Summary sheet
AIFS 2 against IFS and GFS
At 30 DWD stations in Germany, over 174 days, AIFS 2 had a lower error than IFS or GFS in 17 of 28 tested comparisons, a higher error in none, and could not be separated in 11.Error is RMSE against station measurements: 2 m temperature, 10 m wind and solar energy, lead days 1, 3, 5, 7 and 10, 23 May to 12 November 2026. Each comparison was tested with a block bootstrap and a Holm correction.
ERA5 over Europe
a reanalysis, not a measurementNone higher, 4 not separated, 22 without a verdict.
2 m temperature
Germany, AIFS's RMSE against each model, in %Against IFS: 10 to 13% lower (better) at days 1, 3 and 5; not separated at days 7 and 10. Against GFS: 24 to 29% lower (better) at days 1, 3 and 5; not separated at days 7 and 10.
10 m wind
Germany, AIFS's RMSE against each model, in %Against IFS: 7% lower (better) at day 5; not separated at days 1, 3, 7 and 10. Against GFS: 8 to 10% lower (better) at days 3, 5 and 7; not separated at days 1 and 10.
Watch: forecast too rarely
10 m wind of 8 m/s or more, Germany, day 1AIFS forecast it less often than it happened: frequency bias 0.47 (95% interval 0.31 to 0.61; 1 would be right). IFS forecast it 176 times and GFS forecast it 311 times.
A low RMSE and too few forecast events can go together: smooth forecasts score well on RMSE.
Since the last update
12 November 2026 → 16 November 2026, AIFS against IFS at day 3- Germany, 2 m temperature: −12.3% → −13.2% (lower, tested, unchanged)
- Germany, 10 m wind: −2.9% → −3.0% (not separated, unchanged)
- ERA5, 500 hPa height: −8.2% → −7.9% (lower, tested, unchanged)
- ERA5, 2 m temperature: −28.7% → −28.6% (no verdict, unchanged)
Station sample: 174 days, 4 more than before. Minus is a lower AIFS error (better), plus a higher one (worse).
Sample: Summer and autumn 2026. Provisional: no full year yet, winter and spring still to come. Green and red mark tested differences only. The checks are not compared with each other, and Germany is counted again inside Europe. Built from the results files; no sentence on this sheet is typed by hand.
Sources, method and licence
How the verdicts are reached
Each score is a root-mean-square error (RMSE) against measurements or ERA5; lower is better. Every comparison is paired: AIFS against one other model at the same times and places. A difference counts only if its 95% interval excludes zero, its Holm-adjusted p-value is below 0.01 (a level set low because the verdicts are tested again every week), the sample holds at least 15 blocks, and the difference has the same sign in the first and second half of the period. It must also hold when the models are read at the nearest land point instead of by interpolation, and for temperature when the height correction of 6.5 K/km is changed. Intervals come from a stationary block bootstrap over forecast dates with 20,000 resamples and blocks of at least 10 days [1, 2]. Event scores follow [3].
Data version
Data version 16 November 2026, run 20261116T194902Z, code 048c5a9. Germany and Europe through 12 November 2026, Cabauw through 12 November 2026, ERA5 through 9 November 2026. AIFS 2, raw forecasts. The results change with each update.
Method references
- Holm, S. (1979). A simple sequentially rejective multiple test procedure. Scandinavian Journal of Statistics, 6(2), 65–70.
- Politis, D. N. and Romano, J. P. (1994). The stationary bootstrap. Journal of the American Statistical Association, 89(428), 1303–1313. doi:10.1080/01621459.1994.10476870
- Jolliffe, I. T. and Stephenson, D. B., eds. (2012). Forecast Verification: A Practitioner's Guide in Atmospheric Science, 2nd ed. Wiley. doi:10.1002/9781119960003
About
MeteoSignal is independent. It is not affiliated with or endorsed by any organisation whose models or data appear here. Analysis, code and site by Iman Soltanzadeh. Errors and questions: meteosignal.com/contact/. Impressum: meteosignal.com/impressum/.
Sources behind this sheet
- AIFS and IFS forecasts at the stations and for the area section: © 2025–2026 European Centre for Medium-Range Weather Forecasts (ECMWF), www.ecmwf.int, published under CC BY 4.0; decoded and reduced to station and area values of wind speed, temperature and solar energy. ECMWF does not accept any liability whatsoever for any error or omission in the data, their availability, or for any loss or damage arising from their use. AIFS is described in Lang et al. 2024, arXiv:2406.01465.
- GFS forecasts: NOAA/NCEP, from the NOAA Open Data Dissemination archive on AWS, public domain. NOAA does not endorse this site.
- Germany: Source: Deutscher Wetterdienst, 10-minute station observations of wind, of solar and sunshine, and of air temperature for Germany, Version v24.03, CC BY 4.0. Observations from January 2026 onwards are provisional.
- Austria, Denmark, the Netherlands, Switzerland and Finland: GeoSphere Austria, CC BY 4.0. Danish Meteorological Institute (DMI), CC BY 4.0; its 10-minute records were reduced to hourly wind and temperature and daily solar energy. Royal Netherlands Meteorological Institute (KNMI), 10-minute in-situ observations and hourly station data, CC BY 4.0; reduced to hourly wind and temperature and daily solar energy. Source: MeteoSwiss, CC BY 4.0. Finnish Meteorological Institute, CC BY 4.0.
- Cabauw tower: KNMI, CESAR Cabauw tower profiles, validated (cesar_tower_meteo_lb1_t10, v1.3 and v1.2) and unvalidated (cesar_tower_meteo_la1_t10, v1.3), CC BY 4.0.
- ERA5: Contains modified Copernicus Climate Change Service data (ERA5 reanalysis, the latest months preliminary), CC BY 4.0. Neither the European Commission nor ECMWF is responsible for any use made of this information. Described in Hersbach et al. 2020, doi:10.1002/qj.3803. Climatology for the area scores: WeatherBench 2, 1990 to 2019, itself derived from ERA5 (Rasp et al. 2024, doi:10.1029/2023MS004019).
The method page, meteosignal.com/method/, lists every source the site uses.
Licence
cc by 4.0Results and text, this sheet included, are licensed under Creative Commons Attribution 4.0 International. You may copy, share and adapt them, also commercially, if you give credit: "MeteoSignal, Iman Soltanzadeh, meteosignal.com, CC BY 4.0", with a note if you changed anything. The code is under the MIT licence. The source data keep their own terms, listed here.
How to cite
Soltanzadeh, I. (2026). MeteoSignal: verification of open weather forecasts over Europe. Data version 16 November 2026, run 20261116T194902Z. https://meteosignal.com/
Please name the data version, because the results change with each update.
© 2026 Iman Soltanzadeh. Built from the results files of data version 16 November 2026; no sentence on this sheet is typed by hand, apart from the source credits and references, which are those of the method page.