On 31 May 2013 at 6:03pm local time, a tornado developed just west of Oklahoma City, Oklahoma. During the 39 minutes before it dissipated, the tornado crossed two major highways and killed 8 people. But that’s just part of the story. The storm that spawned the lethal tornado also caused flash flooding that killed an additional 13 people. Interviews after the event revealed that people failed to understand the threat of flash flooding, despite the warnings issued.
Tornadoes and flash floods pose substantial threats by themselves, but when they strike simultaneously, the danger is compounded. Another concern for overlapping tornado and flash-flood events (TORFFs) is that the procedures recommended for dealing with each hazard are contradictory: Tornado safety practice advises taking shelter in the lowest central room of a building, and flood safety protocol recommends moving to higher ground. Those conflicting instructions can increase confusion over what to do. One survivor of the 2013 Oklahoma storm said she stayed in her shelter “until the water got too high,” at which point she “just hoped the tornado was over.”
TORFFs also challenge operational forecasters. The early warning signs for a tornado are not typically the same as for a flash flood. For example, tornadoes often have fast-moving convective cells, whereas storms that lead to flash flooding typically are slow-moving. (For more on the environmental characteristics of tornadoes, see the article by Paul Markowski and Yvette Richardson, Physics Today, September 2014, page 26. For more on the ingredients needed for flash flooding, see C. A. Doswell et al., Weather and Forecasting 11, 560, 1996.) Thus, forecasters must monitor environmental conditions and watch for the complex intersection of characteristics that lead to simultaneous tornadoes and flash floods.
Until recently, most research has centered on tornadoes and flash floods as isolated events. Now a group at Colorado State University, which includes my adviser Russ Schumacher and myself, has documented the frequency, geographic distribution, and environmental characteristics of these very dangerous events. Our research reveals that every year about 400 tornado and flash-flood warnings are issued within 30 minutes of each other for the same area, which was more frequent than expected. The findings raise the stakes for meteorologists to understand the characteristics of TORFF events and communicate the risks to the public.
The tornado is one of the most dangerous, but also the most romanticized, extreme weather events. Approximately 1000 tornadoes strike each year in the US, and about 20 000 fatalities have been reported since 1880. On average, the US reports fewer than 100 deaths by tornadoes per year, and the percentage of people impacted by a tornado who are killed has continued to decrease through the 20th century, due largely to advances in radar detection, which has increased warning lead times.
Flooding can be just as dangerous and causes on average 100 deaths a year in the US, with most caused by flash floods. Unlike the number of tornado-related deaths, the number of flood-related deaths has not decreased significantly from the mid 20th century to the early 21st, despite increased outreach and civil mitigation strategies.
Figure 1 shows an example of a potential multithreat flash flood and tornado event and illustrates the overlapping warning scenarios for the town of Harrisburg, Tennessee, the site of a potential TORFF event in May 2010.
Figure 1. Overlapping tornado (red polygon) and flash-flood (green polygon) warnings in the Memphis, Tennessee, county warning area on 2 May 2010 at 9:12pm local time.
To determine how often weather forecasters and public officials deemed that the hazards associated with a potential TORFF event were imminent, we identified spatial intersections of tornado and flash-flood warnings from 2008 to 2014. That method does not confirm that they occurred, but it does estimate the number of times a TORFF threat was communicated to the public. Our focus of study was tornado and flash flood warnings that covered the same area within 30 minutes of one another. The 30-minute threshold represents a reasonable amount of time a person could spend sheltering from a tornado, since the average tornado warning is about 40 minutes in length.
Approximately 2800 tornado and flash-flood warning intersections (roughly 400 per year) occurred within 30 minutes of one another at the same location in the US during those seven years. This number increased to about 5000 (roughly 700 per year) when we looked at warnings that overlapped within an hour, and it decreased to near 1500 (about 200 per year) for a 15-minute separation threshold. Of the 2800 warning intersections that occurred within 30 minutes of one another, the flash-flood warning was issued first in about 1300 instances, or 46% of the time.
Figure 2 depicts the spatial and seasonal distribution of the TORFF warning intersections. The geographic mean center is in the central Mississippi Valley, and the vast majority of intersections occur between the eastern plains and the Appalachian Mountains. The warning intersections generally occur from March through May in the southern part of the country, and in later May through July in the northern part; that finding is generally consistent with US tornado climatology. The diurnal cycles of TORFF warning intersections (not shown) generally match those of tornadoes, with higher frequency in the afternoon to evening hours.
Figure 2. Geographic distribution of concurrent, collocated tornado and flash-flood warnings over the period 2008–14. The black dot represents the geographic mean center of all the events, the pink ellipse represents one spatial standard deviation away from mean center, and the black and blue lines represent National Weather Service Weather Forecast Office and River Forecast Center boundaries, respectively.
Since warning intersections do not corroborate that a TORFF event actually occurred, we also identified locations where a flash-flood observation and tornado track exactly overlapped within three hours, for the years 2008 to 2013. That identification method yields a conservative but confident estimate of verified events, because tornado and flash-flood observations that are slightly separated spatially (even by 1 km) are not verified by this method. In total, we identified 68 verified TORFF events over the period.
We then used these events as the case list to create composites depicting the average conditions present across all identified TORFF events. We also produced event-centered composites from a similarly identified set of 68 randomly sampled tornado-only events over the same period, to ensure equivalent sample sizes. The composites for both the verified TORFF cases and the tornado-only cases were created for environmental fields that have established importance in determining severe weather potential. These characteristics include 850 hPa horizontal temperature warm air advection (WAA), precipitable water (PWAT), most unstable convective available potential energy, and 0–6 km bulk shear.
The results of the event-centered composites are depicted in figure 3. The expanse and magnitude of WAA, which is a proxy for low-level synoptic scale forcing for lift, is larger in TORFF events than in tornado-only events. These conditions would aid in more widespread thunderstorms developing in the TORFF cases, which would lead to more widespread rainfall. The total column moisture (as measured via PWAT) and atmospheric instability (most unstable convective available potential energy) were also higher for the TORFF events than for the tornado-only events in the dataset. This suggests that TORFF environments foster stronger convective storms and can produce higher rainfall rates than the tornado-only cases.
Figure 3. (a) 850 hPa WAA (shading, K/hr, values below 1 K/hr are masked), temperature (blue dashed lines, K), geopotential height (black contours, m), winds (gray wind barbs, kts), and the event location (cyan circle) for TORFF events. (b) Same as panel a, but for the 68-member subsample of tornado-only (TOR) events. (c) PWAT (shading, kg m2, values below 25 kg m2 are masked), MUCAPE (blue contours, J/kg), 0–6 km shear (gray wind barbs, m/s), and the event location (black circle) for TORFF events. (d) Same as panel c, but for the 68-member subsample of TOR events.
While TORFF events were found to occur in stronger synoptically forced and moister environments than storms producing only tornadoes, no uniquely identifying traits have yet been isolated. That makes forecasting the occurrence and evolution of TORFF events extremely difficult. Furthermore, the presence of one weather threat—say, a tornado—could potentially hamper a forecaster’s identification of another kind of hazard. The failure to identify each event in a timely manner can further compound the danger. Meteorologists are therefore motivated to conduct future research to better predict these dangerous TORFF events.
The communication challenges associated with TORFF and other concurrent, collocated multihazard events are just as important to investigate as the meteorological specifics. The perfect forecast does no good if it is not communicated effectively to those in harm’s way.
Great care must be taken to expand social science research and understand the complexities associated with warning communication in TORFFs and similar scenarios. Various methods of communication—word of mouth, social media, radio—plus levels of individual personalization and the extent of the total message received all create different levels of understanding for people in danger. Knowing exactly which and how many threats are imminent can lead to different life-saving actions.
Furthermore, people who are confused about the proper course of action might take too long to act or might take no action at all. That behavior is especially disconcerting for TORFFs because the life-saving instructions are contradictory. Although no clear strategy yet exists to mitigate these communication concerns, we hope our work will initiate improvements to multithreat warning communication through social science research.
Moving forward, the members of our research group, including Gregory Herman, Robert Tournay, John Peters, and Schumacher, will investigate the social and meteorological aspects of TORFFs in the upcoming Verification of the Origins of Rotation in Tornadoes Experiment in the Southeast (VORTEX-SE).
Erik R. Nielsen is a graduate student in atmospheric science at Colorado State University.
To learn more about the TORFF research discussed here, see E. R. Nielsen et al., “Double impact: When both tornadoes and flash floods threaten the same place at the same time,” Weather and Forecasting 30, 1673 (2015).
