When modelling a ground plane as a defined piece of geometry, with a surface reflectance of 20% and a ground reflectance value of -g 0.2 set in gensky I get a discrepancy between the luminance from my phyically modelled ground plane and the groundglow provided by gensky for two of the available sky types generated by gensky:

• Intermediate sky without sun (-i)
• Sunny sky without sun (-s)

For the other sky types, the luminance from my physically modelled ground is identical to that from the groundglow, such that the boundary between the physical ground plane and the groundglow is seamless. For the two sky types mentioned above there is a clear discrepancy, with the ground glow brighter than the modelled ground. I have checked the gensky manual and it includes the following:

Average ground reflectance is rfl. This value is used to compute skyfunc when Dz is negative. Ground plane brightness is the same for −s as for +s. (Likewise for −i and +i, but see the −r option below.)

Further, searching the mailing list yields the following comment from Greg Ward when a user encountered the same thing: http://www.radiance-online.org/piperm...

The glow computed for the ground includes whatever sunlight should be
falling on it. An intermediate sky has some sun, so the solar
irradiance (-R) setting has an effect. (Even if you use -i or -s, gensky still computes the ground radiance based on the sun being out
-- it just leaves off the solar source.)

-Greg

So I understand it is doing what it's supposed to do, my question is why? For what practical purposes is it desirable for the ground glow for these two sky types to be calculated differently from the other sky types? It results in a visible boundary between the modelled ground plane and the groundglow in rendered images.