I was at a pizza joint in Lincoln, Nebraska, when a woman ran up to the bar and announced that the National Weather Service had just issued a tornado warning. It was mid-tornado season and I followed her out the front door (in retrospect, maybe not the best plan) where I could hear a loud siren. The sky was dark and ominous.
I headed back inside and into the basement for shelter. I was the only person doing that and was feeling self-conscious. I called my husband. "I'm either the idiot from out of town, or the sole survivor," I told him.
Severe weather warnings are intended to get people out of harm's way, but as was clear that day in May, responses to these alerts can vary widely. And why that is interests researchers. Whether people immediately take protective measures, seek more evidence that the danger is imminent or simply wait and watch depends on a variety of factors, and understanding them can make severe weather warnings more effective and, hopefully, save lives.
Tornado warnings are just one kind of alert - the National Weather Service also issues warnings about hurricanes, storm surges, severe thunderstorms, floods, extreme winds, winter storms and numerous other types of severe weather. A warning is effective if it motivates people to take a course of action that will protect them from harm, says Robert Drost, a geoscientist at Michigan State University.
"For the most part, people don't disregard weather warnings," says Julie Demuth, who studies risk communication at the National Center for Atmospheric Research. "But that doesn't mean they're always going to do what we want them to do."
Social cues can be influential, Demuth says. If you look around and see your neighbors evacuating in reaction to a hurricane warning, you might think about getting out too. But if you're in a restaurant where no one else is seeking shelter, you might feel self-conscious about heading to the basement.
"If other people aren't reacting, that's giving you a cue about the risk that you'll have to reconcile," Demuth says.
Likewise, seeing a live TV crew or news team standing out in the weather taking photos may give people a false sense of security, Drost says. "They think, he's out there with a camera crew so it must be safe."
Experience can play a big role in how people respond, particularly if the person has previously encountered a traumatic weather-related event, Demuth says. Someone who has had a scary experience with severe weather in the past may be quicker to react to those kinds of events in the future.
On the other hand, if you're in an area where there are lots of tornadoes but they rarely hit the ground or cause destruction, you will have a lot of prior incidents to remember, but you may have a tendency to downplay a new warning, because you weren't affected the last time, Drost says. "And if that's the time when the tornado really hits your area, it can set the stage for something pretty traumatic."
On May 22, 2011, a tornado in Joplin, Missouri, killed 158 people and injured more than 1,000 others, making it one of the deadliest tornadoes in U.S. history. An NOAA assessment of the National Weather Service's warnings and forecasts around the Joplin tornado found that some residents had become desensitized to tornado sirens and warnings, and that "initial siren activation has lost a degree of credibility for many residents."
Credibility, the report concluded, is "one of the most valued characteristics for effective risk communication." (Concerns about credibility are one reason President Donald Trump's false statement about the path of Hurricane Dorian produced concern among meteorologists.)
The high casualty numbers of the Joplin tornado can't be solely blamed on a desensitization to warnings - an evaluation of the numbers by NOAA scientists concluded that the death toll from that tornado was mostly because of high population density in the storm's path. But warnings that are perceived as false alarms can make people dangerously complacent, Drost says.
Drost's mother and sister live in South Carolina. During a recent hurricane, he called his mom, who was sending him selfies of herself having a glass of wine out on the deck.
"I said, 'Hey Mom - I'm kind of knowledgeable in this, and you really need to evacuate,' " Drost recalls. "These things come through all the time," she told him.
"There's a complacency that sets in," Drost says. The last time a hurricane had come through, it rained a lot but nothing really bad happened to them, "so they think it's not going to happen again."
Research has shown that people who are unfamiliar with a particular weather event tend to be more cautious (and more likely to take protective action in response to an alert) compared with people who are familiar with it, says Susan Joslyn, a psychologist who studies decision-making at the University of Washington. I was especially terrified of the tornado in Lincoln because I'd never experienced a tornado alert before. And some research has found that women were more likely than men to evacuate in a hurricane.
When I eventually went back upstairs in that Nebraska pizza place, I asked the bartender whether going into the basement was overreacting. She told me that they get plenty of warnings, but tornadoes never hit Lincoln. She'd heard there was something about the geography of the town that protected it.
Demuth used to live in Lincoln, and she recalls that people talked about it being in a "bubble" protected from tornadoes. But there's not good evidence that that's true. "People have a lot of local beliefs about where they're protected or at risk," Demuth says.
Many places have folklore about where tornadoes do or don't hit, but these ideas are not necessarily scientific, says Kim Klockow-McClain, a behavioral scientist at the NOAA National Severe Storms Laboratory. She has observed that people in Norman, Oklahoma, where she lives seem to believe that the city is very safe, because the community to the north of them has been hit repeatedly by large tornadoes while the heart of Norman has been left alone.
But this pattern is just the result of random storm motion, she says. "It could have easily been us." Several large tornadoes have hit Norman, but they've struck the rural outskirts of town, so "people don't personalize it as much," she says.
In light of all these human factors driving how people respond to weather alerts, how can these warnings be made more effective? One approach that has been studied is focusing warnings on the destruction or impact the weather event might have.
A study published last year looked at whether including information about the impacts that a weather event will have and recommendations about what actions people should take would change how people perceived the threat and how they intended to respond to it.
"We found that impact-based warnings increased the likelihood that people would take protective action," says lead author Philippe Weyrich at the Swiss Federal Institute of Technology in Zurich.
But this study (like many others that have looked at this idea) used hypothetical scenarios, which meant that they were collecting data on intention, and not actual behavior.
So Weyrich's group then took the next step and teamed with the provider of a weather app to deliver either impact-based or nonimpact based weather warnings and then collect data on what people actually did in response.
They're still writing up the results, but Weyrich says that they found the impact-based warnings had "basically no effect" on how people responded to medium- or low-severity events.
He speculates that for these events, which were mostly winter storm warnings, people already had lots of experience and knowledge about what to do in these situations. Where impact-based warnings might have a greater power to change behavior is in situations where there's an event that's unusually severe, he says.
"It's the very high impact ones that are really rare and most people would not know what to do in these cases," he says.
Providing information about the likelihood of the event and the uncertainty surrounding the forecast is another promising approach. Joslyn's research shows that people are pretty capable of understanding numeric likelihoods, so that if they're told there's a 20 or 30 percent chance of something happening, they make better decisions than they might have made without this information.
Her work suggests it is better for people to have the scientific calculations to work from because they're going to make their own estimates based on their past experiences. And if they've experienced some false alarms in the past, they may end up underestimating the risk, Joslyn says. But give the public accurate information about risks and uncertainty, and this misestimating is reduced - and that could save lives.
Every hurricane season, some residents in the storm's path ignore the warnings and underestimate the consequences of the storm surge that can accompany a hurricane. During Hurricane Florence in 2018, for instance, rescuers in the city of New Bern, North Carolina, were forced to save at least 150 people who had not evacuated. Anything that increases a population's likelihood of evacuation could save lives and responder resources.
This article was written by Christie Aschwanden, for The Washington Post.