Identification of Rhus succedanea L. Cultivars Using Elliptic Fourier Descriptors Based on Fruit Shape

Abstract

Abstract We have developed a method to identify cultivars of Rhus succedanea L. based on their fruit contour shape. For this, we collected fruits of five cultivars from three different environments (differing in site and/or year of collection) and the horizontal contour shape of each fruit was expressed by 37 elliptic Fourier descriptors, normalized in terms of size, rotation, shift and starting point of contour tracing. The first six components derived from a principal component analysis of the elliptic Fourier descriptors explained 89% of the variance. The differences among cultivars, environments and the cultivar x environment interaction were significant at the 0.01% probability level for all six principal components according to ANOVA. UPGMA cluster analysis based on the six principal components showed a high degree of clustering and most (but not all) ramets from the same cultivar clustered together. However, results of a UPGMA cluster analysis of Mahalanobis’ generalized distances among cultivars and environments, based on the 37 elliptic Fourier descriptors, showed that samples from the same cultivars clustered together, regardless of the environmental factors. We then applied a ‘similarity probability’ test, based on Mahalanobis’ generalized distances and a randomization test. The similarity probabilities between descriptors in the database and sampled fruits, when the cultivars they represented were included in the database, were >97%. In contrast, for samples representing cultivars that were not included in the database, the probabilities were <46%. These figures also apply to pairs of samples included in the database that represented the same cultivar, and different cultivars, respectively. These results suggest that it is possible to identify R. succedanea cultivars based on fruit contour shape using elliptic Fourier descriptors and similarity probability analysis.