Impacts of Risk Aversion on Forecast-Informed Reservoir Operations for Managed Aquifer Recharge

Increasing stress on water availability, driven by rising water demand and climate variability, presents a global challenge. By leveraging advanced ensemble streamflow forecasts and conjunctive management, reoperating existing dams with Forecast-Informed Reservoir Operations coupled with Managed Aquifer Recharge (FIRO-MAR) offers a cost-effective strategy to achieve long-term water sustainability. However, the inherent uncertainty in streamflow forecasts introduces associated risks, necessitating consideration of risk perception respective to the forecast uncertainties in real-world operation decisions. Here, we explore to what extent the value of forecasts in FIRO-MAR is sensitive to operators’ risk aversion levels, which can help us better identify specific risk aversion levels that limit the use of forecasts. We develop a multi-objective reservoir model integrated with downstream groundwater conjunctive use to support additional groundwater recharge purposes. We conduct historical (1975-2014) and future (2055-2099) simulations for 1183 flood control reservoirs worldwide with three reservoir operation schemes: a baseline operation, a MAR operation that incorporates conjunctive use, and a FIRO-MAR operation that further employs inflow forecasts. River discharge reforecasts from the Global Flood Awareness System (GloFAS) are used to inform operation policies under varying drought risk aversion levels, represented by selected forecast ensemble quantiles. Our results reveal trade-offs between groundwater recharge and non-recharge objectives, such as hydropower generation, while highlighting synergies in water supply through conjunctive management. Under perfect forecasts, FIRO-MAR can boost the global potential of rechargeable water by approximately 14% compared to MAR alone. However, operational forecast uncertainties diminish these gains variably across regions, influenced by reservoir characteristics and background climate. Notably, forecasts prove to be more valuable for risk-averse operators, though this relationship is constrained by forecast skill and reservoir-specific factors such as inflow-to-storage ratios. Our findings elucidate how operators’ risk attitudes influence the strategic use of forecasts in supporting MAR, offering crucial insights for FIRO-MAR implementations.