基于支持向量机的土壤水分入渗参数预测研究  

Research on Soil Infiltration Parameters Prediction Based on Support Vector Machine

作　　者：雷国庆[1] 樊贵盛[1] 

机构地区：[1]太原理工大学水利与科学工程学院,山西太原030024

出　　处：《节水灌溉》2015年第12期28-30,34,共4页Water Saving Irrigation

基　　金：国家自然科学基金项目(40671081);山西省科技攻关项目(20100311124)

摘　　要：为解决土壤水分入渗能力的空间变异性问题,以方山、河津、泽州等地土壤入渗试验为背景,选用两参数Kostiakov入渗模型,建立以土壤密度、体积含水量、黏粒和有机质含量等土壤理化参数为输入变量,土壤水分入渗参数为输出变量的土壤传递函数。通过对入渗参数k、α的土壤理化参数影响因子分析,表明土壤理化参数与土壤入渗参数间存在着相关关系。在此基础上,运用支持向量机理论,将入渗参数的非线性回归问题转化为一个二次凸规划问题,建立了土壤入渗参数k、α的预测模型,通过对预测样本的误差分析,表明基于支持向量机土壤水分入渗参数预测模型的预测效果良好,可实现土壤传递函数的有效建立。In order to solve the spatial variability of soil infiltration parameters, taking the soil infiltration experiments in the counties of Fangshan, Hejin and Zezhou as the background and choosing two-parameters Kostiakov infiltration model, the soil pedotransfer {unction is established, which takes some soil physical and chemical parameters, including soil bulk density, soil moisture content, day and soil organic matter content, as the input variables, and the soil infiltration parameters as output variables. The analysis of the impact factors of infiltration parameters k and a shows the correlation between soil physical and chemical parameters and soil infil tration parameters. Using the theory of support vector machine （SVM）, the infiltration parameters of nonlinear regression problem is transformed into a quadratic convex programming problem. The prediction model for infiltration parameters including k and a is es- tablished. The analysis of the error of the prediction data shows that the prediction results of soil infiltration parameters based on support vector machine （SVM） are good and the established soil pedotransfer functions is effective.

关 键 词：土壤传递函数 土壤理化参数 土壤入渗参数 支持向量机 

分 类 号：S152[农业科学—土壤学] TV93[农业科学—农业基础科学]