Authentication of apples from the Loess Plateau in China based on interannual element fingerprints and multidimensional modelling

Abstract

Apple is an important fruit, and fruit authentication is significant for quality and safety control. The Loess Plateau (LP) in China is an important apple-producing region. However, the geographic authentication of LP apples has not been well studied. In this study, we discriminated LP apples based on multielement analysis. We analysed the differences in 29 elements of 522 samples collected from LP and others in 2018–2020 and constructed discriminant models for LP apple authentication. Linear discriminant analysis, partial least square-discriminant analysis, back-propagation artificial neural networks, and random forest (RF) showed different rates in training and validation accuracy. RF showed better tolerance to the removal of the less-important elements in model optimization. The final RF was optimized on 11 elements, which obtained 95.30% training accuracy for the 2018–2019 samples and 97.29% validation accuracy for the 2020 samples. The multielement-based authentication of LP apples could aid further studies of geographical origins.

Keywords

• Apple
• Multielement
• Geographical verification
• Linear and nonlinear discriminant analysis
• Random forest