Automatic Tornado Prediction with an Improved Mesocyclone-Detection Algorithm

Abstract

A new and improved algorithm for automatic mesocyclone detection is presented and tested on 23 mesocyclonic storms. A small false-alarm rate (4%) and high probability of detection (83%) are achieved for mesocyclone classification. A unique innovation of the algorithm is the automatic assessment of mesocyclone tornado potential. This is accomplished using excess rotational kinetic energy (ERKE), a form of rotational kinetic energy that is tailored for mesocyclonic shear. ERKE provides a measure of low- to midtroposhpheric mesocyclone intensification that is indicative of impending tornado formation. The quantitative determination provided by ERKE is a much better indicator of storm severity than is simple mesocyclone identification. Median lead times of over 30 min are provided for our small sample by ERKE for strong and violent tornadoes with a false-alarm rate of less than 5%.