Comparison of Uncertainty Analysis for Community Based Watershed Models

Predictions from complex watershed models are relied on by policy-makers for resource allocations and development decisions. However, most managers do not have access to effective and practical approaches to represent the uncertainty within these systems and within the complex models; nor can they readily incorporate uncertainty into the decision process. In previous studies, Latin Hypercube Sampling combined with generalized likelihood uncertainty estimation (LHS/GLUE) and optimization was utilized to find reliable load allocations for the watershed of Moore's Creek, Virginia, given parameter uncertainty and uncertainty in the initial loads of fecal coliform bacteria. Although this approach increased load allocation reliability, as compared to a margin-of-safety method, the uncertainty analysis was computationally intensive and research intensive. In light of this rigorous nature, there is concern that many managers and community-based modelers will rely on simpler, less accurate approaches. New methods of uncertainty analysis have been developed that could significantly reduce the amount of computational effort required. This study integrates two methods of uncertainty analysis, LHS/GLUE and Dynamically Dimensioned Search, with the HSPF model and evaluates their performance on the Moore's Creek case study. The purpose of this paper is to present a methodology for comparing computational efficiency, ease of implementation, and accuracy of results. Results from this study will be presented at the conference and will help guide selection of an efficient choice of uncertainty analysis for use by managers and community modeling projects.