Key facts
| Question | Answer |
|---|---|
| Validity | A prediction applies at the gauge that produced it |
| Phase speed | About 10 m/s in 10 m of water, 44 m/s in 200 m |
| Offset example | 20 km at 10 m depth is roughly 33 minutes |
| Route effect | Sills, straits and headlands add delay beyond distance |
| Amplitude | Sheltering, convergence and resonance change the range |
| Official offsets | Subordinate stations use observed time and height corrections |
| Weather | Sustained wind and pressure move timing and level |
| Safe use | Treat a distant prediction as a range, not a clock time |
A prediction belongs to a gauge
The decision is not whether the borrowed table is accurate. It is how much margin the table needs: read the route, estimate the phase error, bracket the event, and pick a mark whose window survives the uncertainty. Every published tide is the sum of harmonic constituents fitted to a water-level record at a named station. The fit carries that station's amplitude, that station's phase and that station's datum, which is why the same day produces different numbers at two ports inside one bay. NOAA publishes the fitted constants station by station and derives predictions from them; the constants are not a regional average that can be moved around the map. A tide table also inherits the station's exposure. A gauge inside a harbour behind a breakwater records a tide that has already been filtered by the entrance. A gauge on an open coast records the swell-influenced water outside. Neither record is wrong. Each is local, and the difference between them is the reason subordinate-station corrections exist. The practical form of the rule is simple. Take the station name, find it on a chart, and measure the water route to the mark. If the route crosses a shallow sill, a narrow passage or a headland, the borrowed prediction has already accumulated error before you apply it.
Phase speed is set by depth
A tide travels as a long shallow-water wave, and its speed depends on depth: the square root of gravitational acceleration times depth. In 10 m of water that comes to about 9.9 m/s, or 36 km/h. In 50 m it is about 22 m/s. In 200 m it is about 44 m/s. Deep water carries the wave quickly and shallow water slows it down. Those speeds convert distance into time. Twenty kilometres at 10 m depth is roughly 33 minutes of phase. The same twenty kilometres at 5 m depth takes about 47 minutes, and at 200 m it takes about eight minutes. Distance alone says nothing useful until the depth of the connecting water is known. The wavelength follows from the speed and the period. The principal lunar semidiurnal tide has a period of about 12.42 hours, roughly 44,700 seconds. At 70 m depth the wave speed is about 26 m/s, so one wavelength is near 1,170 km. That scale explains why a bay a few tens of kilometres long sits inside a single phase contour, while a coast several hundred kilometres long can span a large part of one. Run the arithmetic for a specific trip and the margin becomes concrete. A mark 15 km from the gauge across a 6 m deep bay carries about 38 minutes of phase. The same mark reached around a headland by a 30 km route over 8 m of water carries about 59 minutes. The second route is only twice as long, but the shallow water and the detour together more than double the delay, and no straight-line distance on a chart shows either term. Tide charts compress this information into cotidal lines, which join places where high water arrives at the same time. Where the lines crowd together, the phase is changing quickly and a short move along the coast buys a large timing error. Where they are widely spaced, a borrowed prediction is less dangerous. Reading the spacing is faster than computing the phase speed, and it uses the same physics.
The route matters more than the straight line
A tide does not travel along the straight line between two points on a chart. It follows the channels that hold water. A headland that projects across the shelf forces the wave to refract around it. A shallow bar delays the front. A causeway with one small opening turns a short crossing into a long detour through a narrow throat. This is why two marks at the same distance from one gauge can differ by half an hour. The mark with a deep, open connection tracks the gauge closely. The mark behind a sill lags, sometimes by the better part of an hour, and the lag grows as the tide range falls, because a smaller wave is more sensitive to the same constriction. Estuaries make the effect obvious. The gauge at the mouth records the ocean tide, while a mark ten kilometres upriver records a wave that has been slowed, steepened and partly reflected. The time difference between the two is not a fixed offset through the cycle; it changes with river flow and with the stage of the tide.
Amplitude is a separate error
Phase is the first error and height is the second. A station in an open bay may record a two-metre range while a mark at the head of a funnel records three and a half metres, because the narrowing cross-section concentrates the same volume of water into a smaller area. The reverse happens behind an island or inside a sheltered lagoon, where the range is damped. Resonance adds another term. A semi-enclosed basin whose natural oscillation is close to the tidal period amplifies the tide at its head, and the amplification is largest when the basin length, depth and entrance are tuned to the forcing. Two bays of similar length can therefore produce very different ranges if one is wide and deep and the other is narrow and shallow. Datum differences finish the job. Heights referenced to mean lower low water, to chart datum or to lowest astronomical tide are not interchangeable, and subtracting one from another produces a number with no physical meaning. Before comparing a borrowed height with a mark, confirm that both are on the same reference. Wide bays add a further twist that distance alone cannot capture. In a rotating basin the tide propagates as a Kelvin wave, trapped against one shore and decaying across the bay over a length scale set by the wave speed and the Coriolis parameter. The near shore can run a full range while the far shore runs a fraction of it, and the two can differ in phase as well. A borrowed amplitude from the far side is then wrong in a way that no time offset repairs.
What official offsets can and cannot do
Many agencies publish subordinate stations: a reference port plus a time offset and a height correction. These corrections come from local observation, which makes them better than any distance calculation. Used inside the area where they were derived, they are the right tool. Their limits are geographical. A subordinate correction assumes the same tidal shape as the reference port. It fails where the local geometry changes the regime, where a river adds a non-tidal component, or where the correction was computed for a different stage of the spring-neap cycle. A correction labelled for high water does not automatically apply at low water. Where no subordinate correction exists, the choice is between observing the mark directly and using the nearest station with a deliberate margin. A guessed offset presented as a precise time is worse than a stated uncertainty, because it hides the error instead of budgeting for it.
Bracket the event instead of pinning it
A borrowed prediction is best used as the centre of a window. If the table gives low water at 09:12 and the route suggests thirty minutes of uncertainty, be on the mark at 08:40 and stay until 09:50. That covers the plausible range and costs an hour of standing around, which is cheaper than missing the stage. For a mark that dries at low water, the bracket becomes a safety rule. Arrive while there is still enough water to float and leave before the return route closes. The uncertainty is not an inconvenience to be argued away; it is the reason the margin exists. The bracket also settles the choice of mark. A mark that produces for four hours tolerates a thirty-minute error. A mark that produces for forty minutes does not, and a distant gauge should not be used to plan it.
When the gauge describes different water
Some stations are close on the map and useless in practice. A gauge upstream of a barrage, inside a locked basin or behind a causeway may record a tide that never reaches the mark. In those cases the number is not late or early. It belongs to another water body with another connection to the sea. The test is physical. Watch the mark through one rise and fall and compare the direction of change with the gauge. If the water at the mark rises while the gauge shows a fall, discard the station. If the mark peaks while the gauge is still climbing, keep the station but record the lag. A gauge can also be right about the tide and wrong about the current. Height and flow are separate predictions, and the relationship between them changes with the shape of the inlet. Use a current station for flow and a tide station for depth.
A field check that takes ten minutes
At the mark, note the time when the water stops rising and the height it reaches. Compare both with the borrowed prediction. Repeat on three separate cycles with different wind. A stable offset within ten minutes is usable as a local correction. An offset that changes with wind direction is not a correction at all; it is weather, and the right response is to carry the bracket rather than a number. Record the station name, the distance, the route depth and the observed offset in one line. After three entries the log tells you whether the borrowed table deserves trust or only a margin. The final check is arithmetic, not faith. If the mark's usable window is shorter than twice the estimated phase error, the distant station cannot plan the trip. Move to a gauge inside the same water body or choose a mark with a longer window.
Sources
- Tides and Currents FAQ — NOAA Tides and Currents · T1
- Understanding Tides — NOAA Center for Operational Oceanographic Products and Services · T1
- Tidal Glossary — Permanent Service for Mean Sea Level · T2
- Harmonic Constituents — NOAA Tides and Currents · T1