Robust Adaptation of Multi-Objective Reservoir Operations to Snowpack Decline Under Climate Change

While climate-induced changes in annual runoff volumes remain highly uncertain, snowpack decline due to rising temperatures is projected with high confidence. The continuing shift from snow to rain will move reservoir inflow magnitudes earlier in the water year, increasing vulnerabilities to both floods and droughts and reducing the predictive ability of seasonal snowpack-to-streamflow forecasts. In response to these challenges, this study develops robust adaptations to seasonal reservoir operations to support multiple objectives in snowmelt-driven systems under climate change. Several specific adaptations are explored: a dynamic flood control rule curve, release curtailment rules, and revised snowpack-to-streamflow forecasting methods. We employ a multi-objective policy search to find optimal combinations of these adaptations under a randomly selected subset of end-of-century runoff projections, holding out remaining scenarios for validation. These methods are demonstrated on a case study of northern California's multi-reservoir system, where the policy search maximizes the upstream objectives of flood control, carryover storage, and hydropower generation, along with the downstream objectives of environmental outflows and agricultural water deliveries. Relevant multi-objective tradeoffs are obtained, representing a range of possible climate adaptation policies. Baseline operating rules show significant vulnerabilities to increasing flood and drought risk, which many of the optimized policies are able to overcome using an intra-annual hedging strategy. These candidate policies use combinations of the proposed adaptations to outperform the historical policy in all five objectives. The resulting operations are interpretable and robust in validation across the uncertain ensemble projections. This policy search method for assigning adaptations buffers the effects of changing seasonal runoff patterns due to snowpack decline, while in turn mitigating multi-objective vulnerabilities to uncertainty in future patterns in flood and drought risk.