Model-driven retrieval of fractional snow cover area

Operational snow cover mapping by optical sensors has taken place for more than two decades, but there is still a
demand for improved mapping accuracy. Most operational
products are binary (snow/no-snow). In the work presented here, a new approach has been taken to achieve significant
improvements in the accuracy. Current spectral BRDF
characteristics of the snow and snow-free ground are modeled
locally, per pixel. Spatial functions for these characteristics are established for the region to monitor. The approach opens for snow mapping combining several different sensors independently. It is a solution to the needs for long-term climate monitoring where inter-sensor calibration and introduction of new generations of sensors make it difficult achieving time consistency in the mapping. The method is currently under validation in mountain regions in Norway.