Retrieving surface insolation from narrowband satellite radiances under cloudy skies– A comparison of two different approaches

The Spinning Enhanced Visible and Infra-Red Imager (SEVIRI) on-board the geostationary Meteosat 8 satellite continuously acquires multi-spectral images and enables the retrieval of cloud and radiation products at high temporal and spatial resolution (15 minutes, 3x3 km^2 at nadir) in particular over Europe and Africa.

The Clouds Physical Properties (CPP) and the Surface Insolation under Cloudy Skys (SICCS) retrievals using SEVIRI data are presented, which have been developed at the Royal Netherlands Meteorological Institute. They can derive cloud properties and surface insolation at full pixel resolution in near-real-time, and provide an accurate estimate of the surface insolation and its variability induced by clouds. Both retrievals use lookup tables of detailed 1D radiative transfer calculations and incorporate additional information on atmospheric water vapor and surface albedo. The surface insolation, and optionally other solar radiation components (e.g. the TOA albedo) are determined for these cloud products, and do not directly depend on satellite radiances. This way, consistency between calculated cloud properties and solar radiation components is guaranteed. However, the retrieval is vulnerable to violations of the underlying model assumptions, in particular the independent pixel approximation and 1D radiative transfer theory.

In contrast to our approach, most current retrievals of surface insolation are based on the successive steps of narrow- to broadband (N2BB) conversion, anisotropy correction, and the conversion of the top-of-atmosphere(TOA) albedo to an atmospheric transmission. The mathematical relations for these steps have to be given a priori. In general, they are either assumed to be the same for all atmospheric conditions, or chosen according to a coarse classification of atmospheric state. Hence, the choices for these relations do not correctly represent the influence of cloud properties on solar radiation under all conditions. On the other hand, the N2BB conversion and anisotropy correction are generally determined from measurements, and compensate for 3D effects not accounted by 1D radiative transfer theory.

To clarify the advantages and shortcomings of both approaches under realistic conditions, the accuracies of the SICCS retrieval and a conventional retrieval based on relations published in the literature are compared. This is done not only for the surface insolation, but also for the N2BB conversion and the broadband TOA albedo. As reference, a 1 year dataset of surface measurements from pyranometers throughout the Netherlands, as well as Meteosat's Earth Radiation Budget (GERB) instrument are used. Problematic aspects of both retrievals are highlighted, and suggestions for an improved retrieval strategy are given. As an additional benefit of this work, the suitability of retrieved cloud products for the modelling of a realistic solar radiation budget is established.