Determining Missed Events when Evaluating Probabilistic Convective Outlooks

In the early 2000s, the Storm Prediction Center (SPC) began publicly issuing probabilistic convective outlooks for individual severe weather hazards after only issuing categorical outlooks in the preceding decades. Previous studies have evaluated the accuracy and skill of these categorical convective outlooks, and also forecasts of significant severe weather, using “practically perfect” forecasts as a no-skill baseline by which the skill of SPC forecasts could be assessed. The approach used by these studies was appropriate for forecasts that were designed to include all severe hazards, but for forecasts of heightened risk (e.g. 10% probability of tornadoes), issues arise. This study demonstrates a method by which probabilistic convective outlooks at all risk levels can be evaluated for accuracy and skill by discriminating between observed storm reports relative to each outlook area. Through this approach, “missed event” days can be determined, and information about the rate at which the SPC correctly issues these forecasts is presented. In addition, preliminary measures of skill and accuracy using the proposed method are calculated and compared with that which was used previously.