Operational snow mapping using multitemporal Meteosat SEVIRI imagery

The Spinning Enhanced Visible and Infrared Imager (SEVIRI) onboard Meteosat Second Generation is the first geostationary satellite instrument with all visual and infrared channels that are important for snow mapping. In this paper, we present an algorithm for deriving snow cover maps from SEVIRI data that makes use of the unique combination of adequate spectral resolution and very high frequency. The short interval of 15 min between images makes it possible to extend traditional spectral classification with a detection of changes between images. This improves the detection of clouds and cloud shadows in instantaneous images, because these often display more variation in time than the surface. It therefore allows a more accurate mapping of surface snow cover, as is shown by a validation of the results with ground observations and other satellite data. The accurate classification of each single image allows the generation of temporal composite snow maps in near real-time, which is for example of interest for numerical weather prediction models. When compared to many in situ measurements from the winter of 2005/2006, the accuracy of the algorithm is 95%.