Tornado Alley: where it is and how it has moved

Published September 10, 2026 · about 30 minutes · charts computed from the SPC tornado database, 1950 to 2025

Tornado Alley has no border, no official definition and, depending on how you count, no fixed location. The name began as an Air Force research project in 1952 and became a media shorthand for the southern Plains. Seventy-six years of records now show the Plains producing fewer tornadoes than they did, the Mississippi valley and the Southeast producing more, and the deaths concentrated in a region the name never covered. This article sets out where the term came from, what the database and the peer-reviewed literature show, why the Southeast is deadlier, and what is projected next.

A name without a border

The government's position is short. The National Severe Storms Laboratory: "Tornado Alley is a nickname invented by the media to refer to a broad area of relatively high tornado occurrence in the central United States. Various 'Tornado Alley' maps look different because tornado occurrence can be measured many ways: by all tornadoes, tornado county-segments, strong and violent tornadoes only, and databases with different time periods."^[1] The Storm Prediction Center's FAQ says the same in almost the same words, and adds: "Violent or killer tornadoes do happen outside this Tornado Alley every year. Tornadoes can occur almost anywhere in the U. S., including west of the Rockies and east of the Appalachians."^[2]

The term's origin was untraceable until three National Weather Service meteorologists went looking in 2010. Having asked the field's historians, they wrote, "Nobody could provide a definitive answer to the origin of 'Tornado Alley.'" The one reference they found was an article in Take-off, the Tinker Air Force Base newspaper, on January 16, 1953: "On Feb 15, 1952, Fawbush and Miller set up a new project called, 'Tornado Alley,' in which a concentrated study of severe weather activity was made over an area extending from Lubbock, Texas to Eastern Colorado and Nebraska."^[3] Ernest Fawbush and Robert Miller were the Air Force officers who had issued the first tornado forecast at Tinker in 1948, a story told in the companion article on watches and warnings. Their alley ran up the High Plains, well west of where most maps since have put it.

Harold Brooks, Charles Doswell and Michael Kay put the difficulty plainly in 2003: "Tornado Alley as a distinct geographical location is a popular concept, but one that is historically ill-defined."^[4] The tornado historian Thomas Grazulis, quoted in the 2010 paper, went further: "Any area that seems to experience above average tornado frequency is eventually labeled a 'tornado alley.' There are dozens of such regions."^[3]

Where the tornadoes are

The map below is drawn from the Storm Prediction Center's tornado database, the record of every tornado reported in the United States since 1950. It counts tornadoes rated F1 or EF1 and stronger by the cell of one degree of latitude and longitude in which they began, over the thirty years from 1996 to 2025. The two outlines are the conventional alleys: the Plains corridor from north Texas through Oklahoma, Kansas and Nebraska, and the lower Mississippi valley into Alabama and Tennessee. They are drawn here for orientation and have no official status.^[6]

By state, the same thirty years give two different rankings depending on whether you count every tornado or only those that did at least F1 damage. Texas leads either way by sheer size. Per unit area the order changes completely.

Kansas records the second-most tornadoes in the country and ranks twelfth of these thirteen states per unit area once the weakest are set aside; Mississippi, which never appears on a classic Tornado Alley map, is first by a wide margin. The gap between the "all tornadoes" and "F1 and stronger" columns is itself a finding: Kansas and Iowa report large numbers of brief EF0 tornadoes over open country, in a way the wooded Southeast cannot. Areas are computed from the Census Bureau state boundaries; counts from the SPC database.^[6][7]

A season that moves

Part of the reason no single alley works is that the threat migrates through the year. NSSL: "The U.S. tornado threat shifts from the Southeast in the cooler months of the year, toward the southern and central Plains in May and June, and the northern Plains and Midwest during early summer."^[1] Brooks, Doswell and Kay traced the seasonal march in detail from 1980 to 1999: a threat in mid-February "located in Florida and from Louisiana into southern Alabama"; by early April "in northeastern Texas and southeastern Oklahoma"; by May 20 "over the southern Texas Panhandle"; in June northeastern Colorado with an axis into Iowa; and in mid-November a maximum "over southwestern Mississippi" which "is actually the maximum probability for any location in the United States at any time from the beginning of September through the middle of March."^[4]

That regularity is the basis of the most defensible definition of Tornado Alley anyone has offered. Brooks and his co-authors asked where tornado days are both frequent and predictable in their timing, requiring at least half a tornado day per year on their grid and a peak date whose standard deviation was under 20 days. The answer was a band "from the southern Texas Panhandle through Nebraska and northeastward into eastern North Dakota and Minnesota," which they noted "is somewhat west of many of the popular descriptions of Tornado Alley." The reliability test "eliminates the southeastern part" of the frequency pattern entirely, because in the Southeast the peak date wanders by more than two months from year to year. At Hattiesburg, Mississippi, they wrote, "It is practically impossible to define a tornado 'season'," even though the town has only 20 percent fewer tornado days than Lubbock.^[4]

The same paper carried a warning about what the alley leaves out. Of the 21 tornadoes that killed ten or more people between 1980 and 1999, "Only two of those (Andover, Kansas, on 26 April 1991 and Oklahoma City on 3 May 1999) occurred in the area outlined by the high frequency and repeatable season," where about eight would have been expected from its share of tornadoes.^[4]

Reading the record honestly

Before any claim about change, the database's own history has to be understood, because the number of tornadoes reported has doubled for reasons that have little to do with the weather.

Verbout, Brooks, Leslie and Schultz documented this in 2006: "Over the last 50 yr, the number of tornadoes reported in the United States has doubled from about 600 per year in the 1950s to around 1200 in the 2000s. This doubling is likely not related to meteorological causes alone." The stable series is the one that excludes the weakest tornadoes: "nearly all the doubling of tornado reports over the last 50 yr is most likely due to the increased reporting of F0 tornadoes." The strong series has its own problem, a tendency to over-rate tornadoes in the early record, "with a possible discontinuity in the late 1970s."^[8] Coleman and Dixon add that the Fujita scale "was not adopted for rating tornadoes in their near-immediate aftermath in the United States until 1973," so earlier ratings were assigned years after the fact.^[9]

Every chart of change on this page therefore uses tornadoes rated F1 or stronger, compares periods after the 1950s, and treats the F2-and-stronger series as informative but inflated before the mid-1970s. The reporting biases do not disappear at F1, but they shrink, and they act in the same direction everywhere: more people, more radars and more storm chasers raise counts in every region.

What has changed since the 1950s

With those cautions, the shift is not subtle. The chart below tracks tornadoes rated F1 or stronger in three groups of states since 1954, the first year the database is considered consistent.

The strong-tornado columns need the caveat from the previous section: F2+ counts fell everywhere as rating practice tightened, so the honest comparison is between regions within a period. In 1966 to 1995 the Plains produced 40 percent more F2+ tornadoes a year than Dixie Alley; in 1996 to 2025 Dixie Alley produced 30 percent more than the Plains.

What the literature finds

The pattern in the charts above is the same one the peer-reviewed studies have found, each with a different metric and period, and each with its own caveats.

Agee and colleagues, 2016: the heart has moved

Ernest Agee's group at Purdue split the record into two thirty-year halves, 1954 to 1983 and 1984 to 2013, and counted F1 and stronger tornadoes on a grid. Oklahoma's count "decreased from 1096 in Period I to 713 in Period II for a loss of 383 (or a 35% decrease). Tennessee, however, increased from 275 in Period I to 457 in Period II for a gain of 182 (or a 66% increase)." The busiest grid box moved from southeast Oklahoma and northeast Texas, where it fell by 45 percent, to northern Alabama. Their conclusion: "It is proposed that the new 'heart of Tornado Alley' as based on annual totals (and not on any particular season) is now located in central Tennessee/northern Alabama and not in eastern Oklahoma." They were explicit that "no results have been presented to relate this to climate change."^[10]

Gensini and Brooks, 2018: environments agree with reports

The obvious objection to any trend in reports is that reports depend on people. Victor Gensini and Harold Brooks answered it by testing the atmosphere instead. Using reanalysis data from 1979 to 2017 they computed the significant tornado parameter, a combination of instability, shear, moisture and low cloud bases that forecasters use to discriminate tornado days, at every grid point, and found that it explains "44% of the variance in annual tornado counts." Then they looked for trends in both. "Negative tendencies of tornado occurrence have been noted in portions of the central and southern Great Plains, while robust positive trends have been documented in portions of the Midwest and Southeast United States." Where the two datasets agree, "Both tornado reports and environments indicate significant decreasing trends in frequency over portions of Texas, Oklahoma, and northeast Colorado. Agreement in the sign of the Theil-Sen slope is also noted for significant increasing trends in portions of Mississippi, Alabama, Arkansas, Missouri, Illinois, Indiana, Tennessee, and Kentucky."^[11]

They also drew a line under what the result does not say: "our results do not indicate that maximum United States tornado environment and report frequency is no longer in the Great Plains. Rather, we demonstrate there has been a robust downward trend of tornado environments and reports in portions of the Great Plains concurrent with a simultaneous robust increasing trend in areas surrounding the Mississippi River Valley and the Midwest." And on cause: "it is unclear whether the observed trends in tornado environment and report frequency are due to natural variability or being altered by anthropogenic forcing on the climate system."^[11]

Coleman and Dixon, 2014: risk measured by path

A tornado that travels 40 miles matters more than one that travels two, and Timothy Coleman and Grady Dixon measured risk as kilometers of significant-tornado path near a point, from 1973 to 2011. The result: "the area of highest risk for tornadoes in the United States extends roughly from Oklahoma to Tennessee and northwestern Georgia, with the highest risk in the southeastern United States, from central Arkansas across most of Mississippi and northern Alabama." The very highest values ran "from near Jackson, Mississippi, to Huntsville and Birmingham, Alabama," and the pattern "was already apparent in the data prior to the 2011 tornadoes." They noted that Southeast significant tornadoes have longer paths on average, 18.3 km against 14.8 in the Plains, and closed with a warning about the name: "This area is so clearly different than the tornado alley discussed by the media for many years that many Americans are totally unaware of the tornado risk in some of the areas outlined in this study."^[9]

The variability finding

One more result frames all of the above. Summarizing a 2014 study in Science by Brooks, Carbin and Marsh, Gensini and Brooks wrote that "the annual frequency of United States tornadoes through the most reliable portions of the historical record has remained relatively constant. The most notable trends in tornado frequency are associated with increasing annual variability and a recent tendency for more tornadoes on any given tornado day."^[11][12] The national count is not rising; it is clumping, into fewer days with more tornadoes, and moving east. In the database, 2024 had 18 days with 30 or more tornadoes; no year before 2008 had more than 12.^[6]

Dixie Alley, and why it is deadlier

The name has a documented author. Allen Pearson, then director of the National Severe Storms Forecast Center, "explained that he personally coined the phrase after working the Mississippi Delta outbreak of 21 February 1971. This event had 10 long-track tornadoes that resulted in 121 deaths and over 1,500 injuries."^[3] Gagan, Gerard and Gordon, the NWS meteorologists who traced it, defined Dixie Alley as "all of Arkansas, Louisiana, Mississippi and Alabama, western and central Tennessee, and northern and central Georgia," and compared it with a Plains alley of Nebraska, Kansas, Oklahoma and north and central Texas from 1950 to 2007. The Plains alley had more tornadoes in total and per unit area. But for strong and violent tornadoes per 10,000 square miles the order reversed, 89 in Dixie Alley against 70; Dixie Alley had 371 killer tornadoes of F2 or stronger against 205; and in 1986 to 2005, three quarters of the deaths in the two alleys combined were in the southern one.^[3]

The reasons are well established and reinforce one another.

• Night. In the database, 30 percent of Dixie Alley tornadoes rated F1 or stronger from 1996 to 2025 began between 9 pm and 6 am, against 21 percent in the Plains and the Midwest; those night tornadoes account for 30 percent of Dixie Alley's deaths.^[6] Coleman and Dixon, using sun angle, found "48.6% of all southeastern U.S. tornadoes between 1973 and 2011 occurred at night, compared to 39.3% in the GP."^[9] Ashley, Haberlie and Gensini summarize the consequence: evening and overnight tornadoes "are up to 2.5 times as deadly as their daytime counterparts."^[14]
• The cool season. A third of Dixie Alley's F1+ tornadoes fall between November and February, when days are short, storms move fast and few people are expecting them. Gagan and colleagues found that from October to March Dixie Alley had 8.4 killer strong tornadoes per 10,000 square miles against 1.0 in the Plains.^[3][6]
• Long tracks. Among F2+ tornadoes from 1996 to 2025, 12 percent in Dixie Alley ran 25 miles or more, against 6 percent in the Plains.^[6] The Hackleburg and Mayfield tornadoes are the type.
• Visibility. NSSL's VORTEX-SE field program was created because Southeast tornadoes "tend to occur at night in fast-moving storms," in a region with "larger population densities and poorer visibility compared to other tornado-prone parts of the U.S." Its research "exposed the prevalence of inadequate sheltering and housing types leading to increased vulnerability."^[15] Trees hide the storm and then become its debris.
• Housing. Stephen Strader of Villanova, quoted by E&E News in 2020, put the mobile-home risk at "15 to 20 times more likely to be killed during a tornado" than in a permanent home; the same report counted 42 percent of US tornado deaths since 1995 in mobile homes.^[16] The Enderlin and Fort Pierce deaths of 2025 and 2024 were in exactly such communities.

The other alleys

Once one alley was allowed, others followed. A 2014 University of Akron classification reported in the press added a "Hoosier Alley" of Kentucky, Illinois, Indiana and Ohio and a "Carolina Alley" of the two Carolinas. Harold Brooks of NSSL, asked about them, said he "doesn't accept Carolina or Hoosier Alley as valid concepts," calling them "terms for broad regions" without a "strong seasonal cycle."^[18] None of the terms appears on any NOAA page. What the data supports, as the maps above show, is one continuous region of high frequency from the Plains to the Appalachians, with a Plains core that has the most reliable season and a Southeast core that has the most deaths.

Why it is moving

Three explanations are on the table, and the literature has not chosen among them.

Reporting. More people, the Doppler radar network of the 1990s, warning verification and storm chasing have all raised counts, and not evenly. But the Gensini and Brooks result stands against a purely observational explanation: the atmosphere's own tornado-favorable environments show the same eastward trend as the reports, and the Plains decline appears in both.^[11]

Natural variability. The record since 1950 is short against the timescales of ocean-driven climate cycles, and the two-period comparisons on this page cannot distinguish a trend from a slow oscillation. Agee's group declined to attribute the change; Gensini and Brooks called the cause "unclear."^[10][11]

A warming climate. SPC's FAQ is careful: climate models "can indicate broad-scale shifts in three of the four favorable ingredients for severe thunderstorms (moisture, instability and wind shear)," but "Our physical understanding indicates mixed signals--some ingredients may increase (instability), while others may decrease (shear), in a warmer world."^[2] The most detailed projection is Ashley, Haberlie and Gensini's 2023 study, which simulated supercell thunderstorms at 3.75 km resolution for 1990 to 2005 and for 2085 to 2100 under two emissions scenarios. "Supercells are projected to become more numerous in regions of the eastern United States, while decreasing in frequency in portions of the Great Plains. Supercell risk is expected to escalate outside of the traditional severe storm season, with supercells and their perils likely to increase in late winter and early spring months." Increases were found "east of 95°W" and decreases "from south Texas to South Dakota"; annual supercell counts over the eastern United States rose from 595 to 634 and 682 in the two scenarios; and in the mid-South the overnight supercell footprint "increases by roughly 70%." The authors wrote that the projected pattern "mimics a recent observed trend in tornado environments and reports."^[14]

Whatever the cause, the people who study it agree on the consequence. Agee to E&E News in 2020: "There's definitely been a spatial redistribution of tornadoes."^[16] Gensini and Brooks: the mid-South "already represents a maximum in the occurrence of casualties associated with tornadoes," and "the combination of an increase in risk and exposure could lead to a threefold increase in tornado disaster potential."^[11]

The last few seasons

The recent record reads as if written to illustrate the argument.

• 2011 remains the benchmark: 1,691 tornadoes in the final count, six EF5s, and 553 deaths, "the most in the 62-year period of record." The April 25 to 28 Super Outbreak produced 343 tornadoes and 321 deaths, 240 of them in Alabama.^[19][20]
• 2024 finished with 1,791 tornadoes, the third-most on record behind 2004's 1,813, with outbreaks across the Midwest and Plains in late April and May, 46 tornadoes in Florida from Hurricane Milton, and an EF3 that killed six in a mobile-home community near Fort Pierce.^[19]
• 2025 had 1,559 tornadoes, "127% of the 1991–2020 annual average" and the fifth-largest count, concentrated "across the southern Plains, central Gulf Coast states, and the Mid-Mississippi and Ohio valleys." March set a monthly record of 300. North Dakota broke its annual record with 72 and produced the first EF5 since 2013 at Enderlin. The deadliest single tornado, an EF4 through Somerset and London, Kentucky on May 16, killed 19.^[19]
• 2026 so far. Through July, NCEI counted 1,292 preliminary tornadoes, "about 138% of average," on course for the fourth-highest January to July total on record, and "Illinois continues to lead the United States in preliminary tornado counts from January through July (225), followed by Indiana and Mississippi (tied at 89), and Missouri (86)."^[19] Indiana's 77 by early July was already a state record.^[21] Oklahoma, whose annual average is 59, had 67 by the end of March.^[22]

None of this makes the Plains safe. Oklahoma's 152 tornadoes in 2024 were the most in its record, and the Greenfield and El Reno storms were as violent as any the database holds.^[22] The point of the record is narrower: the region where a tornado is most likely to kill someone is not the region the name points to, and the gap is widening.

Video

Methods

The charts on this page were computed from the Storm Prediction Center's tornado database file covering 1950 to 2025, downloaded in September 2026. National counts use the whole-track record of each tornado; state and regional counts use the state-segment records, so a tornado that crosses a state line is credited to each state it touched. Ratings are as recorded, with the Enhanced Fujita scale treated as continuous with the Fujita scale; tornadoes coded with an unknown rating, which begin in 2016, are counted in "all tornadoes" and excluded from "F1 and stronger." Night is 9 pm to 6 am Central Standard Time, the database's time zone. The three regions are the authors' groupings of whole states, chosen to follow the conventional alleys: Plains (Texas, Oklahoma, Kansas, Nebraska, South Dakota), Dixie (Arkansas, Louisiana, Mississippi, Alabama, Tennessee, Georgia) and Midwest (Missouri, Illinois, Indiana, Iowa, Kentucky, Ohio, Wisconsin, Michigan). State areas are computed from the Census Bureau's cartographic boundaries in an equal-area projection. The 1966 to 1995 and 1996 to 2025 periods were chosen as the two most recent thirty-year spans; the 1954 to 1983 and 1984 to 2013 split used by Agee and colleagues gives the same direction of change.^[6][7]

Related

The profiles on this site cover both alleys. From the Plains: Bridge Creek-Moore, Greensburg, Moore, El Reno, Jarrell and Greenfield. From Dixie Alley: Hackleburg, Smithville, Tuscaloosa and Mayfield. For how the warnings that cover both regions work, see Tornado watch vs. tornado warning.

All articles · Tornado profiles · Search the radar archive by date and time

Sources