Landslide Susceptibility Mapping for Road Corridors Using Coupled InSAR and GIS Statistical Analysis

This study presents landslide susceptibility mapping by using coupled GIS statistical analysis and interferometry synthetic aperture radar (InSAR) data applied to a road corridor in West Sulawesi, Indonesia. Landslide-contributing factors in the road corridor including slope angle, distance to drainage, lithology, distance to road, distance to tectonic fault, and rainfall, were selected based on knowledge of landslide triggering mechanisms, numerically analyzed with the multicriteria decision of the analytical hierarchy process (AHP), and controlled with the variance inflation factor (VIF) in the multicollinearity analysis. Then, the data associated with slope angle, the distance of the road to the slope, and the distance of the road to natural drainage were obtained from the digital elevation model (DEM) of the road corridor. In addition, the data of the lithological type of the area in the road corridor and the distance of the road to the tectonic faults were derived from a geology map, whereas the data of 10 years of daily rainfall were collected from three rainfall stations in the proximity of the road corridor and employed as rainfall data. Causal relations between those contributing factors and landslide occurrences were identified, statistically analyzed with the AHP, and numerically converted as landslide ratings into a GIS statistics–based landslide susceptibility map (LSM). In a parallel way, the interferogram of Sentinel-1’s synthetic aperture radar images of the road corridor was also generated to observe the ground movement rate. The observed ground movement rates were then numerically converted into landslide ratings in the InSAR-based LSM. By overlaying these maps, a coupled GIS statistics and InSAR-based LSM of the road corridor was generated. To validate its accuracy, the landslide density index ($R$-index) was calculated, in which the highly and very highly susceptible zones in the LSM were compared with the actual sliding areas in the landslide inventory data. In comparison with the GIS-AHP-based LSM from a prior study, which had an $R$-index of 91.03%, the GIS statistical analysis and InSAR-based LSM’s $R$-index was determined to be 97.09%, suggesting high accuracy and improved prediction. The results indicated that creating a landslide susceptibility map using GIS statistical analysis and InSAR would be beneficial in reducing the risk of landslides for road infrastructure.