Prediction of soil loss due to erosion using support vector machine model

Soil erosion refers to the detachment and removal of soil particles from land (topsoil), by the natural physical forces (water, glacier and wind). Soil erosion causes soil loss in the catchment or any land areas severely impacting agriculture activity, sedimentation in the dam reservoirs, and hampering developmental activities. Therefore, it is desirable to accurately measure soil loss due to erosion for the development and management of an area. With this objective, a well-known machine learning algorithm Support Vector Machine (SVM) has been used in the development of the soil loss prediction model. Eight erosion affecting variable inputs: ambient temperature Tair, rainfall, Antecedent Moisture Conditions (AMC), rainfall intensity, slope, vegetation cover, soil temperature Tsoil and moisture of the soil. Data on published literature was used in the model study. The accuracy of the proposed SVM was assessed by using three statistical performance evaluation indicators namely Person correlation coefficient (R), Root Mean Squared Error (RMSE), Mean Squared Error (MAE). Partial Dependence Plots (PDP) was used to investigate the dependence of prediction results of eight input variables used in the model study. Model validation results showed that SVM model performed well for the prediction of soil loss for testing (R = 0.8993) and also for training (R=0.9123). Rainfall intensity and vegetation cover were found to be the two most important affecting input parameters for the soil loss prediction.