Modeling soil erosion susceptibility considering morphometric analysis and SWAT application: policy recommendation to achieve SDGs

Soil health promotes the agriculture productivity and support the life system on earth. The soil erosion becomes hazardous when the healthy soil system services come at risk. The Sanjai River basin (Jharkhand) is selected as a vulnerable soil erosion zone where in this subtropical plateau area, the repeated earth movement and river terraces are very common and the morphometric characteristics comes under the climate change threat. Soil erosion from various sub basin has been tracked by the application of SWAT model along with the consideration of morphometric and tectonic parameter. The application of machine learning model (“Random Forest, Artificial Neural Network, Deep Learning Neural Network and Maximum Entropy Model”) has established the soil erosion prone area. The value of AUC of RF, ANN, DLNN and MaxEnt model with the help of the training datasets are 0.801, 0.825, 0.898 and 0.901 respectively where MaxEnt is the most optimal model considering training and validation datasets. All of the indices in the MaxEnt model have values of 0.96, 0.87, 0.92, 0.91, and 0.91 when “training datasets,” “sensitivity, specificity, PPV, NPV, and F score” are included. All of the indices in the MaxEnt model have values of 0.94, 0.85, 0.91, 0.92, and 0.89 when “validation datasets,” “sensitivity, specificity, PPV, NPV, and F score” are taken into consideration. The VIF and TOL ranges are, respectively, 0.66 to 4.00 and 0.25 to 1.51. The SDGs are hampered by the characteristics, which have an impact on the soil erosion caused by rainfall in the Sanjai River Basin. This research will create a plan that will substitute high agricultural production with soil erosion management in order to develop the soil health ecosystem.