A New Approach To Examine the Relationships between Sensory and Gas Chromatography−Olfactometry Data Using Generalized Procrustes Analysis Applied to Six French Chardonnay Wines

Abstract

Six French Chardonnay wines were submitted to both sensory and combined headspace/gas chromatography-olfactometry analyses. The detection frequencies allowed five hierarchical levels to be distinguished: P25, the odorant areas (OAs) having a detection frequency > or =25% (the complete olfactogram without the odor noise); P40, > or =40%; P55, > or =55%; P70, > or =70%; and P85, > or =85%. Moreover, the detection frequencies were analyzed to distinguish 21 discriminative OAs. Wines tested by sensory analysis and the headspace samples analyzed by gas chromatography-olfactometry (GC-O) were described by a heterogeneous vocabulary distributed into nine overall classes of descriptors. The new statistical treatment to examine hierarchical or discriminative OA categories with respect to sensory data used Generalized Procrustes analysis (GPA) from coordinate tables provided by correspondence analysis (CA). The successive data sets supplied by CA were subjected to GPA to yield consensus method maps. The more selective levels of detection frequency (P70 and P85) were responsible for incomplete or distorted information with respect to sensory data. The most appropriate segmentation of the OA distribution (olfactogram) to represent the sensory profile of the six samples would correspond to the intermediate pattern (P40 and P55). The other interest was to study the reasons of distortion due to the dynamic headspace extraction. The highest proportions of the variance were at all times related to the same classes: spicy, herbaceous, and, to a lesser degree, microbiological. This would indicate that the dynamic headspace analysis induces a distortion with respect to sensory data, which systematically affected the perception of both spicy and herbaceous characters of wines.