83rd Annual

Monday, 10 February 2003
A Discussion of the F-Scale of Tornadoes from Quasi-Linear Convective Systems
Elaine S. Godfrey, Univ. of Oklahoma, Norman, OK; and R. J. Trapp, H. E. Brooks, and S. A. Tessendorf
Poster PDF (2.0 MB)
In the meteorological community there exists a pervasive, yet unwritten, theory that tornadoes that form from quasi-linear convective systems (QLCSs) like squall lines and bow echoes are, in general, not as intense or long-lived as those from more isolated cells. A recent study by Tessendorf and Trapp (2000) determined that 19% of all tornadoes between March 1998 and February 1999 developed in QLCS parent-storm types.

There is a strong likelihood that the lowest-intensity tornadoes produced by linear convective storms are underreported. Brooks and Doswell (2001) demonstrated the log-linear distribution of tornadoes by F-scale. Using a similar number of events from the Tessendorf and Trapp study, it was determined that only two percent of all randomly generated datasets would be as similarly distributed and highly correlated as the QLCS and cell-based tornado events in that year. Using the storm-classified data from Tessendorf and Trapp, both cell and QLCS-based tornadoes demonstrate the expected log-linearity with the exception of F0 QLCS tornadoes. It is hypothesized that tornadoes from linear convective systems occur more frequently at night than those from cells, and nighttime weak tornado damage may often be classified as straight-line winds, accounting for the disproportionately small number of F0 QLCS tornadoes.

Examining the three-hour running mean of tornado initiation times from both cell and line-based tornadoes, a strong diurnal cycle is evident only in the cell-type tornado distribution. To determine the statistical significance of the higher occurrences of QLCS tornadoes in the nighttime hours, a Monte-Carlo technique was performed and showed results significant at the 99.9% confidence level. Therefore, there is a statistically significant difference between the observed daily tornado distributions from QLCSs and cell type storms.

It has been difficult to compare tornadoes from QLCSs to those from cell-type storms, given a limited dataset, due to tornado frequency variations between the two storm types. For this reason, a Monte Carlo resampling method was used to evaluate the null hypothesis that the probability of a violent tornado, given a tornadic QLCS, is much less than the probability of a violent tornado, given a tornadic cell. Ten thousand years of simulated tornado activity were generated using a uniform random number generator. Results showed a significantly lower probability of an F3 tornado, given a QLCS tornado, than the probability of an F3 tornado, given a cell-based tornado, at the 98% confidence level. However, no significant conclusions can be drawn about F4 and F5 tornadoes, as the events are particularly rare even in cell parent-storm types. Thus, more type-classified tornado data are needed to determine whether linear convective systems, on the whole, produce weaker tornadoes. Similar studies will be done on two additional years of classified parent storm-type tornado data.

Supplementary URL: