Where the tide predictions come from
Every number on a tide page is NOAA's. This is which NOAA product, how it reaches the page, what the numbers mean — and, most importantly, what they leave out.
The source
Tide data comes from NOAA CO-OPS, the Center for Operational Oceanographic Products and Services — the official United States tide authority. The page asks its predictions API for one thing: the high and low list for that station.
Only the extremes are requested, because only the extremes are published as measurements. The smooth curve drawn between them on the chart is interpolation for readability, not additional data, and should not be read as a prediction of the level at some in-between minute.
It reaches the page two ways, and the difference is worth knowing. The interactive chart and the station finder ask NOAA directly while you are on the page, so they show whatever NOAA publishes at that moment. The seven-day table printed into the page itself is served from a rolling 45-day window of NOAA predictions that a scheduled job fetches once a week and stores with the site — that is what lets a tide page render its table instantly, with no network request, and what lets a search engine read it.
Storing predictions would be a bad idea for a measurement. It is a safe one here because a tide prediction is not a measurement: it is astronomy, solved from decades of observation at that station, and NOAA does not revise it afterwards. A prediction fetched last Tuesday for next Friday is the identical number NOAA would return for next Friday today. Age within the window changes nothing.
What could go wrong is the window running out, so it is built not to fail quietly. Each station's data is only replaced when a fetch succeeds, so an unreachable NOAA leaves slightly older predictions rather than blanks; a run in which no station refreshes fails loudly instead of reporting success; and the window is 45 days against a weekly job, so several consecutive failures would have to go unnoticed before a table could run short. If it ever did, the table would be absent — never wrong.
What "astronomical prediction" excludes — read this one
A tide prediction models the gravitational pull of the moon and the sun on the water at that station, worked out from decades of observation at that specific place. It is genuinely good at that. It knows nothing whatsoever about the weather.
Wind. A sustained onshore wind piles water against the coast; an offshore wind pushes it away. Either can be worth a foot or more.
Barometric pressure. Low pressure lets sea level rise; high pressure presses it down.
Storm surge. During a storm, the difference between predicted and actual is the entire point of a surge warning.
Rain and river discharge. In an estuary, upstream rainfall changes the level days later.
So: do not use these pages for navigation or for any decision where being wrong is dangerous. Not for clearing a bar, not for anchoring, not for judging whether a sandbar walk is safe on the way back. For those, go to NOAA directly and check current conditions with the U.S. Coast Guard. These pages are for knowing roughly when the water will be high on Saturday.
Reading the numbers
Most stations see two highs and two lows a day, of noticeably unequal size — the diurnal inequality. Some places on the Gulf coast get one of each. Both are in the data as NOAA publishes it; neither is an error.
How a city gets matched to a station
Sun and moon pages exist for cities; tide stations exist where NOAA has instruments, which is rarely the middle of a city. The two are matched once, when the site is built: each station is paired with the nearest city that has a sun page, within 35 miles and in the same state. Outside that radius, no link is offered.
That is why most city pages have no tide link and are not missing one. The United States has coasts; Kansas does not. A matching rule that reached further would produce a link that technically resolved and practically misled.
The links are also checked for reciprocity at build time: if a station links to a city, that city must link back, and the build fails if it does not. A one-way link renders perfectly and silently dead-ends anyone following it.