UFS dev PR#232/256

[hertneky]

Changes combined from PRs 232 and 256

Associated fv3atm PRs:
NOAA-EMC/fv3atm #993
NOAA-EMC/fv3atm #885

[scrasmussen]

Once my comment on if Interstitial%tkeh should be initialized to zero or not is answered this looks good to me!

[scrasmussen]

Thanks for changing the initialization from 0 to clear_val, looks good!

[hertneky]

I just need the physics merged for this one @dustinswales @grantfirl

[grantfirl]

@hertneky Please put a reference to the PR from fv3atm into the description. I think that it helps to understand the provenance of these changes for future versions of us. NOAA-EMC/ufsatm#933 and NOAA-EMC/ufsatm#885

[grantfirl]

@hertneky @scrasmussen @mkavulich @dustinswales When we do these ufs/dev catchup PRs, we also need to be checking that our RTs match what is expected from the physics changes and what was seen during the UFS RTs: https://github.com/rhaesung/ufs-weather-model/blob/1715891b34332e025aa161ac8fcfd67ac0b1b809/tests/test_changes.list

[grantfirl]

It looks like the SCM RTs using the following suites differ (see https://github.com/NCAR/ccpp-scm/actions/runs/13842355588/job/38732662755?pr=566):
SCM_GFS_v17_p8_ugwpv1
SCM_GFS_v16_RRTMGP
SCM_GFS_v16
SCM_GFS_v16_no_nsst
SCM_GFS_v17_p8_ugwpv1_no_nsst
SCM_GFS_v17_p8
SCM_GFS_v16_debug

Note that all of these suites use samfdeepcnv and samfshalcnv where the RT differences should originate. Note also that not ALL cases using these suites have different results. This is an important point to remember -- that RTs in the SCM are short and for one column and differences that show up in UFS RTs with millions of columns and more time steps don't always show up in the SCM tests. In this case, some cases happened to trigger the piece of code that resulted in differences while other cases using the same exact suite did not.

[grantfirl]

RT changes are as expected. Thanks @hertneky

[grantfirl]

@hertneky Please feel free to merge after adding the additional info to the PR description as I suggested. Also, since baselines changed, someone with permissions at NCAR needs to upload the newly-created baseline artifacts to mohawk.

[hertneky]

It looks like the SCM RTs using the following suites differ (see https://github.com/NCAR/ccpp-scm/actions/runs/13842355588/job/38732662755?pr=566): SCM_GFS_v17_p8_ugwpv1 SCM_GFS_v16_RRTMGP SCM_GFS_v16 SCM_GFS_v16_no_nsst SCM_GFS_v17_p8_ugwpv1_no_nsst SCM_GFS_v17_p8 SCM_GFS_v16_debug

Note that all of these suites use samfdeepcnv and samfshalcnv where the RT differences should originate. Note also that not ALL cases using these suites have different results. This is an important point to remember -- that RTs in the SCM are short and for one column and differences that show up in UFS RTs with millions of columns and more time steps don't always show up in the SCM tests. In this case, some cases happened to trigger the piece of code that resulted in differences while other cases using the same exact suite did not.

@grantfirl Thanks for the info and I will make note of the expected changes from here on out. I was able to locate the test_changes.list file as well by going to the associated ufs-weather-model PR.

[hertneky]

Can someone merge the physics so I can update the physics pointer and merge this one?

[grantfirl]

Can someone merge the physics so I can update the physics pointer and merge this one?

Done. Commit hash is NCAR/ccpp-physics@7e42e7e