About this data
Where the data comes from
Every number on these five pages comes from the Open Meteo Historical Weather API. The full50 year record comes from ERA5, the reanalysis climate research runs on worldwide. The last few weeks come from a finer model, because ERA5 lands a few days behind. Below we explain when we use which.
How often it updates
A nightly job adds the previous day's weather for every capital that is already live, and keeps filling in older history for capitals still catching up to their full 50 year record. A capital only gets its own page once that record is complete, so every page you can visit is already built on real, finished data.
Which capitals are included
194 national capitals and 51 US states, all with a complete 50 year weather record. The world has about 195 sovereign capitals, so this dataset covers all but one: Vatican City is the only country not included.
Why recent days come from a different model
ERA5 is a reanalysis. Every measurement from a period is run back through one single weather model, so each day is worked out the same way whether it is 1976 or last month. That is what makes the record worth comparing across fifty years. The price is time: ERA5 arrives about five days after the fact, so for yesterday there is nothing yet.
For those last few days we use ECMWF IFS instead. It resolves at 9 kilometres rather than 28, so it is sharper around coastlines and mountains, and it is available the next day. What it cannot do is carry a long record: it only goes back to 2017, and the model itself has improved over those years. A trend read off it would partly be the model's development rather than the climate's.
| View | Source | Why |
|---|---|---|
| 1 day to 30 days | ECMWF IFS, 9 km | runs up to yesterday, ERA5 does not have these days yet |
| 180 days to 50 years | ERA5, 28 km | one model across the whole record, so it compares fairly |
| Looking up a single date | the sharper one that has the day | one day is a measurement, not a trend |
So the same day can carry two slightly different numbers depending on which view you are looking at. Across our whole record the gap averages 0.44 °C, and it does not average away over a longer period, because it belongs to the place. Flat ground like Amsterdam stays within half a degree, while Podgorica reaches 1.5 °C and Reykjavik leans the other way. Mountains and coastlines simply look different at 9 kilometres than at 28.
We do not nudge those numbers towards each other. A correction would hide the difference rather than explain it, and would need adjusting year after year. Instead we keep the two records apart, every view uses exactly one of them, and our API response always states which one answered.
