We can't predict tornadoes with high accuracy. Scientists are trying to change that

It seemed like a nightmare come to life, as tornado after tornado devoured communities across the United States. At one point 30 million people were deemed to be at tornado risk as the series of devastating twisters struck large sections of the Deep South — including Louisiana, Mississippi, Alabama and Georgia — leaving at least 26 people dead and thousands of buildings in ruins. Even as the afflicted Americans attempt to piece together their lives, there remained warnings of even more possible tornados and thunderstorms. Experts agree that, although tornadoes are among the most unpredictable forms of extreme weather, we are better at anticipating them now than ever before.

We owe much of that to real-life storm chasers like Jana Houser.

It was Houser's first storm chasing experience: May 12, 2004. Because tornadoes often do not show up where and when they are expected, storm chasers frequently come home empty-handed. Not Houser. During her maiden voyage in Medicine Lodge, Kansas, she saw seven tornadoes in a single day.

The future Ohio State meteorologist was only a student at the time. She recalled how her college professor, though "usually mild mannered," became "visibly and audibly anxious" as the group of scientists lost sight of one particular tornado they had been chasing in their mobile radar truck. Ominously, the other chase vehicle could see the slippery cyclone right behind a hill, even though Houser's group could not. Then, all of a sudden and out of nowhere, the behemoth was terrifyingly close.

"The lead [scientist] said we were fine, but the tornado got closer and closer," Houser remembered. "Finally we bailed. I was driving the 10-ton radar truck into a 45 mph headwind... peddle to the floor, going 40 mph ourselves. I was so giddy on the high of escaping and seeing a tornado for the first time in my life that I was laughing hysterically!"

Tornadoes — or rotating funnels of air connected to the Earth and either a cumulonimbus or the base of a cumulus cloud — are dangerous because of their immense power, which Houser barely escaped. The famous Fujita Scale exists solely to classify different levels of tornado intensity, with "moderate" tornadoes (F1) having wind speeds of 73-112 mph and "incredible" tornadoes (F5) reaching 261-318 mph. This makes F5s so large that "strong frame houses [are] lifted off foundations and carried considerable distances to disintegrate."

Scientists have only started to understand twisters, and what progress they have made is because — as Iowa State University geology and atmospheric sciences professor William Gallus wrote to Salon — core pieces of tornado-tracking technology like radar and computers are becoming more sophisticated. The only reason it has been harder to predict tornadoes than thunderstorms, blizzards, droughts, hurricanes and other types of extreme weather is that feisty funnels defy traditional weather-tracking methods.

Or, more simply put: twisters do not behave like other types of extreme weather such as hurricanes. 

"Because of [tornadoes'] small-scale, our normal array of weather stations [are] too far apart for us to get the kind of data we need to really help us predict them, and even understand all the atmospheric factors that lead to their creation," Gallus explained.

While it may seem counterintuitive to describe tornadoes as "small-scale," a twister's funnel does not cover the large areas of mileage captured by hurricanes, blizzards, thunderstorms and other extreme weather events. That means that it is difficult to design technology that sits in the right position to measure these tornadoes.

What's more, even if one could accurately anticipate when and where tornadoes will appear, they are generally so destructive that they would demolish normal weather stations.

Want more health and science stories in your inbox? Subscribe to Salon's weekly newsletter The Vulgar Scientist.

"Because we cannot predict them ahead of time, we instead have to wait until the storm has formed and we can see what is happening  from radar," Gallus pointed out. "But even with radar, the radar beam is pointed slightly upward, and this complicates the accuracy of our warnings, since very rapid rotation seen by radar at the base of the clouds may not necessarily mean the rotation extends to the ground — i.e., is a tornado."

Yet radar technicians keep making advances, including "dual-polarization radar" that sends out two beams polarized in different planes, which can identify the nature of precipitation and thereby "predict when during a thunderstorm a tornado is likely to form." Moreover, computer scientists have developed new models that include large numbers grid points so that extreme weather events in general can be seen "on finer scales." That is why, when experts see a thunderstorm rotating, they can issue an informed warning of a possible tornado.

Yet tornado prediction isn't an especially refined science. To improve prediction, Gallus advocates for more weather data on finer scales; increased computer power to run more rigorous models; and ambitious research projects for gathering tornado-based data during storms to better understand how tornadoes are created.

"One big challenge right now is that two storms can be spinning just as fast and look the same on radar, but one will produce a tornado and one will not," Gallus observed. "To understand the reasoning for such differences is almost like finding a needle in a haystack."

Houser had her own list of proposed improvements, telling Salon that "our understanding would really improve if we could better observe temperature and moisture characteristics within the storm and the near environment in a three-dimensional space." Scientists also need to better understand how precipitation and friction interplay with each other to lead to tornado formation, as this information has "major impacts on storm structure and properties like the temperature of the air within the storm, and the nature of the near-ground wind field which is where rotation that becomes tornadic is most critical to understand."

Even though scientists have a long way to go to comprehensively predict tornadoes, that does not mean we also have not come a long way since the deployment of the Doppler radar network in the 1990s.

"In some ways tornadoes are fairly easy to predict nowadays," explained Jason Naylor, an associate professor at the University of Louisville's Department of Geographic and Environmental Sciences, in an email to Salon. "The Storm Prediction Center does an excellent job of identifying days that are favorable for tornadoes, and the general area where tornadoes might occur. However, meteorologists cannot say that a tornado will occur at some specific location at some specific time."

"For example," Naylor continued, "meteorologists can predict that, on a particular day, tornadoes are very likely to occur in central Oklahoma. But they can't predict that a tornado will go through a specific area and a particular time."

Until that happens, there will continue to be tragedies as tornadoes strike unsuspecting victims. Victor Gensini, an associate professor at Northern Illinois University's Department of Earth, Atmosphere, and Environment, recalled one of the more bizarre sights he has beheld in his years of studying these monstrous vortices.

"I've seen things that just don't compute," Gensini recalled. "I mean, an entire house destroyed except for the interior bathroom. And you walk into the interior bathroom and you open the medicine cabinet, and all the medicine pill bottles are all perfectly untouched."