Water runs downhill, so a river in flood is higher upstream than it is two hundred metres further down. Ours is not. The gauge below the bridge reads four tenths of a metre above the one above it, every single time, and only in flood.
I am a flood warden and a retired water industry engineer, which is why I ended up with the job nobody wants.
How a gauge works, because the trick is in the datum.
A river gauge is a sensor in a stilling well, which is a chamber connected to the river by a pipe so the water inside it sits at river level without the waves. The reading is a height above a surveyed datum, and the survey is the whole point: two gauges are comparable only because a surveyor has tied both of them to the same national reference.
Ours were both levelled in, professionally, in 2012, and re-levelled in 2019 after the bridge works.
There are two. One about a hundred and eighty metres above the road bridge and one about twenty metres below it, so two hundred metres apart, on a reach with a fall of about one in four hundred.
In normal flow the upstream one reads about nought point five of a metre higher, which is the fall of the river over that distance, and that is exactly what it should be.
When the river rises past about two metres, they cross over.
At the flood warning level the downstream gauge reads four tenths of a metre higher than the upstream one.
I need to say how wrong that is.
Water has one rule and that is the rule. A water surface can be level, and in a flood with a constriction it can be nearly level, but it cannot be higher downstream, because then the water would be flowing uphill.
Four tenths of a metre over two hundred metres is a gradient of one in five hundred, pointing the wrong way.
What has been checked, and the agency has spent proper money on it because flood data is taken seriously.
Both sensors swapped. Both wells cleaned and the connecting pipes rodded, because a silted pipe gives you a lagging and damped reading, and that is the classic fault.
Both gauges re-levelled in 2019 by a surveyor with a total station, tied to two separate benchmarks. Within eleven millimetres of the 2012 values.
A temporary third gauge, a pressure transducer on a cable, installed in 2020 halfway between them, below the bridge, on the agency's own kit.
In the February 2020 flood the temporary gauge read between the two. Which tells you the surface really is doing this, because three instruments now describe a water surface that rises as it goes downstream.
And the thing I did myself, which is the one I trust most because it has no electronics in it.
In the December 2020 flood I drove a stake in at each gauge at four in the morning, at the waterline, and when the river dropped I had a surveyor friend level the two stakes for the price of a bottle.
Three hundred and eighty one millimetres. Downstream higher.
What people have suggested, and they are all good suggestions.
A constriction at the bridge causing a backwater. That raises the level UPSTREAM of the bridge, which is the opposite, and it is the commonest thing in river engineering.
A tributary joining in between. There is none. There is a surface water outfall from the industrial estate and in a flood it is drowned and carrying nothing.
Wind. Four tenths of a metre of wind setup needs a lake and a gale, not two hundred metres of a river in a town.
Something in the channel. The reach has been surveyed by boat and by drone and there is a shopping trolley and a supermarket pallet.
What the paperwork says now.
The flood warning for this town is issued off the upstream gauge, which was the original design, and there is a note in the procedure that the downstream gauge is not to be used for warning purposes pending investigation.
That note was written in 2016 and the investigation is open.
I have read the file. It has eleven documents in it and the most recent is a 2021 memo that says the anomaly is unexplained, that the upstream gauge is consistent with the hydraulic model, and that no change to the warning arrangements is recommended.
The hydraulic model is the important part of that sentence. The model says the upstream gauge is right. So the warning goes out on the gauge that agrees with the maths, and the other one is in the data and flagged, and nobody is in danger because of it.
What I do.
I put the flood boards in at the church hall and the two bungalows on Ford Street when the upstream gauge hits one point eight, and I have never once been late, and in 2019 I was an hour early and the man at number four gave me a fruit cake.
The thing I stopped doing is standing on the bridge.
A flood warden is supposed to do a visual check, and the bridge is where you do it from, and I did it for nine years with a torch.
In the February 2020 flood I was on that bridge at about half four in the morning with the water two feet under the arch and loud enough that you could not hear a car.
Looking upstream the water was where the upstream gauge said it was. Looking downstream, over the parapet, it was up at the underside of the arch on the far side, four hundred millimetres higher, exactly as the instrument had been saying for four years, sitting there against the stone with nothing pushing it.
I do the visual from the corner of Ford Street now, which is a poorer view and a hundred metres back, and I have written in my own log that this is for safety reasons, and that is true as far as it goes.