Migration of ECMWF forecast data to GRIB2 format: Static sample dataset available in MARS archive

Announcements
, ,
1

We are happy to announce that the first test dataset in full GRIB2 format is now available in MARS archive.

The dataset is provided so that you can start updating and testing your MARS retrieval requests and post-processing chains well before the switch with the implementation of the IFS Cycle 50r2.
Please forward this announcement to your colleagues who are using ECMWF forecast data.

Where to find the data

For users with access to the MARS archive, the data are archived as experiment expver=2250 and can be browsed in the MARS Catalogue and Archive Catalogue (for commercial users).
The dataset in MARS archive currently contains the analysis and forecast from the IFS Medium-range control (ex-HRES) and the ensemble, for both the atmospheric and the wave model. Post-processed, sub-seasonal and Atmospheric Composition (CAMS) sample data will follow later.

For users without access to the archive, data can be retrieved from the static dataset on the ECMWF open data portal: ECMWF MTG2 datasets . This is the same static dataset announced earlier, and will be updated to match the data in MARS in the coming weeks.

Users of the ECMWF HPC can find the static dataset in /ec/vol/mtg2_sample.

Timeline

The IFS Cycle 50r2 test period (e-suite) will start in Q4 2026, producing full GRIB2 alongside the operational IFS Cycle 50r1, and the implementation is planned for Q2 2027. We are planning for at least six months of parallel running.
Once IFS Cycle 50r2 becomes operational, the whole model output, through dissemination and MARS retrieval, will be GRIB2 only.
Data already archived in GRIB1 will remain in GRIB1 and will not be re-encoded.

We have prepared a documentation page that describes this test dataset in MARS as well as the Frequently Asked Question (FAQ) page:

More information can be found in the Migration to GRIB2 Confluence space: Migration to GRIB edition 2 - Information page.

You can also catch up with the planned changes by watching the webinar recording on YouTube: https://www.youtube.com/watch?v=RUV7jRp21T4.

We would welcome your feedback on the test data. For any questions and feedback, please contact us via the ECMWF Support Portal.

ECMWF Support

2

Thank you for the update. I see that the static MTG2 sample datasets are currently available on the Open Data portal.

Once the 6 month IFS Cycle 50r2 parallel test period (e-suite) begins in Q4 2026, will the daily parallel e-suite runs also be published continuously to the Open Data portal alongside the operational 50r1 data? Or will Open Data users only have access to the static samples until the official implementation in Q2 2027?

Thanks

Brian

3

Thank you for the heads-up.

I noticed a long time ago that EMCF tends to provide data in GRIB format. However, all my attempts to read these files were unsuccessful. I’m still using NetCDF. The form of multidimensional cubes fits well my requirements for climate data analytics.
With this new move to GRIBv2 you make me think that the era of NetCDF is over. Is that correct?

If so, what tools would you recommend to read GRIB data (in Windows + R environment)?

4

Hi Brian,

We plan to publish test data from e-suite on the Open Data portal alongside the operational 50r1 data during the months parallel test period as well.

We also plan to update the public static data set to match the one in MARS soon.

Kind regards,
Milana

5

Hi Alexei,

ECMWF must provide the data in GRIB format due to our WMO obligations.

The era of NetCDF is not over, we are involved in projects of harmonisation of GRIB and NetCDF, for example. However we don’t plan to provide the data in other formats than GRIB.

We provide tools to read and process GRIB data, such as ecCodes and EarthKit, however they are not compatible with Windows and R. ecCodes has command line and Python interface, while Earthkit is Python tool.