SWOT-initialized reservoir operation forecasting: Validation in data-rich and application to data-limited regions

Abstract. Challenges in reservoir operation are posed by the fact that reservoir release trajectories depend on both uncertain inflow forecasts and (in cases where in situ observations are not available) imperfect knowledge of the initial reservoir state. The recently launched (science data available since late 2023) Surface Water and Ocean Topography (SWOT) satellite provides global-scale reservoir water level and surface extent observations, creating new opportunities to constrain reservoir storage states for forecast initialization. Here we develop a SWOT-initialized reservoir operation forecasting framework and evaluate it both for two data-rich California reservoirs and a data-limited reservoir in Ethiopia. The operations framework combines SWOT-derived reservoir storage, monthly inflow forecasts, a reservoir operation model, and a downstream agricultural water demand-satisfaction assessment. Results from California reservoirs show that SWOT-derived storage provides useful initial conditions for seasonal reservoir operation forecasts (similar to those available based on in situ data), with water demand satisfaction approaching that obtained using in-situ storage initialization. The resulting storage and release forecasts capture the major seasonal dynamics. Although Ensemble Streamflow Prediction (ESP) (developed using historic observed reservoir inflows) generally provides more accurate inflow and reservoir-operation forecasts, a global reservoir inflow product (GloFAS) produced without local observations yields water demand satisfaction patterns that are broadly consistent with the in situ results Application of the framework to the African reservoir (where essentially no in situ observations are available) demonstrates that, for the two years of SWOT data, nominal irrigation water demands are generally satisfied.