Study On Soil Transfer Function Of Gardner Model Parameters Of Soil Moisture Characteristic Curve

In order to solve this problem and simplify the way of obtaining the curve,this paper is based on the soil water characteristic curve and the Gardner model.The research object is the loess in the agricultural field in Shanxi.The correlation and fitting of the soil water characteristic curve are carried out,and the relevant soil basics are determined.The physical and chemical parameters and the parameters of the model were obtained and the correspondence between them was obtained.Based on the soil transport function theory,the nonlinear multivariate analysis and prediction model,BP neural prediction model,grid search and cross-validation-support vector machine prediction model were established.The main research contents are as follows:(1)Analyze the soil water retention curve and the influencing factors of the Gardner model parameters,and establish the order of relevance between them.Using the method of control variables,the influence of soil basic physical and chemical parameters on the change of soil water retention curve was analyzed in depth,the functional form between the influence factor and the Gardner model parameters was determined,and the concept of “correlation degree” was introduced innovatively,using grey correlation analysis.The method of quantifying the relationship between different influence factors and Gardner model parameters,and determining the order of association and relationship between them.(2)Based on the theory of soil transfer function,a soil transport function prediction model for Gardner model parameters was established.Based on the analysis of soil basic physical and chemical parameters,a nonlinear multivariate analysis and prediction model,BP network prediction model,grid search andcross-validation-support vector machine prediction model for the soil water characteristic curve Gardner model parameters were established.(3)Comparative analysis of predictive models and recommend the most appropriate predictive models.The structure,form,precision and other aspects of the three prediction models are compared and analyzed,and the differences and applicability between them are considered comprehensively.Finally,the most suitable prediction model is recommended.The main conclusions are as follows:(1)Through the mechanism analysis and correlation analysis,the main influencing factors for the Gardner model parameters are: soil texture,soil dry density,soil organic compounds,and soil total salt.The order of correlation between them is soil particle > soil dry density > soil total salt > soil organic compound > soil clay.The relationship between each impact factor and the Gardner model parameters was obtained by single factor analysis: the logarithmic relationship between soil texture and Gardner model parameters;the relationship between soil dry density and Gardner model parameters;soil organic compound content and Gardner model parameter logarithm Relationship;soil total salt content is exponentially related to Gardner model parameters.(2)The average relative error of the modeled samples of the nonlinear multivariate analysis model for Gardner model parameters is 10.1% and 11.8%,respectively,and the control is within 12%.The average relative error of the test samples is 9.5% and 7.7%,respectively.The control is within 10%;the average relative error of the BP network model modeling samples is 9.5% and 0.6%,respectively,and the control is within 10%.The average relative error of the test samples is 5.1% and 0.13%,respectively,and the control is within 6%.The average relative error of the modeled samples of the support vector machine model is 5.7%and 4.9%,respectively,and the control is within 6%.The average relative error of the test samples is 4.0% and 4.77%,respectively,and the control is within 5%.(3)Through the analysis of the theory and results of the model,it can beconcluded that the three models have their own unique meanings for guiding the actual work production,and the three models are analyzed in many aspects.Finally,the support vector machine prediction model is recommended as the most suitable.Prediction model of Gardner model parameters of soil water characteristic curve in loess area.Based on the theory of soil transport function,this paper establishes three prediction models for Gardner model parameters,which not only provides a simple method for obtaining soil moisture characteristic curves for actual farmland workers,but also expands the relevant theory of soil transport functions.However,in the research,the selection of the influencing factors is not comprehensive.Therefore,in the future work and study,this aspect should be further studied and explored,the data of soil types and basic physical and chemical parameters should be expanded,and the predicted model should be optimized.At the same time,an updated soil transport function prediction model is created to further improve the accuracy of the prediction.