Research On The Prediction Of GIS Based Ground Subsidence

Ground subsidence is one of the major geological disasters of Tianjin. This complex environmental geological problem seriously affects the construction of urban infrastructure facilities, so that restricts Tianjin's economic development. It has been widely confirmed that the major course of ground subsidence is the break of geohydrological mechanical equilibrium caused by excessive exploitation of groundwater. Therefore, it is important to establish a set of method to predict the ground subsidence in Tianjin. Thus it can offer scientific basis for effective control of ground subsidence.Firstly, Tanggu District is selected as a model regional in this research, which has complete historical data. Based on optimized ground monitoring points, related analysis of gray system is employed to calculated each aquiferous group and the coefficient of ground subsidence. Kendall examination method is employed to analyze the coefficient of each aquiferous group.Secondly, with input variables of average underground water levels and optimized values of ground subsidence monitoring points to be output variables, a BP neural network model for prediction of ground subsidence is established. On this basis, regularization and early stop methods are used to improve the generalization of this model. Besides, a RBF neural network model is established. With compares between these 3 models, a best model for prediction of ground water in this region is selected.Finally, how to use ArcGIS to design and established the data base and data base map is described in detail, as well as how to use spatial analysis to analyze groundwater contour and ground subsidence. Moreover, a method for seamless integration for Matlab and ArcGIS is developed, which is based on COM technology. It explores a new method for communication between ArcGIS and other software.