Session 7B.2 Proposals for modernizing definitions of tornado and severe thunderstorm outbreaks

Tuesday, 5 October 2004: 4:45 PM
Roger Edwards, SPC/NWS/NOAA, Norman, OK; and R. L. Thompson, C. Crosbie, J. A. Hart, and C. A. Doswell III

Presentation PDF (187.9 kB)

The term “tornado outbreak” has been arbitrarily and inconsistently defined in the literature, and can be highly variable in spatial and temporal coverage. The various definitions also have not accounted for changes in tornado reporting and documentation practices since the 1940s, when some of the earliest quantitative assignments appeared for the term, and since the 1980s, when great changes appeared in reporting trends in response to WSR-88D installations, warning verification efforts and spotter training emphases. In particular, tornado reports have become more profuse, especially on the low end of the damage spectrum (F0). Such trends cast into doubt the utility of outbreak definitions made prior to the WSR-88D era, such as Galway’s 20-tornado threshold and its subsequent derivatives. Extensive use of the words “tornado outbreak” in both the popular media and the meteorological literature, utilizing ambiguous, antiquated and/or highly varying definitions, compels a more consistent, modernized and statistically robust approach. The same principle also is used to motivate describing what constitutes a “severe thunderstorm outbreak,” exclusive of tornadoes.

For tornado outbreaks, we begin with a list of eight non-exclusive criteria by which outbreaks are commonly judged: number of tornadoes, number of violent tornadoes, number of significant (>F2 damage) tornadoes, cumulative path length, deaths, number of killer tornadoes, and Destruction Potential Index (DPI, Thompson and Vescio 1998). Tornado days from 1970-2002 are ranked in each of those eight categories. From these rankings, and relative weighting factors subjectively assigned to each category, we compute an outbreak index (O-index), where positive values (O>0) constitutes a “tornado outbreak". The O-index, in turn, may be used both to rank historical outbreaks and to assess the relative importance of new ones.

Little is available, other than empirical assessment of report maps, by which to judge the “density,” "importance" or “quality” of a severe thunderstorm outbreak on a nationwide basis. In fact, no consistent standard exists for largely nontornadic severe local storm events containing multiple reports of severe hail and damaging convective wind. In developing a solution to this problem, some means must exist to judge the organization of the event (as implied by report concentration), while accounting for the spatial outliers that are far removed from the concentrated event cluster but within the same time domain (i.e., a few severe downbursts in Arizona temporally coincident with a massive derecho event across the Ohio Valley). Therefore we develop a severe thunderstorm outbreak index, and describe utilization of the kernel density estimation (KDE) technique to identify clusters of severe weather reports for each convective day (1200-1200 UTC). We then rank these events according to their respective outbreak indices propose statistically based measures for severe local storm outbreaks that are more meaningful than the simple number of reports.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner