Performance evaluation of a water resources system under varying climatic conditions: Reliability, Resilience, Vulnerability and beyond

Summary As introduced by Hashimoto et al. (1982), Reliability, Resilience, and Vulnerability (RRV) metrics measure different aspects of a water resources system performance. Together, RRV metrics provide one of the most comprehensive approaches for analyzing the probability of success or failure of a system, the rate of recovery (or rebound) of a system from unsatisfactory states, as well as quantifying the expected consequence of being in unsatisfactory states for extended periods. Assessing these comprehensive metrics at current (baseline) and future scenarios provide insight into system performance in changing or varying climatic conditions. Such an approach makes it possible to analyze different scenarios that could include specific mitigation or adaptation strategies to accommodate a varying climate. The method requires a subjective decision defining what constitutes an “unsatisfactory state” depending on acceptable risks. The application of this methodology is demonstrated using Tampa Bay Water’s Enhanced Surface Water System. In this case, for each scenario, a thousand ensembles of 300-years of monthly stream flow traces were first generated by a multi-site rainfall/runoff model. Second, a novel nonlinear disaggregation algorithm was developed to translate monthly outputs into daily values. The daily stream flow traces and their derivatives are then used to drive complex operational models that produce several system variables (e.g., permitted river withdrawals, reservoir storage volumes, and treatment plant production rates) at different locations. Outputs from the operational model were then used to define criteria over which the RRV and other metrics were evaluated. Several mitigation scenarios such as treatment and reservoir capacity expansion, as well as adaptation through operational changes were considered to evaluate system performance under varying climatic conditions. The approach highlights the benefits of comprehensive system performance metrics that are easy to understand by decision makers and stake holders and demonstrates the implementation of seemingly intractable ensemble size and simulation length in a distributed computing environment.