Deformation monitoring and risk assessment of railway transit based on SBAS-InSAR and FR-machine learning

Urban rail transit in Nanning is an important part of urban development. It is exposed to complex environmental conditions. During construction and operation, ground deformation may occur. This deformation can increase the risk of potential geological hazards. Therefore, deformation monitoring and risk assessment are essential for disaster prevention and control. This need is especially important for large-scale linear infrastructure projects and resource development. This study used Small Baseline Subset InSAR (SBAS-InSAR) for wide-area monitoring, we derived deformation results along the rail corridors in Nanning from January 2021 to October 2023. We further applied hotspot analysis and kernel density analysis to rapidly identify areas with large deformation along the rail lines. The results show that: some of the surface subsidence rates around parts of the railroad line exceed − 40 mm/yr, with pronounced subsidence funnels. Most of the area along the route is relatively stable, with localized subsidence rates of -30 mm/yr and a maximum subsidence of -123 mm. We used hotspot analysis to identify high subsidence zones. We then estimated ground subsidence susceptibility along the Nanning metro lines using Frequency Ratio (FR), Random Forest (RF), XGBoost, and LightGBM. The three models all achieved Area Under the Curve(AUC) values consistently above 0.90. We compared R2 and RMSE among the models. Based on R2 and RMSE, LightGBM shows better predictive performance in the study area. The susceptibility maps from the three coupled models show similar spatial patterns. The the “very high” and “high” susceptibility areas are mainly located in Liangqing District.