Response Analysis Of Vegetation Indices To Soil Heavy Metal Anomalies

Remote-sensing hyperspectral imaging spectroscopy can objectively record information hidden in the narrow band of the object and is cost-effective.Many researchers use hyperspectral data to analyze the response characteristics of vegetation indices to soil heavy metal anomalies.Therefore,this paper uses the Xuejiping-Chun Du mining area as a research area to analyze the response characteristics of vegetation indices to heavy metal anomalies.Select and calculate the vegetation indices that represent different physiological characteristics of the vegetation,and analyze the correlations between these indexes and soil Cu,Hg,Pb,and As contents.Then use partial least squares and stepwise regression to establish regression model,and use K-fold cross validation to select a more appropriate modeling method.According to the selected more suitable method,a statistical regression model for heavy metal content in the soil of Xuejiping-Chun Du mining area was established,and the anomalous spatial distribution of heavy metal in the mining area was obtained by threshold segmentation.The results were compared with the actual spatial distribution of heavy metal anomalies.The results are as follows:(1)In the Xuejiping-Harudu mining area,the correlation between Cu,Hg and the vegetation index is higher than the correlation level between As,Pb and the vegetation index.The vegetation indices that are sensitive to Cu,As,Pb,and Hg stress are not exactly the same.There are multiple vegetation indices that are sensitive to the same heavy metal,and a vegetation index that is sensitive to multiple heavy metals.(2)In the stepwise regression model of Cu and NDVI,PRI4 is an effective predictor of soil Cu content in the study area.In the stepwise regression model established between Hg and vegetation index,ARI2,SR and PSSRb are effective predictors of soil Hg content in the study area.PRI4 is an important and effective predictor of Pb in the stepwise regression model of Pb.In As' s stepwise regression model,WI is an effective predictor of As.(3)The method of stepwise regression is better than partial least squares in the fitting effect.The accuracy of the external prediction is also relatively high,but the prediction error is not stable.In each training,Cu,Hg,Pb regression model The prediction errors differ greatly.While the partial least square method of intra-fitting effect is not as good as the stepwise regression method,but its outward predictive ability is more stable than the stepwise regression method,and the prediction error in each cross-validation does not change much.However,whether it is using partial least square method or stepwise regression method,the built-in Cu,Hg regression model of the in-fitting effect,the accuracy of the external prediction and the prediction error are significantly better than the Pb,As regression model.(4)The prediction accuracy and error of the Cu,Hg regression model established by the stepwise regression method is better than the model established by the partial least squares method.The Pb and As regression models established by partial least squares have lower prediction error than the stepwise regression method,and the prediction accuracy and prediction error are more stable.Therefore,the Cu and Hg elements in the soil of Xuejiping-Harbin mining area are suitable to be modeled by stepwise regression method.The soil Pb and As elements in Xuejiping-Harbin mining area are more suitable to be modeled by partial least-squares method.Copper's regression model has the following expression: Y=106.735+PRI4*116.474+MCARI*(-0.512)+ BGI1*7.175+PRI3*492.105+NPI*2.692+ARI2*(-13.425).Hg's regression model has the following expression: Y=-46.783+SR4*(-287.373)+WI*450.942+SIPI*(-273.325)+CTR1*(-0.442).(5)The Cu anomaly in the study area extracted from the model is basically consistent with the distribution of the Cu anomaly in the actual soil in the study area.However,because Cu and other heavy metals often have a certain symbiotic relationship,the Cu anomaly extracted is also related to Pb.And other heavy metal anomalies also overlap in space,but the established Cu regression model can basically reflect the distribution of Cu anomalies in the mining area,and can provide a certain reference for the analysis of changes in soil heavy metal content in the study area.