Effect of Correcting Biases in HRRR Nonprecipitation Forcing Fields for the National Water Model Configuration of WRF-Hydro

Streamflow and other hydrologic variables predicted within the National Water Model (NWM) configuration of WRF-Hydro are a product of the forcing fields input into the model. This study focuses on the non-precipitation inputs from 18-hour HRRR forecasts, including temperature, wind, humidity, and radiation fluxes. We evaluated these fields at ~1,000 stations from several observation networks centered on the contiguous U.S.. Wind speed has a slight low bias of a few tenths m s-1, especially at earlier lead times, while temperature and humidity have more complex biases that vary by lead time, location, annually, and diurnally. Shortwave radiation has a positive bias that is strongly a relationship of solar zenith angle and lead time, which is partially offset by a negative (-9 W m-2) net longwave bias spread out fairly evenly over all hours of the day. We have started by correcting for the first-order biases for each of these fields, and will present results comparing output from hindcast NWM simulations with and without bias corrections to each of these fields. We will then demonstrate which bias corrections resulted in significant improvement in the model output, and what further corrections will be made to further improve the results. In addition to improving streamflow prediction, it is the hope that this study will help shed light on model sensitivities to structural biases within incoming forcing products.