Multivariate Modeling-Enhanced Stable Isotopic Origin Traceability of Qinghai-Tibet Plateau Rape Honey

Qinghai-Tibet Plateau (QTP) rape honey, recognized as a Protected Geographical Indication (PGI) product in China, has faced significant challenges due to fraudulent mislabeling of its origins in the market. To ensure the authenticity of PGI honey products and uphold market integrity, it is crucial to develop a rapid, precise, and efficient geographical traceability technology.

This study investigated the stable isotope signatures of rape honey from the Qinghai-Tibet Plateau (QTP) and the southern region (SR) for identifying key geographical indicators for the origin traceability of rape honey products in QTP. The research compared isotopic differences and elucidated their formation mechanisms across bulk honey, saccharides, and proteins. Finaly, multivariate discrimination models were established for specifically identifying QTP-origin rape honey, with optimized parameters to improve discrimination accuracy.

A total of 208 honey samples were collected from QTP (n = 71) and five provinces in the southern region (SR, n = 137) of China. Stable isotope ratios (δ13C, δ15N, δ2H, and δ18O) of bulk honey, endogenous proteins, and saccharides (glucose, fructose, and sucrose) were measured. One-way analysis of variance (ANOVA) was employed to analyze regional differences among the variables. Partial least-squares discriminant analysis (PLS-DA) and linear discriminant analysis (LDA) models were constructed based on stable isotopic data to discriminate honey sample origins.

ANOVA indicated the geospatial differences (P < 0.05) in δ2H and δ18O of bulk honey, as well as all four ratios of honey protein, are significant between QTP and SR. LDA exhibited superior discrimination performance, with leave-one-out cross-validation accuracies of 87.3% for QTP and 89.1% for SR.

An integrated strategy combining stable isotope ratios analysis with multivariate modeling provides an accurate and effective verification method for geographical origin traceability of high-value honey from QTP. This approach provides a reliable tool to address the issue of fraudulent mislabeling of PGI rape honey.

Stable isotopic signatures of QTP rape honey were discussed. Bulk and component-specific isotopic ratios were informative geospatial indicators. Machine learning algorithms significantly enhanced honey origin discrimination. LDA accuracy for QTP honey samples reached up to 87.3%. This strategy was developed to combat origin mislabeling and ensure food integrity.