Use of energy-dispersive X-ray fluorescence combined with chemometric modelling to classify honey according to botanical variety and geographical origin.

Honey is one of the food commodities most frequently affected by fraud. Although addition of extraneous sugars is the most common type of fraud, analytical methods are also needed to detect origin masking and misdescription of botanical variety. In this work, multivariate analysis of the content of certain macro- and trace elements, determined by energy-dispersive X-ray fluorescence (ED-XRF) without any type of sample treatment, were used to classify honeys according to botanical variety and geographical origin. Principal component analysis (PCA) and partial least squares-discriminant analysis (PLS-DA) were used to create classification models for nine different botanical varieties-orange, robinia, lavender, rosemary, thyme, lime, chestnut, eucalyptus and manuka-and seven different geographical origins-Italy, Romania, Spain, Portugal, France, Hungary and New Zealand. Although characterised by 100% sensitivity, PCA models lacked specificity. The PLS-DA models constructed for specific combinations of botanical variety-country (BV-C) allowed the successful classification of honey samples, which was verified by external validation samples. Graphical abstract.