A tide gauge measures the sea. Ours agree with each other to within a couple of centimetres, all the way up the coast. One of them, twice a year, records a high water forty minutes early and two metres too big.
I look after tide gauges for a living. There are more of them than you would think, a chain of them all round the coast, and most people have never seen one and would not know what they were looking at if they had.
A gauge is a tube. It goes down into the water off a pier or a harbour wall, and inside it a float or a pressure sensor or a radar measures how far away the surface is, every minute, for ever. That number goes to a logger in a small green cabinet and from there down a wire or up to a satellite, and it ends up in a national dataset that is used for flood warnings and for charts and for arguments about sea level.
My job is the cabinets. Batteries, desiccant, corrosion, gulls. I have eleven gauges on my patch and I see each of them four times a year and I have done this for nineteen years.
The important thing to hold onto is this. The tide is predicted, not measured. There is a book, and a set of harmonic constants for each place, and the height of high water on any given day in 2040 is already known and printed. The gauge does not tell us when high water is. The gauge tells us how far the actual sea was from what we predicted, and that difference is called the residual, and the residual is weather. Wind piles water up, pressure drops let it rise, and you get a surge of half a metre, sometimes a metre in a bad storm.
Residuals are small. That is the whole point. A metre is a serious event and gets its own report.
Gauge 07 is on the old pier at a town I will not name, and it has been there since 1971, and it is the best-behaved instrument on my patch.
Its residuals are tiny. It agrees with the two gauges either side of it, forty and sixty miles away, in the ordinary way, all year.
Twice a year it records a high water that is forty minutes early and two point one four metres above prediction.
Not about two metres. Two point one four, and I will come back to that.
I need to explain why a gauge cannot do that, and there are three reasons, and they matter.
The first is that the sea does not have edges. Water two metres higher at that pier would be water higher for miles either side of it, and gauge 06 and gauge 08 would see it, because that is what water is. They see nothing. Their traces on those nights are flat, ordinary, boring.
The second is the shape of it. A gauge fault gives you a spike, or a step, or a drift, and all three look like what they are: a spike is one minute wide, a step never comes back, a drift creeps. What gauge 07 records is a tide. It comes up over about four hours in a proper curve, it stands, and it goes down over about six. It is the shape of water moving, drawn by an instrument that has been drawing that shape correctly for fifty years.
The third is that a real high water two metres over prediction at that place would put the sea over the harbour road and into the amusements, and there would be photographs, and there are not, because it did not happen.
I have the dates. Thirty-eight of them since the record starts, and twenty-two of those are since I took the patch over.
They are not on the same calendar dates. That was the first thing I checked, hoping for something boring like a timer or an annual maintenance cycle.
They are on the two highest spring tides of the year. Not the highest tide, the two highest, one in the spring and one in the autumn, the equinoctial springs, which is exactly when you would expect a real surge to do damage if there were one.
So whatever it is, it knows the calendar of the moon, and it turns up on the two nights of the year when a two-metre surge would matter most, and it arrives forty minutes before the water does.
I have replaced the instrument twice.
In 2011 we took the float gauge out and put a radar in, which is a completely different physical principle. A float measures a float. A radar measures the distance to a surface with a microwave pulse. There is no shared component, no shared cable, no shared anything except the cabinet and the concrete.
The autumn of 2011 recorded two point one four metres above prediction, forty minutes early.
In 2019 we moved the whole installation eleven metres along the pier onto a new bracket with a new cabinet and a new logger and a new power supply and a new telemetry unit.
The autumn of 2019 recorded two point one four metres above prediction, forty minutes early.
I want to say the thing about the number now.
Two point one four metres, every time, to the centimetre, on thirty-eight occasions across three different instruments and two locations eleven metres apart.
Real water does not do that. A real surge in 1974 and a real surge in 2019 would be different by tens of centimetres because the wind was different, because everything is always different.
An identical figure is the signature of a machine, and there is no machine.
I have written that sentence in a report and had it come back with a comment asking whether I had considered a firmware constant.
I have considered a firmware constant. There is no firmware in common between a 1971 float gauge and a 2011 radar and I said so, politely, and the file was closed as no fault found, which in fairness is exactly what it is: the gauge is working perfectly. It passes every calibration. It agrees with everything, all year, except twice.
Now the part I have not put in a report.
In 2014 I was there for one. Not on purpose. I was doing a battery change late because I had been held up, and I was in the cabinet with the door open and a head torch on, and the logger was ticking away the way they do, and I watched the number climb.
The sea was two feet down the pier ladder where it should have been. Flat calm. You could hear it slapping.
And the display went up and up in one-minute steps, and stood, and I stood there with it, and it read two point one four over, and the actual water was where the book said it would be, and I could see it.
I did the only thing I could think of, which was to go down the ladder and put a tape on it, because I am a person who measures things. I got down to the fourth rung and I put my hand on the wet steel below the waterline and it was dry.
Not damp. Dry, and warm, like a wall in the sun, at half past two in the morning in October.
I came back up.
I am due to go out to 07 in three weeks. The autumn springs are on the fourth.
I have started doing something that I know is not my job, which is that I look at the trace for the old pier for the three or four days afterwards, in case something else shows up.
Nothing ever does. It is one tide, it is beautiful, it is impossible, and then everything is normal again for six months.
Except that in 2022 I extended the window and went looking a fortnight either side, and I found the thing I have not been able to put down since.
Forty minutes. That is the number I keep coming back to, because it is the wrong sort of number. It is not a phase error and it is not a clock drift.
Forty minutes before high water at that pier is high water at a place ten miles up the coast where there is nothing now but a shingle bank and the stumps of a jetty, and where, until 1953, there was a village.