| Soil is the relying basis for existence, development and civilization of human. It is not only one of the important natural resources on the earth, but also the lifeline of the entire ecosystem. The large-scale development of coal resource will cause serious heavy metal pollution, which can cause a series of environmental problems as soil degradation, water pollution and vegetation damage and even threaten the life safety of human. In turn, this will also restrict the development of the social economy. Therefore, it is an urgent task for us to have a rapid, real-time monitoring and early warning for the heavy metal pollution in current environment protection of the coal mining areas. Because the traditional soil collecting in field and chemical analysis in the laboratory are time consuming and energy consuming, so it is not suitable for the large scale and rapid monitoring on the heavy metal pollution in soil. Therefore, the development of remote sensing especially hyperspectral remote sensing technology provides a new method for the realization of large-scale, real-time monitoring and early warning for the, heavy metal pollution in soil.For the environmental pollution problems aroused by the development and utilization of coal resources, this study takes Wucaiwan opencast mining area in the Eastern Junggar Coalfield in Xinjiang province as the target area, taking the soil environment as the breakthrough point, reveal the rule of heavy metal(Hg, As, Cu, Ni, Cr, Pb and Zn) pollution in mining area soil and identification the sources of heavy metal pollution, discuss the feasibility of using hyperspectral remote sensing technology to quantitative estimates of heavy metal content in mining area soil, and try to simulate spaceborne multispectral sensors to predict soil heavy metal content. This research mainly draws the following conclusions:(1) The analysis of the spatial distribution’s characteristics of heavy metal elements in soil shows that:there is significant difference in the elements of the spatial autocorrelation, but in general the heavy metal content in the south that near the northern desert is lower than the Kalamali Mountain in the north, the heavy metal content in the central(mining area and surrounding) is high; the distribution of the heavy metal in mining area changes greatly with strong randomness and strongly influenced by human’s activities.(2) The evaluation of heavy metal pollution shows that there is varying degrees of pollution exists in78%Hg sampling points. Among which,10%was polluted extremely serious where the pollution on the Hg of topsoil is the most serious; All sampling points of As sufferred mild, moderate pollution; About10%sampling points of Pb and Cr was slightly polluted. Cu and Ni is not polluted. The sources of heavy metal pollution mainly are:dust, sewage, coal dump, gangue and ash etc.(3) There is distinct difference between the characteristic of Visible and Near Infrared(VNIR,350-2500nm) reflectance spectra and Thermal Infrared(TIR,8～13μm) emissivity spectral features in the soil of mining area. They are mainly influenced by moisture, clay mineral and the soil texture etc. Because the organic matter, iron oxide and the clay mineral can easily accumulate heavy metals, which can explain the theory of using the remote sensing technology to retrieve the heavy metal content in soil. The reflectance spectral of VNIR and emissivity spectral of TIR has low correlation to the heavy metal content in soil, through different forms of spectra after transformation, the correlation was enhanced significantly.(4) Through the analysis of Stepwise Multiple Linear Regression(SMLR), the modeling precision between the original soil spectrum as well as its different transformed forms and seven heavy metal’s content are not high, but when we used the VNIR-TIR spectral combination with the best transformation form, the modeling effect for two methods of SMLR and Partial Least Squares Regression(PLSR)all improved slightly. From the determination coefficient(R2) and the Root Mean Square Error(RMSE) of prediction we can see that:For Ni, the modeling effect is better by using SMLR method, while the PLSR method is relatively more suitable for six elements as Hg, As, Cu, C, Pb and Zn. It is obvious that we can improve the prediction precision on the heavy metal content of soil by transforming the original spectrum processing, selecting best transformation forms for spectral combination of VNIR and TIR and trying different modeling method.(5) To classify the pollution degrees of Hg with the Support Vector Machine(SVM) and have a classified study with the VNIR-TIR spectral combination. The result of the study shows that:the classification accuracy is100%for the polluted soil, is90.91%for the non-polluted soil. It can totally recognize the mild and moderate polluted soil, while the seriously polluted soil may be slighted, and non-polluted soil may be mistaken for polluted soil. Therefore, it is feasible to classify the pollution degrees of heavy metal in soil with the hyperspectral pollution of VNIR and TIR.(6) The effect for predicting the heavy metal content in the soil through the band simulation of satellite-borne multispectral sensors of OLI-TIRS and ASTER is not ideal. Besides, there are many factors that may influence the effect in the practical application and it is hard to realize the quantitative estimation of the heavy metal content in soil with this method, so we must find some other remote sensing data with higher spectral resolution.The study discusses the hyperspectral remote sensing monitoring technology that can be used to predict the heavy metal content in soil, it proposes the technical methods for the large-scale, low-cost and real-time monitoring and early warning of heavy metal pollution in soil as well as the countermeasures for the environmental management and protection in mining area, Besides, it will promotes the development and application of hyperspectral remote sensing technology. |
PDF Full Text Request