Tide accuracy, port by port

In France, our harmonic constants are fitted to Shom/REFMAR's validated hourly observations from January 2023 to June 2025, then compared with seven held-out months of observations, July 2025 to January 2026, never used for the fit. Observations carry the weather (surge, pressure, seiches): an observed high water moves by a few minutes with the day's conditions, so an error measured against the real sea is not a pure computation error.

Elsewhere, the constants come from NOAA in the United States and from TICON-4 (DGFI-TUM) in the United Kingdom, Australia, South Africa and Spain. Times are checked against 28 days of tide gauge readings relayed by UNESCO/IOC (one reading a minute) and, in the United Kingdom, by the Environment Agency (15-minute means); in the United States, against NOAA's official predictions over a few days. A port passes with a bias within ±5 minutes and a mean error under 8 minutes, the level our French ports reach against their gauges.

Le Havre and Ouistreham have a stand of high water (the sea stays high for almost two hours): the exact time of the maximum is inherently fuzzy there. In the Mediterranean the range is around 30 cm and the time of so flat a maximum means little; the app says so. Ports where the time of high water is not a well-defined thing (small tides, the double tides of the central Channel) are set aside, not scored.