Intercomparison of cloud properties retrieved from

Several mechanisms can cause deviations of satellite-derived cloud datasets from the MODIS and SEVIRI instruments. Two mechanisms are investigated here in order to determine their magnitude. First, the MODIS algorithm and the Cloud Physical Properties (CPP) algorithm are compared by using identical radiances obtained from the MODIS instrument as input. Both algorithms show a high correlation of estimated optical thicknesses, with values of 0.993 and 0.978 for water and ice clouds, respectively. However, ice cloud optical thicknesses from the CPP algorithm are biased low by about 20 % versus the MODIS results. The correlation of effective radii is lower, only reaching values of 0.894 and 0.96 for water and ice cloud pixels. Also, a positive bias of 0.9μm and 3μm for water and ice cloud particles is found for the CPP retrieval. A possible explanation is the different choice of absorbing wavelength made in both retrievals. While these results motivate further investigation of these differences, the overall agreement is still encouraging. The instrumental resolution is considered as a second mechanism leading to deviations in retrieved cloud properties. Using high resolution MODIS cloud products, about a quarter of cloudy pixels at SEVIRI resolution are found to contain broken clouds, and about 15-20 % to contain mixed-phase clouds. For these pixels, the retrieval assumption of a plane-parallel homogeneous cloud is clearly violated. However, even for completely cloud-filled pixels, the effects of the plane-parallel albedo bias cause an underestimate of cloud optical thickness of 5 % for water clouds, and an overestimate of cloud effective radius of 0.9μm. Errors in cloud properties increase strongly for broken-cloud pixels. These findings indicate that resolution effects cause deviations between the MODIS and SEVIRI cloud datasets of at least a similar magnitude as those originating from differences in the underlying algorithms