An automatic self-learning cloud-filtering algorithm for Meteosat Second Generation–Spinning Enhanced Visible and Infrared Imager

Abstract

Cloud detection is an important pre-processing step to derive operational products from meteorological satellites. This work presents a new cloud-detection algorithm with Meteosat Second Generation (MSG) images, operative at global scale. The algorithm takes advantage of the spectral and temporal resolution of the Spinning Enhanced Visible and Infrared Imager (SEVIRI) sensor. The algorithm is fully automatic in all its stages, including the thresholds definition by means of a self-learning methodology. These properties remove the need for ancillary data and restrictions in the area of application. This algorithm has been used in order to generate cloud masks during 2009. These cloud masks have been compared to the masks obtained with the National Aeronautics and Space Administration algorithm MOD35 with Terra-Moderate Resolution Imaging Spectroradiometer (MODIS) images and the European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT) algorithm for MSG–SEVIRI in Spain territory. The result shows an 88% agreement with EUMETSAT and a better than 83% agreement with the MOD35 algorithm.