Building a better tornado warning system when every minute counts

The power cut. The air pressure left the room. A security alarm whined. After a particularly large boom from just overhead, Shannon Johnson turned to her husband Keith, “I think the house is gone.” Sitting on the queen bed, she could feel her 4-year-old daughter shaking from adrenaline.

It was nearly 2 a.m. on March 3, 2020. The Johnsons, their two young kids and their dog were tucked away in a basement bedroom of their home in Donelson, Tennessee, about 10 miles east of Nashville. That night, 10 tornadoes moving west to east, including a powerful EF-3 and an even stronger EF-4, touched down in the state, destroying homes and ultimately killing 25 people. The Tennessee Emergency Management Agency estimates damage in Middle Tennessee reached $1.6 billion. It was one of the worst tornado outbreaks in history for a state long accustomed to the destructive storms.

There are about 1,200 tornadoes in the US on average every year, according to the National Severe Storms Laboratory, making them a violent part of life in the parts of the Midwest and Southeast. In 2019 alone, tornadoes were responsible for about $3.1 billion in economic damage across the country. They can turn whole neighborhoods into what looks like piles of popsicle sticks from the air and kill about 70 people every year. And they typically come with less than 15 minutes warning. It had been only a matter of minutes in between the time the couple saw an urgent alert on their phones to seek shelter and when they headed down to the basement of their home — one they’d just spent six months remodeling. 

While technology like Doppler radar has come a long way in helping meteorologists track dangerous storms and warn people of what may be heading their way, there’s less certainty in determining when a tornado will form. Forecasting still relies on interpreting the radar and getting old fashioned eye witness accounts of tornadoes on the ground. And the speed of tornadoes — as fast as 60 miles an hour — makes accurate and timely warnings a matter of life and death for those in the storm’s path. 

But by listening for low-frequency sounds a tornado makes when it begins to form, scientists are hoping to build a better warning system. After decades of languishing from a lack of attention and funding, research into sound waves far below the range of human hearing could help forecasters detect when a tornado touches down, rather than relying on visual reports from people on the ground. Officials could then relay the coming threat far earlier than the average 15-minute warning they can offer now, giving people more time to seek life-saving shelter. One day, this technology could be part of a warning system that could clue forecasters into the presence of a tornado as much as 50 miles away. 

“This won’t save property,” says Roger Waxler, one of the researchers working on this type of tornado detection, “but I’m hoping we can save lives.”

The batting cage from a nearby school tore a hole in the roof above the Johnsons’ master bedroom.

Johnson Family

The Johnsons, though, had only that phone alert. When the rain seemed to quiet, Keith went upstairs. The house was still there, but a mangled and balled up batting cage from a school a half mile away had smashed a roughly 15 by 20 foot hole in the roof of their master bedroom. Bleachers crashed into their sunroom. Outside both cars were totaled and 16 trees in their yard were downed. One had fallen on the porch — that was the huge boom — and the branches shot through the front door when he opened it.

After a sleepless night, Shannon went to look around the neighborhood a few hours later.

“It felt like I was walking in a movie,” she says of seeing neighbors standing amid the rubble of their homes. “It felt like the end of Twister.” 

Listening low

Tornadoes are loud. When they get close, those nearby often describe hearing something that sounds like a freight train bearing down on them. But Waxler was skeptical about whether they also make noises humans can’t hear. So he decided to find out.

Waxler is a professor in the department of physics and astronomy at the University of Mississippi specializing in acoustics, including atmospheric sound propagation. That’s the study of influences on how sound travels through the atmosphere. About 10 years ago, he received funding to study tornadoes and infrasound after his boss Henry Bass, who’d been working on a separate theory of how to detect tornadoes using microphones, passed away.  

Infrasound is sound below the frequency of 20 Hz, which is the lowest frequency that humans can typically hear. The range of human hearing extends up to about 20,000 Hz

Roger Waxler is one of several researchers investigating infrasonic technology to detect tornadoes. 

Roger Waxler

After developing microphones capable of detecting infrasound at the university, Waxler in 2011 sent a team to Oklahoma to capture data from tornado-producing storms. When they looked at what they’d captured, they saw sound waves in the infrasonic range. His initial skepticism faded. 

“It seemed natural,” he says. “You see a tornado and think, ‘Wow, that must be putting out all sorts of crap.’ It’s a violent event.” 

Waxler’s other thought: Alfred Bedard had been right all along. 

The idea that tornadoes might emit an infrasonic signature isn’t new — Bedard, a research scientist at the National Oceanic and Atmospheric Administration, has been working in this area for decades. But Bedard tells me the idea of aiming infrasonic mics at tornadoes had an unlikely genesis. 

In the 1970s, scientists at the Environmental Research Lab in Boulder, Colorado, started researching infrasound and geophysical signatures. Funding for exploring more uses of the infrasonic technology came after the 1963 Nuclear Test Ban Treaty when it was chosen as one of four methods to detect illegal nuclear weapons tests

Though he’d started out in Washington D.C. working with the Department of Defense on the nuclear monitoring network, Bedard eventually got pulled out to Boulder with other team members. 

“We slowly morphed from a global DOD-oriented program to one that was evaluating different geophysical possibilities for hazard mitigation,” says Bedard.

Infrasonic technology may help scientists understand plenty of natural hazards better. Researchers have used microphones to pick up shockwaves given off by the aurora borealis, meteors exploding in the atmosphere and wind whipping off mountain tops. They’ve even used it to detect the sounds ocean waves make when they collide with each other, which one day could help monitor the intensity of hurricanes. 

It was while studying avalanches (and consequently working in a range a bit higher than they had in the past — 0.5-10Hz) that they realized tornadoes could be yet another possible use for infrasonic mics. 

In 2003, Bedard’s team decided to listen. They deployed arrays with infrasonic microphones that could pick up the lower frequency tornado sounds while filtering out interference from wind noise. They put the arrays in three locations: Goodland, Kansas, Boulder, Colorado, and Pueblo, Colorado and paired each one with Doppler radar weather stations. 

The results seemed promising. Although the average warning time for the detected tornadoes was between 7 to 12 minutes, the microphones registered a twister’s infrasonic sound waves about 30 minutes earlier than the radar was seeing it. If this was the case, perhaps meteorologists would be able to learn when a tornado was on the ground with certainty and speed, and not just depend on eye witnesses or suggestive indicators via radar. 

“That’s big, getting an additional 20 minutes of warning for a tornadic storm,” Bedard tells me. 

Despite the findings, other researchers weren’t convinced and funding dried up. Bedard said they operated on fumes the second year of the experiment and then just had to stop. Since 2006, they’ve only been able to do theoretical work on the concept. He didn’t give up on it, though. 

“It’s a persistence and a willingness to hunker down and keep things running, even though you’re not being paid for it that’s kept us going,” he says.  

Tornado tech

Tornadoes can spawn out of several types of storms — thunderstorms, supercells and squall lines. Supercells, though, are the most studied since they tend to be the most intense and longest lived of the three, featuring an area of rotation at the middle levels of the atmosphere. 

But as Jana Houser, an associate professor at Ohio University, tells me on a hot day in June, it’s not completely understood why one supercell might produce a tornado and another might not.

Robert Rodriguez/CNET

“It’s hugely complicated, but our understanding is improving,” Houser says. 

Longtime Nashville meteorologist for NewsChannel 5 Lelan Statom remembers how initially TV stations used repurposed airport radar to look for a signature called a hook echo, which could indicate a possible tornado. Next Generation Doppler Radar, which can scan different levels of the atmosphere, came in 1988. It lets meteorologists get a feel for not just precipitation but also wind conditions. 

On a day when there’s the potential for tornadoes, meteorologists look for those hook echoes, as well as rotation at different levels of the atmosphere, and something called a debris ball, which typically means a tornado is on the ground. A storm’s relative velocity can also indicate rotation, Statom says. They also use computer modeling data to look at areas of the atmosphere where conditions might be ripe for storms. It’s what Houser calls a “first line of defense” ahead of time before anything has actually started.

But confirming that a tornado is actually on the ground is trickier. Houser explains that radar can’t always detect tornadoes because tornadoes are frequently so low level — below 1 kilometer off the ground in the atmosphere — and radar essentially aims at an up angle above the horizon. The farther from the radar, the weaker and higher up the radio wave becomes — sometimes there can be hundreds of miles between adjacent radars. 

Statom says meteorologists are looking for visual confirmation,  or what’s called “ground truth” and often, it comes from actual humans — storm spotters or civilians — who chime in with what they’ve seen. 

“Mother Nature is awesome,” says Statom, “Sometimes that awesome is just watching the clouds on a great day. Sometimes that power comes in these very violent tornadoes.”

Mic drops

An array of arrays could help nail down that ground truth. 

For the past four years during tornado season, Waxler’s team has taken Bedard’s old microphone concept and beefed it up, putting out about 10 microphone arrays through northern Alabama and Mississippi. Each array consists of two sub arrays, each with eight microphones powered by what’s essentially a car battery attached to a solar panel, along with a data acquisition device, GPS and windscreens. 

The mics pick up squiggly sound waves and after some processing, Waxler and his team can look at that data, plus GPS data to see what direction that storm was moving in. Eventually, they’ll need to figure out how to get the raw data, process it and wirelessly get it to meteorologists in real time.

At an American Meteorological Society meeting in Boston in January, Waxler and his group presented findings that the infrasonic signature from a tornado could be picked up “on the order of 100 km” away. 

At Oklahoma State University, Brian Elbing and his team have been studying infrasonic sound and tornadoes since 2015. Stillwater, Oklahoma where the university is located, is a prime spot for setting up arrays because of its location in Tornado Alley, the swath of states like Texas, Oklahoma, Kansas and Nebraska that tend to see the most tornadoes every year. 

A partial view of one of Waxler’s array set ups. 

Roger Waxler

They have two arrays (one with three mics and another with four) installed on OSU’s campus. In 2017, they picked up a signal from a small tornado about eight minutes before it actually formed near Perkins, Oklahoma, roughly 20 miles away. 

With funding from the National Science Foundation as part of a project called CloudMap, Elbing is collaborating with researchers at other schools. He’s also working on developing arrays that can be deployed quickly in places where severe weather might hit. A storm chaser from the Stillwater’s channel 9 will start carrying a mic when things get dicey.

But there’s another part of the country where tornadoes actually tend to be deadier: Dixie Alley, which cuts through Alabama, Georgia, Tennessee and up into parts of Kentucky.

Unlike Tornado Alley, which is generally flat and open, Dixie Alley has more hills and trees making it harder to spot tornadoes. Plus, there’s a higher occurrence of rain-wrapped tornadoes, which are more difficult to detect via radar. Dixie Alley’s higher population density also puts more folks in danger, and if as if all that wasn’t enough, about 47% of tornadoes there occur at night, catching people asleep in their homes. 

Waxler and Elbing hope a future infrasonic warning system could help overcome Dixie Alley’s tornado tracking challenges and give the region’s residents more warning. But there’s still a key mystery that researchers have yet to solve before they can say with full certainty that a system like this would work: They don’t know what exactly in a tornado emits the infrasonic signature. 

“It must be something unique, but we’re still not sure what it is,” Bedard says.

One of the mics Brian Elbing and his team use on OSU’s campus.

Brian Elbing

What’s more, Waxler said some skeptical members of the meteorological community have suggested there’s something else making those infrasonic sounds, which is why he and other researchers are trying to eliminate other possible sources, like thunder. They’re also looking at storms that didn’t produce tornadoes to make sure those didn’t also emit signatures. 

Elbing talked about solving other riddles, as well, like the direction of wind patterns, which can affect how and where sounds are picked up. And, you can’t exactly replicate any of this research in a lab. Down the line, he envisions having lots of arrays and better tools for modeling factors like wind direction, but that aid hasn’t arrived yet. 

Assuming all that gets sorted, there’s the technological challenge of how to one day quickly get the data, process it and send it to the Weather Service when a tornado is actually occurring so that it’s helpful in issuing early warnings. 

“If we can answer these questions,” Elbing said, “and improve warnings in the Southeastern US, that’s where the lives are really going to be saved.”

How the rest of the work on this technology plays out depends somewhat on funding. Waxler’s current funding comes from NOAA’s Vortex Southeast Project. Elbing has some from NSF and NOAA as well. He’s optimistic that there’s increased interest, particularly because of the Southeast’s vulnerability. Waxler thinks that if the money holds out, the Weather Service could be able to use the technology to augment radar within the next few years, and that it could cover threatened areas with lines of arrays 40 km apart along lines of latitude.

Storm warnings

For those who “live in troubled regions,” as Adrienne Rich put it in her 1951 poem Storm Warnings, they’ll need to rely on the tried and true warning systems until the sound technology can be perfected. 

That includes television weather reports — sometimes epic broadcasting marathons of rolled up sleeves and meteorologists passing through the background of the shot, brows furrowed. Then there are tornado sirens — Nashville put up its first in 2003 and now has 133 sirens across the city — and the FCC-run Weather Emergency Alerts system pushes warnings directly to your phone. Some folks own NOAA weather radios, which broadcast directly from the National Weather Service around the clock. Other times, texts and calls from friends and family might be the red flag. Now of course, there’s social media. 

But for anyone in forecasting, a pressing thought is how to effectively reach people, particularly if severe weather is supposed to hit at night.

In good weather and bad weather, NewsChannel 5’s Lelan Statom is a familiar face among Nashvillians.

WTVF, NewsChannel 5

“We’re trying to do what we can to keep people aware, and by doing that, make sure that once the severe weather is over, they’re here to live another day,” Statom says.

For the Johnsons, it was a marked shift in tone from @NashSevereWx, a local account co-run by citizen weather tweeter David Drobny, that finally drove them to their basement with a few minutes to spare. 

“[@NashSevereWx] is the reason we survived,” Shannon says.

While the account is often lighthearted, even suggesting when an extra swipe of deodorant might be necessary in hot weather, it was an all caps plea to seek shelter that drove them out of their room, where the batting cage plowed into the roof.

Drobny tells me that he and the others who work on the account always remember that “underneath the radar are real people.”

Shelter from the storm

One hundred and eight days after that batting cage smashed into their house, the Johnsons finally got a new roof back. The coronavirus pandemic has slowed reconstruction and a derecho storm, with 60 to 80 mile per hour winds in May, damaged their house further. 

As of July, they’ve had their sunroom reframed and the master bedroom replaced. Work is slow but moving, finally. 

They still think about that March night when they ended up right in the path of the storm, their ears popping from the pressure change, and not knowing just how long the tornado was on the ground for before they even knew. 

“The stakes are just so high, anything that can give you a leg up on to stay safe,” Keith says. “We made it downstairs in time but it would have been nice to feel like we weren’t cutting it so close.”

First published Aug. 10.