Authentication of honey floreal origin by using NIR spectroscopy

Honey is a complex natural product that can be classified in different types according to its botanical origin, geographical site of production and season of harvest. Due to the great difference in prices of this product, for the inspection of deceptive practice, it would be useful to develop low-cost and quick multi-analytical methods such as NIRS. At this purpose, polyflora (n = 77), acacia (Robinia) (n = 27) and chestnut (Castanea sativa) (n = 22) honey samples, collected in the spring and summer 2016 in the Veneto region (North-East of Italy) were analysed by NIRS. Spectral data were obtained by a FOSS DS-2500 scanning monochromator using a ring cup equipped with a quartz window and a 0.5 mm optical path gold reflector. Spectral data recorded as log(1/reflectance) were used to build a PLS discriminant analysis (PLS Toolbox, Eigen. Res., USA). To assess the accuracy of the PLS-DA model a cross-validation by means of a venetian blind algorithm was applied (Rubingh et al., 2006). The PLS-DA tested model was not able to reliable discriminate the honey samples according to botanical origin, as the R2CV was lower than 0.50 for the three experimental thesis. The accuracy of discrimination in term of sensitivity (proportion of correctly classified positive cases) was 0.77, 0.63 and 0.59 for polyfloral, acacia and chestnut honey samples, respectively. The outcomes of this study highlighted that the NIRS technique was inadequate to discriminate floral types of honey samples. This was probably due to the high variability and the different effect of the crystallization process on physical and chemical structure of honey within the floral groups (Bakier, 2009). A specific analysis of the VIP (variable importance in prediction) related to short spectral range (Ottavian et al., 2010) and/or a thermic pre-treatment of honey before NIRS data collection might be helpful strategies to improve the performance of the prediction. Acknowledgements – Project CPDA 158894/15 – Padova University (call 2015)