Type at least two letters. Any city worldwide.
Reading the field…
The planet's weather, drawn the way Ceiling draws the sky: NOAA's GFS sea-level pressure on a 2° grid, upsampled and traced as isobars every 4 hPa, so a hurricane in the Gulf and a Bering Sea low both read as rings. Beneath them the model's ten-meter wind, above them the coast in gold and the NHC's live forecast tracks. The wind is drawn as motion: a few thousand traces seeded from the field's valid time and carried through it. Your cursor is a probe, and a click opens a card — three facts read off the data at hand, and a page behind it. The rail is the clock: forward eight days through the forecast, back four years through the analyses, then back through every storm the world has recorded since 1842 — seventeen agencies, 13,593 storms. The record is a drawing of the witness, not of the storms. Scrub back and the tracks thin to the shipping lanes, then to a dozen written accounts, then to nothing; the weather was the constant, and what changed was who was watching. Or take the ride, out from the bay's shoreline to the planet.
Mapped regulatory risk, not current water levels, a surge forecast or an evacuation zone. A city reference is a point, not the whole city. Use the effective FEMA map ↗ for a property determination.
Opening this loads the archive index.
Before the track record, and beside it, a storm exists only as far as someone survived it and wrote it down. Each is one mark on the rail and one point on the map — the place the account describes, not a track. Nothing here is invented; every row carries its source.
An independent experiment, not a warning service. Consult the relevant national weather service for official warnings — in the U.S. the National Hurricane Center and the National Weather Service. A storm's effects can extend far beyond its center and its forecast cone. Modeled gale areas are connected model cells, not diagnosed storm centers, and are not official classifications. An empty feed is never an all-clear.
Tracks, cone and past positions: NOAA / NHC tropical summary. Wind and pressure field: NOAA GFS through CoastWatch ERDDAP, a model, covering 1 December 2022 to about eight days ahead in three-hour steps. The archive: IBTrACS v04r01 (NOAA NCEI), which merges every agency's best tracks — NHC, Tokyo, New Delhi, La Réunion, the Bureau of Meteorology, Nadi, Wellington and the JTWC — into 13,593 storms and 726,804 fixes from 1842 to this season. Retired Atlantic names, quoted on a storm's card: the WMO committee's list as the NHC publishes it, Tropical Cyclone Naming History and Retired Names — every row in the table was read there. Worldwide cyclone reports: GDACS. Alerts: NWS. Flood zones: FEMA NFHL. Topography: USGS National Map. The places behind the city search, which / opens: Open-Meteo / GeoNames.
Two things the archive will not pretend. The record is what was recorded — before satellites nothing beyond the shipping lanes and the coasts was seen, so a blank Southern Ocean in 1880 is no observers, not no storms; the readout says as recorded with the year's storm count and the number of agencies that named an intensity. Intensity is not comparable across agencies — NHC winds are one-minute averages, most others ten-minute, New Delhi's three and Beijing's two, and a great many old fixes carry no wind at all; a track is colored by the wind as given, the agency is named beside it, and where the record has no wind there is no category.
Geography: the map layers of Ceiling (003) — Natural Earth 10 m, public domain, and the bay's shoreline © OpenStreetMap contributors, ODbL. The marching-squares tracer and the isobar styling are Ceiling's too.
By Gregg Perkins, built with AI. Vanilla JavaScript, Canvas 2D, marching squares, an orthographic globe. No framework, map SDK, API key or account.
Loading the record…
As recorded. The record is what was recorded: before satellites nothing beyond the shipping lanes and the coasts was seen, so a year with nothing in it is no observers, not no storms. The twelve accounts at the top of this page are not fixes — they are storms that exist only because somebody survived them and wrote it down.
Intensity is not comparable across agencies. NHC winds are one-minute averages, most others ten-minute, New Delhi's three and Beijing's two, and a great many old fixes carry no wind at all. Every peak below is quoted as its agency gave it, with the averaging period and the agency beside it, and where the record has no wind it says so rather than filling one in. A storm is listed under the calendar year its record begins — which is the year the rail stands in, and in the southern basins often the year before the season IBTrACS names it by — and a year's counts are the storms listed under it and the agencies that named a peak for them.
The record ends where this season does. Choose a row and the rail goes to its first fix.