Every tide gauge in the United States publishes the same thing: a list of thirty-seven waves. Each one is a pull from the moon or the sun, restated as a sine wave with an amplitude and a phase. Add them up and you get the tide — days, decades, centuries ahead, with no measurement of the ocean at all. This page ships the coefficients for — stations and does the adding in your browser. Then it goes looking for the days the sum was wrong.
Every dot below is a tide gauge, placed at its own latitude and longitude. Nothing else is drawn — no basemap, no outlines. The shape of the country appears because the gauges are on the water, and there are enough of them. Press play and each dot is coloured by the water level the model computes for it at that instant, so you are watching the tide sweep the coasts, not an animation of one.
These are the thirty-seven waves for the station you selected, each with the amplitude and phase NOAA measured from years of that gauge's own record. Switch them on one at a time and watch the curve become a tide. The error readout compares your partial sum against all thirty-seven — it is the honest answer to "how much of the tide is just the moon?"
The model above knows nothing about wind, pressure, rainfall or hurricanes — only astronomy. So the difference between what it predicts and what the gauge actually recorded is, almost exactly, the weather's contribution to sea level. That difference has a name: storm surge. Below, the page runs its own thirty-seven-wave prediction across days when things went wrong, and shades the gap.
The Potomac at the foot of Wisconsin Avenue is tidal, and the crest that arrives there is not a local event: it is a wave that entered the Chesapeake at Cape Henry and has been travelling ever since. Because every gauge publishes its M2 phase in the same Greenwich reference, the travel time can simply be read off — no model, no assumption, just subtraction.