Estimating the indices of soil erodibility to wind erosion using pedo- and spectro-transfer functions in calcareous soils

Soil aggregate size and stability are two important factors affecting soil erodibility to wind erosion. In this study, we developed new models to quantify soil erodibility to wind erosion by looking at the mean weighted aggregate diameter (MWD) and the wind erodible fraction (EF) of the soil. These two erodibility indices together with the spectral reflection of soil in the range of Vis- NIR (400-2500 nm) were measured in 511 soil samples. Pedotransfer functions (PTF) and Spectro-transfer functions (STF) were built using the Multiple Linear Regression (MLR), Partial Least Squares Regression (PLSR), and Support Vector Regression (SVR), considering train(70%) and test (30%) datasets. Result showed that shear strength (SS), organic matter (OM), penetration resistance (PR), and clay content had a significant coefficient (p < 0.05) for predicting MWD and EF indices. Vis- NIR spectroscopy method performed better than PTF method and wavelengths of 513, 773, 872, 1256, 1414, 1488, 1908, 2042, 2210, and 2311 nm were introduced as the key wavelengths for the estimation of soil erodibility indices based on the regression coefficient. Results of the predictive models revealed that SVR outperformed (RMSE = 0.12 mm, RPIQ = 2.56 for MWD; and RMSE = 9.32%, RPIQ = 2.24 for EF) PLSR (RMSE = 0.12 mm, RPIQ = 1.98 for MWD; and RMSE = 9.40%, RPIQ = 1.84 for EF). This study proved that Vis- NIR spectroscopy is a promising method for the prediction of MWD and EF indices as soil erodibility indices.

Automated summary

Soil aggregate size and stability play important roles in soil erodibility to wind erosion. New models were developed in this study to quantify soil erodibility based on mean weighted aggregate diameter (MWD) and wind erodible fraction (EF), along with spectral reflection measurements in the Vis-NIR range. Various regression methods were used to build pedotransfer functions (PTF) and spectro-transfer functions (STF), showing that shear strength, organic matter, penetration resistance, and clay content were significant factors in predicting MWD and EF indices. Vis-NIR spectroscopy method outperformed PTF method, and specific wavelengths were identified as key for estimating soil erodibility indices. Support Vector Regression (SVR) yielded better results compared to Partial Least Squares Regression (PLSR), indicating the potential of Vis-NIR spectroscopy for soil erodibility prediction.