|
A Technical Guide for Wine Producers
|
|
RECENT ARTICLES | WYNBOER HOME
Classification of South African wines according to geographical origin using multi-element chemical analysis
by Paul Coetzee1 and Ockert Augustyn2
1Department of Chemistry, University of Johannesburg, Johannesburg,
Introduction
Fingerprinting techniques based on chemical composition and multivariate statistical analysis (1,2) of analytical data, can be used for identification and classification of a specific agricultural product type according to geographical origin. The method assumes that the chemical composition of an agricultural product such as wine (3-10), coffee (11), tea (12), olive oil and fruit juice (13,14) will reflect the composition of the provenance soil, at least for certain elements. The key to the successful application of this technique is the selection of suitable elements that would reflect the link with soil geochemistry and thus have discriminating potential for that particular product. Only a limited number of elements qualify for this purpose. Reliable information on the elemental composition, mostly at trace level, is required to apply this method with any degree of success. One of the most versatile techniques for this purpose is ICP-MS, which can be used to determine the multi-element composition (3-9) of samples or the isotopic ratios (11,15-17) of elements that show variation in their isotopic composition in different geographical regions. Elements with isotopes of radiogenic origin, such as Sr and Pb, fall into this category.
In this study, the trace element composition of South African wines from three major wine-producing regions in the Western Cape Province was compared with the aim of identifying and classifying the wines according to origin. Studies in different wine-producing countries, for example Canada (4), Spain (8), Portugal (5), Germany (3,7), France (18,19), and Switzerland (20), have shown that the method has potential in obtaining information about the origin of wines. The procedure is by no means simple, because many factors may cause the trace element composition of a wine to differ from that related to the elemental composition of the soil alone. These factors include agricultural practices such as the use of fertilizers and pesticides (5) and climatic factors such as heavy rains during the growth season. Pollution (16) and the quality of the irrigation water must also be considered. The vinification process itself may contribute to the trace element composition of a wine. Examples of elements that were found to be little changed during the vinification process, include: Li, B, Mg, Ca, Mn, Zn, Rb, Sr, Cs, and Pb (21), and those that could be affected were: Al, Cd, Co, Cr, Cu, Fe, Mn, Pb, V, and Zn (22). No consensus is reached on this effect and the elements included in the two lists may differ from study to study. Differential uptake of trace elements by different vine varieties or differential uptake of trace elements in the skin and flesh of the grape (23) can cause problems. White and red wine from the same region may differ in trace element composition because for red wine the skin contact is longer after crushing than in the process for making white wine. More elements may therefore leach out into the red grape juice than in the white grape juice.
All these factors may influence the correlation between wine and soil composition and limit the usefulness of the fingerprinting procedure. Some of these factors have been investigated, but the results reported in the literature are still inconclusive and in many cases contradictory. Ideally then, the elements selected for identification should not be affected by the vinification process, agricultural practices and environmental conditions, and should correlate with soil composition.
The specific aims of this study were the determination of the elemental composition of wines from Robertson, Stellenbosch, and Swartland, three main wine producing areas in the Western Cape Province of South Africa and the use of this data to uniquely classify wines from these areas according to a multivariate statistical procedure. A set of indicator elements suitable for discriminant analysis and specific for South African wines was to be determined. The feasibility of including red and white wines in the same data set for provenance determination was evaluated.
Materials and Methods
- Multi-element determinations were carried out with a Perkin Elmer Sciex Elan 5000 or Thermo Electron X7 quadrupole-based ICP-MS instruments equipped with cross-flow nebulizers.
- A sample set of 40 wines was obtained from three major wine-producing regions in South Africa: Stellenbosch, Robertson and Swartland. Each wine could be traced to a specific vineyard where the grapes were grown to produce the wine. Both red and white wines were included. Table 1 lists the origin and soil type pertaining to the selected wines. The white cultivars were about equally divided among chardonnay, sauvignon blanc, and chenin blanc. The red cultivars consisted of cabernet sauvignon, shiraz, pinotage and merlot. None were fermented or aged in wood. All wines were from the 2004 vintage. They were sampled before blending and final bottling in order to preserve the trace to the vineyard of origin.
- Dilution. Wine samples were diluted 1:1 with 0.14 M HNO3. The dilution reduces the ethanol concentration to between 5 and 6%, which is sufficiently low to diminish matrix effects and plasma instability commonly associated with introducing matrices containing organics into the plasma. After 1:1 dilution, concentrations of most elements of interest, except the rare earth elements, remain comfortably above the detection limits of the instrument.
- Microwave digestion. Wine samples were also analysed after microwave digestion (24-26) using a Milestone MLS1200 microwave digestion system. 1.5 mL wine sample + 150 µL 65% HNO3 + 1.5 mL 30% H2O2 were added into the PTFE digestion cells and microwaved for 33 min by increasing the power to 600W in a stepwise fashion. The residues were dissolved in 12 mL of 0.14 M HNO3.
- A standard solution containing 50 µg/L of Be, Co, Rh, In, Pb, Th in 6% ethanol/0.14M HNO3 was used to calibrate the whole mass range for measurement of diluted samples made in semi-quantitative mode. In the case of measuring digested samples, ethanol was omitted in the calibration solution. The internal standard In, was added to the level of 50 µg/L to all samples, standards, and blanks. External standards for quantitative analysis were prepared in 6% ethanol/0.14 M HNO3 for diluted samples and in 0.14 M HNO3 for digested samples. Blanks for the measurement of diluted and digested samples were a 6% ethanol/0.14 M HNO3 solution containing 50 µg/L of the internal standard and a 0.14 M HNO3 procedure blank submitted to the microwave treatment and including the internal standard, respectively. Standards and internal standard were prepared by appropriate dilution from 1000 mg/L Merck ICP standard stock solutions.
- To establish whether the destruction of the organic matrix, in particular the ethanol in the wine samples, would bring any advantages, microwave digested wine samples were measured using the semi-quantitative mode of analysis and the results compared with those obtained from diluted samples. The results for diluted and digested samples compare adequately for most elements.
It was concluded that the 1:1 dilution of the wine samples was to be preferred to the digestion procedure because of the following advantages: lower detection limits, less risk of contamination, less risk of loss of volatile analyte elements, less time-consuming and simple sample preparation. The 1:1 dilution approach was therefore used in the rest of the study.
- Many reports (27-30) can be found in the literature comparing the semi-quantitative and quantitative mode in ICP-MS for multi-element analysis. The conclusions are generally that the results for most elements are comparable.
- To confirm that these conclusions were in fact valid for the analytical procedures and instrumentation used in the current study, results obtained using the semi-quantitative procedure were compared with those obtained using the quantitative procedure. Figure 1 shows the plot of the mean values obtained by the quantitative mode on the x-axis and the semi-quantitative mode on the y-axis. The trendline obtained by least squares regression, y = 1.001x + 0.615, with regression coefficient 0.9943, shows a slope of 1.001, very close to the ideal line.
- These results are similar to previously reported comparative tests (30) and support the use of multi-element concentration data obtained by the semi-quantitative mode of ICP-MS analysis for wine provenance studies.
Results and Discussion
- The results for the analysis of the wine samples are summarised in Table 2 for the 40 elements initially selected for analysis.
- The concentration ranges of most elements overlap within the three regions as can be inferred when comparing the standard deviations of the means for each element in each region. Typically, several combinations of elements could differentiate adequately between any two of the regions, but not all three simultaneously if simple techniques such as scatterplot analyses are used. A more robust statistical analysis procedure such as discriminant analysis is therefore required to differentiate the wines according to region. Because the sample size (number of wines) was relatively small relative to the number of variables (element concentrations), a reduction in the number of variables was necessary to perform useful multivariate statistical analysis. An ANOVA test indicated that the group means of the following elements show significant differences at the 95% confidence level: Li, B, Mg, Al, Si, Cl, Sc, Mn, Ni, Ga, Se, Rb, Sr, Nb, Cs, Ba, La, W, Tl, and U. A few elements (Mg, Si, Cl, Nb, La, and U) were excluded from this group where the analytical uncertainty was large because of high polyatomic background interference or where concentration levels were close to the LOD. Based on these considerations the 15 elements, Li, B, Al, Sc, Mn, Ni, Ga, Se, Rb, Sr, Nb, Cs, Ba, La,W, Tl, and U, were selected to be used in discriminant analysis.
- The discriminant analysis was done by using the statistical package SPSS. Discriminant functions, that is linear combinations of the independent variables (elemental concentrations) were derived that best differentiate among the dependent variables (the three regions Stellenbosch, Robertson, and Swartland). The following canonical discriminant functions were derived by stepwise discriminant analysis:
These functions enabled 100% of wines to be correctly classified. A graphical presentation of the two discriminant functions is given in Figure 2.
It is clear from the plot in Figure 2 that the red and white wines are completely mixed in the three well-separated clusters, and it was therefore not deemed necessary to perform separate discriminant analyses for red and white wines (the sample sizes per wine type being too small for a meaningful discriminant analysis to be performed).
This procedure suggests that it would in principle be possible in future to classify a wine known to be from one of these regions, as coming from Robertson, Stellenbosch, or Swartland. The success rate will depend on how representative the current database is and on factors such as the variability of the soils within a defined region.
Discriminant analysis of the data has shown that red and white wines included in the same data set can be classified correctly according to region of origin. However, it is known that differences may exist in the elemental composition of wines made from different grape cultivars, in particular between red and white wine cultivars. An indication of the nature of these differences can be seen in Table 2 which summarises the means of elemental concentrations for red and white wines in the regions. Red wines show higher elemental concentrations than white wines for most elements. This is consistent with the known fact that the skins of the grapes are enriched in trace elements. In the case of red wines, the crushed grapes remain in contact with the skins for a longer period, allowing more minerals to leach out from the skins.
Because of the seemingly large differences observed in concentrations of some elements in red and white wine from the same region, the question arises whether this could affect the reliability of the fingerprinting procedure when mixed white and red data are used. This would be the case if regional differences were smaller than cultivar differences. To validate the use of one model for red and white wine combined, the discriminant scores of the two discriminant functions were compared by means of a multivariate analysis of variance (MANOVA) test. This test showed that the difference between the means of the three regions was highly significant whereas the difference between the means of the wine types (red and white) was not significant.
Conclusion
The excellent prediction rate of 100% that was achieved in classifying wines from three wine producing areas in South Africa according to their geographical origin, gives further evidence of the ability of multivariate statistical analysis, based on trace element data, to show provenance. Adequately reliable multi-element analytical data were obtained using the semi-quantitative mode of ICP-MS analysis and minimum sample preparation in the form of a 1:1 dilution of the wine samples with 1% HNO3.
Acknowledgements
The authors would like to thank the Vinpro representatives in Stellenbosch, Robertson and Swartland as well as the many producers who ultimately provided the samples. They would also like to thank the Research Fund of Ghent University (Flemish / South African bilateral scientific and technological co-operation, Project No. 011S2403), the South African National Research Foundation, the Flemish Fund for Scientific Research (Research project G.0037.01), and Winetech (Project No. WW08/28), for research funding.
References
Kaufmann, A. Multivariate statistics as a classification tool in the food laboratory. J. AOAC Int. 1997, 3, 665-675.
Martinez, I.; Aursand, M.; Erikson, U.; Singstad, T. E.; Veliyulin, E.; Van der Zwaag, C. Destructive and non-destructive analytical techniques for authentication and composition analyses of foodstuffs. Trends Food Sci. Tech. 2003, 14, 489-498.
Thiel, G.; Geisler, G.; Blechschmidt, I. Determination of trace elements in wines and classification according to their provenance. Anal. Bioanal. Chem. 2004, 378, 1630-1636.
Taylor, V. F.; Longerich, H. P.; Greenough, J. D. Multi-element analysis of Canadian wines by ICPMS and multivariate statistics. J. Agric. Food Chem. 2003, 51, 856-860.
Almeida, C. M. R.; Vasconcelos, M. D. T. S. Multi-element composition of wines and their precursors including provenance soil and their potentialities as fingerprints of wine origin. J. Agric. Food Chem. 2003, 51, 4788-4798.
Baxter, M. J.; Crewes, H.M.; Dennis, M. J.; Goodall, I.; Anderson, D. The determination of the authenticity of wine from its trace element composition. Food Chem. 1997, 60, 4433-450.
Gomez, M. D. M.; Feldmann, I.; Jakubowski, N.; Andersson, J. T. Classification of German wines with certified brand of origin by multielement quantitation and pattern recognition techniques. J. Agric. Food Chem. 2004, 52, 2962-2974.
Rebolo, S,; Pena, R.; Latorre, M. J. Characterization of Galician (NW Spain) Ribeira Sacra wines using pattern recognition analysis. Anal. Chim. Acta. 2000, 417, 211-220.
Jakubowski, N.; Brandt, R.; Stuewer, D.; Eschauer, H. R.; Gortges, S. Analysis of wines by ICP-MS: Is the pattern of the rare earth elements a reliable fingerprint for the provenance? Fresenius J. Anal. Chem. 1999, 364, 424-428.
Arvanitoyannis, I. S.; Katsota, M. N.; Psarra, E. P.; Soufleros, E. H.; Kallithraka, S. Application of quality control methods for assessing wine authenticity: Use of multivariate analysis (chemometrics). Trends in Food Sci.Tech. 1999, 10, 321-336.
Wieser, M.E.; Iyer, S.S.; Krouse, H.R.; Cantagallo, M.I. Variations in the boron isotope composition of Coffea Arabica beans. Appl. Geochem. 2001, 16, 317-322.
Moreda-Pi¤eiro, A.; Fisher, A.; Hill, S. J. The classification of tea according to region of origin using pattern recognition techniques and trace metal data. J. Food Composition Anal. 2003, 16, 195-211.
Day, M. P.; Zhang, B-L.; Martin, G. J. The use of trace element data to complement stable isotope methods in the characterization of grape musts. Am. J. Enol. Vitic. 1994, 5, 79-85.
Ogrinc, N.; Koir, I. J.; Spangenberg, J. E.; Kidric, J. The application of NMR and MS methods for detection of adulteration of wine, fruit juices, and olive oil. Anal. Bioanal. Chem. 2003, 376, 424-430.
Horn, P.; H”lzl, S.; Todt, W.; Matthes, D. Isotope abundance ratios of Sr in wine provenance determinations, in a tree-root activity study, and of Pb in a pollution study on tree-rings. Isotopes in Environ. Health Studies 1998, 34, 31-42.
Rosman, K. J. R.; Chisholm, W.; Jimi, S.; Candelone, J-P.; Boutron, C. F.; Teissedre, P-L.; Adams, F.C. Lead concentrations and isotopic signatures in vintages of French wine between 1950 and 1991. Environ. Res. 1998, 78, 161-167.
Almeida, C. M. R.; Vasconcelos, M. D. T. S. Does the winemaking process influence the wine 87Sr/86Sr? A case study. Food Chem. 2004, 85, 7-12.
Day, M. P.; Zhang, B. L.; Martin, G. J. Determination of the geographical origin of wine using joint analysis of elemental and isotopic composition: Differentiation of the principal production zones in France for the 1990 vintage. J. Sci. Food & Agric. 1995, 67, 113-123.
Martin, G. J.; Mazure, M.; Jouitteau, C. Characterization of the geographic origin of Bordeaux wines by a combined use of isotopic and trace element measurements. Am. J. Enol. & Viticulture 1999, 50, 409-417.
Gremaud, G.; Quaile, S.; Piantini, U. Characterization of Swiss vineyards using isotopic data in combination with trace elements and classical parameters. Eur. Food Res. Tech. 2004, 219, 97-104.
Gomez, M. D. M.; Jakubowski, N.; Andersson, J. T. Changes in the metal composition in German wines through the winemaking process. A study of 63 elements by ICP-MS. J. Agric. Food Chem. 2004, 52, 2953-2961.
Nicolini, G.; Larcher, R.; Pangrazzi, P. Changes in the contents of micro- and trace-elements in wine due to winemaking treatments. VITIS 2004, 43, 41-45.
Eschauer, H.; Jakob, L.; Meierer, H.; Neeb, N. Use and limitations of ICP-OES in wine analysis. Microchim. Acta. 1989, 3, 291-298.
Smith, F.; Arsenault, E. Microwave-assisted sample preparation in analytical chemistry. Talanta 1996, 43, 1207-1268.
Perez-Jordan, M.Y.; Soldevila, J.; Salvador, A.; Pastor, A.; De la Gardia, M. Inductively coupled plasma mass spectrometry analysis of wines. J. Anal. At. Spectrom. 1998, 13, 33-39.
Castineira, M. M.; Brandt, R.; Von Bohlen, A.; Jakubowski, N. Development of a procedure for the multi-element determination of trace elements in wine by ICP-MS. Fresenius J. Anal. Chem. 2001, 370, 553-558.
Almeida, C. M. R.; Vasconcelos, M. D. T. S. Advantages and limitations of the semi-quantitative operation mode in an inductively coupled plasma-mass spectrometer for multi-element analysis of wines. Anal. Chim. Acta. 2002, 463, 165-175.
Barbaste, M.; Robinson, K.; Guilfoyle, S.; Medina, B.; Lobinski, R. Evaluation of the accuracy of the determination of lead isotope ratios in wine by ICP-MS using quadrupole, multicollector magnetic sector and time-of-flight analyzers. Talanta 2001, 54, 307-317.
Laborda, F.; Medrano, J.; Castillo, J. R. Quality of quantitative and semi-quantitative results in ICP-MS. Anal. At. Spectrom. 2001, 16, 732-738.
|
|
Wynboer is incorporated in WineLand, magazine of the SA wine producers.
Visit our sister sites:
|
| UP |
COPYRIGHT (C) 2000 WineLand |