MASC, a qualitative multi-attribute decision model for ex ante assessment of the sustainability of

16 Pages·2017·1.99 MB·English

MASC, a qualitative multi-attribute decision model for
ex ante assessment of the sustainability of cropping
systems
Walid Sadok, Frédérique Angevin, Jacques-Eric Bergez, Christian Bockstaller,
Bruno Colomb, Laurence Guichard, Raymond Reau, Antoine Messéan,
Thierry Doré
To cite this version:
Walid Sadok, Frédérique Angevin, Jacques-Eric Bergez, Christian Bockstaller, Bruno Colomb, et
al.. MASC, a qualitative multi-attribute decision model for ex ante assessment of the sustain-
ability of cropping systems. Agronomy for Sustainable Development, 2009, 29 (3), pp.447-461.
￿10.1051/agro/2009006￿. ￿hal-00886501￿
HAL Id: hal-00886501
https://hal.archives-ouvertes.fr/hal-00886501
Submitted on 1 Jan 2009
HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est
archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
Copyright
Agron.Sustain.Dev.29(2009)447–461 Availableonlineat:
(cid:2)c INRA,EDPSciences,2009 www.agronomy-journal.org
DOI:10.1051/agro/2009006 for Sustainable Development
Research article
MASC, a qualitative multi-attribute decision model for ex ante
assessment of the sustainability of cropping systems
adok ngevin ergez ockstaller olomb
WalidS 1,FrédériqueA 2,Jacques-EricB 3,ChristianB 4,BrunoC 3,
uichard eau essean ore
LaurenceG 5,RaymondR 5,AntoineM ´ 2,ThierryD ´6*
1INRA,UMR211INRA/AgroParisTech,BP01,78850Thiverval-Grignon,France;Presentaddress:Agronomy-PhysiologyLaboratory,
UniversityofFlorida,POBox110965,Gainesville,FL32611-0965,USA
2INRA,UAR1240Eco-Innov,BP01,78850Thiverval-Grignon,France
3INRA,UMR1248AGIRINRA/ENSAT,BP52627Auzeville,31326CastanetTolosan,France
4INRA,UMR1121Nancy-Université-INRANancy-Colmar,BP20507,68021Colmar,France
5INRA,UMR211INRA/AgroParisTech,BP01,78850Thiverval-Grignon,France
6AgroParisTech,UMR211INRA/AgroParisTech,BP01,78850Thiverval-Grignon,France
(Accepted16January2009)
Abstract–Realisticassessmentsofsustainabilityareoftenviewedastypicaldecision-makingproblemsrequiringmulti-criteriadecision-aid
(MCDA) methods taking into account the conflicting objectives underlying the economic, social and environmental dimensions of sustain-
ability,andthedifferentsourcesofknowledgerepresentingthem.SomeMCDA-basedstudieshaveresultedinthedevelopmentofsustainable
agriculturalsystems,butthenewchallengesfacingagricultureandtheincreasingunpredictabilityoftheirdrivingforceshighlighttheneedfor
fasterexante(‘Before-the-event’)assessmentframeworks.Theseframeworksshouldalso(i)provideamorerealisticassessmentofsustainabil-
ity,byintegratingawiderrangeofinformalknowledge,viatheuseofqualitativeinformation;(ii)addressalternativescales,suchascropping
systemlevel,improvinggranularityforthehandlingofsustainabilityissuesand(iii)targetalargerpanelofdecision-makersandcontexts.We
describeheretheMASCmodel,whichisatthecenterofaframeworkaddressingtheseobjectives.TheMASCmodelhasatitscoreadecision
treethatbreaksthesustainabilityassessmentdecisionalproblemdownintosimplerunitsasafunctionofsustainabilitydimensionalstructure
(economic,socialandenvironmental),generatingavectorof32holistic‘mixed’(quantitativeandqualitative)elementarycriteriaratingcrop-
pingsystems.Theassessmentprocessinvolvesthecalculationofthesecriteria,theirhomogenizationintoqualitativeinformationforinputinto
themodelandtheiraggregationthroughoutthedecisiontreebasedon‘If-Then’decisionrules,enteredbytheuser.Wepresentthemodeland
describeitsfirstimplementationfortheevaluationoffourcroppingsystemsgeneratedfromexpertknowledge,anddiscussitsrelevancetothe
objectivescitedabove.TheMASCmodelhasseveraladvantagesoverexistingmethods,duetoitsabilitytohandlequalitativeinformation,its
transparency,flexibilityandfeasibility.
croppingsystem/sustainabilityassessment/exante/MCDA/decisionrules/qualitativemulti-attributedecisionmodels/DEXi
1. INTRODUCTION 1996; Christen, 1998), all highlightingits holistic and multi-
dimensional nature and encompassing economic, social and
environmentalobjectives.
The agriculturalsectoris currentlyattheconfluenceof an
increasing number of new challenges such as market glob- Ever since the emergence of this concept, it has been
alization, food crisis, bio-fuels, environmentalconcerns, and claimedthattheassessmentofmultidimensionalsustainability
changes in legislation at the global and local scales that are isrequiredfortheimplementationofnewformsofsustainable
threateningthefutureofagriculturalsystemsinmanyareasof agriculture(Neher,1992;Schaller,1993;Vereijken,1997;den
theworld.Almosttwodecadesagotheforeseeableuncertain- Biggelaar and Suvedi, 2000; Pacini et al., 2003; Gafsi et al.,
tiesandthreatsconditioningthefutureofagriculturalsystems 2006).Thefirstoperationalstudiesconcludedthat,forsuchas-
resulted in the development of the agricultural sustainabil- sessmentstoberealistic,a‘methodologicalleap’wasneeded
ity concept, which has been defined in many ways (Hansen, (i) to integrate knowledge of different types relating to eco-
nomic, social and environmental objectives and (ii) to han-
*Correspondingauthor:[email protected] dle the conflicting aspects of these objectives (Munda et al.,
ArticlepublishedbyEDPSciences
448 W.Sadoketal.
1994; Dent et al., 1995). One alternative involved consid- aspectsofhisknowledge,includingempirical,non-formalas-
ering sustainability assessment as a typical decision-making pects.
problem that could be handled by multi-criteria decision-aid In a comparative review of the main MCDA families
(MCDA)methods(Mundaetal.,1994;Munda,2005).Asare- focusing on relevance for sustainability assessment, Sadok
sult,overthelasttenyears,anumberofMCDA-basedframe- etal.(2008)suggestedthat‘mixed’or‘nonclassical’MCDA
workshavebeendevelopedtofacilitatetheimplementationof methodswereparticularlyrelevantforhandlingtheabovecon-
sustainableformsofagriculture(Rossingetal.,1997;Zander straints,whilebeingabletoaddressmoreefficientlythemulti-
andKächele,1999;Loyceetal.,2002;Dogliottietal.,2004, dimensionalconstraints of sustainability assessment (i.e., in-
2005). comparability,non compensation and incommensurability of
dimensions). This is particularly true for decision rule-based
However,thecurrentpaceatwhichnewchallengesareap-
methods,whichallowtheuseofqualitativeinputinformation
pearingandshapingagriculturalsustainabilityandtheincreas-
and qualitative reasoning in the decision modeling process.
ing unpredictability of the driving forces behind them make
Programs using these methods — often in the form of deci-
it necessary to find alternative ways to address the questions
sion supportsystems (DSS) — are increasingly designed for
of the representation and assessment of sustainability. First,
use by more diverse decision-makers, with the provision of
these fluctuations make it increasingly necessary to develop
transparentcommunicationinterfaces(Matthiesetal., 2007).
fastex ante (i.e.,‘Before-the-event’)sustainabilityevaluation
Someofthese approacheshavebeenusedsuccessfullytoas-
frameworksthatcanrepresentandassess alargebodyofop-
sessspecificproblemsinagriculturalsystems,suchastheim-
tionsinsilicoandrapidlyidentifyinnovative,alternativesys-
pactofthesesystemsonsoilquality(Bohanecetal.,2007)or
temswithouttheneedforin-fieldassessmentsofallthepossi-
theecological/economicimpactofgeneticallymodifiedcrops
bleoptions(EuropeanCommission,2005;VanIttersumetal.,
(Bohanec et al., 2008). However, to our knowledge, the use
2007;Meynard,2008).
ofsuchapproachestoassessexplicitlyandsimultaneouslythe
Second,thereis a needto improvethe ‘granularity’ofthe social,economicandenvironmentaldimensionsofthesustain-
spatio-temporalscalesgenerallyusedinsustainabilityassess- abilityofagivenagriculturalsystem—inthiscasecropping
mentstudies,makingitpossibletotargetspecificsustainabil- systems—hasneverbeenreported.
ity issues more efficiently. The smallest spatiotemporal unit We describe here the MASC (for multi-attribute assess-
addressedinsustainabilityassessmentstudiesisgenerallythe ment of the sustainability of cropping systems) decision
farmingsystem(e.g.,Hänietal.,2003;Dogliottietal.,2005; model, which was developed within a decision support sys-
Meyer-Aurich,2005;Meuletal.,2008),whereasmuchofthe tem(DEXi),asa keyelementofa frameworkcombiningthe
impact of the environmentand the impact of certain socioe- ex ante generation and assessment of cropping systems. We
conomic factors on farms are deeply rooted within smaller presentanddiscussthedecisionmodelanditsfirstimplemen-
units,suchasthatofthecroppingsystem—thelevelatwhich tation for the assessment of fourcroppingsystems generated
theyaredetermined.Theseunits,definedbySebillotte(1990) fromexpertknowledge.
as ‘a set of management procedures applied to a given, uni-
formly treated area, which may be a field, part of a field or
agroupoffields’arerarelyconsideredinassessmentsofsus- 2. METHODS
tainability (Rosnoblet et al., 2006). Furthermore, in the few
casesinwhichthisscalehasbeenspecificallyaddressed(e.g., 2.1. Theassessmentframework
MazzettoandBonera,2003),thenumberofsustainabilitycri-
Theoverallassessmentframeworkconsistedofathree-step
teriaconsideredisstronglylimitedwithrespecttotheholistic
approach(Fig.1):
natureoftheconceptofsustainability.Amoredetailedassess-
mentofcroppingsystemsmaymakeitpossibletotargetand 1. Thegenerationofcroppingsystems,asinputsforthedeci-
improveeachofthesystemsmakingupthefarmingsystem,in sionmodel.Thisstepcanbeachievedwithspecialcomput-
aspecificmanner. ingtools(Dogliottietal.,2003;Loyceetal.,2002;Bergez
Third,manyoftheavailableMCDA-basedframeworksare et al., 2008), based on expert knowledge (Lançon et al.
restrictive in terms of the type of information required and 2008),orboth.
the types of decision-maker targeted. These restrictions on 2. Representationofeachofthegeneratedcroppingsystems
thetypeofinformationresultfromdiversecontextsthatcan- as a vector of sustainability criteria expressed in quali-
not be addressed quantitatively being overlooked, thus lim- tative terms and the processing of this vector by MASC
iting the range of innovative options assessed (Sadok et al., to obtaina syntheticqualitativeassessment of the overall
2008). The restrictions on the type of decision-maker arise (economic,socialandenvironmental)sustainabilityofthe
fromthe underlyingdecisionalproblembeing addressedin a croppingsystem.
mannerappropriateforonlyasmallsubsetofdecision-makers 3. Analysis and interpretation of the assessment results for
inmanyofthesestudies.Quantitativepreferencemodels,with theselectionofalternativesystemstobetestedinfieldtri-
formalisms that are ‘opaque’ or inaccessible to non-expert als.Ifthedecision-makersarenotsatisfiedwiththeresults,
decision-makers,areoftenused.Dentetal.(1995)suggested the framework offers the possibility of either ‘adjusting’
that an assessment method is likely to be successful only if certaindecisionrulesor,ifnecessary,testinganewsetof
it uses the languageof the decision-makerand represents all improvedcroppingsystems(Fig.1).
MASC,aqualitativemulti-attributedecisionmodelforexanteassessmentofthesustainabilityofcroppingsystems 449
INPUTS DECISION MODEL OUTPUTS
Generation of Representation through sustainability criteria Analysis and
cropping systems Processing of sustainability criteria information interpretation of results
Decision Support System (DEXi)
MASC Model
Sustainability Sustainability:
Cropping system #n
Cropping system #3 ECO SOC ENV CS#1 High
Cropping system #2 Medium
Cropping system #1
CS#2 &#3 Low
[CS#1]
CS#n Very Low
Regeneration DEX Methodology
Adjustment
Prototyping
OR OR
[Best CS]
Figure1.Structureoftheframeworkgeneratingandassessingexantethesustainabilityofcroppingsystems.Aftergeneration(insilicothrough
models,fromexpertknowledgeorboth),thecroppingsystems(inputs)arerepresentedasavectorofvaluesforsustainabilitycriteria,which
are processed by the MASC decision model, implemented within a decision support system (DEXi). Depending on the evaluation results
(outputs),acroppingsystemmaybe(i)selectedforprototypinginthefield(ii)re-submittedtothedecisionmodelwith‘adjusted’choicesin
theevaluationprocessor(iii)discarded/improved,possiblyrequiringanewgenerationprocess.
2.2. MethodologicalbasisofMASC dures based on multiplicative or additive formalisms that do
not realistically capture the multidimensional nature of sus-
MASC is based on DEX methodology, and is imple- tainability assessment and (ii) only quantitative information,
mentedwithin a decisionsupportsystem (DSS) called DEXi hampering the use of empirical, non quantifiable knowledge
(Bohanec,2003).Itcombinesahierarchicalmulti-attributede- that may be relevant for sustainability assessment (Munda
cision model and an expert system shell. Hierarchical multi- et al., 1994; Munda, 2005). By contrast, DEX methodology
attribute decision models are used in many classical MCDA isbasedon(i)qualitative(symbolic)attributesand(ii)utility
approaches, such as the AHP (Analytic Hierarchy Process) functionsbasedon‘If-Then’decisionrulesfortheaggregation
orotherMAUT/MAVT(Multi-AttributeUtility/ValueTheory) of these attributes. Such features make this approach highly
methods (Saaty, 1980; Clemen, 1996; Figueira et al., 2005). suitableforrealisticassessmentsofthesustainabilityofagri-
They make it possible to break a decisional problem (a root culturalsystems(Dentetal.,1995).
attribute)downintosmaller,lesscomplexsubproblemsrepre- A DEX-based evaluation process follows the following
sentedbyvariablesorattributes(X,Fig.2).Theseattributes pathway(Fig.2):
i
are organized hierarchically so that those at higher levels of
the hierarchy depend on those at lower levels. Attributes are (i) Avectorofinputattributesisidentifiedandthedecision
aggregatedthroughutility(oraggregation)functionsF,which problemisrepresentedthroughahierarchyofattributes;
determine the dependence of a given attribute (called aggre- (ii) Eachattributeofthe treeisscoredona qualitativescale
gate attribute) on its immediate descendantsin the hierarchy (e.g.,‘Low’,‘Medium’,‘High’);
(seeexampleinFig.2).Optionsoralternativestobeevaluated (iii) Decision rule-based utility functions are established for
are representedbya vectorofvalues(a) ofattributes, called eachaggregateattribute.InDEXi,atableismadeavail-
i
inputattributes. able for users to enter qualitative values for aggregate
TheaggregationprocedurefollowedbyDEXmethodology attributes (e.g., Y, Fig. 2), as a function of the qualita-
is based on the hierarchical multi-attribute representation of tive values of the attributes to be aggregated (e.g., X ,
4
decisionproblemsusedinmanyclassicalMAUT/MAVTmod- X and X ), thereby defining decision rules (e.g., IF
5 6
els, butdiffers fromthis representationin a numberof ways. X =‘medium’&X =‘verylow’&X =‘low’;THEN
4 5 6
MAUT/MAVT methods generally use (i) aggregation proce- Y = ‘low’).Once the table is filled with these rules, the
450 W.Sadoketal.
Figure2.Typicalstructureandfunctioningofadecisionrule-based,qualitativehierarchicalmulti-attributemodel(adaptedfromBohanecetal.,
2000).Theinsetinthetopleftcornerprovidesanexampleofthewayinwhichautilityfunctionisdevelopedinthedecisionsupportsystem
DEXi(seemaintextfordetails).
softwareprovidesuserswithautilityfunctionrepresent- its currentversion,the modelstructure,sustainability criteria
ing a compilationof their choicesexpressedin terms of andattributescalestructuresarepredefined.Aggregationrules
relativeweightings(i.e.,aggregationrules); areleftopenfordefinitionbytheuserin≈70%ofcases(with
(iv) For the evaluation of an option, users enter the corre- suggested thresholds),for adaptationof the model to the cli-
sponding qualitative values of the input attributes and a matic and agricultural context, with the remaining 30% pre-
qualitativeresultdependingonthe structureofthe deci- determined on the basis of expert knowledge (Sect. 2.4.2).
siontreeandthepredefinedutilityfunctionsisdisplayed The qualitative rating of sustainability criteria is left to the
(valueofY). decision-maker (with suggested guidelines, Sect. 2.4.1). The
steps leading to the development of MASC are outlined in
IntheDEXidecisionsupportsystem,structuressuchasthose thesectionbelow.
shownin Figure 2 are dynamic,in thatchangingthe qualita-
tivevalueofagivenattributeand/ormodifyingafewdecision
rulescanhaveanimmediateeffectontheoverallassessment.
2.3. Workgrouporganizationandmethodology
Thisapproachmakesitpossibletocarryout‘what-if’and‘se-
lectiveexplanation’andsensitivityanalyses.InMASC,input
attributesrepresentsustainabilitycriteriaandprovideaquali- ThedevelopmentofMASCandthefirstexamplesofitsim-
tativeratingofeachcroppingsystem(option).Theaggregate plementationresultfromthecollaborationofaworkgroupthat
attributesarerepresentedbythevariouslevelsofsustainability functionedovertwoyearsaccordingto thespecificorganiza-
assessmentsubproblemsaddressedbythedecisionalmodel.In tionalandmethodologicalprocedurespresentedbelow.
MASC,aqualitativemulti-attributedecisionmodelforexanteassessmentofthesustainabilityofcroppingsystems 451
2.3.1. Modeldevelopment 2.4. FeaturesoftheMASCmodel
TheoverallstructureoftheMASCmodelisshowninFig-
The developmentofDEX/DEXi-basedmodelsrequiresit- ure3.Itincludestwointerdependentcomponents:(i)aninput
erationofthefollowingfoursteps:(i)attributeidentification; informationprocessingunit(Fig.3A) and(ii) a decisiontree
(ii)attributestructuring;(iii)thedefinitionofqualitativemea- (Fig.3B).Thesetwocomponentsdefineadecisionrule-based,
surement scales for each attribute and (iv) the definition of qualitativemulti-attributemodel.
aggregation rules. Typically, such models require collabora-
tion between experts in the considered fields, who suggest
attributes and aggregation rules, and decision analysts, who 2.4.1. Inputinformationprocessing
carryouttheprocessanddefineahierarchicalstructureforthe
attributes(Bohanecetal.,2008). The input information (i.e., sustainability criteria) of the
modelisstructuredintoavectorof32inputattributes(Tab.I,
In our study, these four steps were carried out by eight
Fig. 3). This vector was developedaccording to the working
agronomists. Each contributed his or her own expertise and,
proceduresdescribedinSection2.3.Thevectorhadahetero-
when necessary, exchanges took place with a pool of about
geneousinitialinformationalstructure,combiningqualitative
25 experts in research and development who took part in a
and quantitative information. The input informationprocess-
previousprogram(ReauandLandé,2006;Reauetal.,2006).
ing phase (i) calculates/estimates the values of the input at-
Eachoftheeightagronomistsalsoactedasadecisionanalyst,
allowingthegrouptoexplore,compareandcontrastdifferent tributesbasedonthecorrespondingmethodology(Tab.I)and
(ii)homogenizesthevaluesbyconvertingthemallintoquali-
possible structuresforthe initialdecisionmodelandto iden-
tative(linguistic)variables,forprocessingbythedecisiontree
tify the most consensual and relevant model. Input attributes
wereselected/developedbytheworkgrouponthebasisoftwo (Fig.3A).
mainconsiderations:
(1)‘Thematicrelevance’,correspondingtotheabilityofthe
2.4.1.1.Initialcalculation/estimationoftheinput
attributestoaddressrealisticallyalltherelevantissuesunder-
attributes
lyingtheassessmentofsustainabilityatthescaleofthecrop-
pingsystem.Thisgoalwasachievedbycarryingoutaspecific
Table I summarizes the methods used to calculate the in-
literature review for each of the issues considered, to deter-
putattributes. Allthe calculations/estimationsare carriedout
mine whether an indicator or methodologyfor its estimation
for each year of the rotation and averaged, except for one
hadalreadybeenidentified.
approach (I-Phy indicators, see below for details). During
(2) ‘Operational relevance’, corresponding to the feasi-
this phase, we distinguishedbetween two main groupsof at-
bility of implementing the chosen set of criteria in a multi-
tributes,basedontheinitialtypeofinformation(Fig.3A).
criterion evaluation approachwithout biasing the assessment
Thefirstgroupconsistedof15inputattributesinitiallyesti-
results. This goal was achieved by following the recommen- matedquantitatively,basedonformalismsandmodelsadapted
dations of Keeney and Raiffa (1976), Maystre et al. (1994)
from previous studies (e.g., economic attributes, Contribu-
andBakeretal.(2002).We evaluatedtheextenttowhichthe tiontoLocalEmployment,NH Emissions),ordevelopeddi-
3
selectedcriteriawere:(i)exhaustiveandcomplete(i.e.includ- rectly from expert knowledge (e.g., Phosphorus Use Auton-
ingallgoals),(ii)operationalandmeaningfulintermsof the
omy,SprayedArea,Tab.I).
decision-maker’sunderstandingoftheimplicationsoftheas-
Where necessary, formalisms and models from published
sessment,(iii)abletodiscriminatebetweencroppingsystems,
studies were adapted to the scale of the cropping system.
(iv)nonredundant(asfaraspossible)and(v)realisticallyfew
Thiswasthecase,forexample,foreconomicattributes,such
enoughinnumbertokeepproblemdimensionsmanageable.
as Semi-Net Margin, which was adapted from the work of
Dogliotti et al. (2004) for the estimation of cropping system
profitability.Costsduetothestorage,maintenanceandrepair
ofmachinery,andinfrastructurecostsforbuildings,roadsand
2.3.2. Firstexamplesofimplementation fences—whicharenotrelevantatthecroppingsystemscale
— were not included in calculations of operating costs. Em-
pirical quantitative models were developed,based on the ex-
The four cropping systems used to test the MASC model pertiseoftheworkgroupandthepoolofconsultedexperts,if
wereinitiallydevelopedbasedontheknowledgeof25experts. nosatisfactoryapproachhadbeendescribedinpreviousstud-
One cropping system was designed to be representative of ies. Thiswasthe case forthe attributeestimatingthe levelof
thecroppingsystemscurrentlyprevailinginPicardy/northern PhosphorusUseAutonomyofthecroppingsystems,asmany
France (cropping system #1), whereas the other three repre- studieshaveshownthatphosphorusresourcesare notrenew-
sentedpossiblealternatives(croppingsystems#2,#3and#4; ableandgraduallydecrease(Steen,1998;Stewartetal.,2005).
Reau and Landé, 2006; Reau et al., 2006). The first MASC- The secondgroupconsisted of 17 inputattributes initially
basedassessmentprocessforthesesystemswascarriedoutby estimatedqualitatively,throughtwodifferentapproaches.The
thegroupofeightagronomists,actingasdecisionanalysts. first approach involved the use of indicators readily giving
452 W.Sadoketal.
Figure 3. Sustainability criteria information processing and aggregation by the MASC decision tree, allowing ex ante assessment of the
sustainabilityofcroppingsystems.A.Aftercalculation/estimation(seeTab.Ifordetails),initialunitsoftheinputattributes—eitherquantitative
(QT, black rectangles) or qualitative (QL, black rectangles with L for ‘linguistic’ and O for ‘ordinal’) — are homogenized into linguistic
values(e.g.,‘Low’,‘Medium’,‘High’)on3-to5-pointscales(representedbythenumberofgrayrectangleslocatedtotherightoftheinput
attributes).B.Theinputattributesarethenaggregatedthroughlevelsofaggregateattributes(roundedrectangles),wheretheblackandthethree
grayrectanglesrepresenttheoverallandone-dimensional(economic,socialorenvironmental)evaluationresults,respectively.Theaggregation
rulesareoftwotypes:(i)totallyfree(thickgraylines)whenaggregationisdependentonthespecificcontextofthecroppingsystemandon
the knowledge and views of the decision-maker; and (ii) predetermined (thick black lines) when aggregation isindependent of the context
anddepends insteadonfactual/expert knowledge(seedetailsinthetext).Numericalvaluesinthedecisiontreedisplayedingrayandblack
boxesrepresentweightingscorrespondingtotypes(i)and(ii),respectively.Theyrepresenttheaggregationrulesusedfortheevaluationoffour
examplesofcroppingsystems,theresultsofwhicharereportedinTableIIandFigure4.
qualitative-ordinal scores (‘O’ rectangles in Fig. 3A), in- qualitative-linguistic estimation (e.g., ‘Low’, ‘Medium’,
cluding certain indicators from the I-Phy/INDIGO method ‘High’) of attributes addressing issues for which no satisfac-
(Bockstaller et al., 1997; Bockstaller and Girardin, 2006; tory quantitative or qualitative model could be identified in
e.g., Surface Water and Ground Water pesticide losses, soil previousstudies(‘L’rectanglesinFig.3A).Thiswasthecase,
Organic Matter Content, Energy Consumption), and those for example, for the Water Use Autonomy attribute, which
stemming from the treatment frequency index approach estimates the extent to which water requirements depend on
(Champeaux, 2006; e.g., Health Risks, number of doses for themaintypeofwatersupplytothecroppingsystem.
Insecticides/Fungicides/Herbicides).In this group, the inclu-
sionofvaluesforI-Phyindicatorsattherotationscalewasnot
directlybasedonaveragingyearlyvalues(asforalltheother 2.4.1.2. Qualitative transformation: scale and value
attributes), and the specific developmentof a ‘satellite’ deci- choice
siontree(notshown)wasrequiredforcalculation.
A second approach involves the development of Once calculated or estimated, the initial values of all
bibliographic or expertise-based guidelines for the direct input attributes are transformed into a qualitative-linguistic
MASC,aqualitativemulti-attributedecisionmodelforexanteassessmentofthesustainabilityofcroppingsystems 453
TableI.SummaryofthedifferentknowledgesourcesusedasinputattributesinMASC.1 Inputattributename(attributeorderisthesameas
thatreportedinFig.3).2 Approachesusedtocalculate/estimatetheinputattributes.Quantitativeformalismsareeitherdescribedtextuallyor
mathematicallywhensimple(e.g.waterresourceattributesandeconomicattributes,respectively)orreferredtowhentoocomplextofitinthe
table(e.g.,someIndigoindicators,seecorrespondingreferences).Similarly,qualitativeapproachesareeitherdescribedintext(whennotbased
onspecificformalisms)orreferredtowhenbasedonpublishedindicators.Calculations/estimationsaremadeforeachyearoftherotationthen
averaged,exceptforI-Phyindicators(seetextfordetails).Unitsareindicatedwhennecessary.Unitsofqualitativeattributescanbeexpressed
on the basis of qualitative-ordinal scales (Indigo indicators) or qualitative-linguistic scales. 3 References used. a: Dogliotti et al. (2004); b:
Vilainetal.(2008);c:MSA(2006);d:Bockstalleretal.(1997),BockstallerandGirardin(2006);e:Spooretal.(2003),vandeZande(1991);f:
Champeaux(2006);X:expertise.Whennecessary,importedformalismswereadaptedtothescaleofthecroppingsystem(seetextfordetails).
Input Attribute1 Description / Calculation Method2 Reference3
Profitability SNM = GP - OC SNM: Semi-Net Margin (€ ha-1) a
GP: Gross Product (€ ha-1)
OC: Operational Costs (€ ha-1)
Independence EI = (1-DS/SNM) × 100 EI: Economic Independence (%) b
DS: Direct Subsidies (€ ha-1)
Efficiency EE = (1-OC/GP) × 100 EE: Ecomomic Efficiency (%) b
Specific Equipment Needs Qualitative estimation X
Contribution To Local NTW = EMWU / S NTW: Number of temporary workers ha-1 b
Employment EMWU: Number of external man work units
Physical Constraints Qualitative estimation based on: vibrations frequency, mouvements repetitiveness, frequency c
of heavy loads manipulation, allergies, noise levels
Number Of Crops Indicator of implementation complexity X
Number Of Specific Operations Indicator of implementation complexity f
Health Risks TFIHaz: Treatment Frequency Index calculated for hazardous products X
Surface Water I-PhySW Indigo Method (Qualitative/Ordinal score) d
Ground Water I-PhyGW Indigo Method (Qualitative/Ordinal score) d
NO3 Losses I-NO3 Indigo Method (kg NO3-N ha-1) d
Phosphorus Losses IP Indigo Method (Qualitative/Ordinal score) d
NH3 Emissions INH Indigo Method (kg NH3-N ha-1) d
3
N2O Emissions INO Indigo Method (kg N2O-N ha-1) d
2
Pesticide Emissions I-Phyair Indigo Method (Qualitative/Ordinal score) d
Compaction Risk Qualitative estimation based on texture class, mechanization level, soil wetness, soil work type e
Erosion Risk Qualitative estimation based on soil water status, crusting sensitivity, cover type, soil work type X
Organic Matter Content IMO Indigo Method (Qualitative/Ordinal score) d
Phosphorus Fertility IPsoil Indigo Method (Qualitative/Ordinal score) d
Dry Period Irrigation Needs Indicator of water resource conservation capacity (mm) X
Crop Water Needs Indicator of water resource conservation capacity (mm) X
Water Use Autonomy Qualitative estimation based on water supply type (rainfall, catchment basin, river, aquifer) X
Energy Consumption IEnergy Indigo Method (Qualitative/Ordinal score) d
Energetic Efficiency ENE = OE/IE × 100 ENE: Energetic Efficiency (%) X
OE: Output Energy (GJ)
IE: Input Energy (GJ)
Crop Phosphorus Needs Indicator of phosphorus resource conservation capacity (kg of P) X
Phosphorus Use Autonomy AMP = (1- MP/TP) × 100 AMP: Autonomy with respect to Mineral Phosphorus (%) X
MP: Mineral Phosphorus (kg)
TP: Total Phosphorus (kg)
Crop Diversity IDiv Indigo Method (Qualitative/Ordinal score) d
Sprayed Area PSA = (1- NTC/TCC) × 100 PSA: Proportion of Sprayed Area (%) X
NTC: Number of non-Treated Crop Cycles
TCC: Number of Total Crop Cycles
Insecticides TFIInsecticides TFI calculated for insecticides f
Fungicides TFIFungicides TFI calculated for fungicides f
Herbicides TFIHerbicides TFI calculated for herbicides f
454 W.Sadoketal.
appreciation, using three- to five-value qualitative scales de- economic, social and environmental sustainability) and
finedbytheworkgrouponthebasisofexpertiseanddepend- withinsustainabilitydimensions.
ingontheresolutionofthe attribute(Fig.3A).Typically,the
linguisticscales take the formof a ‘Low/Medium/High’pro- However,dependingonthespecificsocioeconomicandenvi-
ronmental context of the cropping system to be assessed, it
gression, with the addition of ‘Very Low’ or ‘Very High’ in
mayberelevanttoprovidethedecision-makerwithexpertise-
some cases, depending on the attribute (gray rectangles in
basedthresholdsfortheweightingstobeusedintheaggrega-
Fig. 3A). The transformation process may be of one of two
tionprocess.AnexampleofthissituationisprovidedinSec-
types, as a function of the initial type of transformation: (i)
tion2.5.
fromquantitativetolinguistic,(ii)fromordinal(qualitative)to
linguistic.Whentheinitialinformationislinguisticinnature, Predeterminedaggregationistypicaloftheaggregations
notransformationisneeded(e.g.,PhysicalConstraints,Com- developedfromexpertknowledgebytheworkgroup,andmay
pactionRisk, WaterUse Autonomy,Tab.I,Fig.3A).Sugges- beconsideredvalidregardlessofthecontextandcroppingsys-
tionsastothecorrespondencebetweeninitialinformationand tem (Fig. 3B, thick black lines andnumbersin black boxes).
linguisticscalesare madein somecases, especially forsome These aggregations define submodels (e.g. Water Conserva-
I-Phy/INDIGOindicators(e.g.,SurfaceWater,GroundWater, tionandBiodiversityConservationsubmodels),whichcanbe
Phosphorus Losses and Pesticide Emissions), which initially interpretedas‘syntheticindicators’processinginputattributes
displayordinaloutputsthatmustfirstbeinterpretedaccording andlinkingthemwiththedecisiontree.
tothescoregridprovidedbytheINDIGOframework(e.g.,0
to 10 for I-Phy, Tab. I), before transformation into linguistic
2.5. SpecificsettingsforthefirsttestofMASC
information. In other cases, as for economic attributes (e.g.,
Profitability, Independence and Efficiency), interpretation of
Inputinformationprocessingforthesesystemsfollowedthe
thequantitativeresultsandtheirtransformationintolinguistic
steps described in Section 2.4.1. The aggregation rules fol-
assertionsarelefttothedecision-maker’sdiscretion,because
lowed for this specific assessment (numbers in gray boxes
the interpretationprocessin these cases dependsmoreon lo-
in Fig. 3B) were generated through interactions between the
cal/specificconsiderationsandnorms. eightdecision-makers/analystsoftheworkgroup.
Thevaluesusedareoffsetfromthresholdaggregationrules,
whichweredefinedbytheworkgroupformostaggregations,
2.4.2. Thedecisiontree
by 15 to 33%. In the absence of sensitivity analysis, the use
The MASC decision tree (Fig. 3B) is a hierarchicalstruc- ofthesesettingsensuredthatthesustainabilityevaluationpro-
ture aggregating32 input attributes through 22 aggregate at- cess was not distorted or biased by excessive weighting val-
tributes,basedon‘If-Then’decisionrulesdefiningaggregation ues (e.g. completely abolishing a sustainability component,
rulesexpressedas‘weightings’(detailsinSect.2.2andFig.2), usingaweightingof0foroneofthethreemaindimensions).
for assessment of the overall sustainability of cropping sys- These threshold values were established from expert knowl-
tems(OverallSustainabilityattribute).Thehierarchicalstruc- edge,basedon:
ture shown in Figure 3B was developed from expert knowl-
(i) Localconsiderationsrelatingtothecroppingsystems(e.g.
edge(accordingtotheworkingprocessdescribedinSect.2.3)
giventheimportanceofsoilqualityproblemsinPicardy,
and illustrates the decision-modeling process underlying the
minimum thresholds for Physical Quality and Chemical
MASC model. As for input attributes, the scales for aggre-
Qualityattributesweresetat33%inthedecisionmodel)
gateattributesarelinguisticanddefinedonthebasisofexpert
and
knowledge,witha resolutionofthreetofiveunits(grayrect- (ii) Considerations relating to the importance of different
anglesbelow aggregateattributesin Fig. 3). The aggregation
issuesunderlyingsustainabilityingeneral(e.g.giventhe
processinMASCmaybefreeorpredetermined.
importance of the issues relating to the Environmental
Freeaggregationoccursin ≈ 70%ofthe casesin MASC Quality,AbioticResourcesConservationandBiodiversity
(thick gray branches in Fig. 3B). In these cases, the model Conservationattributes,theminimumthresholdvaluefor
does not specify the strict weighting values to be assigned aggregationwas set at 20% for each of these attributes).
to the attributes to be aggregated(numbersin gray boxesre- A similar rationale could be applied to the aggregation
fer to aggregation rules specific to the evaluation presented ofeconomic,socialandenvironmentalsustainabilitysub-
in Sect. 2.5). Instead, the choice is left open to the decision- models,forwhichminimumaggregationthresholdswere
maker,sotheaggregationofattributesdependson: setat15%.
(i) Personal strategic views (e.g., aggregation of economic
attributesorcertainsocialattributes);
3. RESULTSANDDISCUSSION
(ii) Local and specific environmental contexts and norms
(e.g., aggregation within Water Pollution Risks sub- 3.1. FirsttestofMASC:implicationforfuturereal-case
model); studies
(iii) The decision-makers’ broad representations of the sus-
tainability assessment problem for their cropping sys- The coherence of the MASC model was tested by assess-
tems,definingtheprioritiesbetween(e.g.,aggregationof ing four cropping systems generated from expert knowledge
MASC,aqualitativemulti-attributedecisionmodelforexanteassessmentofthesustainabilityofcroppingsystems 455
(see Sect. 2.3), through the specific steps described above. beaggregatedwithinaspecificallydesignedsustainabilityas-
The initial qualitative values of the input attributes, together sessmenttool.
withthoserelatingtoalltheaggregateattributesresultingfrom Classically, studies addressing sustainability assessment
combinationsoftheseinitialvalues,aregiveninTableII,with at larger scales, such as farming systems, benefit from
theaggregationrulesfollowedforthisassessmentreportedin a larger number of input criteria (e.g., frameworks of
Figure3B(numbersingrayandblackboxes). Phillis andAndriantiatsaholiniainan,2001; Häniet al., 2003;
A detailedinterpretationof the resultsobtainedwouldnot VanCauwenberghetal.,2007;Meuletal.,2008).Thus,forthe
be relevant in this study because the cropping systems used specificneedsoftheMASC,weeither(i)directly‘imported’
here were developed for testing the model, rather than for a criteriagenerallyconsideredathigherlevelsor(ii)simplygen-
realcomparisonofsustainability.Nonetheless,theevaluation eratedourspecificcriteriaasafunctionoftheissuestargeted,
results were consistent with the initial representation of the basedonacombinationofliteraturereviewandexpertknowl-
croppingsystems:asexpected,theprevailingcroppingsystem edge.
(CS#1)displayedhigheconomicsustainabilityandlow envi- Criteria were directly imported, for example, for certain
ronmental sustainability, and the 3 alternative cropping sys- economic sustainability factors considered by MASC. This
tems developed on the basis of expertise (CS#2, CS#3 and wasthecaseforthesetofattributesreflectingeconomicinde-
CS#4, Fig. 4) had higher values for environmentaland over- pendenceandefficiency,whichwereimportedfromtheIDEA
allsustainability,rangingfrom‘Medium’to‘VeryHigh’.The framework (Vilain et al., 2008), to reflect medium-term eco-
decisionrulesuseddidnotgenerate‘illogical’outputs,atleast nomicsustainabilityobjectives(inadditiontoshort-termprof-
forthistest,whichprovidesafirstapproximationofthecoher- itability,Fig.3,Tab.I).
ence of the logic behind the disaggregation-aggregation ap- Specificcriteriaweregeneratedincasesinwhichnoready-
proachinMASC.Thisoutcomepavesthewayforamorefor- to-useindicatorswereavailableatcropping-systemscale.This
malassessmentofrealisticcases. was the case for the ‘synthetic indicators’ of MASC (see
Sect. 2.4.2),representedby the submodelsaggregatedon the
basisof expertknowledgefor bothsocial(Complexity ofIm-
plementation)andenvironmental(WaterConservation,Phos-
3.2. Relevanceoftheholisticrepresentation
phorusConservationandBiodiversityConservation)sustain-
ofthesustainabilityconceptatthecropping
abilitydimensions.
systemscale
Classically,thesocialsustainabilitydimensioniseithernot
represented(Bohanec et al., 2008) or only implicitly consid-
MASC explicitlyand specifically functionsat the scale of eredinstudiestargetingcroppingsystems(e.g.,‘operatingdif-
the cropping system, and this is an important characteristic. ficulty’criterionasapartofatechnicalevaluation,Mazzetto
Thecroppingsystemscale hasa granularityintermediatebe- andBonera,2003). InMASC, wechoseto addressthisissue
tweenthoseofthesingleplot/singleyearandfarmingsystem more explicitly, by developing an indicator dealing with the
scales. The use of this scale makes it possible to isolate the Complexity of implementation in the context of the worker’s
effects on overall socioeconomic and environmental perfor- qualityoflifeand,thus,asacomponentofthesocialsustain-
mance of different cropping activities within a farming sys- ability of the cropping system. Furthermore, in line with the
tem, and to take interactions between individualcrops in the recommendationsofCalkeretal.(2007),weexplicitlyconsid-
processintoaccountinanexplicitmanner. eredthehealthofagriculturalworkersinourassessment.Two
A second key feature of MASC is that it explicitly for- complementarystandpoints,eachrepresentedbyasustainabil-
malizesaglobalconceptofsustainability,discretized anden- itycriterion,wereusedforthisanalysis:theactualconstraints
capsulated into a vector of 32 input criteria (Fig. 3, Tab. I). experiencedbytheworker,basedonaninterpretationgridpro-
AlthoughMASC addressesthe relativelylimitedscale of the videdbytheMSA (MutualitéSocialeAgricole,2006,Tab.I)
croppingsystem,acriticalobjectiveofthisstudywastomake andrisksduetocontactwithhazardoussubstances.
thismodelasholisticasreasonablypossible,byaddressingis- FortheWaterConservationattribute,wefocusedonwater
suesnotclassicallyaddressedatthisscalebutconsideredbya consumptionasafunctionoflocalcriticalperiodsandthetype
growingnumberof researchersto be ‘challengingtargetsfor of water supply, rather than on absolute water consumption
future agriculture’ (see Kirchmann and Thorvaldsson, 2000 values or water use efficiency-based factors, as the principal
for review). MASC may therefore be considered a “stand- aim was to estimate the local impact of the croppingsystem
alone” tool, which is not the case for other models analyzed ontheavailabilityofwaterasanincreasinglyrare(andshared)
insimilarstudiesassessingcroppingsystems.Indeed,compa- resource.SimilarapproacheswerefollowedintheRISE(Häni
rablestudieshaveeithertargeted(i)specificobjectivesunder- et al., 2003) and MOTIFS (Meul et al., 2008) sustainabil-
lying sustainability, such as the impact of cropping systems ity assessment frameworks,althoughneither of these models
on soil quality (Bohanecet al., 2007), and the economicand explicitlytookintoaccounttheoccurrenceofcriticalperiods
ecological impacts of genetically modified crops (Bt maize; of water shortage,which are becomingincreasinglyfrequent
Bohanecetal.,2008)or(ii)overallsustainability,butthrough inmanytemperateareasofEurope,andelsewhere.Thesame
a limited numberof sustainability criteria, as in the study by type of reasoning was applied to Phosphorus Conservation.
Mazzetto and Bonera (2003), who considered 9 criteria en- Phosphorususeistakenintoaccountinmanysustainabilityas-
compassing economic,environmentaland technicalissues to sessmentframeworks,mostlyintermsoflosses,useefficiency