Computer support for environmental multiple criteria decision analysis under uncertainty

Abstract

This thesis describes new techniques for Multiple-Criteria Decision Analysis (MCDA) under uncertainty with application to environmental problems. Decision support systems, such as the SEAL (Stochastic Environmental Analysis), REAL (Robust Environmental Analysis), and MEAL (Marginal Distributions for Environmental Analysis) systems, are developed to help decision makers improve their social, economic, and environmental decision making under uncertainty.

A major contribution of this thesis is the investigation of uncertainty approaches, including interval judgments (Saaty and Vargas, 1987), info-gap models (Ben-Haim, 1996), stochastic differential equations (Cox and Miller, 1965), and Bayesian techniques (Ludwig, 1996) in an MCDA context. For example, the proposed MCDA info-gap model approach is completely non-probabilistic; it captures a decision maker's preferences and attitude toward risk without resorting to "non-intuitive probabilistic concepts of gambling and indifference between lotteries" (Barzilai, 1997) and is the first published info-gap model MCDA technique in the literature. Convex modeling is particularly valuable since utility functions are not required (only value functions are necessary) and robust alternatives can be identified.

Significantly, the stochastic water quality models used in the thesis investigates the use of Stratonovich calculus to model the classic interactions among biochemical oxygen demand (BOD), dissolved oxygen (DO), and other environmental variables. In addition, the Streeter-Phelps equations are generalized to more realistically model hydrologic processes. Finally, a practical colored noise approximation is put forth and used to replace the abstract mathematical concept of 'white' (theoretical) noise. Replacing white noise with coloured noise is of great importance in water quality modeling since in almost all cases the white noise assumption is not justified and is used only for mathematical convenience.

Finally, the SEAL decision support system is applied to a wide range of stochastic environmental problems, from water quality modeling to species extinction. Here the 'First Passage Time' problem is considered in detail from an environmental perspective. In the context of fisheries management, it is shown how regulating the 'fishing effort' can significantly reduce the risk of stock extinction.

Finally, it is described how the management of renewable resources, where it has been practiced at all, relies heavily on techniques from optimal control theory, cost-benefit analysis, and maximum sustainable yield (MSY). These approaches are critically reviewed and it is shown that formally modeling both the risk of extinction and the 'preservation value' of a resource can improve the sustainable management of renewable resources.