post-5.4.n02.ctsm5.4.000: Fix for RRTMGP errors caused by high surface temperatures

[olyson]

Description of changes

           ! Add esai to sa_leaf if elai is less than threshold.
           ! Intended to avoid small sa_leaf which leads to small leaf conductance
           ! and high leaf temperature. This in turn can lead to unrealistically
           ! high surface temperatures passed to the atmospheric model (The RRTMGP
           ! component in particular, which returns an error and stops the model
           ! if the surface temperature is greater than 355K).
           ! See https://github.com/ESCOMP/CTSM/issues/3589 for more info.
           ! The 0.1_r8 value is fairly arbitrary but has been effective in
           ! avoiding RRTMGP errors in CESM3 development simulations.

Specific notes

Contributors other than yourself, if any: @swensosc , @slevis-lmwg

CTSM Issues Fixed (include github issue #): #3589

Are answers expected to change (and if so in what way)? Yes (similar climate though)

Any User Interface Changes (namelist or namelist defaults changes)? No

Does this create a need to change or add documentation? Did you do so? Yes. No.

Testing performed, if any:
See #3589 for testing description. I have not run any CTSM system testing.

[slevis-lmwg]

Approving. I will start aux_clm to confirm nothing's broken. Also I will try rebasing to the post-5.4 branch to get this in more efficiently.

[slevis-lmwg]

Rebasing didn't work, as it ended up with 44 files changed, so I reverted to master. I thought rebasing would apply the same changes but starting from post-5.4 instead of starting from master. Clearly there's something about rebasing that I don't get, because sometimes it does what I expect and sometimes not.

Maybe if I update the post_5.4 branch to ctsm5.3.085? I will think about it. I will hold off on aux_clm for now.

[ekluzek]

@olyson thanks so much for figuring this out.

I had some simple suggestions. But, then when I looked closer I saw some other things that I think need to be done. And I then saw a way to make the code both clearer to humans and more maintainable so that the contents of the if blocks didn't need to be matched if something changed. So I think it's all worth doing.

[swensosc]

I think keith has it correct. In some cases he found, stem area was sufficiently large, but leaf area was too small. The boundary layer conductances are calculated separately, so the model doesn't know that the tiny leaf is part of a larger structure that would presumably alter the conductance in this case. So the fix modifies the sa_leaf term for the purpose of increasing the conductance.

…

On Wed, Dec 3, 2025 at 1:48 PM Keith Oleson ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In src/biogeophys/CanopyFluxesMod.F90 <#3643 (comment)>: > @@ -754,6 +754,19 @@ subroutine CanopyFluxes(bounds, num_exposedvegp, filter_exposedvegp, frac_rad_abs_by_stem(p) = 0.0_r8 sa_stem(p) = 0.0_r8 sa_leaf(p) = sa_leaf(p) + esai(p) + else + ! Add esai to sa_leaf if elai is less than threshold. + ! Intended to avoid small sa_leaf which leads to small leaf conductance + ! and high leaf temperature. This in turn can lead to unrealistically + ! high surface temperatures passed to the atmospheric model (The RRTMGP + ! component in particular, which returns an error and stops the model + ! if the surface temperature is greater than 355K). + ! See #3589 for more info. + ! The 0.1_r8 value is fairly arbitrary but has been effective in + ! avoiding RRTMGP errors in CESM3 development simulations. + if(elai(p) .lt. 0.1_r8) then + sa_leaf(p) = sa_leaf(p) + esai(p) From conversations with @swensosc <https://github.com/swensosc> I don't think sa_stem should be set to zero, but pinging @swensosc <https://github.com/swensosc> here to double-check. We still have stems, we are just adding stems to very small leaf area to increase leaf conductance and sensible heat. I don't have a problem with the idea of creating a new function to handle the existing code, allow for unit testing, etc. But this is out of scope for this PR. — Reply to this email directly, view it on GitHub <#3643 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AGRN57EBN7VJY25AXXUVQO3375EC5AVCNFSM6AAAAACN3GZKY2VHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZTKMZWHA3TAOBVGQ> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[slevis-lmwg]

I will try merging the latest from post-5.4 (tag post-5.4.n01.alpha-ctsm5.4.CMIP7.21.ctsm5.3.085) and rebasing to post-5.4. If all that works, then I can start the test-suites.

[slevis-lmwg]

./run_sys_tests -s aux_clm -c post-5.4.n01.alpha-ctsm5.4.CMIP7.21.ctsm5.3.085 -g post-5.4.n02.ctsm5.4.000
derecho and izumi OK: BASELINE post-5.4.n01.alpha-ctsm5.4.CMIP7.21.ctsm5.3.085: DIFF

I looked at the cprnc file of the shortest tests (Ln9, Ld2) e.g.:
SMS_Ln9.ne3pg3_ne3pg3_mt232.I2000Clm60Sp.derecho_gnu.clm-clm60cam7LndTuningMode--clm-nofireemis.GC.1208-175929de_gnu/SMS_Ln9.ne3pg3_ne3pg3_mt232.I2000Clm60Sp.derecho_gnu.clm-clm60cam7LndTuningMode--clm-nofireemis.GC.1208-175929de_gnu.cpl.hi.2000-01-01-16200.nc.cprnc.out
All fields with diffs (greater than roundoff but small) have prefixes lnd[Imp,Exp] and rof[Imp,Exp].

[slevis-lmwg]

Keith and I looked at diffs in a 2-yr test (fields FSH, FIRE, TV, TG), and Keith approved the changes that we saw.