Landslide Displacement Prediction Using Interferometric Synthetic Aperture Radar and Machine Learning Techniques

Landslides represent a major geohazard, often resulting in severe infrastructure damage and economic losses. Conventional ground-based monitoring techniques are limited in spatial coverage and continuity. With the advancement of remote sensing, Small Baseline Subset Interferometric Synthetic Aperture Radar (SBAS-InSAR) provides high-resolution and continuous surface displacement data, enabling large-scale landslide monitoring. This study focuses on the Gunnison region of Colorado, a mountainous area with complex terrain and recurrent slope instabilities. Using SBAS-InSAR data acquired from July 2022 to January 2024, we construct time series displacement maps to support landslide forecasting. Multiple data-driven models, including Random Forest (RF), Extreme Gradient Boosting (XGBoost), Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU), are implemented to predict displacement at the final time step based on the previous data. Model performance is evaluated using RMSE, MAE, and R². Results show that Random Forest achieved the best predictive accuracy with an RMSE of 3.68 mm. The spatial comparison between observed and predicted displacement further confirms the model’s generalization capability. This study demonstrates that the integration of InSAR time series and machine learning models provides an effective approach for landslide early warning and hazard mitigation.