Investigation of the Added Value of Using Statistically Postprocessed GEFS Ensemble Forecasts as Alternative Forcings for the WRF-Hydro National Water Model (NWM)

We present results of a project aiming to improve the characterization of uncertainties in NWM streamflow predictions through the implementation and evaluation of advanced statistical postprocessing methodology for meteorological inputs to the NWM. Forecasts of the eight forcing variables (2-m temperature, surface pressure, u/v wind speed, and precipitation rate, mixing ratio, longwave, and shortwave radiation) from NOAA's Global Ensemble Forecast System (GEFS) are first postprocessed separately for each grid-point and each forecast lead time. Subsequently, a variant of the Ensemble Copula Coupling (ECC) method is used to generate high-resolution, spatially and temporally consistent, and statistically reliable ensemble forecast fields.

These postprocessed and downscaled GEFS ensemble forecasts are used as inputs to the NWM, and the resulting streamflow forecasts are evaluated over different time scales for selected watersheds across California, focusing on gauges where data of observed, unimpaired flow is available. Our evaluation focuses on streamflow at the basin scale and compares to the current operational GFS-based NWM forecasts. We discuss the added value of the probabilistic uncertainty estimation for both the whole analysis time frame and for selected extreme event cases.