Norman, OK 73069
September 1998
1. INTRODUCTION
At present, there is no widely agreed upon method for comparing tornado days or so-called tornado "outbreaks" with each other. The number of tornadoes, number of violent tornadoes, and number of killer tornadoes are often cited as a measures of the "significance" of an outbreak. However, these measures do not reliably reflect both the threat to life and property and the meteorological significance of a tornado outbreak. For example, tornadoes that produce damage in urban areas are nearly always more publicized and scrutinized than equivalent tornadoes that occur in sparsely populated areas. The goal of this study is to improve upon tornado categorization by damage alone with an index that incorporates a measure of tornado intensity with a measure of the damage area.
As a complement to this index, we propose a damage area scale that is analogous to the widely-used Fujita scale to rate tornado damage. Fujita (1971) originally proposed a logarithmic tornado damage area scale. Unfortunately, this area scale did not gain the same acceptance as the F-scale.
2. THE DESTRUCTION POTENTIAL INDEX
In simple terms, the potential for tornado casualties and property damage is a function of tornado intensity, path width, and path length. The Fujita Scale has been used successfully since the 1970s to categorize tornado damage and imply tornado intensity. Unfortunately, the F-scale relies on damage to manmade structures, which tends to preclude high ratings for tornadoes occurring in sparsely populated areas. Instead of relying only on "ratable" tornadoes (i.e., those that hit manmade structures), we propose an index of the total tornado damage area for each day (or any specific time period) multiplied by the weighted mean F-scale for all tornadoes that occur during the time period of interest. This quantity, the Destruction Potential Index (DPI), is a measure of the potential for damage and casualties with a particular outbreak:
where n is the number of tornadoes, a the tornado damage area (path length multiplied by mean path width) for each tornado, and F is the maximum F-scale rating for each tornado. The F-scale rating plus one is used so that F0 tornadoes can produce non-zero values. Clearly, the DPI is heavily weighted toward long-track, violent tornadoes (which typically are the most likely to hit structures), as opposed to large numbers of weak, short-track tornadoes. This index can be computed for particular tornado outbreaks, and for all days with tornadoes.
Sample DPI calculations for several historical tornado outbreaks are presented in Table 1. To begin with, the massive tornado outbreak of 3-4 April 1974 stands out in terms of both number of tornadoes and DPI. The total DPI of 2647 is roughly 40% greater than the DPI associated with the Palm Sunday outbreak from 11 April 1965. Other interesting comparisons involve 31 May 1985 and 13 March 1990. The 31 May 1985 outbreak totaled 40 tornadoes in the United States across portions of Ohio, Pennsylvania, and New York, while the 13 March 1990 outbreak in Oklahoma, Kansas, and Nebraska produced 64 tornadoes. However, DPI values for each outbreak are quite different. The 31 May 1985 DPI of 1338 nearly doubles the 692 value from 13 March 1990, in spite of twenty-four more reported tornadoes in the 1990 outbreak. This difference owes to the fact that the 1985 outbreak produced more long-track, violent tornadoes than the 1990 outbreak, hence the potential for more destruction.
Many other interesting comparisons are possible with the DPI. An outbreak of 60 tornadoes centered on Illinois occurred on 19 April 1996. The large number of tornadoes certainly qualifies 19 April 1996 as a noteworthy day, and significant damage occurred in several cities in Illinois. However, the DPI for this outbreak was only 24 since there were no long track, violent tornadoes reported. In contrast, a single supercell on 5 May 1993 in the open country of southwest Kansas produced a DPI of 175, with 3 individual tornadoes resulting in greater DPI values than the entire 19 April 1996 outbreak. The 3 April 1974 outbreak produced 22 individual tornadoes with DPI values greater than the 19 April 1996 outbreak.
3. TORNADO DAMAGE AREA SCALE
As a complement to the DPI, an additional rating scale for individual tornadoes is proposed. Though many tornadoes do not affect enough structures to reveal maximum intensity of the tornado, path width and path length are reported in Storm Data. We propose combining the path lengths and mean widths into a simple tornado damage area (A) scale. A combination of F-scale and A-scale ratings can be used to more reliably estimate the meteorological "significance" of individual tornadoes.
The approach is similar to Fujita's attempt to characterize tornadoes by both intensity and area, though less involved than the FPP system for separate tornado damage, path width and path length categorization developed by Fujita and Pearson (1973).
The A-scale is modeled after the frequency distribution of F-scale occurrences, such that the relative frequency of "A0" tornadoes will match that of "F0" tornadoes. The frequency distribution of the F and A-scale categories is listed in Table 2.
Table 3 presents the damage area value ranges associated with the A-scale, based on the frequency distribution of F-scale occurrences from 1986-1995. In accordance with the F-scale distribution, roughly 85% of all tornadoes result in damage areas of around .25 mi2 or less. Meanwhile, the top 1% of tornadoes produce damage swaths covering greater than 7.9 mi2.
A large, long-track tornado in open country warrants a high A-scale rating, regardless of problems estimating tornado intensity when little or no manmade structures are damaged. Likewise, a small, short-lived tornado that happens to cause intense damage would be given a small A-scale rating. The most meteorologically significant tornadoes are those that maximize both the area affected and intensity of damage.
Typically, meteorological parameter studies focus on F-scale rating to discriminate between weak and "significant" tornadoes (usually F2 or greater damage). The Penderboro, SC F4 tornado from 7 November 1995 and the Wichita Falls, TX F4 tornado from 10 April 1979 offer an interesting comparison in this regard. Both tornadoes produced damage estimated at F4, though the SC damage was to a single structure while the Wichita Falls tornado resulted in an immense amount of F4 damage. However, the SC F4 produced a damage area of only 0.016 mi2 (minimal A1 category), versus an 23.5 mi2 damage area for the Wichita Falls tornado (A5 rating). From this comparison, it should be obvious that these two tornadoes were not equivalent. A combination of F-scale and A-scale appears to be a more reliable way of separating brief tornadoes from the more dangerous long-lived tornadoes.
Another advantage of the A-scale is the range of area values within each category. Estimates of tornado path length and width are rarely exact, but the range of values in each category will allow some variation in the estimates without necessarily resulting in a different A-scale rating. In contrast, the F-scale forces the damage surveyor to make discrete choices between categories. This can result in variations in subjective F- scale ratings of at least +/- one F-scale category, as discussed by Doswell and Burgess (1988).
4. CONCLUSIONS
The Destruction Potential Index (DPI) is a simple and consistent way to compare different tornado days. The DPI highlights those episodes with large, violent tornadoes, as opposed to just large numbers of tornadoes. Like the F-scale, the DPI cannot completely account for tornadoes that produce no damage in open country, or when the damage is not representative of maximum tornado intensity.
We also present a damage area scale (A-scale) to categorize tornadoes. This A-scale is modeled after the F-scale, such that the relative frequency of a particular A-scale category matches that of the corresponding F-scale category. We believe a combination of F-scale and A-scale ratings allow for more meteorologically sound discrimination between "weak" and "significant" tornadoes.
Acknowledgements: We would like to thank Joe Schaefer and Bob Johns of the Storm Prediction Center for providing the tornado database and for reviewing this manuscript.
5. REFERENCES
Doswell, C. A., III, and D. W. Burgess, 1988: On some issues of United States tornado climatology. Mon. Wea. Rev., 116, 495-501.
Fujita, T. T., 1971: A proposed characterization of tornadoes and hurricanes by area and intensity. SMRP Paper 91, 42 pp., Univ. of Chicago.
_____, and A. D. Pearson, 1973: Results of FPP classification of 1971 and 1972 tornadoes. Preprints, 8th Conf. on Severe Local Storms, Denver, CO, 142-145.