Multivariate Index for Ranking

Abstract

In previous work, severe weather outbreaks have been classified either as major tornado outbreaks or as primarily nontornadic outbreaks, but the large majority of such events are of a mixed character. This study proposes a reproducible method for ranking all types of severe weather outbreaks from the period 1960-2006. Numerous nonmeteorological artifacts exist in the severe weather reports archived during this period, and many of the variables used to formulate the multivariate indices had to be detrended to reduce the effect of secular trends. The resulting outbreak rankings indicate that the methodology presented herein is able to distinguish the most significant severe weather outbreaks from intermediate outbreak days and days with a large amount of geographic scatter in the severe reports. The rankings of the most severe outbreaks and those outbreak days with a large degree of spatial scatter exhibit only limited variability when the selection of parameters and their weights are modified, but a relatively high degree of volatility is noted with the intermediate cases. This result suggests there is relatively little difference in the severity of these intermediate events. However, the particular modes of severe weather in these events can be quite different. A k-means cluster analysis of the outbreak days, using a four-dimensional representation of the multivariate indices developed, indicates that outbreak days can be separated into five groups: major tornado, wind-dominated, hail-dominated, multi-modal, and days with considerable spatial scatter of the severe reports.