Generating Probabilistic Tornado Guidance in a Warn-on-Forecast System

The Warn-on-Forecast (WoF) project is tasked with producing probabilistic guidance of hazards within individual convective storms, including tornadoes. Prototype WoF ensembles run in real-time since 2016 have demonstrated the capability to provide skillful probabilistic forecasts of mesocyclones. However, explicit tornado prediction requires far higher resolution than is currently possible in these real-time WoF systems. Therefore, this study assesses the potential of combining WoF mesocyclone forecasts with near-storm environmental information to create probabilistic tornado forecasts.

The probability a mesocyclone predicted by a WoF system will be tornadic is defined using the climatological relationship between the near-storm significant tornado parameter (STP) and tornado occurrence in right-moving supercells. First, an object-based framework is used to identify mesocyclones associated with right-moving supercells within individual WoF member forecasts. Inflow STP values are then extracted for each mesocyclone object, allowing a probability of tornado occurrence to be prescribed. These individual member probabilities may then be aggregated across the ensemble to produce a single, probabilistic, storm-scale tornado forecast.

The ensemble probability of tornado occurrence is sensitive to several aspects of the methodology, including thresholds used for object identification, the STP extraction method, and aggregation techniques to produce a final ensemble forecast. Therefore, the consistency and quality of probabilistic forecasts produced using several variations of the methodology are assessed using object-based verification against Multi-Radar Multi-Sensor rotation tracks associated with tornado reports. The reliability and contingency-table based skill scores of WoF tornado probabilities generated for over 75 cases from 2017–2019 will be presented.