Spatiotemporal reconstruction and post-failure stability analysis of rainfall-induced landslides via multi-modal SAR-optical data fusion

Monitoring rainfall-induced landslides remains challenging due to the limitations of single-source remote sensing in capturing full-cycle kinematic behaviors, from pre-failure creep to catastrophic collapse. To address this, this study proposes a temporally stratified, multi-modal framework that integrates Distributed Scatterer InSAR (DS-InSAR), dense optical flow (OF) analysis, and three-dimensional discrete element simulations. Applying this framework to the Shaziba landslide (China), DS-InSAR successfully captured early-stage creep signals even in vegetation-covered terrains. The pyramid-optimized OF algorithm achieved a 79% and 61% reduction in displacement Root Mean Square Error (RMSE) compared to traditional Co-registration of Optically Sensed Images and Correlation (COSI-Corr) methods in east–west and north–south directions, respectively, significantly improving rapid failure tracking. Physics-based simulations further reconstructed the dynamic failure process, identifying retrogressive sliding triggered by rainfall-driven crack infiltration into impermeable coal seams. This integrated approach not only reconstructs the full lifecycle of landslide evolution but also identifies post-failure hydro-mechanical feedbacks critical for early warning and targeted mitigation. The framework provides a transferable methodology for landslide monitoring and stability assessment in complex, data-scarce environments.