[BUG] Can't load 'area_t' from ACCESS-OM2

[anton-seaice]

Describe the bug

The access-nri-intake catalog doesn't open the tracer area field from ACCESS-OM2 results. If I try and load the 'area_t' variable from '01deg_jra55v140_iaf_cycle4_jra55v150_extension' I get a failed to load error.

To Reproduce

Run:
catalog['01deg_jra55v140_iaf_cycle4_jra55v150_extension'].search(variable='area_t').to_dask()

Actual Result:
Failed to load dataset with key='ocean_2d_area_t.fx'

Expected Result:
Dataset with x and y dimensions, and area_t variable.

Additional context

I suspect this is because there are multiple files with the area_t variable, but they are all the same (i.e. it's a field for information/analysis, not explicitly model output).

i.e., these files cannot be concatenated (with default settings) by xarray because they are all the same
['/g/data/ik11/outputs/access-om2-01/01deg_jra55v140_iaf_cycle4_jra55v150_extension/output1000/ocean/ocean-2d-area_t.nc', '/g/data/ik11/outputs/access-om2-01/01deg_jra55v140_iaf_cycle4_jra55v150_extension/output1001/ocean/ocean-2d-area_t.nc', '/g/data/ik11/outputs/access-om2-01/01deg_jra55v140_iaf_cycle4_jra55v150_extension/output1002/ocean/ocean-2d-area_t.nc', '/g/data/ik11/outputs/access-om2-01/01deg_jra55v140_iaf_cycle4_jra55v150_extension/output1003/ocean/ocean-2d-area_t.nc', '/g/data/ik11/outputs/access-om2-01/01deg_jra55v140_iaf_cycle4_jra55v150_extension/output1004/ocean/ocean-2d-area_t.nc', '/g/data/ik11/outputs/access-om2-01/01deg_jra55v140_iaf_cycle4_jra55v150_extension/output1005/ocean/ocean-2d-area_t.nc', '/g/data/ik11/outputs/access-om2-01/01deg_jra55v140_iaf_cycle4_jra55v150_extension/output1006/ocean/ocean-2d-area_t.nc', '/g/data/ik11/outputs/access-om2-01/01deg_jra55v140_iaf_cycle4_jra55v150_extension/output1007/ocean/ocean-2d-area_t.nc', '/g/data/ik11/outputs/access-om2-01/01deg_jra55v140_iaf_cycle4_jra55v150_extension/output1008/ocean/ocean-2d-area_t.nc', '/g/data/ik11/outputs/access-om2-01/01deg_jra55v140_iaf_cycle4_jra55v150_extension/output992/ocean/ocean-2d-area_t.nc', '/g/data/ik11/outputs/access-om2-01/01deg_jra55v140_iaf_cycle4_jra55v150_extension/output993/ocean/ocean-2d-area_t.nc', '/g/data/ik11/outputs/access-om2-01/01deg_jra55v140_iaf_cycle4_jra55v150_extension/output994/ocean/ocean-2d-area_t.nc', '/g/data/ik11/outputs/access-om2-01/01deg_jra55v140_iaf_cycle4_jra55v150_extension/output995/ocean/ocean-2d-area_t.nc', '/g/data/ik11/outputs/access-om2-01/01deg_jra55v140_iaf_cycle4_jra55v150_extension/output996/ocean/ocean-2d-area_t.nc', '/g/data/ik11/outputs/access-om2-01/01deg_jra55v140_iaf_cycle4_jra55v150_extension/output997/ocean/ocean-2d-area_t.nc', '/g/data/ik11/outputs/access-om2-01/01deg_jra55v140_iaf_cycle4_jra55v150_extension/output998/ocean/ocean-2d-area_t.nc', '/g/data/ik11/outputs/access-om2-01/01deg_jra55v140_iaf_cycle4_jra55v150_extension/output999/ocean/ocean-2d-area_t.nc']

[dougiesquire]

Thanks for reporting @anton-seaice. You're right that the issue is that the catalog is trying to combine multiple datasets, with no clear dimension coordinate to combine along.

I took a look at how this is dealt with in the cosima-cookbook, but it has the same issue, e.g. try:

import cosima_cookbook as cc

session = cc.database.create_session()

area_t = cc.querying.getvar(
    '01deg_jra55v140_iaf_cycle4_jra55v150_extension',
    'area_t',
    session,
)

The cc.querying.getvar function supports an argument (n) to restrict the total number of files, which I guess is how this is dealt with there. I'm not sure about the best way to deal with this off the top of my head in the catalog (if you have any suggestions?). In the meantime, you could restrict the files in a similar way to the cookbook with, e.g.:

cat['01deg_jra55v140_iaf_cycle4_jra55v150_extension'].search(
    path=".*output1000.*",
    variable='area_t'
).to_dask()

[marc-white]

I wonder if #192 is the same problem... will need to check when Gadi comes back to life.

[anton-seaice]

Yeah all the grid variables will have the same problem.

I think this is the same issue as intake/intake-esm#660