Weather forecast, hourly and 9 days ahead

Because for most of the world that number simply does not exist: weather services publish probability of precipitation only across their own high-resolution domain — the Norwegian one over Scandinavia, the German one over Germany. For Moscow, Jakarta or Adelaide that field is empty in the data. We could invent a percentage, but it would be invented. Instead we state how many of the available models expect precipitation in that hour. Where a real probability exists we show it and label it as coming from the weather service.

It is the spread between the two models: the top of the band is the warmer forecast, the bottom the colder one. Measured across eight cities on five continents, the models usually differ by less than a degree, but in one hour out of ten the gap exceeds 2.8 degrees and it can reach six. The further ahead the day, the wider the band — that is a normal property of forecasting, not a fault.

For the next day or two modern models are typically off by a degree or so; accuracy falls with range, and by day nine a forecast should be read as a tendency rather than a schedule. That is exactly why we show model disagreement: it is the only honest way to see where a forecast is reliable and where it falls apart.

From two national weather services: the Norwegian Meteorological Institute (an ECMWF-based model) and Germany's DWD (the ICON model and MOSMIX station data). Both publish under CC BY 4.0. We do not recompute their forecasts — we bring them together and show where they did not match.

That is how the source publishes it: hourly for the near term, then six-hour blocks. We do not fill the gaps by interpolation, because that would look like knowledge we do not have. The chart switches to the coarser step openly and says so.

It is temperature adjusted for humidity and wind using the Australian Bureau of Meteorology formula. Unlike the usual wind-chill-plus-heat-index pair, it is continuous across the whole range: a muggy 25 °C at 90% humidity feels noticeably warmer than a dry 25 °C, and the tool shows that instead of returning the same number.

Gusts are measured and published only where a weather service runs a high-resolution network. Outside it the field is absent from the data. We do not derive gusts from average wind speed: that would be a guess presented as a measurement, and people make safety decisions based on gusts.

The city's. Every hour, day, sunrise and sunset is computed in the timezone of the place you selected. If you are in London looking up Seoul, "tomorrow" on this page means tomorrow in Seoul.