Impact of 3D cloud structure on satellite retrievals of cloud droplet sizes

The presentation takes a close theoretical look at the radiative impact of cloud 3D structure on retrievals of droplet effective radii. It is distinguished between the errors that are due to sub-pixel inhomogeneity (unresolved variability) and due to the inhomogeneity of neighboring pixels (resolved variability). It is shown that ignoring unresolved variability produces a negative bias in the retrieved effective radius, while ignoring resolved variability, on the contrary, leads to overestimation of the domain average droplet size. The theoretical results are illustrated with examples from LES of Cu and Sc cloud fields. The analysis of cloud drop size distributions retrieved from both LES fields confirms that ignoring shadowing in 1D retrievals results in substantial overestimation of effective radii which is more pronounced for broken Cu than for Sc clouds. Collocated measurements of broken Cu clouds by MODIS and ASTER are used to check simulations and theory with observations.