Advancing Model Diagnostics To Support Hydrologic Prediction And Water Resources Planning Under Uncertainty
Computational models are essential tools for prediction and planning in water resources systems to ensure human water security and environmental health. Water systems models merely approximate the processes by which water moves through natural and built environments; their value depends on assumptions regarding climate, demand, land use, and other uncertain factors that may influence decision making. Numerical techniques to explore the role of these uncertain factors, known as diagnostic methods, can highlight opportunities to improve the accuracy of prediction as well as identify influential uncertainties to inform additional research and policy. This dissertation advances diagnostic methods for water resources models to identify (1) time-varying dominant processes driving modeled hydrologic predictions in flood forecasting, and (2) tradeoffs and vulnerabilities to changing climate and demands in regional urban water supply systems planning for drought. This work proposes diagnostic methods as a key element of a posteriori decision support, in which decision alternatives and vulnerable scenarios are identified following computational modeling and data analysis. Consistent with this theme, this work follows a multi-objective approach in which stakeholders can analyze tradeoffs between conflicting objectives as part of an iterative constructive learning process. For a spatially distributed flood forecasting model, results show that dominant uncertainties vary in space and time, and can inform model-based scientific inference as well as decision making. Similarly, the results of the urban water supply study indicate that sensitivity analysis can suggest costeffective paths to mitigate vulnerability to deeply uncertain future scenarios, for which likelihoods remain unknown or disputed. The multi-objective approach allows stakeholders to explore tradeoffs in their modeled robustness to inform intra-regional policies such as transfer contracts and shared infrastructure investments. Bridging the areas of hydrology and water systems planning is increasingly valuable, as hydrologic modelers begin to incorporate anthropogenic influences on the water cycle, and water systems planners begin to explore uncertainty in hydrologic process representation. In summary, this work develops diagnostic methods to identify time-varying dominant processes in distributed flood forecasting as well as tradeoffs and vulnerabilities under change in regional urban water supply, ultimately seeking to improve model-based planning for extreme floods and droughts in water resources systems.
Water Resources Systems Engineering; Multi-Objective Decision Support; Sensitivity Analysis & Model Diagnostics
Walter,Michael Todd; Stedinger,Jery Russell
Civil and Environmental Engineering
Ph. D., Civil and Environmental Engineering
Doctor of Philosophy
dissertation or thesis