Using CMIP6 data for ocean diagnositcs

[RChapman1407]

Hello,

I am having difficulty getting one of the ocean diagnostic scripts to work with CMIP6. So far it works fine with CMIP5. The recipe is recipe_ocean_example.yml. I am trying to average some ocean salinities over defined volumes before extracting them to use in a conceptual model.
Should I be giving a different route to the datasets? I am currently using the same structure of path as the predefined datasets, and replacing the details for the data I want. I am working within JASMIN.
I have already tried giving the grid as well, but this did not work. I currently get the error 'Failed to run preprocessor function 'volume_statistics' on the data' and in the trace the comment 'Multi-dimensional coordinate not supported: 'lataitude''

Thanks

[zklaus]

@ledm, since you have worked quite a bit with CMIP6 ocean data, could you lend a hand here?

[ledm]

Hi @RChapman1407, this cryptic error message means that ESMValTool can't find the FX files, volcello. This will need to be added to your preprocessor:

https://docs.esmvaltool.org/projects/ESMValCore/en/latest/recipe/preprocessor.html?highlight=fx#fx-variables-as-cell-measures-or-ancillary-variables

Frustratingly, different models put FX files in different places in CMIP6, and ESMValTool still struggles to cope with this. (ESMValGroup/ESMValCore#1138 or ESMValGroup/ESMValCore#1609 are solutions that haven't been merged yet), so in the short term, you may need to split your preprocessor into one for each kind of fx path.

[RChapman1407]

Hi @ledm

Thanks for your answer, I understand the problem now, certainly frustrating.
I have two quick follow up questions as I am stuggling to implement this from the documentation.

Where in the preprocessor should the fx_variables go? I have tried putting this within the volume_statistics as well as directly in the preprocessor but this hasn't worked so far.
And then, how should I find which fx path to give the preprocessor? In my case, I suspect this should be volcello: Omon as in the example in the documentation, but is there a correspondence between this and the badc path to the data?

[ledm]

I typically try to use this approach:

      prep_timeseries_2d:
        area_statistics:
          operator: mean
          fx_variables: [{short_name: areacello, mip: Ofx, }] 

This is for area, but the principle stands!

There's an extra level of complexity here with volcello, as some models have time-varying volume and some have fixed volume. This means that some of the volcello data is in Omon and some is in Ofx. You may need to check how each individual model does it.

For you, i imagine you'll need several different preprocessor, depending on each model set up. ie:

      prep_timeseries_3d_Omon:
        volume_statistics:
          operator: mean
          fx_variables: [{short_name: volcello, mip: Omon, }] 

      prep_timeseries_3d_Ofx:
        volume_statistics:
          operator: mean
          fx_variables: [{short_name: volcello, mip: Ofx, }] 

The BADC path to volcello is: /badc/cmip6/data/CMIP6/CMIP///piControl/Ofx

This may get you closer. Let me know if you hit another snag!

[RChapman1407]

Thank you, this seems to be working so far!