Recipe: recipe_seaice_drift failing

[rbeucher]

Please check log in action workflow

[rbeucher]

Recipe fails because we need the NSDIC datasets.
They need to be formatted from RAWOBS into OBS6 datasets
PB is this will only work with PROJ 9.3 or greater.

esmvaltool data format NSIDC-0116-nh
esmvaltool data format NSIDC-0116-sh

@flicj191 when you have a chance can you try to create the OBS6 datasets with your dev environment?
Should be an easy fix

[rbeucher]

@flicj191 can you check if the dataset is in CT11?

[flicj191]

no its not, its a Tier3 dataset, no Tier3's in ct11

[rbeucher]

Is it in Kj13 then?

[flicj191]

yes kj13

[flicj191]

I got it failing on the second script, first script and preprocessing went through fine once I removed the dates on the area variable datasets.

areacello:
        mip: fx
        preprocessor: extract
        additional_datasets:
          - {dataset: NSIDC-0116-nh, project: OBS6, type: reanaly, version: 4.1,
             tier: 3} #, start_year: 1979, end_year: 2005}
          - {dataset: OSI-450-nh, project: OBS, type: reanaly, version: v2,
             tier: 2} #, start_year: 1979, end_year: 2005}
          - {dataset: PIOMAS, project: OBS, type: reanaly, version: 2.1,
             tier: 2} #, start_year: 1979, end_year: 2005}

though still possibly a projection problem..
shapely.errors.GEOSException: IllegalArgumentException: CGAlgorithmsDD::orientationIndex encountered NaN/Inf numbers

[flicj191]

same as this issue no longer runs: ESMValGroup/ESMValTool#3243
looks like the transformer switches xy so was giving inf numbers. see warning on this page: https://pyproj4.github.io/pyproj/stable/api/transformer.html
currently testing with
self.transformer = Transformer.from_crs("WGS84","North_Pole_Stereographic",always_xy=True)

[rbeucher]

Any luck?

[flicj191]

PR for one that add: ESMValGroup/ESMValTool#3531
Looks like the bot that runs it can't find the data where they are

[rbeucher]

Levante does not have the data?
Gadi has, right @flicj191 ?

[flicj191]

Yep, ran fine for me