The Repository Hook Echo

Hurricane Michael 2018: The Data Made It a Category 5

October 9, 2026 · Runtime 15:28 · Watch on YouTube

Hurricane Michael came ashore at Mexico Beach, Florida, on 10 October 2018, announced as a Category 4. Six months later the National Hurricane Center made it a Category 5 -- by three knots. FEMA footage shows what it left. We go through every instrument the NHC weighed, and read the raw record of the USGS sensor strapped to the Mexico Beach pier, which went under water at 17:20 UTC.

Transcript

Mexico Beach, five days later 0:00

This is Mexico Beach, Florida, five days after Hurricane Michael came ashore. The town had one thousand six hundred and ninety-two buildings. One thousand five hundred and eighty-four were damaged. Eight hundred and nine were destroyed.

From Category 4 to 5 0:16

This is the storm the day it arrived, from an Air Force Reserve aircraft flying through the eye. When Michael made landfall, the National Hurricane Center put its winds at one hundred and fifty-five miles an hour. Category 4. Six months later, after every instrument had been reviewed, the number became one hundred and sixty. Category 5.

So here is the question this episode answers. How do you measure the wind of a hurricane that destroys the instruments put there to measure it? And how do half a dozen numbers that disagree become one number on the record?

What the categories mean 0:52

First, what the categories mean. The Saffir-Simpson scale rates a hurricane by one thing only: its strongest one-minute sustained wind. Category 4 runs from one hundred and thirteen to one hundred and thirty-six knots. Category 5 is one hundred and thirty-seven knots and up, a hundred and fifty-seven miles an hour. The scale says nothing about storm surge or rain.

Thirty knots to 140 1:16

Michael started as a tropical depression early on October 7th, about a hundred and thirty nautical miles south of Cozumel, Mexico. Its winds were thirty knots. Three and a half days later, they were one hundred and forty. In between, it strengthened against wind shear that should have held it back.

The final table in the report shows how fast. At midday on October 9th, ninety knots and nine hundred and sixty-eight millibars. At 18:00 UTC that day, one hundred knots. Twenty-four hours later, one hundred and thirty-five knots, and the pressure had fallen forty-one millibars. It was still strengthening when it reached the coast.

Michael's numbers were revised more than once. On October 8th it passed the western tip of Cuba. A station at Cabo de San Antonio measured ten-minute winds of eighty-four knots as the eyewall went over, but that data did not reach forecasters in real time. When it did, Michael was reassessed as a Category 2 at that point.

The forecasts 2:20

The forecasts got the track right. The official track errors were smaller than the five-year average at every forecast time. The intensity was another story. The hurricane kept strengthening faster than forecast, and the report calls it a large low bias. Even so, the official forecasts beat almost every computer model.

The hurricane warning for the landfall area went out about thirty-six hours before tropical-storm-force winds arrived. That was the warning. The number on the warning, how strong the wind would be at the coast, was the part nobody could pin down, then or later.

The hurricane hunters 2:58

So the aircraft went in. The Air Force Reserve's 53rd Weather Reconnaissance Squadron flew nine missions into Michael, and NOAA flew six. Between them they fixed the centre forty-five times. This is the eyewall from one of those planes: the wall of cloud around the calm centre.

Each plane carries three kinds of wind measurement. The wind at the plane's own altitude. A microwave radiometer that looks down at the foam on the sea and turns its brightness into a surface wind. And dropsondes, instrument packages dropped on parachutes that report as they fall.

On one flight, NOAA tried something new. Early on October 10th, its aircraft released a small drone called a Coyote into the eye and the eyewall. At about two thousand feet, it measured a ten-second wind of one hundred and fifty-nine knots. Converted to the surface, about one hundred and twenty. That was hours before landfall, while the storm was still intensifying.

Landfall on radar 3:58

At about 12:30 in the afternoon, Central time, on October 10th, the eye came ashore near Tyndall Air Force Base, southeast of Panama City. This is the radar at Eglin Air Force Base, about fifty miles to the northwest, watching the eye cross the coast. The National Hurricane Center used this same radar to check the planes.

Number 1: the aircraft 4:19

First number. At 17:23 UTC, seven minutes before landfall, an Air Force plane at about eight thousand feet measured one hundred and fifty-two knots in the southeast eyewall. That is the wind up there. The standard conversion to the surface gives one hundred and thirty-seven knots. Category 5 starts at one hundred and thirty-seven.

Number 2: the radiometer 4:43

Second number. The radiometer reported one hundred and thirty-eight knots at the surface at 17:06. But during that pass through the eyewall, part of its data never reached the ground in real time. Afterwards, NOAA rebuilt the missing seconds from the instrument's raw readings. They gave one hundred and fifty-two knots.

That second number came with a warning attached. The water under the plane there was only about eighty-nine feet deep, where waves start to break and the foam can mislead. And after Irma, Jose and Maria in 2017, there were signs that this radiometer reads too high at exactly these speeds. The research on that was still underway.

Number 3: the radar 5:25

Third number: the radar. Analysts took Eglin's lowest scan and averaged four range bins together, about the distance a plane covers in its ten-second sample. At 17:22, at the same place and height where the plane measured one hundred and fifty-two knots, the radar gave one hundred and fifty-five. And it showed stronger winds just northeast of the plane's path.

Brought down to the surface, the radar gave one hundred and forty-nine knots at 17:25. Averaged over the four scans before landfall, one hundred and forty-four. Small whirls inside the eyewall, called mesovortices, carried one hundred and eighty to two hundred knots aloft as they passed. The radar said Michael was still strengthening right up to the coast.

Not every radar analysis agreed. The Tallahassee radar's eyewall data were limited by communication problems. And an automated method run on Eglin's data suggested about one hundred and sixty knots at three kilometres, which converts to one hundred and thirty-five to one hundred and forty at the surface. The report doubts it: the method had trouble finding the centre of the storm.

One more thing about that radar. We read Eglin's position from the scan files and measured the distance to where the eye came ashore: forty-six miles. At that range, the centre of its lowest beam, tilted half a degree, was about a kilometre up, roughly thirty-four hundred feet. Like the plane, the radar was measuring the wind above the surface, and its numbers also had to be brought down.

Number 4: the ground 6:58

Fourth number: the instruments on the ground. All three of them reported sustained winds below one hundred knots. The station at Tyndall gusted to one hundred and twenty-one knots, and was inside the ring of strongest wind when it last reported. A tower of the Florida Coastal Monitoring Program, a portable mast set up on the base to measure the landfall, recorded a gust of one hundred and twelve, and was knocked down.

A second tower, just southeast of Mexico Beach, survived with a complete record. Its highest gust was one hundred and ten knots. The radar put it just outside the ring of strongest wind. None of the ground instruments sat where the worst wind was, and the report notes that this is typical at landfall.

Number 5: pressure 7:42

Fifth number: pressure. It is easier to measure, because the barometer does not have to be in the worst wind, only in the eye. A dropsonde measured nine hundred and twenty-two millibars. The tower on the base, nine hundred and twenty point two, minutes before it fell. The final assessment is nine hundred and nineteen.

That is the third lowest pressure for any hurricane landfall in the United States since 1900. Only the Labor Day Hurricane of 1935, at eight hundred and ninety-two, and Camille in 1969, at nine hundred, were lower. Three formulas that turn pressure into wind gave one hundred and thirty-nine to one hundred and forty-two knots.

Number 6: satellites 8:25

Sixth: the satellites. The classic method, reading cloud patterns, gave one hundred and forty knots. The automated version, one hundred and forty to one hundred and forty-five. Microwave techniques came in lower, at one hundred and ten to one hundred and thirty-five. A consensus product that weighs them all landed just above one hundred and forty.

The sensor on the pier 8:48

There is one more instrument, and it is the one we measured ourselves. On October 8th, the U.S. Geological Survey strapped two pressure sensors to the pier at Mexico Beach. One to measure the water. One to measure the air.

We downloaded the raw air readings, one every thirty seconds. At 16:00 UTC, nine hundred and seventy-seven millibars. Then the pressure falls, faster and faster. At 17:14, nine hundred and twenty-nine point seven. That is the lowest air pressure this sensor recorded.

Six minutes later, between one reading and the next, the number jumps from nine hundred and thirty-seven to one thousand and ninety-eight. That is not the air. That is the weight of seawater on top of the sensor. At 17:20, the barometer on the pier went under.

Nine hundred and twenty-nine point seven appears in the hurricane centre's report too, credited to a USGS pressure sensor in Mexico Beach, in the eastern eyewall. It is not the lowest pressure of the storm, because the pier was not in the eye. It was in the wall of the strongest wind, which is exactly why it went under.

15.55 feet 10:02

The water sensor kept counting. At 16:00 UTC, the water stood at six point three feet above the reference datum. At 17:39, it peaked at fifteen point five five. Nine point two feet in ninety-nine minutes. Our number matches the one in the hurricane centre's report to the hundredth of a foot.

When the survey team came back on October 15th, the sensor and its mount were gone. Four days later it turned up in the debris, somewhere else, with its data intact. Its logger had kept recording, every thirty seconds, until October 16th. Fifteen and a half feet on that datum means as much as fourteen feet of water over ground that is normally dry. Along this stretch of coast, from just southeast of Tyndall to Port St. Joe, the report estimates nine to fourteen feet.

The surveyors checked it another way. Inside a vacation rental building on the north side of Highway 98, the storm left a line of seeds on the wall, where the water stopped. It was eleven point six feet above the base of the building, which itself sits a few feet above the lagoon behind it.

All the numbers on one table 11:11

So here is everything on one table. The plane, converted: one hundred and thirty-seven. The radiometer: one hundred and thirty-eight, or one hundred and fifty-two if the rebuilt reading is right. The radar: one hundred and forty-four to one hundred and forty-nine. The ground: under one hundred. Pressure and satellites: about one hundred and forty.

The number that goes into the permanent record is called the best track. It is not written during the storm. It is written months later, from a review of the aircraft winds, surface winds, pressures, satellite estimates and radar velocities, including data and analyses that were not available in real time.

The hurricane centre's final answer is one hundred and forty knots, one hundred and sixty miles an hour. Category 5, by three knots. No single instrument measured exactly that. It is a judgement made from all of them, and the report says best-track winds carry an uncertainty of about ten per cent.

What the report admits 12:13

And the report says something most headlines left out. The five-knot increase is small, and well within the normal uncertainty. The Category 5 winds probably covered only a very small area at the coast. And in terms of what happened on the ground, the change is of little practical significance. It also leaves the door open: the landfall number could be revised again, once the research on the radiometer's accuracy at these speeds is finished.

On the record, Michael is now one of only four Category 5 hurricanes to make landfall on the mainland United States since 1900, after the Labor Day storm of 1935, Camille and Andrew. It is the strongest landfall on record in the Florida Panhandle, and the latest in the year that any Category 5 has come ashore.

Tyndall and Bay County 13:00

At Tyndall Air Force Base, the report says every building was damaged, and many were destroyed. About three billion of Michael's twenty-five billion dollars in damage was on the base. This is base housing, filmed from a drone four days after landfall.

Across Bay County, more than forty-five thousand structures were damaged and more than fifteen hundred destroyed, including severe damage to two hospitals. Next door, in Gulf County, nine hundred and eighty-five structures were destroyed. Coastal highways washed out in several places between Panama City and Alligator Point.

On the coast, the water did what the wind could not. Five people drowned in the surge in Florida, three of them in or near Mexico Beach. All five were in storm surge evacuation zones. In all, the report counts sixteen direct deaths in the United States, and forty-three indirect deaths in Florida, from falls during the clean-up, traffic accidents and medical problems made worse by the storm.

Michael did not stop at the coast. It crossed into southwest Georgia still a hurricane. In Jackson County, Florida, around Marianna, four hundred buildings were destroyed. In Seminole County, Georgia, around the town of Donalsonville, ninety-nine per cent of the homes were damaged, and a station there recorded a gust of one hundred knots.

What the record supports 14:24

Here is what the record supports. The strongest wind of Michael was never measured directly at the surface. It was estimated, from a plane above it, a radar fifty miles away, a pressure reading and a satellite. The instruments that were in the right place, on the ground, were knocked down, flooded, or carried away.

And the measurement closest to what actually killed people is not the wind number at all. Five of the seven direct deaths in Florida were drownings in the surge. The number for that is fifteen and a half feet, on a pier piling in Mexico Beach, recorded by a sensor that was found in the debris.

Next: Joplin, 2011 15:02

Next, back to tornadoes. On May 22nd, 2011, the warning for Joplin, Missouri, went out seventeen minutes before the tornado touched down. The federal assessment found that most people did not take shelter at the first warning. They looked for confirmation first. That is the next episode. Joplin, 2011: the warning that came seventeen minutes early.

Description and sources

No instrument measured 140 knots at the surface. The Air Force aircraft at about 8,000 ft measured 152 kt, 137 kt when converted to the surface; the radiometer reported 138 kt, and 152 kt after its missing data were rebuilt, with a possible high bias; the Eglin radar gave 155 kt where the aircraft gave 152; the ground stations reported sustained winds below 100 kt, and one tower was knocked down. The pressure, 919 mb, is the third lowest at a US landfall since 1900. The NHC settled on 140 kt (160 mph) and wrote that the 5-knot increase is well within the normal uncertainty. On the pier, the barometer fell to 929.7 mb at 17:14 UTC and went under at 17:20; the water rose 9.2 ft in 99 minutes, to 15.55 ft NAVD88.

PRIMARY SOURCES

Beven, J. L., Berg, R. & Hagen, A. (2019). National Hurricane Center Tropical Cyclone Report: Hurricane Michael (AL142018). https://www.nhc.noaa.gov/data/tcr/AL142018_Michael.pdf
USGS Short-Term Network (STN), event 287 (2018 Michael), site FLBAY03283, Mexico Beach Pier: https://stn.wim.usgs.gov/STNServices/Sites/3283/Files.json
NOAA NWS NEXRAD Level II, KEVX, 10 October 2018: https://unidata-nexrad-level2.s3.amazonaws.com/?prefix=2018/10/10/KEVX/
NHC, Saffir-Simpson Hurricane Wind Scale: https://www.nhc.noaa.gov/aboutsshws.php

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, U.S. Air Force Reserve (403rd Wing) and U.S. Air Force footage released through DVIDS, public domain. Every radar image is rendered by Hook Echo from NOAA's public archive of KEVX Level II data (NEXRAD, unidata-nexrad-level2 on AWS), reflectivity only, in the National Weather Service colour scale. The pier pressure and water-level series and the radar beam height are our own measurements from public data (USGS Short-Term Network, event 287, site FLBAY03283; KEVX volume headers). Every diagram is drawn in code from the sources listed. Nothing in this video is AI-generated imagery.

#HookEcho #HurricaneMichael #Hurricane