Every weather forecast carries a pressure figure, and almost none of them say a word about it. The number next to it, the UV index, gets a colour, a scale and advice, because the World Health Organization published a scale for it and sites implemented it.
Nobody did that for pressure. The nearest thing is a glossary the Met Office keeps for shipping forecasts, grading how quickly the barometer is moving; this site borrows its wording. But there is no international scale: no index, no colour bands that mean the same thing everywhere. Where a weather site does grade pressure, the scale is its own invention. So the number sits unexplained in a table, while plenty of people look it up on purpose: anyone with a barometer on the wall wants to know whether it is set right, and more people than we expected watch it because they say they can feel a change coming.
So we built a site for that one number. Here is what we learned.
The direction matters more than the number
Ask what the pressure is and you get 1004 hPa, which says little on its own. Ask which way it is going and you learn something: down nine hectopascals since yesterday means a system is arriving, probably with wind and rain. Up six and settling means the opposite.
Almost no site leads with this, so ours does. Every city page states the direction and the size of the change in a sentence, and the graph runs a week back as well as a week ahead, because many readers are trying to make sense of a day that has already happened.
The tropics nearly broke the design
We wanted a list of the cities where pressure was moving fastest. Meteorology has a standard measure for this, the three-hour tendency, so we used it, and the list filled up with cities near the equator in perfectly calm weather.
The reason is a daily rhythm. The sun heats the air in a regular beat, and pressure swells and settles twice a day in response, peaking near ten in the morning and ten at night. Near the equator that swing runs two to four hectopascals, every day, with no weather behind it. Further north it is smaller, but it is still there.
In August 2026 we measured it, city by city, over a fortnight of our own data. For each city we took every three-hour change and checked how well it could be predicted from the local time of day alone. That prediction score is the last column: 96% means the clock almost fully accounts for a three-hour change, 14% means it barely helps.
| City | Latitude | Just the clock |
|---|---|---|
| Nairobi, Jakarta, Bogotá | 1 to 6 | 96% |
| Singapore | 1 | 95% |
| Mumbai | 19 | 91% |
| Manila | 15 | 89% |
| Tokyo | 36 | 45% |
| New York | 41 | 43% |
| London | 52 | 40% |
| Berlin | 53 | 16% |
| Moscow | 56 | 14% |
Even London is 40 per cent clock, so the standard measure was partly ranking cities by what time it was there. Over 24 hours the daily rhythm cancels everywhere on earth. That is why the headline here is the change since the same time yesterday, and why a figure from Jakarta means the same as one from Warsaw. Tropical city pages also explain the shape, because a graph that saws up and down twice a day looks like trouble until you know what it is.
Mexico City reads 786 and that is normal
Pressure falls as you climb, roughly one hectopascal for every eight or nine metres near the ground, so a raw reading tells you mostly about a city's altitude. Weather services correct readings to what they would be at sea level; that is why 1013 is the number everyone recognises and why two cities can be compared at all.
For a high city the corrected figure describes an atmosphere that does not exist. In Mexico City, 2,238 m up, a barometer read 786.4 hPa in our own data while the sea-level figure for the same hour was 1019.0. Neither is wrong, and anyone comparing a barometer there against a forecast could reasonably decide one of them is broken.
Sea level is the default here, because comparing cities is the point of the correction. But every page has a switch to the reading at the city's own height, high cities show both without being asked, and each page says which one is on screen.
Pressure is the number models are best at
We expected forecast uncertainty to be a design problem, so on 11 August 2026 we put the same city and hour, London, through four of the major global models: ECMWF, GFS, ICON and the Met Office. They came back 1023.1, 1023.4, 1023.6 and 1023.8 hPa. The whole spread was 0.7 hPa. Pressure varies broadly and smoothly, which is what models do best.
The agreement does not last all week. The same comparison at the far end of the graph, widened by a fifth model, had them six to nine hectopascals apart by the sixth day: a different low arriving at a different depth. So the site shows one model and names it, treats the first days as the firm part of the graph, and spends its effort on explaining the shape rather than on hedging every value.
Where this comes from
Every figure above was measured on our own data. How this works sets out the model, its resolution, the daily rhythm and what can go wrong with any of it.
This site reports air pressure and how much it has changed. It gives no health advice and makes no prediction about how anyone will feel.
