Assessing gully erosion susceptibility using topographic derived attributes, multi-criteria decision-making, and machine learning classifiers

Gully erosion is an erosive process that contributes considerably to the shape of the earth's surface and is a major contributor to land degradation and soil loss. This study applied a methodology for mapping gully erosion susceptibility using only topographic related attributes derived from a medium-resolution digital elevation model (DEM) and a hybrid analytical hierarchy process (AHP) and the technique for an order of preference by similarity to ideal solutions (TOPSIS) and compare the results with naive Bayes (NB) and support vector machine learning (SVM) algorithms. A transboundary sub-basin in an arid area of southern Iraq was selected as a case study. The performance of the developed models was compared using the receiver operating characteristic curve (ROC). Results showed that the areas under the ROC were 0.933, 0.936, and 0.955 for AHP-TOPSIS, NB, and SVM with radial basis function, respectively, which indicated that the performance of simply derived AHP-TOPSIS model is similar to sophisticated NB and SVM models. Findings indicated that a medium resolution DEM and AHP-TOPSIS are a promising tool for mapping of gully erosion susceptibility.