Esmvaltool: Error running recipe_Validation_CMIP6.yml

[glad94]

Hi, I'm new to ESMValTool and have been running into issues with the "recipe_Validation_CMIP6.yml" recipe. I'm trying to run the recipe without any observation files at the moment because I'm having trouble with downloading from ana4mips and running the cmorizer on raw data (but that's something I'm putting aside for now).

For this recipe test I've copied, modified and renamed the original recipe file to "recipe_Validation_CMIP6_test2.yml" and am using a config file called "config-user-test.yml". The log file I obtain from /esmvaltool_output/recipe_validation_CMIP6_test2_20191008_090010/run/validation/lat_lon/log.txt returns a "local variable 'control' referenced before assignment" error.

I've tried the recipe on another control model but obtained the same results. I'm guessing that I've messed up in excluding an observation file but I'm not sure where.

recipe_validation_CMIP6_test2.txt
main_log_debug.txt
log.txt

Will appreciate any help!

[valeriupredoi]

hello @glad94 and welcome to the ESMValTool gitHub page! Cheers for raising this up, this is something that is legacy to my old implementation when we were using only CMIP5 data and needs to change now to accomodate both CMIP5 and 6 - the problem is in /home/arcdap/anaconda3/envs/esmvaltool/lib/python3.7/site-packages/esmvaltool/diag_scripts/validation.py (or its source, since that one's the target installed one) - if you change _CMIP_TYPE = 'CMIP6' right at the top then you'll be able to run no problemo. I will get rid of the static variable and make it be picked up from the environment so ppl don't have to change this manually in the future :beer:

[glad94]

Hi @valeriupredoi , thanks for the quick response, made the change and the recipe works now for the models I used. Just a question on the default control model used in the recipe (UKESM1-0LL), the programme failed due to a Failed to parse unit "days since 1850-01-01-00-00-00" error which I suppose would be due to the time formatting used in the UKESM1-0LL file's "parent_time_units" key under its attributes.

Notice that the CMIP5 datafiles don't have the "parent_time_units" attributes, so maybe a legacy issue as well? Thanks again!

[valeriupredoi]

yes, that is a known issue with UKESM1 and it's under hammer here https://github.com/ESMValGroup/ESMValCore/pull/231 - we should be able to merge that PR soon and it'll work after :beer:

[glad94]

Good to hear that, thank you!

[valeriupredoi]

once #1364 gets merged Ill close this unless you have more q's :beer:

[glad94]

Sure thing, nothing else directly related to this for now!

[valeriupredoi]

closed via #1364