Change our testing from using the POP land/ocean mask to the MOM t232 land/ocean mask

[ekluzek]

What is the feature/what would you like to discuss?

Since POP has been removed in cesm3 -- it's nonsensical that our testing is done with the POP land/ocean mask rather than the MOM t232 land/ocean mask. In. the past it was also seen as important to run our standalone simulations with the same land/ocean mask and or grid used in coupled simulations. I'm concerned that by running all of our testing and standalone simulations with the POP mask that we will miss important problems in the fully coupled simulations that now are ONLY using the MOM t232 land/ocean mask. Anything we can do to make our testing closer to the fully coupled simulations being done is a good idea in my mind.

Is there anyone in particular you want to be part of this conversation?

@cacraigucar @nusbaume @peverwhee

Will this change (regression test) answers?

No

Will you be implementing this enhancement yourself?

Any CAM SE can do this

[adamrher]

I mentioned this in the VR sub-issue, but I wanted to make sure this was explicit in this main issue -- I think the highest priority task is to ensure that the CESM3 workhorse configurations (B-case, F-case, I-case, G-case, others?) are using the MOM6 ocean mask in the B-case. CAM should prioritize adding a ne30pg3_ne30pg3_mt232 grid alias to ccs_config, and swap this grid in for the relevant CAM7 regression tests.

I also understand that WACCM/CAM-Chem folks want to be running the coarse grid ne16pg3 grid with CAM7. We usually use a coarser ocean grid as the mask, but MOM6 doesn't offer anything coarser than t232 (0.5˚). It would be helpful to understand if these folks plan to run the t232 MOM6 grid in a B-case, or if they have other plans.

Are there issues for removing POP from the code, and should that address removing POP based grid aliases from testing and ccs_config?

[ekluzek]

Thanks @adamrher. Yes you are right the first grid you mention is critical to get in before much of anything.

POP has been removed from the system. I think the only vestiges left are these ocean masks and grids. I'll make sure ccs_config has an issue to remove the POP grids and masks.

[brian-eaton]

I mentioned this in the VR sub-issue, but I wanted to make sure this was explicit in this main issue -- I think the highest priority task is to ensure that the CESM3 workhorse configurations (B-case, F-case, I-case, G-case, others?) are using the MOM6 ocean mask in the B-case. CAM should prioritize adding a ne30pg3_ne30pg3_mt232 grid alias to ccs_config, and swap this grid in for the relevant CAM7 regression tests.

@adamrher, as you pointed out in #1239 there is already an alias for the ne30np4.pg3 grid using the tx2_3v2 mask. The alias is ne30pg3_t232 and the allactive compsets already use this grid. I think the main issue is that we should update our ne30pg3 testing of cam7 compsets to use this alias rather than the one that uses the mg17 mask. I will add this to my testing updates (there is no PR yet for this).

We also do some low res testing using the ne3np4.pg3 grid with the gx3v7 mask. I'm not sure I see any reason to change this mask to a higher resolution MOM mask. The low res test runs are not used to tune anything and the low res grids are not used by any allactive compset configurations. What do you think?

POP has been removed from the system. I think the only vestiges left are these ocean masks and grids. I'll make sure ccs_config has an issue to remove the POP grids and masks.

@ekluzek, I don't think the POP masks should be removed. In addition to low resolution cam7 tests, CAM also does a lot of testing of old physics packages, i.e., cam4, cam5, and cam6. These tests all use the POP masks and I can't think of any good reason to change them to being MOM masks.

[ekluzek]

@ekluzek, I don't think the POP masks should be removed. In addition to low resolution cam7 tests, CAM also does a lot of testing of old physics packages, i.e., cam4, cam5, and cam6. These tests all use the POP masks and I can't think of any good reason to change them to being MOM masks.

Ahhh, that's a good point I'm glad you brought it up so we can state it in the CSEG meeting later today. I think you are right from a science perspective. And the same thing applies to CLM for clm4_5, clm5_0 physics. It would make sense to have the ocean mask stay the same for them as well. On the science side the ocean mask update will change answers around coasts. So it's not enormous, but significant compared to roundoff.

From an SE perspective it DOES make sense to remove the old masks to have cleaner code. And to not allow as many options which has a higher risk of something failing for a strange configuration that's allowed, but maybe not tested. I could see where we figure the code cleanup is more important than keeping the science exact. But, we need to hear from more people and especially the scientists that want to use the old physics options.

Pinging @briandobbins in case he wants to weigh in here.

[briandobbins]

I agree this is a good point; removing them entirely from the code is a no-go then, but I do feel we should aim for all of our testing of CESM3-related configurations to use the new MOM6 masks. That said, I'm still new to a CESM release, and want to hear what others say at today's CSEG meeting.

[ekluzek]

Even if we leave some POP grids, we could delete a lot of them and remove a lot of the combinations. For CLM we only need gx3v7 and gx1v7 with f10, f19, f09. So gx16, tx.01v2, tx0.1v3 could all be removed, and aliases that are are different atm/lnd resolutions from that list could be removed.

I suspect the same for CAM, although I didn't see any testing in CAM for older physics -- but may have missed it...

[brian-eaton]

CAM's testing uses the gx1v7 and gx3v7 masks almost exclusively. A couple of high resolution mpas grids use the tx0.1v3 mask, and some SE refined grids use the tx0.1v2 mask.

Over 90% of CAM's current testing uses older physics, i.e., cam4, cam5, or cam6. Just look at the compset definitions to verify.

[ekluzek]

Thanks @brian-eaton. Ahhh, I assumed that compsets that didn't include a version in them were the latest physics. It does seem to me that our testing needs to switch for the release so that most testing is with cam7 for what we want to make sure is working in the release. This is something we want to do in CLM as well, make most testing with clm6_0, a lower amount with clm5_0, and the least with clm4_5. Since, the latest physics is the one changing it needs more extensive testing to make sure those changes don't break something.

I also propose in that we move mpas grids to MOM masks in #1240 as well as the SE refined grids in #1239. I think those might be things that are done post-release though, so that likely means we need to keep them around for a while. However, we should be able to remove gx1v6! And there might be some compset aliases combinations that can go away.