merge new clubb_mf into cam6_4_180 - #1576
Conversation
…rations on several variables with latest clubb interface and externals.
There was a problem hiding this comment.
Pull request overview
This PR merges updated CLUBB+MF (clubb-mf) changes into CAM, extending the CLUBB interface to support the enhanced integrate_mf call, adding new EDMF diagnostics/outputs, and introducing additional namelist controls for CLUBB+MF behavior (radiation coupling, microphysics, TKE contribution, etc.).
Changes:
- Expanded
clubb_intr.F90CLUBB+MF interface: new pbuf fields, history coords, diagnostic outputs, and enhanced MF plumbing. - Added new CLUBB+MF namelist entries + defaults, and wired them into
build-namelist. - Updated COSP namelist documentation text (but introduced encoding issues that need correction).
Reviewed changes
Copilot reviewed 4 out of 5 changed files in this pull request and generated 18 comments.
| File | Description |
|---|---|
src/physics/cam/clubb_intr.F90 |
Adds/updates CLUBB+MF fields, diagnostics, and MF integration interface wiring. |
bld/namelist_files/namelist_definition.xml |
Adds new CLUBB+MF namelist definitions and edits COSP localtime documentation. |
bld/namelist_files/namelist_defaults_cam.xml |
Adds default values for the new CLUBB+MF namelist entries. |
bld/build-namelist |
Adds add_default(...) plumbing for (most) new CLUBB+MF namelist variables. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| !jt check with Adam These are not bfb with nonclubbmf run | ||
| !jt check with Adam ! Values to use above top_lev, for variables that have not already been | ||
| !jt check with Adam ! set up there. These are mostly fill values that should not actually be | ||
| !jt check with Adam ! used in the run, but may end up in diagnostic output. | ||
| !jt check with Adam !$acc parallel loop gang vector collapse(2) default(present) | ||
| !jt check with Adam do k=1, top_lev-1 | ||
| !jt check with Adam do i=1, ncol | ||
| !jt check with Adam upwp_pbuf(i,k) = 0._r8 | ||
| !jt check with Adam vpwp_pbuf(i,k) = 0._r8 | ||
| !jt check with Adam rcm(i,k) = 0._r8 | ||
| !jt check with Adam wprcp(i,k) = 0._r8 | ||
| !jt check with Adam cloud_frac(i,k) = 0._r8 | ||
| !jt check with Adam rcm_in_layer(i,k) = 0._r8 | ||
| !jt check with Adam zt_output(i,k) = 0._r8 | ||
| !jt check with Adam zi_output(i,k) = 0._r8 | ||
| !jt check with Adam khzm_pbuf(i,k) = 0._r8 | ||
| !jt check with Adam qclvar(i,k) = 2._r8 | ||
| !jt check with Adam end do | ||
| !jt check with Adam end do |
There was a problem hiding this comment.
I'm still looking into this section as one of these variables is causing numberical differences.
| mf_upa_flip(i,k,:clubb_mf_nup) = mf_upa(i,k_clubb,:clubb_mf_nup) | ||
| mf_upw_flip(i,k,:clubb_mf_nup) = mf_upw(i,k_clubb,:clubb_mf_nup) | ||
| mf_upmf_flip(i,k,:clubb_mf_nup) = mf_upmf(i,k_clubb,:clubb_mf_nup) | ||
| mf_upqt_flip(i,k,:clubb_mf_nup) = mf_upqt(i,k_clubb,:clubb_mf_nup) | ||
| mf_upthl_flip(i,k,:clubb_mf_nup) = mf_upthl(i,k_clubb,:clubb_mf_nup) | ||
| mf_upthv_flip(i,k,:clubb_mf_nup) = mf_upthv(i,k_clubb,:clubb_mf_nup) | ||
| mf_upth_flip(i,k,:clubb_mf_nup) = mf_upth(i,k_clubb,:clubb_mf_nup) | ||
| mf_upqc_flip(i,k,:clubb_mf_nup) = mf_upqc(i,k_clubb,:clubb_mf_nup) | ||
| mf_upent_flip(i,k,:clubb_mf_nup) = mf_upent(i,k_clubb,:clubb_mf_nup) | ||
| mf_updet_flip(i,k,:clubb_mf_nup) = mf_updet(i,k_clubb,:clubb_mf_nup) | ||
| mf_upbuoy_flip(i,k,:clubb_mf_nup) = mf_upbuoy(i,k_clubb,:clubb_mf_nup) | ||
| mf_dnw_flip(i,k,:clubb_mf_nup) = mf_dnw(i,k_clubb,:clubb_mf_nup) | ||
| mf_dnthl_flip(i,k,:clubb_mf_nup) = mf_dnthl(i,k_clubb,:clubb_mf_nup) | ||
| mf_dnqt_flip(i,k,:clubb_mf_nup) = mf_dnqt(i,k_clubb,:clubb_mf_nup) |
There was a problem hiding this comment.
Testing alternate solution to flip array.
There was a problem hiding this comment.
Tested and removed these.
| @@ -1,684 +1,3201 @@ | |||
| module clubb_mf | |||
|
|
|||
| ! =============================================================================== ! | |||
There was a problem hiding this comment.
@adamrher You'll probably want to difference this code against your cam6_4_124 clubbmf merge to see my changes. The code is now top-down, bottom-up agnostic so the indexing and loops have all changed. I tried to put some commenting in about the generalization of the vertical coordinates but like the original its still pretty dense. Additionally the vertical dimensions for momentum and thermodynamic arrays are consistent with those grids. It would help if you could make sure the different arrays are dimensioned for the correct grid as well as checking some of the special if logic loops that are not executed by default to make sure the vertical indexing is OK. I've checked all loops that are executed by default and they are roundoff consistent (12-14 digits) consistent with the order of operation differences between this code and your cam6_4_124 clubbmf merge.
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
| @@ -4773,6 +5598,17 @@ subroutine clubb_tend_cam( state, ptend_all, pbuf, hdtime, & | |||
| end do | |||
| end do | |||
|
|
|||
| rcm_macmic(:ncol,nzt_clubb*(macmic_it-1)+1:nzt_clubb*macmic_it) = rcm(:ncol,:nzt_clubb) | |||
| cldfrac_macmic(:ncol,pver*(macmic_it-1)+1:pver*macmic_it) = cld_pbuf(:ncol,:pver) | ||
| wpthlp_macmic(:ncol,pverp*(macmic_it-1)+1:pverp*macmic_it) = wpthlp_output(:ncol,:pverp) | ||
| wprtp_macmic(:ncol,pverp*(macmic_it-1)+1:pverp*macmic_it) = wprtp_output(:ncol,:pverp) | ||
| wpthvp_macmic(:ncol,nzm_clubb*(macmic_it-1)+1:nzm_clubb*macmic_it) = wpthvp_pbuf(:ncol,:nzm_clubb) |
adamrher
left a comment
adamrher
left a comment
There was a problem hiding this comment.
Hi John. I've completed my first round of the code review, getting about about half-way through clubb_intr. The code merge removed a lot of stuff that Gunther cleaned up in cam6_4_144, and I tried to catch all those instances. I also found about 8 or so *_macmic variables that I'd like to remove from the pbuf.
| thlm_zm(:,:) = 0._r8 | ||
| th_zm(:,:) = 0._r8 | ||
| qv_zm(:,:) = 0._r8 | ||
| qc_zm(:,:) = 0._r8 |
There was a problem hiding this comment.
are these zero array initializations unnecessary?
| do i = 1, ncol | ||
| ! Precip accumulates downward across the interfaces | ||
| mf_precc(i,k+1) = mf_precc(i,k) - rho_zt(i,k)*dz_g(i,k)*mf_sqt(i,k) | ||
| end do | ||
| end do | ||
|
|
||
| !--------------------------------------- END integrate_mf call --------------------------------------- |
There was a problem hiding this comment.
This 'END integrate_mf call' comment should be moved up to the end do immediately after the integrate_mf call.
| @@ -3565,6 +4112,7 @@ subroutine clubb_tend_cam( state, ptend_all, pbuf, hdtime, & | |||
| rho_ds_zt = rho_ds_zt(:,nzt_clubb:1:-1) | |||
| invrs_rho_ds_zt = invrs_rho_ds_zt(:,nzt_clubb:1:-1) | |||
| thv_ds_zt = thv_ds_zt(:,nzt_clubb:1:-1) | |||
| khzt = khzt(:,nzt_clubb:1:-1) | |||
There was a problem hiding this comment.
I guess this is technically a bug fix, but note this variable is not used anywhere in the model.
There was a problem hiding this comment.
I fixed it just for consistency sake. As it is an output parameter is will be on the correct levels if we decide to use it later.
There was a problem hiding this comment.
I'll leave it in case khzt is used in the future, it will be on the correct levels.
| clubb_params(i,ic_K) * Lscale(i,k) * sqrt(0.5_r8 * s_aww(i,k)) | ||
| end do | ||
| end do | ||
| else | ||
| !$acc parallel loop gang vector collapse(2) default(present) |
There was a problem hiding this comment.
Done. Indented to 2 spaces like rest of routine.
| mf_qc(:ncol,:nzm_clubb) = mf_qc_nadv(:ncol,:nzm_clubb)/REAL(nadv) | ||
| mf_rcm(:ncol,:nzm_clubb) = mf_rcm_nadv(:ncol,:nzm_clubb)/REAL(nadv) | ||
| mf_cloudfrac(:ncol,:nzm_clubb) = mf_cloudfrac_nadv(:ncol,:nzm_clubb)/REAL(nadv) | ||
| prec_sh(:ncol) = mf_precc_nadv(:ncol)/REAL(nadv) |
There was a problem hiding this comment.
This is a merge bug. This prec_sh pbuf variable needs to be defined before the energy fixer (line ~4696). This entire code block could probably be brought up to just before the energy fixer starts (line ~4649) with an appropriate comment ("process accumulated clubb-mf arrays"). The energy fixer needs prec_sh.
also, to conform with the cam6_4_166 clubb PR, can we append the prec_sh and snow_sh variables with "_pbuf"?
| mf_cloudfrac_zt = zm2zt_api( nzm_clubb, nzt_clubb, ncol, gr, mf_cloudfrac) | ||
| !mf_qc_zt(1:ncol,:) = zm2zt_api( pverp+1-top_lev, ncol, gr, mf_qc(1:ncol,:)) | ||
| !mf_cloudfrac_zt(1:ncol,:) = zm2zt_api( pverp+1-top_lev, ncol, gr, mf_cloudfrac(1:ncol,:)) | ||
|
|
There was a problem hiding this comment.
A few issues with these two variables (1) bring the initializations into the do_clubb_mf block (2) delete the commented out code, (3) these variables are supposed to set mf_qc_output & mf_cloudfrac_output, but those output arrays are not being set to anything. We need to set those output arrays since they are used to set cloud radiative pbuf vars down below (line ~5300).
There was a problem hiding this comment.
It turns out none of the output variables declared in this block (line ~2414) are being set to anything:
real(r8), dimension(pcols,pver) :: mf_thlforcup_output, mf_qtforcup_output, & ! thermodynamic grid
mf_thlforcdn_output, mf_qtforcdn_output, & ! thermodynamic grid
mf_thlforc_output, mf_qtforc_output, & ! thermodynamic grid
mf_ent_output, & ! thermodynamic grid
mf_sqtup_output, mf_sqtdn_output, & ! thermodynamic grid
mf_qc_output, mf_cloudfrac_output ! thermodynamic grid
There was a problem hiding this comment.
I've set all of these missing outputs, used array syntax instead of the do loops and got rid of the flip variables.
| !$acc parallel loop gang vector default(present) | ||
| do i=1, pcols | ||
| zi_output(i,1) = 0._r8 | ||
| end do |
There was a problem hiding this comment.
I don't know what this code block is, and it's possible I added it, but it should be deleted.
| end do | ||
|
|
||
| end if | ||
| end do |
There was a problem hiding this comment.
I think this was pulled out in the cam6_4_166 PR. the edsclr array is size nzt_clubb, and so no 1:top_lev exists. edsclr is not even used after this block.
There was a problem hiding this comment.
Removed this block.
| ! "cloud_frac") and CLUBB+MF cloud fraction ("shalcu") compute the convective cloud | ||
| ! fraction. This follows the formulation found in macrophysics code. Assumes that convective | ||
| ! cloud is all nonstratiform cloud from CLUBB or CLUBB+MF plus the deep convective cloud fraction | ||
| concld_pbuf(i,k) = min(cloud_frac(i,k)-alst_pbuf(i,k)+deepcu_pbuf(i,k)+shalcu_pbuf(i,k),0.80_r8) |
There was a problem hiding this comment.
can you adapt this to follow how this is done in the deepcu block above? Specifically, comment out the current concld_pbuf line and add a new line concld_pbuf(i,k) = min(deepcu_pbuf(i,k)+shalcu_pbuf(i,k),0.80_r8), with a NOTE explaining what we're doing.
| mf_thlforc_output, mf_qtforc_output, & ! thermodynamic grid | ||
| mf_ent_output, & ! thermodynamic grid | ||
| mf_sqtup_output, mf_sqtdn_output, & ! thermodynamic grid | ||
| mf_qc_output, mf_cloudfrac_output ! thermodynamic grid |
There was a problem hiding this comment.
It looks like these are still in the same place?
…ng vs. descending
clean up function arguments in integrate_mf + bug fixes. John will improve the initialization for mcape/mf_cape and correct spelling errors identified by copilot.
This PR merges updated CLUBB+MF (clubb-mf) changes into CAM, extending the CLUBB interface to support the enhanced integrate_mf call, adding new EDMF diagnostics/outputs, and introducing additional namelist controls for CLUBB+MF behavior (radiation coupling, microphysics, TKE contribution, etc.).
Changes:
The majority of the code updates reside in clubb_mf.F90. In addition to the new science, clubb_mf is now agnostic to the direction of the vertical level, adjusting the looping and indexes as needed depending on the top down or bottom up nature of the input arrays. Additionally the number of vertical levels for the new mf arrays are consistent with their associated thermodynamic or momentum grid. There are order of operation differences between this update and Adam's original cam6_4_124 version of clubb_mf which cause roundoff precision differences from the base code. On top of these roundoff differences there are bug fixes which are answer changing.
This PR closes #1370