Ensemble streamflow forecasting across the U.S. middle Atlantic region with a distributed hydrological model forced by GEFS reforecasts

Abstract

he quality of ensemble streamflow forecasts in the U.S. middle Atlantic region (MAR) is investigated for short- to medium-range forecast lead times (6-168 h). To this end, a regional hydrological ensemble prediction system (RHEPS) is assembled and implemented. The RHEPS is in this case comprised by the ensemble meteorological forcing, a distributed hydrological model, and a statistical postprocessor. As the meteorological forcing, precipitation and near surface temperature outputs from the National Oceanic and Atmospheric Administration?s (NOAA?s) National Centers for Environmental Prediction (NCEP) 11-member Global Ensemble Forecast System Reforecast version 2 (GEFSRv2) are used. The Hydrology Laboratory-Research Distributed Hydrologic Model (HL-RDHM) is used as the distributed hydrological model and a statistical auto-regressive model with an exogenous variable is used as the postprocessor. To verify streamflow forecasts from the RHEPS, 8 river basins in the MAR are selected, ranging in drainage area from ~262 to 29,965 km2 and covering some of the major rivers in the MAR.The verification results for the RHEPS show that, at the initial lead times (1-3 days), the hydrological uncertainties have more impact on forecast skill than the meteorological ones. The former become less pronounced, and the meteorological uncertainties dominate, across longer lead times (>3 days). Nonetheless, the ensemble streamflow forecasts remain skillful for lead times of up to 7 days. Additionally, postprocessing increases forecast skills across lead times and spatial scales, particularly for the high flow conditions. Overall, the proposed RHEPS is able to improve streamflow forecasting in the MAR relative to the deterministic (unperturbed GEFSRv2 member) forecasting case.