Ranking and classifying severe weather outbreaks using multivariate indices

Recent studies have investigated the degree to which synoptic-scale processes play a role in the occurrence or absence of major tornado outbreaks. These studies have demonstrated that mesoscale models, initialized with synoptic-scale data, simulate fields of severe weather parameters that were systematically different between tornado outbreaks and primarily nontornadic severe weather outbreaks. Given these promising findings, investigation of outbreaks that fall in between the two categorical extremes is appropriate. To identify such cases, a method to rank and classify severe weather outbreaks of any type is necessary. This study describes one technique to accomplish this task.

The top 30 days each year from 1960-2006, according to the total number of severe reports (tornadoes, hail, wind gusts, and wind damage) in a 24-h period, were selected. A linear-weighted index, using several variables of observed severe weather, was developed to rank the 1410 days. Several nonmeteorological artifacts exist in the dataset, and techniques were implemented to account for these. Results indicated that the highest-ranked outbreak days agreed with subjective perceptions of these events. The most significant outbreak days and days featuring large geographic scatter with the reports were ranked relatively consistently when the weights of the variables were modified. However, rankings of days in between the two extremes exhibited substantial variability. Subjective analysis of these cases suggested the relative severity of these days was similar, but the modes of severe weather observed with these cases could be quite different. As a result, cluster analysis was performed using a four-dimensional decomposition of the multivariate indices. Five types of outbreaks were observed: major tornado, hail-dominant, wind-dominant, mixed-mode, and days with large geographic scatter.