Analysis and prediction of ground deformation in Yinxi Industrial Park based on time-series InSAR technology.

Monitoring ground deformation in industrial parks is of great importance for the economic development of urban areas. However, limited research has been conducted on the deformation mechanism in industrial parks, and there is a lack of integrated monitoring and prediction models. Therefore, this study proposes a comprehensive monitoring and prediction model for industrial parks, utilizing time-series Interferometry Synthetic Aperture Radar (InSAR) technology and the Whale Optimization Algorithm-Back Propagation (WOA-BP) neural network algorithm. Taking Yinxi Industrial Park in Baiyin District as a case study, we used 68 scenes of Sentinel-1A ascending and descending orbit data from June 2018 to April 2021. The Stanford Method for Persistent Scatterers-Permanent Scatterers (StaMPS-PS) and the Small Baseline Subsets-Interferometry Synthetic Aperture Radar (SBAS-InSAR) technologies were employed to obtain the surface deformation information of the park. The deformation information obtained by the two technologies was cross-validated in terms of temporal and spatial distribution, and the vertical and east-west deformation of the park was obtained by combining the ascending and descending orbit data. The results show that the deformation feature points in the line of sight (LOS) direction obtained by the two technologies have a high consistency in spatial distribution, using the ascending orbit data as an example. Additionally, the SBAS-InSAR technology was used to obtain the east-west and vertical deformation results of the park after merging the ascending and descending orbit data for the same period. It was found that the park is mainly affected by vertical deformation, with a maximum subsidence rate of 14.67 mm/yr. The subsidence areas correspond to the deformation positions observed in field survey photos. Based on the ascending orbit deformation data, the two technologies were validated with 585 points of the same latitude and longitude, and the coefficient of determination R was found to be 0.82, with a root mean square error (RMSE) of 2.20 mm/a. The deformation rates were also highly consistent. Due to the 47% increase in the number of sampling points provided by the StaMPS-PS technique compared to the SBAS-InSAR technique, the former was found to be more applicable in the industrial park. Based on the ground deformation mechanism in the park, we combined the StaMPS-PS technique with the WOA-BP neural network to construct a deformation zone prediction model. We conducted predictive studies on the deformation zones of buildings and roads within the park, and the results showed that the WOA-optimized BP neural network achieved higher accuracy and lower overall error compared to the unoptimized network. Finally, we analyzed and discussed the geological conditions and inducing factors of ground deformation in the park, providing a reference for a better understanding of the deformation mechanism and early warning of disasters in the industrial park.