half-life Envinwnental impact


F F 0.0
F U 0.5
U F 0.5
U U 1.0
2232

Table 3. Summary of de&on rules describing the e%bet of the input v&ables Rote and Field half-l@ onthe
hypothetkal module Environmental impact. Truth values of pmmiaes and co&tsions for an epptication of
1.5 kg ha-’of atmzine (FieZd half-life 60 days) are given in bra&eta F = fWourabIe, U = unfavoumble. For
details, see text.

Rate Field haWife Enviroamdai impact


F (0.147) F (0.506) 0.0 (0.147)
F (0.147) u (0.494) 0.5 (0.147)
U (0.853) F (0.506) 0.5 (0.506)
U (0.853) u (0.494) 1.o (0.494)

theher%ideatmzineisappliedat 1.5kg~‘ofactiveingredientaradthattsfiekl~~is60days.The
membershipfUnctions~~aboveallowthecalculationofthetruthvalueoftbepmr6ses,ie.thedegree
of member@ to the f&y subset coneemed (Favourable or Unfavourable) for each input variabk (Fig. 2).
Aoeording to Sugeno’s [23] infemmx method, when the premises arelinkedbymd,thetruthvalueofa
decisionrulecanbedefinedssthesmallestofthetruthvaluesofitspremises(Table3).Thevahteof
Environmental impacf is caloulated as the average of the con&Guts of the de&ion rules, weigh&d by their
tl-llthvalue:
E&romer& impact = /0.0*0.147 + 0.5*0.147 + 0.5*0.506 + 1.0*0.494) = 0.634
(0.147 + 0.147 + 0.506 + 0.494)
Thismethod6tsourreq-, sineeitalIowstobuiklamoduktr Wuoture for our indieator,
providing usable vahtes for the indicator modules. Moreover, the rules are easy to read, and the numer&l
scoresusedfortheirconclusionsareeasytotunetobeinaccordwiththeopinionofthe~.

4. PRESENTATION OF TIIE INDICATOR MODULES

4.1 The module Presence


The indkator module Presence refIeots qusntity of active ingredient applied, it depends on a singIe
input variable: Rate of upplication. A single appl&ion of a pesticide may range f%oma few g ha” of active
&edient for low-dose herbicides to several hundmds of kg tit tbr a soil Watment against nem&des. A
major part of this enormous range of application rates is due to di&mxes in the biological activity of the
active~~sused,andthusalowerwteofapplicationdoesnotnecessarily~,lylessrigkforthe
environment. On the other hand, improved application te&ttiqueS (better targeting) allow lower appheation
l-atesandnXttiyglrml%s~ successful in using application rates below those recomtmmded on thr label by
matchingtimingofapplicationvvithwlnerablepgiodsofthe~o~orby e
withothermwmsofcontrol.Manyexpertsjudgethat,geaeroldy~lowratesofapplieation~
desirable from an environmental point of view [ 11,24-281.
For the input variable Rate of application we did not find htemtum data to deiine the values dehmit&
the transition imerval within which the fuzzy subsets F (Favourable) and U (Untavourable) are
2233

co~~.Asthetramition~should~~awiderangeofvaluesandasobviously
environmentalimpactwiIlnot increese~~withapplicationratewe~t8einput~Rateof
clppJcafi00 as the log,, of the application rate in g ha-‘. Wearbhmrilydecidedtoaasigncomplete
membenhip to F ifRate of upplicution < 1 (10 g ti’) and complete membership to U if Rate of upplicution
> 4 (10,000 g ha-‘).
The vahre of the module Presence depends on the input variable Rate of upphztion according to two
de&ion rules: 1) If the Rate of upplicution is F then Presence is 0.0 (minimal risk, of environmental impact);
2) If the Rate of upplicution is U then Presence is 1.O (maximum risk of environmental impact).

4.2 The module Risk of growhater contamination


The indicator module Rgro (Risk of groundwater contamination) reflecta the potential of a pesticide to
reachgroundwaterthroughleachingandtoaflFectitspotentialuseasasourceofdrinkingwaterforhumans.
ThevahreofRgrodependsontburinputvariab~ 1)pesticideleachingpotemial,2)thepositionof
application of the pesticide (on the crop, on the so4 in the soil), 3) soil leaching risk and 4) the toxicity of
the pesticide to humans.
A number of .authors have proposed indices based on pesticide properties to class@ pesticides
according to their groundwater comamination or leaching potential [22,29-321. We use the groundwamr
ubii score (GUS) as proposed by Gustalhon [17] to quantity the leachiug potential of a pesticide. This
index is simple and can etktively discriminate between pesticiies that leach and pesticides that do not.
GUS is a function of the pesticii characteristics Field half-life (DT50) and Grganic Carbon sorption
constant f&X): GUS = log#TSO) x (4 - log&oc)).
Pesticides detected in groundwater generally have GUS values exceeding 2.8, whereas compounds with
GUS values below 1.8 were not detected in groundwater. We therefore define the limits of the transition
interval within which the fuzzy subsets F (Favourable) and U (Unfkourable) are complementary by
assigning complete membership to F if GUS < 1.8 and complete membership to U if GUS > 2.8.
The position of application of the pesticide strongly affects its leaching potential Dependins on crop
cover, pesticides may be distriited between the crop and the soil slnlhce. Although some of the pesticide
present on the crop may bc washed off and reach the soil [33], we consider application on the crop to be
&vourable. Apart from reducing the amount of pesticide on the so& crop cover in itself reducesleaching
risk, as transpiration by the crop reduces drainage. We therefore define the input variable Position of
application, as a value on scale between 1 (100% of the pesticide on the crop) and 0 (0% on the crop,
pesticide applied on or in the soil). Its value is calculated as: Position of upplicution, = % soil covered by
the crop/lOO. The % soil covered by the crop is to be estimated by the user of the system. We define the
limits of the transition interval by assigning complete membership to F if Position of upplicution, = 1 snd
complete membership to U if Position of upplicution, = 0.
2234

Theridtofpe&iciklossiiomaiieklto grom&mM(~~k)~on~ofthe
soiZofthellllsatunrtedzone~vethewatertablearadoftbe~~~.~Qletbodshavebeen
proposed to assess this risk as a fimction of soil characteristics only (e.g. [34]), or of overall hydrogeologic
settings in&ding soil characteristics (e.g. [35,36]). Which approach is nxxt appropriate will dep& on
local hydrogeologic settings and on the availability of data We express Leaching risk on a scale between 0
(minor leach& risk) and 1 (major leaching risk). The value fix Leaching risk CXUIbe o&&ml by
tram&m&g a score obtained l%omone of the methods cited above or fiorn another appropriate method to a
Oto 1 scale.Ifthisisnotpo~b~,apreliminaryesthnationofleachiagriskcanbebasedonsoilorganic
matter content, which can be considered as the single most important soil characteristic afkting pesticide
leaching (e.g. [371). We deiine the transitiin interval by akgning complete membership to F if Leaching
risk = 0 ad compie$e mmbedip to U if Leaching risk = 1.
Asweconsiderthegto~~~tobeapotetrtialsourceofhumandrinkiogwater,weselectedthe
Acceptable Daily Intake (ADI), which reflects chronic toxicity to humans, as our criterion Ear toxicity. AD1
is de&ed as [38]: “the daily dosage of a chemical which, during an entire lifetime, appears to be w&out
appreciable risk on the basis of all the facts known at the time”. AD1 is expmssed as mg of chemical residue
in food per kg of body weight (mg b’day-‘). Jouany and Dab&e [25] class@ pes&des into 5 classes
according to their ADI: A) <o.OOOlmg kg-Id-‘, B) 0.0001 - 0.001 mg kg+‘, C) 0.001 - 0.01 mg kg-k’, D)
0.01 - 0.1 mg kg%‘, E) > 0.1 mg u’b’. Drawiq from their cksitkation we we 0.01 mg kg-‘&’ as the
n&an value for the tran&ion interval. We deiine the input variable Humon toxicify as the log10of the ADI.

Table 4. Summaxy of decision rules describing the efkt of the input variables GUS, Position of
application, Leuching risk and Humun toxicity on the indicator module Rgro (Risk of grotmdwnter
co&mktion). F = favourable, U = untitvourable. For details, see text.

GUS Poaitioa of Leackimg Htmmn GtWlEdW~ter


application risk toxicity risk
F F F F 0.0
F F F U 0.0
F F U F 0.0
F F U U 0.0
F U F F 0.0
F U F U 0.0
F U U F 0.0
F U U U 0.0
U F F F 0.1
U F F U 0.1
U F U F 0.1
U F U U 0.1
U U F F 0.2
U U F U 0.6
U U U F 0.6
U U U U 1.0
2235

The value of the module Rgro depends on the input miables GUS, Position of appZicafion, Leaching
riskandHvmmrtmicityaccordingtoasetof16decisionrulessunnnarisedinTable4.Theaedecisionrules
refkctour”~“~noftheriskof~~~~nasa~ofapesticide
~~~~TheexpertTeasoningissunnnarisedinFig.3,whichreadsasfollows:ifGUSisF(mn-shaded
box), the vahmeof Z&o is 0.0 (no risk). If GUS is U (shaded box) and Position of application is F, Rgro is
0.1. If GUS is U and Position of application is U, the value of Rgro depends on Leaching risk and,
subsequently, on Human toxici@. If Leaching risk is F, the value of Rgro is 0.2 or 0.6 (depend@ on Human
toxicily), ifLeaching risk is U, the value of Rgro is 0.6 or 1.0.

4.3 Tht module Risk of surjbce water contamination


TheiodicatormoduleRna~ofsurfacewater~~n)reflectstbepotentialofapesticideto
reachrnnface~~throughrunoffordriftandtoharmaquaticor~.Its~depeedson~einput
muiabbx 1)therunoffriskofthefieldsite,2)thedriflpemntageoftheapplieatio~3)thepositionof
application of the pest&de (on the crop, on the so& in the soil), 4) the field half-life of the pest&de and 5)
tbetolricityofthepesticideto~~organisms.Wedidnottakeintoaccountotberpesticide
characteristis such as sohbility or sorption properties as these have been shown not to cons&e&y aft&t

GUS

,...................................................
0.0 0.1 0.2 0.6 0.6 1.0 i
. . . . . . . . . . . . . . . . . . . . ...*.*. . . . . . . . . . . . . . . . . . . . . . . . . . . . . ...?
Risk of groundwater contamination

Fii 3. The eiExt of the iuput variables GUS, Position of application, Leaching risk ad Himnan toxicity
on the value of the comhisions of the decision rules for the indicator module &i-o (Risk of groundmter
contamination) according to their membership to the fimy sets Favourable (non-shaded boxes) and
Un&oucable (shaded boxes). For details see text.
2236

the pesticide’slwloff potential [33,39].

The risk of pesticii transport &om a field to surface water by runoff (R~~risk) depends on many
&tots, e.g. slope s&pness, slope let& soil texture, surthce condition soil particle aggregation and
smbihty, crop cover, dktance to surface water [33,40]. &wet-al methods have been proposed to assess this
risk as a fktion of soil, slope and/or watershed characteristics (e.g. [34,36,41,42]). Which approach is
mostappropriatewi&inthesamemanner as for Leaching ri& depend on the local &ration and on
aw&bihty of dam We express Runoflrisk on a scale between 0 (no runoff risk) and 1 (major runoff risk).
ThevalueforRunoffriskcanbeobtainedbytransformingascoreobtaimdfiromo~of~~cittd
above or Corn another appropriate method to a 0 to 1 scale. If this is not possible, Runofl risk can be.
estimated by the user. We deke the transition interval by assigning complete menkmh@ to F if Runoffrisk
= 0 aud complete membership to U if Runoff risk = 1.
The risk of pesticide transport from a field to surface water by driB depends on many factors, e.g.
dktance to surface water, application technique, crop structure and crop cover and wind speed [ 10,431. We
quantify the drifl potential of a Iield site by its Drifr percentage, de&ted as the pesticide rate reach&g
surf& water expressed as a percentage of the intended rate of field application [43]. The value of Drif)
percentage depends on application technique and distance to surtke water ([43], Table 5). We did not find
literature data to de&. the values delimiting the transition interval for Lkzji percentuge. We therefore
arbitrarily decided to assign complete membership to F if Drip? percentage = O?hand compb& lloembership
to U ifllriiftpercentuge > 1%.
The position of application of the pesticide strongly afkcts its runoff potential. Application on the soil
surface is unliwourable as pesticide concemrations in surface runoff have been shown to be strongly
wrrelated with concentrations in the surface 10 mm of soil [44]. Consequently, application in the soil (that
is, either on the seed or incorporated iu the soil) is favourable. Depend& on crop cover, pesticides may lx
distributedbetweenthecropandthesoilsurface.~~ughsoau:oft~pesticide~onthecropmdybe
washed off and reach the soil [33], we consider application on the crop to be favourable. Apart Iiom
reducing the amount of pesticide on the soil surface, crop cover itself reduces runoff risk. We therefore
defined the input variable Position of application ,asavalueonscalebetweenl (lOO%ofthepesticideon
the soil) and 0 (0% on the so& pesticide applied on the crop or in the soil). Its vahte is cakulated as:
Position of application,, = ( 100 - % soil covered by the crop)/1 00, unless the pesticide is applied in the soil

Table 5. Estimation of drift percentage according to application technique and distance to surIhce water
(adapted Iiom [43]).

Application method % drift at distance to water :


O-2m 2-10 m
Row spraying, knapsack sprayiag 0.5 0.2
Fulltieldspraying,cropheightS25cm 1.0 0.3
Full field spraying, crop height > 25 cm 2.0 0.5
2237

in that case Position of applic4hnm = O.The%soilcoveredbythecropistobe~bytheuserof


thesystem. We~thelimitsofthetransitionintervalbyassigningcorapletemmbe4shiptoFifPosition
of applicath~ = 0 and complete member@ to U ifPosition of applicationw = 1.
Field h&l@? (DT50) a&cts a pest&k’s runoff potential, persistent pe&ides that remain at the soil
surf&x fix longer periods of time have a higher probabibty of runoff than non-per&tent pesticides [33].
Withrespecttotheriskofsurfacewaterco~~nJolumvand~[25]considerapesticidehaving
a DT50 < 8 days as “not untkvourable”, one having a DT50 between 8 and 30 days as “modemtely
unthvourable” and om having a DT50 > 30 days as “unkvourable”. Theii proposition seems reasonable, as
ithasbeenshownthatpesticide~vaioccursmainlyduringthefirstortbefirst~wnmoffeventsafter
application [33]. Thus, drawiug tiom [25], we defined the limits of the transition iuterval by ass@ing
complete membership to F if Field half-life < 1 day and complete membership to U if Field half-lljiz > 30

The input variable Aquatic toxicity is bas4 on biological effects on three aquatic species forming a food
chain: algae (EC50), cmsweam (ECSO) awl fish (LC50), all cxpremd in mg 1-l. EC50 and LC50 beiug the
concentration of the pesticide that would produce a specifk effect (EC50) or death (LCSO) in 50% of a
large population of a test species. In each case the concentration for the most se&ii species tested was
retakl,thusrelatingthe asesment of the risk to the ecosystem to the most sensitk organism [45].
L~etaL[46]considerapesticidewfiichhasanEC50orLC50<1mgT1as~to~,o~withan
EC50 or LC50 between 1 and 10 mg r’ as moderately toxic, one with au EC50 or LC50 between 10 and

Runoff risk Drift?h

0.0 0.1 0.2 0.6 0.6 1.0 :


. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..~
Risk of surface water contamination

Figure 4. Summary of decision rules. The effbct of the input wuiabbzs Runoff risk, Drift percentage,
Pas&m of application, Field haFlife and Aquatic toxicity on the value of the cor&sions of the decision
rulw160rtbeiodicatormoduleRna~ofsurfacewater~~n)acoordiagtotheirmembershipto
the &zy sets Favourable (non-shaded boxes) and Unfavourabk (shaded boxes). For details see text.
2238

100 mg I’ as shghtly toxic and one having an EC50 or LC50 above 100 mg I’ as very slight& toxic. 3ouany
and Dab&e [25] classii pesticides into 5 classes according to their EC50 or LC50 for aquatic oqanisms: a)
<o.OOl mg I’, b) 0.001 - 0.01 mg I’, c) 0.01 - 0.1 mg I’, d) 0.1 - 1 mg I’, e) > 1 mg I’. Drawing from these
revs we de&e the input variable Aquafic toxici@ as the log10 of the EC50 or LC50 of the most
sensitive species of the three considered. We use 0 (log,0 of 1 mg I’) as the median vahre for the transition
inter& and &tins the limits of the transition interwl by assigniug complete membership to F if Aquafic
toxicity > 2 (100 mg kg-‘) and complete membership to U ifAquatic toxicity < -2 (0.01 mg kg-‘).
The vahte of the module Rrur depends on the input variables Run&ri& Driftpercentage, Position of

application, Field hal$l$e and Aquatic toxic@ according to a set of 32 decision rules (not shown). These
decision rules (sumnwi& in Fig. 4) reflect our “expert” perception of the risk of surface water
ummm&Gon as a result of a pesticide application.

4.4 The module Risk of air contamination


The indicator module Rair (Risk of air contamination) reflects the potential of a pesticiie to volatihse
and to contaminate air. The value of Rair depends on four input variables: 1) pesticide v~latihty, 2) the
position of applieation of the pesticide (on the crop, on the so& in the soil), 3) the fiekl half-life of the
pesticide and 4) the toxicity of the pesticide to humans.
Henry’s law constant (IGI, dimemionless), the ratio of vapour pressure to water sohtbihty, is considered
a more appropriate indicator of the volatihsation rate of a pesticide than its vapour pressure alone [47,48].
Compounds with Ku much greater than 2.65 x 1u5 are volatii, compounds with I& much smaller than 2.65
x 10T5are much leas volatile [47,49]. We use KH = 2.65 x 1O-’as the middle of the tmnsition interval, and
define the input variable Volatility as log10 & rather than as I& We define the limits of the transition
interval by assigning complete membership to F if Volatility < Log10 2.65 x 10d and complete membership
to U if Volatility > Log102.65 x 10”.
The position of application of the pesticide strongly affects its volatilisation potential. Application in the
soil is favourable as volatiition is greatly reduced by incorporation into the soil [50]. Consequently,
application on the soil or on the crop is unthvourable. We therefore defiued the input variable Position of
appZiCationoi,as a value on scale between 1 (100% of the pesticide in the soil) and 0 (0% of the. pesticide in
the so& pesticide applied on the soil/crop). Its vahte is &&ted as: Position of applicatiork~ = % of
pesticide applied in the soil/lOO. The % of pesticide applied in the soil is to be estimated by the user of the
system. We define the limits of the trausition interval by ass@ng complete memhe&Q to F if Position of
applicatio~i, = 1 and complete member@ to U if Position of applicatio~i, = 0.
Field haQ=l& (DTSO) atSxti a pesticide’s volatihsation potential, persistent pesticides that remaiu in
the soil will volatikse over longer periods of time than non-pem&ent pe&ides (e.g. j51]). The limits of the
transition interval for Field hal$li$e have been dehed in section 4.3.
2239

L-J Volatilily

,......i.....‘.........‘..........:.........:.........~.....,

0.0 0.1 0.2 1.0 i


. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0.6
. . . . . . . . . .0.6
..................
Risk of air contamination

Fii 5. The em of the input variables Volatility, Position of application, Field half-life and Human
toxicityonthevalueoftheco~~~ofthedecisionruksfortbeindicatormoduleRair(Riskofair
m&&&ion) according to their mmbe&p to the fuzzy sets Favourable (non-sheded boxes) and
U&wmabie (shaded boxes). For details see text.

We selected the Acceptable Daily Intake (ADI), which reflects chronic toxicity to humans, as our
miterion for toxicity. The deiZtion and the limits of the tram&n interval for HwMn toxicity wese given in
section 4.2.
The value of the module R&r depends on the iaput variables Volatility, Position of upplicution, Field
half-life aud Hwnun toxicity accmdbg to a set of 16 decision rules (not shown). These decision rules
(summmkd in Fig. 5) reflect our “expert” pemption of the risk of air water contamination as a mult of a
pesticide application.

5. AGGREGATION OF THE MODULES: AN INDICATOR OF PESTICIDE IMPACT

mu&r&ma amdysis technique, using the modules as evaluakn criteria [6]. An alternative approach would
betoaggregate~fourmodulesinsomewayintoanoverallindicator(Ipest).Thiscanbedoneby
summation, multiplication or a c~mbimtion of both. For exampk Ipest = Presence (Rsur + Rgro + Rair),
or Ipcst = Presence max(lLwr, Rgro, Rair).
2240

water, groundwater and air.In mostindicators of environnYmtal impact surf&x water and groundwater are
given much emphasis, whereas air is rarely takeninto consideration [5]. It is hard to decide on the relative
importance of air pollution by pesticides, because whereas data concerning pesticide concent&ons in
groundwater and surface water are readily a&able, those for air are scarce. However, calculations by
Worthsm (pers. comm. 1996) based on data by Millet [52] show that pesticide concentrations in air in
Alsace (NE France) are such (0.068 pg m3) that pesticide uptake by humans t?om air may be six times
larger than uptake from drink& water (assuming a 0.1 pg 1-lconcentration in drinking water). Consider@
thisevidencewedecidedtogivethesameweighttotheaircompartmentastothesm&cewaterandthe
groundwater compartments in the aggregation of the modules.
The vahte of the indicator Ipest depends on the tmxlules Presence, Rsur, Rgro ad R&r (used as input
variables) according to a set of 16 decision rules (not shown). The detinition of the limits of the tmuskion
interval is the same for the four input variables: we assign complete membership to F if the vahte of the
moduleisOamlcompletemembershiptoUifthevalueofthemod~isl.TBee~reasoningis
summansed in Fig. 6: ifall modules are F, the value of Ipest is 0.0 (no risk). If Presence is F and one or
more of the other three modules is U, the vahte of Ipest is 0.3, (0~ module U, two F) 0.4 (two modules U,

one F) or 0.5 (three modules U). If Presence is U and the other three modules are F, the value of Ipest is
0.1; if one or more of the other three modules is U, the value of Ipest is 0.8, (one module U, two F) 0.9
(twomodulesU,oneF)orl.O(~naod~sU).Theserulesobvio~retlectvaluejudgements,andcan
be mod&d to suit the perception of a user of the system

plesence 0.0 0.3 0.3 0.4 0.4 0.5 :


.. ... ....... ......... ...... ... .............. ........ ..... .. ..... .............
. .... ..... .. ..... ............ ........ . ............. ....... ... ................
0.1 0.8 0.8 0.9 0.9 1.0 :
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..(
Indicator of pesticide environmental impact

Figure 6. The effect of the modules Presence, Rmr, Rgro and Rair on the v&e of the conchrsions ofthe
decision rules for Ipest (imhcator of pesticide environmental impact) according to their m to the
l&y sets Favourable (non boxes) and U&vourable (shaded boxes). For details see text.
2241

6. ILLUSTRATIVE RESUL.TS

InordertoillustratethefUnctioningofthesystem,we~ananahrsisof~sensitivityofthe
iadicatorIpesttovariationinthevaluesofitsinput~~.Each~utvariablewasvariedoverits
transition~~theotherinputvariableswetekept~attbemedianvalueoftheir~n
interval, or at the extremes of the transition interval, i.e. at “Favourable” or at “Unfsvourable” (Figs. 7a, 7b).

+eus
+ Human toxicity j
-Aquatic toxicity !
~Volatilii
--O--Runoffriek
--O--Drift%
- - A - -Leaching risk

0.525

0.500

0.475

0.450

0.425

0.400

0.375

0.350

0.325

0.200 --
,(+---0
0.175 -- *’
P’
0.150 -. #’

.o”
0.125 --
,’
0.100 -- 0’
0.075 -- ,’

0.050 -- .@’
#’

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.6 0.9 1.0

Figure7a~of~~~ofIpesttovariationofallinputvariablesexceptRate.Eachinput
v8riabk3 is varied over its trandion intend from 0.0 (completely hxuable) to 1.0 (completely
unfirvo~)~o~inputvariaMesarekeptatunfavoutaMe(top),atthemedienvalueofthe
trausihn intend (middle) or at unfhourable (bottom). Some curves coincide.
2242

We assumed the pesticide was applied by field spraying on the crop and/or the soil. Consequently, the vahte
of the input variable Position of application depended only on soil cover by the crop, which was varied
between 0% and 100%. The sensitivity analyGs reflects the tunctioning of the system and supplies in
particular some indication of the relative weight the input variables may have in the vahre of ipest. However,
one shoukl be aware that the eflbct of the variation of an input variable over its tram&ion interval on the
value of Ipest depends very much on the value of the other input variables. Therefore, results presented here
should be considered as ilhrstrations of the functioning of the system
The most influent input variable is Rate of application (Fig. 7b), its effect is very large when all other
input variables are unfavourable and much smaller when the other input variables are tavourable. This
obviously is the result of the mode of aggregation (Fig. 6) we adopted, giving a lot of “weight” to the
module Presence (which depends on Rate of application only) when the other three modules are
unlavourable. The extent to which other input variables affect the value of Ipest in this sensitivity analysis
can similarly be deduced from the decision rules involved. For instance: the Muence of Field half-life is
bwge (Fig. 7a), because it is an input variable to three modules (Rru, Rgro and Rair). The effect of Runoff
risk is smaller than that of Drift percentage (Fig. 7a), because the effect of Runoff risk is modulated by
Position of application, Field half-life and Aquatic toxicity, whereas Drifr percentage is modulated by
Aquatic foxicity only (Fig. 4).
Another ill~tion of the functioning of the system is given in Table 6, which shows the value of the
four modules and of Ipest for some pesticide application cases. Results are shown fix different pesticides,
applied at theii recommended rate of application in a field with Leaching risk = 0.5, Dr@ percentage = 0

1.0

0.9

0.6

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

Figure 7b. Aaaiysis of sensitivity of Ipest to variatiin~of the Rateisvariedoverits


tran&ion &erval from 0.0 (completely tiourable) to 1.0 Able)wh%thf?othCX
~~~b~~keptat~~~,~madianraiueof~~~or~o~
2243

Table 6. The values of the modules Presence (&es.), Risk of surface water contamination (Rsur), Risk of
groundwter contamination (Rgro) and Risk of air contamination (Rair), and of the Indicator of Pesticide
Environmental Impact (Ipest) for a wmber of pesticides applied at their rec~mmeded rate in a field with
either Runoflrisk = 0 (no runoff ri&) or Runoflrisk = 1 (major runoff risk), Leaching risk = 0.5 end Dr~ji
percentage = 0. Rfite of application and position of application are given, pesticide properties used for the
cab&&ions are in Table 7.

Pesjkide name Rate Position’ Pres. Rsur Rgro Rair I@


(kg ha-‘) RluKlfE0RluKlE 1 Ru~~ff:ORu&?l
2,4-D 0.300 on&, 50% 0.49 0.00 0.31 0.24 0.00 0.16 0.31
aIachIor 2.400 on c/s, 0% 0.90 0.00 0.74 0.21 0.00 0.17 0.68
atmzine 1.000 on c/s, 0% 0.75 0.00 0.70 0.75 0.00 0.57 0.62
carbofuran 0.600 insoil 0.64 0.00 0.10 0.60 0.00 0.41 0.42
*thrin 0.040 onc/s,lOO% 0.10 0.00 0.10 0.00 0.61 0.21 0.24
EPTC 3.600 insoil 0.95 0.00 0.10 0.00 0.10 0.12 0.15
gIyphosate 4.300 on c/s, 100% 0.96 0.00 0.10 0.00 0.00 0.10 0.12
isoproturon 1.800 on&, 10% 0.86 0.00 0.80 0.26 0.00 0.21 0.68
on c/s, 0% 0.79 0.00 0.98 0.42 0.56 0.58 0.80
on&, 100% 0.49 0.00 0.10 0.00 0.40 0.25 0.27
on c/s, 0% 0.87 0.00 0.93 0.00 0.70 0.61 0.81
on cl& 5% 0.01 0.00 0.37 0.00 0.00 0.00 0.09
‘onc/s:appliedonthecropand/oronthesoilthepercentageindicatesthe~nofsoilcoveredbythe
crop at the time of application; in soil: applied in the soil.

and with e&x Runoff risk = 0 (no nmoff risk) or Runoffrisk = 1 (major runoff risk). This type of output
canbeusedtocompareorrankdifferen~ic~eapplIcationswithrespecttothe~ofoneorseveralof
the modules or of the Ipest indicator. The results obtained here again evidently are the consequence of the
choices we made regard& the selection of input variables, the definition of their transition intervals and the
vahzs given to the ConclWions of the decision rules.

Table 7. Pesticide properties used for the calculation of module values and Ipest values iu Table 6. Data are
taken from the INRA database AGRITOX (F. Bouneb, pezs. corn 19%), except for KH, which was taken
i?om [46]. Data taken f?om other sources are indicated by index letters, C: data Tom [56], E: estimate by
the authors, I? Pesticii Manual [l], R: RIVM data [46], S: the SCWARWCES pesticide properties
database (ILL. Glenn, pers. corn..”1996).
Pestkide name DTSO Koc GUS KB AD1 Aquatox
Q-1
(days) @my) (wk ) (mg l-1)
2,4-D 10 20 2.70 1.4 1O-9 0.3000 0.9
alachlor 15 170 2.08 9.5 lo-’ 0.0005 0.05
60 100 3.56 1.2 IO-’ 0.0005 4.3
zzzzan 50 22 4.52 9.6 1U’ 0.0100 o.015p
cyflukin 3os 100000 -1.47 1.8 lOA o.0200p o.00014R
EPTC 6’ 200 1.32 9.6 10” 0.3160’ 1.4R
glyphosate 47s 1 1275R -0.09 a.5 10dE 0.0500 15R
isoproturon 17p 155 2.23 4.7 1O-9 0.0060 0.004R
400 1100 2.49 6.1 10-5 0.0080 0.034
zn 49 3010R 0.88 2.5 lo-’ 0.005op 0.001 SR
Fthalin 90 5000 0.59 1.5 10-3 0.05 0.138
runsu&on 3p 6@ 1.06 2.5 lo-* 0.0160 l.aR
2244

7. IMF’L.EMRNTATiON OF THE EXPERT SYSTEM

The expert systemdescribed above assesses the environmental impact of a single application of a
~~.Thesystemcanbeusedasadecisionsupportsystemtorankorchoose~~ve
pesticide application options. In this section we will briefly discuss some major issues conWrning the
pm&xl implemeatation of the expert system These issues are: the coustruction of a so-called “agro-
ecological bdicator”, mixtures of pesticides, pesticide metabolites and variabi& of the values of input
variables.

This work was carried out within an ougoing project comxrning “agro-ecological indicators br the
evaluation of tarming systems” as presented by HockstaUer et al. (53). Agro-ecological indicators assess the
effects of cultural practices and farming system (e.g. nitrogen matrage- crop diversity) on the
environment and are destined to be used by farmers, farmer-advisors or other decision makers concerned
with the environmental impact of agriculture. Agro-ecological indicators take a value between 0 and 10, the
value 7 representing the achievement of the minimum requimmems of Integrated Arable Farmiug Systems
(IAFS). Values below 7 indicate that the cultural pm&es do not satisfy minimum IAFS requimmems,
whereas an indicator value above 7 indicates that the farmer does better than these miuimum requirements.
The time scale of calculation is generally about one year: the period between the harvest of the preceding
crop and the harvest of the crop in the current year. The value of Ipest as described in section 5 concerns a
single pesticii application. The agro-ecological indicator IPEST is calculated for all applieatious related to
acropwithinayearas:
IPEST = 10 - k C Ipesti
wherekisaconstantdepend.iugonthecropandtheregionandIpest~isavalueofIpest(betweenOand1)
for a single application of a pesticide. The value of k is chosen such that a value of 7 for IPEST is obtained
for a crop protection programme sati@ng the minimum reqmrements for IAFS according to a group of
experts.
Often a commercial product contains two or more pesticides. For the calculation of Ipest such au
association can either be considered as several separate applications or as a single application. In the latter
case Ipest can be calculated as the weighted mean of the values obtained for each of the component
pesticides at the rate of application of the mixture. For example: 2.88 kg ha“ of a mixture consisting of 70%
of ala&or and 30% of atmzine is applied on the soil in a field with Runoff risk = 0, Leaching risk = 0.5 and
Driftpercentage = 0. Under these circumstances the application of 2.88 kg ha’ of alachlor would result in a

value for Ipest of 0.17, aud the application of 2.88 kg ha-’ of atmzine woukl result iu a value for Ipest of
0.68. The value of Ipest for the mixture is calculated as: Ipest = 0.7*0.17 + 0.3*0.68 = 0.323.
As pesticide degradation products (metabolites) may have envimnmeutally undesirable chara&&tics, the
evaluation of the environmental fate of a pesticide shot&l also take the fate of its major metabolites into
consideration. Unfbrtunately, data availability fix metabolites is often poor. However, if we kuow for
2245

instancethatrhnsulffmMmayyieldupto55%ofametabolitewitha550dayDT50[46],wecantalre~
inso~~bytreetingan~~nof15gha“ofrimsulfUronasa~of15g~‘ofrhnsulfuron
and 8.25 g ha-’of its metabolite, accord@ to the approach outlkd above.
Aspointadoutintbe~nofthispaper,amajorproMeminthe assessment of peskide
emriromnentalimprsctistbeuncertaintyand/or~~nofsomeofthemajorinpHlt~involved
(e.g. DTSO, Koc). Measured~forthesevariablesshowahugevariabilifyreletedtosoilclimaticor
other environmental conditions. The use of some consensus ‘average’ vahte br this type of data is a
pragmatic but far from satisktory sohttion. Provided the variability of input variables can be adequately
cbsracterised(whichisaprobleminitself)tbeaplmoaehinvohring~setsd~~inthispaperisvay
well suited to take into account unce&&y of input data [20].

8. DISCUSSION

Inthedesignofasystemtoassesspesticideemriromnentalimpacttwolnajorquestionshavetobe
aaswered:a)whichinput~~~uldbetaken~oacco~?,b)how~~~~~variabiesbe
aggmgated?Then&hodpresentedinthispaperproposesananswer to both questiork$ however, its
origin&y lies in the answer it provides to the second question. Compared to other methods to assess
pesticide environmental impact (reviewed iu [3-51) our approach contains two new elements: a) the use of
fuzzvsets,b)theuseof~~ntules.Theuseoffu2ZLYSetspfOvideSanelegendsolutiontotheproMemof
deciding on the cut-off values tbr input variables: e.g. the limit between “non-persistent” and “modemtely
persistent” pesticides. The use of decision rules allows an “iutelligent” aggregation of iuput variabk e.g.
Position of application, Leaching risk and Human toxicity afEet Rgro only when GUS is unfavourable (Fig.
3). The combinations of these two concepts in sets of tkzy rules (see Sections 3 and 4) is attractive,
because although the combiuatiins of values of input variables (e.g. Rate of application and Field half-life,
Section 3) arc Snite, a single set of fiuxy rules connects them all.
The system to assess pesticide environmental impact proposed here requires validation, which means
that we should test whether its origii objective has been achieved. The agro-ecological imlicator IPEST is
based on an expert system. The objective of an agro-ecological indicator is to render reality intelligible, and
thus its vabdation consists of determinmg its vahre to potential users [54]. The objectke of au expert system
isthesimulationofahumane~,thustheexpertsystemisvalidatedifitdisplaysunderavrvietyof
conditiins the responses to input that the human expert would display [55].
IPEST and the other agro-ecological indicaton are current@ being tested on 17 commemial arable
~ofthe~~yinFranceandGennany[53]andonthefannsofthreesecondaryschoolsfor
agricultural train& in France. These users will be surveyed to obtain their evahratk and proposed
2246

Tbevalidationofthe~systemsirnilarlyisongo~.AdraRversionofthispeperwitha~n
ofaGrstprototypeofthesystemwassenttoover20expertsinthefiekiofpestic&envirom?x&alimpact
and its m. Most of these experts supplied comments (see A&now&&me&s). Several
modiscationsandadditionsweremade,leadingtotheversionptesentedhae.Thesystemiscurreaatyused
tocomparetheenvironmentalimpactofsioglepesticideapp~aIldofsequeoces of pesticide
applicat~~in~cropsandcroppingsystenrs.Theresultsthpsobtairrsdwiube~~alosgwith
thoseobtainedusingotherpesticideimpact- systenrrsU&aStheEnvirolgaerrtalhXlp&Q@i%nt
[24] or the CLM Environmental Yardstiik for Pesticides [ll]. Experts will be invited to comment on the
results. If there is disagreement between expert perception of pesticide environmental impact and the output
of the expert system we will examinethecausesofthisdivergence,which~ybeoneorseveralofthe
following: choice of input variables, choice of the limits of the tmnsitbn interval, fbrmulation of the decision
rules, values given to the conclusions of the decision rules, or the mode of aggq@ion of the nuxlules. All
ofthesemaybemodi6edaccordingtoexpertcQnsensus.

9. CONCLUSIONS

We propose a t%zzy expert system reflect@ an expert perception of potential pest&ide environmental
imI#sct.Thesystemtakesintoeccountthreetypesofinputvarhtbles:pesticideproperties,~~~
conditions and pesticideapplkation iictors. It can be used as a decision support to rank or choose between

akerna&epesticideapplk&onoptions.Thesystemhasamodularstr&ure andthusprovidesbotha
sylltbetic~or~~overall~asweUasmDredetaibsdintbrmationthrougbitsfourmodules.
New modules can be added ifnecessary Thesystemistlexibleandcaubetunedtoexpertpercq&n.
2247

ACKNOWLEDGEMENTS
Thisresearchwasfundedbya research trainiug fellowship (Contract number AIR3-BM-0022) of the
Commission of the European Cods Research and Techuological Development programm of DG
XII.Thegraphicalrepreserrtationofthedecision~isduetoOlivierRousselAdraftofthispeperwas
submitted to a number of experts in the field of pesticide environmental impact and its asmamnt. The
autlms would like to thank G. Assouliue, E. Baniuso, R Belamie, C. BockstaUer, A. Cavelier, J. Delphin,
P.Gimdin,D.Gusta&on,C.Kempemr,J.Kroes.,L.Levi@RMartin-Clouaire,RNeumm,B.R&xl,J.
I&us, C. Walter, F. Wm and H. Wortham for providing valuable comments on the mammr@t. We
thank F. Bouneb (INlZA-Phytopharmacie) and R.L. Glenn (USDA) for supplying pesticide data_

References

1. Tomlin, C.D.S., 1994. The Pesticide Manual, incorporating The Agrochemicals Handbook 10’ edition.
Britii Crop Protection Council, Far&am, United Kiugdom.
2. Hay% W.J., 1991. Introduction. Pages 1-37 in: Hayes, W.J. aud Laws, E.R (eds.) Handbook of
Pesticide Toxicology. Academic Press, San Die80, CA, USA.
3. Levi&n, L., Merwin, I. and Kovach, J., 1995. Asses&g the relative environmental impacts of agricultoral
pesticides: the quest for a holistic method. Agric., Ecosyst. Environ. 55: 153-168.
4. Shuhaue, A.N. and Inmnn, RC., 1987. Review of available pesticide rating systems. Hazard Assessment
of Chemicals 5: 103-132.
5. Van der Werf; H.M.G., 1996. Evahmtiog the impact of pesticides on the environment Agric., Ecosyst.
Environ. 60: Sl-%.
6. Gira&n Ph., Bockstaller C!h.andvanderWerf;H.M.G,inpress.Evalua&nofthesustainabilityofa
farm by means of indicators. Ix Quel environnement pour le XX?“” siecle. Proceedings, Fontevraud,
September 1996. GERMES, Paris, France.
7. Gras, R, Benoit, M., Del%nt&es, J.P., DUN, M., Lafhrge, M., Langlet, A., and Osty, P.L., 1989. Le
fait technique en agronomie. Activite’agricole, concepts et methodes d’etude. INRA, L’Harmrlttan, Paris,
FranCe.
8. Mitchell, G., May, k aud McDonald, A., 1995. PICABEU a tnethodok@cal -work for the
developnent of indicators of susta&ble development. In?. J. Sust. Dev. World Ecology 2: 104-123.
9. Sevem, D.J. and Ballard, G., 1990. RiswbeneM and regulations. Pages 467-491 in Pesticides in the soil
environment. Soil Sci Sot. of America Book Series, no. 2, Madison, WI, USA.
10. Emans, H.J.B., Be&, M.A. and Linders, J.B.H.J., 1992 Evaluation gstem for pesticides (ESPE) 1.
Agriculturalpesticides. RIVM, report no. 679101004, Biioven, The Netherlands.
11. Reus, J.A.W.A. and Pak, G.A., 1993. An environmental yardstick for pesticides. Med. Fat.
Landbouww. Univ. Gent, 58/2a: 249-255.
12. Rus&, M.H. and Layton, J.L., 1992. Models and modeling in a regulatory setting: considerations,
applications, and problems. Weed Technol. 6: 673-676.
13. Wagenet, RJ. and Rao, P.S.C., 1990. Modeling pesticide fate in soils. Pages 351-399 ix Pesticides in
the soil environment. Soil Sci. Sot. of America Book Series, no. 2, hkdison, WI, USA.
14. EPPO, 1993. De&km-making scheme for the enviromnental risk assessment of plant protection
products, Chapter 3, Soil EPPO Bull. 23: 27-49.
15. G&e%,R, 1995. Modelliug pesticide leaching in soils, main aspects and maio ditIiculties. Eur. .L Agron.
4: 473-484.
16. FOCUS, 1995. Leaching models and EU registration. The 8nal report of the Regulatory Model@
Work Group of FOCUS. DOC. 4952/VI/95. Commission of the E.C., D.G. br Agriculture VI B II-l.
2248

17. Gus&on, D.I., 1995. Use of computer models to assess exposure to ag&&wal ckankaia via drink&
water. Sci. total Environ. 171: 35-42.
18. m@ J., Caklwell, B. and Mohler, C.L., 1994. Evalwing the environmental e&ct of peat&&w: a
critique of the environmental impact quotient. Am. Entomol. 40: 180-184.
19.Ha&L.O.ar.xlKandel,A., 1991.Tbeevolution~mexpertsystenrstofiay~syslRms.Pogp;s3-
21 in: Kaudel, k (Ed.) Fuzzy experr systems theory. CRC Press, Boca Raton, Fkwida, USA.
20. Boucho~Meunies, B., 1993. La Zogiiuefloue. Presses Uni ver&&wde Frauce, Paris, Frauce.
21. Zadeh, L.A., 1%5. Fuzzy sets. Inform. Control 8: 338-353.
22. Guata&on, D.I., 1989. Growlwater ubiquity score: a simple method for asseas@ pesticide lea&@&y.
Environ Toxicoi. Chem. 8: 339-357.
23. Sugeno, M., 1985. An introductory survey of f&y control. Inform. Sciences 36: 59-83.
24. Kovach, J., Petzoldt, C., Degni, J. and Tette, J., 1992. A method to measuretheenviromnentalimpact
of pesticides. New York’sFood L#&Sci. Bull. 139. N.Y. State Agric. EBB.Stat., Geneva, N.Y., USA.
25. Jouany, J.M. and Dal&x, E., 1994. Classements des substances actives plytosanitaires en vue de la
surveillance de la qua&! des ealde b I’kchelle nationale. Direction de IT&ace Rural et de la For&
Minis&e de l’Agricuhure et de la P&&e, Paris, France.
26. Vereijken, P., Wljnands, F. and Stol, W., 1995. Progress Report 2. Designing and testingprotot@es.
Progress reports of the research network on integrated and ecological arable fanning systems for EU
and associated countries (Concerted action AIR 3 - CT920755). AB-DLO, Wageningen, The
NetherlaUdS.
27. Ha&n, E., Gala& S., BrUggemann, R. and Provini, A., 1996. Selection of prior&y propert& to aweas
environmental hazard of pesticii. Chemosphere 33: 1543-1562.
28. Shukla, S., Mos@himi, S., and Bruggemau, A.C., 1996. A risk-based approach for selecting priority
pesticides for groundwater monitoring programs. Trans. ASAE 39: 1379-1390.
29. L,askow&i, D.A., Goring, C.A.I., McCall, P.J. and Swarm, RL., 1982. Terrestrial emriromnent Pages
198-240 in: Conway, RA. (ed.) Environmental risk analysis for chemicals. Van Nostratul Reh&&&.,
NY, USA.
30. Cohen, S.Z., Creeger, S.M., Carsel, RF. and Euliekl, C.G., 1984. Potential for pesticide contamkwtion
of groundwater resulting from agricultural uses. Pages 297-325 in: Kruger, R.F. and seiber, J.N. (eds.)
Treatment and disposal of pesticide wastes. ACS Symp. Series no. 259. American Chem. SOC.,
Wa&ington, DC, USA.
3 1. Jury, W.A., Focht, D.D. and Farmer, W.J., 1987. Evaluation of pesticide groundwater pollution
potential from standard indices of soil chemical adsorption and wn. J: Environ. @al. 16:
422-428.
32. Homsby, A.G., 1992. Site-specific pesticide recommendations: the fmal step in environmental impact
prevention Weed Technol. 6: 736-742.
33. Leonard, R.A., 1990. Movement of pesticides into sur&ce waters. Pages 303-349 iw Pesticides in the
soil environment. Soil Sci. Sot. of Anwrica Book Series, no. 2, Madison, WI, USA.
34. Goss, D. and Wauchope R.D., 1990. The SCS/ARSK!ES pesti&e properties database: II Using it with
soils data in a screen@ prowdue. Pages 471-493 in: Weigman, D.L. (ed.) Pesticides in the wxt
decade: The challenges ahead. Virginia Water Resources Research Center, Biacksburg, VA, USA.
35. Aller, L., Bermett, T., Lehr, J.H. and Petty R.L., 1985. DRASTIC: a star&rdized system for ewlwting
groundwter poltution potential using hydrogeologic settings. U.S.EPA, Environmental Rasearch
laboratory, Ada, OK, USA. U.S.EPA Report 600/2-851018.
36. Hollis, J.M., 1991. Mapping the vulnerability of aqui&rs and surl&e waters to pesticide co&am&ion
at the national/regional scale. Pages 165-174 in: Walken, A., (ed.) Pesticides in soii and nater: current
perspectives. Lavenham Press Ltd., Lavenham, UK. BCPC Monograph nr. 47.
37.VanderZee, S.E.A.T.M. andBoesten, J.J.T.I., 1991.E~ofsoilhete1~geneityonpesticideleachiag
to groundwater. Water Resow: Res. 27: 3051-3063.
38. Edelman, PA., 1991. Preve&onofinjurybypesticides. Pages405-451 in: Hayes, W.J. andLaws, E.R
(eds.) Handbook of Pesticide Toxicology. A&e&c Press, San Diego, CA, USA.
39. Larson, S.J., Capel, P.D., Goolsby, D.A., Zaugg, S.D. and Sam&mm, M.W., 1995. Relations between
pesticide use and ~fluxintheMississippi~basinchemosphere31:3305-3321.
2249

40. Simon, F., 1995. Analyse aks facteurs de risques de transfer de pesticides Asiansles paysages.
Etablissement d’une hierarchic de ces risques: application b a!es bassins versants. Mkmoive de SII
d’kudes. Ec& Nation& Sup&ieure Agronomique, Resmes, France.
41. Papy, F. and Boifiin, J., 1988. Influence des systemes de culture sur les risques d’erosion par
ruissellenwnt concentre. I. E-ion des pos&ilites de ma&rise du ph&rom&ne dans les exploiitions
agricoles. Agronomie 8: 745-756.
42. Aurousseau, P., Squividant, H., Baque, M.C. and Simon, F., 1996. Analyse des facteurs de risques de
transfer& de pesticides darts lespqysages. Etablissement d’une hierarchic de ces risques: apPication au
calcul d’un indice de risque par bassin versant et par parcelle. Rapport fma de la convention 9500.023
avec l’A8ence de l’Eau LoiiBreta8ne. Ecole Nationale Sup&&me Agronomique , Rermes,Frarxe.
43. CLM-MC, 1994. Achtergronden van de milieumeetlat voor bestrijiingsmiddelen j@ckgrouuds of the
environmental yards&k for pesticides]. IKC-AT, Kerngroep MJP-G, Ede, The Netherlands.
44. Leonard, Rk, Langdale, G.W. and Fleming, W.G., 1979. Herbicide runoff fiom upland Piedmont
watemheds - data and implications for modelin8 pesticide transport. J. Environ. Qual. 8: 223-229.
45. Canton, J.H., Linders, J.B.H.J., Luttik, R, Mensink, B.J.W.G., Panman, E., van de Plassche, E.J.,
Spare&~, P.M. awl Tuiustm, J., 1991. Catch-up operation on old pesticides: an inte8ration. RIVM,
report no. 678801002, Bilthoven, The Netherlands.
46. Linders, J.B.H.J., Jansma, J.W., Mensink, B.J.W.G. and Gtermatm, K, 1994. Pesticides: be&action or
Pandora’s box? A synopsis of the environmental aspects of 243 pesticides. RIVM, report no. 679101014,
Bilthoven TheNetherlands.
47. Jury, WA., Farmer, W.J. and Spencer, W.F., 1984. Behavior assessment model for trace organics in
soil: II. chemical classification and parameter se&ii. J. Environ. Qual. 13: 567-572.
48. Spencer, W.F. and Cliath, M.M., 1990. Movement of pesticides from soil to the atmosphere. Pages 1-16
in Kurtz, D.A. (ed.) Long range transport ofpesticides. Lewis publishers, Chehea, Michi8an, USA
49. Clenden&, L.D., Jury, W.A. and Ernst, F.F., 1990. A field mass bslance study of pesticide
volatihzatiou, leaching and persistence. Pa8es 47-60 ix Kurtz, D.A. (ed.) Long range transport of
pesticides. Lewis pubhshers, Chelsea, Michi8an, USA.
50. Taylot, AW. and Spencer, W.F., 1990. Volatilization and vapor transport processes. Pages 213-269 in
Pesticides in the soil environment. Soil Sci. Sot. of America Book Series, no. 2, Madison, WI, USA.
51. Spencer, W.F., Sir@, G., Taylor, C.D., LeMert, RA., Cliath, M.M. and Farmer, W.J., 19%. DDT
pem&enw and volatility as a&ted by maaagement practices after 23 years. J. Env. Qual. 25: 815-821.
52. Millet, M., 1994. Etude de la composition chimique des brouillards et analyse a&pesticides akns les
phases liquide, gazeuse etparticuliere de l’atmosph&e. Thesis. Univ. Louis Pasteur, Strasbour& France.
53. Bocksmller Ch., Girardin, Ph and van der Werf, H.M.G., in press. Use of agro-ecological indicators tbr
the evaluation of farming systems. Eur. J. Agron.
54. Gkardin Ph., Bockstaher, Ch. and van der Werf H.M.G. Indicators: a tool to evaluate the
environmental impact of farming systems. Submitted fix publication to Environmental Management.
55. Plant, RE. and Stone, N.D., 1991. xirowledge-based systems in agriculture. McGraw-Hill Inc., NY,
USA
56. Comite de liaison, 1996. Tableau de don&es de base. Version 16 January 1996, annex to the
document: Clasaement des s&stances actives phytossnitaires en vue de la surveillance de la qualite des
eaux - Fonctiomwment de la n&&ode et recomma&tions pour une application r+ionale. Comite de
liaison “eau-produits ant+ra&kes”, Minis&y of the Environment, Mini&y of Agriculture, Fisheries
and Food and Minis@ of Employment and Social Services, Paris, France.

You might also like