| In recent years, the pollution of heavy metals in Weihe River was more andmore serious with the rapid development of economy. Most of heavy metalswould be enriched in the minerals and organic matters of the sediment, after theywere removed into the water, they were finally accumulated in sediment throughthe river own precipitation and flocculation. The enriched heavy metals mightrelease again with the variation of external conditions, and it caused the secondpollution of river. At the same time, it became a problem to be solved quicklybecause of its difficult degradation and strong toxicity. The content of heavymetals in the sediment cannot reach to the minimum concentration, which couldbe detected by spectrum. Therefore, in this paper, the correlation of infraredspectra, heavy metals, organic matters and minerals were analyzed, to predict thecontent of organic matters and minerals. At the same time, the research focus onthe internal correlation of heavy metals, organic matters, and minerals, in orderto indirectly predict the content of heavy metals by infrared spectrum, which hadno feature of spectra.In Shaanxi section of Weihe River acquisition of the23sediment samples asthe research object. Through the near infrared optical-samples (VNIR) andinfrared (MIR) spectrum data pretreatment, and by the partial least squares(PLSR) method to predicted the contents of Ni, Cr, Cu, Zn, Pb, Cd and organicmatters. This paper discusses the effect of spectral prediction about the heavymetals and organic matters contents. At the same time to discuss the effect of usePLSR method to the scope of different spectral data to predict organic matterscontent, analyzes different range of band about the forecast results influence.Finally discusses the heavy metals, organic matters, minerals between their internal correlation, and use of organic matters, minerals contents predict theeffect of heavy metals is analyzed.The results of this paper are as follows:(1) The content of heavy metals in sediment did not reach to the minimumconcentration of the spectral detection. The values of R2in the PLSR models,which was based on the spectral and heavy metals are nearby0.1, and the valuesof RMSE are very high. The models are ineffective. Therefore, the use ofinfrared spectroscopy to predict the heavy metals content is not feasible.(2) In turn after baseline correction, smoothing, and the first-orderdifferential derivation transform of VNIR spectral data. The value of R2in thePLSR models, which was based on the wave band of1849-1940nm and organicmatters is0.96, and the value of RMSE is very low. The value of R2aftercross-validation can reach to0.90, and the value of RMSEP is much closed tothe value of RMSEC, The model is effective, and has a very good predictionresults.(3) In turn after baseline correction and smoothing transform of MIRspectral data. The value of R2in the PLSR models, which was based on the waveband of769.23～1218.03cm-1and organic matters is0.88, and the value ofRMSE is very low. The value of R2after cross-validation can reach to0.81, andthe value of RMSEP is much closed to the value of RMSEC, The model iseffective, also have a very good prediction results.(4) The model, which was based on the wavelength of473nm and theminerals, such as the content of Fe, is effective. The model, which was based onthe wavenumber of1438.86cm-1and the minerals, such as the content of Si, iseffective. That was to say, the minerals content had a good correlation withspectral. The content of Fe, didn’t have a correlation with heavy metals content.And the content of Si had a correlation with Pb and Ni, but has a bad predictionresults for heavy metals content. Therefore, to use of the content of mineralelements to predict the heavy metals content is not feasible.(5) The organic matters content had a correlation with the content of Cu、Crand Zn, and had a very good prediction results. Therefore, the model, whichbased on the infrared spectroscopy and the organic matters content can used toindirect prediction of heavy metals content. |
PDF Full Text Request