state_vector is NaN when land use is off

[glemieux]

API testing with e3sm_landuse_develop suite for E3SM-Project/E3SM#6353 exposed an issue in which state_vector is being referenced without being defined when use_fates_luh is .false.:

fates/biogeochem/EDPatchDynamicsMod.F90

Lines 472 to 474 in fd7f343

	! if the area of secondary land has just exceeded the minimum below which we ignore things,
	! set the flag to keep track of that.
	if ( (state_vector(secondaryland) .gt. site_in%min_allowed_landuse_fraction) .and. &

The state_vector is set upstream in the same subroutine only when use_fates_luh is true:

fates/biogeochem/EDPatchDynamicsMod.F90

Lines 324 to 328 in fd7f343

	if (hlm_use_luh .eq. itrue) then
	call GetLUHStatedata(bc_in, state_vector)
	site_secondaryland_first_exceeding_min = (state_vector(secondaryland) .gt. site_in%min_allowed_landuse_fraction) &
	.and. (.not. site_in%landuse_vector_gt_min(secondaryland))
	end if

@ckoven for non land use mode run cases, I figured we could simply make sure that state_vector = current_fates_landuse_state_vector based on the very first call at the start of the disturbance_rates subroutine:

fates/biogeochem/EDPatchDynamicsMod.F90

Lines 225 to 226 in fd7f343

	! first calculate the fraction of the site that is primary land
	current_fates_landuse_state_vector = site_in%get_current_landuse_statevector()

This results in the test passing, but I wanted to double check my logic first.

There is one other case that follows this pattern in LoggingMortality_frac:

fates/biogeochem/EDLoggingMortalityMod.F90

Lines 263 to 267 in fd7f343

	if (hlm_use_luh .eq. itrue) then
	call GetLUHStatedata(bc_in, state_vector)
	site_secondaryland_first_exceeding_min = (state_vector(secondaryland) .gt. currentSite%min_allowed_landuse_fraction) &
	.and. (.not. currentSite%landuse_vector_gt_min(secondaryland))
	end if

where state_vector is only used in that subroutine behind a check on site_secondaryland_first_exceeding_min which should always be false for non land use mode per the current logic seen above.

[glemieux]

I should note that the failing test cases on perlmutter are ERS_D_Ld15.f45_g37.IELMFATES.pm-cpu_intel.elm-fates_cold_treedamage and SMS_D_Ld20.f45_f45.IELMFATES.pm-cpu_intel.elm-fates_rd.

The same issue presented itself on izumi via SMS_D_Mmpi-serial_Ld5.5x5_amazon.I2000Clm60FatesRs.izumi_nag.clm-FatesCold.