[Objective] Soil saturated hydraulic conductivity (Ks) serves as a critical parameter characterizing the transport of water and solutes in soil, playing a vital role in understanding and predicting soil moisture movement and erosion processes. Studying the variation characteristics and driving factors of Ks across various vegetation restoration models in gully watershed regions of the Loess Plateau is essential for advancing regional soil erosion mitigation efforts. [Methods] Thirty-eight typical vegetation restoration plots ( 5 bare land, 3 dry land, 8 arbor forest land, 3 other forest land, 5 shrub forest land and 14 other grassland ) in Zhifanggou small watershed were selected as the research objects. The Ks, soil physical and chemical properties and root characteristics of 0-10 cm soil layer in different plots were measured. Spearman correlation analysis, partial least squares regression ( PLSR ) and multiple stepwise regression analysis were used to reveal the influence mechanism of vegetation restoration mode on Ks, and the Ks prediction model was established.[Results] Significant differences in saturated hydraulic conductivity (Ks) were observed across vegetation restoration patterns (P<0.05). The mean Ks values followed the order: shrub forest land (1.46 mm/min) > other forest land (1.36 mm/min) > other grassland (1.23 mm/min) > arbor forest land (1.04 mm/min) > dry land (0.65 mm/min) > bare land (0.15 mm/min). Spearman analysis and PLSR revealed that sand fraction, clay content, bulk density, maximum water retention, non-capillary porosity, root volumetric density, root mass density, and root mass densities in 1-2 mm and 0-1 mm diameter classes were critical determinants of Ks. Stepwise regression showed substantial improvement in model explanatory capacity upon integrating root parameters. The integrated model highlighted that intermediate roots optimized Ks through pore network expansion and compaction effect mitigation. [Conclusion] Vegetation restoration markedly elevates Ks via root-mediated soil structural modifications, particularly through pore architecture optimization. The results offer a scientific foundation for refining ecological restoration strategies and enhancing hydrological model parameterization in Loess Plateau ecosystems.