innovation, or change in ecology or human health. The preference index then leads to a partial ranking of the policy options under consideration and recommendation of an “optimal” set of choices or competitive choices (Brans and Vincke 1985). MCDM has been applied successfully in environmental decision-making (Moffett and Sarkar 2006; Hajkowicz and Collins 2007); however, criterion-specific constituents of the preference index for each policy option are affected by the quality of the science and evidence, scaling, and other factors that can limit validity (Hajkowicz and Collins 2007).

An alternative to single-objective formulations is to provide decision-makers with the Pareto optimal set of nondominated candidate solutions. Essentially, the Pareto optimal set is constructed by identifying decisions that can improve one or more objectives without harming any other. Use of the Pareto optimal set does not determine a single preferred approach but presents decision-makers with a smaller set of options from which to choose. The concept of Pareto optimal sets is not new, but the capacity to apply it in decision-making has been greatly expanded by recent methodologic advances in optimization techniques (most notably multiobjective evolutionary algorithms) and computation of Pareto sets for large complex problems, and this has increased the scope of environmental and other applications (Coello et al. 2007; Nicklow et al. 2010). Rabotyagov et al. (2010) give an example of evolutionary computation for the analysis of tradeoffs between pollution-control costs and nutrient-pollution reductions. Optimal sets of air pollution control measures have been developed that consider aggregate health benefits and inequality in the distribution of those benefits as separate dimensions (Levy et al. 2007). Kasprzyk et al. (2009) demonstrate how multiobjective methods can be used to inform policies for the management of urban water-supply risks that are caused by growing population demands and droughts. Multiobjective optimization in support of environmental-management decisions is especially compelling given the emerging paradigm of managing for multiple ecosystem services and consideration of cumulative risks for human health. Tradeoffs and complementarities can exist between alternative services and between other relevant performance metrics (for example, public and private costs and distribution outcomes by location or income class). Applications of multiobjective optimization methods would promote the explicit specification of preference indices relevant to environmental decision-making and science to quantify outcomes and evaluate tradeoffs; all this would serve to improve the transparency and scientific soundness of decisions.

Addressing Uncertainty in Complex Systems

With any of the solutions-oriented approaches delineated above, regardless of which analytic tools or indicators are used by EPA to support decisions in the future, uncertainty will be an overriding concern. With increasingly complex multifactorial problems and a push for tools that are sufficiently timely and

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement