PS-SMaRT v1.0: A model for automatic clustering of unstable slopes from PS-InSAR time series coupled with soil-moisture anomalies

PS–SMaRT (Persistent scatterer–soil moisture analysis for risk and triggering) v1.0 is an open-source, Python-based model developed to automatically detect and characterize unstable slopes by integrating multi-temporal persistent scatterer interferometry (PS–InSAR) deformation time series with soil-moisture wet anomaly data. Despite advances in multi-temporal InSAR, standardized frameworks integrating ground deformation and soil-moisture anomalies for automated landslide hazard assessment remain limited. The framework bridges geodetic observations and hydrological processes to derive physically interpretable indicators of slope instability that are scalable across landslide-prone regions. The methodology is based on the principle that slope failure is jointly controlled by displacement trend and soil-moisture wet anomalies. Long-term deformation trends obtained from SAR interferometry represent sustained ground motion, whereas soil-moisture anomalies capture surface wetness influencing pore water pressure and shear strength. By jointly analysing these data, PS–SMaRT v1.0 isolates persistent deformation coinciding with anomalously wet conditions, identifying potential zones of incipient instability. The modular workflow includes downslope projection of line-of-sight velocities, statistical filtering of valid persistent and distributed scatterers, DBSCAN-based clustering of coherent deformation, and cross-validation of cluster polygons against soil-moisture anomalies and topographic wetness index (TWI) maps. A composite hazard index (HI) integrating deformation, soil-moisture wet anomaly frequency, and topographic conditioning is then derived to generate spatially explicit hazard maps. Demonstrated over the Petacciato slow-moving landslide in Italy, PS–SMaRT v1.0 delineates hydrologically conditioned deformation clusters consistent with field evidence. The validation analysis yielded a recall accuracy of 70.6%, demonstrating the framework’s capability to delineate unstable slope sectors. Its physically grounded, transparent, and transferable design facilitates adaptation to current and future SAR missions (e.g., NISAR L-band, ROSE-L) and supports reproducible, quantitative landslide-hazard assessment.