Evaluating the impact of WoFS-PHI on real-time severe weather forecasts during the 2024 Spring Forecasting Experiment

Warn-on-Forecast System-Probabilistic Hazard Information (WoFS-PHI) is a real-time machine learning algorithm that predicts the likelihood of severe weather at lead times between 0 and 4 hours based on a blend of WoFS and ProbSevere. During the 2024 Hazardous Weather Testbed Spring Forecasting Experiment (HWT SFE), WoFS-PHI was evaluated in a controlled, real-time severe weather forecasting activity, in which participants created 0-1- and 1-2-h severe weather probability forecasts for tornadoes, wind, and hail due at four fixed times. Each day, one group of participants received access to WoFS-PHI in the WoFS Viewer, while another group did not. These groups were kept separate, and participants rotated groups throughout the week. All participants completed a survey on their experience following the activity. In addition, all forecasts were subjectively evaluated by other SFE participants the next morning.

Preliminary results suggest that forecasts made with WoFS-PHI received subjective ratings about 0.5 points higher (on a 10-point scale) than those made without. Moreover, the 1-2-h forecasts made with WoFS-PHI had similar ratings as the 0-1-h forecasts made without, suggesting WoFS-PHI could help increase forecaster confidence in the severe weather threat at longer lead times. In their surveys, participants indicated that WoFS-PHI influenced both the location and magnitude of their forecast probabilities, and most participants indicated they had a moderate degree of trust in WoFS-PHI after each day. While the SFE activity did not perfectly mirror operations, results from the activity suggest WoFS-PHI is a powerful tool for short-term severe weather forecasting.