A Statistical Approach to Diagnosing Storm Mode

Abstract

Determining storm mode (linear or isolated) is a crucial component of any severe weather forecast. Isolated storms are associated with a greater likelihood of significant (EF2+) tornadoes and very large (2”+) hail, while linear storms are more likely to produce straight-line wind damage. Current operational convection allowing models (CAMs), which are often used to diagnose storm mode, only run up to 48-60 hours into the future and can quickly lose accuracy with increasing lead time. To improve forecast accuracy and messaging on Day 3+ outlooks, a forecast tool was created to predict storm mode using only synoptic-scale variables. The approach uses a blend of theoretical modeling, stochastic modeling, and statistical modeling. The formulation generally performed well with reproducing past events and predicting future events 84+ hours in advance using 0.5° Global Forecasting System (GFS) and 0.5° Global Ensemble Forecasting System (GEFS) outputs.