Saturated hydraulic conductivity is a significant soil hydraulic parameter, which plays an important role in the research of seepage and solute transport. Pdeo-Transfer Functions (PTFs) can indirectly predict this parameter instead of large-scale sampling and indoor and outdoor tests. However, due to the spatial variability of soils and the limitations of methods, previous predictions of PTFs are not accurate enough. This study takes different soil layers in the vadose zone of Yanlou Township, Lankao County, the lower reaches of the Yellow River as the research object. Based on 64 borehole data, nine basic physicochemical properties such as soil particle size distribution (clay, silt and sand content), soil structure fractal dimension, dry bulk density, total porosity, pH value, organic matter and electrical conductivity are considered as the influencing factors of the PTFs. Based on Stepwise Regression, Principal Component Regression and Principal Component Analysis-Geographically Weighted Regression (PCA-GWR), the saturated hydraulic conductivity of different soil layers is regressed separately. After comparison of precision, Geographically Weighted Regression Kriging is carried out for saturated hydraulic conductivity based on PCA-GWR. The results reveal that the prediction accuracy of PCA-GWR method has significant advantages except for the surface silt. The difference of the distribution of saturated hydraulic conductivity and PTFs in different soil layers shows that, due to the rapid and changeable sedimentation process in the lower reaches of the Yellow River, the breach and diversion of the Yellow River and human activities, the phase change of soil layer in the vadose zone is intense, and the saturated hydraulic conductivity also has obvious non-stationary characteristics in the plane. The local geological process and human activities are the deep reasons for the poor regression Kriging estimation results of the second layer of silt.