黄土耕作农田地表土壤(0～2cm)容重的多元非线性预报模型

Multivariate Nonlinear Prediction Model for Soil Bulk Density of Surface Soil(0～2 cm) in Cultivated Loess Fields

为给黄土高原区地表土壤容重参数的获取提供技术支撑,以山西省15个试验点的年度跟踪监测样本为依据,在用灰色关联理论对输入变量进行灰色排序,并对函数进行显著性检验和精度检验的基础上,采用多元非线性方法,建立了以地表(0~2 cm)土壤基本理化参数和累积接收水量为输入变量,以地表(0~2 cm)土壤容重为输出变量的多元土壤传递函数。研究表明:地表(0~2 cm)土壤容重与地表(0~2 cm)土壤基本理化参数和累积接收水量等输入变量间的关联度程度很好;以地表土壤黏粒含量、粉粒含量、累积接收水量、有机质含量、体积含水率和全盐量作为输入变量对地表土壤容重进行多元非线性预报是可行的。建模样本和检验样本相对误差的平均值分别为5.24%和5.88%,预测精度完全在可接受范围。

To provide technical support for access to the parameters of surface soil density of the Loess Plateau,with the annual monitoring samples from fifteen test sites in Shanxi Province,this paper sorts out the input variables according to grey correlation theory and the function is tested about significance and accuracy,and a multiple soil transfer function between surface（0～ 2 cm） soil bulk density and surface（0～2 cm） soil physicochemical parameters and accumulated receiving water is established by using multiple nonlinear methods. Research shows that： the correlation degree between the surface（0～ 2 cm） soil bulk density and surface（0～ 2 cm） soil physicochemical parameters and accumulated receiving water is great. It is feasible to select soil clay content,silt content,accumulated receiving water,organic matter content,volumetric water content and total salt content as input variables to predict surface soil density. The average relative errors of modeling samples and test samples are 5.24% and 5.88%,and the errors are acceptable.