Analysis of Tornado Outbreaks Using Principal Components

Tornado outbreaks occur in geographically extensive regions across the United States. Despite the widespread occurrence, no comprehensive investigation has attempted to discriminate synoptic differences between tornado outbreak and non-tornado outbreaks. Accordingly, that is the focus of this research. The NCEP/NCAR 0000 UTC reanalysis database, from 1970 – 2003, provided initial fields and boundary conditions. A total of 50 tornado outbreak cases and 50 cases of primarily nontornadic severe weather were found for comparison. A multivariate principal component analysis was performed on 5 different reanalysis parameters: temperature, height, meridional and zonal wind components, and relative humidity, simultaneously at all 17 of the reanalysis pressure levels. Biases in the covariance estimates, caused by poleward converging longitude lines, were minimized by using an equally spaced grid on the spherical Earth. Several case studies selected from the tornado and primarily nontornadic outbreak datasets were made with both WRF and MM5 24 hours prior to the event. Key differences between the outbreaks and non-outbreak scenarios in the principal component structures will be shown. The principal component patterns will be used to derive a set of initial fields to drive the numerical simulations to test for their efficacy in discriminating forecasts of tornado outbreaks.