21
Extensions to ranking and classifying severe weather outbreaks using kernel density estimation

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Monday, 3 August 2015
Back Bay Ballroom (Sheraton Boston )
Chad M. Shafer, Univ. of South Alabama, Mobile, AL; and N. S. Grondin and C. A. Doswell III

Recent studies have attempted to rank severe weather outbreaks of any type using kernel density estimation (KDE) as a means of defining outbreak location. Though the results from these studies agree well with subjective notions regarding the most significant outbreaks within the conterminous United States, modifications to the technique are proposed to provide a more comprehensive statistical analysis of these events. For example, the past studies measured outbreak severity by considering severe weather reports occurring within distinct 24-h periods. Though most severe weather events transpire completely within 24 hours, many of the most significant severe weather outbreaks occur over multiple days. To eliminate this time constraint, severe weather reports are considered in overlapping time periods. All regions of severe weather of threshold density (or clustering, as determined by KDE) that intersect during consecutive time periods are defined to be part of the same event.

Additionally, as the Storm Prediction Center forecasts probabilities of individual types of severe weather, ranking outbreaks based on single report types is prudent. Using KDE, report detrending (owing to secular trends in observed severe weather), and a multivariate weighted index, hail events, wind events, and tornado events are ranked based on their perceived meteorological and societal significance. Preliminary findings are discussed, including some examples of how the established rankings can be used to evaluate operational forecasts of severe weather.