Application of elliptic Fourier analysis in understanding leaf shape characters of Anthurium

The neotropical genus Anthurium is the largest and possibly the most complex genus of the Araceae. Conservative estimates suggest that a total of 600-800 species exist worldwide; and undoubtedly, many of these have not yet been described. The genus is taxonomically difficult, exhibiting considerable morphological plasticity in most structures and ordinary brief descriptions are without much value. Also, aroid systematists in general, and those applying numerical methods in particular, frequently use ratios to express the shape of the leaves and spathe in order to standardize for size. However, ratios may have undesirable statistical properties and do not accurately represent continuous shape differences in the leaves of these plants. In this study, however, advances in image analysis and geometric morphometric techniques were applied to discriminate between thirteen varieties of Anthurium and one outgroup, Spathiphylum commutatum; and to allow for complete and uniform quantitative descriptions for more accurate comparisons of the plants. To do this, full-color digital pictures of the leaves of the aroid plants were converted into binary images using an image processing and analysis software. Then, the contours of the leaves were summarized as chain-codes that were later converted into elliptic Fourier coefficients (equivalent to 20 harmonic modes). These coefficients were used as shape descriptors that were employed to automatically reconstruct and produce line drawings of the leaves. Principal component analysis of these elliptic Fourier descriptors revealed a total of 8 principal components that are associated with 8 independent shape characteristics. Results also showed that most of the variations described by the first principal component (PC1) could be attributed to differences in the shapes of the sinus between the posterior lobes on the base of the leaves (parabolic, hippocrepiform and spathulate). On the other hand, PC2 is associated with the leaf-aspect ratio and could be used to group varieties into those that have wide ovate (cordiform), narrow triangular-ovate, triangular-ovate, and ovate leaves. Principal components 3, 5, 6, 7 and 8 describe asymmetries in the shapes of the leaves while PC4 differentiates the varieties and species based on the extent and breadth of the posterior lobes. Further analysis using Kruskal-Wallis (non-parametric ANOVA) revealed that the variations defined by the first seven principal components are statistically significant (P<0.001). Decomposition of the symmetric and asymmetric components of leaf shape variation revealed that among the varieties and species studied, the leaves of the ‘kaumana’ variant are more asymmetrical when compared to the ‘flamingo’ variant. Cluster analysis was also employed on the elliptic Fourier coefficients to determine the systematic relationships of the ariod plants. The results of this study suggest that geometric morphometric analysis should be included among the tools used by taxonomists to objectively quantify and compare the shapes of two-dimensional structures such as the leaves of aroid plants.