Evaluation Of Soil Heavy Metal Pollution In Mining Areas And Research On Inversion Model Of Hyperspectral Remote Sensing

With the rapid development of industrial and mineral enterprises,the soil in mining areas is seriously polluted by heavy metals,which has severely restricted the living and production activities of the mining area people.Therefore,this paper took an iron mining area in Hebei Province as the research target region,the spatial distribution characteristics have been analyzed,and the pollution evaluation and heavy metal hyperspectral remote sensing model inversion research have been carried out by remote sensing and geoscience comprehensive analysis method,which through sample collecting,chemical detecting of heavy metal content,and soil spectrum collecting and processing.The results showed that:（1）The content of heavy metals in the soil of the mining area presents a strong variability（>30%）,and the heavy metal elements are significantly correlated,and there is an obvious trend of clustering.Affected by human and geographic DEM factors,their distribution presents a gradient characteristic in the region which means the similar pattern of formation.（2）Heavy metal pollution have been evaluated in the area by the index method and the improved weighted comprehensive ecological risk assessment method.According to the comprehensive potential ecological risk index RI,the percentage of sample points for each pollution level are very strong（53.03%）>extremely strong（40.91%）>mild（4.55%）>moderate（1.52%）>strong（0%）.The comprehensive ecological risk index I_ERstatistics found that all the samples in the study area are divided into non-alert（4.55%）and severe alarm（95.45%）two levels of early warning categories.The average value of the comprehensive pollution index based on the improved weighted comprehensive ecological risk assessment is 1.9313 which proved that the overall pollution is serious.The heavily and moderately polluted soil samples are located in the stepped dams accumulated in the tailings pond and the edge of the road at the end of the dam.（3）Through model inversion and accuracy evaluation,it is found that the regression prediction model based on the spectral transformation index,especially the second-order derivative index model of the spectrum,effectively improved the accuracy of the heavy metal model inversion.The regression modeling accuracy of Stepwise Multiple Linear Regression（SMLR）and Partial Least Squares Regression（PLSR）based on the second-order derivative of the reciprocal logarithm of the spectrum are as high as 0.795 and 0.802.The BP neural network inversion based on the denoising processing spectrum R showed that the prediction accuracy of the five elements of Fe,Cd,Co,Ni,and Hg are all greater than 0.5,and the inversion accuracy of the heavy metals BP neural network model are Hg>Ni>Co>Fe,Cu>Fe,Pb>Sn,Zn>Sn.Among the three inversion methods,the modeling accuracy of the optimal inversion model for Hg and Sn elements are greater than 0.75,and the prediction accuracy are greater than 0.5.