`cell method` Issue

[rhaegar325]

hur	Amon	Expected 'area: time: mean where air' from registry key 'cell_methods', got 'time: mean'.
ta	Amon	Expected 'area: time: mean where air' from registry key 'cell_methods', got 'time: mean'.
ua	Amon	Expected 'area: time: mean where air' from registry key 'cell_methods', got 'time: mean'.
va	Amon	Expected 'area: time: mean where air' from registry key 'cell_methods', got 'time: mean'.
wap	Amon	Expected 'area: time: mean where air' from registry key 'cell_methods', got 'time: mean'.
zg	Amon	Expected 'area: time: mean where air' from registry key 'cell_methods', got 'time: mean'.
rlus	Amon	Expected 'area: time: mean where sea_ice (mask=siconca)' from registry key 'cell_methods', got 'area: time: mean'.
rsus	Amon	Expected 'area: mean where land' from registry key 'cell_methods', got 'area: time: mean'.
rootd	fx	Expected 'area: mean where land' from registry key 'cell_methods', got 'area: mean'.
snc	LImon	Expected 'area: mean where land time: mean' from registry key 'cell_methods', got 'area: mean: time: mean'.*

where air — atmospheric 3D/column vars (hur, ta, ua, va, wap, zg)
where land — land surface vars (rsus, rootd, snc)
where sea_ice (mask=siconca) — rlus

[rbeucher]

I had a look through the current implementation and this seems to split into two cases.

For most of the variables listed here (hur, ta, ua, va, wap, zg, plus rootd and snc), I do not think this implies a change to how the data are created. In ACCESS-MOPPy we copy cell_methods from the selected CMOR vocabulary/table into the output attrs, so for these variables this looks like a CMIP7 vocabulary / branded-variable selection issue rather than a derivation issue.

A few concrete points from current main:

• CMIP7 runs translate CMIP6-style names to branded CMIP7 names before selecting the CMIP7 vocab/table entry.
• The current bundled CMIP7 tables already have the richer cell_methods for the 3D atmospheric variables, e.g. area: time: mean where air for hur/ta/ua/va/wap/zg.
• rootd and snc also already line up with the current CMIP7 land / landIce table entries (area: mean where land, area: mean where land time: mean).

The one part that still looks potentially real is rlus / rsus. At the moment Amon.rlus and Amon.rsus still map to the unqualified atmos variants (atmos.rlus.tavg-u-hxy-u.mon.GLB / atmos.rsus.tavg-u-hxy-u.mon.GLB), whose CMIP7 table entries still use plain area: time: mean. But CMIP7 now also has masked variants for land / sea-ice / land-ice cases, e.g. land.rsus.tavg-u-hxy-lnd.mon.GLB and seaIce.rlus.tavg-u-hxy-si.mon.GLB with where land / where sea_ice (mask=siconca).

So my take is:

• hur/ta/ua/va/wap/zg/rootd/snc: probably no data-generation change needed; mostly request/vocabulary alignment.
• rlus/rsus: may need actual remapping to different CMIP7 branded variables, and possibly masking logic depending on which CMIP7 request we want to satisfy.

If that sounds right, I think the next step is to narrow this issue to the specific CMIP7 targets we want for rlus and rsus, and treat the rest as already covered by the current CMIP7 tables.