Abstract: With the rapid development of hyperspectral remote sensing, it is feasible to quantitatively investigate the important chemical components in soil. In this study, the values of NIR-Visible spectral reflectance of soil samples collected from the Junggar Basin in Xinjiang, China were measured with an ASD Pro FieldSpec3 hyperspectral meter so as to predict the total soil salinity based on the spectral reflectance of soil, and provide a theoretical basis for monitoring soil salinization. The Pearson correlation coefficient analysis was used to estimate the correlation between total salinity and spectral reflectance of soil. The results showed that there was a good correlation between soil spectral reflectance and total soil salinity. The highest correlation coefficient between second derivative of soil spectral reflectance and total soil salinity was 0.806, and the minimum rootmeansquare error was 1.508. Through the multiple statistics regression analysis based on spectral reflectance, the performance of inversion model of total soil salinity at wavelengths of 1 130 nm, 1 430 nm and 1 930 nm was good, these three wavelengths could be used to develop a regression equation for inversing the values of total soil salinity.

Key words: artificial forest, hypersprctrum, soil salinization, Junggar Basin, Xinjiang