Are We Getting Better at Predicting Tornadoes?

Photo: iStockphoto

In a recent podcast, we took a close look at why it’s so difficult to predict earthquakes, even though several million occur each year around the world. After a series of violent tornadoes tore through much of the American south over the last few days, killing at least 43 people, it’s a good time to ask: how good are we at predicting tornadoes? Even as parts of the country recover from one of the deadliest tornado outbursts in years, a handful of innovations are taking shape that should begin improving our ability to predict when a twister might strike.

First, some context. Tornadoes are almost entirely an American dilemma. Thanks to the unique blend of arctic and tropical air currents that flow across the U.S., no other country gets more tornadoes than the 1,000 or so that touch down in the U.S. each year. As evidence of their sporadic, unpredictable nature, annual death tolls from tornadoes are highly erratic, going along at double-digits with periodic spikes of triple-digit annual causalities every few years. For example, in 1974, 366 people died from tornadoes. In 1975, the number shrunk to 60. (See here for a complete list of annual death tolls from tornadoes.)

While advancements in warning systems over the last 50 years, mostly from the National Weather Service, have significantly cut tornado deaths, there remain periodic spikes. Like in 2008, when 125 people were killed by tornadoes. Of course when evaluating the number of fatalities, we have to consider the broader demographic shifts south and west that have led to increased populations in tornado hotbeds like Texas, Oklahoma, and Missouri over the last decade.

But it appears we may be entering a period of improvement. In Norman, Okla., smack in the middle of Tornado Alley, meteorologists at the National Weather Service have begun testing the largest upgrade to the nation’s weather radars since Doppler was installed in the early 1990s. The new technology, called ‘dual polarization,’ is specifically designed to improve tornado spotting. Over the next two years, the upgrade will be installed at all 169 National Weather Service Doppler radars. By improving forecasting, it should also speed our ability to get the word out through more targeted and timely weather alerts on TV and online.

Since many tornadoes are spawned by violent thunderstorms, senior meteorologist Henry Margusity has developed a classification system for thunderstorms using what he calls a “TS Scale,” similar to the Enhanced Fujita scale used for classifying tornadoes and the Saffir-Simpson scale used for hurricanes. Margusity has been beta-testing the system on his blog and apparently hopes to have the system available to download on mobile devices sometime soon.

The current master of tornado prediction is Dr. Greg Forbes of the Weather Channel, whose recently developed Tornado Condition Index system, or TOR:CON, is widely seen as the most accurate at the moment. TOR:CON calculates a percentage risk of a tornado developing within 50 miles of a specific location due to thunderstorm activity. Forbes has developed somewhat of a cult following this spring. The wall of his Facebook page is filled with comments from his 10,000 fans, asking him questions and thanking him for his predictions.

Amendment: Post has been amended to remove mention of meteorologist Kevin Martin.





It was obvious there was going to be a massive tornado outbreak across the South. SPC had it nailed early on, and by Friday the computer models are all on board with massive shear, strong instability and a negatively tilt trough. Classic setup.


"Tornadoes are almost entirely an American dilemma."

Typical economists trick, pick the metric that supports your argument not the one that best describes the data. :-)

You say "no other country gets more tornadoes than the 1,000 or so that touch down in the U.S. each year" but the US is a large country. Instead of the absolute number of tornadoes, picking the more neutral metric of tornados per unit land area tells a very different story.

Quoting your own source for the figure of 1000 tornados per year :

"In terms of absolute tornado counts, the United States leads the list, with an average of over 1,000 tornadoes recorded each year. A distant second is Canada, with around 100 per year. Other locations that experience frequent tornado occurrences include northern Europe, western Asia, Bangladesh, Japan, Australia, New Zealand, China, South Africa and Argentina. In fact, the United Kingdom has more tornadoes, relative to its land area, than any other country."

The UK gets 3.5 tornadoes per 10,000 square miles per annum, the USA 2.6. If you then also account for population density (the UK's is approximately 4 times that of the USA) the UK resident is 5.4 times more likely to encounter a tornado than a US resident.

However, tornado distribution in the UK is fairly uniform, in the US it is anything but uniform (from 0/10k sq mi in Alaska to 8/10k sq mi in the south west) and in the UK a tornado will rarely have more than 70 miles to run before it hits the sea and dissipates. That has two effects. Firstly, with shorter lifetimes UK tornadoes collect less energy and so mostly do less damage. Secondly, the high density of tornadoes in the US south west makes it more likely that a tornado will reach newsworthiness so that a resident of Washington State (0.15 tornadoes/10k sq mi) will perceive tornadoes as a threat in the US whereas a London resident (3.5 tornadoes/10k sq mi) would laugh at you if you told him that tornadoes were a threat to him.


Jeremy Gibbs

I thought it necessary to point out the lack of due diligence on the part of the author.

Everything else aside, Mr. Kevin Martin is mentioned at the end of the article in context of an expert. This is beyond laughable. Mr. Martin has a long and lurid past on the internet and is widely regarded among the weather community as an unstable, habitual liar. He has zero formal education in meteorology. He has routinely taken products from both government and personal websites and claimed them as his own. The other products he claims as his own are garbage and nothing more than Microsoft-Paint-quality ripoffs of other organizations' weather graphics. He has even urged his followers to ignore official government warnings. I would whol-heartedly stake my professional reputation that there is no T.R.A.M. developed by Mr. Martin.

In addition, Mr. Martin has been removed from many websites due to death threats and other malicious behavior against other members. I can speak to this personally. In May of 2010, I personally contacted Mr. Martin to convey my concerns of his dishonest behavior and how the public's safety could be at risk if they heeded his misinformation. What followed were a slew of threatening emails and facebook messages from both Mr. Martin and his brother, Brian Martin. In these messages, my life was threatened in no uncertain terms. Additionally, Mr. Martin used his position at to print libelous articles and fake user comments about myself. He even made calls to my place of employment to further his harassment. I would be more than happy to provide the author with copies of these emails.

In response to this, I made his employer at aware of his behavior by forwarding the threatening emails. Mr. Martin was subsequently terminated from his position. He then proceeded to call my place of employment again and was subsequently contacted by our attorneys, making it perfectly clear that any further contact would be dealt with legally.

A simple Google search would have made it clear that Mr. Martin is no expert. It is disappointing that a site as respected as this would give such a person an endorsement. Never mind the other inaccuracies and misrepresentations in this article. Shame on you for not doing your due diligence as an author.

Jeremy A. Gibbs
Ph.D. Candidate
School of Meteorology
University of Oklahoma


Matthew Philips

Hi Jeremy,

Thank you for your comments regarding Kevin Martin. All reference to him has been removed from the post. Very sorry for the oversight.

-Matthew Philips

Jeremy Gibbs

Thank you for not giving this guy a voice.


Would we have as many tornadoes if Paul Bunyan hadn't cut down all the trees?


In other novice statistics news, [circumstantial evidence] indisputably disproves [pretty much solid theory].


This was the most idiotic thing I've read on the internet in a very long time, for all the reasons previously mentioned by commenters. There wasn't a real meteorologist in the country who didn't know the atmosphere was going to explode this weekend.

Kevin Martin? He is NOT a meteorologist. KMart is a mentally ill kid in California that pretends to be able to forecast the weather. He is a fraud and a scam artist (just Google, you'll see), and his "work" is going to end up killing gullible people some day. Furthermore, his Facebook page you link to shows that he "likes" Dr. Greg Forbes' page, NOT vice versa.


I learned about this article only after reading a rant about it by a friend:

This article is truly an epic failure of common journalistic sense. Next time, please get a scientist to vet your science stories.


"Are We Getting Better at Predicting Tornadoes?"


Sorry for the snark, but... we've gone from no lead time (your tornado warning was the tornado approaching) to an approx 15 minute lead time (nearly double that for the May 1993 outbreak in Oklahoma).

That's due to... oh yeah: an improvement in tornado forecasting.