Variations, controls and predictions of soil saturated hydraulic conductivity under different land use types in the alpine region of Tibet, China

Understanding the variations and controls of soil saturated hydraulic conductivity (K-s) under different land use types in the alpine region of Tibet is fundamental for predicting K-s in high-altitude regions where sample points are difficult to obtain and is important for establishing the regional soil hydrological model. In this study, soils samples under farmland, forestland, and grassland were collected from an east-west transect in southeastern Tibet, China. Basic soil properties and K-s were measured to analyze the influencing factors of K-s under different land use types and to establish suitable pedotransfer functions (PTFs) for predicting K-s in the alpine region. The average value of K-s followed the decreasing order of grassland (2.33 cm h(-1)) > farmland (2.12 cm h(-1)) > forestland (1.59 cm h(-1)). K-s of grassland (0.25-9.19 cm h(-1)) decreased gradually with increasing soil depth. K-s of farmland (0.11-8.82 cm h(-1)) increased initially and then decreased with increasing soil depth, whereas K-s of forestland (0.12-4.46 cm h(-1)) exhibited the opposite trend. There were significant differences in the main controlling factors of K-s under different land use types. In farmland, soil bulk density (BD), non-capillary porosity (P-n), clay and gravel content (GC) were the main drivers of K-s (P < 0.05). The dominant influencing factors of K-s in forestland were BD and total porosity (P-t) (P < 0.05), while P-n and GC were the main factors controlling K-s in grassland (P < 0.05). Compared with the multiple linear stepwise regression models and published PTFs, the BP neural network transfer model had better prediction accuracy for K-s in the alpine region. Our results not only provide hydraulic parameters for regional hydrological modeling and but also lay a foundation to construct K-s prediction models for other alpine regions in the world.