Advancing inflow forecasting with CIROH and Reclamation

Abstract

Inflow forecasting is central to hydropower operations, where reliable projections inform reservoir management, power generation, and risk mitigation. AI-based forecasts have drawn attention for their potential to enhance forecast skill, but traditional process-based models retain important advantages, such as interpretability and well-established operational use. Hybrid approaches that combine physical and AI-based models are emerging as promising solutions to combine the benefits of the improved accuracy of AI and interpretability of traditional methods.
The Cooperative Institute for Research to Operations in Hydrology (CIROH) is pursuing multiple pathways to improve inflow forecasting, including AI-driven methods, hybrid modeling frameworks, large-sample parameterization strategies, and dynamic multi-model ensemble weighting. In parallel, the U.S. Bureau of Reclamation has invested in exploring alternative forecasting approaches, including sponsoring open competitions to benchmark and advance new techniques and improve snow modeling and data assimilation. Reclamation recently began an effort to develop a forecast portal aimed at collecting and evaluating alternative forecasts for use at reservoirs across the nation.
This presentation will highlight recent advances under CIROH and Reclamation initiatives, discuss the range of methods under exploration, and describe the goals of the forecast portal as a testbed for advancing inflow forecasting in support of reservoir operations.