Forecasting InSAR-derived slope movement from climate records at Baihetan reservoir

Interferometric Synthetic Aperture Radar (InSAR) has become a powerful tool for monitoring hillslope deformation. Recent advances have focused on integrating InSAR with predictive models, yet limited effort has been dedicated to developing scenario-based prediction of hillslope deformation informed by climatological records to assess potential geomorphological hazards under future extreme weather conditions. This study investigates slope instability near the Baihetan Reservoir in China, where notable deformation followed its impoundment in 2021. Using Sentinel-1 images (2021–2024), we applied SBAS and PSI techniques to detect 78 and 65 deformation anomalies from descending and ascending orbits, respectively. A two-dimensional Temporal Convolutional Network (2D-TCN) was developed to predict deformation based on slope angle, precipitation, temperature, and reservoir level. We simulated eight extreme weather scenarios based on 40 years of historical climate data. Results show the model reliably predicts spatiotemporal deformation, with the most hazardous scenarios involving >740 mm precipitation, reservoir level rise, and temperatures >20 °C. For example, the Xiapingzi landslide showed >100 mm deformation within 60 days under one scenario. Although the model itself is not directly transferable to other regions, the framework and workflow are. This approach supports proactive hazard management by quantifying landslide responses to extreme weather, providing a valuable tool for scenario-based risk assessment.