Nitrogen (N) is one of the most important nutrients for plant growth, yet it has high spatial variability because of the effects of topography, climate, and vegetation. Therefore, it is critical to demonstrate and model the distribution of N to enhance our understanding of N variability and related factors. We used Sejila Mountain (elevation of approximately 3950-4350 m) in southeastern Tibet as a model area to examine the spatial pattern of N distribution. We applied a 30 × 50-m grid sampling method and the geostatistical semivariogram analysis to study the spatial variability and distribution of the soil total N (TN), nitrate-N (NN), and ammonium-N (AN) in both valleys and slopes of the Sejila Mountain. The TN, NN, and AN contents in the 0-10cm layer were higher than those in the 10-20cm layer: (3.40±1.19) g/kg and (2.32±0.50) g/kg, respectively, for TN (P < 0.05); (360.55±97.72) mg/kg and (273.15±64.97) mg/kg, respectively, for AN (P < 0.01); and (98.45±22.00) mg/kg and (83.72±33.52) mg/kg, respectively, for NN (not significantly different). AN comprises a greater fraction of the mineral N than NN, and in the 0-10cm layer, the proportions of AN and NN were (78.16±3.97)% and (21.84±3.97)%, respectively. The spatial variability of TN and AN in the 0-10cm layer was higher than that in the 10-20cm layer, but the opposite was found for NN. The coefficients of variation in spatial distribution for TN, AN, and NN in 0-10cm and 10-20cm layers were 34.95% and 21.49% for TN, 27.10% and 23.78% for AN, and 22.35% and 40.04% for NN, respectively. The N content in 0-10cm and 10-20cm layers increased with increasing elevation, but the increase was not significant (P > 0.05). The TN content showed a higher dependency on altitude in the 10-20cm layer than in the 0-10cm layer, whereas the opposite effect was found for NN and AN. The soil N contents in the valleys were higher than those on the slopes, which may have been related to high levels of accumulation and decomposition of vegetation residues in the gully areas. These results imply that the effects of microtopography should be considered when assessing the spatial heterogeneity of N. The distribution of soil TN, AN, and NN showed a moderate spatial correlation. The spatial variability of soil TN followed an exponential function model and the nugget:sill ratio was 50%. Gaussian models were the optimal models for AN and NN, and the nugget:sill ratios were 70.91% and 37.45% for AN and NN, respectively. The spatial autocorrelation of the soil TN, NN, and AN in the study area decreased from NN to TN and AN. The spatial variability of soil NN was affected more by spatial structural factors, whereas soil AN was affected by random factors.