Meteorologists are turning to an unlikely partner -- social scientists -- to try to tackle a troubling question: Why do people tend to shrug off severe weather warnings?

As radar has made revolutionary advances and more people than ever are witnessing and recording tornadoes, a warning system based on Cold War technology -- sirens -- seems out of sync with the public. Even though tornado-related deaths have dropped dramatically in recent decades -- 2011 being an exception -- researchers and weather officials are looking for ways to resolve a high rate of false alarms, siren strategies that vary from place to place, and an apparently increasing tendency for citizens to seek a "second opinion," often unofficial, before responding to warnings.

The effort moved into the spotlight when a review of the tornado that hit Joplin, Mo., last May -- on the same day a tornado ripped through north Minneapolis -- led the National Weather Service to launch experimental new wording for warnings in tornado emergencies. The Joplin tornado killed 158 people, making it the deadliest tornado in nearly 60 years, although sirens were sounded half an hour before the twister hit. The smaller Minneapolis tornado killed one man, after only three minutes elapsed between sighting and touchdown.

"Meteorology is a scientific discipline," said Kenny Blumenfeld, director of projects at ORC International and a visiting assistant professor of geography at the University of Minnesota who has studied urban tornadoes. "Its main practice is to provide scientific information to people who aren't necessarily scientists. We need to do a little work understanding the people we're trying to give the information to."

Top down, bottom up

Julie Demuth, an associate scientist at the National Center for Atmospheric Research (NCAR) in Boulder, Colo., put it another way: Atmospheric scientists approach weather literally from the top down, while social scientists work from the bottom up, starting where people are.

As meteorologists learn more about storms, social scientists' "increasing knowledge about typical patterns and relationships about human thought and behavior is also essential to help protect people from severe weather," Demuth wrote in an e-mail.

"We don't just need better accuracy in forecasts," added Jeff Lazo, her colleague in NCAR's Society Impacts Program. "That's not going to help if it isn't being understood and used. That's where the social sciences come in."

Blumenfeld said language differences, poverty and lack of local experience can all leave large groups of people unaware of the urgency of warnings.

In interviews after the Joplin tornado, the Weather Service found the majority of residents didn't take cover after the first warning, largely because they didn't find the frequent spring warnings credible.

Indeed, 76 percent of tornado warnings are false alarms -- usually meaning a tornado never formed after being noted on radar, never touched down after being sighted, or didn't touch down in the warned area. That has desensitized people, the Weather Service acknowledged. But it also can be perilous, said Tony Hansen, an atmospheric sciences professor at St. Cloud State University.

"People just do not get it -- what you do when you get severe weather warnings," he said. "Disproportionately, they go outside and look around. ... The public doesn't grasp the complexity [of storm predicting]. There's a fine line between covering your butt and overwarning, being the boy who cried 'wolf!'"

In its assessment after the Joplin tornado, the Weather Service found that many residents waited to get to shelter until they got some sort of confirmation of the urgency of the threat, from a second siren, from friends, from broadcast media or social media or from seeing the tornado.

The Joplin assessment used techniques "to scientifically describe cultures and the people within these cultures," and recommended that survey teams be "well-versed in social science and [Weather Service] operations." People's responses to warnings, the assessment said, often depend on their "worldview" -- their experience, education and cultural values. It also examined how people calculate risk, and noted that many in Joplin seemed to employ an "optimism bias" to exempt themselves from a sense of danger.

Stern wording

This year, the National Weather Service is experimenting in Kansas and Missouri with warnings that ramp up the usual wording when a tornado is either sighted or evident from radar. The short-term, "impact-based" warnings will include such statements as "Mass devastation is highly likely, making the area unrecognizable to survivors," or "This storm is not survivable."

Hansen suggested that the false alarm rate for tornadoes may not be the critical factor in public reaction. In Minnesota, surveys in recent winters found that residents frequently brush off winter storm warnings, as well, even though they are far more accurate than tornado warnings. Also, a preliminary St. Cloud student survey finds that residents think "severe" describes conditions that really aren't, so they might tend to underestimate what a severe storm warning might imply.

Todd Krause, warning coordination meteorologist for the Weather Service's Twin Cities office, said he was surprised in interviews after a fatal tornado near Albert Lea in 2010 by how many people told him they had taken shelter. The tornado happened in a rural area, where sirens may not have been easy to hear. Krause said it seemed to him that people responded to warnings because they indicated the tornado had been spotted on the ground, and was not just a figure on radar.

"We need to learn even more about how and why tornadoes form and when they don't, so we can make sure we're not issuing too many warnings based on radar," Krause said.

Bill McAuliffe • 612-673-7646