Multi-parameter optimization for Vis–NIR spectroscopic analysis of multiple indicators of soil heavy metal in the tideland reclamation area of the Pearl River Delta

ABSTRACT

Using visible and near-infrared (Vis–NIR) spectroscopy combined with novel chemometric methods, the rapid reagent-free simultaneous analysis model for Cu, Zn, Ni, and Cr contents in tideland reclamation soil in the Pearl River Delta was established. Based on Savitzky–Golay (SG) smoothing and partial least squares (PLS) regression, a multi-parameter optimization platform (SG-PLS) covering 264 modes was constructed to select appropriate spectral preprocessing mode for each indicator. The equidistant combination PLS (EC-PLS) method with three cycle parameters was adopted for the first large-scale screening of wavelength models. In addition, wavelength phase-out PLS (WSP-PLS) and repetition rate priority combination methods were used as the secondary optimization method. The well-executed competitive adaptive reweighted sampling PLS was also used for comparison. The validation samples that were not involved in modeling were used to validate the selected three group models. For the four indicators, the method of EC-PLS combined with WSP-PLS achieved the best validation effect. In validation, the root mean square error (SEP_V), relative root mean square error (RSEP_V), and correlation coefficients (R_P,V) of prediction were 2.46 mg kg⁻¹, 4.5%, and 0.959 for Cu; 51.37 mg kg⁻¹, 24.6%, and 0.900 for Zn; 3.59 mg kg⁻¹, 9.1%, and 0.739 for Ni; and 9.15 mg kg⁻¹, 8.4%, and 0.843 for Cr, respectively. Results indicated that low relative error and high prediction correlation confirmed the feasibility of using Vis–NIR spectroscopy to analyze soil heavy metal contents. The proposed multi-parameter, multi-stage integrated optimization algorithm could be applied to a wider field of spectral analysis.

KEYWORDS:

• Soil heavy metal
• Cu/Zn/Ni/Cr
• visible and near-infrared spectroscopy analysis
• spectral preprocessing optimization
• wavelength model optimization

Acknowledgments

This work was supported by the National Natural Science Foundation of China (No. 61078040), and the Science and Technology Planning Project of Guangdong Province (No.2014A020213016, No.2014A020212445).

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This research was funded by National Natural Science Foundation of China grant number 61078040 and Science and Technology Planning Project of Guangdong Province grant number 2014A020213016, 2014A020212445