Taking the core demonstration area of Youyi Farm, a typical black soil area in Northeast China, as the study area. Such as soil properties, topography, and remote sensing index were chosen as the environmental variables. Four representative digital soil mapping models, ordinary kriging (OK), geographically weighted regression (GWR), random forest (RF) and random forest-ordinary kriging (RF-OK), were selected to predict the contents and spatial distributions of surface soil pH, organic matter (SOM) and total nitrogen (TN) contents in the demonstration area. And uncertainty maps of spatial distribution were drawn by selecting the optimal model based on model accuracy. The results showed that the average value of pH, SOM, TN in the study area were 6.63, 42.26 g/kg and 1.94 g/kg. The coefficients of variation were 13.67%, 29.50% and 31.98%, respectively, all of which belonged to moderate spatial variation. In terms of the prediction accuracies of the four models, RF-OK model showed the best performance for predicting soil pH (R²=0.83, CCC=0.84, RMSE=0.41) and SOM (R²=0.72, CCC=0.68, RMSE=7.36 g/kg), and RF model achieved the best performance in predicting soil TN (R²=0.59, CCC=0.68, RMSE=0.36 g/kg). The spatial distribution of the three soil attributes in the demonstration area showed strong spatial heterogeneity. The overall trends of the spatial distribution of soil pH, SOM and TN predicted by the four models were basically the same, and all of them showed a spatial pattern of high in the northeast and low in the southwest. This study not only provides data support for precision agriculture management in the demonstration area, but also provides valuable reference for selecting prediction methods of digital soil mapping.