Strategies for Restoring River Ecosystems: Sources of Variability and Uncertainty in Natural and Managed Systems
10. Decision Support Models as Tools for Developing Management Strategies: Examples from the Columbia River Basin
The Independent Scientific Advisory Board
Abstract.—We examine decision-support models designed to help recover salmon Oncorhynchus spp. in the Columbia River Basin as a case study for the use of models to help resolve scientific uncertainty and select management options. The models all have somewhat different objectives, use different data, and deal with a variety of salmon-related issues. Divergence of model outputs has, in the past, been used to justify different policy positions, leading some to conclude that science has failed to provide clarity to salmon recovery planning. Three distinct approaches are represented in the models: decision analysis, statistical, and expert system. Of the three approaches, decision analysis provides the clearest management advice and the most formal method for treating uncertainty. Its success depends on the engagement of decision makers in framing questions, identifying management options under consideration, and assigning values to possible outcomes. However, decision analysis could be very difficult to perform. As an alternative, the statistical model is the traditional scientific approach and it can operate with a large degree of detachment from policy. Statistical models proceed by testing hypotheses and estimating life-cycle parameters with available data. They have the advantage of scientific clarity, rigor, and empirical objectivity. The limitation of a statistical model is that the scope of the questions and their answers are restricted by availability of data, and in a domain that is data-poor, many pressing questions go unanswered. Expert system approaches fill gaps in data with expert opinion. In the context of salmon recovery, expert opinion allows consideration of the most concrete menu of specific options for salmon management. Expert opinion is a weaker basis for scientific prediction than is a mathematical relationship validated with empirical data. However, at the level of spatial resolution and environmental detail required to make salmon management decisions affecting the entire Columbia River Basin, there are no validated mathematical formulae for predicting the effects of management actions on salmon, and no adequate data archive exists for deriving such relationships. Communication between scientists and managers is improved when there is a formal institutional mechanism for summarizing scientific results and clarifying the interpretation of models for policy makers. If a modeling effort is driven by a desire to contribute to a particular decision, it is helpful to initially invest in enough communication to ensure that the model really is addressing the right question. Scientists can help managers craft decision rules that are formalized before analyses are undertaken. Decision rules define what measurements will be made, what statistical operations will be performed, and what threshold magnitudes of estimated quantities at specified levels of certainty will serve as criteria for the decision. Such specifications ensure that model results are properly used in the decision process. Committing to these specifications in advance helps dispel suspicions that analyses may be manipulated to achieve a particular outcome.