Research On Hydraulic Characteristics Of The Soil Based On The Different Construction Methods Of Pedo-Transfer Function In Fengqiu

The research of soil water and solute transport, groundwater level and water quality change has become the hot issue in the study of soil water, and the space change of soil water movement parameter in regional level is the premise and foundation of research on these problems. Due to the test of the parameters of soil water movement indoor or at field is big workload, long cycle and high cost, large area field determination is unrealistic. In this article, taking Fengqiu County in Henan province as the research area, based on the basic soil physicochemical data and indoor measurement of soil hydraulic parameters, using the multiple linear regression method and BP neural network method sets up the indirect calculation method of the soil hydraulic parameters-Pedo-transfer functions, the method is convenient and efficient, strong feasibility, and provides certain basic support for the research of the movement of region soil water and salt and soil pollution problems.The main results of the research are showed as follows:(1) Based on the soil moisture and the corresponding matrix potential data tested in the laboratory, using the least square method constructs the system of nonlinear equations, and using Picard iterative method solves the equations, and determines the parameters of Van Genuchten model-the soil water characteristic curve equation(2) Using multiple linear regression method establishes the point estimation model of saturated soil water content, field moisture capacity and wilting moisture in the study area, and the saturation water content prediction model is ideal; predicting accuracy of field capacity is enhanced by incorporating the wilting moisture as additional independent variables; the wilting moisture content model has relatively poor forecast effect. The results of the prediction about Van Genuchten model parameters using multiple linear regression method are showed, in the area of Fengqiu, clay has better prediction effect, silt loam has the second, and the prediction effect of sand is the worst.(3) At the time of constructing Pedo-transfer functions of Van Genuchten model parameters by BP neural network in Fengqiu, different predictor variables combination methods has different effect to predict soil water characteristic parameters. Compared with multiple linear regression, when predicting the soil moisture at specific matrix potential (saturation moisture content, field moisture capacity and wilting moisture), BP neural network has bad forecast effect and accuracy than the multivariate linear regression method; but BP neural network for soil water retention curve (Van Genuchten model parameters) forecast is better than the multiple linear regression method in the whole level. Therefore, introduced the BP neural network model into the construction and the solving process of Pedo-transfer functions is feasible in the study area.(4) Using space inversion function of the GIS technology, based on the information of basic physicochemical data in the study area, the Pedo-transfer functions has better effect in predicting soil hydraulic parameters. Make the difficulty of soil hydraulic parameters acquisition in the study of soil solute transport solved reasonably and effectively in area large scale, providing certain basis for the study of soil hydraulic parameters further simplify calculation and soil solute movement simulation study.