Decision Making Under Uncertainty: A New Approach to Decision Making in a Multi-Model Environment

A primary role of hydrogeologic models is to support decision making.  However, the uncertainty inherent in these models complicates their use as decision making tools.  Increasingly, multimodel methods are used to develop more robust predictions of future system states and include measures of model associated prediction uncertainties.  Given a set of predictions from a multi-model analysis, one must undertake the process of choosing a prediction to guide decision making.  Outputs from a multi-model analysis are often reduced to single models (e.g., maximum likelihood, worst case), or averaging approaches (e.g., Bayesian model averaging, likelihood weighting) when used to make hydrologic decisions.  With the exception of the worst case method, these approaches consider the range of model predictions, but they do not consider the implications of different predictions for decision making; a consideration that is accepted as standard in formal quantitative risk analysis.  This talk presents an approach that considers the outcomes of decisions based on multi-model hydrologic predictions and demonstrates that it can lead to improved decision making and engineering design under uncertainty.   Examples will be presented from synthetic case studies that illustrate the validity of the method and from real world groundwater flow and transport studies that demonstrate the practical application of the method.