A regional assessment of the operational National Water Model in the San Francisco Bay area for the 2018–2019 water year

This study presents a preliminary assessment of the National Water Model (NWM) operational short-range streamflow forecasts in the San Francisco Bay area, California. Hourly streamflow forecasts with lead times up to 18 h were assessed using a benchmark forecast across varying flow regimes, reservoir regulations, and basin characteristics. The NWM performed well during high flow periods, with effective predictions up to 15 h ahead and correlation coefficients exceeding 0.5 for 9-hour forecasts. However, forecast skill declines after 4 h during low flow periods, primarily due to limitations in baseflow simulation in managed watersheds. The model exhibits systematic overprediction in regions with runoff coefficients below 0.325 and annual precipitation less than 700 mm/year, resulting in relative bias values greater than 1.2 in drier southern areas. In addition, model performance exhibits greater accuracy in northern regions, sustaining useful lead times of up to 8 h, compared to just 4 h in southern regions. While the NWM shows promise for high flow forecasting applications, further calibration and refinements are necessary to enhance baseflow representation and account for complex basin characteristics, particularly in dry and managed watersheds. These improvements are essential for achieving consistent forecast reliability across diverse hydrological conditions.