Variability, Attributes, and Drivers of Optimal Forecast-Informed Reservoir Operating Policies for Water Supply and Flood Control in California.

Reservoirs balance multiple conflicting objectives, including flood control and water supply. In California, shifts in seasonal hydrologic patterns under climate change will amplify the difficulties in balancing flood control with water supply. Current flood control policies are based on fixed seasonal rule curves determined by the observed timing and magnitude of floods in the record. These rule curves generally require the release of wet season inflows, reducing the available stored water for use during the dry season. Here we investigate the potential for forecast-informed reservoir operations (FIRO) to increase water supply availability while minimizing additional flood risk at 14 reservoirs in the Sacramento, San Joaquin, and Tulare river basins. We use a differential evolution algorithm to train risk-based reservoir operation policies with an ensemble of historical forecasts over the period 2013–2023. Results show an average 8.1% increase in storage normalized by capacity, though this varies across reservoirs. The forecast-informed policies also reduce the occurrence of high-magnitude releases throughout the system. The accumulation of benefits is sensitive to the timing and magnitude of flood events, and most of the cumulative benefit is obtained during a few years. Under cross-validation, we find that large floods are needed in the training data to avoid overfitting the policy. We further examine the relationship between reservoir properties and FIRO benefits, finding that the ratio of peak inflow magnitude to maximum safe release correlates with increased storage under the FIRO policy, while the ratio of mean inflow to capacity correlates to the reduction of high-magnitude releases. This study highlights how adaptive reservoir management policies can yield water supply benefits without an increase in flood risk, given adequate historical data for policy training. These policies may be a valuable adaptation to climate change but require careful validation and out-of-sample testing. [ABSTRACT FROM AUTHOR]