Integrating fuzzy multicriteria analysis and uncertainty evaluation in life cycle assessment Enrico Benetto a, * , Christiane Dujet b , Patrick Rousseaux c a ECOINNOVA S.A.S. Consultancy, via G. Verdi 6,10090 Sangano-Turin, Italy b Laboratoire d’Analyse Environnementale des Proce ´de ´s et Syste`mes Industriels (LAEPSI) INSA-Lyon, 9 rue de la physique, 69621 Villeurbanne cedex, France c Laboratoire de Combustion et de De´tonique (LCD, UPR 9028 du CNRS) a ` l’Ecole Nationale Supe´rieure de Me´canique et d’Ae´rotechnique (ENSMA), FUTUROSCOPE-CHASSENEUIL, France article info Article history: Received 19 June 2006 Received in revised form 11 April 2008 Accepted 13 April 2008 Available online 5 June 2008 Keywords: Multicriteria analysis Fuzzy sets NAIADE Life cycle assessment LCA Electricity Noise abstract The interpretation phase of Life Cycle Assessment (LCA) studies is often hampered by the number and the heterogeneity of impact assessment results as well as by the uncertainties arising from data, models and practitioner’s choices. While decision-aiding methods have proven to support this key LCA phase, the inclusion of uncertainty evaluations in these methods is still a new topic in the LCA community. The fuzzy multicriteria method NAIADE [Munda G., 1995. Multicriteria Evaluation in a Fuzzy Environment. Heidelberg: Springer-Verlag.] has already been used in several LCA case studies and its framework is well suited for the consideration of uncertainty evaluations as well. This paper proposes a modified version of NAIADE, which conforms to the LCA requirements and to the type of uncertainties to be addressed, with an application example. The discussion of the rankings obtained showed how the integration of uncertainty results could help focusing on critical LCA issues and could be coupled with dominance, contribution and gravity analysis for the interpretation of LCA results, especially when the rankings are not fully consistent. This could support engineers and scientists working in LCA discipline as well as managers and decision-makers faced with decision-making problems based on LCA. Ó 2008 Elsevier Ltd. All rights reserved. 1. State of the art and goals Due to the rapid technological growth that has dramatically increased environmental pollution, environmental impact assess- ment approaches are often employed in situations in which ap- plications of technologies are known to have a negative impact on the environment (Neto et al., 2008; Salhofer et al., 2007). Envi- ronmental assessment approaches include, among others, global and local scale environmental impact and risk assessment tools. Life Cycle Assessment (LCA) is one of the most common tools and aims at the compilation and evaluation of the mass and energy inputs, outputs and the environmental impacts of a system throughout its life cycle (ISO, 2006; Guine ´ e et al., 2002). All lifecycle processes related to a function, for example the production of 1 GWh of electricity by coal combustion, from coal mining to end- of-life wastes treatment, through coal combustion, transports, treatments etc., are considered and data about the raw materials, energy consumptions and pollutant emissions related are collected. The data collected are then used to evaluate environmental impacts according to several impact categories representing environmental problems, e.g. ‘‘greenhouse effect’’, ‘‘stratospheric ozone depletion’’ and ‘‘eutrophication’’. The number of categories to be considered depends on the impact evaluation method chosen and on the level of aggregation of impact results foreseen. For example, ecotoxicity, eutrophication and acidification impacts can be evaluated sepa- rately or aggregated by evaluating the final damage on ecosystems. Damage (also called endpoint) assessment methods, representing the overall environmental impact of the lifecycle (i.e. aggregating all the intermediate categories) do exist (e.g. Ecoindicator99, Goedkoop and Spriemsma, 1999) but are not consensual yet and depend too much on value choices. Also, it can be very informative to evaluate the different environmental impacts separately, at midpoint categories, such as greenhouse effect, acidification and so on. For these reasons, a LCA essentially ends up with a broad list of environmental impact (or damage) results, which are related to the function studied. These results have to be interpreted and these are not always straightforward, especially in case of comparison be- tween different alternative scenarios fulfilling the same function, for example scenarios of electricity production from different types of coal, combustion plants or flue gas treatments. One scenario can show better performances for some impact categories and worse for others and often it is difficult to identify the best scenario(s). * Corresponding author. Present address: CRP Henri Tudor/CRTE, 66 rue de Luxembourg, L-4002 Esch/Alzette, Luxembourg. E-mail address: enrico.benetto@tudor.lu (E. Benetto). Contents lists available at ScienceDirect Environmental Modelling & Software journal homepage: www.elsevier.com/locate/envsoft 1364-8152/$ – see front matter Ó 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.envsoft.2008.04.008 Environmental Modelling & Software 23 (2008) 1461–1467