Targeted and Nontargeted Wine Analysis by ¹H NMR Spectroscopy Combined with Multivariate Statistical Analysis. Differentiation of Important Parameters: Grape Variety, Geographical Origin, Year of Vintage

TLDR

NMR spectra serve as unique fingerprints of wine, encoding variety, origin, vintage, and processing details. The study aimed to determine grape variety, geographic origin, and vintage of German wines using 1H NMR spectroscopy combined with multivariate statistical analysis. The authors analyzed ~600 wines from five German regions, acquiring 1H NMR spectra with water/ethanol saturation for high SNR, and applied PCA, LDA, MANOVA, cross‑validation, and Monte Carlo resampling to classify samples. The multivariate approach correctly classified grape varieties 95%, vintages 97% and 96%, and geographic origins 89% on average.

Abstract

The authenticity, the grape variety, the geographical origin, and the year of vintage of wines produced in Germany were investigated by 1H NMR spectroscopy in combination with several steps of multivariate data analysis including principal component analysis (PCA), linear discrimination analysis (LDA), and multivariate analysis of variance (MANOVA) together with cross-validation (CV) embedded in a Monte Carlo resampling approach (MC) and others. A total of about 600 wines were selected and carefully collected from five wine-growing areas in the southern and southwestern parts of Germany. Simultaneous saturation of the resonances of water and ethanol by application of a low-power eight-frequency band irradiation using shaped pulses allowed for high receiver gain settings and hence optimized signal-to-noise ratios. Correct prediction of classification of the grape varieties of Pinot noir, Lemberger, Pinot blanc/Pinot gris, Müller-Thurgau, Riesling, and Gewürztraminer of 95% in the wine panel was achieved. The classification of the vintage of all analyzed wines resulted in correct predictions of 97 and 96%, respectively, for vintage 2008 (n = 318) and 2009 (n = 265). The geographic origin of all wines from the largest German wine-producing regions, Rheinpfalz, Rheinhessen, Mosel, Baden, and Württemberg, could be predicted 89% correctly on average. Each NMR spectrum could be regarded as the individual "fingerprint" of a wine sample, which includes information about variety, origin, vintage, physiological state, technological treatment, and others.