Researchon The Prediction Of Spatial And Temporal Variationof Pore Water Level Based On BP-STARMA Model

Spatial-temporal modeling is a proess to predict the properties of unobserved temporal position according to the given spatial-temporal data.For groundwater level prediction modeling,current models most predict the changing of groundwater level just only from the time or space only,rarely predicted in space and time corporatly.Therefore,the study of considering the factors of time and space and modeling space to meet the continuous-time nonlinear discrete temporal variation data have practical significance.In this paper,using the principle and method of the neural network,GIS,spatial and temporal series and statistics to make research on the prediction of groundwater level changes from the study area based on the analysis ofspatial and temporal changes in the nature of physical geography monitoring data.In this paer,the research contents and results are as follows:(1)Temporal and spatial analysis of groundwater levelBased on Yancheng coastal plain research area in the 3rd confined aquifer groundwater monitoring data in 2005 to 2014,respectively from the study area range and individual monitoring wells range analysis the change law of underground water level in time and space.The temporal and spatial variation characteristics of groundwater in this area are obtained.The feature is composed of two parts,which are the spatial and temporal trends of global certainty and the spatial variation of local randomness.(2)BP-STARMA model buildingBy the time and spatial variation characteristics of pore groundwater level analysis,it is found that the temporal and spatial variation of groundwater level in the pore process can be broken down to determine the spatial variability and a local random spatial variability of two parts of a global.Due to BP neural network has good non-linear fitting ability and self-learning ability,this paper will use it to extract deterministic pore groundwater spatiotemporal trends.The residual part of the spatial and temporal trend value which is separated from the monitoring data of the groundwater level is regarded as a stationary sequence,and STARMAhas good fitting and forecasting effect on the stationary time series,so it can be used to fit and predict the random part of the groundwater level.(3)Prediction of groundwater level based on BP-STARMA modelIn Yancheng coastal hydrogeological sub pore groundwater as an example,combined with the area pore confined groundwater occurrence,exploitation and utilization,using BP-STARMA coupling model to predict the process of dynamic changes of groundwater level,and compare the prediction with the results from the BP neural network model and the STARMA model respectively.The fitting and prediction accuracy of this three kind of models are analyzed from temporal and spatial.The resultindicates that the BP-STARMA model is suitable for the prediction of the spatial and temporal process of pore water level.