One-Hour-Lead-Time Tornado Proxies in the Warn-on-Forecast System

Convection-allowing models routinely display some success in predicting the evolution of individual storms or storm clusters on the meso-beta scale. Even individual long-track supercells are realistically depicted at times, with considerable lead time. Experimentation with the Warn-on-Forecast System (WoFS), including real-world real-time feedback, has shown that forecasters gain greater confidence in these model depictions when rapidly-updating ensemble output is available and it can be verified that observed storms are well initialized in the ensemble. In numerous cases since 2017, NWS offices have used WoFS to provide longer lead time messaging on specific thunderstorms and specific tornado threats. The warn-on-forecast vision of warning-style messaging at 1-h or longer lead time appears to be within reach. In fact, several cases are documented in which WoFS tornado proxies at 1-h or even 2-h lead time suggested a high likelihood of tornadoes within 5–15 km of their eventual locations. This skill was particularly evident during the 31 March 2023 tornado outbreak during which multiple concurrent WoFS domains afforded the opportunity to study forecasts of numerous long-track tornadic supercells. This work will study the 31 March event as a starting point for quantifying the performance of WoFS tornado proxies at 1-h lead time for a high shear and moderate buoyancy environment. Of particular interest are WoFS forecasts of 1) supercells present in the 18-member probability matched mean of composite radar reflectivity with pockets of enhanced Significant Tornado Parameter (STP) values to their inflow side, 2) the trend and magnitude of the 90th percentile of 0-2 km vertical vorticity beneath a storm, and 3) WoFS machine-learning based probabilities of at least one tornado occurring in a probabilistic event space (representing the track of an individual storm cluster in WoFS). As WoFS transitions toward operational status it will be useful to identify workflows that forecasters may use to answer specific questions quickly, using only a few products; this work is an exploration of this for identifying tornadic storms.