Landslide Mapping and Susceptibility Assessment with Remote Sensing Data Using Deep Neural Networks

Abstract

Landslides represent a global incident with significant implications for loss of life, property damage, and infrastructure disruption. This research encompasses a multidisciplinary approach, combining geology, geography, satellite imagery, and artificial intelligence to address landslide-related challenges. Recent years have witnessed a surge in landslide occurrences due to extreme weather events, environmental degradation attributed to human activities, and other anthropogenic influences. Landslide susceptibility assessment is pivotal for planning and disaster management in mountainous regions. Over the past two decades, numerous statistical methods have been employed in landslide mapping. newlineInitially, the study commenced in Southern India and delved deeper into a specific location within Kerala, the Idukki district. The dataset utilized for this research was sourced from authorized government websites, remote sensing data, IMD data, and an extensive field visit to validate data from diverse sources. The dataset covers 2011 to 2021, with an updated landslide inventory map available until 2021 for reference. Data preprocessing involved filling in missing data and excluding minor and non-repeated landslide occurrences. Field visits, literature reviews, and excluding less influential factors informed the selection of primary triggering factors newline