Discrimination of Species in Raw Milk by Raman Spectroscopy and Principal Component Analysis

Background

Milk quality and authenticity are fundamental for food safety and consumer confidence. Since milk composition varies among species, it affects nutritional, technological, and sensory properties. Conventional analytical techniques are reliable but time-consuming and unsuitable for rapid, non-destructive monitoring. Raman spectroscopy represents a promising alternative due to its rapidity, non-destructive nature, and minimal sample preparation. This study aimed to characterise raw bovine, ovine, and buffalo milk and explore their discrimination, including mixtures.

 Methods

Raw milk samples included 10 bovine, 10 ovine, and 16 buffalo samples collected from different farms during two sampling periods (April and July). Pure samples and bovine-buffalo mixtures containing 50%, 60%, and 70% bovine milk were analysed. Samples were frozen, thawed for 24 h at 3 °C, and equilibrated to room temperature. Raman spectra were acquired using a fibre-optic probe (210 mW laser power, 10 s integration time; total acquisition time: 210 s per sample). Principal Component Analysis (PCA) was applied to evaluate spectral differences.

 Results

Raman spectra showed characteristic bands associated with lipids, proteins, and lactose. Bovine milk exhibited stronger carbohydrate-related bands (887–1129 cm⁻¹), whereas ovine and buffalo milk showed higher intensity of the lipid-associated CH₂ bending band at 1440 cm⁻¹, consistent with their higher fat content. PCA revealed species-related clustering, with bovine milk showing the clearest separation, while buffalo and ovine milk exhibited partial overlap. PC1, PC2, and PC3 explained 45.26% of the total variance in April and 44.79% in July. No clear separation was observed between sampling periods or among mixed milk samples.

Conclusion

Raman spectroscopy combined with multivariate analysis shows potential for rapid, non-destructive milk authentication and traceability, with future applications in identifying undeclared species substitution and supporting food authenticity, quality control, and food safety. However, partial overlap between buffalo and ovine milk and the lack of clear discrimination among mixed milks represent limitations. Further studies using larger sample sets and validated classification approaches are needed to assess its applicability to milk adulteration detection.