Taxonomical value of leaf shape in Cuban species of the genus Begonia (Begoniaceae)

Abstract

Among the relevant elements to delimit Cuban species of the genus Begonia is the shape of the leaf, although this character alone is considered insufficient to discriminate some of them. The studies that analyze it, use linear dimensions and nominal categories. However, variations of this character are usually continuous, so its quantification using geometric morphometry could be more efficient compared to traditional morphology. Therefore, the objective of this study was the evaluation of the potential of the differences in the shape of the leaves for the identification of Cuban species of the genus Begonia. For this, photographs of the leaves of seven species of the genus were used. Its outline was described using Fourier elliptic coefficients, which were reduced with a Principal Component Analysis. A PERMANOVA test was performed and a neural network was trained to evaluate the potential of this character to identify them. Likewise, the leaf asymmetry index was calculated for each taxon. The graphical reconstruction of the leaf outlines allowed to visualize that the interspecific differences in their shape are mainly found in the degree of ellipsis and asymmetry. The leaves of B. lomensis and B. wrightiana were the most asymmetric. The PERMANOVA test showed significant differences between all species. However, the trained neural network made classification errors for three individuals, although it achieved a perfect identification in five of the seven included taxa.