In this paper, a monitoring model of snow depth was built based on the 4 scenes of NOAA satellite digital images under sunshiny condition and the corresponding ground observation data from 20 meteorological stations during 2 snow disasters from 1996 to 1997 in North Xinjiang. The pixel-based snow coverage rate and snow spatial classification were studied by using linear mixture spectrum disassembling method, and two grid data layers based quantified indices used for estimating snow hazard grade of grassland and animal husbandry were put forward. The results indicated that by using the snow monitoring model and linear mixture spectrum disassembling method, the image cell based snow depth and snow coverage rate could be calculated, and the precision of snow classification could be improved. The image cell based snow hazard index could systematically express the spatial distribution of snow, grass, animal and climate conditions, and reflect the snow hazard grade of grassland and animal husbandry.