Remove direct physcons module dependencies from CCPP-Physics by scrasmussen · Pull Request #381 · ufs-community/ccpp-physics

scrasmussen · 2026-08-15T17:34:54Z

Description of Changes:

Removes direct dependencies on the global physcons module from CCPP physics schemes.
Supplies physical constants through explicit CCPP arguments, with standard names, units, and metadata.
Makes each scheme’s required constants visible in its interface instead of relying on hidden module state.
Improves portability and reuse of physics schemes in different host models and configurations.
Simplifies dependency tracking, testing, and maintenance by centralizing ownership of physical constants in the host/CCPP framework.

Tests Conducted:

ufs-wm Derecho regression tests
- Regression Test Answer Changes
  - control_stochy_gnu failed only the bit-for-bit comparison
    - note: this test for intel, debug-intel, and debug-gnu passed
    - Some initial differences are tiny, around 1e-12, but the regression comparison has no numerical tolerance, so they fail
    - By hour 12, stochastic physics has amplified those perturbations into widespread differences
  - control_csawmg_gnu failed only the bit-for-bit comparison
    - note: this test for intel, debug-intel, and debug-gnu passed
    - The evidence points to the physical-constants refactor in UFSATM/ccpp/physics/physics/CONV/Chikira_Sugiyama/cs_conv.F90:303. Previously, derived quantities such as elocp, emeltocp, cpoel, and fact2 were compile-time parameters. They are now recomputed at runtime from CCPP arguments. Optimized GNU evaluates that arithmetic slightly differently
SCM RTs

Dependencies/Related PRs:

UFSWM - Remove direct physcons module dependencies from CCPP-Physics ufs-weather-model#3323
- UFSATM - Remove direct physcons module dependencies from CCPP-Physics NOAA-EMC/ufsatm#1131
  - ccpp-physics - Remove direct physcons module dependencies from CCPP-Physics #381
    - c3 - Remove direct physcons module dependencies from C3 c3#4
    - MYNN-SFC - Remove direct physcons module dependencies from MYNN NCAR/MYNN-SFC#32

Documentation:

No new documentation is needed, the removal of direct dependencies on the global physcons module falls in line with the already documented principles of the CCPP framework and physics.

…meta

- fpkapx function called before gfuncphys initializes its runtime constants, manually porting function

scrasmussen · 2026-08-15T17:45:34Z

Note to reviewers: my setup had "remove whitespace at the end of lines" automatically turned on, so by the end of this work, removing the whitespace-cleanup changes would have been near impossible. But if you go to "Files changed" -> [Settings Button] -> "Hide whitespace", you can remove those changes so it is easy to visualize the differences

climbfuji · 2026-08-17T13:27:04Z

Note to reviewers: my setup had "remove whitespace at the end of lines" automatically turned on, so by the end of this work, removing the whitespace-cleanup changes would have been near impossible. But if you go to "Files changed" -> [Settings Button] -> "Hide whitespace", you can remove those changes so it is easy to visualize the differences

It's easy to hide whitespaces when reviewing, but the real issue of a greatly increased chance for merge conflicts persists. Is it possible to turn off this setting until we've had the chance to go in and remove all trailing whitespaces from all files in one standalone PR?

scrasmussen · 2026-08-18T16:56:26Z

It's easy to hide whitespaces when reviewing, but the real issue of a greatly increased chance for merge conflicts persists. Is it possible to turn off this setting until we've had the chance to go in and remove all trailing whitespaces from all files in one standalone PR?

Yeah, that totally makes sense! I've added it back in with a commit, I also have a ufs-dev-whitespace-cleanup branch which removes the whitespace from all Fortran and meta files (no physcon changes). I didn't mean to force the whitespace-cleanup on everyone, do you think now would be a good time to submit that whitespace-cleanup branch as a PR?

climbfuji · 2026-08-18T17:01:06Z

It's easy to hide whitespaces when reviewing, but the real issue of a greatly increased chance for merge conflicts persists. Is it possible to turn off this setting until we've had the chance to go in and remove all trailing whitespaces from all files in one standalone PR?

Yeah, that totally makes sense! I've added it back in with a commit, I also have a ufs-dev-whitespace-cleanup branch which removes the whitespace from all Fortran and meta files (no physcon changes). I didn't mean to force the whitespace-cleanup on everyone, do you think now would be a good time to submit that whitespace-cleanup branch as a PR?

I think that's a good idea. Ideally, we'd have the codee formatter or another Fortran linter in place in CI to make sure it stays this way after your whitespace cleanup PR, but we can probably add a duct-tape solution to run in CI ourselves for just this format check?

grantfirl · 2026-08-26T17:06:48Z

@scrasmussen As a followup, we should move the entire physcons.F90/meta file to reside with the host models (UFSATM, CCPP-SCM, ...)

grantfirl · 2026-08-26T17:22:13Z

-      rovcp = rd*rcpd )
+      c5ies=c3ies*(tmelt-c4ies))
+
+     real(kind=kind_phys) :: rd = 1.0E30_kind_phys


What is the purpose of initializing these module-level constants before the init phase is called?

grantfirl · 2026-08-26T17:26:09Z

     &   cnvw(:,:), cnvc(:,:), dt_mf(:,:)
 !
-      real(kind=kind_phys), intent(out) :: ud_mf(:,:)
+      real(kind=kind_phys), intent(out), optional :: ud_mf(:,:)


Why was ud_f changed to optional? It doesn't appear to be optional to me. Also, the metadata doesn't reflect this. I'm guessing that we should just revert this?

grantfirl · 2026-08-26T17:33:39Z

    use cires_ugwp_triggers
    use ugwp_driver_v0
    use drag_suite, only: drag_suite_run, drag_suite_psl
+    use ugwp_common_v0, only: ugwp_common_v0_init


There may be some pushback RE: linking to the CIRES ugwp scheme from the unified ugwp scheme, although the boundaries are often blurry!

grantfirl · 2026-08-26T17:40:18Z


    logical :: tables_are_initialized = .false. ! initialize satuation tables

+    real(kind_phys) :: grav = 1.0E30


Same question RE: initializing to 1000. If it is necessary to initialize them, would it make sense to use something a little less plausible, e.g. -999.9? Something that would definitely cause a crash if somehow the setting of the constants didn't get done in the init phase?

Doh. I missed the extra 0. 10^30 is definitely unplausible enough! The question regarding the need to initialize here remains though.

grantfirl · 2026-08-26T17:45:40Z

       public meltfrz_inst
       public fix_up_clouds_2M
       PUBLIC CLOUD_PTR_STUBS
+#ifdef NEMS_GSM


I'm curious about this NEMS_GSM preprocessor flag. I'm guessing that this harkens back to a decade+ ago when this scheme was first being put into the GSM version of GFS? I can't see anywhere in the ufs-weather-model where this is being set. Are you sure that it is being used?

grantfirl · 2026-08-26T17:49:11Z

        this%pstr(iLev)  = pstr4(iLev)
-        this%pkstr(iLev) = fpkapx(this%pstr(iLev)*100.0)
+        ! following line equivalent to funcphys's fpkapx=(p/1.e5_krealfp)**con_rocp
+        this%pkstr(iLev) = (this%pstr(iLev)*1.0e-3_kind_phys)**con_rocp


What was the reasoning behind making this local instead of invoking the function? Its simplicity?

grantfirl · 2026-08-26T17:52:34Z

      errmsg = ''
      errflg = 0

+      amdw    = con_amd/con_amw


I don't love this in the run phase. It's minor, but it certainly doesn't need to be redone millions of times during a run.

grantfirl · 2026-08-26T17:53:11Z

      errflg = 0

+      ! Set atomic weights
+      amdw    = con_amd/con_amw


Same concern.

grantfirl · 2026-08-26T18:00:59Z

This is probably above-and-beyond, but did you happen to search for literal constant lurking in the code? E.g., if we see 3.14169, 9.81, 287.04, etc. there is a good chance that they should probably use the pi, gravity, dry air gas constant as everybody else. This would probably make one's eyeballs bleed looking at every constant and asking themselves "Is this already being provided by a host model?"

grantfirl · 2026-08-26T18:02:33Z

      real (kind=kind_phys), dimension(:), intent(inout) :: lai, rca
-      real (kind=kind_phys), dimension(:), intent(inout), optional ::   &
-     &       wet1
+      real (kind=kind_phys), intent(inout), optional :: wet1(:)


Saving a newline character?

grantfirl · 2026-08-26T18:03:05Z

            slc(i,k) = slsoil(k)
          enddo
-          wet1(i) = smsoil(1) / smcmax !Sarah Lu added 09/09/2010 (for GOCART)
+          if (present(wet1)) then


Good optional arg catch.

grantfirl · 2026-08-26T18:08:36Z

Really nice and thorough work, @scrasmussen! A couple of minor questions to answer, and I'll approve afterwards.

scrasmussen added 13 commits August 10, 2026 15:10

Removing physcons from tools/

6a2cc74

Remove physcons module from SFC_Models directory

27840a7

Removing physcons from SFC_Layer directory

9b3a084

Removing physcons from PBL directory

bd26a34

Removing physcons from CONV directory

de4d04c

Removing physcons from MP directory

2a34758

Removing physcons from Radiation directory

3d1d4e4

Removing physcons from GWD directory

87388e0

remove unused con_fv variable from smoke_dust/rrfs_smoke_wrapper.F90/…

16d2c4c

…meta

- removing physcons dependency in photochem/

2d87d4a

- fpkapx function called before gfuncphys initializes its runtime constants, manually porting function

Remove physcons from Interstitials

7c6aec5

Fixing physcons meta data, adding variables

574754a

Updating git modules, removing physcons module use

366a8f2

scrasmussen requested review from AnningCheng-NOAA, HelinWei-NOAA, JongilHan66, Qingfu-Liu, RuiyuSun, WeiguoWang-NOAA, XuLi-NOAA, andrewgettelman, barlage, dustinswales, grantfirl, haiqinli, mdtoyNOAA, mjiacono, rhaesung, tanyasmirnova and wd20xw as code owners August 15, 2026 17:34

scrasmussen requested review from climbfuji, lisa-bengtsson and mzhangw as code owners August 15, 2026 17:35

This was referenced Aug 15, 2026

Remove direct physcons module dependencies from CCPP-Physics NOAA-EMC/ufsatm#1131

Open

Remove direct physcons module dependencies from CCPP-Physics ufs-community/ufs-weather-model#3323

Open

This was referenced Aug 15, 2026

Remove references to physcons module from SCM NCAR/ccpp-scm#653

Open

Remove direct physcons module dependencies from CCPP-Physics NCAR/ccpp-physics#1235

Open

Remove physcons Module NCAR/ccpp-scm#652

Open

Add whitespace back in

30db9fd

grantfirl reviewed Aug 26, 2026

View reviewed changes

Comment thread physics/SFC_Models/Land/RUC/lsm_ruc.F90

lisa-bengtsson approved these changes Aug 26, 2026

View reviewed changes


		logical :: tables_are_initialized = .false. ! initialize satuation tables

		real(kind_phys) :: grav = 1.0E30

Uh oh!

Conversation

scrasmussen commented Aug 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description of Changes:

Tests Conducted:

Dependencies/Related PRs:

Documentation:

Uh oh!

scrasmussen commented Aug 15, 2026

Uh oh!

climbfuji commented Aug 17, 2026

Uh oh!

scrasmussen commented Aug 18, 2026

Uh oh!

climbfuji commented Aug 18, 2026

Uh oh!

grantfirl commented Aug 26, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

grantfirl commented Aug 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

grantfirl commented Aug 26, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

scrasmussen commented Aug 15, 2026 •

edited

Loading

grantfirl commented Aug 26, 2026 •

edited

Loading