Automated identification of ground kinematic patterns based on InSAR time series displacement and K-SC clustering

Identifying and understanding ground kinematic patterns is essential for landslide disaster mitigation and early warning. The ground displacement time series obtained through Interferometric Synthetic Aperture Radar (InSAR) technique enables monitoring the evolution of the Earth's surface. However, interpreting enormous volumes of InSAR time series displacement is both labor-intensive and subjective, limiting its applicability across wide regions. In the study, we propose a method for automatically identifying slope kinematic patterns based on InSAR time series displacement and the K-Spectral Centroid (K-SC) clustering algorithm. Five distinct displacement patterns were identified, without accounting for the effects of inconsistent displacement direction and magnitude. The trend components of displacement patterns were then evaluated using least squares fitting, while seasonal components were extracted through empirical mode decomposition (EMD) and a modified separation method. The identified kinematic patterns are characterized by trend variations, periodicity of seasonal components and lag time relative to triggering factors. Our results enhance the applicability of InSAR technique for exploring intrinsic displacement behaviors associated with ground movement. Furthermore, the proposed method for automated identification and analysis of kinematic patterns is applicable to various types of ground movement.