Examining how the spread in the High Resolution Rapid Refresh Ensemble translates into National Water Model streamflow forecasts

The National Water Model (NWM) provides streamflow forecasts for the continental US at high spatial and temporal resolution. The operational short-range forecast configuration is deterministic and is forced by hourly forecasts of precipitation, wind, temperature, radiation, and humidity from the High Resolution Rapid Refresh Model (HRRR) Numerical Weather Prediction Model. In this study, we produce an experimental ensemble of streamflow forecasts for Northern California using the experimental 9-member HRRR Ensemble (HRRRE) to produce nine hydrologic forecasts for an extreme precipitation event that occurred in January 2021. We first examine the performance of the HRRRE, comparing forecasts of the NWM forcing variables to station observations as well as evaluating ensemble spread and uncertainty characteristics. We then evaluate the resulting streamflow forecasts against corresponding gauge observations, as well as examine the spread of the ensemble of streamflow forecasts. Relationships between the uncertainties in the HRRRE inputs and NWM forecast outputs will be discussed. Particular focus will be placed on the timing and location of precipitation phase transitions in the HRRRE forecasts and their impact on NWM streamflow output.