NCAR to main (2023-09-14) #1610
Conversation
Replaces LU_pred to L_diag, since now this logical only controls if diagnostics should be posted.
This commit adds the latest updates to the vertFPmix subroutine after Bill Large did some cleaning. We have highlight places in the code where work must be done.
omega_w2x is the counter-clockwise angle of the wind stress with respect to the horizontal abscissa (x-coordinate) at tracer points [rad]. This variable is needed in the vertPFmix subroutine.
This line of code was lost during the last merge.
* Rename MOM_lateral_boundary_diffusion.F90 to MOM_hor_bnd_diffusion.F90. * Following the suggestion from a reviewer, MOM_lateral_boundary_diffusion has been renamed to MOM_hor_bnd_diffusion. Many submodules related to the 'old; lateral diffusion have been renamed throughout the code. LBD has been replaced to HBD. * Tested that answers for GMOM do not change.
Build and store the HBD grid outside the tracer loop since the same grid is used in all tracers. This makes this module more computationaly efficient. A GMOM case run for 10 days and with 3 tracer is ~ 7.5 % faster.
add desc argument to log_param calls in MOM_CFC_cap
CFC_BC_FILE must be specified if USE_CFC_CAP=.true. use hemispheric averages poleward of 10 degrees latitude linearly interpolate between 10S and 10N correct bug that atm cfc12 was used in cfc11 flux computation
CFC_cap forcing update
* Add subroutine hbd_grid_test, which mimics subroutine hbd_grid but it is only used in the unit tests; * Add unit tests for hbd_grid_test and fix existing tests for fluxes_layer_method; * Delete unused code and fix the format of doxygen throughout the module.
…OM_CFC_cap refs moved out of nuopc cap code, MOM_forcing_type, MOM_variables call CFC_cap_set_forcing in call_tracer_set_forcing add call to call_tracer_set_forcing in nuopc cap add arguments to call_tracer_set_forcing increase width in MOM_CFC_cap unit test output correct typo in oil_tracer
…of_cap migrate nearly all refs to CFC_cap into MOM_tracer_flow_control and MOM_CFC_cap
Improvements in the MOM_lateral_boundary_diffusion module
* changes in nuopc cap, infra, and MOM.F90 to receive ensembe id from the coupler (alternative to FMS ensemble mngr) * multi-instance logfile name correction in nuopc cap * append ensemble suffix to _doc files * changes in rpointer and restart file name handling to accommodate multi-instance CESM runs * remove fms2_io_mod usage in FMS1/MOM_ensemble_manager_infra.F90 * rm whitespace in mom_cap
…ix_draft_18April2023
This commit adds the option to apply a linear decay in the neutral diffusion fluxes within a transition zone defined by the boundary layer depths of adjacent columns. This option is controlled by a new parameter NDIFF_TAPERING, which is only available when NDIFF_INTERIOR_ONLY=True. By default NDIFF_TAPERING=False and answers are bitwise identical.
…0406 Merge changes from GFDL to main (2023-04-06)
Writes useful fields when the diffusivity of viscosity is less than zero. The should help understanding the root cause of such cases and facilitate the necessary adjustments.
…ngmuir_kpp Output relevant fields when diff or visc < 0
Simplifies and reduces the code by adding hbd to the neutral diffusion contril structure. This avoid the need to "extract" hbl multiple times. Answers are bitwise indenticals.
|
Since it runs in debug mode, I'm willing to chalk it up as a compiler issue. All the ESMG tests run. I don't have other compilers to try on "new chinook" with my current software versions. I'll approve assuming that line will get taken out. |
Remove fms_io_mod import
|
It worked fine on R&D machine (b4b results) but I am seeing different results on EMC operational machine. I am doing debug work and will provide some clue soon. |
@alperaltuntas and @gustavo-marques
|
|
Switching to ifort 2023.1 fixed this for me (and broke something else). |
|
@gustavo-marques and @alperaltuntas some updating from UFS: and problem starts from: our regression system only allows each user to run one regression test at the same time, thus it is a slow process for me to provide this clue to you. |
|
This PR passed our regression, but I suspect we will need a bit of time to manually review the changes. I am also a bit concerned about the issues raised by @kshedstrom, although I'm not yet sure which are related to this PR. |
|
@jiandewang, thanks for doing these tests and narrowing down where the change is coming from. I will look into this tomorrow and get back to you. |
@gustavo-marques one more clue for you: UFS quarter degree runs are fine on wcoss2. The half and 1x1 degree runs all failed to retain baseline results. We follow closely to the MOM_input in MOM_examples for all these 3 resolutions, so the issue must be some parameter related but I can't figure out which one it is. |
|
@marshallward; regression tests are failing because |
|
@jiandewang; we are still unable to find where the issue is coming from. |
see https://www.emc.ncep.noaa.gov/gc_wmb/wd20xw/JD/MOM_parameter_doc.short.tar-txt, remove "txt" before untar it |
|
I have tried to look at this, and the only thing that seems to make sense is a regression in The only idea I have at the moment is that the inclusion of My best suggestion for now is to completely remove all of the content related to |
|
@marshallward: thanks for the feedback. |
|
@gustavo-marques: with your latest commit, all UFS jobs retain current baseline now. |
|
@gustavo-marques Did you want to clean up these changes? Or did you want to leave them in so that we can revisit this problem? This is now the second mysterious computational |
|
@marshallward, our computers are down this week. I suggest we take this PR as is and we (NCAR) will revisit this issue in a future PR. |
|
@gustavo-marques Hope you don't mind, I went ahead and merged this for you. Thanks to everyone, this one was particularly tricky. |
|
@marshallward Thanks for merging it |
This PR includes the following:
Features
New parameters
Bug fixes & documentation improvements
Code improvements and cleanup
Obsolete MCT cap
Contributors