Abstract: [Objective] To solve the traditional soil total nitrogen determination methods such as complexity,time-consuming,energy consumption and other issues.[Method] Gray desert soil in Xinjiang was taken as our research object to study the spectral reflection characteristics of different soil total nitrogen and by using classical statistical and spectroscopy method,the remote sensing inversion models of soil total nitrogen were established and validated in the study area.[Result] The results indicated that the content of total nitrogen in soil was similar to those of different spectral reflectance.The spectral reflectance of the soil appeared near 780,1,800 and 2,140 nm wavelength.There was obvious absorption valley in the vicinity of 1,910 nm and the correlation between the total nitrogen content in soil and the original spectral reflectance was poor.The determination coefficients of first-order differential processing to original spectral reflectance were higher than original spectral reflectance and second order differential,the maximum R2 was 0.819,reaching a very significant correlation;By using the first order differential transform to extract the characteristic bands 667 and 1,414 nm,the estimation model of soil total nitrogen content was established,they were:Y =2,698.048 X667-1,062.149 X1414-0.015.The R2 was 0.75;The estimation model validation found that the R2 was 0.80.When the total nitrogen content was too large or too small,the model estimation error was relatively large,and the overall prediction accuracy was quite high.The results has provided a theoretical basis to improve hyperspectral remote sensing monitoring accuracy of soil nitrogen.[Conclusion] High spectral analysis technique has certain significance for prediction of soil total nitrogen content and rapid identification of soil total nitrogen content by using the estimation model can be achieved.