Improving WoFS-PHI Watch-to-Warning Guidance with new Observational Data

Warn-on-Forecast System-Probabilistic Hazard Information (WoFS-PHI) is a real-time machine learning algorithm that predicts individual severe weather hazards (hail, wind, and tornadoes) for up to 4 hours of lead time. The original version of WoFS-PHI used predictors from WoFS and ProbSevere Version 2 (PS2), and previous work found that including both datasets led to more skillful WoFS-PHI forecasts. It was found that the severe hail forecasts benefited most from the combined datasets, and severe wind and tornado forecasts benefited least from the PS2 data. It was hypothesized that incorporating improved observational datasets as WoFS-PHI predictors could enhance the forecast skill for all three hazards, but especially for severe wind and tornadoes. Experiments were conducted to test the impact of using ProbSevere Version 3 (PS3) instead of PS2 as well as Tornado Probability Algorithm (TORP) predictors.

Early results show that using PS3 instead of PS2 predictors improves WoFS-PHI skill, reliability, and sharpness when forecasting all 3 hazards. The inclusion of TORP modestly improves tornado prediction (especially when using PS2 predictors) but does not appreciably impact the wind or hail forecasts (regardless of the ProbSevere version). This work shows the value of combining observational and numerical weather prediction data and demonstrates the importance of using the best predictor data sources available.