PNNL

Abstract

Large-scale hydrological and water resource models (LHMs) require water storage and release schemes to represent the flow regulating effects of reservoirs. Owing to a lack of observed reservoir operations, state-of-the-art LHMs deploy a generic reservoir scheme that may fail to represent important local operating features as well as release decisions during extreme events. Here we introduce ResOpsUS_IP—a new dataset of inferred reservoir operating policies for all 1,906 large reservoirs of conterminous United States (storage capacity > 10MCM). ResOpsUS_IP relies on the ResOpsUS dataset of observed reservoir operations to infer reservoir operating rules for 545 data-rich reservoirs. These functions are developed in a standardized form, allowing for extrapolation of operating schemes to 1,361 data-scarce reservoirs and a first-ever national-scale inventory of empirically derived reservoir operating policies. Evaluation of the new scheme in daily simulation highlights substantial and robust improvement across six performance metrics relative to the popular Hanasaki method. Performance of the extrapolation scheme for data-scare dams is evaluated with a leave-one-out approach and is shown to also offer performance modest gains on average over the Hanasaki approach. ResOpsUSA_IP may be readily adopted in any LHM featuring large dams of the conterminous United States.