Inconsistencies with CMIP6

[sol1105]

Hi @larsbuntemeyer

I just want to list a few inconsistencies I noticed between CORDEX-CMIP6 and CMIP6 CMOR tables. Some are or may be relevant for a successful application of the CMOR library:

• The cell_methods area: mean time:mean in CORDEX-CMIP6 are specified as area: time: mean in CMIP6 CMOR tables.
• The positive attribute for fluxes like hfls, hfss, rlds, rlus, etc. is not defined in the CORDEX-CMIP6 CMOR tables. Even though the standard_name includes the correct orientation, the attribute is important for the CMOR application. CMOR does not write the positive attribute to the file, but uses it to check and potentially switch the orientation.
• In the recent PR #63, you updated the order of dimensions in the CMOR tables. However, in CMIP6, singleton vertical dimensions (single vertical levels such as p220 or height2m) are always listed last, eg. longitude latitude time height2m. Only "normal" vertical dimensions (like alevel, plev19 etc.) are listed before the time dimension. I am not sure if this has relevance for the CMOR application.
• The comment attribute is mostly empty, one could use the comments defined in CMIP6.
• The modeling_realm is set to atmos in all cases, even for variables such as mrso with comment The mass of water in all phases per unit area, summed over all soil layers.. Beside the modeling_realm, also the cell_methods for this kind of variables are different in CMIP6: "area: mean where land time: mean" instead of area: mean time: mean.

Lastly, I noticed a small typo for od550aer: both the label and the out_name for od550aer are followed by a space:

"od550aer_": {
            [...]
            "out_name": "od550aer_",
            [...]

[larsbuntemeyer]

Thanks again @sol1105 very much appreciated! I'll split up your issues and work on that on our data request table. Just some few comments here:

• I really wasn't aware of the difference between "normal" vertical dimensions and singleton vertical dimensions, thanks for that! @gnikulin Could you confirm that the soil variables in CORDEX-CMIP6 (tsl, mrsfl and mrsol) are supposed to be "real" 3D variables (as it is mentioned in the comments here). They weren't part of the CMOR in tables in CMIP5 so i am not sure.

• We discussed the modeling_realm issue in https://github.com/WCRP-CORDEX/data-request-table/issues/4. @gnikulin I think our last decision on that was to simply remove it from the variable attributes?

I will add a test case with a vertical coordinate to the examples to make sure the CMOR tool works with it. The examples are created using the python API without the _cmip6_option (https://github.com/PCMDI/cmor/discussions/679#discussioncomment-3842956)) in the input dataset table.

[larsbuntemeyer]

@gnikulin should we include the original CMIP6 variable attribute "comment"? I would vote for that (although that might depend on the table_id if i remember correctly). Right now, we use the "Comments" column from the original CMIP6 data request.

[larsbuntemeyer]

By comparing cell_methods of current CORDEX data request with CMIP6 data request, i find the following inconsistencies (more details here: https://github.com/WCRP-CORDEX/data-request-table/issues/19)

out_name	frequency	cordex	cmip6	cmip6_realm	cmip6_table
evspsbl	day	area: time: mean	area: mean where land time: mean	land	Eday
tsl	mon	area: time: mean	area: mean where land time: mean	land	Lmon
tsl	day	area: time: mean	area: mean where land time: mean	land	Eday
tsl	6hrPt	area: mean time: point	area: mean where land time: point	land	6hrPlevPt
mrros	mon	area: time: mean	area: mean where land time: mean	land	Lmon
mrros	day	area: time: mean	area: mean where land time: mean	land	Eday
mrro	mon	area: time: mean	area: mean where land time: mean	land	Lmon
mrro	day	area: time: mean	area: mean where land time: mean	land	day
snm	day	area: time: mean	area: mean where land time: mean	landIce land	Eday
mrfso	mon	area: time: mean	area: mean where land time: mean	land landIce	Lmon
mrsfl	mon	area: time: mean	area: mean where land time: mean	land	Emon
mrsfl	day	area: time: mean	area: mean where land time: mean	land	Eday
mrso	mon	area: time: mean	area: mean where land time: mean	land	Lmon
mrso	day	area: time: mean	area: mean where land time: mean	land	day
mrsos	mon	area: time: mean	area: mean where land time: mean	land	Lmon
mrsos	day	area: time: mean	area: mean where land time: mean	land	day
mrsol	mon	area: time: mean	area: mean where land time: mean	land	Emon
mrsol	day	area: time: mean	area: mean where land time: mean	land	Eday
snw	mon	area: time: mean	area: mean where land time: mean	landIce land	LImon
snw	day	area: time: mean	area: mean where land time: mean	landIce land	day
snw	6hrPt	area: mean time: point	area: mean where land time: point	landIce land	6hrPlevPt
snd	mon	area: time: mean	area: mean where land time: mean	landIce land	LImon
snd	day	area: time: mean	area: mean where land time: mean	landIce land	Eday
evspsblpot	mon	area: time: mean	area: mean where land time: mean	land	Emon
evspsblpot	day	area: time: mean	area: mean where land time: mean	land	Eday
mrsofc	fx	area: mean	area: mean where land	land	fx

@gnikulin Should we stick with CMIP6 standard here? Note, that we don't have a modeling_realm in CORDEX for now...

[gnikulin]

* I really wasn't aware of the difference between "normal" vertical dimensions and singleton vertical dimensions, thanks for  that! @gnikulin Could you confirm that the soil variables in CORDEX-CMIP6 (`tsl`, `mrsfl` and `mrsol`) are supposed to be  "real" 3D variables (as it is mentioned in the comments [here](https://docs.google.com/spreadsheets/d/1qUauozwXkq7r1g-L4ALMIkCNINIhhCPx/#gid=1672965248)). They weren't part of the CMOR in tables in CMIP5 so i am not sure.

Yes, all 3 variables (tsl, mrsfl and mrsol) are "real" 3D (each soil layer).

[gnikulin]

@gnikulin should we include the original CMIP6 variable attribute "comment"? I would vote for that (although that might depend on the table_id if i remember correctly). Right now, we use the "Comments" column from the original CMIP6 data request.

Yes, I would also use "Comments" from CMIP6 when possible. We need to make minor adjustments because there are several inconsistencies in the CMIP6 comments attribute for some "similar" variables.

[gnikulin]

By comparing cell_methods of current CORDEX data request with CMIP6 data request, i find the following inconsistencies (more details here: WCRP-CORDEX/data-request-table#19) out_name frequency cordex cmip6 cmip6_realm cmip6_table evspsbl day area: time: mean area: mean where land time: mean land Eday tsl mon area: time: mean area: mean where land time: mean land Lmon tsl day area: time: mean area: mean where land time: mean land Eday tsl 6hrPt area: mean time: point area: mean where land time: point land 6hrPlevPt mrros mon area: time: mean area: mean where land time: mean land Lmon mrros day area: time: mean area: mean where land time: mean land Eday mrro mon area: time: mean area: mean where land time: mean land Lmon mrro day area: time: mean area: mean where land time: mean land day snm day area: time: mean area: mean where land time: mean landIce land Eday mrfso mon area: time: mean area: mean where land time: mean land landIce Lmon mrsfl mon area: time: mean area: mean where land time: mean land Emon mrsfl day area: time: mean area: mean where land time: mean land Eday mrso mon area: time: mean area: mean where land time: mean land Lmon mrso day area: time: mean area: mean where land time: mean land day mrsos mon area: time: mean area: mean where land time: mean land Lmon mrsos day area: time: mean area: mean where land time: mean land day mrsol mon area: time: mean area: mean where land time: mean land Emon mrsol day area: time: mean area: mean where land time: mean land Eday snw mon area: time: mean area: mean where land time: mean landIce land LImon snw day area: time: mean area: mean where land time: mean landIce land day snw 6hrPt area: mean time: point area: mean where land time: point landIce land 6hrPlevPt snd mon area: time: mean area: mean where land time: mean landIce land LImon snd day area: time: mean area: mean where land time: mean landIce land Eday evspsblpot mon area: time: mean area: mean where land time: mean land Emon evspsblpot day area: time: mean area: mean where land time: mean land Eday mrsofc fx area: mean area: mean where land land fx

@gnikulin Should we stick with CMIP6 standard here? Note, that we don't have a modeling_realm in CORDEX for now...

We need to check in detail. For all "soil" variables we indeed should use area: mean where land. However, for example evspsbl in the CMIP5 Amon table has "cell_methods": "area: time: mean" and is provided over both land and water.