Inversion Of Heavy Metal Content In Soil Based On Measured Hyperspectral And Landsat8 Image

Soil is the most basic natural resources that human beings depend on for survival.Soil environment is closely related to biological survival and social and economic development.In recent years,due to the changes of human activities and natural environment,heavy metal pollution in soil is becoming more and more serious,which has become a major environmental problem.The rapid and efficient monitoring of heavy metal content in soil has important practical significance for the treatment of heavy metal pollution in soil and the realization of sustainable agricultural development.The traditional fixed-point chemical monitoring method is inefficient and it is difficult to invert the spatial distribution of heavy metal content in the region comprehensively and dynamically.Therefore,the monitoring and treatment of heavy metals in soil are restricted.The development of remote sensing technology makes it possible to monitor soil heavy metal pollution from point to surface,from static to dynamic,and from instantaneous to continuous.However,due to the low content of heavy metals in soil,the multi-spectral remote sensing technology is difficult to meet the monitoring requirements of heavy metals in soil.Combining the advantages of hyperspectral bands with the advantages of multi-spectral satellite images in large areas,the accurate monitoring of large areas of heavy metals in soil can be realized.Supported by the project of"Chengdu Multi-factor Urban Geological Survey"and the sub-project of",this study aims to explore a feasible method for predicting soil heavy metal content by combining the measured soil hyperspectral data and Landsat8 OLI image.In this study,the southeastern region of Tianfu new district,Chengdu,was taken as the research area.The measured hyperspectral spectrum was used to simulate the spectral reflectance of Landsat8 OLI,and the significantly correlated bands were selected as the characteristic factors by Pearson correlation analysis.The prediction models of As,Cd,Cr,Cu,Hg,Mn,Mo and Pb were established by principal component regression,multiple stepwise regression and partial least squares regression.After the precision comparison,the optimal inversion model of each heavy metal is determined.The mixed pixel decomposition of Landsat8OLI images was carried out to remove the non-soil on the images and replace the compensation as soil information.The non-soil pixel mask was processed.The optimal model was applied to the processed Landsat8 OLI image to realize the inversion of the heavy metal content of the soil in the study area.By comparing the inversion results with the measured interpolation results,the method can reflect the distribution of heavy metal content in the soil.The main conclusions are as follows:(1)The spectral response function of Landsat8 OLI can simulate the measured hyperspectral as the multispectral band of Landsat8 OLI.As,Cd,Cr,Cu,Hg,Mn,Mo,Pb were significantly correlated with simulated multispectral bands LOI3(530-590nm),LOI4(640-670nm),LOI5(850-880nm),LOI6(1570-1650nm),and LOI7(2110-2290nm).As,Cd,Cr,Cu,Hg and Mo were also significantly correlated in LOI1(430-450nm)and LOI2(450-510nm).The simulated multispectral reflectance can be used as a factor in the prediction model of heavy metals in soil.(2)After the end elements were extracted from Landsat8 OLI multi-spectral images,the abundance was calculated by the fully constrained least square method.It considers two constraints at the same time,and the calculated abundance of soil,water,vegetation,building and road is 1,and the abundance of each end is not negative,with clear physical meaning.On this basis,the mixed pixel decomposition is carried out based on the linear decomposition theory,and the image is reconstructed from the perspective of sub-pixel,which can effectively eliminate the non-soil information interference of water,vegetation,buildings and roads and enhance the soil information.(3)Among the three regression models,multiple stepwise regression and partial least squares regression are more accurate than principal component regression.After comparing the different model precision of various metals,the multiple stepwise regression models of heavy metals Mn and Pb were selected as the optimal models,while the partial least squares models of As,Cd,Cr,Cu,Hg and Mo were selected as the optimal models.There is little difference between the inversion and the measured content of the verified sample set of the optimal model,and the accuracy of the optimal model is reliable.The determination coefficient R~2,root mean square error RMSE,mean relative error MRE and the model sample set were verified to be consistent,indicating that the model sample set was representative and the model could be used for subsequent inversion.(4)The optimal model was applied to the mixed pixel decomposition of Landsat8 OLI images,and the spatial distribution of the inversion results was more detailed than the measured interpolation results.The statistical inversion results minus the difference ratio of the measured interpolation results show that the measured and inversion contents differ little in most regions,which indicates that it is feasible to combine the measured hyperspectral and Landsat8 multispectral images to invert soil heavy metals.