Decision Analysis Utilizing Data from Multiple Life‐Cycle Impact Assessment Methods Part I: A Theoretical Basis

Numerous methodologies for the life‐cycle impact assessment (LCIA) step of life‐cycle assessment (LCA) are currently in popular use. These methods, which are based on a single method or level of analysis, are limited to the environmental fates, impact categories, damage functions, and stressors included in the method or model. Because of this, it has been suggested within the LCA community that LCIA data from multiple methods and/or levels of analysis, that is, end‐point and midpoint indicators, be used in LCA‐based decision analysis to facilitate better or, at least more informed, decision making. In this (two‐part) series of articles, we develop and present a series of LCA‐based decision analysis models, based on multiattribute value theory (MAVT), which utilize data from multiple LCIA methods and/or levels of analysis. The key to accomplishing this is the recognition of what LCIA damage indicators represent with respect to decision analysis, namely, decision attributes and, in most cases, proxy attributes. The use of proxy attributes in a decision model, however, poses certain challenges, such as the assessment of decision‐maker preferences for actual consequences that are only known imprecisely because of inherent limits of both LCA and scientific knowledge. In this article (part I), we provide a brief overview of MAVT and examine some of the decision‐theoretic issues and implications of current LCIA methods. We illustrate the application of MAVT to develop a decision model utilizing damage indicators from a single LCIA methodology; and, we identify the decision‐theoretic issues that arise when attempting to combine LCIA indicators from multiple methods and/or levels of analysis in a single decision model. Finally, we introduce the use in our methodology of constructed attributes to combine related end‐point damage indicators into single decision attributes and the concept and evaluation of proxy attributes.