Gauge vs. interpolated stations

Gauge stations

A tide gauge is a physical instrument at a fixed, surveyed point on the water: a float in a stilling well, or a radar or pressure sensor, logging the level every six minutes for decades. In this catalogue the gauges are the 3,450 stations of the NOAA CO-OPS network — Boston and San Francisco (Golden Gate) among them — and they wear a “Gauge station” badge.

That measured record is what an official tide table is made of. NOAA computes the station’s datums from it and publishes its own official predictions on them at tidesandcurrents.noaa.gov. If you need the number a harbourmaster would accept, that is where it lives.

The prediction on our gauge pages is not that number. It is produced exactly like every other page on this site — harmonic constants from the FES2022 global model at the nearest ocean point, heights relative to mean sea level — and it is the model, not the instrument, that we are drawing. What the badge tells you is narrower and still worth knowing:

• The position is a surveyed instrument on the water, not a place name that happens to be near the sea.
• A measured record exists there, and an authoritative table exists with it, so our figures can be checked against something.
• Somebody has cared about the tide at that spot for long enough to build a datum for it.

What it does not tell you is that the heights on the page are NOAA’s, or that they are on NOAA’s zero. They are ours, on ours. How we compute sets out the chain from the model to the page.

Interpolated places

Everything else is a coastal place — a town, a beach, a harbour, a headland — with no gauge of its own. There are tens of thousands of them here against those 3,450 gauges, which is roughly the ratio between the coast people use and the coast anybody has instrumented.

For those we take FES2022’s harmonic constants at the nearest valid ocean point to the place and run the same prediction mathematics, with the same 34 constituents, as we do at a gauge. The model itself is built from satellite altimetry and hydrodynamics, so it knows the deep ocean and the open shelf well and the last kilometre of your particular estuary not at all. Those pages carry an “Interpolated” badge.

The honest summary of the difference is that it is a difference of station, not of method. Winthrop, six kilometres across the harbour from the Boston gauge, is predicted exactly the way Boston is; what Boston has and Winthrop has not is a measured record to argue with.

How much the difference matters

Timing travels well. On an open coast, the times of high and low water from a good global model are usually within a few minutes to a quarter of an hour of a nearby gauge, because the times are set by the astronomy and by the shape of a whole ocean basin — both of which the model has.

Heights travel badly. A model cell is a few kilometres wide, and the nearest wet cell to a beach can be further off than that, so ranges inside bays, rivers and lagoons can be wrong, occasionally by a lot. A funnel that doubles the range along the last stretch of an estuary is exactly the feature the model averages away.

And a height carries its zero with it. Ours is mean sea level, and a chart’s is a low-water datum, so even a perfect height on our page is counted from a different floor than the one printed on the chart in your hand.

That is why a low water reads as a negative number here and as a positive one on the chart, and why you must never subtract one from the other without checking — what is chart datum covers it. The tidal range, on the other hand, compares cleanly between any two sources, because subtracting one height from another cancels the zero out.

Weather sits on top of all of it. Roughly a centimetre of sea level answers to each hectopascal of air pressure, and a storm can move the water half a metre or more for a day at a time. On a day like that the difference between a gauge page and an interpolated one is the smallest of your problems.

When to distrust an interpolated page

Be sceptical wherever the water has to squeeze through something before it reaches you:

• Far up a river, where the tide arrives late and distorted and the flood is shorter than the ebb.
• Deep inside a shallow bay, where friction delays the turn.
• Behind a barrier island, or in a lagoon with a narrow entrance, where the opening throttles the range.
• Anywhere the model’s nearest wet point is several kilometres offshore of the place you picked.

The symptoms are easy to spot once you look. A range that bears no resemblance to the next harbour along the coast, a high water an hour away from what everyone locally says, a low that never uncovers what it should — any of those means the model is describing a different piece of water than the one you are standing beside.

The fix is not to abandon the page but to calibrate it. Find the nearest gauge, compare a week of both, and note the offset in time and in height; on a stable coast that offset is roughly constant, and applying it in your head is what local pilots have always done. If the two disagree in a way no constant can explain, use the gauge and the local knowledge that comes with it.

The model knows the ocean. It has never been up your creek.

Why we publish both

There are a few thousand tide gauges in the world and a great deal more coast that people fish, dive, surf, sail and walk. Publishing only gauges would leave most of that coastline blank, and would send someone planning a morning at a beach to a station tens of kilometres away without telling them how far it was. Publishing a model without labelling it would hide a real difference in confidence behind a uniform-looking number.

So every page says which it is, every page names the zero its heights are counted from, and every page uses the same engine, so two of our stations can be compared with each other without a footnote. Pick a gauge when the answer has to be exact and there is one near enough to be relevant; pick the place nearest your water when what you need is the rhythm of the day — when the water turns, which way it is going, how much of the foreshore will be uncovered.

And read either one as a prediction of the astronomy, not a promise about the sea: use how to read a tide chart for the mechanics, and keep a margin for the weather.

Questions people ask