December 12, 2021

A Critical Gap in Tornado Warning Technology: Lessons of the Recent Tornado Outbreak

A terrible tragedy occurred on Friday evening, as strong tornadoes struck across a multi-state swath from Arkansas through Kentucky.  Early estimates suggest that 50-100 individuals lost their lives and hundreds were injured.

A destroyed candle factory in hard-hit Mayfield, Kentucky

The death toll was undoubtedly enhanced by the nighttime occurrence of these storms and their development during the winter season, which is unusual but not unprecedented.  The storms occurred in two coherent lines oriented southwest to northeast, as shown by the tornado reports (red dots) provided by the NOAA/NWS storm prediction center.  For reference, the tornadoes hit Mayfield, KY around 10 PM Friday evening.

My colleagues at the National Weather Service (NWS), both at the Storm Prediction Center and the local NWS forecast offices (such as Paducah, Kentucky) did an excellent job in predicting the threat and provided timely watches and warnings before the tornadoes struck.

Surely these skillful forecasts and timely warnings saved lives.   But they weren't enough to prevent massive loss of life. 

In this blog, I will review the forecasts/warnings and suggest that new smartphone-based warning systems are both needed and possible.  Technology that could save many lives from such severe storms.

The National Weather Service Forecasts

Earlier that day (1:17 PM), the Paducah National Weather Service Office noted that the expected meteorological conditions were similar to those accompanying previous cool-season tornado outbreaks (see below).

Furthermore, at the same time, the NOAA/NWS Storm Prediction Center (SPC) put out a tornado watch that encompassed Mayfield and other locations that were later hit by tornadoes (see below).

Later in the afternoon, the watches were repeated and became more strident, and as severe thunderstorms, with supercell structure and rotation developed, tornado warnings (which indicate an imminent threat) were communicated (see below for 3:29 PM CST). 

The tornado warning made by the NWS Office at Paducah for Mayfield at 9:03 PM was stunningly good (see below).  This warning was communicated roughly30 minutes before Mayfield was devastated.

I could show you more NWS forecasts, model predictions, and more...and you would be impressed.   The region was given skillful and timely warnings about the potential for strong, damaging tornadoes.  NWS forecasters and the NOAA/NWS Storm Prediction Center staff can be proud of what they did.

But many people still died.     

I would suggest that the technology exists to mitigate the tornado risk substantially.  A technology I have been thinking about for quite a while:  automated smartphone-based warning systems, driven by state of science observations.   But first, let's talk about the nature and limitations of severe storm and tornado forecasting.

Limitations of Tornado Forecasting

Forecasting technology for severe thunderstorms has come a long way during the past decades.  Meteorologists have high-resolution forecasting models (such as NOAA/NWS High-Resolution Rapid Refresh HRRR), improved meteorological radars and satellite systems, improved knowledge of the structure and evolution of severe convective storms and machine-learning techniques now assist severe storm prediction (see model forecast below).
9-h NOAA/NWS HRRR model Forecast at 03 UTC 11 December (9 PM CST 10 Dec)

Improved models and forecast guidance help meteorologists recognize threatening situations and predict the occurrence of severe storms for a region.   However,  roughly three hours or more out, our models cannot predict the exact locations and intensities of incipient severe storms.   Only when the storms have developed and are moving on clear paths can forecasters provide detailed forecasts of tornado occurrence and track....and even then, only for the next few hours.

I should note that NWS tornado warnings have a False Alarm Rate (FAR) of roughly 70%.  So approximately 70% of the warnings don't pan out.

From experience, people and businesses know that the forecasts are not perfect and thus don't necessarily act promptly or effectively to shelter in place or move away from the threat.  And then there is the issue of communication....they may not even know of the threat....  particularly an issue for nighttime storms.

The Predictability of the Recent Event

An important point about the recent event was that the tornadoes were long-lived and moved on a relatively straight, continuous path.   Large, intense tornadoes are often associated with supercell thunderstorms with rotation updrafts.   We can see the large areas or rotation using Doppler radars, which tell us the velocity towards or away from the radar of the precipitation associated with the storm

As shown in the figure below, alternating colors (in this case red and green) in a small region are the sign of rotation in the radar-based Storm Relative Motion (SRM) imagery:

Here is an animation of the SRM imagery from the Paducah radar during the period before the tornado hit Mayfield.   The rotation start was moving straight to the NW with little change in intensity.

We can track this rotation over time, and doing so, the path of rotating storms for the period around the worse tornado occurrence on Friday evening is shown below.  You can see the amazingly straight path of the narrow zone of rotation that extends from Arkansas, into Tennessee, and Kentucky.  There is also a line of rotation associated with other severe storms to the north.

The point of all this is that simple extrapolation of severe storm motion provided an excellent short-term forecast for the next hour or so for this event.

Tornado guidance must be delivered in real-time to individuals or businesses, so they can make optimal, immediate decisions to save life and property.  For tornadoes, folks must determine whether they are threatened at all, and, if so, whether to shelter in place or to get out the way of the storm.  

Smartphone-Based Tornado Warning App

Almost everyone has a smartphone and such phones know exactly where they are since they have GPS.  

Now obviously, we could feed National Weather Service forecasts, including tornado watches and warnings, to smartphones, providing the forecasts appropriate to the location in question.   This is low-hanging fruit...and obvious.  

But in active situations, things are happening too fast and too localized for NWS forecasters to be guiding people....automated software has to take over.

Using  Doppler radar imagery (which is available roughly every 5 minutes), the instantaneous path of the storm could be determined in real-time and the storm's location and path could be made available on the smartphone, with constantly updated information on the probability that the smartphone's location being overrun.  The estimated time of arrival could be noted.  

But there is more.  Modern radars can also track the debris cloud of major tornadoes....and thus have confirmatory information of the position and movement of the tornado vortex

The smartphone could not only provide a warning of imminent tornado passage but could provide information on the best direction to flee if no sheltering location is available--generally at right angles to the storm path.   A distance of a few thousand feet can make a huge difference if there is no place to shelter.

It is also possible that smartphones THEMSELVES can provide critical weather information to facilitate determining tornado location and movement.  Many smartphones possess very capable pressure sensors, something I know about because I have been researching the meteorological applications of such sensors for over a decade.

Tornadoes have a large pressure signature and smartphones could collect and communicate such pressures in real-time.  Such smartphone pressures observations could be used by automated systems to determine tornado intensity and path, and thus improve guidance to individuals making critical decisions.

I have tried talking to Apple and Google about helping collect smartphone pressure data to improve weather forecasts.  So far, they have listened politely and done nothing.  Perhaps that could change.  Or perhaps Amazon could help.

In summary,  a smartphone app, applying real-time meteorological data, could provide extraordinary valuable guidance during tornado outbreaks, and perhaps save many lives.

It could be done.  And I suspect only a government organization would take it on because of liability concerns.


  1. But why this time of year? With insolation at a minimum, how doe s the necessary instability develop?

    1. There was a warm dome of air over the southeast and a cold front coming in from the north. The difference in the temperature drove the storms. I grew up in Arkansas and the really bad storms were always driven by this type of set up. There were thunderstorms (and occasional tornadoes) caused by instability that you asked about in the summer, but those were isolated and generally occurred in the afternoon or evening.

  2. Seems like in Tornado Alley, the building codes ought to call for reinforced concrete or steel construction for all new business and residential buildings...

    1. I'm not sure that any reasonably priced structure could withstand a direct tornado hit. A tornado that hit central Texas many years ago lifted parts of an asphalt road right off the ground! I do know that mobile homes in certain parts of the country are required to be tied down as opposed to relying on gravity to hold them down. Having lived in southern parts of tornado alley in the past, it's an unspoken gamble that you take living in that part of the world. Just like we gamble that Mt Rainier won't erupt or a Cascadia subduction zone won't send a tsunami our way. We can prepare only so much.

  3. Excellent discussion sir!..but with the current anti-government attitudes permeating our society. I wonder of people would even approve, or heed such a warning system!

    1. “Anti-government attitudes permeating our society”? Are you referring to people not wanting the government to force them to take a shot? Because that’s an entirely separate issue from weather forecasting. I don’t think anyone is against the Weather Service.

  4. Excellent post, Cliff, which I have shared. I just completed your UW class as an ACCESS student and wish there was a session today to hear your comments on this. Thanks for the class, your podcast, and this blog.

  5. This is an excellent meteorological review of the recent tragic outbreak, and related discussion regarding the state of the current NWS severe convective storm forecast & warning capabilities. One thing that should not be ignored is how individual (and also localized groups of) human beings act upon such warnings. You did mention the FAR and its impact upon recipient response to the threat (I would think especially in terms of financial considerations, and particularly businesses & medical/nursing facilities). However, even when warnings become more imminent and urgent as events near, people in communities given such knowledge (and also asssuming an ideal communication of the evolving meteorological situation) are always going to respond across a wide range of actions due to health, structural, cultural, and mobility differences alone. On top of that, people just can make bad spontaneous decisions due to both noble and less noble/more selfish reasonings in such high duress moments. It will be very difficult to control for those kind of effects (many of them seemingly random personal choices) that would tend to weigh on how the last 15-30 min before an impact play out....but continual chiseling away at the FAR would at least seeem to allow for more urgent protective actions to be taken by individuals and businesses hours rather than minutes before an event like Mayfield & Dawson Springs. Thanks Cliff, and hope you are well.

  6. SPOILER: They won't.
    Have to echo the sentiment that the weather service did everything exactly right. When political and other convictions outweigh self preservation so often (pandemic as an example) nullify what is already in place, then why spend money on anything more elaborate/intrusive? The citizens of the United States demand the personal freedom to make our own bad decisions.

    1. Yes, we actually do have the right to make our own bad decisions. Thanks so much for acknowledging the most basic of US citizen rights.

    2. Which is fine. Can't save everyone. Hopefully those folks at NOAA in Kentucky believe they have done all they could. Government employees in that position understand that they can't make the horse drink at the watering hole you clearly mapped out for them.

      However, it is not always easy to have the technology in hand to mitigate further death and destruction, but have to resign yourself to the fact that you have to allow Darwin his due just based upon the prevailing national culture. That and there is the invasiveness and the cost with technology and/or building codes that a nation adverse to any concept of civic mindedness will naturally balk at.

      Amazon warehouses are stand up precast with a focus on being cheaply and rapidly built. They are staffed with employees who have to pee in bottles at their workstations. Even if their was a storm shelter, would their corporate culture even allow them to go to it? Well, Bezos made that an easy choice by not providing one. Why should he? There is no law saying he has to included a storm shelter in his buildings and said law might impact placing his warehouse where there are no such rules. Its a race to the bottom.

      Based on who we are, and where this tornado hit, all that could be done WAS done to mitigate this. Cliff is a master of documenting cases of "We can do better. The technology exists!".....but...this is America. Sorry Cliff. You are the One Eyed Man in the Land of the Blind sometimes.

    3. The solution to the issue you raised regarding Amazon is being answered by the increasingly successful efforts to unionize their warehouse workforce. Government edicts won't make any company move faster that unified employees.

  7. I grew up in Arkansas and I'm very familiar with tornadoes. I heard that one tornado was on the ground for over 200 miles. I've never heard of a tornado being on the ground that long. Is this a new phenomenon?

    1. Are you referring to a Derecho? They are similar, but start overland and can last much longer than most tornadoes, like the one that began in Wisconsin and went all the way to Ohio in August of 2020.

  8. Excellent post that I hope will be listened to by the responsible parties. One thing that bothers me is why the large factory and the Amazon warehouse workers didn't get a warning. You would think those organizations would be looking out for their people and had the resources to stay on top of the forecasts.
    I do recall getting a smart phone tornado warning in Edmonds when the threat was actually at Hood canal, but that's something that can be fixed with the GPS you mentioned.

    1. U.S. News:
      Factory Workers Say They Were Told They'd Be Fired If They Left Job Amid Tornado Warning
      Mayfield Consumer Products officials deny the claims, but at least five employees say they were told they couldn't leave the Kentucky candle factory to seek safety.

  9. One aspect of this tornado outbreak that hasn't been reported on enough is that one tornado stayed on the ground in KY for over 200 miles (!). I don't believe that's exactly an anomaly, but it's definitely an unusual one. Another relevant point is that many people living in vulnerable areas (of which I lived in where I grew up, 50 miles west of Chicago) have the opportunity to build low cost tornado shelters that can be constructed within the confines of their homes, rather than the shelters that were build outside and underground (as in the past). Since so many deaths from tornadoes often occur during the nighttime hours when people are usually the most vulnerable, this could help alleviate the most destructive parts of severe weather events of this kind.

  10. Thanks Cliff.
    This is a great effort on your part.
    - - - - - -

    I’ve only had a “smart phone” (still learning) for half a year.
    Three times the phone blurted out an alert for wildfires nowhere near me. The really odd thing is that I had to search up a street address or location on the internet to find out where the danger was. Exp: 25 mile road

    This is in the Great Left Coast State of Washington - - home to some of the best technology and fire fighters in the world.

    They need to do better - - else “the boy who cried wolf” will have a modern meaning.

  11. Local governments have been able to set up alert systems in line with what you are proposing. My phone sounded an alarm along with all the other contractors at a job site a few weeks ago telling us to either evacuate or be prepared to stay in place for several days due to flooding (Skagit County) and the location of where we were working. The alarm was very specific to where we were and was in Spanish for the contractor who spoke Spanish.
    It would be interesting to know how many local county emergency preparedness systems utilize this approach for tornadoes or other rapidly evolving emergency issues.
    There has been funding to develop earthquake warnings to shut off gas and power and give people a few seconds of warning before the shock waves arrive. If that can be done (it has been done in Japan) for earthquakes, it surely could be done for tornadoes.
    If Google maps can tell us how to avoid traffic jams in real time, it seems that telling us how to avoid tornadoes is readily achievable.

  12. We don't even need a special app for this. All smartphones of recent manufacture have the ability to receive broadcast emergency alerts. That technology is still pretty basic, but could be expanded to provide additional information so long as it isn't too bandwidth-consumptive. Works great for Amber Alerts.

  13. I grew up in tornado country. We always had a storm cellar and sheltered in it several times. Now I see videos of destroyed homes that were built with no real place of refuge. I agree more immediate warning might save lives. So would building codes that required a below ground refuge. I would never live in the Midwest without such a shelter. I saw too much damage and too many lives lost to take that risk.

    1. Remember the tornado sirens? When they were testing the system, you would here them wailing away all over the city at high noon.

  14. I grew up in tornado alley. The warning systems in place are sufficient - provided people respond appropriately, or are allowed to. The deaths at the candle factory have been attributed to supervisors' threats to fire anyone who left to take shelter.

  15. Yes, Ballard, you are correct in that the warning systems are sufficient (NY Times link below) but that it was the human response to the warnings that was a major issues for the deaths. Too many tornado skeptics, not unlike what is happening with climate change.

    "Some said they were not sure how seriously to take the dire predictions, or did not think the storm would hit their neighborhood."
    "They Said the Tornado Would Hit at 9:30. It Hit at 9:30"


Please make sure your comments are civil. Name calling and personal attacks are not appropriate.

Mountain-Top Instability

Over the lowlands of Washington State,  today was sunny and generally clear.  Not a drop fell out of the sky. But something very different h...