Analysis And Prediction Of Land Subsidence Based On ARIMA-GRU Model

Land subsidence is an important geological disaster,which will not only destroy buildings,roads and other infrastructure,but also have a negative impact on the urban ecological environment.In recent years,in the process of high-speed urban construction,under the background of the coordinated development of Beijing-Tianjin-Hebei and the strategy of building a sub-center in Beijing,the population is excessively concentrated in the built-up area,and social and economic activities encroach on the ecological space,which will also bring a new round of challenges to the evolution of land subsidence in Tongzhou sub-center Therefore,obtaining long-time series,dynamic and accurate land subsidence monitoring information is helpful to master the law of land subsidence change,which is of great practical significance for dynamic and continuous land subsidence monitoring and effective prediction in Tongzhou District of Beijing.Starting with the factors of urbanization process,this paper discusses the formation mechanism and influencing factors of land subsidence,and summarizes the current research situation at home and abroad.At the same time,this paper also introduces the monitoring method and prediction model of land subsidence,which provides a scientific basis for the prevention and treatment of land subsidence.The main work of this paper is as follows:(1)Summarized the research status of land subsidence monitoring methods and land subsidence prediction methods at home and abroad,and discussed the feasibility of ARIMA(Autoregressive Integrated Moving Average)-GRU(Gated Recurrent Unit)combined model to predict land subsidence;(2)Taking 2015-2021 as the research period,the leveling data of Tongzhou District were measured and calculated by using Beijing land subsidence monitoring system,with a total of95 leveling points and elevation data for 8 consecutive years.Sentinel II data are used for remote sensing images,and socio-economic data are selected from the statistical yearbook data of Tongzhou District,Beijing from 2015 to 2021.Firstly,the improved multilinear model and land use monitoring based on RS and GIS are used to study the process of urban expansion;Secondly,the relationship between urban expansion and land subsidence in Tongzhou district is analyzed by using Sentinel II satellite remote sensing image data and GIS spatial analysis method.Finally,according to the correlation analysis,the relationship between them is studied.(3)Establish an ARIMA-GRU combined model to predict land subsidence.Taking the annual leveling point elevation data from 2005 to 2015 as the reference value,the inverse variance method is used to determine the weight proportion of the optimal solution of the combined model,and an ARIMA-GRU combined model method is established to predict the leveling point elevation from 2016 to 2019.Secondly,two separate models and ARIMA-GRU combined model are used to predict the trend change of future data.(4)According to the results of data analysis and feedback,make a scientific evaluation of the combined model and explain its innovation.Three model indexes,MSE,RMSE and MAPE,are used for quantitative analysis,and the feasibility of the model is verified.