Comparison of Snow Indexes in Estimating Snow Cover Fraction in a Mountainous Area in Northwestern China

The normalized difference snow index (NDSI), a key part of the snow-mapping algorithm for extracting snow information from remotely sensed imageries, has been frequently employed in acquiring snow cover extent in the past decades. A linear regression analysis has been frequently used in estimating snow cover fraction (SCF) on a subpixel basis by developing a statistical relationship between NDSI and SCF. In this letter, a comparison of NDSI, ratio snow index (RSI), and difference snow index (DSI) in estimating SCF is carried out in mountainous area of northwestern China. Three kinds of fitting methods, i.e., linear fitting, logarithmic curve fitting, and exponential curve fitting, are employed in this comparison for the purpose of finding out a best statistical fitting relationship between SCF and snow index. The results show that RSI and DSI are both good substitutes for NDSI in SCF estimation due to high R-squares, up to 0.83, of fitted lines between snow indexes and SCF, respectively. Furthermore, exponential fitting is considered to be of the highest robustness in the three fitting methods of interest for all the snow indexes studied.