Surface subsidence monitoring and prediction along high-speed railways using DS-InSAR and machine learning: A case study of the Jining section of Lunan High-Speed Railway

Surface deformation poses significant safety hazards to high-speed railways, necessitating high-precision monitoring and predictive analysis. Focusing on the Jining section of the Lunan High-Speed Railway (HSR), this study analyzes 36 Sentinel-1A ascending-track images using a DS-InSAR approach. One major subsidence zone was identified along the railway corridor, with a maximum subsidence rate of − 25 mm/a and cumulative subsidence of up to 75 mm during the monitoring period. Based on the monitored deformation time series, a Random Forest regression model was developed to predict subsidence at six representative points. The predictive performance of the Random Forest model was compared with that of the GM (1,1) model, showing substantially lower prediction errors for the Random Forest approach (RMSE ≤ 4.1 mm and MAE ≤ 3.2 mm). These results indicate that the proposed integrated framework effectively combines DS-InSAR monitoring with data-driven prediction, providing a practical and robust reference for subsidence assessment and decision support in the operation and maintenance of high-speed railway corridors.