Unfiltering of the EarthCARE Broadband Radiometer (BBR) observations: the BM-RAD product

Abstract

The methodology to determine the unfiltered solar and thermal radiances from the measured EarthCARE Broadband Radiometer (BBR) shortwave (SW) and total-wave (TW) filtered radiances is presented. Within the EarthCARE ground processing, the correction for the effect of the BBR spectral responses, the unfiltering, is performed by the so-called BM-RAD processor which produces the level-2 BM-RAD product. The BM-RAD product refers to unfiltered broadband radiances that are derived from the BBR and the Multi-Spectral Imager (MSI) instruments on board the forthcoming EarthCARE satellite. The method is based on theoretical regressions between filtered and unfiltered radiances, as is done for the Clouds and the Earth's Radiant Energy System (CERES) and the Geostationary Earth Radiation Budget (GERB) instruments. The regressions are derived from a large geophysical database of spectral radiance curves simulated using radiative transfer models. Based on the radiative transfer computations, the unfiltering error, i.e., the error introduced by the small spectral variations of the BBR instrument response, is expected to remain well below 0.5 % in the shortwave (SW) and 0.1 % in the longwave (LW), at 1 standard deviation. These excellent performances are permitted by the very simple optics used in the BBR instrument: a telescope with a single paraboloid mirror. End-to-end verification of the unfiltering algorithm has been performed by running the BM-RAD processor on modelled level-1 BBR radiances obtained for three EarthCARE orbits simulated by an integrated forecasting and data assimilation system. The resulting unfiltered radiances are eventually compared to the solar and thermal radiances derived by radiative transfer simulations over the three EarthCARE orbits. In addition, this end-to-end verification has provided further evidence on the high accuracy of the unfiltered radiance process, with accuracies better than 0.5 % for SW and better than 0.1 % for LW.