Global Ensemble Forecast System (GEFS) Predictions of Days 1–15 U.S. Tornado and Hail Frequencies

Abstract

U.S. severe convective storms have grown to represent a 10 billion dollar annual peril for the insurance industry, and their accurate prediction remains a challenging task. This study examines days 1–15 severe convective storm predictions from the Global Ensemble Forecast System (GEFS). GEFS forecasts are based on the Supercell Composite Parameter and verified against spatially smoothed tornado and hail reports over the periods 1 March to 31 May 2016–2017. Skill is analyzed for deterministic forecasts (ensemble mean Supercell Composite Parameter exceeding specified thresholds) and probabilistic forecasts (fraction of ensemble members exceeding specified thresholds). Deterministic forecasts of tornado and hail activities are statistically more skillful than a random no-skill reference to days 9 and 12, respectively. Probabilistic forecasts are skillful relative to climatological no-skill reference to day 9 for tornado and day 12 for hail activity. These results provide a useful baseline for further improvement of tornado and hail forecasts at these ranges. Plain Language Summary Tornadoes and hailstorms represent a 10 billion dollar annual peril for the insurance industry. Prediction of these extreme events at long lead times is a challenging task. This research shows that it is possible to use the Global Ensemble Forecast System to anticipate tornado and hail events in the United States at lead times beyond 1 week. This represents an important baseline for the improvement of tornado and hail forecasts at the subseasonal to seasonal time scales.