Final validated ERA5 product to differ from ERA5T for Sep-Dec 2021

This announcement is to the attention of ERA5 users who may have downloaded preliminary data (ERA5T) from September 2021 onwards.

Following continuous monitoring by the ECMWF Reanalysis team, an issue in ERA5T was identified as needing correction due to the assimilation of anomalous snow depth observations over Central Asia.

As a result of this, for September, October, November and at least part of December 2021, the final ERA5 product will differ from ERA5T. Although the affected area of ERA5T is limited to a few relatively small areas and affects mainly surface parameters such as snow depth, soil moisture and to a lesser extent 2-metre temperature and humidity, the results of the final, corrected, release (where the anomalous station observations were now withdrawn) will vary slightly everywhere around the globe for most parameters.

The locally anomalous effect on soil moisture, 2-metre temperature and humidity for October 2021 can be seen from the corresponding C3S monthly bulletin report, as a severe wet/cold anomaly northwest of the Caspian Sea. For other months the effect is milder.

For other locations around the globe differences should typically be within the ERA5 uncertainty estimate, i.e., the local ERA5 Ensemble spread as also provided in the ERA5 catalogue entries.

The effect on larger areas is very small. For instance for September 2021 the ERA5 estimate for global mean two-metre temperature  remained the same within 0.001 deg C, while  it changed from 14.516 to 14.515 deg C for the European mean.

After the final ERA5 data has been released there is no more access to the corresponding preliminary ERA5T data on the Climate Data Store (CDS) online disks.

In practical terms, this means that:

  • If you have accessed ERA5 data for 2 September 2021 or later, this may not correspond to the final validated version. If you think the local effects over Central Asia affect your application we strongly recommend you to re-access the data once the final validated dataset has been released. For September 2021 such data has now been released. For October and later months we would ask you to wait until their final releases are available.
  • To know whether you have accessed ERA5T data or the consolidated set you can look at the value of expver in your data files. It's 0005 for ERA5T and 0001 for the final validated release.
  • The issue was resolved for ERA5T data from 15 November 2021. However, the final validated version is being produced in a parallel stream and as long as it hasn't caught up with ERA5T, both will differ. We will update this announcement when this occurs and both streams have been aligned again (for that date ERA5T will get its initial conditions from the corrected stream rather than from its own first guess; this to ensure a seamless final product). After that date (assuming that no other issues are found) ERA5T will correspond to the final validated version again.

We apologize for the inconvenience this may cause. This is the first time since the introduction of the ERA5T service in December 2019 that it will be temporarily superseded.

This applies to the following ERA5 catalogue entries on the CDS:

ERA5 data on ECMWF MARS tapes accessible via CDS API (i.e. 'reanalysis-era5-complete') had not been made available yet from September 2021 and were therefore not overwritten.

ERA5-Land continues to be based on the final version of ERA5. There is also no need to re-access such data.

In addition, CDS datasets that rely on the ERA5T service may be updated in due time. For most applications the impact is expected to be very small. Please follow the notifications on such datasets carefully.

For reference, this issue has been recorded as a "Known issue" in the ERA5: data documentation.

If you have any questions, you may contact us.

Best regards, 

ECMWF Support

Does it mean that the expver  field will disappear soon from the files? Asking to know if we have to update our production scripts to deal with this field (for example with the surface solar radiation for November 2021) or if we should wait for the finalised version.