You may see the above chart with our preliminary finding that "tornado alley" (the gray shaded area) runs from Oklahoma City to Atlanta, and think we have it oriented the wrong way. Traditionally, "tornado alley" has been defined as the Great Plains of the United States, from Texas up through Oklahoma, Kansas, Nebraska, and Iowa. However, our research at UAH, in coordination with P. Grady Dixon at Mississippi State University, shows otherwise.
We want to show where the area with the greatest risk for tornadoes is. The Storm Prediction Center's tornado database goes all the way back to 1950, and includes almost every reported tornado since then. However, there are some biases in that data that must be addressed first.
1. Unreported tornadoes
The charts below show a) the total number of reported tornadoes, by year, in the U.S. since 1950, and b) the number of reported "significant" tornadoes, or those rated F2 through F5.
Note that the number of total reported tornadoes has been increasing for 60 years. This is because many tornadoes, especially small, weak ones, used to go unreported. Events changed our awareness of tornadoes: Tornado research in the 1950's; developing a spotter network in the 1960's and 1970's; Doppler radar in the 1990's to tell us where to look; and then almost no tornado goes unreported anymore with widespread towercams and cameras on cell phones. The number of reported tornadoes has gone up. They were always there, we just did not know about a lot of them. Especially the weak ones., especially in the southeastern U.S. where trees and mountains block the view, and about half of all tornadoes occur at night.
The bottom chart shows reported significant tornadoes per year. Notice they have been much more steady. F2-F5 tornadoes are wider, last longer, and do more damage than F0-F1 tornadoes, so it is unlikely for very many of those to go unreported. That is apparently the case, since the number of reported F2-F5 tornadoes has been steady or actually dropped. So, in our study, we only use significant tornadoes, a steady dataset.
2. Over-estimated tornadoes
Notice in the lower chart that the number of significant tornadoes decreased suddenly after 1973. The Fujita scale was invented in the early 1970's, and tornadoes 1950-1972 were rated retroactively, using newspaper clippings, etc. that, according to some researchers in Oklahoma (e.g., Schaefer and Edwards 1999; Brooks and Craven 2002), likely biased the damage estimates upward. Therefore, in our study, we only look at data since 1973, since tornadoes were rated on the Fujita scale in near real-time.
The map below uses GIS software and the database of all significant (F2-F5) tornadoes, 1973-2011, to show the total path length (in km) of significant tornadoes, within 25 miles of a point, averaged annually.
Clearly, the largest threat for significant tornadoes since 1973 has been over Arkansas, Mississippi, and north and central Alabama, with a moderate threat extending from central Oklahoma to NW Georgia. The above map is influenced somewhat by the jumbo outbreak of April 27, 2011, but using data 1973-2010 shows a similar overall pattern.
Some may argue that significant tornadoes go unreported in Texas or Nebraska, or are under-rated because they only hit open grassland. However, the number of tornado watches and warnings issued over the past 10-20 years shows that conditions were favorable for tornadoes, or tornadoes were indicated on radar, in roughly the same areas that our statistics show as tornado alley. See below.
There are other statistics that we use to show that our data are correct, but I will reserve those for the peer-reviewed journal article we will be submitting this spring. That means these results are preliminary and have not been peer-reviewed. However, you can be the judge of where tornado alley should be.
Dr. Tim Coleman
UAH Research Meteorologist
Fox 6 Severe Weather Expert and Blogger
Follow me on twitter at timbhm