NIR-based classification of vegetable oils from Amazon rainforest and quantification of adulterants

Abstract

Amazonian vegetable oils are important non-timber forest products supporting local economies and industries, providing sustainable alternatives to logging. However, ensuring the authenticity and integrity of these oils against economic adulteration with cheaper oils necessitates the development of rapid, cost-effective, and environmentally responsible quality control methodologies. This research utilizes one-class classification models (SIMCA, DD-SIMCA, and OCPLS) based on NIR spectroscopy to distinguish ten Amazonian vegetable oils from samples adulterated with soybean, corn, and cottonseed oils, contributing to the quality assurance of these valuable resources. Additionally, Partial Least Squares (PLS) models were developed to quantify oil purity and the content of individual adulterants. DD-SIMCA demonstrated the highest accuracy in classifying oils within their respective target classes and rejecting non-target oil samples. The PLS models predicted the content of adulterant oils (expressed as %ww-1) — corn, soybean, and cotton oils — in samples containing one, two, or three adulterants, yielding RMSEP and R² values of less than 5.1 % and greater than 0.77, respectively. Purity PLS models achieved RMSEP and R² values of less than 4.0 % and greater than 0.95, respectively. The application of NIR-based chemometric models for the classification of Amazonian oils and the evaluation of adulterant content provides a novel methodology. Additionally, the NIR spectral profiles of the majority of the Amazonian oils examined in this study are presented here for the first time.