The Fitzgerald Storm

The Edmund Fitzgerald finished loading 26,116 long tons of taconite pellets at the Burlington Northern dock in Superior, Wisconsin, at a quarter past two on Sunday afternoon, 9 November, and left at once for Detroit at her full speed of 99 revolutions, about 16 miles an hour. The Arthur M. Anderson left Two Harbors two hours later with a similar cargo for Gary, Indiana, making 14.6 miles an hour, and the two ran east ten to twenty miles apart. The wind on the western lake was east-north-east at 10 to 15 knots.

Gale warnings for the whole lake went out at seven that evening. By then the low was over east-central Iowa at 993 millibars and moving at 37 knots. At two in the morning the warnings were raised to storm warnings, with north-east winds of 35 to 50 knots to come round to the north-west. The two masters talked by radiotelephone soon afterwards and agreed to leave the recommended lanes along the Michigan shore for the Canadian side, where a north-east wind leaves a lee. That is the long dog-leg on the map: north-east between Isle Royale and the Keweenaw, east along the Ontario shore, then south-east for Whitefish Bay.

Sunday night

The Fitzgerald's one o'clock report, about twenty miles south of Isle Royale, gave wind from 030 at 52 knots, heavy rain and ten-foot seas. The reanalysis has 29 knots there, gusting 45, with a two-metre sea. Hultquist, Dutter and Schwab's model had 30 to 35 knots and could not reproduce the 52, and two other ships on the western lake reported 32 and 38 at the time; they suggest the figure may be a ship-relative wind reported as true, which for a ship steaming north-east into a north-easter adds her own speed to the reading. The seven o'clock report, 35 miles north of Copper Harbor, gave 050 at 35 knots with the same ten-foot sea; the reanalysis there is 17 to 23 knots from the north-east. She told the company her arrival at the Soo was indefinite because of weather, and that was her last routine report.

The track between those two reports is the least certain part of the map. The reports give a bearing and a distance from a landmark; the reanalysis is sampled along a straight line between them. The ship made good a little over six knots over that stretch, which for a loaded laker punching into a north-east gale is believable, and it is also what a position report rounded to the nearest five miles will do to an average.

Monday

The low's centre crossed the lake at midday, west of Michipicoten Island, and the wind went round through the south. At one o'clock the Anderson, twenty miles north-west of the island and near the centre, reported south-south-east at 20 knots with a twelve-foot sea still running from the night. Fifteen miles to her south-west the Simcoe already had west at 44. The Board describes a wind-shift line, running north-north-east to south-south-west and moving east at 20 to 25 knots, behind which the wind fell below ten knots out of the south and then rose quickly to 35 to 50 from the north-west. At 1:50 the Anderson logged five knots. At 2:45 she logged 42.

The Fitzgerald passed west of Michipicoten's West End light about ten minutes to two, said she would "continue on" though "rolling some", and steered 141 for Whitefish Point. The Anderson hauled west first, for sea room, and followed. At half past three the Fitzgerald, north-east of Caribou Island, reported a fence rail down, two vents lost or damaged and a list, and said she would check down so the Anderson could close. Asked whether her pumps were going: "Yes, both of them." Forty minutes later she reported both radars out and asked the Anderson to keep track of her. At 4:39 Coast Guard Grand Marais told her the Whitefish Point radio beacon was off the air. At 4:52 the Anderson, abeam the north tip of Caribou at six miles, logged 58 knots from 304, the highest of her voyage, in snow, with twelve- to eighteen-foot seas.

Between five and half past, the master spoke to the pilot of the northbound Avafors off Whitefish Point: a bad list, both radars lost, heavy seas over the deck, "one of the worst seas I've ever been in". At seven the Anderson reported north-west 50 knots and sixteen-foot waves; Stannard Rock, sixty miles to windward, had 40 gusting 65. At 7:10 she warned the Fitzgerald of a northbound ship nine miles ahead and asked how she was making out. "We are holding our own." When the Anderson's radar was checked again about 7:20 the target was gone. Shore lights twenty miles off and a ship nineteen miles off were visible between the snow squalls; the Fitzgerald, which should have been about ten miles ahead, was not.

She lies in 530 feet of water at 46°59.9′N, 85°06.6′W, seventeen miles from the entrance to Whitefish Bay, in two pieces.

Reading the map

The clock runs at forty minutes of voyage to the second. The wind under the ships is the reanalysis, interpolated between its hourly fields; the arrows point downwind and lengthen with speed. Watch two things. The first is the lee: for the whole of Sunday night the wind over the Ontario shore is lighter than over the open lake to the south, which is the shelter both masters went north for. The second is the shift, which arrives over Caribou Island at about two in the afternoon as a line, with the lake behind it turning dark in an hour. The Fitzgerald turned south-east for Whitefish Point into the two hours in which that line crossed her.

The low that crossed Lake Superior on 10 November was an ordinary autumn cyclone in its birth and an unusual one in its deepening. It formed over the Oklahoma panhandle on Saturday, was a closed circulation over south-central Kansas by seven on Sunday morning at 1000 millibars, and from there ran north-east at close to 30 knots, losing 22 millibars in thirty-six hours. The Weather Service's first forecast had the centre passing south of the lake; by Sunday afternoon the track had been moved north, over the lake east of Michipicoten Island and on to James Bay, and that is where it went.

The reanalysis against the record

ERA5 carries the storm as the Marine Board recorded it. The table sets the Board's centre positions and pressures beside the reanalysis minimum for the same hour.

The two disagree by four millibars at most, and the reanalysis is the deeper of the two from Wisconsin onward. Its lowest reading is 975.5 millibars at four on Monday afternoon, over the bush north of the lake near 50°N 83°W. Hultquist, Dutter and Schwab's mesoscale run, started from the coarser reanalysis of 2006, reproduced the same pressures to within two millibars, so three independent reconstructions agree on the storm's depth.

The track matters more than the depth. The centre crossed the eastern lake at midday Monday, which put both ships inside the circulation for most of the day: ahead of the centre on Sunday night in a north-east gale, through the centre around one on Monday afternoon in a lull, and behind it in the north-wester for the rest of the day. That order is what the map shows if the clock is run from Sunday noon. The isobars are drawn every four millibars; where they bunch over the eastern lake on Monday afternoon the gradient is about four millibars in fifty miles, and a gradient like that over cold water under cold air is what makes a 45-knot wind.

What the pressure trace says

The trace under the map is the centre's pressure by the hour. It falls steadily through Sunday, steepens overnight as the low crosses Wisconsin, and bottoms out between two and eight on Monday evening. The Fitzgerald was lost at a quarter past seven, inside that window. The ships' barometers tell the same story from underneath: the Anderson's read 28.84 inches, 977 millibars, before noon on Monday, having "dropped rapidly", and that was two hours before the centre was over her.

The dots on the pressure trace are the Board's own figures, and the rings on the map are where it placed the centre. The forecasters of 1975 had the position and depth of this storm right at every synoptic hour, from surface observations alone. What they did not have was the lake itself, which the next four sections are about.

The chart puts the reanalysis wind at the Fitzgerald's position for every hour of her voyage, and after her loss at the wreck site, on one axis with everything that was measured near her. Three things sit on that axis: the reports the two ships made, the readings from Stannard Rock and Whitefish Point that the Board recorded, and the hourly observations of a shore airport the reader can choose.

The shape of it

Sunday evening the wind over the ship builds from 10 to 28 knots out of the north-east by midnight, and the reanalysis gust climbs from 15 to 45. That is the gale the warning named. Through the small hours it holds near 30 knots and then eases as the centre approaches, down to 17 at seven and under ten by eight, while the ship is running the Ontario shore. Between eight and one on Monday the wind at the Fitzgerald is under 14 knots and swinging from north-east through south to west: the centre of the low is passing over her.

The shift arrives between one and two in the afternoon. The reanalysis wind at the ship goes from 20 knots at one to 38 at two and 44 at four, from 290 to 300 degrees, and the hourly gust from 30 to 66. It stays between 40 and 44 knots, gusting 61 to 67, until she is lost, and at the wreck site it holds above 36 knots until ten that night.

Why the ship reads higher

The reanalysis is a 31-kilometre model. Its wind is the average over a cell nineteen miles across and over an hour, and its gust is the model's estimate of the strongest three-second wind in that hour, not a measurement. A ship's anemometer, forty metres over the deck of a loaded laker, reads the real thing at one point and at a height where the wind is some ten per cent stronger than at ten metres. Hultquist, Dutter and Schwab's five-kilometre run found the same thing from the other direction: sustained winds near 50 knots over the south-east lake, local cores above 60, and a maximum of 65 in their one-kilometre nest, where ERA5 has 44. The right reading of the chart is that the reanalysis is the floor of what the Fitzgerald met, and the gust line is nearer the sustained wind on deck than the mean line is.

Two readings on the chart deserve a note. The Fitzgerald's own 52 knots at one in the morning is 20 above the reanalysis and 15 above the ships around her, and the modern re-examination treats it as probably a relative wind. And the Anderson's 58 at 4:52 is printed as 52 in the Safety Board's report; the Marine Board's text, which quotes the log, has 58 "from 304°T, the highest winds recorded during the voyage", and that is the figure used here.

The shore felt a different storm

Switch the station. Thunder Bay, Duluth, Geraldton, White River and Wawa never pass 24 knots in their hourly observations, with gusts to the low thirties; the airports sit in the lee of hills and forest and measure the wind of the land. Sault Ste. Marie is the exception, because the north-wester blows down Whitefish Bay to it: 38 knots at eight that evening with a gust of 59, the strongest gust any shore station in the archive recorded. Marquette, the nearest station to the wreck, has four observations in the window and none on Monday. The storm that mattered was over the water, and in 1975 nothing on the water measured it continuously. The lake's first weather buoy was moored in 1979.

Wind makes waves in proportion to how long and how far it has blown over open water. The distance is the fetch, and on a lake the fetch is set by the shore upwind. That is the whole reason the two masters went north on Sunday night: in a north-east wind the Ontario shore lies a few miles to windward of the Canadian lanes, and a few miles of fetch cannot raise much of a sea however hard it blows. It is also the reason Monday afternoon was different. When the wind came north-west, the water upwind of a ship off Caribou Island ran all the way back to the Keweenaw and beyond, a hundred and thirty miles of it.

Two seas on one track

The committed series carries, for every hour of the Fitzgerald's track, the fetch upwind of her position along the reanalysis wind, the sea that wind and fetch allow under the fetch-limited growth law, and the significant wave height ECMWF's own wave model produced inside the reanalysis.

The two estimates agree within a metre once the north-wester has the lake to work on, and the wave model is the higher of the two after five, because it carries duration as well as fetch: by evening the wind had blown from one quarter for six hours and the sea had caught up with it. The wave model's largest individual wave at the wreck site, the ERA5 field for the highest single crest in a three-hour record, is 12.6 metres at seven, which is what a 6.6-metre significant height implies when the biggest wave in a few hundred is about twice the significant one. Hultquist, Dutter and Schwab's wave model gave significant heights above 7 metres at seven and 7.8 at eight, moving from west to east "nearly perpendicular to the documented track". The Anderson's sixteen feet at seven is 4.9 metres, a ship's estimate at night in snow; her master's eighteen to twenty-five feet south of Caribou Island is 5.5 to 7.6 metres, and the Safety Board took 25 feet as its design sea for the structural analysis.

The morning rows are the other half of the lesson. At seven the fetch-limited sea under the north-easter is a metre and a half, because the Ontario shore is sixty miles upwind and the wind has eased; the wave model still has three metres, the leftover of the night's gale, and the Fitzgerald reported ten feet. Fetch tells you what a wind can build. It does not tell you what is already running.

The angle of it

The wave came from the west-north-west and the Fitzgerald steered 141. That is a sea on the starboard quarter, forty to fifty degrees abaft the beam, which for a 729-foot ship is the angle that rolls her and runs along her deck rather than the angle that stops her. The relative-wind column of the series has the wind 150 to 175 degrees off her bow from two in the afternoon on. Her master's words to the Avafors were that she was taking heavy seas over the deck; the Marine Board's finding was that water came aboard faster than it could leave.

The explorer

Drag the hour and the ship moves along her track with the reanalysis wind under her; the dashed line runs upwind from her position to the first shore, and the readout gives the fetch, the fetch-limited sea, and the wave model's own answer. Take the wind in hand to see what each direction opens. A north-easter at Caribou Island finds the Ontario shore in thirty miles; a north-wester finds nothing for a hundred and sixty. The growth law here is the JONSWAP relation of 1973 capped at a fully developed sea, with no allowance for duration, which overstates a sea that has had less than ten hours to grow and is why the model line is the one to trust at the end.

The honest way to ask how bad a storm was is to measure every other storm the same way. The reanalysis makes that possible for the first time on this lake: its 10 m wind exists for every hour since 1940, over the whole surface, from one model with one set of habits. The climatology here reads it every three hours of every November from 1940 to 2025, over the lake's water cells, and reduces each reading to a lake-wide mean wind, the strongest single cell, and the share of the lake above gale force. A storm episode is a run of readings in which the lake-wide mean stays above 20 knots; there are 394 of them in 86 Novembers, about four and a half a month.

Where 1975 sits

By the lake-wide mean, the hour of four in the afternoon on 10 November 1975 is the strongest in the record: 34.9 knots averaged over the whole lake at once, with 45 in the strongest cell and 62 per cent of the lake above gale force. Out of 20,640 three-hourly readings it is the highest, which puts it at the 99.995th percentile. The nearest challenger is 21 November 1977 at 34.5; the median November's strongest gale is 28 knots.

The Armistice Day storm of 1940, which killed sixty-six people on Lake Michigan, ranks eighteenth here, and the Daniel J. Morrell storm of 1966, a Lake Huron loss, twentieth; both did their worst on other lakes. Change the measure in the widget and the order moves. By the strongest single cell 1975 still leads, at 45 knots; by hours with more than half the lake above gale force, 1977 and several others ran longer. What 1975 had that the others did not was breadth: the whole lake under a 35-knot mean for a time, with the strongest of it over the south-east basin where the two ships were.

What this does not say

A 31-kilometre reanalysis cannot see a squall, a coastal jet or the acceleration off the Huron Mountains that the mesoscale model found, so its strongest cell understates the strongest wind anywhere on the lake in every storm alike. The ranking is of storms as a global model sees them, which is a fair comparison between storms and a poor measure of any one gust. The sweep is November only; the great October storms, including the 955-millibar low of October 2010, are outside it, and so is the 1913 storm, which predates the record. And the reanalysis has grown more observations to assimilate with every decade, so a 1946 storm is reconstructed from less than a 2005 one, which argues for reading the early ranks with a little caution rather than for discounting them.

Hultquist, Dutter and Schwab put return periods on the conditions at the wreck site from a wave-hindcast atlas: about two years for a 50-knot sustained wind over eastern Lake Superior, about twenty for 65, and two years for a significant wave over seven metres. Their conclusion was that the conditions were "not typical" but "not rare" for the lake. Both statements are about a point. Over the whole lake at once, in eighty-six Novembers of one model's record, this storm has no equal.

The Weather Service issued six forecasts for the lake between Sunday morning and Monday afternoon. The table under the widget sets each one beside what the reanalysis says the Fitzgerald then met, from the hour of issue to the next issue, so the forecasts can be read as the people on the bridge had to read them: forward, one at a time.

What they said and what came

The Sunday forecasts had the storm. The seven o'clock gale warning named north-east 25 to 37 knots overnight, 28 to 38 by Monday, and a shift to north-west or north at 30 to 40 by Monday afternoon with five- to ten-foot seas. The reanalysis along the track through that period peaks at 32 knots and 2.7 metres, inside the forecast. The 10:39 revision was the weakest of the set: it took the afternoon wind to west or south-west at 35 to 45, which is the wrong side of the shift for a ship bound south-east, though it raised the seas to 8 to 15 feet. The two o'clock storm warning put that right: north-east 35 to 50 becoming north-west 28 to 38, waves 8 to 15 feet. It was the warning that sent both ships to the Canadian shore.

By Monday morning the forecasters had the afternoon. The 10:34 revision said north to north-west 32 to 48 knots this afternoon, north-west 25 to 48 tonight, waves 8 to 16 feet; the reanalysis along the track gives 44 knots, gusts 66 and 5.7 metres before the next issue. The 4:39 revision, broadcast twenty minutes after the Fitzgerald reported her list, said north-west 38 to 52 with gusts to 60 early tonight and 8 to 16 feet. The ship was lost two and a half hours later in 41 knots, gusting 62 to 67, and a 6.9-metre sea. On the wind the forecast was right. On the sea it was low by about seven feet at the top of its range, and the ships' own reports of 16 to 25 feet say the lake was higher than the reanalysis too.

There is no case here that the warning failed. The storm warning preceded the loss by seventeen hours and named the quarter and the strength that killed her. What the forecasts of 1975 did not carry, and could not, was a wave field: significant height was a single range for the whole eastern lake, and the difference between 8 to 16 feet and a 7-metre sea on a quartering course is the difference a modern wave model draws every hour.

The same voyage, earlier or later

Hultquist, Dutter and Schwab ended their paper with the observation that a ship on the Fitzgerald's course six hours earlier or later would have missed the worst of it. The slider tests that in the reanalysis. The track is held exactly where it was; only the clock is moved, whole hours at a time, and the worst hour along the shifted track is reported.

Earlier holds up. Six hours earlier the Fitzgerald would have been off Whitefish Point before the shift reached her, in 33 knots and a three-metre sea, and twelve hours earlier she would have been through the Soo before the storm warning was issued. Later does not, in this record: three hours later the sea along her track is a few centimetres higher, and six hours later it is still over six metres, because the north-wester held the eastern lake through the night. The curve falls away only by about midnight. That is the plain finding of the reanalysis and it is worth stating against the paper's claim, which came from a model that stopped at one in the morning. A ship that had waited at Two Harbors on Sunday evening would have sailed into the same lake on Monday night.

What would be different now

Everything the reader has scrolled through was computed after the fact. In 1975 the lake had no buoys, Marquette made no observation on the Monday, the Whitefish Point station failed during the evening and the radio beacon there was off. Today the same storm would be run forty times by an ensemble three days out, and the wave field over the eastern lake would be a published chart every hour, with its heights on the quartering course a ship could read off before she turned. None of that would have changed the pressure gradient on Monday afternoon. It would have changed what a master knew at two in the morning about the sea he would meet at six in the evening, and that is the margin this page is about.