Dalal, Nidhi (2024) Authentication and Traceability of Mediterranean Anchovies with NIR spectroscopy, Multielement profile, Stable isotope ratios and Machine Learning. [Tesi di dottorato]

	Testo Dalal_Nidhi_36_COMPLETO.pdf Visibile a [TBR] Amministratori dell'archivio Download (6MB)
Anteprima	Testo Dalal_Nidhi_36_PARZIALE.pdf Download (3MB) | Anteprima

Abstract

Commercialization of seafood industry has led to better accessibility of seafood around the globe and is an important part of global food chain to ensure food requirements worldwide. It forms one of the most complex international food chains and this makes it particularly vulnerable to fraud. Owing to this complex global scenario, traceability of seafood becomes even more important to protect consumer’s rights and ensure safety in food systems. Origin mislabelling includes concealment of geographical origin of illegally harvested fish species whereas species substitution includes replacement of low-value species for a more expensive one for economic gain. Multielement and stable isotope analyses and NIR spectroscopy are reliable analytical techniques providing useful information and thus accurate chemometric-based traceability models. These three techniques were applied for fish authentication. Apart from authentication, we analyzed the fatty acid composition of anchovies from four regions in the Mediterranean Sea. Mediterranean anchovies are a good source of Polyunsaturated and monounsaturated fatty acids, with their concentrations changing with each region. Higher concentrations of omega-3 and omega-6 fatty acids were found in Balearic Sea. NIR spectroscopy when combined with Machine learning algorithms (polynomial-Support vector machine and Random Forest) was able to distinguish anchovies from Tyrrhenian, Adriatic, and Balearic Sea with >95% accuracy. Outlier removal and spectra pre-processing formed important steps in application of NIR. In the P-SVM model, regions of NIR spectra absorbed by fatty acids were most important for discrimination even though fatty acids form a small part of the anchovy’s nutritional profile. Multielement profile was used in conjunction with Linear discriminant analysis and polynomial-Support vector machine to differentiate between anchovies from Adriatic, Balearic, and Tyrrhenian Seas. LDA and P-SVM provided a classification accuracy of 78.10% and 98.93% respectively. Trace elements such as Cu, Zn, Mo, and Rb. Stable isotope data was combined with carbon, nitrogen, hydrogen, and sulphur concentrations to develop a traceability model for anchovies from Adriatic, Balearic, Ionian, and Tyrrhenian Sea. Balearic sea had the least depletion of d13C (-17.931‰) and Tyrrhenian Sea the highest (-20.636‰). As for d14N, Anchovies from Adriatic Sea were least enriched (5.636‰) while those from Balearic Sea most enriched (9.626‰). LDA successfully discriminated between anchovies from these regions with an accuracy of 95.25%.

Downloads

Actions (login required)

Modifica documento