Abstract
The Normalized Difference Snow Index (NDSI) has aroused more and more interests as a snow detecting method, and
its feasibility has been proven in many studies. However, the accuracy of NDSI alone has failed to meet the expected
requirement. The detection of snow cover is not only affected by the atmosphere, but also influenced by the regional terrain
and the underlying surface covered by snow. Due to this reason, in the present work, the snow detection in Akita Prefecture
located in the Tohoku region of Japan was conducted through the combination of NDSI and NDVI (Normalized Difference
Vegetation Index). Because of the large area of mountains and forests in the rugged Akita Prefecture region, the surface
reflectance was retrieved from the top of vegetation after atmospheric and topographic corrections. Furthermore, in order to
reduce the effect of the misclassification of snow and vegetation cover, a Normalized Difference Vegetation Index (NDVI)
model was used to discriminate the snow and forest pixels. Compared to the MOD10_L2 data and NDSI alone results, the
combination of NDSI and NDVI showed high accuracy in snow cover detection.
