You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 6 Next »

On Tuesday 1 July 2025, a first version of the probabilistic model of the Artificial Intelligence Forecasting System (AIFS) will be released and supported operationally. The model version is AIFS ENS v1, which will replace the current experimental model AIFS ENS DIFF. 

AIFS ENS data has undergone internal analysis, demonstrating significant skill and highlighting its transformative impact. We encourage users to read this webpage, which outlines outstanding known issues, before considering using the data for operational use/integration in applications. We would welcome any feedback.

Please note that this release does not impact users of the IFS and AIFS Single models in any way. The current operational version of IFS was successfully implemented on 12 November 2024. The current operational version of AIFS Single was successfully implemented on 25 February 2025.

Join us for the webinar introducing the AIFS ENS model on 9AM UTC! Read more and register here backhand index pointing right  https://events.ecmwf.int/event/487/


Description of the model

The AIFS ENS v1 model is a probabilistic weather forecasting system developed by ECMWF that uses machine learning to generate ensemble forecasts. It produces multiple forecast members by sampling from a learned distribution, capturing the uncertainty in future weather conditions. The model is trained using a version of the Continuous Ranked Probability Score (CRPS), a loss function that helps ensure the forecasts are both accurate and well-calibrated. This training approach accounts for the limitations of using a finite number of ensemble members. AIFS ENS outperforms ECMWF’s traditional physics-based ensemble system in medium-range forecasts and performs competitively for subseasonal forecasts when evaluated as anomalies.

New Methodology

The experimental AIFS ENS model available on ecCharts used a different methodology to AIFS ENS v1 (the experimental version uses diffusion, whereas the operational model optimises the CRPS). Therefore this implementation should not be seen as an update that can be compared with the previous experimental version, but instead it should be seen as the first version of a new model.

Control member

While the AIFS ENS includes a dedicated control member, this has a different interpretation to that of a physics-based ensemble (i.e. the IFS ENS). In traditional physics-based ensembles (like the IFS ensemble), a control member is a deterministic forecast run at high resolution without perturbations. It's often used as a baseline for comparison.

In contrast, AIFS ENS is a stochastic machine learning model that generates ensemble members by sampling from its learned internal distribution. The uncertainty within the AIFS ENS cannot be switched off to run a control run.

In AIFS ensemble the control member is a member of the ensemble that has been initialised using the same initial condition as the IFS control. Therefore it has unperturbed initial conditions but is not a true control member as the model is perturbed.

Timeline of the implementation 

 AIFS ENS will be implemented on the 1 July 2025  06 UTC run.

A short Release Candidate testing Phase (RCP) will start on 23 June 2025.

Datasets affected

This upgrade impacts the ensemble AIFS forecast data set.

Resolution


Component

Horizontal resolution

Vertical resolution
[pressure levels] 

Atmosphere

AIFS ENS

N320

~32 km

13*

*Levels (hPa) 50, 100, 150, 200, 250, 300, 400, 500, 600, 700, 850, 925, 1000

Dissemination schedule 

Schedule for dissemination

  • AIFS data will be available according to the same schedule as IFS data. See Dissemination schedule for further details.
  • Pre-schedule delivery will be available for eligible users, making the data available as soon as they are produced. 

Schedule for open data

  • The delay on open data platforms will not be implemented for AIFS ENS data, making the data available as soon as they are produced. 

Meteorological content 

The variables provided by AIFS ENS are the same as those in AIFS-Single in the atmosphere. At the surface they are largely the same, but AIFS ENS will not provide some precipitation and land variables (see table below for full list).

Tropical cyclone tracks will be available only as graphical products in the first period. Raw BUFR files will be available at the later stage. 

Assimilation

  • AIFS ENS uses the operational IFS ensemble initial conditions, regridded to an N320 grid. In other words, member n of AIFS ENS uses the same initial condition as member n of the IFS ensemble and the control of AIFS ENS uses the same initial condition as the control of the IFS ensemble.

Observations

  • No observations are used to train AIFS ENS.

Model

Fields at step 0

There will be step 0 for all the parameters, which should streamline user workflows.

  • For accumulated fields time step 0 will have a value of 0,
  • For instantaneous fields, these come from the IFS analysis, regridded to an N320 grid. 

Evaluation 

Interactive scorecards presenting the new model performance across period February 2024 to February 2025 have been updated: 

Known issues

Known issues with the AIFS ENS forecast can be found on the following page: Known AIFS ENS Forecasting Issues.
This page will be updated as issues are investigated and resolved.

Key configuration values

A new MARS keyword - model - has been introduced to support AIFS Single v1, and is mandatory to use with AIFS ENS v1 as well. This keyword distinguishes between different AIFS models and will need to be specified in MARS requests for historical data and dissemination requests for real-time data.



AIFS ENS



v1

Basetime & frequency

00/06/12/18 daily

Forecast range


15-days

MARS keywords

Class

ai

Stream

enfo

Model

aifs-ens

Type

cf/pf

Spectral

N/A

Gaussian grid

n320

Horizontal grid resolution

~36 km

Dissemination (LL)

0.25° 

Model Level vertical resolution

13

Stay in the loop!

Want to get updates about future AIFS cycle upgrades?

Join the mailing list

To subscribe or unsubscribe, please send an email to forecast_changes-request@lists.ecmwf.int with either subscribe or unsubscribe as Subject.

Join our FORUM

https://forum.ecmwf.int/ and 'watch' the announcements in IFS, AIFS and OpenIFS category.

Follow the LinkedIn channel for users: world ECMWF Users LinkedIn

Contents of this page

Input and output parameters

The table below shows the parameters that will be input and output by AIFS ENS v1. 

Control and perturbed forecast products

FieldLevel typeInput/Output

Geopotential (z)
Specific humidity (q)
Temperature (t)
U component of wind (u)
V component of wind (v)
Vertical velocity (w)

Pressure level: 50, 100, 150, 200, 250, 300, 400, 500, 600, 700, 850, 925, 1000Both ("Prognostic")
2 metre dewpoint temperature (2d)
2 metre temperature (2t)
10 metre U wind component (10u)
10 metre V wind component (10v)
Mean sea level pressure (msl)
Skin temperature (skt)
Surface pressure (sp)
Total column water (tcw)		SurfaceBoth ("Prognostic")
 Soil temperature (sot), at solid depth 1 and 2Soil layer levelBoth ("Prognostic")

Total precipitation (tp)
100 metre U wind component (100u)
100 metre V wind component (100v)
Surface short-wave (solar) radiation downwards (ssrd)
Surface long-wave (thermal) radiation downwards (strd)
Cloud variables (tcc, hcc, mcc, lcc),
Runoff water equivalent (surface plus subsurface) (rowe)
Snowfall water equivalent (sf)

SurfaceOutput ("Diagnostic")

Standard deviation of sub-gridscale orography (sdor)
Slope of sub-gridscale orography (slor)

Land-sea mask (lsm), orography, insolation, latitude/longitude, time of day/day of year 

SurfaceInput ("Forcings")

 Post-processed products

FieldStatisticLevel type
2 metre temperature
10 metre wind speed
100 metre wind speed
Mean sea level pressure		Ensemble mean and standard deviationSurface
Geopotential, temperature, wind speedEnsemble mean and standard deviationPressure level: 250, 300, 500, 850, 1000
2 metre temperature less than 273.15 K
Total precipitation of less than 0.1 mm
10 metre Wind speed of at least 10 m/s and 15 m/s
Total precipitation of at least 1 mm/5 mm/10 mm/20 mm/25 mm/50 mm/100 mm
Total precipitation rate less than 1 mm/day
Total precipitation rate of at least 3 mm/day
Total precipitation rate of at least 5 mm/day		ProbabilitiesSurface


More detailed information about the post processed parameters in AIFS ENS v1 is provided in the table below.

Ensemble mean and (type=em) standard deviation (type=es)

Param IDShort NameNameUnitsGRIB editionLevel TypeComments
129zGeopotential

m2 s-2

2Pressure level
130tTemperatureK2Pressure level
10wsWind speed

m s-1

2Pressure level
 167 2t2 metre temperature

K

2Surface
 207 10si 10 metre wind speed

m s-1

2Surface
 228249 100si 100 metre wind speed

m s-1

2Surface
 151msl  Mean sea level pressure

Pa

2Surface

Probabilities (type=ep)

Param IDShort NameNameUnitsGRIB editionLevel TypeComments
1310732tl2732 metre temperature less than 273.15 K%2Surface
13106810spg1010 metre Wind speed of at least 10 m/s%2Surface
13106910spg1510 metre Wind speed of at least 15 m/s%2Surface
131060tpg1Total precipitation of at least 1 mm%2Surface
131061tpg5Total precipitation of at least 5 mm%2Surface
131062tpg10Total precipitation of at least 10 mm%2Surface
131063tpg20Total precipitation of at least 20 mm%2Surface
131098

tpg25

Total precipitation of at least 25 mm

%2Surface
131099tpg50Total precipitation of at least 50 mm%2Surface
131085tpg100Total precipitation of at least 100 mm%2Surface
131065tprl1Total precipitation rate less than 1 mm/day%2Surface
131066

tprg3

Total precipitation rate of at least 3 mm/day%2Surface
131067tprg5Total precipitation rate of at least 5 mm/day%2Surface


Technical content

GRIB encoding

Software 

To handle the data of AIFS ENS v1, we recommend use of the ECMWF software packages:
ecCodes 2.40.0
CodesUI 1.8.0 (minimum version 1.7.3)
Magics 4.15.4 (minimum version 4.13.0)
Metview 5.22.1
ecmwf-opendata 0.3.19 (model keyword should explicitly be provided as aifs-ens)

On the ATOS HPC these versions correspond to ecmwf-toolbox/2025.02.0.0.

Older versions of eccodes and the ecmwf-toolbox will still work in terms of reading the data from AIFS ENS v1. 

Availability of AIFS ENS v1 test data

Please note that test data will not be available via ecCharts or Open Charts. Graphical products from AIFS ENS v1 will be available to the user community on implementation day (1 July 2025).

Test data in MARS

Test data is available from 01 December 2024 12z run in the MARS archive with experiment version 103. Use the MARS keywords expver=103, class=ai and model=aifs-ens to retrieve this data. Please note that only users registered with access to MARS are able to access these test datasets. The test data must not be used for operational forecasting. Please report any problems you find with this data via the ECMWF Support Portal.

Test data in dissemination

Test data will be made available on Monday 23 June.

Dissemination file naming convention 

The test data file names end with '0103', corresponding to the experiment version of the test data (see File naming convention and format for real-time data#Naming-AIFSviadissemination for further details about the AIFS file naming convention).

Dissemination requests

A new MARS keyword "model" is required in dissemination requests with class=ai. The model keyword identifies the AI model that was used to produce the forecast. For AIFS ENS v1, the model keyword is "aifs-ens". 

Test data on ECMWF's Open Data platform

Test data on Open Data Platform will be made available soon, before AIFS ENS v1 goes operational on Tuesday, July 1, 2025.

Resources

Webinar

Webinar presenting AIFS-ENS will be held on 10AM BST (11AM CEST). You can register here: https://events.ecmwf.int/event/487/

References


  • No labels