THE EFFECT OF SCALE AND SNOW FRAGMENTATION ON THE ACCURACY OF FRACTIONAL SNOW COVER DATA OVER THE TIBETAN PLATEAU

Three MODIS-based fractional snow cover data are evaluated over the Tibetan Plateau from May, 2013 to April, 2015 with Landsat8/OLI data, including MOD10A1 in MODIS snow products collection 6[1, 2], along with MODSCAG[3] and MODAGE[4] fractional snow cover data which were retrieved based on linear spectral mixture analysis algorithm. The significant difference between MODSCAG and MODAGE is the endmember selection approach. This study compared these three products against the 'true value' derived from Landsat-8/OLI. Forests, grass and soil region were chosen to carry out the evaluation, as well as the Himalaya Mountain due to its acute topographic heterogeneity. We applied both binary and fractional metrics to evaluate all three of them with the spatial resolution increasing from 500m to 1km, 2km and 5km. We also quantitatively depict the patchiness of snow cover under diverse spatial resolutions in order to analyze the effect of snow patchiness to the accuracy of algorithms under different spatial scales.