Studies On Cu And Zn Content Of Agricultural Land Soil Based On The Hyper-spectrum Estimation Model

More and more heavy metal pollutants were released into the environment and unreasonable application of pesticides and fertilizers with the rapid development of industry,which could lead to the fertility of soil was weaken and soil was polluted,affecting the output and quality of agricultural products.Therefore,Analyzing and monitoring the agricultural land and soil pollution condition be of great significance to ensure the China’s food security and accelerate China’s agricultural modernization as well as promote the construction of ecological civilization.However,it is complicated for the operation process of monitoring technology of traditional chemical element in soil,while the near-surface hyperspectral remote sensing technology can build inversion model by spectrum and property of soil,which possess the advantage with saving time.It is a new developing technology to quickly monitor soil quality in recent years.In this paper,the heavy metals of agricultural land soil in Yanyuan County were taken as the research object,and the soil’s reflection spectrum was measured by SVR HR-1024i.On the basis of the study on the characteristics of heavy metal content and spectral characteristics of soil,linear and nonlinear methods,including multivariate linear stepwise regression analysis(MLSR),analysis of the BP neural network based on principal component analysis(PCA-BP)and multiple linear stepwise regression analysis(PCA-MLSR)and particle swarm algorithm to optimize the support vector regression(PSO-SVR),were utilized to build the inversion model for predicting soil heavy metal Cu and Zn content in soil.Finally,the coefficient of decision(R~2),root mean square error(RMSE)and RPD are used as comprehensive evaluation indicators to evaluate the hyperspectral inversion method’s accuracy of the soil heavy metal element,to abtain the inversion model with high inversion accuracy and excellent stability,and further verify the migration generalization property in different spatial scales inversion model.This research will provide technical support for the rapid and accurate acquisition of heavy metal content in soil,and provide a new ideas for the spectral monitoring of other soil heavy metal elements in soil.The main research results are listed as following:(1)The sensitive characteristic wavelength bands of Cu in soil for the study area are mainly distributed in 534-605nm,750-762nm,996-1107nm,1400-1430nm,1900-1931nm,2193-2205nm and 2350-2400nm.The sensitive characteristic wavelength bands of Zn are mainly concentrated in 474-524nm,646-713nm,890-1018nm,1207-1254nm,1811-1822nm,2165-2308nm,and 2421-2457nm.The sensitive wavelength bands of Cu and Zn in soil are similar,showing multiple sensitive bands with similar influence.(2)Hyper-spectrum inversion models were set up by the measured values of soil heavy metal Cu,Zn content and the corresponding reflectance spectra of the first order differential,continuum removal,standard normal variables transform and inverse logarithm,the results show that the first-order differential spectral model is most excellent in four spectral processing methods and heavy metal Cu and Zn content in soil for the study area optimal inversion model are first order differential principal components stepwise regression model.(3)Although the nonlinear mathematical methods of PCA-BP and PSO-SVR can improve the modeling accuracy,compared with the linear mathematical methods of SMLR and PCA-SMLR,the stability and applicability of the model is poorer,and the implementation is more complicated,computational efficiency is lower.In comparison,the MLSR is a more accurate and efficient method,which is suitable for inversion research of heavy metal content in various soils.(4)Although quantitative inversion of heavy metals based on the soil’s hyperspectral data were measured on the ground can establish a regression model with higher statistical accuracy,it is found that the prediction accuracy of the regression model is low for the sampling points in large scale regions.Furthermore,it is difficult to accurately invert the content of heavy metal Cu and Zn in the large-scale test area.These results imply that the process and method of soil heavy metal hyperspectral inversion can be used universally at different spatial scales,but there are some differences in the best inversion model obtained.When performing hyperspectral quantitative inversion of heavy metals in soil,in order to better improve the inversion accuracy of the model,the research area can be divided into multiple small areas according to the actual distribution of the research area before modeling,and quantitative analysis is established by partition model.