Geochemical fingerprinting and machine learning for authenticating sparkling wine origins

Area of Science:

• Geochemistry
• Food Science
• Machine Learning

Background:

• The global wine market, especially high-value sparkling wines, suffers from significant counterfeiting.
• Existing authentication methods like supply chain tracking and geographical indications are inadequate against sophisticated wine fraud.
• Counterfeit wines pose substantial economic risks and potential health hazards to consumers.

Purpose of the Study:

• To develop and validate a scalable geochemical fingerprinting framework for authenticating the origin of sparkling wines.
• To assess the effectiveness of isotopic and elemental analyses combined with machine learning for wine authentication.
• To identify cost-effective analytical methods for widespread adoption in wine fraud detection.

Main Methods:

• Utilized a combination of isotopic analysis (strontium, ⁸⁷Sr/⁸⁶Sr) and elemental analysis (rubidium, Rb).
• Applied advanced machine learning algorithms for classification of wine origins.
• Analyzed 75 French sparkling wine samples from distinct geographical regions (Champagne and Burgundy).

Main Results:

• Achieved 100% classification accuracy in identifying wine origins using strontium isotopic ratios (⁸⁷Sr/⁸⁶Sr).
• Identified rubidium (Rb) concentration as a cost-effective marker, reducing analytical expenses by 75%.
• Maintained over 90% accuracy using the rubidium-based method, demonstrating its practical viability.

Conclusions:

• Geochemical fingerprinting, particularly using strontium isotopes, provides a highly accurate method for sparkling wine authentication.
• Rubidium analysis presents a viable, cost-effective alternative for combating wine fraud, making advanced authentication more accessible.
• The developed framework offers a scalable solution to protect the integrity of the wine market and consumer health.