COMMENTS

Abstract

n a stimulating presentation, Wurman et al. (2007, hereafter W2007) have produced an interesting estimate of the possible effects of violent tornadoes in urban areas by making models of the wind field based on mobile Doppler radar observations. As part of that effort, they have estimated death tolls associated with those modeled wind fields, arriving at estimates from 13,000 up to 63,000 in Chicago, Illinois. Given that the highest death toll in a tornado in U.S. history is 695 in the Tri-State tornado of 1925, and that the last death toll of greater than 100 was in 1953, the validity of these estimates is of some concern. We are certainly in favor of raising the level of awareness of the potential for large casualty figures in future tornadoes. The concentration of our population in urban areas, combined with urban sprawl, has increased the threat of large fatality figures if a large, violent, long-track tornado were to hit a major metropolitan area. Planning for such a catastrophic event requires an estimate of the potential impacts. We fear, however, that the enormity of the fatality estimates in W2007 might discourage emergency managers from planning for such an event, because they would simply be overwhelmed by such a disaster. Two assumptions drive the fatality estimates in W2007. First, W2007 estimate that 10% of occupants within housing subjected to Fujita-scale F4 and greater winds would be killed. Second, W2007 assumes that the maximum intensity and areal extent of high winds are held constant over nearly the entire length of the storm path. Both of these assumptions are open to question and, we believe, represent significant overestimates of the likely highend events. An examination of past tornado events suggests these are simply unrealistic values. W2007’s assumption that 10% of residents in the path of such a catastrophic tornado would be killed is inconsistent with estimates that can be derived from studies of deaths in F4 and F5 tornadoes (Table 1). We infer the percentage of residents killed using the national average household size of 2.7 persons (see 2005 census estimates online at www.census.gov/ popest/states/NST-ann-est.html, and www.census. gov/prod/2006pubs/h150-05.pdf). The 3 May 1999 Oklahoma City, Oklahoma (OKC), and 8 April 1998 Birmingham, Alabama, F5 tornadoes struck metropolitan areas and provide the best-available evidence on fatality rates. The estimates in Table 1 are based on varying levels of damage to homes. The first, from D. A. Speheger (2007, personal communication), is based on a count of homes in the OKC tornado that suffered F4 or F5 damage as rated in the postevent damage survey. The fatality rate in such damaged homes in OKC was only 1.9%, and this estimate most closely approximates W2007’s assumption that all residences in the hypothetical tornado face F4 or F5 winds. But, Speheger et al. (2002) found that only 13% of the total path area affected in the OKC tornado was associated with F4 or F5 damage levels, which is typical of the few F5 tornadoes whose paths have been surveyed reasonably thoroughly. Thus, the other estimates that include either destroyed homes or damaged or destroyed homes [i.e., the Oklahoma Department of Emergency Management, which counted destroyed housing, regardless of the F-scale