Radiative sensitivities for cloud structural properties that are unresolved by conventional GCMs



The Independent Column Approximation (ICA), a stochastic subgrid-scale cloudy column generator, and a dataset produced by an array of two-dimensional cloud system-resolving models are used to estimate radiative sensitivities and uncertainties for parameters that describe the macrostructure of layered cloud at scales unresolved by conventional global climate models (GCMs). Three cloud structural parameters are considered: vertical overlap of fractional cloud, vertical overlap of cloud condensate, and horizontal variability of condensate. Radiative sensitivity is the derivative of top of atmosphere (TOA) fluxes with respect to a cloud structural parameter. Radiative uncertainty is defined as the product of radiative sensitivity and corresponding uncertainty in structural parameter. It is argued that radiative sensitivities and uncertainties can help direct and prioritize development of subgrid-scale parametrizations.

Using estimates of sensitivities and uncertainties for structural parameters, it is shown that uncertainties in TOA short-wave fluxes for cloud fraction overlap and horizontal variability are of similar magnitude averaged over the globe but display distinct spatial differences. Corresponding values for overlap of condensate are smaller and thus less important. For long-wave radiation, cloud horizontal variability is generally the most important of the three parameters considered here, but sensitivities, and thus uncertainties, are smaller than those for short-wave radiation. It is estimated that uncertainties associated with overlap of fractional cloud and horizontal variability are approximately as large as those associated with effective size of cloud particles. These results reinforce prior claims that GCM parametrizations for cloud overlap and horizontal variability should be addressed together, especially at solar wavelengths. These results are echoed in the appendix which examines TOA forcings that would arise in a GCM if the maximum-random overlap of homogeneous clouds model was replaced by a model that addresses the cloud parameters mentioned above. Copyright © 2005 Royal Meteorological Society.