Considering the Effects of Horizontal Heterogeneities in Satellite-Based Large-Scale Statistics of Cloud Optical Properties

This paper explores a new approach to improving satellite measurements of cloud optical thickness and droplet size by considering the radiative impacts of horizontal heterogeneity in boundary-layer cumulus clouds. In contrast to the usual bottom-up approach that retrieves cloud properties for individual pixels and subsequently compiles large-scale statistics, the proposed top-down approach first determines the effect of 3D heterogeneity on large-scale cloud statistics and then distributes the overall effects to individual pixels. The potential of this approach is explored by applying a regression-based scheme to a simulated dataset containing over 3000 scenes generated through large eddy simulations. The results show that the new approach can greatly reduce the errors in widely used bispectral retrievals that assume horizontal homogeneity. Errors in large-scale mean values and cloud variability are typically reduced by factors of two to four for 1 km resolution retrievals-and the reductions remain significant even for a 4 km resolution. The calculations also reveal that over vegetation heterogeneity-caused droplet size retrieval biases are often opposite to the biases found over oceans. Ultimately, the proposed approach shows potential for improving the accuracy of both old and new satellite datasets.