We currently define land surface albedo to be SW_u/SW_d. At night, SW_d is zero, so albedo is not be defined. This is not physical.

Note that SW_u is proportional to SW_d, so our temporary solution is SW_u/max(SW_d, eps(FT)). This would give zero albedo at night and prevent dividing by zero, but it is still not physical.

We should save the fraction of light absorbed by the land surface and fraction reflected, rather than absolute amount absorbed and reflected. Then the fraction of light reflect = albedo, and we avoid dividing by SW_d.

cc @braghiere