A holistic approach of remote sensing, GIS, and machine learning for shallow landslide susceptibility mapping in Gaganbawada region of Western Ghats, India

Every year, the Western Ghats region experiences devastating landslide disasters that are closely linked to the region's wealth and growth, resulting in the loss of lives and damage to private and public property. Thus, it is important to identify the highly vulnerable places to minimize these losses. The major objective of this present research is to create a reliable landslide susceptibility map for the Gaganbawada region of Western Ghats. Support vector machine (SVM) is the machine learning based algorithm used to demarcate landslide susceptibility zones using a holistic approach of remote sensing and geographical information system. At first, the landslide inventory map is produced using Google Earth images and field studies. A total of 170 landslide (85) and non-landslide (85) points are used for the training and testing dataset, with a ratio of 70% and 30%, respectively. Secondly, 15 landslide influencing factors are selected. The predictive capabilities of the influencing factors are assessed using the information gain ratio and Pearson correlation coefficient to choose the optimal subset of influencing factors. Subsequently, a landslide susceptibility map is produced using the radial basis function kernel of the SVM model. As per the result, 23% of Gaganbawada's land area is in the high and very high landslip susceptibility zone, and 50% is in the low zone. The susceptibility map represents that only 8% of the land area is in the very high zone, on the other hand, 70.5% of historical landslides have been recorded. The resulting model is validated using the receiver operating characteristic (ROC), and several statistical evaluation matrices. According to the ROC evaluation result, the SVM model has an area under the curve (AUC) value of 0.88, showing that the present machine learning based model has acceptable prediction effectiveness and landslide susceptibility map result is reliable and effective for implementation.