Geographical origin authentication of beef from Canada, Australia and Mexico using visible-near infrared spectroscopy and chemometric approaches.

Enhancing meat import controls is essential to protect domestic production and ensure fair-trade. This study evaluated the potential of visible-near infrared spectroscopy (Vis-NIRS; 350-2500 nm) combined with chemometrics to classify beef ribeye steaks according to country of origin (CoO): Canada (CAD), Australia (AUS) and Mexico (MEX). In total, 120 ribeye steaks from CAD barley-fed (n = 30), CAD corn-fed (n = 30), MEX grain-fed (n = 30), and AUS grass-fed (n = 30) cattle were classified using partial least squares-discriminant analysis (PLS-DA) and linear/radial support vector machines (L/R-SVM). When modelling CAD barley, MEX and AUS samples, PLS-DA and L-R/SVM correctly classified 95.83-100% and 100% of samples in Train and Test, respectively, across Vis, NIR and Vis-NIR regions. For CAD-corn, MEX and AUS, classification performance ranged from 88.00 to 100% and 75.00-100% in Train and Test, respectively. When broadening applicability by combining CAD feeding systems (CAD mix; n = 15 barley+15 corn), all models correctly classified 91.67-100% (Train) and 83.33-100% (Test) of CAD mix, MEX and AUS samples, except for NIR R-SVM (66.67% for MEX in Test). Dimensionality reduction by variable importance in projection (VIP) from PLS-DA correctly classified 95.83-100% and 83.33-100% of samples in Train and Test, respectively, using models built with 5 wavelengths in the Vis region or 7 bands (6 Vis+1 NIR) in the Vis-NIR region. These wavelengths were mainly associated with meat colour pigments, moisture and fatty acids. These results demonstrate that Vis-NIRS is a reliable and rapid tool for beef CoO authentication. Vis-NIRS implementation could enhance supply chain transparency and support the digitalization of routine import control systems.