InSAR-based landslide detection method with the assistance of C-index

Interferometric Synthetic Aperture Radar (InSAR) has been increasingly used in landslide detection over wide areas. However, atmospheric delays, phase unwrapping errors, and noise, which behave like deformation signals in InSAR deformation results, pose a challenge for semi-automatic and automatic landslide detection. C-index can assistant in landslide interpretation by assessing the rationality of InSAR deformation and filtering out non-real landslide deformation signals. Yet, few studies have analyzed its applicability in identifying landslide from InSAR results. In this study, we develop a method that uses C-index to assist in landslide detection using InSAR. We validate our method in a selected region of the Jinsha River basin, China. We first using multi-temporal InSAR (MT-InSAR) to obtain the deformation rate of the study area. Sixty-nine and 47 suspicious slope deformation areas are extracted from the ascending and descending deformation results, respectively. Next, C-index is used to automatically exclude 28 and 12 false deformation areas from the ascending and descending InSAR results, respectively. Remarkably, all the excluded false deformation areas do not correspond to landslides, suggesting the reliability of the proposed method. Finally, by visually interpreting the optical images of the remaining deformation results, we successfully detect 54 landslides. Among these, 9 landslides can be detected by both the ascending and descending Sentinel-1 data, and 15 landslides are located on the riverbanks, posing a potential risk of river blockage in the event of failure. The presented landslide detection method effectively filters out false deformation areas, ultimately enhancing the efficiency and accuracy of landslide detection over wide areas using InSAR.