Predicting 'outbreak'-level tornado counts and casualties from environmental variables

Abstract:

Environmental variables are routinely used in forecasting when and where tornadoes are likely to occur, but more work is needed to understand how characteristics of severe weather outbreaks vary with the larger-scale environmental factors. Here the authors propose a method to quantify `outbreak'-level tornado and casualty counts with respect to variations in large-scale environmental factors. They do this by fitting negative binomial regression models to cluster-level tornado data that estimate tornado counts and associated casualties on days with at least ten tornadoes. Results show that a 1000 J/kg increase in CAPE corresponds to a 5% increase in tornado counts and a 28% increase in casualties, conditional on at least ten tornadoes, and holding the other variables constant. Further, results show that a 10 m/s increase in deep-layer bulk shear corresponds to a 13% increase in tornado counts and a 98% increase in casualties, conditional on at least ten tornadoes, and holding the other variables constant. The casualty-count model quantifies the decline in the number of casualties per year and indicates that tornado outbreaks have a larger impact in the Southeast than elsewhere after controlling for population and outbreak size.