Study On Spectral Signatures And Estimation Of Heavy Metals In Mine Reclamation Soils

Agricultural lands are given priority to most of Chinese mine reclamation soils. However, due to the particularity of reclamation technology and the complexity of reclamation environment, the heavy mentals carried by filling materials and the mining area environment factors may both lead to the heavy metal pollution of soil, and finally harm to human beings’ health through crops planted on contaminated soil. Monitoring of soil heavy mental pollution status by using conventional methods is not only time-consuming and laborious, but also almost impossible to be applied to large area dynamic monitoring. Therefore, exploring a method that can estimate the heavy mental concentration by using soil spectrum is significant to provide effective way to monitor heavy metal pollution of mine reclamed soil, which also can provide theoretic foundation and technical support to evaluate soil heavy metal pollution and regulate food safety in mining area.Studying the soil heavy metals in Liuxin reclamation area, Xuzhou, this paper explored the potential of hyperspectral remote sensing in quantitative inversion of soil heavy metals(Cd, Cr, Cu, Pb, Zn) concentration and the impact of extraction of spectral features in preprocessing. Analyzed and compared the effectiveness of multiple linear regression, generalized regression neural network and sequential minimal optimization- support vector machines methods for quantitative estimating of soil heavy metals concentration with hyperspectrum, and the results showed that the sequential minimal optimization- support vector machines inversion method was the best method. The main contents and results are as follows:(1) The heavy metal concentration and contamination risk levels of soils in mine reclaimed sites were higher than in control site. The heavy metals concentration of all three sites did not exceed the soil environmental quality standards, and relative to background values of soil, there were individual samples exceeded it of the three sites and the heavy metal concentration of soil in reclaimed sites were all higher than in normal site. The evaluation of pollution situation by Muller index and Nemerow index methods showed that the soil of reclaimed sites with slight to moderate pollution, and the soil of control site with Cu slight pollution. It means that the heavy mentals carried by filling materials may lead to heavy metal pollution of soil.(2) There was significant correlation between soil heavy metal concentration and reflectance spectrum of mine reclaimed soils. Spectral characteristics of mine reclamation soil were enhanced significant after data transformation. The results showed that: the characteristics extraction based on first derivative and continuum removal method can be a good characterization of heavy metals in soil spectrum characteristics. The significant correlation between elements Cd、Cr、Cu、Pb、Zn concentration in soil heavy metal and characteristic variables were no smaller than 0.78, 0.82, 0.7, 0.77 and 0.74, respectively.(3) Spectral estimation models of mine reclamed soil heavy metals were effective. The methods of multiple linear regression, generalized regression neural network and sequential minimal optimization- support vector machines methods were applied to establish the models to estimate heavy metals concentrations in mine reclaimed sites. It was used the R2 to evaluate the stability of model, and MSE and to compute the accuracy of model. The results showed that the stability and accuracy of model established by these three methods was good. The sequential minimal optimization-support vector machines performed best, the R2 of modeling and prediction model of Cd, Cr, Cu, Pb, Zn achieved 0.8628、0.8532、0.7988、0.7901 and 0.7653, respectively.(4) Applying the optimizing model to quantitative estimating the Cd concentration in wheats planted on mine reclaimed soil by using hyperspectral data. The result showed that R2 is 0.6683 and RMSE is 0.0489, which means this method had certain feasibility even though the effect was inferior to soil.