Analytical, Nutritional and Clinical Methods SectionChemometric classification of honeys according to their type. II. Metal content data
Introduction
In recent years, pattern recognition techniques have been widely applied in food chemistry (Brown, Bear, & Blank, 1992Forina & Lanteri, 1984). After the work of Kwan et al. on the classification of wines of Vitis vinifera cv. Pinot Noir from France and the United States (Kwan & Kowalski, 1980, Kwan, Kowalski, & Skogerboe, 1979), a number of examples were reported in the literature concerning a variety of products. These include the geographic classification of olive oils (Derde, Coomans, & Massart, 1984), concentrate orange juices (Bayer, McHard, & Winefordner, 1980) and wines (Herrero & Médina, 1990). More recently, there has been varied work based on diverse types of chemical variables and sensory properties using different statistical tools: principal component analysis, linear discriminant, quadratic discriminant, and SIMCA methods were compared for classification and predictive ability in the separation of two species of fish (Franco, Seeber, Sferlazzo, & Leardi, 1990); chemometric studies on minor and trace elements in cow’s milk were made to differentiate two types of milk according to feeding (Favretto, Vojnovic, & Campisi, 1994); pattern recognition methods have been used in characterization and classification of wine and alcoholic beverages (Etiévant, Schlich, Symonds, Bouvier, & Bertrand, 1988Herrero, Latorre, Garcı́a, & Médina, 1994Maarse, Slump, Tas, & Schaefer, 1987Moret, Scarponi, & Cescon, 1994Vasconcelos & Chaves, 1989).
In a previous work (Herrero & Peña, 1993), we showed that multivariate statistical methods applied to physicochemical parameters can be sucessfully used in order to achieve a correct geographic classification of honey samples from different origins. Also, pattern recognition procedures were used to obtain a classification model based on quality control data to distinguish between natural and processed honey samples (Herrero et al., 1996). In none of these cases were pollen studies necessary to obtain an almost correct assignation of the honey samples.
In this work, we present the results of the application of pattern recognition methods to key metals data in honey to differentiate between processed and industrially commercialized honeys from various origins and natural honeys from Galicia directly obtained from the producers. The basis of the classification procedure was the metal composition of honey samples. Minerals seem to be good candidates for a classification system, as they are stable; however, the number of research papers in the literature concerning the metal content of honey is small (Rodrı́guez-Otero, Paseiro, Simal, Terradillos, & Cepeda, 1992Stein & Umland, 1986Li, Whadat, & Neeb, 1995) and only in one case was the data obtained used to carry out a chemometric classification of honey samples according to their geographical origin (Feller, Vincent, & Beaulieu, 1989). The interest in this classification model is two-fold. On the one hand, processed honeys undergo a heat treatment, generally targeted to handle and homogenize honeys from different origins in the bottle industry; this process causes alterations that can affect the properties and quality of the product. On the other hand, processed honeys from various origins, due to their lower price, can be used as possible substrates for falsification of natural Galician honeys.
Section snippets
Honey samples
Twenty-two natural honey samples from Galicia were provided by the local association of beekepers with guaranteed origin and made by traditional procedures in the producing region. All the natural samples examined were unprocessed honeys of random (mixed) floral type. None of these samples underwent any process that could alter their composition. Twenty processed honey samples from various origins except Galicia were obtained from different supermarkets and commercial areas. Samples were
Results and discussion
The results of the 11 metals determined in honey samples are summarized in Table 1.
The levels obtained in natural Galician honey were similar to those found by other authors in honey samples from Spain (Huidobro, 1983, Sancho, 1990, Serra, 1989) and also in honey samples from Galicia (Rodrı́guez-Otero et al., 1992). The contents of Co and Ni were, in all cases, less than detection limits, 0.02 and 0.05 μg g−1, respectively. The variability of results for all metals in both groups are large,
Conclusion
We have demonstrated that pattern recognition is able to extract useful information for an amount of data. Information was used to relate chemical composition of honeys with their processing and geographic origins. Differentiation and classification of processed and unprocessed honey samples from Galician and non-Galician origin was made possible by using the concentration data of various selected metals and applying multidimensional chemometric techniques. The use of all available features is
Acknowledgements
We express our gratitude to the Asociación Lucense de Apicultura for providing natural honey samples. This work was partly financed by the Instituto Lucense de Desarrollo (INLUDES), Lugo, Spain.
References (30)
- et al.
Chemometric studies on minor and trace elements in cow’s milk
Analytica Chimica Acta
(1994) - et al.
Classification and prediction ability of pattern recognition methods applied to sea-water fish
Analytica Chimica Acta
(1990) Pattern recognition by means of disjoint principal component models
Pattern Recognition
(1976)- et al.
Determination of geographic origin of frozen concentrated orange juices via pattern recognition
Journal of Agricultural and Food Chemistry
(1980) - et al.
Chemometrics
Analytical Chemistry
(1992) - et al.
SIMCA demonstrated with Characterizacion of Italian olive oils
Journal of the Association of Official Analytical Chemists Symposium Series
(1984) - et al.
Varietal and geographic classification of French red wines in terms of elements, amino acids and aromatic alcohols
Journal of the Science of Food and Agriculture
(1988) - et al.
Teneur en minraux et origine géographique des miels du Canada
Apidology
(1989) - Forina, M., & Lanteri, S. (1984). Data analysis in food chemistry. In B. R. Kowalski (Ed.), Chemometrics, mathematics...
- Forina, M., Leardi, R., Armanino, C., & Lanteri, S. (1988). Parvus: an extendable package of programs for data...