Improving NWM forecasts necessitates an ability to diagnose, quantify, and characterize errors in forcings (e.g., precipitation, temperature, and other fields that drive the hydrologic model) and their relationship to hydrologic forecast errors. Such ability is of particular importance considering hydrologic model calibration practices; that is, it is critical that the NWM is not calibrated to compensate for systematic errors in the forcing inputs (which are likely to change by region, phenomena, and model version upgrades). Moreover, diagnosing the contribution of errors in forcings relative to hydrologic model errors as specific functions of region, precipitation type, and forecast lead time is necessary for guiding future development of the NWM.
We will present preliminary results from a regional prototype study that begins to address these challenges by examining relationships between precipitation, temperature, and other forecast forcing errors from the High Resolution Rapid Refresh (HRRR) model and NWM streamflow errors found in short-term HRRR-driven hydrologic forecasts. We will highlight the necessity of combining multiple complementary evaluation approaches (multivariate statistical analysis, continuous-, contingency-, and object-based verification methods, and in-depth case study analysis) in order to adequately understand error types, correlations, and potential root causes.