Traditional statistics, combined with geostatistics, were applied to analyze the spatial distribution of nitrogen(N) in plant and soil on a Stipa bungeana slope in the water-wind erosion crisscross region of the Loess Plateau. The spatial characteristics of the plant-soil system at different spatial scales were studied to reveal characteristics N distribution in the region grassland. Results showed that plant biomass, plant N content, and N stock increased gradually with the decrease of slope position. Soil organic carbon and nitrogen content in the upper and middle positions were similar, lower than those in the bottom position. Soil nitrate nitrogen and ammonium nitrogen contents performed a V-shaped change, with the highest content being on the upper slope. Except nitrate nitrogen and ammonium nitrogen contents, the variation coefficients of those variables measured to be the lowest on the bottom slope, which indicates that N distribution of plant-soil system was uniform on the bottom slope. Furthermore, the best fit models for each variable were precisely consistent at the three scales. Nugget effect increased, while spatial dependence between different sample points and spatial correlation weakened gradually with the increase of analysis scale. These results indicate that N in plant and soil have relatively stable spatial structures at different spatial scales and are characterized by medium or strong spatial correlation, and its spatial variation is mainly caused by structural factors.