Soil hydraulic parameters theta (s), alpha, n, K (s), L and theta (r) of the van Genuchten-Mualem model were estimated using three pedotransfer functions (PTFs) based on soil properties for surface soils of the largest main tributary catchment (the Jialing River) of the upper Yangtze River in China. The soil database was from the second national soil survey of China with a spatial 30 X 30 arc-second resolution. According to the statistical analysis of the differences between the continuous-PTFs-estimated values of soil hydraulic parameters for the study catchment and the reference values for a specific texture class provided in the development database of a specific PTF, best estimations were obtained using the Wosten PTF. The Rawls & Brakenssiek PTF was good estimation for parameter theta (r) that was assumed as zero by Wosten PTF. The established higher theta (r) (0.08%) and lower K (s) (20 cm/d) and theta (s) (0.43%) in the mid-downstream area relative to the other areas of the catchment could lead to larger amounts of surface runoff and consequently provide higher energy to erode soil. Thus, these factors provide a supporting explanation for previously reported severe soil erosion occurring in this area. Spatial heterogeneity analysis for estimated hydraulic parameters in terms of semivariogram showed that the spatial correlation distance was in the range of 50-80 km and that the spatial variability (sill) was not large except for parameters K (s) and L. The semi-variance with the exponential model at the zero distance (nugget) was 30%-50% of the sill. This study provided a practical PTF approach for estimating soil hydraulic properties from soil survey data at a large watershed scale. The estimation results could provide better insight into the mechanism of surface runoff and soil erosion, which is important to better understand and manage erosion in the catchment.