In order to find out the spatial distribution characteristics of soil saturated hydraulic conductivity (Ks), 136 undisturbed soils were collected in Ningdong. The spatial structure characteristics of Ks were analyzed by classical and geostatistical methods. Taking terrain factors and soil properties as auxiliary variables, the regional soil Ks were predicted by the random forest method (RF), ordinary Kriging method (OK) and stepwise regression Kriging method (RK), and the accuracy of the prediction results of the three methods were evaluated. The results showed that Ks ranged from 0.05 to 7.13 mm/min, with an average value of 1.46 mm/min and a coefficient of variation of 106.86%. Ks had autocorrelation and cross-correlation with bulk density, porosity, elevation, slope, aspect, plane curvature and section curvature under different lag distance. The nugget value of soil Ks was 38, indicating that the variability of soil Ks caused by random factor was large, and the spatial heterogeneity ratio was 15.32%, showing strong spatial variability. The prediction accuracy of RF was the highest, and the absolute values of mean relative error (MRE) and root mean square error (RMSE) were both the smallest. Compared with OK and RK, the prediction accuracy of RF was improved by 5.53% and 2.49%, respectively, and the description of local details was more accurate and the simulation effect was the best. RF could accurately predict soil Ks in Ningdong, and provide data reference for understanding soil hydrological process and forest and grass vegetation construction in the study area.