Land Subsidence Monitoring and Forecasting in Mining Areas Based on SBAS-InSAR and SSA-BP Neural Models

This study investigates ground subsidence in the mining areas of Heze City, Shandong Province, using SBAS-InSAR technology and an SSA-BP neural network model. A total of 54 Sentinel-1A SAR images acquired from February 2019 to January 2024 were processed to derive cumulative subsidence and average subsidence rate maps. The results show that severe subsidence mainly occurs in three key coal mining zones, with a maximum annual rate of 182 mm/year and cumulative subsidence reaching − 878 mm. The subsidence area spans approximately 1129.28 km², accounting for 59.33% of the total study area. To predict future deformation trends, a back propagation neural network optimized using the Sparrow Search Algorithm was constructed. Model performance was evaluated against LSTM, SVR, and traditional BP networks. The SSA-BP model achieved the highest accuracy, with RMSE values as low as 1.89 mm. Comparison with leveling measurements validated the reliability of the SBAS-InSAR results. This study demonstrates the feasibility and accuracy of combining SBAS-InSAR and SSA-BP models for subsidence monitoring and prediction in coal mining areas, offering valuable references for early warning and disaster prevention efforts.