mksurfdata_map rounding/truncation of special landunits should happen before wetland filling

[billsacks]

Brief summary of bug

I think the current ordering of the rounding/truncation of special landunits in mksurfdata_map relative to the wetland filling could cause issues in some edge cases. These should be reversed. I haven't actually observed any problems; this comes just from looking at the code.

General bug information

CTSM version you are using: master

Does this bug cause significantly incorrect results in the model's science? No

Configurations affected: Probably unusual conditions in mksurfdata_map

Details of bug

I think this may be important to avoid errors and/or wrong behavior in some edge cases where we have < 0.5% cover of glacier and/or lake. I think the best thing to do would be to move the whole block of truncation code to above the wetland adjustment. i.e., move this block:

CTSM/tools/mksurfdata_map/src/mksurfdat.F90

Lines 737 to 747 in 6c99904

	! Truncate all percentage fields on output grid. This is needed to
	! insure that wt is zero (not a very small number such as
	! 1e-16) where it really should be zero
	do k = 1,nlevsoi
	pctsand(n,k) = float(nint(pctsand(n,k)))
	pctclay(n,k) = float(nint(pctclay(n,k)))
	end do
	pctlak(n) = float(nint(pctlak(n)))
	pctwet(n) = float(nint(pctwet(n)))
	pctgla(n) = float(nint(pctgla(n)))

to above this block:

CTSM/tools/mksurfdata_map/src/mksurfdat.F90

Lines 714 to 735 in 6c99904

	! Assume wetland, glacier and/or lake when dataset landmask implies ocean
	! (assume medium soil color (15) and loamy texture).
	! Also set pftdata_mask here
	if (pctlnd_pft(n) < 1.e-6_r8) then
	pftdata_mask(n) = 0
	soicol(n) = 15
	if (pctgla(n) < 1.e-6_r8) then
	pctwet(n) = 100._r8 - pctlak(n)
	pctgla(n) = 0._r8
	else
	pctwet(n) = max(100._r8 - pctgla(n) - pctlak(n), 0.0_r8)
	end if
	pcturb(n) = 0._r8
	call pctnatpft(n)%set_pct_l2g(0._r8)
	call pctcft(n)%set_pct_l2g(0._r8)
	pctsand(n,:) = 43._r8
	pctclay(n,:) = 18._r8
	organic(n,:) = 0._r8
	else
	pftdata_mask(n) = 1
	end if

In addition to the edge case where pctlak or pctgla is < 0.5%, this could also be relevant in a case like this: pctlak = 5.4%, pctgla = 94.4%. With the current code, we'd first get pctwet = 0.2%. But then, after the rounding, we'd end up with pctlak = 5%, pctgla = 94%, pctwet = 0%. Then later, in normalizencheck_landuse, I think we'd end up with 1% natural veg, which could trigger the 'vegetation found outside the pft mask' abort. (Caveat: I may be missing something here, since there are so many corrections done at different points. It's also possible that there are other problems with doing the rounding first, but I'm not seeing any off-hand.)

(This split off from #1001 )