Climate change is altering global hydroclimatology and threatening the world’s fresh water sources. In addition, population and land use change are stressing water sources in many regions across the Conterminous United States (CONUS). Reservoirs serve various purposes including water supply, flood control, hydropower, and irrigation. Reservoir releases are modified during droughts and floods to meet water demand without increasing downstream risk. Mismanagement of water resources may threaten agriculture and food supply chains, human and environmental health, and regional economies. Most reservoirs operate solely considering a short-range timescale, 1-7 days lead time, which is important for determining daily release values, meeting daily water supply and energy demands, and managing flash floods or storm events. Reservoir demand and use vary across spatial and temporal scales. Subseasonal-to-seasonal forecasts (S2S), 15-90 days lead time, are important for meeting medium-range sector demands, such as energy grid planning, maintaining necessary storage for irrigated agriculture, drought effect mitigation, and meeting ecological demands. This study develops a large-scale framework to address longer lead time forecasts for large-scale water resource management with statistical and machine learning (ML) models. Our large-scale model uses S2S inflow forecasts that are generated by a ML technique using the North American Multi-Model Ensemble (NMME) precipitation forecasts. The reservoir inflow forecasts are fed into a piecewise linear regression tree model (PLRT), based on reservoir characteristics, to optimize reservoir storage. This framework evaluates whether large-scale forecasting methods are viable for water resource managers to operate reservoirs considering longer time horizons.
