Construction and evaluation of pedotransfer functions for saturated hydraulic conductivity in the granite red soil regions of southern China

Abstract

Study region: Six typical regions of granite red soil in southern China. Study focus: Regional large-scale Ks analysis of granite red soil in South China is still limited to a large number of repetitive and expensive workloads. It is necessary to construct Ks PTFs suitable for the granite red soil region of South China. In this study, six different regions granite red soil in South China were used to validate the prediction of classical PTFs for Ks, the best combination of predictors was screened and PTFs developed based on stepwise regression methods and machine learning algorithms. New hydrologic insights for the region: (1) The coefficient of variation for Ks is 69.87% in the granite red soil region of South China, with a large spatial variability. (2) The six typical PTFs of Ks were poor predictors of Ks for soils in southern China, R2 < 0.4. (3) ''Clay+Silt+Sand+BD+SOM+NCP'' is the best combination of predictors of soil Ks in the granite red soil region of South China, which can be used in different machine learning algorithms. (4) The constructed MLR model “Ks= 1.501 + 0.043 ×NCP-0.995 ×BD+ 0.034 ×SOM-0.013 ×Clay” with KNN, ANN and BRT machine algorithm models have a prediction accuracy R2 greater than 0.8. This study provides different prediction models for soil Ks in the granitic regions of southern China and lays a theoretical basis for hydrological research in the these regions.

Keywords

• Soil saturated hydraulic conductivity (Ks)
• Granite red soils
• Pedotransfer functions (PTFs)