Opened 11 months ago

Last modified 11 months ago

#333 accepted Defect

CDS download for ERA5 sdor,cvh,cvl,fsor sometimes fails

Reported by: terhardt Owned by: anphi
Priority: major Milestone: flex_extract_v7.1.3
Component: flex_extract Version: flex_extract_v7.1.2
Keywords: CDS ERA5 Cc:

Description

I ran into the issue, that flex_extract downloads for ERA5 would sometimes fail for the sdor, cvh, cvl and fsor files with a data availability error.

After reaching out to the ECMWF support I was told that the retrieval with either the codes or short names can lead to unpredictable results and that the retrieval with the long parameter names as per the tables on the ERA 5 documentation (https://confluence.ecmwf.int/display/CKB/ERA5%3A+data+documentation) was the only supported way.

For now I have just disable the conversion to codes for this retrieval and it seems to be okay, at least for the data that I am still missing.

I have attached a simple script that illustrates the problem outside of flex_extract

Attachments (1)

access_examples.py (1.3 KB) - added by terhardt 11 months ago.
Example script

Download all attachments as: .zip

Change History (5)

Changed 11 months ago by terhardt

Example script

comment:1 Changed 11 months ago by pesei

  • Owner set to anphi
  • Status changed from new to assigned

Interesting observation, thank you. Did ECMWF Support say anything whether this is CDS-specific or would also apply to regular MARS retrievals? We have never encountered the issue with the latter method (and don't use CDS except for testing).

comment:2 Changed 11 months ago by terhardt

I have only specifically asked about CDS, so it's not clear on how this affects regular MARS retrievals.

In the meantime I have also run into the same issue with retrievals for other variables in flex_extract. So to solve this a general solution for data accessed via the CDS api might be needed.

comment:3 Changed 11 months ago by anphi

  • Status changed from assigned to accepted

Thank you for reporting this issue. It is new to me that long names are necessary when using CDS API. The ERA5 analysis should have consistent names in shortName and parameterID for the whole period. I need some time to investigate and apply a fix for this.
Please report also the other variables which caused problems.

Thank you.

comment:4 Changed 11 months ago by terhardt

This also seems to happen sometimes for the other surface parameters. I haven't observed this yet with the model level parameters.

I am not sure wether that is connected, but incidentally the model level data is not available through the Website, only through the API.

Note: See TracTickets for help on using tickets.
hosted by ZAMG