technology · · 6 min read
Position Z: The Brass Machines That Forecast the Tide for D-Day
In 1943 a tide scientist in Liverpool was asked to predict the sea at an unnamed beach. His mechanical analogue computers, built from gears, cranks and pulleys, helped set the timing of the Normandy landings.
On 9 October 1943, Commander William Ian Farquharson, the Royal Navy's superintendent of tides, sent a three-page handwritten letter marked "MOST URGENT" to a scientist near Liverpool. It held 11 pairs of numbers for a place called only "Position Z", and it asked for hourly tide heights there for four months starting 1 April 1944. "The place is nameless and the constants inferred," Farquharson wrote. He added that he was "gambling on the inferred shallow-water constants giving something like the right answer" [1][6].
The man who got the letter was Arthur Thomas Doodson of the Liverpool Tidal Institute. Position Z was Normandy, and Doodson would answer the request with a machine made of brass gears, cranks and pulleys.
Why the tide decided the plan
Along the French Channel coast the gap between low and high water was always more than 6 metres. At low tide that left long stretches of open sand under German guns [1]. Field Marshal Erwin Rommel expected the Allies to land at high water. From February 1944 he had obstacles built across the intertidal zone: log stakes, some with mines on them; "hedgehogs" made of crossed iron bars; and steel frames called Belgian gates. They were placed so they would be hidden once the tide was about half in [1].
Allied reconnaissance aircraft spotted the obstacles, and the planners changed their approach. The first waves would land soon after low tide, while the obstacles were still exposed and demolition teams could clear lanes through them. The tide also had to be rising, so landing craft could unload and pull off without being stranded. Other limits were added on top: an hour of daylight for the naval bombardment, darkness for the Channel crossing, and a Moon that rose late so paratroopers could find their targets. In June 1944 only the 5th, 6th and 7th met every condition [1]. With a 6-metre range, the water climbs at least about a metre an hour, so the time of low water and the speed of the rise had to be known precisely. Low water also came more than an hour apart at the two ends of the invasion front, Utah and Sword. H-Hour therefore had to be set separately for each beach [1].
A computer made of pulleys
The tide at any port can be described as the sum of many regular waves, called harmonic constituents. Each one comes from a periodic feature of how the Moon and Sun move relative to the Earth. Measure enough of a port's tide and you can work out the height and timing (the amplitude and phase) of each constituent. Add them back together for future dates and you have a prediction. William Thomson, later Lord Kelvin, worked out how to use this kind of harmonic analysis for tides. In 1872 the London instrument maker A. Légé & Co. built his first working tide-predicting machine [3]. Philip Woodworth, who has written a history of these machines, notes that the design came from a device Charles Wheatstone had used in telegraphy, and that it could simulate a tide made of 10 constituents [2].
The machine works like this. Gears turn one crank for each constituent at its own speed. The length of the crank sets the amplitude and its starting angle sets the phase. Each crank moves a pulley up and down. One continuous metal band runs over all the pulleys, so the position of its free end is the sum of every motion, and a pen attached to it draws the tide curve on a turning drum [2]. The Science Museum, which holds Thomson's 1872 machine, says it could draw a full year of tide curves for one harbour in about four hours [3]. Between the 1870s and the 1960s, a little over 30 of these machines were built around the world, most of them in Britain [2]. The United States built its own designs. William Ferrel's machine was finished in 1882. Tide-Predicting Machine No. 2, later nicknamed "Old Brass Brains", followed. It handled 37 constituents and weighed about 2,500 pounds [4][5]. NOAA gives 1910 as the year it was completed; Bruce Parker's account in Physics Today gives 1912 [1][4].
Two machines in two rooms
By the 1940s Doodson's institute at Bidston Observatory had two machines. One was a Kelvin-type machine bought in 1925, with 29 constituents after it was extended. The other was a machine designed by Edward Roberts in 1906, extended to 40 constituents [2]. Published accounts give different dates for the Bidston Kelvin machine. Woodworth's study, based on the institute's records, says it was built in Glasgow by Kelvin, Bottomley and Baird and bought in 1925. Some popular accounts date it to 1872 or 1924 [1][2][6].
The two machines worked in separate rooms so that one bomb could not destroy both. As a precaution, predictions for every port in the Admiralty Tide Tables were finished two years ahead [1][6]. Bombs did land near the observatory, but both machines kept working [1].
Farquharson's constants were themselves partly a guess. The Admiralty had good tidal data for Le Havre and Cherbourg, on either side of the landing area, but none for the beaches themselves. Shallow water changes tides in complicated ways, so simply averaging the two ports would not have worked. Small boats and midget submarines went in at night and collected a few measurements, but Parker writes that this was "much less than is normally required" [1]. Nobody knows exactly how Farquharson got from those scraps to his numbers. Parker thinks he probably adjusted the Le Havre constants to fit the measured readings, and notes that the constants "compare favorably" with modern computer models of the Channel [1]. Doodson ran the constants through his machines. The Hydrographic Office then adjusted the results for each beach and combined them with sunrise, moonrise and twilight times to make planning diagrams [1][6].
Security was tight enough that the two men wrote to each other in code names. Doodson was still not fooled. According to Parker, he admitted years later that he had worked out from the constants that the target was the Normandy coast [1][6].
What the predictions did, and didn't, do
Bad weather moved the landing from 5 June to 6 June. Parker reports that Rommel judged the tides on the 6th unsuitable for a landing, which was one reason German forces were less ready [1]. Once the landings began, the predictions proved "quite accurate" [1]. On Omaha Beach, American demolition teams landed an hour after low water, as planned. They still suffered terribly. More than half the engineers were killed, and of 16 planned lanes they cleared six fully and three partly. Parker argues that without precise timing of the fast-rising tide it would have been worse [1]. At Utah the tide predictions held, but an unexpectedly strong tidal current pushed landing craft about 2 kilometres off course, onto a more lightly defended stretch of beach. The machines could forecast tidal heights, but currents were much harder to predict [1].
Tide-predicting machines kept working for another two decades. The US switched to electronic computers for its predictions in 1966 [4][5]. Bidston used its Roberts machine until 1960 [2]. Its successor, the 42-constituent Doodson–Légé machine delivered in 1950, was tested again in 2016 after being restored. Its predictions for Liverpool that year were very similar to those from modern software [2].
That 2016 test is a reminder of what was at stake in 1943. The machinery was mature and well understood. The weak part was the input: 11 pairs of numbers inferred from a few night-time soundings by a commander who openly called his work a gamble. The machine only added up what Farquharson gave it, and the invasion's timing depended on whether his numbers were right.
Sources
- Bruce Parker, "The tide predictions for D-Day," Physics Today 64, no. 9 (2011). https://physicstoday.aip.org/features/the-tide-predictions-for-d-day
- Philip L. Woodworth, "Tide prediction machines at the Liverpool Tidal Institute," History of Geo- and Space Sciences 11, 15–29 (2020). https://hgss.copernicus.org/articles/11/15/2020/
- Science Museum Group Collection, "William Thomson's Tide Predicting Machine, 1872." https://collection.sciencemuseumgroup.org.uk/objects/co53901/william-thomsons-tide-predicting-machine-1872
- NOAA Tides & Currents, "Tide Predicting Machines." https://tidesandcurrents.noaa.gov/predmach.html
- NOAA National Ocean Service, "150 Years of NOAA Tide Tables." https://oceanservice.noaa.gov/news/nov15/tide-tables-history.html
- Tide and Time, "Tidal predictions for the D-Day landings." https://tide-and-time.uk/node/11