Hyperspectral Remote Sensing Retrieval Of Heavy Metal Content In Soil Of Sanjiangyuan Region

Hyperspectral remote sensing has become a effective means of regionalenvironmental quality monitoring due to its wide range, timeliness and high efficiencyin predicting soil properities contents.As the sources regions of Yangtze River, YellowRiver and Lancang Jiang, the Sanjiang Yuan Regions play a vital role in China’secological environment protection and water resources protection. Changing ineco-environment and water resources of the San Jiang Yuan regions will have a directimpaction on environment situation and sustainable economy development formilldle-downstream regions of China. Among them, the soil environment’s alterationwill result in environment change, sustainable development of animal husbandry. Inthis paper, Yushu County of the Sanjiang Yuan Regions as typical study area, theresearch data source are soil heavy metals content(As, Sb, Bi, Cu, Pb, Zn, Cr, Cd, Hg)which mearsured in laboratory and soil spectral reflectance data which collected withASD Field Spec4in laboratory. The spectral estimation model of soil heavy metalcontent built with partial least squares regression method and BP neural networkmethod, was used for mapping soil heavy metal content with Hyperion Image, and thespatial distribution of As、Sb、Bi、Pb、Zn、Cr、Hg content was aquaried. The maincontents and conclusions are as follows:(1)The results show that the soil heavy metal As was well correlated with the soilMn、Al, while the soil heavy metal Sb was well correlated with the soil organic matter、Fe、Al、Si, while the soil Bi、Pb、Cr was well correlated with the soil Fe、Mn、Al,while the soil Cu was correlated with the soil Si and Mg, while Zn and Cd wascorrelated with the soil organic matter、Fe、Mn and Si, while the soil Hg wascorrelated with the soil organic matter and Al. The adsorption effect of soil heavymetals by the soil organic matter、iron and manganese oxides、clay minerals indifferent degree respectively produced the agglomeration of soil heavy metals.(2) The modeling accuracy and test accuracy of spectral estimation model on soilheavy metal content(As、Sb、Bi、Cu、Pb、Zn、Cd and Hg) built with partial leastsquares regression method and BP neural network method respectively was very close,and the overall accuracy was similar. The modeling decision coefficients of estimationand test on As、Sb、Bi、Pb、Zn、Cd and Hg were all over0.5and the RPD of testsamples were all reached1.4, which had good modeling precision and estimation performance.(3) Compared with the average value of Inversion results and the measuredresults, As, Sb and Pb elements on the low side; Bi, Cd, Hg elements are very close,and Zn element is slightly tall. Hyperion image inversion precision has reached anacceptable range, and it can basically reflect the spatial distribution characteristics ofsoil heavy metal content. As、Sb、Bi、Pb、Zn and Hg content estimated with Hyperionimage was higer near214national highway and309provincial highway, it’s contentmay be influenced with human activities.