Study On Prediction Models Of Gardner Model Parameters Of Soil Moisture Characteristic Curve

Soil unsaturated zone also known as the vadose zone,which is a transition zone between the atmosphere and the underground aquifer for water conversion.It is also the main growth area of the root of crops.The soil water characteristic curve is one of the basic characteristic curves to study the water holding capacity of the unsaturated soil,which describes the complex nonlinear relationship between soil water quantity and energy.In-depth analysis of soil water characteristics curve can be used to provide reference for scientific irrigation of farmland and achieving efficient use of crop water,but also provide support for study the soil moisture evaporation,infiltration and solute transport process of various water movement parameters.Aim at the defects of the test method of SWCC,the difficult operation as well as the time-consuming essence.This paper that based on samples of the Gardner model parameters and the basic physical and chemical parameters of the Loess Plateau on the regional scale of the Loess Plateau and the soil transfer function theory,chose Gardner model as the study model and built non-liner prediction model,gray BP prediction model and SVM prediction model,the analysis of the soil water characteristic curve and the prediction of the Gardner model parameters were carried out.The research includes:(1)Sample of Gardner model parameters and physical and chemical paremeters was built.The soil water characteristics curve of different types of loess soils were measured by 1500 F Pressure Membrane,and the basic physical and chemical parameters such as soil texture,soil volume quality,soil organic matter and inorganic salt were also determined.The sample of Gardner model parameters and the soil physical and chemical parameters of loess soil were established.(2)Analysis of Main and Secondary Influencing Factors of Gardner Model Parameters.Based on the analysis of the influence mechanism of the basic physical and chemical parameters on the Gardner model parameters of the soil moisture and the relationship between the quantitative functions,the relationship between the influence factors and the model parameters was realized by the gray relational degree calculation.And the order of the size of the soil moisture characteristic curve Gardner model parameters of the main and secondary influencing factors was determined.(3)Prediction model on parameters of Gardner model parameters was built.Based on the same sample data,a multivariate nonlinear forecasting model,a gray BP neural network forecasting model and a support vector machine forecasting model were established respectively.The prediction accuracy and model structure of the three forecast models were compared by example analysis.The optimal soil transfer function prediction model was recommended to obtain the soil water characteristic curve model of the tillage layer in the Loess Plateau of Shanxi Province.The conclusions of this paper are as follows:(1)The main influencing factors on Gardner model parameters a and b are soil clay content,grain content,soil volume quality,organic matter content and inorganic salt content.Soil clay content and Gardner model parameters a,b is logarithmic relationship,soil water silt content and Gardner model parameters a,b is logarithmic,soil bulk density and Gardner model parameters a,b is linear relationship,soil organic quality and parameters of Gardner model a,b islogarithmic relationship,the amount of soil containing inorganic salt and Gardner model parameters a,b is logarithmic relationship.(2)The soil moisture characteristic curve of the soil moisture characteristic curve Gardner model parameters a and b is the same as the gray correlation of soil basic physical and chemical parameters.Soil mass fraction> Soil particle mass fraction> Soil sand content> Soil organic matter> Soil containing inorganic salt.Finally,the soil volume quality,soil clay mass fraction,soil particle mass fraction,soil organic matter and inorganic content of soil were selected as the multivariate nonlinear prediction model,gray BP neural network prediction model and support vector machine forecasting model input factors.(3)It is feasible to predict the Gardner model parameters a and b based on the basic soil physical and chemical parameters.The multivariate nonlinear forecasting model is simple in structure and simple in calculation.However,the preliminary workload is large,and the single factor equation and the multiple T value test need to be fitted.The average relative error of the parameters is less than 17% and the prediction accuracy is relatively low.The gray BP neural network forecasting model and the support vector machine forecasting model are complicated,and the forecasting method is completed by computer program.The gray BP neural network forecast model predicts the relative error is controlled within 10% and the prediction accuracy is high,but the BP neural network has the disadvantages of large dependence on the sample and easy to fall into the over-fitting.The prediction accuracy of the support vector machine forecast model and the gray BP neural network forecast model of the relative error is controlled within 10%.In view of the difficulty of selecting the model parameters,this paper explores the new method of selecting the model parameters,and the effect is remarkable.(4)The support vector machine forecast model is the recommended model.The three kinds of forecasting models can meet the basic requirements ofagricultural production activities and can provide reference for the prediction of Gardner model parameters of soil water characteristics curve model in the Loess Plateau.In terms of prediction accuracy,it is recommended to use the support vector machine forecasting model Prediction of Gardner Model Parameters for Soil Moisture Characteristic Curve Model.Questions and Improvements:(1)The sample that paper based on was finished in the student period,because of the combination of the time and soil,it is clear that the sample data are not sufficient and perfect in the process of establishing the forecast model,Therefore,it is necessary to pay attention to the expansion of the sample data and the sample Sex.(2)Paper only consider the main influential factors,but the secondary factors such as soil pH,soil temperature and heavy metal ions are not taken into account,and in the follow-up study,the main factors such as soil pH,soil temperature and heavy metal ions are not taken into account in the selection of input factors.It is necessary to determine the influencing factors comprehensively and rationally,to optimize the microstructure of the forecast model,and to further improve the prediction accuracy of the forecast model.