Be Afraid

05.22.13

Be Afraid: The Future of Tornado Warnings

Moore, Oklahoma, got 16 minutes’ notice. Can’t we do any better? Josh Dzieza on the experts using social sciences to get the word out earlier—and more urgently.
Tornadoes in Moore, Oklahoma, 1999 – 2013

The National Weather Service spotted a vortex forming near Moore, Oklahoma, 16 minutes before the mile-wide tornado touched down, and 36 minutes before it started tearing through homes, schools, and businesses.

That warning gave residents enough time to run for cover, huddling in hallways or storm shelters, if they had them, and is probably part of the reason why the death toll was as low as it was, despite the devastation. Still, meteorologists would like to be able to give warnings even further in advance, and to do so they’re looking less to computer models and physical science, the traditional way to increase lead time on storm warnings, and more to the social sciences. 

“We’re coming to the point where technology and science has progressed so much we can talk about the social side,” says Lans Rothfusz, deputy chief of warning research and development at the Severe Storms Laboratory in Norman, Oklahoma. “Because let’s face it, we’re talking about the intersection of nature and society, and at that intersection it’s not just physical science but the science of how to get people to listen, respond, and act.” 

Rothfusz works on a program called Warn-on-Forecast, which is trying to develop new models and warnings to alert people who might be in the path of a tornado before one has actually formed. The weather service currently has two types of tornado alerts: a “watch,” issued hours or days ahead of time, when models predict that severe thunderstorms could be on the way; and a “warning,” when a vortex has been spotted. That’s when the sirens go off. Advances in radar have allowed forecasters to spot tornadoes about 15 minutes before they touch down, but tornadoes form so quickly and chaotically that they’ve defied attempts to accurately predict them before they form, and computer models powerful enough to do so are probably a decade and several technological breakthroughs away, according to Rothfusz.  

We’ve gotten quite good at modeling the thunderstorms that sometimes spawn tornadoes, however, allowing the weather service to issue watches days in advance. Unfortunately, in Tornado Alley, these storms form so often and watches are issued so frequently and for such large areas, they sometimes lack a sense of urgency. That’s why NOAA is looking for a new type of alert, more specific and urgent than a watch, but with more lead time than a warning, that would tell people that they’re in danger. “Instead of a binary type of warning, either on or off, it would communicate more information about the increasing level of threat as the storm approaches,” says Rothfusz. “That’s where social science comes in.” 

Video screenshot

Relive the tornado that devastated Moore, Oklahoma.

Perhaps not surprisingly, people are more responsive to warnings that include phrases like “complete destruction.”

One possibility would be to map areas where models show the probability of a tornado is increasing. “We don’t expect the public to understand probability, but we do know they respond well to, ‘the risk is increasing, for you, there, in a specific area, the risk is getting greater and greater.’” By ramping up the alert level gradually rather than jumping from watch to warning, people would be able to adjust their behavior: staying near a shelter, perhaps, or deciding not to go play softball, or just keeping an eye on ominous clouds. “The models will catch up and do great things in 10 years,” Rothfusz says, “but this is something we can improve in the near term.” 

Tailoring warnings for the areas at greater risk would also help cut down on false alarms. Right now we issue warnings for large areas, so everyone who is lucky enough not to get hit gets a false sense of security. That’s why, for so many people in Oklahoma, the first reaction to a siren is to go out and look for the storm rather than seek shelter. “You can’t run every time you hear a warning,” a Moore resident told The New York Times. “You’ll be scared your whole life.” Rothfusz thinks that if we can refine the warning area, people will start to pay more attention, because they’ll know it’s meant for them. 

The language of warning also helps get people’s attention. In Moore, the weather service issued a “tornado emergency,” an extra-urgent form of warning that, incidentally, was coined when a different EF5 tornado tore through Moore in 1999. Scott Curl, a forecaster who was in the Norman, Oklahoma, weather service office in 1999 and again on Monday, told the Oklahoman that the office invented the term on the fly when they saw a massive tornado heading toward Oklahoma City and “were trying to do anything that we could at the time to get people’s attention.” 

Now the weather service is playing with words in a more official capacity, studying how people respond when warnings are phrased in terms of possible damage rather than in terms of storm strength. The NWS applies a complicated set of criteria when naming storms—“severe” refers to the speed of the wind and the size of the hail, for instance. Perhaps not surprisingly, people are more responsive to warnings that include phrases like “complete destruction.” The NWS started testing the new “impact-based” warnings in Kansas and Missouri last year, and officials say the new language helped save lives in a tornado outbreak in Kansas last April. “Major house and building damage likely and complete destruction possible” was the sobering message that went out on April 14. “You could be killed if not underground or in a shelter,” read a warning issued on May 19 as a “large and extremely dangerous tornado” neared Wichita before, fortunately, dissipating. “It’s a way of ringing the bell a little louder,” says Mike Hudson, the chief operating officer at the NWS Central Region, which is testing the new warnings. “It paints a mental picture for people in the path of the storm.” After Hurricane Sandy, the NWS concluded that it should use similarly evocative language for storm-surge warnings. 

Eve Gruntfest, a research associate at the Trauma, Health & Hazard Center at University of Colorado, says that better communication will go a long way, but that there’s a lot we don’t know about how people react to warnings. When she studied why people try to drive through flooded streets, she found that they often knew the danger but forged ahead anyway. “The risk of losing their job was perceived as greater than the risk of losing their life,” she says, so in that case it wouldn’t be enough to issue warnings—we should work with employers to give flex time for emergencies. “People are complex and have a lot of other constraints and conflicts and other things they have to do that day,” says Gruntfest. “All this work should be done within the context of how people really use weather information.” 

Rothfusz says the weather service will be looking at how people responded to the Moore warnings, but that early reports show that they reacted well—as well as one would expect in a town that has found itself in the path of enormous tornadoes every couple of years. The tornado watch started days before, so when people got the final warning, they seem to have gone for cover. As tragic as the 24 deaths were, he says, considering the wide swath of devastation it was a testament to the warning system working well.