Snow cover estimation based on spectral unmixing

Abstract

Spectral Unmixing is the most recent method used to recover the Snow Cover Fraction of an area, but it depends particularly on the relevance of the set of endmembers. This communication investigates different strategies for defining set of endmembers for retrieving snow cover fraction with spectral unmixing. Endmembers can be estimated from on site measurements or estimated directly on the image. In this work we propose a set of endmembers associating semantics of field data for snow endmembers with the extraction of a set in a date without snow for other materials. A heterogeneous area in the Alps was considered in the experiment. Considering reference maps of snow available for several dates, Precision and Mean Absolute Error were computed for evaluating the estimated Snow Cover Fractions. Results obtained confirm the soundness of the proposed approach for low snow fraction.