Agron.Sustain.Dev.28(2008)163–174 Availableonlineat: (cid:2)c INRA,EDPSciences,2007 www.agronomy-journal.org DOI:10.1051/agro:2007043 Review article Ex ante assessment of the sustainability of alternative cropping systems: implications for using multi-criteria decision-aid methods. A review adok ngevin ergez ockstaller olomb WalidS 1,FrédériqueA 2,Jacques-ÉricB 3,ChristianB 4,BrunoC 3, uichard eau ore LaurenceG 1,RaymondR 1,ThierryD ´1* 1INRA,UMR211INRA/AgroParisTech,BP01,78850Thiverval-Grignon,France 2INRA,UAREco-Innov,BP01,78850Thiverval-Grignon,France 3INRA,UMR1248AGIRINRA/ENSAT,BP52627,Auzeville,31326CastanetTolosan,France 4INRA,UMRINPL-(ENSAIA)-INRANancy-Colmar,BP20507,68021Colmar,France (Accepted10September2007) Abstract–Sustainabilityisaholisticandcomplexmulti-dimensionalconceptencompassingeconomic,socialandenvironmentalissues,and itsassessment isakeystepintheimplementationofsustainableagriculturalsystems.Realisticassessmentsofsustainabilityrequire:(1)the integration of diverse information concerning economic, social and environmental objectives; and (2) the handling of conflicting aspects of theseobjectivesasafunctionoftheviewsandopinionsoftheindividualsinvolvedintheassessmentprocess.Theassessmentofsustainability isthereforeincreasinglyregardedasatypicaldecision-making problemthatcouldbehandled bymulti-criteriadecision-aid (MCDA)meth- ods. However, the number and variability of MCDAmethods are continually increasing, and these methods are not all equally relevant for sustainabilityassessment.Thedemandsforsuchapproachesarealsorapidlychanging,andfasterexanteassessmentapproachesarerequired, toaddress scalescurrently insufficiently dealt with, suchascropping system level.Researchers regularlycarry out comparative analyses of MCDAmethodsandproposeguidelinesfortheselectionofapriorirelevantmethodsfortheassessmentproblemconsidered.However,many oftheselectioncriteriausedarebasedontechnical/operationalassumptionsthathavelittletodowiththespecificitiesofexantesustainability assessmentofalternativecroppingsystems.WeattemptheretoprovideareasonedcomparativereviewofthemaingroupsofMCDAmethods, basedonconsiderationsrelatedtothosespecificities.Thefollowingmainguidelinesemergefromourdiscussionofthesemethods:(1)decision rule-basedandoutrankingqualitativeMCDAmethodsshouldbepreferred;(2)differentMCDAtoolsshouldbeusedsimultaneously,making itpossibletoevaluateandcomparetheresultsobtained;and(3)arelevantlystructuredgroupofdecision-makersshouldbeestablishedforthe selectionoftoolvariantsofthechoosenMCDAmethods,thedesign/choiceofsustainabilitycriteria,andtheanalysisandinterpretationofthe evaluationresults. multi-criteriadecisionaid/croppingsystem/sustainabilityassessment/qualitativeinformation/decisionrules/outrankingqualitative methods 1. INTRODUCTION integrated while handling a mixture of multiple long-, short- term, interacting and potentially conflictinggoals, depending TheprecisemeaningofSustainableAgricultureisfarfrom onthescaleonwhichsustainabilityisconsidered(farm,land- clear(e.g.,Hansen,1996;SmithandMcDonald,1998;Pannell scape, region, nation, group of nations or global; Kruseman andSchilizzi,1999;RigbyandCaceres,2001),buteffortshave etal.,1996;Meyer-Aurich,2005). been made to produce an integrated definition of this term. Assessingthesustainabilityofagriculturalsystemsisakey AccordingtoIkerd(1993),SustainableAgricultureshouldbe issue for the implementation of policies and practices aim- capable of maintaining its productivity and usefulness to so- ing at revealing sustainable forms of land use (Neher, 1992; ciety in the long term. This implies that it must be environ- Sulser et al., 2001; Pacini et al., 2003). However, if they are mentallysound,resource-conserving,economicallyviableand to be realistic and effective, such assessments must handle socially supportive. Based on this definition, economic, en- thecomplexityoftheconceptof“sustainability”,asdescribed vironmental and social objectives should be analyzed as the above, whilst taking personal and subjective views concern- principaldimensionsofsustainability whensustainable prac- ing the relative importance of priorities into account (Dent ticesareimplementedinagivenagriculturalsystem(Schaller, et al., 1995; Park and Seaton, 1996; Andreoli and Tellarini, 1993;Vereijken,1997;denBiggelaarandSuvedi,2000;Gafsi 2000). The assessment of sustainability is therefore increas- etal.,2006).Iftheseobjectivesaretobeconsideredtogether, ingly regarded as a typical decision-making problem, lead- thenknowledgeandresearchfromrelevantdisciplinesmustbe ingtothedevelopment,bysomeresearchers,ofsustainability *Correspondingauthor:[email protected] assessment decision-aid methods. Most of these approaches Article published by EDP Sciences and available at http://www.agronomy-journal.org or http://dx.doi.org/10.1051/agro:2007043 164 W.Sadoketal. are based on multi-criteria decision-aid (or making) methods andprobablyasmanyselectioncriteriawhicharesetinamore (MCDAorMCDM),andsomehaveresultedinprototypesus- technicalbackground.Asecondaimofthestudyisthustopro- tainablesolutionsinthefield(Rossingetal.,1997;Zanderand videsuggestionsregardingtheparticipatoryprocesstobefol- Kächele,1999;Loyceetal.,2002a,b;Dogliottietal.,2005). lowedbythe decision-makers,starting fromthe finalMCDA However,inpractice,suchassessmentsareconfrontedwith toolchoicetotheanalysis/interpretationoftheexanteassess- two major problems. Firstly, the number of MCDA meth- mentresults. ods and tools available is continually increasing (Bouyssou etal.,1993,2000,2006),andstudiesaimingtoassessthesus- tainability of agricultural systems rarely justify clearly their 2. OVERVIEW AND TAXONOMY OF MULTIPLE choiceofoneMCDAmethodoveranother.Onlyafewstudies CRITERIADECISION-AIDMETHODS have presented a comparative, or at least exploratory, evalu- ation of the principal MCDA approaches available, in terms Multiple-criteria decision aid (MCDA) is a research area of the relevance to the purposes of the assessment. In con- within the field of decision analysis (DA), which aims to de- trast, many authors have concluded that, in typical decision- velop methodsand tools to assist with decision-making,par- aid problems, there is rarely one ideal method and a group ticularlyin termsof the choice,rankingor sortingof options of MCDA methods should therefore be applied (Salminen (i.e.,alternatives,solutions,coursesofaction,etc.),inthepres- etal.,1998;Zanakisetal.,1998;Macharisetal.,2004;Wang enceofmultiple,andoftenconflictingcriteria(Zanakisetal., andTriantaphyllou,2006).Moreover,theguidelinesemerging 1998;Figueiraetal.,2005).MCDAmethodologycanbeseen from comparative studies are generally developed within the as a non-linear recursive process including four main steps: operationalresearchcommunity,basedontechnicallyoriented (i)structuringthedecisionproblem,(ii)articulatingandmod- argumentsandcriteriafromthisfieldofresearch(seeGuitouni eling the preferences,(iii) aggregatingthe alternativeevalua- and Martel, 1998for review)withoutconsideringconstraints tions (preferences), and (iv) making recommendations (Roy, relatedtotheapplicationdomain.Itshouldalsobenoticedthat 1985;Maystreetal.,1994). althoughsomegeneralguidelines,concerningspecificfeatures MCDAmethodshavedevelopedconsiderablyoverthelast of sustainability assessment in most cases, have been devel- 30 years, resulting in a large number of methods and tools oped(RehmanandRomero,1993;Mundaetal., 1994,1995; (Figueiraetal.,2005).Thishasresultedinaneedforthecom- Munda,2005),theyarestillrarelyfollowedexplicitlyinreal- parison of MCDA methods, to identify the most appropriate casecontexts. methodsforthedecisionalproblemconsidered(Zanakisetal., Secondly, demand is increasing among farmers’ groups 1998;BrunnerandStarkl,2004).Manyauthorshavestressed and policy-makers for more innovative sustainability assess- theneedforataxonomyofMCDAmethods,asastartingpoint ment, highlighting a need for (i) faster ex ante assessment for the selection process (MacCrimmon, 1973; Hwang and approaches for rapidly identifying alternative systems with- Yoon, 1981). Dozens of taxonomies are currently available, out assessing the entire initial systems in the field (European basedonanumberofcriteria,including: Commission,2005;VanIttersumetal.,2007),and(ii)theex- – Thenumberofalternativesconsidered:discretevs.contin- pansionofsustainabilityassessmenttoscalesrarelystudiedat uousdistribution of alternatives(Schärlig, 1985;Maystre themoment,suchasthecroppingsystemscale. Indeed,most etal.,1994); published studies have been carried out on a plot scale or on – Informationmeasurementlevelofcriteria–qualitativevs. anevenlargerscale:farm,landscape,stateornation(Bontkes quantitative, and the level of uncertainty (Munda et al., andvanKeulen,2003;Meyer-Aurich,2005).Acroppingsys- 1994,1995); tem consists of a set of management procedures applied to – The methods used to construct the preference model: a given, uniformly treated agricultural area, which may be a mathematical decision analysis approach vs. artificial field, part of a field or a groupof fields (Sebillotte, 1990).A intelligence approach (Nijkamp and Vindigni, 1998; givenfarmingsystem may thereforebecomposedofa group Figueiraetal.,2005); of cropping systems, the sustainability assessment of which – The criteria aggregation mode: complete, partial or local is potentially relevant, as they represent different, uniformly aggregation(Schärlig,1985;Maystreetal.,1994;Vincke, treated units. However,few publishedstudies have described 1989); sustainabilityassessmentexplicitlyatthelevelofthecropping – Thedegreeofcompensationbetweenthecriteria(Hayashi, system, with a given MCDA method (Mazzetto and Bonera, 2000); 2003),andthosedealingwithassessmentsatthislevelcarried – Thedescriptive,prescriptive,constructiveornormativena- outnoinitialcomparativeassessmentofMCDAmethods. tureofdecision-making(Bouyssouetal.,2006). Themajoraimofthepaperistoprovideacomparativere- viewofthemainfamiliesofMCDAmethods,basedoncriteria One of the most integrative taxonomies was established relatedtothespecificitiesofthesustainabilityassessment,for by Hwang and Yoon (1981). This taxonomy distinguished the a prioriselection of groupsof candidateMCDA methods between multiple-objective decision-making (MODM) and forexanteassessmentofthesustainabilityofalternativecrop- multiple-attributedecision-making(MADM)methods,within ping systems. Fine-tuning the selection process to the scale the MCDA area. MODM methods can be used in cases in of individual methods is beyond the scope of our review, as whichthereareaninfinite(continuous)orlargenumberofal- therearedozensofalgorithms/toolsavailableintheliterature ternatives.Theyarebasedonmultiple-objectivemathematical Exanteassessmentofthesustainabilityofalternativecroppingsystems 165 programing models, in which a set of conflicting objectives usedtotranslatetheresponseintoascorefrom1to9.All isoptimizedandsubjectedtoasetofmathematicallydefined pair-wise comparisonsbetween single objectsare used to constraints, for selection of the “best” alternative. MADM constituteapair-wisecomparisonmatrix. methodsare used in cases of discrete, limited numbersof al- 3. Checkingtheconsistencyofthematrixandderivingprior- ternatives,characterizedbymultipleconflictingattributes(cri- itiesfromit. teria). They are based on (i) the aggregation of judgments 4. Aggregationofcriteria,withthehelpofagivenadditiveor for each criterion and alternative, and (ii) the ranking of multiplicativeutilityfunction. the alternatives according to the aggregation rules. MCDA, as used in many published studies, generally refers only to MADM, mainly because of the great number of methods of 2.2. Outrankingmethods thistypeavailable.Indeed,areviewoftheliteraturespanning the last 25 years revealed an increasing number of new and Outranking methods are based on social choice theory. hybridMADM methods,leadingto a greatvariabilityin tax- Thesemethodslacktheaxiomaticbasisofmulti-attributeutil- onomies(Schärlig, 1985;Roy, 1985;Vincke, 1989;Nijkamp ity methods, but are useful in practice (Guitouniand Martel, et al., 1990; Roy and Bouyssou, 1993; Maystre et al., 1994; 1998).“Outranking”isaconceptoriginallydevelopedbyRoy Bouyssouetal.,2000,2006;Figueiraetal.,2005).Neverthe- (1985). It involves comparisons between every possible pair less, a synthesis of these taxonomies revealed that a major- of options considered, to define binary relationships, such as ityofthemostusedMADMmethodscanfallintooneofthe “alternativeaisatleastasgoodasalternativeb”.Procedures followingthree categories:(i) multi-attribute utility methods, based on outranking have two phases. Decision-makers first (ii)outrankingmethods,and(iii)mixedmethods.Thebound- provideinformationabouttheirpreferencesforindividualcri- ariesofthe latter remainfuzzyin thereviewedliteratureto a teria, in the form of indifference and preference thresholds. pointthatledustoprovideourownunderstandingoftheterm Partialbinaryrelationshipsarethencalculatedforallcriteria, (seeSect.3.1.fordiscussion). taking into account the inter-criterion preferences expressed in terms of weightings denoting relative importance. These weightingsdonotrepresentatrade-offbetweencriteriascores 2.1. Multi-attributeutilitymethods (as in MAUT-based methods), as they are used to combine Thesemethodsareessentiallybasedonmulti-attributeutil- preferencerelationshipsratherthanscoresofalternatives.The itytheory(MAUT,KeeneyandRaiffa,1976),whichemerges ELECTRE method (Élimination et choix traduisant la réal- from the philosophical doctrine of Utilitarism. If the deci- ité; Roy, 1968) was the first to use an outranking approach. sion is made in conditionsin which the attributes are known It was followed by many others, including different versions with certainty (deterministic approach), the term “utility” is of ELECTRE (II, III, IV, IS and TRI; Maystre et al., 1994) replaced by “value” (MAVT). The term “utility” is preferred andthePROMETHEEmethods(preferencerankingorganiza- to indicate that the preferences of stakeholders against risk tion methodfor enrichmentevaluations;Brans, 1982).These areformallyincludedintheanalyticalprocedure.TheMAUT methodsarebasedondifferentpreferencestructures. method has three major steps: (i) normalization and evalua- tionoftheperformanceofeachalternativeintermsofitsutil- ity, (ii) identification of the weights statistically representing 2.3. Mixedmethods thedecision-maker’sprioritiesforeachcriterion,and(iii)ag- gregation(basedonadditive,multiplicative,orotherdistribu- ManyapproachesotherthantheMADMmethodsdescribed tionalformalisms,GuitouniandMartel,1998)andrankingof above have been proposed. Some have been referred to as thevariousalternatives. “non-classical” or mixed. There seems to be no common Theanalytichierarchyprocess(AHP)isanothermajorap- definition of these terms within the MCDA community (see proachfirstdevelopedbySaaty(1980),basedonthesameag- Mundaetal.,1994;Maystreetal.,1994;GuitouniandMartel gregationprinciplesasMAUT,butdiffersfromthelatterwith 1998;Figueiraetal.,2005forcomparativereview),butweun- respecttothewaythedecisionalproblemishandled.TheAHP derstandthesetermstocorrespondtoagroupofMADMmeth- comprisesfourmajorsteps: ods(i)abletohandlemixedqualitative-quantitativeorqualita- 1. Disaggregating a complex problem into a hierarchy, in tivecriteriainformationexplicitly,and/or(ii)withapreference whicheachlevelconsistsofspecificelements.Theoverall modeldifferentfromthoseclassicallyusedformulti-attribute objective of the decision lies at the top of the hierarchy, utilityandoutrankingmethods. and the criteria, sub-criteria and decision alternatives are Afirst majorgroupofmixed MADM methodsconsists of placedatdescendinglevelsinthishierarchy. outrankingapproacheshandlingqualitativeormixedinforma- 2. Pair-wise comparisons between all elements at the same tion (Munda et al., 1994; Guitouni and Martel, 1998). There level, based on a method converting verbal and subjec- are many variants among this group, such as the REGIME tive assessments into a set of overall scores or weights. methods (Nijkamp et al., 1990), QUALIFLEX (Paelink, Theconversiondependsonthe decision-maker’sanswers 1978), ORESTE (Roubens, 1982), EVAMIX (Voogd, 1983), to questions of the general form: “How important is cri- MELCHIOR (Leclerc, 1984) and ARGUS (de Keyser and terion A relative to criterion B?” A verbal scale is then Peters,1994). 166 W.Sadoketal. Asecondgroupconsistsofdecisionrule-basedapproaches, (ii) Thesemethodsaremostlyrequiredincasesinwhich“in- which are often generically referred to as ‘expert systems’ finite” alternativesmustbe assessed to identify the “opti- (Kim et al., 1990). These methods were initially developed mal”option(Steuer,1986;Zhouetal.,2006).Inourcase, for the assessment of complex situations that cannot be han- wearemorelikelytobeassessingafinitenumberofalter- dled throughpreferencemodelsbased onconventionalmath- nativecroppingsystems,rankingthemintermsoftheirpo- ematical tools (means, sums, simple weighting and complex tentialsustainability.Suchrankingsallowamoreextensive models; Tixier et al., 2007). In these methods, the prefer- comparativeanalysisoftheoutputsofdifferentassessment ence model can be constructed through learning from exam- methods,potentiallyidentifyingpromisingalternativesnot ples.Theglobalpreferenceisdefinedbysortingtheobjectsof initiallyhighlyranked. analysisintopredefinedcategories(e.g.acceptance,rejection, etc.)throughasetoflogicalstatements,typicallyrepresenting Both these issues highlight the need for more appropriate “if/then”decisionrules,whichareoftenorganizedintheform andrealistic approachesto theexanteassessmentofsustain- of decision trees or decision tables. These decision rules are ability on the scale of the cropping system. The specificity formulatedonthebasisofexpertfactual-heuristicknowledge of the sustainability assessment problem, with the implied (derivedfrominterviewsandliterature)and/orwiththehelpof needforMCDAapproachesotherthanclassicalMODMmeth- data-miningandknowledgediscoverytools(Kimetal.,1990; ods, has already been highlighted by many authors (Voogd, Pawlak,1991;Zupanetal.,1999). 1983;Nijkampetal.,1990;Mundaetal.,1994;Nijkampand Vindigni, 1998). Below, we will define selection criteria for identifyingthemostrelevantofthesemethodsforourpurpose. 3. SELECTION OF MULTIPLE CRITERIA DECISION-AID METHODS FOR EX ANTE 3.2. CriteriaforselectingrelevantMADMmethods ASSESSMENT OF THE SUSTAINABILITY OF CROPPINGSYSTEMS Weconsiderheretwogroupsofcriteriaforidentifyingrele- vantapproachesfromthemanyMADMmethods,basedonthe 3.1. RelevanceofMODMmethods recommendationsofMundaetal.(1994,1995):(i)theability ofthesemethodstohandlethetypicalmulti-dimensionalchar- Mostof thedecision-aidapproachesdevelopedforassess- acteristicsofsustainabilityassessment,and(ii)theirabilityto ing the sustainability of agricultural systems have classically handlemixedmeasurementlevelsofcriteria. been based on multiple-objective decision-making methods The need for methods to handle multi-dimensional char- (MODM)(Meyer-Aurich,2005).Thesemethodsareoftenim- acteristics translates operationally into three requirements: plementedwithinsome“systemsapproach”frameworkscon- (i) incommensurability – an absence of the need for a sisting of (i) systematic and quantitative analysis of agricul- common measure aggregating several dimensions, (ii) non- turalsystemsforthemathematicaldefinitionofobjectivesand compensation – an advantage in one dimension of the eval- constraints, and (ii) the synthesis of optimal “solutions”, us- uationisnottotallyoffsetbyadisadvantage,and(iii)incom- ingoptimizationtechniques(Rossingetal.,1997;Zanderand parability – the method does not offer a single comparative Kächele, 1999; ten Berge et al., 2000; Kropff et al., 2001; termbywhichallalternativescouldberanked(Schärlig,1985; Hengsdijk and van Ittersum, 2002; Bontkes and van Keulen, Maystre et al., 1994; Stewart and Losa, 2003). In realistic 2003;Dogliottietal.,2005).Theexanteevaluationofinnova- evaluationsofthesustainabilityofagriculturalsystems,tack- tivecroppingsystemsustainabilityposestwomajorproblems ling environmental, social and economic dimensions, trans- forMODMmethodsbasedonoptimizationtechniques: latesthenintothefactthatstrongassumptionsaboutthecom- mensurability,compensationandcomparabilityofvaluesmay (i) These methods are known to be sensitive to missing, not be relevant, as the criteria for different sustainability di- inconsistent or mixed (quantitative and qualitative) data mensionsmayhavedifferentunitswithlowlevelsoftrade-off (Dent et al., 1995;Weersink et al., 2002;Dogliotti et al., (O’Neill, 1997; Martinez-Alier et al., 1998). The notions of 2005). In typical ex ante assessments of sustainability – commensurability,compensationandcomparabilityareinter- particularly on the croppingsystem scale, the assessment connected,inthatstrongcommensurabilityimpliesfullcom- of which has not been extensively documented– there is pensation of criteria and a high level of comparability of the unlikely to be sufficient scientific and/or expert quantita- actionsconsidered.Theyarethereforenotconsideredhereto tive knowledge available. Furthermore, as innovative de- beindependentselectioncriteria. mands cannot generally be systematically translated into The second recommendation concerns the ability of these scientificand/orquantitativedata,theuseofqualitativein- methodstohandleheterogeneousmeasurementlevelsofcrite- formationintheassessment processislikelytobeneces- riainformation(i.e.,quantitativevs.qualitative)andtheirun- sary.Inaddition,theuseofqualitativedataisincreasingly certainty(Mundaetal.,1994,1995).Wewillnotconsiderthe considered a rule rather than an exception for the realis- abilityofMADMmethodstohandleuncertaintyinthisreview. tic assessment of the holistic environmental and socioe- Thiscriteriondoesnotseemtobediscriminatory,asalmostall conomic issues underlying sustainability (Maystre et al., MADM and MODM methods can be linked to a procedure 1994;Mundaetal.,1995). handling fuzzy or stochastic uncertainty (Chen and Huang, Exanteassessmentofthesustainabilityofalternativecroppingsystems 167 MMAADDMM mmeetthhooddss MMuullttii--aattttrriibbuuttee uuttiilliittyy OOuuttrraannkkiinngg MMiixxeedd mmeetthhooddss mmeetthhooddss mmeetthhooddss MMAAUUTT//MMAAVVTT AAHHPP EELLEECCTTRREE PPRROOMMEETTHHEEEE DDeecciissiioonn rruullee-- OOuuttrraannkkiinngg mmeetthhooddss mmeetthhooddss mmeetthhooddss mmeetthhooddss bbaasseedd qquuaalliittaattiivvee SSeelleeccttiioonn ccrriitteerriiaa mmeetthhooddss mmeetthhooddss -- -- ++ ++ ++ ++ --TTaacckklliinngg iinnccoommmmeennssuurraabbiilliittyy,, nnoonn--ccoommppeennssaattiioonn,, iinnccoommppaarraabbiilliittyy ooff ddiimmeennssiioonnss -- --//++ -- -- ++ ++ --TTaacckklliinngg qquuaalliittaattiivvee oorr mmiixxeedd iinnffoorrmmaattiioonn ooff ccrriitteerriiaa -- ----//++ --//++ --//++ ++ ++ OOvveerraallll ssuuiittaabbiilliittyy Figure 1. Taxonomy of multiple-attribute decision-making (MADM) methods and selection criteria used for the identification of suitable approachesforexanteassessmentofthesustainabilityofalternativecroppingsystems. Graysymbols aregiven (–; --/+;/–/+;+)toindicatetheaprioriirrelevance, partial relevance or relevance, respectively, ofeach group of methods based on the selection criteriaconsidered. Overall assessments are given for each group of methods, inthe form of dark symbols (–;–/+;–/+;+)indicatingtheoverall levelofsuitability(ranging fromnon-suitable tosuitable)forexanteassessment ofthesustainability ofalternativecroppingsystems.MAUT:multi-attributeutilitytheory;MAVT:multi-attributevaluetheory;AHP:analytichierarchyprocess; ELECTRE:éliminationetchoixtraduisantlaréalité;PROMETHEE:preferencerankingorganizationmethodforenrichmentevaluations. 1992; Munda et al., 1995; Ertugrul Karsak, 2004). The suit- prioritizeinformationtoimprovedecisions(Alphonce,1997). abilityofeachoftheMADM methodsconsideredwillthere- This may explain why AHP is the most used MAUT-based fore be assessed for the interconnected characteristics of in- method for solving agro-environmentaldecisional problems, commensurability,non-compensationandincomparability,to- mainlyexante.Indeed,someauthorshaveusedAHPmodels getherwiththeirabilitytohandlequalitativeormixedcriteria to choose crops so as to determine the best allocation of re- explicitly.Based on these criteria, we discussin detailbelow sources(Alphonce,1997),toevaluatesoilproductivity(Zhang theassessmentprocessforeachofthe3groupsofconsidered etal.,2004)andtoassesstheenvironmental,economicandso- MADMmethodsidentifiedinSection1.Theresultsaresum- cialfactorsrelatingtotheadoptionofsilvopasturetechniques marizedinFigure1. insouth-centralFlorida(Shresthaetal.,2004).Otherauthors haveusedthismethodtorankalternativesforpreservingorin- creasingsocialbenefitsfromthesustainableuseofnaturalre- 3.2.1. Multi-attributeutilitymethods sources,suchasforestrymanagement(Schmoldt,2001),wet- ClassicalMAUT/MAVTmethodsarebasedon(i)atotally landmanagement(Herath,2004)andlandpreservation(Duke andAull-Hyde,2002). compensatoryaggregationofcriteria,and(ii)commensurable judgments,resultinginhighlevelsoftrade-offbetweencrite- The main drawbacks of the AHP method are potential in- ria. Consequently, it is difficult, with most of these methods, ternal inconsistency and the questionable theoretical basis of totakeintoaccounttheincommensurableandpartlycompen- the rigid 1–9 scale, together with the possibility of rank re- satory criteria that often underlie the dimensions of sustain- versalfollowingtheintroductionofanewalternative(French, ability in agricultural systems (Rehman and Romero, 1993). 1988;GoodwinandWright,1998;Macharisetal.,2004).Al- Furthermore, these methods do not take qualitative or mixed ternative methods, such as the MACBETH method (Bana e (qualitative and quantitative) criteria into account efficiently CostaandVansnick,1997)havebeendevelopedtoovercome and explicitly (Munda et al., 1994). Consequently, although some of these objections. However, one of the most often- some authors have reported the use of MAUT methods for cited objectionsto AHP methodsis the totally compensatory some agricultural and environmental assessments (Salminen aggregation procedure, resulting in a high level of trade-offs etal.,1998;Hayashi,2000),thesemodelsmaybeconsidered between criteria (Roy and Bouyssou, 1993; Macharis et al., inappropriateforassessmentofthesustainabilityofalternative 2004). Thus, as for other MAUT approaches, the use of an croppingsystems,accordingtoourobjectives(Fig.1). AHPmethodmaylimit,tosomeextent,thepossibilityoftak- AHP methods offer alternative advantages compared with ing into account incommensurable and partly compensatory classical multi-attribute utility methods, consisting of (i) the criteria, which often underlie the concept of “sustainability”, hierarchicaldecompositionofthedecisionalproblem,and(ii) as applied to agricultural systems. Moreover, although the theuseofsubjectiveandverbalexpressionstodefinethe rel- method uses verbal scales, it is not considered truly quali- ativeimportanceofthecriteria(Macharisetal.,2004).Some tative (Munda et al., 1994). Instead, it is described as semi- authorsconsidermostAHPmodelstobeabletohandlemiss- qualitative(Ayalewetal.,2005)orpurelyquantitative(Moffett ingquantitativedataandalackofprecision,basedonthejudg- and Sarkar, 2006). Consequently, attention must be paid to mentandexperienceofdecision-makers,makingitpossibleto theseadvantagesanddisadvantagesifanAHPmethodisused 168 W.Sadoketal. aloneforexanteevaluationofthemulti-dimensionalsustain- should be considered potentially relevant for ex ante assess- abilityofalternativecroppingsystems(Fig.1). ment of the sustainability of cropping systems. However, it is noteworthy that though many of these methods, and espe- ciallytheREGIMEapproach(Nijkampetal.,1990),arereg- 3.2.2. Outrankingmethods ularlyusedforenvironmentalplanningandmanagementpur- poses, such approaches are rarely applied in the agricultural sector.Nevertheless,thesuccessfuluseofaREGIMEmethod Thekeyfeatureofoutrankingmethods,duetothevaguede- in a real case for evaluation of the sustainability of agricul- terminationofpreferences,isthattheygivelowlevelsofcom- turallanduseintermsofenvironmental,economicandsocial parabilitybetweenoptions(i.e.,performancenotmeasuredon objectivesreportedbyHermanidesandNijkamp(1997)isan- the same cardinal scale), making it possible to deal with in- otherargumentinfavorofitsuseforexanteassessmentofthe commensurability(Maystreetal.,1994).Thisadvantagemay sustainabilityofcroppingsystems. accountfortheirwidespreaduseinagricultural-environmental sustainability evaluation frameworks, on different scales, for Decisionrule-basedMADMmethods dealingwithproblemsofchoicebetweenalternatives.Atpol- icy level, the ELECTRE and PROMETHEE methods have Thesenon-classicalmethodsareconsideredpotentiallyrel- been used to rank different projects for environmental con- evantfor the solutionof variousagriculturaldecisionalprob- servation and the sustainable use of resources for several lems (Dent et al., 1995). Indeed, the decision rules approach specific problems: irrigation management planning (Pillai (i)isintelligibleandusesthelanguageofthedecision-maker, and Raju, 1996), the conservation of multifunctional forests throughsymbolic qualitative variables, (ii) is based on trans- (Kangas et al., 2001), waste management (Salminen et al., parently expressed preference information based on the ob- 1998) and environmental quality (Rogers and Bruen, 1998). servations, views and opinions of the decision-maker, and On the farm and cropping system scales, outranking meth- (iii) offers the possibility of handlinginconsistencies in pref- ods have been used successfully more frequently than other erential information resulting from hesitation on the part of MADMmethodologiesforvariousassessments.VanHuylen- the decision-maker (Bontkes and van Keulen, 2003; Greco broeckandDamasco-Tagarino(1998)usedthePROMETHEE and Matarazzo, 2005). Moreover, decision rule-based meth- method to choose the best crops for the ideal cropping ods can be used for the explicit handling of totally non- timetable for the farmer. Arondel and Girardin (2000) used compensatorydecisionprocesses(Ma,2006),makingiteasier an ELECTRE method to sort cropping systems on the basis to tackle incomparability and incommensurability (O’Neill, oftheirimpactongroundwaterquality.Loyceetal.(2002a,b) 1997; Martinez-Alier et al., 1998; Stewart and Losa, 2003). developedanELECTRE-basedmethod(theBETHAsystem) This makes the decision rule approach more flexible for the forassessingwinterwheatmanagementplanswithrespecttoa modeling of the decision process, as it takes into account a set of conflictingeconomic,environmentalandtechnological largediversityofconsiderationsmuchmoregeneralthanthose requirements.MazzettoandBonera(2003)developedamulti- takenintoaccountbyallotherexistingclassicaldecisionmod- criteriasoftwarepackagederivedfromtheELECTREmethod elsusedwithintheMCDAarea(Figueiraetal.,2005).How- (MEACROS),withtheaimofidentifyingalternativecropping ever, according to Ma (2006), one of the main limitations of systemsmeetingasetoftechnical,economicandenvironmen- this approach is that, in some complex real-life situations, tal criteria. These exampleshighlightthe possibility of using toomanydecision rulesmay be requiredto representthe de- outrankingmethodsforexanteevaluationofthesustainability cisional problem, making this technique cumbersome. Con- ofcroppingsystems,basedontheirabilitytotackletheincom- versely, others would argue that in many real-life situations, mensurability, non-compensation and incomparability of the particularlythoseconcerningthedecisionsfacingfarmhouse- sustainabilitydimensionsmoreefficientlythanclassicalmulti- holds,suchapproachesarefarmorerealisticandpracticalthan attribute utility methods. However, it has never been clearly otherclassicalMCDAmethods(Dentetal.,1995).Inanycase, stated that these outranking methods – in their strictest defi- thelevelofcomplexityofthedecisionrulesprobablydepends nition – are compatible with the use of explicitly qualitative more on the specific features of the decisional problem con- ormixed(qualitative/quantitative)criteria.Consequently,this sideredthanontheapproachitself. possiblelimitationshouldbeborneinmindwhenselectingan In agriculturalcontexts, decision rules have been used for appropriateoutrankingMADMmethodfortheexanteassess- the development of agri-environmentalindicators for assess- mentofalternativecroppingsystems(Fig.1). ing the sustainability of cropping systems in terms of pesti- cide impact(van Der Werf and Zimmer, 1998;Ferraroet al., 2003; Tixier et al., 2007). Those authors combined their ex- 3.2.3. Mixedmethods pert decision rules with fuzzy logic to cope with uncertainty and to avoiding the effect of a knife-edge limit for a given attribute. On the landscape scale, “classical” expert methods Outrankingqualitativemethods havebeenusedtoassesssoilerosionrisks(Cerdanetal.,2002) Given the ability of these methods to tackle incommen- and biodiversity (Crist et al., 2000). Phillis and Andriantiat- surability, non-compensationand incomparability of the sus- saholiniainan (2001) have developed a more integrative con- tainabilitydimensionswhilehandlingqualitativecriteria,they ceptualmethodologybasedonfuzzyexpertdecisionrulesfor Exanteassessmentofthesustainabilityofalternativecroppingsystems 169 evaluating the sustainability of agricultural systems, accord- understanding of this term. For instance, some authors con- ing to their economic, ecological and social goals. However, sidered the REGIME methods to be mixed (Munda et al., this method did not focus explicitly on the cropping system 1995), whereas others considered them to be simple classi- scale,asitevaluatedfarmingsystemsonregionalandnational cal outranking methods (Brunner and Starkl, 2004). Some scales. Agronomy researchers have recently begun to make authors consider the EVAMIX approach to be a mixed out- use of expert tools initially designed for non-agriculturalas- ranking method (Munda et al., 1994), whereas others con- sessment purposes for assessing sustainability-related issues. sider this approachto be neither of the outrankingnorof the Forexample,Bohanecetal.(2004)testedandestablishedthe mixed type (Guitouni and Martel, 1998). Similar discrepan- a prioriusefulnessoftheexperttoolDEXiforevaluatingthe cieshavealsobeenobservedregardingAHPmethods,which ecological and economic sustainability of cropping systems are considered by some authors to be qualitative (Alphonce, based on genetically modified maize (Bt-corn). Such expert 1997), whereas others explicitly consider them to be quanti- tools may thus be relevant for ex ante evaluation of the sus- tative (Moffet and Sarkar, 2006). In each of these situations, tainabilityofcroppingsystems(Fig.1). ourclassificationisbasedonthepredominantviewexpressed As summarized in Figure 1, this review revealed that the in published studies, with particular weight given to classifi- most suitable MADM methodologies for ex ante assessment cations relating to agro-environmentalor environmental sus- ofthesustainabilityofalternativecroppingsystemsareofthe tainability assessment problems (Munda et al., 1994, 1995; “mixed”type (qualitative outrankingand decision rule-based NijkampandVindigni,1998). methods), followed by outranking methods, and then AHP methods,basedoncriteriapresentedatthebeginningofSec- tion3.2. 4.2. RelevanceoftheconsideredMCDAtaxonomyand selectioncriteria 4. GENERALDISCUSSION As recommended by Zanakis et al. (1998), our selection was based on a taxonomy serving more as a tool for elim- 4.1. Bibliographic survey and selection of MCDA ination than for selection of “the right method”. Moreover, methods:difficultiesencountered ratherthanusingselectioncriteriabasedexclusivelyontech- nical/operational assumptions, we based our criteria on as- Inthiswork,wehavediscussedthepre-selectionoffamilies sumptions derived from more realistic situations, reflecting ofMCDAmethodsaccordingtotheirrelevancefortheexante the specific features of the sustainability assessment, as rec- assessment of the sustainability of alternative cropping sys- ommendedbyMundaetal.(1994,1995).Thesecriteriawere tems. We had an idea concerningthe strategy to be followed thentranslatedintomoretechnicalcriteria(incommensurabil- – identification of a relevanttaxonomyof methodsand of an ity, incomparability, non-compensation; mixed information). appropriatesetofselectioncriteria–buttheselectionprocess Though the considered taxonomy and selection criteria are wasnonethelesslaborious.Thelaboriousnessofpre-selection linkedtothespecificpurposeofourassessment,thesetwoini- may explain why so few real-case studies include a compar- tialstepsmightserveasguidelinesforsimilarcases. ative or explorativeevaluation of the main groupsof MCDA approaches available for the specific purposes of the assess- ment(Zanakisetal.,1998;Hayashi,2000). 4.2.1. MADMversusMODM Themaindifficultyinthisprocesswastheidentificationof a relevanttaxonomyof MCDA methods. A reviewof the lit- In our process for selecting potentially relevant MCDA erature published on MCDA over the last 25 years revealed methodsfor ex ante assessment of the sustainability of alter- thatthisresearchfieldhasincreasedindiversityandcomplex- native croppingsystems, we first considered one of the most ity,leadingtoanincreasingnumberofnewandhybridmeth- integrative taxonomies within the MCDA area (MODM vs. ods, resulting in turn in a large number of taxonomies (see MADM),toexcludethelargestpossiblegroupofmethods(see Roy, 1985 and Figueira et al., 2005 for review). The result Sect.3.1). was that in work aiming at selecting relevant MCDA meth- At that stage of MCDA method selection, we were con- ods,theconsideredtaxonomywasoftenfoundnottobeinde- fronted with two opposite approacheswithin the agricultural pendentoftheviewsoftheauthorsandthe specificpurposes sustainabilityresearchcommunity.UsersofMODMmethods of the assessment. This was also the case for more “concep- claimthatonlysuchquantitativemethodscandisentanglethe tual” studies proposing formalized typological tree or expert complex relationships between agricultural production, envi- systemapproachesfortheselectionofrelevantMCDAmeth- ronmentandeconomy,therebyincreasingthetransparencyof ods, while initially based on a specifically established taxon- choicesregardingsustainability(HengsdijkandvanIttersum, omy and thus not appropriate for systematic generalization 2002). Similarly, others even consider that the use of expert (Jelassi and Ozemoy,1988;Guitouniand Martel, 1998).An- rulesandsemi-quantitativeindicatorsinsuchstudiesiscause otherdifficultywas thatsome of these taxonomieswere con- for concern as it is difficult to evaluate such rules, rendering flicting andeven,in somecases, contradictory.Thiswas par- the results of local relevance at best, whereas MODM meth- ticularly true for the “mixed” category, the characteristics of ods are more effective (Dogliotti et al., 2005). Conversely, which were highly variable, according to the authors’ own someauthorsconsiderthatrealisticassessmentoftheholistic 170 W.Sadoketal. anduncertainissuesofsustainabilityrequiresamethodcapa- through fuzzy sets could match the real fuzziness of percep- ble of handling qualitative information (Munda et al., 1994, tionsthathumanstypicallydisplaywithrespecttothecompo- 1995; Hermanides and Nijkamp 1997; Phillis and Andrianti- nentsofdecision problems,and (ii)meansof calibratingand atsaholiniainan, 2001).Based on these elements and the par- manipulatingfuzzyfunctionswithatransparentrationalefrom ticular features of ex ante assessment of the sustainability of the point of view of non-specialists are lacking (UK DTLR, alternativecroppingsystems,asstatedinSection2.1,wehave 2001). thereforediscardedMODMmethodsintheselectionprocess. Some conceptual selection criteria used in some tax- However,itshouldbepointedoutthat,althoughrejectedin onomies were not considered in this work. For instance, we thisstudy,someMODMmethodshavebeenusedforexante did not consider the mode of decision-making, which dis- assessments of alternative farming systems with respect to tinguishes between normative (postulation), descriptive (ob- sustainability-related objectives (e.g., Dogliotti et al., 2004, servation), prescriptive (unveiling) or constructive (reaching 2005; Tré and Lowenberg-Deboer,2005). In these cases, the a consensus) methods (Bouyssou et al., 2006). Indeed, some innovative aspect of these systems consisted of the design authorshave expressed the viewthat thisclassification is not of sustainable production activities, based on the optimiza- really discriminatory, as in practice, methods initially con- tion of an innovative combination of a limited set of quan- sidered normative may be used in a constructive, descrip- titatively measurablecriteria representinginputsand outputs. tive or prescriptive manner, depending on the context in Thisisquitedifferentfromconsideringinnovativesustainabil- which they are applied (Dias and Tsoukias, 2003). Never- ityissuesandobjectivesdirectlytranslatedintoinnovativecri- theless, to some extent, we have considered (besides crite- teria, some of which cannot be measured quantitatively. For riaspecifictoexantesustainabilityassessmentrequirements) instance,thiswouldbethecaseforcriteriarelatedto(i)holis- a decision-making mode-basedselection approachwhen dis- ticissuessuchasbiodiversity,or(ii)subjectiveconsiderations carding MODM methods in favor of MADM ones. Indeed, such as social wellbeing, which are not taken into account theformeraimatreachingoneoptimal(normative)‘solution’ in those quoted studies using a MODM method. However, (i.e., a cropping system) whereas the latter allow for relative this does not mean that optimization approaches are neces- rankingof differentones(see Sect. 2.1),which fitted ourob- sarily unsuitable for purposes similar to ours. Indeed, within jectivesmuchmore. the mathematical programing area, optimization algorithms Other reported selection criteria based on operational as- abletohandlequalitativecriteriahavealreadybeendeveloped sumptions, such as transparency, ease of use, profile of the (Brewka, 2006). With new developments continually occur- decision-maker and number of decision-makers, were not ringintheMCDAfield,thesealgorithmsarelikelytobeinte- takenintoaccountinourreview,asweconsiderthat(i)some gratedintoMODMmethodsinthenearfuture,makingitpos- oftheseconsiderationsdependmoreonthecontrolandreport- sibleforthesemethodstohandlequalitativedata.Ourdecision ingcapabilitiesofthecorrespondingsoftware/toolthanonthe torejectMODMmethodsregardingouraimsandthepresent method itself, and (ii) these considerations are only loosely state ofthe artshouldthereforenotberegardedasdefinitive. connectedtothespecificfeaturesofsustainabilityassessment It will be reconsidered regularly, based on surveys of future (as described in Sect. 2.2). Such an assessment would – at developmentswithintheMCDAarea. leasttheoretically–addressalllevelsofdecision-makers,from stakeholders to policymakers. However, this does not mean thattheseaspects areof secondaryimportancein theprocess 4.2.2. SelectionfromMADMmethods ofexanteevaluationofthesustainabilityofcroppingsystems. Theyaresimplymorerelevanttoconsiderinthestepsfollow- In this study, we considered an integrative taxonomy of ingthepre-selectionofrelevantMADMmethods(seesection MADMmethods,somanymethods’variantsdidnotfindtheir below). wayintothisreview.Indeed,ourpurposewastodiscussgen- eralguidelinesfortheselectionofarelevantMADMmethod forexanteassessmentofthesustainabilityofalternativecrop- 4.3. Recommendednextsteps ping systems, rather than a complete and exhaustive survey of the existing methodsandtheir evaluationforthis purpose. The proposed ranking of candidate MADM methods Inourcase,a moredetailedcomparativereviewofalgorithm (Fig. 1) should be considered as a starting pointfor effective variants within each method group would extend far beyond ex ante assessment of the sustainability of alternative crop- the scope of this paper, as it would require more technical ping systems in order to identify the ones that could be im- backgroundinformationandfewer sustainabilityassessment- plementedinthefield. relatedconsiderations. Ideally,the nextsequenceofstepsshouldbecloselymon- We did not consider here selection criteria based on the itored by a relevant group of decision-makers which should ability of the methods reviewed to tackle information uncer- include researchers and other stakeholders who interact fol- tainty through fuzzy (and/or stochastic) procedures, for two lowingaparticipatory/cooperativeapproach.Thisisessential mainreasons.Firstly,theuncertaintycriterionisnotdiscrim- to avoid a classical researcher-driven process toward a pre- inatory,asallthereviewedmethodscouldbecoupledtosuch determined direction, which is risky especially for sustain- procedures.Secondly,someauthorshavearguedthat(i)there ability assessments as those address intrinsically holistic and isalackofconvincingevidencethattheimprecisioncaptured subjective issues (Brunner and Starkl, 2004). This group of Exanteassessmentofthesustainabilityofalternativecroppingsystems 171 3.Theexanteassessmentofthedesignedcroppingsystems and the analysis/interpretation of the results by the decision- MMAADDMM mmeetthhooddss sseelleeccttiioonn makersbasedontheconsideredsustainabilitycriteriaandthe characteristics of the applied MADM tools. Consistent with therecommendationsofZanakisetal.(1998),Macharisetal. (2004) and Wang and Triantaphyllou(2006), this multi-tool- based analysis may reveal alternative cropping systems that 11 wouldbeconsideredapriorisustainable,independentlyofthe MMAADDMM ttoooollss && methodapplied.However,beforethefinalselectionoftheop- ssooffttwwaarree sseelleeccttiioonn tions,itisrecommendabletoperformsensitivityandexplana- tion analysis of the evaluation results obtained via each con- sideredtool. Duringallthesesteps,itisnecessarytomaintainaregular feedbackwiththeoperationalresearchcommunity,inorderto DDeecciissiioonn ensureacohesiveoperationalframework. --mmaakkeerrss ((rreesseeaarrcchheerrss,, 33 ssttaakkeehhoollddeerrss,,……)) 22 Acknowledgements: ThisstudywasfundedbytheANR(AgenceNa- EExx aannttee CChhooiicceeooff tionalede la Recherche)under theFrenchFederatorProgram ADD(Agri- EExx aannttee ddeessiiggnn ooff ssuussttaaiinnaa-- cultureetDéveloppementDurable)viatheProjectDISCOTECH(Dispositifs eevv&&aall uuiinnaatttteeiioorrppnnrr eeaattnnaaaattiillooyynnssiiss cscsrryyoosspptteeppmmiinnssgg ccbbrriilliiiitttteeyyrriiaa innovants pour la conception et l’évaluationdes systèmestechniques). We thankOlivierCrespoforhisvaluablecommentsonmulti-criteriadecision-aid methods. Figure2.Suggestedstructureoftheframeworktobesetupfollowing themulti-criteriadecision-aid(MADM)methodselectionprocess,in ordertocarryoutexanteassessmentofthesustainabilityofcropping REFERENCES systems. Alphonce C.B. (1997) Application of the analytic hierarchy process in agricultureindevelopingcountries,Agr.Syst.53,97–112. decision-makersshouldthenworkonthebasisofathree-axis AndreoliM.,TellariniV.(2000)Farmsustainabilityevaluation:method- framework(Fig.2),consistingofthefollowingsteps: ologyandpractice,Agr.Ecosyst.Environ.77,43–52. 1. The collective selection of given decision-aid ArondelC.,GirardinP.(2000)Sortingcroppingsystemsonthebasisof tools/software from the most suitable MADM method theirimpactongroundwater quality,Eur.J.Oper.Res.127,467– 482. categories.EachMADMgroupbeingcomposedofnumerous methodvariantsanddozensofcorrespondingtools,acompar- AyalewL.,YamagishiH.,MaruiH.,KannoT.(2005)LandslidesinSado Island of Japan: Part II. GIS-based susceptibility mapping with ative (even restricted) assessment of these methodsand tools comparisons of results fromtwo methods and verifications, Eng. might be necessary before selection. In order for the tool to Geol.81,432–445. beusedbyalargevarietyofdecision-makers,thecomparison Baker D., Bridges D., Hunter R., Johnson G., Krupa J., Murphy J., shouldbe basedon operationaland practicalcriteria, such as Sorenson K. (2002) Guidebook to Decision-Making Methods, (i) the availability of the tool and its documentation, (ii) the WSRC-IM-2002-00002,DepartmentofEnergy,USA. time and manpower resources required for the analysis, and BanaeCostaC.A.,VansnickJ.C.(1997)ApplicationsoftheMACBETH (iii)theeaseofuse,transparencyandreportingcapabilitiesof approach in the framework of an additive aggregation model, J. thetool(UKDTLR,2001) Multi-CriteriaDecisionAnal.6,107–114. 2. The collective ex ante design of the options (i.e., alter- Bohanec M., Džeroski S., Žnidaršiè M., Messean A., Scatasta S., native croppingsystems) to be evaluatedbased on vectorsof Wesseler J. (2004) Multi-attribute modelling of economic and ecological impact of cropping systems, Informatica (Ljubljana, sustainability criteria, with respect to the specific features of Slovenia)28,387–392. the MADM tool considered (Fig. 2). In this key step, it is essential that the knowledge and expertise of the decision- BontkesT.S.,vanKeulenH.(2003)Modellingthedynamicsofagricul- turaldevelopment atfarmand regionallevel,Agr.Syst.76,379– makers encompasses the considered sustainability issues, in 396. order to design (quantitatively and/or qualitatively) relevant BouyssouD.,PernyP.,PirlotM.,TsoukiasA.,VinckeP.(1993)Aman- sustainabilitycriteria(e.g.groundwaterpollution,erosionand ifestoforthenewMCDAera,J.Multi-CriteriaDecisionAnal.2, compactionrisks,impactonbiodiversity,energyconsumption, 125–127. grossmargin,healthrisks,etc.).Inorderforthegrouptopro- motethediscovery/designofalternativesustainabilitycriteria Bouyssou D., Marchant T., Pirlot M., Perny P., Tsoukiàs A., Vincke P. (2000) Evaluation and decision models: a critical perspective, notobviousorapparentatfirstsight,aworkstrategybasedon KluwerAcademic,Boston,London,Dordrecht. brainstormingtoolssuchaslateralthinking,affinitydiagrams Bouyssou D., Marchant T., Pirlot M., Tsoukias A., Vincke P. (2006) and interrelationship diagraphs can be of importance (Baker Evaluationdecisionmodelswithmultiplecriteria.Steppingstones etal.,2002). fortheanalyst,SpringerScience+BusinessMediaInc.,NewYork. 172 W.Sadoketal. BransJ.P.(1982)L’ingénieriedeladécision;Élaborationd’instruments Goodwin P., Wright G. (1998) Decision Analysis for Management d’aide à la décision. La méthode PROMETHEE, in: Nadeau R., Judgement,secondedition,JohnWiley,Chichester. Landry M. (Eds.), L’aide à la décision : Nature, Instruments GrecoS., Matarazzo B. (2005) Decisionruleapproach, in: FigueiraJ., et Perspectives d’Avenir, Presses de l’Université Laval, Québec, pp.183–213. GrecoS.,EhrgottM.(Eds.),MultipleCriteriaDecisionAnalysis: StateoftheArtSurveys,Springer-Verlag,NewYork. BrewkaG.(2006)AnswerSetsandQualitativeOptimization,LogicJnl. IGPL14,413–433. GuitouniA.,MartelJ.M.(1998)Tentativeguidelinestohelpchoosingan appropriateMCDAmethod,Eur.J.Oper.Res.109,501–521. Brunner N.,Starkl M. (2004) Decisionaid systemsfor evaluating sus- HansenJ.W.(1996)Isagriculturalsustainabilityausefulconcept?Agr. tainability:acriticalsurvey,Environ.Impact.Assess.Rev.24,441– 469. Syst.50,117–143. Cerdan O., Souchère V., Lecomte V., Couturier A., Le Bissonnais Y. HayashiK.(2000) Multicriteriaanalysisforagriculturalresourceman- (2002) Incorporatingsoilsurfacecrustingprocessesinan expert- agement:acriticalsurveyandfutureperspectives,Eur.J.Oper.Res. basedrunoffanderosionmodelSTREAM(SealingTransferRunoff 122,486–500. ErosionAgriculturalModification),Catena46,189–205. Hengsdijk H., Van Ittersum M.K. (2002) A goal-oriented approach to Chen S.J., Hwang C.L. (1992) Fuzzy Multiple Attribute Decision identifyandengineerlandusesystems,Agr.Syst.71,231–247. Making:MethodsandApplications,Springer-Verlag,Berlin. HerathG.(2004)Incorporatingcommunityobjectivesinimprovedwet- CristP.J.,KohleyT.W.,OakleafJ.(2000)Assessingland-useimpactson land management: the use of the analytic hierarchy process, J. Environ.Manage.70,263–273. biodiversityusinganexpertsystemstool,LandscapeEcol.15,47– 62. Hermanides G., Nijkamp P. (1997) Multicriteriaevaluation of sustain- deKeyserW.,PetersP.(1994)ARGUS–anewmultiplecriteriamethod ableagriculturallanduse:acaseofLesvos,ResearchMemorandum based on the general idea of outranking, in: Paruccini M. (Ed.), 1997-5. Free University of Amsterdam, Faculty of Economics, BusinessAdministrationandEconometrics. Applyingmultiplecriteriaaidfordecisiontoenvironmentalman- agement,Kluwer,Dordrecht,pp.263–278. Hwang C.L., Yoon K.L. (1981) Multiple Attribute Decision Making: denBiggelaarC.,SuvediM.(2000)Farmers’definitions,goals,andbot- MethodsandApplications,Springer-Verlag,NewYork. tlenecks of sustainable agriculture in the North-Central Region, IkerdJ.E.(1993)Theneedforasystemsapproachtosustainableagricul- Agr.Hum.Val.17,347–358. ture,Agr.Ecosyst.Environ.46,147–160. DentJ.B.,Edward-JonesG.,McGregorM.J.(1995)Simulationofeco- Jelassi M.T.J., Ozemoy V.M. (1988) A framework for building an ex- logical,social,andeconomicfactorsinagriculturalsystems,Agr. pert system for MCDM models selection, in: Lockett A.G., Islei Syst.49,337–351. G.(Eds.),ImprovingDecisionMakinginOrganzations,Springer- DiasL.C.,TsoukiàsA.(2003)Ontheconstructiveandotherapproaches Verlag,NewYork,pp.553–562. indecisionaiding,in:HenggelerAntunesC.A.,FigueiraJ.(Eds.), KangasJ.,KangasA.,LeskinenP.,PykäläinenJ.(2001)MCDMmethods Proceedings of the 57th Meeting of the EURO MCDA Working instrategicplanning offorestryon state-ownedlands inFinland: Group,UniversityofTuscia,Italy. applicationsandexperiences,J.Multi-CriteriaDecisionAnal.10, DogliottiS.,RossingW.A.H.,VanIttersumM.K.(2004)Systematicde- 257–271. signandevaluationofcroprotationsenhancingsoilconservation, Keeney R., Raiffa H. (1976) Decisions with Multiple Objectives: soilfertilityandfarmincome:acasestudyforvegetablefarmsin PerformancesandValueTrade-Offs,Wiley,NewYork. SouthUruguay,Agr.Syst.80,277–302. Kim T.L., Wiggins L., Wright J. (1990) (Eds.) Expert Systems: DogliottiS.,VanIttersumM.K.,RossingW.A.H.(2005)Amethodforex- ApplicationstoUrbanPlanning,Springer-Verlag,NewYork. ploringsustainabledevelopmentoptionsatfarmscale:acasestudy forvegetablefarmsinSouthUruguay,Agric.Syst.86,29–51. Kropff M.J., Bouma J., Jones J.W. (2001) Systems approaches for the designofsustainableagro-ecosystems,Agr.Syst.70,369–393. Duke J., Aull-Hyde R. (2002) Identifying public preferences for land preservationusingtheanalytichierarchyprocess,Ecol.Econ.42, KrusemanG.,RubenR.,KuyvenhovenA.(1996)Analyticalframework 131–145. fordisentanglingtheconceptofsustainablelanduse,Agr.Syst.50, 191–207. ErtugrulKarsakE.(2004)Fuzzymultipleobjectivedecisionmakingap- proachtoprioritizedesignrequirementsinqualityfunctiondeploy- LeclercJ.P.(1984)Propositionsd’extensiondelanotiondedominanceen ment,Int.J.Prod.Res.42,3957–3974. présencederelationsd’ordresurlespseudo-critères:MELCHIOR, Math.SocialSci.8,45–61. European Commission (2005) Impact assessment Guidelines, 15th of June2005withMarch2006Update,SEC(2005)791. LoyceCh.,RellierJ.P.,MeynardJ.M.(2002a)Managementplanningfor winterwheatwithmultipleobjectives(1):theBETHAsystem,Agr. Ferraro D.O., Ghersa C.M., Sznaider G.A. (2003) Evaluation of envi- Syst.72,9–31. ronmentalimpactindicatorsusingfuzzylogictoassessthemixed cropping systems of the Inland pampa, Argentina, Agr. Ecosyst. LoyceCh.,RellierJ.P.,MeynardJ.M.(2002b)Managementplanningfor Environ.96,1–18. winter wheat withmultipleobjectives (2):ethanol-wheat produc- tion,Agr.Syst.72,33–57. FigueiraJ.,GrecoS.,EhrgottM.(2005)(Eds.)MultipleCriteriaDecision Analysis:StateoftheArtSurveys,Springer-Verlag,NewYork. Ma L. (2006) Knowledge Representation Under Inherent Uncertainty in a Multi-Agent System for Land Use Planning, Ph.D. Thesis, FrenchS.(1988)DecisionTheory:anIntroductiontotheMathematicsof EindhovenUniversityofTechnology,Eindhoven,TheNetherlands, Rationality,EllisHorwood,Chichester. 164p. GafsiM.,Legagneux B.,NguyenG.,RobinP.(2006)Towards sustain- MacCrimmon K.R. (1973) An overview of multiple objective decision able farming systems:Effectiveness and deficiency of theFrench making, in: Cochran J.L., Zeleny M. (Eds.), Multiple Criteria procedureofsustainableagriculture,Agr.Syst.90,226–242. DecisionMaking,UniversityofSouthCarolinaPress,Columbia.