Quantification of Uncertainties in Retrieved Cloud Properties from Satellite Imagers

The retrieval of cloud optical thickness and effective radius from meteorological satellite imagers requires inversion of the relation between cloud properties and top-of-atmosphere reflectances. This relation is highly non-linear, and depends on a large number of additional parameters, such as the sun and viewing geometry, and the surface albedo. Here, the application of principal component analysis is investigated to reduce the high dimensionality of the retrieval problem. It is found that 99.96 % of the variability present in the pre-calculated relation of 0.6μm radiance and cloud optical thickness is captured by the first 4 principal components. It is also demonstrated that changes in the error sensitivity of the retrieval can be related to changes in the principal component amplitudes. Results show that the retrieval of cloud optical thickness is not only highly sensitive to random and systematic errors, but also undergoes significant situation-dependent changes. Implications of these finding for the retrieval of diurnal cycle information are discussed. It is shown that a systematic error of 1 % in reflectance can introduce an artificial diurnal cycle with an amplitude of 5-10 % in cloud optical thickness over the course of the day