Quantitative Estimation Of Soil Heavy Metals In Farmland In The Upper Reaches Of Tuojiang River By Hyperspectral Inversion

In recent years,with the rapid development of industry and agriculture in the river basin of Tuojiang,the pollution of heavy metals in the agricultural soil in the basin is becoming more and more serious.When it comes to study the heavy metal content in the soil,the traditional indoor chemical analysis method,which requires a lot of time and effort,and has strong destructiveness,instantaneous and hysteresis,cannot reflect the regional situation quickly and dynamically,is generally used.However,hyperspectral quantitative remote sensing has the characteristics of multi-phase,large area and rapid and intuitive monitoring of soil heavy metal pollution,which is more suitable for the development needs of big data era.In this paper,soil samples were collected from agricultural land in the upper reaches of the Tuojiang River in Deyang City,using laboratory physical and chemical analysis and indoor spectral measurement methods to obtain heavy metals concentrations and spectral curve data of soil samples.By analyzing the characteristic value of heavy metals concentrations,the relationship between elements and the physical and chemical characteristics of the spectral curve,the feasibility of using spectral curve to retrieve heavy metals was explored.In order to extract characteristic bands and band combinations with strong ability to characterize heavy metals concentrations in soil samples,the correlation between the heavy metals concentrations and the spectral data after spectral transformation(First Derivative,Continuum Removal,Inverse-log)were analyzed in detail,as single variable regression,multiple linear regression and partial least squares regression,variable factors in three regression models,so as to establish each heavy metal in the study area.The accuracy of the model was evaluated by the samples of the modeling group and the prediction group,and the optimal estimation model of each heavy metal was selected through the vertical comparison.In addition,the Inverse Distance Weighted was used to generate the continuous distribution of heavy metals in the study area,so as to make an intuitive judgment on the prediction ability of the optimal model.The results of this study showed as follow:1.The source,migration and enrichment of Zn and Cd,Cd and Hg,As and Ni and other elements in the study area are related to each other,among which Cd and Hg in the soil samples of the study area were seriously over standard,and the enrichment degree was 4.89 and 4.04 times of the background value,respectively.2.After the transformation of first derivative,continuum removal and inverse-log,many bands can be significantly related to the heavy metal content.The band combinations obtained by performing difference and ratio operations on the two characteristic bands of Cd,Cu,and Pb elements had better stability and rich response information.The absolute values of the maximum correlation coefficients were 0.62,0.49 and 0.46,respectively.3.The single variable regression method was used to establish the optimal estimation model of As.The stability and prediction accuracy of this model were good,and the determination coefficient reached 0.77.The partial least square regression method was used to establish the best estimation model of Ni and Cu,and coefficients of determination were all 0.40.