Soil Heavy Metal Content Prediction Based On Hyperspectral In The Ugan-kuqa River Oasis

As the progress of industrial development and urbanization,the quantities and types of agricultural chemicals have been increasing,which has led to heavy metal pollution in soil,and the pollution in agricultural soils was especially serious.In this paper,Ugan-Kuqa River oasis was selected to be study area.The amounts of As,Hg,Pb,Cr,Zn and Cu were determined from 57 soil samples collected in Ugan-Kuqa River oasis.The spatial analysis,geo-accumulation index,Pollution load index and potential ecological risk index was adopted to evaluate the pollution characteristics,pollution degree and potential ecological risks of heavy metal,and the multivariate statistical analysis was adopted for the discussion on heavy metal.the stepwise regression method was used to select the characteristic bands.Based on geographic weighted regression(GWR)and ordinary least squares regression(OLS)construct the hyperspectral inversion model of soil heavy metals,the applicability and limitations of the model in the prediction of soil heavy metal hyperspectral were discussed.The results show that:(1)The contents of 6 kinds of heavy metals in soils,i.e.As?Hg?Pb?Cr?Zn and Cu were different,and the average contents were2.41,0,37.64,25.25,39.78 and 15.01mg/kg,respectively;soil heavy metal contents were not higher than national standard level,but Pb content value was higher than the background value of soil in Xinjiang.There is a certain degree of heavy metal pollution in the study area.(2)Through the analysis of geo-accumulation index?Pollution load index and potential ecological risk index,the results showed that the heavy metals for the geo-accumulation index value were sorted as As(-4.59)<Hg(-3.62)<Cu(-1.92)<Cr(-1.44)<Zn(0.59)<Pb(0.73),The PLI value of heavy metals in 6 kinds of soil is less than 0.035,which indicates that the study area is at the level of pollution free and the potential ecological risk index was 22.26,presenting the non-pollution level of soil heavy metal in research area.(3)By the correlation analysis,principal component analysis and cluster analysis,the sources of heavy metals could be classified 2 categories in the study area.The source of Cu,Zn,Cr and Pb might be from the combined influence of the soil parent material and human agricultural activities.As and Hg might be controlled by landform and hydrologic environment factors.(4)Although the ecological risk index and Pollution load index of study area was non-pollution level,there was difference on ecological risk index for different regions;among them,the values of ecological risk index were higher in the areas west-north(WN),north-west(NW)and north-east(NE),and the highest value appeared in the area north-east(NE),which should be on alert.In this paper,the proposed method of analysis can reflect characteristics and source of soil heavy metal quickly and objectively in the oasis,and it can provide scientific basis for evaluation and management of soil environment quality in the oasis.(5)through data transformation of hyperspectral data found that different transform enhancement effect of different spectral characteristics of soil heavy metals,of which two order differential(SD)and the cube root of first order differential(CRFD)transform compared with other transform of the most prominent performance for the maximum correlation coefficient of the two transform and heavy metals are greater than the maximum correlation between the original reflectance and heavy metal content.(6)for GWR and OLS,Hg,Pb model As,Cr,Zn and Cu were compared with the coefficient of determination GWR(R~2)has been significantly improved,respectively0.277,0.593,0.413,0.393,0.498 and 0.293,OLS were 0.164,0.345,0.148,0.406,and 0.376 0.206,while the AIC information criterion and residual sum of squares(RSS)and a OLS model was significantly reduced,the sum of squared residuals respectively 88.49%,35.60%and OLS model 6.193%,26.13%and 28.64%;GWR model to determine the coefficients of As(R~2)increased by 0.113,the sum of squared residuals is reduced by 2.15%.(7)the prediction mean of GWR method and OLS method is basically the same after regression.From the spatial distribution results,the prediction results of GWR coincide with the measured values better than OLS models.It shows that the prediction accuracy of the GWR model is higher than that of the OLS model.From the validation results of the model,the GWR model has certain advantages in estimating soil heavy metal content,providing a reference for the quantitative inversion of soil heavy metals by hyperspectral data.