Several methods of adapting the satellite and model datasets to allow a valid comparison were applied. In particular, a method to filter out optically very thin clouds from the model simulation dataset was developed. This method is based on complementary radiative transfer calculations to estimate the appropriate cloud detection limit of the satellite-retrieval method. The filtering approach was found necessary since the model simulation has no lower limit for the simulated cloud condensate amounts and the associated fractional cloud cover. The satellite observation will not capture clouds with very low optical thicknesses due to very low cloud water and cloud ice amounts and a direct comparison of the two datasets is thus not appropriate. Other adaptations concerned the simulation of satellite viewing effects affecting the interpretation of contributions from Low-, Mid- and High-level clouds and the interpretation of optical thickness categories.
First results indicate both similarities and differences between the model-simulated and the satellite-observed cloud fields, both in terms of the occurrence and in the vertical distribution of clouds. The model appears to produce quite reasonable amounts of total cloud cover (i.e., within a few percent compared to the satellite climatology) on seasonal and annual time scales. However, contributions from the various vertical cloud groups differ substantially from the observations so that medium-level clouds seem to be underestimated while high- and low-level clouds are slightly overestimated. Furthermore, an over-representation of cloud categories with high optical thicknesses is seen for all vertical cloud groups with the largest deviations for low-level clouds in the summer season. Implications for model-simulated radiation budget parameters and other parts of the physical parameterisation (e.g. precipitation processes) will be discussed as well as results from sensitivity studies for testing the robustness of the results.