ekluzek
commented
6 years ago @swensosc there's a potential divide by zero (and I've hit it in testing) in SoilTemperatureMod in the following code snippet...
! dz_h2osfc(c) = max(1.0e-6_r8,1.0e-3*h2osfc(c))
! c_h2osfc(c) = cpliq*denh2o*dz_h2osfc(c) !"areametric" heat capacity [J/K/m^2]
!scs
if (h2osfc(c) > 0._r8) then
c_h2osfc(c) = max(1.0e-6_r8,cpliq*h2osfc(c)/frac_h2osfc(c))
dz_h2osfc(c) = max(1.0e-6_r8,1.0e-3*h2osfc(c)/frac_h2osfc(c))
else
c_h2osfc(c) = 1.0e-6_r8
dz_h2osfc(c) = 1.0e-6_r8
endif
!scs
You check for h2osfc being small, but not the actual number that's the divisor -- frac_h2osfc. There's a few different ways this could be handled, check for them both at the same time, or have a more complex if statement.
It's also recommended that you don't check for greater than exactly zero, but some small value. Doing a check like this would be numerically safe...
`if (h2osfc(c) > epsilon(h2osfc(c)) ) then`
that would translate to about 1.e-15. But, from what you have, it seems like a threshold much higher than that would be appropriate here.
According to coupler diagnostics, there is an energy imbalance in the land in which the land appears to be gaining heat at 0.07 W/m2 (0.25 W/m2 over land only) even when the land is spunup. Specifically, the energy imbalance is 0.0747 W/m2 over 139 years of the b.e20.B1850.f09_g17.pi_control.all.265 simulation. This imbalance was confirmed to within 0.001 W/m2 by processing the CLM history file output from b.e20.B1850.f09_g17.pi_control.all.265 for the balance terms used by the coupler. Offline CLM5 simulations indicate a similar imbalance. CLM4 and CLM4.5 simulations indicate that the imbalance was introduced in CLM4.5, simulations with CLM4 balance to within 0.01 W/m2. Further investigation indicates that the imbalance is due to interactions of snow with surface water and possibly with lakes. A bug fix for snow-surface water interactions has been proposed by S. Swenson and tested in a offline CLM simulation (/glade/p/work/oleson/clm4_5_17_r265/cime/scripts/clm50_r265_1deg_GSWP3V1_iso_400i_ccrit_nosubgrid_bugfix_1850pAD). This bug fix results in an imbalance of -0.01 W/m2 for a twenty year climatology. A related bug fix is currently being tested which indicates that the imbalance increases again, and this remaining appears to be related to snow-lake interactions only. Investigation in progress.