Comparison And Research Of Spatial Data Pretreatment And Interpolation

Tongling ,as an important mining cities in China, the exploitation of its deposits promotes the development of local economy, while the soil heavy metal pollution has been worsening increasingly. Because of the complexity of spatial variability of soil elements, traditional statistical methods can not meet the needs of analysis. Fortunatly, if appropriate method could be chosen for the corresponding element, through spatial interpolation theory which includes many of the interpolation methods, we can effectively simulate the spatial distribution of soil elements in computer. In that case,we could provide the reliable prooves for restoration of soil contamination, environmental remediation and agricultural production.Therefore, in this article, Taking the sample of Tongling soil heavy metals: As, Cd, Cr, Cu, Fe2O3, Hg, MgO, Mn, Mo, Ni, Pb and S, preprocessing the data (by three positive normalization methods: johnson transformation, Box-Cox transformation and ln transformation), making spatial interpolation (using nine kinds of interpolation methods which almost include all the commonly used interpolation methods), checking errors (through cross validation and site inspection), so as to explore those transformation method s'capabilities and their influences on the results of data interpolation, and on this basis, getting the merits of those interpolation methods, and at last in consideration of soil element's correlation analysis and distribution of Tongling deposit points, describing the spatial distribution of elements and pointing out the corresponding ore deposits to every heavy metal element. The conclusions are as follows:(1)The sampling data of soil elements:As, Cd, Cr, Cu, Fe2O3, Hg, MgO, Mn, Mo, Ni, Pb and S in Tongling mining area have strong variability and high skewness, which do not meet normal distribution. Especially for As,Cd,Cu,Hg,Mn,Mo,Pb and S,the spatial variability is so strong that ie indicates that these elements are severely affected by human factors, whcich are likely to be the mining of certain spatial scale and agricultural activities. To reduce erratic behavior in interpolation, the data transformation of pretreament is needed.(2)The comparision indicates that Johnson transformation has a strong ability to make data meet or nearly normal distribution. Ln transformation can not make the data observe normal distribution,and Box-Cox transformation can do that for As while Johsnon for Cd,Cr,Fe2O3,Hg,Mn,Ni and Pb ,seven elements totally. But to the elements which Johnson transformation can not convert to normal distribution, Box-Cox has a better performance. So in practice ,the two methods can be Complementary.(3)The pretreatment effect analysis of spatial interpolation of soil heavy metals in this region shows that for As, Cd, Cu, Hg, Mn, Mo, Pb, and S , which has strong spatial variability , normal transformed interpolation significantly reduced the overall error, but have little effect on the root mean square error, which means the error of estimate of the extreme value does not fall. For As, Cu and Cd elements, Johnson transformation behaves the best; for Hg, Mn and Mo elements, Box-Cox transformation best and for Pb and S ,Johnson and Box-Cox transformation are both effect . But for Cr, Fe2O3, Mg and Ni elements which have weaker variability, transformation can not reduce the interpolation error and there is the possibility of errors increases. So it's better to be no change.(4)The comparision and analysis of the spatial interpolation of the twelve elements in this region shows that for the interpolation methods, in its optimal circumstances, the arrangement of the size of its overall error basically meets: NN> TWL> NA> IDW, RBF, LP, MA> Kriging> MFI; and root mean square error: NN error the largest, TWL slightly second, and the rest of the method error of approximation.(5)In the optimal case,IDW, RBF, LP, Kriging, and MFI displayed a stable and accurate interpolation results; particularly the MFI method ,as a new interpolation method, has the lowest overall error, and its root mean square error is also very low,which could be considered as the best interpolation method for the region soil heavy metals (As, Cd, Cr, Cu, Fe2O3, Hg, Mg, Mn, Mo, Ni, Pb and S).(6)Through the correlation analysis of As, Cd, Cr, Cu, Fe2O3, Hg, Mg, Mn, Mo, Ni, Pb and S,I revealed those elements'intrinsic correlation and made space diagram of elements and the corresponding deposits Combined with optimal interpolation maps and mine point information,and at last analyzed the pollution causation。...