Abstract: It provides important information for soil digital management if soil total nitrogen can be rapidly and accurately estimated. In this research, NIR-Visible spectral reflectance of soil samples was measured using ASD2500 hyperspectral meter. The results indicated that Total Nitrogen (TN) content of black soil had better relationship with the first deviation of reflectance logarithm (LOGR-FD) than the original spectrum in 400-600 nm especially. The optimum model predicting TN of black soil using derivative spectra (LOGR-FD) was established by the method of stepwise regression. The selected bands in the model are 556, 1 642 and 2 491 nm respectively. Normal Difference Index (NDI550, 450) was constructed by spectral reflectance in the bands of 450 and 550 nm, and the model was also established for TN predicting using the NDI550, 450. By validity, the above two models produced better effects: the coefficients of determination (R2) are 0.863 and 0.829 and the Root Mean Square Errors (RMSE) are 0.122 and 0.152 respectively.