The snow cover is an important parameter for water run-off prediction in high mountain areas. At present its seasonal changes can be monitored regularly by satellite. The paper presents a new method for a more accurate classification of snow from multispectral Landsat data with the aid of a digital terrain model (DTM). A reflectance map derived from the DTM was used to register the Landsat data onto the map applying the cross-correlation function. The DTM makes the independent evaluation of snow possible with respect to elevation, exposure and slope angle. For areas with climatologically similar conditions, it becomes possible to extrapolate the snow cover of the total catchment area supposed to be partially obscured by clouds from a small “cloud free” zone.