Considering Model Parameter Uncertainty in Multi-objective Robust Optimizations: Implications for Decision Making in Multi-sector Dynamical Systems
Jared D. Smith,
Laurence Lin, Julianne D. Quinn, and Lawrence E. Band
Abstract: MultiSector Dynamics studies integrate models from environmental, economic, social and other sectors to describe system behaviors. The resulting models are complex, and calibration can result in many equifinal parameter sets. How sensitive are planning and management objective values and subsequent decisions to such uncertainty in model parameter values? This talk presents a catchment reforestation case study wherein optimal locations and numbers of trees to reduce stormwater volumes depend on the spatial distribution of runoff and streamflow in the catchment. Global sensitivity analysis using decision-relevant metrics is used to inform the selection of parameters to calibrate; however, calibration data are only available at the catchment outlet. Equifinal model parameterizations for the outlet can result in uncertainty in the locations and magnitudes of streamflows across the catchment, which can lead to different optimal reforestation locations for different parameterizations. Multi-objective robust optimization (MORO) has been proposed to discover reforestation designs that are robust to such parametric model uncertainty. However, it has not been shown that this actually results in better decisions than optimizing to a single, most likely parameter set, which would be less computationally expensive. This talk evaluates the value of considering parametric uncertainty in designing robust water systems despite the additional computational cost. Implications for robust decision making in multi-sector dynamical systems are postulated to stir discussion within the community.
Biography: Jared Smith is a Postdoctoral Research Associate in Engineering Systems and Environment at the University of Virginia. His research focuses on coupling physical and mathematical models with statistical analyses to inform planning and management decisions for environmental, earth-energy, and water resources systems. This talk will present Jared’s current research on discovering green infrastructure portfolios that are optimized to be robust to uncertainty in Bayesian-calibrated watershed model parameter values. Previous work has addressed the feasibility of direct-use of geothermal energy for the district heating systems of the Cornell University and West Virginia University campuses.