Heavy Metal Pollution Characteristics And Hyperspectral Inversion Of Wheat-soil At Jointing Stage

With the transformation of agricultural planting structure and the rapid development of industry,food security has been greatly threatened by environmental pollution,especially the pollution of heavy metals in soils,in the economically developed coastal areas of eastern China.When heavy metals enter the soil they are difficult to transform and decompose,due to their special physical and chemical properties.After a long time of accumulation,they become more toxic and harmful for a longer time.Heavy metals are transformed and accumulated in food crops after they are absorbed.Heavy metals may endanger human health when human ingestion of food with excessive heavy metals.Under this background,much attention has been paid to the study of heavy metal pollution in soils of grain-growing areas.This study takes Longkou city,a typical coastal industrial city in eastern China,as the study area;and takes the soil-wheat system in the wheat planting area of the northern plain as the study object,systematically collects and analyzes the contents of heavy metals and Fe,Mg and organic matter(SOM)in soils and wheat leaves.Firstly,the spatial distribution of heavy metals in soils is analyzed by geostatistical methods.Soil heavy metals pollution is evaluated by pollution index method and model index method.The enrichment and transfer characteristics of heavy metals in soil-wheat system are analyzed.Sources of heavy metals in soils are analyzed by correlation analysis,self-organization mapping analysis and principal component analysis.Based on the spatial distribution of heavy metals in soils and the characteristics of heavy metal enrichment and transfer in soil-wheat system,wheat sample points are further determined to improve the model accuracy of inversion of heavy metals in wheat.The measured hyperspectral remote sensing data of soils and wheat leaves were collected,and the first derivative reflectance(FDR),second derivative reflectance(SDR),multiple scattering correction(MSC),standard normal variable(SNV),logarithm of the inverse of the spectrum were extracted(Log(1/R)).Then construct partial least squares regression(PLSR)models of heavy metals in soils and wheat leaves.On this basis,the statistical rule between visible-near infrared hyperspectral and soil heavy metal content is analyzed by combining the variable importance projection(VIP)of PLSR model with Spearman correlation analysis,partial correlation analysis and PCA bispectral analysis,which further illustrates the feasibility of soil heavy metals hyperspectral remote sensing inversion.Finally,we used panel data model as reference and improving,introduced the major control factors of heavy metals in soils,such as Fe and SOM;a multivariate linear equation system for inversion of heavy metals in soil-wheat system composed of soil panel hyperspectral model and wheat leaf panel hyperspectral model was established,and we explore a joint inversion model for inversion of heavy metals content in soils and wheat leaves to improve the efficiency of heavy metals inversion.The main contents and results are as follows:(1)The spatial distribution of heavy metals in the six soils in the study area is different.The contents of Cu and Zn are very high in the northeast and west coastal areas,As is high in the Northwest Mining area,Cd and Zn are rich in the southwest,and Cd and Pb are high in the whole Donglai street and at the junction of Xinjia street and Xufu town.The content of the six elements in the soil exceeded the natural background value of the corresponding soil elements in the eastern Shandong Province.The contamination degree of Cu,Cd and Pb is the highest,followed by Zn and As belonging to moderate contamination degree,while Cr belongs to mild contamination.30 of 148 soils sample points reached moderate pollution level,48 of them were light pollution,58 of them were alert,and only 12 of them were safe.Therefore,the inversion of soil heavy metals in this area has a certain basis.(2)The results of correlation analysis between the contents of heavy metals in soil show that there is a strong correlation among Cr,Cd and As,and the correlation coefficient with the parent elements Fe and Mg is strong.These three heavy metals may come from natural factors such as soil parent material.However,there is a strong correlation among Cu,Pb and Zn,but they are almost unrelated to Fe and Mg and may be caused by human factors.The self-organization map shows that Cu,Zn,and Pb may all be anthropogenic sources from different types of human activities.Principal Component Analysis shows that soil heavy metal Cd is a mixture of natural conditions and agricultural production.(3)The hyperspectral inversion model built by PLSR method can invert the content of heavy metals in soil,and after the transformation of the soil hyperspectral,the model accuracy can be improved.The FDR spectral index has the best effect with the model of heavy metals such as Cu,Cd,Pb and Zn in soils,and the SDR spectral index has the best effect with the model of As and Cr in soils.The hyperspectral model built by PLSR method can invert the heavy metals contents in wheat leaves,and the model accuracy can be improved after the soil hyperspectral transformation.The FDR spectral index has the best effect on the model of As and P in wheat leaves,and the SDR spectral index has the best effect on the model of Cd and Zn in Wheat leaves.Trilateral parameters had the best effect on the model of heavy metal Cr and Cu in wheat leaves.(4)Soil heavy metals contents can be estimated from hyperspectral data because Fe and SOM play a major role in controlling soil heavy metals.Cd is closely related to SOM,As and Cr are closely related to soil Fe,and Pb,Cu and Zn are closely related to SOM and Fe concentrations.Therefore,the application of both Fe and SOM as dependent variables can improve the results of hyperspectral inversion.(5)The multivariate linear model of soil heavy metals multispectral index was built with soil spectral index as independent variable and soil heavy metal content as dependent variable.With the spectral index of wheat leaves as the independent variables and the heavy metals contents of wheat leaves as the dependent variables,a multivariate linear model for the multispectral index of heavy metals in wheat leaves was constructed.The soil model and the wheat leaf model were combined to form a multivariate linear equation set for inversion of heavy metals in the soil-wheat system,and the heavy metals contents in soils and wheat leaves in the system was inverted simultaneously.Compared with PLSR model,this model not only improves the efficiency of system modeling,but also improves the accuracy of remote sensing inversion of heavy metals in soils and wheat leaves.