[expatpat]

Discrete models can in fact work well for non-linear continuous systems. That's not the real issue. The real issue is that
(a) they use climate sensitivities to CO2 which are 200-300% larger than values independently calculated from the physics,
(b) they ignore the variations in the sun's effect,
(c) they can't model clouds and cloud formation accurately, which is a big problem since clouds block most of the incoming energy from the sun during the day, and trap heat at night.

[dufekin]

BINGO. We have a winner. The key problem with all climate model output that I’ve seen is this:
(c) they can’t model clouds and cloud formation accurately, which is a big problem since clouds block most of the incoming energy from the sun during the day, and trap heat at night.

When a model can’t simulate cumulus convection accurately, all sorts of bizarre things happen. Severe droughts? Well, yeah, imagine summer in the South without thunderstorms. Tremendous floods? Well, yeah, in the rare event that these models do show thunderstorms, they’re totally disproportionate to real thunderstorms. Extreme hurricanes? Well, yeah, if ordinary thunderstorms don’t form to convect the solar energy from the surface to the upper troposphere, extreme hurricanes would do it explosively.