The night Feltwater Ness had no low water at all
A neap, a hard southwesterly and 992 millibars. The prediction said 0.38 m of range and the gauge never came below a metre. The prediction was not wrong — it was answering a different question.
Two Tuesdays ago the watch took four calls about Feltwater Ness, all of them the same call. The tables said low water 03:14 at 0.4 m. The gauge sat above a metre all night and never turned.
Nobody’s software was broken. Here is what happened, because it is the clearest example we have of what a harmonic prediction is and is not.
Feltwater is the marginal case
Feltwater sits near a node in the semidiurnal wave. M2 there is 0.28 m — an eighth of what it is at Halberry Roads — and S2 is 0.09 m. Mean spring range 0.74 m, mean neap range 0.38 m. That is the entire astronomical signal at that station: less than the freeboard on a dinghy.
If the weather can move the level by half a metre, and the neap range is 0.38 m, what exactly is the tide table predicting?
The answer is: the astronomy, correctly, and nothing else. On that night the barometer was at 992 mb. Twenty-one millibars below standard, at roughly a centimetre of level per millibar, is about 0.21 m of water that the moon had no part in. Add a southwesterly that had been blowing onto that shore for eighteen hours and the residual reaches the 0.71 m we publish as Feltwater’s worst.
0.71 m against a neap range of 0.38 m. The weather was nearly twice the tide.
What we will not do about it
The wrong fix: quietly widen the tolerance
The tempting response is to stop publishing a range for stations like this, or to publish one with the surge already folded in so the number looks better against observations. Both would make the tables read as more reliable and both would be a lie about what is in them.
An eight-constituent harmonic prediction contains no weather. Not “a small amount of weather”, none. Folding an average surge into the published height would mean the number is right on an average day and silently wrong on the two days a year that matter, and there would be no way for the reader to tell which day they were having.
So the range stays as the astronomy gives it, and the residual is published beside it on the accuracy page, where it can be compared with the range and found to be larger. That comparison is the useful fact. It is the one thing a table can tell you about its own limits.
What we did do
Feltwater’s station page now leads with the sentence that the tide there is mostly wind. It is the only station in the eight where that is true, and burying it under a tide table with two decimal places on it was our mistake, not the prediction’s.
And the calls were right to come in
Every one of those four calls is now a line in the record that produces next year’s residual figures. If the water did not do what we said, that is the message we want — the duty watch takes it on 0191 498 0272 or in writing through the contact page.
Every station, with its residual, offline
The published residuals ship with the app. So do the two control stations, because a predictor you cannot check is one you should not be using at night.
Wrackline is an aid to planning and is not a substitute for an official tide table. Do not navigate on it.