Agata Matarazzo, Maria Teresa Clasadonte, Carlo Ingrao, Antonio Zerbo / InternationalJournal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.523-530523 | P a g eCriteria Interaction Modelling In The Framework Of LCAAnalysisAgata Matarazzo, Maria Teresa Clasadonte, Carlo Ingrao, Antonio ZerboDepartment of Economics and Business, University of Catania, Italy- Corso Italia, 55. 95129. Catania.AbstractIn this paper, we emphasize thepossibility to aggregate evaluations by usingdifferent multi-criteria approaches andoperators, actually more suitable forenvironmental problems. More precisely, valuesof different indicators, qualitative andquantitative, can be aggregate in a more flexibleand efficient way, using operators not necessarilycompensatory, allowing partial compensationand/or not transitive preferences, including alsothe very important possibility to modelinteractions among criteria.A single descriptor of multiple attributes, knownas a composite indicator, can be used to reconcileapparently incommensurable criteria into acomparable basis, but the current aggregationand weighting practices in LCA are notsufficiently rigorous, because its methodologicalsimplicity, usually being just weighted sums thatimplicitly require preference independencehypothesis and totally compensatory nature. Wehave explored different MCDA approaches in theliterature that may be suitably applied by LCApractitioners to enhance the technical credibilityand also identify issues affecting the selection andimplementation of more appropriate aggregationapproach. In this direction, particularly attentioncan be devoted to some methods allowing also themodelling of interaction among criteria, wheresuitable techniques can actually implemented inorder to improve the validity of the results ordealing wit ordinal information, very oftenpresent in environmental problems.Keywords: Life Cycle Impact Assessment, impactcategories, MCDA, interactions, weighting.IntroductionProduction and consumption of goods andservices are primary factors causing harmful effectson the environment. For this reason, themanagement of production and consumption is akey area in order to achieve sustainabledevelopment in our society. There is therefore aneed for suitable decision-making tools to evaluateenvironmental options in all the productionprocesses., In this field, Life Cycle Assessment(LCA) appears to be a valuable method forassessing the environmental considerations of aproduct or service throughout its entire Life Cycle,including everythingfrom raw material extraction, processing,transportation, manufacturing, distribution, use, re-use, maintenance and recycling to final disposal(Consoli et al. 1993). LCA is the term that iscurrently widely accepted for environmentalassessment of products or processes on a holisticcradle-to-grave basis (Kooijman 1993); theInternational Standardization Organization iscurrently working on LCA methodologies with theaim of harmonising different methods andpromoting further establishment (Strazza C. et al.,2010).So, LCA models are likely to become thefuture decision support tools and they include aninventory model and an assessment model (DelBorghi 2013) . The Life-Cycle Inventory model(LCI) provides a detailed account of all resourceconsumptions and emissions for an environmentalsystem, as well as for any up-stream or down-streamactivities associated with the impacts. The Life-Cycle Impact Assessment model (LCIA) translatesand aggregates, according to fairly standardizedmethods, all of the detailed information provided bythe LCI into the main resource consumption ofconcern and the main environmental impactcategories (global warming, acidification, etc.). Thesignificance of the aggregated data relative to eachother and to all combined activities in society canusually be obtained by the weighted sum ofnormalized impacts caused by one average person.Weighting of normalized results can be made toidentify the importance of environmental impacts orresource consumptions, but consensus on weightingfactors has not yet been reached (Kirkeby 2007).During the 1990s detailed contents wereassigned to the steps of LCA on the basis of co-operation between investigators in several countries,especially in the Society of EnvironmentalToxicology and Chemistry (SETAC 2008) and theInternational Organization for Standardization(ISO). Life cycle assessment is documented in theform of ISO standards 14040-14044 (ISO 1997; ISO2000a; ISO 2000b; ISO 2006a; ISO 2006b), givinginstructions for LCA practitioners to conduct LCAapplications according to “good practice”. Inventorydata, environmental interventions representing a“cradle-to-grave” perspective, are the core of LCAs.
Agata Matarazzo, Maria Teresa Clasadonte, Carlo Ingrao, Antonio Zerbo / InternationalJournal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.523-530524 | P a g eHowever, from the point of view of a decisionmaker, the inventory results are usually notsufficient for making better comparative decisions.In comparative studies it may be found thatalternative A is better than alternative B with respectto some points of view, but poorer in others. LCIAenables us to interpret the results of the inventory,but it is just the first step to draw the correctconclusions concerning improvement approaches(Saur et al. 1996). The consideration ofenvironmental effects as a consequence ofenvironmental interventions provides additionalinformation, which is not covered by the inventorystep. Several methods exist to establish weightingfactors, because weighting has always been acontroversial issue; it can be difficult, in the contextof developing an LCIA approach, to select onemethod among others. A good weighting methodmust include certain common features, includingtransparency, simplicity in implementation,communicability and magnitude or capacity toassess environmental problems. Moreover, theweighing procedures have to take into considerationa certain number of issues including: the probabilityof the attribute to cause an undesirable consequenceto the environment; the magnitude/severity of thisconsequence, possibly not reversible; the temporalaspect such as duration, time of occurrence, etc.This results in a proliferation of data gathering,storing and analysis techniques within theenvironmental assessment tools and MCDAmethods are more relevant to these tasks. The use ofuncertainty modelling formalisms and the buildingof decision support frameworks allowing theefficient interpretation of the assessment results anddata in an uncertain environmental are greatlyneeded.In this paper, first we outline aclassification and assessment framework of the mainMulti-Criteria Decision Analysis (MCDA) methodsin the light of their ability to take into account thepeculiar needs of LCIA (Benetto 2000). We shortlyanalyze different MCDA approaches to suggest themost appropriate method for aggregating differentattributes values, underling the weighting issue ofeach one. This paper therefore aims to indicate themain guidelines representing a starting point in thedevelopment of a robust weighting method, whichtakes into consideration the importance of clearlyidentifying what part of the responses deals withsubjectivity and objectivity in weighting assessment,to provide the most relevant information during thedecision-making process. In absolute terms, there isnot recognized an ideal weighting method, so theproposed approach combines different weightingtechniques and problems in a single weightingframework.1. The LCIA AnalysisThe LCA provides accurate and robustresults by utilising the very detailed modelsdeveloped in the inventory stage and this phaseneeds greater transparency and standardization ofmethods and calculations (Garrett P., 2013). LCIAis typically divided into five phases: selection ofimpact categories, classification, characterisation,normalisation and weighting. In the selection ofimpact categories (e.g. climate change andacidification), indicators for the categories andmodels to quantify the contributions of differentenvironmental interventions to the impact categoriesare selected. The second phase, classification, is anassignment of the inventory data to some predefinedimpact categories. In the characterisation phase, themodels make an aggregation of possibleinterventions within each impact category. Values ofinterventions are changed into impact categoryindicators resulting from characterisation factors.These factors measure the effect intensity of eachsingle emission on the environmental problem athand.From a decision maker’s perspective,impact category indicators are more manageablethan interventions, but due to their proxycharacteristics they are difficult to interpret. In orderto obtain a more comprehensive view of impactcategory indicator results, some suitablenormalisation can be conducted to better assess theiroverall environmental impact. Finally, the weightingphase relates the normalised magnitude of theindicator results with respect to the different impactcategories to reference values, calculated on thebasis of an inventory of a chosen reference system(Consoli et al. 1993, Wenzel et al. 1997, Finnvedenet al. 1997).Normalisation is needed in practice forcomparative evaluations but it requires trade-offsbetween different category indicators, which arevery difficult to evaluate. These trade-offs aredetermined as particular “weighting factors”(weights) in the weighting phase; we observe thatthese kind of weights depend also on the particularmeasurement scale of each indicator and thereforethey are introduced in order to be able to obtain afinal unique aggregated value (for example, theweighted sum) as an overall evaluation. As aconsequence, the trade-offs do not give us directinformation about the “intrinsic importance” of thecorresponding indicators, since their values dependon the reference system actually chosen, includingthe ranges of the measurement scales. This meansthat if the trade -off between categories A and B isequal to 3, the relative intrinsic importance of A isnot three times the importance of B, but only thatthe “scale factor” to appropriately aggregate thesefactors in an additive value function is 3. For thisreason, weighting is desirable or necessary in many
Agata Matarazzo, Maria Teresa Clasadonte, Carlo Ingrao, Antonio Zerbo / InternationalJournal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.523-530525 | P a g eLCA applications (e.g. Hansen 1999, Bengtsson2000), although determination of weights is basedon subjective value choices. Weights in fact are usedwith different meanings of “trade- off “ in functionalapproach or “coefficients of importance” in the caseof non-compensatory methods (Minciardi et al.2007). In this case, the inter-criteria informationrequired is an evaluation of the relative importancebetween coalitions of criteria; such a concept ofrelative importance is often translated into numbers,called weights (for example in ELECTRE methods).In the case of totally compensatory methods(weighted sum), on the contrary, the weights have tobe considered as scale factors and then theirmeaning is that of a trade-off between each pair ofcriteria (Minciardi et al. 2008).In our opinion, the great attention devotedin LCIA to the weighting problems is related to theuse of the weighted sum as usual, simple andefficient aggregation operator, accepted by themajority of authors, at least implicitly. But thisoperator, that should be applied only when all thedata are homogeneous, although very easy to beunderstood, suffers from a lot of drawbacks. First ofall, it is perfectly compensatory, that is a greatadvantage with respect to some characteristics, andcan always compensate disadvantages, even if veryhuge, with respect to other points of view. In ouropinion, this implicit hypothesis is very dangerousin the framework of environmental analysis, whereit is often not possible to compensate (for example,with high financial cash flows) some very dangeroussituations (concerning public health). Another weakpoint of the weighted sum is that this kind ofaggregation operator can be applied only if there ispreference independence among the consideredimpact indicators. This means that, if categories aand c have the same evaluations with respect to asubset of considered criteria, and categories b and dalso have the same evaluations – but different fromthe evaluations of a and c - with respect the previousconsidered subset of criteria, and category a ispreferred to category b, also c should be preferred tod, independently of their common evaluations withrespect to the remaining criteria. This means that theweighted sum approach is not able to model anyinteraction among criteria, where this characteristicaffects a lot of real environmental situations andshould be taken into consideration (Matarazzo2012).Although there is an approximateconsensus on the procedural framework of LCIA,the methods may vary in LCA applications.Different methods can of course produce differentresults. The results depend, among other things, onthe coverage of impact categories, the chosenimpact category indicators and the models chosenfor characterisation factors. Furthermore, asobserved before, a reference system used innormalisation can dramatically affect theinterpretation and the results, also in the frameworkof the weighted sum operator. In practicalapplication, in order to avoid some assessmentdifficulties and to simplify the implementation ofwell known approaches, “less is better” has beenoften applied in LCIA (Lo Giudice 2011). Thisapproach assumes that all amounts of the samekinds of interventions lumped together causeharmful effects on the environment on the basis oftheir intrinsic hazard characteristics, regardless ofwhere and when they take place, and only whetherthe amounts are above or below certain thresholds(White et al. 1995). Although the advantage of the“only-above-threshold” approach is clear, itsimplementation in practice has been developed onlyslowly, due to the complexity of applying it on a lifecycle scale.In LCIA the values of environmentalinterventions assessed in the inventory analysis areinterpreted on the basis of their potentialcontribution to the overall environmental impact.The term “potential contribution” indicates that theresult of LCIA is not an absolute value, and thatLCIA is a relative approach, part of a morecomplete analysis of environmental assessments.The idea is that comparative studies need moredetailed data on temporal and spatial aspects thanthat required by absolute methods, such asenvironmental risk assessment. Weighting is aprocess which can take place within the differentsteps of LCIA and LCA interpretation to aggregatethe results into a single score; LCIA is used toevaluate the significance of the environmentalinterventions contained in a life cycle.Determination of weights has been acontroversial issue in LCIA because of itssubjectivity. For this reason, for example, the ISOprovides no examples of weighting. Despite theweight assessment being necessary in a lot ofaggregation operators, different approaches havebeen proposed in the specific literature in order toavoid excessive arbitrariness in this so difficultissue, but also taking appropriately into account thesubjective preferences of the decision-maker. LCIAresults for each process of the life cycle have then tobe aggregated in order to get reliable conclusionsand this aggregation framework has to minimize theinformation loss and the compensation betweenimpacts and to reduce the amount of data in order toimprove the intelligibility of results. The impactresults of each scenario cannot be easily comparedwith each other and impacts cannot be aggregated asthey represent very different realities.In the following, we shortly present the mainfeatures of some well known MCDM methods,classified – as usually- in Functional approaches;relational approaches and interactive approaches
Agata Matarazzo, Maria Teresa Clasadonte, Carlo Ingrao, Antonio Zerbo / InternationalJournal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.523-530526 | P a g e2. Functional approachesThis method selects a suitable aggregationfunction, i.e. a real value function (utility or valuefunction) associating to each alternative a realnumber, representing its degree of preference, thatallows us to directly make comparisons and obtainthe final recommendation for choice or rankingproblems.The main features of this approach are thefollowing: inter-criteria information as trade-offweights (substitution rates), usually constant;compensatory logic (only sometimes not perfectlycompensatory). Formally, the utility function isexpressed as U(a) = U(g1(a), g2(a), ...,gm(a)), aA,where U is a function of m variables, increasing inall its argument (marginal utilities) A is the set offeasible alternatives and G the set of consideredcriteria.4.1. Additive modelsThe simplest form of such a utility functionis the additive representation, that isU(a) = Gjjj))a((gv ,where, for each, jG, vj is a non decreasingfunction and))(( agv jjis the marginal utility. But,as we said before, this kind of representation isbased on the hypothesis of preference independence,that is no interaction among criteria, usually nonrealistic particularly in environmental impactanalysis.Very often, a particular case of utility functionis considered, by the hypothesis))(( agv jj=jgj(a),where j0, for each jG, is the (constant) trade offweight, used also to normalize the evaluations ondifferent criteria. Therefore, we obtain the weightedsum utility function,U(a) = Gjjj)a(g ,That is one of the most elementary aggregationoperators, as above recalled, that requires theelicitation of a set of trade-off weights. This verysimple approach is used in a lot of real lifeapplications, such as in LCIA.2.2. Non additive modelsThere are also some non additive utilityfunctions, for example in the form of multiplicativeor polynomial functions. Apart from them, one ofthe most interesting approaches is given by the classof non-additive integrals, also called fuzzy integrals.In this approach, the main very interesting ideaconcerning weighting is that the intrinsic importanceof criteria is represented giving a weight to eachsubset of criteria from G. Then it is possible toexplicitly represent the importance of each coalitionof criteria, and therefore also their interactions.3.3. The Analytic Hierarchy ProcessThe Analytic Hierarchy Process (AHP)structures the decision problem into levels whichcorrespond to DM’s understanding of the situation:objectives, criteria, sub-criteria and alternatives; bybreaking the problem at hand into different levels,the DM can focus on smaller sets of decisions. Thismethod (Saaty 2008) is based on four main axioms:given some alternatives, the DM is able to provide apairwise comparison of these alternatives withrespect to any criterion on a ratio scale, which isreciprocal. One can formulate the decision problemas a hierarchy; all criteria and alternatives whichimpact one decision problem are represented in thehierarchy. After estimating the weights, the DM isalso provided with a measure of the inconsistency ofthe given pairwise comparisons. It is important tonote that the AHP does not require decision makersto be consistent in giving preference information,but rather provides a measure of inconsistency, aswell as a method to reduce this measure if it isdeemed to be too high. After generating a set ofweights for each alternative and for any criterion,the overall priority of the alternative is computed bymeans of a linear, additive function. The AHP is avery widespread approach in many real-worldapplications and it explicitly deals with the issue ofhierarchy in decision problems. AHP can beconsidered a particular case of multi-attribute valuefunctions approach, thus being completelycompensatory in nature. This implies that theweights derived in an AHP framework are always inthe form of trade-offs and never of importancecoefficients.4. Relational approachThis approach is characterized by splittingthe decision analysis in two steps. In the first one,we build up some outranking relations, making adirect comparison between each pair of alternativeswith respect to each criterion, individuallyconsidered. The global outranking relations aSbmeans “a is as least as good as b with respect to allcriteria from G”. Therefore, in the second step, theserelations are exploited in order to obtain the finalrecommendation. The main features of this approachare the following: building of crisp or fuzzyoutranking relations and their exploitation; infra-criteria information by the introduction of suitablethresholds to take into consideration intervalindifference and preference, incomparability andveto situations; inter-criteria information in the formof importance weights; logic usually locally non-compensatory. The outranking relation aSb holds ifand only if there are enough arguments to say that ais at least as good as b, while no serious reason torefute this statement exists (Roy 1985). The finalrecommendation depends on the global outranking
Agata Matarazzo, Maria Teresa Clasadonte, Carlo Ingrao, Antonio Zerbo / InternationalJournal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.523-530527 | P a g erelation, and not directly on the alternativesevaluations with respect to each consideredcriterion. The preference model of this approach isvery rich, but it requires the elicitation of a lot ofparameters, including importance weights, thatexpress the intrinsic relative importance of eachcriterion or coalition of criteria and that then do notdepend on the different scales of measurement.These weights have a crucial role in the building ofthe outranking relations.The most important outranking methods are thefollowing.4.1 ELECTRE methodsThe family of ELECTRE methods (Roy1995., Figueira, et al. 2005), is based on the conceptof concordance and discordance analysis, and eachone is devoted to solving specific decisionalproblems; ELECTRE I for choice problems,ELECTRE II and III for ranking problems,respectively building “nested” outranking (i.e. withdifferent strength) and fuzzy outranking relations;ELECTRE IV that does not require elicitation ofimportance weights; ELECTRE TRI for sortingproblems. The outranking relations are built takinginto consideration the importance of criteriacoalition (i.e. their weights) supporting the idea thata is at least as good as b (concordance test) and thatthere is no strong opposition to this sentence (non-discordance test). A problem specifically connectedwith the outranking approach is that it is necessaryto assess a large number of ad hoc parameters, i.e.indifference and preference thresholds, concordancethreshold, discordance thresholds and weights andthis may cause loss of transparency and consistencyin the model (Norese M.F., 2006).4.2 PROMETHEEIn the PROMETHEE methods (Brans1982), for each criterion gjG and for each pair ofalternatives (a,b) a degree of preferencePj(a,b)[0,1] is defined as a non decreasing functionof the difference of evaluations gj(a) - gj(b). Theshape of this function and the value of its parametersare chosen by the DM among some predefinedtypes.Using the notation kj for the weight of criteriongj, the global preference degree (a,b) is defined as(a,b) =FFjjjjjkbaPk ),(.The final recommendation (ranking) is basedon the concept of outgoing + flow and ingoingflow -, defined as following for each feasiblealternative a:+(a) = Abba,-(a) = Abab,.PROMETHEE I method gives us a partialpreorder of feasible alternatives, (that means thatincomparability between alternatives is allowed) asthe intersection between the complete ascendingpreorder and complete descending preorder, built uprespectively using the values of +(a) and -(a)(Brans and Vincke 1985).PROMETHEE II method proposes the buildingof a complete preorder of feasible alternatives,taking into consideration the so called net flow((Brans 1982, Oberschimidt J. et al., 2010):(a) = +(a)--(a).The PROMETHEE approach is based on asimple mathematical structure and is easy to use. Inthis model, preference intensity can be used, and thedegree of compensability allowed is high.Moreover, weights cannot be considered asimportance coefficients, but they should be derivedas trade-offs. The possibility of rank reversals ishigh (Munda G.2008).4.3 REGIMEREGIME Analysis is a method thatconsiders pure ordinal information, and could beconsidered as the ordinal generalization of theconcordance methods (Hinloopen and Nijkamp,1986; Hinloopen, Nijkamp and Rietveld, 1983). Themain idea is the concordance index Cil concerningthe pair of feasible alternatives ai, al :Cil = iljjw,where wj is the importance weight ofcriterion gj, j=1, 2,…,m, and il, is the subset ofcriteria for which ai is at least as good as al., Themethod considers only the sign of the difference Cil– Cli for each pair of alternatives, and not its value,since the data my be ordinal. In this method, the firststep is the construction of the so-called impactmatrix, composed of the pairwise comparisons ofalternatives, that is then suitably exploited. Thismethod presents the advantage of taking intoaccount ordinal information. It should be recalledthat weights can also be simply ordinal in nature,and therefore from this point of view REGIME canbe considered entirely consistent in its use ofweights as importance coefficients. The onlyassumption required is that of a uniform distributionof the weights along the concordance region, that isnot restrictive. However, when mixed informationon criterion scores is present, the aggregationprocedure becomes cumbersome. To deal with bothqualitative and quantitative information, it isassumed that qualitative information is therepresentation of unknown quantitative information.Building on this assumption, a cardinalizationscheme is applied and then all the ordinalinformation is transformed into quantitative. Besides
Agata Matarazzo, Maria Teresa Clasadonte, Carlo Ingrao, Antonio Zerbo / InternationalJournal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.523-530528 | P a g econsiderations on the way, there is here a basicmethodological problem with weights. In fact, ifweights are connected to quantitative informationand the intensity of preference concept is used,weights can only be trade-off and not importancecoefficients.4.4 NAIADENAIADE (Novel Approach to ImpreciseAssessment and Decision Environments) (Munda1995) is a discrete multi-criteria method whoseimpact matrix may include crisp, stochastic or fuzzymeasurements of the performance of an alternativewith respect to an evaluation criterion, thus it is veryflexible for real world environmental applications. Apeculiarity of NAIADE is the use of conflictanalysis procedures to be integrated with the multi-criteria results. This method can give the followinginformation: ranking of the alternative according tothe set of evaluation criteria; indications of thedistance of the positions of the various interestgroups; ranking of the alternatives according toactors’ impacts or preferences. The whole NAIADEprocedure can be divided into four main steps:pairwise comparison of alternatives according toeach criterion; aggregation of all criteriaevaluations; ranking of alternatives; social conflictanalysis. This method presents many advantages:the possibility of taking into account various formsof mixed information in an equivalent way; thepossibility of determining the degree ofcompensability allowed in the aggregationprocedure; the explicit use of a conflict analysisprocedure, thus distinguishing clearly the technicaland social compromise solutions. However, thismethod suffers also from some serious limitations:the impossibility of using weights explicitly and, ifthey are used, they could only be trade-offs andnever importance coefficients; the large number ofad hoc parameters needed for the elaboration of themulti-criteria impact matrix; the fact that qualitativeinformation can be used only in the form oflinguistic variables and can never be measured on apurely ordinal scale.5. Interactive approachIt consists of a systematic exchange ofinformation between DM and analyst about localpreference evaluations of alternatives and about theconsequent computation results, with the aim ofsearching for a solution (descriptive logic: MultipleCriteria Decision Making– MCDM), or for thecomprehension of a coherent evolution of thedecision process (constructive approach – MCDA).In the first step, this approach requires a minimumnumber of inter-criteria information, such as trade-off weights, given by the DM.The analyst provides a first alternative (or afirst set of alternatives) considered satisfactory topropose to DM, who, during the discussion phase,gives the analyst further information about his/herpreferences. Therefore there is a new computationphase and so on. The procedure stops when the DMis satisfied by the alternative proposed, which is notalways necessarily efficient.A lot of interactive methods use the idea ofa reference point, defined in the criteria space, andsometimes that of ideal point, whose coordinates areusually the maximal values attainable by eachcriterion considered as a single one. There are twomain categories of interactive methods: those thatare based on the hypothesis of the existence of autility function, implicit in the mind of DM, andmethods that try to minimize a particular distancefrom a reference point. Among them we recall theGoal Programming and the STEM method(Benayoun, et al., 1971).Usually in this case someinformation about trade-off weights are explicitly orimplicitly required, as well as the normalization ofalternatives evaluations. Recently, very interestingapproaches have been proposed in this framework,combining the advantages of interactive methodsand genetic algorithms.ConclusionsThe interpretation of the LCA results inenvironmental problems and analyses should beactually aided by the application of MCDAmethods, appropriate from the methodological pointof view to the specific decision problem at hand.Decision analysis is a set of methods of systemsanalysis and operations research, going from theclassical mathematical programming to the newMCDM and MCDA approaches, which should beextensively applied in supporting real life decisionsin the presence of a number of conflicting points ofview. In the LCA literature, decision analysismethods are often known and used, up to now, onlyas tools for the weighting phase of LCIA, especiallyin the context of panel methods in which opinionsabout impact category weights are asked for from anindividual person or a group of persons. However,the use of decision analysis tools has not been verycommon in LCIA applications.This is partly due to the fact that LCApractitioners have been reluctant to use weightings,as a consequence of their subjectivity and difficultyin their assessment. But weighting is only anoptional phase in the ISO standards due to itssubjective results, and in the development of LCIAmethodology during the recent years researchershave overall concentrated on the mandatory phases.Only very few analyses have been conducted usingmore sophisticated and efficient approaches, such asMCDA methods. Possibly, one of the main reasonsfor this behaviour is that only a few LCAresearchers have enough information to use these
Agata Matarazzo, Maria Teresa Clasadonte, Carlo Ingrao, Antonio Zerbo / InternationalJournal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.523-530529 | P a g ekinds of decision analysis approaches in a contextdifferent from assessment of trade-offs.Therefore, our main attention should bedrawn not to the weighting problem in se, but to alldecisional processes, with a clear idea of thedifferent roles of all the actors (scientists, experts,decision-makers, stakeholders) involved in theprocess. Decision analysis is actually a process, withits own dynamic evolution, that has to be coherentwith the preferences and the goals of the decision-maker. In order to support the decision-maker inhis/her difficult activity, a multi-criteria approachcould be very useful, but the particular method to beused must be correctly chosen and applied. Thismeans, in particular, that the method we like toapply needs to be appropriate not only to the multi-criteria decision problem at hand (choice, ranking,classification, sorting), but also to the particular kindof data and information we have (cardinal, ordinal,qualitative, crisp, fuzzy, probabilistic,...).Any MCDA method, on the other hand, asobserved before, requires some peculiar preferenceinformation and technical parameters, among thosevery often weighting, in the form of trade-offs orimportance coefficients. Only few, but veryinteresting approaches, do not require a priori thiskind of information, that sometimes is givenimplicitly by the decision-maker, as in interactiveapproaches, or is obtained as output when we usepreference information in the form of examples ofdecision, like in the dominance-based rough setapproach. In any case, the choice of an MCDAmethod has to take into serious consideration theactual information DM can, wants and is able togive, avoiding forcing him/her to provide technicalinformation required by the method, but that DMdoes not know or cannot give, often also toodifficult or not understandable. It is the method thatmust be suitable and appropriate to the availableinformation, not the information that has to beadapted or forced in order to be used in any case inan approach chosen a priori.The interpretation of LCA results could begreatly improved by the use of suitable aggregationapproaches based on MCDA proper methods, basedon a functional or relational approach or oninteraction between scientist and DM.As a more general conclusion, we wouldsuggest that uncertainty management andquantitative analysis has to be further addressed inthe development and application of MCDAmethods, because uncertainty is also a primordialfactor in decision making, and not only in theenvironmental field.Alternatives approaches to LCA modellingmay also be useful, as well as some additional areaswhere further sensitivity and robustness analysiscould improve presentation of results; MCDAmethods allow very well for sensitivity analysis, atechnique in which a variable is systematicallymodified to determine the impact on the outcome. Inthis case, we are particularly interested in modifyingthe weights assigned to various criteria. Of course,other standardisation of LCA studied should beimproved, from a perspective of both transparencyas well as comprehensiveness of LCA modelling.References. Benayoun, R., J., de Montgolfier, J.,Tergny, and O.I., Larichev, 1971.”LinearProgramming with Multiple ObjectiveFunctions: STEP Method (STEM)”. MathProgram 1 (3):366-375. Bengtsson, M., 2000. “Weighting inpractice”. J Ind Eco 4 (4): 47-60. Benetto E., Rousseaux P., “Dealing withthe uncertainties of decision making in LifeCycle Assessment”. In Proceedings of the2nd International Conference on DecisionMaking in Urban and Civil EngineeringLyon, 20-22 November 2000: 765-776. Brans P., 1982, "L’ingénierie de ladécision: élaboration d’instruments d’aideà la décision. La méthode PROMETHEE.".Presses de l’Université Laval. Brans P. and Vincke P.,1985. "Apreference ranking organization method:The PROMETHEE method for MCDM".Management Science. Consoli, F., D., Allen, I., Boustead, J.,Fava, W., Franklin, A.A., Jensen, N., deOude, R., Parrish, R., Perriman, D.,Postlethwaite, B., Quay, J. Séguin, and B.Vigon, 1993. “Guidelines for Lifecycleassessment: A ’Code of Practice’”. Societyof Environmental Toxicology andChemistry (SETAC), Brussels. Del Borghi A., 2013. LCA andcommunication: Environmental ProductDeclaration. Int J Life Cycle Assess 18:293-295. Figueira, J., S., Greco and M., Ehrgott,2005. Multiple-criteria decision analysis.State of the art surveys. SpringerInternational Series in Operations Researchand Management Science, New York. Finnveden, G., 1999 “A critical rewiew ofoperational valuation/weighting methodsfor life cycle assessment. AFR-REPORT253 AFN, Naturvardsverket SwedishEnvironmental Protection ASgency 106 48Stockolm, Sweden. Garrett P, Ronde K., 2013, “Life cycleassessment of wind power: comprehensiveresults from a state –of-the art approach”,Int J Life Cycle Assess 18: 37-48
Agata Matarazzo, Maria Teresa Clasadonte, Carlo Ingrao, Antonio Zerbo / InternationalJournal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.523-530530 | P a g e. Hansen, O.J. 1999, “Status of Life CycleAssessment (LCA) activities in the Nordicregion”. Int J Life Cycle Assess 4: 315-320. Hinloopen, E., and P. Nijkamp, 1990.“Qualitative multiple criteria choiceanalysis, the dominant regime method”.Qua Quant 24: 37-56. Hinloopen, E., P., Nijkamp and P. Rietveld,1983. “Qualitative discrete multiple criteriachoise models in regional planning”. RegSci Urban Econ, 13: 77-102. ISO (1997) “ISO 14040: Environmentalmanagement - Life cycle assessment -Principles and framework. InternationalOrganization for Standardization”. ISO (2000a) “ISO 14042: Environmentalmanagement - Life cycle assessment - Lifecycle impact assessment. InternationalOrganization for Standardization”. ISO (2000b) “ISO 14043: Environmentalmanagement - Life cycle assessment - Lifecycle impact interpretation. InternationalOrganization for Standardization”. ISO (2006a) Environmental management -Life cycle assessment - Principles andframework. European Standard EN ISO14040. International Organization forStandardization. ISO (2006b) Environmental management -Life cycle assessment – requirements andguidelines. European Standard EN ISO14044. International Organization forStandardization. Kooijman J. 1993. “Environmentalassessment of packaging: sense andsensibility”, J Environ Manage 17: 575-586. Kyrkeby T. J., Birgisdottir H., SinghBhander G., Hauschild M., ChristensenT.H., 2007. “Modelling of environmentalimpacts of solid waste landfilling withinthe life-cycle analysis programEASEWASTE”. Waste Manage 27:961-970. Lo Giudice A., Clasadonte M.T.,Matarazzo A. 2011. LCI PreliminaryResults in the Sicilian durum wheat pastachain production. J Comm S Technol Qual.50 (1): 65-79. Matarazzo A., Clasadonte M.T., 2012.Using Multi-Criteria Decision makingmodels in the framework of LCIAAnalysis. 18th IGWT SymposiumTechnology and Innovationfor a Sustainable Future: a CommodityScience Perspective. Roma. Minciardi R., Robba M., Sacile R. 2007.Decision models for sustainablegroundwater planning and control. ControlEng Pract 8:1013-1029. Munda G. 1995, “Multicriteria evaluationin a fuzzy environment”. Physica-Verlag,Contributions to Economics Series,Heidelberg.. Norese M.F., 2006. “ELECTRE III as asupport for participatory decision-makingon the localization of waste –treatmentplants. Land Use policy”, Land Use Policy.23: 76-85. Oberschimidt J., Geldermenn J., LudwingJ., Schmehl M., 2010. “ModifiedPROMETHEE approach for assessingenergy tecnologies”. Int J Ener SectorManage, 4, 2: 183-212. Roy, B. .1985, “Méthodologie Multicritèred’Aide à la Décision“, Economica, Paris.. Roy, B. and D. Bouyssou, 1993. “AideMulticritère à la Décision: Méthodes etCas“ Economica, Paris.. Saaty T. L., 2008, “Decision making withthe analytic hierarchy process”. Int JService Science, 1, No. 1: 83 -97. Saur, K., J., Gediga, J., Hesselbach, M.,Schuckert, and P., Eyerer, 1996. „Lifecycle assessment as an engineering tool inthe automotive industry”. Int J Life CycleAssess 1(1): 15-21. SETAC Europe LCA Steering Committee(2008) Standardisation efforts to measuregreenhouse gases and carbon foorprinting“for products (editorial). Int J Life CycleAssess 13(2): 87-88. Strazza C., Del Borghi A., Costamagna P.,Traverso A., Santin M. (2010).Comparative LCA of methanol-fuelledSOFCs as auxiliary power systems onboard ships. Applied Energy 87:1670-1678. Wenzel, H., M. Hauschild, and L. Alting,1997. “Environmental assessment ofproducts”, Chapman & Hall, London. White, P., B., De Smet, H. Udo de Haes,and R. Heijungs, 1995. “LCA back ontrack - but is it on track or two ?” LCANews 5 (3): 1-3