Study: Wind patterns in lowest layers of supercell storms key to predicting tornadoes

New research from North Carolina State University has found that wind patterns in the lowest 500 meters of the atmosphere near supercell thunderstorms can help predict whether that storm will generate a tornado. The work may help better predict tornado formation and reduce the number of false alarms during tornado season.

Supercells are a special type of thunderstorm. They last much longer than normal thunderstorms and produce the vast majority of tornadoes and other severe weather. Seventy-five percent of supercell thunderstorms are nontornadic, or don’t cause tornadoes. Difficulty in predicting which storms may produce tornadoes has resulted in a false alarm ratio for tornado warnings that also hovers around 75 percent. When using traditional weather sampling methods, there are no clearly observable differences between tornadic and nontornadic supercells in terms of precipitation echoes, rotating updrafts or surface air circulations.


To address this knowledge gap, researchers involved in the second Verification of the Origins of Rotation in Tornadoes Experiment (VORTEX2) collected data in close proximity to supercell storms. Using data from the 12 best-sampled storms – seven of which produced tornadoes – Brice Coffer, a graduate student in marine, earth, and atmospheric sciences at NC State and lead author of a paper describing the work, ran simulations of supercell storms to determine which factors made tornadogenesis more likely.

“We noticed that the biggest difference between tornadic and nontornadic storms was the wind in the lowest 500 meters near the storm,” Coffer says. “Specifically, it was the difference in the way the air rotated into the storm in the updraft.”

All storms have an updraft, in which air is drawn upward into the storm, feeding it. In supercells, the rising air also rotates due to wind shear, which is how much the wind changes in speed and direction as you go higher in the atmosphere. Coffer’s simulations demonstrated that if wind shear conditions are right in the lowest 500 meters, then the air entering the updraft spirals like a perfectly thrown football. This leads to a supercell that is configured to be particularly favorable for producing a tornado, as broad rotation at the ground is stretched by the updraft’s lift, increasing the speed of the spin and resulting in a tornado.

On the other hand, if the wind shear conditions in the lowest part of the atmosphere are wrong, then the air tumbles into the storm like a football rotating end over end after a kickoff. This results in a disorganized storm that doesn’t produce tornadoes due to a lack of stretching near the ground. Coffer hopes that his results may lead to fewer tornado false alarms.

“This work points to the need for better observational techniques of the low-level winds being drawn into the storms’ updraft,” Coffer says. “Improving this aspect of storm monitoring will improve our predictive abilities when it comes to tornadoes.”


The research appears in Monthly Weather Review. Matthew Parker, professor of marine, earth and atmospheric sciences at NC State, is co-author. The research was supported by the National Science Foundation grant AGS-1156123.

“Simulated supercells in nontornadic and tornadic VORTEX2 environments”

DOI: 10.1175/MWR-D-16-0226.1

Authors: Brice Coffer, Matt Parker, North Carolina State University

The composite near-storm environments of nontornadic and tornadic supercells sampled during the second Verification of the Origins of Rotation in Tornadoes Experiment (VORTEX2) both appear to be generally favorable for supercells and tornadoes. It has not been clear whether small differences between the two environments (e.g. more streamwise horizontal vorticity in the lowest few hundred meters above the ground in the tornadic composite) are actually determinative of storms’ tornadic potential. From the VORTEX2 composite environments, simulations of a nontornadic and a tornadic supercell are used to investigate storm-scale differences that ultimately favor tornadogenesis or tornadogenesis failure. Both environments produce strong supercells with robust mid-level mesocyclones and hook echoes, though the tornadic supercell has a more intense low-level updraft and develops a tornado-like vortex exceeding the EF3 wind speed threshold. In contrast, the nontornadic supercell only produces shallow vortices, which never reach the EF0 wind speed threshold. Even though the nontornadic supercell readily produces subtornadic surface vortices, these vortices fail to be stretched by the low-level updraft. This is due to a disorganized low-level mesocyclone caused by predominately crosswise vorticity in the lowest few hundred meters above ground level within the nontornadic environment. In contrast, the tornadic supercell ingests predominately streamwise horizontal vorticity, which promotes a strong low-level mesocyclone with enhanced dynamic lifting and stretching of surface vertical vorticity. These results support the idea that larger streamwise vorticity leads to a more intense low-level mesocyclone, whereas predominately crosswise vorticity yields a less favorable configuration of the low-level mesocyclone for tornadogenesis.


27 thoughts on “Study: Wind patterns in lowest layers of supercell storms key to predicting tornadoes

  1. Very nice study with material economic implications.
    Low level winds beneath a supercell would be difficult to monitor from space or high altitude.
    Setting up an adequate low level surveillance system will be expensive, even if only tornado prone areas are covered. Wonder what the economic cost of false tornado alerts actually is.

  2. Would a doppler echo signal from a radar scanning just above the horizon differentiate the wind velocity and direction with sufficient resolution to provide this warning signal? If so, at what distance?

    • Doppler can only register things going directly toward and directly away from the radar. It would require somehow linking data from multiple radars. But all of the radars would need to be the same distance from the storm to insure each radar was measuring the altitude at the same resolution. The farther away, the less resolution, which would make it difficult to interpret the data.

      Due to the curvature of the earth, the farther away a radar is from a storm the less of the lower altitude the beam can reach. In some parts of the nation the current Doppler network can scan no lower than 30,000 ft.

      Paris, TX has been hit by at least two tornadoes. Sure enough, Paris, TX is in a gap between coverage where 30,000 ft is the lowest the closest Doppler radar can scan.

      Each year I try to attend the free Skywarn sessions given by the National Weather Service. Most of these programs only present the basic session. Those that also present the advanced session go fairly in depth (for laymen) discussing Doppler radar. Discussion includes the technical aspects of Doppler and the analyzing and interpretation of data.

      • Yes, if this technique works, it would require a rather more dense network of doppler radars to avoid horizon effects in able to pick up low altitude winds.

      • I actually live northwest of Durant, on the cusp of the coverage gap. The local weather reports have always been pretty accurate about a storm’s progress, that is until it gets near my general area. This helps explain it; many storm cells are literally flying under the radar in this vicinity, leaving the weathermen blind to what’s going on at ground level. Storm spotters on the ground help fill in the blanks, but there’s only so many of those guys.

      • “This is due to a disorganized low-level mesocyclone caused by predominately crosswise vorticity in the lowest few hundred meters above ground level within the nontornadic environment.

        Good assessment, especially with regards to the curvature of the earth causing the radar beam to get higher above the surface with distance(tangent line to the surface at the radar site).

        This link provides comprehensive info:

  3. I was struck by the selection of the 12 BEST examples. Were they selected on the basis of their to the theory? Were they selected on their completeness of data? How many of those rejected formed tornadoes? If some did, how many? What is the theory of why these non-BEST formed tornadoes? These are some of my questions.

    • I think they probably had to choose data sets that were more ‘complete’ or had good coverage of the beginning of the storm. Complete data sets from a storms entire life cycle are hard to get.

  4. I always thought that the chance of a tornado increased in a super-cell if the super-cell turned sharply to the right. If the mean wind, which carries the storm, increases with height, vorticity tubes, created by the lowest level wind shear, tend to be perpendicular to the storm movement, your definition of cross-wise vorticity. This would even be true if the wind veers with altitude through the depth of the storm, or even the lower half of the storm.

    Suppose that the lowest level horizontal vorticity, created by speed shear within the lowest level winds remains the same. Now, if the storm were to turn sharply to the right, the storm-relative lowest level vorticity would become less cross-wise and more stream-wise. The updraft in the center of the storm would thus no longer lift the lowest level rotating vorticity tubes like a man lifting up a rope. No longer would the storm tend to turn the vorticity tubes to the vertical on either side of the updraft. With the storm turning sharply to the right, the central updraft of the storm would then take up the lowest level updraft tubes, now stream-wise in a storm-relative context, right into the updraft, causing the updraft and the now vertical vorticity tube to be more closely correlated in location. Thus, the storm updraft would be rotating more efficiently due to first, the deep veering wind-shear in the storm environment, and second, the now stream-wise lowest level vorticity being taken directly into the core of the storm updraft. Not only would this improved ingesting of vorticity tighten and increase the rotation of the updraft, but it would decrease the precipitation loading in the updraft by spinning the precipitation particles out of the updraft core, like a centrifuge.

    So, to summarize, I would think that a sharp turn of a super cell to the right would infer, however indirectly, a significantly increased chance of the storm being able to generate a tornado.

  5. “New research”? Meteorologists have been using Doppler radar indications of low-level circulation (“hook echoes”) and the absence of low-level shear as a sign of imminent funnel cloud formation for years. Tornadoes are just funnel clouds (vortices) that have grown enough to touch ground.

  6. What, no mention of catastrophic warming’s impairment of Doppler radar signals? Surely all that CO2 wafting about in our poisoned atmosphere has the same effect on radar as cataracts on eyeballs. Obviously there’s little hope of additional funding for follow-up work. And I have serious doubts of the simulation’s sophistication because it failed to project a dire future if we don’t abandon all use of fossil fuels.

  7. Back in August two tornadoes struck the Detroit-Windsor area. These were not picked by either U.S. or Canadian radar.Thought to have formed over the Detroit river. This was not a large weather system passing through the area.

    Tornadoes can form so quickly so not picked up by present radar systems. Best to inform the public that weather conditions are right for severe storms. At present, a false alarm is better than no alarm.

    More funding is needed for better detection.

    Back in the good old days you could set your black & white TV on Channel 2 and watch for tornadoes in your immediate area.

    • Barbara, you took my thoughts before I could put them into words, the speed at which these funnel clouds / tornadoes form is fast and also very hard to tell precisely where they will touch down. In my shoes I’d rather be in the cellar and nothing happened than depend on a “precise” forecast ( heck they rarely get the weather forecast correct.). But I still think this a small step forward. I just wish that the competition for funding wouldn’t drive these people so hard as to keep on putting out incomplete and/or bad data, which I think is a huge problem noting that the peer review process has also taken a huge credibility hit the past years especially in this field.

  8. They need to add a visual test. I grew up in Kansas and every tornado that happened nearby came from a green sky. Similar-looking storms (bubbly-bottomed clouds, temperature, etc.) that weren’t green didn’t produce a tornado.

    • Green sky indicates hail.

      And around the confluence of the Great Lakes such as Lake Huron, Lake St.Clair and Lake Erie you can add yellow sky.

      You can also see the formation/appearance of Anvil Clouds over Lakes Huron & Erie.These can be observed over Lake Michigan as well. The larger an Anvil Cloud is the more dangerous it is.

  9. I’ll touch an several posts.

    Mobile radar. Time and distance. And storms never follow the roads. Plus many thunderstorms never become severe and fewer still become Supercells. Meteorologists, researchers and students burn up a lot of fuel every summer in tornado alley trying to get mobile radar and other instruments in the path of storms.

    TV station Doppler radar isn’t comparable to NOAA but it’s better than nothing. But I bet there aren’t many local TV stations in the NOAA gaps. Even if there is, are there enough of them to scan storms from multiple directions?

    In the advanced Skywarn sessions, they present images of storms from multiple radars from different directions and distances. It is a real wake up call to see how different the images taken at the same time look based on direction and elevation.

    False alarms. Amazing how many people totally ignore any warnings. It was either last year or this year that a bunch of people complained to a TV station when it interrupted a reality show with a warning?
    Talk about ‘real’ reality.

    The basic Skywarn session does a great job presenting and quizzing on what is real and what are look a like cloud features.

    If you aren’t able to attend a Skywarn session in your area next spring there are online study guides on storm spotting.

  10. As one who lives in central Oklahoma, tornado capitol of the world, with more local severe weather coverage and available weather education from TV weathermen, than just about anywhere, I find this information interesting but not surprising. I think that the Storm Prediction Center here takes model predictions of low level wind patters into account when publishing their Severe Weather Outlooks, already. That’s based on what I’ve gleaned from the discussion sections, anyway..

Comments are closed.