Abstract: Soil sand content is an import indicator that can reflect soil quality, and it is very meaningful to build the method that uses hyperspectral remote sensing to retrieve soil sand content at regional scale. In this paper, taking the north Xinjiang as the research region, the best band with the fine correlation coefficient and the best simulation method was found based on combining field hyperspectral data and MODIS images, and transferring the original soil reflectance spectrum into several forms. Some good results were gotten by retrieving the sand content of bare soil in north Xinjiang. The results show that 430 nm is the best band to retrieve soil sand content; the soil spectrum data transferred by logarithm’s reciprocal could reach the maximum value at 430 nm, and the correlation coefficient was 0.76. MODIS band 3 could be used to retrieve soil sand content at regional scale; the onedimension threeorder regression model using the original spectrum data at 460 nm as independent variable could get the highest accuracy when retrieving soil sand content, and the accuracy could reach 89.30%. According to the retrieving result from bare soil by using the model built above, the soil sand content of bare soil was relative higher, and it could match the actual conditions well.

Key words: hyperspectral remote sensing, soil sand content, MODIS, north Xinjiang