The Repository Hook Echo

Mayfield 2021: The Tornado That Stayed Down 165 Miles

October 14, 2026 · Runtime 15:08 · Watch on YouTube

On the night of 10 December 2021 one tornado crossed western Kentucky for about 165 miles, in the dark, at about 60 miles an hour. At about the minute it reached Mayfield, the National Weather Service office in Paducah lost its power. Three radars kept recording, and we measured every scan: from 8:59 to 11:45 PM, at least one of them saw debris on the tornado's path in every scan. FEMA drone footage shows Mayfield and Dawson Springs days later.

Chapters

  1. 0:00 Mayfield, three days later
  2. 0:14 The office that lost its power
  3. 0:49 The track, 8:49 to 11:47 PM
  4. 1:15 A December day near 80 degrees
  5. 1:38 One supercell, a family of tornadoes
  6. 2:00 The first EF4, Arkansas to Tennessee
  7. 2:24 The warning that named Cayce
  8. 2:49 The Paducah radar
  9. 3:07 How a radar sees debris
  10. 3:56 Our measurement: 8:52 and 8:59
  11. 4:21 Mayfield, around 9:30
  12. 4:56 150 to 190 mph, from the radar
  13. 5:22 A town in the air
  14. 5:40 Paducah goes dark
  15. 6:49 The radar archive stops
  16. 7:13 Fort Campbell takes the watch
  17. 7:56 Benton and Kentucky Lake
  18. 8:50 Dawson Springs
  19. 9:09 Bremen
  20. 9:25 Louisville, and the end
  21. 10:05 Every scan, all night
  22. 10:32 What radar cannot prove
  23. 10:55 The survey on the ground
  24. 12:20 Where it ranks
  25. 13:04 Tri-State, 1925
  26. 13:26 57 lives
  27. 14:17 What the record supports
  28. 14:38 Next: Alaska, 1964

Transcript

Mayfield, three days later 0:00

This is Mayfield, Kentucky, three days after the night of December 10th, 2021. One tornado did this, in the dark. Then it kept going, on the ground, for about a hundred and sixty-five miles.

The office that lost its power 0:14

Almost nobody saw it. It was night, and it was moving at about sixty miles an hour. At about the minute it reached Mayfield, the Weather Service office that was watching it lost its power. But the radars kept recording. And in every scan, for almost three hours, they saw the same thing: debris, lifted off the ground.

So here is the question this episode answers. How do you know that a tornado nobody could see stayed on the ground for a hundred and sixty-five miles? And what does a radar actually see, when it sees a tornado at night?

The track, 8:49 to 11:47 PM 0:49

First, the map. The tornado began near Woodland Mills, in the northwest corner of Tennessee, at eight forty-nine in the evening, Central time. It crossed into Kentucky almost at once, and then crossed eleven Kentucky counties, from Fulton in the west to Grayson. It ended near Falls of Rough at eleven forty-seven. Mayfield, Benton, Princeton, Dawson Springs and Bremen are all on that line.

A December day near 80 degrees 1:15

December is not tornado season in Kentucky, and that day did not feel like December. The Weather Service office in Memphis describes an unseasonably warm air mass, with temperatures near eighty degrees that afternoon, under a strong wind field. At three in the afternoon a tornado watch went up. That was about six and a half hours before the tornado reached Mayfield.

One supercell, a family of tornadoes 1:38

The storm that made it formed in Arkansas, near Jonesboro. The Weather Service survey team describes it as one supercell that lasted more than eleven hours and travelled about six hundred miles, through four states. It got a nickname, the Quad-State supercell. But one storm is not one tornado. The supercell made a family of them.

The first EF4, Arkansas to Tennessee 2:00

The first long-track tornado of that family touched down at seven minutes past seven, near Bay, Arkansas. It crossed the Missouri bootheel and the Mississippi River, and lifted near Samburg, Tennessee, at eight thirty-six. Eighty point three miles, rated EF4. Thirteen minutes later, not far to the northeast, the tornado this episode is about began.

The warning that named Cayce 2:24

The warnings were already moving with it. At eight twenty-nine, the Paducah office warned of a confirmed large and extremely dangerous tornado over Tiptonville, Tennessee, moving northeast at fifty-five miles an hour. The warning projected its path forward, minute by minute. It said Cayce, Kentucky, around nine oh five. Cayce was hit at nine oh one, by the new tornado.

The Paducah radar 2:49

This is the Paducah radar, about seventy kilometers from where the tornado began. Like every Weather Service radar since about 2013, it measures two things at once: how much the target reflects, and how alike the targets are. The second one is the key to everything that follows.

How a radar sees debris 3:07

Raindrops are nearly all the same shape, so the radar's two pulses, one horizontal and one vertical, return almost the same signal. The correlation stays close to one. Shingles, insulation, branches and pieces of roofs are every shape at once. The correlation collapses. Where the correlation drops below zero point eight inside a strong echo, right at a tornado's rotation, the radar is looking at debris.

The radar also measures motion along its beam. Wind coming toward the radar is drawn in green, wind going away in red. When a patch of bright green sits right next to a patch of bright red, the air between them is turning. That pair is the rotation. The correlation tells you whether that rotation has picked anything up.

Our measurement: 8:52 and 8:59 3:56

We pulled every scan from that night out of NOAA's public archive and measured it ourselves. In each one we looked for that combination, correlation below zero point eight and an echo of at least forty decibels, within three kilometers of the debris center nearest the survey line. At eight fifty-two, three minutes after touchdown, the Paducah radar shows rotation and no debris. At eight fifty-nine, there it is.

Mayfield, around 9:30 4:21

At five minutes past nine, Paducah issued a new tornado warning for Graves County. The text called it a confirmed large and extremely dangerous tornado near Cayce and Clinton, moving northeast at fifty-five miles an hour. And it named the next town on the line, with a time: Mayfield, around nine thirty.

The tornado reached Mayfield at nine twenty-six. That same minute, Paducah upgraded the warning to a tornado emergency, the strongest wording the Weather Service has. Twenty-one minutes had passed since the first warning that named the town.

150 to 190 mph, from the radar 4:56

Forecasters were reading the same radar in real time. At nine twenty-six, the Storm Prediction Center told the Weather Service offices that the Graves County storm most likely had tornadic winds of a hundred and thirty-five to a hundred and eighty miles an hour, based on the last fifteen minutes of radar data. If the rotation held, they wrote, a hundred and fifty to a hundred and ninety. The survey would later rate the peak at a hundred and ninety.

A town in the air 5:22

Here is the scan that began at nine twenty-nine. The debris center sits right over Mayfield, thirty-seven kilometers from the radar, and the lowest beam is passing only about four hundred meters above it. Sixty-four separate radar cells inside it read as debris. That is a town in the air.

Paducah goes dark 5:40

Now the part that is easy to miss. At about twenty-five past nine, commercial power was lost at the Weather Service office in Paducah, and the backup generator failed. The Service Assessment records that radar data stopped reaching the office's main forecasting system. The forecasters kept working on laptops still receiving the raw radar feed.

At about nine fifty-two, the office in Springfield, Missouri, took over Paducah's warnings as full backup. It kept them until almost three in the morning. The tornado emergencies for Princeton, Dawson Springs and Bremen were issued from Springfield, more than four hundred kilometers away.

Read the warnings from that night in order, and they trace the track. Tornado emergency for Benton at nine forty-five. Princeton at ten seventeen. Dawson Springs and Saint Charles at ten thirty-one. Earlington and Anton at ten forty-four. Bremen at ten fifty-eight. The margins were thin. The Dawson Springs emergency went out at ten thirty-one. The tornado entered the town at ten thirty-two.

The radar archive stops 6:49

The Paducah radar's own archive goes on a little longer. Its last scan in the public archive began at nine fifty-eight. The next one is from one sixteen in the morning, more than three hours later. The record does not say why. It only shows where the Paducah radar's watch ends. Power at the Paducah office was not restored until two fifty-two in the morning.

Fort Campbell takes the watch 7:13

But another radar had been watching the same storm the whole time. This is the radar at Fort Campbell, on the Kentucky and Tennessee line, a hundred and twenty-three kilometers from Mayfield. At that distance, the curve of the Earth puts the center of its lowest beam about one point seven kilometers up. And it saw the debris over Mayfield too.

Put the two radars together. Paducah saw debris over Mayfield at about four hundred meters. Fort Campbell saw it at about seventeen hundred. Even the bottom edge of Fort Campbell's beam was more than half a kilometer up. So the debris was not just on the ground. It was being carried hundreds of meters into the air.

Benton and Kentucky Lake 7:56

From Mayfield the tornado kept its heading. It passed along the northwest edge of Benton from nine forty-five to nine fifty, and at nine fifty-seven it reached Kentucky Lake. It crossed the lake, and the last Paducah scan, at nine fifty-eight, still shows debris, near the lake.

At ten fourteen it entered Caldwell County. A neighborhood near the Princeton golf course was hit at ten eighteen, and the tornado passed closest to Princeton at about ten twenty-one. Fort Campbell's radar was now about sixty-five kilometers away, and every scan still showed debris.

Do the arithmetic on the survey's own numbers. A hundred and sixty-five point seven miles, from eight forty-nine to eleven forty-seven, is just under three hours. That is an average of about fifty-six miles an hour, the whole way. The warnings had said fifty-five, and later sixty.

Dawson Springs 8:50

At ten thirty-two, the tornado entered Dawson Springs, in eastern Hopkins County. This is the town from the air, four days later. Fort Campbell's radar was about sixty kilometers away here, the closest it came to the track all night. Fifteen people died in Hopkins County.

Bremen 9:09

At about ten fifty-five it crossed into Muhlenberg County, and at eleven it reached Bremen, a town of about three hundred people. Eleven of them died. The Fort Campbell scan that began at eleven shows sixty-nine debris cells, the most it recorded all night.

Louisville, and the end 9:25

Then the tornado moved into the Louisville office's area, through Ohio, Breckinridge and Grayson counties, rated EF3 on that last stretch. The Louisville radar picks it up here. The survey ends the track near Falls of Rough at eleven forty-seven. The Louisville scan that begins at eleven forty-five still shows debris, thirty-three cells.

And then, one more detail. The Louisville scan at eleven fifty, after the official end, still shows a little debris, eight cells. At eleven fifty-five, none. That is the debris still coming down, after the tornado itself was gone.

Every scan, all night 10:05

So here is the whole night, in one picture. Paducah, eleven scans from eight fifty-nine to its last, all eleven with debris. Fort Campbell, thirty-one scans to eleven forty-five, thirty with debris. Louisville, ten scans on the last stretch, all ten. Put together, from eight fifty-nine to eleven forty-five, there is no gap of more than about six minutes without a scan that saw debris on that line.

What radar cannot prove 10:32

And now the limit of all this. None of those beams touched the ground. They passed between about four hundred meters and two and a half kilometers above it. Debris in the air proves that a tornado is lifting debris. It does not prove that the tornado is touching the ground at every point below. The eleven fifty scan is the proof: debris, after the end.

The survey on the ground 10:55

The number one hundred and sixty-five comes from somewhere else. It comes from people walking the ground. On December 12th, Paducah asked for a quick response team, staffed from the offices in Omaha and Kansas City. The Service Assessment says the survey effort spanned days, and in some places weeks.

What the survey looked for was continuity: damage that does not stop. Along this track it found one, from Woodland Mills to Falls of Rough. Between the end of the first EF4 at Samburg and the start of this one, it found three short, weak tornadoes, each with its own track. That is what makes them a family, and not one tornado.

The survey rated the long tornado EF4, with peak winds of a hundred and ninety miles an hour, and up to one point one miles wide. The survey team itself called it a high-end EF4, and added that the rating might have been higher had it hit more wind-resistant structures. It was also moving fast, at about sixty miles an hour, so the strongest winds stayed over each building only briefly.

The survey has limits of its own, and the Service Assessment writes them down. One of its findings is that some damage surveys were done by Weather Service staff with little survey experience. The Paducah office also used before-and-after satellite imagery from NOAA and NASA. A track this long is too long to walk all at once.

Where it ranks 12:20

So where does it rank? The Paducah office lists it as the ninth longest tornado path ever recorded in the United States, and the Service Assessment calls it the longest tracked December tornado on record. The official record holder is still the Tri-State tornado of 1925, at two hundred and nineteen miles.

And it was not the only long track that night. A different supercell, farther south, produced a tornado that started in Dyer County, Tennessee, and travelled a hundred and twenty-two point seven miles into Kentucky, rated EF3. Two tornadoes of more than a hundred miles each, in one December night. The Service Assessment counts seventy-one tornadoes that night, across eight states.

Tri-State, 1925 13:04

But 1925 had no radar, and no modern survey. A 2013 reanalysis went back to the original damage reports and found gaps in that path. It concluded that about a hundred and seventy-four miles of it was likely continuous. The Mayfield track, mapped with modern tools, is continuous for a hundred and sixty-five.

57 lives 13:26

This tornado killed fifty-seven people. Twenty-four in Graves County, fifteen in Hopkins, eleven in Muhlenberg. More than five hundred were injured. Across the whole outbreak, eighty-nine people died, the deadliest December tornado event on record since 1950.

For the Paducah office, it was the worst night in its history. Fifty-seven deaths in its warning area, the highest number it has ever recorded. And for most of that night, its own building was dark, and its warnings were coming from Springfield.

The warnings did not fail. The Service Assessment found no EF3 or stronger tornado that night without a warning, and an average lead time of over nineteen minutes. Mayfield got twenty-one minutes, from an office that lost its power as the tornado arrived.

What the record supports 14:17

Here is what the record supports. For almost three hours, at night, three radars handed one storm from one to the next, and in every scan, at least one of them saw debris on the line. The radar shows that the tornado never let go of what it lifted. The survey, walked on the ground, shows that it never let go of the ground.

Next: Alaska, 1964 14:38

Next, a disaster measured by something no radar can see. At five thirty-six in the evening on March 27th, 1964, the largest earthquake ever recorded in the United States struck south-central Alaska, magnitude nine point two. Parts of the coast sank. Other parts rose by as much as thirty-eight feet. That is the next episode. Alaska, 1964: the earthquake that lifted a coastline out of the sea.

Description and sources

Our measurement of NOAA's Level II archive (correlation coefficient below 0.8 inside echoes of 40 dBZ or more, within 3 km of the debris centre): Paducah (KPAH) shows debris in all 11 scans from 8:59 PM until its archive stops at 9:58 PM; Fort Campbell (KHPX) in 30 of 31 scans to 11:45 PM; Louisville (KLVX) in 10 of 10 on the last stretch. Over Mayfield at 9:29 PM, KPAH counted 64 debris cells with its beam about 400 m up; KHPX saw the same debris from 123 km away, with its beam centred at 1.7 km. But a radar beam never touches the ground: the 165.7-mile continuous track comes from the NWS damage survey, which also counts the storm's earlier 80.3-mile EF4 as a separate tornado. The Mayfield tornado was rated a high-end EF4 (190 mph), killed 57 people, and is the longest tracked December tornado on record.

PRIMARY SOURCES

Hook Echo reads a disaster through the instruments that recorded it: radar, sensor, survey and report, with the primary sources above.

About the material. Video: FEMA drone and ground footage of Mayfield (13 and 17 December 2021) and Dawson Springs (14 December 2021), released through DVIDS, public domain. Every radar image is rendered by Hook Echo from NOAA's public archive of KPAH, KHPX and KLVX Level II data (NEXRAD, unidata-nexrad-level2 on AWS), in the National Weather Service colour scale; the correlation-coefficient scale is the channel's own, with the debris threshold marked. The debris counts are our own measurement from public data. Every diagram is drawn in code from the sources listed. Nothing in this video is AI-generated imagery.

#HookEcho #MayfieldTornado #Tornado