Read here. Clouds are critical in determining Earth's radiation balance. In order for climate models to accurately project future climate conditions, they absolutely must be able to simulate correctly the many facets of clouds interaction with radiation, both reflection and absorption. The newest peer-reviewed research finds that climate models fail miserably at cloud simulation, which may explain why climate model output ranges from bad to worthless.
"...authors Zhang et al. (2010) note that the different representations of clouds and their feedback processes in Global Climate Models (GCMs) have been identified as major sources of differences in model climate sensitivities, stating that "contemporary GCMs cannot resolve clouds and highly simplified parameterizations are used to represent the interactions between clouds and radiation.".....And what was particularly striking, in the words of Zhang et al., was "the model overestimate of the occurrence frequency of deep convection and the complete absence of cirrus anvils," plus the fact that "the modeled clouds are too reflective in all regimes.".....Since incoming and outgoing radiation are strongly affected by the 3D spatial pattern of clouds of various types, a model that gets the "right" current global temperature with the wrong pattern of clouds must have errors in its radiation and/or heat transfer parameterizations.....results of this study thus suggest that climate modelers' claims of physical realism in their models are not supported by detailed comparisons with the real world..."