A detection method of trace metal elements in black soil based on hyperspectral technology: Geological implications

Abstract

In the case of low content of heavy metals in soil, the hyperspectral characteristic response of heavy metals is very weak, so it is difficult to construct an accurate direct hyperspectral inversion model. In order to solve the above problems, according to the physical and chemical properties of soil chemical variables, the enrichment characteristics of heavy metals are transferred to the related major chemical elements, so that the weak information of heavy metals can be indirectly quantitatively inverted. In this paper, the black soil in Hailun was taken as the research object. Through principal component analysis and cluster analysis, it was confirmed that there was an obvious adsorption occurrence relationship between the major element iron oxide (Fe2O3) and trace heavy metals As, Zn, Cd. The best inversion model of iron oxide content in the study area was established by partial least square method (the determination coefficient is 0.704, the root mean square difference is 0.148, and the F-test is 12.732). Based on the occurrence relationship between iron oxide and As, Zn, CD, a nonlinear fitting model between the predicted value of iron oxide and the real value of heavy metals was constructed by neural network. The fitting results show that the fitting degree of As, Zn and Cd is As>Zn>Cd. The overall correlations are 0.796, 0.732, 0.530 respectively. The study results show that the indirect prediction model based on iron oxide content can better quantitatively predict As, Zn and Cd, which provides a new method for the quantitative analysis of trace heavy metal content. This model provides a basis for hyperspectral remote sensing technology to predict soil heavy metal content, enhances the feasibility of soil trace heavy metal inversion, and is helpful to refine the quality monitoring of natural resource. It is of great significance to deepen the comprehensive analysis and evaluation of geoscience system.

Keywords

• heavy metal
• soil
• occurrence relationship
• hyperspectral
• geochemistry
• nonlinear fitting model
• indirect prediction
• neural networks