New Research Turns Tornado Models Upside Down

As we reflect on the best and worst things about 2018, here’s one reason many in the state of Oklahoma are grateful: no one was killed in a tornado last year.

A shot of the El Reno, Oklahoma EF-3 tornado on May 31, 2013 near maximum width and peak intensity.
Image Credit: Nick Nolte (CC BY-SA 3.0).

Apart from 2018, every year since 2006 has seen tornado deaths reported in Oklahoma, according to the National Weather Service. The 2018 exception is most likely due to random fluctuations, since the number of tornadoes in the state was well below average last year (41 versus 56). But to reduce tornado-related deaths in the long-term and over the full range of areas hit by these violent storms, we need to improve early detection systems—and that means we need a better understanding of tornado formation.

A better understanding may be on its way, thanks to new research led by Dr. Jana Houser at Ohio University. At the December meeting of the American Geophysical Union, Houser presented evidence that the commonly accepted “top-down process” of tornado formation has it wrong. Her research suggests that tornadoes don’t reach down from the sky to impose destruction, their activity starts much closer to the ground. While this theory is not new, the observations to support it are.

About 1,000 tornadoes hit the United States each year (far more than any other country), but quality data on tornado formation is hard to come by. That’s because data collection instruments need to be in just the right place at just the right time to capture the conditions, and predicting the time and place of a tornado in advance is really hard. Researchers often collect hours of data during storms that never produce tornadoes and, on the other hand, miss out on tornado activity because they don’t have time to get to the right location with the proper equipment.

The data we do have is primarily collected by Doppler radar systems. Put simply, these systems send out pulses of energy with wavelengths in the microwave region of the electromagnetic spectrum. The transmitted signal scatters when it hits droplets of water, ice, or snow. A receiver on the radar system records the strength, phase, and time delay of the return signal scattered in its direction. From this data, you can estimate precipitation rate, wind speed, and the distance to a target. The Doppler radar is so-named because the data analysis is based on the Doppler effect.

Scanning radar systems send out pulses and ‘listen’ for return signals at increasing angles above the horizon. This enables them to map the conditions at different heights in the atmosphere. Much of the observational data we have on tornado formation is from a national network of radar systems that complete a full scan about every five minutes and then start over. In the last 20 years, we have also gathered information from mobile radar platforms that can perform scans of smaller, targeted areas in less time.

In the 1990s, a study of data from the national network showed that 67% of tornadoes and nearly all of the tornadoes produced by severe storms known as supercells form in a descending fashion—from the top down. This is the most commonly accepted theory of tornado formation, although it’s not the only one.

The main problem with this study, according to Houser, is that tornado formation isn’t well captured by a five minute cycle. To study the formation process in smaller time intervals, Houser and her team analyzed data collected by the University of Oklahoma’s Rapid X-band Polarimetric Radar—RaXPol. RaXPol is a truck-based system that can scan a sample of the atmosphere about ten times faster than the national network. That means it provides snapshots of what’s happening at intervals of 30 seconds instead of five minutes.

Her team analyzed data from four tornadoes produced by supercells in the United States, two in Oklahoma and two in Kansas. If the top-down process is valid, as many researchers assume, then the first evidence of tornado formation should be strong rotating winds at high altitudes. But that’s not what the researchers saw.

“In four out of four of the detailed data sets that were analyzed for this study, none of the tornadoes formed by this [top-down] mechanism,” Houser said in a press conference. Instead, the results showed rotation happening first at the lowest altitudes where data was collected and then, 30-60 seconds later, the strong rotation extended upward, through the highest altitudes.

To further explore this finding, the team zeroed in on one of the four tornadoes—the deadly El Reno tornado that hit El Reno, Oklahoma on May 31, 2013 and killed eight people, including three storm chasers. The researchers used a software tool called the Tornado Environment Display (TED) created in 2015 by tornado scientist Anton Siemon, programmer Skip Talbot, and meteorologist John Allen. The tool patched together crowdsourced photographs and video footage of the tornado, enabling researchers to view the tornado from different visual perspectives with the same time stamp.

Coupling RaXPol data with visuals of the El Reno tornado forming, the team saw the rotation clearly visible in pictures and videos taken before rotation was present in high altitude Doppler radar data. “We had a visual of a tornado funnel on the ground at a time when there was no evidence in the radar data that this tornado existed except at the very lowest elevation angle,” explained Houser.

In addition, in three of the four cases they studied, the time between the earliest detectable rotation at any altitude and a full tornado was 90 seconds or less. Clearly, snapshots with 5-minute intervals just aren’t capable of telling the full story of tornado formation.

This work is about more than getting the theory right. Forecasters look for evidence of tornadoes based on where scientists think rotation starts, which is currently at high altitudes. This new research suggests that warning systems might do better to focus on what’s happening close to the ground instead. That’s challenging to do with Doppler radar systems because hills, buildings, and other objects get in the way of low-altitude radar signals, but Houser is working with forecasters to consider possible approaches.
In some ways, confirming the top-down process would have been a more convenient result. Not only for forecasting reasons, but also because a top-down process suggests that tornado formation is determined by large-scale processes in the atmosphere. On the other hand, if tornado formation happens close to the ground, that could mean that the formation process is sensitive to local conditions and therefore harder to predict. Science isn’t about getting results that are convenient, though—it’s about getting results that are real.

Kendra Redmond

You may also read these articles