Threshold for explicit snow pack initiation should use frac_sno_eff, not frac_sno

[billsacks]

Brief summary of bug

The threshold for snow pack initiation is currently based on frac_sno(c)*snow_depth(c). It seems this should actually be frac_sno_eff(c)*snow_depth(c)

General bug information

CTSM version you are using: ctsm1.0.dev055 (but also present in earlier versions)

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

I believe this just affects (a) runs with use_subgrid_fluxes = .false. (NOT the default); (b) urban points in all runs. (For runs with use_subgrid_fluxes = .true., columns other than urban and lake have frac_sno == frac_sno_eff; the code in question doesn't operate over lakes, so I think this just affects urban points.) The impact should be fairly small in both cases, since this just changes the timing of when an explicit snow pack is initiated, rather than using an unresolved (0-layer) snow pack.

However, this bears further investigation.

Configurations affected: I think this just affects (a) runs with use_subgrid_fluxes = .false. (NOT the default); (b) urban points in all runs.

Details of bug

This is the code in question:

https://github.com/ESCOMP/ctsm/blob/a87e67da23ab4c7e971351289dc051e2ebc5bf1c/src/biogeophys/SnowHydrologyMod.F90#L746-L751

Based on other, similar code – particularly in CombineSnowLayers – I think frac_sno should actually be frac_sno_eff here. @swensosc agrees.

[billsacks]

I have confirmed that, as expected, in a case with use_subgrid_fluxes = .true. (the default), this fix only changes answers for urban landunits.

I confirmed this by running SMS_D_Ld1_P4x1.f10_f10_musgs.I2000Clm50BgcCropQianRsGs.bishorn_gnu.clm-default before and after the fix, with vector output for the h0 file. I examined the landunits for which FSH differs in this test, and found that it was just landunits 7, 8 and 9.

[billsacks]

And also as expected: in a case with use_subgrid_fluxes = .false., differences are more wide-spread - i.e., across more landunits. (Confirmed via SMS_D_Ld1_P4x1.f10_f10_musgs.I2000Clm45BgcCropQianRsGs.bishorn_gnu.clm-no_subgrid_fluxes with vector output: there are diffs for landunits 1, 2, 4, 8 and 9.)