Evaluation of Quantitative Precipitation Forecasts by the Hurricane Weather Research and Forecast (HWRF) model

Tropical cyclones pose a huge threat to human lives and property. Hurricanes and even weaker tropical storms can produce copious amount of rainfall and cause extensive inland flooding. As an example, tropical storm Allison (2001) produced over 40 inches of rain in the Houston, TX area causing 41 fatalities.

The Hurricane Weather Research and Forecasting (HWRF) model is a state-of-the-art operational NWP suite targeted for tropical cyclones (TCs). HWRF forecasts are primarily used by the NOAA National Hurricane Center and the Joint Typhoon and Warning Center as guidance on forecasting TC track, intensity, and structure. While the HWRF Quantitative Precipitation Forecasts (QPFs) could be useful as additional guidance for disaster management and planning, they are not systematically employed today by the National Weather Service National Centers and Weather Forecast Offices. One of the reasons for this omission is that systematic verification of the HWRF QPFs has not been performed. The aim of this study is to conduct extensive evaluation of the QPFs from the HWRF model to document the model's performance in making these forecasts, compare HWRF's QPF performance against other operational models (such as the GFS), and provide input to the developers for possible improvements.

One of the challenges of verifying TC QPFs is the impact of errors in the track forecasts, which strongly influence the verification results. Marchok et al. (2007) provided some guidance on evaluating QPFs for TCs through application of methods to reduce the impact of track errors on the verification analyses. A similar methodology is adopted in this study to shift the QPF based on track error. The verification is done against CMORPH rainfall estimates. The operational GFS is used as a baseline to ascertain whether a higher-resolution model is able to improve upon a global model. Initial results indicate that HWRF has a higher bias compared to GFS. The equitable threat scores reveal that the HWRF and GFS QPF performance is improved when the impact of track error is minimized. This result indicates that the models are providing useful information; however, the overall QPF performance is degraded by track errors.

This presentation will also include results of a comparison of the high-resolution HWRF forecast-total QPF against Stage IV data over land and and large scale HWRF QPF verification against CMORPH will also be presented.