Updated mom6 tag

[yichengt900]

As titled. This is a regualr update for mom6 and ocean_BGC tags mom6 tag only for now.

This PR from ocean_BGC changes the baseline in NWA and NEP due to the use of a real (floating-point) exponent instead of the original integer exponent, resulting in subtle differences.

[yichengt900]

Unfortunately, the new MOM6 tag (synced with the current dev/gfdl MOM6) produced the following error message in the NEP domain when using the same mask table we used for the production run, mask_table.524.32x80:

FATAL from PE  1024: MPP_DEFINE_DOMAINS(mpp_compute_extent): domain extents must be positive definite.


FATAL from PE  1016: MPP_DEFINE_DOMAINS(mpp_compute_extent): domain extents must be positive definite.


FATAL from PE   955: MPP_DEFINE_DOMAINS(mpp_compute_extent): domain extents must be positive definite.


FATAL from PE   963: MPP_DEFINE_DOMAINS(mpp_compute_extent): domain extents must be positive definite.

I will put this PR on hold and wait for further investigation.

CC @theresa-cordero

Update: Further investigation found that this error started appearing afterhttps://github.com/NOAA-GFDL/MOM6/pull/885

[yichengt900]

OK, further investigation showed that the following section of code is causing the issue we encountered :
https://github.com/NOAA-CEFI-Regional-Ocean-Modeling/MOM6/blob/5f6a7600006a53998104aff3d13410de4c6206c8/src/framework/MOM_domains.F90#L446-L451

This new code attempts to create an unmasked domain and write out unmasked ocean geometry, even if a mask_table is used. For some reason, it doesn't work with the NEP10k 32x80 mask_table. Currently, this section of code is always executed when a mask_table is specified.

Would it be possible to add a runtime parameter to control this behavior temporally? That way, we could disable this section if we don’t want to generate an unmasked ocean geometry file. This would also allow me to merge the latest updates from dev/gfdl into cefi, while buying us more time to fully resolve this issue.

[theresa-cordero]

We could add a parameter to stop this. Should I open a PR to dev/cefi?

[yichengt900]

Sure, that would be great! I can merge it first, then test the latest OBC PR that was just merged yesterday.

[andrew-c-ross]

This same error popped up in older versions of MOM6 with certain layouts of the NWA25 model. I could never figure out why it was happening, but if/when there's a tentative fix I'd be happy to test it out

[theresa-cordero]

I encountered this problem with the NWA12 with the 50x50 mask table, although Yi-Cheng did not.

I added a flag to avoid this code in the MOM6 dev/cefi PR #28

It will be worth knowing why these domain decompositions are causing problems without mask tables.