Decision analysis (DA) is the product of integrating utility theory, probability,
and mathematical optimization (see French, 1990; Morgan and Henrion, 1990; Chechile
and Carlisle, 1991; Keeney and Raiffa, 1993; Kleindorfer *et al*., 1993;
Marshall and Oliver, 1995; Clemen, 1996). The process starts with problem identification
and preparation of a possibly comprehensive list of decision options. Structural
analysis would organize options into a decision tree, carefully distinguishing
decision nodes (points at which the outcome is chosen by the decisionmaker)
and chance nodes (points at which the outcome results from stochastic external
events). Next, uncertainty analysis would assign subjective probabilities to
chance nodes, and utility analysis would stipulate cardinal utilities (in terms
of absolute values) for outcomes. Finally, optimization produces the best outcome
according to a selected criterion, typically maximizing expected utility or
any other that best reflects the risk attitude of the decisionmaker.

Advanced DA provides various extensions of the foregoing conceptual framework and supports a huge diversity of applications. In the literature, some features (sequential decisionmaking, hedging), specific versions (multi-criteria analysis), distinctive applications [risk assessment (RA)], or basic components (multi-attribute utility theory) of DA sometimes are emphasized and taken as separate DAFs, although they all are rooted in the same theoretical framework. As indicated, sequential decisionmaking is an indispensable mode of analysis of climate change in any DAF. It refers to framing of the analysis rather than a distinctive DAF. DA can be performed with a single criterion or with multiple criteria; multi-attribute utility theory provides the conceptual underpinnings for the latter. Finally, DA adapted to managing technological, social, or environmental hazards constitutes part of RA, in which a range of other methods also is available. RA involves estimation of the nature and size of risks. Its objective is to identify quantitative measures of hazards in terms of magnitude and probability. RA methods are diverse; the choice depends on the disciplinary focus and the nature of the hazard to be assessed, but all methods rely on extrapolation (see Kates and Kasperson, 1983). See Chapter 12 for applications of RA in climate impact assessment.

DA is a promising DAF for use in adaptation assessments. Problem formulation in DA allows for consideration of a broad range of uncertain outcomes, different probability distributions assigned to them, and a variety of possible adaptation actions. Structuring the DA model in an intertemporal fashion is helpful for identifying robust adaptation strategies that prove to be effective under a broad range of possible futures and retain a sufficient degree of freedom for course correction.

Other reports in this collection |