From identification to runout prediction: a novel framework for landslide impact assessment along high-speed railways

High-speed railways， as critical transportation infrastructure， are highly vulnerable to landslide hazards. Although Interferometric Synthetic Aperture Radar (InSAR) enables the identification of potential landslides， conventional rate-based criteria often fail to identify landslides with small-magnitude deformation， leading to the underestimation of landslide hazards. Furthermore， the lack of an integrated framework for assessing landslide impacts limits effective railway hazard mitigation. In this study， an automated statistical workflow is proposed to identify potential landslides using the Directional-Root Mean Square Acceleration (D-RMSA) metric. The proposed D-RMSA metric integrates stage-wise accelerations throughout landslide deformation while resolving acceleration sign inconsistencies arising from different displacement directions， thereby enabling reliable identification of potential landslides exhibiting subtle or transient deformation. Consequently， the proposed workflow significantly increases the number of identified potential landslides， extending the capability of InSAR-based landslide identification. To assess the impacts of landslides on railways， landslide runout is predicted using a multi-task cascaded convolutional neural network， which learns the statistical dependency between landslide volume and runout. The proposed framework provides a scientific basis for landslide early warning and risk mitigation along high-speed railways. Moreover， the proposed framework is transferable to the assessment of potential landslide impacts on other major engineering infrastructure.