Determination Of Trace Elements Composition And Stable-isotope Ratio Of Honey And Their Classification

The traceability of food is one of the important aspects in food safety, and analyticaltechniques that are being used for the determination of the origin of food products are veryessential. Trace elements and stable isotope ratios have been widely applied to determinethe origin of food for providing more accurate and reliable information. In this study, amethod for the determination of44trace elements in honey was developed by inductivelycoupled plasma mass spectrometry (ICP-MS). The “bottom-up” approach was used toestimate the sources of uncertainty, such as repeatability, weighing of the sample, recovery,sample volume, sample dilution, stock solution and calibration curve. On the other hand,isotope ratio mass spectrometry (IRMS) was applied to determine the hydrogen, carbon,nitrogen and sulphur stable isotope ratios. Taking carbon stable isotope ratio as anexample, the method for determining the stable carbon isotope ratio value was validatedand measurement uncertainty of this process was also estimated. At the same time,strontium (Sr) and lead (Pb) isotopic ratios were also determined by ICP-MS.On the basis of these,44trace elements, hydrogen, carbon, nitrogen and sulphurstable isotope ratios, as well as Sr and Pb isotopic ratios of253Chinese honey samplesfrom five botanical origins, linden, vitex, rape, lychee and acacia, which were collectedfrom Heilongjiang, Beijing, Hebei, Shandong, Shaanxi and Guangdong have beendetermined. Chemometric approaches, such as principal component analysis (PCA),partial least squares-discriminant analysis (PLS-DA) and back-propagation artificialneural network (BP-ANN), were applied to classify honey samples according to botanicaland geographical origins based on trace element and stable isotope ratios. The researchwas mainly focus on the difference of trace elements and stable isotope ratios in honeysamples from different botanical and geographical origins, and the performance of modelsbased on different variables combined with different chemometric approaches in honeytracing. At last, the proposed models were applied to predict the origin of commercialhoney samples. The details could be found as fellows:When taking trace elements as variables, for both botanical and geographical origin,the performance of BP-ANN was better than PLS-DA. For botanical origin, the total correct classification rates during model training, cross-validation and prediction weremore than98%. For geographical origin, the total correct classification rates of acacia,vitex and all of the honey samples were more than90%,95%and96%, respectively.When taking stable isotope ratios as variables, for botanical origin, the performanceof PLS-DA was better than BP-ANN and the range of total correct classification ratesduring model training, cross-validation and prediction was76.2%-83.4%. However, forgeographical origin, the performance of BP-ANN was better than PLS-DA. For acacia andvitex honey, the ranges of total correct classification rates by BP-ANN were68.4%-93.5%and90.8%-98.5%, respectively. For geographical origin of all of the honey samples, onlyHebei and Heilongjiang have the higher accuracy by BP-ANN.For Sr and Pb isotopic ratios, only84Sr/86Sr and84Sr/88Sr have significant differencein acacia honey samples from Hebei and other origins. No difference of Sr and Pb isotopicratios could be found in vitex honey samples from different geographical origins. Taking84Sr/86Sr and84Sr/88Sr as variables for all of the honey samples, only Hebei andHeilongjiang have the higher accuracy by BP-ANN.Because of the difficulty in collecting the honey samples, taking rape honey as anexample, the relationship of trace elements in rape honey and its corresponding rapeflower and stem were researched. The probabilities of using trace elements in rape flowerand stem instead of rape honey for the traceability of rape honey were studied, and thetotal correct classification rates of BP-ANN model were more than93%.