Grass phenology update

[xuchongang]

This pull request update the phenology for grass. Namely, to drop stems during cold or drought conditions.

Description:

We made the following key modifications:

1. add phen_stem_drop_fraction into the parameter file and EDPhysiologyMod so that the grass can drop their stem during winter or drought;
2. make the leaf phenology for grass size-dependent. This modification is made so that it can accommodate the situation of small salt marsh can overwinter due to the protection of tall dead stems.

Collaborators:

Lu Zhai, [email protected]

Expectation of Answer Changes:

For grass, it will change the phenology if phen_stem_drop_fraction set to values > 0. Otherwise, it should generate the same output as before.

Checklist:

• [] My change requires a change to the documentation.
• I have updated the in-code documentation .AND. (the technical note .OR. the wiki) accordingly.
• [X ] I have read the CONTRIBUTING document.
• FATES PASS/FAIL regression tests were run
• If answers were expected to change, evaluation was performed and provided

Test Results:

Not yet

CTSM (or) E3SM (specify which) test hash-tag:

CTSM (or) E3SM (specify which) baseline hash-tag:

FATES baseline hash-tag:

Test Output:

[glemieux]

My local branch of this PR is failing on the exact restart test for the mpi-serial library. It looks like sapwmemory is being set to NaN instead of zero or the previous value. Currently investigating.

[glemieux]

@LuZhai @xuchongang I noticed in EDInitMod that the init_cohorts subroutine does not include code for sapwmemory or structmemory for deciduous stress:

fates/main/EDInitMod.F90

Lines 506 to 522 in bfe83fd

	if( EDPftvarcon_inst%season_decid(pft) == itrue .and. &
	any(site_in%cstatus == [phen_cstat_nevercold,phen_cstat_iscold])) then
	temp_cohort%laimemory = c_leaf
	temp_cohort%sapwmemory = c_sapw * stem_drop_fraction
	temp_cohort%structmemory = c_struct * stem_drop_fraction
	c_leaf = 0._r8
	c_sapw = (1.0_r8-stem_drop_fraction) * c_sapw
	c_struct = (1.0_r8-stem_drop_fraction) * c_struct
	cstatus = leaves_off
	endif
	if ( EDPftvarcon_inst%stress_decid(pft) == itrue .and. &
	any(site_in%dstatus == [phen_dstat_timeoff,phen_dstat_moistoff])) then
	temp_cohort%laimemory = c_leaf
	c_leaf = 0._r8
	cstatus = leaves_off
	endif

whereas in FatesInventoryInitMod, there is matching code for both stress and seasonality in set_inventory_edcohort_type1:

fates/main/FatesInventoryInitMod.F90

Lines 1045 to 1065 in bfe83fd

	if( EDPftvarcon_inst%season_decid(temp_cohort%pft) == itrue .and. &
	any(csite%cstatus == [phen_cstat_nevercold,phen_cstat_iscold])) then
	temp_cohort%laimemory = b_leaf
	temp_cohort%sapwmemory = b_sapwood * stem_drop_fraction
	temp_cohort%structmemory = b_dead * stem_drop_fraction
	b_leaf = 0._r8
	b_sapwood = (1._r8 - stem_drop_fraction) * b_sapwood
	b_dead = (1._r8 - stem_drop_fraction) * b_dead
	cstatus = leaves_off
	endif
	if ( EDPftvarcon_inst%stress_decid(temp_cohort%pft) == itrue .and. &
	any(csite%dstatus == [phen_dstat_timeoff,phen_dstat_moistoff])) then
	temp_cohort%laimemory = b_leaf
	temp_cohort%sapwmemory = b_sapwood * stem_drop_fraction
	temp_cohort%structmemory = b_dead * stem_drop_fraction
	b_leaf = 0._r8
	b_sapwood = (1._r8 - stem_drop_fraction) * b_sapwood
	b_dead = (1._r8 - stem_drop_fraction) * b_dead
	cstatus = leaves_off
	endif

Should EDInitMod have the initialization code for sapwmemory and structmemory for deciduous stress?

[xuchongang]

@glemieux , good catch and we need to let EDInitMod have the initialization code for sapwmemory and structmemory for deciduous stress.

[glemieux]

@glemieux , good catch and it we need to have EDInitMod have the initialization code for sapwmemory and structmemory for deciduous stress.

Thanks; I'll add it on my local branch. I fixed the restart error as well just now. New PR coming to your branch sometime later today.

[glemieux]

That was interesting. I somehow added labels using keyboard shortcuts. That's what I get for using two screens and not paying attention to both

[xuchongang]

@glemieux, thanks and I did not see the pull request on my branch yet.

[glemieux]

@xuchongang apologies for the delay. I spoke too soon; I wanted to wait for the test to complete, which was held up until the weekend in the Cheyenne queue. The test passed and I've created the PR to your branch.

[xuchongang]

@glemieux, thank you and I have made the merge of your pull request.

[glemieux]

Testing update:

All standard testing passed B4B with the default parameter file (i.e. phen_stem_drop_frac = 0) except for the SMS_Lm6.f45_f45 test which failed on a DIFF to the baseline. @rgknox and I took a look at this yesterday and we're still struggling to determine the reason for the difference. Test output can be found here:

/glade/u/home/glemieux/scratch/clmed-tests/pr554-xuchongang-grass-phen.fates.cheyenne.intel.C5ad8a8d2-Fb9a54836

I also ran a parallel 5 year test at Brazil with and without the stem drop fraction greater than zero (in this case I set all PFTs to a randomly selected value of 0.25). Output comparison plots run through the ACRE suite can be found here: https://drive.google.com/file/d/1Bj5gBPf7a13bvrG7X01ho2V6jCaiimBL/view?usp=sharing

[glemieux]

At @ckoven I plotted out the SMS_Lm6.f45_f45 differences between the baseline and the PR branch.

GPP:

TLAI:

From this and the cprnc output it looks like the problem is just with the one grid cell, for this time point at least. @rgknox suggestion I've got a longer term f45 grid running to see if the errors grow. Barring an expansion of the diff to other grids and/or the increase in the difference, do you think this is reasonable to merge, @xuchongang ?

[xuchongang]

@glemieux , for me, it is fine to merge.....

[glemieux]

@rgknox had another thought on what might be causing the non-B4B results even when the stem drop parameter is zero. The flushing logic looks like it could possibly flush only a portion of the store_c_transfer_frac since it's only checking for non-woody plants: https://github.com/NGEET/fates/pull/554/files#diff-b2580903c042bbb143bb267a7b882e3fR990-R1009

At this suggestion, I added a check during the flush to use all of the store_c_transfer_frac for leaves if the stem drop fraction is zero:
Drought leaf on section: https://github.com/glemieux/fates/blob/ca148d0ff5d4dd0470d17d9e51da2e2e90a005cd/biogeochem/EDPhysiologyMod.F90#L990-L1020
Cold leaf on section: https://github.com/glemieux/fates/blob/ca148d0ff5d4dd0470d17d9e51da2e2e90a005cd/biogeochem/EDPhysiologyMod.F90#L1103-L1135

This indeed resulted in B4B results. The question is, should we be using totalmemory for leaves when stem drop is zero like I have done here? Or is this changing the intent of the hypothesis?

[xuchongang]

@glemieux , I think "using totalmemory for leaves when stem drop is zero" sounds reasonable to me.

[glemieux]

Final testing results. All expected PASS:

/glade/u/home/glemieux/scratch/clmed-tests/pr554-phen-flush-b4b-fix-Cf8f641ab-Fca148d0f.fates.cheyenne.intel