Effects of cloud overlap on radiative feedbacks

The column radiative model (CRM) of the National Center for Atmospheric Res earch Community Climate Model has been used to test feedbacks associated wi th changes in monthly mean cloudiness for atmospheres with two different pr escriptions of how clouds overlap in the vertical. The first specification is the default CRM random overlap assumption. The second, called the nonran dom case, uses an estimate of the observed overlap based upon an analysis o f satellite- and surface-based observations. The results are presented prim arily in terms of the changes in top-of-the-atmosphere net cloud radiative forcing resulting from a 25% increase in total cloud water and separate 16. 5% increases in low, middle and high cloud layer amounts and differences th at occurred during the 1987 El Nino/Southern Oscillation (ENSO) event. Over all, the random model is about 20% more sensitive to 16.5% increases in low clouds than the nonrandom model, but the nonrandom model is about twice as sensitive to increases in middle cloud. Differences in sensitivity for cha nges in high cloud amount and total cloud water are relatively small. In th e areas near the large sea surface temperature anomalies the 1987 ENSO rela ted departures in the nonrandom model are 0.5-2 Wm(-2) greater than for the random model. Thus, this analysis strongly suggests that accurate specific ation of overlap in climate models is critical to the calculation of the ap propriate radiative feedbacks and sensitivities of models to external forci ng such as increased carbon dioxide or sulfate aerosols.