基于有机质光谱特征的土壤重金属Pb估算模型研究

为实现土壤重金属含量的高光谱快速监测,以石家庄地表水源保护区土壤样本为研究对象,基于土壤有机质(SOM)敏感波段对应的多种光谱变换指标,采用多元线性逐步回归和偏最小二乘回归方法,构建土壤重金属Pb含量的间接反演模型。结果表明:研究区土壤样本Pb平均值为19.729mg/kg,SOM与Pb含量的R达到0.740,两者存在一定的吸附赋存关系。SOM在土壤反射率中的敏感波段为798nm,各种光谱变换指标中土壤反射率倒数对数一阶微分(ATFD)与SOM含量的相关性最大,R达到-0.754。基于建模样本和验证样本分析,多光谱变换指标偏最小二乘回归(M-PLSR)模型优于单光谱变换指标多元线性逐步回归(U-MLSR)模型和多光谱变换指标多元线性逐步回归(M-MLSR)模型,建模样本和验证样本R2分别为0.876和0.801。基于SOM光谱诊断特征建立M-PLSR模型来间接反演土壤重金属Pb含量是可行的,可为该地区土壤重金属Pb的快速监测提供参考。

Study on the estimation model of soil heavy metal Pb based on spectral characteristics of soil organic matter

To achieve the purpose of measuring heavy metal content rapidly by hyperspectral techniques,estimation models of Pb were developed adopting multiple linear stepwise regression and partial least squares regression methods on the basis of multiple spectral transformations corresponding to the sensitive bands of soil organic matter (SOM). The research was done in the protection zone of surface water source of Shijiazhuang. The results showed that the average content of Pb in soil samples was19.729 mg/kg ,and R between SOM and Pb reached 0.740 on account of the adsorption effect. The 798 nm was regarded as the sensitive band of SOM in the reflectance spectral. Among the various spectral transformations,absorbance transforms first derivative (ATFD) of soil reflecance and SOM content had the strongest correlation,and R reached -0.754. Based on modeling and validating sample analysis,the multivariate partial least squares regression ( M-PLSR ) model was superior to the univariate multiple linear stepwise regression ( U-MLSR ) model and the multivariate multiple linear stepwise regression ( M-MLSR ) model. The modeling and validating samples’ R2 of M-PLSR were 0.876 and 0.801,respectively. It was feasible for multiple spectral transformations estimation model to estimate the heavy metal Pb content indirectly. M-PLSR model could provide a reference for Pb content monitoring rapidly.