Information presented in a blog post is accurate to the best of my knowledge but has not gone through a rigorous review process typical of articles in a peer reviewed journal. Technical information and or analysis provided is not necessarily intended to supersede conclusions made in the peer reviewed literature. As specified by the NOAA Scientific Integrity Policy, comments I post online through my Weather Underground account, personal website, or any social networking site such as Facebook reflects my personal and/or professional opinion, but may not reflect the official opinion of NOAA and the National Weather Service.
Marshall, T.P., Davis, W., Runnels, S. "Damage Survey of the Joplin Tornado: 22 May 2011." Procedings of the 12th Conference on Severe Local Storms. Nashville, TN. November, 2012. https://ams.confex.com/ams/26SLS/webprogram/Manuscript/Paper211662/Joplinmerger.pdf
We've heard it before: "you won't survive this tornado if you don't get below ground." We've also seen the controversy and further exaggerations (eg. "drive south out of town") from those same people making the claim. Violent (EF4/5) tornadoes, accounting for only ~2% of the total, cause approximately 67% of the fatalities (Ashley, 2007). By the time damage reaches EF4 severity to a single family residence, no walls are left completely standing, with maybe a few small "pockets" remaining in the rubble. At EF5 damage, all the rubble is swept clean away (interior hallways, bathrooms, and all), leaving nothing behind that isn't reinforced such as a safe room. After seeing pictures of the aftermath of violent tornadoes, it can lead one to wonder how anyone could survive such damage. From that, a myth is born... you won't survive this tornado if you don't get below ground or evacuate.
The science, however, does
The peer-reviewed literature already contains
some studies discussing fatality rates from violent tornadoes. In “Low-Level Winds in Tornadoes and
Potential Catastrophic Tornado Impacts in Urban Areas” (Wurman et al.,
2007), the authors estimated the mortality rate at 10% for persons
sheltering in a residence that experiences a violent tornado strike. The authors then used this crude estimate as
a basis for a hypothetical tornado which moved through densely populated
sections of Chicago and resulted in 13,000 or more fatalities. In a comment to Wurman et al. (2007), Brooks et al.
(2008) questioned the core assumptions used in the former study, in
particular the 10% mortality rate and the uniformly large area impacted by
EF4/5 winds in the hypothetical tornado. Using data from the 1998 Brimingham, AL, F5 tornado and the 1999 Moore/OKC, OK, F5 tornado, the authors estimated the the mortality rate from a violent tornado strike as between 0.1-1.9% (Table, below) depending on the grouping of building damage.
Table 1 from Brooks et al. (2008)
In another comment to Wurman et al. (2007), Blumenfeld (2008) questioned a different assumption made by Wurman that was not the focus of Brooks: variability in the violent tornado mortality rates due to time of day and socioeconomic differences. Blumenfeld concludes that tornado fatality estimates are difficult and require assessment of multiple variables. Although certainly true, the data presented in Brooks - and to a lesser extent, the assumptions used in Wurman - strongly suggest that the likelihood of surviving a violent tornado are actually quite high, even in areas with limited access to underground shelter.
In May of 2011, the deadliest tornado in the period of modern tornado records hit Joplin, MO, causing 161 fatalities (including persons seriously injured that died at a later date). Although tornado warnings had been in effect for the Joplin area for 20-30 minutes prior to the tornado, the actual large and violent tornado that hit Joplin formed right on the outskirts of the city, rapidly intensified, and was mostly wrapped in curtains of rain. Earlier warnings for tornadoes impacting areas to the north of Joplin proper precluded the sheltering of most impacted residents, and the formation just on the edge of town limited live television and radio coverage as known during several historic violent tornado strikes to major cities.
Several studies and assessments have already reported on the Joplin tornado. These include the NWS WFO Springfield damage survey and the NWS Central Region Service Assessment. The American Society of Civil Engineers did a comprehensive study of construction techniques of affected structures; the Structural Engineers Association of Kansas and Missouri did a study of their own. Another study, "Technical Investigation of the May 22, 2011, Tornado in Joplin, Missouri" (NIST, 2013) was released by the National Institude for Standards and Technology (NIST) in early November, 2013, for public comment. Of particular interest was the summary of interviews, media reports, and other documents dealing with the fatalities from the tornado. The location of almost every fatality was documented. When possible, an estimate of the number of survivors at a given location was also indicated, which allows us to conclude on the mortality rate for different structures.
In Brooks et al. (2008), the fatality rate was estimated by dividing the number of fatalities by an estimate of total persons residing in the number of homes impacted. Using demographic data, the authors estimated 2.7 persons per detached home. For example, in the Moore/OKC tornado, 17 fatalities were listed as occurring in the F4/5 damage area which also destroyed 334 homes (first row of Table 1). 17 fatalities / (334 homes) x (2.7 persons per home) yields a fatality rate of 1.9%. We can also apply this methodology to the Joplin tornado. As reported by the NIST report, there were 161 fatalities in the Joplin tornado, of these 58% (93) were in areas impacted by EF4/5 winds. It was also noted that 2063 structures were classified as "demolished" (equivalent of approximately EF3 winds or higher). A better estimate of residences impacted by EF4/5 winds can be found in "Damage Survey of the Joplin Tornado: 22 May 2011" (Marshall et al., 2012). The authors surveyed over 7000 residences that were damaged by winds ranging from EF0 to EF5 strength; of those, they concluded that 511 sustained EF4 damage and 22 sustained EF5 damage (533 total). Using the Brooks methodology, the 93 fatalities yields a mortality rate of 6.5%, substantially higher than their summary (but still low). As mentioned by Blumenfeld (2008), this method is rather crude by assuming that all persons impacted were in detached homes, and also assumes that the average home occupancy applies to all areas impacted in Joplin.
Detailed information from the NIST report allows us to break down the fatalities by location. In the table below, the location of each fatality is listed, along with information on total building occupancy, number of injuries, and estimate wind speed, when available. I also divided the building occupants by their location within the building as either "unsafe," "marginal," or "safest." This classification is rather subjective and open to interpretation in other analyses, but was roughly done as follows:
Summary of tornado fatality information from the NIST report on the Joplin tornado. When possible, the number of total occupants was recorded. Values highlighted in yellow were rough estimates and values highlighted in green were the minimum number deduced from the report (using a minimum value would yield a conservative, likely higher, estimate of mortality). The location of each fatality/injury/survival was further sub-divided, albeit somewhat subjectively, by level of shelter taken by the person
In the locations listed in the above, there were 161 fatalities at 18 different locations/location groups. The summary also includes the generalized locations of "outside," "vehicle," and "single family residences" and 10 locations where no fatalities occurred. At these same locations, there were 41 known serious injuries. The total occupancy of each location was estimated as being at least 571 (actual occupancy is very likely higher, but not enough information was provided for a better reasonable estimate). This suggests that the chance of dying and the chance of being seriously injured or dying in the Joplin tornado was less than 28.2% (161/571) and 35.4% (202/571), respectively, for all locations. These values are subject to large uncertainty because not all buildings had estimates for occupancy available. Looking only at locations with both a number of fatalities and an estimate of total occupancy, the estimated mortality rate would be 10.3%; for only areas that experienced EF4/5 winds, 16.8%. The sample size for individual shelter classifications was substantially smaller, but the trend was toward much lower fatality rates for persons classified as "safest" compared to "unsafe."
Past studies have suggested that the majority of persons will survive a direct strike by the strongest portion of a violent tornado if they follow the standard tornado safety advice. The chance of survival increases as the quality of shelter improves. Past tornadoes have demonstrated that being outdoors or in vehicles during a tornado strike is very dangerous. This is especially true when roadways are busy, such as during rush hour periods or during an evacuation. Precise forecasts of a tornado's exact track and intensity and simply not possible, even on very short lead times when evacuations are impractical. Sheltering-in-place remains the best option in the overwhelming majority of situations during a tornado threat.
Review of information on fatalities that occurred from the Joplin, MO, tornado of 2011 seems to corroborate conclusions from previous studies. Complicating this type of analysis, however, is the lack of data on most of the detached homes impacted by the tornado, which dwarf commercial buildings in number and likely had a higher occupancy than the other locations listed. Virtually no information is available to indicate the total number of persons in the various commercial districts impacted because we know little about structures where fatalities did not occur. One estimate quoted in the NIST report suggested that over 20,000 persons may have been in the area impacted by the tornado in Joplin. Although limited analysis can be done with that information, the fact that only 161 persons were killed by a tornado with maximum intensity of EF5 in an urban area with 10s of thousands of persons present strongly suggests that the chance of survival in a violent tornado is high, even in areas where no basement is available.
To improve our conclusions on tornado safety advice, future studies of significant tornadoes should, when possible, gather information on persons who were impacted by the the tornado but survived instead of focusing solely on individuals who perished.
Wurman, J., Alexander, C., Robinson, P., and Richardson, Y. “Low-Level Winds in Tornadoes and Potential Catastrophic Tornado Impacts in Urban Areas.” Bulletin of the American Meteorological Society. January 2007. http://journals.ametsoc.org/doi/pdf/10.1175/BAMS-88-1-31
Brooks, H.E., Doswell, C.A., and Sutter, Daniel. “Comment: Low-Level Winds in Tornadoes and Potential Catastrophic Tornado Impacts in Urban Areas.” Bulletin of the American Meteorological Society. January 2008. http://journals.ametsoc.org/doi/pdf/10.1175/BAMS-89-1-87
Blumenfeld, K.A. “Comment: Low-Level Winds in Tornadoes and
Potential Catastrophic Tornado Impacts in Urban Areas.” Bulletin
of the American Meteorological Society. October 2008. http://journals.ametsoc.org/doi/pdf/10.1175/2008BAMS2670.1