Tailoring Hydrologic Modeling For Improved Water Resources Decision Support: An Approach To Ensemble Modeling
Presenter: Abigail Kahler1
Co-Author(s): -
Advisor(s): Dr. P.A. Ty Ferré
1Department of Hydrology and Atmospheric Sciences, University of Arizona
One of the challenges of hydrologic modeling is quantifying uncertainty. The size and characteristics of aquifers is such that initially small uncertainties magnify over time, increasing the costs of miscalculated decisions. Hydrologic models help predict consequences but are limited by sparse data and uncertainty. This suggests a need for multiple models, and it is worthwhile to pay special attention to less probable, still plausible models that predict consequential outcomes. These are called models of concern (MOCs). The Parameter ESTimation (PEST) suite of programs is the standard for performing uncertainty analysis. PEST iteratively runs models while varying input values to test the outputs for fit to measured data. In Pareto mode, the extent of an outcome is varied while minimizing an objective function. This function expresses the relative importance of contributing data. We define a simple objective function as the misfit between simulated and measured observations. We express a stakeholder’s level of satisfaction with an outcome through a utility function. A prediction may have low or high utility according to the associated consequences. Our objective function is paired with the utility function to identify a threshold below which a consequential outcome is unlikely to occur. Preliminary results suggest a combined ensemble of best-fitting and MOC models more accurately identifies consequential outcomes. Future work will test this method against more complex systems and explore how it may interface with PEST. This work considers the need for, and implications of, ensemble modeling in a new approach to better represent stakeholder concerns.