Surface Deformation Monitoring and Prediction in the Baihetan Hydropower Station Area Based on Small Baseline Subset Interferometric Synthetic Aperture Radar and Grid Search Support Vector Machine

The Baihetan Hydropower Station, as the world's second-largest hydropower station, demands increased attention due to its significance. In our study, we first analyzed the surface deformation in the area using Sentinel-1A satellite data and SBAS-InSAR technology. Subsequently, based on the deformation monitoring results, we conducted ground surveys, confirming an active landslide at the left bank tributary of the Jinsha River. Following this, we evaluated the feasibility of LSTM and GS-SVM models using metrics such as RMSE, MAE, and maximum error. After confirming the models' viability, we forecasted the future 96-day settlement time series of the landslide. The monitoring results showed that there are five significant deformation areas upstream of the Baihetan Hydropower Station, but only one is an active landslide. The model evaluation results indicated that the GS-SVM model exhibited higher stability and accuracy compared to the LSTM model. The prediction results revealed that the landslide area is entering another period of intensified deformation. This study provides important references and guidance for understanding regional surface deformation patterns and geological hazards.