A Comprehensive Study of Machine Learning and Deep Learning methods for Landslide Susceptibility Mapping

Abstract

Landslides are one of the most frequently occurring natural hazards worldwide and cause serious damage to human life, severe social and economic loss. Landslide susceptibility is the likelihood of a landslide occurring in the area determined by local terrain conditions predicting where they are more likely to occur. This paper provides a detailed survey on numerous methods and approaches of determining landslide susceptibility. The key challenges for each method are examined, summarized and discussed with their advantages and disadvantages. Moreover, to improve the accuracy of satellite images several object-based and pixel-based classification methods have been discussed. Based on the detailed survey carried out, our attention is to extend the existing research by considering the root cause and primary triggering factors of landslide to better predict the likelihood in different parts of western ghats region, Karnataka, India. This survey can help the other researchers to know the status on Landslide Susceptibility and provide direction for further research.