Background: Half of all apple production worldwide comes from China. However, the geographic authentication of Chinese apples has not been well studied. We highlight the multi-element-based geographical discrimination of apples from the southwest cold highlands (SCH) of China. 565 samples from the SCH (138) and others (427) were obtained, and the content of fifteen elements were applied to construct models for discrimination.
Results: The SCH apples from 2017 to 2019 had higher concentrations of Mn, Zn, Cr, Cd, Se, Pb, and Fe, but lower concentrations of Na, B, Ni, and P. With sufficient training, linear discriminant analysis (LDA) discriminated the SCH, and the testing accuracy averaged 92.5% and 92.2%. Nonlinear discrimination models were more suitable than the linear models. Optimized random forest analysis was the model with the best fit, and with averaged training and testing it obtained a level of accuracy of 98.2% and 98.5%.
Conclusion: The multielement-based discrimination of SCH apples could aid further studies of geographical origins. © 2021 Society of Chemical Industry.
Keywords: Malus domestica; geographical verification; linear and nonlinear discriminant analysis; mineral elements; random forest analysis.
© 2021 Society of Chemical Industry.