Handling of secondary realm for CMIP7 output

[matthew-mizielinski]

The following just came up for discussion in the DR TISG

The CMIP7 style tables are organised by realm and have a single realm entry available for each table.

Many variables in CMIP6 and CMIP7 have secondary realms as they relate to the interface between model components. An example is the CMIP6 variable Omon/sfdsi (Downward Sea Ice Basal Salt Flux), which had realm "ocean seaice". With the current system the secondary "seaice" realm reference is discarded and in the CMIP7 prototype tables this would appear in the CMIP7_ocean table via the variable sfdsi_tavg-u-hxy-sea.

I don't think with the CMIP7 style tables there is a means to include secondary realms.

Many members of the TISG commented that it would be valuable to retain the secondary realms in the output files -- is this something that could be added to the CMOR library when working with CMIP7?

Given that the realm is now used to organise the variables into tables is it possible to include a secondary realm field for variables that could be populated if additional realms are needed?

I don't think that extending the global attribute realm is problematic as table_id is being used as part of the DRS (I think).

@durack1 @JamesAnstey @taylor13 -- thoughts welcome, but it might be worth resolving this before v3.12 is released.

[durack1]

Thinking about this, it might be a good idea to get CMOR 3.12 released, and add this, and any other issues raised once 3.12 is available to a new 3.13 milestone (there's already a placeholder - see here).

In addition to code changes, an ambition of 3.13+ was to start to augment the documentation and example libraries in the repo (we now have numerous examples, however these are distributed outside of this/cmor repo), to aid users. It'd be great to get some help with this from users that are starting to engage with the CMIP7 prototype pieces.

It might be best that for ongoing CMOR development, we try and nimbly (and quickly) iterate with versions that make small changes, which means these changes can be implemented and tested in "production" and tweaks made if required.

If it wasn't obvious, we also have a CMOR nightly channel, with conda builds available at https://anaconda.org/PCMDI/cmor/files, which I have been testing across some of my examples before we get to a release marker - if folks are not testing these pre-release, probably a good idea to do so