What is already known:
What this study adds:
The National Weather Service issues hundreds of hazardous weather advisories, such as tornado warnings each year. However, the current warning paradigm has led to false alarm rates of 70% - 80% (Simmons and Sutter, 2011) and a stagnancy in warning lead times. In an attempt to mitigate these issues, the Warn-on-Forecast (WoF) project is developing a regional, high-resolution, storm-scale prediction system capable of predicting hazardous weather phenomena on the 0-3 h time scale. A prototype system, the NSSL Experimental Warn-on-Forecast System for ensembles (NEWS-e) combines, surface observations, radar data, and satellite retrievals in an ensemble data assimilation framework to create rapidly-updating probabilistic forecasts. Testing of the real-time configuration of the NEWS-e was completed during the 2017 Hazardous Weather Testbed Spring Forecast Experiment (HWT-SFE).
Two separate experiments were performed on three events to evaluate how changes in the configuration affect the performance of the NEWS-e. The first was a data denial experiment where the METAR surface observations were removed from the data assimilation. The second was a switch from the NSSL two-moment to the Thompson microphysics scheme. Quantitative metrics, namely the probability of detection (POD) and false alarm rate (FAR) for rotational objects, were computed to examine the overall performance of each NEWS-e forecast. The real-time run, which includes the assimilation of all surface observations and employs the NSSL two-moment microphysics scheme, produced POD values ~ 20% higher compared to both experimental runs. Qualitative comparisons of reflectivity values and updraft speed between the real-time and Thompson runs are also presented. Lastly, the impacts of removing the METAR observations from the 16 May 2017 run are discussed.