Food authentication: Determination of the geographical origin of almonds (Prunus dulcis MILL.) via near-infrared spectroscopy

The aim of the present study was the prediction of the geographical origin of almonds (Prunus dulcis MILL.) via Fourier transform near-infrared (FT-NIR) spectroscopy. For this purpose, 250 almond samples from six different countries were analyzed. As the year of harvest has a major impact on the metabolome, three different crop years (2017-2019) were considered. In order to predict the geographical origin, a support vector machine (SVM) model was trained. The SVM achieved a mean classification accuracy of 80.3% (1.5%). In particular one of the economically relevant questions - the distinction between Mediterranean almonds and American almonds - can be answered with this model. Combining the Spanish and Italian almonds to one Mediterranean class the overall classification accuracy is increased to up to 88.2% +/- 1.0%. These results confirmed the suitability of NIR screening for the determination of the geographical origin of almonds and may pave the way for future analytical applications. With regard to potential future applications, the transferability of the developed NIR method to blanched almonds was discussed and evaluated: Even if the classification accuracy of unblanched almonds is higher than the prediction based on blanched almonds, the determination of the geographical origin still seems to be possible with this type of processed almonds.

Automated summary

This study aimed to predict the geographical origin of almonds using FT-NIR spectroscopy, achieving a mean classification accuracy of 80.3% with the SVM model. By combining Spanish and Italian almonds into a Mediterranean class, the overall classification accuracy increased to up to 88.2%, demonstrating the suitability of NIR screening for determining the geographical origin of almonds.