Skip to content

Are Tornadoes Getting Stronger?

December 13, 2013

By Paul Homewood


WUWT drew attention to a presentation by Professor James Elsner of Florida State at the AGU meeting this week. Elsner put up the above graph, which suggested that tornadoes were becoming stronger.

Scientific American have the story:


Sometimes scientists can’t help themselves from showing dramatic curves, even though they have so many caveats that no firm conclusions can be made from the data. James Elsner at Florida State University has a killer curve, and lots of caveats. The curve indicates that tornadoes in the U.S. may be getting stronger. The caveats indicate they may not be.

“If I were a betting man I’d say tornadoes are getting stronger,” he noted on Tuesday during a lecture at the annual American Geophysical Union fall meeting in San Francisco. But when asked directly at a press conference whether that is the case, he would not commit. “I’m not doing this [work] to establish the future intensity of tornadoes,” he explained, but to establish a method that someday could indeed determine if the storms are becoming more powerful.

Because the lecture was titled “Are tornadoes getting stronger?” the audience expected an answer. And their consternation rose when Elsner showed his final graph, adding up the kinetic energy of tornadoes each year from 1994 to 2012. The curve is flat from 1994 to about 2006 but then spikes upward through 2012. It was reminiscent of the now famous “hockey stick” graph produced by Michael Mann and colleagues a decade ago, indicating that Earth’s temperature had been flat for 1,000 years and began spiking upward in the mid-1800s. But Mann had 1,000 years of data; Elsner has 18. His data begin in 1994 because that’s when Doppler radar, the best at tracking tornadoes, began covering the entire U.S.

The point of the curve, however, is to show that measuring the length and width of a tornado’s damage path gives an accurate indication of its strength, which is driven by the storm’s peak wind speed. It is difficult if not impossible to measure that speed directly, as is done for hurricanes by ground instruments and planes that fly into the storms.


Elsner’s presentation raises a number of questions.



1) There are already well established methods for measuring the strength of tornadoes, the Enhanced Fujita Scale, which estimates wind speeds from damage on the ground. This method shows clearly that the long term trend of stronger tornadoes has been decreasing since reliable records started to be kept around 1970. (More on this below).

It is not clear why this system needs to be replaced.


2) Elsner’s method is based on length and width of tornadoes, on the basis that bigger tornadoes tend to be stronger.Let’s leave aside the matter of whether this correlation is a real and reliable one, and also the question of why we need the proxy of length and width to tell us what we already know.

How reliable is historical data on tornado length and widths? Elsner himself accepts that his analysis can only start from 1994, when Doppler began to track tornadoes. But this straightaway raises the question – why does his graph above start in the mid 1970’s and not 1994? (Sorry for the quality of the graph).

(And I might also add – why does the graph not start in 1970? More later)


3) Even with a start point of 1994, a number of problems remain with the Elsner approach:-

a) Although Doppler began to be rolled out in 1992, the system was not fully operational until 1997. This effectively reduces the period of Elsner’s analysis to just 15 years.

b) Until 1994, it was the mean widths of tornadoes that were measured. After 1994, it has been maximum width. This means that the period prior to 1994, which Elsner shows on his graph is not just dubious, it is totally incompatible with the post 1994 period.

Was he aware of this change of practice? If he was, why has he produced a positively misleading graph.


4) How reliable are past records of tornado width and length?

As NOAA describe, the survey of tornadoes has always been , and still is, very much hit and miss:

Who surveys tornado damage? What’s the criteria for the National Weather Service to do a survey? This varies from place to place; and there are no rigid criteria. The responsibility for damage survey decisions at each NWS office usually falls on the Warning-Coordination Meteorologist (WCM) and/or the Meteorologist in Charge (MIC). Budget constraints keep every tornado path from having a direct ground survey by NWS personnel; so spotter, chaser and news accounts may be used to rate relatively weak, remote or brief tornadoes. Killer tornadoes, those striking densely populated areas, or those generating reports of exceptional damage are given highest priority for ground surveys. Most ground surveys involve the WCM and/or forecasters not having shift responsibility the day of the survey. For outbreaks and unusually destructive events–usually only a few times a year–the NWS may support involvement by highly experienced damage survey experts and wind engineers from elsewhere in the country. Aerial surveys are expensive and usually reserved for tornado events with multiple casualties and/or massive degrees of damage. Sometimes, local NWS offices may have a cooperative agreement with local media or police to use their helicopters during surveys.


Path width is usually determined from these damage reports, but can we rely on the accuracy of this when there is no consistent approach?

Also, into this mix we can throw mobile Doppler radar. For instance, the official report on the El Reno tornado, earlier this year, states:

The maximum tornado width was 2.6 miles. However, the damaging wind swath was much larger, as non-tornadic downdraft winds extended for at least a mile south of the tornado. Given the difficulty of separating this damage from tornadic damage, the Oklahoma University RaXPol radar was used to help determine the width.

This particular radar system is relatively new, and obviously can record only a few tornadoes each year. Without this system, would the estimated width have been less?

It is worth noting, with regard to the El Reno tornado, that although it was originally categorised as EF-5, it is now officially just an EF-3. The storm report explains:

The RaXPol radar data shows winds of at least 295 mph very close to the surface. These intense winds were present in very small sub-vortices within the larger tornado circulation. An analysis of the high resolution radar data combined with the results of the ground damage survey indicates that none of these intense sub-vortices impacted any structures in rural Canadian County. So despite the measured wind speeds, surveyors could not find any damage that would support a rating higher than EF3 based solely on the damage indicators used with the EF scale.

Put simply, modern radar can find things that earlier systems could not.


5) There are examples of earlier tornadoes, where the “official width” is much less than the detail report states. From a 5 minute scan, I found two examples, which I summarised here.

One was in Hale County, Texas, in 1968, where the width is given as “two to three miles wide”. You may recall that the El Reno tornado is claimed as the “widest on record” at 2.6 miles.

The second example was also in Texas, but in 1997 – in other words, after the change in procedure to maximum width in 1994, and after the introduction of Doppler. The official listing states a width of 650 yards, yet the detail gives a figure of three quarters of a mile.

The existence of these discrepancies, particularly the latter, must cast grave doubt on the reliability of even recent data records.


6) Urban sprawl. Determination of tornado width relies on damage to observe, so where there is little to damage, there is nothing to observe. As urban sprawl increases, we are likely to see tornadoes getting “bigger”.


7) An interesting example of the above point is the Mulhall, OK tornado of 1999. From Wikipedia:

However, a possible contender for the widest tornado as measured by radar was the F4 Mulhall tornado in north-central Oklahoma which occurred during the 1999 Oklahoma tornado outbreak. The diameter of the maximum winds (over 110 mph (49 m/s)) was over 5,200 feet (1,600 m) as measured by a DOW radar. Although the tornado passed largely over rural terrain, the width of the wind swath capable of producing damage was as wide as 4 mi.

(The official width given for the Mulhall tornado is 1760 yards) 


However, let’s ignore these problems and take Elsner’s conclusions at face value, that since 1994 tornadoes have been getting stronger. Well, if you look at the conventional data, you see exactly that. There has been an increasing number of EF –3 + tornadoes, largely influenced, it must be said, by the spike in 2011.




But, of course, we know that is not the whole story. When we go back to 1954, we find that stronger tornadoes were much more common up to the mid 1970’s (when Elsner’s graph conveniently starts), and that, with the exception of 2011, the last decade has not been abnormal in any way.





I appreciate that Elsner is trying to establish a method for measuring tornado intensity for use in future. Such a method, though, would need a consistent, reliable and quality assured method of measuring width and length.

It must, however, be borne in mind that such measurements were not made in the past for climatological purposes, but rather for local and practical reasons. Any comparison with past years should, therefore, be treated with a good deal of caution.

It also seems more than a little ludicrous to be making any claims about trends of just 15 years. Would Elsner agree that global warming had ceased to exist, simply because of the 17 year standstill in temperatures? Any analysis of trends would surely need to be based over at least 60 years, so as to cover the full cycles of both PDO and AMO. Is it just a coincidence that strong tornadoes were more common during the cold phase of the PDO up to the mid 1970’s, and are now starting to increase again with the recent PDO switch?  ( Roy Spencer has a detailed explanation of the mechanics behind why this is so).

The presentation of a graph which starts in the mid 1970’s, despite Elsner admitting that data prior to that is unreliable, is highly misleading, particularly since it does not include the years prior to 1976 when strong tornadoes were much more frequent. It is hard to avoid the conclusion that Elsner wanted to present a “hockey stick”, which would not have been possible with a start point of 1997. The long “stick” is, of course, needed to generate an impression of “normality” prior to the uptick of the blade.

The graph, as it stands, should be withdrawn.


As for the question, “Are tornadoes getting stronger?”, perhaps Mr Elsner might care to look at the work Roy Spencer has done, which suggests that tornadoes are worse in a cold world, not a warm one.

  1. Brian H permalink
    December 13, 2013 7:14 pm

    It’s going to be funny: as the world cools and storms get worse, Warmists will be declaring it must really be getting warmer because they’ve been trying to tie storms to warming all along. Personally, I hope for warming and its reduced storminess.

  2. Andy DC permalink
    December 14, 2013 12:01 am

    It would be interesting to see what they would have come up with respect to the 1925 tornado that killed around 700 people without hitting a major city.

    Also, all the graphs seem to show is that 2011 was an outlier during an otherwise quiet tornado period.


  1. At AGU 2013: Are tornadoes getting stronger? No. | Watts Up With That?

Comments are closed.

%d bloggers like this: