Snow mapping using multi-temporal Meteosat-8 data

Abstract

Meteosat-8 is the first geostationary satellite that possesses channels at all bandwidths that are of use for snow mapping. It therefore offers new possibilities for multi-temporal snow mapping as well as for snow mapping at short time intervals, which is for example required for numerical weather prediction models. The spectral capabilities of Meteosat-8 allow an optimal spectral separation of clouds and surface snow cover, but there always may be clouds with spectral characteristics similar to those of snow. In this study, we propose a classification technique that uses temporal information, in addition to spectral information, to detect these clouds. This technique is based on the temporal differences which often exist between subsequent values of a cloud-covered pixel. The inclusion of temporal information in the classification process improves the classification of images from a three-day test period in March 2004.