Research On Prediction Of Soil Water Characteristic Curve Van-Genuechten Model Based On Soil Basic Physical And Chemical Parameters

The soil water characteristic curve is the relation curve of soil water suction(matrix potential)and the soil moisture,the relationship between energy and quantity of soil water.As an index of reflecting the basic soil hydraulic characteristics,soil water characteristic curve is important for research on soil moisture migration and retention,especiallyhave important guiding significance for agricultural management.But there are lots of problems for many people,especially for basic water conservancy worker to measurate soil water characteristic curve is time-consuming,laborious and some problems such as soil spatial variability,so this study is based on theory of soil transfer function,implementation to acquisition of basic physical and chemical parameters easily to predict to obtain the model parameters of soil water characteristic curve,therefore derive reasonable accurate soil water characteristic curve model parameter is complete the key technology of this study.This study is based series of indoor test of soil water characteristic curve on loess plateau area,the reasonable selection of high precision,applicable widely Van-Genuechten model as the fitting model.By use of RETC and MATLAB software to fit Van-Genuechten modelparameter? and n,and the basic physical and chemical parameters were mensurated at the same time.Finally,the 108 sets of data samples were constructed based on a test process that lasted more than one year with Van-Genuechten model parameter ? and n and the basic physical and chemical parameters.On the basis of analysising every basic physical and chemical parameters and Van-Genuechten model parameter ? and n,this research realize nonlinear prediction,BP prediction and SVM prediction by using the basic physical and chemical parameters as input variables,the model parameter ? and nas out variables.The main results and conclusions as follows:(1)The main basic physical and chemical factors affecting the soil water characteristic curve are soil texture,dry density,organic matter content and salt content,Through the single factor analysis,this study determined respectively the function relation between Van-Genuechten model parameter ? and n and basic physical and chemical parameters of the soil.Parameter ?:It's linear with texture,logarithmic with dry bulk density,exponential with organic matter and logarithmic with salinity.parameter n: It's linear with texture and dry bulk density,logarithmic with organic matter,logarithmic with salinity.(2)It'sfeasible topredict the Van-Genuechten model parameter ? and nby using conventional physical and chemical parameters and establish nonlinear prediction,BP prediction and SVM prediction.This research take soil texture,dry density,organic matter content and salt content as input variables,the Van-Genuechten model parameter ? and n as the output variable,which is based on uniform data sample,including 100 sets of modeling samples and 8 sets of validation samples.Nonlinear prediction: the average relative error of modeling samples of parameters ? and nrespectively are 10.23% and 7.20%,the average relative error of samples of validation respectively are 7.65% and 7.65%;BP neural network forecast model: the average relative error of modeling samples of parameters ? and n respectively are 1.52% and 0.67%,the average relative error of samples of validation respectively are 1.21% and 0.28%;Support vector machine(SVM)forecast model: the average relative error of modeling samples of parameters ? and n respectively are 8.32% and 7.48%,the average relative error of samples of validation respectively are 6.99% and 4.54%.This research establish the soil transfer function of SWCC by using the three forecast model based on they own characteristics and laws,and we achieved ideal result on the three models.(3)The nonlinearforecast model maybe the first choice model.By comparing the nonlinear,BP,and the support vector machine forecasting model,we found that the nonlinear prediction model is simple in the form,physical meaning is clear,but the precision is slightly inferior;BP forecast model accuracy is high,can accurately realize the Van-Genuechtenmodel parameter prediction,but model form is complex,and its application is demanding to use personnel;The support vector machine(SVM)model,as first introduced to predict soil water characteristic curve model parameter,its precision meets the requirement of parameter prediction,and has the ideal forecast effect,but the theory of support vector machine(SVM)model is heavy and complicated,and the depth is higher,so in the practical application it's likely to be inferior in nonlinear and BP prediction model.Therefore,this study recommended nonlinear and BP prediction model as the rational prediction model,the nonlinearforecast model maybe the best model,but at the same time we decided to use what kind of forecast model should accord to different circumstances.This study establish the soil transfer function between soil water characteristic curve Van-Genuechten model parameters ? and n and basic soil physical and chemical parameters,realizes themodel forecast of three kinds of nonlinear,BP,support vector machine(SVM),but about the selection of input variables,this study is based on the the most commonly used basic soil physical and chemical parameters,so its representative is not strong,there may be in the future research should be further discussed,so as to improve the representative and the feasibility of the study,improve the prediction precision.In addition,for support vector machine(SVM)model of the depth is not enough,should further strengthen in the future,so as to improve the prediction accuracy.