Short-Wave Near-Infrared Spectroscopy and Chemometrics to Detect Adulteration of Honey with Sucrose Syrup

Abstract

Near-infrared spectroscopy (NIRS) coupled with multivariate statistical analysis was assessed as a rapid method of detecting honey adulteration in the Philippines. The sample set contained 288 spectra of authentic and adulterated honey samples obtained from six farms in the provinces of Benguet and La Union. Spectral data were pre-processed using ParLeS software in which partial least squares regression (PLSR) analysis was employed. Multivariate analyses like PLSR, principal component analysis (PCA) and linear discriminant analysis (LDA) were performed to predict the level of adulteration, reducing sugar content, and apparent sucrose content. Calibration models for predicting adulteration level (from none to highly adulterated) gave the best results using PLSR on spectral data without pretreatment (calibration R2 > 0.93, validation R2 > 0.94). Principal components (PCs) of the spectral data were extracted using PCA wherein the NIR absorbance bands for sugar (836-840 nm and 978-980 nm) and water (938-940 nm, 978-980 nm, 984-986 nm and 992-994 nm) in honey were identified. Using LDA, calibration models were able to classify honey based on level of adulteration and apparent sucrose; overall accuracies of 99.5% and 100% were observed, respectively. When validated, however, the LDA models could not detect pure honey (out of 6 and 11 samples, respectively), while 86.7% and 100% of adulterated honey could be detected, respectively. Further sampling is recommended to strengthen LDA models.