Datasets

A subset of ECMWF real-time forecast data from the IFS and AIFS models is made available to the public free of charge. Their use is governed by the Creative Commons CC-BY-4.0 licence and the ECMWF Terms of Use.

ECMWF is now running version 2 of its Artificial Intelligence Forecasting System (AIFS). The AIFS consists of a deterministic model, AIFS Single, and an ensemble model, AIFS ENS.

The deterministic model has been running operationally since 25 February 2025 and was upgraded from AIFS Single v1.1 to AIFS Single v2 on 12 May 2026. Further details can be found on the dedicated Implementation of AIFS Single v1 page.

These products are available to the African Center of Meteorological Application for Development (ACMAD) countries.

The products outlined below are disseminated via EUMETCast

Please refer to WMO Additional products  for the equivalent products available on a global domain via the ECMWF's DCPC FTP server.

This is an extension of the i4ql experiment covering the period 2025-07-27 to 2023-08-10

Examples

Control experiment (CTRL) for an assimilation test with near-real time TROPOMI total column CO data (exp=hmib) using the CAMS CY47R3 configuration. In CTRL, MOPITT TIR and IASI TCCO are assimilated, but TROPOMI TCCO data are passive.

Examples

Retrieval of total column CO analysis

Analysis experiment (ASSIM) testing the assimilation of near-real time TROPOMI total column CO data in the CAMS CY47R3 system for the period 6/7-31/12/2021 (in addition to already assimilated MOPITT TIR and IASI TCCO).

Examples

Retrieval of total column CO analysis

Control experiment for a series of SO2 assimilation tests for the 2019 Raikoke eruption with the CAMS system which assess the impact of assimilating SO2 layer height data (V3.1 FP_ILM). In BLexp (hhu5) ESA NRT SO2 data were assimilated.

Examples

Retrieval of total column SO2 forecast at step 24

Layer height experiment for a series of SO2 assimilation tests for the 2019 Raikoke eruption with the CAMS system which assess the impact of assimilating SO2 layer height data (V3.1 FP_ILM). In LH1.4 (hgz7) SO2 Layer Height data produced with FP_ILM (V3.1) were assimilated with SO2 background error standard deviation values of 1.4e-7 kg/kg and horizontal background error correlation length scale of 100 km.

Examples

Layer height experiment for a series of SO2 assimilation tests for the Raikoke eruption with the CAMS system which assess the impact of assimilating SO2 layer height data (V3.1 FP_ILM). In LH100 (hhtm) SO2 Layer Height data produced with FP_ILM (V3.1) were assimilated with SO2 background error standard deviation values of 1e-7 kg/kg and horizontal background error correlation length scale of 100 km.

Examples

Layer height experiment for a series of SO2 assimilation tests for the 2019 Raikoke eruption with the CAMS system which assess the impact of assimilating SO2 layer height data (V3.1 FP_ILM). In LH250 (hhtn) SO2 Layer Height data produced with FP_ILM (V3.1) were assimilated with SO2 background error standard deviation values of 1e-7 kg/kg and horizontal background error correlation length scale of 250 km.

Examples

Layer height experiment for a series of SO2 assimilation tests for the 2019 Raikoke eruption with the CAMS system which assess the impact of assimilating SO2 layer height data (V3.1 FP_ILM). In LH50 (hhbu) SO2 Layer Height data produced with FP_ILM (V3.1) were assimilated with SO2 background error standard deviation values of 1e-7 kg/kg and horizontal background error correlation length scale of 50 km.

Examples

Layer height experiment for a series of SO2 assimilation tests for the 2019 Raikoke eruption with the CAMS system which assess the impact of assimilating SO2 layer height data (V3.1 FP_ILM). In LHexp (hgze) SO2 Layer Height data produced with FP_ILM (V3.1) were assimilated with SO2 background error standard deviation values of 0.7e-7 kg/kg and horizontal background error correlation length scale of 100 km.

Examples

Probabilistic 15-day TCo1279L137 forecast with CY49R1.1. 8 ENS members. Control for the nudged experiments. 00/12UTC start from 1 December 2024 to 28 February 2025.

Examples

Probabilistic 15-day TCo1279L137 forecast with CY49R1.1. 8 ENS members. Control for the nudged experiments. 00/12UTC start from 1 July 2024 to 12 November 2024.

Examples

Probabilistic 15-day TCo1279L137 forecast with CY49R1.1. 8 members. Vorticity and virtual temperature below tropopause nudged to equivalent members from model level AIFS-CRPS ML model for total wavenumbers <21. 00/12UTC start from 1 July 2024 to 30 September 2024. Only perturbed forecasts should be analysed.

Examples

Probabilistic 15-day TCo1279L137 forecast with CY49R1.1. 8 members. Vorticity and virtual temperature below tropopause nudged to equivalent members from model level AIFS-CRPS ML model for total wavenumbers <21. 00/12UTC start from 1 December 2024 to 28 February 2025. Only perturbed forecasts should be analysed.

Examples

Probabilistic 15-day TCo1279L137 forecast with CY49R1.1. 8 members. Vorticity and virtual temperature below tropopause nudged to equivalent members from model level AIFS-CRPS ML model for total wavenumbers <21. 00/12UTC start from 1 October 2024 to 12 November 2024. Only perturbed forecasts should be analysed.

Examples

Four-month 15-member ensemble hind-casts initialised on 1st of October 1992-2020. Fully radiatively interactive ozone. produced by the hybrid linear ozone scheme. IFS cycle 47r1 used. Output for pressure-level and surface fields is 6-hourly.

Examples

Four-month 15-member ensemble hind-casts initialised on 1st of October 1992-2020. Ozone monthly-mean zonal-mean climatology used in radiation. IFS cycle 47r1 used. Output for pressure-level and surface fields is 6-hourly.

Examples

This experiment contains 20 years of re-forecasts with 9 start dates twice a week between 12 December and 12 January of the following year. ensemble size is 11 members.

Examples

This experiment contains 20 years of re-forecasts with 9 start dates twice a week between 12 December and 12 January of the following year. ensemble size is 11 members.

Examples

This experiment contains 20 years of re-forecasts with 9 start dates twice a week between 12 December and 12 January of the following year. ensemble size is 11 members.

Examples

This experiment contains 20 years of re-forecasts with 9 start dates twice a week between 12 December and 12 January of the following year. ensemble size is 11 members.

Examples

This experiment contains 20 years of re-forecasts with 9 start dates twice a week between 12 December and 12 January of the following year. ensemble size is 11 members.

Examples

This experiment contains 20 years of re-forecasts with 9 start dates twice a week between 12 December and 12 January of the following year. ensemble size is 11 members.

Examples