Soil erodibility assessment and model improvement based on soil microstructure in a typical watershed of the loess hilly-gully region, China

Soil erodibility is a key parameter that characterizes soil susceptibility to raindrop splash and runoff scouring, yet conventional empirical models based on soil texture and organic matter often fail to capture vegetation-induced improvements in soil structure following Grain for Green restoration. Five land use types were selected in a typical watershed of the loess hilly-gully region, and traditional empirical models were revised by incorporating soil microstructural variables across the 0–20, 20–40, and 40–60 cm soil layers to improve the assessment of soil erodibility under different land use types and elucidate its controlling mechanisms. The results showed that incorporating soil microstructure improved the K factor estimation, with a stronger revision effect for the EPIC model than for the TORRI and SHIRAZI models. Relative to the original estimates, the revised K values derived from the EPIC model decreased by 10.05%, 5.64%, and 1.84% in the 0–20, 20–40, and 40–60 cm soil layers, respectively, indicating that the effect was most pronounced in surface soils. Compared with cropland, the revised K values in the surface soils of woodland, shrubland, and grassland decreased by 15.74%, 11.08%, and 7.28%, respectively, and revised K values generally increased with soil depth across land use types. Root mass density, soil chemical properties, and aggregate stability jointly explained 98.35% of the variability in soil erodibility, including both their independent and interactive effects. Roots had the largest total negative effect on soil erodibility ( SPC = −0.865), and their indirect effects mediated through improved soil chemical properties and enhanced aggregate stability were 1.70 times greater than the direct effect. Soil chemical properties exerted a total negative effect on soil erodibility ( SPC = −0.489) that was secondary only to that of roots, acting through both direct regulation of soil erodibility and indirect enhancement of aggregate stability, whereas aggregate stability exhibited the strongest direct negative effect ( SPC = −0.411). Overall, vegetation restoration reduces soil erodibility primarily by improving soil structural stability through root development, changes in soil chemical properties, and enhanced aggregate stability. Incorporating structural variables into empirical models therefore improves the accuracy and applicability of soil erodibility prediction in restored ecosystems.