Tornado Alley Is Drifting Southeast,
Not Racing East

The claim everyone has seen

You have probably seen the map: a blob labeled "Tornado Alley" sliding off the Great Plains and into the Southeast, with a caption saying it is moving east at 10 to 20 km a year. It has run in major newspapers and is backed in part by real research on where tornadoes form. The worry behind it is genuine. The Southeast is far more dangerous ground for the same tornado: more people per square mile, more mobile homes, more tornadoes at night, and more trees hiding them until they are close.

We wanted to measure the migration directly, in our own copy of the full tornado record, with every analysis choice locked down in writing before we looked at the answer. What we found is that the direction in the headline is right but the number is badly wrong, and the real story is more interesting than the one being shared.

Why we only counted the strong ones

The US tornado record has a trap in it. We record far more tornadoes today than in 1950, but mostly because we got much better at spotting them: storm-spotter networks, Doppler radar, smartphones, dashcams, and security cameras. That improvement falls almost entirely on the weakest tornadoes. The number of recorded EF0 tornadoes (the weakest rung of the Enhanced Fujita damage scale) roughly quadrupled between the 1950s and today. The number of strong ones did not.

So if you throw every tornado into the calculation, the "center" of tornado activity partly tracks where people and cameras grew, not where tornadoes actually are. The fix is simple: count only significant tornadoes, rated EF2 or stronger. A violent tornado that levels houses was never going to be missed, even in 1950. Across 76 years there are 13,400 of them, and their geography is honest. That is the population behind every number on this page.

The center is moving southeast, and slowly

For each year we found the geographic center of significant-tornado activity (weighting each tornado by how long its track was, since long-track tornadoes are the hardest to miss). Then we watched that center move. It drifts east at 2.7 km a year and south at 2.3 km a year. Those two speeds are essentially equal. Added up over 76 years, the center has moved about 203 km east and 179 km south: a near-45-degree diagonal toward the Southeast, not a straight march east.

The southward half of that motion is as solid as the eastward half. Drop any single decade from the record and it barely moves. Whoever wrote "moving east" was telling you about half of the trip.

The viral number is three to seven times too fast

Our measured eastward speed, 2.7 km a year, sits a factor of three to seven below the 10-to-20 figure in circulation. Even the most generous reading our data allow, about 3.6 km a year, is less than 40 percent of the slowest version of the popular claim. We are measuring where strong tornadoes actually touch down; if the viral number is measuring something else, such as the eastern edge of the atmosphere's tornado-friendly zone, then it is simply not the same quantity people think it is. Either way, the place where the worst tornadoes happen is moving far more slowly than the headline suggests.

What is really happening: the Plains empty, the Southeast fills

The diagonal drift is not a solid alley sliding across the map. It is two separate things happening at once. The Great Plains (Texas, Oklahoma, Kansas, Nebraska) are in real decline: their significant tornadoes fell from about 74 a year in the early record to 32 a year recently, and their share of the national total dropped from 36 percent to 24 percent. The Southeast (Dixie Alley: Arkansas, Louisiana, Mississippi, Alabama, Tennessee, Georgia, Kentucky, Missouri) did the opposite, climbing from 25 to 40 percent of the national total while holding steady in absolute numbers.

There is one obvious objection, and we checked it. The total number of significant tornadoes nationwide also fell over this period, partly because the rating scale changed in 2007. Could the "Plains decline" just be that nationwide drop showing up locally? No. If it were, every region would fall together. Instead one region rose while the other fell, and the Plains fell about 1.6 times faster than the country as a whole. A uniform national decline cannot make one region go up and another go down. The split is real.

No sudden jump, just a slow slide

We also looked for a single year where the pattern "switched on," the kind of clean break that makes a tidy story. There is not one. The best candidate year is different depending on how you count tornadoes (1963 one way, 1983 another), and a model with a sudden jump barely beats a plain straight line. The honest description is a gradual, continuous slide over the whole 76 years, not a recent regime change.

Why the slow version still matters

It would be easy to read "the viral number is wrong" as "nothing to worry about." That is not the lesson. The drift is slow, but it points the wrong way. The Southeast already suffers more tornado deaths than the Plains for an identical storm, because of its people, its housing, its nighttime tornadoes, and its terrain. A migration of the strongest tornadoes toward that ground, even a creeping one, stacks risk on top of existing vulnerability. The right headline is not that the alley is racing east. It is that the country's most violent tornadoes are slowly concentrating where they do the most harm.

Reproducibility

Every number on this page comes from the tornado-alley-east workspace: extraction from the National Oceanic and Atmospheric Administration (NOAA) Storm Prediction Center Tornado History database (1950 to 2025), a frozen pre-registration of all methods, path-length-weighted yearly centroids, and a cluster bootstrap that resamples whole storm-days (because tornadoes on the same day are not independent: the 13,400 significant tornadoes collapse to 5,044 distinct storm-days). The analysis was reviewed adversarially before publication. The full results.json, scripts, and the paper are linked below.