Abstract: Based on LiDAR data of Liangshui National Nature Reserve, digital elevation model (DEM) was constructed and both primary terrain attributes (slope, aspect, profile curvature, etc.) and secondary terrain attributes (wetness index, sediment transport index, relative stream power index, etc.) were extracted. According to the theory of soil formation, geographically weighted regression (GWR) was applied to predict soil total nitrogen (TN) of the area, and the predicted results were compared with those of three traditional interpolation methods including inverse distance weighting (IDW), ordinary Kriging (OK) and universal Kriging (UK). Results showed that the prediction accuracy of GWR (77.4%) was higher than that of other three interpolation methods and the accuracy of IDW (69.4%) was higher than that of OK (63.5%) and UK (60.6%). The average of TN predicted by GWR reached 4.82 g·kg^-1 in the study area and TN tended to be higher in the region with higher elevation, bigger wetness index and stronger relative stream power index than in other areas. Further, TN also varied partly with various aspects and slopes. Thus, local model using terrain attributes as independent variables was effective in predicting soil attribute distribution.