Landslide mapping using a combination of sentinel-2 multi spectral instruments and GIS data at Namwon, Jeollabuk-do, South Korea

With the recent development of satellite, aerial, and remote sensing technologies, it is easy to produce landslide inventory maps over a large area. In this study, the object-based framework was designed to address the limitations inherent in the pixel-based deep learning (DL) methodology. This framework explores the potential of combining Sentinel-2 MultiSpectral Instrument (MSI) satellite imagery and digital elevation models (DEMs) to enhance shallow landslide mapping across diverse terrains comprehensively. The study area for analysis and verification was selected as Jucheon-myeon, Namwon-si, and Jeollabuk-do, where significant large-scale landslides and slope failures occurred in 2020. As a result, the application of this framework led to the classification of 68 candidate sites spanning an area of 0.5 hectares or more. Site surveying was conducted on 20 random sites with a 1ha or more scale. Furthermore, six sites were selected where satellite imagery could discern the damaged areas. At these locations, the damaged area estimated by the framework was compared with the actual observed damaged area to assess accuracy. These rapid and cost-effective landslide mapping techniques can accurately estimate the location and extent of landslides and enhance the precision of sensitivity models and land management strategies.