Comparison between NIR, MIR, concatenated NIR and MIR analysis and hierarchical PLS model. Application to virgin olive oil analysis

This work investigates the potential use of simultaneously near-infrared (NIR) and mid-infrared (MIR) spectroscopies for the quantitative analysis of fatty acids and triacylglycerols and for identifying the Registered Designation of Origin (RDO) of extra virgin olive oils. The two spectral ranges were used separately using PLS and PLS-DA regressions. To combine both information, concatenated matrix was used at the first time, multiblock method using H-PLS models were constructed at the second time. The models were compared in terms of prediction errors. The results obtained with MIR spectroscopy are better than the ones obtained with NIR spectroscopy. The H-PLS methodology seems to be very interesting for quantitative analysis with the use of additional information in the NIR range, which is not present in the MIR one. For RDO identification by discriminant analysis, the use of multiblock method was less efficient.