investigate snow properties such as grain size or contaminants. Meanwhile, the research on the spatial distribution of avalanches has remained marginal. A method for automatic classification of avalanches in satellite or aerial imagery would greatly enlarge the possibilities of avalanche modelling and forecast evaluation.
Snow avalanches originate from mountain slopes further uphill where a different and usually colder temperature regime prevails. During the descent of an avalanche, different layers are mixed, the snow agglomerates and occasionally soil and vegetation material is accumulated. Avalanche deposits differ from undisturbed snow by their densities, grain sizes and contaminations by impurities. In this study, more than 400 spectra of nine avalanche deposits and adjacent undisturbed snow cover have been measured in the Davos region in the Swiss Alps in January and April 2008 using an ASD Field Spectroradiometer. Normalisation of the spectra was required due to high variance in reflectance values caused by shadow effects and different avalanche parameters. Continuum removal of absorption features highlighted significant relative differences in reflectance and allowed for local separation between snow and avalanche spectra. Three groups of dirty, wet and recent deposits could be distinguished. The comparability between the deposits is low due to high variance of avalanche spectra; nevertheless an overall classification accuracy of 86% could be reached using area and depth of specific absorption features.