Advanced Chemometrics-Driven Screening Approach for Orange Juice Authentication Using Dual Handheld NIR Spectrometers

Citrus fruits, particularly oranges, are widely consumed globally due to their appealing sensory properties and nutritional value [1]. However, their popularity has made citrus juices, and orange juice in particular, a prime target for adulteration and fraud. Adulteration of orange juice typically involves dilution with water or pulp wash, addition of organic acids such as citric, tartaric, malic, and citrate acids, sugars, other additives, and even juice-to-juice adulteration. The sugar to acid ratio or Brix to acid ratio is commonly used to describe the taste or tartness of fruit juices. Higher Brix or Brix to acid ratios indicate a higher sugar content, resulting in a sweeter and less tart juice. In the juice industry, this ratio is often manipulated by adding pulp-wash to adulterate the juice with the aim of achieving lower tartness levels [2, 3]. Given the significance of this issue, this study investigated the feasibility of using two handheld NIR spectrometers as rapid screening techniques, in combination with class modelling (DDSIMCA and soft-PLS-DA) and discrimination strategies (Ensemble learning and hard-PLS-DA), for the first time, to authenticate orange juice samples and identify levels of Brix to citric acid ratio in pulp-wash as adulterants. Both NIR spectrometers coupled with DD-SIMCA demonstrated 100% sensitivity and specificity in calibration and prediction sets. Furthermore, ensemble learning approaches such as Gradient Boosting Tree (GBT) and Adaptive Boosting (Adaboost) coupled with the NIR Tellspec spectrometer were able to perfectly predict the levels of adulterants with a limit of detection (LOD) of 2% and 5% for Brix to citric acid ratio and pulp-wash, respectively. This outperformed hard-PLS-DA, which is the most commonly used technique in food control studies.