When we added rain and snow to our Circulating Cells program, we removed the slow descent of microscopic water droplets, saying that their movement would be insignificant compared with that caused by convection. This is indeed the case when an equilibrium with plenty of atmospheric convection is established.
Nevertheless, we have found in our recent tests, in which we are allowing the clouds to radiate heat directly into space, that clouds can form and sit directly upon the surface, where they block the Sun's light. The surface cools beneath these clouds, and the clouds themselves cool by radiating into space, and we have seen them sit there fore hundreds of hours. This is unrealistic, because in a hundred hours, a cloud will sink by at least a few hundred meters.
So we restored the sinking of cloud droplets to our simulation, at 3 mm/s, which is realistic for cloud droplets 20 μm in diameter.
When we restored the sinking, we noticed that our previous implementation had allowed the clouds to sink only when the cells containing them took part in a convection circulation. As a result, the clouds were sinking through our simulated atmosphere a hundred times slower than they should have been. The fast-sinking implemented by CC9 were in fact sinking at 3 mm/s instead of 300 mm/s, and the slow-sinking clouds were sinking at only 0.03 mm/s instead of 3 mm/s. Thus our fast-sinking clouds were a more realistic simulation of the manner in which actual clouds would sink, while our slow-sinking clouds were unrealistically slow. We run our sinking cloud experiments with a corrected version of CC9, and the new slow-sinking result looked like the former fast-sinking result.
When clouds sink at 3 mm/s, they can sit on the surface for a few hours, but after a hundred hours, they disappear. The droplets will sink 1000 m in a hundred hours, which is three times the height of a gas cells resting upon the surface. The slow descent of the droplets removes clouds that would otherwise freeze the surface, and therefore plays an important role in our simulation, despite the fact that convection, rain, and snow cause movements that are thousands of times faster.