Identification and Prediction Inversion of Mining Area Subsidence by Integrating SBAS-InSAR and EMD-ARIMA Model

Due to the use of heavy industry equipment and frequent mining activities, the rock strata and the surface of Dexing Copper Mine were prone to settlement and deformation. In order to accurately analyze the time series settlement variation of the surface, this paper used Small Baseline Subset InSAR (SBAS-InSAR) technology to obtain 23 Sentinel-1A orbit SAR images covering Dexing Copper Mine for temporal settlement monitoring. The monitoring results show that there are three obvious settlement funnels within the mining area, and the total settlement area is about 7.63 km2, located in the southwest of the Copper Factory, the southwest of Fujiawu Stope and between Duyang Lake and the Copper Factory, the maximum cumulative settlement in the area is −1131.9 mm, and the maximum annual average settlement rate is −375.5 mm/a. In order to analyze the change trend of settlement in mining area timely and accurately, make full use of monitoring data to carry out disaster prevention and warning in mining area, and solve the problem that the prediction accuracy of traditional settlement prediction model is not high, this paper proposes a time series settlement prediction model that takes into account the periodic and seasonal variation of settlement-EMD-ARIMA (Empirical Mode Decomposition-Autoregressive Integrated Moving Average Model), further, perform retrospective predictive analysis of the monitoring data. The results show that compared with the traditional GM (1,1) and ARIMA models, the average RMSE of EMD-ARIMA model is increased by 78.5% and 49.5%, and the average MAE is increased by 74.4% and 52.4%, respectively. The average MAPE increased by 81.2% and 47.9%, respectively. It shows that the EMD-ARIMA model has high prediction accuracy, strong stability and adaptability, and can be used to predict the time series data of mining settlement, which can provide scientific guidance for the safety production of mining areas.