The spatial distribution of snow cover area is a crucial input to models of hydrology and climate in alpine and other seasonally snow covered areas. The objective in our study is to develop a rapidly automatic and high accuracy snow cover mapping algorithm applicable for the Tibetan Plateau which is the most sensitive about climatic change. Monitoring regional snow extent requires higher temporal frequency-moderate spatial resolution imagery. Our algorithm is based AVHRR and MODIS data and will provide long-term fraction snow cover area map. We present here a technique is based on the multiple endmembers approach and by taking advantages of current approaches, we developed a technique for automatic selection of local reference spectral endmembers