Studies On Cu And Pb Content Of Paddy Soil In Chengdu Plain Based On The Hyper-spectrum Estimation Model

This paper chose the paddy soil in Chengdu plain as the study object to analyze the soil samples and collect the hyper-spectrum data. Based on preprocessing, the authors extracted the original soil spectrum, first-order derivative reflectance spectrum> inverse-log spectrum and band depth, so as to establish the inverse model by multiple stepwise regression method and partial least square method, for fast and accurate access to the Chengdu Plain copper and lead content in paddy soils. The main results as follows.(1)The copper content of soil was 24.74-77.68mg/kg in the study area, the lead content was 6.48-159.56mg/kg, soil in the study area was the average of 1.63 and 1.64 times around the background value, showed significant accumulation characteristics. The relation between the copper and lead content in paddy soils with soil organic matter and total iron content were significant, among them, the correlation coefficient with copper content of organic matter and total iron content reached 0.457 and 0.413, with lead content of organic matter and total iron content reached 0.520 and 0.407.(2) After the original soil spectrum by first-order derivative and inverse-log, continuum removal transformation, enhanced spectral variables associated with copper, lead content in soils, the spectral characteristics of the band was increasingly outstanding, among them, the relativity between first-order derivative reflectance spectrum and soil copper, lead content was most relevant, Correlation coefficients were 0.697 and 0.680, continuum removal spectrum of variables is the second, correlation coefficients were 0.646 and 0.585.(3) In the study area,the spectral characteristics of the band located in 400 nm,605 nm,634 nm,750 nm,810 nm,906 nm,1005 nm, which copper content of heavy metals in soil spectral response characteristics of the band focused on 605 nm,757nm,906nm, lead content in the spectral response characteristic bands focused on 605 nm,906nm.(4) Spectral inversion model established by the soil heavy metals copper, lead content of the measured value and soil reflectance spectra, first-order derivative reflectance spectrum, inverse-log spectrum, continuum removal spectrum, the results show, first-order derivative spectrum variable model is optimal effect in four-variable model, among them, copper content of first-order derivative spectra variable model’s test precision reach 86.9%, the lead content of first-order derivative spectra variable model’s test precision reach 64.0%.(5) Build the model between the soil spectra of characteristic of copper, lead content by multivariate stepwise regression and partial least squares, the results show, Partial least squares modeling accuracy of copper and lead content in soil was better in the study area, among them, partial least squares models highest coefficient of determination R2 of copper and lead content were 0.716 and 0.760, however, partial least squares and multiple regression model’s test precision of copper and lead content is insignificant, among them, the average relative error of the two models differ minimum copper content was 0.2% and the average relative error of the lowest lead content of 8.8%.