Study On Missing Data Imputation Mixed Model Of Groundwater Monitoring Water Level

Water is the source of life and the basis for the survival and development of plants and animals.Groundwater,as an important water supply resource,has been plagued by problems such as over-exploitation and pollution,resulting in frequent occurrence of a series of environmental geological problems,such as falling groundwater level,land subsidence and salinity of fresh water.Long-term monitoring of groundwater level is an important basis for understanding the status quo of groundwater development and utilization and planning for sustainable development of water resources.As a typical spatio-temporal sequence data,groundwater monitoring water level data is characterized by nonlinearity,instability and strong spatio-temporal correlation,which is the basis of groundwater dynamic analysis,groundwater level simulation and hydrogeological 3d modeling.However,due to the influence of human or natural factors,long-term monitoring data of groundwater level are often missing to varying degrees.However,the existing traditional restoration methods of missing groundwater level data are poor in efficiency and accuracy,while the methods based on spatiotemporal geostatistics and machine learning only focus on time or space factors when repairing the spatiotemporal sequence data,and lack the consideration of spatiotemporal correlation.Therefore,the design of a highly applicable and high-precision model for the spatial and temporal data loss and repair of groundwater table has become an urgent problem to be solved in the analysis and research of groundwater resources and the protection of hydrogeological resources.Under the current academic background,by studying the spatial and temporal distribution characteristics of groundwater level,the establishment of the missing value restoration model with both spatial and temporal elements has become the direction to solve the above problems.Based on the above requirements,this paper studies the mixed repair model of missing data that takes into account space-time.The main research contents and obtained results are as follows:(1)characteristic analysis of groundwater level monitoring value.By analyzing the missing status,discretization,stationarity and spatio-temporal correlation of groundwater level monitoring values,the high missing rate,high spatio-temporal correlation and non-stationary characteristics of groundwater level monitoring values were obtained,which pointed out the direction for the selection and construction of spatio-temporal missing data restoration model and the verification of the accuracy of complete data set after restoration.(2)construction of a mixed model for the restoration of time-space loss of groundwater level monitoring values.Based on the characteristics of data and algorithm,the universal Kriging interpolation method(UK)and support vector regression(SVR)for spatial defect restoration were selected.According to the combination method of time and space elements,a hybrid model UK-SVR for separable spatial-temporal defect restoration was constructed.(3)comparative evaluation of the groundwater level spatio-temporal missing data repair model.The reliability and accuracy of UK-SVR,UK,SVR and nearest neighbor(KNN)models were compared and evaluated using the cross-validation algorithm,which proved that the mixed model UK-SVR was more reliable and more accurate than other single repair models.