Comparison of CFMIP GCM cloud statistics to long-term ARM data

Description

Cloud occurrence and radiative forcing statistics derived from long-term, continuous ground observations from the ARM SGP and TWP Darwin sites are compared to similar statistics derived from general circulation models submitted to CFMIP2 in order to show accuracies and inaccuracies in the representation of clouds. Quantities considered include the annual cycles of cloud cover, volumetric cloud fractional occurrence, solar and IR cloud forcing at the surface and TOA, and solar and IR atmospheric cloud forcing. During a period of twelve years of monthly averages, the models rarely produced accurate descriptions of cloud cover and volumetric cloud fraction. The models often underestimated cloud cover, especially for low- and mid-level clouds, sometimes only producing less than 40% of the cloud cover that is observed at these levels. High clouds were somewhat more accurately modeled, and in some cases, the models produced too much cirrus. However, despite the fact that cloudiness was not well-modeled, the models produced cloud forcing values throughout the year that compared very well to the column calculations and observations. Therefore, it is concluded that the models inaccurately represent the cloud processes and physics in their cloud parameterizations.