Rapid geographical traceability and quality identification of black tea via integration of ambient mass spectrometry and machine learning.

Accurate geographical traceability and comprehensive adulteration assessment of black tea are essential for quality control and market regulation, but remain challenging due to subtle metabolic differences and complex adulteration practices. In this study, a unified analytical framework integrating ambient mass spectrometry with advanced data-driven algorithms was developed to rapidly achieve origin traceability, adulterant identification, and quantitative adulteration prediction. To rapidly acquire rich metabolomic profiles, a cell-disruption-assisted solvent direct extraction strategy was employed to allow effective characterization of chemical variability among black teas from diverse geographic origins. Moreover, clear biochemical evidence for regional discrimination were systematically screened and validated. To achieve accurate geographical classification, random forest (RF) model was employed to extract informative patterns from complex metabolomic datasets, yielding classification accuracies ranging from 95.00% to 100%. More importantly, 11 standardized adulteration systems were constructed to simulate realistic blending scenarios and generate representative datasets for machine learning-based modeling. By integrating an XGBoost classification model, reliable qualitative identification of adulterants was achieved, with accuracies ranging from 93.3% to 100%. Furthermore, quantitative prediction of adulteration levels was accomplished using RF regression model, which demonstrated high predictive accuracy and stability, yielding root mean square error (RMSE) values below 6.78% and R2 values exceeding 0.9606. Overall, this study presents the comprehensive analytical framework that integrates origin tracing, adulterant identification, and adulteration quantification, representing a significant advancement in black tea authentication and regulatory supervision.