Random Forest Model to Assess Predictor Importance and Nowcast Severe Storms using High-Resolution Radar–GOES Satellite–Lightning Observations

Abstract Few studies have assessed combined satellite, lightning, and radar databases to diagnose severe storm potential. The research goal here is evaluate next-generation, 60-second update frequency geostationary satellite lightning information with ground-based isolate which variables, when used in concert, provide skillful discriminatory for identifying (hail ≥2.5 cm diameter, winds ≥25 m s –1 , tornadoes) versus non-severe storms. focus of this study predicting thunderstorm tornado warnings. A total 2,004 storms 2014–2015 were objectively tracked 49 potential predictor fields related May, daytime Great Plains convective All occurred 1-min Geostationary Operational Environmental Satellite (GOES)–14 “super rapid scan” data available. three importance methods assess warnings, random forests a model skill evaluation measuring the ability predict Three show that GOES mesoscale atmospheric motion vector derived cloud-top divergence above anvil cirrus plume presence most satellite-based power diagnosing Other important include Earth Networks Total Lightning flash density, estimated vorticity, overshooting-top presence. Severe warning predictions are significantly improved at 95% confidence level few fields, only model. This provides basis including within machine-learning models help forecast weather.