Soil organic carbon (SOC) has great impacts on global warming, land degradation and food security. Classic statistical and geostatistical methods were used to characterize and compare the spatial heterogeneity of SOC and related factors, such as topography, soil type and land use, in the Liudaogou watershed on the Loess Plateau of North China. SOC concentrations followed a log-normal distribution with an arithmetic and geometric means of 23.4 and 21.3 g kg-1, respectively, were moderately variable (CV = 75.9%), and demonstrated a moderate spatial dependence according to the nugget ratio (34.7%). The experimental variogram of SOC was best-fitted by a spherical model, after the spatial outliers had been detected and subsequently eliminated. Lower SOC concentrations were associated with higher elevations. Warp soils and farmland had the highest SOC concentrations, while aeolian sand soil and shrublands had the lowest SOC values. The geostatistical characteristics of SOC for the different soil and land use types were different. These patterns were closely related to the spatial structure of topography, and soil and land use types. Soil organic carbon (SOC) has great impacts on global warming, land degradation and food security. Classic statistical and geostatistical methods were used to characterize and compare the spatial heterogeneity of SOC and related factors, such as topography, soil type and land use, in the Liudaogou watershed on the Loess Plateau of North China. SOC concentrations followed a log-normal distribution with an arithmetic and geometric means of 23.4 and 21.3 g kg-1, respectively, were moderately variable (CV = 75.9%), and demonstrated a moderate Spatial dependence according to the nugget ratio (34.7%). The experimental variogram of SOC was best-fitted by a spherical model, after the spatial outliers had been detected and subsequently eliminated. Lower SOC concentrations were associated with higher elevations. Warp soils and farmland had the highest SOC concentrations, while aeolian sand soil and shrublands had the lowest SOC values. The geostatistical characteristics of SOC for the different soil and land use types were different. These patterns were closely related to the spatial structure of topography, and soil and land use types.