Predicting The Abundance Of Soil Copper Of The Tong Kuang Yu Tailings In Shanxi Province Using Reflectance Spectroscopy And Remote Sensing Imagery

With the increasing human demand for mineral resources,large-scale mining activities,resulting in large-scale accumulation of waste and slag to form the tailings,which not only damaged land resources,but also caused the deterioration of the soil ecological environment around the tailings in mining areas,resulting in increasing soil heavy metal pollution.In addition,traditional methods of soil heavy metal monitoring and assessment have serious shortcomings in terms of scale and timeliness,while hyperspectral remote sensing technology offers a new methodological approach to soil heavy metal content inversion with its high timeliness,wide field of view,wide measurement information and spectral properties.In this paper,68 soil samples were collected from the field,the reflectance spectra of soil samples from visible to short-wave infrared interval were determined using an indoor spectrometer,and the concentration of heavy metal copper elements in the soil was determined in the laboratory.The sensitive bands of spectral response were selected according to the characteristics of reflection spectrum,and the copper element content inversion model was established by piecewise partial least squares regression(P-PLSR),and the accuracy of the model was verified by soil samples.The model was then directly applied to the high-score5 hyperspectral image data and Sentinel-2 multispectral image data,respectively,to map the copper content of the study area and explore the feasibility of using remote sensing images to map the copper content of the soil.The main research results obtained in this paper are as follows:1.Piecewise partial least squares regression is proposed to address the problem that large variations in Cu content and limited soil samples result in ineffective predictions of Cu content using basic partial least squares modelling.The estimation accuracy of R~2for the predicted Cu content of the measured spectral model by piecewise partial least squares regression for resampling to GF5 data reached 0.93(RPD=3.72);the estimation accuracy of R~2for the predicted Cu content of the measured spectral model for resampling to Sentinel-2data reached 0.92(RPD=3.34).The method was shown to be effective in improving the wide range of variation in the copper content of the samples,improving the prediction accuracy of the model,resulting in better stability,better fitting and better prediction accuracy of the final model.2.The model established by the piecewise partial least squares regression is not only effective for the measured spectra,but also can be effectively applied to the GF5hyperspectral remote sensing image and Sentinel-2 multispectral remote sensing image,and the model was applied to the GF5 image data inversion,and the estimation accuracy of R~2for the predicted Cu content reached 0.80(RPD=2.07);the estimation accuracy of R~2for the predicted Cu content reached 0.77(RPD=2.00)when applied to the Sentinel-2 image data inversion,thus obtaining a more accurate spatial distribution of the Cu content in the region and providing a new and fast solution for large-scale soil environmental monitoring.3.Through the comparative analysis and evaluation of the spatial distribution results of soil Cu content in the study area of the GF5 and Sentinel-2 remote sensing image data,we obtained the GF5 data with its high spectral resolution(4nm,7nm),high signal-to-noise ratio(SNR 200:1,100:1),large width(60km swath width)and a large number of bands,and the final inversion results distribution map is more accurate than that of Sentinel-2.