Monitoring And Analysis And Prediction Of Land Subsidence In Beijing Plain Based On Multi-temporal InSAR Technology

As a new technology for remote sensing earth observation,synthetic aperture radar interferometry has the advantages of wide monitoring range,all-weather and all-time observation,but it also has many technical limitations,such as spatio-temporal incoherence and atmospheric delay effect.In order to overcome the impact of the above factors on surface deformation monitoring,multi-temporal InSAR technology came into being.The generation of multi-temporal InSAR technology not only improves the accuracy of deformation monitoring,but also can obtain the time series results of surface deformation.However,in the face of massive new SAR images,the traditional multi-temporal InSAR technology takes a long time to process and cannot be monitored dynamically.In view of the above problems,The progressive SBAS-InSAR technology integrating sequential adjustment came into being.Taking the Beijing plain as the study area,the land subsidence in the study area is analyzed and monitored based on multi-temporal InSAR technology and progressive SBAS-InSAR on the basis of sequential adjustment,and the relationship between land subsidence and groundwater flow field in the study area is quantitatively analyzed by Cross Wavelet and wavelet coherence spectrum.Finally,various models are used to simulate and predict the land subsidence in the study area.The main research contents and achievements are as follows:(1)This paper uses PS-InSAR and SBAS-InSAR technology to monitor and obtain the spatial-temporal evolution information of land subsidence in the study area from 2017 to 2019 based on sentinel-1A orbit lifting data,and the influencing factors of land subsidence in the study area are qualitatively analyzed in combination with relevant data.The results show that during the monitoring period,the main settlement areas in the study area are located in Tongzhou District and Chaoyang District in the east of the study area,and the annual settlement rates obtained by PS-InSAR and SBAS-InSAR are 115mm/a and120mm/a respectively;The land subsidence in the study area is affected by the change of groundwater level,regional base structure,thickness of compressible cohesive soil and geological disasters.(2)In this paper,a progressive SBAS-InSAR technology integrating sequential adjustment is used to dynamically,accurately and efficiently monitor the surface deformation information in the study area.Based on the existing achievements,this technology only needs to incrementally solve a small number of "newer" SAR images in the existing images and new SAR images,and then use the idea of sequential adjustment to obtain the surface deformation of all SAR images.Compared with the traditional SBAS-InSAR technology,this technology greatly improves the efficiency of surface deformation monitoring;The comparison between the surface deformation monitoring results of progressive SBAS and traditional SBAS shows that the monitoring results obtained by progressive SBAS-InSAR technology are highly reliable.(3)In this paper,the Cross Wavelet and wavelet coherence spectrum methods are used to analyze the relationship between land subsidence and groundwater level change in time and frequency domain.It is found that there is a lag or positive correlation between land subsidence and groundwater level,and the lag time varies with the study area.(4)This paper takes the annual settlement of the study area from 2015 to 2021 as the dependent variable and the groundwater level change,rainfall and human engineering activities as the independent variables,this paper establishes GM(1,1)model based on the annual settlement,the univariate linear regression model based on the groundwater level change,the multivariate linear regression model and BP neural network model and GA-BP neural network model which improved by genetic algorithm all based on groundwater level change,rainfall and human engineering activities.The model fitting prediction results show that when the settlement change in the study area is a nonlinear trend,the ideal prediction fitting model is GA-BP neural network model.When the settlement change in the study area is a linear trend,the better fitting prediction model is multiple linear regression model.