Tornado Damage Probabilities Derived from WSR-88D Data

Prior work by Smith et al. (2015) documented a sample of reported tornadoes across the contiguous United States from 2009-2013, and their manually-calculated rotational velocity magnitudes were based on the lowest elevation radar scans.  Their work demonstrated clear trends for higher tornado damage ratings (per the EF scale) with stronger rotational velocities.  We have extended that original work to include rotational velocity with reported tornadoes from 2014-2015, as well as "null" severe thunderstorms (≥1" diameter hail, thunderstorm wind gusts ≥50kt, or reported wind damage) with low-level cyclonic rotation, but no reported tornadoes.  The addition of the null sample allows the computation of tornado probabilities across a spectrum of organized severe thunderstorms (both right-moving supercells and QLCSs).

Preliminary analysis of the 2014 sample shows that tornado probabilities increase as circulation diameter and height above radar level (range) decrease, for a given bin of rotational velocity.  Additionally, tornado probability distributions vary substantially (for  similar sample sizes) when comparing the Southeast United States, which has a high density of damage indicators, to the Great Plains, where damage indicators are more sparse.  Velocity data quality plays a major role in the proper identification of null cases, with a heavy emphasis on dual-polarization data.  We plan to further investigate combined ranges of circulation diameter and height above radar level with the addition of cases from 2015, as well as variations in the distributions related to the scan strategies with more frequent low-level data collection.  Our goal is for this unique sample of radar characteristics of both tornadic and nontornadic severe storms to lead to improvements in real-time tornado detection, and associated software tool development to aid in National Weather Service tornado warning decisions.