This paper made a coherence analysis to access the ability of snow extraction based on the Terra-SAR X-HH data, and develop a method using multi-temporal coherence combined with the traditional method. Two pairs of Terra-SAR data with the same temporal baseline and similar spatial baseline were chosen, before and after snowfall. Traditional method was used to remove the snow free area with relative higher coherence. Differences between two coherence data was taken to distinguish snow covered from the snow free area which shows low coherence value under X-band condition. Backscatter coefficient data was used as well. The result was verified using GF-1 optical image. An accuracy of 82.44% was achieved if we consider the result from GF-1 optical image as `ground true'. Google Earth image was also used as verification which shows a good estimation of the snow covered areas.