【Objective】Soil organic carbon content in China and its variation plays a very important role in the global carbon cycle and variation of atmospheric carbon dioxide concentration. Even minor changes in soil organic carbon content may affect the global carbon budget, atmospheric carbon dioxide concentration, and long-term sustainability of the ecosystem. Moreover, soil spatial information at the national scale is the basis for the study of changes in soil carbon storage and agricultural macro-decision-making. The processes of soil formation, development and erosion are subject to impacts of the complex and volatile environmental factors in the surroundings of the soil, so soil organic carbon is of strong spatial variability and dependence. Due to invisibility and concealment of soil properties, traditional field sampling methods require a lot of manpower and material resources, and can hardly obtain sufficient information to characterize continuous spatial distribution of soil organic carbon. Therefore, soil environment models are the main approaches to the research on digital soil mapping. However, most of these studies focus on small-scale and small-sized watersheds, while for large-scale areas the traditional area product method is still used in most cases. How to establish correlations between environmental variables and soil organic carbon on the national scale is the main research content of this paper. 【Method】Based on the data of soil organic carbon contents in the surface layers of 2 473 soil profiles collected during the Second National Soil Survey, this paper explored influences of factors, like topography, vegetation and climate on spatial distribution of soil organic carbon; With the ordinary Kriging method as control, geographically weighted regression, geographically weighted regression Kriging, multiple linear regression and regression Kriging were used separately to modeling for spatial prediction of soil organic carbon; indices, like mean absolute estimation error (MAE), mean relative error (MRE), root mean square error (RMSE) and Pearson correlation coefficient (r) were used to evaluate performance of these models; and soil organic carbon spatial distribution prediction maps were drawn separately. 【Result】Results show: (1) Soil organic carbon varied in the range from 1.62 g•kg ^-1 to 223.88 g•kg ^-1in content and averaged to be 22.28 g•kg ^-1in the country. Its variation coefficient reached 96.10%, which indicates that organic carbon in the soil varies very sharply in range, and is of strong spatial heterogeneity; (2)Soil organic carbon content was significantly related to annual mean precipitation, ≥10℃ annual accumulated temperature, elevation, slope, aspect, normalized difference vegetation index, annual average temperature, topographic wetness index, topographic position index and topographic roughness index. Among them, slope, elevation, aspect, topographic roughness index, annual average precipitation and normalized vegetation index were positively related, while topographic position index, topographic wetness index, annual average temperature and ≥10℃ accumulated temperature were negatively related; (3) Multiple linear regression coefficients might reflect influences of the environmental variables on soil organic carbon globally, whereas the geographically weighted regression coefficient map might do clearly those of different environmental variables on soil organic carbon in different geographical locations; (4) The mean absolute estimation error, root mean square error, mean relative error and Pearson correlation coefficient of the model were used as model validation indices and indicated that the geographically weighted regression is higher than the other models in prediction accuracy, and hence can be used to plot soil organic carbon spatial distribution characteristics maps of large scales areas; and (5) Areas relatively high in soil organic carbon content were mainly distributed in the northeast and southwest of the studied region, and patches in the southeast, while areas relatively low were in the northwest. 【Conclusion】The geographically weighted regression is higher than the ordinary Kriging, multiple linear regression, regression Kriging and geographically weighted regression Kriging in prediction accuracy. In this paper, efforts have been made to explore correlations between soil properties and environmental variables on large scales in an attempt to provide certain solutions and ideas for soil properties spatial mapping.