CCPPize Hack shallow convection #195

jimmielin · 2025-01-27T17:02:52Z

Originator(s): @jimmielin

Summary (include the keyword ['closes', 'fixes', 'resolves'] and issue number):

Closes CCPP-ize Hack shallow convection #136: CCPPize Hack shallow convection scheme (cmfmca) and related interstitials
Move ZM namelist read to a separate zm_conv_options scheme so not all of ZM has to run while ZM namelist options are read for zm_conv_evap (needed for complete shallow convection)
Closes ZM convective mass flux (atmosphere_convective_mass_flux_due_to_deep_convection) units inconsistent with snapshot #197: Changes ZM convective mass flux units to kg m-2 s-1
Closes Add errmsg, errflg to cloud_fraction_fice_run #201: Adds errmsg, errflg to cloud_fraction_fice sxheme
Fixes Adjust trop_cloud_top_lev for cam4 runs CAM-SIMA#284: Updates top_lev in cloud_fraction_fice to a registry quantity that can be defined by macrophysics initialization
Fixes ZM reserved liquid unit metadata to be m s-1
Freeze ChangeLog

Describe any changes made to the namelist:

new hack shallow convection namelist options cmfmca_namelist.xml
move ZM namelist options to zm_conv_options_namelist.xml

List all files eliminated and why: N/A

List all files added and what they do:

CCPPize Hack shallow convection
A       schemes/hack_shallow/cmfmca.F90
A       schemes/hack_shallow/cmfmca.meta
A       schemes/hack_shallow/cmfmca_namelist.xml
A       test/test_suites/suite_convect_shallow_hack.xml
A       schemes/sima_diagnostics/convect_shallow_diagnostics.F90
A       schemes/sima_diagnostics/convect_shallow_diagnostics.meta

Interstitial schemes for zm_conv_evap
A       schemes/hack_shallow/set_general_conv_fluxes_to_shallow.F90
A       schemes/hack_shallow/set_general_conv_fluxes_to_shallow.meta
A       schemes/hack_shallow/set_shallow_conv_fluxes_to_general.F90
A       schemes/hack_shallow/set_shallow_conv_fluxes_to_general.meta

Interstitial schemes to combine shallow and deep convective fluxes
A       schemes/hack_shallow/convect_shallow_sum_to_deep.F90
A       schemes/hack_shallow/convect_shallow_sum_to_deep.meta

Move ZM namelist options to zm_conv_options scheme
A       schemes/zhang_mcfarlane/zm_conv_options.F90
A       schemes/zhang_mcfarlane/zm_conv_options.meta

Update cloud_fraction_fice to accept top_lev registry quantity (vertical_layer_index_of_cloud_fraction_top)
Add new init-phase only scheme to assign top_lev to trop_cloud_top_lev for CAM5+ macrophysics
M       schemes/cloud_fraction/cloud_fraction_fice.F90
M       schemes/cloud_fraction/cloud_fraction_fice.meta
A       schemes/cloud_fraction/set_cloud_fraction_top.F90
A       schemes/cloud_fraction/set_cloud_fraction_top.meta

Change CMFMC_DP units for consistency with snapshot; update rliq metadata unit
M       schemes/zhang_mcfarlane/zm_convr.F90
M       schemes/zhang_mcfarlane/zm_convr.meta
M       schemes/sima_diagnostics/zm_diagnostics.F90
M       schemes/sima_diagnostics/zm_diagnostics.meta

List all existing files that have been modified, and describe the changes:
(Helpful git command: git diff --name-status development...<your_branch_name>)

Freeze ChangeLog
M       doc/ChangeLog

Move ZM namelist options to zm_conv_options scheme
R100    schemes/zhang_mcfarlane/zm_convr_namelist.xml   schemes/zhang_mcfarlane/zm_conv_options_namelist.xml
M       suites/suite_cam7.xml
M       test/test_suites/suite_zhang_mcfarlane.xml

Update cloud_fraction_fice to accept top_lev registry quantity (vertical_layer_index_of_cloud_fraction_top)
M       schemes/cloud_fraction/cloud_fraction_fice.F90
M       schemes/cloud_fraction/cloud_fraction_fice.meta
M       suites/suite_cam7.xml
M       test/test_suites/suite_zhang_mcfarlane.xml

List any test failures:

Is this a science-changing update? New physics package, algorithm change, tuning changes, etc?
New Hack shallow convection scheme

…_to_deep updates

…ter uses none as units

…_saturation

…m_tend

peverwhee

nothing massive - just a few requests!

thanks @jimmielin woo! hack shallow!

peverwhee · 2025-01-28T05:01:33Z

schemes/hack_shallow/cmfmca.F90

+! Hack shallow convective scheme.
+!
+! Original Author: J. Hack
+! CCPPized: Haipeng Lin, October 2024


Can I request porting the full comment over from CAM, to preserve that history?

Suggested change

! Hack shallow convective scheme.

!

! Original Author: J. Hack

! CCPPized: Haipeng Lin, October 2024

!-----------------------------------------------------------------------

!

! Purpose:

! Moist convective mass flux procedure:

!

! Method:

! If stratification is unstable to nonentraining parcel ascent,

! complete an adjustment making successive use of a simple cloud model

! consisting of three layers (sometimes referred to as a triplet)

!

! Code generalized to allow specification of parcel ("updraft")

! properties, as well as convective transport of an arbitrary

! number of passive constituents (see q array). The code

! is written so the water vapor field is passed independently

! in the calling list from the block of other transported

! constituents, even though as currently designed, it is the

! first component in the constituents field.

!

! Author: J. Hack

!

! BAB: changed code to report tendencies in cmfdt and dq, instead of

! updating profiles. Cmfdq contains water only, made it a local variable

! made dq (all constituents) the argument.

!

! CCPPized: Haipeng Lin, October 2024

!-----------------------------------------------------------------------

Done, thanks! Added to above cmfmca_run (lightly edited to collapse in comment by BAB)

peverwhee · 2025-01-28T05:14:46Z

schemes/hack_shallow/cmfmca.F90

+    ! in-module parameters (hardcoded)
+    dzmin = 0.0_kind_phys
+    betamn = 0.10_kind_phys
+
+    ! specify that relaxation timescale should be applied to column as opposed to triplets individually
+    rlxclm = .true.


can you just set these above at the module level, as you do ssfac, etc?

Thanks, done!

peverwhee · 2025-01-28T05:15:20Z

schemes/hack_shallow/cmfmca.F90

+    integer,         intent(in)     :: pcnst              ! number of ccpp constituents
+    type(ccpp_constituent_prop_ptr_t), &
+                     intent(in)     :: const_props(:)     ! ccpp constituent properties pointer
+    integer,         intent(in)     :: nstep              ! current time step index


I don't think this is used - can remove from argument list and metadata.

Done, thanks!

peverwhee · 2025-01-28T05:19:15Z

schemes/hack_shallow/cmfmca.F90

+    ! temporary for debug
+    use cam_logfile,               only: iulog


Should this be an input argument to cmfmca_run ?

Yes, thanks! Updated

peverwhee · 2025-01-28T05:22:06Z

schemes/hack_shallow/cmfmca.meta

@@ -0,0 +1,281 @@
+[ccpp-table-properties]
+  name = cmfmca
+  type = scheme


i believe this should include the wv_saturation dependency

dependencies = ../../to_be_ccppized/wv_saturation.F90

Added, thank you!

peverwhee · 2025-01-28T05:25:41Z

schemes/hack_shallow/cmfmca.F90

+    const_check_loop: do m = 1, pcnst
+      call const_props(m)%standard_name(const_standard_name)
+      if (const_standard_name == 'water_vapor_mixing_ratio_wrt_moist_air_and_condensed_water') then
+        const_wv_idx = m
+        exit const_check_loop
+      endif
+    enddo const_check_loop


I'd recommend using the ccpp_const_get_idx in ccpp_const_utils that you added. I hope that'll make it easier to find and replace with the framework version eventually...

Thanks, updated!

peverwhee · 2025-01-28T05:33:16Z

schemes/hack_shallow/cmfmca_namelist.xml

+      <!-- FIXME: This is a resolution-dependent value -->
+      <value>1.0e-4</value>


is this a "FIXME before merging this PR" or a "FIXME eventually when we have some capability"? If it's the latter, I'd suggest a comment describing what we need to implement

Good question... it was more of a reminder that eventually the resolution-dependent options need to be ported, but from the CAM-SIMA docs on namelist XML files it looks like this work is implied: so I got rid of the comment.

When populating the value fields, you don't need to port all of them at this time. You need to put in the value that is used by default by CAM, but the rest will be ported once all schemes have been ported.

peverwhee · 2025-01-28T05:46:59Z

schemes/sima_diagnostics/convect_shallow_diagnostics.F90

+    call history_out_field('CMFDQR', cmfdqr)
+
+    ! Calculate fractional occurrence of shallow convection
+    ! FIXME (hplin): this definition looks counter-intuitive but is replicated verbatim from convect_shallow.F90. To check.


did you check? :)

Thanks for the reminder, I did not! @adamrher, could you please help check if this is correct? This is from convect_shallow.F90 - The definition of this just feels backwards to me, since it marks a column as having shallow convection when the shallow convection cloud mass flux is zero.

freqsh(:) = 0._kind_phys do i = 1, ncol if(maxval(cmfmc_sh(i,:)) <= 0._kind_phys) then freqsh(i) = 1._kind_phys endif enddo call history_out_field('FREQSH', freqsh)

Hi @jimmielin sorry I missed this git prompt. That is indeed strange -- a value of 1 should be assigned to the column when there is shallow convection. The comment above it "! Modification : I should check whether below computation of freqsh is correct." suggests we aren't the only one who thinks so (can you git blame this comment?).

I looked through hk_conv.F90 and verified that the array (cmfmc2, which I presume is cmfmc_sh in your code) cannot be negative. There are explicit limiters on the components that are summed and multiplied to give cmfmc2: beta and eta:

https://github.com/ESCOMP/CAM/blob/6eb3046284ef57385ca7612165465668c97ec1f5/src/physics/cam/hk_conv.F90#L697.

And beta is limited using beta(i) = max(0.0_r8,beta(i)) statements anytime it's modified. So I believe this is a bug in this particular diagnostic, and that the logic should be:

if (maxval(cmfmc_sh(i,:)) > 0._kind_phys) then

Thanks @adamrher! git blame shows that comment is before the Git era, unfortunately.

I've opened an issue in CAM: ESCOMP/CAM#1254. I can work on a PR to fix this (probably a separate one from the CCPPization PR, so that the bit-for-bit property of the CCPPization PR can be preserved.)

I went back to svn and a blame there yields:
mvr 4/29/2010 11:49:39 AM updated the physics with code from the cam5 development branch; some changes to the dynamics, including the enabling of fourth-order divergence damping
I don't remember who mvr is, but that person was the person who did the very first svn commit in CAM, so an experienced programmer. Maybe Matt Rothstein (or maybe I'm making up names!)

That was Matt's username. Oh well. Diagnostics can be tough to trouble shoot as the simulation may still be bfb.

Thanks @cacraigucar and @swrneale. It looks like FREQSH is not in the default field list for CAM4 so indeed if I change the definition of this diagnostic it will still show up as b4b in the regression tests. Anyway, I have fixed it on the CAM-SIMA end first (which has separate diagnostics code from current CAM) and can address it in ESCOMP/CAM separately.

On second thought, it might not be worth the extra time fixing this diagnostic in ESCOMP/CAM, before you merge the PR to ESCOMP/CAM to call Hack from the atmospheric_physics version of convect_shallow.F90.

But I'm confused on one aspect. If we were running non-CCPP' cam5 physics in ESCOMP/CAM, which is also impacted by this bug, would it then use the convect_shallow in atmospheric_physics? Or are you separating just the cam4 parts out of convect_shallow for this Hack shallow task?

My apologies for the confusion! For this freqsh definition in particular, the code used for calculating it in CAM-SIMA is separate because it is in the "diagnostic" scheme, where the non-portable calls to the CAM history code reside. (schemes/sima_diagnostics/convect_shallow_diagnostics.F90, line ~191). The freqsh definition in ESCOMP/CAM will remain in convect_shallow.F90. So I would have to fix it in both atmospheric_physics and ESCOMP/CAM.

There is no direct equivalent to convect_shallow.F90 in atmos_phys -- hk_conv.F90 (subroutine cmfmca) for Hack shallow is CCPPized (into cmfmca.F90), so the only part that is shared by CAM and CAM-SIMA is the cmfmca_run and cmfmca_init subroutines.

There's a variety of other schemes in this PR, e.g., schemes/hack_shallow/convect_shallow_sum_to_deep.F90 (combine shallow and deep convective fluxes to create "all" convective fluxes) and the diagnostics scheme (calls history_add_field and outfld) that do similar things as what convect_shallow.F90 and convect_diagnostics.F90 do, but in the SIMA way, so they're not used in ESCOMP/CAM (because history calls are different).

I'm happy to elaborate further if you have any questions. I realize this may be easier to explain once I open the ESCOMP/CAM PR and you will be able to see what's going on.

OK I understand now. Thanks for straightening me out and I agree we'll need fix it in ESCOMP/CAM.

I actually really like this idea of dissolving the whole concept convect_shallow into portable interfaces and interstitials. It really helps clarify what parts of the code are used.

…with snapshot

jimmielin added 30 commits January 24, 2025 17:42

Initial commit of cmfmca_init and cmfmca_run

3e1026e

Update _sh suffixes to out args; kind_phys from r8

4db372e

Remove pverp; note hack for qpert change in-scheme

6c97a5b

Fixes for B4B in CAM

a2d5a6a

Temporary check-in of convect_shallow (hack)

16d73bd

Fill out remaining standard names

25ba7f8

Update standard names based on pbuf and snapshot entries; add_shallow…

d23b650

…_to_deep updates

Update to use constituent properties ptr and tendency updater

65a5d4d

Add errmsg, errflg

98470fc

Remove dependency on deep conv. quantities; constituent tendency upda…

2184851

…ter uses none as units

Make qpert horiz only; initialize constituents (needs slipstreamed fix)

947072b

Fix for qpert remove higher dims; initialize (temp) to-be-ccppized wv…

41e972b

…_saturation

Add to_be_ccppized_temporary shared with ZM

396f5ee

Add debug diagnostics from cmfmca

044d9e9

Remove dependency on q being in index 1

354189e

Add geopotential_temp scheme after heating_rate

1b799b0

Sketch out diagnostics (wip)

075758e

Sketch out convect_shallow_diagnostics take 2

6d4a788

Add interstitial schemes in preparation for zm_conv_evap

9ad9652

Build time fixes

ec432da

Add ZMDLF and SHDLF convective reserved water m s-1 from rk_stratifor…

28d66f6

…m_tend

Add zm_conv_evap for convect_shallow_hack

f3744be

Add cloud_fraction_fice call for zm_conv_evap

232952f

Move zm namelist read to separate scheme for sharing by SDFs

54ef3c9

Add ZMDLF and SHDLF convective reserved water m s-1 from rk_stratifor…

18230c1

…m_tend

Add check_energy_chng bindings to cmfmca scheme

cec9680

Remove units.F90 from to_be_ccppized

015ea23

Remove changes to initialize_constituents

a1253e5

Remove unused interstitial

dae06a7

Remove ZM ChangeLog -- ChangeLog now frozen in time

3ed4ea7

jimmielin added 3 commits January 27, 2025 11:25

Move diagnostics under debug_verbose logical

a7a0b38

Finish up sum_to_deep and convect_shallow diagnostics

8016400

Add zm_conv_options to CAM7 suite

591481d

jimmielin added the enhancement New feature or request label Jan 27, 2025

jimmielin self-assigned this Jan 27, 2025

Fix remaining b4b issues with sum-to-deep interstitial

a45b68d

jimmielin requested review from cacraigucar and peverwhee January 27, 2025 20:51

peverwhee requested changes Jan 28, 2025

View reviewed changes

jimmielin added 2 commits January 28, 2025 12:42

Address review comments

54cf50f

Address review comments (try 2)

bfac2a8

peverwhee requested a review from nusbaume January 29, 2025 17:29

jimmielin added 12 commits January 30, 2025 13:50

Add CAM4 suite; fix cmfmc to use vertical_interface_dimension

4d4ca68

Convert cmfmc_dp to kg m-2 s-1 within ZM scheme for consistent units …

ffeb80b

…with snapshot

Fix ZM rliq units (and note that they were converted to precip units)

d460908

jctop, bot is integer from ZM scheme

aaa10fb

Add use_shfrc support for interfacing with cloud_fraction scheme

ebcda86

Add errmsg, errflg to cloud_fraction_fice scheme

fc1e812

Correct freqsh calculation.

fb6275c

Fix regression tests; add missing debug output

8b32b28

Update top_lev for cloud_fraction_fice; improve cmfmca_init

1564399

Fix metadata for new cmfmca_init

7e5e8e1

Merge branch 'development' into hplin/hk_conv_rebase

414b6a6

Update trop_cloud_top_lev standard name

ac51af8

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

CCPPize Hack shallow convection #195

CCPPize Hack shallow convection #195

jimmielin commented Jan 27, 2025 •

edited

Loading

peverwhee left a comment

peverwhee Jan 28, 2025

jimmielin Jan 28, 2025

peverwhee Jan 28, 2025

jimmielin Jan 28, 2025

peverwhee Jan 28, 2025

jimmielin Jan 28, 2025

peverwhee Jan 28, 2025

jimmielin Jan 28, 2025

peverwhee Jan 28, 2025

jimmielin Jan 28, 2025

peverwhee Jan 28, 2025

jimmielin Jan 28, 2025

peverwhee Jan 28, 2025 •

edited

Loading

jimmielin Jan 28, 2025

peverwhee Jan 28, 2025

jimmielin Jan 28, 2025

adamrher Feb 10, 2025

jimmielin Feb 10, 2025

cacraigucar Feb 10, 2025

swrneale Feb 10, 2025

jimmielin Feb 10, 2025

adamrher Feb 11, 2025

jimmielin Feb 11, 2025 •

edited

Loading

adamrher Feb 11, 2025

-! Hack shallow convective scheme.
-!
-! Original Author: J. Hack
-! CCPPized: Haipeng Lin, October 2024
+!-----------------------------------------------------------------------
+!
+! Purpose:
+! Moist convective mass flux procedure:
+!
+! Method:
+! If stratification is unstable to nonentraining parcel ascent,
+! complete an adjustment making successive use of a simple cloud model
+! consisting of three layers (sometimes referred to as a triplet)
+!
+! Code generalized to allow specification of parcel ("updraft")
+! properties, as well as convective transport of an arbitrary
+! number of passive constituents (see q array).  The code
+! is written so the water vapor field is passed independently
+! in the calling list from the block of other transported
+! constituents, even though as currently designed, it is the
+! first component in the constituents field.
+!
+! Author: J. Hack
+!
+! BAB: changed code to report tendencies in cmfdt and dq, instead of
+! updating profiles. Cmfdq contains water only, made it a local variable
+! made dq (all constituents) the argument.
+!
+! CCPPized: Haipeng Lin, October 2024
+!-----------------------------------------------------------------------

		<!-- FIXME: This is a resolution-dependent value -->
		<value>1.0e-4</value>

CCPPize Hack shallow convection #195

Are you sure you want to change the base?

CCPPize Hack shallow convection #195

Conversation

jimmielin commented Jan 27, 2025 • edited Loading

peverwhee left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

peverwhee Jan 28, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jimmielin Feb 11, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jimmielin commented Jan 27, 2025 •

edited

Loading

peverwhee Jan 28, 2025 •

edited

Loading

jimmielin Feb 11, 2025 •

edited

Loading