Merging vibrational spectroscopic data for wine classification according to the geographic origin.

Area of Science:

• Food Science
• Analytical Chemistry
• Spectroscopy

Background:

• Wine authenticity is crucial, with geographic origin being a key characteristic.
• Spectroscopic techniques are explored for wine origin classification.
• Official systems exist to catalogue wines by provenance.

Purpose of the Study:

• To compare the effectiveness of near-infrared (NIR), mid-infrared (MIR), and Raman spectroscopy for classifying wine geographic origin.
• To evaluate the performance of combined spectral datasets for origin classification.

Main Methods:

• Analysis of NIR, MIR, and Raman spectra from four Portuguese wine regions.
• Application of partial least squares discriminant analysis (PLS-DA) for classification.
• Development of multiblock partial least squares (MB-PLS) models to explore combined spectral data.

Main Results:

• MIR spectroscopy achieved the highest correct prediction rate (87.7%).
• NIR (60.4%) and Raman (60.8%) spectroscopy showed lower performance.
• Combining MIR and NIR spectra yielded 86.7% correct predictions, with MIR data showing higher contribution.

Conclusions:

• MIR spectroscopy is the most suitable technique for classifying Portuguese wine geographic origin among the methods tested.
• Joint use of spectral datasets did not significantly improve classification accuracy compared to MIR alone.
• MIR spectroscopy demonstrates strong potential for ensuring wine authenticity and traceability.