The Tri-State tornado of March 18, 1925

F5  Missouri, Illinois and Indiana · surveyed by the United States Weather Bureau

On a Wednesday afternoon in March 1925 a tornado formed in the Ozark hills of southeast Missouri and moved east-northeast at close to 60 mph for three and a half hours. It crossed the Mississippi, destroyed Gorham, cut through Murphysboro, De Soto, West Frankfort and Parrish, crossed the Wabash, levelled Griffin and struck Princeton before it died in the fields of Pike County, Indiana. It killed 695 people, more than any tornado before or since in the United States. No forecast preceded it, no warning was possible, no photograph of it exists, and a century later the question of whether it was one tornado or several is still not settled.

De Soto, Illinois the day after the tornado: a flat field of splintered lumber, stripped tree trunks and a line of cars on a road at left, with the wreckage of the town along the horizon.
De Soto, Illinois, March 19, 1925. The day after. Seventy-two people died in De Soto township, most of them in the town, which had 900 residents. Photo: US National Archives, via Wikimedia Commons, public domain.[4]
Rating
F5 assigned retrospectively; no scale existed in 1925
Path length
219 mi 1925 survey; 174 mi probable in the 2013 reanalysis
Width
½ to 1 mi wider in parts of Hamilton County
Start
~1:00 pm CST western Reynolds County, Missouri
Forward speed
59 mph average; 57, 59 and 68 by state
Duration
3½ h
Deaths
695 742 in the 1925 Weather Bureau count
Injuries
2,027 2,756 in the 1925 count
Damage
$16.5M 1925 USD; about $1.4B in 1997 terms
Warning
"About five minutes" after first seeing the cloud
Surveyed
7 days by car Illinois and Indiana only
Outbreak
7 tornadoes 792 deaths across the day

The setup

The storm system behind the deadliest tornado in American history was not, by the measures available, an unusual one. Robert Maddox and colleagues, who reconstructed the day in 2013 from the twice-daily Weather Bureau observations, kite and balloon soundings and ships' reports of 1925, found the surface low was "not unusually intense," with a central pressure near 998 hPa while the tornado was on the ground. The lows of the 1965 Palm Sunday outbreak and the 1974 Super Outbreak were about 15 hPa deeper.[3]

What the low did was move. It had come south-southeast from Montana to northern Oklahoma, then turned and accelerated northeast; at 7 am it was over extreme northeastern Oklahoma, by noon it was in the data void over southern Missouri, and by 4 pm it was over southern Illinois. The tornado formed near the triple point, where the low, the dryline and the warm front met, and then moved east-northeast at 59 mph, about 11 mph faster than the low and some 15 degrees to its right. Because the warm front advanced north as the storm and dryline raced east, the supercell "remained very near the warm front" throughout its life. "Apparently, this consistent time and space concatenation of the supercell, warm front, and dryline for more than three hours was extremely unusual."[3]

A 1925 Weather Bureau map of the central United States with isobars and temperature lines drawn in red and the tracks of seven tornadoes drawn as black arrows, the longest running from southeast Missouri across Illinois into Indiana.
The Weather Bureau's map of the day. The tornado tracks of March 18 as Alfred Henry published them in the Monthly Weather Review, with the Tri-State track marked (a), from "12:55 p" to "4:30 p", ahead of the low. A modern editor has added the legend. Maddox and colleagues found the 1925 isobars were drawn well enough, and that a widely reproduced 1966 reanalysis which put the tornado in the cold air behind the low was "seriously in error." Map: US Weather Bureau, via NOAA Photo Library and Wikimedia Commons, public domain.[3][5]

The tornado's parent supercell, by Maddox's estimate, lasted about seven hours and traveled more than 400 miles. The 2013 damage reanalysis concluded it was a classic supercell in Missouri that became a high-precipitation storm in Illinois and Indiana, and that the tornado rode "on the right side and near the back end of the storm," not the front, as a 1966 study had inferred from the hail and rain that followed it.[1][3] Henry, editing the Monthly Weather Review a month later, wrote that he could not recall "an authoritative record of the occurrence of a tornado in or very close to the center of a general cyclone up to the present case."[5]

No forecast, no warning

There was no tornado forecast because there were no tornado forecasts. "During the first half of the 20th century, the U.S. Weather Bureau did not try to forecast tornado events; in fact, the use of the word 'tornado' in forecasts was explicitly banned." The Bureau had no severe storm reporting procedure either; if a tornado struck within 25 miles of an office, the official in charge was to investigate and note it in the monthly record. Comprehensive surface observations were taken twice a day, at 7 am and 7 pm. The only Weather Bureau station near the path was Evansville, Indiana, 25 miles south of its end.[3] The forecast for the day, as the National Weather Service's Paducah office now summarizes it, called for "rains and strong shifting winds."[8]

The Weather Bureau's own surveyors, Clarence Root of Springfield and William Barron of Cairo, asked the country people how much warning they had. "It would seem that they had about five minutes' warning after first noting the cloud. Asked as to the length of time in which the destruction took place, opinions varied but most persons thought about two minutes." Their explanation of the death toll, written within weeks, has not been improved on: "In the first place the path was of great length and was wider than usual, thus embracing an unusual area. The tornado advanced at a rapid rate, with great destructive force. There are relatively few basements in this region and surprisingly few storm caves. Where could the people take refuge? Many did not realize the danger present, thinking it merely a severe thunderstorm. Some entered the houses to take shelter from the rain."[4]

"Back then, people had no way of knowing this was coming. There were no severe weather forecasts, no tornado watches, no warnings, no radio or television coverage, no social media, no sirens. All they could do was react to what they saw in front of them."

Christine Wielgos, warning coordination meteorologist, NWS Paducah, to the Southern Illinoisan, March 2025[10]

Chronology

The Weather Bureau reconstructed the tornado's timetable from the railroads: "By calling on the train dispatchers, either in person or through the agents, we ascertained the time the wires went out as the storm crossed the several railroads." Times below are 1925 Central Standard Time as Root and Barron gave them, with stopped clocks and railroad logs as their sources; they wrote that the times were "approximately" correct. The 2013 reanalysis warned against using the sparse clock times to estimate timing along shorter segments.[1][4]

Murphysboro

Murphysboro was a railroad and mining town of about 12,000, and the tornado took its center. Root and Barron's accounting a few weeks later: 234 dead, 800 injured, $10 million of the tornado's $16.5 million in losses, 1,200 dwellings damaged or destroyed. "The storm covered 152 city blocks, representing seventy per cent of the residence section and sixty per cent of the entire city. After the storm, fire consumed eleven blocks."[4] A resident's account written decades later described how the fires spread: "One house seemed to catch fire from another; there was nothing to stop the domino effect. By nightfall we were to see the whole town afire."[15]

The brick shell of a large school building with its roof and upper floors collapsed inward, surrounded by rubble and stripped trees.
Longfellow School, Murphysboro. "Several large schools were wrecked by the wind, resulting in considerable loss of life," the Weather Bureau wrote. The school site is now Longfellow Park, where the centenary marker was dedicated in 2025. Photo: via Wikimedia Commons, public domain (NOAA collection).[4][10]
A flattened industrial yard beside a street: brick wall fragments, boilers, a leaning smokestack and scattered timbers under a pale sky.
The Mobile & Ohio yards. The railroad shops had employed about 1,100 people and were wiped out. Photo: via Wikimedia Commons, public domain by age.[11]

A Mobile & Ohio engineer, Michael Kiley, told the St. Louis Post-Dispatch: "We were pulling into Murphysboro from the south at 2:35 p.m., when the tornado struck. Then we saw the terrific funnel-shaped cloud approached from the west and saw houses rumbling before it. The air was full of wreckage."[2] Relief came by rail: Pullman cars carrying doctors, and Red Cross tents for the homeless.[11] The town did not regain what it lost. "Murphysboro is a great small town," Susan Adorjan said at the centenary. "It's been through so much, but it keeps rebuilding."[10]

The remains of a brick church in Murphysboro: one tower and a section of wall standing, the roof gone, with debris in the street in front.
The Baptist church, Murphysboro. One of the photographs an engineering survey published in 1925. Photo: via Wikimedia Commons, public domain (NOAA collection).

De Soto, West Frankfort and Parrish

De Soto had 900 people. "From the wreck of the two-story brick school 38 children were taken out dead. In De Soto township, including the town and rural sections, there were 72 deaths, and 97 out of about 175 who were injured were taken to hospitals."[4] The National Weather Service today gives the school's toll as 33, a record for a single school in an American tornado; the 1925 report's 38 is the higher of the two figures, and neither document explains the other.[8]

West Frankfort was a mining town, and the timing was the difference between a disaster and a catastrophe. "The dead numbered 127 and the injured were estimated at 450 of whom 117 were in hospitals," with 925 dwellings damaged or destroyed and 3,000 people left homeless. The tornado cut the power to the Old Ben mine, and "the 800 men, 500 feet down in the mine, had to climb up the escapement. Had the storm occurred an hour or so later at least many of them would have been above ground." The mine's barograph recorded the pressure "then dropped suddenly 0.23 inch to 28.87 inches, recovering the 0.23 immediately," the only barograph trace ever made within a mile of the tornado.[4][5] Maddox later found the instrument was not well calibrated and its clock ran about fifteen minutes slow.[3]

Seven miles on, "after killing 24, injuring 21, and tearing up $112,000 worth of farm property in the intervening territory, the tornado struck the town of Parrish, where about 275 people resided. Here 30 houses out of 34 were broken up, and no buildings escaped except the school and the church. Two families were entirely wiped out." Twenty-two died.[4] The path from southwestern Franklin County to eastern White County, 61 miles, is the longest stretch of the whole track without a single gap in the damage.[1]

Griffin and Princeton

The tornado crossed the Wabash a few minutes after four, "three hours after its start in Missouri." Griffin, "a village of 388 population," lost 34 people in the town and 150 were injured; the school was badly damaged, "but school had been dismissed and most of the pupils had left." Across Posey County the dead numbered 58.[4] "Griffin was absolutely demolished," Princeton's mayor and historian Greg Wright said a century later. "There was nothing left."[12]

At Princeton the path curved north and the tornado took the south end of the city: "Two hundred homes were destroyed, and 100 more damaged. There were 20 deaths, 200 injured, and 1500 were homeless." The Heinz plant and the Southern Railway shops were destroyed. "The tornado was described as a blackness moving across the south end of the city, with the air full of timbers."[4][12] Wright believes the city was spared a far worse toll by an accident of commerce: a downtown prize drawing had emptied the south side. "If the folks wouldn't have left their homes to go downtown to be present to win in the drawings, we would've had hundreds, hundreds more killed."[12] A mile and a half past Princeton the path narrowed to a quarter of a mile, stayed "severe for seven miles," and faded out near Petersburg.[4]

What the Weather Bureau observed

The Weather Bureau of 1925 had a network of stations reporting twice a day, at 7 am and 7 pm, kites and pilot balloons at four sites in the region, and observers who wrote down what they saw. The Cairo observer's form for the day reads, in full on the subject: "Second thunderstorm--Only one peal of thunder was heard--at 2:50 pm. The storm was north of the station." Cairo was more than 50 miles south of the tornado at the time. At Evansville, 26 miles from Princeton, the barograph dipped slightly as the tornado passed and large hail fell on the southeast side of the city around 4:15.[3]

The storm's structure is therefore inferred from its damage. The 2013 reanalysis put it plainly: "With no radar information available, we cannot know for sure if the supercell changed from classic to HP structure along the damage path."[1] Henry, in 1925, was gloomier about the prospects: "It must also be reluctantly admitted that there is little hope that the actual conditions that initiate a tornado vortex will ever be experimentally observed."[5]

How long it was

The 219 miles is a 1925 figure. "The tornado under discussion covered 86 miles in Missouri, 92 in Illinois, and 41 in Indiana, or a total of 219 miles," Root and Barron wrote, having driven the Illinois and Indiana portions in seven days by automobile. "It can positively be stated that there was only one tornado in Illinois, and that it was continuous from Missouri to beyond Princeton, Indiana." The Missouri figure was not theirs: it came from the state's section director and "was not based on a driving survey, and little was known about what happened across large segments of the track--one of which was 64 km (40 mi) long."[1][4]

219 mi or 174 mi the Weather Bureau's 1925 path, against the segment the 2013 reanalysis found "likely continuous": from central Madison County, Missouri to Pike County, Indiana, with no gap over two miles Johns, Burgess, Doswell, Gilmore, Hart and Piltz 2013. The same study found 151 miles it was more confident of, and 235 miles of damage in all, from a family of at least three tornadoes.[1]

The doubt was there from the start, in a different form. In 1925 "a multi-mile gap with no damage was considered to be 'skipping', rather than an ending followed by a new tornado." Charles Doswell and Donald Burgess, writing in 1988, found the Tri-State's length so far outside the record that "we find it easier to believe that the Tri-state tornado is misclassified than to accept the observed trend as valid," and observed that surveyors "with less extensive data and without today's knowledge of tornadoes to guide the survey, might mistake a similar series of tornadoes for the track of a single tornado."[1][6]

Burgess and Doswell then spent years with Robert Johns and others finding out. Their 2013 paper mapped every damage point that could still be located, from county histories, newspapers, plat books and interviews with the last survivors, and extended the known damage west into Shannon County and east past Petersburg to 235 miles. Within that path they found 32 gaps of a mile or more. At both ends the evidence pointed to separate tornadoes: in Shannon County the damage ran in a different direction from the Reynolds County damage, and in Pike County an eyewitness saw "an 'arm coming down out of the clouds'" reach the ground before the gap ended. The gaps through the Ozarks could not be judged either way. From central Madison County east there were no gaps over two miles, "more strongly suggesting that the tornado was likely continuous for that path segment," and from central Bollinger County to the edge of Pike County the 151 miles "can be considered likely a continuous path."[1]

Their conclusion was a refusal to conclude: "our findings cannot determine whether there was just one continuous primary tornado or how many separate tornadoes occurred in association with the long-lived supercell." Whatever the answer, the primary tornado remains the longest on record, and the same supercell produced a further 20-mile tornado in Washington and Jackson counties, Indiana, 75 minutes after the main path ended. The whole family covered about 320 miles in five and a half hours.[1]

How many died

The accepted toll is 695, "generally is accepted as the deadliest single tornado in United States recorded history."[1] It is not the Weather Bureau's number. Root and Barron's report of April 1925 opens: "In its course it caused 742 deaths, the injury of 2,756 persons, and estimated property losses of $16,500,000. These figures create new records for American tornadoes." Henry's table in the Monthly Weather Review gives the same 742 dead and 2,771 injured, and 792 dead for all seven tornadoes of the day.[4][5] The 695 traces to a 1966 study; no document read for this page explains what was subtracted. The modern injury figure of 2,027 is lower than either 1925 count.[3][8]

By state, the 1925 report counted 11 dead in Missouri, 631 in Illinois and 100 in Indiana; centenary coverage in Indiana gives 70 or 71 for that state.[4][12] Fifteen thousand homes were destroyed along the path.[8][13] The reanalysis counted 48 damage points that were schools.[1]

"It strikes one as remarkable that a meteorological phenomenon can be 219 miles in length and yet less than one mile in width. Equally strange is the fact that a mile on either side of such complete devastation can be found perfectly normal conditions and nothing harmed."

Clarence Root and William Barron, US Weather Bureau, April 1925[4]

The rating

The F5 is retrospective. Fujita's scale dates from 1971; the National Weather Service lists the tornado as "F5 tornado with winds exceeding 200 mph" without saying on what basis, and neither 2013 paper assigns a rating or a wind speed, describing the tornado as "violent."[1][3][8] What the record contains is damage consistent with the top of the scale as it is now understood: "In some cases residences were carried from the foundations with scarcely a board left in the immediate vicinity," and a Bollinger County witness "commented on scoured ground, with grass and some top soil removed."[2][4] Doswell and Burgess's caution applies: "if a house is swept off its foundation, that is F5 damage by definition, with no implication of the windspeeds needed to accomplish the observed damage."[6]

No wind was measured. The nearest anemometers, at Cairo and Evansville, recorded a fastest mile of 66 mph that afternoon as the low passed to the north, which says something about the day and nothing about the tornado.[3] Popular accounts give the tornado winds of 300 mph; no source read for this page supports a figure.

Why nobody saw a funnel

The surveyors found something they had not expected. "Over the west half of Illinois, very few could be found who thought they saw anything like a funnel-shaped cloud. Farther east some thought they saw such a cloud, especially those out of the path of the storm to the north, but these people were not very definite. All agreed, however, that two clouds came together, one from the right and one from the left. One mentioned a 'fog' rolling toward him, and many described the cloud as a turbulent boiling mass. Others mentioned greenish or purplish colors."[4] Root's explanation in 1925 was that the cloud "was so close to the earth there was no room for the usual pendant portion. It might be likened to an inverted truncated cone."[4]

The modern explanation is the same observation with a different name. "When the first tornado started with the Tri-State supercell in Shannon County, MO, it was observed to be funnel-shaped and was not large. By the time the Tri-State supercell was entering Bollinger County, the primary tornado was already wedge-shaped; to many people it appeared as a big black cloud rolling along the ground. Since it was so large, many people observed a big black storm 'cloud', 'smoke', 'fog', or 'big black mass' that was 'rolling' or 'swirling' around on the ground as it approached. As a result, some people did not recognize it as a tornado." In places, including Griffin, it was a multiple-vortex tornado.[1] A Bollinger County boy of nine remembered "a big black cloud rolling along the ground"; a sixteen-year-old was caught in the open because he could not recognize its "black fog" form.[2] "The sounds were likened to the noise of several freight trains or a large number of airplanes in action."[4]

People in 1920s clothing picking through a pile of splintered lumber and household debris beside two large fallen tree trunks, with stripped trees behind.
West of Biehle, Missouri. The remains of the Joseph Blechle home in Perry County, where the tornado passed at about 2 pm and was already, in the words of the reanalysis, "wedge-shaped." Photo: Missouri Digital Heritage, via Wikimedia Commons, public domain.[1]

No photograph of the tornado is known to exist. The Murphysboro Daily Independent lamented it within the month: "Had Mr. Short had a camera, the picture of America's greatest cyclo-tornado could have been made."[2][8]

A century on

The centenary was marked in March 2025 across the three states. Jackson County, Illinois held six days of events: a screening at the Liberty Theater, a day of National Weather Service presentations at Murphysboro High School, cemetery walks at De Soto and Tower Grove, new historical markers in Gorham and De Soto, and on the anniversary itself a marker dedication at Longfellow Park, where the school stood.[9][10] The Jackson County Historical Society's exhibit includes a child's shoes and a tree trunk with a pine board driven through it. "You know the numbers: 200 mph winds. It was a mile wide," Mary Riseling, who coordinated the remembrance, told the Associated Press. "But the force that it took to put that pine board into that maple tree, it really puts it all in perspective."[11][13][14]

In Princeton, Indiana a state historical marker was unveiled at West Garfield and South Madison avenues. Sylvia Carvell, 81, told the AP her mother, a survivor, "couldn't tell which direction was home."[11][12] The tornado's most-read account, Wallace Akin's "The Forgotten Storm" (2002), was written by a Murphysboro survivor who was two years old when the tornado carried him and his house away.[16]

"We just want them to take away what occurred 100 years ago - how amazing it is that Murphysboro is here today after something that devastating occurred."

Scott Johnson, Jackson County Historical Society, to WSIU, March 2025[13]

Videos

Documentary treatments, all via YouTube's privacy-enhanced embeds. No footage of the tornado exists.

100 Years. WFCN's full-length centenary documentary.
The meteorology. Broadcast meteorologist Alan Sealls on the storm.
American Nightmare. Carly Anna WX's documentary treatment.
Case study. MeteoTechWx on the setup and the reanalysis.

The Hackleburg-Phil Campbell tornado of 2011 had its 132-mile survey path cut to 102 by a reanalysis using the same reasoning about gaps. The Xenia tornado of 1974 was the deadliest of the outbreak this one is measured against. For a tornado that killed people in a town at a rate not seen since, see Joplin.

All tornado profiles · Search the radar archive by date and time · How to read reflectivity and velocity

Sources

Path length, deaths and per-town figures in the text follow the Weather Bureau's 1925 survey unless attributed to the 2013 reanalysis; the vitals table gives both where they differ. Times are 1925 Central Standard Time as the survey recorded them.

  1. R. H. Johns, D. W. Burgess, C. A. Doswell III, M. S. Gilmore, J. A. Hart and S. F. Piltz, The 1925 Tri-State Tornado Damage Path and Associated Storm System, Electronic Journal of Severe Storms Meteorology 8(2), 2013.
  2. Johns et al., Supplemental material to the above: county-by-county damage points, eyewitness interviews and newspaper extracts.
  3. R. A. Maddox, M. S. Gilmore, C. A. Doswell III, R. H. Johns, C. A. Crisp, D. W. Burgess, J. A. Hart and S. F. Piltz, Meteorological Analyses of the Tri-State Tornado Event of March 1925, Electronic Journal of Severe Storms Meteorology 8(1), 2013.
  4. C. J. Root and W. E. Barron, The Tri-State Tornado of March 18, 1925, Climatological Data, Illinois Section, US Weather Bureau, March 1925 (scanned PDF).
  5. A. J. Henry, The Tornadoes of March 18, 1925, Monthly Weather Review 53(4), April 1925.
  6. C. A. Doswell III and D. W. Burgess, On Some Issues of United States Tornado Climatology, Monthly Weather Review 116(2), 1988.
  7. H. E. Brooks and C. A. Doswell III, Normalized Damage from Major Tornadoes in the United States: 1890–1999, Weather and Forecasting 16(1), 2001. The wealth-adjusted figure of $1.39 billion is in 1997 dollars.
  8. NWS Paducah, The 1925 Tri-State Tornado. The office's former statistics, track and photograph subpages no longer resolve.
  9. NWS Paducah, Tri-State Tornado 100th Anniversary Event, March 15, 2025.
  10. P. Wilcoxen, Jackson County marks 100 years since Tri-State Tornado, The Southern Illinoisan, March 13, 2025.
  11. J. O'Connor, Associated Press, 100 years ago this week, nearly 700 people were killed in the deadliest U.S. tornado, PBS NewsHour, March 17, 2025.
  12. J. Buchanan, Great Tri-State Tornado hit 100 years ago, WFIE 14News, March 18, 2025; and J. Lyman and J. Yaney, New historical marker set to be unveiled on 100-year anniversary of Tri-State Tornado, WFIE, March 7, 2025.
  13. WSIU, Tri-State Tornado 100 year anniversary, March 12, 2025, and Tri-State Tornado commemoration, March 13, 2025.
  14. T. Chaney, Jackson County Historical Society hosts events to commemorate 100-year anniversary of Tri-State Tornado, KFVS12, March 2025.
  15. W. T. Kimmel, The 1925 tornado, a resident's account (1992), Jackson County ILGenWeb.
  16. B. Dowd, The Forgotten Storm by Wallace Akin, book review, History Nerds United, July 28, 2020; and Wikipedia, 1925 Tri-State tornado, consulted September 2026, used as a finding aid.

Where sources disagree on a figure, the difference is noted in the text. Corrections: contact@weatherovertime.com.