Large-scale clustered landslides triggered by intense rainfall or earthquakes often occur rapidly across residential areas, roads, engineered slopes, and mountainous terrain, creating major challenges for emergency investigation and risk assessment. Optical remote sensing, UAV and field surveys provide detailed landslide boundaries and damage information, but their early post-event use is often constrained by weather, accessibility, flight coverage, and processing time. This study developed and implemented an integrated satellite SAR-based framework for rapid landslide mapping and deformation analysis in urban mountainous environments.
For landslide mapping, an amplitude–texture fusion method was established using an asymmetric temporal-window design. Multiple pre-event SAR amplitude images were used to generate a stable mean or median backscatter background, reducing speckle noise, scene-specific fluctuations, and local outliers. The first one or two post-event SAR images were then averaged to enable rapid post-disaster mapping. Abnormal amplitude changes were extracted using dB difference, log-ratio, and standardized anomaly indices. The resulting change images were further analyzed using gray-level co-occurrence matrix texture features, including contrast, entropy, and correlation, to characterize surface roughness, boundary sharpness, spatial heterogeneity, and patch structure. Principal component analysis was applied to reduce feature redundancy, and the first three principal components were used to generate an RGB composite that enhanced the separability between landslide-disturbed patches and background surfaces. The workflow successfully supported landslide candidate mapping, landslide density assessment, and UAV survey prioritization.
For deformation analysis, Sentinel-1 time-series InSAR was conducted in the same study area to retrieve pre- and post-event surface deformation. By integrating InSAR deformation with SAR amplitude-based mapping, the study distinguished newly triggered shallow landslides, reactivated slopes with pre-existing slow movement, and areas with continuing post-failure deformation and secondary risk. Deformation time-series clustering was also performed to identify different temporal deformation patterns. Displacement series were first interpolated, smoothed and normalized, then evaluated using silhouette scores, and finally clustered using K-means. The integrated results demonstrate that combining SAR amplitude changes, texture features, and time-series InSAR provides an effective framework for rapid mapping, activity interpretation, and post-disaster assessment of clustered landslides in urban mountainous environments.
 

SAR,InSAR,Change detection,Clustered landslide,Geohazard