The model league
This month, the Atlas Forecast has the smallest error on tomorrow's temperature in 42 of the 51 provinces with data; in 17 of them, with a lead that is not down to chance.
Which weather model is most accurate in each Spanish province, measured against weather stations over the last 30 days. When the data cannot tell who wins, we say so.
By province
Who is ahead on tomorrow's temperature.
- A Coruña Atlas Forecast, no clear lead
- Álava Atlas Forecast, no clear lead
- Albacete Atlas Forecast, no clear lead
- Alicante Google AI (WeatherNext 2), no clear lead
- Almería ICON, with a clear lead
- Asturias Atlas Forecast, with a clear lead
- Ávila Atlas Forecast, no clear lead
- Badajoz Atlas Forecast, no clear lead
- Balearic Islands Google AI (WeatherNext 2), with a clear lead
- Barcelona Atlas Forecast, no clear lead
- Biscay ICON, no clear lead
- Burgos Atlas Forecast, with a clear lead
- Cáceres Atlas Forecast, with a clear lead
- Cádiz Model median, with a clear lead
- Cantabria Atlas Forecast, with a clear lead
- Castellón Atlas Forecast, no clear lead
- Ceuta no data yet
- Ciudad Real Atlas Forecast, no clear lead
- Córdoba Atlas Forecast, with a clear lead
- Cuenca Atlas Forecast, no clear lead
- Gipuzkoa Atlas Forecast, no clear lead
- Girona Atlas Forecast, with a clear lead
- Granada Google AI (WeatherNext 2), no clear lead
- Guadalajara Atlas Forecast, with a clear lead
- Huelva Atlas Forecast, with a clear lead
- Huesca Atlas Forecast, no clear lead
- Jaén Atlas Forecast, with a clear lead
- La Rioja Atlas Forecast, no clear lead
- Las Palmas ICON, no clear lead
- León Atlas Forecast, no clear lead
- Lleida Atlas Forecast, no clear lead
- Lugo Atlas Forecast, with a clear lead
- Madrid Atlas Forecast, with a clear lead
- Málaga ICON, no clear lead
- Melilla Model median, no clear lead
- Murcia Atlas Forecast, with a clear lead
- Navarre Atlas Forecast, no clear lead
- Ourense Atlas Forecast, with a clear lead
- Palencia Atlas Forecast, no clear lead
- Pontevedra Atlas Forecast, no clear lead
- Salamanca Atlas Forecast, with a clear lead
- Santa Cruz de Tenerife Atlas Forecast, no clear lead
- Segovia Atlas Forecast, no clear lead
- Seville Atlas Forecast, with a clear lead
- Soria Atlas Forecast, with a clear lead
- Tarragona Atlas Forecast, no clear lead
- Teruel Atlas Forecast, no clear lead
- Toledo Atlas Forecast, no clear lead
- Valencia Atlas Forecast, no clear lead
- Valladolid Atlas Forecast, no clear lead
- Zamora Atlas Forecast, with a clear lead
- Zaragoza Atlas Forecast, no clear lead
All of Spain
Last 30 days (11 October to 9 November) · 524 stations from AEMET and Meteocat (XEMA)
Temperature
For today (0–24 h)
Across Spain, Météo-France's ARPEGE is the most accurate this month: its average error on today's temperature is 1.2 °C, and its lead over the second is not down to chance.
- 1ARPEGE (Météo-France)1.2 °C
- 2Model median1.3 °C
- 3ICON1.3 °C
- 4ECMWF IFS1.5 °C
- 5UKMO (Met Office)1.5 °C
- 6ECMWF AIFS (AI)1.6 °C
- 7GFS1.8 °C
- 8Google AI (WeatherNext 2) (estimated)1.9 °C
519 stations · 30 days
For tomorrow (24–48 h)
Across Spain, the model median, Météo-France's ARPEGE and ICON do about equally well on tomorrow's temperature (the first has an average error of 1.3 °C): the difference between them could be chance.
- 1Model median1.3 °C
- 2ARPEGE (Météo-France)1.3 °C
- 3ICON1.3 °C
- 4ECMWF IFS1.6 °C
- 5UKMO (Met Office)1.6 °C
- 6ECMWF AIFS (AI)1.7 °C
- 7GFS1.8 °C
- 8Google AI (WeatherNext 2) (estimated)1.9 °C
519 stations · 30 days
Average error in the temperature for each hour, at the stations in the area. The shorter the bar, the better. Highlighted: the leaders, where the difference between them cannot be put down to chance.
Wind
For today (0–24 h)
Across Spain, ICON and the model median do about equally well on today's wind (the first has an average error of 3.6 km/h): the difference between them could be chance.
- 1ICON3.6 km/h
- 2Model median3.7 km/h
- 3UKMO (Met Office)4.0 km/h
- 4ECMWF IFS4.1 km/h
- 5ECMWF AIFS (AI)4.2 km/h
- 6GFS4.6 km/h
395 stations · 30 days
For tomorrow (24–48 h)
Across Spain, the model median and ICON do about equally well on tomorrow's wind (the first has an average error of 3.7 km/h): the difference between them could be chance.
- 1Model median3.7 km/h
- 2ICON3.7 km/h
- 3ECMWF IFS4.2 km/h
- 4UKMO (Met Office)4.2 km/h
- 5ECMWF AIFS (AI)4.2 km/h
- 6GFS4.6 km/h
395 stations · 30 days
Rain
For today (0–24 h)
Across Spain, the model median is the most accurate this month: its average error on today's rain is 1.1 l/m² per day, and its lead over the second is not down to chance.
- 1Model median1.1 l/m² per day
- 2UKMO (Met Office)1.1 l/m² per day
- 3ECMWF AIFS (AI)1.3 l/m² per day
- 4ARPEGE (Météo-France)1.3 l/m² per day
- 5ICON1.3 l/m² per day
- 6GFS1.3 l/m² per day
- 7ECMWF IFS1.3 l/m² per day
518 stations · 30 days
For tomorrow (24–48 h)
Across Spain, the model median and ICON do about equally well on tomorrow's rain (the first has an average error of 1.3 l/m² per day): the difference between them could be chance.
- 1Model median1.3 l/m² per day
- 2ICON1.4 l/m² per day
- 3GFS1.5 l/m² per day
- 4ARPEGE (Météo-France)1.5 l/m² per day
- 5ECMWF AIFS (AI)1.5 l/m² per day
- 6UKMO (Met Office)1.5 l/m² per day
- 7ECMWF IFS1.7 l/m² per day
518 stations · 30 days
Error in the total rain for each day, in litres per square metre.
Google's AI (WeatherNext 2) is now in the league. As we only compare it every 6 hours (that is how often it runs) and measure it at one station per province, its error is an estimate: the model median's error plus how much more or less Google's AI misses than the median at those hours.
How it is calculated
We compare what each model had forecast for the same day (0–24 h) and for the next day (24–48 h) with what the Meteocat (in Catalonia) and AEMET weather stations measured over the last 30 days. They all compete over the same hours. Being first is not enough to win: the difference has to pass a day-by-day statistical test (otherwise it could be chance) and at least 14 days are needed.
Winning this month does not mean winning every month: every model has its good and bad days. More detail in Accuracy and, for your exact spot, in «Which model is right here?» in the dashboard.