ERA5 CDS data corruption

We have found that some ERA5 data from the Climate Data Store (CDS) has become corrupted.

The corrupted data are in the following ERA5 CDS catalogue entries (datasets):

A page listing the details of the affected fields has been produced -

361 corrupted fields (out of a total of 3.1 billion ERA5 fields) have been identified and replaced.

This corruption issue was resolved on 2021-04-14.

Users that have downloaded any of these affected fields prior to 2021-04-15, are recommended to re-download them.

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

Please note, this issue does NOT affect:

  • ERA5 data downloaded from the MARS archive (such as CDS API requests using 'reanalysis-era5-complete')
  • ERA5 Monthly CDS catalogue entries (datasets);  these were derived directly from the ERA5 data on MARS, so are unaffected.

There is no need to re-download these data.

We apologise for any inconvenience this may have caused you.


ECMWF Support

Is there a date, prior to which the data was NOT corrupt?  In other words, if I want to guarantee that I do not have corrupt data, do I need to re-download any affected datasets that were originally downloaded any time after the initial release of ERA5?  I hope that's not the case.  But if users are interpretting the announcement that way, there is going to be a massive surge of requests coming.

This is difficult to find out, unfortunately. We know date stamps of the affected files, but we cannot guarantee that the corruption didn't occur earlier.

Additional ERA5 fields were found to be corrupted and they are now corrected. Please see this page for details:

ERA5 CDS data corruption - additional data corrected on 21 Jul 2021

Michela Giusti I found a corrupted file for AgERA5, specifically for Min temperature (-70 Celcius). It seems that is the issue of ERA5 since other values in time-series seem to be reasonable! Where can I report the bug and upload the clipped netCDF file (size 36 KB)?

Thank you!

Dear Lukas. The disk corruption exposes itself as latitude lines of constant values. Is that what you see?

No... That's not it. This error is related to minimum temperature values (-70 Celcius record for several grid cells). Hmm, where should I post these wrong values?


Hi Lukas Tuma

could you please open a query and attach the data file and the script used to retrieve the data: