Research On The Identification System And Prediction Evaluation Model Of Land Subsidence Factors In Heze City

Land subsidence induced by multiple factors is an important natural geological disaster,which affects the development of many urban areas in the world.The uneven land subsidence can lead to the development of corresponding disasters such as ground fissures.It causes damage to urban buildings and infrastructure,and risks to human production and life.The Heze City of Shandong Province was selected as the research object.By fully collecting land subsidence related historical and current data and actual monitoring data,combined with field surveys,based on geographic information systems(GIS),a general land subsidence evaluation model was constructed using theoretical analysis,machine learning,mathematical modeling,and numerical calculations.The model uses machine learning methods,which can efficiently and conveniently deal with the problem of multi factor identification,prediction and evaluation of land subsidence.The main factors affecting the development of land subsidence in Heze city are determined quantitatively by using the established model,and the development trend of land subsidence in Heze city is analyzed and studied.The main research work and results of this paper are as follows:(1)The physical geographical conditions,regional basic geology,hydrogeology and engineering geology conditions of Heze City were collected and investigated in detail.Then combined with the existing relevant literature,the history and current situation of land subsidence in the study area are studied and analyzed,and the influencing factors of land subsidence are determined.Twelve static factors and three dynamic factors affecting land subsidence in Heze were extracted from multi-source data sets(digital elevation model,satellite remote sensing big data,thematic data maps of hydrological bureau and water resources bureau,water resources report,etc.).At the same time,the extraction method of each factor is introduced in detail,and the influencing factor database of Heze land subsidence is established in GIS system.(2)Based on the remote sensing satellite data set and the land subsidence monitoring system of "four networks in one"(leveling monitoring network,GNSS monitoring network,stratified and bedroom-rock monitoring network,and groundwater level monitoring network)in Shandong Province,the original learning sample data of land subsidence in Heze City from 2017 to 2020 were extracted by using InSAR remote sensing data processing technology.The original learning sample data set of land subsidence data was divided into two parts,one accounting for 70%as model training sample data set,and the other accounting for 30%as model test sample data set.All data sets were imported into the random forest model for continuous iterative training,and a more accurate identification and prediction evaluation model of land subsidence factors in Heze City was obtained.(3)According to the characteristics of rainfall time series monitoring data in Heze city,different rainfall SARIMA time series prediction models were designed and established for each region.In this paper,the continuous monthly mean monitoring data of rainfall in Heze from 2008 to 2020 are simulated and analyzed,and the behavior patterns of historical rainfall in different regions are determined.After data preparation and preprocessing,' data conversion,parameter identification and model verification,the SARIMA model has a high degree of goodness of fit.It has a low Ljung-box Q value and a P value significantly greater than 0.05.and the residual of the model is white noise,which shows that the model has a high accuracy and good predictive ability.(4)By analyzing the hydrogeological conditions of the study area,the structural characteristics of the underground aquifer group and the flow state of groundwater,the conceptual model of groundwater in Heze City was established.According to the actual situation of the study area,the differential mathematical model of groundwater movement was established by referring to the boundary conditions and parameter properties of the conceptual model,and the three-dimensional unsteady flow model of groundwater in the study area was determined to be heterogeneous and anisotropic.In the GIS system,the high-resolution remote sensing geological data,borehole data,groundwater monitoring data and other data were digitized and converted into a unified format to construct a groundwater numerical model database.Combined with the MODFLOW program package,the groundwater movement changes from 2017 to 2020 were simulated and analyzed.The results showed that the quantitative groundwater flow model can predict the dynamic changes of groundwater flow field scientifically and reasonably.Under the current planning conditions,the shallow groundwater level slowly decreases,the deep groundwater level gradually rises in most areas,and only the groundwater at the funnel slowly drops.(5)With the support of GIS platform,the dynamic prediction and analysis model of land subsidence was established by constructing the land subsidence influence factor database,combining the SARIMA model,MODFLOW model and machine learning model.Based on the model,the distribution of land subsidence in Heze City in 2025 was predicted and analyzed.The results showed that the subsidence analysis model proposed in this paper presents a good simulation effect in the prediction and prevention and control of land subsidence in Heze City.It can realize the overall prevention and control of land subsidence risk in Heze City from the aspects of disaster risk prediction,distribution characteristics and development trends.