cesm3_0_beta05 case.build MOM error: DO_GEOTHERMAL

What version of the code are you using?
cesm3_0_beta05-1-g8e7888c
in /glade/work/raeder/Models/cesm3_0_beta05


Have you made any changes to files in the source tree?
I added in ESCOMP/DART_interface, which Alper Altuntas developed in 2023 using POP as an example.
This involved adding a 'dart' section to .gitmodules, which @jedwards thought looks usable.
? Is it likely to cause a problem that the DART_interface was developed for POP,
but not extended to other components or MOM?

I modified config_archive.xml files; I haven't reached the point of trying to use them yet.


Describe every step you took leading up to the problem:
On derecho:
Downloaded CESM from ESCOMP and checked out the cesm3_0_beta05 tag
(/glade/work/raeder/Models/cesm3_0_beta05)
Forked the DART_interface from ESCOMP to GitHub - kdraeder/DART_interface: CESM-DART interface; required for adding DART files to the st_archive list..
Cloned the interface into components/dart and modified .gitmodules to include it.
Modified components/*/cime_config/config_archive.xml to handle DART file name extensions.

$ setenv cesmroot /glade/work/raeder/Models/cesm3_0_beta05
$ setenv CIMEROOT /glade/work/raeder/Models/cesm3_0_beta05/cime
The compset used here is a simplified version of BHISTC_LTso (stub land ice and wave).
I also tried HIST_CAM60_CLM50%BGC-CROP_CICE_MOM6_MOSART_SGLC_WW3_BGC%BDRD unmodified,
but got the same result.
$ /glade/work/raeder/Models/cesm3_0_beta05/cime/scripts/create_newcase
--case /glade/work/raeder/Exp/St_cesm3_noDART
--machine derecho
--res f19_g17_r05
--project P86850054
--queue main
--walltime 0:10
--pecount 128x1
--ninst 3
-d
--compset HISTC_CAM70%LT_CLM60%BGC-CROP_CICE_MOM6_MOSART_SGLC_SWAV
--run-unsupported
--multi-driver
This appeared to work, and it instructed me to run case.setup:
---------------------------------------------------
2025-03-07 16:43:41: case.setup starting
---------------------------------------------------
2025-03-07 16:43:47: case.setup success

When I run `case.build -d` I get the exerpts in the attached file (I can't attach the whole file because it's too big).


Describe your problem or question:
I'm trying to build something close to "out of the box", which I can use for testing mods to the st_archive xml files,
which will handle DART-related output. I'm requesting a small ensemble and low, standard resolution.
I chose a supported CAM compset but changed the land ice and wave models to stub, to match the possible output file set from DART runs.

I'm not familiar with MOM namelists, and couldn't find any helpful information about DO_GEOTHERMAL,
and mom_cime doesn't appear in any file names or files in components/mom or cime,
so I'm hoping that someone can point me in the right direction to identifying and fixing the problem.
As noted above, it may be related to the DART_interface.

I removed the DART_interface,
moved the components/dart directory out of the way,
ran `./bin/git-fleximod update`, and created a new case using
HISTC_CAM70%LT_CLM60%BGC-CROP_CICE_MOM6_MOSART_SGLC_SWAV.
case.setup ran without error.
case.build failed in the same way as above.

It seems unhappy with the 'data' dictionary in cime/CIME/ParamGen/paramgen.py line 392.
The contents of 'data' above there are

_reduce_recursive data = {
'module': 'ocean_model',
'packing': '= 1 if $TEST else 2',
'lists': [['geolon', 'geolat', 'geolon_c', 'geolat_c', 'geolon_u', 'geolat_u', 'geolon_v', 'geolat_v',
'deptho', 'wet', 'wet_c', 'wet_u', 'wet_v', 'Coriolis', 'areacello', 'areacello_cu', 'areacello_cv', 'areacello_bu',
'sin_rot', 'cos_rot', 'C_P', 'Rho_0']],
'lists2': {'$DO_GEOTHERMAL == "True"': [['Geo_heat']]}
}

Click to expand...

It looks strange that the error message is "KeyError: 'DO_GEOTHERMAL", but DO_GEOTHERMAL is not a key,
it's a variable in the dictionary entry(?).

Then in the loop over keys it prints entries for the first 3, but can't print lists2:

_reduce_recursive reduced data[key] = ocean_model
key module

_reduce_recursive reduced data[key] = 2
key packing

_reduce_recursive reduced data[key] = [['geolon', 'geolat', 'geolon_c', 'geolat_c', 'geolon_u', 'geolat_u', 'geolon_v', 'geolat_v', 'deptho', 'wet', 'wet_c', 'wet_u', 'wet_v', 'Coriolis', 'areacello', 'areacello_cu', 'areacello_cv', 'areacello_bu', 'sin_rot', 'cos_rot', 'C_P', 'Rho_0']]
key lists

Click to expand...

The next messages are variations on

RUN: git -C /glade/work/raeder/Exp/St_cesm3_flexi_noDART ls-files --deleted --exclude-standard
FROM: /glade/work/raeder/Exp/St_cesm3_flexi_noDART

Click to expand...

Then the traceback.

When I print each 'key' before the each'data[key]' it's clear that the key 'lists2' is still OK, but the entry can't be printed.
I'm guessing that the "$DO_GEOTHERMAL" expansion failed.

I finally found components/mom/standalone/examples/cesm_t232_with_MARBL/MOM_input,
which says that the variables in it should only be changed through the user_nl_mom namelist(s) (or SourceMods).
I put 'DO_GEOTHERMAL = False' in each of my namelists, and that problem disappeared.
Next it can't find NIGLOBAL, so I'm guessing that there's a failure to make a default mom namelist somewhere.

It appears that components/mom/param_templates/json/MOM_input.json contents are being read into buildnml,
but 313 of its variables are not appearing in the MOM_input_#### files (see attached for the list),
including DO_GEOTHERMAL and NIGLOBAL.
It seems unlikely that users should need to figure out which of those need to be put into a user_nl_mom file.

This might be a multi-instance problem, based on some of the code in buildnml.

I tried building a couple MOM compsets, to see what ends up in those namelists.
Both G1850MARBL_JRA (_MOM6%MARBL_) and G_JRA_RYF (_MOM6_) also failed on $DO_GEOTHERMAL.

Building a single instance case of G_JRA_RYF doesn't sidestep the problem.

Now I see that DO_GEOTHERMAL is actually read from MOM_input.json in MOM's buildnml:prep_input:
'DO_GEOTHERMAL': {'description': '"[Boolean] default = FalsenIf true, apply geothermal heating."n', 'datatype': 'logical', 'units': 'Boolean', 'value': {'$OCN_GRID == "tx2_3v2"': True, '$OCN_GRID == "tx0.25v1"': True}},
But the 'value' is 'None' because the grid I'm trying to use is not one of the two listed,
and apparently no default is set to handle that case, despite what 'description' says.
So DO_GEOTHERMAL is not written to the MOM_input namelist and is not available to fill in $DO_GEOTHERMAL variables.
Where should the default value be set?

It turns out that the problem is not that default values are not being set or used,
it's that I'm trying to use a grid that's incompatible with MOM6.
I was counting on create_newcase or case.build to tell me that, but that's not a feature yet.
@altuntas says "gx1v7 is not a supported MOM6-CESM grid. I recommend using our standard grid, tx2_3v2."
The choices seem to be

• TL319_t232[+wave options], which requires a DATM compset,
• T62_t232[+wave options], with compsets that are not _CAM
• f09_t232[+glacier options],
• ne30pg3_t23[+glacier options]

So I consider this thread to be finished, but hopefully create_newcase or case.build will be augmented
to give meaningful error messages when users ask for something that can't work.