Test recipe_autoassess_landsurface_surfrad with CERES-EBAF_Surface updated dataset

[valeriupredoi]

We changed to that dataset when we moved to new names for obs4mips in #4321 but had to revert to older CERES-EBAF in in #4377 bc that dataset don't work with the metrics - we need to investigate that. Myself and Alistair, at some point, hopefully before 2.15.

[alistairsellar]

The only versions of CERES-EBAF that I can see in Obs4MIPs at https://metagrid.esgf-west.org/search are CERES-EBAF-4-2 and CERES-EBAF-4-2-1. As @bouweandela notes here, in these versions all the variables are part of the same dataset again (no CERES-EBAF_Surface).

Therefore, the reverted version of recipe_autoassess_landsurface_surfrad.yml following #4377 works OK with CERES-EBAF-4-2-1, and results are scientifically equivalent to v2.13.0.

[alistairsellar]

I'm not sure about best approach now that versions earlier than CERES-EBAF-4-2 are not available on obs4MIPs, given that Levante has an earlier version that needs recipe changes (for this recipe). Should I try to get hold of the earlier version for consistency with Levante data pool, or test all the recipes that use CERES-EBAF against 4-2-1?

[bouweandela]

I would recommend updating the recipe to the latest version, if possible.

The data you can find with the metagrid application is probably only data that is formatted according to CMIP6 standards and even then there appear to be differences bewteen the different index nodes (e.g. https://esgf-metagrid.cloud.dkrz.de shows different datasets from the American index node you used). I am not sure if obs4MIPs are planning to republish older data that was formatted according to the CMIP5 standard on the new ESGF infrastructure. I asked @hb326 who is on the obs4MIPs steering panel, but have not received an answer yet. See also ESMValGroup/ESMValCore#2711.