Idly looking for something to write about this morning I sidled over to Geophysical Research Letters where my eye alighted on the abstract of a paper by Cheruy et al. It concerns the CMIP5 climate models and considers clouds and the transfer of moisture from land to atmosphere. Here's the abstract:
Over land, most state-of-the-art climate models contributing to Coupled Model Intercomparison Project Phase 5 (CMIP5) share a strong summertime warm bias in midlatitude areas, especially in regions where the coupling between soil moisture and atmosphere is effective. The most biased models overestimate solar incoming radiation, because of cloud deficit and have difficulty to sustain evaporation. These deficiencies are also involved in the spread of the summer temperature projections among models in the midlatitude; the models which simulate a higher-than-average warming overestimate the present climate net shortwave radiation which increases more-than-average in the future, in link with a decrease of cloudiness. They also show a higher-than-average reduction of evaporative fraction in areas with soil moisture-limited evaporation regimes. Over these areas, the most biased models in the present climate simulate a larger warming in response to climate change which is likely to be overestimated.
If I'm understanding this correctly, when you consider climate models' simulation of clouds and land-atmosphere coupling, the warmer models show the most marked biases against observations.
But don't worry, they are just fine and dandy for informing policymakers.